Difficulties with Cell optimization

[Lucas Bandeira]

Dear CP2K Community,

I'm performing cell optimization for a gold slab varying the basis sets used. The system is composed by 72 Au atoms. the basis I'm using are DZVP, TZVP and TZV2P all of them are MOLOPT. In the output, the message I get after some optimization iterations is "run CONVERGED", but I suppose the default message is "GEOMETRY OPTIMIZATION COMPLETED". What is the difference between them? Something went awry with my simulation? Moreover, not all the four convergence criteria were satisfied at the last convergence cycle.

More exactly the message printed is:

  ***********************************************
 * Specific L-BFGS convergence criteria
 * WANTED_PROJ_GRADIENT and WANTED_REL_F_ERROR
 * satisfied .... run CONVERGED!
 ***********************************************

The template used as an input is:

&FORCE_EVAL
  METHOD Quickstep
  STRESS_TENSOR ANALYTICAL
  &DFT
  &PRINT
    &E_DENSITY_CUBE LOW
      &EACH
        CELL_OPT 50
      &END EACH
      ADD_LAST SYMBOLIC
    &END E_DENSITY_CUBE
  &END
    BASIS_SET_FILE_NAME BASIS_MOLOPT
    BASIS_SET_FILE_NAME BASIS_MOLOPT_UCL
    POTENTIAL_FILE_NAME GTH_POTENTIALS
    &QS
      EPS_DEFAULT 1.0E-7
      METHOD GPW
      EXTRAPOLATION ASPC
      EXTRAPOLATION_ORDER 3
    &END QS
    &MGRID
      CUTOFF XXcutXX
      NGRIDS 4
    &END
    &SCF
      MAX_SCF 20
      SCF_GUESS RESTART
      EPS_SCF 1.0E-5
      &OT
        PRECONDITIONER  FULL_SINGLE_INVERSE
        MINIMIZER  CG
      &END
      &OUTER_SCF
        MAX_SCF 50
        EPS_SCF 1.0E-5
      &END
      &PRINT
      &END
    &END SCF
    &XC
      &XC_FUNCTIONAL XXfunctionalXX
      &END XC_FUNCTIONAL
    &END XC
    &POISSON
       PERIODIC XYZ
       POISSON_SOLVER ANALYTIC
    &END
  &END DFT
  &SUBSYS
    &CELL
      ABC  XXaXX  XXbXX  XXcXX          #Marcadores das dimensões do slab
      PERIODIC XYZ
      SYMMETRY ORTHORHOMBIC
      MULTIPLE_UNIT_CELL 1 1 1
    &END CELL
    &TOPOLOGY
      COORD_FILE_NAME XXsurfaceXXSlab.xyz
      COORDINATE xyz
      MULTIPLE_UNIT_CELL 1 1 1
    &END
    &KIND Au
      BASIS_SET XXbasisXX
      POTENTIAL GTH-XXfunctionalXX
    &END KIND
  &END SUBSYS
&END FORCE_EVAL
&MOTION
   &CELL_OPT
     OPTIMIZER LBFGS ! Good choice for 'small' systems (use LBFGS for large systems)
     MAX_ITER  100
     KEEP_SYMMETRY
     KEEP_ANGLES
     MAX_DR    [bohr] 0.003 ! adjust target as needed
     &LBFGS
     &END
   &END CELL_OPT
   &PRINT
     &RESTART

        ADD_LAST SYMBOLIC
     &END RESTART
   &END PRINT
&END MOTION
&GLOBAL
  PRINT_LEVEL LOW
  PROJECT Au_XXprojectXX
  RUN_TYPE CELL_OPT
&END GLOBAL

I hope you can help me.

Sincerely yours,

Lucas Bandeira

[Thomas Kühne]

Dear Lucas, 

contrary to SD, CG and BFGS, the LBFGS optimizer is terminating according to 

the WANTED_PROJ_GRADIENT and WANTED_REL_F_ERROR criteria instead 

of the RMS and MAX values. The latter are only reported, but do not impact the 

optimization. Hence, this does not imply that there must be s.th. wrong with your 

simulation … 

Cheers, 

Thomas

P.S. If you want to to play it safe simply restart the final geometry of your LBFGS 

optimization using a different method. 

Am 26.03.2020 um 19:15 schrieb Lucas Bandeira <bandeir...@gmail.com>:

Dear CP2K Community,

I'm performing cell optimization for a gold slab varying the basis sets used. The system is composed by 72 Au atoms. the basis I'm using are DZVP, TZVP and TZV2P all of them are MOLOPT. In the output, the message I get after some optimization iterations is "run CONVERGED", but I suppose the default message is "GEOMETRY OPTIMIZATION COMPLETED". What is the difference between them? Something went awry with my simulation? Moreover, not all the four convergence criteria were satisfied at the last convergence cycle.

More exactly the message printed is:

  ***********************************************
 * Specific L-BFGS convergence criteria
 * WANTED_PROJ_GRADIENT and WANTED_REL_F_ERROR
 * satisfied .... run CONVERGED!
 ***********************************************

I hope you can help me.

Sincerely yours,

Lucas Bandeira

--
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/3ba76d03-524d-4167-9e81-cda8363f1c8b%40googlegroups.com.

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

Thomas D. Kühne

Dynamics of Condensed Matter

Chair of Theoretical Chemistry

University of Paderborn

Warburger Str. 100

D-33098 Paderborn

Germany

tdku...@mail.upb.de

+49/(0)5251/60-5726