Dear Q-Chem community,
I was wondering if there’s a way to run the gradients necessary for a VPT2 computation (or any other method for computing anharmonic frequencies) as separate jobs on a computing cluster with several nodes. For example, if I had a VPT2 computation that required around 4000 gradients and had access to around 40 nodes, could I split up the computation so that each node runs ~100 gradients each? If so, is there any restriction on which methods I can do this with (HF, MP2, DFT, etc.)?
I have also been experimenting with running some computations on a single node with an older version of Q-Chem (5.1.2), and was wondering about a couple of details (one of my sample inputs is included at the bottom):
- I noticed that the anharmonic procedure was also computing the frequencies for Transition-Optimized Shifted Hermite Theory, in addition to VPT2. Are these frequencies computed at no additional cost once I’ve collected all the energies/gradients/hessians necessary for VPT2 (if not, is there a way to disable TOSH)?
- I tried changing the FDIFF_DER keyword from 0 to 1, 2, and 3 to see how that changed the anharmonic frequencies when using energies, analytic gradients, analytic hessians, etc. However, I was getting the same anharmonic frequencies regardless of my choice of FDIFF_DER (see below), and the output files appeared to show the same procedures being performed. The example I attached below is a quicker example with water and HF, though I have noticed this issue persist with trying another method (RI-MP2) and with increasing the size of the molecule. I just wanted to know if there’s a bug with this older version that’s been fixed in the most recent release, or if there are some additional keywords I forgot to include to change the finite difference procedure?
TOSH[ 1] 1659.48 VPT2[ 1] 1657.58
TOSH[ 2] 3867.21 VPT2[ 2] 3809.18
TOSH[ 3] 3934.02 VPT2[ 3] 3934.02
Here is a sample input of what I had been running on a single node, in case I am missing any critical keywords:
$molecule
0 1
O -0.049712339387 -0.037671587095 0.000000000000
H -0.025545113489 0.920598188333 0.000000000000
H 0.872386367085 -0.299583639498 0.000000000000
$end
$rem
JOBTYPE opt
METHOD hf
BASIS 6-31G
SYMMETRY false
SCF_CONVERGENCE 11
MEM_TOTAL 25000
$end
@@@
$molecule
READ
$end
$rem
JOBTYPE FREQ
METHOD hf
BASIS 6-31G
scf_guess read
skip_scfman true ! no need to run SCF again
SYMMETRY false
SCF_CONVERGENCE 11
MEM_TOTAL 25000
$end
@@@
$molecule
READ
$end
$rem
JOBTYPE FREQ
METHOD hf
BASIS 6-31G
scf_guess read
skip_scfman false ! if true, doesn’t run SCF for finite difference steps!
skip_drvman 1 ! read in the already-computed harmonic frequencies
SYMMETRY false
ANHAR TRUE
VCI 0
FDIFF_DER 0
SCF_CONVERGENCE 11
MEM_TOTAL 25000
$end
Thanks in advance!