Article ID | Journal | Published Year | Pages | File Type |
---|---|---|---|---|
5372993 | Chemical Physics | 2016 | 9 Pages |
â¢Determination of a suitable reference for high-level benchmark studies.â¢Assessment of the local errors due to the incremental scheme.â¢Computation of highly accurate CCSD(T)/CBS benchmark values for binding energies.â¢Assessment of different state-of-the-art methods for binding energies.
In this work we test different strategies to compute high-level benchmark energies for medium-sized molecular clusters. We use the incremental scheme to obtain CCSD(T)/CBS energies for our test set and carefully validate the accuracy for binding energies by statistical measures. The local errors of the incremental scheme are <1Â kJ/mol. Since they are smaller than the basis set errors, we obtain higher total accuracy due to the applicability of larger basis sets. The final CCSD(T)/CBS benchmark values are ÎE=-278.01Â kJ/mol for (H2O)10,ÎE=-221.64Â kJ/mol for (HF)10,ÎE=-45.63Â kJ/mol for (CH4)10,ÎE=-19.52Â kJ/mol for (H2)20 and ÎE=-7.38Â kJ/mol for (H2)10.Furthermore we test state-of-the-art wave-function-based and DFT methods. Our benchmark data will be very useful for critical validations of new methods.We find focal-point-methods for estimating CCSD(T)/CBS energies to be highly accurate and efficient. For foQ-i3CCSD(T)-MP2/TZ we get a mean error of 0.34Â kJ/mol and a standard deviation of 0.39Â kJ/mol.
Graphical abstractDownload high-res image (82KB)Download full-size image