Applications and high performance computing-based method developments for plasmon-adsorbate interactions

Abstract

Plasmonic nanoparticles have been shown to facilitate bond breakage under mild conditions on molecules that are traditionally difficult to activate. However, the exact mechanism through which plasmonic nanoparticles drive reactions on nearby molecules is unclear. Modeling plasmonic photocatalysis is computationally demanding, typically requiring methods beyond the workhorse ground-state density functional theory (DFT). Our efforts focus on modeling this process, developing new methods to model this process, and enabling these methods to run faster and scale better in high performance computing environments. Herein, we first present investigations of plasmon-adsorbate interactions via time-dependent DFT and the non-adiabatic Ehrenfest dynamics method, finding that the dissociation process we investigate is symmetry- and electric field-dependent. Next, we apply a method we developed in the massively parallel BerkeleyGW code, called “subspace summation,” to the problem of H₂ on gold to study the effects that individual bands produce on the collective modes of the system. Following this application, we describe our development efforts for the subspace summation method. Finally, we describe our efforts to alleviate parallel file I/O bottlenecks within the BerkeleyGW code, which enables significantly faster time-to-solution for systems requiring large amounts of compute resources.