Order of atoms in Fist input file changes result?

[Matthias Krack]

Dear Johannes,

I think this is due to the initialisation of the atomic velocities which is 
eventually based on pseudo-random numbers. If you switch two atoms in the 
list, then you switch also their initial velocities and their positions 
will be different in the next MD step. Consequently the total energy will 
be different, tool. The pseudo-random numbers are deterministic and thus 
for the same atomic sequence the result is reproducable, if you alter the 
sequence, however, this is no longer true.

Matthias


On Thursday, 5 March 2015 18:51:13 UTC+1, Johannes Wagner wrote:
>
> Dear all,
>
> maybe I am missing something but I recently had to realise that when I 
> change the order of coordinates of my inpuf file in a FIST run, but leave 
> anything else the same, I get another result (quite substantially from time 
> to time)
>
> e.g. in the al2o3.inp from /cp2k/tests/Fist/al2o3.inp
>
> If I run this file for one step I get
> #     Step Nr.          Time[fs]        Kin.[a.u.]          
> Temp[K]            Pot.[a.u.]        Cons Qty[a.u.]        UsedTime[s]
>          0            0.000000         1.768042507       
> 298.000000000      -281.499021201      -279.730978694         0.000000000
>          1            3.000000         1.760113969       
> 296.663661049      -281.491144428      -279.731030458         0.80785918
>
> If I run it again, putting all the Oxygens and their coordinates first, I 
> get
> #     Step Nr.          Time[fs]        Kin.[a.u.]          
> Temp[K]            Pot.[a.u.]        Cons Qty[a.u.]        UsedTime[s]
>          0            0.000000         1.768042507       
> 298.000000000      -281.499021201      -279.730978694         0.000000000
>          1            3.000000         1.777233556       
> 299.549132838      -281.508050366      -279.730816810         0.802598953
>
>
> Is this expected behaviour? Have I missed something? I am using the latest 
> svn checkout. I figured since everything is index with the Atomic Kind, it 
> shouldn't matter where in the coordinate list an atom appears, maybe this 
> was wrong.
> If this is normal, what is the "correct" way of ordering the atomic 
> coordinates? 
>
> thank you all for any hints
>
> all the best,
> Johannes
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20150305/909aa776/attachment.htm>