Changelog

TeMFpy 0.4 (in development)

Gutzwiller projection routines now correctly keep the effect of non-inplace Array operations such as drop_charge() and gauge_total_charge(). #30
Gutzwiller projection routines add a complete set of dummy Schmidt values for compatibility with canonical_form_infinite1() in TeNPy v1.1.0. #31
Fix a bug preventing temfpy.slater.C_to_iMPS() to be used with complex correlation matrices. #32 & #34
Removed unnecessary and erroneous setflags() calls from temfpy.pfaffian.MPSTensorData.from_schmidt_vectors(). #33

Fixed an error in temfpy.gutzwiller.abrikosov() and temfpy.gutzwiller.abrikosov_ph() for infinite MPS when return_canonical=True by preventing an unsupported cutoff argument and resolving the correct canonical form via boundary conditions. #29

The offset parameter is now applied to the gauge-fixing matrix in temfpy.slater.C_to_iMPS(). #28

By default, the iMPS creation routines temfpy.slater.C_to_iMPS(), temfpy.slater.H_to_iMPS(), and temfpy.iMPS.MPS_to_iMPS() now subtract an estimate of the total U(1) charge to the left of the extracted unit cell from the quantum numbers on the virtual legs of the iMPS. The amount of this offset can be controlled with the new offset parameter; the old behaviour can be restored using offset=0. #20

Gutzwiller projections temfpy.gutzwiller.abrikosov() and temfpy.gutzwiller.abrikosov_ph() allow targeting different virtual charge/parity sectors, which may give access to different topological sectors. #18
All functions that create an MPS support an optional unit_cell_width argument for specifying the physical length of the system being simulated. Sensible defaults are also provided. #21

Gutzwiller projections temfpy.gutzwiller.abrikosov() and temfpy.gutzwiller.abrikosov_ph() now handle most infinite fermion MPS correctly. #18
A check in temfpy.pfaffian.SchmidtModes.from_correlation_matrix() assumed that at most half of all Nambu eigenvalues are 1/2, which is not always the case. These checks now pass. #24
A bug in the construction of Nambu eigenvectors with eigenvalue 1/2 is now fixed. #25
In certain edge cases involving many 1/2 Nambu-Schmidt eigenvalues, the preference of eigensolvers to return the identity matrix as eigen/singular vectors of the identity matrix resulted in zero overlaps between Boguliubov vacua when constructing temfpy.pfaffian.MPSTensorData objects. This issue is mitigated by rotating the real Nambu eigenvectors with eigenvalue 1/2 with a quasi-random orthogonal matrix. This results in a unitary rotation of degenerate Boguliubov-Schmidt states and hence a gauge rotation of the MPS tensors. However, this should have no observable consequences. #26

First stable version without severe bugs.