(323d) Towards the Design of Active Site Requirements for the Selective Hydrogenation of Different Functionalities

Selective hydrogenation of certain functionalities in certain molecules in mixed reactant streams are important in the chemical industry. By targeting specific functionalities unique to varying molecules, it is possible remove impurities in a stream without modifying the relevant reactants before continuing to further downstream processes. For example, selective hydrogenation of acetylene to ethylene can remove the acetylene impurity in ethylene prior to polymerization before the polymerization catalyst can be poisoned by acetylene. Currently in industry, Pd-based catalysts are used to catalyze the selective hydrogenation of acetylene, but the selectivity for ethylene is low.

Bimetallic catalysts are often used to improve selectivity performance as compared to single component catalysts. However, surface segregation of randomly distributed alloys is unavoidable in reaction conditions, which makes selective hydrogenation elusive. Intermetallics compounds can offer tunable site electronics and ensemble structure for selective hydrogenation catalysis. By tuning the composition of intermetallic compounds, it is possible to determine the site requirements needed for different functionalities. Our group successfully utilized Pd-Zn γ-brass intermetallic compounds to isolate the active site ensemble and demonstrate the site requirement of ethylene hydrogenation on such surfaces. Herein, we expanded our intermetallic materials to those available from the Materials Project to investigate the site requirement of other functionalities for selective hydrogenation.

In this talk, we will present how we use DFT and machine learning to identify key descriptors of catalytic surfaces which promote the selective hydrogenation of certain functionalities, while deterring the selective hydrogenation of others. The adsorption energies of probe molecules on varying bimetallic surfaces with different active site ensembles will be calculated and collated into a database. The databases will be used to identify descriptors, such as active site nuclearity and geometry and bimetallic surface composition, through a machine learning algorithm to determine which properties from the bimetallic surfaces influenced the adsorption of different functionalities. By comparing the differences in descriptors for different functionalities, it is possible to identify what active site ensemble is needed to selective hydrogenate a certain functionality while leaving the other functionality untouched.