### shielding vibrational average

Posted:

**18 Sep 2014, 08:01**I have been running the vibrational averaging of shielding constants, as described in the manual, for N-Me-Acetamide. The vibrational averaging at the effective geometry (i.e. non-stationary point) produces a NaN for the last frequency, which (presumably) leads to NaN's in all the vibrational averaged shieldings.

I have not been able to produce the error with smaller systems (FH, H2O, CH4). Basis sets of 3-21G or larger gives the error, STO-3G does not. The DFT-KT3 method produces the error, but DFT-BLYP or HF do not. Increasing or decreasing the numerical stepsize by a factor of two has no effect. Running in parallel or on one-core (attached files) makes no difference. Running a VIBANA works fine (but of course gives a number of imaginary frequencies).

A side comment: when the averaging works, one more energy calculation is performed after the averaging, and this gives a much lower energy than those in the numerical differentiation, and also much lower than the energy at the optimized geometry. What is this energy and how can it be a couple of Hartree's lower than the optimized energy?

I have not been able to produce the error with smaller systems (FH, H2O, CH4). Basis sets of 3-21G or larger gives the error, STO-3G does not. The DFT-KT3 method produces the error, but DFT-BLYP or HF do not. Increasing or decreasing the numerical stepsize by a factor of two has no effect. Running in parallel or on one-core (attached files) makes no difference. Running a VIBANA works fine (but of course gives a number of imaginary frequencies).

A side comment: when the averaging works, one more energy calculation is performed after the averaging, and this gives a much lower energy than those in the numerical differentiation, and also much lower than the energy at the optimized geometry. What is this energy and how can it be a couple of Hartree's lower than the optimized energy?