Triplet-Tuning: A Novel Family of Non-Empirical Exchange-Correlation Functionals

In the framework of DFT, the lowest triplet excited state, T$_1$, can be evaluated using multiple formulations, the most straightforward of which are UDFT and TDDFT. Assuming the exact XC functional is applied, UDFT and TDDFT provide identical energies for T$_1$ ($E_{\rm T}$), which is also a constraint that we require our XC functionals to obey. However, this condition is not satisfied by most of the popular XC functionals, leading to inaccurate predictions of low-lying, spectroscopically and photochemically important excited states, such as T$_1$ and S$_1$. Inspired by the optimal tuning strategy for frontier orbital energies [Stein, Kronik, and Baer, {\it J. Am. Chem. Soc.} {\bf 2009}, 131, 2818], we proposed a novel and non-empirical prescription of constructing an XC functional in which the agreement between UDFT and TDDFT in $E_{\rm T}$ is strictly enforced. Referred to as "triplet tuning", our procedure allows us to formulate the XC functional on a case-by-case basis using the molecular structure as the exclusive input, without fitting to any experimental data. The first triplet tuned XC functional, TT-$ω$PBEh, is formulated as a long-range-corrected hybrid of PBE and HF functionals [Rohrdanz, Martins, and Herbert, {\it J. Chem. Phys.} {\bf 2009}, 130, 054112] and tested on four sets of large organic molecules. Compared to existing functionals, TT-$ω$PBEh manages to provide more accurate predictions for key spectroscopic and photochemical observables, including but not limited to $E_{\rm T}$, $E_{\rm S}$, $ΔE_{\rm ST}$, and $I$, as it adjusts the effective electron-hole interactions to arrive at the correct excitation energies. This promising triplet tuning scheme can be applied to a broad range of systems that were notorious in DFT for being extremely challenging.