Do frequencies dependent on the dertype setting?

While trying to reproduce some gaussian calculations with psi4, I noticed that the frequecies dependent on the dertype setting for the version of psi4 I am using:

$ rpm -qif /usr/bin/psi4 | egrep '^Version|^Source'
Version : 1.1
Source RPM : psi4-1.1-2.add49b9git.fc26.src.rpm

Is this as expected? I whould expect the frequencies to be more or less the same. The input file I used is:

molecule h2o {
units angstrom
H  0.000000  0.760300  -0.196100
O  0.000000  0.000000  0.392200
H  0.000000  -0.760300  -0.196100

set    basis 6-311G**
set e_convergence 9
set g_convergence gau_verytight
set scf_type pk

optimize('scf', molecule=h2o)

scf_e, scf_wfn = frequencies('scf', dertype='energy', return_wfn=True)
scf_e, scf_wfn = frequencies('scf', dertype='gradient', return_wfn=True)
scf_e, scf_wfn = frequencies('scf', dertype='hessian', return_wfn=True)

And the generated freqs:

$ grep -A 3 'Vib. Freq' 'hf-6-311Gstarstar-freq.out'
  No.    Vib. Freq. [cm^-1]      Vib. Temp. [K]
    1              1750.943            2519.214
    2              4142.104            5959.557
    3              4237.400            6096.666
  No.    Vib. Freq. [cm^-1]      Vib. Temp. [K]
    1              1750.925            2519.188
    2              4142.111            5959.567
    3              4237.427            6096.704
  No.    Vib. Freq. [cm^-1]      Vib. Temp. [K]
    1              1731.766            2491.623
    2              4136.243            5951.124
    3              4232.474            6089.579

The freqs generated by gaussian are:

Frequencies -- 1750.9454 4142.1029 4237.3716

If someone could clarify this, that would be appreciated.

Hmm, I wouldn’t expect the exact Hessian to be that different. Can @andysim take a look?

Yeah, this looks like something I screwed up somewhere. I won’t be able to get to it for a few days at least, but will definitely poke around and figure out what’s wrong. Thanks for the bug report.

Thanks both. Should I create an issue ticket on github?

That would help a bit with tracking, thanks!

I have created: