Prediction of Negative Thermal Expansion in 2D-Materials by Multistep Machine Learning Using Structural Descriptors

Area of Science:

• Materials Science
• Condensed Matter Physics
• Computational Materials Science

Background:

• Discovering new negative thermal expansion (NTE) metamaterials is experimentally challenging and computationally intensive.
• Predicting thermal expansion coefficient (TEC) and NTE maxima (αmax) is crucial for materials design.

Purpose of the Study:

• To develop a machine learning (ML) approach for efficient prediction of NTE maxima and TEC in 2D materials.
• To rapidly screen a large number of 2D materials for NTE properties.

Main Methods:

• A multistep machine learning (ML) model was utilized, taking structural and tunable features as input.
• Predictions were validated against first-principles calculations using the quasi-harmonic approximation (QHA).
• Blind tests on materials from the 2DMatPedia database confirmed model robustness.

Main Results:

• The ML model demonstrated high correlation with QHA calculations for predicted target attributes.
• Out of 234 investigated 2D materials, 194 were identified as exhibiting NTE between 0-1000 K.
• Key features influencing NTE were identified, guiding future material design.

Conclusions:

• The developed ML approach offers a systematic and efficient method for screening 2D materials for NTE.
• This work accelerates the discovery and design of novel 2D NTE materials.
• Identified features provide valuable insights for tailoring material properties for specific thermal expansion behaviors.