Benchmarking SOPPA-based methods for the calculation of static and dynamic polarizabilities

Abstract

Static and frequency-dependent polarizabilities were computed for 41 molecules using RPA, RPA(D), HRPA, HRPA(D), SOPPA, SOPPA(CC2), and SOPPA(CCSD) with the aug-cc-pVTZ basis set and benchmarked against CCSD reference values and available experimental data. The analysis reveals a pronounced distinction between the performance of these methods for aromatic versus non-aromatic molecules. Across all frequencies, HRPA consistently yields substantially larger deviations from CCSD than the other approaches, whereas HRPA(D) and SOPPA(CCSD) provide the most accurate results overall. For static polarizabilities, HRPA(D) performs best for non-aromatic systems, followed by SOPPA(CCSD) and RPA(D), while SOPPA(CCSD) is most accurate for aromatic molecules. In the frequency-dependent regime, HRPA(D) remains the most accurate method for non-aromatic molecules, although RPA(D) shows greater consistency. For aromatic molecules, SOPPA(CCSD) performs best at low frequencies, with RPA offering intermediate accuracy but higher consistency than most other methods; at higher frequencies, RPA becomes the most accurate approach, followed by RPA(D), while SOPPA(CCSD) deteriorates. These trends highlight the importance of doubles corrections in RPA(D) and HRPA(D), which achieve accuracy comparable to or better than SOPPA(CCSD) at lower computational cost. The strong performance of RPA for aromatic molecules is attributed to its characteristic overestimation of the lowest electronic excitation energy. Comparison with experimental data confirms SOPPA(CCSD) as the most reliable method for static polarizabilities, while RPA and HRPA(D) provide the best agreement for frequency-dependent polarizabilities of aromatic systems.