University
stringclasses 19
values | Text
stringlengths 458
20.7k
|
|---|---|
UWA
|
transformed to an inertial frame of reference, led to derived embedment depths that
were in error by up to 27%. In contrast, embedment depths derived from IMU data
interpreted from within an inertial frame of reference through application of the
interpretation framework were shown to be in excellent agreement with independent
direct measurements. The framework was shown to be a suitable means of establishing
anchor embedment.
The IMU measurements verified the appropriateness of an embedment prediction model
for dynamically installed anchors based on strain rate enhanced shear resistance and
fluid mechanics drag resistance, and facilitated refinement and calibration of the model.
9.2.2 Free-fall and dynamic embedment in soil
Back analysis of the motion data measured by the IMU during anchor installation using
the interpretation framework described in Chapter 5, enabled profiles of net anchor
resistance and velocity during free-fall in water and soil penetration to be established.
The main findings from this analysis are summarised in the follow:
๏ง The fluid drag resistance acting on the DEPLA during free-fall in water may be
approximated using a constant mean drag coefficient of C = 0.7 for Re > 1.0x106
d,a
(v > 5.9 m/s), which is in good agreement with the upper range of values typically
reported for dynamically installed anchors and free-fall penetrometers.
๏ง DEPLA impact velocities were in the ranges: 4.9 to 6.4 m/s (1:12 scale), 2.7 to 8.2
m/s (1:7.2 scale), 8.5 to 10.5 m/s (1:4.5 scale, at Lough Erne) and 5.6 to 12.9 m/s
(1:4.5 scale, at Firth of Clyde), corresponding to anchor release heights in the of: 3
to 8.2 m (4 to 10.9L, 1:12 scale), 0.5 to 6.2 m (0.4 to 5L, 1:7.2 scale), 6.1 to 9.1 m
f f
(3.1 to 4.6L, 1:4.5 scale, at Lough Erne), 2.3 to 42.5 m (1.2 to 21.3L, 1:4.5 scale, at
f f
Firth of Clyde). For the larger release heights the DEPLA reached terminal velocity
but then slowed slightly due to the increasing fluid drag resistance afforded by the
236
|
UWA
|
increasing length of mooring and follower recovery lines that mobilised in the
water.
๏ง DEPLA tip embedment were in the ranges: 1.6 to 1.8 m (2.1 to 2.4L, 1:12 scale),
f
2.8 to 3.7 m (2.2 to 2.9L, 1:7.2 scale), 6.4 to 6.9 m (3.2 to 3.5L, 1:4.5 scale, Lough
f f
Erne) and 2.6 to 4 m (1.3 to 2L, 1:4.5 scale, Firth of Clyde). These tip embedments
f
are in good agreement with the centrifuge data (1.6 to 2.8L) and reported field data
f
experience of full scale dynamically installed anchors with a broadly similar
geometry. The much shallower embedments measured at the Firth of Clyde tests
reflect the much higher strength gradient at this site.
๏ง The results indicated that anchor embedment increases with increasing impact
velocity, which is sensible as the anchor possesses more kinetic energy when it
impacts the soil.
๏ง The appropriateness of an embedment prediction model for DEPLA based on strain
rate enhanced shear resistance and fluid mechanics drag resistance was
demonstrated through comparison with the IMU measurements.
๏ง For Lough Erne, the best agreement was achieved between the embedment depth
prediction model and the IMU measurements using a power function with a strain
rate parameter of ฮฒ = 0.06, which is within the typical range of that reported in
variable rate penetrometer testing (ฮฒ = 0.06 to 0.08), together with friction ratios of
ฮฑ = 0.46 (1:12 scale), 0.45 (1:7.2 scale) and 0.32 (1:4.5 scale). These ฮฑ values are
generally consistent with the range inferred from cyclic piezocone tests. In the case
of Firth of Clyde, the best agreement was achieved using a power law strain
parameter of ฮฒ = 0.08 and ฮฑ = 0.26 (average value inferred from cyclic piezocone
tests).
๏ง Qualitatively, the embedment depth prediction model captures the dynamic response
adequately for both test locations and is capable of predicting the final embedments
237
|
UWA
|
to within 10% of the measurements using parameters that are routinely used
offshore geotechnical design, in conjunction less common parameters (ฮฒ and C )
d,a
which are both related to the increase in penetration resistance with increasing
anchor velocity. These โdynamicโ parameters account for the increase in
geotechnical resistance due to strain-rate effects (ฮฒ) and the increase in fluid
mechanics resistance due to pressure drag (C ).
d,a
๏ง Back analysis IMU data indicated that the power function is capable of describing
the strain rate dependence of undrained shear strength at the very high strain rates
associated with the dynamic embedment process, although complexities associated
with initial embedment (up to one anchor length) are not captured by the model.
However, these shortcomings do not appear to unduly lessen the capability of the
embedment model to predict the measured velocity profiles and importantly the
final anchor embedment depth.
๏ง ROV video captured at Firth of Clyde showed an open crater at the anchor drop sites
which suggest that the hole caused by the passage of the anchor remains open. This
observation is consistent with numerical analysis and other experimental work,
which suggests that for the strength ratios encountered at this location, the hole will
remain open to an anchor tip embedment of 3.85 m, which is the limit of the
embedments achieved at this test location. Hence, an open hole was assumed in the
embedment depth prediction model.
9.2.3 Follower extraction
DEPLA follower extraction was conducted in the centrifuge and at the Lough Erne test
site. The maximum follower extraction resistance was 2.2 to 3 times the follower dry
mass in the centrifuge tests and 2.9 to 4.1 times the follower dry mass in Lough Erne.
These ranges reflect the range in tip embedment depths achieved and hence the range in
undrained shear strength mobilised during follower extraction. The follower extraction
238
|
UWA
|
resistance is slightly higher in Lough Erne than measured in the centrifuge tests due to
the higher tip embedments achieved in the field tests.
9.2.4 Plate anchor keying response and capacity
The keying and pullout of reduced scale model DEPLAs was successfully carried out in
the centrifuge and at both field locations. The embedment depth loss of the plate anchor
during keying was quantified, and anchor capacity for a given plate geometry and
seabed strength profile was determined. The main findings are as summarised in the
follow:
๏ง Results for dynamic and jacked DEPLA installations were similar in both the
centrifuge and field indicating that the dynamic installation process does not affect
either the keying response or the plate anchor capacity.
๏ง The keying response in the field was qualitatively consistent with the DEPLA
centrifuge data, and data from tests on square anchors installed on the vertical
orientation. However, despite the geometrical similarity between the field and
centrifuge DEPLAs, the loss in embedment, โz = 1D to 1.8D in Lough Erne
e,plate
(could not be determined in Firth of Clyde) was much higher than that in the
centrifuge tests (โz = 0.5D to 0.7D). This was attributed to the soil strength
e,plate
ratio, s /ฮณ'D, which was higher in the field tests (s /ฮณ'D = 4.5 to 7.5 in Lough Erne)
u0 u0
than in the centrifuge tests (s /ฮณ'D = 0.6 to 1.4) as numerical analyses has shown
u0
that higher local strength results in higher embedment depth loss during keying.
๏ง The field tests showed that the plate anchor capacity, F , increased linearly with
vp
depth, which reflects the linear increase in undrained shear strength at both test
locations.
๏ง Similar anchor capacity was achieved at both test sites despite the much shallower
embedment from the Firth of Clyde tests associated with the much higher strength
gradient. This is because the capacity of dynamically installed anchors such as the
239
|
UWA
|
DEPLA is governed by the soil strength at the final embedment depth. This will be
higher in weaker soils and lower in stronger soils but with approximately the same
strength at the final plate embedment depth in either case.
๏ง Field test data indicated that the end of keying coincides with the peak anchor
capacity and the plate anchor bearing capacity factor, N , decreases with reducing
c
load inclination for plate embedment ratios less than four.
๏ง Back analysed peak bearing capacity factors from the centrifuge and field tests were
in the range N = 14.2 to 15.8 (centrifuge), N =14.3 to 14.6 (Lough Erne) and N =
c c c
4.2 to 12.9 (Firth of Clyde).
๏ง At deep embedment, where the plate embedment is at least 2.5D, the bearing
capacity factors derived from the centrifuge and field results reach limiting values
that are in quite good agreement with numerical no-breakaway values for DEPLA.
These are much higher than the exact solution for a deeply embedded vanishingly
thin smooth circular plate. This is due to the combined effect of the fluke thickness
and cruciform arrangement on the DEPLA sleeve, which results in a larger failure
surface (and hence a higher capacity factor) compared with a flat vanishingly thin
circular plate. At shallow embedments where the plate embedment is at less than
2.5D, the experimental capacity factors lie within the bounds of the numerical
immediate breakaway and no-breakaway cases for DEPLA. This is because at
shallow embedment the failure mechanism extends to the soil surface, such that the
gap formed during breakaway may be filled with the free water above the mudline.
๏ง A simple design procedure, based on a design chart presented in terms of the total
energy of the anchor at impact with the mudline for a range of strength gradients
and soil sensitivities, was shown to be a useful approach for scaling a DEPLA for a
given design load.
240
|
UWA
|
Low, H. E., Lunne, T., Andersen, K. H., Sjursen, M. A., Li, X and Randolph, M. F.
(2010). Time effects on the stressโstrain behaviour of natural soft clays.
Gรฉotechnique, Vol. 60, No. 11, pp. 843โ859.
Low, H. E., Randolph M. F., De Jong, J. T., and Yafrate, N. J. (2008). Variable rate full
flow penetration tests in intact and remoulded soil. Proceedings of the 3rd
International Conference on Geotechical and Geophysical Site Characterization,
Taipei, Taiwan, 1โ4 April 2008, pp. 1087โ1092.
Lunne, T. and Andersen, K. H. (2007). Soft clay shear strength parameters for
deepwater geotechnical design. Proceedings of the 6th International Offshore and
Site Investigation and Geotechnics Conference, London, United Kingdom, 11โ13
September 2007, pp. 151โ176.
Martin, C. M. and Randolph, M. F. (2001). Applications of the lower and upper bound
theorems of plasticity to collapse of circular foundations. Proceedings of the 10th
International Conference of the International Computer Methods and Advances in
Geomechanics, Tucson, Arizona, USA, 7โ12 January 2001, Vol. 2, pp. 1417โ1428.
Massey, S. J. (2000). Feasibility study of a new concept for the anchorage of offshore
floating production platforms. Honours thesis, Curtin University of Technology,
Perth, Australia.
Medeiros, C. J. (2001). Torpedo anchor for deep water. Proceedings of the Deepwater
Offshore Technology Conference, Rio de Janeiro, Brazil, October 2001.
Medeiros, C. J. (2002). Low cost anchor system for flexible risers in deep waters.
Proceedings of the Offshore Technology Conference, Houston, Texas, USA, 6โ9
May 2002, OTC 14151.
255
|
UWA
|
THESIS DECLARATION
I, Zhikao Li, certify that:
This thesis has been substantially accomplished during enrolment in the degree.
This thesis does not contain material which has been submitted for the award of any other degree or
diploma in my name, in any university or other tertiary institution.
No part of this work will, in the future, be used in a submission in my name, for any other degree or
diploma in any university or other tertiary institution without the prior approval of The University of
Western Australia and where applicable, any partner institution responsible for the joint-award of this
degree.
This thesis does not contain any material previously published or written by another person, except
where due reference has been made in the text and, where relevant, in the Declaration that follows.
The work(s) are not in any way a violation or infringement of any copyright, trademark, patent, or
other rights whatsoever of any person.
This thesis contains published work and work prepared for publication, some of which has been co-
authored.
Signature:
Date: April 05, 2018
2
|
UWA
|
ABSTRACT
Natural gas (NG) is considered to be one of the most environmentally-friendly energy sources for the
coming century due to its lower rate of carbon emissions to the environment compared to coal and
petroleum. The growth in global demand for NG and the depletion of conventional resources have
boosted the development of sub-quality natural gas reserves that contain high levels of impurities,
such as nitrogen gas (N ) (>4%). The removal of N from natural gas remains one of the most
2 2
challenging tasks in the natural gas upgrading process. Currently, cryogenic distillation technology is
the principal technology for separating N from methane (CH ) on a large scale (> 20 MMscfd).
2 4
However, this technology is very energy-intensive. Alternative technologies to remove N from NG in
2
a more economical and environmentally-friendly way are required.
In this thesis, an extensive review is conducted to study the mechanisms and prospective processes
by which N and CH could be separated. The physical and chemical differences of N and CH are the
2 4 2 4
foundation of any such separation and are used to critically analyze the state of the art of conventional
separation approaches as well as several emerging separation approaches. New experimental
approaches, namely N separation from CH by differences in adsorption kinetics, N capture by
2 4 2
transition metal complexes (TMC) solutions, and N capture by lithium metal, were developed and
2
conducted.
Adsorption based processes are one of the well-established technologies to separate gas mixtures.
Adsorption equilibria and kinetics are two sets of properties crucial to the design and simulation of
adsorption-based gas separation processes. The adsorption equilibria and kinetics of N and CH on
2 4
four commercial adsorbents were experimentally studied in this work. The adsorption measurements
were carried out using a commercial volumetric apparatus, which was operated in its rate of
adsorption mode. Calibration experiments were conducted using helium to correct for the impact of
gas expansion on the observed uptake dynamics. Correcting the rate of adsorption data for N and
2
CH using the non-isothermal Fickian diffusion (FD) model was also found to be essential. The
4
measured sorption kinetics had no dependence on the gas pressure but their temperature
dependence was consistent with an Arrhenius-type relation. The effective sorption rates extracted
using the FD model were able to resolve inconsistencies in the literature for similar measurements.
A continuous recirculating absorption process operating at ambient temperature would have clear
advantages over cryogenic distillation processes. The key to the development of such an absorption
process is to find a N -selective solvent. Although this can be extremely challenging as N is very stable
2 2
and inert at ambient conditions, N can function as a weak ligand and be bound to certain TMCs. An
2
aqueous solution consisting of a โtask-specificโ TMC which can selectively and reversibly bond N is
2
3
|
UWA
|
reported. The absorption equilibrium capacities of N in this TMC aqueous solution were measured
2
using a custom-built volumetric apparatus at temperatures of 293.15, 303.15 and 313.15 K and
pressures ranging from 0 to 4000 kPa. The results show that the ruthenium-based (Ru-based) TMC
aqueous solution could selectively bond N over CH with a specific capacity of up to 0.5 mole N per
2 4 2
mole of RuII, which is half of the stoichiometric amount. The reversibility of N bonding in this TMC
2
solution was also verified by desorption tests. The calculated N absorption energy (20-70 kJ/mol) was
2
moderate, indicating a viable regeneration energy requirement.
Finally, the ability of lithium metal (Li) to capture N from natural gas was investigated experimentally.
2
The Li and N reaction was characterized in this work by SEM, TGA-DSC and synchrotron XRD
2
measurements. In addition, the separation performance was examined using a custom-built flow-
through apparatus. Compared to conventional adsorbents, Li metal showed significant advantages for
separating N from CH : (1) the theoretical loading capacity of N on Li is 24 mmol N /g, which is an
2 4 2 2
order of magnitude higher than the best reported N selective adsorbents[1]; and (2) Li does not react
2
with CH at room temperature, which indicates that Li could have a near infinite selectivity for N over
4 2
CH . Two approaches to recycle Li from Li N are discussed: (1) lithium recycling by thermal
4 3
decomposition of Li N; (2) lithium recycling by a closed chemical loop.
3
In conclusion, while the kinetic separation of N from CH on various adsorbents only delivered a
2 4
moderate performance, the other two more innovative approaches have both clear advantages and
significant challenges that need to overcome. Based on the performance of the separation of N from
2
CH using Ru-based TMC solution, an attractive and promising alternative approach is proposed.
4
Although the limited availability of RuII and the low concentration of this Ru-based TMC in aqueous
solution will likely impede its industrial-scale application, this study serves as a proof-of-concept for
the future development of economically viable nitrogen bonding TMCs. Lastly, while the tremendous
potential of Li to selectively capture N are demonstrated in this thesis, the in situ thermal
2
regeneration of Li was proved to be impractical. An alternative batch mode regeneration process
involving a chemical loop was then proposed, which was estimated to be technically and economically
more feasible.
4
|
UWA
|
ACKNOWLEDGEMENTS
First, I would like to thank my primary supervisor, Professor Eric May, for giving me the opportunity
to pursue science in this research group. Your guidance and support have helped me to develop
essential skills to go through this journey, and your dedication and vision in research is truly an
inspiration to me. I really appreciate all the efforts and time you have put into this work.
I would like to thank Dr. Gongkui (James) Xiao, Dr. Gang (Kevin) Li and Dr. Brendan Graham for guiding
me for the past four years. Thank you for all the insightful discussions and all the skills I have learned
from you. Without your support, I would not have been able to finish this study and none of the work
completed in this thesis would have been possible. I would like to thank our technicians Mr Craig
Grimm and Mr David Amm, and all our group members. You are always there whenever I look for help.
I would also like to thank all my friends. Thank you for all the happy and sad moments that we have
had together in this city. Thank you for supporting me and tolerating me during hardships and when
my spirit was low. The journey of completing a Ph.D is not easy, but I am lucky enough to have you
guys here as companions through this adventure.
Finally, I would like to thank my family. Thank you for your understanding, your patience, your
encouragement, and your love during these pivotal four years.
This research was funded by Chevron Cooperation and supported by scholarship for international
research fees (SIRF), university international stipend (UIS) and UIS scholarship.
The synchrotron XRD studied was supported by the Australian Synchrotron. The author gratefully
acknowledges Dr. Qinfen Gu for his help and insightful discussion.
The author gratefully acknowledges the use of a simultaneous thermogravimetric analyzer and
differential scanning calorimeter (Model: SDT Q600) from Professor Dongke Zhang and a
thermogravimetric analyzer Q50 (TGA Q50) from Professor Hong Yang.
The author gratefully acknowledges the use of a scanning electron microscope & energy-dispersive X-
ray spectroscopy (SEM-EDS) from the Centre for Microscopy, Characterisation and Analysis at
University of Western Australia.
14
|
UWA
|
Chapter 1: Introduction
1.1 The Need for Energy and Natural Gas
As the worldโs population and productivity increase, the global economy is expected to almost double
by 2035.[2] Matched with improvements in efficiency, world energy consumption is predicted to rise
by 28% between 2015 and 2040.[3] Most of the new consumption will come from non-OECD
countries, such as China and India. Within these countries, energy consumption is expected to rise by
41% between 2015 and 2040 due to strong economic growth, increased access to energy markets,
and the demand from a quickly growing population.[3] The largest share of this energy consumption
will continue to come from the industrial sector, such as mining, agriculture, manufacturing, and
construction.[3, 4]
Figure 1.1. (a) Primary energy consumption by fuel; (b) Share of primary energy[5]
Historically, the industrial sector has relied heavily on fossil fuels, mostly oil and coal as an energy
source, but the fuel mix has been gradually changing. Figure 1.1 shows that coal is increasingly being
replaced by natural gas, renewables, hydro and nuclear power. This trend of weakening demand for
coal can be observed in many sectors, such as electricity generation. [5]
Renewable energy, such as wind, solar, geothermal, biomass, and biofuels, is the fast-growing source
of energy at a rate of 7.1% annually. However, this would still put its share at around 10% of primary
energy supply by 2035.[4] Fossil fuels, such as oil, gas, and coal remain as dominant sources of energy
powering the world economy.
1
|
UWA
|
While the growth of coal is projected to decline to 0.2% annually, and the growth of oil is projected at
0.7% annually and expected to slow gradually, natural gas will become the worldโs fastest-growing
fossil fuel, increasing by 1.4 โ 1.6% annually.[3, 4] The use of natural gas is predicted to rise by 45% by
2040, led by demand from mostly industrial and electric power sectors, which would account for a
quarter of global energy demand in 2040.[6]
The appeal of using natural gas for new power plants is not only because of the lower capital costs,
favourable heat rates, and relatively low fuel cost,[3, 6] but also because of the combustion of natural
gas emiting significantly less carbon dioxide and negligible amounts of CO, NOx, SOx and particulates
compared to the combustion of oil or coal.[7] As such, while the renewables sector continues to ramp
up, utilizing natural gas would help to achieve the transition to clean energy as mandated by many
environmental regulations set up by governments and their environmental agencies.[8]
Natural gas trade occurs mostly through pipelines; however, most of the current and future demand
for energy consumption will come from non-OECD countries. As these countries are not necessarily
natural gas producers, it becomes increasingly important to be able to transport natural gas to more
distant destinations. Almost 90% of long distance gas trade will occur through the transport of
liquefied natural gas (LNG) to 2040.[6] By cooling the natural gas to -161ยฐC to reach its liquid state, the
volume occupied by liquid methane is reduced by more than 600 times, allowing the practical storage
and transport of LNG.[1]
In Australia, natural gas has become increasingly important both domestically and for export. Within
Australia, the share of natural gas as primary energy is expected to increase from 26% in 2013 to 34%
in 2050, making it the second largest source of energy after oil.[9] Australia also continues to export
its gas via LNG production, mostly to Japan and South Korea, and in the process, it is about to overtake
Qatar as the worldโs top LNG exporter.[10]
1.2 The Need to Separate Nitrogen from Natural Gas
Nitrogen is an almost inevitable component in natural gas, and its amount in the mixture varies with
different sources of natural gas reservoirs. Due to its inert nature, nitrogen is challenging to be
removed, and it is usually not separated as long as its content remains below the sales specification:
about 3% for pipeline gas or about 1 % for LNG.[1] High nitrogen content natural gas can be blended
with low nitrogen content natural gas to meet such specifications, but this requires a low nitrogen
content gas reservoir to be available nearby.[7] However, with the depletion of low nitrogen content
natural gas and the increasing world demand for LNG, the production of LNG from high nitrogen
content gas reservoirs, which historically have been considered as sub-quality gas reservoirs, becomes
2
|
UWA
|
economically feasible and necessary.[11, 12] Although detailed data about the quality of natural gas
reserves are limited on the market, it was reported in 1998 that in general, 24 trillion cubic feet of
natural gas in the US contains > 4 mol % nitrogen gas, and which were considered as sub-quality gas
reservoirs.[7] At that time, this represented around 16% of the proved reserves of natural gas in the
world. Furthermore, while the recently developed technologies of enhanced gas recovery (EGR) and
nitrogen fracturing have improved the production of natural gas significantly, they have at the same
time aggravated the nitrogen contamination in the natural gas[13]. Thus, there is a significant
motivation to develop economical separation technologies to remove nitrogen from natural gas.
For the production of LNG, nitrogen has to be removed from the natural gas mixture for three reasons.
Firstly, it is an inert gas which does not contribute any heating value. Secondly, the liquefaction of
natural gas is an extremely energy intensive process and the presence of excess nitrogen results in
more energy consumption and larger sizes of the low-temperature equipment and vessels. The higher
the nitrogen concentration, the higher the additional operational and capital costs have to be. Lastly,
a high content nitrogen presence in LNG tanks (>1%) can cause safety hazards. Since LNG is a multi-
component liquid, a high content of nitrogen, which is more volatile than methane in LNG tank,
increases the risk of rollover which is a potentially fatal safety hazard. Studies have shown that if the
nitrogen content in LNG is less than 1%, such rollover risks can be eliminated.[14]
Both pipeline gas and LNG production would benefit from the development of a process that can
remove nitrogen from natural gas economically. To be adopted, such process should also have a
relatively small capital and operating cost. Ideally, the nitrogen separation technology should be
scalable so that it can be applied to both small plants as well as in large-scale facilities.
1.3 The Challenges of Separating Nitrogen and Methane
1.3.1 Physical Properties
To separate a binary gas mixture, the two components must have one or more relatively large
differences in their properties: the greater the difference in these properties, the more efficient the
separation process based on these properties will be. Unfortunately, nitrogen and methane have very
similar physical properties, as shown in Table 1.1.[1]
3
|
UWA
|
Table 1.1 Key physical properties of nitrogen and methane [15, 16] (Adapted from ref. 15 and 16)
Molecule T (K) ฯ (ร
) ฮฑ (ร
3) ยต (D) ฮ (D*ร
)
b
N 77.3 3.64 1.710 0.000 1.54
2
CH 111.7 3.80 2.448 0.000 0.02
4
T Normal boling point;
B:
ฯ: Kinetic diameter
ฮฑ: Polarisibility
ยต: Dipole momnent
ฮ: Quadrapole moment
The difference in nitrogen and methaneโs normal boiling points is relatively large, and thus an effective
distillation process can be developed to efficiently separate this this mixture. The major issue with this
approach, however, is that the required cryogenic conditions are extremely energy-intensive to
achieve. To make this process economical, a large flow rate of the natural gas is typically required.
Nevertheless, cryogenic distillation is currently the dominant technology to separate nitrogen from
methane.
Both nitrogen and methane do not have an overall dipole moment and thus they are non-polar
molecules. Their solubilities in common polar solvents, such as water, are low. Methane has a
relatively larger solubility in organic solvents than nitrogen does, such as in lean oil mixtures (discussed
further in section 2.4.2). No solvent has been reported to have larger physical solubilities for nitrogen
than methane; the only reported pure solvent that has relatively large nitrogen solubility (it is still
methane-selective) is liquid ammonia (discussed further in section 2.4.3). Based on solubility
differences, a continuous circulating absorption process can be designed.
Polarizability, dipole moment and quadrupole moment are parameters contributing to the interacting
potential between gas species and a solid surface. Separation processes that exploit these three
parameters include adsorption and membrane technologies. Polarisability contributes to both van der
Waals (non-electrostatic interaction) and electrostatic interactions. Dipole moments contribute to
electrostatic interaction but the dipole moments for both of nitrogen and methane are zero.
4
|
UWA
|
Quadrapole moments contribute to electrostatic interactions when a gradient in the electric field is
present (i.e. near a solid surface). Methane has larger polarisability and nitrogen has a larger
quadrapole moment, thus the overall interaction difference between nitrogen and methane is usually
relatively small.
A moleculeโs kinetic diameter provides a measure of the likelihood of a collision with another molecule
or a solid pore wall. Gas mixtures can therefore be separated based on differences in their kinetic
diameters by the so-called molecular sieve effect. However, nitrogen and methane have similar kinetic
diameters, which makes molecular sieving extraordinarily difficult. Molecular sieves with micropores
between 3.68 ร
and 3.8 ร
are required to separate such a gas mixture.[15] The detailed mechanism
will be discussed in section 2.2.4.
1.3.2 Chemical Properties
Nitrogen and methane do have significant differences in their chemical properties. Methane can be
converted to other chemicals by various approaches in which nitrogen is essential inert; while nitrogen
can be fixed under different conditions when methane remains inactive. The typical chemical
properties of methane and nitrogen are summarized here. However, considering the amount of
methane in natural gas that needs to be converted, it is likely that chemically converting nitrogen will
be more feasible.
Methane conversion to other chemicals that are more valuable than methane:
(1) Methane conversion to methanol[17, 18]
๐๐๐๐๐๐๐๐๐๐๐๐๐๐๐๐
1
(2) Oxidative coupling of methane[๐ถ๐ถ19๐ถ๐ถ,4 2 0+] ๐๐2 ๏ฟฝโฏโฏโฏโฏโฏ๏ฟฝ ๐ถ๐ถ๐ถ๐ถ3๐๐๐ถ๐ถ
2
๐๐๐๐๐๐๐๐๐๐๐๐๐๐๐๐
1
(3) Methane steam reforming[22๐ถ๐ถ1๐ถ๐ถ, 242 ]+ ๐๐2 ๏ฟฝโฏโฏโฏโฏโฏ๏ฟฝ ๐ถ๐ถ๐ถ๐ถ3๐ถ๐ถ๐ถ๐ถ3+๐ถ๐ถ2๐๐
2
Nitrogen fixation reactions: ๐ถ๐ถ๐ถ๐ถ4 + ๐ถ๐ถ2๐๐ โ ๐ถ๐ถ๐๐ + 3๐ถ๐ถ2
(1) Biological nitrogen fixation by nitrogenase[23-26]
๐๐๐๐๐๐๐๐๐๐๐๐๐๐๐๐๐๐๐๐๐๐
2โ + โ โ
(2) Nitrogen fi ๐๐xa2ti +on 1 b 6y ๐ด๐ด t ๐ด๐ดr ๐๐ansit +ion 8 ๐ถ๐ถmet ๏ฟฝa โฏl โฏ c โฏo โฏโฏm โฏโฏp โฏle ๏ฟฝx 2(T ๐๐M ๐ถ๐ถC3) +16๐ด๐ด๐ด๐ด๐๐ +16๐ถ๐ถ2๐๐๐๐4 +๐ถ๐ถ2
(A) Reversible nitrogen bonding on TMC[27-32]
๐๐2+๐ด๐ด๐๐๐ถ๐ถ ๏ฟฝโฏโฏโฏ๏ฟฝ ๐๐2โ๐ด๐ด๐๐๐ถ๐ถ
5
|
UWA
|
TMC bonds nitrogen at high pressure and low temperature and release nitrogen at low
pressure and high temperature.
(B) Nitrogen reduction by TMC which function as a catalyst[33, 34]
๐๐๐๐๐๐
(3) Nitrogen fixation by Haber-Bosch proc ๐๐e2ss +[3 35 ๐ถ๐ถ, 326 ] ๏ฟฝ โฏ๏ฟฝ 2๐๐๐ถ๐ถ3
๐๐๐๐๐๐๐๐๐๐๐๐๐๐๐๐
The most common catalyst is an ir ๐๐o2n +-ba 3s ๐ถ๐ถe2d c ๏ฟฝa โฏt โฏa โฏโฏly โฏs ๏ฟฝt. T 2h ๐๐is
๐ถ๐ถ
p3r ocess requires high pressure and
high temperature.
(4) Electrochemical Nitrogen fixation[37-40]
๐ธ๐ธ๐๐๐๐๐๐๐๐๐๐๐๐๐๐โ๐๐๐๐๐๐๐๐๐๐๐๐
(5) Nitrogen fixation by lithium me ๐๐t2al +[41 3- ๐ถ๐ถ426 ] ๏ฟฝโฏโฏโฏโฏโฏโฏโฏโฏโฏโฏโฏ๏ฟฝ 2๐๐๐ถ๐ถ3
๐๐2+6๐ฟ๐ฟ๐๐ โ 2๐ฟ๐ฟ๐๐3๐๐
1.3.3 Separation Processes and Mechanisms
Typically, there are four types of process that have been developed and been employed for nitrogen
and methane separation: distillaiton, adsorption, absorption and membrane. These mainly utilize the
differences in nitrogen and methane`s physical properties. The chemical conversion of methane or
nitrogen to other chemicals have not been systematically studied and represents the basis of a fifth
separation process. However, the chemical conversion of nitrogen or methane to other chemicals
could require a novel process design somewhat different to the existing four processes. Table 1-2
shows a matrix showing the gas component`s property which serves as the mechanism upon which
the associated separation processes is based.
6
|
UWA
|
Table 1.2 Property differences and separation mechanism matrix for N and CH
2 4
Distillation Membrane Adsorption Absorption Conversion
Volatility โ
Kinetic
โ โ
diameter
Physical
Polarizability โ โ โ
property
Quadrupole
โ
moment
Solubility โ โ
N as a weak
2
โ โ โ
ligand
Chemical
N conversion โ
2
property
CH
4
โ
conversion
1.4 Thesis Objectives and Outline
This thesis seeks to explore both conventional and emerging technologies with the potential to
overcome the challenges of separating gaseous mixtures of nitrogen and methanes. In Chapter 2, a
detailed review and analysis of technologies for separating nitrogen from natural gas will be
presented. The review will cover both conventional technologies as well as emerging technologies,
and will outline the advantages and challenges of each technology. The work of this thesis begins in
Chapter 3, where a new measurement approach is demonstrated to provide consistent results for
adsorption kinetics on various adsorbents as is needed for conventional separation technologies
based on physical adsorption. Mass transfer coefficient data are reported for several adsorbents that
can be used for nitrogen โ methane separations. In Chapter 4 and Chapter 5, two emerging
technologies for the separation of nitrogen from natural gas are explored. First the use of transition
metal complexes (TMC) to selectively absorb the nitrogen into a liquid solution is investigated in
Chapter 4. Then, in Chapter 5, the use of lithium to to remove nitrogen from the natural gas mixture
7
|
UWA
|
Chapter 2: Background and Overview
As the global demand for energy grows, the need for natural gas as an energy source becomes even
more necessary. Currently, one of the forefront challenges in the field of natural gas is to remove
nitrogen (N ), which is an inert gas with no heating value, so that the transport and storage of liquefied
2
natural gas (LNG) can be done more efficiently.
In this chapter, technologies that have potentials in separating nitrogen from natural gas will be
presented and critically analyzed. Below are some general criteria on which the technologies will be
assessed on.
1. Energy efficient
As the natural gas feed is coming at high pressure, it would be advantageous the resulting
methane (CH ) feed could be kept at high pressure as it gets processed further into liquefied
4
natural gas. To do so, the technology needs to be nitrogen selective, as if it was methane
selective, additional energy would be required to depressurize and pressurize the methane
gas, which will result in additional steps and higher operational costs.
2. Scalable
Since current natural gas processing plants process volumes to the order of several billions of
cubic feet per day, the flow rate at which the natural gas feed is coming in can vary. It is
important for the viable technology to be robust and able to be scaled up easily to handle
larger volume.
3. Meets safety standard
The ideal technology must also meet safety requirement and operate at moderate
temperature and pressure, so that it would be easily adaptable to current industrial
processing plants.
To answer these criteria, assessment of various technologies, both conventional and emerging will be
performed. The list of all technologies covered in this chapter is given in Table 2.1. Conventional
technologies have mainly relied on the physical properties of nitrogen and methane. The conventional
technologies include cryogenic distillation, membrane, adsorption and absorption processes. In
addition, emerging technologies with the potential to be developed to separate nitrogen from natural
gas will also be reviewed. These emerging technologies mainly rely on the distinct chemical properties
of the nitrogen, and include biological nitrogen capture, transition metal complex (TMC) nitrogen
capture, Haber-Bosch process and electrochemical nitrogen capture, which are all inspired by the
9
|
UWA
|
nitrogen fixation from atmosphere studies. Additionally, a novel technology to separate nitrogen by
lithium metal will also be introduced.
Table 2.1 Conventional and emerging technologies for the separation of nitrogen from natural gas
covered in this chapter
Conventional technology Emerging technology
Biological nitrogen fixation
o
Cryogenic distillation
o Transitional metal complex (TMC)
o
Membranes โข Mixed matrix membranes
o
โข Adsorption
Adsorption
o
โข Absorption
Absorption
o
Haber Bosch process
o
Electrochemical nitrogen fixation
o
Lithium
o
Following the introduction of each technology, the benefits and challenges will be discussed.
Ultimately, according to the defined criteria, three technologies will be selected to be investigated
further in the remainder of this thesis.
2.1 Separation of Nitrogen from Natural gas by Conventional Technologies
Conventional methods used to separate nitrogen from natural gas include cryogenic distillation,
membrane, adsorption and absorption processes. These methods are all based on the difference of
physical properties between nitrogen and methane. Although these methods have a number of
limitations, they are well established in the industry and still have some distinct merits that will be
discussed further in each of the subsection.
2.1.1 Cryogenic Distillation
Distillation is one of the most well-established separation technologies, with the driving force being
the difference in physical properties of the molecules, namely its volatility. The volatility difference
between nitrogen (normal boiling point = -195.8 ยฐC) and methane (normal boiling point = -161.5 ยฐC) is
large enough to drive an effective distillation separation process. The process has been reported to
achieve high purity gases โ up to 98% methane recovery with less than 1% nitrogen content in the
resulting methane feed, and less than 1% of methane in the nitrogen vent stream. However, the
distillation process would need to be operated under cryogenic conditions which require (1) pre-
10
|
UWA
|
treatment of the feed gas to remove moisture and carbon dioxide (CO ); (2) high capital cost of the
2
cryogenic equipment; (3) additional operational costs to liquefy nitrogen in the natural gas mixture.
The cryogenic requirement has also made it difficult to meet the criteria for the ideal nitrogen
separation technology, which include energy efficiency and safety.
Currently, nitrogen rejection units (NRU) are already available at the end of liquefaction process for
LNG productions to operate under cryogenic conditions, but considerable capital and operational
costs are still needed to handle the unnecessary large volume of nitrogen component. The current
focus of the cryogenic distillation is process optimization of the heat exchanger network, which can
recover the โcoldโ to a certain extent. However, at the same time, such process optimization will add
complexity in the process design and raise operational issues, especially when nitrogen content varies
as the reservoir is being depleted.[47, 48] Subjected to the high capital and operational costs,
cryogenic distillation process is only economical in LNG production with a flow rate higher than 15
MMSCFD.[16]
2.1.2 Membranes
2.1.2.1 Membranes in General
Membranes serve as semi-permeable barriers to separate a feed gas stream into an enriched
permeate stream and an enriched retentate stream. The driving force of this separation technique is
the difference in the partial pressures between the feed side and the permeate side. The
commercialization of the first H separation membrane โ Prismโ in 1980 has boosted the
2
development of this technology and motivated the scientists to expand its application to different
areas.[49] This technology has several intrinsic advantages: (1) no phase changeโ nitrogen can be
removed under moderate temperature and no cryogenic conditions are required; (2) simple process
flow scheme, compact footprint and high adaptability especially for offshore application; (3) relatively
low capital costs, user-friendly and and easy operation with inexpensive running costs.[1, 50, 51]
There are three types of membranes based on the composition of the material, all of which have been
studied to separate the gas mixture of nitrogen and methane: (1) Organic polymer membrane; (2)
Inorganic molecular sieve membrane; (3) Mixed matrix membrane. Organic polymer membrane and
mixed matrix membrane can be either nitrogen selective or methane selective, while inorganic
molecular sieve membrane is mainly nitrogen selective due to the molecular sieve effect. Table 2.2
summarizes the types of membranes types and main applications.
11
|
UWA
|
Table 2.2 Overview of types of membranes and their applications
Polymer membrane Inorganic
Mixed matrix
Rubbery polymer Glassy polymer molecular sieve
membrane
membrane membrane membrane
CH selective โ โ
4
N selective โ โ โ
2
2.1.2.2 Polymer Membrane
Within the category of organic polymer membranes, they can be further classified by their transition
temperature into two types: rubbery polymer and glassy polymer. When the process at which the
polymer membrane is used is above the glass transition temperature of the polymer, the polymer
chains become flexible, and it is called a rubbery polymer membrane. Due to the flexibility of the
polymer chains, rubbery polymer membranes show weak ability to distinguish the molecular size
around 3.8ร
and gives similar diffusion coefficients to either nitrogen or methane. The diffusion
selectivity of nitrogen over methane on rubbery polymer membrane ranges from 1.5-2.2.
When the operating temperature is below the glass transition temperature of the polymer, the
membrane, the polymer chain becomes relatively rigid, and it is called glassy polymer membrane. The
rigid structure in glassy polymer membrane enables improved differentiation on the size of nitrogen
and methane and thus gives relatively higher diffusion selectivity, typically ranging from 3-6. For
sorption selectivity, there are no big differences between rubbery polymer membrane and glassy
polymer membrane. Typically, the nitrogen over methane sorption selectivity ranges between 0.2-0.5,
as shown in Table 2.3. It can be seen from Table 2.3 that the permselectivity (CH /N ) for CH selective
4 2 4
rubber membrane is usually under 4;[50] while simulation studies show that the required
permselectivity for one stage CH selective membrane process is 6.[52]
4
12
|
UWA
|
Table 2.3 Summary of separation properties of polymer membranes[53]
Rubbery Glassy Membrane material
polymer polymer
0.2-0.5 0.2-0.5
๐พ๐พ๐๐2โ๐พ๐พ๐๐๐ถ๐ถ4
1.5-2.0 3-6
Highest ๐ด๐ด๐๐2โ๐ด๐ด๐๐๐ถ๐ถ r4 eported 2.3[54] Polyimide (6FDA-BAHF)*
๐ผ๐ผ๐๐2โ๐๐๐ถ๐ถ4
Polyamideโpolyether block
Highest reported 4.2[55]
copolymer (Pebaxยฎ 2533)
๐ผ๐ผ๐๐๐ถ๐ถ4โ๐๐2
โข There is an editing error in reference 11. The highest should be acchieved on
Polyimide (6FDA-BAHF) ๐ผ๐ผ๐๐2โ๐๐๐ถ๐ถ4
Multi-stage process design can help to achieve sale gas specification based on the performance of
existing membranes. So far, the NitroSepTM membrane process developed by MTR, Inc is the sole
example of industrial methane selective membrane technology.[56, 57] This process employs either
single stage or two-stage membrane units under low-temperature conditions to reject nitrogen for
flowrate ranging from 0.005 to 30 MMscfd. Field demonstrations showed that natural gas streams
with a nitrogen concentration lower than 12% could achieve 93% CH recovery. However, when the
4
nitrogen concentration is higher than 30%, this system becomes economically unfeasible.[53, 56-59]
Also, the methane selective membrane approach suffers a drawback in energy efficiency as the
wellhead pressure must be sacrificed. Since the methane stream loses its high pressure after passing
through the membrane, additional energy and costs are required to recompress the enriched
methane stream to a sale gas pressure.
To address this issue, a nitrogen selective membrane process which can remove the nitrogen from
methane and at the same time maintain the high head pressure of methane would be preferred. In
the 1970s, Kim and Koros et al. reported glass polymer membranes owning a โreverse selectivityโ,
meaning that nitrogen would permeate such membranes faster than methane does.[60] The
membrane has a relatively rigid structure due to the limited mobility of glassy polymer segments and
13
|
UWA
|
is able to distinguish the molecular size more effectively, making the diffusion-selective effect
predominant over the sorption effect. [53]
There are three limitations of the glassy polymer membrane: (1) The trade-off between nitrogen
permeability and the selectivity; as in high permeability is usually achieved by sacrificing the
selectivity.[61, 62] (2) Plasticization of the membrane; as over usage under high pressure and flow
rate of methane and other impurities in the natural gas stream, such as carbon dioxide and heavy
hydrocarbons, the polymeric membrane would plasticize and swell, raising issues resulting in a
decrease of diffusion selectivity. Lastly, simulation studies have shown that a > 17 is required
to treat a 10% N
2
and 90% CH
4
binary mixture with a single stage membrane p๐ผ๐ผro๐๐c2e/๐๐s๐ถ๐ถs 4to meet pipeline
specification (<3% N ).[52] The required high nitrogen over methane selectivity is due to the relatively
2
low nitrogen partial pressure compared to that of methane in the feed stream. However, the best
performing glass polymer membranes with acceptable permeabilities reported so far are only able to
provide a โ 2.3,[54] which is still too small to be economically feasible.
๐ผ๐ผ๐๐2/๐๐๐ถ๐ถ4
2.1.2.3 Inorganic Membrane
In contrast to organic polymer membranes, inorganic membranes have more rigid and controllable
porous structures. Inorganic membranes, such as zeolite, metal-organic framework (MOF) and carbon
molecular sieve membranes can be artificially designed to have micropores that can distinguish
nitrogen from methane more effectively. The molecular sieve effect in such materials dominates the
diffusion separation performance. The equation for diffusion selectivity, , is shown in Equation
๐ท๐ท๐๐2
2.1.[12, 63-65] ๐ท๐ท๐ถ๐ถ๐ถ๐ถ4
Equation 2.1
2
๐ด๐ด๐๐2 ๐๐๐๐2 โ๐๐๐ท๐ท(๐๐2โ๐๐๐ถ๐ถ4) โ๐ถ๐ถ๐ท๐ท(๐๐2โ๐๐๐ถ๐ถ4)
๐ด๐ด๐๐๐ถ๐ถ4 = ๏ฟฝ ๐๐2 ๐๐๐ถ๐ถ4๏ฟฝโ๏ฟฝexp (
๐
๐
)๏ฟฝโ๏ฟฝexp (โ
๐
๐
๐ด๐ด
)๏ฟฝ
Jump length of component I, which is the distance between two adjacent
๐๐๐๐ micropores
The difference in diffusion transition state entropy of nitrogen and methane
โ๐๐๐ท๐ท(๐๐2โ๐๐๐ถ๐ถ4)
The difference in diffusion transition state enthalpy of nitrogen and methane
โ๐ถ๐ถ๐ท๐ท(๐๐2โ๐๐๐ถ๐ถ4)
R Gas constant
14
|
UWA
|
T Absolute temperature
The diffusion selectivity defined in Equation 2.5 can further divide into three terms: the jump length
selectivity, the enthalpic selectivity, and the entropic selectivity.
Jump length selectivity
2
๐๐๐๐2
๏ฟฝ 2 ๏ฟฝ
๐๐๐๐๐ถ๐ถ4
Entropic selectivity
โ๐๐๐ท๐ท(๐๐2โ๐๐๐ถ๐ถ4)
๏ฟฝexp ( ๏ฟฝ
๐
๐
Enthalpic selectivity
โ๐ถ๐ถ๐ท๐ท(๐๐2โ๐๐๐ถ๐ถ4)
๏ฟฝโ ๏ฟฝ
๐
๐
๐ด๐ด
Within the inorganic membrane matrix, diffusion will occur when long-range segmental motions open
a gap of sufficient size into which the penetrant can jump. However, the jump length for nitrogen and
methane are almost the same, which results in the jump length selectivity, , to be close to unity.
2
๐๐๐๐2
2
The diffusion entropy is a function of the shape and size of the gas species.๏ฟฝ D๐๐๐ถ๐ถu๐ถ๐ถe4 ๏ฟฝto the slimmer shape
of nitrogen compared to methane, the entropic selectivity, , would be the main
โ๐๐๐ท๐ท(๐๐2โ๐ถ๐ถ๐ถ๐ถ4)
contributor to a high diffusion selectivity if the pore gate can b๏ฟฝeex cpo (ntroll๐
๐
ed pre๏ฟฝcisely.[65] Finally, the
diffusion enthalpy is the sum of the repulsion enthalpy from the pore gate of the membrane pores
and the desorption enthalpy from the membrane pores. Usually, the diffusion enthalpy selectivity,
, is nitrogen favorable and the sorption enthalpy selectivity varies depending on the
โ๐ถ๐ถ๐ท๐ท(๐๐2โ๐ถ๐ถ๐ถ๐ถ4)
๏ฟฝnโature o๐
๐
f๐๐ the i๏ฟฝnorganic materials.
In sum, the inorganic membranes give much better nitrogen and methane separation performance
for both permeance and selectivity, compared to the polymer membranes. The performance of these
inorganic membranes is summarised in Table 2.4.
15
|
UWA
|
Table 2.4 Comparison of N /CH separations through inorganic membranes[66] (Table adapted from
2 4
ref. 25)
Material ฮ N permeance (GPU) Thickness (ฮผm)
2
Carbon molecular sieve 7.7 0.1 70 ยฑ 15
โผ
SSZ-13 13 66 7.8
SAPO-34 5โ7 300 2.0โ3.0
SAPO-34 8 500 6.2
SAPO-34 prepared 6.5โ7.4 880โ1300 3.0โ4.2
Although the separation performance has been improved, there is a major drawback of inorganic
membranes that hinders them from meeting all three criteria of energy efficiency, scalability, and
safety. Since the inorganic membranes have fragile mechanical properties, it is extremely challenging
to engineer large-scale defect-free inorganic membranes, therefore, it would be difficult to scale up.
Continued work in the field of membrane science are underway to address this issue, namely by
developing the carbon molecular sieve (CMS) membranes which are prepared from polymer
membrane precursors through pyrolysis, and thus, would be relatively easier to process[12, 65],
however, much progress is still required to realize this technology to be fully scalable and
commercially viable.
2.1.2.4 Mixed Matrix Membrane
Integration of inorganic molecular sieves into a polymer matrix is a strategy to utilize both the
advantages of these two membranes. The resulting composite membrane is called a mixed matrix
membrane, which would have the improved separation performance as well as the strong and flexible
mechanical properties. However, the implementation of this idea has been very challenging to do in
practice due to the poor compatibility between these two materials. As a result, the overall selectivity
can only be enhanced slightly compared to the mother polymer matrix, still far from the required
values (For one stage membrane process, the simulation shows a selectivity of =17 is
required.[52]). ๐ผ๐ผ๐๐2/๐๐๐ถ๐ถ4
In sum, although the membrane technology offers a promising alternative in separation technique, as
well as being scalable, safe, and energy efficient (if the membrane is nitrogen selective), much
16
|
UWA
|
advancement in the field of material science, both for the selections of filler and matrix, as well as the
mixing and casting technique to obtain membranes with satisfactory morphology, are required before
the separation performance of these membranes is able to compete against other mature
conventional technologies, such as the cryogenic distllation, in being viable to be used as the main
technology to remove the nitrogen content from natural gas.
2.1.3 Adsorption
Adsorption based separation processes are well-established in the industry for a variety of
applications in separating gas mixtures, namely as leading technology used in the purification of air
[67], production of hydrogen gas [68-70], and the capture of carbon dioxide from flue gases[1, 71-74].
In the past several decades, adsorption-based processes for separating nitrogen and methane have
also attracted significant attention in the natural gas processing industry [1, 75], mainly because of
the potentially huge savings in energy costs and capital investment when compared to other mature
technology, such as cryogenic distillation.
However, although modern adsorption-based technologies have been used in large industrial scale, it
is still limited to only handle relatively small volume in separating nitrogen from natural gas. For
instance, the processing capacity of the cutting edge adsorption-based technologies is only around 15
MMscfd of natural gas feed, while a commercial cyclic pressure swing adsorption processes are
already producing high purity hydrogen gas from a 200 MMscfd feed of steam methane reformers.[76]
The reason behind the two significantly different volumes in the throughputs of gases is the
composition of feed gases. For the purification of hydrogen, the feed stream was mainly composed of
hydrogen with a small amount of carbon dioxide, nitrogen, and methane, which could be adsorbed by
the adsorbents, allowing the non-adsorbing hydrogen gas to simply flow through the adsorption bed.
Hence, this requires a relatively small inventory of adsorbents. However, for the removal of nitrogen
from natural gas, the feed stream is mainly composed of methane balanced with nitrogen. Since
commonly, methane is more selectively adsorbed gas, this results in a prohibitively large inventory of
adsorbents, which in turn leads to larger than normal adsorbers. The two key conceptual strategies
that may help develop adsorption-based processes for nitrogen removal from natural gas at LNG
scales or gas fields with very high contaminant concentrations are (1) modifications to pressure swing
adsorption (PSA) process configurations, and (2) improved performance, cost and reliability of
adsorbents.
In this chapter, the focus of the review will be on the application of adsorption-based processes for
nitrogen removal from natural gas, which covers the fundamental principles of adsorption processes,
17
|
UWA
|
the state of the art of adsorbents for nitrogen and methane separation, and current adsorption based
technologies for nitrogen removal from natural gas.
2.1.3.1 Adsorption in General
For a separation of any given mixture, the first step in designing an adsorption-based process to
achieve the required separation performance is the selection of adsorbents. Specifically, the
adsorbents selected must have optimal adsorption equilibria and kinetics for the desired components
to be adsorbed on the adsorbents. To optimize the adsorption equilibria, high adsorption capacity and
selectivity, ease of regeneration, and fast adsorption and desorption kinetics must all be obtained.
Achieving an optimal adsorption equilibria could be a challenging task, as there might be a trade-off
between adsorption equilibrium equilibria and kinetics; for instance, some adsorbents might have
higher equilibrium adsorption capacities for the component, but prohibitively slow adsorption kinetics
in real applications.
The selectivity of the adsorbents for the components could be defined as either the equilibrium
selectivity (ratio of adsorption amount at equilibrium) or the kinetic selectivity (ratio of adsorption
rates). For nitrogen and methane separation, most current adsorbents have higher methane over
nitrogen equilibrium selectivity, and only a selected few have a higher kinetic selectivity of nitrogen
over methane.
2.1.3.2 Adsorbent Selectivity
The equilibrium selectivity of an adsorbent, , for two components in a gas mixture (i is the more
adsorbed component and j the less adsorbed๐ผ๐ผ c๐๐o๐๐mponent ) is defined in Equation 2.2 [77]:
Equation 2.2
๐ฅ๐ฅ๐๐ ๐ฆ๐ฆ๐๐
๐ผ๐ผ๐๐๐๐ = ๏ฟฝ ๏ฟฝ๏ฟฝ ๏ฟฝ
๐ฅ๐ฅ๐๐ ๐ฆ๐ฆ๐๐
where, y and x are the mole fractions of a component in the vapor and adsorbed phase, respectively.
Separations based on differences in sorption rates may still be possible even if . To quantify
this, it, is convenient to define a kinetic selectivity factor, , which incorporate๐ผ๐ผs ๐๐t๐๐hโคe e1ffects of each
componentโs sorption mass transfer coefficient, shown in E๐ฝ๐ฝq๐๐๐๐uation 2.3.
Equation 2.3
๐๐๐๐
๐ฝ๐ฝ๐๐๐๐ = ๐ผ๐ผ๐๐๐๐๏ฟฝ
๐๐๐๐
18
|
UWA
|
As discussed by Ruthven[78], the kinetic selectivity depends on both the diffusivity ratio (assuming
) and the equilibrium selectivity. Thus, evaluating is an important part of screening
๐๐po๐๐tโen๐ด๐ดti๐๐a,๐๐l adsorbents for a given gas separation application. ๐ฝ๐ฝ๐๐๐๐
2.1.3.3 Adsorbents
In addition to the equilibrium capacity, selectivity, and kinetics of the adsorbents, the selection of
adsorbents for a given process objective is also subjected to the impact of other factors such as the
cost and availability of the material. Zeolites and carbon-based adsorbents have successfully been
used in natural gas processing for the separation of nitrogen from methane. Most zeolites and carbon
adsorbents are known to be methane selective based on thermodynamic equilibrium, and the
equilibrium selectivity of methane over nitrogen can be 3 to 4 for these materials. Certain zeolites and
carbon adsorbents also exhibit much larger adsorption kinetic rates for nitrogen than methane, which
could be explored in the design of adsorption processes using kinetic separations. In this section, these
two classes of adsorbents will be discussed in further detail.
The adsorption capacities of methane and nitrogen have been reported for many different types of
zeolites including mordenite[79], ZSM-5 [80, 81], ฮฒ-zeolite [82, 83], chabazite [84], silicate [85, 86],
and 13X [87]. The molecular structures of these zeolite frameworks are well known and characterized,
such as a 12-ring window of the 13X framework, which when balanced by a sodium cation, results in
a cage opening of 7.8 ร
[88]. Zeolites such as 4A, clinoptilolites and the adsorbent ETS-4 (a
titanosilicate material) have smaller pore openings that are near the kinetic diameters of methane
(3.758 ร
) and nitrogen (3.64-3.8 ร
) which can result in an appreciable kinetic selectivity for the smaller
nitrogen molecule, although these materials are often methane equilibrium-selective. [89-91]
A new zeolite subclass that has recently been reported is the ionic liquid zeolite, which combines the
high capacity and fast kinetics of zeolites with the high selectivity of ionic liquids to produce an
adsorbent that exceeded the separation performance of either of the base materials.[92] This
adsorbent was synthesized within our laboratory by ion exchange of sodium Y-type zeolite with a
solution containing tetra- methyl-ammonium (TMA) cations. The resulting TMA-Y adsorbent was
found to have increased methane and decreased nitrogen adsorption compared to the base Na-Y
adsorbent, significantly increasing its methane selectivity. Preliminary measurements with pure fluids
indicated this material had a methane-nitrogen selectivity around 5.
Activated carbons are produced from carbonaceous materials such as coal, charcoal, wood, coconut
husks, or nut shells which are pyrolyzed in the absence of oxygen or any halogen to convert the
material to carbon before they are activated with an oxidizing agent[93]. The base material and the
19
|
UWA
|
conditions under which they are created control the particular properties of the final adsorbent
material. Advanced synthesis techniques can manufacture carbon adsorbents that have the
exceptionally high surface area, as achieved for example by the Maxwell activated carbon [94]. This
adsorbent has a measured surface area of 2250 m2โg-1, a total pore volume of 1.15 cm3โg-1 and fast
kinetics for methane (2.9 s-1) and nitrogen (8.3 s-1); ideal properties for the nitrogen-methane
separation, although it only has an average methane selectivity (~2.8). While many activated carbon
adsorbents have a large adsorption capacity and fast kinetics of both methane and nitrogen, they tend
to have a selectivity less than three which makes it difficult to design an effective separation
process[95]. Molecular sieving carbons (MSC), also known as carbon molecular sieve (CMS), are
activated carbons that are mainly composed of microspores less than 20 ร
in width. These small pores
are often of a similar scale to the adsorbing molecules, which causes a significantly large number of
collisions between the molecule and the pore mouths. The relative size and surface interactions of the
molecules can result in a significantly different rate of diffusion into the pores for similar molecules
[93]. Previous investigations of a carbon molecular sieve produced by Japan EnviroChemicals (formerly
Takeda) have found that the sorption rate of methane is hundreds of times smaller than that of
nitrogen.[96, 97] Measurements of the pore size of MSC 3K-161 have shown that the largest
contribution to the total pore volume was from pores in the range (3.7-4) ร
[98] which is near the
kinetic diameters of methane (3.758 ร
) and nitrogen (3.64-3.8 ร
). These materials will absorb a larger
quantity of methane than nitrogen if left for long enough; however, the difference in kinetics can be
exploited so that the material absorbs more nitrogen within a certain period.
2.1.3.4 Adsorption Processes
Adsorption processes can be classified into temperature swing adsorption (TSA) and pressure swing
adsorption (PSA) according to the methods employed to regenerate the adsorbents. TSA is widely
used in natural gas processing for gas dehydration with adsorbents being zeolite molecular sieve and
silica gel. In TSA gas dehydration units, one TSA cycle (including the heating and cooling the bed) can
take several hours (or even days). This TSA process is feasible for dehydration because the water
content in the gas streams is much smaller than that of other components and that the adsorbents
have a large capacity and high selectivity for water compared to other natural gas components.
However, TSA is not a feasible solution to the separation of nitrogen and methane in the context of
natural gas processing because of the high concentrations of nitrogen and methane in the gas streams
and the much smaller capacities of adsorbents for nitrogen and methane than that for water. PSA is
the potential adsorption process for nitrogen and methane separation.
20
|
UWA
|
In the PSA method, adsorption occurs at elevated pressure and desorption occurs at a pressure lower
than the adsorption pressure, making use of the difference in adsorption amount between these two
pressures. Using the PSA method, adsorbent beds can be depressurized and re-pressurized rapidly,
allowing cycle times of several minutes or even several seconds to be utilized. Accordingly, the amount
of adsorbent required for PSA processes can be much smaller than for an equivalent TSA processes.
There have been only a few industrial PSA processes for nitrogen and methane separation, and a list
of these commercial nitrogen-methane separation processes can be found in a review by Rufford
et.al[1]. Most of these commercial processes use methane equilibrium selective adsorbents except
for one that uses nitrogen kinetic-selective adsorbent[99], carbon molecular sieve.
In summary, the adsorption based process, in particular PSA, offers the capability to separate nitrogen
from the natural gas feed that could also address the three criteria listed earlier in the chapter by
being energy efficient, scalable, and industrially safe. Despite the existing current challenges of
selecting the right kind of adsorbents materials for this application, the technology is largely promising
and worth to be explored further.
2.1.4 Absorption
2.1.4.1 Absorption in General
Absorption is another mature, well-developed, and widely applied separation process, which utilizes
a liquid solution to separate a gas mixture in an absorption tower. In the tower, the lean liquid solution
flows down from the top and makes contact with the gas mixture that is bubbled from the bottom of
the tower. Gas components with higher solubility in the liquid are absorbed, while the remaining
purified gas components are collected at the top of the tower. The enriched liquid solution is pumped
to another regeneration tower called a stripper to be regenerated by temperature swing, pressure
swing, or the combination of these two. The absorbed gas component is then liberated and collected.
The most well-known example of an absorption process in the natural gas processing plant is the use
of amine solution to strip carbon dioxide gas. An aqueous solution of amines have high carbon dioxide
capacity and can effectively remove carbon dioxide from the natural gas feed. Currently, the amine
absorption process remains the leading technology of carbon dioxide removal from natural gas.
To implement a gas-liquid absorption process for the separation of nitrogen from natural gas akin to
the carbon dioxide adsorption process, the desired liquid solution should have a higher capacity of
one component of the feed gas mixtures (either the N or CH ) than the other. Moreover, there are
2 4
three main cost factors: (1) the required liquid circulation rate, which is determined by the amount of
impurity that must be rejected from the feed gas stream and the impurity loading capacity in the
21
|
UWA
|
sorbent (2) the energy required to regenerate the sorbent; (3) and the recompression energy needed
to compress the CH stream to a sale pressure.[1]The nitrogen rejection using liquid absorption-based
4
technology is not widely employed in the natural gas plants mainly due to the difficulty of finding a
proper liquid solution that can either absorb methane or nitrogen effectively and can meet the above
three cost factors. Nevertheless, there are two kinds of absorption process for nitrogen rejection that
have been reported so far: a lean oil CH -selective absorption process and a liquid ammonia N -
4 2
selective absorption process.
2.1.4.2 Lean oil CH -Selective Absorption Process
4
The lean-oil CH -selective absorption process is a physical absorption process where the loading
4
capacities of CH and N in the solvent follow Henryโs law. The Henry constant of methane is much
4 2
larger in organic solvents than that ofnitrogen, therefore the majority of methane goes to lean oil
leaving nitrogen as the vent gas. The separation performance of this process improves at high pressure
and low temperature. Accordingly, the lean-oil can be regenerated under low pressure or high
temperature in a stripper tower. Advanced Extraction Technologies, Inc. (AET) has designed and
constructed a lean-oil based absorption process with a capacity of 2-30 MMscfd.[100] This process
requires chilling the feed gas to -30 to maximize the methane loading capacity. The CH recovery
4
can reach up to 90% with the enriched CH streams contains less than 4% N . The enriched N stream
โ 4 2 2
which is maintained at feed pressure and low temperature can be recycled to chill the feed gas and/or
be recycled for Enhanced Oil Recovery operations if the stream can be economically pumped to a
wellhead. The CH enriched stream liberated from the stripper is at low pressure, and thus, energy is
4
required to recompress the stream to sale gas pressure. As CH is the majority in the feed gas, a
4
considerable lean-oil solvent circulation rate is necessary to achieve high CH recovery.
4
The costs of these two factors, high recompression energy and large solvent circulation rate, are
prohibitive when it comes employing this CH absorption process for LNG production. Since this
4
technology does not meet the energy efficiency requirement defined as one of the three criteria
earlier in this chapter, it will not be discussed further in this chapter.
2.1.4.3 Liquid ammonia N -Selective Absorption Process
2
In natural gas streams, N is the minority component (<50%), and thus a process selectively removing
2
N is preferred. In such process, a relatively small amount of solvent is required and the associated
2
equipment size is reduced compared to the methane selective absorption process. Furthermore, the
purified CH stream remains at the high feed gas pressure, and thus, no further recompression is
4
needed.
22
|
UWA
|
Studies were conducted between the 1930s to 1950s focusing on using liquid ammonia to separate
N from natural gas.[101, 102] In the absorption tower, the ammonia was pressurized to around 50
2
bar at the temperature range from 21 to43 (ammonia stays in liquid form at this condition) to
promote the N loading capacity.[102] The methane stream collected from the top of the absorption.
2 โ
The N enriched ammonia stream left the absorption tower bottom and was chilled to between -100
2
to -30 by a series of cooler and chiller system. At the same time, this stream was depressurized to
โ
20-25 psi and then goes into a stripper. As ammonia is relatively volatile and toxic, for safety and
โ
operation cost consideration, it requires an ammonia recovery section after the stripper and before
N stream is released. The methane and nitrogen solubilities in liquid ammonia are shown in Figure
2
2.1. Because of the lack of solubility data at the same temperature, the solubility data can only be
shown at similar temperatures โ the solubility of methane is shown at 40 ยฐC and solubility of nitrogen
is shown at 38 ยฐC. The extracted Henryโs selectivity of nitrogen over methane is around 0.25 at the
temperature range of 38-40 ยฐC.[103, 104]
Figure 2.1 The solubilities of methane and nitrogen in liquid ammonia. Due to the lack of solubilities
data at the same temperature, the solubilities of nitrogen (๏) were shown at 38 ยฐC and the
solubilities of methane (๏ข) were shown at 40 ยฐC. The extracted Henry`s selectivity of nitrogen over
methane is around 0.25.
There are three main drawbacks to this process: (1) it involves hazardous operation issues, especially
in handling liquid ammonia at high pressure and low temperature in the cooler and chiller systems;
(2) the high cost of the refrigeration cycle for the cooler and chiller systems and the extra ammonia
recovery sections make this process not economically feasible; and (3) the small selectivity of nitrogen
23
|
UWA
|
over methane sacrifices the methane recovery and additional steps are required to deal with the
nitrogen stream, which contains significant amount of methane.
As described earlier in this chapter, the ideal technology for nitrogen removal from natural gas feed
must meet the criteria of energy efficiency, scalability, and safety. The three drawbacks described
above have made the ammonia absorption process unsuitable choice, given the energy inefficiency
and high costs as well as the safety hazards. Nowadays, no reports indicate that natural gas plants are
employing this process.
2.2 Separation of Nitrogen by Emerging Technologies
All of the conventional technologies are based on the physical property differences of nitrogen and
methane. However, since the differences are not great, it has been challenging to separate them
efficiently by such conventional technologies. In the following section, some emerging technologies
that mainly utilize chemical property differences will be examined.
2.2.1 Biological Nitrogen Fixation (BNF)
It has long been known that molecular nitrogen in the air can be captured and converted to fertilizer
by certain microorganisms. This biological nitrogen fixation (BNF) is one of the most important natural
processes in the world. BNF can occur under atmospheric temperature and pressure due to the high
catalytic ability of nitrogenase โ the bio-catalyst for BNF. To date, studies have characterized three
distinct kinds of nitrogenase enzymes which contain different metal cofactors: molybdenum-iron
complex, vanadiumโiron complex, or sole iron-complex constituting their active sites.[105]
Among these enzymes, the molybdenum-iron complex (FeMo) based nitrogenase is the predominant
form, constituting an electron delivery Fe protein and a catalytic FeMo protein, shown in Figure
2.2.[25, 26, 106] The overall stoichiometry equation of nitrogen reduction by nitrogenase is shown in
Equation 2.4. For each electron delivery, the Fe protein consumes two adenosine triphosphate (ATP)
to attach to the FeMo protein and then transfer one electron. ATP is a compound consisting of an
adenosine molecule bonded with three phosphate groups. The breakage of one phosphate group (P)
i
from ATP forming adenosine diphosphate (ADP) releases 34 kJ/mol of energy for many metabolic
processes, shown in Equation 2.5.[107] ATP is present in all forms of life and functions as the
"molecular unit of currency" of intracellular energy transfer.[108] Overall, 8 electrons and 16 ATP are
obligatory to reduce one nitrogen molecule.[109] Regarding kinetics, the overall nitrogen fixation rate
cannot be significantly increased, with the disconnection of the Fe protein from FeMo protein after
24
|
UWA
|
electron delivery thought to be the rate-limiting step in the whole BNF process, which has a relatively
fixed turn-over rate.[110]
Analogous to the biological CO fixation,[111, 112] the intensively studied BNF[24, 109] provides an
2
alternative, green way to capture N from natural gas under moderate conditions. There has been no
2
reported study of the possibility of using this BNF to separate nitrogen from methane yet. However,
following critical analysis of the feasibility of this approach is conducted.
Equation 2.4
2โ โ + โ โ
๐๐2+16๐ด๐ด๐ด๐ด๐๐ +8๐๐ +8๐ถ๐ถ โ 2๐๐๐ถ๐ถ3+16๐ด๐ด๐ด๐ด๐๐ +16๐ถ๐ถ2๐๐๐๐4 +๐ถ๐ถ2
Equation 2.5
๐ด๐ด๐ด๐ด๐๐+ ๐ถ๐ถ2๐๐ โ๐ด๐ด๐ด๐ด๐๐+ ๐๐๐๐
Although it has been intensively studied over the past decades, the mechanism of nitrogenase
catalytic reduction of nitrogen remains a mystery due to its abstruse nature and the difficulty of
trapping the reaction intermediates.[23-25, 109, 113-115]. For the sake of brevity, in this section, the
detailed catalytic mechanism will not be examined; rather the three โmacro-aspectsโ of this BNF will
be analyzed: (1) Capacity, (2) Kinetics: fixing rate, (3) Energy input.
25
|
UWA
|
Figure 2.2 Schematic of FeMo nitrogenase. (A)MoFe protein complex with the Fe protein homodimer
shown in tan, the MoFe protein ฮฑ subunit in green, and the ฮฒ subunit in cyan. (B) Space-filling and
stick models for the 4Feโ4S cluster (F), P-cluster (P), and FeMo-co (M). [109]
2.2.1.1 Nitrogenase Capacity
Capacity is one of the key parameters that determine whether the BNF process could be practical for
nitrogen capture from natural gas. It describes how much nitrogen can be captured per unit
nitrogenase. In this report, the nitrogenase composed of Fe-Mo cofactor is taken as an example.
Theoretically, one Fe-Mo cofactor can only bond one N molecule, and one nitrogenase only contains
2
one Fe-Mo cofactor. Without immersing into the details of the mechanisms, the overall capacity of
N on Fe-Mo cofactor and nitrogenase has been calculated, shown in Table 2.5. The molecular weight
2
of Fe-Mo cofactor and nitrogenase are calculated or estimated from the literature.[116-120] These
absolute values could vary with different Fe-Mo cofactors or nitrogenase, but they would be of the
same order.
26
|
UWA
|
Table 2.5 The loading capacities of nitrogen on Fe-Mo cofactor and nitrogenase and the required
amount of Fe-Mo cofactor and nitrogenase to absorb the nitrogen in 1 m3 of natural gas (nitrogen
concentration = 10%)
Fe-Mo cofactor Nitrogenase
MW 1395 g/mol 344, 000 g/mol
Capacity 0.7168 mol/kg 0.0029 mol/kg
Amount required to treat 1 m3 of
Natural Gas 6.7 kg 1600 kg
(10% N , 1bar, 273.15K)
2
It is evident from Table 2.5 that the nitrogen capacity on nitrogenase is extremely small: 0.0029
mol/kg. This small capacity is mainly due to the huge molecular weight of the nitrogenase, which
contains not only Fe-Mo cofactor but also associated proteins. The nitrogen capacity of Fe-Mo
cofactor and nitrogenase are too small compared to the reported adsorbents and too small to make
this process industrially practical.
2.2.1.2 The Natural Gas Production by BNF with Current BNF Kinetics
Another important parameter to consider is the kinetics, which indicates how fast a process would
occur. This parameter is just as crucial as the equilibrium capacity in determining whether a method
would be suitable to be adopted in the industry. By understanding the kinetics of BNF, the maximum
amount of natural gas that can be processed by global BNF annually can be theoretically estimated.
Table 2.6 shows the annually global BNF in 2005,[121] which is around is around 180 Tg, including 60
Tg anthropogenic sources and 220 Tg natural sources. Two assumptions are made here to estimate
the natural gas production by global BNF: (1) the content of nitrogen in the raw natural gas stream is
10%; (2) the efficiency of natural gas production by BNF is 1%.
27
|
UWA
|
2.2.1.3 Energy Consumption
Another concern with BNF is its high energy consumption. Although nitrogenase can reduce nitrogen
to ammonia under moderate conditions, it consumes 16 moles of ATP, or 488 kJ energy, to reduce
each mole of nitrogen to ammonia.[25, 109] Such high energy consumption is due to the inert triple
covalent bond of nitrogen.[126, 127] In addition, the energy form in the reduction process of nitrogen
is ATP, which is mainly biosynthesized from cellular respiration and cannot be artificially provided by
commonly used industrial energy forms, such as heat, electricity or mechanical energy.
One of the most common sources of ATP is glucose. The schematic process of oxidizing glucose by
aerobic respiration and anaerobic respiration to produce ATP is shown in Figure 2.3. In total, 38 ATP
and 2 ATP molecules can be produced from one glucose molecule decomposition by aerobic
respiration and anaerobic respiration respectively. The high energy consumption of BNF (488 kJ/mol
N ) and the highly specific form of energy (ATP) required make the approach of BNF impractical to fix
2
nitrogen from natural gas.
(a) Aerobic Respiration (b) Anaerobic Respiration
Figure 2.3 Cellular respiration: (a) Aerobic Respiration; (b) Anaerobic Respiration. ATP stands for
adenosine triphosphate, ADP stands for adenosine diphosphate and stands for phosphate.One
glucose moleule can produce 38 ATP through aerobic respiration o ๐๐r ๐๐2 ATP through anaerobic
respiration.
To summarize, although the BNF process is an elegant approach inspired by nature, it is determined
through critical analysis that this would not be practical for the application of nitrogen removal from
natural gas. The main drawbacks of this technology are its small capacity to be scaled up, its limited
availability, its high energy consumption and the required specific energy form. As described earlier in
this chapter, the ideal technology for nitrogen removal from natural gas feed must meet the criteria
of energy efficiency, scalability, and safety. The three challenges described above have made the BNF
process unsuitable choice, and therefore will not be explored further throughout of this work.
However, the approach by which it works remains an inspiring motivation to scientists, and will be
discussed in the following subsections 2.2.2 โ 2.2.4.
29
|
UWA
|
2.2.2 Transition Metal Complex (TMC) Based Processes
2.2.2.1 TMC-N Coordinating Chemistry
2
The previous section has introduced FeMo cofactor as the active site for the binding of nitrogen in
BNF process, and its challenges to be directly adapted as a technology for nitrogen removal. However,
given this knowledge, researchers have been attempting to replicate the function of FeMo cofactor as
the active site using smaller molecules known as transition metal complexes (TMC) for decades, with
the main goal of fixing nitrogen under moderate conditions.
In general, a TMC is a bulky molecule composed of a transition metal ion center and associated various
supporting ligands. A transition metal is an element that has an incomplete sub-shell in its valence
electronic configuration. Ligands are neutral or anionic non-metallic species which can bond to the
metal center by sharing electron density with it. The properties of TMCs strongly depend on the
properties of the metal center and the supporting ligands. To be more specific, the nitrogen bonding
ability of the TMC depends on the electronegativity, the oxidative level, the geometry of the metal
center, and the electron affinity and the steric hindrance effect of the ligands. In principle, careful
tuning the combination of metal center and the associated ligands can lead to achieve the aim of
developing a TMC that can reversibly bind nitrogen under moderate conditions.
The nitrogen molecule has two lone electron pairs which are available to initiate a dative single ฯ bond
with the metal center of TMC. In the dative single ฯ bond, nitrogen molecule contributes more
electron density than the TMC does. Although this ฯ bond is too weak to fix a nitrogen molecule to
the TMC, it is robust enough to draw nitrogen close enough to TMC metal center to initiate a back ฯ
bond by letting the nitrogen molecule attract more electron density from the high energy ฯ orbital of
the metal center to its low energy ฯ* orbital. The imbalance in the electron sharing through a ฯ bond
and a back ฯ bond between nitrogen and TMC could serve as the mechanism of nitrogen bonding to
TMCs.
The extent of electron density transfer from the TMC metal center to the nitrogen through the back ฯ
bound depends on the energy level of the d orbital of the metal center and the geometry of the metal
center. In general, the higher energy d orbital has, the more electron density would be transferred
from the metal center to the nitrogen molecule. As the d-orbital energy decreases from left to right
for the transition metals in the periodic table, the TMCs further to the right tend to have weaker
interaction with nitrogen ligands compared to the TMCs formed from the left transition metals.[128]
Also, for the same transition metal, the lower oxidation state of the metal center is, the higher the d-
orbital energy would be. Usually, a low oxidation state of the metal center leads to a strong back ฯ
30
|
UWA
|
bonding.[113, 129-132] In addition, the geometry of the metal center of TMCs also plays an important
role in determining the distribution of electron density in the complexes. Certain geometric shapes,
for example, trigonal-planar and tetrahedral geometries, tend to have stronger ฯ back bonding.[133]
The supporting ligands are essential in determining the strength of the nitrogen-TMC bonds.
Compared to Cp-type supporting ligands which draw electron density from the metal center, ฯ-donor
supporting ligands, such as aryloxide and amido ligands,[134, 135] can donate electron density to the
metal center and thus increase the energy level of its d-orbital, which eventually leads to a strong ฯ
back bond between the nitrogen molecule and the TMC. Transition metals with low d orbital energies
can bind nitrogen molecule only when arming with active ฯ-donor supporting ligands. [136-138]
However, such strong ฯ-donor supporting ligands are usually very active and sometimes become
unstable under moderate conditions. For example, the phosphine ligands are strong ฯ-donor ligands
and can facilitate Fe center to bond nitrogen, but they are hypergolic when exposing to air.[27, 139]
Intensive screening studies are required to explore TMCs that can bind nitrogen at moderate
conditions and at the same time remain stable under operating conditions in natural gas processing
plants.
Nitrogen can bind to TMC in different modes and in total four N -TMC bonding modes have been
2
proposed,[131] and the most common and relevant mode for the application of nitrogen and methane
separation is a mode called mononuclear end-on, as shown in Figure 2.4. It involves a ฯ-donation from
the lone electron pairs of nitrogen molecule to the empty or orbitals of the metal center
2 2 2
and a back-ฯ-donation from the filled , , or o๐๐r๐ง๐งbitals๐๐ o๐ฅ๐ฅf โm๐๐etal center to the vacant ฯ*
orbitals of nitrogen molecule.[131] ๐๐๐ฅ๐ฅ๐ง๐ง ๐๐๐๐๐ง๐ง ๐๐๐ฅ๐ฅ๐๐
Figure 2.4 End-on bonding mode (Michael and Samuel, 2000). The shading means the two ฯ orbitals
(on nitrogen or on metal) are in an orthogonal position.
2.2.2.2 TMC-CH Coordinating Chemistry
4
Unlike the nitrogen molecule, the methane molecule exhibits a tetrahedral configuration with all
electrons bound to the C and H atom via single (ฯ) bonds. Therefore, it does not have any additional
lone pairs of electrons to form ฯ bond with the TMC or an empty ฯ* orbital to accept electrons from
the TMC to form a ฯ back-bond. As a result, binding methane to TMC is a far less energetically favored
31
|
UWA
|
process.[131] This difference between nitrogen and methane is one key parameter that can be
exploited to develop a separation process that selectively removes nitrogen from methane.
2.2.2.3 TMC-based Membranes Process
The first approach in leveraging TMC for nitrogen removal from natural gas is TMC-based membranes.
As mentioned in section 2.1.2, it is challenging to find membrane materials with higher physical
sorption of N due to the poor condensability of N compared to CH . One possibly feasible way to
2 2 4
improve the sorption selectivity of nitrogen over methane ( is to integrate the nitrogen
binding TMCs into polymer matrix and develop a mixed ma๐พ๐พt๐๐ri2x/ ๐๐m๐ถ๐ถ4e)mbrane. Xue & Koros et al.
conducted an initial study of this strategy.[64] Although no nitrogen sorption improvement was
observed in this initial study, this finding has opened a new avenue to improve . To further
advance this approach, a rigorous screening study of such TMC-polymer-solvent ๐พ๐พc๐๐o2m/๐๐b๐ถ๐ถin4ation is still
required.
One of such example is the Cr-MIL (Cr = chromium; MIL = Matรฉriaux de l'Institut Lavoisier) composite
membranes. After pre-treating at elevated temperature, Cr-MIL-100 and Cr-MIL-101 showed
improved nitrogen capacities due to the nitrogen binding on the unsaturated Cr sites.[140, 141] In
addition, Cr-MIL-101 (non-pre-treated under high temperature) has been shown to have better
compatibility with certain polymers and has been successfully integrated into a polysulfone
membrane. The resulted mixed matrix membrane gave improved gas separation performance.[142]
The mixed matrix membranes composed pre-treated nitrogen binding Cr-MIL-101 and polysulfone or
other mother-polymers serves as one of the potential candidates that can be explored further.
In sum, the TMC-polymer mixed matrix membrane strategy remains a promising approach to remove
N from natural gas. As introduced in the conventional technologies section, membrane technology
2
would be able to address the criteria requirement of being energy efficient, scalable, and safe.
However, the research is still in very early stage and requires deep expertise in material science to be
able to address the poor compatibility issue and improve separation performance significantly.
2.2.2.4 TMC-based Adsorption Process
As discussed before, nitrogen and methane are difficult to be distinguished efficiently only by their
physical sorption interactions with adsorbents. Their van der Waals and electrostatic interactions with
the adsorbates are similar, and usually slightly methane-favorable. The nitrogen`s coordinating
interaction with unsaturated transition metal sites on adsorbents can serve as an additional
interaction that contributes to a higher nitrogen selectivity, shown in Equation 2.10.
32
|
UWA
|
Table 2.8 Successful examples of sorbents based on TMC for different applications of binary gas
mixture separation. Such TMC adsorbents are named as ฯ-complexation.
Adsorbate TMC Substrate Reference
Al O [143]
2 3
CO from syngas CuCl monolayer Carbon or Coked Al O [144]
2 3
NaY zeolite [145]
Ag+ exchanged
Exchanged resins [146, 147]
resin
r-Al O [146, 147]
Olefin/paraffin 2 3
CuCl monolayer
pillared clays [148]
AgNO monolayer SiO [149, 150]
3 2
PdCl monolayer SiO [151]
Aromatics/Apliphatics 2 2
benzene/cyclohexane
AgY zeolite Y zeolite [152, 153]
AgY zeolite Y zeolite [154]
Dienes/Olefins
CuY Zeolite Y zeolite [155]
Nitrogen as a weak binding ligand is not as well-studied as other gaseous ligands mentioned in Table
2.8. Nevertheless, it has been shown that under moderate temperature and low pressure, nitrogen
can bind to transition metal ion-exchanged zeolites, such as Ni-exchanged MFI zeolite,[156] Cu+
exchanged MFI zeolite[156-160] and Cu+ exchanged ZSM zeolite.[161, 162] On Cu+ exchanged MFI
zeolite, the adsorbed nitrogen to Cu+ ratio can reach as high as 0.8 at 15 kPa.[160] The adsorption
enthalpy of nitrogen on Cu+ exchanged MFI zeolite is between from 30 to 60 kJ/mol which indicates
that only moderate regeneration energy would be required.[157] Even though this would have met
the criteria of being energy efficient, the performance is still far too low to be economically feasible
for industrial application. The overall adsorption capacities reported are lower than 1 mmol/g up to
100 kPa, and thus no continued study has been conducted for the nitrogen and methane separation
on these materials yet.
Another type of adsorbent is metal-organic frameworks (MOFs). In 2014, Kyuho Lee and his co-
workers conducted quantum mechanical computations, aiming to predict a particular metal-organic
34
|
UWA
|
framework that can selectively bind nitrogen without tremendous experimental screening.[163] Their
results show that a MOF structure with unsaturated-coordinated V(II) sites can selectively bind
nitrogen through a coordinating bond between the nitrogen and the V(ii)`s dorbital. However, the
experimental synthesis of this particular V-MOF proves to be unattainable so far. In 2017, MOFs based
on chromium rather than vanadium, Cr-MIL-100 and Cr-MIL-101, were synthesized with unsaturated
Cr sites that can reversibly bind nitrogen under moderate conditions.[164] A significant higher overall
affinity to nitrogen than to methane had been observed on this type of MOF through the pressure
range of 0 kPa up to 500 kPa, which is mainly due to the strong coordinating interaction between
nitrogen and the unsaturated Cr sites. However, when the pressure is higher than 500 kPa, the
methane adsorption capacities exceed the ones of nitrogen. There are two reasons for this: (1) the
ratio of nitrogen to the task-specific unsaturated Cr sites can only be up to one and high pressures
would not contribute to binding more nitrogen on these sites once they are saturated; (2) the physical
interactions start to play a dominant role at high pressure, which usually favor methane adsorption.
Since these materials become methane selective at high pressure, it is not scalable. Thus, it does not
meet our defined criteria for nitrogen removal from natural gas, and therefore will not be pursued
further in this thesis.
2.2.2.5 TMC-based Absorption Process
A nitrogen selective absorption process has attractive advantages compared to a methane selective
absorption process, which has been stimulating scientists to explore liquid solutions that can
chemically and selectively absorb N since the 1990s. Two research institutes studied such TMC
2
solutions in the 1990s.[29, 32, 165, 166] Stanford Research Institute International (SRI) is one of the
two companies leading this research.[165, 166] They have summarized the preferred properties of
potential TMC solutions as follow:
1. Reversibility โ can bind nitrogen at high pressure/low temperature and desorb
nitrogen at low pressure/high temperature
2. High nitrogen capacity โ high nitrogen loading at feed gas conditions can reduce the
TMC solution flow rate and thus reduce the operation cost
3. High selectivity โ such TMC should not react with methane; methane solubility in the
solvent should be as small as possible to maximize high methane recovery.
35
|
UWA
|
4. Fast kinetics โ reaction kinetics of TMC-nitrogen binding is preferentially faster than
the gas transportation kinetics in the TMC solution to minimize the absorption tower
diameter and height, and thus reduce capital cost.
5. Thermal and pressure stability โ the TMC should maintain its activity under operation
conditions.
6. Tolerance โ the TMC should retain its activity in the presence of natural gas impurities,
such as carbon dioxide, hydrogen sulfide, and oxygen.
The best TMC system identified by SRI is (bis)tricyclohexylphosphine molybdenum tricarbonyl in
toluene solution. This TMC system can selectively bind nitrogen forming a precipitate from the
solution and then can be regenerated by heating and/or vacuuming. However, severe issues with
regeneration of this TMC was encountered in the demonstrated batch process: firstly, the precipitate
size is uncontrollable which leads to absorbents lost by bypassing the filtration step or tube blockage
by large precipitates; secondly, this TMC degrades at an unexpectedly fast rate. Furthermore, the
aromatic solvent โ toluene is not desirable due to the high solubility of methane in it, which leads to
a low nitrogen selectivity over methane. Nevertheless, their economic analysis based on this
molybdenum compound, assuming the regeneration issue could be fixed, indicates that such a
nitrogen absorption process could be competitive with the current cryogenic process when TMC
lifetime could be extended to 5 years.
Bend Research Inc. (BRI) is the other organization that systematically studied TMC solution based
absorption process.[29, 32] They reported two kinds of TMC systems: Ru-EDTA system and iron(ii)-
phosphine system. The Ru-based TMC has a capacity of 0.6mol N per mol Ru2+ at around 20 bar and
2
21 ยฐC and shows reversibility under low pressure and high temperature. However, this TMC system
cannot be scaled up due to the limited Ruthenium production โ approximately 12 tons per year which
makes this TMC material too expensive to be feasible.[167] The best iron(ii)-phosphine TMC water
solution has a nitrogen capacity of 0.5 mol N per mol Fe2+ at 20 ยฐC and 10 bar, with a nitrogen
2
selectivity over methane reaching 6. Also, these TMCs systems show stability to maintain their
nitrogen capacity up to 100 cycles at the presence of CO , H S. in addition, the solvent for this TMC
2 2
system is water which is claimed to be the best solvent. The desorption process can be achieved
through a pressure swing or a combination of a pressure swing process and a temperature swing.
So far, there have been no reports showing the commercialized application of these TMC-based
chemical absorption process for removing nitrogen from natural gas. Three main reasons may be
responsible for this. Firstly, the types of TMCs reported are too expensive to be scaled up. Ruthenium
36
|
UWA
|
only has a limited production; the phosphine ligands in Iron-based TMCs are challenging to synthesize.
Secondly, iron-phosphine TMCs involve pyrophoric ligands which may cause serious safety hazards in
the natural gas processing plants. Thirdly, before the recent boost of large-scale LNG production, there
was not strong enough driving force to develop efficient nitrogen rejection technologies under
moderate conditions.
However, with the development of LNG production, natural gas plants are more eager for economic
nitrogen rejection processes, which could provide a strong enough driving force for a more rigorous
screening of TMCs. Since the technology is largely promising and has the potential to meet the criteria
described earlier in this chapter, it will be studied in further details in Chapter 4 of this thesis.
2.2.3 Haber - Bosch Process
Every year, the industrial Haber-Bosch processes produce more than 150 million tons of ammonia
from nitrogen and hydrogen.[105] The well-established process, as well as the potential to produce
ammonia to be sold, could be considered to be adapted for the process of nitrogen removal from
natural gas.
Nowadays, the most common catalysts for this reaction are iron-based catalysts which require
extreme reaction conditions, with a temperature higher than 400 , and pressure between 200 and
300 bar. Even the most advanced plants, performing the Kellogg Advanced Ammonia Process (KAAP)
โ
based on an iron-ruthenium mixed catalyst, need a total pressure around 90 bar.[105] For example,
assuming that the N concentration is natural gas stream is 10%, the natural gas stream needs to be
2
at least 225 bar and 400 for the KAAP process, which is much harsher than the current natural gas
processing conditions (50 to 70 bar and at ambient temperature).[1] Furthermore, the Haber Bosch
โ
process is extremely energy intensive and consumes 1% of the global power production.[168]
Therefore, the possibility of converting the nitrogen in the natural gas stream directly to ammonia by
mimicking the Haber-Bosch process is not feasible so far due to the high N partial pressure and high
2
temperature required.
In conclusion, although this process could be scaled up if it were able to be adapted to remove
nitrogen from natural gas, it would have not been able to meet the criteria of energy efficiency and
safety. Therefore, it will not be pursued further in this thesis.
2.2.4 Electrochemical Nitrogen Fixation
The electrochemical route is another possible alternative technology, similar to TMC catalysis, to
produce ammonia. This technology has been claimed to save more than 20% of the energy
37
|
UWA
|
consumption compared to the Haber - Bosch process,[37, 169] and thus shows the potential to be
modified to capture nitrogen from natural gas. In this section, its feasibility for processing large natural
gas streams is considered.
Electrochemical reduction of nitrogen to ammonia happens on equipment called electrolytic cells
which are composed of two electrodes on each side and electrolytes in the middle. A schematic
illustration of the electrochemical conversion of nitrogen to ammonia utilizing a natural gas stream as
the source of nitrogen has been proposed and is shown in Figure 2.5. A natural gas stream is sent to
the cathode chamber to provide nitrogen and another hydrogen stream is sent to the anode chamber.
By applying an external voltage, the hydrogen molecule loses electrons and become protons that
transfer through the electrolyte to the cathode chamber. In the cathode chamber, protons, electrons
and nitrogen react on the cathode and produce ammonia. The product of ammonia can be easily
separated from methane due to the significant difference in their volatilities.
Figure 2.5 Schematic illustration of electrochemical conversion of nitrogen to ammonia using a
natural gas stream as the source of nitrogen. A natural gas stream mainly containing nitrogen and
methane is sent to the nitrogen chamber where nitrogen is converted to ammonia. The effluent
stream from the nitrogen chamber mainly contains the produced ammonia and the remaining
methane. Due to the large difference in their physical properties, the binary gas mixture of ammonia
and methane can be separated relatively easier compared to the binary gas mixture of nitrogen and
methane.
The electrochemical reduction of nitrogen to ammonia is still in its infantile phase, and so far there is
no reported study of using this method to capture nitrogen from natural gas. In order to estimate its
feasibility for the application of nitrogen capture from natural gas, the parameter called ammonia
38
|
UWA
|
production rate would be examined in this section, which indicates how fast nitrogen can be captured
from natural gas and then converted to ammonia. Similar to the analysis of BNF, a fast enough
ammonia production rate is essential to process large natural gas streams with a reasonable scale of
the electrolytic cells.
The highest ammonia production rate so far is achieved by electrolytic cells using polymer proton
exchange membrane (PEM) as the electrolyte.[170] Because of their high proton conductivity at low
temperature and the well-established knowledge and techniques of cell construction and assembly in
the area of the fuel cell, PEM based electrochemical cells show the greatest potential for the
electrochemical reduction of nitrogen from natural gas streams. Kordali et al. are among the first
scientists who study ammonia synthesis under atmospheric pressure and below 100 ยฐC and the
ammonia production rate they observed is 2.12*10-11 mol cm-2 s-1 at 90ยฐC with a low current efficiency
of 0.24%.100ยฐC.[171] [172] Xu et al. reported the highest ammonia production rate of 1.13*10-8 mol
cm-2 s-1 with a current efficiency of 90% at 80ยฐC, which was achieved by using Nafionยฎ membrane 102
as the electrolyte, SmFe Cu Ni O ฮด as the cathode and Ni-samarium doped ceria (Ce Sm O ฮด)
0.7 0.1 0.2 3โ 0.8 0.2 2โ
(Ni-SDC) as the anode.[170, 172] Following this work, although other combinations of anode, cathode
and electrolyte have been investigated, [170] [173] 1.13*10-8 mol cm-2 s-1 remains the highest
ammonia production rate. Because the PEM in this ammonia production electrolytic cell is similar to
that in the well-studied fuel cell, the ammonia production rate thus has the potential to rise to the
range of 4.3 โ 8.7* 10-7 mol cm-2 s-1 when the current densities could achieve 0.25 โ 0.5 A cm-2 with a
50% current efficiency.[37]
The highest reported ammonia production rate and the highest potential ammonia production rate
will be used to estimate the feasibility of using electrolytic cells to capture nitrogen from methane.
Given a natural gas flow rate which is assumed to contain 10% nitrogen, the required area of the PEM
can be estimated from the ammonia production rate. Table 2.9 summarises the required areas of
electrolytic cell based on the above two ammonia production rates to process different natural gas
flow rates. The price of PEM, Nafionยฎ membrane 102, varies but locates within the range of 1000 โ
3000 USD/m2 on the current market (USD: United States Dollar). The capital cost of the PEM only is
shown in Table 2.9. It is apparent from Table 2.9 that electrochemical route technically has the
potential to capture N from natural gas and then convert it to ammonia, but the capital cost of the
2
PEM only is already far too high to make it economically feasible.
Despite the promising theoretical performance, this technology is still in its early stage and the actual
current ammonia production rate is still far from the theoretical one. Furthermore, the required area
of PEM is too large to be processed and the scale-up of the anode and cathode catalysts also pose
39
|
UWA
|
Table 2.9 The estimated areas of electrolytic membrane and their capital cost to handle different
natural gas flow rates with different ammonia production rates
Ammonia Production Natural Gas Flow rate Area of Nafionยฎ Capital cost of Nafionยฎ
Rate (10 mol % Nitrogen) membraneโข membrane
mol/s/cm2 MMscfd 103 m2 million USD
1 24.5 24.49
1.13E-08โ 15 367 367.29
75 1840 1836.45
1 0.318 0.32
8.70E-07โก 15 4.77 4.77
75 23.9 23.85
โ The highest ammonia production rate reported so far.
โก The ammonia production rate that could be potentially achieved at the current of 0.25 โ 0.5 A cm-
2 with a 50% current efficiency, similar to that has been accomplished in fuel cells.
โขNafionยฎ membrane is a proton conductivity membrane and functions the electrolyte
2.2.5 Lithium Based Process
Lithium reactions with different gas species, including nitrogen, have been studied for decades. The
primary reason for this is the safety hazards caused by lithium spills in fusion reactors[174, 175], in
which lithium functions as a tritium breeder blanket and as a coolant.[176, 177] Recently, lithium has
also been proposed as carriers in energy circuits based on renewable energy.[178-180] Lithium reacts
with CO and N , the flue gas from power plants to generate electricity, and the resulting Li N and
2 2 3
Li CO would go through a series of treatments to become LiCl and eventually be electrolyzed back to
2 3
lithium metal by seasonal renewable energy (wind energy, solar energy). All the studies mentioned
above are focusing on the reactions of lithium with nitrogen at high temperatures and some of them
involves lithium combustion in nitrogen atmosphere usually at temperatures above 1000 ยฐC.[46] Such
41
|
UWA
|
harsh conditions would pose serious safety hazards to the natural gas processing plants, and thus
these high-temperature reactions are not suitable to be utilized to separate nitrogen from methane.
The reaction of lithium with nitrogen at low temperatures, especially at temperatures lower than the
melting point of lithium (180.6 ยฐC), are relatively less documented and the mechanism remains a
mystery. In the presence of moisture, lithium can react with nitrogen smoothly under moderate
conditions.[41, 46, 181-183] However, when water is absent, it becomes complicated. Some studies
show that, at the absence of water moisture, lithium remains stable in dry nitrogen and even dry air
for days under room temperature;[41, 184] while, others claim that the lithium can react with dry
nitrogen readily under moderate temperatures.[42, 185, 186] Here, dry nitrogen usually means
nitrogen that contains less than 10 ppm water moisture, but usually the nitrogen are pretreated with
the various method and the water contents should be smaller than 2 ppm.[182] Interestingly to notice
the lithium metal that can not react with dry nitrogen under moderate temperatures are usually have
a storage history, either in hydrocarbon solvents[41], or in Ar atmosphere. In contrast, the lithium that
can react with dry nitrogen is in-situ freshly made, either by recrystallization from molten lithium,[42]
vapor lithium,[185] or by electrodeposition.[186] Or, the lithium metal has fresh cut edges where the
reaction always started.[183] After the reaction of lithium with nitrogen is initiated, it becomes self-
sustained and gives a sigmoid growth curve with time.[42, 183, 185]
One of the applications of the reaction of lithium with nitrogen at moderate temperatures is a lithium
battery. Some researchers claimed that a passivation layer of lithium nitride on the surface of the
lithium anode could prevent the dendritic growth, and thus improve the performance of lithium metal
battery significantly.[186] While, other researchers reported that under room temperature the
charge-discharge cycles in the lithium-ion battery could facilitate the reaction of lithium and nitrogen,
which could serve as a novel method for the preparation of lithium nitride. Traditionally, the
preparation of lithium nitride usually involves high temperatures where lithium metal stays in liquid
form, and thus poses safety hazards during operation.[187-189]
In the 1990s, the use of lithium metal started to show up in the area of gas separation, patented to
separate trace amounts of nitrogen impurities from crude argon.[190, 191] The crude argon passes
through a nitrogen removal unit which contains lithium supported on high surface area materials or
staying in a molten state. The preferred operating temperature ranges from 100ยฐC to 200ยฐC (low
temperatures for high surface area materials supporting lithium and high temperatures for molten
lithium). Due to lithium`s high reactivity under such conditions, nitrogen can be removed and
converted to lithium nitride effectively.
42
|
UWA
|
A similar process to the purification process of argon can be designed to remove nitrogen from natural
gas using the reaction of lithium with nitrogen under moderate conditions. This nitrogen separation
process by lithium metal can be operated in a batch mode which is akin to a pressure or temperature
swing adsorption process. A significant advantage of using lithium to separate nitrogen is that the
theoretical loading capacity of nitrogen on lithium has the potential reach 24mmol/g, which is an
encouragingly high capacity compared to those of traditional adsorbents, which are smaller than
1mmol/g at 100 kPa.[190] Moreover, because lithium does not react with methane, the selectivity of
nitrogen over methane on lithium would be significantly high which would lead to a full recovery of
methane in such a nitrogen separation process. The two challenges of such process are (1) making the
reaction of lithium with nitrogen happen under moderate conditions; (2) regeneration of lithium metal
from produced lithium nitride which might need many harsh conditions.[43, 192]
No studies of utilizing lithium to remove nitrogen from natural gas have been reported so far, yet given
the right operating condition, lithium-based technology to remove nitrogen from natural gas could be
potential to meet the criteria of being energy efficient, scalable, and safe. Additionally, lithium has
more than an order magnitude higher capacity of nitrogen loading compared to traditional
adsorbents. As such, an intensive study of using lithium to capture nitrogen from methane will be
discussed in Chapter 5.
2.3 Conclusions
In this chapter, a review of conventional and emerging technologies to remove nitrogen from natural
gas has been given. Additionally, a set of criteria to examine these technologies have also been
presented, namely being energy efficient, scalable, and safe according to industrial standards. A
review of conventional and emerging technologies of nitrogen separation from natural gas covered in
this chapter has been summarized in Table 2.10.
Through the critical analysis laid out in this chapter, three technologies are selected to be studied in
more depth within this thesis: (1) Adsorption, the most promising amongst the other conventional
technologies. The adsorption method is largely mature, however have much potential for optimization
within the particular application of nitrogen removal from natural gas, (2) TMC based Absorption, an
emerging technology which is inspired by BNF and able to deliver good performance even at higher
flowrate of natrual gas feed, but would have been better and much more attractive for industrial
application if the right TMC and solvent pair could be determined, and (3) Lithium metal based
separation, an emerging technology from the 1990s that worth further reevaluation due to its
prominent benefit of having a large nitrogen loading capacity and the pressing and growing global
43
|
UWA
|
Chapter 3: Adsorption Equilibria and Kinetics of CH and N on
4 2
Commercial Zeolites and Carbons
3.1 Forward
This chapter forms the basis of a journal paper published on Adsorption (January 2017, Volume 23,
Issue 1, pp 131โ147) and so has been written in that format.
3.2 Abstract
Adsorption equilibria and kinetics are two sets of properties crucial to the design and simulation of
adsorption based gas separation processes. The adsorption equilibria and kinetics of N and CH on
2 4
commercial activated carbon Norit RB3, zeolite 13X, zeolite 4A and molecular sieving carbon MSC-3K
172 were studied experimentally at temperatures of (273 and 303) K in the pressure range of (5 to
120) kPa. These measurements were in part motivated by the lack of consistent adsorption kinetic
data available in the literature for these systems, which forces the use of empirical estimates with
large uncertainties in process designs. The adsorption measurements were carried out on a
commercial volumetric apparatus. To obtain reliable kinetic data, the apparatus was operated in its
rate of adsorption mode with calibration experiments conducted using helium to correct for the
impact of gas expansion on the observed uptake dynamics. Analysis of the corrected rate of adsorption
data for N and CH using the non-isothermal Fickian diffusion (FD) model was also found to be
2 4
essential; the FD model was able to describe the dynamic uptake observed to better that 1 % in all
cases, while the more commonly applied isothermal Linear Driving Force model was found to have a
relative root mean square deviation of around 10 %. The measured sorption kinetics had no
dependence on gas pressure but their temperature dependence was consistent with an Arrhenius-
type relation. The effective sorption rates extracted using the FD model were able to resolve
inconsistencies in the literature for similar measurements.
3.3 Introduction
Adsorption based processes are well-established technologies for the separation of gas mixtures in,
for example, the air separation industry [67], the hydrogen production industry [68-70], and the
capture of carbon dioxide from flue gases [1, 71-74]. In the past several decades, adsorption based
processes for separating nitrogen and methane have attracted significant attention in areas such as
47
|
UWA
|
natural gas production [1, 75], landfill gas upgrading [193, 194], coalbed methane enrichment [195],
and coal mine methane/ventilation air methane purification [196].
In the design of an adsorption process for nitrogen and methane separation, estimates of the lengths
of the saturation and mass transfer zones are crucial to specify the height of the adsorption bed.
Equilibria data for nitrogen and methane mixtures, which are often available either from literature
reports or direct measurements, are the foremost information required for such estimates.
Information about the effective sorption kinetics is also essential to properly estimate the length of
the mass transfer zone in the bed because, for many adsorption based applications, the mass transfer
from gas phase to solid phase is limited by sorption kinetics. This mass transfer limitation arises from
the fact that the diffusion of gases into the porous interior of the adsorbent is restricted and that the
intrinsic adsorption rate is usually much faster than the diffusion rate [197]. The kinetics are especially
important when the adsorbents do not possess high equilibrium selectivity for the mixture
components because for such scenarios the mass transfer fronts for different adsorbates could be
very close to each other. Adsorption kinetics are even more vital to the design of pressure swing
adsorption (PSA) separations for nitrogen and methane mixtures that exploit differences in their
sorption rates, and are crucial parameters in the development of accurate and reliable simulations of
industrial PSA processes intended to separate N and CH . However, kinetic parameters must often be
2 4
estimated empirically for such applications since the relevant data in the literature are limited and
inconsistent, even for commercial adsorbents. This is partly because of the difficultly associated with
accurate measurements of adsorption kinetics [198].
In principle, adsorption kinetics can be measured experimentally using a variety of techniques, for
example, by monitoring the time-dependence of volumetric or gravimetric sorption capacity [199,
200], by the combined pressure-swing and volume-swing frequency response technique [201, 202],
using a dynamic column breakthrough apparatus [203], and via zero length column (ZLC) experiments
[204, 205]. However, the results of kinetic measurements made by different groups and/or with
different techniques tend to have large deviations. A survey of literature (Figure 3.1) revealed that the
reported effective sorption rates for N and CH on commercial adsorbents similar to those measured
2 4
in this work have large variations, ranging in some cases over two orders of magnitude.
48
|
UWA
|
Figure 3.1 Variations in effective sorption rate (D/r2) reported in the literature for N and CH on
2 4
activated carbon (AC) [94, 200, 203, 206], zeolite 13X [207-209], carbon molecular sieve (CMS) [199,
202, 210] and zeolite 4A [211-213]
One possible reason for the large deviations present in the literature data is that kinetic measurements
are often analysed under the assumption of a constant temperature throughout the adsorbent,
usually on the basis that the sample mass is small (โค 1 g). However, the adsorption rate observed in
such measurements is generally non-isothermal because heat is evolved during the process and
cannot be removed instantaneously due to heat transfer limitations. Temperature rises due to
sorption can impact the apparent kinetics in different ways, such as increasing gas diffusivities while
decreasing equilibrium capacities [214]. The non-isothermal effects associated with these adsorption
kinetic measurements are often erroneously overlooked by extracting adsorption kinetics from the
experimental data using the isothermal linear driving force model [203, 215, 216], which is widely
implemented in simulations of adsorption processes. In particular, for volumetric sorption kinetics
measurements the delay in signal response caused by value effect or gas diffusion resistance were
often ignored although several workers in the past have aimed to account for the dynamics of sorption
uptake in volumetric experiments [217, 218].
In this work, the adsorption equilibria and kinetics of methane and nitrogen on four commercial
adsorbents (Norit RB3, zeolite 13X, molecular sieve MSC-3K 172 and zeolite 4A) were measured to
assess these adsorbents for potential applications in the separation of methane and nitrogen. The
sample size (mass) of each adsorbent used was carefully chosen to give sufficient signal to noise ratio
49
|
UWA
|
after it was confirmed that any heat and mass transfer limitations due to sample size were negligible.
The kinetics were obtained from rate of adsorption measurements made using the volumetric method
at two temperatures of 273 K and 303 K with gas pressures up to 120 kPa. Two important steps were
applied to the analysis of the dynamic uptake data to extract reliable values of the effective sorption
rate (D/r2) for methane and nitrogen on each adsorbent: calibration of the effects at short time scales
due to gas expansion using helium, and regression of a non-isothermal kinetic model to the measured
data. The impact of adsorption heat on the apparent kinetics was evaluated by comparing the results
of isothermal and non-isothermal kinetic models. The effect of temperature on the adsorption kinetics
was also studied and compared to an Arrhenius-type correlation.
3.4 Experimental Section
3.4.1 Method and Materials
The Norit RB3 extrude pellet used in this work is the same activated charcoal studied by Rufford et al.
[203], supplied by IMCD Australia Ltd. Zeolite 13X APG extrude pellet was supplied by Shanghai MLC
Molecular Sieve Co., Ltd. The Linde zeolite 4A extrude pellet was purchased from Sigma-Aldrich (Castle
Hill, NSW, Australia). The granular molecular sieving carbon MSC-3K 172 was obtained from Japan
EnviroChemicals, Ltd. (Osaka, Japan). The properties of these adsorbents relevant to the rate of
adsorption measurements are summarised in Table 3.1; other properties for these materials can be
found in elsewhere [203, 219, 220]. The sample mass to be used in these experiments was determined
by first conducting ROA measurements for N on zeolite 4A at 283 K. Two sample masses of 0.7704 g
2
and 0.3694 g were tested for the measurements and the uptake data for these two masses are
presented in the supporting information (SI Figure 1). It is clear from the figure that the signal-to-noise
for each of these two masses is much larger than for the other results presented in this work. One
reason for these noisy data is the low overall N adsorption that occurs for that amount of sample at
2
283 K, which is comparable to the measurementโs uncertainty. Thus, for measurements at 303 K, one
would expect the signals to be even noisier than those measured at 283 K if the same amount of
adsorbent were used, since the adsorption capacity at 303 K is lower than that at 283 K. We then used
1 gram of zeolite 4A for the N ROA measurement at 303 K and it was found this was an appropriate
2
sample size for ROA measurements. Accordingly, we used about 1 gram of zeolite 4A for all our
measurements and used a similar mass for all other adsorbents for the sake of comparison on an
equivalent basis. All gases used for analysis were of high purity supplied by Coregas Australia, with the
following specified molar purities: CH : 99.995 %; N : 99.999 % and He: 99.995%.
4 2
50
|
UWA
|
Table 3.1 Physical properties of the adsorbents
Norit RB3 Zeolite 13X Zeolite 4A MSC-3K 172
Sample mass (g) 1.0077 0.8092 1.0328 1.0511
Particle Diameter* (mm) 3.2 1.4 2.0 1.0
Particle length* (mm) 4.0 4.0 3.6 2.0
Pore volume (cm3/g) 0.21 0.33 0.28 0.22
Pellet density* (g/cm3) 0.85 1.12 1.09 0.90
*Assuming cylindrical, averaged over measurements of 20 particles
Equilibrium adsorption capacity in terms of pellet mass and adsorption rates of CH and N on four
4 2
adsorbents were measured at two temperatures (273 and 303) K using a volumetric adsorption
measurement system, the Micromeritics ASAP 2020, in Rate of Adsorption (ROA) mode. Prior to these
measurements, each of the adsorbents was degassed under vacuum (1 Pa) in a sample tube for 12
hours at the recommended temperatures for the given adsorbents: 473 K for Norit RB3 and MSC-3K
172, 623 K for zeolite 13X and zeolite 4A [72, 203, 220, 221]. After degassing, the sample tube was
backfilled with helium and transferred to the analysis port for measurement of the adsorption amount
and rate of pure gases. The measurements were then carried out from low pressure (5 kPa) to high
pressure (120 kPa) incrementally to ensure that upon a step pressure change the driving force of
adsorption was sufficiently small to be considered linear as assumed by the Fickian diffusion model
[218, 222]. In the ROA mode of the ASAP2020, the pressure in the sample tube was automatically
monitored and recorded by the data acquisition software on the instrument once a certain amount of
gas was dosed. The pressure change in the sample tube was then converted to molar amount of gas
adsorbed per unit mass of the adsorbent using the reference equation of state for the respective pure
gases [223, 224]. In this work, the pressure change in the sample tube was recorded every 0.4 s
immediately following the introduction of a gas dose (0.223 mmol gas per gram of adsorbent). The 0.4
s data acquisition period, which limits the fastest sorption rate that can be measured with the
apparatus, is the smallest meaningful time interval that can be studied with the instrument even
though the operating software allows a 0.1 s interval to be specified, because the data acquisition
system on the ASAP 2020 executes a 400 ms averaging of the analogue pressure transducer signals.
For all the rate of adsorption measurements, 1000 pressure data points (the maximum number
allowable by the data acquisition software) were recorded following the introduction of the gas dose.
Representative uptake curves of three consecutive equilibrium pressure points for โslowโ and โfastโ
cases are shown in Figure 3.2. For systems with slow adsorption, the rate of adsorption over a pressure
51
|
UWA
|
step was monitored for only the first 400 seconds (0.4 s ร 1000) and then the uptake rate on the
adsorbent was no longer recorded while the system was allowed to proceed towards equilibrium.
Once the equilibrium criterion (a pressure change of less than 0.01 % in a time interval of 30 s) was
satisfied, the measurement of the next pressure point was automatically initiated. For systems with
faster adsorption where equilibrium could be obtained within 400 seconds, all of the uptake data of
gas on the adsorbent were recorded.
Figure 3.2 Examples of rate of adsorption measurements for CH on Norit RB3 (fast) and zeolite 4A
4
(slow) for three consecutive pressure steps at 303 K. The solid curves correspond to ROA
measurements, while the dashed curve indicates the pressure evolution that occurred prior to the
attainment of equilibrium.
3.4.2 Calibration with Helium
To properly analyse the data acquired during the rate of adsorption measurements, blank experiments
with helium were carried out for each of the adsorbents with exactly the same run files as for CH and
4
N at the same temperatures. This enabled measurement of the pressure in the sample tube following
2
the opening of the dose valve due to the gas expansion and in the absence of any adsorption (assuming
the adsorption of helium on the adsorbents was negligible [225]). The blank experiments were
particularly important for two reasons: first, the very first pressure data point recorded for each rate
of adsorption measurement by the ASAP 2020 data acquisition system actually corresponded to the
pressure in the reference volume (solid red line in Figure 3.3) before Valve 12 (Figure 3.3) was opened
to introduce gas into the sample tube and was not representative of the actual pressure in the sample
tube. Second, for most measurements of gas adsorption kinetics based on the volumetric technique
it is important to account for the time required for gas flow and expand into the sample tube.
Determining the correct initial pressure in the sample tube is critical because it sensitively affects a
52
|
UWA
|
key parameter (ฮฑ*) in the non-isothermal dynamic model to which the dynamic pressure data are
regressed, as discussed further in Section 3.
Figure 3.3. Schematic of the volumetric adsorption system: red solid line represents the apparatusโ
reference volume, used to calculate the amount of gas dosed into the sample volume below V12.
The four pressure transducers could also be isolated (valves not shown) from the manifold whenever
the pressure exceeded their full scale.
3.5 Models for Adsorption Kinetics
The linear driving force (LDF) and Fickian diffusion (FD) equations are two kinetic models commonly
used to extract adsorption kinetics information from experimentally measured rate of adsorption data
[200, 207, 226, 227]. Commonly, the LDF model is used instead of the FD model to represent
adsorption kinetics within adsorption process simulations even though LDF predictions often show
larger deviations from experimentally measured kinetic data than those made with FD models. This
results from the fact that in adsorption process simulations, the LDF model allows elimination of the
integration step at the particle level which is required for the FD model and, thus, use of the LDF model
significantly reduces computational time [226]. However, the widespread availability of high
performance computers nowadays may enable the use of more accurate FD models to simulate cyclic
adsorption processes even though they are more computationally intensive.
The linear driving force model was initially developed by Glueckauf and Coates [228] for interpreting
the incomplete equilibrium of the front boundary in adsorption chromatography under isothermal
conditions, and it has since been adopted for describing adsorption kinetics as summarised by Sircar
[226]. The general form of the LDF model for a single adsorbate is given by:
53
|
UWA
|
Eqn. 1
๐
๐
๐๐๐๐
๐
๐
๐๐ = ๐๐[๐๐โโ๐๐๐๐]
Here is the molar adsorbed amount of adsorbate on the adsorbent particle at time , and is
the e ๐๐qu๐๐ilibrium adsorbed amount of adsorbate at the gas phase pressure and th ๐ก๐กe adso ๐๐rbโent
temperature . The is the effective LDF mass transfer coefficient for the equilibrium adsorbed
๐๐
amount of . The analytical solution of the LDF model [226] is used for the regression of data
๐ด๐ด ๐๐
obtained vi ๐๐a โconstant volume-variable pressure experiments when the adsorption process can be
assumed an isothermal process:
Eqn. 2
๐๐๐๐โ๐๐๐๐
โ
๐๐โโ๐๐๐๐ = ๐๐โ๐๐๐๐๐๐[โ(๐๐+๐ถ๐ถ )๐๐๐๐]
Eqn. 3
โ 0+ โ
๐ผ๐ผ ๐๐ โ๐๐
โ 0+ 0โ
1+๐ผ๐ผ = ๐๐ โ๐๐
Here is the equilibrium adsorbed amount on the adsorbent at the start of a pressure step, is
โ
dimen ๐๐si0onless and related to the ratio of the amount gas adsorbed during the pressure step to
๐ผ๐ผ
the
amount of gas introduced to the system at time zero of the pressure step as indicated in Eqn. 3,
0+
is the gas phase pressure immediately after opening the dosing valve, is the final equilibrium
๐๐
โ
pressure for a pressure step, is the gas phase pressure before opening the dosing valve. The mass
๐๐
0โ
transfer coefficient can be correlated with the sorption rate using the commonly
๐๐
2
recommended relation for adsorption process design [229].
๐๐ ๐ด๐ดโ๐๐
2
๐๐ = 15๐ด๐ดโ๐๐
The Fickian diffusion (FD) model describes the mass transfer between the gas and adsorbed phases
using Fickโs law of diffusion in terms of the sorption rate (D/r2) for a gas on given adsorbents. For a
constant volume system, and under the assumptions (1) that the gas diffusivity is independent of the
amount of adsorbed phase, and (2) the uptake is isothermal, the analytical solution to the dynamic
mass balance equation is given as follows [218]:
Eqn. 4
โ ๐๐๐ซ๐ซ
๐๐๐๐ โ ๐๐(๐๐+๐ถ๐ถ )๐๐๐๐๐๐๏ฟฝโ๐๐๐๐ ๐๐๐๐๐๐๏ฟฝ
๐๐โ = ๐๐โโ๐๐=๐๐ ๐๐๐ถ๐ถโ (๐๐+๐ถ๐ถโ )+๐๐๐๐๐๐
where
Eqn. 5
2 โ
๐๐๐๐ = (1+๐๐๐๐โ3๐ผ๐ผ )โ๐๐๐๐๐ก๐ก๐๐๐๐
54
|
UWA
|
Here is the root of transcendental equation (Eqn. 5). Eqn. 4 is strictly limited to the analysis of
isothe ๐๐r๐๐mal uptake data for cases where the heat resulting from the adsorption can be considered as
being instantaneously dissipated to the environment. To account for situations where the heat
released by adsorption cannot be dissipated fast enough to achieve an isothermal condition, the non-
isothermal kinetic model developed by Kocirik et. al. [218] can be used. This model was obtained for
non-isothermal, constant-volume but variable-pressure conditions, by deriving an analytical solution
to the simultaneous mass and heat transport equations. The main assumptions for this model are that
the adsorption rate was controlled by intracrystalline gas diffusion and heat (arising from heat of
adsorption) transfer from the adsorbent surface to the surrounding environment. The analytical
solution is as follows:
2
โ โ ๐๐๐๐ 2 ๐ด๐ด Eqn. 6
9(1+๐ผ๐ผ )๏ฟฝ 2๏ฟฝ exp๏ฟฝโ๐๐๐๐ 2๐ก๐ก๏ฟฝ
๐๐๐๐ โ๐๐๐๐ ๐๐
= 1โ ๏ฟฝ โ
๐๐โ
๐๐=1
1
โ
+3 ๐ฝ๐ฝ โ๏ฟฝ๐๐๐๐๐๐๐๐๐ก๐ก๐๐๐๐๏ฟฝ๐๐๐๐ 2๏ฟฝ+1๏ฟฝ+3๐ผ๐ผ 4๐ต๐ต๐๐
โ
๐ฝ๐ฝ๐๐ 2๐ฝ๐ฝ๐๐ ๐๐๐๐ 2๐๐๐๐๐ฝ๐ฝ๐๐
Here is given by the roots of the following equation:
๐๐๐๐
Eqn. 7
โ
2 3๐ผ๐ผ 2
(โ๐๐๐๐+๐ผ๐ผ)+3๐ฝ๐ฝ๐๐๐๐โ 2 (โ๐๐๐๐+๐ผ๐ผ)๐๐๐๐ = 0
๐๐๐๐
with:
Eqn. 8
๐๐๐๐ = ๐๐๐๐๐๐๐๐๐ก๐ก๐๐๐๐โ1
and with the additional parameters defined as follows:
Eqn. 9
โ๐๐ ๐ด๐ด
๐ผ๐ผ = ๏ฟฝ 2
๐๐๐๐๐๐๐๐ ๐๐
Eqn. 10
โ
โ๐ถ๐ถ ๐๐๐๐
๐ฝ๐ฝ = ๏ฟฝ ๏ฟฝ
๐๐๐๐ ๐๐๐ด๐ด ๐๐๐๐(0),๐๐0
Eqn. 11
โ
1 1 3๐ผ๐ผ ๐๐๐๐
โ = ๏ฟฝ1โ 2 ๏ฟฝ
๐ฝ๐ฝ๐๐ ๐ฝ๐ฝ ๐๐๐๐
Eqn. 12
2 2 2
๐ต๐ต๐๐ = ๐๐๐๐[(๐๐๐๐โฮฑ)๐๐๐๐๐๐๐๐๐ก๐ก๐๐๐๐โ2ฮฑ]+๐๐๐๐(๐๐๐๐โฮฑ)
55
|
UWA
|
Here is the overall heat transfer coefficient between the external surface of the adsorbent and its
surroundings, is the external surface area per unit volume of the adsorbent, is the adsorbent
โ
density, is th ๐๐e heat capacity of the adsorbent at constant pressure, and is h ๐๐e๐๐at of adsorption.
Both the๐๐ i๐๐ sothermal LDF model (Eqn. 2) and non-isothermal FD model (Eqnโ.๐ถ๐ถ 6) were regressed to the
experimental uptake curves by adjusting the parameter in the LDF model and by adjusting the
2
parameters ฮฑ, ฮฒ and in the non-isothermal FD to minimise the root mean squared deviation
๐ด๐ดโ๐๐
2
(RMSD) between the dynamic uptake data and the model according to Eqn.13.
๐ด๐ดโ๐๐
๐๐ 2
โ๐๐=1๏ฟฝ(1โ๐๐๐ก๐กโ๐๐โ)๐๐๐๐๐๐โ(1โ๐๐๐ก๐กโ๐๐โ)๐๐๐๐๐๐๐๐๐๐๏ฟฝ
๐
๐
๐๐๐๐๐ด๐ด = ๏ฟฝ ๐๐ Eqn. 13
where N is the number of experimental data points. Since is normalised and
dimensionless, the RMSD should have a value between 0 (indicati (n 1g โa p ๐๐e๐๐rf โe ๐๐ctโ f )it) and 1 (indicating a
very bad fit).
3.6 Results and Discussion
3.6.1 Adsorption Equilibria
Figure 3.4 shows the measured pure component adsorption isotherms for CH and N on the four
4 2
adsorbents, and Table 3.2 lists the data measured for these isotherms, together with their combined
standard uncertainties which were calculated according to uncertainty propagation formula shown in
Eqn. 14 [230]. Assuming that the input quantities are not correlated, the combined standard
uncertainty in the adsorption capacity for the present measurements can be calculated as
Eqn. 14
๐๐๐ธ๐ธ๐๐ ๐๐ ๐๐๐ธ๐ธ๐๐ ๐๐ ๐๐๐ธ๐ธ๐๐ ๐๐ ๐๐๐ธ๐ธ๐๐ ๐๐
๐๐(๐ธ๐ธ๐๐)= ๏ฟฝ๏ฟฝ๏ฟฝ๐๐๐๐๏ฟฝ๐๐(๐๐)๏ฟฝ +๏ฟฝ๏ฟฝ๐๐๐๐๏ฟฝ๐๐(๐๐)๏ฟฝ +๏ฟฝ๏ฟฝ๐๐๐๐๏ฟฝ๐๐(๐๐)๏ฟฝ +๏ฟฝ๏ฟฝ๐๐๐๐๏ฟฝ๐๐(๐๐)๏ฟฝ
where denotes the standard uncertainty of a quantity x and V is sample tube volume. The
uncertainty in the sample mass was estimated from the balance resolution, the uncertainty of the
๐๐(๐๐)
pressure was taken to be 0.15 % of the equilibrium reading, and the uncertainty of the temperature
was set by the stability of the liquid bath. The uncertainty in the volume was estimated from the
standard deviation in the free gas volume (โ20 mL) determined automatically by the apparatus
corresponding to each equilibrium pressure measurement. Numerical values of these component
uncertainties are listed in Table 3.2.
56
|
UWA
|
303.15 16.37 0.1867 0.0027 303.15 20.62 0.0572 0.0027
303.15 22.10 0.2410 0.0032 303.15 27.47 0.0737 0.0032
303.15 27.89 0.2920 0.0035 303.15 34.32 0.0894 0.0035
303.15 33.72 0.3408 0.0039 303.15 41.18 0.1043 0.0039
303.15 39.71 0.3882 0.0042 303.15 48.24 0.1186 0.0042
303.15 45.74 0.4338 0.0045 303.15 55.11 0.1319 0.0045
303.15 51.83 0.4777 0.0048 303.15 62.13 0.1446 0.0048
303.15 57.98 0.5202 0.0050 303.15 69.05 0.1564 0.0050
303.15 64.18 0.5615 0.0053 303.15 76.06 0.1674 0.0053
303.15 70.31 0.6008 0.0055 303.15 83.07 0.1778 0.0055
303.15 76.56 0.6396 0.0057 303.15 90.07 0.1870 0.0057
303.15 82.87 0.6774 0.0059 303.15 97.06 0.1956 0.0059
303.15 89.17 0.7142 0.0061 303.15 104.11 0.2036 0.0061
303.15 95.49 0.7497 0.0063 303.15 111.14 0.2108 0.0063
303.15 101.86 0.7846 0.0065
303.15 108.23 0.8185 0.0067
303.15 114.64 0.8517 0.0069
Figure 3.4 clearly shows that all of these adsorbents have higher CH capacities than N capacities at
4 2
the same temperature and pressure. At 303 K, MSC-3K 172 and Norit RB3 exhibited a similar
adsorption capacity for methane, which is about 40% higher than the corresponding methane
capacities observed for zeolite 4A and zeolite 13X at around 100 kPa. The nitrogen adsorption capacity
for MSC-3K 172, Norit RB3, zeolite 4A and zeolite 13X were essentially identical at 303 K. At 273 K,
both MSC-3K 172 and Norit RB3 displayed slightly higher (~20%) capacities for methane than those
for the other adsorbents at pressures up to 40 kPa. The nitrogen adsorption capacity for MSC-3K 172,
zeolite 4A and zeolite 13X were essentially identical at 273 K, and about 30% higher than that for Norit
RB3.
3.6.2 Kinetic Measurement Calibration
To accurately measure sorption kinetics using a volumetric system, the time delay in recorded data
caused by the effects of the valve and the sensorsโ response needs to be accounted for when
processing the uptake data. In the work of Hu et al. [217], for example, the initial 5 seconds of pressure
data were discarded by the authors presumably because, as we found, the dynamics of the early
transient is not captured correctly by existing models. The objective of this work is to explicitly account
for those very early dynamics through the use of a helium blank run at the corresponding pressures
62
|
UWA
|
where measurements with CH and N (which both adsorb but also experience similar transient
4 2
dynamics) were acquired. Helium was used to calibrate the effects of gas expansion upon dosing for
all the kinetic measurements conducted in this work to enable correction of the initial condition (time
origin and pressure) for the subsequent analysis of the dynamic uptake. Figure 3.5 shows an example
of why the helium calibration was necessary to obtain accurate sorption kinetics from the raw
pressure readings acquired with the apparatus, through the comparison of ROA data obtained for
helium and methane on Norit RB3 at 80 kPa and 273 K. At 273 K, the methane pressure dropped
quickly from 86.74 kPa at t = 0 to 80.63 kPa at t = 0.8 s, with the final equilibrium pressure being 80.08
kPa. These raw pressure data superficially indicate that 92 % of the sorption capacity at 273 K was
reached within 0.8 seconds of gas dosage, which could be interpreted erroneously as extremely fast
sorption kinetics if these raw data were used. However, comparison of these raw pressure data for
methane with those obtained during the corresponding helium calibration experiment shows that
those initial rapid pressure changes were more likely caused by gas expansion rather than adsorption
given that the experiment with helium also exhibited a similar rapid pressure drop within the first 0.8
seconds, with a negligible pressure decrease thereafter. Clearly then, the pressure drop occurring in
the first 0.8 seconds is dominated by gas expansion, and the condition from which the initial time and
pressure datum should be taken for the analysis of sorption kinetics is t = 0.8 s and p โ 80.63 kPa. To
capture a more precise estimate of the initial pressure, which accounts for any possible methane
adsorption that did occur during the first 0.8 s, the pressure ratio obtained for helium between t = 0.8
s and t = 0 s can be applied to the methane pressure recorded at t = 0 s. Across all the experiments
conducted, the time scale of the gas expansion effect as measured with helium ranged from 0.8 to 5.6
s, and it was found to be important to perform helium calibration runs for each adsorbent.
Figure 3.5 Pressure readings in methane rate of adsorption measurements on Norit RB3 at around
80 kPa at 273 K together with corresponding helium calibration pressure readings.
63
|
UWA
|
3.6.3 Sorption Rate Determination from Dynamic Uptake with Kinetic Models
The objective of this paper is to demonstrate a new robust method of extracting effective reliable
sorption kinetic data for materials that may be used in industrial adsorption process design.
Accordingly, we do not discuss extensively the mechanisms governing the observed uptake rates, in
part because it is well known that the controlling mechanism for nitrogen and methane diffusion in
Norit RB3 and zeolite 13X is macropore diffusion [203, 207, 217] and that for zeolite 4A and molecular
sieve carbon is micropore diffusion [199, 211].
Figure 3.6 shows pairs of representative CH and N uptake data obtained at 273 K and about 100 kPa
4 2
for Norit RB3, zeolite 13X, MSC 3K-172 and zeolite 4A. For both gases, the non-isothermal Fickian
diffusion models resulting from the regression to the CH and N uptake data are presented along with
4 2
the experimental data. Additionally, the results of the isothermal LDF model regression to the CH
4
data are shown.
The LDF model is the most widely-used correlation for representing gas adsorption kinetics in
adsorption process simulations. Therefore, fits of this model to the gas uptake data were tested for
each of the adsorbents. However, the LDF fit had particularly large deviations from the experimental
data for Norit RB3 (RMSD = 12.26 %) and zeolite 13X (RMSD = 7.44 %) measured in this study. The LDF
model tends to predict an overall faster approach to equilibrium for these large pore sized adsorbents
but a slower adsorption rate at the initial stage of the adsorption. The deviations of the LDF model
from the experimental data can be explained by the fact that in the experiments, the approach to
adsorption equilibrium was initially accelerated by faster gas diffusion due to the temperature rise
associated with heat-transfer limitations, before the reduction in the adsorbentโs equilibrium capacity
adversely affected the sorption driving force [222]. The isothermal LDF model cannot capture such
heat related effects, and instead assumes that the rate of adsorption depends solely on the initial
driving force for adsorption. The deviations of the LDF model from the experimental data for MSC-3K
172 (RMSD = 1.27 %) and zeolite 4A (RMSD = 2.6 %) are smaller compared to those for the other two
adsorbents, which can be attributed to the small micropores within MSC-3K 172 and zeolite 4A, which
slow gas diffusion therein. This enables the heat of adsorption to be dissipated to the environment
relatively rapidly in comparison with the time scale for sorption, thereby maintaining the system at
near constant temperature.
64
|
UWA
|
Figure 3.6 Experimental and theoretical (black solid lines representing FD model for both CH and N ,
4 2
and blue dash line representing LDF model for CH ) uptake curves of CH and N on a) Norit RB3, b)
4 4 2
Zeolite 13X, c) carbon molecular sieve MSC-3K 172, and d) zeolite 4A pellet at 273 K with pressure
around 100 kPa
When the non-isothermal FD model was used to regress the experimental uptake data, the respective
RMSDs were much smaller than that for LDF model: 0.91 % for Norit RB3, 0.82 % for zeolite 13X, 0.95
% for MSC-3K 172 and 0.96 % for zeolite 4A. The effective sorption rates obtained with the LDF model
were 10 to 20 times smaller than those obtained with the FD model for Norit RB3 and zeolite 13X, and
differed by about a factor of two for MSC-3K 172 and zeolite 4A. Clearly, the non-isothermal FD model
provided a much better description of the experimental uptake curves on all four adsorbents than the
LDF model and for this reason we exclusively used the FD model in all subsequent analysis.
3.6.4 Kinetic Adsorption Results
Figure 3.7 shows the effective sorption rate of CH and N on Norit RB3, zeolite 13X, zeolite 4A and
4 2
MSC-3K 172 as a function of pressure at 303 K and 273 K. The effective sorption rates for CH are of
4
the same order of magnitude for adsorbents Norit RB3 and zeolite 13X, which are approximately 103
times faster than that of CH in zeolite 4A and 104 times faster than that of CH in MSC-3K 172. The
4 4
effective sorption rates for N are also of the same order of magnitude for adsorbents Norit RB3 and
2
65
|
UWA
|
zeolite 13X, which are approximately 102 times faster than that of N in zeolite 4A and 103 times faster
2
than that of N in MSC-3K 172. The effective sorption rate showed only a weak dependence on
2
pressure, which is consistent with the assumption of the FD model that the sorption rates are
independent of loading of adsorbed phase on the adsorbents [218] in the range of pressures studied
in this work.
Figure 3.7 Effective diffusion time constant for CH and N on Norit RB3, zeolite 13X, zeolite 4A and
4 2
carbon molecular sieve MSC-3K 172 obtained by non-isothermal Fickian diffusion model at 303 K
(solid) and 273 K (hollow) at various pressures: a) effective sorption rate constant for CH , b)
4
effective sorption rate constant for N
2
Values of the effective sorption rate for Norit RB3 are listed in Table 3.3, ranging from (0.13 to 0.15) s-
1 for CH and being no smaller than 0.16 s-1 for N at 273 K. At 303 K they ranged from (0.31 to 0.35) s-1
4 2
for CH but the rate of N adsorption was too fast (more than 90 % of adsorption capacity over the
4 2
pressure step was achieved within 1 second) to obtain meaningful values given the limited temporal
resolution of the apparatus. To the best of our knowledge, there has been only one work reporting
the kinetics of CH and N on activated carbon Norit RB3 by Rufford et al. [203]. They used an
4 2
isothermal linear driving force model to extract mass transfer coefficients for CH and N from
4 2
breakthrough experiments conducted in a fixed bed column. However, even the largest effective
sorption rate derived from their reported mass transfer coefficients are about 100 times smaller than
our results. We suspect that the apparent sorption rates extracted from the measurements of Rufford
et al. [21] were significantly afflicted by heat transfer limitations in the fixed bed, and we have since
updated our approach to such DCB sorption rate measurements [231, 232]. Meanwhile, our results
are similar to that of a work by Malek and Farooq [206] for CH on an activated carbon. Their
4
measurements were performed at temperatures from (299 to 338) K and pressures in the range of
(199 to 651) kPa with dynamic breakthrough experiments using dilute CH mixtures. The effective
4
66
|
UWA
|
sorption rate reported by them for these conditions ranged from (0.41 to 0.66) s-1. In addition, our
measurements are consistent with effective sorption rates obtained from N +CH breakthrough
2 4
experiments conducted on a Maxsorb activated carbon at 300 K and 1 atm by Sheikh et. al. [94] (0.197
s-1 for CH and 0.55 s-1 for N ). More recently, Ju et al. [200] reported the kinetics of pure CH and N
4 2 4 2
on cylindrical activated carbon granules with effective sorption rate of (0.06 to 0.08) s-1 for CH in the
4
pressure range of (25 to 78) kPa and (0.06 to 0.08) s-1 for N in the pressure range of (20 to 90) kPa at
2
308 K. These values for methane are about 30 to 50 times smaller than our results at 303 K. However,
we note that the dynamic uptake data shown graphically in reference [200] for CH on activated
4
carbon are reasonably consistent with our dynamic uptake data. Figure 3 of reference [200] suggests
that 90 % of the uptake occurred in about 20 seconds, which is even faster than our observations of
80 % uptake within 20 seconds for CH on activated carbon. As indicated above, systematically low
4
estimates of the effective sorption rate can occur if the LDF model is used to analyse data with
appreciable heating effects. However, Ju et al. [200] stated they used the non-isothermal FD model to
regress their data. Bulow [233] has suggested alternative reasons for the apparently low sorption rates
reported by Ju et al. [200], such as neglecting to deduct the overall response time of the apparatus or
the delay caused by forcing the gas to pass through a 0.5 micron ceramic frit on top of the adsorbent.
67
|
UWA
|
Table 3.3. Effective sorption rate and associated parameters for the non-isothermal FD model in eqs
(6)-(12) for CH and N on activated carbon Norit RB3. Values are not reported for conditions where
4 2
90% of the uptake occurred within 1 s.
Norit RB3
Methane Nitrogen
273.15 K 273.15 K
P D/r2 u(D/r2) P D/r2 u(D/r2)
(kPa) โ๐๐(โs-๐๐1)๐๐ ๐๐๐๐ ๐ฝ๐ฝ (s-1) (s-1) (kPa) โ๐๐(โs-๐๐1)๐๐ ๐๐๐๐ ๐ฝ๐ฝ (s-1) (s-1)
21.41 0.033 0.249 0.14 0.02 24.81 โ โ โ โ
41.57 0.033 0.345 0.15 0.02 43.78 โ โ โ โ
63.22 0.032 0.541 0.14 0.01 63.178 โ โ โ โ
80.08 0.033 0.653 0.14 0.01 82.66 0.034 0.257 0.16 0.04
103.02 0.034 0.73 6 0.13 0.01 102.32 0.034 0.253 0.16 0.03
303.15 K 303.15 K
22.10 0.032 0.155 0.34 0.04 20.62 โ โ โ โ
39.71 0.034 0.203 0.34 0.04 41.18 โ โ โ โ
64.18 0.034 0.334 0.32 0.05 62.13 โ โ โ โ
82.87 0.035 0.38 4 0.32 0.05 83.07 โ โ โ โ
101.86 0.032 0.358 0.31 0.05 104.11 โ โ โ โ
The effective sorption rates measured for CH and N on zeolite 13X are listed in Table 3.4, and range
4 2
from (0.11 to 0.29) s-1 and (0.13 to 0.43) s-1, respectively. These are broadly consistent with kinetic
results reported in the literature. Delgado et al. used pulse experiments to measure the effective
sorption rate for CH at 289 K and 324 K on a zeolite 13X with results ranging from (0.18 and 0.33) s-1
4
[209]. Dantas et al. [208] obtained N mass transfer coefficients on a zeolite 13X using binary gas
2
breakthrough experiments at 1 bar of pressure at various temperatures, and at about 303K, the
effective sorption rate reported for N was 0.18 s-1. However, Park et al. [207] recently reported the
2
kinetics of pure CH and N on spherical pelletised UOP zeolite 13X with effective time of 0.05-0.07 s-
4 2
1 for CH and 0.01-0.08 s- 1 for N in the pressure range of 10-80 kPa and temperature range of (293 to
4 2
323) K. These values are about 3~20 times smaller than our results, which might be explained by the
fact that Park et al. [207] used the same apparatus and method as Ju et al. [200].
68
|
UWA
|
Table 3.4 Effective sorption rate of CH and N on zeolite 13X obtained from the non-isothermal FD
4 2
model Values are not reported for conditions where 90% of the uptake occurred within 1 s.
Zeolite 13X
Methane Nitrogen
273.15 K 273.15 K
P D/r2 u(D/r2) P D/r2 u(D/r2)
(kPa) โ๐๐(โs-๐๐1)๐๐ ๐๐๐๐ ๐ฝ๐ฝ (s-1) (s-1) (kPa) โ๐๐(โs-๐๐1)๐๐ ๐๐๐๐ ๐ฝ๐ฝ (s-1) (s-1)
18.9 0.027 0.258 0.11 0.03 19.5 โ โ โ โ
42.1 0.029 0.404 0.14 0.02 39.4 0.029 0.271 0.19 0.05
60.6 0.027 0.484 0.12 0.02 59.9 0.028 0.334 0.16 0.03
79.6 0.026 0.468 0.11 0.01 80.6 0.027 0.380 0.16 0.03
103.2 0.029 0.583 0.11 0.01 101.7 0.028 0.338 0.13 0.02
303.15 K 303.15 K
20.6 โ โ โ โ 21.8 โ โ โ โ
41.7 โ โ โ โ 38.5 โ โ โ โ
63.2 0.026 0.169 0.29 0.09 60.9 0.030 0.178 0.43 0.08
79.2 0.025 0.167 0.27 0.07 83.4 0.029 0.149 0.40 0.07
100.7 0.027 0.217 0.23 0.05 100.3 0.028 0.138 0.34 0.05
The values of effective sorption rate for MSC-3K 172 are listed in Table 3.5, and at pressures from (20
to120) kPa range from (0.95 to 27.8) ร 10-5 s-1 for CH and (3.61 to 5.57) ร 10-4 s-1 for N at 273 K, and
4 2
from (1.72 to 2.95) ร 10-5 s-1 for CH and (0.88 to 1.35) ร 10-3 s-1 for N at 303 K. Our results are
4 2
consistent with the data of Bae and Lee [210] for CH and N on a CMS-T3A carbon molecular sieve.
4 2
Their measurements were performed at 303 K with pressures in the range of 0-1500 kPa with a
volumetric-type apparatus. The effective sorption rate reported by them for these conditions ranged
from (0.05 to 8) ร 10-5 s-1 for CH and (2 to 30) ร 10-4 s-1 for N . More recently, Yang et. al. [199] reported
4 2
the kinetics of pure CH and N on carbon molecular sieve CMS-131510 with pellet diameters of (0.11
4 2
to 0.13) cm using a magnetic suspension microbalance. The effective time constants reported were
(4.25 to 6.71) ร 10-6 s-1 for CH at 343 K and (1.44 to 2.56) ร 10-4 s-1 for N at 303 K in the pressure range
4 2
of (0 to 100) kPa. In addition, Hossain [202] measured the effective sorption rate for N using volume
2
69
|
UWA
|
They used the isothermal Fickian diffusion model to extract the โself-diffusivitiesโ of CH and N which
4 2
are effectively the sorption kinetics as we used in this work. They measured the diffusion of pure N
2
and pure CH at three temperatures and obtained sorption rates of 3.0 ร 10-3 s-1 for N and 2.5 ร 10-4
4 2
s-1 for CH at 273 K. They also measured binary N /CH diffusion at the same temperature and obtained
4 2 4
values of (2.34 to 2.61) ร 10-4 s-1 for CH and (2.86 to 3.52) ร 10-3 s-1 for N . These measured sorption
4 2
rates are reasonably consistent with our results considering the adsorbents were from different
manufacturers. No obvious pressure dependence in the effective sorption rate was observed by Mohr
et al. [212] for either CH or N .
4 2
Table 3.6 Effective sorption rate of CH and N on Zeolite 4A obtained from the non-isothermal FD
4 2
model
Zeolite 4A
Methane Nitrogen
273.15 K 273.15 K
P D/r2ร104 u(D/r2) ร104 P D/r2ร103 u(D/r2) ร103
(kPa) โ๐๐(sโ-1๐๐)๐๐ ๐๐๐๐ ๐ฝ๐ฝ (s-1) (s-1) (kPa) โ๐๐(sโ-๐๐1)๐๐ ๐๐๐๐ ๐ฝ๐ฝ (s-1) (s-1)
21.00 0.027 27.77 1.30 0.03 17.49 0.024 0.769 1.15 0.08
42.35 0.024 19.3 6 1.34 0.01 41.82 0.023 0.604 1.15 0.06
59.15 0.028 19.3 9 1.48 0.01 60.71 0.025 0.567 1.12 0.05
82.13 0.027 16.1 4 1.46 0.01 79.82 0.024 0.951 1.14 0.04
99.87 0.028 11.0 7 1.37 0.01 99.03 0.028 0.456 1.11 0.04
30 3.15 K 303.15 K
18.13 0.024 5.692 2.68 0.02 18.92 0.025 0.233 3.16 0.38
43.68 0.024 5.009 3.29 0.02 39.45 0.024 0.298 3.14 0.50
63.30 0.029 5.520 3.01 0.02 60.28 0.026 0.304 3.12 0.47
82.93 0.029 5.221 3.11 0.02 81.27 0.026 0.338 2.98 0.45
102.72 0.028 4.244 3.12 0.03 102.37 0.028 0.385 2.96 0.50
The temperature dependence of the effective sorption rate for the four adsorbents is shown in Figure
3.8 with higher sorption rates observed at higher temperatures. The sorption rate of CH for all the
4
adsorbents showed a strong temperature dependence, as did the sorption rate of N for the three
2
71
|
UWA
|
adsorbents measured at both temperatures (zeolite 13X, zeolite 4A and MSC-3K 172). The activation
energies of CH and N diffusion were estimated using an Arrhenius type equation [211], and were
4 2
found to be (10 โ 20) kJโ
mol-1, which is comparable with the enthalpy of adsorption for these systems.
Figure 3.8 Temperature dependence of effective sorption rate
3.7 Conclusions
Equilibrium capacities and sorption kinetics for pure CH and N were measured for several widely-
4 2
available adsorbents using a commercial volumetric system at pressures from (5 to 100) kPa and
temperatures of (273 and 303) K. Literature values for the effective sorption rate, D/r2 of both gases
on these adsorbents varied in all cases by nearly an order of magnitude or more. Accurate
measurements of the sorption rate in this work were found to require two key elements in the method
and analysis of the dynamic uptake data. First, correction of the dynamic uptake data for the effects
of gas expansion by calibrating the system response with helium was important to the accurate
determination of the initial condition. Second, use of the non-isothermal Fickian diffusion model was
found to be essential for reliable analysis of the ROA data, even for the small sample masses studied
here. For adsorbents with small heat-to-mass transfer ratios (i.e. small values of ฮฑ as defined in Eq
(9)), such as activated carbon Norit RB3 and zeolite 13X, use of the isothermal linear driving force
model results in apparent sorption rate values at least an order of magnitude too small. A small
pressure dependence was observed only for the adsorbents MSC-3K 172 and zeolite 4A, with an
increased (but still small) effective sorption rate observed at higher pressures. The sorption rates of
both CH and N for all four adsorbents showed a clear temperature dependence with Arrhenius-type
4 2
activation energies around (10 - 20) kJโ
mol-1.
For the adsorbents studied here, Norit RB3 and zeolite 13X have such fast and similar kinetics that any
separation of N and CH would rely on their equilibrium selectivity for CH over N . However, while
2 4 4 2
72
|
UWA
|
Chapter 4: Separation of Nitrogen from Methane Using A
Transitional Metal Complex (TMC) Solution
4.1 Forward
In this chapter, in total 14 transition metal complexes (TMCs) have been rigorously screened based on
the absorption performance of nitrogen, and the results have been summarized in Section 4.2. Among
these 14 TMCs, the TMC named K[RuII(EDTA)] in an aqueous solution gave the best performance. And
thus, the main content (section 4.3 to 4.8) of this chapter will be based on this TMC and written in the
format of a journal article which will be submitted after the submission of this thesis. Section 4.4
describes the motivations of developing alternative technologies to separate nitrogen from methane
as introduced in Chapter 2, it can be skipped if Chapter 2 has been read.
4.2 Summary of the Screened TMC Solutions
All the 14 transitional metal complexes that have studied and the nitrogen absorption performance
of their solutions are summarized in Figure 4.1 and Table 4.1 The measurement methods and the
apparatus are the same as the ones used for K[RuII(EDTA)] aqueous solution that are discussed in
Section 4.5 and 4.6. The synthesis methods for TMC 1 to 4 and TMC 6 to 12 are essentially the same
as the one for TMC 2 (K[RuII(EDTA)]) which is listed in Section A. 1.1. The synthesis method for TMC 14
is different and is summarised separately in Section A.2. TMC 4 and TMC 13 are used as received from
the providers.
Based on the transitional metal center, these 14 TMC solutions can be classified into four groups: (1)
Ru(II) based TMCs, (2) Fe(II) based TMCs, (3) V(II) and (III) based TMCs, and (4) Ag(I) based TMCs. The
reasons for exploring Ru(II) based TMCs is that RuII is the first reported transitional metal center that
can absorb nitrogen molecule and have been studied for the separation of nitrogen from methane.[29,
32, 236] Therefore, Ru-based TMCs are intensively studied in this work. Three types of ligands were
chosen to support Ru, naming EDTA, HEDTA and NTA. EDTA (Ethylenediaminetetraacetic acid) is one
of most common ligands which has been used in cosmetics[237], medicine[238, 239], and other
industry[237, 240-242], and thus it is chosen as the first ligand to be studied. HEDTA has a very similar
structure and properties as EDTA, but with one carboxylic group replaced by a hydroxyl group which
leads to a slightly weaker affinity to cations and a higher solubility in aqueous solution. NTA
(Nitrilotriacetic acid) is another type of commercially available ligands, which can support Ru or Fe to
bind gas molecules, such as NO.[243]
74
|
UWA
|
The limitation of such Ru based TMC solution is that its annual global production of Ruthenium is only
around 20 tonnes,[167] which is too small and confines the application of this solution from an
industrial scale. Fe is located within the same chemical element group as Ru and has similar chemical
properties to Ru. In addition, there are plenty of reported studies showing that FeII baed TMCs can
absorb N molecules under various conditions. Therefore Fe-based TMCs have the potential to capture
2
nitrogen from methane. However, most of the reported Fe-based TMCs are not soluble in aqueous
solution,[105, 244-247] and only a few can dissolve in an aqueous solution which involves highly
flammable phosphine ligands.[248] The Fe-based TMCs combined with low cost and safe ligands are
disired and screened in this work. Except for EDTA, HEDTA and NTA, another two supporting ligands
are studied as well: Lactate and Triethyl phosphate. In the molecule of FeIILactate, the FeII is supported
by two lactate anions and this structure has a potential to absorb ฯ donor and ฯ acceptor. However,
the solubility of this compound in aqueous solution is small. Akin to phosphine ligands, Triethyl
phosphate tends to donate an electron to FeII as well. [249-251] More importantly, the easy availability
and relatively less flammability of Triethyl phosphate drive us to study its combination with FeII.
The aqueous solutions of VII or VIII based TMCs were reported having the ability to reduce nirogen to
ammonia[130, 252, 253] and thus are also selected in our study. AgI ion-exchanged zeolite can absorb
nitrogen molecule after auto-reduction at elevated temperature,[254-256] and thus it is worthwhile
to explore AgI centered TMC as well.
From Figure 4.1 it is clear that only three Ru-based TMCs can bind nitrogen under testing conditions
of 30 ยฐC and 0-3000 kPa. RuII[NTA] aqueous solution gives a lightly higher capacity of nitrogen than the
pure solvent of water. RuII[HEDTA] aqueous solution gives around 5 timers high capacity of nitrogen.
The TMC solution with the best performance is K[RuII(EDTA)] aqueous solution which gives 8 times
higher capacity of nitrogen than the pure solvent does, and thus this TMC solution will be intensively
studied in this work. The Fe, V and Ag based TMC solutions do not show enhancement of the nitrogen
capacity, and thus will not be discussed in details in this work.
75
|
UWA
|
P(EtO) : Triethyl phosphate
3
Figure 4.1 Summary of the performance of the TMC solutions studied in this project.
Red: the capacity of nitrogen in pure solvents (pure water or pure TEG); Blue: the increased
capacities of nitrogen in TMC solutions
4.3 Abstract
Nitrogen is an inherent impurity in natural gas. Currently, the principle technology for the separation
of nitrogen from natural gas is a cryogenic distillation, a highly energy-intensive process. A continuous
recirculating absorption process operating at ambient temperature is a highly preferred alternative,
which could be analogous to the stripping process for the removal of carbon dioxide (CO ). The key
2
step to developing such an absorption process for removal of nitrogen is to find a solution that can
selectively and reversibly absorb nitrogen over methane under moderate conditions. Here, we report
an aqueous solution consisting of a โtask-specificโ transition metal complex (TMC) which can meet
such requirements. The TMC, K[RuII (EDTA)], was synthesized in this work and was characterized by
elemental analysis and infrared spectroscopy (IR) to confirm its structure. The absorption equilibrium
capacities of nitrogen in this aqueous TMC solution were measured using a custom-built volumetric
apparatus at temperatures of 20 ยฐC, 30 ยฐC and 40 ยฐC and at pressures ranging from (0 to 4000) kPa.
The results showed that this RuII based TMC aqueous solution is able to selectively capture nitrogen
77
|
UWA
|
over methane with a specific capacity up to 0.5 mole nitrogen per mole of RuII โ half of the
stoichiometric amount. A desorption hyteresis was observed under the measured conditions, which
suggests a combination of pressure swing and temperature swing process is required to recycle the
TMC solution. The calculated absorption enthalpy of nitrogen (30-60 kJ/mol) was moderate compared
to that of carbon dioxide (~90 kJ/mol) in aqueous monoethanolamine (MEA) solutions (30 wt %),
indicating a viable regeneration energy requirement.
4.4 Introduction
Nitrogen is a ubiquitous impurity in natural gas that has no heating value but poses a safety concern
in the transportation and storage of liquefied natural gas (LNG). While the product specification for
LNG is no more than 1 mol % nitrogen [257], the inert nature and the similarity of its physical
properties to methane (CH ) make the separation of nitrogen from natural gas challenging[1].
4
Currently, cryogenic distillation is the most frequently used commercial technology for separating
nitrogen from natural gas in LNG production[257]. This technology utilizes the difference in the boiling
points of nitrogen and methane to separate the binary mixture at cryogenic temperatures, and has a
track record of being economically viable (95-98 % methane recovery) but only for large-scale gas flow
rates (>15 MMscfd) [1, 257]. However, the cryogenic process is energetically parasitic because a
significant fraction of nitrogen gas needs to be cooled to cryogenic temperatures unnecessarily.
Furthermore, the cryogenic Nitrogen Rejection Unit (NRU) is located at the end of the LNG production
process. Consequently, the low-temperature gas processing facilities have to be designed to handle a
greater volume of gas than if the nitrogen could have been removed at ambient temperature.
Unfortunately, no other conventional technology that can effectively separate nitrogen under
moderate temperature on a large scale exists.
Potential alternative technologies for nitrogen separation from natural gas include pressure swing
adsorption (PSA), membranes, and absorption processes. PSA using molecular gate adsorbents has
been demonstrated for nitrogen rejection in small scales. This process has been shown to be effective
for nitrogen rejection (95 โ 98% methane recovery) at low flow rates (2 - 15 MMscfd) [15]. However,
the technology is still in its early stages of commercialization and struggles to accommodate high gas
flow rates (>15 MMscfd). Also, PSA process typically employs multiple beds, requiring complicated
switching controls between beds. Thus, the capital cost and operational cost remains high for large
flow rate systems.[50]
78
|
UWA
|
Membrane technology has attracted much attention because of its ability to separate gas species
without undergoing a phase change, its low energy input and labor intensity, and the relatively simple
process with fewer pieces of moving parts [1, 50]. One successful implementation of this technology
is the hydrogen (H ) purification process using inorganic membranes. This success is mainly attributed
2
to the significant difference in the physical properties of H from that of other gas species, such as
2
carbon monoxide (CO), carbon dioxide (CO ) and nitrogen (N ), which leads to a high H
2 2 2
selectivity.[258] However, the lack of nitrogen-selective membranes has hindered the applicability of
membrane technology fornitrogen separation. Prior studies show that N /CH selectivity of 15 is
2 4
required to make this process economically viable.[50] Currently, the highest reported N /CH
2 4
selectivity is only around 8 from carbon molecular sieve membranes[12, 259] and 2-3 for polymer
membranes.[50] Such N /CH selectivity is too low to make this process economically feasible.
2 4
The absorption process for separation is based on the difference in the solubility of different gases in
the circulating solvent. This process has been widely used in natural gas industry for carbon dioxide
removal with various aqueous solutions of amine. If one could find a sufficiently robust and affordable
nitrogen-selective solvent, it would be possible to deploy an absorption process for the nitrogen
rejection from natural gas. However nitrogen is very stable and inert at room temperature and
materials which are reactive with nitrogen are rare. Fortunately, certain transitional metal complexes
(TMCs) have been reported to be able to bind nitrogen.[29-34, 128, 247, 260-262] Therein, the TMC
functions in a similar way as the metal co-factor in the nitrogenase functions for nitrogen bio-fixation
[24, 25, 109]. By dissolving the TMC in a solvent, the resulting solution will have the potential for
nitrogen rejection. The binding of nitrogen with TMCs under various conditions has been previously
studied, showing that nitrogen can bind to TMCs to a diverse extent depending on the type of the
transition metal center and the type of its supporting ligands. [27, 34, 114, 128, 130, 131, 247, 262-
264] However, most TMCs require an organic solvent, which has an inherently high solubility of
methane, making the solution unsuitable for nitrogen-selective absorption. Therefore, a task-specific
TMC that has the ability to reversibly bind nitrogen under moderate conditions (T=~30 ยฐC, P=0-3000
kPa) over methane in a proper solvent is desired. Here, a โproper solventโ should have (1) a high
solubility for the TMCs achieve higher absorption capacity for nitrogen and low capacity for methane,
(2) a low viscosity for better mass transfer properties, and (3) low volatility to avoid solvent loss. An
absorption process based on such TMC solutions would be analogous to CO capture with aqueous
2
amine solutions, having the potential to remove high-concentration nitrogen from a large-scale
natural gas flow rates (>30 MMscfd) with low capital and operational cost compared to the cryogenic
process. Three groups of researchers have reported the possibility of utilizing this technology to
79
|
UWA
|
remove nitrogen from natural gas [27, 29-32, 139]. However, they did not provide either detailed
solubility data with quantitative uncertainties or the absorption heat for nitrogen on corresponding
TMC solutions, which are both essential for absorption process design.
In this work, a nitrogen-selective TMC has been synthesized and its aqueous solution has been
prepared, in which water served as the solvent. The absorption equilibria of nitrogen in this TMC-
water solution were measured using a custom-designed static solubility apparatus under different
conditions to assess its potential application for the removal of nitrogen from natural gas. To validate
this method, baseline experiments of nitrogen in the water were performed using the proposed
solubility apparatus system, which would also serve as the loading baseline for possible nitrogen
loading in TMC solutions. The absorption equilibria measurements of nitrogen in TMC solution was
conducted over the temperature range of 20, 30 and 40 ยฐC, and the pressure range of 100 kPa to 3000
kPa. From the overall nitrogen loading capacity, the specific absorption capacities of nitrogen were
extracted and compared with the reported data. To estimate the regeneration cost, the enthalpy of
absorption was extracted. A desorption study at 30 ยฐC has also been conducted to characterize the
reversibility of the binding of nitrogen in TMC. Moreover, we present a quantitative method to
estimate the uncertainties of these absorption capacities.
4.5 Materials and Apparatus
4.5.1 Materials Preparation
Anhydrous toluene (99.9 wt%), magnesium metal (chips, 4-30 mesh, 99.98 wt%),
ethylenediaminetetraacetic acid tri-potassium salt dihydrate (K EDTAโ2H O, 98 wt%) and Ruthenium
3 2
(III) Chloride Hydrate (RuCl โxH O) were purchased from Sigma-Aldrich (Australia) and used without
3 2
further treatment. Hydrochloric acid aqueous solution (HCl, 37 vol%) and potassium hydroxide pellets
(KOH, 85 wt%) were purchased from Sigma-Aldrich as well and dissolved in deionized (DI) water. The
pH values of the solutions were measured by an Oakton pH 5+ Handheld Meter with a pH Probe (John
Morris Scientific Pty Ltd). The nitrogen (N ), methane (CH ), and Argon (Ar) gases used in this work
2 4
were purchased from Coregas with the following claimed purities: N 99.999 mol%, CH 99.995 mol%,
2 4
and Ar 99.995 mol%.
The synthesis of K[RuIII(EDTA)(Cl)] โ2H O was carried out according to a method reported in literature.
2
[244, 265] A mass of 4.43 g (~0.01mol) of K EDTAโ2H O was first placed in 20.00 mL of deionized water
3 2
in a 100.00 mL glass beaker, and the resultant mixture was gently heated up on a hot plate until all
solid dissolved to give a clear solution. A ruthenium (III) chloride water solution composed of 2.08 g
80
|
UWA
|
(~0.01 mol) RuCl โx H O and 20.00 mL of DI water was added to the K EDTA solution while swirling the
3 2 3
beaker. The pH value of this solution was adjusted and maintained in the range of 4-5 by adding 0.10
M KOH aqueous solution or 0.10 M HCl aqueous solution. The resulted mixture solution was gently
boiled off at 140-145 ยฐC until no more solid precipitated out. Then, another 40.00 mL DI water was
added to the beaker to dissolve the solid precipitate fully. Following that, the water in the solution
was evaporated again. This dissolving-precipitating process was carried out several times until the
color of the solid precipitate became light yellow. The collected solid product was washed thoroughly
with ice water until it was free of RuIII ions (transparent residual solution without the brown color).
Afterwards, the product was washed with ethanol twice, pre-dried using filter paper, and thoroughly
dried in an oven at 80 ยฐC for 10 hours. A mass of 3.52 g of dry K[RuIII(EDTA)(Cl)] 2H O was obtained
2
with an overall yield of ~70% based on ruthenium mass balance calculation. A Spectrum One FT-IR
Spectrometer (PerkinElmer) equipped with an attenuated total reflection (ATR) sampler and a
deuterated triglycine sulfate (DTGS) detector was used to record the IR spectra of both the raw
K EDTAโ2H O and the synthesized ruthenium complex. The synthesized K[RuIII(EDTA)(Cl)] gave IR
3 2
peaks at 3410 (-OH), 1724 (free -COOH) and 1610 (coordinated โCOO-) cm-1 which were consistent
with literature data [266-268]; while the raw K EDTAโ2H O gave IR peaks at 3410, 1634 and 1595 cm-
3 2
1. An Elementary Vario Macro was used to analyze the total carbon and nitrogen contents. The
theoretical weight percentage (wt %) of carbon (C) and nitrogen (N) for K[RuIII(EDTA)Cl] โ2H O were C
2
23.98 % and N 5.59 %; the measured values were C 24.90 %, N 5.82 %, which matched well with the
theoretical values.
To prepare K[RuII(EDTA)] aqueous solution for N absorption, 2.93 g dry K[RuIII(EDTA)(Cl)] was first
2
dissolved in 58.86 g of degassed DI water; the pH of the solution was adjusted to 7 by adding 0.10 M
KOH solution and 0.10 M HCl solution as requested with a total added mass of 0.95 g. This procedure
was conducted in a glove box filled with argon. Next, 0.42 g of magnesium chips (an excess amount
compared to that required for reducing Ru(III) due to the slow reaction of magnesium and the side
reaction with water) was added to the resultant solution to reduce the RuIII in K[RuIII(EDTA)(Cl)] to RuII
[29] in the glovebox. The solution with the added magnesium was kept in the glove box for 10 hours
before the nitrogen absorption measurements. The overall ruthenium element concentration in the
solution was calculated to be 0.10 M by mass balance. The density of K[RuII(EDTA)] aqueous solution
was obtained independantly from volumetric methods using 5 mL and 10 mL volumetric flasks, which
resulted in an averaged density of 1.02 g/mL.
To transfer the K[RuII(EDTA)] aqueous solution into an absorption cell (described in Section 4.4.2) for
the nitrogen absorption measurements, 10-15 g of the solution was measured to a precision of 4
81
|
UWA
|
decimal places using a digital balance in the glove box. The absorption cell was then sealed with a
stainless-steel lid on which a 1/8 inch VCR gland was sealed with PARAFILM (purchased from Sigma-
Aldrich). The completely sealed absorption cell was transferred out from the glove box and quickly
connected to the absorption apparatus through VCR fittings.
4.5.2 Apparatus
The sorption capacities of pure gasses in solutions were measured with a volumetric method using a
custom-designed apparatus. A schematic diagram of the apparatus is shown in
Figure 4.2. Pure gases were first transferred to and stored in a syringe pump (ISCO 260 D) which has a
maximum volume of 266.05 mL and is capable of delivering gas at constant pressure (up to 51,710
kPa) by volume displacement. Then, a required amount of gas was transferred from the syringe pump
to the accumulation cell (AC). When the temperature and pressure of the gas in the accumulation cell
became steady, the valve between the accumulation cell and solubility cell (SC) was opened to
introduce the gas to the SC and start the sorption measurements. The AC, SC, and their associated
pressure transmitters and platinum resistance thermometers were all located inside a well-insulated
air bath with a temperature uniformity and stability of 0.1 ยฐC. The tapered seal valves for controlling
gas flow to or from the accumulation and solubility cell were also located inside the air bath, with their
long shafts accessible from the outside of the air bath. The solubility cell is equipped with a mini
magnetic stirrer to ensure excellent mixing of the solution and gas species after they were introduced
into the cell. A vacuum pump is attached to the solubility measurement system to evacuate the system
before introducing gas.
82
|
UWA
|
V3 V4
Va Vb
V1
Controller Vc P1 P2
Vd Ve
V2
T T
Gas Gas Vacuum
T T TMC Pump Vent
solution
AC SC
Gas Cylinder ISCO PUMP
Figure 4.2 Schematic of the air-bath static gas solubility apparatus. V: valve; AC: Accumulation cell;
SC: solubility cell. An ISCO pump is used to inject the gas component into the AC to a specific
pressure. An air oven (shown as the blue rectangle) is used to control and maintain the experimental
temperatures. Four high pressure needle valves are used to connect the two cells and the associated
line, which are extended to outside of the oven. A valve ( V ) is used to depressurize the system and
e
a vacuum pump is used to evacuate the system.
The Digiquartzยฎ pressure transmitter for the accumulation cell has a full scale of 13,790 kPa and 0.01%
uncertainty of reading over this range. The four 100 โฆ platinum resistance thermometers on the
accumulation cell and solubility cell were calibrated over the range of 0 ยฐC to 70 ยฐC with an uncertainty
of 0.1 ยฐC according to the International Temperature Scale of 1990 (ITS-90).
4.6 Analysis and Uncertainty
The volumetric method for measuring gas absorption is based on the transfer of a known amount of
gas from the accumulation cell (AC) to the solubility cell (SC) that contains the TMC solution. Because
the two cells and the associated lines constitute a closed system, the mass balance of the gas in the
system can be expressed by Equation 4.1 below.
abs
AG
C
AG
C AC AC
Equation 4.1
๐๐ = ๐๐ โ
G
๐๐ ๏ฟฝ๐๐๐๐( ),๐ด๐ดL๐๐( )๏ฟฝ
G
SC L SC SC SC
SC ๐๐SC SC
+๏ฟฝ๐๐ โ ๏ฟฝ๐๐ ๏ฟฝ๐๐๐๐( ),๐ด๐ด๐๐( )๏ฟฝ
G G๐๐ ๏ฟฝ๐๐๐๐( ),๐ด๐ด๐๐( )๏ฟฝ
AC AC AC AC
โ ๐๐ โ๐๐ ๏ฟฝ๐๐๐๐( L),๐ด๐ด๐๐( )๏ฟฝ
G G
SC L SC SC SC
SC ๐๐SC SC
โ๏ฟฝ๐๐ โ๐๐ ๏ฟฝ๐๐๐๐( ),๐๐๐๐( )๏ฟฝ๏ฟฝ๐๐ ๏ฟฝ๐๐๐๐( ),๐ด๐ด๐๐( )๏ฟฝ
83
|
UWA
|
Subscripts โabsโ, โACโ, โSCโ, โiโ and โfโ refer to absorbed gas, accumulation cell, solubility cell, initial
value and final values, respectively; the superscripts โGโ and โLโ refer to gas phase and liquid phase,
respectively; V is the measured volume; P is the measured pressure; T is the measured temperature;
ฯG is the molar gas density determined using an equation of state at the measured P and T; ฯL is the
density of the solution, mL is the mass of the solvent; and n is the moles of gas absorbed in the
abs
solvent.
From Equation 4.1, it is evident that the accuracy of the measured absorption capacity relies on three
factors: (1) the volumes of the accumulation section (V ) and the solubility section (V ), (2) the
AC SC
volume of the transition metal complex solution, and (3) the density of the gas phase at corresponding
temperature and pressures. The V and V were measured via the displaced volume in the ISCO pump
AC SC
(Figure 4.2) at a constant pressure mode. Specifically, the ISCO pump and the associated line until V
1
was filled with nitrogen at 5000 kPa and operated in constant pressure mode at room temperature. A
vacuum pump was used to evacuate the absorption apparatus, and then V , V and V were closed to
2 3 4
isolated AC from SC and the atmosphere. A certain amount of nitrogen was transferred from the ISCO
pump into the AC by opening V . After few minutes, the equilibrium of gas between the pump and the
1
AC was reached and the decrease of volume in the ISCO pump was recorded as V Ten repeated
AC.
measurements of the volume of AC were performed, and the averaged volume of V was 33.10 ยฑ 0.02
AC
mL. The same method was used to measure V and the averaged volume was 32.59 ยฑ 0.02 mL. The
SC,
density of the TMC solution was measured twice using a 5 mL and a 10 mL volumetric measuring flasks
at room temperature and pressure, and the obtained density was assumed to remain constant
throughout the experiment conditions. The densities of the gas phase in the AC and SC at different
pressures and temperatures were calculated by using equation of state (0.01% uncertainty at 0-77 ยฐC
and < 120 bar for the density of nitrogen; 0.03% uncertainty at < 77 ยฐC and < 120 bar for density of
methane) implemented in REFPROP software (version 9.1).[224, 269] The equilibrium temperatures
were measured by the four 100 โฆ platinum resistance thermometers, and the equilibrium pressures
were measured by the two Digiquartzยฎ pressure transmitters. By substituting these data into Equation
1, we can calculate the amount of gas absorbed by a particular TMC solution under a certain pressure
and temperature.
The uncertainties for each component of the sorption measurements, u(x), were calculated by taking
the individual uncertainties associated with the static solubility apparatus into account, as listed in
Table 4.2 below. We assumed that contributions of chemical and gas impurities to the estimated
uncertainty were negligible.
84
|
UWA
|
The solution density, L , was derived from the solution volume and solution
SC SC SC
weight, which has tak๐ข๐ขe๏ฟฝn๐๐ int๏ฟฝo๐๐ ๐๐a,๐๐c(co)u,n๐ด๐ดt๐๐ ,๐๐t(he) ๏ฟฝu๏ฟฝncertainties of solution mass and solution volume as
measured using the volumetric flask.
4.7 Results and Discussion
4.7.1 Validation Measurements
To validate the performance of the static gas solubility apparatus, control experiments of the
absorption of nitrogen in water at 30 ยฐC were conducted. Additionally, the amount of nitrogen
absorbed in water serves as the baseline for experiments of nitrogen absorptions in aqueous TMC
solutions. The experiment results of absorption of nitrogen in pure water were compared with the
predicted results from MultiFlash (MF) as shown in Figure 4.3 and Table 4.3. The absolute deviations
were below 1 ร 10-3 mol/L within the measured pressure range from 0 to 3000 kPa, which was almost
one order of magnitude smaller than the nitrogen absorption amount in water under the experimental
conditions (30 ยฐC, 500-3000 kPa). This indicates that the apparatus is reliable to provide accurate
solubility data, even when the TMC solutions can only absorb nitrogen at a level of 1ร10-3 mol/L.
Figure 4.3 Isotherms of nitrogen loading in water at 30 ยฐC. Absorption (๏) and desorption (๏) data
were measured based on a cumulative method. The prediction data (dash line) were obtained from
CPA-Infochem model implemented within MultiFlash (MF).[270]
86
|
UWA
|
Table 4.3 Loading capacities of nitrogen in water with the associated uncertainties and their deviations
from MF. MF: multiflash.
Capacity
Capacity
(MF) Deviation
T P Uncertainty
from MF
N N
2 2
10-Mar 10-Mar 10-Mar 10-Mar
ยฐC kPa mol N / mol N / mol N /L mol N /L
2 2 2 2
L water L solution
30.6 407 2.44 1.75 0.58 -0.69
30.6 925 5.53 4.53 1.36 -1.00
Absorption 30.5 1431 8.52 7.77 2.33 -0.75
30.6 1967 11.60 10.71 3.5 -0.89
30.6 2829 16.60 16.71 5.13 0.11
30.6 1969 11.64 11.82 5.91 0.18
30.5 1518 9.02 9.22 6.3 0.19
30.6 1018 6.08 5.79 6.49 -0.30
Desorption
30.6 762 4.56 4.01 6.59 -0.55
30.6 554 3.32 2.85 6.64 -0.47
30.6 298 1.78 0.81 6.66 -0.97
4.7.2 Nitrogen Absorption in K[RuII (EDTA)] Aqueous Solution
4.7.2.1 Overall Absorption Isotherms
The binding of a nitrogen molecule to the vacant site of a TMC can be explained by a ฯ-donor/ฯ-
acceptor coordination model. It involves a ฯ-donation of electrons from the nitrogen non-bonding
electron pair to the metal vacant dz2 or dx2-y2 orbitals, and a back-ฯ-donation of electrons from the
filled metal dxz, dyz, or dxy orbitals to the vacant ฯ* orbitals of nitrogen. However, the nitrogen
molecule is both a poor ฯ-donor and a poor ฯ-acceptor because of the stable nitrogen lone pair and
the large highest occupied molecular orbital-lowest unoccupied molecular orbital (HOMO-LUMO) gap
which reduce the overlap between the metal and nitrogen orbitals.[129, 131] Nevertheless, a large
number of TMCs have been synthesized with the ability to bind nitrogen, although unfortunately most
88
|
UWA
|
of them require organic solvents.[128] Ruthenium is the first transition metal that has been shown to
bind nitrogen in aqueous solutions,[236] which have been reported to bind nitrogen reversibly by
tuning the nitrogen pressure.[29] A schematic of nitrogen binding to Ru(II) supported by the ligand of
EDTA is shown in Figure 4.5 below.
Figure 4.5 Schematic of nitrogen binding to [Ru(EDTA)]-. The shaded balloon stands for the vacant
site of [Ru(EDTA)]- which accepts nitrogen at high pressure; while at low pressure, nitrogen escapes
from [Ru(EDTA)]- and makes this nitrogen-binding process reversible.
In this work, the nitrogen loading capacities in ruthenium-based aqueous solution, K[RuII (EDTA)] water
solution, were measured at 20 ยฐC, 30 ยฐC and 40 ยฐC within the pressure range of 500 kPa to 4000 kPa.
All experiments used the molar concentration of K[RuII (EDTA)] in the aqueous solution of 0.10 M. The
absorption results under three temperatures are summarized in Figure 4.6, together with the baseline
of nitrogen absorption capacities in pure water at corresponding temperatures. The uncertainty bars
shown in the graph are calculated uncertainties using the error propagation method by taking into
account every factor listed in Table 4.2 that might have affected the solubility measurement.
Therefore, the final uncertainty is a conservative estimate. These results show that the K[RuII (EDTA)]
water solutions have much higher absorption capacities than that of the solvent (water) at the same
conditions, which confirms the reported N -Ru chemisorption under elevated pressures [29, 32, 271,
2
272].
At 40 ยฐC, the isotherm is almost linear and the nitrogen loading capacity reaches 0.04 mol N /L solution
2
at 3000 kPa. For the isotherm at 30 ยฐC, a sharp uptake appears at around 1000 kPa, and the nitrogen
loading capacity reaches 0.05 mol /L at 3000 kPa. At 20 ยฐC, the isotherm gives a sharper uptake at even
lower pressure (around 500 kPa) and eventually reaches 0.06 mol /L at 3000 kPa. Moreover, it is
evident from Figure 4.6 that the nitrogen loading capacities increased with the decrease of solution
temperature, indicating that the TMC-N binding reaction is an exothermic reaction,[29] and that low
2
temperatures are preferred to achieve high nitrogen loading capacities.
89
|
UWA
|
Figure 4.6 The loading capacities of nitrogen in pure water and K[RuII (EDTA)] aqueous solutions at
the pressure range of 0 to 3000 kPa and at the temperatures of 20, 30 and 40 ยฐC. The capacities of
nitrogen in pure water are shown with hollowed markers; the capacities of nitrogen in K[RuII (EDTA)]
aqueous solutions are shown with solid markers.
4.7.2.2 Specific Absorption Isotherms
The overall nitrogen loading capacity in the TMC solution depends on both the TMC concentration
and the specific nitrogen absorption capacity. At a specific TMC concentration, increasing the specific
nitrogen absorption capacity would lead to a higher overall nitrogen loading capacity.
Stoichiometrically, each K[RuII(EDTA)] molecule has one vacant site accepting incoming nitrogen and
thus can only absorb one nitrogen molecule when saturated. The nitrogen specific absorption capacity
measurements under different temperatures and pressures are critical to evaluate the overall
nitrogen loading performance of a TMC solution. The nitrogen specific capacities per RuII with
associated uncertainties at 20 ยฐC, 30 ยฐC and 40 ยฐC are summarized in Figure 4.7 together with two sets
of reported data at 21 ยฐC and 41 ยฐC [29]. It is apparent that the measured results are comparable to
and aligned reasonably well with the literature results at 20 ยฐC, but are slightly different at 40 ยฐC [273].
The differences in nitrogen absorption capacity between our measurements and the literature data
at 40ยฐC are negligible in the low-pressure region and increase with pressure. This could be attributed
to an experimental error in the estimated solution volume or density that could cause slope difference
in the isotherms.
90
|
UWA
|
Figure 4.7 Comparison of the specific capacities of nitrogen on RuII solution in this work and in the
literature for a pressure range from 0 kPa to 3000 kPa. Three temperatures (20, 30 and 40 ยฐC) were
studied in this work; while only two temperatures (20 and 40 ยฐC) were reported by Friesen, et al.,
(1993). The results in this work are shown with the solid markers; the results reported in literatures
were shown with hollow markers.
Although the prepared metal complex solutions in this work has lower concentrations (0.10M) than
that stated in the literature (0.25M), it is clear that the specific nitrogen loading capacity per RuII was
not affected. The measured data shows that the solution absorbs about 0.30 moles of nitrogen per
mole of the Ru compound at 30 ยฐC and 3000 kPa and about 0.45 moles of nitrogen per mole of the
compound at 20 ยฐC at a similar pressure. The N specific capacity in this TMC solution at 20 ยฐC is similar
2
to that of CO capture by monoethanolamine (MEA) solutions, in which CO to MEA molar ratio in the
2 2
CO -rich amine is about 0.3-0.5 [274-277]. Because the N specific capacity in this TMC solution can
2 2
reach a level comparable to the CO specific capacity in MEA solution, which has been well-established
2
on an industrial scale to separate CO from natural gas, the N specific capacity would not be a
2 2
potential limiting factor of this K[RuII(EDTA)] aqueous solution to be scaled up for nitrogen removal
from natural gas. However, the low overall capacity would require high TMC solution flowrate when
using an absorption process to remove the same amount of nitrogen from natural gas. The low overall
capacity is mainly due to the low concentration of this bulk TMC which lacks polar functional groups.
The carboxyl groups of EDTA ligands are polar groups, but are still not able to provide K[RuII(EDTA)]
with high enough solubility in the aqueous solution. This EDTA-based TMC concentration (0.1mol/L
for our work, 0.25mol/L for the literature) is one order of magnitude smaller than the typical
concentration of MEA aqueous solution that has been used in the natural gas sweetening plants
91
|
UWA
|
(ranging from 3-7mol/L[278]), which is a limitation of this TMC system for industrial absorption
applications.
Until now, the only reported absorption process that selectively captures nitrogen from natural gas
by a physical solvent uses liquid ammonia.[102] The Henry selectivity of nitrogen over CH in liquid
4
ammonia is around 0.25 and the nitrogen loading at 3400 kPa is around 0.1 mol/L.[101, 103, 104] In
this study, the chemical solvent of K[RuII(EDTA)] aqueous solution gives a nitrogen over methane
selectivity of around 2 at 3000 kPa with a nitrogen loading capacity of 0.06 mol/L, shown in Figure 4.8
and Table 4.4. Compared to the liquid ammonia, the K[RuII(EDTA)] aqueous solution gives a moderate
nitrogen loading capacity but with much higher nitrogen over methane selectivity. In addition, the
K[RuII(EDTA)] aqueous solution does not involve severe handling concerns which complicate the liquid
ammonia process. The K[RuII(EDTA)] aqueous solution which chemically and selectively absorbs
nitrogen, would seem to have a greater potential than the ammonia-based physical solvent to
separate nitrogen from methane in terms selectivity and loading capacities of nitrogen up to 3000
kPa.
Figure 4.8 The comparison of the capacity of nitrogen in K[RuII(EDTA)] aqueous solution and the
capacities of N and CH in pure water at the pressure range from 0 to 3000 kPa
2 4
and at the temperature of 30 ยฐC.
92
|
UWA
|
this work and also in the literature [29] are calculated using the following Gibbs-Helmholtz equation
as shown in Equation 4.3, and the results are summarized in Figure 4.9.[229, 280]
Equation 4.3
โ๐๐๐๐๐๐๐ถ๐ถ ๐๐ln๐๐
wh๐
๐
ere= R โis ๏ฟฝth๐๐(e1 โg๐๐a)s๏ฟฝ constant, R = 8.314 J /mol K
Figure 4.9 The absorption enthalpies of the nitrogen on RuII Calculated with Equation 4.3. The solid
circle stands for the absorption enthalpy of nitrogen on RuII calculated from the absorption
capacities measured in this work at three temperatures (20, 30 and 40 ยฐC); the hollow circle stands
for the absorption enthalpy of nitrogen on RuII calculated from the absorption capacities reported in
literature at two temperatures (20 and 41 ยฐC).[29, 32]
In the literature, the enthalpy of absorption was not reported directly; the results shown in this work
were calculated from the N specific capacity on RuII at 21 ยฐC and 41 ยฐC. The absorption enthalpy from
2
this work and the calculated absorption enthalpy from literature matched with each other when the
specific absorption capacity was higher than 0.3mol N /mol RuII. At lower absorption capacities in the
2
region (< 0.3 mol N /mol RuII), the enthalpy calculated from the literature data was 7-15% higher than
2
the one from this work. Both of the measured data in this work and the literature data show that the
absorption enthalpy decreases with the increase of N specific absorption capacity, which follows a
2
similar trend of CO absorption enthalpy in amine solutions [281]. However, at a specific N absorption
2 2
capacity of 0.2 mol N /mol RuII, a small peak appears with the highest absorption enthalpy of -67
2
kJ/mol N for results calculated from the literature and -57 kJ/mol N for this work. At a specific
2 2
94
|
UWA
|
capacity of 0.55 mol N /mol RuII (in this work, at 3000 kPa and 20 ยฐC), the absorption enthalpy was
2
estimated to be about -30 kJ/mol for both literature and this work. This absorption enthalpy is
considered to be moderate when compared with the CO scrubbing by aqueous MEA (30 wt %)
2
solution (around -90kJ/mol) [275, 282]. Therefore, it would require less energy to regenerate the N
2
rich solution than that for CO desorption.
2
Table 4.5 The absorption enthalpies of the nitrogen on RuII.[29, 32]
ฮH kJ/mol
Specific capacity
Measured in this work Calculated from literature
0.10 52.78 64.56
0.15 57.12 67.41
0.20 57.45 66.55
0.30 52.16 59.12
0.40 42.66 47.58
0.50 30.82 33.73
4.7.2.4 Reversibility of the nitrogen sorption in K[RuII(EDTA)] aqueous solution
To evaluate the reversibility of nitrogen sorption in this K[RuII(EDTA)] aqueous solution, desorption
tests were conducted at the moderate temperature of 30 ยฐC over a pressure range of 0 to 3000 kPa.
Figure 4.10 shows that the desorption capacities are higher than the absorption ones under the
measured experimental conditions, giving a pronounced hysteresis loop which might indicate that the
absorption of N on RuII is not fully reversible. When the pressure decreased from 2075 kPa to 812
2
kPa, the specific capacity of nitrogen remains essentially the same, around 0.38 mol N /mol of RuII.
2
Further decreasing the pressure to 112 kPa, the specific capacity of nitrogen reduces to 0.27 mol
N /mol RuII, which indicates that the coordination bond between nitrogen and RuII is partially
2
reversible under the experimental temperature. Similar hysteresis has been observed on other
chemisorption systems as well, for example, the chemisorption of CO on Li SiO or amine-grafted
2 8 6
zeolite 13 X,[283, 284], and the chemisorption of H on carbon nanotubes.[285] The nitrogen specific
2
95
|
UWA
|
Table 4.6 The absorption/desorption of nitrogen on RuII with an uncertainty of 0.01 mol N /mol RuII.
2
At a pressure range from 0 to 3000 kPa and at temperatures of 30 ยฐC and 60 ยฐC.
Temperature Pressure Specific Capacity of N
2
ยฐC kPa mol N /mol RuII
2
29.5 446 0.07
29.6 937 0.23
Absorption 29.6 1461 0.29
29.6 1950 0.36
29.5 2873 0.43
29.5 2075 0.39
29.5 1555 0.38
29.4 1053 0.38
29.4 812 0.37
Desorption
29.5 592 0.34
30.1 336 0.29
29.9 112 0.27
59.0 143 0.11
4.8 Conclusion
A nitrogen-selective transition metal complex, K[RuIII(EDTA)(Cl)] โ2H O, has been successfully
2
synthesized in this work. The equilibrium capacities and their associated uncertainties of pure nitrogen
absorption on this RuII-based TMC aqueous solution were measured experimentally at three different
temperatures using a custom-built solubility apparatus. The apparatus has a capacity ranging from 1
x 10-3 to 1 mol/L and working conditions ranging from 20ยฐC to 40ยฐC and from 0 kPa to 5,000 kPa. The
TMC aqueous solution showed the ability to selectively absorb nitrogenover methane, with an overall
N capacity of 6.11 x 10-2 mol/L and a N /CH selectivity of 2 at ~30ยฐC and ~3000 kPa when the
2 2 4
concentration of RuII is 0.1M. Our work shows that the specific equilibrium capacities of N per RuII
2
remain essentially constant at different TMC concentrations of 0.1 M and 0.25M, which means that
both the overall N capacity and N /CH selectivity in this RuII-based TMC aqueous solution are
2 2 4
proportional to the TMC concentration. The calculated absorption enthalpy (30-60 kJ/mol N ) was
2
relatively moderate compared to CO absorption enthalpies in aqueous MEA (30 wt %) solution
2
(around ~90 kJ/mol CO ), meaning that this TMC aqueous solution would require less regeneration
2
energy for N scrubbing than what is required for the MEA aqueous solution used in CO scrubbing.
2 2
97
|
UWA
|
Chapter 5: Separation of Nitrogen from Natural Gas by Lithium
Metal
5.1 Introduction
Lithium is the only metal and one of the few elements that can react with nitrogen directly, slowly at
ambient temperatures and violently at elevated temperatures(>200 ยฐC).[42, 46]Lithium reactions with
different gas species, including nitrogen, have been studied for decades. The primary reason has been
the safety hazards caused by lithium spill in fusion reactors where lithium functions as a tritium
breeder blanket and as a coolant [174-177]. Recently, lithium has also been proposed as energy-
carrier to store renewable energies.[178-180] In such a process, lithium reacts with carbon dioxide
and nitrogen from the flue gas of power plants to generate electricity. The resulting Li N and Li CO
3 2 3
would go through a series of treatments to be converted to LiCl, and eventually to be electrolyzed
back to lithium metal using seasonal renewable energy (solar energy or wind energy). In the 1990s,
the application of lithium metal for gas separation was reported through a patent, where it was used
to separate trace amounts of nitrogen and oxygen impurities from crude argon.[43] Due to the high
reactivity of lithium with nitrogenat temperatures higher than its melting point (180.6 ยฐC), it was
claimed that the concentration of nitrogen in the Ar stream could be reduced to less than 1ppm.
Inspired by the Ar purification process, the use of lithium metal to separate nitrogen from natural gas
is studied in this chapter. The separation process could be operated in a batch mode analogous to an
adsorption process: lithium captures nitrogen from the natural gas stream to produce a pure methane
stream; then, the resulted product of lithium nitride is regenerated back to lithium metal by various
methods.[43, 46, 192, 286] Compared to conventional adsorbents that have been studied for the
separation of nitrogen from natural gas, lithium metal shows significant advantages: (1) the
theoretical uptake of for nitrogen on lithium is 24 mmol N /g Li (calculated from the reaction equation
2
shown in Equation 5.1), which is an order of magnitude higher than the best reported nitrogen-
selective adsorbents.[1] This uptake is about twice of the update of H O on molecular sieves, which
2
have been successfully employed in a liquefied natural gas (LNG) production plants.[287] This high
separation performance suggests that a lithium-based process could be adapted to natural gas
processing plants. (2) Moreover, lithium does not react with methane which implies that lithium could
have a very large selectivity for nitrogen over methaneallowing full methane recovery in principle.
However, the reactions of lithium with nitrogen mentioned above are all at temperatures higher than
the melting point of lithium metal (180.6 ยฐC). Using melted lithium poses safety concerns in natural
99
|
UWA
|
gas processing plants, and it would be desirable to operate the reaction of lithium with nitrogen at a
temperature below its melting point.
Equation 5.1
6๐ฟ๐ฟ๐๐+๐๐2 = 2๐ฟ๐ฟ๐๐3๐๐
In the presence of moisture, lithium can react with nitrogen smoothly under ambient temperature
[46, 181, 182]. However, the involvement of water could cause two potential concerns: (1) the
reaction of lithium with water could become violent and uncontrollable when the water concentration
in the stream is higher than a certain level (1 mol%)[186]; (2) lithium metal would be wasted if it reacts
with water rather than with the target nitrogen.
When water is absent, the reaction of lithium with nitrogen is poorly understood. Some studies
showed that in the absence of water moisture, lithium remains stable in dry nitrogen or even dry air
for days at room temperature.[41, 184] Other studies have claimed that lithium can react with dry
nitrogen under moderate temperatures.[42, 185, 186] Here, dry nitrogen usually means nitrogen that
contains less than 10 ppm water content, but usually the nitrogen used contained <2 ppm water due
to further dehydration by various methods.[182] It is interesting to note that reports of lithium metal
not reacting with dry nitrogen under moderate temperatures have been the ones which were stored
either in hydrocarbon solvents[41], or in Ar atmosphere (like the ones in this study). In contrast, when
lithium metal that is freshly made in-situ, either by recrystallization from molten lithium,[42] vapor
lithium,[185] or by electrodeposition,[186] it was reported to be able to react with dry nitrogen. The
main difference between the freshly made lithium metal and the lithium that had been exposed to
other compounds is that there are likely more active edge sites on the former rather than on the latter.
Such active edge sites are expected to be rapidly consumed by materials surrounding them, and thus
the lithium metal with an exposure/storage history has lost its activity toward dry nitrogen. Because
the in-situ production of lithium involves high cost, it would be desirable to explore a simple and
economical method to activate contaminated lithium metal so that it can react with dry nitrogen.
In this work, an activation method using moisture to activate the lithium metal was developed. Various
techniques were used to investigate this method, and then a mechanism for the reaction of lithium
with dry nitrogen was proposed. The uptakes of nitrogen and methane on different lithium samples
were measured individually through pure fluid measurements and the reaction of enthalpies were
extracted as well. A flow-based breakthrough study using a binary gas mixture of nitrogen and
methane on lithium samples was conducted to assess the potential of lithium for separating of
methane and nitrogen. Thermal regeneration of lithium metal from lithium nitride was considered
100
|
UWA
|
further by both thermodynamic calculation and experimental exploration. For the chemical loop
process, as each step of it is already well-established, only an economic estimation was conducted.
Finally, to achieve a cyclic process, the lithium metal would need to be regenerated from lithium
nitride if the whole process is to be cyclic. Various methods can potentially achieve the regeneration
of lithium metal from lithium nitride: thermal decomposition of lithium nitride; chemical looping;
direct electrolysis of lithium nitride. In principle, lithium can be recycled from the thermal
decomposition of lithium nitride by increasing the temperature and by reducing the partial pressure
of nitrogen. Several studies have shown that under certain conditions lithium metal can be
regenerated from lithium nitride.[192, 288-290] The most recent study has experimentally produced
lithium metal from lithium nitride at 382 ยฐC with a partial pressure of nitrogen at 10-3 Pa. A rigorous
study needs to be conducted to explore this approach to regenerate lithium metal. A chemical loop
involving H O, CO and HCl for recycling lithium from lithium nitride has been proposed.[46] However,
2 2
there is no reported study estimating its economic feasibility in the application of separating nitrogen
from natural gas. Electrolysis of lithium nitride to recycle lithium has been proposed due to the low
electrical decomposition potential of lithium nitride.[43] However, no experimental exploration of this
approach has been reported so far, and thus it was not included in this study.
5.2 Experimental Section
5.2.1 Materials and Preparation of Samples
5.2.1.1 Materials
In the experiments conducted with a scanning electron microscope & energy-dispersive X-ray
spectroscopy (SEM-EDS), thermogravimetric analyzer (TGA) and the ASAP2020, the lithium metal was
purchased from Sigma-Aldrich with a product number: 444456. The lithium was granular in shape with
a size of 4-10 mesh. The purity of the lithium metal was 99 % trace metal basis. Each granule weighted
5-10 mg. The lithium samples which were used as received without any treatment were named as
non-activated lithium samples; the lithium samples which were subsequently activated with water
moisture were named as the pre-activated lithium samples.
In the experiments conducted with synchrotron X-ray diffraction (XRD) and binary gas mixture
breakthrough, the lithium metal was also purchased from Sigma-Aldrich but with a different product
number: 266000. The lithium was in a ribbon shape with a thickness of 1.5 mm. The purity of the
lithium is 99.9 % trace metal basis. The reason to choose the lithium with a ribbon shape was that the
101
|
UWA
|
ribbon was easier to cut into the desired shapes needed for the apparatus compared to the lithium
with granular shape. For example, in the mixture gas breakthrough experiment, the lithium samples
needed to be cut into a shape with a thickness of 1.5 mm, a width of 2-3 mm and a length of 5 mm.
In the synchrotron XRD experiments, the lithium metal needed to be scraped into tiny slices with a
thickness less than 1mm, a width less than 1mm and a length of ~2mm. Except where specifically
mentioned, all the operations that involved lithium metal were conducted in an Ar-filled glove box.
In the experiments conducted with the TGA, ASAP2020 and the binary gas mixture breakthrough
apparatus, nitrogen, methane, helium and argon were purchased from Coregas Pty Ltd with a purity
of 99.999 vol. %. The water moisture concentrations in these gases were all less than 2ppm. For the
synchrotron XRD experiments, the nitrogen was purchased from BOC with a purity of 99.99% in which
the water concentration was less than 10 ppm. The reason for choosing nitrogen with a higher water
content was that the size of the lithium samples used in the study of synchrotron XRD was tiny, with
a mass of only 1e-3 mg. Thus it was not possible to activate the lithium samples with a separate
pretreatment in water moisture. The nitrogen with a purity of 99.99% provided more moisture needed
to activate the lithium samples in-situ.
5.2.1.2 Preparation of the Lithium Samples
The pre-activated lithium samples were prepared as shown in Figure 5.1. Two round bottom flasks,
which were connected by a plastic tube (ID = 6.35 mm) with an isolation valve in between, were put
into a glovebox which had been filled with Ar (Coregas, 99.999 vol. %). One of the flasks was filled with
deionized water (DI) water, and the other was loaded with the non-activated lithium samples as
received. When a black coating was observed on the surface of lithium metal, the lithium samples
were ready for the nitridation reaction. The empirical time found necessary for the activation of
lithium particles was around 8-12 hours.
102
|
UWA
|
Figure 5.1 Schematic of the activation of lithium samples by water moisture. The two flasks were
located in an Ar-filled glove box. An isolation valve was used to stop moisture flowing after the
lithium samples were activated. The left flask was filled with DI water (light blue color); the righ flask
was filled with as received lithium samples (light grey color).
In the synchrotron XRD experiments, quartz capillaries with an outside diameter (OD) of 1mm were
purchased from the Charlessupper Company. The lithium metal was first scraped into tiny slices and
then these slices were loaded into quartz capillaries, which were sealed with wax. All the operations
were conducted in a glove-box filled with Ar which contained less than 2 ppm of water moisture and
less than 4 ppm of nitrogen. The capillaries were fixed to the beamline apparatus and connected to a
nitrogen stream. In the flow-through cell experiments, the ends of the capillaries were cut off, so
nitrogen could keep flowing through; in the static cell experiments, no further action was required.
5.2.2 Sample Characterisation
The surface morphology and surface chemical elemental compositions of the lithium metal and
lithium nitride samples were studied with an SEM-EDS (FEI Verios XHR SEM). The in-situ identification
of the transformation of crystalline phases from lithium to lithium nitride was measured by
synchrotron powder X-ray diffraction (XRD) (powder diffraction beamline Australia Synchrotron). The
surface area of the samples of lithium and lithium nitride were measured by ASAP2020, using a BET
analysis at 77 K.
5.2.3 The Reaction of Lithium with Dry Nitrogen
The reactions of different lithium samples with dry nitrogen containing less than 2 ppm water were
examined using a simultaneous thermogravimetric analyzer and differential scanning calorimeter
(Model: SDT Q600). This Q600 can simultaneously measure the weight change and the true differential
heat flow of the same sample with an operating temperature range from ambient to 1500 ยฐC. The
103
|
UWA
|
integrated mass flow controller with automatic gas switching enables the selection of two different
streams with certain flow rates. In this study, the two gas streams were an Ar stream and a nitrogen
or methane stream. Alumina sample crucibles with a capacity of 90 ฮผL which is intended for operation
from ambient to 1500ยฐC were chosen for the experiments of nitrogen/methane uptake. The uptake
of nitrogen and methane can be calculated from the weight change of the samples. The enthalpies of
reaction can be directly extracted from the measured heat flow.
Equilibrium sorption isotherms of nitrogen and methane on lithium samples were measured using a
volumetric method by a commercial Micromeritics ASAP 2020 as well to confirm the uptakes
measured by the TGA. This ASAP2020 was also used to measure the surface areas of lithium samples
and produced lithium nitride samples.
The regeneration of lithium from lithium nitride was examined with a thermogravimetric analyzer Q50
(TGA Q50). Alumina sample crucibles with a capacity of 90 ฮผL, which is intended for operation at
temperatures from ambient to 1500ยฐC were chosen. A platinum crucible with a capacity of 100 ฮผL
which is intended for operation from ambient to 1000 ยฐC was also used to conduct the lithium
regeneration experiments.
5.2.4 Binary Gas Mixture Breakthrough
A custom-designed flow-through apparatus was built to measure lithium`s ability to separate nitrogen
from a binary gas mixture of nitrogen and methane, shown in Figure 5.2. The flow-through apparatus
consists of a stainless steel column (130 mm long and 22.2 mm internal diameter) that can be packed
with up to ~4 g of lithium samples. This column and its associated lines were located in an oven with
a controllable temperature range from 30 to 400 ยฐC. The feed gas was supplied through three mass
flow controllers (MFCs) where a manual valve controls whether helium or a combination of methane
and nitrogen flowed to the column. A pre-heating coil located before the lithium column inside the air
oven was used to heat the gas mixture to the desired temperature. The effluent gas flow rate was
measured by a mass flow meter (MFM), which was calibrated according to a method reported by
Hofman et al..[291] The pressure of the system was adjusted by a needle valve located next to the
MFC. The composition of effluent gas was measured by a universal gas analyzer (UGA) from Stanford
Research Systems, which was calibrated in advance with a method described by Hofman et al..[291]
A water sensor and a drier filled with silica gel were installed to prevent any high concentrations of
water from being carried into the UGA.
104
|
UWA
|
5.2.4.1 Experimental Procedure
The gas mixture breakthrough experiments were performed with similar method as described
previously albeit using a differently sized column.[291, 292] In a typical run, the column was packed
with samples of activated lithium metal inside an Ar-filled glovebox before being isolated under
atmospheric pressure of Ar. After the column was connected back to the flow through apparatus, the
flow of He was set to 30 sccm, allowing the system to be purged through the bypass line. After the air
in the system was fully purged, the He flow was switched to the column to purge Ar out of the system.
Then the binary mixture of nitrogen and methane with the desired composition was pre-mixed and
sent to purge the system through the bypass line. At the same time the He flow was stopped by the
manual valve. When the feed composition recorded by UGA became stable, the binary gas mixture
was switched to go through the column. The UGA and the MFM were programmed to record the
effluent compositions and the effluent flow rate respectively. When the effluent gas composition and
the flow rate reached the original values of the feed gas, the reaction of lithium with nitrogen was
considered to be finished.
5.2.4.2 The Determination of Uptake
The uptakes of nitrogen and methane on lithium metal were determined by an analysis of mass
balance across the breakthrough column :
Equation 5.2
๐๐๐๐,๐๐๐๐๐๐๐๐๐๐๐๐๐๐๐๐๐๐๐๐๐๐๐๐ = ๐๐๐๐,๐๐๐๐๐๐๐๐๐๐ ๐๐๐๐ โ๐๐๐๐,๐๐๐๐๐๐๐๐๐๐ ๐๐๐๐๐๐
Equation 5.3
๐๐
๐๐๐๐,๐๐๐๐๐๐๐๐ ๐๐๐๐ = ๏ฟฝ ๐น๐น0๐ฆ๐ฆ๐๐,๐๐0๐๐๐ก๐ก
๐๐0
Equation 5.4
๐๐
๐๐๐๐,๐๐๐๐๐๐๐๐ ๐๐๐๐๐๐ = ๏ฟฝ ๐น๐น๐ฆ๐ฆ๐๐,๐๐๐๐๐ก๐ก
๐๐0
106
|
UWA
|
: the starting point of the experiment when the binary gas mixture was
๐ก๐กs0witched to the column
: the flowrate of the binary gas mixture (meaured by MFM)
๐น๐น0
: the composition of componet in feed gas (measured by UGA)
๐ฆ๐ฆ๐๐,๐๐0 ๐๐
: the total flowrate of effluent
๐น๐น : the composition of component at the time of
๐ฆ๐ฆ๐๐,๐๐ ๐๐ ๐ก๐ก.
The accumulation of component consists of two parts: the accumulation of component in the gas
phase in the system and the uptake of component by lithium metal, as shown in Equation 5.5.
๐๐ ๐๐
๐๐
Equation 5.5
๐๐๐๐๐๐ ๐๐๐๐๐๐๐๐๐ข๐ข๐๐
๐๐๐๐,๐๐๐๐๐๐๐๐๐๐๐๐๐๐๐๐๐๐๐๐๐๐๐๐ = โ๐๐๐๐ + โ๐๐๐๐
The uptake of component on lithium metal can be calculated by combining Equation 5.2 to Equation
5.5, with the results shown in Equation 5.6.
๐๐
Equation 5.6
๐๐ ๐๐
๐๐๐๐๐๐๐๐๐ข๐ข๐๐ ๐๐๐๐๐๐
โ๐๐๐๐ = ๏ฟฝ ๐น๐น0๐ฆ๐ฆ๐๐,๐๐0๐๐๐ก๐กโ ๏ฟฝ ๐น๐น๐ฆ๐ฆ๐๐,๐๐๐๐๐ก๐กโ โ๐๐๐๐
๐๐0 ๐๐0
The accumulation of component in the gas phase in the system can be determined from the void
volume of the system. As helium doesnโt react with lithium, the displaced volume of helium equals
๐๐
the void volume of the system, as shown in Equation 5.7 and Equation 5.8.
Equation 5.7
๐๐
โ๐๐๐ถ๐ถ๐๐ = ๏ฟฝ ๐น๐น๐ฆ๐ฆ๐ถ๐ถ๐๐,๐๐๐๐๐ก๐ก
๐๐0
Equation 5.8
โ๐๐๐ถ๐ถ๐๐๐
๐
๐ด๐ด๐๐๐๐๐๐๐๐๐๐๐๐๐๐
๐๐๐ฃ๐ฃ๐๐๐๐๐๐ =
๐๐๐๐๐๐๐๐๐๐๐๐๐๐๐๐
: the gas constant
๐
๐
: the He mole fraction in the effluent at time
๐ฆ๐ฆ๐ถ๐ถ๐๐,๐๐ ๐ก๐ก
The needle valve used in the apparatus could not maintain the system pressure precisely, and
consequently the initial pressure and the final pressure in the system would vary. Thus, the total moles
107
|
UWA
|
of the gas mixture and the accumulation of gas component in the void space of the system at the
finial stage were calculated using Equation 5.9 and Equation 5.10, respectively.
๐๐
Equation 5.9
(๐๐๐๐๐๐) ๐๐๐๐๐๐๐๐๐๐๐๐๐๐๐ฃ๐ฃ๐๐๐๐๐๐ (๐๐๐๐๐๐) ๐๐๐๐๐๐๐๐๐๐๐๐ ๐ด๐ด๐๐๐๐๐๐๐๐๐๐๐๐๐๐
๐๐๐๐๐๐๐๐๐๐๐๐,๐๐๐๐๐๐๐๐๐๐ = = โ๐๐๐ถ๐ถ๐๐
๐
๐
๐ด๐ด๐๐๐๐๐๐๐๐๐๐ ๐๐๐๐๐๐๐๐๐๐๐๐๐๐๐๐ ๐ด๐ด๐๐๐๐๐๐๐๐๐๐
(๐๐๐๐๐๐) (๐๐๐๐๐๐) ๐๐๐๐๐๐๐๐๐๐๐๐๐๐๐ฃ๐ฃ๐๐๐๐๐๐
Equation 5.10
๐๐๐๐,๐๐๐๐๐๐๐๐๐๐ = ๐๐๐๐๐๐๐๐๐๐๐๐,๐๐๐๐๐๐๐๐๐๐๐ฆ๐ฆ๐๐,๐๐๐๐๐๐๐๐๐๐ = ๐ฆ๐ฆ๐๐,๐๐๐๐๐๐๐๐๐๐
๐
๐
๐ด๐ด๐๐๐๐๐๐๐๐๐๐
(๐๐๐๐๐๐) ๐๐๐๐๐๐๐๐๐๐๐๐ ๐ด๐ด๐๐๐๐๐๐๐๐๐๐๐๐๐๐
= โ๐๐๐ถ๐ถ๐๐ ๐ฆ๐ฆ๐๐,๐๐๐๐๐๐๐๐๐๐
๐๐๐๐๐๐๐๐๐๐๐๐๐๐๐๐ ๐ด๐ด๐๐๐๐๐๐๐๐๐๐
Thus, the amount of component that was loaded on the lithium sample can be calculated by
Equation 5.11, as follows:
๐๐
Equation 5.11
๐๐ ๐๐
๐๐๐๐๐๐๐๐๐ข๐ข๐๐ (๐๐๐๐๐๐) ๐๐๐๐๐๐๐๐๐๐๐๐ ๐ด๐ด๐๐๐๐๐๐๐๐๐๐๐๐๐๐
โ๐๐๐๐ = ๏ฟฝ ๐น๐น0๐ฆ๐ฆ๐๐,0๐๐๐ก๐กโ ๏ฟฝ ๐น๐น๐ฆ๐ฆ๐๐๐๐๐ก๐กโ โ๐๐๐ถ๐ถ๐๐ ๐ฆ๐ฆ๐๐
๐๐0 ๐๐0 ๐๐๐๐๐๐๐๐๐๐๐๐๐๐๐๐ ๐ด๐ด๐๐๐๐๐๐๐๐๐๐
The uptake of gas component on lithium can then be calculated from Equation 5.12.
๐๐
Equation 5.12
๐๐๐๐๐๐๐๐๐ข๐ข๐๐
โ๐๐๐๐
๐๐๐๐ = ๏ฟฝ ๐๐
5.3 The Results of the Reaction of Lithium with Nitrogen
5.3.1 Synchrotron XRD
Measurements using the Synchrotron XRD apparatus have been conducted to obtain an in-situ
characterization of the crystal switching from lithium metal to lithium nitride. A static cell and a flow-
through were set up to control the lithium surface chemistry. In the static cell, there was only a trace
amount of moisture (less than 10 ppm) in the nitrogen atmosphere, which meant that moisture would
have less/slower effect to the surface of lithium metal. In the flow-through cell, as nitrogen gas was
flowing through continuously, a larger amount of moisture passed across the surface of lithium metal,
thus creating more active edge sites.
The results of the Synchrotron XRD measurements at 90 ยฐC in atmospheric nitrogen from both the
static and the flow through cell are shown in Figure 5.3 (a) and (b). For both cases, strong peaks of
lithium at 19, 27.3, and 33.5 could be seen at the initial time of the experiment, which intensity faded
over time as new peaks at 12.3, 15.1, 25.3, 26.1, 29.0, and 36.0, representing the presence of lithium
108
|
UWA
|
nitride, began forming. This observation proved that the phase transformation from lithium to lithium
nitride did occur at 90ยฐC and atmospheric nitrogen condition.
Additionally, the LiOH phase, with a peak at 17.3, was always detected before the formation of the
lithium nitride phase in both cases. This might be due to the higher reactivity of water moisture to
lithium and lithium nitride, which would result in the side product of LiOH. The side reaction of lithium
and water moisture is beneficial, as it would result in creation and reveal of more active edge sites,
which is crucial for the nitrogen and lithium reaction.
Lastly, although in both cells the phase transformation from lithium to lithium nitride was observed,
the initiation times differed: 132 minutes for static cell and 32 minutes for the flowthrough cell. With
more water moisture flowing across the surface of the lithium metal forming LiOH in the flow-through
cell, more active edge sites could be created, which lead to a faster initiation time of the reaction
between lithium and nitrogen.
109
|
UWA
|
5.3.2 Thermogravimetric Measurements of Uptake
The uptakes of nitrogen and methane on different lithium samples were measured on the TGA (Model:
SDT Q600). Both the non-activated and pre-activated lithium samples were heated at a temperature
ramp rate of 10 ยฐC/min and kept at 60 ยฐC for 10 mins under an argon flow (100 ml/min) before the gas
was switched to nitrogen or methane (100 ml/min). The uptakes were calculated from the weight
change of the lithium samples and the reaction enthalpies were extracted from the measured heat
flow directly.
The uptake of nitrogen and methane on non-activated and pre-activated lithium samples at 60 ยฐC and
138 kPa are shown in Figure 5.4. There was no uptake of nitrogen observed on the non-activated
lithium sample within the experimental time scale (200 mins). On the pre-activated lithium sample,
significant uptake of nitrogen of 24 mmol/g was observed, which indicates a full conversion of lithium
to lithium nitride. Similar to the results observed in the Synchroton XRD measurements, an initial delay
period was observed before the weight of the sample started to increase. Morever, the initiation
period was of around 45 minutes โ which is within the same magnitude as the results from Synchroton
XRD studies. The uptakes of methane on both non-activated and pre-activated lithium samples were
also measured. Negligible uptake of methane were observed, indicating a significantly large selectivity
of nitrogen over methane on lithium metal.
Figure 5.4 The uptakes of nitrogen and methane as a function of time on both non-activated and
pre-activated lithium samples at 60ยฐC and 138 kPa.
To obtain the enthalpy of the reaction, the heat flows of the reactions of both non-activated lithium
and pre-activated lithium with nitrogen under at 60 ยฐC are measured and shown in Figure 5.5.
Negligible heat flow was observed for the non-activated lithium, while significant heat flow was
111
|
UWA
|
detected for the reaction of pre-activated lithium with nitrogen. From the heat flow curves, the
enthalpy of reaction of the pre-activated lithium with nitrogen at 60 ยฐC and 138 kPa was extracted.
The resulting enthalpy from the heat flow curves was 55 kJ/mol (lithium), which is consistent with the
theoretical value of 54 kJ/mol.[46]
Figure 5.5 Heat flow of the reactions of different lithium samples with nitrogen at the experimental
temperature of 60 ยฐC and pressure of 138 kPa. The integrated enthalpy for the reaction of pre-
activated lithium with nitrogen is 55 kJ/mol (Li); the integrated enthalpy of the reaction of non-
activated lithium with nitrogen is 0 kJ/mol (Li).
5.3.3 ASAP2020 Measurements of Uptake
The uptakes of nitrogen and methane on the non-activated and pre-activated lithium samples at 60ยฐC
were also measured using the ASAP2020 apparatus via the static volumetric method, and the results
are shown in Figure 5.6. The reaction of the pre-activated lithium sample with nitrogen started when
the pressure reached 33 kPa at 60 ยฐC, and continued until the uptake of nitrogen on the pre-activated
lithium eventually reached 21.5 mmol/g. The color of the sample changed from metallic luster to dark
purple, which indicates the formation of lithium nitride. Titration with an aqueous solution of the
produced lithium nitride showed an alkaline level consistent with the amount of lithium reacted with
nitrogen (1 pellet of lithium nitride fully dissociate in 160 ml cold water, giving pH of 11.8, equivalent
to 0.006719 g lithium). There was no uptake of nitrogen on non-activated lithium across the
experimental pressure up to 100 kPa. There were no observed uptakes of methane on either pre-
112
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.