espressomd-users
[Top][All Lists]

## Re: [ESPResSo-users] Choice of timestep and langevin-gamma

 From: Georg Rempfer Subject: Re: [ESPResSo-users] Choice of timestep and langevin-gamma Date: Tue, 19 Sep 2017 21:42:44 +0200

There is a feature called LANGEVIN_PER_PARTICLE, that allows you to set different gammas for different particles.

On Tue, Sep 19, 2017 at 4:38 PM, Clemens Jochum wrote:
Dear all,

I just noticed a mistakeI made. In the user-guide it is stated that
gamma is not the inverse relaxation time tau anymore, but it is now
defined as mass / tau. In this case if I want to achieve tau^-1 = 1.5
ps^-1 for a particle of mass2 = 660, I have to set gamma = 100, which
seems high.

At the same time the other particles with mass1 = 20 will have tau^-1 =
50 ps^-1 (mass2 / mass1 = 33). But with the way the Langevin-thermostat
is implemented in ESPResSo it is not possible to prevent this large
difference between the two particle types with different masses.

Could this lead to some problems? Or is there a way to assign a fixed
value to tau and tau^-1 for all masses in ESPResSo?

Best,
Clemens

> The other parameter I am concerned about is gamma of the
> langevin-thermostat. So far I just used gamma = 1. = 10 ps^-1, which as
> I recently read is a little bit high. In my system I simulate a charged
> macromolecule with counter-ions in an implicit solvent (water). Is a
> value of gamma = 0.2 = 2 ps^-1 a good fit?

On 19.09.2017 13:11, Clemens Jochum wrote:
> Dear all,
>
> I read previously that for a system with Temp = 1. and mass = 1. a good
> value for the timestep is 0.01. I have Temp  = 2.5 and mass1 = 20 and
> mass2 = 660. Is it correct to assume that I can scale the timestep
> inversely to the thermal velocity with a factor of (m / kT)^(1/2) ~ 3
> (using the mass1)? I have been using a timestep of 0.05 = 5e-15 s so far
> without any major problems.
>
> The other parameter I am concerned about is gamma of the
> langevin-thermostat. So far I just used gamma = 1. = 10 ps^-1, which as
> I recently read is a little bit high. In my system I simulate a charged
> macromolecule with counter-ions in an implicit solvent (water). Is a
> value of gamma = 0.2 = 2 ps^-1 a good fit?
>
> Best,
> Clemens
>
>
>
>
>
>