(362o) Rational Catalyst Design at Atomic Scales through Physics and Machine Learning-Based Insights Towards Energy and Sustainable Applications

In the realm of heterogeneous catalysis, whether thermal or electrochemical, physical models supplemented by Machine learning (ML) models can accelerate the discovery of inexpensive and abundant catalysts with high activity, selectivity, and stability. These models come very handy since the chemical space of materials to explore can be quite large, as well as difficult and time-taking to be tested experimentally. Therefore, I propose physical insights complemented ML approaches for rational catalyst design through an understanding of a) surface restructuring and sintering and, b) segregation behaviors in High-entropy or multimetallic alloys (MMAs) to predict relative surface concentrations of the metals under experimentally relevant conditions (temperature, coverage of molecular reacting species etc.) towards heterogeneous catalysis contexts. Overall, I will show how combining physical models, high-throughput simulations, density functional theory (DFT), and machine learning (ML) can elucidate different aspects of materials discovery and catalyst design (of MMAs) in thermal and electrocatalysis for energy and sustainable applications.

High-entropy alloys (HEA), defined as near-equimolar alloys of five or more elements, have only recently been recognized, and there are several potential reactions that might be successfully accelerated by HEA catalysts due to their extensive composition space and distinctive high entropy structures. The distribution of active sites of HEA catalysts needs to be identified, understood, and then rationally designed and controlled. The successful screening of HEA catalysts can be made possible by computationally assisted design of HEA catalysts. This will be especially true in estimating the binding energy and activity of materials and choosing the best metal combinations. Additionally, computational methods can aid in decoupling the multielement synergistic effects to better understand HEA catalysts and direct additional research in this direction. Extension of physics-based alloy stability models^1–3,5 or descriptive ML models⁴ can aid in predicting the configurational energies, shapes and thereafter the binding of adsorbates on these multi metallic sites as a function of the local geometry, coordination number (CN) and the metal identities. The alloy stability model has ample ingredients for parameterization to model more than five elements and the associated parameters.

Therefore, in my talk, I illustrate the case for building an understanding of surface restructuring and sintering in heterogeneous catalytic contexts since the durability and morphologies/shapes of active metal sites under reaction conditions are areas of great significance for catalyst design. For this purpose, I build a training set of simple first-principle DFT calculations of activation energies for atomic diffusions/migrations over various metal FCC surfaces. From these, I then extract coordination-based parameters to fit to the energies of atoms at non-equilibrium distances. Such extension of our previously developed coordination-based schemes¹ breaks down any sintering process into individual atomic diffusions such that they can be fed to a kinetic Monte Carlo (kMC) approach for modeling the migration of nanoparticles on various supports through random fluctuations in the nanoparticle structure. This can be then compared with experimental sintering patterns. Since surface energies and the distribution of site morphologies are vital for optimization of catalytic activity and selectivity, using mentioned CN schemes with structural and site information, we also evaluate surface energies and predict relative stabilities of various catalytic surfaces with periodic slabs as well as various metal nanoparticle (MNP) shapes. Finally, we also propose to extend these models to multimetallic surfaces (alloys). This can lead to effective screening of thermodynamic stabilities of alloy MNPs. Finally, the extension of the surface energy and activation barrier calculations to multimetallic surfaces (alloys) are also discussed. The approach is also extended to describe segregation behaviors in multimetallic alloys (MMAs)/HEAs to predict relative surface concentrations of the constituent metals under experimentally relevant conditions (temperature, coverage of molecular reacting species etc.). As such, the resulting physical model can be applied to calculate the energetics of any nanoparticle morphology and chemical composition, thus significantly accelerating design of durable nanoalloys.

On a bigger picture, ML extracted human-interpretable models along with derived physical insights can accelerate materials discovery and design as well as foster new collaborations through the sharing of ideas via easy access to online database of theoretical and experimental physicochemical properties.

(1) Roling, L. T.; Li, L.; Abild-Pedersen, F. Configurational Energies of Nanoparticles Based on Metal–Metal Coordination. J. Phys. Chem. C 2017, 121 (41), 23002–23010. https://doi.org/10.1021/acs.jpcc.7b08438.

(2) Streibel, V.; Choksi, T. S.; Abild-Pedersen, F. Predicting Metal–Metal Interactions. I. The Influence of Strain on Nanoparticle and Metal Adlayer Stabilities. The Journal of Chemical Physics 2020, 152 (9), 94701–94701. https://doi.org/10.1063/1.5130566.

(3) Choksi, T. S.; Streibel, V.; Abild-Pedersen, F. Predicting Metal-Metal Interactions. II. Accelerating Generalized Schemes through Physical Insights. J Chem Phys 2020, 152 (9), 094702. https://doi.org/10.1063/1.5141378.

(4) Lamoureux, P. S.; Choksi, T. S.; Streibel, V.; Abild-Pedersen, F. Combining Artificial Intelligence and Physics-Based Modeling to Directly Assess Atomic Site Stabilities: From Sub-Nanometer Clusters to Extended Surfaces. Phys. Chem. Chem. Phys. 2021, 23 (38), 22022–22034. https://doi.org/10.1039/D1CP02198B.

(5) Deo, S.; Abild-Pedersen, F. Towards swift predictions of metal ad-atom diffusion barriers for enhanced understanding of sintering and catalysts durability (manuscript submitted)