[CP2K-user] [CP2K:12185] modelling protein crystals

[hut... at chem.uzh.ch]

Hi

yes, LS_SCF is for very large systems, but
- it has a large prefactor that is strongly dependent on the
  method and system (basis set etc)
- it needs much more care and should only be used after
  extensively study the system properties using smaller systems.

Depending on the QM method you are using (I assume here a GGA KS-DFT
with a DZVP basis) I would guess

- protein crystal of 10000 atoms -> use OT
- 100000 atom system -> use LS_SCF and a very large supercomputer

regards

Juerg Hutter
--------------------------------------------------------------
Juerg Hutter                         Phone : ++41 44 635 4491
Institut für Chemie C                FAX   : ++41 44 635 6838
Universität Zürich                   E-mail: hut... at chem.uzh.ch
Winterthurerstrasse 190
CH-8057 Zürich, Switzerland
---------------------------------------------------------------

-----cp... at googlegroups.com wrote: -----
To: cp... at googlegroups.com
From: "Pierre Cazade" 
Sent by: cp... at googlegroups.com
Date: 09/06/2019 10:44AM
Subject: Re: [CP2K:12185] modelling protein crystals

Dear Juerg,

Thank you for your answer. Could you please elaborate? Is not Linear 
Scaling SCF meant to make it feasible running QM calculation for large 
systems? It was my understanding from the workshops presentations and 
the tutorials. What would you recommend for a protein crystal of 10000 
atoms and another of 100000 with different heme groups (A and B) and 
ions bridges? Simply OT?

Thank you again,
Pierre

On 06/09/2019 09:37, hut... at chem.uzh.ch wrote:
> Hi
>
> I would not start using the LS_SCF optimizer, but use a standard
> OT setup.
>
> regards
>
> Juerg Hutter
> --------------------------------------------------------------
> Juerg Hutter                         Phone : ++41 44 635 4491
> Institut für Chemie C                FAX   : ++41 44 635 6838
> Universität Zürich                   E-mail: hut... at chem.uzh.ch
> Winterthurerstrasse 190
> CH-8057 Zürich, Switzerland
> ---------------------------------------------------------------
>
> -----cp... at googlegroups.com wrote: -----
> To: "cp2k" <cp... at googlegroups.com>
> From: "Pierre-André Cazade"
> Sent by: cp... at googlegroups.com
> Date: 09/04/2019 12:29PM
> Subject: [CP2K:12182] modelling protein crystals
>
> Dear CP2K users,
>
> Aside for my previous post, I am also interested in using CP2K to model protein crystals. Some of the proteins I am interested in are metalloproteins. Using again my test system, beta glycine, I am simply adding the following block to an input that works to preform a CELL_OPT. Please find part of this input below:
>
>   &FORCE_EVAL
>     METHOD  QS
>     STRESS_TENSOR  ANALYTICAL
>     &DFT
>       BASIS_SET_FILE_NAME /usr/local/src/cp2k/data/GTH_BASIS_SETS
>       POTENTIAL_FILE_NAME /usr/local/src/cp2k/data/GTH_POTENTIALS
>       MULTIPLICITY  0
>       CHARGE  0
>       &SCF
>         MAX_SCF  20
>         EPS_SCF  1.e-7
>         SCF_GUESS  ATOMIC
>         &OT  T
>           MINIMIZER CG
>           PRECONDITIONER FULL_SINGLE_INVERSE
>         &END OT
>         &OUTER_SCF  T
>           EPS_SCF  1.e-6
>           MAX_SCF  50
>         &END OUTER_SCF
>       &END SCF
>       &QS
>         LS_SCF T
>         EPS_DEFAULT     1.0000000000000000E-10
>         EPS_PGF_ORB     1.0000000000000000E-08
>         METHOD  GPW
>       &END QS
>       &MGRID
>         NGRIDS  4
>         CUTOFF     5.0000000000000000E+02
>         REL_CUTOFF     8.0000000000000000E+01
>       &END MGRID
>       &XC
>         DENSITY_CUTOFF     1.0000000000000000E-10
>         GRADIENT_CUTOFF     1.0000000000000000E-10
>         TAU_CUTOFF     1.0000000000000000E-10
>         &XC_FUNCTIONAL  NO_SHORTCUT
>           &PBE  T
>           &END PBE
>         &END XC_FUNCTIONAL
>       &END XC
>       &POISSON
>         POISSON_SOLVER  PERIODIC
>         PERIODIC  XYZ
>       &END POISSON
>     &END DFT
>
> So I tried to add this block:
>
>       &LS_SCF
>         PURIFICATION_METHOD TRS4
>         EPS_FILTER 1E-6
>         EPS_SCF    1E-7
>         MAX_SCF    20
>         S_PRECONDITIONER ATOMIC
>       &END
>
> and it does not work. Then, I tried:
>
>       &LS_SCF
>         PURIFICATION_METHOD TRS4
>         EPS_FILTER 1E-6
>         EPS_SCF    1E-7
>         MAX_SCF    30
>         S_PRECONDITIONER NONE
>         &CURVY_STEPS
>         &END CURVY_STEPS
>       &END
>
> And again, it failed. In both cases, at first the optimization goes well, I reach a point where the system is almost optimized and then everything goes wrong. If anyone can tell me what I am doing wrong, I would be grateful. More generally, if anyone could tell me which algorithms I should use to maximize the efficiency of the calculation for large systems (up to about 100000 atoms), while having decent accuracy and stability, I would be grateful. I have checked the various options in SCF, OT, OUTER_SCF, LS_SCF and MOTION, but it is unclear which ones are the most suitable and can work well together.
>
> Thank you,
> Pierre
>
>    
>    --
>   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 cp... at googlegroups.com.
>   To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/ea4192e6-fe57-4ecd-99eb-e1cb6430c365%40googlegroups.com.
>   
>

-- 
Dr Pierre Cazade, PhD
AD3-023, Bernal Institute,
University of Limerick,
Plassey Park Road,
Castletroy, co. Limerick,
Ireland

-- 
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 cp... at googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/47f4bdba-74cc-ab74-bf18-53a2dea287ea%40gmail.com.