Oxidation State Assignments for Metal Centers from Crystallographic Data

A machine learning model trained on chemist-assigned oxidation states encoded in chemical names from the Cambridge Crystallographic Database. The approach considers the immediate local chemical environment around a metal center and achieves a prediction accuracy exceeding 98%. This work by Kevin Maik Jablonka of École Polytechnique Fédérale de Lausanne demonstrates how collective knowledge can be harvested to create a tool for chemists.

Use Cases

• Predicting oxidation states for metal centers in complex materials based on local chemical environment.
• Identifying potentially incorrect oxidation state assignments in crystallographic databases.
• Developing chemical intuition-driven models without explicit heuristic programming.

Strengths

• Model prediction accuracy is reported to be greater than 98%.
• Method is described as robust to common experimental uncertainties like incorrect protonation.
• Derived from the authoritative Cambridge Crystallographic Database, representing collective chemist knowledge.

Limitations

• Row count, column definitions, and sample data are unknown, limiting suitability assessment.
• Last update date is unknown; freshness unverified.
• Data may reflect the biases inherent to the source database and the historical assignments by chemists.

Provenance

Source

Cambridge Crystallographic Database

Collection Method

Machine learning algorithm trained on oxidation state assignments encoded in chemical names.