[CP2K-user] [CP2K:10927] GAPW convergence for Cu and Ni

[hut... at chem.uzh.ch]

Hi

there are a few parameters that determine the numerical accuracy
of a GAPW calculation. We tried to pick good default values for
them, but it is well possible that for certain elements/basis set
combinations they do not work.
HARD_EXP_RADIUS sets the GAPW atomic sphere radius. The smaller
this value the higher the PW cutoff you will need. It shifts
the border between the soft and hard part of the basis set.
The same is also true for EPSFIT. This parameter sets the accuracy
of the fitting of the different densities at the GAPW sphere radius.
For heavy elements EPSISO can also be of importance (determines which
basis functions are completely within the GAPW sphere).

Usually, one can see the sensitivity of a calculation by changing these
parameters and then search for stable values.

best

Juerg Hutter
--------------------------------------------------------------
Juerg Hutter                         Phone : ++41 44 635 4491
Institut für Chemie C                FAX   : ++41 44 635 6838
Universität Zürich                   E-mail: hut... at chem.uzh.ch
Winterthurerstrasse 190
CH-8057 Zürich, Switzerland
---------------------------------------------------------------

-----cp... at googlegroups.com wrote: -----
To: "cp2k" <cp... at googlegroups.com>
From: "Maxime Van den Bossche" 
Sent by: cp... at googlegroups.com
Date: 11/10/2018 04:51PM
Subject: [CP2K:10927] GAPW convergence for Cu and Ni

Dear all,

I would like to use the GAPW method in CP2K to study
(late) transition metal clusters, as it would offer
quite an increase in speed compared to GPW (because
GAPW requires less high plane-wave cutoffs for dealing
with such "harder" pseudopotentials).

I had no problems performing GAPW runs for fifth- and 
sixth-row elements (Pd, Ag, Pt, Au), but for the 
fourth-row ones (Ni, Cu), I ran into severe SCF 
convergence problems (strongly oscillatory behaviour).

I managed to create a minimal example, involving just
the Cu dimer, where this behaviour is also present 
(see attached input and output files). In this particular 
case, the SCF does converge when using the DIRECT_P_MIXING
mixing scheme, but this solution is not applicable to 
my original problem, where I need e.g. Broyden mixing.

I then found that lowering the HARD_EXP_RADIUS parameter
to 1.2 Bohr radii or below (compared to the default value
of 1.512) makes the SCF cycle converge without problems.
I realize that the GAPW mode is still a bit more "experimental"
compared to GPW, so I'm fine with having to tinker with
such parameters, but I wanted to check with you:

Does it indeed make sense to you that the default value for 
HARD_EXP_RADIUS is not appropriate here? Would this e.g. 
be related to the shorter 3d-orbital radius for Cu compared to 
Ag-4d and Au-5d?

Best regards,
Maxime
  
  -- 
 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... at googlegroups.com.
 To post to this group, send email to cp... at googlegroups.com.
 Visit this group at https://groups.google.com/group/cp2k.
 For more options, visit https://groups.google.com/d/optout.