Performing MRCC Calculation with Brueckner orbitals

1

Is it possible to save the Brueckner orbitals for use in a subsequent MRCC calculation?

2

First, disclaimer that I haven’t done any numerical tests of this myself.

That said, it should work. Get a wfn object from a Brueckner calculation, and pass that as the ref_wfn keyword argument to energy('mrcc'). The Brueckner orbitals are saved to the wavefunction object, and Psi uses the wavefunction orbitals to transform the integrals as part of the MRCC procedure.

3

Thanks! I’ll try doing that.

4

This was my attempt to implement what you suggested:
set {
reference rohf
basis aug-cc-pVTZ
print_MOs true
print 2
scf_type pk
guess sad
freeze_core true
}
benergy, b_wfn = energy(‘bccd(t)’,return_wfn = True)
energy(‘mrccsdt’, ref_wfn=b_wfn)

but after the bccd(t) calculation, it crashes with:

Traceback (most recent call last):
File “/home/kipeters/psi4/build/stage/bin/psi4”, line 333, in
exec(content)
File ““, line 34, in
File “/home/kipeters/psi4/build/stage/lib/psi4/driver/driver.py”, line 597, in energy
wfn = procedures[‘energy’][lowername](lowername, molecule=molecule, **kwargs)
File “/home/kipeters/psi4/build/stage/lib/psi4/driver/procrouting/proc.py”, line 4563, in run_mrcc
vscf = core.variable(‘SCF TOTAL ENERGY’)
File “/home/kipeters/psi4/build/stage/lib/psi4/driver/p4util/python_helpers.py”, line 859, in _core_variable
raise KeyError(“psi4.core.variable: Requested variable ” + key + ” was not set!\n”)
KeyError: ‘psi4.core.variable: Requested variable SCF TOTAL ENERGY was not set!\n’
Printing out the relevant lines from the Psithon → Python processed input file:
core.set_global_option(“PRINT”, 2)
core.set_global_option(“SCF_TYPE”, “pk”)
core.set_global_option(“GUESS”, “sad”)
core.set_global_option(“FREEZE_CORE”, “true”)
benergy, b_wfn = energy(‘bccd(t)’,return_wfn = True)
→ energy(‘mrccsdt’, ref_wfn=b_wfn)
!---------------------------------------------------------------------------!
! !
! ‘psi4.core.variable: Requested variable SCF TOTAL ENERGY was not set!\n’ !
! !
!---------------------------------------------------------------------------!

5

Thanks for the bug report. I’ve filed an issue. This looks like a simple one to fix, but no promises.

6

The bug has been fixed on the developer version of Psi4. If you want this feature, you’ll need to use that.

All I can promise is that Psi will get to the line of code where control hands over to MRCC, and I can triage any further issues you find. I don’t have the time to do the in-depth investigation I would need to say that this feature is working as intended.

7

Thank you so much! I’ll be trying it out.

8

Hi Jonathon,

looks like it now gets into preparing the input for MRCC just fine but dies with a segmentation fault when it starts the integral transformation.

-Kirk

9

Ah, thank you for the report. I naively assumed that making my minimal RHF example work would be enough, but there’s an additional ROHF-specific bug I wasn’t aware of.

My initial debugging suggests the problem is with how Psi’s CC code handles the wavefunction for Brueckner CC on top of an ROHF reference: RHF-Brueckner and ROHF-SCF both work fine when I pass them into MRCC. @crawdad may be able to either tap somebody to investigate this one or offer wisdom?