How to fix the difference of phonon frequency calculated with different occupations?

[jgSi1113]

Dear Dr. Kawamura,

Recently, I started my superconductivity calculations with sctk. Before this, I often used QE and EPW for superconductivity calculations and I repeated the examples for Al and MgB2 in the sctk folder.

For my case, I have finished the superconducting calculations via analytically solving the Allen-Dynes-modified McMillan formula. Everything is going well and there is no imaginary frequency (occupations=gauss & deguass=0.05). When it comes to the SCDFT (or sctk), the code requires the occupations=tetrahedra_opt. I changed occupations from gauss into tetrahedra_opt, and I found that there are some imaginary frequency in the matdyn files.

I have read the PRB 89, 094515 (2014) and found the occupations=tetrahedra_opt can reduce the number k or q points, but the frequency is also descreasing significantly (Fig 5 in PRB 89, 094515 (2014) ).

It is too bad for me. Would you like to share me some suggestions kindly to fix the imaginary frequency induced by the different occupations? I have resubmitted the JOB with denser k/q-points mesh, but the imaginary is robust existted.

Looking forward to your reply. Ang comments and suggestions are welcomed.

Thanks and Regards,
Jianguo Si

[jgSi1113]

BTW，if I ignore the existence of imaginary frequency, the calculated Tc (91.72 K) is very closed to the AD-McMillan one (~91.4 K). But the gap is too small (0.35129E-05) and the lambda it too big (0.75456E+01).

Waiting for you kindly reply.

(The e-mail can not connect you correctly. )