OT for added MOS

97 views
Skip to first unread message

Lorenzo Agosta

unread,
May 25, 2023, 7:40:59 AM5/25/23
to cp2k
Dear developers,

I know that is not possible to perform SCF calculations using the OT scheme when extra molecular orbitals are requested.

The problem is that for many not trivial systems the OT scheme is always able to reach a SCF convergence, while the standard DIAGONALIZATION algorithms ( no matter which combinations of parameters are used for the mixing ) fail, even starting from a  fully converged electron density reached with the OT scheme.
Strict diagonalization algorithms  are also required in various subroutines like KPOINT, XAS and GW for example.

Having the possibility to calculate added MOS using the OT scheme would be enormously beneficial in terms of applicability of the CP2K software and a considerable gain in computational resources. 

So, I was wondering if there is a specific reason why extra MO cannot be computed with the OT scheme, or more simply it has not been implemented yet.

Thank you in advance for the help and clarifications.

Kind regards,
Lorenzo Agosta

Jürg Hutter

unread,
May 26, 2023, 4:50:31 AM5/26/23
to cp...@googlegroups.com
Hi

OT is a direct energy optimization method. As the energy doesn't depend on the virtual states, they cannot
be calculated directly within the optimization.
However, you can calculate additional states (LUMO +++) after an OT optimization. Several property options
allow for such calculations (e.g. PRINT/MO_CUBES). Another possibility is to restart an optimized structure
and requesting a single diagonalization step.

regards
JH

________________________________________
From: cp...@googlegroups.com <cp...@googlegroups.com> on behalf of Lorenzo Agosta <pha...@gmail.com>
Sent: Thursday, May 25, 2023 1:40 PM
To: cp2k
Subject: [CP2K:18853] OT for added MOS
--
You received this message because you are subscribed to the Google Groups "cp2k" group.
To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+uns...@googlegroups.com<mailto:cp2k+uns...@googlegroups.com>.
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/1d8bf378-a519-464e-893d-4fdeced35373n%40googlegroups.com<https://groups.google.com/d/msgid/cp2k/1d8bf378-a519-464e-893d-4fdeced35373n%40googlegroups.com?utm_medium=email&utm_source=footer>.

Lorenzo Agosta

unread,
May 29, 2023, 5:54:00 AM5/29/23
to cp2k
Hi Jurg,

Thank you a lot for the nice clarification.
This solved a lot of doubts for me.

Best,
Lorenzo
Reply all
Reply to author
Forward
0 new messages