Hello,
I have a question if the calculation of second derivatives are available for DC-DFT calculations and also for MBD dispersion. I am trying to explore this analysis to get force constant but I got a bit lost on the manual and did not understand very much if it’s available or not.
Thanks in advance
Analytic first derivatives for DC-DFT are available although I don’t think that the coupled-perturbed equations are parallelized. Analytic 2nd derivatives are not available but frequencies could be computed by finite-difference of gradients.
Thanks for the reply. So if I want to calculate the force constant for ex of an DC-DFT + MBD dispersion it would be possible? Because for the functional that I want to run (r2scan) the combination of HF-r2scan + MBD is not available, than I don’t know how I would do to obtain the force constant.
Q-Chem will always invoke numerical frequencies for JOBTYPE = FREQ, if the analytic Hessian is not available. For DC-DFT this would be finite difference of analytic gradients but you can also try double finite difference of energies by also setting IDERIV=0 in $rem. If there’s a specific combination of features that fails, please post it here.
Okay I think I understood the part of the DC-DFT. But regarding MBD, it would be also possible to compute the second derivative of MBD? because I already tested that the combination of HF-R2SCAN + MBD does not work so for example to compute forces I was calculating the gradient of the hf-r2scan and hf-mbd separately and than summing the two of them. So I don’t know if this approach would work with the second derivative for MBD.
Set IDERIV=0 (with JOBTYPE=FREQ) to force frequencies by double finite-difference. It’s not cheap, (6*NAtoms)**2 energy calculations, but it should work. In principle this should happen automatically if you set JOBTYPE=FREQ and the appropriate analytic derivatives are not available, but if for some reason that doesn’t happen, you can force it with IDERIV that tells Q-Chem the maximum order of analytic derivative that it is allowed to use.
Okay, So doing a HF (method) setting the job type as freq and Ideriv=0 it will compute also the contribution of the MBD to the derivatives right? should I set the MBD flag (MBDVDW) as MBD_FORCES ?
Why don’t you try it and pay careful attention to what is printed in the output.
Sure, I will do some test. Thanks a lot