Critical point search and linear response theory for computing electronic excitation energies of molecular systems. Part I: General framework, application to Hartree-Fock and DFT
Laura Grazioli, Yukuan Hu, Eric Cancès · 2025
Read Critical point search and linear response theory for computing electronic excitation energies of molecular systems. Part I: General framework, application to Hartree-Fock and DFT by Laura Grazioli, Yukuan Hu, Eric Cancès (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-2506-16420.txt.
Introduction
Laura Grazioli, Yukuan Hu, and Eric Cancès set out a shared frame for computing the excited states of molecules. Their paper is from two thousand twenty-five. Excited states of many-body systems are a core hard task in physics and chemistry. Past methods split into two camps: search for critical points, or linear response.
They use the geometry of Kähler manifolds to hold both camps in one frame, fit for models from Hartree-Fock to full configuration methods and time-based density methods. The frame gives a clean path to the linear response equations of nonlinear models. For mean-field models it yields a simpler route than the classic Casida path. Part one ends with theory checks and sums at the Hartree-Fock level.
Catalog record
| CREATOR | Laura Grazioli, Yukuan Hu, Eric Cancès |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 66 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2506-16420
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Critical point search and linear response theory for computing electronic excitation energies of molecular systems. Part I: General framework, application to Hartree-Fock and DFT Laura Grazioli, Yukuan Hu, Eric Cancès Abstract Computing excited states of many-body quantum Hamiltonians is a fundamental challenge in computational physics and chemistry, with state-of-the-art methods broadly classified into variational (critical point search) and linear response approaches. This article introduces a unified framework based on the Kähler manifold formalism, which naturally accommodates both strategies for a wide range of variational models, including Hartree-Fock, CASSCF, Full CI, and adiabatic TDDFT. In particular, this formalism leads to a systematic and straightforward way to obtain the final equations of linear response theory for nonlinear models, which provides, in the case of mean-field models (Hartree-Fock and DFT), a simple alternative to Casida’s derivation. We detail the mathematical structure of Hamiltonian dynamics on Kähler manifolds, establish connections to standard quantum chemistry equations, and provide theoretical and numerical comparisons of excitation energy computation schemes at the Hartree-Fock level. 1 Introduction Computing excited states of many-body quantum Hamiltonians is one of the biggest challenges in computational physics and chemistry 1 . The main approaches for computing approximate excited states can be classified into two categories: 1. variational or bivariational methods: computation of critical points of energy functionals on differentiable manifolds or algebraic varieties. This category encompasses state-specific and state-averaged methods 2 , 3 , 4 , 5 , 6 , 7 , 8 ; 2. methods based on linear response theory. Linear-response Hartree-Fock 9 , 10 , 11 , 12 , 13 , 14 , 15 (LR-HF), Complete Active Space Self-Consistent Field 16 , 17 , 18 , 19 (LR-CASSCF), time-dependent density functional theory 20 , 21 , 22 , 23 , 24 , 25 (LR-TDDFT), and equation-of-motion coupled cluster 26 , 27 , 28 , 29 , 30 , 31 , 32 , 33 , 34 (EOM-CC) fall into this category. All these approaches coincide for the reference electronic Schrödinger equation set on the continuous position space \bigwedge^{N}L^{2}({\mathbb{R}}^{3};{\mathbb{C}}^{2}) , as well as for the Full Configuration Interaction (FCI) approximation 35 , 36 , but differ when further approximations are made. This article aims at providing a unified and compact framework covering critical point search (CP) and linear response theory (LR) from a theoretical viewpoint. We will use the Kähler manifold formalism allowing one to conveniently formulate CP and LR for any variational model, including any variant of HF (restricted, unrestricted, general, restricted open-shell), CASSCF, and FCI methods, in a unified and compact framework. It also applies to adiabatic TDDFT 37 , 38 , as this method has an intrinsic Hamiltonian Kähler structure, although it is not a variational approximation of the reference electronic Schödinger equation. We will concentrate on the FCI and HF theories in this study. The cases of more advanced variational approximations such as CASSCF 39 and bivariational approximations such as CC 40 , 41 , 42 will be dealt with in a future work. If \mathcal{M} is a Kähler manifold (the definition and properties of this mathematical structure will be recalled in Section 2 ), and \mathcal{E}:\mathcal{M}\to{\mathbb{R}} an energy functional over \mathcal{M} , the Hamiltonian Kähler dynamics is defined as {\frac{dx}{dt}(t)=J_{x(t)}{\rm grad}_{\mathcal{M}}\mathcal{E}(x(t))}, (1) where {\rm grad}_{\mathcal{M}}\mathcal{E}(x) is the Riemannian gradient of \mathcal{E} at point x\in\mathcal{M} , and J_{x} the so-called symplectic operator at point x . The critical points (steady states) therefore satisfy {{\rm grad}_{\mathcal{M}}\mathcal{E}(x_{\star})=0}. (2) The CP approach for computing excitation energies consists in: 1. finding critical points x_{0},x_{1},x_{2},\ldots of \mathcal{E} on \mathcal{M} that can be identified with approximate ground and excited states, labelled in such a way that \mathcal{E}(x_{0})\leq\mathcal{E}(x_{1})\leq\mathcal{E}(x_{2})\leq\cdots ; 2. computing the k -th excitation energy as the difference \omega^{\rm CP}_{k}:=\mathcal{E}(x_{k})-\mathcal{E}(x_{0}), (3) between the energies of the approximation x_{k} of the k -th excited state and the approximation x_{0} of the ground state, i.e., the (presumably) global minimum of \mathcal{E} on \mathcal{M} . Step 1 is quite involved. Indeed, in contrast with linear models such as FCI, not all critical points of nonlinear models such as HF or CASSCF are suitable approximations of ground or excited states of the reference electronic Schrödinger equation. In general, the energy functional \mathcal{E} can have many spurious critical points on \mathcal{M} . Worse still, there can be mismatch between the k -th excited states and the critical points with a Morse index k , i.e., critical points where the Riemannian Hessian has exactly k negative eigenvalues. This phenomenon is illustrated on a 2D toy model in Fig. 1 , as well as on an H 4 molecule in Section 5.2 . In Ref. 43 , Lewin introduced a nonlinear min-max characterization of physically-relevant critical points for CASSCF with no internal orbitals. In Ref. 44 , an algorithm for computing CASSCF approximations of first excited states based on this characterization was proposed and tested on an H 2 molecule. Figure 1 : Contour of a 2D function E:(0,1)\times(0,1)\to{\mathbb{R}} with one global minimizer ( A ), one local, non-global minimizer ( A^{\prime} ) and one saddle point with Morse index 1 ( B ). The restriction \mathcal{E} of this function to the one-dimensional manifold \mathcal{M} (indicated by the orange curve) has 3 global minimizers ( a_{1} , a_{2} , a^{\prime} ), 0 local, non-global, minimizers, and 2 critical points ( b , b^{\prime} ) with a Morse index 1, which are also local maxima since \mathcal{M} is one-dimensional. The points a_{1} and a_{2} can be considered as reasonable approximations of A , a^{\prime} as a reasonable approximation of A^{\prime} , b as a reasonable approximation of B . The point b^{\prime} is a spurious critical point, which does not approximate well any saddle point of E . In LR, the excitation energies are identified with the resonant energies of the linearized dynamics around the global minimum x_{0} of \mathcal{E} on \mathcal{M} . This dynamics is given by {\frac{dy}{dt}(t)=J_{x_{0}}{\rm Hess}_{\mathcal{M}}\mathcal{E}(x_{0})y(t)}, (4) where {\rm Hess}_{\mathcal{M}}\mathcal{E}(x_{0}) is the Riemannian Hessian of \mathcal{E} at x_{0} . Under a nondegeneracy assumption, it follows that the approximate excitation energies \omega^{\rm LR}_{k} computed by LR are in fact the symplectic eigenvalues of {\rm Hess}_{\mathcal{M}}\mathcal{E}(x_{0}) : \{\omega_{k}^{\rm LR}\}_{1\leq k\leq\dim_{{\mathbb{C}}}(\mathcal{M})}:=\{\mbox{symplectic eigenvalues of }{\rm Hess}_{\mathcal{M}}\mathcal{E}(x_{0})\}. (5) This approach provides a systematic and conceptually simple way to derive linear-response approximations to excitation energies for any variational model with Kähler structure. In Section 2 , we present the abstract framework of Hamiltonian dynamics on a generic Kähler manifold, and define precisely the mathematical objects arising in the fundamental Eqs. ( 1 )-( 5 ). In Section 3 , we consider the specific case of Hamiltonian dynamics on Grassmann manifolds \mathcal{M}={\rm Gr}_{\mathbb{C}}(r,n):=\left\{\gamma\in{\mathbb{C}}^{n\times n}_{\rm herm}\;|\;\gamma^{2}=\gamma,\;{\rm Tr\,}(\gamma)=r\right\}, (6) allowing one to deal with FCI (or more generally any linear Schrödinger or Liouville equation on a finite-dimensional state-space), HF, and adiabatic TDDFT. In Section 3.1 , we describe the Kähler structure of Grassmann manifolds. In Section 3.2 , we give more explicit formulations of Eqs. ( 1 )-( 5 ) in the special case when \mathcal{M} is a Grassmann manifold and apply this formalism to a molecular system with N_{\rm b} atomic spin-orbitals and N electrons. In Section 3.3 , we deal with the FCI framework, where n=\binom{N_{\rm b}}{N} , r=1 , \Gamma\in{\rm Gr}_{\mathbb{C}}(1,n)\subset{\mathbb{C}}_{\rm herm}^{n\times n} represents the