Re: [ESPResSo-users] a kind request for help

[Axel Arnold]

Dear Dusan,

please use the Espresso mailing list for any questions regarding Espresso, 
other people might also be interested in the answers.

On Thursday 09 August 2012 14:44:52 you wrote:
> Well, the question is if there is a possibility to change the "seed" for
> random number generations.

There are in fact three random generators involved, two that Espresso 
implements and the standard Tcl one. The two Espresso RNGs are documented on 
page 126ff of the User's guide, and both support a "seed" command to set the 
seed(s), according to the guide. The plural applies to multi-node simulations, 
where you have to specify the seed for each processor. The Tcl RNG (expr 
rand()) is seeded by expr seed(n) and not parallel. The Tcl RNG is initialized 
randomly, while the Espresso RNGs rely on the user to choose a reasonable seed 
(e.g. [pid]).

> I performed a series of MD simulations on polymers with different
> stiffnesses and I observed
> the same time evolution of fluctuations - the peaks lie at the same
> positions in time, however they
> have a smaller amplitude.

If anything changes, you probably don't use the same random numbers...
 
> I talked already about with Peter Kosovan. He admited, the seed is an
> internal parameter, but we
> agreed the observed evolution was peculiar.

The seed is not an internal parameter, see above.

Axel

-- 
JP Dr. Axel Arnold      Tel: +49 711 685 67609
ICP, Universität Stuttgart      Email: address@hidden
Pfaffenwaldring 27
70569 Stuttgart, Germany