DFTD3 scaling - possible bug

[Bram Van der Linden]

Dear,

According to the manual, the D3_SCALING is 0 by default, which will make cp2k determine s6,sr6,s8 automatically. However, if you do specify this parameter as D3_SCALING     0.0000000000000000E+00    0.0000000000000000E+00    0.0000000000000000E+00, zeroes are taken for all coefficients. 

With kind regards,

Bram

[Jürg Hutter]

Hi

the situation is a bit different in this case. CP2K takes always the parameters defined in the input (even if they are 0).
If no explicit input is given, it takes method dependent default values.

Automatic parameter determination is triggered by 'no values given in the input' and not 0 values.

regards
JH

________________________________________
From: cp...@googlegroups.com <cp...@googlegroups.com> on behalf of Bram Van der Linden <bramvdl...@gmail.com>
Sent: Monday, February 2, 2026 7:33 PM
To: cp2k
Subject: [CP2K:22077] DFTD3 scaling - possible bug

--
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 visit https://groups.google.com/d/msgid/cp2k/198e83c0-d69f-422e-ae14-bcae3385d5c8n%40googlegroups.com<https://groups.google.com/d/msgid/cp2k/198e83c0-d69f-422e-ae14-bcae3385d5c8n%40googlegroups.com?utm_medium=email&utm_source=footer>.

[Bram Van der Linden]

Dear Jurg,

Thanks for the clarification. I think the manual may be confusing here: " if set to zero CP2K attempts to guess the xc functional from REFERENCE_FUNCTIONAL". So, I interpreted this as that you can set the coefficients safely to 0. It may be better to replace "if set to zero" by "if the parameter is omitted".

With kind regards,

Bram

Op dinsdag 3 februari 2026 om 09:40:27 UTC+1 schreef Jürg Hutter: