convergency problem using SCAN+rVV10

[huy pham]

Hi All,

I have the problem with the convergence when using SCAN+rVV10 functional for the very simple system with only one water molecule.
The convergence just fluctuate around 0.002, and can not decrease more (see below).

     1 OT DIIS     0.80E-01   54.9     0.00325654       -17.2027773693 -4.08E-07
     2 OT DIIS     0.80E-01   27.5     0.00276549       -17.2027717490  5.62E-06
     3 OT DIIS     0.80E-01   27.6     0.00376733       -17.2027529642  1.88E-05
     4 OT DIIS     0.80E-01   27.5     0.00424108       -17.2027547590 -1.79E-06
     5 OT DIIS     0.80E-01   27.6     0.00315550       -17.2027413567  1.34E-05
     6 OT DIIS     0.80E-01   27.4     0.00191592       -17.2027409117  4.45E-07
     7 OT SD       0.80E-01   27.5     0.00388064       -17.2027478523 -6.94E-06
     8 OT DIIS     0.80E-01   27.6     0.00287957       -17.2027087049  3.91E-05
     9 OT DIIS     0.80E-01   27.6     0.00156285       -17.2027379497 -2.92E-05
    10 OT DIIS     0.80E-01   27.5     0.00129428       -17.2027445179 -6.57E-06
    11 OT DIIS     0.80E-01   27.2     0.00161973       -17.2027517827 -7.26E-06
    12 OT DIIS     0.80E-01   27.5     0.00220730       -17.2027652153 -1.34E-05
    13 OT DIIS     0.80E-01   28.0     0.00127484       -17.2027571303  8.09E-06
    14 OT DIIS     0.80E-01   27.5     0.00260157       -17.2027619579 -4.83E-06
    15 OT DIIS     0.80E-01   27.8     0.00124040       -17.2027613316  6.26E-07
    16 OT DIIS     0.80E-01   27.4     0.00161009       -17.2027627322 -1.40E-06
    17 OT DIIS     0.80E-01   27.7     0.00270573       -17.2027633849 -6.53E-07
    18 OT DIIS     0.80E-01   27.5     0.00143321       -17.2027609365  2.45E-06
    19 OT DIIS     0.80E-01   27.6     0.00949434       -17.2027630484 -2.11E-06
    20 OT DIIS     0.80E-01   27.6     0.00206430       -17.2027666772 -3.63E-06
    21 OT DIIS     0.80E-01   27.3     0.00327081       -17.2027624981  4.18E-06
    22 OT DIIS     0.80E-01   27.6     0.00140816       -17.2027603709  2.13E-06
    23 OT DIIS     0.80E-01   27.6     0.00194449       -17.2027617267 -1.36E-06
    24 OT DIIS     0.80E-01   27.5     0.00248205       -17.2027603408  1.39E-06
    25 OT DIIS     0.80E-01   28.1     0.00317999       -17.2027635590 -3.22E-06
    26 OT DIIS     0.80E-01   27.6     0.00376279       -17.2027604966  3.06E-06
    27 OT DIIS     0.80E-01   27.4     0.00175908       -17.2027571292  3.37E-06
    28 OT DIIS     0.80E-01   27.6     0.00113761       -17.2027575317 -4.02E-07
    29 OT DIIS     0.80E-01   27.5     0.00631357       -17.2027592196 -1.69E-06
    30 OT DIIS     0.80E-01   27.7     0.00129080       -17.2027564625  2.76E-06
    31 OT DIIS     0.80E-01   27.3     0.00296235       -17.2027626012 -6.14E-06
    32 OT DIIS     0.80E-01   27.6     0.00267161       -17.2027653970 -2.80E-06
    33 OT DIIS     0.80E-01   27.6     0.00196123       -17.2027620008  3.40E-06
    34 OT DIIS     0.80E-01   27.6     0.00344508       -17.2027683804 -6.38E-06

I tried many changes that was recommended from different posts in the forum (changing MINIMIZER, PRECONDITIONER, ENERGY_GAP, ...), but nothing works.

I only have problem with SCAN+rVV10, I tried other functionals (B97M-rV which is also MGGA+rVV10, revPBE-D3, BLYP-D3, rVV10) and they all work. I don't know what was the problem. Could you please give me some suggestions? Has anyone perform the calculation with SCAN/SCAN+rVV10 successfully, if yes, can you send me one example of the input?
I installed cp2k following this post: https://groups.google.com/forum/#!topic/cp2k/yJP-HRqx4Y0

Thanks,
Huy

Below is my input
--------------------------------
&GLOBAL
  PROJECT WATER
  RUN_TYPE ENERGY
  IOLEVEL  LOW
&END GLOBAL

&FORCE_EVAL
 
  METHOD Quickstep

  &DFT
   
    BASIS_SET_FILE_NAME GTH_BASIS_SETS
    POTENTIAL_FILE_NAME GTH_POTENTIALS

    &MGRID
       CUTOFF 1200
       NGRIDS 5
    &END

    &QS
      EPS_DEFAULT 1.0E-18
      WF_INTERPOLATION ASPC
    &END QS

    &SCF                          
  
      SCF_GUESS RESTART
      MAX_SCF 40
      EPS_SCF 1.0E-9

      &OT
        MINIMIZER DIIS
        PRECONDITIONER FULL_ALL
        ENERGY_GAP 0.002
      &END OT

      &OUTER_SCF
        MAX_SCF 40
        EPS_SCF 1.0E-9
      &END OUTER_SCF

      &PRINT
        &RESTART
          &EACH
            MD 0
          &END
        &END RESTART
      &END PRINT
   &END SCF

      &POISSON
        PERIODIC NONE
        PSOLVER  MT
      &END POISSON

    &XC
      &XC_FUNCTIONAL
        &LIBXC
          FUNCTIONAL MGGA_X_SCAN
        &END LIBXC
        &LIBXC
          FUNCTIONAL MGGA_C_SCAN
        &END LIBXC
      &END XC_FUNCTIONAL

      &vdW_POTENTIAL
        DISPERSION_FUNCTIONAL NON_LOCAL
        &NON_LOCAL
            TYPE RVV10
            PARAMETERS 15.7 0.0093
            KERNEL_FILE_NAME rVV10_kernel_table.dat
        &END NON_LOCAL
      &END vdW_POTENTIAL

      &XC_GRID
         XC_DERIV NN50_SMOOTH
      &END XC_GRID
    &END XC

  &END DFT
 
  &SUBSYS

    &CELL
      ABC 25.0 25.0 25.0
      PERIODIC NONE
    &END CELL

    &TOPOLOGY
      COORD_FILE_NAME file.xyz
      COORD_FILE_FORMAT XYZ
      &CENTER_COORDINATES
      &END
    &END

    &KIND H                             
      BASIS_SET aug-QZV3P-GTH
      POTENTIAL GTH-PBE-q1            
    &END KIND

    &KIND O
      BASIS_SET aug-QZV3P-GTH
      POTENTIAL GTH-PBE-q6
    &END KIND

    &KIND C
      BASIS_SET aug-QZV3P-GTH
      POTENTIAL GTH-PBE-q4
    &END KIND

  &END SUBSYS

&END FORCE_EVAL

&MOTION
  &GEO_OPT
    MAX_ITER 2000
    RMS_FORCE 1.0E-6
    MAX_FORCE 3.0E-6
    RMS_DR 1.0E-5
    MAX_DR 3.0E-5
  &END
&END

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

Hi

yes, there seems to be a problem with the SCAN functional. The
behavior points to an inconsistency of potential and energy.
If this is a problem of the libxc library or our new interface
is not clear. However, it seems to work smoothly for other
MGGA functionals (TPSS).
Is there a fundamental difference between SCAN and TPSS?

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: huy pham
Sent by: cp...@googlegroups.com
Date: 03/01/2018 07:43AM
Subject: [CP2K:10044] convergency problem using SCAN+rVV10

--
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.
For more options, visit https://groups.google.com/d/optout.

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

Hi

it seems that the SCAN functional is even more sensitive to
the small density/gradient/tau regime.
For systems with large empty space with very low electron density
it is very difficult to get tight convergence.

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: huy pham
Sent by: cp...@googlegroups.com
Date: 03/01/2018 07:43AM
Subject: [CP2K:10044] convergency problem using SCAN+rVV10

[huy pham]

Hi Juerg,

Thank you for your email.
I actually tried with different options and found that using MIXING METHOD DIRECT_P_MIXING helps to converge the energy.
Please part of the input below.

Anyway, I have to check very carefully the result of the SCAN functional.

Thanks,
Huy


      SCF_GUESS RESTART
      MAX_SCF 4000
      EPS_SCF 1.0E-7

      ADDED_MOS 400
      CHOLESKY OFF
      &SMEAR ON
        METHOD energy_window
        WINDOW_SIZE 0.02
      &END SMEAR
     &DIAGONALIZATION
          ALGORITHM STANDARD
     &END DIAGONALIZATION
     &MIXING
          METHOD DIRECT_P_MIXING
          ALPHA   0.1
          BETA    1.5
          NBROYDEN  8
     &END

[Yongbin Zhuang]

Hi Huy,

    I found a same problem with SCAN when I optimize the slab model.

    this is very helpful