MSDE

View Article Online

PAPER View Journal

Exploring the limits of adsorption-based CO2

capture using MOFs with PVSA – from molecular
Published on 23 September 2019. Downloaded by University of Toronto on 1/3/2020 1:27:15 AM.

Cite this: DOI: 10.1039/c9me00102f

design to process economics†
ab cd
David Danaci, Mai Bui, Niall Mac Dowell *cd and Camille Petit *ab

Metal–organic frameworks (MOFs) have taken the materials science world by storm, with potentials of near
infinite possibilities and the panacea for adsorption-based carbon capture. Yet, no pilot-scale (or larger-
scale) study exists on MOFs for carbon capture. Beyond material scalability issues, this clear gap between
the scientific and engineering literature relates to the absence of suitable and accessible assessment of
MOFs in an adsorption process. Here, we have developed a simple adsorbent screening tool with process
economics to evaluate adsorbents for post-combustion capture, while also considering factors relevant to
industry. Specifically, we have assessed the 25 adsorbents (22 MOFs, 2 zeolites, 1 activated carbon) against
performance constraints – i.e. CO2 purity and recovery – and cost. We have considered four different CO2
capture scenarios to represent a range of CO2 inlet concentrations. The cost is compared to that of
amine-based solvents for which a corresponding model was developed. Using the model developed, we
have conceptually assessed the materials properties and process parameters influencing the purity, recov-
ery and cost in order to design the ‘best’ adsorbent. We have also set-up a tool for readers to screen their
own adsorbent. In this contribution, we show that minimal N2 adsorption and moderate enthalpies of ad-
sorption are key in obtaining good process performance and reducing cost. This stands in contrast to the
popular approaches of maximizing CO2 capacity or surface area. Of the 22 MOFs evaluated, UTSA-16
shows the best performance and lowest cost for post-combustion capture, having performance in-line
Received 12th August 2019, with the benchmark, i.e. zeolite 13X. Mg-MOF-74 performs poorly. The cost of using the adsorbents re-
Accepted 13th September 2019
mains overall higher than that of an amine-based absorption process. Ultimately, this study provides spe-
cific directions for material scientists to design adsorbents and assess their performance at the process
DOI: 10.1039/c9me00102f
scale. This works intends to bridge the gap between scientific and engineering studies to accelerate CO2
rsc.li/molecular-engineering capture adsorbents development. The approach can be extended to other molecular separations.

Design, System, Application

Adsorption-based processes represent one of the proposed routes for post combustion CO2 capture. The design and operation of an adsorption process,
and subsequently its feasibility, centres around the adsorbent, i.e. the porous material used to carry out the separation between CO2 and N2 molecules. As
such, evaluating adsorbents and identifying adsorbent properties that yield both technically and economically feasible designs is paramount to test the
potential of adsorption-based carbon capture at large scale. This task is particularly important given the thousands of adsorbents that exist. Adsorption pro-
cesses are inherently highly coupled systems. For example, an adsorbent with a very high capacity will have a correspondingly high temperature change
upon adsorption, which will reduce the equilibrium loading of that sorbent. The total energy released upon adsorption is the product of amount adsorbed
and enthalpy of adsorption. Thus, an optimum would exist for capacity which will balance this trade-off. Our work relates adsorbent properties with pro-
cess performance both from a technical, and economic perspective. Including the link between molecular-level material properties and process cost expe-
dites the design of new fit-for-purpose adsorbents. This work can be used to guide, or change the focus of, future adsorbent development for post-
combustion capture applications.

a d
Department of Chemical Engineering, Imperial College London, South Centre for Process Systems Engineering, Imperial College London, South
Kensington, London, SW7 2AZ, UK. E-mail: [email protected]; Kensington, London, SW7 2AZ, UK
Tel: +44 (0)20 7594 3182 † Electronic supplementary information (ESI) available: Descriptions of the ad-
b
Barrer Centre, Imperial College London, South Kensington, London, SW7 2AZ, sorption and absorption models, dual-site Langmuir isotherm parameters and
UK physical properties of the adsorbents investigated, supplementary results, the
c
Centre for Environmental Policy, Imperial College London, South Kensington, digitised isotherm data, and a surrogate model for the readers' use. See DOI:
London, SW7 1NE, UK. E-mail: [email protected]; Tel: +44 (0)20 7594 9298 10.1039/c9me00102f

This journal is © The Royal Society of Chemistry 2019 Mol. Syst. Des. Eng.
 View Article Online

Paper MSDE

Introduction exhibit a “fast” rate of mass transfer (non-limiting) and be

“easily” regenerated under moderate conditions.
Despite nearly two decades of intense research effort, some Comparatively to the number of lab scale research studies,
could argue that the field of so-called “designer adsorbents” progress towards scale-up remains very slow. Hence, it be-
is still more of a scientific curiosity than a fully realised tool. comes clear that in order to transition to the next stage or
Designer adsorbents allow one to tune materials for a given technology readiness level (TRL), one cannot rely solely on
application, e.g. CO2 capture in the context of the present the current practice of post-hoc rationalisation of materials to
study. Heavily researched exemplars of designer adsorbents maximise CO2 uptake, CO2/N2 selectivity or working capacity.
are metal organic frameworks (MOFs). A number of MOFs Such an approach makes it impossible to reliably and quickly
have been tested at lab-scale.1–19 These studies typically place
Published on 23 September 2019. Downloaded by University of Toronto on 1/3/2020 1:27:15 AM.

determine the most promising materials (and material attri-

a large emphasis on unveiling the molecular separation butes), especially when those are highly tuneable as in MOFs.
mechanisms and enhancing CO2 uptake, selectivity (e.g. ideal To address this knowledge gap, links must be drawn between
selectivity (eqn (1)), or Henry selectivity (eqn (2))), or working the adsorbent design and the adsorption process. In other
capacity (eqn (3)). words, adsorbent properties should be screened/evaluated

 using a process performance index. The screening tool
nA P feed ; T feed ·yfeed
A=B ¼
Sideal
Afeed feed  Bfeed (1) should provide a framework based on selected metrics that
nB P B ; T ·yA define a range of acceptable values for each of the process pa-
rameters considered. Typical parameters for the evaluation of
where nj is the uptake of component j at the partial pressure
{adsorbent + adsorption process} systems are: purity, recov-
of j in the feed (Pfeed
j ) at the feed temperature T feed, and yfeed
j
ery, energy consumption and cost. In the case of CCS, a CO2
is the molar fraction of j in the feed gas.

 purity of ≥ 95% mol is desired, and the highest recovery pos-
H A T feed sible; 90% mol is generally taken. However, when comparing
Henry
SA=B ¼
feed  (2)
HB T alternatives within a technology (i.e. various adsorbents) or be-
tween technologies (i.e. adsorption, absorption, membranes,
where Hj is the Henry constant for component j at the feed distillation), the cost of undertaking the separation is the next
temperature, T feed. key consideration. Process economics allows for trade-offs be-
tween capital and operating costs to be explicitly described.
WCpure
A = nA(P feed·yfeed
A ,T
feed
) − nA(Pdes·yfeed
A ,T
feed
) (3) For instance, an adsorbent with poor working capacity may
have higher capital costs attributed to adsorbent amount and
where nA(P,T) is the uptake of component A at a partial pres- adsorption vessels, however, it may show better performance
sure P and temperature T, P feed and P des are the feed and de- at lower vacuum levels (higher absolute pressure), saving capi-
sorption pressures respectively, T feed is the feed temperature, tal and operating costs for vacuum generation.
and yfeed
A is the molar fraction of A in the feed gas. The evaluation of adsorbents accounting for these, and
While these are necessary features to investigate, they are other, parameters is crucial, as it gives a more accurate repre-
in no way sufficient. For instance, an adsorbent should also sentation of the attainable performance at scale. In the ab-
sence of experimental determination, these parameters can be
accounted for by rigorous adsorption process simulation,20–29
or by reduced-order adsorption process models,30–34 as de-
Dr Camille Petit is a Reader in scribed in a number of CO2 capture studies. Rigorous cyclic
the Chemical Engineering De- modelling can be arduous, especially regarding the cycle de-
partment at Imperial College sign and optimization. For what we aim to address here, i.e.
London. Dr Petit was a postdoc- swift adsorbent evaluation and screening, this approach is pro-
toral researcher at Columbia hibitive owing to a typical paucity of available data. Reduced-
University and received her PhD order models are typically 0-D or 1-D equilibrium models
in Chemistry in 2011 from the which account for the aforementioned issues. These adsorp-
City University of New York. Her tion models are ideal for adsorbent screening as they are rapid
research focuses on elucidating and require minimum input data that can be easily obtained
the fundamentals of porous ma- using common laboratory apparatus.
terials formation, structure, and When applying these models, one should select the spe-
chemistry to exploit them in cific adsorption process, i.e. pressure swing adsorption (PSA),
Dr Camille Petit interfacial applications, e.g. sep- pressure vacuum swing adsorption (PVSA) temperature swing
aration of molecules. Her work adsorption (TSA) or a combination of PSA and TSA (PTSA).
also investigates the implications of using these materials at the All of these adsorption processes are cyclic, and in their sim-
large-scale. Dr Petit has published >60 peer-reviewed articles. She plest implementation, a mixture is passed through a packed
is the recipient of an ERC Starting Grant and the 2017 AIChE's 35 bed of adsorbent where the more strongly adsorbed (heavy)
under 35 award. component is removed, and the other (light) component

Mol. Syst. Des. Eng. This journal is © The Royal Society of Chemistry 2019
 View Article Online

MSDE Paper

passes through. Upon saturation of the adsorbent, the feed is teristics, and to generate a dataset that can be used by others
diverted to another vessel and the saturated vessel is regene- to simply evaluate the performance of their adsorbents.
rated. In the case of PSA or PVSA this is by a reduction in Our work evaluates adsorbents at scale without the neces-
pressure, and the more strongly adsorbed is collected as the sity of intermediate scale-up steps. Those that are promising
other product. PSA or PVSA are commonly proposed as they could be candidates for larger scale investigation, and rigor-
have shorter cycle times than TSA processes, due to the ous adsorption modelling. This is accomplished while ac-
lengthy cool down times of TSA beds. Given the vast volume counting for a range of industrially relevant factors, including
of flue gases that are associated with large-scale post-combus- separation performance, process equipment requirements,
tion capture, short cycle times are key in reducing the and cost. The association of adsorbent properties with these
Published on 23 September 2019. Downloaded by University of Toronto on 1/3/2020 1:27:15 AM.

amount of CO2 that must be captured/stored per cycle – as factors forms a link between materials science and industry;
there are the associated adsorbent requirements. PVSA – the a pertinent requirement to accelerate the implementation of
approach selected for this work – also has the potential for adsorption-based carbon capture technologies.
much lower regeneration energy requirements than amine-
based absorption processes.35 This is due to the absence of Methods
both chemical reactions to be reversed, and the sensible
heating requirements of the amine solution. A possible area Materials selection and cyclic adsorption model
of concern regarding packed-bed processes are the pressure We provide here a general overview of the model and de-
drop limitations. Given the high volumetric flow rates, man- scribe the details in the ESI.† The model includes three main
agement of the pressure drop to avoid issues such as gas sections, the PVSA model, equipment enumeration and
channelling and adsorbent crushing may result in a large sizing, and process economics.
number of vessels, as will be investigated in this work. An im- The PVSA model determines the adsorbent performance
portant aspect to consider when analysing a PSA/PVSA pro- under cyclic and adiabatic conditions. For this, we employ
cess is the variation of the adsorbent bed temperature. As ad- the 0-D equilibrium model developed by Maring and
sorption is an exothermic phenomenon, the bed experiences Webley,30 albeit with 500 steps used in the solution of the
a temperature increase during the adsorption step and a cor- equations, rather than 100. This helps improve the accuracy
responding decrease during the desorption step. These tem- of the result for materials with steep isotherms. The valida-
perature changes hamper the working capacity that the ad- tion of the adsorption model was conducted in the original
sorbent has. Upon adsorption, the amount adsorbed is less work, which found an acceptable match between experimen-
than the isothermal (or feed temperature) amount. During tal data and rigorous simulation results.
desorption, the gas recovered is less than the isothermal The model is equilibrium-based and its benefits over a
amount. This can result in a situation where the non- kinetic-based model include: minimal input data, rapid solu-
isothermal working capacity is significantly lower than that tions, and improved accuracy over simple metrics. Yet, opting
under isothermal conditions (or read off a single isotherm). for an equilibrium-based adsorption model has associated
While the desorption step is taking place, the composition of compromises. A kinetic model allows for bed profiles (i.e.
the gas and adsorbed phase also changes. As gas is removed temperature, pressure, composition) to be obtained, which
from the adsorbent, this mixes with the existing gas in the leads to a more accurate simulation, and flexibility around
bed, causing the composition to change. This means that the the cycle operation – not possible with an equilibrium model.
concentration of CO2 at any point of desorption process is The timing of cycle steps can be optimised to maximise pro-
not the same as the feed concentration; thus, simple iso- cess performance. Another advantage of kinetic models is
therm evaluation cannot give a reliable estimate. This is fur- that all adsorbents can be modelled, however, experimentally
ther complicated by the fact that the ‘end composition’ is determined mass transfer coefficients are required and these
not known a priori. The CO2 concentration in the bed at the are generally obtained from breakthrough measurements.
end of the desorption step is higher than the feed, thus the Breakthrough apparatus are fairly uncommon, and analysis
partial pressure is higher than imagined, also reducing the of the data must be conducted carefully. In the absence of re-
amount that can be removed from the adsorbent. liable adsorption kinetics data, efforts in cycle optimisation
In our study, we have applied a PVSA model to screen 22 may be futile. Equilibrium models are valid only for adsor-
MOFs for CO2/N2 separation from post-combustion flue-gas bents with sufficiently fast kinetics. It is therefore inappropri-
sources from natural gas combustion, coal combustion, ce- ate to apply the model for adsorbents, which are known to
ment production, and steel production. We have screened have slow adsorption kinetics, such as K-A, K- or Cs-CHA,
MOFs using data reported in the literature and compared CMS, and supported amines at ambient temperatures. In
their performance and cost to that of commercial zeolites summary, an equilibrium-based model allows a first assess-
and activated carbon. Following this approach, we have deter- ment of {adsorbent + process} systems, while a kinetic-based
mined links between adsorbent properties that influence as- model enables a refined evaluation and a possible optimiza-
pects of the separation process (i.e. purity, recovery, and cap- tion of the process.
ture cost). We have also conducted a parametric evaluation of Adsorbents features and inputs. The PVSA model requires
adsorbent properties to identify optimum adsorbent charac- isotherms as a function of temperature and pressure for CO2

This journal is © The Royal Society of Chemistry 2019 Mol. Syst. Des. Eng.
 View Article Online

Paper MSDE

and N2, the density and void fraction of the bed, and the ad- The reliance on reported isotherms as well estimated
sorbent heat capacity as inputs. One must have experimen- physical properties gives rise to uncertainties. Of the physical
tally measured isotherm data while the other inputs can be properties required in this work, the approach used for the
approximated by various means, as detailed below. The avail- MOF heat capacity has the most uncertainty. Using a molar
ability of isotherm data in the literature at three tempera- fraction contribution of the metal atoms and ligands ignores
tures for both CO2 and N2 represent the limiting factor in the contribution of the lattice/crystal vibrational modes, thus
this work. In total, we investigated 22 MOFs along with an ac- underestimating the heat capacity. Conversely, the loss of
tivated carbon, zeolite 13X and zeolite 5A. The 22 MOFs are some degrees of freedom of the ligand upon coordination is
in contrast to the approximately 86 that any gas adsorption also not reflected, overestimating the heat capacity. The rela-
Published on 23 September 2019. Downloaded by University of Toronto on 1/3/2020 1:27:15 AM.

data for CO2 and N2 were found in the literature, and to the tive contributions of these factors are not known and will
many hundreds of reported MOFs without any reported gas likely vary on a case-by-case basis. We therefore carried out a
adsorption data. The isotherm data, sources, and correspond- sensitivity analysis on the adsorbent physical properties and
ing dual-site Langmuir isotherm fits are provided in the ESI.† their influence on performance metrics (i.e. purity, recovery,
As a result of the limited information available, some of the and cost) and we discuss the findings in the results and dis-
adsorbents evaluated have stability issues in the presence of cussion section. Uncertainty also exists in isotherm
moisture, and other acid gas components in the flue gases. measurements,39–43 the root causes of this are multifaceted
Yet, as will be explained in the cost analysis section, the encompassing material synthesis, apparatus, and method.
annualised capture costs are substantial regardless of the The utility of a sensitivity analysis on this is contentious, as
MOFs, therefore, the inclusion of flue gas pre-treatment such isotherms repeated on the same sample on the same appara-
as dehydration and desulphurisation will not alter the final tus typically show little variation.
conclusions. As our model relies on reported experimental isotherms,
The physical properties data for the adsorbents – i.e. den- this approach takes advantage of the existing literature and
sity, bed void fraction, and heat capacity – are often accounts for ‘real’ structures rather than model ones. Yet,
unavailable and must be estimated. We used the desolvated the requirement of having CO2 and N2 isotherms at 3 temper-
crystallographic information files of the MOFs to derive some atures means that only a few materials can be investigated
parameters. We determined the porosity of the adsorbent here.
(εads) using the solvent accessible pore volume; a probe with Adsorption process features, inputs and outputs. We cov-
Connolly radius of 1.3 Å (helium) was used to obtain the oc- ered a wide range of CO2 capture scenarios: a natural gas
cupied and free volume with the porosity obtained by apply- combined cycle flue gas (NGCC) a high rank pulverised coal
ing eqn (4). From this, we calculated the total void space of ultra-supercritical flue gas (PC), a cement plant (cement), and
the packed bed (εbed) by applying eqn (5), taking 0.37 (ran- an integrated steel mill (steel). The details of each scenario
domly packed spheres) as the packing void fraction (εpack). are provided in Table 1, with the sizes of each being typical
We estimated the pelletised adsorbent density (ρpellet) by ap- scales of those plants. We determined the power plant flue
plying a scaling factor of 80% to the crystallographic density; gas flow rates and compositions using a procedure outlined
this relationship was established by Wu et al.36 We then in the ESI.† The composition and flow rate data regarding the
quantified the density of the packed bed (ρbed) using eqn (6). industrial capture applications are ‘end-of-pipe’ capture op-
tions proposed by the IEAGHG,44,45 where all flue gases are
V free consolidated in to one stream.
εads ¼ (4)
V free þ V occupied We considered an inlet gas pressure of 1.05 bara at a tem-
perature of 313.15 K. The gas is then compressed to 1.50 bara
and cooled to 313.15 K for adsorption. We investigated the
εbed = εpack + (1 − εpack)·εads (5) influence of desorption pressure on adsorbent performance
at five levels: 0.15 bara, 0.10 bara, 0.05 bara, 0.03 bara and
ρbed = (1 − εpack)·ρpellet (6) 0.01 bara. These low vacuum levels are not necessarily indus-
trially viable. We chose these levels in order to obtain the
The estimation of the heat capacity does not have an best possible performance (purity and recovery) from the
established method. We calculated the heat capacity of the
MOFs by summing the molar fraction contributions of the
metal atoms and the organic ligands. We used the molar heat Table 1 Summary of post-combustion capture scenarios investigated in
capacity of the metal atoms from Rumble,37 and approxi- this work
mated the molar heat capacity of the organic ligands at
CO2 concentration Flow rate CO2 emissions
313.15 K following the method described by Goodman
Scenario Size [mol mol−1] [kg s−1] [MM tCO2 y−1]
et al.38 A sample calculation is provided in the ESI.† For the
Natural gas 400 MWe 4.38 607 1.28
three reference materials, i.e. zeolite 13X, 5A, and activated
Coal 500 MWe 12.5 377 2.18
carbon, the sources for their physical property data are pro- Cement 1 MM tpa 21.0 152 1.41
vided in the ESI.† Steel 4 MM tpa 25.5 621 6.86

Mol. Syst. Des. Eng. This journal is © The Royal Society of Chemistry 2019
 View Article Online

MSDE Paper

adsorbents. If even at these vacuum levels, the performance neric manufacturers, with costs being up to 3.5–5 £ kg−1 for
is not reaching the desirable level, a change in the process is the state-of-the-art materials from the eminent manufac-
required. turers. If a MOF is deemed suitable for carbon capture appli-
We used the output from the PVSA model to size and enu- cations and is commercialised, raw material production can
merate the required process equipment. The PVSA model be scaled to the requisite level with existing technologies. An
returns values on the basis of 1 kg of adsorbent and the pro- assigned value of 1.5 £ kg−1 assumes that MOF production
cess is scaled to the feed flow rate of each scenario, based on could be scaled efficiently and is the best-case scenario cost
the total feed required to complete one adsorption cycle for 1 for a synthetic adsorbent. As per the physical properties, the
kg. As the PVSA model implemented here is an equilibrium influence of adsorbent cost on capture cost was investigated
Published on 23 September 2019. Downloaded by University of Toronto on 1/3/2020 1:27:15 AM.

model, cycle times are imposed ex post facto. We specified a in our sensitivity analysis (see Results and discussion).
1-hour total cycle consisting of 45 minutes adsorption/feed We converted the electrical requirements to an equivalent
time, and 15 minutes desorption time. This is compared to cost of natural gas that would need to be combusted in an
≤15 minutes for PSA systems used for H2 production or air on-site boiler to produce that electrical power. We did not in-
separation. The cycle time is the main contributor to process clude these outside battery limits (OBL) costs in the separa-
intensification: fewer adsorbent vessels are required if the cy- tion plant capital costs.
cle time is reduced. This is due to the requirement to accom- We determined the installed costs using the IChemE/fac-
modate the adsorbing species in the feed for the duration of torial cost estimation method46,47 with equipment purchase
the adsorption time. The volumetric flow rate of the feed gas costs taken from a range of sources detailed in the ESI.† We
dictates the line and valve sizes, which for large-scale post- adjusted the costs for economic factors using the Chemical
combustion capture, is inherently high. This necessitates Engineering Plant Cost Index (CEPCI). Values reported in this
large pipe diameters and valves, generally motor actuated, work are based on the october 2018 value of 616.3, and a
with slow actuation times (5–10 minutes). Thus, switching USD to GBP exchange rate of 0.76 £ $−1. There are limitations
flow paths between cycle steps is time consuming. In the in- in the costing approach proposed here (accuracy is ±30%).
stance where one bed is being switched from adsorption to Yet, such approach allows us to make a more informed con-
regeneration, the feed must be diverted to the regenerated clusions with the data that is currently available.
column prior to shutting off the saturated one. These factors To determine the cost per tonne of CO2 captured, we used
culminate in a situation where cycle times <30 minutes for the total annualised cost (TAC), eqn (7), where CRF is the
these situations are highly unlikely. As detailed design data capital recovery factor and CCap and COp are the total capital
is unknown, we have assumed a cycle time of 1 hour to ac- costs and annual operating costs, respectively.
count for these factors while also allowing for some margin
in the operation. TAC = CRF·CCap + COp (7)

For the capital recovery factor (eqn (8)), we assumed a cost

Process economics of capital (i) of 10%, and a repayment period (n) of 25 years.
The primary equipment items included in the capital cost cal-
culations are the feed gas compressor, feed cooling ex- i·ð1 þ iÞn
CRF ¼ (8)
changers, adsorbent vessels, adsorbent, and vacuum pumps. ð1 þ iÞn − 1
The minor equipment items included are the cooling water
pumps, vacuum pump motors, and control valves directly as- The costs of capture for each scenario using amine ab-
sociated with the adsorbent vessels. The operating costs cover sorption to achieve 90% CO2 capture were also determined as
the electrical requirements for the rotating machinery, cooling a point of comparison. Costs for 30% wt monoethanolamine
water requirements, and adsorbent replacement costs. (MEA) and a ‘new generation’ (‘NG’) absorbent with lower re-
For the purposes of comparison, we assigned all adsor- generation energy requirements such as Cansolv.48 A cost of
bents a cost of 1.5 £ kg−1 and a lifetime of 5 years. Some 1100 US$ t−1 was used for both amine absorbents. The details
MOFs have chemical and thermal stability issues and/or are of the model are provided in the ESI.† A conventional process
costly to synthesise due to the requirement for exotic organic topology is used with multiple unit operations were required
ligands, rare metals, and solvent use. Therefore, it may be na- due to equipment size limitations (heat exchanger areas and
ive to assume that all MOFs have the same lifetime of 5 years. column fabrication limitations).
Adsorbents currently applied industrially such as zeolites, ac- The flue-gases are assumed to be provided dry and with-
tivated carbons, and silicas and aluminas, display lifetimes out contaminants, with the drawback that pre-treatment
of 7–10 years if the process is operated sensibly. As there are equipment and costs are not accounted for in this work. The
no long-term studies of MOFs being applied at the pilot scale implications of this are not significant in this case as
for adsorption-based processes, we have opted for a conserva- explained in the results and discussion section.
tive estimate based on existing knowledge, while also giving The majority of the work on post-combustion capture via
the MOF materials the best chance. The cost of 1.5 £ kg−1 is adsorption has been focused on coal-fired sources, likely due
in-line with industrial purchase costs for zeolite 13X from ge- to their more manageable CO2 content. Indeed, the higher

This journal is © The Royal Society of Chemistry 2019 Mol. Syst. Des. Eng.
 View Article Online

Paper MSDE

CO2 content in coal-fired flue gases means that the separation icant variation. Indexing of costs to current day values was also
is easier to achieve. These reported economic assessments49,50 not mentioned in either case.
provide comparison points for our work. We have observed
significant variation between these reported estimates and our Results and discussion
work (3 to 10 times). The cost differences between our work
and that of Susarla et al.50 likely arose from the differences in Adsorbent evaluation and screening
the cost of vessels and vacuum pumps. In our case, we calcu- Applying the model described above, we have screened 22
lated the cost of adsorbent vessels based on mass of steel re- MOFs, 2 zeolites and an activated carbon for the CO2 capture
quired, which is determined using the relevant pressure vessel scenarios mentioned earlier (Table 1). We show the outcomes
Published on 23 September 2019. Downloaded by University of Toronto on 1/3/2020 1:27:15 AM.

standards. The vessels must withstand vacuum, which results for all scenarios in Fig. 1. We describe in detail the NGCC
in high vessel thicknesses and a correspondingly high mass. case first and then provide a broader overview on the other
For vacuum pumps, we opted for an equipment cost correla- scenarios. In Fig. 1, we have grouped the adsorbents in the
tion from Garrett.51 Susarla et al. used cost correlations for the following way, Mg-MOF-74 through UTSA-16 are MOFs with
vessels based on volume, which ordinarily do not account for open metal sites, MOF-177 through ZnIJBPDC)IJBPP) are those
the pressure vessel requirements, and they obtained costs for with fully saturated coordination sites, ZIF-8 through ZIF-82
vacuum pumps from vendors. Considering vessels and vac- are the sub-classification of MOFs known as zeolitic
uum pumps comprise the greatest proportion of capital costs, imidazolate frameworks, and activated carbon through zeo-
differences in the determination of these could result in signif- lite 13X are commercial adsorbents for reference.

Fig. 1 Separation performance and cost evaluation of 25 adsorbents for CO2/N2 separation for the four scenarios investigated as a function of
desorption pressure: (a) NGCC, (b) coal-fired, (c) cement, (d) steel. In all cases, the cost of the MOFs is taken as £1.5 kg−1.

Mol. Syst. Des. Eng. This journal is © The Royal Society of Chemistry 2019
 View Article Online

MSDE Paper

NGCC case. Zeolite 13X and UTSA-16 display the best re- to S12†). The increased feed concentration from approxi-
sults in terms of purity and recovery (69% mol and 81% mol mately 4% mol CO2 not only increases the CO2 working ca-
respectively at 0.01 bara). Mg-MOF-74, one of the most often pacity for adsorbents with more linear isotherms, the associ-
cited MOFs for CO2 capture,17 appears as a very poor adsor- ated decrease in N2 concentration means N2 adsorption is
bent for this process (24% mol and 37% mol respectively at also reduced. The relationship to capture cost is not as
0.01 bara). These results highlight particularly well the limita- straightforward since the flow rates between each case are
tions of simply considering CO2 uptake as a relevant metric; not consistent, and neither are the obtained purity and recov-
this aspect is discussed further below. For all adsorbents, ery values. In general, the increased CO2 recovery for most
deep vacuum is required to obtain any noticeable improve- adsorbents is a major contributor to the reductions in cost,
Published on 23 September 2019. Downloaded by University of Toronto on 1/3/2020 1:27:15 AM.

ments in CO2 recovery, and purities are generally low. and further capital cost reductions are seen due to the lower
The most striking output is the capture cost, with most gas flow rate. Yet, the cement and steel scenarios display sim-
data points >300 £ t−1, while amine absorption is below 150 ilar cost despite significantly different flow rates. Increasing
£ t−1. These costs may seem initially higher in comparison to CO2 concentration has limited benefit on M-MOF-74 adsor-
other reports.52–57 In all those cases, single equipment items bents (i.e. steep isotherms) as they transition into the satu-
are used and sized without limitations. In our work, limits rated area of the isotherm. The initial improvement in perfor-
are imposed on maximum column diameter, heat exchanger mance – from the NGCC to the coal scenario – stems from
area, feed blower flow rate etc. resulting in multiple unit op- the transition from the saturated region of the isotherm to
erations. The full details of this are provided in the model de- the steeper region during desorption, thereby increasing the
scription in the ESI,† however, the result of this is that capital working capacity. Most of the other MOFs display gradual
costs using our approach are high. isotherms and these higher partial pressures result in im-
The adsorption high costs are capital driven, and a break- proved working capacities. For the cement and steel scenar-
down of the costs are presented in the ESI† (Fig. S1 to S4). ios, the 0.01 bara desorption points are grouped fairly well to-
One reason for this is the feed gas is low density and corre- gether. This is mostly coincidental as there is no trend in the
spondingly the volumetric flow rate is significantly high. Ac- numerical data that suggests it is due to a given factor. The
commodating the recommended gas velocities and pressure specific combination of CO2 working capacity, CO2 recovery,
drops in the adsorption columns necessitates several col- and vacuum flow rate required for each adsorbent has
umns in parallel to distribute the feed. Low velocities (i.e. resulted in this grouping with improvements in one value be-
pressure drop per unit length) will result in gas channelling ing offset by another.
and poor distribution of gas through the bed. Conversely, In addition to purity and recovery, another commonly
high overall pressure drops (i.e. total pressure drop of the used performance indicator is the energy penalty, or parasitic
packed bed) can crush the adsorbent pellets. Vacuum pumps energy of the plant. We have included plots of energy penalty
also contribute significantly to the capital costs. Normally, for all scenarios in the ESI† (Fig. S5 to S8). For UTSA-16, the
large-scale vacuum generation is achieved using steam ejec- total power and specific energy range from 83 MW and 1.96
tors, however, the product becomes mixed with steam. The GJe tCO2−1 in the cement scenario to 377 MW and 1.82 GJe
limitation of mechanical vacuum pumps is the maximum tCO2−1 for the steel scenario at 0.01 bara desorption pressure.
achievable flow rate. The variant opted for here has a maxi- The power requirements are affected by the feed flow rate dif-
mum flow rate of ≈15 000 act m3 h−1. However, given that ferences, and the specific energy requirements may be a bet-
the desorption step is a fraction of an hour (in this case 15 ter comparison between separations. For the natural gas and
minutes), multiple parallel pumps are required to achieve the coal scenarios, specific energies of 4.1 and 2.4 GJe tCO2−1, re-
nameplate flow rate. Furthermore, at 0.01 bara, a given num- spectively, are required.
ber of moles of gas occupies ≈100 times more volume than The scenarios investigated here do not highlight any
at standard conditions. grouping within the MOF families. This suggests that there is
The adsorbent working capacity also contributes to the not a singular adsorbent factor or characteristic that defines
number of columns. Adsorbents such as MOF-177 and ZIF- the suitability of an adsorbent for a given application. This is
8 have low capacities and linear isotherms in the pressure not necessarily surprising as working capacity, recovery, and
range of interest.58,59 This results in a low working capacity working selectivity are the parameters that define process
which necessitates a considerable adsorbent mass, and corre- performance and cost. It is not unforeseeable that a multi-
sponding vessel volume to adsorb the captured CO2. tude of isotherms could yield the same results. For instance
Other cases. We now turn our attention to the other sce- as previously mentioned, UTSA-16 and zeolite 13X display
narios. We provide a more brief analysis for those since the very similar performance, however, they are very different in
same theories as the NGCC case apply (Fig. 1). All cases ex- nature. This hypothesis is tested later in this study.
hibit better purities, recoveries and costs compared to NGCC. Comparison of performance and cost with respect to tar-
This observation is globally attributed to the higher CO2 con- gets. Of the 22 MOFs investigated, none under any of the
centration in the flue gas, resulting in an easier separation. conditions investigated here display CO2 purities ≥95% mol,
This aspect is also reflected in the working selectivities with which is the level generally considered acceptable for CCS.
many more adsorbents displaying values above unity (Fig. S9 UTSA-16 displays similar performance to 13X in this work,

This journal is © The Royal Society of Chemistry 2019 Mol. Syst. Des. Eng.
 View Article Online

Paper MSDE

and 13X has been shown to achieve purities ≥95% mol and greater number of columns to continuously accommodate
recoveries ≥90% mol (ref. 60 and 61) for coal-fired post-com- the feed. Yet, this may be offset by the absence of vacuum
bustion capture. This comparison highlights the possible im- costs. Whether suitable process intensification can be
provement brought by cycle optimization, an aspect we dis- achieved will determine the feasibility of conventional TSA
cuss in a latter section. The choice, and more importantly, processes for large-scale applications. If recent developments
the optimization of the adsorption cycle used significantly in TSA technology such as Inventys' VeloxoTherm™ process70
impact the attainable results.62 The same cycle (i.e. sequence are successful, it may enable industrial scale application of
of steps) that is operated or designed (i.e. flow direction, step TSA for post-combustion capture.
times, intermediate pressures, etc.) by different people will From the screening exercise above, we conclude that: (i)
Published on 23 September 2019. Downloaded by University of Toronto on 1/3/2020 1:27:15 AM.

yield different results. A complex cycle poorly optimised may MOFs performed poorly relative to amine-based absorption
yield worse results than a finely tuned simpler cycle. Complex for the selected CO2 capture scenarios, though we observe
cycles add significant capital cost due to the multiplicity of improvements for more concentrated feeds and further im-
columns required. If the ratio between total feed time and de- provements could be brought by process optimization; (ii)
sorption time are equal, i.e. 75% of total cycle time is adsorp- UTSA-16 performs the best out of all the MOFs screened
tion time, a 6-bed cycle would require six times the number while Mg-MOF-74 performs unsatisfactorily; (iii) the former
of beds of the cycle used here. Considering UTSA-16 for the observation highlights how the evaluation methods used in
coal case requires 47 beds at best, although it may be techni- the existing literature are typically unfitting to claim a given
cally feasible, it will not be economically or spatially feasible MOF is suitable for CO2 capture applications.
for these high flow rate applications.
We now provide a perspective on costs with comparison to
the costs of amine absorption. All scenarios here cost more UTSA-16 vs. Mg-MOF-74
than amine adsorption, though we find exceptions for ce- We investigate here in more details the underlying reasons
ment and steel at ‘high’ desorption pressure (0.10–0.15 bara). for the distinct performances of these two MOFs. This ‘exer-
For the lowest desorption pressure (0.01 bara) for which the cise’ allows us to get a first grasp on the links between mate-
best performance in terms of purity and recovery are seen, rial properties and process performance. The isotherms of
the costs remain significantly higher than the benchmark. USTA-16 and Mg-MOF-74 are shown in Fig. 2. Mg-MOF-74 ex-
There are two main contributions to this. The first is the su- hibits steep isotherms and a high enthalpy of adsorption,
perior gas handling capability of absorption columns relative leading to two compounding effects. The steep isotherms re-
to packed columns. A given diameter can accommodate a quire very deep vacuum to desorb the adsorbed gas. Since
higher flow rate with lower pressure loss. For the 0.01 bara the process is adiabatic, a high enthalpy of adsorption causes
case, the vast majority of the costs are vacuum equipment greater temperature increase in the bed upon adsorption, re-
capital costs, which are unavoidable in a PVSA system. Fi- ducing the capacity, and the corresponding decrease upon
nally, the amine absorption process achieves a recovery of desorption retains the adsorbed gas. The bed temperature
90% mol and purities ≥ 95% mol CO2, it is therefore an un- swing, i.e. the difference between the highest and lowest tem-
fair comparison to a certain extent but further demonstrates perature experienced, is 25 K for Mg-MOF-74, vs. 5 K for
that for these large-scale applications, PVSA may not be the UTSA-16 at 0.01 bara desorption. Comparing the isotherms in
optimal choice. The use of MEA based absorbents as a bench- Fig. 2, it also becomes apparent that ultimate CO2 capacity is
mark is also somewhat outdated as there are new systems not necessarily significant. In fact, working capacity drives
based on mixtures of methyl diethanolamine (MDEA) and pi- the low recoveries seen in many cases. Indeed, although Mg-
perazine (PZ), and 2-amino-2-methyl-1-propanol (AMP) and pi- MOF-74 has excellent CO2 capacity, it does not perform well
perazine, which display more desirable characteristics63 and when regenerability is accounted for. Considering the cyclic
may have lowered capture costs even more. Given the costs nature of the adsorption process, regenerability/desorption
presented thus far, pretreatment and product compression has equal importance as adsorption.
costs are not an immediate concern. However, examples of The excellent CO2 adsorption of Mg-MOF-74 is also accom-
these costs can be found in a range of resources.64–66 panied by high N2 adsorption. This is due to the large pore
From an adsorbent perspective, of the MOFs evaluated, apertures and high surface areas, which result in the adsor-
UTSA-16 does appear to perform the best for the post- bent being non-discriminatory. In such a case where there is
combustion capture scenarios investigated using PVSA. Its ≈96% mol N2 in the feed, N2 adsorption becomes an area of
performance is on par with zeolite 13X over all factors con- concern. The adsorbed N2 is collected with the CO2 product,
sidered here. Considering its water co-adsorption and regen- and as such, any N2 adsorbed dilutes the CO2 product. This
eration characteristics are more desirable than zeolite is further enhanced by the fact that N2 is easier to desorb
13X,67–69 it may be a preferred candidate assuming that long- than CO2, and in many cases the working capacity of N2 can
term cyclic stability can be verified. It is possible that the be greater than CO2. The greater purity results displayed by
MOFs investigated here display better performance in a TSA UTSA-16 are not solely due to better CO2 working capacity,
process. However, conventional TSA processes have long cycle but also due to excellent selectivity. For the purposes of com-
times due to the cooling of the vessel. This results in a parison, a plot of the ratio between the CO2 and N2 working

Mol. Syst. Des. Eng. This journal is © The Royal Society of Chemistry 2019
 View Article Online

MSDE Paper

factors: higher temperature isotherms are less steep, which

improves the working capacity, and the nitrogen loading is
suppressed. Beyond 373 K, the CO2 purity reduces due to the
CO2 working capacity decreasing at a rate faster than the re-
duction in void volume. The improvement in recovery with
temperature is also due to the isotherms becoming less
steep. The minimum capture cost is seen at ≈373 K, i.e.
when the improved working capacity and recovery and the
corresponding reduction in capital costs offset the increased
Published on 23 September 2019. Downloaded by University of Toronto on 1/3/2020 1:27:15 AM.

operating costs to supply steam for feed heating. Beyond this

point, the increased heat exchanger area and steam utility re-
quired to heat the feed are further disadvantaged by reduc-
tions in working capacity and recovery – resulting in a higher
cost per tonne captured. The CO2 emissions from the genera-
tion of the steam should also be kept in mind. The optimum
temperature for UTSA-16 is seen at 313 K, i.e. the current op-
erating temperature. Changes in temperature either side will
reduce the CO2 working capacity, resulting in poorer purity,
recovery, and capture cost. For completeness, we present an
extended temperature range in the ESI† (Fig. S16 to S20) for
all adsorbents. Only results for purity and recovery are
shown, as the requisite allowances to calculate the cost under
cryogenic operation are not present in the model. In sum-
mary, tuning the adsorption temperature may allow to reach
higher performance levels, though in the present cases, the
outputs remain below the targets.
Fig. 3d shows the fraction of operating costs over total an-
nual costs as a function of desorption pressure. This fraction
reduces with decreasing desorption pressure and is solely
Fig. 2 CO2 (solid lines) and N2 (dashed lines) adsorption isotherms of
due to the increasing capital costs due to the vacuum require-
the two ‘extreme’ MOF adsorbents in terms of separation ments. A range of adsorbents, including Mg-MOF-74, displays
performance: Mg-MOF-74 (top, poorly performing MOF) and UTSA-16 a maxima, caused by vacuum pump capital costs. Higher de-
(bottom, highly performing MOF) at 273 K (red), 303 K (green), and 333 sorption pressures reduce the suction flow rate of the vac-
K (blue). uum pump selected. Consequently, a greater number of
pumps are needed to achieve the required flow rate. Data for
the other adsorbents is included in Fig. S21.†
capacities, also known as the working selectivity, is included Fig. 3e presents the influence of adsorbent cost. As other
in the ESI† (Fig. S9 to S12) for all adsorbents. It shows that significant capital costs contribute to the process, the cost of
most of the MOFs are hampered by poor selectivity. This the adsorbent does not become significant until 50–100 £
analysis highlights that one should look at enhancing CO2/N2 kg−1. In systems where adsorption is currently applied, e.g.
selectivity and decreasing the N2 uptake rather than maximiz- air separation and H2 production, the capital costs are signif-
ing CO2 uptake. icantly lower, so the cost of the adsorbent can impact the
separation cost. However, for post-combustion capture, the
current MOF cost target value71 of 10 US$ kg−1 does not have
Effects of process parameters to be reached to minimise the separation cost. It should be
Taking Mg-MOF-74 and USTA-16 as examples, we then inves- noted that this cost was proposed by the US DOE for CH4
tigated the effect of process and adsorbent parameters on storage in vehicles, however, it seems to have propagated
performance and capture cost, namely: adsorption tempera- through the community. We determined the cost of UTSA-16
ture, desorption pressure, adsorbent cost and adsorbent (196 £ kg−1, vertical line on the plot) and Mg-MOF-74 (3070 £
density. kg−1) using prices for bulk chemical supply from common
We first investigated the impact of adsorption tempera- scientific suppliers and yields reported in the literature. We
ture, i.e. the temperature the feed is adjusted to following calculated the cost of zeolite 13X (66 £ kg−1) following the
compression, on the purity, recovery and cost. The results are same method. Using the latter cost as an indicator of econo-
presented in Fig. 3a–c. Considering all three outputs, Mg- mies of scale to the industrial 13X price of 1.5 £ kg−1, this im-
MOF-74's performance is optimum between 358 and 373 K. plies that UTSA-16 might be produced at a cost of 5 £ kg−1.
CO2 purity increases with temperature up to 373 K due to two The largest limitation in the synthesis of UTSA-16 is the

This journal is © The Royal Society of Chemistry 2019 Mol. Syst. Des. Eng.
 View Article Online

Paper MSDE
Published on 23 September 2019. Downloaded by University of Toronto on 1/3/2020 1:27:15 AM.

Fig. 3 Evaluation of the effect of process parameters on separation performance and cost for Mg-MOF-74 (red) and UTSA-16 (green). Influence of
adsorption temperature on: (a) CO2 purity, (b) CO2 recovery, and (c) capture cost. (d) Influence of desorption pressure on annual fraction of oper-
ating costs, where ‘NG’ represents Cansolv. (e) Influence of adsorbent cost on capture cost. (f) Influence of adsorbent density on capture cost. All
for the NGCC scenario. All at 0.01 bara desorption pressure, except (d).

metal source, cobalt acetate, as all other raw materials are may be 1.5–2× this value. The default comparison to a range
readily available at scale. It may therefore be worth investigat- of stadia has also been included. Jest aside, reduction in the
ing the possibility of isoreticular chemistry for UTSA-16, and plant area is paramount in the successful implementation of
whether a more accessible metal source such as Zn2+ or Fe2+ adsorption for post-combustion capture. Efforts should be
is possible while maintaining similar adsorption properties. placed towards process intensification and improvements in
Quantitative results for the impact of adsorbent density mechanical vacuum generation technology.
on cost are shown in Fig. 3f. As addressed during the sensi-
tivity analysis, the cost increases as the adsorbent density de-
creases. The benefits of a dense adsorbent are seen for UTSA- Sensitivity analysis
16. For Mg-MOF-74, significant cost reductions are not neces- As previously mentioned, there may be some uncertainty in
sarily attainable even if pelletisation were used to increase the adsorbent physical properties as they are not
density. This is ultimately due to the poor CO2 recovery of
Mg-MOF-74. The recovery defines the total mass of CO2 that
is captured from the feed, and the working capacity defines
the amount of adsorbent required to capture that amount of
CO2. Hence, the poor recovery implies a small mass of adsor-
bent. As the density increases, the vessel volume required to
store that adsorbent reduces. A point is reached for Mg-MOF-
74 such that it is not possible to reduce the number or size
of the vessels further while still meeting the velocity and
pressure drop requirements.
Finally, we evaluated the impact of feed flow rate on the
approximate site area required for the plant. We conducted
this study in the context of the NGCC scenario and using
USTA-16 properties. As seen in Fig. 4, the 400 MWe case calls
for an area of ≈18 000 m2. This area only accounts for the
number of adsorption columns and vacuum pumps required,
while allowing a 0.5× spacing between the equipment items. Fig. 4 Area occupied by vacuum pumps and adsorption columns as a
Once all other equipment items such as compressors, heat function of power generation capacity (flue gas flow rate) for the
exchangers, pumps, and OBL plant is included, the total area NGCC scenario, using UTSA-16 as the adsorbent.

Mol. Syst. Des. Eng. This journal is © The Royal Society of Chemistry 2019
 View Article Online

MSDE Paper

experimentally measured or verified values but estimated versa. An increase in density also improves CO2 purity. The
ones. To investigate the implications of inaccuracies in bed largest impacts are observed at the lower desorption pres-
density, void fraction, and heat capacity, we carried out a sen- sure. Although a lower desorption pressure recovers more
sitivity analysis over a ±30% range for each property while CO2 from the adsorbent positively contributing to purity, the
keeping the others constant (Fig. 5 and S13 to S15†). Due to lower pressures also recover more of the approximately 96%
the interrelated nature of adsorption processes, we investi- N2 present in the void space.
gated the effect of the three physical property inputs for Unlike density and void space, variations in the heat ca-
both Mg-MOF-74 and UTSA-16 at two desorption pressures. pacity significantly affect the performance of Mg-MOF-74 at
Upon first observation, variations in adsorbent density and 0.01 bara. The effect is not as severe at higher desorption
Published on 23 September 2019. Downloaded by University of Toronto on 1/3/2020 1:27:15 AM.

void fraction do not have significant influence on any of the pressures, nor is it significant for UTSA-16. To explain this
three performance metrics, while heat capacity does to some observation, we compare the adsorbents and operating condi-
extent. tions. The first contribution is due to the difference in en-
We now examine these observations in greater detail. The thalpy of adsorption between the two adsorbents. Mg-MOF-74
density and void fraction both contribute to the total amount has an enthalpy of adsorption at zero loading for CO2 of at
of void space present in the bed for a given mass of adsor- least 47 kJ mol−1,17,72 whereas UTSA-16 is ≈32–35 kJ mol−1.73
bent. The void space impacts the dilution of the desorbed If the actual heat capacity is lower than the estimated value
product gas (CO2) by the feed gas occupying the void space. (negative x-values), the bed experiences a larger temperature
Therefore, in general, a larger void volume results in poorer swing for a given energy input/output. As these swings are al-
product purity. This is seen for all scenarios presented below, ready large for Mg-MOF-74, a further increase has even worse
a reduction in void fraction improves CO2 purity, and vice implications on the CO2 purity and recovery, reducing the

Fig. 5 Sensitivity analysis of adsorbent physical properties (density, void fraction, and heat capacity) on CO2 purity for Mg-MOF-74 (top) and
UTSA-16 (bottom) at 0.01 bara desorption (left) and 0.15 bara desorption (right).

This journal is © The Royal Society of Chemistry 2019 Mol. Syst. Des. Eng.
 View Article Online

Paper MSDE

working capacity even further. The reverse applies for the im- Optimising adsorbent properties
provements in purity and recovery seen in the right direction Having developed a reasonable understanding of the process
of the axis, a higher adsorbent heat capacity limits the tem- and the factors which influence process performance, we will
perature swing. The contribution of desorption pressure to now direct our attention to the adsorbents themselves. Our
this observation is related to the working capacity/amount goal here is to identify and quantify the adsorbents character-
desorbed. At higher desorption pressures, fewer moles of gas istics needed to push the performance and capture cost to-
desorb, thus requiring smaller thermal energy input from the wards the targets. This theoretical exercise, conducted on
surroundings to do so; therefore, variations in heat capacity Mg-MOF-74 and USTA-16, will provide directions to chemists
are less significant. As the amount desorbed is not a linear and materials scientists in the design and synthesis of highly
Published on 23 September 2019. Downloaded by University of Toronto on 1/3/2020 1:27:15 AM.

function of pressure (logarithmic isotherm shape), the influ- performing adsorbents. This has been attempted by others to
ence of heat capacity does not have a linear relationship with a certain extent,30,49,74,75 where hypothetical adsorbents (or
desorption pressure. isotherms) are proposed to optimise separation performance.
In some cases, these effects combine in a beneficial (and Aside from the impact on process economics, which the ma-
convoluted manner), such as the CO2 recovery for UTSA-16 at jority do not address, we propose some synthesis routes by
0.15 bara (Fig. S14†). The feed required to complete one cycle which the optimum properties could be obtained. We note
decreases due to the lower heat capacity of the adsorbent that limiting the analysis to the extreme cases of Mg-MOF-74
resulting in a greater temperature swing of the bed. However, and USTA-16 will not necessarily exclude any of the MOFs
a natural feedback loop exists where desorption becomes from the original list. Indeed, Mg-MOF-74 has large surface
more difficult as it proceeds, due to the concurrent reduction area and porosity and strong open metal centres, whereas
in temperature. As desorption is taking place at a higher UTSA-16 is a small pore adsorbent with low(er) porosity and
pressure, the working capacity of CO2 is not significantly re- more moderate enthalpy of adsorption. Thus, the majority of
duced as a balance is met between higher temperature the other adsorbents should be incorporated in that range.
change from reduced heat capacity, and lower temperature Adsorption isotherms provide a reflection of an adsorbent's
change due to higher desorption pressure. As CO2 recovery is physical and chemical properties. Assuming our ability to tune
defined as the ratio of CO2 product to total CO2 feed, this cul- the adsorption properties of the adsorbent in any way desired,
minates in higher CO2 recovery. This is not observed for all we can investigate the influence of various aspects of the ad-
adsorbents and is normally caused by the combination of sorbent/isotherm. In the model, we described the adsorption
heat of adsorption, and the relative amounts of gas removed isotherms by the dual-site Langmuir isotherm (eqn (9)).
from the void space and from the adsorbent. This is seen in  
ΔH j
2 mj ·b0;j · exp
X R·T ·P
the effect of void fraction on CO2 recovery for UTSA-16 at
0.15 bara desorption. A lower void fraction results in lower n DSL
¼   (9)
ΔH j
j¼1 1 þ b0;j · exp R·T ·P
CO2 recovery, which suggests that a reasonable portion of the
‘working capacity’ is coming from the feed CO2 occupying
the void space, rather than what is collected from the where mj is the saturation loading for site j, b0, j is the pre-
adsorbent. exponential factor of the Langmuir constant for site j, ΔHj is
Considering now capture cost, lower densities increase the enthalpy of adsorption for site j, and P is the partial pres-
costs due to: (i) the larger size of the vessels needed to ac- sure of the gas species of interest. The dual-site Langmuir
commodate the mass of adsorbent, and (ii) the greater isotherm is based on two distinct, homogenous, adsorption
amount of vacuum capacity required. In general, a lower void sites.
fraction reduces capital costs since the adsorbent occupies a First, we artificially increased the CO2 capacity by multi-
greater portion of the adsorption vessel volume and requires plying m (eqn (9)) by a factor and we quantified the impact
reduced vacuum capacity. The UTSA-16 result for 0.15 bara on the CO2 purity, recovery and capture cost (Fig. 6a–c). This
desorption represents an exception. Due to the aforemen- exercise somewhat equals to increasing the selectivity of the
tioned reduction in recovery, a greater mass of adsorbent is adsorbent by increasing CO2 capacity. In nearly every in-
required to capture the given feed flow rate, resulting in stance, increasing the CO2 capacity of each adsorbent even
higher capital costs for vessels and adsorbent. by a factor of 20 has negligible impact. UTSA-16's CO2 purity
Overall, variations in the adsorbent density and void frac- slightly improves up to a factor of 4, due to an increased
tion do not influence significantly the performance and costs working capacity. Beyond that, thermal effects offset the in-
of the capture process. Heat capacity, on the other hand, may crease in CO2 capacity. Improving the working capacity re-
have an impact for adsorbents with high enthalpies of ad- sults in less adsorbent mass requirements and thereby a drop
sorption, such as Mg-MOF-74. We note that these implica- in cost. However, this effect remains small and restricted to
tions remain relative. For instance, although an improvement low multiplication factor.
of ≈46% can be seen in the CO2 purity at 0.01 bara if the heat Next, we investigated the implications of reducing N2 ca-
capacity were underestimated by 30%, the ultimate outcome pacity (Fig. 6d–f). This was achieved by dividing the corre-
of this is that the CO2 purity increases from 23.7% mol to sponding m parameters in eqn (9) by a factor. As before, this
34.5% mol, which remains low. can be thought of improving the selectivity by reducing N2

Mol. Syst. Des. Eng. This journal is © The Royal Society of Chemistry 2019
 View Article Online

MSDE Paper
Published on 23 September 2019. Downloaded by University of Toronto on 1/3/2020 1:27:15 AM.

Fig. 6 Influence of increasing CO2 uptake (top, a–c), decreasing N2 adsorption (middle, d–f), and enthalpy of adsorption (bottom, g–i) for Mg-
MOF-74 (red) and UTSA-16 (green) on CO2 purity (left), CO2 recovery (middle), and capture cost (right) for the NGCC scenario.

adsorption. This area has received little to no attention in the is not seen for CO2 recovery due to significantly different N2
literature. We observe significant gains for the three outputs adsorption characteristics of the baseline materials.
for the two MOFs, though the extent of improvement is less Finally, we assessed the effects of scaling the enthalpy of
pronounced for UTSA-16. CO2 purity improves due to a reduc- adsorption by varying ΔH (eqn (9)) for both CO2 and N2 by
tion in co-adsorption. The relative gain in CO2 purity for the same factor. We present the results in Fig. 6g–i. For CO2
UTSA-16 is not as large since the baseline sample does not purity, the maximum for Mg-MOF-74 is at ≈0.858×, and for
adsorb significant amounts of N2. Reducing N2 adsorption UTSA-16 is at ≈1.01×. Taking the ΔH ads values referred to ear-
improves CO2 recovery for Mg-MOF-74. The N2 working ca- lier, the optimum values become ≈40 kJ mol−1 for CO2 on
pacity on Mg-MOF-74 is rather high since Mg-MOF-74 ad- Mg-MOF-74, and the baseline value of ≈32–35 kJ mol−1 for
sorbs a significant amount of N2 considering the partial pres- UTSA-16. The overall trend in CO2 purity as a function of en-
sure of N2 for this application (Fig. 2). In fact, the working thalpy of adsorption is due to the trade-off between two sce-
capacity of N2 is higher than CO2, as indicated by the work- narios. At lower ΔH, the isotherm is insufficiently steep to ad-
ing selectivity values less than unity (Fig. S9 to S12†). The en- sorb a useful amount at these low partial pressures, while at
ergetics of N2 adsorption and desorption result in bed tem- higher ΔH, the isotherm becomes too steep and working ca-
perature swings. Reducing the portion of the temperature pacity suffers. The optimum value would therefore shift to-
swing attributed to N2 adsorption improves the CO2 working wards lower ΔH for higher feed concentrations with all other
capacity, and thus the recovery. This factor also enhances factors being equal. As the CO2 recovery is a function of total
CO2 purity. The improvements in recovery for Mg-MOF-74 feed amount, the influence of ΔH is multifaceted. A change
culminate in significant capture cost reductions due to the in ΔH influences the amount adsorbed at a given partial pres-
smaller adsorbent mass required. For UTSA-16, CO2 purity sure. Lower ΔH reduces the total amount adsorbed, thus the
improves for the same reasons mentioned above. The effect recovery is not impacted as dramatically as purity. At higher

This journal is © The Royal Society of Chemistry 2019 Mol. Syst. Des. Eng.
 View Article Online

Paper MSDE

ΔH, the total amount adsorbed increases, while the amount maximum performance for the given scenarios? In response,
that can be recovered decreases, resulting in a sharp decline we applied the model to generate a dataset by sweeping over
in recovery. The minima in cost align with the maxima in re- a range of isotherm input parameters. Similar work has been
covery as a given mass of adsorbent is more productive, re- previously undertaken,30,74 however, we assess a range of flue
quiring less infrastructure. At lower ΔH, the capture cost in- gas sources over a wider range of isotherms, as well as con-
creases due to poor working capacity attributed to the low sidering capture cost. To reduce computation and data pro-
amount adsorbed, requiring more adsorbent and vessels. At cessing time, we used a single-site Langmuir isotherm (eqn
higher ΔH, poor working capacity arising from severe thermal (10)) in place of the dual-site Langmuir isotherm (eqn (9)),
effects increases cost. thereby reducing the number of isotherm input parameters
Published on 23 September 2019. Downloaded by University of Toronto on 1/3/2020 1:27:15 AM.

Having assessed the effect of tuning various adsorbent from 12 to 6. To further reduce the number of input parame-
properties, the question becomes: is this form of omnipotent ters, we made assumptions making use of the physical inter-
control even possible? To some extent it may be. Increasing pretation of the Langmuir isotherm model. Firstly, the satu-
CO2 adsorption capacity while operating in a physisorption ration capacity, m, is made equal for both nCO2 and nN2 since
regime would likely be attainable by increasing adsorbent the number of adsorption sites is constant for a given adsor-
surface area. The drawback to this is that N2 adsorption ca- bent. The parameter b0 is related to the entropy of adsorption
pacity would correspondingly increase due to the non- and assuming that the adsorbed phase is liquid-like allows
selective nature of raw surface area. Selectively decreasing N2 the application of Trouton's rule,88 fixing the value of b0 to
adsorption may be possible by reducing pore aperture, creat- e −10.5 for both CO2 and N2. These assumptions allow the ad-
ing a molecular sieving style effect. Small pore MOFs with sorption of CO2 and N2 to be described by three input param-
lower surface areas are uncommon and is an area which eters: mmax, ΔHCO2, and ΔHN2. Therefore, in effect, the amount
could benefit from additional research. Control over the ΔH adsorbed at a given pressure is defined by the value of ΔH.
may be possible for both MOFs investigated here. Consider- From the Langmuir isotherm theory, this is because a higher
ing that the unsaturated metal sites of Mg-MOF-74 contribute ΔH means a larger energy barrier must be overcome for the
most to the high adsorption enthalpy, two potential avenues adsorbate to return to the gas phase. We swept the parame-
arise. The first avenue is using other metal analogues of M- ters over the following range, 1 ≤ mmax ≤ 16 mol kg−1, 24 ≤
MOF-74 to modulate the M–CO2 interactions and steric ef- ΔHCO2 ≤ 60 kJ mol−1, and 11 ≤ ΔHN2 ≤ 20 kJ mol−1, in steps
fects. This has been reported by Queen et al.76 and based on of 1, 2, and 1, respectively. We fixed the adsorbent physical
this study, Co-, Fe-, Mn-, and Zn-MOF-74 may be more suit- properties throughout this process to the values of UTSA-16:
able than Mg-MOF-74. Partially metal substituted forms have bed density = 787 kg m−3, total void fraction = 0.61, and heat
also been developed,77 and such techniques may yield finer capacity = 878 J kg−1 K−1. Including the variation of these pa-
control. The second avenue is the grafting of organic mole- rameters over a useful range would exponentially increase
cules to the unsaturated metal sites.78–80 The approach could the number of data points and make data presentation un-
be applied to control the enthalpy of adsorption by using sightly. It may also be an unnecessary complication given the
mono or single-ended amines such as ethylamine for the sensitivity analysis on physical properties discussed around
most significant reduction, and functionalised forms such as Fig. 3. The desorption pressure was fixed at 0.01 bara for all
aminoethanethiol, or ethanolamine for intermediate cases. cases.
An uncommon feature of UTSA-16 is the presence of charge-
ΔH 
balancing cations (K+) in the structure. A family of MOFs mmax ·b0 · exp R·T ·P
nSSL ¼
ΔH  (10)
known as ZMOFs (zeolite like MOFs) also demonstrates this 1 þ b0 · exp R·T ·P
feature.81,82 This feature could be exploited to influence the
adsorbate–adsorbent interactions as is common in Although the use of the single-site Langmuir isotherm
zeolites,83–86 for instance via exchange with Li+ ions.87 Li+ ze- means that some measured type-I adsorption isotherms can-
olites exhibit the strongest adsorbate–adsorbent interactions not be fit very accurately, our aim is to uncover the relation-
due to the high charge density of the Li+ ion. A complete ion ship between the isotherm characteristics and process perfor-
exchange would increase the heat of adsorption of UTSA-16. mance. For reference, the values of the three parameters for
Whether this has the counter-effect of increasing the accessi- Mg-MOF-74 and UTSA-16 are provided in Table 2. Mg-MOF-
ble pore size and N2 adsorption remains to be seen. The au- 74's CO2 isotherms cannot be accurately fit with the single-
thors also found that the Cs+ form displayed the highest en- site model, however, the range between 0 ≤ P ≤ 0.40 bara
thalpy of adsorption, which is contrary to the order displayed and T ≥ 303 K displays good fitting performance. This is for-
in zeolites. This area seems to warrant further investigation tunate as it matches the range relevant to this work.
and rigorous gas adsorption studies. The outputs for CO2 purity, recovery, and capture cost are
shown in Fig. 7 for all four scenarios. We first discuss CO2
purity. The regions of good performance (red & grey) increase
The ideal isotherm in size as the feed concentration increases indicating an
The previous section naturally leads to the question: what easier separation. Taking the 11 kJ mol−1 plane, the region
combination of CO2 and N2 isotherms are required to achieve defined by the yellow-green transition around ΔHCO2 of 48 kJ

Mol. Syst. Des. Eng. This journal is © The Royal Society of Chemistry 2019
 View Article Online

MSDE Paper

Table 2 Reduced single-site Langmuir isotherm fitting parameters for the cement and steel scenarios, the dark blue–violet–black re-
Mg-MOF-74 and UTSA-16
gions exhibit lower CO2 purities than the feed concentration.
Parameter Mg-MOF-74 UTSA-16 This stems from high N2 adsorption combined with poor
−1 CO2 working capacity.
mmax [mol kg ] 7.54 4.16
ΔHCO2 [kJ mol−1] 34.6 30.4 We now look at the variation in CO2 recovery, i.e. the ratio
ΔHN2 [kJ mol−1] 21.3 17.8 between the CO2 obtained in the product relative to the total
CO2 fed to the process. At the lower CO2 concentrations,
there are fewer regions of good recovery. Additionally, there
mol−1 remains mostly unchanged when increasing CO2 con- is an appreciable amount of N2 adsorption due to its compar-
Published on 23 September 2019. Downloaded by University of Toronto on 1/3/2020 1:27:15 AM.

centration. This represents the limitation imposed by ther- atively high partial pressure. The ad- and desorption of CO2
mal effects of the bed and regenerability. Even if the N2 ad- is not the sole contributor to the thermal effects of the bed,
sorption is negligible, a steep CO2 isotherm will hamper all species do. High, or increased, adsorption of N2 leads to
performance. The regions in the lower left (low ΔHCO2, high worse recovery of CO2 as its desorption is hampering CO2 de-
mmax) also increase in size with increasing CO2 concentra- sorption. This is reflected in two ways, any vertical line
tion. In the regions of low ΔHCO2, a high capacity (mmax) is re- through the planes of ΔHN2 has reducing CO2 recovery as N2
quired in order to have an appreciable adsorption and work- adsorption increases, and a given ΔHN2 plane has reducing
ing capacity at low feed concentrations. Hence, as the feed CO2 recovery as mmax increases.
concentration increases, adsorbents with lower capacity also The capture cost contours highlight the areas of best CO2
become feasible alternatives. For the 17 and 20 kJ mol−1 working capacity and recovery. Cross-scenario comparisons
planes, the good performance regions increase noticeably in are not straightforward due to the differences in feed flow
size with feed CO2 concentration. This increased allowance rate. The regions of lowest cost (black) increase in size with
for N2 adsorption from a pure-component isotherm perspec- feed CO2 concentration, somewhat expected due to the in-
tive stems from the lower N2 partial pressure in the feed. For creasing ease of separation, driven mostly by improvements

Fig. 7 Contour plots displaying the influence of CO2 and N2 isotherms on CO2 purity and recovery, and capture cost for the four post-
combustion scenarios investigated. Only selected ΔHN2 layers are shown for greater clarity.

This journal is © The Royal Society of Chemistry 2019 Mol. Syst. Des. Eng.
 View Article Online

Paper MSDE

in CO2 working capacity. If CO2 working capacity did not purity for a given adsorbent, as discussed earlier. Zeolite 13X
have a significant effect, the regions of lowest cost would for the coal scenario at 0.01 bara desorption returned an 84%
align with those of highest CO2 recovery. In this model, as mol CO2 purity. Separate studies showed that purities ≥95%
CO2 purity is proportional to CO2 working capacity, we ob- mol and recoveries ≥90% mol are achievable using 13X.60,61
serve similarities between the regions of lowest cost and Hence, using zeolite 13X as a reference, we hypothesise that
highest purity (red & grey) for a given scenario. isotherms with CO2 purity values ≥85% mol could achieve
We now return to our initial aim to define the ideal iso- the purity and recovery targets in a more complex cycle. The
therm. In the framework of the single-site Langmuir iso- validity of this assumption depends on the concentration of
therm, the parameter mmax would be proportional to the spe- CO2 in the feed gas. At feed concentrations equal to or
Published on 23 September 2019. Downloaded by University of Toronto on 1/3/2020 1:27:15 AM.

cific surface area (pore size and porosity), and the ΔH greater than for the coal combustion case (12.5% mol), there
parameters to the respective adsorbate–adsorbent interac- is little concern. At lower feed concentrations, it becomes
tions (surface chemistry). It becomes possible to determine more difficult to achieve the purity and recovery targets even
whether there is an isotherm type, and hence material prop- with a more complex cycle. We could not locate experimental
erties, that are most suited for these separations. The utility work or rigorous simulations for a PVSA process for post-
of this exercise is to guide material development and enable combustion capture from natural gas. Therefore, we cannot
a binary (yes/no) initial assessment of an adsorbent with min- quantitatively describe the uncertainty, or make an equiva-
imal experimental input. Taking the natural gas scenario as lent comparison for the natural gas scenario. In the absence
an example, we generated a plot of all the isotherms achiev- of a sounder reference, we assume that adsorbents or iso-
ing a CO2 purity of 85% mol (Fig. 8). We observe no trend or therms which display CO2 purity results ≥85% mol will
commonality between the CO2 isotherms that result in good achieve the required purity and recovery for CCS applications,
process performance. Therefore, we cannot simply recom- albeit at a higher cost.
mend a particular isotherm shape (and corresponding molec-
ular scale properties), or CO2 capacity, which yields good per-
formance, further highlighting the limitations of using CO2 Establishing a surrogate model
adsorption capacity as a performance metric. It could be said Considering our previous analysis, we developed a surrogate
in general that, given the N2 loading of UTSA-16 at 273.15 K model to simplify the evaluation of adsorbents. A surrogate
and 1 bara is 0.28 mol kg−1, N2 adsorption greater than this model (i.e. black-box model) consists of one or more alge-
would render the adsorbent unsuitable for post-combustion braic expressions fit to data, which enables regression of that
capture using PVSA. For these styles of separations (i.e. post- data. The data generated in the previous section serves as the
combustion capture using adsorption), the co-adsorption of input, or training, data for the model. The data was fit using
N2 has the greatest influence on process performance. Conse- ALAMO.89 ALAMO generates ‘data-driven’ surrogate models
quently, efforts should be directed towards minimizing N2 by combining multiple function types to fit the data using its
adsorption and this is in-line with previous work.30,74 solver, BARON. This surrogate model is meant to be used by
Prima facie, for the NGCC scenario, one cannot achieve researchers to quickly evaluate the performance of their ad-
CO2 purities ≥95% mol within the bounds investigated. How- sorbents. We supply it here as ESI.† An example of the
ever, the choice of adsorption cycle impacts the attainable goodness-of-fit for the natural gas scenario is shown in Fig. 9
by means of parity plots for three output variables. The com-
plete results of the statistical analysis for all the scenarios are
provided in the ESI† (Fig. S22 to S33). Given the robust
goodness-of-fit of the surrogate model, we are confident that
any adsorbent that is within the evaluated/swept bounds will
be represented accurately.
A surrogate model advantageously provides reasonable es-
timate of an adsorbent's performance, without the need to
implement the complete model. To make use of the surro-
gate model, the user must measure CO2 and N2 isotherms at
three temperatures and fit to the single-site Langmuir iso-
therm (eqn (10)). We note that this equation may not per-
fectly fit the measured isotherm data (usually CO2 data).
However, as discussed earlier, as long as the data points in
the pressure and temperature range of interest can be fit ac-
curately, there are no ill-effects. While the number of iso-
therms needed may seem large, they are key to obtain a reli-
Fig. 8 Representation of CO2 isotherms that correspond to a CO2
purity ≥85% mol for the NGCC scenario. The red line indicates the
able evaluation of any adsorbent so the thermal effects can
‘average’ isotherm, with the shaded region showing the 99% be elucidated. The other concern may be the adsorbent physi-
confidence interval. Calculated for 273.15 K. cal properties, as mentioned the data points are generated

Mol. Syst. Des. Eng. This journal is © The Royal Society of Chemistry 2019
 View Article Online

MSDE Paper
Published on 23 September 2019. Downloaded by University of Toronto on 1/3/2020 1:27:15 AM.

Fig. 9 Parity plots of surrogate model fits for the natural-gas scenario data points to indicate the goodness-of-fit.

using the physical properties for UTSA-16. We addressed the isotherm has little capacity at low CO2 partial pressures such
implications of this assumption in the sensitivity analysis as those experienced in post-combustion capture, and it can-
(Fig. 5). For an initial evaluation to gauge whether an adsor- not be too high such that the working capacity and regene-
bent warrants detailed investigation, the repercussions are rability is hampered.
not severe. Finally, we have attempted to design an ‘ideal adsorbent’
by numerically screening all isotherms defined in a given
Conclusions range. Ultimately, it was not possible to define a single iso-
therm, or a group of similar isotherms, that meet the CO2 pu-
We have screened 22 MOFs along with two zeolites and an rity and recovery requirements. However, we have used the
activated carbon for CO2/N2 separation in four different CO2 data to develop a surrogate model and allow readers to
capture scenarios, i.e.: NGCC, pulverised coal power plant, ce- swiftly evaluate their adsorbents for the four scenarios inves-
ment plant and integrated steel mill. The screening ‘metrics’ tigated without needing to implement the full model. This
involved CO2 purity, CO2 recovery and capture cost (OPEX model requires users to input CO2 and N2 isotherms of their
and CAPEX). Of the 22 MOFs evaluated, UTSA-16 performs materials at 3 different temperatures each, in order to deter-
the best. Its low N2 uptake and less steep isotherms prevent mine CO2 purity and recovery.
product dilution by co-adsorbed N2 and promote regene- The natural progression from here is to consider TSA
rability. UTSA-16's performance aligns with that of zeolite given the technology advancements in the area, and the de-
13X. Mg-MOF-74 represents the worst choice of adsorbent for velopment of an associated model is underway to understand
the targeted separation over all scenarios, due to its high ad- and evaluate TSA processes for post-combustion capture.
sorption enthalpy and high N2 adsorption capacity. In our
model, the performance displayed by UTSA remains below Conflicts of interest
the targets in terms of purity and recovery, yet cycle optimisa-
tion would likely allow reaching these targets. Nonetheless, There are no conflicts to declare.
the process footprint requirements – at least 18 000 m2 is re-
quired for the natural gas scenario – represent an area of Acknowledgements
concern, imposing limitations on, or discounts, retrofit possi- We thank Mr. Wouter Arts for his efforts in surveying the lit-
bilities. In addition, the cost of the separation remains gener- erature for available isotherm data. We are grateful to the
ally high compared to the use of amine-based solvents. Inter- EPSRC for providing the funding to undertake this research
estingly, we have found that the cost of the adsorbent via the UK Carbon Capture and Storage Research Centre
significantly impacts that of the process only if it is above 50 (UKCCSRC, grant EP/P026214/1).
£ kg−1. A sensitivity analysis on adsorbent physical properties
was also carried out to understand the impacts of uncertainty References
on these parameters. In general, uncertainties in heat capac-
ity have the greatest influence on results, however, the mag- 1 K. Sumida, S. Horike, S. S. Kaye, Z. R. Herm, W. L. Queen,
nitude is variable depending on the situation. C. M. Brown, F. Grandjean, G. J. Long, A. Dailly and J. R.
In the quest to identify optimum adsorbent characteristics Long, Hydrogen storage and carbon dioxide capture in an
that give best performance for CO2 purity and recovery, and iron-based sodalite-type metal–organic framework (Fe-BTT)
capture cost, we have ‘optimised’ UTSA-16 and Mg-MOF-74. discovered via high-throughput methods, Chem. Sci.,
Reducing N2 adsorption has the greatest impact in increasing 2010, 1, 184.
CO2 purity and recovery, and reducing capture cost. This 2 P. Nugent, V. Rhodus, T. Pham, B. Tudor, K. Forrest, L.
should be a key area of research in materials development in Wojtas, B. Space and M. Zaworotko, Enhancement of CO2
the future. In addition, we have shown that an optimum heat selectivity in a pillared pcu MOM platform through pillar
of adsorption exists. Values cannot be too low such that the substitution, Chem. Commun., 2013, 49, 1606.

This journal is © The Royal Society of Chemistry 2019 Mol. Syst. Des. Eng.
 View Article Online

Paper MSDE

3 Q.-Q. Dang, C.-Y. Liu, X.-M. Wang and X.-M. Zhang, Novel Coordination Polymer with Cylindrical Pores, J. Am. Chem.
Covalent Triazine Framework for High-Performance CO2 Soc., 2008, 130, 10870–10871.
Capture and Alkyne Carboxylation Reaction, ACS Appl. 18 Z. Zhao, Z. Li and Y. S. Lin, Adsorption and Diffusion of
Mater. Interfaces, 2018, 10, 27972–27978. Carbon Dioxide on Metal−Organic Framework (MOF-5), Ind.
4 V. Guillerm, Ł. J. Weseliński, M. Alkordi, M. I. H. Mohideen, Eng. Chem. Res., 2009, 48, 10015–10020.
Y. Belmabkhout, A. J. Cairns and M. Eddaoudi, Porous 19 A. O. Yazaydın, R. Q. Snurr, T.-H. Park, K. Koh, J. Liu, M. D.
organic polymers with anchored aldehydes: a new platform LeVan, A. I. Benin, P. Jakubczak, M. Lanuza, D. B. Galloway,
for post-synthetic amine functionalization en route for en- J. J. Low and R. R. Willis, Screening of Metal−Organic
hanced CO2 adsorption properties, Chem. Commun., Frameworks for Carbon Dioxide Capture from Flue Gas
Published on 23 September 2019. Downloaded by University of Toronto on 1/3/2020 1:27:15 AM.

2014, 50, 1937. Using a Combined Experimental and Modeling Approach,

5 W. Lu, D. Yuan, J. Sculley, D. Zhao, R. Krishna and H.-C. J. Am. Chem. Soc., 2009, 131, 18198–18199.
Zhou, Sulfonate-Grafted Porous Polymer Networks for 20 E. S. Kikkinides, R. T. Yang and S. H. Cho, Concentration
Preferential CO2 Adsorption at Low Pressure, J. Am. Chem. and recovery of carbon dioxide from flue gas by pressure
Soc., 2011, 133, 18126–18129. swing adsorption, Ind. Eng. Chem. Res., 1993, 32, 2714–2720.
6 Y. Zeng, R. Zou and Y. Zhao, Covalent Organic Frameworks 21 K. T. Chue, J. N. Kim, Y. J. Yoo, S. H. Cho and R. T. Yang,
for CO2 Capture, Adv. Mater., 2016, 28, 2855–2873. Comparison of Activated Carbon and Zeolite 13X for CO2
7 D.-H. Park, K. S. Lakhi, K. Ramadass, M.-K. Kim, S. N. Recovery from Flue Gas by Pressure Swing Adsorption, Ind.
Talapaneni, S. Joseph, U. Ravon, K. Al-Bahily and A. Vinu, Eng. Chem. Res., 1995, 34, 591–598.
Energy Efficient Synthesis of Ordered Mesoporous Carbon 22 W.-K. Choi, T.-I. Kwon, Y.-K. Yeo, H. Lee, H. K. Song and
Nitrides with a High Nitrogen Content and Enhanced CO2 B.-K. Na, Optimal operation of the pressure swing
Capture Capacity, Chem. - Eur. J., 2017, 23, 10753–10757. adsorption (PSA) process for CO2 recovery, Korean J. Chem.
8 A. Alabadi, S. Razzaque, Y. Yang, S. Chen and B. Tan, Highly Eng., 2003, 20, 617–623.
porous activated carbon materials from carbonized biomass 23 S. P. Reynolds, A. D. Ebner and J. A. Ritter, Stripping PSA
with high CO2 capturing capacity, Chem. Eng. J., 2015, 281, Cycles for CO2 Recovery from Flue Gas at High Temperature
606–612. Using a Hydrotalcite-Like Adsorbent, Ind. Eng. Chem. Res.,
9 J. Chen, J. Yang, G. Hu, X. Hu, Z. Li, S. Shen, M. Radosz and 2006, 45, 4278–4294.
M. Fan, Enhanced CO2 Capture Capacity of Nitrogen-Doped 24 P. Xiao, J. Zhang, P. Webley, G. Li, R. Singh and R. Todd,
Biomass-Derived Porous Carbons, ACS Sustainable Chem. Capture of CO2 from flue gas streams with zeolite 13X by
Eng., 2016, 4, 1439–1445. vacuum-pressure swing adsorption, Adsorption, 2008, 14,
10 R. Krishna and J. M. van Baten, A comparison of the CO2 575–582.
capture characteristics of zeolites and metal–organic 25 J. A. Delgado, M. A. Uguina, J. L. Sotelo, V. I. Águeda, A. Sanz
frameworks, Sep. Purif. Technol., 2012, 87, 120–126. and P. Gómez, Numerical analysis of CO2 concentration and
11 P. Harlick and F. Tezel, Equilibrium Analysis of Cyclic recovery from flue gas by a novel vacuum swing adsorption
Adsorption Processes: CO2 Working Capacities with NaY, cycle, Comput. Chem. Eng., 2011, 35, 1010–1019.
Sep. Sci. Technol., 2005, 40, 2569–2591. 26 Z. Liu, C. A. Grande, P. Li, J. Yu and A. E. Rodrigues, Multi-
12 D.-X. Xue, A. J. Cairns, Y. Belmabkhout, L. Wojtas, Y. Liu, bed Vacuum Pressure Swing Adsorption for carbon dioxide
M. H. Alkordi and M. Eddaoudi, Tunable Rare-Earth fcu- capture from flue gas, Sep. Purif. Technol., 2011, 81, 307–317.
MOFs: A Platform for Systematic Enhancement of CO2 Ad- 27 R. Haghpanah, A. Majumder, R. Nilam, A. Rajendran, S.
sorption Energetics and Uptake, J. Am. Chem. Soc., Farooq, I. A. Karimi and M. Amanullah, Multiobjective
2013, 135, 7660–7667. Optimization of a Four-Step Adsorption Process for
13 A. Phan, C. J. Doonan, F. J. Uribe-Romo, C. B. Knobler, M. Postcombustion CO2 Capture Via Finite Volume Simulation,
O'Keeffe and O. M. Yaghi, Synthesis, Structure, and Carbon Ind. Eng. Chem. Res., 2013, 52, 4249–4265.
Dioxide Capture Properties of Zeolitic Imidazolate 28 Y. Shen, Y. Zhou, D. Li, Q. Fu, D. Zhang and P. Na, Dual-
Frameworks, Acc. Chem. Res., 2010, 43, 58–67. reflux pressure swing adsorption process for carbon dioxide
14 S. Salehi and M. Anbia, High CO2 Adsorption Capacity and capture from dry flue gas, Int. J. Greenhouse Gas Control,
CO2/CH4 Selectivity by Nanocomposites of MOF-199, Energy 2017, 65, 55–64.
Fuels, 2017, 31, 5376–5384. 29 G. N. Nikolaidis, E. S. Kikkinides and M. C. Georgiadis, An
15 A. Demessence, D. M. D'Alessandro, M. L. Foo and J. R. Integrated Two-Stage P/VSA Process for Postcombustion CO2
Long, Strong CO2 Binding in a Water-Stable, Triazolate- Capture Using Combinations of Adsorbents Zeolite 13X and
Bridged Metal−Organic Framework Functionalized with Mg-MOF-74, Ind. Eng. Chem. Res., 2017, 56, 974–988.
Ethylenediamine, J. Am. Chem. Soc., 2009, 131, 8784–8786. 30 B. J. Maring and P. A. Webley, A new simplified pressure/
16 J. Liu, P. K. Thallapally, B. P. McGrail, D. R. Brown and J. vacuum swing adsorption model for rapid adsorbent
Liu, Progress in adsorption-based CO2 capture by metal–or- screening for CO2 capture applications, Int. J. Greenhouse
ganic frameworks, Chem. Soc. Rev., 2012, 41, 2308–2322. Gas Control, 2013, 15, 16–31.
17 S. R. Caskey, A. G. Wong-Foy and A. J. Matzger, Dramatic 31 G. D. Pirngruber, L. Hamon, S. Bourrelly, P. L. Llewellyn, E.
Tuning of Carbon Dioxide Uptake via Metal Substitution in a Lenoir, V. Guillerm, C. Serre and T. Devic, A Method for

Mol. Syst. Des. Eng. This journal is © The Royal Society of Chemistry 2019
 View Article Online

MSDE Paper

Screening the Potential of MOFs as CO2 Adsorbents in 46 D. J. Brennan, Process Industry Economics: An International
Pressure Swing Adsorption Processes, ChemSusChem, Perspective, IChemE, Rugby, UK, 1997.
2012, 5, 762–776. 47 D. J. Brennan and K. A. Golonka, New Factors for Capital
32 L. Joss, M. Gazzani, M. Hefti, D. Marx and M. Mazzotti, Cost Estimation in Evolving Process Designs, Chem. Eng.
Temperature Swing Adsorption for the Recovery of the Heavy Res. Des., 2002, 80, 579–586.
Component: An Equilibrium-Based Shortcut Model, Ind. Eng. 48 P.-E. Just, Advances in the development of CO2 capture
Chem. Res., 2015, 54, 3027–3038. solvents, Energy Procedia, 2013, 37, 314–324.
33 M. Hefti, L. Joss, Z. Bjelobrk and M. Mazzotti, On the 49 M. T. Ho, G. W. Allinson and D. E. Wiley, Reducing the Cost
potential of phase-change adsorbents for CO2 capture by of CO2 Capture from Flue Gases Using Pressure Swing
Published on 23 September 2019. Downloaded by University of Toronto on 1/3/2020 1:27:15 AM.

temperature swing adsorption, Faraday Discuss., 2016, 192, Adsorption, Ind. Eng. Chem. Res., 2008, 47, 4883–4890.
153–179. 50 N. Susarla, R. Haghpanah, I. A. Karimi, S. Farooq, A.
34 V. Subramanian Balashankar, A. K. Rajagopalan, R. de Pauw, Rajendran, L. S. C. Tan and J. S. T. Lim, Energy and cost
A. M. Avila and A. Rajendran, Analysis of a Batch Adsorber estimates for capturing CO2 from a dry flue gas using
Analogue for Rapid Screening of Adsorbents for Postcombustion pressure/vacuum swing adsorption, Chem. Eng. Res. Des.,
CO2 Capture, Ind. Eng. Chem. Res., 2019, 58, 3314–3328. 2015, 102, 354–367.
35 P. A. Webley, A. Qader, A. Ntiamoah, J. Ling, P. Xiao and Y. 51 D. E. Garrett, Chemical Engineering Economics, van Nostrand
Zhai, A New Multi-bed Vacuum Swing Adsorption Cycle for Reinhold, New York, 1989.
CO2 Capture from Flue Gas Streams, Energy Procedia, 52 M. R. M. Abu-Zahra, J. P. M. Niederer, P. H. M. Feron and
2017, 114, 2467–2480. G. F. Versteeg, CO2 capture from power plants, Int. J.
36 H. Wu, W. Zhou and T. Yildirim, High-capacity methane Greenhouse Gas Control, 2007, 1, 135–142.
storage in metal-organic frameworks M2(dhtp): the impor- 53 A. Raksajati, M. T. Ho and D. E. Wiley, Reducing the Cost of
tant role of open metal sites, J. Am. Chem. Soc., 2009, 131, CO2 Capture from Flue Gases Using Aqueous Chemical
4995–5000. Absorption, Ind. Eng. Chem. Res., 2013, 52, 16887–16901.
37 in CRC Handbook of Chemistry and Physics, ed. J. R. Rumble, 54 M. Vaccarelli, R. Carapellucci and L. Giordano, Energy and
CRC Press/Taylor & Francis, Boca Raton, FL, 100th edn Economic Analysis of the CO2 Capture from Flue Gas of
(Internet Version 2019), 2019. Combined Cycle Power Plants, Energy Procedia, 2014, 45,
38 B. T. Goodman, W. V. Wilding, J. L. Oscarson and R. L. 1165–1174.
Rowley, Use of the DIPPR Database for Development of 55 G. Manzolini, E. Sanchez Fernandez, S. Rezvani, E. Macchi,
Quantitative Structure−Property Relationship Correlations: E. L. V. Goetheer and T. J. H. Vlugt, Economic assessment of
Heat Capacity of Solid Organic Compounds †, J. Chem. Eng. novel amine based CO2 capture technologies integrated in
Data, 2004, 49, 24–31. power plants based on European Benchmarking Task Force
39 J. Park, J. D. Howe and D. S. Sholl, How Reproducible Are methodology, Appl. Energy, 2015, 138, 546–558.
Isotherm Measurements in Metal–Organic Frameworks?, 56 K. Li, W. Leigh and P. Feron, H. Yu and M. Tade, Systematic
Chem. Mater., 2017, 29, 10487–10495. study of aqueous monoethanolamine (MEA)-based CO2
40 R. Han, K. S. Walton and D. S. Sholl, Does Chemical capture process: Techno-economic assessment of the MEA pro-
Engineering Research Have a Reproducibility Problem?, cess and its improvements, Appl. Energy, 2016, 165, 648–659.
Annu. Rev. Chem. Biomol. Eng., 2019, 10, 43–57. 57 G. Guandalini, M. C. Romano, M. Ho, D. Wiley, E. S. Rubin
41 S. Mohammad, J. Fitzgerald, R. L. Robinson and K. A. M. and J. C. Abanades, A sequential approach for the economic
Gasem, Experimental Uncertainties in Volumetric Methods evaluation of new CO2 capture technologies for power
for Measuring Equilibrium Adsorption, Energy Fuels, plants, Int. J. Greenhouse Gas Control, 2019, 84, 219–231.
2009, 23, 2810–2820. 58 J. A. Mason, K. Sumida, Z. R. Herm, R. Krishna and J. R.
42 Y. Gensterblum, P. van Hemert, P. Billemont, A. Busch, D. Long, Evaluating metal–organic frameworks for post-
Charriére, D. Li, B. M. Krooss, G. de Weireld, D. Prinz and combustion carbon dioxide capture via temperature swing
K.-H. A. A. Wolf, European inter-laboratory comparison of adsorption, Energy Environ. Sci., 2011, 4, 3030.
high pressure CO2 sorption isotherms. I: Activated carbon, 59 Z. Zhang, S. Xian, Q. Xia, H. Wang, Z. Li and J. Li,
Carbon, 2009, 47, 2958–2969. Enhancement of CO2 Adsorption and CO2/N2 Selectivity on
43 Y. Gensterblum, P. van Hemert, P. Billemont, E. Battistutta, ZIF-8 via Postsynthetic Modification, AIChE J., 2013, 59,
A. Busch, B. M. Krooss, G. De Weireld and K.-H. A. A. Wolf, 2195–2206.
European inter-laboratory comparison of high pressure CO2 60 R. Haghpanah, R. Nilam, A. Rajendran, S. Farooq and I. A.
sorption isotherms II: Natural coals, Int. J. Coal Geol., Karimi, Cycle synthesis and optimization of a VSA process
2010, 84, 115–124. for postcombustion CO2 capture, AIChE J., 2013, 59,
44 IEA Greenhouse Gas R&D Programme (IEA GHG), CO2 4735–4748.
Capture in the Cement Industry, 2008. 61 D. Diagne, M. Goto and T. Hirose, Parametric Studies on
45 IEA Greenhouse Gas R&D Programme (IEA GHG), Iron and CO2 Separation and Recovery by a Dual Reflux PSA Process
Steel CCS Study (Techno-Economics Integrated Steel Mill), Consisting of Both Rectifying and Stripping Sections, Ind.
2013. Eng. Chem. Res., 1995, 34, 3083–3089.

This journal is © The Royal Society of Chemistry 2019 Mol. Syst. Des. Eng.
 View Article Online

Paper MSDE

62 A. Agarwal, L. T. Biegler and S. E. Zitney, A superstructure- frameworks M2(dobdc) (M = Mg, Mn, Fe, Co, Ni, Cu, Zn),
based optimal synthesis of PSA cycles for post-combustion Chem. Sci., 2014, 5, 4569–4581.
CO2 capture, AIChE J., 2009, 56, 1813–1828. 77 Y. Jiao, C. R. Morelock, N. C. Burtch, W. P. Mounfield, J. T.
63 G. T. Rochelle, in Absorption-Based Post-combustion Capture Hungerford and K. S. Walton, Tuning the Kinetic Water
of Carbon Dioxide, Elsevier, 2016, pp. 35–67. Stability and Adsorption Interactions of Mg-MOF-74 by Par-
64 E. S. Rubin, S. Yeh, M. Antes, M. Berkenpas and J. Davison, tial Substitution with Co or Ni, Ind. Eng. Chem. Res.,
Use of experience curves to estimate the future cost of power 2015, 54, 12408–12414.
plants with CO2 capture, Int. J. Greenhouse Gas Control, 78 S. Choi, T. Watanabe, T.-H. Bae, D. S. Sholl and C. W. Jones,
2007, 1, 188–197. Modification of the Mg/DOBDC MOF with Amines to
Published on 23 September 2019. Downloaded by University of Toronto on 1/3/2020 1:27:15 AM.

65 M. M. F. Hasan, E. L. First, F. Boukouvala and C. A. Floudas, A Enhance CO2 Adsorption from Ultradilute Gases, J. Phys.
multi-scale framework for CO2 capture, utilization, and seques- Chem. Lett., 2012, 3, 1136–1141.
tration: CCUS and CCU, Comput. Chem. Eng., 2015, 81, 2–21. 79 X. Su, L. Bromberg, V. Martis, F. Simeon, A. Huq and T. A.
66 C. Kolster, E. Mechleri, S. Krevor and N. Mac Dowell, The Hatton, Postsynthetic Functionalization of Mg-MOF-74 with
role of CO2 purification and transport networks in carbon Tetraethylenepentamine: Structural Characterization and En-
capture and storage cost reduction, Int. J. Greenhouse Gas hanced CO 2 Adsorption, ACS Appl. Mater. Interfaces,
Control, 2017, 58, 127–141. 2017, 9, 11299–11306.
67 A. Masala, J. G. Vitillo, G. Mondino, C. A. Grande, R. Blom, 80 T. M. McDonald, W. R. Lee, J. A. Mason, B. M. Wiers, C. S.
M. Manzoli, M. Marshall and S. Bordiga, CO2 Capture in Dry Hong and J. R. Long, Capture of Carbon Dioxide from Air
and Wet Conditions in UTSA-16 Metal–Organic Framework, and Flue Gas in the Alkylamine-Appended Metal–Organic
ACS Appl. Mater. Interfaces, 2017, 9, 455–463. Framework mmen-Mg2IJdobpdc), J. Am. Chem. Soc.,
68 V. K. Chuikina, A. V. Kiselev, L. V. Mineyeva and G. G. 2012, 134, 7056–7065.
Muttik, Heats of adsorption of water vapour on NaX and 81 F. Nouar, J. Eckert, J. F. Eubank, P. Forster and M.
KNaX zeolites at different temperatures, J. Chem. Soc., Eddaoudi, Zeolite- like Metal−Organic Frameworks (ZMOFs)
Faraday Trans. 1, 1976, 72, 1345. as Hydrogen Storage Platform: Lithium and Magnesium Ion-
69 K. N. Son, T.-M. J. Richardson and G. E. Cmarik, Exchange and H2-(rho-ZMOF) Interaction Studies, J. Am.
Equilibrium Adsorption Isotherms for H2O on Zeolite 13X, Chem. Soc., 2009, 131, 2864–2870.
J. Chem. Eng. Data, 2019, 64, 1063–1071. 82 G. Calleja, J. A. Botas, M. Sánchez-Sánchez and M. G. Orcajo,
70 Inventys Inc., Technology, http://inventysinc.com/ Hydrogen adsorption over Zeolite-like MOF materials modi-
technology/, (accessed 14 July 2019). fied by ion exchange, Int. J. Hydrogen Energy, 2010, 35,
71 US Department of Energy, MOVE Program Overview, https:// 9916–9923.
arpa-e.energy.gov/sites/default/files/documents/files/MOVE_ 83 D. W. Breck, Zeolite Molecular Sieves: Structure, Chemistry
ProgramOverview.pdf, (accessed 28 July 2019). and Use, John Wiley & Sons Inc., New York, 1974.
72 Z. Bao, L. Yu, Q. Ren, X. Lu and S. Deng, Adsorption of CO2 84 R. M. Barrer and R. M. Gibbons, Zeolitic carbon dioxide:
and CH4 on a magnesium-based metal organic framework, energetics and equilibria in relation to exchangeable cations
J. Colloid Interface Sci., 2011, 353, 549–556. in faujasite, Trans. Faraday Soc., 1965, 61, 948.
73 A. Masala, F. Grifasi, C. Atzori, J. G. Vitillo, L. Mino, F. 85 J. Zhang, R. Singh and P. A. Webley, Alkali and alkaline-
Bonino, M. R. Chierotti and S. Bordiga, CO2 Adsorption earth cation exchanged chabazite zeolites for adsorption
Sites in UTSA-16: Multitechnique Approach, J. Phys. Chem. based CO2 capture, Microporous Mesoporous Mater.,
C, 2016, 120, 12068–12074. 2008, 111, 478–487.
74 A. K. Rajagopalan and A. Rajendran, The effect of nitrogen 86 K. S. Walton, M. B. Abney and M. Douglas LeVan, CO2
adsorption on vacuum swing adsorption based post- adsorption in Y and X zeolites modified by alkali metal
combustion CO2 capture, Int. J. Greenhouse Gas Control, cation exchange, Microporous Mesoporous Mater., 2006, 91,
2018, 78, 437–447. 78–84.
75 R. Zhao, S. Deng, S. Wang, L. Zhao, Y. Zhang, B. Liu, H. Li 87 X. Zhang, Z. Chen, X. Yang, M. Li, C. Chen and N. Zhang,
and Z. Yu, Thermodynamic research of adsorbent materials The fixation of carbon dioxide with epoxides catalyzed by
on energy efficiency of vacuum-pressure swing adsorption cation-exchanged metal-organic framework, Microporous
cycle for CO2 capture, Appl. Therm. Eng., 2018, 128, 818–829. Mesoporous Mater., 2018, 258, 55–61.
76 W. L. Queen, M. R. Hudson, E. D. Bloch, J. A. Mason, M. I. 88 A. L. Myers and F. Siperstein, Characterization of adsorbents
Gonzalez, J. S. Lee, D. Gygi, J. D. Howe, K. Lee, T. A. by energy profile of adsorbed molecules, Colloids Surf., A,
Darwish, M. James, V. K. Peterson, S. J. Teat, B. Smit, J. B. 2001, 187–188, 73–81.
Neaton, J. R. Long and C. M. Brown, Comprehensive study 89 Z. T. Wilson and N. V. Sahinidis, The ALAMO approach to
of carbon dioxide adsorption in the metal–organic machine learning, Comput. Chem. Eng., 2017, 106, 785–795.

Mol. Syst. Des. Eng. This journal is © The Royal Society of Chemistry 2019