Cutoff convergence for SCAN functional

157 lượt xem
Chuyển tới thư đầu tiên chưa đọc

Rajorshi Chattopadhyay

chưa đọc,
19:11:52 16 thg 3, 202316/3/23
đến cp...@googlegroups.com
Dear all,

I am trying to do cutoff convergence with the SCAN functional in a aqueous H2O-H2SO4 system. I need an energy convergence of 10**-6 Ha or lower. However, I find that even at a cutoff of 4000Ry (REL_CUTOFF:60), energy convergence is only 10**-4 Ha. The input script is as follows:

&GLOBAL

    PROJECT Sulphate_amb_dens

    RUN_TYPE ENERGY_FORCE

    PRINT_LEVEL MEDIUM

&END GLOBAL

&FORCE_EVAL

&DFT

      MULTIPLICITY 1

      BASIS_SET_FILE_NAME ./BASIS_MOLOPT

      POTENTIAL_FILE_NAME ./SCAN_POTENTIALS

  &MGRID

      CUTOFF 3800

      REL_CUTOFF 60

      NGRIDS 4

  &END MGRID

  &SCF

     SCF_GUESS ATOMIC

     MAX_SCF 3000

     EPS_SCF 1.00000000E-007

     &OT

        MINIMIZER DIIS

        PRECONDITIONER FULL_SINGLE_INVERSE

     &END OT

  &END SCF

  &XC

     &XC_FUNCTIONAL

         &LIBXC

          FUNCTIONAL MGGA_X_SCAN

         &END LIBXC

         &LIBXC

           FUNCTIONAL MGGA_C_SCAN

           &END LIBXC

     &END XC_FUNCTIONAL

    &END XC

   &END DFT

  &SUBSYS

    &TOPOLOGY

     &CENTER_COORDINATES TRUE

     &END CENTER_COORDINATES

    &END TOPOLOGY


    &CELL

       ABC 12.32 12.32 12.32

       PERIODIC XYZ

    &END CELL

    &COORD

      @include ../coord_eql

   &END COORD

   &KIND H

      BASIS_SET DZVP-MOLOPT-SR-GTH

      POTENTIAL GTH-SCAN-q1

   &END

   &KIND O

      BASIS_SET DZVP-MOLOPT-SR-GTH

      POTENTIAL GTH-SCAN-q6 

   &END

  &KIND S

      BASIS_SET DZVP-MOLOPT-GTH

      POTENTIAL GTH-SCAN-q6

   &END


  &END SUBSYS

 &END FORCE_EVAL


I use the pseudopotentials provided by Dr. Joerg Hutter in his github.

Am I missing something in the input script ? Is it possible to achieve such tight energy convergences using SCAN the functional?

I thank you in advance for your help/suggestions. 

--
With Regards,
Rajorshi Chattopadhyay,
PhD Researcher in Mineralogy/Crystallography,
Institut für Geologie und Mineralogie,
Universität zu Köln, Germany

Fabian

chưa đọc,
05:18:15 17 thg 3, 202317/3/23
đến cp2k
Dear Rajorshi,

The original SCAN functional is known to be very sensitive to the integration grid used during calculation. 
An easy fix would be to switch to r2SCAN (acs.jpclett.0c02405) instead: 

      &XC
        &XC_FUNCTIONAL
          &MGGA_X_R2SCAN
          &END
          &MGGA_C_R2SCAN
          &END
        &END
      &END XC

Typically, the cutoff should be converged at 1200Ry with SCAN.

Kind regards,
Fabian

Rajorshi Chattopadhyay

chưa đọc,
07:28:01 17 thg 3, 202317/3/23
đến cp...@googlegroups.com
Hello Fabian,

I thank you for your reply.

When I do cutoff convergence test with r2SCAN, I get the following energy values till 1500Ry (REL_CUTOFF:60):

Cutoff

Rel cutoff

Energy

Charge

600

60

-1125.17821812053535

-0.0000000141

700

60

-1125.17827283731867

-0.0000000147

800

60

-1125.17815235733997

-0.0000000121

900

60

-1125.17818711771633

-0.0000000128

1000

60

-1125.17816274805796

-0.0000000131

1100

60

-1125.17813293059771

-0.0000000132

1200

60

-1125.17820271100982

-0.0000000140

1300

60

-1125.17820271126402

-0.0000000141

1400

60

-1125.17817412922477

-0.0000000128

1500

60

-1125.17817458499053

-0.0000000139


I find that the convergence is only till 10**-3 Ha.


--
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 view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/a68b39f8-10cc-4a1f-a6a4-a3aa8c0893ffn%40googlegroups.com.

Jürg Hutter

chưa đọc,
07:37:50 17 thg 3, 202317/3/23
đến cp...@googlegroups.com
Hi

Maybe:

EPS_DEFAULT 1.E-14

REL_CUTOFF 100

helps.
regards
JH

________________________________________
From: cp...@googlegroups.com <cp...@googlegroups.com> on behalf of Rajorshi Chattopadhyay <rajors...@gmail.com>
Sent: Friday, March 17, 2023 12:27 PM
To: cp...@googlegroups.com
Subject: Re: [CP2K:18549] Re: Cutoff convergence for SCAN functional
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/a68b39f8-10cc-4a1f-a6a4-a3aa8c0893ffn%40googlegroups.com<https://groups.google.com/d/msgid/cp2k/a68b39f8-10cc-4a1f-a6a4-a3aa8c0893ffn%40googlegroups.com?utm_medium=email&utm_source=footer>.


--
With Regards,
Rajorshi Chattopadhyay,
PhD Researcher in Mineralogy/Crystallography,
Institut für Geologie und Mineralogie,
Universität zu Köln, Germany

​

--
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/CACTJSRfhgJHLbZRTFmKnor5O58G%3DNriRQVaHm-2KK_xbuK-gEg%40mail.gmail.com<https://groups.google.com/d/msgid/cp2k/CACTJSRfhgJHLbZRTFmKnor5O58G%3DNriRQVaHm-2KK_xbuK-gEg%40mail.gmail.com?utm_medium=email&utm_source=footer>.

Rajorshi Chattopadhyay

chưa đọc,
15:21:01 17 thg 3, 202317/3/23
đến cp...@googlegroups.com
Hi Prof. Hutter,

I thank you for your reply.

I tried out with the settings as suggested by you. The data is as follows:

Cutoff

Rel cutoff

Energy

Charge

600

100

-1125.17821814826993

-0.0000000138

700

100

-1125.17827294369863

-0.0000000147

800

100

-1125.17815245523525

-0.0000000120

900

100

-1125.17818718982608

-0.0000000115

1000

100

-1125.17816278218288

-0.0000000122

1100

100

-1125.17813191141613

-0.0000000121

1200

100

-1125.17820280116666

-0.0000000130

1300

100

-1125.17820280127080

-0.0000000125

1400

100

-1125.17817423150291

-0.0000000108

1500

100

-1125.17817423131828

-0.0000000112

1600

100

-1125.17813474004834

-0.0000000119


I still find the highest convergence of 10**-3 Ha.

To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+uns...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/ZR0P278MB075931AD020B7A19C3CA23E99FBD9%40ZR0P278MB0759.CHEP278.PROD.OUTLOOK.COM.

Rajorshi Chattopadhyay

chưa đọc,
10:53:44 18 thg 3, 202318/3/23
đến cp...@googlegroups.com
Hi,
I confused EPS_DEFAULT with EPS_SCF. The convergence data in the previous email were done with wrong settings. The data with EPS_DEFAULT 1E-14 and REL_CUTOFF 100 is as follows:

Cutoff

Rel cutoff

Energy

Charge

600

100

-1125.17821665946303

-0.0000000000

700

100

-1125.17827145715273

0.0000000000

800

100

-1125.17815095995729

-0.000000000

900

100

-1125.17818569659994

-0.000000000

1000

100

-1125.17816128538243

-0.000000000

1100

100

-1125.17813041874047

-0.000000000

1200

100

-1125.17820131280564

-0.000000000

1300

100

-1125.17820131281405

-0.000000000

1400

100

-1125.17817273531045

-0.000000000

1500

100

-1125.17817273531455

-0.000000000


I see an improvement in the total charge on the grids with these settings. However, the energy convergence is still 10**-3 Ha.

I apologise for the confusion with the settings.
Trả lời tất cả
Trả lời tác giả
Chuyển tiếp
0 tin nhắn mới