scale keyword in fix langevin

Hello everybody,

I am simulating a system of polymers and nanoparticles (NPs) using brownian dynamics implemented via fix nve + fix langevin.

At the beginning, I set a single ‘damp’ parameter for all the species in the systems (monomers and NPs), but then I realized that by doing so the viscosity of the system will not be well-defined, because the small monomers and the large NPs will experience the same friction, and for the Stokes law the fritction coefficient (m/damp) should be proportional to the radius of the particle.

I then decided to scale the ‘damp’ parameter with a factor m/R (m=mass, R=radius) using the ‘scale’ keyword of fix langevin. This should give a well-defined, unique viscosity to the implicit solvent I am simulating. To be more precise, a free monomer and a free NP freely diffusing now satisfy the Stokes-Einstein relation with the same value of the viscosity (which I verified).

But then, reading many articles about simulations of polydisperse systems, many of which realized with LAMMPS, I found no mention of the need to use different ‘damp’ parameters for particles of different size when using a langevin thermostat.

So I began wondering: is my reasoning correct?

Thank you very much,


Not really a LAMMPS Q, but maybe someone will answer …


Not really a LAMMPS Q, but maybe someone will answer ...

​i don't have an answer to offer but a speculation. using fix langevin to
model an implicit solvent is a quite large approximation as it is, so i can
imagine that some people consider applying this scale factor as not
significant. there also is the chance that a) the publications predate the
implementation of the scale keyword in LAMMPS or b) that the people
publishing were not aware of it. the LAMMPS documentation is large and some
people have a tendency to only study it thoroughly, when they have problems
(and some don't even do that :-().

on the other hand, LAMMPS also offers more sophisticated generalized
langevin thermostat/integrators, e.g. in fix gld or fix gle, that may be
worth considering when worrying about the plain langevin ​implementation.


Thank you for your answer. Yes, it may be that the approximation is already so severe that they don’t really care…

In any case, I will take a look at those other thermostats.

All the best,