From Densities to Potentials: Benchmarking Local Exchange-Correlation Approximations
Visagan Ravindran, Clio Johnson, Neil D. Drummond, Stewart J. Clark, Nikitas I. Gidopoulos · 2025
Read From Densities to Potentials: Benchmarking Local Exchange-Correlation Approximations by Visagan Ravindran, Clio Johnson, Neil D. Drummond, Stewart J. Clark, Nikitas I. Gidopoulos (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2511-02744.txt.
Introduction
Visagan Ravindran and four colleagues test how well common methods in density functional theory match near-exact results. Chemists use these methods to predict electron behavior, but each one builds a different effective potential. Their 2025 study turns the problem around: starting from quantum Monte Carlo densities, they invert the Kohn-Sham equations to recover benchmark potentials.
The benchmarks cover a range of insulators and semiconductors. Different methods gave similar electron densities despite clearly different potentials, with most gaps coming from the exchange and correlation part. The gap from the Monte Carlo density was usually larger than the methods' gaps, Hartree-Fock excepted. The authors also warn that including semicore states shifts the gap, so care is needed when comparing with experiment.
Catalog record
| CREATOR | Visagan Ravindran, Clio Johnson, Neil D. Drummond, Stewart J. Clark, Nikitas I. Gidopoulos |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 74 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2511-02744
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
From Densities to Potentials: Benchmarking Local Exchange-Correlation Approximations Visagan Ravindran, Clio Johnson, Neil D. Drummond, Stewart J. Clark, Nikitas I. Gidopoulos Abstract Using the Kohn-Sham (KS) inversion method of Hollins et al. [J. Phys.: Condens. Matter 29 , 04LT01 (2017)], we invert densities from variational and diffusion quantum Monte Carlo (QMC) calculations to obtain benchmark QMC-KS potentials for a range of insulators and semiconductors, which we then compare to the KS potentials of popular density functional approximations (DFAs). Our results show that different DFAs yield similar electron densities, despite differences in their KS potentials, which originate primarily from the exchange and correlation contribution. We also find that the KS gap from the QMC density is typically larger than the KS gaps of most DFAs, with the exception of Hartree-Fock. Finally, the KS gap is sensitive to the inclusion of semicore states in the pseudopotentials, such that comparison with experiment should be done with caution. I Introduction Kohn-Sham (KS) density functional theory (DFT) Kohn and Sham 1965 ; Hohenberg and Kohn 1964 is ubiquitous in ab initio electronic structure calculations within condensed matter physics, chemistry, and materials science. In KS theory, an interacting system of electrons is mapped to an auxiliary noninteracting system of particles moving in a mean-field effective potential v_{s}(\mathbf{r}) that is constructed such that the ground state densities of the interacting and noninteracting systems are the same. However, the exact exchange-correlation (XC) contribution E_{xc}[\rho] to the KS energy functional is unknown, and by extension the exact XC contribution v_{xc}[\rho](\mathbf{r})=\frac{\delta E_{xc}[\rho]}{\delta\rho(\mathbf{r})} (1) to the KS potential v_{s}(\mathbf{r}) is also unknown, necessitating approximations for both E_{xc}[\rho] and v_{xc}(\mathbf{r}) in practice. Numerous density functional approximations (DFAs) Becke 1988 ; Perdew et al. 1996a ; Perdew et al. 2008 ; Tao et al. 2003 ; Sun et al. 2015 ; Bartók and Yates 2019 ; Furness et al. 2020 ; Becke 1993 ; Heyd et al. 2003 ; Perdew et al. 1996b have been proposed over the decades but the direct assessment of the accuracy of E_{xc}[\rho] and v_{s}(\mathbf{r}) is a formidable challenge, not least because neither are experimentally observable quantities. Density inversion van Leeuwen and Baerends 1994 ; Görling 1992 ; Zhao et al. 1994 ; Chan et al. 1997 ; Savin et al. 1998 ; Peirs et al. 2003 ; Kadantsev and Stott 2004 ; Ryabinkin and Staroverov 2012 ; Jensen and Wasserman 2018 ; Shi and Wasserman 2021 ; Kumar et al. 2019 ; Erhard et al. 2022 ; Aouina et al. 2023 ; Aouina et al. 2024 ; Kaiser and Kümmel 2025 offers a possible methodology to assess v_{s}(\mathbf{r}) and in particular v_{xc}(\mathbf{r}) by tackling the inverse KS problem, where for a given target density \rho_{\mathrm{t}}(\mathbf{r}) one finds the KS potential v_{s}(\mathbf{r}) such that the ground-state density of the KS auxiliary system is equal to \rho_{\mathrm{t}}(\mathbf{r}) . Consequently, the inversion of numerically accurate densities enables one to gain insight into the behavior of the exact v_{s}(\mathbf{r}) . Quantum Monte Carlo (QMC) Foulkes et al. 2001 methods are a well-established family of many-body techniques, which have played a historically significant role in the development of DFAs, notably in facilitating parameterizations Perdew and Wang 1992 ; Perdew and Zunger 1981 for the local density approximation (LDA) through QMC simulations of the homogeneous electron gas (HEG) Ceperley and Alder 1980 . Among such methods are the variational and diffusion Monte Carlo (VMC and DMC) methods. In the VMC method, electron configurations are sampled from the square modulus of a trial wave function using the Metropolis algorithm, with estimators of observables of interest being averaged over those configurations. The trial wave function contains free parameters that are optimized by minimizing either the energy expectation value or the variance of the energy. In the DMC method Ceperley and Alder 1980 a population of electron configurations (also known as walkers) is evolved according to the Schrödinger equation in imaginary time to project out the ground-state component of the trial wave function. Fermionic antisymmetry is maintained by fixing the complex phase of the DMC wave function at that of the VMC trial wave function. In this work, we use the VMC and DMC methods to calculate the ground-state electronic densities of various semiconductors and insulators. We then invert these densities to find the KS potentials v_{s}(\mathbf{r}) that give rise to them in the ground state. Using the QMC densities and QMC-derived KS potentials as a benchmark, we assess the quality of various DFAs using a range of metrics, including KS band gaps, integrated absolute density differences, and integrated potential differences weighted by density differences. Following Burke and coworkers Kim et al. 2013 ; Wasserman et al. 2017 ; Sim et al. 2018 ; Vuckovic et al. 2019 , we also analyze the total energy error of each DFA by examining the density-driven and functional error contributions. The rest of this article is structured as follows. In Sec. II.1 we provide details of our QMC calculations, in particular the calculation of densities, while Sec. II.2 outlines our algorithm to invert densities and obtain the KS potential. We describe the errors and uncertainties in our charge densities in Sec. III . We compare our exact Kohn-Sham potential with other local potentials in Sec. IV , while the KS band gaps for each DFA along with the XC derivative discontinuity \Delta_{xc} are described in Sec. V . Finally, we draw our conclusions in Sec. VI . We use Hartree atomic units (a.u.), in which the reduced Planck’s constant \hbar , the electron mass m_{\text{e}} , the magnitude of the electronic charge e , and 4\pi\varepsilon_{0} are 1 a.u, where \varepsilon_{0} is the permittivity of free space. II Theory and computational details II.1 Obtaining QMC charge densities II.1.1 QMC calculations The VMC and DMC calculations reported in this work were performed using the casino program Needs et al. 2020 , and the DFT calculations were carried out using the castep plane-wave-basis-set code Clark et al. 2005 . For the materials studied we used experimental lattice parameters taken from Ref. Madelung 1996 unless otherwise stated (see Table 1 ). Trail-Needs (TN) Dirac-Fock pseudopotentials Trail and Needs 2005 were used to represent atomic cores. The s channel was chosen to be the local channel of the pseudopotential in each case to avoid possible issues with ghost states Drummond et al. 2016 that can arise due to the Kleinman-Bylander Kleinman and Bylander 1982 representation of the pseudopotentials in plane-wave DFT calculations. Our QMC trial wave functions were of Slater-Jastrow (SJ) form, \Psi_{\text{T}}\left(\mathbf{R}\right)=S_{\uparrow}\left(\mathbf{R}\right)S_{\downarrow}\left(\mathbf{R}\right)\exp\left(J\left(\mathbf{R}\right)\right), (2) where \mathbf{R} is the 3N -dimensional electron configuration vector and S_{\uparrow/\downarrow}\left(\mathbf{R}\right) are Slater determinants of single-particle orbitals for spin-up and spin-down electrons. The Jastrow exponent J\left(\mathbf{R}\right) consisted of electron–electron, electron–ion, and electron–electron–ion polynomials, as well as electron–electron plane-wave expansions, with the coefficients being optimizable parameters Drummond et al. 2004 . The orbitals in the Slater determinants were generated using castep with the Perdew-Burke-Ernzerhof (PBE) Perdew et al. 1996a XC functional and re-represented in a B-spline (blip) basis for use in casino Hernández et al. 1997 ; Alfè and Gillan 2004 . For greater accuracy in fixed-node DMC calculations for a particular system, Slater-Jastrow-backflow (SJB) Kwon et al. 1998 ; López Ríos et al. 2006 wave functions can be used.