Discovering Melting Temperature Prediction Models of Inorganic Solids by Combining Supervised and Unsupervised Learning

Vahe Gharakhanyan,Luke J. Wirth,Jose A. Garrido Torres,Ethan Eisenberg,Ting Wang,Dallas R. Trinkle,Snigdhansu Chatterjee,Alexander Urban

2024-03-06

Abstract:The melting temperature is important for materials design because of its relationship with thermal stability, synthesis, and processing conditions. Current empirical and computational melting point estimation techniques are limited in scope, computational feasibility, or interpretability. We report the development of a machine learning methodology for predicting melting temperatures of binary ionic solid materials. We evaluated different machine-learning models trained on a data set of the melting points of 476 non-metallic crystalline binary compounds, using materials embeddings constructed from elemental properties and density-functional theory calculations as model inputs. A direct supervised-learning approach yields a mean absolute error of around 180~K but suffers from low interpretability. We find that the fidelity of predictions can further be improved by introducing an additional unsupervised-learning step that first classifies the materials before the melting-point regression. Not only does this two-step model exhibit improved accuracy, but the approach also provides a level of interpretability with insights into feature importance and different types of melting that depend on the specific atomic bonding inside a material. Motivated by this finding, we used a symbolic learning approach to find interpretable physical models for the melting temperature, which recovered the best-performing features from both prior models and provided additional interpretability.

Materials Science

What problem does this paper attempt to address?

The paper aims to address the problem of predicting the melting points of inorganic solid materials. Specifically, the researchers have developed a method that combines supervised learning and unsupervised learning to improve the accuracy of melting point predictions and enhance the interpretability of the model. ### Problems the Paper Attempts to Solve: 1. **Limitations of Existing Methods**: Current melting point prediction techniques (including empirical methods and computational methods) have issues such as limited scope, insufficient computational feasibility, or poor interpretability. 2. **Lack of Data**: For many non-metallic inorganic materials, especially high melting point materials, experimental measurement data are relatively scarce, and there is a lack of comprehensive public databases. 3. **Applicability to Different Material Types**: Existing theories and techniques are mostly limited to specific categories of materials, such as metals and their alloys, while their application to non-metallic compounds is relatively limited. 4. **Improving Prediction Accuracy and Interpretability**: By introducing an unsupervised learning step to classify materials, the melting point regression model's predictive performance can be improved, and a certain level of physical mechanism explanation can be provided. ### Research Focus: - **Material Types**: The primary focus is on binary inorganic solid materials, especially those with melting points as high as 4000°C, as these materials have significant technological applications. - **Model Construction**: By combining supervised learning and unsupervised learning, using material embedding features (based on elemental properties and density functional theory calculation results) as input, a machine learning model is trained to predict melting points. - **Model Evaluation**: The model's predictive performance is evaluated, and symbolic regression methods are used to find analytical expressions for the melting points, thereby increasing the model's interpretability. In summary, this paper aims to address the challenges in predicting the melting points of inorganic solid materials through innovative machine learning methods, improving prediction accuracy and model interpretability.

Discovering Melting Temperature Prediction Models of Inorganic Solids by Combining Supervised and Unsupervised Learning

Melting temperature prediction using a graph neural network model: From ancient minerals to new materials

Melting temperature prediction via first principles and deep learning

MeltNet: Predicting alloy melting temperature by machine learning

Deciphering melting behaviors of energetic compounds using interpretable Machine learning for melt-castable applications

Structure classification and melting temperature prediction in octet AB solids via machine learning

A Generic and Automated Methodology to Simulate Melting Point

A melting temperature database and a neural network model for melting temperature prediction

Machine learning with systematic density-functional theory calculations: Application to melting temperatures of single and binary component solids

Machine-Learning Model Prediction of Ionic Liquids Melting Points

Accurate melting point prediction through autonomous physics-informed learning

Machine Learning Models for Melting Point Prediction of Ionic Liquids: CatBoost Approach

Machine Learning-Assisted Materials Design and Discovery of Low-Melting-Point Inorganic Oxides for Low-Temperature Cofired Ceramic Applications

Machine-Learning Prediction of Curie Temperature from Chemical Compositions of Ferromagnetic Materials

Predicting melting temperature of inorganic crystals via crystal graph neural network enhanced by transfer learning

Enhancing Magnetocaloric Material Discovery: A Machine Learning Approach Using an Autogenerated Database by Large Language Models

Computational Design of Low Melting Eutectics of Molten Salts: A Combined Machine Learning and Thermodynamic Modeling Approach

Accelerating ab initio melting property calculations with machine learning: Application to the high entropy alloy TaVCrW

A comprehensive investigation on the accuracy and efficiency of methods for melting temperature calculation using molecular dynamics simulations

Machine Learning Predictions of High-Curie-Temperature Materials

Predictive Modeling of High-Entropy Alloys and Amorphous Metallic Alloys Using Machine Learning