REVIEW 3 major objections 4 minor 34 references
The Variance of the Number of Zeros for Complex Random Polynomials Spanned by OPUC
T0 review · 3 major / 4 minor · reviewed 2026-08-14 · deepseek-v4-flash
Pith's one-line read For random combinations of Nevai-class OPUC with Gaussian coefficients, the limiting variance of the number of zeros in an annulus avoiding the unit circle is an explicit rational function of the inner and outer radii.
desk verdict Solid extension of random polynomial variance to OPUC in the Nevai class, with a fixable sign error in the exterior annulus derivation. read the letter →
The pith
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
The reading
What carries the argument
The load-bearing object is the pair correlation function $\rho_n^{(2)}(z,w)$, expressed through the Christoffel-Darboux kernel $K_n(z,w)=\sum_{j=0}^n \varphi_j(z)\varphi_j(w)$ and its mixed derivatives. The permanent-determinant formula for correlation functions of Gaussian analytic functions turns $\rho_n^{(2)}$ into ratios of these kernels; the Christoffel-Darboux identity and the Nevai-class ratio asymptotics $\varphi_n(z)/\varphi_n^*(z)\to0$ locally uniformly in $\mathbb{D}$ (with the reciprocal tending to $0$ outside $\mathbb{D}$) let every kernel ratio converge to the hyperbolic Bergman kernel. Integrating the resulting universal kernel over $A(s,t)$, using the geometric series for $(1-xy)^{-2}$ inside and outside the disk, produces the closed variance formulas.
What would settle it
Take the monomial basis $\varphi_k(z)=z^k$, draw i.i.d. standard complex Gaussian coefficients, and estimate by Monte Carlo the variance of the number of zeros in $A(0,1/2)$ for large degree $n$; the theorem predicts the limit $4/15$ for $\operatorname{Var}[N_n]$. Equivalently, estimate the two-point correlation for two interior points $z,w$ and compare it with the kernel stated in Theorem 3; a systematic mismatch beyond Monte Carlo error would refute the pair-correlation limit and the variance formula derived from it.
Extended reading notes
Core claim
Let $P_n(z)=\sum_{k=0}^n \eta_k\varphi_k(z)$ with i.i.d. standard complex Gaussian $\eta_k$. Theorem 3 states that when the $\varphi_k$ are real-valued on the real line and belong to the Nevai class, the second correlation function of zeros converges locally uniformly for $z,w$ both in $\mathbb{D}$ or both in $\mathbb{C}\setminus\overline{\mathbb{D}}$ to $$\lim_{n\to\infty} \$rho_n^{{(2)}}$(z,w)=\frac{1}{\$pi^{2}$}\left(\frac{1}{(1-|z|^2)^2(1-|w|^2)^2}-\frac{1}{|1-zw|^4}\right).$$ Theorem 4 integrates this kernel over an annulus $A(s,t)=\{z:0\le s<|z|<t\}$ that avoids the unit circle. For $A(s,t)\subsetneq\mathbb{D}$ the limiting variance is $$\frac{($t^{2}$-$s^{2}$)(1-$s^{2}$($t^{4}$(2+$s^{2}$)-2))}{(1-$t^{4}$)(1-$s^{4}$)(1-(st)^2)},$$ and for $A(s,t)\subsetneq\mathbb{C}\setminus\overline{\mathbb{D}}$ the same expression holds with the roles of $s$ and $t$ interchanged. A corollary singled out by the paper is the disk case $s=0$: $\lim \operatorname{Var}[N_n(D(0,t))]=t^2/(1-t^4)$. For sectors crossing the unit circle, Theorem 2 gives $\operatorname{Var}[N_n]/n^2=O(\sqrt{\log n/n})$ for generalized Jacobi weights and $O(\max\{\sqrt{\log n/n},\varepsilon_n^{1/4}\})$ for Ullman-Stahl-Totik regular OPUC.
Load-bearing premise
The load-bearing assumption is that the OPUC basis is in the Nevai class with real coefficients, meaning the ratio of each polynomial to its reversed conjugate tends to zero locally uniformly inside the unit disk (and the reciprocal tends to zero outside). If a basis leaves this class, the pair-correlation kernel need not converge to the universal limit, and the variance formulas in Theorem 4 are not guaranteed.
Editorial extensions
If this is right
- For a disk $D(0,t)$ strictly inside the unit circle, the limiting variance is $t^2/(1-t^4)$, a single-radius formula suitable for direct numerical checks.
- The negative term in the pair-correlation kernel gives short-distance repulsion between zeros, so the limiting zero process is not Poisson: nearby zeros are suppressed.
- In sectors crossing the unit circle, $\operatorname{Var}[N_n]/n^2\to0$ at the stated rates, so the fraction of zeros in a sector converges to the sector angle divided by $2\pi$.
- The variance formulas are universal across the Nevai class: any real-coefficient OPUC basis in this class yields the same limiting constants.
Reading between the lines
- A natural extension is to replace annuli by angular sectors away from the unit circle; the same integration method with Fourier series should give explicit variance formulas, though the paper notes the result takes a complicated shape.
- The proof only needs the Christoffel-Darboux kernel to converge to the hyperbolic kernel, which suggests the formulas may survive for non-OPUC bases with the same kernel asymptotics; a direct check on weighted Bergman bases could test this.
- The exterior formula is obtained from the interior one by exchanging $s$ and $t$; that reflection symmetry is not named in the paper and hints at an underlying $z\mapsto 1/z$ invariance of the limiting zero statistics.
- Because the monomial basis is a degenerate case of the Nevai class, a Monte Carlo variance estimate for $P_n(z)=\sum \eta_k z^k$ in $|z|<1/2$ with $n\sim200$ gives a sharp numerical target ($4/15$) for the closed-form prediction.
Signed reviews
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper studies the number of zeros of random linear combinations P_n(z)=sum eta_k phi_k(z), where {phi_k} are orthonormal polynomials on the unit circle (OPUC) and the coefficients are random. The main results are: (i) Theorem 2, a quantitative decay bound for Var[N_n(A_r(alpha,beta))]/n^2 for sectors crossing the unit circle under either Ullman-Stahl-Totik regularity or generalized Jacobi weights; (ii) Theorem 3, a limiting formula for the two-point correlation function of zeros when the OPUC are real-valued on the real line and belong to the Nevai class, with i.i.d. complex Gaussian coefficients; and (iii) Theorem 4, explicit limiting variance formulas for the number of zeros in annuli A(s,t) that are strictly inside or strictly outside the unit disk. The proofs use the Hough-Krishnapur-Peres-Virag correlation formulas, the Christoffel-Darboux formula for OPUC, and Nevai-class ratio asymptotics.
Significance. If the results are correct, Theorem 4 gives the first explicit variance formulas for zero counts of random OPUC in annuli away from the unit circle, and Theorem 2 provides quantitative variance bounds in the sector case. The final exterior variance formula in Theorem 4 passes the inversion-symmetry check Var_ext(s,t)=Var_int(1/t,1/s) in the monomial case and is supported by direct quadrature, which is evidence that the target formula is right. The proofs rely on standard external tools (Hough-Krishnapur-Peres-Virag, Simon) and the author's earlier results as input; I found no circularity in the derivation of the variance formula. The central results are significant in extending zero-statistics asymptotics beyond the monomial basis, though the manuscript needs corrections in the statement and proof before the claims are fully supported.
major comments (3)
- [Theorem 3, Eq. (14)] As printed, Eq. (14) states that the limiting pair correlation contains the term -1/|1-zw|^4. This cannot be correct for the monomial basis phi_k(z)=z^k, for which the zero process is rotationally symmetric: the pair correlation must depend on Re(z \bar w), not Re(zw), so the singular term must be -1/|1-z\bar w|^4. The proof of Theorem 4, Eq. (77), uses exactly the Hermitian expression (1-z\bar w)^{-2}(1-\bar z w)^{-2}, confirming that the intended kernel is |1-z\bar w|. Please correct Eq. (14) and the corresponding final display after Eq. (75), and make the notation for z\bar w consistent throughout the proof of Theorem 3.
- [Proof of Theorem 4, exterior-annulus case] The displayed chain in the exterior case, after Eq. (79), contains a sign error. For t>s>1, the sum \sum_{k=0}^\infty (t^{-2k-2}-s^{-2k-2})^2 is positive and equals -(t^2-s^2)^2(1+(st)^2)/((1-t^4)(1-s^4)(1-(st)^2)), because 1/(t^4-1)= -t^4/(1-t^4), 1/(s^4-1)= -s^4/(1-s^4), and -2/((st)^2-1)=2(st)^2/(1-(st)^2). The manuscript instead writes the intermediate expression t^4/(1-t^4) - 2(st)^2/(1-(st)^2) + s^4/(1-s^4), which is the negative of the sum. The final exterior variance formula is consistent with the correct sign, but the derivation as displayed is not valid; a reader following the chain of equalities cannot obtain the stated result. Please correct this line and re-check the analogous interior computation.
- [Proof of Theorem 3, Eqs. (35)-(75)] The proof of Theorem 3 is a long and intricate computation, and several cancellations are asserted rather than demonstrated. In particular, the step 'simplifying the sum of expressions (51), (52), and (53)' that yields Eq. (54), and the analogous simplifications leading to Eqs. (70) and (74), involve substantial algebra and cancellations of o(1) terms that are not shown. Because the theorem statement also has the conjugation issue noted above, the reader cannot verify the proof line by line. Please expand these computations or provide a supplementary appendix with the missing algebraic details, and ensure that every occurrence of 'zw' in the asymptotic displays is consistent with the Hermitian inner product used in Eq. (19).
minor comments (4)
- [Abstract and references] There are several typographical errors: the abstract has 'on the the variance'; 'Farahmond' should be 'Farahmand'; 'Boomolny' should be 'Bogomolny'; 'Grandville' should be 'Granville'; 'Zeldtich' should be 'Zelditch'; and 'trignometric' should be 'trigonometric' in several places.
- [Proof of Theorem 2, Eqs. (17)-(18)] The chain after Eq. (17) uses an equality sign after an upper bound; the equality should be read as a continuing upper bound. Please insert '\le' where the transition from the bound on E[(N_n/n)^2] to the variance is made, to avoid confusion about whether an identity is being asserted.
- [Eq. (33) and Nevai class] The paper defines the Nevai class through the local-uniform ratio limit \phi_n(z)/\phi_n^*(z)\to 0. This is nonstandard; the usual Nevai class is defined by the recurrence coefficients tending to zero. Please either cite the equivalence or state explicitly that the ratio property is the standing assumption for Theorems 3 and 4.
- [Notation near Eq. (35)] The text refers to 'the denominator of \pi\rho^{(2)}_n(z,w)' but the subsequent expression has the cube of the determinant; please clarify whether the factor is \pi^2 or \pi and check the notation for consistency.
Circularity Check
No significant circularity: the new variance formula is derived from external correlation formulas and stated Nevai-class asymptotics, not from its own target.
full rationale
Walking the derivation chain: Theorem 4's limiting variance is obtained by applying the Hough-Krishnapur-Peres-Virag correlation formulas [15] and the Christoffel-Darboux kernel (21), under the stated Nevai-class hypothesis (33) for OPUC that are real-valued on the real line. The first-correlation limit (13) is imported from the author's prior result [34], and the sector variance estimates in Lemma 1 and Theorem 2 use prior estimates from [26]; in both cases these are published, parameter-free results whose hypotheses do not include the target variance formula, so the self-citations function as lemmas rather than as a covert assumption of the conclusion. No parameter is fitted to a subset of the data and then called a prediction, no quantity is defined in terms of the variance being computed, and no uniqueness theorem from the authors' own work is invoked to force the choice of model. The only internally inconsistent step found is in the exterior-annulus proof of Theorem 4: the displayed identity "∑(t^{-2k-2}-s^{-2k-2})^2 = t^4/(1-t^4)-2(st)^2/(1-(st)^2)+s^4/(1-s^4)" has the wrong sign (the sum is positive), and the following line flips sign again; this is an algebraic/correctness defect in an intermediate computation, not circularity, since the final closed form is consistent with the inversion symmetry V_ext(s,t)=V_int(1/t,1/s). Hence no significant circularity.
Assumptions & free parameters
assumptions (7)
- standard math The general correlation function formula (19) for zeros of Gaussian analytic functions (Corollary 3.4.2 of [15]) is valid for any polynomial basis.
- standard math The Christoffel-Darboux formula (21) and the recurrence relations for OPUC from Simon [29].
- domain assumption The Nevai class property (33): φ_n(z)/φ_n^*(z) → 0 locally uniformly in D, and the reciprocal in C\D via real coefficients.
- domain assumption The OPUC are real-valued on the real line, i.e., have real coefficients.
- domain assumption The random coefficients in Section 2.1 satisfy uniform moment bounds (3) and (4).
- domain assumption The measure μ is regular in the sense of Ullman-Stahl-Totik (5) or has the generalized Jacobi weight form (7).
- standard math The prior L^1 estimate from [26, Theorem 3.1] and the first correlation limit from [34, Corollary 2.2].
Cite this review
Pith. "Pith review of The Variance of the Number of Zeros for Complex Random Polynomials Spanned by OPUC." pith.science (2026). https://pith.science/paper/HUKRPPXJ
@misc{pith2026190802234,
author = {Pith},
title = {Pith review of: The Variance of the Number of Zeros for Complex Random Polynomials Spanned by OPUC},
year = {2026},
howpublished = {\url{https://pith.science/paper/HUKRPPXJ}},
note = {Machine review of arXiv:1908.02234}
}
abstract
Let $\{\varphi_k\}_{k=0}^\infty $ be a sequence of orthonormal polynomials on the unit circle (OPUC) with respect to a probability measure $ \mu $. We study the variance of the number of zeros of random linear combinations of the form $$ P_n(z)=\sum_{k=0}^{n}\eta_k\varphi_k(z), $$ where $\{\eta_k\}_{k=0}^n $ are complex-valued random variables. Under the assumption that the distribution for each $\eta_k$ satisfies certain uniform bounds for the fractional and logarithmic moments, for the cases when $\{\varphi_k\}$ are regular in the sense of Ullman-Stahl-Totik or are such that the measure of orthogonality $\mu$ satisfies $d\mu(\theta)=w(\theta)d\theta$ where $w(\theta)=v(\theta)\prod_{j=1}^J|\theta - \theta_j|^{\alpha_j}$, with $v(\theta)\geq c>0$, $\theta,\theta_j\in [0,2\pi)$, and $\alpha_j>0$, we give a quantitative estimate on the the variance of the number of zeros of $P_n$ in sectors that intersect the unit circle. When $\{\varphi_k\}$ are real-valued on the real-line from the Nevai class and $\{\eta_k\}$ are i.i.d.~complex-valued standard Gaussian, we prove a formula for the limiting value of variance of the number of zeros of $P_n$ in annuli that do not contain the unit circle.
Reference graph
Works this paper leans on
-
[1]
P. Bleher and X. Di, Correlations between zeros of a random polyn omial, J. Statist. Phys. 88 (1997), 269–305
work page 1997
-
[2]
A. Bloch and G. P´ olya, On the roots of a certain algebraic equatio n, Proc. Lond. Math. Soc. 33 (1932), 102–114
work page 1932
-
[3]
A. T. Bharucha-Reid and M. Sambandham, Random polynomials, Ac ademic Press, Orlando, 1986
work page 1986
-
[4]
E. Bogomolny, O. Bhigas, and P. Leboeuf, Quantum chaotic dyna mics and random polynomials, J. Statist. Phys. 85 (1996), 639–679
work page 1996
-
[5]
A. Edelman and E. Kostlan, How many zeros of a random polynomial are real?, Bull. Amer. Math. Soc. 32 (1995), 1-37
work page 1995
-
[6]
P. Erd˝ os and A. Offord, On the number of real roots of a rando m algebraic equation, Proc. Lond. Math. Soc. 6 (1956), 139–160
work page 1956
-
[7]
P. Erd˝ os and P. Tur´ an, On the distribution of roots of polynom ials, Ann. Math. 51 (1950), 105–119
work page 1950
-
[8]
Farahmand, Topics in random polynomials, Pitman Res
K. Farahmand, Topics in random polynomials, Pitman Res. Notes Ma th. 393, 1998
work page 1998
Show all 34 references
-
[9]
Farahmand, On the variance of the number of real zeros of a random trignometric polynomial, J
K. Farahmand, On the variance of the number of real zeros of a random trignometric polynomial, J. Appl. Math. Stochastic Anal. 14 (2001), 265–274
2001
-
[10]
Fairley, Roots of random polynomials, Ph.D
W. Fairley, Roots of random polynomials, Ph.D. Dissertation, Har vard Univeristy, Cambridge, Mas- sachusetts, 1968
1968
-
[11]
Forrester and G
P. Forrester and G. Honner, Exact statistical properties of complex random polynomials, J. Phys. A: Math. Gen. 32 (1999), 2961–2981
1999
-
[12]
Granville and I
A. Granville and I. Wigman, The zeros of random trignometric poly nomials, Amer. J. Math. 133 (2011), 295–357
2011
-
[13]
Hammersley, The zeros of a random polynomial, Proc
J. Hammersley, The zeros of a random polynomial, Proc. of the T hird Berk. Sym. on Math. Stat. and Prob. 1954-1955 vol. II, University of Cal. Press, Berkeley and Lo s Angeles (1956) 89–111. 17
1956
-
[14]
Hannay, Chaotic analytic zero points: exact statistics for t hose of a random spin state, J
J. Hannay, Chaotic analytic zero points: exact statistics for t hose of a random spin state, J. Phys. A, 29 (1996), L101–L105
1996
-
[15]
Hough, M
J. Hough, M. Krishnapur, Y. Peres, B. Vir´ ag, Zeros of Gauss ian Analytic Functions and Determinantal Point Processes, Univ. Lect. Ser. 51. American Mathematical Soc iety, Providence, RI, 2009
2009
-
[16]
Jamrom, On the average number of real roots of a random a lgebraic polynomial, Vestnik Leningrad
B. Jamrom, On the average number of real roots of a random a lgebraic polynomial, Vestnik Leningrad. Univ. Ser. Math. Meh. 4 (1971), no. 19, 152–156
1971
-
[17]
Jamrom, The average number of real zeros of random polyn omials, Soviet Math
B. Jamrom, The average number of real zeros of random polyn omials, Soviet Math. Dokl. 13 (1972), 1381-1383
1972
-
[18]
Kac, On the average number of real roots of a random algeb raic equation, Bull
M. Kac, On the average number of real roots of a random algeb raic equation, Bull. Amer. Math. Soc. 49 (1943), 314–320
1943
-
[19]
Littlewood and A
J. Littlewood and A. Offord, On the distribution of the zeros and a-values of a random integral function I, J. Lond. Math. Soc. 20 (1945), 120–136
1945
-
[20]
Littlewood and A
J. Littlewood and A. Offord, On the distribution of the zeros and values of a random integral function II, Ann. Math. 49 (1948), 885–952. Errata, 50 (1949), 990–99 1, 976, 35–58
1948
-
[21]
Littlewood and A
J. Littlewood and A. Offord, On the number of real roots of a ra ndom algebraic equation, J. Lond. Math. Soc. 13 (1938), 288–295
1938
-
[22]
Littlewood and A
J. Littlewood and A. Offord, On the number of real roots of a ra ndom algebraic equation II, Proc. Cambridge Philos. Soc. 35 (1939), 133–148
1939
-
[23]
Littlewood and A
J. Littlewood and A. Offord, On the number of real roots of a ra ndom algebraic equation III, Rec. Math. [Mat. Sbornik] N.S. 54 (1943), 277–286
1943
-
[24]
Maslova, The variance of the number of real roots of rando m polynomials, Theor
N. Maslova, The variance of the number of real roots of rando m polynomials, Theor. Probab. Appl. 19 (1974), 35–52
1974
-
[25]
Nguyen and V
H. Nguyen and V. Vu, Random polynomials: central limit theorems for the real roots, (2019) arXiv:1904.04347
2019 arXiv
-
[26]
Pritsker and A
I. Pritsker and A. Yeager, Zeros of polynomials with random coe fficients, J. Approx. Theory 189 (2015), 88–100
2015
-
[27]
Rice, Mathematical theory of random noise, Bell System Tec h J
S. Rice, Mathematical theory of random noise, Bell System Tec h J. 25 (1945), 46–156
1945
-
[28]
Shiffman and S
B. Shiffman and S. Zelditch, Number variance of random zeros on complex manifolds, Geo. and Funct. Anal. 18 no. 4 (2008), 1422–1475
2008
-
[29]
Simon, Orthogonal Polynomials on the Unit Circle
B. Simon, Orthogonal Polynomials on the Unit Circle. American Mat hematical Society Colloquium Publications, Vol. 54, Part I, Providence, RI, 2005
2005
-
[30]
Stevens, The average and variance of the number of real z eros of random functions
D. Stevens, The average and variance of the number of real z eros of random functions. Ph.D. Disserta- tion, New York University, New York, 1965
1965
-
[31]
Su and Q
Z. Su and Q. Shao, Asymptotics of the variance of the number r eal roots of random trignometic poly- nomials, Science China Mathematics 55 no. 11 (2012), 2347–2366
2012
-
[32]
Wang, Bounds on the average number of real roots of a ran dom algebraic equation (Chinese), Chinese Ann
Y. Wang, Bounds on the average number of real roots of a ran dom algebraic equation (Chinese), Chinese Ann. Math. Ser. A 4 no. 5 (1983), 601–605. An English summary app ears in Chinese Ann. Math. Ser. B 4 no 4 (1983), 527
1983
-
[33]
Wilkins, An asymptotic expansion for the expected number of real zeros of a random polynomial, Proc
J. Wilkins, An asymptotic expansion for the expected number of real zeros of a random polynomial, Proc. Amer. Math. Soc. 103 (1988), 1249–1258. 18
1988
-
[34]
Yeager, Zeros of random orthogonal polynomials with comple x Gaussian coefficients, Rocky Mout
A. Yeager, Zeros of random orthogonal polynomials with comple x Gaussian coefficients, Rocky Mout. J. Math. (2018) 48 no. 7, 2385–2403. Aaron M. Yeager, Department of Mathematics, College of Coastal Georgia, Bru nswick, Georgia 31520 E-mail address : ayeager@ccga.edu 19
2018
Reviewed August 14, 2026 · model on record in the stance chip above.
Discussion (0). Continue with ORCID to comment.