Thin palladium layer deposited on ceramic materials

202

Int. J. Surface Science and Engineering, Vol. 2, Nos. 3/4, 2008

Thin palladium layer deposited on ceramic materials: application in hydrogen transport and catalytic membrane process Stéphane Haag, Michel Burgard, Jamal Amer and Barbara Ernst* Laboratoire des Procédés de Séparation (LPS), IPHC-DSA, ULP, CNRS 7178, Ecole Européenne de Chimie, Polymères et Matériaux (ECPM), 25 rue Becquerel 67087 Strasbourg Cedex 2, France Fax: +33 390 242 747 E-mail: [email protected] E-mail: [email protected] E-mail: [email protected] E-mail: [email protected] *Corresponding author Abstract: The objective of this work was to develop a thin, highly permeable, composite inorganic membrane based on palladium supported on a porous ceramic for high temperature hydrogen separation. The palladium layer (1–2 µm thick) was deposited by electroless plating on an asymmetric tubular alumina support. High temperature hydrogen permeation tests were done showing interesting hydrogen permselectivity properties with a H2/N2 separation factor at 60 (400°C) with a trans-membrane pressure difference of 1 bar. The transport of the hydrogen through the Pd membrane is mixed, solution-diffusion through the metal bulk and surface diffusion/Knudsen diffusion through the pores/defects of the film. As predicted, the membrane reactor for the dry reforming of methane shifts the equilibrium of the reaction in the direction of a higher hydrogen production. An enhancement of the methane conversion of 18% was observed in the membrane reactor configuration due to the selective removal of the hydrogen during the reaction as well as a limitation of side reactions. Keywords: inorganic membrane; palladium; hydrogen separation; membrane reactor; methane reforming; thin layer; electroless plating. Reference to this paper should be made as follows: Haag, S., Burgard, M., Amer, J. and Ernst, B. (2008) ‘Thin palladium layer deposited on ceramic materials: application in hydrogen transport and catalytic membrane process’, Int. J. Surface Science and Engineering, Vol. 2, Nos. 3/4, pp.202–221. Biographical notes: Stéphane Haag, a young French researcher, obtained his PhD in Chemistry in 2003 (Louis Pasteur University, France). His doctoral and postdoctoral research (Dechema e.V. Frankfurt/Germany, IRCE Lyon/France) is mainly focused on Chemical Engineering, membrane technology and heterogeneous catalysis. The research fields of Michel Burgard, Professor in Chemical Engineering, are: Selective transport through liquid membranes using macrocycles as carriers (bulk liquid membranes and supported liquid membranes). Application of metal Copyright © 2008 Inderscience Enterprises Ltd.

Thin palladium layer deposited on ceramic materials

203

ion separations. Use of membrane contactors (hollow fibres) in liquid extraction processes involving metal ions. Mechanism, selectivity and mass transfer modelling. Catalytic membrane reactors for gas separation and improvement of reaction yields. The research field of Barbara Ernst, Associate Professor in Chemical Engineering (1998), Louis Pasteur University (Strasbourg), France, is Membrane Science. With a formation in Material Science and Heterogeneous Catalysis (PhD (1997) in Fischer–Tropsch synthesis), the development of inorganic membranes applied for high temperature and catalytic processes is the heart of her activity.

1

Introduction

Most of the studies involving metallic membranes for high-temperature hydrogen separation deal with dense palladium or palladium alloy membranes (Pd–Ag (Uemiya et al., 1991a, 1991b; Shu et al., 1993; Keuler and Lorenzen, 2002; Hou and Hughes, 2003; Dittmeyer et al., 2001), Pd–Cu (Nam and Lee, 2001; Roa et al., 2003)). To obtain a high and selective flux of hydrogen, asymmetric membranes, composed of a metallic layer deposited on a support, have been developed. The support is generally the element providing high-thermal stability and mechanical strength of the membrane. For gas separation using palladium membranes, supports such as Vycor glass (Yeung et al., 1995), ceramics (alumina (Collins and Way, 1993) or titania (Wu et al., 2000)), stainless steel (Lin and Rei, 2001) or silver disks (Govind and Atnoor, 1991) have been used. Ceramics, in general, provide the best thermal stability, combined with good mechanical strength under different transmembrane pressures. To deposit palladium on a porous substrate, several techniques such as sputter deposition, Chemical Vapor Deposition (CVD), electroplating or electroless plating (ELP) are used. The most common deposition methods CVD and ELP have been used. CVD allows control of the thickness of the metallic layer and leads to the formation of uniform films (Nam and Lee, 2001; Kikuchi and Chen, 1997; Kajiwara et al., 2000; Xomeriatis and Lin, 1996), but it necessitates a specific experimental device and products of high purity. Because of the tubular geometry of our specific support, the electroless plating technique was chosen as the most suitable; it consists of an autocatalytic reduction of a metastable metallic complex (or salt) on the target surface (Lowenheim, 1974). A characteristic that may vary with the nature of a metal film is the mechanism of hydrogen permeation. Usually, through a dense metal membrane, the mechanism is solution-diffusion, separation resulting from differences in both solubility and diffusivity. Dense membranes are generally 6–50 µm thick, very selective with very few defects, but they show low hydrogen fluxes. The thin Pd membrane investigated in this work was intended to be both selective and to show high hydrogen permeance at high temperatures. The interest in developing such metallic membranes lies in their potential applications in catalysis. In the design of a membrane reactor for the dry reforming of methane reaction, the anticipated function of the Pd/ceramic membrane was that of a selective extractor allowing the shift of the equilibrium and inducing a higher yield when compared with a conventional packed bed reactor. Most studies of membrane reactors for methane reforming have focused on the steam reforming reaction (Kikuchi et al., 2000;

204

S. Haag et al.

Shu et al., 1994; Hou et al., 2000; Laegsgaard et al., 1995; Marigliano et al., 2003) (CH4 + H2O ↔ CO + 3H2; ∆H0298K = 206 kJ.mol−1). Although the steam reforming of methane is a very endothermic reaction, it is the most common way to produce syngas (a mixture of hydrogen and carbon monoxide, which can be used as a raw material in many chemical processes). Recently, interest has been shown in the utilisation of hydrogen permselective membrane reactors for the dry reforming of methane (Raybold and Huff, 2002; Munera et al., 2003; Galuszka et al., 1998; Kikuchi and Chen, 1997; Ferreira-Aparicio et al., 2002; Prabhu and Oyama, 2000; Ponelis and van Zyl, 1997) (CH4 + CO2 ↔ 2CO + 2H2; ∆H0298K = 247 kJ.mol−1). Use of this highly endothermic reaction is still limited to small units for the exploitation of alternative methane sources (biomass) (Rapagna et al., 2002; Courson et al., 2002) or for the production of acetic acid or formaldehyde (Edwards and Maitra, 1995), despite the attractive feature of the conversion of two greenhouse gases. Presently, the Pd/ceramic composite membrane, working as hydrogen extractor, was used in a high-temperature membrane reactor for the dry reforming of methane. A nickel-based catalyst, active at low temperatures (400–450°C) for the reaction, was tested in both the membrane reactor and a packed bed reactor. The beneficial effects of the membrane reactor were apparent in the physico-chemical characterisation of the Pd composite membrane and its hydrogen permselectivity properties.

2

Experimental

2.1 Materials The palladium film was deposited on a porous alumina support, a MembraloxR-type membrane (T1.70) provided by Pall Exekia. The geometry of the asymmetric alumina support was tubular, composed of four concentric layers with the thickness and pore size decreasing towards the interior. Three alpha-alumina layers with thickness of, respectively, 1500 µm, 40 µm, 0.2 µm, supported a mesoporous gamma-alumina film 3–4 µm thick with an average pore size of 4.7 nm (permeametry).

2.2 Metal deposition by electroless plating The electroless plating was conducted in three steps. The first step was cleaning of the support surface. The support was sonicated in a water/ethanol solution (25% ethanol) at room temperature for 30 min. Then, it was dried for 12 h at 120°C and weighed before the activation step. The inner part of the membrane was activated by the deposition of palladium nuclei to initiate the autocatalytic reduction. A classical sensitisation step with tin chloride was followed by an activation step with a tetraammine palladium nitrate, Sn(II) reducing the Pd(II) to metallic Pd (Feldstein and Weiner, 1972). At the end of the treatment, the support was dried for 12 h at 120°C and weighed before the ELP step. The key step was the deposition of the metal on the activated support. The outer part of the membrane was covered by Teflon and put in the ELP bath. The composition of the bath and the experimental conditions are given in Table 1. The deposition time varied from 1 h to 4 h and the temperature of the ELP bath was 77°C, as suggested by


205

Keuler et al. (1997). After the deposition, the membrane was rinsed with demineralised water, dried 12 h at 120°C and weighed again. Table 1

Composition of the electroless plating bath

Compounds

Concentration

Pd(NH3)4(NO3)2 (5% Pd in weight)

55 g/L

cPd = 0.0258 mol/L

Na2EDTA

288 g/L

nEDTA/nPd = 30

NH4OH

187 mL/L

Hydrazine

0.0186 mol/L

nHydrazine/nPd = 0.72

2.3 Characterisation of the Pd/ceramic membrane The Pd/ceramic composite membrane was characterised by X-Ray Diffraction (XRD) to determine its structure. The data were collected at room temperature on a Siemens D5000 diffractometer using CuKα radiation. The microstructure and the thickness of the palladium layer were evaluated by Scanning Electron Microscopy (SEM) on a JEOL device.

2.4 Hydrogen permselectivity The hydrogen flux and the selectivity of the membrane were evaluated for single gas and gas mixtures. The influence of several parameters (transmembrane pressure difference, sweep gas flow rate, temperature) was tested. The experimental device is shown in Figure 1. The same device was used for the catalytic tests. It contains three gas entries (N2, H2/CH4, CO2) regulated by mass flow controllers BROOKS 5850 TR, a co-current sweep gas (argon) in the external compartment, a stainless steel reactor in which a quartz tube or a membrane could be placed, an oven composed of THERMOCOAX temperature-resistant heating wire with a WEST 820 temperature regulator (the temperature measurement was done with a type K thermocouple between the reactor and the heating wire), an EDWARDS ASG 2000 pressure gauge, a KELLER numerical manometer (1–30 bar), a NUPRO pressure regulator NUPRO and an online AGILENT Technologies M200 gas microchromatograph. The flux of a gas through the membrane was established from the permeate flow rate. The selectivity of the membrane was evaluated by the calculation of the separation factor (αi/j) from the molar ratios of the species i and j in the permeate and in the feed:

αi j =

( xi / x j ) permeate ( xi / x j )feed

.

(1)

The separation factors were compared with those expected for Knudsen diffusion. In this case, the separation factor between gas i and j can be written as follows:

αi j = M j M i

with M j > M i .

(2)

For example, the H2/N2 separation factor for a classical Knudsen diffusion mechanism is 3.74.

206 Figure 1

S. Haag et al. Experimental device for hydrogen permeation and catalytic tests

2.5 Catalyst preparation and characterisation The catalyst Ni/Al2O3 was prepared by co-precipitation of the metal oxalates followed by thermal decomposition. Hydrated Ni(II) and Al(III) nitrates were combined in quantities corresponding to the final desired composition of 33 wt% Ni/Al2O3. Each nitrate was separately dissolved in ethanol and the solution heated to 70°C before being mixed. Then, under vigorous stirring, oxalic acid dihydrate was added to cause precipitation of the mixed oxalates. The mixture of the precipitate and the solvent was slowly evaporated to dryness. The prepared sample was then dried under mild vacuum at 60°C for 4 h and finally calcined at 550°C for 7 h. The catalyst after reaction was examined by high-resolution transmission microscope (Topcon 002B). The composition was determined by Energy Dispersive X-Ray Spectroscopy (EDS).

2.6 Catalytic tests The catalytic tests were done in both reactor configurations, in a packed bed reactor (without membrane/with a quartz tube) and in a membrane reactor (Figure 2). Prior to the reaction, the catalyst was reduced in situ at 500°C under 50% H2/50% N2 for 10 h for both types of reactors. The reactant flow rate was 50 mL min−1 composed of 10% CH4, 10% CO2 and 80% N2 (carrier gas), while the catalyst was placed in a fixed bed configuration with a contact time of 0.45 s−1 and a GHSV of 3000 h−1. The conversions of the reactants were calculated with the following formulae:

Thin palladium layer deposited on ceramic materials •

for the conventional reactor CH 4 (%) =

•

207

∆J CH4 J CH4 ,in

× 100 =

J CH4 ,in − J CH4 ,out J CH4 ,in

× 100

(3)

for the membrane reactor CH 4 (%) =

J CH4 ,in − J CH4 ,permeate − J CH4 ,retentate J CH4 ,in

× 100

(4)

with J CH4 : molar flux of methane. Figure 2

3

Principle of the membrane reactor

Results and discussion

3.1 Preparation and characterisation of the composite Pd/membrane As demonstrated by Keuler et al. (1997), the use of tetraammine palladium nitrate as metal precursor with hydrazine as reducing agent leads to a good deposition on porous substrates. Disodium ethylenediamine tetra-acetic acid (Na2EDTA) is commonly used as a complexant (Shu et al., 1993). Here, the molar ratio of EDTA to Pd(II) was adjusted to 30 to avoid bath decomposition phenomena and adhesion problems leading to a non-uniform metallic layer (Keuler et al., 1997). The Nernst equation for the reactions involved can be written: E(Pd 2+ /Pd) = E 0 (Pd 2+ /Pd) + (RT/2F) ln [Pd 2+ ]

(5)

with [Pd2+]: free Pd2+ concentration and E0(Pd2+/Pd) = 0.83 V E(N 2 /N 2 H 4 ) = E 0 (N 2 /N 2 H 4 ) + (RT/2F) ln ([H+]4 /[N 2 H 4 ])

(6)

with E0(N2/N2H4) = 1.16 V. Clearly, hydrazine becomes a more effective reductant as the pH rises. Given the strong complexation of Pd(II) by EDTA (log KEDTA = 18.5 at 25°C) and the use of a large excess of EDTA, equation (5) can be simplified to: E(Pd 2+ /Pd) = 0.2831 + (RT/2F) ln ([Pd 2+ ]0 /[EDTA])

(7)

with [Pd2+]0: Initial Pd2+ concentration and the optimisation of plating conditions consists of maximising the difference E(Pd2+/Pd)−E(N2/N2H4). As expected, the gain in weight became more important when the temperature increased while the deposition seemed to be optimal after 4 h in the ELP bath (Figure 3).

208

S. Haag et al.

A macroscopic observation of the membrane after ELP showed that the deposition led to the formation of a grey metallic layer covering the whole alumina support (Figure 4). Metallic palladium (JCPDS: 46–1043) was detected by XRD after ELP (Figure 5). The particle size of palladium, estimated using the Debye Scherrer formula, was 23 nm after 1 h of ELP treatment, 35 nm and 36 nm after 3 and 4 h, respectively. The values of the grain size could be underestimated both because of instrumental line broadening effects and because of the additional contribution to line broadening of microstrains existing in such electroless deposited thin films. By SEM, the presence of a palladium layer was confirmed. This layer seemed to cover the whole alumina surface (Figure 6(a)), but its surface was not uniform (Figure 6(b)). The film thickness was estimated to be 1–2 µm thick (Figure 6(c)). The metal was deposited on the γ-alumina surface without alteration of the mesoporous layer. Moreover, the membrane microstructure was quite compact (Figure 6(d)), although some sub-surface defects were noticed in Figure 6(c). EDS analysis gave mass percentages of 97% Pd and 3% Sn. To stabilise the metallic layer, a heat treatment at 400°C (heating rate: 1°C/min, treatment time: 12 h) under a reducing gas mixture (5% H2/95% He) was done. This treatment was expected to lead to a reorganisation of the Pd surface by migration of the metallic atoms and to avoid a sintering of the Pd particles, which composes the layer, during the permeation measurement under pure hydrogen. No major changes in the structure of the membrane were seen (Figure 5), but the relative intensity of the Pd peaks to the α-alumina peaks is higher after the heat treatment. Moreover, the microstructure seemed to be more compact (Figure 7). Because of its better uniformity, the membrane made with a 4 h deposition time was chosen for the high-temperature hydrogen permeation tests. Figure 3

Gain in weight after electroless plating as a function of the deposition time

Figure 4

(a) Macroscopic observation of the Pd/ceramic composite membrane after activation treatment and (b) after electroless plating (see online version for colours)

Thin palladium layer deposited on ceramic materials Figure 5

XRD patterns of the Pd/ceramic membrane as a function of the deposition time and after heat treatment at 400°C

Figure 6

Scanning electron micrographs of the composite Pd/ceramic membrane obtained after ELP

209

210

S. Haag et al.

Figure 7

Scanning electron micrograph of the composite Pd/ceramic membrane after heat treatment at 400°C

3.2 Hydrogen permselectivity properties of the Pd/ceramic membrane Measurement of the hydrogen flux through the Pd/ceramic membrane enabled determination of its density and the detection of any defects or pores. The permeability highly depends on the nature of the membrane, porous or dense; it defines the ability of the permeate to go through the membrane, generally affected by temperature and pressure effects. The permeability is deduced from the permeate flux experimentally measured. The flux of a gas i through a membrane is given at the steady state by Fick’s law: Ji =

Qi n n A( Pi ,feed − Pi ,permeate ) l

(8)

with Ji: flux of gas i (mol.s−1) Qi: l: A: Pi,feed, Pi,permeate: n:

Permeability of the gas i (mol.m.m−2.s−1.Pa−n) Membrane thickness (m) Membrane surface (m2) Partial pressure of the gas i in the feed and permeate compartment Exponent correlated to the diffusion mechanism of the gaseous species through the membrane.

For a solution-diffusion mechanism, the transport of hydrogen through a metallic membrane is considered to involve adsorption and dissociation of the hydrogen at the membrane surface, dissolution of the atomic hydrogen at the membrane surface, diffusion of the atomic hydrogen through the metal bulk and desorption and recombination of the atomic hydrogen on the other side of the membrane. If the rate-limiting step is the diffusion of the atomic hydrogen through the bulk metal, the flux is then proportional to the square root of the partial pressures (n = 0.5), also known as Sievert’s law (Hurlbert and Konecny, 1961). At room temperature and atmospheric pressure in the internal compartment (feed) and without sweep gas in the external compartment (permeate), with a very low transmembrane pressure difference (∆P = 5 mbar), the Pd/ceramic membrane was totally impermeable to each of the gaseous species used in this work (H2, N2, CH4 and CO2) for a single gas feed flow rate of 10–100 mL min−1. When the temperature increases (20–400°C), the membrane was still impermeable to N2 as to H2 (CH4 and CO2 were not


211

tested to be sure that there was no coke formation before the catalytic tests). When a transmembrane pressure difference of 110 mbar was applied, a hydrogen flux was observed in the permeate side at 300°C (1.6 × 10−5 mol m−2 s−1) and 400°C (3.9 × 10−4 mol m−2 s−1) but no detectable nitrogen flux. The behaviour of the thin Pd/ceramic composite membrane towards H2 and N2 was studied as a function of the temperature (200, 300 and 400°C) and the transmembrane pressure difference (single gas feed flow rate: 50 mL min−1). The nitrogen flux through the membrane became more significant when the total pressure increased in the internal compartment, probably owing to the presence of small pores in the palladium film. Assuming that N2 could diffuse through the membrane only via defects and not through the Pd layer, the N2 flux was taken as an approximate measure of the H2 flux, so that an estimate of the separate components of the total H2 flow owing to the passage through defects and to Knudsen diffusion could be made. The H2 and N2 permeances are given in Table 2 as a function of the temperature. Jayaraman and Lin (1995) have given an expression for the permeance through the pores, with no interaction between the pore wall and the molecules, as a combination of Knudsen diffusion kavg and viscous flux βavg: Q / l = kavg + βavg ∆P . Table 2

(9)

Evaluation of Knudsen diffusion on H2 and N2 permeances

Temperature ∆P (mbar) (°C)

Permeance (mol.m−2.s−1.Pa−1) (standard deviation) H2

N2 −8

Permeance (Knudsen) (mol.m−2.s−1.Pa−1) H2

10

200

350–1520

1.71 × 10 (4.52 × 10−9)

6.74 × 10− (8.31 × 10−11)

2.52 × 10−9

300

240–1500

2.80 × 10−8 (6.55 × 10−10)

9.35 × 10−10 (1.77 × 1010)

3.50 × 10−9

400

110–1500

3.54 × 10−8 (8.74 × 10−10)

1.17 × 10−9 (1.81 × 1010)

4.38 × 10−9

Both coefficients are correlated to the characteristics of the membrane (porosity, tortuosity, pore diameter) and to the properties of the gaseous species (molar weight, viscosity). As a first approximation, it is supposed that the contribution of the viscous flux is negligible. The permeance of the hydrogen owing to Knudsen diffusion can be calculated by the following equation: (Q / l ) H2 = (Q / l ) N2 × M N2 / M

H2

.

(10)

Thus, it was estimated that 12% of the total hydrogen permeance was due to Knudsen diffusion (Table 2). It is clear that the 1 µm thick Pd/membrane was not completely dense and that nitrogen and hydrogen were able to diffuse through some small defects. The H2/N2 single gas separation factors for 200, 300 and 400°C are shown in Figure 8. With transmembrane pressure differences lower than 300 mbar, the H2/N2 separation factor decreased when ∆P increased despite the low hydrogen flux. When the transmembrane pressure difference increased from 300 mbar to 1500 mbar, the selectivity remained stable at 200°C (αH2/N2 = 35) while the separation factor increased from 27 to 62

212

S. Haag et al.

and from 42 to 65 at 300°C and 400°C, respectively. As the ∆P increased, permeation through defects caused lower separation factors but these defects/pores seemed also to be hydrogen selective. In the literature, the H2/N2 separation factor values are typically between 6 and 200 for metallic membranes based on palladium with a layer thickness under 2.2 µm (Table 3). The H2 fluxes found in this work are quite similar to those obtained in the literature for a transmembrane pressure difference of 1 bar. A high hydrogen selectivity (>1000) was generally obtained for film thicknesses of 10–40 µm, but this gain in selectivity was combined with a big cut in H2 permeability. The molar fluxes for each temperature are given in Figure 9 as a function of ∆P. The best correlation was found for n = 0.85. Although n should show a temperature-dependence owing to changes in hydrogen solubility (the diffusion in the metal bulk and the adsorption at the membrane surface are temperature-dependent), here, as also observed by Collins and Way (1993), it remained invariant when the temperature increased. This may be explained by a mixed transport mechanism in which H2 permeates both through the bulk metal and through defects and pores. Table 4 compares literature value of n with those for Pd membranes with different thicknesses on different supports. Our value of n, despite being high, remains typical of a very thin palladium layer. Hydrogen transport through a metallic membrane is an activated process. For a dissolution-diffusion mechanism, the apparent activation energy can be written as the addition of a hydrogen dissolution molar enthalpy and a diffusion activation energy (Shu et al., 1991). If n is temperature-independent, the relation between the hydrogen permeability Q and the temperature follows an Arrhénius law and can be written as: Q = Q0 exp( − Ea / RT )

with Eα(J mol−1): activation energy of the hydrogen permeation Q0 (mol m m−2 s−1 Pa): pre-exponential factor R (J K−1 mol−1): perfect gas constant T(K): temperature. Figure 8

Single gas H2/N2 separation factor for the Pd/ceramic composite membrane as a function of the transmembrane pressure difference

(11)

Thin palladium layer deposited on ceramic materials Table 3

213

High-temperature H2/N2 separation using thin Pd membranes Thickness Temperature (µm) (°C)

Membrane

∆P (bar)

Permeation flux (mol.m−2.s−1)

αH2/N2

−3

Pd/Al2O3-γ (tube) (this work)

1

400

1

60

Pd/Al2O3 (disk) (Zhao et al., 2000)

1

402

1

–

40*

Pd–Ag/Al2O3 (disk) (Jayaraman and Lin, 1995)

0.5

250

1

1.25 × 10−2

5.7

Pd–Cu/Al2O3 (tube) (Roa et al., 2003)

1.5

350

3.45

0.1

7.7*

Pd–Ag/Vycor (tube) (Chang, 1999)

1.2

400

2

0.1

–

Pd et Pd–Ag/Al2O3 (tube) (Dittmeyer et al., 2001)

2.2

410

– (+ sweep gas)

1.37

200

3.52 × 10

–No information. *Permselectivity. Figure 9

Table 4

Permeation fluxes of H2 as a function of the temperature and the transmembrane pressure difference

Values of the exponent n as a function of the membrane thickness

Nature of the Nature of metallic layer the support

Geometry of Thickness ∆P the support (µm) (bar)

Value of n Reference

Pd

Alumina

Tube

1

1.5

0.85

Pd–Ag

Alumina

Disk

0.5

1.4

1

This work

Pd

Alumina

Disk

1

1

1

Pd–Ag

Alumina

Tube

7.5

6

0.61

Pd

Alumina

Tube

4–8

1.2

1

Pd

Alumina

Tube

10.3

1.7

0.65

Li et al. (2000)

Pd

Alumina

Tube

17

17

0.57

Collins and Way (1993)

Pd

Stainless steel

Tube

18

3

0.50

Mardilovich et al. (2002)

Pd

Stainless steel

Tube

20–30

6

0.50

Mardilovich et al. (1998)

Jayaraman and Lin (1995) Zhao et al. (2000) Weyten et al. (2000) Yan et al. (1994)

214 Table 4

S. Haag et al. Values of the exponent n as a function of the membrane thickness (continued) Geometry of Thickness ∆P the support (µm) (bar)

Nature of the Nature of metallic layer the support

Value of n Reference

Pd

Porous glass

Tube

20

2.9

0.76

Uemiya et al. (1991)

Pd

Stainless steel

Disk

20

27

0.80

DeRosset (1960)

Pd

Alumina

Tube

15

3.5

0.65

Hoellein et al. (2001)

Stainless steel

Tube

60

5.2

0.50

Pd

Stainless steel

Disk

24

6.2

0.68

Hurlbert and Konecny (1961)

Pd

Stainless steel

Tube

70

2

0.50

Basile et al. (2001)

Pd

–

Disk

1000

30

0.62

Morreale et al. (2003)

The activation energy was estimated to be 8.8 kJ mol−1 (Figure 10), similar to values obtained elsewhere (Table 5). However, it can be seen that the activation energy decreased with decreasing film thickness. Figure 10 Evaluation of the activation energy

Table 5

Activation energies as a function of the membrane thickness

Thickness of the metallic layer (µm) 1

Value of n 0.85

Temperature Activation energy Reference (°C) Ea (kJ.mol−1) 200–400

8.8

This work

3

0.50

450

10.3

Cheng et al. (2002)

7.5

0.61

400–500

11–12

Weyten et al. (2000)

10.3

0.65

320–577

12.3

15

0.65

350–650

10

17

0.57

450–600

14.5

20

0.76

400

10

Uemiya et al. (1991)

24

0.68

350

12

Hurlbert and Konecny (1961)

25

–

250–350

10.6

Elkina and Meldon (2002)

70

0.50

350–500

29.7

Basile et al. (2001)

Li et al. (2000) Hoellein et al. (2001) Collins and Way (1993)


215

When very small external pressure difference (5 mbar) imposed with a co-current sweep gas (argon) in the external compartment is used, the hydrogen permeation fluxes through the membrane were much lower than those obtained with an applied ∆P. At room temperature, no hydrogen was detected in the permeate with a sweep gas flow rate ≤200 mL.min−1, but when the temperature increased, both gases, H2 and N2, were found on the permeate side. Nevertheless, H2 fluxes at 300°C and 400°C were ten times lower than those obtained with an applied transmembrane pressure difference of 1 bar (Figure 11). The separation factor H2/N2, which varies as a function of the sweep gas flow rate, varied from 16 to 25 at 300°C and from 12 to 16 at 400°C for single gas measurements. If a binary mixture H2/N2 was used as a feed gas, the separation factors were of the same magnitude in this configuration. αH2/N2 varied from 12 to 24 at 300°C and from 21 to 28 at 400°C, depending on the sweep gas flow rate (20–200 mL min−1). The higher the sweep gas flow rate, the higher the H2/N2 separation factor. At 500°C, the hydrogen and nitrogen flux were, respectively, 4.5 and 3 times higher than at 400°C. Figure 11 H2 permeation fluxes through the Pd/membrane as a function of the temperature and the co-current sweep gas flow rate

It seems that the main diffusion mechanisms of the hydrogen through the Pd/ceramic membrane were surface diffusion through the defects/pores (n ≠ 0.5) and solution-diffusion through the bulk of the metal (n ≠ 1). It is difficult to evaluate the amount of hydrogen that came from each mechanism (solution-diffusion and surface diffusion). Hydrogen and nitrogen also permeated via a Knudsen diffusion mechanism through the defects: 12% of the hydrogen permeation in single gas measurements was estimated to be due to this process. The existence of simultaneous multiple mechanisms of diffusion for hydrogen could be explained by the irregularity of Pd layer thickness. Moreover, thin films manufactured by electroless deposition could present specific microstructure often displaying some anisotropy in their properties, induced by columnar grain growth. The possibility that diffusion along grain boundaries contributes to the overall surface diffusion mechanism could be considered. These results show that the Pd/ceramic composite membrane is hydrogen selective and has the potential to be used in a membrane reactor for the dry reforming of methane. However, at 500°C, a temperature above that used in the pre-treatment (400°C), an increased hydrogen flux coupled to a decreased selectivity indicated the possible creation of nanofissures or alteration of the palladium layer. The pressure–composition phase diagram of hydrogen and palladium system (Palczewska, 1975) shows two phases, one (α-phase) stable at low hydrogen

216

S. Haag et al.

concentrations corresponding to the formation of an H–Pd solid solution and one (Pd hydride β-phase) stable at higher H/Pd ratios where H atoms occupy the interstitial positions of the Pd lattice. Transition between these phases could be the origin of the fragility of the metallic Pd layer, but it can be avoided by working at temperatures higher than 290°C with a hydrogen pressure higher than 20 atm.

3.3 Application of the dry reforming of methane in a membrane reactor A preliminary test without catalyst has showed that the Pd/ceramic membrane was inactive for the dry methane reforming reaction (no hydrogen and carbon monoxide were detected by µ-GC) in the temperature range of 400–450°C. Reactions in the packed bed and membrane reactors were then compared if there was a beneficial effect owing to the presence of the membrane. A catalyst of high Ni content (33 wt% Ni/Al2O3) was chosen to reach the maximal methane conversion fixed by the equilibrium in a conventional packed bed reactor under our operating conditions. The first experiment was done without a sweep gas and without a small-applied transmembrane pressure difference (5 mbar) with the Ni/Al2O3 catalyst in a packed bed configuration. The results in the membrane reactor and in the packed bed reactor were very similar because reactants and products could not permeate through the membrane during the reaction (or the flux was so low that it had no effect on conversion). When an argon sweep gas was used, with a transmembrane pressure of 5 mbar, the results remained similar for this configuration. Table 6 gives the methane conversion and the composition of both sides, the permeate and the retentate, during the reaction. The membrane was hydrogen selective, but the amount of hydrogen in the permeate was too small to have an impact on the methane conversion. Table 6

Catalytic results obtained with the packed bed reactor and the palladium membrane reactor (sweep gas in the external compartment) Production (µmol s−1)

Conversion (%) T (°C)

CH4

CO2

RCH4/CO2

H2retentate

H2permeate

COretentate

COpermeate

RH2/CO

−1

Pd/ceramic membrane reactor (sweep gas Ar: 100 mL.min ) 400

11.5

17.2

0.67

0.625

0.0063

0.847

0.0007

0.81

450

20

27

0.74

1.389

0.1042

1.667

0.0140

0.88

Packed bed reactor 400

11.7

17.4

0.67

0.653

–

0.875

–

0.75

450

19.7

27.5

0.72

1.465

–

1.715

–

0.85

When an applied pressure difference of 500 mbar was imposed, without any sweep gas, the catalytic results (Table 7) were much better. The amount of hydrogen in the permeate considerably exceeded that obtained with a sweep gas and this high hydrogen permeability, associated with a hydrogen selectivity, led to a shift of the equilibrium in the direction of higher H2 production. The estimated separation factors during the reaction were 6.2, 7.4 and 13.3 at 450°C for H2/CH4, H2/CO and H2/CO2, respectively. An increase in the methane conversion of 18% was calculated, compared with what was observed in a packed bed reactor. Moreover, the RCH4/CO2 and RH2/CO ratios were very close to unity in the membrane reactor but not in the conventional reactor. The carbon


217

balance was also equilibrated (ΣCout/ΣCin = 0.98) in the membrane reactor configuration. These results indicate that the secondary reactions of the reverse water gas shift (CO2 + H2 ↔ H2O + CO) and the Boudouard equilibrium (2CO ↔ C + CO2) were limited by the preferential removal of the hydrogen during the reaction in the membrane reactor. In a packed bed reactor, with this catalyst, these two reactions were important and led to coke formation on the surface of the Ni/Al2O3 catalyst. No graphitic carbon was observed by XRD on the catalyst after the catalytic test in the membrane reactor. The EDS analysis of the surface of the Ni/Al2O3 catalyst confirmed that the amount of C was very low (Figure 12). Table 7

Catalytic results obtained with the palladium membrane reactor (transmembrane pressure difference: 500 mbar) Packed bed reactor Conversion (%)

Pd/ceramic composite membrane −1

Production (µmol.s )

Conversion (%)

Production (µmol.s−1)

T (°C)

CH4

CO2

RCH4/CO2

H2

CO

RH2/CO

CH4

CO2

RCH4/CO2

H2

CO

RH2/CO

400

11.7

17.4

0.67

0.653

0.875

0.75

11.2

15.3

0.73

0.694

0.854

0.81

450

19.7

27.5

0.72

1.465

1.715

0.85

23.3

25.5

0.91

1.583

1.618

0.97

Figure 12 Transmission electron micrograph and EDS analysis of the Ni/Al2O3 catalysts after reaction in the palladium membrane reactor

These results can be compared with those in the literature (Table 8) for the dry reforming of methane in a Pd/membrane reactor. The membranes used in these works were 10–50 µm thick, 10 or 50 times thicker than the membrane used here, but the conclusions remain similar. In all cases, an enhancement of the methane conversion was found, though little information about the permeability and the selectivity of these membranes was given. It seems that even with a relatively low selectivity, a beneficial effect can be obtained. However, Galuszka et al. (1998) concluded that in spite of a better methane conversion, a Pd membrane reactor is not suitable for the dry reforming of methane at high temperatures because of an alteration of the palladium layer at 550°C during the reaction owing to the formation of carbon filaments leading to the destruction of their membrane. In our case, the membrane was not proper at high temperatures (T ≥ 500°C), possibly owing to the embrittlement of the palladium layer in the presence of hydrogen

218

S. Haag et al.

resulting from the phase change mentioned earlier. Nevertheless, no carbon was detected by XRD (Figure 13) and SEM after the reaction. Moreover, the membrane was stable at 450°C for a few days and could be used for several tests. Table 8

Membrane Pd

Dry reforming of methane in palladium membrane reactors Thickness (µm)

Catalyst

CH4

1

Ni/Al2O3

450

23.3 (19.7)

20

Ru/Al2O3

500

49.5 (–)

Pd

Pd Pd/Ag

CO2 43.1 (–)

62.2 (–)

45.3 (–)

Pd/Al2O3

55.3 (–)

52.0 (–)

61.3 (–)

10

Pd/Al2O3

550

37.5 (17.2)

50 (disk)

Pt/Al2O3

550

40 (28)

–

Rh/La2O3

550

33.9 (26.0)

Pt/La2O3

31.3 (26.3)

Reference

25.5 (27.5) This work

Rh/Al2O3 Pt/Al2O3

Pd

Conversion (%)

Temperature (°C)

Kikuchi and Chen (1997)

52.5 (–) 51.0 (24.6) Galuszka et al. (1998) –

Raybold and Huff (2002)

41.0 (41.9) Munera 41.3 (38.2) et al. (2003)

(–): conventional packed bed reactor conversion. Figure 13 XRD pattern of the Pd/ceramic composite membrane after the catalytic tests

4

Conclusion

In this work, a very thin (1–2 µm thick) Pd/ceramic composite membrane was prepared by electroless plating. The Pd/ceramic membrane was found to be hydrogen permselective and the ideal H2/N2 separation factor (single gas) was 60 at 400°C (∆P = 1 bar). Hydrogen transport through the membrane appeared to be complex, involving a solution-diffusion mechanism through the bulk of the Pd, surface diffusion in the pores of the separative layer and Knudsen diffusion, the last contribution 12% of the total hydrogen permeance. Despite relatively low hydrogen selectivity but with good permeability, the thin Pd/ceramic membrane was used successfully in a membrane


219

reactor for the dry reforming of methane. The preferential removal of hydrogen during the reaction led to an enhancement of the methane conversion combined with a higher syngas production. Moreover, the membrane reactor allowed the elimination of the secondary reactions (reverse water gas shift and Boudouard reaction), important in the packed bed reactor configuration, and consequently had a beneficial effect by inhibiting coke formation on the catalyst and on the membrane surface.

Acknowledgements The authors express their gratitude to Pr. Jean Guille from the IPCMS (Strasbourg, France) for his professional help in SEM observations. Special thanks are due to Pr. Jack Harrowfield, from ISIS (Strasbourg, France), for his help in elaborating this paper.

References Basile, A., Partuzo, L. and Lagana, F. (2001) ‘The partial oxidation of methane to syngas on a palladium membrane reactor: simulation and experimental studies’, Catal. Today, Vol. 67, p.65. Cheng, Y.S., Pena, M.A., Fierro, J.L., Hui, D.C.W. and Yeung, K.L. (2002) ‘Performance of alumina, zeolite, palladium, Pd-alloy membranes for hydrogen separation from towngas mixture’, J. Membr. Sci., Vol. 204, p.329. Collins, J.P. and Way, J.D. (1993) ‘Preparation and characterization of a composite palladium-ceramic membrane’, Ind. Eng. Chem. Res., Vol. 32, p.3006. Courson, C., Udron, L., Swierczynski, D., Petit, C. and Kiennemann, A. (2002) ‘Hydrogen production from biomass gasification on nickel catalysts: tests for dry reforming of methane’, Catal. Today, Vol. 76, p.75. DeRosset, A.J. (1960) ‘Diffusion of hydrogen through palladium membranes’, Ind. Eng. Chem., Vol. 52, p.525. Dittmeyer, R., Hoellein, V. and Daub, K. (2001) ‘Membrane reactors for hydrogenation and dehydrogenation processes based on supported palladium’, J. Mol. Catal. A: Chemical, Vol. 173, p.135. Edwards, J.H. and Maitra, A.M. (1995) ‘The chemistry of methane reforming with carbon dioxide and its current and potential applications’, Fuel Process. Technol., Vol. 42, p.269. Elkina, I.B. and Meldon, J.H. (2002) ‘Hydrogen transport in palladium membranes’, Desalination, Vol. 147, p.445. Feldstein, N. and Weiner, J.A. (1972) ‘The effectiveness of Tin Sensitizer solutions’, Plating, p.140. Ferreira-Aparicio, P., Rodriguez-Ramos, I. and Guerrero-Ruiz, A. (2002) ‘On the applicability of membrane technology to the catalysed dry reforming of methane’, Appl. Catal. A: General, Vol. 237, p.239. Galuszka, J., Pandey, R.N. and Ahmed, S. (1998) ‘Methane conversion to syngas in a palladium membrane reactor’, Catal. Today, Vol. 46, p.83. Govind, R. and Atnoor, D. (1991) ‘Development of a composite palladium membrane for selective hydrogen separation at high temperature’, Ind. Eng. Chem. Res., Vol. 30, p.591. Hoellein, V., Thornton, M., Quicker, P. and Dittmeyer, R. (2001) ‘Preparation and characterization of palladium composite membranes for hydrogen removal in hydrocarbon dehydrogenation membrane reactors’, Catal. Today, Vol. 67, p.33.

220

S. Haag et al.

Hou, K. and Hughes, R. (2003) ‘Preparation of thin and highly stable Pd/Ag composite membranes and simulative analysis of transfer resistance for hydrogen separation’, J. Membr. Sci., Vol. 214, p.43. Hou, K., Fowles, M. and Hughes, R. (2000) ‘The effect of hydrogen removal during methane steam reforming in membrane reactors in the presence of hydrogen sulphide’, Catal. Today, Vol. 56, p.13. Hurlbert, R.C. and Konecny, J.O. (1961) ‘Diffusion of hydrogen through palladium’, J. Chem. Phys., Vol. 34, p.655. Jayaraman, V. and Lin, Y.S. (1995) ‘Synthesis and hydrogen permeation properties of ultrathin palladium-silver alloy membranes’, J. Membr. Sci., Vol. 104, p.251. Kajiwara, M., Uemiya, S., Kojima, T. and Kikuchi, E. (2000) ‘Hydrogen permeation properties through composite membranes of platinum supported on porous alumina’, Catal. Today, Vol. 56, p.65. Keuler, J.N. and Lorenzen, L. (2002) ‘Developing a heating procedure to optimise hydrogen permeance though Pd-Ag membranes of thickness less than 2.2 µm’, J. Membr. Sci., Vol. 195, p.203. Keuler, J.N., Lorenzen, L., Sanderson, R.N. and Linkov, V. (1997) ‘Optimzing palladium conversion in electroless palladium plating of alumina membranes’, Plating and Surf. Finish., Vol. 84, p.34. Kikuchi, E. and Chen, Y. (1997) ‘Low-temperature syngas formation by CO2 reforming of methane in a hydrogen-permselective membrane reactor’, Stud. Surf. Sci. and Catal., Vol. 107, p.547. Kikuchi, E., Nemoto, Y., Kajiwara, M., Uemiya, S. and Kojima, T. (2000) ‘Steam reforming of methane in membrane reactors: comparison of electroless-plating and CVC membranes and catalyst packing modes’, Catal. Today, Vol. 56, p.75. Laegsgaard, S., Hojlund Nielsen, P. and Lehrmann, P. (1995) ‘Steam reforming of methane in a membrane reactor’, Catal. Today, Vol. 25, p.303. Li, A., Liang, W. and Hughes, R. (2000) ‘Fabrication of dense palladium composite membranes for hydrogen separation’, Catal. Today, Vol. 56, p.45. Lin, Y.M. and Rei, M.H. (2001) ‘Separation of hydrogen from the gas mixture out of catalytic reformer by using supported palladium membrane’, Sep. and Purif. Technol., Vol. 25, p.87. Lowenheim, F.A. (1974) Modern Electroplating, John Wiley & Sons, New York, NY, pp.342–357. Mardilovich, I.P., Engwall, E. and Ma, Y.H. (2002) ‘Dependence of hydrogen flux on the pore size and plating surface topology of asymmetric Pd-porous stainless steel membranes’, Desalination, Vol. 144, p.85. Mardilovich, I.P., She, Y., Rei, M.H. and Ma, Y.H. (1998) ‘Defect-free palladium membranes on porous stainless-steel support’, AIChE J., Vol. 44, p.310. Marigliano, G., Barbieri, G. and Drioli, E. (2003) ‘Equilibrium conversion for a Pd-based membrane reactor – dependence on the temperature and pressure’, Chem. Eng. and Process, Vol. 42, p.231. Morreale, B.D., Ciocco, M.V., Enick, R.M., Morsi, B.I., Howard, B.H., Cugini, A.V. and Rothenberger, K.S. (2003) ‘The permeability of hydrogen in bulk palladium at elevated temperatures and pressures’, J. Membr. Sci., Vol. 212, p.87. Munera, J., Irusta, S., Cornaglia, L. and Lombardo, E. (2003) ‘CO2 reforming of methane as a sourve of hydrogen using a membrane reator’, Appl. Catal. A: General, Vol. 245, p.383. Nam, S.E. and Lee, K.H. (2001) ‘Hydrogen separation by Pd alloy composite membranes: introduction of a diffusion-barrier’, J. Membr. Sci., Vol. 192, p.177. Palczewska, W. (1975) ‘Catalytic reactivity of hydrogen on palladium and nickel hydride phases’, Adv. Catal., Vol. 24, p.248. Ponelis, A.A. and van Zyl, P. (1997) ‘CO2 reforming of methane in a membrane reactor’, Stud. Surf. Sci. and Catal., Vol. 107, p.555.


221

Prabhu, A.K. and Oyama, S.T. (2000) ‘Highly hydrogen selective ceramic membranes: application to the transformation of greenhouse gases’, J. Membr. Sci., Vol. 176, p.233. Rapagna, S., Provendier, H., Petit, C., Kiennemann, A. and Foscolo, P.U. (2002) ‘Development of catalysts suitable for hydrogen or syn-gas production from biomass gasification’, Biomass and Bioenergy, Vol. 22, p.377. Raybold, T.M. and Huff, M.C. (2002) ‘Analyzing enhancement of CO2, reforming of CH4, in Pd membrane reactors’, AIChE J., Vol. 48, p.1051. Roa, F., Way, J.D., McCormick, R.L. and Paglieri, S.N. (2003) ‘Preparation and characterization of Pd-Cu composite membranes for hydrogen separation’, Chem. Eng. J., Vol. 93, p.11. Shu, J., Grandjean, B.P.A. and Kaliaguine, S. (1994) ‘Methane steam reforming in asymmetric Pd- and Pd-Ag/porous SS membrane reactors’, Appl. Catal. A: General, Vol. 119, p.305. Shu, J., Grandjean, B.P.A., Ghali, E. and Kaliaguine, S. (1993) ‘Simultaneous deposition of Pd and Ag on porous stainless steel by electroless plating’, J. Membr. Sci., Vol. 77, p.181. Shu, J., Grandjean, B.P.A., Van Neste, A. and Kaliaguine, S. (1991) ‘Catalytic palladium-based membrane reactors. A review’, Can. J. Chem. Eng., Vol. 69, p.1036. Uemiya, S., Matsuda, T. and Kikuchi, E. (1991a) ‘Hydrogen permeable palladium-silver alloy membrane supported on porous ceramics’, J. Membr. Sci., Vol. 56, p.315. Uemiya, S., Sato, N., Ando, H. and Kikuchi, E. (1991b) ‘The water gas shift reaction assisted by a palladium membrane reactor’, Ind. Eng. Chem. Res., Vol. 30, p.585. Weyten, H., Luyten, J., Keizer, K., Willems, L. and Leysen, R. (2000) ‘Membrane performance: the key issues for dehydrogenation reactions in a catalytic membrane reactor’, Catal. Today, Vol. 56, p.3. Wu, L.Q., Xu, N. and Shi, J. (2000) ‘Preparation of a palladium composite-membrane by an improved electroless plating technique’, Ind. Eng. Chem. Res., Vol. 39, p.342. Xomeriatis, G. and Lin, Y.S. (1996) ‘Fabrication of a thin palladium membrane supported in a porous ceramic substrate by chemical vapor deposition’, J. Membr. Sci., Vol. 120, p.261. Yan, S., Maeda, H., Kusakabe, K. and Morooka, S. (1994) ‘Thin palladium membrane formed in support pores by metal-organic chemical vapor deposition method and application to hydrogen separation’, Ind. Eng. Chem. Res., Vol. 33, p.616. Yeung, K.L., Sebastian, J.M. and Varma, A. (1995) ‘Novel preparation of Pd/Vycor composite membranes’, Catal. Today, Vol. 25, p.231. Zhao, H.B. Xiong, G.X. and Baron, G.V. (2000) ‘Preparation and characterization of palladium-based composite membranes by electroless plating and magnetron sputtering’, Catal. Today, Vol. 56, p.89.