open shell atoms with X2Cmmf + CCSD(T) ?

[Peterson, Kirk A]

Dear Dirac experts,

 

Unless I’m doing something very wrong, using X2Cmmf  with CCSD(T) for open shell atoms seems to yield incorrect results without seemingly any hints in the output that things are going wrong (except for the resulting energies). I attach an example for the neutral Lr atom which has a single open-shell p1/2 spinor.  As noted in the output for RELCCSD, the SCF spinor eigenvalues are used and the Fock matrix elements are not recomputed. I believe the latter might be the problem. Based on previous conversations on this list, I thought the latter was only a problem when the HOMO-LUMO gap ended up changing sign, but then the program dies with an error. As you’ll note in the attached output, everything looks well-behaved (to me) except the (T) contribution is over -300 mEh when it should be about -50 mEh (based on 4c calculations). The CCSD correlation energy is also about 40 mEh lower than it should be. But perhaps I’ve done something wrong?

 

Since X2Cmmf appears to be the only way to get correlated results in Dirac for the effects of Gaunt or QED, it would be of great interest to know if this is indeed possible or not for open-shell systems. I didn’t find anything in the manual that said it couldn’t but….

 

Best regards,  -Kirk

[Miroslav Iliaš]

Hello Kirk,

your case attracted my interest,  my few cents:

x2cmmf without recomputing Fock matrix elements is not recommended for open-shell systems (https://diracprogram.org/doc/master/manual/wave_function/relccsd.html#norecmp) , it alters RelCC energies. 

But it should be judged based on concrete computed property(ies) rather than on differences between "pure" and altered (X2cmmf) CC runs.

Replacement of the X2cmmf Hamiltonian is the X2c+Gaunt, see 

 https://github.com/miroi/open-collection/tree/master/theoretical_chemistry/software/dirac/runs/Lr

Miro

--
You received this message because you are subscribed to the Google Groups "dirac-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to dirac-users...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/dirac-users/SJ2PR01MB8052A95C952A2B77C0686815D60FA%40SJ2PR01MB8052.prod.exchangelabs.com.

[peterso...@gmail.com]

Hi Miro,

 

It is my understanding from a previous conversation on this list (Johann P.) that recomputing the Fock matrix element for X2Cmmf is not supported, so it seems this means X2Cmmf should not be used at all for open shell systems, except perhaps with careful benchmarking.  In regards to just X2c+Gaunt, we’ve use that quite a bit but have found that at least for actinides the resulting contribution from Gaunt is not very accurate. Perhaps the new X2c routines would improve this, but unfortunately there are currently problems with using this with CCSD (Trond is working on this).

 

Best regards,  -Kirk

To view this discussion visit https://groups.google.com/d/msgid/dirac-users/CA%2BGrSJwLEqQ%2ByKQ%3Dt8pag%2Br-njZb1e0hQwYKRwEaU-9j47YwYQ%40mail.gmail.com.