it seems to me that you have very different corrections for some occupied states, see e.g. at gamma:
you have positive or negative correction depending on the bands (negative for the deeper and positive for the others). For other k points you are in a different situation e.g. k=59, nearly degenerate bands (69 and 70),
one with a positive and one with a negative correction (see below). I suspect that this makes different band crossing at GW level.
Before looking at the interpolated bands I suggest you inspect your GW energies, I suspect that also the bands extracted from the calculated points will have similar problems.
From the report file, I can see you are using pseudopotential with Non-linear core corrections. I suggest you to use pseudopotential without NLCC (e.g. sg15 set), or if you want to use pseudo with NLCC you need to activate in the input file the flag:
Code: Select all
UseNLCC
It appears in the input file when creating files with high verbosity (-V all) and you need to uncomment it.
If you want to try this, you need to delete the ndb.HF* and ndb.QP file and repeat your calculation (you can keep the ndb.pp file as this is not affected by the nlcc).
Best,
Daniele
Gamma:
# K-point Band Eo E-Eo Sc|Eo
1.00000 67.00000 -1.11448 -0.33846 1.51109
1.00000 68.00000 -1.11448 -0.33827 1.51108
1.00000 69.00000 -0.76919 0.29526 1.53658
1.00000 70.00000 -0.76919 0.29494 1.53658
1.00000 71.00000 -0.37559 0.27921 1.41679
1.00000 72.00000 -0.37559 0.27965 1.41679
K= 0.142857 0.238095 0.000000 (rlu)
59.00000 67.00000 -1.85707 -0.82703 2.21691
59.00 68.00 -1.760 0.4978E-2 2.148
59.00000 69.00000 -1.67270 -0.52412 2.06564
59.00000 70.00000 -1.64314 0.44321 2.10633
59.00000 71.00000 -1.11250 -0.14020 1.77366
59.00000 72.00000 -0.80520 -0.07430 1.58187