Predicting the phase diagram of titanium dioxide with random search and pattern recognition

Area of Science:

• Computational materials science
• Solid-state chemistry
• Crystallography

Background:

• Predicting crystal polymorph stability is crucial but computationally intensive.
• Traditional methods require extensive free-energy calculations to account for temperature effects.
• Discovering new phases often relies on prior knowledge or exhaustive searches.

Purpose of the Study:

• To develop an efficient framework for predicting phase stabilities of crystal polymorphs.
• To leverage information from random crystal structure searches.
• To accelerate the discovery and characterization of new materials.

Main Methods:

• Automated clustering, classification, and visualization of crystal structures.
• Machine learning models for estimating enthalpy and entropy.
• Application to the titanium dioxide (TiO2) system without prior phase knowledge.

Main Results:

• Identification of novel TiO2 polymorphs.
• Prediction of the TiO2 phase diagram and metastability at 1600 K.
• Validation of the framework against rigorous free-energy calculations.

Conclusions:

• The developed framework significantly enhances the prediction of crystal phase stability.
• It offers a powerful tool for materials discovery by efficiently exploring the materials space.
• This approach reduces the computational burden associated with traditional methods.