(376be) First Principles-Based Multiscale Atomistic Methods for Next-Generation Electrocatalysts

Recent dramatic developments in first-principles quantum mechanics (QM) methods have enabled accurate predictions for the properties of various systems. However, such QM methods are limited to several hundreds of atoms for dozens of picoseconds (ps) (10^−12 seconds). This capacity may grow by an order of magnitude over the next few years, but pure QM methods will remain far from meeting the requirement to inform the continuum degrees of freedom (e.g. solvent effect in electrochemical systems), which require millions of atoms (10⁶) and microseconds (10^−3 seconds) of simulations. Thus, there is an enormous gap between the scale of current QM methods and that of the real applications. In my research, I develop and validate computational tools to fill this gap while simultaneously applying them to developing new generations of materials.

A great step toward achieving this goal is the development of novel hybrid QM/molecular mechanics (QM/MM) frameworks. However, applications of the most current QM/MM approaches have been limited by lacking flexible, reactive, and polarizable force fields (FFs) with variable charge transfer (particularly at the QM/MM boundary) and accurate coupling of QM and MM with minimal computational cost.

In this talk, I present a new generation of polarizable reactive FFs (RexPoN) that I have developed based entirely on QM with no empirical information. The resulting RexPoN FF has already made significant breakthroughs in discovering the new properties of materials that have been hidden to QM and experiment. Indeed, the simulations of water system using RexPoN FF lead to the most accurate properties ever predicted by a FF (and more accurate than DFT). At 298 K, RexPoN finds an average of 2.1 Strong Hydrogen Bonds (SHB) with an average lifetime of 93 fs. Connecting these SHBs leads to a one-dimensional polymer with occasional branches to sidechains. This established the revolutionary new paradigm for the structure of water as a dynamic polydisperse branched polymer (DynPol). In addition, there has been confusion and debates about the nature of anomalies in supercooled water near 230 K over the last 40 years. Using RexPoN FF, I show the changes in the structure and properties of liquid water as it is supercooled from 298 K to 200 K. Vibrational frequency and SHB calculations show that there is a topological transformation in which water changes from one-dimensional DynPol above 230 K to a different structural network at 230 K. We find that this topological transition accounts for the anomalous properties long associated with supercooled water.

In addition, I present a novel hybrid QM/MM framework called RexPoN embedded QM (REQM) with dual boundary regions. REQM dynamically communicates forces and energies between the RexPoN FF and QM regions via a python program. This allows accurate long-range interactions and polarization effects of the FF (for the application size solvent) to be included with the QM region to study complex reactions. The results of REQM simulations for the electrochemical CO₂ reduction on the Pt and Cu surface using explicit water solvent are validated against explicit DFT-D3 calculations. This means that REQM allows physical size and time scales to be used for charge flow and polarization, while crucial catalytic reactions and rate determining steps be determined by short but accurate QM (PBE flavor of DFT with D3 Grimme dispersion forces). The results of this research to reveal the real catalytic mechanisms in electrochemical processes should help the design of novel catalysts with much high performance than the traditional ones.