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 h
R : Type u
inst✝ : CommRing R
I : Ideal R
a b : R
S ι : Type v
x : ι → R ⧸ I
i : ι
⊢ AddHom.toFun
{
toAddHom :=
{
toFun := fun x =>
Quotient.liftOn' x (fun f i => (Quotient.mk I) (f i))
(_ :
∀ (a b : ι → R),
Setoid.r a b → (fun f i => (Quotient.mk I) (f i)) a = (fun f i => (Quotient.mk I) (f i)) b),
map_add' :=
(_ :
∀ (x y : (ι → R) ⧸ pi I ι),
(fun x =>
Quotient.liftOn' x (fun f i => (Quotient.mk I) (f i))
(_ :
∀ (a b : ι → R),
Setoid.r a b →
(fun f i => (Quotient.mk I) (f i)) a = (fun f i => (Quotient.mk I) (f i)) b))
(x + y) =
(fun x =>
Quotient.liftOn' x (fun f i => (Quotient.mk I) (f i))
(_ :
∀ (a b : ι → R),
Setoid.r a b →
(fun f i => (Quotient.mk I) (f i)) a = (fun f i => (Quotient.mk I) (f i)) b))
x +
(fun x =>
Quotient.liftOn' x (fun f i => (Quotient.mk I) (f i))
(_ :
∀ (a b : ι → R),
Setoid.r a b →
(fun f i => (Quotient.mk I) (f i)) a = (fun f i => (Quotient.mk I) (f i)) b))
y) },
map_smul' :=
(_ :
∀ (r : R ⧸ I) (x : (ι → R) ⧸ pi I ι),
AddHom.toFun
{
toFun := fun x =>
Quotient.liftOn' x (fun f i => (Quotient.mk I) (f i))
(_ :
∀ (a b : ι → R),
Setoid.r a b →
(fun f i => (Quotient.mk I) (f i)) a = (fun f i => (Quotient.mk I) (f i)) b),
map_add' :=
(_ :
∀ (x y : (ι → R) ⧸ pi I ι),
(fun x =>
Quotient.liftOn' x (fun f i => (Quotient.mk I) (f i))
(_ :
∀ (a b : ι → R),
Setoid.r a b →
(fun f i => (Quotient.mk I) (f i)) a = (fun f i => (Quotient.mk I) (f i)) b))
(x + y) =
(fun x =>
Quotient.liftOn' x (fun f i => (Quotient.mk I) (f i))
(_ :
∀ (a b : ι → R),
Setoid.r a b →
(fun f i => (Quotient.mk I) (f i)) a =
(fun f i => (Quotient.mk I) (f i)) b))
x +
(fun x =>
Quotient.liftOn' x (fun f i => (Quotient.mk I) (f i))
(_ :
∀ (a b : ι → R),
Setoid.r a b →
(fun f i => (Quotient.mk I) (f i)) a =
(fun f i => (Quotient.mk I) (f i)) b))
y) }
(r • x) =
(RingHom.id (R ⧸ I)) r •
AddHom.toFun
{
toFun := fun x =>
Quotient.liftOn' x (fun f i => (Quotient.mk I) (f i))
(_ :
∀ (a b : ι → R),
Setoid.r a b →
(fun f i => (Quotient.mk I) (f i)) a = (fun f i => (Quotient.mk I) (f i)) b),
map_add' :=
(_ :
∀ (x y : (ι → R) ⧸ pi I ι),
(fun x =>
Quotient.liftOn' x (fun f i => (Quotient.mk I) (f i))
(_ :
∀ (a b : ι → R),
Setoid.r a b →
(fun f i => (Quotient.mk I) (f i)) a =
(fun f i => (Quotient.mk I) (f i)) b))
(x + y) =
(fun x =>
Quotient.liftOn' x (fun f i => (Quotient.mk I) (f i))
(_ :
∀ (a b : ι → R),
Setoid.r a b →
(fun f i => (Quotient.mk I) (f i)) a =
(fun f i => (Quotient.mk I) (f i)) b))
x +
(fun x =>
Quotient.liftOn' x (fun f i => (Quotient.mk I) (f i))
(_ :
∀ (a b : ι → R),
Setoid.r a b →
(fun f i => (Quotient.mk I) (f i)) a =
(fun f i => (Quotient.mk I) (f i)) b))
y) }
x) }.toAddHom
((fun x => (Quotient.mk (pi I ι)) fun i => Quotient.out' (x i)) x) i =
x i
|
/-
Copyright (c) 2018 Kenny Lau. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Kenny Lau, Chris Hughes, Mario Carneiro, Anne Baanen
-/
import Mathlib.Algebra.Ring.Fin
import Mathlib.Algebra.Ring.Prod
import Mathlib.LinearAlgebra.Quotient
import Mathlib.RingTheory.Congruence
import Mathlib.RingTheory.Ideal.Basic
import Mathlib.Tactic.FinCases
#align_import ring_theory.ideal.quotient from "leanprover-community/mathlib"@"949dc57e616a621462062668c9f39e4e17b64b69"
/-!
# Ideal quotients
This file defines ideal quotients as a special case of submodule quotients and proves some basic
results about these quotients.
See `Algebra.RingQuot` for quotients of non-commutative rings.
## Main definitions
- `Ideal.Quotient`: the quotient of a commutative ring `R` by an ideal `I : Ideal R`
-/
universe u v w
namespace Ideal
open Set
open BigOperators
variable {R : Type u} [CommRing R] (I : Ideal R) {a b : R}
variable {S : Type v}
-- Note that at present `Ideal` means a left-ideal,
-- so this quotient is only useful in a commutative ring.
-- We should develop quotients by two-sided ideals as well.
/-- The quotient `R/I` of a ring `R` by an ideal `I`.
The ideal quotient of `I` is defined to equal the quotient of `I` as an `R`-submodule of `R`.
This definition is marked `reducible` so that typeclass instances can be shared between
`Ideal.Quotient I` and `Submodule.Quotient I`.
-/
@[reducible]
instance : HasQuotient R (Ideal R) :=
Submodule.hasQuotient
namespace Quotient
variable {I} {x y : R}
instance one (I : Ideal R) : One (R ⧸ I) :=
⟨Submodule.Quotient.mk 1⟩
#align ideal.quotient.has_one Ideal.Quotient.one
/-- On `Ideal`s, `Submodule.quotientRel` is a ring congruence. -/
protected def ringCon (I : Ideal R) : RingCon R :=
{ QuotientAddGroup.con I.toAddSubgroup with
mul' := fun {a₁ b₁ a₂ b₂} h₁ h₂ => by
rw [Submodule.quotientRel_r_def] at h₁ h₂ ⊢
have F := I.add_mem (I.mul_mem_left a₂ h₁) (I.mul_mem_right b₁ h₂)
have : a₁ * a₂ - b₁ * b₂ = a₂ * (a₁ - b₁) + (a₂ - b₂) * b₁ := by
rw [mul_sub, sub_mul, sub_add_sub_cancel, mul_comm, mul_comm b₁]
rwa [← this] at F }
#align ideal.quotient.ring_con Ideal.Quotient.ringCon
instance commRing (I : Ideal R) : CommRing (R ⧸ I) :=
inferInstanceAs (CommRing (Quotient.ringCon I).Quotient)
#align ideal.quotient.comm_ring Ideal.Quotient.commRing
-- Sanity test to make sure no diamonds have emerged in `commRing`
example : (commRing I).toAddCommGroup = Submodule.Quotient.addCommGroup I := rfl
-- this instance is harder to find than the one via `Algebra α (R ⧸ I)`, so use a lower priority
instance (priority := 100) isScalarTower_right {α} [SMul α R] [IsScalarTower α R R] :
IsScalarTower α (R ⧸ I) (R ⧸ I) :=
(Quotient.ringCon I).isScalarTower_right
#align ideal.quotient.is_scalar_tower_right Ideal.Quotient.isScalarTower_right
instance smulCommClass {α} [SMul α R] [IsScalarTower α R R] [SMulCommClass α R R] :
SMulCommClass α (R ⧸ I) (R ⧸ I) :=
(Quotient.ringCon I).smulCommClass
#align ideal.quotient.smul_comm_class Ideal.Quotient.smulCommClass
instance smulCommClass' {α} [SMul α R] [IsScalarTower α R R] [SMulCommClass R α R] :
SMulCommClass (R ⧸ I) α (R ⧸ I) :=
(Quotient.ringCon I).smulCommClass'
#align ideal.quotient.smul_comm_class' Ideal.Quotient.smulCommClass'
/-- The ring homomorphism from a ring `R` to a quotient ring `R/I`. -/
def mk (I : Ideal R) : R →+* R ⧸ I where
toFun a := Submodule.Quotient.mk a
map_zero' := rfl
map_one' := rfl
map_mul' _ _ := rfl
map_add' _ _ := rfl
#align ideal.quotient.mk Ideal.Quotient.mk
instance {I : Ideal R} : Coe R (R ⧸ I) :=
⟨Ideal.Quotient.mk I⟩
/-- Two `RingHom`s from the quotient by an ideal are equal if their
compositions with `Ideal.Quotient.mk'` are equal.
See note [partially-applied ext lemmas]. -/
@[ext 1100]
theorem ringHom_ext [NonAssocSemiring S] ⦃f g : R ⧸ I →+* S⦄ (h : f.comp (mk I) = g.comp (mk I)) :
f = g :=
RingHom.ext fun x => Quotient.inductionOn' x <| (RingHom.congr_fun h : _)
#align ideal.quotient.ring_hom_ext Ideal.Quotient.ringHom_ext
instance inhabited : Inhabited (R ⧸ I) :=
⟨mk I 37⟩
#align ideal.quotient.inhabited Ideal.Quotient.inhabited
protected theorem eq : mk I x = mk I y ↔ x - y ∈ I :=
Submodule.Quotient.eq I
#align ideal.quotient.eq Ideal.Quotient.eq
@[simp]
theorem mk_eq_mk (x : R) : (Submodule.Quotient.mk x : R ⧸ I) = mk I x := rfl
#align ideal.quotient.mk_eq_mk Ideal.Quotient.mk_eq_mk
theorem eq_zero_iff_mem {I : Ideal R} : mk I a = 0 ↔ a ∈ I :=
Submodule.Quotient.mk_eq_zero _
#align ideal.quotient.eq_zero_iff_mem Ideal.Quotient.eq_zero_iff_mem
theorem eq_zero_iff_dvd (x y : R) : Ideal.Quotient.mk (Ideal.span ({x} : Set R)) y = 0 ↔ x ∣ y := by
rw [Ideal.Quotient.eq_zero_iff_mem, Ideal.mem_span_singleton]
-- Porting note: new theorem
theorem mk_eq_mk_iff_sub_mem (x y : R) : mk I x = mk I y ↔ x - y ∈ I := by
rw [← eq_zero_iff_mem, map_sub, sub_eq_zero]
theorem zero_eq_one_iff {I : Ideal R} : (0 : R ⧸ I) = 1 ↔ I = ⊤ :=
eq_comm.trans <| eq_zero_iff_mem.trans (eq_top_iff_one _).symm
#align ideal.quotient.zero_eq_one_iff Ideal.Quotient.zero_eq_one_iff
theorem zero_ne_one_iff {I : Ideal R} : (0 : R ⧸ I) ≠ 1 ↔ I ≠ ⊤ :=
not_congr zero_eq_one_iff
#align ideal.quotient.zero_ne_one_iff Ideal.Quotient.zero_ne_one_iff
protected theorem nontrivial {I : Ideal R} (hI : I ≠ ⊤) : Nontrivial (R ⧸ I) :=
⟨⟨0, 1, zero_ne_one_iff.2 hI⟩⟩
#align ideal.quotient.nontrivial Ideal.Quotient.nontrivial
theorem subsingleton_iff {I : Ideal R} : Subsingleton (R ⧸ I) ↔ I = ⊤ := by
rw [eq_top_iff_one, ← subsingleton_iff_zero_eq_one, eq_comm, ← (mk I).map_one,
Quotient.eq_zero_iff_mem]
#align ideal.quotient.subsingleton_iff Ideal.Quotient.subsingleton_iff
instance : Unique (R ⧸ (⊤ : Ideal R)) :=
⟨⟨0⟩, by rintro ⟨x⟩; exact Quotient.eq_zero_iff_mem.mpr Submodule.mem_top⟩
theorem mk_surjective : Function.Surjective (mk I) := fun y =>
Quotient.inductionOn' y fun x => Exists.intro x rfl
#align ideal.quotient.mk_surjective Ideal.Quotient.mk_surjective
instance : RingHomSurjective (mk I) :=
⟨mk_surjective⟩
/-- If `I` is an ideal of a commutative ring `R`, if `q : R → R/I` is the quotient map, and if
`s ⊆ R` is a subset, then `q⁻¹(q(s)) = ⋃ᵢ(i + s)`, the union running over all `i ∈ I`. -/
theorem quotient_ring_saturate (I : Ideal R) (s : Set R) :
mk I ⁻¹' (mk I '' s) = ⋃ x : I, (fun y => x.1 + y) '' s := by
ext x
simp only [mem_preimage, mem_image, mem_iUnion, Ideal.Quotient.eq]
exact
⟨fun ⟨a, a_in, h⟩ => ⟨⟨_, I.neg_mem h⟩, a, a_in, by simp⟩, fun ⟨⟨i, hi⟩, a, ha, Eq⟩ =>
⟨a, ha, by rw [← Eq, sub_add_eq_sub_sub_swap, sub_self, zero_sub]; exact I.neg_mem hi⟩⟩
#align ideal.quotient.quotient_ring_saturate Ideal.Quotient.quotient_ring_saturate
instance noZeroDivisors (I : Ideal R) [hI : I.IsPrime] : NoZeroDivisors (R ⧸ I) where
eq_zero_or_eq_zero_of_mul_eq_zero {a b} := Quotient.inductionOn₂' a b fun {_ _} hab =>
(hI.mem_or_mem (eq_zero_iff_mem.1 hab)).elim (Or.inl ∘ eq_zero_iff_mem.2)
(Or.inr ∘ eq_zero_iff_mem.2)
#align ideal.quotient.no_zero_divisors Ideal.Quotient.noZeroDivisors
instance isDomain (I : Ideal R) [hI : I.IsPrime] : IsDomain (R ⧸ I) :=
let _ := Quotient.nontrivial hI.1
NoZeroDivisors.to_isDomain _
#align ideal.quotient.is_domain Ideal.Quotient.isDomain
theorem isDomain_iff_prime (I : Ideal R) : IsDomain (R ⧸ I) ↔ I.IsPrime := by
refine' ⟨fun H => ⟨zero_ne_one_iff.1 _, fun {x y} h => _⟩, fun h => inferInstance⟩
· haveI : Nontrivial (R ⧸ I) := ⟨H.2.1⟩
exact zero_ne_one
· simp only [← eq_zero_iff_mem, (mk I).map_mul] at h ⊢
haveI := @IsDomain.to_noZeroDivisors (R ⧸ I) _ H
exact eq_zero_or_eq_zero_of_mul_eq_zero h
#align ideal.quotient.is_domain_iff_prime Ideal.Quotient.isDomain_iff_prime
theorem exists_inv {I : Ideal R} [hI : I.IsMaximal] :
∀ {a : R ⧸ I}, a ≠ 0 → ∃ b : R ⧸ I, a * b = 1 := by
rintro ⟨a⟩ h
rcases hI.exists_inv (mt eq_zero_iff_mem.2 h) with ⟨b, c, hc, abc⟩
rw [mul_comm] at abc
refine' ⟨mk _ b, Quot.sound _⟩
simp only [Submodule.quotientRel_r_def]
rw [← eq_sub_iff_add_eq'] at abc
rwa [abc, ← neg_mem_iff (G := R) (H := I), neg_sub] at hc
#align ideal.quotient.exists_inv Ideal.Quotient.exists_inv
open Classical
/-- The quotient by a maximal ideal is a group with zero. This is a `def` rather than `instance`,
since users will have computable inverses in some applications.
See note [reducible non-instances]. -/
@[reducible]
protected noncomputable def groupWithZero (I : Ideal R) [hI : I.IsMaximal] :
GroupWithZero (R ⧸ I) :=
{ inv := fun a => if ha : a = 0 then 0 else Classical.choose (exists_inv ha)
mul_inv_cancel := fun a (ha : a ≠ 0) =>
show a * dite _ _ _ = _ by rw [dif_neg ha]; exact Classical.choose_spec (exists_inv ha)
inv_zero := dif_pos rfl }
#align ideal.quotient.group_with_zero Ideal.Quotient.groupWithZero
/-- The quotient by a maximal ideal is a field. This is a `def` rather than `instance`, since users
will have computable inverses (and `qsmul`, `rat_cast`) in some applications.
See note [reducible non-instances]. -/
@[reducible]
protected noncomputable def field (I : Ideal R) [hI : I.IsMaximal] : Field (R ⧸ I) :=
{ Quotient.commRing I, Quotient.groupWithZero I with }
#align ideal.quotient.field Ideal.Quotient.field
/-- If the quotient by an ideal is a field, then the ideal is maximal. -/
theorem maximal_of_isField (I : Ideal R) (hqf : IsField (R ⧸ I)) : I.IsMaximal := by
apply Ideal.isMaximal_iff.2
constructor
· intro h
rcases hqf.exists_pair_ne with ⟨⟨x⟩, ⟨y⟩, hxy⟩
exact hxy (Ideal.Quotient.eq.2 (mul_one (x - y) ▸ I.mul_mem_left _ h))
· intro J x hIJ hxnI hxJ
rcases hqf.mul_inv_cancel (mt Ideal.Quotient.eq_zero_iff_mem.1 hxnI) with ⟨⟨y⟩, hy⟩
rw [← zero_add (1 : R), ← sub_self (x * y), sub_add]
refine' J.sub_mem (J.mul_mem_right _ hxJ) (hIJ (Ideal.Quotient.eq.1 hy))
#align ideal.quotient.maximal_of_is_field Ideal.Quotient.maximal_of_isField
/-- The quotient of a ring by an ideal is a field iff the ideal is maximal. -/
theorem maximal_ideal_iff_isField_quotient (I : Ideal R) : I.IsMaximal ↔ IsField (R ⧸ I) :=
⟨fun h =>
let _i := @Quotient.field _ _ I h
Field.toIsField _,
maximal_of_isField _⟩
#align ideal.quotient.maximal_ideal_iff_is_field_quotient Ideal.Quotient.maximal_ideal_iff_isField_quotient
variable [CommRing S]
/-- Given a ring homomorphism `f : R →+* S` sending all elements of an ideal to zero,
lift it to the quotient by this ideal. -/
def lift (I : Ideal R) (f : R →+* S) (H : ∀ a : R, a ∈ I → f a = 0) : R ⧸ I →+* S :=
{ QuotientAddGroup.lift I.toAddSubgroup f.toAddMonoidHom H with
map_one' := f.map_one
map_mul' := fun a₁ a₂ => Quotient.inductionOn₂' a₁ a₂ f.map_mul }
#align ideal.quotient.lift Ideal.Quotient.lift
@[simp]
theorem lift_mk (I : Ideal R) (f : R →+* S) (H : ∀ a : R, a ∈ I → f a = 0) :
lift I f H (mk I a) = f a :=
rfl
#align ideal.quotient.lift_mk Ideal.Quotient.lift_mk
theorem lift_surjective_of_surjective (I : Ideal R) {f : R →+* S} (H : ∀ a : R, a ∈ I → f a = 0)
(hf : Function.Surjective f) : Function.Surjective (Ideal.Quotient.lift I f H) := by
intro y
obtain ⟨x, rfl⟩ := hf y
use Ideal.Quotient.mk I x
simp only [Ideal.Quotient.lift_mk]
#align ideal.quotient.lift_surjective_of_surjective Ideal.Quotient.lift_surjective_of_surjective
/-- The ring homomorphism from the quotient by a smaller ideal to the quotient by a larger ideal.
This is the `Ideal.Quotient` version of `Quot.Factor` -/
def factor (S T : Ideal R) (H : S ≤ T) : R ⧸ S →+* R ⧸ T :=
Ideal.Quotient.lift S (mk T) fun _ hx => eq_zero_iff_mem.2 (H hx)
#align ideal.quotient.factor Ideal.Quotient.factor
@[simp]
theorem factor_mk (S T : Ideal R) (H : S ≤ T) (x : R) : factor S T H (mk S x) = mk T x :=
rfl
#align ideal.quotient.factor_mk Ideal.Quotient.factor_mk
@[simp]
theorem factor_comp_mk (S T : Ideal R) (H : S ≤ T) : (factor S T H).comp (mk S) = mk T := by
ext x
rw [RingHom.comp_apply, factor_mk]
#align ideal.quotient.factor_comp_mk Ideal.Quotient.factor_comp_mk
end Quotient
/-- Quotienting by equal ideals gives equivalent rings.
See also `Submodule.quotEquivOfEq` and `Ideal.quotientEquivAlgOfEq`.
-/
def quotEquivOfEq {R : Type*} [CommRing R] {I J : Ideal R} (h : I = J) : R ⧸ I ≃+* R ⧸ J :=
{ Submodule.quotEquivOfEq I J h with
map_mul' := by
rintro ⟨x⟩ ⟨y⟩
rfl }
#align ideal.quot_equiv_of_eq Ideal.quotEquivOfEq
@[simp]
theorem quotEquivOfEq_mk {R : Type*} [CommRing R] {I J : Ideal R} (h : I = J) (x : R) :
quotEquivOfEq h (Ideal.Quotient.mk I x) = Ideal.Quotient.mk J x :=
rfl
#align ideal.quot_equiv_of_eq_mk Ideal.quotEquivOfEq_mk
@[simp]
theorem quotEquivOfEq_symm {R : Type*} [CommRing R] {I J : Ideal R} (h : I = J) :
(Ideal.quotEquivOfEq h).symm = Ideal.quotEquivOfEq h.symm := by ext; rfl
#align ideal.quot_equiv_of_eq_symm Ideal.quotEquivOfEq_symm
section Pi
variable (ι : Type v)
/-- `R^n/I^n` is a `R/I`-module. -/
instance modulePi : Module (R ⧸ I) ((ι → R) ⧸ I.pi ι) where
smul c m :=
Quotient.liftOn₂' c m (fun r m => Submodule.Quotient.mk <| r • m) $ by
intro c₁ m₁ c₂ m₂ hc hm
apply Ideal.Quotient.eq.2
rw [Submodule.quotientRel_r_def] at hc hm
intro i
exact I.mul_sub_mul_mem hc (hm i)
one_smul := by
rintro ⟨a⟩
convert_to Ideal.Quotient.mk (I.pi ι) _ = Ideal.Quotient.mk (I.pi ι) _
congr with i; exact one_mul (a i)
mul_smul := by
rintro ⟨a⟩ ⟨b⟩ ⟨c⟩
convert_to Ideal.Quotient.mk (I.pi ι) _ = Ideal.Quotient.mk (I.pi ι) _
congr 1; funext i; exact mul_assoc a b (c i)
smul_add := by
rintro ⟨a⟩ ⟨b⟩ ⟨c⟩
convert_to Ideal.Quotient.mk (I.pi ι) _ = Ideal.Quotient.mk (I.pi ι) _
congr with i; exact mul_add a (b i) (c i)
smul_zero := by
rintro ⟨a⟩
convert_to Ideal.Quotient.mk (I.pi ι) _ = Ideal.Quotient.mk (I.pi ι) _
congr with _; exact mul_zero a
add_smul := by
rintro ⟨a⟩ ⟨b⟩ ⟨c⟩
convert_to Ideal.Quotient.mk (I.pi ι) _ = Ideal.Quotient.mk (I.pi ι) _
congr with i; exact add_mul a b (c i)
zero_smul := by
rintro ⟨a⟩
convert_to Ideal.Quotient.mk (I.pi ι) _ = Ideal.Quotient.mk (I.pi ι) _
congr with i; exact zero_mul (a i)
#align ideal.module_pi Ideal.modulePi
/-- `R^n/I^n` is isomorphic to `(R/I)^n` as an `R/I`-module. -/
noncomputable def piQuotEquiv : ((ι → R) ⧸ I.pi ι) ≃ₗ[R ⧸ I] ι → (R ⧸ I) where
toFun := fun x ↦
Quotient.liftOn' x (fun f i => Ideal.Quotient.mk I (f i)) fun a b hab =>
funext fun i => (Submodule.Quotient.eq' _).2 (QuotientAddGroup.leftRel_apply.mp hab i)
map_add' := by rintro ⟨_⟩ ⟨_⟩; rfl
map_smul' := by rintro ⟨_⟩ ⟨_⟩; rfl
invFun := fun x ↦ Ideal.Quotient.mk (I.pi ι) fun i ↦ Quotient.out' (x i)
left_inv := by
rintro ⟨x⟩
exact Ideal.Quotient.eq.2 fun i => Ideal.Quotient.eq.1 (Quotient.out_eq' _)
right_inv := by
intro x
ext i
|
obtain ⟨_, _⟩ := @Quot.exists_rep _ _ (x i)
|
/-- `R^n/I^n` is isomorphic to `(R/I)^n` as an `R/I`-module. -/
noncomputable def piQuotEquiv : ((ι → R) ⧸ I.pi ι) ≃ₗ[R ⧸ I] ι → (R ⧸ I) where
toFun := fun x ↦
Quotient.liftOn' x (fun f i => Ideal.Quotient.mk I (f i)) fun a b hab =>
funext fun i => (Submodule.Quotient.eq' _).2 (QuotientAddGroup.leftRel_apply.mp hab i)
map_add' := by rintro ⟨_⟩ ⟨_⟩; rfl
map_smul' := by rintro ⟨_⟩ ⟨_⟩; rfl
invFun := fun x ↦ Ideal.Quotient.mk (I.pi ι) fun i ↦ Quotient.out' (x i)
left_inv := by
rintro ⟨x⟩
exact Ideal.Quotient.eq.2 fun i => Ideal.Quotient.eq.1 (Quotient.out_eq' _)
right_inv := by
intro x
ext i
|
Mathlib.RingTheory.Ideal.Quotient.359_0.TwNAv7Pc4PYOWjX
|
/-- `R^n/I^n` is isomorphic to `(R/I)^n` as an `R/I`-module. -/
noncomputable def piQuotEquiv : ((ι → R) ⧸ I.pi ι) ≃ₗ[R ⧸ I] ι → (R ⧸ I) where
toFun
|
Mathlib_RingTheory_Ideal_Quotient
|
case h.intro
R : Type u
inst✝ : CommRing R
I : Ideal R
a b : R
S ι : Type v
x : ι → R ⧸ I
i : ι
w✝ : R
h✝ : Quot.mk Setoid.r w✝ = x i
⊢ AddHom.toFun
{
toAddHom :=
{
toFun := fun x =>
Quotient.liftOn' x (fun f i => (Quotient.mk I) (f i))
(_ :
∀ (a b : ι → R),
Setoid.r a b → (fun f i => (Quotient.mk I) (f i)) a = (fun f i => (Quotient.mk I) (f i)) b),
map_add' :=
(_ :
∀ (x y : (ι → R) ⧸ pi I ι),
(fun x =>
Quotient.liftOn' x (fun f i => (Quotient.mk I) (f i))
(_ :
∀ (a b : ι → R),
Setoid.r a b →
(fun f i => (Quotient.mk I) (f i)) a = (fun f i => (Quotient.mk I) (f i)) b))
(x + y) =
(fun x =>
Quotient.liftOn' x (fun f i => (Quotient.mk I) (f i))
(_ :
∀ (a b : ι → R),
Setoid.r a b →
(fun f i => (Quotient.mk I) (f i)) a = (fun f i => (Quotient.mk I) (f i)) b))
x +
(fun x =>
Quotient.liftOn' x (fun f i => (Quotient.mk I) (f i))
(_ :
∀ (a b : ι → R),
Setoid.r a b →
(fun f i => (Quotient.mk I) (f i)) a = (fun f i => (Quotient.mk I) (f i)) b))
y) },
map_smul' :=
(_ :
∀ (r : R ⧸ I) (x : (ι → R) ⧸ pi I ι),
AddHom.toFun
{
toFun := fun x =>
Quotient.liftOn' x (fun f i => (Quotient.mk I) (f i))
(_ :
∀ (a b : ι → R),
Setoid.r a b →
(fun f i => (Quotient.mk I) (f i)) a = (fun f i => (Quotient.mk I) (f i)) b),
map_add' :=
(_ :
∀ (x y : (ι → R) ⧸ pi I ι),
(fun x =>
Quotient.liftOn' x (fun f i => (Quotient.mk I) (f i))
(_ :
∀ (a b : ι → R),
Setoid.r a b →
(fun f i => (Quotient.mk I) (f i)) a = (fun f i => (Quotient.mk I) (f i)) b))
(x + y) =
(fun x =>
Quotient.liftOn' x (fun f i => (Quotient.mk I) (f i))
(_ :
∀ (a b : ι → R),
Setoid.r a b →
(fun f i => (Quotient.mk I) (f i)) a =
(fun f i => (Quotient.mk I) (f i)) b))
x +
(fun x =>
Quotient.liftOn' x (fun f i => (Quotient.mk I) (f i))
(_ :
∀ (a b : ι → R),
Setoid.r a b →
(fun f i => (Quotient.mk I) (f i)) a =
(fun f i => (Quotient.mk I) (f i)) b))
y) }
(r • x) =
(RingHom.id (R ⧸ I)) r •
AddHom.toFun
{
toFun := fun x =>
Quotient.liftOn' x (fun f i => (Quotient.mk I) (f i))
(_ :
∀ (a b : ι → R),
Setoid.r a b →
(fun f i => (Quotient.mk I) (f i)) a = (fun f i => (Quotient.mk I) (f i)) b),
map_add' :=
(_ :
∀ (x y : (ι → R) ⧸ pi I ι),
(fun x =>
Quotient.liftOn' x (fun f i => (Quotient.mk I) (f i))
(_ :
∀ (a b : ι → R),
Setoid.r a b →
(fun f i => (Quotient.mk I) (f i)) a =
(fun f i => (Quotient.mk I) (f i)) b))
(x + y) =
(fun x =>
Quotient.liftOn' x (fun f i => (Quotient.mk I) (f i))
(_ :
∀ (a b : ι → R),
Setoid.r a b →
(fun f i => (Quotient.mk I) (f i)) a =
(fun f i => (Quotient.mk I) (f i)) b))
x +
(fun x =>
Quotient.liftOn' x (fun f i => (Quotient.mk I) (f i))
(_ :
∀ (a b : ι → R),
Setoid.r a b →
(fun f i => (Quotient.mk I) (f i)) a =
(fun f i => (Quotient.mk I) (f i)) b))
y) }
x) }.toAddHom
((fun x => (Quotient.mk (pi I ι)) fun i => Quotient.out' (x i)) x) i =
x i
|
/-
Copyright (c) 2018 Kenny Lau. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Kenny Lau, Chris Hughes, Mario Carneiro, Anne Baanen
-/
import Mathlib.Algebra.Ring.Fin
import Mathlib.Algebra.Ring.Prod
import Mathlib.LinearAlgebra.Quotient
import Mathlib.RingTheory.Congruence
import Mathlib.RingTheory.Ideal.Basic
import Mathlib.Tactic.FinCases
#align_import ring_theory.ideal.quotient from "leanprover-community/mathlib"@"949dc57e616a621462062668c9f39e4e17b64b69"
/-!
# Ideal quotients
This file defines ideal quotients as a special case of submodule quotients and proves some basic
results about these quotients.
See `Algebra.RingQuot` for quotients of non-commutative rings.
## Main definitions
- `Ideal.Quotient`: the quotient of a commutative ring `R` by an ideal `I : Ideal R`
-/
universe u v w
namespace Ideal
open Set
open BigOperators
variable {R : Type u} [CommRing R] (I : Ideal R) {a b : R}
variable {S : Type v}
-- Note that at present `Ideal` means a left-ideal,
-- so this quotient is only useful in a commutative ring.
-- We should develop quotients by two-sided ideals as well.
/-- The quotient `R/I` of a ring `R` by an ideal `I`.
The ideal quotient of `I` is defined to equal the quotient of `I` as an `R`-submodule of `R`.
This definition is marked `reducible` so that typeclass instances can be shared between
`Ideal.Quotient I` and `Submodule.Quotient I`.
-/
@[reducible]
instance : HasQuotient R (Ideal R) :=
Submodule.hasQuotient
namespace Quotient
variable {I} {x y : R}
instance one (I : Ideal R) : One (R ⧸ I) :=
⟨Submodule.Quotient.mk 1⟩
#align ideal.quotient.has_one Ideal.Quotient.one
/-- On `Ideal`s, `Submodule.quotientRel` is a ring congruence. -/
protected def ringCon (I : Ideal R) : RingCon R :=
{ QuotientAddGroup.con I.toAddSubgroup with
mul' := fun {a₁ b₁ a₂ b₂} h₁ h₂ => by
rw [Submodule.quotientRel_r_def] at h₁ h₂ ⊢
have F := I.add_mem (I.mul_mem_left a₂ h₁) (I.mul_mem_right b₁ h₂)
have : a₁ * a₂ - b₁ * b₂ = a₂ * (a₁ - b₁) + (a₂ - b₂) * b₁ := by
rw [mul_sub, sub_mul, sub_add_sub_cancel, mul_comm, mul_comm b₁]
rwa [← this] at F }
#align ideal.quotient.ring_con Ideal.Quotient.ringCon
instance commRing (I : Ideal R) : CommRing (R ⧸ I) :=
inferInstanceAs (CommRing (Quotient.ringCon I).Quotient)
#align ideal.quotient.comm_ring Ideal.Quotient.commRing
-- Sanity test to make sure no diamonds have emerged in `commRing`
example : (commRing I).toAddCommGroup = Submodule.Quotient.addCommGroup I := rfl
-- this instance is harder to find than the one via `Algebra α (R ⧸ I)`, so use a lower priority
instance (priority := 100) isScalarTower_right {α} [SMul α R] [IsScalarTower α R R] :
IsScalarTower α (R ⧸ I) (R ⧸ I) :=
(Quotient.ringCon I).isScalarTower_right
#align ideal.quotient.is_scalar_tower_right Ideal.Quotient.isScalarTower_right
instance smulCommClass {α} [SMul α R] [IsScalarTower α R R] [SMulCommClass α R R] :
SMulCommClass α (R ⧸ I) (R ⧸ I) :=
(Quotient.ringCon I).smulCommClass
#align ideal.quotient.smul_comm_class Ideal.Quotient.smulCommClass
instance smulCommClass' {α} [SMul α R] [IsScalarTower α R R] [SMulCommClass R α R] :
SMulCommClass (R ⧸ I) α (R ⧸ I) :=
(Quotient.ringCon I).smulCommClass'
#align ideal.quotient.smul_comm_class' Ideal.Quotient.smulCommClass'
/-- The ring homomorphism from a ring `R` to a quotient ring `R/I`. -/
def mk (I : Ideal R) : R →+* R ⧸ I where
toFun a := Submodule.Quotient.mk a
map_zero' := rfl
map_one' := rfl
map_mul' _ _ := rfl
map_add' _ _ := rfl
#align ideal.quotient.mk Ideal.Quotient.mk
instance {I : Ideal R} : Coe R (R ⧸ I) :=
⟨Ideal.Quotient.mk I⟩
/-- Two `RingHom`s from the quotient by an ideal are equal if their
compositions with `Ideal.Quotient.mk'` are equal.
See note [partially-applied ext lemmas]. -/
@[ext 1100]
theorem ringHom_ext [NonAssocSemiring S] ⦃f g : R ⧸ I →+* S⦄ (h : f.comp (mk I) = g.comp (mk I)) :
f = g :=
RingHom.ext fun x => Quotient.inductionOn' x <| (RingHom.congr_fun h : _)
#align ideal.quotient.ring_hom_ext Ideal.Quotient.ringHom_ext
instance inhabited : Inhabited (R ⧸ I) :=
⟨mk I 37⟩
#align ideal.quotient.inhabited Ideal.Quotient.inhabited
protected theorem eq : mk I x = mk I y ↔ x - y ∈ I :=
Submodule.Quotient.eq I
#align ideal.quotient.eq Ideal.Quotient.eq
@[simp]
theorem mk_eq_mk (x : R) : (Submodule.Quotient.mk x : R ⧸ I) = mk I x := rfl
#align ideal.quotient.mk_eq_mk Ideal.Quotient.mk_eq_mk
theorem eq_zero_iff_mem {I : Ideal R} : mk I a = 0 ↔ a ∈ I :=
Submodule.Quotient.mk_eq_zero _
#align ideal.quotient.eq_zero_iff_mem Ideal.Quotient.eq_zero_iff_mem
theorem eq_zero_iff_dvd (x y : R) : Ideal.Quotient.mk (Ideal.span ({x} : Set R)) y = 0 ↔ x ∣ y := by
rw [Ideal.Quotient.eq_zero_iff_mem, Ideal.mem_span_singleton]
-- Porting note: new theorem
theorem mk_eq_mk_iff_sub_mem (x y : R) : mk I x = mk I y ↔ x - y ∈ I := by
rw [← eq_zero_iff_mem, map_sub, sub_eq_zero]
theorem zero_eq_one_iff {I : Ideal R} : (0 : R ⧸ I) = 1 ↔ I = ⊤ :=
eq_comm.trans <| eq_zero_iff_mem.trans (eq_top_iff_one _).symm
#align ideal.quotient.zero_eq_one_iff Ideal.Quotient.zero_eq_one_iff
theorem zero_ne_one_iff {I : Ideal R} : (0 : R ⧸ I) ≠ 1 ↔ I ≠ ⊤ :=
not_congr zero_eq_one_iff
#align ideal.quotient.zero_ne_one_iff Ideal.Quotient.zero_ne_one_iff
protected theorem nontrivial {I : Ideal R} (hI : I ≠ ⊤) : Nontrivial (R ⧸ I) :=
⟨⟨0, 1, zero_ne_one_iff.2 hI⟩⟩
#align ideal.quotient.nontrivial Ideal.Quotient.nontrivial
theorem subsingleton_iff {I : Ideal R} : Subsingleton (R ⧸ I) ↔ I = ⊤ := by
rw [eq_top_iff_one, ← subsingleton_iff_zero_eq_one, eq_comm, ← (mk I).map_one,
Quotient.eq_zero_iff_mem]
#align ideal.quotient.subsingleton_iff Ideal.Quotient.subsingleton_iff
instance : Unique (R ⧸ (⊤ : Ideal R)) :=
⟨⟨0⟩, by rintro ⟨x⟩; exact Quotient.eq_zero_iff_mem.mpr Submodule.mem_top⟩
theorem mk_surjective : Function.Surjective (mk I) := fun y =>
Quotient.inductionOn' y fun x => Exists.intro x rfl
#align ideal.quotient.mk_surjective Ideal.Quotient.mk_surjective
instance : RingHomSurjective (mk I) :=
⟨mk_surjective⟩
/-- If `I` is an ideal of a commutative ring `R`, if `q : R → R/I` is the quotient map, and if
`s ⊆ R` is a subset, then `q⁻¹(q(s)) = ⋃ᵢ(i + s)`, the union running over all `i ∈ I`. -/
theorem quotient_ring_saturate (I : Ideal R) (s : Set R) :
mk I ⁻¹' (mk I '' s) = ⋃ x : I, (fun y => x.1 + y) '' s := by
ext x
simp only [mem_preimage, mem_image, mem_iUnion, Ideal.Quotient.eq]
exact
⟨fun ⟨a, a_in, h⟩ => ⟨⟨_, I.neg_mem h⟩, a, a_in, by simp⟩, fun ⟨⟨i, hi⟩, a, ha, Eq⟩ =>
⟨a, ha, by rw [← Eq, sub_add_eq_sub_sub_swap, sub_self, zero_sub]; exact I.neg_mem hi⟩⟩
#align ideal.quotient.quotient_ring_saturate Ideal.Quotient.quotient_ring_saturate
instance noZeroDivisors (I : Ideal R) [hI : I.IsPrime] : NoZeroDivisors (R ⧸ I) where
eq_zero_or_eq_zero_of_mul_eq_zero {a b} := Quotient.inductionOn₂' a b fun {_ _} hab =>
(hI.mem_or_mem (eq_zero_iff_mem.1 hab)).elim (Or.inl ∘ eq_zero_iff_mem.2)
(Or.inr ∘ eq_zero_iff_mem.2)
#align ideal.quotient.no_zero_divisors Ideal.Quotient.noZeroDivisors
instance isDomain (I : Ideal R) [hI : I.IsPrime] : IsDomain (R ⧸ I) :=
let _ := Quotient.nontrivial hI.1
NoZeroDivisors.to_isDomain _
#align ideal.quotient.is_domain Ideal.Quotient.isDomain
theorem isDomain_iff_prime (I : Ideal R) : IsDomain (R ⧸ I) ↔ I.IsPrime := by
refine' ⟨fun H => ⟨zero_ne_one_iff.1 _, fun {x y} h => _⟩, fun h => inferInstance⟩
· haveI : Nontrivial (R ⧸ I) := ⟨H.2.1⟩
exact zero_ne_one
· simp only [← eq_zero_iff_mem, (mk I).map_mul] at h ⊢
haveI := @IsDomain.to_noZeroDivisors (R ⧸ I) _ H
exact eq_zero_or_eq_zero_of_mul_eq_zero h
#align ideal.quotient.is_domain_iff_prime Ideal.Quotient.isDomain_iff_prime
theorem exists_inv {I : Ideal R} [hI : I.IsMaximal] :
∀ {a : R ⧸ I}, a ≠ 0 → ∃ b : R ⧸ I, a * b = 1 := by
rintro ⟨a⟩ h
rcases hI.exists_inv (mt eq_zero_iff_mem.2 h) with ⟨b, c, hc, abc⟩
rw [mul_comm] at abc
refine' ⟨mk _ b, Quot.sound _⟩
simp only [Submodule.quotientRel_r_def]
rw [← eq_sub_iff_add_eq'] at abc
rwa [abc, ← neg_mem_iff (G := R) (H := I), neg_sub] at hc
#align ideal.quotient.exists_inv Ideal.Quotient.exists_inv
open Classical
/-- The quotient by a maximal ideal is a group with zero. This is a `def` rather than `instance`,
since users will have computable inverses in some applications.
See note [reducible non-instances]. -/
@[reducible]
protected noncomputable def groupWithZero (I : Ideal R) [hI : I.IsMaximal] :
GroupWithZero (R ⧸ I) :=
{ inv := fun a => if ha : a = 0 then 0 else Classical.choose (exists_inv ha)
mul_inv_cancel := fun a (ha : a ≠ 0) =>
show a * dite _ _ _ = _ by rw [dif_neg ha]; exact Classical.choose_spec (exists_inv ha)
inv_zero := dif_pos rfl }
#align ideal.quotient.group_with_zero Ideal.Quotient.groupWithZero
/-- The quotient by a maximal ideal is a field. This is a `def` rather than `instance`, since users
will have computable inverses (and `qsmul`, `rat_cast`) in some applications.
See note [reducible non-instances]. -/
@[reducible]
protected noncomputable def field (I : Ideal R) [hI : I.IsMaximal] : Field (R ⧸ I) :=
{ Quotient.commRing I, Quotient.groupWithZero I with }
#align ideal.quotient.field Ideal.Quotient.field
/-- If the quotient by an ideal is a field, then the ideal is maximal. -/
theorem maximal_of_isField (I : Ideal R) (hqf : IsField (R ⧸ I)) : I.IsMaximal := by
apply Ideal.isMaximal_iff.2
constructor
· intro h
rcases hqf.exists_pair_ne with ⟨⟨x⟩, ⟨y⟩, hxy⟩
exact hxy (Ideal.Quotient.eq.2 (mul_one (x - y) ▸ I.mul_mem_left _ h))
· intro J x hIJ hxnI hxJ
rcases hqf.mul_inv_cancel (mt Ideal.Quotient.eq_zero_iff_mem.1 hxnI) with ⟨⟨y⟩, hy⟩
rw [← zero_add (1 : R), ← sub_self (x * y), sub_add]
refine' J.sub_mem (J.mul_mem_right _ hxJ) (hIJ (Ideal.Quotient.eq.1 hy))
#align ideal.quotient.maximal_of_is_field Ideal.Quotient.maximal_of_isField
/-- The quotient of a ring by an ideal is a field iff the ideal is maximal. -/
theorem maximal_ideal_iff_isField_quotient (I : Ideal R) : I.IsMaximal ↔ IsField (R ⧸ I) :=
⟨fun h =>
let _i := @Quotient.field _ _ I h
Field.toIsField _,
maximal_of_isField _⟩
#align ideal.quotient.maximal_ideal_iff_is_field_quotient Ideal.Quotient.maximal_ideal_iff_isField_quotient
variable [CommRing S]
/-- Given a ring homomorphism `f : R →+* S` sending all elements of an ideal to zero,
lift it to the quotient by this ideal. -/
def lift (I : Ideal R) (f : R →+* S) (H : ∀ a : R, a ∈ I → f a = 0) : R ⧸ I →+* S :=
{ QuotientAddGroup.lift I.toAddSubgroup f.toAddMonoidHom H with
map_one' := f.map_one
map_mul' := fun a₁ a₂ => Quotient.inductionOn₂' a₁ a₂ f.map_mul }
#align ideal.quotient.lift Ideal.Quotient.lift
@[simp]
theorem lift_mk (I : Ideal R) (f : R →+* S) (H : ∀ a : R, a ∈ I → f a = 0) :
lift I f H (mk I a) = f a :=
rfl
#align ideal.quotient.lift_mk Ideal.Quotient.lift_mk
theorem lift_surjective_of_surjective (I : Ideal R) {f : R →+* S} (H : ∀ a : R, a ∈ I → f a = 0)
(hf : Function.Surjective f) : Function.Surjective (Ideal.Quotient.lift I f H) := by
intro y
obtain ⟨x, rfl⟩ := hf y
use Ideal.Quotient.mk I x
simp only [Ideal.Quotient.lift_mk]
#align ideal.quotient.lift_surjective_of_surjective Ideal.Quotient.lift_surjective_of_surjective
/-- The ring homomorphism from the quotient by a smaller ideal to the quotient by a larger ideal.
This is the `Ideal.Quotient` version of `Quot.Factor` -/
def factor (S T : Ideal R) (H : S ≤ T) : R ⧸ S →+* R ⧸ T :=
Ideal.Quotient.lift S (mk T) fun _ hx => eq_zero_iff_mem.2 (H hx)
#align ideal.quotient.factor Ideal.Quotient.factor
@[simp]
theorem factor_mk (S T : Ideal R) (H : S ≤ T) (x : R) : factor S T H (mk S x) = mk T x :=
rfl
#align ideal.quotient.factor_mk Ideal.Quotient.factor_mk
@[simp]
theorem factor_comp_mk (S T : Ideal R) (H : S ≤ T) : (factor S T H).comp (mk S) = mk T := by
ext x
rw [RingHom.comp_apply, factor_mk]
#align ideal.quotient.factor_comp_mk Ideal.Quotient.factor_comp_mk
end Quotient
/-- Quotienting by equal ideals gives equivalent rings.
See also `Submodule.quotEquivOfEq` and `Ideal.quotientEquivAlgOfEq`.
-/
def quotEquivOfEq {R : Type*} [CommRing R] {I J : Ideal R} (h : I = J) : R ⧸ I ≃+* R ⧸ J :=
{ Submodule.quotEquivOfEq I J h with
map_mul' := by
rintro ⟨x⟩ ⟨y⟩
rfl }
#align ideal.quot_equiv_of_eq Ideal.quotEquivOfEq
@[simp]
theorem quotEquivOfEq_mk {R : Type*} [CommRing R] {I J : Ideal R} (h : I = J) (x : R) :
quotEquivOfEq h (Ideal.Quotient.mk I x) = Ideal.Quotient.mk J x :=
rfl
#align ideal.quot_equiv_of_eq_mk Ideal.quotEquivOfEq_mk
@[simp]
theorem quotEquivOfEq_symm {R : Type*} [CommRing R] {I J : Ideal R} (h : I = J) :
(Ideal.quotEquivOfEq h).symm = Ideal.quotEquivOfEq h.symm := by ext; rfl
#align ideal.quot_equiv_of_eq_symm Ideal.quotEquivOfEq_symm
section Pi
variable (ι : Type v)
/-- `R^n/I^n` is a `R/I`-module. -/
instance modulePi : Module (R ⧸ I) ((ι → R) ⧸ I.pi ι) where
smul c m :=
Quotient.liftOn₂' c m (fun r m => Submodule.Quotient.mk <| r • m) $ by
intro c₁ m₁ c₂ m₂ hc hm
apply Ideal.Quotient.eq.2
rw [Submodule.quotientRel_r_def] at hc hm
intro i
exact I.mul_sub_mul_mem hc (hm i)
one_smul := by
rintro ⟨a⟩
convert_to Ideal.Quotient.mk (I.pi ι) _ = Ideal.Quotient.mk (I.pi ι) _
congr with i; exact one_mul (a i)
mul_smul := by
rintro ⟨a⟩ ⟨b⟩ ⟨c⟩
convert_to Ideal.Quotient.mk (I.pi ι) _ = Ideal.Quotient.mk (I.pi ι) _
congr 1; funext i; exact mul_assoc a b (c i)
smul_add := by
rintro ⟨a⟩ ⟨b⟩ ⟨c⟩
convert_to Ideal.Quotient.mk (I.pi ι) _ = Ideal.Quotient.mk (I.pi ι) _
congr with i; exact mul_add a (b i) (c i)
smul_zero := by
rintro ⟨a⟩
convert_to Ideal.Quotient.mk (I.pi ι) _ = Ideal.Quotient.mk (I.pi ι) _
congr with _; exact mul_zero a
add_smul := by
rintro ⟨a⟩ ⟨b⟩ ⟨c⟩
convert_to Ideal.Quotient.mk (I.pi ι) _ = Ideal.Quotient.mk (I.pi ι) _
congr with i; exact add_mul a b (c i)
zero_smul := by
rintro ⟨a⟩
convert_to Ideal.Quotient.mk (I.pi ι) _ = Ideal.Quotient.mk (I.pi ι) _
congr with i; exact zero_mul (a i)
#align ideal.module_pi Ideal.modulePi
/-- `R^n/I^n` is isomorphic to `(R/I)^n` as an `R/I`-module. -/
noncomputable def piQuotEquiv : ((ι → R) ⧸ I.pi ι) ≃ₗ[R ⧸ I] ι → (R ⧸ I) where
toFun := fun x ↦
Quotient.liftOn' x (fun f i => Ideal.Quotient.mk I (f i)) fun a b hab =>
funext fun i => (Submodule.Quotient.eq' _).2 (QuotientAddGroup.leftRel_apply.mp hab i)
map_add' := by rintro ⟨_⟩ ⟨_⟩; rfl
map_smul' := by rintro ⟨_⟩ ⟨_⟩; rfl
invFun := fun x ↦ Ideal.Quotient.mk (I.pi ι) fun i ↦ Quotient.out' (x i)
left_inv := by
rintro ⟨x⟩
exact Ideal.Quotient.eq.2 fun i => Ideal.Quotient.eq.1 (Quotient.out_eq' _)
right_inv := by
intro x
ext i
obtain ⟨_, _⟩ := @Quot.exists_rep _ _ (x i)
|
convert Quotient.out_eq' (x i)
|
/-- `R^n/I^n` is isomorphic to `(R/I)^n` as an `R/I`-module. -/
noncomputable def piQuotEquiv : ((ι → R) ⧸ I.pi ι) ≃ₗ[R ⧸ I] ι → (R ⧸ I) where
toFun := fun x ↦
Quotient.liftOn' x (fun f i => Ideal.Quotient.mk I (f i)) fun a b hab =>
funext fun i => (Submodule.Quotient.eq' _).2 (QuotientAddGroup.leftRel_apply.mp hab i)
map_add' := by rintro ⟨_⟩ ⟨_⟩; rfl
map_smul' := by rintro ⟨_⟩ ⟨_⟩; rfl
invFun := fun x ↦ Ideal.Quotient.mk (I.pi ι) fun i ↦ Quotient.out' (x i)
left_inv := by
rintro ⟨x⟩
exact Ideal.Quotient.eq.2 fun i => Ideal.Quotient.eq.1 (Quotient.out_eq' _)
right_inv := by
intro x
ext i
obtain ⟨_, _⟩ := @Quot.exists_rep _ _ (x i)
|
Mathlib.RingTheory.Ideal.Quotient.359_0.TwNAv7Pc4PYOWjX
|
/-- `R^n/I^n` is isomorphic to `(R/I)^n` as an `R/I`-module. -/
noncomputable def piQuotEquiv : ((ι → R) ⧸ I.pi ι) ≃ₗ[R ⧸ I] ι → (R ⧸ I) where
toFun
|
Mathlib_RingTheory_Ideal_Quotient
|
R : Type u
inst✝¹ : CommRing R
I : Ideal R
a b : R
S ι✝ : Type v
ι : Type u_1
inst✝ : Finite ι
ι' : Type w
x : ι → R
hi : ∀ (i : ι), x i ∈ I
f : (ι → R) →ₗ[R] ι' → R
i : ι'
⊢ f x i ∈ I
|
/-
Copyright (c) 2018 Kenny Lau. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Kenny Lau, Chris Hughes, Mario Carneiro, Anne Baanen
-/
import Mathlib.Algebra.Ring.Fin
import Mathlib.Algebra.Ring.Prod
import Mathlib.LinearAlgebra.Quotient
import Mathlib.RingTheory.Congruence
import Mathlib.RingTheory.Ideal.Basic
import Mathlib.Tactic.FinCases
#align_import ring_theory.ideal.quotient from "leanprover-community/mathlib"@"949dc57e616a621462062668c9f39e4e17b64b69"
/-!
# Ideal quotients
This file defines ideal quotients as a special case of submodule quotients and proves some basic
results about these quotients.
See `Algebra.RingQuot` for quotients of non-commutative rings.
## Main definitions
- `Ideal.Quotient`: the quotient of a commutative ring `R` by an ideal `I : Ideal R`
-/
universe u v w
namespace Ideal
open Set
open BigOperators
variable {R : Type u} [CommRing R] (I : Ideal R) {a b : R}
variable {S : Type v}
-- Note that at present `Ideal` means a left-ideal,
-- so this quotient is only useful in a commutative ring.
-- We should develop quotients by two-sided ideals as well.
/-- The quotient `R/I` of a ring `R` by an ideal `I`.
The ideal quotient of `I` is defined to equal the quotient of `I` as an `R`-submodule of `R`.
This definition is marked `reducible` so that typeclass instances can be shared between
`Ideal.Quotient I` and `Submodule.Quotient I`.
-/
@[reducible]
instance : HasQuotient R (Ideal R) :=
Submodule.hasQuotient
namespace Quotient
variable {I} {x y : R}
instance one (I : Ideal R) : One (R ⧸ I) :=
⟨Submodule.Quotient.mk 1⟩
#align ideal.quotient.has_one Ideal.Quotient.one
/-- On `Ideal`s, `Submodule.quotientRel` is a ring congruence. -/
protected def ringCon (I : Ideal R) : RingCon R :=
{ QuotientAddGroup.con I.toAddSubgroup with
mul' := fun {a₁ b₁ a₂ b₂} h₁ h₂ => by
rw [Submodule.quotientRel_r_def] at h₁ h₂ ⊢
have F := I.add_mem (I.mul_mem_left a₂ h₁) (I.mul_mem_right b₁ h₂)
have : a₁ * a₂ - b₁ * b₂ = a₂ * (a₁ - b₁) + (a₂ - b₂) * b₁ := by
rw [mul_sub, sub_mul, sub_add_sub_cancel, mul_comm, mul_comm b₁]
rwa [← this] at F }
#align ideal.quotient.ring_con Ideal.Quotient.ringCon
instance commRing (I : Ideal R) : CommRing (R ⧸ I) :=
inferInstanceAs (CommRing (Quotient.ringCon I).Quotient)
#align ideal.quotient.comm_ring Ideal.Quotient.commRing
-- Sanity test to make sure no diamonds have emerged in `commRing`
example : (commRing I).toAddCommGroup = Submodule.Quotient.addCommGroup I := rfl
-- this instance is harder to find than the one via `Algebra α (R ⧸ I)`, so use a lower priority
instance (priority := 100) isScalarTower_right {α} [SMul α R] [IsScalarTower α R R] :
IsScalarTower α (R ⧸ I) (R ⧸ I) :=
(Quotient.ringCon I).isScalarTower_right
#align ideal.quotient.is_scalar_tower_right Ideal.Quotient.isScalarTower_right
instance smulCommClass {α} [SMul α R] [IsScalarTower α R R] [SMulCommClass α R R] :
SMulCommClass α (R ⧸ I) (R ⧸ I) :=
(Quotient.ringCon I).smulCommClass
#align ideal.quotient.smul_comm_class Ideal.Quotient.smulCommClass
instance smulCommClass' {α} [SMul α R] [IsScalarTower α R R] [SMulCommClass R α R] :
SMulCommClass (R ⧸ I) α (R ⧸ I) :=
(Quotient.ringCon I).smulCommClass'
#align ideal.quotient.smul_comm_class' Ideal.Quotient.smulCommClass'
/-- The ring homomorphism from a ring `R` to a quotient ring `R/I`. -/
def mk (I : Ideal R) : R →+* R ⧸ I where
toFun a := Submodule.Quotient.mk a
map_zero' := rfl
map_one' := rfl
map_mul' _ _ := rfl
map_add' _ _ := rfl
#align ideal.quotient.mk Ideal.Quotient.mk
instance {I : Ideal R} : Coe R (R ⧸ I) :=
⟨Ideal.Quotient.mk I⟩
/-- Two `RingHom`s from the quotient by an ideal are equal if their
compositions with `Ideal.Quotient.mk'` are equal.
See note [partially-applied ext lemmas]. -/
@[ext 1100]
theorem ringHom_ext [NonAssocSemiring S] ⦃f g : R ⧸ I →+* S⦄ (h : f.comp (mk I) = g.comp (mk I)) :
f = g :=
RingHom.ext fun x => Quotient.inductionOn' x <| (RingHom.congr_fun h : _)
#align ideal.quotient.ring_hom_ext Ideal.Quotient.ringHom_ext
instance inhabited : Inhabited (R ⧸ I) :=
⟨mk I 37⟩
#align ideal.quotient.inhabited Ideal.Quotient.inhabited
protected theorem eq : mk I x = mk I y ↔ x - y ∈ I :=
Submodule.Quotient.eq I
#align ideal.quotient.eq Ideal.Quotient.eq
@[simp]
theorem mk_eq_mk (x : R) : (Submodule.Quotient.mk x : R ⧸ I) = mk I x := rfl
#align ideal.quotient.mk_eq_mk Ideal.Quotient.mk_eq_mk
theorem eq_zero_iff_mem {I : Ideal R} : mk I a = 0 ↔ a ∈ I :=
Submodule.Quotient.mk_eq_zero _
#align ideal.quotient.eq_zero_iff_mem Ideal.Quotient.eq_zero_iff_mem
theorem eq_zero_iff_dvd (x y : R) : Ideal.Quotient.mk (Ideal.span ({x} : Set R)) y = 0 ↔ x ∣ y := by
rw [Ideal.Quotient.eq_zero_iff_mem, Ideal.mem_span_singleton]
-- Porting note: new theorem
theorem mk_eq_mk_iff_sub_mem (x y : R) : mk I x = mk I y ↔ x - y ∈ I := by
rw [← eq_zero_iff_mem, map_sub, sub_eq_zero]
theorem zero_eq_one_iff {I : Ideal R} : (0 : R ⧸ I) = 1 ↔ I = ⊤ :=
eq_comm.trans <| eq_zero_iff_mem.trans (eq_top_iff_one _).symm
#align ideal.quotient.zero_eq_one_iff Ideal.Quotient.zero_eq_one_iff
theorem zero_ne_one_iff {I : Ideal R} : (0 : R ⧸ I) ≠ 1 ↔ I ≠ ⊤ :=
not_congr zero_eq_one_iff
#align ideal.quotient.zero_ne_one_iff Ideal.Quotient.zero_ne_one_iff
protected theorem nontrivial {I : Ideal R} (hI : I ≠ ⊤) : Nontrivial (R ⧸ I) :=
⟨⟨0, 1, zero_ne_one_iff.2 hI⟩⟩
#align ideal.quotient.nontrivial Ideal.Quotient.nontrivial
theorem subsingleton_iff {I : Ideal R} : Subsingleton (R ⧸ I) ↔ I = ⊤ := by
rw [eq_top_iff_one, ← subsingleton_iff_zero_eq_one, eq_comm, ← (mk I).map_one,
Quotient.eq_zero_iff_mem]
#align ideal.quotient.subsingleton_iff Ideal.Quotient.subsingleton_iff
instance : Unique (R ⧸ (⊤ : Ideal R)) :=
⟨⟨0⟩, by rintro ⟨x⟩; exact Quotient.eq_zero_iff_mem.mpr Submodule.mem_top⟩
theorem mk_surjective : Function.Surjective (mk I) := fun y =>
Quotient.inductionOn' y fun x => Exists.intro x rfl
#align ideal.quotient.mk_surjective Ideal.Quotient.mk_surjective
instance : RingHomSurjective (mk I) :=
⟨mk_surjective⟩
/-- If `I` is an ideal of a commutative ring `R`, if `q : R → R/I` is the quotient map, and if
`s ⊆ R` is a subset, then `q⁻¹(q(s)) = ⋃ᵢ(i + s)`, the union running over all `i ∈ I`. -/
theorem quotient_ring_saturate (I : Ideal R) (s : Set R) :
mk I ⁻¹' (mk I '' s) = ⋃ x : I, (fun y => x.1 + y) '' s := by
ext x
simp only [mem_preimage, mem_image, mem_iUnion, Ideal.Quotient.eq]
exact
⟨fun ⟨a, a_in, h⟩ => ⟨⟨_, I.neg_mem h⟩, a, a_in, by simp⟩, fun ⟨⟨i, hi⟩, a, ha, Eq⟩ =>
⟨a, ha, by rw [← Eq, sub_add_eq_sub_sub_swap, sub_self, zero_sub]; exact I.neg_mem hi⟩⟩
#align ideal.quotient.quotient_ring_saturate Ideal.Quotient.quotient_ring_saturate
instance noZeroDivisors (I : Ideal R) [hI : I.IsPrime] : NoZeroDivisors (R ⧸ I) where
eq_zero_or_eq_zero_of_mul_eq_zero {a b} := Quotient.inductionOn₂' a b fun {_ _} hab =>
(hI.mem_or_mem (eq_zero_iff_mem.1 hab)).elim (Or.inl ∘ eq_zero_iff_mem.2)
(Or.inr ∘ eq_zero_iff_mem.2)
#align ideal.quotient.no_zero_divisors Ideal.Quotient.noZeroDivisors
instance isDomain (I : Ideal R) [hI : I.IsPrime] : IsDomain (R ⧸ I) :=
let _ := Quotient.nontrivial hI.1
NoZeroDivisors.to_isDomain _
#align ideal.quotient.is_domain Ideal.Quotient.isDomain
theorem isDomain_iff_prime (I : Ideal R) : IsDomain (R ⧸ I) ↔ I.IsPrime := by
refine' ⟨fun H => ⟨zero_ne_one_iff.1 _, fun {x y} h => _⟩, fun h => inferInstance⟩
· haveI : Nontrivial (R ⧸ I) := ⟨H.2.1⟩
exact zero_ne_one
· simp only [← eq_zero_iff_mem, (mk I).map_mul] at h ⊢
haveI := @IsDomain.to_noZeroDivisors (R ⧸ I) _ H
exact eq_zero_or_eq_zero_of_mul_eq_zero h
#align ideal.quotient.is_domain_iff_prime Ideal.Quotient.isDomain_iff_prime
theorem exists_inv {I : Ideal R} [hI : I.IsMaximal] :
∀ {a : R ⧸ I}, a ≠ 0 → ∃ b : R ⧸ I, a * b = 1 := by
rintro ⟨a⟩ h
rcases hI.exists_inv (mt eq_zero_iff_mem.2 h) with ⟨b, c, hc, abc⟩
rw [mul_comm] at abc
refine' ⟨mk _ b, Quot.sound _⟩
simp only [Submodule.quotientRel_r_def]
rw [← eq_sub_iff_add_eq'] at abc
rwa [abc, ← neg_mem_iff (G := R) (H := I), neg_sub] at hc
#align ideal.quotient.exists_inv Ideal.Quotient.exists_inv
open Classical
/-- The quotient by a maximal ideal is a group with zero. This is a `def` rather than `instance`,
since users will have computable inverses in some applications.
See note [reducible non-instances]. -/
@[reducible]
protected noncomputable def groupWithZero (I : Ideal R) [hI : I.IsMaximal] :
GroupWithZero (R ⧸ I) :=
{ inv := fun a => if ha : a = 0 then 0 else Classical.choose (exists_inv ha)
mul_inv_cancel := fun a (ha : a ≠ 0) =>
show a * dite _ _ _ = _ by rw [dif_neg ha]; exact Classical.choose_spec (exists_inv ha)
inv_zero := dif_pos rfl }
#align ideal.quotient.group_with_zero Ideal.Quotient.groupWithZero
/-- The quotient by a maximal ideal is a field. This is a `def` rather than `instance`, since users
will have computable inverses (and `qsmul`, `rat_cast`) in some applications.
See note [reducible non-instances]. -/
@[reducible]
protected noncomputable def field (I : Ideal R) [hI : I.IsMaximal] : Field (R ⧸ I) :=
{ Quotient.commRing I, Quotient.groupWithZero I with }
#align ideal.quotient.field Ideal.Quotient.field
/-- If the quotient by an ideal is a field, then the ideal is maximal. -/
theorem maximal_of_isField (I : Ideal R) (hqf : IsField (R ⧸ I)) : I.IsMaximal := by
apply Ideal.isMaximal_iff.2
constructor
· intro h
rcases hqf.exists_pair_ne with ⟨⟨x⟩, ⟨y⟩, hxy⟩
exact hxy (Ideal.Quotient.eq.2 (mul_one (x - y) ▸ I.mul_mem_left _ h))
· intro J x hIJ hxnI hxJ
rcases hqf.mul_inv_cancel (mt Ideal.Quotient.eq_zero_iff_mem.1 hxnI) with ⟨⟨y⟩, hy⟩
rw [← zero_add (1 : R), ← sub_self (x * y), sub_add]
refine' J.sub_mem (J.mul_mem_right _ hxJ) (hIJ (Ideal.Quotient.eq.1 hy))
#align ideal.quotient.maximal_of_is_field Ideal.Quotient.maximal_of_isField
/-- The quotient of a ring by an ideal is a field iff the ideal is maximal. -/
theorem maximal_ideal_iff_isField_quotient (I : Ideal R) : I.IsMaximal ↔ IsField (R ⧸ I) :=
⟨fun h =>
let _i := @Quotient.field _ _ I h
Field.toIsField _,
maximal_of_isField _⟩
#align ideal.quotient.maximal_ideal_iff_is_field_quotient Ideal.Quotient.maximal_ideal_iff_isField_quotient
variable [CommRing S]
/-- Given a ring homomorphism `f : R →+* S` sending all elements of an ideal to zero,
lift it to the quotient by this ideal. -/
def lift (I : Ideal R) (f : R →+* S) (H : ∀ a : R, a ∈ I → f a = 0) : R ⧸ I →+* S :=
{ QuotientAddGroup.lift I.toAddSubgroup f.toAddMonoidHom H with
map_one' := f.map_one
map_mul' := fun a₁ a₂ => Quotient.inductionOn₂' a₁ a₂ f.map_mul }
#align ideal.quotient.lift Ideal.Quotient.lift
@[simp]
theorem lift_mk (I : Ideal R) (f : R →+* S) (H : ∀ a : R, a ∈ I → f a = 0) :
lift I f H (mk I a) = f a :=
rfl
#align ideal.quotient.lift_mk Ideal.Quotient.lift_mk
theorem lift_surjective_of_surjective (I : Ideal R) {f : R →+* S} (H : ∀ a : R, a ∈ I → f a = 0)
(hf : Function.Surjective f) : Function.Surjective (Ideal.Quotient.lift I f H) := by
intro y
obtain ⟨x, rfl⟩ := hf y
use Ideal.Quotient.mk I x
simp only [Ideal.Quotient.lift_mk]
#align ideal.quotient.lift_surjective_of_surjective Ideal.Quotient.lift_surjective_of_surjective
/-- The ring homomorphism from the quotient by a smaller ideal to the quotient by a larger ideal.
This is the `Ideal.Quotient` version of `Quot.Factor` -/
def factor (S T : Ideal R) (H : S ≤ T) : R ⧸ S →+* R ⧸ T :=
Ideal.Quotient.lift S (mk T) fun _ hx => eq_zero_iff_mem.2 (H hx)
#align ideal.quotient.factor Ideal.Quotient.factor
@[simp]
theorem factor_mk (S T : Ideal R) (H : S ≤ T) (x : R) : factor S T H (mk S x) = mk T x :=
rfl
#align ideal.quotient.factor_mk Ideal.Quotient.factor_mk
@[simp]
theorem factor_comp_mk (S T : Ideal R) (H : S ≤ T) : (factor S T H).comp (mk S) = mk T := by
ext x
rw [RingHom.comp_apply, factor_mk]
#align ideal.quotient.factor_comp_mk Ideal.Quotient.factor_comp_mk
end Quotient
/-- Quotienting by equal ideals gives equivalent rings.
See also `Submodule.quotEquivOfEq` and `Ideal.quotientEquivAlgOfEq`.
-/
def quotEquivOfEq {R : Type*} [CommRing R] {I J : Ideal R} (h : I = J) : R ⧸ I ≃+* R ⧸ J :=
{ Submodule.quotEquivOfEq I J h with
map_mul' := by
rintro ⟨x⟩ ⟨y⟩
rfl }
#align ideal.quot_equiv_of_eq Ideal.quotEquivOfEq
@[simp]
theorem quotEquivOfEq_mk {R : Type*} [CommRing R] {I J : Ideal R} (h : I = J) (x : R) :
quotEquivOfEq h (Ideal.Quotient.mk I x) = Ideal.Quotient.mk J x :=
rfl
#align ideal.quot_equiv_of_eq_mk Ideal.quotEquivOfEq_mk
@[simp]
theorem quotEquivOfEq_symm {R : Type*} [CommRing R] {I J : Ideal R} (h : I = J) :
(Ideal.quotEquivOfEq h).symm = Ideal.quotEquivOfEq h.symm := by ext; rfl
#align ideal.quot_equiv_of_eq_symm Ideal.quotEquivOfEq_symm
section Pi
variable (ι : Type v)
/-- `R^n/I^n` is a `R/I`-module. -/
instance modulePi : Module (R ⧸ I) ((ι → R) ⧸ I.pi ι) where
smul c m :=
Quotient.liftOn₂' c m (fun r m => Submodule.Quotient.mk <| r • m) $ by
intro c₁ m₁ c₂ m₂ hc hm
apply Ideal.Quotient.eq.2
rw [Submodule.quotientRel_r_def] at hc hm
intro i
exact I.mul_sub_mul_mem hc (hm i)
one_smul := by
rintro ⟨a⟩
convert_to Ideal.Quotient.mk (I.pi ι) _ = Ideal.Quotient.mk (I.pi ι) _
congr with i; exact one_mul (a i)
mul_smul := by
rintro ⟨a⟩ ⟨b⟩ ⟨c⟩
convert_to Ideal.Quotient.mk (I.pi ι) _ = Ideal.Quotient.mk (I.pi ι) _
congr 1; funext i; exact mul_assoc a b (c i)
smul_add := by
rintro ⟨a⟩ ⟨b⟩ ⟨c⟩
convert_to Ideal.Quotient.mk (I.pi ι) _ = Ideal.Quotient.mk (I.pi ι) _
congr with i; exact mul_add a (b i) (c i)
smul_zero := by
rintro ⟨a⟩
convert_to Ideal.Quotient.mk (I.pi ι) _ = Ideal.Quotient.mk (I.pi ι) _
congr with _; exact mul_zero a
add_smul := by
rintro ⟨a⟩ ⟨b⟩ ⟨c⟩
convert_to Ideal.Quotient.mk (I.pi ι) _ = Ideal.Quotient.mk (I.pi ι) _
congr with i; exact add_mul a b (c i)
zero_smul := by
rintro ⟨a⟩
convert_to Ideal.Quotient.mk (I.pi ι) _ = Ideal.Quotient.mk (I.pi ι) _
congr with i; exact zero_mul (a i)
#align ideal.module_pi Ideal.modulePi
/-- `R^n/I^n` is isomorphic to `(R/I)^n` as an `R/I`-module. -/
noncomputable def piQuotEquiv : ((ι → R) ⧸ I.pi ι) ≃ₗ[R ⧸ I] ι → (R ⧸ I) where
toFun := fun x ↦
Quotient.liftOn' x (fun f i => Ideal.Quotient.mk I (f i)) fun a b hab =>
funext fun i => (Submodule.Quotient.eq' _).2 (QuotientAddGroup.leftRel_apply.mp hab i)
map_add' := by rintro ⟨_⟩ ⟨_⟩; rfl
map_smul' := by rintro ⟨_⟩ ⟨_⟩; rfl
invFun := fun x ↦ Ideal.Quotient.mk (I.pi ι) fun i ↦ Quotient.out' (x i)
left_inv := by
rintro ⟨x⟩
exact Ideal.Quotient.eq.2 fun i => Ideal.Quotient.eq.1 (Quotient.out_eq' _)
right_inv := by
intro x
ext i
obtain ⟨_, _⟩ := @Quot.exists_rep _ _ (x i)
convert Quotient.out_eq' (x i)
#align ideal.pi_quot_equiv Ideal.piQuotEquiv
/-- If `f : R^n → R^m` is an `R`-linear map and `I ⊆ R` is an ideal, then the image of `I^n` is
contained in `I^m`. -/
theorem map_pi {ι : Type*} [Finite ι] {ι' : Type w} (x : ι → R) (hi : ∀ i, x i ∈ I)
(f : (ι → R) →ₗ[R] ι' → R) (i : ι') : f x i ∈ I := by
|
classical
cases nonempty_fintype ι
rw [pi_eq_sum_univ x]
simp only [Finset.sum_apply, smul_eq_mul, map_sum, Pi.smul_apply, map_smul]
exact I.sum_mem fun j _ => I.mul_mem_right _ (hi j)
|
/-- If `f : R^n → R^m` is an `R`-linear map and `I ⊆ R` is an ideal, then the image of `I^n` is
contained in `I^m`. -/
theorem map_pi {ι : Type*} [Finite ι] {ι' : Type w} (x : ι → R) (hi : ∀ i, x i ∈ I)
(f : (ι → R) →ₗ[R] ι' → R) (i : ι') : f x i ∈ I := by
|
Mathlib.RingTheory.Ideal.Quotient.377_0.TwNAv7Pc4PYOWjX
|
/-- If `f : R^n → R^m` is an `R`-linear map and `I ⊆ R` is an ideal, then the image of `I^n` is
contained in `I^m`. -/
theorem map_pi {ι : Type*} [Finite ι] {ι' : Type w} (x : ι → R) (hi : ∀ i, x i ∈ I)
(f : (ι → R) →ₗ[R] ι' → R) (i : ι') : f x i ∈ I
|
Mathlib_RingTheory_Ideal_Quotient
|
R : Type u
inst✝¹ : CommRing R
I : Ideal R
a b : R
S ι✝ : Type v
ι : Type u_1
inst✝ : Finite ι
ι' : Type w
x : ι → R
hi : ∀ (i : ι), x i ∈ I
f : (ι → R) →ₗ[R] ι' → R
i : ι'
⊢ f x i ∈ I
|
/-
Copyright (c) 2018 Kenny Lau. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Kenny Lau, Chris Hughes, Mario Carneiro, Anne Baanen
-/
import Mathlib.Algebra.Ring.Fin
import Mathlib.Algebra.Ring.Prod
import Mathlib.LinearAlgebra.Quotient
import Mathlib.RingTheory.Congruence
import Mathlib.RingTheory.Ideal.Basic
import Mathlib.Tactic.FinCases
#align_import ring_theory.ideal.quotient from "leanprover-community/mathlib"@"949dc57e616a621462062668c9f39e4e17b64b69"
/-!
# Ideal quotients
This file defines ideal quotients as a special case of submodule quotients and proves some basic
results about these quotients.
See `Algebra.RingQuot` for quotients of non-commutative rings.
## Main definitions
- `Ideal.Quotient`: the quotient of a commutative ring `R` by an ideal `I : Ideal R`
-/
universe u v w
namespace Ideal
open Set
open BigOperators
variable {R : Type u} [CommRing R] (I : Ideal R) {a b : R}
variable {S : Type v}
-- Note that at present `Ideal` means a left-ideal,
-- so this quotient is only useful in a commutative ring.
-- We should develop quotients by two-sided ideals as well.
/-- The quotient `R/I` of a ring `R` by an ideal `I`.
The ideal quotient of `I` is defined to equal the quotient of `I` as an `R`-submodule of `R`.
This definition is marked `reducible` so that typeclass instances can be shared between
`Ideal.Quotient I` and `Submodule.Quotient I`.
-/
@[reducible]
instance : HasQuotient R (Ideal R) :=
Submodule.hasQuotient
namespace Quotient
variable {I} {x y : R}
instance one (I : Ideal R) : One (R ⧸ I) :=
⟨Submodule.Quotient.mk 1⟩
#align ideal.quotient.has_one Ideal.Quotient.one
/-- On `Ideal`s, `Submodule.quotientRel` is a ring congruence. -/
protected def ringCon (I : Ideal R) : RingCon R :=
{ QuotientAddGroup.con I.toAddSubgroup with
mul' := fun {a₁ b₁ a₂ b₂} h₁ h₂ => by
rw [Submodule.quotientRel_r_def] at h₁ h₂ ⊢
have F := I.add_mem (I.mul_mem_left a₂ h₁) (I.mul_mem_right b₁ h₂)
have : a₁ * a₂ - b₁ * b₂ = a₂ * (a₁ - b₁) + (a₂ - b₂) * b₁ := by
rw [mul_sub, sub_mul, sub_add_sub_cancel, mul_comm, mul_comm b₁]
rwa [← this] at F }
#align ideal.quotient.ring_con Ideal.Quotient.ringCon
instance commRing (I : Ideal R) : CommRing (R ⧸ I) :=
inferInstanceAs (CommRing (Quotient.ringCon I).Quotient)
#align ideal.quotient.comm_ring Ideal.Quotient.commRing
-- Sanity test to make sure no diamonds have emerged in `commRing`
example : (commRing I).toAddCommGroup = Submodule.Quotient.addCommGroup I := rfl
-- this instance is harder to find than the one via `Algebra α (R ⧸ I)`, so use a lower priority
instance (priority := 100) isScalarTower_right {α} [SMul α R] [IsScalarTower α R R] :
IsScalarTower α (R ⧸ I) (R ⧸ I) :=
(Quotient.ringCon I).isScalarTower_right
#align ideal.quotient.is_scalar_tower_right Ideal.Quotient.isScalarTower_right
instance smulCommClass {α} [SMul α R] [IsScalarTower α R R] [SMulCommClass α R R] :
SMulCommClass α (R ⧸ I) (R ⧸ I) :=
(Quotient.ringCon I).smulCommClass
#align ideal.quotient.smul_comm_class Ideal.Quotient.smulCommClass
instance smulCommClass' {α} [SMul α R] [IsScalarTower α R R] [SMulCommClass R α R] :
SMulCommClass (R ⧸ I) α (R ⧸ I) :=
(Quotient.ringCon I).smulCommClass'
#align ideal.quotient.smul_comm_class' Ideal.Quotient.smulCommClass'
/-- The ring homomorphism from a ring `R` to a quotient ring `R/I`. -/
def mk (I : Ideal R) : R →+* R ⧸ I where
toFun a := Submodule.Quotient.mk a
map_zero' := rfl
map_one' := rfl
map_mul' _ _ := rfl
map_add' _ _ := rfl
#align ideal.quotient.mk Ideal.Quotient.mk
instance {I : Ideal R} : Coe R (R ⧸ I) :=
⟨Ideal.Quotient.mk I⟩
/-- Two `RingHom`s from the quotient by an ideal are equal if their
compositions with `Ideal.Quotient.mk'` are equal.
See note [partially-applied ext lemmas]. -/
@[ext 1100]
theorem ringHom_ext [NonAssocSemiring S] ⦃f g : R ⧸ I →+* S⦄ (h : f.comp (mk I) = g.comp (mk I)) :
f = g :=
RingHom.ext fun x => Quotient.inductionOn' x <| (RingHom.congr_fun h : _)
#align ideal.quotient.ring_hom_ext Ideal.Quotient.ringHom_ext
instance inhabited : Inhabited (R ⧸ I) :=
⟨mk I 37⟩
#align ideal.quotient.inhabited Ideal.Quotient.inhabited
protected theorem eq : mk I x = mk I y ↔ x - y ∈ I :=
Submodule.Quotient.eq I
#align ideal.quotient.eq Ideal.Quotient.eq
@[simp]
theorem mk_eq_mk (x : R) : (Submodule.Quotient.mk x : R ⧸ I) = mk I x := rfl
#align ideal.quotient.mk_eq_mk Ideal.Quotient.mk_eq_mk
theorem eq_zero_iff_mem {I : Ideal R} : mk I a = 0 ↔ a ∈ I :=
Submodule.Quotient.mk_eq_zero _
#align ideal.quotient.eq_zero_iff_mem Ideal.Quotient.eq_zero_iff_mem
theorem eq_zero_iff_dvd (x y : R) : Ideal.Quotient.mk (Ideal.span ({x} : Set R)) y = 0 ↔ x ∣ y := by
rw [Ideal.Quotient.eq_zero_iff_mem, Ideal.mem_span_singleton]
-- Porting note: new theorem
theorem mk_eq_mk_iff_sub_mem (x y : R) : mk I x = mk I y ↔ x - y ∈ I := by
rw [← eq_zero_iff_mem, map_sub, sub_eq_zero]
theorem zero_eq_one_iff {I : Ideal R} : (0 : R ⧸ I) = 1 ↔ I = ⊤ :=
eq_comm.trans <| eq_zero_iff_mem.trans (eq_top_iff_one _).symm
#align ideal.quotient.zero_eq_one_iff Ideal.Quotient.zero_eq_one_iff
theorem zero_ne_one_iff {I : Ideal R} : (0 : R ⧸ I) ≠ 1 ↔ I ≠ ⊤ :=
not_congr zero_eq_one_iff
#align ideal.quotient.zero_ne_one_iff Ideal.Quotient.zero_ne_one_iff
protected theorem nontrivial {I : Ideal R} (hI : I ≠ ⊤) : Nontrivial (R ⧸ I) :=
⟨⟨0, 1, zero_ne_one_iff.2 hI⟩⟩
#align ideal.quotient.nontrivial Ideal.Quotient.nontrivial
theorem subsingleton_iff {I : Ideal R} : Subsingleton (R ⧸ I) ↔ I = ⊤ := by
rw [eq_top_iff_one, ← subsingleton_iff_zero_eq_one, eq_comm, ← (mk I).map_one,
Quotient.eq_zero_iff_mem]
#align ideal.quotient.subsingleton_iff Ideal.Quotient.subsingleton_iff
instance : Unique (R ⧸ (⊤ : Ideal R)) :=
⟨⟨0⟩, by rintro ⟨x⟩; exact Quotient.eq_zero_iff_mem.mpr Submodule.mem_top⟩
theorem mk_surjective : Function.Surjective (mk I) := fun y =>
Quotient.inductionOn' y fun x => Exists.intro x rfl
#align ideal.quotient.mk_surjective Ideal.Quotient.mk_surjective
instance : RingHomSurjective (mk I) :=
⟨mk_surjective⟩
/-- If `I` is an ideal of a commutative ring `R`, if `q : R → R/I` is the quotient map, and if
`s ⊆ R` is a subset, then `q⁻¹(q(s)) = ⋃ᵢ(i + s)`, the union running over all `i ∈ I`. -/
theorem quotient_ring_saturate (I : Ideal R) (s : Set R) :
mk I ⁻¹' (mk I '' s) = ⋃ x : I, (fun y => x.1 + y) '' s := by
ext x
simp only [mem_preimage, mem_image, mem_iUnion, Ideal.Quotient.eq]
exact
⟨fun ⟨a, a_in, h⟩ => ⟨⟨_, I.neg_mem h⟩, a, a_in, by simp⟩, fun ⟨⟨i, hi⟩, a, ha, Eq⟩ =>
⟨a, ha, by rw [← Eq, sub_add_eq_sub_sub_swap, sub_self, zero_sub]; exact I.neg_mem hi⟩⟩
#align ideal.quotient.quotient_ring_saturate Ideal.Quotient.quotient_ring_saturate
instance noZeroDivisors (I : Ideal R) [hI : I.IsPrime] : NoZeroDivisors (R ⧸ I) where
eq_zero_or_eq_zero_of_mul_eq_zero {a b} := Quotient.inductionOn₂' a b fun {_ _} hab =>
(hI.mem_or_mem (eq_zero_iff_mem.1 hab)).elim (Or.inl ∘ eq_zero_iff_mem.2)
(Or.inr ∘ eq_zero_iff_mem.2)
#align ideal.quotient.no_zero_divisors Ideal.Quotient.noZeroDivisors
instance isDomain (I : Ideal R) [hI : I.IsPrime] : IsDomain (R ⧸ I) :=
let _ := Quotient.nontrivial hI.1
NoZeroDivisors.to_isDomain _
#align ideal.quotient.is_domain Ideal.Quotient.isDomain
theorem isDomain_iff_prime (I : Ideal R) : IsDomain (R ⧸ I) ↔ I.IsPrime := by
refine' ⟨fun H => ⟨zero_ne_one_iff.1 _, fun {x y} h => _⟩, fun h => inferInstance⟩
· haveI : Nontrivial (R ⧸ I) := ⟨H.2.1⟩
exact zero_ne_one
· simp only [← eq_zero_iff_mem, (mk I).map_mul] at h ⊢
haveI := @IsDomain.to_noZeroDivisors (R ⧸ I) _ H
exact eq_zero_or_eq_zero_of_mul_eq_zero h
#align ideal.quotient.is_domain_iff_prime Ideal.Quotient.isDomain_iff_prime
theorem exists_inv {I : Ideal R} [hI : I.IsMaximal] :
∀ {a : R ⧸ I}, a ≠ 0 → ∃ b : R ⧸ I, a * b = 1 := by
rintro ⟨a⟩ h
rcases hI.exists_inv (mt eq_zero_iff_mem.2 h) with ⟨b, c, hc, abc⟩
rw [mul_comm] at abc
refine' ⟨mk _ b, Quot.sound _⟩
simp only [Submodule.quotientRel_r_def]
rw [← eq_sub_iff_add_eq'] at abc
rwa [abc, ← neg_mem_iff (G := R) (H := I), neg_sub] at hc
#align ideal.quotient.exists_inv Ideal.Quotient.exists_inv
open Classical
/-- The quotient by a maximal ideal is a group with zero. This is a `def` rather than `instance`,
since users will have computable inverses in some applications.
See note [reducible non-instances]. -/
@[reducible]
protected noncomputable def groupWithZero (I : Ideal R) [hI : I.IsMaximal] :
GroupWithZero (R ⧸ I) :=
{ inv := fun a => if ha : a = 0 then 0 else Classical.choose (exists_inv ha)
mul_inv_cancel := fun a (ha : a ≠ 0) =>
show a * dite _ _ _ = _ by rw [dif_neg ha]; exact Classical.choose_spec (exists_inv ha)
inv_zero := dif_pos rfl }
#align ideal.quotient.group_with_zero Ideal.Quotient.groupWithZero
/-- The quotient by a maximal ideal is a field. This is a `def` rather than `instance`, since users
will have computable inverses (and `qsmul`, `rat_cast`) in some applications.
See note [reducible non-instances]. -/
@[reducible]
protected noncomputable def field (I : Ideal R) [hI : I.IsMaximal] : Field (R ⧸ I) :=
{ Quotient.commRing I, Quotient.groupWithZero I with }
#align ideal.quotient.field Ideal.Quotient.field
/-- If the quotient by an ideal is a field, then the ideal is maximal. -/
theorem maximal_of_isField (I : Ideal R) (hqf : IsField (R ⧸ I)) : I.IsMaximal := by
apply Ideal.isMaximal_iff.2
constructor
· intro h
rcases hqf.exists_pair_ne with ⟨⟨x⟩, ⟨y⟩, hxy⟩
exact hxy (Ideal.Quotient.eq.2 (mul_one (x - y) ▸ I.mul_mem_left _ h))
· intro J x hIJ hxnI hxJ
rcases hqf.mul_inv_cancel (mt Ideal.Quotient.eq_zero_iff_mem.1 hxnI) with ⟨⟨y⟩, hy⟩
rw [← zero_add (1 : R), ← sub_self (x * y), sub_add]
refine' J.sub_mem (J.mul_mem_right _ hxJ) (hIJ (Ideal.Quotient.eq.1 hy))
#align ideal.quotient.maximal_of_is_field Ideal.Quotient.maximal_of_isField
/-- The quotient of a ring by an ideal is a field iff the ideal is maximal. -/
theorem maximal_ideal_iff_isField_quotient (I : Ideal R) : I.IsMaximal ↔ IsField (R ⧸ I) :=
⟨fun h =>
let _i := @Quotient.field _ _ I h
Field.toIsField _,
maximal_of_isField _⟩
#align ideal.quotient.maximal_ideal_iff_is_field_quotient Ideal.Quotient.maximal_ideal_iff_isField_quotient
variable [CommRing S]
/-- Given a ring homomorphism `f : R →+* S` sending all elements of an ideal to zero,
lift it to the quotient by this ideal. -/
def lift (I : Ideal R) (f : R →+* S) (H : ∀ a : R, a ∈ I → f a = 0) : R ⧸ I →+* S :=
{ QuotientAddGroup.lift I.toAddSubgroup f.toAddMonoidHom H with
map_one' := f.map_one
map_mul' := fun a₁ a₂ => Quotient.inductionOn₂' a₁ a₂ f.map_mul }
#align ideal.quotient.lift Ideal.Quotient.lift
@[simp]
theorem lift_mk (I : Ideal R) (f : R →+* S) (H : ∀ a : R, a ∈ I → f a = 0) :
lift I f H (mk I a) = f a :=
rfl
#align ideal.quotient.lift_mk Ideal.Quotient.lift_mk
theorem lift_surjective_of_surjective (I : Ideal R) {f : R →+* S} (H : ∀ a : R, a ∈ I → f a = 0)
(hf : Function.Surjective f) : Function.Surjective (Ideal.Quotient.lift I f H) := by
intro y
obtain ⟨x, rfl⟩ := hf y
use Ideal.Quotient.mk I x
simp only [Ideal.Quotient.lift_mk]
#align ideal.quotient.lift_surjective_of_surjective Ideal.Quotient.lift_surjective_of_surjective
/-- The ring homomorphism from the quotient by a smaller ideal to the quotient by a larger ideal.
This is the `Ideal.Quotient` version of `Quot.Factor` -/
def factor (S T : Ideal R) (H : S ≤ T) : R ⧸ S →+* R ⧸ T :=
Ideal.Quotient.lift S (mk T) fun _ hx => eq_zero_iff_mem.2 (H hx)
#align ideal.quotient.factor Ideal.Quotient.factor
@[simp]
theorem factor_mk (S T : Ideal R) (H : S ≤ T) (x : R) : factor S T H (mk S x) = mk T x :=
rfl
#align ideal.quotient.factor_mk Ideal.Quotient.factor_mk
@[simp]
theorem factor_comp_mk (S T : Ideal R) (H : S ≤ T) : (factor S T H).comp (mk S) = mk T := by
ext x
rw [RingHom.comp_apply, factor_mk]
#align ideal.quotient.factor_comp_mk Ideal.Quotient.factor_comp_mk
end Quotient
/-- Quotienting by equal ideals gives equivalent rings.
See also `Submodule.quotEquivOfEq` and `Ideal.quotientEquivAlgOfEq`.
-/
def quotEquivOfEq {R : Type*} [CommRing R] {I J : Ideal R} (h : I = J) : R ⧸ I ≃+* R ⧸ J :=
{ Submodule.quotEquivOfEq I J h with
map_mul' := by
rintro ⟨x⟩ ⟨y⟩
rfl }
#align ideal.quot_equiv_of_eq Ideal.quotEquivOfEq
@[simp]
theorem quotEquivOfEq_mk {R : Type*} [CommRing R] {I J : Ideal R} (h : I = J) (x : R) :
quotEquivOfEq h (Ideal.Quotient.mk I x) = Ideal.Quotient.mk J x :=
rfl
#align ideal.quot_equiv_of_eq_mk Ideal.quotEquivOfEq_mk
@[simp]
theorem quotEquivOfEq_symm {R : Type*} [CommRing R] {I J : Ideal R} (h : I = J) :
(Ideal.quotEquivOfEq h).symm = Ideal.quotEquivOfEq h.symm := by ext; rfl
#align ideal.quot_equiv_of_eq_symm Ideal.quotEquivOfEq_symm
section Pi
variable (ι : Type v)
/-- `R^n/I^n` is a `R/I`-module. -/
instance modulePi : Module (R ⧸ I) ((ι → R) ⧸ I.pi ι) where
smul c m :=
Quotient.liftOn₂' c m (fun r m => Submodule.Quotient.mk <| r • m) $ by
intro c₁ m₁ c₂ m₂ hc hm
apply Ideal.Quotient.eq.2
rw [Submodule.quotientRel_r_def] at hc hm
intro i
exact I.mul_sub_mul_mem hc (hm i)
one_smul := by
rintro ⟨a⟩
convert_to Ideal.Quotient.mk (I.pi ι) _ = Ideal.Quotient.mk (I.pi ι) _
congr with i; exact one_mul (a i)
mul_smul := by
rintro ⟨a⟩ ⟨b⟩ ⟨c⟩
convert_to Ideal.Quotient.mk (I.pi ι) _ = Ideal.Quotient.mk (I.pi ι) _
congr 1; funext i; exact mul_assoc a b (c i)
smul_add := by
rintro ⟨a⟩ ⟨b⟩ ⟨c⟩
convert_to Ideal.Quotient.mk (I.pi ι) _ = Ideal.Quotient.mk (I.pi ι) _
congr with i; exact mul_add a (b i) (c i)
smul_zero := by
rintro ⟨a⟩
convert_to Ideal.Quotient.mk (I.pi ι) _ = Ideal.Quotient.mk (I.pi ι) _
congr with _; exact mul_zero a
add_smul := by
rintro ⟨a⟩ ⟨b⟩ ⟨c⟩
convert_to Ideal.Quotient.mk (I.pi ι) _ = Ideal.Quotient.mk (I.pi ι) _
congr with i; exact add_mul a b (c i)
zero_smul := by
rintro ⟨a⟩
convert_to Ideal.Quotient.mk (I.pi ι) _ = Ideal.Quotient.mk (I.pi ι) _
congr with i; exact zero_mul (a i)
#align ideal.module_pi Ideal.modulePi
/-- `R^n/I^n` is isomorphic to `(R/I)^n` as an `R/I`-module. -/
noncomputable def piQuotEquiv : ((ι → R) ⧸ I.pi ι) ≃ₗ[R ⧸ I] ι → (R ⧸ I) where
toFun := fun x ↦
Quotient.liftOn' x (fun f i => Ideal.Quotient.mk I (f i)) fun a b hab =>
funext fun i => (Submodule.Quotient.eq' _).2 (QuotientAddGroup.leftRel_apply.mp hab i)
map_add' := by rintro ⟨_⟩ ⟨_⟩; rfl
map_smul' := by rintro ⟨_⟩ ⟨_⟩; rfl
invFun := fun x ↦ Ideal.Quotient.mk (I.pi ι) fun i ↦ Quotient.out' (x i)
left_inv := by
rintro ⟨x⟩
exact Ideal.Quotient.eq.2 fun i => Ideal.Quotient.eq.1 (Quotient.out_eq' _)
right_inv := by
intro x
ext i
obtain ⟨_, _⟩ := @Quot.exists_rep _ _ (x i)
convert Quotient.out_eq' (x i)
#align ideal.pi_quot_equiv Ideal.piQuotEquiv
/-- If `f : R^n → R^m` is an `R`-linear map and `I ⊆ R` is an ideal, then the image of `I^n` is
contained in `I^m`. -/
theorem map_pi {ι : Type*} [Finite ι] {ι' : Type w} (x : ι → R) (hi : ∀ i, x i ∈ I)
(f : (ι → R) →ₗ[R] ι' → R) (i : ι') : f x i ∈ I := by
classical
|
cases nonempty_fintype ι
|
/-- If `f : R^n → R^m` is an `R`-linear map and `I ⊆ R` is an ideal, then the image of `I^n` is
contained in `I^m`. -/
theorem map_pi {ι : Type*} [Finite ι] {ι' : Type w} (x : ι → R) (hi : ∀ i, x i ∈ I)
(f : (ι → R) →ₗ[R] ι' → R) (i : ι') : f x i ∈ I := by
classical
|
Mathlib.RingTheory.Ideal.Quotient.377_0.TwNAv7Pc4PYOWjX
|
/-- If `f : R^n → R^m` is an `R`-linear map and `I ⊆ R` is an ideal, then the image of `I^n` is
contained in `I^m`. -/
theorem map_pi {ι : Type*} [Finite ι] {ι' : Type w} (x : ι → R) (hi : ∀ i, x i ∈ I)
(f : (ι → R) →ₗ[R] ι' → R) (i : ι') : f x i ∈ I
|
Mathlib_RingTheory_Ideal_Quotient
|
case intro
R : Type u
inst✝¹ : CommRing R
I : Ideal R
a b : R
S ι✝ : Type v
ι : Type u_1
inst✝ : Finite ι
ι' : Type w
x : ι → R
hi : ∀ (i : ι), x i ∈ I
f : (ι → R) →ₗ[R] ι' → R
i : ι'
val✝ : Fintype ι
⊢ f x i ∈ I
|
/-
Copyright (c) 2018 Kenny Lau. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Kenny Lau, Chris Hughes, Mario Carneiro, Anne Baanen
-/
import Mathlib.Algebra.Ring.Fin
import Mathlib.Algebra.Ring.Prod
import Mathlib.LinearAlgebra.Quotient
import Mathlib.RingTheory.Congruence
import Mathlib.RingTheory.Ideal.Basic
import Mathlib.Tactic.FinCases
#align_import ring_theory.ideal.quotient from "leanprover-community/mathlib"@"949dc57e616a621462062668c9f39e4e17b64b69"
/-!
# Ideal quotients
This file defines ideal quotients as a special case of submodule quotients and proves some basic
results about these quotients.
See `Algebra.RingQuot` for quotients of non-commutative rings.
## Main definitions
- `Ideal.Quotient`: the quotient of a commutative ring `R` by an ideal `I : Ideal R`
-/
universe u v w
namespace Ideal
open Set
open BigOperators
variable {R : Type u} [CommRing R] (I : Ideal R) {a b : R}
variable {S : Type v}
-- Note that at present `Ideal` means a left-ideal,
-- so this quotient is only useful in a commutative ring.
-- We should develop quotients by two-sided ideals as well.
/-- The quotient `R/I` of a ring `R` by an ideal `I`.
The ideal quotient of `I` is defined to equal the quotient of `I` as an `R`-submodule of `R`.
This definition is marked `reducible` so that typeclass instances can be shared between
`Ideal.Quotient I` and `Submodule.Quotient I`.
-/
@[reducible]
instance : HasQuotient R (Ideal R) :=
Submodule.hasQuotient
namespace Quotient
variable {I} {x y : R}
instance one (I : Ideal R) : One (R ⧸ I) :=
⟨Submodule.Quotient.mk 1⟩
#align ideal.quotient.has_one Ideal.Quotient.one
/-- On `Ideal`s, `Submodule.quotientRel` is a ring congruence. -/
protected def ringCon (I : Ideal R) : RingCon R :=
{ QuotientAddGroup.con I.toAddSubgroup with
mul' := fun {a₁ b₁ a₂ b₂} h₁ h₂ => by
rw [Submodule.quotientRel_r_def] at h₁ h₂ ⊢
have F := I.add_mem (I.mul_mem_left a₂ h₁) (I.mul_mem_right b₁ h₂)
have : a₁ * a₂ - b₁ * b₂ = a₂ * (a₁ - b₁) + (a₂ - b₂) * b₁ := by
rw [mul_sub, sub_mul, sub_add_sub_cancel, mul_comm, mul_comm b₁]
rwa [← this] at F }
#align ideal.quotient.ring_con Ideal.Quotient.ringCon
instance commRing (I : Ideal R) : CommRing (R ⧸ I) :=
inferInstanceAs (CommRing (Quotient.ringCon I).Quotient)
#align ideal.quotient.comm_ring Ideal.Quotient.commRing
-- Sanity test to make sure no diamonds have emerged in `commRing`
example : (commRing I).toAddCommGroup = Submodule.Quotient.addCommGroup I := rfl
-- this instance is harder to find than the one via `Algebra α (R ⧸ I)`, so use a lower priority
instance (priority := 100) isScalarTower_right {α} [SMul α R] [IsScalarTower α R R] :
IsScalarTower α (R ⧸ I) (R ⧸ I) :=
(Quotient.ringCon I).isScalarTower_right
#align ideal.quotient.is_scalar_tower_right Ideal.Quotient.isScalarTower_right
instance smulCommClass {α} [SMul α R] [IsScalarTower α R R] [SMulCommClass α R R] :
SMulCommClass α (R ⧸ I) (R ⧸ I) :=
(Quotient.ringCon I).smulCommClass
#align ideal.quotient.smul_comm_class Ideal.Quotient.smulCommClass
instance smulCommClass' {α} [SMul α R] [IsScalarTower α R R] [SMulCommClass R α R] :
SMulCommClass (R ⧸ I) α (R ⧸ I) :=
(Quotient.ringCon I).smulCommClass'
#align ideal.quotient.smul_comm_class' Ideal.Quotient.smulCommClass'
/-- The ring homomorphism from a ring `R` to a quotient ring `R/I`. -/
def mk (I : Ideal R) : R →+* R ⧸ I where
toFun a := Submodule.Quotient.mk a
map_zero' := rfl
map_one' := rfl
map_mul' _ _ := rfl
map_add' _ _ := rfl
#align ideal.quotient.mk Ideal.Quotient.mk
instance {I : Ideal R} : Coe R (R ⧸ I) :=
⟨Ideal.Quotient.mk I⟩
/-- Two `RingHom`s from the quotient by an ideal are equal if their
compositions with `Ideal.Quotient.mk'` are equal.
See note [partially-applied ext lemmas]. -/
@[ext 1100]
theorem ringHom_ext [NonAssocSemiring S] ⦃f g : R ⧸ I →+* S⦄ (h : f.comp (mk I) = g.comp (mk I)) :
f = g :=
RingHom.ext fun x => Quotient.inductionOn' x <| (RingHom.congr_fun h : _)
#align ideal.quotient.ring_hom_ext Ideal.Quotient.ringHom_ext
instance inhabited : Inhabited (R ⧸ I) :=
⟨mk I 37⟩
#align ideal.quotient.inhabited Ideal.Quotient.inhabited
protected theorem eq : mk I x = mk I y ↔ x - y ∈ I :=
Submodule.Quotient.eq I
#align ideal.quotient.eq Ideal.Quotient.eq
@[simp]
theorem mk_eq_mk (x : R) : (Submodule.Quotient.mk x : R ⧸ I) = mk I x := rfl
#align ideal.quotient.mk_eq_mk Ideal.Quotient.mk_eq_mk
theorem eq_zero_iff_mem {I : Ideal R} : mk I a = 0 ↔ a ∈ I :=
Submodule.Quotient.mk_eq_zero _
#align ideal.quotient.eq_zero_iff_mem Ideal.Quotient.eq_zero_iff_mem
theorem eq_zero_iff_dvd (x y : R) : Ideal.Quotient.mk (Ideal.span ({x} : Set R)) y = 0 ↔ x ∣ y := by
rw [Ideal.Quotient.eq_zero_iff_mem, Ideal.mem_span_singleton]
-- Porting note: new theorem
theorem mk_eq_mk_iff_sub_mem (x y : R) : mk I x = mk I y ↔ x - y ∈ I := by
rw [← eq_zero_iff_mem, map_sub, sub_eq_zero]
theorem zero_eq_one_iff {I : Ideal R} : (0 : R ⧸ I) = 1 ↔ I = ⊤ :=
eq_comm.trans <| eq_zero_iff_mem.trans (eq_top_iff_one _).symm
#align ideal.quotient.zero_eq_one_iff Ideal.Quotient.zero_eq_one_iff
theorem zero_ne_one_iff {I : Ideal R} : (0 : R ⧸ I) ≠ 1 ↔ I ≠ ⊤ :=
not_congr zero_eq_one_iff
#align ideal.quotient.zero_ne_one_iff Ideal.Quotient.zero_ne_one_iff
protected theorem nontrivial {I : Ideal R} (hI : I ≠ ⊤) : Nontrivial (R ⧸ I) :=
⟨⟨0, 1, zero_ne_one_iff.2 hI⟩⟩
#align ideal.quotient.nontrivial Ideal.Quotient.nontrivial
theorem subsingleton_iff {I : Ideal R} : Subsingleton (R ⧸ I) ↔ I = ⊤ := by
rw [eq_top_iff_one, ← subsingleton_iff_zero_eq_one, eq_comm, ← (mk I).map_one,
Quotient.eq_zero_iff_mem]
#align ideal.quotient.subsingleton_iff Ideal.Quotient.subsingleton_iff
instance : Unique (R ⧸ (⊤ : Ideal R)) :=
⟨⟨0⟩, by rintro ⟨x⟩; exact Quotient.eq_zero_iff_mem.mpr Submodule.mem_top⟩
theorem mk_surjective : Function.Surjective (mk I) := fun y =>
Quotient.inductionOn' y fun x => Exists.intro x rfl
#align ideal.quotient.mk_surjective Ideal.Quotient.mk_surjective
instance : RingHomSurjective (mk I) :=
⟨mk_surjective⟩
/-- If `I` is an ideal of a commutative ring `R`, if `q : R → R/I` is the quotient map, and if
`s ⊆ R` is a subset, then `q⁻¹(q(s)) = ⋃ᵢ(i + s)`, the union running over all `i ∈ I`. -/
theorem quotient_ring_saturate (I : Ideal R) (s : Set R) :
mk I ⁻¹' (mk I '' s) = ⋃ x : I, (fun y => x.1 + y) '' s := by
ext x
simp only [mem_preimage, mem_image, mem_iUnion, Ideal.Quotient.eq]
exact
⟨fun ⟨a, a_in, h⟩ => ⟨⟨_, I.neg_mem h⟩, a, a_in, by simp⟩, fun ⟨⟨i, hi⟩, a, ha, Eq⟩ =>
⟨a, ha, by rw [← Eq, sub_add_eq_sub_sub_swap, sub_self, zero_sub]; exact I.neg_mem hi⟩⟩
#align ideal.quotient.quotient_ring_saturate Ideal.Quotient.quotient_ring_saturate
instance noZeroDivisors (I : Ideal R) [hI : I.IsPrime] : NoZeroDivisors (R ⧸ I) where
eq_zero_or_eq_zero_of_mul_eq_zero {a b} := Quotient.inductionOn₂' a b fun {_ _} hab =>
(hI.mem_or_mem (eq_zero_iff_mem.1 hab)).elim (Or.inl ∘ eq_zero_iff_mem.2)
(Or.inr ∘ eq_zero_iff_mem.2)
#align ideal.quotient.no_zero_divisors Ideal.Quotient.noZeroDivisors
instance isDomain (I : Ideal R) [hI : I.IsPrime] : IsDomain (R ⧸ I) :=
let _ := Quotient.nontrivial hI.1
NoZeroDivisors.to_isDomain _
#align ideal.quotient.is_domain Ideal.Quotient.isDomain
theorem isDomain_iff_prime (I : Ideal R) : IsDomain (R ⧸ I) ↔ I.IsPrime := by
refine' ⟨fun H => ⟨zero_ne_one_iff.1 _, fun {x y} h => _⟩, fun h => inferInstance⟩
· haveI : Nontrivial (R ⧸ I) := ⟨H.2.1⟩
exact zero_ne_one
· simp only [← eq_zero_iff_mem, (mk I).map_mul] at h ⊢
haveI := @IsDomain.to_noZeroDivisors (R ⧸ I) _ H
exact eq_zero_or_eq_zero_of_mul_eq_zero h
#align ideal.quotient.is_domain_iff_prime Ideal.Quotient.isDomain_iff_prime
theorem exists_inv {I : Ideal R} [hI : I.IsMaximal] :
∀ {a : R ⧸ I}, a ≠ 0 → ∃ b : R ⧸ I, a * b = 1 := by
rintro ⟨a⟩ h
rcases hI.exists_inv (mt eq_zero_iff_mem.2 h) with ⟨b, c, hc, abc⟩
rw [mul_comm] at abc
refine' ⟨mk _ b, Quot.sound _⟩
simp only [Submodule.quotientRel_r_def]
rw [← eq_sub_iff_add_eq'] at abc
rwa [abc, ← neg_mem_iff (G := R) (H := I), neg_sub] at hc
#align ideal.quotient.exists_inv Ideal.Quotient.exists_inv
open Classical
/-- The quotient by a maximal ideal is a group with zero. This is a `def` rather than `instance`,
since users will have computable inverses in some applications.
See note [reducible non-instances]. -/
@[reducible]
protected noncomputable def groupWithZero (I : Ideal R) [hI : I.IsMaximal] :
GroupWithZero (R ⧸ I) :=
{ inv := fun a => if ha : a = 0 then 0 else Classical.choose (exists_inv ha)
mul_inv_cancel := fun a (ha : a ≠ 0) =>
show a * dite _ _ _ = _ by rw [dif_neg ha]; exact Classical.choose_spec (exists_inv ha)
inv_zero := dif_pos rfl }
#align ideal.quotient.group_with_zero Ideal.Quotient.groupWithZero
/-- The quotient by a maximal ideal is a field. This is a `def` rather than `instance`, since users
will have computable inverses (and `qsmul`, `rat_cast`) in some applications.
See note [reducible non-instances]. -/
@[reducible]
protected noncomputable def field (I : Ideal R) [hI : I.IsMaximal] : Field (R ⧸ I) :=
{ Quotient.commRing I, Quotient.groupWithZero I with }
#align ideal.quotient.field Ideal.Quotient.field
/-- If the quotient by an ideal is a field, then the ideal is maximal. -/
theorem maximal_of_isField (I : Ideal R) (hqf : IsField (R ⧸ I)) : I.IsMaximal := by
apply Ideal.isMaximal_iff.2
constructor
· intro h
rcases hqf.exists_pair_ne with ⟨⟨x⟩, ⟨y⟩, hxy⟩
exact hxy (Ideal.Quotient.eq.2 (mul_one (x - y) ▸ I.mul_mem_left _ h))
· intro J x hIJ hxnI hxJ
rcases hqf.mul_inv_cancel (mt Ideal.Quotient.eq_zero_iff_mem.1 hxnI) with ⟨⟨y⟩, hy⟩
rw [← zero_add (1 : R), ← sub_self (x * y), sub_add]
refine' J.sub_mem (J.mul_mem_right _ hxJ) (hIJ (Ideal.Quotient.eq.1 hy))
#align ideal.quotient.maximal_of_is_field Ideal.Quotient.maximal_of_isField
/-- The quotient of a ring by an ideal is a field iff the ideal is maximal. -/
theorem maximal_ideal_iff_isField_quotient (I : Ideal R) : I.IsMaximal ↔ IsField (R ⧸ I) :=
⟨fun h =>
let _i := @Quotient.field _ _ I h
Field.toIsField _,
maximal_of_isField _⟩
#align ideal.quotient.maximal_ideal_iff_is_field_quotient Ideal.Quotient.maximal_ideal_iff_isField_quotient
variable [CommRing S]
/-- Given a ring homomorphism `f : R →+* S` sending all elements of an ideal to zero,
lift it to the quotient by this ideal. -/
def lift (I : Ideal R) (f : R →+* S) (H : ∀ a : R, a ∈ I → f a = 0) : R ⧸ I →+* S :=
{ QuotientAddGroup.lift I.toAddSubgroup f.toAddMonoidHom H with
map_one' := f.map_one
map_mul' := fun a₁ a₂ => Quotient.inductionOn₂' a₁ a₂ f.map_mul }
#align ideal.quotient.lift Ideal.Quotient.lift
@[simp]
theorem lift_mk (I : Ideal R) (f : R →+* S) (H : ∀ a : R, a ∈ I → f a = 0) :
lift I f H (mk I a) = f a :=
rfl
#align ideal.quotient.lift_mk Ideal.Quotient.lift_mk
theorem lift_surjective_of_surjective (I : Ideal R) {f : R →+* S} (H : ∀ a : R, a ∈ I → f a = 0)
(hf : Function.Surjective f) : Function.Surjective (Ideal.Quotient.lift I f H) := by
intro y
obtain ⟨x, rfl⟩ := hf y
use Ideal.Quotient.mk I x
simp only [Ideal.Quotient.lift_mk]
#align ideal.quotient.lift_surjective_of_surjective Ideal.Quotient.lift_surjective_of_surjective
/-- The ring homomorphism from the quotient by a smaller ideal to the quotient by a larger ideal.
This is the `Ideal.Quotient` version of `Quot.Factor` -/
def factor (S T : Ideal R) (H : S ≤ T) : R ⧸ S →+* R ⧸ T :=
Ideal.Quotient.lift S (mk T) fun _ hx => eq_zero_iff_mem.2 (H hx)
#align ideal.quotient.factor Ideal.Quotient.factor
@[simp]
theorem factor_mk (S T : Ideal R) (H : S ≤ T) (x : R) : factor S T H (mk S x) = mk T x :=
rfl
#align ideal.quotient.factor_mk Ideal.Quotient.factor_mk
@[simp]
theorem factor_comp_mk (S T : Ideal R) (H : S ≤ T) : (factor S T H).comp (mk S) = mk T := by
ext x
rw [RingHom.comp_apply, factor_mk]
#align ideal.quotient.factor_comp_mk Ideal.Quotient.factor_comp_mk
end Quotient
/-- Quotienting by equal ideals gives equivalent rings.
See also `Submodule.quotEquivOfEq` and `Ideal.quotientEquivAlgOfEq`.
-/
def quotEquivOfEq {R : Type*} [CommRing R] {I J : Ideal R} (h : I = J) : R ⧸ I ≃+* R ⧸ J :=
{ Submodule.quotEquivOfEq I J h with
map_mul' := by
rintro ⟨x⟩ ⟨y⟩
rfl }
#align ideal.quot_equiv_of_eq Ideal.quotEquivOfEq
@[simp]
theorem quotEquivOfEq_mk {R : Type*} [CommRing R] {I J : Ideal R} (h : I = J) (x : R) :
quotEquivOfEq h (Ideal.Quotient.mk I x) = Ideal.Quotient.mk J x :=
rfl
#align ideal.quot_equiv_of_eq_mk Ideal.quotEquivOfEq_mk
@[simp]
theorem quotEquivOfEq_symm {R : Type*} [CommRing R] {I J : Ideal R} (h : I = J) :
(Ideal.quotEquivOfEq h).symm = Ideal.quotEquivOfEq h.symm := by ext; rfl
#align ideal.quot_equiv_of_eq_symm Ideal.quotEquivOfEq_symm
section Pi
variable (ι : Type v)
/-- `R^n/I^n` is a `R/I`-module. -/
instance modulePi : Module (R ⧸ I) ((ι → R) ⧸ I.pi ι) where
smul c m :=
Quotient.liftOn₂' c m (fun r m => Submodule.Quotient.mk <| r • m) $ by
intro c₁ m₁ c₂ m₂ hc hm
apply Ideal.Quotient.eq.2
rw [Submodule.quotientRel_r_def] at hc hm
intro i
exact I.mul_sub_mul_mem hc (hm i)
one_smul := by
rintro ⟨a⟩
convert_to Ideal.Quotient.mk (I.pi ι) _ = Ideal.Quotient.mk (I.pi ι) _
congr with i; exact one_mul (a i)
mul_smul := by
rintro ⟨a⟩ ⟨b⟩ ⟨c⟩
convert_to Ideal.Quotient.mk (I.pi ι) _ = Ideal.Quotient.mk (I.pi ι) _
congr 1; funext i; exact mul_assoc a b (c i)
smul_add := by
rintro ⟨a⟩ ⟨b⟩ ⟨c⟩
convert_to Ideal.Quotient.mk (I.pi ι) _ = Ideal.Quotient.mk (I.pi ι) _
congr with i; exact mul_add a (b i) (c i)
smul_zero := by
rintro ⟨a⟩
convert_to Ideal.Quotient.mk (I.pi ι) _ = Ideal.Quotient.mk (I.pi ι) _
congr with _; exact mul_zero a
add_smul := by
rintro ⟨a⟩ ⟨b⟩ ⟨c⟩
convert_to Ideal.Quotient.mk (I.pi ι) _ = Ideal.Quotient.mk (I.pi ι) _
congr with i; exact add_mul a b (c i)
zero_smul := by
rintro ⟨a⟩
convert_to Ideal.Quotient.mk (I.pi ι) _ = Ideal.Quotient.mk (I.pi ι) _
congr with i; exact zero_mul (a i)
#align ideal.module_pi Ideal.modulePi
/-- `R^n/I^n` is isomorphic to `(R/I)^n` as an `R/I`-module. -/
noncomputable def piQuotEquiv : ((ι → R) ⧸ I.pi ι) ≃ₗ[R ⧸ I] ι → (R ⧸ I) where
toFun := fun x ↦
Quotient.liftOn' x (fun f i => Ideal.Quotient.mk I (f i)) fun a b hab =>
funext fun i => (Submodule.Quotient.eq' _).2 (QuotientAddGroup.leftRel_apply.mp hab i)
map_add' := by rintro ⟨_⟩ ⟨_⟩; rfl
map_smul' := by rintro ⟨_⟩ ⟨_⟩; rfl
invFun := fun x ↦ Ideal.Quotient.mk (I.pi ι) fun i ↦ Quotient.out' (x i)
left_inv := by
rintro ⟨x⟩
exact Ideal.Quotient.eq.2 fun i => Ideal.Quotient.eq.1 (Quotient.out_eq' _)
right_inv := by
intro x
ext i
obtain ⟨_, _⟩ := @Quot.exists_rep _ _ (x i)
convert Quotient.out_eq' (x i)
#align ideal.pi_quot_equiv Ideal.piQuotEquiv
/-- If `f : R^n → R^m` is an `R`-linear map and `I ⊆ R` is an ideal, then the image of `I^n` is
contained in `I^m`. -/
theorem map_pi {ι : Type*} [Finite ι] {ι' : Type w} (x : ι → R) (hi : ∀ i, x i ∈ I)
(f : (ι → R) →ₗ[R] ι' → R) (i : ι') : f x i ∈ I := by
classical
cases nonempty_fintype ι
|
rw [pi_eq_sum_univ x]
|
/-- If `f : R^n → R^m` is an `R`-linear map and `I ⊆ R` is an ideal, then the image of `I^n` is
contained in `I^m`. -/
theorem map_pi {ι : Type*} [Finite ι] {ι' : Type w} (x : ι → R) (hi : ∀ i, x i ∈ I)
(f : (ι → R) →ₗ[R] ι' → R) (i : ι') : f x i ∈ I := by
classical
cases nonempty_fintype ι
|
Mathlib.RingTheory.Ideal.Quotient.377_0.TwNAv7Pc4PYOWjX
|
/-- If `f : R^n → R^m` is an `R`-linear map and `I ⊆ R` is an ideal, then the image of `I^n` is
contained in `I^m`. -/
theorem map_pi {ι : Type*} [Finite ι] {ι' : Type w} (x : ι → R) (hi : ∀ i, x i ∈ I)
(f : (ι → R) →ₗ[R] ι' → R) (i : ι') : f x i ∈ I
|
Mathlib_RingTheory_Ideal_Quotient
|
case intro
R : Type u
inst✝¹ : CommRing R
I : Ideal R
a b : R
S ι✝ : Type v
ι : Type u_1
inst✝ : Finite ι
ι' : Type w
x : ι → R
hi : ∀ (i : ι), x i ∈ I
f : (ι → R) →ₗ[R] ι' → R
i : ι'
val✝ : Fintype ι
⊢ f (∑ i : ι, x i • fun j => if i = j then 1 else 0) i ∈ I
|
/-
Copyright (c) 2018 Kenny Lau. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Kenny Lau, Chris Hughes, Mario Carneiro, Anne Baanen
-/
import Mathlib.Algebra.Ring.Fin
import Mathlib.Algebra.Ring.Prod
import Mathlib.LinearAlgebra.Quotient
import Mathlib.RingTheory.Congruence
import Mathlib.RingTheory.Ideal.Basic
import Mathlib.Tactic.FinCases
#align_import ring_theory.ideal.quotient from "leanprover-community/mathlib"@"949dc57e616a621462062668c9f39e4e17b64b69"
/-!
# Ideal quotients
This file defines ideal quotients as a special case of submodule quotients and proves some basic
results about these quotients.
See `Algebra.RingQuot` for quotients of non-commutative rings.
## Main definitions
- `Ideal.Quotient`: the quotient of a commutative ring `R` by an ideal `I : Ideal R`
-/
universe u v w
namespace Ideal
open Set
open BigOperators
variable {R : Type u} [CommRing R] (I : Ideal R) {a b : R}
variable {S : Type v}
-- Note that at present `Ideal` means a left-ideal,
-- so this quotient is only useful in a commutative ring.
-- We should develop quotients by two-sided ideals as well.
/-- The quotient `R/I` of a ring `R` by an ideal `I`.
The ideal quotient of `I` is defined to equal the quotient of `I` as an `R`-submodule of `R`.
This definition is marked `reducible` so that typeclass instances can be shared between
`Ideal.Quotient I` and `Submodule.Quotient I`.
-/
@[reducible]
instance : HasQuotient R (Ideal R) :=
Submodule.hasQuotient
namespace Quotient
variable {I} {x y : R}
instance one (I : Ideal R) : One (R ⧸ I) :=
⟨Submodule.Quotient.mk 1⟩
#align ideal.quotient.has_one Ideal.Quotient.one
/-- On `Ideal`s, `Submodule.quotientRel` is a ring congruence. -/
protected def ringCon (I : Ideal R) : RingCon R :=
{ QuotientAddGroup.con I.toAddSubgroup with
mul' := fun {a₁ b₁ a₂ b₂} h₁ h₂ => by
rw [Submodule.quotientRel_r_def] at h₁ h₂ ⊢
have F := I.add_mem (I.mul_mem_left a₂ h₁) (I.mul_mem_right b₁ h₂)
have : a₁ * a₂ - b₁ * b₂ = a₂ * (a₁ - b₁) + (a₂ - b₂) * b₁ := by
rw [mul_sub, sub_mul, sub_add_sub_cancel, mul_comm, mul_comm b₁]
rwa [← this] at F }
#align ideal.quotient.ring_con Ideal.Quotient.ringCon
instance commRing (I : Ideal R) : CommRing (R ⧸ I) :=
inferInstanceAs (CommRing (Quotient.ringCon I).Quotient)
#align ideal.quotient.comm_ring Ideal.Quotient.commRing
-- Sanity test to make sure no diamonds have emerged in `commRing`
example : (commRing I).toAddCommGroup = Submodule.Quotient.addCommGroup I := rfl
-- this instance is harder to find than the one via `Algebra α (R ⧸ I)`, so use a lower priority
instance (priority := 100) isScalarTower_right {α} [SMul α R] [IsScalarTower α R R] :
IsScalarTower α (R ⧸ I) (R ⧸ I) :=
(Quotient.ringCon I).isScalarTower_right
#align ideal.quotient.is_scalar_tower_right Ideal.Quotient.isScalarTower_right
instance smulCommClass {α} [SMul α R] [IsScalarTower α R R] [SMulCommClass α R R] :
SMulCommClass α (R ⧸ I) (R ⧸ I) :=
(Quotient.ringCon I).smulCommClass
#align ideal.quotient.smul_comm_class Ideal.Quotient.smulCommClass
instance smulCommClass' {α} [SMul α R] [IsScalarTower α R R] [SMulCommClass R α R] :
SMulCommClass (R ⧸ I) α (R ⧸ I) :=
(Quotient.ringCon I).smulCommClass'
#align ideal.quotient.smul_comm_class' Ideal.Quotient.smulCommClass'
/-- The ring homomorphism from a ring `R` to a quotient ring `R/I`. -/
def mk (I : Ideal R) : R →+* R ⧸ I where
toFun a := Submodule.Quotient.mk a
map_zero' := rfl
map_one' := rfl
map_mul' _ _ := rfl
map_add' _ _ := rfl
#align ideal.quotient.mk Ideal.Quotient.mk
instance {I : Ideal R} : Coe R (R ⧸ I) :=
⟨Ideal.Quotient.mk I⟩
/-- Two `RingHom`s from the quotient by an ideal are equal if their
compositions with `Ideal.Quotient.mk'` are equal.
See note [partially-applied ext lemmas]. -/
@[ext 1100]
theorem ringHom_ext [NonAssocSemiring S] ⦃f g : R ⧸ I →+* S⦄ (h : f.comp (mk I) = g.comp (mk I)) :
f = g :=
RingHom.ext fun x => Quotient.inductionOn' x <| (RingHom.congr_fun h : _)
#align ideal.quotient.ring_hom_ext Ideal.Quotient.ringHom_ext
instance inhabited : Inhabited (R ⧸ I) :=
⟨mk I 37⟩
#align ideal.quotient.inhabited Ideal.Quotient.inhabited
protected theorem eq : mk I x = mk I y ↔ x - y ∈ I :=
Submodule.Quotient.eq I
#align ideal.quotient.eq Ideal.Quotient.eq
@[simp]
theorem mk_eq_mk (x : R) : (Submodule.Quotient.mk x : R ⧸ I) = mk I x := rfl
#align ideal.quotient.mk_eq_mk Ideal.Quotient.mk_eq_mk
theorem eq_zero_iff_mem {I : Ideal R} : mk I a = 0 ↔ a ∈ I :=
Submodule.Quotient.mk_eq_zero _
#align ideal.quotient.eq_zero_iff_mem Ideal.Quotient.eq_zero_iff_mem
theorem eq_zero_iff_dvd (x y : R) : Ideal.Quotient.mk (Ideal.span ({x} : Set R)) y = 0 ↔ x ∣ y := by
rw [Ideal.Quotient.eq_zero_iff_mem, Ideal.mem_span_singleton]
-- Porting note: new theorem
theorem mk_eq_mk_iff_sub_mem (x y : R) : mk I x = mk I y ↔ x - y ∈ I := by
rw [← eq_zero_iff_mem, map_sub, sub_eq_zero]
theorem zero_eq_one_iff {I : Ideal R} : (0 : R ⧸ I) = 1 ↔ I = ⊤ :=
eq_comm.trans <| eq_zero_iff_mem.trans (eq_top_iff_one _).symm
#align ideal.quotient.zero_eq_one_iff Ideal.Quotient.zero_eq_one_iff
theorem zero_ne_one_iff {I : Ideal R} : (0 : R ⧸ I) ≠ 1 ↔ I ≠ ⊤ :=
not_congr zero_eq_one_iff
#align ideal.quotient.zero_ne_one_iff Ideal.Quotient.zero_ne_one_iff
protected theorem nontrivial {I : Ideal R} (hI : I ≠ ⊤) : Nontrivial (R ⧸ I) :=
⟨⟨0, 1, zero_ne_one_iff.2 hI⟩⟩
#align ideal.quotient.nontrivial Ideal.Quotient.nontrivial
theorem subsingleton_iff {I : Ideal R} : Subsingleton (R ⧸ I) ↔ I = ⊤ := by
rw [eq_top_iff_one, ← subsingleton_iff_zero_eq_one, eq_comm, ← (mk I).map_one,
Quotient.eq_zero_iff_mem]
#align ideal.quotient.subsingleton_iff Ideal.Quotient.subsingleton_iff
instance : Unique (R ⧸ (⊤ : Ideal R)) :=
⟨⟨0⟩, by rintro ⟨x⟩; exact Quotient.eq_zero_iff_mem.mpr Submodule.mem_top⟩
theorem mk_surjective : Function.Surjective (mk I) := fun y =>
Quotient.inductionOn' y fun x => Exists.intro x rfl
#align ideal.quotient.mk_surjective Ideal.Quotient.mk_surjective
instance : RingHomSurjective (mk I) :=
⟨mk_surjective⟩
/-- If `I` is an ideal of a commutative ring `R`, if `q : R → R/I` is the quotient map, and if
`s ⊆ R` is a subset, then `q⁻¹(q(s)) = ⋃ᵢ(i + s)`, the union running over all `i ∈ I`. -/
theorem quotient_ring_saturate (I : Ideal R) (s : Set R) :
mk I ⁻¹' (mk I '' s) = ⋃ x : I, (fun y => x.1 + y) '' s := by
ext x
simp only [mem_preimage, mem_image, mem_iUnion, Ideal.Quotient.eq]
exact
⟨fun ⟨a, a_in, h⟩ => ⟨⟨_, I.neg_mem h⟩, a, a_in, by simp⟩, fun ⟨⟨i, hi⟩, a, ha, Eq⟩ =>
⟨a, ha, by rw [← Eq, sub_add_eq_sub_sub_swap, sub_self, zero_sub]; exact I.neg_mem hi⟩⟩
#align ideal.quotient.quotient_ring_saturate Ideal.Quotient.quotient_ring_saturate
instance noZeroDivisors (I : Ideal R) [hI : I.IsPrime] : NoZeroDivisors (R ⧸ I) where
eq_zero_or_eq_zero_of_mul_eq_zero {a b} := Quotient.inductionOn₂' a b fun {_ _} hab =>
(hI.mem_or_mem (eq_zero_iff_mem.1 hab)).elim (Or.inl ∘ eq_zero_iff_mem.2)
(Or.inr ∘ eq_zero_iff_mem.2)
#align ideal.quotient.no_zero_divisors Ideal.Quotient.noZeroDivisors
instance isDomain (I : Ideal R) [hI : I.IsPrime] : IsDomain (R ⧸ I) :=
let _ := Quotient.nontrivial hI.1
NoZeroDivisors.to_isDomain _
#align ideal.quotient.is_domain Ideal.Quotient.isDomain
theorem isDomain_iff_prime (I : Ideal R) : IsDomain (R ⧸ I) ↔ I.IsPrime := by
refine' ⟨fun H => ⟨zero_ne_one_iff.1 _, fun {x y} h => _⟩, fun h => inferInstance⟩
· haveI : Nontrivial (R ⧸ I) := ⟨H.2.1⟩
exact zero_ne_one
· simp only [← eq_zero_iff_mem, (mk I).map_mul] at h ⊢
haveI := @IsDomain.to_noZeroDivisors (R ⧸ I) _ H
exact eq_zero_or_eq_zero_of_mul_eq_zero h
#align ideal.quotient.is_domain_iff_prime Ideal.Quotient.isDomain_iff_prime
theorem exists_inv {I : Ideal R} [hI : I.IsMaximal] :
∀ {a : R ⧸ I}, a ≠ 0 → ∃ b : R ⧸ I, a * b = 1 := by
rintro ⟨a⟩ h
rcases hI.exists_inv (mt eq_zero_iff_mem.2 h) with ⟨b, c, hc, abc⟩
rw [mul_comm] at abc
refine' ⟨mk _ b, Quot.sound _⟩
simp only [Submodule.quotientRel_r_def]
rw [← eq_sub_iff_add_eq'] at abc
rwa [abc, ← neg_mem_iff (G := R) (H := I), neg_sub] at hc
#align ideal.quotient.exists_inv Ideal.Quotient.exists_inv
open Classical
/-- The quotient by a maximal ideal is a group with zero. This is a `def` rather than `instance`,
since users will have computable inverses in some applications.
See note [reducible non-instances]. -/
@[reducible]
protected noncomputable def groupWithZero (I : Ideal R) [hI : I.IsMaximal] :
GroupWithZero (R ⧸ I) :=
{ inv := fun a => if ha : a = 0 then 0 else Classical.choose (exists_inv ha)
mul_inv_cancel := fun a (ha : a ≠ 0) =>
show a * dite _ _ _ = _ by rw [dif_neg ha]; exact Classical.choose_spec (exists_inv ha)
inv_zero := dif_pos rfl }
#align ideal.quotient.group_with_zero Ideal.Quotient.groupWithZero
/-- The quotient by a maximal ideal is a field. This is a `def` rather than `instance`, since users
will have computable inverses (and `qsmul`, `rat_cast`) in some applications.
See note [reducible non-instances]. -/
@[reducible]
protected noncomputable def field (I : Ideal R) [hI : I.IsMaximal] : Field (R ⧸ I) :=
{ Quotient.commRing I, Quotient.groupWithZero I with }
#align ideal.quotient.field Ideal.Quotient.field
/-- If the quotient by an ideal is a field, then the ideal is maximal. -/
theorem maximal_of_isField (I : Ideal R) (hqf : IsField (R ⧸ I)) : I.IsMaximal := by
apply Ideal.isMaximal_iff.2
constructor
· intro h
rcases hqf.exists_pair_ne with ⟨⟨x⟩, ⟨y⟩, hxy⟩
exact hxy (Ideal.Quotient.eq.2 (mul_one (x - y) ▸ I.mul_mem_left _ h))
· intro J x hIJ hxnI hxJ
rcases hqf.mul_inv_cancel (mt Ideal.Quotient.eq_zero_iff_mem.1 hxnI) with ⟨⟨y⟩, hy⟩
rw [← zero_add (1 : R), ← sub_self (x * y), sub_add]
refine' J.sub_mem (J.mul_mem_right _ hxJ) (hIJ (Ideal.Quotient.eq.1 hy))
#align ideal.quotient.maximal_of_is_field Ideal.Quotient.maximal_of_isField
/-- The quotient of a ring by an ideal is a field iff the ideal is maximal. -/
theorem maximal_ideal_iff_isField_quotient (I : Ideal R) : I.IsMaximal ↔ IsField (R ⧸ I) :=
⟨fun h =>
let _i := @Quotient.field _ _ I h
Field.toIsField _,
maximal_of_isField _⟩
#align ideal.quotient.maximal_ideal_iff_is_field_quotient Ideal.Quotient.maximal_ideal_iff_isField_quotient
variable [CommRing S]
/-- Given a ring homomorphism `f : R →+* S` sending all elements of an ideal to zero,
lift it to the quotient by this ideal. -/
def lift (I : Ideal R) (f : R →+* S) (H : ∀ a : R, a ∈ I → f a = 0) : R ⧸ I →+* S :=
{ QuotientAddGroup.lift I.toAddSubgroup f.toAddMonoidHom H with
map_one' := f.map_one
map_mul' := fun a₁ a₂ => Quotient.inductionOn₂' a₁ a₂ f.map_mul }
#align ideal.quotient.lift Ideal.Quotient.lift
@[simp]
theorem lift_mk (I : Ideal R) (f : R →+* S) (H : ∀ a : R, a ∈ I → f a = 0) :
lift I f H (mk I a) = f a :=
rfl
#align ideal.quotient.lift_mk Ideal.Quotient.lift_mk
theorem lift_surjective_of_surjective (I : Ideal R) {f : R →+* S} (H : ∀ a : R, a ∈ I → f a = 0)
(hf : Function.Surjective f) : Function.Surjective (Ideal.Quotient.lift I f H) := by
intro y
obtain ⟨x, rfl⟩ := hf y
use Ideal.Quotient.mk I x
simp only [Ideal.Quotient.lift_mk]
#align ideal.quotient.lift_surjective_of_surjective Ideal.Quotient.lift_surjective_of_surjective
/-- The ring homomorphism from the quotient by a smaller ideal to the quotient by a larger ideal.
This is the `Ideal.Quotient` version of `Quot.Factor` -/
def factor (S T : Ideal R) (H : S ≤ T) : R ⧸ S →+* R ⧸ T :=
Ideal.Quotient.lift S (mk T) fun _ hx => eq_zero_iff_mem.2 (H hx)
#align ideal.quotient.factor Ideal.Quotient.factor
@[simp]
theorem factor_mk (S T : Ideal R) (H : S ≤ T) (x : R) : factor S T H (mk S x) = mk T x :=
rfl
#align ideal.quotient.factor_mk Ideal.Quotient.factor_mk
@[simp]
theorem factor_comp_mk (S T : Ideal R) (H : S ≤ T) : (factor S T H).comp (mk S) = mk T := by
ext x
rw [RingHom.comp_apply, factor_mk]
#align ideal.quotient.factor_comp_mk Ideal.Quotient.factor_comp_mk
end Quotient
/-- Quotienting by equal ideals gives equivalent rings.
See also `Submodule.quotEquivOfEq` and `Ideal.quotientEquivAlgOfEq`.
-/
def quotEquivOfEq {R : Type*} [CommRing R] {I J : Ideal R} (h : I = J) : R ⧸ I ≃+* R ⧸ J :=
{ Submodule.quotEquivOfEq I J h with
map_mul' := by
rintro ⟨x⟩ ⟨y⟩
rfl }
#align ideal.quot_equiv_of_eq Ideal.quotEquivOfEq
@[simp]
theorem quotEquivOfEq_mk {R : Type*} [CommRing R] {I J : Ideal R} (h : I = J) (x : R) :
quotEquivOfEq h (Ideal.Quotient.mk I x) = Ideal.Quotient.mk J x :=
rfl
#align ideal.quot_equiv_of_eq_mk Ideal.quotEquivOfEq_mk
@[simp]
theorem quotEquivOfEq_symm {R : Type*} [CommRing R] {I J : Ideal R} (h : I = J) :
(Ideal.quotEquivOfEq h).symm = Ideal.quotEquivOfEq h.symm := by ext; rfl
#align ideal.quot_equiv_of_eq_symm Ideal.quotEquivOfEq_symm
section Pi
variable (ι : Type v)
/-- `R^n/I^n` is a `R/I`-module. -/
instance modulePi : Module (R ⧸ I) ((ι → R) ⧸ I.pi ι) where
smul c m :=
Quotient.liftOn₂' c m (fun r m => Submodule.Quotient.mk <| r • m) $ by
intro c₁ m₁ c₂ m₂ hc hm
apply Ideal.Quotient.eq.2
rw [Submodule.quotientRel_r_def] at hc hm
intro i
exact I.mul_sub_mul_mem hc (hm i)
one_smul := by
rintro ⟨a⟩
convert_to Ideal.Quotient.mk (I.pi ι) _ = Ideal.Quotient.mk (I.pi ι) _
congr with i; exact one_mul (a i)
mul_smul := by
rintro ⟨a⟩ ⟨b⟩ ⟨c⟩
convert_to Ideal.Quotient.mk (I.pi ι) _ = Ideal.Quotient.mk (I.pi ι) _
congr 1; funext i; exact mul_assoc a b (c i)
smul_add := by
rintro ⟨a⟩ ⟨b⟩ ⟨c⟩
convert_to Ideal.Quotient.mk (I.pi ι) _ = Ideal.Quotient.mk (I.pi ι) _
congr with i; exact mul_add a (b i) (c i)
smul_zero := by
rintro ⟨a⟩
convert_to Ideal.Quotient.mk (I.pi ι) _ = Ideal.Quotient.mk (I.pi ι) _
congr with _; exact mul_zero a
add_smul := by
rintro ⟨a⟩ ⟨b⟩ ⟨c⟩
convert_to Ideal.Quotient.mk (I.pi ι) _ = Ideal.Quotient.mk (I.pi ι) _
congr with i; exact add_mul a b (c i)
zero_smul := by
rintro ⟨a⟩
convert_to Ideal.Quotient.mk (I.pi ι) _ = Ideal.Quotient.mk (I.pi ι) _
congr with i; exact zero_mul (a i)
#align ideal.module_pi Ideal.modulePi
/-- `R^n/I^n` is isomorphic to `(R/I)^n` as an `R/I`-module. -/
noncomputable def piQuotEquiv : ((ι → R) ⧸ I.pi ι) ≃ₗ[R ⧸ I] ι → (R ⧸ I) where
toFun := fun x ↦
Quotient.liftOn' x (fun f i => Ideal.Quotient.mk I (f i)) fun a b hab =>
funext fun i => (Submodule.Quotient.eq' _).2 (QuotientAddGroup.leftRel_apply.mp hab i)
map_add' := by rintro ⟨_⟩ ⟨_⟩; rfl
map_smul' := by rintro ⟨_⟩ ⟨_⟩; rfl
invFun := fun x ↦ Ideal.Quotient.mk (I.pi ι) fun i ↦ Quotient.out' (x i)
left_inv := by
rintro ⟨x⟩
exact Ideal.Quotient.eq.2 fun i => Ideal.Quotient.eq.1 (Quotient.out_eq' _)
right_inv := by
intro x
ext i
obtain ⟨_, _⟩ := @Quot.exists_rep _ _ (x i)
convert Quotient.out_eq' (x i)
#align ideal.pi_quot_equiv Ideal.piQuotEquiv
/-- If `f : R^n → R^m` is an `R`-linear map and `I ⊆ R` is an ideal, then the image of `I^n` is
contained in `I^m`. -/
theorem map_pi {ι : Type*} [Finite ι] {ι' : Type w} (x : ι → R) (hi : ∀ i, x i ∈ I)
(f : (ι → R) →ₗ[R] ι' → R) (i : ι') : f x i ∈ I := by
classical
cases nonempty_fintype ι
rw [pi_eq_sum_univ x]
|
simp only [Finset.sum_apply, smul_eq_mul, map_sum, Pi.smul_apply, map_smul]
|
/-- If `f : R^n → R^m` is an `R`-linear map and `I ⊆ R` is an ideal, then the image of `I^n` is
contained in `I^m`. -/
theorem map_pi {ι : Type*} [Finite ι] {ι' : Type w} (x : ι → R) (hi : ∀ i, x i ∈ I)
(f : (ι → R) →ₗ[R] ι' → R) (i : ι') : f x i ∈ I := by
classical
cases nonempty_fintype ι
rw [pi_eq_sum_univ x]
|
Mathlib.RingTheory.Ideal.Quotient.377_0.TwNAv7Pc4PYOWjX
|
/-- If `f : R^n → R^m` is an `R`-linear map and `I ⊆ R` is an ideal, then the image of `I^n` is
contained in `I^m`. -/
theorem map_pi {ι : Type*} [Finite ι] {ι' : Type w} (x : ι → R) (hi : ∀ i, x i ∈ I)
(f : (ι → R) →ₗ[R] ι' → R) (i : ι') : f x i ∈ I
|
Mathlib_RingTheory_Ideal_Quotient
|
case intro
R : Type u
inst✝¹ : CommRing R
I : Ideal R
a b : R
S ι✝ : Type v
ι : Type u_1
inst✝ : Finite ι
ι' : Type w
x : ι → R
hi : ∀ (i : ι), x i ∈ I
f : (ι → R) →ₗ[R] ι' → R
i : ι'
val✝ : Fintype ι
⊢ ∑ x_1 : ι, x x_1 * f (fun j => if x_1 = j then 1 else 0) i ∈ I
|
/-
Copyright (c) 2018 Kenny Lau. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Kenny Lau, Chris Hughes, Mario Carneiro, Anne Baanen
-/
import Mathlib.Algebra.Ring.Fin
import Mathlib.Algebra.Ring.Prod
import Mathlib.LinearAlgebra.Quotient
import Mathlib.RingTheory.Congruence
import Mathlib.RingTheory.Ideal.Basic
import Mathlib.Tactic.FinCases
#align_import ring_theory.ideal.quotient from "leanprover-community/mathlib"@"949dc57e616a621462062668c9f39e4e17b64b69"
/-!
# Ideal quotients
This file defines ideal quotients as a special case of submodule quotients and proves some basic
results about these quotients.
See `Algebra.RingQuot` for quotients of non-commutative rings.
## Main definitions
- `Ideal.Quotient`: the quotient of a commutative ring `R` by an ideal `I : Ideal R`
-/
universe u v w
namespace Ideal
open Set
open BigOperators
variable {R : Type u} [CommRing R] (I : Ideal R) {a b : R}
variable {S : Type v}
-- Note that at present `Ideal` means a left-ideal,
-- so this quotient is only useful in a commutative ring.
-- We should develop quotients by two-sided ideals as well.
/-- The quotient `R/I` of a ring `R` by an ideal `I`.
The ideal quotient of `I` is defined to equal the quotient of `I` as an `R`-submodule of `R`.
This definition is marked `reducible` so that typeclass instances can be shared between
`Ideal.Quotient I` and `Submodule.Quotient I`.
-/
@[reducible]
instance : HasQuotient R (Ideal R) :=
Submodule.hasQuotient
namespace Quotient
variable {I} {x y : R}
instance one (I : Ideal R) : One (R ⧸ I) :=
⟨Submodule.Quotient.mk 1⟩
#align ideal.quotient.has_one Ideal.Quotient.one
/-- On `Ideal`s, `Submodule.quotientRel` is a ring congruence. -/
protected def ringCon (I : Ideal R) : RingCon R :=
{ QuotientAddGroup.con I.toAddSubgroup with
mul' := fun {a₁ b₁ a₂ b₂} h₁ h₂ => by
rw [Submodule.quotientRel_r_def] at h₁ h₂ ⊢
have F := I.add_mem (I.mul_mem_left a₂ h₁) (I.mul_mem_right b₁ h₂)
have : a₁ * a₂ - b₁ * b₂ = a₂ * (a₁ - b₁) + (a₂ - b₂) * b₁ := by
rw [mul_sub, sub_mul, sub_add_sub_cancel, mul_comm, mul_comm b₁]
rwa [← this] at F }
#align ideal.quotient.ring_con Ideal.Quotient.ringCon
instance commRing (I : Ideal R) : CommRing (R ⧸ I) :=
inferInstanceAs (CommRing (Quotient.ringCon I).Quotient)
#align ideal.quotient.comm_ring Ideal.Quotient.commRing
-- Sanity test to make sure no diamonds have emerged in `commRing`
example : (commRing I).toAddCommGroup = Submodule.Quotient.addCommGroup I := rfl
-- this instance is harder to find than the one via `Algebra α (R ⧸ I)`, so use a lower priority
instance (priority := 100) isScalarTower_right {α} [SMul α R] [IsScalarTower α R R] :
IsScalarTower α (R ⧸ I) (R ⧸ I) :=
(Quotient.ringCon I).isScalarTower_right
#align ideal.quotient.is_scalar_tower_right Ideal.Quotient.isScalarTower_right
instance smulCommClass {α} [SMul α R] [IsScalarTower α R R] [SMulCommClass α R R] :
SMulCommClass α (R ⧸ I) (R ⧸ I) :=
(Quotient.ringCon I).smulCommClass
#align ideal.quotient.smul_comm_class Ideal.Quotient.smulCommClass
instance smulCommClass' {α} [SMul α R] [IsScalarTower α R R] [SMulCommClass R α R] :
SMulCommClass (R ⧸ I) α (R ⧸ I) :=
(Quotient.ringCon I).smulCommClass'
#align ideal.quotient.smul_comm_class' Ideal.Quotient.smulCommClass'
/-- The ring homomorphism from a ring `R` to a quotient ring `R/I`. -/
def mk (I : Ideal R) : R →+* R ⧸ I where
toFun a := Submodule.Quotient.mk a
map_zero' := rfl
map_one' := rfl
map_mul' _ _ := rfl
map_add' _ _ := rfl
#align ideal.quotient.mk Ideal.Quotient.mk
instance {I : Ideal R} : Coe R (R ⧸ I) :=
⟨Ideal.Quotient.mk I⟩
/-- Two `RingHom`s from the quotient by an ideal are equal if their
compositions with `Ideal.Quotient.mk'` are equal.
See note [partially-applied ext lemmas]. -/
@[ext 1100]
theorem ringHom_ext [NonAssocSemiring S] ⦃f g : R ⧸ I →+* S⦄ (h : f.comp (mk I) = g.comp (mk I)) :
f = g :=
RingHom.ext fun x => Quotient.inductionOn' x <| (RingHom.congr_fun h : _)
#align ideal.quotient.ring_hom_ext Ideal.Quotient.ringHom_ext
instance inhabited : Inhabited (R ⧸ I) :=
⟨mk I 37⟩
#align ideal.quotient.inhabited Ideal.Quotient.inhabited
protected theorem eq : mk I x = mk I y ↔ x - y ∈ I :=
Submodule.Quotient.eq I
#align ideal.quotient.eq Ideal.Quotient.eq
@[simp]
theorem mk_eq_mk (x : R) : (Submodule.Quotient.mk x : R ⧸ I) = mk I x := rfl
#align ideal.quotient.mk_eq_mk Ideal.Quotient.mk_eq_mk
theorem eq_zero_iff_mem {I : Ideal R} : mk I a = 0 ↔ a ∈ I :=
Submodule.Quotient.mk_eq_zero _
#align ideal.quotient.eq_zero_iff_mem Ideal.Quotient.eq_zero_iff_mem
theorem eq_zero_iff_dvd (x y : R) : Ideal.Quotient.mk (Ideal.span ({x} : Set R)) y = 0 ↔ x ∣ y := by
rw [Ideal.Quotient.eq_zero_iff_mem, Ideal.mem_span_singleton]
-- Porting note: new theorem
theorem mk_eq_mk_iff_sub_mem (x y : R) : mk I x = mk I y ↔ x - y ∈ I := by
rw [← eq_zero_iff_mem, map_sub, sub_eq_zero]
theorem zero_eq_one_iff {I : Ideal R} : (0 : R ⧸ I) = 1 ↔ I = ⊤ :=
eq_comm.trans <| eq_zero_iff_mem.trans (eq_top_iff_one _).symm
#align ideal.quotient.zero_eq_one_iff Ideal.Quotient.zero_eq_one_iff
theorem zero_ne_one_iff {I : Ideal R} : (0 : R ⧸ I) ≠ 1 ↔ I ≠ ⊤ :=
not_congr zero_eq_one_iff
#align ideal.quotient.zero_ne_one_iff Ideal.Quotient.zero_ne_one_iff
protected theorem nontrivial {I : Ideal R} (hI : I ≠ ⊤) : Nontrivial (R ⧸ I) :=
⟨⟨0, 1, zero_ne_one_iff.2 hI⟩⟩
#align ideal.quotient.nontrivial Ideal.Quotient.nontrivial
theorem subsingleton_iff {I : Ideal R} : Subsingleton (R ⧸ I) ↔ I = ⊤ := by
rw [eq_top_iff_one, ← subsingleton_iff_zero_eq_one, eq_comm, ← (mk I).map_one,
Quotient.eq_zero_iff_mem]
#align ideal.quotient.subsingleton_iff Ideal.Quotient.subsingleton_iff
instance : Unique (R ⧸ (⊤ : Ideal R)) :=
⟨⟨0⟩, by rintro ⟨x⟩; exact Quotient.eq_zero_iff_mem.mpr Submodule.mem_top⟩
theorem mk_surjective : Function.Surjective (mk I) := fun y =>
Quotient.inductionOn' y fun x => Exists.intro x rfl
#align ideal.quotient.mk_surjective Ideal.Quotient.mk_surjective
instance : RingHomSurjective (mk I) :=
⟨mk_surjective⟩
/-- If `I` is an ideal of a commutative ring `R`, if `q : R → R/I` is the quotient map, and if
`s ⊆ R` is a subset, then `q⁻¹(q(s)) = ⋃ᵢ(i + s)`, the union running over all `i ∈ I`. -/
theorem quotient_ring_saturate (I : Ideal R) (s : Set R) :
mk I ⁻¹' (mk I '' s) = ⋃ x : I, (fun y => x.1 + y) '' s := by
ext x
simp only [mem_preimage, mem_image, mem_iUnion, Ideal.Quotient.eq]
exact
⟨fun ⟨a, a_in, h⟩ => ⟨⟨_, I.neg_mem h⟩, a, a_in, by simp⟩, fun ⟨⟨i, hi⟩, a, ha, Eq⟩ =>
⟨a, ha, by rw [← Eq, sub_add_eq_sub_sub_swap, sub_self, zero_sub]; exact I.neg_mem hi⟩⟩
#align ideal.quotient.quotient_ring_saturate Ideal.Quotient.quotient_ring_saturate
instance noZeroDivisors (I : Ideal R) [hI : I.IsPrime] : NoZeroDivisors (R ⧸ I) where
eq_zero_or_eq_zero_of_mul_eq_zero {a b} := Quotient.inductionOn₂' a b fun {_ _} hab =>
(hI.mem_or_mem (eq_zero_iff_mem.1 hab)).elim (Or.inl ∘ eq_zero_iff_mem.2)
(Or.inr ∘ eq_zero_iff_mem.2)
#align ideal.quotient.no_zero_divisors Ideal.Quotient.noZeroDivisors
instance isDomain (I : Ideal R) [hI : I.IsPrime] : IsDomain (R ⧸ I) :=
let _ := Quotient.nontrivial hI.1
NoZeroDivisors.to_isDomain _
#align ideal.quotient.is_domain Ideal.Quotient.isDomain
theorem isDomain_iff_prime (I : Ideal R) : IsDomain (R ⧸ I) ↔ I.IsPrime := by
refine' ⟨fun H => ⟨zero_ne_one_iff.1 _, fun {x y} h => _⟩, fun h => inferInstance⟩
· haveI : Nontrivial (R ⧸ I) := ⟨H.2.1⟩
exact zero_ne_one
· simp only [← eq_zero_iff_mem, (mk I).map_mul] at h ⊢
haveI := @IsDomain.to_noZeroDivisors (R ⧸ I) _ H
exact eq_zero_or_eq_zero_of_mul_eq_zero h
#align ideal.quotient.is_domain_iff_prime Ideal.Quotient.isDomain_iff_prime
theorem exists_inv {I : Ideal R} [hI : I.IsMaximal] :
∀ {a : R ⧸ I}, a ≠ 0 → ∃ b : R ⧸ I, a * b = 1 := by
rintro ⟨a⟩ h
rcases hI.exists_inv (mt eq_zero_iff_mem.2 h) with ⟨b, c, hc, abc⟩
rw [mul_comm] at abc
refine' ⟨mk _ b, Quot.sound _⟩
simp only [Submodule.quotientRel_r_def]
rw [← eq_sub_iff_add_eq'] at abc
rwa [abc, ← neg_mem_iff (G := R) (H := I), neg_sub] at hc
#align ideal.quotient.exists_inv Ideal.Quotient.exists_inv
open Classical
/-- The quotient by a maximal ideal is a group with zero. This is a `def` rather than `instance`,
since users will have computable inverses in some applications.
See note [reducible non-instances]. -/
@[reducible]
protected noncomputable def groupWithZero (I : Ideal R) [hI : I.IsMaximal] :
GroupWithZero (R ⧸ I) :=
{ inv := fun a => if ha : a = 0 then 0 else Classical.choose (exists_inv ha)
mul_inv_cancel := fun a (ha : a ≠ 0) =>
show a * dite _ _ _ = _ by rw [dif_neg ha]; exact Classical.choose_spec (exists_inv ha)
inv_zero := dif_pos rfl }
#align ideal.quotient.group_with_zero Ideal.Quotient.groupWithZero
/-- The quotient by a maximal ideal is a field. This is a `def` rather than `instance`, since users
will have computable inverses (and `qsmul`, `rat_cast`) in some applications.
See note [reducible non-instances]. -/
@[reducible]
protected noncomputable def field (I : Ideal R) [hI : I.IsMaximal] : Field (R ⧸ I) :=
{ Quotient.commRing I, Quotient.groupWithZero I with }
#align ideal.quotient.field Ideal.Quotient.field
/-- If the quotient by an ideal is a field, then the ideal is maximal. -/
theorem maximal_of_isField (I : Ideal R) (hqf : IsField (R ⧸ I)) : I.IsMaximal := by
apply Ideal.isMaximal_iff.2
constructor
· intro h
rcases hqf.exists_pair_ne with ⟨⟨x⟩, ⟨y⟩, hxy⟩
exact hxy (Ideal.Quotient.eq.2 (mul_one (x - y) ▸ I.mul_mem_left _ h))
· intro J x hIJ hxnI hxJ
rcases hqf.mul_inv_cancel (mt Ideal.Quotient.eq_zero_iff_mem.1 hxnI) with ⟨⟨y⟩, hy⟩
rw [← zero_add (1 : R), ← sub_self (x * y), sub_add]
refine' J.sub_mem (J.mul_mem_right _ hxJ) (hIJ (Ideal.Quotient.eq.1 hy))
#align ideal.quotient.maximal_of_is_field Ideal.Quotient.maximal_of_isField
/-- The quotient of a ring by an ideal is a field iff the ideal is maximal. -/
theorem maximal_ideal_iff_isField_quotient (I : Ideal R) : I.IsMaximal ↔ IsField (R ⧸ I) :=
⟨fun h =>
let _i := @Quotient.field _ _ I h
Field.toIsField _,
maximal_of_isField _⟩
#align ideal.quotient.maximal_ideal_iff_is_field_quotient Ideal.Quotient.maximal_ideal_iff_isField_quotient
variable [CommRing S]
/-- Given a ring homomorphism `f : R →+* S` sending all elements of an ideal to zero,
lift it to the quotient by this ideal. -/
def lift (I : Ideal R) (f : R →+* S) (H : ∀ a : R, a ∈ I → f a = 0) : R ⧸ I →+* S :=
{ QuotientAddGroup.lift I.toAddSubgroup f.toAddMonoidHom H with
map_one' := f.map_one
map_mul' := fun a₁ a₂ => Quotient.inductionOn₂' a₁ a₂ f.map_mul }
#align ideal.quotient.lift Ideal.Quotient.lift
@[simp]
theorem lift_mk (I : Ideal R) (f : R →+* S) (H : ∀ a : R, a ∈ I → f a = 0) :
lift I f H (mk I a) = f a :=
rfl
#align ideal.quotient.lift_mk Ideal.Quotient.lift_mk
theorem lift_surjective_of_surjective (I : Ideal R) {f : R →+* S} (H : ∀ a : R, a ∈ I → f a = 0)
(hf : Function.Surjective f) : Function.Surjective (Ideal.Quotient.lift I f H) := by
intro y
obtain ⟨x, rfl⟩ := hf y
use Ideal.Quotient.mk I x
simp only [Ideal.Quotient.lift_mk]
#align ideal.quotient.lift_surjective_of_surjective Ideal.Quotient.lift_surjective_of_surjective
/-- The ring homomorphism from the quotient by a smaller ideal to the quotient by a larger ideal.
This is the `Ideal.Quotient` version of `Quot.Factor` -/
def factor (S T : Ideal R) (H : S ≤ T) : R ⧸ S →+* R ⧸ T :=
Ideal.Quotient.lift S (mk T) fun _ hx => eq_zero_iff_mem.2 (H hx)
#align ideal.quotient.factor Ideal.Quotient.factor
@[simp]
theorem factor_mk (S T : Ideal R) (H : S ≤ T) (x : R) : factor S T H (mk S x) = mk T x :=
rfl
#align ideal.quotient.factor_mk Ideal.Quotient.factor_mk
@[simp]
theorem factor_comp_mk (S T : Ideal R) (H : S ≤ T) : (factor S T H).comp (mk S) = mk T := by
ext x
rw [RingHom.comp_apply, factor_mk]
#align ideal.quotient.factor_comp_mk Ideal.Quotient.factor_comp_mk
end Quotient
/-- Quotienting by equal ideals gives equivalent rings.
See also `Submodule.quotEquivOfEq` and `Ideal.quotientEquivAlgOfEq`.
-/
def quotEquivOfEq {R : Type*} [CommRing R] {I J : Ideal R} (h : I = J) : R ⧸ I ≃+* R ⧸ J :=
{ Submodule.quotEquivOfEq I J h with
map_mul' := by
rintro ⟨x⟩ ⟨y⟩
rfl }
#align ideal.quot_equiv_of_eq Ideal.quotEquivOfEq
@[simp]
theorem quotEquivOfEq_mk {R : Type*} [CommRing R] {I J : Ideal R} (h : I = J) (x : R) :
quotEquivOfEq h (Ideal.Quotient.mk I x) = Ideal.Quotient.mk J x :=
rfl
#align ideal.quot_equiv_of_eq_mk Ideal.quotEquivOfEq_mk
@[simp]
theorem quotEquivOfEq_symm {R : Type*} [CommRing R] {I J : Ideal R} (h : I = J) :
(Ideal.quotEquivOfEq h).symm = Ideal.quotEquivOfEq h.symm := by ext; rfl
#align ideal.quot_equiv_of_eq_symm Ideal.quotEquivOfEq_symm
section Pi
variable (ι : Type v)
/-- `R^n/I^n` is a `R/I`-module. -/
instance modulePi : Module (R ⧸ I) ((ι → R) ⧸ I.pi ι) where
smul c m :=
Quotient.liftOn₂' c m (fun r m => Submodule.Quotient.mk <| r • m) $ by
intro c₁ m₁ c₂ m₂ hc hm
apply Ideal.Quotient.eq.2
rw [Submodule.quotientRel_r_def] at hc hm
intro i
exact I.mul_sub_mul_mem hc (hm i)
one_smul := by
rintro ⟨a⟩
convert_to Ideal.Quotient.mk (I.pi ι) _ = Ideal.Quotient.mk (I.pi ι) _
congr with i; exact one_mul (a i)
mul_smul := by
rintro ⟨a⟩ ⟨b⟩ ⟨c⟩
convert_to Ideal.Quotient.mk (I.pi ι) _ = Ideal.Quotient.mk (I.pi ι) _
congr 1; funext i; exact mul_assoc a b (c i)
smul_add := by
rintro ⟨a⟩ ⟨b⟩ ⟨c⟩
convert_to Ideal.Quotient.mk (I.pi ι) _ = Ideal.Quotient.mk (I.pi ι) _
congr with i; exact mul_add a (b i) (c i)
smul_zero := by
rintro ⟨a⟩
convert_to Ideal.Quotient.mk (I.pi ι) _ = Ideal.Quotient.mk (I.pi ι) _
congr with _; exact mul_zero a
add_smul := by
rintro ⟨a⟩ ⟨b⟩ ⟨c⟩
convert_to Ideal.Quotient.mk (I.pi ι) _ = Ideal.Quotient.mk (I.pi ι) _
congr with i; exact add_mul a b (c i)
zero_smul := by
rintro ⟨a⟩
convert_to Ideal.Quotient.mk (I.pi ι) _ = Ideal.Quotient.mk (I.pi ι) _
congr with i; exact zero_mul (a i)
#align ideal.module_pi Ideal.modulePi
/-- `R^n/I^n` is isomorphic to `(R/I)^n` as an `R/I`-module. -/
noncomputable def piQuotEquiv : ((ι → R) ⧸ I.pi ι) ≃ₗ[R ⧸ I] ι → (R ⧸ I) where
toFun := fun x ↦
Quotient.liftOn' x (fun f i => Ideal.Quotient.mk I (f i)) fun a b hab =>
funext fun i => (Submodule.Quotient.eq' _).2 (QuotientAddGroup.leftRel_apply.mp hab i)
map_add' := by rintro ⟨_⟩ ⟨_⟩; rfl
map_smul' := by rintro ⟨_⟩ ⟨_⟩; rfl
invFun := fun x ↦ Ideal.Quotient.mk (I.pi ι) fun i ↦ Quotient.out' (x i)
left_inv := by
rintro ⟨x⟩
exact Ideal.Quotient.eq.2 fun i => Ideal.Quotient.eq.1 (Quotient.out_eq' _)
right_inv := by
intro x
ext i
obtain ⟨_, _⟩ := @Quot.exists_rep _ _ (x i)
convert Quotient.out_eq' (x i)
#align ideal.pi_quot_equiv Ideal.piQuotEquiv
/-- If `f : R^n → R^m` is an `R`-linear map and `I ⊆ R` is an ideal, then the image of `I^n` is
contained in `I^m`. -/
theorem map_pi {ι : Type*} [Finite ι] {ι' : Type w} (x : ι → R) (hi : ∀ i, x i ∈ I)
(f : (ι → R) →ₗ[R] ι' → R) (i : ι') : f x i ∈ I := by
classical
cases nonempty_fintype ι
rw [pi_eq_sum_univ x]
simp only [Finset.sum_apply, smul_eq_mul, map_sum, Pi.smul_apply, map_smul]
|
exact I.sum_mem fun j _ => I.mul_mem_right _ (hi j)
|
/-- If `f : R^n → R^m` is an `R`-linear map and `I ⊆ R` is an ideal, then the image of `I^n` is
contained in `I^m`. -/
theorem map_pi {ι : Type*} [Finite ι] {ι' : Type w} (x : ι → R) (hi : ∀ i, x i ∈ I)
(f : (ι → R) →ₗ[R] ι' → R) (i : ι') : f x i ∈ I := by
classical
cases nonempty_fintype ι
rw [pi_eq_sum_univ x]
simp only [Finset.sum_apply, smul_eq_mul, map_sum, Pi.smul_apply, map_smul]
|
Mathlib.RingTheory.Ideal.Quotient.377_0.TwNAv7Pc4PYOWjX
|
/-- If `f : R^n → R^m` is an `R`-linear map and `I ⊆ R` is an ideal, then the image of `I^n` is
contained in `I^m`. -/
theorem map_pi {ι : Type*} [Finite ι] {ι' : Type w} (x : ι → R) (hi : ∀ i, x i ∈ I)
(f : (ι → R) →ₗ[R] ι' → R) (i : ι') : f x i ∈ I
|
Mathlib_RingTheory_Ideal_Quotient
|
R : Type u_1
inst✝ : Ring R
p q : R[X]
comm : leadingCoeff p * leadingCoeff q = leadingCoeff q * leadingCoeff p
h : natDegree p ≤ natDegree q
hq : 0 < natDegree q
⊢ natDegree (cancelLeads p q) < natDegree q
|
/-
Copyright (c) 2020 Aaron Anderson. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Aaron Anderson
-/
import Mathlib.Data.Polynomial.Degree.Definitions
import Mathlib.Data.Polynomial.Degree.Lemmas
import Mathlib.Tactic.ComputeDegree
#align_import data.polynomial.cancel_leads from "leanprover-community/mathlib"@"70fd9563a21e7b963887c9360bd29b2393e6225a"
/-!
# Cancel the leading terms of two polynomials
## Definition
* `cancelLeads p q`: the polynomial formed by multiplying `p` and `q` by monomials so that they
have the same leading term, and then subtracting.
## Main Results
The degree of `cancelLeads` is less than that of the larger of the two polynomials being cancelled.
Thus it is useful for induction or minimal-degree arguments.
-/
namespace Polynomial
noncomputable section
open Polynomial
variable {R : Type*}
section Ring
variable [Ring R] (p q : R[X])
/-- `cancelLeads p q` is formed by multiplying `p` and `q` by monomials so that they
have the same leading term, and then subtracting. -/
def cancelLeads : R[X] :=
C p.leadingCoeff * X ^ (p.natDegree - q.natDegree) * q -
C q.leadingCoeff * X ^ (q.natDegree - p.natDegree) * p
#align polynomial.cancel_leads Polynomial.cancelLeads
variable {p q}
@[simp]
theorem neg_cancelLeads : -p.cancelLeads q = q.cancelLeads p :=
neg_sub _ _
#align polynomial.neg_cancel_leads Polynomial.neg_cancelLeads
theorem natDegree_cancelLeads_lt_of_natDegree_le_natDegree_of_comm
(comm : p.leadingCoeff * q.leadingCoeff = q.leadingCoeff * p.leadingCoeff)
(h : p.natDegree ≤ q.natDegree) (hq : 0 < q.natDegree) :
(p.cancelLeads q).natDegree < q.natDegree := by
|
by_cases hp : p = 0
|
theorem natDegree_cancelLeads_lt_of_natDegree_le_natDegree_of_comm
(comm : p.leadingCoeff * q.leadingCoeff = q.leadingCoeff * p.leadingCoeff)
(h : p.natDegree ≤ q.natDegree) (hq : 0 < q.natDegree) :
(p.cancelLeads q).natDegree < q.natDegree := by
|
Mathlib.Data.Polynomial.CancelLeads.52_0.8Zq2gl3suYkxnmI
|
theorem natDegree_cancelLeads_lt_of_natDegree_le_natDegree_of_comm
(comm : p.leadingCoeff * q.leadingCoeff = q.leadingCoeff * p.leadingCoeff)
(h : p.natDegree ≤ q.natDegree) (hq : 0 < q.natDegree) :
(p.cancelLeads q).natDegree < q.natDegree
|
Mathlib_Data_Polynomial_CancelLeads
|
case pos
R : Type u_1
inst✝ : Ring R
p q : R[X]
comm : leadingCoeff p * leadingCoeff q = leadingCoeff q * leadingCoeff p
h : natDegree p ≤ natDegree q
hq : 0 < natDegree q
hp : p = 0
⊢ natDegree (cancelLeads p q) < natDegree q
|
/-
Copyright (c) 2020 Aaron Anderson. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Aaron Anderson
-/
import Mathlib.Data.Polynomial.Degree.Definitions
import Mathlib.Data.Polynomial.Degree.Lemmas
import Mathlib.Tactic.ComputeDegree
#align_import data.polynomial.cancel_leads from "leanprover-community/mathlib"@"70fd9563a21e7b963887c9360bd29b2393e6225a"
/-!
# Cancel the leading terms of two polynomials
## Definition
* `cancelLeads p q`: the polynomial formed by multiplying `p` and `q` by monomials so that they
have the same leading term, and then subtracting.
## Main Results
The degree of `cancelLeads` is less than that of the larger of the two polynomials being cancelled.
Thus it is useful for induction or minimal-degree arguments.
-/
namespace Polynomial
noncomputable section
open Polynomial
variable {R : Type*}
section Ring
variable [Ring R] (p q : R[X])
/-- `cancelLeads p q` is formed by multiplying `p` and `q` by monomials so that they
have the same leading term, and then subtracting. -/
def cancelLeads : R[X] :=
C p.leadingCoeff * X ^ (p.natDegree - q.natDegree) * q -
C q.leadingCoeff * X ^ (q.natDegree - p.natDegree) * p
#align polynomial.cancel_leads Polynomial.cancelLeads
variable {p q}
@[simp]
theorem neg_cancelLeads : -p.cancelLeads q = q.cancelLeads p :=
neg_sub _ _
#align polynomial.neg_cancel_leads Polynomial.neg_cancelLeads
theorem natDegree_cancelLeads_lt_of_natDegree_le_natDegree_of_comm
(comm : p.leadingCoeff * q.leadingCoeff = q.leadingCoeff * p.leadingCoeff)
(h : p.natDegree ≤ q.natDegree) (hq : 0 < q.natDegree) :
(p.cancelLeads q).natDegree < q.natDegree := by
by_cases hp : p = 0
·
|
convert hq
|
theorem natDegree_cancelLeads_lt_of_natDegree_le_natDegree_of_comm
(comm : p.leadingCoeff * q.leadingCoeff = q.leadingCoeff * p.leadingCoeff)
(h : p.natDegree ≤ q.natDegree) (hq : 0 < q.natDegree) :
(p.cancelLeads q).natDegree < q.natDegree := by
by_cases hp : p = 0
·
|
Mathlib.Data.Polynomial.CancelLeads.52_0.8Zq2gl3suYkxnmI
|
theorem natDegree_cancelLeads_lt_of_natDegree_le_natDegree_of_comm
(comm : p.leadingCoeff * q.leadingCoeff = q.leadingCoeff * p.leadingCoeff)
(h : p.natDegree ≤ q.natDegree) (hq : 0 < q.natDegree) :
(p.cancelLeads q).natDegree < q.natDegree
|
Mathlib_Data_Polynomial_CancelLeads
|
case h.e'_3
R : Type u_1
inst✝ : Ring R
p q : R[X]
comm : leadingCoeff p * leadingCoeff q = leadingCoeff q * leadingCoeff p
h : natDegree p ≤ natDegree q
hq : 0 < natDegree q
hp : p = 0
⊢ natDegree (cancelLeads p q) = 0
|
/-
Copyright (c) 2020 Aaron Anderson. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Aaron Anderson
-/
import Mathlib.Data.Polynomial.Degree.Definitions
import Mathlib.Data.Polynomial.Degree.Lemmas
import Mathlib.Tactic.ComputeDegree
#align_import data.polynomial.cancel_leads from "leanprover-community/mathlib"@"70fd9563a21e7b963887c9360bd29b2393e6225a"
/-!
# Cancel the leading terms of two polynomials
## Definition
* `cancelLeads p q`: the polynomial formed by multiplying `p` and `q` by monomials so that they
have the same leading term, and then subtracting.
## Main Results
The degree of `cancelLeads` is less than that of the larger of the two polynomials being cancelled.
Thus it is useful for induction or minimal-degree arguments.
-/
namespace Polynomial
noncomputable section
open Polynomial
variable {R : Type*}
section Ring
variable [Ring R] (p q : R[X])
/-- `cancelLeads p q` is formed by multiplying `p` and `q` by monomials so that they
have the same leading term, and then subtracting. -/
def cancelLeads : R[X] :=
C p.leadingCoeff * X ^ (p.natDegree - q.natDegree) * q -
C q.leadingCoeff * X ^ (q.natDegree - p.natDegree) * p
#align polynomial.cancel_leads Polynomial.cancelLeads
variable {p q}
@[simp]
theorem neg_cancelLeads : -p.cancelLeads q = q.cancelLeads p :=
neg_sub _ _
#align polynomial.neg_cancel_leads Polynomial.neg_cancelLeads
theorem natDegree_cancelLeads_lt_of_natDegree_le_natDegree_of_comm
(comm : p.leadingCoeff * q.leadingCoeff = q.leadingCoeff * p.leadingCoeff)
(h : p.natDegree ≤ q.natDegree) (hq : 0 < q.natDegree) :
(p.cancelLeads q).natDegree < q.natDegree := by
by_cases hp : p = 0
· convert hq
|
simp [hp, cancelLeads]
|
theorem natDegree_cancelLeads_lt_of_natDegree_le_natDegree_of_comm
(comm : p.leadingCoeff * q.leadingCoeff = q.leadingCoeff * p.leadingCoeff)
(h : p.natDegree ≤ q.natDegree) (hq : 0 < q.natDegree) :
(p.cancelLeads q).natDegree < q.natDegree := by
by_cases hp : p = 0
· convert hq
|
Mathlib.Data.Polynomial.CancelLeads.52_0.8Zq2gl3suYkxnmI
|
theorem natDegree_cancelLeads_lt_of_natDegree_le_natDegree_of_comm
(comm : p.leadingCoeff * q.leadingCoeff = q.leadingCoeff * p.leadingCoeff)
(h : p.natDegree ≤ q.natDegree) (hq : 0 < q.natDegree) :
(p.cancelLeads q).natDegree < q.natDegree
|
Mathlib_Data_Polynomial_CancelLeads
|
case neg
R : Type u_1
inst✝ : Ring R
p q : R[X]
comm : leadingCoeff p * leadingCoeff q = leadingCoeff q * leadingCoeff p
h : natDegree p ≤ natDegree q
hq : 0 < natDegree q
hp : ¬p = 0
⊢ natDegree (cancelLeads p q) < natDegree q
|
/-
Copyright (c) 2020 Aaron Anderson. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Aaron Anderson
-/
import Mathlib.Data.Polynomial.Degree.Definitions
import Mathlib.Data.Polynomial.Degree.Lemmas
import Mathlib.Tactic.ComputeDegree
#align_import data.polynomial.cancel_leads from "leanprover-community/mathlib"@"70fd9563a21e7b963887c9360bd29b2393e6225a"
/-!
# Cancel the leading terms of two polynomials
## Definition
* `cancelLeads p q`: the polynomial formed by multiplying `p` and `q` by monomials so that they
have the same leading term, and then subtracting.
## Main Results
The degree of `cancelLeads` is less than that of the larger of the two polynomials being cancelled.
Thus it is useful for induction or minimal-degree arguments.
-/
namespace Polynomial
noncomputable section
open Polynomial
variable {R : Type*}
section Ring
variable [Ring R] (p q : R[X])
/-- `cancelLeads p q` is formed by multiplying `p` and `q` by monomials so that they
have the same leading term, and then subtracting. -/
def cancelLeads : R[X] :=
C p.leadingCoeff * X ^ (p.natDegree - q.natDegree) * q -
C q.leadingCoeff * X ^ (q.natDegree - p.natDegree) * p
#align polynomial.cancel_leads Polynomial.cancelLeads
variable {p q}
@[simp]
theorem neg_cancelLeads : -p.cancelLeads q = q.cancelLeads p :=
neg_sub _ _
#align polynomial.neg_cancel_leads Polynomial.neg_cancelLeads
theorem natDegree_cancelLeads_lt_of_natDegree_le_natDegree_of_comm
(comm : p.leadingCoeff * q.leadingCoeff = q.leadingCoeff * p.leadingCoeff)
(h : p.natDegree ≤ q.natDegree) (hq : 0 < q.natDegree) :
(p.cancelLeads q).natDegree < q.natDegree := by
by_cases hp : p = 0
· convert hq
simp [hp, cancelLeads]
|
rw [cancelLeads, sub_eq_add_neg, tsub_eq_zero_iff_le.mpr h, pow_zero, mul_one]
|
theorem natDegree_cancelLeads_lt_of_natDegree_le_natDegree_of_comm
(comm : p.leadingCoeff * q.leadingCoeff = q.leadingCoeff * p.leadingCoeff)
(h : p.natDegree ≤ q.natDegree) (hq : 0 < q.natDegree) :
(p.cancelLeads q).natDegree < q.natDegree := by
by_cases hp : p = 0
· convert hq
simp [hp, cancelLeads]
|
Mathlib.Data.Polynomial.CancelLeads.52_0.8Zq2gl3suYkxnmI
|
theorem natDegree_cancelLeads_lt_of_natDegree_le_natDegree_of_comm
(comm : p.leadingCoeff * q.leadingCoeff = q.leadingCoeff * p.leadingCoeff)
(h : p.natDegree ≤ q.natDegree) (hq : 0 < q.natDegree) :
(p.cancelLeads q).natDegree < q.natDegree
|
Mathlib_Data_Polynomial_CancelLeads
|
case neg
R : Type u_1
inst✝ : Ring R
p q : R[X]
comm : leadingCoeff p * leadingCoeff q = leadingCoeff q * leadingCoeff p
h : natDegree p ≤ natDegree q
hq : 0 < natDegree q
hp : ¬p = 0
⊢ natDegree (C (leadingCoeff p) * q + -(C (leadingCoeff q) * X ^ (natDegree q - natDegree p) * p)) < natDegree q
|
/-
Copyright (c) 2020 Aaron Anderson. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Aaron Anderson
-/
import Mathlib.Data.Polynomial.Degree.Definitions
import Mathlib.Data.Polynomial.Degree.Lemmas
import Mathlib.Tactic.ComputeDegree
#align_import data.polynomial.cancel_leads from "leanprover-community/mathlib"@"70fd9563a21e7b963887c9360bd29b2393e6225a"
/-!
# Cancel the leading terms of two polynomials
## Definition
* `cancelLeads p q`: the polynomial formed by multiplying `p` and `q` by monomials so that they
have the same leading term, and then subtracting.
## Main Results
The degree of `cancelLeads` is less than that of the larger of the two polynomials being cancelled.
Thus it is useful for induction or minimal-degree arguments.
-/
namespace Polynomial
noncomputable section
open Polynomial
variable {R : Type*}
section Ring
variable [Ring R] (p q : R[X])
/-- `cancelLeads p q` is formed by multiplying `p` and `q` by monomials so that they
have the same leading term, and then subtracting. -/
def cancelLeads : R[X] :=
C p.leadingCoeff * X ^ (p.natDegree - q.natDegree) * q -
C q.leadingCoeff * X ^ (q.natDegree - p.natDegree) * p
#align polynomial.cancel_leads Polynomial.cancelLeads
variable {p q}
@[simp]
theorem neg_cancelLeads : -p.cancelLeads q = q.cancelLeads p :=
neg_sub _ _
#align polynomial.neg_cancel_leads Polynomial.neg_cancelLeads
theorem natDegree_cancelLeads_lt_of_natDegree_le_natDegree_of_comm
(comm : p.leadingCoeff * q.leadingCoeff = q.leadingCoeff * p.leadingCoeff)
(h : p.natDegree ≤ q.natDegree) (hq : 0 < q.natDegree) :
(p.cancelLeads q).natDegree < q.natDegree := by
by_cases hp : p = 0
· convert hq
simp [hp, cancelLeads]
rw [cancelLeads, sub_eq_add_neg, tsub_eq_zero_iff_le.mpr h, pow_zero, mul_one]
|
by_cases h0 :
C p.leadingCoeff * q + -(C q.leadingCoeff * X ^ (q.natDegree - p.natDegree) * p) = 0
|
theorem natDegree_cancelLeads_lt_of_natDegree_le_natDegree_of_comm
(comm : p.leadingCoeff * q.leadingCoeff = q.leadingCoeff * p.leadingCoeff)
(h : p.natDegree ≤ q.natDegree) (hq : 0 < q.natDegree) :
(p.cancelLeads q).natDegree < q.natDegree := by
by_cases hp : p = 0
· convert hq
simp [hp, cancelLeads]
rw [cancelLeads, sub_eq_add_neg, tsub_eq_zero_iff_le.mpr h, pow_zero, mul_one]
|
Mathlib.Data.Polynomial.CancelLeads.52_0.8Zq2gl3suYkxnmI
|
theorem natDegree_cancelLeads_lt_of_natDegree_le_natDegree_of_comm
(comm : p.leadingCoeff * q.leadingCoeff = q.leadingCoeff * p.leadingCoeff)
(h : p.natDegree ≤ q.natDegree) (hq : 0 < q.natDegree) :
(p.cancelLeads q).natDegree < q.natDegree
|
Mathlib_Data_Polynomial_CancelLeads
|
case pos
R : Type u_1
inst✝ : Ring R
p q : R[X]
comm : leadingCoeff p * leadingCoeff q = leadingCoeff q * leadingCoeff p
h : natDegree p ≤ natDegree q
hq : 0 < natDegree q
hp : ¬p = 0
h0 : C (leadingCoeff p) * q + -(C (leadingCoeff q) * X ^ (natDegree q - natDegree p) * p) = 0
⊢ natDegree (C (leadingCoeff p) * q + -(C (leadingCoeff q) * X ^ (natDegree q - natDegree p) * p)) < natDegree q
|
/-
Copyright (c) 2020 Aaron Anderson. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Aaron Anderson
-/
import Mathlib.Data.Polynomial.Degree.Definitions
import Mathlib.Data.Polynomial.Degree.Lemmas
import Mathlib.Tactic.ComputeDegree
#align_import data.polynomial.cancel_leads from "leanprover-community/mathlib"@"70fd9563a21e7b963887c9360bd29b2393e6225a"
/-!
# Cancel the leading terms of two polynomials
## Definition
* `cancelLeads p q`: the polynomial formed by multiplying `p` and `q` by monomials so that they
have the same leading term, and then subtracting.
## Main Results
The degree of `cancelLeads` is less than that of the larger of the two polynomials being cancelled.
Thus it is useful for induction or minimal-degree arguments.
-/
namespace Polynomial
noncomputable section
open Polynomial
variable {R : Type*}
section Ring
variable [Ring R] (p q : R[X])
/-- `cancelLeads p q` is formed by multiplying `p` and `q` by monomials so that they
have the same leading term, and then subtracting. -/
def cancelLeads : R[X] :=
C p.leadingCoeff * X ^ (p.natDegree - q.natDegree) * q -
C q.leadingCoeff * X ^ (q.natDegree - p.natDegree) * p
#align polynomial.cancel_leads Polynomial.cancelLeads
variable {p q}
@[simp]
theorem neg_cancelLeads : -p.cancelLeads q = q.cancelLeads p :=
neg_sub _ _
#align polynomial.neg_cancel_leads Polynomial.neg_cancelLeads
theorem natDegree_cancelLeads_lt_of_natDegree_le_natDegree_of_comm
(comm : p.leadingCoeff * q.leadingCoeff = q.leadingCoeff * p.leadingCoeff)
(h : p.natDegree ≤ q.natDegree) (hq : 0 < q.natDegree) :
(p.cancelLeads q).natDegree < q.natDegree := by
by_cases hp : p = 0
· convert hq
simp [hp, cancelLeads]
rw [cancelLeads, sub_eq_add_neg, tsub_eq_zero_iff_le.mpr h, pow_zero, mul_one]
by_cases h0 :
C p.leadingCoeff * q + -(C q.leadingCoeff * X ^ (q.natDegree - p.natDegree) * p) = 0
·
|
exact (le_of_eq (by simp only [h0, natDegree_zero])).trans_lt hq
|
theorem natDegree_cancelLeads_lt_of_natDegree_le_natDegree_of_comm
(comm : p.leadingCoeff * q.leadingCoeff = q.leadingCoeff * p.leadingCoeff)
(h : p.natDegree ≤ q.natDegree) (hq : 0 < q.natDegree) :
(p.cancelLeads q).natDegree < q.natDegree := by
by_cases hp : p = 0
· convert hq
simp [hp, cancelLeads]
rw [cancelLeads, sub_eq_add_neg, tsub_eq_zero_iff_le.mpr h, pow_zero, mul_one]
by_cases h0 :
C p.leadingCoeff * q + -(C q.leadingCoeff * X ^ (q.natDegree - p.natDegree) * p) = 0
·
|
Mathlib.Data.Polynomial.CancelLeads.52_0.8Zq2gl3suYkxnmI
|
theorem natDegree_cancelLeads_lt_of_natDegree_le_natDegree_of_comm
(comm : p.leadingCoeff * q.leadingCoeff = q.leadingCoeff * p.leadingCoeff)
(h : p.natDegree ≤ q.natDegree) (hq : 0 < q.natDegree) :
(p.cancelLeads q).natDegree < q.natDegree
|
Mathlib_Data_Polynomial_CancelLeads
|
R : Type u_1
inst✝ : Ring R
p q : R[X]
comm : leadingCoeff p * leadingCoeff q = leadingCoeff q * leadingCoeff p
h : natDegree p ≤ natDegree q
hq : 0 < natDegree q
hp : ¬p = 0
h0 : C (leadingCoeff p) * q + -(C (leadingCoeff q) * X ^ (natDegree q - natDegree p) * p) = 0
⊢ natDegree (C (leadingCoeff p) * q + -(C (leadingCoeff q) * X ^ (natDegree q - natDegree p) * p)) = 0
|
/-
Copyright (c) 2020 Aaron Anderson. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Aaron Anderson
-/
import Mathlib.Data.Polynomial.Degree.Definitions
import Mathlib.Data.Polynomial.Degree.Lemmas
import Mathlib.Tactic.ComputeDegree
#align_import data.polynomial.cancel_leads from "leanprover-community/mathlib"@"70fd9563a21e7b963887c9360bd29b2393e6225a"
/-!
# Cancel the leading terms of two polynomials
## Definition
* `cancelLeads p q`: the polynomial formed by multiplying `p` and `q` by monomials so that they
have the same leading term, and then subtracting.
## Main Results
The degree of `cancelLeads` is less than that of the larger of the two polynomials being cancelled.
Thus it is useful for induction or minimal-degree arguments.
-/
namespace Polynomial
noncomputable section
open Polynomial
variable {R : Type*}
section Ring
variable [Ring R] (p q : R[X])
/-- `cancelLeads p q` is formed by multiplying `p` and `q` by monomials so that they
have the same leading term, and then subtracting. -/
def cancelLeads : R[X] :=
C p.leadingCoeff * X ^ (p.natDegree - q.natDegree) * q -
C q.leadingCoeff * X ^ (q.natDegree - p.natDegree) * p
#align polynomial.cancel_leads Polynomial.cancelLeads
variable {p q}
@[simp]
theorem neg_cancelLeads : -p.cancelLeads q = q.cancelLeads p :=
neg_sub _ _
#align polynomial.neg_cancel_leads Polynomial.neg_cancelLeads
theorem natDegree_cancelLeads_lt_of_natDegree_le_natDegree_of_comm
(comm : p.leadingCoeff * q.leadingCoeff = q.leadingCoeff * p.leadingCoeff)
(h : p.natDegree ≤ q.natDegree) (hq : 0 < q.natDegree) :
(p.cancelLeads q).natDegree < q.natDegree := by
by_cases hp : p = 0
· convert hq
simp [hp, cancelLeads]
rw [cancelLeads, sub_eq_add_neg, tsub_eq_zero_iff_le.mpr h, pow_zero, mul_one]
by_cases h0 :
C p.leadingCoeff * q + -(C q.leadingCoeff * X ^ (q.natDegree - p.natDegree) * p) = 0
· exact (le_of_eq (by
|
simp only [h0, natDegree_zero]
|
theorem natDegree_cancelLeads_lt_of_natDegree_le_natDegree_of_comm
(comm : p.leadingCoeff * q.leadingCoeff = q.leadingCoeff * p.leadingCoeff)
(h : p.natDegree ≤ q.natDegree) (hq : 0 < q.natDegree) :
(p.cancelLeads q).natDegree < q.natDegree := by
by_cases hp : p = 0
· convert hq
simp [hp, cancelLeads]
rw [cancelLeads, sub_eq_add_neg, tsub_eq_zero_iff_le.mpr h, pow_zero, mul_one]
by_cases h0 :
C p.leadingCoeff * q + -(C q.leadingCoeff * X ^ (q.natDegree - p.natDegree) * p) = 0
· exact (le_of_eq (by
|
Mathlib.Data.Polynomial.CancelLeads.52_0.8Zq2gl3suYkxnmI
|
theorem natDegree_cancelLeads_lt_of_natDegree_le_natDegree_of_comm
(comm : p.leadingCoeff * q.leadingCoeff = q.leadingCoeff * p.leadingCoeff)
(h : p.natDegree ≤ q.natDegree) (hq : 0 < q.natDegree) :
(p.cancelLeads q).natDegree < q.natDegree
|
Mathlib_Data_Polynomial_CancelLeads
|
case neg
R : Type u_1
inst✝ : Ring R
p q : R[X]
comm : leadingCoeff p * leadingCoeff q = leadingCoeff q * leadingCoeff p
h : natDegree p ≤ natDegree q
hq : 0 < natDegree q
hp : ¬p = 0
h0 : ¬C (leadingCoeff p) * q + -(C (leadingCoeff q) * X ^ (natDegree q - natDegree p) * p) = 0
⊢ natDegree (C (leadingCoeff p) * q + -(C (leadingCoeff q) * X ^ (natDegree q - natDegree p) * p)) < natDegree q
|
/-
Copyright (c) 2020 Aaron Anderson. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Aaron Anderson
-/
import Mathlib.Data.Polynomial.Degree.Definitions
import Mathlib.Data.Polynomial.Degree.Lemmas
import Mathlib.Tactic.ComputeDegree
#align_import data.polynomial.cancel_leads from "leanprover-community/mathlib"@"70fd9563a21e7b963887c9360bd29b2393e6225a"
/-!
# Cancel the leading terms of two polynomials
## Definition
* `cancelLeads p q`: the polynomial formed by multiplying `p` and `q` by monomials so that they
have the same leading term, and then subtracting.
## Main Results
The degree of `cancelLeads` is less than that of the larger of the two polynomials being cancelled.
Thus it is useful for induction or minimal-degree arguments.
-/
namespace Polynomial
noncomputable section
open Polynomial
variable {R : Type*}
section Ring
variable [Ring R] (p q : R[X])
/-- `cancelLeads p q` is formed by multiplying `p` and `q` by monomials so that they
have the same leading term, and then subtracting. -/
def cancelLeads : R[X] :=
C p.leadingCoeff * X ^ (p.natDegree - q.natDegree) * q -
C q.leadingCoeff * X ^ (q.natDegree - p.natDegree) * p
#align polynomial.cancel_leads Polynomial.cancelLeads
variable {p q}
@[simp]
theorem neg_cancelLeads : -p.cancelLeads q = q.cancelLeads p :=
neg_sub _ _
#align polynomial.neg_cancel_leads Polynomial.neg_cancelLeads
theorem natDegree_cancelLeads_lt_of_natDegree_le_natDegree_of_comm
(comm : p.leadingCoeff * q.leadingCoeff = q.leadingCoeff * p.leadingCoeff)
(h : p.natDegree ≤ q.natDegree) (hq : 0 < q.natDegree) :
(p.cancelLeads q).natDegree < q.natDegree := by
by_cases hp : p = 0
· convert hq
simp [hp, cancelLeads]
rw [cancelLeads, sub_eq_add_neg, tsub_eq_zero_iff_le.mpr h, pow_zero, mul_one]
by_cases h0 :
C p.leadingCoeff * q + -(C q.leadingCoeff * X ^ (q.natDegree - p.natDegree) * p) = 0
· exact (le_of_eq (by simp only [h0, natDegree_zero])).trans_lt hq
|
apply lt_of_le_of_ne
|
theorem natDegree_cancelLeads_lt_of_natDegree_le_natDegree_of_comm
(comm : p.leadingCoeff * q.leadingCoeff = q.leadingCoeff * p.leadingCoeff)
(h : p.natDegree ≤ q.natDegree) (hq : 0 < q.natDegree) :
(p.cancelLeads q).natDegree < q.natDegree := by
by_cases hp : p = 0
· convert hq
simp [hp, cancelLeads]
rw [cancelLeads, sub_eq_add_neg, tsub_eq_zero_iff_le.mpr h, pow_zero, mul_one]
by_cases h0 :
C p.leadingCoeff * q + -(C q.leadingCoeff * X ^ (q.natDegree - p.natDegree) * p) = 0
· exact (le_of_eq (by simp only [h0, natDegree_zero])).trans_lt hq
|
Mathlib.Data.Polynomial.CancelLeads.52_0.8Zq2gl3suYkxnmI
|
theorem natDegree_cancelLeads_lt_of_natDegree_le_natDegree_of_comm
(comm : p.leadingCoeff * q.leadingCoeff = q.leadingCoeff * p.leadingCoeff)
(h : p.natDegree ≤ q.natDegree) (hq : 0 < q.natDegree) :
(p.cancelLeads q).natDegree < q.natDegree
|
Mathlib_Data_Polynomial_CancelLeads
|
case neg.a
R : Type u_1
inst✝ : Ring R
p q : R[X]
comm : leadingCoeff p * leadingCoeff q = leadingCoeff q * leadingCoeff p
h : natDegree p ≤ natDegree q
hq : 0 < natDegree q
hp : ¬p = 0
h0 : ¬C (leadingCoeff p) * q + -(C (leadingCoeff q) * X ^ (natDegree q - natDegree p) * p) = 0
⊢ natDegree (C (leadingCoeff p) * q + -(C (leadingCoeff q) * X ^ (natDegree q - natDegree p) * p)) ≤ natDegree q
|
/-
Copyright (c) 2020 Aaron Anderson. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Aaron Anderson
-/
import Mathlib.Data.Polynomial.Degree.Definitions
import Mathlib.Data.Polynomial.Degree.Lemmas
import Mathlib.Tactic.ComputeDegree
#align_import data.polynomial.cancel_leads from "leanprover-community/mathlib"@"70fd9563a21e7b963887c9360bd29b2393e6225a"
/-!
# Cancel the leading terms of two polynomials
## Definition
* `cancelLeads p q`: the polynomial formed by multiplying `p` and `q` by monomials so that they
have the same leading term, and then subtracting.
## Main Results
The degree of `cancelLeads` is less than that of the larger of the two polynomials being cancelled.
Thus it is useful for induction or minimal-degree arguments.
-/
namespace Polynomial
noncomputable section
open Polynomial
variable {R : Type*}
section Ring
variable [Ring R] (p q : R[X])
/-- `cancelLeads p q` is formed by multiplying `p` and `q` by monomials so that they
have the same leading term, and then subtracting. -/
def cancelLeads : R[X] :=
C p.leadingCoeff * X ^ (p.natDegree - q.natDegree) * q -
C q.leadingCoeff * X ^ (q.natDegree - p.natDegree) * p
#align polynomial.cancel_leads Polynomial.cancelLeads
variable {p q}
@[simp]
theorem neg_cancelLeads : -p.cancelLeads q = q.cancelLeads p :=
neg_sub _ _
#align polynomial.neg_cancel_leads Polynomial.neg_cancelLeads
theorem natDegree_cancelLeads_lt_of_natDegree_le_natDegree_of_comm
(comm : p.leadingCoeff * q.leadingCoeff = q.leadingCoeff * p.leadingCoeff)
(h : p.natDegree ≤ q.natDegree) (hq : 0 < q.natDegree) :
(p.cancelLeads q).natDegree < q.natDegree := by
by_cases hp : p = 0
· convert hq
simp [hp, cancelLeads]
rw [cancelLeads, sub_eq_add_neg, tsub_eq_zero_iff_le.mpr h, pow_zero, mul_one]
by_cases h0 :
C p.leadingCoeff * q + -(C q.leadingCoeff * X ^ (q.natDegree - p.natDegree) * p) = 0
· exact (le_of_eq (by simp only [h0, natDegree_zero])).trans_lt hq
apply lt_of_le_of_ne
·
|
compute_degree!
|
theorem natDegree_cancelLeads_lt_of_natDegree_le_natDegree_of_comm
(comm : p.leadingCoeff * q.leadingCoeff = q.leadingCoeff * p.leadingCoeff)
(h : p.natDegree ≤ q.natDegree) (hq : 0 < q.natDegree) :
(p.cancelLeads q).natDegree < q.natDegree := by
by_cases hp : p = 0
· convert hq
simp [hp, cancelLeads]
rw [cancelLeads, sub_eq_add_neg, tsub_eq_zero_iff_le.mpr h, pow_zero, mul_one]
by_cases h0 :
C p.leadingCoeff * q + -(C q.leadingCoeff * X ^ (q.natDegree - p.natDegree) * p) = 0
· exact (le_of_eq (by simp only [h0, natDegree_zero])).trans_lt hq
apply lt_of_le_of_ne
·
|
Mathlib.Data.Polynomial.CancelLeads.52_0.8Zq2gl3suYkxnmI
|
theorem natDegree_cancelLeads_lt_of_natDegree_le_natDegree_of_comm
(comm : p.leadingCoeff * q.leadingCoeff = q.leadingCoeff * p.leadingCoeff)
(h : p.natDegree ≤ q.natDegree) (hq : 0 < q.natDegree) :
(p.cancelLeads q).natDegree < q.natDegree
|
Mathlib_Data_Polynomial_CancelLeads
|
case neg.a.a
R : Type u_1
inst✝ : Ring R
p q : R[X]
comm : leadingCoeff p * leadingCoeff q = leadingCoeff q * leadingCoeff p
h : natDegree p ≤ natDegree q
hq : 0 < natDegree q
hp : ¬p = 0
h0 : ¬C (leadingCoeff p) * q + -(C (leadingCoeff q) * X ^ (natDegree q - natDegree p) * p) = 0
⊢ natDegree q - natDegree p + natDegree p ≤ natDegree q
|
/-
Copyright (c) 2020 Aaron Anderson. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Aaron Anderson
-/
import Mathlib.Data.Polynomial.Degree.Definitions
import Mathlib.Data.Polynomial.Degree.Lemmas
import Mathlib.Tactic.ComputeDegree
#align_import data.polynomial.cancel_leads from "leanprover-community/mathlib"@"70fd9563a21e7b963887c9360bd29b2393e6225a"
/-!
# Cancel the leading terms of two polynomials
## Definition
* `cancelLeads p q`: the polynomial formed by multiplying `p` and `q` by monomials so that they
have the same leading term, and then subtracting.
## Main Results
The degree of `cancelLeads` is less than that of the larger of the two polynomials being cancelled.
Thus it is useful for induction or minimal-degree arguments.
-/
namespace Polynomial
noncomputable section
open Polynomial
variable {R : Type*}
section Ring
variable [Ring R] (p q : R[X])
/-- `cancelLeads p q` is formed by multiplying `p` and `q` by monomials so that they
have the same leading term, and then subtracting. -/
def cancelLeads : R[X] :=
C p.leadingCoeff * X ^ (p.natDegree - q.natDegree) * q -
C q.leadingCoeff * X ^ (q.natDegree - p.natDegree) * p
#align polynomial.cancel_leads Polynomial.cancelLeads
variable {p q}
@[simp]
theorem neg_cancelLeads : -p.cancelLeads q = q.cancelLeads p :=
neg_sub _ _
#align polynomial.neg_cancel_leads Polynomial.neg_cancelLeads
theorem natDegree_cancelLeads_lt_of_natDegree_le_natDegree_of_comm
(comm : p.leadingCoeff * q.leadingCoeff = q.leadingCoeff * p.leadingCoeff)
(h : p.natDegree ≤ q.natDegree) (hq : 0 < q.natDegree) :
(p.cancelLeads q).natDegree < q.natDegree := by
by_cases hp : p = 0
· convert hq
simp [hp, cancelLeads]
rw [cancelLeads, sub_eq_add_neg, tsub_eq_zero_iff_le.mpr h, pow_zero, mul_one]
by_cases h0 :
C p.leadingCoeff * q + -(C q.leadingCoeff * X ^ (q.natDegree - p.natDegree) * p) = 0
· exact (le_of_eq (by simp only [h0, natDegree_zero])).trans_lt hq
apply lt_of_le_of_ne
· compute_degree!
|
rwa [Nat.sub_add_cancel]
|
theorem natDegree_cancelLeads_lt_of_natDegree_le_natDegree_of_comm
(comm : p.leadingCoeff * q.leadingCoeff = q.leadingCoeff * p.leadingCoeff)
(h : p.natDegree ≤ q.natDegree) (hq : 0 < q.natDegree) :
(p.cancelLeads q).natDegree < q.natDegree := by
by_cases hp : p = 0
· convert hq
simp [hp, cancelLeads]
rw [cancelLeads, sub_eq_add_neg, tsub_eq_zero_iff_le.mpr h, pow_zero, mul_one]
by_cases h0 :
C p.leadingCoeff * q + -(C q.leadingCoeff * X ^ (q.natDegree - p.natDegree) * p) = 0
· exact (le_of_eq (by simp only [h0, natDegree_zero])).trans_lt hq
apply lt_of_le_of_ne
· compute_degree!
|
Mathlib.Data.Polynomial.CancelLeads.52_0.8Zq2gl3suYkxnmI
|
theorem natDegree_cancelLeads_lt_of_natDegree_le_natDegree_of_comm
(comm : p.leadingCoeff * q.leadingCoeff = q.leadingCoeff * p.leadingCoeff)
(h : p.natDegree ≤ q.natDegree) (hq : 0 < q.natDegree) :
(p.cancelLeads q).natDegree < q.natDegree
|
Mathlib_Data_Polynomial_CancelLeads
|
case neg.a
R : Type u_1
inst✝ : Ring R
p q : R[X]
comm : leadingCoeff p * leadingCoeff q = leadingCoeff q * leadingCoeff p
h : natDegree p ≤ natDegree q
hq : 0 < natDegree q
hp : ¬p = 0
h0 : ¬C (leadingCoeff p) * q + -(C (leadingCoeff q) * X ^ (natDegree q - natDegree p) * p) = 0
⊢ natDegree (C (leadingCoeff p) * q + -(C (leadingCoeff q) * X ^ (natDegree q - natDegree p) * p)) ≠ natDegree q
|
/-
Copyright (c) 2020 Aaron Anderson. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Aaron Anderson
-/
import Mathlib.Data.Polynomial.Degree.Definitions
import Mathlib.Data.Polynomial.Degree.Lemmas
import Mathlib.Tactic.ComputeDegree
#align_import data.polynomial.cancel_leads from "leanprover-community/mathlib"@"70fd9563a21e7b963887c9360bd29b2393e6225a"
/-!
# Cancel the leading terms of two polynomials
## Definition
* `cancelLeads p q`: the polynomial formed by multiplying `p` and `q` by monomials so that they
have the same leading term, and then subtracting.
## Main Results
The degree of `cancelLeads` is less than that of the larger of the two polynomials being cancelled.
Thus it is useful for induction or minimal-degree arguments.
-/
namespace Polynomial
noncomputable section
open Polynomial
variable {R : Type*}
section Ring
variable [Ring R] (p q : R[X])
/-- `cancelLeads p q` is formed by multiplying `p` and `q` by monomials so that they
have the same leading term, and then subtracting. -/
def cancelLeads : R[X] :=
C p.leadingCoeff * X ^ (p.natDegree - q.natDegree) * q -
C q.leadingCoeff * X ^ (q.natDegree - p.natDegree) * p
#align polynomial.cancel_leads Polynomial.cancelLeads
variable {p q}
@[simp]
theorem neg_cancelLeads : -p.cancelLeads q = q.cancelLeads p :=
neg_sub _ _
#align polynomial.neg_cancel_leads Polynomial.neg_cancelLeads
theorem natDegree_cancelLeads_lt_of_natDegree_le_natDegree_of_comm
(comm : p.leadingCoeff * q.leadingCoeff = q.leadingCoeff * p.leadingCoeff)
(h : p.natDegree ≤ q.natDegree) (hq : 0 < q.natDegree) :
(p.cancelLeads q).natDegree < q.natDegree := by
by_cases hp : p = 0
· convert hq
simp [hp, cancelLeads]
rw [cancelLeads, sub_eq_add_neg, tsub_eq_zero_iff_le.mpr h, pow_zero, mul_one]
by_cases h0 :
C p.leadingCoeff * q + -(C q.leadingCoeff * X ^ (q.natDegree - p.natDegree) * p) = 0
· exact (le_of_eq (by simp only [h0, natDegree_zero])).trans_lt hq
apply lt_of_le_of_ne
· compute_degree!
rwa [Nat.sub_add_cancel]
·
|
contrapose! h0
|
theorem natDegree_cancelLeads_lt_of_natDegree_le_natDegree_of_comm
(comm : p.leadingCoeff * q.leadingCoeff = q.leadingCoeff * p.leadingCoeff)
(h : p.natDegree ≤ q.natDegree) (hq : 0 < q.natDegree) :
(p.cancelLeads q).natDegree < q.natDegree := by
by_cases hp : p = 0
· convert hq
simp [hp, cancelLeads]
rw [cancelLeads, sub_eq_add_neg, tsub_eq_zero_iff_le.mpr h, pow_zero, mul_one]
by_cases h0 :
C p.leadingCoeff * q + -(C q.leadingCoeff * X ^ (q.natDegree - p.natDegree) * p) = 0
· exact (le_of_eq (by simp only [h0, natDegree_zero])).trans_lt hq
apply lt_of_le_of_ne
· compute_degree!
rwa [Nat.sub_add_cancel]
·
|
Mathlib.Data.Polynomial.CancelLeads.52_0.8Zq2gl3suYkxnmI
|
theorem natDegree_cancelLeads_lt_of_natDegree_le_natDegree_of_comm
(comm : p.leadingCoeff * q.leadingCoeff = q.leadingCoeff * p.leadingCoeff)
(h : p.natDegree ≤ q.natDegree) (hq : 0 < q.natDegree) :
(p.cancelLeads q).natDegree < q.natDegree
|
Mathlib_Data_Polynomial_CancelLeads
|
case neg.a
R : Type u_1
inst✝ : Ring R
p q : R[X]
comm : leadingCoeff p * leadingCoeff q = leadingCoeff q * leadingCoeff p
h : natDegree p ≤ natDegree q
hq : 0 < natDegree q
hp : ¬p = 0
h0 : natDegree (C (leadingCoeff p) * q + -(C (leadingCoeff q) * X ^ (natDegree q - natDegree p) * p)) = natDegree q
⊢ C (leadingCoeff p) * q + -(C (leadingCoeff q) * X ^ (natDegree q - natDegree p) * p) = 0
|
/-
Copyright (c) 2020 Aaron Anderson. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Aaron Anderson
-/
import Mathlib.Data.Polynomial.Degree.Definitions
import Mathlib.Data.Polynomial.Degree.Lemmas
import Mathlib.Tactic.ComputeDegree
#align_import data.polynomial.cancel_leads from "leanprover-community/mathlib"@"70fd9563a21e7b963887c9360bd29b2393e6225a"
/-!
# Cancel the leading terms of two polynomials
## Definition
* `cancelLeads p q`: the polynomial formed by multiplying `p` and `q` by monomials so that they
have the same leading term, and then subtracting.
## Main Results
The degree of `cancelLeads` is less than that of the larger of the two polynomials being cancelled.
Thus it is useful for induction or minimal-degree arguments.
-/
namespace Polynomial
noncomputable section
open Polynomial
variable {R : Type*}
section Ring
variable [Ring R] (p q : R[X])
/-- `cancelLeads p q` is formed by multiplying `p` and `q` by monomials so that they
have the same leading term, and then subtracting. -/
def cancelLeads : R[X] :=
C p.leadingCoeff * X ^ (p.natDegree - q.natDegree) * q -
C q.leadingCoeff * X ^ (q.natDegree - p.natDegree) * p
#align polynomial.cancel_leads Polynomial.cancelLeads
variable {p q}
@[simp]
theorem neg_cancelLeads : -p.cancelLeads q = q.cancelLeads p :=
neg_sub _ _
#align polynomial.neg_cancel_leads Polynomial.neg_cancelLeads
theorem natDegree_cancelLeads_lt_of_natDegree_le_natDegree_of_comm
(comm : p.leadingCoeff * q.leadingCoeff = q.leadingCoeff * p.leadingCoeff)
(h : p.natDegree ≤ q.natDegree) (hq : 0 < q.natDegree) :
(p.cancelLeads q).natDegree < q.natDegree := by
by_cases hp : p = 0
· convert hq
simp [hp, cancelLeads]
rw [cancelLeads, sub_eq_add_neg, tsub_eq_zero_iff_le.mpr h, pow_zero, mul_one]
by_cases h0 :
C p.leadingCoeff * q + -(C q.leadingCoeff * X ^ (q.natDegree - p.natDegree) * p) = 0
· exact (le_of_eq (by simp only [h0, natDegree_zero])).trans_lt hq
apply lt_of_le_of_ne
· compute_degree!
rwa [Nat.sub_add_cancel]
· contrapose! h0
|
rw [← leadingCoeff_eq_zero, leadingCoeff, h0, mul_assoc, X_pow_mul, ← tsub_add_cancel_of_le h,
add_comm _ p.natDegree]
|
theorem natDegree_cancelLeads_lt_of_natDegree_le_natDegree_of_comm
(comm : p.leadingCoeff * q.leadingCoeff = q.leadingCoeff * p.leadingCoeff)
(h : p.natDegree ≤ q.natDegree) (hq : 0 < q.natDegree) :
(p.cancelLeads q).natDegree < q.natDegree := by
by_cases hp : p = 0
· convert hq
simp [hp, cancelLeads]
rw [cancelLeads, sub_eq_add_neg, tsub_eq_zero_iff_le.mpr h, pow_zero, mul_one]
by_cases h0 :
C p.leadingCoeff * q + -(C q.leadingCoeff * X ^ (q.natDegree - p.natDegree) * p) = 0
· exact (le_of_eq (by simp only [h0, natDegree_zero])).trans_lt hq
apply lt_of_le_of_ne
· compute_degree!
rwa [Nat.sub_add_cancel]
· contrapose! h0
|
Mathlib.Data.Polynomial.CancelLeads.52_0.8Zq2gl3suYkxnmI
|
theorem natDegree_cancelLeads_lt_of_natDegree_le_natDegree_of_comm
(comm : p.leadingCoeff * q.leadingCoeff = q.leadingCoeff * p.leadingCoeff)
(h : p.natDegree ≤ q.natDegree) (hq : 0 < q.natDegree) :
(p.cancelLeads q).natDegree < q.natDegree
|
Mathlib_Data_Polynomial_CancelLeads
|
case neg.a
R : Type u_1
inst✝ : Ring R
p q : R[X]
comm : leadingCoeff p * leadingCoeff q = leadingCoeff q * leadingCoeff p
h : natDegree p ≤ natDegree q
hq : 0 < natDegree q
hp : ¬p = 0
h0 : natDegree (C (leadingCoeff p) * q + -(C (leadingCoeff q) * X ^ (natDegree q - natDegree p) * p)) = natDegree q
⊢ coeff
(C (leadingCoeff p) * q +
-(C (leadingCoeff q) * (p * X ^ (natDegree p + (natDegree q - natDegree p) - natDegree p))))
(natDegree p + (natDegree q - natDegree p)) =
0
|
/-
Copyright (c) 2020 Aaron Anderson. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Aaron Anderson
-/
import Mathlib.Data.Polynomial.Degree.Definitions
import Mathlib.Data.Polynomial.Degree.Lemmas
import Mathlib.Tactic.ComputeDegree
#align_import data.polynomial.cancel_leads from "leanprover-community/mathlib"@"70fd9563a21e7b963887c9360bd29b2393e6225a"
/-!
# Cancel the leading terms of two polynomials
## Definition
* `cancelLeads p q`: the polynomial formed by multiplying `p` and `q` by monomials so that they
have the same leading term, and then subtracting.
## Main Results
The degree of `cancelLeads` is less than that of the larger of the two polynomials being cancelled.
Thus it is useful for induction or minimal-degree arguments.
-/
namespace Polynomial
noncomputable section
open Polynomial
variable {R : Type*}
section Ring
variable [Ring R] (p q : R[X])
/-- `cancelLeads p q` is formed by multiplying `p` and `q` by monomials so that they
have the same leading term, and then subtracting. -/
def cancelLeads : R[X] :=
C p.leadingCoeff * X ^ (p.natDegree - q.natDegree) * q -
C q.leadingCoeff * X ^ (q.natDegree - p.natDegree) * p
#align polynomial.cancel_leads Polynomial.cancelLeads
variable {p q}
@[simp]
theorem neg_cancelLeads : -p.cancelLeads q = q.cancelLeads p :=
neg_sub _ _
#align polynomial.neg_cancel_leads Polynomial.neg_cancelLeads
theorem natDegree_cancelLeads_lt_of_natDegree_le_natDegree_of_comm
(comm : p.leadingCoeff * q.leadingCoeff = q.leadingCoeff * p.leadingCoeff)
(h : p.natDegree ≤ q.natDegree) (hq : 0 < q.natDegree) :
(p.cancelLeads q).natDegree < q.natDegree := by
by_cases hp : p = 0
· convert hq
simp [hp, cancelLeads]
rw [cancelLeads, sub_eq_add_neg, tsub_eq_zero_iff_le.mpr h, pow_zero, mul_one]
by_cases h0 :
C p.leadingCoeff * q + -(C q.leadingCoeff * X ^ (q.natDegree - p.natDegree) * p) = 0
· exact (le_of_eq (by simp only [h0, natDegree_zero])).trans_lt hq
apply lt_of_le_of_ne
· compute_degree!
rwa [Nat.sub_add_cancel]
· contrapose! h0
rw [← leadingCoeff_eq_zero, leadingCoeff, h0, mul_assoc, X_pow_mul, ← tsub_add_cancel_of_le h,
add_comm _ p.natDegree]
|
simp only [coeff_mul_X_pow, coeff_neg, coeff_C_mul, add_tsub_cancel_left, coeff_add]
|
theorem natDegree_cancelLeads_lt_of_natDegree_le_natDegree_of_comm
(comm : p.leadingCoeff * q.leadingCoeff = q.leadingCoeff * p.leadingCoeff)
(h : p.natDegree ≤ q.natDegree) (hq : 0 < q.natDegree) :
(p.cancelLeads q).natDegree < q.natDegree := by
by_cases hp : p = 0
· convert hq
simp [hp, cancelLeads]
rw [cancelLeads, sub_eq_add_neg, tsub_eq_zero_iff_le.mpr h, pow_zero, mul_one]
by_cases h0 :
C p.leadingCoeff * q + -(C q.leadingCoeff * X ^ (q.natDegree - p.natDegree) * p) = 0
· exact (le_of_eq (by simp only [h0, natDegree_zero])).trans_lt hq
apply lt_of_le_of_ne
· compute_degree!
rwa [Nat.sub_add_cancel]
· contrapose! h0
rw [← leadingCoeff_eq_zero, leadingCoeff, h0, mul_assoc, X_pow_mul, ← tsub_add_cancel_of_le h,
add_comm _ p.natDegree]
|
Mathlib.Data.Polynomial.CancelLeads.52_0.8Zq2gl3suYkxnmI
|
theorem natDegree_cancelLeads_lt_of_natDegree_le_natDegree_of_comm
(comm : p.leadingCoeff * q.leadingCoeff = q.leadingCoeff * p.leadingCoeff)
(h : p.natDegree ≤ q.natDegree) (hq : 0 < q.natDegree) :
(p.cancelLeads q).natDegree < q.natDegree
|
Mathlib_Data_Polynomial_CancelLeads
|
case neg.a
R : Type u_1
inst✝ : Ring R
p q : R[X]
comm : leadingCoeff p * leadingCoeff q = leadingCoeff q * leadingCoeff p
h : natDegree p ≤ natDegree q
hq : 0 < natDegree q
hp : ¬p = 0
h0 : natDegree (C (leadingCoeff p) * q + -(C (leadingCoeff q) * X ^ (natDegree q - natDegree p) * p)) = natDegree q
⊢ leadingCoeff p * coeff q (natDegree p + (natDegree q - natDegree p)) + -(leadingCoeff q * coeff p (natDegree p)) = 0
|
/-
Copyright (c) 2020 Aaron Anderson. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Aaron Anderson
-/
import Mathlib.Data.Polynomial.Degree.Definitions
import Mathlib.Data.Polynomial.Degree.Lemmas
import Mathlib.Tactic.ComputeDegree
#align_import data.polynomial.cancel_leads from "leanprover-community/mathlib"@"70fd9563a21e7b963887c9360bd29b2393e6225a"
/-!
# Cancel the leading terms of two polynomials
## Definition
* `cancelLeads p q`: the polynomial formed by multiplying `p` and `q` by monomials so that they
have the same leading term, and then subtracting.
## Main Results
The degree of `cancelLeads` is less than that of the larger of the two polynomials being cancelled.
Thus it is useful for induction or minimal-degree arguments.
-/
namespace Polynomial
noncomputable section
open Polynomial
variable {R : Type*}
section Ring
variable [Ring R] (p q : R[X])
/-- `cancelLeads p q` is formed by multiplying `p` and `q` by monomials so that they
have the same leading term, and then subtracting. -/
def cancelLeads : R[X] :=
C p.leadingCoeff * X ^ (p.natDegree - q.natDegree) * q -
C q.leadingCoeff * X ^ (q.natDegree - p.natDegree) * p
#align polynomial.cancel_leads Polynomial.cancelLeads
variable {p q}
@[simp]
theorem neg_cancelLeads : -p.cancelLeads q = q.cancelLeads p :=
neg_sub _ _
#align polynomial.neg_cancel_leads Polynomial.neg_cancelLeads
theorem natDegree_cancelLeads_lt_of_natDegree_le_natDegree_of_comm
(comm : p.leadingCoeff * q.leadingCoeff = q.leadingCoeff * p.leadingCoeff)
(h : p.natDegree ≤ q.natDegree) (hq : 0 < q.natDegree) :
(p.cancelLeads q).natDegree < q.natDegree := by
by_cases hp : p = 0
· convert hq
simp [hp, cancelLeads]
rw [cancelLeads, sub_eq_add_neg, tsub_eq_zero_iff_le.mpr h, pow_zero, mul_one]
by_cases h0 :
C p.leadingCoeff * q + -(C q.leadingCoeff * X ^ (q.natDegree - p.natDegree) * p) = 0
· exact (le_of_eq (by simp only [h0, natDegree_zero])).trans_lt hq
apply lt_of_le_of_ne
· compute_degree!
rwa [Nat.sub_add_cancel]
· contrapose! h0
rw [← leadingCoeff_eq_zero, leadingCoeff, h0, mul_assoc, X_pow_mul, ← tsub_add_cancel_of_le h,
add_comm _ p.natDegree]
simp only [coeff_mul_X_pow, coeff_neg, coeff_C_mul, add_tsub_cancel_left, coeff_add]
|
rw [add_comm p.natDegree, tsub_add_cancel_of_le h, ← leadingCoeff, ← leadingCoeff, comm,
add_right_neg]
|
theorem natDegree_cancelLeads_lt_of_natDegree_le_natDegree_of_comm
(comm : p.leadingCoeff * q.leadingCoeff = q.leadingCoeff * p.leadingCoeff)
(h : p.natDegree ≤ q.natDegree) (hq : 0 < q.natDegree) :
(p.cancelLeads q).natDegree < q.natDegree := by
by_cases hp : p = 0
· convert hq
simp [hp, cancelLeads]
rw [cancelLeads, sub_eq_add_neg, tsub_eq_zero_iff_le.mpr h, pow_zero, mul_one]
by_cases h0 :
C p.leadingCoeff * q + -(C q.leadingCoeff * X ^ (q.natDegree - p.natDegree) * p) = 0
· exact (le_of_eq (by simp only [h0, natDegree_zero])).trans_lt hq
apply lt_of_le_of_ne
· compute_degree!
rwa [Nat.sub_add_cancel]
· contrapose! h0
rw [← leadingCoeff_eq_zero, leadingCoeff, h0, mul_assoc, X_pow_mul, ← tsub_add_cancel_of_le h,
add_comm _ p.natDegree]
simp only [coeff_mul_X_pow, coeff_neg, coeff_C_mul, add_tsub_cancel_left, coeff_add]
|
Mathlib.Data.Polynomial.CancelLeads.52_0.8Zq2gl3suYkxnmI
|
theorem natDegree_cancelLeads_lt_of_natDegree_le_natDegree_of_comm
(comm : p.leadingCoeff * q.leadingCoeff = q.leadingCoeff * p.leadingCoeff)
(h : p.natDegree ≤ q.natDegree) (hq : 0 < q.natDegree) :
(p.cancelLeads q).natDegree < q.natDegree
|
Mathlib_Data_Polynomial_CancelLeads
|
C : Type u_2
inst✝ : Category.{u_1, u_2} C
X : C
S T : Presieve X
⊢ FactorsThruAlong S T (𝟙 X) ↔ FactorsThru S 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]
|
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T := by
|
Mathlib.CategoryTheory.Sites.Coverage.81_0.qkZFgqEgDC2P633
|
@[simp]
lemma factorsThruAlong_id {X : C} (S T : Presieve X) :
S.FactorsThruAlong T (𝟙 X) ↔ S.FactorsThru T
|
Mathlib_CategoryTheory_Sites_Coverage
|
C : Type u_2
inst✝ : Category.{u_1, u_2} C
X : C
S T : Presieve X
h : S ≤ T
Y : C
g : Y ⟶ X
hg : S g
⊢ 𝟙 Y ≫ 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 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
|
Mathlib.CategoryTheory.Sites.Coverage.86_0.qkZFgqEgDC2P633
|
lemma factorsThru_of_le {X : C} (S T : Presieve X) (h : S ≤ T) :
S.FactorsThru T
|
Mathlib_CategoryTheory_Sites_Coverage
|
C : Type u_2
inst✝ : Category.{u_1, u_2} C
X : C
S : Presieve X
T : Sieve X
h : FactorsThru S T.arrows
⊢ S ≤ T.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
|
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
|
Mathlib.CategoryTheory.Sites.Coverage.90_0.qkZFgqEgDC2P633
|
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T
|
Mathlib_CategoryTheory_Sites_Coverage
|
C : Type u_2
inst✝ : Category.{u_1, u_2} C
X : C
S : Presieve X
T : Sieve X
h : FactorsThru S T.arrows
Y : C
f : Y ⟶ X
hf : f ∈ S
⊢ f ∈ T.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
|
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T := by
rintro Y f hf
|
Mathlib.CategoryTheory.Sites.Coverage.90_0.qkZFgqEgDC2P633
|
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T
|
Mathlib_CategoryTheory_Sites_Coverage
|
case intro.intro.intro.intro
C : Type u_2
inst✝ : Category.{u_1, u_2} C
X : C
S : Presieve X
T : Sieve X
h : FactorsThru S T.arrows
Y W : C
i : Y ⟶ W
e : W ⟶ X
h1 : T.arrows e
hf : i ≫ e ∈ S
⊢ i ≫ e ∈ T.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 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
|
Mathlib.CategoryTheory.Sites.Coverage.90_0.qkZFgqEgDC2P633
|
lemma le_of_factorsThru_sieve {X : C} (S : Presieve X) (T : Sieve X) (h : S.FactorsThru T) :
S ≤ T
|
Mathlib_CategoryTheory_Sites_Coverage
|
C : Type u_2
inst✝ : Category.{u_1, u_2} C
X : C
S T : Presieve X
P : Cᵒᵖ ⥤ Type w
H : FactorsThru S T
hS : IsSheafFor P S
h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ R, IsSeparatedFor P R ∧ FactorsThruAlong R 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 *
|
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
|
Mathlib.CategoryTheory.Sites.Coverage.99_0.qkZFgqEgDC2P633
|
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
|
Mathlib_CategoryTheory_Sites_Coverage
|
C : Type u_2
inst✝ : Category.{u_1, u_2} C
X : C
S T : Presieve X
P : Cᵒᵖ ⥤ Type w
H : FactorsThru S T
h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ R, IsSeparatedFor P R ∧ FactorsThruAlong R S f
hS :
IsSeparatedFor P S ∧
∀ (x : FamilyOfElements P S), FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
⊢ IsSeparatedFor P T ∧
∀ (x : FamilyOfElements P T), 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
|
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 *
|
Mathlib.CategoryTheory.Sites.Coverage.99_0.qkZFgqEgDC2P633
|
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
|
Mathlib_CategoryTheory_Sites_Coverage
|
C : Type u_2
inst✝ : Category.{u_1, u_2} C
X : C
S T : Presieve X
P : Cᵒᵖ ⥤ Type w
h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ R, IsSeparatedFor P R ∧ FactorsThruAlong R S f
hS :
IsSeparatedFor P S ∧
∀ (x : FamilyOfElements P S), FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
W : ⦃Z : C⦄ → ⦃g : Z ⟶ X⦄ → S g → C
i : ⦃Z : C⦄ → ⦃g : Z ⟶ X⦄ → (a : S g) → Z ⟶ W a
e : ⦃Z : C⦄ → ⦃g : Z ⟶ X⦄ → (a : S g) → W a ⟶ X
h1 : ∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄ (a : S g), T (e a)
h2 : ∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄ (a : S g), i a ≫ e a = g
⊢ IsSeparatedFor P T ∧
∀ (x : FamilyOfElements P T), 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 => ?_⟩
|
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
|
Mathlib.CategoryTheory.Sites.Coverage.99_0.qkZFgqEgDC2P633
|
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
|
Mathlib_CategoryTheory_Sites_Coverage
|
case refine_1
C : Type u_2
inst✝ : Category.{u_1, u_2} C
X : C
S T : Presieve X
P : Cᵒᵖ ⥤ Type w
h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ R, IsSeparatedFor P R ∧ FactorsThruAlong R S f
hS :
IsSeparatedFor P S ∧
∀ (x : FamilyOfElements P S), FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
W : ⦃Z : C⦄ → ⦃g : Z ⟶ X⦄ → S g → C
i : ⦃Z : C⦄ → ⦃g : Z ⟶ X⦄ → (a : S g) → Z ⟶ W a
e : ⦃Z : C⦄ → ⦃g : Z ⟶ X⦄ → (a : S g) → W a ⟶ X
h1 : ∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄ (a : S g), T (e a)
h2 : ∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄ (a : S g), i a ≫ e a = g
⊢ IsSeparatedFor 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
|
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 => ?_⟩
·
|
Mathlib.CategoryTheory.Sites.Coverage.99_0.qkZFgqEgDC2P633
|
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
|
Mathlib_CategoryTheory_Sites_Coverage
|
case refine_1
C : Type u_2
inst✝ : Category.{u_1, u_2} C
X : C
S T : Presieve X
P : Cᵒᵖ ⥤ Type w
h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ R, IsSeparatedFor P R ∧ FactorsThruAlong R S f
hS :
IsSeparatedFor P S ∧
∀ (x : FamilyOfElements P S), FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
W : ⦃Z : C⦄ → ⦃g : Z ⟶ X⦄ → S g → C
i : ⦃Z : C⦄ → ⦃g : Z ⟶ X⦄ → (a : S g) → Z ⟶ W a
e : ⦃Z : C⦄ → ⦃g : Z ⟶ X⦄ → (a : S g) → W a ⟶ X
h1 : ∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄ (a : S g), T (e a)
h2 : ∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄ (a : S g), i a ≫ e a = g
x : FamilyOfElements P T
y₁ y₂ : P.obj (Opposite.op X)
h₁ : FamilyOfElements.IsAmalgamation x y₁
h₂ : FamilyOfElements.IsAmalgamation x y₂
⊢ y₁ = y₂
|
/-
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 => ?_)
|
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
|
Mathlib.CategoryTheory.Sites.Coverage.99_0.qkZFgqEgDC2P633
|
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
|
Mathlib_CategoryTheory_Sites_Coverage
|
case refine_1
C : Type u_2
inst✝ : Category.{u_1, u_2} C
X : C
S T : Presieve X
P : Cᵒᵖ ⥤ Type w
h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ R, IsSeparatedFor P R ∧ FactorsThruAlong R S f
hS :
IsSeparatedFor P S ∧
∀ (x : FamilyOfElements P S), FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
W : ⦃Z : C⦄ → ⦃g : Z ⟶ X⦄ → S g → C
i : ⦃Z : C⦄ → ⦃g : Z ⟶ X⦄ → (a : S g) → Z ⟶ W a
e : ⦃Z : C⦄ → ⦃g : Z ⟶ X⦄ → (a : S g) → W a ⟶ X
h1 : ∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄ (a : S g), T (e a)
h2 : ∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄ (a : S g), i a ≫ e a = g
x : FamilyOfElements P T
y₁ y₂ : P.obj (Opposite.op X)
h₁ : FamilyOfElements.IsAmalgamation x y₁
h₂ : FamilyOfElements.IsAmalgamation x y₂
Y : C
g : Y ⟶ X
hg : S g
⊢ P.map g.op y₁ = P.map g.op y₂
|
/-
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 _)]
|
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 => ?_)
|
Mathlib.CategoryTheory.Sites.Coverage.99_0.qkZFgqEgDC2P633
|
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
|
Mathlib_CategoryTheory_Sites_Coverage
|
case refine_2
C : Type u_2
inst✝ : Category.{u_1, u_2} C
X : C
S T : Presieve X
P : Cᵒᵖ ⥤ Type w
h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ R, IsSeparatedFor P R ∧ FactorsThruAlong R S f
hS :
IsSeparatedFor P S ∧
∀ (x : FamilyOfElements P S), FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
W : ⦃Z : C⦄ → ⦃g : Z ⟶ X⦄ → S g → C
i : ⦃Z : C⦄ → ⦃g : Z ⟶ X⦄ → (a : S g) → Z ⟶ W a
e : ⦃Z : C⦄ → ⦃g : Z ⟶ X⦄ → (a : S g) → W a ⟶ X
h1 : ∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄ (a : S g), T (e a)
h2 : ∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄ (a : S g), i a ≫ e a = g
x : FamilyOfElements P T
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 _))
|
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 _)]
|
Mathlib.CategoryTheory.Sites.Coverage.99_0.qkZFgqEgDC2P633
|
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
|
Mathlib_CategoryTheory_Sites_Coverage
|
case refine_2
C : Type u_2
inst✝ : Category.{u_1, u_2} C
X : C
S T : Presieve X
P : Cᵒᵖ ⥤ Type w
h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ R, IsSeparatedFor P R ∧ FactorsThruAlong R S f
hS :
IsSeparatedFor P S ∧
∀ (x : FamilyOfElements P S), FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
W : ⦃Z : C⦄ → ⦃g : Z ⟶ X⦄ → S g → C
i : ⦃Z : C⦄ → ⦃g : Z ⟶ X⦄ → (a : S g) → Z ⟶ W a
e : ⦃Z : C⦄ → ⦃g : Z ⟶ X⦄ → (a : S g) → W a ⟶ X
h1 : ∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄ (a : S g), T (e a)
h2 : ∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄ (a : S g), i a ≫ e a = g
x : FamilyOfElements P T
hx : FamilyOfElements.Compatible x
y : FamilyOfElements P S := fun Y g hg => P.map (i hg).op (x (e hg) (_ : T (e 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]
|
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 _))
|
Mathlib.CategoryTheory.Sites.Coverage.99_0.qkZFgqEgDC2P633
|
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
|
Mathlib_CategoryTheory_Sites_Coverage
|
C : Type u_2
inst✝ : Category.{u_1, u_2} C
X : C
S T : Presieve X
P : Cᵒᵖ ⥤ Type w
h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ R, IsSeparatedFor P R ∧ FactorsThruAlong R S f
hS :
IsSeparatedFor P S ∧
∀ (x : FamilyOfElements P S), FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
W : ⦃Z : C⦄ → ⦃g : Z ⟶ X⦄ → S g → C
i : ⦃Z : C⦄ → ⦃g : Z ⟶ X⦄ → (a : S g) → Z ⟶ W a
e : ⦃Z : C⦄ → ⦃g : Z ⟶ X⦄ → (a : S g) → W a ⟶ X
h1 : ∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄ (a : S g), T (e a)
h2 : ∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄ (a : S g), i a ≫ e a = g
x : FamilyOfElements P T
hx : FamilyOfElements.Compatible x
y : FamilyOfElements P S := fun Y g hg => P.map (i hg).op (x (e hg) (_ : T (e hg)))
⊢ FamilyOfElements.Compatible 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
|
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
|
Mathlib.CategoryTheory.Sites.Coverage.99_0.qkZFgqEgDC2P633
|
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
|
Mathlib_CategoryTheory_Sites_Coverage
|
C : Type u_2
inst✝ : Category.{u_1, u_2} C
X : C
S T : Presieve X
P : Cᵒᵖ ⥤ Type w
h✝ : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ R, IsSeparatedFor P R ∧ FactorsThruAlong R S f
hS :
IsSeparatedFor P S ∧
∀ (x : FamilyOfElements P S), FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
W : ⦃Z : C⦄ → ⦃g : Z ⟶ X⦄ → S g → C
i : ⦃Z : C⦄ → ⦃g : Z ⟶ X⦄ → (a : S g) → Z ⟶ W a
e : ⦃Z : C⦄ → ⦃g : Z ⟶ X⦄ → (a : S g) → W a ⟶ X
h1 : ∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄ (a : S g), T (e a)
h2 : ∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄ (a : S g), i a ≫ e a = g
x : FamilyOfElements P T
hx : FamilyOfElements.Compatible x
y : FamilyOfElements P S := fun Y g hg => P.map (i hg).op (x (e hg) (_ : T (e hg)))
Y₁ Y₂ Z : C
g₁ : Z ⟶ Y₁
g₂ : Z ⟶ Y₂
f₁ : Y₁ ⟶ X
f₂ : Y₂ ⟶ X
h₁ : S f₁
h₂ : S f₂
h : g₁ ≫ f₁ = g₂ ≫ f₂
⊢ P.map g₁.op (y f₁ h₁) = P.map g₂.op (y 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]
|
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
|
Mathlib.CategoryTheory.Sites.Coverage.99_0.qkZFgqEgDC2P633
|
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
|
Mathlib_CategoryTheory_Sites_Coverage
|
C : Type u_2
inst✝ : Category.{u_1, u_2} C
X : C
S T : Presieve X
P : Cᵒᵖ ⥤ Type w
h✝ : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ R, IsSeparatedFor P R ∧ FactorsThruAlong R S f
hS :
IsSeparatedFor P S ∧
∀ (x : FamilyOfElements P S), FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
W : ⦃Z : C⦄ → ⦃g : Z ⟶ X⦄ → S g → C
i : ⦃Z : C⦄ → ⦃g : Z ⟶ X⦄ → (a : S g) → Z ⟶ W a
e : ⦃Z : C⦄ → ⦃g : Z ⟶ X⦄ → (a : S g) → W a ⟶ X
h1 : ∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄ (a : S g), T (e a)
h2 : ∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄ (a : S g), i a ≫ e a = g
x : FamilyOfElements P T
hx : FamilyOfElements.Compatible x
y : FamilyOfElements P S := fun Y g hg => P.map (i hg).op (x (e hg) (_ : T (e hg)))
Y₁ Y₂ Z : C
g₁ : Z ⟶ Y₁
g₂ : Z ⟶ Y₂
f₁ : Y₁ ⟶ X
f₂ : Y₂ ⟶ X
h₁ : S f₁
h₂ : S f₂
h : g₁ ≫ f₁ = g₂ ≫ f₂
⊢ P.map (g₁ ≫ i h₁).op (x (e h₁) (_ : T (e h₁))) = P.map (g₂ ≫ i h₂).op (x (e h₂) (_ : T (e 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
|
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]
|
Mathlib.CategoryTheory.Sites.Coverage.99_0.qkZFgqEgDC2P633
|
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
|
Mathlib_CategoryTheory_Sites_Coverage
|
case a
C : Type u_2
inst✝ : Category.{u_1, u_2} C
X : C
S T : Presieve X
P : Cᵒᵖ ⥤ Type w
h✝ : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ R, IsSeparatedFor P R ∧ FactorsThruAlong R S f
hS :
IsSeparatedFor P S ∧
∀ (x : FamilyOfElements P S), FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
W : ⦃Z : C⦄ → ⦃g : Z ⟶ X⦄ → S g → C
i : ⦃Z : C⦄ → ⦃g : Z ⟶ X⦄ → (a : S g) → Z ⟶ W a
e : ⦃Z : C⦄ → ⦃g : Z ⟶ X⦄ → (a : S g) → W a ⟶ X
h1 : ∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄ (a : S g), T (e a)
h2 : ∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄ (a : S g), i a ≫ e a = g
x : FamilyOfElements P T
hx : FamilyOfElements.Compatible x
y : FamilyOfElements P S := fun Y g hg => P.map (i hg).op (x (e hg) (_ : T (e hg)))
Y₁ Y₂ Z : C
g₁ : Z ⟶ Y₁
g₂ : Z ⟶ Y₂
f₁ : Y₁ ⟶ X
f₂ : Y₂ ⟶ X
h₁ : S f₁
h₂ : S f₂
h : g₁ ≫ f₁ = g₂ ≫ f₂
⊢ (g₁ ≫ i h₁) ≫ e h₁ = (g₂ ≫ i h₂) ≫ e 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]
|
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
|
Mathlib.CategoryTheory.Sites.Coverage.99_0.qkZFgqEgDC2P633
|
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
|
Mathlib_CategoryTheory_Sites_Coverage
|
case refine_2
C : Type u_2
inst✝ : Category.{u_1, u_2} C
X : C
S T : Presieve X
P : Cᵒᵖ ⥤ Type w
h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ R, IsSeparatedFor P R ∧ FactorsThruAlong R S f
hS :
IsSeparatedFor P S ∧
∀ (x : FamilyOfElements P S), FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
W : ⦃Z : C⦄ → ⦃g : Z ⟶ X⦄ → S g → C
i : ⦃Z : C⦄ → ⦃g : Z ⟶ X⦄ → (a : S g) → Z ⟶ W a
e : ⦃Z : C⦄ → ⦃g : Z ⟶ X⦄ → (a : S g) → W a ⟶ X
h1 : ∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄ (a : S g), T (e a)
h2 : ∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄ (a : S g), i a ≫ e a = g
x : FamilyOfElements P T
hx : FamilyOfElements.Compatible x
y : FamilyOfElements P S := fun Y g hg => P.map (i hg).op (x (e hg) (_ : T (e hg)))
hy : FamilyOfElements.Compatible y
⊢ ∃ 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
|
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]
|
Mathlib.CategoryTheory.Sites.Coverage.99_0.qkZFgqEgDC2P633
|
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
|
Mathlib_CategoryTheory_Sites_Coverage
|
case refine_2
C : Type u_2
inst✝ : Category.{u_1, u_2} C
X : C
S T : Presieve X
P : Cᵒᵖ ⥤ Type w
h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ R, IsSeparatedFor P R ∧ FactorsThruAlong R S f
hS :
IsSeparatedFor P S ∧
∀ (x : FamilyOfElements P S), FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
W : ⦃Z : C⦄ → ⦃g : Z ⟶ X⦄ → S g → C
i : ⦃Z : C⦄ → ⦃g : Z ⟶ X⦄ → (a : S g) → Z ⟶ W a
e : ⦃Z : C⦄ → ⦃g : Z ⟶ X⦄ → (a : S g) → W a ⟶ X
h1 : ∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄ (a : S g), T (e a)
h2 : ∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄ (a : S g), i a ≫ e a = g
x : FamilyOfElements P T
hx : FamilyOfElements.Compatible x
y : FamilyOfElements P S := fun Y g hg => P.map (i hg).op (x (e hg) (_ : T (e hg)))
hy : FamilyOfElements.Compatible y
left✝ : IsSeparatedFor P S
h2' : ∀ (x : FamilyOfElements P S), FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x 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
|
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
|
Mathlib.CategoryTheory.Sites.Coverage.99_0.qkZFgqEgDC2P633
|
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
|
Mathlib_CategoryTheory_Sites_Coverage
|
case refine_2.intro
C : Type u_2
inst✝ : Category.{u_1, u_2} C
X : C
S T : Presieve X
P : Cᵒᵖ ⥤ Type w
h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ R, IsSeparatedFor P R ∧ FactorsThruAlong R S f
hS :
IsSeparatedFor P S ∧
∀ (x : FamilyOfElements P S), FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
W : ⦃Z : C⦄ → ⦃g : Z ⟶ X⦄ → S g → C
i : ⦃Z : C⦄ → ⦃g : Z ⟶ X⦄ → (a : S g) → Z ⟶ W a
e : ⦃Z : C⦄ → ⦃g : Z ⟶ X⦄ → (a : S g) → W a ⟶ X
h1 : ∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄ (a : S g), T (e a)
h2 : ∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄ (a : S g), i a ≫ e a = g
x : FamilyOfElements P T
hx : FamilyOfElements.Compatible x
y : FamilyOfElements P S := fun Y g hg => P.map (i hg).op (x (e hg) (_ : T (e hg)))
hy : FamilyOfElements.Compatible y
left✝ : IsSeparatedFor P S
h2' : ∀ (x : FamilyOfElements P S), FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
z : P.obj (Opposite.op X)
hz : FamilyOfElements.IsAmalgamation y z
⊢ ∃ 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 => ?_⟩
|
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
|
Mathlib.CategoryTheory.Sites.Coverage.99_0.qkZFgqEgDC2P633
|
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
|
Mathlib_CategoryTheory_Sites_Coverage
|
case refine_2.intro
C : Type u_2
inst✝ : Category.{u_1, u_2} C
X : C
S T : Presieve X
P : Cᵒᵖ ⥤ Type w
h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ R, IsSeparatedFor P R ∧ FactorsThruAlong R S f
hS :
IsSeparatedFor P S ∧
∀ (x : FamilyOfElements P S), FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
W : ⦃Z : C⦄ → ⦃g : Z ⟶ X⦄ → S g → C
i : ⦃Z : C⦄ → ⦃g : Z ⟶ X⦄ → (a : S g) → Z ⟶ W a
e : ⦃Z : C⦄ → ⦃g : Z ⟶ X⦄ → (a : S g) → W a ⟶ X
h1 : ∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄ (a : S g), T (e a)
h2 : ∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄ (a : S g), i a ≫ e a = g
x : FamilyOfElements P T
hx : FamilyOfElements.Compatible x
y : FamilyOfElements P S := fun Y g hg => P.map (i hg).op (x (e hg) (_ : T (e hg)))
hy : FamilyOfElements.Compatible y
left✝ : IsSeparatedFor P S
h2' : ∀ (x : FamilyOfElements P S), FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
z : P.obj (Opposite.op X)
hz : FamilyOfElements.IsAmalgamation y z
Y : C
g : Y ⟶ X
hg : T g
⊢ P.map g.op z = 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
|
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 => ?_⟩
|
Mathlib.CategoryTheory.Sites.Coverage.99_0.qkZFgqEgDC2P633
|
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
|
Mathlib_CategoryTheory_Sites_Coverage
|
case refine_2.intro.intro.intro
C : Type u_2
inst✝ : Category.{u_1, u_2} C
X : C
S T : Presieve X
P : Cᵒᵖ ⥤ Type w
h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ R, IsSeparatedFor P R ∧ FactorsThruAlong R S f
hS :
IsSeparatedFor P S ∧
∀ (x : FamilyOfElements P S), FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
W : ⦃Z : C⦄ → ⦃g : Z ⟶ X⦄ → S g → C
i : ⦃Z : C⦄ → ⦃g : Z ⟶ X⦄ → (a : S g) → Z ⟶ W a
e : ⦃Z : C⦄ → ⦃g : Z ⟶ X⦄ → (a : S g) → W a ⟶ X
h1 : ∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄ (a : S g), T (e a)
h2 : ∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄ (a : S g), i a ≫ e a = g
x : FamilyOfElements P T
hx : FamilyOfElements.Compatible x
y : FamilyOfElements P S := fun Y g hg => P.map (i hg).op (x (e hg) (_ : T (e hg)))
hy : FamilyOfElements.Compatible y
left✝ : IsSeparatedFor P S
h2' : ∀ (x : FamilyOfElements P S), FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
z : P.obj (Opposite.op X)
hz : FamilyOfElements.IsAmalgamation y z
Y : C
g : Y ⟶ X
hg : T g
R : Presieve Y
hR1 : IsSeparatedFor P R
hR2 : FactorsThruAlong R S g
⊢ P.map g.op z = 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
|
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
|
Mathlib.CategoryTheory.Sites.Coverage.99_0.qkZFgqEgDC2P633
|
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
|
Mathlib_CategoryTheory_Sites_Coverage
|
case refine_2.intro.intro.intro
C : Type u_2
inst✝ : Category.{u_1, u_2} C
X : C
S T : Presieve X
P : Cᵒᵖ ⥤ Type w
h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ R, IsSeparatedFor P R ∧ FactorsThruAlong R S f
hS :
IsSeparatedFor P S ∧
∀ (x : FamilyOfElements P S), FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
W : ⦃Z : C⦄ → ⦃g : Z ⟶ X⦄ → S g → C
i : ⦃Z : C⦄ → ⦃g : Z ⟶ X⦄ → (a : S g) → Z ⟶ W a
e : ⦃Z : C⦄ → ⦃g : Z ⟶ X⦄ → (a : S g) → W a ⟶ X
h1 : ∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄ (a : S g), T (e a)
h2 : ∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄ (a : S g), i a ≫ e a = g
x : FamilyOfElements P T
hx : FamilyOfElements.Compatible x
y : FamilyOfElements P S := fun Y g hg => P.map (i hg).op (x (e hg) (_ : T (e hg)))
hy : FamilyOfElements.Compatible y
left✝ : IsSeparatedFor P S
h2' : ∀ (x : FamilyOfElements P S), FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
z : P.obj (Opposite.op X)
hz : FamilyOfElements.IsAmalgamation y z
Y : C
g : Y ⟶ X
hg : T g
R : Presieve Y
hR1 : IsSeparatedFor P R
WW : ⦃Z : C⦄ → ⦃g : Z ⟶ Y⦄ → R g → C
ii : ⦃Z : C⦄ → ⦃g : Z ⟶ Y⦄ → (a : R g) → Z ⟶ WW a
ee : ⦃Z : C⦄ → ⦃g : Z ⟶ Y⦄ → (a : R g) → WW a ⟶ X
hh1 : ∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (a : R g), S (ee a)
hh2 : ∀ ⦃Z : C⦄ ⦃g_1 : Z ⟶ Y⦄ (a : R g_1), ii a ≫ ee a = g_1 ≫ g
⊢ P.map g.op z = 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 => ?_)
|
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
|
Mathlib.CategoryTheory.Sites.Coverage.99_0.qkZFgqEgDC2P633
|
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
|
Mathlib_CategoryTheory_Sites_Coverage
|
case refine_2.intro.intro.intro
C : Type u_2
inst✝ : Category.{u_1, u_2} C
X : C
S T : Presieve X
P : Cᵒᵖ ⥤ Type w
h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ R, IsSeparatedFor P R ∧ FactorsThruAlong R S f
hS :
IsSeparatedFor P S ∧
∀ (x : FamilyOfElements P S), FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
W : ⦃Z : C⦄ → ⦃g : Z ⟶ X⦄ → S g → C
i : ⦃Z : C⦄ → ⦃g : Z ⟶ X⦄ → (a : S g) → Z ⟶ W a
e : ⦃Z : C⦄ → ⦃g : Z ⟶ X⦄ → (a : S g) → W a ⟶ X
h1 : ∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄ (a : S g), T (e a)
h2 : ∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄ (a : S g), i a ≫ e a = g
x : FamilyOfElements P T
hx : FamilyOfElements.Compatible x
y : FamilyOfElements P S := fun Y g hg => P.map (i hg).op (x (e hg) (_ : T (e hg)))
hy : FamilyOfElements.Compatible y
left✝ : IsSeparatedFor P S
h2' : ∀ (x : FamilyOfElements P S), FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
z : P.obj (Opposite.op X)
hz : FamilyOfElements.IsAmalgamation y z
Y : C
g : Y ⟶ X
hg : T g
R : Presieve Y
hR1 : IsSeparatedFor P R
WW : ⦃Z : C⦄ → ⦃g : Z ⟶ Y⦄ → R g → C
ii : ⦃Z : C⦄ → ⦃g : Z ⟶ Y⦄ → (a : R g) → Z ⟶ WW a
ee : ⦃Z : C⦄ → ⦃g : Z ⟶ Y⦄ → (a : R g) → WW a ⟶ X
hh1 : ∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (a : R g), S (ee a)
hh2 : ∀ ⦃Z : C⦄ ⦃g_1 : Z ⟶ Y⦄ (a : R g_1), ii a ≫ ee a = g_1 ≫ g
Q : C
t : Q ⟶ Y
ht : R t
⊢ P.map t.op (P.map g.op z) = P.map t.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]
|
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 => ?_)
|
Mathlib.CategoryTheory.Sites.Coverage.99_0.qkZFgqEgDC2P633
|
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
|
Mathlib_CategoryTheory_Sites_Coverage
|
case refine_2.intro.intro.intro
C : Type u_2
inst✝ : Category.{u_1, u_2} C
X : C
S T : Presieve X
P : Cᵒᵖ ⥤ Type w
h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ R, IsSeparatedFor P R ∧ FactorsThruAlong R S f
hS :
IsSeparatedFor P S ∧
∀ (x : FamilyOfElements P S), FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
W : ⦃Z : C⦄ → ⦃g : Z ⟶ X⦄ → S g → C
i : ⦃Z : C⦄ → ⦃g : Z ⟶ X⦄ → (a : S g) → Z ⟶ W a
e : ⦃Z : C⦄ → ⦃g : Z ⟶ X⦄ → (a : S g) → W a ⟶ X
h1 : ∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄ (a : S g), T (e a)
h2 : ∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄ (a : S g), i a ≫ e a = g
x : FamilyOfElements P T
hx : FamilyOfElements.Compatible x
y : FamilyOfElements P S := fun Y g hg => P.map (i hg).op (x (e hg) (_ : T (e hg)))
hy : FamilyOfElements.Compatible y
left✝ : IsSeparatedFor P S
h2' : ∀ (x : FamilyOfElements P S), FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
z : P.obj (Opposite.op X)
hz : FamilyOfElements.IsAmalgamation y z
Y : C
g : Y ⟶ X
hg : T g
R : Presieve Y
hR1 : IsSeparatedFor P R
WW : ⦃Z : C⦄ → ⦃g : Z ⟶ Y⦄ → R g → C
ii : ⦃Z : C⦄ → ⦃g : Z ⟶ Y⦄ → (a : R g) → Z ⟶ WW a
ee : ⦃Z : C⦄ → ⦃g : Z ⟶ Y⦄ → (a : R g) → WW a ⟶ X
hh1 : ∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (a : R g), S (ee a)
hh2 : ∀ ⦃Z : C⦄ ⦃g_1 : Z ⟶ Y⦄ (a : R g_1), ii a ≫ ee a = g_1 ≫ g
Q : C
t : Q ⟶ Y
ht : R t
⊢ P.map (ii ht ≫ i (_ : S (ee ht))).op (x (e (_ : S (ee ht))) (_ : T (e (_ : S (ee ht))))) = P.map t.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
|
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]
|
Mathlib.CategoryTheory.Sites.Coverage.99_0.qkZFgqEgDC2P633
|
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
|
Mathlib_CategoryTheory_Sites_Coverage
|
case refine_2.intro.intro.intro.a
C : Type u_2
inst✝ : Category.{u_1, u_2} C
X : C
S T : Presieve X
P : Cᵒᵖ ⥤ Type w
h : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, T f → ∃ R, IsSeparatedFor P R ∧ FactorsThruAlong R S f
hS :
IsSeparatedFor P S ∧
∀ (x : FamilyOfElements P S), FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
W : ⦃Z : C⦄ → ⦃g : Z ⟶ X⦄ → S g → C
i : ⦃Z : C⦄ → ⦃g : Z ⟶ X⦄ → (a : S g) → Z ⟶ W a
e : ⦃Z : C⦄ → ⦃g : Z ⟶ X⦄ → (a : S g) → W a ⟶ X
h1 : ∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄ (a : S g), T (e a)
h2 : ∀ ⦃Z : C⦄ ⦃g : Z ⟶ X⦄ (a : S g), i a ≫ e a = g
x : FamilyOfElements P T
hx : FamilyOfElements.Compatible x
y : FamilyOfElements P S := fun Y g hg => P.map (i hg).op (x (e hg) (_ : T (e hg)))
hy : FamilyOfElements.Compatible y
left✝ : IsSeparatedFor P S
h2' : ∀ (x : FamilyOfElements P S), FamilyOfElements.Compatible x → ∃ t, FamilyOfElements.IsAmalgamation x t
z : P.obj (Opposite.op X)
hz : FamilyOfElements.IsAmalgamation y z
Y : C
g : Y ⟶ X
hg : T g
R : Presieve Y
hR1 : IsSeparatedFor P R
WW : ⦃Z : C⦄ → ⦃g : Z ⟶ Y⦄ → R g → C
ii : ⦃Z : C⦄ → ⦃g : Z ⟶ Y⦄ → (a : R g) → Z ⟶ WW a
ee : ⦃Z : C⦄ → ⦃g : Z ⟶ Y⦄ → (a : R g) → WW a ⟶ X
hh1 : ∀ ⦃Z : C⦄ ⦃g : Z ⟶ Y⦄ (a : R g), S (ee a)
hh2 : ∀ ⦃Z : C⦄ ⦃g_1 : Z ⟶ Y⦄ (a : R g_1), ii a ≫ ee a = g_1 ≫ g
Q : C
t : Q ⟶ Y
ht : R t
⊢ (ii ht ≫ i (_ : S (ee ht))) ≫ e (_ : S (ee ht)) = 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]
|
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
|
Mathlib.CategoryTheory.Sites.Coverage.99_0.qkZFgqEgDC2P633
|
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
|
Mathlib_CategoryTheory_Sites_Coverage
|
C : Type ?u.23525
inst✝ : Category.{?u.23529, ?u.23525} C
J : GrothendieckTopology C
⊢ ∀ ⦃X Y : C⦄ (f : Y ⟶ X),
∀ S ∈ (fun X => {S | Sieve.generate S ∈ GrothendieckTopology.sieves J X}) X,
∃ T ∈ (fun X => {S | Sieve.generate S ∈ GrothendieckTopology.sieves J X}) 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)
|
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
|
Mathlib.CategoryTheory.Sites.Coverage.158_0.qkZFgqEgDC2P633
|
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
|
Mathlib_CategoryTheory_Sites_Coverage
|
C : Type ?u.23525
inst✝ : Category.{?u.23529, ?u.23525} C
J : GrothendieckTopology C
X Y : C
f : Y ⟶ X
S : Presieve X
hS : Sieve.generate S ∈ GrothendieckTopology.sieves J X
⊢ ∃ T ∈ (fun X => {S | Sieve.generate S ∈ GrothendieckTopology.sieves J X}) 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⟩
|
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)
|
Mathlib.CategoryTheory.Sites.Coverage.158_0.qkZFgqEgDC2P633
|
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
|
Mathlib_CategoryTheory_Sites_Coverage
|
C : Type ?u.23525
inst✝ : Category.{?u.23529, ?u.23525} C
J : GrothendieckTopology C
X Y : C
f : Y ⟶ X
S : Presieve X
hS : Sieve.generate S ∈ GrothendieckTopology.sieves J X
⊢ (Sieve.pullback f (Sieve.generate S)).arrows ∈ (fun X => {S | Sieve.generate S ∈ GrothendieckTopology.sieves J X}) 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
|
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⟩
|
Mathlib.CategoryTheory.Sites.Coverage.158_0.qkZFgqEgDC2P633
|
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
|
Mathlib_CategoryTheory_Sites_Coverage
|
C : Type ?u.23525
inst✝ : Category.{?u.23529, ?u.23525} C
J : GrothendieckTopology C
X Y : C
f : Y ⟶ X
S : Presieve X
hS : Sieve.generate S ∈ GrothendieckTopology.sieves J X
⊢ Sieve.generate (Sieve.pullback f (Sieve.generate S)).arrows ∈ GrothendieckTopology.sieves J 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]
|
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
|
Mathlib.CategoryTheory.Sites.Coverage.158_0.qkZFgqEgDC2P633
|
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
|
Mathlib_CategoryTheory_Sites_Coverage
|
C : Type ?u.23525
inst✝ : Category.{?u.23529, ?u.23525} C
J : GrothendieckTopology C
X Y : C
f : Y ⟶ X
S : Presieve X
hS : Sieve.generate S ∈ GrothendieckTopology.sieves J X
⊢ Sieve.pullback f (Sieve.generate S) ∈ GrothendieckTopology.sieves J 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
|
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]
|
Mathlib.CategoryTheory.Sites.Coverage.158_0.qkZFgqEgDC2P633
|
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
|
Mathlib_CategoryTheory_Sites_Coverage
|
C : Type u_2
inst✝ : Category.{u_1, u_2} C
X Y : C
S T : Sieve X
h : S ≤ T
f : Y ⟶ X
hf : S.arrows f
⊢ Sieve.pullback f 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
|
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤ := by
|
Mathlib.CategoryTheory.Sites.Coverage.189_0.qkZFgqEgDC2P633
|
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤
|
Mathlib_CategoryTheory_Sites_Coverage
|
case h
C : Type u_2
inst✝ : Category.{u_1, u_2} C
X Y : C
S T : Sieve X
h : S ≤ T
f : Y ⟶ X
hf : S.arrows f
Z : C
g : Z ⟶ Y
⊢ (Sieve.pullback f T).arrows g ↔ ⊤.arrows 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]
|
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
|
Mathlib.CategoryTheory.Sites.Coverage.189_0.qkZFgqEgDC2P633
|
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤
|
Mathlib_CategoryTheory_Sites_Coverage
|
case h
C : Type u_2
inst✝ : Category.{u_1, u_2} C
X Y : C
S T : Sieve X
h : S ≤ T
f : Y ⟶ X
hf : S.arrows f
Z : C
g : Z ⟶ Y
⊢ T.arrows (g ≫ 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
|
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]
|
Mathlib.CategoryTheory.Sites.Coverage.189_0.qkZFgqEgDC2P633
|
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤
|
Mathlib_CategoryTheory_Sites_Coverage
|
case h.a
C : Type u_2
inst✝ : Category.{u_1, u_2} C
X Y : C
S T : Sieve X
h : S ≤ T
f : Y ⟶ X
hf : S.arrows f
Z : C
g : Z ⟶ Y
⊢ S.arrows (g ≫ 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
|
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
|
Mathlib.CategoryTheory.Sites.Coverage.189_0.qkZFgqEgDC2P633
|
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤
|
Mathlib_CategoryTheory_Sites_Coverage
|
case h.a.x
C : Type u_2
inst✝ : Category.{u_1, u_2} C
X Y : C
S T : Sieve X
h : S ≤ T
f : Y ⟶ X
hf : S.arrows f
Z : C
g : Z ⟶ Y
⊢ S.arrows 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 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
|
Mathlib.CategoryTheory.Sites.Coverage.189_0.qkZFgqEgDC2P633
|
lemma eq_top_pullback {X Y : C} {S T : Sieve X} (h : S ≤ T) (f : Y ⟶ X) (hf : S f) :
T.pullback f = ⊤
|
Mathlib_CategoryTheory_Sites_Coverage
|
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
X : C
S T : Sieve X
h : S ≤ T
hS : saturate K X S
⊢ saturate K 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
|
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
|
Mathlib.CategoryTheory.Sites.Coverage.197_0.qkZFgqEgDC2P633
|
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
|
Mathlib_CategoryTheory_Sites_Coverage
|
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
X : C
S T : Sieve X
h : S ≤ T
hS : saturate K X S
⊢ ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, S.arrows f → saturate K Y (Sieve.pullback f 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
|
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
|
Mathlib.CategoryTheory.Sites.Coverage.197_0.qkZFgqEgDC2P633
|
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
|
Mathlib_CategoryTheory_Sites_Coverage
|
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
X : C
S T : Sieve X
h : S ≤ T
hS : saturate K X S
Y : C
g : Y ⟶ X
hg : S.arrows g
⊢ saturate K Y (Sieve.pullback g 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)]
|
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
|
Mathlib.CategoryTheory.Sites.Coverage.197_0.qkZFgqEgDC2P633
|
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
|
Mathlib_CategoryTheory_Sites_Coverage
|
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
X : C
S T : Sieve X
h : S ≤ T
hS : saturate K X S
Y : C
g : Y ⟶ X
hg : S.arrows g
⊢ saturate 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
|
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)]
·
|
Mathlib.CategoryTheory.Sites.Coverage.197_0.qkZFgqEgDC2P633
|
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
|
Mathlib_CategoryTheory_Sites_Coverage
|
case hf
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
X : C
S T : Sieve X
h : S ≤ T
hS : saturate K X S
Y : C
g : Y ⟶ X
hg : S.arrows g
⊢ S.arrows 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
|
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
·
|
Mathlib.CategoryTheory.Sites.Coverage.197_0.qkZFgqEgDC2P633
|
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
|
Mathlib_CategoryTheory_Sites_Coverage
|
C : Type ?u.27900
inst✝ : Category.{?u.27904, ?u.27900} C
K : Coverage C
⊢ ∀ ⦃X Y : C⦄ ⦃S : Sieve X⦄ (f : Y ⟶ X), S ∈ saturate K X → Sieve.pullback f S ∈ saturate 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
|
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
|
Mathlib.CategoryTheory.Sites.Coverage.205_0.qkZFgqEgDC2P633
|
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
|
Mathlib_CategoryTheory_Sites_Coverage
|
C : Type ?u.27900
inst✝ : Category.{?u.27904, ?u.27900} C
K : Coverage C
X Y : C
S : Sieve X
f : Y ⟶ X
hS : S ∈ saturate K X
⊢ Sieve.pullback f S ∈ saturate 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
|
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
|
Mathlib.CategoryTheory.Sites.Coverage.205_0.qkZFgqEgDC2P633
|
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
|
Mathlib_CategoryTheory_Sites_Coverage
|
C : Type ?u.27900
inst✝ : Category.{?u.27904, ?u.27900} C
K : Coverage C
X Y : C
S : Sieve X
f : Y ⟶ X
hS : S ∈ saturate K X
⊢ Sieve.pullback f S ∈ saturate 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
|
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
|
Mathlib.CategoryTheory.Sites.Coverage.205_0.qkZFgqEgDC2P633
|
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
|
Mathlib_CategoryTheory_Sites_Coverage
|
case of
C : Type ?u.27900
inst✝ : Category.{?u.27904, ?u.27900} C
K : Coverage C
X✝ : C
S✝ : Sieve X✝
X : C
S : Presieve X
hS : S ∈ covering K X
Y : C
f : Y ⟶ X
⊢ Sieve.pullback f (Sieve.generate S) ∈ saturate 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]
|
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
|
Mathlib.CategoryTheory.Sites.Coverage.205_0.qkZFgqEgDC2P633
|
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
|
Mathlib_CategoryTheory_Sites_Coverage
|
case of
C : Type ?u.27900
inst✝ : Category.{?u.27904, ?u.27900} C
K : Coverage C
X✝ : C
S✝ : Sieve X✝
X : C
S : Presieve X
hS : S ∈ covering K X
Y : C
f : Y ⟶ X
⊢ Sieve.pullback f (Sieve.generate S) ∈ saturate 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
|
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 =>
|
Mathlib.CategoryTheory.Sites.Coverage.205_0.qkZFgqEgDC2P633
|
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
|
Mathlib_CategoryTheory_Sites_Coverage
|
case of.intro.intro
C : Type ?u.27900
inst✝ : Category.{?u.27904, ?u.27900} C
K : Coverage C
X✝ : C
S✝ : Sieve X✝
X : C
S : Presieve X
hS : S ∈ covering K X
Y : C
f : Y ⟶ X
R : Presieve Y
hR1 : R ∈ covering K Y
hR2 : Presieve.FactorsThruAlong R S f
⊢ Sieve.pullback f (Sieve.generate S) ∈ saturate 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)
|
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
|
Mathlib.CategoryTheory.Sites.Coverage.205_0.qkZFgqEgDC2P633
|
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
|
Mathlib_CategoryTheory_Sites_Coverage
|
case of.intro.intro
C : Type ?u.27900
inst✝ : Category.{?u.27904, ?u.27900} C
K : Coverage C
X✝ : C
S✝ : Sieve X✝
X : C
S : Presieve X
hS : S ∈ covering K X
Y : C
f : Y ⟶ X
R : Presieve Y
hR1 : R ∈ covering K Y
hR2 : Presieve.FactorsThruAlong R S f
⊢ Sieve.generate R ≤ Sieve.pullback f (Sieve.generate S)
|
/-
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⟩
|
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)
|
Mathlib.CategoryTheory.Sites.Coverage.205_0.qkZFgqEgDC2P633
|
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
|
Mathlib_CategoryTheory_Sites_Coverage
|
case of.intro.intro.intro.intro.intro.intro
C : Type ?u.27900
inst✝ : Category.{?u.27904, ?u.27900} C
K : Coverage C
X✝ : C
S✝ : Sieve X✝
X : C
S : Presieve X
hS : S ∈ covering K X
Y : C
f : Y ⟶ X
R : Presieve Y
hR1 : R ∈ covering K Y
hR2 : Presieve.FactorsThruAlong R S f
Z : C
g : Z ⟶ Y
W : C
i : Z ⟶ W
e : W ⟶ Y
h1 : R e
h2 : i ≫ e = g
⊢ (Sieve.pullback f (Sieve.generate S)).arrows 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
|
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⟩
|
Mathlib.CategoryTheory.Sites.Coverage.205_0.qkZFgqEgDC2P633
|
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
|
Mathlib_CategoryTheory_Sites_Coverage
|
case of.intro.intro.intro.intro.intro.intro.intro.intro.intro.intro
C : Type ?u.27900
inst✝ : Category.{?u.27904, ?u.27900} C
K : Coverage C
X✝ : C
S✝ : Sieve X✝
X : C
S : Presieve X
hS : S ∈ covering K X
Y : C
f : Y ⟶ X
R : Presieve Y
hR1 : R ∈ covering K Y
hR2 : Presieve.FactorsThruAlong R S f
Z : C
g : Z ⟶ Y
W : C
i : Z ⟶ W
e : W ⟶ Y
h1 : R e
h2 : i ≫ e = g
WW : C
ii : W ⟶ WW
ee : WW ⟶ X
hh1 : S ee
hh2 : ii ≫ ee = e ≫ f
⊢ (Sieve.pullback f (Sieve.generate S)).arrows 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, ?_⟩
|
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
|
Mathlib.CategoryTheory.Sites.Coverage.205_0.qkZFgqEgDC2P633
|
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
|
Mathlib_CategoryTheory_Sites_Coverage
|
case of.intro.intro.intro.intro.intro.intro.intro.intro.intro.intro
C : Type ?u.27900
inst✝ : Category.{?u.27904, ?u.27900} C
K : Coverage C
X✝ : C
S✝ : Sieve X✝
X : C
S : Presieve X
hS : S ∈ covering K X
Y : C
f : Y ⟶ X
R : Presieve Y
hR1 : R ∈ covering K Y
hR2 : Presieve.FactorsThruAlong R S f
Z : C
g : Z ⟶ Y
W : C
i : Z ⟶ W
e : W ⟶ Y
h1 : R e
h2 : i ≫ e = g
WW : C
ii : W ⟶ WW
ee : WW ⟶ X
hh1 : S ee
hh2 : ii ≫ ee = e ≫ f
⊢ (i ≫ ii) ≫ ee = g ≫ 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]
|
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, ?_⟩
|
Mathlib.CategoryTheory.Sites.Coverage.205_0.qkZFgqEgDC2P633
|
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
|
Mathlib_CategoryTheory_Sites_Coverage
|
case top
C : Type ?u.27900
inst✝ : Category.{?u.27904, ?u.27900} C
K : Coverage C
X✝ : C
S : Sieve X✝
X Y : C
f : Y ⟶ X
⊢ Sieve.pullback f ⊤ ∈ saturate 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
|
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]
|
Mathlib.CategoryTheory.Sites.Coverage.205_0.qkZFgqEgDC2P633
|
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
|
Mathlib_CategoryTheory_Sites_Coverage
|
case top
C : Type ?u.27900
inst✝ : Category.{?u.27904, ?u.27900} C
K : Coverage C
X✝ : C
S : Sieve X✝
X Y : C
f : Y ⟶ X
⊢ Sieve.pullback f ⊤ ∈ saturate 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
|
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 =>
|
Mathlib.CategoryTheory.Sites.Coverage.205_0.qkZFgqEgDC2P633
|
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
|
Mathlib_CategoryTheory_Sites_Coverage
|
case transitive
C : Type ?u.27900
inst✝ : Category.{?u.27904, ?u.27900} C
K : Coverage C
X✝ : C
S✝ : Sieve X✝
X : C
R S : Sieve X
a✝ : saturate K X R
hS : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R.arrows f → saturate K Y (Sieve.pullback f S)
H1 : ∀ ⦃Y : C⦄ (f : Y ⟶ X), Sieve.pullback f R ∈ saturate K Y
a_ih✝ :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f → ∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y), Sieve.pullback f_1 (Sieve.pullback f S) ∈ saturate K Y_1
Y : C
f : Y ⟶ X
⊢ Sieve.pullback f S ∈ saturate 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
|
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
|
Mathlib.CategoryTheory.Sites.Coverage.205_0.qkZFgqEgDC2P633
|
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
|
Mathlib_CategoryTheory_Sites_Coverage
|
case transitive
C : Type ?u.27900
inst✝ : Category.{?u.27904, ?u.27900} C
K : Coverage C
X✝ : C
S✝ : Sieve X✝
X : C
R S : Sieve X
a✝ : saturate K X R
hS : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R.arrows f → saturate K Y (Sieve.pullback f S)
H1 : ∀ ⦃Y : C⦄ (f : Y ⟶ X), Sieve.pullback f R ∈ saturate K Y
a_ih✝ :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f → ∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y), Sieve.pullback f_1 (Sieve.pullback f S) ∈ saturate K Y_1
Y : C
f : Y ⟶ X
⊢ Sieve.pullback f S ∈ saturate 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
|
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 _ =>
|
Mathlib.CategoryTheory.Sites.Coverage.205_0.qkZFgqEgDC2P633
|
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
|
Mathlib_CategoryTheory_Sites_Coverage
|
case transitive.a
C : Type ?u.27900
inst✝ : Category.{?u.27904, ?u.27900} C
K : Coverage C
X✝ : C
S✝ : Sieve X✝
X : C
R S : Sieve X
a✝ : saturate K X R
hS : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R.arrows f → saturate K Y (Sieve.pullback f S)
H1 : ∀ ⦃Y : C⦄ (f : Y ⟶ X), Sieve.pullback f R ∈ saturate K Y
a_ih✝ :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f → ∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y), Sieve.pullback f_1 (Sieve.pullback f S) ∈ saturate K Y_1
Y : C
f : Y ⟶ X
⊢ saturate K Y ?transitive.R
case transitive.a
C : Type ?u.27900
inst✝ : Category.{?u.27904, ?u.27900} C
K : Coverage C
X✝ : C
S✝ : Sieve X✝
X : C
R S : Sieve X
a✝ : saturate K X R
hS : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R.arrows f → saturate K Y (Sieve.pullback f S)
H1 : ∀ ⦃Y : C⦄ (f : Y ⟶ X), Sieve.pullback f R ∈ saturate K Y
a_ih✝ :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f → ∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y), Sieve.pullback f_1 (Sieve.pullback f S) ∈ saturate K Y_1
Y : C
f : Y ⟶ X
⊢ ∀ ⦃Y_1 : C⦄ ⦃f_1 : Y_1 ⟶ Y⦄, ?transitive.R.arrows f_1 → saturate K Y_1 (Sieve.pullback f_1 (Sieve.pullback f S))
case transitive.R
C : Type ?u.27900
inst✝ : Category.{?u.27904, ?u.27900} C
K : Coverage C
X✝ : C
S✝ : Sieve X✝
X : C
R S : Sieve X
a✝ : saturate K X R
hS : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R.arrows f → saturate K Y (Sieve.pullback f S)
H1 : ∀ ⦃Y : C⦄ (f : Y ⟶ X), Sieve.pullback f R ∈ saturate K Y
a_ih✝ :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f → ∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y), Sieve.pullback f_1 (Sieve.pullback f S) ∈ saturate K Y_1
Y : C
f : Y ⟶ X
⊢ Sieve 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
|
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
|
Mathlib.CategoryTheory.Sites.Coverage.205_0.qkZFgqEgDC2P633
|
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
|
Mathlib_CategoryTheory_Sites_Coverage
|
case transitive.a
C : Type ?u.27900
inst✝ : Category.{?u.27904, ?u.27900} C
K : Coverage C
X✝ : C
S✝ : Sieve X✝
X : C
R S : Sieve X
a✝ : saturate K X R
hS : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R.arrows f → saturate K Y (Sieve.pullback f S)
H1 : ∀ ⦃Y : C⦄ (f : Y ⟶ X), Sieve.pullback f R ∈ saturate K Y
a_ih✝ :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f → ∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y), Sieve.pullback f_1 (Sieve.pullback f S) ∈ saturate K Y_1
Y : C
f : Y ⟶ X
⊢ ∀ ⦃Y_1 : C⦄ ⦃f_1 : Y_1 ⟶ Y⦄,
(Sieve.pullback f R).arrows f_1 → saturate K Y_1 (Sieve.pullback f_1 (Sieve.pullback f S))
|
/-
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
|
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
|
Mathlib.CategoryTheory.Sites.Coverage.205_0.qkZFgqEgDC2P633
|
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
|
Mathlib_CategoryTheory_Sites_Coverage
|
case transitive.a
C : Type ?u.27900
inst✝ : Category.{?u.27904, ?u.27900} C
K : Coverage C
X✝ : C
S✝ : Sieve X✝
X : C
R S : Sieve X
a✝ : saturate K X R
hS : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R.arrows f → saturate K Y (Sieve.pullback f S)
H1 : ∀ ⦃Y : C⦄ (f : Y ⟶ X), Sieve.pullback f R ∈ saturate K Y
a_ih✝ :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f → ∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y), Sieve.pullback f_1 (Sieve.pullback f S) ∈ saturate K Y_1
Y : C
f : Y ⟶ X
Z : C
g : Z ⟶ Y
hg : (Sieve.pullback f R).arrows g
⊢ saturate K Z (Sieve.pullback g (Sieve.pullback f S))
|
/-
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]
|
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
|
Mathlib.CategoryTheory.Sites.Coverage.205_0.qkZFgqEgDC2P633
|
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
|
Mathlib_CategoryTheory_Sites_Coverage
|
case transitive.a
C : Type ?u.27900
inst✝ : Category.{?u.27904, ?u.27900} C
K : Coverage C
X✝ : C
S✝ : Sieve X✝
X : C
R S : Sieve X
a✝ : saturate K X R
hS : ∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄, R.arrows f → saturate K Y (Sieve.pullback f S)
H1 : ∀ ⦃Y : C⦄ (f : Y ⟶ X), Sieve.pullback f R ∈ saturate K Y
a_ih✝ :
∀ ⦃Y : C⦄ ⦃f : Y ⟶ X⦄,
R.arrows f → ∀ ⦃Y_1 : C⦄ (f_1 : Y_1 ⟶ Y), Sieve.pullback f_1 (Sieve.pullback f S) ∈ saturate K Y_1
Y : C
f : Y ⟶ X
Z : C
g : Z ⟶ Y
hg : (Sieve.pullback f R).arrows g
⊢ saturate K Z (Sieve.pullback (g ≫ f) S)
|
/-
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
|
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]
|
Mathlib.CategoryTheory.Sites.Coverage.205_0.qkZFgqEgDC2P633
|
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
|
Mathlib_CategoryTheory_Sites_Coverage
|
C : Type ?u.30764
inst✝ : Category.{?u.30768, ?u.30764} C
K : Coverage C
J : GrothendieckTopology 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
|
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
|
Mathlib.CategoryTheory.Sites.Coverage.248_0.qkZFgqEgDC2P633
|
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 _
|
Mathlib_CategoryTheory_Sites_Coverage
|
case mp
C : Type ?u.30764
inst✝ : Category.{?u.30768, ?u.30764} C
K : Coverage C
J : GrothendieckTopology 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
|
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
·
|
Mathlib.CategoryTheory.Sites.Coverage.248_0.qkZFgqEgDC2P633
|
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 _
|
Mathlib_CategoryTheory_Sites_Coverage
|
case mp
C : Type ?u.30764
inst✝ : Category.{?u.30768, ?u.30764} C
K : Coverage C
J : GrothendieckTopology C
H : toGrothendieck C K ≤ J
X : C
S : Presieve X
hS : S ∈ covering K X
⊢ S ∈ covering (ofGrothendieck C 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
|
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
|
Mathlib.CategoryTheory.Sites.Coverage.248_0.qkZFgqEgDC2P633
|
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 _
|
Mathlib_CategoryTheory_Sites_Coverage
|
case mpr
C : Type ?u.30764
inst✝ : Category.{?u.30768, ?u.30764} C
K : Coverage C
J : GrothendieckTopology C
⊢ K ≤ ofGrothendieck C J → toGrothendieck C K ≤ 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
|
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
·
|
Mathlib.CategoryTheory.Sites.Coverage.248_0.qkZFgqEgDC2P633
|
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 _
|
Mathlib_CategoryTheory_Sites_Coverage
|
case mpr
C : Type ?u.30764
inst✝ : Category.{?u.30768, ?u.30764} C
K : Coverage C
J : GrothendieckTopology C
H : 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
|
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
|
Mathlib.CategoryTheory.Sites.Coverage.248_0.qkZFgqEgDC2P633
|
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 _
|
Mathlib_CategoryTheory_Sites_Coverage
|
case mpr
C : Type ?u.30764
inst✝ : Category.{?u.30768, ?u.30764} C
K : Coverage C
J : GrothendieckTopology C
H : 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
|
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
|
Mathlib.CategoryTheory.Sites.Coverage.248_0.qkZFgqEgDC2P633
|
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 _
|
Mathlib_CategoryTheory_Sites_Coverage
|
case mpr.of
C : Type ?u.30764
inst✝ : Category.{?u.30768, ?u.30764} C
K : Coverage C
J : GrothendieckTopology C
H : 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
|
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
|
Mathlib.CategoryTheory.Sites.Coverage.248_0.qkZFgqEgDC2P633
|
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 _
|
Mathlib_CategoryTheory_Sites_Coverage
|
case mpr.of
C : Type ?u.30764
inst✝ : Category.{?u.30768, ?u.30764} C
K : Coverage C
J : GrothendieckTopology C
H : 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
|
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 =>
|
Mathlib.CategoryTheory.Sites.Coverage.248_0.qkZFgqEgDC2P633
|
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 _
|
Mathlib_CategoryTheory_Sites_Coverage
|
case mpr.top
C : Type ?u.30764
inst✝ : Category.{?u.30768, ?u.30764} C
K : Coverage C
J : GrothendieckTopology C
H : 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
|
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
|
Mathlib.CategoryTheory.Sites.Coverage.248_0.qkZFgqEgDC2P633
|
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 _
|
Mathlib_CategoryTheory_Sites_Coverage
|
case mpr.top
C : Type ?u.30764
inst✝ : Category.{?u.30768, ?u.30764} C
K : Coverage C
J : GrothendieckTopology C
H : 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
|
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 =>
|
Mathlib.CategoryTheory.Sites.Coverage.248_0.qkZFgqEgDC2P633
|
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 _
|
Mathlib_CategoryTheory_Sites_Coverage
|
case mpr.transitive
C : Type ?u.30764
inst✝ : Category.{?u.30768, ?u.30764} C
K : Coverage C
J : GrothendieckTopology C
H : 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
|
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
|
Mathlib.CategoryTheory.Sites.Coverage.248_0.qkZFgqEgDC2P633
|
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 _
|
Mathlib_CategoryTheory_Sites_Coverage
|
case mpr.transitive
C : Type ?u.30764
inst✝ : Category.{?u.30768, ?u.30764} C
K : Coverage C
J : GrothendieckTopology C
H : 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
|
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 =>
|
Mathlib.CategoryTheory.Sites.Coverage.248_0.qkZFgqEgDC2P633
|
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 _
|
Mathlib_CategoryTheory_Sites_Coverage
|
C : Type ?u.30764
inst✝ : Category.{?u.30768, ?u.30764} C
J : GrothendieckTopology C
X : C
S : Sieve X
hS : S ∈ GrothendieckTopology.sieves J X
⊢ S ∈ GrothendieckTopology.sieves (toGrothendieck C (ofGrothendieck C 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]
|
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
|
Mathlib.CategoryTheory.Sites.Coverage.248_0.qkZFgqEgDC2P633
|
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 _
|
Mathlib_CategoryTheory_Sites_Coverage
|
C : Type ?u.30764
inst✝ : Category.{?u.30768, ?u.30764} C
J : GrothendieckTopology C
X : C
S : Sieve X
hS : S ∈ GrothendieckTopology.sieves J X
⊢ Sieve.generate S.arrows ∈ GrothendieckTopology.sieves (toGrothendieck C (ofGrothendieck C 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
|
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]
|
Mathlib.CategoryTheory.Sites.Coverage.248_0.qkZFgqEgDC2P633
|
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 _
|
Mathlib_CategoryTheory_Sites_Coverage
|
case hS
C : Type ?u.30764
inst✝ : Category.{?u.30768, ?u.30764} C
J : GrothendieckTopology C
X : C
S : Sieve X
hS : S ∈ GrothendieckTopology.sieves J X
⊢ S.arrows ∈ covering (ofGrothendieck C 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]
|
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
|
Mathlib.CategoryTheory.Sites.Coverage.248_0.qkZFgqEgDC2P633
|
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 _
|
Mathlib_CategoryTheory_Sites_Coverage
|
case hS
C : Type ?u.30764
inst✝ : Category.{?u.30768, ?u.30764} C
J : GrothendieckTopology C
X : C
S : Sieve X
hS : S ∈ GrothendieckTopology.sieves J X
⊢ Sieve.generate S.arrows ∈ 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]
|
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]
|
Mathlib.CategoryTheory.Sites.Coverage.248_0.qkZFgqEgDC2P633
|
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 _
|
Mathlib_CategoryTheory_Sites_Coverage
|
C : Type u_1
inst✝ : Category.{u_2, u_1} C
K : Coverage C
⊢ toGrothendieck C K = sInf {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
|
/--
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
|
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
⊢ toGrothendieck C K ≤ sInf {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
|
/--
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
·
|
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
⊢ ∀ b ∈ {J | K ≤ ofGrothendieck C J}, toGrothendieck C K ≤ b
|
/-
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
|
/--
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;
|
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}
⊢ toGrothendieck C K ≤ 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
|
/--
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
|
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
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.