12692
J. Phys. Chem. B 2008, 112, 12692–12695
Permeation Hysteresis in PdCu Membranes Lixiang Yuan, Andreas Goldbach,* and Hengyong Xu Dalian Institute of Chemical Physics, Chinese Academy of Sciences, Zhongshan Road 457, 116023 Dalian, China ReceiVed: June 4, 2008; ReVised Manuscript ReceiVed: July 31, 2008
H2 permeation hysteresis has been observed during cycling of a 3 µm thick supported PdCu membrane with ∼50 atom % Pd through the fcc/bcc (face-centered cubic/body-centered cubic) miscibility gap between 723 and 873 K. Structural investigations after annealing of membrane fragments under H2 at 823 K reveal retardation of the fcc(H) f bcc(H) transition, which is attributed to the occurrence of metastable hydrogenated fcc PdCu(H) phases. The H2 flux at 0.1 MPa H2 pressure difference in the well-annealed bcc single phase regime below 723 K can be described by JH2 ) (1.3 ( 0.2) mol · m-2 · s-1 exp[(-11.1 ( 0.6) kJ · mol-1/(RT)] and that in the fcc single phase regime above 873 K by JH2 ) (7 ( 2) mol · m-2 · s-1 exp[(-30.3 ( 2.5) kJ · mol-1/(RT)]. Introduction Supported PdCu membranes have received growing attention lately because they combine enhanced resistance to embrittlement in hydrogen1 with a remarkable tolerance to H2S2,3 while allowing a reduction of the alloy layer thickness and noble metal costs.4,5 The PdCu alloy containing 47.3 atom % Pd is of particular interest for low-temperature separation applications such as CO elimination from H2 for PEM fuel cells6 and methanol steam reforming,7 since it exhibits a relatively large H2 permeability below 673 K and stability in H2 at lower temperatures. However, this permeability drops by more than 50% if the alloy stoichiometry deviates by only 3 atom % in either direction from the optimum value.8 The narrow permeability maximum originates from an ordered body-centered cubic (bcc) phase, which forms below 873 K at Pd concentrations of approximately 35-50 atom %,9 and exhibits much higher low-temperature hydrogen diffusivity than face-centered cubic (fcc) Pd and Pd alloys.10 Intriguingly, the bcc Pd47.3Cu52.7 alloy with optimum H2 permeability borders to the miscibility gap that separates bcc phases from surrounding disordered fcc PdCu phases, because the hydrogen solubility declines rapidly with increasing Cu content.3,11 Obviously, this precarious situation needs to be well understood with regard to applications, but fundamentally interesting phenomena are also arising from it. In a previous study, we found that cycling of a borderline bcc Pd48Cu52 alloy through the fcc/bcc miscibility gap in H2 leads to stoichiomteric heterogeneity on micrometer scales.12 Here, we investigate the reversibility of this structural transformation by monitoring the H2 permeation through a similar PdCu membrane during cycling through the fcc/bcc mixed phase region. In addition, we studied the impact of the precursor phase on the structural segregation during annealing of the membrane within the miscibility gap. Experimental Section The preparation of the membrane has been reported elsewhere as one example for a novel method that enables control of the alloy composition during sequential electroless plating of metals by monitoring the gas evolution.13 The membrane consisted of * To whom correspondence should be addressed. Telephone: + 86-4118437-9229. Fax: +86-411-8469-1570. E-mail:
[email protected].
a 50 mm long PdCu layer at one end of a 370 mm long tubular ceramic microfiltration support (o.d. ) 12.2 mm, i.d. ) 8 mm, Nanjing University of Technology), with that tube end and the remainder of the tube being sealed by a glaze. The stoichiometry and thickness estimated from gas evolution were Pd54Cu46 and 3.0 µm, respectively, with the latter value agreeing reasonably with that from subsequent SEM measurements (3.5 µm).13 The gas permeation setup has been described before.14 The membrane was mounted into the separator using a graphite seal at the open cold end of the tube. Single gas fluxes were measured between 473 and 953 K at pressure differences of 0.1 MPa, feeding H2 or N2 (99.999% purity each) to the shell side. The permeate side was kept at atmospheric pressure, and no sweep gas was used. The permeated flow was measured with a bubble flowmeter. H2 fluxes were corrected for leak flow components with the help of N2 permeation data considering Knudsen diffusion and viscous flow according to wellestablished procedures.15 The membrane was broken after conclusion of the permeation study, and the thickness, composition, and phase of the alloy layer were determined by scanning electron microscopy (SEM, QUANTA 200F, FEI), energy dispersive spectroscopic analysis (EDS), and X-ray diffraction (XRD, Philips PANalytical, Cu KR ) 0.15406 nm at 40 mA and 40 kV). In addition, membrane fragments were subjected to annealing at 823 K in H2 after structural homogenization at 693 and 1073 K in the bcc and fcc single phase regions, respectively, to probe the impact of the precursor phase on the compositional and structural segregation during the bcc(H) f fcc(H) and the reverse phase transition under H2. Results After 14 h of initial annealing in H2 at 773 K, the H2 and N2 fluxes through the membrane amounted to 0.18 and 0.001 mol · m-2 · s-1, respectively. The temperature was then raised to 873 K, and the H2 flux JH2 was measured down to 473 K in 100 K decrements. Figure 1 shows JH2 as a function of dwell time at each temperature. It was stable during 24 h of annealing at 873 K, which indicated the formation of a homogeneous fcc PdCu alloy. Upon cooling to 773 K, JH2 increased markedly before it became stable at 0.16 mol · m-2 · s-1 after 75 h. At 673 K, it took 2 days for JH2 to stabilize while it was constant during the initial 24 h dwell time at both 573 and 473 K. The leak
10.1021/jp8049119 CCC: $40.75 2008 American Chemical Society Published on Web 09/11/2008
H2 Permeation Hysteresis in PdCu Membranes
J. Phys. Chem. B, Vol. 112, No. 40, 2008 12693
JH2 ) (7 ( 2) mol · m-2 · s-1 × exp[(-30.3 ( 2.5) kJ · mol-1 ⁄ (RT)]
Figure 1. H2 flux as a function of dwell time at different temperatures during cooling (open triangles) and heating (full circles).
When cooling once more, H2 fluxes deviated from that curve again below 873 K, stabilizing within 1 day but remaining below the values measured during the previous heating cycle (open squares in Figure 2). Clearly, JH2 followed different trend lines during cooling and heating between 723 and 873 K. Note also that the N2 flux had significantly increased after the excursion to 953 K, reducing the H2/N2 selectivity from initially 180 to less than 40 at 773 K at the end of the permeation experiments. This deterioration of the membrane integrity is mainly caused by defects forming at the junction between the glaze and the metal layer above 823 K as a result of the mismatch between the thermal expansion coefficients of both materials.13 XRD analysis of seven membrane fragments (typically 10-25 mm2) revealed both fcc and bcc single as well as mixed phase morphologies (Table 1, F1-F7). fcc compositions were determined from the XRD lattice constants afcc employing a relationship previously derived from literature data.12 Figure 3 shows literature bcc PdCu lattice constants16 as a function of alloy composition. The dashed line in Figure 3 is a linear fit yielding
abcc ) 0.2845 nm + 2.88 × 10-4 nm × xPd
Figure 2. H2 flux as a function of temperature during cooling (open triangles and squares) and heating (full circles). Solid lines are guides to the eye.
flow-corrected, stable H2 fluxes during this cooling sequence are displayed as open triangles in the Arrhenius plot in Figure 2. Between 473 and 673 K, the activated JH2 can be expressed by the following equation:
JH2 ) (1.3 ( 0.2) mol · m-2 · s-1 × exp[(-11.1 ( 0.6) kJ · mol-1 ⁄ (RT)]
(1)
After testing at 473 K, the membrane was heated up to 673 K again and in smaller temperature increments further to 953 K to assess the permeation behavior more accurately (full circles in Figure 2). JH2 deviated slightly from eq 1 above 723 K, and at 773 K it started to decline after a short dwell time. Within 16 h, it dropped from 0.22 to 0.20 mol · m-2 · s-1 but did not change further during the following 24 h (Figure 1). This value was ∼25% higher than that of the stable H2 flux obtained at 773 K during the initial cooling cycle. Upon further heating, JH2 kept declining but stabilized in general within 24 h. Above 873 K, the H2 fluxes increased again and became stable after even shorter dwell times. Between 873 and 953 K, the activated H2 flux also exhibited Arrhenius-type temperature dependence with
(2)
(3)
The XRD compositions were derived by averaging the lattice constants of the four (fcc) or five (bcc) most intense Bragg reflections resulting in the compositional errors indicated in Table 1. The overall accuracy of the XRD compositions including instrumental error (