"Reference energy" with scftype=uks lower than scftype=uhf

More
6 months 3 weeks ago #1064 by benedek1231
Dear Developers,

I would like to know if the "Reference energy" calculated from Kohn-Sham (like scftype=uks dft=b3lyp calc=ccsd) can be directly compared to the HF energy. As far as I am concerned, if one calculates the orbital from KS and creates the reference Slater determinant from them, its energy MUST be higher than the HF energy (because HF is based on the variational principle).

In most of my calculations the statement above is fulfilled, but I found some embarassing examples.
If I run a CC calculation for the BH3 molecule with both scftype=rhf and scftype=rks
- when I use B3LYP functional, I get similar CC energies and the "Reference energy" with RKS is a little bit higher (as expected, so this is fine)
See files rhftest_bh3.out rkstest_b3lyp_bh3.out
- BUT when I use HSE06, both the reference energies and CC energies differ by more than 1 Ha, what is more, the "Reference energy" calculated from HSE06 orbitals is lower. I'm confused because I think this is theoreticvally incorrect.
See files rhftest_bh3.out rkstest_hse06_bh3.out
(I also tested using UHF/UKS and I got the same.)
See files uhftest_bh3.out ukstest_hse06_bh3.out ukstest_b3lyp_bh3.out

Another example:
I also calculated B2 with ROHF/ROKS and the obtained reference energies do not have to do anything with each other (like -29 Ha for ROKS and -48 Ha for ROHF).
See files rohftest_b2.out rokstest_b2.out
However, when I change to UHF/UKS, the energies become similar and - as I would expect - the reference energy from UKS is a little bit higher than UHF.
See files uhftest_b2.out ukstest_b2.out

So, my question is: why is it that in some cases the HF and KS reference energies look comparable, while in other cases not? In theory, when can they be directly compared?

Many thanks in advance,
Zsolt

  








 
Attachments:

Please Log in or Create an account to join the conversation.

More
6 months 3 weeks ago #1065 by kallay
Dear Zsolt,
Thank you for reporting this. Unfortunately, the calculation of the HF energy was not compatible with range-separated functionals.
You should look up the line 
          chfx=-1.d0
in subroutine scf in file scf.f and insert the following two lines after that:
          clrhfx=0.d0
          csrhfx=0.d0
Then, you should recompile the code.

Best regards,
Mihaly Kallay

Please Log in or Create an account to join the conversation.

More
6 months 3 weeks ago #1066 by benedek1231
Thank you for your quick reply, I will modify the code.
Is it also possible to make ROHF and ROKS energies comparable?
(for example for B2, as discussed in my previous post).

Please Log in or Create an account to join the conversation.

Time to create page: 0.019 seconds
Powered by Kunena Forum