Comprehensive evaluation of medium and long range correlated density functionals in TD-DFT investigation of DNA bases and base pairs: gas phase and water solution study

Abstract

A comprehensive analysis of the performance of the TD-DFT method using different density functionals including recently developed medium and long-range correlation corrected density functionals have been carried out for lower-lying electronic singlet valence transitions of nucleic acid bases and the Watson–Crick base pairs in the gas phase and in the water solution. The standard 6-311++G(d,p) basis set was used. Ground state geometries of bases and base pairs were optimized at the M05-2X/6-311++G(d,p) level. The nature of potential energy surfaces (PES) was ascertained through the harmonic vibrational frequency analysis; all geometries were found to be minima at the respective PES. Electronic singlet vertical transition energies were also computed at the CC2/def2-TZVP level in the gas phase. The effect of state-specific water solvation on TD-DFT computed transition energies was considered using the PCM model. For the isolated bases the performance of the B3LYP functional was generally found to be superior among all functionals, but it measurably fails for charge-transfer states in the base pairs. The CC2/def2-TZVP computed transition energies were also revealed to be inferior compared with B3LYP results for the isolated bases. The performance of the ωB97XD, CAM-B3LYP and BMK functionals were found to be similar and comparable with the CC2 results for the isolated bases. However, for the Watson–Crick adenine–thymine and guanine–cytosine base pairs the performance of the ωB97XD functional was found to be the best among all the studied functionals in the present work in predicting the locally excited transitions as well as charge transfer states.

Keywords:

• nucleic acid bases
• base pairs
• time-dependent density functional theory
• long range correlated density functional
• state-specific salvation

Acknowledgements

The authors are grateful for financial supports from NSF-CREST grant No. HRD-0833178, and NSF EPSCoR grant No. 440900362427-02. The authors thank the Mississippi Center for Supercomputing Research (MCSR) for the generous use of the computational facility.