EMAX_SPLINE too small

[salah]

Dear Laino,
first of all, thank you very much for the very detailed answer.
I would like to start with your comment on the neutrality of the
system. I'm doing some tests on different polymorphs of SiO2 (bulk).
So I set the charges exactly as shown if the distribution test files,
UO2 for instance, and here a copy-past from the regtest-7 files in
Fist directory:
##########################
       &CHARGE
        ATOM U
        CHARGE 4.0
      &END
      &CHARGE
        ATOM O
        CHARGE -2.0
      &END
      &SHELL U
        CORE_CHARGE   -2.840000
        SHELL_CHARGE 6.840000
        MASS_FRACTION  0.1
        SPRING  [eV*angstrom^-2] 171.556
        MAX_DISTANCE [angstrom] 1.0
      &END SHELL
      &SHELL O
        CORE_CHARGE 1.186267
        SHELL_CHARGE -3.186267
        MASS_FRACTION 0.2
        SPRING [eV*angstrom^-2] 70.824
        MAX_DISTANCE [angstrom] 1.0
      &END SHELL
###################################
as you can see the charge comes from the atom definition and the shell
definition as well.
I checked the output of my simple run and it says :

------------> CHARGE_INFO| Total Charge of the Classical
System:                     0.000000
I've tried to set the charge of O to zero and kept the charge coming
from the shell-core part but cp2k displays a positive total charge for
the system.
so, I do not see what do you mean by setting charge of O to zero and
keep the charge flowing from the shell-core part !!!!!!

second point is the EMAX_SPLINE itself. I've tried to increase it but
the simulation start running only with a value higher than 300
hartree !!! does this really need to be that high ????

third point, the 61 61 61, of course it's a crazy value, I was
increasing it automatically and the file I posted was the last one, I
do not think that I need that much, a GMAX of 21 will be sufficient I
guess.

fourth point, a timestep of 0.2 fs should be enough to capture the
movements of the shell-core couples especially with a high force
constant (using DLPOLY for example, simulations with a shell model and
with 0.2 fs is stable without drift of energy).

last point, I changed everything as you advised but the system
collapsed.

So, can you please do me a favor and check this again
closely ??????????????????

Thanks in advance

salah



On 2 oct, 11:45, Laino Teodoro <teodor... at gmail.com> wrote:
> ... and now most important few comments on your input file:
>
>
>
> >   &MM
> >     &FORCEFIELD
> >       &CHARGE
> >         ATOM Si
> >         CHARGE 2.7226
> >       &END CHARGE
> >       &CHARGE
> >         ATOM O
> >         CHARGE -1.3613
> >       &END CHARGE
> >       &SHELL O
> >         CORE_CHARGE 1.91981
> >         SHELL_CHARGE -3.28111
> >         MASS_FRACTION 0.1
> >         SPRING [eV*angstrom^-2] 256.71027
> >         MAX_DISTANCE [angstrom] 1.0
> >       &END SHELL
>
> don't you see anything strange from the charge balance of your system???
> So .. you have a Si atom which has a charge of 2.7226
> and an oxygen which has a charge of -1.3613.
> SiO2 must be neutral.. unless you don't do some exotic type of SiO2.
> Let's assume for the moment that you do the standard SiO2.
> So the balance of charge is 0= 2.7226-2*1.3613 without considering
> the contribution of
> core-shells...
>
> Strange enough you put then a core-shell on the oxygen with the
> following charges:
> CORE    1.91981
> SHELL   -3.28111
>
> the core-shell gives an additional charge on the O atom of -1.3613..
> So.. can you explain me what are you simulating?
> SiO2 with your settings has a charge of -1.3613
>
> Think about it....
> Maybe you want to set the charge of O to zero and keep the charge
> coming from the core-shell...
> It must be you to fix this controversial settings!
>
> >       &EWALD
> >         EWALD_TYPE spme
> >         ALPHA 0.4
> >         GMAX 61 61 61 ! just for trying
> >        !O_SPLINE 2
> >       &END EWALD
> >     &END POISSON
> >   &END MM
> >   &SUBSYS
> >     &CELL
> >       ABC 14.4 14.4 14.4
> >       UNIT ANGSTROM
> >     &END CELL
>
> a mesh of 61x61x61 for 14 angstrom box is maybe a bit exaggerated!
> why do you want to have such an accuracy on the electrostatic and keep
> the spline accuracy to 10^-6 - 10^-7.. ??
>
> Think about it...
>
> > &MOTION
> >   &MD
> >     ENSEMBLE NPT_F
> >     STEPS 1000
> >     TIMESTEP 0.2
>
> TIMESTEP of 0.2 is already a quite good setting.. BUT... keep in mind
> that the core-shell may require even smaller integration time steps,
> due to the high frequency of the core-shell model..
> It's up to you to check that you have a properly conserved total
> energy in your simulation...
>
>
>
> Teo

On 2 oct, 11:45, Laino Teodoro <teodor... at gmail.com> wrote:
> ... and now most important few comments on your input file:
>
>
>
> >   &MM
> >     &FORCEFIELD
> >       &CHARGE
> >         ATOM Si
> >         CHARGE 2.7226
> >       &END CHARGE
> >       &CHARGE
> >         ATOM O
> >         CHARGE -1.3613
> >       &END CHARGE
> >       &SHELL O
> >         CORE_CHARGE 1.91981
> >         SHELL_CHARGE -3.28111
> >         MASS_FRACTION 0.1
> >         SPRING [eV*angstrom^-2] 256.71027
> >         MAX_DISTANCE [angstrom] 1.0
> >       &END SHELL
>
> don't you see anything strange from the charge balance of your system???
> So .. you have a Si atom which has a charge of 2.7226
> and an oxygen which has a charge of -1.3613.
> SiO2 must be neutral.. unless you don't do some exotic type of SiO2.
> Let's assume for the moment that you do the standard SiO2.
> So the balance of charge is 0= 2.7226-2*1.3613 without considering
> the contribution of
> core-shells...
>
> Strange enough you put then a core-shell on the oxygen with the
> following charges:
> CORE    1.91981
> SHELL   -3.28111
>
> the core-shell gives an additional charge on the O atom of -1.3613..
> So.. can you explain me what are you simulating?
> SiO2 with your settings has a charge of -1.3613
>
> Think about it....
> Maybe you want to set the charge of O to zero and keep the charge
> coming from the core-shell...
> It must be you to fix this controversial settings!
>
> >       &EWALD
> >         EWALD_TYPE spme
> >         ALPHA 0.4
> >         GMAX 61 61 61 ! just for trying
> >        !O_SPLINE 2
> >       &END EWALD
> >     &END POISSON
> >   &END MM
> >   &SUBSYS
> >     &CELL
> >       ABC 14.4 14.4 14.4
> >       UNIT ANGSTROM
> >     &END CELL
>
> a mesh of 61x61x61 for 14 angstrom box is maybe a bit exaggerated!
> why do you want to have such an accuracy on the electrostatic and keep
> the spline accuracy to 10^-6 - 10^-7.. ??
>
> Think about it...
>
> > &MOTION
> >   &MD
> >     ENSEMBLE NPT_F
> >     STEPS 1000
> >     TIMESTEP 0.2
>
> TIMESTEP of 0.2 is already a quite good setting.. BUT... keep in mind
> that the core-shell may require even smaller integration time steps,
> due to the high frequency of the core-shell model..
> It's up to you to check that you have a properly conserved total
> energy in your simulation...
>
>
>
> Teo