XC_FUNCTIONAL SCAN

[Vibhav Yadav]

Dear CP2K admins/users,

I am trying run a GEO_OPT calculation on InP. with pseudo-hydrogen passivation along-with the bulk being fixed using the SCAN XC_FUNCTIONAL.
I am facing the problem of the SCF cycles not converging, or dropping below 1.0E-2.

Could you let me know what's the problem, or additional criteria I should be incorporating.

I have attached the input file below.
Few of the things that cannot be removed from the input file are:
1.) Pseudo-hydrogen, vacuum and the CONSTRAINT subsection.

Thank you for your time and consideration.

Regards, Vibhav Yadav

Institut für Physiklaische und Theoretische Chemie, 

Universtiat Tübingen

[Marcella Iannuzzi]

Hello Vibhav Yadav

Try a smaller ALPHA for the MIXING, something like 0.005

Remove OUTER_SCF

Increase MAX_SCF

Regards

Marcella

[Jürg Hutter]

Hi

in addition you might want to try the r2SCAN functional
(a numerically more stable version of SCAN).

regards
JH


________________________________________
From: cp...@googlegroups.com <cp...@googlegroups.com> on behalf of Marcella Iannuzzi <marci...@gmail.com>
Sent: Friday, July 19, 2024 2:45 PM
To: cp2k
Subject: [CP2K:20464] Re: XC_FUNCTIONAL SCAN

--
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/d770053a-f093-47e8-9386-c7886d30180en%40googlegroups.com<https://groups.google.com/d/msgid/cp2k/d770053a-f093-47e8-9386-c7886d30180en%40googlegroups.com?utm_medium=email&utm_source=footer>.

[Vibhav Yadav]

Hi, 
I did try the following,
Removing OUTER_SCF, increasing MAX_SCF, with SCAN and included R2SCAN separately. 
Still didn't converge for the former and gave  incorrect XC_FUNCTIONAL error for the later.

Thank you for your time and consideration.

With regards, Vibhav Yadav

Institut für Physiklaische und Theoretische Chemie, 

Universtiat Tübingen

[Frederick Stein]

Hi,
What Jürg meant was to use the following XC_FUNCTIONAL section
&XC_FUNCTIONAL
  &MGGA_X_R2SCAN
  &END
  &MGGA_C_R2SCAN
  &END
&END

Do not pass a keyword to the &XC_FUNCTIONAL section because the R2SCAN functional is not predefined within CP2K. This setup requires support for LibXC (default with toolchain and most center-provided distributions).

If this still does not help, could you please provide your new input file (and potentially also output files) such that we are on the same page as you.

HTH,
Frederick

[Vibhav Yadav]

Hi, I have tried both the methods as stated below:

1.) My first trial was to use SCAN without LIBXC support as you stated above, filenames are inp2, out2. 

out2 is the slurm file not the CP2K out file.
CP2K version would be 7.1.

 The run never initiates since "unkown subsection" is printed in the slurm.out file.

2.) SCAN with LIBXC support, the problem encountered was in the SCF not converging. I have  attached the input/output files with inp1, out1 naming.
Tried both 7.1 and 9.1 CP2K versions for method 2.

Regards, Vibhav Yadav

[Marcella Iannuzzi]

Hi Vibhav Yadav

The problem is not the functional, but the charge -2.

The SCF converges decently setting the system as neutral. 

If charging the system is necessary, you could try some other tricks,

such as initialising the electronic structure differently, in order to localise the electrons somewhere,

or constraining it.

Regards

Marcella

[Vibhav Yadav]

Dear Marcella, 
The calculation still doesn't converge if you remove the CHARGE -2 keyword. 
Charging of the system is essential, otherwise it won't be neutral. So one needs to add a background charge to maintain the NET CHARGE as 0, coz we are terminating the bottom layer, saturating it with pseudo-H
What did you include in your input file for it to converge?

Thank you for your time and consideration

With regards, Vibhav Yadav

[Jürg Hutter]

Hi

I attached an input and output that worked. It is not perfect, but
might be used as a starting point.

regards
JH

________________________________________
From: cp...@googlegroups.com <cp...@googlegroups.com> on behalf of Vibhav Yadav <vibhav...@gmail.com>
Sent: Monday, July 29, 2024 11:05 AM
To: cp2k
Subject: Re: [CP2K:20511] Re: XC_FUNCTIONAL SCAN

To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/2c9928ac-118b-4168-8ee6-117f560dc73fn%40googlegroups.com<https://groups.google.com/d/msgid/cp2k/2c9928ac-118b-4168-8ee6-117f560dc73fn%40googlegroups.com?utm_medium=email&utm_source=footer>.