[lammps-users] Thermal conductivity of Ar-Cu nanofluids usingfix thermal/conductivity command

OK, thanks.
I’ve tried “compute heat/flux” to reproduce the result of the paper, but I failed.

I use compute heat/flux to compute the thermal conductivity of Ar at 85K with dimendionless density as 0.844, and the result is about 0.167, which is higher than Sarkar’s result 0.127 and experiment result 0.132.
But when I use the old edition of lammps, I got the result about 0.13, I don’t know what’s wrong with the latest edition of “compute heat/flux command”.
And also even if I use the old edition of “compute heat/flux”, the thermal conductivity of Ar-Cu nanofluids is about 10 which is far more than Sarkar’s result (0.15~0.2), I don’t konw the reason.
The input script is attached.

10年4月15日,周四, Tengfei Luo <[email protected]…64…> 写道:

Ar.in (1.52 KB)

ArCu.in (1.87 KB)

They updated the “compute heat/flux” by using the correct velocities. It’s hard to get an accurate K through a single GK run. You need to do ensemble average and you have to find the correct convergence. Try to run your job with different initial velocities for 10 times. Also, you forgot to substract the enthalpy term in the heat flux for your liquid system.


I notice that Sarkar compute average enthalpy as the sum of the average kinetic energy, potential energy, and average virial terms per particle of each species, so I modify the compute_heat_flux.cpp to include the enthalpy items in the following:

But I don’t konw if I’m right, so could you please help me check it, thanks.

And I also notice that the melting point of Ar is 83.80K, and its boiling point is 87.30K, so as for 85K in my simulation, the state of Ar is liquid. But to use Muller Plathe method, or other NEMD method, to establish temperature in the simulation system, I think the hot and cold will all outside the liquid state, i.e., the hot end maybe gas state and cold end may be solid. What should I do to use NEMD method to compute the thermal conductivity of Ar at 85K?

In this case, you’d better stick to GK. Any specific reason why you need to use NEMD? I’m not that into c++, so I cannot check the code. But I’m sure there are experts in the list who can check it for you.



OK, thanks, the reason is that I can’t get good result using GK, and then I turn to NEMD, but the result is also not so satisfactory.

Xiaoliang Zhang

10年4月15日,周四, Tengfei Luo <[email protected]…64…> 写道: