Catalytic Membrane Reactors involving
Inorganic Membranes

– A short overview –

Anne JULBE and André AYRAL

Institut Européen des Membranes (ENSCM, UM2, CNRS), Université Montpellier 2 (cc047) Place Eugène Bataillon, 34095 Montpellier cedex 5FRANCE

Introduction

Membrane reactors (MRs) as a concept, dates back to 1960s and a large number of papers have been published on this multidisciplinary vibrant subject at the frontier between catalysis, membrane science and chemical engineering [1-14]. In such an integrated process, the membrane is used as an active participant in a chemical transformation for increasing the reaction rate, selectivity and yield. The membrane does not only play the role as a separator but also as part of the reactor. Membrane reactors, combining in the same unit a conversion effect (catalyst) and a separation effect (membrane), already showed various potential benefits (increased reaction rate, selectivity and yield) for a range of reactionsinvolving the membrane as extractor, distributor or fluid-solid contactor.Due to the generally severe conditions of heterogeneous catalysis,most MR applications use inorganic membranes, which can be dense or porous, inert or catalytically active.

The interest of MRs has been largely demonstrated at the laboratory scale, namely for hydrogenation, dehydrogenation, decomposition and oxidation reactions including partial oxidation and oxidative coupling of methane. Though some small industrial installations already exist, the concept has yet to find widespread industrial applications. One of the important factors hindering further commercial development of MRs are the membrane themselves, their support and the surrounding modules (performance, stability, cost, sealing,..) which still need optimization and new developments. After a rapid overview of the working concepts of MRs, several examples of current research and developments in the field of inorganic membrane reactors will be described.

General considerations on inorganic membranes for MRs

Inorganic membranes for MRs can be inert or catalytically active, they can be either dense or porous, made from metals, carbon, glass or ceramics. They can be uniform in composition or composite, with a homogeneous or asymmetric porous structure. Membranes can be supported on porous glass, sintered metal, granular carbon or ceramics such as alumina.

Different membrane shapes can be used such as flat discs, plates, corrugated systems, tubes (dead-end or not), capillaries, hollow fibers, or monolithic multi-channel elements for ceramic membranes, but also foils, spirals or helix for metallic membranes. The shape of the separative element induces a specific surface/volume ratio for the reactor, which needs to be maximized, typically above 500 m2/m3, for industrial applications [14]. Apart from the evident need for low cost, resistant and efficient membranes for the process, highly permeable membranes are requiredfor all applications. This parameter is directly related to the membrane structure which can be dense or porous, and which defines the transport mechanisms.

Dense membranes have been largely and successfully investigated in MRs, for reactions consumingor generating H2 or O2[4]. Indeed these membranes,exclusively permselective to either H2 [15] or O2[16,17], aregenerally used either as efficient H2 extractors or asO2 distributors. H2 permselective dense membranes include Pd and Pd alloys membranes, whichare commercially available (Johnson–Matthey). Thin supported films and new alloyed compositions have been recently developed in order to reduce membrane cost, sensitivity to sulfur speciesand embrittlement upon aging. Dense ceramicmembranes arealso consideredfor H2 separation, e.g. with proton conducting membranes based on zirconate and cerate perovskite systems [18].

O2 permselectivedense membranes include metallic (Ag) or ceramic membranes (e.g. Yttrium-stabilized zirconia (YSZ), BiMeVOx, La2NiO4+ or (La-Sr)(Fe-Co)O3-perovskites and related oxides). Gas transportin dense metallic membranes occurs via a solution/diffusion mechanism. In stabilized zirconia, ionicconduction is involved at high temperature (800-1000°C);an electronic current is needed in such electrochemical reactors(Fig 1a).The mixed ionic and electronic conductivity ofperovskites avoids the need of an external electrodefor these materials (Fig. 1b). The gradient of partial pressure PO2 on both sides of the membranes is the driving force for ion transport. Composite materials involving a mixture of an ion conducting oxide with an electronic conducting phase (e.g. metal) is also considered (Fig. 1c). One of the common drawbacksto these attractive dense ceramic membranes is theirlimited permeability and thermo-chemicalstability upon aging in operating conditions. Dense membranesperformance has been improved namely by decreasingmembrane thickness (supported membranes), byincreasing the surface roughness and by developing new materials [16].

Although being less permselective than dense membranes, porous membranes offer a higher permeabilityand have been extensively used in catalytic reactors.The gas transport mechanisms in porous membranescan be related to the ratio between the pore sizes andthe mean free path length of gas molecules [19]. The typical gas transport mechanisms in porous membranes are: molecular diffusion and viscous flow (macropores and mesopores), capillary condensation (mesopores), Knudsen diffusion (mesopores), surface diffusion (mesopores and micropores) and micropore activated diffusion. The contribution of the different mechanisms is dependent on the properties of the membranes and the gases as well on the operating conditions of temperature and pressure. At high temperature, when adsorption is no more effective, capillary condensation and surface diffusion are nomore involved. In these conditions, the permselectivity increases when pore sizes decrease and pressure can be used to control the transport through membranes with macropores or big mesopores. As far as the permselectivity is not always a key factor in membrane reactors, membrane research for MRs focus on both microporous and mesoporous materials, with a large range of porous texture and compositions adaptable to a large number of applications.

The main membrane functions in MRs

The concept of combining membranes and reactors is being explored in various configurations, which can be classified in three groups, related to the role ofthe membrane in the process [1]. As shown in figure 2, the membrane can actas:

(a) an extractor, where the removal of the product(s) increases the reactionconversion by shifting the reaction equilibrium;

(b) a distributor, where thecontrolled addition of reactant(s) limits side reactions; and

(c) an active contactor,where the controlled diffusion of reactants to the catalyst can lead to an engineeredcatalytic reaction zone.

In the first two cases, the membrane is usually catalyticallyinert and is coupled with a conventional fixed bed of catalyst placed on one ofthe membrane sides.

Figure 2.The three main membrane functions in membrane reactors [5].

The extractor mode corresponds to the earlier applications of MRs and has been applied to increasethe conversion of a number of equilibrium limitedreactions, such as alkane dehydrogenation, by selectivelyextracting the hydrogen produced. Other H2producing reactions such as the water gas shift, the steam reforming of methane and the decomposition of H2S and HI, have been also successfully investigated with the MR extractor mode. The H2 permselectivity of the membrane and its permeability are two important factors controlling the efficiency of the process. Although most extractor applications feature H2 removal,several decomposition reactions in which O2 is removed have been also considered [6].

An example of a compact and economical on-site H2 production unit based on steam methane reforming coupled with membrane technology is shown in figure 3. This type of membrane reformer has been developed in 2002 in Japan by the New Energy and Industrial Technology Development Organization (NEDO) & Japan Gas association. This system typically works at 500-550°C (instead of 800°C in classical methane reformers)and is able to produce 20-40 Nm3/h [20].The Pd-based membrane selectively extracts the H2, then shifting the equilibrium to the production side (CH4 + H2O CO + 3 H2).The residual CO separatedfrom H2 is burned to CO2and the generated heat is utilized for the reforming reaction. In newly developed systems, the CO is used in the water gas shift reaction (CO + H2O  CO2 + H2) in orderto increase the H2 yield.

In H2 producing reaction, substantial conversion improvements can be obtained with H2 permselective porous membrane extractors such as Pd-based membranes, or almost dense silica membranes prepared by chemical vapor deposition/ infiltration or sol-gel process [21]. Carbon molecular sieves [22] are also considered as H2 extractors, except cannot in oxidative atmospheres. We have to note that silica membranes have a limited stability upon aging above 400°C in the presence of steam, although this problem has been largely improved recently with composite sol-gel derived membranes such as Ni-SiO2 or C-SiO2[21]. Dense proton conducting membranes (as ceramic-metal composite materials) are under also under study as infinitively selective H2 extractors [4].

Figure 3. Membrane reformer developed in 2002 by the NEDO & Japan GasAssociation (including researchers from Tokyo Gas Co., Ltd and Osaka Gas Co., Ltd).

The distributor mode is typically adapted to consecutive parallel reaction systems such as partial oxidation or oxy-dehydrogenation of hydrocarbons or oxidative coupling of methane. For these applications the membrane, separating the alkane from O2, is generally used to control the supply of O2 in a fixed bed of catalyst in order to by-pass the flammability area,to optimize the O2 profile concentration along the reactor, and to maximize the selectivity in the desired oxygenate product [2,5]. This concept has also a beneficial role in mitigating the temperature rise in exothermic reactions. In such reactors, the O2 permselectivityof the membrane is an important economic factor because air can be used instead of pure O2. However, the limited permeability of dense O2 permselective ceramic membranes below 800°C and problems of long-term stability has limited their commercial development. In spite of their poor permselectivity, meso and macroporous membranes remain attractiveoxygen distributors for oxidative reactions below 700°C [2,5].

An example of membrane distributor concept used for converting methane to syngas by partial oxidation (CH4 + 1/2O2 CO + 2H2) is shown in figure 4 [23]. The dense ion conducting membrane is able to selectively transport oxygen above 700°C. They are exclusively permselective to oxygen ions and deliver highly reactive oxygen species (O*) on the reaction side. This type of reformer, resulting in significant cost savings, is currently considered worldwide in a number of industrially driven consortia and namely by Air Products and Chemicals, Inc.

Figure 4.Syngas production with aion transport membrane combining air separation and methane partial oxidation into a single unit operation [23].

In the active contactor mode, the membrane acts as a diffusion barrier and does not need to be permselective but catalytically active. The concept can be used with a forced flow-mode or with an opposing reactant mode. The forced flow contactor mode, largely investigated for enzyme-catalyzed reactions, has been also applied to the total oxidation of volatile organic compounds [2]. The opposing reactant contactor modeapplies to both equilibrium and irreversible reactions [2,5], if the reaction is sufficiently fast compared to transport resistance (diffusion rate of reactants in themembrane). In such case a small reaction zone formsin the membrane (if sufficiently thick and symmetric)in which reactants are in stoechiometric ratio. Triphasic (gas/liquid/solid) reactions, which are limited by the diffusion of thevolatile reactant (e.g. olefin hydrogenation), can alsobe improved by using this concept. Indeed the volatilereactant does not have to diffuse through a liquid film,as far as a gas/liquid interface is created inside thepores, in direct contact with the catalyst [3,5].

An example of a catalytic contactor process developed for the oxidation of waste water is shown in figure 5. This process has been recently developed and patented by SINTEFwithin the frame of a European project called Watercatox [24]. An overpressure of air or oxygen is applied to the outside wall of the tube. The overpressure (5-15 bar) forces the gas-liquid interface to a position close to the catalytic layer.

Figure 5. Principle of the WATERCATOX process developed for waste water treatment [24]

Other strategiesare considered in our group for the continuous catalytic degradation of VOCs in water. For example, a mesoporous photocatalytic TiO2-basedmembranecan be used to reject the macromolecules and colloids contained in the aqueous effluent, although small organic molecules go across the membrane. They are degraded to CO2 and H2O thank to UV-visible illumination of the membrane back-side (Fig. 6)[25].

Figure 6. The photocatalytic membrane reactor concept developed for the treatment of aqueous effluents (from [25])

The different types of membranes and membrane/catalyst
arrangements used in MRs

The different types of MR configurations can be also classified according to the relative placement ofthe two most important elements of this technology: the membrane and the catalyst. Three main configurations can be considered (Fig. 7): the catalyst is physically separated from the membrane; the catalyst is dispersed in the membrane; or the membrane is inherently catalytic. The first configuration is often called ‘inert membrane reactor’ (IMR) by opposition to the two other ones which are ‘catalytic membrane reactors’ (CMRs) [5].

Figure 7.The main membrane/catalyst combinations: (a) bed of catalyst on an inert membrane; (b) catalyst dispersed in an inert membrane; and (c) inherently catalytic membrane.

Catalyst physically separated from an inert membrane

In most cases an inert membrane compartmentalizes the reactor but is not directly involved in the catalyticreaction. The catalyst pellets are usually packed or fluidizedon the membrane (Fig. 7a) which acts as anextractor and/or as a distributor (fractionation of productsor controlled addition of a reactant). This type ofconfiguration is probably one of the most promisedto development as far as only the separative functionof the membrane has to be controlled. Inert membranereactors have been largely studied in the literaturewith both dense permselective membranes andporous membranes.

Other important field of application for inert porous membranes concerned their use as oxygen distributorsin partial oxidation or oxy-dehydrogenation of alkanes, or in the oxidative coupling of methane.The membranes used for this type of application need to play therole as a barrier achieving the desired transmembraneflux while avoiding the back-diffusion of the secondreactant in the oxygen rich compartment. When using meso or macro-porous membranes, this latterfunction can be achieved by an imposed pressuregradient [5].

Catalyst dispersed in an inert porous membrane

When the catalyst is immobilized within the pores of an inert membrane (Fig. 7b), the catalytic and separationfunctions are engineered in a very compact fashion. In classical reactors, the reaction conversionis often limited by the diffusion of reactants into thepores of the catalyst or catalyst carrier pellets. If thecatalyst is inside the pores of the membrane the combinationof the open pore path and transmembranepressure provides easier access of the reactants to thecatalyst (Fig. 8). Two contactor configurations: forcedflow-mode or opposing reactant mode, can be usedwith these catalytic membranes, which does not necessarilyneed to be permselective. It is estimated thata membrane catalyst could be 10 times more activethan in the form of pellets, provided that themembrane thickness and porous texture, as well as the quantity and location of the catalyst in the membrane,are adapted to the reaction kinetics.

Figure 8. Comparison of the contact reactant/catalyst situation in: a) a classical bed reactor configuration and b) an active membrane contactor configuration (from [5]).

In biphasic applications (gas/catalyst) the poroustexture of the membrane must favor gas-wall (catalyst) interactions to ensure a maximum contact of thereactant with the catalyst surface. In the case of catalytic consecutive-parallel reaction systems, such asthe selective oxidation of hydrocarbons, the gas-gas molecular interactions must be limited because theyare non-selective and lead to a total oxidation of reactants and products. Because of these reasons, smallpores mesoporous or microporous membranes, in which the dominant gas transport is Knudsen or microporeactivated diffusion, are typically favored for contactor applications in biphasic reactions. Greaterpore sizes (10-25 nm) are preferred for triphasiccontactor applications (e.g. hydrogenation of liquid alkenes or VOCs oxidation in water as shown in Fig. 5) with an opposing reactant mode.The gas phase combustion of VOCshas been successfully investigated with a Pt--Al2O3membrane with a flow-through contactor mode [2].

Inherently catalytic membranes

The last types of membranes used in MR applications are called inherently catalytic membranes(Fig. 7c). In this highly challenging case, the membranematerial serves as both separator and catalyst,and controls the two most important functions of the reactor. As in the previous case such porous catalyticmembranes are used as active contactors to improvethe access of the reactants to the catalyst. A numberof meso- and micro-porous inorganic membrane materialshave been investigated for their intrinsic catalyticproperties such as alumina, titania, zeolites with acid sites, rheniumoxide, LaOCl, RuO2-TiO2 and RuO2-SiO2, VMgO, or La-based perovskites [5].nThe mesoporous TiO2 photocatalytic membranes currently studied in our group for the continuous degradation of VOCs in water is a typical example of an inherently catalytic membrane. The morphology of the nanophase photocatalytic TiO2 mesoporous membranes prepared by the sol-gel process[25] is shown in figure 9.

In most contactors, the catalytic membrane does not need to be permselective but needs to be highly activefor the considered reaction, to contain a sufficient quantity of active sites, to have a sufficiently lowoverall permeability and to operate in the diffusioncontrolled regime. In most cases, new synthesismethods have to be developed for preparing these catalytically active membranes, namely when the optimumcatalyst composition is complex. The catalytic membrane composition, activity and porous texturehave to be optimized for each considered reactionand keep stable upon use. This challenge explains thelimited number of examples given in the literature for the development of inherently catalytic membranes.