| Abstract Scope |
Computing high-temperature materials properties is a difficult task using first principles or machine learning methods. We have demonstrated the feasibility of predicting melting temperature on a large scale and in a rapid manner using an integrated approach of density functional theory calculations and deep learning. We have identified the material with the highest melting temperature and potential refractory materials. To make our methodology accessible to the public, we have developed a cyber infrastructure that allows users to calculate melting temperatures of up to 10000 materials in a single API call with a processing speed of 0.03 seconds/material. Our models and databases are freely available to the public. We have also created deep learning models for predicting other material properties, such as bulk modulus, volume, and fusion enthalpy. Our framework, Materials Properties Prediction (MAPP), offers a diverse array of materials properties and has the potential for iterative improvement and model integration. |