Developing High-performance Hydrogen Purification Membranes For High Temperature Operation Using Ternary Metal Films David Sholl*, Chen Ling, Lymarie Semidey-Flecha School of Chemical and Biomolecular Engineering Georgia Tech, Atlanta, GA *Email: david.sholl@chbe.gatech.edu Sabina Gade, J. Douglas Way Dept. of Chemical Engineering, Colorado School of Mines Kent Coulter Southwest Research Institute Financial support: DOE National Energy Technology Laboratory Chemical and Biomolecular Engineering

Biomolecular Engineering Ternary Alloys for Metal Membranes Thin metal membranes are a useful technology for H2 purification at elevated temperatures (e.g. for coal/biomass gasification) Binary alloys have well known advantages relative to pure Pd - avoid membrane embrittlement by making hydride phase less stable - some binary alloys have higher permeability than pure Pd - some binary alloys have better impurity resistance than pure Pd Ternary alloys have potential for improving upon binary alloys, but choosing appropriate alloys to test is a significant challenge Collaborative work at Georgia Tech/SWRI/CSM Selection of ternary alloy compositions using quantitative first-principles calculations of membrane permeability Fabrication and testing of freestanding ultra-thin ternary alloy membranes Chemical and Biomolecular Engineering

Biomolecular Engineering Theoretical Methods: First-Principles Calculations Density Functional Theory (DFT) is well suited for quantitative calculations of interstitial H in metals without experimental input. - Accurate prediction of interstitial binding energies1-3 - Accurate prediction of site to site diffusion barriers1-3 - Isotope effects1 and (at low T) tunneling effects4 - Free energies for solid phase hydride reactions5 DFT data must be combined with coarse-grained statistical mechanics descriptions to give information on macroscopic properties.1,4 P. Kamakoti, D.S. Sholl, J.Membr. Sci. 225, 145-154 (2003) P. Kamakoti, D.S. Sholl, Phys. Rev. B 71, 014301 (2005) P. Kamakoti, et al., Science, 2005, 307, 569 B. Bhatia, D. S. Sholl, Phys. Rev. B 72, 224302 (2005) S. Alapati et al., J. Phys. Chem. C 111 1584 (2007) Chemical and Biomolecular Engineering

Modeling Membrane Properties Using First-Principles Data Understand and predict intrinsic membrane properties Given: Alloy phase diagram Hierarchial Approach Use quantum chemical methods (Density Functional Theory) to calculate atomic level information: i.e. hydrogen binding energies, energy barriers, etc. Input for coarse grained lattice gas models to calculate long-range transport properties Predict macroscopic properties that can be directly compared with experimental data P. Kamakoti, D.S. Sholl, J.Membr. Sci. 225, 145-154 (2003) P. Kamakoti, D.S. Sholl, Phys. Rev. B 71, 014301 (2005) P. Kamakoti, et al., Science, 2005, 307, 569 Chemical and Biomolecular Engineering

DFT Data for Ternary Alloys An important challenge for screening ternary alloys is to generate DFT datasets that span the range of physically realizable sites on a moderate time scale We have developed heuristic algorithms to adapt known configurations from detailed binary alloy calculations to ternary alloys using only (relatively inexpensive) energy minimizations. We have calculated the properties of hundreds of transition states with this approach Chemical and Biomolecular Engineering

Cluster Expansions for Energies of Intersititial H Description of interstitial H must include O sites, T sites, and transition states (TS). Previous work1,2 used simple linear correlations: e.g., EO site = E1 + nNN E2 + nNNN E3 It is difficult to demonstrate the statistical validity of this approach. # of Pd atoms in surrounding shells A more general method is to use a Cluster Expansion (CE)3: 4 body +….. P. Kamakoti, D.S. Sholl, Phys. Rev. B 71, 014301 (2005) P. Kamakoti, et al., Science, 2005, 307, 569 J. M. Sanchez et al., Physica A, 1984, 128, 734 Chemical and Biomolecular Engineering

DFT + Cluster Expansion Methods DFT calculations for individual sites Least squares fit for all possible truncated cluster expansions Select best model using Leave One Out analysis 1 CE per alloy per site Binary alloys Pd96M4 21 O sites 88 T sites 92 TS Ternary alloys Pd70Cu26M4 54 O sites 108 T sites 432 TS Binary alloys Pd96M4 16,396 O site expansions 32,752 T site expansions 268,435,427 TS expansions Ternary alloys Pd70Cu26M4 1,048,555 O site expansions 8,388,584 T site expansions ≈ 67,108,837 TS expansions Best model for Pd96Ag4 O-sites includes 4 terms (inc. two 2 body and one 3 body) Mean-squared error relative to DFT data set: 0.00057 eV Previous simple correlation gave error of 0.0048 eV Chemical and Biomolecular Engineering

Validation of Cluster Expansion Models Define substitutionally random alloy (66,325 atoms in volume) 66,325 O sites Define each site with the CE model for the specific site and compare with DFT data. Chemical and Biomolecular Engineering

Experiments performed by Yoshihara et al. Solubility in Binary Alloys Semidey-Flecha and Sholl, J. Chem. Phys., in press Sievert’s constant obtained from a statistical mechanics treatment of the O site using CE models for each alloy Experimental work done by Boureau et al. on Pd90Ag10 showed these alloys to have greater solubility than pure Pd1 Experiments performed by Yoshihara et al. H in Pd at 1000 K Qexp = 0.009, our results Qcomp =0.011 H in Pd96Cu4 at 663 K Qexp = 0.02, our results Qcomp =0.03 Results are normalized with respect to pure Pd 1. Boureau, G., O.J. Kleppa, and K.C. Hong, J. Chem. Phys, 1977, 67(8), 3437. 2. Yoshihara, M., McLellan, R.B, Acta. Metall, 1983, 31, 66. Chemical and Biomolecular Engineering

Diffusion and Permeability: Binary Alloys Semidey-Flecha and Sholl, J. Chem. Phys., in press Permeability: 1 Pd90Ag10 (Exp) 2 Pd86Cu16 (Exp) Diffusion: All results are normalized with respect to pure Pd 1. Gryaznov, V., Sep. Purif. Methods, 2000, 29(2), 171. 2. McKinley, D., US Patent 3439474. 1969 Chemical and Biomolecular Engineering

Predicted H Solubility in Binary and Ternary Alloys All results are normalized with respect to pure Pd Enhanced solubility in PdCuAg relative to PdCu suggests Ag may be a useful “additive” to PdCu Chemical and Biomolecular Engineering

Predicted H Diffusivity in Binary and Ternary Alloys All results are normalized with respect to pure Pd Enhanced diffusivity in PdCuAg relative to PdCu suggests Ag may be a useful “additive” to PdCu Chemical and Biomolecular Engineering

Biomolecular Engineering Collaborative work at Georgia Tech/SWRI/CSM Selection of ternary alloy compositions using quantitative first-principles calculations of membrane permeability Fabrication and testing of freestanding ultra-thin ternary alloy membranes Chemical and Biomolecular Engineering

Fabrication of Freestanding Thin Foils Thin films of desired alloy composition sputter deposited onto Si substrate Removing film gives thin foil Typical samples to date are < 10 mm thick Chemical and Biomolecular Engineering

Testing Ternary Films – Preliminary Data ~8 micron thick foil with composition Pd80Cu17Ag4 (wt.%) Chemical and Biomolecular Engineering

Comparing Ternary and Binary Films – Preliminary Data Two foils tested under identical conditions Pd80Cu17Ag4 and Pd86Cu14 (wt.%) T (oC) H2/N2 Selectivity Permeability Ratio Binary Ternary 250 45 > 640 1.24 350 74 >1320 1.22 400 134 >1790 1.21 Adding a small amount of Ag appears to improve permeability of binary alloy, although observed increase is less dramatic than predicted by theory for similar compositions Theory suggests that permeability ratio will increase as T is lowered – this is not observed in the current experiments Chemical and Biomolecular Engineering

Surface Resistances and Surface Segregation Ling and Sholl, J. Membrane Sci., 2007 Previous DFT calculations for Pd75Cu25 (at.%) predicted surface resistances may contribute to net flux when T < 300 oC for films thinner than 10 mm Diffusion limited Diffusion + desorption Our next round of experiments will use somewhat thicker foils than our current experiments to avoid possible surface resistances Ag and Au expected to surface segregation in binary Pd alloys This effect is likely to be stronger in PdCu alloys (based on lattice strain) Surface segregation may be partially reversed in presence of H Chemical and Biomolecular Engineering

Summary and Ongoing Work We have developed a rigorous strategy for predicting H permeability through ternary alloys using first-principles calculations Results shown today: Pd-Cu-Ag and Pd-Cu-Au Calculations in progress: Pd-Cu-Ni and Pd-Cu-Pt We have used sputter deposition to fabricate freestanding alloy films of Pd, Pd-Cu, and Pd-Cu-Ag Testing of pure Pd and Pd-Cu alloys give permeabilities in reasonable agreement with prior literature Preliminary tests indicate that adding ~4 wt.% Ag improves permeability of Pd-Cu by ~20-25%. Further experimental testing with thicker foils (10-15 mm) will be used to characterize ternary foils Financial support: DOE National Energy Technology Laboratory Chemical and Biomolecular Engineering

Beyond Crystalline Films Shiqiang Hao, Mike Widom and David Sholl, Phys. Rev. Lett., submitted Amorphous films may have useful processing advantages over crystalline films (see, e.g. Ockwig and Nenoff, Chem. Rev. 107 (2007) 4078) We have developed first-principles methods to characterize H transport through amorphous films Results below compare amorphous and crystalline Fe3B amorphous crystalline Chemical and Biomolecular Engineering

Biomolecular Engineering Acknowledgements Georgia Tech: Lymarie Semidey-Flecha, Chen Ling, Shiqiang Hao Southwest Research Institute: Kent Coulter Colorado School of Mines: Abby Gade, Doug Way Financial support: DOE National Energy Technology Laboratory Chemical and Biomolecular Engineering