'fix Langvein' command fails to control temperature

Hello everyone!
Failure of temperature control In reproducing a literature on calculating thermal conductivity, the ehex velocity rescaling method NEMD was used to calculate the thermal conductivity, and the temperature gradient distribution was normal, but since the temperature difference between the two ends could not be controlled to be the same in repeating the experiments, it was switched to the langvein temperature difference to calculate the thermal conductivity.
However, in the process of simulation, it was found that the temperature distribution was always very turbulent, resulting in the calculation of the results are not useful at all!

yours sincerely


cnt_atom.data (23.5 KB)
edge_20_all.temp (47.5 KB)
edge_langvein.in (7.7 KB)

What is a turbulent temperature distribution?

Please note that a graph without axis is useless, and that fluctuations are normal in molecular simulations. If you want to convince people that there is an actual problem with the implementation of fix langevin, you will need to show that the fluctuations you are observing are truly out of the ordinary.

Dear simongravelle
I’m so sorry, “distribution was always very turbulent” is my expression error, it means the temperature distribution is so chaotic that it does not exhibit a gradient distribution as I expected.
The y-axis is temperature(K) and x-axis is the chunk number (Along the carbon nanotubes with scale) . By my expected simulation results, it should be similar to the temperature distribution in the following figure. However, the result is so terrible…

Thanks again for the detailed explanation!

If you don’t get expected results there are two possibilities: 1) there is a bug in the software or 2) your simulation setup is bad (geometry, settings, parameters, force field).

Both of the fixes you reference have seen extensive testing and used successfully many times, especially fix langevin. Thus the - by far - most likely case is 2). That in turn means, that it is your job to correct bad choices. You cannot expect that someone here will take your inputs and will debug and correct them for you.

Let me reiterate; the fact that you don’t get the results you expect, does not make it automatically a LAMMPS problem.

Dear akohlmey
Thanks for your detailed explanation!
For the second possibility, the geometry, settings, force field in case are both refer to the literature and LAMMPS examples( in.langvein). As u say, I should care about the accuracy of my case and I will try to find the error again.
yours sincerely

I’m encountering a similar issue with thermal conductivity calculations. The temperature distribution appears erratic and doesn’t exhibit the expected gradient.