-
Notifications
You must be signed in to change notification settings - Fork 56
List of questions
For example/Si-PBEsol, you should get something like below. When I intentionally break xml structure of vasprun.xml of no 3, I get the following result:
% phono3py --cf3 vasprun_xmls/disp-00{001..111}/vasprun.xml
_ _____
_ __ | |__ ___ _ __ ___|___ / _ __ _ _
| '_ \| '_ \ / _ \| '_ \ / _ \ |_ \| '_ \| | | |
| |_) | | | | (_) | | | | (_) |__) | |_) | |_| |
| .__/|_| |_|\___/|_| |_|\___/____/| .__/ \__, |
|_| |_| |___/
1.21.0
Python version 3.7.6
Spglib version 1.16.0
Displacement dataset was read from "disp_fc3.yaml".
Number of displacements: 111
Number of supercell files: 111
counter (file index): 1 2 3 Traceback (most recent call last):
File "/Users/togo/.miniconda/envs/dev/bin/phono3py", line 7, in <module>
exec(compile(f.read(), __file__, 'exec'))
File "/Users/togo/code/phono3py/scripts/phono3py", line 43, in <module>
main(**argparse_control)
...
Now the "counter" stops at no 3. So you can doubt the vaspurn.xml is broken. In the previous versions of phono3py, the number may be shifted +/-1.
My calculated thermal conductivity value disagrees with experiments/calculations in literature. Why?
Calculation results depend on many parameters, e.g., theoretical approximations, input parameters of force calculators (energy cutoff etc), or details of implementation of employed codes. So if we want to discuss, the detailed calculation parameters have to be provided. However, just copying and pasting those input files on the mailing list are not accepted. Before submitting a question, we have to consider our problem carefully and summarize it easy to understand for people who want to answer.
Supercell has to be large enough to well describe interaction among three atoms (fc3). This depends on each compound and so we have to test changing from smaller supercell size to larger systematically. Without specifying --dim-fc2, atomic pair interaction (fc2) is also calculated from the same supercell size for fc3. We can start with ~30 atoms supercell and increase the size gradually.
We may expect the interaction range is shorter in fc3 than fc2. So we can separately calculate them. This has to be determined by checking the convergence of interested value over fc2 supercell size. ~100 atoms supercell for fc2 is often a good choice for lattice thermal conductivity calculation, and then we should increase the supercell size gradually for more accuracy. Of course there are many exceptional cases. One of them is the long-range interaction. Dipole-dipole interaction is usually long-ranged and can be reasonably corrected by including non-analytical term correction. For crystals having soft acoustic modes, acoustic branch near Gamma point is calculated to show imaginary frequency. This may be fixed by increasing fc2 supercell size very large.
We can systematically increase --cutoff-pair distance for a fixed fc3 supercell size. When increasing the distance, only newly appearing displacement pairs are necessary to calculate. So there is no loss of computational resources when increasing. Therefore we can start with a
small --cutoff-pair distance and gradually increase it. The detailed procedure is written here, https://phonopy.github.io/phono3py/cutoff-pair.html. An example is found in the appendix of this paper https://doi.org/10.1103/PhysRevB.97.224306 (open access).
FORCES_FC2 is optional. fc2.hdf5 is generated from FORCES_FC3. When we want to use a larger supercell for fc2 than that for fc3, i.e., different supercell sizes for fc2 and fc3, FORCES_FC2 is used together with the --dim-fc2 option.
To understand the cutoff-pair procedure well, please examine the steps written for Si-PBE (https://phonopy.github.io/phono3py/cutoff-pair.html#the-test). The data necessary to run is found in example/Si-PBE directory.