doc/tutorialsexercises/opticalresponse/gw_tutorial/check_gw.py fails G0W0G check
See check_gw.py. New number are:
[1.1579354 2.63426896 6.40692348 6.40678709]
Old numbers:
[1.147, 2.640, 6.417, 6.418]
Issue closed - new values reflect an accidental bug fix.
In !1430 (merged), the G0W0Kernel
was changed to use the actual grid of irreducible BZ k-points from the QPointDescriptor
which defines the q-point integration of the G0W0Calculator
, instead of using a grid from the ResponseGroundStateAdapter
as it was the case in the old code. These grids were assumed to be the same (since the G0W0Calculator
needs the G0W0Kernel
to calculate the kernel for exactly the q-points it is interested in), but in fact, individual q-points of the old code could be shifted by a reciprocal lattice vector with respect to the intended q-point. Hence, we unknowingly fixed a bug with !1430 (merged), leading to a slight change in test values for GWG
calculations.