Quantum Localization

• Quantum localization is a phenomenon where quantum states and dynamics become confined due to disorder, interference, geometric constraints, or controlled dissipation.
• Diagnostic techniques such as the inverse participation ratio, level statistics, and spectral analysis precisely measure localization and distinguish phases like Anderson and many-body localization.
• Quantum-enhanced localization leverages entangled probing and advanced algorithms to improve sensor accuracy and computational efficiency in position estimation.

Quantum localization denotes a family of non-equivalent notions that share a common theme: a quantum state, quantum dynamics, observable algebra, or operational capability becomes confined relative to a physically relevant structure. In condensed-matter and many-body physics, it refers to the suppression of transport and relaxation due to interference or disorder; in relativistic quantum theory, it refers to localization of observables rather than Born-type localization of states; in more recent work, it also denotes equatorial concentration of probability on $S^2$, localization of a state within a prescribed subspace, localization of unclonable quantum information at a spacetime point, and quantum-enhanced estimation of spatial position through entangled ranging (Kudo, 2023, Balachandran, 2016, Ye et al., 29 Apr 2026, Salcedo, 14 Jan 2026, Bartusek et al., 29 May 2026, He et al., 2024).

1. Disorder, interference, and spectral confinement

In the disorder-driven setting, quantum localization refers to the suppression of transport and relaxation in quantum systems due to interference or disorder. The canonical single-particle form is Anderson localization: in noninteracting systems with quenched disorder, single-particle eigenstates become exponentially localized, with hallmarks that include absence of diffusion, lack of thermalization of local observables, Poissonian level statistics, and vanishing long-time transport. Its interacting extension is many-body localization (MBL): in strongly disordered interacting systems, many-body eigenstates can be localized, leading to failure of thermalization even at finite energy density, memory of initial conditions, logarithmically slow entanglement growth after a quench, Poissonian level statistics in the localized regime, and a disorder-driven crossover or transition between thermal and localized behavior (Kudo, 2023).

This spectral viewpoint extends beyond point particles. In “Localization of Extended Quantum Objects” the motion of a string is reduced to a lower-dimensional many-body localization problem along its internal coordinate, so that localization of point-like internal modes implies localization of the entire string. The same hierarchical logic is applied to membranes: membranes move via strings, which move via kinks, and sufficiently strong disorder localizes the internal modes and freezes the global motion. The paper formulates a convergent “string locator expansion” and identifies a string-localized phase by out-of-time-order string correlators, thereby generalizing localization-protected behavior to domain walls, loop excitations, and flux lines (Pretko et al., 2017).

Localization also appears without quenched disorder. In periodically driven chaotic maps, dynamical localization halts classically diffusive momentum growth after a finite break time and yields an exponentially localized stationary distribution in momentum space. For the quantum sawtooth map, the stationary profile is

$$W_m \approx \frac{1}{\ell}\exp\!\left[-\frac{2|m-m_0|}{\ell}\right],$$

with $t^\star \approx \ell \approx D$, so localization is encoded in the Floquet spectrum rather than in a static random Hamiltonian (Pizzamiglio et al., 2021). In discrete-time coined quantum walks, localization is controlled by the spectral type of the one-step unitary $U$: for walks with coins that are eventually periodic on the left and right, eigenvalues arise precisely when an eigenphase lies in a spectral gap of both asymptotic periodic operators and satisfies a finite matching condition across the interface. Those eigenvalues generate exponentially decaying bound states and non-vanishing time-averaged probabilities (Kiumi, 2021).

A recurring misconception is that quantum localization is synonymous with Anderson localization. The cited literature shows a broader landscape: disorder-induced localization, Floquet dynamical localization, interface localization in quantum walks, and localization of extended objects all instantiate confinement of quantum dynamics, but they do so through distinct spectral and geometric mechanisms.

2. Diagnostics and quantitative measures

The localization literature is characterized by a large diagnostic repertoire. For single-particle states, a standard measure is the inverse participation ratio

$\mathrm{IPR}=\sum_i|\psi_i|^4,$

which is $O(1)$ for localized states and scales as $1/L$ for extended states (Ozawa et al., 2019). In MBL, level statistics, local autocorrelators that fail to decay, LIOM representations of the Hamiltonian, and logarithmic-in-time entanglement growth are standard signatures (Pretko et al., 2017). In “Localization protected quantum order,” adjacent gap ratios

$$r_n=\frac{\min(\delta_n,\delta_{n+1})}{\max(\delta_n,\delta_{n+1})}$$

separate Poissonian localized spectra from Wigner–Dyson ergodic spectra, and a distinct paired-versus-unpaired spectral transition appears inside the MBL regime when edge-mode splitting competes with the many-body level spacing (Huse et al., 2013).

Several diagnostics tie localization to experimentally economical observables. In a disordered transverse-field Ising chain on a noisy quantum computer, the magnetization

$$M_z=\langle \psi|\hat S^z|\psi\rangle,\qquad \hat S^z=\sum_{j=1}^{L}\sigma_j^z,$$

serves as a memory diagnostic, while the twist overlap

$$z=\langle \psi|\hat U_{\rm twist}|\psi\rangle,\qquad |z|^2\in[0,1],$$

acts as an eigenstate-sensitive probe: in thermal eigenstates $$W_m \approx \frac{1}{\ell}\exp\!\left[-\frac{2|m-m_0|}{\ell}\right],$$0, whereas in localized eigenstates $$W_m \approx \frac{1}{\ell}\exp\!\left[-\frac{2|m-m_0|}{\ell}\right],$$1 remains finite. The two quantities are computed from the same measurement record $$W_m \approx \frac{1}{\ell}\exp\!\left[-\frac{2|m-m_0|}{\ell}\right],$$2, but they probe distinct physics and differ sharply in noise sensitivity (Kudo, 2023).

A more general linear-response framework relates localization to spectroscopy. Under a weak periodic perturbation $$W_m \approx \frac{1}{\ell}\exp\!\left[-\frac{2|m-m_0|}{\ell}\right],$$3, the integrated excitation rate obeys

$$W_m \approx \frac{1}{\ell}\exp\!\left[-\frac{2|m-m_0|}{\ell}\right],$$4

For $$W_m \approx \frac{1}{\ell}\exp\!\left[-\frac{2|m-m_0|}{\ell}\right],$$5, this yields

$$W_m \approx \frac{1}{\ell}\exp\!\left[-\frac{2|m-m_0|}{\ell}\right],$$6

so spectroscopy measures a localization-sensitive variance without imaging. In the 1D Anderson model, the scaling prediction

$$W_m \approx \frac{1}{\ell}\exp\!\left[-\frac{2|m-m_0|}{\ell}\right],$$7

implies $$W_m \approx \frac{1}{\ell}\exp\!\left[-\frac{2|m-m_0|}{\ell}\right],$$8 (Ozawa et al., 2019).

Resource-theoretic coherence yields another diagnostic language. For a Hamiltonian eigenbasis $$W_m \approx \frac{1}{\ell}\exp\!\left[-\frac{2|m-m_0|}{\ell}\right],$$9, the paper “Quantum coherence and the localization transition” shows that the time-averaged escape probability equals the $t^\star \approx \ell \approx D$0-coherence,

$t^\star \approx \ell \approx D$1

and that coherence-generating power detects the uniformity of the transition matrix between the Hamiltonian and configuration bases. In the same framework, a differential coherence metric equals dynamical conductivity, linking localization directly to response functions (Styliaris et al., 2019).

In phase space, Rényi occupations provide a unifying construction. For a probability distribution $t^\star \approx \ell \approx D$2 on a measure space $t^\star \approx \ell \approx D$3,

$t^\star \approx \ell \approx D$4

The participation ratio is $t^\star \approx \ell \approx D$5, while $t^\star \approx \ell \approx D$6 gives the exponential of the Shannon entropy. In bounded spaces $t^\star \approx \ell \approx D$7; in unbounded spaces, a bounded reference subspace must be chosen, so the definition of maximal delocalization is contextual (Villaseñor et al., 2021).

3. Localization-protected phases and engineered localization

Localization can protect phases that equilibrium thermodynamics would destroy. In disordered Ising, Majorana, and Dirac chains, highly excited many-body eigenstates can break symmetries or display topological order even at energy densities where the thermally equilibrated system is disordered. In the MBL spin-glass regime,

$t^\star \approx \ell \approx D$8

while in the topological Majorana chain an edge operator with exponentially small splitting $t^\star \approx \ell \approx D$9 survives despite the absence of a many-body gap. The central claim is that localization protects order by preventing the mobile defects that would otherwise disorder the phase (Huse et al., 2013).

The same work argues that localized systems can move between ordered and disordered localized phases via non-thermodynamic transitions in the properties of the many-body eigenstates. In the disordered transverse-field Ising chain this yields an MBL paramagnet, an MBL spin glass, an infinite-randomness critical point, and a spectral transition from paired to unpaired Poisson statistics inside the localized regime. These phenomena are eigenstate-structural rather than thermodynamic (Huse et al., 2013).

Dissipation introduces a distinct possibility: localization need not simply be destroyed by openness. In “Localization in open quantum systems,” a disordered tight-binding chain is coupled to phase-engineered local jump operators

$U$0

and the resulting Lindblad dynamics can drive the system into steady states with tunable localization properties. For $U$1 and $U$2, the steady-state weights concentrate near lower-band-edge Anderson modes; for $U$3 and $U$4, they concentrate near upper-band-edge modes; for $U$5 and $U$6, they select band-center modes. Quantum trajectories in the localized steady regime exhibit intermittent dynamics consisting of long-time sticking events near selected modes interrupted by jumps between them (Yusipov et al., 2016).

This body of work makes clear that localization is not merely a diagnostic of spectral confinement. It can reorganize phase structure, stabilize edge modes without a bulk many-body gap, and, under engineered dissipation, become a controllable property of the steady state rather than a purely Hamiltonian phenomenon.

4. Geometric, phase-space, and subspace localization

A distinct line of work studies localization as concentration relative to geometry rather than disorder. On $U$7, the highest-weight spherical harmonics

$U$8

produce a polar marginal

$U$9

which sharpens near $\mathrm{IPR}=\sum_i|\psi_i|^4,$0 as $\mathrm{IPR}=\sum_i|\psi_i|^4,$1 grows. The angular width scales as $\mathrm{IPR}=\sum_i|\psi_i|^4,$2, the full width at half maximum is

$\mathrm{IPR}=\sum_i|\psi_i|^4,$3

and the exact vertical spread is

$\mathrm{IPR}=\sum_i|\psi_i|^4,$4

The paper introduces the equatorial rigidity index

$\mathrm{IPR}=\sum_i|\psi_i|^4,$5

so the exact finite-$\mathrm{IPR}=\sum_i|\psi_i|^4,$6 value is proportional to the Wallis partial product, and the semiclassical limit $\mathrm{IPR}=\sum_i|\psi_i|^4,$7 recovers Wallis’s formula for $\mathrm{IPR}=\sum_i|\psi_i|^4,$8 (Ye et al., 29 Apr 2026).

Phase-space localization raises a different issue: the reference space may be unbounded. In the four-dimensional phase space of the Dicke model, Husimi-based Rényi occupations were defined on two bounded references: the finite atomic phase space $\mathrm{IPR}=\sum_i|\psi_i|^4,$9 and the classical energy shell $O(1)$0. The same eigenstate can have $O(1)$1 and $O(1)$2, because projection onto the atomic subspace integrates out bosonic localization while energy-shell restriction remains sensitive to the full phase-space distribution. The paper’s central conclusion is that in unbounded spaces the definition of maximal delocalization requires a bounded reference subspace, and different choices lead to contextual answers (Villaseñor et al., 2021).

Subspace localization abstracts this idea further. For a density matrix $O(1)$3 and a subspace $O(1)$4 with projector $O(1)$5, “Localization of quantum states within subspaces” defines a unique decomposition

$O(1)$6

with $O(1)$7 and $O(1)$8. The localization probability within $O(1)$9 is

$1/L$0

which is not the same as the overlap probability

$1/L$1

The inequality

$1/L$2

is strict in general, and equality holds when $1/L$3. In block form with respect to $1/L$4,

$1/L$5

so the localized weighted component is the Schur complement of $1/L$6 (Salcedo, 14 Jan 2026).

These geometric and algebraic formulations broaden the meaning of localization. The localized object need not be an exponentially confined eigenstate in real space; it may be a probability cloud concentrating on a classical ring, a Husimi distribution relative to a bounded reference shell, or the maximal state component fully supported in a prescribed subspace.

5. Relativistic, algebraic, and spacetime localization

In relativistic quantum theory, localization becomes a question about observables and spacetime regions. The review “Localisation in Quantum Field Theory” argues that Born localization of states is incompatible with relativity and causality and is replaced by symplectic localization of observables. In AQFT one assigns to each causally complete spacetime region $1/L$7 a von Neumann algebra $1/L$8 satisfying isotony, covariance, and locality; microcausality is expressed by

$1/L$9

for spacelike-separated regions. This notion underlies modular localization, the Bisognano–Wichmann property, the spin–statistics theorem, and the Unruh effect (Balachandran, 2016).

Recent comparative analyses sharpen this point. “Localization in Quantum Field Theory” examines three schemes—Newton–Wigner, AQFT, and the modal scheme—and emphasizes that only AQFT provides a fundamental concept of localization rooted in the axiomatic formulation. Newton–Wigner localization retains orthogonality of disjointly localized states and is useful mainly in the nonrelativistic regime, while the modal scheme based on positive-frequency Klein–Gordon solutions is incompatible with AQFT localization. The same review highlights that Reeh–Schlieder violates independence between state preparations and observable measurements in spacelike separated regions, but causality remains unviolated, a point explicitly compared to quantum teleportation (Falcone et al., 2023).

For inertial and accelerated observers, the distinction becomes sharper. “Localization in Quantum Field Theory for inertial and accelerated observers” studies Newton–Wigner, AQFT, and modal localization in Minkowski and Rindler frames and concludes that only the AQFT scheme obeys causality and physical invariance under diffeomorphisms. In the nonrelativistic limit of quantum fields in the Rindler frame, the AQFT and modal schemes converge and the Born notion of localization of states and observables emerges. The same paper finds that independence between preparation of states and measurements is not guaranteed when both experimenters are accelerated and the background state is different from Rindler vacuum, or when one of the two experimenters is inertial (Falcone et al., 2024).

A different proposal constructs a bilinear covariant spacetime operator

$$r_n=\frac{\min(\delta_n,\delta_{n+1})}{\max(\delta_n,\delta_{n+1})}$$0

together with a real Klein–Gordon wave function propagated by the Wheeler propagator rather than the Feynman propagator. In that framework, the operator is local in both $$r_n=\frac{\min(\delta_n,\delta_{n+1})}{\max(\delta_n,\delta_{n+1})}$$1- and $$r_n=\frac{\min(\delta_n,\delta_{n+1})}{\max(\delta_n,\delta_{n+1})}$$2-space, Hermitian with respect to the relativistic inner product, extends to higher powers, and fits into a “Weyl–Heisenberg–Poincaré” algebra; a perfectly localized relativistic probability density is exhibited for a massless wave packet (Westra, 2010).

Spacetime localization also appears in quantum cryptography. “How To Track Qubits Through Space and Time” introduces quantum localization as a stronger notion than position verification: there must be a specified, unclonable state at the verified spacetime point, and this state can be found nowhere else. The paper defines entanglement localization, state localization, trajectory verification, and functionality localization, all based on quantum anchor states that generalize coset states from unclonable cryptography. The security statements are formulated through local extractors acting on the quantum registers present in a small neighborhood of the claimed spacetime point (Bartusek et al., 29 May 2026).

A common source of confusion is to treat all of these relativistic schemes as variants of position measurement. The AQFT and cryptographic frameworks instead localize observables, algebras, or unclonable information in spacetime; they do not presuppose a Lorentz-covariant Born-style position operator.

6. Quantum-enhanced localization technologies

In sensing and information processing, “quantum localization” has acquired an operational meaning tied to estimating position with quantum resources. “QuERLoc: Towards Next-Generation Localization with Quantum-Enhanced Ranging” defines quantum localization as estimating a sensor’s position by leveraging quantum-enhanced ranging with entangled probe states. For an even-sized anchor set $$r_n=\frac{\min(\delta_n,\delta_{n+1})}{\max(\delta_n,\delta_{n+1})}$$3 with balanced signs $$r_n=\frac{\min(\delta_n,\delta_{n+1})}{\max(\delta_n,\delta_{n+1})}$$4, controlled evolution yields the phase

$$r_n=\frac{\min(\delta_n,\delta_{n+1})}{\max(\delta_n,\delta_{n+1})}$$5

Because $$r_n=\frac{\min(\delta_n,\delta_{n+1})}{\max(\delta_n,\delta_{n+1})}$$6, the quadratic $$r_n=\frac{\min(\delta_n,\delta_{n+1})}{\max(\delta_n,\delta_{n+1})}$$7 term cancels after expanding $$r_n=\frac{\min(\delta_n,\delta_{n+1})}{\max(\delta_n,\delta_{n+1})}$$8, and the localization problem becomes linear,

$$r_n=\frac{\min(\delta_n,\delta_{n+1})}{\max(\delta_n,\delta_{n+1})}$$9

With noisy measurements, the paper solves the weighted least-squares problem

$$M_z=\langle \psi|\hat S^z|\psi\rangle,\qquad \hat S^z=\sum_{j=1}^{L}\sigma_j^z,$$0

with closed-form solution

$$M_z=\langle \psi|\hat S^z|\psi\rangle,\qquad \hat S^z=\sum_{j=1}^{L}\sigma_j^z,$$1

The reported numerical results state a minimum reduction of $$M_z=\langle \psi|\hat S^z|\psi\rangle,\qquad \hat S^z=\sum_{j=1}^{L}\sigma_j^z,$$2 in RMSE and $$M_z=\langle \psi|\hat S^z|\psi\rangle,\qquad \hat S^z=\sum_{j=1}^{L}\sigma_j^z,$$3 in time consumption compared to baselines (He et al., 2024).

A related proposal considers TDoA. “Quantum Ranging Enhanced TDoA Localization” uses entangled probes to measure a linear combination of target–anchor distances in one shot rather than subtracting two separate ToA measurements. For a signed set of paths,

$$M_z=\langle \psi|\hat S^z|\psi\rangle,\qquad \hat S^z=\sum_{j=1}^{L}\sigma_j^z,$$4

and the corresponding localization estimator is based on

$$M_z=\langle \psi|\hat S^z|\psi\rangle,\qquad \hat S^z=\sum_{j=1}^{L}\sigma_j^z,$$5

The paper reports “over 50% gains on average” relative to classical TDoA in numerical simulations (He et al., 2024).

Quantum algorithms have also been proposed for robot localization. In “Robot localization aided by quantum algorithms,” 2D localization on a discretized costmap is mapped to unstructured search over

$$M_z=\langle \psi|\hat S^z|\psi\rangle,\qquad \hat S^z=\sum_{j=1}^{L}\sigma_j^z,$$6

candidate positions, with Grover iteration count

$$M_z=\langle \psi|\hat S^z|\psi\rangle,\qquad \hat S^z=\sum_{j=1}^{L}\sigma_j^z,$$7

Theoretical complexity becomes $$M_z=\langle \psi|\hat S^z|\psi\rangle,\qquad \hat S^z=\sum_{j=1}^{L}\sigma_j^z,$$8 oracle calls rather than classical $$M_z=\langle \psi|\hat S^z|\psi\rangle,\qquad \hat S^z=\sum_{j=1}^{L}\sigma_j^z,$$9 scans. Simulators show amplitude amplification of the correct pose, but current noisy hardware does not realize the practical advantage because error rates dominate larger circuits (Antero et al., 31 Jan 2025).

These technological usages are conceptually distinct from Anderson, MBL, or AQFT localization. Here the localized quantity is not a confined eigenstate or a local algebra; it is an unknown position inferred by entangled ranging, convexified inverse geometry, or quantum search. The shared term reflects the target of the inference, not a common microscopic mechanism.