Exact Constraint of Density Functional Approximations at the Semiclassical Limit
Yunzhi Li, Chen Li · 2025
Read Exact Constraint of Density Functional Approximations at the Semiclassical Limit by Yunzhi Li, Chen Li (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-2501-02243.txt.
Introduction
Density functional theory is the workhorse method for computing how electrons behave in molecules, but its shortcuts fail for some hard cases. Yunzhi Li and Chen Li study those shortcuts in 2025 at an extreme edge: the semiclassical limit, where a key constant of quantum physics is sent toward zero. Exact theory says the energy should stay finite there.
The pair show that mainstream shortcuts instead blow up in that limit, breaking the exact rule. They link the same flaw to the poor energies these methods give for strongly bound metal pairs. The test gives makers of new shortcuts a sharp check to pass. It may point the way to better tools for hard molecules.
Catalog record
| CREATOR | Yunzhi Li, Chen Li |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 46 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2501-02243
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Exact Constraint of Density Functional Approximations at the Semiclassical Limit Yunzhi Li, Chen Li Abstract ABSTRACT: We introduce the semiclassical limit to electronic systems by taking the limit ℏ → 0 \hbar\rightarrow 0 in the solution of Schrödinger equations. We show that this limit is closely related to one type of strong correlation that is particularly challenging from conventional multi-configurational perspective but can be readily described through semiclassical analysis. Furthermore, by studying the performance of density functional approximations (DFAs) in the semiclassical limit, we find that mainstream DFAs have erroneous divergent energy behaviors as ℏ → 0 \hbar\rightarrow 0 , violating the exact constraint of finite energy. Importantly, by making connection of the significantly underestimated DFA energies of many strongly correlated transition-metal diatomic molecules to their rather small estimated ℏ eff \hbar_{\text{eff}} , we demonstrate the usefulness of our semiclassical analysis and its promise for inspiring better DFAs. I Introduction Density functional theory (DFT) with the existing approximations has been extremely successful in describing weakly correlated systems, ranging from atoms, molecules to metallic solids. Through introducing an auxiliary non-interacting system, Kohn and Sham 1965 Kohn-Sham DFT is endowed with the physical picture of independent electrons moving in a mean field, which is essentially valid for most of the main-group compounds of chemical interests. In such a way, the complicated many-body wave function is effectively reduced to N occupied orbitals. This crucial simplification makes life so much easier in terms of both theory and computation that chemists cannot live without the concept of orbitals since the early days of quantum mechanics until the present day. However, there is one class of challenging systems where the single-electron orbital picture breaks down, known as strongly correlated systems. Typical examples of strongly correlated systems include stretched molecules and transition-metal compounds, Chan et al. 2019 ; Cohen et al. 2008 ; Kulik and Marzari 2010 where the correlation cannot be captured by perturbative treatment to any single Slater determinant, and one cannot find a dominant determinant in the configuration interaction (CI) expansion of the many-body wave function. Cohen et al. 2012 ; Becke 2014 ; Mardirossian and Head-Gordon 2017 ; Bulik et al. 2015 Strong correlation can also occur in solid materials, such as Mott insulators, Mott and Peierls 1937 ; Mott 1968 ; Imada et al. 1998 ; Boer and Verwey 1937 ; Zaanen et al. 1985 ; Kulik 2015 where the strong electron–electron interaction leads to the breakdown of the band structure theory. Kotliar and Vollhardt 2004 It is crucial to note that the above-mentioned challenges should not be regarded as failures of density functional theory, which by the Hohenberg-Kohn formulation Hohenberg and Kohn 1964 is formally exact. Instead, problem arises in the exchange-correlation (xc) functional in the Kohn-Sham decomposition of the total energy, whose exact form cannot be obtained explicitly and in practical calculations has to be treated by density functional approximations (DFAs). It is the qualitatively incorrect behaviors of these DFAs, characterized by the violation of some important conditions or constraints, that are responsible for their failures in strongly correlated systems. Among other exact constraints, one important constraint relevant for strong correlation is the fractional spin (FS) condition. Cohen et al. 2008 ; Cohen et al. 2008 In particular, the strong correlation problem of a stretched molecule is understood from the perspective that local spin densities of an isolated fragment integrate to a fractional number, a problematic scenario where DFAs tend to overestimate the energy. Along this line of thinking, functionals have been constructed to restore the local FS condition, which can then properly describe molecular dissociation. Su et al. 2018 However, for unstretched molecules such as well-bonded metal oxides, how one should perform the FS analysis remains a major challenge. Figure 1: Schematic illustration of the weakly and strongly correlated picture: (a) two electrons independently orbit around the helium nucleus, representative of a weakly correlated picture; (b) two oxygen atoms rotate around the carbon atom in collective motion in a CO 2 molecule, representative of a strongly correlated picture. To gain more insight into the problem of strong correlation, let us take one step back and review the physical nature of weak and strong correlation through an example of a two-particle system. In a weakly correlated picture such as the electrons in a helium atom in its ground state, the electronic motions are almost independent from each other, pictorially shown as in Fig. 1 (a). It follows from the probability theory that the joint probability of finding one electron at \mathbf{r}_{1} and the other at \mathbf{r}_{2} is given by the following product, \rho(\mathbf{r}_{1},\mathbf{r}_{2})\approx\rho_{1s}(\mathbf{r}_{1})\rho_{1s}(\mathbf{r}_{2}) , where \rho_{1s}(\mathbf{r}) is the probability of finding an electron in the 1s orbital. In the case of strong correlation, in contrast, the joint probability \rho(\mathbf{r}_{1},\mathbf{r}_{2}) becomes a \delta function. This is because the motions of the two electrons are strongly entangled; if one knows the position of one particle, then one immediately knows the position of the other. Interpreted physically, having motions of collective rather than of individual nature, is what distinguishes strong correlation from weak correlation. Such a picture is reminiscent of atoms in a molecule within the semiclassical treatment. We can remind ourselves of the linear CO 2 molecule as shown in Fig. 1 (b), where the two oxygen atoms are strongly entangled as they always appear equidistantly on the opposite sides of the carbon atom. The fact that the strong correlation limit and the semiclassical limit has a lot in common is not a coincidence. Here let us remind ourselves that in the most general sense semiclassical limit is defined by taking a small \hbar -related dimensionless quantity to zero; Landsman 2007 this quantity can be \hbar itself or a combination of \hbar and other parameters such as the nuclear mass. In CO 2 , because the nuclear kinetic energy T_{n} is much smaller than the potential energy V (due to the much heavier nuclear mass compared to the electronic mass), as a good approximation by Born and Oppenheimer, Born and Oppenheimer 1927 we can neglect T_{n} and treat the nuclei as clamped at their classical positions. This forms a major contrast with the electrons, for which the kinetic energy T is for most cases on the same order of magnitude as V . However, there are exceptions. A famous example is the uniform electron gas (UEG) in the low density limit, where V dominates over T such that electrons behave like semiclassical particles that crystalize to form a lattice, known as the Wigner crystal. Wigner 1934 ; Wigner 1938 ; Carr 1961 Fig. 2 (a) is an illustration of the Wigner crystal for a body-centered cubic (bcc) lattice. In the design of the simplest local density approximation (LDA) to the correlation energy by Vosko, Wilk and Nusair (VWN5) Ceperley and Alder 1980 ; Vosko et al. 1980 , the correct limiting behaviors for low as well as high density UEG have already been built in the functional. In this sense, it comes as no surprise that LDA can properly describe metallic solids even though some metals are presumably strongly correlated- such systems are not so distinct from the UEG. Figure 2: Schematic illustration comparing (a) Wigner crystal and (b) a 5-electron “Wigner