Ionic crystals with GFN-xTB

[mpol]

Dear CP2k developers and users,

Have anyone encountered a problem with GFN-xTB when modeling ionic crystals? If I set

CHECK_ATOMIC_CHARGES .TRUE.

When performing CELL_OPT of a CaF_2 supercell, the program stops with the following warning:

*** WARNING in xtb_matrices.F:1088 :: Atomic charges outside chemical   ***

 *** range were detected. Switch-off CHECK_ATOMIC_CHARGES keyword in the ***

 *** &xTB section if you want to force to continue the calculation.      ***

Setting CHECK_ATOMIC_CHARGES .FALSE. makes the calculation divergent. As I get it, the charges on Ca and F become more and more unphysical in every SCF procedure, and, concomitantly, the total energy becomes lower and lower. I have attached the input and output of the divergent calculation to this message.

Could anyone give a hint on what to do in this situation? I am interested in MD of solid-liquid interfaces with ionic crystals. GFN-xTB seems the best choice.

With best regards,

Mikhail

[hut...@chem.uzh.ch]

Hi

we also found that xTB is rather unstable for some systems/atom types.
We think this is mainly due the 3rd order Coulomb term and the
value of gamma (its proportionality factor).
Currently, I don't have an easy solution ready. In your case I would
start with some test calculations on the simple molecules, e.g. CaF2.
You can then try to re-parameterize the atoms (or just the metal?).
You can add the parameters to the input file (see examples in the regtests)
for easy changes.

regards

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

-----cp...@googlegroups.com wrote: -----
To: "cp2k" <cp...@googlegroups.com>
From: "mpol"
Sent by: cp...@googlegroups.com
Date: 07/04/2019 10:57AM
Subject: [CP2K:11946] Ionic crystals with GFN-xTB

--
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.
To post to this group, send email to cp...@googlegroups.com.
Visit this group at https://groups.google.com/group/cp2k.
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/2e624605-fb53-4994-aa7b-b5eec0d12eba%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.


[attachment "test1.inp" removed by Jürg Hutter/at/UZH]
[attachment "opt.cif" removed by Jürg Hutter/at/UZH]
[attachment "opt.xyz" removed by Jürg Hutter/at/UZH]
[attachment "test1.out" removed by Jürg Hutter/at/UZH]

[mpol]

Dear Juerg,

Thanks for the reply. Notably, I had no such problem with small clusters and molecules.

Unfortunately, I have to pass on the idea to use GFN-xTB for now, in this case.

With best regards,

Mikhail

четверг, 4 июля 2019 г., 19:01:42 UTC+3 пользователь jgh написал:

 To unsubscribe from this group and stop receiving emails from it, send an email to cp...@googlegroups.com.

[Maxim Papusha]

Hello everyone,

i had the same diffuculties with xtb in my calculations.

I was not able to fully solve the problem but i found some tricks to reduce this effect.

The following keywords were helpful in my case:

&DFT
&SCF
SCF_GUESS MOPAC
ADDED_MOS 1000
&SMEAR
ELECTRONIC_TEMPERATURE 5000
&END SMEAR

I would assume xtb is more sensetive to the starting occupation than other methods.

In my experience the first few iteration steps are crucial and afterwards an reasenoble guess can be made based on the previous iterations.

With best regards,

Maxim

[qingh...@gmail.com]

Hello,

In case someone needs it. I also have an alternative one.

First, I follow the suggestion, turn off the charge checking, run some steps of geometry optimization,

in which a new wavefunction file is generated.

         &xTB
               CHECK_ATOMIC_CHARGES .FALSE.
         &END xTB

Use the newly generated wavefunction file, and turn on the charge checking, start the simulation again. At least it worked for me.

Maxim was right, xTB is sensitive to the starting occupation.

All the best,

Qinghua