Inferring energy-composition relationships with Bayesian optimization enhances exploration of inorganic materials.

Area of Science:

• Materials Science
• Computational Chemistry
• Data Science

Background:

• Computational materials science accelerates discovery by exploring compositional spaces.
• Current high-throughput methods often trade accuracy for speed in energy evaluations.
• Efficient sampling is crucial for discovering novel materials.

Purpose of the Study:

• To present an alternative computational approach for materials discovery.
• To optimize the sampling of compositional spaces for accelerated material synthesis.
• To improve the probability of discovering target materials accurately and efficiently.

Main Methods:

• Developed a learning algorithm named PhaseBO.
• PhaseBO focuses on effective sampling of the compositional space.
• Optimizes material stoichiometry while ensuring accurate energy evaluations.

Main Results:

• PhaseBO accelerates the discovery of novel materials.
• The approach enhances the probability of finding target materials.
• Accuracy of energy evaluation is maintained.

Conclusions:

• PhaseBO offers an effective strategy for computational materials discovery.
• This method balances speed and accuracy in exploring material compositions.
• PhaseBO guides synthetic research towards novel material realization.