Materials Science

ML for formation energy, band gaps, elastic properties, and materials discovery — from crystal structure featurisation to interatomic potentials.

Courses

3 resources

ML for Materials Science — JARVIS Tutorials

NIST's tutorial series covering DFT datasets, electronic structure, and ML interatomic potentials using the JARVIS infrastructure.

Computational Materials Science with Python — matminer Docs

Documentation and tutorial notebooks for crystal structure featurisation using Magpie descriptors, site fingerprints, and composition features.

LBL · Free notebooks

Introduction to Machine Learning for Atomistic Simulations

Overview of ML interatomic potentials: force fields, equivariant networks (NequIP, MACE), and active learning for potential fitting.

atomistic.software · Free

Tutorials

2 resources

Materials Project API Tutorial

Official Python API (mp-api) tutorial for querying crystal structures, formation energies, and band gaps from the Materials Project database.

materialsproject.org · Free

matminer Example Notebooks

Notebooks demonstrating structure featurisation and property prediction using matminer descriptors and scikit-learn models.

Papers

3 resources

Jain et al. — "Commentary: The Materials Project" (2013)

Original Materials Project paper describing the DFT workflow and database for inorganic materials.

APL Materials 2013

Ward et al. — "A general-purpose machine learning framework for predicting properties of inorganic materials" (2016)

Introduces the Magpie descriptor set and demonstrates ML for formation energy and band gap prediction.

npj Computational Materials 2016

Batatia et al. — "MACE: Higher Order Equivariant Message Passing Neural Networks for Fast and Accurate Force Fields" (2022)

State-of-the-art equivariant ML interatomic potential achieving near-DFT accuracy at force-field cost.

NeurIPS 2022 · arXiv:2206.07697

Tools

4 resources

Materials Project (mp-api)

Open database of 150,000+ inorganic materials with DFT-computed properties. Python API via pymatgen and mp-api.

Python library for crystal structure featurisation: 200+ descriptors integrating with Materials Project and AFLOW.

Open Catalyst Project (OCP)

Meta AI's large-scale dataset (1.3M+ DFT structures) and model zoo for catalyst surface ML.

Python Materials Genomics library for crystal structure analysis, DFT input/output, and phase diagram construction.

Datasets

3 resources

Materials Project

150,000+ inorganic materials with DFT-computed formation energies, band gaps, elastic constants, and magnetic properties. The standard benchmark for materials ML.

3.5M+ compounds with DFT electronic structure, thermodynamic, and mechanical properties from the AFLOW high-throughput framework.

Duke University · Free

Open Catalyst Project (OC20)

1.3M+ DFT relaxations of adsorbates on catalyst surfaces. The largest publicly available dataset for ML interatomic potentials in catalysis.

Meta AI · Free