6
$\begingroup$

I am trying to calculate the stability of CO at HF/3-21g level using G09. I am trying to reproduce an issue that was observed in PySCF, and I have made the following observations:

  1. If I use the default arguments stable=opt hf/3-21g the calculations would work normally, and the wavefunction would converge to a UHF state, since the code finds an RHF->UHF instability.
  2. However, If I try to use stable=(opt,fc) hf/3-21g the code will crash with an error message
 Operation on file out of range.
 FileIO: IOper= 2 IFilNo(1)=  -554 Len=          77 IPos=           0 Q=   23003655819672

I need to use a frozen core to compare my calculations from G09 and PySCF. Any pointers on why this is happening would be much appreciated.

$\endgroup$
2
  • 1
    $\begingroup$ Is stable=(opt,fc) a valid option? I don't see fc as a possible option to add to stable on the Gaussian website: gaussian.com/stable. $\endgroup$
    – Tyberius
    Jun 30, 2023 at 13:18
  • 1
    $\begingroup$ Ideally the error message would be better. My answer tries to convey that "frozen core" doesn't make sense for HF. $\endgroup$ Jun 30, 2023 at 13:46

3 Answers 3

7
$\begingroup$

A "frozen core" is used to freeze "core" electrons from correlating.

There's no correlation in HF calculations, so it would make sense for a frozen core to only have a visible effect on post-HF calculations, which the code that you've shown doesn't specify at all.

Apart from that, there seems to be an index out of bounds which suggests a bug in G09 or an error in your input. Maybe someone with G16 can see if this bug persists there.

$\endgroup$
5
$\begingroup$

I think, the answer of @NikeDattani is clear enough a frozen core option should not be used for SCF methods as HF and even for DFT except for Double-Hybrid methods : MP Methods. Therefore your calculation would not work even if your input was rightly formulated. In fact, for performing frozen core calculations your route section might be like this : #p stable=opt MP(fc,semidirect)/3-21g ...

This will certainly output an error if your method is not a Post-SCF method. Even if you code works in G16, the option "fc" is ignored. I cannot tell more about this option because I am not used to perform this type of calculation.

$\endgroup$
1
$\begingroup$

To amend the previous answers, there is no reason a priori why frozen core stability checks would not make sense. My intuition would be that the deep lying orbitals can be safely ignored in stability analyses, since the energy penalty of not occupying them is so high.

However, stability checks are typically carried out with all orbitals, since the frozen core approximation would not lead to significant savings in the computational time: you only have o*v rotations, which is similar to CI singles theory.

$\endgroup$

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .