[gmx-users] timestep

[Erik Lindahl]

David van der Spoel wrote:
> On Mon, 25 Mar 2002, Anton Feenstra wrote:
> 
> 
>>Rainer Boeckmann wrote:
>>
>>>maybe that's the problem. I did in fact only change the timestep. But: how do
>>>parameters like nstlist affect the energies at step 0? The kinetic energy at step
>>>0 is different fot two runs starting from the same gro-file but with different
>>>delta t. Thought these parameters were only important for later steps...
>>>Thanks a lot,
>>>
>>I can't see any relation between nstlist and properties at step 0,
>>but dt will actually affect these. The reason is that in the trajectory
>>(or .gro file) coordinates are stored at whole time steps and velocities
>>at half. Changing the timestep will change the 'interval' between the
>>time of the velocities and the coordinates. This might give rise to some
>>artefacts...
>>
> There is one relation: initial shake.

Another potential culprit is the temperature coupling algorithm; 
Nose-Hoover will give you a closer equilibrium temperature, but each 
time you start your simulation the dynamical heat bath variable is set 
to zero, and you might see some fluctuations before it has settled down.

Cheers,

Erik