[v1.0.7] - 2025-10-30

Release Notes

A bunch of smaller new features and bug fixes.

Python versions < 3.10 are no longer officially supported and built for.

Several bugs were fixed that have silently caused wrong behavior in the past. In particular this had affected Hofstadter models with symmetric gauge, mixed xk models in edge cases, some MPO methods if explicit_plus_hc was used and DMRG with chi_list.

Changelog

Added

Added the subsites_start parameter to exponentially decaying couplings, making the kind of couplings it can cover even more general. See add_exponentially_decaying_coupling().
Added add_centered_exponentially_decaying_term()
Added plus_identity() for MPO.
Added tenpy.networks.purification_mps.PurificationMPS.from_density_matrix()
Implemented Krylov subspace expansion for TDVP.

Fixed

Fixed a bug in the hofstadter models, where the symmetric gauge was not implemented correctly and yielded a different physical model. Removed the config options for controlling the magnetic unit cell size, since in all well-configured cases they had no effect. See PR #527 for details.
Fixed a bug (see issue #446 ) where segment DMRG failed with an error.
Fixed a bug where add_exponentially_decaying_coupling() with plus_hc=True did not hc the op_string
Fixed a bug in single-site DMRG, where mixer_activate raised an error if mixer = None
Fixed a bug where DMRG using a chi_list caused the environment sweeps to use the default chi_max, which is 100 and implies unintentional heavy truncation in most settings.
Fixed a typo (wrong keyword) for QRBasedTEBDEngine with nontrivial qtotal of MPS B tensors
Fixed a typo in the mixed_xk models, where from_charges_of_orbitals() ignored the ring_order.
Fix tenpy.networks.mpo.MPO.is_equal() if explicit_plus_hc=True.
Fixed a bug where CouplingMPOModel could not configure NLegLadder lattices automatically, from only the model_params, because that lattice has a different __init__ call structure.