Benchmarking Computational Approaches to Predict the UV-Vis Spectra of Organic Photocatalysts

Abstract

With this work, we wish to facilitate further developments in photocatalysis by proposing reliable methods for the computational prescreening of potential photocatalysts. To this end we introduce a new benchmark strategy and apply it to evaluate the predictions given by two wavefunction and several density functional theory (DFT) methods for the UV-vis absorption spectra of recently developed organic photocatalyst molecules. The novelty in our benchmark framework is that it focuses on evaluating the real-world applicability of computational methods and does not penalize errors that do not contribute to spectral shapes. We employ a spectral fitting process where the calculated excitations are convoluted with Gaussians using two parameters for broadening and wavelength scaling. This way, most methods can sufficiently reproduce the experimental spectra but they differ in how much adjustment they require from the parameters. B2PLYP and M06 are the best functionals that offer highest predictive power as they require practically no scaling. B2PLYP is exceptionally good in estimating the excitation energies with almost 90% of the fitted spectra falling into the ±10% scaling window. This is the same level of accuracy as what the more demanding STEOM-DLPNO-CCSD method provides. M06 compensates its slightly less consistent performance with its lower computational demand and availability in nearly all computational codes. Therefore, we recommend the use of these three methods for developing novel photocatalysts and we also highlight that the M06 functional can be used as a black-box method even by those who are non-experts in computational chemistry. The developed protocol and a user-friendly notebook to assist the analysis are available on github.