[CP2K:1416] NPT simulations

[Laino Teodoro]

Dear Tomaso,

> I did a simulation on 128 H2O box applying both thermostat and
> barostat, so NPT ensamble.
> I am not using the latest cp2k (21 july 2008 version).
the suggestion of course is to try to use always an up2date CVS  
snapshot,
whenever you feel like there's something strange in your numbers..

> I have 2 questions: the first one is about the barostat. Although I
> went through the discussion there was already in this group (under
> "barostat"), and read the reference paper (Martyna et al, JCP 101,
> 4177 (1994)), and used the suggested setup I found in the test file
> cp2k/tests/Fist/H2O-32_NPT.inp
> I see in my restart file run-1.restart at the end of the run that I
> have 2 thermostat, one in the BAROSTAT section and one in the
> THERMOSTAT section, with different coors, masses and velocities, and
> honestly I still do not understand why one has 2 thermostats. My input
> file is at the end of this post.

The isothermal-isobaric ensemble is generated with the use of an  
extended Lagrangian
for the barostat. This means that the barostat is seen like an  
additional degree of freedom
with its own masses.
Particles and barostat are often coupled to the same chains of  
thermostats, although
for generality we prefer to have them coupled to independent  
thermostats.
All these things are  fully explained in the paper you mentioned.
Anyway the answer to your question ("why 2?") is for being more  
flexible and general..

> The second question is about some strange energy behaviour during the
> run. I took the simulations coordinates and velocities from a previous
> NVT simulations, and it looks like the setup I chose (BLYP, DZVP
> basis) results in a more dense liquid. As the box shrinks there are
> some large jumps in the total energy: for example this is a part of
> the run-1.ener file:
>
>         33              16.500         0.577002245
> 317.150009889     -2199.101322770     -2198.365847076
> 82.407300000
>         34              17.000         0.579470429
> 318.506650669     -2199.103374256     -2198.365814097
> 93.484800000
>         35              17.500         0.584554875
> 321.301322756     -2199.107827624     -2198.365870421
> 94.342100000
>         36              18.000         0.591370574
> 325.047580217     -2198.642282517     -2197.894050509
> 101.114600000
>         37              18.500         0.598376710
> 328.898511501     -2198.649038164     -2197.894118726
> 122.682700000
>         38              19.000         0.604204462
> 332.101743054     -2198.654572558     -2197.894131860
> 111.490300000
>
> as you see from step 35 to 36 the energy makes a huge jump.
> I cannot attach the whole output files, but there are other jumps in
> energies, and they always happen at given values of the cell size: if,
> due to fluctuations, the cell size crosses again the value at which I
> had seen an energy jump, the energy goes back to its previous value:
> again from the energy file (see steps 1123-1124 and 1133-1134)

These jumps are due to the plane waves used to evaluate the coulomb
interactions..
if you increase your cutoff jumps will become smaller..
I'm sure Juerg or other people will be happy to comment extensively  
on this point..

Teo


>
>
> Hope this is enough to get some help.
> thanks
> tomaso