-
Membrane in a Reactor: A Functional Perspective
Kamalesh K. Sirkar,* Purushottam V. Shanbhag, and A. Sarma
Kovvali
Center for Membrane Technologies, Department of Chemical
Engineering, Chemistry and EnvironmentalScience, New Jersey
Institute of Technology, Newark, New Jersey 07102
Membrane reactors have found utility in a broad range of
applications including biochemical,chemical, environmental, and
petrochemical systems. The variety of membrane separationprocesses,
the novel characteristics of membrane structures, and the
geometrical advantagesoffered by the membrane modules have been
employed to enhance and assist reaction schemesto attain higher
performance levels compared to conventional approaches. In these,
membranesperform a wide variety of functions, often more than one
function in a given context. Anunderstanding of these various
membrane functions will be quite useful in future developmentand
commercialization of membrane reactors. This overview develops a
functional perspectivefor membranes in a variety of reaction
processes. Various functions of the membranes in a reactorcan be
categorized according to the essential role of the membranes. They
can be employed tointroduce/separate/purify reactant(s) and
products, to provide the surface for reactions, to providea
structure for the reaction medium, or to retain specific catalysts.
Within these broad contexts,the membranes can be
catalytic/noncatalytic, polymeric/inorganic, and ionic/nonionic and
havedifferent physical/chemical structures and geometries. The
functions of the membrane in areaction can be enhanced or increased
also by the use of multiple membrane-based schemes.This overview
develops a perspective of each membrane function in a reactor to
facilitate a betterappreciation of their role in the improvement of
overall process performance.
1. Introduction
Membrane reactors have been investigated since the1970s. The
early investigations employed primarilypolymeric membranes and
enzymatic reactions. Laterinvestigations show an abundance of
petrochemicallyrelevant systems and inorganic membranes. Whole
cellfermentation-based chemical and biochemical produc-tions as
well as degradation of pollutants biologicallyor otherwise have
also been studied in membranereactors. Polymer membrane-based
reactors have beenblessed with some commercial success. Much of
thisresearch has been discussed in a number of reviews.1-9
In these investigations a variety of membrane separa-tion
processes as well as membranes have been used.More importantly, the
membrane inside the reactor hasserved a variety of functions. In
some studies, themembrane has a single well-defined function. In
others,the membrane allows two or more functions to becarried out.
The variety of functions achievable via amembrane in a reactor is
very broad. An understandingof the breadth of the roles capable of
being performedby a membrane is likely to be quite useful in the
futuredevelopment of membrane reactors. This overviewproposes to
develop a functional perspective of a mem-brane(s) in a reactor.
This perspective is developed byemploying a variety of contexts
including differentmembrane separation processes, different
membranes,chemical/electrochemical reactions, enzymatic
processes,fermentations, catalyst immobilization/segregation,
cata-lytic membranes, integration of functions, etc.
A brief enumeration of different membrane separationprocesses
and different classes of membranes investi-
gated in the literature is useful in the Introductionbefore we
present the functional perspective. Of themany types of membrane
separation processes andmembrane-based equilibrium separation
processes avail-able for separation,10 membrane reactors have
beenstudied using the following: reverse osmosis
(RO),nanofiltration (NF), ultrafiltration (UF),
microfiltration(MF), electrodialysis (ED), liquid membranes
(LM),pervaporation (PV), gas permeation, vapor permeation,molecular
sieving, Knudsen diffusion (and moleculardiffusion), gas membrane,
membrane solvent extraction,and membrane gas
absorption/stripping.
An extraordinary variety of membranes have alsobeen used.
Membranes are employed in gross physicalforms as flat films, hollow
fibers, tubules, and tubes,while their physical structures can be
as follows: mi-croporous symmetric and asymmetric membranes,
non-porous membranes, and composite membranes. Mem-branes can be of
the polymeric variety or be inorganicin nature, which would include
zeolitic, ceramic, andmetallic membranes. Membranes can also
conductelectrical charges and can be chosen from one of follow-ing
categories: ion-exchange membranes, bipolar mem-branes, mixed
conducting membranes, proton-conduct-ing membranes, etc. In many
cases, the membraneshave catalysts incorporated in their porous
structure oron the surfaces. The membranes in such cases aretermed
as catalytic membranes. Of course, the mem-brane can be catalytic
by itself without the addition ofany catalyst materials from
external sources. The termcatalytic membrane reactor sometimes
includes theabove cases as well as a catalytic reactor enclosed by
amembrane, which is noncatalytic.2
In the next section, we will first present a compactlist of
membrane functions in a reactor. Often, thegeneric membrane
function identified will affect thereaction processes in different
ways. Such effects on the
* To whom correspondence should be addressed. Tel.:
(973)596-8447. Fax: (973) 642-4854. E-mail:
[email protected].
Current address: Compact Membrane Systems, Inc.,Wilmington, DE
19804.
3715Ind. Eng. Chem. Res. 1999, 38, 3715-3737
10.1021/ie990069j CCC: $18.00 1999 American Chemical
SocietyPublished on Web 09/18/1999
-
reaction will be identified. Each generic membranefunction will
then be illustrated in a separate subsectionusing examples from the
literature. This illustrativeexposition will employ different
membrane separationprocesses, reaction-phase systems, and other
distin-guishing features to elaborate briefly on particularmembrane
reactors. The goal is to develop a perspectiveon the range and the
utility of each membrane functionin a membrane reactor rather than
a review of allinvestigations. Frequently, a membrane (or two
mem-branes) incorporated in a reactor serves more than onedesired
function, only one of which may involve amembrane separation
process where membrane fluxand selectivity are important. The
development of sucha multifunctional perspective of membranes in a
reactoris an additional objective of this paper.
The reaction processes of interest in this paper mayinvolve the
production of a particular chemical or abiochemical product.
Alternately, it may involve thedestruction of some organic species
in a phase for thepurpose of controlling environmental pollution.
Reactionprocesses are sometimes employed to purify a
particularfluid stream without destroying the undesirable
speciesinto simple compounds such as CO2, H2O, etc. (e.g.,enzymatic
resolution processes). Although such pro-cesses are within the
scope of this paper, our main focuswill be on those reactive
processes where a particularcompound or two are produced by the
reactions. Wespecifically exclude those processes where reactions
areused to enhance separation of a mixture.
2. Membrane Functions in a Reactor
Figure 1 schematically identifies many of the majorgeneric
functions performed by a membrane in a reactor.One should not
conclude from the figure that a given
membrane in a given reactor is capable of all
functionsidentified in the figure. However, a given membraneunder
appropriate circumstances can perform more thanone generic
function. The introduction of another mem-brane into the reactor
can increase the number ofgeneric membrane functions in the reactor
or achievethe same generic membrane function vis-a`-vis someother
species. Figure 1 also indicates other activitiesconcurrently
taking place in the so-called nonreactor (orpermeate) side of the
membrane as well as in the reactorside of the membrane. A list of
the generic membranefunctions performed by a membrane or two in a
reactoris provided next:
2.1. Separation of products from the reaction mixture2.2.
Separation of a reactant from a mixed stream forintroduction into
the reactor2.3. Controlled addition of one reactant or two
reac-tants2.4. Nondispersive phase contacting (with reaction atthe
phase interface or in the bulk phases)2.5. Segregation of a
catalyst (and cofactor) in areactor2.6. Immobilization of a
catalyst in (or on) a mem-brane2.7. Membrane is the catalyst2.8.
Membrane is the reactor2.9. Solid-electrolyte membrane supports the
elec-trodes, conducts ions, and achieves the reactions onits
surfaces2.10. Transfer of heat2.11. Immobilizing the liquid
reaction medium
Membranes in a reactor existing as membrane lami-nates or
physically separated membranes with a fluidphase between have also
been studied. They can provideparticular combinations of the above
functions some-times with added and novel benefits;5,11 these
novelbenefits include product separation and
simultaneousconcentration, separation of multiple products,
reactionintensification, and physically containing the
reactionmedium in multiphase reaction systems.
Before proceeding further, it is necessary to point outthat
there are many studies where the membrane isphysically located in a
device external to the reactorproper (structure). The reaction
medium is then circu-lated over the membrane and back to the
reactor in arecycle mode (Figure 2). This configuration is
frequentlyemployed in reaction processes based on enzymes andwhole
cells; it is also being proposed for organic syn-theses. The
reactor vessel in such case is sometimesoperated as a batch reactor
or more frequently as acontinuous stirred tank reactor (CSTR). In
many cir-cumstances, the system behavior here can be considered
Figure 1. Schematic of possible functions of a membrane in
areactor.
Figure 2. Schematic of a recycle-based configuration of a
coupledreactor and membrane separator.
3716 Ind. Eng. Chem. Res., Vol. 38, No. 10, 1999
-
to be de facto equivalent to that with a membrane insidea
reactor. Therefore, such recycle membrane reactorswill also be
included in the following treatment: pri-mary emphasis will however
be on systems where amembrane (or two) is physically located in the
reactor.
One must recognize the major advantages of thesedifferent
arrangements:
(1) The mixing conditions and the flow velocities (andtherefore
the extent of consequent concentration polar-ization in membrane
devices involving liquid-phasesystems) can be maintained at
different levels in thereactor and the membrane separator if
recycle mem-brane reactors are employed; conditions can be
opti-mized for each. The reactor may require long residencetimes
whereas the membrane device may need a shortresidence time.
(2) Building a reactor with a membrane in it or usinga membrane
device as the reactor can sometimes be verydemanding on the
membrane, especially for highertemperature systems. The recycle
membrane reactorallows the reactor and the membrane unit to
operateat two different temperatures by using heat exchangersin
between.
(3) Recycle membrane reactors allow use of existingequipment,
namely, a separate reactor and a separatemembrane device.
(4) For fast reactions, the membrane in a reactor islikely to be
a more desirable configuration.
2.1. Separation of Products from the ReactionMixture. Separation
of products from the reactionmixture is one of the most common
functions of amembrane in a reactor. The separation may be
purifica-tion, enrichment, or concentration. Consider the
follow-ing elementary reversible reaction (see Figure 1):
where D is a product needed to be removed via themembrane to the
permeate side. The separation processemployed may produce a
permeate side stream wherethe mole fraction of D is much higher
than that in thereactor side. In the case of species D being H2 and
apalladium membrane, pure H2 is obtained in the perme-ate side. The
H2 partial pressure on the permeate sidewill be lower than the
partial pressure of H2 on the feedside. If species C also permeates
to some extent throughthe membrane, the permeate stream is enriched
in Dvis-a`-vis the product species C: permeation of D leadsto
partial purification of the product C in the reactoroutlet
stream.
Removal of D via the separation function of themembrane has the
following effects on reaction (1) andthe reactor performance:
(a) The equilibrium condition indicated in the revers-ible
reaction (1) is shifted to the right, i.e., leading tohigher
equilibrium conversion of A and B to C and D.
(b) If there is an undesirable side reaction as shownbelow,
taking place in the reactor (see Figure 1), the separationof
product D from the reaction mixture reduces the lossof reactant B
to the side reaction, increasing theselectivity of conversion to
product C (or D) (modeledby Whu et al.12 for nanofiltration-aided
liquid-phaseorganic synthesis). An experimental study by Raich
andFoley13 of ethanol dehydrogenation in a palladium
membrane reactor whereby the product H2 is withdrawnthrough the
palladium membrane to shift the reaction
to the right showed that the deleterious effect of the
sidereaction
can be drastically reduced, provided the reaction (3a)is shifted
to the right via the Cu/SiO2-aq catalyst andH2 removal by the
membrane.
(c) In consecutive catalytic reactions,
where B is the desired intermediate product, if the rateconstant
for reaction 4b is significantly larger than therate constant for
reaction 4a, it is difficult to achieve ahigh selectivity to B
using a conventional packed bed,plug flow reactor. By using an
inert sweep gas on theoutside of a permeable tube having the
catalysts andthe reaction taking place inside the tube, the
intermedi-ate product B may be selectively removed from thereaction
zone, leading to increased selectivity.14 Removalof the
intermediate products (methanol/formaldehyde)via a membrane in the
partial oxidation of methane isan example; this strategy will
prevent further oxidationof these products to CO and CO2.
(d) In fermentation processes, one of the products maybe
inhibitory to the fermentation process. Removal ofthe product from
the fermentation broth via a mem-brane can substantially reduce
product inhibition andincrease volumetric productivity of the
fermentor.15,16Further, one can use higher concentrations of
thesubstrate in the feed (e.g., glucose for ethanol fermenta-tion)
since the product is being removed as it is beingformed.17
The separation of a reaction product(s) (C or D orboth) can be
implemented using a variety of membraneprocesses. The nature of the
membrane process isobviously influenced by the phase of the
reactionmedium exposed to the membrane and the desiredphase of the
permeated product stream. Examples ofsuch processes will be
provided under two categories,namely, (1) liquid reaction
medium/liquid feed phaseand (2) gaseous reactions/gaseous feed
phase.
2.1.1. Separation from a Liquid Reaction Mix-ture. We will
briefly mention and/or illustrate the useof the following membrane
separation processes forremoving products from the liquid reaction
medium:reverse osmosis, nanofiltration, ultrafiltration,
pervapo-ration, gas membranes, electrodialysis, and liquid
mem-branes.
2.1.1.1. Reverse Osmosis. Vasudevan et al.18 havedescribed a
membrane sandwich reactor in which theSaccharomyces cerevisae (ATCC
4126) cells were ef-fectively placed between an ultrafiltration
(UF) mem-brane and a reverse osmosis (RO) membrane; thereactor was
fed with a solution of glucose at a highpressure from the UF
membrane side and the productsolution was forced out through the RO
membrane. Theproduct solution concentration progressively
increasedin ethanol; the glucose in the feed solution was
ef-fectively rejected by the RO membrane. The reactorstructure
shown in Figure 3 has a microfiltration
C2H5OH T CH3CHO + H2 (3a)
C2H5OH + CH3CHO T CH3COOC2H5 + H2 (3b)
A f B (4a)
B f C (4b)
A + B T C + D (1)
B + D T E (2)
Ind. Eng. Chem. Res., Vol. 38, No. 10, 1999 3717
-
membrane and a coarse filter paper on two sides of theyeast
cells between the UF and the RO membrane toimmobilize and provide
physical protection in a high-pressure (up to 400 psig)
environment. The membranesin this reactor separate the product
ethanol from thereactant glucose and effectively immobilize the
biocata-lyst (whole cell).
2.1.1.2. Nanofiltration. Whu et al.12 have modeledthe
performance of a semibatch/batch reactor coupledto an external
nanofiltration (NF) unit for the synthesisof the desired product C
(a hydroxyester of MW 400)from reactants A (a diketone of MW 400)
and B (analkoxide of MW 40-100) present in an organicsynthesis
solvent, methanol. The membrane removesthe low-molecular-weight
product, D (MW 40-100)and the solvent which significantly improves
the selec-tivity if the reaction system consists of the reactions
(1)and A + D T E. Figure 4 illustrates the role of themembrane in
improving the selectivity of the reactionfor the product C. This
figure shows that a much higher
selectivity is achieved when the semibatch reactor isexternally
coupled with a nanofiltration unit to removethe solvent and the
product D (in the manner of Figure2). Operation as a continuous
flow stirred tank reactoras in Figure 2, coupled with a NF unit,
could alsoprovide a way to increase the concentration of
thereactants in the reactor from a dilute feed if thereactants are
rejected by the NF membrane. In such acase, separation of products
by the NF unit mayfacilitate conversion of the reactants.
2.1.1.3. Ultrafiltration. Cheryan and Mehaia6 haveprovided a
comprehensive review of enzyme-based andwhole-cell-based membrane
bioreactors where ultrafil-tration is often the predominant mode of
membraneseparation. Many systems have been described. Someexamples
are hydrolysis of proteins leading to modifiedproteins with smaller
molecular weights appearing inthe permeate; hydrolysis of
carbohydrates, e.g., starch,cellulose to produce lower molecular
weight sugars;hydrolysis of sugars, e.g., lactose. Perhaps the
earliestexperimental study was carried out in a stirred tankreactor
coupled to an ultrafiltration membrane cell forthe hydrolytic
breakdown of cellulose to the membranepermeable product glucose
using cellulase enzymes.19The utility of a thin channel membrane
reactor linedwith UF membranes for the enzymatic reduction ofstarch
to glucose is shown in Figure 5. The figureillustrates the
lengthwise variation of the performanceindicator fA of the plug
flow membrane reactor definedby
with that for a solid tube reactor; the membrane reactorhas a
much higher conversion for the cases analyzedwhere the enzyme is
completely rejected.20
2.1.1.4. Pervaporation. The pervaporation (PV)process is used to
remove volatile reaction products fromthe reaction mixture;
generally, a vacuum on the
Figure 3. Membrane sandwich: (1) UF membrane, (2) coarsefilter
paper (>10 m), (3) cell mass, (4) fine filter (0.2
m,microporous), and (5) RO membrane (reprinted from Vasudevanet
al.18 with permission).
Figure 4. Selectivity with respect to the desired product C
(SC)as a function of time. (SC, semibatch-coupled with membrane;
SU,semibatch-uncoupled; BU, batch-uncoupled) (reprinted from Whuet
al.12 Copyright 1999, with permission from Elsevier Science).
Figure 5. Reactor performance as a function of enzyme
concen-tration and distance along the membrane for enzymatic
reductionof starch to glucose; uj0 ) 1 cm/s (reprinted with
permission fromClosset et al.20 Copyright 1973, John Wiley &
Sons, Inc.).
fA )(inlet flow of reactant ( flow of reactantacross the tube
wall - flow of reactant
at a given axial distance)inlet flow of reactant
(5)
3718 Ind. Eng. Chem. Res., Vol. 38, No. 10, 1999
-
permeate side is employed to create the needed partialpressure
difference. The most common reaction systemstudied for the
application of pervaporation is anesterification reaction between
an alcohol and an acidin the presence of a highly acidic catalyst
(e.g., concen-trated sulfuric acid):
This reversible reaction in industrial processing isdriven to
high conversion by adding a large excess ofalcohol. Adding a
poly(vinyl alcohol) (PVA)-based water-selective PV membrane to the
esterification reactorallows one to shift the equilibrium in
reaction 6 to theright (thus reducing the need for excess alcohol
beyondthat needed for solubilization of the acid). Figure
6illustrates how the conversion to the ester is increasedin a batch
reactor with time as the ratio of themembrane surface area (S) to
the reaction volume (V)is increased for the esterification reaction
betweenpropionic acid and 1-propanol studied by David et al.21Note
that in this case the equilibrium value of theconversion in the
absence of any product removal is 0.7;the membrane allows a much
higher conversion to beattained much more rapidly. If the
temperature of theesterification reaction is high, it will be
necessary toemploy vapor permeation membranes to remove H2Ovapor.
Another application of pervaporation studied22involved selective
removal of alcohol from a fermenta-tion broth in a recycle
configuration using siliconecapillary membranes. Silicone (PDMS) is
a highlybiocompatible material. Further, some of the
otherbyproducts of fermentation (which can inhibit thefermentation
if their concentrations build up in a recyclesystem) like
acetaldehyde, butanol, etc. are also easilyremoved. Removal of
ethanol decreased product inhibi-tion and increased fermentor
productivity.
2.1.1.5. Gas Membranes. A hydrophobic microporousor porous
membrane having gas-filled pores and twononwetting aqueous
solutions on two sides will allowspontaneous transfer of volatile
species from one solu-tion to the other solution through the pore
as long asthere is a partial pressure difference.23 Such a gas
membrane-mediated selective transfer of volatile spe-cies has
been used to remove volatile product speciesfrom a reactor solution
to that on the receiving side.Twardowski and McGilvey24 have used a
porous poly-(tetrafluoroethylene) (PTFE) membrane to
transferproduct ClO2, from the reactant stream in a reactor tothe
aqueous solution on the other side of the membranewhere it is
ultimately used to bleach wood pulp, etc.Removal of the ClO2 from
the reactor solution isnecessary to reduce the high partial
pressure of ClO2inevitably occurring in the gas space of commercial
ClO2generators which leads to localized decomposition ofClO2.
2.1.1.6. Electrodialysis. In electrodialysis processesusing
bipolar membranes, a solution of a salt, e.g., NaCl,is converted to
a pure solution of NaOH and anotherpure solution of HCl. This acid
and base production iscarried out first by the production of H+ and
OH- ionsfrom water and their collection into separate
aqueoussolutions into which are fed the corresponding ions,
viz.,Cl- and Na+, respectively, via the electrodialysis pro-cess.
The splitting of water into H+ and OH- ions inseparate aqueous
solutions is carried out in a bipolarion-exchange membrane-based
reactor shown in Figure7.25 The thin space between a
cation-exchange mem-brane and an anion-exchange membrane
laminatedtogether and placed between a cathode and an anode
isfilled with water. Any ions, e.g., Na+ and Cl-, in thiswater are
quickly removed through the correspondingion-exchange membrane.
This leads to deionized waterin the space between the two laminated
ion-exchangemembranes. The resistance of the aqueous
solutionbecomes very high, which leads to H+ and OH-
ionsparticipating in the transport of electrical chargesthrough the
membranes. In the water dissociationequilibrium process,
As the ions H+ and OH- are removed through thecation-exchange
membrane and anion-exchange mem-
Figure 6. Influence of the variation of the membrane area to
thesolution volume ratio on the esterification rate of propionic
acidwith 1-propanol. T ) 50 C; 1 wt % catalyst; Noac/Noalc ) 1
(withpermission21).
no. 1 2 3 4S/V (cm-1) 1 2 4 8
R1COOHacid
+ R2OHalcohol
{\}Cat-H+ R1COOR2
ester+ H2O v (6)
Figure 7. Schematic diagram showing the configuration and
thebasic function of a bipolar membrane for water splitting
(reprintedfrom Strathmann.25 Copyright 1992, Kluwer Academic
Publish-ers).
H2O T H+ + OH- (7)
Ind. Eng. Chem. Res., Vol. 38, No. 10, 1999 3719
-
brane, respectively, more water is dissociated. Thus,separation
of products, H+ and OH-, through themembranes is essential to
continue the water-splittingreaction.
2.1.1.7. Liquid Membranes. Enantioselective hy-drolysis of
(R,S)-phenylalanine isopropyl esters by theenzyme subtilisin
Carlsberg in solution in a reactor wasshown by Ricks et al.26 to
selectively convert the (S)-ester to (S)-Phe-COO- while leaving the
(R)-esteressentially unchanged. Part of the enzymatic solutionwas
circulated on the shell side of a hollow fiber modulein which a
liquid membrane of N,N-diethyldodecana-mide in dodecane is
immobilized as a supported liquidmembrane in the pores of
hydrophobic Celgard polypro-pylene hollow fibers. The (R)-ester is
permeable throughthis liquid membrane into an acidic strip solution
whereit gets protonated and cannot partition back into thefeed
phase. The (S)-Phe-COO- is essentially imperme-able through the
liquid membrane and is recovered fromthe feed solution. A part of
the feed solution is recircu-lated through the reactor and the
membrane separatorin the recycle mode. If the (R)-ester accumulates
in thereaction media, it can inhibit the rate of conversion ofthe
(S)-ester, which can also permeate through themembrane. The
retention time of any ester fed to thereaction vessel is adjusted
such that hydrolysis isessentially completed (12 min) before the
solution isintroduced in the membrane module in the
recirculationmode (Figure 8). This is an example of an
unconvertedreactant being removed from the reaction mass as if
itwere an inhibitory product. Further, a reaction (in thiscase
protonation) is carried out in the permeate phase(Figure 1),
to maintain a large driving force for the permeation ofD.
2.1.2. Separation from a Gaseous Reaction Me-dium. The membrane
processes employed include thoseusing liquid membranes, Knudsen
diffusion, gas per-meation, molecular sieving, etc. We will briefly
identifysome typical examples to illustrate product removal
bymembranes in each case.
2.1.2.1. Liquid Membrane. One of the earlieststudies by Ollis et
al.27 involved acetaldehyde synthesisin a multiphase catalytic
liquid membrane-based reac-tor. The feed gaseous mixture of O2 and
C2H4 as theinner gas phase surrounded by an aqueous catalyticliquid
membrane layer was prepared by dispersing thegas as bubbles into an
aqueous liquid membranereservoir containing the palladium-based
catalyst PdCl2and the oxidizing agent CuCl2. The individual
gasbubbles surrounded by the aqueous liquid membranelayer then
spontaneously rose through a solvent layerkept above the aqueous
catalyst reservoir. Any reactionproduct was extracted into the
outer solvent layer. Theoverall reaction and the continuous
extraction of theproduct acetaldehyde into a solvent phase (e.g.,
ethylacetate, n-butanol, etc.) can be indicated by
The product acetaldehyde is recovered from the solventin a
separate flash drum, while the solvent phase isrecycled back to the
reactor to form the outer continuoussolvent phase in the
double-emulsion liquid membraneemployed in the reactor. The liquid
membrane in thiscase allows not only product separation but also
thesegregation of the soluble catalyst in the aqueousmembrane phase
(function 2.6) in addition to being theactual site where the
reaction is taking place (function2.8). The product separation here
does not lead to a pureproduct for which an additional step (flash)
is necessary.
2.1.2.2. Knudsen Diffusion. Since the mid 1980s, alarge number
of studies have been carried out usingreactants in the gas phase
and an inorganic membranethrough which one or more of the gaseous
products(usually H2) is withdrawn. The membranes used oftenwere
microporous/mesoporous, e.g., -alumina, Vycorglass, etc., with or
without a catalyst deposited in thepores. Tubular noncatalytic
membranes were also usedpacked with catalyst particles.
Sun and Khang28 have compared the performancesof a catalytic
membrane reactor, an inert membranereactor, and an ordinary reactor
without any membrane
Figure 8. Apparatus for the resolution of (R,S)-Phe-O-iPr-HCl
with subtilisin, using a hollow fiber SLM module (reprinted
withpermission from Ricks et al.26 Copyright 1992, American
Chemical Society).
D + F f G (8)
O2 (w) + C2H4 (w) f CH3CHO (w) fCH3CHO (s) (9)
3720 Ind. Eng. Chem. Res., Vol. 38, No. 10, 1999
-
for the dehydrogentation of cyclohexane:
They employed a porous Vycor glass membrane tubewith and without
a Pt catalyst in the pores (whichexhibited Knudsen flow for H2 and
N2 but had surfacediffusion and other effects for C6H6 and C6H12).
For thenoncatalytic membrane, catalyst particles were packedin the
tube. Their simulation results shown in Figure 9indicate that at
high space-time operations the cata-lytic membrane reactor is
superior to the inert mem-brane reactor which performs better than
a conventionalequilibrium-limited reactor without any selective
sepa-ration of products. Propane dehydrogenation studies ina
packed-bed membrane reactor by Ziaka et al.29employed a 40 -Al2O3
membrane in a porous multi-layered alumina tube containing
Pt/-Al2O3 catalystparticles; the membrane reactor at a residence
time of10 s provided 1.8 times higher propylene yield than
thecorresponding equilibrium conversion.
Numerical studies30 of catalytic dehydrogenation ofethylbenzene
using microporous ceramic membranespossessing Knudsen diffusion
behavior indicate only ag5% increase in styrene yield over the
thermodynamiclimit by a hybrid system, i.e., a fixed-bed
reactorfollowed by a membrane reactor. Numerical analysis byMohan
and Govind31 of a cocurrent packed bed mem-brane reactor suggest
among others, the use of amembrane with high permselectivity to
achieve largevalues of conversion; the loss of reactants to the
perme-ate side through a membrane with not-too-high aselectivity is
a serious concern. Hsieh1 has provided adetailed review of the many
investigations based onporous/mesoporous membranes.
2.1.2.3. Gas Permeation/Molecular Sieving. Aconsiderable number
of membrane reactor studies withgas-phase reactions have employed
denser membraneswith selectivities much higher than Knudsen
diffusionmembranes; these membranes include metallic mem-branes of
Pd, molecular sieve membranes, and densesilica membranes among
others.1 Catalytic dehydroge-nation of ethanol to acetaldehyde in a
palladium mem-brane reactor with H2 removal through the
membraneincreased ethanol conversion from 60% to nearly 90%with a
commensurate rise in selectivity to acetaldehyde
from 35% to 70%, moving the yield from 21% to 63%.13In the
catalytic dehydrogenation of propane using silicamembranes (having
H2/N2 selectivity of 10-19, muchhigher than Knudsen diffusion-based
selectivity), apropylene yield of 39.6% was obtained at 823 K
com-pared to a yield of 29.6% in a conventional packed-bedreactor
operated at the same flow rate.32 Catalyticisobutane
dehydrogenation in a dense silica membranereactor yielded somewhat
higher isobutene yield andselectivity than a conventional
reactor.33
Catalytic decomposition of NH3 in a gas feed to N2and H2 in a
composite Pd-ceramic membrane reactorachieved an NH3 conversion of
over 94% at 873 K and1618 kPa compared to 53% in a conventional
reactor.34A water-gas shift reaction carried out at 673 K in
apalladium membrane reactor enclosing an iron-chro-mium oxide
catalyst achieved higher CO conversionthan the equilibrium level
due to selective removal ofH2.35 A metal membrane-based catalytic
membranereactor for facilitating the water-gas shift reaction ata
temperature of 500 C was run successfully andcontinuously for 6
months for the economical productionof H2 in a 0.4 ft2
plate-and-frame module.36 Recovery ofH2 isotopes (tritium,
deuterium) is being studied in Pd/Ag-based membrane reactors from
impurities presentin fusion reactor exhaust streams.37 Dixon et
al.38 havemodeled the performance of membrane reactors havingmixed
conducting O2 permeable ceramic membranes forthe very high
temperature reactions:
Removal of O2 through the membrane resulted indramatic
improvements in conversion over the non-membrane tubular
reactors.
Armor8 has provided a critical review of the needs inmany
dehydrogenation membrane reactor applications.The problem areas
are: defects in metallic membranesat higher temperatures, phase
changes in metallicmembranes causing catastrophic failure, leakage
be-tween the membrane and device, low surface area perunit volume
of commonly used membranes, severemass-transfer limitations, very
low feed flow ratesresulting in high residence times, carbon
deposition onmembrane pores and surfaces, no method
(currentlyavailable) for repairing defective membranes in situ,
andthe low turnover number of commercially availabledehydrogenation
catalysts. The approach adopted byRezac et al.39 is of great
interest in this respect. Theypropose existing plug flow reactors
and heat-exchangeequipment to be used in series with an optimized
higher-temperature stable polyimide-ceramic composite H2removal
membrane module. Thus, C4H10 dehydrogena-tion is carried out at 480
C, the product mixture iscooled to 180 C, and H2 selectively
permeated throughthe polyimide-ceramic composite membrane
(H2/C4H10selectivity > 75). Unreacted C4H10 is passed on to
thesecond-stage reactor after heat exchange and so on. Thisresulted
in a substantial increase in conversion and ahigh selectivity for
the production of n-C4H8; membranepoisoning is also substantially
avoided.
2.2. Separation of a Reactant from a MixedStream for
Introduction into the Reactor. Figure1 identified a particular
function of the membrane aspurify reactant A from species F before
addition to the
Figure 9. Conversion vs temperature for high
space-timeoperations with a Vycor glass membrane for cyclohexane
dehy-drogenation (reprinted with permission from Sun and
Khang.28Copyright 1988, American Chemical Society).
C6H12 {\}Pt
C6H6 + 3H2 (10)
CO2 T CO +1/2O2 (11a)
NO2 T NO +1/2O2 (11b)
Ind. Eng. Chem. Res., Vol. 38, No. 10, 1999 3721
-
reactor on the left-hand side. The effect of this separa-tion on
the reaction system is generally quite differentfrom that of a
reaction product from the reactionmixture. The purification may
lead to pure A beingintroduced into the reactor; a direct effect of
this isprevention of dilution of the reaction mixture. It can
alsolead to rejection of a class of compounds by the mem-brane
while reactant species (one or a class) may beintroduced by the
membrane preferentially into thereactor from the feed stream; the
species rejected caninhibit the reaction. An additional possibility
involvessimultaneous operation of two different reactions on
twosides of the membrane wherein the products of onereaction feeds
the other and vice versa; the latter couldbe in a coupled mode as
well (to be explained later).
2.2.1. Pure O2 from Air. Mixed conducting denseceramic membranes
(see sections 2.3 and 2.9 as well)allow O2 transport from air to a
lower O2 partialpressure side without allowing N2 to be
transportedthrough the membrane. When such a membrane is usedwith
air on one side and CH4 on the other side, partialoxidation of CH4
to syngas can be carried out at a hightemperature (800 C) without
contaminating the reac-tion gas mixture with N2. Further, the need
for an O2plant is eliminated,40 improving the economics
consider-ably. When a similar dense ceramic membrane in
asolid-electrolyte-cell reactor (see section 2.9) is used,HCN is
produced in a tubular reactor fed with a NH3 +CH4 mixture as O2
permeates through the membrane41as an anion O2- into the reaction
zone; in this case airis fed on the outside of the tubular reactor
to produceoxygen throughout the reactor length (function 2.3).
Hadair been used directly, the auxiliary byproduct of H2 +CO, a
high-quality fuel stream, would have been dilutedwith N2.
2.2.2. Purify Organic Pollutants from Wastewa-ter for
Biodegradation. Point-source industrial waste-waters containing a
variety of priority pollutants wereoften considered recalcitrant
from a biodegradationperspective. These industrial wastewaters are
fre-quently released from organic synthesis and containhigh salts,
extreme pHs, and residual catalysts, all ofwhich are either singly
or jointly very harmful to thegrowth of microbial cultures used for
biodegradation.Brookes and Livingston42 have employed a
siliconecapillary membrane-based device to extract organicpriority
pollutants from these demanding wastewaters.The biological reaction
medium is circulated between abioreactor and one side of this
membrane device. Onthe other side of the silicone capillary
membranes flowsthe wastewater. Priority organic pollutants, e.g.,
aniline,4-chloroaniline, 3,4-dichloroaniline, etc., from the
waste-
water (having high pH 9-11) are partitioned throughthe silicone
membrane into the biological reactionmedium. Reactors operated over
3000 h show very highreductions of the pollutants without any form
of pre-treatment, pH adjustment, or dilution of the wastewa-ter.
The membrane essentially isolates the bioreactorenvironment from
the vagaries of the industrial waste-water properties as the
pollutants get extracted out intothe bioreaction medium for
degradation by the ap-propriate microorganisms. Successful pilot
plant studieshave been conducted using this technique.
2.2.3. Purify Organic Compound from a Synthe-sis Medium. To
prepare a concentrated aqueous solu-tion of diltiazem malate by
reacting diltiazem (sparinglysoluble in water) with malic acid, a
liquid membranewas utilized to recover diltiazem from an
aqueousreaction mixture containing diltiazem, NaCl, NaHCO3,etc. A
contained liquid membrane of decanol wasutilized by Basu and
Sirkar43 to extract diltiazem intothe reaction zone where it
reacted with L-malic acid inwater to form diltiazem malate (in
solution). When avery high concentration of L-malic acid is
maintained,the aqueous concentration of diltiazem malate
achievedcould be 3 orders of magnitude higher than the very
lowconcentration of diltiazem in the feed solution. Themembrane not
only facilitated production of a highlyconcentrated and purified
solution of diltiazem malatebut also avoided two steps of solvent
extraction and backextraction which was very problematic due to the
ordersof magnitude difference in the two aqueous phase flowrates,
namely, the very high flow rate of a very dilutefeed solution and
the very concentrated reactor effluenthaving a low flow rate.
2.2.4. Coupling of Two Chemical Reactions. Inenzymatic methods
for the production of pure enanti-omers from achiral precursors
using, say, dehydroge-nase enzymes, specialized and costly
reagents, e.g.,nicotinamide cofactors (NAD+ or NADH), are
required.A process for efficient and continuous regeneration
ofthese nicotinamide cofactors by a chemical reactionisolated from
the main reaction by a membrane parti-tion is illustrated in Figure
10.44 The membrane em-ployed is a gas membrane (see 2.1.1.5)
utilizing amicroporous hydrophobic polypropylene membrane
whosepores (0.03 m) are gas-filled. In the main reactioncompartment
on the left side, D-lactic acid is producedin an aqueous solution
by catalyzing the reduction ofpyruvic acid using D-lactic acid
dehydrogenase. TheNAD+ produced by this reaction reacts with
ethanol toregenerate NADH required for the main reaction. In
anadjacent compartment, sodium borohydride reductionof acetaldehyde
leads to ethanol which, being volatile,
Figure 10. Two reactions coupled through a gas membrane;
membrane-assisted synthesis (reprinted from Van Eikeren et al.44
withpermission).
3722 Ind. Eng. Chem. Res., Vol. 38, No. 10, 1999
-
enters the main reaction chamber whereas acetaldehydeproduced in
the regeneration of NADH in the mainreaction chamber enters the
adjacent chamber (throughthe membrane). Thus, the membrane allows
separationof the volatile species from the rest of the
nonvolatilereaction mixtures in both compartments and feeds
eachvolatile species at a controlled rate into the adjacentreaction
chamber (function 2.3).
A weaker coupling between two reactions on two sidesof a
membrane (see Figure 1 for A + B T C + D and D+ F f G where the
membrane controls the addition ofreactant D to the permeate side
reaction) was explorednumerically by Itoh and Govind.45 On the
basis of earlierproposals by Gryaznov and Itoh, they modeled
thecatalytic dehydrogenation of 1-butene to butadiene inthe main
packed bed reactor. The permeate side wasbeing continuously fed
with O2 in air which oxidizes thepermeated H2 to produce water and
liberate heat by anexothermic reaction. This reaction is
surface-catalyzedby the palladium membrane used to remove
hydrogenfrom the main reactor. The heat so generated is con-ducted
through the palladium membrane to the mainendothermic
dehydrogenation reaction. The membranehas many functions: separate
product of main reaction,H2 (function 2.1); provide purified
reactant (H2) tosecond reaction (function 2.2); add reactant H2 to
secondreaction in a controlled way (function 2.3); supply heatto
main endothermic reaction (function 2.10) and therebycontrol both
reactions; and catalyze the second reaction(function 2.7).
2.3. Controlled Addition of a Reactant or TwoReactants. Control
of the reaction pathway is a majorconcern in reaction engineering.
Partial oxidation reac-tions of hydrocarbons are especially
relevant here. Inparticular cases, possibilities of thermal runaway
andcatalyst poisoning do exist. In biodegradation processesfor
toxic organics, microorganism growth may be af-fected by inhibition
from the toxic organics unless theirconcentrations are controlled.
In an aerobic wastewatertreatment process, high O2 utilization with
minimumwaste to the atmosphere requires controlled but
efficientintroduction of O2 to the system. In processes
usingreactants having limited half-lives, e.g., ozonation
ofwastewater or for water purification, efficient andlocalized
introduction of O3 at a controlled rate can leadto higher O3
utilization. Using a membrane to introducea reactant or two in a
controlled fashion in the reactorcan facilitate achievement of the
desired reaction condi-tions. A number of examples provided below
will il-lustrate the role of a membrane in such processes.
2.3.1. Gas-Phase Reactions. Gas-phase reactionswhere reactants
are introduced into the reaction zoneby membranes at a controlled
rate(s) (see Figure 1,purify reactant A from species F before
addition) caninvolve three types of membranes: (1) inert
porousmembranes which provide no selectivity; (2)
microporousmembranes with some selectivity; (3)
nonporous/densemembranes having much higher selectivity. The
firsttwo types of membranes may employ catalysts in thepores; the
nonporous membrane can be inherentlycatalytic. All three types of
membranes have beenstudied with O2 as one of the reactants
introduced in acontrolled manner for partial oxidation or
oxidativedehydrogenation reactions.
Tonkovich et al.46,47 employed 50 -Al2O3 (effectivelayer) or 200
R-Al2O3 membrane tubes packed with amagnesium oxide catalyst doped
with samarium oxide
to study the oxidative dehydrogenation of C2H6 to C2H4at 600 C.
Air was introduced via permeation through-out the length of the
reactor from the outside of the tube.Controlling the ratio of C2H6
to O2 was found to becrucial to selectivity (with respect to CO2,
CO, etc.) andconversion. At low-to-moderate C2H6 to O2 feed
ratios(98%, CO selectivitywas 90%, and H2 produced was twice that
of CO.
Zaspalis et al.49,50 have experimentally studied
thedehydrogenation of methanol in a microporous -Al2O3membrane with
methanol fed from one side and O2 fromthe other (Figure 1,
reactants A and B introduced intothe pore from opposite sides).
This arrangement avoideddirect contact between the two streams
(i.e., an alcoholand an oxidant), thereby minimizing undesirable
gas-phase reactions. Further, the inner surface areas of
themembranes were used. Zaspalis et al.50 used silverparticles
deposited on the -Al2O3 membrane poresurfaces; CH3OH was fed from
one side and O2 fromthe other so that activation of the Ag catalyst
occurredsimultaneously with the methanol conversion to
form-aldehyde. This opposed flow mode of feeding two reac-tants
must be carried out in a controlled manner so asnot to decrease the
selectivity for the desired product.
2.3.2. Gas-Liquid Reactions. Catalytic reactionsbetween gas and
liquid phases pumped concurrentlydown a bed of catalyst particles
in conventional reactorsencounter mass-transport limitations due to
intrapar-ticle mass-transfer resistance, liquid film
resistance,liquid maldistribution, channeling, etc., apart from
hotspots and undesirable side reactions. To overcome theseproblems,
Cini and Harold51 have studied hydrogenationof R-methylstyrene
(diluted in mesitylene) to cumenein a porous (6 m) tubular -Al2O3
membrane impreg-nated with a Pd catalyst on the pore surface area.
H2was supplied on one side of the porous membrane andthe liquid
reactant flowed on the other side of the
Ind. Eng. Chem. Res., Vol. 38, No. 10, 1999 3723
-
membrane (Figure 1). The results demonstrated anefficient supply
of the volatile reactant, H2, which wasalso the limiting reactant
when compared with thecatalyst pellets conventionally used in a
trickle-bedreactor. There were no operational difficulties
encoun-tered in the membrane-based operation; further, therate
increased by up to a factor of 20. From a study ofhydrogenation of
nitrobenzene to produce aniline in atubular -Al2O3 membrane (50
pore size) having a Ptcatalyst deposited on the pore surfaces,
Torres et al.52have also concluded that catalytic membrane
reactorsare efficient for three-phase reactions as a result of
theeasy access of the gas to the catalytically active phase.
Biodegradation processes to destroy organic andinorganic
contaminants in air or water employ organ-isms in a biofilm
attached to a support. The efficientsupply of nutrients, e.g., O2
and pollutants, to thebiofilm is demanding without incurring
excessive powerconsumption and with minimum loss of O2 to
theatmosphere. Hydrophobic porous hollow fibers mem-branes are
especially useful. For example, Brindle etal.53 have immobilized a
biofilm on the fiber outerdiameter. The shell side is fed with,
say, the wastewatercontaining NH4+ which is oxidized by the
nitrifyingbiofilm into nitrite and then nitrate; the tube side is
fedwith pure O2. Exceptionally high O2 utilization efficien-cies
were achieved via efficient interfacial oxygentransfer to the
biocatalyst in the biofilm, even at lowO2 supply rates. Parvatiyar
et al.54 have demonstratedefficient biodegradation of toluene in a
hollow fibermembrane-based biofilter: air containing toluene wasfed
on the lumen side of porous polysulfone hollow fiberson the outer
surface of which the biofilm was im-mobilized. Nutrients were
supplied via an aqueousstream on the shell side. High conversion of
toluene wasachieved because of efficient contact of the
biomasscatalyst with O2 and toluene through the membranepores at
controlled rates of supply of the reactants inthe gas phase.
Noncatalytic gas-liquid reactions are employed usingO3 to
destroy pollutants and bacteria in water purifica-tion and
wastewater treatment processes. Since O3 hasa very low solubility
in water, kla (volumetric mass-transfer coefficient) controls the
mass-transfer rate ofO3 and thereby the reaction rate. Further, O3
has a verylimited half-life in a moist gaseous phase.
Membrane-based nondispersive ozonation studied by Shanbhag
etal.55,56 provides a much higher value of kla (g5 times)compared
to conventional dispersive bubble-based ozo-nation, eliminating the
possibility of gas-phase degrada-tion of the unstable O3 in bubbles
and efficientlybringing O3 in contact with the aqueous pollutants
alongthe length of reactor.
2.3.3. Liquid-Phase Reactions. Lee et al.57 carriedout
simultaneous biodegradation of the pollutants tolu-ene and p-xylene
in a completely mixed and convention-ally aerated bioreactor using
the microorganismPseudomonas putida. Under aerobic conditions,
themicroorganisms utilize toluene and xylene as carbonsources. The
pollutant species (toluene and p-xylene)were introduced into the
reactor in a controlled mannerthrough silicone capillary membranes.
The removalefficiency of these species increased at the
beginningwith an increase in the transfer rate of the
pollutantmixture (increased by, for example, the impeller speedand
not by the aeration rate or the circulation rate ofthe solvents in
the capillary); however, beyond a certain
rate, the removal efficiency decreased since the
limitingsubstrate shifted from carbon to O2. At this time,
thesolvent loss in the exit gas also increased. For givenimpeller
speeds, the membrane can be designed tocontrol the rate of
introduction of organic pollutants tothe biomass-containing
medium.
2.4. Nondispersive Phase Contacting. In manyreactions, aqueous
and organic phases are frequentlyused together. One phase is
dispersed as drops in theother phase followed by coalescence after
the process isover. This can be problematic if there are
tendenciesfor emulsification. Microporous/porous membranes canbe
particularly useful here since the two immisciblephases can be kept
on two sides of the membrane withtheir phase interfaces immobilized
at the membranepore mouths. Solvent extraction is conventionally
usedto isolate and concentrate dilute organic productsobtained from
whole cell-based fermentation processes.If the fermentation suffers
from product inhibition, thenextraction of the product(s) during
fermentation in-creases the fermentor productivity. However,
solventdispersion can lead to a phase-level toxicity58 problemfor
the whole cells. Nondispersive phase contactingusing
microporous/porous membranes can resolve thisproblem.
In nondispersive phase contacting employing mi-croporous/porous
hydrophobic membranes, the organicphase wets the membrane pores;
the aqueous phase ismaintained outside the pores at a pressure
equal to orhigher than that of the organic phase. As long as
thisexcess pressure does not exceed a breakthrough pres-sure, the
aqueous-organic interface remains immobi-lized on the aqueous side
of the membrane with eachphase flowing on a particular side of the
membrane.59,60For hydrophilic microporous/porous membranes,
theaqueous phase is inside the pores; the organic phase iskept
outside the pores at a pressure higher than thatof the aqueous
phase (Figure 11).
This technique has been employed in four types ofreaction
systems: fermentor-extractor; enzymatic fatsplitting; phase
transfer catalysis; extractive membranebioreactor for enzymatic
resolution of isomers. Theadvantages of these techniques are: no
dispersion andtherefore no need for coalescence; no need for
densitydifference between the two phases; known interfacialarea;
modular systems leading to easy scale-up; mass-transfer rates
independent of interfacial tension; noflooding and no loading,
allowing widely different phaseflow rate ratios to be used.
Further, the membrane mayprovide a very large interfacial area per
unit reactorvolume.
Nondispersive phase contacting advantages are alsopresent in
gas-liquid systems already discussed under
Figure 11. Membrane as a phase separator/contactor for
thereaction-extraction processes.
3724 Ind. Eng. Chem. Res., Vol. 38, No. 10, 1999
-
the category of gas-liquid reactions (section 2.3.2) forthe
controlled addition of one or two reactants. Toachieve
nondispersive operation, one has to maintainthe proper pressure in
each phase. For example, forhydrophobic microporous membranes, the
aqueous phaseis usually outside the pores (which are gas-filled)
andmaintained at a pressure23 higher than that of the gas.For
hydrophilic membranes, the aqueous or organicphase is usually
inside the pores and the outside gasphase is at a higher
pressure.
2.4.1. Fermentor-Extractor. In fermentation pro-cesses for
producing ethanol, acetone-butanol-ethanol(ABE), etc., microporous
hydrophobic hollow fiber mem-branes have been introduced into
tubular reactors inwhich whole cells are immobilized on appropriate
sup-ports on the shell side; through the bore of the hydro-phobic
hollow fibers, O2 is supplied in ethanol fermen-tation and N2 is
supplied in ABE fermentation.15,61,62The gases supplied help the
cells grow, maintain theneeded anaerobic condition, and remove
gases such asCO2 and H2 produced by fermentation. When a
sub-stantial concentration of the desired product has beenachieved
in the shell-side broth, an organic solvent,passed through the
fiber lumen at a pressure lower thanthat in the broth, extracts the
products (e.g., ethanol,ABE, etc.) nondispersively (Figure 12).
This reducesproduct inhibition and can lead to considerably
in-creased volumetric fermentor productivity.16 Other in-hibitory
side products, e.g., acetic acid, etc., are simul-taneously
extracted out. Although the fermentation-extraction process has not
been commercialized yet, themembrane-based solvent extraction
technique is alreadybeing commercially employed in at least two
largeinstallations (one in Europe and the other in Japan).
2.4.2. Enzymatic Fat Splitting. In enzymatic split-ting of olive
oil using hydrophobic microporous mem-branes and the enzyme lipase
immobilized at theaqueous-organic interface63 (Figure 13), olive
oil flowson one side of the membrane and wets the pores; anaqueous
buffer solution flows on the other side at ahigher pressure which
immobilizes the aqueous-organicinterface. The enzyme Candida
cylindracea lipase,spontaneously adsorbed at the aqueous-organic
inter-face of the microporous membrane, catalyzes the fol-lowing
hydrolysis reaction:
Glycerine is removed in the aqueous phase; fatty acids
are removed in the oil phase. The membrane allowsaqueous-organic
phase immobilization, enzyme im-mobilization, and localized product
separation into theappropriate phase. Molinari et al.64 have shown
thatthis reactor was better than a conventional emulsion-based
reactor: the specific enzymatic activity washigher, the specific
rate was more constant with time,and the two products were
separated after the reaction.
2.4.3. Phase Transfer Catalysis. Stanley andQuinn65 have studied
the reaction of bromooctane in thesolvent chlorobenzene with
aqueous iodide to form thedisplacement products, iodooctane in the
solvent chlo-robenzene and aqueous bromide. The phase
transfercatalyst (PTC), tetrabutylammonium (TBA) ion, wasintroduced
as the bromide salt in the organic feed. Theaqueous feed was passed
on one side of the microporoushydrophobic flat membrane of
poly(tetrafluoroethylene)(PTFE); the organic phase wetting the
membrane poreswas passed on the opposite side at a pressure lower
thanthat of the aqueous phase. Conventional
emulsification/coalescence problems were avoided in this
PTC-facili-tated reaction. Further, since the membrane
area(therefore the aqueous-organic interfacial area) isknown,
operation of the reactor can be carried out withgreater
flexibility.
2.4.4. Extractive Membrane Bioreactor for En-zymatic Resolution.
In a multiphase/extractive en-zyme membrane reactor66 used for the
industrial pro-duction of diltiazem chiral intermediate, an
asymmetricwater-filled hydrophilic hollow fiber membrane of
acopolymer of acrylonitrile having a 30000 MWCO (mo-lecular weight
cutoff) skin on the fiber internal diameteris employed. The enzyme
is immobilized in the poroussubstructure from the shell side via
ultrafiltration(function 2.6). The organic phase in which the
enzymehas limited solubility flows on the shell side at a
higherpressure, immobilizing the aqueous-organic interfaceat the
pore mouth, thus containing the enzyme in the
Figure 12. Bifunctional nature of hydrophobic membranes. Shown
above is the end view of the hollow fiber membranes with
biocatalystparticles surrounding them. (a) The membrane can be used
to supply gas throughout the reactor volume while at the same time
removingcarbon dioxide and hydrogen. (b) When a solvent is passed
through the membrane lumen under correct pressure conditions,
solventextraction can provide integrated product recovery while
removing carbon dioxide (reprinted with permission from Frank and
Sirkar.61Copyright 1986, John Wiley & Sons).
triglycerides + 3H2O {\}enzyme
glycerine + 3 fatty acids (12)
Figure 13. Enzymatic fat splitting in a hollow fiber
bioreactor(reprinted with permission from Hoq et al.63 Copyright
1985,American Oil Chemists Society).
Ind. Eng. Chem. Res., Vol. 38, No. 10, 1999 3725
-
water-filled substructure bounded by the membraneskin
impermeable to the enzymes. If the enzyme activityis reduced
substantially, the enzyme can be flushed outeasily and fresh enzyme
loaded in the absence of theorganic phase. The organic flow is then
restarted.Stereoselective enzymatic hydrolysis of the
undesiredisomer from a racemic mixture of glycidic esters and
itsremoval in an aqueous buffer is carried out continu-ously. Here,
the membrane provides reversible im-mobilization of the hydrolytic
enzymes in the pores ofthe hydrophilic asymmetric hollow fiber,
immobilizationof the aqueous-organic interface on the fiber
outsidediameter, reaction product extraction, and a very
largeaqueous-organic interfacial area.
2.5. Segregation of the Catalyst (and Cofactor)in a Reactor. A
membrane incorporated in a catalyticreaction system can perform,
among others, a numberof functions related to the catalyst. If the
catalyst ismobile in the reaction fluid, the membrane can
preventits escape from the system. If the catalyst is to
beimmobilized with easy access to the reactants andconvenient exit
for the products, a porous/microporousmembrane structure may have
the catalyst immobilizedon/within its structure (function 2.6).
Alternately, themembrane material itself may act as the
catalyst(function 2.7). We focus here on cases where the catalystis
mobile in the reaction fluid. Examples of suchcatalysts are enzymes
(and cofactors where applicable),whole cells, and homogeneous
catalysts (in organicsynthesis). The segregation of particulate
heterogeneouscatalysts by filters is not under consideration.
The production of organic compounds by synthesis inorganic
solvents or an aqueous-organic biphasic reac-tion medium is very
common. Many use homogeneouscatalysts whose molecular weights are
considerable(e.g., in the range of, say, 300-800). Very few
mem-branes are capable of retaining such species whileallowing the
organic solvent to pass at appreciable rates.Nanofiltration
membranes, just becoming available,have the necessary solvent
resistance and rejectionbehavior in a few cases. Whu et al.12 have
identifiedsome of these capabilities including retaining the
ho-mogeneous catalysts while passing the organic solvents.More
extensive use of such nanofiltration membraneswill allow their use
in organic synthesis for, amongothers, retaining the homogeneous
catalysts.
The use of membranes to segregate enzymes used ascatalysts for
biosynthesis or biocatalysis is practiced ina wide scale.
Primarily, ultrafiltration membranes hav-ing molecular weight
cutoffs in the range of 5000-100000 are used to retain the enzymes
(molecularweight range 10000-100000 Da) in the CSTR reactoras
smaller products are removed with water throughthe membrane (Figure
2). Originally suggested byMichaels67 and implemented as early as
1970,19 thistechnique is used commercially for the production
ofamino acids. Jandel et al.68 have illustrated
continuousproduction of L-alanine from fumaric acid in a
two-stagemembrane reactor using the enzyme aspartase in thefirst
stage,
and L-aspartate--decarboxylase in the second stage,
Some enzymatic synthesis reactions are carried outin hollow
fiber ultrafiltration membrane devices in whatis called the
perfusion reactor mode. The enzymes orthe whole cells are packed on
the shell side with theshell side ports being closed off;
substrate-containingfeed is pumped through the tube side. The
substratediffuses through the membrane pores to the shell sideand
reacts with the enzymes/whole cells, and thesmaller molecular
weight products diffuse back to thetube side and are carried away.
The enzymes or wholecells may also be kept in the tube side with
both endsclosed off: the substrate-containing solution will
thenflow into and out of the shell side. This latter mode isnot
commonly used. A review of membrane bioreactorswherein the
membranes segregate enzymes/whole cellsis available in Cheryan and
Mehaia.6
Some enzyme-based reactions, however, require
low-molecular-weight coenzymes or cofactors in addition tothe main
enzyme to carry out the overall enzymaticreaction. Typical examples
of such coenzymes are nico-tinamide adenine dinucleotide (NAD+),
the reducedform of NAD+, viz., NADH, NADPH (the reduced formof NAD
phosphate), etc. Figure 10 shows one suchreaction where the
D-lactate dehydrogenase enzymeneeds NADH for the conversion of
pyruvate to D-lactate.In the process, the oxidized form of NADH,
viz., NAD+,is produced. Unless this NADH is regenerated fromNAD+,
one has to continuously supply NADH fromexternal sources. NADH is
costly ($1000/mol). A numberof strategies have been explored to
solve this problem(see Cheryan and Mehaia6).
The strategy shown in Figure 10 is an important one,viz., an
additional enzymatic reaction employing aregenerating enzyme, in
this case an alcohol dehydro-genase, to regenerate the NADH from
NAD+. However,this requires a cosubstrate (e.g., ethanol) and
yields acoproduct (e.g., acetaldehyde). To design a
continuousprocess to retain the enzymes as well as the cofactorsin
the system by a membrane as the substrates comeinto the reaction
chamber and the product and coprod-uct leave the reaction chamber,
one needs a very specificmembrane. Figure 10 provides a special
example forvolatile cosubstrate and coproduct. We consider here
aseparate situation, but one more commonly encountered;viz.,
enzymes have molecular weights > 40000 Da andthe substrates are
5-12 carbon sugars; note that thecofactor molecular weights are
around 700 Da. Thus, ifwe use too tight of a membrane, the
substrate introduc-tion into the reaction zone will encounter
considerableresistance, although NAD+ and NADH will be retainedby
the membrane.
To solve the problem, Nidetzky et al.69,70 have selecteda
charged nanofiltration (NF) membrane having -vecharges on the
surface (as shown in Figure 14) thatpreferably retains the
cofactors almost completely with-out binding the enzymes or
cofactors. However, the NFmembrane had a size exclusion (1 kD)
slightly higherthan those molecules having the molecular mass of
thecofactor; this ensured higher substrate fluxes. Thecharge of the
NF membrane (-ve) in this case allows forcofactor retention since
both NAD and NADH carrynegative net charge at pH values greater
than 3. Thisfumaric acid + NH3 {\}
aspartaseL-aspartic acid (13a)
L-aspartic acid98L-aspartate--decarboxylase
L-alanine + CO2 (13b)
3726 Ind. Eng. Chem. Res., Vol. 38, No. 10, 1999
-
process has been demonstrated for the conversion offructose to
mannitol by mannitol dehydrogenase69 (orxylose to xylitol by xylose
reductase70) as the cosubstrateglucose was converted by the
regenerating enzyme,glucose dehydrogenase, to glucono--lactone
(ultimatelyconverted to gluconic acid in a separate device).
Pilotscale results show a TTN (total turnover number) ofcoenzyme to
be in the range of 75000-100000.
This approach to cofactor retention or segregation bythe
membrane appears to be considerably simpler thanthe earlier
strategy of binding the cofactors to a mac-romolecule, e.g.,
poly(ethylene glycol), dextran, etc.,which allowed a regular UF
membrane to retain thecofactor within the reactor.71 (A similar
strategy72 hasbeen pursued for the enantioselective addition of
diethylzinc to benzaldehyde using a homogeneously solublecatalyst
retained by a solvent-stable polyaramide UFmembrane within the
reaction vessel; the R,R-diphenyl-1-prolinol used as the chiral
ligand was coupled to acopolymer made from 2-hydroxyethyl
methacrylate andoctadecyl methacrylate (96000 MW) and the
solventused was hexane.) However, the charged NF membrane-based
retention of NAD+ is at this time somewhat lower,around 0.65-0.85.
Optimization of the NF membranevis-a`-vis the cofactor retention is
still needed unless theproduct happens to have a very high value so
that alower TTN (1000-5000) may be tolerated.
2.6. Immobilization of a Catalyst in (or on) aMembrane. Four
basic types of catalysts are relevant:(a) enzymes and (b) whole
cells for biocatalysis; (c)oxides and (d) metals for nonbiological
synthesis. Bio-catalysts will be considered first since their
immobiliza-tion in (or on) the membrane was explored much
earlier.Five techniques have been studied in varying degrees.They
are (1) enzyme contained in the spongy fibermatrix; (2) enzyme
immobilized on the membranesurface by gel polarization; (3) enzyme
adsorbed on themembrane surface; (4) enzyme immobilized in
themembrane pore by covalent bonding; (5) enzyme im-mobilized in
the membrane during membrane formationby the phase inversion
process of membrane making.
Of these, technique (1) is used commercially. Anenzyme solution
is ultrafiltered from the spongy side(porous substructure) of an
asymmetric ultrafiltrationmembrane to the skin side; this
introduces the enzymesin the spongy matrix. Feed solution is also
supplied from
the same side; product solution goes out through themembrane
skin (see Jones et al.73 for a brief introduc-tion to the
literature of hollow fiber enzymatic reactorsemploying this method
of immobilization). An exampleof this type of immobilization in the
context of aqueous-organic enzymatic processing has been
illustrated insection 2.4.4. Technique (2) whereby the enzymes
areimmobilized on the pressurized ultrafiltration mem-brane surface
where an adherent gel layer of concen-trated enzymes are formed was
suggested by Drioli andScardi.74
Lipase enzymes (e.g., C. cylindracea) are spontane-ously
adsorbed (technique (3)) on a hydrophobic mi-croporous
polypropylene membrane surface and areutilized in aqueous-organic
enzymatic processing (sec-tion 2.4.2). Enzymes can be covalently
bound to hydro-philic membranes by the cyanogen bromide
procedure.This technique has been known for a long time and hasbeen
used, for example, for binding chymotrypsin to aMillipore filter
membrane by Matson and Quinn75 intheir membrane reactor studies.
Site-directed mutagen-esis has been introduced into enzymes so that
the activesite of the enzyme is away from the surfaces of
mem-branes such as hydrophobic poly(ether)sulfone; im-mobilized
mutant enzymes on such membranes havemuch higher activity than
randomly immobilized en-zymes for catalytic conversions.76 Finally,
it is possibleto incorporate particular enzymes into the
organiccasting solution for polymeric membranes and thenprepare a
membrane by the phase inversion processwherein the final membrane
has the enzyme dispersedthroughout the membrane structure.
Chopped microporous hydrophobic or hydrophilic hol-low fibers
have been used to grow whole cells in the fiberlumen and the fiber
outside surface.77 Such choppedhollow fibers with immobilized cells
were later utilizedin a tubular fermentor to carry out yeast-based
ethanolfermentation for an extended period. The chopped fibersof
appropriate lengths provide adequate nutrients aswell as O2 to the
immobilized cell mass; such a biore-actor could also have
continuous lengths of hydrophobicmicroporous hollow fibers for gas
supply and removalas well as for in situ product extraction by
dispersion-free solvent extraction.16
Nonbiological synthesis of most products involvechemical and
thermal conditions too harsh for almostall of the current polymeric
membranes available;therefore, the membranes investigated are
primarilyinorganic in nature, either ceramic or metallic.
Thecatalysts are predominantly oxides and/or metals. Ofthe numerous
oxides and metals used as catalysts,membrane reactor studies, where
the membrane hadimmobilized catalysts on/in it, have primarily used
thoseemployed for dehydrogenation reactions, viz.,
platinum,palladium, etc., on metal oxides such as alumina
andsilica.1 Examples are a Pt-impregnated Vycor glasstube28 for
cyclohexane dehydrogenation, Ag-modified 40 -Al2O3 membrane
obtained via wet impregnation ofAgNO3, and then calcination under a
reduced atmo-sphere for dehydrogenation of methanol.50 For the
Clausreaction,
350-nm R-Al2O3 membrane pores were impregnatedwith aluminum
nitrate and urea solution in water,dried, and calcined.78 For
multiphase hydrogenation
Figure 14. The substrates and products flow through
thenanofiltration (NF) membrane, whereas the enzyme and thecoenzyme
stay within the enzyme reactor. The membrane used isasymmetric with
a supporting matrix of polyetherketone andpolyethersulfone with a
thin-coated layer of sulfonated polysulfone.The fixed charge
density is 1.5 mequiv g-1. NAD+, nicotinamideadenine dinucleotide;
NADH, reduced NAD+; UF, ultrafiltration(reprinted with permission
from Nidetzky et al.69 Copyright 1996,American Chemical
Society).
2H2S + SO2 T3/8S8 + 2H2O (14)
Ind. Eng. Chem. Res., Vol. 38, No. 10, 1999 3727
-
studies, palladium was deposited on microporous -Al2O3deposited
on 6 m R-Al2O3 by soaking the tube in anaqueous solution of
ammonium tetrachloropalladium,drying it, and calcining it.51 For
nitrobenzene hydroge-nation, the platinum catalyst was deposited in
a 50 -Al2O3 membrane tube by ion exchange with H2PtCl6.52Although
the effect of catalyst distribution in a mem-brane-enclosed
catalytic reactor has been studied,79 nostudies have been made with
a variation of catalystdistribution in the catalytic membrane
itself.
Polymeric membrane-based studies involving thedistribution of
nanosized clusters or micron-sized par-ticles in porous or
nonporous polymeric membranes hasbeen carried out at GKSS
(Geesthacht, Germany) byFritsch;80 10% Pd on charcoal powder
particles (40 m)were used as catalysts and dispersed in the
castingsolution for the preparation of porous polyetherimide(PEI)
membranes. Transfer hydrogenation of N-CBZ-L-phenylalanine to
L-phenylalanine was successfullycarried out at low temperatures.
Hydrogenation ofpropene to propane was carried out by using a
nonpo-rous membrane containing nanoclusters of Pd in a
densepoly(amide-imide) membrane. The mechanism of sucha membrane
function has been described in section 2.8.
Bellobono et al.81 have described the performance
ofmicrofiltration membranes formed with 30% TiO2 aswell as some 6%
of particular photocatalysts. Themonomer, prepolymer blend with the
semiconductorTiO2 and the photcatalysts were photografted onto
aperforated polyester support; the pore size of photosyn-thesized
membranes were 2.5-4 m. The membraneswere placed coaxially with a
UV lamp in a stainless steelcasing having a mirror-like surface for
reflection. Thesolutions processed permeated through the
membrane;before entering the reactor, the wastewater solutionswere
saturated with O2 or O3.
2.7. Membrane Is the Catalyst. Most catalyticmembrane reactors
for higher temperature operationsemploy ceramic membranes in the
pores/micropores ofwhich catalysts were deposited. The base
membranes,e.g., silica and alumina, are generally not catalysts
forthe reactions studied. There are, however, a number ofmembranes
which are inherently catalytic for particularreactions; no catalyst
needs to be deposited on or in themembrane. Particular examples are
cation-exchangemembranes, Nafion membranes, palladium membranes,and
zeolite membranes.
Consider a cation-exchange membrane and an esteri-fication
reaction (15) in which a carboxylic acid, R1-COOH (e.g., oleic
acid), reacts with an alcohol, R2OH(e.g., methanol), under the
influence of an acidic catalystproviding a proton:
In the schematic shown in Figure 15, CH3OH and H2-SO4 are on one
side of the membrane and the reactants,viz., CH3OH and oleic acid,
are on the other side of themembrane.82 In this case protons are
the counterionsin the cation-exchange membrane introduced from
theleft-hand side of the membrane; naturally, a layer ofprotons
appears on the other membrane surface exposedto the reaction
mixture of the alcohol and the acid. Theseprotons catalyze the
esterification reaction. No separatecatalyst, e.g., H2SO4,
p-toluenesulfonic acid, etc., isrequired, eliminating the need for
catalyst separationafter reaction if homogeneous catalysts are
employed
(anions, e.g., HSO-4, do not get transported through
thecation-exchange membrane). The cation-exchange mem-brane also
separates a reaction product, viz., water, fromthe reaction
mixture. Water is transported from thereaction mixture to the
catalyzing mixture, thus achiev-ing equilibrium shift to the right
in reaction 15 (section2.1). If solid ion-exchange beads were used
as thecatalyst, such a function would not have been possiblein a
continuous process.
The acid form of a Nafion membrane is also capableof carrying
out both functions.83 The acid form of themembrane prepared by
boiling tubular Nafion mem-branes in concentrated HNO3 and then in
water wasfound to catalyze the esterification of methanol
orn-butanol to methyl acetate or butyl acetate, respec-tively, with
acetic acid. These membranes were alsofound to be effective in
separating water from n-butanol(water/alcohol selectivity 8.0;
water/acetic acid selec-tivity 9.0). The Cs+ form of the Nafion
membrane wasfound to have a much higher selectivity for both
water/alcohol (71) and water/acetic acid (149). Simulta-neous
catalysis and separation by a catalytically activemembrane can
potentially increase the membrane fluxfor the products produced and
removed since the reac-tions occur within the membrane compared to
a passivemembrane over which the solution is passed after
thereaction is completed.
Palladium is known to be a catalyst for hydrogenationand
dehydrogenation reactions. A palladium membraneis also infinitely
selective for H2. Thus, for dehydroge-nation reactions, a palladium
membrane simultaneouslyacts as a catalyst and allows the product H2
to beremoved through the membrane and obtained in thepure form (see
section 2.1 for references). Zeolites arewell-known as catalysts.
Thin zeolite membranes arebeing developed for the selective
transmission of speciespreferentially adsorbed or smaller than the
pore size.A number of reactions have been studied.2
2.8. Membrane Is the Reactor. In a membranereactor, catalysts
are used frequently. The membranemay physically segregate the
catalyst in the reactor(function 2.5) or have the catalyst
immobilized in theporous/microporous structure or on the
membranesurface (function 2.6). The membrane having the cata-lyst
immobilized in/on it functions almost in the sameway as a catalyst
particle in a reactor does exceptseparation of the product(s)
(function 2.1) takes place,in addition, through the membrane to the
permeateside. All such configurations involve the bulk flow of
thereaction mixture along the reactor length while diffusionof the
reactants/products takes place generally in aperpendicular
direction to/from the porous/microporouscatalyst.
R1COOH + R2OH {\}Cat-H+
R1COOR2 + H2O (15)
Figure 15. Concentration profiles around a
cation-exchangemembrane. Example at the beginning of run no. 11.
Catalyzingsolution, 0.9 g of H2SO4 in 180 mL of methanol; reacting
mixture,25 g of oleic acid and 75 g of methanol (reprinted from
Chemsed-dine and Audinos.82 Copyright 1996, with permission from
ElsevierScience).
3728 Ind. Eng. Chem. Res., Vol. 38, No. 10, 1999
-
When the bulk flow of a reaction mixture takes placethrough the
membrane from one membrane surface tothe other, the membrane is the
reactor. Generally, themembrane in such a case will be
porous/microporous toreduce the pressure drop for practical flow
rates. Thelength of the pores/transport corridors is the
reactorlength; the reactions may take place on the surfaces ofsuch
macropores or there may be radial diffusion ofreactants into the
micropores and products out of themicropores into the main porous
corridors where con-vective motion occurs. The convective motion of
thereaction mixture through the membrane created by anapplied
pressure drop across the membrane thicknessmay involve Knudsen
diffusion, Poiseuille flow, or atransitional regime for gaseous
reaction mixtures.
Such a reactor can have an exceptionally high valueof reactor
L/d for even thin membranes since d valuescan be very small (from
0.2 m to 4 nm). As a result,extremely high conversions are
possible, as shown byrecent theoretical studies.84 Such a reactor
can beidentified as a pore flow through reactor (PFTR) sinceeach
macropore that traverses the membrane thicknessis a reactor tube.
Further, the mass-transfer resistanceencountered by the reactants
to reach the catalytic sitesare significantly reduced because of
the bulk flowthrough the membrane pores.85 Pina et al.85
haveexperimentally demonstrated extremely efficient re-moval of
toluene and methyl ethyl ketone from air bylow-temperature
oxidation (100-320 C) using a Pt/-Al2O3 catalytic membrane
operating under the Knudsendiffusion regime, supporting the earlier
work by Pinaet al.86
The membrane in this case does not possess appar-ently the
separation capability characteristic of a mem-brane in a tubular
reactor enclosed by a membrane. Themembrane can still separate, for
example, particles froma gas mixture as the gas mixture enters the
membraneand is convected through the pores. One can have stacksof
disks of such membrane reactors with spaces betweenthe disks being
used for product separation or reactantaddition or heat
exchange.
Fritsch80 deposited nanoclusters of catalysts such asPd, Pt/Ag,
and Pd/Co in the size range of 1-3 nm inpolymeric membranes such as
poly(amide-imide) andpoly(dimethylsiloxane) (PDMS). Reactions were
carriedout in the flow through mode except that the mem-branes were
essentially nonporous. In this case, themembrane treats every gas
in the feed mixture accord-ing to the normal gas permeation
selectivity displayedin the polymeric membrane. For example, PDMS
has aselectivity of 8.4 from pure gas measurements for
C3H6(propene) over H2. Therefore, a feed gas ratio of at least8.4
for H2/C3H6 is required for complete conversion ofC3H6 to C3H8. In
fact, Fritsch80 has illustrated completeconversion of C3H6 to C3H8
using Pd cluster-containingpore-free membranes of PDMS. The
metallic catalyticnanoclusters were prepared during the polymeric
mem-brane formation itself.
Such thin membranes acting as the reactor are quiteuseful for
fast exothermic reactions or for reactionswhere one of the
intermediates should be the mainproduct. They possess very high
catalyst surface areaper unit membrane area. However, the
throughput perunit membrane area would be low. For example,
Fritsch80has observed permeance values in the range of 1-3.2 10-6
cm3/cm2 s cmHg for PDMS membranes. Thereactor, therefore, has to
have a large area and will have
the shape of a thin disk of large diameter. A standardindustrial
reactor will have the shape of a long narrowtube.
In an alternate strategy, Wu et al.87 had used asomewhat similar
membrane as a reactor via interphasecontacting. They employed PDMS
membranes modifiedappropriately and containing titanium silicalite
zeoliteas a catalyst. Oxyfunctionalization of n-hexane to amixture
of hexanol and hexanone was carried out bybringing in n-hexane and
30 wt % of an aqueous H2O2solution: the silicone membrane acted as
the reactionmedium and the reactor.
2.9. Solid-Electrolyte Membrane Supports theElectrode, Conducts
Ions, and Achieves the Reac-tions on the Surface. Solid
electrolytes are solid-statematerials possessing ionic
conductivity. The two ionsof the greatest relevance are H+ and O2-,
although otherions, Cl-, F-, Ag+, etc., have been found to be
conductedas well. Solid polymer electrolytes such as
perfluori-nated ionomer membranes (e.g., Nafion) allow transportof
H+ ions in the presence of water and are often
calledproton-exchange membranes. Solid solutions of oxidesof di- or
trivalent cations (e.g., Y2O3) in oxides oftetravalent metals such
as ZrO2 can conduct O2- overa wide temperature range. Nonporous
disks of such asolid electrolyte can act as membranes for such
ionicspecies and are quite useful for fuel cells and as
O2-conductors.
Consider a solid-polymer-electrolyte fuel cell: porousgraphite
gas-diffusion electrodes hot pressed onto bothsides of a thin
polymer membrane (e.g., Nafion) aboveits glass transition
temperature. This cell is fed withwet air on one side and wet H2 on
the other side (derivedfrom the reforming of CH3OH in an adjacent
reformerand therefore contains CO2 also). The membrane andthe
electrode assembly is schematically shown in Figure16.88 The
gas-diffusion electrode is made from porousgraphite impregnated
with a Pt catalyst. H2 gas diffusesthrough the porous electrode and
is oxidized on Ptcatalyst sites at the anode in a three-phase
region(Figure 16) containing a polymer electrolyte,
gaseousreactants, and a carbon matrix89
Protons transferred through the membrane react withO2 at similar
catalyst sites at the cathode to form water
Figure 16. Membrane and electrode assembly. The five regionsof
the model are shown (not to scale) (adopted from refs 88
and89).
H2 T 2H+ + 2e- (16)
O2 + 4H+ + 4e- T 2H2O (17)
Ind. Eng. Chem. Res., Vol. 38, No. 10, 1999 3729
-
Thus, the membrane supports the electrodes on twosides,
transports H+, and achieves reactions on itssurfaces. The electrons
in the external circuit obtainedfrom the chemical energy of the
oxidation of H2 providethe required current.
A solid ceramic proton conductor tube from a
stron-tium-ceria-ytterbia (SCY) perovskite of the
formSrCe0.95Yb0.05O3 has been employed with two
porouspolycrystalline palladium films deposited on the twosides of
the ceramic tube to carry out NH3 synthesis atatmospheric
pressure.90 The two electrodes were con-nected to an external
galavanostat-potentiostat bywhich the appropriate current was
applied. At theanode, gaseous H2 is converted to H+:
The protons are transported through the solid electro-lyte to
the cathode where the half-cell reaction
takes place to complete the overall reaction
The reaction rate in this case was strictly controlled bythe
rate of H+ supply since H2 was the limiting reactant,N2 being
present in abundance. The membrane func-tions in this case again
are quite similar to those in thefuel cell example given earlier.
At 570 C and atmo-spheric pressure, greater than 78% of the
electrochemi-cally supplied H2 was converted into NH3. This
rateexceeds that in a conventional catalytic reactor
underequivalent conditions of compositions, pressure,
andtemperature by 3 orders of magnitude.
Using a yttria-stabilized zirconia-based (YSZ-based)conducting
solid electrolyte which transports O2-, McK-enna et al.41 have
studied the synthesis of HCN in asolid-electrolyte-cell reactor. At
the cathode, an O2-containing gas (containing N2) passes; adsorbed
O2 getsconverted to O2- via
which is then conducted to the anode surface. At theanode, a
mixture of CH4 and NH3 is supplied, whichleads to the following
reaction,
producing HCN. In such a scheme, N2 in the air on thecathode
side is rejected by the membrane.
For partial oxidation reactions using, for example,CH4, it is
not necessary to apply a voltage. Instead, oneemploys mixed
conducting materials having both ionicand electronic
conductivities. A gradient of O2 partialpressure across the ceramic
membrane induces a gradi-ent of O2- in the same direction and
electrons in theopposite direction. The mixed conducting membrane
hasa catalyst on the O2 side to facilitate the formation ofO2-. The
catalytic properties on the other surface of themembrane allows CH4
to react in the manner of
generating the electrons which diffuse to the O2 side tomaintain
charge neutrality in the ionic lattice structure
of the ceramic membranes. See Figure 17 from EltronResearch for
CH4 conversion to synthesis gas.91 Themembrane in this case
separates O2 from air throughthe membrane, while supporting the
catalyst for re-forming CH4 and distributing O2 in a controlled
fashionthroughout the reactor. The study jointly sponsored byDOE
and Argonne National Laboratories40 used arhodium-based partial
oxidation catalyst inside themixed-conducting ceramic tube with O2
from the airbeing present on the outside of the tube. Such
arrange-ments also reduce the possibilities of an
explosivemixture.
2.10. Transfer of Heat. The most recent studies ofmembrane
reactors have been in the context of thepetrochemical industry.8
They take place at highertemperatures (>200 C) and there likely
is a need forconsiderable heat transfer because the reaction may
beexothermic or endothermic. Dehydrogenation reactionsstudied
frequently are endothermic. The membrane, ifinert, is in a
catalytic reactor, packed bed, or fluidizedbed.92 Thus, the
membrane may have to participate inheat transfer. Itoh and Govind45
have studied an en-dothermic dehydrogenation reaction on one side
of apalladium membrane coupled with an exothermic hy-drogen
oxidation on the other side. Heat was transferredfrom the oxidation
reaction side to the dehydrogenationside through the membrane. They
have concluded thatheat transfer across the membrane leading to
anadiabatic reactor resulted in a higher conversion thanwhat was
possible under isothermal conditions.
In actual practice, there will be one particular reactiongoing
on and heat is going to be supplied from a firedheater, molten salt
baths, or thermal fluid jackets.Therefore, the membrane is most
likely going to bedecoupled from the heat transfer process. A
commonconfiguration of some interest in a packed bed mem-brane
reactor consists of multiple membrane tubesinside tubular catalyst
beds, placed in turn, in anotherenclosure for heat exchange.7,92
Thermal expansionproperties of the membrane tube, sealing at the
header,and protection from abrasional damage from catalystparticles
are of much greater importance.8
2.11. Immobilizing the Reaction Medium. Manyreactions are
carried out in an organic solvent. Theseinclude two-phase
reactions, e.g., those encountered inphase transfer catalysis,
gas-liquid reactions, etc. Aporous/microporous membrane can
immobilize an ap-propriate reaction medium in the pores. The two
dif-ferent phases containing reactants can be brought to
3H2 f 6H+ + 6e- (18)
N2 + 6H+ + 6e- f 2NH3 (19)
N2 + 3H2 T 2NH3 (20)
3/2O2 + 6e- f 3O2- (21)
CH4 + NH3 + 3O2- f HCN + 3H2O + 6e
- (22)
CH4 + O2- f CO + 2H2 + 2e
- (23)
Figure 17. Methane conversion to synthesis gas (Eltron
ResearchInc., with permission91).
3730 Ind. Eng. Chem. Res., Vol. 38, No. 10, 1999
-
the two sides of the membrane. As long as the two feedphases are
immiscible with the reaction medium, reac-tants can partition into
the reaction medium and reactand then the products can partition
back into theflowing phases on opposite sides of the
membrane.Unfortunately, such a configuration, usually termed asthe
supported liquid membrane (SLM), has limitedstability93 because of
a variety of reasons including afinite solubility of the reaction
medium in the twodifferent reactant-containing phases. Therefore,
as dis-cussed in the next section, Guha and Sirkar,94 Chen etal.,95
and Guha et al.11,96 have employed two differenthollow fiber
membranes for bringing two reactant-containing streams into the
membrane reactor while thereactants partitioned into the liquid
reaction mediumcontained (confined) in the shell side between the
twosets of hollow fibers.
Kim and Datta97 have used a porous support disk toimmobilize a
homogeneous catalyst in a high-boilingorganic solvent. However,
they did not completely wetthe pores so that the gas space was left
in the pore. Thereaction was hydroformylation of ethylene to give
pro-pionaldehyde. With an appropriate support membrane,the
capabilities of the reaction medium within themembrane could be
substantially enhanced. The onlyissue is the lifetime of the
reaction medium.
3. Functions of Multiple Membranes in aReactor
Although more than two different types of membranescan be
accommodated in a reactor, this section willconsider primarily the
functions of only two membranes(differe