# Selection of appropriate Langevin damping parameter for MD of solid metal

I am performing an MD simulation involving solid metal. The system is in the NVT ensemble, and I am using the Langevin thermostat (fix Langevin in LAMMPS) to enforce this.

I have previously used the Langevin thermostat on ions present in an implicit solvent. The damping parameter was straightforward to set, because I was able to connect it to the viscosity of the fluid.

However, I am trying to understand what the analog of viscosity would be for the atomic sites making up a solid metal. Does anyone have experience/pointers with determining an appropriate Langevin damping constant in this case?

• +1 for the question! I've sent this question to Prof. Michele Ceriotti, who knows a lot about Langevin dynamics and MD. Hopefully you'll get an answer from him, or someone else soon! Commented Feb 2, 2021 at 22:44
• Commented Mar 10, 2021 at 16:36
• Also related on making a Langevin thermostat work efficiently for a solid: mattermodeling.stackexchange.com/q/9374/4515 Commented Jul 21, 2022 at 21:37

If you wish to model dynamical properties of the system, the issue is much tougher. You correctly note that there exists a connection with the viscosity of the surrounding medium, but this doesn't make all too much sense on/inside a metal. In a liquid, diffusion and viscosity are correlated, but in metals the diffusion of point defects is actually a governed by discrete jumps. So this connection breaks down. Generally speaking I would avoid calculating dynamical properties with Langevin dynamics, and use a global thermostat instead (like fix nvt or fix temp/csvr).
To return to the first application, there are ways to more systematically optimize the performance of the thermostat, so that it most efficiently samples thermodynamic averages. Whereas a normal Langevin thermostat (fix langevin or fix temp/csld in LAMMPS) adds random white noise to the dynamics, you can also add colored noise, using fix gle. This method implements several ideas developed in the group of Michele Ceriotti. One possible application is "optimal sampling", which guarantees optimal coupling of the thermostat to the system, but there are also options to mimic nuclear quantum effects or vibrational nonequilibrium. There is a dedicated web site that has all the references and documentation.