I am using fix_langevin and I am trying to understand the origin of the fluctuation-dissipation relation in the white noise. Looking at the source code I think LAMMPS uses
24*kB*m*T*xi
instead of the more commonly used
2*kB*m*T*xi
or sometimes
6*kB*m*T*xi
where xi is the friction term or the inverse of the damping period. Is my interpretation of the source code correct? Are there any citations that would help me to resolve the apparent discrepancies?
Gary Grest helped me out here. LAMMPS uses
uniform random numbers for Langevin
thermostatting (instead of Gaussian) b/c they're cheaper.
I believe 24 is the correct normalization factor in that case.
This paper has details, but I can't get easy access to it:
4. Brownian dynamics simulations without Gaussian random numbers.
Dunweg, B.; Paul, W.
Source: International Journal of Modern Physics C (Physics and
Computers); Sept. 1991; vol.2, no.3, p.817-27