Effects of Carbon-to-Zeolite Ratio

on Layered Bed H 2 PSA for Coke Oven Gas

Chang-Ha Lee, Jaeyoung Yang, and Hyungwoong Ahn
Dept. of Chemical Engineering, Yonsei University, Seoul 120-749, South Korea

Effects of carbon-to-zeolite ratio on a layered bed H2 PSA using acti®ated carbon

and zeolite 5A were studied experimentally and theoretically. Coke o®en gas (56.4 ®ol.
% H2 , 26.6 ®ol. % CH4 , 8.4 ®ol. % CO, 5.5 ®ol. % N2 , and 3.1 ®ol. % CO2 ) was used
as a feed gas for the se®en-step two-bed PSA process incorporating a backfill step. In
these experiments, the effects of three operating ®ariables such as adsorption pressure,
feed rate and purge rate on the performance of a layered bed PSA were in®estigated.
The layered bed ga®e better purity than the single-adsorbent bed at the same operating
condition, except at low purge rate. Since e®ery component had its own front ®elocity at
each layer, a carbon-to-zeolite ratio affected product purity at a gi®en reco®ery or
throughput. Moreo®er, for a high-purity H2 product from coke o®en gas, an optimum
carbon-to-zeolite ratio had to be determined to control a leading wa®efront of N2 . In
layered bed PSA processes, the temperature ®ariations inside the bed reflected a kind of
inflection or plateau at which a roll-up phenomenon occurred and showed the dynam-
ics of adsorption well at each step during a cycle. Simulated results of the dynamic
model incorporating mass, energy and momentum balances agreed well with the PSA
experimental results.

Introduction
Increasing demand for hydrogen, particularly in a Single-adsorbent H 2 PSA for binary or ternary mixtures
petroleum refinery and in the petrochemical process, has with only major impurities have been employed in many pre-
provided a strong economic motivation to develop a process vious works ŽDoong and Yang, 1986; Kumar, 1994; Kim et
for recovering hydrogen from refinery fuel gases, coke oven al., 1995; Yang et al., 1997a,b.. However, a practical necessity
gases ŽCOG., other similar sources, and from more tradi- to obtain a high purity product from a multicomponent feed
tional sources such as off-gas ŽRuthven et al., 1994; Yang, in commercialized H 2 PSA processes has prompted a simul-
1987.. In many industrial processes using hydrogen, a high- taneous use of various kinds of adsorbents because different
purity hydrogen product of at least 99.99% is required and adsorbents show different selectivity depending on the adsor-
higher than 99.999% is often required as well. Since hydro- bate. Therefore, an important characteristic of the H 2 PSA
gen is adsorbed much less strongly than almost any other process is that many different adsorbents are used either in
components, a well-designed pressure swing system can meet multiple-layered beds, where each bed is composed of differ-
this challenge. Indeed, this is one application for which pres- ent adsorbents, or in two groups of adsorption beds with dif-
sure swing adsorption ŽPSA. has a clear advantage over al- ferent adsorbents. In the typical H 2 PSA with layered beds,
most all other possible approaches, for many these purity lev- the activated carbon layer at the bottom of the bed separates
els are unattainable. a bulk stream as a separator and the zeolite layer purifies the
raffinate stream from the activated carbon separator. Also,
the backfill step or product pressurization step in the com-
Correspondence concerning this article should be addressed to C.-H. Lee. mercial H 2 PSA has widely been used to increase product
Current address of J. Yang: Sunkyong Engineering and Construction Ltd., Seoul,
South Korea. purity ŽLiow and Kenney, 1990; Chou and Huang, 1994..

AIChE Journal March 1999 Vol. 45, No. 3 535

 In spite of the wide utilization of a layered bed H 2 PSA, and the overall mass balance was represented as follows
only a few studies have been done up to now. Kikkinides et
al. Ž1993. and Kumar Ž1994. studied the influence of adsor- ­ 2P ­P ­u ­P
bents in a PSA process. Chlendi et al. Ž1995. investigated a y DL q qP qu
six-step layered bed PSA process to recover H 2 from a ­z 2
­t ­z ­z
pseudo-cracking natural gas, which was carried out by numer- ­ 2
1 ­ 1 ­ 1
ical simulation without energy balance and which assumed
CO and N2 as a single component. Recently, Yang and Lee
q PT y DL
­zž / ž /
2
T
q
­t T
qu
­z T ž /
Ž1998. studied adsorption dynamics of a layered bed using
coke oven gas. Through breakthrough experiments and a dy- ­ 1 ­P 1y e n ­q j
y2 DLT ž / q r p RT Ý s0 Ž2.
namic model incorporating mass, energy, and momentum ­z T ­z e js 1 ­t
balances, they found that controlling a leading wave front of
N2 played a very important role in obtaining a high purity
Assuming a thermal equilibrium between fluid and particles,
product and the breakthrough times of every component var-
the energy balance for gas and solid phases was given by
ied with a carbon ratio. They also pointed out there will be
an optimum carbon ratio of a layered bed H 2 PSA. In that
study, they defined a carbon ratio Žc.r.. as a ratio of an acti- ­ 2T ­T ­T
y KL q ar g Ž C p . g q r B Ž C p . s
Ž . ­ t q rg Ž Cp . g e u ­ z
vated carbon layer length to a bed length, because the capac- ­z 2

ity of an adsorbent had to be compared with others under the

packing state.
n ­q i 2 hi
y rB Ý Qi q Ž T yTw . s 0 Ž 3 .
A basic study on the cyclic behavior of a layered bed PSA i ­t R Bi
process is very important in understanding well the perfor-
mance of the processes under various conditions. In this study,
where K L is the effective axial thermal conductivity ŽJrm ? s ?
by using a layered bed packed with activated carbon and zeo-
K. ŽSuzuki, 1990; Malek and Farooq, 1997..
lite 5A, the effects of carbon-to-zeolite ratio on a layered
Since the diameter of the adsorption bed used in the pre-
two-bed PSA process with a backfill step were investigated
sent study was rather small, the heat loss through a wall and
experimentally and theoretically for the H 2 recovery from
heat accumulation in the wall could not be neglected. There-
COG. The effects of various operating variables such as ad-
fore, another energy balance for the wall of the adsorption
sorption pressure, feed flow rate, and purge rate on product
bed was used by neglecting the axial conduction in the wall.
purity and recovery were also investigated with a change in
the carbon ratio of the bed. Then, the results were compared
with those of single-adsorbent beds. The experimental results ­ Tw
rw Ž C p . w A w s 2p R Bi h i Ž T yTw . y2p R Bo h o Ž Tw yTat m .
and dynamics of these H 2 PSA processes were predicted by a ­t
model simulation, where the effect of operating variables, as Ž 4a .
well as the optimum carbon-to-zeolite ratio, had all been
taken into account. A w sp Ž R 2Bo y R 2Bi . Ž 4b .

Mathematical Modeling The Ergun’s equation shown below was applied to the
A nonisothermal dynamic model including a component pressure drop across the bed ŽLu et al., 1993; Yang et al.,
balance, overall mass balance, energy balance and momen- 1998.
tum balance was used to simulate a PSA process with the
following assumptions: Ži. the flow pattern in the bed can be dP
y s a mn q brn < n < Ž 5a .
described by an axial dispersion plug-flow model; Žii. the in- dz
traparticle mass-transfer and adsorption equilibrium can be 2
described by a linear driving force ŽLDF. model and Lang- 150 Ž 1y e . Ž 1y e .
as , bs1.75 Ž 5b .
muir-Freundlich ŽL-F. model, respectively; Žiii. the solid and 4 R 2p e3 2 Rpe 3
gas phases reach thermal equilibrium instantaneously; Živ. ra-
dial concentration and temperature gradients in the adsorp-
where n is superficial velocity Žmrs..
tion bed are negligible; Žv. axial conduction in the wall can be
The sorption rate into an adsorbent pellet was described
neglected; Žvi. the gas phase behaves as an ideal gas mixture.
by the following LDF model, with a single lumped mass-
Using an axially dispersed plug flow and an ideal gas law,
transfer parameter v Žsy1 . ŽRuthven et al., 1994.
the material balance for the bulk phase in the adsorption col-
umn was given by
­q i
s v i Ž qUi yqi . Ž6.
­ yi
2
­ yi ­ yi ­t
y DL q qu
­ z2 ­t ­z
The well-known Danckwerts boundary conditions were ap-
RT 1y e ­ qi n ­ qj plied to the steps which have influent streams. The boundary
q
P e
rp
ž ­t
y yi Ý
js 1 ­t / s0 Ž1. conditions used in the PSA simulation were in the following
forms:

536 March 1999 Vol. 45, No. 3 AIChE Journal

 Boundary Conditions for Feed Pressurization and Adsorption DL 20
Steps s q0.5 Ž 11.
2 uR p Re Sc

­ yi
y DL ž / s u Ž yi < zs0y y yi < zs0q . ; The effective axial thermal conductivity K L was used by
­z zs 0 the following empirical correlation ŽYagi et al., 1960; Kunii
­ yi and Smith, 1960; Suzuki, 1990. instead of the axial thermal
ž /
­z zs L
s0 Ž 7a . dispersion coefficient ŽYang et al., 1997b.

­T K Lrk g s K L0rk g q d Pr Re Ž 12a .

y KL ž / s r g Ž C p . g u Ž T < zs0y yT < zs0q . ;
­z zs 0 1y e
K L0rk g s e q Ž 12b .
­T f q Ž 2r3 . Ž k grk s .
ž /
­z zs L
s0 Ž 7b .
e y0.26
f s f 2 qŽ f1y f 2 . ž / for 0.260 F e F 0.476
where yi < zs0y means feed composition for component i. 0.216
Boundary Conditions for Purge, Pressurizing Pressure Equal- Ž 12c .
ization, and Backfill Steps
The bed porosities for activated carbon and zeolite 5A beds
­ yi were both within the range of Eq. 12c, as shown in Table 1.
y DL ž / s u Ž yi < zsL q y yi < zsL y . ;
­z zs L
Three parameters Ž d , f 1, and f 2 . of these equations for both
activated carbon bed and zeolite 5A bed were 0.75, 0.2, and
­ yi 0.1, respectively ŽSuzuki, 1990.. The constant effective axial
ž /
­z zs 0
s0 Ž 8a .
thermal conductivity, which was estimated by using Eq. 12
under the experimental condition of the adsorption step, was
­T used for the simulation of that run.
y KL ž / s r g Ž C p . g u Ž T < zsL q yT < zsL y . ;
­z zs L For input parameters of model simulation, the adsorption
isotherm data and the sorption rate by a linear driving force
­T ŽLDF. model of CH 4 , CO, N2 , CO 2 , and H 2 on the activated
ž /
­z zs 0
s0 Ž 8b .
carbon and zeolite 5A were the same as those in the previous
work ŽYang and Lee, 1998.. The isosteric heat of adsorption
where yi < zsL q means a volume-averaged composition of the for each adsorbate-adsorbent pair was calculated by using a
effluent stream during the adsorption step for purge and Clausius-Clapeyron equation ŽYang et al., 1997b..
backfill steps, and a temporal effluent stream’s composition To develop a mathematical model for a layered bed, it was
during a depressurizing pressure equalization step for the assumed that a double layered bed consisted of two single-
pressurizing pressure equalization step, respectively. The fluid adsorbent beds linked in series. This mathematical model and
velocity is inherently negative during these steps. numerical simulation for a layered bed PSA process were de-
Boundary Conditions for Depressurizing Pressure Equalization scribed elsewhere in details ŽYang and Lee, 1998..
and Countercurrent Depressurization Steps
Experimental Studies
­ yi ­T
The COG feed used in this study was H 2 Ž56.4 vol. %.
ž /
­z zs 0
s ž /
­z zs0
s 0;
mixture gas with CH 4 Ž26.6 vol. %., CO Ž8.4 vol. %., N2 Ž5.5
­ yi ­T
ž /
­z zs L
s ž /
­z zs L
s0 Ž9.
Table 1. Characteristics of Adsorption Bed and
Adsorbents
The equilibrium of mixtures was predicted by following the
Adsorption Bed
extended Langmuir-Freundlich model Length 100 cm
Inside diameter 3.71 cm
qmi Bi Pin i Outside diameter 4.245 cm
qi s n Ž 10a . Heat capacity of wall 0.12 calrg ? K
1q Ý Bj Pjn j Activated Carbon Bed Zeolite Bed
js 1 Bulk Žbed. density 0.482 grcm3 0.764 grcm3
External void fraction 0.433 0.357
qm s k 1 q k 2 T , Bs k 3 e k 4rT , ns k 5 q k 6rT Ž 10b . Total void fraction 0.78 0.77
Adsorbents
The axial dispersion coefficient DL Žm2rs. was calculated Activated Carbon Zeolite 5A
by a Wakao equation using interstitial feed velocity at the Pellet size 6]16 mesh 4]8 mesh
adsorption pressure and this constant value was used for all Pellet density 0.85 grcm3 1.16 grcm3
Heat capacity 0.25 calrg ? K 0.22 calrg ? K
the steps ŽWakao and Funazkri, 1978.

AIChE Journal March 1999 Vol. 45, No. 3 537

vol. %., and CO 2 Ž3.1 vol. %. purchased from a gas company
ŽDaesung Sanso Co., South Korea..
The adsorption beds were 100 cm long and 3.71 cm I.D.
Four resistance temperature detectors ŽRTD, Pt 100V . were
installed at the positions of 10, 30, 50, and 75 cm from the
feed end to measure temperature variations inside the ad-
sorption bed. For layered bed experiments, the activated car-
bon ŽPCB, 6]16 mesh. manufactured by Calgon Carbon Co.
ŽU.S.A.. was packed into the bottom of the bed, because a
zeolite 5A is almost saturated with CO 2 at low pressure. Then,
after putting a metal screen of 0.3 mm thickness on top of
the activated carbon layer, zeolite 5A Ž4]8 mesh. manufac-
tured by W. R. Grace Co. ŽU.S.A.. was added. The character-
istics of the bed and adsorbents are shown in Table 1.
Feed and purge flow rates were controlled by mass-flow
controllers ŽBronkhorst high-tech, F-201C.. To calculate re-
covery accurately, the amount of gas flowing into and out of
the PSA system was measured with a mass-flow meter
ŽBronkhorst high-tech, F-112ac-HA-55-V. and a wet gas me-
ter ŽShinagawa, W-NK-1B.. Gas samples taken from a prod-
uct tank were analyzed by using a mass spectrometer ŽBal-
zers, QME 200. and a GC ŽHP, GC 5890 II. for N2 and CO. Figure 1. Cyclic approach of gas composition of efflu-
The PSA system was fully automated by interfacing with a ent stream to dynamic steady state in a lay-
personal computer and all the measured data were saved on ered bed process with 0.65 carbon-to-zeolite
the computer. The details for the system are described in ratio under 10 atm adsorption pressure, 8
Yang and Lee’s Ž1998. study. LSTP//min feed rate, and 0.7 LSTP//min purge
The following cycle sequence and step time for this two-bed rate.
seven-step PSA process were used: Ž1. feed pressurization of
a partially pressurized bed by a previous pressurizing pres-
sure equalization ŽFP: 4s.; Ž2. high-pressure adsorption ŽAD: in this figure. After approximately 10 cycles, the difference
180s.; Ž3. depressurizing pressure equalization ŽDPE: 20s.; Ž4. between the performance of the last two cycles was less than
countercurrent depressurization ŽDP: 8s.; Ž5. purge with a 0.01 %. And, it was also confirmed by the 25-cycle PSA ex-
light product, H 2 ŽPG: 180s.; Ž6. pressurizing pressure equal- periments and 100-cycle simulations. In the present study, all
ization ŽPPE: 20s.; Ž7. a backfill step with light product up to the experimental data were collected at above 15 cycles.
the average pressure of the adsorption pressure and final In Figure 1, constant purity product at the adsorption step
pressure of pressurizing pressure equalization ŽBF: 4s.. The after a cyclic steady state was obtained because the operating
effluent in step II ŽAD. is high purity H 2 product and this conditions were favorable. During the depressurizing pres-
experimental and theoretical study concentrated upon the sure equalization step ŽDPE., the emitting gases from the
quality and quantity of this light product rather than those of product end of the adsorption bed with a steep decrease in
heavy product. total pressure made concentration wavefronts disperse and
proceed quickly. Also, this step made the adsorbed compo-
Results and Discussion nents desorb and led to a decrease in H 2 mole fraction of
effluent stream as shown in Figure 1. However, Yang et al.
Cyclic steady state of PSA process Ž1997b. pointed out that a decrease in H 2 concentration of
Figure 1 showed a representative cyclic behavior of a lay- effluent stream during the DPE step contaminated the prod-
ered bed PSA process, plotting a simulated composition vari- uct end and had a bad effect on product purity. Therefore,
ation of effluent stream from the adsorption bed with time. the product purity at this step was not high due to the unfa-
The hydrogen concentration of the effluent was obtained from vorable operating condition. At the countercurrent depres-
the product end during the adsorption step and depressuriz- surization step, the rapid decrease in total pressure led to
ing pressure equalization step. And, it was also obtained from further desorption of the adsorbed components and an in-
the feed end during the countercurrent depressurization step crease in mole fraction of strongly adsorbed components in
and purge step. an effluent stream. Also, desorption continued at the purge
In the case of the frozen solid-phase model, it was impossi- step. Therefore, the composition of strongly adsorbed com-
ble to predict a transient variation of the effluent stream dur- ponents rose at the early stage of a purge step until the influ-
ing pressurization and depressurization steps because the ent stream of higher purity H 2 reached the feed end. The
solid-phase concentration was assumed to remain unchanged dynamics of each step in the layered bed PSA process will be
at the end of the previous step ŽFarooq et al., 1988; Lu et al., explained later in detail.
1993.. However, the LDF model used in the present study Along with the concentration of the effluent stream, the
showed a reasonable prediction of effluent composition for temperature variation inside the adsorption bed represented
all the steps in Figure 1. The composition of the effluent a cyclic behavior of a PSA process. The temperature varia-
stream approached rapidly to a cyclic steady state as shown tions at four locations along the layered bed with 0.65 car-

538 March 1999 Vol. 45, No. 3 AIChE Journal

 Figure 3. Effect of carbon ratio on (a) purity and (b)
recovery at three adsorption pressures, 7
LSTP//min feed rate, and 0.7 LSTP//min purge
Figure 2. Transient variation of temperature at (a) z =10 rate.
cm, (b) z=30 cm, (c) z=50 cm, and (d) z=75
cm for a layered bed process with 0.65 car-
bon-to-zeolite ratio under 10 atm adsorption adsorption beds with various carbon-to-zeolite ratios under
pressure, 8 LTP//min feed rate, and 0.7 LTP//min three different adsorption pressures. As shown in Figure 3a
purge rate. and Table 2, there existed a carbon-to-zeolite ratio which gave
the highest product purity at each adsorption pressure. Also,
the purity difference upon carbon-to-zeolite ratio appeared
bon-to-zeolite ratio were presented in Figure 2. It could be much less sharp as adsorption pressure increased. At 5 atm
seen that a PSA process approached rapidly a cyclic steady adsorption pressure, the layered bed with around 0.5 carbon-
state as revealed by the predicted concentration of an efflu- to-zeolite ratio gave best product purity and the activated
ent stream ŽFigure 1.. The range of temperature swing was carbon bed gave higher purity than the zeolite bed. At this
largest at the feed end section, because this section behaved adsorption pressure, the reason for purity change by carbon
as a separator where a considerable amount of adsorbates ratio was explained in Figure 4. While the main impurities in
underwent adsorption and desorption during a cycle. How- the H 2 product of the processes using zeolite-rich bed were
ever, the range of temperature swing at the product end sec-
tion was smallest because this section purified the raffinate
stream from the activated carbon separator. It was notewor- Table 2. H 2 Product Purity and Recovery Under 11
thy that the two temperature sensors of 50 and 75 cm at the atm Adsorption Pressure, 7 LSTPrrmin Feed Rate and
initial cycles showed relatively little extent of temperature 0.7 LSTPr
rmin Purge Rate
swing. This implied that, at an early stage of operation, the Carbon-to-Zeolite
concentration wavefronts did not proceed much from feed Ratio 0.0 0.2 0.35 0.5 0.65 1.0
end due to the initial condition of the clean bed. H 2 Purity Ž%. 99.983 99.986 99.985 99.991 99.981 99.988
Ž99.986. Ž99.987. Ž99.988. Ž99.999. Ž99.995. Ž99.978.
Effects of adsorption pressure and carbon-to-zeolite ratio H 2 Recovery Ž%. 71.96 72.09 68.75 69.38 66.62 66.62
Ž71.13. Ž68.25. Ž67.10. Ž66.15. Ž65.74. Ž62.66.
To study the effect of carbon-to-zeolite ratio of a layered
bed in detail, many PSA runs were experimented by using Simulated results are in parentheses.

AIChE Journal March 1999 Vol. 45, No. 3 539

 99.99% H 2 purity product even at 11 atm by a single-ad-
sorbed bed process. The carbon-to-zeolite ratio to obtain the
highest purity at 11 atm was between 0.35 and 0.5, experi-
mentally, and between 0.4 and 0.5, theoretically, giving
99.991q% and 99.999% H 2 purity, respectively.
In Figure 3b, the recovery of H 2 product showed an inflec-
tional decrease with increasing carbon-to-zeolite ratio at each
adsorption pressure, especially at low adsorption pressure.
The recovery in the range of 0.4]0.6 carbon-to-zeolite ratio
which gave the highest product purity at a given adsorption
pressure was higher than those of the activated carbon-rich
bed processes which gave lower product purity under the same
condition. Also, in single-adsorbent bed processes that gave
similar product purity, the recovery at the zeolite bed process
was higher than that at the activated carbon bed process and
the difference in recovery between two single-adsorbent bed
processes increased with an increase in the adsorption pres-
sure.
Through a simulation at higher than 11 atm adsorption
Figure 4. Average composition variation of products pressure, it was found that recovery decreased more steeply
with carbon ratio under 5 atm adsorption and purity approached more slowly to the desired product
pressure, 7 LSTP//min feed rate, and 0.7 LSTP// purity Ž99.99q%. by increasing adsorption pressure in a zeo-
min purge rate. lite bed process than in any other bed processes. Therefore,
at higher than 99.99% H 2 purity from COG, a layered bed
PSA proved to be better than activated carbon or zeolite 5A
bed PSA in respect to H 2 recovery.
CH 4 and CO, N2 became a major impurity in a product as
carbon-to-zeolite ratio increased. Although the adsorption
pressure was different and the adsorption bed was not clean
in a cyclic PSA process, this result agreed well with that ob- Effects of feed rate and carbon-to-zeolite ratio
tained from breakthrough experiments ŽYang and Lee, 1998.. The effect of feed rate on the performance of a PSA pro-
Due to the adsorption capacity of the activated carbon bed cess was studied with varying carbon ratio under the adsorp-
for CH 4 , the breakthrough time of CH 4 was longer than that tion pressure of 11 atm and the purge rate of 0.7 L STPrmin.
of CO and N2 . Therefore, CH 4 did not act as one of main Considering the adsorption capacity of a given adsorption
impurities for an activated carbon-bed process. On the other bed, the effect of increasing feed rate was basically the same
hand, in the case of a zeolite bed, CH 4 breakthrough time as that of decreasing adsorption pressure. As shown in Fig-
was comparable with that of CO and N2 . Therefore, CH 4 ure 5a, the H 2 purity curve had a plateau at some carbon
could be one of the main impurities in a zeolite bed process, ratio range and then declined. The carbon-to-zeolite ratio to
especially when the adsorption pressure was low. As a result, obtain the highest purity under 10 LSTPrmin feed was be-
the carbon-to-zeolite ratio to improve H 2 purity moved to a tween 0.5 and 0.8. However, this ratio for purity moved to
low carbon-to-zeolite ratio as N2 composition in the feed in- the left as the feed rate was reduced. For PSA processes with
creased. On the other hand, the carbon-to-zeolite ratio for 10 LSTPrmin feed rate, the effect of carbon-to-zeolite ratio
the high H 2 purity moved to high carbon-to-zeolite ratio as on purity was prominent at low values. Therefore, it could be
the composition of CH 4 and CO in a product increased. said that activated carbon played a very important role in
However, in the case of 8 atm adsorption pressure, the controlling the extent of breakthroughs of major impurities
simulated result in Figure 3a showed that the carbon ratio over low carbon-to-zeolite ratio range of the adsorption bed.
for the highest product purity shifted a little to low carbon Also, compared to zeolite, activated carbon had larger ad-
ratio in comparison with the 5 atm case. At this condition, sorption capacity for CH 4 per unit volume of an adsorption
even in a zeolite bed process, the major impurity became N2 bed. Also, the adsorption capacity of activated carbon for CO
with a relatively small amount of CO instead of CH 4 . Since was slightly larger than that of zeolite 5A at higher than 5
the increased adsorption capacity for CH 4 and CO with an atm adsorption pressure. Therefore, with an increase in feed
increase in the adsorption pressure was relatively higher than rate, the H 2 purity in the activated carbon bed process
that for N2 , the importance of CH 4 and CO in product purity dropped less than that in the zeolite bed process.
became small. Even though the effect of carbon-to-zeolite ra- As expected, the zeolite bed gave the highest recovery at
tio on purity was not clearly shown for the 11 atm case in each feed rate condition in Figure 5b, which was similar to
Figure 3a, this effect played an important role in obtaining the effect of the adsorption pressure mentioned earlier.
high purity in H 2 product Ž )99.99q%. from COG by a PSA However, at higher than 99.99% H 2 purity, the layered bed
process. The role of a layered bed in the PSA process was PSA gave the highest recovery of H 2 like the results men-
clearly shown in Table 2, which represented a performance tioned in the previous section. In general, experimental re-
comparison of activated carbon, zeolite and layered beds. As sults gave higher recoveries than simulated results and a max-
shown in this table, it was very hard to obtain more than imum of 4% error was found at 7 and 8 LSTPrmin feed rates.

540 March 1999 Vol. 45, No. 3 AIChE Journal

Figure 5. Effects of carbon ratio on (a) purity and (b) Figure 6. Effects of carbon ratio on (a) purity and (b)
recovery at three feed rates, 11 atm adsorp- recovery at three purge rates, 11 atm adsorp-
tion pressure, and 0.7 LSTP//min purge rate. tion pressure and 7 LSTP//min feed rate.

LSTPrmin was obtained from around 0.5 carbon ratio bed

Effect of purge rate and carbon-to-zeolite ratio due to the adsorption capacity of the zeolite layer for N2 .
In Figure 6b, for recovery, all the experimental data re-
Unlike the above two cases, the effects of carbon-to-zeolite
sulted in slightly higher recovery than predicted. Also, the
ratio with various purge rates on PSA performance were
zeolite bed at each purge rate condition was better than any
rather complicated. As shown in Figure 6, at low purge rate,
other beds like the results of the adsorption pressure and the
0.1 LSTPrmin, the activated carbon bed gave the best H 2
feed rate. However, at high H 2 purity Ž99.99q%. similar re-
purity. Under the condition of the pressure swinging from 1
covery results were obtained as in the previous operating
atm to 11 atm, the working capacity of the activated carbon
variables.
bed, especially for CH 4 , was larger than that of the zeolite
5A bed ŽYang and Lee, 1998.. Therefore, the activated car-
bon bed could be regenerated in some measure by the de- Effects of carbon-to-zeolite ratio on layered bed dynamics
pressurization step itself. However, to acquire the needed It was noted that the layered bed gave higher product pu-
working capacity for the next cycle, partial pressure of a rity than two single-adsorbent beds at the same operating
strongly adsorbed component during a purge step needed to condition. This meant that the carbon-to-zeolite ratio to ob-
be reduced more in the zeolite 5A bed than in the activated tain the highest purity was present under the same operating
carbon bed. Therefore, the effect of the purge step with low condition. To clarify the fact, concentration profiles in the
purge rate on the activated carbon-rich bed was greater than gas phases at the end of the adsorption step were presented
that on the zeolite-rich bed. However, in Figure 6a, the im- in Figure 7 for three different layered beds. As shown in this
provement of purity in the zeolite-rich bed with an increase figure, although CH 4 was a major impurity gas in the feed,
in purge rate was higher than that in the activated carbon-rich the most important components in the H 2 PSA process for
bed. The reason was that the concentration wavefronts of im- COG were CO and N2 . Because the adsorption capacity of
purities in the layered bed by an incomplete purge of strongly activated carbon for N2 was less than that of zeolite 5A, N2
adsorbed components under a high purge rate became simi- became a major impurity in the product stream when carbon-
lar to those of the activated carbon bed and the product end to-zeolite ratio was increased.
of the layered bed was kept cleaner than that of the activated In Figure 7a, a small but broad roll-up of CH 4 was shown
carbon bed. As a result, the highest product purity at 0.7 due to the competitive adsorption with a more strongly ad-

AIChE Journal March 1999 Vol. 45, No. 3 541

 CH 4 wavefront, the carbon-to-zeolite ratio was increased to
0.7 and the results were presented in Figure 7c. According to
the increased carbon-to-zeolite ratio, the wavefront of CH 4
remained in the carbon layer and the other part of the ad-
sorption bed could be used to adsorb CO. Therefore, in the
product stream, the composition of CO can be reduced with
an increase of N2 composition because the fast and dispersed
MTZ of CO led to a long leading front on the N2 wave front.
Such behavior of wavefronts might become more complicated
at lower adsorption pressure, because the CH 4 wavefront did
not any longer only reside in a carbon layer and was mixed
with CO and N2 wavefronts. As a result, it was very impor-
tant to find out the superposing configuration of adsorbents
to obtain high product purity at a given recovery.
Along with concentration profiles inside the adsorption
bed, the temperature variations with time were also useful in
understanding the effect of carbon-to-zeolite ratio on bed dy-
namics. Experimental temperature variations with time at
four locations along the bed under 11 atm adsorption pres-
sure, 7 LSTDrmin feed rate, and 0.7 LSTDrmin were pre-
sented with the simulated results in Figure 8]10. The tem-
perature variations in these figures reflected a kind of inflec-
tion or plateau at which a roll-up phenomenon occurred as in
Figure 7.
In the case of the activated carbon bed process shown in
Figure 8a, an abrupt rise of temperature at a short feed pres-
surization step was followed by a successive rise and drop of
temperature from the feed end at the adsorption step. This
was because of the propagation of distinctive concentration
wavefronts mentioned in Figure 7. However, in the zeolite
bed process, the temperature excursion at the adsorption step
was continuously increased up to the third RTD. Since the
dispersed and advanced wavefront of CH 4 by CO 2 pushed
forward the CO wavefront at the adsorption step in the zeo-
lite bed process, mass-transfer zones of N2 and CO were
mixed at the product end section. According to this, the high-
Figure 7. Concentration profiles at a cyclic steady state est-temperature excursion in the zeolite bed occurred near
for layered bed processes with (a) 0.2, (b) 0.5, the third RTD position and was sharper than that in the acti-
and (c) 0.7 carbon-to-zeolite ratio at the end vated carbon bed.
of the adsorption step under 7 atm adsorption During the depressurizing pressure equalization step, Fig-
pressure, 7 LSTP//min feed rate, and 0.7 LSTP// ure 8 showed that the temperatures in both single-adsorbent
min purge rate. beds dropped because of the endothermic desorption. This
implied that some amount of adsorbates adsorbed during the
feed pressurization and adsorption steps was desorbed at this
sorbed component CO 2 . Particularly, the concentration pro- step. The temperatures in both single-adsorbent bed proc-
file of CO 2 passing over the activated carbon layer gave esses dropped significantly during the countercurrent depres-
detrimental effects on the zeolite layer due to the strong ad- surization step and at the beginning of the purge step. How-
sorption of CO 2 . Therefore, the product stream of zeolite-rich ever, after this temperature decrease, a small temperature
bed contained much CO as well as N2 because the dispersed rise in the activated carbon bed was shown at the purge step.
and advanced wavefront of CH 4 pushed forward the CO It was due to the heating by the purge gas, which was around
wavefront. 292 K. During the pressurizing pressure equalization step,
The layered bed with increased carbon ratio could elimi- temperature rose again in both single-adsorbent beds.
nate the above disadvantages of the zeolite-rich bed by im- Experimental temperature variations for the zeolite bed in
posing on each layer a different role. Namely, while activated Figure 8b showed some discrepancy with the simulated one.
carbon adsorbed mainly CH 4 , zeolite 5A adsorbed CO and Especially, the temperatures showed a relatively slow and
N2 . The corresponding concentration profiles of a layered bed small decrease during the countercurrent depressurization
with a 0.5 carbon-to-zeolite ratio were shown in Figure 7b. step and at the beginning of the purge step. Also, the experi-
This layered bed eliminated, in some measure, the disadvan- mental temperatures during the purge step decreased contin-
tages of a zeolite-rich bed. However, the forefront of the CH 4 uously, which was different from the activated carbon bed.
wavefront still caused a negative effect on a zeolite layer. This is possibly because of the slow desorption at the purge
Therefore, in order to protect the zeolite 5A layer from the step and the dispersed wavefronts of the adsorbates remain-

542 March 1999 Vol. 45, No. 3 AIChE Journal

 Figure 9. Temperature variations at a cyclic steady state
for layered bed PSA processes with (a) 0.2
and (b) 0.3 carbon-to-zeolite ratio under 11
Figure 8. Temperature variations at a cyclic steady state atm adsorption pressure, 7 LSTP//min feed rate
for single-adsorbent bed PSA processes with and 0.7 LSTP//min purge rate.
(a) activated carbon and (b) zeolite 5A under Expt.: ` 10 cm, ^ 30 cm, I 50 cm, \ 75 cm; Pred.:
10 cm, 30 cm, ] ? ] ? ] 50 cm, ] ? ? ] ? ? ] 75 cm.
11 atm adsorption pressure, 7 LSTP//min feed
rate, and 0.7 LSTP//min purge rate.
Expt.: ` 10 cm, ^ 30 cm, I 50 cm, \ 75 cm; Pred.:
10 cm, 30 cm, ] ? ] ? ] 50 cm, ] ? ? ] ? ? ] 75 cm.
that on the activated carbon. Also, the temperatures in the
position of second and third RTDs from the feed end were
continuously increased until the depressurizing pressure
ing in the bed after the regeneration steps. Also, since the equalization step, which was due to the dispersed concentra-
temperature in the bed during the desorption steps was not tion wavefronts of CH 4 and CO in the zeolite layer. How-
less than that of the atmosphere, there was no effect of heat ever, with increased carbon-to-zeolite ratio, the temperature
input from the surroundings. As a result, insufficient regen- variations of the layered bed process became similar to those
eration affected the adsorption and depressurizing pressure of the activated carbon bed process.
equalization steps which also showed some discrepancy be- In Figures 9 and 10, during the depressurizing pressure
tween experimental and simulated results. equalization step, the two temperature profiles at the prod-
Figures 9 and 10 showed temperature variations for each uct end rose because of the fast movement of wavefronts,
layered bed process with a different carbon-to-zeolite ratio. while the two temperatures at the feed end dropped. At that
The layered bed process with low carbon-to-zeolite ratio step, the zeolite-rich bed process showed sharper tempera-
showed quite different temperature variations from the zeo- ture excursion than the activated carbon-rich bed process due
lite bed process due to different concentration wavefronts of to the roll-up of CH 4 at the product end. However, the lay-
strong adsorbates in the activated carbon layer. As shown in ered bed process with 0.5 carbon-to-zeolite ratio in Figure
Figure 9a, an abrupt rise of temperature in the feed end oc- 10a showed much smaller temperature excursion at that step
curred due to the adsorption of strong adsorbates at a short than any other layered bed processes. Especially, the temper-
feed pressurization step. Then, the temperature excursion of ature variations for this layered bed process show an inter-
the second RTD Žzeolite layer. from the feed end was smaller mediate behavior of the single-adsorbent bed process shown
than that in the activated carbon layer because the adsorp- in Figure 8. While the temperature variations of two RTDs in
tion capacity of zeolite 5A for CH 4 and CO was smaller than the feed end at all steps were very similar to those of the

AIChE Journal March 1999 Vol. 45, No. 3 543

 to the synergistic effects of two adsorbents, the recovery of a
layered bed process was between the limits of two single-ad-
sorbent bed processes. In this experimental range, the car-
bon-to-zeolite ratio to improve purity moved to a lower ratio
with an increase in the adsorption pressure and the purge
rate and with a decrease in the feed rate. Under the operat-
ing condition producing a high purity H 2 product, the main
impurities in the zeolite-rich bed process were CO and N2 ,
while N2 became a major impurity in the product as the car-
bon-to-zeolite ratio increased. In the case of the two-bed H 2
PSA system used in the present study, the operating condi-
tions for obtaining 99.99q% high purity H 2 product were 11
atm adsorption pressure, 7 LSTPrmin feed rate, and 0.7
LSTPrmin purge rate.
The temperature variations in the bed showed well the in-
flection or plateau caused by the roll-up phenomenon in the
bed. The result of the dynamics of adsorption through tem-
perature variations agreed well with the carbon-to-zeolite ra-
tio obtained from the effects of operating variables. Espe-
cially, the temperature variations of the layered bed process
with 0.5 carbon-to-zeolite ratio showed an intermediate be-
havior of the single-adsorbent bed processes.
Since every component had its own front velocity at each
layer, carbon-to-zeolite ratio affected product purity at a given
recovery or throughput. Even though N2 was a minor impu-
rity in coke oven gas, controlling the leading wavefront of N2
played a very important role in obtaining a high purity prod-
uct and in determining the optimum carbon-to-zeolite ratio
of a PSA process for high purity H 2 recovery. As a result, to
obtain high product purity, it was very important to find out
Figure 10. Temperature variations at a cyclic steady
the superposing configuration of the adsorbents.
state for layered bed PSA processes with (a)
0.5 and (b) 0.65 carbon-to-zeolite ratio under
11 atm adsorption pressure, 7 LSTP//min feed Acknowledgment
rate, and 0.7 LSTP//min purge rate. The financial assistance and support of Sunkyong Engineering &
Expt.: ` 10 cm, ^ 30 cm, I 50 cm, \ 75 cm; Pred.: Construction Limited and R&D Management Center for Energy and
10 cm, 30 cm, ] ? ] ? ] 50 cm, ] ? ? ] ? ? ] 75 cm. Resources are gratefully acknowledged.

Notation
activated carbon bed process, the temperature variation in Ascross-sectional area, m2
the product end was similar to that of the zeolite bed process BsLangmuir-Freundlich isotherm parameter, atmy1
mainly due to the adsorption of nitrogen. As a result, the C ps heat capacity, Jrkg ? K
layered bed process with this carbon-to-zeolite ratio could hsheat-transfer coefficient, Jrm2 ? s ? K
give better product purity than any other process at that op- k gs thermal conductivity of fluid, Jrm ? s ? K
k ss thermal conductivity of particle, Jrm ? s ? K
erating condition. This was because the product end of this nsLangmuir-Freundlich isotherm parameter
layered bed was contaminated less by the N2 wavefront than Ps pressure, atm
that of other processes. This result for the highest purity PrsPrandtl number, Ž C p . g mrk g
agreed well with the experimental carbon ratio obtained from qm s Langmuir-Freundlich isotherm parameter, molrkg
qs volume-averaged adsorbed phase concentration, molrkg
the effects of operating variables.
qUs equilibrium adsorbed phase concentration, molrkg
Qsaverage isosteric heat of adsorption, Jrmol
Conclusions Rsradius, m
ResReynolds number, r g n Ž2 R p .rm
The effects of carbon-to-zeolite ratio on the seven-step ScsSchumidt number, mr grDm
two-bed PSA processes using activated carbon and zeolite 5A ts time, s
were studied to produce high purity H 2 product from COG. t ssstoichiometric breakthrough time, s
Tssolid phase and gas phase temperature, K
To develop the model, the layered bed was assumed to con- Tatm sambient temperature, K
sist of two independent beds with a single adsorbent in each. usinterstitial velocity, mrs
The predicted results by a mathematical model including ys mole fraction in gas phase
mass, energy, and momentum balances agreed well with the zsaxial position in an adsorption bed, m
astotal void fraction
experimental results. es interparticle void fraction
While the product purity of a layered bed process was not ms viscosity, mrkg ? s
between the limits of two single adsorbent bed processes due rsdensity, m3rkg

544 March 1999 Vol. 45, No. 3 AIChE Journal

Subscripts Kunii, D., and J. M. Smith, ‘‘Heat Transfer Characteristics of Porous
Rocks,’’ AIChE J., 6, 71 Ž1960..
Bsbed Liow, J.-L., and C. N. Kenney, ‘‘The Backfill Cycle of the Pressure
iscomponent i Swing Adsorption Process,’’ AIChE J., 36, 53 Ž1990..
ps pellet Lu, Z. P., J. M. Loureiro, A. E. Rodrigues, and M. D. LeVan, ‘‘Pres-
gsgas phase surization and Blowdown of Adsorption Beds}II. Effect of the
sssolid phase Momentum and Equilibrium Relations on Isothermal Operations,’’
ws wall Chem. Eng. Sci., 48, 1699 Ž1993..
Malek, A., and S. Farooq, ‘‘Kinetics of Hydrocarbon Adsorption on
Activated Carbon and Silica Gel,’’ AIChE J., 43, 761 Ž1997..
Literature Cited Ruthven, D. M., S. Farooq, and K. S. Knaebel, Pressure Swing Ad-
Chlendi, M., D. Tondeur, and F. Rolland, ‘‘A Method to Obtain a sorption, VCH Publishers, New York Ž1994..
Compact Representation of Process Performances from a Numeri- Suzuki, M., Adsorption Engineering, Elsevier, Amsterdam Ž1990..
cal Simulator: Example of Pressure Swing Adsorption for Pure Hy- Wakao, N., and T. Funazkri, ‘‘Effect of Fluid Dispersion Coefficients
drogen Production,’’ Gas Sep. Purif., 9, 125 Ž1995.. on Particle-to-Fluid Mass Transfer Coefficients in Packed Beds,’’
Chou, C.-T., and W.-C. Huang, ‘‘Simulation of a Four-Bed Pressure Chem. Eng. Sci., 33, 1375 Ž1978..
Swing Adsorption Process for Oxygen Enrichment,’’ Ind. Eng. Yagi, S., D. Kunii, and N. Wakao, ‘‘Studies on Axial Effective Ther-
Chem. Res., 33, 1250 Ž1994.. mal Conductivities in Packed Beds,’’ AIChE J., 6, 543 Ž1960..
Doong, S. J., and R. T. Yang, ‘‘Bulk Separation of Multicomponent Yang, R. T., Gas Separation by Adsorption Processes, Butterworths,
Gas Mixtures by Pressure Swing Adsorption: PorerSurface Diffu- Boston Ž1987..
sion and Equilibrium Model,’’ AIChE J., 37, 1027 Ž1986.. Yang, J., C. Cho, K. Baek, and C.-H. Lee, ‘‘Comparison of One-Bed
Farooq, S., M. M. Hassan, and D. M. Ruthven, ‘‘Heat Effects in and Two-Bed H 2 PSA Using Zeolite 5A,’’ HWAHAK KONGHAK,
Pressure Swing Adsorption Systems,’’ Chem. Eng. Sci., 43, 1017 35, 545 Ž1997a..
(1998). Yang, J., J.-W. Chang, and C.-H. Lee, ‘‘Separations of Hydrogen
Kikkinides, E. S., R. T. Yang, and S. H. Cho, ‘‘Concentration and Mixtures by a Two-Bed Pressure Swing Adsorption Process Using
Recovery of CO 2 from Flue Gas by Pressure Swing Adsorption,’’ Zeolite 5A,’’ Ind. Eng. Chem. Res., 36, 2789 Ž1997b..
Ind. Eng. Chem. Res., 32, 2714 Ž1993.. Yang, J., M.-W. Park, J.-W. Chang, and C.-H. Lee, ‘‘Effects of Pres-
Kim, W. G., J. Yang, S. Han, C. Cho, C.-H. Lee, and H. Lee, ‘‘Ex- sure Drop in a PSA Process,’’ Korean J. Chem. Eng., 15, 211 Ž1998..
perimental and Theoretical Study on H 2rCO 2 Separation by a Yang, J., and C.-H. Lee, ‘‘Adsorption Dynamics of a Layered Bed
Five-Step One-Column PSA Process,’’ Korean J. Chem. Eng., 12, PSA for H 2 Recovery from Coke Oven Gas,’’ AIChE J., 44, 1325
503 Ž1995.. Ž1998..
Kumar, R., ‘‘Pressure Swing Adsorption Process: Performance Opti-
mum and Adsorbent Selection,’’ Ind. Eng. Chem. Res., 33, 1600
Ž1994.. Manuscript recei®ed July 27, 1998, and re®ision recei®ed No®. 12, 1998.

AIChE Journal March 1999 Vol. 45, No. 3 545