XDFT: an efficient first-principles method for neutral excitations in molecules

TL;DR

XDFT is a new, computationally efficient first-principles method for calculating neutral excitations in molecules, capable of capturing single and double excitations with accuracy comparable to TDDFT.

Contribution

The paper introduces XDFT, a novel constrained DFT-based approach that efficiently computes neutral excitations, including double excitations, surpassing traditional methods in speed and applicability.

Findings

Accurately predicts singlet and triplet excitation energies.

Successfully captures double excitations beyond TDDFT capabilities.

Operates at a computational cost similar to standard DFT.

Abstract

State-of-the-art methods for calculating neutral excitation energies are typically demanding and limited to single electron-hole pairs and their composite plasmons. Here we introduce excitonic density-functional theory (XDFT) a computationally light, generally applicable, first-principles technique for calculating neutral excitations based on generalized constrained DFT. In order to simulate an M-particle excited state of an N-electron system, XDFT automatically optimizes a constraining potential to confine N-M electrons within the ground-state Kohn-Sham valence subspace. We demonstrate the efficacy of XDFT by calculating the lowest single-particle singlet and triplet excitation energies of the well-known Thiel molecular test set, with results which are in excellent agreement with time-dependent DFT. Furthermore, going beyond the capability of adiabatic time-dependent DFT, we show that XDFT can successfully capture double excitations. Overall our method makes optical gaps, excition bindings and oscillator strengths readily accessible at a computational cost comparable to that of standard DFT. As such, XDFT appears as an ideal candidate to work within high-throughput discovery frameworks and within linear-scaling methods for large systems.