An electrochemical series for materials.

Area of Science:

• Materials Science
• Computational Chemistry
• Electrochemistry

Background:

• The standard electrochemical series, primarily based on aqueous solutions, has limitations in predicting oxidation states for solid-state inorganic materials.
• Existing models struggle to accurately represent the complex oxidation state behavior in diverse inorganic compounds.

Purpose of the Study:

• To develop a novel, machine learning-driven electrochemical series for inorganic materials.
• To create a more accurate and reliable tool for predicting oxidation states in solid-state materials.
• To enhance applications in materials discovery and electrochemistry.

Main Methods:

• Utilized machine learning to construct an electrochemical series from tens of thousands of entries in the Inorganic Crystal Structure Database.
• Developed and parameterized a physical, human-interpretable model for predicting oxidation states from material composition.
• Compared the model's accuracy against a state-of-the-art transformer-based neural network.

Main Results:

• The new machine learning-based electrochemical series shows greater consistency with oxidation states in solid-state materials than traditional aqueous series.
• The developed model accurately predicts oxidation states from composition, outperforming advanced neural network models.
• Demonstrated successful applications in structure prediction, materials discovery, and materials electrochemistry.

Conclusions:

• Machine learning offers a powerful approach to generating robust electrochemical series for inorganic materials.
• The developed model and freely available resources (website, API) facilitate advancements in materials science and electrochemistry.
• This work provides a more accurate foundation for understanding and predicting material properties.