Lkind="full"

[Xiaoming Wang]

Dear Davide,

1 - removing the t-rev, did it change only the dispersion with Lkind="full" or also with Lkind="Lbar"?

It changed the dispersion with both.

2 - is the epsilon obtained with "Lbar" the same as the one obtained with "full"?

No

Best,
Xiaoming

[Davide Sangalli]

Thanks. I'm still checking point 1.

For point 2, the two should coincide in the "coupling" case (and I think also in the "resonant" case).
Instead in the "retarded" case the eps obtained with the Lkind="full" procedure does not make much sense.

Best,
D.

[Xiaoming Wang]

Hi Davide,

Here is what I obtained for epsilon with Lbar and Lfull ( resonant case)

Capture.PNG

And the eigenvalues for Lbar

Code: Select all

     2.2026873         0.35157913E-09      1.0000000
     2.20345712        0.190970852E-3      2.00000000
     2.20741916        0.242707565E-8      3.00000000
     2.2080593         0.89593766E-09      4.0000000
     2.21173120         1.00000000         5.00000000
     2.21229005        0.153034243E-7      6.00000000
     2.2589130         0.11822803E-10      7.0000000
     2.2991512         0.47022434E-12      8.0000000

For Lfull

Code: Select all

     2.2026865         0.27045106E-12      1.0000000
     2.20345855        0.563315552E-6      2.00000000
     2.2074187         0.49781387E-13      3.0000000
     2.2080586         0.61604606E-12      4.0000000
     2.2122898         0.43542668E-12      5.0000000
     2.2589135         0.46481737E-12      6.0000000
     2.2991514         0.14118894E-12      7.0000000
     2.36617661        0.999999940         8.00000000

Best,
Xiaoming

[Daniele Varsano]

Dear xiaomimg,
Thanks for reporting. This is without symmetries, right? So, it seems there is still something not working properly.

Best,

Daniele

[Xiaoming Wang]

Dear Daniele,

Yes, I set 'noinv=T' in the nscf input and all the k points were calculated.

Best,
Xiaoming