I was having trouble running a calculation. I found a workaround, but either I am making a mistake somewhere else (likely), or the code does something unintended (less likely).
Typically when running EPW I first do it in steps. First the wannierization, check it, then do interpolation, then run the Migdal-Eliashberg part, and finally re-run the code just to get the phonon self-energy along a select path. This time the code worked fine until the Eliashberg part, but when re-started with phonselfen = .true. I got that all the linewidths are zero, which couldn't be.
It was unsurprising considering that in the standard output I had
DOS = -0.000000 states/spin/eV/Unit Cell at Ef= 0.000000 eV
Despite the code correctly recognizing that the Fermi energy of the coarse grid was ~8.3 eV
For reference, in the Eliashberg part I had:
DOS(states/spin/eV/Unit Cell) = 2.9165175653E+00
I found a workaround by directly enforcing the fine-mesh Fermi energy when I restart the calculation to get the phonon selfernergy, i.e.
efermi_read = .true.
fermi_energy = 8.296993
Which returned the right DOS and in general a sensible result.
But it doesn't feel like a very elegant solution. Any ideas on what can cause this?
ps. I cannot share the input files publicly right now
Thanks for all your hard work,
Simone Di Cataldo
Sapienza University of Rome