Nonperiodic Poisson solvers

Matt W mattwa... at gmail.com
Sat Sep 30 20:26:49 UTC 2017


Hi,

I'm a zombie after 12 hours on a plane, so may contain mistakes:

On Wednesday, September 27, 2017 at 5:29:47 PM UTC+4, Jadzia wrote:
>
> Hi everyone, 
>
> I am trying to simulate a system with an isolated molecule, therefore I 
> need a non-periodic setup. I have read the previous posts about the topic, 
> but there are a number of questions open. They are both for the wavelet and 
> the MT Poisson solver. 
>
> *1) System size*
> My simulations will be quite long (both MM and QM) and due to the movement 
> of the molecule, to my understanding it needs to be avoided that the 
> molecule leaves the box. Will it be fine if I set the box size to 1000 
> Angstrom for security reasons or is this not recommended? 
>
> Definitely not fine for QM, because you will run out of memory - we store 
densities etc on grids that grow with the volume of the system. MM you 
could be OK if totally non-periodic.
 

> *2) Periodic cells*
> This question is closely related to question 1). If I set the Poisson 
> solver to non-periodic, could one set the CELL to periodic, and thus solve 
> the problem that the molecule leaves the box?
>
> This would work for MM, I think. I suspect that MT doesn't care where the 
molecule is in the box, but haven't every tried it.
 

> *3) Ewald methods*
> Also closely related to question 1). If I set the box size to 1000, even 
> with Ewald as Poisson solver the system should be effectively non-periodic 
> due to the large distance, right? Would this be a plausible alternative to 
> the MT or wavelet solvers?
>
> Yes, at least for neutral systems, but it would be extremely expensive.
 

> *4) Mixed force evaluations*
> In the case that I have a mixed force evaluation setup and use the Wavelet 
> solver (see the config file below), do I need to include the 
> &CENTER_COORDINATES section also in the mixed force eval section, or only 
> in the real force evaluation section?
>
> No idea. Test _very_ carefully.
 

> *5) MT solver for periodic systems*
> Is the MT solver only for nonperiodic systems in CP2K, or does it also 
> work for periodic systems? In the publication it is at least written it 
> works for periodic systems and even with ewald/spme. 
>
> Don't know, try it. 
 
Matt

Looking forward to hearing from you.
>
> Best wishes,
> Jadzia
>
>
> &GLOBAL
>   PROJECT system
>   RUN_TYPE md
>   PRINT_LEVEL medium
> &END global
>
> &MOTION
>   &MD
>     ENSEMBLE langevin
>     &LANGEVIN 
>       GAMMA [fs^-1] 1.0E-3                        # "Langevin" gamma 
> (adapt for your system) 
>     &END LANGEVIN
>     STEPS 5000000
>     TIMESTEP 0.5
>     TEMPERATURE 300.0
>   &END MD  
> &END MOTION
>
> &MULTIPLE_FORCE_EVALS
>    FORCE_EVAL_ORDER 2 3
>    MULTIPLE_SUBSYS true
> &END
>
> &FORCE_EVAL
>   STRESS_TENSOR analytical
>   METHOD mixed
>   &MIXED
>     MIXING_TYPE genmix
>     &GENERIC
>       VARIABLES a b
>       MIXING_FUNCTION a+b
>     &END
>     @include cp2k.in.mapping
>   &END
>   &SUBSYS
>     &TOPOLOGY
>       &CENTER_COORDINATES 
>       &END 
>       CONNECTIVITY off
>       COORDINATE pdb
>       COORD_FILE_NAME system.pdb
>     &END TOPOLOGY
>     &CELL
>       PERIODIC NONE
>       ABC 20 20 20 
>     &END CELL
>   &END SUBSYS
> &END FORCE_EVAL
>
> &FORCE_EVAL
>   STRESS_TENSOR analytical
>   METHOD fist
>   &MM
>     &FORCEFIELD
>       PARM_FILE_NAME system1.prm
>       PARMTYPE chm
>       &SPLINE
>         EMAX_SPLINE 100000
>         RCUT_NB 12.
>       &END
>     &END FORCEFIELD    
>     &POISSON
>       POISSON_SOLVER WAVELET
>       PERIODIC NONE
>       &EWALD 
>         EWALD_TYPE NONE 
>       &END EWALD 
>     &END POISSON
>   &END MM
>   &SUBSYS
>     &CELL
>       PERIODIC NONE
>       ABC 20 20 20
>     &END CELL
>     &TOPOLOGY
>       &CENTER_COORDINATES 
>       &END 
>       COORDINATE pdb
>       COORD_FILE_NAME system1.pdb
>       CONN_FILE_FORMAT psf
>       CONN_FILE_NAME system1.psf
>     &END TOPOLOGY
>   &END SUBSYS
> &END FORCE_EVAL
>
> &FORCE_EVAL
>   STRESS_TENSOR analytical
>   METHOD fist
>   &MM
>     &FORCEFIELD
>       PARM_FILE_NAME system2.prm
>       PARMTYPE chm
>       &SPLINE
>         EMAX_SPLINE 100000
>         RCUT_NB 12.
>       &END
>     &END FORCEFIELD    
>     &POISSON
>       POISSON_SOLVER WAVELET
>       PERIODIC NONE
>       &EWALD 
>         EWALD_TYPE NONE 
>       &END EWALD 
>     &END POISSON
>   &END MM
>   &SUBSYS
>     &CELL
>       PERIODIC NONE
>       ABC 20 20 20
>     &END CELL
>     &TOPOLOGY
>       &CENTER_COORDINATES 
>       &END 
>       COORDINATE pdb
>       COORD_FILE_NAME system2.pdb
>       CONN_FILE_FORMAT psf
>       CONN_FILE_NAME system2.psf
>     &END TOPOLOGY
>   &END SUBSYS
> &END FORCE_EVAL
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20170930/06d85a74/attachment.htm>


More information about the CP2K-user mailing list