Accuracy of interpolated eigenvalues

[NFH]

Dear users, dear Carla, dear Samuel

Was there a change in plot_band.f90 or accompanied modules between EPW 4.0 and 4.1?

When extracting the wannier-interpolated electronic eigenvalues on a given path with "band_plot = true", the eigenvalues seem not to be determined/calculated with the same accuracy as they are printed. The print output suggests an accuracy of 0.1meV, which would be sufficient. Nevertheless the "real" accuracy seems to be ~10meV. I've never noticed smt similar in older EPW versions and couldn't find striking changes in the code.

An example of a zoomed in band:

http://www.directupload.net/file/d/4556 ... mc_png.htm

all the best

Nicki
CAMd, DTU Lyngby

[NFH]

I'll reply to myself

Problem appears if (and only if) kpath = true AND nkf is still set.
This of course was not clean inputting of myself, but no error occurred and I was sure that kpath = true would be the stronger argument and default nkf = false. The interpolated eigenvalues are then those corresponding to the nkf-mesh, while the kpoint-coord. of the kpath-file are shown. At least this is what I think what happens.

Problem solved for me.
Nicki

[sponce]

Hello Nicki,

Thank you for sharing the solution !

I could add a warning error indeed.

I have added the following two lines:

Code: Select all

  IF (band_plot .and. filkf .ne. ' ' .and. (nkf1 > 0 .or. nkf2 > 0 .or. nkf3 > 0) ) CALL errore('epw_readin', &
                &'You should define either filkf or nkf when band_plot',1)
  IF (band_plot .and. filqf .ne. ' ' .and. (nqf1 > 0 .or. nqf2 > 0 .or. nqf3 > 0) ) CALL errore('epw_readin', &
                &'You should define either filqf or nqf when band_plot',1)

in the file epw_readin.f90, line 501.

Could you check that this would solve similar issue to your?
Thank you.

Best,

Samuel