Changelog

All notable changes to this project will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

[Unreleased]

Changed

• merge_negative_weight! throws an ArgumentError for a total weight within rounding noise of zero instead of returning an empty PolesSum (#286) (f226887)
• init_system requires H_int::Operator{Float64} and ϵ_imp::Float64, which the ground-state calculation needed already (#292) (d123a1e)
• anderson_matrix, and with it inverse and self_energy_schur, no longer set the BLAS thread count to the Julia thread count (#293) (ecfdf8f)
• find_chemical_potential and hybridization_function_local warn when an integral stops above n_tol or tol (#294) (54e66b2)
• find_chemical_potential starts the bisection from the band range of H_k + Σ_stat instead of the pole range of Σ_dyn by default (#299) (95b3356)

Removed

• tol argument of PolesSumBlock(locs, amps) and tol_location keyword of greens_function_local, which threw for two clusters on either side of zero; merge nearby locations with merge_degenerate_poles! instead (#290) (85f79a3)

Fixed

• ndet2 counts two holes in the valence sites or two electrons in the conduction sites correctly, which was wrong whenever nfilled ≠ nempty (#280) (866b33a)
• PolesContinuedFractionBlock(::PolesSumBlock) keeps negative and complex entries of the scale instead of setting them to zero or throwing (#282) (210516c)
• merge_negative_weight! conserves the first moment whenever the mean lies within the locations, which was lost when an interior split pushed an outermost pole negative (#286) (f226887)
• self_energy_schur no longer returns a zero self-energy for energies in large units, such as meV (#287) (64e0cd2)
• block_lanczos_full_ortho, and with it PolesContinuedFraction(::PolesSum) and PolesContinuedFractionBlock(::PolesSumBlock), no longer collapses poles at small energy scales (#288) (66f766d)
• merge_small_weight! keeps the last pole instead of throwing a BoundsError when all poles have weight <= tol (#289) (8002e51)
• grid_interpolate, greens_function_bethe_grid, and greens_function_bethe_grid_hubbard3 accept integer grids instead of throwing an InexactError (#291) (c475393)
• grid_interpolate throws an ArgumentError for fewer than 2 points instead of a BoundsError (#291) (c475393)
• PolesContinuedFractionBlock(locs, amps) accepts matrices that are Hermitian up to rounding and copies its input, as the form with a scale does (#295) (2f3be5c)

[0.12.0] - 2026-09-25

Added

• quasiparticle_weight_optimum_regularization to find the regularization parameter λ at the first plateau of $Z(\lambda)$ on a logarithmic axis, with a prominence filter to discard shallow wiggles (#240) (05c4c0d)
• NaturalImpurityOrbital describing the natural impurity orbital basis by its impurity-mirror block and its valence and conduction chains, together with the accessors n_valence and n_conduction (#242) (addb83a)
• hybridization_function_local to obtain the mean-field impurity level and the hybridization function of the correlated orbitals on a grid from the lattice Hamiltonian and self-energy, evaluating the correlated block in the upper half-plane instead of diagonalizing the block arrowhead matrix at every k-point (#266) (277c1db)

Changed

• AbstractPoles, AbstractPolesSum, and AbstractPolesContinuedFraction are now parameterized by (A, B) (#232) (ec0b81a)
• amplitude constructor of PolesSumBlock accepts zero amplitudes, consistent with the weight-list constructor (#236) (8385d55)
• rename to natural impurity orbitals (#242) (591fbf3)
  • to_natural_orbitals → natural_impurity_orbital
  • natural_orbital_operator → natural_impurity_orbital_operator
  • natural_orbital_ras_operator → natural_impurity_orbital_ras_operator
• natural_impurity_orbital takes the hybridization function Δ::PolesSum instead of an arrowhead matrix, and returns a NaturalImpurityOrbital instead of (H_nat, n_occ). Both operator builders drop their n_occ argument accordingly (#242) (addb83a)
• chain hoppings are positive, fixing the gauge LAPACK leaves free (#242) (addb83a)
• natural_impurity_orbital and init_system take the mean-field impurity level ϵ_mf that defines the reference basis, next to the bare level ϵ_imp entering the Hamiltonian (#242) (0fbc17c)
• natural_impurity_orbital requires every pole of Δ to carry weight and throws an ArgumentError otherwise. (#242) (675e34f)
• ground_state! on an already-shifted Hamiltonian returns the remaining correction, not the running total (#247) (1c6a88a)
• self_energy_IFG → self_energy_schur (#259) (61d7bad)
• find_chemical_potential evaluates the filling as a contour integral along the imaginary axis and updates the resolvent with the Woodbury identity (#265) (17ac045)

Removed

• discretize_similar_weight, discretize_to_grid (#225) (e043ebb)
• submodule Debug from the public API (#228) (8c953db)
• custom LAPACK wrappers sytrd! and orgtr! (#242) (27f4c30)
• get_RAS_parameters as parameters can be inferred directly (#242) (79cec2a)
• greens_function_bethe_equal_weight and hybridization_function_bethe_equal_weight, (#248) (b1278c8)

Fixed

• loss of particle-hole symmetry in natural_impurity_orbital for a symmetric bath with an odd number of sites, where the level at the Fermi energy used to be assigned to the conduction chain as a whole. Its weight is now shared evenly between both chains (#242) (addb83a)
• natural_impurity_orbital throws on an empty valence or conduction sector instead of fabricating a site (#242) (addb83a)
• correct eigenvectors and their adjoint in the self-energy from Schur complement (#235) (c2ea559)
• clear ArgumentError from size on empty blocks of poles (#236) (105f69b)
• natural_impurity_orbital_ras_operator allows zero valence or conduction vector sites, but not both (#245) (e0336a0)
• spectral_function_loggaussian broadens a pole at zero with a Gaussian instead of dropping its weight (#249) (4bd2786)
• find_chemical_potential and greens_function_local accept a real self-energy together with a complex dispersion (#250) (19edfbd)
• init_system stops after 100 restarts instead of looping forever when the requested variance is unreachable (#253) (a9149aa)
• greens_function_bethe_analytic(-0.0) returns the retarded branch (#254) (022c2f1)
• natural_impurity_orbital_ras_operator accepts an empty valence or conduction chain (#255) (2335f63)
• find_chemical_potential counts a level at the Fermi level half (#256) (4bd5028)
• Löwdin orthonormalization in anderson_matrix, PolesContinuedFractionBlock, and block_lanczos_full_ortho stays orthonormal for ill-conditioned, rank-deficient, and zero blocks (#270) (60a9db0)
• anderson_matrix throws an ArgumentError for rank-deficient amplitudes, such as self_energy_schur at $U = 0$, instead of adding a spurious site or throwing DimensionMismatch (#270) (ba91343)

[0.11.0] - 2026-08-14

Added

• moments of complex function (#164) (da721d2)
• transformation from sum of poles to Anderson form (#166) (1729012)
• generalize self-energy to any 2n×2n block sum of poles (#167) (22f40e7)
• get location of poles directly (#181) (78be1d5)
• iterate through AbstractPolesSum (#182) (12f5719)
• calculate filling for AbstractPolesSum (#183) (68c6489)
• shift spectrum of AbstractPolesSum (#186) (d191feb)
• calculate trace of AbstractPolesSum (#187) (5cf51be)
• calculate thin rectangular amplitude (#195) (bfc7348)
• local lattice Greens function in resolvent formalism (#201) (5b8c250)
• enforce canonical representation for AbstractPolesSum (strictly increasing with no duplicates) (#203) (3ea3db4)
• evaluate a pole representation at an arbitrary complex frequency z (#209) (273a6b0)

Changed

• add regularization parameter for quasiparticle weight (#169) (8c3a40b)
• PolesSum and PolesSumBlock are always sorted by increasing pole locations. Degenerate poles are automatically merged. (#175) (4f876e1)
• explicitly specify matrix representation for given pole representation (#196) (80d9ae7) New methods are anderson_matrix, arrowhead_matrix, or tridiagonal_matrix.
• Base.inv(::PolesSum) → inverse (#215) (7dbfc91)
• natural_orbital_ci_operator → natural_orbital_ras_operator (#216) (d515549)

Removed

• module ED (#198) (86ed14b)
• finite broadening methods (#207) (f4b5a8d)
  • imagKK
  • realKK
  • greens_function_partial
  • quasiparticle_weight_gaussian
  • self_energy_IFG_gaussian
  • self_energy_IFG_lorentzian
  • δ_gaussian
• equal_weight_discretization (#217) (0d6582e)

Fixed

• lognormal broadening on a grid of type AbstractVector (#163) (2cb92dd)
• allow zero length for AbstractPoles (#174) (7167664)

[0.10.0] - 2025-12-11

Changed

• rename package to RAS_DMFT (#161) (fb3e137)

[0.9.0] - 2025-12-04

Added

• calculate self-energy using Schur complement (#131) (caeee45)
• calculate quasiparticle weight using quasiparticle_weight (#132) (80f7666)
• add pole at location 0 with weight 0 (#137) (c7b6cd8)
• calculate quasiparticle weight with Gaussian broadening using quasiparticle_weight_gaussian (#145) (ef631cd)
• calculate quasiparticle weight on a grid using quasiparticle_weight (#149) (ba2a2eb)
• merge negative locations for block pole sum (#150) (133e0ae)
• put block sum of poles on given grid (#151) (49db03f)
• flip spectrum for block sum of poles (#152) (13be8e3)

Changed

• merge_small_poles! → merge_small_weight! (#130) (35dd38d)

• rename starting wave function (#133) (709e917)

  • starting_Wavefunction → Wavefunction_singlet
  • starting_CIWavefunction → CIWavefunction_singlet

• ground_state → ground_state! (#134) (a34264e)

• require Julia >=1.12 (#153) (494ffac)

• logarithmic broadening keeps pole at zero frequency unchanged (#156) (1b72fba)

Removed

• self_energy_FG_lorentzian (#128) (bc6f5de)
• remove_small_poles! (#130) (35dd38d)

Fixed

• addition of PolesSumBlock did not create true copies (#129) (85827bc)
• discretize_similar_weight did not return sorted poles (#148) (f1ba260)

[0.8.0] - 2025-07-24

Added

• weight for Poles (#108) (dd5a5da)
• DMFT with poles on a fixed grid (#112) (ecf0680)
• Core.Array for Poles with complex amplitudes (#113) (1ba6b19)
• moment of function defined on a grid (#114) (44c6495)
• PolesSum, PolesSumBlock which will replace Poles (#116) (0570d85) (#117) (4445404)
• PolesContinuedFraction (#118) (7dea995)
• PolesContinuedFractionBlock (#119) (edd8060)
• conversion between different pole representations (#123) (f15b907)

Changed

• rename η_gaussian to δ_gaussian (#110) (c8911ae)
• grid_log needs positive initial value (#111) (ebef3fc)
• δ_gaussian works on scalar (#115) (eed69b5)
• old Poles is split into PolesSum, PolesSumBlock (#126) (87b2fb5)

Fixed

• type-stability when taking square root of matrix (#120) (969d322)
• type-stability for reading in Number (#121) (1831991)

Removed

• struct Poles is split into PolesSum, PolesSumBlock (#127) (04189d7)

[0.7.0] - 2025-06-04

Added

• merge equal poles (#54) (66b6dc2)
• logarithmic grid and log-gaussian broadening (#55) (f3a09b0)
• Hubbard-III approximation (#64) (e5dbed7)
• warn if wrong spectral weight exists (#66) (25da3b0)
• merge poles with small weight to neighbors (#72) (7ab44a7)
• getters for Poles (#81) (add4ec6)
• weight(s) of each pole (#82) (b1f1780)
• moments of Poles (#83) (1917ce2)
• similar weight discretization of Poles (#86) (0987324)
• Base.issorted for Poles (#89) (58cd09a)
• Base.allunique for Poles (#90) (b3d8e52)
• moving poles with negative location to zero (#91) (5dd9f13)
• flip locations of Poles (#95) (5717654) (#97) (b560d33)
• shift locations of Poles (#96) (5b85142) (#97) (0b0ec8f)
• correlator_plus(), correlator_minus() (#98) (3a78427)
• general block correlator (#99) (4f20faa)

Changed

• rename Pole → Poles, self_energy_pole → self_energy_poles (#76) (ee98b96)
• rename move_negative_weight_to_neighbors! → merge_negative_weight! (#77) (5d88c89)
• rename merge_equal_poles! → merge_degenerate_poles! (#79) (c5d2b27)
• remove to_grid_sqr! (#80) (cc71c12)
• calculation of self-energy using Poles does not enforce original grid (#84) (b976fcd)
• calculation of new hybridization function does not enforce original grid (#85) (bba8809)
• merge poles that are tol apart (#88) (ad832db)
• merging negative weight function private (#93) (5491f81)
• remove update_hybridization_function for grid (#102) (dc23810)
• stabilize Lanczos algorithm (#103) (607426d)

Removed

• DMFT step with Lanczos (#69) (34633dd)
• DMFT step with block Lanczos (#105) (2b7710d)
• solve_impurity (#106) (c6c210f)

Fixed

• errors in DMFT self-consistency loop (#57) (b10573a)
• loss of particle-hole symmetry in natural impurity orbitals (#58) (#59) (37ad203)
• length on Pole containing vector and matrix (#67) (9a6dc41)

[0.6.0] - 2025-04-16

Added

• Green's function and hybridization function on a given grid (#46) (c30556d)
• subtraction of Pole (#47) (8ba021b)
• inversion of Pole (#48) (c34ff69)
• update hybridization function in pole representation (#50) (c0725f5)
• calculate DMFT self-consistency in pure Pole representation (#52) (18c783b)

Changed

• unify function names of Green's function and hybridization function (#41) (d095219) (#45) (d4a4bf6)
• rename functions (#49) (6fcefff )
  • self_energy → self_energy_IFG
  • self_energy_gauss → self_energy_IFG_gauss
  • update_weiss_field → update_hybridization_function

Fixed

• block Lanczos must return Hermitian matrices (#42) (d0b066a)
• correlator FR in improved self-energy (#51) (f374b44)

[0.5.1] - 2025-04-02

Fixed

• documentation (#40) (032ed29)

[0.5.0] - 2025-04-01

Added

• calculate Kondo temperature (#35) (b1fe8ff)
• local Green's function from dispersion relation and optional self-energy (#36) (925c512)
• partial Green's function (#37) (1d05c37)
• find chemical potential for desired filling (#38) (4547335)
• non-interacting spectral function with Gaussian broadening (#39) (527b4ab)

Changed

• rename Greensfunction to Pole to reflect that it is more generic (#27) (5631a33)
• unify IO under read_hdf5, write_hdf5 (#29)
• update Fermions dependency to v0.12.0 (#30)
• enforce Pole pole locations to be on the real axis (#34) (d094356)
• create module Debug and clean up namespace (#34) (c92e6a8)