Thermal conductivity by BTE

Dear. Julian

Based on the example67, I conducted the calculation compiled with alamode.

I just changed the input parameter of ala_processor from 4 to 1.

the other information of input parameter is all the same.

Thanks for your update, the calculation with alamode 1.3 was done.

But, I got the example67_rta.kl that is different from output in the example

There is a huge difference of conductivity about a factor of 1000 times

my result is

Temperature [K], Thermal Conductivity (xx, xy, xz, yx, yy, yz, zx, zy, zz) [W/mK]

300.00     14211.7841       -64.3881       -64.3881       -64.3881     14211.7841       -64.3881       -64.3881       -64.3881     14211.7841

example result of example67 is

Temperature [K], Thermal Conductivity (xx, xy, xz, yx, yy, yz, zx, zy, zz) [W/mK]

300.00       140.3922        -0.0982        -0.0982        -0.0982       140.3922        -0.0982        -0.0982        -0.0982       140.3922

I don’t know why this huge difference come out.

and, in the other large cell consisting of over 400 atoms ,

I got the other error during force constant calculation like this:

Doing Displacement =   6371 of   6372 for atoms  217 321
Doing Displacement =   6372 of   6372 for atoms  217 321

Cubic displacements finished

!! ERROR : attempt to run alm for force constants failed

Program terminated by processor 0 in gen_fc_alamode

If the cell would be so large, do this problem comes out ? and Is this error related with computer memory due to the large cell?


I’ve had a check of things on my laptop and here are some responses to the 2 main issues highlighted:

  1. For example67 there does seem to be some variability in the results for the thermal conductivity even between multiple runs for the same number of cores, though I’ve only seen changes in the range of 140-141 or thereabouts for the first value. I didn’t find anything like the extreme case you seem to have had and so I’m not sure what’s going. On investigation, the variability arises because of the use of a shell model leading to differences in the forces at the level of the 9th decimal place. This is because the shell model involves having to perform an optimisation as part of every force evaluation to relax the shells and so the level of numerical noise is far higher than for other models.
  2. For the ERROR termination, this is because the execution of Alamode has failed and so you need to look into the log files written by this code, such as anphon.log, to find out what happened. It might be a segmentation fault (which I have seen) or running out of memory, as you suggest, is possible if the system gets too large.
    Hope this helps a bit,
1 Like


I appreciate so much your comment.