Excited State Densities from Time-Dependent Density Functional Response Theory

Anna A. Baranova, Neepa T. Maitra · 2025

Read Excited State Densities from Time-Dependent Density Functional Response Theory by Anna A. Baranova, Neepa T. Maitra (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-2506-05082.txt.

Introduction

Anna Baranova and Neepa Maitra derive in 2025 a direct formula for the electron density of excited states in a leading quantum chemistry method. The method is widely used for excited state energies, but the quality of the matching densities in real space has drawn little study. Their expression also covers a class of improved approximations, and for the first time yields densities for states with double excitation character.

Under clear simplifying steps, they compare two basic approximations that sit at opposite extremes on simple model systems, for local and charge transfer excitations. A dressed version of the method gave good densities for the hard double excitation states. The work guides users who need densities, not just energies.

Catalog record

CREATORAnna A. Baranova, Neepa T. Maitra
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE56 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2506-05082

Browse the full library · About the Tower of Babel

Full text

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

Excited State Densities from Time-Dependent Density Functional Response Theory Anna A. Baranova, Neepa T. Maitra Abstract While the variational principle for excited-state energies leads to a route to obtaining excited-state densities from time-dependent density functional theory, relatively little attention has been paid to the quality of the resulting densities in real space obtained with different exchange-correlation functional approximations, nor how non-adiabatic approximations developed for energies of states of double excitation character perform for their densities. Here we derive an expression directly in real space for the excited-state density, that includes the case of non-adiabatic kernels, and consequently is able, for the first time, to yield densities of states of double-excitation character. Under some well-defined simplifications, we compare the performance of the local-density approximation and exact-exchange approximation, which are in a sense at opposite extremes of the fundamental functional approximations, on local and charge-transfer excitations in one-dimensional model systems, and show that the dressed TDDFT approach gives good densities of double-excitations. keywords American Chemical Society, L a T e X † † email: [email protected] † † affiliation: Department of Physics, Rutgers University, Newark 07102, New Jersey USA † † abbreviations: IR,NMR,UV 1 Introduction The advent of time-dependent density functional theory (TDDFT) forty years ago Runge and Gross 1984 ; Ullrich 2012 ; Marques et al. 2012 has enabled the calculation of electronic spectra, response properties, and general non-perturbative dynamics, for large and complex systems that would otherwise not be possible. This is due to the one-to-one mapping between time-dependent potentials and densities, for a specified initial state, opening the possibility of finding a non-interacting (Kohn-Sham) system that reproduces the same density while evolving in the computationally simpler one-body Schrödinger equation. While the approach would formally yield exact observables of the interacting system, in practice approximations must be made for the exchange-correlation potential. The majority of applications operate in the linear response regime, where one considers a time-dependent perturbation on the system, from whose response excitation energies and transition properties out of the ground-state can be extracted, and, with the approximations in use today, an unprecedented balance between accuracy and efficiency has been achieved. When thinking about the linear response of a ground-state system, it can be understood that properties related to transition densities between the ground and excited states of the true interacting system, \langle\Psi_{I}|\hat{n}({\bf r})|\Psi_{0}\rangle , can be directly obtained, but it is not immediately obvious that excited-state densities \langle\Psi_{I}|\hat{n}({\bf r})|\Psi_{I}\rangle can be extracted from the formalism. However, with the recognition that excited states are stationary points of the Hamiltonian, then using time-independent perturbation theory instead of time-dependent perturbation theory offers a route to extracting their densities from energies computed from linear response; equivalently, framing the problem as a variational one Furche 2001 ; Furche and Ahlrichs 2002 . (We note that excited state densities can also be obtained from extrema of the purely the ground-state energy functional Perdew and Levy 1985 ; Levi et al. 2020 ). In this paper, we work directly in real space instead of in the space of single-excitations that the earlier pioneering work was framed in Furche 2001 ; Furche and Ahlrichs 2002 , and going beyond the use of an adiabatic kernel that the earlier work was limited to. The latter extension enables us to obtain densities for states of double-excitation character. We explore the small-matrix approximation limit of our expressions on a series of one-dimensional model systems that cover local excitations, charge-transfer excitations, and double excitations. Motivated by the works Grabarz and Ośmiałowski 2021 ; Guido et al. 2018 ; Hait and Head-Gordon 2018 ; Verma and Truhlar 2017 ; Sarkar et al. 2021 ; Jacquemin 2016 ; Hickey and Rowley 2014 ; Herbert 2024 suggesting that the hybrid exchange-correlation functionals with a large fraction of exact exchange yield accurate excited-state properties, we explore the performance of the excited-state densities in two opposite limits of approximation – in the adiabatic local density approximation (LDA), and adiabatic exact exchange (EXX). We find that the choice of ground-state approximation is important in getting good Kohn-Sham densities in the first place, on top of which TDDFT provides corrections that bring the density close to the exact. These corrections involve a sum over all orbitals, whose accuracy depends on the choice of exchange-correlation functional, and we study the convergence of this sum. Finally, we show that the DSMA and DSPA frequency-dependent approximations provide good approximations to the densities of states of double-excitation character, while the adiabatic approximation fails. 2 Excited-state densities from time-independent perturbation theory A standard result from time-independent perturbation theory of quantum mechanics is that the first-order change to an excited-state energy is the expectation value of the perturbation evaluated in that excited state. Applied to many-electron systems when this perturbation is in the externally applied one-body potential, this then yields \delta E_{I}=\int n_{I}({\bf r})\delta v_{\rm ext}({\bf r})d^{3}r , where n_{I}({\bf r})=\sum_{\sigma_{1}..\sigma_{N}}\int|\Psi_{I}({\bf r}\sigma_{1},{\bf r}_{2}\sigma_{2}...{\bf r}_{N}\sigma_{N})|^{2}d^{3}r_{2}..d^{3}r_{N} is the density of the unperturbed excited state I . Thus, knowledge of the excited state energy and how it varies with the potential yields the excited-state density itself: n_{I}({\bf r})=\frac{\delta E_{I}}{\delta v_{\rm ext}({\bf r})}\,. (1) Eq. 1 provides the cornerstone for obtaining excited state densities from TDDFT response: while ground-state DFT provides the ground-state energy E_{0} , the TDDFT linear response formalism provides excitation energies \omega_{I} , so taking the functional derivative with respect to the external potential of the sum of these yields the excited-state density, exact in principle and dependent on functional approximations made in practise, as we shall see shortly. Thus, we write the excited state energy in two parts: E_{I}=E_{0}+\omega_{I} (2) where E_{0}=\left\langle\Psi_{0}|\hat{T}+\hat{W}|\Psi_{0}\right\rangle+\int drn_{0}(r)v_{\mathrm{ext}}(r) (with \hat{T} and \hat{W} as kinetic and electron-interaction operators), and \omega_{I}=E_{I}-E_{0} is the excitation frequency of the excited state I . The latter are eigenvalues of the TDDFT matrix equation Casida 1995 ; Casida 1996 ; Petersilka et al. 1996 ; Grabo et al. 2000 : \Omega(\omega_{I})\mathbf{G}_{I}=\omega_{I}^{2}\mathbf{G}_{I} (3) where the matrix elements of \Omega(\omega) are expressed in the single-excitation basis, q=i\rightarrow a,\;q^{\prime}=j\rightarrow b , as \Omega_{qq^{\prime}}(\omega_{I})=\nu^{2}_{q}\delta_{qq^{\prime}}+4\sqrt{\nu_{q}\nu_{q^{\prime}}}f_{{\scriptscriptstyle\rm HXC},qq^{\prime}}[n_{0}](\omega_{I})\,. (4) Here \nu_{q}={\epsilon}_{a}-{\epsilon}_{i},\;\nu_{q^{\prime}}={\epsilon}_{b}-{\epsilon}_{j} are the corresponding KS transition frequencies, and f_{{\scriptscriptstyle\rm HXC},qq^{\prime}} is Hartree-exchange-correlation matrix element, defined through f_{{\scriptscriptstyle\rm HXC},qq^{\prime}}[n_{0}](\omega_{I})=\iint d^{3}xd^{3}x^{\prime}\Phi_{q}(\mathbf{x})f_{\scriptscriptstyle\rm HXC}[n_{0}]({\bf x},{\bf x}^{\prime},\omega_{I})\Phi_{q^{\prime}}(\mathbf{x}^{\prime}) (5) where f_{\scriptscriptstyle\rm HXC}[n_{0}]({\bf x},{\bf x}^{\prime},\omega)=\frac{1}{\left|{\bf x}-{\bf x}^{\prime}\right|}+\int d(t-t^{\prime})e^{{\rm i}\omega(t-t^{\prime})}\left.\frac{\delta v_{\scriptscriptstyle\rm XC}({\bf x},t)}{\delta n({\bf x}^{\prime},t^{\prime})}\right|_{n=n_{0}} (6) and \Phi_{q}(\mathbf{x})=\phi_{i}(\mathbf{x})\phi_{a}(\mathbf{x}) represents a KS single-excitation transition density. Evaluating the derivative in Eq. ( 1 ) from Eq. 2 we have n_{I}({\bf r})=n_{0}({\bf r})+\frac{\delta\omega_{I}}{\delta v_{\rm ext}({\bf r})} (7) where the first term, the ground-state density, comes from the derivative of the ground-state energy. We focus here on the excited state density difference \Delta n_{I}(\mathbf{r})\equiv n_{I}({\bf r})-n_{0}({\bf r})=\frac{\delta\omega_{I}}{\delta v_{\mathrm{ext}}({\bf r})}=\frac{1}{2\omega_{I}}\frac{\delta\omega^{2}_{I}}{\delta v_{\mathrm{ext}}(\mathbf{r})}, (8) obtaining the derivative from Eq. ( 3 ) in the following way. Multiplying by \mathbf{G}_{I}^{\dagger} on the left, and then taking the derivative, yields the left-hand side as \displaystyle\frac{\delta}{\delta v_{\text{ext}}({\bf r})}\left(\mathbf{G}_{I}^{\dagger}\Omega(\omega_{I})\mathbf{G}_{I}\right) \displaystyle=\frac{\delta\mathbf{G}_{I}^{\dagger}}{\delta v_{\text{ext}}({\bf r})}\Omega(\omega_{I})\mathbf{G}_{I}+\mathbf{G}_{I}^{\dagger}\Omega(\omega_{I})\frac{\delta\mathbf{G}_{I}}{\delta v_{\text{ext}}({\bf r})}+\mathbf{G}_{I}^{\dagger}\frac{\delta\Omega(\omega_{I})}{\delta v_{\text{ext}}({\bf r})}\mathbf{G}_{I} (9) \displaystyle=\omega^{2}_{I}\frac{\delta}{\delta v_{\rm ext}({\bf r})}\left(\mathbf{G}_{I}^{\dagger}\mathbf{G}_{I}\right)+\mathbf{G}_{I}^{\dagger}\frac{\delta\Omega(\omega_{I})}{\delta v_{\rm ext}({\bf r})}\mathbf{G}_{I}, (where we have used the Hermitian property of \Omega(\omega) ) while the right-hand-side gives \frac{\delta}{\delta v_{\rm ext}({\bf r})}\left(\omega_{I}^{2}\mathbf{G}_{I}^{\dagger}\mathbf{G}_{I}\right)=\omega^{2}_{I}\frac{\delta}{\delta v_{\rm ext}(r)}\left(\mathbf{G}_{I}^{\dagger}\mathbf{G}_{I}\right)+\frac{\delta\omega^{2}_{I}}{\delta v_{\rm ext}(r)}\mathbf{G}_{I}^{\dagger}\mathbf{G}_{I}. (10) Equating the two sides gives \Delta n_{I}({\bf r})=\frac{1}{2\omega_{I}}\mathbf{G}_{I}^{\dagger}\frac{\delta\Omega(\omega_{I})}{\delta v_{\rm ext}({\bf r})}\left(\mathbf{G}_{I}^{\dagger}\right)^{-1}=\frac{1}{2\omega_{I}}\mathbf{G}_{I}^{-1}\frac{\delta\Omega(\omega_{I})}{\delta v_{\rm ext}({\bf r})}\mathbf{G}_{I} (11) This expression gives us the exact density-difference, once we compute the functional derivative of the TDDFT matrix (see

Read the complete text