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.

Feature Optimization for Atomistic Machine Learning Yields A Data-Driven Construction of the Periodic Table of the Elements

Atom-Density Representations for Machine Learning

Distributed Representations of Atoms and Materials for Machine Learning

An Atomistic Machine Learning Package for Surface Science and Catalysis

Experiment-driven atomistic materials modeling: A case study combining X-ray photoelectron spectroscopy and machine learning potentials to infer the structure of oxygen-rich amorphous carbon

Learning Atoms from Crystal Structure

Highly transferable atomistic machine-learning potentials from curated and compact datasets across the periodic table

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

Atomic Structure Optimization with Machine-Learning Enabled Interpolation between Chemical Elements

Discovering the Building Blocks of Atomic Systems using Machine Learning

Designing a Periodic Table for Alloy Design: Harnessing Machine Learning to Navigate a Multiscale Information Space

AtomSets -- A Hierarchical Transfer Learning Framework for Small and Large Materials Datasets

Optimal design of experiments in the context of machine-learning inter-atomic potentials: improving the efficiency and transferability of kernel based methods

Combining Machine Learning Potential and Structure Prediction for Accelerated Materials Design and Discovery

Quantum machine learning using atom-in-molecule-based fragments selected on-the-fly

Representations and Strategies for Transferable Machine Learning Models in Chemical Discovery

Synergy of semiempirical models and machine learning in computational chemistry

Atom2Vec: learning atoms for materials discovery

AtomAgents: Alloy design and discovery through physics-aware multi-modal multi-agent artificial intelligence

Decomposing Chemical Space: Applications to the Machine Learning of Atomic Energies

Machine learning with persistent homology and chemical word embeddings improves prediction accuracy and interpretability in metal-organic frameworks