CCSD(T) Harmonic and Anharmonic Frequencies

Questions
#1

Hello everyone,

I was wondering if someone would help me better understand how CCSD(T) frequencies are computed using Q-Chem 6.2. It is my understanding that while CCSD gradients are implemented, CCSD(T) gradients are not. I’m curious to get a better understanding as to how exactly the numeric gradients, hessians, and third/fourth derivatives are calculated in a CCSD(T) VPT2 calculation, as well as a harmonic calculation. So, specifically, which gradients are computed how (i.e. is the entire gradient computed numerically, or is the CCSD gradient computed numerically and the (T) tacked on numerically), is a 3-point or 5-point approach used to obtain numeric derivatives, etc. Any help on better understanding this would be much appreciated!

#2

You are correct that CCSD(T) gradients are not implemented in Q-Chem. Thus, the gradient is evaluated automatically by finite-difference of CCSD(T) energies. (There is no way to just “tack on” the triples correction.) Harmonic frequencies, if requested, would be evaluated via double finite-difference of energies. This will be catastrophically expensive except for very small molecules and basis sets. The 3rd- and 4th-order derivatives needed for VPT2 would also need to be evaluated by finite-difference.