v0.5.0

TNRKit v0.5.0

Diff since v0.4.0

TNRKit v0.5.0 adds new features and significantly unifies the model interface.

Breaking Changes

Unified Model Interface (#149)

All models that previously had a separate functionname_symmetric variant for symmetry-enhanced tensors
have been unified into a single function that takes the symmetry type as the first positional argument.
The default behavior of each function now returns a tensor with the maximum available symmetry.

Old (v0.4.0)	New (v0.5.0)
classical_ising(β)	classical_ising(Trivial, β)
classical_ising_symmetric(β)	classical_ising(β) or classical_ising(Z2Irrep, β)
classical_ising_3D(β)	classical_ising_3D(Trivial, β)
classical_ising_symmetric_3D(β)	classical_ising_3D(β) or classical_ising_3D(Z2Irrep, β)
classical_ising_triangular(β)	classical_ising_triangular(Trivial, β)
classical_ising_triangular_symmetric(β)	classical_ising_triangular(β) or classical_ising_triangular(Z2Irrep, β)
classical_potts(q, β)	classical_potts(Trivial, q, β)
classical_potts_symmetric(q, β)	classical_potts(q, β) or classical_potts(ZNIrrep{q}, q, β)
classical_clock(q, β)	classical_clock(Trivial, q, β)
classical_clock_symmetric(q, β)	classical_clock(q, β) or classical_clock(ZNIrrep{q}, q, β)
classical_XY_U1_symmetric(β, n)	classical_XY(β, n) or classical_XY(U1Irrep, β, n)
classical_XY_O2_symmetric(β, n)	classical_XY(CU1Irrep, β, n)
phi4_real(K, μ0, λ)	phi4_real(Trivial, K, μ0, λ)
phi4_real_symmetric(K, μ0, λ)	phi4_real(K, μ0, λ) or phi4_real(Z2Irrep, K, μ0, λ)
phi4_complex(K, μ0, λ)	phi4_complex(Trivial, K, μ0, λ)
phi4_complex_symmetric(K, μ0, λ)	phi4_complex(K, μ0, λ) or phi4_complex(U1Irrep, K, μ0, λ)

The following exported names have been removed:

• classical_ising_symmetric
• classical_ising_symmetric_3D
• classical_ising_triangular_symmetric
• classical_potts_symmetric
• classical_clock_symmetric
• classical_XY_U1_symmetric
• classical_XY_O2_symmetric
• phi4_real_symmetric
• phi4_complex_symmetric

All model functions now also accept a T::Type{<:Number} keyword argument to control the element type of the output tensor.

LoopTNR run! Interface (#144)

The six-argument run! call for LoopTNR that took separate truncation and stopping criterion arguments for the entanglement optimization has been replaced by the LoopParameters struct, which bundles all loop-optimization parameters:

# Old (v0.4.0)
run!(scheme, trscheme, truncentanglement, criterion, entanglement_criterion, loop_criterion)

# New (v0.5.0)
run!(scheme, trscheme, criterion, LoopParameters())

The simplified two-argument form run!(scheme, trscheme, criterion) still works and uses LoopParameters() defaults internally.

New Methods

Correlation Functions

• CorrelationHOTRG: Computes two-point correlation functions using HOTRG. Calculates correlators between two impurities at a horizontal distance of 2ⁿ sites (#139).

Honeycomb Lattice CTM

• c3vCTM_honeycomb: C3v-symmetric Corner Transfer Matrix for honeycomb lattices (#141).

New Models

Honeycomb Lattice Ising Model (#141)

• classical_ising_honeycomb: Ising model on the honeycomb lattice
• New constants: ising_βc_honeycomb and f_onsager_honeycomb

XY Model

• classical_XY: Unified XY model function (replaces classical_XY_U1_symmetric and classical_XY_O2_symmetric)
• New constant: XY_βc

Added Functionality

Nuclear Norm Regularized LoopTNR (#144)

• LoopParameters: New exported struct to configure loop optimization, including support for the nuclear norm regularization (NNR-TNR).
• VN_entropy: Calculate the von Neumann entropy of a transfer matrix.
• loop_entropy: Calculate the loop entropy of a LoopTNR scheme.

Bugfixes

• Fix wrong twist factors and switch from svd_trunc to eigh_trunc for projectors in 3D HOTRG (#145).
• Fix broken onesite CTM implementation (#155).
• Fix incorrect symmetric complex φ⁴ tensor construction, now accessible via phi4_complex(U1Irrep, ...) (#142).
• Remove incorrect mentions of external field h from complex φ⁴ docstrings and ensure all tensors use ComplexF64 (#147).

Merged pull requests:

• CorrelationHOTRG (#139) (@JaridPiceu)
• Implement c3vCTM on the honeycomb lattice (#141) (@sanderdemeyer)
• Fix complex phi4 (#142) (@JaridPiceu)
• Update README with new phi4 models (#143) (@JaridPiceu)
• Nuclear norm regularized LoopTNR (#144) (@Chenqitrg)
• Fix twist and use eigh_trunc in 3D HOTRG (#145) (@Yue-Zhengyuan)
• Add Code Coverage (#146) (@VictorVanthilt)
• Small Improvement Complex phi4 model (#147) (@JaridPiceu)
• Small improvements to the LoopTNR docstrings (#148) (@Chenqitrg)
• [Breaking] Redefine functions to take symmetries as argument (#149) (@borisdevos)
• LoopTNR docstring typos (#150) (@Chenqitrg)
• Added the reference and test of classical XY model (#152) (@Chenqitrg)
• Fix onesite CTM (#155) (@VictorVanthilt)
• Adding two references about CFT data calculation (#156) (@Yue-Zhengyuan)
• Test Krylov and NNR LoopTNR methods (#157) (@VictorVanthilt)

Merged pull requests:

• CorrelationHOTRG (#139) (@JaridPiceu)
• Implement c3vCTM on the honeycomb lattice (#141) (@sanderdemeyer)
• Fix complex phi4 (#142) (@JaridPiceu)
• Update README with new phi4 models (#143) (@JaridPiceu)
• Nuclear norm regularized LoopTNR (#144) (@Chenqitrg)
• Fix twist and use eigh_trunc in 3D HOTRG (#145) (@Yue-Zhengyuan)
• Add Code Coverage (#146) (@VictorVanthilt)
• Small Improvement Complex phi4 model (#147) (@JaridPiceu)
• Small improvements to the LoopTNR docstrings (#148) (@VictorVanthilt)
• [Breaking] Redefine functions to take symmetries as argument (#149) (@borisdevos)
• LoopTNR docstring typos (#150) (@VictorVanthilt)
• Added the reference and test of classical XY model (#152) (@Chenqitrg)
• Fix onesite CTM (#155) (@VictorVanthilt)
• Adding two references about CFT data calculation (#156) (@Chenqitrg)
• Test Krylov and NNR LoopTNR methods (#157) (@VictorVanthilt)

Closed issues:

• Hi Yue, thanks for this! I want to refactor the whole way we are calculating cft data to unify it and make it easier for the user to understand. Right now the cft methods are often type unstable, or return Dict{Any, Any} by design and this really should not be the case. (#153)