[CP2K-user] Some inconsistencies in reproducing the virial kinetic energies

[am...@gmail.com]

As an update, I could figure it. With Eq (5) in the provided paper, the 
virial kinetic energies and the values which CP2K would be equal if for the 
term 3/2KT the temperature T is always almost a fixed value and equal to 
the temperature at which the simulation is done. 
It means there should be some bug or different meaning for the temperature 
and kinetic energy which CP2K prints in the first three columns.


On Sunday, May 30, 2021 at 6:21:11 PM UTC+2 am... at gmail.com wrote:

>
> Hi everyone,
>
> In running some path integral MD simulation in cp2k I have observed some 
> strange issues and would be grateful if someone can comment on that.
>
> After running the attached input script main.inp, the energy terms are 
> printed in a file named “md-01-energy-1.dat”. 
>
> The first issue which I encountered is that the kinetic energy printed out 
> in the third column can no way be related to the calculated temperatures 
> printed in the 5th column. Regardless of the degree of freedom and unit 
> conversion factors, at least there should be a correlation between them 
> which doesn’t exist here.
>
> The second and more important issue: The calculated virial kinetic energy 
> which is printed in the fourth column of “md-01-energy-1.dat” file also 
> seems kind of strange. This value is calculated via the virial estimator 
> written as equation (5) in the attached pdf file. 
>
> To reproduce what CP2K prints, I manually calculated the second term in 
> right hand side of eq (5) using the position of the centroids and beads 
> printed out by the script and for the gradient of the potential energy 
> required by eq(5) I also used the forces which CP2K prints. 
>
> Something really surprising and unusual which I observed is that the 
> second term in right hand side of eq (5) which I manually calculated 
> *before* adding 3/2KT to it (as should be done based on the virial 
> estimator) shows a perfect correlation with the total virial kinetic energy 
> which CP2K prints (R=1.00) (although for this the virial kinetic energies 
> which CP2K prints are ca. 438 times the values which I calculate and I 
> couldn’t find the reason even with trying various possible unit 
> conversions). 
>
> If I also add the data I obtain with 3/2KT which should be done to yield 
> the total kinetic energies, then the correlation coefficient dramatically 
> reduced. 
>
> I would highly appreciate your comments/helps.
>
> Best regards,
>
> Amin
>
>  
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20210531/a958a6ba/attachment.htm>