# Re: Langevin damping coefficient

From: Anton Arkhipov (anton_at_ks.uiuc.edu)
Date: Wed May 06 2009 - 10:25:49 CDT

Hi Gianluca,

The damping coefficient introduces friction, so it does influence an
MD simulation. In principle, you want to set the damping coefficient
as low as possible, because if its value is high, atoms will
experience too much unnatural friction. However, you cannot make the
coefficient too low, because in that case your system will fluctuate
too far away from the temperature you want.

> How much does for example a value of 10/ps slow down dynamics versus
> a value of 5/ps?

Friction is going to be 2 times stronger.

I saw some people using the damping coefficient as low as 0.5 ps^(-1),
and like you said as high as 10 ps^(-1). In my opinion, 5 ps^(-1) is
probably the highest value one can use for an all-atom simulation. The
reason is that with damping coefficient set to 0, water diffusion (for
TIP3P) is a few times faster than it should be (compared to
experimental measurements), while experience in our group shows that
using 5 ps^(-1) results in the water diffusion coefficient being close
to the experimental value. So, if you are using damping coefficient
higher than 5 ps^(-1), you water diffusion is going to be slower than
it should be.

On the other hand, even though 5 ps^(-1) gives you water diffusion
that is about right, it may slow down other processes. So it still
might be too high. I personally have been using 2 ps^(-1) for a while,
and it seems to work pretty well, even though water may diffuse a bit
too fast.

Anton.

On 5 May 2009, at 16:49, Gianluca Interlandi wrote:

> How strongly does the Langevin damping coefficient influence a
> molecular dynamics simulation?
>
> After quickly browsing the literature I see that values between 1/ps
> and 10/ps are used where most people use 5/ps. How much does for
> example a value of 10/ps slow down dynamics versus a value of 5/ps?
>
> Thanks,
>
> Gianluca

This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:52:44 CST