A benchmark of anisotropic polarizabilities has been carried out for 14 (hetero)-
aromatic molecules using the methods: RPA, RPA(D), HRPA, HRPA(D), SOPPA,
SOPPA(CC2), SOPPA(CCSD), CC2, CCSD and CC3. While this benchmark, to a
large extend, shows similar tendencies as found for isotropic polarizabilities, it also reveals some differences between isotropic and anisotropic polarizabilities. CCSD is found to be the method performing closest to CC3 as it also was for isotropic polarizabilities. For static anisotropic polarizabilities SOPPA(CCSD) performs incredibly close to CCSD, however, the less demanding HRPA(D) follows shortly after in precision. For dynamic anisotropic polarizabilities SOPPA(CCSD) is again the method least deviating from CC3, beside CCSD, but its standard deviation is worse than for RPA, which gives results only slightly more deviating from the CC3 results than SOPPA(CCSD). While the HRPA model is seen to perform incomparably worse than any of the other methods, the simpler RPA is on the other hand thus performing notably well. The finding of this good performance of the relatively simpler and cheaper methods, RPA and HRPA(D), permits calculation of much larger systems without sacrificing the quality of the calculation.