Sie sind auf Seite 1von 15

Chemical Engineering Science 174 (2017) 174–188

Contents lists available at ScienceDirect

Chemical Engineering Science


journal homepage: www.elsevier.com/locate/ces

Computationally based analysis of the energy efficiency of a CO2 capture


process
Bojan Vujic, Alexander P. Lyubartsev ⇑
Stockholms University, Department of Materials and Environmental Chemistry, Division of Physical Chemistry, Stockholm University, SE 106 91 Stockholm, Sweden

h i g h l i g h t s g r a p h i c a l a b s t r a c t

 A model for thermodynamic


evaluation of the energy efficiency of
CO2 capture.
 Accounting for the working capacity,
regenerability and purity of the
captured CO2.
 Determination of the most efficient
desorption conditions for CO2
capture.
 Grand Canonical Monte Carlo
simulations for more than 1000
zeolite structures.

a r t i c l e i n f o a b s t r a c t

Article history: We propose a model for thermodynamic evaluation of the energy efficiency of a CO2 capture in a
Received 4 October 2016 temperature-pressure swing adsorption. Major parts of this model are computational prediction of the
Received in revised form 8 August 2017 adsorbed gas loading as a function of temperature and partial CO2 pressure, evaluation of the energy
Accepted 1 September 2017
expenses under specified conditions for the working capacity, regenerability of the sorbent and purity
Available online 7 September 2017
of the captured CO2, as well as determination of the most optimal desorption conditions in terms of des-
orption pressure and temperature. The proposed model can be applied for fast evaluation of the energy
Keywords:
costs of the CO2 capture process with the use of both experimental or simulation adsorption data with
Gas separation
Adsorption
respect to pressure and temperature. We tested this model analyzing data obtained from Grand
Carbon capture and storage Canonical Monte Carlo simulations for more than thousand different zeolite structures. Within our
Monte Carlo simulations approach it is possible to evaluate a theoretical limit of the energy expenses for each specific material
Pressure-temperature swing adsorption and to use the proposed method in screening different structures for the most efficient sorbent material
from the energy efficiency point of view under specified requirements for the working capacity of the
process, regenerability and purity of captured CO2. We show that setting realistic from the industrial
point of view parameters of the CO2 capture cycle leads to substantial reduction of the number of suitable
zeolite structures, and to increase of the energy penalty of the CO2 capture compared to evaluations based
on minimization of the parasitic energy only.
Ó 2017 Elsevier Ltd. All rights reserved.

1. Introduction special emphasis is placed particularly on carbon dioxide, as the


main greenhouse gas. Increasing concentration of CO2 in the atmo-
One of the major problems the modern era is facing is a rapid sphere leads to global warming, and may also have a negative
increase of greenhouse gases concentration in the atmosphere. A impact on human health (Patz et al., 2005; Milly et al., 2005;
McMichael et al., 2006). Because of this reason, science and tech-
⇑ Corresponding author. nology are facing a problem of finding new ways to fast and yet
E-mail address: alexander.lyubartsev@mmk.su.se (A.P. Lyubartsev). efficiently reduce CO2 emission. Carbon capture and storage

http://dx.doi.org/10.1016/j.ces.2017.09.006
0009-2509/Ó 2017 Elsevier Ltd. All rights reserved.
B. Vujic, A.P. Lyubartsev / Chemical Engineering Science 174 (2017) 174–188 175

(CCS) technology (Bacsik et al., 2016; Hedin et al., 2013; Xu and pressure (conventionally 150 bar). These energy contributions
Hedin, 2016) is expected to play an important role in the reduction can be evaluated from the adsorption isotherms determined in
of greenhouse gases emissions into the atmosphere (Miles and computer simulations of considered materials. By varying thermo-
Kapos, 2008; Paustian et al., 1998; Choi et al., 2009; Ming et al., dynamic conditions of the desorption state, a minimum of the par-
2016; Songolzadeh et al., 2012; Smit et al., 1986). asitic energy (per mol or per gram of the captured CO2) can be
CCS can be defined as the separation and capture of CO2 pro- found. The most optimal desorption conditions are defined by
duced at stationary sources, followed by transport and storage in the minimum of the parasitic energy. Computations can be done
geological reservoirs or in the ocean (Damen et al., 2006). Carbon for any material with known atomic structure, thus the method
dioxide can be captured directly from industrial sources by one becomes suitable for screening many possible candidates for effi-
of three main methods: pre-combustion capture (via oxygen- cient CO2 adsorbents.
blown gasification), post-combustion capture, and oxyfuel process The model described in works (Lin et al., 2012; Huck et al.,
(Songolzadeh et al., 2012; D’Alessandro et al., 2010; Benson and 2014) predicted the theoretical minimum of the parasitic energy
Orr, 2008). Post-combustion capture in power plants involves the from the basic relationships of thermodynamics. However, that
separation of CO2 from flue gas mixtures, followed by its compres- methodology does not consider several important parameters of
sion and sequestration in geological formations (Metz et al., 2005; the adsorption-desorption cycle which are important from the
Massood et al., 2007). industrial applications point of view: working capacity (which is
There are several techniques for separation of carbon dioxide defined as the difference between adsorption and desorption load-
from flue gas mixture, such as adsorption (Martin-Calvo et al., ings and equals to amount of CO2 captured in a single cycle), purity
2014; Sethia et al., 2010; Maurin et al., 2005; Hedin et al., 2013; of the captured CO2 and regenerability (percent of CO2 remaining
Akhtar and Bergström, 2011), absorption (Yan et al., 2008; in the sorbent). Our computations show that in many cases work-
Conway et al., 2015), cryogenic membrane (Pfaff and Kather, ing conditions defined by the minimum of the parasitic energy
2009; Ye et al., 2014) and micro algal bio-fixation (Pellerano alone correspond to a very small working capacity (which would
et al., 2009). Practical choices can vary, but in general, adsorption require many adsorption/desorption cycles for the same amount
can be a very good choice in terms of efficiency, possibility of of CO2 captured and thus make the whole capturing process slow
large-scale applications and recycling of the adsorbent material. and expensive), to a low regenerability of the sorbent framework
Considering the overall cost of CO2 capture, adsorption process in and to poor purity of the captured CO2. Amount of the parasitic
comparison with other methods is the most efficient one energy related to the realistic working conditions can thus be
(Pellerano et al., 2009; Liu et al., 2010). highly underestimated. We therefore propose in this work to com-
Adsorption relates to uptake of sorbate molecules onto the sur- plement the search of the optimal conditions for the minimum of
face or in pours of a solid adsorbent, to which they adhere via two the parasitic energy by additional conditions for the minimal
types of forces, weaker van der Waals and dipole forces in the case acceptable working capacity, purity and regenerability.
of physisorption, or stronger covalent bonding in the case of Another problem we address to is how to make the computer
chemisorption (Hedin et al., 2013). Absorption in contrasts with search for the most suitable adsorbents more efficient. For precise
adsorption is the process where sorbate molecules dissolve into computation of the parasitic energy, it is required to have a large
the bulk of the material itself (Bhown and Freeman, 2011). The most number of fugacity steps (desorption points) for each adsorption
used physisorbents, such as active carbon (Ribeiro et al., 2008; isotherm, as well as a large number of adsorption isotherms in a
Berlier, 1997), metal-organic frameworks (MOFs) (Bastin et al., reasonable temperature range. Such computations require hun-
2008; Bao et al., 2015; Lee et al., 2015; McDonald et al., 2015; dreds of simulations for each considered material which makes
Verdegaal et al., 2016) and zeolites (Liu et al., 2010; Cheung et al., the screening process expensive in terms of CPU time. Here we pro-
2012; Siriwardane et al., 2005; García-Pérez et al., 2007; García- pose a thermodynamical model (based on the Langmuir adsorption
Sánchez et al., 2009), are better options than chemisorbents, such model), which expresses the adsorption load of CO2 and N2 as a
as amine-modified mesoporous silica (Bacsik et al., 2010; Hedin function of partial gas pressures and temperature. Parameters of
et al., 2010) due to relatively easy regeneration of the sorbent mate- this function can be fitted by running simulations for about ten
rial by the change of temperature or pressure. temperature–pressure points, which allows reducing the number
Whatever technique is used for CO2 capture, it inevitably of computations (relative to the computation of the whole set of
requires certain amount of energy, which exert an additional load adsorption isotherms) by more than one order of magnitude. Fur-
on the power plant and which is called for this reason parasitic thermore, within the model the parasitic energy itself is expressed
energy. Within existing techniques parasitic energy can account as an analytical function of the parameters which facilitates evalu-
for about 30% of the total energy produced by a power plant ation of the parasitic energy, and makes it possible to use the model
(Bhown and Freeman, 2011; Lin et al., 2012) and reducing this for experimental data. Using this methodology we are able to fast
energy cost is of primary importance for further development of and yet accurately evaluate the minimum of the parasitic energy
CCS technology. Not surprisingly, much effort in recent years, in and thus find the most optimal desorption conditions (in terms of
both experimental (Jensen et al., 2012; Harlick and Tezel, 2004) desorption temperature and pressure) for working capacity, regen-
and computational (Gomez-Alvarez et al., 2016; Babarao and erability and purity of captured CO2 specified in advance.
Jiang, 2008; Deroche et al., 2008; Krishna and Long, 2011;
Yazaydin et al., 2009; Kim et al., 2012; Hasan et al., 2013;
Matito-Martos et al., 2014; Matito-Martos et al., 2015) studies have 2. Materials and computational methods
been devoted to investigation of the adsorption properties of vari-
ous materials like zeolites and MOFs, as well as the energy costs 2.1. Zeolites as CO2 adsorbents
and efficiency of CCS.
In recent computational studies (Lin et al., 2012; Berger and Among many materials which can be used as adsorbents (Hedin
Bhown, 2013; Huck et al., 2014) a model for computation of para- et al., 2013; Ribeiro et al., 2008; Cheung et al., 2012; Hedin et al.,
sitic energy in a typical temperature–pressure swing process has 2010; Xu and Hedin, 2016), zeolites have shown promising results
been suggested which included two main contributions: thermal for separating carbon dioxide from flue gas mixtures in the pres-
energy to desorb CO2 and thus regenerate the adsorbent, and com- sure swing adsorption (PSA) (Chue et al., 1995), temperature swing
pression energy to compress captured CO2 to the transportation adsorption (TSA) (Merel et al., 2008; Ntiamoah et al., 2016) as well
176 B. Vujic, A.P. Lyubartsev / Chemical Engineering Science 174 (2017) 174–188

as in the combined techniques. Zeolites are three-dimensional (Metz et al., 2005; Bhown and Freeman, 2011). While it can be dif-
microporous materials which are built from tetrahedral blocks. ficult to affect the cost of transport and storage, there can exist
Central atoms (T-atoms) in these blocks are silicon, but quite often good possibilities to significantly reduce the cost of capture by a
aluminum and/or phosphorus. Depending on the composition of T proper choice of adsorbent.
atoms inside a zeolite structure, zeolites can be divided in general Following previous works, (Lin et al., 2012; Huck et al., 2014)
into three main groups: alumino-silicates, alumino-phosphates we distinguish three energy components needed to arrange a full
and silico-alumino-phosphates. adsorption-desorption cycle in a combined pressure–temperature
At some point in the zeolite framework, Al3+ ion can replace Si4+ swing capture process: (1) energy to heat the material to the des-
and thus the framework becomes negatively charged (Čejka et al., orption temperature; (2) energy to supply the heat of desorption
2007) For each Si4+ replaced by Al3+ the charge must be balanced (equal to the minus enthalpy of adsorption); (3) energy required
by other positive ions such as Na+, K+ and Ca2+, or with protons, to pressurize captured CO2 to 150 bar. We modified approach
H+. According to Löwenstein’s rule (Löwenstein, 1954), the distri- accepted in the previous works by providing a more realistic com-
bution of aluminum-oxygen tetrahedra is such that they tend to putation of energy required to perform the compression part by
connect to four silicon-oxygen tetrahedra. In another words, oxy- considering separately ranges of pressure below and above 1 atm
gen bridges between two aluminum atoms (Al–O–Al) are forbid- and taking into account the degree of purity of captured CO2 as
den. This leads to the conclusion that in zeolites, the minimal Si: well as considering a multistage compression train.
Al ratio is equal to one. Still, in some zeolites, depending on the The first two parts of the energy requirements represent ther-
space group of the crystal system, it is not possible to create a mal energy (Q), which can be obtained in a power plant diverting
structure with Si:Al = 1 without violation of the Löwenstein’s rule. steam from a turbine. Since not the whole thermal energy is con-
Thus in zeolite structures the minimal Si:Al ratio is often greater verted to the output mechanical work, the parasitic load on the
that 1. power plant caused by diverting the steam can be evaluated by
Since carbon dioxide emissions are frequently associated with multiplication of the thermal energy by the efficiency of turbine
large amounts of other gases such as nitrogen, an adsorbent selec- (e) and Carnot efficiency (g) (Lin et al., 2012). The total parasitic
tivity for one of these compounds is required. These adsorbents energy (Ep ) will be then
should also be selective even at high temperatures, i.e., tempera-
Ep ¼ egQ thermal þ W comp: þ W v acuum ð1Þ
tures typical of carbon dioxide emission sources (Dantas et al.,
2012). In principle, a carbon dioxide adsorbent must have a high where W v acuum is the vacuum work to create a low pressure for the
adsorption capacity, high selectivity, adequate adsorption/desorp- desorption process, W comp is the compression work to pressurize
tion kinetics, and possess good thermal and mechanical stability, captured CO2 ; e is the efficiency of turbine (taken to be equal to
remaining stable after several adsorption/desorption cycles 0.75). Carnot efficiency (g) is determined by the temperature ratio:
(Helwani et al., 2012). Besides all these requirements which should
T ads
be fulfilled by a perspective adsorbent material, the requirement of g¼1 ð2Þ
an acceptable energy cost for the adsorption/desorption cycle
T des
should also be satisfied in practical use. Thus an adsorbent material where T ads and T des denote the adsorption and desorption tempera-
needs to provide economically efficient adsorption with a good tures, respectively.
working capacity, high purity of captured CO2, high regenerability, The thermal work (thermal energy expenses) is determined by
as well as low cost energy requirements. In this work we concen- the temperature difference and amount of the adsorbed gases (CO2
trate on equilibrium thermodynamic properties of adsorption and N2)
which determine the theoretical limit for the energy cost of CO2  
capture, leaving issues related to adsorption kinetics and frame-
cp msorbent ðT des  T ads Þ þ DnCO2 DhCO2 þ DnN2 DhN2
Q thermal ¼ ð3Þ
work stability for future studies. mCO2
A total number of theoretically predicted zeolite framework where cp and msorbent are the heat capacity and mass of the sorbent,
types is enormously large (Pophale et al., 2011) compared to pre-
respectively; DhCO2 and DhN2 are the heats of adsorption of the flue
sently known 225 zeolite frameworks (Baerlocher et al., 2007). In
gas components; DnCO2 and DnN2 are the amount (in mol) of the des-
this work we consider only known IZA zeolite structures, noting
orbed (captured in one adsorption/desorption cycle) gas; mCO2 is the
that only few of them are so far industrially commercialized
amount of captured CO2 in kg per 1 kg of adsorbent
(Čejka et al., 2007). With varying composition of Si and Al atoms
(mCO2 ¼ DnCO2  MCO2 ). In practical computation we assume ‘‘aver-
we screen each of IZA zeolite frameworks with 8 different Si:Al 1
ratios (2, 3, 5, 7, 10, 20, 40 and all silica structures), resulting with age” heat capacity of the zeolite framework C p ¼ 1 kJkg K1
1800 structures in total. (Hemingway and Robie, 1984; Deem et al., 2009; Čejka et al., 2007)
According to Eqs. (2) and (3), if there is no change in tempera-
2.2. Energy expenses in CCS process ture of a flue gas (DT ¼ T des  T ads ¼ 0 which corresponds to pres-
sure swing adsorption cycle), thermal work ðQ Þ will be equal to
Flue gas produced in power plants is often a mixture of nitro- zero. The parasitic energy evaluated in that way depends only on
gen, carbon dioxide, water and other gases. Partial pressure (con- the vacuum and compression works.
centration) of some components can vary, but frequently the The amount of captured CO2 and N2 (in moles) is determined as
composition of flue gas is as follows: 70–75% N2, 10–15% CO2, the difference in the load at the adsorption and desorption points
8–10% H2O, 3–4% O2, with traces of SOx, NOx and other compounds DnCO2 ¼ nads and DnN2 ¼ nads
CO2  nCO2 N2  nN2 ð4Þ
des des
(Bhown and Freeman, 2011; Feron, 2010). Flue gases are normally
at atmospheric pressure (1 bar), but the temperature can vary from The difference between adsorption and desorption loading
30° up to a 100 °C (303–373 K). Following previous works (Lin defines the working capacity, that is the amount of CO2 or N2 cap-
et al., 2012; Huck et al., 2014), we in this paper assume a flue tured in a single adsorption–desorption cycle and given usually in
gas as a mixture of nitrogen and carbon dioxide in a molar ratio terms of mol of the captured gas per unit mass of the sorbent
85:15, and temperature 40 °C. framework. The adsorption point is determined by the physical
Among the three parts of CCS, the separation and capture pro- properties (pressure and temperature) of the flue gases in the tur-
cess of carbon dioxide from flue gas mixtures is the most expensive bine outlet. Typically, pads and T ads , are 100 kPa (1 bar) and 40 °C
B. Vujic, A.P. Lyubartsev / Chemical Engineering Science 174 (2017) 174–188 177

respectively (Bhown and Freeman, 2011; Lin et al., 2012) which is thalpy isotherms, C p and C v of gases downloaded from the NIST
also accepted in this work. Unlike the adsorption point, thermody- database (Linstrom and Mallard, 2005) (summarized and available
namic conditions at the desorption point can vary. The subsequent from the authors upon request). Thus, real gas properties have
desorption can be achieved by decreasing the pressure (Pressure been used in all calculations regarding the compression work.
Swing Adsorption, or PSA), by increasing the temperature (Temper- Assuming that the efficiency of the compression process is 0.85,
ature Swing Adsorption, or TSA) or by simultaneous change of tem- amount of work required for compression of captured gas from
perature and pressure which is called hybrid temperature/pressure 1 bar to 150 bar, including 0.2 bar pressure loss and 5% heat-
swing adsorption (TPSA). Partial contributions from all compo- exchanger loss between subsequent stages, is equal to 406 kJ(kg
nents of the parasitic energy depend on the equilibrium adsorption CO2)1. For comparison, ideal compression under single-stage
load at the adsorption and desorption conditions. Thus for calcula- isothermal conditions computed by Eq. (5) is equal to 337 kJ(kg
tion of the parasitic energy the most important data are the CO2)1. Assuming instead that the efficiency of isothermal com-
adsorption isotherms, which together with the isosteric heat of pression process is equal to 70.55% (instead of 85%), W isothermal
comp:
adsorption provide all necessary data to compute the parasitic
becomes equal to W adiabatic
comp: .
energy.
Regarding the compression work in Eq. (1), we first note that Regardless of which compression method is used, for an ideal
compression of the desorbed CO2 from the desorption point (which case when the purity of the captured CO2 is equal to 100%, the dif-
is typically below the atmospheric pressure) to the atmospheric ference between any two specific compression pathways is a con-
pressure (1 bar) does not in fact require energy expenses, and in stant which can be of order of several tens of kJ(kg CO2)1.
further calculations we assume the compression work starting Therefore, deviation in the computed compression work from com-
from 1 bar pressure in cases when the desorption pressure is below pression work in real industrial applications will be the same for
1 bar. Further, the work to compress the captured gas to the stor- any adsorbent material and not important solely for comparison
age pressure can be evaluated, in the first approximation, for a of different zeolite structures in the process of screening zeolites
closed system at constant temperature determined in the ideal for gas separation. We argue that for the screening purposes it is
gas approximation: not essential which compression method will be used. We note
  further that in all our calculations we assumed adiabatic compres-
RT ads p sion composed of 7 stages with efficiency of heat-exchanger in the
W isothermal
comp: ¼ ln 2 ð5Þ
0:85M gas p1 interstage cooling process equal to 95% and a pressure drop of
0.2 bar between subsequent compression stages.
where p1 and p2 are inlet and outlet pressure respectively; Mgas is an
As we already noted, compression of the captured CO2 from the
average molar mass of the captured gas (given in kgmol1); R is the
1
desorption point (below the atmospheric pressure) to the atmo-
gas constant (8.314 Jmol K1) and factor 0.85 takes into account spheric pressure does not require energy expenses. On the other
the average efficiency of the compression process (Lin et al., 2012) hand, certain energy is required for generating low pressure (work
A typical industrial compression is usually an open flow system against atmospheric pressure to create vacuum). We therefore pro-
with increase of temperature from an inlet pressure to an outlet pose a model for calculation of the amount of work to create low
pressure, which can be approximated by adiabatic compression: (under 1 bar) pressure needed to put adsorbent framework with
" c1 #
RT ads c p2 c
W adiabatic ¼ 1 ð6Þ
comp:
0:85M gas c  1 p1

where c is the average ratio between molar heat capacities


(c ¼ Cp=C v ). The average molar mass of the captured gas is calcu-
lated according to the expression:
M gas ¼ Pr  MCO2 þ ð1  Pr Þ  MN2 ð7Þ
where MCO2 and MN2 are molar masses of CO2 and N2, taken to be 44
and 28 gmol1, respectively; Pr is the purity of captured CO2 given
as
DnCO2
Pr ¼ ð8Þ
DnCO2 þ DnN2
An average ratio of heat capacities c is evaluated as:
Cpgas
cgas ¼ ð9Þ
C v gas
where

Cpgas ¼ Pr  CpCO2 þ ð1  Pr Þ  CpN2


ð10Þ
C v gas ¼ Pr  C v CO2 þ ð1  Pr Þ  C v N2
In industrial conditions compression is usually decomposed
into several stages of adiabatic compression with subsequent cool- Fig. 1. Schematic presentation of a typical 6 stage-compression of CO2, and
ing, creating so called single compression train. A typical 6 stage compression train assumed in this work are shown on a pressureenthalpy
compression train (Vermeulen, 2011) shown in Fig. 1 requires diagram. Horizontal lines represent interstage cooling when compressed gas is
around 378 kJ(kg CO2)1 of energy to compress captured CO2 to cooled back to the inlet temperature (in our study 40  C). Efficiency of the heat
exchanger in interstage cooling was assumed to be 95% and a pressure drop of
123 bar. We have computed the work to compress 1 kg CO2 gas 0.2 bar between subsequent compression stages is included in the calculations.
from 1 bar to 150 bar using a 7 stage compression train shown in Tabulated pressure–enthalpy isotherms are downloaded from the NIST database
Fig. 1 by black line and using numerical values for pressure–en- (Linstrom and Mallard, 2005).
178 B. Vujic, A.P. Lyubartsev / Chemical Engineering Science 174 (2017) 174–188

adsorbed CO2 for gas desorption and sorbent regeneration. Given


the fact that for low pressures under 1 bar the gas state is well
described by the ideal gas equation, and that due to low heat
capacity of rarefies gas an isothemic process can be a more appro-
priate way of description, we compute the ”vacuum” work assum-
ing isothermic process and ideal gas equation of state. Denoting p1
as desorption pressure and p0 ¼ 1bar, we have for 1 mol of gas:
Z V1 Z V1
W v acuum ¼ DpðVÞdV ¼ ðp0  pðVÞÞdV
V0 V0
Z V1  
RT ð11Þ
¼ p0  dV
V0 V
V1
¼ p0 ðV 1  V 0 Þ  RT ln
V0
By using V 1 ¼ V 0  p0 =p1 , we have
p0 V 0 p0 V1 Fig. 2. Schematic presentation of adsorption and desorption stages.
W v acuum ¼  p0 V 0  RT ln ð12Þ
p1 V0
Taking p0 V 0 ¼ RT, for vacuum work (for 1 mol of captured gas) In the calculations, adsorbed loadings of each particular gas are
finally we have: determined by the partial pressure of the given gas and adsorption
  isotherms. Desorption loading, however, is predicted by the Ideal
p0 V1
W v acuum ¼ RT  1  RT ln Adsorbed Solution Theory (IAST), where the molar ratio of CO2
p1 V0
  ð13Þ and N2 molecules is assumed to be equal to that in the adsorbed
p0 p0
¼ RT  1  ln phase. By taking desorption pressure (pdes ) as a sum of partial
p1 p1 CO2 and N2 pressures
Taking the efficiency of the vacuum process to be equal to 0.75,
pdes ¼ ppartial
CO2 þ ppartial
N2 ð15Þ
vacuum work for 1 kg of captured gas can be estimated as follows:
  we assume that partial pressures of CO2 and N2 are equal to
RT des pads p
W v acuum ¼  1  ln ads ð14Þ
0:75M gas pdes pdes ppartial
CO2 ¼ xCO2  pdes and ppartial
N2 ¼ xN2  pdes ð16Þ
Thus instead of using only compression work, we introduce two where xCO2 and xN2 are the mole fractions of each gas, given as
separate functions: vacuum work, Eq. (13) is used for flue gas pres- follows:
sures below 1 atmosphere (in general below adsorption pressure)
and compression work, Eq. (5) for pressurization of captured gas nads
CO2 nads
N2
xCO2 ¼ and xN2 ¼ ð17Þ
from atmospheric pressure to 150 bar. nads
CO2 þ nads
N2 nads
CO2 þ nads
N2

2.3. Adsorption – desorption cycle The amount of adsorbed CO2 and N2 molecules remaining in the
sorbent framework under the desorption pressure (ndes des
CO2 and nN2 )

Adsorption and desorption processes are assumed as follows. was then computed using the fitting Eq. (25) (see details in the sec-
Composition of the flue gas is assumed as a mixture of CO2 and tion below), for partial pressures of each given gas and applying the
N2 in molar ratio 15:85, respectively. Adsorption takes place at flue IAST model (Myers and Prausnitz, 1965; Walton and Sholl, 2015;
gas temperature (T ads ¼ 313 K and atmospheric pres- Simon et al., 2016; Kadoura et al., 2016). To test the validity of
surepads ¼ 100 kPa). During the adsorption (Fig. 2) valves 1 and 2 the IAST model versus competitive adsorption of a binary gas mix-
are kept open, while valve 3 is closed. Due to the different break- ture, we performed a test simulations on randomly chosen 50 zeo-
through curves of CO2 and N2 (Dantas et al., 2009; Ye et al., lite structures. Only in a few cases the estimated error in parasitic
2013; Liu et al., 2011) valve 2 is kept open until CO2 molecules energy calculated as
reach this valve, after which it becomes closed. After the adsorbent jEpðcompetitiv eÞ  EpðIASTÞ j
reaches the saturation uptake the bed inlet (valve 1) is closed. The Eperror ¼  100% ð18Þ
Epðcompetitiv eÞ
amount of adsorbed CO2 and N2 (nads ads
CO2 and nN2 ) is defined by com-
puted adsorption isotherms of the given gas mixture. A pressuriza- was more than 1%, but always below 2.5%.
tion stage used sometimes in PSA is not assumed in this work. We
can argue that pressurizing CO2/N2 mixture would require more 2.4. Monte Carlo calculation of adsorption isotherms
work than for (almost) pure captured CO2.
The next step is the desorption process in which the adsorbed Adsorption isotherms in this work are computed by the Grand
gas is extracted from the sorbent framework and the sorbent is Canonical Monte Carlo (GCMC) (Stenqvist et al., 2013; Purton
regenerated. In the hybrid temperature-pressure swing process et al., 2013; Martin, 2013; Yiannourakou et al., 2013; Morales
the adsorption bed is first heated to the desorption temperature et al., 2013) simulation method as implemented in the package
(T des ). Upon the heating, adsorbed molecules start to desorb, which RASPA (Dubbeldam et al., 2013, 2016; Dubbeldam and Snurr,
is allowed by connecting the outlet bed (opening valve 3) with the 2007). We also used Materials Studio Accelrys (Materials Studio,
vacuum pump. The vacuum pump decreases the pressure to the 2013; Akkermans et al., 2013) for preparation of input files and
desorption value P des , and the process of desorption goes until a Music (Gupta et al., 2003; Chempath et al., 2013) software package
thermodynamic equilibrium is reached between the extracted for some of production runs. In GCMC simulations, alongside with
gas at temperature and pressure, and the remaining gas in the Monte Carlo steps moving the gas molecules, steps with insertion
adsorption bed. and deletion of a single molecule are performed with probabilities
B. Vujic, A.P. Lyubartsev / Chemical Engineering Science 174 (2017) 174–188 179

determined by the given chemical potential (Leach, 2001). In this coordinates and sodium ions were allowed to move during the
work we used the Peng–Robinson equation of state (Peng and simulations. CO2 and N2 molecules were treated as flexible. A typ-
Robinson, 1976) for conversion between gas pressure and fugacity. ical simulation consisted of 3  105 MC cycles (one MC cycle con-
The result of a CGMC simulation is the average number of mole- sists of attempts to move in average each molecule in the system)
cules in the simulation cell. Other thermodynamic parameters of where initial one third of cycles was used for equilibration and the
the system, such as adsorption enthalpy, are also determined from rest of simulation for collecting the averages. We found that the
the simulations. average number of adsorbed molecules in zeolite unit cell usually
We have carried out CGMC simulations of CO2/N2 gas mixture does not change significantly after about 105 MC cycles indicating
in a large number of zeolite-like frameworks available from the that convergence was reached.
Zeolite data base (Baerlocher et al., 2007). For each structure the In all of the adsorption isotherms, loading of CO2 and N2 mole-
unit cell was repeated periodically several times so that the system cules in the zeolite framework is given in the unit mmolg1. Since
size was always above 24 Å in each direction. Further, in each of GCMC simulations provide results as number of guest molecules
the framework types a part of T-atoms (which by default represent per unit cell, we convert it into mmolg1 by using expression
Si) were substituted in a random manner by Al atoms to have
desirable composition. The position of Al3+ ions in the zeolite nCO2 Nmolecules=cell Nmolecules=cell  103
nmmol=g ¼ ¼ ¼ ð19Þ
framework are usually unknown, that is why in our study we used mcell M cell  Da  N A M cell
a random method for distribution of Al3+ under the only condition
where Mcell is mass of the simulation cell framework in atomic
that Löwenstein’s rule is satisfied. After that, Na+ counter-ions
were added to neutralize the charge. Gas phase molecules (CO2 units; Da is unified atomic mass (equal to 1:66  1027 kg) and NA
and N2) were treated in the frame of a grand-canonical ensemble. is the Avogadro number (equal to 6:022  1023 mol1).
Transferable force field for adsorption of CO2, N2, O2 and Ar in zeo-
lites (Vujić and Lyubartsev, 2016), recently developed in our group, 3. Results and discussion
was used in the simulations. This force field was optimized to
reproduce CO2 and N2 (among other gases) experimental adsorp- 3.1. Minimization of the parasitic energy
tion isotherms of a large set of zeolite structures. Electrostatic
interactions were treated by the Ewald summation method In a typical zeolite framework and partial gas pressure below
(Frenkel and Smit, 2002; Leach, 2001; Karasawa and Goddard, 100 kPa one can generally distinguish three categories of adsorp-
1989). Van der Waals interactions where treated with the 12–6 tion isotherms depending on the value of Henry coefficient (Lin
Lennard-Jones potential. The potential was truncated at the cut- et al., 2012). In adsorption isotherms of category one (Fig. 3a,
off distance of 12 Å and long range tail correction was used. Zeolite SAO zeolite), the Henry coefficient has the highest value in com-
framework atoms were fixed at their initial crystallographic parison with two other zeolite frameworks, JSR and SAS. Thus,

Fig. 3. (a) Three categories of adsorption isotherms represented by adsorption of CO2 in SAO (red), JSR (blue) and SAS (green) zeolites. (b–d) Plots of the parasitic energy
(hexagons: 313 K blue, 333 K red and 353 K green) and working capacity (stars: 313 K blue, 333 K red and 353 K green) versus desorption pressure (as given by Eq. (15)). For
each graph (b–d), working capacity is given on the secondary, and parasitic energy on the primary y-axes. Si:Al ratio in all three zeolites (SAO, JSR, SAS) is equal to 2. (For
interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
180 B. Vujic, A.P. Lyubartsev / Chemical Engineering Science 174 (2017) 174–188

the adsorption isotherm can be described by the Henry law in a culate the minimum of the parasitic energy within the constraints
very low pressure range, after which the loading reaches the satu- that purity and regenerability have to be larger than 95 and 90%,
ration point. The amount of captured carbon dioxide in this cate- respectively. Theoretical minimum of the parasitic energy, within
gory of adsorption isotherms is usually high and saturation these constraints, can be used as a reliable indicator of zeolite per-
loading is reached faster than in two other categories of adsorption formance as a CO2 adsorbent in the CCS process.
isotherms. Regeneration of the sorbent may require in this case
rather high temperatures and/or low pressures, that is why para- 3.2. Predicting the adsorption isotherms of CO2 and N2
sitic energy can be relatively high, as demonstrated in Fig. 3b.
In the second category of adsorption isotherm, represented by To reduce the time for calculation of the adsorption isotherms
the JSR zeolite, the amount of adsorbed gas at the adsorption point in GCMC simulations, we have developed a simple Langmuir-like
(1 bar) is still high, but amount of the adsorbed gas is decreasing model which expresses the adsorption load as a function of gas
with a pressure drop faster than in the adsorption isotherm of cat- pressure and temperature. There are several methods for fitting
egory 1. This means that regeneration of the material can be done data from adsorption isotherms with respect to pressure (Čejka
at lower temperatures (T des P T ads ) or higher pressures (pdes < pads ) et al., 2007). According to the commonly used Langmuir model,
compared with isotherm of category 1, which as a result may have amount of adsorbed gas (loading) can be described using single
lower parasitic energy for the same working capacity, as demon- ði ¼ 1Þ or dual binding site ði ¼ 2Þ Langmuir isotherm
strated in Fig. 3c.
The adsorption isotherm of category three (Fig. 3a, SAS zeolite)
X
2
Kic
ni ¼ ni;mono ð21Þ
follows the Henry law in a wide pressure range. The amount of i¼1
1 þ Kic
captured CO2 with respect to the partial CO2 pressure is typically
lower than in the first two examples, and thus the working capac- where ni is loading of flue gas component at concentration
ity is low. To reach acceptable working capacity (if possible at all), ðc ¼ p=RTÞ; ni;mono is the saturation loading (called also monolayer
a very low desorption pressure or a high desorption temperature is capacity); and K i is the adsorption coefficient, related also to the
required, which would lead to a higher value of the parasitic adsorption Henry law constant and adsorption free energy.
energy, as shown in Fig. 3d. From these simple observations we Previous works (Lin et al., 2012; Hasan et al., 2013) have used
can conclude that the optimal performance in terms of parasitic two states Langmuir model to describe adsorption loading. Here
energy can be expected for materials showing adsorption iso- we have employed both, a single site Langmuir model for struc-
therms of category two. tures which show this type of adsorption isotherms, as well as dual
Another observation from Fig. 3 is that the graphs for parasitic site for other structures. Furthermore, in contrast to previous stud-
energy as a function of desorption pressure look similar, since in ies we complemented the Langmuir adsorption model by a tem-
the case of PSA the parasitic energy is determined only by the des- perature dependence of the monolayer capacity.
orption pressure and purity of captured CO2. However, working For a constant temperature we describe adsorption load of CO2
capacities of these three cases are different (note also different as a function of pressure by
scale of the axes). We demonstrate also in Fig. 3 that the minimum X i
2
nimono kCO2 pCO2
CO2 ¼
nads ð22Þ
of parasitic energy is achieved when the desorption point
i
approaches the adsorption point (15 kPa CO2, 85 kPa N2 and i¼1 1 þ kCO2 pCO2
313 K, in the present case). At the same conditions, the working i
capacity tends to zero, which means that only a minor part of where kCO2 is adsorption Henry coefficient, nimono is the monolayer
the adsorbed CO2 become captured. Same conclusion can be made capacity (saturation loading), and pCO2 is the partial pressure of car-
from the analysis of the parasitic energy and working capacities for bon dioxide.
other zeolite structures, examples of 24 of which are presented in For each adsorption isotherm, the monolayer capacity and
the Supplementary Information, Figs. S1 and S2. As another conse- Henry coefficient can be obtained by fitting the computed data
quence this leads to a low regenerability of CO2, given as points to the expression (22). For a temperature dependency of
the Henry coefficient, we used a known relationship relating the
DnCO2 ndes
CO latter to the excess molar Gibbs free energy (chemical potential
Rg ¼ ¼ 1  ads2 ð20Þ
ads
nCO2 nCO2 l) in the limit of low concentration:
 l
Low regenerability from the perspective of large-scale industrial kCO2 ¼ c  exp  ð23Þ
RT
applications is not economical, because it is directly related to the
amount of remaining (not captured) CO2. The trend that conditions where c and l are fitting parameters.
leading to the lowest parasitic energy lead to lower working capac- There are a number of different flexible functions which we
ity and low regenerability holds in all three presented cases. tested for evaluation of dependency of saturation capacity with
The parasitic energy computed within the present approach is respect to temperature, and in the end a simple exponential form
the theoretical minimum implying establishment of the thermody- similar to that used for the Henry coefficient was found to provide
namic equilibrium both at the adsorption and at the desorption good results for various categories of adsorption isotherms
points, which requires a certain amount of time depending on  
b
the adsorption kinetics. Too small working capacity would require nmono ¼ a  exp ð24Þ
T
too many adsorption-desorption cycles. That is why it seems rea-
sonable to set an acceptable minimum of the working capacity, where a and b are fitting parameters. Since the first fitting parame-
and after that define conditions of the parasitic energy minimum ter ðaÞ has the unit of mmol=g, it can be interpreted as a monolayer
according to this constraint. Moreover, besides setting a minimum (saturation) capacity of material at infinite temperature. Physical
of the working capacity, it is even more important from the envi- meaning of the parameter b can be interpreted then as a chemical
ronmental point of view to set an acceptable minimum for the potential (divided by the gas constant R) corresponding to a satu-
regenerability, as well as the purity of captured CO2. In the follow- rated state of the adsorbed monolayer.
ing, we will define a series of the working capacities, starting from Combining Eqs. (22)–(24), a final expression for the single site
0.25, up to 4.0 mmolg1 and for each given working capacity cal- loading nðp; TÞ function can be presented as follows:
B. Vujic, A.P. Lyubartsev / Chemical Engineering Science 174 (2017) 174–188 181

   l
b c exp  RT pCO2 an analytical approach from the fitted data can be found in the
nCO2 ðpCO2 ; TÞ ¼ a exp  l ð25Þ
T 1 þ c exp  RT pCO2 Supplementary Information.
Examples of fitting simulation adsorption isotherm of CO2 by
In cases when single site Langmuir equation did not provide a using Eq. (25) for the three categories of adsorption isotherms
good fitting of the adsorption isotherms, a dual-site Langmuir are shown in Fig. 4. Fitting lines show generally a very good agree-
model was used with two additional fitting parameters (b and l0 ):
0
ment with the simulation points in the whole considered range of
2 0  3 pressures and temperatures. Some deviations between fitted and
b l  l0
exp  exp bT  RT simulation results can be seen for adsorption isotherms of category
n00CO2 ¼ acpCO2 4 T RT 
l þ  l0  5 ð26Þ
1 þ c exp  RT pCO2 1 þ c exp  RT pCO2 one (zeolites showing a large Henry coefficient value). Although
some errors in the calculated parasitic energy with the use of fitted
where n00CO2 (as a function of pCO2 and T) indicates that the dual site isotherms thus can appear, they would be noticeable mostly for
low desorption pressures, much below 2 kPa, which are of less
Langmuir equation is used.
interest because of large parasitic energy corresponding to such
To determine whether we should use a single or a dual site
desorption points due to large vacuum work. From our experience,
Langmuir model we compared the errors of both models:
most zeolites have shown the minimum of parasitic energy for
N  2
1X nj  nsim;j capturing at least 1 mmol/g of CO2 at partial CO2 desorption pres-
erri ¼ ð27Þ sures above 1 kPa. Nonetheless, these deviations in predicted par-
N j¼1 nsim;j
asitic energy appears only for zeolites with a large Henry
coefficient, while in other cases Eq. (25) provides a good agreement
where erri is the error of dual or single Langmuir model, N is num-
ber of data points included, nj is loading predicted using dual or sin- between GCMC simulations and fitted results.
Eq. (25) can be used not only for fitting of simulated adsorption
gle model and nsim;j is the gas loading obtained directly from
isotherms, Fig. 4, but also for analyses based on experimental data.
molecular simulation. We have used the single site expression in
An example of fitting experimental adsorption isotherms of CO2 in
cases when dual site fitting did not provide better than 5% improve-
ment compared to the single site model.
In all considered cases, N2 adsorption isotherms were found to
be of category 3 (Henry-like). That is why the nitrogen loading as a
function of N2 partial pressure was described by the Henry law
equation

nN2 ðpN2 ; TÞ ¼ kN2  pN2 ð28Þ

where kN2 is the Henry coefficient of nitrogen adsorption, pN2 is par-


tial pressure of nitrogen, and nN2 is loading of nitrogen at adsorption
or desorption point. As in the case of CO2 adsorption, here we
implemented the temperature dependency of kN2 in the form of
 
0 T
kN2 ¼ kN2 exp  ð29Þ
dN2

where kN2 ¼ 1 mmolg1 kPa1 is a constant ensuring the correct


0

units and being the same for all zeolite structures, and dN2 is a fit-
ting parameter. Thus, loading of nitrogen at any temperature and
partial N2 pressure below 85 kPa is predicted with only two fitting
Fig. 5. An example of fitting experimental (Ohlin et al., 2013) adsorption isotherms
parameters.
of CO2 in Na-ZSM-5, measured at 308, 323, 358 and 393 K. The fitting lines were
An example of evaluation of the fitting parameters of the single- obtained with fitting parameters in Eq. (25): a ¼ 5:0  102 [mmolg1],
site model for NES zeolite, and evaluation of the parasitic energy by b ¼ 1234:39 [K], c ¼ 5:0  107 [kPa1], l ¼ 25:65 [kJmol1].

Fig. 4. Examples of fitting simulation adsorption isotherm of CO2 using the proposed model, for the three categories of adsorption isotherms. Representative zeolites have the
following framework code: JOZ, NES and SAS. Si:Al ratio in all three cases were around 2. For JOZ zeolite a dual site adsorption isotherm is used, were for other two structures
a single site isotherm was used instead. Markers: GCMC simulation results, lines: fitting according to Eq. (25).
182 B. Vujic, A.P. Lyubartsev / Chemical Engineering Science 174 (2017) 174–188

Na-ZSM-5 shown in Fig. 5 demonstrates almost perfect matching can not be considered as efficient, even though parasitic energy
to the experimentally measured data. might have a low value at that specific point.
Besides the given example, we also tested the fitting model on In the next selection we have analyzed available zeolite struc-
experimental data from other works (Moura et al., 2015; Cheung tures with respect to the minimum of the parasitic energy at a
and Hedin, 2014). In overall the fitting model was found to be very given working capacity (1, 2 and 3 mmolg1) and the purity of
well applicable. The fitted adsorption isotherms together with the the captured CO2, and displayed the statistics on Fig. 7a. Data for
heat of adsorption, measured in experiments, can also be used for other working capacities are given in Figs. S3–S7 of the Supple-
evaluation of the parasitic energy. This can be used as a further val- mentary Information. Each point in Fig. 7a represents the mini-
idation of the adsorbents which were found to be efficient from the mum of the parasitic energy obtained by a given zeolite at
computer screening point of view. condition of the given working capacity, relative to the purity of
captured CO2 when the desorption conditions correspond to the
minimum of the parasitic energy. As the required working capacity
3.3. Working capacity and purity of captured CO2
increases (from 1 up to 3 mmolg1), the number of zeolite struc-
tures is decreasing. Another observation is that purity of captured
The first step in the process of screening zeolite frameworks is
CO2 at the minimum of parasitic energy for a certain working
to make a selection of zeolite structures based on a required work-
capacity is increasing as the Si:Al ratio decreases. Therefore, con-
ing capacity. Theoretically the maximal working capacity achiev-
sidering parasitic energy with two additional constraints, working
able for a given structure is equal to the amount of adsorbed gas
capacity and the purity of captured CO2, zeolites with Si:Al ratio of
at the adsorption point. Since the vacuum work given in Eq. (14)
2 perform best.
exponentially increases with respect to decrease in desorption
In the following we accept the purity of 95% (meaning that 95%
pressure, the total parasitic energy related to the maximal working
of captured gas is CO2 and only 5% of the N2) to be an acceptable
capacity becomes significantly high. Thus, in the process of selec-
level.
tion of potentially good CO2 adsorbents, working capacity is an
important parameter determining on the one hand the amount
of the captured gas, and on the other hand strongly affecting the 3.4. Parasitic energy and regenerability of CO2
parasitic energy. A screening of zeolite structures is made for work-
ing capacities of CO2 in steps of 0.25 mmolg1, Fig. 6. One more important aspect in the process of defining the opti-
Our computations show that a rather high percentage of zeolite mal CO2 capture working condition is to include regenerability as
structures within a given Si:Al ratio are able to capture at least an additional constraint. Regenerability of CO2, calculated accord-
0.25 mmolg1 of CO2. As the working capacity increases, it inevita- ing to Eq. (20), is a direct measure of the amount CO2 left not cap-
bly leads to decrease in the number of zeolites that can achieve this tured, and from the environmental point of view it is of primary
working capacity. From Fig. 6 a conclusion can be drawn, that the importance to have a good regenerability. Here we assume the
higher the Si:Al ratio is, the lower is the number of zeolite struc- minimum of Rg being equal to 90%, which means that at least
tures for a given working capacity. One can also see in Fig. 6 that 90% of CO2 from flue gases has to be captured.
for intermediate range of working capacities (0:5 6 DnCO2 6 Fig. 7b presents the minimum of the parasitic energy for each
1:0 mmolg1), zeolites with Si:Al = 7 have the highest number of specific zeolite structure as a function of the corresponding Rg at
structures. Structures with Si:Al = 5 performs better for working a given minimum value of the working capacity. Additional plots
capacity in the range of 1.25–1.75 mmolg1. For even higher can be found in the Supplementary Information, Figs. S8–S12.
working capacity, zeolites with Si:Al = 3 and less performs best. The evaluated minimum of Ep clearly increases with increase in
In other words, the higher the aluminum content, the higher the Rg . The lowest unconstrained parasitic energy within the consid-
maximal working capacity which can be achieved with a given ered model is around 540 kJ(kg CO2)1. However, the minimum
structure. Thus, if we need to achieve a working capacity of of Ep within constraints DnCO2 P 2:0 mmolg1 and Rg P 90% is
4 mmolg1, the only choice would be Si:Al = 2, since there are four about 770 kJ(kg CO2)1, which is ca. 230 kJ(kg CO2)1 higher than
zeolite structures with Si:Al = 2 which can achieve this working the theoretical minimum.
capacity. Fig. 8 shows histograms of the number of zeolite structures
Another important factor to consider in CCS is the purity of cap- with respect to P r of captured CO2 for zeolites with working capac-
tured CO2. If the purity of captured CO2 at certain working condi- ity above 1.5 mmolg1 (a detailed procedure of the selection of
tions (desorption pressure and temperature) is poor the process structures and extended data for other working capacities are pre-
sented in Section S6, Figs. S13–S28 of the Supplementary Informa-
tion). The number of zeolites within each selection satisfying also
criteria Rg 6 90% is given on each plot. The previously made selec-
tion of zeolite structures shown in Fig. 6 changes in such a way that
the number of zeolites within a given Si:Al ratio decreases with
each additional constraint (Pr and Rg ). To demonstrate our reason-
ing we give Si:Al = 5 as an example, Fig. 8. The number of zeolites
within this particular Si:Al ratio that are able to capture at least
1.5 mmolg1 of CO2 is equal to 122. 23 of them satisfy condition
for the purity Pr 6 95%. Yet, there are only five zeolites having
Emin:
p 6 1000 kJ(kg CO2)1 and Pr P 95%, and none if condition
for the regenerability Rg 6 90% is also included.
To illustrate further the relation between Rg and Pr for the cor-
responding minimum of Ep at a specified working capacity, we pre-
sent statistics in Fig. 9. Data for other minimal working capacities
are given in the Supplementary Information, Figs. S29–S44. Each
Fig. 6. Number of zeolite structures within a given Si:Al ratio, which are able to
capture a certain amount of CO2. The total number of zeolites within each Si:Al ratio
specific point on a graph is colored according to the actual value
is equal to 225. of parasitic energy (Emin:
p ) computed at the only constraint of work-
B. Vujic, A.P. Lyubartsev / Chemical Engineering Science 174 (2017) 174–188 183

Fig. 7. (A) Minimum of parasitic energy (Ep ) in relation to purity (P r ) of captured CO2 for three working capacities. For more data refer to SI related with this article. (B) The
minimum of parasitic energy in relation to regenerability of CO2 for three working capacities is presented.

Fig. 8. Histogram of distribution of zeolite structures within a certain Si:Al ratio versus purity of captured CO2, for DnCO2 P 1:5 mmolg1 and Rg P 90%. Value for P r relates to
the working condition (pdes ; T des ) at which Ep have a minimum for a given working capacity. Dark blue areas correspond to the number of zeolites within additional condition
Ep 6 1000 kJ(kg CO2)1.

ing capacity. The minimum of parasitic energy for a given working then excluded structures which, for the desorption conditions cor-
capacity does not necessarily provide acceptable values of Rg and responding to the minimum of parasitic energy did not provide
P r for CO2 capture. Inevitably, increased requirements of regenera- requirements for the minimal purity of captured CO2 and/or regen-
bility and purity of captured CO2 decrease substantially the num- erability. Still, for many structures the last two criteria can be sat-
ber of suitable structures. isfied by choosing other desorption conditions with expense of the
parasitic energy. From the point of view of industrial applications,
3.5. Procedure for evaluation of the most efficient adsorbents for all four criteria are important: working capacity, regenerability,
industrial applications purity of captured CO2 and finally the parasitic energy. We there-
fore made selection of the most efficient adsorbents according to
In the previous subsections we evaluated, for each structure, the the following procedure. For each zeolite structure, we generated
minimum of Ep under constraint of a given working capacity, and about 105 desorption conditions, considering pdes from 0.1 kPa to
184 B. Vujic, A.P. Lyubartsev / Chemical Engineering Science 174 (2017) 174–188

Fig. 9. Relation between regenerability and purity of CO2 for DnCO2 P 1:5 mmolg1. Within a given Si:Al ratio, each scatter point represents one zeolite structure. Points are
colored according to the corresponding value of the minimum parasitic energy for a given working capacity (see secondary axis).

a point at which DnCO2 tends to zero (maximum 100 kPa), and T des
in a slightly wider range from 313 up to 473 K (previously given
from 313 to 413 K). Among all the desorption points for each zeo-
lite framework, we selected only those that satisfied given require-
ments in terms of DnCO2 ; P r and Rg . Then the parasitic energy was
calculated for the remaining set of desorption points and the min-
imum of Ep was determined.
An example of two-dimensional maps showing dependences of
the parasitic energy, working capacity, regenerability and purity on
the desorption pressure and temperature for a FAU zeolite is
shown in Fig. 10. One can see that fulfilling of the condition for a
low parasitic energy requires lower desorption temperatures while
conditions for high working capacity, regenerability and purity
require higher desorption temperatures. Furthermore, the lines Fig. 11. Number of zeolite structures within given Si:Al ratio, able to capture a
with equal parasitic energy in Fig. 10a reach maximum tempera- certain amount of CO2 satisfying requirements for the purity P r P 95% and
ture in the desorption pressure range 10–20 kPa. This mean that regenerability Rg P 90%.
the desorption conditions providing minimum parasitic energy
under specified constraint conditions correspond to the partial Rg P 90% for different values of working capacity is shown in
pressure which is similar to that at the adsorption point, and Fig. 11. One can note considerable decrease of the number of struc-
higher temperature. This observation, made for the case of FAU tures compared with data in Fig. 6 where conditions for the purity
zeolite, holds also for other structures showing good adsorption and regenerability were not set.
properties: the minimum of parasitic energy determined in this The minimum of the parasitic energy for CCS in the current
approach tends towards higher desorption temperature and higher state of art industrial applications gives an estimate of 1060 kJ
pressure, ultimately to TSA. per kg CO2 (Lin et al., 2012; Huck et al., 2014; Berger and Bhown,
The total number of zeolite frameworks within each Si:Al ratio 2013). Previous works reported possibility of many hundreds
that satisfied conditions for the purity Pr P 95% and regenerability structures within this energy penalty. The number of zeolites with

Fig. 10. Two-dimensional maps of parasitic energy (A), working capacity (B), purity (C) and regenerability (D) as functions of the desorption temperature and desorption
pressure, for CO2 capture in FAU zeolite with Si:Al = 2.
B. Vujic, A.P. Lyubartsev / Chemical Engineering Science 174 (2017) 174–188 185

Pr ¼ 97% and Rg ¼ 91%. The reason is because FAU zeolite shows


isotherm of category 1 for CO2 adsorption, meaning that although
the amount of adsorbed CO2 at the adsorption conditions is high, it
is rather difficult and energy demanding to regenerate the material
after the adsorption, and thus met required conditions in working
capacity and regenerability, as presented in Fig. 10. The parasitic
energy value associated with the three examples on Fig. 12 is in
the range  860—960 kJ(kg CO2)1, which in complement to previ-
ously reported values (Lin et al., 2012; Berger and Bhown, 2013)
gives a realistic estimate of the minimal energy penalty in the
CCS process.

4. Conclusion

In this work we have developed a new method for thermody-


namic evaluation of the theoretical minimum of the energy
expenses in the CCS process under realistic industrial conditions.
The method provides a fast evaluation of the amount of captured
CO2 and N2, as well as the energetically optimal adsorption and
desorption conditions under specified requirements for the work-
ing capacity, regenerability of he sorbent framework and purity
of captured CO2. The minimum of thermodynamic energy expenses
in a hybrid pressure–temperature swing CO2 capture process (or in
a pure pressure or temperature swing process) thus can be evalu-
ated within the given constraints. Mechanical and thermal stability
Fig. 12. Zeolite structures which provide the lowest parasitic energy under
of a zeolite framework, energy losses of non-thermodynamics ori-
requirements for the minimal working capacity (3.5 mmolg1), purity 95% and gin, as well as kinetics of the adsorption and sorbent regeneration
regenerability 90%. Exact values of the parasitic energy, working capacity, purity were out of the scope of this work.
and regenerability for these four zeolites are given in Table 1. Na+ ions are omitted. For a fast evaluation of adsorption isotherms in a temperature
Figure created using VMD (Humphrey et al., 1996; Stone et al., 2001; Varshney
range of interest we have developed a fitting model with four
et al., 1994; Sanner et al., 1995) (IRR, JSR) and Materials Studio Accelrys (Materials
Studio, 2013) (RWY, NPT) software. parameters for CO2 (or six using the dual site Langmuir model)
and two parameters for N2, which can be optimized from a small
number of Grand Canonical Monte Carlo calculations. The model
was tested on a large number of zeolite structures by comparing
different Si:Al ratios, which exhibits the minimum of Ep 6 1060 kJ
fitted data with detailed computation of adsorption isotherms by
(kg CO2)1 under conditions Pr P 95%; Rg P 90%, is given in the
Grand Canonical Monte Carlo simulations. It can be applied also
Figs. S45–S60 of the Supplementary Information. In comparison
for fitting of experimental adsorption data enabling evaluation of
to the total number of zeolites within a given Si:Al ratio, only a
the energy expenses in CCS using data from laboratory experimen-
small portion is capable of capturing CO2 under the specified
tal measurements.
conditions.
Within our approach, it is possible to evaluate the theoretical
All 225 known zeolite framework types, each with several com-
limit of the energy expenses for each specific material which can
positions (resulting with 1800 structures in total), have been tested
be used as an adsorbent in the CCS process, under specified in
in GCMC simulations to determine parameters for computation of
advance requirements for the working capacity, regenerability
adsorption isotherms and parasitic energy. According to our find-
and purity, and use the proposed method for screening of various
ings, the number of the most efficient structures is ultimately less
structures for the most efficient adsorbent. For this purpose we
than 1% of the total number of structures within each Si:Al ratio.
performed adsorption calculations for 225 known zeolite frame-
Examples of zeolite structures which have shown the lowest para-
work types varying composition in each of them, and created a
sitic energy under the given constraints of the working capacity,
short list of the structures with could be potentially good adsor-
purity and regenerability are presented in Fig. 12, whereas the
bents, which can be used for further analysis. It follows from our
exact values of the desorption conditions and parameters
observations that when conditions for the working capacity, regen-
DnCO2 ; Pr and Rg are summarized in Table 1.
erability and purity are set, the optimal from the equilibrium ther-
Surprisingly, a commonly used FAU zeolite is not found in the modynamics point of view desorption conditions tend towards
top of the most optimal structures since its lowest parasitic energy higher desorption temperature and higher pressure, ultimately to
is 1070 kJ per kg of captured CO2 for DnCO2 ¼ 3 mmolg1 ; TSA.

Table 1
Zeolite structures heaving the lowest parasitic energy for working capacity of at least 3.5 mmolg1 CO2, purity 95% and regenerability 90%.

Zeolite framework Si:Al ratio Ep T des pdes DnCO2 Pr Rg


[kJ(kg CO2)1] [K] [kPa] [mmolg1] [%] [%]
RWY 2 856.4 440 14.30 4.39 96.05 90.06
JSR 2 875.4 472 10.10 4.41 96.91 90.02
IRR 2 915.5 479 10.05 4.17 97.93 90.04
NPT 5 955.9 492 20.10 3.55 95.79 90.00
186 B. Vujic, A.P. Lyubartsev / Chemical Engineering Science 174 (2017) 174–188

The development method can be further extended for studies of modeling of fixed-bed adsorption. Sep. Sci. Technol. 45 (1), 73–84. http://dx.doi.
org/10.1080/01496390903401762.
water effect on CO2 capture by generalizing the model to ternary
Dantas, T., Rodrigues, A., Moreira, R., 2012. Separation of carbon dioxide from flue
mixtures CO2/N2/H2O, to other sorbent materials such as metallo- gas using adsorption on porous solids, greenhouse gases – capturing, utilization
organic frameworks, or for separation of other gas mixtures such and reduction, Dr Guoxiang Liu (Ed.), InTech. http://dx.doi.org/10.5772/31917.
as CO2/CH4. URL <http://www.intechopen.com/books/greenhouse-gases-capturing-
utilization-and-reduction/separation-of-carbon-dioxide-from-flue-gas-using-
adsorption-on-porous-solids>.
Appendix A. Supplementary material Deem, M.W., Pophale, R., Cheeseman, P.A., Earl, D.J., 2009. Computational discovery
of new zeolite-like materials. J. Phys. Chem. C 113 (51), 21353–21360. http://dx.
doi.org/10.1021/jp906984z.
Supplementary data associated with this article can be found, in Deroche, I., Gaberova, L., Maurin, G., Llewellyn, P., Wright, C.M.P., 2008. Adsorption
the online version, at http://dx.doi.org/10.1016/j.ces.2017.09.006. of carbon dioxide in SAPO STA-7 and AlPO-18: grand canonical Monte Carlo
simulations and microcalorimetry measurements. Adsorption 14, 207–213.
Dubbeldam, D., Snurr, R.Q., 2007. Recent developments in the molecular modeling
References of diffusion in nanoporous materials. Mol. Simul. 33 (4–5), 305–325. http://dx.
doi.org/10.1080/08927020601156418.
Akhtar, F., Bergström, L., 2011. Colloidal processing and thermal treatment of Dubbeldam, D., Torres-Knoop, A., Walton, K.S., 2013. On the inner workings of
binderless hierarchically porous zeolite 13X monoliths for CO2 capture. J. Am. Monte Carlo codes. Mol. Simul. 39 (14–15), 1253–1292. http://dx.doi.org/
Ceram. Soc. 94 (1), 92–98. http://dx.doi.org/10.1111/j.1551-2916.2010.04044.x. 10.1080/08927022.2013.819102.
Akkermans, R.L., Spenley, N.A., Robertson, S.H., 2013. Monte Carlo methods in Dubbeldam, D., Calero, S., Ellis, D.E., Snurr, R.Q., 2016. RASPA: molecular simulation
materials studio. Mol. Simul. 39 (14–15), 1153–1164. http://dx.doi.org/ software for adsorption and diffusion in flexible nanoporous materials. Mol.
10.1080/08927022.2013.843775. Simul. 42 (2), 81–101. http://dx.doi.org/10.1080/08927022.2015.1010082.
Babarao, R., Jiang, J., 2008. Molecular screening of metal-organic frameworks for Feron, P., 2010. Exploring the potential for improvement of the energy performance
CO2 storage. Langmuir 24, 6270–6278. of coal fired power plants with post-combustion capture of carbon dioxide. Int.
Bacsik, Z., Atluri, R., Garcia-Bennett, A., Hedin, N., 2010. Temperature-induced J. Greenh. Gas Contr. 4, 152–160.
uptake of CO2 and formation of carbonates in mesocaged silica modified with n- Frenkel, D., Smit, B., 2002. Understanding Molecular Simulations: from Algorithms
propylamines. Langmuir 26, 10013–10024. to Application. Academic Press, San Diego.
Bacsik, Z., Cheung, O., Vasiliev, P., Hedin, N., 2016. Selective separation of CO2 and García-Pérez, E., Parra, J., Ania, C., García-Sánchez, A., van Baten, J., Krishna, R.,
CH4 for biogas upgrading on zeolite NaKA and sapo-56. Appl. Energy 162, 613– Dubbeldam, D., Calero, S., 2007. A computational study of CO2, N2, and CH4
621. http://dx.doi.org/10.1016/j.apenergy.2015.10.109. URL <http:// adsorption in zeolites. Adsorption 13 (5–6), 469–476.
www.sciencedirect.com/science/article/pii/S0306261915013458>. García-Sánchez, A., Ania, C., Parra, J., Dubbeldam, D., Vlugt, T., Krishna, R., Calero, S.,
Baerlocher, C., McCusker, L., Olson, D., 2007. Atlas of Zeolite Framework Types. 2009. Transferable force field for carbon dioxide adsorption in zeolites. J. Phys.
Elsevier, Amsterdam. Chem. C 113 (20), 8814–8820. http://dx.doi.org/10.1021/jp810871f.
Bao, Y., Martin, R.L., Simon, C.M., Haranczyk, M., Smit, B., Deem, M.W., 2015. In silico Gomez-Alvarez, P., Hamad, S., Haranczyk, M., Ruiz-Salvador, A.R., Calero, S., 2016.
discovery of high deliverable capacity metal-organic frameworks. J. Phys. Chem. Comparing gas separation performance between all known zeolites and their
C 119 (1), 186–195. http://dx.doi.org/10.1021/jp5123486. zeolitic imidazolate framework counterparts. Dalton Trans. 45, 216–225. http://
Bastin, L., Brcia, P., Hurtado, E., Silva, J., Rodrigues, A., Chen, B., 2008. A microporous dx.doi.org/10.1039/C5DT04012D.
metal-organic framework for separation of CO2/N2 and CO2/CH4 by fixed-bed Gupta, A., Chempath, S., Sanborn, M.J., Clark, L.A., Snurr, R.Q., 2003. Object-oriented
adsorption. J. Phys. Chem. C 112, 1575–1581. programming paradigms for molecular modeling. Mol. Simul. 29 (1), 29–46.
Benson, S., Orr, F., 2008. Carbon dioxide capture and storage. MRS Bull. 33, 303–305. http://dx.doi.org/10.1080/0892702031000065719.
Berger, A.H., Bhown, A.S., 2013. Optimizing solid sorbents for CO2 capture. Energy Harlick, P.J., Tezel, F.H., 2004. An experimental adsorbent screening study for CO2
Proc. 37 (0), 25–32. http://dx.doi.org/10.1016/j.egypro.2013.05.081. gHGT-11. removal from N2. Micropor. Mesopor. Mater. 76, 71–79.
URL <http://www.sciencedirect.com/science/article/pii/S187661021300091X>. Hasan, M.M.F., First, E.L., Floudas, C.A., 2013. Cost-effective CO2 capture based on in
Berlier, M.F.K., 1997. Adsorption of CO2 on microporous materials. 1. On activated silico screening of zeolites and process optimization. PCCP 15, 17601–17618.
carbon and silica gel. J. Chem. Eng. Data 42, 533–537. http://dx.doi.org/10.1039/C3CP53627K.
Bhown, A., Freeman, B., 2011. Analysis and status of post-combustion carbon Hedin, N., Chen, L., Laaksonen, A., 2010. Sorbents for CO2 capture from flue gas-
dioxide capture technologies. Environ. Sci. Technol. 45, 8624–8632. aspects from materials and theoretical chemistry. Nanoscale 2, 1819–1841.
Čejka, J., van Bekkum, H., Corma, A., Schüth, F., 2007. Introduction to Zeolite Science http://dx.doi.org/10.1039/C0NR00042F.
and Practice. Elsevier, Amsterdam. Hedin, N., Andersson, L., Bergström, L., Yan, J., 2013. Adsorbents for the post-
Chempath, S., Düren, T., Sarkisov, L., Snurr, R.Q., 2013. Experiences with the publicly combustion capture of CO2 using rapid temperature swing or vacuum swing
available multipurpose simulation code. Music Mol. Simul. 39 (14–15), 1223– adsorption. Appl. Energy 104 (0), 418–433. http://dx.doi.org/10.1016/j.
1232. http://dx.doi.org/10.1080/08927022.2013.819103. apenergy.2012.11.034. URL <http://www.sciencedirect.com/science/article/pii/
Cheung, O., Hedin, N., 2014. Zeolites and related sorbents with narrow pores for CO2 S0306261912008276>.
separation from flue gas. RSC Adv. 4, 14480–14494. http://dx.doi.org/10.1039/ Helwani, M., Wiheeb, A., Kim, J., Othman, M., 2012. Improved carbon dioxide
C3RA48052F. capture using metal reinforced hydrotalcite under wet conditions. Int. J. Greenh.
Cheung, O., Liu, Q., Bacsik, Z., Hedin, N., 2012. Silicoaluminophosphates as CO2 Gas Contr. 7, 127–136.
sorbents. Micropor. Mesopor. Mater. 156 (0), 90–96. http://dx.doi.org/10.1016/ Hemingway, B., Robie, R., 1984. Thermodynamic properties of zeolites: low-
j.micromeso.2012.02.003. URL <http://www.sciencedirect.com/science/article/ temperature heat capacities and thermodynamic functions for phillipsite and
pii/S1387181112000625>. clinoptilolite. Estimates of the thermochemical properties of zeolitic water at
Choi, S., Drese, J., Jones, C., 2009. Adsorbent materials for carbon dioxide capture low temperature. Am. Mineral. 69 (7–8), 692–700. URL <http://www.
from large anthropogenic point sources. ChemSusChem 2 (9), 796–854. http:// minsocam.org/ammin/AM69/AM69_692.pdf>.
dx.doi.org/10.1002/cssc.200900036. Huck, J.M., Lin, L.-C., Berger, A.H., Shahrak, M.N., Martin, R.L., Bhown, A.S.,
Chue, K.T., Kim, J.N., Yoo, Y.J., Cho, S.H., Yang, R.T., 1995. Comparison of activated Haranczyk, M., Reuter, K., Smit, B., 2014. Evaluating different classes of
carbon and zeolite 13X for CO2 recovery from flue gas by pressure swing porous materials for carbon capture. Energy Environ. Sci. 7, 4132–4146.
adsorption. Ind. Eng. Chem. Res. 34 (2), 591–598. http://dx.doi.org/10.1021/ http://dx.doi.org/10.1039/C4EE02636E.
ie00041a020. Humphrey, W., Dalke, A., Schulten, K., 1996. VMD – visual molecular dynamics. J.
Conway, W., Beyad, Y., Richner, G., Puxty, G., Feron, P., 2015. Rapid {CO2} absorption Mol. Graph. 14, 33–38.
into aqueous benzylamine (BZA) solutions and its formulations with Jensen, N.K., Rufford, T.E., Watson, G., Zhang, D.K., Chan, K.I., May, E.F., 2012.
monoethanolamine (MEA), and 2-amino-2-methyl-1-propanol (AMP) as Screening zeolites for gas separation applications involving methane, nitrogen,
components for post combustion capture processes. Chem. Eng. J. 264 (0), and carbon dioxide. J. Chem. Eng. Data 57, 106–113.
954–961. http://dx.doi.org/10.1016/j.cej.2014.11.040. URL <http:// Kadoura, A., Nair, A.K.N., Sun, S., 2016. Adsorption of carbon dioxide, methane, and
www.sciencedirect.com/science/article/pii/S1385894714014958>. their mixture by montmorillonite in the presence of water. Micropor.
D’Alessandro, D., Smit, B., Long, J., 2010. Carbon dioxide capture: prospects for new Mesopor. Mater. 225, 331–341. http://dx.doi.org/10.1016/j.
materials. Angew. Chem. Int. Ed. 49 (35), 6058–6082. http://dx.doi.org/10.1002/ micromeso.2016.01.010. URL <http://www.sciencedirect.com/science/article/
anie.201000431. pii/S1387181116000202>.
Damen, K., van Troost, M., Faaij, A., Turkenburg, W., 2006. A comparison of Karasawa, N., Goddard, W.A., 1989. Acceleration of convergence for lattice sums. J.
electricity and hydrogen production systems with CO2 capture and storage. Part Phys. Chem. 93 (21), 7320–7327. http://dx.doi.org/10.1021/j100358a012.
A: Review and selection of promising conversion and capture technologies. Kim, J., Lin, L., Swisher, J., Haranczyk, M., Smit, B., 2012. Predicting large CO2
Prog. Energy Combust. Sci. 32, 215–246. URL <http://www.sciencedirect. adsorption in aluminosilicate zeolites for postcombustion carbon dioxide
com/science/article/pii/S0360128505000626>. capture. J. Am. Chem. Soc. 134, 18940–18943.
Dantas, T.L.P., Amorim, S.M., Luna, F.M.T., Silva Jr., I.J., de Azevedo, D.C.S., Rodrigues, Krishna, R., Long, J., 2011. Screening metal-organic frameworks by analysis of
A.E., Moreira, R.F.P.M., 2009. Adsorption of carbon dioxide onto activated carbon transient breakthrough of gas mixtures in a fixed bed adsorber. J. Phys. Chem. C
and nitrogen-enriched activated carbon: surface changes, equilibrium, and 115, 12941–12950. URL <http://pubs.acs.org/doi/abs/10.1021/jp202203c>.
B. Vujic, A.P. Lyubartsev / Chemical Engineering Science 174 (2017) 174–188 187

Leach, A., 2001. Molecular Modelling, Principles and Applications. Pearson Ohlin, L., Bazin, P., Thibault-Starzyk, F., Hedlund, J., Grahn, M., 2013. Adsorption of
Education. CO2, CH4, and H2O in zeolite ZSM-5 studied using in situ ATR-FTIR spectroscopy.
Lee, K., Howe, J.D., Lin, L.-C., Smit, B., Neaton, J.B., 2015. Small-molecule adsorption J. Phys. Chem. C 117 (33), 16972–16982. http://dx.doi.org/10.1021/jp4037183.
in open-site metal organic frameworks: a systematic density functional theory Patz, J.A., Campbell-Lendrum, D., Holloway, T., Foley, J.A., 2005. Impact of regional
study for rational design. Chem. Mater. 27 (3), 668–678. http://dx.doi.org/ climate change on human health. Nature 438 (7066), 310–317. URL <http://
10.1021/cm502760q. www.nature.com.ezp.sub.su.se/nature/journal/v438/n7066/full/nature04188.
Lin, L., Berger, A., Martin, R., Kim, J., Swisher, J., Jariwala, K., Rycroft, C., Bhown, A., html>.
Deem, M., Haranczyk, M., Smit, B., 2012. In silico screening of carbon-capture Paustian, K., Cole, C.V., Sauerbeck, D., Sampson, N., 1998. CO2 mitigation by
materials. Nat. Mater. 11, 633–641. agriculture: an overview. Clim. Change 40 (1), 135–162. http://dx.doi.org/
Linstrom, P.J., Mallard, W.G. (Eds.), 2005. NIST Chemistry WebBook, NIST Standard 10.1023/A:1005347017157.
Reference Database Number 69. National Institute of Standards and Technology, Pellerano, M., Pré, P., Kacem, M., Delebarre, A., 2009. CO2 capture by adsorption on
Gaithersburg MD, 20899, 2005. URL <http://webbook.nist.gov>. activated carbons using pressure modulation. Energy Proc. 1, 647–653.
Liu, Q., Mace, A., Bacsik, Z., Sun, J., Laaksonen, A., Hedin, N., 2010. NaKA sorbents Peng, D.-Y., Robinson, D.B., 1976. A new two-constant equation of state. Ind. Eng.
with high CO2-over-N2 selectivity and high capacity to adsorb CO2. Chem. Chem. Fundam. 15 (1), 59–64. http://dx.doi.org/10.1021/i160057a011.
Commun. 46, 4502–4504. http://dx.doi.org/10.1039/C000900H. Pfaff, I., Kather, A., 2009. Comparative thermodynamic analysis and integration
Liu, Z., Grande, C.A., Li, P., Yu, J., Rodrigues, A.E., 2011. Adsorption and desorption of issues of CCS steam power plants based on oxy-combustion with cryogenic or
carbon dioxide and nitrogen on zeolite 5a. Sep. Sci. Technol. 46 (3), 434–451. membrane based air separation. Energy Proc. 1 (1), 495–502. http://dx.doi.org/
http://dx.doi.org/10.1080/01496395.2010.513360. 10.1016/j.egypro.2009.01.066 (Greenhouse Gas Control Technologies 9
Löwenstein, W., 1954. The distribution of aluminum in the tetrahedra of silicates Proceedings of the 9th International Conference on Greenhouse Gas Control
and aluminates. Am. Miner. 39, 92–96. Technologies (GHGT-9), 16–20 November 2008, Washington DC, USA). URL
Martin, M.G., 2013. Mcccs towhee: a tool for Monte Carlo molecular simulation. <http://www.sciencedirect.com/science/article/pii/S1876610209000678>.
Mol. Simul. 39 (14–15), 1212–1222. http://dx.doi.org/10.1080/ Pophale, R., Cheeseman, P., Deem, M., 2011. A database of new zeolite-like
08927022.2013.828208. materials. PCCP 13, 12407–12412. http://dx.doi.org/10.1039/C0CP02255A.
Martin-Calvo, A., Parra, J.B., Ania, C.O., Calero, S., 2014. Insights on the anomalous Purton, J., Crabtree, J., Parker, S., 2013. Dl_monte: a general purpose program for
adsorption of carbon dioxide in LTA zeolites. J. Phys. Chem. C 118 (44), 25460– parallel Monte Carlo simulation. Mol. Simul. 39 (14–15), 1240–1252. http://dx.
25467. http://dx.doi.org/10.1021/jp507431c. doi.org/10.1080/08927022.2013.839871.
Massood, R., Timothy, J., Yansakala, N., Liljedahl, G., 2007. Carbon Dioxide Capture Ribeiro, R., Sauer, T., Lopes, F., Moreira, R., Grande, C., Rodrigues, A., 2008.
from Existing Coal-Fired Power Plants. National Energy Technology Laboratory, Adsorption of CO2, CH4, and N2 in activated carbon honeycomb monolith. J.
US Department of Energy. Chem. Eng. Data 53, 2311–2317.
Materials Studio, 2013. Materials Studio, Accelerys Software Inc., San Diego. Sanner, M., Olsen, A., Spehner, J.-C., 1995. Fast and robust computation of molecular
Matito-Martos, I., Martin-Calvo, A., Gutierrez-Sevillano, J.J., Haranczyk, M., Doblare, surfaces. In: Proceedings of the 11th ACM Symposium on Computational
M., Parra, J.B., Ania, C.O., Calero, S., 2014. Zeolite screening for the separation of Geometry. ACM, New York, pp. C6–C7.
gas mixtures containing SO2 CO2 and CO. PCCP 16, 19884–19893. http://dx.doi. Sethia, G., Pillai, R.S., Dangi, G.P., Somani, R.S., Bajaj, H.C., Jasra, R.V., 2010. Sorption
org/10.1039/C4CP00109E. of methane, nitrogen, oxygen, and argon in ZSM-5 with different SiO2/Al2O3
Matito-Martos, I., Alvarez-Ossorio, J., Gutierrez-Sevillano, J.J., Doblare, M., Martin- ratios: grand canonical Monte Carlo simulation and volumetric measurements.
Calvo, A., Calero, S., 2015. Zeolites for the selective adsorption of sulfur Ind. Eng. Chem. Res. 49 (5), 2353–2362. http://dx.doi.org/10.1021/ie900280w.
hexafluoride. Phys. Chem. Chem. Phys. 17, 18121–18130. http://dx.doi.org/ Simon, C.M., Smit, B., Haranczyk, M., 2016. pyIAST: Ideal adsorbed solution theory
10.1039/C5CP02407B. (IAST) python package. Comput. Phys. Commun. 200, 364–380. http://dx.doi.
Maurin, G., Llewellyn, P.L., Bell, R.G., 2005. Adsorption mechanism of carbon dioxide org/10.1016/j.cpc.2015.11.016. URL <http://www.sciencedirect.com/science/
in faujasites: grand canonical Monte Carlo simulations and microcalorimetry article/pii/S0010465515004403>.
measurements. J. Phys. Chem. B 109 (33), 16084–16091. http://dx.doi.org/ Siriwardane, R., Shen, M., Fisher, E., Losch, J., 2005. Adsorption of CO2 on zeolites at
10.1021/jp052716s. moderate temperatures. Energy Fuels 19, 1153–1159.
McDonald, T.M., Mason, J.A., Kong, X., Bloch, E.D., Gygi, D., Dani, A., Crocellá, V., Smit, B., Reimer, J.R., Oldenburg, C.M., Bourg, I.C., 1986. Introduction to Carbon
Giordanino, F., Odoh, S.O., Drisdell, W.S., Vlaisavljevich, B., Dzubak, A.L., Poloni, Capture and Sequestration. Imperial College Press., London.
R., Schnell, S.K., Planas, N., Lee, K., Pascal, T., Wan, L.F., Prendergast, D., Neaton, J. Songolzadeh, M., Ravanchi, M., Soleimani, M., 2012. Carbon dioxide capture and
B., 2015. Cooperative insertion of CO2 in diamine-appended metal-organic storage: a general review on adsorbents. Int. Sci. Index 6 (10), 213–220. URL
frameworks. Nature 519 (7543), 303–308. URL <https://ezp.sub.su.se/login? <http://waset.org/Publications?p=70>.
url=http://search.ebscohost.com/login.aspx?direct=true&db=aph&AN=1016286 Stenqvist, B., Thuresson, A., Kurut, A., Vácha, R., Lund, M., 2013. Faunus – a flexible
21&site=ehost-live&scope=site>. framework for Monte Carlo simulation. Mol. Simul. 39 (14–15), 1233–1239.
McMichael, A.J., Woodruff, R.E., Hales, S., 2006. Climate change and human health: http://dx.doi.org/10.1080/08927022.2013.828207.
present and future risks. The Lancet 367 (9513), 859–869. http://dx.doi.org/ Stone, J., Gullingsrud, J., Grayson, P., Schulten, K., 2001. A system for interactive
10.1016/S0140-6736(06)68079-3. URL <http://www.sciencedirect.com/ molecular dynamics simulation. In: Hughes, J.F., Séquin, C.H. (Eds.), 2001 ACM
science/article/pii/S0140673606680793>. Symposium on Interactive 3D Graphics. ACM SIGGRAPH, New York, pp. 191–
Merel, J., Clausse, M., Meunier, F., 2008. Experimental investigation on CO2 post- 194.
combustion capture by indirect thermal swing adsorption using 13X and 5A Varshney, A., Brooks, F.P., Wright, W.V., 1994. Linearly scalable computation of
zeolites. Ind. Eng. Chem. Res. 47 (1), 209–215. http://dx.doi.org/10.1021/ smooth molecular surfaces. IEEE Comput. Graph. Appl. 14, 19–25.
ie071012x. Verdegaal, W.M., Wang, K., Sculley, J.P., Wriedt, M., Zhou, H.-C., 2016. Evaluation of
Metz, B., Davidson, O., de Coninck, H., Loos, M., Meyer, L., 2005. IPCC special report metal-organic frameworks and porous polymer networks for CO2-capture
on carbon dioxide capture and storage. In: Intergovernmental Panel on Climate applications. ChemSusChem 9 (6), 636–643. http://dx.doi.org/10.1002/
Change (IPCC). cssc.201501464.
Miles, L., Kapos, V., 2008. Reducing greenhouse gas emissions from deforestation Vermeulen, T., 2011. Knowledge Sharing Report. CO2 Liquid Logistics Shipping
and forest degradation: global land-use implications. Science 320 (5882), 1454– Concept (LLSC): Overall Supply Chain Optimization. Global CCS Institute,
1455. http://dx.doi.org/10.1126/science.1155358. arXiv:http://science. Anthony Veder, Vopak.
sciencemag.org/content/320/5882/1454.full.pdf URL <http://science. Vujić, B., Lyubartsev, A., 2016. Transferable force-field for modelling of CO2, N2, O2
sciencemag.org/content/320/5882/1454>. and Ar in all silica and Na+ exchanged zeolites. Modell. Simul. Mater. Sci. Eng. 24
Milly, P.C.D., Dunne, K.A., Vecchia, A.V., 2005. Global pattern of trends in streamflow (4), 045002. URL <http://stacks.iop.org/0965-0393/24/i=4/a=045002>.
and water availability in a changing climate. Nature 438 (7066), 347–350. URL Walton, K.S., Sholl, D.S., 2015. Predicting multicomponent adsorption: 50 years of
<https://ezp.sub.su.se/login?url=http://search.ebscohost.com/login.aspx? the ideal adsorbed solution theory. AIChE J. 61 (9), 2757–2762. http://dx.doi.
direct=true&db=aph&AN=18893099&site=ehost-live&scope=site>. org/10.1002/aic.14878.
Ming, T., deRichter, R., Shen, S., Caillol, S., 2016. Fighting global warming by Xu, C., Hedin, N., 2016. Ultramicroporous {CO2} adsorbents with tunable mesopores
greenhouse gas removal: destroying atmospheric nitrous oxide thanks to based on polyimines synthesized under off-stoichiometric conditions.
synergies between two breakthrough technologies. Environ. Sci. Pollut. Res. 23 Micropor. Mesopor. Mater. 222, 80–86. http://dx.doi.org/10.1016/j.
(7), 6119–6138. http://dx.doi.org/10.1007/s11356-016-6103-9. micromeso.2015.09.041. URL <http://www.sciencedirect.com/science/article/
Morales, A.D.C., Economou, I.G., Peters, C.J., Siepmann, J.I., 2013. Influence of pii/S1387181115005302>.
simulation protocols on the efficiency of gibbs ensemble Monte Carlo Yan, S., Fang, M., Zhang, W., Zhong, W., Luo, Z., Cen, K., 2008. Comparative analysis
simulations. Mol. Simul. 39 (14–15), 1135–1142. http://dx.doi.org/10.1080/ of CO2 separation from flue gas by membrane gas absorption technology and
08927022.2013.828209. chemical absorption technology in china. Energy Convers. Manage. 49 (11),
Moura, P.A.S., Bezerra, D.P., Vilarrasa-Garcia, E., Bastos-Neto, M., Azevedo, D.C.S., 3188–3197. http://dx.doi.org/10.1016/j.enconman.2008.05.027 (Special Issue
2015. Adsorption equilibria of CO2 and CH4 in cation-exchanged zeolites 13x. 3rd International Conference on Thermal Engineering: Theory and
Adsorption 22 (1), 71–80. http://dx.doi.org/10.1007/s10450-015-9738-9. Applications). URL <http://www.sciencedirect.com/science/article/pii/
Myers, A.L., Prausnitz, J.M., 1965. Thermodynamics of mixed-gas adsorption. AIChE S0196890408002240>.
J. 11 (1), 121–127. http://dx.doi.org/10.1002/aic.690110125. Yazaydin, A., Snurr, R., Park, T.H., Koh, K., Liu, J., LeVan, M., Benin, A., Jakubczak, P.,
Ntiamoah, A., Ling, J., Xiao, P., Webley, P.A., Zhai, Y., 2016. Co2 capture by Lanuza, M., Galloway, D., Low, J., Willis, R., 2009. Screening of metal-organic
temperature swing adsorption: use of hot co2-rich gas for regeneration. Ind. frameworks for carbon dioxide capture from flue gas using a combined
Eng. Chem. Res. 55 (3), 703–713. http://dx.doi.org/10.1021/acs.iecr.5b01384. experimental and modeling approach. J. Am. Chem. Soc. 13, 18198–18199.
188 B. Vujic, A.P. Lyubartsev / Chemical Engineering Science 174 (2017) 174–188

Ye, S., Jiang, X., Ruan, L.-W., Liu, B., Wang, Y.-M., Zhu, J.-F., Qiu, L.-G., 2013. Post- International Zeolite Membrane Meeting). URL <http://www.sciencedirect.
combustion CO2 capture with the HKUST-1 and MIL-101(Cr) metal-organic com/science/article/pii/S138718111300471X>.
frameworks: adsorption, separation and regeneration investigations. Micropor. Yiannourakou, M., Ungerer, P., Leblanc, B., Ferrando, N., Teuler, J.-M., 2013.
Mesopor. Mater. 179 (0), 191–197. http://dx.doi.org/10.1016/j. Overview of medeaÒ-gibbs capabilities for thermodynamic property
micromeso.2013.06.007. URL <http://www.sciencedirect.com/science/article/ calculation and VLE behaviour description of pure compounds and mixtures:
pii/S1387181113002813>. application to polar compounds generated from ligno-cellulosic biomass. Mol.
Ye, P., Sjöberg, E., Hedlund, J., 2014. Air separation at cryogenic temperature using Simul. 39 (14–15), 1165–1211. http://dx.doi.org/10.1080/
MFI membranes. Micropor. Mesopor. Mater. 192 (0), 14–17. http://dx.doi.org/ 08927022.2013.830182.
10.1016/j.micromeso.2013.09.016 (Special Issue Devoted to the 6th