espressomd-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [ESPResSo-devel] Mass Dependent Diffusion in Langevin Thermostat


From: Joost de Graaf
Subject: Re: [ESPResSo-devel] Mass Dependent Diffusion in Langevin Thermostat
Date: Thu, 18 Jun 2015 14:21:10 +0200

Dear All,

After a bit of further discussion, the following has become clear. Apparently, the number which you specify in the thermostats is the inverse relaxation time, which is mass dependent. So if you want to get the same friction coefficient, you'll need to take the mass into account. After browsing through the User's Guide some more, I'm pretty sure that this is not the behavior ESPResSo advertises. Moreover, I am fairly certain that there are now inconsistencies in the code, because some features (a.o. ENGINE) have based their mass dependence on the idea that the "friction" is a proper friction, not an "inverse relaxation time", to make sure that there is no odd mass dependency. Pretty much a fail all round. I'll open a bug report on it.

Kind Regards,

Joost

On 18 June 2015 at 14:08, Ulf Schiller <address@hidden> wrote:
I agree and the obvious source is the Langevin noise - one only has to
look at the units to see that it's wrong...

On 18/06/15 10:42, Joost de Graaf wrote:
> Dear All,
>
> Just a quick check before I open a bug report. In my understanding the
> friction coefficient in the Langevin thermostat is supposed to represent
>
> $\zeta = 6 \pi \eta R$
>
> and therefore, the diffusion coefficient in LD should not depend on the
> mass of the particle. However, in ESPResSo, the behavior is different,
> see attached MSD.png. Is this because we choose a different way to
> define the input parameter for the Langevin thermostat? I did not find
> anything on it in the UG, so I think the ESPResSo behavior is wrong --
> but if anyone can come up with a good argument as to why it should be
> this way, I'm listening.
>
> If such fundamental things are broken, we should check the rotational
> diffusion, and other simple aspects of ESPResSo as well.
>
> Kind Regards, Joost


--
Dr Ulf D Schiller
Centre for Computational Science
University College London
20 Gordon Street
London WC1H 0AJ
United Kingdom

Phone: +44 (0)20 7679 5300



reply via email to

[Prev in Thread] Current Thread [Next in Thread]