Abstract:Having access to accurate electron densities in chemical systems, especially for dynamical systems involving chemical reactions, ion transport, and other charge transfer processes, is crucial for numerous applications in materials chemistry. Traditional methods for computationally predicting electron density data for such systems include quantum mechanical (QM) techniques, such as density functional theory. However, poor scaling of these QM methods restricts their use to relatively small system sizes and short dynamic time scales. To overcome this limitation, we have developed a deep neural network machine learning formalism, which we call deep charge density prediction (DeepCDP), for predicting charge densities by only using atomic positions for molecules and condensed phase (periodic) systems. Our method uses the weighted smooth overlap of atomic positions to fingerprint environments on a grid-point basis and map it to electron density data generated from QM simulations. We trained models for bulk systems of copper, LiF, and silicon; for a molecular system, water; and for two-dimensional charged and uncharged systems, hydroxyl-functionalized graphane, with and without an added proton. We showed that DeepCDP achieves prediction R2 values greater than 0.99 and mean squared error values on the order of 10−5e2 Å−6 for most systems. DeepCDP scales linearly with system size, is highly parallelizable, and is capable of accurately predicting the excess charge in protonated hydroxyl-functionalized graphane. We demonstrate how DeepCDP can be used to accurately track the location of charges (protons) by computing electron densities at a few selected grid points in the materials, thus significantly reducing the computational cost. We also show that our models can be transferable, allowing prediction of electron densities for systems on which it has not been trained but that contain a subset of atomic species on which it has been trained. Our approach can be used to develop models that span different chemical systems and train them for the study of large-scale charge transport and chemical reactions.

A Transferable Machine-Learning Model of the Electron Density

Transferable Machine-Learning Model of the Electron Density

Learning electron densities in the condensed phase

Transferable and extensible machine learning derived atomic charges for modeling hybrid nanoporous materials

Machine Learning Electron Density Prediction Using Weighted Smooth Overlap of Atomic Positions

A Recipe for Charge Density Prediction

Machine learning for accuracy in density functional approximations

Self-consistent Coulomb interactions for machine learning interatomic potentials

Transferability in Machine Learning for Electronic Structure via the Molecular Orbital Basis

Efficient Sampling for Machine Learning Electron Density and Its Response in Real Space

Deep Charge: A Deep Learning Model of Electron Density from One-Shot Density Functional Theory Calculation

Equivariant graph neural networks for fast electron density estimation of molecules, liquids, and solids

Atom-Density Representations for Machine Learning

Transferability of Atom-Based Neural Networks

Efficient Machine Learning Force Field for Large-Scale Molecular Simulations of Organic Systems

A representation-independent electronic charge density database for crystalline materials

Constructing Accurate and Efficient General-Purpose Atomistic Machine Learning Model with Transferable Accuracy for Quantum Chemistry

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

Electronic excited states from physically-constrained machine learning

MLTB: Enhancing Transferability and Extensibility of Density Functional Tight Binding Theory with Multi-body Interaction Corrections