Electric Dipole Descriptor for Machine Learning Prediction of Catalyst Surface-Molecular Adsorbate Interactions

Area of Science:

• Materials Science
• Computational Chemistry
• Surface Science

Background:

• Rational catalyst design requires accurate evaluation of surface-molecular adsorbate interactions.
• Developing a descriptor that is both experimentally measurable and theoretically computable is crucial.

Purpose of the Study:

• To identify and validate a descriptor for surface-adsorbate interactions.
• To establish structure-property relationships for molecular adsorbates on metal catalysts.
• To develop a machine learning model for predicting adsorption energy and charge transfer.

Main Methods:

• Utilized first-principles calculations to generate a large dataset.
• Trained a machine learning neural network using electric dipole moment as a descriptor.
• Tested model transferability across different metal substrates (Au(111), Au(001), Ag(111)) and adsorbates (NO, CO).

Main Results:

• The electric dipole moment accurately predicts molecular adsorption energy and transferred charge.
• Machine learning model achieved quick and accurate predictions.
• Model demonstrated excellent transferability to new substrates, validating the descriptor's effectiveness.

Conclusions:

• The electric dipole moment serves as a convenient and accurate descriptor for surface-adsorbate interactions.
• Machine learning models trained with this descriptor offer an efficient approach for catalyst design.
• This work provides a pathway for accelerating the discovery of new catalysts.