processes

Article
Modeling and Optimization of High-Performance
Polymer Membrane Reactor Systems for Water–Gas
Shift Reaction Applications
Andrew J. Radcliffe 1 , Rajinder P. Singh 2 , Kathryn A. Berchtold 2 and Fernando V. Lima 1, *
1 Department of Chemical Engineering, West Virginia University, Morgantown, WV 26506, USA;
[email protected]
2 Carbon Capture and Separations for Energy Applications, Materials Physics and Applications Division,
Los Alamos National Laboratory, Los Alamos, NM 87545, USA; [email protected] (R.P.S.);
[email protected] (K.A.B.)
* Correspondence: [email protected]; Tel.: +1-304-293-2353; Fax: +1-304-293-4139

Academic Editor: Masoud Soroush

Received: 2 December 2015; Accepted: 21 March 2016; Published: 1 April 2016

Abstract: In production of electricity from coal, integrated gasification combined cycle plants
typically operate with conventional packed bed reactors for the water-gas shift reaction, and a
Selexol process for carbon dioxide removal. Implementation of membrane reactors in place of these
two process units provides advantages such as increased carbon monoxide conversion, facilitated
CO2 removal/sequestration and process intensification. Proposed H2 -selective membranes for
these reactors are typically of palladium alloy or ceramic due to their outstanding gas separation
properties; however, on an industrial scale, the cost of such materials may become exorbitant.
High-performance polymeric membranes, such as polybenzimidazoles (PBIs), present themselves
as low-cost alternatives with gas separation properties suitable for use in such membrane reactors,
given their significant thermal and chemical stability. In this work, the performance of a class of
high-performance polymeric membranes is assessed for use in integrated gasification combined cycle
(IGCC) units operated with carbon capture, subject to constraints on equipment and process streams.
Several systems are considered for use with the polymeric membranes, including membrane reactors
and permeative stage reactors. Based upon models developed for each configuration, constrained
optimization problems are formulated which seek to more efficiently employ membrane surface area.
From the optimization results, the limiting membrane parameter for achieving all carbon capture
and H2 production specifications for water–gas shift reactor applications is determined to be the
selectivity, αH2 {CO2 , and thus a minimum value of this parameter which satisfies all the constraints
is identified for each analyzed configuration. For a CO2 capture value of 90%, this value is found
to be α = 61 for the membrane reactor and the 3-stage permeative stage reactor and α = 62 for the
2-stage permeative stage reactor. The proposed systems approach has the potential to be employed
to identify performance limitations associated with membrane materials to guide the development of
future polymeric and other advanced materials with desired membrane characteristics for energy
and environmental applications.

Keywords: polymer membranes; water-gas shift membrane reactors; optimization

1. Introduction and Prior Work

As the world transitions to a more environmentally conscious economy, the importance of
hydrogen (H2 ) production processes is paramount. Hydrocarbons such as petroleum, natural
gas, coal and biomass serve as the principal sources of H2 , which will see use as a feedstock in
myriad clean energy and chemical production processes. As H2 production from hydrocarbons

Processes 2016, 4, 8; doi:10.3390/pr4020008 www.mdpi.com/journal/processes

Processes 2016, 4, 8 2 of 19

generates carbon dioxide (CO2 ), processes incorporating carbon capture technologies are necessary
to achieve the objective of reduction of CO2 emissions in accordance with protocols that seek to
mitigate global climate change. Based upon extrapolation of the rates of consumption and available
reserves, projections posit that petroleum resources may be depleted within 50 years and natural
gas resources within 100; however, coal resources may exhibit their current availability for a couple
hundred years [1]. Consequently, emerging energy technologies that utilize coal as the feedstock, such
as integrated gasification combined cycle (IGCC) power plants operated with carbon capture, are
particularly promising.
Coal-based IGCC units produce electricity through a synthesis gas (syngas) intermediate, which
is subjected to the water-gas shift (WGS) reaction to maximize H2 produced prior to the stream
being sent to the gas turbine portion of the unit. An IGCC process scheme with carbon capture
typically utilizes packed-bed WGS reactors followed by CO2 removal by a Selexol process [2,3]. An
alternative to this method of syngas conversion utilizes membrane reactors (MRs) equipped with
H2 -selective membranes, which grant advantages such as increased carbon monoxide (CO) conversion,
facilitated CO2 removal/sequestration (CO2 -rich effluent is produced at high pressure), and process
intensification through a reduction to the total number of process units [2,4].
There are challenges inherent to the use of MRs for such an application as the H2 -selective
membranes must be stable under high-temperature and extreme pressure conditions in the presence
of water and contaminants such as hydrogen sulfide (H2 S). H2 -selective membranes commonly
considered for this application are as follows: (i) zeolite-based molecular sieves; (ii) dense metals such
as Pd; and (iii) polymeric membranes. Of these potential membrane materials, (i) and (ii) possess highly
favorable gas separation properties in terms of selectivity and flux, but the cost for these materials may
be prohibitive for industrial-scale application. Only some polymeric membranes can be considered for
the WGS application, as the elevated operating temperature of the MR unit is often outside the stability
limits of the membrane material or the membrane material exhibits limited gas separation properties at
the operating temperatures defined by the WGS-MR. However, if the aforementioned performance and
stability challenges are addressed, polymeric membranes possessing suitable gas separation properties
offer the potential to greatly reduce the cost of industrial-scale-MR implementation.
Polybenzimidazoles (PBIs) represent one such class of high performance polymers having
exceptional chemical and physical characteristics enabling H2 /CO2 separation in challenging
thermo-chemical environments. These materials exhibit molecular-sieving mechanisms analogous to
those observed in zeolite-based membranes, which imbues these materials with attractive H2 /CO2
selectivity for syngas separations. High-performance polymeric materials have also been found to
exhibit good thermal stability up to 400 ˝ C and chemical stability in the presence of common syngas
contaminants [5,6].
One objective of this study is to assess the feasibility of the state-of-the-art high-performance
polymeric materials for use in membrane reactor systems with respect to performance constraints set
forth by the U.S. Department of Energy (DOE) for pre-combustion CO conversion/CO2 separation
processes within IGCC units [3]. In this study, the performance characteristics of PBI-based membranes,
as demonstrated by Berchtold and coworkers in [5], are used to develop the benchmark case for the
polymer membrane-based MR process schemes investigated and developed herein. These PBI-based
membrane materials have demonstrated industrially attractive H2 /CO2 separation characteristics
including ideal H2 permeabilities between 58 and 78 barrer and H2 /CO2 selectivities between 23 and
43 at 250 ˝ C [5,7]. Additionally, this study seeks to determine the minimum membrane characteristics
needed to satisfy the DOE’s performance constraints by considering process models for several reactor
designs. The performance of the various reactors is assessed in the base case conformations, which are
then modified by considering different catalyst/membrane placement about the axial axis. Alternative
reactor designs are developed by seeking to maximize reactor performance (H2 recovery) for the
minimum reactor cost as determined by the required membrane surface area. As demonstrated
previously, an optimization problem is formulated to guide these designs [2].
Processes 2016, 4, 8 3 of 19

With regard to the systems analysis, there are several MR models (utilizing H2 -selective
membranes) related to the WGS reaction available in the literature, encompassing the range from
1-D/isothermal to 2-D/non-isothermal. Also available in the literature are MR models that employ
H2 -selective membranes relating to widely varied applications (see [2] for a summary for MR models,
efforts and applications). A review of literature shows a few computational modeling studies based on
membrane reactors employing polymeric membranes [8,9], due in part to the temperature limitations
imposed by available polymers. However, the literature suggests a lack of studies on optimization
of polymer-based MR configurations. Recent and continued development and demonstration of
high performance polymers such as PBIs for potential use in challenging membrane separation
environments, such as those encountered in the vicinity of the WGS reaction, presents an opportunity
to derive a MR model for a system utilizing such H2 -selective polymers and subsequently evaluate
their potential in this challenging separations role [5,7,10]. This study is focused on H2 -selective
membranes due to their advantages over CO2 -selective membranes in IGCC process schemes, as
discussed by [11].
Moreover, several optimization studies relating to packed-bed MRs and reactor systems
employing membrane separators are available in the literature. These studies have utilized H2 -selective
membranes (ceramic or Pd) to formulate optimization problems that examine staged membrane
reactors [12–14] and traditional MRs [2]. In the case of the staged membrane reactors, the optimization
problems were formulated with the objective of maximizing methane conversion, H2 recovery or
H2 yield in a steam methane reforming (SMR) process employing a Pd-based membrane. These
studies considered a permeative stage membrane reactor (PSMR) with a fixed number of stages, or a
staged membrane reactor (continuous membrane, catalyst packing with inert stages). The decision
variables were composed of the catalyst/membrane stage lengths, but the problem was not subject to
performance constraints. For the case of the traditional MR performing the WGS reaction, the study
formulated an optimization problem in terms of economic variables that maximizes performance (H2
recovery) for the minimum cost (membrane surface area) subject to multiple constraints on reactor
effluent streams by considering alternative catalyst/membrane placement about the axial axis of the
reactor. With regard to the available literature, it is worth noting that computational studies of SMR
or WGS processes that use Pd/micro-porous ceramic membranes have H2 selectivity values that are
comparatively larger than those of polymeric membranes.
Thus, the computational study performed here of MR systems employing novel polymeric
membrane materials provides insight into their feasibility for WGS reaction applications. Additionally,
such a study may be used to identify performance limitations associated with the material, which may
be used to guide the development of future polymeric materials with desired membrane characteristics.
To this end, mathematical models are developed for traditional MRs and PSMRs using the performance
characteristics of PBI membranes; these models are subsequently employed to develop reactor designs
that satisfy the set of performance constraints set forth by the U.S. DOE for pre-combustion CO
conversion (WGS reaction)/CO2 separation processes within IGCC units. Using these process
models for the MR and PSMR cases, constrained optimization problems are formulated that seek
to maximize performance (H2 recovery) through minimization of membrane surface area—this is
achieved by considering alternate membrane placement about the axial axis of the reactor. Through the
formulation of two optimization problems, the performance-limiting membrane parameter is identified
and a minimum value that satisfies all equipment/stream constraints is successfully calculated for
each configuration. This study contributes insight into identifying and prioritizing the membrane
parameters that should be the focus of future polymeric membrane development efforts, and provides
a minimum value for key parameters that satisfy the set of six performance constraints; to this end,
it is worth noting that the minimum selectivity value (one such key parameter) presented here is
unique to the operating temperature and pressure of the process units, and the syngas feed/steam
sweep flowrates.
Processes 2016, 4, 8  4 of 19 

2. Systems Analysis—Process Modeling, Simulation and Optimization Approach 
Processes 2016, 4, 8 4 of 19
2.1. Membrane Modeling 
2. Systems Analysis—Process
The  membrane  Modeling,
reactor  model  Simulation
employed  and
for  the  Optimization
performance  Approach
assessment  and  optimization 
studies is a one‐dimensional, isothermal model in which operation is steady‐state and the ideal gas 
2.1. Membrane Modeling
law is assumed to hold. The 1‐D and isothermal model assumptions are reasonable for a laboratory‐
scale The
membrane 
membrane reactor 
reactor[15]. 
modelThis  model  was 
employed developed 
for the performancebased  on  the  WGS‐MR 
assessment model  in 
and optimization [2].  A 
studies
summary of the development is presented below; refer to [2] for additional detail. Assuming plug‐
is a one-dimensional, isothermal model in which operation is steady-state and the ideal gas law is
flow  operation, 
assumed to hold.the 
Themembrane  reactor  model 
1-D and isothermal modelconsists  of  species 
assumptions are mole  balances 
reasonable for for  co‐current  and 
a laboratory-scale
counter‐current cases: 
membrane reactor [15]. This model was developed based on the WGS-MR model in [2]. A summary of
Mole balance, tube: 
the development is presented below; refer to [2] for additional detail. Assuming plug-flow operation,
the membrane reactor model consists of species ,
mole balances for co-current and counter-current cases:
Mole balance, tube:    
dFi,t
“ ri At ´ Ji πdt
where Fi,t is the flow rate in the tube, ri is the species reaction rate, A t is the cross‐sectional area of the 
dz
tube, Ji is the molar flux across the tube wall, and dt is the tube diameter. Additionally, ri = rCO for 
where Fi,t is the flow rate in the tube, ri is the species reaction rate, At is the cross-sectional area of the
i=CO, H2O; ri = ‐rCO for i = CO2, H2 and ri = 0 for i = N2. The reaction rate, rCO, is the rate associated with 
tube, Ji is the molar flux across the tube wall, and dt is the tube diameter. Additionally, ri = rCO for
the Cu/ZnO/Al2O3 catalyst [16]. 
i=CO, H2 O; ri = -rCO for i = CO2 , H2 and ri = 0 for i = N2 . The reaction rate, rCO , is the rate associated
Mole balance, shell: 
with the Cu/ZnO/Al2 O3 catalyst [16].
Mole balance, shell: ,
 
dFi,s  
p˘q “ Ji πdt
dz
where Fi,s is the flow rate in the shell. The positive coefficient corresponds to co‐current operation and 
where Fi,s is the flow rate in the shell. The positive coefficient corresponds to co-current operation
negative to counter‐current. For the permeative stage membrane reactor, the only differences in the 
and
model  are:  Ji to
negative counter-current.
=  0  For theand 
in  the  reactor  stages,  permeative
ri  =  0  in stage membraneseparator 
the  membrane  reactor, the only differences
stages.  in
The  resulting 
the model are: Jmodel 
mathematical  i = 0 in the reactor
consists  of stages, and rsystem 
an  ODE  i = 0 in the membrane separator
corresponding  stages.value 
to  an  initial  The resulting
problem   
mathematical model consists of an ODE system corresponding to an initial
(co‐current) or a boundary value problem (counter‐current), both of which may be solved using the  value problem (co-current)
or a boundary value problem (counter-current), both of which may be solved using the MATLAB
MATLAB subroutines ode15s or bvp4c, respectively. A schematic of the counter‐current MR design 
subroutines ode15s or bvp4c, respectively. A schematic of the counter-current MR design employed here
employed here is shown in Figure 1; co‐current operation of the unit would align the feed/sweep in 
is shown in Figure 1; co-current operation of the unit would align the feed/sweep in the same direction
the same direction with respect to the axial axis. A comparison using models developed in Aspen 
with respect to the
Plus  considering  the  axial axis. Aof comparison
scenarios  an  IGCC  plant  usingwith models developedprocess 
the  WGS‐MR  in Aspen Plus the 
against  considering
CO  shift 
the scenarios of an IGCC plant with the WGS-MR process against the CO shift followed
followed  by  physical  absorption  (e.g.,  Selexol)  technologies  for  CO2  capture  will  be  analyzed  in  a  by physical
absorption (e.g., Selexol) technologies for CO2 capture will be analyzed in a future publication.
future publication. 

 
Figure 1. The membrane reactor consists of a shell and tube setup in which the tube is packed with 
Figure 1. The membrane reactor consists of a shell and tube setup in which the tube is packed with
catalyst and the membrane is fixed to the tube wall; reaction/permeation occur simultaneously. 
catalyst and the membrane is fixed to the tube wall; reaction/permeation occur simultaneously.

The flux through the high‐performance polymer membrane is assumed to be Fickian activated 
The flux
diffusion,  through
which  the high-performance
is  proportionate  polymer partial 
to  the  component  membrane is assumed
pressure  to be
difference  Fickian
across  the   
activated diffusion, which is proportionate to the component partial pressure difference across the
membrane [17]; and is described by: 
membrane [17]; and is described by:
∆  
Ji “ Qi ∆p
 
i

where ∆pi is the partial pressure difference of the component across the membrane. The permeance
of a component i, Qi , is determined by the membrane properties that are taken from test systems
Processes 2016, 4, 8 5 of 19

available in the literature. The permeability (Pi ) of a component through polymers is considered to be
the product of the diffusion coefficient, Di , and solubility coefficient, Si [18].
Noting that selectivity (α) of species i to j is defined as the ratio of the permeability of species i to
j, such that:
P
αij “ i
Pj
and the permeance of a species is:
Pi
Qi “
δmem
where δmem is the membrane thickness.
In this paper, we focus on a class of high-performance PBI-based H2 -selective polymeric
membranes utilizing their demonstrated separation performance characteristics in multiple platforms
including flat sheet, tubular, and hollow fiber [5,7,10] for an assumed industrially relevant selective
layer thickness range of 100-200 nm. Thus, the following membrane characteristics are used in
this study:

‚ QH2 = 250 GPU

‚ αH2 /CO2 = 20–28

Also, for this study we assume that at high temperatures, these membranes have high permeability
to water (αH2 /H2 O = 0.33) and low permeability to the other species considered here. The H2 /CO
selectivity in this study was assumed to be 99, similar to experimentally measured H2 /N2 selectivity
of 99. However, based on the size difference between CO (kinetic diameter = 3.76 Å) and N2 (3.64 Å),
a H2 /CO selectivity greater than 99 would be possible. In particular, both H2 /CO2 and H2 /CO
permselectivities must be high for this application as one desires to produce a purified CO2 effluent
from the reactor side.

2.2. Simulation Set Up

The reactor feed composition/molar flow rate and sweep composition/molar flow rate are drawn
from [2]; the feed corresponds to a syngas stream from the gasifier after steam injection (it is assumed
that sulfurous compounds and other impurities have been removed) while the sweep composition is
pure steam. The feed/sweep compositions are summarized in Table 1.

Table 1. Molar composition of reactor inlet streams, given in mole fraction.

Component Syngas Feed Steam Sweep

H2 0.1933 0
CO2 0.0568 0
H2 O 0.4886 1
CO 0.2443 0
N2 0.017 0

The performance of each reactor system is evaluated in terms of the three performance goals set
forth by the U.S. DOE for CO conversion, H2 recovery and CO2 capture in addition to three constraints
on the reactor effluent streams as defined [2]:

‚ CO conversion (XCO )
` ˘
CO converted FCO, f ´ FCO,r ` FCO,p
XCO “ “ ě 98%
CO in feed FCO, f
Processes 2016, 4, 8 6 of 19

‚ H2 recovery (RH2 )
H2 in permeate FH2 ,p
RH2 “ “ ě 95%
pH2 ` COq in feed FH2 , f ` FCO, f

‚ CO2 Capture (CCO2 )

carbon in retentate FCO,r ` FCO2 ,r

CCO2 “ “ ě 90%
carbon in feed FCO, f ` FCO2 , f

‚ CO2 + H2 O purity in the retentate

purityCO2 `H2 O,r ě 95%

‚ H2 mole fraction in the retentate

yH2 ,r ď 4%

‚ H2 purity in the permeate

purityH2 ,p ě 44%

The reactor designs considered in the performance assessment and the optimization problems
(as the initial guess) are a 2-stage PSMR, 3-stage PSMR and a conventional MR. Reactor feed/sweep
molar flow rate, composition and flow arrangement (counter-current) are kept constant across all
simulations, as are all other reactor operating conditions such as temperature (constant at 300 ˝ C),
tube/shell pressure (47.63 atm/25.86 atm, respectively), mass of catalyst (20 mg), and tube/shell
diameter (1.02 cm/6.12 cm, respectively). Flow arrangement was fixed as counter-current as co-current
results were consistently unable to satisfy any constraint other than CO conversion; this result was
observed in [2] and was verified in the co-current simulations performed as part of this study. With
regard to the membrane properties, maintenance of constant temperature, and fixed permeance (and
selectivity) values for the analyzed polymer material are considered. For base case performance studies,
total reactor length is kept constant at 300 cm (thereby making membrane surface area a constant) for
MRs, and for the PSMRs the catalyst/membrane are divided into their components and arranged in
equally sized pieces such that the total length (L) of the 2/3-stage PSMRs are 600 cm—this corresponds
to catalyst/membrane lengths of 150 cm per stage in the 2-stage PSMR and 100 cm per stage in the
3-stage PSMR.

2.3. Optimization Problem Formulations

The first of the two formulated optimization problems seeks to maximize reactor performance,
expressed by H2 production, subject to the six performance constraints, by considering alternative
catalyst/membrane placement while minimizing the total membrane surface area (Sm ) required. To
this end, cost parameters were associated with RH2 and Sm in accordance with the method set forth
by [2]. The main difference in this analysis is the cost of the high-performance polymer membrane.
The polymer membrane cost is estimated to range between $5–200/m2 depending on the membrane
module platform [19], where an all polymeric hollow fiber platform typically provides the best
economics, i.e., lowest cost, and a porous inorganic supported composite tubular membrane platform
is typically the highest cost option. The application of robust stainless steel porous supports with
weldability and correspondingly perceived lower risk for incorporation into MR configuration can
further increase the cost of the resulting polymer/inorganic tubular membranes [20]. For the purposes
of this work, we have chosen the highly robust tubular membrane platform as our benchmark. As
such, a cost of $m = $500/m2 has been assumed as an upper bound estimate for the cost of the polymer
membrane selective layer on a tubular stainless steel support (the platform utilized by [5] in their year
plus evaluations of this membrane in elevated temperature separation environments).
Processes 2016, 4, 8  7 of 19 
Processes 2016, 4, 8 7 of 19
membrane selective layer on a tubular stainless steel support (the platform utilized by [5] in their 
year plus evaluations of this membrane in elevated temperature separation environments). 
The objective function formulated for this optimization problem is defined as:
The objective function formulated for this optimization problem is defined as: 
ϕφ“ min
min costm ´ creditH2  
“ ‰
 
x

In which: 
In which:
HHVH2 $$H2 Op
creditH2 “ FH2,p HHV
,
Op   
costm,2,´stage PSMR “ $$m πd
πdt pl2 ´ l1 ` l4 ´ l3 q  
costm,3,´stage PSMR
“ $$m πd
πdt pl2 ´ l1 ` l4 ´ l3 ` l6 ´ l5 q  

costm,MR
, “ $$m πd
πdt pl6 ´ l5 ` l8 ´ l7 ` L ´ l9 q  
where F
where FHH2,p(mol/s) is the molar flow rate of H
2 ,p
(mol/s) is the molar flow rate of2 in the permeate, HHV
H2 in the permeate, HHV H2 (BTU/mol) is the higher heating 
H2 (BTU/mol) is the higher
value of H , $  ($/BTU) is the monetary credit associated with the heating value, and Op is the 1‐year 
heating value of H2 , $H2 ($/BTU) is the monetary credit associated with the heating value, and Op is
2 H2

operating period in seconds. 
the 1-year operating period in seconds.
In particular, for the 2‐stage PSMR design depicted in Figure 2, the vector of decision variables 
In particular, for the 2-stage PSMR design depicted in Figure 2, the vector of decision variables is
is as follows: 
as follows:
x2´stage PSMR “ rl1 l2 l3 l4 sT    

 
Figure 2.2.  Arrangement
Figure Arrangement  of
of  decision
decision  variables
variables  about
about  the
the  axial
axial axis
axis of
of the
the 2-stage
2‐stage permeative
permeative stage
stage 
membrane reactor (PSMR). 
membrane reactor (PSMR).

Subject to the dimensional constraints on the catalyst stage lengths: 
Subject to the dimensional constraints on the catalyst stage lengths:
0,    
l1 ą 0, l3 ě l2
and for the membrane stage lengths: 
and for the membrane stage lengths: , , L   
The initial guess for the 2‐stage PSMR optimization problem corresponds to four equally sized 
l2 ě l1 , l4 ě l3 , l4 ď L
stages (two catalyst, two membrane) consisting of the same membrane surface area and catalyst mass 
as the conventional MR, given by: 
The initial guess for the 2-stage PSMR optimization problem corresponds to four equally sized
stages (two catalyst, two membrane) consisting of the150 same300membrane
450 600 surface
  area and catalyst mass
,  
as the conventional MR, given by:
The vector of decision variables for the 3‐stage PSMR design shown in Figure 3 is as follows: 
x2´stage PSMR,initial “ r150 300 450 600sT
   
The vector of decision variables for the 3-stage PSMR design shown in Figure 3 is as follows:
Subject to the dimensional constraints on the catalyst stage lengths: 
0, “ rl l, l l l l sT
x3´stage PSMR  
1 2 3 4 5 6
and for the membrane stage lengths: 
Subject to the dimensional constraints on the catalyst stage lengths:
, , , L   
l1 ą 0, l3 ě l2 , l5 ě l4
Processes 2016, 4, 8 8 of 19

and for the membrane stage lengths:

l2 ě l1 , l4 ě l3 , l6 ě l5 , l6 ď L
Processes 2016, 4, 8  8 of 19 
Processes 2016, 4, 8  8 of 19 

 
Figure 3. Arrangement of decision variables about the axial axis of a 3‐stage PSMR.   
Figure 3. Arrangement of decision variables about the axial axis of a 3-stage PSMR.
Figure 3. Arrangement of decision variables about the axial axis of a 3‐stage PSMR. 
The  initial  guess  for  the  3‐stage  PSMR  corresponds  to  six  equally  sized  stages  of 
The initial
The  guess
initial  forfor 
guess  thethe 3-stage
3‐stage  PSMR
PSMR  corresponds
corresponds  to tosix 
sixequally 
equally sized
sized 
catalyst/membrane, using the same membrane surface area and catalyst mass as the 2‐stage PSMR, is  stages
stages  of  of
catalyst/membrane, using the same membrane surface area and catalyst mass as the 2-stage PSMR, is
catalyst/membrane, using the same membrane surface area and catalyst mass as the 2‐stage PSMR, is 
as follows: 
as follows: 
as follows:
, 100 200 300 400 500 600    
x3´stage PSMR,initial
,
200300
100 200
“ r100 300400
400500 600s T
500600  
The  vector  of  decision  variables  corresponding  to  the  MR  design  presented  in  Figure  4  is   
The vector of decision
The  vector 
as follows:  variables
of  decision  corresponding
variables  to theto MR
corresponding  the design presented
MR  design  in Figure
presented  4 is as4 follows:
in  Figure  is   
as follows: 
   
xMR “ rl1 l2 l3 l4 l5 l6 l7 l8 l9 sT
   

 
 
Figure 4. Arrangement of decision variables about the axial axis of the MR. 
Figure 4. Arrangement of decision variables about the axial axis of the MR. 
Figure 4. Arrangement of decision variables about the axial axis of the MR.
Subject to the dimensional constraints on the reaction zone: 
Subject to the dimensional constraints on the reaction zone: 
0, , , , L   
Subject to the dimensional constraints 0, on the , reaction , zone:
, L   
and in the permeation zone: 
and in the permeation zone:  l ą
1 0, l2 ě l,1 , l3 ě, l2 , l4 ě , l3 , l4, ď LL   
0, , , , , L   
The initial guess for the MR is a conventional case in which catalyst/membrane are present along 
and in the permeation zone:
The initial guess for the MR is a conventional case in which catalyst/membrane are present along 
the whole axial length, the vector for which is: 
the whole axial length, the vector for which is: 
l5 ě, 0, l6 ě 100
l5 , l100
7 ě 200l6 , l200
8 ě 0l7100
, l9 100ě l8200
, l9 200
ďL    
, 100 100 200 200 0 100 100 200 200    
A  second  optimization  problem  is  formulated  to  verify  the  hypothesis  that  minimization  of 
The initial guess
A  second 
membrane  forarea 
thealso 
MRproblem 
optimization 
surface  is a conventional
is  formulated 
corresponds  caseto inverify 
to  maximization  which catalyst/membrane
of the 
the  hypothesis  areparameter 
presentof 
that  minimization 
limiting  performance  along
the whole axial surface 
membrane  length, area 
the vector for which to 
also  corresponds  is: maximization  of  the  limiting  performance  parameter 
(CCO2), using the five remaining nonlinear constraints on reactor performance with the same linear 
(C CO2), using the five remaining nonlinear constraints on reactor performance with the same linear 
constraints as  the  cost  optimization  problem  described  above.  The  difference  in  this  case is  in  the 
constraints as  xMR,initial “ r100
the  cost  optimization  100 described 
problem  200 200 0 above. 
formulation of the objective function, which is defined as:  100 100The  200sT in  this  case is  in  the 
200difference 
formulation of the objective function, which is defined as: 
φ min     of
A second optimization problem is formulated to verify the hypothesis that minimization
φ min    
membrane surface area also corresponds to maximization of the limiting performance parameter
Processes 2016, 4, 8 9 of 19

(CCO2 ), using the five remaining nonlinear constraints on reactor performance with the same linear
constraints as the cost optimization problem described above. The difference in this case is in the
formulation of the objective function, which is defined as:
“ ‰
ϕ “ min ´CCO2
x

As both optimization problems possess nonlinear objective functions subject to a set of nonlinear
constraints, solutions may be obtained through the MATLAB fmincon subroutine employing the
“active-set” algorithm.

3. Systems Analysis Results

3.1. 2-Stage Permeative Stage Membrane Reactor Performance, Optimization

The performance of the high-performance polymeric membrane is first assessed as part of a
2-stage PSMR, using permeance (QH2 = 250 GPU) and H2 /CO2 selectivity (αH2 /CO2 = 28) values. The
results of this simulation are summarized in Table 2, which employs the feed/sweep flow rate, flow
composition, flow arrangement (counter-current), temperature and pressure conditions as defined
above. The placement of catalyst/membrane correspond to x2´stage PSMR,initial or the base case design.

Table 2. Performance of polymer membrane in a 2-stage PSMR (QH2 = 250 GPU, αH2 /CO2 = 28).

Parameter Value (%) Target (%)

XCO 98.95 98
RH2 98.60 95
CCO2 74.71 90
purityCO2 `H2 O,r 95.98 95
purityH2 ,p 41.71 44
yH2 ,r 0.77 ď4

Thus, two stages for this base case satisfies all but carbon capture and H2 purity constraints for
this reactor configuration. Also through simulations, the limiting membrane characteristic for these
performance parameters is identified as the H2 /CO2 selectivity (though reduction to total membrane
surface area improves carbon capture, sufficient reductions cannot be performed should one desire
to satisfy the remaining five performance constraints). Thus, an incremental variation of H2 /CO2
selectivity is performed for the range of α = 25–75 with the objective of determining the minimum
H2 /CO2 selectivity that would satisfy the carbon capture and permeate hydrogen purity constraints in
an optimized (minimum membrane surface area) 2-stage PSMR. Utilizing an α increment of 5, eleven
optimization problems were formulated and solved using the technique described above. Designs
employing values of αH2 /CO2 greater than 30 satisfy all but the carbon capture constraint; the first
design to satisfy all six constraints in a 2-stage PSMR falls within the range αH2 /CO2 = 60–65, and
occurs at a H2 /CO2 selectivity value of approximately 62. As the solutions represent a maximization
of carbon capture attainable in a 2-stage PSMR, while satisfying the other five performance constraints,
the results may be used to determine the minimum H2 /CO2 selectivity needed to satisfy a given
carbon capture constraint. The carbon capture resulting from varying selectivity on the range of 25–75
in an optimized 2-stage PSMR employing the minimum membrane area is shown in Figure 5.
The solution vector for the case corresponding to αH2 /CO2 = 60 is shown below; this case is closest
to satisfying CCO2 ě 0.90 in the 2-stage PSMR.

x2´stage PSMR,final,α “ r156.28 295.85 454.85 579.93sT

H2
“60
CO2
Processes 2016, 4, 8  10 of 19 

Processes 2016, 4, 8 10 of 19
Processes 2016, 4, 8  10 of 19 

 
 
Figure  5.  Maximization  of  carbon  capture  in  an  optimized  2‐stage  PSMR;  only  the  design 
Figure  5.  Maximization  of  carbon  capture  in  an  optimized  2‐stage  PSMR;  only  the  design 
Figure 5. Maximization
corresponding to α of carbon capture in an optimized  otherwise the remaining five constraints are 
H2/CO2 = 25 fails to satisfy 
2-stage PSMR; only the design corresponding
corresponding to αH2/CO2 = 25 fails to satisfy  ,   otherwise the remaining five constraints are 
,
to αH2 /CO2 = 25 fails toCOsatisfy
satisfied for all cases; C purityH2 ,p otherwise the remaining five constraints are satisfied for all
satisfied for all cases; C2 ≥ 90% corresponds to the target value used in this study. 
CO2 ≥ 90% corresponds to the target value used in this study. 
cases; CCO2 ě 90% corresponds to the target value used in this study.

Upon examination of the optimization results representing a minimization of membrane surface 
Upon examination of the optimization results representing a minimization of membrane surface 
Upon examination of the optimization results representing a minimization of membrane surface
area, a pattern is noted in the optimal 2‐stage PSMR catalyst/membrane placement. In each design, 
area, a pattern is noted in the optimal 2‐stage PSMR catalyst/membrane placement. In each design, 
area, a pattern is noted in the optimal 2-stage PSMR catalyst/membrane placement. In each design,
there is an approximately equal distribution of catalyst mass across the two stages; the membrane 
there is an approximately equal distribution of catalyst mass across the two stages; the membrane 
there is an approximately equal distribution of catalyst mass across the two stages; the membrane
stages are unevenly distributed, with the first membrane stage slightly larger than the second for all 
stages are unevenly distributed, with the first membrane stage slightly larger than the second for all 
stages are unevenly distributed, with the first membrane stage slightly larger than the second
cases. The membrane placement likely results from the relatively high partial pressure of H for all
2 in the 
cases. The membrane placement likely results from the relatively high partial pressure of H 2 in the 
stream exiting the first reactor stage (see Figure 6), thus, placement of more membrane directly after 
cases. The membrane placement likely results from the relatively high partial pressure of H2 in the
stream exiting the first reactor stage (see Figure 6), thus, placement of more membrane directly after 
the first reactor allows for greater utilization of the permeation driving force and consequently greater 
stream exiting the first reactor stage (see Figure 6), thus, placement of more membrane directly after the
the first reactor allows for greater utilization of the permeation driving force and consequently greater 
H   recovery  in  this  region  (relative  to  of
the  second  membrane  separator  stage).  Figure  7  shows  the 
first reactor allows
2 for greater utilization the permeation driving force and consequently greater H 2
H2  recovery  in  this 
profiles for the H region  (relative  to  the  second  membrane  separator  stage).  Figure  7  shows 
2 reaction and diffusion rates as function of axial axis for this optimized 2‐stage PSMR. 
the   
recovery in this region (relative to the second membrane separator stage). Figure 7 shows the profiles
profiles for the H2 reaction and diffusion rates as function of axial axis for this optimized 2‐stage PSMR. 
for the H2 reaction and diffusion rates as function of axial axis for this optimized 2-stage PSMR.

 
Figure 6. Species concentration (mol/cm ) as function of axial length for optimized 2‐stage PSMR for   
3

αH2/CO2 = 60, corresponding to the solution vector shown above; solid blue and solid green lines denote 
feed and sweep streams, respectively. 
 
Figure 6. Species concentration (mol/cm
Figure 6. Species concentration (mol/cm
3) as function of axial length for optimized 2‐stage PSMR for 
3 ) as function of axial length for optimized 2-stage PSMR  
αHfor 2 = 60, corresponding to the solution vector shown above; solid blue and solid green lines denote 
2/COα
H2 /CO2 = 60, corresponding to the solution vector shown above; solid blue and solid green lines
feed and sweep streams, respectively. 
denote feed and sweep streams, respectively.
Processes 2016, 4, 8 11 of 19
Processes 2016, 4, 8  11 of 19 

 
Figure 7. H22 reaction (r
Figure 7. H reaction (rH ) and diffusion (JH2a, a ≡ 4/d
H22) and diffusion (JH2
a, a ” 4/d t ) rates as function of axial axis for the optimized
t) rates as function of axial axis for the optimized 
2-stage PSMR (α
2‐stage  PSMR  (αH 2  2
/CO
H22/CO
= 60); dimensionless quantities
=  60);  dimensionless  quantities  scaled scaledby 
by the 
the maximum 
maximum H H22  reaction 
reaction and 
and
diffusion rates, respectively.
diffusion rates, respectively. 

3.2. 3-Stage Permeative Stage Membrane Reactor Performance, Optimization

3.2. 3‐Stage Permeative Stage Membrane Reactor Performance, Optimization 
The performance of the polymeric membrane is assessed next as a 3-stage PSMR,H2using
The performance of the polymeric membrane is assessed next as a 3‐stage PSMR, using Q  = 250 
Q H2 = 250 GPU,
GPU, and α and αH2 /CO2 = 28 once again as base case. The results of this simulation, which
H2/CO2 = 28 once again as base case. The results of this simulation, which uses precisely the 

uses precisely
same  the same process
process  conditions  as  the conditions as theare 
2‐stage  PSMR,  2-stage PSMR, are
summarized  in summarized
Table  3;  the inplacement 
Table 3; the
of 
placement of catalyst/membrane in the 3-stage
catalyst/membrane in the 3‐stage PSMR correspond to PSMR correspond
to x .  PSMR,initial
, 3´stage .

Table 3. Performance of polymeric membrane in a 3-stage PSMR (QH2H=

Table 3. Performance of polymeric membrane in a 3‐stage PSMR (Q 250 GPU, αHH22/CO
2 = 250 GPU, α = 28).
/CO2 = 28). 
2

Parameter 
Parameter Value (%)
Value (%) Target (%)
Target (%)
99.44  98 
XCO 99.44 98
R H2 98.82 
98.82 95 
95
CCO2 74.73 
74.73 90 
90
purity
purityCO2 `H2 O,r , 96.17
96.17  95
95 
purity
purityH2 ,p , 41.82
41.82  44
44 
yHy2 ,r 0.71 ď4
, 0.71  ≤4 

Once 
Once more, 
more, given 
given the 
the membrane 
membrane properties, 
properties, the 
the material 
material did 
did not 
not to to satisfy 
satisfy the  carbon 
the carbon
capture/permeate  hydrogen purity
capture/permeate hydrogen purity  constraint 
constraint whilewhile  satisfying 
satisfying the otherthe 
four other  four  performance 
performance constraints.
constraints. 
Similar to theSimilar 
previousto case,
the  previous  case,  variation
an incremental an  incremental  variation 
of H2 /CO 2 of 
selectivity H was
2 /CO   selectivity 
performed
2 forwas 
the
performed for the range of α = 25–75 so as to determine the minimum H /CO  selectivity that would 
range of α = 25–75 so as to determine the minimum H2 /CO2 selectivity that would satisfy the carbon
2 2

satisfy the carbon capture and permeate hydrogen purity constraints in an optimized 3‐stage PSMR. 
capture and permeate hydrogen purity constraints in an optimized 3-stage PSMR. These simulations
These simulations indicate that designs employing H
indicate that designs employing H2 /CO2 selectivity values 2/CO2 selectivity values greater than 30 satisfy 
greater than 30 satisfy all but the carbon
all but the carbon capture constraint, with the first design satisfying all six constraints falling in the 
capture constraint, with the first design satisfying all six constraints falling in the αH2 /CO2 range of
α H2/CO2 with
60–65 range 
theof  60–65  with 
minimum the 2 minimum 
H2 /CO H2/CO
selectivity that 2  selectivity 
satisfies that  satisfies 
all constraints all  constraints 
of approximately of 
61. The
approximately 61. The carbon capture resulting from varying selectivity on the range of 25–75 in an 
carbon capture resulting from varying selectivity on the range of 25–75 in an optimized 3-stage PSMR
optimized 3‐stage PSMR employing the minimum membrane area is shown in Figure 8, along with 
employing the minimum membrane area is shown in Figure 8, along with the previously obtained
the previously obtained result for the 2‐stage reactor. 
result for the 2-stage reactor.
Processes 2016, 4, 8 12 of 19
Processes 2016, 4, 8  12 of 19 

 
Figure 8. Maximization of carbon capture in an optimized 3‐stage PSMR (blue) as well as optimized 
Figure 8. Maximization of carbon capture in an optimized 3-stage PSMR (blue) as well as optimized
2‐stage  PSMR  (black);  only  the  designs  corresponding  to  αH2/CO2  =  25  fail  to  satisfy  , 
2-stage PSMR (black); only the designs corresponding to αH2 /CO2 = 25 fail to satisfy purityH2,,p ,
otherwise the remaining five constraints are satisfied for all cases; CCO2 ≥ 90% corresponds to the target 
otherwise the remaining five constraints are satisfied for all cases; CCO2 ě 90% corresponds to the
value used in this study. 
target value used in this study.

The solution vector for the case corresponding to αH2/CO2 = 60 is shown below; this case is closest 
The solution
to satisfying C vector for the case corresponding to αH2 /CO2 = 60 is shown below; this case is closest
CO2 ≥ 0.90 in the 3‐stage PSMR. 
to satisfying CCO2 ě 0.90 in the 3-stage PSMR.
, , 107.26 192.26 309.39 399.86 499.83 583.20    
x3´stage PSMR,final,α “ r107.26 192.26 309.39 399.86 499.83 583.20sT
H2
From the optimization results corresponding to a minimization of membrane area for the given 
“60
CO2
range  of  selectivity,  the  designs  conformed  to  a  general  pattern  (as  viewed  from  left  to  right  in   
Figure 3): an uneven catalyst distribution that preferentially placed the most catalyst in the second 
From the optimization results corresponding to a minimization of membrane area for the given
reactor stage, and slightly more catalyst in the first reactor stage than the third reactor stage (in this 
range of selectivity, the designs conformed to a general pattern (as viewed from left to right in Figure 3):
case differences in catalyst amount corresponded to less than 10% of the total catalyst mass). As for 
an uneven catalyst distribution that preferentially placed the most catalyst in the second reactor stage,
the membrane area placement, a pattern was also observed across the range of selectivity values in 
and slightly more catalyst in the first reactor stage than the third reactor stage (in this case differences
which 
in more 
catalyst membrane 
amount was  utilized 
corresponded to lessin  the 10%
than second 
of thestage 
totalthan  the mass).
catalyst first/third  stages; 
As for however, area
the membrane the 
difference in total membrane area between each stage was small (the second membrane separator 
placement, a pattern was also observed across the range of selectivity values in which more membrane
stage 
was utilized 
utilized in 5%–15% 
the second more 
stagemembrane  than  the stages;
than the first/third first/third  stages). 
however, theThe  difference 
difference in  membrane
in total membrane 
placement is likely due to the relatively large difference in partial pressure presented in the second 
area between each stage was small (the second membrane separator stage utilized 5%–15% more
membrane stage (see Figure 9); at this point, the sweep gas has a relatively low H
membrane than the first/third stages). The difference in membrane placement2 mole fraction while 
is likely due to the
the reactor stream has a relatively larger H
relatively 2 mole fraction (having passed over approximately ⅔ of 
large difference in partial pressure presented in the second membrane stage (see Figure 9); at
the  point,
this total  catalyst 
the sweepmass  (two 
gas has reactor  stages), 
a relatively but  fraction
low H2 mole only  one  membrane 
while the reactor separator 
stream hasstage).  When 
a relatively
2
comparing the performance of the 2‐stage PSMR to that of the 3‐stage PSMR, it is worth noting that 
larger H2 mole fraction (having passed over approximately {3 of the total catalyst mass (two reactor
the use of three reaction/permeation stages allowed for increased CO conversion and H
stages), but only one membrane separator stage). When comparing the performance of 2 recovery by 
the 2-stage
relieving equilibrium limitations on the WGS reaction (the first membrane separator is implemented 
PSMR to that of the 3-stage PSMR, it is worth noting that the use of three reaction/permeation stages
after the reactor feed sees approximately ⅓ of the total catalyst, rather than ½ the total catalyst as in 
allowed for increased CO conversion and H2 recovery by relieving equilibrium limitations on the WGS
(the first membrane separator is implemented after the reactor feed sees approximately 1 {3
the 2‐stage PSMR). Additionally, the 3‐stage design permitted increases to carbon capture through 
reaction
more 
of efficient 
the total membrane 
catalyst, utilization 
rather than (optimized 
½ the total catalyst3‐stage  PSMRs 
as in the 2-stageemployed  slightly  less  the
PSMR). Additionally, membrane 
3-stage
area than optimized 2‐stage PSMRs). Figure 10 depicts the profiles for the H
design permitted increases to carbon capture through more efficient membrane 2 reaction and diffusion 
utilization (optimized
rates as function of axial axis for this optimized 3‐stage PSMR. 
3-stage PSMRs employed slightly less membrane area than optimized 2-stage PSMRs). Figure 10
depicts the profiles for the H2 reaction and diffusion rates as function of axial axis for this optimized
3-stage PSMR.
Processes 2016, 4, 8 13 of 19
Processes 2016, 4, 8  13 of 19 

Processes 2016, 4, 8  13 of 19 

 
3  
Figure 9. Species
Figure  concentration
9.  Species  (mol/cm
concentration  (mol/cm)3) profiles
profiles inin the
the optimized
optimized 3-stage
3‐stage  PSMR for α
PSMR  for  αHH2/CO
2 /CO
2  = 
= 60,
2 60, 
Figure  9.  Species  concentration  (mol/cm3)  profiles  in  the  optimized  3‐stage  PSMR  for  αH /CO   =  60,  2 2
corresponding to the solution vector shown above; solid blue and solid green lines
corresponding to the solution vector shown above; solid blue and solid green lines denote feed and denote feed and
corresponding to the solution vector shown above; solid blue and solid green lines denote feed and 
sweep streams, respectively.
sweep streams, respectively. 
sweep streams, respectively. 

 
 
Figure 10. H2 reaction (rH2) and diffusion (JH2a, a ≡ 4/dt) rates for the optimized 3‐stage PSMR (αH2/CO2 = 60); 
Figure 10. H2 reaction (rH2 ) and diffusion (JH2 a, a ” 4/dt ) rates for the optimized 3-stage
Figure 10. H2 reaction (rH2) and diffusion (JH2a, a ≡ 4/dt) rates for the optimized 3‐stage PSMR (α
dimensionless quantities scaled by the maximum H H2/CO2 = 60); 
2 reaction and diffusion rates, respectively. 
PSMR (αH2 /CO2 = 60); dimensionless quantities scaled by the maximum H2 reaction and diffusion
dimensionless quantities scaled by the maximum H2 reaction and diffusion rates, respectively. 
rates, respectively.
3.3. Membrane Reactor Performance, Optimization 
3.3. Membrane Reactor Performance, Optimization 
Finally, the performance of the polymeric membrane (QH  = 250 GPU, αH /CO  = 28) is assessed as 
2 2 2
3.3. Membrane Reactor Performance, Optimization
a  traditional  membrane  reactor.  The  results  of  this  simulation,  which  uses  the  same  process 
Finally, the performance of the polymeric membrane (QH2 = 250 GPU, αH2/CO2 = 28) is assessed as 
conditions 
a Finally, as  the  2,3‐stage 
the performance of thePSMRs, 
The are 
polymeric summarized 
of  this in 
membrane (QTable  4;  the  reactor  design  is  that  of  a 
traditional  membrane  reactor.  results  H2 = 250 GPU,
simulation,  which αHuses 
2 /CO2
= 28)
the  is assessed
same  process  as
conventional MR, which corresponds to 
conditions  as  the  2,3‐stage  , . 
a traditional membrane reactor.PSMRs,  are  of
The results summarized 
this in  Table 
simulation, which4;  the 
usesreactor 
the samedesign 
process is  that  of  a 
conditions
Noting that the material did not satisfy the carbon capture and hydrogen purity constraints, the 
conventional MR, which corresponds to 
as the 2,3-stage PSMRs, are summarized in Table , .  the reactor design is that of a conventional MR,
4;
same procedure outlined above for the 2‐stage and 3‐stage PSMRs was performed for the MR; that 
which is:  Noting that the material did not satisfy the carbon capture and hydrogen purity constraints, the 
corresponds x MR,initial
incremental to variation  of .H2/CO2  selectivity  was  performed  for  the  range  of  α  =  25–75  so  as  to 
same procedure outlined above for the 2‐stage and 3‐stage PSMRs was performed for the MR; that 
Noting thatthe 
theminimum 
materialHdid
determine  2/COnot satisfy the
2  selectivity  that carbon capture
would  satisfy  the and hydrogen
carbon  capture purity constraints,
and  permeate 
the is: 
sameincremental 
procedure variation 
outlined of above
H2/CO2for   selectivity 
the 2-stagewas and
hydrogen purity constraints in an optimized MR. All values of α
performed 
3-stagefor  the  range 
PSMRs of  α  =  25–75 
was performed so the
for
H /CO  greater than 30 satisfied the 
2 2
as  to 
MR;
determine  the  minimum  H 2/CO2  selectivity  that  would  satisfy  the  carbon  capture  and  permeate 
that is:permeate hydrogen purity constraint, and the carbon capture constraint is satisfied in the range of 
incremental variation of H2 /CO2 selectivity was performed for the range of α = 25–75 so as
hydrogen purity constraints in an optimized MR. All values of α H2/CO2 greater than 30 satisfied the 
αH /CO  = 60–65, occurring at a value of approximately 61. The carbon capture resulting from varying 
to determine2 2
the minimum H2 /CO2 selectivity that would satisfy the carbon capture and permeate
permeate hydrogen purity constraint, and the carbon capture constraint is satisfied in the range of 
selectivity on the range of 25–75 in an optimized MR utilizing the minimum membrane area is shown 
hydrogen purity constraints in an optimized MR. All values of αH2 /CO2 greater than 30 satisfied the
αH2/CO 2 = 60–65, occurring at a value of approximately 61. The carbon capture resulting from varying 
in Figure 11. 
permeate hydrogen purity constraint, and the carbon capture constraint is satisfied in the range of
selectivity on the range of 25–75 in an optimized MR utilizing the minimum membrane area is shown 
/CO2 = 60–65, occurring at a value of approximately 61. The carbon capture resulting from varying
αH2in Figure 11. 
Processes 2016, 4, 8 14 of 19

selectivity on the range of 25–75 in an optimized MR utilizing the minimum membrane area is 14 of 19 
Processes 2016, 4, 8  shown
in Figure 11.
Table 4. Performance of polymeric membrane (QH2 = 250 GPU, αH2/CO2 = 28) in a conventional MR for 
Table 4. Performance of polymeric membrane (QH2 = 250 GPU, αH2 /CO2 = 28) in a conventional MR
the same conditions of the PSMR. 
Processes 2016, 4, 8  14 of 19 
for the same conditions of the PSMR.
Parameter Value (%) Target (%)
Table 4. Performance of polymeric membrane (QH2 = 250 GPU, αH2/CO2 = 28) in a conventional MR for 
Parameter
the same conditions of the PSMR.  99.36 
Value (%) Target98 (%)
98.38  95 
XCOParameter 99.36
Value (%) 98
Target (%)
RH2 98.3875.77  90 
99.36  98 95
CCO2 , 75.7796.05 
98.38  95  95 
90
purityCO2 `H2 O,r , 96.0542.05 
75.77  90 95
44 
purityH2 ,p 42.05 1.01 
96.05  44
95  ≤4 
, ,
yH2 ,r 1.0142.05  44 ď4
,
, 1.01  ≤4 

 
 
Figure 11. Maximization of carbon capture in an optimized MR (red) as well as 3‐stage PSMR (blue); 
Figure 11. Maximization of carbon capture in an optimized MR (red) as well as 3-stage PSMR (blue);
Figure 11. Maximization of carbon capture in an optimized MR (red) as well as 3‐stage PSMR (blue); 
only the design corresponding to αH /CO  = 25 fails to satisfy 
2 2 , , otherwise the remaining five 
only the design corresponding
only the design corresponding to α to α HH
2 /CO
constraints are satisfied for all cases; C = 25 fails to satisfy
2 = 25 fails to satisfy 
purity , otherwise the remaining five
H2, ,p, otherwise the remaining five 
2  ≥ 90% corresponds to the target value used in this study. 
2/COCO 2

constraints are satisfied for all cases; CCO

constraints are satisfied for all cases; C ě 90% corresponds to the target value used in this study.
CO22 ≥ 90% corresponds to the target value used in this study. 
As for the arrangement of the membrane about the axial axis of the reactor, the optimal solutions 
(representing a minimization of membrane surface area) were all of the same general form as shown 
AsAs for the arrangement of the membrane about the axial axis of the reactor, the optimal solutions 
for the arrangement of the membrane about the axial axis of the reactor, the optimal solutions
in  Figure  12  The  cause  for  variation  in  total  membrane  surface  area  was  due  to  differences  in 
(representing a minimization of membrane surface area) were all of the same general form as shown 
(representing a minimization of membrane
selectivity,  with  lower  values  of  αH2/CO2 surface area)
allowing  were allof ofthe 
for  dilution  thepermeate 
same general form
with  CO as shown in
2,  which 
in  Figure 
Figure 12 12  The 
The cause
improved H cause  for  variation  in  total  membrane  surface 
for variation in total2 partial pressure in the shell. 
2 recovery by lowering H
area  was  due  to  differences 
membrane surface area was due to differences in selectivity, in 
selectivity, 
with with  lower 
lower values of αH2values  of  αH /COfor
/CO2 allowing   allowing 
dilution for 
2 dilution 
of the
2 of  the 
permeate withpermeate  with improved
CO2 , which CO2,  which 
H2
improved H
recovery 2 recovery by lowering H
by lowering 2 partial pressure in the shell. 
H2 partial pressure in the shell.

 
Figure 12. General optimized MR design; the solution specific to each H2/CO2 selectivity (on range 
25–75) value falls within the range presented. 

 
Figure 12. General optimized MR design; the solution specific to each H
Figure /CO22 selectivity (on range 
12. General optimized MR design; the solution specific to each H22/CO selectivity (on range
25–75) value falls within the range presented. 
25–75) value falls within the range presented.
Processes 2016, 4, 8  15 of 19 
Processes 2016, 4, 8 15 of 19

The solution vector for the case corresponding to αH2/CO2 = 60 is shown below; this case is closest 
to  satisfying  CCO2  ≥  0.90 
The solution vectorin for
the 
theMR 
case(see  Figure  13  to
corresponding for 
αHconcentration  profiles 
= 60 is shown and this
below; Figure 
case 14 
is for 
2 /CO2
reaction/diffusion rate profiles associated with this case). 
closest to satisfying CCO2 ě 0.90 in the MR (see Figure 13 for concentration profiles and Figure 14 for
reaction/diffusion rate profiles associated with this case).
, , 98.43 98.43 142.53 257.30 15.22 90.54 96.88 185.21 200.20    
T
xMR,final,α “ r98.43 98.43 142.53 257.30 15.22 90.54 96.88 185.21 200.20s
H2
Each  optimal  design  consisted 
“60 of  a  short  pre‐shift  zone  lacking  membrane;  following  the   
CO2
pre‐shift zone, a region resembling a conventional MR exists until approximately 150 cm. From 150 
cm to 250 cm, catalyst is absent and only membrane is placed so as to remove the reaction H
Each optimal design consisted of a short pre-shift zone lacking membrane; following the pre-shift 2 product 

zone, a region resembling a conventional MR exists until approximately 150 cm. From 150 cm to
from the tube side (though, there is a small section of membrane removed between 180 and 200 cm). 
250 cm, catalyst is absent and only membrane is placed so as to remove the reaction H2 product from
From 250 cm to 300 cm, the design is once more that of the conventional MR, indicating that further 
the tube side (though, there is a small section of membrane removed between 180 and 200 cm). From
CO conversion is best achieved after removal of a significant portion of the reaction products. The 
250 cm to 300 cm, the design is once more that of the conventional MR, indicating that further CO
resulting  design  can  be  explained  by  the  more  efficient  membrane  area  utilization  resulting  from 
conversion is best achieved after removal of a significant portion of the reaction products. The resulting
increased  H2  partial  pressure  in  the  tube  side  achieved  through  use  of  a  pre‐shift  followed  by  a 
design can be explained by the more efficient membrane area utilization resulting from increased H2
conventional MR. In essence, the use of the combined pre‐shift zone and conventional MR increases 
partial pressure in the tube side achieved through use of a pre-shift followed by a conventional MR. In
H2  partial 
essence,pressure 
the use ofin 
thethe  reaction 
combined side,  but 
pre-shift zoneas  products 
and build 
conventional MR up  the  thermodynamic 
increases limitation 
H2 partial pressure in
associated 
the reaction side, but as products build up the thermodynamic limitation associated with the WGS is and 
with  the  WGS  is  increases.  At  this  point,  membrane  is  added  to  remove  products 
alleviate  this Atlimitation. 
increases. Following 
this point, membrane the  pre‐shift 
is added to removezone  in and
products which  product 
alleviate concentration 
this limitation. Followingis  not 
sufficiently  high, 
the pre-shift nearly 
zone continuous 
in which removal  of is
product concentration product  through 
not sufficiently the nearly
high, membrane  permits 
continuous increased 
removal of
product through the membrane permits increased CO conversion with respect to the staged reactors.
CO conversion with respect to the staged reactors. 

 
Figure 13. Species concentration (mol/cm
Figure 13. Species concentration (mol/cm) profiles as function of axial length in tube (solid blue line) 
33
) profiles as function of axial length in tube (solid blue line)
and shell (solid green line) for the solution vector shown above (optimized MR with α
and shell (solid green line) for the solution vector shown above (optimized MR with αH2 /CO  = 60). 
= 60).
H2/CO2
2
Processes 2016, 4, 8 16 of 19
Processes 2016, 4, 8  16 of 19 

 
Figure 14. H22 reaction (r
Figure 14. H reaction (rH2 ) and diffusion (JH2H2
H2) and diffusion (J a, a ” 4/d
a, a ≡ 4/d t ) rates, dimensionless quantities scaled by
t) rates, dimensionless quantities scaled by their 
their maximum rates, for optimized MR (α = 60); the ratio of reaction to diffusion H2
maximum rates, for optimized MR (αH2/COH2 = 60); the ratio of reaction to diffusion rate (r
2 /CO2
rate (ra, defined 
/ JH2 H2 /JH2 a,
defined only where both catalyst and membrane are present) provides insight into the catalyst
only where both catalyst and membrane are present) provides insight into the catalyst packing near  packing
near the end of the reactor.
the end of the reactor. 

The optimized MRs exhibited slightly higher CO conversions than were obtained in the 2,3‐stage 
The optimized MRs exhibited slightly higher CO conversions than were obtained in the 2,3-stage
PSMRs, asas well
PSMRs, well asas  attained 
attained slightly 
slightly higher 
higher H2Hrecovery
2  recovery  values; 
values; these 
these results 
results forfor 
αHα H2/CO2 =
2 /CO2
= 60
60 are
are 
summarized in Table 5. 
summarized in Table 5.

Table 5. CO conversion and H2 recovery

Table 5. CO conversion and H values for optimized MR, 2-stage PSMR and 3-stage PSMR at
2 recovery values for optimized MR, 2‐stage PSMR and 3‐stage PSMR 
αat α
H2 /CO = 60.
2  = 60. 
H2/CO2

Reactor Configuration
Reactor Configuration XCO (%)
X CO (%) RH
H2 (%)
2
(%)
2-stage2‐stage PSMR 
PSMR 98.68 98.68  96.17 
96.17
3-stage3‐stage PSMR 
PSMR 99.16 99.16  96.07 
96.07
MR MR  99.62 99.62  96.21
96.21 

The increase in CO conversion is due to the continuous removal of products (CO
The increase in CO conversion is due to the continuous removal of products (CO2 ,2H , H
2 )2) in portions 
in portions
of the MR, which constitutively relieves equilibrium limitations on the WGS reaction; the increase in 
of the MR, which constitutively relieves equilibrium limitations on the WGS reaction; the increase
inHH 2  recovery 
2 recoveryis isdue 
dueto 
to the 
the higher 
higher H H22  partial
partial pressure
pressure differences
differences between
between reactor/sweep
reactor/sweep streams
streams 
achieved through
achieved through selective
selective placement
placement ofof membrane
membrane inin a a  reactor 
reactor inin  which 
which reaction/diffusion occur
reaction/diffusion occur 
simultaneously. With regard to carbon capture, the MR achieves higher values at lower selectivity 
simultaneously. With regard to carbon capture, the MR achieves higher values at lower selectivity
(α(αHH22/CO
/CO2 ≤ 50) due to lower CO
2
ď 50) due to lower CO2 partial pressure differences between the tube/shell (a positive factor 
2 partial pressure differences between the tube/shell (a positive factor
for carbon capture, but this works against the operator for H
for carbon capture, but this works against the operator for H22 recovery); at higher values of selectivity, 
recovery); at higher values of selectivity,
the 3‐stage PSMR and MR produce nearly the same values for carbon capture, with the MR exceeding 
the 3-stage PSMR and MR produce nearly the same values for carbon capture, with the MR exceeding
the 3‐stage PSMR for α
the 3-stage PSMR for αHH22/CO /CO2 = 55–75 by a very small margin. 
2
= 55–75 by a very small margin.
From the 2‐stage and 3‐stage results, improved performance is the result of increasing the stage 
From the 2-stage and 3-stage results, improved performance is the result of increasing the stage
number, which allows for implementation of membrane stages at increased H
number, which allows for implementation of membrane stages at increased H2 2partial  partial pressures (see 
pressures (see
Figure 6 vs. Figure 9 and Figure 7 vs. Figure 10 for such configurations) As the stage number increases 
Figure 6 vs. Figures 7 and 9 vs. Figure 10 for such configurations) As the stage number increases
dramatically to the point at which the number of stages bring the reactor design to the limiting case 
dramatically to the point at which the number of stages bring the reactor design to the limiting case
(infinitesimally small
(infinitesimally small stage
stage lengths),
lengths), the
the design
design equation
equation associated
associated with
with plug-flow
plug‐flow becomes
becomes thatthat 
associated with a continuous‐flow stirred tank reactor (CSTR). As each stage operates as a CSTR and 
associated with a continuous-flow stirred tank reactor (CSTR). As each stage operates as a CSTR and
anan infinite sequence of algebraic stages (infinite stage number reactor) may be taken to represent the 
infinite sequence of algebraic stages (infinite stage number reactor) may be taken to represent the
differential reactor (the MR), large stage numbers cause the PSMR design to approach that of the MR, 
where the MR represents the maximum achievable performance for a given set of conditions. The 
Processes 2016, 4, 8 17 of 19

differential reactor (the MR), large stage numbers cause the PSMR design to approach that of the MR,
where the MR represents the maximum achievable performance for a given set of conditions. The
trend for increasing stage number in PSMRs leading to operation resembling that of an MR is present
as early as 2/3 stages for the case considered herein (Figures 8 and 11); four or more stages should
yield results that increasingly resemble the MR. The ratio of reaction to diffusion rate in the membrane
reactor configuration (see Figure 14) suggests that for cases in which minimization of membrane area
is desired at fixed catalyst mass, more densely packing the catalyst (compared to spreading the fixed
mass across the entire reactor length, as in the base case) can achieve this aim, provided temperature
limitations are not present; alternatively for cases in which there is unlimited catalyst, dense packing
throughout the entire reactor may serve to improve reactor performance.
In general, the supported tubular membrane platform benchmarked here is a desirable platform
for the MR configuration as it allows efficient catalyst packing and the mechanical strength required to
contain catalyst material. However, based on the results of this study, the performance of an optimized
3-staged PSMR can be comparable to that of a MR. This result presents an exciting opportunity for
lower cost high performance hollow fiber membranes in this application. Given the high surface area to
volume ratio of hollow fiber membrane modules and their resulting dramatically reduced containment
vessel/module size, their cost per m2 is estimated to be an order of magnitude lower than the tubular
platform benchmarked here. As the optimization results presented here are sensitive to membrane
cost, the influence of such membrane cost reduction opportunities on the process optimization will be
explored in future studies.

4. Conclusions
One-dimensional isothermal models were developed for traditional MRs and PSMRs, and used
to assess high-performance polymeric membrane reactor systems. Constrained cost optimization
problems were formulated so as to systematically determine optimal reactor designs through more
efficient membrane placement. As the solutions to these optimization problems also corresponded
to a maximization of the limiting performance parameter, CCO2 , an incremental search of H2 /CO2
selectivity was then performed with the intent of determining CCO2 as a function of αH2 /CO2 at constant
permeance. These designs were generated through the cost minimization optimization problem and the
result that the economic optimum corresponds to maximization of carbon capture was verified by way
of the second optimization problem formulation (utilizing identical decision variables), which sought
to maximize CCO2 subject to the other five performance constraints. This analysis was successfully
completed for the three reactor designs considered (2-stage PSMR, 3-stage PSMR, MR). Graphs of
carbon capture as a function of selectivity for fixed conditions were produced, which can guide the
development of polymeric membrane materials to achieve all the desired specifications for their
implementation in IGCC WGS environments.
Using the unit design framework considered herein, one may generalize from the process
conditions of feed/sweep molar flow rate, flow composition/arrangement, reactor operating
conditions (temperature, pressure), catalyst mass, tube dimensions and membrane properties to
grant insight into future membrane material development by identifying the limiting parameter and
determining a minimum value that satisfies all imposed constraints. Having identified a minimum
value for a given parameter (αH2 /CO2 in this case), a clear goal can be set for researchers in material
development (should it be desired to use the process designs considered here). As the optimization
results presented here are sensitive to membrane cost, it is desirable to investigate membranes with
varied cost (i.e., hollow fibers) to further understand the resulting outcomes in terms of required
performance characteristics and optimized PSMR design. To that end, the presented modeling
framework can be extended to evaluate performance of membrane materials in a systematic manner
by considering several process designs in which material placement (catalyst, membrane) is guided
by economic considerations and/or satisfaction of a set of performance constraints. The formulated
Processes 2016, 4, 8 18 of 19

optimization problem can also be extended to consider different operating conditions (temperature,
pressure) for each reaction/separation module.

Acknowledgments: The authors (Andrew J. Radcliffe and Fernando V. Lima) would like to acknowledge
West Virginia University for their financial assistance on this work. The authors (Rajinder P. Singh and
Kathryn A. Berchtold) gratefully acknowledge the U.S. DOE/NETL-Strategic Center for Coal: Carbon Capture
Program for financial support of the project under contract LANL-FE-308-13. Los Alamos National Laboratory
is operated by Los Alamos National Security, LLC for DOE/NNSA under Contract DE-AC52-06NA25396.
Additionally, the authors would like to thank Juan Carlos Carrasco for his assistance on illustrations.
Author Contributions: This paper is a collaborative work among the authors. A.J.R. performed all simulations
and wrote the paper. F.V.L. helped with the paper writing and supervised all the technical aspects of the work.
R.P.S. and K.A.B. assisted in manuscript preparation and provided technical guidance for the performed work.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Marbán, G.; Valdés-Solís, T. Towards the hydrogen economy? Int. J. Hydrogen Energy 2007, 32, 1625–1637.
[CrossRef]
2. Lima, F.V.; Daoutidis, P.; Tsapatsis, M.; Marano, J.J. Modeling and optimization of membrane reactors for
carbon capture in integrated gasification combined cycle units. Ind. Eng. Chem. Res. 2012, 51, 5480–5489.
[CrossRef]
3. Woods, M.C.; Capicotto, P.J.; Haslbeck, J.L.; Kuehn, N.J.; Matuszewski, M.; Pinkerton, L.L.; Rutkowski, M.D.;
Schoff, R.L.; Vaysman, V. Cost and Performance Baseline for Fossil Energy Plants. Volume 1: Bituminous Coal and
Natural Gas to Electricity; Final Report, Technical Report Revision 1, DOE/NETL-2007/1281. U.S. Department
of Energy: Washington, DC, USA, 2007.
4. Lima, F.V.; Daoutidis, P.; Tsapatsis, M. Modeling, optimization, and cost analysis of an IGCC plant with a
membrane reactor for carbon capture. AIChE J. 2016. [CrossRef]
5. Berchtold, K.A.; Singh, R.P.; Young, J.S.; Dudeck, K.W. Polybenzimidazole composite membranes for high
temperature synthesis gas separations. J. Membr. Sci. 2012, 415–416, 265–270. [CrossRef]
6. Singh, R.P.; Berchtold, K.A. Chapter 6—H2 Selective Membranes for Precombustion Carbon Capture. In Novel
Materials for Carbon Dioxide Mitigation Technology; Morreale, F.S., Ed.; Elsevier: Amsterdam, The Netherlands,
2015; pp. 177–206.
7. Li, X.; Singh, R.P.; Dudeck, K.W.; Berchtold, K.A.; Benicewicz, B.C. Influence of polybenzimidazole main
chain structure on H2 /CO2 separation at elevated temperatures. J. Membr. Sci. 2014, 461, 59–68. [CrossRef]
8. Zou, J.; Ho, W.S.W. Hydrogen purification for fuel cells by carbon dioxide removal membrane followed by
water gas shift reaction. J. Chem. Eng. Japan 2007, 40, 1011–1020. [CrossRef]
9. Huang, J.; El-Azzami, L.; Ho, W.S.W. Modeling of CO2 -selective water-gas-shift membrane reactor for
fuel cell. J. Membr. Sci. 2005, 261, 67–75. [CrossRef]
10. Singh, R.P.; Dahe, G.J.; Dudeck, K.W.; Welch, C.F.; Berchtold, K.A. High temperature polybenzimidazole
hollow fiber membranes for hydrogen separation and carbon dioxide capture from synthesis gas.
Energy Procedia 2014, 63, 153–159. [CrossRef]
11. Merkel, T.C.; Zhou, M.; Baker, R.W. Carbon dioxide capture with membranes at an IGCC power plant.
J. Membr. Sci. 2012, 389, 441–450. [CrossRef]
12. Caravella, A.; di Maio, F.P.; di Renzo, A. Optimization of membrane area and catalyst distribution in
a permeative-stage membrane reactor for methane steam reforming. J. Membr. Sci. 2008, 321, 209–221.
[CrossRef]
13. Caravella, A.; di Maio, F.P.; di Renzo, A. Computation study of staged membrane reactor configurations for
methane steam reforming. I. Optimization of stage lengths. AIChE J. 2010, 56, 248–258. [CrossRef]
14. Caravella, A.; di Maio, F.P.; di Renzo, A. Computation study of staged membrane reactor configurations for
methane steam reforming. II. Effect of number of stages and catalyst amount. AIChE J. 2010, 56, 259–267.
[CrossRef]
15. Adrover, M.E.; López, E.; Borio, D.O.; Pedernera, M.N. Theoretical study of a membrane reactor for the
water gas shift reaction under nonisothermal conditions. AIChE J. 2009, 55, 3206–3213. [CrossRef]
Processes 2016, 4, 8 19 of 19

16. Choi, Y.; Stenger, H.G. Water gas shift reaction kinetics and reactor modeling for fuel cell grade hydrogen.
J. Power Sources 2003, 124, 432–439. [CrossRef]
17. Kärger, J.; Ruthven, D.M. Diffusion in Zeolites and Other Microporous Solids; Wiley-Interscience: New York, NY,
USA, 1992.
18. Ghosal, K.; Freeman, B.D. Gas separation using polymer membranes: An overview. Polym. Adv. Technol.
1994, 5, 673–697. [CrossRef]
19. Baker, R.W. Membrane Technology and Applications; John Wiley & Sons: Hoboken, NJ, USA, 2012; p. 152.
20. Vora, S.; Brickett, L. DOE/NETL Advanced Carbon Dioxide Capture R & D Program: Technology Update;
U.S. Department of Energy/National Energy Technology Laboratory: Washington, DC, USA, 2013.
21. Marano, J.J. Integration of H2 Separation Membranes with CO2 Capture and Compression; Report to DoE, Contract
No. DE-AC2605NT41816, DOE/NETL-401/113009; U.S. Department of Energy/National Energy Technology
Laboratory: Washington, DC, USA, 2009.

© 2016 by the authors; licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons by Attribution
(CC-BY) license (http://creativecommons.org/licenses/by/4.0/).