Pith. sign in

REVIEW 5 minor 2 cited by

Dynamics of rotationally invariant polynomial root sets under iterated differentiations

T0 review · 0 major / 5 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 For grid-sampled polynomials, repeated differentiation has an explicit limiting measure.

desk verdict Solid paper; the stress-test's Lemma 4.6 concern is a misreading of a typo, and the main theorem stands as a genuinely new result on the Hoskins-Kabluchko PDE. read the letter →

arxiv 2506.06263 v2 pith:LI7VBPW4 submitted 2025-06-06 math.PR math.APmath.DS

André Galligo , Joseph Najnudel , Truong Vu This is my paper
classification math.PRmath.APmath.DS MSC 30C1560B1060F05
keywords repeateddifferentiationpolynomialrootsempiricalmeasuresrotationallyinvariantmeasurequantilefunctionstransportPDEinterlacingrandompolynomials
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

The paper studies what happens to the roots of a polynomial when one differentiates it many times: specifically, the roots of the $\lfloor tN\rfloor$-th derivative of a degree-$N$ polynomial. Its central claim is that for a special but natural sampling of a rotationally invariant measure, this dynamics has a simple provable limit. The roots are placed on $n$ concentric circles, $m_n$ equally spaced points per circle, and the radii form an empirical measure converging to the radial part $\nu_0$ of the starting measure. Provided $m_n/(n\log n)\to\infty$, the empirical measure of the derivative roots converges to $\mu_t=\nu_t\otimes \mathrm{unif}$, where $\nu_t$ is described by the quantile identity $q_{(1-t)\nu_t}(x)=\frac{x}{x+t}q_{\nu_0}(x+t)$. A sympathetic reader should care because this is a rigorous, two-dimensional instance of a conjectured transport equation, and it shows explicitly how the choice of sampling can change the limit.

What carries the argument

The engine is a reduction of the two-dimensional problem to a one-dimensional radius flow. Lemma 3.1 shows that $P_{n,m}(z)=\prod_{j=1}^n(z^m-r_j^m)$ satisfies $P'_{n,m}(z)=z^{m-1}S(z^m)$ with $S$ a real-rooted polynomial whose positive roots interlace with the $r_j^m$, so one differentiation keeps the root set on $n$ circles with the same angular grid and reduces the radius list by one entry. Iterating $m$ differentiations reduces $n$ to $n-1$ circles. The quantitative control is provided by Lemmas 4.3 and 4.4, which bound the motion of the interlacing roots in terms of $\alpha=\max_j r_j/r_{j+1}$; the condition $m_n/(n\log n)\to\infty$ makes the error factor $\gamma=e^{3n\log n/m_n}$ tend to $1$, so the bounds close. The limit is read from the empirical quantile function of the radii, not from solving the PDE.

What would settle it

Take a smooth initial radial measure $\nu_0$, such as uniform on $[1,2]$, set $m_n=n\log^2 n$, and numerically compute the empirical radial distribution of the roots of the $\lfloor n m_n t\rfloor$-th derivative of $\prod_{j=1}^n(z^{m_n}-(r_j^{(n)})^{m_n})$; any systematic discrepancy from the law of $(1-t/V_t)q_{\nu_0}(V_t)$ as $n$ grows would refute Theorem 4.1, while agreement in the unproved regime $m_n\sim n$ would show the growth condition is only an artifact.

Watch

Extended reading notes

Core claim

The discovery is a provable variant of the derivative-root conjecture for rotationally invariant measures formulated in [10]. For $P_{n,m_n}(z)=\prod_{j=1}^n(z^{m_n}-(r_j^{(n)})^{m_n})$, with $\frac1n\sum_j\delta_{r_j^{(n)}}\to\nu_0$ and $m_n/(n\log n)\to\infty$, the empirical measure of the roots of the $\lfloor n m_n t\rfloor$-th derivative converges in probability to $\mu_t=\nu_t\otimes\mathrm{unif}$, where $\nu_t$ is the law of $(1-t/V_t)\,q_{\nu_0}(V_t)$ with $V_t$ uniform on $[t,1]$. Equivalently, the quantile functions satisfy $q_{(1-t)\nu_t}(x)=\frac{x}{x+t}q_{\nu_0}(x+t)$ for $x\in[0,1-t]$, which is the inversion identity conjectured for the general process. If $\nu_0$ has a continuous positive density, the distribution function $\Psi_t$ of $(1-t)\nu_t$ solves the PDE $\partial_t\Psi_t = x(\partial_x\Psi_t)/\Psi_t - 1$, and if the density is also differentiable, the density itself solves the transport equation $\partial_t\psi=\partial_x(\psi/(\frac1x\int_0^x\psi))$. The proof does not solve the PDE directly; it derives the limiting quantile relation from a one-dimensional radius flow and then differentiates it.

Load-bearing premise

The load-bearing assumption is that the initial roots sit exactly on a perfect grid, $n$ concentric circles with $m_n$ equally spaced points on each, with $m_n/(n\log n)\to\infty$; if $m_n$ grows only as fast as $n$, the error estimates stop closing and the proof gives nothing, though the paper conjectures the result still holds in that regime.

Editorial extensions

If this is right

  • For any compactly supported $\nu_0$, the limiting radial measure is explicitly computable from the quantile function of the initial radii, so predictions about derivative-root flows can be made without solving a PDE.
  • Under the stated regularity, the PDEs that were only conjectured for general i.i.d. samplings are rigorously obtained for the grid sampling, giving a concrete class of examples where the transport equation is valid.
  • The limit is sampling-dependent: when $\nu_0=\delta_1$, the grid construction yields a nontrivial density on $[0,1-t]$, while a roots-of-unity sampling makes every iterated-derivative root equal to zero, so no universal sampling-independent statement can hold.
  • On the real line, the situation is different: Proposition 2.2 shows that any two root configurations with the same limiting empirical measure have the same derivative-root limit, because the one-dimensional root map is monotone and Lipschitz in the L\'evy metric.
  • The dynamics preserves the $m$-fold rotational symmetry at every step, reducing the whole process to a deterministic map on ordered positive radii.

Reading between the lines

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

  • The role of the $m_n/(n\log n)$ condition looks technical rather than intrinsic: the estimates lose control only through the factor $e^{3n\log n/m_n}$, and the paper's own conjecture that $m_n$ proportional to $n$ suffices could likely be attacked by exploiting cancellations between the positive and negative terms in the interlacing equation.
  • A plausible extension is that any sampling whose angular points become equidistributed on a growing set of circles, not just exact roots of unity, will produce the same $\nu_t$, because the proof only needs the angular measure to converge to uniform; this would give a bridge to i.i.d. angular samples without a mean-field argument.
  • The $\nu_0=\delta_1$ comparison suggests that the arguments' arithmetic structure, not just their uniformity, can change the limit; a useful test is to perturb the grid angles by independent noise of size $1/m$ and see at what noise level the dynamics switches from the grid limit to the roots-of-unity collapse.
  • The circular-derivative example in Section 6.4 conserves the geometric mean of the radii, whereas ordinary differentiation does not; tracking this quantity for the grid model might yield a Lyapunov function and quantitative rates for the convergence to $\nu_t$.
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

0 major / 5 minor

Summary. The paper studies the behavior of root sets of polynomials under iterated differentiation in a two-dimensional, rotationally invariant setting. The authors introduce a special grid sampling: n concentric circles, each carrying m equally spaced points, with radii r_j^(n) chosen so that their empirical measure converges to a compactly supported measure ν0 on R_+. For N=nm the polynomial is P_{n,m}(z)=∏_{j=1}^n (z^m - (r_j^(n))^m). The main result, Theorem 4.1, states that if m_n/(n log n) tends to infinity, then the empirical measure of the roots of the ⌊n m_n t⌋-th derivative of P_{n,m_n} converges to μ_t = ν_t ⊗ unif, where ν_t is the distribution of (1 - t/V_t) q_{ν0}(V_t) with V_t uniform on [t,1] and q_{ν0} the quantile function of ν0. The authors derive the quantile relation (7), and under increasing regularity assumptions they obtain the distribution-function PDE (8) and the density PDE (9), matching the Hoskins-Kabluchko reformulation of the O'Rourke-Steinerberger conjecture. The paper also proves independently useful one-dimensional results: Theorem 2.1 establishes monotonicity and an n/(n-1)-Lipschitz property for roots of derivatives of real-rooted polynomials with positive weights, and Proposition 2.2 shows sampling-independence of the limiting measure in the real case. The growth condition m_n/(n log n) → ∞ is explicitly flagged by the authors as restrictive, and they conjecture that it can be relaxed to m of order n.

Significance. If correct, the paper gives the first rigorous proof of the conjectured PDE dynamics in a nontrivial two-dimensional model, albeit for a structured sampling rather than an i.i.d. one. The proof is essentially self-contained and transparent: the limit measure is computed explicitly from the quantile function of ν0, and the PDEs are derived downstream from the quantile relation rather than assumed. The paper's strengths include detailed lemmas (Lemmas 4.3-4.6), a clear statement of the resulting measure via (5) and (7), and an honest discussion of the limitations of the main theorem, including numerical evidence for the m≤n regime. The work does not settle the original i.i.d. conjecture, but it constitutes substantial progress and clarifies the role of the sampling in two dimensions, in contrast to the sampling-independent one-dimensional case. The main caveat is the strong growth assumption on m_n, which the authors explicitly acknowledge and conjecture can be relaxed.

minor comments (5)
  1. [Lemma 4.6] The definition of γ is typeset ambiguously, apparently as γ = exp(3 m_n^{-1} n log n). If that were the definition, the bound γ^{j+ℓ} ≤ exp(3 n log n/m_n) would be false, since γ^{j+ℓ} ≤ γ^n = exp(3 n^2 log n/m_n). However, the subsequent identities γ^{-m_n} = exp(-3 log n) and m_n log γ = 3 log n show that the intended definition is γ = exp(3 log n/m_n), for which γ^n = exp(3 n log n/m_n) and the estimates close. Please rewrite the display with explicit braces, e.g., γ_n = exp(3 log n / m_n), to remove this ambiguity.
  2. [Section 4.2] The sentence 'By moving a negligible part of the measure' is informal; the preceding argument should state quantitatively that the δ0 contribution has mass O(1/n) and that the angular averaging moves points by O((A+1)/m_n), so the Lévy-Prokhorov distance indeed tends to zero.
  3. [Lemma 4.6] The notation r_j^{(n,m,ℓ-q/m)} uses a non-integer third argument without a formal definition; please define it explicitly, since it denotes the radii obtained after ℓm-q differentiations and is central to the proof.
  4. [Section 5 and abstract] The growth condition is written informally as 'm/(n log n)'; please use m_n/(n log n) with parentheses to match Theorem 4.1 and avoid ambiguity.
  5. [Throughout] The spelling of 'Lévy' is inconsistent; please unify the accents, and check for similar minor typographical inconsistencies in the references and figure captions.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the limiting measure is proved from explicit root-motion estimates and quantile convergence, with the conjectured PDEs derived downstream rather than assumed.

full rationale

The paper is self-contained in its main derivation. Theorem 4.1 is proved from Lemma 4.6, which bounds the radii of roots of the iterated derivative by explicit factors involving the initial radii, and then from a quantitative convergence argument for the empirical measures. The limiting measure νt is defined directly from the quantile function of ν0, and the PDEs (8) and (9) are derived only after this convergence, from the quantile identity (7); they are not imposed as inputs. The conjectures of Hoskins–Kabluchko [10], O'Rourke–Steinerberger [21], and Campbell–O'Rourke–Renfrew [5] are used as motivation and as a benchmark, not as premises in the proof. The only self-citation that appears, reference [8], is listed among related prior work on derivative root dynamics and is not load-bearing for the argument. The paper even explicitly notes its restriction m_n/(n log n) → ∞ and offers it as a topic for future improvement, which further shows the theorem is not tailored circularly to a predetermined conclusion. A possible issue in Lemma 4.6 regarding the exponent in the factor γ would be a correctness concern, not a circularity, and does not affect this assessment.

Assumptions & free parameters 0 free parameters · 6 assumptions · 0 invented entities

All parameters, including n, m, t, and nu0, are inputs or construction choices, not fitted values. The paper introduces no new particles, fields, or unobserved entities. Its only nonstandard assumption is the class of grid samplings, which is explicitly constructed and not hidden.

assumptions (6)
  • standard math Standard facts about weak convergence, Levy-Prokhorov distance, and quantile functions of probability measures on R+.
    Used throughout the proof of Theorem 4.1, especially in the final empirical measure convergence argument in Section 4.2.
  • standard math Interlacing and monotonicity of roots of derivative-type polynomials, and strict monotonicity of the rational function sum w_j/(z-z_j) between consecutive roots.
    Basis for Theorem 2.1 and Lemmas 4.2 through 4.4.
  • standard math The implicit function theorem and inverse function theorem on open sets.
    Used in Section 4 to go from the quantile relation (7) and (14) to differentiability of Psi_t and the PDEs (8) and (9).
  • domain assumption The radial grid sampling model: P_{n,m}(z) = product over j of (z^m - r_j^m), with r_j an increasing sequence whose empirical measure converges to compactly supported nu0.
    This is the specific class of initial root sets for which the theorem is proved in Section 3 and Theorem 4.1. It is not the original i.i.d. sampling.
  • domain assumption Growth condition m_n/(n log n) tending to infinity.
    Needed in Lemma 4.6 to make gamma = exp(3 n log n / m_n) tend to 1 and the bounds epsilon_n tend to 0; the paper states in Section 5 that relaxing it is open.
  • domain assumption For the PDE part, nu0 is absolutely continuous with continuous strictly positive density, and C^1 for the second PDE.
    Assumed in the second and third parts of Theorem 4.1 to ensure Psi_t is a bijection and differentiable, and that the two-variable PDE has classical meaning.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Dynamics of rotationally invariant polynomial root sets under iterated differentiations." pith.science (2026). https://pith.science/paper/LI7VBPW4

@misc{pith2026250606263,
  author       = {Pith},
  title        = {Pith review of: Dynamics of rotationally invariant polynomial root sets under iterated differentiations},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/LI7VBPW4}},
  note         = {Machine review of arXiv:2506.06263}
}
abstract

We associate to an $N$-sample of a given rotationally invariant probability measure $\mu_0$ with compact support in the complex plane, a polynomial $P_N$ with roots given by the sample. Then, for $t \in (0,1)$, we consider the empirical measure $\mu_t^{N}$ associated to the root set of the $\lfloor t N\rfloor$-th derivative of $P_N$. A question posed by O'Rourke and Steinerberger [21], reformulated as a conjecture by Hoskins and Kabluchko [10], and recently reaffirmed by Campbell, O'Rourke and Renfrew [5], states that under suitable conditions of regularity on $\mu_0$, for an i.i.d. sample, $\mu_t^{N}$ converges to a rotationally invariant probability measure $\mu_t$ when $N$ tends to infinity, and that $(1-t)\mu_t$ has a radial density $x \mapsto \psi(x,t)$ satisfying the following partial differential equation: \begin{equation} \label{PDErotational} \frac{ \partial \psi(x,t) }{\partial t} = \frac{ \partial}{\partial x} \left( \frac{ \psi(x,t) }{ \frac{1}{x} \int_0^x \psi(y,t) dy } \right). \end{equation} In [10], this equation is reformulated as an equation on the distribution function $\Psi_t$ of the radial part of $(1-t) \mu_t$: \begin{equation} \label{equationPsixtabstract} \frac{\partial \Psi_t (x)}{\partial t} = x \frac{\frac{\partial \Psi_t (x)}{\partial x} } {\Psi_t(x)} - 1. \end{equation} Restricting our study to a specific family of $N$-samplings, we are able to prove a variant of the conjecture above. We also emphasize the important differences between the two-dimensional setting and the one-dimensional setting, illustrated in our Theorem 2.1.

Figures

Figures reproduced from arXiv: 2506.06263 by the authors.

Figure 1
Figure 1. Root sets of P, P (2) , P (15), and P (45) [PITH_FULL_IMAGE:figures/full_fig_p006_1.png] view at source ↗
Figure 2
Figure 2. m = 20, with n = 20 at left and n = 40 at right after 100 differentiations. 6.2. Very small perturbations. We consider small angular perturbations of order 1/m, on our model of sample with m = n = 10, [PITH_FULL_IMAGE:figures/full_fig_p019_2.png] view at source ↗
Figure 3
Figure 3. m = 10, n = 10, 0 and 30 differentiations on an initially perturbed sample. 6.3. Uniform distribution on each circle. In this subsection, we sample the uniform distribu￾tion on n circles by n sets of m i.i.d. uniform random variables on each circle, so in total mn random variables taking values in [0, 2π) for the argument of the points. The left of [PITH_FULL_IMAGE:figures/full_fig_p019_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: m = 20, n = 20, i.i.d. uniform distribution of the arguments. illustrate this behavior with the following [PITH_FULL_IMAGE:figures/full_fig_p020_4.png]
Figure 5
Figure 5. Figure 5: Action of 0, 2, 30, and 200 circular differentiations. We notice an intricate dynamics. This is due to the fact that even if at each differentiation we have a polynomial of the form QN j=1(z m − z m j ) multiplied by a constant, in this setting these zj may become real…

Discussion (0). Sign in to comment.

Forward citations

Cited by 2 Pith papers

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

  1. Repeated differentiation of deterministic polynomials with asymptotically radial root distributions

    math.PR 2026-07 conditional novelty 6.0 of 10

    Roots of polynomials with asymptotically radial root sets move inward under repeated differentiation according to an explicit quantile formula; this paper simplifies the proof and extends it to z^a(d/dz)^b and to fixed m.

  2. Root Dynamics of Differentiated Polynomials with Rotationally Invariant Structure

    math.PR 2026-07 accept novelty 6.0 of 10

    Empirical root measures of structured rotationally invariant polynomials converge under differentiation as soon as m_n / log n → ∞, via sharper single-step root-magnitude bounds.

Reference graph

Works this paper leans on

23 extracted references · 20 canonical work pages · cited by 2 Pith papers

  1. [1]

    Alazard, O

    T. Alazard, O. Lazar, and Q. H. Nguyen. On the dynamics of the roots of polynomials under differentiation. J. Math. Pures Appl. (9), 162:1–22, 2022

    work page 2022

  2. [2]

    Arizmendi, K

    O. Arizmendi, K. Fujie, and Y. Ueda. New combinatorial identity for the set of partitions and limit theorems in finite free probability theory. Int. Math. Res. Not. IMRN, (13):10450–10484, 2024

    work page 2024

  3. [3]

    Arizmendi, J

    O. Arizmendi, J. Garza-Vargas, and D. Perales. Finite free cumulants: Multiplicative convolutions, genus ex- pansion and infinitesimal distributions, 2024

    work page 2024

  4. [4]

    S.-S. Byun, J. Lee, and T. R. Reddy. Zeros of random polynomials and their higher derivatives. Trans. Amer. Math. Soc., 375(9):6311–6335, 2022

    work page 2022

  5. [5]

    Campbell, S

    A. Campbell, S. O’Rourke, and D. Renfrew. The fractional free convolution of R-diagonal elements and random polynomials under repeated differentiation. International Mathematics Research Notices, 2024(13):10189–10218, 2024

    work page 2024

  6. [6]

    Feng and D

    R. Feng and D. Yao. Zeros of repeated derivatives of random polynomials. Analysis & PDE, 12(6):1489–1512, 2019

    work page 2019

  7. [7]

    A. Galligo. Modeling complex root motion of real random polynomials under differentiation. In ISSAC ’22— Proceedings of the 2022 International Symposium on Symbolic and Algebraic Computation, pages 245–253. ACM, New York, [2022] ©2022

    work page 2022

  8. [8]

    Dynamics of roots of randomized derivative polynomials

    A. Galligo and J. Najnudel. Dynamics of roots of randomized derivative polynomials. arXiv preprint arXiv:2503.06650, 2025

    work page Pith review arXiv 2025

Show all 23 references
  1. [9]

    Galligo, J

    A. Galligo, J. Najnudel, and T. Vu. Anti-concentration applied to roots of randomized derivatives of polynomials. Electronic Journal of Probability, 29:1–20, 2024

    2024

  2. [10]

    Hoskins and Z

    J. Hoskins and Z. Kabluchko. Dynamics of zeroes under repeated differentiation. Exp. Math., pages 1–27, 2021

    2021

  3. [11]

    J. G. Hoskins and S. Steinerberger. A semicircle law for derivatives of random polynomials. Int. Math. Res. Notices, 2022(13):9784–9809, 2022

    2022

  4. [12]

    C. P. Hughes and A. Nikeghbali. The zeros of random polynomials cluster uniformly near the unit circle. Compos. Math., 144(3):734–746, 2008

    2008

  5. [13]

    Ibragimov and D

    I. Ibragimov and D. Zaporozhets. On distribution of zeros of random polynomials in complex plane. In Prokhorov and Contemporary Probability Theory: In Honor of Yuri V. Prokhorov, pages 303–323. Springer, 2012

    2012

  6. [14]

    Jalowy, Z

    J. Jalowy, Z. Kabluchko, and A. Marynych. Zeros and exponential profiles of polynomials i: Limit distributions, finite free convolutions and repeated differentiation. arXiv preprint arXiv:2504.11593, 2025

    2025 arXiv

  7. [15]

    Kabluchko

    Z. Kabluchko. Critical points of random polynomials with independent identically distributed roots. Proc. Amer. Math. Soc., 143(2):695–702, 2015

    2015

  8. [16]

    Kabluchko

    Z. Kabluchko. Repeated differentiation and free unitary Poisson process. arXiv preprint arXiv:2112.14729, 2021

    2021 arXiv

  9. [17]

    Kabluchko and D

    Z. Kabluchko and D. Zaporozhets. Asymptotic distribution of complex zeros of random analytic functions.Annals of Probability, 42(4):1374–1395, 2014

    2014

  10. [18]

    Kiselev and C

    A. Kiselev and C. Tan. The flow of polynomial roots under differentiation. Ann. PDE, 8(2):Paper No. 16, 69, 2022

    2022

  11. [19]

    Michelen and X.-T

    M. Michelen and X.-T. Vu. Almost sure behavior of the zeros of iterated derivatives of random polynomials. Electronic Communications in Probability, 29:1–10, 2024

    2024

  12. [20]

    Michelen and X.-T

    M. Michelen and X.-T. Vu. Zeros of a growing number of derivatives of random polynomials with independent roots. Proceedings of the American Mathematical Society, 152(06):2683–2696, 2024

    2024

  13. [21]

    O’Rourke and S

    S. O’Rourke and S. Steinerberger. A nonlocal transport equation modeling complex roots of polynomials under differentiation. Proc. Amer. Math. Soc., 149(4):1581–1592, 2021

    2021

  14. [22]

    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

  15. [23]

    Steinerberger

    S. Steinerberger. A nonlocal transport equation describing roots of polynomials under differentiation. Proc. Amer. Math. Soc., 147(11):4733–4744, 2019. Universit´e Cˆote d’Azur, Laboratoire de Math´ematiques ”J.A. Dieudonn´e”, UMR CNRS 7351, Inria, Nice, France. Email address:...

    2019

Pith tools

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