Regionally Localized Entanglement

• Regionally localized entanglement is the confinement of quantum correlations to a specified region, interface, or junction rather than being uniformly global.
• It is characterized by diverse diagnostic methods—including vacuum subtraction, entanglement minima, and replica-twist measures—depending on subsystem definitions and physical settings.
• The concept has practical applications in quantum field theory, discrete-time quantum walks, many-body localization, and measurement-based protocols for concentrating entanglement.

Regionally localized entanglement denotes a family of constructions in which entanglement is tied to a specified region, interface, junction, or selected subsystem rather than being treated as an undifferentiated global resource. In the cited literature, this includes vacuum-subtracted entanglement of spatial regions containing localized excitations, entanglement minima that diagnose topological interface states in discrete-time quantum walks, genuine multipartite entanglement confined to a correlation-length neighborhood of subsystem-boundary junctions, spatio-temporally heterogeneous entanglement in many-body localized systems, and measurement-based concentration of entanglement onto a chosen subset of qubits (Mondal, 2015, Chandrashekar et al., 2015, Iizuka et al., 18 Feb 2026, Krishnan et al., 7 Aug 2025, Vairogs et al., 9 Oct 2025). A recurring theme is that “localization” is not a single universal observable: it depends on how subsystems are defined, which entanglement monotone is used, and whether the problem is static, dynamical, or operational.

1. Conceptual scope and recurring definitions

The literature uses several inequivalent but structurally related notions of regional localization. In one line of work, a spatial region $R$ is the primary subsystem, and the relevant object is the reduced density matrix $\rho_R$, often with explicit subtraction of the vacuum contribution in QFT (Mondal, 2015). In another, a region is a two-qubit neighborhood $\{1,i\}$ in a larger network, and the task is to maximize the average entanglement that can be concentrated on that region by measuring the complement (Krishnan et al., 7 Aug 2025). In yet another, a region is a geometric feature such as a topological interface or a multipartite junction, and localization is inferred from entropy minima or from saturation to a finite plateau controlled by the correlation length (Chandrashekar et al., 2015, Iizuka et al., 18 Feb 2026).

Setting	Region notion	Representative diagnostic
Local QFT	Spatial region $R$	$S_n(R)-S_n^{(vac)}$
DTQW	Interface at $x=0$	$S(\rho_c)$ or $\mathcal N(\rho)$ minima
Gapped free fermions	Junction of subsystem boundaries	$GM_2^{(q)}$
Quantum networks	Two-qubit region $\{1,i\}$	$\rho_R$0, $\rho_R$1
Large multipartite states	Target subsystem $\rho_R$2	LE and EA via average $\rho_R$3-tangle

A common misconception is that more entanglement always signals stronger localization. The split-step quantum-walk analysis shows the opposite in a precise setting: strong topological localization is accompanied by pronounced local minima of coin-position entanglement (Chandrashekar et al., 2015). A second misconception is that multipartite entanglement in gapped systems is generically distributed along all subsystem boundaries. The junction-law results instead show exponential suppression when no point is adjacent to all participating regions (Iizuka et al., 18 Feb 2026).

2. Spatial regions, localized particles, and localized orbitals

Mondal considered states in a local QFT that describe two entangled particles localized in disjoint regions $\rho_R$4 and $\rho_R$5. For a one-particle wave packet with field operator $\rho_R$6, the smeared creation operator is

$\rho_R$7

For a two-particle entangled state of the form

$\rho_R$8

with $\rho_R$9 nearly supported in $\{1,i\}$0 and $\{1,i\}$1 nearly supported in $\{1,i\}$2, the main leading-order result is

$\{1,i\}$3

equivalently

$\{1,i\}$4

Thus the Rényi entropy of a spatial region containing one localized particle reproduces the ordinary quantum-mechanical particle entanglement after subtraction of the vacuum contribution, up to corrections controlled by wave-function overlap across the boundary (Mondal, 2015).

This framework is technically significant because it identifies the vacuum subtraction $\{1,i\}$5 as indispensable rather than optional. The QFT vacuum already has nonzero regional entanglement, so the physically relevant quantity for localized excitations is the excess over the vacuum. Mondal also emphasized that the overlap corrections are never identically zero in a local QFT, but can be made exponentially small when the wave packets are well separated compared to their widths (Mondal, 2015). This suggests a precise operational meaning of localized-particle entanglement for indistinguishable excitations.

A closely related orbital formulation was developed by Ding, Dünnweber and Schilling for realistic fermionic systems. There the subsystem is not a macroscopic half-chain but a pair of localized orbitals, and physically accessible entanglement must respect local superselection rules. For two spatial orbitals $\{1,i\}$6 and $\{1,i\}$7, they first construct the number-parity-SSR projection

$\{1,i\}$8

and then evaluate the entanglement of the projected state. In free electron chains and in an interacting $\{1,i\}$9 ring, this produces an orbital-resolved map of where entanglement resides, including long-distance entanglement at sufficiently low or high filling fractions (Ding et al., 2023). A plausible implication is that “regional” localization in fermionic matter is often best understood in a basis of localized orbitals rather than only through contiguous spatial cuts.

Related constructions extend the regional idea beyond ordinary spatial slices. A modular-operator approach associates a causal diamond $R$0 with a local von Neumann algebra $R$1, modular operator $R$2, and modular Hamiltonian $R$3, allowing concurrence and entanglement entropy to be formulated for causally closed regions (Gallaro et al., 2022). Another localization scheme for scalar fields on curved spacetime argues that separable states correspond to flat spacetime while nonzero entanglement induces a nontrivial metric coefficient $R$4 and nonzero scalar curvature (Kiosses, 2020). These are specialized frameworks, but they show that regional localization can be algebraic and geometric as well as spatial.

3. Topological interface states in discrete-time quantum walks

In one-dimensional discrete-time quantum walks, localized states can be engineered by placing different coin parameters on the two sides of a position-space interface. The split-step walk is

$R$5

with

$R$6

and the interface is created by choosing $R$7 for $R$8 and $R$9 for $S_n(R)-S_n^{(vac)}$0 (Chandrashekar et al., 2015). Both the split-step and double-split-step walks obey chiral symmetry

$S_n(R)-S_n^{(vac)}$1

with two independent winding numbers $S_n(R)-S_n^{(vac)}$2. When $S_n(R)-S_n^{(vac)}$3 changes across $S_n(R)-S_n^{(vac)}$4, bulk-boundary correspondence yields a topologically protected eigenstate at quasienergy $S_n(R)-S_n^{(vac)}$5 or $S_n(R)-S_n^{(vac)}$6, localized at the interface (Chandrashekar et al., 2015).

The entanglement diagnostic is based on the reduced coin state $S_n(R)-S_n^{(vac)}$7 and the von Neumann entropy

$S_n(R)-S_n^{(vac)}$8

An alternative is the negativity

$S_n(R)-S_n^{(vac)}$9

where $x=0$0 are the eigenvalues of the partial transpose $x=0$1. Sweeping left and right coin parameters and computing $x=0$2 after a fixed number of steps, the numerical landscape exhibits pronounced local minima precisely in parameter regions where the topological invariant changes across the interface. The paper states that “the deepest minima (dark blue) occur precisely where the bulk winding number $x=0$3 changes across $x=0$4, indicating a strongly localized $x=0$5 edge state” (Chandrashekar et al., 2015).

The physical interpretation is specific and somewhat counterintuitive: a strongly localized topological state reduces the coin-position entanglement because the walker’s state is concentrated into a sharply localized bound-state component with minimal leakage into the bulk. This is why minima, not maxima, provide the “witness” for localization. The same paper further models uncorrelated bit-flip noise by

$x=0$6

and shows that the entanglement minima remain robust for small $x=0$7, because the localized interface component is an eigenstate of $x=0$8 (Chandrashekar et al., 2015). The authors note that such an entanglement-based diagnostic is directly applicable in forthcoming experiments in photonic, cold-atom, and trapped-ion platforms.

4. Junction laws for genuine multipartite entanglement

A distinct use of localization appears in the 2026 junction-law analysis of genuine multipartite entanglement in gapped $x=0$9-dimensional free-fermion lattices. The diagnostic is the Rényi-2 genuine multi-entropy $S(\rho_c)$0, constructed from replica-twist multi-entropies by subtracting all lower-order contributions so that $S(\rho_c)$1 and it vanishes on states whose correlations can be built from at most $S(\rho_c)$2-party entanglement (Iizuka et al., 18 Feb 2026). For $S(\rho_c)$3 and $S(\rho_c)$4, the paper gives explicit linear combinations of $S(\rho_c)$5, $S(\rho_c)$6, $S(\rho_c)$7, and $S(\rho_c)$8.

For partitions consisting of $S(\rho_c)$9 wedges meeting at a single point, the numerics show the scaling law

$\mathcal N(\rho)$0

Here $\mathcal N(\rho)$1 is the linear subsystem size and $\mathcal N(\rho)$2 is the correlation length set by the staggered mass $\mathcal N(\rho)$3. The interpretation given in the paper is that for $\mathcal N(\rho)$4, genuine $\mathcal N(\rho)$5-partite entanglement is confined to an $\mathcal N(\rho)$6 neighborhood of the junction, so enlarging the wedges further cannot increase $\mathcal N(\rho)$7 (Iizuka et al., 18 Feb 2026).

The contrast with no-junction partitions is sharp. If no point is adjacent to all $\mathcal N(\rho)$8 regions, then irreducible $\mathcal N(\rho)$9-body correlators must bridge distances of order $GM_2^{(q)}$0, and numerically

$GM_2^{(q)}$1

In the large-$GM_2^{(q)}$2 regime, the signal drops below numerical resolution (Iizuka et al., 18 Feb 2026). This is one of the clearest cases in which “localization” is not a metaphor: the multipartite quantity itself is exponentially suppressed away from the geometrical junction.

The tripartite and quadripartite cases both exhibit universal collapse as functions of $GM_2^{(q)}$3. For $GM_2^{(q)}$4, the normalized quantity $GM_2^{(q)}$5 approaches a universal curve $GM_2^{(q)}$6, linear for $GM_2^{(q)}$7 and saturating near unity for $GM_2^{(q)}$8. For $GM_2^{(q)}$9, both $\{1,i\}$0 and $\{1,i\}$1 show the same growth-and-saturation pattern, although $\{1,i\}$2 may be slightly negative in the crossover (Iizuka et al., 18 Feb 2026). The accompanying holographic picture interprets this through a “minimal multiway-cut” that can access the IR cap only when the boundary regions meet at a sufficiently small junction.

5. Dynamical localization and heterogeneity in many-body systems

In many-body localized systems, regional localization is fundamentally dynamical. Nanduri, Kim, and Huse studied the XXZ chain with random longitudinal fields and showed that half-chain entanglement grows as

$\{1,i\}$3

where $\{1,i\}$4 is the localization length governing the decay $\{1,i\}$5 of effective l-bit interactions, and $\{1,i\}$6 is the long-time entanglement density per spin set primarily by the initial state (Nanduri et al., 2014). In this phenomenology, by time $\{1,i\}$7 only l-bits within $\{1,i\}$8 of the cut have become appreciably entangled. The entanglement front is therefore confined to a logarithmically growing neighborhood rather than spreading ballistically or diffusively (Nanduri et al., 2014).

A more local characterization was given by Bera and collaborators through the on-site concurrence

$\{1,i\}$9

where $\rho_R$00 is the usual two-site concurrence, and the site-dependent relaxation time

$\rho_R$01

They then defined the spatial correlation function

$\rho_R$02

and extracted an entanglement length $\rho_R$03 from

$\rho_R$04

This yields dynamically correlated “active” and “inactive” clusters, with cluster-size histogram $\rho_R$05 deep in the MBL phase (Artiaco et al., 2021). The notable conclusion is that $\rho_R$06 grows as disorder increases or as the initial energy approaches the spectral edges, even though localization suppresses transport.

Xu and Zhang extended the dynamical picture to structured initial states in the random-field XXZ model. Using both the second-Rényi entropy and the Wehrl–Rényi entropy $\rho_R$07, they showed that for $\rho_R$08-polarized product states the net entanglement growth under MBL dynamics is non-monotonic in the initial scrambling time $\rho_R$09: it first rises because finite local magnetization must relax, then falls once $\rho_R$10 and only inter-site correlations remain (Xu et al., 20 May 2026). For $\rho_R$11-polarized product states, where $\rho_R$12 from the outset, the growth decreases monotonically with $\rho_R$13 (Xu et al., 20 May 2026). The paper interprets this as entanglement build confined to clusters of size $\rho_R$14 set by LIOM dressing.

An even more radical state-dependent notion of localization appears in Lee’s relatively local Hamiltonians. There the effective geometry is determined by bi-local variables

$\rho_R$15

and a state has local structure when these entanglement bonds fall off rapidly with an emergent lattice distance (Lee, 2018). For initial nearest-neighbor bond strength $\rho_R$16, the coordinate speed of entanglement spread scales as $\rho_R$17, while exact product states with $\rho_R$18 have $\rho_R$19 and remain strictly ultra-local at all times (Lee, 2018). This suggests that regional confinement of entanglement can arise not only from disorder or topology but from the initial-state-dependent emergence of locality itself.

6. Measurement-based concentration onto selected regions

In multipartite quantum networks, regionally localized entanglement has been defined operationally as the average entanglement that can be concentrated on a two-qubit region by measuring away the remainder. For an $\rho_R$20-qubit state $\rho_R$21 on $\rho_R$22, with qubit $\rho_R$23 designated as the hub, the entanglement on region $\rho_R$24 is

$\rho_R$25

where the paper uses the rescaled negativity

$\rho_R$26

The total regionally localized entanglement is

$\rho_R$27

This quantity is bounded using the block-localizable entanglement

$\rho_R$28

or, when $\rho_R$29, the ordinary block entanglement $\rho_R$30 (Krishnan et al., 7 Aug 2025).

For pure permutation-symmetric states, the paper proves

$\rho_R$31

when $\rho_R$32, and

$\rho_R$33

when $\rho_R$34 (Krishnan et al., 7 Aug 2025). The symmetric gGHZ states saturate the upper bound, while the $\rho_R$35-qubit $\rho_R$36 state satisfies $\rho_R$37. By contrast, for generalized-$\rho_R$38 states in a fixed magnetization sector,

$\rho_R$39

with the sharper bounds

$\rho_R$40

The paper emphasizes that no single bound simultaneously describes the permutation-symmetric and fixed-magnetization classes (Krishnan et al., 7 Aug 2025).

A high-dimensional asymptotic variant studies localization onto a target region $\rho_R$41 by measuring its complement $\rho_R$42. The figures of merit are the entanglement of assistance

$\rho_R$43

and the localizable entanglement

$\rho_R$44

where $\rho_R$45 is the $\rho_R$46-tangle (Vairogs et al., 9 Oct 2025). A key simplification is the fidelity formula

$\rho_R$47

For Haar-random states, if $\rho_R$48 then both global and local quantities scale to zero as $\rho_R$49, whereas if $\rho_R$50 then $\rho_R$51 but $\rho_R$52 (Vairogs et al., 9 Oct 2025). The paper describes this as a phase-transition-like dependence on whether more or less than half of the qubits are measured. It also proves, for random graph states, that the condition $\rho_R$53 is equivalent to solvability of the linear system $\rho_R$54 over $\rho_R$55, yielding an $\rho_R$56 decision procedure (Vairogs et al., 9 Oct 2025).

Taken together, these operational results show that regional localization can mean concentration rather than confinement: entanglement is not merely observed to live in a region, but is actively focused there by measurements. They also show that the locality of the measurement protocol is itself decisive. In large systems, global measurements may localize substantial multipartite entanglement onto $\rho_R$57 even when local measurements asymptotically fail (Vairogs et al., 9 Oct 2025).