Pith. sign in

REVIEW 3 major objections 4 minor 24 references

Repeated differentiation of deterministic polynomials with asymptotically radial root distributions

T0 review · 3 major / 4 minor · reviewed 2026-08-01 · deepseek-v4-flash

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

Pith's one-line read This paper proves an explicit limiting root law for repeatedly differentiated polynomials with radially symmetric roots, using a sharp coefficient-ratio bound, and extends it to all operators z^a(d/dz)^b.

desk verdict Worth engaging with: a genuinely simpler proof of the GNV/NV theorem plus a solid extension to z^a(d/dz)^b, but with a real proof gap when the limiting measure has an atom at the origin. read the letter →

arxiv 2607.16954 v1 pith:J7M6UFQF submitted 2026-07-18 math.PR math-phmath.MPmath.OA

Brian C. Hall , Daniel Perales This is my paper
classification math.PRmath-phmath.MPmath.OA MSC 30C1546L54
keywords repeateddifferentiationradialrootdistributionquantilefunctioncoefficientratiosfinitefreeprobabilitymultiplicativeconvolutionpolynomialrootsdifferentialoperators
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 establishes an explicit formula for the limiting root distribution when a deterministic polynomial of the form p(z^m), whose roots lie on radially symmetric rays, is repeatedly differentiated — and, more generally, when any operator z^a(d/dz)^b is repeatedly applied. Assuming the number m of rays grows faster than log n and the initial root distribution converges to a radial measure μ0, the authors show that after ⌊nt⌋ steps the empirical root measure converges to a measure whose radial quantile function is exactly Q_μ0(α)(1−t(b−a)/α)^{b/(b−a)} (with an exponential factor when a=b). The entire derivation rests on a simple, sharp bound comparing the k-th largest root to the ratio of consecutive coefficients. This yields a substantially shorter proof of earlier results and directly gives the weaker growth condition m≫log n.

What carries the argument

The engine is a sharp coefficient-ratio bound (Theorem 1.3): for a degree-n polynomial with real positive roots and rescaled coefficients e_k, the k-th largest root λ_k satisfies (1/k)e_k/e_{k−1} ≤ λ_k ≤ (n−k+1)e_k/e_{k−1}. Because applying z^a(d/dz)^b to p(z^m) multiplies the e_k's by explicit falling-factorial products, the bound becomes a pair of upper and lower estimates that squeeze together after taking the m-th root and letting n,m→∞. A second ingredient, Lemma 3.2, guarantees that the real-non-negative-root property survives each application, so the bound can be iterated.

What would settle it

Take p_n(x)=∏_{k=1}^n (x−(k/n)^{M_n}) with M_n growing like log²n, compute Q_{n,1/2} for the repeated-differentiation case, and compare the empirical radial quantile at α=0.75 to Q_μ0(0.75)(1−0.5/0.75) as n→∞; a sustained mismatch would disprove the formula.

Watch

Extended reading notes

Core claim

The central discovery is that the evolution of radial root counts under repeated differential operators is governed entirely by a transfer formula for quantile functions: with αmin=max(0,t(b−a)), one has Q_σt(α)=0 for α≤αmin and Q_σt(α)=Q_μ0(α)(1−t(b−a)/α)^{b/(b−a)} for α>αmin (or Q_μ0(α)e^{−at/α} when a=b). The authors prove this for every sequence P_n(z)=p_n(z^{m_n}) with p_n having real non-negative roots and m_n/log n→∞, whenever the empirical root measure of P_n converges to a compactly supported radial μ0. They also identify, in the fixed- m regime, the limit as a free multiplicative convolution with a Bernoulli measure — so the same evolution is exactly a free-probability operation.

Load-bearing premise

The entire argument leans on every relevant polynomial having only real, non-negative roots — the starting p_n and every polynomial produced along the way — because the coefficient-ratio bound and its iteration have no analogue otherwise.

Editorial extensions

If this is right

  • The limiting root law is universal within this class: it depends only on the initial radial quantile function, not on further details of the polynomial sequence.
  • The same quantile-transfer formula covers all operators z^a(d/dz)^b, unifying the degree-decreasing (a<b), degree-preserving (a=b), and degree-increasing (a>b) cases.
  • The coefficient-ratio bound is sharp and of independent interest; it converts the whole problem into coefficient bookkeeping, which is why the proof is much shorter than previous ones.
  • In the fixed-m regime, the limit equals γ0⊠(B_t)^{⊠m} — the initial radial measure freely multiplied by a Bernoulli convolution — connecting deterministic root evolution to free probability.

Reading between the lines

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

  • If the same coefficient-ratio control holds for sequences that only approximate the p(z^m) form, the quantile-transfer formula might hold for a much wider deterministic class — a testable extension the paper does not pursue.
  • The explicit root-trajectory picture (each root moves radially inward on the curve z_j(1−t(b−a)/α)^{b/(b−a)}) suggests a deterministic transport map; the rates of convergence to this map as n→∞ are not addressed and could be examined numerically.
  • For fixed m, the free-convolution description and, for m→∞, the differential-operator formula are two limits of one process; interpolating between them may give a one-parameter family of evolution laws worth exploring.
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 studies repeated application of differential operators of the form z^a(d/dz)^b to deterministic polynomials P_n(z)=p_n(z^{m_n}), where p_n has real non-negative roots and m_n/log n → ∞. Theorem 1.1 states that if the empirical root measure of P_n converges to a compactly supported radial measure μ_0, then the empirical root measure of Q_{n,t}(z)=z^{(b-a)m_n⌊nt⌋}(z^a(d/dz)^b)^{m_n⌊nt⌋}P_n(z) converges to a measure σ_t whose radial quantile function is given by the explicit formula (1.4). The proof strategy is a substantial simplification of earlier work by Galligo–Najnudel–Vu and Najnudel–Vu: it is based on an elementary root–coefficient bound (Theorem 1.3) plus coefficient comparison and squeeze arguments. The paper also treats the fixed-m case through free multiplicative convolution, identifying the limit as γ_0 ⊠ (ν_{a,b;t})^⊠m. The central derivation is clear and largely self-contained, with the fixed-m result relying on the external black-box [AFPU26, Theorem 1.1].

Significance. If the result holds, the paper makes a useful contribution: it provides a much simpler proof of a known theorem in the case a=0,b=1, extends it to a natural family of differential operators, and identifies the fixed-m limit in free-probabilistic terms. The coefficient-ratio bound in Theorem 1.3 is elegant and of independent interest, and the overall structure (coefficient estimates → root bounds → squeeze) is transparent and reproducible. The paper is honest about its reliance on existing results in the fixed-m case and on [HHJK26] for consistency checks. The main mathematical content appears sound after the corrections below; the issues are mostly in auxiliary lemmas and in an omitted proof case rather than in the core asymptotic argument.

major comments (3)
  1. [§4.1, c_q formula before Lemma 4.2] The displayed formula for the zero multiplicity c_q is incorrect when p itself has roots at 0. For example, take a=0, b=1, m=5, l=2, and p(w)=w(w−1)(w−2), so c_p=1 and Δ=−1. The paper gives c_q = c_p + l|Δ| = 3, but a direct computation gives q(w)=A w^3 + B w^2, so c_q=2 = max(l|Δ|, c_p). A similar counterexample holds for Δ≥0: for a=b=1 and p(w)=w, one obtains c_q=c_p, not c_p+1. The correct formula in general is c_q = max(l|Δ|, c_p) for Δ<0 and c_q = c_p for Δ≥0 (up to the convention for the prefactor z^{-lmΔ}). This is not merely a typo: the range k≤n−c_q in Lemma 4.3 and the argument that k>n−c_q forces λ_k(q)=0 in Theorem 1.1 depend on this quantity. The proof is asymptotically repairable because c_p is fixed while l|Δ|/n→t|Δ|, but the statement as written is false and should be corrected, and the proof of Lemma 4.3 should be rephrased using the correct multiplicity.
  2. [§4.2, proof of Theorem 1.1, case α<α_min] The proof says 'The analysis of the case α < α_min is similar to the corresponding case in the proof of Theorem 3.7 and is omitted.' This is not fully satisfactory, especially when the limiting measure μ_0 has an atom at the origin. In that situation, for α<β (β the atom mass), one must show that the k-th largest root of q^{⟨1/m⟩} tends to zero. For α<α_min, k/n → 1−α > 1−α_min, so k exceeds n−c_q asymptotically, and the conclusion follows from the zero multiplicity of q; but this needs to be stated explicitly, and the boundary point α=α_min, where k/n is comparable to 1−α_min, requires a separate limiting argument. The formula (1.4) is plausibly correct at those points, but the written proof does not establish it. I request an explicit treatment of the case α≤α_min.
  3. [§3.1, proof of Lemma 3.2] There is a sign error in the proof of Lemma 3.2. After noting that ilde p(0)>0 and ilde p(λ_n)<0, the text states that because p crosses from negative to positive at λ_{n−1}, one has ilde p(λ_{n−1})<0. In fact p'(λ_{n−1})>0 and λ_{n−1}>0, so ilde p(λ_{n−1}) = m λ_{n−1} p'(λ_{n−1}) > 0. The conclusion that the signs alternate over the n intervals remains correct after this correction: the sign at λ_{n−1} is positive, then negative at λ_{n−2}, and so on. This is a local flaw in an otherwise standard argument, but it should be fixed.
minor comments (4)
  1. [§3.2, Lemma 3.4] In equation (3.2), the product is written with factors (n−k−l+i/m)/(n−k+i/m). After the cancelation of m, this is correct, but the displayed notation is a bit confusing with i/m. Please write the factors as (m(n−k−l)+i)/(m(n−k)+i) or similar.
  2. [Throughout] A number of typos and minor wording issues: 'measaures' (Remark 4.6), 'measue' (singular), 'q⟨1/m⟩y' typesetting glitch in §3.3, 'nononical'? Also the references [AFPU26] and [HHJK26] are cited as '2026' works; since the manuscript itself is dated 2026, please confirm the publication status or use preprint identifiers to avoid ambiguity.
  3. [§4.3, Theorem 4.5 proof] The conclusion uses [AFPU26, Theorem 1.1], but the text does not explicitly state the hypotheses on γ_0 that are needed for that black-box theorem (e.g., the convergence of the relevant coefficient ratios). Please add a sentence clarifying that the assumptions of [AFPU26, Theorem 1.1] are satisfied in this setting.
  4. [§1.2, Remark 1.2] The connection to [HHJK26] is stated as 'the same as in [HHJK26]' but the precise dictionary (the exponential profile g and the identification q(α)=e^{-g'(α)}) is not fully explained. Since this is a consistency remark rather than a proof step, it is acceptable, but a short paragraph could improve readability.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: Theorem 1.1 is proved from first principles via the independent root-coefficient bound (Theorem 1.3); self-citations are not load-bearing for the main claim.

full rationale

The main theorem (Theorem 1.1) is derived self-containedly. Its engine is Theorem 1.3, proved inside the paper from Vieta's formulas (Section 2), with the proof given in full. The iterated-real-root preservation step (Lemma 3.2) is quoted from [GNV25] but its proof is included in §3.1; the sign typo there ('\tilde p(λ_{n−1}) < 0') is a harmless slip, not a circular reduction. The coefficient-ratio identities (Lemma 3.4, Lemma 4.2) are direct computations from P(z)=p(z^m), and the root bounds (Lemmas 3.5, 4.3) follow from Theorem 1.3 with explicit constants that are killed by the 1/m power via Lemma 3.6. The limiting quantile formula emerges from the squeeze theorem and a Riemann-sum estimate (Lemma 4.4), not from any fitted parameter or from assuming the conclusion. Self-citations do appear ([AFPU26] includes Perales; [HHJK26] includes Hall), but [AFPU26, Theorem 1.1] is used only in the auxiliary fixed-m results (Theorems 3.8 and 4.5) as an external S-transform limit theorem, and [HHJK26] appears only in Remark 1.2 and the heuristic background, not in the proof of Theorem 1.1. These citations are real independent support, not circular. The paper does contain genuine gaps, but they are correctness issues rather than circularity: the zero-multiplicity formula c_q = c_p + l|Δ| in §4.1 is false when p already has a zero root (e.g., a=0, b=1, m=5, l=2, p(w)=w(w−1)(w−2) gives c_q=2, not 3), and the proof of Theorem 1.1 omits the case α < α_min ('The analysis of the case α < αmin is similar ... and is omitted'), which matters when μ0 has an atom at the origin. Neither issue involves a prediction reducing to an input by construction.

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

No numbers are fitted; the only 'parameters' are asymptotic limits (t, l/n, m/log n). The paper relies on standard Vieta and weak-convergence facts, the stated structural hypotheses on the polynomials, S-transform machinery from free probability, and one external theorem [AFPU26, Theorem 1.1] (self-cited, but independent of the new results). No invented entities.

assumptions (6)
  • standard math Vieta's formulas and elementary symmetric polynomial bounds (Theorem 1.3 proof)
    Used to bound the k-th largest root by ratios of consecutive coefficients; requires roots real and non-negative so all symmetric sums are positive.
  • standard math Weak convergence of measures on R is equivalent to pointwise convergence of quantile functions at continuity points
    Invoked in Theorem 3.7 to convert root-wise limits into measure limits.
  • domain assumption The polynomial p has degree n with real, non-negative roots
    Hypothesis of Theorem 1.1; needed for Theorem 1.3 to apply since all symmetric sums must be positive.
  • domain assumption m_n/log(n) → ∞ and the empirical root measure of P_n converges to a compactly supported radial μ0 with quantile function Q_{μ0}
    These are the stated hypotheses of Theorem 1.1; the rate m≫log n is used in Lemma 3.6 to control k^{1/m}.
  • domain assumption [AFPU26, Theorem 1.1]: ratios of consecutive coefficients converge to the S-transform of the limiting measure
    Used in §3.4 and §4.3 (fixed-m case) to identify limits as free multiplicative convolutions; this is a published theorem, partially self-cited (Perales is a coauthor), but independent of the paper's new results.
  • domain assumption S-transform theory for free multiplicative convolution and [Ari18, Corollary 8.10] for m-symmetric measures
    Used to translate the coefficient-limit into the free convolution statement in Theorem 3.8/4.5 and Remarks 3.9/4.7.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Repeated differentiation of deterministic polynomials with asymptotically radial root distributions." pith.science (2026). https://pith.science/paper/J7M6UFQF

@misc{pith2026260716954,
  author       = {Pith},
  title        = {Pith review of: Repeated differentiation of deterministic polynomials with asymptotically radial root distributions},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/J7M6UFQF}},
  note         = {Machine review of arXiv:2607.16954}
}
abstract

Recent works of Galligo, Najnudel, and Vu (2025) and Najnudel and Vu (2026) study repeated differentiation for polynomials of the form $P(z)=p(z^m)$, where $p$ is a deterministic polynomial of degree $n$ with real, non-negative roots, in the regime where $m$ and $n$ are large. If $m\gg \log(n)$ and the root distribution of $P$ converges to a compactly supported, radial probability measure $\mu_0$, these works show that for $0\le t<1$, the root distribution of the $\lfloor nmt\rfloor$-th derivative of $P$ converges to a compactly supported probability measure $\mu_t$ given by an explicit formula for its radial quantile function. We give a substantially simplified proof of this result and also extend the result from repeated differentiation to repeated applications of the differential operator $z^a(d/dz)^b$. We also compute the limiting root distribution in the case when $m$ is fixed and $n$ tends to infinity.

Figures

Figures reproduced from arXiv: 2607.16954 by the authors.

Figure 1
Figure 1. The small blue dots show all the derivatives of P N 0 up to time t and the larger red dots show the roots at time t. On the left, the smaller roots move radially inward and die at the origin before time t. On the right, the larger roots move radially inward and survive until time t. Shown for t = 0.5 and N = 100, starting from a polynomial with roots that are approximately uniform on the unit disk. [HK23], Campbell,… view at source ↗
Figure 2
Figure 2. Roots of the polynomial P(z) = p(z m) approximating the radial measure for which r = |z| is uniformly distributed between 0 and 1. Here p(z) is the polynomial with roots (i/n) m, for i = 1, . . . , n. Shown for n = 10 and m = 15. and suppose the empirical root measure of Pn converges as n → ∞ to a compactly supported probability measure µ0 with radial quantile function Qµ0 . For 0 ≤ t < 1, define Qn,t(z) = z ⌊nmnt⌋ … view at source ↗

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

24 extracted references · 12 canonical work pages

  1. [1]

    Advances in Mathematics , volume=

    S-transform in finite free probability , author=. Advances in Mathematics , volume=. 2026 , publisher=

    2026

  2. [2]

    Advances in Applied Mathematics , volume=

    k-divisible random variables in free probability , author=. Advances in Applied Mathematics , volume=. 2018 , publisher=

    2018

  3. [3]

    Integral Equations and Operator Theory , volume=

    The Atoms of the Free Multiplicative Convolution of Two Probability Distributions , author=. Integral Equations and Operator Theory , volume=. 2003 , publisher=

    2003

  4. [4]

    Indiana University Mathematics Journal , VOLUME =

    Bercovici, Hari and Voiculescu, Dan , TITLE =. Indiana University Mathematics Journal , VOLUME =. 1993 , NUMBER =. doi:10.1512/iumj.1993.42.42033 , URL =

    work page doi:10.1512/iumj.1993.42.42033 1993

  5. [5]

    Byun, Sung-Soo and Lee, Jaehun and Reddy, Tulasi Ram , TITLE =. Trans. Amer. Math. Soc. , FJOURNAL =. 2022 , NUMBER =. doi:10.1090/tran/8674 , URL =

    work page doi:10.1090/tran/8674 2022

  6. [6]

    Campbell, Andrew and O'Rourke, Sean and Renfrew, David , TITLE =. Int. Math. Res. Not. IMRN , FJOURNAL =. 2024 , VOLUME =. doi:10.1093/imrn/rnae062 , URL =

    work page doi:10.1093/imrn/rnae062 2024

  7. [7]

    Feng, Renjie and Yao, Dong , TITLE =. Anal. PDE , FJOURNAL =. 2019 , NUMBER =. doi:10.2140/apde.2019.12.1489 , URL =

    work page doi:10.2140/apde.2019.12.1489 2019

  8. [8]

    Law of large numbers for roots of finite free multiplicative convolution of polynomials , author=. SIGMA. Symmetry, Integrability and Geometry: Methods and Applications , volume=. 2023 , publisher=

    2023

Show all 24 references
  1. [9]

    arXiv preprint arXiv:2506.06263 , year=

    Dynamics of rotationally invariant polynomial root sets under iterated differentiations , author=. arXiv preprint arXiv:2506.06263 , year=

    arXiv

  2. [10]

    Operator Algebra and Dynamics

    The law of large numbers for the free multiplicative convolution , author=. Operator Algebra and Dynamics. Springer Proc. Math. Stat. , volume=. 2013 , publisher=

    2013

  3. [11]

    Transactions of the American Mathematical Society, Series B , volume=

    Roots of polynomials under repeated differentiation and repeated applications of fractional differential operators , author=. Transactions of the American Mathematical Society, Series B , volume=

  4. [12]

    Hoskins, Jeremy and Kabluchko, Zakhar , TITLE =. Exp. Math. , FJOURNAL =. 2023 , NUMBER =. doi:10.1080/10586458.2021.1980752 , URL =

    2023

  5. [13]

    Kabluchko, Zakhar , TITLE =. Proc. Amer. Math. Soc. , FJOURNAL =. 2015 , NUMBER =. doi:10.1090/S0002-9939-2014-12258-1 , URL =

    2015 doi

  6. [14]

    Kabluchko, Zakhar and Zaporozhets, Dmitry , TITLE =. Ann. Probab. , FJOURNAL =. 2014 , NUMBER =. doi:10.1214/13-AOP847 , URL =

    2014 doi

  7. [15]

    Michelen, Marcus and Vu, Xuan-Truong , TITLE =. Proc. Amer. Math. Soc. , FJOURNAL =. 2024 , NUMBER =. doi:10.1090/proc/16794 , URL =

    2024 doi

  8. [16]

    2026 , eprint=

    Root Dynamics of Differentiated Polynomials with Rotationally Invariant Structure , author=. 2026 , eprint=

    2026

  9. [17]

    Zeros and exponential profiles of polynomials

    Jalowy, Jonas and Kabluchko, Zakhar and Marynych, Alexander , journal=. Zeros and exponential profiles of polynomials

  10. [18]

    O'Rourke, Sean and Steinerberger, Stefan , TITLE =. Proc. Amer. Math. Soc. , FJOURNAL =. 2021 , NUMBER =. doi:10.1090/proc/15314 , URL =

    2021 doi

  11. [19]

    Electron

    O'Rourke, Sean and Williams, Noah , TITLE =. Electron. J. Probab. , FJOURNAL =. 2020 , PAGES =. doi:10.1214/20-ejp499 , URL =

    2020 doi

  12. [20]

    Real Roots of Hypergeometric Polynomials via Finite Free Convolution , journal =

    Mart. Real Roots of Hypergeometric Polynomials via Finite Free Convolution , journal =. 2024 , month =. doi:10.1093/imrn/rnae120 , url =

    2024 doi

  13. [21]

    Steinerberger, Stefan , TITLE =. Proc. Amer. Math. Soc. , FJOURNAL =. 2019 , NUMBER =. doi:10.1090/proc/14699 , URL =

    2019 doi

  14. [22]

    Totik, Vilmos , TITLE =. Trans. Amer. Math. Soc. , FJOURNAL =. 2019 , NUMBER =. doi:10.1090/tran/7667 , URL =

    2019 doi

  15. [23]

    , TITLE =

    Tucci, Gabriel H. , TITLE =. Indiana University Mathematics Journal , VOLUME =. 2010 , NUMBER =. doi:10.1512/iumj.2010.59.3775 , URL =

    2010 doi

  16. [24]

    Journal of Operator Theory , pages=

    Multiplication of certain non-commuting random variables , author=. Journal of Operator Theory , pages=. 1987 , publisher=

    1987

Pith tools

Reviewed August 1, 2026 · model on record in the stance chip above.