Some questions about what I have met when I use the THERMAL CONDUCTIVITY SCRIPTC simulated in Forcite module
Dear all, There are some questions about what I have met when I use the THERMAL CONDUCTIVITY SCRIPT, which run on the Materials Studio software. I hope that you can give me some suggestions to help me solve the problems. Thank you very much!
Firstly, I use the Materials Studio2017 Forcite module to measure the ability of the SWCNT (Single-Walled Carbon Nanotubes) to transfer heat. In other words, it’s the thermal conductivity of the SWCNT. The simulation is driven by script written in the programming language PERL, referring to the example of argon at density 0.2652 g/cc and temperature 500 K that Reiniera provides.(Appendix)
The thermal conductivity of (6,6) single-walled carbon nanotube (SWNT) with the length of 245.95nm, the diameter of 8.14 Å, the thickness of 3.34Å(diameter of carbon atom and I change the cellarea to 85.37 Å in line 201) is 197.70W/m•K. That’s different from the value of 341.07 W/m•K which I do the simulation using the Lammps software with the same SWNT model.
Then, I change the temperature (line 108 in the Perl script) to study how the thermal conductivity of the SWCNT change with the temperature. Disappointingly, the thermal conductivity has almost no change with the temperature change(show in the figure1).Actually, we use the LAMMPS to simulate and get the result that the thermal conductivity of the SWCNT decreases with the increase of temperature(show in the figure2).
So sorry! The length is 245.95Å. I used "Universal" forcefield in Forcite module in all simulations. I just changed the temperature (line 108 in the Perl script). The method was the Reverse Non Equilibrium Molecular Dynamics (RNEMD) method.
The appendix files include the CNT model,Perl script and output.
The temperature profile looks rather noisy, with a temperature variation of just a few degree over the length of the tube:
I think you need much longer runs (NumExchangesEqui, NumExchangesProd), and possibly more frequent exchanges (NumberOfSteps) to get an accurate gradient.
This should give a smoother λ(T) profile, but I am not sure it would give much better agreement with LAMMPS. Did you use the same cross sectional area (85.37 Å²)? Does LAMMPS provide the gradient and flux data? That might tell us something.