Langevin MD simulation friction variation

[Lim changmin]

Hello,

I'm currently performing melt-quench molecular dynamics simulations to generate amorphous structures using the Langevin thermostat. In the Langevin method, there's a friction parameter (with a default value of 0.01).

When I use the default value, the system's temperature rapidly rises to unphysically high values like 1e12 K or 1e11 K. However, when I increase the friction coefficient to 0.05 or 0.1, the temperature starts to fluctuate reasonably well around the reservoir temperature.

Still, for some structures, the temperature control is not stable enough. So I'm considering increasing the friction coefficient further to 0.15 or even 0.2.

Here are my questions:

Would increasing the friction coefficient to this extent affect the resulting amorphous structure?

From a theoretical standpoint, how much can I increase the friction coefficient before it starts to significantly affect the dynamics and structural outcomes? Is there a practical or theoretical upper limit?

Kindly waiting for the response.

[Anders Blom]

I recommend reading https://www.tandfonline.com/doi/abs/10.1080/08927022.2020.1791858.
It doesn't look like there any major problems to increase up to 1 ps^-1, but do note that a higher friction necessitates a lower time step, and each application needs to be evaluated individually.