EHEX

Dear LAMMPS users,

I tried to calculate thermal conductivity by NEMD simulation by applying Ehex and Hex methods.
I modified the example code to use real units for all-atom polymer models.

I used the value of energy flux into the reservoir https://lammps.sandia.gov/doc/fix_ehex.htmlhttps://lammps.sandia.gov/doc/fix_ehex.html

equal 0.075 and 0.15 as suggested in the LAMMPS documentation.

I obtained rather good temperature difference for cold and hot layers on second equilibration step,

Step Temp c_Thot c_Tcold

96000 249.81691 309.96814 187.69341
97000 249.8308 315.17475 189.87639
98000 250.45888 308.53876 188.77604
99000 250.15357 310.0092 191.03955
100000 249.88901 314.70286 186.27202
101000 249.82089 314.98229 189.73962

however, when thermal conductivity is calculated

Step Temp c_Thot c_Tcold v_tdiff f_ave

211000 250.34183 318.81244 183.16881 -2.7927217 -1.5338256
212000 249.46583 327.51738 178.23532 -4.088393 -1.5566343
213000 250.5441 324.96023 180.77809 0.72385968 -1.5364529
214000 249.70081 328.91091 181.15441 0.451866 -1.5190115
215000 249.94985 316.56335 179.66115 -0.8909564 -1.5135502
216000 249.91148 323.04664 181.80206 -1.3746355 -1.5123526
217000 248.94808 326.45275 184.7369 0.6222239 -1.4941084
218000 249.12776 320.9257 183.63537 0.79814159 -1.4746825
219000 249.783 322.86491 181.29182 -2.1519487 -1.4803738
220000 249.44366 318.10454 179.22659 -5.3308305 -1.512461
221000 249.64709 324.88001 181.65959 -5.7936214 -1.5478425
222000 249.32393 312.35475 179.54857 -4.3969411 -1.5711957
223000 249.14977 329.25245 180.41405 -3.4935655 -1.5868248
224000 249.21498 324.93523 181.52636 -3.9016567 -1.6054928
225000 250.39509 324.27827 179.266 -0.31080066 -1.5951352
226000 249.55423 319.02419 182.36446 -2.4408183 -1.601847

there are no temperature gradient in the system

Chunk Coord1 Ncount v_temp

1 -13.7079 3109.88 248.85
2 -9.68368 3130.43 247.449
3 -5.65946 3113.68 250.083
4 -1.63525 3108.77 252.999
5 2.38897 3098.47 249.15
6 6.41318 3067.66 248.884
7 10.4374 3093.61 248.539
8 14.4616 3127.21 250.038
9 18.4858 3131.94 249.452
10 22.51 3120.9 248.054
11 26.5343 3118.81 246.409
12 30.5585 3072.47 249.121
13 34.5827 3043.94 253.163
14 38.6069 3070.93 252
15 42.6311 3104.19 249.459
16 46.6553 3135.22 247.556
17 50.6795 3111.98 248.382
18 54.7038 3080.89 253.826
19 58.728 3068.22 251.01
20 62.7522 3090.8 250.958

How should I choose the value of the parameter “energy flux into the reservoir” using the real units so that a linear temperature gradient arises in the system?
Is it possible to somehow determine this parameter from the simulation?

Best regards,
Victor

I suggest you also tally the temperature
profile between the hot/cold regions during
equilibration, or during the latter part of it.

If you are not seeing a well-developed, static slope
gradient at the beginning
of when you start calculating kappa, then
you are not fully equilibrated.

Steve

Greetings,

In my experience it usually takes a couple million timesteps, where the timestep is 1fs, for a system like Silicon to approximate a steady temperature profile. I’m not sure what the case will be for your polymer. Additionally the temperature may fluctuate indefinitely, meaning as long as you are willing to practically simulate, in certain areas if your system geometry or boundary conditions are rather inhomogeneous. However if you average the temperature over several tens of thousands of steps, or more, you can circumvent this fluctuation. Additionally, if your systems has interfaces between types of polymers etc. you may see sudden drops in temperature that will not be averaged out, i.e. thermal boundary resistance.

Adrian Diaz

Dear Sirs,

Thank you for your comments.

Best regards,
Victor

Без вирусов. www.avast.ru

пт, 3 июл. 2020 г. в 00:30, Diaz,Adrian <adriandiaz@…1447…>: