DETAILED DESCRIPTION OF THE INVENTION

[0076] Hydrogen transport membranes function for transport of hydrogen from a hydrogen source to a hydrogen sink and allow hydrogen to be separated from other gases. A membrane has a side facing the hydrogen source unto which hydrogen molecules adsorb and are dissociated, and a side facing the hydrogen sink from which hydrogen molecules are desorbed. A hydrogen-permeable metallic layer(s) formed between the surfaces function for hydrogen transport. The membranes of this invention are designed to maximize the flux of hydrogen, while resisting poisoning and degradation by the components of the hydrogen source gas and preferably to minimize mechanical stress which will result in longer useful life.

[0077] Membranes of this invention comprise some material that functions for the dissociation of molecular hydrogen. This function may be provided by certain metals employed in the membrane which exhibit catalytic properties for the dissociation of hydrogen, such as palladium and its alloys. Metal alloys which contain Co, Fe, Rh, Ru, Pt, Mo, W or Ni can also function as catalysts for the dissociation of hydrogen. Additionally, certain ceramic materials can function as catalysts for the dissociation of hydrogen. Alternatively or in combination, one or more catalyst may be provided in the or on the membrane, e.g., at the hydrogen source surface, which dissociate molecular hydrogen and resist poisoning. The membranes herein also optionally provide materials at the hydrogen sink surface which minimize hydrogen desorption energy.

[0078] Thin metallic layers are formed, for example, in the membranes by deposition of metals or mixtures of metals in the pores of a ceramic. Alternatively, thin foils of metals are used, a ceramic adhesive or paste is applied to a foil to form a porous ceramic, or cermets are fabricated by sintering together powders of metal and ceramic. The metallic layers formed are preferably sufficiently thin so that bulk diffusion of hydrogen is not rate limiting.

[0079] In some of the specific embodiments, the hydrogen transport membranes of this invention are cermet composite membranes in which a mixture of metal or metal alloy particles and ceramic particles are sintered together to form layers including layers that are hydrogen-permeable preferably 50 microns or thinner. In preferred embodiments the porous ceramic and the metal or mixture of metals are selected to maximize matching of the crystal lattice constants of the ceramic and metal materials. Matching of the crystal lattice constants of the ceramic with the deposited metal or metals or with the metal or metals mixed into the cermet provides for a good epitaxial/endotaxial fit between the two materials to minimize mechanical strains and to improved mechanical strength of the composite membrane.

[0080] A concentration gradient of hydrogen provides the ultimate thermodynamic driving force for the transport of hydrogen across membranes. A concentration gradient must be maintained across the membrane at all times during operation. The membrane transports hydrogen from the hydrogen source to a hydrogen sink where a low concentration of hydrogen is maintained. Hydrogen concentration is kept low by physical removal of hydrogen, for example, by application of a vacuum, by use of a sweep gas, or by chemical reaction of the hydrogen after it is transported. To maximize the driving force, the concentration difference between the source and the sink should be made as large as possible.

[0081] Hydrogen transport mediated through the membrane is believed to function by the following steps:

[0082] 1. Diffusion of hydrogen molecules from the hydrogen source to the membrane surface;

[0083] 2. Adsorption of hydrogen molecules on the membrane surface facing the source;

[0084] 3. Dissociation of hydrogen molecules to hydrogen atoms on the membrane surface to form hydrogen atoms, followed by loss of electrons and formation of H+ ions;

[0085] 4. Transport of H+ ions and electrons through the membrane;

[0086] 5. Recombination of H+ and electrons and formation of hydrogen molecules at the surface facing the hydrogen sink;

[0087] 6. Desorption of hydrogen molecules from the surface facing the sink; and

[0088] 7. Diffusion of hydrogen molecules away from the surface.

[0089] The rate of hydrogen transport can be limited by any one of the listed steps or by a combination of steps. The rate limiting step may vary depending upon the exact membrane design. Diffusion of hydrogen molecules in the gas phase to and from the membrane surface (Steps 1 and 7 ) is normally very rapid compared to the other steps. However, if differential pressure across the membrane is extreme, and it becomes necessary to make the porous layers very thick, the diffusion of molecular hydrogen through the tortuous pores can become rate limiting. Step 2 may become rate limiting if the surface area of the hydrogen dissociation catalyst is limited or if adsorption sites are occupied by other components of the gas mixture, such as CO. Desorption of the other components from the surface of the catalyst may be rate limiting and may be necessary before hydrogen adsorbs and dissociates. Electron transfer reactions: H⇄H++e− are usually extremely rapid, however, ionization of hydrogen atoms to protons is highly endothermic and requires considerable energy, and Step 3 could be rate limiting if all other steps are rapid. Recombination of protons with electrons (the reverse reaction of Step 5 ) on the hydrogen sink side of the membrane should be very rapid and not be rate limiting. Step 4 , diffusion of protons through the metal, will be rate limiting if the metal is thick, and all other steps are rapid. Although much of the literature describes hydrogen transport in the form of protons, transport as hydrogen atoms, transport as proton-electron pairs, or transport as hydride ions cannot be ruled out in all cases. Step 6 , desorption of hydrogen molecules from the surface facing the sink, will be endothermic in all cases and will require energy. Step 6 can become rate limiting if all other steps are rapid.

[0090] For hydrogen to adsorb on a metal surface held above cryogenic temperatures, hydrogen molecules must dissociate into hydrogen atoms and chemisorb. Thus, Steps 2 and 3 can be combined into one dissociative chemisorption step:

H ₂ (gas)+2*→2H (ad); ΔH _adsorption negative.

[0091] Similarly, desorption Steps 5 and 6 are combined as follows:

2H (ad)→H ₂ (gas)+2*; ΔH _desorption positive.

[0092] The “*” indicates an unoccupied surface site and H(ad) represents adsorbed hydrogen atoms on the metal surface.

[0093] Diffusion of hydrogen across dense metal membranes occurs by diffusion of hydrogen in a dissociated form and not by diffusion of hydrogen molecules. It is thus essential that hydrogen molecules be dissociated first into atoms on the surface of the membrane facing the hydrogen source. The surface of the membrane facing the hydrogen source should be capable of catalytically dissociating hydrogen into adsorbed atoms. This implies that pairs of adjacent surface sites should be maintained in reasonable concentrations to facilitate dissociative adsorption of hydrogen. It is desirable that occupation of surface sites by sulfur, CO, carbon, or other adsorbates, which block adsorption and dissociation of hydrogen, should be minimized. In this invention, membranes are preferably provided with catalysts, particularly those containing cobalt, cobalt-molybdenum, iron, magnetite, lanthanum strontium cobalt oxide, Pt, Ir, WS ₂ or MoS ₂ which are resistant to poisoning and particularly resistant to poisoning by sulfur.

[0094] We have found that hydrogen flux in the membranes of this invention is proportional to the square root of hydrogen partial pressure. This implies that hydrogen is transported in a dissociated form as noted above. Dissociated forms of hydrogen include protons (H ⁺ ), hydride ions (H ⁻ ), neutral atoms, or as proton-electron pairs with proton/electron separation sufficiently great that the proton and electron are considered a pair rather than a neutral ion.

[0095] Although, we currently consider proton transfer to be the most likely mechanism for hydrogen transport, we do not wish in any way to be limited by this proposed mechanism. Further, hydrogen transfer may occur by different mechanisms in different materials employed in the membranes of this invention or hydrogen transfer may occur by several mechanisms in a given material.

[0096] After transport through the membrane and electron transfer, the dissociated hydrogen must recombine into hydrogen molecules and desorb from the membrane surface facing the hydrogen sink. Desorption requires input of energy. To facilitate desorption, the surface of the membrane facing the hydrogen sink should have surface sites having the lowest possible desorption energy for hydrogen. If metal hydrogen transport layers in the membrane are sufficiently thin, and bulk diffusion and other steps are no longer rate limiting, then desorption from the membrane can become rate limiting.

[0097] Table 1 lists chemisorption energies of common metals which are useful for hydrogen desorption (See Benzinger 1991). 1

[0098] All of the listed metals have desorption energies of less than about 270 kJ/mol. Note that desorption energy varies with crystal face. While all of the listed metals in Table 1 will function for hydrogen desorption. Silver has by far the lowest desorption energy for hydrogen of the listed metals. Silver sites at the surface of the membrane facing the hydrogen sink will enhance hydrogen desorption rate and enhance the overall rate of hydrogen transport through a membrane. Surface chemistry of other metals that function for hydrogen transport can be improved by alloying with silver. In such mixtures, silver at least in part segregates to the surface of the metal layer and thus can facilitate desorption of hydrogen from the membrane surface facing the hydrogen sink. Other common metals such as Fe, W, Mo, Nb, and Ru have energies for hydrogen desorption ranging from 276-293 kJ/mol and are less well suited for desorption of hydrogen compared to the metals (and alloys thereof) listed in Table 1. However, alloying Fe, W, Mo, Nb, or Ru with any of Ag, Co, Cu or Pt would improve their ability to desorb hydrogen.

[0099] Palladium is recognized as the most commercially successful hydrogen transport membrane material. However, it is rapidly poisoned by sulfur and it does not function as a hydrogen dissociation catalyst or for hydrogen transport if sulfur adsorbs at its surface. According to Amandusson (2000), only one third of a monolayer of sulfur is sufficient to completely poison hydrogen adsorption on Pd (111). Assuming sulfur contamination of 500 ppb, equivalent to a partial pressure of 3.8×10 ⁻⁴ torr, a palladium surface will be poisoned in less than one second, assuming that a surface is completely saturated with a gas in one second if the partial pressure is 1×10 ⁻⁶ torr and the sticking coefficient is one (i.e. every molecule which strikes the surface adsorbs).

[0100] Table 2 lists exemplary activation energies for bulk diffusion of hydrogen through metals based upon data of D. N. Beshers, (1973). Vanadium, niobium and tantalum all have superior bulk diffusion properties for hydrogen compared to palladium. However, the catalytic properties of these metals for hydrogen dissociation are inferior to those of palladium. In addition, these metals are susceptible to oxidation and to carbide and nitride formation by reaction with carbon and ammonia which may be present in hydrogen source gas. 2

[0101] In general, the energies for bulk diffusion of hydrogen ions through the exemplified metals as seen in Table 2, are quite low (5.6-45 kJ/mol) compared to the desorption energies of hydrogen molecules from metals (218-293 kJ/mol). As the metallic layers in membranes are reduced in thickness, desorption of hydrogen molecules from the side of the membrane facing the hydrogen sink will be come rate limiting and the rate of diffusion of hydrogen through the bulk material will be become less important compared to surface desorption. When thin metallic layers are used to transport hydrogen, it is possible to replace palladium with less expensive metals, such as nickel, cobalt and iron and alloys thereof, having superior surface properties or with vanadium, niobium or tantalum coated with catalysts.

[0102] For clean surfaces, desorption of hydrogen may become rate limiting as the thickness of the metallic layer is decreased below one micron in the case of palladium. In contrast, when gases are present in the hydrogen source that poison the catalysts that facilitate dissociation of molecular hydrogen, dissociation of molecular hydrogen will become rate limiting. This invention employs catalysts containing cobalt, molybdenum or iron which are resistant to poisoning by sulfur. Although sulfur adsorbs on cobalt and transforms the surface into a sulfide of cobalt, the sulfided surfaces retain catalytic activity for breaking hydrogen-hydrogen bonds. Iron based water-gas-shift catalysts such as 90 to 95 weight % Fe ₂ O ₃ with 5 to 10 weight % Cr ₂ O ₃ with sulfur tolerance can also be used to dissociate hydrogen in the membranes of this invention, as can cobalt analogs of the water-gas shift catalyst.

[0103] Carbon monoxide can be a component of the hydrogen source gas. For example, carbon monoxide is generated along with hydrogen and carbon dioxide during reforming of natural gas, coal, and petroleum. Carbon monoxide poisons catalysts including Pd by occupying surface sites needed for dissociation of hydrogen. The more strongly CO adsorbs to a metallic surface, the more significant detrimental effect it will have on catalysis. Table 3 (from data of D. N. Beshers, 1973) lists the desorption energies of CO from exemplary metals. Palladium has the highest heat of desorption for CO of the metals listed and therefore will be most susceptible to poisoning by CO. Improvements can be achieved in resistance to poisoning by CO by replacing palladium (in whole or in part) by metals with lower heats of desorption for CO, for example, Ag, Cu, Co or Ni. Addition of Ag to Pd should result in improved resistance to CO poisoning of the catalyst. However, if sulfur is also present, Ag at the metal layer surface will be rapidly transformed into silver sulfide. Copper with a relatively low CO desorption energy should also exhibit resistance to CO poisoning. Cobalt should be a good balance given improved resistance to CO poisoning compared to palladium and also providing for sulfur resistance as discussed above. 3

[0104] Adsorption of CO can also poison a catalyst surface by depositing carbonaceous residues by the reaction: 2CO→C+CO ₂ . This detrimental reaction can be countered by adding a large excess of steam, carbon dioxide or hydrogen to the system to remove surface carbon.

[0105] Carbon dioxide can also poison the catalyst surface by the formation of stable carbonates. To avoid or minimize such poisoning, metals which form stable carbonates should be avoided. Cobalt carbonate decomposes at 52° C. and will not be stable under water-gas-shift temperatures of 350-450° C.

[0106] In summary, hydrogen transport membranes and particularly those that are compatible for integration with sulfur tolerant water-gas-shift catalysis should have the following properties:

[0107] 1. The membrane surface facing the hydrogen source should be capable of adsorbing and dissociating hydrogen and should be resistant to sulfur, CO, CO ₂ , ammonia, steam, and carbon;

[0108] 2. The membrane surface facing the hydrogen sink should comprise a metal or metal alloy having a low desorption energy for hydrogen; and

[0109] 3. The membrane material should have a low activation energy for bulk diffusion of hydrogen or should be made sufficiently thin (preferably less than one micron) that bulk diffusion is not rate limiting.

[0110] The first two properties are obtained by selection of materials, particularly metals or metal alloys, for use in the membranes herein. In order to make metal membranes sufficiently thin so that bulk diffusion is not rate limiting, this invention employs thin deposits of metals or metal alloys formed and supported in the pores of porous ceramic membranes. The metals or metal alloys that facilitate hydrogen transport and desorption are chemically deposited into the ceramic pores to plug them. In specific embodiments, metals are deposited into the pores of ceramic by infusing aqueous solutions of corresponding metal salts into the pores and reducing the metal ions of the infused salt with heating to deposit metallic layers in the pores. In a preferred embodiment, metals are deposited into the pores by chemical vapor deposition, in which volatile compounds of the metals are decomposed to deposit the metal. In other preferred embodiments, metal foils or cermets are used.

[0111] Thin foils of hydrogen-permeable metals or cermets containing hydrogen-permeable metals may be employed to obtain hydrogen-permeable membranes. Various composite membranes structures can be prepared.

[0112] Thin foils of hydrogen-permeable metal can be employed in combination with porous supports. The foils can be applied or attached to a porous metal (or alloy) or ceramic support, or can be positioned or held (e.g., by clamps or other holders) in contact with a porous support or held between two porous supports. The supports can be made of metal (or alloy), ceramic or other inorganic material or an organic polymer or resin. In another alternative method, the thin foil may be coated on one or both sides with a material which forms a porous support. For example, the foil may be coated with a ceramic adhesive or ceramic paste which forms a porous ceramic. In another example, the foil may be coated with an organic polymer or polymer precursor which forms a porous polymer or resin support.

[0113] A cermet is a composite materials that has a metal component (containing one or more metals or an alloy) and a ceramic component. Cermet membranes of this invention can be formed in several ways. For example, powders of metal (or alloy) and ceramic are combined in a desired ratio (preferably the range of metal or alloy employed ranges from about 40-60 volume %), optionally with one or more binders and sintered. The sintered cermet may be made sufficiently thick to be self-supporting (e.g., preferably 100-500 microns thick). Alternatively, the mixed metal (or alloy), ceramic and binder(s) can be formed into an applique, preferably ranging from 10-50 microns thick, which is applied to a porous ceramic support in the green state. The green support with cermet applique is then sintered together to form a cermet containing composite membrane. In another alternative method, slurries of cermet powders can be coated onto porous supports by dip-coating followed by sintering to form hydrogen-permeable membranes.

[0114] Hydrogen transport membranes of this invention preferably exhibit mechanical strength such that they have an extended useful lifetime and are resistant to mechanical stress which results in cracking and leakage.

[0115] It has been found that the porous ceramic and the metal or metal alloy to be deposited in the pores of the ceramic or coated on the porous ceramic, or the materials used to fabricate a cermet, can be selected to minimize mechanical stress in the composite membrane by matching the crystal lattice constants of the ceramic to the metal. This can be done, for example, by selecting an appropriate ceramic with lattice constants to substantially match those of a metal or metal alloy that provides desired high permeability for hydrogen. In a specific embodiment, the stoichiometry of a mixed metal oxide is adjusted so that the crystal lattice constants of the ceramic formed from it will substantially match the lattice constants of the selected metal or metal alloy.

[0116] Lattice constants in crystalline materials are routinely measured using x-ray diffraction, although electron diffraction is also used. Commercially available, x-ray powder diffractometers are convenient for the measurements. In these instruments, polycrystalline samples are exposed to an approximately monochromatic x-ray beam. Angles of x-ray diffraction maxima are measured. If two materials have diffraction maxima at the same measured angle, then this implies that they both have an identical lattice spacing corresponding to this diffraction maximum. From the known wavelength of the incident x-rays, and the measured angle of diffraction maximum, the atomic distances between crystal lattice planes can be calculated.

[0117] Composite materials used as membranes for gas separation are greatly improved by controlling the number of dislocations at internal interfaces. Interfacial dislocations enhance diffusion of many substances through dense membranes. However, in the case of composite membranes in which it is necessary to separate and purify hydrogen from gas mixtures contaminated with oxygen, nitrogen, carbon and sulfur, it is desired to minimize the number of dislocations at the interfaces of the composite so as to minimize diffusion of the contaminants through the membrane. Interfacial dislocations are minimized by matching the crystallographic lattices of the composite materials. In lattice matching, the pair of materials in the composite is chosen to have crystallographic planes of similar symmetry as well as closely matched interatomic lattice spacings. Lattice symmetry and lattice spacings are determined by x-ray or electron diffraction. To minimize dislocations at interfaces, the lattice mismatch is ideally kept below 1-2%, although mismatches up to 15% may be tolerated, depending upon interatomic forces and film thickness, which determine the stress at the interface. In composite membranes used to separate oxygen from air, diffusion of oxygen is enhanced by intentionally increasing the number of interfacial dislocations in a composite by increasing the misfit above 15%. Control of interfacial dislocations between thin films and their substrates through control of lattice match has long been used in the semiconductor industry.

[0118] In defining misfit and mismatch of lattices, the convention of van der Merwe (1984) is adopted. Misfit specifically refers to quantification of dimensional differences, including differences introduced by thermal expansion. Mismatch includes also misorientation and differences in symmetry between substrate and overlayer. One-dimensional misfit, f, can be mathematically defined as f=(o−s)/s where o is the distance between lattice atoms in a particular crystallographic direction in the overlayer, o, and s is the distance between lattice atoms in a parallel crystallographic direction in the substrate, s.

[0119] For example, for palladium deposited into the pores of a ceramic, the ceramic substrate is chosen to match both the crystallographic symmetry and the lattice constants of palladium. Elemental palladium has the face centered cubic crystal structure with a cube edge of 3.89 Å at room temperature. Because of the symmetry in the cubic system, if the cube edges match, the major other lattice spacings will also match. In the preferred embodiment, a ceramic substrate is chosen with cubic crystal symmetry and with a lattice spacing in a cube face close to 3.89 Å and preferably within the range 3.80 Å<x<3.96 Å to yield a misfit of less than 2%. A specific example is palladium deposited in the pores of the cubic perovskite material La _0.5 Sr _0.5 CoO _3−z .

[0120] In the more preferred embodiment, the ceramic substrate is chosen to have lattice constants of palladium at the operating temperature of the membrane. In the specific case of using membranes to extract hydrogen from a water-gas shift reaction mixture, the preferred temperatures are between 200 and 500° C. In a specific example, the composition of La _x Sr _1−x CoO _3−z is varied between x=0.8 and x=0.4 to vary the lattice spacing of the perovskite to match that of palladium at the operating temperature. In a further preferred embodiment, the composition of the perovskite substrate is varied between x=0.8 and x=0.4 to match that of palladium which has changed its lattice constants due to absorption of hydrogen, in addition to changes due to temperature.

[0121] Although misfits of less than 2% are most desired, it is possible for misfits up to 15% to be tolerated without the formation of dislocations. For the case where membranes are used for separation of hydrogen from a contaminated gas mixture containing oxygen, carbon, sulfur and nitrogen, and where very pure hydrogen is desired, it is desired to eliminate dislocations at the metal-ceramic interface by minimizing lattice misfit. However, if it is desired to separate larger atoms, specifically oxygen, from a gas mixture, specifically air, then it is desired to maximize the number of dislocations at the metal-ceramic interface to allow enhanced diffusion of oxygen. In the case in which dislocations are desired, lattice misfit is intentionally increased to values where dislocations spontaneously form.

[0122] Theoretically, there are an infinite number of lattice constants in any crystalline material, and it would not be practical to match all lattice constants between two materials unless both have the identical crystal structure. In the example of lattice matching between palladium and other materials with cubic crystal symmetry, if one cube edge matches, then by symmetry, all cube edges automatically match as well as all cube face diagonals and cube diagonals and many other lattice constants. For good lattice matching in the case of two cubic materials sharing the same symmetry there should be good lattice matching (less than about 15% mismatch and preferably less than 10% mismatch and more preferably less than about 2% mismatch) in all of the most important crystallographic planes of low Miller index.

[0123] In the more complicated case of lattice matching body centered cubic metals such as Nb, V and Ta to an alumina substrate, mismatch is small only on a few select planes such as the (011) plane of niobium deposited atop the (1120) planes of alumina. Never-the-less, interfaces with a minimum of dislocations can be produced. Lattice matching in these complicated cases is approximated by calculating a one-dimensional lattice matching in a particular crystallographic plane and calculating a second one-dimensional lattice matching in a crystallographic direction perpendicular to the first, but in the same plane. This type of approximation is consistent with methods commonly used in the literature of epitaxial growth and allows the selection of compatible materials.

[0124] Lattice matching can be performed between carrier materials that are refractory materials, ceramics, metals or metal alloys and the metal or metal alloys that are to be introduced into the pores of the carrier or coated on the porous carrier or support. Lattice matching does not depend upon the size of the pores in the carrier material and so can be employed with any porous materials.

[0125] Lattice matching does not apply to composite membranes of this invention that employ non-crystalline organic polymers or resins as components.

[0126] A specific example of a lattice matched composite system for minimizing interfacial dislocations and thus minimizing diffusion of oxygen, nitrogen, carbon and sulfur in a hydrogen transport membrane is palladium metal supported in a porous perovskite ceramic of La _0.5 Sr _0.5 CoO _3−z . Both materials have cubic crystal symmetry. The lattice constants of this perovskite material have been adjusted to match those of palladium, within 1-2% by varying the relative amounts of La and Sr. Other materials with the perovskite crystal structure can also be synthesized to match the lattice constants of palladium. Alternatively, the lattice constants of the Pd can be varied to match ceramic substrates by alloying the Pd with other metals. Because of the similarity of platinum to palladium, a second example of a hydrogen transport composite membrane is platinum supported by La _1−x Sr _x CoO _3−z . The ceramic need not be limited to perovskites. Further examples of lattice matched cermets for hydrogen transport membranes include niobium on porous alumina, tantalum on porous alumina, and molybdenum on porous alumina and vanadium on porous aluminum. In all of these cases the (011) crystallographic planes of the body centered cubic metals are very well lattice matched to the (1120) planes of the Al ₂ O ₃ . Thermal expansion between these metals and alumina is also well matched.

[0127] Table 4 provides a short list of perovskite ceramics which are well lattice matched to palladium at 298 K. This list is by no means exhaustive, and many other perovskites (of general formula, A _1−x A′ _x B _1−y B′ _y O _3−z , where z is a number that rendered the compound neutral) with various metal elements substituted into the “A” and “B” lattice sites can be synthesized to give similar lattice matching. Table 4 lists values only for Pd(100)//perovskite (100) and Pd[100]//perovskite[100], using standard crystallographic notation for crystallographic planes and crystallographic directions. Note that some of the perovskites such as LaFeO _3−x and

[0128] SrTiO _3−x in Table 4 have basically zero mismatch with palladium and have perfect lattice matching at 298 K. Both the palladium and perovskite have cubic symmetry, and therefore there exist many crystal planes which have good lattice matching.

[0129] Palladium supported on a-alumina has been used for hydrogen-permeable membranes. Even though palladium which is face centered cubic and alumina which is hexagonal do not share the same crystallographic symmetry, it is possible to find some planes of reasonable lattice match between cubic palladium and hexagonal alumina. For example, lattice matches can be found for the crystallographic planes of Pd(111)//Al ₂ O ₃ (1120) and the crystallographic directions of Pd[110]//Al ₂ O ₃ [0001]. Using lattice constants for hexagonal alumina of a=4.76 Å and c=13.01 Å (using standard crystallographic values at 298 K and standard notation for the a and c axes of a hexagonal crystal) and that a=3.8902 Å for the cube edge of face centered cubic palladium, the cube diagonal of palladium has a distance of 5.50 Å. Two such cube diagonals would span a distance of 11.0 Å, which would match the c distance of α-alumina of 13.01 Å with a misfit of (11.0-13.01)/13.01×100%=−15.4%. In the perpendicular Pd[112] direction Pd palladium has a lattice constant of 4.76 Å in the Pd(111) plane, which matches exactly with the a spacing of alumina of 4.76 Å. However, this perfect lattice match in one direction would be combined with the relatively poor match of −15.5%, in the perpendicular direction, which is near the limits of acceptable lattice match. 4

[0130] Of the materials in Table 4, LaFeO _3−z , LaCrO _3−z and mixtures of LaFe _1−y Cr _y O _3−z (which form solid solutions from 0 to 100% Fe, balance Cr, and which all have lattice constants expected at 3.88 to 3.89 Å), BaTiO _3−z provide hydrogen-permeable membranes in combination with Pd and its alloys (or other hydrogen-permeable metals and alloys to which they are lattice matched) that exhibit improved mechanical stability and operating lifetime. The titanates CaTiO _3−z , SrTiO _3−z will also provide improved hydrogen-permeable membranes in combination with Pd metal.

[0131] Some porous ceramic, metal or metal alloy carriers are available commercially or can be prepared using known methods or by routine adaptation of known methods. Methods for coating or deposition of metals or metal alloys onto porous ceramics are known in the art. For example, deposition of palladium and palladium alloys onto commercially available porous alumina and porous stainless steel is well known in the art. Specific examples herein employ V, Ta, or Nb, on alumina, and Pd on various perovskites. Metal and metal alloy foils of thickness appropriate for use in this invention are commercially available or can be made by art-known methods. Cermet materials useful in the present invention may be commercially available or can be prepared by art-known techniques in view of the teachings herein. Ceramic pastes and/or adhesives may be commercially available or may be prepared by methods known in the art. For example, commercially available high temperature alumina paste or cement, such as Cotronics 903HP ceramic adhesive, can be used to form porous ceramic supports.

[0132] Organic polymers, including organic resins, can be employed in the membranes of this invention as porous supports for hydrogen-permeable metal layers or to block pores of porous hydrogen-permeable materials. The polymers or resins employed must maintain mechanical integrity at the selected operation temperature contemplated for the membrane. Judicious selection of the polymer resin material for use at elevated temperatures is required. Suitable polymers or resins exhibit stability and retain mechanical integrity after initial setting or hardening for long-term use (preferably 100's of hours, and more preferably 1000's of hours) at operational temperatures (e.g., at or above about 300° C.). Suitable polymers or resins do not exhibit substantial decomposition and do not exhibit substantial deformation at selected operational temperatures.

[0133] Organic polymers and more specifically organic resins for impregnation or blocking of pores of a porous support should have a viscosity which allows the polymer or resin to freely flow into support pores. Preferred polymers or resins have viscosities in the range of about 100 to 1000 centipoise and meet this requirement. The polymer or resin system used to block pores must have a suitable “working life” during which the viscosity remains sufficiently low before the polymer or resin set or hardens in order to flow into pores over the surface or surfaces of the support that will be exposed to gases. The length of this working life is particularly important when blocking pores over a large surface area. Resins that exhibit low viscosity over a long time period (1-60 hrs) are preferred in this application.

[0134] Polyimides (see Ghosh and Mittal (1996) and/or Wilson & Stenzenberger (1990)) are a class of polymeric materials which have found widespread use in a number of high-temperature applications, including aerospace structural and engine parts, automotive exhaust and engine components and industrial parts exposed to high temperatures, including some electronic components. Polyimides exhibiting stability and mechanical integrity at operational temperature, e.g., above about 300° C. are specifically useful for forming membrane support materials. Polyimides which in addition exhibit sufficiently low melt viscosity to flow and facilitate impregnation of a porous substrate are useful for blocking pores in the membranes of this invention.

[0135] Some polyimides have been developed which have viscosities that make them particularly suitable for impregnation of porous substrates. Moreover, these materials display excellent mechanical properties at elevated temperatures. PETI-298 and PETI-330 polyimide resins available from Eikos Inc., exhibit low and stable melt viscosities and offer solvent free processing. These resins have glass transition temperatures of 298° and 330° C., respectively and are examples of polyimide resins exhibiting a glass transition temperature of about 300° C. or more. Long-term use for greater than 1000 hours at temperatures of 288° C. in air has been demonstrated for these exemplary resins.

[0136] PETI-298 and PETI-330 are copolymers prepared from 1,3-Bis(4-aminophenoxy) benzene, (1,3,4-APB)3, 4′-oxydianiline, (3,4′-ODA), 3, 3′,4, 4′-biphenyl tetracarboxylic dianhydride, (s-BPDA) and end-capped with 4-phenylethynylphthalic anhydride (4-PEPA): 1

[0137] Solvent borne polyimides are also commercially available as 50% solid solutions. These polyimides are typically rated for higher temperature use than neat melt processable polyimides. Polyimide resins such as RP46 (Unitech, LLC. Hampton Va.), a PMR-type polyimide(see U.S. Pat. No. 5,171,822 (Pater)) and the Skybond (TM) 700 series (IST, Industrial Summit Technologies, Japan) of aromatic polyimides (available as solutions of polyimide precursors) have been recommended for use at 370° C. and 350° C., respectively. Methods for forming resin elements employing these and other resins, particularly polyimide resins are known in the art.

[0138] Porous polymers for use as supports or carriers in membranes of this invention can be made, for example, by co-polymerizing a polyimide (or other highly thermally stable polymer or resin) with a thermally labile material such as polystyrene, polypropylene oxide, or a polymethyl methacrylate.

[0139] In one general embodiment, a porous support matrix of ceramic or material other than the metals or metal alloy which are permeable to hydrogen, is fabricated separately, and the pores are then blocked with a metal or alloy, which is permeable to hydrogen. As examples, porous alumina is fabricated first, then pores are blocked by mechanically clamping thin metal or alloy foils (e.g., of V, Nb, or Ta or their alloys) on to or between porous ceramic (e.g., alumina) or using a commercial high temperature ceramic paste or adhesive (e.g., alumina adhesive) to form a bond between the foil and ceramic substrate. Further examples include electroless deposition of palladium or other hydrogen-permeable metals or alloys into porous perovskite materials, which are lattice matched to the palladium or other metal or alloy.

[0140] In another embodiment, (see FIG. 1 ) fine powders of V, Nb, Ta, Zr, Pd, and their alloys are mixed with fine powders of ceramic, and are then pressed and sintered to form dense cermets. Preferably continuous matrices of both metal and ceramic are formed in the cermet after the material is sintered to facilitate hydrogen-permeability. Powders of metal (preferably from about 40-about 60 vol %) mixed with a ceramic provide a cermet that is hydrogen permeable. It is believed that the use of this range of metal and ceramic facilitates production of the desired continuous metal matrix. Again it is believed that the continuous matrix of the hydrogen-permeable metal provides permeability for hydrogen and allows transport of hydrogen, whereas the continuous phase of ceramic provides mechanical support and provides re-enforcement of the metals, which may become embrittled by hydrogen.

[0141] In certain embodiments, the ceramic employed is a proton conductor, such as a perovskite proton conductor, which may serve a dual purpose and may also transport hydrogen as well as act as a mechanical support. In the most preferred embodiments, the ceramic and metal of the cermets are lattice matched at the atomic level, in order to form a coherent interface between the ceramic and metal. Lattice matching minimizes interfacial stress. Interfacial stress can lead to the formation of dislocations which give rise to potential leak paths or which can initiate cracks. Specific combinations for the formation of lattice matched cermets include the following. Powders of niobium, tantalum or vanadium are mixed with powders of alumina and are sintered together to form dense cermets with minimum pore volume. In all of these cases the (011) crystallographic planes of the body centered cubic metals are very well lattice matched to the (1120) planes of the Al ₂ O ₃ . Thermal expansion is also very well matched in these combinations of metals and ceramic. In an additional example, a powder of palladium is mixed and sintered with a powder of the perovskite LaFe _0.9 Cr _0.1