Marina Filip

Abstract:

The properties of excitons, or correlated electron-hole pairs, are of paramount importance to optoelectronic applications of materials. A central component of exciton physics is the electron-hole interaction, which is commonly treated as screened solely by electrons within a material. However, nuclear motion can screen this Coulomb interaction as well, with several recent studies developing model approaches for approximating the phonon screening of excitonic properties. While these model approaches tend to improve agreement with experiment, they rely on several approximations that restrict their applicability to a wide range of materials, and thus far they have neglected the effect of finite temperatures. Here, we develop a fully first-principles, parameter-free approach to compute the temperature-dependent effects of phonon screening within the ab initio [Formula: see text]-Bethe-Salpeter equation framework. We recover previously proposed models of phonon screening as well-defined limits of our general framework, and discuss their validity by comparing them against our first-principles results. We develop an efficient computational workflow and apply it to a diverse set of semiconductors, specifically AlN, CdS, GaN, MgO, and [Formula: see text]. We demonstrate under different physical scenarios how excitons may be screened by multiple polar optical or acoustic phonons, how their binding energies can exhibit strong temperature dependence, and the ultrafast timescales on which they dissociate into free electron-hole pairs.

Abstract:

Using both time-dependent density functional theory (TDDFT) and the “single-shot” GW plus Bethe-Salpeter equation (GW-BSE) approach, we compute optical band gaps and optical absorption spectra from first principles for eight common binary and ternary closed-shell metal oxides (MgO, Al2O3, CaO, TiO2, Cu2O, ZnO, BaSnO3, and BiVO4), based on the nonempirical Wannier-localization-based, optimally tuned, screened range-separated hybrid functional. Overall, we find excellent agreement between our TDDFT and GW-BSE results and experiment, with a mean absolute error smaller than 0.4 eV, including for Cu2O and ZnO that are traditionally considered to be challenging for both methods.

Abstract:

Excitons are prevalent in semiconductors and insulators, and their binding energies are critical for optoelectronic applications. The state-of-the-art method for first-principles calculations of excitons in extended systems is the ab initio GW-Bethe-Salpeter equation (BSE) approach, which can require a fine sampling of reciprocal space to accurately resolve solid-state exciton properties. Here we show, for a range of semiconductors and insulators, that the commonly employed approach of uniformly sampling the Brillouin zone can lead to underconverged exciton binding energies, as impractical grid sizes are required to achieve adequate convergence. We further show that nonuniform sampling of the Brillouin zone, focused on the region of reciprocal space where the exciton wave function resides, enables efficient rapid numerical convergence of exciton binding energies at a given level of theory. We propose a well-defined convergence procedure, which can be carried out at relatively low computational cost and which in some cases leads to a correction of previous best theoretical estimates by almost a factor of 2, qualitatively changing the predicted exciton physics. These results call for the adoption of nonuniform sampling methods for ab initio GW-BSE calculations and for revisiting previously computed values for exciton binding energies of many systems.