Metal composite membranes for hydrogen separation

Journal of Membrane Science, 94 (1994) 299-311 Elsevier Science B.V., Amsterdam

299

Metal composite membranes

for hydrogen separation

A.L. Athayde, R.W. Baker*, P. Nguyen Membrane Technology and Research, Inc., 1360 Willow Road, Menlo Park, CA 94025, USA

(Received March 11,1993; accepted in revised form January 5,1994)

Abstract Further advances in the field of gas separation membranes will require materials with higher permeation selectivities than the currently available polymers. Metal membranes offer such an alternative. Because the mechanism for gas permeation through metals differs from that through polymers, metal membranes with very high selectivities can be made. Gas permeation through these membranes is very slow, however, and membranes with adequate fluxes are difficult to make. High membrane cost has also limited development because some of the best membrane materials are precious metals. Membrane Technology and Research, Inc. is pursuing a new approach to preparing metal membranes that offer the high selectivity inherent to permeation through metals with the high fluxes typical of polymeric membranes. We have prepared ultrathin metal composite membranes by sputter-deposition of a 76 atom% palladium/24 atom% silver alloy layer onto a conventional polymeric gas separation membrane. Preparation of membranes wiih different thicknesses of the metal layer (250-l,OOOA) under different deposition conditions (75-260 Almin) showed that the best membranes are obtained at high metal deposition rates. These membranes have a hydrogen flux of 1 x lo@ cm3 (STP) /cm2 s cmHg at room temperature (25’ C ) and a hydrogen/carbon dioxide selectivity greater than 100. In contrast, the best commercial polymeric membranes have a hydrogen/carbon dioxide selectivity of 6. The hydrogen flux through the membrane increases with temperature, but the fluxes of all other gases are unchanged. The net effect is an increase in membrane hydrogen flux and selectivity at high temperature. We have demonstrated that the membranes are stable for over 6 weeks continuous operation at room temperature, Key words: composite membranes; membrane preparations silver alloys; gas separations; sputter deposition

1. Introduction Since the commercialization of the Prism@’ membrane by Monsanto in the early 19808, gas separation using membranes has gained considerable attention in the chemical processing and refining industries [ 11. Membranes are in *Corresponding (415)328-6580.

author.

Tel.:

(415)328-2228;

Fax:

and structure; metal membranes; palladium/

use for the recovery of hydrogen from purge streams and for the production of nitrogen-enriched air. Membranes have failed, however, to make any significant impact on other important gas separations, such as the production of oxygen from air or the separation of hydrogen from carbon dioxide-rich streams. The main obstacle is limited membrane selectivity; the selectivity of the best polymeric membranes for

0376-7388/94/$01.00 0 1994 Elsevier Science B.V. All rights reserved. SSDZ 0376-7388(94)00042-W

300

A.L. Athayde et al. /J. Membrane Sci. 94 (1994) 299-311

oxygen/nitrogen is -6-8, and for hydrogen/ carbon dioxide, 6. A major current focus of research in the membrane gas separation area is, therefore, the search for highly selective membrane materials. Advances are being made, particularly in the synthesis of highly selective polymeric materials [ 21. The difficulty of making further advances, however, increases as the limits of the gas separation mechanism in polymer materials are reached. If dramatically improved selectivities are to be achieved, membrane materials that separate gases by a completely different mechanism are required. Metals are an example of such membrane materials. Study of gas permeation through metals and alloys began with Thomas Graham’s observation of hydrogen permeation through palladium. In the 1950s and 608, interest in the use of palladium-alloy membranes for producing high-purity hydrogen led to the installation of a pilot plant, by Union Carbide, for separating hydrogen from a refinery off-gas stream containing methane and ethane [ 31. Although the membranes were demonstrated successfully at several sites, the process proved uneconomical, because the membranes were too thick and the hydrogen flux was too low, and was eventually abandoned. The only commercial metal membrane system is Johnson Matthey’s palladiumalloy membrane for the laboratory-scale production of ultrapure hydrogen (99.999999 mol% ) [ 41. Metal membranes continue to be of great interest, however, because of their potential for permeation selectivities that are orders of magnitude greater than the selectivities of the best polymeric membranes. Palladium and its alloys have been studied extensively as potential membrane materials, in part because of their very high level of hydrogen sorption. Pure palladium absorbs 600 times its volume of hydrogen at room temperature [ 51. Hydrogen permeates most metals, including tantalum, niobium, vanadium, nickel,

iron, copper, cobalt, and platinum, but in most cases the membrane must be operated at a high temperature ( > 300’ C) to obtain a viable permeation rate [ 61. Oxygen-permeable metals are also known; oxygen permeation through silver membranes has been demonstrated [ 71. Metal oxides such as zirconia [8] and perovskites [9,101 are also highly permeable to oxygen and have been the focus of recent membrane-development efforts. This paper focuses on the development of highly selective palladium/silver-alloy membranes with high hydrogen permeability. 1.1. Gas transport in metal membranes The gas transport mechanism through metal membranes is the key to their high permeation selectivity. Gas permeation through a metal membrane is believed to follow a multistep process [ 111. Fig. 1 illustrates the permeation of hydrogen through a palladium-alloy membrane. Hydrogen molecules from the feed gas are sorbed on the membrane surface, where they dissociate into hydrogen atoms. Hydrogen atoms emerging at the permeate side of the membrane reassociate to form hydrogen molecules, then desorb, completing the permeation process. Only hydrogen is transported through the membrane by this mechanism; all other gases are excluded. Theoretically, the membrane has Hydrogen

molecules

.k I Metal membran*

Hydrogen

Btoms

I

Fig. 1. Permeation of hydrogen through metal membranes.

I

301

A.L. Athayde et al. /J. Membrane Sci. 94 (1994) 299-31 I

infinite selectivity, but the palladium membrane is typically a polycrystalline film with a fine grain structure. The grain boundaries form pathways for the diffusion of other gases, limiting the attainable selectivity. If the sorption and dissociation of hydrogen molecules is a rapid process, the hydrogen atoms on the membrane surface are in equilibrium with the gas phase. The concentration of hydrogen atoms on the metal surface, c, is given by Sieverts’s law, c= KPO.~

(1)

where K is Sieverts’s constant andp is the hydrogen pressure in the gas phase. At high temperatures ( > 300’ C ), the surface sorption and dissociation processes are fast, and the ratecontrolling step is diffusion of atomic hydrogen through the metal lattice. This is supported by the data of Holleck [ 121 and others, who have observed that the hydrogen flux through the metal membrane is proportional to the difference of the square roots of the hydrogen pressures on either side of the membrane. At lower temperatures, however, the sorption and dissociation of hydrogen on the membrane surface become the rate-controlling steps and the permeation characteristics of the membrane deviate from Sieverts’s law predictions. 1.2. Palladium-alloy membranes Most metal membrane research has focused on palladium membranes because of the very high hydrogen permeation rates that can be achieved at modest temperatures ( < 500°C). Hydrogen is readily sorbed into palladium, and diffusion of hydrogen through the palladium lattice is a rapid process. Below a critical temperature of 300 oC and a critical pressure of 20 atm, the hydrogen/palladium system exhibits two-phase behavior as shown by the hydrogen sorption isotherms plotted in Fig. 2. At 30°C and N 18 mmHg pressure, the hydrogen/pal-

Pressure 20 atm

11 atm

2atm

16 mmHg 0.1

0.2

0.3

0.4

Hydrogen/palladium

0.5

0.6

0.7

0.6

ratio

Fig. 2. Absorption relationship in the palladium-hydrogen system.

ladium ratio is 0.03 for the hydrogen-poor cy phase and 0.57 for the hydrogen-rich /3 phase

WI. Early experiments with palladium membranes resulted in mechanical failure of the membranes due to the expansion and contraction of the metal lattice as hydrogen was absorbed and released. Alloying palladium with silver, however, depresses the two-phase region below ambient temperature (25’ C ), stabilizing the membranes. Hunter reported X-ray diffraction measurements that substantiate the stabilizing effect of silver on the palladium/silver metal lattice [ 111. The dimension of the unit cell of the palladium/silver-alloy lattice expands from 3.92 to 3.96 A, upon saturation with hydrogen. Under the same conditions, the dimension of the unit cell in the palladium lattice changes from 3.89 to 4.02 A. The equilibrium sorption of hydrogen in palladium/silver alloys increases as the relative amount of silver in the alloy increases. The diffusion coefficient of hydrogen in the alloy, however, decreases with increasing silver content. As a result of these two opposing trends, the hydrogen permeability for a palladium/silver alloy has its maximum value at a silver content of 24 atom%. Stable membranes have also been made from palladium alloys with rare earth metals, particularly yttrium. Hydrogen sorption and diffu-

302

sion are both enhanced in palladium/yttrium membranes, resulting in higher permeation rates than for palladium/silver alloys [ 13,141. 1.3. Metal composite membranes The key issue in metal membrane development is to make the metal layer thin enough for practical use. If diffusion through the metal layer is the rate-controlling step, then the gas flux through the membrane is inversely proportional to the membrane thickness. Membranes must be made extremely thin to achieve high fluxes, but must also be strong enough to withstand the operating pressures. Palladiumalloy membranes as thin as 5 pm have been prepared, but these membranes must be operated above 300°C to obtain appreciable gas fluxes ( N 10m5 cm3 (STP)/cm2 s cmHg). The high diffusion coefficient of hydrogen in palladium ( 10-4-10-5 cm”/s) at these temperatures then offsets the limitation of the membrane thickness. Both Union Carbide and Johnson Matthey relied on high-temperature operation to achieve viable gas fluxes. If the gas is produced at less than 300 ’ C, then heating the feed gas stream to the operating temperature of the membrane adds to the energy and capital costs of the process. The membrane material cost is another incentive to reduce the membrane thickness. Metal membranes are typically made from expensive precious metals and rare earths that are 50-100 times more expensive than the highperformance polymers used in polymeric membranes. For a palladium-alloy membrane system, the cost of the alloy is likely to be the most significant part of the total capital cost. The hydrogen flux through a metal membrane is inversely proportional to the membrane thickness if the rate-determining step is the diffusion of hydrogen through the metal layer. For a fixed separation requirement, the membrane area is inversely proportional to the flux. The

A.L. Athayde et al. j J. Membrane Sci. 94 (1994) 299-311

total volume of metal required, which is the product of the thickness of the metal layer and the membrane area, is proportional to the square of the membrane thickness. Consequently, the system cost is proportional to the square of the metal layer thickness, and precious metal membranes must be ultrathin to be cost-effective. Our approach to producing ultrathin metal membranes is to assign the permeation and strength characteristics to different parts of a composite membrane [ 151. Palladium and palladium-alloy composite membranes prepared on polymeric [ 161, glass [ 171, and ceramic [ 181 support membranes using a very similar approach have been described in the literature. The metal membrane can be made very thin without compromising its mechanical strength. Our multilayer composite membrane, shown in Fig. 3, consists of a microporous polymer support membrane coated with one or more layers of suitable polymers to form a smooth, porefree substrate. A very thin layer (250-1,000 A) of metal or alloy is deposited on the substrate by a vacuum sputter-coating process. A protective polymer coating on top of the metal layer prevents abrasion of the metal layer during handling or testing. The polymeric substrate and protective overcoat give the composite membrane its mechanical strength. These highly permeable layers do not offer any resistance to gas permeation, relative to the metal layer. Therefore, the

/

Fig. 3. Structure of MTR’s multilayer, ultrathin metal composite membrane.

A.L. Athyde

303

et al. /J. Membrane Sci. 94 (1994) 299-311

gas flux and selectivity characteristics of the composite membrane are determined solely by the metal layer. The hydrogen flux through the composite membrane is high at ambient temperature because the metal layer is very thin; operating at higher temperatures is feasible but not necessary. Ultrathin metal composite membranes use a very small amount of metal and the materials cost is comparable to that of state-of-the-art polymeric membranes.

2. Experimental 2.1. Membrane preparation The support for the metal composite membranes is a conventional multilayer composite separation membrane. A polygas (dimethylsiloxane ) sealing layer was added to cover any pores, cracks, or defects, forming a smooth substrate for the metal layer. The poly (dimethylsiloxane ) -coated support membranes had a hydrogen flux of 7~ 10m5 cm3 (STP) /cm” s cmHg and a carbon dioxide flux of 1 x 10d4cm3 (STP) /cm” s cmHg. The sealed support membranes (substrate) were sputter-coated with a thin layer of 76 atom% palladium/24 atom% silver alloy to form the metal membrane. Before sputtercoating, the sealed support membranes were soaked in a 10 wt% nitric acid bath overnight, rinsed with deionized water, and vacuum dried, The membrane surface was cleaned with a commercial cleaner, Dust Off@ Plus, or ionized air to remove any adhering dust particles. Pieces of membrane were mounted on aluminum plates and placed in the sputtering chamber. During sputter-deposition, metal atoms, dislodged from a palladium/silver-alloy target by ions from an energized plasma, travel towards the substrate. The metal atoms are adsorbed on the substrate surface and diffuse across the

surface towards low-energy sites, where they form metal islands. The islands grow and merge to form a continuous coating. The sputter-deposition rate can be controlled very precisely and thin, high-quality films can be made reproducibly. The membranes for this study were sputter-coated on a DC magnetron sputter coater (Perkin-Elmer 2400, Perkin-Elmer Corp., Norwalk, CT), operated by Scientific Coating Lab, Inc. (Santa Clara, CA), which can produce membranes in 25 x 15 cm2 sheets. The effect of sputtering conditions on membrane quality was determined by performing sputter-coating runs at varying sputter-deposition rate, chamber pressure, and thickness of the deposited film. The values of the sputtering parameters used in the coating process are listed in Table 1. The resulting metal membranes must be treated with great care to prevent damage to the very thin (250-1,000 A) metal layer. To protect the metal layer against abrasion, we added a thin polymer protective overcoat of a highly permeable material. We used either poly (dimethylsiloxane ) or polyimide protective coating layers. Poly (dimethylsiloxane ) coated membranes showed no change in the hydrogen flux relative to the uncoated metal membranes. The hydrogen flux through polyimide-coated membranes was up to 50% less than that of the uncoated metal membranes.

TABLE 1 Sputter-deposition parameters Parameter

Value

Working gas Base pressure Sputtering pressure Voltage Cathode current Target to substrate distance Deposition rate

Argon 5x10-7t.orr 1.2x10-3torr 300-400 v 0.3-1.2 A 1Ocm o 75-260 A/min at 20 rpm

A.L. Athayde et al. JJ. Membrane Sci. 94 (1994) 299-311

304

2.2. Membrane testing

Measurements of gas permeation properties are the simplest and most reliable method of characterizing membranes. All of the membranes prepared in this program were tested with pure gases and gas mixtures to determine the gas permeability and permeation selectivity. The gas permeation apparatus shown schematically in Fig. 4 was used to measure the permeation properties of disks of the metal composite membranes. Pure-gas permeation tests were carried out with hydrogen, carbon dioxide, nitrogen, oxygen, and helium. The hydrogen gas was semiconductor purity grade (99.999 + % ); all other gases were industrial grade purity (99.99 + % ) . Disks of the membrane, 4.9 cm in diameter, were cut out and mounted in the test cells with the metal layer on the high-pressure side and the support membrane on the low-pressure side. The feed side of the membrane was pressurized with the test gas to 100 psig and the permeate side of the membrane was maintained at atmospheric pressure. The permeate flow rate was measured with a soap-film flow meter. The gas flux through the membrane was calculated from the flow rate and then normalized with respect

Permeate

PWmeet~

to the pressure difference across the membrane. The effective selectivity of the membrane for a given gas pair was calculated as the ratio of the normalized permeation rates. All membranes were tested at room temperature (25 ’ C ) ; some measurements were also made at high temperatures (up to 150°C) to determine the effect of temperature on the membrane properties. Membranes were also tested with equimolar hydrogen/carbon dioxide gas mixtures, using the same cell and a feed-side pressure of 100 psig. In the mixture tests, gas was continuously vented from the feed side to ensure sufficient mixing in the cell. The feed-gas flow rate through the cell was always maintained at least 100 times greater than the permeation rate through the membrane. We can therefore assume that the gas composition on the feed side of the membrane was uniform across the membrane surface. The feed and permeate streams were analyzed using an on-line gas chromatograph with a thermal conductivity detector. The atomic composition of the sputter-deposited metal layer was determined using Xray photoelectron spectroscopy (XPS-ESCA) . ESCA is a surface-sensitive technique that provides elemental and chemical bonding infor-

Permeate

Fig. 4. Schematic diagram of permeation apparatus for evaluating membrane disks.

A. L. A thayde et al. / J. Membrane Sci. 94 (1994) 299-311

mation for all elements of atomic weight greater than helium. The relative amounts of palladium and silver can be determined, as well as the presence of contaminants such as carbon, sulfur, and chlorine. The ESCA analysis has a sensitivity of - 1 atom% and provides information on a 100-A thick film extending down from the surface. The metal layer was sputteretched away in 200-A steps, and the exposed surface was analyzed to determine composition as a function of the membrane thickness. Larger-scale features of the membranes were examined using scanning electron microscopy. A glass slide was placed in the deposition chamber along with the polymer membrane substrates during sputter-deposition. The thickness of the metal film deposited on the glass slide was measured using a Dektak 3030 displacement measurement instrument. The thickness of the metal film on the glass slide agreed with the estimates of the metal membrane thickness from the scanning electron micrographs of the composite membrane. 3. Results 3.1. Structure and composition of the metal membranes

Ultrathin metal composite membranes were prepared by sputter-coating polymer membranes with a 76 atom% palladium/24 atom% silver-alloy target. Metal coatings 250-1,000-A thick were prepared by controlling the sputterdeposition rate and the deposition time. The composite membranes could be handled without damaging the metal film, indicating that the metal coating adhered well to the polymer substrate. A field emission source scanning electron micrograph of the cross section of the metal composite membrane is shown in Fig. 5 The 600-A thick metal layer is visible on the left edge of the micrograph, the l- to 2-p thick dense sealing layer is to the immediate right of the

Fig. 5. Scanning electron micrograph of the cross section of a metal composite membrane.

metal layer, and the polymeric support membrane extends to the right of the micrograph. We were unable to observe the grain structure of the metal film with the scanning electron microscope. The ESCA results for two membrane samples are reported in Table 2. The metal membranes were prepared without the protective overcoat so the metal layer could be analyzed directly. Only the surface layer of the first membrane sample was analyzed. The second membrane sample was analyzed at the surface, after etching away the metal layer to a depth of 200 A from the top surface, and again after etching to a depth of 400 A from the top surface. The ESCA analysis shows that the metal membranes are composed of palladium, silver, and a few impurities. The palladium/silver ratio at the surface is 3.3 : 1 to 3.9: 1, which is close to the ratio of the alloy target, 3 : 1. The alloy is richer in palladium at depths of 200 and 400 A from the surface, with a palladium/silver ration of 8.2 : 1 to 8.7 : 1. Carbon, oxygen, silicon, and small amounts of sulfur are present as impurities on the surface of both membranes. Sulfur, oxygen, and chlorine contamination does not extend into the bulk of the metal layer. Carbon, oxygen and sil-

A.L. Athayde et al. /J. Membrane Sci. 94 (1994) 299-311

306 TABLE 2

ESCA resultsz elemental composition data measured from the surface (approximately the top 100 A) of each sample and expressed in atomic percent units for the elements detected Sample Membrane #224 Membrane #224 Membrane #224 Membrane #226

(surface) (after 200-A etch) (after 400-A etch) (surface)

0

Si

S

Cl

C

Pd

Ag

19

22 4.0

2.0 -

-

14

11

2.2

1.2

46 11 7.6 39

9.8 75 83 25

2.5 9.2 9.5 7.6

icon are present in the ratio 2 : 1: 1. which is the composition of dimethylsiloxane. It is possible that the poly (dimethylsiloxane) from the support membrane may have migrated to the surface of the metal film during its formation. The dimethylsiloxane may also be deposited from diffusion pump fluid vapors that leak into the sputtering chamber. We also analyzed membrane samples by a wet chemical method to determine the bulk alloy composition of the metal layer. The metal layer was dissolved in acid and the solution was analyzed by inductively coupled plasma emission spectroscopy (ICP) . The analysis showed that the metal layer had an average composition of 84 atom% palladium/l6 atom% silver, confirming that the deposited metal layers are palladium-rich relative to the target. 3.2. Permeation of hydrogen and carbon dioxide through metal membranes

All the metal composite membranes were characterized by measuring their gas permeation properties. Fig. 6 is a plot of the pressurenormalized fluxes of hydrogen and carbon dioxide through the membranes. Data are presented for a number of different membranes, sputter-deposited from a 76 atom% palladium/ 24 atom% silver-alloy target, at different deposition rates, and with coatings of other metals. All of these membranes have a poly (dimethylsiloxane ) coating layer. The solid

lines correspond to membrane hydrogen/carbon dioxide selectivities, cy, of 1, 10, 100, and 1,000. The thickness of the metal layer, which ranges from 250 to 1,000 A, was estimated from the sputter-deposition rate and the sputtering time. The actual thickness of the metal layer will vary locally as a function of the deposition conditions and the roughness of the support membrane surface. The support membranes are prepared by a solution-coating process and should have a very smooth surface. We examined the support membrane surface using scanning electron microscopy but did not see any significant structure variation. The poly(dimethylsiloxane) -sealed support membrane had a pressure-normalized flux of - 10e4 cm3 (STP) /cm” s cmHg for both hydrogen and carbon dioxide. The support membrane was slightly more permeable to carbon dioxide than to hydrogen, with a hydrogen/carbon dioxide selectivity of 0.7-0.8. After the membranes were coated with the palladium/ silver-alloy layer, the hydrogen flux decreased lo- to loo-fold to 10-5-10-6 cm3(STP)/cm2 s cmHg. The permeation of carbon dioxide through the metal-coated membranes was severely limited and the pressure-normalized flux of carbon dioxide dropped to 10-6-10-7 cm3 (STP) /cm” s cmHg. The hydrogen/carbon dioxide selectivity of the membranes increased to lo-100 after deposition of the metal layer. A few membranes showed a hydrogen/carbon

A.L. Athayde et al. /J. Membrane Sci. 94 (1994) 299-311

307

0

support

0 ZOA Pd!Ag, 75kmin Hydrogen preeeurenormellxed flUXXlOb (cm3 (STP)/ cm2+cmHg)

A SOOAPdiAg, 75A/min 0 1,OOOAPdIAg, 75hmln A SOOAPd/Ag, 200A/min

10

n 1,OgOAPdIAg, 2WA/min l 60oA PdlAg, 25OA/min X WA Pd!Ag, ZOg&mln *SOA W, M/min 0 5ggA PdIAg, 75kmln +1WA R, lWs/min

Carbon dloxlde preeeure-normalized flux x106 (cm3 (STP)/om%cmHg)

Fig. 6. Gas flux through ultrathin metal membranes prepared under different deposition conditions.

dioxide selectivity greater than 100, which is two orders of magnitude better than that observed for polymeric membranes. Defect-free palladium/silver-alloy films greater than 10,000-A thick would be permeable to hydrogen only and would have an infinite hydrogen/carbon dioxide selectivity [ 191. Because our membranes have measurable carbon dioxide fluxes, we infer that there are defects in the metal layer. Hydrogen permeates the membrane by diffusion through the film as well as through the pores. Carbon dioxide and other gases can only permeate the pores. The high hydrogen permeability of the metal layer, however, ensures that the membranes have excellent permeation selectivity despite the presence of defects. For example, the support membrane is equally permeable to hydrogen and helium gases. After the metal layer is added, the composite membrane is lo-100 times more permeable to hydrogen than to helium even though the helium atoms are smaller (2.6 A) than the hydrogen molecules (2.89 A). The deposition rate of the deposited metal layer plays an important role in determining the flux and selectivity characteristics of the composite membrane. We prepared palladium/sil-

ver-alloy membranes at metal deposition rates of 75,200, and 260 A/min, corresponding to 0.1, 0.3, and 0.5 kW sputtering power. The data in Fig. 6 indicate that the carbon dioxide pressure-normalized flux through the membranes decreases as the deposition rate is increased. The hydrogen flux is not a strong function of the deposition rate and does not show a significant change as the deposition rate is varied. Therefore, membranes prepared at higher deposition rates have greater hydrogen/carbon dioxide permeation selectivity. The effect of varying the metal layer deposition rate is best illustrated by the hydrogen and carbon dioxide flux data for the 500- and 1,000-A thick palladium/silver-alloy membranes, prepared at deposition rates of 75 and 200 A/min. The carbon dioxide flux decreases as the deposition rate is increased for membranes of the same thickness. The deposition rate has no significant effect on the hydrogen flux. We have not fully explored the relationship between the metal deposition rate and the permeation properties of the metal membrane. Earlier studies of sputter-deposited films have shown that the structure of the film, particularly the grain size, number of lattice disloca-

308

tions, and extent of atomic rearrangement are functions of the deposition rate [ 20,211. The effect of coatings of other metals on the hydrogen flux was investigated by depositing other metals on top of the palladium/silver membrane. Results were mixed, for example, a 500-A thick palladium/silver membrane with 75-A thick palladium or 100-A thick platinum coatings did not show any significant change in the hydrogen flux relative to the palladium-alloy membrane. Coating the membranes with less permeable metals reduced the hydrogen flux through the membrane. Palladium/silveralloy membranes coated with a 100-A thick titanium layer had no detectable hydrogen flux at room temperature. 3.3. Pressure dependence of the hydrogen flux The hydrogen flux through thick ( > 10,000 A) palladium-alloy films is reported to be proportional to the difference in the square root of the hydrogen pressure on either side of the membrane, in accordance with Sieverts’s law. We studied the effect of feed pressure on the hydrogen flux to determine whether Sieverts’s law is applicable to our composite metal membranes. The membranes were tested with hydrogen at feed pressures varying from 25 to 175 psig and a permeate pressure of Opsig. The hydrogen flux measured for a 500-A thick palladium/silver-alloy membrane without any protective layer is plotted as a function of the feed pressure in Fig. 7. The data were fit to a simple Fick’s law model for diffusion through a composite membrane. The data agreed well with the model, and the pressure-normalized flux for the composite membrane estimated from the model is 2 x 10e5 cm3 (STP) /cm2 s cmHg. The pressure-normalized hydrogen flux of the support measured before depositing the metal layer was 7 x 10m5 cm3 (STP) /cm2 s cmHg. The support membrane contributes one third of the total transport resistance, as estimated by resis-

A.L. Athuyde et al. /J. Membrane Sci. 94 (1994) 299-311

tance model calculations for the Fick’s law model. We also fit the data to a two-layer composite membrane model in which permeation through the metal layer was governed by Sieverts’s law, and permeation through the polymer support membrane was governed by Fick’s law. The hydrogen flux through the two-layer composite membrane, Q, is given by Q+(pi5

-pp”)

=

f 0

(PI

-Pd

(2)

S

where PM is the permeability coefficient for the metal layer, ZMis the thickness of the metal layer, (P/l) s is the pressure-normalized flux of the polymeric support membrane, pn is the hydrogen pressure at the upstream side of the membrane, pI is the equivalent hydrogen pressure at the metal-support membrane interface, and pL is the hydrogen pressure at the downstream side of the membrane. The only adjustable parameter used to fit the data is the permeability coefficient for the metal layer, which was set to 65 x lo-” cm3 (STP) cm/cm2 s cmHe5. Published permeability data for palladium-alloy films were, however, obtained with thicker films and at higher temperatures, and no meaningful comparison can be made with our observations. Sieverts’s law behavior is observed when diffusion through the metal layer is the rate-controlling process. At temperatures below 200’ C, however, the sorption and dissociation steps at the metal surface are the rate-controlling steps, and the hydrogen flux is no longer inversely proportional to membrane thickness [ 221. There is not much difference between the flux predicted by the two models over the range of pressure studied, 20-200 psig. This may account for the good tit of the data by both models. Further work is required to determine the effect of the support membrane and protective

A.L. Athayde et al. /J. Membrane Sci. 94 (1994) 299-311

wdrw en‘,” xx ,os

15,000 -

309

Comporlt~nmnbrana

(cm~(STP)/cd.s) 10,000 Fick’s law model

Feed pro6sun

Fig. 7. Hydrogen flux through a 500-A thick palladium/silver

layer on the kinetics of hydrogen uptake and release by the metal at the polymer/metal interfaces. 3.4. Effect of temperature on membrane performance Permeation tests were carried out in the range 25100°C to determine the effect of temperature on membrane performance. Fig. 8 shows that the hydrogen flux through the membrane increases with temperature. The data are presented as a plot of the gas flux versus the reciprocal of the absolute temperature (K- ’ ), illustrating the Arrhenius-type relationship between the hydrogen flux and temperature. “C

Pressure-normalized flux x 106 (cm3(STP)/cm2~s-cmHg)

50

75

100

25

t

uI

1

I

0.0026

0.0026

0.0030

0.0032

0.0034

l/T (“K-‘)

Fig. 8. Hydrogen and carbon dioxide fluxes through an ultrathin metal composite membrane cs a function of membrane temperature. Membrane: 500-A thick palladium/silver composite membrane with a l-pm thick polyimide protective layer.

(psig)

metal composite membrane at 25°C.

Carbon dioxide cannot permeate the metal and must diffuse through the grain boundaries and defects in the membrane. The carbon dioxide flux variation with temperature is small; our data show no appreciable change in carbon dioxide flux over the temperature range 25100 oC. The hydrogen/carbon dioxide selectivity increases from 240 at 25’ C to 1,200 at 100” C. Membranes that have been heated above 70°C typically show some loss of flux upon cooling back to room temperature as shown in Fig. 8. The membrane initially had a hydrogen flux of 1.2 x lop5 cm3 ( STP)/cm2 s cmHg at 25 “C. After heating the membrane to 100” C and cooling it back to 25’ C, the hydrogen flux dropped to 3.7 x 10m6cm3 ( STP)/cm2 s cmHg. Metal composite membranes with a polyimide protective layer showed no further decline in hydrogen flux over successive thermal cycles. Metal membranes with a poly(dimethylsiloxane) protective layer were not stable to repeated thermal cycling and showed further loss of hydrogen flux. There was no change in the carbon dioxide flux after a single heating and cooling cycle. ’ Membranes that were tested through multiple thermal cycles showed up to 100% increase in carbon dioxide flux, probably due to the formation of defects in the metal layer.

A.L. Athayde et al. /J. Membrane Sci. 94 (1994) 299-311

310

3.5. Membrane stability

Metal membranes must demonstrate longterm stability if they are to be viable for industrial gas separations. The typical permeation test for our membranes lasted 48-72 hours; we have accumulated a vast amount of short-term stability data. Most of the membranes showed a decline in hydrogen flux over the first few hours of testing. Fig. 9 shows, however, permeation measurements for a poly(dimethylsiloxane) -coated palladium/silver-alloy membrane over a period of 1,200 h. As shown in Fig. 9a, the hydrogen flux is initially 2.8 x 10m6cm3 (STP) /cm” s cmHg but drops to 1.4 x 10v6over the first 24 h. Thereafter, the hydrogen flux stabilizes to 1x 10m6cm3 (STP) / cm2 s cmHg. After the initial drop in hydrogen

(a)

lo.w

0.00

I

1 0

(b)

100.0, 99.6

I-

99.0 96.6

Pmrmeme

hydrogen

concentration 0

I

. 200

. c

98.0 97.5

400

6w

6oa

l.m

1.m

1.4w

The(h)

I

I .’ .

n

.

-

l-.

g,,. 96.5 96.0

I

1

flux, the membranes are stable for at least 6 weeks. The carbon dioxide flux was originally 1x 10m8cm3 (STP) /cm” cmHg and increased to 2 x lo-’ by the end of the test period. The slow increase in flux is believed to be due to the formation of defects in the metal layer. The data in Fig. 9 were obtained with a 50/ 50 mol% hydrogen/carbon dioxide feed gas. The hydrogen and carbon dioxide concentrations in the permeate during the test period are shown in Fig. 9b. The hydrogen content in the permeate was originally above 98.5% and decreased to 98% by the end of the test. The hydrogen/carbon dioxide selectivity of the membrane remained above 50 over the duration of the test. 4. Conclusion Metal composite membranes represent an intermediate stage between today’s state-ofthe-art polymer membranes and the high-temperature metal membranes under development. Because of the very thin metal layer ( N 500 A), the membranes are inexpensive and can be operated at ambient temperature. They offer the high selectivity inherent to metal membranes at the operating pressures and temperatures of polymeric membrane systems. Palladium/silver-alloy membranes are particularly effective in separating hydrogen from other gases, including carbon dioxide and helium, separations that are difficult to achieve with polymeric membranes. Membrane selectivities greater than 100 have been measured and the membrane has been shown to be stable for over 6 weeks of continuous operation.

Time (h) Fig. 9. Long-term stability of palladium/silver alloy membranes. (a) Pressure-normalized gas fluxes as a function of time. Feed pressure: 1OOpsig. Feed: 50/50 mol% hydrogen1 carbon dioxide. Temperature: 25°C. Membrane: 500-A thick Pd/Ag metal composite membrane with a l-pm poly (dimethylsiloxane ) protective overcoat. (b ) Permeate composition as a function of time.

5. Acknowledgement The work is being supported by an SBIR grant from the U.S. Department of Energy Office of Basic Energy Sciences, Contract Number DE-FG03-90ER80949.

A.L. Athayde et al. /J. Membrane Sci. 94 (1994) 299-311

References [ 1] J. Volkl and G. Alefeld, Diffusion of hydrogen in metals, in G. Alefeld and J. Volkl (Eds.), Hydrogen in Metals, Springer Verlag, Berlin, 1978. [ 21R.A. Hayes, Polyimide gas separation membranes, US Pat. 4,717,394 (January 1988). [ 31 R.B. McBride and D.L. McKinley, A new hydrogen recovery route, Chem. Eng. Prog., 61 (1965) 81. [ 41W. Chu, Hydrogen purifiers: reliable technology in demanding times, Am. Lab., (February 1989) 22. [ 51 J.R. Partington, General and Inorganic Chemistry, MacMillan, London, 1954. [ 61 A.S. Darling, Hydrogen separation by diffusion through palladium alloy metals, in J.M. Pirie (Ed.), Proc. Symp. Less Common Means of Separation, April 1963, Inst. Chem. Eng., London, 1963, p. 103. [7]R.A. Outlaw, W.K. Peregoy and G.B. Hoflund, Permeation of oxygen through high purity, large grain silver, NASA Tech. Paper 2755 (1987). [8] J.W. Suitor, Zirconia cell oxygen source, in V.P. Kothari and J.R. Longanbach (Eds.), Proc. 8th Ann. Gasification and Gas Stream Cleanup Systems Contractors Review Meeting, May 1988, METC-88/6092, Vol. 1, Department of Energy, Morgantown, WV. [9]Y. Teraoka, H.-M. Zhang, S. Furukawa and N. Yamazoe, Oxygen permeation through perovskite-type oxides, Chem. Lett., (1985) 1743. [ lO]Y. Teraoka, T. Nobunaga and N. Yamazoe, Effect of cation substitution on the oxygen semipermeability of perovskite-type oxides, Chem. Lett., (1988) 503. [ 111 J.B. Hunter, Ultrapure hydrocarbon by diffusion through palladium alloys, paper presented at the Symposium on the Production of Hydrogen, ACS Fall Meeting, New York, NY, 1963. [ 121 G.L. Holleck, Diffusion and solubility of hydrogen in palladium and palladium-silver alloys, J. Phys. Chem., 74 (1970) 503.

311 [ 131 A.G. Knapton, Palladium alloys for hydrogen diffusion membranes, Platinum Met. Rev., 21 (1977) 44. [ 14]D. Fort, J.P.G. Farr and I.R. Harris, Comparison of palladium-silver and palladium-yttrium alloys as hydrogen separation membranes, J. Less-Common Met., 39 (1975) 295. [15]R.W. Baker, J. Louie, P.H. Pfromm and J.G. Wijmans, Ultrathin composite metal membranes, US Pat. 4,857,080 (August 1989). [ 161 P.V. Mercea, D. Silipas and V. Mecca, Separation of a gas mixture through a polymer membrane metallized with palladium, Gas Sep. Purif., 4 (1990) 137. [ 171 S. Uemiya, N. Sato, H. Ando, Y. Kude, T. Matsuda and E. Kikuchi, Separation of hydrogen through palladium thin-film supported on a porous glass tube, J. Membrane Sci., 56 (1991) 303. [ 181 S. Uemiya, T. Matsuda and E. Kikuchi, Hydrogen permeable palladium-silver alloy membrane supported on porous ceramics, J. Membrane Sci., 56 (1991) 315. [ 19lD.J. Edlund, D.T. Friesen, J.W. Brooke and E.D. LaChapelle, Separation of hydrogen at high temperature with a membrane based process, paper presented at the Fourth National Meeting of the North American Membrane Society, San Diego, CA, May 2931,199l. [ 201 G.K. Wehner and G.S. Anderson, The nature of physical sputtering in L.I. Meissel and R. Glang (Eds. ), Handbook of Thin Film Technology, McGraw Hill, New York, 1970. [21]J.E. Greene and C.E. Wickersham, Structural and electrical characteristics of InSb thin films grown by rf sputtering, J. Appl. Phys., 47 (1976) 3630. [ 221 J. Chabot, J. LeComte, C. Grumet and J. Sannier, Fuel clean-up system: poisoning of palladium-silver membranes by gaseous impurities, Fusion Technol., 14 (1988) 614.