Insights into lithium manganese oxide–water interfaces using machine learning potentials

Abstract

Unraveling the atomistic and the electronic structure of solid–liquid interfaces is the key to the design of new materials for many important applications, from heterogeneous catalysis to battery technology. Density functional theory (DFT) calculations can, in principle, provide a reliable description of such interfaces, but the high computational costs severely restrict the accessible time and length scales. Here, we report machine learning-driven simulations of various interfaces between water and lithium manganese oxide (LixMn2O4), an important electrode material in lithium ion batteries and a catalyst for the oxygen evolution reaction. We employ a high-dimensional neural network potential to compute the energies and forces several orders of magnitude faster than DFT without loss in accuracy. In addition, a high-dimensional neural network for spin prediction is utilized to analyze the electronic structure of the manganese ions. Combining these methods, a series of interfaces is investigated by large-scale molecular dynamics. The simulations allow us to gain insights into a variety of properties, such as the dissociation of water molecules, proton transfer processes, and hydrogen bonds, as well as the geometric and electronic structure of the solid surfaces, including the manganese oxidation state distribution, Jahn–Teller distortions, and electron hopping.

Investigation of Lithium Manganese Oxides Using High-Dimensional Neural Networks

“…Despite the different values of the finite band gaps, the PBE0, HSE06 and PBE0r DOSs (Figs. 4.2 (b) to (d)) agree that the lower band of the Mn d states is similarly occupied for all Mn atoms in the system. [236] The highest band below the Fermi level is occupied only for every second Mn atom in LiMn 2 O 4 , i.e., for four of the eight Mn ions shown in Figs.…”

“…During this work, a scheme to adjust the parameter η of the angular ACSFs was developed. [4] For each element triple, a set of values η is calculated as the arithmetic mean of the sets used for the radial ACSFs of the respective element pairs,…”

“…This initialization of the weights has performed well in particular for atomic neural networks with more than two hidden layers and funnel-like architectures. [2][3][4][5] To adapt this initialization to the HDNNS method, the mean atomic spin of each element is used instead of the mean atomic energy contribution. Additionally, the atomic spin range of each element is employed instead of the atomic energy range.…”

“…[236] The highest band below the Fermi level is occupied only for every second Mn atom in LiMn 2 O 4 , i.e., for four of the eight Mn ions shown in Figs. 4.2 (b) to (d). [236] According to the PBE0r DOSs of single Mn atoms, these two groups can be assigned to Mn IV with the orbital occupation t 3 2g e 0 g and high-spin Mn III with the orbital occupation t 3 2g e 1 g .…”

“…Two main groups of MnO 6 octahedra can be identified in Figs. 4.9 (a) and (b). The atomic spins of one group are between about 1.5 to 1.7 with a mean of 1.58 .…”