Abstract:Computational modelling of materials using machine learning, ML, and historical data has become integral to materials research. The efficiency of computational modelling is strongly affected by the choice of the numerical representation for describing the composition, structure and chemical elements. Structure controls the properties, but often only the composition of a candidate material is available. Existing elemental descriptors lack direct access to structural insights such as the coordination geometry of an element. In this study, we introduce Local Environment-induced Atomic Features, LEAFs, which incorporate information about the statistically preferred local coordination geometry for atoms in crystal structure into descriptors for chemical elements, enabling the modelling of materials solely as compositions without requiring knowledge of their crystal structure. In the crystal structure, each atomic site can be described by similarity to common local structural motifs; by aggregating these features of similarity from the experimentally verified crystal structures of inorganic materials, LEAFs formulate a set of descriptors for chemical elements and compositions. The direct connection of LEAFs to the local coordination geometry enables the analysis of ML model property predictions, linking compositions to the underlying structure-property relationships. We demonstrate the versatility of LEAFs in structure-informed property predictions for compositions, mapping of chemical space in structural terms, and prioritising elemental substitutions. Based on the latter for predicting crystal structures of binary ionic compounds, LEAFs achieve the state-of-the-art accuracy of 86 per cent. These results suggest that the structurally informed description of chemical elements and compositions developed in this work can effectively guide synthetic efforts in discovering new materials.

Revealing Hidden Patterns through Chemical Intuition and Interpretable Machine Learning: A Case Study of Binary Rare-Earth Intermetallics RX

Machine-learning enabled thermodynamic model for the design of new rare-earth compounds

Deciphering Cryptic Behavior in Bimetallic Transition Metal Complexes with Machine Learning

Information-Theoretic Approach for the Discovery of Design Rules for Crystal Chemistry

Leveraging Transfer Learning and Chemical Principles toward Interpretable Materials Properties

Representations and Strategies for Transferable Machine Learning Models in Chemical Discovery

Machine-Learning-based Investigation of Atomic Packing Effects: Chemical Pressures at the Extremes of Intermetallic Complexity

Integrating Machine Learning Potential and X-ray Absorption Spectroscopy for Predicting the Chemical Speciation of Disordered Carbon Nitrides

Predicting Miscibility in Binary Compounds: A Machine Learning and Genetic Algorithm Study

Interpretable structure-property correlation in X-Ray diffraction patterns of HfZrO thin films via machine learning

On-The-Fly Machine-Learning for High-Throughput Experiments: Search for Rare-Earth-Free Permanent Magnets

Quantifying chemical short-range order in metallic alloys

Phase prediction in high entropy alloys with a rational selection of materials descriptors and machine learning models

Rational design of chemically complex metallic glasses by hybrid modeling guided machine learning

Discovering the Building Blocks of Atomic Systems using Machine Learning

Unsupervised Analysis of Optical Imaging Data for the Discovery of Reactivity Patterns in Metal Alloy

Harnessing interpretable and unsupervised machine learning to address big data from modern X-ray diffraction

Machine-Learning Based Selection and Synthesis of Candidate Metal-Insulator Transition Metal Oxides

Resolving transition metal chemical space: feature selection for machine learning and structure-property relationships

Discovering Unusual Structures from Exception Using Big Data and Machine Learning Techniques.

Learning Atoms from Crystal Structure