Hi julian,
Sorry for bothering you. Currently, I am trying to invoke ALAMODE to calculate the thermal conductivity, referring to example67 and modifying the input file as follows:
Unfortunately, the files si222_rta.kl and si222_rta.result were not generated. However, running example67.gin can generate them. The output log shows as follows:
Hi LiMa
I’m afraid I can’t directly test your run since I’m travelling and don’t have access to Alamode.
Did example67 run correctly before you made changes? Note that GULP was tested for Alamode up to version 1.3.0 and so if you are running a later version this might cause issues.
The files you mention weren’t generated because there was an earlier error in the run, so there’s no way to generate these (sensibly). GULP doesn’t normally a require a change to TOLERANCE since if you generated the structure using symmetry in GULP then it will be accurate to within this anyway. If you really want to add it then you’ll need to edit the inputs that are written out and re-run.
Regards,
Julian
PS You don’t need the “shopt” keyword since there are no shells in GFN-FF.
Hi julian,
I’m really sorry for interrupting your trip. Before making modifications, Example 67 ran correctly. However, if I only change the force field to GFNFF and delete the “shopt” keyword, Example 67 runs successfully but the output thermal conductivity is 0. I suspect that the model parameters of Example 67 include shell parameters. Therefore, I tried using a shell-free and geometrically optimized crystal for calculation, but the output thermal conductivity is still 0. Are there any limitations when using the GFNFF force field to calculate thermal conductivity?
Wish you a pleasant journey.
Sincerely,
LiMa
Hi LiMa,
I’ve managed to get Alamode 1.4.2 compiled and so I’ve now tested what happens if you run the equivalent of example67 but with GFN-FF. In short, everything runs as expected and all the correct files are produced with sensible output. However, the thermal conductivity does come out as zero. There shouldn’t be a restriction on GFN-FF for this kind of calculation since the main restriction is that phonons can only be computed at gamma in 6.1, but Alamode does the k points using a supercell, so that’s fine. The only real difference between example67 and the GFN-FF is that the frequency spectrum is different (and so are the force constants). Why this leads to thermal conductivity of zero (or below the range of the output) is an open question for further investigation. At least everything runs correctly in terms of the GULP-Alamode interaction as far as I can see.
Regards,
Julian
Hi julian,
Thank you very much for your response. If I understand correctly, you mean that in GULP 6.1 version, it is not possible to simultaneously use Alamode and GFN-FF due to the limitation of GFN-FF being only able to compute phonons at gamma point. So, does it mean that I only need to use other potential functions?
Regards,
LiMa
Hi LiMa,
That’s actually not correct. GULP 6.1 only does gamma point for GFN-FF, but this shouldn’t be a problem. for Alamode. Everything seems to work fine, except the thermal conductivity comes out at zero. It needs analysis of the force constants and how Alamode uses them to work out why this is the case for this force field.
Regards,
Julian
