Discrepancy between TDHF and TIMEEV

[ashwin_r]

I'm trying to compare the outputs of a set of time-dependent calculations on LiF using TDHF and TIMEEV algorithms. I'm using a global hybrid XC-functional (AEXX=0.5) in these calculations. (Vasp version 6.4.2; Intel compiler+MKL)

From running a few tests, it appears that, minimally, one has to explicitly set LADDER=.TRUE. for the TIMEEV case otherwise it just reproduces the independent-particle spectrum. With that done, the TIMEEV spectrum now seems to capture the energy of the first excitonic peak correctly but the height is quite wrong. The overall shape of the spectrum with TIMEEV is also quite different from the TDHF case.

My files are attached.

LiF-AEXX0.5.png

[pedro_melo]

Dear ashwin_r,

Thank you for reporting this issue. I am currently working on it and will come back to you with an answer as soon as possible.

Kind regards,
Pedro Melo

[pedro_melo]

Dear ashwin_r,

I have found that several default variables are reported with strange values in the OUTCAR you obtained during the time-evolution run. I have also reproduced the issue with my own runs. I am looking into this issue.

Kind regards,
Pedro Melo

[ashwin_r]

Dear Pedro,

I'll recheck my runs. But for now it seems that the problem is reproducible so I'll look forward to the bug-fix .

Best,
Ashwin

[ashwin_r]

Just following up to inquire if this bug-fix is expected soon. Thanks!

[pedro_melo]

Dear Ashwin,

As far as we have planned this issue should be fixed in the next release of VASP.

Kind regards,
Pedro

[ashwin_r]

Dear Pedro,

I see that a new release of VASP (6.5) is now available. Has this bug with TIMEEV and hybrid functionals been fixed in this latest release?

Thanks,
Ashwin