m
109
Zeolites
Selectivities
com
L-10
ing
ata
xces
ed t
ing
sures are increased significantly above 2 bar. For a realistic comparison of the separation characteristics
sidera
s (MOF
ive ad
s [1–1
urificat
(CO2, CO, and CH4) are then recovered from the column by desorp-
tion at lower pressures. The CO2/CH4/CO purge gas is normally
used for combustion purposes in a steam reformer. In view of
the current concerns about CO2 emissions there is an incentive to
remove CO2 from the purge gas mixture. After selective adsorption
deed, in several recent investigations [1–3,11,15,27,28], the exper-
imental data have been first converted from excess to absolute
loadings, prior to isotherm fitting and subsequent use in IAST
calculations of Sads, and breakthrough simulations of PSA perfor-
mance. Regretfully, the excess loading data for CO2, CH4, and CO
isotherms in CuBTC, MIL-101, and Zn(bdc)dabco have been fitted
directly by Chowdhury et al. [21] and Mishra et al. [22] without
first conversion to absolute loadings. This implies that their re-
ported isotherm fit parameters are not directly usable.
⇑ Tel.: +31 20 6270990; fax: +31 20 5255604.
Microporous and Mesoporous Materials 156 (2012) 217–223
Contents lists available at
e
lse
E-mail address: r.krishna@uva.nl
Significantly fewer studies have been devoted to separation of
gas mixtures containing CO [16–23]. The adsorptive separation of
CO2 from CO is of particular interest to NASA’s MARS in-situ
resource utilization program [23]. The industrial need for separa-
tion of CO2/CH4/CO gas mixtures has been underlined by Hamon
et al. [20]. Hydrogen is mainly produced by steam reforming of
natural gas, a process which generates a synthesis gas mixture con-
taining H2, CO2, CO, and CH4. In order to obtain pure H2, pressure
swing adsorption (PSA) is used to remove these impurities from
the synthesis gas mixture. In practice, the adsorbed impurities
NaX zeolite is available in the works of Belmabkhout et al. [16] and
Cavenati et al. [24]. Isotherm fitting with fundamental models,
such as Langmuir and Toth, should be done using absolute, not
excess, loadings; this has been stressed by Myers and Monson
[25]. The poorer the adsorption strength, the larger is the differ-
ence between the absolute and excess loadings; the differences
are particularly significant for poorly adsorbing gases such as H2
and N2. Isotherm fits of absolute loadings need to be used for
calculations of the adsorption selectivity, Sads, using say the Ideal
Adsorbed Solution Theory (IAST) of Myers and Prausnitz [26]. In-
1. Introduction
In recent years there has been con
of zeolites, metal organic framework
olate frameworks (ZIFs) for select
CO2/H2, CO2/CH4 and CO2/N2 mixture
as natural gas sweetening, and H2 p
sures can range up to 60 bar [15].
1387-1811/$ - see front matter � 2012 Elsevier Inc. A
doi:10.1016/j.micromeso.2012.02.034
in a fixed bed adsorber unit, transient breakthrough calculations were performed for an equilibrium
packed bed adsorber. For a specified purity of CO2 exiting the packed bed adsorber, the best CO2 removal
performance is obtained with CuBTC. Our studies highlight the relative importance of adsorption selec-
tivities and capacities in the performance of fixed bed adsorbers, and underline the significant advantage
of MOFs over traditionally used zeolites.
� 2012 Elsevier Inc. All rights reserved.
ble research on the use
s), and zeolitic imidaz-
sorption of CO2 from
4]. In applications such
ion the operating pres-
of CO2 from the purge gas, the recovered CO and CH4 are usable as
fuel gas in the steam reformer.
For evaluation of suitable adsorbents for selective removal of
CO2 from mixtures containing CO and CH4, pure component
adsorption isotherm data are required. Experimental isotherm
data, in terms of excess loadings, are available for a variety of MOFs:
CuBTC [21], MIL-101 [21], and Zn(bdc)dabco [22] for pressures
ranging to 100 bar. The corresponding data for conventionally used
CO2 removal
Pressure swing adsorption
Metal organic frameworks
CO2/CO selectivities are significantly higher with NaX than for the three MOFs. The working capacity
for CO2 adsorption, on the other hand, is significantly higher for MOFs than for NaX zeolite as the pres-
Short Communication
Adsorptive separation of CO2/CH4/CO gas
Rajamani Krishna ⇑
Van ‘t Hoff Institute for Molecular Sciences, University of Amsterdam, Science Park 904,
a r t i c l e i n f o
Article history:
Received 30 November 2011
Received in revised form 13 February 2012
Accepted 18 February 2012
Available online 3 March 2012
Keywords:
a b s t r a c t
The major objective of this
works (MOFs): CuBTC, MI
CO2 from mixtures contain
sures ranging to 60 bar. D
are available in terms of e
type models, these data ne
Solution Theory (IAST), us
Microporous and M
journal homepage: www.e
ll rights reserved.
ixtures at high pressures
8 XH Amsterdam, The Netherlands
munication is to compare the performance of three metal–organic frame-
1, and Zn(bdc)dabco, with that of NaX zeolite for selective adsorption of
CH4 and CO in a pressure swing adsorption (PSA) unit operating at pres-
on the pure component adsorption isotherms in the published literature
s loadings. For purposes of isotherm fitting with fundamental Langmuir-
o be converted to absolute loadings. Calculations using the Ideal Adsorbed
the fitted isotherm data on absolute loadings, show that the CO2/CH4 and
SciVerse ScienceDirect
soporous Materials
vier .com/locate /micromeso
poro
The present communication has three main objectives. The first
objective is to demonstrate the need for converting experimental
data from excess to absolute loadings before fitting with Lang-
muir-type models. We aim to show that this conversion is essential
for calculations of Sads, and isosteric heats of adsorption, Qst; this
aspect appears to be insufficiently appreciated in some of the
published literature analyzing high pressure adsorption data
[13,21,22], while others have been careful in converting excess
loadings to absolute loadings for the purposes of calculating Qst
and Sads [2,3,11,15,28–30]. The second objective is to underscore
the fallacy in selecting materials purely based on selectivity con-
siderations; this need arise in view of the growing research in
MOFs for which selectivity considerations are often used for
screening purposes [31]. The third objective is to emphasize the
importance of capacity considerations especially for high pressure
separations. To achieve these objectives we examine the separa-
tion performance of CuBTC, MIL-101, Zn(bdc)dabco, and NaX for
CO2 removal from CO2/CH4/CO mixtures by performing break-
through calculations in a packed bed adsorber, representative of
the adsorption phase of a pressure swing adsorption (PSA) unit.
Details and methodology of isotherm fitting, tabulated isotherm
fit parameters, along with calculations of Sads, Qst, and adsorber
breakthrough for the chosen microporous materials have been pro-
vided in the Supplementary material accompanying this publica-
tion. The discussions below serve to highlight the major issues
that emerge from our studies.
2. Absolute and excess loadings, calculations of Qst and Sads
The differences between absolute and excess loading depends
on a variety of factors that include pore volume, and bulk gas pres-
sure. As illustration, Fig. 1 compares absolute and excess loadings
Nomenclature
L length of packed bed adsorber, m
pi partial pressure of species i in mixture, Pa
pt total system pressure, Pa
qi component molar loading of species i, mol kg�1
Qst isosteric heat of adsorption, J mol�1
R gas constant, 8.314 J mol�1 K�1
Sads adsorption selectivity, dimensionless
t time, s
T absolute temperature, K
u superficial gas velocity in packed bed adsorber, m s�1
z distance along the adsorber, m
218 R. Krishna /Microporous and Meso
for adsorption of CO in CuBTC, MIL-10, and Zn(bdc)dabco that have
pore volumes, respectively, of 0.75, 1.38, and 0.67 cm3 g�1. For
MIL-101, that has the highest pore volume the absolute loadings
are about 50% larger than the excess loadings of CO for pressures
higher than 60 bar; a similar conclusion holds for adsorption of
CH4. For CO2, that has a higher adsorption strength, the differences
between absolute and excess loadings are expected to be lower; see
Supplementary material for quantitative information. The experi-
mental data of Moellmer et al. [30] shows that for adsorption of
weakly adsorbing H2 in CuBTC, the absolute loadings can be more
than twice the excess loadings for pressures approaching 50 MPa.
One consequence of the foregoing is that estimations of the
adsorption selectivity, defined by
Sads ¼ q1=q2p1=p2
ð1Þ
will be overly optimistic for CO2/CO and CO2/CH4 mixtures if these
are based on isotherm fits of excess loadings. To illustrate this, we
compared the estimates of Sads for equimolar binary CO2/CO mix-
ture (p1 = p2) for Zn(bdc)dabco, using our fits of absolute loadings,
with those reported by Mishra et al. [22], which were based on Viri-
al fits of excess loadings; see Fig. 2. At a bulk gas pressure of 10 bar,
our estimates of Sads is 9, as compared to their estimates which are
about 60% higher.
The isosteric heat of adsorption, Qst, are usually estimated from
the isotherm fits using the Clausius–Clapeyron equation
Qst ¼ �RT2
@ ln p
@T
� �
q
ð2Þ
Strictly speaking, the derivative in Eq. (2) is to be determined
with the absolute loading, q, held constant; see Myers and Monson
[25]. Moellmer et al. [30] have been meticulous in their calcula-
tions of the isosteric heats of adsorption of CO2, CH4, N2 and H2
in CuBTC and the absolute loadings are held constant in applying
Eq. (2). However, in the work of Chowdhury et al. [21] and Mishra
et al. [22], the isosteric heats are calculated using fits of excess
loadings. By implication, their values are obtained with derivatives
in which the excess loadings are held constant. Their calculated val-
ues of the isosteric heats of adsorption are expected to be different
to those obtained in this work. This expectation is fulfilled as evi-
denced in Fig. 3 that compares the calculations of Qst obtained in
this work, with those reported by Chowdhury et al. [21] and
Mishra et al. [22]. There are significant differences in the two sets
of values for all three MOFs. For CuBTC and Zn(bdc)dabco the dif-
ferences are apparent only at higher loadings. For MIL-101, there
are significant differences over the entire range of loadings. The
reason for this can be traced to the significantly higher pore vol-
ume for MIL-101 which results in a significantly large difference
between excess and absolute loadings over a wide range of
Greek letters
e voidage of packed bed, dimensionless
q framework density, kg m�3
s time, dimensionless
sbreak breakthrough time, dimensionless
Subscripts
i referring to component i
t referring to total mixture
us Materials 156 (2012) 217–223
operating pressures.
3. Evaluation of adsorbents for CO2/CH4/CO separation
We now address the question: which is the best adsorbent for
selective removal of CO2 from the ternary CO2/CH4/CO mixture?
The most commonly used procedure for comparing adsorbents is
on the basis of the adsorption selectivity. For a arbitrarily chosen
temperature of 300 K, Fig. 4a and b shows the IAST calculations
of CO2/CH4 and CO2/CO selectivities for a range of operating pres-
sures using four different materials. In both cases, NaX zeolite is
found to have the highest selectivities over the entire range of
pressures. The conventional wisdom would be to assume that
NaX is the best adsorbent. However, the performance of a PSA unit
is dictated not only by Sads but also by the working capacity, as has
been emphasized in several publications [1,2,15,27,32–34]. Let us
define the working capacity for selective adsorption of CO2 from
CO2/CH4/CO gas mixtures as the component loading of CO2 in the
poro
R. Krishna /Microporous and Meso
mixture at the operating pressure minus the value at a desorption
pressure of 0.1 bar. Fig. 5 compares the working capacities for
CuBTC, MIL-101, Zn(bdc)dabco, and NaX, expressed as the number
Bulk gas phase pressure, pi /bar
0 5 10 15 20 25 30
Lo
ad
in
g,
q
i /
m
ol
k
g-
1
0
1
2
3
4
5
6
1-site Langmuir
Virial fit
294 K: absolute
294 K: excess
Zn(bdc)dabco; 294 K;
Mishra excess data;
pure CO
Bulk gas phase pressure, pi /bar
0 10 20 30 40 50 60 70
Lo
ad
in
g,
q
i /
m
ol
k
g-
1
0
2
4
6
8
10
12 2-site Langmuir: this work
2-site Langmuir: Chowdhury
295 K: absolute
295 K: excess
MIL-101; 295 K;
Comparison of fits;
pure CO
Bulk gas phase pressure, pi /bar
0 10 20 30 40 50 60 70 80
Lo
ad
in
g,
q
i /
m
ol
k
g-
1
0
2
4
6
8
10
12
2-site Langmuir: this work
Virial-Langmuir: Chowdhury
295 K: absolute
295 K: excess
CuBTC; 295 K;
Comparison of fits;
pure CO
(a)
(b)
(c)
Fig. 1. Pure component isotherm data for CO in (a) CuBTC, (b) MIL-101, (c)
Zn(bdc)dabco. The absolute and excess loadings are plotted in these diagrams.
The continuous solid lines represent the fits of the absolute loadings from this work.
The dashed lines are the fits of Chowdhury et al. [21] and Mishra et al. [22] for the
excess loadings.
of moles of CO2 in the adsorbed mixture per L of adsorbent. With
increasing pressures, the working capacity of NaX tends to reach
plateau values, because saturation conditions are reached. This is
because of the relatively low pore volume of NaX (0.28 cm3 g�1)
compared to the much higher pore volumes of the three MOFs.
The information presented in Figs. 4 and 5 underline the dilem-
ma faced in adsorbent selection. NaX has the best selectivities, but
is limited by capacity considerations for operation at pressures
exceeding 1 bar. In order to resolve the adsorbent selection dilem-
ma we need to consider the breakthrough characteristics in a
packed bed adsorber. The breakthrough calculations were per-
formed using the methodology described in our earlier papers
[27,35]. Assuming plug flow of CO2/CH4/CO gas mixture through
a fixed bed maintained under isothermal conditions, the partial
pressures in the gas phase at any position and instant of time are
obtained by solving the following set of partial differential equa-
tions for each of the species i in the gas mixture.
Total gas pressure / bar
0 2 4 6 8 10
C
O
2/C
O
a
ds
or
pt
io
n
se
le
ct
iv
ity
, S
ad
s
0
2
4
6
8
10
12
14
IAST with fits of absolute loadings
IAST with fits of excess loadings
Zn(bdc)dabco;
binary CO2/CO mixtures
294 K;
Fig. 2. Calculations of the CO2/CO adsorption selectivities, Sads, for equimolar
CO2/CO gas mixtures at 294 K using Zn(bdc)dabco. The calculations are based on the
Ideal Adsorbed Solution Theory (IAST) using the pure component isotherm fits of
either absolute loadings (this work) or excess loadings (obtained from the work of
Mishra et al. [22]).
us Materials 156 (2012) 217–223 219
1
RT
e
@piðt; zÞ
@t
¼ � 1
RT
@ðuðt; zÞpiðt; zÞÞ
@z
� ð1� eÞq @qiðt; zÞ
@t
; i ¼ 1;2;3
ð3Þ
In Eq. (3), t is the time, z is the distance along the adsorber, q is
the framework density, e is the bed voidage, and u is the superficial
gas velocity. The adsorber bed is initially free of adsorbates, i.e. we
have the initial condition
t ¼ 0; qið0; zÞ ¼ 0 ð4Þ
At time, t = 0, the inlet to the adsorber, z = 0, is subjected to a
step input of the ternary gas mixture and this step input is main-
tained till the end of the adsorption cycle when steady-state con-
ditions are reached.
t � 0; pið0; tÞ ¼ pi0; uð0; tÞ ¼ u0 ð5Þ
where u0 is the superficial gas velocity at the inlet to the adsorber.
Invoking the constraint of negligible pressure drop, the overall
material balance is obtained by summing Eq. (3) over the three
component species
1
RT
pt
@ðuðt; zÞÞ
@z
¼ �ð1� eÞq
X3
i¼1
@qiðt; zÞ
@t
ð6Þ
Eq. (6) allows the calculation of the superficial gas velocity
along the length of the adsorber.
(a)
porous Materials 156 (2012) 217–223
l-1 30(a)
220 R. Krishna /Microporous and Meso
The molar loadings of the species i, qi(z,t) at any position z, and
time t are determined from IAST calculations. Eq. (3) is first sub-
jected to finite volume discretization. Typically, the adsorber
length L is divided into 100 slices. The number of slices is deter-
Loading, qi /mol kg
-1
1 10
Is
os
te
ric
h
ea
t o
f a
ds
or
pt
io
n,
-
Q
st
/
kJ
m
ol
-1
0
10
20
30
40
50
CO2: This work
CO: This work
CH4: This work
CO2: Chowdhury
CO: Chowdhury
CH4: Chowdhury
MIL-101; Comparison with Chowdhury
200.2
Loading, qi /mol kg
-1
0.1 1 10
Is
os
te
ric
h
ea
t o
f a
ds
or
pt
io
n,
-
Q
st
/
kJ
m
ol
-1
0
5
10
15
20
25
CO2: This work
CO: This work
CH4: This work
CO2: Mishra
CO: Mishra
CH4: Mishra
Zn(bdc)dabco;
Comparison with Mishra data
Loading, qi /mol kg
-1
1 10
Is
os
te
ric
h
ea
t o
f a
ds
or
pt
io
n,
-
Q
st
/
kJ
m
o
0
5
10
15
20
25
CO2: This work
CO: This work
CH4: This work
CO2: Chowdhury data
CH4: Chowdhury data
CO: Chowdhury data
200.2
CuBTC; Comparison with Chowdhury
(b)
(c)
Fig. 3. The isosteric heat of adsorption for CO2, CH4, and CO in (a) CuBTC, (b) MIL-
101, and (c) Zn(bdc)dabco at 300 K. Our calculations are compared to those
presented in the works of Chowdhury et al. [21] and Mishra et al. [22].
(b)
Fig. 4. Calculations of the (a) CO2/CH4, and (b) CO2/CO adsorption selectivities, Sads,
for equimolar CO2/CH4/CO ternary gas mixtures at 300 K using CuBTC, MIL-101,
Zn(bdc)dabco, and NaX zeolite. The calculations are based on the Ideal Adsorbed
Solution Theory (IAST) using the pure component isotherm fits.
Total gas pressure / bar
0 10 20 30 40 50 60 70
W
or
ki
ng
c
ap
ac
ity
fo
r C
O
2 /
m
ol
L
-1
0
2
4
6
8
10
12
14 CuBTCNaX
MIL-101
Zn(bdc)dabco
CO2/CH4/CO mixtures
300 K;
desorption pressure =0.1 bar
Fig. 5. Calculations of working capacities per L of adsorbent for selective adsorption
of CO2 from equimolar CO2/CH4/CO ternary gas mixtures at 300 K using CuBTC, MIL-
101, Zn(bdc)dabco, and NaX zeolite. The calculations are based on the Ideal
Adsorbed Solution Theory (IAST) using the pure component isotherm fits with
parameters specified in the Supplementary material. The desorption pressure is
assumed to be 0.1 bar.
mined by checking that the obtained breakthrough results do not
change on increasing it. Combination of the discretized PDEs along
with the algebraic IAST equilibriummodel, results in a set of differ-
ential-algebraic equations (DAEs), which are solved using BESIRK
[36]. BESIRK is a sparse matrix solver, based on the semi-implicit
Runge–Kutta method originally developed by Michelsen [37], and
extended with the Bulirsch–Stoer extrapolation method [38]. Use
of BESIRK improves the numerical solution efficiency in solving
the set of DAEs. The evaluation of the sparse Jacobian required in
the numerical algorithm is largely based on analytic expressions
[39]. Further details of the adsorber model, along with the numer-
ical procedures used in this work, are provided in our earlier works
[39–41]. Typical computation times for a binary gas mixture break-
through are less than 100 s, allowing such transient adsorber cal-
culations to be routinely used for screening purposes.
The solution to Eq. (3) yields piðt; zÞ, from which we can calcu-
late the mol% CO2 in the outlet gas mixture as a function of time.
The comparison of the breakthrough characteristics of CuBTC,
MIL-101, Zn(bdc)dabco, and NaX for p10 = p20 = p30 = 10 bar are
provided in Fig. 6b.
For a specified purity of CO2 in the outlet gas, arbitrarily chosen
as 0.05 mol% CO2, the dimensionless breakthrough times, sbreak,
can be determined for a range of inlet total gas pressures. Fig. 6c
compares the dimensionless breakthrough times, sbreak, for differ-
ent materials. Higher values of sbreak are desirable because longer
times will be available for uptake of CO2 before the bed needs to
be regenerated. The dimensionless breakthro