Chemical Engineering Science 66 (2011) 303–317

Contents lists available at ScienceDirect

Chemical Engineering Science

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

Activated carbon for hydrogen purification by pressure swing adsorption:

Multicomponent breakthrough curves and PSA performance
Filipe V.S. Lopes, Carlos A. Grande, Alı́rio E. Rodrigues n
Laboratory of Separation and Reaction Engineering (LSRE), Associate Laboratory LSRE-LCM, Department of Chemical Engineering, Faculty of Engineering,
University of Porto (FEUP), Rua Dr. Roberto Frias, s/n, 4200-465 Porto, Portugal

a r t i c l e in f o abstract

Article history: In this work, multicomponent breakthrough experiments (binary H2-CO2, ternary H2-CO2-CO and five-
Received 20 May 2010 component H2-CO2-CO-CH4-N2) were performed under different operating conditions in activated carbon
Received in revised form extrudates to validate the mathematical model. A 10 steps one-column VPSA experiment was also
19 October 2010
performed. These experiments allow experimental validation of adsorption equilibrium predicted by the
Accepted 26 October 2010
multicomponent extension of the Virial isotherm and a fixed-bed mathematical model. In the VPSA
Available online 31 October 2010
experiment, a 99.981% hydrogen purity stream (with 63 ppm of CO contamination) was obtained with a
Keywords: 1 1
hydrogen recovery of 81.6% and an adsorbent productivity of 101 molH2 kgads day .
Pressure swing adsorption The mathematical model was also employed to assess the effect of operating conditions and the
Adsorption
influence of step times and pressure equalizations in the PSA unit. It was verified that high-purity
Separations
hydrogen ( 499.99%) can be obtained using this adsorbent with recoveries higher than 75% and unit
Steam methane reforming off-gases 1 1
Mathematical modeling productivities of 160 molH2 kgads day .
Simulation & 2010 Elsevier Ltd. All rights reserved.

1. Introduction
Low cost with high efficiency generation of electricity; suitable
for megawatt-scale distribution in power applications and with
almost no CO2 emissions (and emissions of other greenhouse gases)
these are some characteristics of power systems in the future
(Khelifa et al., 1999). Recently, the increased demand for the supply
of pure hydrogen gained strength, especially in the fuel cell
industry. The main reason for this is that hydrogen fuel cells are
extremely efficient energy conversion devices (Kärger and
Ruthven, 1992). In its natural and elemental form hydrogen is
scarce, making crucial its production. Steam methane reforming
(SMR) remains one of the most economical routes to produce
hydrogen (Ruthven et al., 1994). However, the SMR reactor
produces hydrogen with several impurities such as water vapor,
carbon dioxide, methane, carbon monoxide and, in some cases,
nitrogen (Sircar et al., 1999). For fuel cell applications, hydrogen
with a purity of over 99.99% is required and hence these impurities
must be removed. In particular, carbon monoxide must be reduced
to less than 10 ppm for the purified hydrogen to be used in some
types of fuel cells (DOE, 2000; OECD/IEA, 2004). Pressure swing
adsorption (PSA) units are commonly used in hydrogen purification
from SMR off-gases (Wagner, 1969; Ruthven et al., 1994; Yang
et al., 1995; Malek and Farooq, 1998; Yang and Lee, 1998; Ahn et al.,
n
Corresponding author. Tel.: + 351 22 508 1671; fax: + 351 22 508 1674.
E-mail address: arodrig@fe.up.pt (A.E. Rodrigues).
1999; Sircar et al., 1999; Park et al., 2000; Sircar and Golden, 2000;
Esteves and Mota, 2002; Zhou et al., 2002; Bec, 2005; Lee et al.,
2008). This process involves a large number of design and operation
parameters. As a consequence, for study of H2 purification through
the PSA process it is important to determine the adsorption
equilibria and kinetics data of single and multicomponent systems
in order to develop the completed design of the separation unit
(Ritter and Yang, 1987).
The purpose of this work is to provide adsorption data of
multicomponent mixtures of SMR off-gases. Based on single-
component data already established from several commercial
adsorbents (Grande et al., 2008; Lopes et al., 2009a) it is possible
to estimate the behaviour of the mixture through the extension of
results obtained for pure gases. However, it is convenient to
validate these data, especially in the case of a mixture with five
components. Therefore, breakthrough curves were tested, consid-
ering different feed compositions and feed flow rates in order to
determine some multicomponent adsorption data. As mentioned
before, the multicomponent mixtures of SMR off-gases comprise
several gases and that they require a detailed study. Since there
are available commercial adsorbents with enhanced adsorptive
capacities for different gases (but not for all contaminants
simultaneously), it is crucial to study different adsorbents and
also evaluate their adsorption properties (Grande et al., 2008; Lopes
et al., 2009a, 2009b, 2010). To measure the adsorption data of
multicomponent mixtures, an experimental setup composed of a
one-column PSA unit and a gas chromatograph analysis system for
determination of bed outlet concentrations of multicomponent gas

0009-2509/$ - see front matter & 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ces.2010.10.034
304 F.V.S. Lopes et al. / Chemical Engineering Science 66 (2011) 303–317

mixtures was built, calibrated and tested for SMR off-gases
(H2, CO2, CH4, CO and N2). Once the analysis system began
operating in the upgraded facility, adsorption equilibrium and
kinetic behaviour of binary H2-CO2, ternary H2-CO2-CO and five-
component H2-CO2-CH4-CO-N2 mixtures were measured. Further
modeling of the multicomponent adsorption data was performed
testing the mathematical model to obtain suitable predictions.
Besides, a 10 steps one-column vacuum pressure swing adsorption
(VPSA) experiment was performed for the complete validation of
the model. Finally, a cycle sequence of 6 steps along with another
cycle of 10 steps was considered for the simulation of two one-
column VPSA and PSA processes in order to achieve high H2
purification process performances.
of H2-CO2-CH4-CO-N2. The bed outlet concentrations of multi-
component gas mixtures of hydrogen, carbon dioxide, carbon
monoxide, methane and nitrogen were determined for each loop
by use of the TCD detector and only after storage of the complete
storage of the 15 loops. This apparatus was used for the study
of hydrogen purification from SMR off-gases (H2, CO2, CH4, CO
and N2).
Breakthrough experiments on binary (H2-CO2), ternary
(H2-CO2-CO) and five-component (H2-CO2-CH4-CO-N2) mixtures
were performed in a fixed-bed filled with a modified activated
carbon AC5-KS (previously studied by Lopes et al., 2009b). The
production of this material was performed by additional physical
activation with CO2 of a commercial activated carbon sample

Table 1
2. Experimental Characteristics of the column used in the multicomponent breakthrough curves and
VPSA experiments and adsorbent physical properties of the material employed.

Breakthrough curves and VPSA experiments were performed in Fixed-bed characteristics

an one-column PSA experimental setup able to operate with gas Mass of adsorbent (kg) 0.2456
mixtures of three (or more) components and with feed flow rates Bed height (m) 0.267
between 1.67  10  6 and 8.33  10  5 m3 s  1 (measured at 298 K). Bed volume (m3) 5.24  10  4
Bed porosity, e 0.404
This unit can work in the pressure range of 0.1-5 bar and at
Thermocouple distance from feed inlet (m)
temperatures between 298 and 423 K. The schematic view of the TTBottom 0.059
complete unit is shown in Fig. 1. The gas chromatograph analysis TTMiddle 0.151
system installed in this setup is composed of a gas chromatograph TTTop 0.239
(Varian CP-3800, Netherlands) equipped with a multi-port valve
(Vici, Valco) able to store up to 15 samples for further analysis. A Adsorbent physical properties
packed column (Supelco 60/80 Carboxen-1000 12390-U, USA) was Pellet porosity, ep 0.627
placed in the oven of the gas chromatograph, operating at 453 K Adsorbent density, rp (kg m  3) 786
Solid density, rs (kg m  3) 2110
and using an inert carrier (He) flow rate of 8.33  10  7 m3 s  1 Average pellet radius, Rp (m) 1.45  10  3
(measured at 298 K) to allow the complete component separation

Fig. 1. Setup for one-column PSA experiments and for multicomponent fixed-bed breakthroughs connected to a GC analysis system to measure the bed outlet concentration.
 F.V.S. Lopes et al. / Chemical Engineering Science 66 (2011) 303–317 305

(Norit R2030, Netherlands) in order to enhance its capacity toward experiment was performed at 303 K (an experimental temperature
carbon dioxide (Lopes et al., 2009b). The physical activation near the room temperature) that is closer to the temperature of steam
promoted the formation of additional micropores in the carbon reformer off-gases (after passing through various heat exchangers).
structure to selectively adsorb more CO2. Temperature variations A cycle sequence of 10 elementary steps was employed (cycle B
along the bed are recorded by three thermocouples (TT) located at reported in Fig. 2): a co-current feed, three pressure equalizations–
0.059 (TTBottom), 0.151 (TTMiddle) and 0.239 m (TTTop) from the feed depressurizations, a counter-current blowdown, a counter-current
inlet. To degas the adsorbent sample before the experiments, a inert purge with hydrogen, three pressure equalizations–pressurizations
gas (He) flow of 5.00  10  6 m3 s  1 (measured at 298 K) was used and a counter-current pressurization with hydrogen. Cycle B is a
overnight at 423 K, after heating the sample at heating rate of Skarstrom cycle (Skarstrom, 1960) with the integration of three
1 K min  1. The breakthrough experiments were carried out at pressure equalization steps used in PSA columns to increase H2
323 K and the total pressure of 5 bar was kept constant with a back- recovery. The cycle started with the co-current feed step with the
pressure regulator. The dimensions of the column used in these consideration that the column was filled with an inert gas (helium) at
measurements and the adsorbent physical properties are given in the feed temperature and feed pressure. Both inlet streams of the
Table 1. The experimental conditions of the fixed-bed break- pressurization and purge steps were composed of a pure less adsorbed
through curves are reported in Table 2. gas (in this case, hydrogen). Pressure equalizations were achieved by
The same column employed in the fixed-bed breakthrough curves successive depressurization of the column, in different steps in
was used in a VPSA experiment (see properties in Table 1). This manner similar to that of three different equalization steps. To
achieve a pressure increase in the column in the equalization steps,
pure hydrogen was employed. The operating parameters of the VPSA
Table 2
Experimental conditions of the fixed-bed and VPSA experiments. experiment are also detailed in Table 2.
The gases used in this study were provided by Air Liquide:hy-
Fixed-bed experiments drogen N60, carbon dioxide N24, carbon monoxide N47, methane
Run 1 2 3 4 5 N35, nitrogen N45 and helium N50 (purities greater than 99.9999%,
Feed molar fraction, yi (dimensionless)
99.4%, 99.997%, 99.95%, 99.995% and 99.999%, respectively).
H2 0.70 0.90 0.20 0.70 0.73
CO2 0.30 0.10 0.80 0.26 0.23
CO – – – 0.04 0.012 3. Theoretical
CH4 – – – – 0.021
N2 – – – – 0.007
PFeed ¼5 bar; T¼323 K; QFeed ¼5  10  5 m3 s  1a Adsorption equilibrium and kinetics of pure hydrogen, carbon
dioxide, carbon monoxide, methane and nitrogen on modified
VPSA experiment
Feed molar fraction, yi (dimensionless) H2: 0.79; CO2: 0.17; CH4: 0.021;
activated carbon AC5-KS were previously reported by Lopes et al.
CO: 0.012; N2: 0.007 (2009b). The Virial adsorption model was used to describe adsorp-
Step times (s) tFeed ¼ 101 s; tBlow ¼ 201 s; tPurge ¼ 40 s; tion equilibria of these gases at 303 and 323 K and in the pressure
tPress ¼ 21 s; range 0-7 bar (Lopes et al., 2009b).
tED1 ¼tED2 ¼tED3 ¼ tEP3 ¼tEP2 ¼tEP1 ¼ 16 s
The predictions for multicomponent fixed-bed experiments of
Phigh ¼ 5 bar; Pinter1 ¼3.875 bar; Pinter2 ¼ 2.75 bar; Pinter3 ¼ 1.625 bar;
Plow ¼0.5 bar; T¼303 K; QFeed ¼5  10  5 m3 s  1a H2-CO2, H2-CO2-CO and H2-CO2-CH4-CO-N2 were made applying
the extended Virial isotherm model (Kiselev, 1971; Barrer, 1981;
a
Measured at 298 K. Taqvi and LeVan, 1997) using the parameters previously obtained

Fig. 2. Schematic diagram of the cycle sequences used in the PSA simulations for H2 purification: Cycle A: (1) co-current feed step; (2) pressure equalization–depressurization
step; (3) counter-current blowdown step; (4) counter-current purge step with hydrogen; (5) pressure equalization–pressurization step; (6) counter-current pressurization
step with hydrogen; Cycle B: (1) co-current feed step; (2) first, (3) second and (4) third pressure equalization–depressurization steps; (5) counter-current blowdown step;
(6) counter-current purge step with hydrogen; (7) third, (8) second and (9) first pressure equalization–pressurization steps; and (10) counter-current pressurization step with
hydrogen.

Table 3
Virial isotherm fitting parameters of pure H2, CO2, CO, CH4 and N2 at 303–323 K.

Adsorbate KN (–DH) (kJ mol  1) A0  10  5 A1  10  6 B0  10  11 B1  10  11

(mol kg  1 bar  1) (m2 mol  1) (m2 K mol  1) (m4 mol  2) (m4 K mol  2)

H2 2.12  10  3 7.26  1.917 0.1994  4.015 7315.4

CO2 8.06  10  5 28.02 0.081 45.31  0.365 92.247
CO 7.63  10  5 23.23  0.149  105.46  2.407 528.18
CH4 2.29  10  4 23.18 12.808  294.44  6.101 1745.7
N2 3.06  10  4 18.27 11.989  278.79  13.758 3988.6
306 F.V.S. Lopes et al. / Chemical Engineering Science 66 (2011) 303–317

from pure component data (see Table 3) (Lopes et al., 2009b). The
dynamic behaviour of multicomponent adsorption was described
taking into consideration:
(a) ideal gas behaviour throughout the fixed-bed,
(b) constant porosity along the bed,
(c) bi-disperse adsorbent where macro and micropore mass trans-
fer is expressed by a bi-linear driving force (bi-LDF) model,
(d) axial dispersed plug flow mathematical model composed of
mass, momentum and energy balances,
(e) no temperature gradients inside each particle. Radial gradients
of temperature and concentration were also neglected,

Table 4
Mathematical model used for fixed-bed, VPSA and PSA simulations.

Virial adsorption isotherm model

 
The Virial isotherm and Virial coefficients q 2 3 P
1 A
m P1 B
m
P¼ exp Aq þ 2 Bq2 þ    ; A ¼ m
, B¼ m
KH S 2S m¼0T m¼0T
 
Van’t Hoff equation DH
KH ¼ K1 exp
Rg T
!
The Virial extended isotherm and mixing Virial qi 2 P N 3 P N PN ðAi þ Aj Þ ðBi þ Bj þ Bk Þ
coefficients Pi ¼ exp Aij qj þ 2 Bijk qj qk ; Aij ¼ , Bijk ¼
KH i S j¼1 2S j ¼ 1 k ¼ 1 2 3

System of partial differential equations

 
Material balance for each component @ @yi @ @ Cg,i
e Dax Cg,T  ðu0 Cg,i Þe ð1eÞap kf ðCg,i Cs,i Þ ¼ 0
@z @z @z @t
Ergun equation @P 150 mð1eÞ 2 1:75ð1eÞrg
 ¼ u0 þ 9u0 9u0
@z e3 dp2 e3 dp
 
Gas phase energy balance @ @Tg @Tg @Cg,T 4 hw @Tg
l u0 Cg,T Cp þ eRg Tg ð1eÞap hf ðTg Tp Þ ðTg Tw Þe Cg,T CV ¼0
@z @z @z @t dwi @t
LDF equation for the macropores @Cm,i Om Dp,i rp @qi
¼ ðCs,i Cm,i Þ
@t R2p ep @t
LDF equation for the micropores @qi Oc Dc,i 
¼ ðqi qi Þ
@t r 2
" c #
Solid phase energy balance P
n P
n @Tp @Cm,T P
n @q
ð1eÞ ep Cm,i Cv,i þ rp qi Cv,ads,i þ rp C^ p,s ¼ ð1eÞep Rg Tp þ rb ðDHÞi i þ ð1eÞap hf ðTg Tp Þ
i¼1 i¼1 @t @t i¼1 @t
Solid surface concentration of each component Bii Cg, i þ Cm, i
Cs ,i ¼
1þ Bii
Biot number of each component ap kf R2p
Bii ¼
ep Om Dp,i
Column wall energy balance @T
rw C^ p,w w ¼ aw hw ðTg Tw Þaw‘ UðTw T1 Þ
@t
with aw ¼ dwi/[e(dwi + e)] and aw‘ ¼ 1=½ðdwi þ eÞlnððdwi þ eÞ=dwi Þ

Table 5
Correlations and dimensionless parameters used for estimation of mass and heat parameters.

Correlations

Axial mass dispersion and mass transfer coefficients eDax

¼ 20 þ 0:5 Sc Re
Dm
 
Micropore diffusivity Dc D0 Ea
¼ 2c exp 
rc2 rc Rg T
Heat dispersion and convective coefficients l
¼ 7 þ 0:5Pr Re
kg
Correlation based on the Chilton–Colburn analogy Nu ¼ 2:0 þ 1:1Re0:6 Pri
1=3

Global heat transfer coefficient 1 1 e dwi dwi dwe dwi

¼ þ þ with d‘n ¼
U hw lw d‘n dwe hex lnðdwe =dwi Þ
Internal convective heat transfer coefficient hw dwi d2
¼ 140 þ 0:013396 wi Re
kg dp kg
External convective heat transfer coefficient hex L 0:67Ra1=4 Tw þ T1
¼ 0:68 þ at Tfilm ¼
kg,ex ½1 þ ð0:492=PrÞ9=16 4=9 2

Dimensionless parameters
Schmidt number m
Sc ¼
rg Dm
Prandtl number C^ p m
Pr ¼
kg
Nusselt number hf d p
Nu ¼
kg
Grashof number g bðTw T1 ÞL3
Gr ¼
na
Reynolds number rg u0 dp
Re ¼
m
Sherwood number kf dp
Sh ¼
Dm
Rayleigh number Ra ¼ GrPr
 F.V.S. Lopes et al. / Chemical Engineering Science 66 (2011) 303–317 307

Table 6
Transport parameters used in the fixed-bed, VPSA and PSA simulations.

Fixed-bed simulations VPSA and PSA simulations

Run 1 2 3 4 5 1–4 5–14 15–17

Dax  105 (m2 s  1) 1.46 1.46 1.46 1.37 1.33 1.26 0.628 0.679
l (J s-1 m  1 K  1) 0.766 1.109 0.296 0.778 0.823 0.881 0.881 0.921
2 1
kf  10 (m s ) 1.83 1.69 2.04 1.53 1.40 1.31 0.65 0.72
hf (J s  1 m  2 K  1) 123 161 57 125 130 136 136 141
hw (J s  1 m  2 K  1) 55 55 55 55 55 55 55 55
U (J s  1 m  2 K  1) 50 50 50 50 50 50 50 50

Dp  106 (m2 s  1)a

H2 7.36 7.36 7.36 7.54 7.51 6.77 3.38 3.78
CO2 7.36 7.36 7.36 6.29 6.40 5.78 2.89 3.23
CO – – – 4.26 4.43 4.51 2.26 2.53
CH4 – – – – 4.56 4.58 2.29 2.57
N2 – – – – 4.38 4.45 2.23 2.49

D0c  1 H2 46.35 ; 8.51

(s ); Ea (kJ mol  1)
rc2
CO2 35.64 ; 17.6
CO 4.955 ; 10.8
CH4 35.52 ; 26.4
N2 349.7 ; 19.8

a
Values at inlet conditions.

Table 7 Table 8
Equations for the VPSA and PSA processes performance: hydrogen purity and VPSA and PSA simulations.
recovery and adsorbent productivity.
Run tFeed tBlow tPurge H2 pur. CO cont. H2 rec. Ads. prod.
R tFeed 1
CH u0 9z ¼ L dt (s) (s) (s) (%) (ppm) (%) (molH2 kgads
H2 purityð%Þ ¼ Pn t ¼ 0R tFeed2 day  1)
i ¼ 0 t ¼ 0 Ci u0 9z ¼ L dt

R tFeed R tPress R tPurge Cycle A

t¼0 CH2 u0 9z ¼ L dt CH2 u0 9z ¼ L dt CH2 u0 9z ¼ L dt 1a,e 100 200 30 99.992 15.2 70.96 91.6
H2 recoveryð%Þ ¼ RttFeed
¼0 t¼0

t¼0 CH2 u0 9z ¼ 0 dt 2a,e 100 200 45 99.999 0.7 68.68 85.6

5b,e 100 200 30 100.00 0.0 47.10 60.5
 
molH2 Abed R tFeed 6b,e 120 200 30 100.00 0.0 57.17 84.7
Ads productivity ¼ CH u0 9z ¼ L dt
kgads day tcycle wads t ¼ 0 2 tED ¼tEP ¼ 45 s; tPress ¼ 20 s
Z tPress Z tPurge 
 CH2 u0 9z ¼ L dt CH2 u0 9z ¼ L dt
t¼0 t¼0 Cycle B
3c,e 100 200 30 99.956 160 82.40 106.3
4c,e 100 200 45 99.993 18.0 79.98 99.4
7d,e 100 200 30 100.00 0.0 67.50 86.4
8d,e 120 200 30 99.998 5.0 74.17 109.4
d,e
9 120 150 30 99.997 6.4 74.34 122.7
(f) no mass, heat or velocity gradients in the radial direction, and 10d,e 120 150 20 99.991 26.0 75.77 128.8
(g) external mass and heat transfer resistances that were 11 d,e
110 150 20 99.999 3.0 72.55 115.6
expressed with the film model. 12d,e 110 100 20 99.998 4.5 72.77 133.1
13d,e 110 70 20 99.998 4.9 72.88 146.0
14d,e 120 70 20 99.987 37.4 76.14 161.6
Therefore, mass, momentum and energy balance equations can
15d,f 100 200 30 100.00 0.0 70.26 90.3
be written as shown in Table 4 (Table A1 summarizes the fixed-bed 16 d,f
120 70 20 99.972 112 78.48 165.9
initial and boundary conditions used to solve this system of partial 17 d,f
120 70 35 99.993 27.1 76.34 154.7
differential equations (Da Silva et al., 1999; Da Silva and Rodrigues, tED1 ¼tED2 ¼ tED3 ¼ tEP3 ¼tEP2 ¼ tEP1 ¼ 15 s ; tPress ¼ 20 s
2001). Table 5 shows the correlations used in estimations of mass Feed yH2 ¼ 0:79; yCO2 ¼ 0:17; yCH4 ¼ 0:021; yCO ¼ 0:012; yN2 ¼ 0:007
Q ¼5  10  5 m3 s  1 (measured at 298 K)
and heat parameters for the same experiments (Wasch and
Froment, 1972; Wakao and Funazkri, 1978; Ruthven, 1984; a
Phigh ¼5 bar; Pinter ¼ 2.75 bar ; Plow ¼0.5 bar.
Yang, 1987; Incropera and Witt, 1996; Da Silva, 1999; Bird et al., b
Phigh ¼ 10 bar; Pinter ¼5.5 bar ; Plow ¼ 1 bar.
c
2002; Grande et al., 2008). The simulations were performed using Phigh ¼ 5 bar; Pinter1 ¼3.875 bar; Pinter2 ¼2.75 bar; Pinter3 ¼1.625 bar;
the transport parameters reported in Table 6. This mathematical Plow ¼ 0.5 bar.
d
Phigh ¼ 10 bar; Pinter1 ¼7.75 bar; Pinter2 ¼5.5 bar; Pinter3 ¼3.25 bar;
model was previously employed in simulations of fixed-bed
Plow ¼ 1 bar.
behaviour and/or PSA applications (Da Silva et al., 1999; Grande e
T¼ 303 K.
and Rodrigues, 2005; Cavenati et al., 2006; Grande et al., 2008). f
T¼323 K.
The model described in Table 4 was also applied in the
simulations of one-column VPSA and PSA processes (with the
initial and boundary conditions summarized in Table A1), using PSA processes using only one and three pressure equalization steps,
the correlations shown in Table 5 (Wasch and Froment, 1972; respectively (Skarstrom, 1960). The performances of the different
Wakao and Funazkri, 1978; Ruthven, 1984; Yang, 1987; Incropera simulations of the two VPSA and PSA processes were evaluated
and Witt, 1996; Da Silva, 1999; Bird et al., 2002; Grande et al., 2008). according to three performance parameters: product purity, product
The mathematical model was employed to simulate two different recovery and adsorbent productivity (Rota and Wankat, 1990).
cycles reported in Fig. 2. Cycles A and B are Skarstrom-type cycles of These parameters are defined in Table 7. The simulations were
308 F.V.S. Lopes et al. / Chemical Engineering Science 66 (2011) 303–317

performed using the operation conditions described in Table 8 and finite elements method (40 elements with two interior collocation
the transport parameters reported in Table 6. The mathematical points, 1  10  5 of absolute and relative tolerances). The simulations
model was solved in gPROMS using the orthogonal collocation on of the depressurization steps were performed using an exponential

Fig. 3. H2 ( ) and CO2 (&, —) molar flow rate histories (a) and temperature histories (b) at the column outlet obtained for binary H2-CO2 breakthrough curves at 323 K and
5 bar on AC5-KS; mixtures: CO2 content (balanced with H2) of 30% (1), 10% (2) and 80% (3); temperatures measured at 0.059 m (bottom ( )), 0.151 m (middle ( )) and
0.239 m (top ( )) from feed inlet; flow rate of 5.00  10  5 m3 s  1; experimental data (symbols) simulation (lines); operation conditions described in Table 2.
 F.V.S. Lopes et al. / Chemical Engineering Science 66 (2011) 303–317 309

Fig. 4. H2 ( ) and CO2 (&, —)), CH4 ( ), CO ( ) and N2 ( ) molar flow rate histories (a) and temperature histories (b) at the column outlet obtained for
ternary H2-CO2-CO (1) and for multicomponent H2-CO2-CH4-CO-N2 (2) breakthrough curves at 323 K and 5 bar on AC5-KS; mixtures: 26% of CO2 and 4% of CO (1) and 23% of
CO2, 2.1% of CH4, 1.2% of CO and 0.7% of N2 (2) both balanced with H2; temperatures measured at 0.059 m (bottom ( )), 0.151 m (middle ( )) and 0.239 m (top ( ))
from feed inlet; flow rate of 5.00  10  5 m3 s  1; experimental data (symbols); simulation (lines); operation conditions described in Table 2.
310 F.V.S. Lopes et al. / Chemical Engineering Science 66 (2011) 303–317

decay function until a pre-defined pressure was achieved (Pinter).
When three successive pressure equalizations take place, three
different pressure intervals are defined using exponential functions
with the same exponential decay rate constant (bP).
4. Results and discussion
4.1. Fixed-bed experiments
Prediction of multicomponent adsorption equilibrium data is
normally carried out by extending the mathematical models
(Virial, Langmuir, multi-site Langmuir, etc.) using the coefficients
determined for pure components. One of the objectives of this work
was to validate the results predicted by the extension of the
multicomponent Virial isotherm based on previously reported data
of pure components (Lopes et al., 2009b). Experimental validation
was performed by multicomponent breakthrough experiments in a
fixed-bed.
Binary H2-CO2 breakthrough curves were performed for mix-
tures with different CO2 content (balanced with H2): 30% (run 1),
10% (run 2) and 80% (run 3). All these experiments were carried out
at 323 K of temperature, 5 bar of total feed pressure and using a
feed flow rate of 5.00  10  5 m3 s  1. The results are reported in
Fig. 3. For each breakthrough experiment the temperature profile
was measured at three different positions of the column (0.059,
0.151 and 0.239 m from the feed inlet). It can be observed that
when the concentration of CO2 increases, the heat generated by
adsorption of CO2 rises resulting in higher temperatures inside the
column. Since H2 is only slightly adsorbed, it breaks through the
column very fast. The CO2 profiles at the column exit are very sharp,
especially at higher CO2 partial pressures due to the non-linearity
of the isotherm. The mathematical model could satisfactorily
describe the binary data. This binary system is particularly
important because one gas is the desired product and the other
gas is the main contaminant. It should be mentioned however,
that parts of these curves were used to adjust thermal parameters
(hw and U).
The results of the ternary H2-CO2-CO breakthrough curve (run 4,
26% of CO2, 4% of CO balanced with H2) are shown in Fig. 4. Ternary
mixtures with carbon monoxide should be well described since the
specifications for high-quality hydrogen are in the ppm level. It can
be observed in Fig. 4 that the breakthrough time of CO can be
predicted by the model confirming the validity of the multi-
component extension of the Virial model for this ternary mixture
(at least for this composition). Since CO is the intermediate
compound adsorbed, it presents two different plateaus. Although
it is not possible to see it in detail in Fig. 4, the adsorption of CO
produces an independent temperature peak before the larger peak
due to adsorption of CO2.
Finally, a five-component H2-CO2-CO-CH4-N2 breakthrough
curve (23% of CO2, 2.1% of CH4, 1.2% of CO and 0.7% of N2 balanced
with H2) at the same feed flow rate was performed (run 5). The
experimental and simulated results obtained are reported in Fig. 4.
Again, for the five-component mixture, the mathematical model can
successfully predict the data. In this experiment the concentrations
of N2, CO and CH4 are quite small so is difficult to observe the
successive plateaus of the lighter components (N2 and CO). However,
it is clear that there is a good separation zone between hydrogen and
the rest of the contaminants. The first contaminant to breakthrough

Fig. 5. Pressure (a; J,—) and temperature (b) histories; cycle sequence: (1) co-current feed step; (2) first, (3) second and (4) third pressure equalization–depressurization
steps; (5) counter-current blowdown step; (6) counter-current purge step with hydrogen; (7) third, (8) second and (9) first pressure equalization–pressurization steps;
(10) counter-current pressurization step with hydrogen; temperatures measured at 0.059 m (bottom ( )), 0.151 m (middle ( )) and 0.239 m (top ( )) from feed
inlet; experimental data (symbols); simulation (lines); step times and operation conditions described in Table 2.
 F.V.S. Lopes et al. / Chemical Engineering Science 66 (2011) 303–317 311

Fig. 6. H2 ( ), CO2 (&, —), CH4 ( ), CO ( ) and N2 ( ) molar flow rate histories (a) obtained for cycle 1 (1) and 37 (2) at bed outlet; Temperature histories
(b) measured for cycle 1 (1) and 37 (2) at 0.059 m (bottom ( )), 0.151 m (middle ( )) and 0.239 m (top ( )) from feed inlet; flow rate of 5.00  10  5 m3 s  1;
experimental data (symbols); simulation (lines); step times and operation conditions described in Table 2.
312 F.V.S. Lopes et al. / Chemical Engineering Science 66 (2011) 303–317

the column is nitrogen, followed by carbon monoxide, methane and
finally carbon dioxide. In this case also, the adsorption of the
different gases produces independent temperature peaks.
Note that when the feed gas stream reaches the adsorbent at the
beginning of each experiment, the heat of adsorption of the faster
species present in the mixture starts to increase (locally) the
temperature of the adsorbent. Consequently, the temperature of
the column wall increases in the same area where the temperature
sensor of the flexible insulated heater was placed (at the beginning
of the column, between the heater and the column wall). Therefore,
the heater does not generate more energy (it switches off) for
heating of the column and thus the temperature in the remaining
column decreases. For this reason, it can be observed in both Figs. 3
and 4 that a steep decrease of temperature occurs at the initial
period of time.
4.2. VPSA experiment
After confirmation of the validity of multicomponent adsorp-
tion equilibrium and kinetic data, a VPSA experiment was per-
formed to validate the mathematical model in a cyclic experiment.
A one-column VPSA experiment was performed using the cycle
scheduling illustrated in Fig. 2b. The experimental run was
performed at 303 K at a feed flow rate of 5.00  10  5 m3 s  1 and
the following composition: 79% of H2, 17% of CO2, 2.1% of CH4, 1.2%
of CO and 0.7% of N2. The feed (Phigh) pressure was fixed at 5 bar and
the blowdown pressure (Plow) was 0.5 bar. The experimental
conditions are shown in Table 2. In this one-column experiment,
the pressure equalization was divided into three different steps to
simulate the behaviour of an industrial PSA unit where several
pressure equalization steps are required to increase H2 recovery.
The pressure history of cycle 1 and the temperature histories of
the first ten cycles are shown in Fig. 5. The H2, CO2, CH4, CO and N2
molar flow rate histories obtained at cycles 1 and 37 at bed outlet
are reported in Fig. 6. According to the simulation, the process takes
240 cycles to reach the cyclic steady state (CSS). We have assumed
that CSS was achieved when the adsorbed phase concentration of
each gas in the column became equal for two consecutive cycles.
In these two cycles, a good agreement between the simulations and
the experimental molar flow rates was observed. During the
pressure equalization steps a considerable amount of hydrogen
is released from the column, which can be used for pressure
equalization or for purging. The temperature predicted by the
mathematical model is in reasonable agreement with the experi-
mental results, with deviations lower than 1 K. It should be
observed that temperature variations are larger in the initial
portion of the column, where CO2 is adsorbed. In the rest of the
column (used to remove the other contaminants) the temperature
oscillations are smaller than 4 K.
The performance of the 10 steps one-column VPSA process
experiment was evaluated according to the parameters reported in
Table 7. It was observed that 81.6% of hydrogen was recovered with
a purity of 99.981% contaminated with 63 ppm of carbon monoxide.
1 1
The adsorbent productivity attained was 101 molH2 kgads day .
Ribeiro et al. (2008) reported a hydrogen recovery of 51.84% for an 8
steps one-column layered bed (activated carbon and zeolite) PSA
system that produces 99.9994% of pure hydrogen operating at a feed
pressure of 7 bar and temperature of 303 K. For a similar system
(H2, CO2, CH4 and CO; feed: 1.17  10-4 m3 s  1, 9.12 bar), Lee et al.
(2008) reported 6 steps one-column PSA simulations that were
able to purify 74-76% of hydrogen with much lower purities of
99.4–99.7%. It should be mentioned that the feed pressure of this

Fig. 7. Pressure (a) and molar flow rate (b, c) histories obtained for simulation 13 at the bed outlet at cyclic steady state: (1) co-current feed step; (2) first, (3) second and
(4) third pressure equalization–depressurization steps; (5) counter-current blowdown step; (6) counter-current purge step with hydrogen; (7) third, (8) second and (9) first
pressure equalization–pressurization steps; (10) counter-current pressurization step with hydrogen; step times and operation conditions described in Table 8.
 F.V.S. Lopes et al. / Chemical Engineering Science 66 (2011) 303–317 313

experiment is quite low when compared to most reported values predict the experimental results and thus we can use the model to
from literature (using normally 20 bar). The main conclusion from gain a deeper understanding of the cycle and of the important
this experiment is that the mathematical model can successfully operating variables of the process.

Fig. 8. Solid phase concentration profiles of H2 (a), CO2 (b), CH4 (c), CO (d) and N2 (e) obtained for simulation 13 along the bed of each step at cyclic steady state; step times and
operation conditions described in Table 8.
314 F.V.S. Lopes et al. / Chemical Engineering Science 66 (2011) 303–317
4.3. VPSA and PSA simulations
In the design of a PSA for H2 purification, two different types of
variables can be differentiated. The first ones are the operating
variables, particularly the ones related to feed conditions (composi-
tion, pressure and temperature). The second ones are related to the
design of the unit (process variables), which are the step times and
the blowdown and purge pressures as well as the cycle proposed.
With regard to the operating variables, we have performed
experiments at low pressures and we are also interested in
exploring the performance of the process at higher pressures.
Among the design variables, the effect of pressure equalization
steps should be accounted for. If a cycle with three pressure
equalization steps is employed, a large number of columns should
be used. However, while using three pressure equalizations, the
pressure reduction in the blowdown step is only 1/4 the ratio of
Phigh/Plow (from 1.625 to 0.500 bar) seen in the previous example. If
only one pressure equalization step is used, the ratio increases to
1/2, releasing much more H2 in the blowdown step, thereby losing it.
We have performed different simulations for a mixture with
feed composition of 0.79 H2/0.17 CO2/0.021 CH4/0.012 CO/0.007 N2
and at a feed flow rate of 5.00  10  5 m3 s  1 available at 303 or at
323 K. The same column properties as those used in the experi-
ments were used in the simulations (see details in Table 1). The full
set of operating conditions of the simulations is reported in Table 8,
together with the performance parameters of the cycle in CSS (H2
purity, H2 recovery and adsorbent productivity).
Initially, a 6 steps (cycle A) one-column VPSA process was
simulated (only one pressure equalization step). Two different purge
times (simulations 1 and 2) were tested to analyse their influence on
the hydrogen purity stream at the end of the feed step. It was
observed that by increasing the purge time by 50%, a significant
reduction in the contamination by carbon monoxide in the product
stream was attained. For a cycle sequence that comprises a single
pressure equalization (cycle A), the blowdown starts from an
intermediate pressure that is very high and, thus, the amount of
hydrogen ‘‘lost’’ in this step is significant. The step times used in
simulation 1 were close to the ones used in the VPSA experiment (see
Table 8). It can be observed that with only one pressure equalization,
the purity of the H2 obtained is much higher at the expense of losing
an additional 10% of H2 in the blowdown. In fact, when more
pressure equalizations are employed, the concentration of contami-
nants increases at the product end, contaminating the stream of H2.
To reduce the amount of hydrogen lost in the blowdown step, a PSA
cycle with three pressure equalizations was used; cycle B is shown in
Fig. 2. Simulations 3 and 4 were performed utilizing the same step
times of simulations 1 and 2, respectively, with the consideration that
the total pressure equalization step times were kept constant, i.e., tED
(cycle A) is equal to the sum of tED1, tED2 and tED3 (cycle B) at the same
step pressures and the same temperature. As expected, the hydrogen
recovery attained in this process was higher (increased 16%). Never-
theless, the purity of the hydrogen produced decreased considerably,
reaching levels of CO contamination two orders of magnitude higher.
The effect of purging can also be evaluated in these simulations; by
increasing the purge time by 50% a drastic reduction in CO contam-
ination can be achieved, while losing less than 3% of H2.
Alternatively, another pressure range was studied for both VPSA
processes in order to avoid the use of vacuum in the blowdown and
purge steps. In simulation 5, the high and low pressures applied in the
6 steps (cycle A) one-column PSA process were 10 and 1 bar,
respectively. As can be observed from Table 8, almost 100% pure H2
was produced. However the amount of hydrogen recovered and the
adsorbent productivity significantly decreased. Even when the feed
step time was increased by 20% (simulation 6), both process perfor-
mance parameters, hydrogen recovery and adsorbent productivity,
remained very low. In similar simulations using the cycle sequence B
(simulations 7 and 8, respectively), considering higher H2 recoveries
were obtained at the expense of reduction in their purity. However, in
most of the simulations, the H2 purity is higher than 99.99%, which is
the target purity. In these simulations, since feed step time increased, it
was seen that higher unit productivities were obtained.
Afterwards, the blowdown, purge and feed step times were
analysed. First, by reducing the blowdown step time by 25%
(to 150 s), it was verified that the hydrogen purity was insign-
ificantly affected (but that the adsorbent productivity increased by
5% as shown in simulation 9). A decrease of 33% in the purge step
time results in better unit productivity and recovery, but at the
expense of retaining 26 ppm of CO in the product.
Another important consideration is the temperature of the feed
stream. We have also performed some simulations considering the
feed temperature to be 323 K. In this case, using step times with
greater similarity than in previous simulations (compare simulation
15 with simulation 7) a higher contamination by CO was found. In
fact, at a higher temperature the loading of all contaminants is smaller
and these can breakthrough faster, contaminating the product. When
the blowdown step time is further reduced, the CO contamination
increases yet again, following the same trend observed in the
simulations at 303 K. Higher purity can also be obtained by increasing
the purge time, with the expense of 2% less in hydrogen recovery.
In order to understand the simulated results in detail, in Figs. 7–10
we present several results of the one-column PSA process simulation
13 that is reported in Table 8. The pressure evolution along one cycle
at the outlet of the bed is shown for the simulation 13 in Fig. 7. The
same figure shows the molar flow rates obtained for the five gases of
the feed mixture during one cycle in the cyclic steady state (CSS). As
previously stated, the high amounts of pure hydrogen released from
the column during the pressure equalization–depressurization steps
can be recycled, improving the recovery of the process. However, a
significant amount of hydrogen is lost in the blowdown step. More-
over, for the pressurization of the column a fraction of the hydrogen
produced is used, decreasing the recovery of the process. Therefore,
the reuse of the hydrogen released from the column during the
pressure equalization–depressurization steps is crucial to attain high
hydrogen process recoveries.
The solid phase concentration profiles obtained along the bed at
the end of each step at CSS are presented in Fig. 8. Observing these
figures, it can be concluded that the impurities detected at the feed
outlet stream were CH4, CO and, mainly, N2. This fact is in accordance
with the adsorption capacities and diffusivity time constants of the
Fig. 9. Hydrogen purity and recovery evolution along the cycles of simulation 13;
step times and operation conditions described in Table 8.
 F.V.S. Lopes et al. / Chemical Engineering Science 66 (2011) 303–317 315

Fig. 10. Evolution of the gas temperature (a) and nitrogen gas phase concentration (b) profiles with the cycle number at the end of the feed step obtained for simulation 13;
step times and operation conditions described in Table 8.

pure gases of this activated carbon as reported by Lopes et al. (2009a, Several VPSA and PSA simulations were also performed for
2009b); carbon dioxide has the highest adsorption capacity followed different cycles and under operating conditions. It was observed
by CH4, CO, N2 and finally H2; hydrogen has the highest micropore that by using three different pressure equalizations it is possible to
diffusivity followed by N2, CO, CH4 and finally CO2 (Lopes et al., achieve H2 recoveries 15% higher than those achieved by using only
2009b). Fig. 8 shows that the solid phase concentration profile is only one pressure equalization. However, when multiple equalizations
slightly affected in the equalization–depressurization steps; the short are used, the concentration of the contaminants reaches the
duration of these steps does not allow significant desorption of pre- product end faster, contaminating the product. According to
adsorbed gases from the activated carbon. simulation results increasing the feed pressure makes it possible
Fig. 9 shows the evolution of the hydrogen purity and recovery to increase the unit productivity while keeping H2 purity over
along the cycle number in simulation 13. In the same simulation, 99.99% with carbon monoxide contamination at a level lower than
the evolution of the gas temperature profile and gas phase 10 ppm.
concentration profile of the less adsorbed contaminant specie
(N2), at the end of the feed step, with the cycle number is reported
in Fig. 10. Observing Fig. 10, it can led us to think that CSS was Nomenclature
reached after 50 cycles. However, the ‘‘very’’ small variations in the
stabilization of the temperatures and gas phase concentrations Latin letters
profiles within the bed delayed the achievement of the CSS, thus ap particle specific area (m  1)
generating long simulation times. For this reason, and as it is A Virial coefficients (m2 mol  1)
possible to observe in Fig. 9, the CSS was reached only after B Virial coefficients (m4 mol  2)
approximately 200 cycles. Bii mass Biot number of component i
C concentration (mol m  3)
Cg,i component i gas phase concentration (mol m  3)
5. Conclusions Cg,T total gas phase concentration (mol m  3)
Cm,i average concentration of component i in the macropores
In this work, multicomponent breakthrough experiments (bin- (mol m  3)
ary H2-CO2, ternary H2-CO2-CO and five-component H2-CO2-CO- Cm,T average total concentration in the macropores (mol m  3)
CH4-N2) were performed under different operating conditions in Cp gas mixture molar specific heat at constant pressure
activated carbon extrudates to validate the single-component (J mol  1 K  1)
adsorption equilibrium. The experiments show that the multi- C^ p,s particle specific heat at constant pressure [per mass unit]
component extension of the Virial model could successfully (J kg  1 K  1)
predict the adsorption equilibrium capacities for all mixtures ^C p,w wall specific heat at constant pressure [per mass unit]
(using single-component parameters). Furthermore, the mathe- (J kg  1 K  1)
matical model used for the fixed-bed could also be used to predict Cs,i component i concentration at the solid surface (mol m  3)
the column behaviour of the multicomponent mixtures. Cv gas mixture molar specific heat at constant volume
A 10 steps one-column VPSA experiment was performed for (J mol  1 K  1)
the complete validation of the mathematical model under cyclic Cv,i molar specific heat of component i at constant volume
conditions. A cycle sequence comprising feed, three different depres- (J mol  1 K  1)
surizations, blowdown, purge, three pressure equalizations and pres- Cv,ads,i molar specific heat of component i in the adsorbed phase
surization was used. In the experiment, a H2 stream with 99.981% at constant volume (J mol  1 K  1)
hydrogen purity (63 ppm CO contaminated) was produced, with a dp particle diameter (m)
1 1
recovery of 81.6% and unit productivity of 101 molH2 kgads day . dwe wall external diameter (m)
316 F.V.S. Lopes et al. / Chemical Engineering Science 66 (2011) 303–317

dwi wall internal diameter (m) rg density of the gas phase (kg m  3)
Dax axial dispersion coefficient (m2 s  1) rp adsorbent (particle) density (kg m  3)
Dc micropore diffusivity (m2 s  1) rw wall density (kg m  3)
D0c limiting diffusivity at infinite temperature (m2 s  1)
Dm molecular diffusivity (m2 s  1)
Dp macropore diffusivity (m2 s  1) Acknowledgments
e wall thickness (m)
Ea activation energy of micropore diffusion (kJ mol  1) F.V.S. Lopes gratefully acknowledges Fundac- a~ o para a Ciência e a
g acceleration due to gravity (m s  2) Tecnologia (FCT, Portugal) for his Ph.D. Research Fellowship, no.
Gr Grashof number SFRH/BD/44563/2008.
hex external convective heat transfer coefficient (J s  1 m  2 K  1)
hf film heat transfer coefficient between the gas phase and
particle (J s  1 m  2 K  1) Appendix A
hw film heat transfer coefficient between the gas phase and
wall (J s  1 m  2 K  1) See Table A1.
kf film mass transfer coefficient (m s  1)
kg gas conductivity (J s  1 m  1 K  1) Table A1
KH Henry constant (mol kg  1 bar  1) Initial and boundary conditions for fixed-bed, VPSA and PSA simulations.
KN adsorption constant at infinite temperature
Initial conditions for fixed-bed, VPSA and PSA simulations
(mol kg  1 bar  1)
yi ¼ Cp,i ¼ qi ¼ 0 for i a inert, yinert ¼1; Cm,inert ¼ Cg,T ; Tg ¼Tp ¼Tw ¼Tinlet
L column length (m)
Nu Nusselt number Boundary conditions for fixed-bed simulations
z ¼0 (inlet) z ¼L (outlet)
P pressure (Pa)
@y @Cg,i
Pr Prandtl number u0 inlet Cinlet,i ¼ u0 Cg,i eDax Cg,T i ¼0
@z @z
q absolute adsorbed phase concentration (mol kg  1) u0inletCinlet,T ¼ u0Cg,T P¼ Pexit
qi particle averaged adsorbed concentration (mol kg  1) u0inlet Cinlet,T Cp Tinlet ¼ u0 Cg,T Cp Tg l
@Tg @Tg
¼0
@z @z
qi adsorbed concentration in equilibrium with Cm,i
(mol kg  1) Boundary conditions for VPSA and PSA simulations
Co-current feed
rc micropore radius (m) z ¼0 (inlet) z ¼L (outlet)
Rg universal gas constant (J mol  1 K  1) @y @Cg,i
u0inlet Cinlet,i ¼ u0 Cg,i eDax Cg,T i ¼0
Rp radius of the adsorbent extrudates (m) @z @z
Ra Rayleigh number P ¼Pinlet P¼ Pexit
@Tg @Tg
Re Reynolds number u0inlet Cinlet,T Cp Tinlet ¼ u0 Cg,T Cp Tg l
@z @z
¼0
S adsorbent specific area (m2 kg  1)
Pressure equalization–depressurization
Sc Schmidt number z ¼0 z ¼L (outlet)
Sh Sherwood number @Cg,i @Cg,i
¼0 ¼0
t time (s) @z @z
u0 ¼0 P ¼ Phigh þ ðPinter Phigh Þ½1expðbP tÞ
T temperature (K)
@Tg @Tg
Tfilm film temperature (K) @z
¼0
@z
¼0
Tg gas temperature (K)
Counter-current blowdown
Tp particle temperature (K) z ¼0 (outlet) z ¼L
Tw wall temperature (K) @Cg,i @Cg,i
¼0 ¼0
TN external temperature (K) @z @z
P ¼ Pinter þ ðPlow Pinter Þ½1expðbP tÞ u0 ¼ 0
u0 superficial velocity (m  s  1)
@Tg
U overall heat transfer coefficient (J s  1 m  2 K  1) @z
¼0
yi molar fraction of component i @Tg
¼0
z axial position @z
Counter-current purge with hydrogen
z ¼0 (outlet) z ¼L (inlet)
@Cg,i @yi
¼0 u0inlet Cinlet,i ¼ u0 Cg,i eDax Cg,T
Greek letters @z @z
a external thermal diffusivity (W m  1 K  1) P ¼Pexit u0inletCinlet,T ¼ u0Cg,T
@Tg @Tg
aw ratio of the internal surface area to the volume of the
@z
¼0 u0inlet Cinlet,T Cp Tinlet ¼ u0 Cg,T Cp Tg l
@z
column wall (m  1)
aw‘ ratio of the logarithmic mean surface area of the column
Pressure equalization–pressurization
shell to the volume of the column wall (m  1) z ¼0 z ¼L (inlet)
b thermal expansion coefficient (m m  1 K  1) @Cg,i @yi
¼0 u0inlet Cinlet,i ¼ u0 Cg,i eDax Cg,T
bP exponential decay rate constant @z @z
u0 ¼0 u0inletCinlet,T ¼ u0Cg,T
 DH isosteric heat of adsorption (kJ mol  1)
@Tg @Tg
e bed porosity @z
¼0 u0inlet Cinlet,T Cp Tinlet ¼ u0 Cg,T Cp Tg l
@z
ep particle porosity
Counter-current pressurization with hydrogen
l heat axial dispersion coefficient (J s  1 m  1 K  1) z ¼0 z ¼L (inlet)
lw wall conductivity (J s  1 m  1 K  1) @Cg,i @Tg
¼0 u0inlet Cinlet,i ¼ u0 Cg,T Cp Tg l
m gas viscosity (Pa s) @z @z
u0 ¼0 P¼ Pinter + (Phigh  Pinter)[t/tpress]
n kinematic viscosity of the external gas (m2 s  1) @y
@Tg
rb bulk density of the bed (kg m  3) ¼0 u0inlet Cinlet,T Cp Tinlet ¼ u0 Cg,i eDax Cg,T i
@z @z
 F.V.S. Lopes et al. / Chemical Engineering Science 66 (2011) 303–317
References
Ahn, H., Lee, C.-H., Seo, B., Yang, J., Baek, K., 1999. Backfill cycle of a layered bed H2
PSA process. Adsorption 5, 419–433.
Barrer, R.M., 1981. Sorption in porous crystals: equilibria and their interpretation.
Journal of Chemical Technology and Biotechnology 31, 71–85.
Bec, R.L., 2005. Method for purifying hydrogen-based gas mixtures using calcium
X-zeolite. U.S. Patent 6,849,106 B2.
Bird, R.B., Stewart, W.E., Lightfoot, E.N., 2002. Transport Phenomena. Wiley
International, Singapore.
Cavenati, S., Grande, C.A., Rodrigues, A.E., 2006. Separation of CH4/CO2/N2 mixtures
by layered pressure swing adsorption for upgrade of natural gas. Chemical
Engineering Science 61, 3893–3906.
Da Silva, F.A., 1999. Cyclic Adsorption Processes: Application to Propane/Propylene
Separation. University of Porto, Porto, Portugal.
Da Silva, F.A., Rodrigues, A.E., 2001. Vacuum swing adsorption for propylene/
propane separation with 4A zeolite. Industrial and Engineering Chemistry
Research 40, 5758–5774.
Da Silva, F.A., Silva, J.A., Rodrigues, A.E., 1999. A general package for the simulation of
cyclic adsorption processes. Adsorption 5, 229–244.
DOE, 2000. Fuel Cell Handbook. DOE, EG&G Services, Parsons, Inc. Science Applica-
tions International Corporation. Contract no. DE-AM26-99FT40575.
Esteves, I.A.A.C., Mota, J.P.B., 2002. Simulation of a new hybrid membrane/pressure
swing adsorption process for gas separation. Desalination 148, 275–280.
Grande, C.A., Lopes, F.V.S., Ribeiro, A.M., Loureiro, J.M., Rodrigues, A.E., 2008.
Adsorption of off-gases from steam methane reforming (H2, CO2, CH4, CO and
N2) on activated carbon. Separation Science and Technology 43, 1338–1364.
Grande, C.A., Rodrigues, A.E., 2005. Propane/propylene separation by pressure swing
adsorption using zeolite 4A. Industrial and Engineering Chemistry Research 44,
8815–8829.
Incropera, F.P., Witt, D.P.D., 1996. Fundamentals of Heat and Mass Transfer. John
Wiley & Sons, New York.
Kärger, J., Ruthven, D.M., 1992. Diffusion in Zeolites and other Microporous Solids.
John Wiley & Sons, New York.
Khelifa, A., Derriche, Z., Bengueddach, A., 1999. Sorption of carbon dioxide by zeolite
X exchanged with Zn2 + and Cu2. Microporous and Mesoporous Materials 32,
199–209.
Kiselev, A.V., 1971. Vapor adsorption on zeolites considered as crystalline species
adsorbents. Molecular Sieve Zeolites-II, ACS Series 102, 37.
Lee, J.-J., Kim, M.-K., Lee, D.-G., Ahn, H., Kim, M.-J., Lee, C.-H., 2008. Heat-exchange pressure
swing adsorption process for hydrogen separation. AIChE Journal 54, 2054–2064.
Lopes, F.V.S., Grande, C.A., Ribeiro, A.M., Loureiro, J.M., Evaggelos, O., Nikolakis, V.,
Rodrigues, A.E., 2009a. Adsorption of H2, CO2, CH4, CO, N2 and H2O in activated
carbon and zeolite for hydrogen production. Separation Science and Technology
44, 1045–1073.
Lopes, F.V.S., Grande, C.A., Ribeiro, A.M., Oliveira, E.L.G., Loureiro, J.M., Rodrigues,
A.E., 2009b. Enhancing capacity of activated carbons for hydrogen production.
Industrial and Engineering Chemistry Research 48, 3978–3990.
317
Lopes, F.V.S., Grande, C.A., Ribeiro, A.M., Vilar, V.J.P., Loureiro, J.M., Rodrigues, A.E.,
2010. Effect of ion exchange on the adsorption of SMR off-gases on zeolite 13X.
Journal of Chemical and Engineering Data 55, 184–195.
Malek, A., Farooq, S., 1998. Hydrogen purification from refinery fuel gas by pressure
swing adsorption. AIChE Journal 44, 1985–1992.
OECD/IEA, 2004. Hydrogen & Fuel cells: Review of National R&D Programs.
Organisation for Economic Co-operation and Development & International
Energy Agency. ISBN: 92-64-108-831.
Park, J.-H., Kim, J.-N., Cho, S.-H., 2000. Performance analysis of a four-bed H2 PSA
process using layered beds. AIChE Journal 46, 790–802.
Ribeiro, A.M., Grande, C.A., Lopes, F.V.S., Loureiro, J.M., Rodrigues, A.E., 2008. A
parametric study of layered bed PSA for hydrogen purification. Chemical
Engineering Science 63, 5258–5273.
Ritter, J.A., Yang, R.T., 1987. Equilibrium adsorption of multicomponent gas mixtures
at elevated pressures. Industrial and Engineering Chemistry Research 26,
1679–1686.
Rota, R., Wankat, P.C., 1990. Intensification of pressure swing adsorption processes.
AIChE Journal 36, 1299–1312.
Ruthven, D.M., 1984. Principles of Adsorption and Adsorption Processes. John Wiley
& Sons, New York.
Ruthven, D.M., Farooq, S., Knabel, K.S., 1994. Pressure Swing Adsorption. VCH
Publishers, New York.
Sircar, S., Golden, T.C., 2000. Purification of hydrogen by pressure swing adsorption.
Separation Science and Technology 35, 667–687.
Sircar, S., Waldron, W.E., Rao, M.B., Anand, M., 1999. Hydrogen production by hybrid
SMR–PSA–SSF membrane system. Separation and Purification Technology 17,
11–20.
Skarstrom, C.W., 1960. Fractionating gas mixtures by adsorption. U.S. Patent
2,944,627.
Taqvi, S.M., LeVan, M.D., 1997. Virial description of two-component adsorption
on homogeneous and heterogeneous surfaces. Industrial and Engineering
Chemistry Research 36, 2197–2206.
Wagner, J.L., 1969. Selective adsorption process. U.S. Patent 3,430,418.
Wakao, N., Funazkri, T., 1978. Effect of fluid dispersion coefficients on particle-to-
fluid mass transfer coefficients in packed beds—correlation of Sherwood
numbers. Chemical Engineering Science 33, 1375–1384.
Wasch, A.P.D., Froment, G.F., 1972. Heat transfer in packed beds. Chemical
Engineering Science 27, 567.
Yang, J., Han, S., Cho, C., Lee, C.-H., Lee, H., 1995. Bulk separation of
hydrogen mixtures by a one-column PSA process. Separations Technology 5,
239–349.
Yang, J., Lee, C.-H., 1998. Adsorption dynamics of a layered bed PSA for H2 recovery
from coke oven gas. AIChE Journal 44, 1325–1334.
Yang, R.T., 1987. Gas Separation by Adsorption Processes. Butterworths, Boston.
Zhou, L., Lü, C.-Z., Bian, S.-J., Zhou, Y.-P., 2002. Pure hydrogen from the dry gas of
refineries via a novel pressure swing adsorption process. Industrial and
Engineering Chemistry Research 41, 5290–5297.