calculation with spin-orbit coupling of polar material

[ZhichaoLiu-0306]

Dear Nakib,

The input file of 2nd Force constants is "espresso.ifc2", but for SOC calculation in EPW needed is espresso.ifc2.xml which cannot be read by src/phonon.f90 in elphbolt. Can the current version be used for SOC calculation?

If calculating polar materials, should I add "lpolar = .true." and "lifc = .true." in "epw.in" to read "Read Z* and epsilon", or just make NAC correction in elphbolt rather than in epw?

Thanks!

[nakib]

Hi Zhichao

Thank you for your question!

In this case you can run the phonon calculation twice. Once generating the ifc2 file in the "normal" format to be passed to elphbolt. And another time to to generate it in the xml format to be used with the SOC enabled calculation with EPW. In the future I will consider implementing an xml format parser in elphbolt.

For polar materials, setting the flag lpolar in epw.in and epsil in ph.in to .true. will suffice. The NAC correction will be performed in elphbolt only if you set the flag polar to .true. in the input.nml file.

Make sure to pay attention to the key spindeg in elphbolt for your SOC enabled calculation.

I have not run any SOC enabled calculations, so I am curious to know how your calculations go. It is always a good idea to compare the electron and phonon bands and the e-ph matrix elements plotted along a path between EPW and elphbolt. Within numerical errors, they should be the same. I'd really appreciate it if you send feedback.

Best,
Nakib

[nakib]

Hi Zhichao,

Thanks for digging deeper into the use_ws flag. I am traveling right now, but will take a closer look once I'm back. It would be great if you could also tell me what you see for the electronic band structure calculation with the plot_along_path flag when using use_ws = .true. in EPW.

Best,
Nakib

[ZhichaoLiu-0306]

Dear nakib,

Thanks! Wish you a pleasant journey!

The energy band is good using use_ws flag. For phonon dispersions, two TA branches from EPW and elphbolt are not degenerate along G-X-W with use_ws = .false. in fig2 (a), while using use_ws flag it is improved as shown in fig2(b).

Best,
Zhichao

[nakib]

Dear Zhichao,

Thanks a lot for sharing these. If possible, would you please share your elphbolt input.nml file?

I have created an issue (#77) and try to get to it soon, but, unfortunately, can't give a timeline.

Best regards,
Nakib

[ZhichaoLiu-0306]

Dear Nakib,

Thanks a lot !

The input.nml has been uploaded for SOC and no-SOC calculations (https://github.com/ZhichaoLiu-0306/Si-test).

Best regards,
Zhichao

[nakib]

Thanks!