state
stringlengths 0
159k
| srcUpToTactic
stringlengths 387
167k
| nextTactic
stringlengths 3
9k
| declUpToTactic
stringlengths 22
11.5k
| declId
stringlengths 38
95
| decl
stringlengths 16
1.89k
| file_tag
stringlengths 17
73
|
|---|---|---|---|---|---|---|
case a.a
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
J : GrothendieckTopology C
hJ : J ∈ {J | K ≤ ofGrothendieck C J}
X : C
S : Sieve X
hS : S ∈ GrothendieckTopology.sieves (toGrothendieck C K) X
⊢ S ∈ GrothendieckTopology.sieves J X
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
|
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
|
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
|
Mathlib.CategoryTheory.Sites.Coverage.271_0.qkZFgqEgDC2P633
|
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J }
|
Mathlib_CategoryTheory_Sites_Coverage
|
case a.a
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
J : GrothendieckTopology C
hJ : J ∈ {J | K ≤ ofGrothendieck C J}
X : C
S : Sieve X
hS : S ∈ GrothendieckTopology.sieves (toGrothendieck C K) X
⊢ S ∈ GrothendieckTopology.sieves J X
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
|
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
|
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
|
Mathlib.CategoryTheory.Sites.Coverage.271_0.qkZFgqEgDC2P633
|
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J }
|
Mathlib_CategoryTheory_Sites_Coverage
|
case a.a.of
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
J : GrothendieckTopology C
hJ : J ∈ {J | K ≤ ofGrothendieck C J}
X✝ : C
S✝ : Sieve X✝
X : C
S : Presieve X
hS : S ∈ covering K X
⊢ Sieve.generate S ∈ GrothendieckTopology.sieves J X
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
|
| of X S hS => apply hJ; assumption
|
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
|
Mathlib.CategoryTheory.Sites.Coverage.271_0.qkZFgqEgDC2P633
|
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J }
|
Mathlib_CategoryTheory_Sites_Coverage
|
case a.a.of
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
J : GrothendieckTopology C
hJ : J ∈ {J | K ≤ ofGrothendieck C J}
X✝ : C
S✝ : Sieve X✝
X : C
S : Presieve X
hS : S ∈ covering K X
⊢ Sieve.generate S ∈ GrothendieckTopology.sieves J X
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS =>
|
apply hJ
|
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS =>
|
Mathlib.CategoryTheory.Sites.Coverage.271_0.qkZFgqEgDC2P633
|
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J }
|
Mathlib_CategoryTheory_Sites_Coverage
|
case a.a.of.a
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
J : GrothendieckTopology C
hJ : J ∈ {J | K ≤ ofGrothendieck C J}
X✝ : C
S✝ : Sieve X✝
X : C
S : Presieve X
hS : S ∈ covering K X
⊢ S ∈ covering K X
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ;
|
assumption
|
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ;
|
Mathlib.CategoryTheory.Sites.Coverage.271_0.qkZFgqEgDC2P633
|
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J }
|
Mathlib_CategoryTheory_Sites_Coverage
|
case a.a.top
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
J : GrothendieckTopology C
hJ : J ∈ {J | K ≤ ofGrothendieck C J}
X : C
S : Sieve X
X✝ : C
⊢ ⊤ ∈ GrothendieckTopology.sieves J X✝
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
|
| top => apply J.top_mem
|
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
|
Mathlib.CategoryTheory.Sites.Coverage.271_0.qkZFgqEgDC2P633
|
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J }
|
Mathlib_CategoryTheory_Sites_Coverage
|
case a.a.top
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
J : GrothendieckTopology C
hJ : J ∈ {J | K ≤ ofGrothendieck C J}
X : C
S : Sieve X
X✝ : C
⊢ ⊤ ∈ GrothendieckTopology.sieves J X✝
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top =>
|
apply J.top_mem
|
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top =>
|
Mathlib.CategoryTheory.Sites.Coverage.271_0.qkZFgqEgDC2P633
|
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J }
|
Mathlib_CategoryTheory_Sites_Coverage
|
case a.a.transitive
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
J : GrothendieckTopology C
hJ : J ∈ {J | K ≤ ofGrothendieck C J}
X✝ : C
S✝ : Sieve X✝
X : C
R S : Sieve X
a✝¹ : saturate K X R
a✝ : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R.arrows f → saturate K Y (Sieve.pullback f S)
H1 : R ∈ GrothendieckTopology.sieves J X
H2 : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R.arrows f → Sieve.pullback f S ∈ GrothendieckTopology.sieves J Y
⊢ S ∈ GrothendieckTopology.sieves J X
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
|
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
|
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
|
Mathlib.CategoryTheory.Sites.Coverage.271_0.qkZFgqEgDC2P633
|
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J }
|
Mathlib_CategoryTheory_Sites_Coverage
|
case a.a.transitive
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
J : GrothendieckTopology C
hJ : J ∈ {J | K ≤ ofGrothendieck C J}
X✝ : C
S✝ : Sieve X✝
X : C
R S : Sieve X
a✝¹ : saturate K X R
a✝ : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R.arrows f → saturate K Y (Sieve.pullback f S)
H1 : R ∈ GrothendieckTopology.sieves J X
H2 : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R.arrows f → Sieve.pullback f S ∈ GrothendieckTopology.sieves J Y
⊢ S ∈ GrothendieckTopology.sieves J X
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 =>
|
exact J.transitive H1 _ H2
|
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 =>
|
Mathlib.CategoryTheory.Sites.Coverage.271_0.qkZFgqEgDC2P633
|
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J }
|
Mathlib_CategoryTheory_Sites_Coverage
|
case a
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
⊢ sInf {J | K ≤ ofGrothendieck C J} ≤ toGrothendieck C K
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
·
|
apply sInf_le
|
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
·
|
Mathlib.CategoryTheory.Sites.Coverage.271_0.qkZFgqEgDC2P633
|
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J }
|
Mathlib_CategoryTheory_Sites_Coverage
|
case a.a
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
⊢ toGrothendieck C K ∈ {J | K ≤ ofGrothendieck C J}
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
|
intro X S hS
|
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
|
Mathlib.CategoryTheory.Sites.Coverage.271_0.qkZFgqEgDC2P633
|
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J }
|
Mathlib_CategoryTheory_Sites_Coverage
|
case a.a
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
X : C
S : Presieve X
hS : S ∈ covering K X
⊢ S ∈ covering (ofGrothendieck C (toGrothendieck C K)) X
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
|
apply saturate.of _ _ hS
|
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
|
Mathlib.CategoryTheory.Sites.Coverage.271_0.qkZFgqEgDC2P633
|
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J }
|
Mathlib_CategoryTheory_Sites_Coverage
|
C : Type ?u.32430
inst✝ : Category.{?u.32434, ?u.32430} C
x y : Coverage C
⊢ ∀ ⦃X Y : C⦄ (f : Y ⟶ X),
∀ S ∈ (fun B => covering x B ∪ covering y B) X,
∃ T ∈ (fun B => covering x B ∪ covering y B) Y, Presieve.FactorsThruAlong T S f
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
|
rintro X Y f S (hx | hy)
|
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
|
Mathlib.CategoryTheory.Sites.Coverage.288_0.qkZFgqEgDC2P633
|
instance : SemilatticeSup (Coverage C) where
sup x y
|
Mathlib_CategoryTheory_Sites_Coverage
|
case inl
C : Type ?u.32430
inst✝ : Category.{?u.32434, ?u.32430} C
x y : Coverage C
X Y : C
f : Y ⟶ X
S : Presieve X
hx : S ∈ covering x X
⊢ ∃ T ∈ (fun B => covering x B ∪ covering y B) Y, Presieve.FactorsThruAlong T S f
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
·
|
obtain ⟨T, hT⟩ := x.pullback f S hx
|
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
·
|
Mathlib.CategoryTheory.Sites.Coverage.288_0.qkZFgqEgDC2P633
|
instance : SemilatticeSup (Coverage C) where
sup x y
|
Mathlib_CategoryTheory_Sites_Coverage
|
case inl.intro
C : Type ?u.32430
inst✝ : Category.{?u.32434, ?u.32430} C
x y : Coverage C
X Y : C
f : Y ⟶ X
S : Presieve X
hx : S ∈ covering x X
T : Presieve Y
hT : T ∈ covering x Y ∧ Presieve.FactorsThruAlong T S f
⊢ ∃ T ∈ (fun B => covering x B ∪ covering y B) Y, Presieve.FactorsThruAlong T S f
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
|
exact ⟨T, Or.inl hT.1, hT.2⟩
|
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
|
Mathlib.CategoryTheory.Sites.Coverage.288_0.qkZFgqEgDC2P633
|
instance : SemilatticeSup (Coverage C) where
sup x y
|
Mathlib_CategoryTheory_Sites_Coverage
|
case inr
C : Type ?u.32430
inst✝ : Category.{?u.32434, ?u.32430} C
x y : Coverage C
X Y : C
f : Y ⟶ X
S : Presieve X
hy : S ∈ covering y X
⊢ ∃ T ∈ (fun B => covering x B ∪ covering y B) Y, Presieve.FactorsThruAlong T S f
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
·
|
obtain ⟨T, hT⟩ := y.pullback f S hy
|
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
·
|
Mathlib.CategoryTheory.Sites.Coverage.288_0.qkZFgqEgDC2P633
|
instance : SemilatticeSup (Coverage C) where
sup x y
|
Mathlib_CategoryTheory_Sites_Coverage
|
case inr.intro
C : Type ?u.32430
inst✝ : Category.{?u.32434, ?u.32430} C
x y : Coverage C
X Y : C
f : Y ⟶ X
S : Presieve X
hy : S ∈ covering y X
T : Presieve Y
hT : T ∈ covering y Y ∧ Presieve.FactorsThruAlong T S f
⊢ ∃ T ∈ (fun B => covering x B ∪ covering y B) Y, Presieve.FactorsThruAlong T S f
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
· obtain ⟨T, hT⟩ := y.pullback f S hy
|
exact ⟨T, Or.inr hT.1, hT.2⟩
|
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
· obtain ⟨T, hT⟩ := y.pullback f S hy
|
Mathlib.CategoryTheory.Sites.Coverage.288_0.qkZFgqEgDC2P633
|
instance : SemilatticeSup (Coverage C) where
sup x y
|
Mathlib_CategoryTheory_Sites_Coverage
|
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
P : Cᵒᵖ ⥤ Type w
⊢ IsSheaf (toGrothendieck C K) P ↔ ∀ {X : C}, ∀ R ∈ covering K X, IsSheafFor P R
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
· obtain ⟨T, hT⟩ := y.pullback f S hy
exact ⟨T, Or.inr hT.1, hT.2⟩ }
toPartialOrder := inferInstance
le_sup_left _ _ _ := Set.subset_union_left _ _
le_sup_right _ _ _ := Set.subset_union_right _ _
sup_le _ _ _ hx hy X := Set.union_subset_iff.mpr ⟨hx X, hy X⟩
@[simp]
lemma sup_covering (x y : Coverage C) (B : C) :
(x ⊔ y).covering B = x.covering B ∪ y.covering B :=
rfl
end Coverage
open Coverage
namespace Presieve
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
|
constructor
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
|
Mathlib.CategoryTheory.Sites.Coverage.313_0.qkZFgqEgDC2P633
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R)
|
Mathlib_CategoryTheory_Sites_Coverage
|
case mp
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
P : Cᵒᵖ ⥤ Type w
⊢ IsSheaf (toGrothendieck C K) P → ∀ {X : C}, ∀ R ∈ covering K X, IsSheafFor P R
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
· obtain ⟨T, hT⟩ := y.pullback f S hy
exact ⟨T, Or.inr hT.1, hT.2⟩ }
toPartialOrder := inferInstance
le_sup_left _ _ _ := Set.subset_union_left _ _
le_sup_right _ _ _ := Set.subset_union_right _ _
sup_le _ _ _ hx hy X := Set.union_subset_iff.mpr ⟨hx X, hy X⟩
@[simp]
lemma sup_covering (x y : Coverage C) (B : C) :
(x ⊔ y).covering B = x.covering B ∪ y.covering B :=
rfl
end Coverage
open Coverage
namespace Presieve
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
·
|
intro H X R hR
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
·
|
Mathlib.CategoryTheory.Sites.Coverage.313_0.qkZFgqEgDC2P633
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R)
|
Mathlib_CategoryTheory_Sites_Coverage
|
case mp
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
P : Cᵒᵖ ⥤ Type w
H : IsSheaf (toGrothendieck C K) P
X : C
R : Presieve X
hR : R ∈ covering K X
⊢ IsSheafFor P R
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
· obtain ⟨T, hT⟩ := y.pullback f S hy
exact ⟨T, Or.inr hT.1, hT.2⟩ }
toPartialOrder := inferInstance
le_sup_left _ _ _ := Set.subset_union_left _ _
le_sup_right _ _ _ := Set.subset_union_right _ _
sup_le _ _ _ hx hy X := Set.union_subset_iff.mpr ⟨hx X, hy X⟩
@[simp]
lemma sup_covering (x y : Coverage C) (B : C) :
(x ⊔ y).covering B = x.covering B ∪ y.covering B :=
rfl
end Coverage
open Coverage
namespace Presieve
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
|
rw [Presieve.isSheafFor_iff_generate]
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
|
Mathlib.CategoryTheory.Sites.Coverage.313_0.qkZFgqEgDC2P633
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R)
|
Mathlib_CategoryTheory_Sites_Coverage
|
case mp
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
P : Cᵒᵖ ⥤ Type w
H : IsSheaf (toGrothendieck C K) P
X : C
R : Presieve X
hR : R ∈ covering K X
⊢ IsSheafFor P (Sieve.generate R).arrows
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
· obtain ⟨T, hT⟩ := y.pullback f S hy
exact ⟨T, Or.inr hT.1, hT.2⟩ }
toPartialOrder := inferInstance
le_sup_left _ _ _ := Set.subset_union_left _ _
le_sup_right _ _ _ := Set.subset_union_right _ _
sup_le _ _ _ hx hy X := Set.union_subset_iff.mpr ⟨hx X, hy X⟩
@[simp]
lemma sup_covering (x y : Coverage C) (B : C) :
(x ⊔ y).covering B = x.covering B ∪ y.covering B :=
rfl
end Coverage
open Coverage
namespace Presieve
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
|
apply H _ <| saturate.of _ _ hR
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
|
Mathlib.CategoryTheory.Sites.Coverage.313_0.qkZFgqEgDC2P633
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R)
|
Mathlib_CategoryTheory_Sites_Coverage
|
case mpr
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
P : Cᵒᵖ ⥤ Type w
⊢ (∀ {X : C}, ∀ R ∈ covering K X, IsSheafFor P R) → IsSheaf (toGrothendieck C K) P
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
· obtain ⟨T, hT⟩ := y.pullback f S hy
exact ⟨T, Or.inr hT.1, hT.2⟩ }
toPartialOrder := inferInstance
le_sup_left _ _ _ := Set.subset_union_left _ _
le_sup_right _ _ _ := Set.subset_union_right _ _
sup_le _ _ _ hx hy X := Set.union_subset_iff.mpr ⟨hx X, hy X⟩
@[simp]
lemma sup_covering (x y : Coverage C) (B : C) :
(x ⊔ y).covering B = x.covering B ∪ y.covering B :=
rfl
end Coverage
open Coverage
namespace Presieve
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
·
|
intro H X S hS
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
·
|
Mathlib.CategoryTheory.Sites.Coverage.313_0.qkZFgqEgDC2P633
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R)
|
Mathlib_CategoryTheory_Sites_Coverage
|
case mpr
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
P : Cᵒᵖ ⥤ Type w
H : ∀ {X : C}, ∀ R ∈ covering K X, IsSheafFor P R
X : C
S : Sieve X
hS : S ∈ GrothendieckTopology.sieves (toGrothendieck C K) X
⊢ IsSheafFor P S.arrows
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
· obtain ⟨T, hT⟩ := y.pullback f S hy
exact ⟨T, Or.inr hT.1, hT.2⟩ }
toPartialOrder := inferInstance
le_sup_left _ _ _ := Set.subset_union_left _ _
le_sup_right _ _ _ := Set.subset_union_right _ _
sup_le _ _ _ hx hy X := Set.union_subset_iff.mpr ⟨hx X, hy X⟩
@[simp]
lemma sup_covering (x y : Coverage C) (B : C) :
(x ⊔ y).covering B = x.covering B ∪ y.covering B :=
rfl
end Coverage
open Coverage
namespace Presieve
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
|
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
|
Mathlib.CategoryTheory.Sites.Coverage.313_0.qkZFgqEgDC2P633
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R)
|
Mathlib_CategoryTheory_Sites_Coverage
|
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
P : Cᵒᵖ ⥤ Type w
H : ∀ {X : C}, ∀ R ∈ covering K X, IsSheafFor P R
X : C
S : Sieve X
hS : S ∈ GrothendieckTopology.sieves (toGrothendieck C K) X
this : ∀ ⦃Y : C⦄ (f : Y ⟶ X), IsSheafFor P (Sieve.pullback f S).arrows
⊢ IsSheafFor P S.arrows
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
· obtain ⟨T, hT⟩ := y.pullback f S hy
exact ⟨T, Or.inr hT.1, hT.2⟩ }
toPartialOrder := inferInstance
le_sup_left _ _ _ := Set.subset_union_left _ _
le_sup_right _ _ _ := Set.subset_union_right _ _
sup_le _ _ _ hx hy X := Set.union_subset_iff.mpr ⟨hx X, hy X⟩
@[simp]
lemma sup_covering (x y : Coverage C) (B : C) :
(x ⊔ y).covering B = x.covering B ∪ y.covering B :=
rfl
end Coverage
open Coverage
namespace Presieve
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
|
simpa using this (f := 𝟙 _)
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
|
Mathlib.CategoryTheory.Sites.Coverage.313_0.qkZFgqEgDC2P633
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R)
|
Mathlib_CategoryTheory_Sites_Coverage
|
case mpr
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
P : Cᵒᵖ ⥤ Type w
H : ∀ {X : C}, ∀ R ∈ covering K X, IsSheafFor P R
X : C
S : Sieve X
hS : S ∈ GrothendieckTopology.sieves (toGrothendieck C K) X
⊢ ∀ ⦃Y : C⦄ (f : Y ⟶ X), IsSheafFor P (Sieve.pullback f S).arrows
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
· obtain ⟨T, hT⟩ := y.pullback f S hy
exact ⟨T, Or.inr hT.1, hT.2⟩ }
toPartialOrder := inferInstance
le_sup_left _ _ _ := Set.subset_union_left _ _
le_sup_right _ _ _ := Set.subset_union_right _ _
sup_le _ _ _ hx hy X := Set.union_subset_iff.mpr ⟨hx X, hy X⟩
@[simp]
lemma sup_covering (x y : Coverage C) (B : C) :
(x ⊔ y).covering B = x.covering B ∪ y.covering B :=
rfl
end Coverage
open Coverage
namespace Presieve
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
|
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← op_comp, h₁, h₂]
simpa only [Sieve.pullback_apply, Category.assoc] using hgg
let y : ∀ ⦃Z : C⦄ (g : Z ⟶ Y),
((S.pullback (g ≫ f)).pullback (𝟙 _)).arrows.FamilyOfElements P :=
fun Z g ZZ gg hgg => x (gg ≫ g) (by simpa using hgg)
have hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), (y g).Compatible := by
intro Z g Y₁ Y₂ ZZ g₁ g₂ f₁ f₂ h₁ h₂ h
rw [hx]
rw [reassoc_of% h]
choose z hz using fun ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : R.pullback f g) =>
H2' hg (𝟙 _) (y g) (hy g)
let q : (R.pullback f).arrows.FamilyOfElements P := fun Z g hg => z hg
have hq : q.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
apply (H2 h₁ g₁).ext
intro ZZ gg hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← P.map_comp, ← op_comp, ← op_comp, hz, hz]
· dsimp; congr 1; simp only [Category.assoc, h]
· simpa [reassoc_of% h] using hgg
· simpa using hgg
obtain ⟨t, ht⟩ := H1' f q hq
refine ⟨t, fun Z g hg => ?_⟩
refine (H1 (g ≫ f)).ext (fun ZZ gg hgg => ?_)
rw [← types_comp_apply _ (P.map gg.op), ← P.map_comp, ← op_comp, ht]
swap; simpa using hgg
refine (H2 hgg (𝟙 _)).ext (fun ZZZ ggg hggg => ?_)
rw [← types_comp_apply _ (P.map ggg.op), ← P.map_comp, ← op_comp, hz]
swap; simpa using hggg
refine (H2 hgg ggg).ext (fun ZZZZ gggg _ => ?_)
rw [← types_comp_apply _ (P.map gggg.op), ← P.map_comp, ← op_comp]
apply hx
simp
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
|
Mathlib.CategoryTheory.Sites.Coverage.313_0.qkZFgqEgDC2P633
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R)
|
Mathlib_CategoryTheory_Sites_Coverage
|
case mpr
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
P : Cᵒᵖ ⥤ Type w
H : ∀ {X : C}, ∀ R ∈ covering K X, IsSheafFor P R
X : C
S : Sieve X
hS : S ∈ GrothendieckTopology.sieves (toGrothendieck C K) X
⊢ ∀ ⦃Y : C⦄ (f : Y ⟶ X), IsSheafFor P (Sieve.pullback f S).arrows
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
· obtain ⟨T, hT⟩ := y.pullback f S hy
exact ⟨T, Or.inr hT.1, hT.2⟩ }
toPartialOrder := inferInstance
le_sup_left _ _ _ := Set.subset_union_left _ _
le_sup_right _ _ _ := Set.subset_union_right _ _
sup_le _ _ _ hx hy X := Set.union_subset_iff.mpr ⟨hx X, hy X⟩
@[simp]
lemma sup_covering (x y : Coverage C) (B : C) :
(x ⊔ y).covering B = x.covering B ∪ y.covering B :=
rfl
end Coverage
open Coverage
namespace Presieve
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
|
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← op_comp, h₁, h₂]
simpa only [Sieve.pullback_apply, Category.assoc] using hgg
let y : ∀ ⦃Z : C⦄ (g : Z ⟶ Y),
((S.pullback (g ≫ f)).pullback (𝟙 _)).arrows.FamilyOfElements P :=
fun Z g ZZ gg hgg => x (gg ≫ g) (by simpa using hgg)
have hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), (y g).Compatible := by
intro Z g Y₁ Y₂ ZZ g₁ g₂ f₁ f₂ h₁ h₂ h
rw [hx]
rw [reassoc_of% h]
choose z hz using fun ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : R.pullback f g) =>
H2' hg (𝟙 _) (y g) (hy g)
let q : (R.pullback f).arrows.FamilyOfElements P := fun Z g hg => z hg
have hq : q.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
apply (H2 h₁ g₁).ext
intro ZZ gg hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← P.map_comp, ← op_comp, ← op_comp, hz, hz]
· dsimp; congr 1; simp only [Category.assoc, h]
· simpa [reassoc_of% h] using hgg
· simpa using hgg
obtain ⟨t, ht⟩ := H1' f q hq
refine ⟨t, fun Z g hg => ?_⟩
refine (H1 (g ≫ f)).ext (fun ZZ gg hgg => ?_)
rw [← types_comp_apply _ (P.map gg.op), ← P.map_comp, ← op_comp, ht]
swap; simpa using hgg
refine (H2 hgg (𝟙 _)).ext (fun ZZZ ggg hggg => ?_)
rw [← types_comp_apply _ (P.map ggg.op), ← P.map_comp, ← op_comp, hz]
swap; simpa using hggg
refine (H2 hgg ggg).ext (fun ZZZZ gggg _ => ?_)
rw [← types_comp_apply _ (P.map gggg.op), ← P.map_comp, ← op_comp]
apply hx
simp
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
|
Mathlib.CategoryTheory.Sites.Coverage.313_0.qkZFgqEgDC2P633
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R)
|
Mathlib_CategoryTheory_Sites_Coverage
|
case mpr.of
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
P : Cᵒᵖ ⥤ Type w
H : ∀ {X : C}, ∀ R ∈ covering K X, IsSheafFor P R
X✝ : C
S✝ : Sieve X✝
X : C
S : Presieve X
hS : S ∈ covering K X
⊢ ∀ ⦃Y : C⦄ (f : Y ⟶ X), IsSheafFor P (Sieve.pullback f (Sieve.generate S)).arrows
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
· obtain ⟨T, hT⟩ := y.pullback f S hy
exact ⟨T, Or.inr hT.1, hT.2⟩ }
toPartialOrder := inferInstance
le_sup_left _ _ _ := Set.subset_union_left _ _
le_sup_right _ _ _ := Set.subset_union_right _ _
sup_le _ _ _ hx hy X := Set.union_subset_iff.mpr ⟨hx X, hy X⟩
@[simp]
lemma sup_covering (x y : Coverage C) (B : C) :
(x ⊔ y).covering B = x.covering B ∪ y.covering B :=
rfl
end Coverage
open Coverage
namespace Presieve
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
|
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
|
Mathlib.CategoryTheory.Sites.Coverage.313_0.qkZFgqEgDC2P633
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R)
|
Mathlib_CategoryTheory_Sites_Coverage
|
case mpr.of
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
P : Cᵒᵖ ⥤ Type w
H : ∀ {X : C}, ∀ R ∈ covering K X, IsSheafFor P R
X✝ : C
S✝ : Sieve X✝
X : C
S : Presieve X
hS : S ∈ covering K X
⊢ ∀ ⦃Y : C⦄ (f : Y ⟶ X), IsSheafFor P (Sieve.pullback f (Sieve.generate S)).arrows
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
· obtain ⟨T, hT⟩ := y.pullback f S hy
exact ⟨T, Or.inr hT.1, hT.2⟩ }
toPartialOrder := inferInstance
le_sup_left _ _ _ := Set.subset_union_left _ _
le_sup_right _ _ _ := Set.subset_union_right _ _
sup_le _ _ _ hx hy X := Set.union_subset_iff.mpr ⟨hx X, hy X⟩
@[simp]
lemma sup_covering (x y : Coverage C) (B : C) :
(x ⊔ y).covering B = x.covering B ∪ y.covering B :=
rfl
end Coverage
open Coverage
namespace Presieve
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
|
intro Y f
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
|
Mathlib.CategoryTheory.Sites.Coverage.313_0.qkZFgqEgDC2P633
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R)
|
Mathlib_CategoryTheory_Sites_Coverage
|
case mpr.of
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
P : Cᵒᵖ ⥤ Type w
H : ∀ {X : C}, ∀ R ∈ covering K X, IsSheafFor P R
X✝ : C
S✝ : Sieve X✝
X : C
S : Presieve X
hS : S ∈ covering K X
Y : C
f : Y ⟶ X
⊢ IsSheafFor P (Sieve.pullback f (Sieve.generate S)).arrows
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
· obtain ⟨T, hT⟩ := y.pullback f S hy
exact ⟨T, Or.inr hT.1, hT.2⟩ }
toPartialOrder := inferInstance
le_sup_left _ _ _ := Set.subset_union_left _ _
le_sup_right _ _ _ := Set.subset_union_right _ _
sup_le _ _ _ hx hy X := Set.union_subset_iff.mpr ⟨hx X, hy X⟩
@[simp]
lemma sup_covering (x y : Coverage C) (B : C) :
(x ⊔ y).covering B = x.covering B ∪ y.covering B :=
rfl
end Coverage
open Coverage
namespace Presieve
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
|
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
|
Mathlib.CategoryTheory.Sites.Coverage.313_0.qkZFgqEgDC2P633
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R)
|
Mathlib_CategoryTheory_Sites_Coverage
|
case mpr.of.intro.intro
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
P : Cᵒᵖ ⥤ Type w
H : ∀ {X : C}, ∀ R ∈ covering K X, IsSheafFor P R
X✝ : C
S✝ : Sieve X✝
X : C
S : Presieve X
hS : S ∈ covering K X
Y : C
f : Y ⟶ X
T : Presieve Y
hT1 : T ∈ covering K Y
hT2 : FactorsThruAlong T S f
⊢ IsSheafFor P (Sieve.pullback f (Sieve.generate S)).arrows
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
· obtain ⟨T, hT⟩ := y.pullback f S hy
exact ⟨T, Or.inr hT.1, hT.2⟩ }
toPartialOrder := inferInstance
le_sup_left _ _ _ := Set.subset_union_left _ _
le_sup_right _ _ _ := Set.subset_union_right _ _
sup_le _ _ _ hx hy X := Set.union_subset_iff.mpr ⟨hx X, hy X⟩
@[simp]
lemma sup_covering (x y : Coverage C) (B : C) :
(x ⊔ y).covering B = x.covering B ∪ y.covering B :=
rfl
end Coverage
open Coverage
namespace Presieve
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
|
apply Presieve.isSheafFor_of_factorsThru (S := T)
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
|
Mathlib.CategoryTheory.Sites.Coverage.313_0.qkZFgqEgDC2P633
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R)
|
Mathlib_CategoryTheory_Sites_Coverage
|
case mpr.of.intro.intro.H
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
P : Cᵒᵖ ⥤ Type w
H : ∀ {X : C}, ∀ R ∈ covering K X, IsSheafFor P R
X✝ : C
S✝ : Sieve X✝
X : C
S : Presieve X
hS : S ∈ covering K X
Y : C
f : Y ⟶ X
T : Presieve Y
hT1 : T ∈ covering K Y
hT2 : FactorsThruAlong T S f
⊢ FactorsThru T (Sieve.pullback f (Sieve.generate S)).arrows
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
· obtain ⟨T, hT⟩ := y.pullback f S hy
exact ⟨T, Or.inr hT.1, hT.2⟩ }
toPartialOrder := inferInstance
le_sup_left _ _ _ := Set.subset_union_left _ _
le_sup_right _ _ _ := Set.subset_union_right _ _
sup_le _ _ _ hx hy X := Set.union_subset_iff.mpr ⟨hx X, hy X⟩
@[simp]
lemma sup_covering (x y : Coverage C) (B : C) :
(x ⊔ y).covering B = x.covering B ∪ y.covering B :=
rfl
end Coverage
open Coverage
namespace Presieve
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
·
|
intro Z g hg
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
·
|
Mathlib.CategoryTheory.Sites.Coverage.313_0.qkZFgqEgDC2P633
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R)
|
Mathlib_CategoryTheory_Sites_Coverage
|
case mpr.of.intro.intro.H
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
P : Cᵒᵖ ⥤ Type w
H : ∀ {X : C}, ∀ R ∈ covering K X, IsSheafFor P R
X✝ : C
S✝ : Sieve X✝
X : C
S : Presieve X
hS : S ∈ covering K X
Y : C
f : Y ⟶ X
T : Presieve Y
hT1 : T ∈ covering K Y
hT2 : FactorsThruAlong T S f
Z : C
g : Z ⟶ Y
hg : T g
⊢ ∃ W i e, (Sieve.pullback f (Sieve.generate S)).arrows e ∧ i ≫ e = g
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
· obtain ⟨T, hT⟩ := y.pullback f S hy
exact ⟨T, Or.inr hT.1, hT.2⟩ }
toPartialOrder := inferInstance
le_sup_left _ _ _ := Set.subset_union_left _ _
le_sup_right _ _ _ := Set.subset_union_right _ _
sup_le _ _ _ hx hy X := Set.union_subset_iff.mpr ⟨hx X, hy X⟩
@[simp]
lemma sup_covering (x y : Coverage C) (B : C) :
(x ⊔ y).covering B = x.covering B ∪ y.covering B :=
rfl
end Coverage
open Coverage
namespace Presieve
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
|
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
|
Mathlib.CategoryTheory.Sites.Coverage.313_0.qkZFgqEgDC2P633
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R)
|
Mathlib_CategoryTheory_Sites_Coverage
|
case mpr.of.intro.intro.H.intro.intro.intro.intro
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
P : Cᵒᵖ ⥤ Type w
H : ∀ {X : C}, ∀ R ∈ covering K X, IsSheafFor P R
X✝ : C
S✝ : Sieve X✝
X : C
S : Presieve X
hS : S ∈ covering K X
Y : C
f : Y ⟶ X
T : Presieve Y
hT1 : T ∈ covering K Y
hT2 : FactorsThruAlong T S f
Z : C
g : Z ⟶ Y
hg : T g
W : C
i : Z ⟶ W
e : W ⟶ X
h1 : S e
h2 : i ≫ e = g ≫ f
⊢ ∃ W i e, (Sieve.pullback f (Sieve.generate S)).arrows e ∧ i ≫ e = g
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
· obtain ⟨T, hT⟩ := y.pullback f S hy
exact ⟨T, Or.inr hT.1, hT.2⟩ }
toPartialOrder := inferInstance
le_sup_left _ _ _ := Set.subset_union_left _ _
le_sup_right _ _ _ := Set.subset_union_right _ _
sup_le _ _ _ hx hy X := Set.union_subset_iff.mpr ⟨hx X, hy X⟩
@[simp]
lemma sup_covering (x y : Coverage C) (B : C) :
(x ⊔ y).covering B = x.covering B ∪ y.covering B :=
rfl
end Coverage
open Coverage
namespace Presieve
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
|
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
|
Mathlib.CategoryTheory.Sites.Coverage.313_0.qkZFgqEgDC2P633
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R)
|
Mathlib_CategoryTheory_Sites_Coverage
|
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
P : Cᵒᵖ ⥤ Type w
H : ∀ {X : C}, ∀ R ∈ covering K X, IsSheafFor P R
X✝ : C
S✝ : Sieve X✝
X : C
S : Presieve X
hS : S ∈ covering K X
Y : C
f : Y ⟶ X
T : Presieve Y
hT1 : T ∈ covering K Y
hT2 : FactorsThruAlong T S f
Z : C
g : Z ⟶ Y
hg : T g
W : C
i : Z ⟶ W
e : W ⟶ X
h1 : S e
h2 : i ≫ e = g ≫ f
⊢ 𝟙 Z ≫ g = g
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
· obtain ⟨T, hT⟩ := y.pullback f S hy
exact ⟨T, Or.inr hT.1, hT.2⟩ }
toPartialOrder := inferInstance
le_sup_left _ _ _ := Set.subset_union_left _ _
le_sup_right _ _ _ := Set.subset_union_right _ _
sup_le _ _ _ hx hy X := Set.union_subset_iff.mpr ⟨hx X, hy X⟩
@[simp]
lemma sup_covering (x y : Coverage C) (B : C) :
(x ⊔ y).covering B = x.covering B ∪ y.covering B :=
rfl
end Coverage
open Coverage
namespace Presieve
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by
|
simp
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by
|
Mathlib.CategoryTheory.Sites.Coverage.313_0.qkZFgqEgDC2P633
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R)
|
Mathlib_CategoryTheory_Sites_Coverage
|
case mpr.of.intro.intro.hS
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
P : Cᵒᵖ ⥤ Type w
H : ∀ {X : C}, ∀ R ∈ covering K X, IsSheafFor P R
X✝ : C
S✝ : Sieve X✝
X : C
S : Presieve X
hS : S ∈ covering K X
Y : C
f : Y ⟶ X
T : Presieve Y
hT1 : T ∈ covering K Y
hT2 : FactorsThruAlong T S f
⊢ IsSheafFor P T
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
· obtain ⟨T, hT⟩ := y.pullback f S hy
exact ⟨T, Or.inr hT.1, hT.2⟩ }
toPartialOrder := inferInstance
le_sup_left _ _ _ := Set.subset_union_left _ _
le_sup_right _ _ _ := Set.subset_union_right _ _
sup_le _ _ _ hx hy X := Set.union_subset_iff.mpr ⟨hx X, hy X⟩
@[simp]
lemma sup_covering (x y : Coverage C) (B : C) :
(x ⊔ y).covering B = x.covering B ∪ y.covering B :=
rfl
end Coverage
open Coverage
namespace Presieve
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
·
|
apply H
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
·
|
Mathlib.CategoryTheory.Sites.Coverage.313_0.qkZFgqEgDC2P633
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R)
|
Mathlib_CategoryTheory_Sites_Coverage
|
case mpr.of.intro.intro.hS.a
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
P : Cᵒᵖ ⥤ Type w
H : ∀ {X : C}, ∀ R ∈ covering K X, IsSheafFor P R
X✝ : C
S✝ : Sieve X✝
X : C
S : Presieve X
hS : S ∈ covering K X
Y : C
f : Y ⟶ X
T : Presieve Y
hT1 : T ∈ covering K Y
hT2 : FactorsThruAlong T S f
⊢ T ∈ covering K Y
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
· obtain ⟨T, hT⟩ := y.pullback f S hy
exact ⟨T, Or.inr hT.1, hT.2⟩ }
toPartialOrder := inferInstance
le_sup_left _ _ _ := Set.subset_union_left _ _
le_sup_right _ _ _ := Set.subset_union_right _ _
sup_le _ _ _ hx hy X := Set.union_subset_iff.mpr ⟨hx X, hy X⟩
@[simp]
lemma sup_covering (x y : Coverage C) (B : C) :
(x ⊔ y).covering B = x.covering B ∪ y.covering B :=
rfl
end Coverage
open Coverage
namespace Presieve
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H;
|
assumption
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H;
|
Mathlib.CategoryTheory.Sites.Coverage.313_0.qkZFgqEgDC2P633
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R)
|
Mathlib_CategoryTheory_Sites_Coverage
|
case mpr.of.intro.intro.h
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
P : Cᵒᵖ ⥤ Type w
H : ∀ {X : C}, ∀ R ∈ covering K X, IsSheafFor P R
X✝ : C
S✝ : Sieve X✝
X : C
S : Presieve X
hS : S ∈ covering K X
Y : C
f : Y ⟶ X
T : Presieve Y
hT1 : T ∈ covering K Y
hT2 : FactorsThruAlong T S f
⊢ ∀ ⦃Y_1 : C⦄ ⦃f_1 : Y_1 ⟶ Y⦄,
(Sieve.pullback f (Sieve.generate S)).arrows f_1 → ∃ R, IsSeparatedFor P R ∧ FactorsThruAlong R T f_1
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
· obtain ⟨T, hT⟩ := y.pullback f S hy
exact ⟨T, Or.inr hT.1, hT.2⟩ }
toPartialOrder := inferInstance
le_sup_left _ _ _ := Set.subset_union_left _ _
le_sup_right _ _ _ := Set.subset_union_right _ _
sup_le _ _ _ hx hy X := Set.union_subset_iff.mpr ⟨hx X, hy X⟩
@[simp]
lemma sup_covering (x y : Coverage C) (B : C) :
(x ⊔ y).covering B = x.covering B ∪ y.covering B :=
rfl
end Coverage
open Coverage
namespace Presieve
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
·
|
intro Z g _
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
·
|
Mathlib.CategoryTheory.Sites.Coverage.313_0.qkZFgqEgDC2P633
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R)
|
Mathlib_CategoryTheory_Sites_Coverage
|
case mpr.of.intro.intro.h
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
P : Cᵒᵖ ⥤ Type w
H : ∀ {X : C}, ∀ R ∈ covering K X, IsSheafFor P R
X✝ : C
S✝ : Sieve X✝
X : C
S : Presieve X
hS : S ∈ covering K X
Y : C
f : Y ⟶ X
T : Presieve Y
hT1 : T ∈ covering K Y
hT2 : FactorsThruAlong T S f
Z : C
g : Z ⟶ Y
a✝ : (Sieve.pullback f (Sieve.generate S)).arrows g
⊢ ∃ R, IsSeparatedFor P R ∧ FactorsThruAlong R T g
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
· obtain ⟨T, hT⟩ := y.pullback f S hy
exact ⟨T, Or.inr hT.1, hT.2⟩ }
toPartialOrder := inferInstance
le_sup_left _ _ _ := Set.subset_union_left _ _
le_sup_right _ _ _ := Set.subset_union_right _ _
sup_le _ _ _ hx hy X := Set.union_subset_iff.mpr ⟨hx X, hy X⟩
@[simp]
lemma sup_covering (x y : Coverage C) (B : C) :
(x ⊔ y).covering B = x.covering B ∪ y.covering B :=
rfl
end Coverage
open Coverage
namespace Presieve
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
|
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
|
Mathlib.CategoryTheory.Sites.Coverage.313_0.qkZFgqEgDC2P633
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R)
|
Mathlib_CategoryTheory_Sites_Coverage
|
case mpr.of.intro.intro.h.intro.intro
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
P : Cᵒᵖ ⥤ Type w
H : ∀ {X : C}, ∀ R ∈ covering K X, IsSheafFor P R
X✝ : C
S✝ : Sieve X✝
X : C
S : Presieve X
hS : S ∈ covering K X
Y : C
f : Y ⟶ X
T : Presieve Y
hT1 : T ∈ covering K Y
hT2 : FactorsThruAlong T S f
Z : C
g : Z ⟶ Y
a✝ : (Sieve.pullback f (Sieve.generate S)).arrows g
R : Presieve Z
hR1 : R ∈ covering K Z
hR2 : FactorsThruAlong R T g
⊢ ∃ R, IsSeparatedFor P R ∧ FactorsThruAlong R T g
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
· obtain ⟨T, hT⟩ := y.pullback f S hy
exact ⟨T, Or.inr hT.1, hT.2⟩ }
toPartialOrder := inferInstance
le_sup_left _ _ _ := Set.subset_union_left _ _
le_sup_right _ _ _ := Set.subset_union_right _ _
sup_le _ _ _ hx hy X := Set.union_subset_iff.mpr ⟨hx X, hy X⟩
@[simp]
lemma sup_covering (x y : Coverage C) (B : C) :
(x ⊔ y).covering B = x.covering B ∪ y.covering B :=
rfl
end Coverage
open Coverage
namespace Presieve
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
|
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
|
Mathlib.CategoryTheory.Sites.Coverage.313_0.qkZFgqEgDC2P633
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R)
|
Mathlib_CategoryTheory_Sites_Coverage
|
case mpr.top
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
P : Cᵒᵖ ⥤ Type w
H : ∀ {X : C}, ∀ R ∈ covering K X, IsSheafFor P R
X : C
S : Sieve X
X✝ : C
⊢ ∀ ⦃Y : C⦄ (f : Y ⟶ X✝), IsSheafFor P (Sieve.pullback f ⊤).arrows
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
· obtain ⟨T, hT⟩ := y.pullback f S hy
exact ⟨T, Or.inr hT.1, hT.2⟩ }
toPartialOrder := inferInstance
le_sup_left _ _ _ := Set.subset_union_left _ _
le_sup_right _ _ _ := Set.subset_union_right _ _
sup_le _ _ _ hx hy X := Set.union_subset_iff.mpr ⟨hx X, hy X⟩
@[simp]
lemma sup_covering (x y : Coverage C) (B : C) :
(x ⊔ y).covering B = x.covering B ∪ y.covering B :=
rfl
end Coverage
open Coverage
namespace Presieve
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
|
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
|
Mathlib.CategoryTheory.Sites.Coverage.313_0.qkZFgqEgDC2P633
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R)
|
Mathlib_CategoryTheory_Sites_Coverage
|
case mpr.top
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
P : Cᵒᵖ ⥤ Type w
H : ∀ {X : C}, ∀ R ∈ covering K X, IsSheafFor P R
X : C
S : Sieve X
X✝ : C
⊢ ∀ ⦃Y : C⦄ (f : Y ⟶ X✝), IsSheafFor P (Sieve.pullback f ⊤).arrows
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
· obtain ⟨T, hT⟩ := y.pullback f S hy
exact ⟨T, Or.inr hT.1, hT.2⟩ }
toPartialOrder := inferInstance
le_sup_left _ _ _ := Set.subset_union_left _ _
le_sup_right _ _ _ := Set.subset_union_right _ _
sup_le _ _ _ hx hy X := Set.union_subset_iff.mpr ⟨hx X, hy X⟩
@[simp]
lemma sup_covering (x y : Coverage C) (B : C) :
(x ⊔ y).covering B = x.covering B ∪ y.covering B :=
rfl
end Coverage
open Coverage
namespace Presieve
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top =>
|
intros
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top =>
|
Mathlib.CategoryTheory.Sites.Coverage.313_0.qkZFgqEgDC2P633
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R)
|
Mathlib_CategoryTheory_Sites_Coverage
|
case mpr.top
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
P : Cᵒᵖ ⥤ Type w
H : ∀ {X : C}, ∀ R ∈ covering K X, IsSheafFor P R
X : C
S : Sieve X
X✝ Y✝ : C
f✝ : Y✝ ⟶ X✝
⊢ IsSheafFor P (Sieve.pullback f✝ ⊤).arrows
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
· obtain ⟨T, hT⟩ := y.pullback f S hy
exact ⟨T, Or.inr hT.1, hT.2⟩ }
toPartialOrder := inferInstance
le_sup_left _ _ _ := Set.subset_union_left _ _
le_sup_right _ _ _ := Set.subset_union_right _ _
sup_le _ _ _ hx hy X := Set.union_subset_iff.mpr ⟨hx X, hy X⟩
@[simp]
lemma sup_covering (x y : Coverage C) (B : C) :
(x ⊔ y).covering B = x.covering B ∪ y.covering B :=
rfl
end Coverage
open Coverage
namespace Presieve
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros;
|
simpa using Presieve.isSheafFor_top_sieve _
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros;
|
Mathlib.CategoryTheory.Sites.Coverage.313_0.qkZFgqEgDC2P633
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R)
|
Mathlib_CategoryTheory_Sites_Coverage
|
case mpr.transitive
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
P : Cᵒᵖ ⥤ Type w
H : ∀ {X : C}, ∀ R ∈ covering K X, IsSheafFor P R
X✝ : C
S✝ : Sieve X✝
X : C
R S : Sieve X
a✝¹ : saturate K X R
a✝ : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R.arrows f → saturate K Y (Sieve.pullback f S)
H1 : ∀ ⦃Y : C⦄ (f : Y ⟶ X), IsSheafFor P (Sieve.pullback f R).arrows
H2 :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f → ∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y), IsSheafFor P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows
⊢ ∀ ⦃Y : C⦄ (f : Y ⟶ X), IsSheafFor P (Sieve.pullback f S).arrows
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
· obtain ⟨T, hT⟩ := y.pullback f S hy
exact ⟨T, Or.inr hT.1, hT.2⟩ }
toPartialOrder := inferInstance
le_sup_left _ _ _ := Set.subset_union_left _ _
le_sup_right _ _ _ := Set.subset_union_right _ _
sup_le _ _ _ hx hy X := Set.union_subset_iff.mpr ⟨hx X, hy X⟩
@[simp]
lemma sup_covering (x y : Coverage C) (B : C) :
(x ⊔ y).covering B = x.covering B ∪ y.covering B :=
rfl
end Coverage
open Coverage
namespace Presieve
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
|
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← op_comp, h₁, h₂]
simpa only [Sieve.pullback_apply, Category.assoc] using hgg
let y : ∀ ⦃Z : C⦄ (g : Z ⟶ Y),
((S.pullback (g ≫ f)).pullback (𝟙 _)).arrows.FamilyOfElements P :=
fun Z g ZZ gg hgg => x (gg ≫ g) (by simpa using hgg)
have hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), (y g).Compatible := by
intro Z g Y₁ Y₂ ZZ g₁ g₂ f₁ f₂ h₁ h₂ h
rw [hx]
rw [reassoc_of% h]
choose z hz using fun ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : R.pullback f g) =>
H2' hg (𝟙 _) (y g) (hy g)
let q : (R.pullback f).arrows.FamilyOfElements P := fun Z g hg => z hg
have hq : q.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
apply (H2 h₁ g₁).ext
intro ZZ gg hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← P.map_comp, ← op_comp, ← op_comp, hz, hz]
· dsimp; congr 1; simp only [Category.assoc, h]
· simpa [reassoc_of% h] using hgg
· simpa using hgg
obtain ⟨t, ht⟩ := H1' f q hq
refine ⟨t, fun Z g hg => ?_⟩
refine (H1 (g ≫ f)).ext (fun ZZ gg hgg => ?_)
rw [← types_comp_apply _ (P.map gg.op), ← P.map_comp, ← op_comp, ht]
swap; simpa using hgg
refine (H2 hgg (𝟙 _)).ext (fun ZZZ ggg hggg => ?_)
rw [← types_comp_apply _ (P.map ggg.op), ← P.map_comp, ← op_comp, hz]
swap; simpa using hggg
refine (H2 hgg ggg).ext (fun ZZZZ gggg _ => ?_)
rw [← types_comp_apply _ (P.map gggg.op), ← P.map_comp, ← op_comp]
apply hx
simp
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
|
Mathlib.CategoryTheory.Sites.Coverage.313_0.qkZFgqEgDC2P633
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R)
|
Mathlib_CategoryTheory_Sites_Coverage
|
case mpr.transitive
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
P : Cᵒᵖ ⥤ Type w
H : ∀ {X : C}, ∀ R ∈ covering K X, IsSheafFor P R
X✝ : C
S✝ : Sieve X✝
X : C
R S : Sieve X
a✝¹ : saturate K X R
a✝ : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R.arrows f → saturate K Y (Sieve.pullback f S)
H1 : ∀ ⦃Y : C⦄ (f : Y ⟶ X), IsSheafFor P (Sieve.pullback f R).arrows
H2 :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f → ∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y), IsSheafFor P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows
⊢ ∀ ⦃Y : C⦄ (f : Y ⟶ X), IsSheafFor P (Sieve.pullback f S).arrows
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
· obtain ⟨T, hT⟩ := y.pullback f S hy
exact ⟨T, Or.inr hT.1, hT.2⟩ }
toPartialOrder := inferInstance
le_sup_left _ _ _ := Set.subset_union_left _ _
le_sup_right _ _ _ := Set.subset_union_right _ _
sup_le _ _ _ hx hy X := Set.union_subset_iff.mpr ⟨hx X, hy X⟩
@[simp]
lemma sup_covering (x y : Coverage C) (B : C) :
(x ⊔ y).covering B = x.covering B ∪ y.covering B :=
rfl
end Coverage
open Coverage
namespace Presieve
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
|
intro Y f
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
|
Mathlib.CategoryTheory.Sites.Coverage.313_0.qkZFgqEgDC2P633
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R)
|
Mathlib_CategoryTheory_Sites_Coverage
|
case mpr.transitive
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
P : Cᵒᵖ ⥤ Type w
H : ∀ {X : C}, ∀ R ∈ covering K X, IsSheafFor P R
X✝ : C
S✝ : Sieve X✝
X : C
R S : Sieve X
a✝¹ : saturate K X R
a✝ : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R.arrows f → saturate K Y (Sieve.pullback f S)
H1 : ∀ ⦃Y : C⦄ (f : Y ⟶ X), IsSheafFor P (Sieve.pullback f R).arrows
H2 :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f → ∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y), IsSheafFor P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows
Y : C
f : Y ⟶ X
⊢ IsSheafFor P (Sieve.pullback f S).arrows
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
· obtain ⟨T, hT⟩ := y.pullback f S hy
exact ⟨T, Or.inr hT.1, hT.2⟩ }
toPartialOrder := inferInstance
le_sup_left _ _ _ := Set.subset_union_left _ _
le_sup_right _ _ _ := Set.subset_union_right _ _
sup_le _ _ _ hx hy X := Set.union_subset_iff.mpr ⟨hx X, hy X⟩
@[simp]
lemma sup_covering (x y : Coverage C) (B : C) :
(x ⊔ y).covering B = x.covering B ∪ y.covering B :=
rfl
end Coverage
open Coverage
namespace Presieve
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
|
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
|
Mathlib.CategoryTheory.Sites.Coverage.313_0.qkZFgqEgDC2P633
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R)
|
Mathlib_CategoryTheory_Sites_Coverage
|
case mpr.transitive
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
P : Cᵒᵖ ⥤ Type w
X✝ : C
S✝ : Sieve X✝
X : C
R S : Sieve X
a✝¹ : saturate K X R
a✝ : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R.arrows f → saturate K Y (Sieve.pullback f S)
Y : C
f : Y ⟶ X
H :
∀ {X : C},
∀ R ∈ covering K X,
IsSeparatedFor P R ∧
∀ (x : FamilyOfElements P R), FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H1 :
∀ ⦃Y : C⦄ (f : Y ⟶ X),
IsSeparatedFor P (Sieve.pullback f R).arrows ∧
∀ (x : FamilyOfElements P (Sieve.pullback f R).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H2 :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f →
∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y),
IsSeparatedFor P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows ∧
∀ (x : FamilyOfElements P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
⊢ IsSeparatedFor P (Sieve.pullback f S).arrows ∧
∀ (x : FamilyOfElements P (Sieve.pullback f S).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
· obtain ⟨T, hT⟩ := y.pullback f S hy
exact ⟨T, Or.inr hT.1, hT.2⟩ }
toPartialOrder := inferInstance
le_sup_left _ _ _ := Set.subset_union_left _ _
le_sup_right _ _ _ := Set.subset_union_right _ _
sup_le _ _ _ hx hy X := Set.union_subset_iff.mpr ⟨hx X, hy X⟩
@[simp]
lemma sup_covering (x y : Coverage C) (B : C) :
(x ⊔ y).covering B = x.covering B ∪ y.covering B :=
rfl
end Coverage
open Coverage
namespace Presieve
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
|
choose H1 H1' using H1
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
|
Mathlib.CategoryTheory.Sites.Coverage.313_0.qkZFgqEgDC2P633
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R)
|
Mathlib_CategoryTheory_Sites_Coverage
|
case mpr.transitive
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
P : Cᵒᵖ ⥤ Type w
X✝ : C
S✝ : Sieve X✝
X : C
R S : Sieve X
a✝¹ : saturate K X R
a✝ : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R.arrows f → saturate K Y (Sieve.pullback f S)
Y : C
f : Y ⟶ X
H :
∀ {X : C},
∀ R ∈ covering K X,
IsSeparatedFor P R ∧
∀ (x : FamilyOfElements P R), FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H2 :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f →
∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y),
IsSeparatedFor P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows ∧
∀ (x : FamilyOfElements P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H1 : ∀ ⦃Y : C⦄ (f : Y ⟶ X), IsSeparatedFor P (Sieve.pullback f R).arrows
H1' :
∀ ⦃Y : C⦄ (f : Y ⟶ X) (x : FamilyOfElements P (Sieve.pullback f R).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
⊢ IsSeparatedFor P (Sieve.pullback f S).arrows ∧
∀ (x : FamilyOfElements P (Sieve.pullback f S).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
· obtain ⟨T, hT⟩ := y.pullback f S hy
exact ⟨T, Or.inr hT.1, hT.2⟩ }
toPartialOrder := inferInstance
le_sup_left _ _ _ := Set.subset_union_left _ _
le_sup_right _ _ _ := Set.subset_union_right _ _
sup_le _ _ _ hx hy X := Set.union_subset_iff.mpr ⟨hx X, hy X⟩
@[simp]
lemma sup_covering (x y : Coverage C) (B : C) :
(x ⊔ y).covering B = x.covering B ∪ y.covering B :=
rfl
end Coverage
open Coverage
namespace Presieve
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
|
choose H2 H2' using H2
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
|
Mathlib.CategoryTheory.Sites.Coverage.313_0.qkZFgqEgDC2P633
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R)
|
Mathlib_CategoryTheory_Sites_Coverage
|
case mpr.transitive
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
P : Cᵒᵖ ⥤ Type w
X✝ : C
S✝ : Sieve X✝
X : C
R S : Sieve X
a✝¹ : saturate K X R
a✝ : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R.arrows f → saturate K Y (Sieve.pullback f S)
Y : C
f : Y ⟶ X
H :
∀ {X : C},
∀ R ∈ covering K X,
IsSeparatedFor P R ∧
∀ (x : FamilyOfElements P R), FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H1 : ∀ ⦃Y : C⦄ (f : Y ⟶ X), IsSeparatedFor P (Sieve.pullback f R).arrows
H1' :
∀ ⦃Y : C⦄ (f : Y ⟶ X) (x : FamilyOfElements P (Sieve.pullback f R).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H2 :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f → ∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y), IsSeparatedFor P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows
H2' :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f →
∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y) (x : FamilyOfElements P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
⊢ IsSeparatedFor P (Sieve.pullback f S).arrows ∧
∀ (x : FamilyOfElements P (Sieve.pullback f S).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
· obtain ⟨T, hT⟩ := y.pullback f S hy
exact ⟨T, Or.inr hT.1, hT.2⟩ }
toPartialOrder := inferInstance
le_sup_left _ _ _ := Set.subset_union_left _ _
le_sup_right _ _ _ := Set.subset_union_right _ _
sup_le _ _ _ hx hy X := Set.union_subset_iff.mpr ⟨hx X, hy X⟩
@[simp]
lemma sup_covering (x y : Coverage C) (B : C) :
(x ⊔ y).covering B = x.covering B ∪ y.covering B :=
rfl
end Coverage
open Coverage
namespace Presieve
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
|
refine ⟨?_, fun x hx => ?_⟩
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
|
Mathlib.CategoryTheory.Sites.Coverage.313_0.qkZFgqEgDC2P633
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R)
|
Mathlib_CategoryTheory_Sites_Coverage
|
case mpr.transitive.refine_1
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
P : Cᵒᵖ ⥤ Type w
X✝ : C
S✝ : Sieve X✝
X : C
R S : Sieve X
a✝¹ : saturate K X R
a✝ : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R.arrows f → saturate K Y (Sieve.pullback f S)
Y : C
f : Y ⟶ X
H :
∀ {X : C},
∀ R ∈ covering K X,
IsSeparatedFor P R ∧
∀ (x : FamilyOfElements P R), FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H1 : ∀ ⦃Y : C⦄ (f : Y ⟶ X), IsSeparatedFor P (Sieve.pullback f R).arrows
H1' :
∀ ⦃Y : C⦄ (f : Y ⟶ X) (x : FamilyOfElements P (Sieve.pullback f R).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H2 :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f → ∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y), IsSeparatedFor P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows
H2' :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f →
∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y) (x : FamilyOfElements P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
⊢ IsSeparatedFor P (Sieve.pullback f S).arrows
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
· obtain ⟨T, hT⟩ := y.pullback f S hy
exact ⟨T, Or.inr hT.1, hT.2⟩ }
toPartialOrder := inferInstance
le_sup_left _ _ _ := Set.subset_union_left _ _
le_sup_right _ _ _ := Set.subset_union_right _ _
sup_le _ _ _ hx hy X := Set.union_subset_iff.mpr ⟨hx X, hy X⟩
@[simp]
lemma sup_covering (x y : Coverage C) (B : C) :
(x ⊔ y).covering B = x.covering B ∪ y.covering B :=
rfl
end Coverage
open Coverage
namespace Presieve
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
·
|
intro x t₁ t₂ h₁ h₂
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
·
|
Mathlib.CategoryTheory.Sites.Coverage.313_0.qkZFgqEgDC2P633
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R)
|
Mathlib_CategoryTheory_Sites_Coverage
|
case mpr.transitive.refine_1
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
P : Cᵒᵖ ⥤ Type w
X✝ : C
S✝ : Sieve X✝
X : C
R S : Sieve X
a✝¹ : saturate K X R
a✝ : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R.arrows f → saturate K Y (Sieve.pullback f S)
Y : C
f : Y ⟶ X
H :
∀ {X : C},
∀ R ∈ covering K X,
IsSeparatedFor P R ∧
∀ (x : FamilyOfElements P R), FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H1 : ∀ ⦃Y : C⦄ (f : Y ⟶ X), IsSeparatedFor P (Sieve.pullback f R).arrows
H1' :
∀ ⦃Y : C⦄ (f : Y ⟶ X) (x : FamilyOfElements P (Sieve.pullback f R).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H2 :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f → ∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y), IsSeparatedFor P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows
H2' :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f →
∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y) (x : FamilyOfElements P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
x : FamilyOfElements P (Sieve.pullback f S).arrows
t₁ t₂ : P.obj (Opposite.op Y)
h₁ : FamilyOfElements.IsAmalgamation x t₁
h₂ : FamilyOfElements.IsAmalgamation x t₂
⊢ t₁ = t₂
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
· obtain ⟨T, hT⟩ := y.pullback f S hy
exact ⟨T, Or.inr hT.1, hT.2⟩ }
toPartialOrder := inferInstance
le_sup_left _ _ _ := Set.subset_union_left _ _
le_sup_right _ _ _ := Set.subset_union_right _ _
sup_le _ _ _ hx hy X := Set.union_subset_iff.mpr ⟨hx X, hy X⟩
@[simp]
lemma sup_covering (x y : Coverage C) (B : C) :
(x ⊔ y).covering B = x.covering B ∪ y.covering B :=
rfl
end Coverage
open Coverage
namespace Presieve
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
|
refine (H1 f).ext (fun Z g hg => ?_)
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
|
Mathlib.CategoryTheory.Sites.Coverage.313_0.qkZFgqEgDC2P633
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R)
|
Mathlib_CategoryTheory_Sites_Coverage
|
case mpr.transitive.refine_1
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
P : Cᵒᵖ ⥤ Type w
X✝ : C
S✝ : Sieve X✝
X : C
R S : Sieve X
a✝¹ : saturate K X R
a✝ : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R.arrows f → saturate K Y (Sieve.pullback f S)
Y : C
f : Y ⟶ X
H :
∀ {X : C},
∀ R ∈ covering K X,
IsSeparatedFor P R ∧
∀ (x : FamilyOfElements P R), FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H1 : ∀ ⦃Y : C⦄ (f : Y ⟶ X), IsSeparatedFor P (Sieve.pullback f R).arrows
H1' :
∀ ⦃Y : C⦄ (f : Y ⟶ X) (x : FamilyOfElements P (Sieve.pullback f R).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H2 :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f → ∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y), IsSeparatedFor P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows
H2' :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f →
∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y) (x : FamilyOfElements P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
x : FamilyOfElements P (Sieve.pullback f S).arrows
t₁ t₂ : P.obj (Opposite.op Y)
h₁ : FamilyOfElements.IsAmalgamation x t₁
h₂ : FamilyOfElements.IsAmalgamation x t₂
Z : C
g : Z ⟶ Y
hg : (Sieve.pullback f R).arrows g
⊢ P.map g.op t₁ = P.map g.op t₂
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
· obtain ⟨T, hT⟩ := y.pullback f S hy
exact ⟨T, Or.inr hT.1, hT.2⟩ }
toPartialOrder := inferInstance
le_sup_left _ _ _ := Set.subset_union_left _ _
le_sup_right _ _ _ := Set.subset_union_right _ _
sup_le _ _ _ hx hy X := Set.union_subset_iff.mpr ⟨hx X, hy X⟩
@[simp]
lemma sup_covering (x y : Coverage C) (B : C) :
(x ⊔ y).covering B = x.covering B ∪ y.covering B :=
rfl
end Coverage
open Coverage
namespace Presieve
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
|
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
|
Mathlib.CategoryTheory.Sites.Coverage.313_0.qkZFgqEgDC2P633
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R)
|
Mathlib_CategoryTheory_Sites_Coverage
|
case mpr.transitive.refine_1
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
P : Cᵒᵖ ⥤ Type w
X✝ : C
S✝ : Sieve X✝
X : C
R S : Sieve X
a✝¹ : saturate K X R
a✝ : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R.arrows f → saturate K Y (Sieve.pullback f S)
Y : C
f : Y ⟶ X
H :
∀ {X : C},
∀ R ∈ covering K X,
IsSeparatedFor P R ∧
∀ (x : FamilyOfElements P R), FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H1 : ∀ ⦃Y : C⦄ (f : Y ⟶ X), IsSeparatedFor P (Sieve.pullback f R).arrows
H1' :
∀ ⦃Y : C⦄ (f : Y ⟶ X) (x : FamilyOfElements P (Sieve.pullback f R).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H2 :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f → ∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y), IsSeparatedFor P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows
H2' :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f →
∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y) (x : FamilyOfElements P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
x : FamilyOfElements P (Sieve.pullback f S).arrows
t₁ t₂ : P.obj (Opposite.op Y)
h₁ : FamilyOfElements.IsAmalgamation x t₁
h₂ : FamilyOfElements.IsAmalgamation x t₂
Z : C
g : Z ⟶ Y
hg : (Sieve.pullback f R).arrows g
ZZ : C
gg : ZZ ⟶ Z
hgg : (Sieve.pullback (𝟙 Z) (Sieve.pullback (g ≫ f) S)).arrows gg
⊢ P.map gg.op (P.map g.op t₁) = P.map gg.op (P.map g.op t₂)
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
· obtain ⟨T, hT⟩ := y.pullback f S hy
exact ⟨T, Or.inr hT.1, hT.2⟩ }
toPartialOrder := inferInstance
le_sup_left _ _ _ := Set.subset_union_left _ _
le_sup_right _ _ _ := Set.subset_union_right _ _
sup_le _ _ _ hx hy X := Set.union_subset_iff.mpr ⟨hx X, hy X⟩
@[simp]
lemma sup_covering (x y : Coverage C) (B : C) :
(x ⊔ y).covering B = x.covering B ∪ y.covering B :=
rfl
end Coverage
open Coverage
namespace Presieve
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
|
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
|
Mathlib.CategoryTheory.Sites.Coverage.313_0.qkZFgqEgDC2P633
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R)
|
Mathlib_CategoryTheory_Sites_Coverage
|
case mpr.transitive.refine_1
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
P : Cᵒᵖ ⥤ Type w
X✝ : C
S✝ : Sieve X✝
X : C
R S : Sieve X
a✝¹ : saturate K X R
a✝ : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R.arrows f → saturate K Y (Sieve.pullback f S)
Y : C
f : Y ⟶ X
H :
∀ {X : C},
∀ R ∈ covering K X,
IsSeparatedFor P R ∧
∀ (x : FamilyOfElements P R), FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H1 : ∀ ⦃Y : C⦄ (f : Y ⟶ X), IsSeparatedFor P (Sieve.pullback f R).arrows
H1' :
∀ ⦃Y : C⦄ (f : Y ⟶ X) (x : FamilyOfElements P (Sieve.pullback f R).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H2 :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f → ∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y), IsSeparatedFor P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows
H2' :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f →
∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y) (x : FamilyOfElements P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
x : FamilyOfElements P (Sieve.pullback f S).arrows
t₁ t₂ : P.obj (Opposite.op Y)
h₁ : FamilyOfElements.IsAmalgamation x t₁
h₂ : FamilyOfElements.IsAmalgamation x t₂
Z : C
g : Z ⟶ Y
hg : (Sieve.pullback f R).arrows g
ZZ : C
gg : ZZ ⟶ Z
hgg : S.arrows (gg ≫ g ≫ f)
⊢ P.map gg.op (P.map g.op t₁) = P.map gg.op (P.map g.op t₂)
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
· obtain ⟨T, hT⟩ := y.pullback f S hy
exact ⟨T, Or.inr hT.1, hT.2⟩ }
toPartialOrder := inferInstance
le_sup_left _ _ _ := Set.subset_union_left _ _
le_sup_right _ _ _ := Set.subset_union_right _ _
sup_le _ _ _ hx hy X := Set.union_subset_iff.mpr ⟨hx X, hy X⟩
@[simp]
lemma sup_covering (x y : Coverage C) (B : C) :
(x ⊔ y).covering B = x.covering B ∪ y.covering B :=
rfl
end Coverage
open Coverage
namespace Presieve
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
|
simp only [← types_comp_apply]
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
|
Mathlib.CategoryTheory.Sites.Coverage.313_0.qkZFgqEgDC2P633
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R)
|
Mathlib_CategoryTheory_Sites_Coverage
|
case mpr.transitive.refine_1
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
P : Cᵒᵖ ⥤ Type w
X✝ : C
S✝ : Sieve X✝
X : C
R S : Sieve X
a✝¹ : saturate K X R
a✝ : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R.arrows f → saturate K Y (Sieve.pullback f S)
Y : C
f : Y ⟶ X
H :
∀ {X : C},
∀ R ∈ covering K X,
IsSeparatedFor P R ∧
∀ (x : FamilyOfElements P R), FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H1 : ∀ ⦃Y : C⦄ (f : Y ⟶ X), IsSeparatedFor P (Sieve.pullback f R).arrows
H1' :
∀ ⦃Y : C⦄ (f : Y ⟶ X) (x : FamilyOfElements P (Sieve.pullback f R).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H2 :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f → ∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y), IsSeparatedFor P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows
H2' :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f →
∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y) (x : FamilyOfElements P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
x : FamilyOfElements P (Sieve.pullback f S).arrows
t₁ t₂ : P.obj (Opposite.op Y)
h₁ : FamilyOfElements.IsAmalgamation x t₁
h₂ : FamilyOfElements.IsAmalgamation x t₂
Z : C
g : Z ⟶ Y
hg : (Sieve.pullback f R).arrows g
ZZ : C
gg : ZZ ⟶ Z
hgg : S.arrows (gg ≫ g ≫ f)
⊢ (P.map g.op ≫ P.map gg.op) t₁ = (P.map g.op ≫ P.map gg.op) t₂
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
· obtain ⟨T, hT⟩ := y.pullback f S hy
exact ⟨T, Or.inr hT.1, hT.2⟩ }
toPartialOrder := inferInstance
le_sup_left _ _ _ := Set.subset_union_left _ _
le_sup_right _ _ _ := Set.subset_union_right _ _
sup_le _ _ _ hx hy X := Set.union_subset_iff.mpr ⟨hx X, hy X⟩
@[simp]
lemma sup_covering (x y : Coverage C) (B : C) :
(x ⊔ y).covering B = x.covering B ∪ y.covering B :=
rfl
end Coverage
open Coverage
namespace Presieve
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
simp only [← types_comp_apply]
|
rw [← P.map_comp, ← op_comp, h₁, h₂]
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
simp only [← types_comp_apply]
|
Mathlib.CategoryTheory.Sites.Coverage.313_0.qkZFgqEgDC2P633
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R)
|
Mathlib_CategoryTheory_Sites_Coverage
|
case mpr.transitive.refine_1.h
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
P : Cᵒᵖ ⥤ Type w
X✝ : C
S✝ : Sieve X✝
X : C
R S : Sieve X
a✝¹ : saturate K X R
a✝ : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R.arrows f → saturate K Y (Sieve.pullback f S)
Y : C
f : Y ⟶ X
H :
∀ {X : C},
∀ R ∈ covering K X,
IsSeparatedFor P R ∧
∀ (x : FamilyOfElements P R), FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H1 : ∀ ⦃Y : C⦄ (f : Y ⟶ X), IsSeparatedFor P (Sieve.pullback f R).arrows
H1' :
∀ ⦃Y : C⦄ (f : Y ⟶ X) (x : FamilyOfElements P (Sieve.pullback f R).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H2 :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f → ∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y), IsSeparatedFor P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows
H2' :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f →
∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y) (x : FamilyOfElements P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
x : FamilyOfElements P (Sieve.pullback f S).arrows
t₁ t₂ : P.obj (Opposite.op Y)
h₁ : FamilyOfElements.IsAmalgamation x t₁
h₂ : FamilyOfElements.IsAmalgamation x t₂
Z : C
g : Z ⟶ Y
hg : (Sieve.pullback f R).arrows g
ZZ : C
gg : ZZ ⟶ Z
hgg : S.arrows (gg ≫ g ≫ f)
⊢ (Sieve.pullback f S).arrows (gg ≫ g)
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
· obtain ⟨T, hT⟩ := y.pullback f S hy
exact ⟨T, Or.inr hT.1, hT.2⟩ }
toPartialOrder := inferInstance
le_sup_left _ _ _ := Set.subset_union_left _ _
le_sup_right _ _ _ := Set.subset_union_right _ _
sup_le _ _ _ hx hy X := Set.union_subset_iff.mpr ⟨hx X, hy X⟩
@[simp]
lemma sup_covering (x y : Coverage C) (B : C) :
(x ⊔ y).covering B = x.covering B ∪ y.covering B :=
rfl
end Coverage
open Coverage
namespace Presieve
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← op_comp, h₁, h₂]
|
simpa only [Sieve.pullback_apply, Category.assoc] using hgg
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← op_comp, h₁, h₂]
|
Mathlib.CategoryTheory.Sites.Coverage.313_0.qkZFgqEgDC2P633
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R)
|
Mathlib_CategoryTheory_Sites_Coverage
|
case mpr.transitive.refine_2
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
P : Cᵒᵖ ⥤ Type w
X✝ : C
S✝ : Sieve X✝
X : C
R S : Sieve X
a✝¹ : saturate K X R
a✝ : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R.arrows f → saturate K Y (Sieve.pullback f S)
Y : C
f : Y ⟶ X
H :
∀ {X : C},
∀ R ∈ covering K X,
IsSeparatedFor P R ∧
∀ (x : FamilyOfElements P R), FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H1 : ∀ ⦃Y : C⦄ (f : Y ⟶ X), IsSeparatedFor P (Sieve.pullback f R).arrows
H1' :
∀ ⦃Y : C⦄ (f : Y ⟶ X) (x : FamilyOfElements P (Sieve.pullback f R).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H2 :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f → ∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y), IsSeparatedFor P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows
H2' :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f →
∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y) (x : FamilyOfElements P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
x : FamilyOfElements P (Sieve.pullback f S).arrows
hx : FamilyOfElements.Compatible x
⊢ ∃ t, FamilyOfElements.IsAmalgamation x t
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
· obtain ⟨T, hT⟩ := y.pullback f S hy
exact ⟨T, Or.inr hT.1, hT.2⟩ }
toPartialOrder := inferInstance
le_sup_left _ _ _ := Set.subset_union_left _ _
le_sup_right _ _ _ := Set.subset_union_right _ _
sup_le _ _ _ hx hy X := Set.union_subset_iff.mpr ⟨hx X, hy X⟩
@[simp]
lemma sup_covering (x y : Coverage C) (B : C) :
(x ⊔ y).covering B = x.covering B ∪ y.covering B :=
rfl
end Coverage
open Coverage
namespace Presieve
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← op_comp, h₁, h₂]
simpa only [Sieve.pullback_apply, Category.assoc] using hgg
|
let y : ∀ ⦃Z : C⦄ (g : Z ⟶ Y),
((S.pullback (g ≫ f)).pullback (𝟙 _)).arrows.FamilyOfElements P :=
fun Z g ZZ gg hgg => x (gg ≫ g) (by simpa using hgg)
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← op_comp, h₁, h₂]
simpa only [Sieve.pullback_apply, Category.assoc] using hgg
|
Mathlib.CategoryTheory.Sites.Coverage.313_0.qkZFgqEgDC2P633
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R)
|
Mathlib_CategoryTheory_Sites_Coverage
|
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
P : Cᵒᵖ ⥤ Type w
X✝ : C
S✝ : Sieve X✝
X : C
R S : Sieve X
a✝¹ : saturate K X R
a✝ : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R.arrows f → saturate K Y (Sieve.pullback f S)
Y : C
f : Y ⟶ X
H :
∀ {X : C},
∀ R ∈ covering K X,
IsSeparatedFor P R ∧
∀ (x : FamilyOfElements P R), FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H1 : ∀ ⦃Y : C⦄ (f : Y ⟶ X), IsSeparatedFor P (Sieve.pullback f R).arrows
H1' :
∀ ⦃Y : C⦄ (f : Y ⟶ X) (x : FamilyOfElements P (Sieve.pullback f R).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H2 :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f → ∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y), IsSeparatedFor P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows
H2' :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f →
∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y) (x : FamilyOfElements P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
x : FamilyOfElements P (Sieve.pullback f S).arrows
hx : FamilyOfElements.Compatible x
Z : C
g : Z ⟶ Y
ZZ : C
gg : ZZ ⟶ Z
hgg : (Sieve.pullback (𝟙 Z) (Sieve.pullback (g ≫ f) S)).arrows gg
⊢ (Sieve.pullback f S).arrows (gg ≫ g)
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
· obtain ⟨T, hT⟩ := y.pullback f S hy
exact ⟨T, Or.inr hT.1, hT.2⟩ }
toPartialOrder := inferInstance
le_sup_left _ _ _ := Set.subset_union_left _ _
le_sup_right _ _ _ := Set.subset_union_right _ _
sup_le _ _ _ hx hy X := Set.union_subset_iff.mpr ⟨hx X, hy X⟩
@[simp]
lemma sup_covering (x y : Coverage C) (B : C) :
(x ⊔ y).covering B = x.covering B ∪ y.covering B :=
rfl
end Coverage
open Coverage
namespace Presieve
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← op_comp, h₁, h₂]
simpa only [Sieve.pullback_apply, Category.assoc] using hgg
let y : ∀ ⦃Z : C⦄ (g : Z ⟶ Y),
((S.pullback (g ≫ f)).pullback (𝟙 _)).arrows.FamilyOfElements P :=
fun Z g ZZ gg hgg => x (gg ≫ g) (by
|
simpa using hgg
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← op_comp, h₁, h₂]
simpa only [Sieve.pullback_apply, Category.assoc] using hgg
let y : ∀ ⦃Z : C⦄ (g : Z ⟶ Y),
((S.pullback (g ≫ f)).pullback (𝟙 _)).arrows.FamilyOfElements P :=
fun Z g ZZ gg hgg => x (gg ≫ g) (by
|
Mathlib.CategoryTheory.Sites.Coverage.313_0.qkZFgqEgDC2P633
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R)
|
Mathlib_CategoryTheory_Sites_Coverage
|
case mpr.transitive.refine_2
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
P : Cᵒᵖ ⥤ Type w
X✝ : C
S✝ : Sieve X✝
X : C
R S : Sieve X
a✝¹ : saturate K X R
a✝ : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R.arrows f → saturate K Y (Sieve.pullback f S)
Y : C
f : Y ⟶ X
H :
∀ {X : C},
∀ R ∈ covering K X,
IsSeparatedFor P R ∧
∀ (x : FamilyOfElements P R), FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H1 : ∀ ⦃Y : C⦄ (f : Y ⟶ X), IsSeparatedFor P (Sieve.pullback f R).arrows
H1' :
∀ ⦃Y : C⦄ (f : Y ⟶ X) (x : FamilyOfElements P (Sieve.pullback f R).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H2 :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f → ∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y), IsSeparatedFor P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows
H2' :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f →
∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y) (x : FamilyOfElements P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
x : FamilyOfElements P (Sieve.pullback f S).arrows
hx : FamilyOfElements.Compatible x
y : ⦃Z : C⦄ → (g : Z ⟶ Y) → FamilyOfElements P (Sieve.pullback (𝟙 Z) (Sieve.pullback (g ≫ f) S)).arrows :=
fun Z g ZZ gg hgg => x (gg ≫ g) (_ : (Sieve.pullback f S).arrows (gg ≫ g))
⊢ ∃ t, FamilyOfElements.IsAmalgamation x t
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
· obtain ⟨T, hT⟩ := y.pullback f S hy
exact ⟨T, Or.inr hT.1, hT.2⟩ }
toPartialOrder := inferInstance
le_sup_left _ _ _ := Set.subset_union_left _ _
le_sup_right _ _ _ := Set.subset_union_right _ _
sup_le _ _ _ hx hy X := Set.union_subset_iff.mpr ⟨hx X, hy X⟩
@[simp]
lemma sup_covering (x y : Coverage C) (B : C) :
(x ⊔ y).covering B = x.covering B ∪ y.covering B :=
rfl
end Coverage
open Coverage
namespace Presieve
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← op_comp, h₁, h₂]
simpa only [Sieve.pullback_apply, Category.assoc] using hgg
let y : ∀ ⦃Z : C⦄ (g : Z ⟶ Y),
((S.pullback (g ≫ f)).pullback (𝟙 _)).arrows.FamilyOfElements P :=
fun Z g ZZ gg hgg => x (gg ≫ g) (by simpa using hgg)
|
have hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), (y g).Compatible := by
intro Z g Y₁ Y₂ ZZ g₁ g₂ f₁ f₂ h₁ h₂ h
rw [hx]
rw [reassoc_of% h]
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← op_comp, h₁, h₂]
simpa only [Sieve.pullback_apply, Category.assoc] using hgg
let y : ∀ ⦃Z : C⦄ (g : Z ⟶ Y),
((S.pullback (g ≫ f)).pullback (𝟙 _)).arrows.FamilyOfElements P :=
fun Z g ZZ gg hgg => x (gg ≫ g) (by simpa using hgg)
|
Mathlib.CategoryTheory.Sites.Coverage.313_0.qkZFgqEgDC2P633
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R)
|
Mathlib_CategoryTheory_Sites_Coverage
|
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
P : Cᵒᵖ ⥤ Type w
X✝ : C
S✝ : Sieve X✝
X : C
R S : Sieve X
a✝¹ : saturate K X R
a✝ : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R.arrows f → saturate K Y (Sieve.pullback f S)
Y : C
f : Y ⟶ X
H :
∀ {X : C},
∀ R ∈ covering K X,
IsSeparatedFor P R ∧
∀ (x : FamilyOfElements P R), FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H1 : ∀ ⦃Y : C⦄ (f : Y ⟶ X), IsSeparatedFor P (Sieve.pullback f R).arrows
H1' :
∀ ⦃Y : C⦄ (f : Y ⟶ X) (x : FamilyOfElements P (Sieve.pullback f R).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H2 :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f → ∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y), IsSeparatedFor P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows
H2' :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f →
∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y) (x : FamilyOfElements P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
x : FamilyOfElements P (Sieve.pullback f S).arrows
hx : FamilyOfElements.Compatible x
y : ⦃Z : C⦄ → (g : Z ⟶ Y) → FamilyOfElements P (Sieve.pullback (𝟙 Z) (Sieve.pullback (g ≫ f) S)).arrows :=
fun Z g ZZ gg hgg => x (gg ≫ g) (_ : (Sieve.pullback f S).arrows (gg ≫ g))
⊢ ∀ ⦃Z : C⦄ (g : Z ⟶ Y), FamilyOfElements.Compatible (y g)
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
· obtain ⟨T, hT⟩ := y.pullback f S hy
exact ⟨T, Or.inr hT.1, hT.2⟩ }
toPartialOrder := inferInstance
le_sup_left _ _ _ := Set.subset_union_left _ _
le_sup_right _ _ _ := Set.subset_union_right _ _
sup_le _ _ _ hx hy X := Set.union_subset_iff.mpr ⟨hx X, hy X⟩
@[simp]
lemma sup_covering (x y : Coverage C) (B : C) :
(x ⊔ y).covering B = x.covering B ∪ y.covering B :=
rfl
end Coverage
open Coverage
namespace Presieve
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← op_comp, h₁, h₂]
simpa only [Sieve.pullback_apply, Category.assoc] using hgg
let y : ∀ ⦃Z : C⦄ (g : Z ⟶ Y),
((S.pullback (g ≫ f)).pullback (𝟙 _)).arrows.FamilyOfElements P :=
fun Z g ZZ gg hgg => x (gg ≫ g) (by simpa using hgg)
have hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), (y g).Compatible := by
|
intro Z g Y₁ Y₂ ZZ g₁ g₂ f₁ f₂ h₁ h₂ h
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← op_comp, h₁, h₂]
simpa only [Sieve.pullback_apply, Category.assoc] using hgg
let y : ∀ ⦃Z : C⦄ (g : Z ⟶ Y),
((S.pullback (g ≫ f)).pullback (𝟙 _)).arrows.FamilyOfElements P :=
fun Z g ZZ gg hgg => x (gg ≫ g) (by simpa using hgg)
have hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), (y g).Compatible := by
|
Mathlib.CategoryTheory.Sites.Coverage.313_0.qkZFgqEgDC2P633
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R)
|
Mathlib_CategoryTheory_Sites_Coverage
|
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
P : Cᵒᵖ ⥤ Type w
X✝ : C
S✝ : Sieve X✝
X : C
R S : Sieve X
a✝¹ : saturate K X R
a✝ : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R.arrows f → saturate K Y (Sieve.pullback f S)
Y : C
f : Y ⟶ X
H :
∀ {X : C},
∀ R ∈ covering K X,
IsSeparatedFor P R ∧
∀ (x : FamilyOfElements P R), FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H1 : ∀ ⦃Y : C⦄ (f : Y ⟶ X), IsSeparatedFor P (Sieve.pullback f R).arrows
H1' :
∀ ⦃Y : C⦄ (f : Y ⟶ X) (x : FamilyOfElements P (Sieve.pullback f R).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H2 :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f → ∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y), IsSeparatedFor P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows
H2' :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f →
∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y) (x : FamilyOfElements P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
x : FamilyOfElements P (Sieve.pullback f S).arrows
hx : FamilyOfElements.Compatible x
y : ⦃Z : C⦄ → (g : Z ⟶ Y) → FamilyOfElements P (Sieve.pullback (𝟙 Z) (Sieve.pullback (g ≫ f) S)).arrows :=
fun Z g ZZ gg hgg => x (gg ≫ g) (_ : (Sieve.pullback f S).arrows (gg ≫ g))
Z : C
g : Z ⟶ Y
Y₁ Y₂ ZZ : C
g₁ : ZZ ⟶ Y₁
g₂ : ZZ ⟶ Y₂
f₁ : Y₁ ⟶ Z
f₂ : Y₂ ⟶ Z
h₁ : (Sieve.pullback (𝟙 Z) (Sieve.pullback (g ≫ f) S)).arrows f₁
h₂ : (Sieve.pullback (𝟙 Z) (Sieve.pullback (g ≫ f) S)).arrows f₂
h : g₁ ≫ f₁ = g₂ ≫ f₂
⊢ P.map g₁.op (y g f₁ h₁) = P.map g₂.op (y g f₂ h₂)
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
· obtain ⟨T, hT⟩ := y.pullback f S hy
exact ⟨T, Or.inr hT.1, hT.2⟩ }
toPartialOrder := inferInstance
le_sup_left _ _ _ := Set.subset_union_left _ _
le_sup_right _ _ _ := Set.subset_union_right _ _
sup_le _ _ _ hx hy X := Set.union_subset_iff.mpr ⟨hx X, hy X⟩
@[simp]
lemma sup_covering (x y : Coverage C) (B : C) :
(x ⊔ y).covering B = x.covering B ∪ y.covering B :=
rfl
end Coverage
open Coverage
namespace Presieve
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← op_comp, h₁, h₂]
simpa only [Sieve.pullback_apply, Category.assoc] using hgg
let y : ∀ ⦃Z : C⦄ (g : Z ⟶ Y),
((S.pullback (g ≫ f)).pullback (𝟙 _)).arrows.FamilyOfElements P :=
fun Z g ZZ gg hgg => x (gg ≫ g) (by simpa using hgg)
have hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), (y g).Compatible := by
intro Z g Y₁ Y₂ ZZ g₁ g₂ f₁ f₂ h₁ h₂ h
|
rw [hx]
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← op_comp, h₁, h₂]
simpa only [Sieve.pullback_apply, Category.assoc] using hgg
let y : ∀ ⦃Z : C⦄ (g : Z ⟶ Y),
((S.pullback (g ≫ f)).pullback (𝟙 _)).arrows.FamilyOfElements P :=
fun Z g ZZ gg hgg => x (gg ≫ g) (by simpa using hgg)
have hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), (y g).Compatible := by
intro Z g Y₁ Y₂ ZZ g₁ g₂ f₁ f₂ h₁ h₂ h
|
Mathlib.CategoryTheory.Sites.Coverage.313_0.qkZFgqEgDC2P633
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R)
|
Mathlib_CategoryTheory_Sites_Coverage
|
case a
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
P : Cᵒᵖ ⥤ Type w
X✝ : C
S✝ : Sieve X✝
X : C
R S : Sieve X
a✝¹ : saturate K X R
a✝ : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R.arrows f → saturate K Y (Sieve.pullback f S)
Y : C
f : Y ⟶ X
H :
∀ {X : C},
∀ R ∈ covering K X,
IsSeparatedFor P R ∧
∀ (x : FamilyOfElements P R), FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H1 : ∀ ⦃Y : C⦄ (f : Y ⟶ X), IsSeparatedFor P (Sieve.pullback f R).arrows
H1' :
∀ ⦃Y : C⦄ (f : Y ⟶ X) (x : FamilyOfElements P (Sieve.pullback f R).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H2 :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f → ∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y), IsSeparatedFor P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows
H2' :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f →
∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y) (x : FamilyOfElements P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
x : FamilyOfElements P (Sieve.pullback f S).arrows
hx : FamilyOfElements.Compatible x
y : ⦃Z : C⦄ → (g : Z ⟶ Y) → FamilyOfElements P (Sieve.pullback (𝟙 Z) (Sieve.pullback (g ≫ f) S)).arrows :=
fun Z g ZZ gg hgg => x (gg ≫ g) (_ : (Sieve.pullback f S).arrows (gg ≫ g))
Z : C
g : Z ⟶ Y
Y₁ Y₂ ZZ : C
g₁ : ZZ ⟶ Y₁
g₂ : ZZ ⟶ Y₂
f₁ : Y₁ ⟶ Z
f₂ : Y₂ ⟶ Z
h₁ : (Sieve.pullback (𝟙 Z) (Sieve.pullback (g ≫ f) S)).arrows f₁
h₂ : (Sieve.pullback (𝟙 Z) (Sieve.pullback (g ≫ f) S)).arrows f₂
h : g₁ ≫ f₁ = g₂ ≫ f₂
⊢ g₁ ≫ f₁ ≫ g = g₂ ≫ f₂ ≫ g
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
· obtain ⟨T, hT⟩ := y.pullback f S hy
exact ⟨T, Or.inr hT.1, hT.2⟩ }
toPartialOrder := inferInstance
le_sup_left _ _ _ := Set.subset_union_left _ _
le_sup_right _ _ _ := Set.subset_union_right _ _
sup_le _ _ _ hx hy X := Set.union_subset_iff.mpr ⟨hx X, hy X⟩
@[simp]
lemma sup_covering (x y : Coverage C) (B : C) :
(x ⊔ y).covering B = x.covering B ∪ y.covering B :=
rfl
end Coverage
open Coverage
namespace Presieve
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← op_comp, h₁, h₂]
simpa only [Sieve.pullback_apply, Category.assoc] using hgg
let y : ∀ ⦃Z : C⦄ (g : Z ⟶ Y),
((S.pullback (g ≫ f)).pullback (𝟙 _)).arrows.FamilyOfElements P :=
fun Z g ZZ gg hgg => x (gg ≫ g) (by simpa using hgg)
have hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), (y g).Compatible := by
intro Z g Y₁ Y₂ ZZ g₁ g₂ f₁ f₂ h₁ h₂ h
rw [hx]
|
rw [reassoc_of% h]
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← op_comp, h₁, h₂]
simpa only [Sieve.pullback_apply, Category.assoc] using hgg
let y : ∀ ⦃Z : C⦄ (g : Z ⟶ Y),
((S.pullback (g ≫ f)).pullback (𝟙 _)).arrows.FamilyOfElements P :=
fun Z g ZZ gg hgg => x (gg ≫ g) (by simpa using hgg)
have hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), (y g).Compatible := by
intro Z g Y₁ Y₂ ZZ g₁ g₂ f₁ f₂ h₁ h₂ h
rw [hx]
|
Mathlib.CategoryTheory.Sites.Coverage.313_0.qkZFgqEgDC2P633
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R)
|
Mathlib_CategoryTheory_Sites_Coverage
|
case mpr.transitive.refine_2
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
P : Cᵒᵖ ⥤ Type w
X✝ : C
S✝ : Sieve X✝
X : C
R S : Sieve X
a✝¹ : saturate K X R
a✝ : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R.arrows f → saturate K Y (Sieve.pullback f S)
Y : C
f : Y ⟶ X
H :
∀ {X : C},
∀ R ∈ covering K X,
IsSeparatedFor P R ∧
∀ (x : FamilyOfElements P R), FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H1 : ∀ ⦃Y : C⦄ (f : Y ⟶ X), IsSeparatedFor P (Sieve.pullback f R).arrows
H1' :
∀ ⦃Y : C⦄ (f : Y ⟶ X) (x : FamilyOfElements P (Sieve.pullback f R).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H2 :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f → ∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y), IsSeparatedFor P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows
H2' :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f →
∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y) (x : FamilyOfElements P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
x : FamilyOfElements P (Sieve.pullback f S).arrows
hx : FamilyOfElements.Compatible x
y : ⦃Z : C⦄ → (g : Z ⟶ Y) → FamilyOfElements P (Sieve.pullback (𝟙 Z) (Sieve.pullback (g ≫ f) S)).arrows :=
fun Z g ZZ gg hgg => x (gg ≫ g) (_ : (Sieve.pullback f S).arrows (gg ≫ g))
hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), FamilyOfElements.Compatible (y g)
⊢ ∃ t, FamilyOfElements.IsAmalgamation x t
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
· obtain ⟨T, hT⟩ := y.pullback f S hy
exact ⟨T, Or.inr hT.1, hT.2⟩ }
toPartialOrder := inferInstance
le_sup_left _ _ _ := Set.subset_union_left _ _
le_sup_right _ _ _ := Set.subset_union_right _ _
sup_le _ _ _ hx hy X := Set.union_subset_iff.mpr ⟨hx X, hy X⟩
@[simp]
lemma sup_covering (x y : Coverage C) (B : C) :
(x ⊔ y).covering B = x.covering B ∪ y.covering B :=
rfl
end Coverage
open Coverage
namespace Presieve
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← op_comp, h₁, h₂]
simpa only [Sieve.pullback_apply, Category.assoc] using hgg
let y : ∀ ⦃Z : C⦄ (g : Z ⟶ Y),
((S.pullback (g ≫ f)).pullback (𝟙 _)).arrows.FamilyOfElements P :=
fun Z g ZZ gg hgg => x (gg ≫ g) (by simpa using hgg)
have hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), (y g).Compatible := by
intro Z g Y₁ Y₂ ZZ g₁ g₂ f₁ f₂ h₁ h₂ h
rw [hx]
rw [reassoc_of% h]
|
choose z hz using fun ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : R.pullback f g) =>
H2' hg (𝟙 _) (y g) (hy g)
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← op_comp, h₁, h₂]
simpa only [Sieve.pullback_apply, Category.assoc] using hgg
let y : ∀ ⦃Z : C⦄ (g : Z ⟶ Y),
((S.pullback (g ≫ f)).pullback (𝟙 _)).arrows.FamilyOfElements P :=
fun Z g ZZ gg hgg => x (gg ≫ g) (by simpa using hgg)
have hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), (y g).Compatible := by
intro Z g Y₁ Y₂ ZZ g₁ g₂ f₁ f₂ h₁ h₂ h
rw [hx]
rw [reassoc_of% h]
|
Mathlib.CategoryTheory.Sites.Coverage.313_0.qkZFgqEgDC2P633
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R)
|
Mathlib_CategoryTheory_Sites_Coverage
|
case mpr.transitive.refine_2
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
P : Cᵒᵖ ⥤ Type w
X✝ : C
S✝ : Sieve X✝
X : C
R S : Sieve X
a✝¹ : saturate K X R
a✝ : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R.arrows f → saturate K Y (Sieve.pullback f S)
Y : C
f : Y ⟶ X
H :
∀ {X : C},
∀ R ∈ covering K X,
IsSeparatedFor P R ∧
∀ (x : FamilyOfElements P R), FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H1 : ∀ ⦃Y : C⦄ (f : Y ⟶ X), IsSeparatedFor P (Sieve.pullback f R).arrows
H1' :
∀ ⦃Y : C⦄ (f : Y ⟶ X) (x : FamilyOfElements P (Sieve.pullback f R).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H2 :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f → ∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y), IsSeparatedFor P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows
H2' :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f →
∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y) (x : FamilyOfElements P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
x : FamilyOfElements P (Sieve.pullback f S).arrows
hx : FamilyOfElements.Compatible x
y : ⦃Z : C⦄ → (g : Z ⟶ Y) → FamilyOfElements P (Sieve.pullback (𝟙 Z) (Sieve.pullback (g ≫ f) S)).arrows :=
fun Z g ZZ gg hgg => x (gg ≫ g) (_ : (Sieve.pullback f S).arrows (gg ≫ g))
hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), FamilyOfElements.Compatible (y g)
z : ⦃Z : C⦄ → ⦃g : Z ⟶ Y⦄ → (Sieve.pullback f R).arrows g → P.obj (Opposite.op Z)
hz : ∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : (Sieve.pullback f R).arrows g), FamilyOfElements.IsAmalgamation (y g) (z hg)
⊢ ∃ t, FamilyOfElements.IsAmalgamation x t
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
· obtain ⟨T, hT⟩ := y.pullback f S hy
exact ⟨T, Or.inr hT.1, hT.2⟩ }
toPartialOrder := inferInstance
le_sup_left _ _ _ := Set.subset_union_left _ _
le_sup_right _ _ _ := Set.subset_union_right _ _
sup_le _ _ _ hx hy X := Set.union_subset_iff.mpr ⟨hx X, hy X⟩
@[simp]
lemma sup_covering (x y : Coverage C) (B : C) :
(x ⊔ y).covering B = x.covering B ∪ y.covering B :=
rfl
end Coverage
open Coverage
namespace Presieve
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← op_comp, h₁, h₂]
simpa only [Sieve.pullback_apply, Category.assoc] using hgg
let y : ∀ ⦃Z : C⦄ (g : Z ⟶ Y),
((S.pullback (g ≫ f)).pullback (𝟙 _)).arrows.FamilyOfElements P :=
fun Z g ZZ gg hgg => x (gg ≫ g) (by simpa using hgg)
have hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), (y g).Compatible := by
intro Z g Y₁ Y₂ ZZ g₁ g₂ f₁ f₂ h₁ h₂ h
rw [hx]
rw [reassoc_of% h]
choose z hz using fun ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : R.pullback f g) =>
H2' hg (𝟙 _) (y g) (hy g)
|
let q : (R.pullback f).arrows.FamilyOfElements P := fun Z g hg => z hg
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← op_comp, h₁, h₂]
simpa only [Sieve.pullback_apply, Category.assoc] using hgg
let y : ∀ ⦃Z : C⦄ (g : Z ⟶ Y),
((S.pullback (g ≫ f)).pullback (𝟙 _)).arrows.FamilyOfElements P :=
fun Z g ZZ gg hgg => x (gg ≫ g) (by simpa using hgg)
have hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), (y g).Compatible := by
intro Z g Y₁ Y₂ ZZ g₁ g₂ f₁ f₂ h₁ h₂ h
rw [hx]
rw [reassoc_of% h]
choose z hz using fun ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : R.pullback f g) =>
H2' hg (𝟙 _) (y g) (hy g)
|
Mathlib.CategoryTheory.Sites.Coverage.313_0.qkZFgqEgDC2P633
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R)
|
Mathlib_CategoryTheory_Sites_Coverage
|
case mpr.transitive.refine_2
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
P : Cᵒᵖ ⥤ Type w
X✝ : C
S✝ : Sieve X✝
X : C
R S : Sieve X
a✝¹ : saturate K X R
a✝ : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R.arrows f → saturate K Y (Sieve.pullback f S)
Y : C
f : Y ⟶ X
H :
∀ {X : C},
∀ R ∈ covering K X,
IsSeparatedFor P R ∧
∀ (x : FamilyOfElements P R), FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H1 : ∀ ⦃Y : C⦄ (f : Y ⟶ X), IsSeparatedFor P (Sieve.pullback f R).arrows
H1' :
∀ ⦃Y : C⦄ (f : Y ⟶ X) (x : FamilyOfElements P (Sieve.pullback f R).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H2 :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f → ∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y), IsSeparatedFor P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows
H2' :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f →
∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y) (x : FamilyOfElements P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
x : FamilyOfElements P (Sieve.pullback f S).arrows
hx : FamilyOfElements.Compatible x
y : ⦃Z : C⦄ → (g : Z ⟶ Y) → FamilyOfElements P (Sieve.pullback (𝟙 Z) (Sieve.pullback (g ≫ f) S)).arrows :=
fun Z g ZZ gg hgg => x (gg ≫ g) (_ : (Sieve.pullback f S).arrows (gg ≫ g))
hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), FamilyOfElements.Compatible (y g)
z : ⦃Z : C⦄ → ⦃g : Z ⟶ Y⦄ → (Sieve.pullback f R).arrows g → P.obj (Opposite.op Z)
hz : ∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : (Sieve.pullback f R).arrows g), FamilyOfElements.IsAmalgamation (y g) (z hg)
q : FamilyOfElements P (Sieve.pullback f R).arrows := fun Z g hg => z hg
⊢ ∃ t, FamilyOfElements.IsAmalgamation x t
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
· obtain ⟨T, hT⟩ := y.pullback f S hy
exact ⟨T, Or.inr hT.1, hT.2⟩ }
toPartialOrder := inferInstance
le_sup_left _ _ _ := Set.subset_union_left _ _
le_sup_right _ _ _ := Set.subset_union_right _ _
sup_le _ _ _ hx hy X := Set.union_subset_iff.mpr ⟨hx X, hy X⟩
@[simp]
lemma sup_covering (x y : Coverage C) (B : C) :
(x ⊔ y).covering B = x.covering B ∪ y.covering B :=
rfl
end Coverage
open Coverage
namespace Presieve
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← op_comp, h₁, h₂]
simpa only [Sieve.pullback_apply, Category.assoc] using hgg
let y : ∀ ⦃Z : C⦄ (g : Z ⟶ Y),
((S.pullback (g ≫ f)).pullback (𝟙 _)).arrows.FamilyOfElements P :=
fun Z g ZZ gg hgg => x (gg ≫ g) (by simpa using hgg)
have hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), (y g).Compatible := by
intro Z g Y₁ Y₂ ZZ g₁ g₂ f₁ f₂ h₁ h₂ h
rw [hx]
rw [reassoc_of% h]
choose z hz using fun ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : R.pullback f g) =>
H2' hg (𝟙 _) (y g) (hy g)
let q : (R.pullback f).arrows.FamilyOfElements P := fun Z g hg => z hg
|
have hq : q.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
apply (H2 h₁ g₁).ext
intro ZZ gg hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← P.map_comp, ← op_comp, ← op_comp, hz, hz]
· dsimp; congr 1; simp only [Category.assoc, h]
· simpa [reassoc_of% h] using hgg
· simpa using hgg
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← op_comp, h₁, h₂]
simpa only [Sieve.pullback_apply, Category.assoc] using hgg
let y : ∀ ⦃Z : C⦄ (g : Z ⟶ Y),
((S.pullback (g ≫ f)).pullback (𝟙 _)).arrows.FamilyOfElements P :=
fun Z g ZZ gg hgg => x (gg ≫ g) (by simpa using hgg)
have hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), (y g).Compatible := by
intro Z g Y₁ Y₂ ZZ g₁ g₂ f₁ f₂ h₁ h₂ h
rw [hx]
rw [reassoc_of% h]
choose z hz using fun ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : R.pullback f g) =>
H2' hg (𝟙 _) (y g) (hy g)
let q : (R.pullback f).arrows.FamilyOfElements P := fun Z g hg => z hg
|
Mathlib.CategoryTheory.Sites.Coverage.313_0.qkZFgqEgDC2P633
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R)
|
Mathlib_CategoryTheory_Sites_Coverage
|
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
P : Cᵒᵖ ⥤ Type w
X✝ : C
S✝ : Sieve X✝
X : C
R S : Sieve X
a✝¹ : saturate K X R
a✝ : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R.arrows f → saturate K Y (Sieve.pullback f S)
Y : C
f : Y ⟶ X
H :
∀ {X : C},
∀ R ∈ covering K X,
IsSeparatedFor P R ∧
∀ (x : FamilyOfElements P R), FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H1 : ∀ ⦃Y : C⦄ (f : Y ⟶ X), IsSeparatedFor P (Sieve.pullback f R).arrows
H1' :
∀ ⦃Y : C⦄ (f : Y ⟶ X) (x : FamilyOfElements P (Sieve.pullback f R).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H2 :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f → ∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y), IsSeparatedFor P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows
H2' :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f →
∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y) (x : FamilyOfElements P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
x : FamilyOfElements P (Sieve.pullback f S).arrows
hx : FamilyOfElements.Compatible x
y : ⦃Z : C⦄ → (g : Z ⟶ Y) → FamilyOfElements P (Sieve.pullback (𝟙 Z) (Sieve.pullback (g ≫ f) S)).arrows :=
fun Z g ZZ gg hgg => x (gg ≫ g) (_ : (Sieve.pullback f S).arrows (gg ≫ g))
hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), FamilyOfElements.Compatible (y g)
z : ⦃Z : C⦄ → ⦃g : Z ⟶ Y⦄ → (Sieve.pullback f R).arrows g → P.obj (Opposite.op Z)
hz : ∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : (Sieve.pullback f R).arrows g), FamilyOfElements.IsAmalgamation (y g) (z hg)
q : FamilyOfElements P (Sieve.pullback f R).arrows := fun Z g hg => z hg
⊢ FamilyOfElements.Compatible q
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
· obtain ⟨T, hT⟩ := y.pullback f S hy
exact ⟨T, Or.inr hT.1, hT.2⟩ }
toPartialOrder := inferInstance
le_sup_left _ _ _ := Set.subset_union_left _ _
le_sup_right _ _ _ := Set.subset_union_right _ _
sup_le _ _ _ hx hy X := Set.union_subset_iff.mpr ⟨hx X, hy X⟩
@[simp]
lemma sup_covering (x y : Coverage C) (B : C) :
(x ⊔ y).covering B = x.covering B ∪ y.covering B :=
rfl
end Coverage
open Coverage
namespace Presieve
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← op_comp, h₁, h₂]
simpa only [Sieve.pullback_apply, Category.assoc] using hgg
let y : ∀ ⦃Z : C⦄ (g : Z ⟶ Y),
((S.pullback (g ≫ f)).pullback (𝟙 _)).arrows.FamilyOfElements P :=
fun Z g ZZ gg hgg => x (gg ≫ g) (by simpa using hgg)
have hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), (y g).Compatible := by
intro Z g Y₁ Y₂ ZZ g₁ g₂ f₁ f₂ h₁ h₂ h
rw [hx]
rw [reassoc_of% h]
choose z hz using fun ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : R.pullback f g) =>
H2' hg (𝟙 _) (y g) (hy g)
let q : (R.pullback f).arrows.FamilyOfElements P := fun Z g hg => z hg
have hq : q.Compatible := by
|
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← op_comp, h₁, h₂]
simpa only [Sieve.pullback_apply, Category.assoc] using hgg
let y : ∀ ⦃Z : C⦄ (g : Z ⟶ Y),
((S.pullback (g ≫ f)).pullback (𝟙 _)).arrows.FamilyOfElements P :=
fun Z g ZZ gg hgg => x (gg ≫ g) (by simpa using hgg)
have hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), (y g).Compatible := by
intro Z g Y₁ Y₂ ZZ g₁ g₂ f₁ f₂ h₁ h₂ h
rw [hx]
rw [reassoc_of% h]
choose z hz using fun ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : R.pullback f g) =>
H2' hg (𝟙 _) (y g) (hy g)
let q : (R.pullback f).arrows.FamilyOfElements P := fun Z g hg => z hg
have hq : q.Compatible := by
|
Mathlib.CategoryTheory.Sites.Coverage.313_0.qkZFgqEgDC2P633
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R)
|
Mathlib_CategoryTheory_Sites_Coverage
|
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
P : Cᵒᵖ ⥤ Type w
X✝ : C
S✝ : Sieve X✝
X : C
R S : Sieve X
a✝¹ : saturate K X R
a✝ : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R.arrows f → saturate K Y (Sieve.pullback f S)
Y : C
f : Y ⟶ X
H :
∀ {X : C},
∀ R ∈ covering K X,
IsSeparatedFor P R ∧
∀ (x : FamilyOfElements P R), FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H1 : ∀ ⦃Y : C⦄ (f : Y ⟶ X), IsSeparatedFor P (Sieve.pullback f R).arrows
H1' :
∀ ⦃Y : C⦄ (f : Y ⟶ X) (x : FamilyOfElements P (Sieve.pullback f R).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H2 :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f → ∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y), IsSeparatedFor P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows
H2' :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f →
∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y) (x : FamilyOfElements P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
x : FamilyOfElements P (Sieve.pullback f S).arrows
hx : FamilyOfElements.Compatible x
y : ⦃Z : C⦄ → (g : Z ⟶ Y) → FamilyOfElements P (Sieve.pullback (𝟙 Z) (Sieve.pullback (g ≫ f) S)).arrows :=
fun Z g ZZ gg hgg => x (gg ≫ g) (_ : (Sieve.pullback f S).arrows (gg ≫ g))
hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), FamilyOfElements.Compatible (y g)
z : ⦃Z : C⦄ → ⦃g : Z ⟶ Y⦄ → (Sieve.pullback f R).arrows g → P.obj (Opposite.op Z)
hz : ∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : (Sieve.pullback f R).arrows g), FamilyOfElements.IsAmalgamation (y g) (z hg)
q : FamilyOfElements P (Sieve.pullback f R).arrows := fun Z g hg => z hg
Y₁ Y₂ Z : C
g₁ : Z ⟶ Y₁
g₂ : Z ⟶ Y₂
f₁ : Y₁ ⟶ Y
f₂ : Y₂ ⟶ Y
h₁ : (Sieve.pullback f R).arrows f₁
h₂ : (Sieve.pullback f R).arrows f₂
h : g₁ ≫ f₁ = g₂ ≫ f₂
⊢ P.map g₁.op (q f₁ h₁) = P.map g₂.op (q f₂ h₂)
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
· obtain ⟨T, hT⟩ := y.pullback f S hy
exact ⟨T, Or.inr hT.1, hT.2⟩ }
toPartialOrder := inferInstance
le_sup_left _ _ _ := Set.subset_union_left _ _
le_sup_right _ _ _ := Set.subset_union_right _ _
sup_le _ _ _ hx hy X := Set.union_subset_iff.mpr ⟨hx X, hy X⟩
@[simp]
lemma sup_covering (x y : Coverage C) (B : C) :
(x ⊔ y).covering B = x.covering B ∪ y.covering B :=
rfl
end Coverage
open Coverage
namespace Presieve
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← op_comp, h₁, h₂]
simpa only [Sieve.pullback_apply, Category.assoc] using hgg
let y : ∀ ⦃Z : C⦄ (g : Z ⟶ Y),
((S.pullback (g ≫ f)).pullback (𝟙 _)).arrows.FamilyOfElements P :=
fun Z g ZZ gg hgg => x (gg ≫ g) (by simpa using hgg)
have hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), (y g).Compatible := by
intro Z g Y₁ Y₂ ZZ g₁ g₂ f₁ f₂ h₁ h₂ h
rw [hx]
rw [reassoc_of% h]
choose z hz using fun ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : R.pullback f g) =>
H2' hg (𝟙 _) (y g) (hy g)
let q : (R.pullback f).arrows.FamilyOfElements P := fun Z g hg => z hg
have hq : q.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
|
apply (H2 h₁ g₁).ext
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← op_comp, h₁, h₂]
simpa only [Sieve.pullback_apply, Category.assoc] using hgg
let y : ∀ ⦃Z : C⦄ (g : Z ⟶ Y),
((S.pullback (g ≫ f)).pullback (𝟙 _)).arrows.FamilyOfElements P :=
fun Z g ZZ gg hgg => x (gg ≫ g) (by simpa using hgg)
have hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), (y g).Compatible := by
intro Z g Y₁ Y₂ ZZ g₁ g₂ f₁ f₂ h₁ h₂ h
rw [hx]
rw [reassoc_of% h]
choose z hz using fun ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : R.pullback f g) =>
H2' hg (𝟙 _) (y g) (hy g)
let q : (R.pullback f).arrows.FamilyOfElements P := fun Z g hg => z hg
have hq : q.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
|
Mathlib.CategoryTheory.Sites.Coverage.313_0.qkZFgqEgDC2P633
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R)
|
Mathlib_CategoryTheory_Sites_Coverage
|
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
P : Cᵒᵖ ⥤ Type w
X✝ : C
S✝ : Sieve X✝
X : C
R S : Sieve X
a✝¹ : saturate K X R
a✝ : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R.arrows f → saturate K Y (Sieve.pullback f S)
Y : C
f : Y ⟶ X
H :
∀ {X : C},
∀ R ∈ covering K X,
IsSeparatedFor P R ∧
∀ (x : FamilyOfElements P R), FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H1 : ∀ ⦃Y : C⦄ (f : Y ⟶ X), IsSeparatedFor P (Sieve.pullback f R).arrows
H1' :
∀ ⦃Y : C⦄ (f : Y ⟶ X) (x : FamilyOfElements P (Sieve.pullback f R).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H2 :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f → ∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y), IsSeparatedFor P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows
H2' :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f →
∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y) (x : FamilyOfElements P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
x : FamilyOfElements P (Sieve.pullback f S).arrows
hx : FamilyOfElements.Compatible x
y : ⦃Z : C⦄ → (g : Z ⟶ Y) → FamilyOfElements P (Sieve.pullback (𝟙 Z) (Sieve.pullback (g ≫ f) S)).arrows :=
fun Z g ZZ gg hgg => x (gg ≫ g) (_ : (Sieve.pullback f S).arrows (gg ≫ g))
hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), FamilyOfElements.Compatible (y g)
z : ⦃Z : C⦄ → ⦃g : Z ⟶ Y⦄ → (Sieve.pullback f R).arrows g → P.obj (Opposite.op Z)
hz : ∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : (Sieve.pullback f R).arrows g), FamilyOfElements.IsAmalgamation (y g) (z hg)
q : FamilyOfElements P (Sieve.pullback f R).arrows := fun Z g hg => z hg
Y₁ Y₂ Z : C
g₁ : Z ⟶ Y₁
g₂ : Z ⟶ Y₂
f₁ : Y₁ ⟶ Y
f₂ : Y₂ ⟶ Y
h₁ : (Sieve.pullback f R).arrows f₁
h₂ : (Sieve.pullback f R).arrows f₂
h : g₁ ≫ f₁ = g₂ ≫ f₂
⊢ ∀ ⦃Y_1 : C⦄ ⦃f_1 : Y_1 ⟶ Z⦄,
(Sieve.pullback g₁ (Sieve.pullback (f₁ ≫ f) S)).arrows f_1 →
P.map f_1.op (P.map g₁.op (q f₁ h₁)) = P.map f_1.op (P.map g₂.op (q f₂ h₂))
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
· obtain ⟨T, hT⟩ := y.pullback f S hy
exact ⟨T, Or.inr hT.1, hT.2⟩ }
toPartialOrder := inferInstance
le_sup_left _ _ _ := Set.subset_union_left _ _
le_sup_right _ _ _ := Set.subset_union_right _ _
sup_le _ _ _ hx hy X := Set.union_subset_iff.mpr ⟨hx X, hy X⟩
@[simp]
lemma sup_covering (x y : Coverage C) (B : C) :
(x ⊔ y).covering B = x.covering B ∪ y.covering B :=
rfl
end Coverage
open Coverage
namespace Presieve
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← op_comp, h₁, h₂]
simpa only [Sieve.pullback_apply, Category.assoc] using hgg
let y : ∀ ⦃Z : C⦄ (g : Z ⟶ Y),
((S.pullback (g ≫ f)).pullback (𝟙 _)).arrows.FamilyOfElements P :=
fun Z g ZZ gg hgg => x (gg ≫ g) (by simpa using hgg)
have hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), (y g).Compatible := by
intro Z g Y₁ Y₂ ZZ g₁ g₂ f₁ f₂ h₁ h₂ h
rw [hx]
rw [reassoc_of% h]
choose z hz using fun ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : R.pullback f g) =>
H2' hg (𝟙 _) (y g) (hy g)
let q : (R.pullback f).arrows.FamilyOfElements P := fun Z g hg => z hg
have hq : q.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
apply (H2 h₁ g₁).ext
|
intro ZZ gg hgg
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← op_comp, h₁, h₂]
simpa only [Sieve.pullback_apply, Category.assoc] using hgg
let y : ∀ ⦃Z : C⦄ (g : Z ⟶ Y),
((S.pullback (g ≫ f)).pullback (𝟙 _)).arrows.FamilyOfElements P :=
fun Z g ZZ gg hgg => x (gg ≫ g) (by simpa using hgg)
have hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), (y g).Compatible := by
intro Z g Y₁ Y₂ ZZ g₁ g₂ f₁ f₂ h₁ h₂ h
rw [hx]
rw [reassoc_of% h]
choose z hz using fun ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : R.pullback f g) =>
H2' hg (𝟙 _) (y g) (hy g)
let q : (R.pullback f).arrows.FamilyOfElements P := fun Z g hg => z hg
have hq : q.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
apply (H2 h₁ g₁).ext
|
Mathlib.CategoryTheory.Sites.Coverage.313_0.qkZFgqEgDC2P633
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R)
|
Mathlib_CategoryTheory_Sites_Coverage
|
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
P : Cᵒᵖ ⥤ Type w
X✝ : C
S✝ : Sieve X✝
X : C
R S : Sieve X
a✝¹ : saturate K X R
a✝ : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R.arrows f → saturate K Y (Sieve.pullback f S)
Y : C
f : Y ⟶ X
H :
∀ {X : C},
∀ R ∈ covering K X,
IsSeparatedFor P R ∧
∀ (x : FamilyOfElements P R), FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H1 : ∀ ⦃Y : C⦄ (f : Y ⟶ X), IsSeparatedFor P (Sieve.pullback f R).arrows
H1' :
∀ ⦃Y : C⦄ (f : Y ⟶ X) (x : FamilyOfElements P (Sieve.pullback f R).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H2 :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f → ∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y), IsSeparatedFor P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows
H2' :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f →
∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y) (x : FamilyOfElements P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
x : FamilyOfElements P (Sieve.pullback f S).arrows
hx : FamilyOfElements.Compatible x
y : ⦃Z : C⦄ → (g : Z ⟶ Y) → FamilyOfElements P (Sieve.pullback (𝟙 Z) (Sieve.pullback (g ≫ f) S)).arrows :=
fun Z g ZZ gg hgg => x (gg ≫ g) (_ : (Sieve.pullback f S).arrows (gg ≫ g))
hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), FamilyOfElements.Compatible (y g)
z : ⦃Z : C⦄ → ⦃g : Z ⟶ Y⦄ → (Sieve.pullback f R).arrows g → P.obj (Opposite.op Z)
hz : ∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : (Sieve.pullback f R).arrows g), FamilyOfElements.IsAmalgamation (y g) (z hg)
q : FamilyOfElements P (Sieve.pullback f R).arrows := fun Z g hg => z hg
Y₁ Y₂ Z : C
g₁ : Z ⟶ Y₁
g₂ : Z ⟶ Y₂
f₁ : Y₁ ⟶ Y
f₂ : Y₂ ⟶ Y
h₁ : (Sieve.pullback f R).arrows f₁
h₂ : (Sieve.pullback f R).arrows f₂
h : g₁ ≫ f₁ = g₂ ≫ f₂
ZZ : C
gg : ZZ ⟶ Z
hgg : (Sieve.pullback g₁ (Sieve.pullback (f₁ ≫ f) S)).arrows gg
⊢ P.map gg.op (P.map g₁.op (q f₁ h₁)) = P.map gg.op (P.map g₂.op (q f₂ h₂))
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
· obtain ⟨T, hT⟩ := y.pullback f S hy
exact ⟨T, Or.inr hT.1, hT.2⟩ }
toPartialOrder := inferInstance
le_sup_left _ _ _ := Set.subset_union_left _ _
le_sup_right _ _ _ := Set.subset_union_right _ _
sup_le _ _ _ hx hy X := Set.union_subset_iff.mpr ⟨hx X, hy X⟩
@[simp]
lemma sup_covering (x y : Coverage C) (B : C) :
(x ⊔ y).covering B = x.covering B ∪ y.covering B :=
rfl
end Coverage
open Coverage
namespace Presieve
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← op_comp, h₁, h₂]
simpa only [Sieve.pullback_apply, Category.assoc] using hgg
let y : ∀ ⦃Z : C⦄ (g : Z ⟶ Y),
((S.pullback (g ≫ f)).pullback (𝟙 _)).arrows.FamilyOfElements P :=
fun Z g ZZ gg hgg => x (gg ≫ g) (by simpa using hgg)
have hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), (y g).Compatible := by
intro Z g Y₁ Y₂ ZZ g₁ g₂ f₁ f₂ h₁ h₂ h
rw [hx]
rw [reassoc_of% h]
choose z hz using fun ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : R.pullback f g) =>
H2' hg (𝟙 _) (y g) (hy g)
let q : (R.pullback f).arrows.FamilyOfElements P := fun Z g hg => z hg
have hq : q.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
apply (H2 h₁ g₁).ext
intro ZZ gg hgg
|
simp only [← types_comp_apply]
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← op_comp, h₁, h₂]
simpa only [Sieve.pullback_apply, Category.assoc] using hgg
let y : ∀ ⦃Z : C⦄ (g : Z ⟶ Y),
((S.pullback (g ≫ f)).pullback (𝟙 _)).arrows.FamilyOfElements P :=
fun Z g ZZ gg hgg => x (gg ≫ g) (by simpa using hgg)
have hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), (y g).Compatible := by
intro Z g Y₁ Y₂ ZZ g₁ g₂ f₁ f₂ h₁ h₂ h
rw [hx]
rw [reassoc_of% h]
choose z hz using fun ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : R.pullback f g) =>
H2' hg (𝟙 _) (y g) (hy g)
let q : (R.pullback f).arrows.FamilyOfElements P := fun Z g hg => z hg
have hq : q.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
apply (H2 h₁ g₁).ext
intro ZZ gg hgg
|
Mathlib.CategoryTheory.Sites.Coverage.313_0.qkZFgqEgDC2P633
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R)
|
Mathlib_CategoryTheory_Sites_Coverage
|
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
P : Cᵒᵖ ⥤ Type w
X✝ : C
S✝ : Sieve X✝
X : C
R S : Sieve X
a✝¹ : saturate K X R
a✝ : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R.arrows f → saturate K Y (Sieve.pullback f S)
Y : C
f : Y ⟶ X
H :
∀ {X : C},
∀ R ∈ covering K X,
IsSeparatedFor P R ∧
∀ (x : FamilyOfElements P R), FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H1 : ∀ ⦃Y : C⦄ (f : Y ⟶ X), IsSeparatedFor P (Sieve.pullback f R).arrows
H1' :
∀ ⦃Y : C⦄ (f : Y ⟶ X) (x : FamilyOfElements P (Sieve.pullback f R).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H2 :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f → ∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y), IsSeparatedFor P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows
H2' :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f →
∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y) (x : FamilyOfElements P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
x : FamilyOfElements P (Sieve.pullback f S).arrows
hx : FamilyOfElements.Compatible x
y : ⦃Z : C⦄ → (g : Z ⟶ Y) → FamilyOfElements P (Sieve.pullback (𝟙 Z) (Sieve.pullback (g ≫ f) S)).arrows :=
fun Z g ZZ gg hgg => x (gg ≫ g) (_ : (Sieve.pullback f S).arrows (gg ≫ g))
hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), FamilyOfElements.Compatible (y g)
z : ⦃Z : C⦄ → ⦃g : Z ⟶ Y⦄ → (Sieve.pullback f R).arrows g → P.obj (Opposite.op Z)
hz : ∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : (Sieve.pullback f R).arrows g), FamilyOfElements.IsAmalgamation (y g) (z hg)
q : FamilyOfElements P (Sieve.pullback f R).arrows := fun Z g hg => z hg
Y₁ Y₂ Z : C
g₁ : Z ⟶ Y₁
g₂ : Z ⟶ Y₂
f₁ : Y₁ ⟶ Y
f₂ : Y₂ ⟶ Y
h₁ : (Sieve.pullback f R).arrows f₁
h₂ : (Sieve.pullback f R).arrows f₂
h : g₁ ≫ f₁ = g₂ ≫ f₂
ZZ : C
gg : ZZ ⟶ Z
hgg : (Sieve.pullback g₁ (Sieve.pullback (f₁ ≫ f) S)).arrows gg
⊢ (P.map g₁.op ≫ P.map gg.op) (z h₁) = (P.map g₂.op ≫ P.map gg.op) (z h₂)
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
· obtain ⟨T, hT⟩ := y.pullback f S hy
exact ⟨T, Or.inr hT.1, hT.2⟩ }
toPartialOrder := inferInstance
le_sup_left _ _ _ := Set.subset_union_left _ _
le_sup_right _ _ _ := Set.subset_union_right _ _
sup_le _ _ _ hx hy X := Set.union_subset_iff.mpr ⟨hx X, hy X⟩
@[simp]
lemma sup_covering (x y : Coverage C) (B : C) :
(x ⊔ y).covering B = x.covering B ∪ y.covering B :=
rfl
end Coverage
open Coverage
namespace Presieve
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← op_comp, h₁, h₂]
simpa only [Sieve.pullback_apply, Category.assoc] using hgg
let y : ∀ ⦃Z : C⦄ (g : Z ⟶ Y),
((S.pullback (g ≫ f)).pullback (𝟙 _)).arrows.FamilyOfElements P :=
fun Z g ZZ gg hgg => x (gg ≫ g) (by simpa using hgg)
have hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), (y g).Compatible := by
intro Z g Y₁ Y₂ ZZ g₁ g₂ f₁ f₂ h₁ h₂ h
rw [hx]
rw [reassoc_of% h]
choose z hz using fun ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : R.pullback f g) =>
H2' hg (𝟙 _) (y g) (hy g)
let q : (R.pullback f).arrows.FamilyOfElements P := fun Z g hg => z hg
have hq : q.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
apply (H2 h₁ g₁).ext
intro ZZ gg hgg
simp only [← types_comp_apply]
|
rw [← P.map_comp, ← P.map_comp, ← op_comp, ← op_comp, hz, hz]
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← op_comp, h₁, h₂]
simpa only [Sieve.pullback_apply, Category.assoc] using hgg
let y : ∀ ⦃Z : C⦄ (g : Z ⟶ Y),
((S.pullback (g ≫ f)).pullback (𝟙 _)).arrows.FamilyOfElements P :=
fun Z g ZZ gg hgg => x (gg ≫ g) (by simpa using hgg)
have hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), (y g).Compatible := by
intro Z g Y₁ Y₂ ZZ g₁ g₂ f₁ f₂ h₁ h₂ h
rw [hx]
rw [reassoc_of% h]
choose z hz using fun ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : R.pullback f g) =>
H2' hg (𝟙 _) (y g) (hy g)
let q : (R.pullback f).arrows.FamilyOfElements P := fun Z g hg => z hg
have hq : q.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
apply (H2 h₁ g₁).ext
intro ZZ gg hgg
simp only [← types_comp_apply]
|
Mathlib.CategoryTheory.Sites.Coverage.313_0.qkZFgqEgDC2P633
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R)
|
Mathlib_CategoryTheory_Sites_Coverage
|
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
P : Cᵒᵖ ⥤ Type w
X✝ : C
S✝ : Sieve X✝
X : C
R S : Sieve X
a✝¹ : saturate K X R
a✝ : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R.arrows f → saturate K Y (Sieve.pullback f S)
Y : C
f : Y ⟶ X
H :
∀ {X : C},
∀ R ∈ covering K X,
IsSeparatedFor P R ∧
∀ (x : FamilyOfElements P R), FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H1 : ∀ ⦃Y : C⦄ (f : Y ⟶ X), IsSeparatedFor P (Sieve.pullback f R).arrows
H1' :
∀ ⦃Y : C⦄ (f : Y ⟶ X) (x : FamilyOfElements P (Sieve.pullback f R).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H2 :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f → ∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y), IsSeparatedFor P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows
H2' :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f →
∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y) (x : FamilyOfElements P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
x : FamilyOfElements P (Sieve.pullback f S).arrows
hx : FamilyOfElements.Compatible x
y : ⦃Z : C⦄ → (g : Z ⟶ Y) → FamilyOfElements P (Sieve.pullback (𝟙 Z) (Sieve.pullback (g ≫ f) S)).arrows :=
fun Z g ZZ gg hgg => x (gg ≫ g) (_ : (Sieve.pullback f S).arrows (gg ≫ g))
hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), FamilyOfElements.Compatible (y g)
z : ⦃Z : C⦄ → ⦃g : Z ⟶ Y⦄ → (Sieve.pullback f R).arrows g → P.obj (Opposite.op Z)
hz : ∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : (Sieve.pullback f R).arrows g), FamilyOfElements.IsAmalgamation (y g) (z hg)
q : FamilyOfElements P (Sieve.pullback f R).arrows := fun Z g hg => z hg
Y₁ Y₂ Z : C
g₁ : Z ⟶ Y₁
g₂ : Z ⟶ Y₂
f₁ : Y₁ ⟶ Y
f₂ : Y₂ ⟶ Y
h₁ : (Sieve.pullback f R).arrows f₁
h₂ : (Sieve.pullback f R).arrows f₂
h : g₁ ≫ f₁ = g₂ ≫ f₂
ZZ : C
gg : ZZ ⟶ Z
hgg : (Sieve.pullback g₁ (Sieve.pullback (f₁ ≫ f) S)).arrows gg
⊢ y f₁ (gg ≫ g₁) ?h = y f₂ (gg ≫ g₂) ?h
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
· obtain ⟨T, hT⟩ := y.pullback f S hy
exact ⟨T, Or.inr hT.1, hT.2⟩ }
toPartialOrder := inferInstance
le_sup_left _ _ _ := Set.subset_union_left _ _
le_sup_right _ _ _ := Set.subset_union_right _ _
sup_le _ _ _ hx hy X := Set.union_subset_iff.mpr ⟨hx X, hy X⟩
@[simp]
lemma sup_covering (x y : Coverage C) (B : C) :
(x ⊔ y).covering B = x.covering B ∪ y.covering B :=
rfl
end Coverage
open Coverage
namespace Presieve
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← op_comp, h₁, h₂]
simpa only [Sieve.pullback_apply, Category.assoc] using hgg
let y : ∀ ⦃Z : C⦄ (g : Z ⟶ Y),
((S.pullback (g ≫ f)).pullback (𝟙 _)).arrows.FamilyOfElements P :=
fun Z g ZZ gg hgg => x (gg ≫ g) (by simpa using hgg)
have hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), (y g).Compatible := by
intro Z g Y₁ Y₂ ZZ g₁ g₂ f₁ f₂ h₁ h₂ h
rw [hx]
rw [reassoc_of% h]
choose z hz using fun ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : R.pullback f g) =>
H2' hg (𝟙 _) (y g) (hy g)
let q : (R.pullback f).arrows.FamilyOfElements P := fun Z g hg => z hg
have hq : q.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
apply (H2 h₁ g₁).ext
intro ZZ gg hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← P.map_comp, ← op_comp, ← op_comp, hz, hz]
·
|
dsimp
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← op_comp, h₁, h₂]
simpa only [Sieve.pullback_apply, Category.assoc] using hgg
let y : ∀ ⦃Z : C⦄ (g : Z ⟶ Y),
((S.pullback (g ≫ f)).pullback (𝟙 _)).arrows.FamilyOfElements P :=
fun Z g ZZ gg hgg => x (gg ≫ g) (by simpa using hgg)
have hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), (y g).Compatible := by
intro Z g Y₁ Y₂ ZZ g₁ g₂ f₁ f₂ h₁ h₂ h
rw [hx]
rw [reassoc_of% h]
choose z hz using fun ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : R.pullback f g) =>
H2' hg (𝟙 _) (y g) (hy g)
let q : (R.pullback f).arrows.FamilyOfElements P := fun Z g hg => z hg
have hq : q.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
apply (H2 h₁ g₁).ext
intro ZZ gg hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← P.map_comp, ← op_comp, ← op_comp, hz, hz]
·
|
Mathlib.CategoryTheory.Sites.Coverage.313_0.qkZFgqEgDC2P633
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R)
|
Mathlib_CategoryTheory_Sites_Coverage
|
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
P : Cᵒᵖ ⥤ Type w
X✝ : C
S✝ : Sieve X✝
X : C
R S : Sieve X
a✝¹ : saturate K X R
a✝ : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R.arrows f → saturate K Y (Sieve.pullback f S)
Y : C
f : Y ⟶ X
H :
∀ {X : C},
∀ R ∈ covering K X,
IsSeparatedFor P R ∧
∀ (x : FamilyOfElements P R), FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H1 : ∀ ⦃Y : C⦄ (f : Y ⟶ X), IsSeparatedFor P (Sieve.pullback f R).arrows
H1' :
∀ ⦃Y : C⦄ (f : Y ⟶ X) (x : FamilyOfElements P (Sieve.pullback f R).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H2 :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f → ∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y), IsSeparatedFor P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows
H2' :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f →
∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y) (x : FamilyOfElements P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
x : FamilyOfElements P (Sieve.pullback f S).arrows
hx : FamilyOfElements.Compatible x
y : ⦃Z : C⦄ → (g : Z ⟶ Y) → FamilyOfElements P (Sieve.pullback (𝟙 Z) (Sieve.pullback (g ≫ f) S)).arrows :=
fun Z g ZZ gg hgg => x (gg ≫ g) (_ : (Sieve.pullback f S).arrows (gg ≫ g))
hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), FamilyOfElements.Compatible (y g)
z : ⦃Z : C⦄ → ⦃g : Z ⟶ Y⦄ → (Sieve.pullback f R).arrows g → P.obj (Opposite.op Z)
hz : ∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : (Sieve.pullback f R).arrows g), FamilyOfElements.IsAmalgamation (y g) (z hg)
q : FamilyOfElements P (Sieve.pullback f R).arrows := fun Z g hg => z hg
Y₁ Y₂ Z : C
g₁ : Z ⟶ Y₁
g₂ : Z ⟶ Y₂
f₁ : Y₁ ⟶ Y
f₂ : Y₂ ⟶ Y
h₁ : (Sieve.pullback f R).arrows f₁
h₂ : (Sieve.pullback f R).arrows f₂
h : g₁ ≫ f₁ = g₂ ≫ f₂
ZZ : C
gg : ZZ ⟶ Z
hgg : (Sieve.pullback g₁ (Sieve.pullback (f₁ ≫ f) S)).arrows gg
⊢ x ((gg ≫ g₁) ≫ f₁) (_ : S.arrows (((gg ≫ g₁) ≫ f₁) ≫ f)) = x ((gg ≫ g₂) ≫ f₂) (_ : S.arrows (((gg ≫ g₂) ≫ f₂) ≫ f))
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
· obtain ⟨T, hT⟩ := y.pullback f S hy
exact ⟨T, Or.inr hT.1, hT.2⟩ }
toPartialOrder := inferInstance
le_sup_left _ _ _ := Set.subset_union_left _ _
le_sup_right _ _ _ := Set.subset_union_right _ _
sup_le _ _ _ hx hy X := Set.union_subset_iff.mpr ⟨hx X, hy X⟩
@[simp]
lemma sup_covering (x y : Coverage C) (B : C) :
(x ⊔ y).covering B = x.covering B ∪ y.covering B :=
rfl
end Coverage
open Coverage
namespace Presieve
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← op_comp, h₁, h₂]
simpa only [Sieve.pullback_apply, Category.assoc] using hgg
let y : ∀ ⦃Z : C⦄ (g : Z ⟶ Y),
((S.pullback (g ≫ f)).pullback (𝟙 _)).arrows.FamilyOfElements P :=
fun Z g ZZ gg hgg => x (gg ≫ g) (by simpa using hgg)
have hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), (y g).Compatible := by
intro Z g Y₁ Y₂ ZZ g₁ g₂ f₁ f₂ h₁ h₂ h
rw [hx]
rw [reassoc_of% h]
choose z hz using fun ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : R.pullback f g) =>
H2' hg (𝟙 _) (y g) (hy g)
let q : (R.pullback f).arrows.FamilyOfElements P := fun Z g hg => z hg
have hq : q.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
apply (H2 h₁ g₁).ext
intro ZZ gg hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← P.map_comp, ← op_comp, ← op_comp, hz, hz]
· dsimp;
|
congr 1
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← op_comp, h₁, h₂]
simpa only [Sieve.pullback_apply, Category.assoc] using hgg
let y : ∀ ⦃Z : C⦄ (g : Z ⟶ Y),
((S.pullback (g ≫ f)).pullback (𝟙 _)).arrows.FamilyOfElements P :=
fun Z g ZZ gg hgg => x (gg ≫ g) (by simpa using hgg)
have hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), (y g).Compatible := by
intro Z g Y₁ Y₂ ZZ g₁ g₂ f₁ f₂ h₁ h₂ h
rw [hx]
rw [reassoc_of% h]
choose z hz using fun ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : R.pullback f g) =>
H2' hg (𝟙 _) (y g) (hy g)
let q : (R.pullback f).arrows.FamilyOfElements P := fun Z g hg => z hg
have hq : q.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
apply (H2 h₁ g₁).ext
intro ZZ gg hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← P.map_comp, ← op_comp, ← op_comp, hz, hz]
· dsimp;
|
Mathlib.CategoryTheory.Sites.Coverage.313_0.qkZFgqEgDC2P633
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R)
|
Mathlib_CategoryTheory_Sites_Coverage
|
case e_f
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
P : Cᵒᵖ ⥤ Type w
X✝ : C
S✝ : Sieve X✝
X : C
R S : Sieve X
a✝¹ : saturate K X R
a✝ : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R.arrows f → saturate K Y (Sieve.pullback f S)
Y : C
f : Y ⟶ X
H :
∀ {X : C},
∀ R ∈ covering K X,
IsSeparatedFor P R ∧
∀ (x : FamilyOfElements P R), FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H1 : ∀ ⦃Y : C⦄ (f : Y ⟶ X), IsSeparatedFor P (Sieve.pullback f R).arrows
H1' :
∀ ⦃Y : C⦄ (f : Y ⟶ X) (x : FamilyOfElements P (Sieve.pullback f R).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H2 :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f → ∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y), IsSeparatedFor P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows
H2' :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f →
∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y) (x : FamilyOfElements P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
x : FamilyOfElements P (Sieve.pullback f S).arrows
hx : FamilyOfElements.Compatible x
y : ⦃Z : C⦄ → (g : Z ⟶ Y) → FamilyOfElements P (Sieve.pullback (𝟙 Z) (Sieve.pullback (g ≫ f) S)).arrows :=
fun Z g ZZ gg hgg => x (gg ≫ g) (_ : (Sieve.pullback f S).arrows (gg ≫ g))
hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), FamilyOfElements.Compatible (y g)
z : ⦃Z : C⦄ → ⦃g : Z ⟶ Y⦄ → (Sieve.pullback f R).arrows g → P.obj (Opposite.op Z)
hz : ∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : (Sieve.pullback f R).arrows g), FamilyOfElements.IsAmalgamation (y g) (z hg)
q : FamilyOfElements P (Sieve.pullback f R).arrows := fun Z g hg => z hg
Y₁ Y₂ Z : C
g₁ : Z ⟶ Y₁
g₂ : Z ⟶ Y₂
f₁ : Y₁ ⟶ Y
f₂ : Y₂ ⟶ Y
h₁ : (Sieve.pullback f R).arrows f₁
h₂ : (Sieve.pullback f R).arrows f₂
h : g₁ ≫ f₁ = g₂ ≫ f₂
ZZ : C
gg : ZZ ⟶ Z
hgg : (Sieve.pullback g₁ (Sieve.pullback (f₁ ≫ f) S)).arrows gg
⊢ (gg ≫ g₁) ≫ f₁ = (gg ≫ g₂) ≫ f₂
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
· obtain ⟨T, hT⟩ := y.pullback f S hy
exact ⟨T, Or.inr hT.1, hT.2⟩ }
toPartialOrder := inferInstance
le_sup_left _ _ _ := Set.subset_union_left _ _
le_sup_right _ _ _ := Set.subset_union_right _ _
sup_le _ _ _ hx hy X := Set.union_subset_iff.mpr ⟨hx X, hy X⟩
@[simp]
lemma sup_covering (x y : Coverage C) (B : C) :
(x ⊔ y).covering B = x.covering B ∪ y.covering B :=
rfl
end Coverage
open Coverage
namespace Presieve
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← op_comp, h₁, h₂]
simpa only [Sieve.pullback_apply, Category.assoc] using hgg
let y : ∀ ⦃Z : C⦄ (g : Z ⟶ Y),
((S.pullback (g ≫ f)).pullback (𝟙 _)).arrows.FamilyOfElements P :=
fun Z g ZZ gg hgg => x (gg ≫ g) (by simpa using hgg)
have hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), (y g).Compatible := by
intro Z g Y₁ Y₂ ZZ g₁ g₂ f₁ f₂ h₁ h₂ h
rw [hx]
rw [reassoc_of% h]
choose z hz using fun ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : R.pullback f g) =>
H2' hg (𝟙 _) (y g) (hy g)
let q : (R.pullback f).arrows.FamilyOfElements P := fun Z g hg => z hg
have hq : q.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
apply (H2 h₁ g₁).ext
intro ZZ gg hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← P.map_comp, ← op_comp, ← op_comp, hz, hz]
· dsimp; congr 1;
|
simp only [Category.assoc, h]
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← op_comp, h₁, h₂]
simpa only [Sieve.pullback_apply, Category.assoc] using hgg
let y : ∀ ⦃Z : C⦄ (g : Z ⟶ Y),
((S.pullback (g ≫ f)).pullback (𝟙 _)).arrows.FamilyOfElements P :=
fun Z g ZZ gg hgg => x (gg ≫ g) (by simpa using hgg)
have hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), (y g).Compatible := by
intro Z g Y₁ Y₂ ZZ g₁ g₂ f₁ f₂ h₁ h₂ h
rw [hx]
rw [reassoc_of% h]
choose z hz using fun ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : R.pullback f g) =>
H2' hg (𝟙 _) (y g) (hy g)
let q : (R.pullback f).arrows.FamilyOfElements P := fun Z g hg => z hg
have hq : q.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
apply (H2 h₁ g₁).ext
intro ZZ gg hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← P.map_comp, ← op_comp, ← op_comp, hz, hz]
· dsimp; congr 1;
|
Mathlib.CategoryTheory.Sites.Coverage.313_0.qkZFgqEgDC2P633
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R)
|
Mathlib_CategoryTheory_Sites_Coverage
|
case h
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
P : Cᵒᵖ ⥤ Type w
X✝ : C
S✝ : Sieve X✝
X : C
R S : Sieve X
a✝¹ : saturate K X R
a✝ : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R.arrows f → saturate K Y (Sieve.pullback f S)
Y : C
f : Y ⟶ X
H :
∀ {X : C},
∀ R ∈ covering K X,
IsSeparatedFor P R ∧
∀ (x : FamilyOfElements P R), FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H1 : ∀ ⦃Y : C⦄ (f : Y ⟶ X), IsSeparatedFor P (Sieve.pullback f R).arrows
H1' :
∀ ⦃Y : C⦄ (f : Y ⟶ X) (x : FamilyOfElements P (Sieve.pullback f R).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H2 :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f → ∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y), IsSeparatedFor P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows
H2' :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f →
∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y) (x : FamilyOfElements P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
x : FamilyOfElements P (Sieve.pullback f S).arrows
hx : FamilyOfElements.Compatible x
y : ⦃Z : C⦄ → (g : Z ⟶ Y) → FamilyOfElements P (Sieve.pullback (𝟙 Z) (Sieve.pullback (g ≫ f) S)).arrows :=
fun Z g ZZ gg hgg => x (gg ≫ g) (_ : (Sieve.pullback f S).arrows (gg ≫ g))
hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), FamilyOfElements.Compatible (y g)
z : ⦃Z : C⦄ → ⦃g : Z ⟶ Y⦄ → (Sieve.pullback f R).arrows g → P.obj (Opposite.op Z)
hz : ∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : (Sieve.pullback f R).arrows g), FamilyOfElements.IsAmalgamation (y g) (z hg)
q : FamilyOfElements P (Sieve.pullback f R).arrows := fun Z g hg => z hg
Y₁ Y₂ Z : C
g₁ : Z ⟶ Y₁
g₂ : Z ⟶ Y₂
f₁ : Y₁ ⟶ Y
f₂ : Y₂ ⟶ Y
h₁ : (Sieve.pullback f R).arrows f₁
h₂ : (Sieve.pullback f R).arrows f₂
h : g₁ ≫ f₁ = g₂ ≫ f₂
ZZ : C
gg : ZZ ⟶ Z
hgg : (Sieve.pullback g₁ (Sieve.pullback (f₁ ≫ f) S)).arrows gg
⊢ (Sieve.pullback (𝟙 Y₂) (Sieve.pullback (f₂ ≫ f) S)).arrows (gg ≫ g₂)
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
· obtain ⟨T, hT⟩ := y.pullback f S hy
exact ⟨T, Or.inr hT.1, hT.2⟩ }
toPartialOrder := inferInstance
le_sup_left _ _ _ := Set.subset_union_left _ _
le_sup_right _ _ _ := Set.subset_union_right _ _
sup_le _ _ _ hx hy X := Set.union_subset_iff.mpr ⟨hx X, hy X⟩
@[simp]
lemma sup_covering (x y : Coverage C) (B : C) :
(x ⊔ y).covering B = x.covering B ∪ y.covering B :=
rfl
end Coverage
open Coverage
namespace Presieve
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← op_comp, h₁, h₂]
simpa only [Sieve.pullback_apply, Category.assoc] using hgg
let y : ∀ ⦃Z : C⦄ (g : Z ⟶ Y),
((S.pullback (g ≫ f)).pullback (𝟙 _)).arrows.FamilyOfElements P :=
fun Z g ZZ gg hgg => x (gg ≫ g) (by simpa using hgg)
have hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), (y g).Compatible := by
intro Z g Y₁ Y₂ ZZ g₁ g₂ f₁ f₂ h₁ h₂ h
rw [hx]
rw [reassoc_of% h]
choose z hz using fun ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : R.pullback f g) =>
H2' hg (𝟙 _) (y g) (hy g)
let q : (R.pullback f).arrows.FamilyOfElements P := fun Z g hg => z hg
have hq : q.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
apply (H2 h₁ g₁).ext
intro ZZ gg hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← P.map_comp, ← op_comp, ← op_comp, hz, hz]
· dsimp; congr 1; simp only [Category.assoc, h]
·
|
simpa [reassoc_of% h] using hgg
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← op_comp, h₁, h₂]
simpa only [Sieve.pullback_apply, Category.assoc] using hgg
let y : ∀ ⦃Z : C⦄ (g : Z ⟶ Y),
((S.pullback (g ≫ f)).pullback (𝟙 _)).arrows.FamilyOfElements P :=
fun Z g ZZ gg hgg => x (gg ≫ g) (by simpa using hgg)
have hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), (y g).Compatible := by
intro Z g Y₁ Y₂ ZZ g₁ g₂ f₁ f₂ h₁ h₂ h
rw [hx]
rw [reassoc_of% h]
choose z hz using fun ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : R.pullback f g) =>
H2' hg (𝟙 _) (y g) (hy g)
let q : (R.pullback f).arrows.FamilyOfElements P := fun Z g hg => z hg
have hq : q.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
apply (H2 h₁ g₁).ext
intro ZZ gg hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← P.map_comp, ← op_comp, ← op_comp, hz, hz]
· dsimp; congr 1; simp only [Category.assoc, h]
·
|
Mathlib.CategoryTheory.Sites.Coverage.313_0.qkZFgqEgDC2P633
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R)
|
Mathlib_CategoryTheory_Sites_Coverage
|
case h
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
P : Cᵒᵖ ⥤ Type w
X✝ : C
S✝ : Sieve X✝
X : C
R S : Sieve X
a✝¹ : saturate K X R
a✝ : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R.arrows f → saturate K Y (Sieve.pullback f S)
Y : C
f : Y ⟶ X
H :
∀ {X : C},
∀ R ∈ covering K X,
IsSeparatedFor P R ∧
∀ (x : FamilyOfElements P R), FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H1 : ∀ ⦃Y : C⦄ (f : Y ⟶ X), IsSeparatedFor P (Sieve.pullback f R).arrows
H1' :
∀ ⦃Y : C⦄ (f : Y ⟶ X) (x : FamilyOfElements P (Sieve.pullback f R).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H2 :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f → ∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y), IsSeparatedFor P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows
H2' :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f →
∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y) (x : FamilyOfElements P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
x : FamilyOfElements P (Sieve.pullback f S).arrows
hx : FamilyOfElements.Compatible x
y : ⦃Z : C⦄ → (g : Z ⟶ Y) → FamilyOfElements P (Sieve.pullback (𝟙 Z) (Sieve.pullback (g ≫ f) S)).arrows :=
fun Z g ZZ gg hgg => x (gg ≫ g) (_ : (Sieve.pullback f S).arrows (gg ≫ g))
hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), FamilyOfElements.Compatible (y g)
z : ⦃Z : C⦄ → ⦃g : Z ⟶ Y⦄ → (Sieve.pullback f R).arrows g → P.obj (Opposite.op Z)
hz : ∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : (Sieve.pullback f R).arrows g), FamilyOfElements.IsAmalgamation (y g) (z hg)
q : FamilyOfElements P (Sieve.pullback f R).arrows := fun Z g hg => z hg
Y₁ Y₂ Z : C
g₁ : Z ⟶ Y₁
g₂ : Z ⟶ Y₂
f₁ : Y₁ ⟶ Y
f₂ : Y₂ ⟶ Y
h₁ : (Sieve.pullback f R).arrows f₁
h₂ : (Sieve.pullback f R).arrows f₂
h : g₁ ≫ f₁ = g₂ ≫ f₂
ZZ : C
gg : ZZ ⟶ Z
hgg : (Sieve.pullback g₁ (Sieve.pullback (f₁ ≫ f) S)).arrows gg
⊢ (Sieve.pullback (𝟙 Y₁) (Sieve.pullback (f₁ ≫ f) S)).arrows (gg ≫ g₁)
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
· obtain ⟨T, hT⟩ := y.pullback f S hy
exact ⟨T, Or.inr hT.1, hT.2⟩ }
toPartialOrder := inferInstance
le_sup_left _ _ _ := Set.subset_union_left _ _
le_sup_right _ _ _ := Set.subset_union_right _ _
sup_le _ _ _ hx hy X := Set.union_subset_iff.mpr ⟨hx X, hy X⟩
@[simp]
lemma sup_covering (x y : Coverage C) (B : C) :
(x ⊔ y).covering B = x.covering B ∪ y.covering B :=
rfl
end Coverage
open Coverage
namespace Presieve
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← op_comp, h₁, h₂]
simpa only [Sieve.pullback_apply, Category.assoc] using hgg
let y : ∀ ⦃Z : C⦄ (g : Z ⟶ Y),
((S.pullback (g ≫ f)).pullback (𝟙 _)).arrows.FamilyOfElements P :=
fun Z g ZZ gg hgg => x (gg ≫ g) (by simpa using hgg)
have hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), (y g).Compatible := by
intro Z g Y₁ Y₂ ZZ g₁ g₂ f₁ f₂ h₁ h₂ h
rw [hx]
rw [reassoc_of% h]
choose z hz using fun ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : R.pullback f g) =>
H2' hg (𝟙 _) (y g) (hy g)
let q : (R.pullback f).arrows.FamilyOfElements P := fun Z g hg => z hg
have hq : q.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
apply (H2 h₁ g₁).ext
intro ZZ gg hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← P.map_comp, ← op_comp, ← op_comp, hz, hz]
· dsimp; congr 1; simp only [Category.assoc, h]
· simpa [reassoc_of% h] using hgg
·
|
simpa using hgg
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← op_comp, h₁, h₂]
simpa only [Sieve.pullback_apply, Category.assoc] using hgg
let y : ∀ ⦃Z : C⦄ (g : Z ⟶ Y),
((S.pullback (g ≫ f)).pullback (𝟙 _)).arrows.FamilyOfElements P :=
fun Z g ZZ gg hgg => x (gg ≫ g) (by simpa using hgg)
have hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), (y g).Compatible := by
intro Z g Y₁ Y₂ ZZ g₁ g₂ f₁ f₂ h₁ h₂ h
rw [hx]
rw [reassoc_of% h]
choose z hz using fun ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : R.pullback f g) =>
H2' hg (𝟙 _) (y g) (hy g)
let q : (R.pullback f).arrows.FamilyOfElements P := fun Z g hg => z hg
have hq : q.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
apply (H2 h₁ g₁).ext
intro ZZ gg hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← P.map_comp, ← op_comp, ← op_comp, hz, hz]
· dsimp; congr 1; simp only [Category.assoc, h]
· simpa [reassoc_of% h] using hgg
·
|
Mathlib.CategoryTheory.Sites.Coverage.313_0.qkZFgqEgDC2P633
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R)
|
Mathlib_CategoryTheory_Sites_Coverage
|
case mpr.transitive.refine_2
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
P : Cᵒᵖ ⥤ Type w
X✝ : C
S✝ : Sieve X✝
X : C
R S : Sieve X
a✝¹ : saturate K X R
a✝ : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R.arrows f → saturate K Y (Sieve.pullback f S)
Y : C
f : Y ⟶ X
H :
∀ {X : C},
∀ R ∈ covering K X,
IsSeparatedFor P R ∧
∀ (x : FamilyOfElements P R), FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H1 : ∀ ⦃Y : C⦄ (f : Y ⟶ X), IsSeparatedFor P (Sieve.pullback f R).arrows
H1' :
∀ ⦃Y : C⦄ (f : Y ⟶ X) (x : FamilyOfElements P (Sieve.pullback f R).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H2 :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f → ∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y), IsSeparatedFor P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows
H2' :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f →
∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y) (x : FamilyOfElements P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
x : FamilyOfElements P (Sieve.pullback f S).arrows
hx : FamilyOfElements.Compatible x
y : ⦃Z : C⦄ → (g : Z ⟶ Y) → FamilyOfElements P (Sieve.pullback (𝟙 Z) (Sieve.pullback (g ≫ f) S)).arrows :=
fun Z g ZZ gg hgg => x (gg ≫ g) (_ : (Sieve.pullback f S).arrows (gg ≫ g))
hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), FamilyOfElements.Compatible (y g)
z : ⦃Z : C⦄ → ⦃g : Z ⟶ Y⦄ → (Sieve.pullback f R).arrows g → P.obj (Opposite.op Z)
hz : ∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : (Sieve.pullback f R).arrows g), FamilyOfElements.IsAmalgamation (y g) (z hg)
q : FamilyOfElements P (Sieve.pullback f R).arrows := fun Z g hg => z hg
hq : FamilyOfElements.Compatible q
⊢ ∃ t, FamilyOfElements.IsAmalgamation x t
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
· obtain ⟨T, hT⟩ := y.pullback f S hy
exact ⟨T, Or.inr hT.1, hT.2⟩ }
toPartialOrder := inferInstance
le_sup_left _ _ _ := Set.subset_union_left _ _
le_sup_right _ _ _ := Set.subset_union_right _ _
sup_le _ _ _ hx hy X := Set.union_subset_iff.mpr ⟨hx X, hy X⟩
@[simp]
lemma sup_covering (x y : Coverage C) (B : C) :
(x ⊔ y).covering B = x.covering B ∪ y.covering B :=
rfl
end Coverage
open Coverage
namespace Presieve
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← op_comp, h₁, h₂]
simpa only [Sieve.pullback_apply, Category.assoc] using hgg
let y : ∀ ⦃Z : C⦄ (g : Z ⟶ Y),
((S.pullback (g ≫ f)).pullback (𝟙 _)).arrows.FamilyOfElements P :=
fun Z g ZZ gg hgg => x (gg ≫ g) (by simpa using hgg)
have hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), (y g).Compatible := by
intro Z g Y₁ Y₂ ZZ g₁ g₂ f₁ f₂ h₁ h₂ h
rw [hx]
rw [reassoc_of% h]
choose z hz using fun ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : R.pullback f g) =>
H2' hg (𝟙 _) (y g) (hy g)
let q : (R.pullback f).arrows.FamilyOfElements P := fun Z g hg => z hg
have hq : q.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
apply (H2 h₁ g₁).ext
intro ZZ gg hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← P.map_comp, ← op_comp, ← op_comp, hz, hz]
· dsimp; congr 1; simp only [Category.assoc, h]
· simpa [reassoc_of% h] using hgg
· simpa using hgg
|
obtain ⟨t, ht⟩ := H1' f q hq
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← op_comp, h₁, h₂]
simpa only [Sieve.pullback_apply, Category.assoc] using hgg
let y : ∀ ⦃Z : C⦄ (g : Z ⟶ Y),
((S.pullback (g ≫ f)).pullback (𝟙 _)).arrows.FamilyOfElements P :=
fun Z g ZZ gg hgg => x (gg ≫ g) (by simpa using hgg)
have hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), (y g).Compatible := by
intro Z g Y₁ Y₂ ZZ g₁ g₂ f₁ f₂ h₁ h₂ h
rw [hx]
rw [reassoc_of% h]
choose z hz using fun ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : R.pullback f g) =>
H2' hg (𝟙 _) (y g) (hy g)
let q : (R.pullback f).arrows.FamilyOfElements P := fun Z g hg => z hg
have hq : q.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
apply (H2 h₁ g₁).ext
intro ZZ gg hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← P.map_comp, ← op_comp, ← op_comp, hz, hz]
· dsimp; congr 1; simp only [Category.assoc, h]
· simpa [reassoc_of% h] using hgg
· simpa using hgg
|
Mathlib.CategoryTheory.Sites.Coverage.313_0.qkZFgqEgDC2P633
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R)
|
Mathlib_CategoryTheory_Sites_Coverage
|
case mpr.transitive.refine_2.intro
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
P : Cᵒᵖ ⥤ Type w
X✝ : C
S✝ : Sieve X✝
X : C
R S : Sieve X
a✝¹ : saturate K X R
a✝ : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R.arrows f → saturate K Y (Sieve.pullback f S)
Y : C
f : Y ⟶ X
H :
∀ {X : C},
∀ R ∈ covering K X,
IsSeparatedFor P R ∧
∀ (x : FamilyOfElements P R), FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H1 : ∀ ⦃Y : C⦄ (f : Y ⟶ X), IsSeparatedFor P (Sieve.pullback f R).arrows
H1' :
∀ ⦃Y : C⦄ (f : Y ⟶ X) (x : FamilyOfElements P (Sieve.pullback f R).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H2 :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f → ∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y), IsSeparatedFor P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows
H2' :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f →
∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y) (x : FamilyOfElements P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
x : FamilyOfElements P (Sieve.pullback f S).arrows
hx : FamilyOfElements.Compatible x
y : ⦃Z : C⦄ → (g : Z ⟶ Y) → FamilyOfElements P (Sieve.pullback (𝟙 Z) (Sieve.pullback (g ≫ f) S)).arrows :=
fun Z g ZZ gg hgg => x (gg ≫ g) (_ : (Sieve.pullback f S).arrows (gg ≫ g))
hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), FamilyOfElements.Compatible (y g)
z : ⦃Z : C⦄ → ⦃g : Z ⟶ Y⦄ → (Sieve.pullback f R).arrows g → P.obj (Opposite.op Z)
hz : ∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : (Sieve.pullback f R).arrows g), FamilyOfElements.IsAmalgamation (y g) (z hg)
q : FamilyOfElements P (Sieve.pullback f R).arrows := fun Z g hg => z hg
hq : FamilyOfElements.Compatible q
t : P.obj (Opposite.op Y)
ht : FamilyOfElements.IsAmalgamation q t
⊢ ∃ t, FamilyOfElements.IsAmalgamation x t
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
· obtain ⟨T, hT⟩ := y.pullback f S hy
exact ⟨T, Or.inr hT.1, hT.2⟩ }
toPartialOrder := inferInstance
le_sup_left _ _ _ := Set.subset_union_left _ _
le_sup_right _ _ _ := Set.subset_union_right _ _
sup_le _ _ _ hx hy X := Set.union_subset_iff.mpr ⟨hx X, hy X⟩
@[simp]
lemma sup_covering (x y : Coverage C) (B : C) :
(x ⊔ y).covering B = x.covering B ∪ y.covering B :=
rfl
end Coverage
open Coverage
namespace Presieve
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← op_comp, h₁, h₂]
simpa only [Sieve.pullback_apply, Category.assoc] using hgg
let y : ∀ ⦃Z : C⦄ (g : Z ⟶ Y),
((S.pullback (g ≫ f)).pullback (𝟙 _)).arrows.FamilyOfElements P :=
fun Z g ZZ gg hgg => x (gg ≫ g) (by simpa using hgg)
have hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), (y g).Compatible := by
intro Z g Y₁ Y₂ ZZ g₁ g₂ f₁ f₂ h₁ h₂ h
rw [hx]
rw [reassoc_of% h]
choose z hz using fun ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : R.pullback f g) =>
H2' hg (𝟙 _) (y g) (hy g)
let q : (R.pullback f).arrows.FamilyOfElements P := fun Z g hg => z hg
have hq : q.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
apply (H2 h₁ g₁).ext
intro ZZ gg hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← P.map_comp, ← op_comp, ← op_comp, hz, hz]
· dsimp; congr 1; simp only [Category.assoc, h]
· simpa [reassoc_of% h] using hgg
· simpa using hgg
obtain ⟨t, ht⟩ := H1' f q hq
|
refine ⟨t, fun Z g hg => ?_⟩
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← op_comp, h₁, h₂]
simpa only [Sieve.pullback_apply, Category.assoc] using hgg
let y : ∀ ⦃Z : C⦄ (g : Z ⟶ Y),
((S.pullback (g ≫ f)).pullback (𝟙 _)).arrows.FamilyOfElements P :=
fun Z g ZZ gg hgg => x (gg ≫ g) (by simpa using hgg)
have hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), (y g).Compatible := by
intro Z g Y₁ Y₂ ZZ g₁ g₂ f₁ f₂ h₁ h₂ h
rw [hx]
rw [reassoc_of% h]
choose z hz using fun ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : R.pullback f g) =>
H2' hg (𝟙 _) (y g) (hy g)
let q : (R.pullback f).arrows.FamilyOfElements P := fun Z g hg => z hg
have hq : q.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
apply (H2 h₁ g₁).ext
intro ZZ gg hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← P.map_comp, ← op_comp, ← op_comp, hz, hz]
· dsimp; congr 1; simp only [Category.assoc, h]
· simpa [reassoc_of% h] using hgg
· simpa using hgg
obtain ⟨t, ht⟩ := H1' f q hq
|
Mathlib.CategoryTheory.Sites.Coverage.313_0.qkZFgqEgDC2P633
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R)
|
Mathlib_CategoryTheory_Sites_Coverage
|
case mpr.transitive.refine_2.intro
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
P : Cᵒᵖ ⥤ Type w
X✝ : C
S✝ : Sieve X✝
X : C
R S : Sieve X
a✝¹ : saturate K X R
a✝ : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R.arrows f → saturate K Y (Sieve.pullback f S)
Y : C
f : Y ⟶ X
H :
∀ {X : C},
∀ R ∈ covering K X,
IsSeparatedFor P R ∧
∀ (x : FamilyOfElements P R), FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H1 : ∀ ⦃Y : C⦄ (f : Y ⟶ X), IsSeparatedFor P (Sieve.pullback f R).arrows
H1' :
∀ ⦃Y : C⦄ (f : Y ⟶ X) (x : FamilyOfElements P (Sieve.pullback f R).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H2 :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f → ∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y), IsSeparatedFor P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows
H2' :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f →
∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y) (x : FamilyOfElements P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
x : FamilyOfElements P (Sieve.pullback f S).arrows
hx : FamilyOfElements.Compatible x
y : ⦃Z : C⦄ → (g : Z ⟶ Y) → FamilyOfElements P (Sieve.pullback (𝟙 Z) (Sieve.pullback (g ≫ f) S)).arrows :=
fun Z g ZZ gg hgg => x (gg ≫ g) (_ : (Sieve.pullback f S).arrows (gg ≫ g))
hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), FamilyOfElements.Compatible (y g)
z : ⦃Z : C⦄ → ⦃g : Z ⟶ Y⦄ → (Sieve.pullback f R).arrows g → P.obj (Opposite.op Z)
hz : ∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : (Sieve.pullback f R).arrows g), FamilyOfElements.IsAmalgamation (y g) (z hg)
q : FamilyOfElements P (Sieve.pullback f R).arrows := fun Z g hg => z hg
hq : FamilyOfElements.Compatible q
t : P.obj (Opposite.op Y)
ht : FamilyOfElements.IsAmalgamation q t
Z : C
g : Z ⟶ Y
hg : (Sieve.pullback f S).arrows g
⊢ P.map g.op t = x g hg
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
· obtain ⟨T, hT⟩ := y.pullback f S hy
exact ⟨T, Or.inr hT.1, hT.2⟩ }
toPartialOrder := inferInstance
le_sup_left _ _ _ := Set.subset_union_left _ _
le_sup_right _ _ _ := Set.subset_union_right _ _
sup_le _ _ _ hx hy X := Set.union_subset_iff.mpr ⟨hx X, hy X⟩
@[simp]
lemma sup_covering (x y : Coverage C) (B : C) :
(x ⊔ y).covering B = x.covering B ∪ y.covering B :=
rfl
end Coverage
open Coverage
namespace Presieve
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← op_comp, h₁, h₂]
simpa only [Sieve.pullback_apply, Category.assoc] using hgg
let y : ∀ ⦃Z : C⦄ (g : Z ⟶ Y),
((S.pullback (g ≫ f)).pullback (𝟙 _)).arrows.FamilyOfElements P :=
fun Z g ZZ gg hgg => x (gg ≫ g) (by simpa using hgg)
have hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), (y g).Compatible := by
intro Z g Y₁ Y₂ ZZ g₁ g₂ f₁ f₂ h₁ h₂ h
rw [hx]
rw [reassoc_of% h]
choose z hz using fun ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : R.pullback f g) =>
H2' hg (𝟙 _) (y g) (hy g)
let q : (R.pullback f).arrows.FamilyOfElements P := fun Z g hg => z hg
have hq : q.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
apply (H2 h₁ g₁).ext
intro ZZ gg hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← P.map_comp, ← op_comp, ← op_comp, hz, hz]
· dsimp; congr 1; simp only [Category.assoc, h]
· simpa [reassoc_of% h] using hgg
· simpa using hgg
obtain ⟨t, ht⟩ := H1' f q hq
refine ⟨t, fun Z g hg => ?_⟩
|
refine (H1 (g ≫ f)).ext (fun ZZ gg hgg => ?_)
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← op_comp, h₁, h₂]
simpa only [Sieve.pullback_apply, Category.assoc] using hgg
let y : ∀ ⦃Z : C⦄ (g : Z ⟶ Y),
((S.pullback (g ≫ f)).pullback (𝟙 _)).arrows.FamilyOfElements P :=
fun Z g ZZ gg hgg => x (gg ≫ g) (by simpa using hgg)
have hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), (y g).Compatible := by
intro Z g Y₁ Y₂ ZZ g₁ g₂ f₁ f₂ h₁ h₂ h
rw [hx]
rw [reassoc_of% h]
choose z hz using fun ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : R.pullback f g) =>
H2' hg (𝟙 _) (y g) (hy g)
let q : (R.pullback f).arrows.FamilyOfElements P := fun Z g hg => z hg
have hq : q.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
apply (H2 h₁ g₁).ext
intro ZZ gg hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← P.map_comp, ← op_comp, ← op_comp, hz, hz]
· dsimp; congr 1; simp only [Category.assoc, h]
· simpa [reassoc_of% h] using hgg
· simpa using hgg
obtain ⟨t, ht⟩ := H1' f q hq
refine ⟨t, fun Z g hg => ?_⟩
|
Mathlib.CategoryTheory.Sites.Coverage.313_0.qkZFgqEgDC2P633
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R)
|
Mathlib_CategoryTheory_Sites_Coverage
|
case mpr.transitive.refine_2.intro
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
P : Cᵒᵖ ⥤ Type w
X✝ : C
S✝ : Sieve X✝
X : C
R S : Sieve X
a✝¹ : saturate K X R
a✝ : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R.arrows f → saturate K Y (Sieve.pullback f S)
Y : C
f : Y ⟶ X
H :
∀ {X : C},
∀ R ∈ covering K X,
IsSeparatedFor P R ∧
∀ (x : FamilyOfElements P R), FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H1 : ∀ ⦃Y : C⦄ (f : Y ⟶ X), IsSeparatedFor P (Sieve.pullback f R).arrows
H1' :
∀ ⦃Y : C⦄ (f : Y ⟶ X) (x : FamilyOfElements P (Sieve.pullback f R).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H2 :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f → ∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y), IsSeparatedFor P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows
H2' :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f →
∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y) (x : FamilyOfElements P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
x : FamilyOfElements P (Sieve.pullback f S).arrows
hx : FamilyOfElements.Compatible x
y : ⦃Z : C⦄ → (g : Z ⟶ Y) → FamilyOfElements P (Sieve.pullback (𝟙 Z) (Sieve.pullback (g ≫ f) S)).arrows :=
fun Z g ZZ gg hgg => x (gg ≫ g) (_ : (Sieve.pullback f S).arrows (gg ≫ g))
hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), FamilyOfElements.Compatible (y g)
z : ⦃Z : C⦄ → ⦃g : Z ⟶ Y⦄ → (Sieve.pullback f R).arrows g → P.obj (Opposite.op Z)
hz : ∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : (Sieve.pullback f R).arrows g), FamilyOfElements.IsAmalgamation (y g) (z hg)
q : FamilyOfElements P (Sieve.pullback f R).arrows := fun Z g hg => z hg
hq : FamilyOfElements.Compatible q
t : P.obj (Opposite.op Y)
ht : FamilyOfElements.IsAmalgamation q t
Z : C
g : Z ⟶ Y
hg : (Sieve.pullback f S).arrows g
ZZ : C
gg : ZZ ⟶ Z
hgg : (Sieve.pullback (g ≫ f) R).arrows gg
⊢ P.map gg.op (P.map g.op t) = P.map gg.op (x g hg)
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
· obtain ⟨T, hT⟩ := y.pullback f S hy
exact ⟨T, Or.inr hT.1, hT.2⟩ }
toPartialOrder := inferInstance
le_sup_left _ _ _ := Set.subset_union_left _ _
le_sup_right _ _ _ := Set.subset_union_right _ _
sup_le _ _ _ hx hy X := Set.union_subset_iff.mpr ⟨hx X, hy X⟩
@[simp]
lemma sup_covering (x y : Coverage C) (B : C) :
(x ⊔ y).covering B = x.covering B ∪ y.covering B :=
rfl
end Coverage
open Coverage
namespace Presieve
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← op_comp, h₁, h₂]
simpa only [Sieve.pullback_apply, Category.assoc] using hgg
let y : ∀ ⦃Z : C⦄ (g : Z ⟶ Y),
((S.pullback (g ≫ f)).pullback (𝟙 _)).arrows.FamilyOfElements P :=
fun Z g ZZ gg hgg => x (gg ≫ g) (by simpa using hgg)
have hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), (y g).Compatible := by
intro Z g Y₁ Y₂ ZZ g₁ g₂ f₁ f₂ h₁ h₂ h
rw [hx]
rw [reassoc_of% h]
choose z hz using fun ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : R.pullback f g) =>
H2' hg (𝟙 _) (y g) (hy g)
let q : (R.pullback f).arrows.FamilyOfElements P := fun Z g hg => z hg
have hq : q.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
apply (H2 h₁ g₁).ext
intro ZZ gg hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← P.map_comp, ← op_comp, ← op_comp, hz, hz]
· dsimp; congr 1; simp only [Category.assoc, h]
· simpa [reassoc_of% h] using hgg
· simpa using hgg
obtain ⟨t, ht⟩ := H1' f q hq
refine ⟨t, fun Z g hg => ?_⟩
refine (H1 (g ≫ f)).ext (fun ZZ gg hgg => ?_)
|
rw [← types_comp_apply _ (P.map gg.op), ← P.map_comp, ← op_comp, ht]
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← op_comp, h₁, h₂]
simpa only [Sieve.pullback_apply, Category.assoc] using hgg
let y : ∀ ⦃Z : C⦄ (g : Z ⟶ Y),
((S.pullback (g ≫ f)).pullback (𝟙 _)).arrows.FamilyOfElements P :=
fun Z g ZZ gg hgg => x (gg ≫ g) (by simpa using hgg)
have hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), (y g).Compatible := by
intro Z g Y₁ Y₂ ZZ g₁ g₂ f₁ f₂ h₁ h₂ h
rw [hx]
rw [reassoc_of% h]
choose z hz using fun ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : R.pullback f g) =>
H2' hg (𝟙 _) (y g) (hy g)
let q : (R.pullback f).arrows.FamilyOfElements P := fun Z g hg => z hg
have hq : q.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
apply (H2 h₁ g₁).ext
intro ZZ gg hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← P.map_comp, ← op_comp, ← op_comp, hz, hz]
· dsimp; congr 1; simp only [Category.assoc, h]
· simpa [reassoc_of% h] using hgg
· simpa using hgg
obtain ⟨t, ht⟩ := H1' f q hq
refine ⟨t, fun Z g hg => ?_⟩
refine (H1 (g ≫ f)).ext (fun ZZ gg hgg => ?_)
|
Mathlib.CategoryTheory.Sites.Coverage.313_0.qkZFgqEgDC2P633
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R)
|
Mathlib_CategoryTheory_Sites_Coverage
|
case mpr.transitive.refine_2.intro
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
P : Cᵒᵖ ⥤ Type w
X✝ : C
S✝ : Sieve X✝
X : C
R S : Sieve X
a✝¹ : saturate K X R
a✝ : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R.arrows f → saturate K Y (Sieve.pullback f S)
Y : C
f : Y ⟶ X
H :
∀ {X : C},
∀ R ∈ covering K X,
IsSeparatedFor P R ∧
∀ (x : FamilyOfElements P R), FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H1 : ∀ ⦃Y : C⦄ (f : Y ⟶ X), IsSeparatedFor P (Sieve.pullback f R).arrows
H1' :
∀ ⦃Y : C⦄ (f : Y ⟶ X) (x : FamilyOfElements P (Sieve.pullback f R).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H2 :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f → ∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y), IsSeparatedFor P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows
H2' :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f →
∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y) (x : FamilyOfElements P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
x : FamilyOfElements P (Sieve.pullback f S).arrows
hx : FamilyOfElements.Compatible x
y : ⦃Z : C⦄ → (g : Z ⟶ Y) → FamilyOfElements P (Sieve.pullback (𝟙 Z) (Sieve.pullback (g ≫ f) S)).arrows :=
fun Z g ZZ gg hgg => x (gg ≫ g) (_ : (Sieve.pullback f S).arrows (gg ≫ g))
hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), FamilyOfElements.Compatible (y g)
z : ⦃Z : C⦄ → ⦃g : Z ⟶ Y⦄ → (Sieve.pullback f R).arrows g → P.obj (Opposite.op Z)
hz : ∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : (Sieve.pullback f R).arrows g), FamilyOfElements.IsAmalgamation (y g) (z hg)
q : FamilyOfElements P (Sieve.pullback f R).arrows := fun Z g hg => z hg
hq : FamilyOfElements.Compatible q
t : P.obj (Opposite.op Y)
ht : FamilyOfElements.IsAmalgamation q t
Z : C
g : Z ⟶ Y
hg : (Sieve.pullback f S).arrows g
ZZ : C
gg : ZZ ⟶ Z
hgg : (Sieve.pullback (g ≫ f) R).arrows gg
⊢ q (gg ≫ g) ?mpr.transitive.refine_2.intro.h = P.map gg.op (x g hg)
case mpr.transitive.refine_2.intro.h
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
P : Cᵒᵖ ⥤ Type w
X✝ : C
S✝ : Sieve X✝
X : C
R S : Sieve X
a✝¹ : saturate K X R
a✝ : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R.arrows f → saturate K Y (Sieve.pullback f S)
Y : C
f : Y ⟶ X
H :
∀ {X : C},
∀ R ∈ covering K X,
IsSeparatedFor P R ∧
∀ (x : FamilyOfElements P R), FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H1 : ∀ ⦃Y : C⦄ (f : Y ⟶ X), IsSeparatedFor P (Sieve.pullback f R).arrows
H1' :
∀ ⦃Y : C⦄ (f : Y ⟶ X) (x : FamilyOfElements P (Sieve.pullback f R).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H2 :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f → ∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y), IsSeparatedFor P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows
H2' :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f →
∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y) (x : FamilyOfElements P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
x : FamilyOfElements P (Sieve.pullback f S).arrows
hx : FamilyOfElements.Compatible x
y : ⦃Z : C⦄ → (g : Z ⟶ Y) → FamilyOfElements P (Sieve.pullback (𝟙 Z) (Sieve.pullback (g ≫ f) S)).arrows :=
fun Z g ZZ gg hgg => x (gg ≫ g) (_ : (Sieve.pullback f S).arrows (gg ≫ g))
hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), FamilyOfElements.Compatible (y g)
z : ⦃Z : C⦄ → ⦃g : Z ⟶ Y⦄ → (Sieve.pullback f R).arrows g → P.obj (Opposite.op Z)
hz : ∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : (Sieve.pullback f R).arrows g), FamilyOfElements.IsAmalgamation (y g) (z hg)
q : FamilyOfElements P (Sieve.pullback f R).arrows := fun Z g hg => z hg
hq : FamilyOfElements.Compatible q
t : P.obj (Opposite.op Y)
ht : FamilyOfElements.IsAmalgamation q t
Z : C
g : Z ⟶ Y
hg : (Sieve.pullback f S).arrows g
ZZ : C
gg : ZZ ⟶ Z
hgg : (Sieve.pullback (g ≫ f) R).arrows gg
⊢ (Sieve.pullback f R).arrows (gg ≫ g)
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
· obtain ⟨T, hT⟩ := y.pullback f S hy
exact ⟨T, Or.inr hT.1, hT.2⟩ }
toPartialOrder := inferInstance
le_sup_left _ _ _ := Set.subset_union_left _ _
le_sup_right _ _ _ := Set.subset_union_right _ _
sup_le _ _ _ hx hy X := Set.union_subset_iff.mpr ⟨hx X, hy X⟩
@[simp]
lemma sup_covering (x y : Coverage C) (B : C) :
(x ⊔ y).covering B = x.covering B ∪ y.covering B :=
rfl
end Coverage
open Coverage
namespace Presieve
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← op_comp, h₁, h₂]
simpa only [Sieve.pullback_apply, Category.assoc] using hgg
let y : ∀ ⦃Z : C⦄ (g : Z ⟶ Y),
((S.pullback (g ≫ f)).pullback (𝟙 _)).arrows.FamilyOfElements P :=
fun Z g ZZ gg hgg => x (gg ≫ g) (by simpa using hgg)
have hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), (y g).Compatible := by
intro Z g Y₁ Y₂ ZZ g₁ g₂ f₁ f₂ h₁ h₂ h
rw [hx]
rw [reassoc_of% h]
choose z hz using fun ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : R.pullback f g) =>
H2' hg (𝟙 _) (y g) (hy g)
let q : (R.pullback f).arrows.FamilyOfElements P := fun Z g hg => z hg
have hq : q.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
apply (H2 h₁ g₁).ext
intro ZZ gg hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← P.map_comp, ← op_comp, ← op_comp, hz, hz]
· dsimp; congr 1; simp only [Category.assoc, h]
· simpa [reassoc_of% h] using hgg
· simpa using hgg
obtain ⟨t, ht⟩ := H1' f q hq
refine ⟨t, fun Z g hg => ?_⟩
refine (H1 (g ≫ f)).ext (fun ZZ gg hgg => ?_)
rw [← types_comp_apply _ (P.map gg.op), ← P.map_comp, ← op_comp, ht]
|
swap
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← op_comp, h₁, h₂]
simpa only [Sieve.pullback_apply, Category.assoc] using hgg
let y : ∀ ⦃Z : C⦄ (g : Z ⟶ Y),
((S.pullback (g ≫ f)).pullback (𝟙 _)).arrows.FamilyOfElements P :=
fun Z g ZZ gg hgg => x (gg ≫ g) (by simpa using hgg)
have hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), (y g).Compatible := by
intro Z g Y₁ Y₂ ZZ g₁ g₂ f₁ f₂ h₁ h₂ h
rw [hx]
rw [reassoc_of% h]
choose z hz using fun ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : R.pullback f g) =>
H2' hg (𝟙 _) (y g) (hy g)
let q : (R.pullback f).arrows.FamilyOfElements P := fun Z g hg => z hg
have hq : q.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
apply (H2 h₁ g₁).ext
intro ZZ gg hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← P.map_comp, ← op_comp, ← op_comp, hz, hz]
· dsimp; congr 1; simp only [Category.assoc, h]
· simpa [reassoc_of% h] using hgg
· simpa using hgg
obtain ⟨t, ht⟩ := H1' f q hq
refine ⟨t, fun Z g hg => ?_⟩
refine (H1 (g ≫ f)).ext (fun ZZ gg hgg => ?_)
rw [← types_comp_apply _ (P.map gg.op), ← P.map_comp, ← op_comp, ht]
|
Mathlib.CategoryTheory.Sites.Coverage.313_0.qkZFgqEgDC2P633
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R)
|
Mathlib_CategoryTheory_Sites_Coverage
|
case mpr.transitive.refine_2.intro.h
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
P : Cᵒᵖ ⥤ Type w
X✝ : C
S✝ : Sieve X✝
X : C
R S : Sieve X
a✝¹ : saturate K X R
a✝ : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R.arrows f → saturate K Y (Sieve.pullback f S)
Y : C
f : Y ⟶ X
H :
∀ {X : C},
∀ R ∈ covering K X,
IsSeparatedFor P R ∧
∀ (x : FamilyOfElements P R), FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H1 : ∀ ⦃Y : C⦄ (f : Y ⟶ X), IsSeparatedFor P (Sieve.pullback f R).arrows
H1' :
∀ ⦃Y : C⦄ (f : Y ⟶ X) (x : FamilyOfElements P (Sieve.pullback f R).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H2 :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f → ∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y), IsSeparatedFor P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows
H2' :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f →
∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y) (x : FamilyOfElements P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
x : FamilyOfElements P (Sieve.pullback f S).arrows
hx : FamilyOfElements.Compatible x
y : ⦃Z : C⦄ → (g : Z ⟶ Y) → FamilyOfElements P (Sieve.pullback (𝟙 Z) (Sieve.pullback (g ≫ f) S)).arrows :=
fun Z g ZZ gg hgg => x (gg ≫ g) (_ : (Sieve.pullback f S).arrows (gg ≫ g))
hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), FamilyOfElements.Compatible (y g)
z : ⦃Z : C⦄ → ⦃g : Z ⟶ Y⦄ → (Sieve.pullback f R).arrows g → P.obj (Opposite.op Z)
hz : ∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : (Sieve.pullback f R).arrows g), FamilyOfElements.IsAmalgamation (y g) (z hg)
q : FamilyOfElements P (Sieve.pullback f R).arrows := fun Z g hg => z hg
hq : FamilyOfElements.Compatible q
t : P.obj (Opposite.op Y)
ht : FamilyOfElements.IsAmalgamation q t
Z : C
g : Z ⟶ Y
hg : (Sieve.pullback f S).arrows g
ZZ : C
gg : ZZ ⟶ Z
hgg : (Sieve.pullback (g ≫ f) R).arrows gg
⊢ (Sieve.pullback f R).arrows (gg ≫ g)
case mpr.transitive.refine_2.intro
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
P : Cᵒᵖ ⥤ Type w
X✝ : C
S✝ : Sieve X✝
X : C
R S : Sieve X
a✝¹ : saturate K X R
a✝ : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R.arrows f → saturate K Y (Sieve.pullback f S)
Y : C
f : Y ⟶ X
H :
∀ {X : C},
∀ R ∈ covering K X,
IsSeparatedFor P R ∧
∀ (x : FamilyOfElements P R), FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H1 : ∀ ⦃Y : C⦄ (f : Y ⟶ X), IsSeparatedFor P (Sieve.pullback f R).arrows
H1' :
∀ ⦃Y : C⦄ (f : Y ⟶ X) (x : FamilyOfElements P (Sieve.pullback f R).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H2 :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f → ∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y), IsSeparatedFor P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows
H2' :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f →
∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y) (x : FamilyOfElements P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
x : FamilyOfElements P (Sieve.pullback f S).arrows
hx : FamilyOfElements.Compatible x
y : ⦃Z : C⦄ → (g : Z ⟶ Y) → FamilyOfElements P (Sieve.pullback (𝟙 Z) (Sieve.pullback (g ≫ f) S)).arrows :=
fun Z g ZZ gg hgg => x (gg ≫ g) (_ : (Sieve.pullback f S).arrows (gg ≫ g))
hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), FamilyOfElements.Compatible (y g)
z : ⦃Z : C⦄ → ⦃g : Z ⟶ Y⦄ → (Sieve.pullback f R).arrows g → P.obj (Opposite.op Z)
hz : ∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : (Sieve.pullback f R).arrows g), FamilyOfElements.IsAmalgamation (y g) (z hg)
q : FamilyOfElements P (Sieve.pullback f R).arrows := fun Z g hg => z hg
hq : FamilyOfElements.Compatible q
t : P.obj (Opposite.op Y)
ht : FamilyOfElements.IsAmalgamation q t
Z : C
g : Z ⟶ Y
hg : (Sieve.pullback f S).arrows g
ZZ : C
gg : ZZ ⟶ Z
hgg : (Sieve.pullback (g ≫ f) R).arrows gg
⊢ q (gg ≫ g) ?mpr.transitive.refine_2.intro.h = P.map gg.op (x g hg)
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
· obtain ⟨T, hT⟩ := y.pullback f S hy
exact ⟨T, Or.inr hT.1, hT.2⟩ }
toPartialOrder := inferInstance
le_sup_left _ _ _ := Set.subset_union_left _ _
le_sup_right _ _ _ := Set.subset_union_right _ _
sup_le _ _ _ hx hy X := Set.union_subset_iff.mpr ⟨hx X, hy X⟩
@[simp]
lemma sup_covering (x y : Coverage C) (B : C) :
(x ⊔ y).covering B = x.covering B ∪ y.covering B :=
rfl
end Coverage
open Coverage
namespace Presieve
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← op_comp, h₁, h₂]
simpa only [Sieve.pullback_apply, Category.assoc] using hgg
let y : ∀ ⦃Z : C⦄ (g : Z ⟶ Y),
((S.pullback (g ≫ f)).pullback (𝟙 _)).arrows.FamilyOfElements P :=
fun Z g ZZ gg hgg => x (gg ≫ g) (by simpa using hgg)
have hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), (y g).Compatible := by
intro Z g Y₁ Y₂ ZZ g₁ g₂ f₁ f₂ h₁ h₂ h
rw [hx]
rw [reassoc_of% h]
choose z hz using fun ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : R.pullback f g) =>
H2' hg (𝟙 _) (y g) (hy g)
let q : (R.pullback f).arrows.FamilyOfElements P := fun Z g hg => z hg
have hq : q.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
apply (H2 h₁ g₁).ext
intro ZZ gg hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← P.map_comp, ← op_comp, ← op_comp, hz, hz]
· dsimp; congr 1; simp only [Category.assoc, h]
· simpa [reassoc_of% h] using hgg
· simpa using hgg
obtain ⟨t, ht⟩ := H1' f q hq
refine ⟨t, fun Z g hg => ?_⟩
refine (H1 (g ≫ f)).ext (fun ZZ gg hgg => ?_)
rw [← types_comp_apply _ (P.map gg.op), ← P.map_comp, ← op_comp, ht]
swap;
|
simpa using hgg
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← op_comp, h₁, h₂]
simpa only [Sieve.pullback_apply, Category.assoc] using hgg
let y : ∀ ⦃Z : C⦄ (g : Z ⟶ Y),
((S.pullback (g ≫ f)).pullback (𝟙 _)).arrows.FamilyOfElements P :=
fun Z g ZZ gg hgg => x (gg ≫ g) (by simpa using hgg)
have hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), (y g).Compatible := by
intro Z g Y₁ Y₂ ZZ g₁ g₂ f₁ f₂ h₁ h₂ h
rw [hx]
rw [reassoc_of% h]
choose z hz using fun ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : R.pullback f g) =>
H2' hg (𝟙 _) (y g) (hy g)
let q : (R.pullback f).arrows.FamilyOfElements P := fun Z g hg => z hg
have hq : q.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
apply (H2 h₁ g₁).ext
intro ZZ gg hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← P.map_comp, ← op_comp, ← op_comp, hz, hz]
· dsimp; congr 1; simp only [Category.assoc, h]
· simpa [reassoc_of% h] using hgg
· simpa using hgg
obtain ⟨t, ht⟩ := H1' f q hq
refine ⟨t, fun Z g hg => ?_⟩
refine (H1 (g ≫ f)).ext (fun ZZ gg hgg => ?_)
rw [← types_comp_apply _ (P.map gg.op), ← P.map_comp, ← op_comp, ht]
swap;
|
Mathlib.CategoryTheory.Sites.Coverage.313_0.qkZFgqEgDC2P633
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R)
|
Mathlib_CategoryTheory_Sites_Coverage
|
case mpr.transitive.refine_2.intro
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
P : Cᵒᵖ ⥤ Type w
X✝ : C
S✝ : Sieve X✝
X : C
R S : Sieve X
a✝¹ : saturate K X R
a✝ : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R.arrows f → saturate K Y (Sieve.pullback f S)
Y : C
f : Y ⟶ X
H :
∀ {X : C},
∀ R ∈ covering K X,
IsSeparatedFor P R ∧
∀ (x : FamilyOfElements P R), FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H1 : ∀ ⦃Y : C⦄ (f : Y ⟶ X), IsSeparatedFor P (Sieve.pullback f R).arrows
H1' :
∀ ⦃Y : C⦄ (f : Y ⟶ X) (x : FamilyOfElements P (Sieve.pullback f R).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H2 :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f → ∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y), IsSeparatedFor P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows
H2' :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f →
∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y) (x : FamilyOfElements P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
x : FamilyOfElements P (Sieve.pullback f S).arrows
hx : FamilyOfElements.Compatible x
y : ⦃Z : C⦄ → (g : Z ⟶ Y) → FamilyOfElements P (Sieve.pullback (𝟙 Z) (Sieve.pullback (g ≫ f) S)).arrows :=
fun Z g ZZ gg hgg => x (gg ≫ g) (_ : (Sieve.pullback f S).arrows (gg ≫ g))
hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), FamilyOfElements.Compatible (y g)
z : ⦃Z : C⦄ → ⦃g : Z ⟶ Y⦄ → (Sieve.pullback f R).arrows g → P.obj (Opposite.op Z)
hz : ∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : (Sieve.pullback f R).arrows g), FamilyOfElements.IsAmalgamation (y g) (z hg)
q : FamilyOfElements P (Sieve.pullback f R).arrows := fun Z g hg => z hg
hq : FamilyOfElements.Compatible q
t : P.obj (Opposite.op Y)
ht : FamilyOfElements.IsAmalgamation q t
Z : C
g : Z ⟶ Y
hg : (Sieve.pullback f S).arrows g
ZZ : C
gg : ZZ ⟶ Z
hgg : (Sieve.pullback (g ≫ f) R).arrows gg
⊢ q (gg ≫ g) (_ : (Sieve.pullback f R).arrows (gg ≫ g)) = P.map gg.op (x g hg)
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
· obtain ⟨T, hT⟩ := y.pullback f S hy
exact ⟨T, Or.inr hT.1, hT.2⟩ }
toPartialOrder := inferInstance
le_sup_left _ _ _ := Set.subset_union_left _ _
le_sup_right _ _ _ := Set.subset_union_right _ _
sup_le _ _ _ hx hy X := Set.union_subset_iff.mpr ⟨hx X, hy X⟩
@[simp]
lemma sup_covering (x y : Coverage C) (B : C) :
(x ⊔ y).covering B = x.covering B ∪ y.covering B :=
rfl
end Coverage
open Coverage
namespace Presieve
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← op_comp, h₁, h₂]
simpa only [Sieve.pullback_apply, Category.assoc] using hgg
let y : ∀ ⦃Z : C⦄ (g : Z ⟶ Y),
((S.pullback (g ≫ f)).pullback (𝟙 _)).arrows.FamilyOfElements P :=
fun Z g ZZ gg hgg => x (gg ≫ g) (by simpa using hgg)
have hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), (y g).Compatible := by
intro Z g Y₁ Y₂ ZZ g₁ g₂ f₁ f₂ h₁ h₂ h
rw [hx]
rw [reassoc_of% h]
choose z hz using fun ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : R.pullback f g) =>
H2' hg (𝟙 _) (y g) (hy g)
let q : (R.pullback f).arrows.FamilyOfElements P := fun Z g hg => z hg
have hq : q.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
apply (H2 h₁ g₁).ext
intro ZZ gg hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← P.map_comp, ← op_comp, ← op_comp, hz, hz]
· dsimp; congr 1; simp only [Category.assoc, h]
· simpa [reassoc_of% h] using hgg
· simpa using hgg
obtain ⟨t, ht⟩ := H1' f q hq
refine ⟨t, fun Z g hg => ?_⟩
refine (H1 (g ≫ f)).ext (fun ZZ gg hgg => ?_)
rw [← types_comp_apply _ (P.map gg.op), ← P.map_comp, ← op_comp, ht]
swap; simpa using hgg
|
refine (H2 hgg (𝟙 _)).ext (fun ZZZ ggg hggg => ?_)
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← op_comp, h₁, h₂]
simpa only [Sieve.pullback_apply, Category.assoc] using hgg
let y : ∀ ⦃Z : C⦄ (g : Z ⟶ Y),
((S.pullback (g ≫ f)).pullback (𝟙 _)).arrows.FamilyOfElements P :=
fun Z g ZZ gg hgg => x (gg ≫ g) (by simpa using hgg)
have hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), (y g).Compatible := by
intro Z g Y₁ Y₂ ZZ g₁ g₂ f₁ f₂ h₁ h₂ h
rw [hx]
rw [reassoc_of% h]
choose z hz using fun ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : R.pullback f g) =>
H2' hg (𝟙 _) (y g) (hy g)
let q : (R.pullback f).arrows.FamilyOfElements P := fun Z g hg => z hg
have hq : q.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
apply (H2 h₁ g₁).ext
intro ZZ gg hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← P.map_comp, ← op_comp, ← op_comp, hz, hz]
· dsimp; congr 1; simp only [Category.assoc, h]
· simpa [reassoc_of% h] using hgg
· simpa using hgg
obtain ⟨t, ht⟩ := H1' f q hq
refine ⟨t, fun Z g hg => ?_⟩
refine (H1 (g ≫ f)).ext (fun ZZ gg hgg => ?_)
rw [← types_comp_apply _ (P.map gg.op), ← P.map_comp, ← op_comp, ht]
swap; simpa using hgg
|
Mathlib.CategoryTheory.Sites.Coverage.313_0.qkZFgqEgDC2P633
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R)
|
Mathlib_CategoryTheory_Sites_Coverage
|
case mpr.transitive.refine_2.intro
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
P : Cᵒᵖ ⥤ Type w
X✝ : C
S✝ : Sieve X✝
X : C
R S : Sieve X
a✝¹ : saturate K X R
a✝ : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R.arrows f → saturate K Y (Sieve.pullback f S)
Y : C
f : Y ⟶ X
H :
∀ {X : C},
∀ R ∈ covering K X,
IsSeparatedFor P R ∧
∀ (x : FamilyOfElements P R), FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H1 : ∀ ⦃Y : C⦄ (f : Y ⟶ X), IsSeparatedFor P (Sieve.pullback f R).arrows
H1' :
∀ ⦃Y : C⦄ (f : Y ⟶ X) (x : FamilyOfElements P (Sieve.pullback f R).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H2 :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f → ∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y), IsSeparatedFor P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows
H2' :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f →
∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y) (x : FamilyOfElements P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
x : FamilyOfElements P (Sieve.pullback f S).arrows
hx : FamilyOfElements.Compatible x
y : ⦃Z : C⦄ → (g : Z ⟶ Y) → FamilyOfElements P (Sieve.pullback (𝟙 Z) (Sieve.pullback (g ≫ f) S)).arrows :=
fun Z g ZZ gg hgg => x (gg ≫ g) (_ : (Sieve.pullback f S).arrows (gg ≫ g))
hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), FamilyOfElements.Compatible (y g)
z : ⦃Z : C⦄ → ⦃g : Z ⟶ Y⦄ → (Sieve.pullback f R).arrows g → P.obj (Opposite.op Z)
hz : ∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : (Sieve.pullback f R).arrows g), FamilyOfElements.IsAmalgamation (y g) (z hg)
q : FamilyOfElements P (Sieve.pullback f R).arrows := fun Z g hg => z hg
hq : FamilyOfElements.Compatible q
t : P.obj (Opposite.op Y)
ht : FamilyOfElements.IsAmalgamation q t
Z : C
g : Z ⟶ Y
hg : (Sieve.pullback f S).arrows g
ZZ : C
gg : ZZ ⟶ Z
hgg : (Sieve.pullback (g ≫ f) R).arrows gg
ZZZ : C
ggg : ZZZ ⟶ ZZ
hggg : (Sieve.pullback (𝟙 ZZ) (Sieve.pullback (gg ≫ g ≫ f) S)).arrows ggg
⊢ P.map ggg.op (q (gg ≫ g) (_ : (Sieve.pullback f R).arrows (gg ≫ g))) = P.map ggg.op (P.map gg.op (x g hg))
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
· obtain ⟨T, hT⟩ := y.pullback f S hy
exact ⟨T, Or.inr hT.1, hT.2⟩ }
toPartialOrder := inferInstance
le_sup_left _ _ _ := Set.subset_union_left _ _
le_sup_right _ _ _ := Set.subset_union_right _ _
sup_le _ _ _ hx hy X := Set.union_subset_iff.mpr ⟨hx X, hy X⟩
@[simp]
lemma sup_covering (x y : Coverage C) (B : C) :
(x ⊔ y).covering B = x.covering B ∪ y.covering B :=
rfl
end Coverage
open Coverage
namespace Presieve
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← op_comp, h₁, h₂]
simpa only [Sieve.pullback_apply, Category.assoc] using hgg
let y : ∀ ⦃Z : C⦄ (g : Z ⟶ Y),
((S.pullback (g ≫ f)).pullback (𝟙 _)).arrows.FamilyOfElements P :=
fun Z g ZZ gg hgg => x (gg ≫ g) (by simpa using hgg)
have hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), (y g).Compatible := by
intro Z g Y₁ Y₂ ZZ g₁ g₂ f₁ f₂ h₁ h₂ h
rw [hx]
rw [reassoc_of% h]
choose z hz using fun ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : R.pullback f g) =>
H2' hg (𝟙 _) (y g) (hy g)
let q : (R.pullback f).arrows.FamilyOfElements P := fun Z g hg => z hg
have hq : q.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
apply (H2 h₁ g₁).ext
intro ZZ gg hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← P.map_comp, ← op_comp, ← op_comp, hz, hz]
· dsimp; congr 1; simp only [Category.assoc, h]
· simpa [reassoc_of% h] using hgg
· simpa using hgg
obtain ⟨t, ht⟩ := H1' f q hq
refine ⟨t, fun Z g hg => ?_⟩
refine (H1 (g ≫ f)).ext (fun ZZ gg hgg => ?_)
rw [← types_comp_apply _ (P.map gg.op), ← P.map_comp, ← op_comp, ht]
swap; simpa using hgg
refine (H2 hgg (𝟙 _)).ext (fun ZZZ ggg hggg => ?_)
|
rw [← types_comp_apply _ (P.map ggg.op), ← P.map_comp, ← op_comp, hz]
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← op_comp, h₁, h₂]
simpa only [Sieve.pullback_apply, Category.assoc] using hgg
let y : ∀ ⦃Z : C⦄ (g : Z ⟶ Y),
((S.pullback (g ≫ f)).pullback (𝟙 _)).arrows.FamilyOfElements P :=
fun Z g ZZ gg hgg => x (gg ≫ g) (by simpa using hgg)
have hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), (y g).Compatible := by
intro Z g Y₁ Y₂ ZZ g₁ g₂ f₁ f₂ h₁ h₂ h
rw [hx]
rw [reassoc_of% h]
choose z hz using fun ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : R.pullback f g) =>
H2' hg (𝟙 _) (y g) (hy g)
let q : (R.pullback f).arrows.FamilyOfElements P := fun Z g hg => z hg
have hq : q.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
apply (H2 h₁ g₁).ext
intro ZZ gg hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← P.map_comp, ← op_comp, ← op_comp, hz, hz]
· dsimp; congr 1; simp only [Category.assoc, h]
· simpa [reassoc_of% h] using hgg
· simpa using hgg
obtain ⟨t, ht⟩ := H1' f q hq
refine ⟨t, fun Z g hg => ?_⟩
refine (H1 (g ≫ f)).ext (fun ZZ gg hgg => ?_)
rw [← types_comp_apply _ (P.map gg.op), ← P.map_comp, ← op_comp, ht]
swap; simpa using hgg
refine (H2 hgg (𝟙 _)).ext (fun ZZZ ggg hggg => ?_)
|
Mathlib.CategoryTheory.Sites.Coverage.313_0.qkZFgqEgDC2P633
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R)
|
Mathlib_CategoryTheory_Sites_Coverage
|
case mpr.transitive.refine_2.intro
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
P : Cᵒᵖ ⥤ Type w
X✝ : C
S✝ : Sieve X✝
X : C
R S : Sieve X
a✝¹ : saturate K X R
a✝ : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R.arrows f → saturate K Y (Sieve.pullback f S)
Y : C
f : Y ⟶ X
H :
∀ {X : C},
∀ R ∈ covering K X,
IsSeparatedFor P R ∧
∀ (x : FamilyOfElements P R), FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H1 : ∀ ⦃Y : C⦄ (f : Y ⟶ X), IsSeparatedFor P (Sieve.pullback f R).arrows
H1' :
∀ ⦃Y : C⦄ (f : Y ⟶ X) (x : FamilyOfElements P (Sieve.pullback f R).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H2 :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f → ∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y), IsSeparatedFor P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows
H2' :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f →
∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y) (x : FamilyOfElements P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
x : FamilyOfElements P (Sieve.pullback f S).arrows
hx : FamilyOfElements.Compatible x
y : ⦃Z : C⦄ → (g : Z ⟶ Y) → FamilyOfElements P (Sieve.pullback (𝟙 Z) (Sieve.pullback (g ≫ f) S)).arrows :=
fun Z g ZZ gg hgg => x (gg ≫ g) (_ : (Sieve.pullback f S).arrows (gg ≫ g))
hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), FamilyOfElements.Compatible (y g)
z : ⦃Z : C⦄ → ⦃g : Z ⟶ Y⦄ → (Sieve.pullback f R).arrows g → P.obj (Opposite.op Z)
hz : ∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : (Sieve.pullback f R).arrows g), FamilyOfElements.IsAmalgamation (y g) (z hg)
q : FamilyOfElements P (Sieve.pullback f R).arrows := fun Z g hg => z hg
hq : FamilyOfElements.Compatible q
t : P.obj (Opposite.op Y)
ht : FamilyOfElements.IsAmalgamation q t
Z : C
g : Z ⟶ Y
hg : (Sieve.pullback f S).arrows g
ZZ : C
gg : ZZ ⟶ Z
hgg : (Sieve.pullback (g ≫ f) R).arrows gg
ZZZ : C
ggg : ZZZ ⟶ ZZ
hggg : (Sieve.pullback (𝟙 ZZ) (Sieve.pullback (gg ≫ g ≫ f) S)).arrows ggg
⊢ y (gg ≫ g) ggg ?mpr.transitive.refine_2.intro.h = P.map (ggg ≫ gg).op (x g hg)
case mpr.transitive.refine_2.intro.h
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
P : Cᵒᵖ ⥤ Type w
X✝ : C
S✝ : Sieve X✝
X : C
R S : Sieve X
a✝¹ : saturate K X R
a✝ : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R.arrows f → saturate K Y (Sieve.pullback f S)
Y : C
f : Y ⟶ X
H :
∀ {X : C},
∀ R ∈ covering K X,
IsSeparatedFor P R ∧
∀ (x : FamilyOfElements P R), FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H1 : ∀ ⦃Y : C⦄ (f : Y ⟶ X), IsSeparatedFor P (Sieve.pullback f R).arrows
H1' :
∀ ⦃Y : C⦄ (f : Y ⟶ X) (x : FamilyOfElements P (Sieve.pullback f R).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H2 :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f → ∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y), IsSeparatedFor P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows
H2' :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f →
∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y) (x : FamilyOfElements P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
x : FamilyOfElements P (Sieve.pullback f S).arrows
hx : FamilyOfElements.Compatible x
y : ⦃Z : C⦄ → (g : Z ⟶ Y) → FamilyOfElements P (Sieve.pullback (𝟙 Z) (Sieve.pullback (g ≫ f) S)).arrows :=
fun Z g ZZ gg hgg => x (gg ≫ g) (_ : (Sieve.pullback f S).arrows (gg ≫ g))
hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), FamilyOfElements.Compatible (y g)
z : ⦃Z : C⦄ → ⦃g : Z ⟶ Y⦄ → (Sieve.pullback f R).arrows g → P.obj (Opposite.op Z)
hz : ∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : (Sieve.pullback f R).arrows g), FamilyOfElements.IsAmalgamation (y g) (z hg)
q : FamilyOfElements P (Sieve.pullback f R).arrows := fun Z g hg => z hg
hq : FamilyOfElements.Compatible q
t : P.obj (Opposite.op Y)
ht : FamilyOfElements.IsAmalgamation q t
Z : C
g : Z ⟶ Y
hg : (Sieve.pullback f S).arrows g
ZZ : C
gg : ZZ ⟶ Z
hgg : (Sieve.pullback (g ≫ f) R).arrows gg
ZZZ : C
ggg : ZZZ ⟶ ZZ
hggg : (Sieve.pullback (𝟙 ZZ) (Sieve.pullback (gg ≫ g ≫ f) S)).arrows ggg
⊢ (Sieve.pullback (𝟙 ZZ) (Sieve.pullback ((gg ≫ g) ≫ f) S)).arrows ggg
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
· obtain ⟨T, hT⟩ := y.pullback f S hy
exact ⟨T, Or.inr hT.1, hT.2⟩ }
toPartialOrder := inferInstance
le_sup_left _ _ _ := Set.subset_union_left _ _
le_sup_right _ _ _ := Set.subset_union_right _ _
sup_le _ _ _ hx hy X := Set.union_subset_iff.mpr ⟨hx X, hy X⟩
@[simp]
lemma sup_covering (x y : Coverage C) (B : C) :
(x ⊔ y).covering B = x.covering B ∪ y.covering B :=
rfl
end Coverage
open Coverage
namespace Presieve
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← op_comp, h₁, h₂]
simpa only [Sieve.pullback_apply, Category.assoc] using hgg
let y : ∀ ⦃Z : C⦄ (g : Z ⟶ Y),
((S.pullback (g ≫ f)).pullback (𝟙 _)).arrows.FamilyOfElements P :=
fun Z g ZZ gg hgg => x (gg ≫ g) (by simpa using hgg)
have hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), (y g).Compatible := by
intro Z g Y₁ Y₂ ZZ g₁ g₂ f₁ f₂ h₁ h₂ h
rw [hx]
rw [reassoc_of% h]
choose z hz using fun ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : R.pullback f g) =>
H2' hg (𝟙 _) (y g) (hy g)
let q : (R.pullback f).arrows.FamilyOfElements P := fun Z g hg => z hg
have hq : q.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
apply (H2 h₁ g₁).ext
intro ZZ gg hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← P.map_comp, ← op_comp, ← op_comp, hz, hz]
· dsimp; congr 1; simp only [Category.assoc, h]
· simpa [reassoc_of% h] using hgg
· simpa using hgg
obtain ⟨t, ht⟩ := H1' f q hq
refine ⟨t, fun Z g hg => ?_⟩
refine (H1 (g ≫ f)).ext (fun ZZ gg hgg => ?_)
rw [← types_comp_apply _ (P.map gg.op), ← P.map_comp, ← op_comp, ht]
swap; simpa using hgg
refine (H2 hgg (𝟙 _)).ext (fun ZZZ ggg hggg => ?_)
rw [← types_comp_apply _ (P.map ggg.op), ← P.map_comp, ← op_comp, hz]
|
swap
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← op_comp, h₁, h₂]
simpa only [Sieve.pullback_apply, Category.assoc] using hgg
let y : ∀ ⦃Z : C⦄ (g : Z ⟶ Y),
((S.pullback (g ≫ f)).pullback (𝟙 _)).arrows.FamilyOfElements P :=
fun Z g ZZ gg hgg => x (gg ≫ g) (by simpa using hgg)
have hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), (y g).Compatible := by
intro Z g Y₁ Y₂ ZZ g₁ g₂ f₁ f₂ h₁ h₂ h
rw [hx]
rw [reassoc_of% h]
choose z hz using fun ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : R.pullback f g) =>
H2' hg (𝟙 _) (y g) (hy g)
let q : (R.pullback f).arrows.FamilyOfElements P := fun Z g hg => z hg
have hq : q.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
apply (H2 h₁ g₁).ext
intro ZZ gg hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← P.map_comp, ← op_comp, ← op_comp, hz, hz]
· dsimp; congr 1; simp only [Category.assoc, h]
· simpa [reassoc_of% h] using hgg
· simpa using hgg
obtain ⟨t, ht⟩ := H1' f q hq
refine ⟨t, fun Z g hg => ?_⟩
refine (H1 (g ≫ f)).ext (fun ZZ gg hgg => ?_)
rw [← types_comp_apply _ (P.map gg.op), ← P.map_comp, ← op_comp, ht]
swap; simpa using hgg
refine (H2 hgg (𝟙 _)).ext (fun ZZZ ggg hggg => ?_)
rw [← types_comp_apply _ (P.map ggg.op), ← P.map_comp, ← op_comp, hz]
|
Mathlib.CategoryTheory.Sites.Coverage.313_0.qkZFgqEgDC2P633
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R)
|
Mathlib_CategoryTheory_Sites_Coverage
|
case mpr.transitive.refine_2.intro.h
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
P : Cᵒᵖ ⥤ Type w
X✝ : C
S✝ : Sieve X✝
X : C
R S : Sieve X
a✝¹ : saturate K X R
a✝ : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R.arrows f → saturate K Y (Sieve.pullback f S)
Y : C
f : Y ⟶ X
H :
∀ {X : C},
∀ R ∈ covering K X,
IsSeparatedFor P R ∧
∀ (x : FamilyOfElements P R), FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H1 : ∀ ⦃Y : C⦄ (f : Y ⟶ X), IsSeparatedFor P (Sieve.pullback f R).arrows
H1' :
∀ ⦃Y : C⦄ (f : Y ⟶ X) (x : FamilyOfElements P (Sieve.pullback f R).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H2 :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f → ∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y), IsSeparatedFor P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows
H2' :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f →
∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y) (x : FamilyOfElements P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
x : FamilyOfElements P (Sieve.pullback f S).arrows
hx : FamilyOfElements.Compatible x
y : ⦃Z : C⦄ → (g : Z ⟶ Y) → FamilyOfElements P (Sieve.pullback (𝟙 Z) (Sieve.pullback (g ≫ f) S)).arrows :=
fun Z g ZZ gg hgg => x (gg ≫ g) (_ : (Sieve.pullback f S).arrows (gg ≫ g))
hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), FamilyOfElements.Compatible (y g)
z : ⦃Z : C⦄ → ⦃g : Z ⟶ Y⦄ → (Sieve.pullback f R).arrows g → P.obj (Opposite.op Z)
hz : ∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : (Sieve.pullback f R).arrows g), FamilyOfElements.IsAmalgamation (y g) (z hg)
q : FamilyOfElements P (Sieve.pullback f R).arrows := fun Z g hg => z hg
hq : FamilyOfElements.Compatible q
t : P.obj (Opposite.op Y)
ht : FamilyOfElements.IsAmalgamation q t
Z : C
g : Z ⟶ Y
hg : (Sieve.pullback f S).arrows g
ZZ : C
gg : ZZ ⟶ Z
hgg : (Sieve.pullback (g ≫ f) R).arrows gg
ZZZ : C
ggg : ZZZ ⟶ ZZ
hggg : (Sieve.pullback (𝟙 ZZ) (Sieve.pullback (gg ≫ g ≫ f) S)).arrows ggg
⊢ (Sieve.pullback (𝟙 ZZ) (Sieve.pullback ((gg ≫ g) ≫ f) S)).arrows ggg
case mpr.transitive.refine_2.intro
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
P : Cᵒᵖ ⥤ Type w
X✝ : C
S✝ : Sieve X✝
X : C
R S : Sieve X
a✝¹ : saturate K X R
a✝ : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R.arrows f → saturate K Y (Sieve.pullback f S)
Y : C
f : Y ⟶ X
H :
∀ {X : C},
∀ R ∈ covering K X,
IsSeparatedFor P R ∧
∀ (x : FamilyOfElements P R), FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H1 : ∀ ⦃Y : C⦄ (f : Y ⟶ X), IsSeparatedFor P (Sieve.pullback f R).arrows
H1' :
∀ ⦃Y : C⦄ (f : Y ⟶ X) (x : FamilyOfElements P (Sieve.pullback f R).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H2 :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f → ∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y), IsSeparatedFor P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows
H2' :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f →
∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y) (x : FamilyOfElements P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
x : FamilyOfElements P (Sieve.pullback f S).arrows
hx : FamilyOfElements.Compatible x
y : ⦃Z : C⦄ → (g : Z ⟶ Y) → FamilyOfElements P (Sieve.pullback (𝟙 Z) (Sieve.pullback (g ≫ f) S)).arrows :=
fun Z g ZZ gg hgg => x (gg ≫ g) (_ : (Sieve.pullback f S).arrows (gg ≫ g))
hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), FamilyOfElements.Compatible (y g)
z : ⦃Z : C⦄ → ⦃g : Z ⟶ Y⦄ → (Sieve.pullback f R).arrows g → P.obj (Opposite.op Z)
hz : ∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : (Sieve.pullback f R).arrows g), FamilyOfElements.IsAmalgamation (y g) (z hg)
q : FamilyOfElements P (Sieve.pullback f R).arrows := fun Z g hg => z hg
hq : FamilyOfElements.Compatible q
t : P.obj (Opposite.op Y)
ht : FamilyOfElements.IsAmalgamation q t
Z : C
g : Z ⟶ Y
hg : (Sieve.pullback f S).arrows g
ZZ : C
gg : ZZ ⟶ Z
hgg : (Sieve.pullback (g ≫ f) R).arrows gg
ZZZ : C
ggg : ZZZ ⟶ ZZ
hggg : (Sieve.pullback (𝟙 ZZ) (Sieve.pullback (gg ≫ g ≫ f) S)).arrows ggg
⊢ y (gg ≫ g) ggg ?mpr.transitive.refine_2.intro.h = P.map (ggg ≫ gg).op (x g hg)
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
· obtain ⟨T, hT⟩ := y.pullback f S hy
exact ⟨T, Or.inr hT.1, hT.2⟩ }
toPartialOrder := inferInstance
le_sup_left _ _ _ := Set.subset_union_left _ _
le_sup_right _ _ _ := Set.subset_union_right _ _
sup_le _ _ _ hx hy X := Set.union_subset_iff.mpr ⟨hx X, hy X⟩
@[simp]
lemma sup_covering (x y : Coverage C) (B : C) :
(x ⊔ y).covering B = x.covering B ∪ y.covering B :=
rfl
end Coverage
open Coverage
namespace Presieve
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← op_comp, h₁, h₂]
simpa only [Sieve.pullback_apply, Category.assoc] using hgg
let y : ∀ ⦃Z : C⦄ (g : Z ⟶ Y),
((S.pullback (g ≫ f)).pullback (𝟙 _)).arrows.FamilyOfElements P :=
fun Z g ZZ gg hgg => x (gg ≫ g) (by simpa using hgg)
have hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), (y g).Compatible := by
intro Z g Y₁ Y₂ ZZ g₁ g₂ f₁ f₂ h₁ h₂ h
rw [hx]
rw [reassoc_of% h]
choose z hz using fun ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : R.pullback f g) =>
H2' hg (𝟙 _) (y g) (hy g)
let q : (R.pullback f).arrows.FamilyOfElements P := fun Z g hg => z hg
have hq : q.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
apply (H2 h₁ g₁).ext
intro ZZ gg hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← P.map_comp, ← op_comp, ← op_comp, hz, hz]
· dsimp; congr 1; simp only [Category.assoc, h]
· simpa [reassoc_of% h] using hgg
· simpa using hgg
obtain ⟨t, ht⟩ := H1' f q hq
refine ⟨t, fun Z g hg => ?_⟩
refine (H1 (g ≫ f)).ext (fun ZZ gg hgg => ?_)
rw [← types_comp_apply _ (P.map gg.op), ← P.map_comp, ← op_comp, ht]
swap; simpa using hgg
refine (H2 hgg (𝟙 _)).ext (fun ZZZ ggg hggg => ?_)
rw [← types_comp_apply _ (P.map ggg.op), ← P.map_comp, ← op_comp, hz]
swap;
|
simpa using hggg
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← op_comp, h₁, h₂]
simpa only [Sieve.pullback_apply, Category.assoc] using hgg
let y : ∀ ⦃Z : C⦄ (g : Z ⟶ Y),
((S.pullback (g ≫ f)).pullback (𝟙 _)).arrows.FamilyOfElements P :=
fun Z g ZZ gg hgg => x (gg ≫ g) (by simpa using hgg)
have hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), (y g).Compatible := by
intro Z g Y₁ Y₂ ZZ g₁ g₂ f₁ f₂ h₁ h₂ h
rw [hx]
rw [reassoc_of% h]
choose z hz using fun ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : R.pullback f g) =>
H2' hg (𝟙 _) (y g) (hy g)
let q : (R.pullback f).arrows.FamilyOfElements P := fun Z g hg => z hg
have hq : q.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
apply (H2 h₁ g₁).ext
intro ZZ gg hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← P.map_comp, ← op_comp, ← op_comp, hz, hz]
· dsimp; congr 1; simp only [Category.assoc, h]
· simpa [reassoc_of% h] using hgg
· simpa using hgg
obtain ⟨t, ht⟩ := H1' f q hq
refine ⟨t, fun Z g hg => ?_⟩
refine (H1 (g ≫ f)).ext (fun ZZ gg hgg => ?_)
rw [← types_comp_apply _ (P.map gg.op), ← P.map_comp, ← op_comp, ht]
swap; simpa using hgg
refine (H2 hgg (𝟙 _)).ext (fun ZZZ ggg hggg => ?_)
rw [← types_comp_apply _ (P.map ggg.op), ← P.map_comp, ← op_comp, hz]
swap;
|
Mathlib.CategoryTheory.Sites.Coverage.313_0.qkZFgqEgDC2P633
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R)
|
Mathlib_CategoryTheory_Sites_Coverage
|
case mpr.transitive.refine_2.intro
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
P : Cᵒᵖ ⥤ Type w
X✝ : C
S✝ : Sieve X✝
X : C
R S : Sieve X
a✝¹ : saturate K X R
a✝ : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R.arrows f → saturate K Y (Sieve.pullback f S)
Y : C
f : Y ⟶ X
H :
∀ {X : C},
∀ R ∈ covering K X,
IsSeparatedFor P R ∧
∀ (x : FamilyOfElements P R), FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H1 : ∀ ⦃Y : C⦄ (f : Y ⟶ X), IsSeparatedFor P (Sieve.pullback f R).arrows
H1' :
∀ ⦃Y : C⦄ (f : Y ⟶ X) (x : FamilyOfElements P (Sieve.pullback f R).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H2 :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f → ∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y), IsSeparatedFor P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows
H2' :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f →
∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y) (x : FamilyOfElements P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
x : FamilyOfElements P (Sieve.pullback f S).arrows
hx : FamilyOfElements.Compatible x
y : ⦃Z : C⦄ → (g : Z ⟶ Y) → FamilyOfElements P (Sieve.pullback (𝟙 Z) (Sieve.pullback (g ≫ f) S)).arrows :=
fun Z g ZZ gg hgg => x (gg ≫ g) (_ : (Sieve.pullback f S).arrows (gg ≫ g))
hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), FamilyOfElements.Compatible (y g)
z : ⦃Z : C⦄ → ⦃g : Z ⟶ Y⦄ → (Sieve.pullback f R).arrows g → P.obj (Opposite.op Z)
hz : ∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : (Sieve.pullback f R).arrows g), FamilyOfElements.IsAmalgamation (y g) (z hg)
q : FamilyOfElements P (Sieve.pullback f R).arrows := fun Z g hg => z hg
hq : FamilyOfElements.Compatible q
t : P.obj (Opposite.op Y)
ht : FamilyOfElements.IsAmalgamation q t
Z : C
g : Z ⟶ Y
hg : (Sieve.pullback f S).arrows g
ZZ : C
gg : ZZ ⟶ Z
hgg : (Sieve.pullback (g ≫ f) R).arrows gg
ZZZ : C
ggg : ZZZ ⟶ ZZ
hggg : (Sieve.pullback (𝟙 ZZ) (Sieve.pullback (gg ≫ g ≫ f) S)).arrows ggg
⊢ y (gg ≫ g) ggg (_ : (Sieve.pullback (𝟙 ZZ) (Sieve.pullback ((gg ≫ g) ≫ f) S)).arrows ggg) =
P.map (ggg ≫ gg).op (x g hg)
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
· obtain ⟨T, hT⟩ := y.pullback f S hy
exact ⟨T, Or.inr hT.1, hT.2⟩ }
toPartialOrder := inferInstance
le_sup_left _ _ _ := Set.subset_union_left _ _
le_sup_right _ _ _ := Set.subset_union_right _ _
sup_le _ _ _ hx hy X := Set.union_subset_iff.mpr ⟨hx X, hy X⟩
@[simp]
lemma sup_covering (x y : Coverage C) (B : C) :
(x ⊔ y).covering B = x.covering B ∪ y.covering B :=
rfl
end Coverage
open Coverage
namespace Presieve
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← op_comp, h₁, h₂]
simpa only [Sieve.pullback_apply, Category.assoc] using hgg
let y : ∀ ⦃Z : C⦄ (g : Z ⟶ Y),
((S.pullback (g ≫ f)).pullback (𝟙 _)).arrows.FamilyOfElements P :=
fun Z g ZZ gg hgg => x (gg ≫ g) (by simpa using hgg)
have hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), (y g).Compatible := by
intro Z g Y₁ Y₂ ZZ g₁ g₂ f₁ f₂ h₁ h₂ h
rw [hx]
rw [reassoc_of% h]
choose z hz using fun ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : R.pullback f g) =>
H2' hg (𝟙 _) (y g) (hy g)
let q : (R.pullback f).arrows.FamilyOfElements P := fun Z g hg => z hg
have hq : q.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
apply (H2 h₁ g₁).ext
intro ZZ gg hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← P.map_comp, ← op_comp, ← op_comp, hz, hz]
· dsimp; congr 1; simp only [Category.assoc, h]
· simpa [reassoc_of% h] using hgg
· simpa using hgg
obtain ⟨t, ht⟩ := H1' f q hq
refine ⟨t, fun Z g hg => ?_⟩
refine (H1 (g ≫ f)).ext (fun ZZ gg hgg => ?_)
rw [← types_comp_apply _ (P.map gg.op), ← P.map_comp, ← op_comp, ht]
swap; simpa using hgg
refine (H2 hgg (𝟙 _)).ext (fun ZZZ ggg hggg => ?_)
rw [← types_comp_apply _ (P.map ggg.op), ← P.map_comp, ← op_comp, hz]
swap; simpa using hggg
|
refine (H2 hgg ggg).ext (fun ZZZZ gggg _ => ?_)
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← op_comp, h₁, h₂]
simpa only [Sieve.pullback_apply, Category.assoc] using hgg
let y : ∀ ⦃Z : C⦄ (g : Z ⟶ Y),
((S.pullback (g ≫ f)).pullback (𝟙 _)).arrows.FamilyOfElements P :=
fun Z g ZZ gg hgg => x (gg ≫ g) (by simpa using hgg)
have hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), (y g).Compatible := by
intro Z g Y₁ Y₂ ZZ g₁ g₂ f₁ f₂ h₁ h₂ h
rw [hx]
rw [reassoc_of% h]
choose z hz using fun ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : R.pullback f g) =>
H2' hg (𝟙 _) (y g) (hy g)
let q : (R.pullback f).arrows.FamilyOfElements P := fun Z g hg => z hg
have hq : q.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
apply (H2 h₁ g₁).ext
intro ZZ gg hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← P.map_comp, ← op_comp, ← op_comp, hz, hz]
· dsimp; congr 1; simp only [Category.assoc, h]
· simpa [reassoc_of% h] using hgg
· simpa using hgg
obtain ⟨t, ht⟩ := H1' f q hq
refine ⟨t, fun Z g hg => ?_⟩
refine (H1 (g ≫ f)).ext (fun ZZ gg hgg => ?_)
rw [← types_comp_apply _ (P.map gg.op), ← P.map_comp, ← op_comp, ht]
swap; simpa using hgg
refine (H2 hgg (𝟙 _)).ext (fun ZZZ ggg hggg => ?_)
rw [← types_comp_apply _ (P.map ggg.op), ← P.map_comp, ← op_comp, hz]
swap; simpa using hggg
|
Mathlib.CategoryTheory.Sites.Coverage.313_0.qkZFgqEgDC2P633
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R)
|
Mathlib_CategoryTheory_Sites_Coverage
|
case mpr.transitive.refine_2.intro
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
P : Cᵒᵖ ⥤ Type w
X✝ : C
S✝ : Sieve X✝
X : C
R S : Sieve X
a✝¹ : saturate K X R
a✝ : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R.arrows f → saturate K Y (Sieve.pullback f S)
Y : C
f : Y ⟶ X
H :
∀ {X : C},
∀ R ∈ covering K X,
IsSeparatedFor P R ∧
∀ (x : FamilyOfElements P R), FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H1 : ∀ ⦃Y : C⦄ (f : Y ⟶ X), IsSeparatedFor P (Sieve.pullback f R).arrows
H1' :
∀ ⦃Y : C⦄ (f : Y ⟶ X) (x : FamilyOfElements P (Sieve.pullback f R).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H2 :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f → ∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y), IsSeparatedFor P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows
H2' :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f →
∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y) (x : FamilyOfElements P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
x : FamilyOfElements P (Sieve.pullback f S).arrows
hx : FamilyOfElements.Compatible x
y : ⦃Z : C⦄ → (g : Z ⟶ Y) → FamilyOfElements P (Sieve.pullback (𝟙 Z) (Sieve.pullback (g ≫ f) S)).arrows :=
fun Z g ZZ gg hgg => x (gg ≫ g) (_ : (Sieve.pullback f S).arrows (gg ≫ g))
hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), FamilyOfElements.Compatible (y g)
z : ⦃Z : C⦄ → ⦃g : Z ⟶ Y⦄ → (Sieve.pullback f R).arrows g → P.obj (Opposite.op Z)
hz : ∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : (Sieve.pullback f R).arrows g), FamilyOfElements.IsAmalgamation (y g) (z hg)
q : FamilyOfElements P (Sieve.pullback f R).arrows := fun Z g hg => z hg
hq : FamilyOfElements.Compatible q
t : P.obj (Opposite.op Y)
ht : FamilyOfElements.IsAmalgamation q t
Z : C
g : Z ⟶ Y
hg : (Sieve.pullback f S).arrows g
ZZ : C
gg : ZZ ⟶ Z
hgg : (Sieve.pullback (g ≫ f) R).arrows gg
ZZZ : C
ggg : ZZZ ⟶ ZZ
hggg : (Sieve.pullback (𝟙 ZZ) (Sieve.pullback (gg ≫ g ≫ f) S)).arrows ggg
ZZZZ : C
gggg : ZZZZ ⟶ ZZZ
x✝ : (Sieve.pullback ggg (Sieve.pullback (gg ≫ g ≫ f) S)).arrows gggg
⊢ P.map gggg.op (y (gg ≫ g) ggg (_ : (Sieve.pullback (𝟙 ZZ) (Sieve.pullback ((gg ≫ g) ≫ f) S)).arrows ggg)) =
P.map gggg.op (P.map (ggg ≫ gg).op (x g hg))
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
· obtain ⟨T, hT⟩ := y.pullback f S hy
exact ⟨T, Or.inr hT.1, hT.2⟩ }
toPartialOrder := inferInstance
le_sup_left _ _ _ := Set.subset_union_left _ _
le_sup_right _ _ _ := Set.subset_union_right _ _
sup_le _ _ _ hx hy X := Set.union_subset_iff.mpr ⟨hx X, hy X⟩
@[simp]
lemma sup_covering (x y : Coverage C) (B : C) :
(x ⊔ y).covering B = x.covering B ∪ y.covering B :=
rfl
end Coverage
open Coverage
namespace Presieve
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← op_comp, h₁, h₂]
simpa only [Sieve.pullback_apply, Category.assoc] using hgg
let y : ∀ ⦃Z : C⦄ (g : Z ⟶ Y),
((S.pullback (g ≫ f)).pullback (𝟙 _)).arrows.FamilyOfElements P :=
fun Z g ZZ gg hgg => x (gg ≫ g) (by simpa using hgg)
have hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), (y g).Compatible := by
intro Z g Y₁ Y₂ ZZ g₁ g₂ f₁ f₂ h₁ h₂ h
rw [hx]
rw [reassoc_of% h]
choose z hz using fun ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : R.pullback f g) =>
H2' hg (𝟙 _) (y g) (hy g)
let q : (R.pullback f).arrows.FamilyOfElements P := fun Z g hg => z hg
have hq : q.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
apply (H2 h₁ g₁).ext
intro ZZ gg hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← P.map_comp, ← op_comp, ← op_comp, hz, hz]
· dsimp; congr 1; simp only [Category.assoc, h]
· simpa [reassoc_of% h] using hgg
· simpa using hgg
obtain ⟨t, ht⟩ := H1' f q hq
refine ⟨t, fun Z g hg => ?_⟩
refine (H1 (g ≫ f)).ext (fun ZZ gg hgg => ?_)
rw [← types_comp_apply _ (P.map gg.op), ← P.map_comp, ← op_comp, ht]
swap; simpa using hgg
refine (H2 hgg (𝟙 _)).ext (fun ZZZ ggg hggg => ?_)
rw [← types_comp_apply _ (P.map ggg.op), ← P.map_comp, ← op_comp, hz]
swap; simpa using hggg
refine (H2 hgg ggg).ext (fun ZZZZ gggg _ => ?_)
|
rw [← types_comp_apply _ (P.map gggg.op), ← P.map_comp, ← op_comp]
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← op_comp, h₁, h₂]
simpa only [Sieve.pullback_apply, Category.assoc] using hgg
let y : ∀ ⦃Z : C⦄ (g : Z ⟶ Y),
((S.pullback (g ≫ f)).pullback (𝟙 _)).arrows.FamilyOfElements P :=
fun Z g ZZ gg hgg => x (gg ≫ g) (by simpa using hgg)
have hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), (y g).Compatible := by
intro Z g Y₁ Y₂ ZZ g₁ g₂ f₁ f₂ h₁ h₂ h
rw [hx]
rw [reassoc_of% h]
choose z hz using fun ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : R.pullback f g) =>
H2' hg (𝟙 _) (y g) (hy g)
let q : (R.pullback f).arrows.FamilyOfElements P := fun Z g hg => z hg
have hq : q.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
apply (H2 h₁ g₁).ext
intro ZZ gg hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← P.map_comp, ← op_comp, ← op_comp, hz, hz]
· dsimp; congr 1; simp only [Category.assoc, h]
· simpa [reassoc_of% h] using hgg
· simpa using hgg
obtain ⟨t, ht⟩ := H1' f q hq
refine ⟨t, fun Z g hg => ?_⟩
refine (H1 (g ≫ f)).ext (fun ZZ gg hgg => ?_)
rw [← types_comp_apply _ (P.map gg.op), ← P.map_comp, ← op_comp, ht]
swap; simpa using hgg
refine (H2 hgg (𝟙 _)).ext (fun ZZZ ggg hggg => ?_)
rw [← types_comp_apply _ (P.map ggg.op), ← P.map_comp, ← op_comp, hz]
swap; simpa using hggg
refine (H2 hgg ggg).ext (fun ZZZZ gggg _ => ?_)
|
Mathlib.CategoryTheory.Sites.Coverage.313_0.qkZFgqEgDC2P633
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R)
|
Mathlib_CategoryTheory_Sites_Coverage
|
case mpr.transitive.refine_2.intro
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
P : Cᵒᵖ ⥤ Type w
X✝ : C
S✝ : Sieve X✝
X : C
R S : Sieve X
a✝¹ : saturate K X R
a✝ : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R.arrows f → saturate K Y (Sieve.pullback f S)
Y : C
f : Y ⟶ X
H :
∀ {X : C},
∀ R ∈ covering K X,
IsSeparatedFor P R ∧
∀ (x : FamilyOfElements P R), FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H1 : ∀ ⦃Y : C⦄ (f : Y ⟶ X), IsSeparatedFor P (Sieve.pullback f R).arrows
H1' :
∀ ⦃Y : C⦄ (f : Y ⟶ X) (x : FamilyOfElements P (Sieve.pullback f R).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H2 :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f → ∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y), IsSeparatedFor P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows
H2' :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f →
∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y) (x : FamilyOfElements P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
x : FamilyOfElements P (Sieve.pullback f S).arrows
hx : FamilyOfElements.Compatible x
y : ⦃Z : C⦄ → (g : Z ⟶ Y) → FamilyOfElements P (Sieve.pullback (𝟙 Z) (Sieve.pullback (g ≫ f) S)).arrows :=
fun Z g ZZ gg hgg => x (gg ≫ g) (_ : (Sieve.pullback f S).arrows (gg ≫ g))
hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), FamilyOfElements.Compatible (y g)
z : ⦃Z : C⦄ → ⦃g : Z ⟶ Y⦄ → (Sieve.pullback f R).arrows g → P.obj (Opposite.op Z)
hz : ∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : (Sieve.pullback f R).arrows g), FamilyOfElements.IsAmalgamation (y g) (z hg)
q : FamilyOfElements P (Sieve.pullback f R).arrows := fun Z g hg => z hg
hq : FamilyOfElements.Compatible q
t : P.obj (Opposite.op Y)
ht : FamilyOfElements.IsAmalgamation q t
Z : C
g : Z ⟶ Y
hg : (Sieve.pullback f S).arrows g
ZZ : C
gg : ZZ ⟶ Z
hgg : (Sieve.pullback (g ≫ f) R).arrows gg
ZZZ : C
ggg : ZZZ ⟶ ZZ
hggg : (Sieve.pullback (𝟙 ZZ) (Sieve.pullback (gg ≫ g ≫ f) S)).arrows ggg
ZZZZ : C
gggg : ZZZZ ⟶ ZZZ
x✝ : (Sieve.pullback ggg (Sieve.pullback (gg ≫ g ≫ f) S)).arrows gggg
⊢ P.map gggg.op (y (gg ≫ g) ggg (_ : (Sieve.pullback (𝟙 ZZ) (Sieve.pullback ((gg ≫ g) ≫ f) S)).arrows ggg)) =
P.map (gggg ≫ ggg ≫ gg).op (x g hg)
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
· obtain ⟨T, hT⟩ := y.pullback f S hy
exact ⟨T, Or.inr hT.1, hT.2⟩ }
toPartialOrder := inferInstance
le_sup_left _ _ _ := Set.subset_union_left _ _
le_sup_right _ _ _ := Set.subset_union_right _ _
sup_le _ _ _ hx hy X := Set.union_subset_iff.mpr ⟨hx X, hy X⟩
@[simp]
lemma sup_covering (x y : Coverage C) (B : C) :
(x ⊔ y).covering B = x.covering B ∪ y.covering B :=
rfl
end Coverage
open Coverage
namespace Presieve
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← op_comp, h₁, h₂]
simpa only [Sieve.pullback_apply, Category.assoc] using hgg
let y : ∀ ⦃Z : C⦄ (g : Z ⟶ Y),
((S.pullback (g ≫ f)).pullback (𝟙 _)).arrows.FamilyOfElements P :=
fun Z g ZZ gg hgg => x (gg ≫ g) (by simpa using hgg)
have hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), (y g).Compatible := by
intro Z g Y₁ Y₂ ZZ g₁ g₂ f₁ f₂ h₁ h₂ h
rw [hx]
rw [reassoc_of% h]
choose z hz using fun ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : R.pullback f g) =>
H2' hg (𝟙 _) (y g) (hy g)
let q : (R.pullback f).arrows.FamilyOfElements P := fun Z g hg => z hg
have hq : q.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
apply (H2 h₁ g₁).ext
intro ZZ gg hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← P.map_comp, ← op_comp, ← op_comp, hz, hz]
· dsimp; congr 1; simp only [Category.assoc, h]
· simpa [reassoc_of% h] using hgg
· simpa using hgg
obtain ⟨t, ht⟩ := H1' f q hq
refine ⟨t, fun Z g hg => ?_⟩
refine (H1 (g ≫ f)).ext (fun ZZ gg hgg => ?_)
rw [← types_comp_apply _ (P.map gg.op), ← P.map_comp, ← op_comp, ht]
swap; simpa using hgg
refine (H2 hgg (𝟙 _)).ext (fun ZZZ ggg hggg => ?_)
rw [← types_comp_apply _ (P.map ggg.op), ← P.map_comp, ← op_comp, hz]
swap; simpa using hggg
refine (H2 hgg ggg).ext (fun ZZZZ gggg _ => ?_)
rw [← types_comp_apply _ (P.map gggg.op), ← P.map_comp, ← op_comp]
|
apply hx
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← op_comp, h₁, h₂]
simpa only [Sieve.pullback_apply, Category.assoc] using hgg
let y : ∀ ⦃Z : C⦄ (g : Z ⟶ Y),
((S.pullback (g ≫ f)).pullback (𝟙 _)).arrows.FamilyOfElements P :=
fun Z g ZZ gg hgg => x (gg ≫ g) (by simpa using hgg)
have hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), (y g).Compatible := by
intro Z g Y₁ Y₂ ZZ g₁ g₂ f₁ f₂ h₁ h₂ h
rw [hx]
rw [reassoc_of% h]
choose z hz using fun ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : R.pullback f g) =>
H2' hg (𝟙 _) (y g) (hy g)
let q : (R.pullback f).arrows.FamilyOfElements P := fun Z g hg => z hg
have hq : q.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
apply (H2 h₁ g₁).ext
intro ZZ gg hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← P.map_comp, ← op_comp, ← op_comp, hz, hz]
· dsimp; congr 1; simp only [Category.assoc, h]
· simpa [reassoc_of% h] using hgg
· simpa using hgg
obtain ⟨t, ht⟩ := H1' f q hq
refine ⟨t, fun Z g hg => ?_⟩
refine (H1 (g ≫ f)).ext (fun ZZ gg hgg => ?_)
rw [← types_comp_apply _ (P.map gg.op), ← P.map_comp, ← op_comp, ht]
swap; simpa using hgg
refine (H2 hgg (𝟙 _)).ext (fun ZZZ ggg hggg => ?_)
rw [← types_comp_apply _ (P.map ggg.op), ← P.map_comp, ← op_comp, hz]
swap; simpa using hggg
refine (H2 hgg ggg).ext (fun ZZZZ gggg _ => ?_)
rw [← types_comp_apply _ (P.map gggg.op), ← P.map_comp, ← op_comp]
|
Mathlib.CategoryTheory.Sites.Coverage.313_0.qkZFgqEgDC2P633
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R)
|
Mathlib_CategoryTheory_Sites_Coverage
|
case mpr.transitive.refine_2.intro.a
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
P : Cᵒᵖ ⥤ Type w
X✝ : C
S✝ : Sieve X✝
X : C
R S : Sieve X
a✝¹ : saturate K X R
a✝ : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R.arrows f → saturate K Y (Sieve.pullback f S)
Y : C
f : Y ⟶ X
H :
∀ {X : C},
∀ R ∈ covering K X,
IsSeparatedFor P R ∧
∀ (x : FamilyOfElements P R), FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H1 : ∀ ⦃Y : C⦄ (f : Y ⟶ X), IsSeparatedFor P (Sieve.pullback f R).arrows
H1' :
∀ ⦃Y : C⦄ (f : Y ⟶ X) (x : FamilyOfElements P (Sieve.pullback f R).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
H2 :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f → ∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y), IsSeparatedFor P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows
H2' :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f →
∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y) (x : FamilyOfElements P (Sieve.pullback f_1 (Sieve.pullback f S)).arrows),
FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
x : FamilyOfElements P (Sieve.pullback f S).arrows
hx : FamilyOfElements.Compatible x
y : ⦃Z : C⦄ → (g : Z ⟶ Y) → FamilyOfElements P (Sieve.pullback (𝟙 Z) (Sieve.pullback (g ≫ f) S)).arrows :=
fun Z g ZZ gg hgg => x (gg ≫ g) (_ : (Sieve.pullback f S).arrows (gg ≫ g))
hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), FamilyOfElements.Compatible (y g)
z : ⦃Z : C⦄ → ⦃g : Z ⟶ Y⦄ → (Sieve.pullback f R).arrows g → P.obj (Opposite.op Z)
hz : ∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : (Sieve.pullback f R).arrows g), FamilyOfElements.IsAmalgamation (y g) (z hg)
q : FamilyOfElements P (Sieve.pullback f R).arrows := fun Z g hg => z hg
hq : FamilyOfElements.Compatible q
t : P.obj (Opposite.op Y)
ht : FamilyOfElements.IsAmalgamation q t
Z : C
g : Z ⟶ Y
hg : (Sieve.pullback f S).arrows g
ZZ : C
gg : ZZ ⟶ Z
hgg : (Sieve.pullback (g ≫ f) R).arrows gg
ZZZ : C
ggg : ZZZ ⟶ ZZ
hggg : (Sieve.pullback (𝟙 ZZ) (Sieve.pullback (gg ≫ g ≫ f) S)).arrows ggg
ZZZZ : C
gggg : ZZZZ ⟶ ZZZ
x✝ : (Sieve.pullback ggg (Sieve.pullback (gg ≫ g ≫ f) S)).arrows gggg
⊢ gggg ≫ ggg ≫ gg ≫ g = (gggg ≫ ggg ≫ gg) ≫ g
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
· obtain ⟨T, hT⟩ := y.pullback f S hy
exact ⟨T, Or.inr hT.1, hT.2⟩ }
toPartialOrder := inferInstance
le_sup_left _ _ _ := Set.subset_union_left _ _
le_sup_right _ _ _ := Set.subset_union_right _ _
sup_le _ _ _ hx hy X := Set.union_subset_iff.mpr ⟨hx X, hy X⟩
@[simp]
lemma sup_covering (x y : Coverage C) (B : C) :
(x ⊔ y).covering B = x.covering B ∪ y.covering B :=
rfl
end Coverage
open Coverage
namespace Presieve
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← op_comp, h₁, h₂]
simpa only [Sieve.pullback_apply, Category.assoc] using hgg
let y : ∀ ⦃Z : C⦄ (g : Z ⟶ Y),
((S.pullback (g ≫ f)).pullback (𝟙 _)).arrows.FamilyOfElements P :=
fun Z g ZZ gg hgg => x (gg ≫ g) (by simpa using hgg)
have hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), (y g).Compatible := by
intro Z g Y₁ Y₂ ZZ g₁ g₂ f₁ f₂ h₁ h₂ h
rw [hx]
rw [reassoc_of% h]
choose z hz using fun ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : R.pullback f g) =>
H2' hg (𝟙 _) (y g) (hy g)
let q : (R.pullback f).arrows.FamilyOfElements P := fun Z g hg => z hg
have hq : q.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
apply (H2 h₁ g₁).ext
intro ZZ gg hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← P.map_comp, ← op_comp, ← op_comp, hz, hz]
· dsimp; congr 1; simp only [Category.assoc, h]
· simpa [reassoc_of% h] using hgg
· simpa using hgg
obtain ⟨t, ht⟩ := H1' f q hq
refine ⟨t, fun Z g hg => ?_⟩
refine (H1 (g ≫ f)).ext (fun ZZ gg hgg => ?_)
rw [← types_comp_apply _ (P.map gg.op), ← P.map_comp, ← op_comp, ht]
swap; simpa using hgg
refine (H2 hgg (𝟙 _)).ext (fun ZZZ ggg hggg => ?_)
rw [← types_comp_apply _ (P.map ggg.op), ← P.map_comp, ← op_comp, hz]
swap; simpa using hggg
refine (H2 hgg ggg).ext (fun ZZZZ gggg _ => ?_)
rw [← types_comp_apply _ (P.map gggg.op), ← P.map_comp, ← op_comp]
apply hx
|
simp
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← op_comp, h₁, h₂]
simpa only [Sieve.pullback_apply, Category.assoc] using hgg
let y : ∀ ⦃Z : C⦄ (g : Z ⟶ Y),
((S.pullback (g ≫ f)).pullback (𝟙 _)).arrows.FamilyOfElements P :=
fun Z g ZZ gg hgg => x (gg ≫ g) (by simpa using hgg)
have hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), (y g).Compatible := by
intro Z g Y₁ Y₂ ZZ g₁ g₂ f₁ f₂ h₁ h₂ h
rw [hx]
rw [reassoc_of% h]
choose z hz using fun ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : R.pullback f g) =>
H2' hg (𝟙 _) (y g) (hy g)
let q : (R.pullback f).arrows.FamilyOfElements P := fun Z g hg => z hg
have hq : q.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
apply (H2 h₁ g₁).ext
intro ZZ gg hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← P.map_comp, ← op_comp, ← op_comp, hz, hz]
· dsimp; congr 1; simp only [Category.assoc, h]
· simpa [reassoc_of% h] using hgg
· simpa using hgg
obtain ⟨t, ht⟩ := H1' f q hq
refine ⟨t, fun Z g hg => ?_⟩
refine (H1 (g ≫ f)).ext (fun ZZ gg hgg => ?_)
rw [← types_comp_apply _ (P.map gg.op), ← P.map_comp, ← op_comp, ht]
swap; simpa using hgg
refine (H2 hgg (𝟙 _)).ext (fun ZZZ ggg hggg => ?_)
rw [← types_comp_apply _ (P.map ggg.op), ← P.map_comp, ← op_comp, hz]
swap; simpa using hggg
refine (H2 hgg ggg).ext (fun ZZZZ gggg _ => ?_)
rw [← types_comp_apply _ (P.map gggg.op), ← P.map_comp, ← op_comp]
apply hx
|
Mathlib.CategoryTheory.Sites.Coverage.313_0.qkZFgqEgDC2P633
|
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R)
|
Mathlib_CategoryTheory_Sites_Coverage
|
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K L : Coverage C
P : Cᵒᵖ ⥤ Type w
⊢ IsSheaf (toGrothendieck C (K ⊔ L)) P ↔ IsSheaf (toGrothendieck C K) P ∧ IsSheaf (toGrothendieck C L) P
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
· obtain ⟨T, hT⟩ := y.pullback f S hy
exact ⟨T, Or.inr hT.1, hT.2⟩ }
toPartialOrder := inferInstance
le_sup_left _ _ _ := Set.subset_union_left _ _
le_sup_right _ _ _ := Set.subset_union_right _ _
sup_le _ _ _ hx hy X := Set.union_subset_iff.mpr ⟨hx X, hy X⟩
@[simp]
lemma sup_covering (x y : Coverage C) (B : C) :
(x ⊔ y).covering B = x.covering B ∪ y.covering B :=
rfl
end Coverage
open Coverage
namespace Presieve
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← op_comp, h₁, h₂]
simpa only [Sieve.pullback_apply, Category.assoc] using hgg
let y : ∀ ⦃Z : C⦄ (g : Z ⟶ Y),
((S.pullback (g ≫ f)).pullback (𝟙 _)).arrows.FamilyOfElements P :=
fun Z g ZZ gg hgg => x (gg ≫ g) (by simpa using hgg)
have hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), (y g).Compatible := by
intro Z g Y₁ Y₂ ZZ g₁ g₂ f₁ f₂ h₁ h₂ h
rw [hx]
rw [reassoc_of% h]
choose z hz using fun ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : R.pullback f g) =>
H2' hg (𝟙 _) (y g) (hy g)
let q : (R.pullback f).arrows.FamilyOfElements P := fun Z g hg => z hg
have hq : q.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
apply (H2 h₁ g₁).ext
intro ZZ gg hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← P.map_comp, ← op_comp, ← op_comp, hz, hz]
· dsimp; congr 1; simp only [Category.assoc, h]
· simpa [reassoc_of% h] using hgg
· simpa using hgg
obtain ⟨t, ht⟩ := H1' f q hq
refine ⟨t, fun Z g hg => ?_⟩
refine (H1 (g ≫ f)).ext (fun ZZ gg hgg => ?_)
rw [← types_comp_apply _ (P.map gg.op), ← P.map_comp, ← op_comp, ht]
swap; simpa using hgg
refine (H2 hgg (𝟙 _)).ext (fun ZZZ ggg hggg => ?_)
rw [← types_comp_apply _ (P.map ggg.op), ← P.map_comp, ← op_comp, hz]
swap; simpa using hggg
refine (H2 hgg ggg).ext (fun ZZZZ gggg _ => ?_)
rw [← types_comp_apply _ (P.map gggg.op), ← P.map_comp, ← op_comp]
apply hx
simp
/--
A presheaf is a sheaf for the Grothendieck topology generated by a union of coverages iff it is a
sheaf for the Grothendieck topology generated by each coverage separately.
-/
theorem isSheaf_sup (K L : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
(Presieve.IsSheaf ((K ⊔ L).toGrothendieck C)) P ↔
(Presieve.IsSheaf (K.toGrothendieck C)) P ∧ (Presieve.IsSheaf (L.toGrothendieck C)) P := by
|
refine ⟨fun h ↦ ⟨Presieve.isSheaf_of_le _ ((gi C).gc.monotone_l le_sup_left) h,
Presieve.isSheaf_of_le _ ((gi C).gc.monotone_l le_sup_right) h⟩, fun h ↦ ?_⟩
|
/--
A presheaf is a sheaf for the Grothendieck topology generated by a union of coverages iff it is a
sheaf for the Grothendieck topology generated by each coverage separately.
-/
theorem isSheaf_sup (K L : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
(Presieve.IsSheaf ((K ⊔ L).toGrothendieck C)) P ↔
(Presieve.IsSheaf (K.toGrothendieck C)) P ∧ (Presieve.IsSheaf (L.toGrothendieck C)) P := by
|
Mathlib.CategoryTheory.Sites.Coverage.388_0.qkZFgqEgDC2P633
|
/--
A presheaf is a sheaf for the Grothendieck topology generated by a union of coverages iff it is a
sheaf for the Grothendieck topology generated by each coverage separately.
-/
theorem isSheaf_sup (K L : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
(Presieve.IsSheaf ((K ⊔ L).toGrothendieck C)) P ↔
(Presieve.IsSheaf (K.toGrothendieck C)) P ∧ (Presieve.IsSheaf (L.toGrothendieck C)) P
|
Mathlib_CategoryTheory_Sites_Coverage
|
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K L : Coverage C
P : Cᵒᵖ ⥤ Type w
h : IsSheaf (toGrothendieck C K) P ∧ IsSheaf (toGrothendieck C L) P
⊢ IsSheaf (toGrothendieck C (K ⊔ L)) P
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
· obtain ⟨T, hT⟩ := y.pullback f S hy
exact ⟨T, Or.inr hT.1, hT.2⟩ }
toPartialOrder := inferInstance
le_sup_left _ _ _ := Set.subset_union_left _ _
le_sup_right _ _ _ := Set.subset_union_right _ _
sup_le _ _ _ hx hy X := Set.union_subset_iff.mpr ⟨hx X, hy X⟩
@[simp]
lemma sup_covering (x y : Coverage C) (B : C) :
(x ⊔ y).covering B = x.covering B ∪ y.covering B :=
rfl
end Coverage
open Coverage
namespace Presieve
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← op_comp, h₁, h₂]
simpa only [Sieve.pullback_apply, Category.assoc] using hgg
let y : ∀ ⦃Z : C⦄ (g : Z ⟶ Y),
((S.pullback (g ≫ f)).pullback (𝟙 _)).arrows.FamilyOfElements P :=
fun Z g ZZ gg hgg => x (gg ≫ g) (by simpa using hgg)
have hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), (y g).Compatible := by
intro Z g Y₁ Y₂ ZZ g₁ g₂ f₁ f₂ h₁ h₂ h
rw [hx]
rw [reassoc_of% h]
choose z hz using fun ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : R.pullback f g) =>
H2' hg (𝟙 _) (y g) (hy g)
let q : (R.pullback f).arrows.FamilyOfElements P := fun Z g hg => z hg
have hq : q.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
apply (H2 h₁ g₁).ext
intro ZZ gg hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← P.map_comp, ← op_comp, ← op_comp, hz, hz]
· dsimp; congr 1; simp only [Category.assoc, h]
· simpa [reassoc_of% h] using hgg
· simpa using hgg
obtain ⟨t, ht⟩ := H1' f q hq
refine ⟨t, fun Z g hg => ?_⟩
refine (H1 (g ≫ f)).ext (fun ZZ gg hgg => ?_)
rw [← types_comp_apply _ (P.map gg.op), ← P.map_comp, ← op_comp, ht]
swap; simpa using hgg
refine (H2 hgg (𝟙 _)).ext (fun ZZZ ggg hggg => ?_)
rw [← types_comp_apply _ (P.map ggg.op), ← P.map_comp, ← op_comp, hz]
swap; simpa using hggg
refine (H2 hgg ggg).ext (fun ZZZZ gggg _ => ?_)
rw [← types_comp_apply _ (P.map gggg.op), ← P.map_comp, ← op_comp]
apply hx
simp
/--
A presheaf is a sheaf for the Grothendieck topology generated by a union of coverages iff it is a
sheaf for the Grothendieck topology generated by each coverage separately.
-/
theorem isSheaf_sup (K L : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
(Presieve.IsSheaf ((K ⊔ L).toGrothendieck C)) P ↔
(Presieve.IsSheaf (K.toGrothendieck C)) P ∧ (Presieve.IsSheaf (L.toGrothendieck C)) P := by
refine ⟨fun h ↦ ⟨Presieve.isSheaf_of_le _ ((gi C).gc.monotone_l le_sup_left) h,
Presieve.isSheaf_of_le _ ((gi C).gc.monotone_l le_sup_right) h⟩, fun h ↦ ?_⟩
|
rw [isSheaf_coverage, isSheaf_coverage] at h
|
/--
A presheaf is a sheaf for the Grothendieck topology generated by a union of coverages iff it is a
sheaf for the Grothendieck topology generated by each coverage separately.
-/
theorem isSheaf_sup (K L : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
(Presieve.IsSheaf ((K ⊔ L).toGrothendieck C)) P ↔
(Presieve.IsSheaf (K.toGrothendieck C)) P ∧ (Presieve.IsSheaf (L.toGrothendieck C)) P := by
refine ⟨fun h ↦ ⟨Presieve.isSheaf_of_le _ ((gi C).gc.monotone_l le_sup_left) h,
Presieve.isSheaf_of_le _ ((gi C).gc.monotone_l le_sup_right) h⟩, fun h ↦ ?_⟩
|
Mathlib.CategoryTheory.Sites.Coverage.388_0.qkZFgqEgDC2P633
|
/--
A presheaf is a sheaf for the Grothendieck topology generated by a union of coverages iff it is a
sheaf for the Grothendieck topology generated by each coverage separately.
-/
theorem isSheaf_sup (K L : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
(Presieve.IsSheaf ((K ⊔ L).toGrothendieck C)) P ↔
(Presieve.IsSheaf (K.toGrothendieck C)) P ∧ (Presieve.IsSheaf (L.toGrothendieck C)) P
|
Mathlib_CategoryTheory_Sites_Coverage
|
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K L : Coverage C
P : Cᵒᵖ ⥤ Type w
h : (∀ {X : C}, ∀ R ∈ covering K X, IsSheafFor P R) ∧ ∀ {X : C}, ∀ R ∈ covering L X, IsSheafFor P R
⊢ IsSheaf (toGrothendieck C (K ⊔ L)) P
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
· obtain ⟨T, hT⟩ := y.pullback f S hy
exact ⟨T, Or.inr hT.1, hT.2⟩ }
toPartialOrder := inferInstance
le_sup_left _ _ _ := Set.subset_union_left _ _
le_sup_right _ _ _ := Set.subset_union_right _ _
sup_le _ _ _ hx hy X := Set.union_subset_iff.mpr ⟨hx X, hy X⟩
@[simp]
lemma sup_covering (x y : Coverage C) (B : C) :
(x ⊔ y).covering B = x.covering B ∪ y.covering B :=
rfl
end Coverage
open Coverage
namespace Presieve
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← op_comp, h₁, h₂]
simpa only [Sieve.pullback_apply, Category.assoc] using hgg
let y : ∀ ⦃Z : C⦄ (g : Z ⟶ Y),
((S.pullback (g ≫ f)).pullback (𝟙 _)).arrows.FamilyOfElements P :=
fun Z g ZZ gg hgg => x (gg ≫ g) (by simpa using hgg)
have hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), (y g).Compatible := by
intro Z g Y₁ Y₂ ZZ g₁ g₂ f₁ f₂ h₁ h₂ h
rw [hx]
rw [reassoc_of% h]
choose z hz using fun ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : R.pullback f g) =>
H2' hg (𝟙 _) (y g) (hy g)
let q : (R.pullback f).arrows.FamilyOfElements P := fun Z g hg => z hg
have hq : q.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
apply (H2 h₁ g₁).ext
intro ZZ gg hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← P.map_comp, ← op_comp, ← op_comp, hz, hz]
· dsimp; congr 1; simp only [Category.assoc, h]
· simpa [reassoc_of% h] using hgg
· simpa using hgg
obtain ⟨t, ht⟩ := H1' f q hq
refine ⟨t, fun Z g hg => ?_⟩
refine (H1 (g ≫ f)).ext (fun ZZ gg hgg => ?_)
rw [← types_comp_apply _ (P.map gg.op), ← P.map_comp, ← op_comp, ht]
swap; simpa using hgg
refine (H2 hgg (𝟙 _)).ext (fun ZZZ ggg hggg => ?_)
rw [← types_comp_apply _ (P.map ggg.op), ← P.map_comp, ← op_comp, hz]
swap; simpa using hggg
refine (H2 hgg ggg).ext (fun ZZZZ gggg _ => ?_)
rw [← types_comp_apply _ (P.map gggg.op), ← P.map_comp, ← op_comp]
apply hx
simp
/--
A presheaf is a sheaf for the Grothendieck topology generated by a union of coverages iff it is a
sheaf for the Grothendieck topology generated by each coverage separately.
-/
theorem isSheaf_sup (K L : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
(Presieve.IsSheaf ((K ⊔ L).toGrothendieck C)) P ↔
(Presieve.IsSheaf (K.toGrothendieck C)) P ∧ (Presieve.IsSheaf (L.toGrothendieck C)) P := by
refine ⟨fun h ↦ ⟨Presieve.isSheaf_of_le _ ((gi C).gc.monotone_l le_sup_left) h,
Presieve.isSheaf_of_le _ ((gi C).gc.monotone_l le_sup_right) h⟩, fun h ↦ ?_⟩
rw [isSheaf_coverage, isSheaf_coverage] at h
|
rw [isSheaf_coverage]
|
/--
A presheaf is a sheaf for the Grothendieck topology generated by a union of coverages iff it is a
sheaf for the Grothendieck topology generated by each coverage separately.
-/
theorem isSheaf_sup (K L : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
(Presieve.IsSheaf ((K ⊔ L).toGrothendieck C)) P ↔
(Presieve.IsSheaf (K.toGrothendieck C)) P ∧ (Presieve.IsSheaf (L.toGrothendieck C)) P := by
refine ⟨fun h ↦ ⟨Presieve.isSheaf_of_le _ ((gi C).gc.monotone_l le_sup_left) h,
Presieve.isSheaf_of_le _ ((gi C).gc.monotone_l le_sup_right) h⟩, fun h ↦ ?_⟩
rw [isSheaf_coverage, isSheaf_coverage] at h
|
Mathlib.CategoryTheory.Sites.Coverage.388_0.qkZFgqEgDC2P633
|
/--
A presheaf is a sheaf for the Grothendieck topology generated by a union of coverages iff it is a
sheaf for the Grothendieck topology generated by each coverage separately.
-/
theorem isSheaf_sup (K L : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
(Presieve.IsSheaf ((K ⊔ L).toGrothendieck C)) P ↔
(Presieve.IsSheaf (K.toGrothendieck C)) P ∧ (Presieve.IsSheaf (L.toGrothendieck C)) P
|
Mathlib_CategoryTheory_Sites_Coverage
|
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K L : Coverage C
P : Cᵒᵖ ⥤ Type w
h : (∀ {X : C}, ∀ R ∈ covering K X, IsSheafFor P R) ∧ ∀ {X : C}, ∀ R ∈ covering L X, IsSheafFor P R
⊢ ∀ {X : C}, ∀ R ∈ covering (K ⊔ L) X, IsSheafFor P R
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
· obtain ⟨T, hT⟩ := y.pullback f S hy
exact ⟨T, Or.inr hT.1, hT.2⟩ }
toPartialOrder := inferInstance
le_sup_left _ _ _ := Set.subset_union_left _ _
le_sup_right _ _ _ := Set.subset_union_right _ _
sup_le _ _ _ hx hy X := Set.union_subset_iff.mpr ⟨hx X, hy X⟩
@[simp]
lemma sup_covering (x y : Coverage C) (B : C) :
(x ⊔ y).covering B = x.covering B ∪ y.covering B :=
rfl
end Coverage
open Coverage
namespace Presieve
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← op_comp, h₁, h₂]
simpa only [Sieve.pullback_apply, Category.assoc] using hgg
let y : ∀ ⦃Z : C⦄ (g : Z ⟶ Y),
((S.pullback (g ≫ f)).pullback (𝟙 _)).arrows.FamilyOfElements P :=
fun Z g ZZ gg hgg => x (gg ≫ g) (by simpa using hgg)
have hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), (y g).Compatible := by
intro Z g Y₁ Y₂ ZZ g₁ g₂ f₁ f₂ h₁ h₂ h
rw [hx]
rw [reassoc_of% h]
choose z hz using fun ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : R.pullback f g) =>
H2' hg (𝟙 _) (y g) (hy g)
let q : (R.pullback f).arrows.FamilyOfElements P := fun Z g hg => z hg
have hq : q.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
apply (H2 h₁ g₁).ext
intro ZZ gg hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← P.map_comp, ← op_comp, ← op_comp, hz, hz]
· dsimp; congr 1; simp only [Category.assoc, h]
· simpa [reassoc_of% h] using hgg
· simpa using hgg
obtain ⟨t, ht⟩ := H1' f q hq
refine ⟨t, fun Z g hg => ?_⟩
refine (H1 (g ≫ f)).ext (fun ZZ gg hgg => ?_)
rw [← types_comp_apply _ (P.map gg.op), ← P.map_comp, ← op_comp, ht]
swap; simpa using hgg
refine (H2 hgg (𝟙 _)).ext (fun ZZZ ggg hggg => ?_)
rw [← types_comp_apply _ (P.map ggg.op), ← P.map_comp, ← op_comp, hz]
swap; simpa using hggg
refine (H2 hgg ggg).ext (fun ZZZZ gggg _ => ?_)
rw [← types_comp_apply _ (P.map gggg.op), ← P.map_comp, ← op_comp]
apply hx
simp
/--
A presheaf is a sheaf for the Grothendieck topology generated by a union of coverages iff it is a
sheaf for the Grothendieck topology generated by each coverage separately.
-/
theorem isSheaf_sup (K L : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
(Presieve.IsSheaf ((K ⊔ L).toGrothendieck C)) P ↔
(Presieve.IsSheaf (K.toGrothendieck C)) P ∧ (Presieve.IsSheaf (L.toGrothendieck C)) P := by
refine ⟨fun h ↦ ⟨Presieve.isSheaf_of_le _ ((gi C).gc.monotone_l le_sup_left) h,
Presieve.isSheaf_of_le _ ((gi C).gc.monotone_l le_sup_right) h⟩, fun h ↦ ?_⟩
rw [isSheaf_coverage, isSheaf_coverage] at h
rw [isSheaf_coverage]
|
intro X R hR
|
/--
A presheaf is a sheaf for the Grothendieck topology generated by a union of coverages iff it is a
sheaf for the Grothendieck topology generated by each coverage separately.
-/
theorem isSheaf_sup (K L : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
(Presieve.IsSheaf ((K ⊔ L).toGrothendieck C)) P ↔
(Presieve.IsSheaf (K.toGrothendieck C)) P ∧ (Presieve.IsSheaf (L.toGrothendieck C)) P := by
refine ⟨fun h ↦ ⟨Presieve.isSheaf_of_le _ ((gi C).gc.monotone_l le_sup_left) h,
Presieve.isSheaf_of_le _ ((gi C).gc.monotone_l le_sup_right) h⟩, fun h ↦ ?_⟩
rw [isSheaf_coverage, isSheaf_coverage] at h
rw [isSheaf_coverage]
|
Mathlib.CategoryTheory.Sites.Coverage.388_0.qkZFgqEgDC2P633
|
/--
A presheaf is a sheaf for the Grothendieck topology generated by a union of coverages iff it is a
sheaf for the Grothendieck topology generated by each coverage separately.
-/
theorem isSheaf_sup (K L : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
(Presieve.IsSheaf ((K ⊔ L).toGrothendieck C)) P ↔
(Presieve.IsSheaf (K.toGrothendieck C)) P ∧ (Presieve.IsSheaf (L.toGrothendieck C)) P
|
Mathlib_CategoryTheory_Sites_Coverage
|
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K L : Coverage C
P : Cᵒᵖ ⥤ Type w
h : (∀ {X : C}, ∀ R ∈ covering K X, IsSheafFor P R) ∧ ∀ {X : C}, ∀ R ∈ covering L X, IsSheafFor P R
X : C
R : Presieve X
hR : R ∈ covering (K ⊔ L) X
⊢ IsSheafFor P R
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
· obtain ⟨T, hT⟩ := y.pullback f S hy
exact ⟨T, Or.inr hT.1, hT.2⟩ }
toPartialOrder := inferInstance
le_sup_left _ _ _ := Set.subset_union_left _ _
le_sup_right _ _ _ := Set.subset_union_right _ _
sup_le _ _ _ hx hy X := Set.union_subset_iff.mpr ⟨hx X, hy X⟩
@[simp]
lemma sup_covering (x y : Coverage C) (B : C) :
(x ⊔ y).covering B = x.covering B ∪ y.covering B :=
rfl
end Coverage
open Coverage
namespace Presieve
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← op_comp, h₁, h₂]
simpa only [Sieve.pullback_apply, Category.assoc] using hgg
let y : ∀ ⦃Z : C⦄ (g : Z ⟶ Y),
((S.pullback (g ≫ f)).pullback (𝟙 _)).arrows.FamilyOfElements P :=
fun Z g ZZ gg hgg => x (gg ≫ g) (by simpa using hgg)
have hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), (y g).Compatible := by
intro Z g Y₁ Y₂ ZZ g₁ g₂ f₁ f₂ h₁ h₂ h
rw [hx]
rw [reassoc_of% h]
choose z hz using fun ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : R.pullback f g) =>
H2' hg (𝟙 _) (y g) (hy g)
let q : (R.pullback f).arrows.FamilyOfElements P := fun Z g hg => z hg
have hq : q.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
apply (H2 h₁ g₁).ext
intro ZZ gg hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← P.map_comp, ← op_comp, ← op_comp, hz, hz]
· dsimp; congr 1; simp only [Category.assoc, h]
· simpa [reassoc_of% h] using hgg
· simpa using hgg
obtain ⟨t, ht⟩ := H1' f q hq
refine ⟨t, fun Z g hg => ?_⟩
refine (H1 (g ≫ f)).ext (fun ZZ gg hgg => ?_)
rw [← types_comp_apply _ (P.map gg.op), ← P.map_comp, ← op_comp, ht]
swap; simpa using hgg
refine (H2 hgg (𝟙 _)).ext (fun ZZZ ggg hggg => ?_)
rw [← types_comp_apply _ (P.map ggg.op), ← P.map_comp, ← op_comp, hz]
swap; simpa using hggg
refine (H2 hgg ggg).ext (fun ZZZZ gggg _ => ?_)
rw [← types_comp_apply _ (P.map gggg.op), ← P.map_comp, ← op_comp]
apply hx
simp
/--
A presheaf is a sheaf for the Grothendieck topology generated by a union of coverages iff it is a
sheaf for the Grothendieck topology generated by each coverage separately.
-/
theorem isSheaf_sup (K L : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
(Presieve.IsSheaf ((K ⊔ L).toGrothendieck C)) P ↔
(Presieve.IsSheaf (K.toGrothendieck C)) P ∧ (Presieve.IsSheaf (L.toGrothendieck C)) P := by
refine ⟨fun h ↦ ⟨Presieve.isSheaf_of_le _ ((gi C).gc.monotone_l le_sup_left) h,
Presieve.isSheaf_of_le _ ((gi C).gc.monotone_l le_sup_right) h⟩, fun h ↦ ?_⟩
rw [isSheaf_coverage, isSheaf_coverage] at h
rw [isSheaf_coverage]
intro X R hR
|
cases' hR with hR hR
|
/--
A presheaf is a sheaf for the Grothendieck topology generated by a union of coverages iff it is a
sheaf for the Grothendieck topology generated by each coverage separately.
-/
theorem isSheaf_sup (K L : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
(Presieve.IsSheaf ((K ⊔ L).toGrothendieck C)) P ↔
(Presieve.IsSheaf (K.toGrothendieck C)) P ∧ (Presieve.IsSheaf (L.toGrothendieck C)) P := by
refine ⟨fun h ↦ ⟨Presieve.isSheaf_of_le _ ((gi C).gc.monotone_l le_sup_left) h,
Presieve.isSheaf_of_le _ ((gi C).gc.monotone_l le_sup_right) h⟩, fun h ↦ ?_⟩
rw [isSheaf_coverage, isSheaf_coverage] at h
rw [isSheaf_coverage]
intro X R hR
|
Mathlib.CategoryTheory.Sites.Coverage.388_0.qkZFgqEgDC2P633
|
/--
A presheaf is a sheaf for the Grothendieck topology generated by a union of coverages iff it is a
sheaf for the Grothendieck topology generated by each coverage separately.
-/
theorem isSheaf_sup (K L : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
(Presieve.IsSheaf ((K ⊔ L).toGrothendieck C)) P ↔
(Presieve.IsSheaf (K.toGrothendieck C)) P ∧ (Presieve.IsSheaf (L.toGrothendieck C)) P
|
Mathlib_CategoryTheory_Sites_Coverage
|
case inl
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K L : Coverage C
P : Cᵒᵖ ⥤ Type w
h : (∀ {X : C}, ∀ R ∈ covering K X, IsSheafFor P R) ∧ ∀ {X : C}, ∀ R ∈ covering L X, IsSheafFor P R
X : C
R : Presieve X
hR : R ∈ covering K X
⊢ IsSheafFor P R
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
· obtain ⟨T, hT⟩ := y.pullback f S hy
exact ⟨T, Or.inr hT.1, hT.2⟩ }
toPartialOrder := inferInstance
le_sup_left _ _ _ := Set.subset_union_left _ _
le_sup_right _ _ _ := Set.subset_union_right _ _
sup_le _ _ _ hx hy X := Set.union_subset_iff.mpr ⟨hx X, hy X⟩
@[simp]
lemma sup_covering (x y : Coverage C) (B : C) :
(x ⊔ y).covering B = x.covering B ∪ y.covering B :=
rfl
end Coverage
open Coverage
namespace Presieve
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← op_comp, h₁, h₂]
simpa only [Sieve.pullback_apply, Category.assoc] using hgg
let y : ∀ ⦃Z : C⦄ (g : Z ⟶ Y),
((S.pullback (g ≫ f)).pullback (𝟙 _)).arrows.FamilyOfElements P :=
fun Z g ZZ gg hgg => x (gg ≫ g) (by simpa using hgg)
have hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), (y g).Compatible := by
intro Z g Y₁ Y₂ ZZ g₁ g₂ f₁ f₂ h₁ h₂ h
rw [hx]
rw [reassoc_of% h]
choose z hz using fun ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : R.pullback f g) =>
H2' hg (𝟙 _) (y g) (hy g)
let q : (R.pullback f).arrows.FamilyOfElements P := fun Z g hg => z hg
have hq : q.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
apply (H2 h₁ g₁).ext
intro ZZ gg hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← P.map_comp, ← op_comp, ← op_comp, hz, hz]
· dsimp; congr 1; simp only [Category.assoc, h]
· simpa [reassoc_of% h] using hgg
· simpa using hgg
obtain ⟨t, ht⟩ := H1' f q hq
refine ⟨t, fun Z g hg => ?_⟩
refine (H1 (g ≫ f)).ext (fun ZZ gg hgg => ?_)
rw [← types_comp_apply _ (P.map gg.op), ← P.map_comp, ← op_comp, ht]
swap; simpa using hgg
refine (H2 hgg (𝟙 _)).ext (fun ZZZ ggg hggg => ?_)
rw [← types_comp_apply _ (P.map ggg.op), ← P.map_comp, ← op_comp, hz]
swap; simpa using hggg
refine (H2 hgg ggg).ext (fun ZZZZ gggg _ => ?_)
rw [← types_comp_apply _ (P.map gggg.op), ← P.map_comp, ← op_comp]
apply hx
simp
/--
A presheaf is a sheaf for the Grothendieck topology generated by a union of coverages iff it is a
sheaf for the Grothendieck topology generated by each coverage separately.
-/
theorem isSheaf_sup (K L : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
(Presieve.IsSheaf ((K ⊔ L).toGrothendieck C)) P ↔
(Presieve.IsSheaf (K.toGrothendieck C)) P ∧ (Presieve.IsSheaf (L.toGrothendieck C)) P := by
refine ⟨fun h ↦ ⟨Presieve.isSheaf_of_le _ ((gi C).gc.monotone_l le_sup_left) h,
Presieve.isSheaf_of_le _ ((gi C).gc.monotone_l le_sup_right) h⟩, fun h ↦ ?_⟩
rw [isSheaf_coverage, isSheaf_coverage] at h
rw [isSheaf_coverage]
intro X R hR
cases' hR with hR hR
·
|
exact h.1 R hR
|
/--
A presheaf is a sheaf for the Grothendieck topology generated by a union of coverages iff it is a
sheaf for the Grothendieck topology generated by each coverage separately.
-/
theorem isSheaf_sup (K L : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
(Presieve.IsSheaf ((K ⊔ L).toGrothendieck C)) P ↔
(Presieve.IsSheaf (K.toGrothendieck C)) P ∧ (Presieve.IsSheaf (L.toGrothendieck C)) P := by
refine ⟨fun h ↦ ⟨Presieve.isSheaf_of_le _ ((gi C).gc.monotone_l le_sup_left) h,
Presieve.isSheaf_of_le _ ((gi C).gc.monotone_l le_sup_right) h⟩, fun h ↦ ?_⟩
rw [isSheaf_coverage, isSheaf_coverage] at h
rw [isSheaf_coverage]
intro X R hR
cases' hR with hR hR
·
|
Mathlib.CategoryTheory.Sites.Coverage.388_0.qkZFgqEgDC2P633
|
/--
A presheaf is a sheaf for the Grothendieck topology generated by a union of coverages iff it is a
sheaf for the Grothendieck topology generated by each coverage separately.
-/
theorem isSheaf_sup (K L : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
(Presieve.IsSheaf ((K ⊔ L).toGrothendieck C)) P ↔
(Presieve.IsSheaf (K.toGrothendieck C)) P ∧ (Presieve.IsSheaf (L.toGrothendieck C)) P
|
Mathlib_CategoryTheory_Sites_Coverage
|
case inr
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K L : Coverage C
P : Cᵒᵖ ⥤ Type w
h : (∀ {X : C}, ∀ R ∈ covering K X, IsSheafFor P R) ∧ ∀ {X : C}, ∀ R ∈ covering L X, IsSheafFor P R
X : C
R : Presieve X
hR : R ∈ covering L X
⊢ IsSheafFor P R
|
/-
Copyright (c) 2023 Adam Topaz. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Adam Topaz
-/
import Mathlib.CategoryTheory.Sites.SheafOfTypes
/-!
# Coverages
A coverage `K` on a category `C` is a set of presieves associated to every object `X : C`,
called "covering presieves".
This collection must satisfy a certain "pullback compatibility" condition, saying that
whenever `S` is a covering presieve on `X` and `f : Y ⟶ X` is a morphism, then there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`.
The main difference between a coverage and a Grothendieck pretopology is that we *do not*
require `C` to have pullbacks.
This is useful, for example, when we want to consider the Grothendieck topology on the category
of extremally disconnected sets in the context of condensed mathematics.
A more concrete example: If `ℬ` is a basis for a topology on a type `X` (in the sense of
`TopologicalSpace.IsTopologicalBasis`) then it naturally induces a coverage on `Opens X`
whose associated Grothendieck topology is the one induced by the topology
on `X` generated by `ℬ`. (Project: Formalize this!)
## Main Definitions and Results:
All definitions are in the `CategoryTheory` namespace.
- `Coverage C`: The type of coverages on `C`.
- `Coverage.ofGrothendieck C`: A function which associates a coverage to any Grothendieck topology.
- `Coverage.toGrothendieck C`: A function which associates a Grothendieck topology to any coverage.
- `Coverage.gi`: The two functions above form a Galois insertion.
- `Presieve.isSheaf_coverage`: Given `K : Coverage C` with associated
Grothendieck topology `J`, a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if
it is a sheaf for `J`.
# References
We don't follow any particular reference, but the arguments can probably be distilled from
the following sources:
- [Elephant]: *Sketches of an Elephant*, P. T. Johnstone: C2.1.
- [nLab, *Coverage*](https://ncatlab.org/nlab/show/coverage)
-/
set_option autoImplicit true
namespace CategoryTheory
variable {C : Type _} [Category C]
namespace Presieve
/--
Given a morphism `f : Y ⟶ X`, a presieve `S` on `Y` and presieve `T` on `X`,
we say that *`S` factors through `T` along `f`*, written `S.FactorsThruAlong T f`,
provided that for any morphism `g : Z ⟶ Y` in `S`, there exists some
morphism `e : W ⟶ X` in `T` and some morphism `i : Z ⟶ W` such that the obvious
square commutes: `i ≫ e = g ≫ f`.
This is used in the definition of a coverage.
-/
def FactorsThruAlong {X Y : C} (S : Presieve Y) (T : Presieve X) (f : Y ⟶ X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g ≫ f
/--
Given `S T : Presieve X`, we say that `S` factors through `T` if any morphism in `S`
factors through some morphism in `T`.
The lemma `Presieve.isSheafFor_of_factorsThru` gives a *sufficient* condition for a
presheaf to be a sheaf for a presieve `T`, in terms of `S.FactorsThru T`, provided
that the presheaf is a sheaf for `S`.
-/
def FactorsThru {X : C} (S T : Presieve X) : Prop :=
∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄, S g →
∃ (W : C) (i : Z ⟶ W) (e : W ⟶ X), T e ∧ i ≫ e = g
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
simp [FactorsThruAlong, FactorsThru]
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T :=
fun Y g hg => ⟨Y, 𝟙 _, g, h _ hg, by simp⟩
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
obtain ⟨W, i, e, h1, rfl⟩ := h hf
exact T.downward_closed h1 _
lemma factorsThru_top {X : C} (S : Presieve X) : S.FactorsThru ⊤ :=
factorsThru_of_le _ _ le_top
lemma isSheafFor_of_factorsThru
{X : C} {S T : Presieve X}
(P : Cᵒᵖ ⥤ Type w)
(H : S.FactorsThru T) (hS : S.IsSheafFor P)
(h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ (R : Presieve Y),
R.IsSeparatedFor P ∧ R.FactorsThruAlong S f):
T.IsSheafFor P := by
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose W i e h1 h2 using H
refine ⟨?_, fun x hx => ?_⟩
· intro x y₁ y₂ h₁ h₂
refine hS.1.ext (fun Y g hg => ?_)
simp only [← h2 hg, op_comp, P.map_comp, types_comp_apply, h₁ _ (h1 _ ), h₂ _ (h1 _)]
let y : S.FamilyOfElements P := fun Y g hg => P.map (i _).op (x (e hg) (h1 _))
have hy : y.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
rw [← types_comp_apply (P.map (i h₁).op) (P.map g₁.op),
← types_comp_apply (P.map (i h₂).op) (P.map g₂.op),
← P.map_comp, ← op_comp, ← P.map_comp, ← op_comp]
apply hx
simp only [h2, h, Category.assoc]
let ⟨_, h2'⟩ := hS
obtain ⟨z, hz⟩ := h2' y hy
refine ⟨z, fun Y g hg => ?_⟩
obtain ⟨R, hR1, hR2⟩ := h hg
choose WW ii ee hh1 hh2 using hR2
refine hR1.ext (fun Q t ht => ?_)
rw [← types_comp_apply (P.map g.op) (P.map t.op), ← P.map_comp, ← op_comp, ← hh2 ht,
op_comp, P.map_comp, types_comp_apply, hz _ (hh1 _),
← types_comp_apply _ (P.map (ii ht).op), ← P.map_comp, ← op_comp]
apply hx
simp only [Category.assoc, h2, hh2]
end Presieve
variable (C) in
/--
The type `Coverage C` of coverages on `C`.
A coverage is a collection of *covering* presieves on every object `X : C`,
which satisfies a *pullback compatibility* condition.
Explicitly, this condition says that whenever `S` is a covering presieve for `X` and
`f : Y ⟶ X` is a morphism, then there exists some covering presieve `T` for `Y`
such that `T` factors through `S` along `f`.
-/
@[ext]
structure Coverage where
/-- The collection of covering presieves for an object `X`. -/
covering : ∀ (X : C), Set (Presieve X)
/-- Given any covering sieve `S` on `X` and a morphism `f : Y ⟶ X`, there exists
some covering sieve `T` on `Y` such that `T` factors through `S` along `f`. -/
pullback : ∀ ⦃X Y : C⦄ (f : Y ⟶ X) (S : Presieve X) (_ : S ∈ covering X),
∃ (T : Presieve Y), T ∈ covering Y ∧ T.FactorsThruAlong S f
namespace Coverage
instance : CoeFun (Coverage C) (fun _ => (X : C) → Set (Presieve X)) where
coe := covering
variable (C) in
/--
Associate a coverage to any Grothendieck topology.
If `J` is a Grothendieck topology, and `K` is the associated coverage, then a presieve
`S` is a covering presieve for `K` if and only if the sieve that it generates is a
covering sieve for `J`.
-/
def ofGrothendieck (J : GrothendieckTopology C) : Coverage C where
covering X := { S | Sieve.generate S ∈ J X }
pullback := by
intro X Y f S (hS : Sieve.generate S ∈ J X)
refine ⟨(Sieve.generate S).pullback f, ?_, fun Z g h => h⟩
dsimp
rw [Sieve.generate_sieve]
exact J.pullback_stable _ hS
lemma ofGrothendieck_iff {X : C} {S : Presieve X} (J : GrothendieckTopology C) :
S ∈ ofGrothendieck _ J X ↔ Sieve.generate S ∈ J X := Iff.rfl
/--
An auxiliary definition used to define the Grothendieck topology associated to a
coverage. See `Coverage.toGrothendieck`.
-/
inductive saturate (K : Coverage C) : (X : C) → Sieve X → Prop where
| of (X : C) (S : Presieve X) (hS : S ∈ K X) : saturate K X (Sieve.generate S)
| top (X : C) : saturate K X ⊤
| transitive (X : C) (R S : Sieve X) :
saturate K X R →
(∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R f → saturate K Y (S.pullback f)) →
saturate K X S
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
ext Z g
simp only [Sieve.pullback_apply, Sieve.top_apply, iff_true]
apply h
apply S.downward_closed
exact hf
lemma saturate_of_superset (K : Coverage C) {X : C} {S T : Sieve X} (h : S ≤ T)
(hS : saturate K X S) : saturate K X T := by
apply saturate.transitive _ _ _ hS
intro Y g hg
rw [eq_top_pullback (h := h)]
· apply saturate.top
· assumption
variable (C) in
/--
The Grothendieck topology associated to a coverage `K`.
It is defined *inductively* as follows:
1. If `S` is a covering presieve for `K`, then the sieve generated by `S` is a covering
sieve for the associated Grothendieck topology.
2. The top sieves are in the associated Grothendieck topology.
3. Add all sieves required by the *local character* axiom of a Grothendieck topology.
The pullback compatibility condition for a coverage ensures that the
associated Grothendieck topology is pullback stable, and so an additional constructor
in the inductive construction is not needed.
-/
def toGrothendieck (K : Coverage C) : GrothendieckTopology C where
sieves := saturate K
top_mem' := .top
pullback_stable' := by
intro X Y S f hS
induction hS generalizing Y with
| of X S hS =>
obtain ⟨R,hR1,hR2⟩ := K.pullback f S hS
suffices Sieve.generate R ≤ (Sieve.generate S).pullback f from
saturate_of_superset _ this (saturate.of _ _ hR1)
rintro Z g ⟨W, i, e, h1, h2⟩
obtain ⟨WW, ii, ee, hh1, hh2⟩ := hR2 h1
refine ⟨WW, i ≫ ii, ee, hh1, ?_⟩
simp only [hh2, reassoc_of% h2, Category.assoc]
| top X => apply saturate.top
| transitive X R S _ hS H1 _ =>
apply saturate.transitive
apply H1 f
intro Z g hg
rw [← Sieve.pullback_comp]
exact hS hg
transitive' X S hS R hR := .transitive _ _ _ hS hR
instance : PartialOrder (Coverage C) where
le A B := A.covering ≤ B.covering
le_refl A X := le_refl _
le_trans A B C h1 h2 X := le_trans (h1 X) (h2 X)
le_antisymm A B h1 h2 := Coverage.ext A B <| funext <|
fun X => le_antisymm (h1 X) (h2 X)
variable (C) in
/--
The two constructions `Coverage.toGrothendieck` and `Coverage.ofGrothendieck` form
a Galois insertion.
-/
def gi : GaloisInsertion (toGrothendieck C) (ofGrothendieck C) where
choice K _ := toGrothendieck _ K
choice_eq := fun _ _ => rfl
le_l_u J X S hS := by
rw [← Sieve.generate_sieve S]
apply saturate.of
dsimp [ofGrothendieck]
rwa [Sieve.generate_sieve S]
gc K J := by
constructor
· intro H X S hS
exact H _ <| saturate.of _ _ hS
· intro H X S hS
induction hS with
| of X S hS => exact H _ hS
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
/--
An alternative characterization of the Grothendieck topology associated to a coverage `K`:
it is the infimum of all Grothendieck topologies whose associated coverage contains `K`.
-/
theorem toGrothendieck_eq_sInf (K : Coverage C) : toGrothendieck _ K =
sInf {J | K ≤ ofGrothendieck _ J } := by
apply le_antisymm
· apply le_sInf; intro J hJ
intro X S hS
induction hS with
| of X S hS => apply hJ; assumption
| top => apply J.top_mem
| transitive X R S _ _ H1 H2 => exact J.transitive H1 _ H2
· apply sInf_le
intro X S hS
apply saturate.of _ _ hS
instance : SemilatticeSup (Coverage C) where
sup x y :=
{ covering := fun B ↦ x.covering B ∪ y.covering B
pullback := by
rintro X Y f S (hx | hy)
· obtain ⟨T, hT⟩ := x.pullback f S hx
exact ⟨T, Or.inl hT.1, hT.2⟩
· obtain ⟨T, hT⟩ := y.pullback f S hy
exact ⟨T, Or.inr hT.1, hT.2⟩ }
toPartialOrder := inferInstance
le_sup_left _ _ _ := Set.subset_union_left _ _
le_sup_right _ _ _ := Set.subset_union_right _ _
sup_le _ _ _ hx hy X := Set.union_subset_iff.mpr ⟨hx X, hy X⟩
@[simp]
lemma sup_covering (x y : Coverage C) (B : C) :
(x ⊔ y).covering B = x.covering B ∪ y.covering B :=
rfl
end Coverage
open Coverage
namespace Presieve
/--
The main theorem of this file: Given a coverage `K` on `C`,
a `Type*`-valued presheaf on `C` is a sheaf for `K` if and only if it is a sheaf for
the associated Grothendieck topology.
-/
theorem isSheaf_coverage (K : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
Presieve.IsSheaf (toGrothendieck _ K) P ↔
(∀ {X : C} (R : Presieve X), R ∈ K X → Presieve.IsSheafFor P R) := by
constructor
· intro H X R hR
rw [Presieve.isSheafFor_iff_generate]
apply H _ <| saturate.of _ _ hR
· intro H X S hS
-- This is the key point of the proof:
-- We must generalize the induction in the correct way.
suffices ∀ ⦃Y : C⦄ (f : Y ⟶ X), Presieve.IsSheafFor P (S.pullback f).arrows by
simpa using this (f := 𝟙 _)
induction hS with
| of X S hS =>
intro Y f
obtain ⟨T, hT1, hT2⟩ := K.pullback f S hS
apply Presieve.isSheafFor_of_factorsThru (S := T)
· intro Z g hg
obtain ⟨W, i, e, h1, h2⟩ := hT2 hg
exact ⟨Z, 𝟙 _, g, ⟨W, i, e, h1, h2⟩, by simp⟩
· apply H; assumption
· intro Z g _
obtain ⟨R, hR1, hR2⟩ := K.pullback g _ hT1
refine ⟨R, (H _ hR1).isSeparatedFor, hR2⟩
| top => intros; simpa using Presieve.isSheafFor_top_sieve _
| transitive X R S _ _ H1 H2 =>
intro Y f
simp only [← Presieve.isSeparatedFor_and_exists_isAmalgamation_iff_isSheafFor] at *
choose H1 H1' using H1
choose H2 H2' using H2
refine ⟨?_, fun x hx => ?_⟩
· intro x t₁ t₂ h₁ h₂
refine (H1 f).ext (fun Z g hg => ?_)
refine (H2 hg (𝟙 _)).ext (fun ZZ gg hgg => ?_)
simp only [Sieve.pullback_id, Sieve.pullback_apply] at hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← op_comp, h₁, h₂]
simpa only [Sieve.pullback_apply, Category.assoc] using hgg
let y : ∀ ⦃Z : C⦄ (g : Z ⟶ Y),
((S.pullback (g ≫ f)).pullback (𝟙 _)).arrows.FamilyOfElements P :=
fun Z g ZZ gg hgg => x (gg ≫ g) (by simpa using hgg)
have hy : ∀ ⦃Z : C⦄ (g : Z ⟶ Y), (y g).Compatible := by
intro Z g Y₁ Y₂ ZZ g₁ g₂ f₁ f₂ h₁ h₂ h
rw [hx]
rw [reassoc_of% h]
choose z hz using fun ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (hg : R.pullback f g) =>
H2' hg (𝟙 _) (y g) (hy g)
let q : (R.pullback f).arrows.FamilyOfElements P := fun Z g hg => z hg
have hq : q.Compatible := by
intro Y₁ Y₂ Z g₁ g₂ f₁ f₂ h₁ h₂ h
apply (H2 h₁ g₁).ext
intro ZZ gg hgg
simp only [← types_comp_apply]
rw [← P.map_comp, ← P.map_comp, ← op_comp, ← op_comp, hz, hz]
· dsimp; congr 1; simp only [Category.assoc, h]
· simpa [reassoc_of% h] using hgg
· simpa using hgg
obtain ⟨t, ht⟩ := H1' f q hq
refine ⟨t, fun Z g hg => ?_⟩
refine (H1 (g ≫ f)).ext (fun ZZ gg hgg => ?_)
rw [← types_comp_apply _ (P.map gg.op), ← P.map_comp, ← op_comp, ht]
swap; simpa using hgg
refine (H2 hgg (𝟙 _)).ext (fun ZZZ ggg hggg => ?_)
rw [← types_comp_apply _ (P.map ggg.op), ← P.map_comp, ← op_comp, hz]
swap; simpa using hggg
refine (H2 hgg ggg).ext (fun ZZZZ gggg _ => ?_)
rw [← types_comp_apply _ (P.map gggg.op), ← P.map_comp, ← op_comp]
apply hx
simp
/--
A presheaf is a sheaf for the Grothendieck topology generated by a union of coverages iff it is a
sheaf for the Grothendieck topology generated by each coverage separately.
-/
theorem isSheaf_sup (K L : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
(Presieve.IsSheaf ((K ⊔ L).toGrothendieck C)) P ↔
(Presieve.IsSheaf (K.toGrothendieck C)) P ∧ (Presieve.IsSheaf (L.toGrothendieck C)) P := by
refine ⟨fun h ↦ ⟨Presieve.isSheaf_of_le _ ((gi C).gc.monotone_l le_sup_left) h,
Presieve.isSheaf_of_le _ ((gi C).gc.monotone_l le_sup_right) h⟩, fun h ↦ ?_⟩
rw [isSheaf_coverage, isSheaf_coverage] at h
rw [isSheaf_coverage]
intro X R hR
cases' hR with hR hR
· exact h.1 R hR
·
|
exact h.2 R hR
|
/--
A presheaf is a sheaf for the Grothendieck topology generated by a union of coverages iff it is a
sheaf for the Grothendieck topology generated by each coverage separately.
-/
theorem isSheaf_sup (K L : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
(Presieve.IsSheaf ((K ⊔ L).toGrothendieck C)) P ↔
(Presieve.IsSheaf (K.toGrothendieck C)) P ∧ (Presieve.IsSheaf (L.toGrothendieck C)) P := by
refine ⟨fun h ↦ ⟨Presieve.isSheaf_of_le _ ((gi C).gc.monotone_l le_sup_left) h,
Presieve.isSheaf_of_le _ ((gi C).gc.monotone_l le_sup_right) h⟩, fun h ↦ ?_⟩
rw [isSheaf_coverage, isSheaf_coverage] at h
rw [isSheaf_coverage]
intro X R hR
cases' hR with hR hR
· exact h.1 R hR
·
|
Mathlib.CategoryTheory.Sites.Coverage.388_0.qkZFgqEgDC2P633
|
/--
A presheaf is a sheaf for the Grothendieck topology generated by a union of coverages iff it is a
sheaf for the Grothendieck topology generated by each coverage separately.
-/
theorem isSheaf_sup (K L : Coverage C) (P : Cᵒᵖ ⥤ Type w) :
(Presieve.IsSheaf ((K ⊔ L).toGrothendieck C)) P ↔
(Presieve.IsSheaf (K.toGrothendieck C)) P ∧ (Presieve.IsSheaf (L.toGrothendieck C)) P
|
Mathlib_CategoryTheory_Sites_Coverage
|
f : ℕ →. ℕ
hf : Partrec f
⊢ Partrec₂ fun a m =>
Part.map (fun x => x + m) (Nat.rfind fun n => (fun m => decide (m = 0)) <$> f (Nat.pair a (n + m)))
|
/-
Copyright (c) 2018 Mario Carneiro. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Mario Carneiro
-/
import Mathlib.Computability.Partrec
#align_import computability.partrec_code from "leanprover-community/mathlib"@"6155d4351090a6fad236e3d2e4e0e4e7342668e8"
/-!
# Gödel Numbering for Partial Recursive Functions.
This file defines `Nat.Partrec.Code`, an inductive datatype describing code for partial
recursive functions on ℕ. It defines an encoding for these codes, and proves that the constructors
are primitive recursive with respect to the encoding.
It also defines the evaluation of these codes as partial functions using `PFun`, and proves that a
function is partially recursive (as defined by `Nat.Partrec`) if and only if it is the evaluation
of some code.
## Main Definitions
* `Nat.Partrec.Code`: Inductive datatype for partial recursive codes.
* `Nat.Partrec.Code.encodeCode`: A (computable) encoding of codes as natural numbers.
* `Nat.Partrec.Code.ofNatCode`: The inverse of this encoding.
* `Nat.Partrec.Code.eval`: The interpretation of a `Nat.Partrec.Code` as a partial function.
## Main Results
* `Nat.Partrec.Code.rec_prim`: Recursion on `Nat.Partrec.Code` is primitive recursive.
* `Nat.Partrec.Code.rec_computable`: Recursion on `Nat.Partrec.Code` is computable.
* `Nat.Partrec.Code.smn`: The $S_n^m$ theorem.
* `Nat.Partrec.Code.exists_code`: Partial recursiveness is equivalent to being the eval of a code.
* `Nat.Partrec.Code.evaln_prim`: `evaln` is primitive recursive.
* `Nat.Partrec.Code.fixed_point`: Roger's fixed point theorem.
## References
* [Mario Carneiro, *Formalizing computability theory via partial recursive functions*][carneiro2019]
-/
open Encodable Denumerable Primrec
namespace Nat.Partrec
open Nat (pair)
theorem rfind' {f} (hf : Nat.Partrec f) :
Nat.Partrec
(Nat.unpaired fun a m =>
(Nat.rfind fun n => (fun m => m = 0) <$> f (Nat.pair a (n + m))).map (· + m)) :=
Partrec₂.unpaired'.2 <| by
|
refine'
Partrec.map
((@Partrec₂.unpaired' fun a b : ℕ =>
Nat.rfind fun n => (fun m => m = 0) <$> f (Nat.pair a (n + b))).1
_)
(Primrec.nat_add.comp Primrec.snd <| Primrec.snd.comp Primrec.fst).to_comp.to₂
|
theorem rfind' {f} (hf : Nat.Partrec f) :
Nat.Partrec
(Nat.unpaired fun a m =>
(Nat.rfind fun n => (fun m => m = 0) <$> f (Nat.pair a (n + m))).map (· + m)) :=
Partrec₂.unpaired'.2 <| by
|
Mathlib.Computability.PartrecCode.50_0.A3c3Aev6SyIRjCJ
|
theorem rfind' {f} (hf : Nat.Partrec f) :
Nat.Partrec
(Nat.unpaired fun a m =>
(Nat.rfind fun n => (fun m => m = 0) <$> f (Nat.pair a (n + m))).map (· + m))
|
Mathlib_Computability_PartrecCode
|
f : ℕ →. ℕ
hf : Partrec f
⊢ Partrec (unpaired fun a b => Nat.rfind fun n => (fun m => decide (m = 0)) <$> f (Nat.pair a (n + b)))
|
/-
Copyright (c) 2018 Mario Carneiro. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Mario Carneiro
-/
import Mathlib.Computability.Partrec
#align_import computability.partrec_code from "leanprover-community/mathlib"@"6155d4351090a6fad236e3d2e4e0e4e7342668e8"
/-!
# Gödel Numbering for Partial Recursive Functions.
This file defines `Nat.Partrec.Code`, an inductive datatype describing code for partial
recursive functions on ℕ. It defines an encoding for these codes, and proves that the constructors
are primitive recursive with respect to the encoding.
It also defines the evaluation of these codes as partial functions using `PFun`, and proves that a
function is partially recursive (as defined by `Nat.Partrec`) if and only if it is the evaluation
of some code.
## Main Definitions
* `Nat.Partrec.Code`: Inductive datatype for partial recursive codes.
* `Nat.Partrec.Code.encodeCode`: A (computable) encoding of codes as natural numbers.
* `Nat.Partrec.Code.ofNatCode`: The inverse of this encoding.
* `Nat.Partrec.Code.eval`: The interpretation of a `Nat.Partrec.Code` as a partial function.
## Main Results
* `Nat.Partrec.Code.rec_prim`: Recursion on `Nat.Partrec.Code` is primitive recursive.
* `Nat.Partrec.Code.rec_computable`: Recursion on `Nat.Partrec.Code` is computable.
* `Nat.Partrec.Code.smn`: The $S_n^m$ theorem.
* `Nat.Partrec.Code.exists_code`: Partial recursiveness is equivalent to being the eval of a code.
* `Nat.Partrec.Code.evaln_prim`: `evaln` is primitive recursive.
* `Nat.Partrec.Code.fixed_point`: Roger's fixed point theorem.
## References
* [Mario Carneiro, *Formalizing computability theory via partial recursive functions*][carneiro2019]
-/
open Encodable Denumerable Primrec
namespace Nat.Partrec
open Nat (pair)
theorem rfind' {f} (hf : Nat.Partrec f) :
Nat.Partrec
(Nat.unpaired fun a m =>
(Nat.rfind fun n => (fun m => m = 0) <$> f (Nat.pair a (n + m))).map (· + m)) :=
Partrec₂.unpaired'.2 <| by
refine'
Partrec.map
((@Partrec₂.unpaired' fun a b : ℕ =>
Nat.rfind fun n => (fun m => m = 0) <$> f (Nat.pair a (n + b))).1
_)
(Primrec.nat_add.comp Primrec.snd <| Primrec.snd.comp Primrec.fst).to_comp.to₂
|
have : Nat.Partrec (fun a => Nat.rfind (fun n => (fun m => decide (m = 0)) <$>
Nat.unpaired (fun a b => f (Nat.pair (Nat.unpair a).1 (b + (Nat.unpair a).2)))
(Nat.pair a n))) :=
rfind
(Partrec₂.unpaired'.2
((Partrec.nat_iff.2 hf).comp
(Primrec₂.pair.comp (Primrec.fst.comp <| Primrec.unpair.comp Primrec.fst)
(Primrec.nat_add.comp Primrec.snd
(Primrec.snd.comp <| Primrec.unpair.comp Primrec.fst))).to_comp))
|
theorem rfind' {f} (hf : Nat.Partrec f) :
Nat.Partrec
(Nat.unpaired fun a m =>
(Nat.rfind fun n => (fun m => m = 0) <$> f (Nat.pair a (n + m))).map (· + m)) :=
Partrec₂.unpaired'.2 <| by
refine'
Partrec.map
((@Partrec₂.unpaired' fun a b : ℕ =>
Nat.rfind fun n => (fun m => m = 0) <$> f (Nat.pair a (n + b))).1
_)
(Primrec.nat_add.comp Primrec.snd <| Primrec.snd.comp Primrec.fst).to_comp.to₂
|
Mathlib.Computability.PartrecCode.50_0.A3c3Aev6SyIRjCJ
|
theorem rfind' {f} (hf : Nat.Partrec f) :
Nat.Partrec
(Nat.unpaired fun a m =>
(Nat.rfind fun n => (fun m => m = 0) <$> f (Nat.pair a (n + m))).map (· + m))
|
Mathlib_Computability_PartrecCode
|
f : ℕ →. ℕ
hf : Partrec f
this :
Partrec fun a =>
Nat.rfind fun n =>
(fun m => decide (m = 0)) <$> unpaired (fun a b => f (Nat.pair (unpair a).1 (b + (unpair a).2))) (Nat.pair a n)
⊢ Partrec (unpaired fun a b => Nat.rfind fun n => (fun m => decide (m = 0)) <$> f (Nat.pair a (n + b)))
|
/-
Copyright (c) 2018 Mario Carneiro. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Mario Carneiro
-/
import Mathlib.Computability.Partrec
#align_import computability.partrec_code from "leanprover-community/mathlib"@"6155d4351090a6fad236e3d2e4e0e4e7342668e8"
/-!
# Gödel Numbering for Partial Recursive Functions.
This file defines `Nat.Partrec.Code`, an inductive datatype describing code for partial
recursive functions on ℕ. It defines an encoding for these codes, and proves that the constructors
are primitive recursive with respect to the encoding.
It also defines the evaluation of these codes as partial functions using `PFun`, and proves that a
function is partially recursive (as defined by `Nat.Partrec`) if and only if it is the evaluation
of some code.
## Main Definitions
* `Nat.Partrec.Code`: Inductive datatype for partial recursive codes.
* `Nat.Partrec.Code.encodeCode`: A (computable) encoding of codes as natural numbers.
* `Nat.Partrec.Code.ofNatCode`: The inverse of this encoding.
* `Nat.Partrec.Code.eval`: The interpretation of a `Nat.Partrec.Code` as a partial function.
## Main Results
* `Nat.Partrec.Code.rec_prim`: Recursion on `Nat.Partrec.Code` is primitive recursive.
* `Nat.Partrec.Code.rec_computable`: Recursion on `Nat.Partrec.Code` is computable.
* `Nat.Partrec.Code.smn`: The $S_n^m$ theorem.
* `Nat.Partrec.Code.exists_code`: Partial recursiveness is equivalent to being the eval of a code.
* `Nat.Partrec.Code.evaln_prim`: `evaln` is primitive recursive.
* `Nat.Partrec.Code.fixed_point`: Roger's fixed point theorem.
## References
* [Mario Carneiro, *Formalizing computability theory via partial recursive functions*][carneiro2019]
-/
open Encodable Denumerable Primrec
namespace Nat.Partrec
open Nat (pair)
theorem rfind' {f} (hf : Nat.Partrec f) :
Nat.Partrec
(Nat.unpaired fun a m =>
(Nat.rfind fun n => (fun m => m = 0) <$> f (Nat.pair a (n + m))).map (· + m)) :=
Partrec₂.unpaired'.2 <| by
refine'
Partrec.map
((@Partrec₂.unpaired' fun a b : ℕ =>
Nat.rfind fun n => (fun m => m = 0) <$> f (Nat.pair a (n + b))).1
_)
(Primrec.nat_add.comp Primrec.snd <| Primrec.snd.comp Primrec.fst).to_comp.to₂
have : Nat.Partrec (fun a => Nat.rfind (fun n => (fun m => decide (m = 0)) <$>
Nat.unpaired (fun a b => f (Nat.pair (Nat.unpair a).1 (b + (Nat.unpair a).2)))
(Nat.pair a n))) :=
rfind
(Partrec₂.unpaired'.2
((Partrec.nat_iff.2 hf).comp
(Primrec₂.pair.comp (Primrec.fst.comp <| Primrec.unpair.comp Primrec.fst)
(Primrec.nat_add.comp Primrec.snd
(Primrec.snd.comp <| Primrec.unpair.comp Primrec.fst))).to_comp))
|
simp at this
|
theorem rfind' {f} (hf : Nat.Partrec f) :
Nat.Partrec
(Nat.unpaired fun a m =>
(Nat.rfind fun n => (fun m => m = 0) <$> f (Nat.pair a (n + m))).map (· + m)) :=
Partrec₂.unpaired'.2 <| by
refine'
Partrec.map
((@Partrec₂.unpaired' fun a b : ℕ =>
Nat.rfind fun n => (fun m => m = 0) <$> f (Nat.pair a (n + b))).1
_)
(Primrec.nat_add.comp Primrec.snd <| Primrec.snd.comp Primrec.fst).to_comp.to₂
have : Nat.Partrec (fun a => Nat.rfind (fun n => (fun m => decide (m = 0)) <$>
Nat.unpaired (fun a b => f (Nat.pair (Nat.unpair a).1 (b + (Nat.unpair a).2)))
(Nat.pair a n))) :=
rfind
(Partrec₂.unpaired'.2
((Partrec.nat_iff.2 hf).comp
(Primrec₂.pair.comp (Primrec.fst.comp <| Primrec.unpair.comp Primrec.fst)
(Primrec.nat_add.comp Primrec.snd
(Primrec.snd.comp <| Primrec.unpair.comp Primrec.fst))).to_comp))
|
Mathlib.Computability.PartrecCode.50_0.A3c3Aev6SyIRjCJ
|
theorem rfind' {f} (hf : Nat.Partrec f) :
Nat.Partrec
(Nat.unpaired fun a m =>
(Nat.rfind fun n => (fun m => m = 0) <$> f (Nat.pair a (n + m))).map (· + m))
|
Mathlib_Computability_PartrecCode
|
f : ℕ →. ℕ
hf : Partrec f
this :
Partrec fun a => Nat.rfind fun n => Part.map (fun m => decide (m = 0)) (f (Nat.pair (unpair a).1 (n + (unpair a).2)))
⊢ Partrec (unpaired fun a b => Nat.rfind fun n => (fun m => decide (m = 0)) <$> f (Nat.pair a (n + b)))
|
/-
Copyright (c) 2018 Mario Carneiro. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Mario Carneiro
-/
import Mathlib.Computability.Partrec
#align_import computability.partrec_code from "leanprover-community/mathlib"@"6155d4351090a6fad236e3d2e4e0e4e7342668e8"
/-!
# Gödel Numbering for Partial Recursive Functions.
This file defines `Nat.Partrec.Code`, an inductive datatype describing code for partial
recursive functions on ℕ. It defines an encoding for these codes, and proves that the constructors
are primitive recursive with respect to the encoding.
It also defines the evaluation of these codes as partial functions using `PFun`, and proves that a
function is partially recursive (as defined by `Nat.Partrec`) if and only if it is the evaluation
of some code.
## Main Definitions
* `Nat.Partrec.Code`: Inductive datatype for partial recursive codes.
* `Nat.Partrec.Code.encodeCode`: A (computable) encoding of codes as natural numbers.
* `Nat.Partrec.Code.ofNatCode`: The inverse of this encoding.
* `Nat.Partrec.Code.eval`: The interpretation of a `Nat.Partrec.Code` as a partial function.
## Main Results
* `Nat.Partrec.Code.rec_prim`: Recursion on `Nat.Partrec.Code` is primitive recursive.
* `Nat.Partrec.Code.rec_computable`: Recursion on `Nat.Partrec.Code` is computable.
* `Nat.Partrec.Code.smn`: The $S_n^m$ theorem.
* `Nat.Partrec.Code.exists_code`: Partial recursiveness is equivalent to being the eval of a code.
* `Nat.Partrec.Code.evaln_prim`: `evaln` is primitive recursive.
* `Nat.Partrec.Code.fixed_point`: Roger's fixed point theorem.
## References
* [Mario Carneiro, *Formalizing computability theory via partial recursive functions*][carneiro2019]
-/
open Encodable Denumerable Primrec
namespace Nat.Partrec
open Nat (pair)
theorem rfind' {f} (hf : Nat.Partrec f) :
Nat.Partrec
(Nat.unpaired fun a m =>
(Nat.rfind fun n => (fun m => m = 0) <$> f (Nat.pair a (n + m))).map (· + m)) :=
Partrec₂.unpaired'.2 <| by
refine'
Partrec.map
((@Partrec₂.unpaired' fun a b : ℕ =>
Nat.rfind fun n => (fun m => m = 0) <$> f (Nat.pair a (n + b))).1
_)
(Primrec.nat_add.comp Primrec.snd <| Primrec.snd.comp Primrec.fst).to_comp.to₂
have : Nat.Partrec (fun a => Nat.rfind (fun n => (fun m => decide (m = 0)) <$>
Nat.unpaired (fun a b => f (Nat.pair (Nat.unpair a).1 (b + (Nat.unpair a).2)))
(Nat.pair a n))) :=
rfind
(Partrec₂.unpaired'.2
((Partrec.nat_iff.2 hf).comp
(Primrec₂.pair.comp (Primrec.fst.comp <| Primrec.unpair.comp Primrec.fst)
(Primrec.nat_add.comp Primrec.snd
(Primrec.snd.comp <| Primrec.unpair.comp Primrec.fst))).to_comp))
simp at this;
|
exact this
|
theorem rfind' {f} (hf : Nat.Partrec f) :
Nat.Partrec
(Nat.unpaired fun a m =>
(Nat.rfind fun n => (fun m => m = 0) <$> f (Nat.pair a (n + m))).map (· + m)) :=
Partrec₂.unpaired'.2 <| by
refine'
Partrec.map
((@Partrec₂.unpaired' fun a b : ℕ =>
Nat.rfind fun n => (fun m => m = 0) <$> f (Nat.pair a (n + b))).1
_)
(Primrec.nat_add.comp Primrec.snd <| Primrec.snd.comp Primrec.fst).to_comp.to₂
have : Nat.Partrec (fun a => Nat.rfind (fun n => (fun m => decide (m = 0)) <$>
Nat.unpaired (fun a b => f (Nat.pair (Nat.unpair a).1 (b + (Nat.unpair a).2)))
(Nat.pair a n))) :=
rfind
(Partrec₂.unpaired'.2
((Partrec.nat_iff.2 hf).comp
(Primrec₂.pair.comp (Primrec.fst.comp <| Primrec.unpair.comp Primrec.fst)
(Primrec.nat_add.comp Primrec.snd
(Primrec.snd.comp <| Primrec.unpair.comp Primrec.fst))).to_comp))
simp at this;
|
Mathlib.Computability.PartrecCode.50_0.A3c3Aev6SyIRjCJ
|
theorem rfind' {f} (hf : Nat.Partrec f) :
Nat.Partrec
(Nat.unpaired fun a m =>
(Nat.rfind fun n => (fun m => m = 0) <$> f (Nat.pair a (n + m))).map (· + m))
|
Mathlib_Computability_PartrecCode
|
x✝ : Code.const 0 = Code.const 0
⊢ 0 = 0
|
/-
Copyright (c) 2018 Mario Carneiro. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Mario Carneiro
-/
import Mathlib.Computability.Partrec
#align_import computability.partrec_code from "leanprover-community/mathlib"@"6155d4351090a6fad236e3d2e4e0e4e7342668e8"
/-!
# Gödel Numbering for Partial Recursive Functions.
This file defines `Nat.Partrec.Code`, an inductive datatype describing code for partial
recursive functions on ℕ. It defines an encoding for these codes, and proves that the constructors
are primitive recursive with respect to the encoding.
It also defines the evaluation of these codes as partial functions using `PFun`, and proves that a
function is partially recursive (as defined by `Nat.Partrec`) if and only if it is the evaluation
of some code.
## Main Definitions
* `Nat.Partrec.Code`: Inductive datatype for partial recursive codes.
* `Nat.Partrec.Code.encodeCode`: A (computable) encoding of codes as natural numbers.
* `Nat.Partrec.Code.ofNatCode`: The inverse of this encoding.
* `Nat.Partrec.Code.eval`: The interpretation of a `Nat.Partrec.Code` as a partial function.
## Main Results
* `Nat.Partrec.Code.rec_prim`: Recursion on `Nat.Partrec.Code` is primitive recursive.
* `Nat.Partrec.Code.rec_computable`: Recursion on `Nat.Partrec.Code` is computable.
* `Nat.Partrec.Code.smn`: The $S_n^m$ theorem.
* `Nat.Partrec.Code.exists_code`: Partial recursiveness is equivalent to being the eval of a code.
* `Nat.Partrec.Code.evaln_prim`: `evaln` is primitive recursive.
* `Nat.Partrec.Code.fixed_point`: Roger's fixed point theorem.
## References
* [Mario Carneiro, *Formalizing computability theory via partial recursive functions*][carneiro2019]
-/
open Encodable Denumerable Primrec
namespace Nat.Partrec
open Nat (pair)
theorem rfind' {f} (hf : Nat.Partrec f) :
Nat.Partrec
(Nat.unpaired fun a m =>
(Nat.rfind fun n => (fun m => m = 0) <$> f (Nat.pair a (n + m))).map (· + m)) :=
Partrec₂.unpaired'.2 <| by
refine'
Partrec.map
((@Partrec₂.unpaired' fun a b : ℕ =>
Nat.rfind fun n => (fun m => m = 0) <$> f (Nat.pair a (n + b))).1
_)
(Primrec.nat_add.comp Primrec.snd <| Primrec.snd.comp Primrec.fst).to_comp.to₂
have : Nat.Partrec (fun a => Nat.rfind (fun n => (fun m => decide (m = 0)) <$>
Nat.unpaired (fun a b => f (Nat.pair (Nat.unpair a).1 (b + (Nat.unpair a).2)))
(Nat.pair a n))) :=
rfind
(Partrec₂.unpaired'.2
((Partrec.nat_iff.2 hf).comp
(Primrec₂.pair.comp (Primrec.fst.comp <| Primrec.unpair.comp Primrec.fst)
(Primrec.nat_add.comp Primrec.snd
(Primrec.snd.comp <| Primrec.unpair.comp Primrec.fst))).to_comp))
simp at this; exact this
#align nat.partrec.rfind' Nat.Partrec.rfind'
/-- Code for partial recursive functions from ℕ to ℕ.
See `Nat.Partrec.Code.eval` for the interpretation of these constructors.
-/
inductive Code : Type
| zero : Code
| succ : Code
| left : Code
| right : Code
| pair : Code → Code → Code
| comp : Code → Code → Code
| prec : Code → Code → Code
| rfind' : Code → Code
#align nat.partrec.code Nat.Partrec.Code
-- Porting note: `Nat.Partrec.Code.recOn` is noncomputable in Lean4, so we make it computable.
compile_inductive% Code
end Nat.Partrec
namespace Nat.Partrec.Code
open Nat (pair unpair)
open Nat.Partrec (Code)
instance instInhabited : Inhabited Code :=
⟨zero⟩
#align nat.partrec.code.inhabited Nat.Partrec.Code.instInhabited
/-- Returns a code for the constant function outputting a particular natural. -/
protected def const : ℕ → Code
| 0 => zero
| n + 1 => comp succ (Code.const n)
#align nat.partrec.code.const Nat.Partrec.Code.const
theorem const_inj : ∀ {n₁ n₂}, Nat.Partrec.Code.const n₁ = Nat.Partrec.Code.const n₂ → n₁ = n₂
| 0, 0, _ => by
|
simp
|
theorem const_inj : ∀ {n₁ n₂}, Nat.Partrec.Code.const n₁ = Nat.Partrec.Code.const n₂ → n₁ = n₂
| 0, 0, _ => by
|
Mathlib.Computability.PartrecCode.108_0.A3c3Aev6SyIRjCJ
|
theorem const_inj : ∀ {n₁ n₂}, Nat.Partrec.Code.const n₁ = Nat.Partrec.Code.const n₂ → n₁ = n₂
| 0, 0, _ => by simp
| n₁ + 1, n₂ + 1, h => by
dsimp [Nat.add_one, Nat.Partrec.Code.const] at h
injection h with h₁ h₂
simp only [const_inj h₂]
|
Mathlib_Computability_PartrecCode
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.