Nonlocal density functional theory (DFT) calculations are used to examine alternative mechanisms for the hydrogenolysis of acetic acid to ethanol over Pd. The overall surface reaction energies, at low surface coverage, are computed for a number of different possible paths by which acetic acid may be converted to ethanol over Pd(111). Binding energies of the various oxygenated C2 intermediates formed along these paths are also reported. The overall reaction energies were used to propose a plausible mechanism for acetic acid hydrogenolysis. In the postulated mechanism, acetic acid dissociates to form an acetyl surface intermediate. The acetyl intermediate is then subsequently hydrogenated to ethanol via the formation of an acetaldehyde surface intermediate. Detailed reaction coordinate calculations were used to isolate the transition states and calculate activation barriers for acetic acid dissociation to acetyl (ΔEact = +142 kJ/mol) and acetyl hydrogenation to acetaldehyde (ΔEact = +66 kJ/mol) over Pd(111). Experimental observations and DFT calculations suggest that these two steps are likely to be rate determining in acetic acid hydrogenolysis. The barriers and overall reaction energies for these same steps are also computed on Re(0001) and pseudomorphic overlayers of Pd on Re (PdML/Re(0001)) as well. The results suggest that the C-OH bond-dissociation reaction is more favored over Re(0001) since it has a more open d band. However, bond-association reactions such as acetyl hydrogenation are more favored over PdML/Re(0001), which has an electronic d-band structure similar to that of a noble metal. The optimal balance may require a Pd/Re alloy. Calculations performed over a Pd0.66Re0.33 alloy demonstrate a nominal barrier for both C-OH bond breaking and C-H bond formation. This may be ideal for acetic acid hydrogenolysis to ethanol. Rhenium ensembles, however, should be avoided as they lead to acetic acid decarboxylation.
Bibliographical noteFunding Information:
We are grateful to the DuPont Chemical Company and the National Science Foundation (CTS-9702762) for their partial support of this work. We also thank the National Computational Science Alliance for a portion of the computational resources necessary to carry out this work. We are also grateful to Professor Jurgen Hafner and Dr. Georg Kresse of the Universitaet Wien for the use of their code VASP.