1098 Ind. Eng. Chem. Res.

2002, 41, 1098-1106

Identification of Transport Mechanism in Adsorbent Micropores

from Column Dynamics
S. Farooq,* Huang Qinglin, and I. A. Karimi
Department of Chemical and Environmental Engineering, The National University of Singapore,
4 Engineering Drive 4, Singapore 117576

The possibility of distinguishing between barrier and pore diffusional resistances in adsorbent
micropores from dynamic column response has been explored. The modeling of transport through
a barrier resistance confined at a micropore mouth is mathematically equivalent to the linear
driving force (LDF) model. By comparing the pore diffusion and LDF models for dynamic column
response for a linear isotherm, it is shown that the extent of agreement between the two models
depends on the product of the two parameters R [) (1 - p)Kc/p] and γ () DcL/rc2vo). By an
appropriate choice of operating conditions, it is indeed possible to clearly distinguish between
the two mechanisms. The proposed criteria have been verified experimentally for the break-
through of oxygen and nitrogen in a 4A zeolite and two carbon molecular sieve samples.

Introduction Chemical Company calls its product molecular sieving

carbon (MSC). In this paper, products from both manu-
It is commonly assumed that, unlike uptake in an facturers will be termed CMS.
adsorbent particle, the breakthrough dynamics obtained
CMSs have a bidisperse pore structure with clearly
using the linear driving force (LDF) approximation with
distinguishable macropore and micropore resistances to
the LDF rate constant k, given by eq 1, is practically
the transport of sorbates. However, the controlling
indistinguishable from that obtained using the full pore
resistance for the uptake of sorbates is typically diffu-
diffusion model with constant diffusivity D.
sion in the micropores. The design of a kinetically
controlled process using CMSs is entirely dependent on
D
k ) 15 2 (1) the fundamental understanding of the diffusion of gases
Rp in the micropores of the adsorbent. The transport
mechanism in CMS micropores is still not fully under-
Equation 1 is commonly referred to as the Glueckauf1 stood. Distributed pore diffusional resistance5-9 or bar-
approximation. Through a moment analysis of the pulse rier resistance confined at the pore mouth10-13 or a
response from a chromatographic column model that combination of two11,14 might be encountered.
explicitly allowed for linear adsorption equilibrium and The modeling of transport through a barrier resist-
external film, macropore, and micropore resistances, ance confined at a micropore mouth is mathematically
Haynes and Sharma2 showed that equivalent to the linear driving force (LDF) model.15 The
wide success of the Gleuckauf approximation in column
Rp2K rc2 dynamics calculations has given rise to a misconception
1 RpK
) + + (2) that the details of the transport mechanism in the
k 3kf 15pDp 15Dc micropores of an adsorbent in which the controlling step
is uptake in the micropores are not captured in the
Equation 2 is actually an extension of the Glueckauf breakthrough response from a column. Although it has
approximation for systems in which more than one mass been pointed out in some studies that the Glueckauf
transfer resistance is significant.3 Although eq 2 is approximation breaks down for short columns or short
strictly valid for a linear isothermal system, it is also contact times,16,17 the possibility of using this observa-
known to work reasonably well for nonlinear systems tion as an alternative route for understanding the
with K replaced by qo/Co, where Co is the feed concen- transport mechanism in the micropores has not been
tration of the adsorbate in the gas phase and qo is the explored.
corresponding equilibrium in the adsorbed phase.4 In this study, it is shown that the extent of agree-
The separation of gases and vapors by adsorption can ment between the pore diffusion and barrier models
be broadly classified into two categories, namely, equi- depends on the product of two parameters, namely, R
librium-controlled and kinetically controlled. Most of the [) (1 - p)Kc/p] and γ () DcL/rc2vo). Through the
known kinetically selective processes use carbon mo- appropriate choice of operating conditions, it is possible
lecular sieves (CMSs), although such processes based to distinguish between the two transport mechanisms
on zeolite adsorbents are also known. Takeda Chemical in CMS micropores from breakthrough experiments.
Company of Japan and Bergbau Forschung (BF) of The application of this method seems to indicate that
Germany are the leading manufacturers of CMSs used the transport of oxygen and nitrogen in the micropores
all over the world. In the product literature, Takeda of both BF and Takeda samples is controlled by barrier
resistance confined at the pore mouth. The break-
* Author to whom correspondence should be addressed. through of oxygen and nitrogen was also studied in a
Tel.: (65) 874 6545. Fax: (65) 779 1936. E-mail: chesf@ 4A zeolite sample as a reference point for unambiguous
nus.edu.sg. pore diffusional behavior.
10.1021/ie0104621 CCC: $22.00 © 2002 American Chemical Society
Published on Web 10/04/2001
 Ind. Eng. Chem. Res., Vol. 41, No. 5, 2002 1099

Table 1. Equations Describing the Two Models

distributed pore diffusional resistance barrier resistance at the pore mouth
fluid-phase material balance and boundary conditions

h
∂C
)
1 ∂2C
∂τ Pe ∂χ2
h ∂C
-
h
∂χ
- 3ψβ
∂Chp
∂η | η)1

h
∂C
∂χ| χ)0
) -Pe(C
h |χ)0- - C|χ)0+);
h
∂C
∂χ χ)1 |
)0

macropore material balance and boundary conditions macropore material balance and boundary conditions

∂Chp
∂τ
)β
∂2C
∂η2
(
h p 2 ∂C
+
hp
η ∂η
- 3Rγ
∂Y
∂ξ ) |ξ)1
∂Chp
∂τ
)β
∂2C
h p 2 ∂C
∂η2
+ (
hp
η ∂η
- 15Rγ(Y
h* - Y
h) )
∂Chp
∂η |
η)0
) 0;
∂C
∂η|
hp
η)1
) ∆(C
h -C
h p|η)1)
∂Chp
∂η | η)0
) 0;
∂Chp
∂η | η)1
) ∆(C
h -C
h p|η)1)

micropore material balance and boundary conditions micropore material balance

∂Y
∂τ
)γ + (
∂2Y 2 ∂Y
∂ξ2 ξ ∂ξ ) h
∂Y
∂τ
) 15γ(Y
h* - Y
h)

∂Y
∂ξ | ξ)0
) 0; Y|ξ)1 ) C
hp

Model for Predicting Column Dynamics summarized in Table 1. The equations contain the
following dimensionless variables
Assumptions. The mathematical model for the break-
through of an adsorbate in a bed that is packed with a C Cp q j
q z
bidisperse adsorbent and subjected to a change (step or h )
C hp ) , Y ) , Y
, C h ) , χ ) z/L,
Co Co qo qo L
pulse) in the feed concentration of the adsorbate has
three main components: (i) adsorbate material balance R r
η ) , and ξ )
for the fluid phase, (ii) adsorbate material balance in Rp rc
the macropores, and (iii) adsorbate material balance in
the micropores. where C is the adsorbate concentration in the fluid
Transport in the macropores is represented by mo- phase, Cp is the adsorbate concentration in the macro-
lecular-diffusion-controlled pore diffusion, where the pore, Co is the adsorbate concentration in the feed, q is
macropore diffusivity is given by the bulk molecular the adsorbate concentration in the micropore, q j is the
diffusivity corrected for pore tortuosity. Both possibili- average adsorbate concentration in the micropore, qo
ties of barrier resistance confined at the pore mouth and () KcCo) is the adsorbate concentration in the micropore
pore diffusional resistance distributed in the micropore in equilibrium with Co, z is the distance along the bed
interior are considered for micropore transport in this length, R is the distance along the macroparticle radius,
comparative study. The steady-state assumption at the r is the distance along the microparticle radius, L is the
fluid-solid interface forms the link between the fluid- bed length, Rp is the macroparticle radius, and rc is the
and solid-phase balance equations. In other words, the microparticle radius. In addition, the equations also
diffusive flux of the sorbate across the external gas film contain the following dimensionless groups
around the adsorbent particle is assumed to be equal
to that at the macropore surface. In most gas adsorption 1 - p Dp L Dc L kb L
studies, intraparticle transport of the adsorbate is the
R) Kc, β ) 2 , γ ) 2 or ,
p R vo r vo 15vo p c
slower step, and it is reasonable to assume negligible
gas-side resistance. Nevertheless, in this study, the k fR p 1- voL
∆) , ψ) p, and Pe )
external film resistance is included in the model for pDp  DL
completeness. However, it is assumed that the micro-
particle surface is in equilibrium with the macropore In the above dimensionless groups,  is the bed voidage,
gas. The analysis is restricted to a linear isotherm. It p is the particle voidage, Kc is the dimensionless
is further assumed that a single adsorbate is fed to the Henry’s constant based on the micropore volume, vo is
bed in a stream of inert carrier. Moreover, the stream the interstitial fluid velocity in the bed, kf is the external
is dilute in adsorbate concentration so that the system film mass transfer coefficient, kb is the barrier mass
is isothermal and the bulk velocity remains practically transfer coefficient, Dc is the micropore diffusivity, DL
unchanged along the column length. An axially dis- is the axial dispersion coefficient, and Dp is the macropore
persed plug-flow pattern is assumed for the fluid phase. diffusivity. According to the assumption made about
Frictional pressure drop is neglected, and the column macropore diffusion, Dp is the bulk molecular diffusivity
is assumed to operate at constant pressure. (Dm) corrected for particle tortuosity (τp). L, Rp, and rc
Model Equations. Subject to the assumptions dis- have already been introduced.
cussed in the preceding section, the dimensionless forms Numerical Solution. The model equations were
of the model equations and the boundary conditions are solved numerically by the method of orthogonal collo-
1100 Ind. Eng. Chem. Res., Vol. 41, No. 5, 2002

Figure 1. Comparison between the breakthrough responses of

the two transport models. The shaded area represents I defined
by eq 3. R ) 100, β ) 250, γ ) 0.0005, Pe ) 250, Ψ ) 0.5, ∆ ) 10.

cation. The partial differential equations were dis-

cretized in space, which converted them into a system
of ordinary differential equations in the time domain.
Eleven collocation points were used along the bed
length, and seven points were used along the macro-
particle/microparticle radii. The ordinary differential
equations were integrated in the time domain using
Gear’s variable-step integration algorithm as provided
in the FORSIM package.18
Although an analytical solution is available for the
model that considers distributed pore diffusional resist-
ance,19 numerical evaluation of the associated slowly
converging complicated functions did not appear to be
advantageous over the numerical solution adopted here.
In fact, it was much simpler to modify similar routines
for column dynamics calculations with nonlinear iso-
therms that were developed earlier in this laboratory.20
Comparative Study. R, β, γ, ∆, ψ, and Pe are the
six dimensionless groups that constitute the inputs for
the two models. These parameters were systematically
varied to study their effects on the difference in the
solutions of the two models when kb ) 15Dc/rc2. The
shaded area between the solutions of the two models
shown in Figure 1 was normalized with respect to R and
taken as a measure of the difference. The shaded area, Figure 2. (a) Plot of I vs γ showing dependence on R and
I, is given by the integral negligible effects of β and Peclet number; (b) plot of E vs γ showing
that the peaks are separated by the ratio of their R values; (c)

( )
∫0∞| CCo|χ)1 pore - ( | ) |
plot of E vs Rγ showing that all of the results converge to a single
C
I) dτ (3) curve.
Co χ)1 barrier
Results and Discussion
Note that the integrand is taken as an absolute value. The results of the simulations are summarized in
The trapezoidal rule with a dimensionless time interval Figure 2a. From Figure 2a, it is clear that the difference
of 1 was used to calculate the area. between the two models, denoted by I according to eq
Range of Investigation. R contains the equilibrium 3, depends on R and γ and is practically insensitive to
constant Kc and was varied from 10 to 100. β and Pe the parameters β and Pe. The area difference goes
are measures of the significance of macropore resistance through a maximum with increasing γ and practically
and axial dispersion, respectively. β was varied from 10 vanishes at the two ends of the range covered. For the
to 500, and Pe from 50 to 500 to cover the range typically same γ value, the area difference between the two
encountered in a system controlled by transport in the models increases with increasing R, i.e., with increasing
micropores. Values greater than 500 for those two adsorption capacity.
parameters are not significant in micropore-controlled Figure 2a apparently carries the impression that, for
systems. ∆ is a measure of the ratio of the external film comparable γ values, the difference will be more clearly
transport coefficient to the macropore transport coef- identifiable for a more strongly adsorbed system. On
ficient. In the limit of no flow (i.e., Sherwood number, closer examination, it was found that the increase in
kf2Rp/Dm ) 2), ∆ reduces to τp/p, which is typically taken the integral defined by eq 3 with increasing R was due
as 10. This value is already too high to have any effect to longer breakthrough times coming from increased
on the solution. Typical values of  and p are 0.4 and column capacities. It was, therefore, concluded that a
0.33, respectively, which makes ψ ≈ 0.5. Hence, values more meaningful measure of the difference between the
of ∆ and ψ were fixed at 10 and 0.5, respectively. γ is two models would be the area, I, normalized with
the most crucial parameter, and it was varied in this respect to R. We denote this normalized area as E ()
comparative study from 10-6 to 1. I/R). Plots of E vs R are shown in Figure 2b.
 Ind. Eng. Chem. Res., Vol. 41, No. 5, 2002 1101

Table 2. Summary of the Experimental Runs

operating column interstitial operating
pressure, P feed length, L velocity, v0 temperature, T
run (bar) composition (cm) (cm/s) (°C) R γ Rγ Pe β ψ ∆
RS-10 zeolitea
1 1.4 4.2% O2 60 9.4 -10 20.4 0.04 0.82 731.5 261.4
2 1.4 4.2% O2 60 10.1 10 17.5 0.055 0.96 716.1 275.8
3 1.4 3.9% O2 60 5.1 26 13.3 0.14 1.86 485.4 600.9
0.65 10
4 1 4.8% N2 60 1.8 -10 68 0.0013 0.09 221.0 2531.1
5 1 4.8% N2 60 1.9 10 41.9 0.0038 0.16 208.8 2712.0
6 1 4.8% N2 60 2 26 31.5 0.0067 0.21 202.6 2812.9
BF CMSb
7 1.4 2.7% O2 40 8.3 -30 82.6 0.0039 0.32 246.8 25.8
8 1.4 2.8% O2 40 4.7 1 40.4 0.015 0.61 201.0 55.5
9 1.4 3.1% O2 40 5.1 29 19.3 0.039 0.75 195.3 60.2
0.8 10
10 1.4 2.9% N2 40 0.9 1 34.7 0.004 0.14 85.6 271.6
11 1.4 3.0% N2 40 1 29 17.9 0.014 0.25 82.3 286.6
12 1.4 4.7% N2 40 1.2 70 14 0.039 0.55 80.3 296.1
Takeda CMSb
13 1.4 2.7% O2 40 8.3 -30 98.1 0.0044 0.43 246.8 25.8
14 1.2 3.0% O2 40 4.7 1 48.7 0.021 1.02 190.0 75.6
15 1.2 3.0% O2 40 5.1 29 25.4 0.05 1.27 184.1 82.0
0.61 10
16 1.4 2.8% N2 40 1.9 1 45.3 0.0022 0.10 138.0 128.6
17 1.4 3.1% N2 40 2 29 22.4 0.0066 0.15 129.9 143.3
18 1 4.7% N2 80 1.8 70 13.5 0.064 0.86 169.0 773.9
a For RS-10, Rp ) 0.06 cm, p ) 0.29 (Farooq22). b For CMSs, Rp ) 0.15 cm, p ) 0.33 (assumed).

Figure 3. Comparison of experimental and theoretical breakthrough of oxygen in RS-10 zeolite. See Table 2 for experimental conditions
and values of various parameters.

The normalized plots in Figure 2b revealed an inter- to the two possible transport mechanisms in the ad-
esting feature in that they were identical in shape but sorbent micropores. The level of difference that is
shifted from one another by the ratio of their R values. visually distinguishable will be discussed in the next
This clue paved the way for further improvement. As section, where the experimental results are analyzed.
can be seen from Figure 2c, all of the results converged At this point, it is important to recall that the
to one curve when E was plotted as a function of the comparison made in this study is based on kb ) 15Dc/
product of R and γ. Figure 2c provides a measure of the rc2. Perhaps more important is the fact that, when the
expected difference in the dynamic column response due two models differ, one cannot be made to fit the other
1102 Ind. Eng. Chem. Res., Vol. 41, No. 5, 2002

even if the micropore parameter is varied. In other

words, an unambiguous distinction between the two
transport mechanisms is possible throuh the appropri-
ate choice of operating conditions.
Significance of rγ. The concentration wave velocity
is given by

vo
ωc )
1-
1+ Kp

where Kp ) p + (1 - p)Kc ) p(1 + R). The following
alternative expression can be obtained for γ by substi- Figure 4. Comparison of experimental and theoretical break-
tuting the interstitial fluid velocity in the column, vo, through of nitrogen in RS-10 zeolite showing the fit of the pore
with an equivalent expression in terms of concentration model. For experimental conditions and values of various param-
wave velocity, ωc eters, see Table 2.

[ ]
Dc L 1
γ)
r 2 ωc 1-
c 1+  (1 + R)
 p
or

[ ]
Dc L R
Rγ ) 2 ω
rc c 1 + 1 - 
 (1 + R)
 p
The term in the square brackets on the right-hand side
of the above equation varies from 1.55 to 1.96. Hence,
it is not unreasonable to assume that this quantity is
approximately constant. One can then take Rγ ∝ (Dc/
rc2)(L/ωc), which can be viewed as the dimensionless
residence time of an adsorbing molecule in the bed.
Experimental Verification. Several breakthrough
experiments of oxygen and nitrogen in one zeolite
(modified 4A called RS-10) and two CMS (BF and
Takeda) samples were conducted to verify the theoreti-
cal results presented in Figure 2c. These experiments
are summarized in Table 2. In all runs, the column
pressure was nearly atmospheric, and the concentration
of adsorbate in the feed was kept low (<5%) to ensure
operation in the linear range. The experimental column
was kept in a heavily insulated constant-temperature
bath, which was controlled with a Lauda thermostat
(model RK 8 KS). The working temperature range of
the thermostat was from -45 to 150 °C, and it had a
cooling power of 120 W at -40 °C. It is the cooling power
at the desired low temperature that determines the
stability of low-temperature operation. Methanol was
the cooling medium during subzero operations.
The equilibrium capacity based on the micropore
volume, Kc, is related to that based on the particle
Figure 5. Comparison of experimental and theoretical break-
volume, Kp, by the equation through of oxygen in the BF CMS. The experimental conditions
and values of various parameters are given in Table 2.
Kp ) p + (1 - p)Kc (4)
estimated from the Chapman-Enskog correlation.21
Material balance of a transient adsorption column that Similarly, the axial dispersion coefficient, DL, was
has been saturated starting from clean initial conditions estimated using the equation3
yields

(
∫0∞ 1 - CCo ) dt )
L
(
vo
1+
1-

Kp ) (5)
DL ) 0.7Dm + 0.5vo(2Rp)

Hence, the values of R, β, and Pe were known for each

(6)

The left-hand side of eq 5 was obtained from the run. It is important to recall that the parameters β and
experimental breakthrough curves. The parameters L, Pe had negligible effects on the model solution in the
vo, and  were directly measured. A typical value of 0.33 micropore-controlled regime covered in the experimental
was assumed for p. The molecular diffusivity, Dm, was runs. Dc/rc2 or kb (in the parameter γ) was, therefore,
 Ind. Eng. Chem. Res., Vol. 41, No. 5, 2002 1103

Figure 6. Comparison of experimental and theoretical breakthrough of nitrogen in the BF CMS. For experimental conditions and values
of various parameters, see Table 2.

the only unknown that was used as the fitting param-

eter to match the simulations with the experimental
runs.
The breakthroughs of oxygen and nitrogen in a bed
packed with RS-10 are shown in Figures 3 and 4,
respectively, for three different γ values. In these runs,
the γ value was controlled by changing the temperature.
The micropore transport parameter (and hence γ)
decreased, but the equilibrium constant (and hence R)
increased with decreasing temperature. For oxygen,
even at -10 °C, Rγ was in the range where the two
models just begin to become distinguishable. A close look
at the magnified inset in Figure 3 reveals that the initial
breakthrough is better represented by the pore model.
On the other hand, Rγ values for the nitrogen runs at
all three temperatures were in the clearly distinguish-
able range. Whereas the pore model gives excellent fit
of the experimental data, the lack of fit of the barrier
model is evident. The difference is very pronounced in
the early part and does not seem to improve by adjust-
ing kb.
Having established that the proposed approach to
distinguish between the two possible transport mech-
anisms in adsorbent micropores yields expected results
for a zeolite adsorbent, let us now focus our attention
on the results obtained for the BF and Takeda CMS
samples presented in Figures 5-8. In contrast to the
results obtained for RS-10, it is clear from Figures 6
and 8 that, at both 1 and 29 °C, the nitrogen runs are
in very good agreement with the barrier model. In fact,
the Rγ values for these runs fall in the range where Figure 7. Comparison of experimental and theoretical break-
large differences are expected between the two models. through of oxygen in the Takeda CMS. The experimental condi-
Here, the pore model completely fails to fit these tions and values of various parameters are given in Table 2.
1104 Ind. Eng. Chem. Res., Vol. 41, No. 5, 2002

Figure 8. Comparison of experimental and theoretical breakthrough of nitrogen in the Takeda CMS. For experimental conditions and
values of various parameters, see Table 2.

experimental results for a significant part of the break- the same sorbates using the volumetric method. The
through. The results of runs conducted at 70 °C are also manufacturers of the samples used in these studies
shown in these figures. Faster transport in the micro- were, however, not disclosed. In these studies, the fit of
pores at high temperature pushed the Rγ value into the the barrier resistance model was very clear. The diffu-
region where the two models are supposed to be very sion of oxygen and nitrogen was also measured by
close according to Figure 2. Indeed, the results obtained Loughlin et al.14 in the BF CMS using volumetric,
are as expected. gravimetric, and pulse chromatographic methods and
The oxygen breakthrough results in the BF and by Rynders et al.13 using the isotope exchange technique
Takeda CMSs and the corresponding model fits are in the Takeda CMS. Whereas the former study advo-
presented in Figures 5 and 7. The fact that the two cated a dual resistance model, barrier model fitted the
models fit the experimental results equally well at 1 and data presented in the latter study. Liu and Ruthven12
29 °C is consistent with the expectations according to gravimetrically measured the diffusion of oxygen, ni-
the Rγ values for these runs. Some extreme measures trogen, argon, methane, and carbon dioxide in the BF
were necessary to reach the region where the transport CMS. They found that, for oxygen, nitrogen, and argon,
mechanism would be clearly distinguishable. These the data were consistent with distributed pore diffu-
extreme measures included lowering the operating sional resistance. The data for methane were consistent
temperature to -30 °C and increasing the operating with barrier resistance, and carbon dioxide showed a
velocity to a few times its original value. These results transition from barrier resistance control at lower
are also shown in Figures 5 and 7. Although the temperatures to diffusion control at higher tempera-
deviation of the pore model is not yet very large, the tures. They concluded that the results perhaps sug-
excellent fit of the barrier model is indeed a clear gested a dual resistance model with varying importance
indication of the dominant mechanism. of the two components depending on pressure and
Kawazoe et al.5 and Chihara et al.6 measured the temperature.
diffusions of nitrogen and propylene, respectively, in From the breakthrough results in the linear range of
Takeda MSC 5A using a pulse chromatographic method. the isotherm analyzed in this study, it appears that the
They concluded that the micropore resistance was transport of oxygen and nitrogen in the micropores of
distributed in the micropore interior. Ruthven7,8 and both CMS samples used in this study is controlled by a
Chen et al.9 (1994) gravimetrically measured the diffu- barrier resistance confined at the pore mouth. The
sion of oxygen and nitrogen in the BF CMS and were barrier model always provided a good overall fit of the
able to fit their data with a pore diffusion model. LaCava experimental results, and the extent of distinction from
et al.10 used gravimetric and batch column adsorption the pore model could be varied by controlling the
methods to measure the diffusion of oxygen and nitro- experimental conditions, as expected from the theoreti-
gen, and Srinivasan et al.11 measured the diffusion of cal analysis. It is, therefore, reasonable to say that any
 Ind. Eng. Chem. Res., Vol. 41, No. 5, 2002 1105

transition in transport mechanism is not evident in the kf ) external film mass transfer coefficient
temperature range covered. K ) Henry’s constant
The possibility of a dual resistance, however, cannot Kc ) Henry’s constant based on micropore volume
be excluded on the basis of the breakthrough analysis Kp ) Henry’s constant based on particle volume
presented here. Pore diffusional resistance, if any, will L ) bed length
normally become important in the later part of the Pe ) Peclet number (voL/DL)
uptake, i.e., in the later part of the breakthrough q ) adsorbate concentration in the micropore
response. While the pore model deviates significantly qo ) adsorbate concentration in the micropore in equilib-
from the experimental data in the early part of break- rium with Co
through, the results (particularly the magnified insets
j ) average adsorbate concentration in the micropore
q
in Figures 6 and 8) also seem to suggest that the pore
r ) distance along microparticle radius
model can be made to fit the experimental data very
well in the range C/Co ) 0.9-1.0. The long asymptotic rc ) microparticle radius
nature of the response in this region makes the distinc- R ) distance along macroparticle radius
tion between the two transport mechanisms less con- Rp ) macroparticle radius
clusive. In this respect, the uptake data from a static t ) time
method is expected to provide a better resolution. vo ) interstitial fluid velocity
Y ) dimensionless adsorbate concentration in the micro-
Conclusion pore (q/qo)
h ) dimensionless average adsorbate concentration in the
Y
A detailed comparative analysis of the column dy- micropore (qj /qo)
namics of micropore-diffusion-controlled bidisperse ad- h* ) Y
Y h in equilibrium with C hp
sorption has been performed for two different transport z ) distance along the bed length
mechanisms, namely, pore diffusional resistance dis- 0(0+, 0-) ) column inlet (inside, outside)
tributed in the micropore interior and barrier resistance
confined at the micropore mouth. It is shown that the Greek Letters
extent of agreement between the two models depends
on the product of two parameters, R and γ. Moreover, R ) dimensionless parameter [(1 - p)Kc/p]
by appropriate choice of operating conditions, it is β ) dimensionless parameter (DpL/Rp2vo)
possible to magnify the distinction between the two γ ) dimensionless parameter (DcL/rc2vo)
model solutions. These theoretical findings are sum- ∆ ) dimensionless parameter (kfRp/pDp)
marized in the form of an easy-to-read plot in Figure 2. τ ) dimensionless time (tvo/L)
It has been verified with oxygen and nitrogen break- τp ) particle tortuosity
through experiments in RS-10 zeolite, BF CMS, and ψ ) dimensionless parameter [p(1 - )/]
Takeda CMS that the proposed criteria can be success-  ) bed voidage
fully used to identify the transport mechanism in p ) particle voidage
adsorbent micropores. The results with RS-10 zeolite ξ ) dimensionless distance along micropore radius (r/rc)
provide a validation for the theoretical analysis by
η ) dimensionless distance along macropore radius (R/Rp)
reconfirming the expected distributed pore diffusional
χ ) dimensionless bed distance (z/L)
behavior. Experiments conducted in the region where
the distinction is prominent suggest that the transport
mechanism of oxygen and nitrogen in the micropores Literature Cited
of both CMS samples is controlled by barrier resistance
confined at the pore mouth. Closer looks at the re- (1) Glueckauf, E.; Coates, J. J. Theory of chromatography. Part
sponses in the range C/Co ) 0.9-1.0 also seem to IV. The influence of incomplete equilibrium on the front boundary
support the presence of some pore diffusional resistance of chromatograms and on the effectiveness of separation. J. Chem.
Soc. 1947, 1315.
in the later part of uptake.
(2) Haynes, J. R.; Sharma, H. W. The determination of effective
The experimental verification of the theoretical re- diffusivity by gas chromatography time domain solutions. Chem.
sults was confined to the right-hand side of the maxima Eng. Sci. 1975, 30, 955.
in the plot of E vs Rγ. From the analysis of the (3) Ruthven, D. M. Principles of Adsorption and Adsorption
experimental results, it appears that the maximum Processes; John Wiley & Sons: New York, 1984.
value of Rγ at which the two models became clearly (4) Hassan, M. M.; Raghavan, N. S.; Ruthven, D. M.; Boniface,
distinguishable visually was ∼0.4. H. A. Pressure swing adsorption II. AIChE J. 1985, 31, 2008.
(5) Kawazoe, K.; Suzuki, M.; Chihara, K. Chromatographic
study of diffusion in molecular-sieving carbon. J. Chem. Eng. Jpn.
Notation 1974, 7, 151.
C ) adsorbate concentration in the fluid phase (6) Chihara, K.; Suzuki, M.; Kawazoe, K. Adsorption rate on
Co ) adsorbate concentration in the feed molecular-sieving carbon by chromatography. AIChE J. 1978, 24,
Cp ) adsorbate concentration in the macropore 237.
h ) dimensionless adsorbate concentration (C/Co)
C (7) Ruthven, D. M.; Raghavan, N. S.; Hassan, M. M. Adsorption
and diffusion of nitrogen and oxygen in a carbon molecular sieve.
h p ) dimensionless adsorbate concentration in the macro-
C
Chem. Eng. Sci. 1986, 41, 1325.
pore (Cp/Co)
(8) Ruthven, D. M. Diffusion of oxygen and nitrogen in carbon
D ) diffusivity molecular sieve. Chem. Eng. Sci. 1992, 47, 4305.
Dc ) micropore diffusivity (9) Chen, Y. D.; Yang, R. T.; Uawithya, P. Diffusion of oxygen,
DL ) axial dispersion coefficient nitrogen and their mixtures in carbon molecular sieve. AIChE J.
Dm ) molecular diffusivity 1994, 40, 577.
Dp ) bulk molecular diffusivity (10) LaCava, A. I.; Koss, V. A.; Wickens, D. Non-Fickian
k ) rate constant adsorption rate behaviour of some carbon molecular sieves: I. Slit-
kb ) barrier mass transfer coefficient potential rate model. Gas Sep. Purif. 1989, 3, 180.
1106 Ind. Eng. Chem. Res., Vol. 41, No. 5, 2002

(11) Srinivasan, R.; Auvil, S. R.; Schork, J. M. Mass transfer (18) Forsim: A Fortran Package for the Automated Solution of
in carbon molecular sievessAn interpretation of Langmuir kinet- Coupled Partial and/or Ordinary Differential Equation Systems;
ics. Chem. Eng. J. 1995, 57, 137. Atomic Energy of Canada Limited: Chalk River, Ontario, Canada,
(12) Liu, H.; Ruthven, D. M. Diffusion in Carbon Molecular 1976.
Sieves. In Proceedings of the 5th International Conference on the (19) Rasmuson, A. Time domain solution of a model for
Fundamentals of Adsorption; LaVan, M. D., Ed.; Kluwer Academic transport processes in bidisperse structured catalysts. Chem. Eng.
Publishers: Norwell, MA, 1996. Sci. 1982, 37, 787.
(13) Rynders, R. M.; Rao, M. B.; Sircar, S. Isotope exchange (20) Zhong, G. Kinetically Controlled Pressure Swing Adsorp-
technique for measurement of gas adsorption equilibria and tion. M.Eng. Dissertation, The National University of Singapore,
kinetics. AIChE J. 1997, 43, 2456. Singapore, 2000.
(14) Loughlin, K. F.; Hassan, M. M.; Fatehi, A. I.; Zahur, M. (21) Sherwood, T. K.; Pigford, R. L.; Wilke, C. R. Mass Transfer;
Rate and equilibrium sorption parameters for nitrogen and McGraw-Hill Book Company: Singapore, 1975.
methane on carbon molecular sieve. Gas Sep. Purif. 1993, 7, 264. (22) Farooq, S. Sorption and diffusion of oxygen and nitrogen
(15) Kärger, J.; Ruthven, D. M. Diffusion in Zeolites and Other in molecular sieve RS-10. Gas Sep. Purif. 1995, 9, 205.
Microporous Solids; John Wiley & Sons: New York, 1992.
(16) Liaw, C. H.; Wang, J. S. P.; Greenkorn, R. A.; Chao, K. C. Received for review May 22, 2001
Kinetics of fixed bed adsorption. AIChE J. 1979, 25, 376. Revised manuscript received August 13, 2001
(17) Raghavan, N. S.; Ruthven, D. M. Simulation of chromato- Accepted August 15, 2001
graphic response in columns packed with bidisperse structured
particles. Chem. Eng. Sci. 1985, 40, 699. IE0104621