The solver was unstable. The main reason: DC opposes U correction. E_int-E_DC is around an order of magnitude lower than the two. So now the DC for the hartree solver gets computed directly in the solver following the DFT+U philosophy.
Added a DC interface with solid_dmft to allow all DC tags to work with the new solver
The solver struggled with restarting: every time the solver is initialized (for example during a restart and during csc), the Sigma inside the solver was initialized to zero, which made the solver unstable. Added an interface to solid_dmft to initialize the Sigma matrix to a given value.
renamed f -> compute_sigma_hartree
Suggestions/doubts:
The DC uses a copy of the triqs_dft_tools DC function. We should think about passing the DC as function to the solver. In this way the user can specify its own DC formula.
Changelog:
The solver was unstable. The main reason: DC opposes U correction. E_int-E_DC is around an order of magnitude lower than the two. So now the DC for the hartree solver gets computed directly in the solver following the DFT+U philosophy.
Added a DC interface with solid_dmft to allow all DC tags to work with the new solver
The solver struggled with restarting: every time the solver is initialized (for example during a restart and during csc), the Sigma inside the solver was initialized to zero, which made the solver unstable. Added an interface to solid_dmft to initialize the Sigma matrix to a given value.
renamed f -> compute_sigma_hartree
Suggestions/doubts: