Expert-level materials informatics covering materials databases, machine learning for property prediction, high-throughput computation, and data-driven materials discovery.
Materials Project: DFT-calculated properties for over 150,000 inorganic compounds. AFLOW: high-throughput DFT database with thermodynamic and electronic properties. OQMD: open quantum materials database, formation energies and stability. NOMAD: repository for experimental and computational materials data. Cambridge Structural Database: crystal structures from X-ray diffraction.
Composition features: element fractions, stoichiometry-based descriptors. Structure features: radial distribution function, coordination environment. Electronic structure: band gap, DOS features from DFT calculations. SOAP: smooth overlap of atomic positions, invariant descriptor for local environments. Graph neural networks: represent crystal structure as graph, learn from topology.
CGCNN: crystal graph convolutional neural network, predict formation energy. MEGNet: graph network for molecules and crystals, multi-property prediction. Gaussian process: uncertainty quantification, active learning integration. Transfer learning: pre-train on large DFT dataset, fine-tune on small experimental.
VASP and Quantum ESPRESSO: DFT codes for electronic structure calculations.