Orbital-Free Density Functional Theory for Periodic Solids: Construction of the Pauli Potential

Sangita Majumdar, Zekun Shi, Giovanni Vignale · 2025

Read Orbital-Free Density Functional Theory for Periodic Solids: Construction of the Pauli Potential by Sangita Majumdar, Zekun Shi, Giovanni Vignale (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-2503-15959.txt.

Introduction

Sangita Majumdar, Zekun Shi, and Giovanni Vignale push orbital-free density functional theory toward real materials. The standard Kohn-Sham method computes electron motion through helper orbitals, which is exact but costly. A direct link from electron density to kinetic energy would be faster, yet finding it for repeating solids has lagged.

Their 2025 paper gives an algorithm that directly minimizes the energy for periodic densities, the repeating patterns of crystals. A second routine extracts the matching potential, the Pauli side of the problem named in the title. The code also detects the potential's jump when an electron is added, which separates insulators from conductors. Tricks like an adaptive basis set keep the computation quick.

Catalog record

CREATORSangita Majumdar, Zekun Shi, Giovanni Vignale
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE74 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2503-15959

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Orbital-Free Density Functional Theory for Periodic Solids: Construction of the Pauli Potential Sangita Majumdar, Zekun Shi, Giovanni Vignale Abstract The practical success of density functional theory (DFT) is largely credited to the Kohn-Sham approach, which enables the exact calculation of the non-interacting electron kinetic energy via an auxiliary noninteracting system. Yet, the realization of DFT’s full potential awaits the discovery of a direct link between the electron density, n n , and the non-interacting kinetic energy, T S ​ [ n ] T_{S}[n] . In this work, we address two key challenges towards this objective. First, we introduce a new algorithm for directly solving the constrained minimization problem yielding T S ​ [ n ] T_{S}[n] for periodic densities – a class of densities that, in spite of its central importance for materials science, has received limited attention in the literature. Second, we present a numerical procedure that allows us to calculate the functional derivative of T S ​ [ n ] T_{S}[n] with respect to the density at constant electron number, also known as the Kohn-Sham potential V S ​ [ n ] ​ ( 𝐫 ) V_{S}[n]({\bf r}) . Lastly, the algorithm is augmented with a subroutine that computes the “derivative discontinuity”, i.e., the spatially uniform jump in V S ​ [ n ] ​ ( 𝐫 ) V_{S}[n]({\bf r}) which occurs upon increasing or decreasing the total number of electrons. This feature allows us to distinguish between “insulating” and “conducting” densities for non interacting electrons. The code integrates key methodological innovations, such as the use of an adaptive basis set (“equidensity orbitals”) for wave function expansion and the QR decomposition to accelerate the implementation of the orthogonality constraint. Notably, we derive a closed-form expression for the Pauli potential in one dimension, expressed solely in terms of the input density, without relying on Kohn-Sham eigenvalues and eigenfunctions. We validate this method on one-dimensional periodic densities, achieving results within “chemical accuracy”. † † affiliation: Institute for Functional Intelligent Materials (I-FIM), National University of Singapore, 4 Science Drive 2, Singapore 117544 † † affiliation: SEA AI Lab, Singapore † † affiliation: School of Computing, National University of Singapore, 13 Computing Drive, Singapore 117417 † † affiliation: Institute for Functional Intelligent Materials (I-FIM), National University of Singapore, 4 Science Drive 2, Singapore 117544 † † email: [email protected] 1 Introduction The original density functional theory (DFT) of Hohenberg and Kohn Hohenberg and Kohn (1964) purported to replace the variational principle of quantum mechanics, in which the energy is minimized with respect to a many-body wave function, by a much simpler variational principle in which the energy is minimized with respect to the electronic density, a function of position denoted by n({\bf r}) . In practice, that bold vision never materialized. The second landmark paper in the field (Kohn-Sham, KS) Kohn and Sham (1965) already introduced a representation of the electronic density in terms of orthogonal single-particle orbitals (still much simpler than the full many-body wave function) and the minimization of the energy with respect to the density was replaced by the solution of self-consistent equations for the single-particle orbitals, which are now known as the Kohn-Sham equations. By doing that, they not only ensured that the electronic density would be compatible with the requirements of the Pauli exclusion principle, but also that one of the largest components of the energy, the non-interacting kinetic energy, defined precisely below, would be treated exactly. From then onwards, much of the research in DFT focused on devising good approximations for the genuinely many-body part of the energy functional, the exchange-correlation energy functional E_{xc}[n] and its functional derivative with respect to density, known as the exchange-correlation potential, V_{xc}[n]({\bf r}) . While tremendous progress was made in this direction, many problems that lurked under the smooth surface of the formalism gradually came into focus (self-interaction error, derivative discontinuity, band gaps, strong correlations, degenerate ground states, broken symmetry) Cohen et al. (2012) ; Verma and Truhlar (2020) and the community learned to cope with them with varying degrees of success. However, the quest for a practical orbital-free implementation of DFT was never abandoned Karasiev and Trickey (2012) . At the heart of the quest lies the formidable problem of constructing the universal energy functional F[n] (sometimes referred to as the “Holy Grail” of DFT) Lieb (1983) ; Perdew and Zunger (1981) ; Lewin et al. (2023) whose formal definition is [Levy-Lieb] Levy (1979) ; Lieb (1983) F[n]\equiv\min_{\Psi\to n}\langle\Psi|\hat{T}+\hat{U}|\Psi\rangle (1) where |\Psi\rangle is an N -particle state with wave function \Psi of the correct symmetry (completely antisymmetric for fermions), \hat{T} is the kinetic energy operator, \hat{U} is the two-body interaction energy operator. The minimization is carried over the set of wave functions that yield the prescribed density n({\bf r}) (positive, integrating to the total particle number N ) 1 1 1 This definition can be extended to allow the search to run over ensembles of states that yield the prescribed density. Such an extension is, of course, mandatory in the case of fractional particle numbers. The appeal of this definition is that F[n] is universal , i.e., in principle, it needs to be calculated only once and applies to all electronic systems, atoms, molecules, or solids. Different systems differ only by the external potential V({\bf r}) : their ground state energy, E_{0} , is obtained by solving the minimization problem E_{0}=\min_{n}\left\{F[n]+\int n({\bf r})V({\bf r})d{\bf r}\right\}\,. (2) The F functional can further be separated into three components F[n]=T_{s}[n]+E_{H}[n]+E_{xc}[n]\,, (3) where T_{s}[n] is the non-interacting kinetic energy defined as T_{s}[n]\equiv\min_{\Psi\to n}\langle\Psi|\hat{T}|\Psi\rangle\,, (4) E_{H}[n]=\frac{1}{2}\int d{\bf r}\int d{\bf r}^{\prime}n({\bf r})U({\bf r},{\bf r}^{\prime})n({\bf r}^{\prime}) is the classical potential of the particle distribution, (excluding correlation and including an unphysical self-interaction of each particle with itself) E_{xc}[n] is the remainder. In the Kohn-Sham theory, the functional T_{s}[n] does not appear explicitly: it is implicitly determined as the sum of the kinetic energies associated with the orbitals of the self-consistent solution of Kohn-Sham equations: hence most efforts focused on approximating E_{xc} and V_{xc} . There are, however, some drawbacks to the Kohn-Sham procedure. First of all, the solution of the self-consistent field (SCF) equation is relatively slow (computational time scaling as N^{3} ). Second, not knowing T_{s}[n] as a functional of density precludes us from extracting E_{xc}[n] for benchmarking purposes from accurate quantum mechanical solutions of many-body systems (e.g., configuration interaction Leininger et al. (1997) ; Sherrill and Schaefer III (1999) , quantum Monte Carlo Hammond et al. (1994) ; Nightingale and Umrigar (1998) ; Foulkes et al. (2001) ; Santos (2003) ). A more basic reason for trying to explicitly compute T_{s}[n] and its functional derivative ( \frac{\delta T_{s}[n]}{\delta n({\bf r})}=-V_{s}[n]({\bf r}) , also known as the Kohn-Sham potential) emerges in the context of Machine Learning (ML) and Artificial Intelligence (AI) as applied to electronic structure and materials science Sun and Chen (2024) ; Zhong et al. (2024) . It appears that the T_{s} functional could be “learned” by a neural network, given the relatively simple nature of the basic variable n({\bf r}) and the relative ease of producing large amounts of training data by simply solving the Kohn-Sham equation – an operation that is routinely performed thousands of times every day. Snyder et al. (2012) ;

Read the complete text