HCTH407 GTH pseudopotential set?

[Matthias Krack]

Hi Hanning,

the final segmentation fault may have various reasons and I cannot
tell you exactly the reason.
However, here are some suggestions concerning the fitting procedure:

- The atomic all-electron reference calculation is usually performed
for the relativistic atom. Such a run will produce more information
which is later used as a reference during the fitting of the new
pseudopotential.

- use the final psp.par and weights.par files of an already optimised
pseudopotential, e.g. C HCTH120 as a starting point. These parameters
are most likely a very good guess.

- don't try to optimise all parameters, just check the FITPAR file for
C from another functional, e.g. again HCTH120

Best regards,

Matthias

On Aug 30, 6:58 am, NUCP2K <chenh... at gmail.com> wrote:
> Dear CP2K community,
>
>   Could someone share me the GTH pseodopotentials for Carbon and
> Nitrogen using the HCTH407 functional ? It seems that only the files
> for Oxygen and Hydrogen are on the repository.
>
>   I tried to use the following "atoms.dat" file to build up the
> pseudopotential for Carbon.
>
>          C
>    HCTH407
>    nonrel
>        60.0     10.0 30.0      rmax,aa,bb
>        1.3d0   50.0            rcov,rprb
>     1    3                     number of core and valence orbitals
>     2    0     2.00      0.00
>     2    1     2.00      0.00
>     3    0     0.00      0.00
>
> Some all-electron files were successfully generated together with the
> "psp.par"
>
>   10   2.0     ng, rij (initial guess)
>    1.3000000000  50.0000000000 rcov, rprb
>  non relativistic calculation
>  HCTH407      XC-functional
>   6.000  4.000  0.325  0.0 0.0 0.0 0.0    znuc,zpseudo,rloc,gpot()
>            2  lpx
>   0.325  0.0 0.0 0.0 0.0 0.0 0.0 r_l(), hsep()
>   0.325  0.0 0.0 0.0 0.0 0.0 0.0 r_l(), hsep()
>   0.325  0.0 0.0 0.0 0.0 0.0 0.0 r_l(), hsep()
>
> and "weights.par".
>
>  Initial weight-guess from atomic-program
>  ----------------------------------------
>     .1E+03     weight_psir0
>     n   l  so   eigval  chrg    dchrg  ddchrg  res    rnode dnode
> ddnode
>    2   0 0.00   1.0e0   1.0e0   0.0e0  0.0e0   0.0e0  1.0e0 0.0e0
> 0.0e0
>    2   1 0.00   1.0e0   1.0e0   0.0e0  0.0e0   0.0e0  1.0e0 0.0e0
> 0.0e0
>    3   0 0.00   0.0e0   0.0e0   0.0e0  0.0e0   0.0e0  0.0e0 0.0e0
> 0.0e0
>
> Thereafter, I created the file "FITPAR" as instructed:
>
>   Fit                        Parameter
>  t    t t t t                rloc,gpot(1,..,4)
>             2                lmax
>  t    t t t t t t            r_l(1),h11,h12,h22,h13,h23,h33 (l=0)
>  t    t t t t t t            r_l(2),h11,h12,h22,h13,h23,h33 (l=1)
>  f    f f f f f f            r_l(3),h11,h12,h22,h13,h23,h33 (l=2)
>  f    f f f f f f            r_l(4),h11,h12,h22,h13,h23,h33 (l=3)
>
> However, when I typed "/pseudo.x -orth -c100 -n600", the program
> crashed with the following messages:
>
>  *********************************************
>  ***           pseudo_2.2                  ***
>  ***           fitting of                  ***
>  ***   goedecker type pseudopotentials     ***
>  ***   last changes: 14.7.98               ***
>  *********************************************
>
>  orthogonalize the projectors
>          100 fit cycles
>          600 max. simplex iterations
>  ***        Reading data from atom.ae      ***
>  pseudo states =            3
>  znuc          =    6.0000000000000000
>  zpseudo       =    4.0000000000000000
>  r_covalent    =    1.3000000000000000
>  r_confining   =    50.000000000000000
>  ispp          = n
>  icorr         = HCTH407
>  gridpoints    =          477
>   nl    s      occ        eigenvalue     charge(rcov)
>  2s   0.0    2.0000   -0.5169586012    0.4040927822
>  2p   0.0    2.0000   -0.2003521860    0.3587996075
>  3s   0.0    0.0000   -0.0090950680    0.0182791586
>  ***        Quantum numbers        ***
>  AE             <->  PSEUDO
>  n,l(AE):            2           0       n,l(PS):
> 1           0
>  n,l(AE):            2           1       n,l(PS):
> 1           1
>  n,l(AE):            3           0       n,l(PS):
> 2           0
>  ***        Reading data from psp.par
> ***
>
>   Initial Pseudpotential Parameter from psp.par
>   ---------------------------------------------
>      1.300    50.000         rcov,rprb
>                              non relativistic calculation
>  HCTH407                     XC-functional
>  local part
>    6.   4.00  0.325  0.000E+00  0.000E+00  0.000E+00  0.000E+00
> znuc,zion, rloc, gpot()
>  nonlocal
> part
>    2                                                       lpx,
> (Projectors for l=0..lpx)
>  l=
> 0
>   0.325  0.000E+00  0.000E+00  0.000E+00  0.000E+00  0.000E+00  0.000E
> +00  r_l(),hsep(),   so=0
>  l=
> 1
>   0.325  0.000E+00  0.000E+00  0.000E+00  0.000E+00  0.000E+00  0.000E
> +00  r_l(),hsep(),   so=0
>  l=
> 2
>   0.325  0.000E+00  0.000E+00  0.000E+00  0.000E+00  0.000E+00  0.000E
> +00  r_l(),hsep(),   so=0
>
>  Reading actual weights from file weights.par
>
>  Basis:
>  ------
>  amin,a0in,amax 0.1625E+00 0.3250E+00 0.1300E+01
>  gaussian 0.1249E+02 0.8242E+01 0.5438E+01 ....  0.4484E+00 0.2959E+00
>  gaussians:          10
>  r-grid:  0.4063E-03 0.4520E-03 0.5029E-03 ....  0.1079E+03 0.1200E+03
>  gridpoints:         119
>  T T T T T
>  T T T T T T T
>  T T T T T T T
>  F F F F F F F
>  fitting parameter determined by 'FITPAR'
>   l=           0
>  warning! transformation of hsep(): no fit of h(1,2)
>   l=           0
>  warning! transformation of hsep(): no fit of h(1,3)
>   l=           0
>  warning! transformation of hsep(): no fit of h(2,3)
>   l=           1
>  warning! transformation of hsep(): no fit of h(1,2)
>   l=           1
>  warning! transformation of hsep(): no fit of h(1,3)
>   l=           1
>  warning! transformation of hsep(): no fit of h(2,3)
>  fitting parameter:
>  rloc
>  gpot(1)
>  gpot(2)
>  gpot(3)
>  gpot(4)
>  r_l(1)
>  hsep(1,1)
>  hsep(1,3)
>  hsep(1,6)
>  r_l(2)
>  hsep(2,1)
>  hsep(2,3)
>  hsep(2,6)
>  maxdim,nfit:          30          13
>  nfit=          13
>  penalties for start simplex:
> Segmentation fault
>
> Can someone tell me what went wrong?
>
> Thanks.
>
> Hanning