[CP2K-user] [CP2K:22237] Re: Why is DZVP-MOLOPT-GTH much slower than DZVP-MOLOPT-SR-GTH in CP2K?

[Jürg Hutter]

Hi
as a rule of thumb, the costs for integrals goes with

(1/a)^3/2. where a is the value of the most diffuse exponent in the basis.

For hybrid functionals this becomes (1/a)^3, making it clear why the ADMM method is so much faster.

regards
JH

________________________________________
From: cp2k at googlegroups.com <cp2k at googlegroups.com> on behalf of Marcella Iannuzzi <marci.akira at gmail.com>
Sent: Sunday, May 17, 2026 4:31 PM
To: cp2k
Subject: [CP2K:22236] Re: Why is DZVP-MOLOPT-GTH much slower than DZVP-MOLOPT-SR-GTH in CP2K?


Hi

The output reports the total number of atomic orbitals, i.e., the size of the matrix that is  diagonalised. If the calculation is dominated by the diagonalization step, this size has a strong impact on the timings.
And yes, as you can check from the BASIS_MOLOPT library, the short range  basis set is  less diffuse
C DZVP-MOLOPT-GTH DZVP-MOLOPT-GTH-q4
1

 2 0 2 7 2 2 1

      6.132624767898 -0.105576563700  0.024850587600  0.035098108400 -0.056712999000  0.037469425800

      2.625196064782 -0.174866621100  0.147440856000  0.114197930900 -0.076743951500  0.071047574800

      1.045456957247  0.064464981400 -0.421418379200  0.215908137300 -0.528056757000  0.619330561200

      0.478316330874  0.830447035300  0.823051336800  0.355929151000  0.394594045700  0.710730857000

      0.178617414302  0.371911373800  0.184880712100  0.271541842000  0.721247475800  0.389148235000

      0.075144725465  0.006174464000 -0.742086051600  0.056256557700  0.733902531700  0.346374124100

      0.030286753006  0.008966477300  0.083382192000  0.004998059700  0.248172175500 -0.005142430900



 C DZVP-MOLOPT-SR-GTH DZVP-MOLOPT-SR-GTH-q4

 1

 2 0 2 5 2 2 1

      5.605330751705  0.111532881900  0.466507483200 -0.064868978500 -0.055475693400  0.029952774600

      2.113016390533  0.153142194100  0.144113212800 -0.200470255700 -0.216165175200  0.070784458800

      0.769911454810 -0.321396879400  0.029364880900 -0.413141730100 -0.436624272300  0.674601058600

      0.348157086385 -0.610926421600  0.929470388400 -0.415337139200  0.784000698400  0.173652061800

      0.128212254710 -0.133990862800  1.066914331200 -0.288527673600  0.987908135000  1.036317024700

Regards
Marcella
On Saturday, May 16, 2026 at 5:01:11 PM UTC+2 Donovan K wrote:

Hi all,

I noticed that DZVP-MOLOPT-GTH is significantly slower than DZVP-MOLOPT-SR-GTH in my CP2K calculations, even though the setup is otherwise identical.


=========================================
 CP2K| version string:                                       CP2K version 2025.1
 CP2K| source code revision number:                                  git:9635df4
 CP2K| cp2kflags: omp libint fftw3 libxc libgrpp pexsi elpa parallel scalapack c
 CP2K|            osma quip xsmm plumed2 spglib libdftd4 sirius libvori libbqb l
 CP2K|            libtorch libvdwxc hdf5 trexio
 CP2K| is freely available from                            https://www.cp2k.org/
 CP2K| Program compiled at                          Tue Jan 14 12:16:35 UTC 2025
 CP2K| Program compiled on                                          7282285e16d1
 CP2K| Program compiled for                                                local
 CP2K| Data directory path                                        /opt/cp2k/data
 CP2K| Input file name                                      xtal-molopt-sr-9.inp

=========================================

The only difference is the presence of SR.

Example:

   &KIND C
       ELEMENT C
       BASIS_SET DZVP-MOLOPT-GTH
       POTENTIAL GTH-PBE-q4
  &END KIND

vs

   &KIND C
       ELEMENT C
       BASIS_SET DZVP-MOLOPT-SR-GTH
       POTENTIAL GTH-PBE-q4
  &END KIND

=================

&DFT
  BASIS_SET_FILE_NAME BASIS_MOLOPT
  POTENTIAL_FILE_NAME POTENTIAL

  &QS
    EPS_DEFAULT 1.0E-10
  &END QS

  &MGRID
    CUTOFF 500
    REL_CUTOFF 50
  &END MGRID

  &XC
  &XC_FUNCTIONAL PBE
  &END XC_FUNCTIONAL

&END XC

&SCF
  SCF_GUESS ATOMIC
  EPS_SCF 1.0E-7
  MAX_SCF 300
&END SCF

&END DFT

=================


Computing time:

1. DZVP-MOLOPT-GTH
  Step     Update method      Time    Convergence         Total energy    Change
     1 P_Mix/Diag.   13.3     1.37995875      -343.8583303534 -3.44E+02

1. DZVP-MOLOPT-SR-GTH
  Step     Update method      Time    Convergence         Total energy    Change
     1 P_Mix/Diag. 0.40E+00    1.6     0.97935146      -341.1328952493 -3.41E+02

Is this expected behavior?
Does DZVP-MOLOPT-GTH introduce substantially more diffuse functions or a larger overlap matrix compared to the SR version?

Thanks!



--
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+unsubscribe at googlegroups.com<mailto:cp2k+unsubscribe at googlegroups.com>.
To view this discussion visit https://groups.google.com/d/msgid/cp2k/c859cff8-23f1-42b2-b9f2-4bd107824d49n%40googlegroups.com<https://groups.google.com/d/msgid/cp2k/c859cff8-23f1-42b2-b9f2-4bd107824d49n%40googlegroups.com?utm_medium=email&utm_source=footer>.

-- 
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+unsubscribe at googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/cp2k/ZR0P278MB0759D3C5CCBE4893CDF321EC9F032%40ZR0P278MB0759.CHEP278.PROD.OUTLOOK.COM.