This article was downloaded by: [Fondren Library, Rice University ]

On: 12 November 2014, At: 15:59

Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House,
37-41 Mortimer Street, London W1T 3JH, UK

Separation Science and Technology

Publication details, including instructions for authors and subscription information:
http://www.tandfonline.com/loi/lsst20

Efficient Cycle Recovery of Hydrogen from a Low

Concentration Pyrolysis Gas Stream by Pressure Swing
Adsorption – An Experimental Evaluation
a a
Eliasu Teiseh & Sergio Capareda
a
Texas A&M University , College Station , TX , USA
Accepted author version posted online: 03 Apr 2012.Published online: 12 Jul 2012.

To cite this article: Eliasu Teiseh & Sergio Capareda (2012) Efficient Cycle Recovery of Hydrogen from a Low Concentration
Pyrolysis Gas Stream by Pressure Swing Adsorption – An Experimental Evaluation, Separation Science and Technology, 47:10,
1522-1530, DOI: 10.1080/01496395.2012.655834

To link to this article: http://dx.doi.org/10.1080/01496395.2012.655834

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained
in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no
representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the
Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and
are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and
should be independently verified with primary sources of information. Taylor and Francis shall not be liable for
any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever
or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of
the Content.

This article may be used for research, teaching, and private study purposes. Any substantial or systematic
reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any
form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://
www.tandfonline.com/page/terms-and-conditions
 Separation Science and Technology, 47: 1522–1530, 2012
Copyright # 2012 Texas A&M University
ISSN: 0149-6395 print=1520-5754 online
DOI: 10.1080/01496395.2012.655834
Efficient Cycle Recovery of Hydrogen from a Low
Concentration Pyrolysis Gas Stream by Pressure Swing
Adsorption – An Experimental Evaluation
Eliasu Teiseh and Sergio Capareda
Texas A&M University, College Station, TX, USA
Downloaded by [Fondren Library, Rice University ] at 15:59 12 November 2014
Breakthrough curves, cycle mass balances, and cycle bed produc-
tivities (mg H2 per gram of adsorbent) on three dual adsorbent
amounts (g) of 2,892, 1,963, and 1,013 respectively each filling
200 cm, 135 cm, and 70 cm of a 5.0 cm internal diameter stainless
steel pipe were performed. The approximate optimum (sludge pyrol-
ysis) synthesis gas with composition in volume % of 45% H2/35%
CO/20% CH4 was used as the feed gas with molecular sieve 5 Å
and activated carbon as adsorbents. Impurity breakthroughs
occurred at
14.9, 12.3, and 5.0 minutes respectively for % cycle
recoveries of 72.2, 65.0, and 60.2 using 2,892, 1,962, and 1,013 g
of adsorbent respectively. Our results indicated that basing %
recycle recovery on cycle bed productivity can enable efficient hydro-
gen recovery with savings on adsorbent amount. An optimum cycle
bed productivity of 2.3 mg H2/g of adsorbent corresponded to a cycle
recovery of 66.2% for 2,300 g of adsorbent used. Only 1.7 mg H2/g of
adsorbent was obtained for a cycle recovery of 72.2% requiring up to
2,800 g of adsorbent. This makes economic sense in the pressure
swing adsorption separation of hydrogen from traditionally low
hydrogen concentration biomass sources.
Keywords breakthrough curves; cycle bed productivity; cycle
mass balance; % cycle recovery; low concentration;
pressure swing adsorption; pyrolysis
INTRODUCTION
As an energy carrier, hydrogen is a clean fuel with a
higher heating value(HHV) of 142 MJ=Kg at standard
temperature and pressure versus gasoline whose HHV is
44 MJ=Kg (Mahfud et al. 2007). Hydrogen can therefore
contribute as an alternative renewable energy carrier in
meeting the increasing energy needs of a world faced with
limited fossil fuel supply.
Hydrogen is used in several applications such as in the
industrial production of ammonia, catalytic hydrocracking,
hydrogenation unit operations in refineries and petrochem-
icals industries. Hydrogen has currently also witnessed
increased applications in fuel cells (1,5). The demand for
pure hydrogen will continue to skyrocket as new paradigms
Received 31 July 2011; accepted 5 January 2012.
Address correspondence to Eliasu Teiseh, Texas A&M Univer-
sity, College Station, TX, USA. E-mail: eliasy1@yahoo.com
1522
such as the Hydrogen Economy continue to favor the
development of alternative energy sources or carriers. The
production and separation of pure hydrogen from synthesis
gas therefore becomes of importance. PSA separation tech-
nology recovers the desired product based on the differ-
ences in the equilibrium physisorption interactions of
adsorbate species with the adsorbents at higher separation
pressures. The more strongly adsorbed adsorbates consti-
tute the impurities or ‘‘rejects’’ which are released with a
swing in the pressure (12,17).
Cryogenics and membrane technologies can also be used
to recover useful gases from impure mixtures. While cryo-
genic techniques are highly energy intensive, membrane
approaches have their drawbacks in the pressure drops
across membranes and concentration polarization
especially for high concentrations of the desired species in
the feed (16). This drop in pressure reduces the selectivities
of membrane based separation systems. PSA offers the
advantages of reduced net energy consumption, no pressure
drop experienced with membranes, use of product gases to
build the needed pressure in equalization steps and the
flexibility provided by the availability of a wide variety of
adsorbents (6,10,16).
Single adsorbent beds are widely used in the separation
of gases from binary mixtures such as hydrogen from H2=
CO and H2=CO2, methane from CH4=N2, and oxygen from
O2=N2 in which the gases have wide differences in selectiv-
ities (5,9,11,14,15). Purities as high as 99.99% and recoveries
as high as 67.5% have been reported for one-column PSA
systems (5,9,16). Sorption isotherms have been used
to determine the selectivities of different adsorbents for dif-
ferent pure gases (2,3). For a single adsorbent sorption
isotherms have not been established for mixed multiple-
component gas mixtures.
When separation involves a multiple-component gas
mixture the possibility for a narrow difference in selectiv-
ities between the gases might dictate the use of composite
adsorbent beds especially for single bed columns with
reduced % recoveries. A composite bed of activated carbon
and molecular sieve 5 Å have been used to purify hydrogen
 EFFICIENT CYCLE RECOVERY OF HYDROGEN 1523

from the steam reforming of methane (SRM) off-gas TABLE 2.1

with the composition, H2 (72.2 v%), CO2 (21.6 v%), CO Physical properties of adsorbents – manufacturers’ report
(2.03 v%), CH4 (4.17 v%), (1,4). This improves selectivity
and eliminates the inherent conflict between the effects of Activated
equilibrium and kinetics based separation mechanisms for 5 Å carbon
gases with similar critical diameters such as carbon monox- Properties (1=16’’ pellets) (granular) Unit
ide and methane by splitting the adsorption of each Specific surface area – 1200 m2=g
impurity between the two adsorbents (17). Because of the Bulk density 0.71 0.47 g=cc
size of hydrogen and its relatively lower interaction with a Moisture content 1.0 2.0 wt%
wide variety of adsorbents it is usually easier to purify min. CCl4 activity – 60 –
hydrogen from multiple-component gas mixtures (5). Load- Ash content – 3.0 wt%
ing adsorbent supports such as activated carbon with tran- Particle diameter 0.16 – cm
sition metals has made it possible to also selectively remove Water capacity 21 – wt%
carbon monoxide from other gases such as methane and
Downloaded by [Fondren Library, Rice University ] at 15:59 12 November 2014

nitrogen (7,13).
The size of hydrogen notwithstanding, there is always the
requirement for its concentration in the feed stream to be
60 þ v% because of losses to adsorbents especially in one-
or two-column PSA setups, and because of the use of the
hydrogen in the purge streamto provide pressure in equali-
zation steps in multiple bed systems (12). SRM meets this
requirement since the H2=CO mole ratio is 3:1. Even when
this is not met(as is the case in coal gasification) the water-
gas-shift reaction can be used to enhance the amount of
hydrogen in the off-gas to meet the pressure requirements.
Pyrolysis of sludge biomass at high temperatures under
anaerobic environments produces synthesis gas with almost
no carbon dioxide in the off-gas stream. Using PSA to sep-
arate hydrogen from an optimum pyrolysis synthesis gas
composition could save costs in a downstream hydrogen-
ation of a mixture of oxygenates to produce gasoline range
hydrocarbons. In this work, we use breakthrough curves
and cycle mass balances for different bed lengths and there-
fore different adsorbent weights to show that the cycle
recovery for such low concentration feeds could be enhanced
by selecting the bed height or amount of adsorbent that gives
optimum cycle bed productivity (CBP). We base our find-
ings on experimental data exclusively since stringent indus-
trial standards still require that experimental data be used
to validate PSA model-based findings (6).
setup used to do studies on impurities breakthroughs, cycle
mass balances, and CBP during runs with continuous
hydrogen withdrawal.
Synthesis gas (with its components in their approximate
highest concentrations) obtained from the pyrolysis of the
MixAlco process sludge was used as the feed gas with the com-
position (v=v) of 45% H2=35% CO=20% CH4. The feed was
delivered from a tank into an adsorption separation column
holding a dual bed of adsorbents, namely activated carbon
(at the bottom) and molecular sieve 5 Å (at the top). The feed
gas went through a pressure regulator, a pressure gauge, and a
digital mass flow controller before entering at the bottom of
the adsorbents bed. At the top of the bed, another pressure
gauge was installed to measure potential pressure drop.
To maintain the bed pressure, the exit gas flow was
controlled by a digital hydrogen mass flow controller. The
feed gas pressure was regulated using a pressure regulator
supplied by McMaster-Carr (Aurora, Ohio). The effluent
gas pressure and the pressure dynamic of the bed were read
from digital pressure gauges supplied by Cole-Parmer
(Vernon Hills, Illinois). The feed flow was then controlled
by a programmable digital mass flow controller supplied
by Alicat Scientific (Tucson, Arizona). Downstream the
column, a ball valve, the same mass flow controller, and
pressure gauge (in make and manufacturer) were installed.

EXPERIMENTAL
Description of Setup and Experimental Approach
A single column pressure swing adsorption unit was used
to simulate the performance of a two-bed PSA system.The
adsorbents used were activated carbon and molecular sieve
5 Å.The activated carbon, a 4  8 mesh size coca 60 was
supplied by Activated Carbon Corporation (Compton,
California). The molecular sieve was supplied by UOP
LLC (Des Plaines, Illinois).The column was a 5.0 cm inter-
nal diameter stainless steel pipe. The manufacturer reported
physical properties of the adsorbents were as shown on
Table 2.1. Figure 2.1. is a schematic of the experimental FIG. 2.1. Process scheme for a single-column PSA system.
 1524 E. TEISEH AND S. CAPAREDA
Separation and recovery of hydrogen from impurity
carbon monoxide and methane was through a series of cycle
steps such as pressurization with feed (with no H2 discharged),
pressurization with feed (with H2 discharged), co-current
depressurization, counter-current depressurization, and exter-
nal purge with hydrogen described in detail in an appropriate
section of the paper. Experiments to determine cycle recov-
eries from cycle bed productivities for three different adsorb-
ent amounts were preceded by preliminary experiments to test
separation effectiveness of selected adsorbent pair.
Preliminary Experimental Runs
Syngas feedflow rate was studied at two levels-7 and 10
standard lpm while the pressure was studied at three
Downloaded by [Fondren Library, Rice University ] at 15:59 12 November 2014
levels-310, 610, and 710 kPa in 15 experimental runs. A
200 cm stainless steel column was first filled to 120 cm with
1,585 g of activated carbon, and then topped with about
1,307 g of molecular sieve 5 Å. The adsorption column was
pressurized until a predetermined pressure was reached. At
that pressure, sampling port 9-1 was opened and the effluent
analyzed using an SRI 8610C GC supplied by SRI Instru-
ments (Torrance, California). A random factorial statistical
design using the Design Expert software was performed for
all combinations of feed rate and pressure to study the
impact of each input parameter on traditional outputpara-
meters such as purity (%), recovery (%), and hydrogen dis-
charge rate (standard lpm). Experiments were performed
in triplicates for each combination of feed flow and pressure-
with the intention of selecting the feed rate and pressure that
gave the best hydrogen purity and per-pass recovery for use
in actual experiments meant to efficiently recover hydrogen
from the syngas with the specified composition.The Design
Expert software was used to analyze the data obtained.
It appeared that both high pressure and feed flow rate
improved the percent recovery and hydrogen discharge.
A higher bed pressure of 809 kPa and 10 standard lpm
were thereforeusedin more investigative experiments that
aimed at determining the cycle recovery from cycle bed
productivity determinations.
Investigative Experiments on Three Different Adsorbent
Amounts at 809 kPa and 10 Standard lpm
Because of the positive influence of both the pressure and
the feed rate on the discharge rate and the % recovery, a higher
pressure of 809 kPa and a mass flow controller full range feed
rate of 10 standard lpm were therefore selected to conduct
separations in continuous run mode for the purposes of deter-
mining impurity breakthrough times (from breakthrough
curves), cycle mass balances, and bed productivities at a con-
stant hydrogen draw rate of
4.5 standard lpm.This draw rate
was selected because it approximately balanced the rate at
which hydrogen was fed into the bed, giving a constant press-
ure dynamic condition. A total of 9 experiments, in triplicates
(to determine breakthrough times for CO and CH4) were then
performed using 200 cm, 135 cm, and 70 cm columns each
containing 2,892 g, 1,962 g, and 1,013 g of adsorbent, respect-
ively, with a feed rate of 10 standard lpm and 809 kPa
pressure. Another 9 experiments, in triplicates, were per-
formed at the same condition of pressure, flow rate, and
adsorbent amounts to carry out cycle mass balances and hence
determine cycle recovery and cycle bed productivities but with
the times for pressurization (with hydrogen discharge) being
less than the breakthrough times to prevent the Co and CH4
impurity front from reaching the top of the column.
Cycle Description and Breakthrough Times (from
Breakthrough Curves)
In order to characterize pressure drops in dynamic sys-
tems Sundaram and Wankat (1988) combined Darcy’s
law with the continuity equation and arrived at a dimen-
sionless characteristic constant f, which varies inversely
with the bed length. Their rule of thumb was that for values
of f greater than 0.5, the pressure drops can be assumed
insignificant. PSA beds can be run under dynamic or static
conditions, and the selection of f may become relevant in
dynamic systems. For f < 0.5, a certain bed height results
in a significant pressure drop that limits impurities adsorp-
tion which a direct impact on the breakthrough times.
The value of f was not specifically determined although
experiments were performed under dynamic conditions. To
account for this, breakthrough times were determined for
each adsorbent bed height and cycle mass balances were
conducted for pressurization steps with durations less than
the breakthrough times. Large adsorbent particle sizes
minimized bulk mass transfer limitations hence eliminating
pressure drop across the beds. A robust dynamic pressur-
ized bed was obtained with an inlet pressure of 827 kPa
and a nearly constant bed pressure of 809 kPa.
The effluent leaving port 9-2 was analyzed at specific
time intervals. With all pressure units in kPa, the cycle
steps involved were:
1. a pressurization with feed syngas (with no hydrogen
leaving bed) from atmospheric pressure to a pressure
of 809 kPa.
2. a continuous pressurization at 809 kPa with feed (as
hydrogen was discharged at the top of the column).
3. a co-currrent depressurization (from 809 kPa to 379,
448, and 482 kPa respectively for 200, 135, and 70 cm
beds). During this step, there was no feed stream but
hydrogen was released at the top of the column. The
hydrogen from steps 2 and 3 were collected into the stor-
age tank and later wasted.
4. a counter-current depressurization step further decreased
the bed pressure to 101 kPa.
5. pure hydrogen was delivered counter-currently (through
valve 5-3) from the hydrogen purge tank and used to
pressurize the bed from 101 kPa to 310 kPa.
 EFFICIENT CYCLE RECOVERY OF HYDROGEN 1525

6. the bed pressure was finally reduced to 101 kPa by open- concentration ðv%Þ of
ing valve 5-2. The effluents of steps 4 and 6 were wasted hydrogen in the product
by directing to the outside through the fume hood. stream at sampling time
purity ð%Þ ¼  100
The overall cycle times for the 200, 135, and 70 cm col- sum of the individual
umns each holding a total of 2,892 g, 1,962 g, and 1,013 g of concentrations ðv%Þ of species
adsorbent, respectively, were 23.8, 16.3, and 7.8 minutes, in the product stream
respectively, and step 2, continuous pressurization at
ð2:2Þ
809 kPa took place for a total time that was less than the
time at which the impurities (CO and CH4) broke through
In order to perform cycle mass balances for the respective
to prevent the impurity front from reaching the top of the
bed heights or adsorbent amounts, the beds were pressurized
column. The time in steps 1 and 2 were combined to get the
to 809 kPa at a feed rate of 10 standard lpm. This was followed
time during which hydrogen was fed into the columns. A
by a co-current pressurization but with hydrogen released.
detailed cycle step data are shown in Table 2.2.
The combined time for these pressurization steps was useful
Downloaded by [Fondren Library, Rice University ] at 15:59 12 November 2014

in calculating the hydrogen ‘‘mass in.’’ The hydrogen ‘‘mass

Parameters Estimation
The % per-pass recovery and purity were obtained as
defined in Eqs. (2.1) and (2.2). This definition of recovery
is in accord with the traditional definition of product recov-
ery where the flow rate of the raffinate product is expressed
as a percentage of the flow of the feed stream for the desired
species. The hydrogen discharge rate was characterized by
the numerator of Eq. (2.1) (17). Since this rate increases
with increase in the bed pressure, the value at the maximum
pressure was the maximum hydrogen discharge rate.
max hydrogen flow rate ðmg=sÞ
in products tream
recovery ð%Þ ¼  100
total feed flow ðmg=sÞ  hydrogen
feed concentration ðv%Þ
ð2:1Þ
TABLE 2.2
Cycle steps, times (minutes) and pressure changes (kPa) for
the three bed lengths
out’’ was obtained from the hydrogen discharge rate and
the time in steps 2 and 3. Figure 3.3 is a plot of the hydrogen
purity against time, with the actual volumetric flow rate (or
discharge rate) being the product of constant discharge rate
of 4.5 standard lpm and the percent purity of hydrogen.
To get the actual hydrogen mass in or out, a plot of
hydrogen flow rates versus time was obtained (see
Fig. 3.4), and the Matlab software was used to calculate
the total area by the Trapezoidal Rule.This was numerically
equal to the volume of the hydrogen discharged over there-
spective times in the relevant cycle steps. The product of the
density of hydrogen at standard conditions and the dis-
charge rate gave the mass flow rate of hydrogen out. With
focus on the amount of hydrogen ‘‘caught,’’ the hydrogen
adsorption or accumulation on the adsorbent pores was
considered as losses and ignored largely because of the dif-
ficulties associated with measuring with accuracy the flows
of gas mixtures with varying compositions. The terms in
the mass balance are shown in Eq. (2.3) while the definitions
of cycle recovery (%) and cycle productivity are shown in
Eqs. (2.4) and (2.5), respectively. The error margins per
cycle of boththe mean mass of hydrogen, and of the recov-
ery (%) for each bed height were within 4.0%.

Column Time cycle mass of hydrogen out (g) ¼ cycle mass of hydrogen

height (cm) 1 2 3 4 5 6 7 B in (g)  cycle mass of hydrogen lost to the adsorbents (g)
200 7.7 12.0 3.3 0.2 0.3 0.3 23.8 15.0 ð2:3aÞ
135 4.6 9.0 2.2 0.2 0.2 0.2 16.3 12.3
70 2.4 4.0 1.0 0.2 0.1 0.1 7.8 5.4 Zt Zt
qQðoutÞ pdt ¼ /qQðinÞ dt
Pressure change
0 0
200 101–809 809–379 379–101 101–310 310–101  cycle mass of hydrogenlost ð2:3bÞ
135 809–448 448–101
70 809–482 482–101 mass of hydrogen
1 ¼ feed pressurization (no discharge); 2 ¼ co-current pressuri- caught per cycle
zation (with discharge); 3 ¼ co-current depressurization; 4 ¼ % cycle recovery ¼  100 ð2:4Þ
mass of hydrogen fed
counter-current depressurization; 5 ¼ H2 pressurization; 6 ¼

purge; 7 ¼ cycle time2 ¼ breakthrough time (CO and CH4). per cycle
 1526 E. TEISEH AND S. CAPAREDA


mg
cycle productivity
g
ð2:5Þ
mass of hydrogen caught per cycle
¼
combined weight of adsorbents

RESULTS AND DISCUSSIONS

Preliminary Experiments on 2,892 g of Adsorbent
The Design Expert generated summary results for all
combinations of input parametersduring preliminary
screening experiments for the 200 cm bed are shown in
Table 3.1. The mean values of the three parameters were
as follows: % purity (99.9), % recovery (55), and hydrogen
discharge rate(2.1). This means that a little over half of
Downloaded by [Fondren Library, Rice University ] at 15:59 12 November 2014

the hydrogen in the feed stream will be recovered as the pro-

duct with the balance lost to the adsorbent matrix. This
translates to a mean hydrogen discharge rate of 2.1standard
lpm across the range of pressures and feed rates studied.
The productivity increased with bed pressure and feed rate
from 35.2  4.4 at 310 kPa and 7 standard lpm to
1270.2  35.4 standard lpm=kg of adsorbent=minute at
710 kPa and 10 standard lpm. The results showed that the
composite adsorbent bed can be used to separate hydrogen
from a low hydrogen concentration pyrolysis gas stream
with hydrogen, methane and carbon monoxide at their opti-
mum production compositions of 45 v% H2, 35 v% CO and
20 v% CH4. The variations in the output parameters inves-
tigated with changes in the input parameters are shown in
Figure 3.1. Since 15 out of 18 factorial experiments were
FIG. 3.1. Changes in % recovery and throughput with changes in feed
performed (excluding the triplicate from the condition of flow rate (slpm) and bed pressure (kPa) – top [% recovery]; bottom
7 standard lpm and 710 kPa), the point on Fig. 3.1 for this [throughput (slpm) of hydrogen produced]. (Color figure available
condition is a poorly predicted non-experimental point and online)
was considered an outlier. This is evident by the fact that
unlike other points, it lacks an error bar.
The two factors studied, the interaction between them Yang et al. (2009) studied the dynamics of bed recovery
and the generated model, were statistically significant using superficial velocity, rather than feed flow. Arriving at
(p-value << 0.05). For a given feed rate, the percent recov- a different conclusion, they explained their fall in recov-
ery and discharge rate increased with increase in bed press- ery with increase in superficial velocity by arguing that
ure, but the increase in recovery seemed to be more at there is a broadening of the mass transfer zone with
higher feed rates compared to lower feed rates for a given increase in feed rate or superficial velocity. For a given
bed pressure. This could be explained by the definition of pressure, increasing the feed flow (superficial velocity)
% recovery and the fact that pressure is a stronger recovery reduces the residence time of the least adsorbed H2 thereby,
factor as can be seen from the bars of the top part of reducing the impact of its co-adsorption on the adsorbent.
Fig. 3.1. The bar heights represent predicted mean values Higher pressures means an increase in the concentrations
from a software generated best fit-model. of CO and CH4 each having a higher affinity for the

TABLE 3.1
Summary results for all preliminary runs (200 cm – software generated)
Response Name Unit Analysis Min. Max. Mean STD Model
Y1 Purity % Factorial 99.9 99.9 99.9 0 2 factor
Y2 Recovery % Factorial 22.7 85.6 55.0 25.3 2 factor
Y3 Discharge Slpm Factorial 1.0 3.9 2.1 2.1 2 factor
 EFFICIENT CYCLE RECOVERY OF HYDROGEN 1527

adsorbent. Both compete more favorably with H2 as the

number of adsorption sites reduces. Hence an increase in
the amount of enriched H2 leaving at the top of the bed.-
Our result of an increase in recovery at higher pressure
and low flow could be explained by the favorable combi-
nation of high pressure and lower viscosities in the mass
flux regime between that modeled by Knudsen diffusion
and bulk molecular diffusion [Poisseuille Eq. (3.3)].

Breakthrough Curves and Breakthrough Times

The effluent concentrations ofhydrogen, methane, and
carbon monoxide for the three bed lengths or adsorbent
amounts are shown in Fig. 3.2. Figure 3.3 shows the hydro-
gen purity time profile for the three column heights each
Downloaded by [Fondren Library, Rice University ] at 15:59 12 November 2014

holding a different amount of the two adsorbents investi-

gated. The breakthrough times as deduced from the
product gas composition curves are shown as parts of
Table 2.2 and Table 3.2. It should be noted here that, the
concentrations (v%) of the three gases (together with those
of O2 and N2 carried over in the sampling bag during sam-
pling) added up to 100%. In PSA beds, the dynamics of
kinetically and thermodynamically controlled physisorp-
tion interactions of the adsorbates with the adsorbent sur-
face creates three distinct zones along the length of the
adsorption bed. When the fresh feed gets into the bed,
the impurity front generated under pressure is largely a
saturated zone. Further down the middle of the bed, a con-
centration gradient between the gas phase inside the pores
and the pore surfaces creates a mass transfer zone. The bed
terminates with a fresh zone of unsaturated adsorbent at
the top. To enhance product purity in PSA separation,
the impurity front must be pushed to the foot of the bed
in one cycle step (counter-current depressurization, blow
down, or hydrogen purge steps).
In terms of critical diameter the ranking for methane,
carbon monoxide, and hydrogen is as follows: metha-
ne>carbon monoxide>hydrogen. While the pore structure
of activated carbon is unimodal, that of the molecular sieve
(5 Å) is bimodal with microspores interconnected by
macropores. The size of hydrogen and to some extent that
of carbon monoxide permit their mass transfer flux to be
FIG. 3.2. Concentrations of product gas streams against time of run at
largely dominated by bothbulk molecular and Knudsen
809 kPa for a feed rate of 10 slpm: a) 2,892 g, b) 1,962 and c) 1,013 g of
diffusions (through activated carbon and 5 Å molecular adsorbent used.
sieve) according to Eqs. (3.1) and (3.2) (17).
Knudsen diffusion through the macropores especially at
T1:7 lower pressures. The flux of each adsorbate moves at a dif-
Dm / pffiffiffiffiffi ð3:1Þ
P M ferent rate (and hence breakthrough at different times) due
pffiffiffiffiffiffiffiffiffiffiffi to differences in molecular weight. In the intermediate
DK / r T=M ð3:2Þ region, mass flux in PSA is governed by the equivalent
Poiseuille diffusivity that depends largely on pressure, since
Recovery of hydrogen at lower pressures is largely due the influence of viscosity is undermined by the minimal dif-
to faster bulk molecular diffusion through both adsor- ferences in the viscosities of the gases under identical press-
bents. Larger molecules such as CO and CH4 move by ure and temperature conditions. Hence improved recovery
 1528 E. TEISEH AND S. CAPAREDA

TABLE 3.2
Performance data for the 3 adsorbent amounts (10 slpm and pressure of 809 kPa)
Bed height Adsorbent height Adsorbent Adsorbent H2 mass H2 mass Cycle Impurity breakthrough
(cm) in bed (cm) type weight (g) in (g) out (g) recovery (%) time (min.)
200 120 AC 1585 7.9 5.7 72.2 14.95  2.1
80 5 Å 1307
135 73 AC 1067 6.7 4.4 65.4 12.35  1.0
62 5 Å 896
70 38 AC 555 2.6 1.6 60.2 5.4  1.3
32 5 Å 458

is obtained at higher pressure as illustrated in Poiseuille separation and the %cycle recovery. According to
Downloaded by [Fondren Library, Rice University ] at 15:59 12 November 2014

Eq. (17) Table 3.2, decreasing the adsorbent amount from 2,893
to 1,963 g caused the recovery to dive from 72.2% to
Pr2 65.4%. A further decrease of the adsorbent amount to
D¼ ð3:3Þ
8l 1,013 g caused the recovery to fall to about 60.0%. This
The persistent micropores in unimodal adsorbents such could be explained by a reduction in the retention time
as activated carbon and in bimodal heterogeneous zeolites for CO and CH4 removal from the gas phase with decreas-
such as 5 Å create a medium for micropore diffusion for each ing overall bed volume. An increase in retention time has
adsorbent. In micropores mass transfer is largely dependent been shown to boost % recovery, Yang et al. (2009) (4).
on the concentration gradient and in steric hindrances or The areas under the graphs in Figure 3.4. represent the
activation energy barriers that can be created due to differ- mass of hydrogen recovered per cycle. This area increased
entials in molecular kinetic diameters. For a given micropore
channel, only species with a given critical diameter (less than
the diameter of the micropore) will move across and become
separated because they experience low steric hindrances.
Those with larger critical diameters will experience higher
steric hindrance, requiring a higher activation energy. The
channels of zeolites therefore kinetically and selectively sieve
out only smaller critical diameter molecules.

Hydrogen Recoveries and Cycle Bed Productivities

From the curves in Figs. 3.2 and 3.3, we find that the
smaller the adsorbent amount, the shorter the break-
through time (tbr) for CO and CH4 for a given flow. Hence
the greater the amount of adsorbent used, the better the

FIG. 3.4. Cycle hydrogen out versus separation time: a) 200 and 135 cm,
FIG. 3.3. Hydrogen purity versus run. b) 70 cm.
 EFFICIENT CYCLE RECOVERY OF HYDROGEN
with an increase in the bed length or amount of adsorbent
used, thereby corresponding to an increase in the total
mass of hydrogen discharged at the end of the cycle.
Figure 3.5. shows that although the cycle recovery seems
to increase with an increase in bed height and hence the
amount of adsorbent needed, the cycle bed productivity
(CBP) maintained that trend only to a certain adsorbent
amount beyond which it (CBP) peaked, and then decreased
with increasing adsorbent amount. Our experimental data
fits a quadratic polynomial neatly. A maximum cycle bed
productivity of 2.25 mg H2=g adsorbent was achieved using

2,300 g of adsorbent. This CBP value corresponded to

66.4% cycle recovery lower than the 72.2% for
2,800 g
of adsorbent. A high cycle recovery of 72.2% only leads to
Downloaded by [Fondren Library, Rice University ] at 15:59 12 November 2014
about 1.7 g of hydrogen per gram of adsorbent compared
to 2.3 g of hydrogen per gram of adsorbent obtained at a
cycle recovery of 66.2%. Hence assessing % recovery via
CBP can be a more effective way of measuring the cycle
recovery especially for low concentration syngas sources
such as biomass sludge. This is because the pyrolysis of bio-
mass for syngas production produces a relatively lower
amount of hydrogen versus that obtained from the indus-
trial steam reforming of methane.It would thus appear that
not basing the % recovery on CBP might result in the recov-
ery being arelativelyweak evaluation metric. The CBP gives
a recovery with a potential economic connotation and there-
fore should be used in assessing PSA bed performances in
multi-bed PSA systems. This is because it directly relates
to the amount of adsorbent, and hence the total cost asso-
ciated with product recovery especially for low hydrogen
concentration feed streams.
Integrating Pyrolysis and PSA for Optimum Hydrogen
Recovery
The product gas is usually needed to provide the needed
pressure in the next pressurization step in vessels that
FIG. 3.5. Effect of column height on PSA bed productivity.
1529
FIG. 3.6. Recovering hydrogen from a low concentration pyrolysis
process using a PSA unit. (Color figure available online)
previously underwent a blowdown in PSA systems with a
cascade of vessels in series.It is for this reason that one of
the requirements is that the product species to be captured
should be in a higher concentration in the feed streams,
preferably in excess of 65 v% (12). The primordial PSA sys-
tem was the Skarstrom system with two beds alternately
undergoing pressurization and depressurization in a single
cycle (17). Many modifications to this system have resulted
in the connection of multiple beds in series to improve per-
cent recovery and purity. While this is difficult with low
concentrations of the desired species in the feed stream,
such multiple cascading of the beds results only in a slight
increase in the percent purity although an increase in recov-
ery of about 50% has been reported (8).
Figure 3.6 shows how combining a PSA system with opti-
mum bed productivity could be used to harness hydrogen
produced from relatively lower hydrogen concentration
sources such as ones derived from the pyrolysis of biomass
sludge. This will lead to an improved recovery with yet a
minimum number of beds versus the multiple beds used for
product recovery in PSA systems with the attendant costs.
Our pyrolysis experiments showed that for a ton of
sludge from the MixAlco process that is pyrolyzed, up to
5,990 kg of hydrogen could be obtained in the syngas with
constituents [(H2, CO) and CH4] at approximate maximum
concentrations used as the feed in this separation study.
Applying 66.2% recovery means up to 3,964 g of hydrogen
will be recovered with the CO and CH4 recycled.
CONCLUSIONS
The key findings of this work can be captured in the
following points:
 An increase in the percent cycle recovery with
increasing amount of adsorbent does not necessar-
ily translate into an increase in the overall process
CBP.
 1530 E. TEISEH AND S. CAPAREDA
 CBP, a stronger evaluation criterion increases with
adsorbent amount but soon reaches diminishing
returns. Hence there is a critical amount of
adsorbent for some bed size or height that gives
optimum CBP.
 Low hydrogen concentration synthesis gas
streams from the pyrolysis of biomass under opti-
mum conditions could be recovered with a 99.9%
purity in a single-bed PSA system.
ACKNOWLEDGEMENTS
Funding for this research was jointly provided by the
following institutions to which we sincerely express our
appreciation: the US Department of Energy, the US
Downloaded by [Fondren Library, Rice University ] at 15:59 12 November 2014
Department of Defense, the DARPA Program, and the
Texas A& M Agrilife and Research.
My sincere thanks also go to Dr. M. El-Halwagi of the
Department of Chemical Engineering, Texas A & M
University for providing the needed orientation to carry
out this work.
NOMENCLATURE
q: density of hydrogen at standard conditions (g=L).
ø: volumetric fraction of hydrogen in the feed stream
(v=v%).
Q: volumetric flow rate of the feed (standard lpm)
p: hydrogen purity (%)
f: dimensionless parameter relating axial distance, bed
length and pressure drop.
Dm: molecular diffusivity (m2=s)
DK: Knudsen diffusivity (m2=s)
T: temperature (k)
M: molecular weight (g)
P: absolute pressure in the column (kPa)
r: radial coordinate in microparticle
D: diffusivity (Poiseuille) (m2=s)
l: viscosity (gcm1s1)
REFERENCES
1. Yang, S.-I.; Choi, D.-Y.; Jang, S.-C. (2008) Hydrogen separation by
multi-bed pressure swing adsorption of synthesis gas. Adsorption,
14: 583–590.
2. Nam, G.-M.; Jeong, B.-M.; Kang, H.-S.; Lee, K.-B.; Choi, D.-K.
(2005) Equilibrium isotherms of CH4, C2H6,C2H4,N2 and H2 on
zeolite 5A using a static volumetric method. J. Chem. Eng. Data,
50: 72–76.
3. Jeong, B.-M.; Ahn, E.-S.; Yun, J.-H.; Lee, C.-H.; Choi, D.-K. (2007)
Ternary adsorption equilibrium of H2=CH4=C2H4 onto activated
carbon. Separation and Purification Technology, 55: 335–342.
4. Yang, S.-I.; Park, J.-Y.; Choi, D.-K.; Kim, S.-H. (2009) Effects of
the residence time in four-bed pressure swing adsorption process.
Separation Science and Technology, 44: 1023–1044
5. Yang, J.; Han, S.; Cho, C.; Lee, C.-H.; Lee, H. (1995) Bulk separation
of hydrogen mixtures by one-column PSA process. Separations
Technology, 5: 239–249.
6. Sircar, S. (2002) Pressure swing adsorption. Ind. Eng. Chem. Res., 41:
1389–1382.
7. Iyuke, S.E.; Mohamad, A.B.; Daud, W.R.W.; Kadhum, A.A.H.
(2000) Removal of CO from process gas with Sn-activated carbon
in pressure swing adsorption. Journal of ChemicalTechnology and
Biotechnology, 75: 803–811.
8. Nikolic, D.; Giovanoglou, A.; Georgiadis, M.C.; Kikinides, E.S.
Modelling and simulation of multi-bed pressure swing adsorption
processes. 17th European symposium on computer aided process
engineering-ESCAPE 17. Edited by Plesu V and Agachi P.S.
9. Teague, Jr., K.G.; Edgar, T.F. (1999) Predictive dynamic model of a
small pressure swing adsorption air separation unit. Ind. Eng. Chem.
Res., 38: 3761–3775.
10. Sircar, S.; Golden, T.C.; Rao, M.B. (1996) Activated carbon for gas
separation and storage. Carbon, 34 (1): 1–12.
11. Fatehi, A.I.; Loughlin, K.F.; Hassan, M.M. (1995) Separation
of methane-nitrogen mixtures by pressure swing adsorption using a
carbon molecular sieve. Gas. Sep. Purif., 9 (3): 199–204.
12. Stöcker, J.; Wysall, M.; Antwerp; Miller, G. Q. 30 years of PSA tech-
nology for hydrogen purification. Not peer reviewed.
13. Tamon, H.; Kitamura, K.; Okazaki, M. (1996) Adsorption of carbon
monoxide on activated carbon impregnated with metal halide. AIchE
Journal, 42: 2.
14. Richards, J.A.; Watanabe, K.; Austin, N.; Stapleton, M.R. (1995)
Computer simulation of gas separation properties of zeolite Li-X.
Journal of Porous Materials, 2: 43–49.
15. Zheng, D.-X.; Xie, Z.-J.; Fan, Z. (1988) Prediction of breakthrough
curves of oxygen-nitrogen coadsorption system on molecular sieves.
Gas Separation & Purification, vol. 2 Dec.
16. Miyajima, H.; Kodama, A.; Goto, M.; Hirose, T. (2005) Improved
purge step in pressure swing adsorption for CO purification. Adsorp-
tion, 11: 625–630.
17. Douglas, M.R.; Shamsuzzaman, F.; Kent, S.K. (1994) Pressure Swing
Adsorption, 1st Ed.; VCH Publishers, Inc.
18. Sundaram, N.; Wankat, P.C. (1988) Pressure drop effects in the
pressurization and blowdown steps of pressure swing adsorption.
Chem. Eng. Sci., 43: 123.
19. Mahfuh, F.H.; Bussemaker, S.; Kooi, B.J.; Ten Brink, G.H.; Heeres,
H.J. (2007) The application of water soluble ruthenium catalysts for
the hydrogenation of dichloromethane soluble fraction of fast pyrol-
ysis oil and related model compounds in a two-phase aqueous-organic
system. Journal of Molecular Catalysis A: Chemical, 277: 127–136.