Research

Developing implicit solvation and electrolyte models (Islam)

Modeling electrocatalytic processes using DFT requires a reasonably accurate description of the electrochemical interface between the catalyst surface and the electrolyte. Implicit electrolyte models provide a computationally efficient way of doing this by representing the electrolyte in terms of one or more continuum fields such as polarization density or ionic concentration. We have recently implemented an implicit electrolyte model utilizing a nonlinear description of the electrolyte response and a nonlocal description of the electrolyte cavity. The resulting VASPsol++ code is a rewrite of the original VASPsol developed by the Hennig group at University of Florida and enables the use of an implicit electrolyte model in VASP calculations. In particular the code is extremely efficient and robust, requiring only slightly more computational effort than gas phase calculations. The source code is publicly available at github.com.

We are continually developing VASPsol++, with the next steps being to implement a model based on classical (fluid) DFT and the incorporation of quantum chemical coupling between the explicit and implicit systems to account for hydrogen bonding and other charge transfer interactions.

Identifying fundamental electrochemical mechanisms (Kingsley, Ukuku)

Compared to gas phase catalysis, the molecular mechanisms by which electrocatalytic processes occur are much less understood. On the computational side, this is mostly due to the difficulty in modeling transition states involved in electrocatalytic processes. The key question that remains to be answered is whether electrocatalytic processes occur by similar surface-mediated mechanisms as gas phase catalytic processes or by entirely new electrolyte-mediated mechanisms. To answer these questions, we are examining CO2 electrolysis to identify the conditions under which electrolyte-mediated processes are prevalent. As part of this work, we have developed an approach for routinely computing kinetic activation barriers for electrocatalytic processes on metal surfaces.

Additionally, we are examining the oxygen evolution reaction (OER) on transition metal oxide surfaces to understand how the active site structure can be tuned to “optimize” the Bronsted-Evans-Polanyi relationship between the kinetic activation barrier of a step and its thermodynamic driving force. This requires the development of new computational approaches to model the electrochemical interface of semiconducting metal oxides, which are significantly more challenging to model than metal electrodes due to the presence of localized charge defects.

Exploring orbital transformations underlying catalytic reaction steps (Loaiza Orduz)

One of the ultimate goals of theoretical heterogeneous catalysis is being able to atomically design an optimal active site for a given reaction. While descriptor-based screening approaches for this are commonly used in the community, we introduce an alternative approach based on first identifying the orbital transformations underlying the key steps in a mechanism and then examining how the structure of the active site affects the energetics of these orbital transformations. We have implemented two computational approaches for doing this. The first is the quasiatomic orbital method that was originally developed in the group of Klaus Ruedenberg and implemented by us into the VASP code. This method generates a minimal atomic orbital basis set that exactly reproduces the occupied bands of a plane wave DFT calculation, allowing for standard population and energy decomposition techniques developed in the computational chemistry community to be applied to condensed matter systems. The second method we utilize is constrained-orbital density functional theory (CO-DFT) that we have developed and implemented into VASP. This method allows the user to manipulate individual orbitals in a plane wave DFT calculation in order to decompose the complex orbital transformations occurring during a reaction step into simpler elementary transformations.

Currently, we are applying these methods to study C-O bond cleavage on transition metal phosphides in order to understand the electronic origin of their special selectivity for cleaving at the more substituted carbon.

Developing low-cost electronic structure methods with DFT accuracy (Meadows, Sarder)

Although implicit electrolyte models can efficiently represent the electrochemical interface in DFT calculations, it is difficult to determine the accuracy of the computed free energies. A far more accurate treatment involves the explicit simulation of the water molecules and ions in the electrolyte at the same level as the catalyst surface, which is extremely expensive at the DFT level. To circumvent this, we are developing a linear-scaling tight-binding model that is automatically parameterized using the quasiatomic orbital Hamiltonian computed in plane wave DFT calculations. This approach utilizes far more data from plane wave DFT than just the energy, forces, and band energies, allowing for parameterization using far fewer of these expensive calculations. Linear scaling is achieved using an operator expansion form of the wave function, rather than the more typical orbital-based representation. The end goal is a method that achieves near-DFT accuracy while being able to compute the energy and forces on over 10,000 configurations per cpu-hour on a system with up to 10,000 atoms.