Matt,
I’ve been reading over your paper. It’s a very interesting topic, thanks for posting the link. Would it be possible to use Muller Pathe to determine thermal conductivity using the SRD method? In your model you swap max and min velocities in the x-direction only to create a velocity profile. Could you in principle swap the hottest and coldest momentum to create a thermal gradient from which you could derive the thermal conductivity of the system? This of course a standard method for atomistic simulations but would it have physical meaning in this context? For instance, I presume it ignores any thermal transfer between the particle and the solvent.
I’d be interested to hear your thoughts.
Thanks,
Dave
Yes, you can use M-P with an SRD system. As I recall,
it works better is you swap momenta of the nano-particles
not the (zillions of tiny) SRD particles. You can also
do a NEMD model of a nanoparticle/SRD system in
a sheared box to get a viscosity in the usual way.
Steve
Steve - I think it's the other way around. That is, the momentum swap works better among the SRD particles. I don't remember how well it worked when just the nanoparticle momentum was swapped.
Matt
You might be right Matt. If there are too few nanoparticles,
then swapping on them could be overkill. The mental picture
I have is a few big bowling balls in a sea of ping-pong balls.
The M-P in LAMMPS has the option to do multiple swaps
per invocation, so if the SRD particles swaps are not
effective at imposing a drift velocity on the bowling balls,
you could always ramp up the number of swaps. At some
point though you could overwhelm the velocity rotation
operator acting on the SRDs and alter the background
fluid properties.
Steve
. At some
point though you could overwhelm the velocity rotation
operator acting on the SRDs and alter the background
fluid properties.
There was also the effect on the measured viscosity from the chosen velocities of the swapped SRD particles. If I recall correctly, we found that the measured viscosity converged to the theoretical viscosity only when many swaps of very low velocity particles were used for a given total momentum swap per timestep. And we never came up with a convincing reason for that. Anyway, I think there is likely to be a related problem with the thermal conductivity method given how similar it is to the viscosity method.
Matt