[CP2K-user] Zeolite structure and cutoff not converging

Travis polla... at gmail.com
Tue Oct 22 17:52:25 UTC 2019


Hi,

Several corrections should help you out...

EPS_DEFAULT should be a smaller,
  &QS
    EPS_DEFAULT 1.0E-12
    EXTRAPOLATION ASPC
  &END QS

A reasonable grid setup,
  &MGRID
   NGRIDS 5 # faster MOLOPT calculations
   CUTOFF 450 # should be at least largest exponent in basis set multiplied 
by REL_CUTOFF, round up just a bit
   REL_CUTOFF 50 # or 60 is fine too, jump to 80 for flexible cell 
calculations
  &END MGRID

Switch to the MOLOPT basis sets. Faster and probably better quality than 
the Pade XC optimized basis set.
   &KIND X
    BASIS_SET DZVP-MOLOPT-SR-GTH
    POTENTIAL GTH-PBE
   &END KIND

Faster geometry optimization than CG,
  &GEO_OPT
    TYPE MINIMIZATION
    MAX_FORCE 4e-5
    MAX_ITER 200
    OPTIMIZER BFGS # CG should be used in pathological cases
  &END GEO_OPT

Faster SCF,
  &SCF
    SCF_GUESS RESTART # defaults to ATOMIC without wfn present, no need to 
deviate from this unless using k-points
    EPS_SCF 1.0E-06 # slightly tighter
    MAX_SCF 30 # range of 20-40 is fine
    &OT  T # use instead of full diagonalization for non-metals
     MINIMIZER  DIIS
     PRECONDITIONER  FULL_ALL
     ENERGY_GAP  0.002
    &END OT
    &OUTER_SCF  T
     EPS_SCF  1e-6
     MAX_SCF  20 # will update preconditioner up to 20 times before giving 
up
    &END OUTER_SCF 
  &END SCF

-T


On Tuesday, October 22, 2019 at 10:17:51 AM UTC-3, Carlos Bornes wrote:
>
> Hi there,
>
> I'm quite new to CP2K, I've decided to move from the 'HOW TO' examples 
> from CP2K's website and apply what I've learn in the materials of my 
> research but I got stuck both in the optimization of the Cutoff and the 
> Geom Opt. I started by replacing one of the Si atoms by and Al atom and 
> adding a H close to the Al.
> I've tried to optimize optimize the Cutoff while Real Cutoff = 60 but it 
> seems that the energy does not converge.
>
> After that I've decided to move one with Cuttoff=300 and Real_Cutoff=60 
> and the calculation seems to stop in the middle of the 9 iteration without 
> converging 
>
> The input and output files for Geo Opt and Cutoff results are attached
>
> Thank you
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20191022/7ee6cf01/attachment.htm>


More information about the CP2K-user mailing list