Atomistic neural network representations for chemical dynamics simulations of molecular, condensed phase, and interfacial systems: Efficiency, representability, and generalization

Abstract

Machine learning techniques have been widely applied in many fields of chemistry, physics, biology, and materials science. One of the most fruitful applications is machine learning of the complicated multidimensional function of potential energy or related electronic properties from discrete quantum chemical data. In particular, substantial efforts have been dedicated to developing various atomistic neural network (AtNN) representations, which refer to a family of methods expressing the targeted physical quantity as a sum of atomic components represented by atomic NNs. This class of approaches not only fully preserves the physical symmetry of the system but also scales linearly with respect to the size of a system, enabling accurate and efficient chemical dynamics and spectroscopic simulations in complicated systems and even a number of variably sized systems across the phases. In this review, we discuss different strategies in developing highly efficient and representable AtNN potentials, and in generalizing these scalar AtNN models to learn vectorial and tensorial quantities with the correct rotational equivariance. We also review active learning algorithms to generate practical AtNN models and present selected examples of AtNN applications in gassurface systems to demonstrate their capabilities of accurately representing both molecular systems and condensed phase systems. We conclude this review by pointing out remaining challenges for the further development of more reliable, transferable, and scalable AtNN representations in more application scenarios.

Theoretical Insights into Structure Sensitivity in Formate Decomposition Dynamics on Cu Surfaces

“…The EANN PES is invariant with respect to translation, rotation, and permutation. The key advantage of this EANN method is that the density-like descriptors given in eq scale linearly with respect to the number of neighboring atoms …”

Biomass carbon mining to develop nature-inspired materials for a circular economy

“…Applying methods that have been proved successful in other areas to biobased molecules and materials is expected to revolutionize the development of nature-inspired computational materials for a circular economy. Examples are the use of active learning for free energy calculations, 217 efficient analysis of high-throughput nanopore data, 218 chemical dynamics simulations of interfacial systems, 219 or physics-informed ML models. 220 …”

Rotationally Inelastic Scattering Dynamics of NO from Ag(111): Influence of Interaction Potentials

“…The embedded atom neural network (EANN) approach , was used to construct the two PESs. As an atomistic neural network approach, the total energy of the system is regarded as the sum of the atomic energies. Each atomic energy is an output by an element-wise neural network with an input vector of embedded atom density (EAD) descriptors {ρ i }­ Specifically, each component of {ρ i } is an electron density contribution at the position of atom i provided by surrounding atoms that can effectively describe the atomic environment.…”

“…The Brillouin zone integration was sampled by a 4 × 4 × 1 58,59 was used to construct the two PESs. As an atomistic neural network approach, 60 the total energy of the system is regarded as the sum of the atomic energies. Each atomic energy is an output by an element-wise neural network with an input vector of embedded atom density (EAD) descriptors {ρ i }…”