Correlated Disorder in Condensed Matter

• Correlated Disorder is a phenomenon where disorder exhibits nontrivial, structured correlations across space, layers, or channels rather than independent random fluctuations.
• It modifies phase transitions, synchronization, and localization by renormalizing effective couplings and altering the interplay of disorder strength and correlation length.
• This disorder can stabilize new coherent or topological regimes, impacting systems from laser arrays to quantum lattices and influencing measurable observables.

Correlated disorder denotes disorder whose realizations are not independent, but instead exhibit nontrivial statistical structure across space, layers, fields, sites, or disorder channels. In the literature, this includes spatially correlated random potentials with finite correlation length, scale-free disorder with power-law correlations, quasi-periodic disorder with random phase, dimer-type short-range correlations, anti-symmetric correlations between coupled subsystems, and deterministic correlations between distinct random variables such as oscillator frequencies and couplings. Across these settings, correlated disorder is not a perturbative refinement of uncorrelated randomness: it can suppress disorder-induced phases, stabilize new coherent or topological regimes, alter universality classes and critical points, reshape localization and transport, and imprint mesoscopic or layer-to-layer memory into nominally disordered materials (Hong et al., 2016).

1. Statistical definitions and canonical forms

A common starting point is the contrast with uncorrelated disorder, where correlations are short-ranged or delta-like. In the Anderson model, uncorrelated on-site energies satisfy

$$C(\ell)=\langle \varepsilon_{\mathbf{i}}\varepsilon_{\mathbf{i}+\ell}\rangle=\frac{W^2}{12}\delta_{\ell,0},$$

whereas scale-free correlated disorder is defined by

$C(\ell)\propto |\ell|^{-\alpha},$

with smaller $\alpha$ corresponding to longer-range correlations (Croy et al., 2011). In laser arrays, quenched detunings are assigned a Gaussian spatial correlator,

$$\langle \Omega_{ij}\Omega_{i'j'}\rangle=(\Omega_{RMS})^2 e^{-[(i-i')^2+(j-j')^2]/\xi^2},$$

so that the correlation length $\xi$ becomes an experimentally tunable parameter (Pando et al., 2023). In topological and Floquet settings, correlated random potentials are often modeled as

$$\langle V(\mathbf{r})V(\mathbf{r}')\rangle=W^2 \exp\!\left[-\frac{|\mathbf{r}-\mathbf{r}'|^2}{2\sigma^2}\right],$$

with $\sigma$ controlling smoothness in space (Zheng et al., 2023).

Correlations also arise in forms that are not purely spatial. In mean-field oscillator models, the sign of the coupling is deterministically correlated with the oscillator’s natural frequency, so that attractive and repulsive oscillators are arranged asymmetrically or symmetrically across the frequency spectrum rather than independently (Hong et al., 2016). In many-body localization studies, correlated random fields are drawn from a multivariate Gaussian distribution with covariance

$C_{ij}=h^2[\delta_{ij}+\alpha(1-\delta_{ij})],$

where $\alpha=0$ yields independent disorder and $\alpha=1$ yields maximally correlated fields within a sample (Samanta et al., 2021). In quasi-periodic settings, correlated disorder may be generated deterministically through a potential of the form

$C(\ell)\propto |\ell|^{-\alpha},$0

so that the orientation of the quasi-periodicity becomes physically consequential (Okugawa et al., 2022).

Correlation structure	Representative form	Representative context
Finite-range Gaussian	$C(\ell)\propto |\ell|^{-\alpha},$1 or $C(\ell)\propto |\ell|^{-\alpha},$2	Laser arrays; Floquet and topological disorder (Pando et al., 2023, Zheng et al., 2023)
Scale-free power law	$C(\ell)\propto |\ell|^{-\alpha},$3	Anderson metal-insulator transition (Croy et al., 2011)
Cross-variable deterministic correlation	coupling sign tied to $C(\ell)\propto |\ell|^{-\alpha},$4	Phase oscillators (Hong et al., 2016)
Multivariate correlated fields	$C(\ell)\propto |\ell|^{-\alpha},$5	MBL–ergodic transition (Samanta et al., 2021)
Quasi-periodic correlated potential	$C(\ell)\propto |\ell|^{-\alpha},$6	Magnetically doped TI thin films (Okugawa et al., 2022)

These constructions show that correlated disorder is best understood as a family of ensembles rather than a single model class. A plausible implication is that “correlation length” alone is often insufficient; the sign structure, anisotropy, inter-channel correlations, and spectral content of the disorder can be equally decisive.

2. Observables and theoretical diagnostics

Because correlated disorder modifies both geometry and effective couplings, its signatures are distributed across several classes of observables. In synchronization problems, the standard Kuramoto order parameter

$C(\ell)\propto |\ell|^{-\alpha},$7

measures global phase coherence, while a weighted order parameter

$C(\ell)\propto |\ell|^{-\alpha},$8

tracks the effective mean field generated by attractive and repulsive oscillators (Hong et al., 2016). In coupled laser arrays, synchronization is quantified by the far-field inverse participation ratio, interpreted as the coherence area or effective number $C(\ell)\propto |\ell|^{-\alpha},$9 of synchronized lasers (Pando et al., 2023).

Localization and transport are commonly characterized by participation-based measures. In split-ring-resonator arrays, the participation ratio

$\alpha$0

distinguishes localized from extended eigenmodes, while the mean square displacement

$\alpha$1

and its long-time scaling $\alpha$2 resolve localized, diffusive, and super-diffusive regimes (Molina, 2020). In disordered quantum lattices and Anderson-type problems, inverse participation ratios, transfer-matrix localization lengths, and finite-size scaling of $\alpha$3 are standard tools (Croy et al., 2011, Liu, 2020).

Topological diagnostics depend on symmetry class and drive protocol. In disordered Floquet systems, the disorder-averaged Bott index

$\alpha$4

is used to identify transitions among Floquet phases (Zheng et al., 2023). In non-Hermitian chains, the real-space winding number

$\alpha$5

captures the point-gap topology associated with the non-Hermitian skin effect (Jin et al., 2023). In topological insulators, conductance plateaus, disorder-averaged $\alpha$6, and renormalized masses extracted from self-energy decompositions are central to identifying disorder-driven phase boundaries (Girschik et al., 2012).

Correlated disorder can also be diagnosed structurally. In van der Waals heterostructures, layer-to-layer or angular memory is quantified by a real-space statistical correlation function,

$\alpha$7

whose non-exponential decay signals long-range correlation in orientation disorder (Laanait et al., 2016). More broadly, the literature repeatedly combines numerical diagonalization, self-consistent Born approximation, effective-medium theory, transfer-matrix calculations, exact diagonalization, and X-ray diffuse or reciprocal-space imaging to separate the effects of disorder amplitude from those of disorder correlation.

3. Correlated disorder as a source of coherence and synchronization

In mean-field oscillator systems, correlated disorder can generate collective order precisely where uncorrelated disorder forbids it. For

$\alpha$8

with $\alpha$9 and a Lorentzian frequency distribution, uncorrelated disorder with equal attractive and repulsive populations yields no phase coherence. When coupling sign is deterministically correlated with frequency, two distinct regimes appear. For symmetrically correlated disorder, oscillators near $$\langle \Omega_{ij}\Omega_{i'j'}\rangle=(\Omega_{RMS})^2 e^{-[(i-i')^2+(j-j')^2]/\xi^2},$$0 are attractive and tail oscillators are repulsive, and phase coherence emerges below

$$\langle \Omega_{ij}\Omega_{i'j'}\rangle=(\Omega_{RMS})^2 e^{-[(i-i')^2+(j-j')^2]/\xi^2},$$1

For asymmetrically correlated disorder, the system develops coherent traveling waves with nonzero rotating order parameters below

$$\langle \Omega_{ij}\Omega_{i'j'}\rangle=(\Omega_{RMS})^2 e^{-[(i-i')^2+(j-j')^2]/\xi^2},$$2

with $$\langle \Omega_{ij}\Omega_{i'j'}\rangle=(\Omega_{RMS})^2 e^{-[(i-i')^2+(j-j')^2]/\xi^2},$$3 near threshold (Hong et al., 2016).

The same theme appears experimentally in laser arrays, but with a different outcome: correlations need not promote coherence monotonically. In a $$\langle \Omega_{ij}\Omega_{i'j'}\rangle=(\Omega_{RMS})^2 e^{-[(i-i')^2+(j-j')^2]/\xi^2},$$4 array of 400 coupled lasers with tunable Talbot coupling $$\langle \Omega_{ij}\Omega_{i'j'}\rangle=(\Omega_{RMS})^2 e^{-[(i-i')^2+(j-j')^2]/\xi^2},$$5, quenched frequency disorder of strength $$\langle \Omega_{ij}\Omega_{i'j'}\rangle=(\Omega_{RMS})^2 e^{-[(i-i')^2+(j-j')^2]/\xi^2},$$6, and spatial correlation length $$\langle \Omega_{ij}\Omega_{i'j'}\rangle=(\Omega_{RMS})^2 e^{-[(i-i')^2+(j-j')^2]/\xi^2},$$7, the effect of correlated disorder depends on the comparison between $$\langle \Omega_{ij}\Omega_{i'j'}\rangle=(\Omega_{RMS})^2 e^{-[(i-i')^2+(j-j')^2]/\xi^2},$$8 and the synchronized cluster size $$\langle \Omega_{ij}\Omega_{i'j'}\rangle=(\Omega_{RMS})^2 e^{-[(i-i')^2+(j-j')^2]/\xi^2},$$9. When $\xi$0, the synchronized cluster size scales as

$\xi$1

so increasing $\xi$2 suppresses synchronization. When $\xi$3,

$\xi$4

so increasing $\xi$5 aids synchronization. At intermediate $\xi$6, the dependence is non-monotonic, with a minimum in synchronization for weak disorder (Pando et al., 2023).

These results establish that correlated disorder is not reducible to “smoother disorder.” In oscillator models, correlation between distinct disorder channels selects which degrees of freedom can lock. In laser arrays, spatial correlations restructure the effective random-walk excursions that delimit phase-locked domains. This suggests that correlation architecture can act either as a coherence filter or as a coherence frustrator, depending on whether it isolates synchronizable degrees of freedom or coherently couples detuning across regions comparable to the cluster size.

4. Topology, localization, and transport under correlated disorder

A major theme in condensed-matter applications is that correlated disorder can either destroy or favor topological phases, depending on how it renormalizes masses, hoppings, and bulk-state connectivity. In two-dimensional topological insulators modeled on HgTe/CdTe quantum wells, uncorrelated disorder can induce the topological Anderson insulator, but spatially correlated disorder can entirely suppress that phase. The conductance plateau narrows or disappears, and a quantum percolation transition occurs as bulk states percolate along smooth disorder contours and connect opposite edges. A generalized self-consistent Born approximation incorporates the Fourier transform $\xi$7 of the correlated disorder and reproduces the numerically observed phase boundaries (Girschik et al., 2012).

In driven systems, the direction of the effect can reverse. In a two-dimensional Floquet model with honeycomb geometry, both uncorrelated and correlated disorder drive transitions from Floquet topological insulator phases to the anomalous Floquet topological insulator, and correlated disorder does so at smaller disorder strengths. For point-like gap closings, self-consistent Born analysis captures the renormalized masses and hoppings; for ring-shaped gaps, the Born approximation fails because the transition is controlled by strong mixing of near-degenerate ring states and Berry-curvature restructuring rather than simple parameter renormalization (Zheng et al., 2023).

Correlated disorder also reshapes localization without necessarily changing density-of-states physics dramatically. In the three-dimensional Anderson model with scale-free correlations $\xi$8, increasing long-range correlations increases the localization length, expands the metallic region, and shifts the mobility edge. At $\xi$9, the critical disorder rises from

$$\langle V(\mathbf{r})V(\mathbf{r}')\rangle=W^2 \exp\!\left[-\frac{|\mathbf{r}-\mathbf{r}'|^2}{2\sigma^2}\right],$$0

while the critical exponent at fixed energy remains close to the uncorrelated value $$\langle V(\mathbf{r})V(\mathbf{r}')\rangle=W^2 \exp\!\left[-\frac{|\mathbf{r}-\mathbf{r}'|^2}{2\sigma^2}\right],$$1 (Croy et al., 2011). In one-dimensional magnetic metamaterials, short-range dimer-type correlations increase the average participation ratio, change transmission from exponential to power-law decay for weak disorder, and produce nonzero transport exponents $$\langle V(\mathbf{r})V(\mathbf{r}')\rangle=W^2 \exp\!\left[-\frac{|\mathbf{r}-\mathbf{r}'|^2}{2\sigma^2}\right],$$2, including super-diffusive behavior, in contrast to the uncorrelated case with $$\langle V(\mathbf{r})V(\mathbf{r}')\rangle=W^2 \exp\!\left[-\frac{|\mathbf{r}-\mathbf{r}'|^2}{2\sigma^2}\right],$$3 (Molina, 2020).

Quasi-periodic correlated disorder in magnetically doped topological-insulator thin films introduces an additional anisotropic control parameter: orientation. For diagonal or longitudinal quasi-periodicity, disorder can induce QAHI, QSCI, QSHI, and topological Anderson-insulator regimes before Anderson insulation at large $$\langle V(\mathbf{r})V(\mathbf{r}')\rangle=W^2 \exp\!\left[-\frac{|\mathbf{r}-\mathbf{r}'|^2}{2\sigma^2}\right],$$4; diagonal quasi-periodicity also yields quantized transport from extended bulk states. For purely transverse quasi-periodicity, the QSHI and QSCI phases persist to arbitrarily strong disorder potential. Except for the transition to the Anderson insulator, these phase boundaries are described by a self-consistent Born approximation adapted to the correlated disorder structure (Okugawa et al., 2022).

Taken together, these studies show that correlated disorder modifies topology through at least three distinct mechanisms: smoothing and momentum filtering of the random potential, correlation-induced renormalization of effective Hamiltonian parameters, and geometrical reorganization of extended bulk pathways. A plausible implication is that the sign of the correlation effect cannot be inferred from disorder smoothness alone; one must also resolve the relevant gap geometry, orientation, and percolative connectivity.

5. Interacting, non-Hermitian, and rare-event regimes

In interacting and non-Hermitian systems, correlated disorder can alter stability thresholds and even change the qualitative mechanism of localization. For vortex-bound Majorana zero modes in a two-dimensional $$\langle V(\mathbf{r})V(\mathbf{r}')\rangle=W^2 \exp\!\left[-\frac{|\mathbf{r}-\mathbf{r}'|^2}{2\sigma^2}\right],$$5 topological superconductor, the disorder-averaged wavefunction obeys

$$\langle V(\mathbf{r})V(\mathbf{r}')\rangle=W^2 \exp\!\left[-\frac{|\mathbf{r}-\mathbf{r}'|^2}{2\sigma^2}\right],$$6

leading to a disorder-enhanced localization length

$$\langle V(\mathbf{r})V(\mathbf{r}')\rangle=W^2 \exp\!\left[-\frac{|\mathbf{r}-\mathbf{r}'|^2}{2\sigma^2}\right],$$7

Correlated disorder is most detrimental when the correlation length $$\langle V(\mathbf{r})V(\mathbf{r}')\rangle=W^2 \exp\!\left[-\frac{|\mathbf{r}-\mathbf{r}'|^2}{2\sigma^2}\right],$$8 is comparable to the superconducting coherence length $$\langle V(\mathbf{r})V(\mathbf{r}')\rangle=W^2 \exp\!\left[-\frac{|\mathbf{r}-\mathbf{r}'|^2}{2\sigma^2}\right],$$9; for $\sigma$0, fluctuations self-average, while for $\sigma$1, an MZM can survive in a favorable domain (Christian et al., 2020).

In non-Hermitian lattices, anti-symmetric disorder correlations between coupled chains can induce delocalization rather than merely reshape localization length. A strongly disordered Hatano–Nelson chain coupled to a disordered Hermitian chain remains localized for symmetric disorder correlations, but with anti-symmetric correlations $\sigma$2, increasing the inter-chain coupling $\sigma$3 can restore Anderson delocalization and a reentrant non-Hermitian skin effect for arbitrarily large bare disorder strength. The transition is tracked by inverse participation ratios, mean center of mass, and a real-space winding number (Jin et al., 2023).

In many-body localization with correlated random fields, the effective disorder strength is encoded by the average sample variance

$\sigma$4

Gap statistics and extremal entanglement-spectrum measures collapse when plotted against $\sigma$5, allowing an analytic phase boundary,

$\sigma$6

with increasing correlations favoring ergodicity by suppressing intra-sample randomness (Samanta et al., 2021).

Correlated disorder also changes rare-event physics. In one-dimensional tight-binding lattices with correlated diagonal disorder, the effective correlation length can minimize the local disorder strength and enhance rogue-wave-like extreme amplitudes; the right tail of the amplitude distribution follows a Gumbel law, and the largest events occur at intermediate correlations rather than in either limiting regime (Buarque et al., 2022). In waveguide QED, frequency disorder in a chain of qubits can generate antibunching and nearly perfect photon blockade in transmission and reflection through near-complete destructive interference of photon-scattering paths, an effect absent in ordered chains (Tian et al., 13 Oct 2025).

These examples show that correlated disorder often acts through scale matching: $\sigma$7 for Majorana modes, $\sigma$8 for many-body localization, and interference-path matching in photonics and non-Hermitian chains. This suggests that correlation effects are frequently strongest when the disorder structure resonates with a preexisting intrinsic length or scattering manifold.

6. Structural realizations, criticality, and effective theories

Correlated disorder is equally important as a structural descriptor of real materials. In epitaxial Bi$\sigma$9Se$C_{ij}=h^2[\delta_{ij}+\alpha(1-\delta_{ij})],$0/graphene/SiC heterostructures, X-ray diffraction microscopy and reciprocal-space imaging reveal spatially correlated structural disorder coexisting with high crystallographic order along the growth direction. Substrate step edges induce lattice tilts, in-plane angular disorder retains long-range memory across layers, and the resulting stack acquires a “pseudo-3D character” rather than behaving as a collection of nearly independent two-dimensional units (Laanait et al., 2016). In epitaxially stabilized $C_{ij}=h^2[\delta_{ij}+\alpha(1-\delta_{ij})],$1, tuneable correlated disorder arises from a mismatch between the preferred symmetry of the crystallographic basis and the global lattice. Diffuse X-ray scattering identifies short-range correlated displacive disorder, while grazing-incidence inelastic X-ray scattering and ab initio molecular dynamics show strong disorder–phonon coupling and strong suppression of phonon lifetimes beyond alloying alone (Chaney et al., 2020).

Several theories formalize these effects beyond independent-defect approximations. A generalized coherent potential approximation for elastic systems replaces the single-impurity description by an $C_{ij}=h^2[\delta_{ij}+\alpha(1-\delta_{ij})],$2-impurity or cluster approximation with jointly sampled defect variables, leading to a matrix self-consistency condition $C_{ij}=h^2[\delta_{ij}+\alpha(1-\delta_{ij})],$3. Applied to a rigidity-percolation model for colloidal gels, correlations lower the critical packing fraction but leave the critical coordination number isostatic, with

$C_{ij}=h^2[\delta_{ij}+\alpha(1-\delta_{ij})],$4

in the two-dimensional example discussed (Escobar-Agudelo et al., 2 Oct 2025). In the $C_{ij}=h^2[\delta_{ij}+\alpha(1-\delta_{ij})],$5-state Potts model with slowly decaying disorder correlations,

$C_{ij}=h^2[\delta_{ij}+\alpha(1-\delta_{ij})],$6

Monte Carlo simulations at infinite disorder identify the magnetic scaling dimension of a correlated-percolation fixed point, while finite disorder introduces an ultimately unstable crossover with enormous crossover lengths and a Griffiths phase (Chatelain, 2016).

Correlated disorder can also coexist with partial order rather than replacing order outright. In the $C_{ij}=h^2[\delta_{ij}+\alpha(1-\delta_{ij})],$7 Heisenberg antiferromagnet on the $C_{ij}=h^2[\delta_{ij}+\alpha(1-\delta_{ij})],$8-distorted triangular lattice, the weakly frustrated regime $C_{ij}=h^2[\delta_{ij}+\alpha(1-\delta_{ij})],$9 exhibits a partially disordered ground state: a honeycomb subsystem retains $\alpha=0$0 Néel order, while central spins remain disordered but develop short-range ferromagnetic correlations

$\alpha=0$1

The proposed mechanism is a Casimir-like effect mediated by zero-point fluctuations of the ordered subsystem (Gonzalez et al., 2018). In Kondo lattices, two-particle impurity correlations encoded by a form factor $\alpha=0$2 weaken the suppression of the coherence temperature $\alpha=0$3, especially when $\alpha=0$4, providing a mean-field explanation for the observed robustness of coherence in some disordered heavy-fermion materials (Dzero et al., 2011).

Across these structural and critical settings, correlated disorder appears not merely as randomness with memory, but as an organizing principle that can preserve hidden constraints, create pseudo-dimensionality, generate disorder–phonon coupling, or produce effective interactions through fluctuations. A plausible implication is that the most consequential role of correlated disorder is often mesoscale: it determines which local motifs, domains, or clusters are statistically permitted to persist and communicate across a system.