Separation and Purication Technology 77 (2011) 251260

Contents lists available at ScienceDirect

Separation and Purication Technology

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

Dynamic bed measurements of CO adsorption on microporous adsorbents at high

pressures for hydrogen purication processes
M. Bastos-Neto a , A. Moeller a , R. Staudt b, , J. Bhm c , R. Glser a,c
a
Institut fr Nichtklassische Chemie e.V., Permoserstr. 15, 04318 Leipzig, Germany
b
Fakultt Maschinenbau und Verfahrenstechnik (M+V), Hochschule Offenburg, Badstrae 24, 77652 Offenburg, Germany
c
Institut fr Technische Chemie, Universitt Leipzig, Linnstr. 3, 04103 Leipzig, Germany

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

Article history: Regarding the importance of adsorptive removal of carbon monoxide from hydrogen-rich mixtures
Received 1 October 2010 for novel applications (e.g. fuel cells), this work provides a series of experimental data on adsorption
Received in revised form 7 December 2010 isotherms and breakthrough curves of carbon monoxide. Three recently developed 5A zeolites and one
Accepted 10 December 2010
commercial activated carbon were used as adsorbents. Isotherms were measured gravimetrically at tem-
peratures of 278313 K and pressures up to 0.85 MPa. Breakthrough curves of CO were obtained from
Keywords:
dynamic column measurements at temperatures of 298301 K, pressures ranging from 0.1 MPa to ca.
Adsorption
6 MPa and concentrations of CO in H2 /CO mixtures of 517.5 mol%. A simple mathematical model was
Fixed bed
Simulation
developed to simulate breakthrough curves on adsorbent beds using measured and calculated data as
Modeling inputs. The number of parameters and the use of correlations to evaluate them were restricted in order
Separation to focus the importance of measured values. For the given assumptions and simplications, the results
Hydrogen purication show that the model predictions agree satisfactorily with the experimental data at the different operating
Carbon monoxide conditions applied.
2010 Elsevier B.V. Open access under the Elsevier OA license.

1. Introduction gel to essentially adsorb H2 O; the second is composed of acti-

vated carbon, which adsorbs CH4 , CO, CO2 and traces of sulfur
In spite of not being a primary energy source hydrogen needs to components; and as a third layer, zeolites are used for improved
be produced and consequently puried. The current main source adsorption of CO, N2 and other trace components [8,9]. Therefore,
of hydrogen is the processing of fossil derivates like methane, coal factors like the choice of material, the relative length (ratio) of
and heavy hydrocarbons. Other sources include electrolysis and the layers, composition of the feed gas and interactions gassolid
biological processes [13]. can signicantly inuence the yield and efciency of the process
The most common and economical way to produce hydrogen is [5,810].
through steam reforming of natural gas combined with a watergas Despite the remarkable growth in practical applications of
shift reaction, from which a hydrogen-rich stream (7080%) con- adsorptive gas separation, the commercial design and optimiza-
taining impurities, namely hydrogen sulde (traces), water vapor tion of these processes still largely remain an experimental effort.
(<1%), nitrogen (<1%), methane (36%), carbon monoxide (13%) This is primarily due to the inherent complex nature of the practical
and carbon dioxide (1525%), is produced [46]. In order to obtain adsorption systems [11].
hydrogen with the desired purity these contaminants must be The most important challenges in this area today are (i) a deeper
removed. The removal process is industrially carried out by pres- understanding of the equilibrium, the dynamics and thermal effects
sure swing adsorption (PSA), which is usually designed to obtain a involved in the separation of mixtures; and (ii) the development
hydrogen stream containing 9899.999 mol% H2 [6,7]. of less expensive and less time-consuming models to describe the
In practice, PSA units for hydrogen purication use up to three inuence of the equilibrium, kinetics and heat effects on such
different adsorbent layers. The rst layer reached by the feed processes. Todays models frequently cannot predict data with
mixture, usually a guard bed, is composed of alumina or silica the accuracy and reliability required by industry [1113]. A more
detailed covering of the contemporary research needs with inter-
esting examples of the current industrial design requirements can
Abbreviations: AC, activated carbon; PSA, pressure swing adsorption; SATP,
be found elsewhere [11].
Standard Ambient Temperature and Pressure (25 C and 100.000 kPa). To achieve the desired reliability of the model it is necessary to
Corresponding author. Tel.: +49 0341 235 2405; fax: +49 0341 235 2701. have a considerable amount of experimental data, including mainly
E-mail address: Reiner.Staudt@fh-offenburg.de (R. Staudt).

1383-5866 2010 Elsevier B.V. Open access under the Elsevier OA license.
doi:10.1016/j.seppur.2010.12.015
252 M. Bastos-Neto et al. / Separation and Purication Technology 77 (2011) 251260

adsorption isotherms and breakthrough curves, which can pro-

Nomenclature vide a direct and realistic interpretation of the adsorption process.
By measuring breakthrough curves, one can evaluate the amount
A bed sectional area (m2 ) adsorbed, the inuence of other species and the interval required
Cg,i component concentration in the gas phase by each step (adsorptiondesorption).
(mol kg1 ) Most of the available data on breakthrough curves of CH4 , CO,
cpg,i component molar specic heat at constant pressure CO2 and N2 in hydrogen are results of simulations, mainly those at
(J mol1 K1 ) high pressures (above 2.0 MPa), providing few experimental data.
cps particle specic heat at constant pressure The current literature lacks measured data at different pressure and
(J kg1 K1 ) temperature conditions, especially in newly developed materials
cpw wall specic heat at constant pressure (J kg1 K1 ) [12,13].
Dax axial mass dispersion coefcient (m2 s1 ) Carbon monoxide was chosen as a component to be removed
di column internal diameter (m) from H2 -rich mixtures for this study, since only few experimen-
Dm molecular diffusivity (m2 s1 ) tal data are available from the literature and the importance of its
ew wall thickness (m) concentration on H2 for fuel cells applications. CO tends to act as
hw lm heat transfer coefcient between the gas and a severe poison towards the platinum catalyst in a PEM fuel cell
wall (W m2 K1 ) and for this reason its concentration must be generally reduced to
keff effective mass transfer coefcient (s1 ) levels as low as 10 ppm [14].
kg gas thermal conductivity (W m1 K1 ) The main goal of this study was to provide a series of experimen-
Ki Toth isotherm parameter afnity tal data for breakthrough curves and adsorption isotherms of CO on
K0,i afnity at reference temperature T0 different materials, including an activated carbon and new com-
L bed length (m) mercial zeolites, with different feed compositions and pressures
mE adsorbent mass (kg) ranging from 0.1 to 6 MPa. Then, a simple model was applied using
ni Toth heterogeneity parameter measured parameters (obtained from adsorption isotherms, bed
n0,i heterogeneity parameter at reference temperature geometry, etc.) and properties evaluated according to correspond-
T0 ing literature data, for instance heat capacity and conductivity, as
p total pressure (MPa) input values. After that, this model was validated by comparing
q i component average concentration on the adsorbed simulations with experiments.
phase (mol kg1 ) With the modeling and experimental investigation, this study
Qi heat of adsorption for fractional loading equal to also aimed a deeper insight into the adsorptive separation of carbon
zero (J mol1 ) monoxide from hydrogen on xed beds, by analyzing the inuence
qi adsorbed concentration in equilibrium with Cg,i of the adsorption equilibrium and operational variables such as the
(mol kg1 ) feed ow rate, pressure and composition on the performance of a
qmax,i Toth isotherm parameter maximal concentration separation unit.
(mol kg1 )
qmax0,i maximum adsorbed amount at reference tempera-
ture T0 (mol kg1 ) 2. Model
R ideal gas constant (J mol1 K1 )
R2 coefcient of determination for the isotherm model In this section we provide a simple approach to the calculation of
tting xed bed adsorption. A comprehensive summary of models devel-
rp average particle radius (m) oped to analyze data and predict results for a variety of adsorption
t time (s) systems has been reported in the literature [15].
T room temperature (K) In order to describe the dynamic behavior of separation pro-
T0 reference temperature for isotherm tting (K) cesses in xed beds a general mathematical model was used, which
T bulk phase temperature (K) consists of the coupled mass and energy balances with suitable
Tw wall temperature (K) boundary and initial conditions. The set of differential equations
u advection (interstitial) velocity (m s1 ) was solved by numerical integration using the Euler method, writ-
Ug overall heat transfer coefcient (W m2 K1 ) ten in Borland Delphi 3.0. The model was designed based on the
yi component molar fraction system schematically illustrated in Fig. 1 with assumptions com-
z axial position (m) monly found in the literature [4,1619] and simplied for reasons
Hi isosteric heat of adsorption for component i to be discussed below.
(J mol1 )
w ratio of the internal surface area to the volume of 2.1. Model assumptions
the column wall (m1 )
wL ratio of the log mean surface to the volume of col- 1. Ideal gas behavior of the adsorptive throughout the column.
umn wall (m1 ) 2. No mass, heat or velocity gradients in the radial direction.
bed porosity 3. Plug ow with axial mass dispersion.
p particle porosity 4. Negligible external mass transfer effect (i.e. lm mass transfer
 fractional loading ( = qi /qmax,i ) and macropore diffusion).
i parameter for the temperature dependency of the 5. Mass transfer into the particle described according to the linear
Toth exponent driving force (LDF) model [7,16,19].
 uid density (kg m3 ) 6. No temperature gradient inside the particles and thermal equi-
ap particle average density (kg m3 ) librium between the gas and the adsorbent.
w wall density (kg m3 ) 7. The system is assumed to operate adiabatically.
i the coefcient for the temperature dependency of 8. Constant heat transfer coefcients.
the maximum capacity 9. Constant and homogeneous bed porosity along the bed length.
 M. Bastos-Neto et al. / Separation and Purication Technology 77 (2011) 251260 253

The adsorption equilibrium was described by the Toth model:

Ki Cg,i
qi = qmax,i (6)
1/ni
(1 + (Ki Cg,i )ni )

The temperature dependence of K, qmax,i and ni was evaluated

according to Do [21] as described below:
Q 1 1

i
Ki = K0,i exp (7)
R T T0
  T

qmax,i = qmax0,i exp i 1 (8)
T0
 T0

ni = n0,i + i 1 (9)
T
The axial mass dispersion coefcient was calculated as
described in Ref. [22] using the following correlation:

Dax = (0.45 + 0.55)Dm + 0.35rp u (10)

where Dm , the molecular diffusivity, was calculated for each exper-

imental run at feed temperature and pressure conditions using the
Fig. 1. Schematic representation of the xed bed.
ChapmanEnskog equation [23].
Heat capacities of H2 and CO were evaluated with a polyno-
10. No pressure drop was considered. mial function of the temperature [24] and the heat capacity of the
mixture was calculated as follows:
2.2. Model equations and parameters 
cpg = (cpg,i yi ) (11)
As no radial dispersion is considered and the volumetric ow i

rate remains constant along the bed (u/ t = 0), the mass balance The lm heat transfer coefcient between the gas and wall, hw ,
can be written as follows: was calculated according to the following equation:
 
Cg,i Cg,i 2 Cg,i (1 ) Cg,i q i Nu kg
+u Dax + p + ap =0 (1) hw = (12)
t z z 2 t t di

The ideal gas law is used to correlate the partial pressures to the The Nusselt number, Nu, was set to 5.77, which corresponds to
concentrations: plug ow with constant wall temperature in a circular tube [23] and
 the gas mixture thermal conductivity, kg , was calculated as reported
p=RT Cg,i (2) in Refs. [23,24] at feed conditions and assumed to be constant along
i the bed operation.
The following boundary and initial conditions were used:
Ignoring the external mass transfer effect and using the LDF
approximation for intra-pellet diffusion, the rate in which the aver- Dax Cg,i
Cg,i (z = 0, t) = + C0
age concentration changes is expressed as follows: u z
Cg,i (z > 0, t = 0) = 0
q i
= keff (qi q i ) (3) Cg,i
t =0
z z=L
The effective mass transfer coefcient, keff , was assumed to be
constant for a given operation and tted to each breakthrough
T (z = 0, t) = T0
curve.
The assumption of thermal equilibrium between gas and adsor- T (z > 0, t = 0) = T0
bent in each axial position is reported to be very reasonable [4].
Additionally, no dispersive energy transfer is taken into account. T
=0
Another simplication is that the adsorbed phase heat capacity is z z=L
negligible [16], as its magnitude is much lower than the sum of the The boundary and initial conditions dene an empty adsorber,
other heat capacities (uid phase, adsorbent and wall). Hence, we
have two energy balances: which is thermally equilibrated with the environment at the begin-
  ning, with no concentration and temperature gradients within the
 T
 1 T
bed outlet. The well known Dankwerts boundary condition was
(Cg,i cpg,i ) u + (Cg,i cvg,i ) + ap cs applied at the inlet of the adsorber to consider the effects of disper-
z t
i i sion on the concentration.
4hw 1
 qi
 The model parameters were analyzed for the relative sensitivity
+ (T Tw ) + ap Hi =0 (4) on the process dynamics and inuence on the simulation. Model
di t
i validation was accomplished by comparing simulations with the
and experimental data.

Tw
w cpw = w hw (T Tw ) wL Ug (Tw T ) (5) 3. Materials and methods
t
where w and wL are geometrical factors (see nomenclature) and Hydrogen (99.9992%) and carbon monoxide (99.997%), used in
the overall heat transfer coefcient, Ug , is set to zero (adiabatic this study, were supplied by Air Products (Germany). The used
system). adsorbents were three zeolites 5A, namely zeolite A and zeolite B
254 M. Bastos-Neto et al. / Separation and Purication Technology 77 (2011) 251260

Table 1 Adsorption isotherms were measured with a magnetic suspen-

Textural characteristics of the samples.
sion balance from Rubotherm (Germany) and evaluated according
Sample Specic surface Total pore volume Average pore to literature Refs. [20,21,2529].
area [m2 g1 ] [cm3 g1 ] size [nm]

AC D55/2 C PSA 765.4 0.414 0.54 4. Results and discussions

Zeolite A 669.9 0.274 0.41
Zeolite B 669.9 0.282 0.42
KSTROLITH 5ABF 680.0 0.279 0.41 In industrial separation units the adsorbent column is usually
lled with a carbon layer, the rst to be percolated by the mixture,
and a zeolite layer. Part of the CO is retained in the carbon bed and
the rest breaks through it to the zeolite bed. The information of
how much is adsorbed in each step is an important input for unit
operation as well as for simulation purposes.
Fig. 3 presents the CO isotherms at temperatures ranging from
279 to 313 K on the different materials. For a better comparison,
Fig. 4 was plotted with the isotherms of all samples at 293 K.
All isotherms show typical type I behavior and the experimen-
tal data were very well matched by the Toth model (Eq. (6)). The
parameters of the Toth isotherm model are summarized in Table 4
for each sample.
The adsorption enthalpy, necessary for the non-isothermal
description of the dynamic process, was obtained from the Toth
isotherm tting parameter Qi , which is equal to the isosteric heat
of adsorption, when the fractional loading ( = qi /qmax,i ) is zero
[21]:

Hi = Qi (13)
=0

Zeolites clearly show higher adsorption capacities in compar-

ison with the carbon sample for CO (Fig. 4), being the zeolite
KSTROLITH 5ABF slightly better at pressures above ca. 0.5 MPa.
Additionally, the isotherms of zeolites exhibit a more pronounced
Fig. 2. Illustration of the experimental system for the measurement of breakthrough non-linearity, which has a direct inuence on the complexity of a
curves. model to simulate adsorptive processes.
The curvature of isotherms plays an important role on the posi-
provided by Mahler AGS GmbH (Germany) and KSTROLITH 5ABF tion of the experimental breakthrough time. For linear isotherms
(binder-free) produced by CWK Chemiewerk Bad Kstritz GmbH and isothermal conditions the breakthrough time should be the
(Germany) and one activated carbon sample, AC D55/2 C PSA, man- same for different concentrations, because the adsorbed amount
ufactured by CarboTech AC GmbH (Germany), all of them assumed (loading) and the adsorbate concentration (or partial pressure)
to be suitable for H2 purication processes. Textural characteristics increase by the same factor (i.e. constant relationship). Non-linear
of the samples, evaluated through N2 isotherms at 77 K, are shown isotherms, like the ones described by Langmuir or Toth isotherm
in Table 1. models, may shift the breakthrough curves to shorter times and
The experimental setup for measuring breakthrough curves, the curves become steeper with increasing concentration, for an
represented in Fig. 2, consists of a stainless steel column, where isothermal case [19,30]. This inuence is qualitatively illustrated in
the adsorbent sample was packed; two mass ow controllers (max. Fig. 5 with a theoretical bed of zeolite sample KSTROLITH 5ABF,
ow of 5.0 cm3 s1 (SATP) and 1.7 cm3 s1 (SATP)); a pressure using the respective isotherm tted parameters as input on the
controller (max. 6.5 MPa) from Brooks Instrument (USA); and a proposed model and keeping all other parameters constant.
continuous gas analyzer with an Infra-Red detector from ABB From this point on, for the sake of a acceptable model validation,
Hartmann & Braun (Germany). two approaches are generally considered: (i) the non-linear region
The adsorbent bed was tested for the capacity and kinetics of of the isotherm, i.e. high partial pressures or concentrations, are
CO adsorption by measuring breakthrough curves at distinct pres- avoided in the study or (ii) more correlations and parameters are
sures and concentrations. Table 2 lists the column characteristics used, increasing the model complexity. On the one hand, operat-
and physical properties of the adsorbents and Table 3 shows the ing conditions of practical importance and use for the industry are
operating conditions of each experimental run for all adsorbent ignored, and on the other hand the strategy of using more empirical
samples. parameters or correlations could be questionable for their validity.
The regeneration conditions were 5 h at 423 K for the activated Fig. 6 presents the experimental breakthrough curves and the
carbon sample and about 10 h at 673 K for the zeolites, in both respective simulations for each sample at the operating conditions
cases with approximately 1.0 cm3 s1 (SATP) helium ow. Subse- listed in Table 3.
quently, the bed was purged with pure hydrogen for at least 30 min A general analysis of the simulations with the proposed model
to assure the absence of helium during the experiments. After the suggests that the predictions were very satisfactory. For most of the
regeneration step, cyclic experiments (24 cycles) were carried out cases, however, the breakthrough time is slightly underestimated.
for a given concentration and pressure at room temperature. Feed This underestimation is more pronounced for the experiments car-
concentrations of mixtures H2 /CO were set by adjusting the gas ried out at higher pressures and lower concentrations. This is the
ow controllers. Between each cycle, there was a desorption step case for the zeolites at conditions correspondent to the experimen-
to purge the bed with pure H2 at the same ow rate used in the tal runs A1, A2, B2, B3, E3, E4, F2, F3 and F4.
adsorption step, accomplished by closing the CO-stream valve after The main reason for this small difference on the prediction of
bed saturation. the breakthrough time is attributed to the assumption of an adia-
 M. Bastos-Neto et al. / Separation and Purication Technology 77 (2011) 251260 255

Table 2
Column and bed characteristics.

Adsorbent sample

AC D55/2 C PSA Zeolite A Zeolite B KSTROLITH 5ABF

(C)a (A)a (B)a (E)/(F)a

Form Cylindrical Spherical Spherical Spherical

Bed length [102 m] 18.2 18.0 18.5 18.2/8.0
Bed massb [103 kg] 37.03 50.25 50.70 47.35/21.4
Bed porosity 0.38 0.42 0.47 0.45/0.43
Particle porosity 0.44 0.41 0.41 0.41
Average particle radius [102 m] 0.15 0.1 0.1 0.1
Particle average density [kg m3 ] 783 1191 1301 1209
Particle specic heat [J kg1 K1 ] 820 920 920 920
Column internal diameter [102 m] 2.2
Wall thickness [103 m] 3
Wall density [kg m3 ] 7860
Wall specic heat [J kg1 K1 ] 477
a
Operating conditions.
b
Non-activated sample as supplied.

Table 3
Operating conditions of the breakthrough curves experiments and simulations.

Adsorbent: AC D55/2 C PSA Experimental run

C1 C2 C3

Feed temperature [K] 298 299 299

Feed pressure [MPa] 0.50 0.10 1.99
Feed molar fraction of CO [%] 6.6 13.1 13.1
Feed ow rate [cm3 s1 at SATP] 2.79 2.84 2.84

Adsorbent: zeolite A Experimental run

A1 A2 A3

Feed temperature [K] 296 296 295

Feed pressure [MPa] 2.02 3.98 5.97
Feed molar fraction of CO [%] 17.5 17.5 17.5
Feed ow rate [cm3 s1 at SATP] 2.99 2.98 2.99

Adsorbent: zeolite B Experimental run

B1 B2 B3 B4

Feed temperature [K] 296 298 293 297

Feed pressure [MPa] 0.10 2.00 3.98 5.98
Feed molar fraction of CO [%] 17.5 17.5 17.5 17.5
Feed ow rate [cm3 s1 at SATP] 2.99 2.99 2.99 2.98

Adsorbent: KSTROLITH 5ABF Experimental run

E1 E2 E3 E4

Feed temperature [K] 301 301 297 300

Feed pressure [MPa] 0.20 0.20 1.00 1.00
Feed molar fraction of CO [%] 5.0 10.0 5.0 10.0
Feed ow rate [cm3 s1 at SATP] 3.35 3.35 3.35 3.35

Adsorbent: KSTROLITH 5ABF Experimental run

F1 F2 F3 F4

Feed temperature [K] 299 300 300 300

Feed pressure [MPa] 0.10 1.99 1.99 3.97
Feed molar fraction of CO [%] 17.5 17.5 9.0 9.0
Feed ow rate [cm3 s1 at SATP] 2.99 2.99 2.86 2.86

Table 4
Fitting parameters of the Toth model for the adsorbent samples.

Adsorbent T0 [K] qmax0,i [mol kg1 ] K0,i [10 MPa1 ] n0,i i i Qi [J mol1 ] R2

AC D55/2 C PSA 298 6.272 0.112 0.662 0.000 0.443 14592 0.9998
Zeolite A 298 7.730 1.974 0.324 0.179 0.000 30463 0.9993
Zeolite B 298 5.815 1.330 0.415 0.177 0.000 31597 0.9995
KSTROLITH 5ABF 293 7.584 1.600 0.353 0.083 0.000 26676 0.9997
256 M. Bastos-Neto et al. / Separation and Purication Technology 77 (2011) 251260

a 3.0 b 3.5
KSTROLITH 5ABF 283 K
AC D 55/2 C PSA

adsorbed amount [mol kg ]

3.0
adsorbed amount [mol kg ]

293 K

-1
-1

2.5 283 K
293 K
298 K 2.5 313 K
2.0
313 K 2.0
1.5
1.5
1.0
1.0
0.5 experimental points 0.5 experimental points
fit with Toth model fit with Toth model
0.0 0.0
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
pressure [MPa] pressure [MPa]

c 3.5 d 3.5
zeolite A 278 K zeolite B 278 K

adsorbed amount [mol kg ]

adsorbed amount [mol kg ]

-1
3.0 3.0 293 K
-1

293 K
298 K 298 K
2.5 313 K 2.5 313 K

2.0 2.0

1.5 1.5

1.0 1.0
experimental points experimental points
0.5 fit with Toth model
0.5
fit with Toth model
0.0 0.0
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
pressure [MPa] pressure [MPa]

Fig. 3. Isotherms of carbon monoxide on the samples used in the present study.

batic operating condition on the model. Although real adsorptive to generate more heat, effect which is also proportional to the feed
processes operate near to an adiabatic condition [31], this study partial pressure. For an adiabatic operation, the system is submitted
made used of a lab-scale xed bed with dimensions small enough to a higher temperature increase, which poses a negative impact on
to allow a certain degree of heat transfer to the environment during the adsorption capacity, dropping the amount of gas (CO) that can
the experimental runs. be adsorbed, as one can discern from the isotherms in Fig. 3. Con-
Referring to Table 4, one can observe that the heat of adsorp- sequently, the bed gets saturated faster and CO breaks through the
tion is signicantly higher for the zeolites in comparison with the column earlier.
activated carbon sample. Therefore, the process with zeolites tends

1.0
3.5
CO Adsorption
relative concentration (C/C0)

1.0 MPa
3.0 T = 293 K 0.8
adsorbed amount [mol kg ]
-1

0.8 MPa
2.5
0.6 0.6 MPa
2.0 0.4 MPa
0.4
1.5 0.2 MPa

1.0 0.2
AC D 55/2 C PSA
KSTROLITH 5ABF
KSTROLITH 5ABF
0.5 zeolite A carbon monoxide
zeolite B
0.0
0.0 0 800 1600 2400 3200 4000
0.0 0.2 0.4 0.6 0.8 1.0
time [s]
pressure [MPa]
Fig. 5. Inuence of the isotherm curvature on the breakthrough time with simulated
Fig. 4. Comparison of the adsorption capacities of the samples at 293 K. curves for different feed partial pressures of carbon monoxide.
 M. Bastos-Neto et al. / Separation and Purication Technology 77 (2011) 251260 257

a 1.0
b
1.0
zeolite A
carbon monoxide

relative concentration (C/C0)

relative concentration (C/C0)

0.8
C2 0.8 A3
C1 A2
0.6 A1
C3 0.6

0.4
0.4
AC D 55/2 C PSA
0.2 carbon monoxide
0.2
model
model
0.0
0.0
0 800 1600 2400 3200 4000 4800
0 2000 4000 6000 8000 10000
time [s] time [s]

c
1.0 1.0
relative concentration (C/C0)

relative concentration (C/C0)

0.8 0.8
B3 E2 E4
B1
E1
E3
0.6 B4 0.6
B2

0.4 0.4

0.2 zeolite B KSTROLITH 5ABF

carbon monoxide
0.2
carbon monoxide
model model
0.0 0.0
0 2000 4000 6000 8000 10000 12000 0 2000 4000 6000 8000 10000 12000
time [s] time [s]

e 1.0
relative concentration (C/C0)

0.8 F3

F1
0.6 F2

F4

0.4

KSTROLITH 5ABF
0.2 carbon monoxide
model

0.0
0 1200 2400 3600 4800 6000 7200

time [s]

Fig. 6. Breakthrough curves of carbon monoxide on different samples. Points were experimentally obtained and solid lines are simulations with the proposed model.
Operating conditions for each curve are listed in Table 3.

In the case of the experiments with activated carbon (C1, C2 and carbon. For the operating condition F3 with the zeolite KSTROLITH
C3), heat should be generated in a lower rate. The same argument is 5ABF, which is even milder than C3 considering the bed height
also valid for the cases of B1, E1, E2 and F1, where the feed is much and feed concentration, the temperature increases 8.2 K. For the
diluted on CO. set of conditions F4 the temperature increase reaches a value of
To illustrate the heat effects on the temperature increase, Fig. 7 10.5 K, which is quite considerable to affect the bed adsorptive
was plotted with the simulated temperature proles at the col- capacity. As one can observe from the same plot, the simulated tem-
umn exit for some representative cases, namely C3, F3 and F4. The perature rises and remains practically constant since the system
temperature increases only 2.6 K for the run C3 with the activated is considered to operate adiabatically. Hence, the relative con-
258 M. Bastos-Neto et al. / Separation and Purication Technology 77 (2011) 251260

312
simulated temperature
310 adiabatic process
F4

308
exit temperature [K]

F3
306

304 8.2 K 10.5 K

302

300 2.6 K

298 C3

0 1200 2400 3600 4800 6000 7200 8400

time [s] Fig. 8. Trend of the effective mass transfer coefcient in relation to the isotherm
curvature along the total pressure.
Fig. 7. Comparison of simulated temperature proles at the column exit for an
adiabatic process.
Where the curvature is evaluated from Eq. (6) in terms of pres-
sure at nite loadings, according to Ref. [21]:

qi Ki qmax,i
isoterm curvature = = (14)
p (1/ni +1)
(1 + (Ki p)ni )
centration of the simulated breakthrough curves also reaches the
unity at the same rate that the temperature reaches the maximum. In Fig. 8, points represent the keff values, which were connected
Unlikely to the experiments, for which it was already observed that by means of an empiric exponential decay function of the type
heat is transferred to the environment, the temperature reaches a y = axb , and the lines correspond to the isotherm curvature for the
maximum and then starts to decrease and consequently the exper- zeolites. Although no direct match is found, the purpose of this
imental curves present a slow approach to unity. illustration is to observe the relationship between keff and qi / p
Other minor mismatches are also found in the mass transfer by the trend shown with the pressure.
zone (MTZ), more specically on the prediction of the upper region The neglect of the pressure drop also seems to be very rea-
of some breakthrough curves (e.g. A1, A2, B1, B2, B3, E3 and E4). sonable for the conditions studied. Despite the common use of an
This is likely to be caused by a tailing effect stemmed from the additional equation to describe the pressure drop along the bed,
heat of adsorption, which was also observed and discussed in more the relative small lengths of the adsorbers in the present study do
details in references [3234]. The deviation between simulation not seem to be susceptible to signicant speed variations due to
and experimental results may come from the isotherm variation by changes in the total pressure, which is supposed to be controlled
temperature change and the velocity gap between MTZ and tem- by an appropriate device.
perature proles. Additionally, changes on the temperature also If the system experiences pressure drops, it is expected that the
lead to changes in the kinetics and such effect was not considered ow rate becomes higher and, therefore, the component should
on the model. earlier breakthrough. No obvious correlation between bed length
The ow rate, inuenced by the operating pressure, also plays and the occurrence of considerable pressure drop could be observed
a role on the process dynamics. What one could also expect from in this work.
the isotherms in Fig. 3 is that the higher the pressure, the higher CO It is difcult to compare the experimental performance data
adsorption and accumulation in the adsorber. Consequently, there of H2 /CO separation between the samples studied here because
could be a misinterpretation of the graphs in Fig. 6. But what actu- of different operating conditions and system parameters. That the
ally happens is an indirect inuence of the pressure on the retention activated carbon presents a less efcient performance in relation
time, as the volume is compressed reducing the ow speed and to the zeolites is an expected fact, since its adsorption capacity for
increasing the contact time between gas and solid. CO is far lower. But in view of a cyclic process, the cyclic capacities
The mass ow controllers provide a volume ow under stan- of the samples are fairly comparable, depending on working pres-
dard conditions. Therefore an increasing of pressure by a known sure range, seeing that the shapes of the isotherms are practically
factor would reduce the ow, in an ideal case, also by the same fac- identical. The most important factor regarding cyclic capacities is
tor. As a consequence, the contact time of solid and gas phases is the slope of the isotherm within the working pressure range.
reduced with increasing pressure. Neglecting the slight differences A comparison between the zeolites only is even harder, never-
on the temperature and ow rate, the experiments with zeolite A theless we attempted an evaluation of the performance based on
and zeolite B illustrate the problem. the experimental stoichiometric time corrected with respect to the
The mass transport coefcient affects the steepness and the size mass of sample. Similar sets of conditions are desirable for such
of the mass transfer zone of a breakthrough curve [7,19]. A faster procedure and therefore we selected A1 for the zeolite A, B2 for the
mass transport (i.e. kinetics) is highly desired for a more efcient zeolite B and F2 for the KSTROLITH 5ABF, which correspond to ca.
use of the adsorbent bed. Regardless of our assumption of constant 2 MPa, ca. 298 K and 17.5% CO in the feed. The time (in s) divided by
behavior, this parameter is actually related to the material trans- the sample mass (in g) give us 98.07 for zeolite A, 98.00 for zeolite
port into the pores of the adsorbent, which by itself depends on B and 98.00 for the KSTROLITH 5ABF. These results are identical,
the concentration gradient, temperature and pressure, and is asso- considering the temperature difference between each set of condi-
ciated to the curvature of the correspondent isotherm [19]. This tions and experimental deviations, and this is a clear indication that
association is illustrated in Fig. 8 with the plot of the tted values the key for optimizing the performance of H2 /CO separation lies in
of keff , which are listed in Table 5, along with the isotherm curvature the adsorption equilibrium, which was presented by the isotherms
against the total pressure. in advance.
 M. Bastos-Neto et al. / Separation and Purication Technology 77 (2011) 251260 259

Table 5
Fitted values of the effective mass transport coefcient.

Adsorbent: AC D55/2 C PSA Experimental run

C1 C2 C3

keff [s1 ] 1.07 2.67 2.67 102

Adsorbent: zeolite A Experimental run

A1 A2 A3
1 2 3
keff [s ] 1.50 10 6.25 10 3.25 103

Adsorbent: zeolite B Experimental run

B1 B2 B3 B4

keff [s1 ] 1.75 101 1.50 102 6.75 103 3.75 103

Adsorbent: KSTROLITH 5ABF Experimental run

E1 E2 E3 E4
1 2 1 2
keff [s ] 7.50 10 1.25 10 2.50 10 2.50 102

Adsorbent: KSTROLITH 5ABF Experimental run

F1 F2 F3 F4

keff [s1 ] 2.50 101 1.38 102 1.63 102 5.75 103

5. Conclusion Cientco e Tecnolgico - Brazil), DAAD (Deutscher Akademis-

cher Austausch Dienst - Germany), CarboTech AC GmbH, Mahler
Breakthrough curves of carbon monoxide were measured on AGS GmbH and CWK Chemiewerk Bad Kstritz GmbH. The
three different 5A zeolites and one activated carbon at different authors also acknowledge aid of Ulf Roland from the Zen-
operating conditions. A simple mathematical model was used to trum fr Umweltforschung (UFZ Leipzig, Germany) with sample
simulate the dynamic behavior of the adsorbent bed using inputs characterization.
of previously measured adsorption isotherms, heat of adsorption
and physical properties evaluated according to literature reports.
Despite the simplications, the model was capable of simulat- References
ing the breakthrough curves of carbon monoxide with satisfactory
agreement with experimental data. In this study, it was observed [1] D. Simbeck, E. Chang, Hydrogen Supply: Cost Estimate for Hydrogen Pathways
Scoping Analysis, Subcontractor Report, Mountain View, California, 2002,
that the heat effects also play an important role on the evaluation http://www.nrel.gov/docs/fy03osti/32525.pdf (online address).
of the breakthrough time. Additionally, the assumption of adiabatic [2] J. Benemann, Hydrogen biotechnology: progress and prospects, Nat. Biotech-
operation appears not to be completely valid, as the simulations of nol. 14 (1996) 11011103.
[3] D. Das, T.N. Veziroglu, Hydrogen production by biological processes: a survey
most conditions slightly underestimate the breakthrough time. of literature, Int. J. Hydrogen Energy 26 (2001) 1328.
The proposed model may be further developed to take into [4] A.M. Ribeiro, C.A. Grande, F.V.S. Lopes, J.M. Loureiro, A.E. Rodrigues, A para-
account non-constant ow rates through the bed, pressure drop metric study of layered bed PSA for hydrogen purication, Chem. Eng. Sci. 63
(2008) 52585273.
along the column, temperature dependence of the effective mass
[5] R.T. Yang, Gas Separation by Adsorption Processes, Imperial College Press, Lon-
transport coefcient in the future. However, it is important to keep don, 1997.
in mind that the use of a more sophisticated model requires the [6] S. Sircar, T.C. Golden, Purication of hydrogen by pressure swing adsorption,
knowledge of more parameters as well as their inuence on the Sep. Sci. Technol. 35 (5) (2000) 667687.
[7] D.M. Ruthven, Principles of Adsorption and Adsorption Processes, Wiley, New
dynamics of adsorptive processes. York, 1984.
Ideally, every model input should be experimentally attained. [8] J.H. Park, J.N. Kim, S.H. Cho, J.D. Kim, R.T. Yang, Adsorber dynamics and optimal
But due to experimental limitations such as time, nancial aspects design of layered beds for multicomponent gas adsorption, Chem. Eng. Sci. 53
(23) (1998) 39513963.
or infrastructure, if only some of the model parameters are available [9] M. Chlendi, D. Tondeur, Dynamic behaviour of layered columns in pressure
from experimental data, one should nd other ways to obtain them. swing adsorption, Gas Sep. Purif. 9 (4) (1995) 231242.
That is typically achieved by using correlations but whether they [10] S. Cavenati, C.A. Grande, A.E. Rodrigues, Separation of CH4 /CO2 /N2 mixtures by
layered pressure swing adsorption for upgrade of natural gas, Chem. Eng. Sci.
can be properly applied or not is a topic for further discussion. 61 (2006) 38933906.
An important conclusion of this work is that simplied mod- [11] S. Sircar, Basic research needs for design of adsorptive gas separation processes,
els are capable of properly simulating breakthrough curves under Ind. Eng. Chem. Res. 45 (2006) 54355448.
[12] Mahler AGS GmbH Internal Note, 2008.
the studied conditions, provided that enough experimental data
[13] Chemiewerk Bad Kstritz (CWK) Internal Note, 2008.
are available. The indiscriminate use of parameters can lead to [14] X. Cheng, Z. Shi, N. Glass, L. Zhang, J. Zhang, D. Song, Z.S. Liu, H. Wang, J. Shen,
an unnecessary model complexity of questionable practical value. A review of PEM hydrogen fuel cell contamination: impacts, mechanisms, and
mitigation, J. Power Sources 165 (2) (2007) 739756.
Simple models may, however, not properly work for systems with
[15] C. Tien, Adsorption Calculations and Modeling, Butterworth-Heinemann,
high adsorption and heat effects and considerable derivations from Boston, 1994.
ideal gas behavior like CO2 rich mixtures. [16] D. Bathen, M. Breitbach, Adsorptionstechnik, Springer, Berlin, 2001.
[17] J.A. Delgado, M.A. Uguina, J.R. Sotelo, B. Ruz, J.M. Gmez, Fixed-bed adsorption
of carbon dioxide/methane mixtures on silicalite pellets, Adsorption 12 (2006)
Acknowledgements 518.
[18] S. Karcher, Eignung verschiedener Sorbentien zur Entfernung von Reaktivfarb-
stoffen aus Abwasser, Dissertation, Technische Universitt Berlin, Berlin, 2000
The authors acknowledge the support from INNOWATT - Project (in German).
Az: IW070135, CNPq (Conselho Nacional de Desenvolvimento [19] W. Kast, Adsorption aus der Gasphase, VCH, Weinheim, 1988.
260 M. Bastos-Neto et al. / Separation and Purication Technology 77 (2011) 251260

[20] M. Bastos-Neto, A.E.B. Torres, D.C.S. Azevedo, C.L. Cavalcante Jr., Methane [28] R.E. Bazan, M. Bastos-Neto, R. Staudt, H. Papp, D.C.S. Azevedo, C.L. Cavalcante
adsorption storage using microporous carbons obtained from coconut shells, Jr., Adsorption equilibria of natural gas components on activated carbon: pure
Adsorption 11 (2005) 911915. and mixed gas isotherms, Adsorpt. Sci. Technol. 26 (5) (2008) 323332.
[21] D.D. Do, Adsorption Analysis: Equilibria, Kinetics, Series on Chemical Engineer- [29] R. Staudt, S. Bohn, F. Dreisbach, J.U. Keller, Gravimetric and volumetric mea-
ing, vol. 2, Imperial College Press, London, 1998. surements of helium adsorption equilibria on different porous solids, in: B.
[22] F.V.S. Lopes, C.A. Grande, A.M. Ribeiro, J.M. Loureiro, O. Evaggelos, V. Nikolakis, McEnaney, et al. (Eds.), Characterization of Porous Solids, vol. 4, Royal Society
A.E. Rodrigues, Adsorption of H2 , CO2 , CH4 , CO, N2 and H2 O in activated carbon of Chemistry, London, 1997, pp. 261266.
and zeolite for hydrogen production, Sep. Sci. Technol. 44 (2009) 10451073. [30] M. Bastos-Neto, A. Mller, R. Staudt, J. Bhm, R. Glser, Breakthrough curves of
[23] R.B. Bird, W.E. Stewart, E.N. Lightfoot, Transport Phenomena, revised second methane at high pressures for H2 purication processes, Chem. Ing. Tech., 83
ed., Wiley International, New York, 2006. (12), doi:10.1002/cite.201000155.
[24] R.H. Perry, D.W. Green, J.O. Maloney (Eds.), Perrys Chemical Engineers Hand- [31] R.T. Yang, P.L. Cen, Improved pressure swing adsorption processes for gas sep-
book, 7th ed., McGraw-Hill, New York, 1999. aration: by heat exchange between adsorbers and by using high-heat-capacity
[25] F. Dreisbach, H.W. Lsch, P. Harting, Highest pressure adsorption equilibria inert additives, Ind. Eng. Chem. Proc. Des. Dev. 25 (1) (1986) 5459.
data: measurement with magnetic suspension balance and analysis with a new [32] J.Y. Yang, C.H. Lee, Adsorption dynamics of a layered bed PSA for H-2 recovery
adsorbent/adsorbate-volume, Adsorption 8 (2) (2002) 95109. from coke oven gas, AIChE J. 44 (1998) 13251334.
[26] A. Herbst, P. Harting, Thermodynamic description of excess isotherms in high- [33] J.G. Jee, M.B. Kim, C.H. Lee, Adsorption characteristics of hydrogen mixtures in
pressure adsorption of methane, argon and nitrogen, Adsorption 8 (2002) a layered bed: binary, ternary, and ve-component mixtures, Ind. Eng. Chem.
111123. Res. 40 (2001) 868878.
[27] F. Dreisbach, R. Staudt, J.U. Keller, High pressure adsorption data of methane, [34] J.J. Lee, M.K. Kim, D.G. Lee, H. Ahn, M.J. Kim, C.H. Lee, Heat-exchange pres-
nitrogen, carbon dioxide and their binary and ternary mixtures on activated sure swing adsorption process for hydrogen separation, AIChE J. 54 (2008)
carbon, Adsorption 5 (1999) 215227. 20542064.