Mo/ller–Plesset third order calculations with large basis sets

Abstract

The value of the Mo/ller–Plesset third order calculations is examined. An efficient method for the evaluation of the gradient of the MP3 energy is reported, and it has been programmed for both restricted and unrestricted Hartree–Fock wave functions. Large basis set calculations (TZ2P or better) are reported for the optimization of geometries and the determination of harmonic frequencies (which are obtained by finite differences of analytic gradients). The molecules selected are NH2, PH2, AsH2, H2O, NH3, H2CO, HCN, and C2H2. For the closed shell systems, the RMP3 predictions for bond lengths are inferior (≈0.006 Å) to RMP2 predictions (≈0.003 Å) for single bonds, and for multiple bonds the RMP3 bond lengths are too short by approximately the same amount (≈0.01 Å) that RMP2 are too long. For the open shell systems, the UMP3 geometrical parameters show only a marginal improvement over UMP2, except for PH2 where the bond length error is reduced to 0.003 Å. The results for harmonic frequencies show a similar comparison between MP2 and MP3. On the basis of this experience, it appears that large scale calculations at the MP3 level are not recommended; MP2 calculations with a large basis set are much cheaper and provide results with a similar, if not superior, accuracy.