Deep potentials for materials science

Abstract

To fill the gap between accurate (and expensive) ab initio calculations and efficient atomistic simulations based on empirical interatomic potentials, a new class of descriptions of atomic interactions has emerged and been widely applied; i.e., machine learning potentials (MLPs). One recently developed type of MLP is the Deep Potential (DP) method. In this review, we provide an introduction to DP methods in computational materials science. The theory underlying the DP method is presented along with a step-by-step introduction to their development and use. We also review materials applications of DPs in a wide range of materials systems. The DP Library provides a platform for the development of DPs and a database of extant DPs. We discuss the accuracy and efficiency of DPs compared with ab initio methods and empirical potentials.

Moiré Phonons in Magic-Angle Twisted Bilayer Graphene

“…In such a context, in this work we study the phonon properties of magic-angle TBG based on ab initio deep potential molecular dynamics (DPMD) method. , To be specific, a machine-learning-based reactive potential method , is adopted, which allows for an accurate multibody description for the interatomic potentials of large-scale systems such as magic-angle TBG (with ∼11 000 atoms in each moiré primitive cell). The accuracy of the calculated total energies and forces based on such DPMD method is comparable to that from first-principles calculations based on density functional theory (DFT), which has been verified in various previous studies. − Using this method, we have calculated phonon band structures and phonon density of states at the magic angle, and have systematically analyzed the phonon eigenmodes at high-symmetry points in the moiré Brillouin zone. In particular, at the moiré Γ point, we have discovered a number of soft phonon modes with frequencies ∼0.05–0.1 THz, which exhibit various intriguing vibrational patterns on the moiré length scale.…”

Towards predictive design of electrolyte solutions by accelerating ab initio simulation with neural networks

“…The underlying theoretical details of these methods are beyond the scope of this review and we refer the reader to several excellent reviews of these methods that have recently been published for more detail. [45][46][47][48][49][50][51] At a high level these methods work by dening a mapping between atomic coordinates to energies and forces (occasionally virial tensors also). This mapping contains a large number of parameters (weights and biases) that can be systematically adjusted to minimise the error on a set of training data, combined with an algorithm to systematically optimise the parameters (Backpropagation).…”

Towards predictive design of electrolyte solutions by accelerating ab initio simulation with neural networks.

“…The underlying theoretical details of these methods are beyond the scope of this review and we refer the reader to several excellent reviews of these methods that have recently been published for more detail. [45][46][47][48][49][50][51] At a high level these methods work by defining a mapping between atomic coordinates to energies and forces (occasionally virial tensors also). This mapping contains a large number of parameters (weights and biases) that can be systematically adjusted to minimise the error on a set of training data, combined with an algorithm to systematically optimise the parameters (Backpropagation).…”