My C++ skills are about 0, so I doubt I could pull it off myself. However, I came across a paper which appears to fix some of the problems with the RNEMD method of Muller-Plathe. Instead of swaps, it uses scalings, but is essentially very similar. Wondering if/when this would be put into lammps (or if someone has written something to augment my current copy).
Kuang S, JD Gezelter “A gentler approach to RNEMD: Nonisotropic velocity scaliong for computing thermal conductivity and shear viscosity” J. Chem. Phys v:133, 164101 (2010).
My C++ skills are about 0, so I doubt I could pull it off myself. However,
I came across a paper which appears to fix some of the problems with the
RNEMD method of Muller-Plathe. Instead of swaps, it uses scalings, but is
essentially very similar. Wondering if/when this would be put into lammps
(or if someone has written something to augment my current copy).
please remember that in most community supported open source projects
people implement what they need personally, or what seems useful, provided
that that the effort is not too large. thus, if you want somebody to implement
a feature for you in such a package, you have to make a compelling case
or offer some incentives.
I haven't looked at the new paper. But if you
are worried about "gentle" then if you read the
fix viscosity doc page carefullly, there are several
settings you can play with to make it more gentle.
E.g. N, swap, vtarget.
Also, you may want to try just using regular NEMD
in LAMMPS. Lots of people do that for polymeric
systems. See section 4.13 of the Manual.