On the Subsystem Formulation of Linear-Response Time-Dependent DFT

A new and thorough derivation of linear-response subsystem time-dependent density functional theory (TD-DFT) is presented and analyzed in detail. Two equivalent derivations are presented and naturally yield self-consistent subsystem TD-DFT equations. One reduces to the subsystem TD-DFT formalism of Neugebauer [J. Chem. Phys. 126, 134116 (2007)10.1063/1.2713754]. The other yields Dyson type equations involving three types of subsystem response functions: coupled, uncoupled, and Kohn-Sham. The Dyson type equations for subsystem TD-DFT are derived here for the first time. The response function formalism reveals previously hidden qualities and complications of subsystem TD-DFT compared with the regular TD-DFT of the supersystem. For example, analysis of the pole structure of the subsystem response functions shows that each function contains information about the electronic spectrum of the entire supersystem. In addition, comparison of the subsystem and supersystem response functions shows that, while the correlated response is subsystem additive, the Kohn-Sham response is not. Comparison with the non-subjective partition DFT theory shows that this non-additivity is largely an artifact introduced by the subjective nature of the density partitioning in subsystem DFT.

💡 Research Summary

This paper presents a comprehensive and rigorous derivation of linear‑response subsystem time‑dependent density‑functional theory (TD‑DFT). Two formally equivalent routes are followed. The first route starts from a variational principle applied to a subsystem‑specific Lagrangian, leading to subsystem Kohn‑Sham equations that contain an averaged embedding potential. The resulting self‑consistent equations are shown to be identical to the subsystem TD‑DFT formalism introduced by Neugebauer in 2007, thereby confirming that the established approach is a special case of the more general framework developed here.

The second route adopts a response‑function perspective. Three distinct subsystem response functions are defined: (i) the coupled (or fully interacting) response, (ii) the uncoupled (or isolated) response, and (iii) the Kohn‑Sham (non‑interacting) response. By inserting these definitions into the Dyson equation for the total system, a set of Dyson‑type equations specific to subsystems is derived for the first time. These equations reveal how the total correlated response χ can be expressed as a sum of uncoupled subsystem responses χ⁰ together with a convolution of χ⁰ with the exchange‑correlation kernel fₓc. Importantly, the coupled subsystem response contains poles at all excitation energies of the full supersystem, while the uncoupled response only carries the poles of the individual fragment. Consequently, each subsystem response function encodes the complete electronic spectrum of the whole system, albeit with fragment‑dependent residues that reflect inter‑fragment coupling.

A detailed pole‑structure analysis demonstrates that the correlated response is additive over subsystems (χ = Σ_i χ_i), whereas the Kohn‑Sham response is not. The non‑additivity originates from the subjective nature of the density partitioning employed in subsystem DFT: the assignment of electron density to fragments is not unique, leading to overlapping or missing density contributions that manifest as extra terms in the Kohn‑Sham response. By contrasting these findings with non‑subjective partition DFT, the authors show that when a unique, physically motivated partition is used, the Kohn‑Sham response regains additivity. This comparison highlights that the apparent deficiency of subsystem TD‑DFT is not a fundamental flaw but a consequence of the chosen partitioning scheme.

The paper also discusses practical implementation aspects. In a typical calculation, each fragment’s Kohn‑Sham orbitals and potentials are obtained independently, after which the Dyson equations are solved iteratively to incorporate the embedding kernel. The choice of an accurate coupling kernel and a reliable model for electron‑transfer between fragments are identified as critical for achieving quantitative agreement with full‑system TD‑DFT.

Overall, the work clarifies the mathematical structure of subsystem TD‑DFT, unifies the Neugebauer formulation with a novel Dyson‑type formalism, and provides new insight into how subsystem response functions carry global excitation information. These results lay a solid theoretical foundation for applying TD‑DFT to large, heterogeneous systems such as biomolecules, polymers, and surface‑adsorbate complexes, where a full‑system treatment would be computationally prohibitive. Future directions suggested include the development of non‑subjective partitioning strategies and higher‑order approximations to the exchange‑correlation kernel, both of which promise to improve the accuracy and robustness of subsystem TD‑DFT.