Prediction of Adsorption Energies for Chemical Species on Metal Catalyst Surfaces Using Machine Learning

Abstract

Computational catalyst screening has the potential to significantly accelerate heterogeneous catalyst discovery. Typically, this involves developing microkinetic reactor models that are based on parameters obtained from density functional theory and transition-state theory. To reduce the large computational cost involved in computing various adsorption and transition-state energies of all possible surface states on a large number of catalyst models, linear scaling relations for surface intermediates and transition states have been developed that only depend on a few, typically one or two descriptors, such as the carbon atom adsorption energy. As a result, only the descriptor values have to be computed for various active site models to generate volcano curves in activity or selectivity. Unfortunately, for more complex chemistries the predictability of linear scaling relations is unknown. Also, the selection of descriptors is essentially a trial and error process. Here, using a database of adsorption energies of the surface species involved in the decarboxylation and decarbonylation of propionic acid over eight monometalic transition-metal catalyst surfaces (Ni, Pt, Pd, Ru, Rh, Re, Cu, Ag), we tested if nonlinear machine learning (ML) models can outperform the linear scaling relations in prediction accuracy when predicting the adsorption energy for various species on a metal surface based on data from the rest of the metal surfaces. We found linear scaling relations to hold well for predictions across metals with a mean-absolute error of 0.12 eV, and ML methods being unable to outperform linear scaling relations when the training dataset contains a complete set of energies for all of the species on various metal surfaces. Only when the training dataset is incomplete, namely, contains a random subset of species’ energies for each metal, a currently unlikely scenario for catalyst screening, do kernel-based ML models significantly outperform linear scaling relations. We also found that simple coordinate-free species descriptors, such as bond counts, achieve as good results as sophisticated coordinate-based descriptors. Finally, we propose an approach for automatic discovery of appropriate metal descriptors using principal component analysis.

Data‐Driven Materials Science: Status, Challenges, and Perspectives

“…This idea of "materials cartography" has been used to identify common features for high-temperature superconductor materials, [218] to group molecules into intuitive maps that can reveal key structure-property relations, [211] find phase transitions in complex systems, [289] or establish the key descriptors in the catalytic properties of metal surfaces. [290] The final application area is related to advancing materials modelling with automated construction of surrogate models directly from data. These surrogate models can replace the laborious fitting of semiempirical models, and if trained with highly accurate data are able to reproduce complex chemical phenomena with very low computational cost by sacrificing some of the accuracy.…”

Identifying Active Sites for CO2 Reduction on Dealloyed Gold Surfaces by Combining Machine Learning with Multiscale Simulations

“…To place our model accuracy in a more straightforward context, we compared our errors to a similar work in predicting CO adsorption energy in Thiolated Ag-alloyed Au nanoclusters 38 , which finds a much higher RMSE at ~0.17eV using over 2000 data points for training. Another work using machine learning for predicting adsorption energies of CH4 related species (CH3, CH2, CH, C, and H) on the Cu-based alloys 39 reported the best performance of RMSEs around 0.3 eV after an extra tree regression algorithm. Our model complexity (determined by feature representation and neural network structure) and data set size have the best balance, giving much smaller errors compared to previous works.…”

Machine Learning for Computational Heterogeneous Catalysis

“…Predictive models of adsorption energies of common catalytic descriptors like OH*, CO*, or NO* on multimetallic surfaces have flourished in the past years. These models can be roughly assigned to physics-based [72,73,[127][128][129] and datadriven [57,59,69,[130][131][132][133][134][135][136][137][138] approaches. While the former approaches start their predictions from pre-determined physical models, like the d-band theory, [53] bond order conservation, [139] or electrostatic interactions, [140] the latter use large sets of materials properties and try to identify and use trends that might not have a known analytical expression.…”