pith. sign in

arxiv: 2605.31054 · v1 · pith:66UVSMD6new · submitted 2026-05-29 · 🧮 math.PR

On the empirical spectral distribution of matrix perpetuities

Bartosz Ko{\l}odziejek , Kamil Szpojankowski This is my paper

Pith reviewed 2026-06-28 21:21 UTC · model grok-4.3

classification 🧮 math.PR
keywords matrix perpetuitiesempirical spectral distributionpower-law tailsfree perpetuityorthogonal invariancepositive semidefinite matricesBeta prime distributionhigh-dimensional convergence
0
0 comments X

The pith

For positive semidefinite orthogonally invariant matrix perpetuities the expected empirical spectral distribution has power-law tails governed by the largest eigenvalue and converges in high dimensions to the corresponding free perpetuity.

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

The paper studies solutions to the matrix fixed-point equation X equals in distribution A X A transpose plus B, with A and B independent of X. It first links existence and uniqueness to classical vector perpetuities and builds tools that keep the matrix structure intact under orthogonal invariance. For positive semidefinite orthogonally invariant models it derives power-law tail asymptotics for the expected empirical spectral distribution, with the tail controlled by the largest eigenvalue. In the subcritical regime the expected empirical spectral distribution converges weakly to the law of the associated free perpetuity as dimension tends to infinity. The results are shown explicitly for matrix Beta prime perpetuities that possess closed-form limiting spectral distributions.

Core claim

Under the positive semidefinite and orthogonally invariant assumptions the expected empirical spectral distribution of the solution to the matrix perpetuity equation admits power-law tail asymptotics whose leading behavior is set by the largest eigenvalue; moreover, in the subcritical regime this distribution converges weakly in the high-dimensional limit to the distribution of the corresponding free perpetuity.

What carries the argument

The affine fixed-point equation X =_d A X A^T + B together with orthogonal invariance, which preserves the matrix structure needed to track the empirical spectral distribution.

If this is right

  • The tail of the expected empirical spectral distribution is asymptotically determined by the largest eigenvalue of the driving random matrix.
  • In the subcritical regime the high-dimensional limit of the expected empirical spectral distribution is exactly the distribution of the free perpetuity.
  • Explicit limiting spectral distributions become available for concrete families such as matrix Beta prime perpetuities.

Where Pith is reading between the lines

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

  • The free perpetuity therefore acts as a universal high-dimensional attractor for the spectral measure of such recursions under the stated invariance.
  • The same tail mechanism may govern the largest-eigenvalue statistics of other orthogonally invariant affine recursions that arise in covariance estimation or stability analysis.

Load-bearing premise

The models must be positive semidefinite and orthogonally invariant.

What would settle it

A direct computation or simulation for an orthogonally invariant positive semidefinite model in which the tail of the expected empirical spectral distribution fails to follow a power law set by the largest eigenvalue, or in which the high-dimensional weak limit is not the free perpetuity in the subcritical regime.

read the original abstract

We study matrix perpetuities, that is, solutions to affine fixed-point equations of the form \[ \mathbf{X} \stackrel{d}{=} \mathbf{A}\,\mathbf{X} \,\mathbf{A}^\top+\mathbf{B},\qquad (\mathbf{A},\mathbf{B})\mbox{ and }\mathbf{X} \mbox{ are independent}, \] with particular emphasis on the empirical spectral distribution of the solution. We first establish existence and uniqueness results by relating the problem to classical vector perpetuities, and then develop tools that preserve the matrix structure under orthogonal invariance. For positive semidefinite, orthogonally invariant models, we obtain power-law tail asymptotics for the expected empirical spectral distribution and show that the tail is governed by the largest eigenvalue. We also prove that, in the subcritical regime, the expected empirical spectral distribution of matrix perpetuities converges weakly, as the dimension tends to infinity, to the distribution of the corresponding free perpetuity. Our results are illustrated by matrix Beta prime perpetuities, for which explicit limiting spectral distributions are available.

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

2 major / 2 minor

Summary. The paper studies solutions X to the matrix perpetuity equation X =_d A X A^T + B (with (A,B) independent of X), with emphasis on the empirical spectral distribution (ESD) of X. It establishes existence and uniqueness by reduction to vector perpetuities, develops tools preserving structure under orthogonal invariance, derives power-law tail asymptotics for the expected ESD (governed by the largest eigenvalue) for positive semidefinite orthogonally invariant models, and proves weak convergence of the expected ESD to the law of the corresponding free perpetuity in the subcritical regime as dimension tends to infinity. Results are illustrated via matrix Beta prime perpetuities admitting explicit limiting spectral distributions.

Significance. If the derivations hold, the work provides a rigorous bridge between classical vector perpetuity theory, free probability, and high-dimensional random matrix theory. The orthogonal-invariance tools and explicit tail results for the expected ESD (controlled by the top eigenvalue) are potentially useful for spectral analysis of affine matrix recursions; the Beta prime example supplies a concrete case with closed-form limits. The restriction to positive semidefinite orthogonally invariant models is stated explicitly, avoiding hidden assumptions.

major comments (2)
  1. [§2 (existence/uniqueness)] The abstract states that existence/uniqueness follows from relating the matrix equation to vector perpetuities, but the precise moment conditions (e.g., E[log ||A||] < 0 or integrability of B) required for the reduction are not visible in the provided abstract; if these are only implicit in §2, the load-bearing step for all subsequent claims needs an explicit statement of the minimal assumptions.
  2. [tail asymptotics section] The power-law tail claim for the expected ESD is stated to be governed by the largest eigenvalue under orthogonal invariance; however, without seeing the precise statement of the tail exponent (e.g., whether it equals the top Lyapunov exponent or involves the spectral radius of E[A ⊗ A]), it is unclear whether the result is parameter-free or requires additional eigenvalue bounds that could affect the subcritical convergence.
minor comments (2)
  1. [Introduction / §3] Ensure the subcritical regime is defined at the first use (likely via a spectral-radius or Lyapunov condition) rather than only in the convergence theorem, to aid readability.
  2. [examples section] In the Beta prime illustration, clarify whether the explicit limiting spectral distribution is obtained by direct computation or by invoking known free-probability results; a short derivation sketch would strengthen the example.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the careful reading, positive assessment, and constructive comments on our manuscript. We address each major comment below and will incorporate the requested clarifications in the revised version.

read point-by-point responses

  1. Referee: [§2 (existence/uniqueness)] The abstract states that existence/uniqueness follows from relating the matrix equation to vector perpetuities, but the precise moment conditions (e.g., E[log ||A||] < 0 or integrability of B) required for the reduction are not visible in the provided abstract; if these are only implicit in §2, the load-bearing step for all subsequent claims needs an explicit statement of the minimal assumptions.

    Authors: We agree that the minimal assumptions should be stated explicitly. In the revised manuscript we will update the abstract to include the precise moment conditions (E[log ||A||] < 0 together with the integrability requirements on B) that justify the reduction to vector perpetuities, and we will restate these conditions prominently at the opening of Section 2 so that they are visible before any subsequent claims. revision: yes

  2. Referee: [tail asymptotics section] The power-law tail claim for the expected ESD is stated to be governed by the largest eigenvalue under orthogonal invariance; however, without seeing the precise statement of the tail exponent (e.g., whether it equals the top Lyapunov exponent or involves the spectral radius of E[A ⊗ A]), it is unclear whether the result is parameter-free or requires additional eigenvalue bounds that could affect the subcritical convergence.

    Authors: Under the orthogonal-invariance assumption the power-law tail of the expected ESD is governed by the largest eigenvalue, with the precise exponent given by the associated top Lyapunov exponent. The result is not parameter-free; it is controlled by this Lyapunov exponent and holds under the subcritical regime already assumed in the paper. No further eigenvalue bounds are required. We will revise the tail-asymptotics section to state the exponent explicitly and to confirm that the high-dimensional convergence statement remains valid under the stated subcritical conditions. revision: yes

Circularity Check

0 steps flagged

No significant circularity; derivations rely on external vector perpetuities and free probability

full rationale

The paper establishes existence/uniqueness by relating matrix perpetuities to classical vector perpetuities, develops orthogonal-invariance tools, and derives power-law tail asymptotics plus weak convergence to the free perpetuity under the explicit assumptions of positive semidefiniteness and orthogonal invariance. These steps use independent external frameworks (vector fixed-point theory and free probability) rather than reducing any central claim to a self-referential definition, fitted parameter renamed as prediction, or load-bearing self-citation chain. The Beta-prime illustration invokes independently available explicit distributions. The structure is self-contained against external mathematical benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

Only the abstract is available, so the ledger reflects the high-level relations stated there: existence via vector perpetuities and orthogonal invariance for spectral results. No free parameters or invented entities are mentioned.

axioms (1)
  • domain assumption Existence and uniqueness results for classical vector perpetuities hold under the conditions needed to transfer to the matrix setting
    Invoked to establish matrix existence and uniqueness

pith-pipeline@v0.9.1-grok · 5708 in / 1246 out tokens · 35648 ms · 2026-06-28T21:21:07.659304+00:00 · methodology

discussion (0)

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

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Finite free perpetuities

    math.PR 2026-06 unverdicted novelty 7.0

    Finite free perpetuities are defined as degree-n monic polynomials solving a truncated perpetuity equation; the paper proves existence, uniqueness, real nonnegative zeros for admissible (A,B), and weak convergence of ...

Reference graph

Works this paper leans on

24 extracted references · 1 canonical work pages · cited by 1 Pith paper

  1. [1]

    Anderson, Alice Guionnet, and Ofer Zeitouni.An introduction to random matrices, volume 118 ofCambridge Studies in Advanced Mathematics

    Greg W. Anderson, Alice Guionnet, and Ofer Zeitouni.An introduction to random matrices, volume 118 ofCambridge Studies in Advanced Mathematics. Cambridge University Press, Cambridge, 2010

    2010

  2. [2]

    Matsumoto-Yor and Dufresne type theorems for a random walk on positive definite matrices.Ann

    Jonas Arista, Elia Bisi, and Neil O’Connell. Matsumoto-Yor and Dufresne type theorems for a random walk on positive definite matrices.Ann. Inst. Henri Poincar´ e Probab. Stat., 60(2):923–945, 2024

    2024

  3. [3]

    Belinschi, B

    S. Belinschi, B. Ko lodziejek, and K. Szpojankowski. Free Perpetuities I: Existence, Subordination and Tail Asymptotics.arXiv:2503.10319, 2025

    work page arXiv 2025

  4. [4]

    Belinschi and Alexandru Nica

    Serban T. Belinschi and Alexandru Nica. On a remarkable semigroup of homo- morphisms with respect to free multiplicative convolution.Indiana Univ. Math. J., 57(4):1679–1713, 2008

    2008

  5. [5]

    Strict stationarity of generalized autoregressive processes.Ann

    Philippe Bougerol and Nico Picard. Strict stationarity of generalized autoregressive processes.Ann. Probab., 20(4):1714–1730, 1992

    1992

  6. [6]

    Buraczewski, E

    D. Buraczewski, E. Damek, and T. Mikosch.Stochastic models with power-law tails. Springer Series in Operations Research and Financial Engineering. Springer, [Cham],

  7. [7]

    Matsumoto-Yor processes on Jordan algebras

    Reda Chhaibi and Manon Defosseux. Matsumoto-Yor processes on Jordan algebras. Probab. Theory Relat. Fields, 2025

    2025

  8. [8]

    Damek and S

    E. Damek and S. Mentemeier. Analysing heavy-tail properties of stochastic gradient descent by means of stochastic recurrence equations.J. Appl. Probab., page 1–25, 2026

    2026

  9. [9]

    D. F´ eral. On large deviations for the spectral measure of discrete Coulomb gas. In S´ eminaire de probabilit´ es XLI, volume 1934 ofLecture Notes in Math., pages 19–49. Springer, Berlin, 2008

    1934

  10. [10]

    Charles M. Goldie. Implicit renewal theory and tails of solutions of random equations. Ann. Appl. Probab., 1(1):126–166, 1991

    1991

  11. [11]

    Guivarc’h and ´E

    Y. Guivarc’h and ´E. Le Page. Spectral gap properties for linear random walks and Pareto’s asymptotics for affine stochastic recursions.Ann. Inst. Henri Poincar´ e Probab. Stat., 52(2):503–574, 2016

    2016

  12. [12]

    Hiai and D

    F. Hiai and D. Petz.The semicircle law, free random variables and entropy, vol- ume 77 ofMathematical Surveys and Monographs. American Mathematical Society, Providence, RI, 2000

    2000

  13. [13]

    Liam Hodgkinson, Zhichao Wang, and Michael W. Mahoney. Models of heavy- tailed mechanistic universality. InForty-second International Conference on Machine Learning, 2025

    2025

  14. [14]

    Horn and Charles R

    Roger A. Horn and Charles R. Johnson.Matrix analysis. Cambridge University Press, Cambridge, second edition, 2013

    2013

  15. [15]

    In- verse problems for regular variation of linear filters, a cancellation property forσ-finite measures and identification of stable laws.Ann

    Martin Jacobsen, Thomas Mikosch, Jan Rosi´ nski, and Gennady Samorodnitsky. In- verse problems for regular variation of linear filters, a cancellation property forσ-finite measures and identification of stable laws.Ann. Appl. Probab., 19(1):210–242, 2009. SPECTRAL DISTRIBUTION OF MATRIX PERPETUITIES 39

    2009

  16. [16]

    H. Kesten. Random difference equations and renewal theory for products of random matrices.Acta Math., 131:207–248, 1973

    1973

  17. [17]

    Ko lodziejek

    B. Ko lodziejek. The Matsumoto-Yor property and its converse on symmetric cones. J. Theoret. Probab., 30(2):624–638, 2017

    2017

  18. [18]

    An independence property for the product of GIG and gamma laws.Ann

    G´ erard Letac and Jacek Weso lowski. An independence property for the product of GIG and gamma laws.Ann. Probab., 28(3):1371–1383, 2000

    2000

  19. [19]

    The Matsumoto-Yor property and the struc- ture of the Wishart distribution.J

    H´ el` ene Massam and Jacek Weso lowski. The Matsumoto-Yor property and the struc- ture of the Wishart distribution.J. Multivariate Anal., 97(1):103–123, 2006

    2006

  20. [20]

    C. M. Newman. The distribution of Lyapunov exponents: exact results for random matrices.Comm. Math. Phys., 103(1):121–126, 1986

    1986

  21. [21]

    On the multiplication of freeN-tuples of noncommutative random variables.Amer

    Alexandru Nica and Roland Speicher. On the multiplication of freeN-tuples of noncommutative random variables.Amer. J. Math., 118(4):799–837, 1996

    1996

  22. [22]

    E. B. Saff and V. Totik.Logarithmic potentials with external fields, volume 316 of Grundlehren der mathematischen Wissenschaften [Fundamental Principles of Math- ematical Sciences]. Springer-Verlag, Berlin, 1997. Appendix B by Thomas Bloom

    1997

  23. [23]

    Fractional free convolution powers.Indiana Univ

    Dimitri Shlyakhtenko and Terence Tao. Fractional free convolution powers.Indiana Univ. Math. J., 71(6):2551–2594, 2022

    2022

  24. [24]

    H. Yoshida. Remarks on a free analogue of the beta prime distribution.J. Theoret. Probab., 33(3):1363–1400, 2020. Bartosz Ko lodziejek: Faculty of Mathematics and Information Sciences, Warsaw Uni- versity of Technology, Koszykowa 75, 00-662 Warsaw, Poland Email address:bartosz.kolodziejek@pw.edu.pl Kamil Szpojankowski: Institute of Mathematics of the Poli...

    2020