Hyperuniform Disorder Overview

• Hyperuniform disorder is defined by the suppression of infinite-wavelength fluctuations, resulting in crystal-like stability at large scales despite local disorder.
• It uses diagnostics such as the structure factor and spectral density to classify fluctuation scaling into distinct classes, unifying point patterns and two-phase media.
• This concept underpins practical advances in photonics, acoustics, and material science by enabling control over scattering, transport, and rigidity.

Hyperuniform disorder denotes disordered many-particle or heterogeneous states in which infinite-wavelength fluctuations are anomalously suppressed: the system remains amorphous, statistically isotropic, and free of Bragg peaks at short and intermediate scales, yet behaves like a crystal in the limit of large length scales. In reciprocal space, the defining signature is the vanishing of the structure factor or spectral density as the wavevector tends to zero; in direct space, the corresponding number or volume-fraction fluctuations grow more slowly than the observation-window volume. The concept unifies point patterns, two-phase media, random fields, and nonequilibrium steady states, and it has become a central framework for jammed matter, driven suspensions, photonic media, acoustic arrays, defected solids, and inverse-designed materials (Torquato, 2018, Torquato, 2016).

1. Definition, classification, and scope

For statistically homogeneous point configurations of number density $\rho$, the standard reciprocal-space diagnostic is

$S(\mathbf{k}) = 1+\rho\,\widetilde h(\mathbf{k}),$

with hyperuniformity defined by

$\lim_{|\mathbf{k}|\to 0} S(\mathbf{k}) = 0.$

The equivalent direct-space statement is that the number variance $\sigma_N^2(R)$ inside a spherical window of radius $R$ grows more slowly than $R^d$. When the small-$k$ behavior obeys $S(k)\sim |k|^\alpha$ with $\alpha>0$, the large-$R$ asymptotics fall into three classes: Class I for $S(\mathbf{k}) = 1+\rho\,\widetilde h(\mathbf{k}),$0, with $S(\mathbf{k}) = 1+\rho\,\widetilde h(\mathbf{k}),$1; Class II for $S(\mathbf{k}) = 1+\rho\,\widetilde h(\mathbf{k}),$2, with $S(\mathbf{k}) = 1+\rho\,\widetilde h(\mathbf{k}),$3; and Class III for $S(\mathbf{k}) = 1+\rho\,\widetilde h(\mathbf{k}),$4, with $S(\mathbf{k}) = 1+\rho\,\widetilde h(\mathbf{k}),$5 (Torquato, 2018, Chen et al., 2022).

For two-phase heterogeneous media, the relevant object is not necessarily the point-center structure factor but the spectral density

$S(\mathbf{k}) = 1+\rho\,\widetilde h(\mathbf{k}),$6

where $S(\mathbf{k}) = 1+\rho\,\widetilde h(\mathbf{k}),$7 is the autocovariance of the phase indicator. Hyperuniformity then means

$S(\mathbf{k}) = 1+\rho\,\widetilde h(\mathbf{k}),$8

equivalently, the local volume-fraction variance decays faster than $S(\mathbf{k}) = 1+\rho\,\widetilde h(\mathbf{k}),$9. This formulation is essential for packings and cellular materials in which the geometry of the particle and void phases, rather than only the point centers, controls the large-scale fluctuations (Torquato, 2016, Zachary et al., 2010).

The concept also admits broader generalizations. Hyperuniformity has been extended to interfacial-area fluctuations in two-phase media, random scalar fields, divergence-free random vector fields, and statistically anisotropic systems. In anisotropic settings, the correct formulation can be directional: the relevant spectral function may vanish only along selected approaches to the origin in Fourier space, which leads to the notion of directional hyperuniformity (Torquato, 2016).

A further distinction separates static from dynamic hyperuniformity. Static hyperuniformity refers to time-independent or time-averaged structures such as jammed packings. Dynamic hyperuniformity refers to nonequilibrium fluctuating steady states in which each instantaneous configuration remains disordered, but the dynamics continually filters out low-$\lim_{|\mathbf{k}|\to 0} S(\mathbf{k}) = 0.$0 density modes; in such cases equal-time structure factors and, in some systems, spatiotemporal spectra suppress low-$\lim_{|\mathbf{k}|\to 0} S(\mathbf{k}) = 0.$1, low-$\lim_{|\mathbf{k}|\to 0} S(\mathbf{k}) = 0.$2 fluctuations (Castillo et al., 2018, Lei et al., 2024).

2. Diagnostics and observables

The most common diagnostic remains the small-wavevector structure factor. For point patterns it tests $\lim_{|\mathbf{k}|\to 0} S(\mathbf{k}) = 0.$3 directly; for packings and composites the analogous quantity is the spectral density $\lim_{|\mathbf{k}|\to 0} S(\mathbf{k}) = 0.$4. In maximally random jammed packings, for example, local-volume-fraction fluctuations provide a more faithful descriptor than point-center statistics, because polydisperse or nonspherical packings can have $\lim_{|\mathbf{k}|\to 0} S(\mathbf{k}) = 0.$5 for the centers while still satisfying $\lim_{|\mathbf{k}|\to 0} S(\mathbf{k}) = 0.$6 as two-phase media (Zachary et al., 2010).

Real-space observables remain important, but they are more delicate. Number variance and local-volume-fraction variance encode the same long-wavelength physics asymptotically, yet finite-size effects can obscure the expected scaling. In the vibrated granular layer near the liquid-to-solid critical point, the structure factor proved more robust than particle-number variance: boundary effects and experimental drift affect $\lim_{|\mathbf{k}|\to 0} S(\mathbf{k}) = 0.$7 and $\lim_{|\mathbf{k}|\to 0} S(\mathbf{k}) = 0.$8 equally and thus cancel in $\lim_{|\mathbf{k}|\to 0} S(\mathbf{k}) = 0.$9, whereas sublinear number-variance scaling required very large systems to resolve (Castillo et al., 2018).

Several operational metrics have been introduced for near-hyperuniform or experimentally imperfect systems. A practical metric for nearly stealthy behavior is

$\sigma_N^2(R)$0

with $\sigma_N^2(R)$1 indicating nearly stealthy hyperuniformity; this was used for self-organized TiO$\sigma_N^2(R)$2 nanodisk arrays that reached $\sigma_N^2(R)$3 and $\sigma_N^2(R)$4 (Piechulla et al., 2021). Another real-space framework is Hyperuniformity Disorder Length Spectroscopy. It defines a hyperuniformity disorder length $\sigma_N^2(R)$5 by treating fluctuations as if only particles within a shell of thickness $\sigma_N^2(R)$6 around the boundary of a measuring window contribute. For hypercubic windows,

$\sigma_N^2(R)$7

where $\sigma_N^2(R)$8 is the variance ratio relative to a random pattern. Random patterns have $\sigma_N^2(R)$9, whereas strongly hyperuniform patterns satisfy $R$0 for large $R$1 (Chieco et al., 2017, Durian, 2017).

Stealthy hyperuniformity constitutes a stricter subtype. Here one enforces

$R$2

or, in finite constructions, for all reciprocal vectors inside a cutoff disk $R$3. This creates an exclusion region around the origin of Fourier space and has direct consequences for scattering, band-gap formation, and array-factor suppression in photonic and acoustic implementations (Torquato et al., 2018, Tang et al., 2023).

3. Formation mechanisms and constructive routes

One of the best-established mechanisms is jamming-controlled void regularization. In strictly jammed, saturated maximally random jammed packings, the void space is fragmented into small pockets determined by the contact network. Because arbitrarily large pores are forbidden, the spectral density obeys $R$4 with a vanishing intercept, and the real-space two-point correlations exhibit quasi-long-range tails of the form $R$5 (Zachary et al., 2010). Slight unjamming reconnects the void network, produces larger irregular pores, and restores $R$6.

A second mechanism is critical slowing and spatial blocking in driven nonequilibrium matter. In a vertically vibrated quasi-two-dimensional granular layer, patches of square-crystalline order appear below the continuous liquid-to-solid transition. Their characteristic size $R$7 and lifetime $R$8 diverge according to an Ornstein-Zernike form, and within these high-order patches particles become effectively caged and subdiffusive. Density modes at smaller $R$9 then encounter more such high-friction islands and relax more slowly. Experimentally and in the associated continuum model, the structure factor takes the form

$R^d$0

with $R^d$1 in two dimensions and $R^d$2 at criticality, yielding dynamic hyperuniformity (Castillo et al., 2018).

The continuum formulation of that mechanism couples a conserved density field $R^d$3 to a fluctuating scalar order or friction field $R^d$4 through

$R^d$5

together with a critical stochastic evolution for $R^d$6. As $R^d$7, the $R^d$8 fluctuations become scale-free, creating solid-like regions in which $R^d$9 and blocking long-wavelength density relaxation (Castillo et al., 2018).

A third mechanism is topological transformation of an ordered hyperuniform parent. In defected triangular-lattice inherent structures built from bound dislocations, free dislocations, and disclinations, the large-scale density fluctuations are dominated by elastic displacement fields produced by individual defects. When the defect-induced displacement kernel $k$0 satisfies $k$1 and $k$2, the small-$k$3 structure factor retains

$k$4

so the disordered inherent structure remains Class I hyperuniform even after substantial loss of translational and bond-orientational order (Chen et al., 2021).

Several constructive algorithms impose hyperuniformity directly. The tessellation-based procedure partitions space into bounded cells and inserts particles so that the local-cell packing fraction equals the global packing fraction in every cell. Under the bounded-cell condition, only boundary-intersecting cells contribute to large-window fluctuations, which yields Class I behavior in direct space and small-$k$5 spectral densities scaling as $k$6 or $k$7, depending on the relation between particle centroids and cell centers (Kim et al., 2019, Kim et al., 2019). A different route uses periodic Voronoi tessellations to iteratively rescale particle volumes so that each new particle volume equals the Voronoi-cell volume, thereby enforcing hyperuniform number-variance scaling while interleaving mechanical relaxation to recover jammed soft-sphere packings (Dale et al., 2022). Stealthy point patterns can also be generated by minimizing the collective-coordinate potential

$k$8

which drives $k$9 to zero inside the prescribed cutoff region (Tang et al., 2023).

4. Principal realizations across physical systems

Hyperuniform disorder was first established as a universal trait of maximally random jammed hard-particle packings, including monodisperse and polydisperse, spherical and nonspherical cases, when viewed as two-phase media. In these systems the small-$S(k)\sim |k|^\alpha$0 linear spectral density, the associated $S(k)\sim |k|^\alpha$1 quasi-long-range correlations, and the constrained void statistics jointly characterize the state (Zachary et al., 2010).

Granular and absorbing-state systems provide canonical nonequilibrium realizations. The vibrated granular layer at the critical point of the liquid-to-solid transition displays dynamic hyperuniformity through a critical friction-field mechanism (Castillo et al., 2018). More broadly, critical hyperuniformity appears in absorbing-state transitions such as random organization, the Manna model, and the conserved lattice gas. The review of nonequilibrium dynamic hyperuniform states reports $S(k)\sim |k|^\alpha$2 in $S(k)\sim |k|^\alpha$3 and $S(k)\sim |k|^\alpha$4 in $S(k)\sim |k|^\alpha$5 for these critical absorbing states, while non-equilibrium active fluids, chiral spinners, and related driven-dissipative systems can exhibit $S(k)\sim |k|^\alpha$6 (Lei et al., 2024).

Phase-separating media furnish a two-phase counterpart. Model-B and Model-H spinodal decomposition under critical quenches produce concentration-field spectral densities $S(k)\sim |k|^\alpha$7 at small $S(k)\sim |k|^\alpha$8, and active generalizations such as ECH, AMB, and AMB+ preserve the same small-$S(k)\sim |k|^\alpha$9 scaling at the level of two-point statistics (Lei et al., 2024). This places hyperuniform disorder within coarsening dynamics as well as in particle-based steady states.

Solid-state materials now provide an extensive family of disordered hyperuniform realizations. Reported examples include amorphous graphene, amorphous 2D silica, defected transition metal dichalcogenides, defected pentagonal 2D materials, amorphous carbon nanotubes, and medium/high-entropy alloys (Chen et al., 2022). In the review of solid-state materials, amorphous graphene obtained from experimental ADF-STEM images has $\alpha>0$0 and $\alpha>0$1, amorphous 2D silica has scalar-field spectral density $\alpha>0$2 with $\alpha>0$3, and equimolar Si-Ge-Sn alloys can be inverse-designed so that partial structure factors satisfy multihyperuniform targets (Chen et al., 2022).

Dimensional reduction does not eliminate the phenomenon. In quasi-one-dimensional carbon nanotubes with randomly distributed Stone-Wales defects, a generalized weighted structure factor

$\alpha>0$4

shows that single-walled and multi-walled amorphous nanotubes remain Class I hyperuniform for all examined diameters, chiralities, rolling axes, wall numbers, and defect fractions (Chen et al., 2022). The same review notes that projection preserves hyperuniformity for these isotropic parent structures, while anisotropic patterns require greater care.

5. Functional consequences in optics, acoustics, transport, and mechanics

In photonics, disordered hyperuniform dielectric networks can combine isotropy with photonic functionality usually associated with crystals. Two-dimensional hyperuniform cellular networks possess complete photonic band gaps comparable in size to photonic crystals while maintaining statistical isotropy, which enables waveguide geometries not possible with photonic crystals (Torquato et al., 2018). Hyperuniform disorder is also relevant to band-edge rare states: for a Gaussian random potential with $\alpha>0$5, the Lifshitz-tail density of states near a quadratic band edge obeys

$\alpha>0$6

so hyperuniform correlations modify the rate at which disorder fills a nominal band gap (Karcher et al., 2024).

Real photonic structures are not lossless, and intrinsic non-Hermiticity alters the hyperuniform design rules. In photonic crystal slabs with a complex effective mass $\alpha>0$7, the disorder-induced scattering loss no longer follows a pure $\alpha>0$8 law. Instead,

$\alpha>0$9

with a finite offset

$R$0

and $R$1 (Zhang et al., 4 Mar 2026). Thus a plausible implication is that hyperuniformity can still reduce added scattering, but radiative loss imposes a nonzero floor in realistic slab platforms.

In acoustics, hyperuniform disorder has been implemented in parametric loudspeaker arrays. A hyperuniform disordered array suppresses grating lobes and maintains a minimal radiation region around the main lobe for primary waves; the same structure benefits the secondary frequency wave in canceling grating lobes regardless of the primary frequencies (Tang et al., 2023). In the reported 200-element array, the measured $R$2 beamwidth remained $R$3, the exclusion zone satisfied $R$4 at $R$5, the peak sidelobe level was $R$6 at $R$7 versus $R$8 for the periodic array, and no grating lobes appeared even under beam steering to $R$9 (Tang et al., 2023).

In nanophotonics, nearly hyperuniform TiO$S(\mathbf{k}) = 1+\rho\,\widetilde h(\mathbf{k}),$00 nanodisk arrays demonstrate how structure factor and form factor can simultaneously shape scattering. Relative to uncorrelated arrays, nearly hyperuniform arrays suppress small-angle scattering, produce ring-shaped angle-resolved scattering with maxima at finite angles, and maintain broadband confinement of the scattering ring over $S(\mathbf{k}) = 1+\rho\,\widetilde h(\mathbf{k}),$01 for wavelengths $S(\mathbf{k}) = 1+\rho\,\widetilde h(\mathbf{k}),$02–$S(\mathbf{k}) = 1+\rho\,\widetilde h(\mathbf{k}),$03 (Piechulla et al., 2021). Mode-selective azimuthal anisotropy was amplified up to a $S(\mathbf{k}) = 1+\rho\,\widetilde h(\mathbf{k}),$04 intensity contrast, and simulations predicted $S(\mathbf{k}) = 1+\rho\,\widetilde h(\mathbf{k}),$05 with $S(\mathbf{k}) = 1+\rho\,\widetilde h(\mathbf{k}),$06 in the forward hemisphere for one parameter set (Piechulla et al., 2021).

Transport and elasticity in hyperuniform cellular materials are likewise exceptional. For low-density two-dimensional networks derived from stealthy hyperuniform parent patterns, effective conductivities and elastic moduli attain or nearly attain homogenization-theory bounds. In particular, stealthy $S(\mathbf{k}) = 1+\rho\,\widetilde h(\mathbf{k}),$07 Voronoi, Delaunay, and Delaunay-centroidal networks give $S(\mathbf{k}) = 1+\rho\,\widetilde h(\mathbf{k}),$08, $S(\mathbf{k}) = 1+\rho\,\widetilde h(\mathbf{k}),$09, and $S(\mathbf{k}) = 1+\rho\,\widetilde h(\mathbf{k}),$10, respectively, with tortuosities near unity, placing them within about $S(\mathbf{k}) = 1+\rho\,\widetilde h(\mathbf{k}),$11 of the optimal dilute-limit value $S(\mathbf{k}) = 1+\rho\,\widetilde h(\mathbf{k}),$12 (Torquato et al., 2018).

Hyperuniformity can also restructure vibrational spectra. Hyperuniform jammed soft-sphere packings generated by Voronoi-based rescaling exhibit a true low-frequency phononic gap, followed by an isolated narrow band of highly collective “sloshing” modes and then the usual higher-frequency plateau (Dale et al., 2022). For $S(\mathbf{k}) = 1+\rho\,\widetilde h(\mathbf{k}),$13 at packing fraction $S(\mathbf{k}) = 1+\rho\,\widetilde h(\mathbf{k}),$14, the gap extends up to $S(\mathbf{k}) = 1+\rho\,\widetilde h(\mathbf{k}),$15, and the sloshing band lies roughly in $S(\mathbf{k}) = 1+\rho\,\widetilde h(\mathbf{k}),$16 (Dale et al., 2022). In one-dimensional tight-binding chains, enhanced hyperuniformity can even induce delocalization: strong hyperuniform disorder yields a sharp delocalization transition, mobility edges, and ballistic transport in a finite energy window, in contrast to the conventional expectation that all 1D disordered states localize (Jeon et al., 27 Sep 2025).

6. Conceptual issues, measurement caveats, and open problems

A persistent source of confusion is the distinction between point-pattern and two-phase criteria. For packings with size or shape dispersity, the point-center structure factor can fail to vanish at the origin even when local-volume-fraction fluctuations are hyperuniform. The physically relevant criterion in such cases is $S(\mathbf{k}) = 1+\rho\,\widetilde h(\mathbf{k}),$17, not necessarily $S(\mathbf{k}) = 1+\rho\,\widetilde h(\mathbf{k}),$18 for the centers (Zachary et al., 2010). This is why void statistics, pore-size bounds, and phase geometry are central in heterogeneous materials.

A second issue concerns observability. Number variance is asymptotically equivalent to the small-$S(\mathbf{k}) = 1+\rho\,\widetilde h(\mathbf{k}),$19 structure factor, but it is often less reliable in finite experiments or simulations. The granular-layer study found that structure-factor measurements reveal the onset of hyperuniformity more clearly than particle-number variance, while Hyperuniformity Disorder Length Spectroscopy offers a complementary real-space spectrum $S(\mathbf{k}) = 1+\rho\,\widetilde h(\mathbf{k}),$20 that distinguishes random, weakly hyperuniform, and strongly hyperuniform arrangements (Castillo et al., 2018, Chieco et al., 2017).

Thermal and vibrational fluctuations can also mask the underlying order. In defect-generated inherent structures, hyperuniformity is preserved by localized static displacement fields, but thermal vibrations introduce long-ranged displacement correlations that give $S(\mathbf{k}) = 1+\rho\,\widetilde h(\mathbf{k}),$21 a finite intercept. The consequence is that one must decouple positional degrees of freedom from vibrational degrees of freedom, or quench to the inherent structure, to reveal the hidden disordered hyperuniformity (Chen et al., 2021).

Projection and anisotropy require additional care. The quasi-1D nanotube analysis shows that projection of an isotropic hyperuniform parent can preserve hyperuniformity, but an illustrative anisotropic example demonstrates that different projection directions can yield either a 1D Poisson pattern or a 1D lattice (Chen et al., 2022). More generally, the generalized theory of directional hyperuniformity shows that for anisotropic point patterns, two-phase media, or divergence-free vector fields, the vanishing of the spectral function may depend on the direction of approach to the origin in Fourier space (Torquato, 2016).

Current work suggests several unresolved directions. Non-equilibrium systems reveal multiple dynamical routes to hyperuniformity, but their universality classes depend sensitively on conservation laws, noise symmetries, and criticality (Lei et al., 2024). Solid-state materials indicate that disordered hyperuniform states can be energetically favored and electronically distinct from conventional disorder models, but the structural conditions that stabilize them remain system-specific (Chen et al., 2022). In photonic slabs, intrinsic non-Hermiticity implies that increasing the hyperuniformity exponent $S(\mathbf{k}) = 1+\rho\,\widetilde h(\mathbf{k}),$22 cannot eliminate the finite scattering floor $S(\mathbf{k}) = 1+\rho\,\widetilde h(\mathbf{k}),$23 when $S(\mathbf{k}) = 1+\rho\,\widetilde h(\mathbf{k}),$24 (Zhang et al., 4 Mar 2026). Taken together, these results suggest that hyperuniform disorder is less a single material class than a structural principle: liquid-like disorder at short range, crystal-like suppression of fluctuations at long range, and highly system-dependent consequences for waves, transport, rigidity, and dynamics.