Pith. sign in

REVIEW 3 major objections 4 minor 30 references

Critical points of random polynomials and finite free cumulants

T0 review · 3 major / 4 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read Repeatedly differentiating random polynomials gives Gaussian fluctuations around Hermite limits.

desk verdict A promising method with a misstated central CLT: the proof gives -ell(ell-1)/2 He_{ell-2}, not +ell/2 He_{ell-2}, so Theorem 2.3 is false as written but easily corrected. read the letter →

arxiv 2506.08910 v2 pith:6VIHSX2M submitted 2025-06-10 math.PR math.OA

Octavio Arizmendi , Andrew Campbell , Katsunori Fujie This is my paper
classification math.PRmath.OA MSC 60B2060F0546L5430C15
keywords randompolynomialscriticalpointsfinitefreeprobabilitycumulantsHermiteAppellsequencesinfinitelydivisibledistributionscentrallimittheorem
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

It is known that repeatedly differentiating a random polynomial with independent mean-zero, variance-one roots stabilizes, after rescaling, to the Hermite polynomial. This paper establishes central limit theorems for the fluctuations around that deterministic limit: the rescaled differences for the polynomial and for its ordered roots converge to Gaussian distributions whose covariance depends on the root law only through its fourth moment. It also removes the assumption that the roots have a finite second moment, showing that when the roots belong to the domain of attraction of a general infinitely divisible law the Hermite limit is replaced by a random Appell sequence built from the Lévy triple. The proofs use finite free cumulants to track how root fluctuations propagate through differentiation.

What carries the argument

Finite free cumulants $\kappa^N_j(p)$, the coefficients of the finite $R$-transform $R_p(s) = -P'(Ns)/P(Ns)$ for $P(d/dx)x^N = p(x)$, linearize the finite free additive convolution $\boxplus_N$. The load-bearing identity is $\kappa^k_j(\partial_{k|N}p) = (k/N)^{j-1}\kappa^N_j(p)$, which lets the authors read off how cumulants of the original roots are scaled by differentiation; the moment-cumulant formula of Lemma 1.3 then converts the classical CLT for empirical moments into a CLT for finite free cumulants, and the $\Delta$ method transfers that CLT from moments to polynomial coefficients and roots.

What would settle it

Evaluate the coefficient of $x^{\ell-2}$ on both sides of Theorem 2.3 for $\ell=4$ using the paper's own expansion in equation (3.14): the stated limit has coefficient $(\ell/2)\sqrt{m_4(\mu)-1}\,Z = 2\sqrt{m_4(\mu)-1}\,Z$, while equation (3.14) gives $-\ell(\ell-1)/2\,\sqrt{m_4(\mu)-1}\,Z = -6\sqrt{m_4(\mu)-1}\,Z$; if this calculation is correct, the proof does not establish the stated amplitude for $\ell>2$.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central discovery is Theorem 2.3 together with Theorem 2.1: for fixed $\ell$, $\sqrt{N}[\tilde{p}_{\ell,N}(x) - \mathrm{He}_\ell(x)]$ converges in distribution to $\sqrt{m_4(\mu)-1}\,Z\,(\ell/2)\,\mathrm{He}_{\ell-2}(x)$, and the ordered roots satisfy $\sqrt{N}[z(\tilde{p}_{\ell,N}) - z(\mathrm{He}_\ell)] \Rightarrow \mathcal{N}(0,\Sigma^{(\ell),z})$ with a covariance matrix built from the fourth moment of $\mu$ and the Hermite moments. The second discovery, Theorem 2.6, replaces the Gaussian domain of attraction by an arbitrary Lévy triple $(c,\sigma^2,\nu)$: the limiting polynomial is $f(d/dx)x^\ell$, where $f$ is a random entire function $e^{-Yz - \sigma^2 z^2/2}\prod_{j\neq 0}(1-\alpha_j z)e^{\alpha_j z}$, expressed as a finite free convolution of a random monomial, a Hermite term, and a Poisson-point-process factor.

Load-bearing premise

The central limit theorems require that the empirical root moments of the original random polynomial fluctuate at scale $N^{-1/2}$ toward a Gaussian, which only holds when the root distribution has enough finite moments; if the fourth moment is infinite, the variance formulas in Theorems 2.1 and 2.3 fail.

Editorial extensions

If this is right

  • The fluctuation variance of the critical-point process is universal across root distributions sharing mean 0, variance 1, and fourth moment $m_4(\mu)$.
  • For a fixed $\ell$, the polynomial fluctuation limit is a Gaussian multiple of $\mathrm{He}_{\ell-2}$, so the zeros of the fluctuation field asymptotically sit at the roots of $\mathrm{He}_{\ell-2}$.
  • Removing the finite-variance assumption does not destroy the limit: the Hermite polynomial is replaced by a random Appell sequence determined by the Lévy triple of the root distribution.
  • In the Bernoulli/Poisson case the limiting polynomial is a Laguerre polynomial of random parameter, so discrete root models produce explicitly identifiable limits.
  • The same cumulant argument applies to other real-rooted random polynomial ensembles with understood root statistics, such as characteristic polynomials of random matrices.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The factorization in equation (2.9) suggests a finite-free analogue of a Gaussian perturbation: to first order, differentiation acts on the limiting noise by finite free convolution with a quadratic polynomial, which could make the CLT coefficient a combinatorial statistic of pairings rather than a separate calculation.
  • Because Theorem 2.6 depends only on the Lévy triple, the same limiting Appell sequence should arise from any real-rooted polynomial ensemble whose roots converge to the same infinitely divisible law; the paper does not fully spell this out, but its method appears ready for it.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 4 minor

Summary. The paper uses finite free cumulants to analyse the roots of random polynomials under repeated differentiation. Its first contribution is a central limit theorem for the fluctuations of the rescaled differentiated polynomial around the limiting Hermite polynomial, together with a CLT for the fluctuations of the ordered roots. Its second contribution is a limit theorem for the differentiated polynomials when the iid roots are not in the Gaussian domain of attraction, with the limit described through a random Appell sequence and an infinitely divisible distribution. The proofs are based on finite free cumulants, moment–cumulant formulas, the Delta method, and convergence of point processes.

Significance. The finite free cumulant framework is well chosen for this problem and the paper makes a useful connection between repeated differentiation of random polynomials and finite free probability. The second result, Theorem 2.6, is a substantial extension of the Hoskins–Steinerberger theorem to infinite-variance and Poisson-type root distributions, and the Appell-sequence description is natural. The paper also gives compact proofs and explicitly identifies the fourth-moment dependence of the fluctuations. However, the central polynomial CLT, Theorem 2.3, is stated with an incorrect coefficient, and the proof contains a scaling-factor error that is load-bearing for the same theorem. Because the main advertised result is false as stated, the manuscript cannot be accepted in its current form, although the corrected version appears to follow from the authors' own proof.

major comments (3)
  1. [Theorem 2.3 and Eq. (3.14)] The statement of Theorem 2.3 is contradicted by the proof. In Eq. (3.14), the Delta method gives the r-th term of the limit as (-1)^r * ell! / ((ell-2r)! (r-1)! 2^r) * sqrt(m4(mu)-1) Z x^{ell-2r}, and summing r=1,...,floor(ell/2) yields -ell(ell-1)/2 * sqrt(m4(mu)-1) Z He_{ell-2}(x), not (ell/2) sqrt(m4(mu)-1) Z He_{ell-2}(x). For ell=2, this changes the sign of the constant term. Thus Theorem 2.3 is false as stated; the coefficient in (2.6) and in the final line of (3.14) must be corrected to -ell(ell-1)/2.
  2. [Eq. (1.2)] The definition of the differentiation operator is normalized incorrectly. For a monic degree-N polynomial p_N, the (N-k)-th derivative has leading coefficient N!/k!, so the definition in (1.2) gives a leading coefficient of (N-k)!/k! for the resulting degree-k polynomial. After applying D_{\sqrt N} in (1.4), the leading coefficient is N^{k/2}(N-k)!/k!, which diverges as N grows. This contradicts Proposition 1.1 and every limit in the paper. The intended normalization appears to be k!/N! (or an equivalent convention that makes the derivative monic after the standard scaling); please correct and re-verify the subsequent formulas under that convention.
  3. [Eq. (3.12)] The scaling formula in the proof of Theorem 3.3 has the wrong exponent. Eq. (3.12) states kappa^ell_j(\tilde p_{ell,N}) = ell^{j-1} N^{j/2-1} kappa^N_j(p_N). Since D_{\sqrt N} scales roots by \sqrt N and Lemma 1.2 contributes a factor (ell/N)^{j-1}, the correct exponent is N^{1-j/2}. With the printed exponent, for j >= 3 the quantity \sqrt{N} kappa^ell_j would diverge for generic mu with nonzero kappa_j(mu), contradicting the conclusion of Theorem 3.3 that only the j=2 entry has a non-zero limiting variance. Please correct the exponent and adjust the surrounding argument.
minor comments (4)
  1. [Introduction] There is a typo in the sentence introducing Theorem 2.6: 'See Theorem, 2.6 below' should read 'See Theorem 2.6 below'.
  2. [Example 2.7(2)] The text contains a double comma in 'standard exponential random variables,, ε1'; this should be cleaned up.
  3. [Remark 2.4] The sentence 'the right-hand side of (2.1) vanishes at the roots of He_{ell-2}' is terse; (2.1) is the covariance matrix, and the intended statement is that the polynomial fluctuations are of smaller order at those points. Please clarify the wording.
  4. [References] Several cited works are preprints ([1], [6], [7]); if any have appeared or been updated, the reference entries should reflect that.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the proof chain uses independently established finite free cumulant lemmas; the algebraic sign error in (3.14) is a correctness issue, not circularity.

full rationale

The derivation is self-contained after importing standard, independently established results of finite free probability. The central CLT proof (Theorem 3.3) starts from the iid moments CLT (Lemma 3.1), the in-paper proof that finite free cumulants agree with free cumulants up to O(1/N) (Lemma 3.2), and the prior derivative-scaling lemma for finite free cumulants (Lemma 1.2). No parameter is fitted and no conclusion is assumed by construction. Corollary 3.4, Theorem 2.1, and Theorem 2.3 are then delta-method consequences of the cumulant CLT. The self-citations to [1], [2], [3], and [6] supply algebraic lemmas whose statements do not include Theorems 2.1, 2.3, or 2.6; they are parameter-free, externally checkable facts and therefore count as real evidence rather than circular support. The paper's own display (3.14) does contain an algebraic error: the displayed sum equals -ell(ell-1)/2 sqrt(m4(mu)-1) Z He_{ell-2}(x), not +(ell/2) sqrt(m4(mu)-1) Z He_{ell-2}(x), so Theorem 2.3 as stated is false for ell>2 and has the wrong sign even for ell=2. That is an internal mathematical mistake in the last equality, not a reduction of the theorem to its inputs, and it does not constitute circularity. No self-definitional, fitted-input, uniqueness-import, ansatz-smuggling, or renaming pattern is present.

Assumptions & free parameters 0 free parameters · 4 assumptions · 0 invented entities

The paper relies on standard tools from finite free probability, classical CLT, and Levy processes. No free parameters are fitted. No new entities are postulated; the random Appell polynomials are constructed from the given Levy triple, not invented ad hoc.

assumptions (4)
  • standard math Finite free cumulants are defined by (1.8) and satisfy Lemmas 1.2, 1.3, and 1.5 from [1] and [3].
    These are established results in finite free probability, cited and not rederived in the paper.
  • standard math The empirical moments of iid roots satisfy a CLT with covariance Sigma_{ij}=m_{i+j}-m_i m_j (Lemma 3.1).
    Standard multivariate CLT for sample moments, requiring finite moments up to order 2ell.
  • standard math The Delta method applies to the moment-to-root map and the moment-to-cumulant map.
    Used in the proofs of Theorems 2.1 and 2.3; the relevant Jacobians are invertible at the limiting Hermite polynomial.
  • standard math The Levy-Khintchine representation and the Ferguson-Klass representation of infinitely divisible laws without Gaussian components are valid.
    Used in the construction of the random entire function f in (2.17) and in Lemma 4.1.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Critical points of random polynomials and finite free cumulants." pith.science (2026). https://pith.science/paper/6VIHSX2M

@misc{pith2026250608910,
  author       = {Pith},
  title        = {Pith review of: Critical points of random polynomials and finite free cumulants},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/6VIHSX2M}},
  note         = {Machine review of arXiv:2506.08910}
}
read the original abstract

A result of Hoskins and Steinerberger [Int. Math. Res. Not., (13):9784-9809, 2022] states that repeatedly differentiating a random polynomials with independent and identically distributed mean zero and variance one roots will result, after an appropriate rescaling, in a Hermite polynomial. We use the theory of finite free probability to extend this result in two natural directions: (1) We prove central limit theorems for the fluctuations around these deterministic limits for the polynomials and their roots. (2) We consider a generalized version of the Hoskins and Steinerberger result by removing the finite second moment assumption from the roots. In this case the Hermite polynomials are replaced by a random Appell sequence conveniently described through finite free probability and an infinitely divisible distribution. We use finite free cumulants to provide compact proofs of our main results with little prerequisite knowledge of free probability required.

Figures

Figures reproduced from arXiv: 2506.08910 by the authors.

Figure 1
Figure 1. Histogram of 100 realizations of the roots of ˜pℓ,N illus￾trating the fluctuation described in Theorem 2.1. For this simula￾tion, N = 100, ℓ = 6, and we used µ ∼ N (0, 1). and hence by Lemma 1.5 p˜ℓ,N (x) ≈ Heℓ(x) ⊞ℓ x ℓ−2 [PITH_FULL_IMAGE:figures/full_fig_p006_1.png] view at source ↗
Figure 2
Figure 2. Plot of a sample of 15 trials of the random polynomials p˜8,200(x) − He8(x). Definition 2.5. Let {qN }∞ N=1 be a sequence of real rooted polynomials indexed by their degree of the form qN (x) = Y N j=1 (x − Xj,N ). (2.11) We say a sequence of polynomials {qN } is generated by the L´evy triple (c, σ2 , ν) if (1) For any N ∈ N, X1,N , X2,N , . . . , XN,N are iid random variables. (2) There exists σ ≥ 0, c ∈ R, and a m… view at source ↗

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

30 extracted references · 19 canonical work pages

  1. [1]

    Arizmendi, K

    O. Arizmendi, K. Fujie, D. Perales, and Y. Ueda. S-transform in Finite Free Probability. Available at arXiv:2408.09337, 2024. Preprint

    arXiv 2024

  2. [2]

    Arizmendi, J

    O. Arizmendi, J. Garza-Vargas, and D. Perales. Finite free cumulants: multiplicative convolu- tions, genus expansion and infinitesimal distributions. Trans. Amer. Math. Soc., 376(6):4383– 4420, 2023

    work page 2023

  3. [3]

    Arizmendi and D

    O. Arizmendi and D. Perales. Cumulants for finite free convolution. J. Combin. Theory Ser. A, 155:244–266, 2018

    2018

  4. [4]

    Assiotis and J

    T. Assiotis and J. Najnudel. The boundary of the orbital beta process. Mosc. Math. J. , 21(4):659–694, 2021

    work page 2021

  5. [5]

    Bercovici and V

    H. Bercovici and V. Pata. Stable laws and domains of attraction in free probability theory. Ann. of Math. (2) , 149(3):1023–1060, 1999. With an appendix by Philippe Biane

    work page 1999

  6. [6]

    Campbell

    A. Campbell. Free infinite divisibility, fractional convolution powers, and Appell polynomials. Available at arXiv:2412.20488, 2025. Preprint

    work page arXiv 2025

  7. [7]

    Campbell, S

    A. Campbell, S. O’Rourke, and D. Renfrew. Universality for roots of derivatives of entire functions via finite free probability. Available at arXiv:2410.06403, 2024. Preprint

    arXiv 2024

  8. [8]

    T. S. Ferguson and M. J. Klass. A representation of independent increment processes without Gaussian components. Ann. Math. Statist. , 43:1634–1643, 1972

    work page 1972

Show all 30 references
  1. [9]

    Gorin and V

    V. Gorin and V. Kleptsyn. Universal objects of the infinite beta random matrix theory. J. Eur. Math. Soc. (JEMS) , 26(9):3429–3496, 2024

    2024

  2. [10]

    Gorin and A

    V. Gorin and A. W. Marcus. Crystallization of random matrix orbits. Int. Math. Res. Not. IMRN, (3):883–913, 2020

    2020

  3. [11]

    Hall, C.-W

    B. Hall, C.-W. Ho, J. Jalowy, and Z. Kabluchko. Roots of polynomials under repeated differentiation and repeated applications of fractional differential operators. Available at arXiv:2312.14883, 2023. Preprint

    2023 arXiv

  4. [12]

    B. Hanin. Correlations and pairing between zeros and critical points of Gaussian random polynomials. Int. Math. Res. Not. IMRN , (2):381–421, 2015

    2015

  5. [13]

    B. Hanin. Pairing of zeros and critical points for random polynomials. Ann. Inst. Henri Poincar´ e Probab. Stat., 53(3):1498–1511, 2017

    2017

  6. [14]

    Hoskins and Z

    J. Hoskins and Z. Kabluchko. Dynamics of zeroes under repeated differentiation.Experimental Mathematics, 0(0):1–27, 2021

    2021

  7. [15]

    J. G. Hoskins and S. Steinerberger. A semicircle law for derivatives of random polynomials. Int. Math. Res. Not. IMRN , (13):9784–9809, 2022

    2022

  8. [16]

    Kabluchko and H

    Z. Kabluchko and H. Seidel. Distances between zeroes and critical points for random poly- nomials with i.i.d. zeroes. Electron. J. Probab., 24:Paper No. 34, 25, 2019

    2019

  9. [17]

    Kallenberg

    O. Kallenberg. Foundations of modern probability . Probability and its Applications (New York). Springer-Verlag, New York, second edition, 2002

    2002

  10. [18]

    A. W. Marcus. Polynomial convolutions and (finite) free probability. Available at arXiv:2108.07054, 2021

    2021 arXiv

  11. [19]

    A. W. Marcus, D. A. Spielman, and N. Srivastava. Interlacing families I: Bipartite Ramanujan graphs of all degrees. Ann. of Math. (2) , 182(1):307–325, 2015

    2015

  12. [20]

    A. W. Marcus, D. A. Spielman, and N. Srivastava. Interlacing families II: Mixed characteristic polynomials and the Kadison-Singer problem. Ann. of Math. (2) , 182(1):327–350, 2015

    2015

  13. [21]

    A. W. Marcus, D. A. Spielman, and N. Srivastava. Finite free convolutions of polynomials. Probab. Theory Related Fields, 182(3-4):807–848, 2022

    2022

  14. [22]

    Michelen and X.-T

    M. Michelen and X.-T. Vu. Almost sure behavior of the zeros of iterated derivatives of random polynomials. Electron. Commun. Probab., 29:Paper No. 27, 10, 2024. 14 OCTA VIO ARIZMENDI, ANDREW CAMPBELL, AND KATSUNORI FUJIE

    2024

  15. [23]

    Michelen and X.-T

    M. Michelen and X.-T. Vu. Zeros of a growing number of derivatives of random polynomials with independent roots. Proc. Amer. Math. Soc. , 152(6):2683–2696, 2024

    2024

  16. [24]

    J. A. Mingo and R. Speicher. Free probability and random matrices , volume 35 of Fields Institute Monographs . Springer, New York; Fields Institute for Research in Mathematical Sciences, Toronto, ON, 2017

    2017

  17. [25]

    O’Rourke and N

    S. O’Rourke and N. Williams. Pairing between zeros and critical points of random polynomials with independent roots. Trans. Amer. Math. Soc. , 371(4):2343–2381, 2019

    2019

  18. [26]

    O’Rourke and N

    S. O’Rourke and N. Williams. On the local pairing behavior of critical points and roots of random polynomials. Electron. J. Probab., 25:Paper No. 100, 68, 2020

    2020

  19. [27]

    Pemantle and I

    R. Pemantle and I. Rivin. The distribution of zeros of the derivative of a random polynomial. In Advances in combinatorics, pages 259–273. Springer, Heidelberg, 2013

    2013

  20. [28]

    Pemantle and S

    R. Pemantle and S. Subramanian. Zeros of a random analytic function approach perfect spacing under repeated differentiation. Trans. Amer. Math. Soc. , 369(12):8743–8764, 2017

    2017

  21. [29]

    S. I. Resnick. Heavy-tail phenomena . Springer Series in Operations Research and Financial Engineering. Springer, New York, 2007. Probabilistic and statistical modeling

    2007

  22. [30]

    Shlyakhtenko and T

    D. Shlyakhtenko and T. Tao. Fractional free convolution powers. Indiana University Mathe- matics Journal , 71(6):2551–2594, 2022. Centro de Investigacion en Matem ´aticas. A.C., Jalisco S/N, Col. V alenciana CP: 36023 Guanajuato, Gto, Mexico Email address : octavius@cimat.mx I...

    2022

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.