Li-Haldane Correspondence
- Li-Haldane correspondence is the principle that the low-lying entanglement spectrum mirrors the structure and degeneracies of physical edge excitations in topological phases.
- It reveals that virtual boundaries obtained through bipartitioning quantum many-body states emulate effective edge dynamics in systems like fractional quantum Hall and spin-chain models.
- Recent studies have refined the concept through analytical proofs, tensor-network realizations, and counterexamples, clarifying its limits and applicability in both chiral and non-chiral settings.
The Li-Haldane correspondence is the proposition that the low-lying entanglement spectrum of a quantum many-body ground state, obtained from a bipartition and the reduced density matrix of one part, reflects the spectrum and state counting of physical edge excitations. In its canonical form, the correspondence arose from topological phases in $2+1$ dimensions, especially fractional quantum Hall states, where a virtual entanglement cut behaves as an effective boundary. Subsequent work has both strengthened and qualified this viewpoint: it has been analytically substantiated for broad classes of model fractional quantum Hall wavefunctions, explicitly realized in spin-chain symmetry-protected topological phases, operationalized in tensor-network and synthetic-quantum-system settings, and subjected to counterexamples and reformulations in non-chiral, boundary-perturbed, and critical systems (Chandran et al., 2011, Roy et al., 2020, Zache et al., 2021, Guo et al., 24 Sep 2025).
1. Formal statement and entanglement-Hamiltonian framework
For a bipartition of a pure ground state , the reduced density matrix is written as
where is the entanglement Hamiltonian, and the entanglement spectrum consists of the entanglement energies defined through
The Li-Haldane correspondence asserts that the low-lying part of this spectrum mirrors the structure, quantum numbers, and degeneracies of physical edge modes associated with the same topological phase (Zache et al., 2021).
Two entanglement spectra play a central role in the literature. The orbital entanglement spectrum (OES) is obtained by partitioning the single-particle orbital space, so that the cut mimics a spatial edge; this is the spectrum most directly associated with the Li-Haldane proposal. The particle entanglement spectrum (PES) is obtained by tracing out particles instead of orbitals, and is related to bulk quasihole physics. In the fractional quantum Hall setting, later work established a microscopic relation between these two spectra, thereby turning the original conjectural bulk-edge interpretation into an analytic statement about counting in the thermodynamic limit (Chandran et al., 2011).
A widely used field-theoretic rationale for the correspondence is the quasi-locality of . In the Bisognano-Wichmann form,
and on lattices one writes
This makes the entanglement cut a virtual boundary governed by a spatially deformed Hamiltonian, which is the technical basis for interpreting low-lying entanglement levels as edge-like excitations (Zache et al., 2021).
2. Fractional quantum Hall origin and analytic substantiation
The most developed analytic foundation of the correspondence is in model fractional quantum Hall states defined as unique zero modes of pseudopotential Hamiltonians. For these states, a one-to-one map was constructed between the thermodynamic-limit counting of the PES and the OES. Because PES counting equals the counting of bulk quasiholes, while OES counting is tied to edge physics, this map provides a microscopic bulk-edge correspondence in entanglement spectra (Chandran et al., 2011).
The derivation uses clustering operators originating in conformal-field-theory operator expansions. For 0-clustering model states, the wavefunction obeys
1
and the corresponding model Hamiltonian can be written as
2
These constraints control the rank structure of the entanglement matrices and imply that, in the relevant sectors, the OES counting must match the counting of bulk quasihole zero modes. The latter equals the counting of edge modes at a hard-wall boundary placed on the sample (Chandran et al., 2011).
The result applies to Read-Rezayi states and other clustering model states, and it was shown to remain valid even for conformal-field-theory states likely to be bulk gapless, such as the Gaffnian. In this sense, the correspondence is not merely numerical pattern matching but a consequence of the algebraic structure of the wavefunction space (Chandran et al., 2011).
A related lowest-Landau-level analysis later emphasized that Haldane pair amplitudes 3, introduced as angular-momentum-resolved generalizations of Tan’s contact, provide a complete description of translation-invariant and rotation-invariant lowest-Landau-level states. That work noted that the same data encode universal features such as edge spectra and entanglement, placing the correspondence in a broader microscopic framework for quantum Hall correlations (Hofmann et al., 2022).
3. Spin chains, Haldane phases, and symmetry-protected topological order
In one-dimensional symmetry-protected topological phases, the correspondence takes the form of an explicit relation between the entanglement Hamiltonian of a periodic system and the physical Hamiltonian of an open chain. For the spin-1 AKLT and bilinear-biquadratic Hamiltonians in the Haldane phase, the low-lying entanglement spectrum is described by effective spin-4 degrees of freedom localized at the virtual ends of the subsystem, exactly paralleling the emergent edge spins of the open chain (Roy et al., 2020).
For a contiguous block in the AKLT chain, the entanglement Hamiltonian takes the form
5
with 6 acting on the effective edge spins. The entanglement spectrum then reproduces the singlet-triplet structure of two interacting edge spin-7 variables, and the entanglement gap scales with the same correlation length that governs the physical open-chain edge splitting. This is the precise content of the bulk-edge correspondence in the Haldane phase (Roy et al., 2020).
For non-contiguous partitions, the entanglement Hamiltonian generalizes to a ring of interacting edge spins,
8
which suggests a relation between symmetry-protected topological phases and conformal field theory through the effective spin-9 Heisenberg description. The same work connected this structure to the string order parameter, showing that the entanglement gap and string order decay with the same correlation length at the AKLT point (Roy et al., 2020).
This spin-chain realization clarifies a common point of terminology. In these systems, the Li-Haldane correspondence does not mean a statement about chiral edge conformal field theories, but rather that the virtual edge degrees of freedom exposed by the entanglement cut are the same fractionalized modes that appear at a physical boundary of the open system (Roy et al., 2020).
4. Sector-resolved state counting and tensor-network realizations
The correspondence is frequently applied diagnostically by comparing low-lying entanglement state counting with the representation content of an anticipated edge conformal field theory. This use is particularly prominent in projected entangled pair states (PEPS) for 0-dimensional topological phases, where the entanglement spectrum can be organized by topological sector and momentum around a cylinder (Arildsen et al., 2022).
A detailed counterexample to naive diagnostic use was given for an 1 spin-liquid PEPS with 2 topological order and nine minimally entangled sectors. Although the state is non-chiral, it strongly breaks time-reversal and reflection while preserving their product, leading to entanglement spectra with both right- and left-moving branches but with vastly different velocities. In the topologically trivial sector, and in certain other sectors, the low-lying entanglement spectrum precisely follows the Li-Haldane state counting of a chiral 3 conformal field theory; for example, the singlet sector exhibits the counting
4
even though the underlying topological order is doubled and non-chiral (Arildsen et al., 2022).
The explanation is that the full entanglement spectrum is a non-chiral tensor-product structure built from a “high-velocity” chiral 5 conformal field theory sector and a “low-velocity” chiral 6 sector. When the velocity ratio is large, descendant states of the fast branch are pushed to higher entanglement energy, so a restricted low-energy window can look exactly chiral. Sector-by-sector analysis of all nine anyon sectors is therefore necessary; counting in a single sector, especially the trivial sector, is insufficient to distinguish a truly chiral phase from a non-chiral but close-to-chiral state (Arildsen et al., 2022).
This episode established an important refinement of the correspondence: Li-Haldane counting is a strong indicator of edge structure, but not by itself a sufficient criterion for chirality.
5. Experimental access and entanglement-Hamiltonian reconstruction
The quasi-locality of the entanglement Hamiltonian has motivated direct experimental strategies for probing the correspondence in synthetic quantum systems. A concrete proposal introduced two protocols: Entanglement Hamiltonian Tomography (EHT) for Gaussian systems and Quantum Variational Learning (QVL) for interacting systems. Both exploit the fact that 7 can be approximated by a deformed local Hamiltonian with a manageable number of parameters, rather than by a generic nonlocal operator (Zache et al., 2021).
For a lattice integer quantum Hall state in the Harper-Hofstadter model, the reconstructed entanglement spectrum on ladders such as 8 and 9 reproduces the expected chiral-boson counting 0, as well as the linear spatial profile of the Bisognano-Wichmann couplings near the cut. For an interacting one-dimensional symmetry-protected topological phase realized in a bond-alternating XXZ chain, the learned entanglement spectrum shows the characteristic four-fold degeneracy in the topological phase and its absence in the trivial phase (Zache et al., 2021).
These proposals emphasize a shift in how the correspondence is used. Instead of treating the entanglement spectrum solely as a numerical diagnostic extracted from a wavefunction, they treat 1 as an experimentally learnable and implementable operator. Entanglement spectroscopy can then be performed through quench dynamics under the learned entanglement Hamiltonian, making the Li-Haldane relation testable rather than merely inferential (Zache et al., 2021).
The same framework also reinforces the theoretical role of quasi-locality. The feasibility of EHT and QVL depends precisely on the expectation that low-lying entanglement physics is controlled by deformations of local edge dynamics, which is the structural premise behind most uses of the correspondence (Zache et al., 2021).
6. Limits, breakdowns, and counterexamples
The modern literature no longer treats the Li-Haldane correspondence as universally valid. One line of counterexamples comes from the two-dimensional AKLT model with a tunable boundary on the square-octagon lattice. There, quantum Monte Carlo calculations of large-scale entanglement spectra show that the low-lying entanglement spectrum does not always track the energy spectrum of the subsystem edge. In particular, when a perturbation is applied only on the environment boundary, the physical edge of the subsystem can remain gapless while the entanglement spectrum becomes gapped. This led to the proposal of a more general “wormhole mechanism,” based on the replica path integral for 2, in which low-lying entanglement levels are controlled by near-boundary worldlines from both subsystem and environment, weighted by their coupling across the cut (Liu et al., 2023).
A second line of counterexamples arises in the momentum-resolved entanglement spectrum of spin-3 ladders in the Haldane phase. Exact diagonalization up to 40 spins resolved two distinct low-energy branches at 4 and 5, rather than a single des Cloizeaux-Pearson-type 6 mode. Upon breaking 7 symmetry with XXZ anisotropy, the entanglement Hamiltonian undergoes an entanglement quantum phase transition at 8, while the bulk Haldane-to-stripe-Néel transition occurs only at 9. In the easy-plane regime the entanglement ground state becomes quasi-degenerate across 0 sectors in the thermodynamic limit, consistent with spontaneous 1 symmetry breaking; in the easy-axis regime it becomes doubly degenerate, consistent with spontaneous 2 symmetry breaking. The authors interpret this as direct evidence that the entanglement Hamiltonian is sufficiently nonlocal to evade conventional one-dimensional constraints such as the Lieb-Schultz-Mattis and Mermin-Wagner theorems, and conclude that the Li-Haldane correspondence can fail even at the level of low-energy dispersions (Tzeng et al., 3 Sep 2025).
These counterexamples do not merely show finite-size or numerical limitations. They indicate that the entanglement spectrum may depend on the structure of the environment, on the geometry of the bipartition, and on the nonlocality of 3 in ways that are not captured by a simple edge-theory analogy. A common misconception is therefore that any low-lying entanglement spectrum should reproduce the physical boundary spectrum of the subsystem; the counterexamples show that this is not generally true.
7. Generalized formulations and current perspective
Recent work has also extended the correspondence rather than only restricting it. For critical free-fermion systems protected by global on-site symmetries, an exact relation has been established between the bulk entanglement spectrum and the boundary energy spectrum at topological criticality in arbitrary dimensions. In this setting, the single-particle entanglement eigenvalues 4 are related to the entanglement-Hamiltonian single-particle energies 5 by
6
and the zero modes of the physical Hamiltonian under open boundary conditions correspond one-to-one to mid-spectrum entanglement modes at 7. The argument uses a quasi-local Gaussian circuit that maps the physical Hamiltonian to the entanglement Hamiltonian while preserving the localized zero-mode content (Guo et al., 24 Sep 2025).
This “generalized Li-Haldane correspondence” differs from the original gapped-phase formulation in two ways. First, it applies at criticality, where conventional topological invariants are often ill-defined. Second, it identifies topology not through low-lying dispersions alone but through protected entanglement mid-spectrum degeneracies tied to boundary zero modes. The result is supported by lattice simulations in one, two, and three dimensions, including robustness to symmetry-preserving disorder (Guo et al., 24 Sep 2025).
The current perspective is therefore layered rather than unitary. In fractional quantum Hall model states and in some symmetry-protected topological chains, the correspondence can be made precise as a bulk-edge equivalence of counting and low-energy structure. In PEPS and related tensor-network states, it remains a powerful but sector-sensitive diagnostic. In boundary-perturbed or extensively bipartitioned systems, it can fail because the entanglement Hamiltonian is influenced by both subsystem and environment and may be intrinsically nonlocal. In critical free-fermion systems, a reformulated version survives as an exact boundary-zero-mode to entanglement-mid-spectrum relation. The phrase “Li-Haldane correspondence” thus now denotes a family of related claims about how topological edge information is encoded in entanglement spectra, ranging from rigorous state-counting theorems to conditional diagnostic heuristics and symmetry-protected critical correspondences (Chandran et al., 2011, Arildsen et al., 2022, Liu et al., 2023, Tzeng et al., 3 Sep 2025, Guo et al., 24 Sep 2025).