Non-Uniform Entanglement

• Non-uniform entanglement is defined by the uneven distribution of quantum correlations arising from spatial, phase, and modal asymmetries.
• It manifests in diverse settings, from random state ensembles with specific concurrence benchmarks to many-body systems exhibiting local, GHZ-like, and W-state–like structures.
• Non-uniform couplings, mode mismatches, and spatial inhomogeneities critically impact observable metrics such as fidelity, metrological gain, and entanglement-spectrum statistics.

Non-uniform entanglement denotes situations in which entanglement is not adequately represented as a homogeneous resource fixed by a single spectrum, a uniform geometry, or a permanently chosen subsystem split. In contemporary work, the phrase appears in several distinct but related senses: non-flat distributions of entanglement over random-state ensembles; phase-dependent mixtures of local, GHZ-like, and W-state-like correlations in many-body systems; entanglement dynamics under spatially inhomogeneous Hamiltonians; collective and modal entanglement under non-uniform couplings or changing mode definitions; and entanglement sectors that remain hidden from local tomography but persist in holistic degrees of freedom (Biswas et al., 2021, Odavić et al., 2022, Wen et al., 2018, Hu et al., 2015, Baldijão et al., 18 Feb 2026). This suggests that the unifying content of the term is structural asymmetry: the relevant entanglement depends on where it is located, how it is accessed, which modes define the parties, or which operational constraints are imposed.

1. Distributional and structural senses of non-uniformity

In random-state studies, non-uniformity can refer literally to the distribution of entanglement values over an ensemble. For Haar-uniform pure two-qubit states, concurrence is not uniformly spread over $[0,1]$: the reported mean concurrence is approximately $0.589$ and the standard deviation approximately $0.230$. For Haar-random pure three-qubit states, the Jungnitsch–Moroder–Gühne monotone has mean about $0.35$ and standard deviation $0.068$. Disorder in state parameters drives an “inhibition of spread”: with Gaussian, uniform, or Cauchy–Lorentz perturbations, the standard deviation decreases, more disordered parameters strengthen the effect, and Cauchy–Lorentz disorder produces the strongest localization among the cases studied (Biswas et al., 2021).

In that setting, the relevant point is not that disorder makes entanglement “more uniform,” but the opposite: it compresses the entanglement histogram around intermediate values. Low-entanglement states tend to move upward and high-entanglement states downward toward the global mean of the clean Haar ensemble (Biswas et al., 2021). This use of non-uniformity is therefore distributional.

A different use is structural rather than statistical. In many-body, field-theoretic, and operational settings, non-uniform entanglement means that the entanglement content is distributed unevenly across space, phases, collective modes, momentum sectors, or operationally accessible subspaces. The same system can then exhibit entanglement that is local in one regime, GHZ-like in another, W-state-like in a third, or even invisible to all product measurements despite being globally present (Odavić et al., 2022, Baldijão et al., 18 Feb 2026).

2. Phase-dependent entanglement structure in many-body systems

A clear many-body realization occurs in the one-dimensional transverse-field Ising chain with frustrated boundary conditions,

$$H = J \sum_{l=1}^{N} \sigma_{l}^{x} \sigma_{l+1}^{x} - h \sum_{l=1}^{N} \sigma_{l}^{z},$$

with odd $N=2M+1$. Three macroscopic phases are distinguished: paramagnetic (PARA), ferromagnetic (FM), and topologically frustrated antiferromagnetic (AFM). The analysis combines Rényi-2 entropy,

$S_2(\rho_A)=-\log_2 \operatorname{Tr}(\rho_A^2),$

nearest-neighbor concurrence, and entanglement-spectrum spacing ratios. The level statistics are mostly Poissonian in every phase, with $\bar r_{\rm Poisson}=2\ln2-1\simeq0.386$, despite strongly different entanglement dynamics under a Metropolis-like cooling algorithm using local gates

$$U_l^{(k)}=\exp(i O_l^{(k)}\Delta t), \qquad \Delta t=\pi/10,$$

accepted with probability

$0.589$0

and averaged over $0.589$1 trajectories (Odavić et al., 2022).

The central result is that the entanglement spectrum alone is incomplete as a characterization tool. In the paramagnetic phase, the ground state is adiabatically connected to a fully factorized classical state, so its entanglement is mainly local and is substantially reduced by universal local gates. In the ferromagnetic phase, the ground state behaves as a locally deformed GHZ state; the cooling protocol removes local dressing but leaves a robust global component, producing a residual plateau near $0.589$2. In the frustrated AFM phase, the ground state is described as a locally deformed W-state-like superposition of single-kink states; its multipartite entanglement is reduced more effectively than in the FM case, but not eliminated, and the final baseline becomes increasingly robust with system size (Odavić et al., 2022).

This phase dependence is the concrete many-body content of non-uniform entanglement in this line of work. PARA is mostly local, FM contains local dressing plus robust GHZ-like nonlocal structure, and frustrated AFM contains local dressing plus robust W-state-like multipartite structure. A further notable result is that Wigner–Dyson entanglement-spectrum statistics can emerge after cooling even while the final state remains area-law-like rather than volume-law. The implication is that entanglement complexity depends on locality, multipartite structure, frustration, and gate accessibility, not only on spectral statistics (Odavić et al., 2022).

3. Spatial inhomogeneity and dynamical reorganization

In conformal field theory, non-uniform entanglement dynamics arises when the post-quench Hamiltonian is spatially inhomogeneous. For a quench from a uniform CFT on $0.589$3 with open boundary conditions to the sine-square deformed Hamiltonian

$0.589$4

the subsystem entropy of $0.589$5 is frozen at early times,

$0.589$6

up to a crossover time $0.589$7, which for $0.589$8 is

$0.589$9

At late times,

$0.230$0

independent of $0.230$1 and $0.230$2, and without revival. The position-dependent quasiparticle velocity

$0.230$3

vanishes near the boundaries, so the SSD CFT behaves effectively as an infinite system. Möbius deformation interpolates continuously between uniform and SSD dynamics, with oscillation period

$0.230$4

and for $0.230$5,

$0.230$6

(Wen et al., 2018).

A subsystem-based formulation of dynamical criticality sharpens the same theme. For a bipartition $0.230$7, with reduced density matrix

$0.230$8

the entanglement echo is

$0.230$9

where $0.35$0 is the entanglement ground state. Zeros of $0.35$1 define dynamical phase transitions in the subsystem. In non-uniform quenches, different spatial regions can undergo different transition classes at different times, because short-time dynamics is locally determined by the local quench history. One region can exhibit an entanglement-type transition while another exhibits a bulk-type Loschmidt transition under the same global protocol (Pöyhönen et al., 2021).

Spatial non-uniformity therefore affects entanglement in two distinct ways. It can reshape propagation velocities and recurrence structure, as in SSD and Möbius CFTs, or it can produce spatially resolved reorganization of the subsystem entanglement ground state, detectable through the entanglement echo and associated fluctuation observables (Wen et al., 2018, Pöyhönen et al., 2021).

4. Non-uniform couplings, fields, and mode definitions

For atomic ensembles under non-uniform atom–light coupling,

$0.35$2

the relevant collective degree of freedom is a weighted bright mode rather than the uniformly symmetric spin. In the nearly polarized regime, one introduces effective collective operators

$0.35$3

with

$0.35$4

The non-uniform ensemble is then equivalent to a uniformly coupled ensemble with reduced effective coupling $0.35$5 and reduced effective atom number $0.35$6. Entanglement is not destroyed; rather, the light selects a weighted collective mode, while orthogonal dark modes remain unobserved (Hu et al., 2015).

If state preparation and readout use different coupling profiles $0.35$7 and $0.35$8, the observable entanglement is controlled by the mode overlap

$0.35$9

For the first Dicke state, matched preparation and readout give $0.068$0. For standing-wave preparation and uniform readout, $0.068$1 and $0.068$2. When $0.068$3, the Wigner function becomes everywhere nonnegative. For spin-squeezed states, the metrological gain is bounded by

$0.068$4

and near-Heisenberg scaling is lost when mismatch exceeds the inverse atom-number scale (Hu et al., 2015).

A closely related inhomogeneity appears in an $0.068$5 spin-$0.068$6 pair under non-uniform transverse fields,

$0.068$7

Here the field difference

$0.068$8

is the controlling parameter. For $0.068$9, the entangled $$H = J \sum_{l=1}^{N} \sigma_{l}^{x} \sigma_{l+1}^{x} - h \sum_{l=1}^{N} \sigma_{l}^{z},$$0 eigenstates have concurrence

$$H = J \sum_{l=1}^{N} \sigma_{l}^{x} \sigma_{l+1}^{x} - h \sum_{l=1}^{N} \sigma_{l}^{z},$$1

and the entangled ground-state condition is

$$H = J \sum_{l=1}^{N} \sigma_{l}^{x} \sigma_{l+1}^{x} - h \sum_{l=1}^{N} \sigma_{l}^{z},$$2

At finite temperature, the thermal entanglement threshold depends only on $$H = J \sum_{l=1}^{N} \sigma_{l}^{x} \sigma_{l+1}^{x} - h \sum_{l=1}^{N} \sigma_{l}^{z},$$3, producing a separability stripe

$$H = J \sum_{l=1}^{N} \sigma_{l}^{x} \sigma_{l+1}^{x} - h \sum_{l=1}^{N} \sigma_{l}^{z},$$4

in the $$H = J \sum_{l=1}^{N} \sigma_{l}^{x} \sigma_{l+1}^{x} - h \sum_{l=1}^{N} \sigma_{l}^{z},$$5 plane (Rios et al., 2016).

Mode dependence gives a further refinement. Ordinary mode entanglement can disappear under a change of orthonormal mode basis, but mode-independent entanglement persists under every such transformation. The experimentally realized two-photon family

$$H = J \sum_{l=1}^{N} \sigma_{l}^{x} \sigma_{l+1}^{x} - h \sum_{l=1}^{N} \sigma_{l}^{z},$$6

was certified against the global separable fidelity bound

$$H = J \sum_{l=1}^{N} \sigma_{l}^{x} \sigma_{l+1}^{x} - h \sum_{l=1}^{N} \sigma_{l}^{z},$$7

Across $$H = J \sum_{l=1}^{N} \sigma_{l}^{x} \sigma_{l+1}^{x} - h \sum_{l=1}^{N} \sigma_{l}^{z},$$8, the reported average quantum-state fidelity was $$H = J \sum_{l=1}^{N} \sigma_{l}^{x} \sigma_{l+1}^{x} - h \sum_{l=1}^{N} \sigma_{l}^{z},$$9, with certification up to six standard deviations. This provides a limiting case in which non-uniformity under mode changes is eliminated by construction: the state remains entangled for all orthonormal mode bases (Held et al., 29 Jun 2026).

5. Motion, symmetry, and holistic degrees of freedom

Non-uniform motion can act as an entangling transformation. For quantum fields in moving cavities, Bogoliubov mode mixing generated by piecewise inertial and uniformly accelerated motion functions as a multipartite quantum gate. In bosonic cavities,

$N=2M+1$0

and genuine multipartite entanglement is detected whenever the first-order coefficient $N=2M+1$1. For repeated motion blocks satisfying

$N=2M+1$2

one obtains resonant enhancement,

$N=2M+1$3

within the perturbative regime $N=2M+1$4. Fermionic variants generate Dicke-type and $N=2M+1$5-type states rather than bosonic qutrit-like structures, reflecting Pauli constraints and monogamy (Friis et al., 2012).

A cavity with explicitly position-dependent coupling,

$N=2M+1$6

shows a different non-uniform effect. In a two-photon Tavis–Cummings-type model with dynamical Stark shift, atom–field entanglement is quantified by the purity $N=2M+1$7 of the reduced atomic state. Increasing the Stark shift decreases entanglement, i.e. raises purity, while weak-coupling chaotic motion drives the purity toward $N=2M+1$8 (Chotorlishvili et al., 2011).

Non-uniformity can also be algebraic. In $N=2M+1$9, the coproduct

$S_2(\rho_A)=-\log_2 \operatorname{Tr}(\rho_A^2),$0

is non-cocommutative for $S_2(\rho_A)=-\log_2 \operatorname{Tr}(\rho_A^2),$1. Although the single-qubit Hamiltonian remains $S_2(\rho_A)=-\log_2 \operatorname{Tr}(\rho_A^2),$2, the two-qubit coproduct Hamiltonian becomes intrinsically nonlocal,

$S_2(\rho_A)=-\log_2 \operatorname{Tr}(\rho_A^2),$3

The unitary $S_2(\rho_A)=-\log_2 \operatorname{Tr}(\rho_A^2),$4 has closed-form operator entanglement

$S_2(\rho_A)=-\log_2 \operatorname{Tr}(\rho_A^2),$5

with $S_2(\rho_A)=-\log_2 \operatorname{Tr}(\rho_A^2),$6 for $S_2(\rho_A)=-\log_2 \operatorname{Tr}(\rho_A^2),$7 and generically nonzero for $S_2(\rho_A)=-\log_2 \operatorname{Tr}(\rho_A^2),$8. For Haar-uniform product inputs, the entangling power is exactly

$S_2(\rho_A)=-\log_2 \operatorname{Tr}(\rho_A^2),$9

so the entangling capability is fully determined by the operator entanglement itself (Arzano et al., 31 Dec 2025).

Observer-dependent non-uniformity appears in accelerated quantum field theory. In a quantum twin-paradox setup with Unruh–DeWitt detectors, changes in the direction of acceleration imprint peaks and dips in transition rates and produce nontrivial time dependence in negativity and mutual information; the spacelike-to-timelike change in separation does not by itself determine the entanglement behavior (Hari et al., 11 Dec 2025). In a non-uniform Rindler construction with time-dependent acceleration,

$\bar r_{\rm Poisson}=2\ln2-1\simeq0.386$0

the perceived particle density becomes a time-dependent modulation of the standard Unruh spectrum, and the Minkowski vacuum is represented as a one-mode squeezed state rather than the usual two-mode squeezed Rindler form (Fernández et al., 29 Dec 2025).

6. Certification, operational boundaries, and related notions

Operational claims about entanglement can become unreliable when the underlying non-uniformity is ignored. In quantum teleportation with a non-uniform prior over qubit inputs, the appropriate classical benchmark depends on the prior itself rather than on the uniform-prior value $\bar r_{\rm Poisson}=2\ln2-1\simeq0.386$1. For a von Mises–Fisher prior

$\bar r_{\rm Poisson}=2\ln2-1\simeq0.386$2

Bayesian measure-and-prepare optimization yields an optimal projective measurement axis on the equator,

$\bar r_{\rm Poisson}=2\ln2-1\simeq0.386$3

and a mean-fidelity benchmark

$\bar r_{\rm Poisson}=2\ln2-1\simeq0.386$4

This reduces to $\bar r_{\rm Poisson}=2\ln2-1\simeq0.386$5 as $\bar r_{\rm Poisson}=2\ln2-1\simeq0.386$6, rises above $\bar r_{\rm Poisson}=2\ln2-1\simeq0.386$7 for concentrated priors, and approaches $\bar r_{\rm Poisson}=2\ln2-1\simeq0.386$8 as $\bar r_{\rm Poisson}=2\ln2-1\simeq0.386$9. For multi-qubit spin-coherent inputs, the uniform-prior benchmark $$U_l^{(k)}=\exp(i O_l^{(k)}\Delta t), \qquad \Delta t=\pi/10,$$0 is the minimum of the optimized non-uniform-prior fidelities, not the generic threshold. Entanglement is therefore demonstrated only when the observed fidelity exceeds the correct prior-dependent classical ceiling (Opatrný et al., 24 May 2026).

Measurement-device-independent entanglement witnesses exhibit a related sensitivity to non-uniformity in the trusted inputs. If the noisy input map is uniform and its adjoint preserves separable positive semidefinite operators, then separable states remain nonnegative under the witness functional. By contrast, non-uniform input-dependent noise and entangling input noise can produce false positives, including negative witness values for product shared states (Sen et al., 2020).

A deeper conceptual distinction appears when tomographic locality fails. In generalized probabilistic theories with holistic composite degrees of freedom,

$$U_l^{(k)}=\exp(i O_l^{(k)}\Delta t), \qquad \Delta t=\pi/10,$$1

entanglement decomposes into tomographically-local (TL) entanglement and tomographically-nonlocal (TNL) entanglement. TNL entanglement is useless for Bell nonlocality, steering, and teleportation, because product effects annihilate the holistic component; however, it is sufficient for dense coding and perfectly secure data hiding. Real quantum theory supplies explicit examples in which the holistic term $$U_l^{(k)}=\exp(i O_l^{(k)}\Delta t), \qquad \Delta t=\pi/10,$$2 carries TNL entanglement (Baldijão et al., 18 Feb 2026).

This operational boundary is important for distinguishing entanglement from broader nonlocal information-processing phenomena. Product-state ensembles can exhibit nonlocality without entanglement: they may be globally distinguishable but not optimally distinguishable by LOCC, and separable measurements can be locally unimplementable. Such behavior occurs not only in quantum theory but also in generalized probabilistic theories, yet it is not entanglement in the state itself (Bhattacharya et al., 2019).

Taken together, these developments show that non-uniform entanglement is not a single invariant object but a family of phenomena in which entanglement depends on structural asymmetries: phase, geometry, coupling profile, motion, mode definition, or the gap between global and locally tomographic descriptions. The recurring lesson is that uniform benchmarks, uniform partitions, and spectrum-only diagnostics can be insufficient precisely when the entanglement is non-uniform.