testing on simulation cell

[huan... at mail.huji.ac.il]

Dear All,

I did some testing on the size of simulation cell on my benzene-water 
cluster. Actually, my system is not a periodic system, so my setting is 
PERIODIC NONE in both &POISSON and &CELL sections.

The structure of my benzene-water cluster was optimized by Gaussian09 and 
can be put into a cell with size as 5*5*5. I tried 7*7*7, 10*10*10, 
15*15*15 and 20*20*20 cells for NVE calculation in CP2K with PBE 
functional. However, only cell with 20*20*20 can maintain the structure of 
benzene-water cluster. For the rest of results with smaller cell, all the 
structures break due to the dramatic increasing of temperature from 50 K to 
10000 K after 30 steps.

I have no idea about these testing results. Why could not use smaller cell?

I consulted the CP2K reference and found a description on &MT section. *Sets 
up parameters of Martyna-Tuckerman poisson solver. Note that exact results 
are only guaranteed if the unit cell is twice as large as charge density 
(and serious artefacts can result if the cell is much smaller). 
*
In addition, I noticed that the REL_CUTOFF value is related to the &MGRID 
section. *Specify the multiplicative factor for the CUTOFF keyword in 
MULTI_GRID section. The result gives the cutoff at which the 1/r 
non-periodic FFT3D is evaluated.Default is 2.0*
*
***However, both of these lines do not have any detail explanations or 
examples. The CP2K reference is really too poor for a beginner to learn how 
to use CP2K, comparing the Gaussian reference. I do not understand what is 
the meaning of "if the cell is twice as large as charge density". How to 
check the charge density of my system? And also, is there any rule of thumb 
to choose the value of REL_CUTOFF?

Best regards,
Huan




**
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20131028/a78d1c11/attachment.htm>