Dear Gregor,
thanks for the reply. Hope you had a good vacation.
I did perform some tests using your input, and also adding local orbitals for the 5s and 5p Gd states (attached inputs).
What I noticed is that the calculations with the local orbitals are the following: they depend on the size of the spheres, if I increase the size of the Ga muffin-tin radius the calculations crashes for some q-values; it also crashes (for some q-values) when considering relativisticCorrections="T".
Now, the trend observed for the energy as a function of q (with and without local-orbitals) are completely different (see figure).
With Gd 5s and 5p states in the core, the lowest energy of the system is ferromagnetic (q=(0,0,0)). When I consider these states as local orbitals, the spin-spiral states have lower energies in the path considered for the calculations.
I wonder: which result would be more reliable, if any?
In the calculations I consider a full k-mesh and only the Identity symmetry.
(The calculations are not very well converged - k-mesh is 6x6x6 - but don't know if just this justifies such different trends).
I'll test later with lda+U and see if these trends are the same.