Extended Entanglement Entropy

• Extended entanglement entropy is a framework that generalizes standard entanglement measures by embedding non-factorizable physical Hilbert spaces into an extended space that includes edge modes and gauge constraints.
• It decomposes the total entropy into classical, dimension, and quantum (EPR) components, enabling precise quantification of both nonlocal correlations and operationally accessible entanglement.
• The approach serves as a robust diagnostic of topological order and fracton phases, providing universal invariants validated through lattice gauge theories, TQFTs, and string theoretic models.

Extended entanglement entropy refers to a class of entanglement measures and structural decompositions that arise when the standard framework of subregion factorization fails or is nontrivially generalized. This concept plays a central role in the characterization of quantum correlations in systems with gauge constraints (lattice gauge theory, TQFTs), systems with nontrivial boundary or edge-mode dynamics, topologically ordered phases (notably in higher dimensions and fracton models), and in situations where the Hilbert-space factorization is replaced by an “extended” or “enlarged” Hilbert space to accommodate nonlocal constraints or edge excitations. Extended entanglement entropy frameworks provide rigorous means to identify universal invariants, including robust subleading corrections beyond the conventional “area law,” and systematically account for emergent boundary degrees of freedom.

1. Extended Hilbert Space and Edge Mode Formalism

In ordinary quantum systems on a lattice, the Hilbert space factorizes as $$\mathcal{H} = \mathcal{H}_A \otimes \mathcal{H}_{\bar{A}}$$ for a partition into regions $A$ and $\bar{A}$, and the reduced density matrix is $\rho_A = \operatorname{Tr}_{\bar{A}}\, \rho$. However, in gauge theories or models with nontrivial local constraints (such as lattice gauge theories, TQFTs, and string field theory), the physical Hilbert space $\mathcal{H}_{\rm phys}$ does not admit such tensor decomposition due to Gauss's law or other constraints coupling the boundary degrees of freedom across the entanglement cut. The “extended Hilbert space” method resolves this by embedding $\mathcal{H}_{\rm phys}$ into a larger $H_{\rm ext} = H_A \otimes H_{\bar{A}}$, where gauge invariance is relaxed at the entangling surface, or by explicitly introducing edge-mode degrees of freedom to account for the residual symmetry action and boundary data (Aoki et al., 2015, Soni et al., 2015, Moitra et al., 2018, Geiller et al., 2019).

This facilitates a well-defined reduced density matrix and a von Neumann entropy that encapsulates intrinsically nonlocal or symmetry-enforced correlations. For example, in lattice gauge theory (Abelian or non-Abelian), edge modes correspond to allowed configurations of electric flux or their non-Abelian Casimirs at the boundary. Similarly, in topological quantum field theories (TQFTs) and string theory, the extended Hilbert space accounts for dynamical modes localized at the entanglement cut, such as stringy edge states or open-string zero modes (Balasubramanian et al., 2018).

These constructions ensure that the entanglement entropy thus defined satisfies the standard information-theoretic properties (positivity, subadditivity, strong subadditivity), matches replica-path-integral calculations, and incorporates both bulk and boundary contributions arising from the interplay of constraints and gauge invariance.

2. Structure of Extended Entanglement Entropy: Decomposition and Universality

A salient feature of extended entanglement entropy is its characteristic decomposition into multiple contributions, which reflect the mixture of classical and quantum correlations enforced by gauge or topological constraints. The canonical decomposition, valid for both Abelian and non-Abelian lattice gauge theories and two-dimensional TQFTs, is

$$S_{\rm EE} = S_{\rm cl} + S_{\rm dim} + S_{\rm EPR}$$

where:

• $S_{\rm cl}$ is the classical Shannon entropy of the boundary superselection sector distribution (e.g., quantum numbers labeling normal flux, representation class, or other sectoral data at the entanglement surface).
• $S_{\rm dim}$, which appears in non-Abelian and topological systems, encodes the “dimension” or “color entanglement” due to the multiplicity of irreducible representations associated with boundary degrees of freedom.
• $S_{\rm EPR}$ is the genuinely quantum, “distillable” Bell-pair entanglement accessible within each fixed superselection sector (1705.01549, Soni et al., 2015, Donnelly et al., 2018).

The edge-mode structure enforces a block-diagonal form on the reduced density matrix, $\rho_A = \bigoplus_i p_i \rho_i$, where $p_i$ is the probability for superselection sector $i$, and $\rho_i$ is the density matrix within that sector. The different terms are then transparent: $S_{\rm cl} = -\sum_i p_i \log p_i$, $S_{\rm dim} = \sum_i p_i \sum_a \log d^i_a$ (sum over boundary representation dimensions), $$S_{\rm EPR} = -\sum_i p_i \operatorname{Tr} \rho_i \log \rho_i$$.

Only the “quantum” term $S_{\rm EPR}$ contributes to distillable Bell pairs or can be exchanged under LOCC protocols, while $S_{\rm cl}$ and (in non-Abelian contexts) $S_{\rm dim}$ are fundamentally superselection-induced and are not operationally accessible (Soni et al., 2015, Moitra et al., 2018).

In two-dimensional (open/closed) TQFTs, the extended entanglement entropy can be recast in the algebraic language of Frobenius algebra structures and the Moore–Segal axioms, with an “entanglement brane” axiom capturing the physical irrelevance of certain boundary conditions and further clarifying the role of edge modes (Donnelly et al., 2018).

3. Extended Topological Entanglement Entropy in Higher Dimensions

A paradigmatic application of extended entanglement entropy is the diagnosis of topological order in higher-dimensional systems, notably three-dimensional fracton phases. In contrast to two-dimensional topological phases, where the universal entanglement invariant is a constant (topological entanglement entropy), in certain fracton stabilizer codes (e.g., X-cube model, Haah's code), the universal correction scales linearly with subsystem size along a preferred axis: $S(A) = A_1 R^2 - \gamma R + c + o(1),$ where $R$ is the linear subsystem size, $A_1$ is a non-universal area-law coefficient, $\gamma$ is a universal “topological” coefficient quantifying the number of nonlocal constraints or logical operators, and $c$ is an offset (Ma et al., 2017).

This “extended” topological term $\gamma R$ is robust under arbitrary local perturbations (preserving the gap, as shown via Schrieffer–Wolff flows), and remains a universal diagnostic of fracton order, functionally tying ground-state wavefunction constraints to new entropic invariants. In such models, the subleading linear term arises from the counting of independently nonlocal stabilizer constraints (membrane or loop operators) intersecting the region and is directly accessible through generalized Kitaev–Preskill/Levin–Wen prescriptions (ABC and PQWT schemes), which systematically cancel area-law and corner effects to isolate topological contributions.

A summary of results for two archetypal fracton models is as follows:

Model	Prescription	$\gamma$	Topological Entropy
X-cube	ABC	1	$S^{ABC}_{\rm topo} = -R$
X-cube	PQWT	2	$S^{PQWT}_{\rm topo} = -2R-1$
Haah's code	ABC	2	$S^{ABC}_{\rm topo} = -2R+2$
Haah's code	PQWT	4	$S^{PQWT}_{\rm topo} = -4R+12$

This universal linear-in-size term is a robust fingerprint of 3D topological order and extends even to localization-protected excited states in the presence of strong disorder (Ma et al., 2017).

4. Dynamics of Edge Modes and Boundary Symplectic Structure

Beyond lattice and block-diagonal constructions, the inclusion of edge modes as dynamical degrees of freedom is realized via an extended action approach. In the presence of boundaries, the action is enlarged by boundary terms involving Stückelberg or edge-mode fields, and a modified variational principle yields an extended (pre-)symplectic form: $$\Omega_{\rm ext} = \int_\Sigma \delta\theta + \int_{\partial\Sigma} \delta\vartheta,$$ where $\theta$ is the bulk symplectic potential, and $\vartheta$ captures the boundary dynamics (Geiller et al., 2019). Integrating out the bulk with fixed edge-mode data leaves a nontrivial effective boundary theory, which governs the edge contributions to the entanglement entropy.

For Abelian Chern–Simons theory, this formalism recovers the topological entropy $S_{\rm EE} = -\frac12 \ln k$ (with $k$ the level), with boundary current algebras matching the underlying topological edge structure, and similar results hold for Abelian BF theory.

The edge-mode contribution provides a universal, regulator-independent addition to the entropy, rigorously derived from the boundary state space dimension or algebraic structure, and admits generalization to all gauge theories and gravity, where horizon or boundary “soft modes” are entangled in a highly nontrivial way (Geiller et al., 2019).

5. Extended Approaches in TQFTs, String Theory, and Non-Local QFTs

In two-dimensional TQFT and Yang–Mills, the extended entanglement entropy is formulated via open–closed TQFT with a coupling between open- and closed-sector state spaces and a special “entanglement brane” object designed to encapsulate the invisibility of certain cut boundary labels (Donnelly et al., 2018). The entropy again splits into Shannon and edge (dimension) parts, and the modular Hamiltonian, negativity, and area-law analogues are all computable in this formalism.

In covariant string field theory, the extended phase space is constructed on the space of open-string configurations, with dynamical BRST edge modes present at the subregion boundary (Balasubramanian et al., 2018). The entanglement entropy of a spatial subregion is then computed as a sum over all string states, with modular invariance regularizing the sum and providing a finite answer once tachyonic divergences are removed. The Hilbert space does not factorize on naive subregions in the string field context and demands an extended construction with edge-mode gluing conditions.

6. Physical Implications and Universal Invariants

Extended entanglement entropy frameworks rigorously extract universal, nonlocal invariants associated with topological phases (e.g., total quantum dimension in non-Abelian toric codes, group-theoretic multiplicities in fusion basis representations), and provide robust diagnostics of nontrivial order beyond symmetry breaking. In fracton models, the extended entropy identifies a new class of “entropic invariants” proportional to subsystem size, surviving disorder and local perturbations and classifying phases lacking a continuum TQFT description (Ma et al., 2017).

The separation of the entropy into classical (non-distillable) and quantum (distillable) parts has operational consequences for entanglement extraction, quantum information protocols, and the physical meaning of mutual information or relative entropy in gauge and gravitational systems. Notably, only the quantum part survives in mutual or relative entropies between well-separated regions, as the classical edge-mode contribution is shown to be strictly UV-dominated and cancels in continuum limits (Moitra et al., 2018).

The machinery of extended entanglement entropy is foundational for a complete understanding of entanglement in quantum field theory, quantum gravity, and beyond, providing a unified formalism linking bulk, boundary, and edge degrees of freedom, and identifying nonlocal invariants with deep physical and mathematical significance (Aoki et al., 2015, Soni et al., 2015, Donnelly et al., 2018, Ma et al., 2017).