Gs symmetries and testsuite
Many bugfixes for the symmetries in solids. Major improvements are:
- The detection of the symmetries and reduction of the BZ are now working properly for all spacegroups (tested against abinit) and not cubic materials only
- The real space symmetrization is now working properly, tested against LDA, GGA amd MGGA cases, for non-trivial spacegroups
- The testsuite now tests one example of each test group to check the symmetries, kpoints coordinates and weight.
- The testsuite now tests that few arbitrary solids give the same results up to the tolerance with symmetrization than without symmetries
Note however few important points:
- Spin-polarized and spinors cases are not concerned by these bugfixes and have not been properly tested. Symmetries should not be used in these cases. For instance the time-reversal symmetry should be modified and therefore will lead to wrong results in the present implementation.
- The TD case is not concerned by these bugfixes. There is in particular a know bug (I reported it as an issue already) for the combination MGGA+etrs self-consistent+symmetries, which might indicate problems with propagators.
- Symmetries for magnetic systems are not treated for the moment. For AFM, this can be done by hand.
- Fractional translations are not considered for the moment.
- Many observables might not be symmetrized (in real space) or might not work with symmetries. Only the current, density, and the kinetic energy density have been tested.
- The combination symmetries + time reversal symmetries + RPP09 MGGA seems to introduce a very small energy difference compared to symmetries without TR or no symmetries cases (~tens of meV), which does not seems to vanish with convergence. For TD calculations, the TR is anyway not used.