Accelerated Computer-Aided Screening of Optical Materials: Investigating the Potential of Δ‑SCF Methods to Predict Emission Maxima of Large Dye Molecules

Accurate simulation of electronic excited states of large
chromophores
is often difficult due to the computationally expensive nature of
existing methods. Common approximations such as fragmentation methods
that are routinely applied to ground-state calculations of large molecules
are not easily applicable to excited states due to the delocalized
nature of electronic excitations in most practical chromophores. Thus,
special techniques specific to excited states are needed. Δ-SCF
methods are one such approximation that treats excited states in a
manner analogous to that for ground-state calculations, accelerating
the simulation of excited states. In this work, we employed the popular
initial maximum overlap method (IMOM) to avoid the variational collapse
of the electronic excited state orbitals to the ground state. We demonstrate
that it is possible to obtain emission energies from the first singlet
(S1) excited state of many thousands of dye molecules without
any external intervention. Spin correction was found to be necessary
to obtain accurate excitation and emission energies. Using thousands
of dye-like chromophores and various solvents (12,318 combinations),
we show that the spin-corrected initial maximum overlap method accurately
predicts emission maxima with a mean absolute error of only 0.27 eV.
We further improved the predictive accuracy using linear fit-based
corrections from individual dye classes to achieve an impressive performance
of 0.17 eV. Additionally, we demonstrate that IMOM spin density can
be used to identify the dye class of chromophores, enabling improved
prediction accuracy for complex dye molecules, such as dyads (chromophores
containing moieties from two different dye classes). Finally, the
convergence behavior of IMOM excited state SCF calculations is analyzed
briefly to identify the chemical space, where IMOM is more likely
to fail.