## Info

a Data from Kucharski and Bartlett22 b Data from Ljubic and Sabljic23

a Data from Kucharski and Bartlett22 b Data from Ljubic and Sabljic23

The simple picture with ozone as a resonance structure between ionic and biradical forms suggests that a two-configuration wave function should be able to give a qualitatively correct description. The [2,2]-CASSCF and [2,2]-CASPT2 results, however, show that dynamical correlation is also very important. The poor RHF reference wave function is clearly seen by the MPn results, with the MP2 value being in error by a factor of 2 for the asymmetric stretch, and the MP4 result is in error by ~500cm-1 for n2, despite reproducing n1 and n3 to within 30 cm-1. The coupled cluster methods are less sensitive to the quality of the HF wave function, and are in somewhat better agreement with the experimental values. The CCSD(T) results are within ~20cm 1 of the experimental values, but part of this agreement is accidental as seen by the CCSDT and CCSDT(Q) results, and even the CCSDT(Q) model has errors of ~25cm-1. Part of this discrepancy may be due to basis set errors, although the results for the CASPT2 method indicate that larger basis sets will further increase the value of the vibrational frequencies.23 The DFT methods perform well, yielding results comparable to those at the CCSD or CCSD(T) levels, at a fraction of the computational cost. Even the local density functional gives acceptable results, but this is a case where the hybrid DFT methods (B3LYP and PBE0) perform worse than the pure DFT ones. It can be noted that the cc-pVTZ basis set is sufficiently large that the DFT results are essentially converged, and the results in Table 11.22 thus reflect the intrinsic accuracy of the different DFT methods.

## Post a comment