REVIEW 2 major objections 5 minor 33 references
Selberg's Central Limit Theorem weighted by Linear Statistics of Zeta Zeros
T0 review · 2 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash
Pith's one-line read At leading order the logarithm of zeta and the counting statistics of its zeros are asymptotically independent Gaussians, so weighting by nearby zero density leaves Selberg's central limit theorem intact.
desk verdict Main theorem is new and valuable, but the proof of Proposition 4.2 skips the exact combinatorics that carries the main constant; likely routine, yet currently the least secure load-bearing step. 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 argument runs on a weighted moment calculation with three ingredients. Lemma 2.1 is an explicit formula that rewrites the centered zero statistic as the prime-power sum $S_\varphi(t)=-(2/\log T)\sum_n \Lambda(n)/\sqrt{n}\,\hat\varphi(\log n/\log T)\cos(t\log n)$ up to an error $O(1/\log T)$, so the zero statistic becomes a Dirichlet polynomial whose coefficients are supported on primes and prime squares. The second ingredient is the classical Dirichlet polynomial $P_x(t)=\sum_{p\le x}p^{-1/2-it}$, which approximates $\log\zeta(1/2+it)$ in mean square; Propositions 3.1 and 3.2 reduce the mixed moments of $\log\zeta$ and $N_\varphi$ to moments of $P_x$ and $S_\varphi^*$, with the Riemann Hypothesis needed only for the wider support in the imaginary part. The load-bearing combinatorial step is Proposition 4.2: after expanding the product and integrating, only the diagonal terms survive, and the prime-power equalities split into ten possible pairing relations, of which only two contribute at leading order: $P_x$ pairing with itself, contributing $h!(\log\log T)^h$, and $S_\varphi^*$ pairing with itself, contributing $\mu_k\sigma_\varphi^k$ through the variance integral $\sigma_\varphi^2=\int\min(|u|,1)\,\hat\varphi(u)^2\,du$. The factorization into these two dominant sums is exactly what produces asymptotic independence.
What would settle it
Compute the $h=\ell=1$, $k=2$ mixed moment $\frac{1}{T}\int_T^{2T}|\log\zeta(1/2+it)|^2(N_\varphi(t)-\hat\varphi(0))^2\,dt$ for a fixed $\varphi$ with $\operatorname{supp}\hat\varphi\subset(-1/2,1/2)$, and divide by $\log\log T$; the theorem forces this ratio to converge to $\sigma_\varphi^2$, so a numerical evaluation at large $T$ showing the ratio drifting away from $\sigma_\varphi^2$ would refute the claimed asymptotic independence.
Extended reading notes
Core claim
On its own terms, the paper's central claim is Theorem 1.1: for fixed $h,\ell,k$ with $k$ even, and for any even real-valued $\varphi$ with smooth compactly supported Fourier transform satisfying $\operatorname{supp}\hat\varphi\subseteq(-2/(k+2),\,2/(k+2))$, the joint moment $\frac{1}{T}\int_T^{2T}(\log\zeta(1/2+it))^h(\overline{\log\zeta(1/2+it)})^\ell(N_\varphi(t)-\hat\varphi(0))^k\,dt$ equals $\mathbf{1}(h=\ell)\,\mu_k\sigma_\varphi^k\,h!(\log\log T)^h$ plus an error $O((\log\log T)^{(h+\ell-1)/2})$. This is the moment sequence of a standard complex Gaussian multiplied by the moment sequence of a real Gaussian with variance $\sigma_\varphi^2=\int\min(|u|,1)\,\hat\varphi(u)^2\,du$, so the two random variables are independent at leading order. Passing from moments to distributions gives Corollary 1.3: if $\tau$ is sampled with probability weight $|N_\varphi(\tau)-\hat\varphi(0)|^k$, then $\log\zeta(1/2+i\tau)/\sqrt{\log\log T}$ still converges to a standard complex Gaussian. Under the Riemann Hypothesis, Theorem 1.2 achieves the same factorization for the imaginary part with the full support $\eta<2/k$; the imaginary part, normalized by $\sqrt{(1/2)\log\log T}$, converges to a standard real Gaussian under the weighted measure. Finally, Proposition 1.5 identifies the next-order correlation: under RH, $\frac{1}{T}\int_T^{2T}\log\zeta(1/2+it)(N_\varphi(t)-\hat\varphi(0))\,dt=-\varphi(0)/2+O(\sqrt{\log\log T}/\log T)$, so the correlation coefficient is asymptotic to $-\varphi(0)/(2\sigma_\varphi\sqrt{\log\log T})$.
Load-bearing premise
The load-bearing assumption is the Riemann Hypothesis, used for the full support $\eta<2/k$ in the imaginary-part theorem and for the correlation formula $-\varphi(0)/2$; without it, the unconditional theorem survives only with the narrower support $\eta<2/(k+2)$.
Editorial extensions
If this is right
- If the theorem is right, choosing $t$ with probability proportional to $|N_\varphi(t)-\hat\varphi(0)|^k$ does not change the limiting Gaussian distribution of $\log\zeta(1/2+it)$; Selberg's central limit theorem survives this conditioning.
- Under the Riemann Hypothesis, the same holds for $\Im\log\zeta$ with Fourier support up to the natural barrier $\eta<2/k$, so the imaginary part has a real Gaussian limit under the weighted measure.
- The leading mixed moments of $\log\zeta$ and $N_\varphi$ factor into the product of their individual Gaussian and mock-Gaussian moments, meaning the two are asymptotically independent at the scale of Selberg's theorem.
- The correlation estimate in Proposition 1.5 shows the independence is only asymptotic: the correlation coefficient between $\log\zeta$ and $N_\varphi-\hat\varphi(0)$ is of size $1/\sqrt{\log\log T}$, so the approach to independence is slow.
- Setting $h=\ell=0$ in the moment computation recovers the mock-Gaussian moment result for the one-level density with the sharp cutoff $\omega=\mathbf{1}_{[0,1]}$, so the weighted theorem contains that result as a special case.
Reading between the lines
- The same mechanism suggests a robust principle: any fixed polynomial in short-range zero statistics should fail to shift the leading Gaussian fluctuations of $\log\zeta$, so the factorization should persist for weighted measures built from several one-level densities.
- Because the slow correlation decay matches the hybrid prime-zero product picture in which the zero product contributes only at second order, one would expect the $-\varphi(0)/2$ term to be visible in short-interval averages too; computing the covariance on intervals of length $T^\theta$ rather than $[T,2T]$ is a concrete way to test how the correlation develops.
- If the Riemann Hypothesis is false, Theorem 1.2 and Proposition 1.5 lose their footing; a natural test is whether a much weaker zero-density hypothesis, rather than the full Riemann Hypothesis, is enough to push the imaginary-part support to $\eta<2/k$.
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper studies the joint distribution of log ζ(1/2+it) and the linear statistics N_φ(t) of the nontrivial zeros of ζ. The main results are an unconditional mixed-moment asymptotic (Theorem 1.1) for a power of N_φ−φ̂(0) times powers of log ζ and its conjugate, with Fourier support η < 2/(k+2); an RH-conditional version (Theorem 1.2) for ℑ log ζ with the larger support η < 2/k; and an RH-conditional correlation estimate (Proposition 1.5) showing that the first mixed moment of log ζ and N_φ−φ̂(0) is −φ(0)/2 + O(√log log T / log T). The proof follows the standard Selberg–Tsang route: an explicit formula for N_φ, approximation of log ζ by a short Dirichlet polynomial, reduction to diagonal moments, and a combinatorial evaluation of the diagonal sum. The paper also derives weighted central limit theorems (Corollaries 1.3 and 1.4) as consequences of the moment estimates.
Significance. If the proofs are correct, this is a substantial contribution: it establishes a genuine joint version of Selberg's CLT and the Hughes–Rudnick mock-Gaussian moment results, with explicit constants and no free parameters. The unconditional range η < 2/(k+2) is natural given the Hölder-based approximation, and the RH-conditional extension for ℑ log ζ reaches the conjecturally sharp barrier η < 2/k. The correlation estimate of Proposition 1.5 is a nice concrete illustration of the slow decay of independence. The methods are standard but the combination is new. The main theorems are clearly stated and the paper is well organized. However, as detailed below, a load-bearing step in the proof of the central Theorem 1.1 is asserted rather than proved, so the paper needs revision before it is fully convincing.
major comments (2)
- [§4, Proposition 4.2, after Eq. (4.8)] The proof of Proposition 4.2 asserts that every solution of the diagonal equality ∏ p_i ∏_{j∈J} n_j = ∏ q_i ∏_{j∈Jc} n_j 'has to be in one of the relations' (4.9)–(4.10), and that the left-hand side of (4.8) 'factors as a product' of the corresponding sums. This step directly produces the leading constant μ_k σ_φ^k h! in Theorem 1.1, yet it is not proved. In particular, the constants C(u,J) are introduced but never computed or bounded; the exclusion of all configurations with u1 < h or with u7+u10 > 0 is only sketched; and the claim that the number of p–q matchings is 'simply h!' requires qualification when primes repeat, because the same tuple can then arise from several bijections. A rigorous enumeration is needed, including a verification that all configurations outside u1 = h = ℓ and u4 = k/2 contribute at most O((log log T)^((h+ℓ−1)/2)). Since Theorem 1.2 and Corollaries 1.3–1.4 inherit this step, the gap is load-bearing.
- [§4, Eq. (4.11) and subsequent paragraph] The bounds for the individual sums S1,...,S10 are stated with 'one verifies', and the subsequent counting argument is informal. The paper does not write down the linear relations between h, ℓ, k and the multiplicities u1,...,u10; such an accounting is necessary to justify the assertions that u2 = u3 = u5 = u6 = u8 = u9 = 0 when h = ℓ = u1, and that u7 = u10 = 0 unless the contribution is absorbed into the error term. Without these details, the proof of the main constant in Proposition 4.2 is not fully verifiable. Please provide a complete count of the relation types and a proof that the only leading-order configuration is the one with all p's paired to q's and all n's paired among themselves.
minor comments (5)
- [§4, Eqs. (4.5)–(4.6)] The letter h is used both for the number of p-variables and for the nonzero integer difference in the diagonal approximation; this notation clash is confusing. Please use a different symbol (e.g., m) for the integer difference.
- [§3, Proposition 3.1] The bound for E2 is described only as a 'routine calculation' after an application of Hölder's inequality; please spell out the mixed-moment estimate for ℜ E_x and ℑ E_x, especially the role of the (k+2)-th moment of S*_φ in fixing the support condition η < 2/(k+2).
- [§3, Proposition 3.2] The estimates for the weighted moments of A3 and R are stated without proof ('by a similar calculation'). Since these bounds are needed for Theorem 1.2, a few explanatory lines would make the argument easier to check.
- [§2, proof of Proposition 1.5, Eq. (2.6)] The error term after applying Goldston's formula (2.5) should be O((log T / T) ∑_{n≤T^η} Λ(n)), not O((1/T) ∑ Λ(n)); the displayed bound is too optimistic by a factor log T. The final estimate is still acceptable, but the displayed line should be corrected.
- [§1, Theorem 1.1] For k = 0 the support condition on φ̂ is unnecessary, since S*_φ does not appear in the moment; stating the k = 0 case separately (or noting that the condition is vacuous for that case) would avoid an artificial restriction.
Circularity Check
No circular derivation: the joint CLT is derived from external ingredients (Selberg-Tsang, Hughes-Rudnick, Goldston); the few self-citations are contextual and not load-bearing.
full rationale
The derivation chain is: explicit formula (Lemma 2.1) expresses N_phi via prime sums; Propositions 3.1 and 3.2 reduce log-zeta moments to Dirichlet-polynomial moments using Selberg-Tsang's unweighted moment estimates; Proposition 4.1 isolates the diagonal contribution; Proposition 4.2 evaluates that diagonal combinatorially; Theorems 1.1 and 1.2 combine these steps. None of these steps fits a parameter to the target moment or assumes the target asymptotic. The boundedness of the S*_phi moments is imported from Hughes-Rudnick [14], not derived from the theorem being proved, and Goldston's correlation formula [12] is used for Proposition 1.5 with the RH assumption made explicit. Self-citations appear (Bettin-Fazzari [3] in Lemma 2.1 and several Fazzari papers in the introduction), but they are contextual or replaceable by standard arguments; no load-bearing step reduces to a self-citation chain. The forward reference in the introduction to the sharp-cutoff Hughes-Rudnick result is not circular: Theorem 1.1 is proved without assuming that sharp-cutoff result, and the h=l=0 case of the new theorem recovers it. The combinatorial enumeration in Proposition 4.2 contains compressed assertions, since the constants C(u,J) are not computed except in the special case u1=h and u4=k/2, and a miscount there would affect the leading constant. That is a possible rigor gap, not circularity, because the constants are counts of matchings rather than inputs fitted from the claimed Gaussian moments. There is also no renaming of a known result: combining Selberg's CLT with Hughes-Rudnick's mock-Gaussian moments into a joint weighted CLT is new content. The score reflects only minor, non-load-bearing self-citations.
Assumptions & free parameters
assumptions (5)
- domain assumption Riemann Hypothesis
- standard math Selberg-Tsang approximation estimates (3.6)
- standard math Hughes-Rudnick moment bound for S_φ
- standard math Goldston's correlation estimate (2.5)
- standard math Prime Number Theorem
Cite this review
Pith. "Pith review of Selberg's Central Limit Theorem weighted by Linear Statistics of Zeta Zeros." pith.science (2026). https://pith.science/paper/GTRZRG3L
@misc{pith2026250704150,
author = {Pith},
title = {Pith review of: Selberg's Central Limit Theorem weighted by Linear Statistics of Zeta Zeros},
year = {2026},
howpublished = {\url{https://pith.science/paper/GTRZRG3L}},
note = {Machine review of arXiv:2507.04150}
}
abstract
We consider the value distribution of the logarithm of the Riemann zeta function on the critical line, weighted by the local statistics of zeta zeros. We show that, with appropriate normalization, it satisfies a complex Central Limit Theorem, provided that the Fourier support of the test function in the linear statistics is sufficiently small. For the imaginary part, we extend this support condition up to its natural barrier under the Riemann Hypothesis. Finally, we prove that the correlation between $\log \zeta$ and the one-level density, while negligible on the level of Selberg's Central Limit Theorem, only decays at a rather slow rate if the Riemann Hypothesis is assumed. Our results can be viewed as a combination of Selberg's Central Limit Theorem with work of Hughes and Rudnick on mock-Gaussian behavior of the local statistics.
Reference graph
Works this paper leans on
-
[1]
L. P. Arguin and E. Bailey. Large deviation estimates of Selberg’s Central Limit Theorem and applications. Int. Math. Res. Not. IMRN, 2023(23):20574–20612, 2023
work page 2023
-
[2]
L. P. Arguin and E. Bailey. Lower bounds for the large deviations of Selberg’s central limit theorem. Mathematika, 71(1):Paper No. e70002, 40, 2025
work page 2025
-
[3]
S. Bettin and A. Fazzari. A weighted one-level density of the non-trivial zeros of the Riemann zeta-function. Math. Z., 307(2):32, 2024
work page 2024
-
[4]
Billingsley.Probability and Measure
P. Billingsley.Probability and Measure. John Wiley and Sons, New York, 3rd edition, 1995
work page 1995
-
[5]
E. Bombieri and D. Hejhal. On the distribution of zeros of linear combinations of Euler products. Duke Math. J., 80(3), 821–862, 1995
work page 1995
-
[6]
H. M. Bui, N. Evans, S. Lester, and K. Pratt. Weighted central limit theorems for central values ofL-functions. J. Eur. Math. Soc. (JEMS), 27(6):2477–2529, 2025
work page 2025
-
[7]
A. Fazzari. A weighted central limit theorem forlog|ζ(1/2 +it)|. Mathematika, 67(2): 324–341, 2021
work page 2021
-
[8]
A. Fazzari. Weighted value distributions of the Riemann zeta function on the critical line. Forum Math., 33(3):579–592, 2021
work page 2021
Show all 33 references
-
[9]
A. Fazzari. A weighted one-level density of families ofL-functions. Algebra Number Theory, 18(1):87–132, 2024
2024
-
[10]
A. Fazzari. On the joint second moment of zeta and its logarithmic derivative.Acta Arith., 217(4):379–394, 2025
2025
-
[11]
Fazzari and M
A. Fazzari and M. Gerspach. The third moment of the logarithm of zeta and a twisted pair correlation conjecture. Preprint, arXiv:2412.20099, 2024
2024 arXiv
-
[12]
D. A. Goldston. On the functionS(T ) in the theory of the Riemann zeta-function.Journal of Number Theory, 27(2):149–177, 1987
1987
-
[13]
S. M. Gonek, C. P. Hughes, and J. P. Keating. A hybrid Euler-Hadamard product for the Riemann zeta function.Duke Math. J., 136, no. 3, pp 507-549, 2007
2007
-
[14]
C. P. Hughes and Z. Rudnick. Linear statistics for zeros of Riemann’s zeta function.C.R. Acad. Sci. Paris, Ser I. 335, 667–670, 2002
2002
-
[15]
C. P. Hughes and Z. Rudnick. Linear statistics of low-lying zeros ofL-functions. Q. J. Math. 54, no. 3, 309–333, 2003. SELBERG’S CENTRAL LIMIT THEOREM WEIGHTED BY LINEAR STATISTICS OF ZETA ZEROS 16
2003
-
[16]
Iwaniec and E
H. Iwaniec and E. Kowalski.Analytic Number Theory, volume 53 ofColloq. Publ., Am. Math. Soc.Providence, RI: American Mathematical Society (AMS), 2004
2004
-
[17]
D. Joyner. Distribution theorems ofL-functions. Longman Scientific, Harlow, 1986
1986
-
[18]
Kowalski
E. Kowalski. An Introduction to Probabilistic Number Theory. Cambridge Studies in Advanced Mathematics. Cambridge University Press, 2021
2021
-
[19]
Laurinčikas
A. Laurinčikas. A limit theorem for the Riemann zeta-function on the critical line.II. Litovsk. Mat. Sb., 27(3):489–500, 1987
1987
-
[20]
Laurinčikas.Limit Theorems for the Riemann Zeta-Function
A. Laurinčikas.Limit Theorems for the Riemann Zeta-Function. Springer, 1996
1996
-
[21]
M. M. Lugar, M. B. Milinovich, and E. Quesada-Herrera. On the number variance of zeta zeros and a conjecture of Berry.Mathematika, 69(2):303–348, 2023
2023
-
[22]
H. L. Montgomery. The pair correlation of zeros of the zeta function. InAnalytic number theory (Proc. Sympos. Pure Math., Vol. XXIV, St. Louis Univ., Mo.), volume Vol. XXIV of Proc. Sympos. Pure Math., pages 181–193. Amer. Math. Soc., Providence, RI, 1973
1973
-
[23]
Radziwiłł
M. Radziwiłł. Large deviations in Selberg’s central limit theorem.Preprint, arxiv:1108.5092, 2011
2011 arXiv
-
[24]
Radziwiłł and K
M. Radziwiłł and K. Soundararajan. Selberg’s central limit theorem forlog|ζ( 1 2 +it)|. L’Enseignement Mathématique (2) 63, 1–19, 2017
2017
-
[25]
Rudnick and P
Z. Rudnick and P. Sarnak. Then–level correlations of zeros of the zeta function.C. R. Acad. Sci. Paris Sér. I Math. 319, 1027–1032, 1994
1994
-
[26]
Rudnick and P
Z. Rudnick and P. Sarnak. Zeros of principalL-functions and random matrix theory.Duke Math. J. 81, no. 2, 269–322, 1996
1996
-
[27]
A. Selberg. Contributions to the theory of the Riemann zeta-function.Arch. Math. Naturvid., 48(5):89–155, 1946
1946
-
[28]
A. Selberg. Old and new conjectures and results about a class of Dirichlet series.In Pro- ceedings of the Amalfi Conference on Analytic Number Theory, September 1989, Università di Salerno, 367-385; Collected Papers, Vol. II, Springer-Vedag, Berlin, 1991, 47-63, 1992
1989
-
[29]
Sugiyama
S. Sugiyama. Low-lying zeros of symmetric powerL-functions weighted by symmetric square L-values. To appear in Journal of the Mathematical Society of Japan, 2025
2025
-
[30]
Sugiyama and A
S. Sugiyama and A. I. Suriajaya. Weighted one-level density of low-lying zeros of Dirichlet L-functions. Res. Number Theory, 8, Article number: 55, 2022
2022
-
[31]
E. C. Titchmarsh. On the remainder in the formula forN(T ), the number of zeros ofζ(s) in the strip0<t<T . Proceedings of the London Mathematical Society, s2-27(1):449–458, 1928
1928
-
[32]
K. M. Tsang.The distribution of the values of the Riemann zeta function. PhD thesis, Princeton University, 1984
1984
-
[33]
S. Zhao. Weighted low-lying zeros ofL-functions attached to Siegel modular forms.Pacific J. Math., 335(1):183–210, 2025. Département de mathématiques et de statistique, Université de Montréal. CP 6128, succ. Centre-ville. Montreal, QC H3C 3J7, Canada Email address: alessandro....
2025
Reviewed August 6, 2026 · model on record in the stance chip above.
Discussion (0). Continue with ORCID to comment.