pith. machine review for the scientific record. sign in

arxiv: 2604.22447 · v1 · submitted 2026-04-24 · ❄️ cond-mat.stat-mech · cond-mat.other· math-ph· math.MP· physics.data-an

Recognition: unknown

Long-Range Correlated Random Matrices

Abbas Ali Saberi , Roderich Moessner
Authors on Pith no claims yet

Pith reviewed 2026-05-08 09:32 UTC · model grok-4.3

classification ❄️ cond-mat.stat-mech cond-mat.othermath-phmath.MPphysics.data-an
keywords random matrix theoryeigenvalue distributionlong-range correlationspercolationspectral densitykurtosissemicircle lawgeneralized t-distribution
0
0 comments X

The pith

Power-law correlations between random matrix entries induce a transition in eigenvalue distributions at the critical exponent H equals 3/4.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

This work demonstrates that algebraic correlations in the entries of random matrices, decaying as the inverse power of distance with exponent 2H, modify the statistics of the matrix eigenvalues in a controlled way. Using a percolation-based construction to impose these correlations, the analysis reveals that the eigenvalue probability density changes character as H is varied. Below the threshold value of 3/4 the distribution is a fat-tailed generalized t form, at the threshold Gaussian behavior appears, and above it the shape approaches the semicircle law while the excess kurtosis changes sign. Such results are relevant for any field that employs random matrices to describe systems with spatial or long-range dependencies.

Core claim

Motivated by the role of correlations in random matrices modeling various phenomena, algebraic correlations between matrix elements that decay as a power law proportional to r to the minus 2H are introduced via a long-range correlated percolation model. As the exponent H varies, the eigenvalue distribution and excess kurtosis change qualitatively. At the threshold H sub c equals 3/4, where Gaussian statistics emerge, a sign change in excess kurtosis marks the transition from a fat-tailed generalized t-distribution to distributions that approach the standard semicircle law for H much larger than H sub c. Analytical results from scaling analysis are confirmed by numerical simulations.

What carries the argument

A long-range correlated percolation model that generates entries with algebraic correlations proportional to r to the minus 2H, allowing the exponent H to be varied and its impact on eigenvalue statistics to be tracked via scaling analysis and simulations.

If this is right

  • The eigenvalue distribution is fat-tailed and follows a generalized t-distribution when H is less than 3/4.
  • Emergent Gaussian statistics appear precisely at H equals 3/4, identified by the zero crossing of excess kurtosis.
  • For large H the spectral density converges to the Wigner semicircle law of standard random matrix theory.
  • New regimes for correlated random matrix ensembles are identified through this construction.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Applications to physical systems with long-range interactions, such as certain spin models or neural networks, could use this to predict their spectral properties.
  • The connection between percolation and random matrix spectra may allow similar models for other correlation forms.
  • Testing the predictions in finite systems would require careful control of correlation length to match the assumed power-law decay.

Load-bearing premise

The percolation model accurately implements algebraic correlations that fall off exactly as the power law r to the minus 2H, and the scaling analysis accounts for the eigenvalue statistics without significant unaccounted corrections from finite size or higher-order terms.

What would settle it

Constructing many random matrices with inter-element correlations decaying as r to the minus 2H for H set near 0.75 and checking if the eigenvalue histogram matches a generalized t or Gaussian form, and whether for H greater than 1 it matches the semicircle; a mismatch in the location of the kurtosis sign change or the limiting shape would falsify the described transition.

Figures

Figures reproduced from arXiv: 2604.22447 by Abbas Ali Saberi, Roderich Moessner.

Figure 1
Figure 1. Figure 1: FIG. 1. Numerical simulations of the variance of eigenvalues view at source ↗
Figure 2
Figure 2. Figure 2: FIG. 2. Numerical simulations of the excess kurtosis (EK) as view at source ↗
Figure 3
Figure 3. Figure 3: FIG. 3. Probability density function view at source ↗
read the original abstract

Motivated by the importance ascribed to correlations in random matrices used to model phenomena in various scientific disciplines, we report how algebraic correlations between matrix elements affect the eigenvalue statistics and spectral density of random matrices. These correlations, introduced through a long-range correlated percolation model, decay as a power law $\propto r^{-2H}$, with exponent $H > 0$. As $H$ varies, both the eigenvalue distribution and excess kurtosis undergo qualitative changes. At the threshold $H_c = 3/4$, characterized by emergent Gaussian statistics, a sign change in excess kurtosis marks a transition from a fat-tailed generalized $t$-distribution to one that gradually approaches the standard semicircle law for $H \gg H_c$. Our analytical results, based on scaling analysis and supported by extensive numerical simulations, provide clear predictions and uncover novel spectral regimes in random matrix theory. Our results connect techniques from statistical physics, percolation theory, and random matrix analysis, offering a new perspective on universality in correlated ensembles.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

3 major / 2 minor

Summary. The manuscript introduces long-range algebraic correlations ∝ r^{-2H} (H>0) into random matrices via a percolation-based construction. Scaling analysis of the fourth moment combined with numerical simulations is used to argue that the eigenvalue distribution and excess kurtosis undergo a qualitative change at the threshold H_c=3/4: the statistics are Gaussian there, excess kurtosis changes sign, the distribution is a fat-tailed generalized t-distribution for H<H_c, and it gradually approaches the Wigner semicircle for H≫H_c.

Significance. If the scaling derivation is rigorous and the simulations confirm that lattice artifacts do not shift the kurtosis zero-crossing, the result would be significant: it supplies a single tunable exponent H that controls the entire spectral regime, including an emergent Gaussian point and a crossover to semicircle statistics. The explicit bridge between percolation theory and random-matrix universality classes is a strength, as is the combination of scaling arguments with extensive numerics that yields falsifiable predictions for the kurtosis sign change.

major comments (3)
  1. [Abstract / scaling analysis] The scaling analysis that produces the exact threshold H_c=3/4 is only asserted in the abstract; no explicit steps, moment calculation, or equation showing how the fourth-moment scaling yields a kurtosis sign change precisely at 3/4 are supplied. Because this derivation is load-bearing for the central claim of a sharp transition, the full steps (including any assumptions about the correlation function) must be provided, e.g., in a dedicated section or appendix.
  2. [Model / percolation implementation] The percolation construction is stated to realize exact algebraic decay ∝ r^{-2H}. On a finite lattice, however, cluster geometry necessarily introduces short-range cutoffs and higher-order dependencies. The manuscript must demonstrate (via explicit correlation-function measurement or analytic argument) that these corrections remain sub-dominant for the eigenvalue statistics and do not displace the kurtosis zero-crossing away from H=3/4.
  3. [Numerical simulations] Numerical support is invoked for the transition, yet no information is given on lattice sizes, ensemble sizes, or finite-size scaling procedure used to locate the kurtosis sign change. Without these controls, it is impossible to rule out that the apparent transition is an artifact of the discrete lattice rather than the continuum scaling limit assumed in the analytic argument.
minor comments (2)
  1. [Abstract] The phrase 'generalized t-distribution' should be defined or referenced to a standard form (e.g., via its pdf or moment properties) so that readers can verify the claimed fat-tailed behavior.
  2. [Abstract] Notation for the excess kurtosis (e.g., whether it is normalized by the Gaussian value or left un-normalized) should be stated explicitly when the sign-change claim is introduced.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the careful reading and constructive comments on our manuscript. The points raised identify important aspects of the scaling derivation, model validation, and numerical controls that require clearer exposition. We address each major comment below and will incorporate the requested clarifications and additions in a revised version.

read point-by-point responses

  1. Referee: [Abstract / scaling analysis] The scaling analysis that produces the exact threshold H_c=3/4 is only asserted in the abstract; no explicit steps, moment calculation, or equation showing how the fourth-moment scaling yields a kurtosis sign change precisely at 3/4 are supplied. Because this derivation is load-bearing for the central claim of a sharp transition, the full steps (including any assumptions about the correlation function) must be provided, e.g., in a dedicated section or appendix.

    Authors: We agree that the explicit derivation of H_c=3/4 from fourth-moment scaling is central and currently insufficiently detailed. In the revision we will add a dedicated section (or appendix) that presents the full moment calculation step by step. This will include the expression for the fourth moment under the algebraic correlation ∝ r^{-2H}, the assumptions of isotropy and power-law decay (with no additional higher-order correlations), and the scaling argument that shows the excess kurtosis changes sign exactly at H=3/4, corresponding to the emergence of Gaussian statistics. The revised text will make the analytic threshold fully transparent and self-contained. revision: yes

  2. Referee: [Model / percolation implementation] The percolation construction is stated to realize exact algebraic decay ∝ r^{-2H}. On a finite lattice, however, cluster geometry necessarily introduces short-range cutoffs and higher-order dependencies. The manuscript must demonstrate (via explicit correlation-function measurement or analytic argument) that these corrections remain sub-dominant for the eigenvalue statistics and do not displace the kurtosis zero-crossing away from H=3/4.

    Authors: We acknowledge that finite-lattice percolation clusters can introduce short-range cutoffs. To address this we will add explicit numerical measurements of the two-point correlation function of the matrix elements on the lattices used in the simulations. These measurements will confirm that the algebraic tail ∝ r^{-2H} dominates for the relevant distances, with short-range corrections affecting only a vanishing fraction of pairs as system size grows. We will also supply a brief analytic argument showing that such corrections enter the fourth-moment scaling only as sub-leading terms that do not shift the location of the kurtosis zero-crossing in the thermodynamic limit. revision: yes

  3. Referee: [Numerical simulations] Numerical support is invoked for the transition, yet no information is given on lattice sizes, ensemble sizes, or finite-size scaling procedure used to locate the kurtosis sign change. Without these controls, it is impossible to rule out that the apparent transition is an artifact of the discrete lattice rather than the continuum scaling limit assumed in the analytic argument.

    Authors: We regret the omission of these methodological details. In the revised manuscript we will add a dedicated subsection describing the numerical protocol: lattice linear sizes L ranging from 128 to 1024, ensemble sizes of at least 10^4 independent realizations per H value, and the finite-size scaling procedure (including data collapse and extrapolation of the kurtosis zero-crossing) used to confirm convergence to H=3/4 in the continuum limit. We will also include supplementary plots demonstrating that the sign-change location stabilizes with increasing L, thereby ruling out lattice artifacts. revision: yes

Circularity Check

0 steps flagged

Scaling analysis yields independent derivation of H_c threshold without reduction to inputs

full rationale

The paper derives the threshold H_c = 3/4 and associated changes in eigenvalue statistics and excess kurtosis via scaling analysis applied to the fourth moment of the long-range correlated percolation model whose correlations decay as r^{-2H}. This analytical step is presented as producing predictions that are then checked against numerics, with no evidence that the claimed distributions or the location of the kurtosis sign change are obtained by fitting parameters to the target quantities or by any self-definitional loop. The abstract and claims treat the scaling result as logically prior to the reported regimes (fat-tailed t-distribution below H_c, Gaussian at threshold, semicircle approach above), and no load-bearing self-citation or ansatz smuggling is required for the central chain. The derivation therefore remains self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim rests on the assumption that the percolation construction produces the stated power-law correlations and that scaling arguments suffice to determine the eigenvalue statistics.

axioms (1)
  • domain assumption Matrix elements possess long-range algebraic correlations that decay as r^{-2H} and are realized through a percolation model.
    This is the modeling choice stated in the abstract that defines the ensemble under study.

pith-pipeline@v0.9.0 · 5478 in / 1372 out tokens · 70593 ms · 2026-05-08T09:32:28.557217+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

56 extracted references

  1. [1]

    EK(λ)∝ R |λ|4−(f+1)dλ∝ |λ| 4−f, which indeed diverges for f≤4 as|λ| → ∞

    The symbols represent numerical results forL= 2 6 to 2 12, while the solid lines correspond to the theoretical prediction∝L −(H+1/4). EK(λ)∝ R |λ|4−(f+1)dλ∝ |λ| 4−f, which indeed diverges for f≤4 as|λ| → ∞. To ensure a finite and well-behaved probability density function forλ≪1, in particular avoiding the singulari- ties present in the pure power-law, we ...

  2. [2]

    The symbols represent numerical computations forL= 2 6 to 212, the solid lines the theoretical prediction Eq. (2). The logarithmic scal- ing predicted forH= 1 4 is fitted by the dashed line. 4 FIG. 3. Probability density functionP(λ) of scaled eigenvalues forH= 1 8 , 1 4 , 3 8 , 1 2 , 5 8 ,and 3

  3. [3]

    For sizes fromL= 2 6 to 212, the data collapse well. For each fixedH, the solid curves for allLlie on top of one another within the line width, so most traces are visually indistinguishable and only the largest-Lcurve is clearly visible. The black dashed lines represent the analytical predictions Eqs. (3)-(5); the blue dashed line atH= 3 4 shows the Gauss...

  4. [4]

    (2) predicts a logarithmic dependence onL, which is consis- tent with our numerics, as illustrated by the dashed line in Fig

    Notably, forH= 1 4, Eq. (2) predicts a logarithmic dependence onL, which is consis- tent with our numerics, as illustrated by the dashed line in Fig. 2. Finally, Fig. 3 shows the probability density function of the scaled eigenvalues for the same values ofH. For eachH, the simulations were carried out across different Lvalues, ranging from 2 6 to 2 12. Al...

  5. [5]

    Strongly Correlated Extreme Fluctuations

    A full account of these correlation results, including deriva- tions and comprehensive numerics, will appear in a sep- arate paper. Acknowledgments— This work was funded by the Deutsche Forschungsgemeinschaft (DFG, German Re- search Foundation) under Project No. 557852701 (A.A.S.), and in part by the Deutsche Forschungsge- meinschaft via Research Unit FOR...

    2024

  6. [6]

    T. Guhr, A. Muller-Groeling, and H. A. Weidenmuller, Physics Reports299, 189 (1998)

    1998

  7. [7]

    E. P. Wigner, Mathematical Proceedings of the Cam- bridge Philosophical Society47, 790 (1951)

    1951

  8. [8]

    Laloux, P

    L. Laloux, P. Cizeau, M. Potters, and J.-P. Bouchaud, International Journal of Theoretical and Applied Finance 3, 391 (2000)

    2000

  9. [9]

    Plerou, P

    V. Plerou, P. Gopikrishnan, B. Rosenow, L. A. Ama- ral, and H. E. Stanley, Physical Review Letters83, 1471 (1999)

    1999

  10. [10]

    Wainrib and J

    G. Wainrib and J. Touboul, Physical Review Letters110, 118101 (2013)

    2013

  11. [11]

    Couillet and M

    R. Couillet and M. Debbah,Random matrix methods for wireless communications(Cambridge University Press, 2011)

    2011

  12. [12]

    A. M. Tulino and S. Verdu,Random matrix theory and wireless communications(Now Publishers Inc., 2004)

    2004

  13. [13]

    Allesina and S

    S. Allesina and S. Tang, Nature483, 205 (2012)

    2012

  14. [14]

    J. W. Baron, T. J. Jewell, C. Ryder, and T. Galla, Phys- ical Review Letters130, 137401 (2023)

    2023

  15. [15]

    M. S. Santhanam and P. K. Patra, Physical Review E 64, 016102 (2001)

    2001

  16. [16]

    ¨Onder, A

    Y. ¨Onder, A. A. Saberi, and R. Moessner, Physical Re- view Letters136, 087103 (2026)

    2026

  17. [17]

    Stanley, S

    H. Stanley, S. Buldyrev, A. Goldberger, Z. Goldberger, S. Havlin, R. N. Mantegna, S. Ossadnik, C.-K. Peng, and M. Simons, Physica A: Statistical Mechanics and its Applications205, 214 (1994)

    1994

  18. [18]

    Rangarajan and M

    G. Rangarajan and M. Ding,Processes with long-range correlations: Theory and applications, Vol. 621 (Springer, 2008)

    2008

  19. [19]

    E. G. Altmann, G. Cristadoro, and M. D. Esposti, Pro- ceedings of the National Academy of Sciences109, 11582 (2012)

    2012

  20. [20]

    Seba, Physical Review Letters91, 198104 (2003)

    P. Seba, Physical Review Letters91, 198104 (2003)

    2003

  21. [21]

    Zhang and B

    W. Zhang and B. Dietz, Physical Review B104, 064310 (2021)

    2021

  22. [22]

    Haque, P

    M. Haque, P. A. McClarty, and I. M. Khaymovich, Phys- ical Review E105, 014109 (2022)

    2022

  23. [23]

    A. M. Garc´ ıa-Garc´ ıa, T. Nosaka, D. Rosa, and J. J. Ver- baarschot, Physical Review D100, 026002 (2019)

    2019

  24. [24]

    J. Wang, M. H. Lamann, J. Richter, R. Steinigeweg, A. Dymarsky, and J. Gemmer, Physical Review Letters 128, 180601 (2022)

    2022

  25. [25]

    J. Che, X. Zhang, W. Zhang, B. Dietz, and G. Chai, Physical Review E106, 014211 (2022)

    2022

  26. [26]

    X. Wang, J. Wang, and W.-g. Wang, arXiv e-prints , arXiv (2025)

    2025

  27. [27]

    M´ ezard, G

    M. M´ ezard, G. Parisi, and A. Zee, Nuclear Physics B 559, 689 (1999)

    1999

  28. [28]

    I. M. Johnstone, The Annals of statistics29, 295 (2001)

    2001

  29. [29]

    H. J. Sommers, A. Crisanti, H. Sompolinsky, and Y. Stein, Physical review letters60, 1895 (1988)

    1988

  30. [30]

    Akemann and G

    G. Akemann and G. Vernizzi, Nuclear Physics B660, 532 (2003)

    2003

  31. [31]

    J. W. Baron, T. J. Jewell, C. Ryder, and T. Galla, Phys- ical Review Letters128, 120601 (2022)

    2022

  32. [32]

    Penrose,Random geometric graphs, Vol

    M. Penrose,Random geometric graphs, Vol. 5 (OUP Ox- ford, 2003)

    2003

  33. [33]

    Burda, J

    Z. Burda, J. Jurkiewicz, and B. Wac law, Physical Review E71, 026111 (2005)

    2005

  34. [34]

    Altland, K

    A. Altland, K. W. Kim, T. Micklitz, M. Rezaei, J. Son- ner, and J. J. Verbaarschot, Physical Review Research6, 033286 (2024)

    2024

  35. [35]

    A. C. Bertuola, O. Bohigas, and M. P. Pato, Phys. Rev. E70, 065102 (2004)

    2004

  36. [36]

    Toscano, R

    F. Toscano, R. O. Vallejos, and C. Tsallis, Phys. Rev. E 69, 066131 (2004)

    2004

  37. [37]

    A. Y. Abul-Magd, Phys. Rev. E71, 066207 (2005). 6

    2005

  38. [38]

    K. A. Muttalib and J. R. Klauder, Phys. Rev. E71, 055101 (2005)

    2005

  39. [39]

    Bohigas, J

    O. Bohigas, J. X. de Carvalho, and M. P. Pato, Phys. Rev. E77, 011122 (2008)

    2008

  40. [40]

    Bohigas and M

    O. Bohigas and M. P. Pato, Phys. Rev. E84, 031121 (2011)

    2011

  41. [41]

    Akemann and P

    G. Akemann and P. Vivo, J. Stat. Mech. , P09002 (2008)

    2008

  42. [42]

    Sahimi, Journal de Physique I4, 1263 (1994)

    M. Sahimi, Journal de Physique I4, 1263 (1994)

    1994

  43. [43]

    Prakash, S

    S. Prakash, S. Havlin, M. Schwartz, and H. E. Stanley, Physical Review A46, R1724 (1992)

    1992

  44. [44]

    ForH <1, the slow power-law decay of correlations en- sures that occupancy at one site influences occupancy at distant sites, leading to persistent correlations over large distances. In contrast, in random percolation, the span- ning clusters appear only at the critical threshold due to a statistical phase transition, not due to correlations between site ...

  45. [45]

    Weinrib, Physical Review B29, 387 (1984)

    A. Weinrib, Physical Review B29, 387 (1984)

    1984

  46. [46]

    A. B. Harris, Journal of Physics C: Solid State Physics 7, 1671 (1974)

    1974

  47. [47]

    Weinrib and B

    A. Weinrib and B. I. Halperin, Physical Review B27, 413 (1983)

    1983

  48. [48]

    Sahimi,Applications of percolation theory(CRC Press, 1994)

    M. Sahimi,Applications of percolation theory(CRC Press, 1994)

    1994

  49. [49]

    Coniglio and W

    A. Coniglio and W. Klein, Journal of Physics A: Mathe- matical and General13, 2775 (1980)

    1980

  50. [50]

    Coniglio, Physical review letters62, 3054 (1989)

    A. Coniglio, Physical review letters62, 3054 (1989)

    1989

  51. [51]

    A. A. Saberi, Applied Physics Letters97(2010)

    2010

  52. [52]

    Saber and A

    S. Saber and A. A. Saberi, Physical Review E105, L022102 (2022)

    2022

  53. [53]

    Malekan, S

    A. Malekan, S. Saber, and A. A. Saberi, Chaos: An In- terdisciplinary Journal of Nonlinear Science32, 023112 (2022)

    2022

  54. [54]

    A. A. Saberi, S. Saber, and R. Moessner, Physical Review B110, L180102 (2024). [50]https://www.pks.mpg.de/asg2024

    2024

  55. [55]

    N. M. Temme, Error functions, dawson’s and fresnel in- tegrals. (2010)

    2010

  56. [56]

    R. J. Adler,The geometry of random fields(SIAM, 2010). End Matter Appendix A:Scaling Analysis of the Eigenvalue vari- ance—Consider a symmetric random matrixMof size L×L, with diagonal elementsM ii =±1 and off-diagonal elementsM ij =±1 or 0. The correlation between ele- mentsM ij andM kl is given by: E[MijMkl]∼ p (i−k) 2 + (j−l) 2 −2H =r −2H , whereH >0. ...

    2010