Characterizing Termination-Dependent TiS₂/H₂O Interfaces using Deep-Neural-Network-Assisted Molecular Dynamics

TiS₂ electrodes are promising materials for water desalination devices. However, a fundamental understanding of the TiS₂ interface with liquid water is still lacking. For instance, it remains unclear how the physicochemical properties of water are affected by different surface terminations of TiS₂. This work describes a series of atomic-scale simulations of liquid water in contact with four different terminations of TiS₂: Armchair, Zigzag, Zigzag-L and Zigzag-R. The potential energy surface of these systems is described with a first-principles-based deep neural network potential (DP) trained on molecular dynamics (MD) simulations with forces from density functional theory (DFT) using the SCAN+rVV10 exchange-correlation functional. The DP provides good agreement with experimental results available on bulk TiS₂. In addition, the DP accurately reproduces the density distribution of interfacial water near TiS₂ predicted by ab initio MD. The density distribution profile of interfacial water depends on the TiS₂ surface termination exposed to water. Water is found to spontaneously dissociate only on Zigzag-L, the only surface exposing both 4-fold and 1-fold coordinated Ti (Ti_4c) and S (S_1C) atoms, respectively. The Armchair, Zigzag and Zigzag-R surfaces contain molecular water strongly bound to undercoordinated Ti atoms, but they have a different influence on the depletion region between first- and second-layer water adsorbed on TiS₂. The results reported in this work will help further design and improvement of TiS₂-based technologies for capacitive deionization.