Warming up ground-state orbitals: A route to excitation energies?

Linear-response time-dependent density functional theory (TDDFT) is the workhorse of spectral prediction, yet it can struggle to predict or even describe important low-lying excitation energies. An alternative is to use static formalisms to target excited-state energies directly from variational density functional theory (DFT). This may involve using custom excited-state functionals or feeding excited-state orbitals and densities into standard ground-state approximations, and can often be effective, even for describing states for which adiabatic TDDFT fails. However, this comes at the cost of requiring specialized algorithms to find stationary points. In this exploratory work, we demonstrate that finite-temperature DFT can provide an effective auxiliary framework for generating excited-state-like orbitals from ground-state orbitals for modern ensemble and state-specific functionals, such as the GX24 ground-state (G) and excited-state (X) functional. Notably, the necessary fixed-point algorithms are already available and well optimized in many electronic-structure codes.