Subscribe to the gold package and get unlimited access to Shamra Academy
Register a new userAccurate and numerically efficient r$^2$SCAN meta-generalized gradient approximation
71
0
0.0
(
0
)
Authors
James W. Furness
Ask ChatGPT about the research
No Arabic abstract
The recently proposed rSCAN functional [J. Chem. Phys. 150, 161101 (2019)] is a regularized form of the SCAN functional [Phys. Rev. Lett. 115, 036402 (2015)] that improves SCANs numerical performance at the expense of breaking constraints known from the exact exchange-correlation functional. We construct a new meta-generalized gradient approximation by restoring exact constraint adherence to rSCAN. The resulting functional maintains rSCANs numerical performance while restoring the transferable accuracy of SCAN.
rate research
Read More
We develop numerical methods to calculate electron dynamics in crystalline solids in real-time time-dependent density functional theory employing exchange-correlation potentials which reproduce band gap energies of dielectrics; a meta generalized gradient approximation (meta-GGA) proposed by Tran and Blaha [Phys. Rev. Lett. 102, 226401 (2009)] (TBm-BJ) and a hybrid functional proposed by Heyd, Scuseria, and Ernzerhof [J. Chem. Phys. 118, 8207 (2003)] (HSE). In time evolution calculations employing the TB-mBJ potential, we have found it necessary to adopt a predictor-corrector step for stable time-evolution. Since energy functional is not known for the TB-mBJ potential, we propose a method to evaluate electronic excitation energy without referring to the energy functional. Calculations using the HSE hybrid functional is computationally expensive due to the nonlocal Fock-like term. We develop a computational method for the operation of the Fock-like term in Fourier space, for which we employ massively parallel computers equipped with graphic processing units. To demonstrate significances of utilizing potentials providing correct band gap energies, we compare electronic excitations induced by femtosecond laser pulses using the TB-mBJ, HSE, and a simple local density approximation (LDA). At low laser intensities, electronic excitations are found to be sensitive to the band gap energy: results using TB-mBJ and HSE are close to each other, while the excitation of the LDA calculation is more intensive than the others. At high laser intensities close to a damage threshold, we have found that electronic excitation energies are similar among the three cases.
The development of semilocal models for the kinetic energy density (KED) is an important topic in density functional theory (DFT). This is especially true for subsystem DFT, where these models are necessary to construct the required non-additive embedding contributions. In particular, these models can also be efficiently employed to replace the exact KED in meta-Generalized Gradient Approximation (meta-GGA) exchange-correlation functionals allowing to extend the subsystem DFT applicability to the meta-GGA level of theory. Here, we present a two-dimensional scan of semilocal KED models as linear functionals of the reduced gradient and of the reduced Laplacian, for atoms and weakly-bound molecular systems. We find that several models can perform well but in any case the Laplacian contribution is extremely important to model the local features of the KED. Indeed a simple model constructed as the sum of Thomas-Fermi KED and 1/6 of the Laplacian of the density yields the best accuracy for atoms and weakly-bound molecular systems. These KED models are tested within subsystem DFT with various meta-GGA exchange-correlation functionals for non-bonded systems, showing a good accuracy of the method.
Unlike the local density approximation (LDA) and the generalized gradient approximation (GGA), calculations with meta-generalized gradient approximations (meta-GGA) are usually done according to the generalized Kohn-Sham (gKS) formalism. The exchange-correlation potential of the gKS equation is non-multiplicative, which prevents systematic comparison of meta-GGA bandstructures to those of the LDA and the GGA. We implement the optimized effective potential (OEP) of the meta-GGA for periodic systems, which allows us to carry out meta-GGA calculations in the same KS manner as for the LDA and the GGA. We apply the OEP to several meta-GGAs, including the new SCAN functional [Phys. Rev. Lett. 115, 036402 (2015)]. We find that the KS gaps and KS band structures of meta-GGAs are close to those of GGAs. They are smaller than the more realistic gKS gaps of meta-GGAs, but probably close to the less-realistic gaps in the band structure of the exact KS potential, as can be seen by comparing with the gaps of the EXX+RPA OEP potential. The well-known grid sensitivity of meta-GGAs is much more severe in OEP calculations.
We propose a generalized gradient approximation (GGA) for the angle- and system-averaged exchange-correlation hole of a many-electron system. This hole, which satisfies known exact constraints, recovers the PBEsol (Perdew-Burke-Ernzerhof for solids) exchange-correlation energy functional, a GGA that accurately describes the equilibrium properties of densely packed solids and their surfaces. We find that our PBEsol exchange-correlation hole describes the wavevector analysis of the jellium exchange-correlation surface energy in agreement with a sophisticated time-dependent density-functional calculation (whose three-dimensional wavevector analysis we report here).
The Strongly Constrained and Appropriately Normed (SCAN) functional is a non-empirical meta-generalized-gradient approximation (meta-GGA) functional that satisfies all the known constraints that a meta-GGA functional can, but it also exhibits a great degree of sensitivity to numerical grids. Its numerical complexities are amplified when used in Perdew-Zunger (PZ) self-interaction correction (SIC) which requires evaluating energies and potentials using orbital densities that vary far more rapidly than spin densities. Recent regularization of the SCAN functional (rSCAN) simplifies numerical complexities of SCAN at the expense of violation of some exact constraints. To develop a good understanding of the performance of rSCAN and the effect of loss of an exact constraint at the limit of slowly varying density, we have compared its performance against SCAN for vibrational frequencies, infra-red and Raman intensities of water clusters, electric dipole moments, spin magnetic moments of a few molecular magnets, weak interaction energies of dimers, barrier heights of reactions, and atomization energies for benchmark sets of molecules. Likewise, we examined the performance of SIC-rSCAN using the PZ-SIC method by studying atomic total energies, ionization potentials and electron affinities, molecular atomization energies, barrier heights, and dissociation and reaction energies. We find that rSCAN requires a much less dense numerical grid and gives very similar results as SCAN for all properties examined with the exception of atomization energies which are somewhat worse in rSCAN. On the other hand, SIC-rSCAN gives marginally better performance than SIC-SCAN for almost all properties studied in this work.
Log in to be able to interact and post comments
comments
Fetching comments
Sign in to be able to follow your search criteria
