Benchmarking doubles-corrected random-phase approximation methods for frequency dependent polarizabilities: Aromatic molecules calculated at the RPA, HRPA, RPA(D), HRPA(D), and SOPPA levels

Abstract

The performance of different polarization propagator methods, such as RPA, RPA(D), HRPA, HRPA(D), and SOPPA, have been tested against CC3 values for both static and dynamic polarizabilities. The test set consists of 14 (hetero-)aromatic medium-sized organic molecules, mostly with a high degree of symmetry. The benchmark of the methods remarkably reveals that RPA and HRPA(D) yield results comparable with the CC3 values and that they outperform SOPPA for these molecules. For a subset of the molecules, a comparison could be made to experimental values. The comparison for static polarizabilities proves that RPA and HRPA(D) as well as RPA(D) reproduce experimental values to a satisfying precision, whereas the SOPPA method compared to these three methods appears to perform only adequately. An investigation of the performance of Sadlej’s polarized triple zeta basis set against Dunning’s aug-cc-pVTZ basis set was also carried out. It is found that in contrast to other methods, Sadlej’s basis set did not perform sufficiently compared to the larger aug-cc-pVTZ basis set for the RPA based methods.

Random phase approximation for gapped systems: Role of vertex corrections and applicability of the constrained random phase approximation

“…This relation between the size of the gap and the applicability of the RPA diagrams is borne out by the comparison with experimental dielectric constants by Shishkin and Kresse [5]: Good agreement between the RPA and experiment is found in wide-gap systems (MgO, C, and LiF) and the largest disagreements occur for systems with a small gap (ZnO, GaAs, ZnS, and CdS). Good agreement has also been found between the RPA and experimental molecular polarizabilities in aromatic molecules [7] (cf. Appendix H), which also have a substantial gap.…”

“…The evaluation of the RPA screening can be done in several steps, generating an intermediate partially dressed interaction by integrating out some electronic states and then using this as the bare interaction for the next downfolding step. This multitiered approach 7 is illustrated in Fig. 5, where T 1 (all The partially screened interaction of a specific tier is calculated by including all screening processes except those that occur entirely within that tier.…”

“…Introduced by Bohm and Pines [1][2][3], it provides a self-consistent microscopic view on the Coulomb interaction between electrons. Nowadays, the approximation is used in ab initio methods to calculate energetics [4], dielectric properties [5], plasmon spectra [6], the polarizability of molecules [7] and solids, and the effective interaction strengths [8] in low-energy models of correlated matter.…”

Comparison of Optical Response from DFT Random Phase Approximation and Low-Energy Effective Model: Strained Phosphorene

“…In this case, the Kohn-Sham equations can be evaluated to obtain the dielectric response of the material. [28] In most DFT packages, such as GPAW, QUANTUM ESPRESSO, and VASP, a random phase approximation (RPA) [20][21][22] is implemented to evaluate the dielectric response, or permittivity tensor. [23,29] Unfortunately, this approximation neglects the exchange-correlation contribution and can lead to unphysical modifications to the original band structure obtained through a specific functional and its associated exchange-correlation.…”

“…In this paper, we compute each component of the permittivity tensor of phosphorene subject to in-plane strain. Two methods are used: One involves DFT combined with a random phase approximation (DFT-RPA), [20][21][22][23][24][25][26][27] and the other uses a low-energy model Hamiltonian with the Green's function current-current correlator.…”