Exploring negative thermal expansion materials with bulk framework structures and their relevant scaling...

Area of Science:

• Materials Science
• Computational Materials Science
• Machine Learning in Materials Discovery

Background:

• Experimental discovery of negative thermal expansion (NTE) materials is challenging.
• Machine learning (ML) offers a data-driven approach to accelerate materials discovery.
• Existing material databases can be leveraged for identifying novel NTE candidates.

Purpose of the Study:

• To employ a multi-step machine learning method for identifying potential NTE materials.
• To predict the coefficients of negative thermal expansion (CNTE) and operational temperature ranges.
• To establish relationships between material properties and NTE behavior for design guidance.

Main Methods:

• Utilized a multi-step machine learning approach with data augmentation and cross-validation.
• Screened materials from the Inorganic Crystal Structure Database (ICSD) and other sources.
• Employed first-principles calculations with quasi-harmonic approximation (QHA) for validation.

Main Results:

• Identified approximately 1000 candidate materials (oxides, fluorides, cyanides) with bulk framework structures.
• Predicted CNTE values and temperature ranges for the identified materials.
• Approximately 57 materials showed a 100% predicted probability for NTE.
• ML predictions showed good agreement with first-principles calculations.
• Established three universal relationships for CNTE based on average electronegativity, porosity, and temperature range.

Conclusions:

• The ML approach is effective for discovering and predicting NTE materials.
• Identified critical values and relationships can guide the design of new NTE materials.
• This work expands the library of known and potential NTE materials.