Explicit solvation modeling in the accurate TD-DFT prediction of absorption spectra for natural nucleobases and fluorescent nucleobase analogues

Computational modeling of the absorption and emission properties of organic fluorophores is very useful for understanding their photophysics and improving fluorophore design, but it is challenging to do accurately. Methodologies based on density functional theory have been tested for accuracy in the quantitative simulation of experimental absorbance spectra of the longest wavelength π→π^* and n→π^* transitions for several natural nucleobases and nucleobase analogues to arrive at a tractable approach for relatively reliable prediction of excitation wavelengths in new analogues. Good performance was obtained for TD-B3LYP/aug-cc-pVDZ//B3LYP-D3(BJ)/cc-pVDZ calculations, with SMD implicit solvation plus explicit solvent at likely H-bonding sites. For spectra in aqueous solution, the explicit waters at sites with substantial difference in HOMO/LUMO shift the calculated absorption wavelengths by as much as 20 nm. This methodology yielded a root mean square error in absorbance λ_max values of 10 nm with a maximum absolute error of 16 nm for 16 transitions in aqueous solutions of 14 molecules (including the natural nucleobase monomers) in the range 219-442 nm; errors are even smaller for a more limited set of transitions in 1,4-dioxane. Qualitative trends in absorbance intensities are faithfully modeled, although the root mean square relative error in the absolute intensities is 76%. Although limited to non-charge-transfer transitions, the methodology is easily implemented and successfully describes the long-wavelength absorption spectra of many fluorescent nucleobase analogues such as tricyclic cytosines featuring substituents with a wide range of Hammett parameters.