Hydrogen generation by photocatalytic water splitting is a promising solution to mitigate terawatt energy demand gap. Earth abundant semiconductor materials that can efficiently harness solar spectrum are preferred for this application, such as Cu2O. However, the overall solar to hydrogen efficiency of a semiconductor photocatalyst is depended on several properties such as efficiency of light absorption, charge mobility, extent of electron-hole recombination, particle size, morphology, particle geometry, etc. Among these, particle size and geometry have a significant influence on the catalytic activity and in this project, we attempt to optimize the size and geometry of Cu2O nanoparticles using Neural Network Potential (NNP) constructed using data obtained from Density Functional Theory (DFT) calculations. The advantage of employing an NNP is that the calculations from NNP are orders of magnitude faster than DFT calculations, enabling simulations of larger systems for longer timescales that are prohibitively expensive otherwise. We constructed an NNP using molecular dynamics and metadynamics trajectories of Cu2O bulk and low index surfaces (100, 110, and 111). Preliminary tests show that the NNP predicted energies are in close agreement with DFT for a metadynamics trajectory. We are expanding our study to stepped surfaces and small clusters in order to train an NNP able to predict energies of nanoparticles. Our goal is to build an NNP that can predict the energetics of Cu2O nanoparticles in the presence of adsorbates relevant in water splitting.

Optimizing the particle size and geometry of Cu2O nanoparticles for water splitting using DFT trained neural network potential

2019 Spring National Meeting

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.