Quantum Chemistry & DFT

QM9, ANI-1, and ML interatomic potentials — using DFT datasets to train and benchmark neural network models for molecular energetics.

Courses

3 resources

Introduction to Quantum Chemistry for Computationalists

Free LibreTexts quantum chemistry textbook covering wave functions, the Schrödinger equation, and the foundations of DFT.

LibreTexts · Free

ML Interatomic Potentials — Review

Overview of neural network potentials, equivariant architectures, and active learning for potential fitting.

Nature Computational Science 2021

Density Functional Theory — Koch & Holthausen

Accessible introduction to DFT for chemists. Covers the Hohenberg-Kohn theorems, exchange-correlation functionals, and practical DFT calculations.

Tutorials

2 resources

Official PySCF tutorial for setting up and running HF, DFT, and MP2 calculations on small molecules from Python.

pyscf.org · Free

QM9 with PyTorch Geometric

Guide to loading QM9 in PyG, selecting target properties, and building a GNN for quantum chemistry property prediction.

PyG Docs · Free

Papers

4 resources

Ramakrishnan et al. — "Quantum chemistry structures and properties of 134 kilo molecules" (2014)

Original QM9 paper describing the dataset: 134k molecules with DFT-computed electronic, thermodynamic, and geometric properties.

Scientific Data 2014

Smith et al. — "ANI-1: An extensible neural network potential with DFT accuracy" (2017)

ANI-1 dataset and neural network potential for organic molecules — 20M DFT calculations spanning conformational space.

Chemical Science 2017

Schütt et al. — "SchNet: A continuous-filter convolutional neural network for modeling quantum interactions" (2017)

Equivariant neural network for molecular energy and force prediction using continuous-filter convolutions over pairwise distances.

NeurIPS 2017 · arXiv:1706.08566

Batatia et al. — "MACE: Higher Order Equivariant Message Passing Neural Networks" (2022)

State-of-the-art ML interatomic potential combining message passing with higher-order equivariant features.

Tools

4 resources

Lightweight Python quantum chemistry package for DFT, HF, MP2, and coupled-cluster. Scriptable and integrable with ML workflows for on-the-fly labelling.

ASE (Atomic Simulation Environment)

Python library for setting up, running, and analysing atomistic simulations. Interfaces with PySCF, VASP, Quantum ESPRESSO, and ML calculators.

QM9 dataset (via PyG)

134k small organic molecules with 19 quantum chemical properties. Available directly through PyTorch Geometric.

20M+ DFT energies and forces for small organic molecules across conformational space. Primary dataset for training ML potentials.

Datasets

3 resources

134k small organic molecules with 19 DFT-computed properties at B3LYP/6-31G(2df,p) level. The foundational benchmark for neural network potentials and molecular GNNs.

ANI-1ccx Dataset

500k organic molecules with CCSD(T)/CBS-quality energies derived via transfer learning. The benchmark for high-accuracy ML potentials beyond DFT.

Isayev group · Free

3.5M molecules with B3LYP/6-31G* optimised geometries and electronic properties computed from PubChem structures. Large-scale DFT dataset for pre-training.