Searcharxiv⌕ Search

arXiv · 2611.23332

Gram's law for moving targets of the Riemann zeta-function

Abstract

We study the solutions of $ζ(s)=f(s)$ for targets $f$ that grow with $|s|$, and Gram's law for them: the statement that each shifted Gram interval contains exactly one solution. The deviation from Gram's law is an integer phase-slip counter, equal up to rounding to $-\frac{1}{2}S(t;x_c)$, where $S(t;x)=π^{-1}\argζ(x+it)$ and $x_c\approx 1/2+\log|f(it)|/\log t$ is a balance abscissa determined by the target. If $|f|\asymp t^κ$ with $κ>κ_1=0.5339\ldots$, where $\sum_p\arcsin(p^{-1/2-κ_1})=π$, and $\arg f$ converges, then every sufficiently high shifted Gram interval contains exactly one solution, which is simple; this localises the fixed points of $ζ$. The threshold is sharp: for $f=c(-s)^κ$ with $1/2<κ<κ_1$, Gram's law fails in $\asymp T$ intervals up to height $T$, and for $0<κ\le 1/2$ it fails infinitely often but with density zero. In both ranges the counter has an explicit limiting distribution. For targets growing more slowly than any power, such as $\log(-s)$, the solutions lie on average $(1+o(1))\log|f(iT)|/\log T$ to the left of the critical line. If $\log|f(iT)|/\sqrt{\log\log T}\to\infty$, the proportion of solutions to the right of the line tends to $0$, whereas for fixed targets Selberg conjectured a quarter, and the counting error at the shifted Gram points is asymptotically Gaussian with variance $(8π^2)^{-1}\log(\log T/\log|f(iT)|)+O(1)$. For $\log(-s)$ the solutions follow neither $θ(T)/π$ nor the zeros of $ζ$. All results are unconditional; numerical computations for $\log(-s)$ illustrate how slowly they take effect.

Explore related subjects

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Ryan Goulden. 2026-11-16. Gram's law for moving targets of the Riemann zeta-function. https://searcharxiv.com/abs/2611.23332

Citations retain the original authors and DOI and include this site’s record URL. If this discovery service contributed to your research, you can also cite Searcharxiv by Indiciae: BibTeX · RIS.

KEEP EXPLORING

Related papers

Generalized Centralizers and Polynomial Equivalence of Matrices: Answers to Questions of Dolinar, Guterman, Kuzma, and Markova

For $n\times n$ matrices $A$ and $B$ over a field and a scalar $ω\ne0$, let $\mathcal C_ω(A)=\{X:AX=ωXA\}$. If $ω$ is a root of unity of order $q\ge2$, then $\mathcal C_{ω^j}(A)=\mathcal C_{ω^j}(B)$ for all integers $j$ if and only if each of $A$ and $B$ is a polynomial in the other with exponents $\equiv1\pmod q$. Over algebraically closed fields, for $ω\ne1$, all $A$, and scalars $\varepsilon\ne0$, we describe all $Y$ with $XY=\varepsilon YX$ for all $X\in\mathcal C_ω(A)$. Let $a(μ)$ be the size of the largest Jordan block of $A$ at $μ$, or $0$ if there is none. If $\varepsilon$ is a power of $ω$, all such $Y$ are polynomials in $A$ if and only if $a(ωμ)=a(μ)$ for all $μ\ne0$, or this fails only at some $ν$ and at $ω^{-1}ν$, where $A$ has a single largest Jordan block at $ν$ and $a(ων)=a(ω^{-1}ν)=a(ν)-1$. This answers two questions of Dolinar, Guterman, Kuzma, and Markova. If $ω\ne1$ has finite order, then for such $A$, $\mathcal C_ω(A)=\mathcal C_ω(B)$ exactly when $A$ and $B$ are polynomials in each other of this form. For all other $A$ some $B$ that is not a polynomial in $A$ satisfies $\mathcal C_ω(A)=\mathcal C_ω(B)$ if the field has more than $n^2+1$ elements. In the $ω$-commuting digraph of Raja and Vaezpour, with an edge from $A$ to $B$ if $AB=ωBA$, vertices with edges to and from the same vertices are twins. Over fields with more than $2n^2+1$ elements we determine the vertices $A$ whose twins are all polynomials in $A$, and describe these twins. We find all $A$, $B$, and $k\ge2$ for which the kernel of $(X\mapsto[A,X])^k$ is the set of $X$ with $[X,[X,\ldots,[X,B]\ldots]]=0$ ($k$ brackets), where $[X,Y]=XY-YX$. Apart from scalar $B$, this happens only in characteristic $\ne2$, for $k=2$ and $2\times2$ matrices that generate the same quadratic extension.

math.GM↗

Theta-Eta Mellin Transforms, Shifted Reflections, and Rigorous Approximation Theory for a Variant Theta Construction

We develop a rigorous analytic framework for investigating a proposed functional equation associated with a variant theta construction. We prove the classical Mellin identity connecting the Jacobi kernel $1-θ_4(0\mid ix^2)$ to the completed Dirichlet eta function, and extend its Mellin transform meromorphically to the whole plane with a single simple pole at the origin. Mellin uniqueness identifies $x^{3/8}(1-θ_4(0\mid ix^2))$ as the unique kernel in a natural weighted $L^1$ class whose Mellin transform is the shifted completed eta function. We distinguish this Gaussian kernel from the weighted exponential series $Θ^B(s)=\sum_{n\geq1}ne^{-ns}$, whose Mellin transform is $Γ(s)ζ(s-1)$ and obeys a different, explicitly derived reflection relation. We show that every affine reflection $s\mapsto c-s$ can be generated for a shifted completed eta or zeta function by choosing the shift appropriately; consequently, the occurrence of $\tfrac14-s$ alone cannot identify the original variant with a shifted eta function. The corrected analytic product defined here is an elementary meromorphic multiplier of the completed zeta function and therefore has a precise $1-s$ functional equation. We prove an explicit eta-tail bound, a sharper alternating-tail asymptotic, a uniformly convergent Euler-transformed eta series with a geometric error bound, a uniform zeta-approximation bound away from the line $\Re s=1$, and a combined truncation estimate when both the zeta and weighted-exponential factors are truncated. We also determine the principal Laurent data and the cancellation of the gamma poles at the trivial zeta zeros for the corrected product. Reproducible high-precision experiments verify the stated error scales. The exact identification of the original variant with a shifted completed eta function remains open.

math.GM↗

An Algebraic Approach to the Goldbach Conjecture

We prove the Goldbach Conjecture using p-adic analysis and algebraic methods, requiring no knowledge of prime gaps or distribution. For $a \in \mathbb{N}$, let $\mathcal{P}_a$ be the set of primes up to $a$. We prove that if a counter-example $2a$ exists, there is a monic polynomial $\mathcal{G}_-(z) \in \mathbb{Z}[z]$ where $\mathcal{G}_-(2a) = 0$. Applying Hensel's Lemma, we then show this polynomial structure forces each prime $p_i \in \mathcal{P}_a$ to satisfy $2a = p_i^{α_i} + p_i$ for a unique $α_i \in \mathbb{N} \cup \{0\}$. This produces a contradiction for $a > 4$ since each $2a - p_i > 1$ requires each $α_i > 0$ and the primorial, $a\#$, to satisfy $a\# \mid 2a$, which is impossible since $2a < a\#$ for all $a > 4$.

math.GM↗