Machine learning unifies the modeling of materials and molecules.

Area of Science:

• Computational chemistry
• Materials science
• Machine learning

Background:

• Atomistic modeling is crucial for understanding chemical and material properties.
• Predicting molecular and condensed phase stability is a key challenge.

Purpose of the Study:

• To develop a unified machine-learning framework for predicting atomic-scale properties.
• To capture quantum mechanical effects and improve stability predictions.

Main Methods:

• Utilized a machine-learning model based on local chemical environments.
• Employed Bayesian statistical learning for property prediction.

Main Results:

• Successfully predicted silicon surface reconstructions governed by quantum mechanics.
• Achieved chemical accuracy in predicting molecular stability.
• Differentiated active and inactive protein ligands with >99% reliability.

Conclusions:

• The developed framework offers a unified and systematic approach to predicting potential energy surfaces.
• Provides new insights into the behavior of materials and molecules at the atomic scale.