Shifted Hankel determinants of Riemann zeta screw-function moments

edit · history · discussion · files · long url · screw functions Hankel determinants moments L-function number theory
Numbers
$n$ 
$E_n$
0:
0.04622851856867953250013300124690386295746054628405804775271983534384274965723136441306201472404825257
1:
0.01710335630979576002485584351102085731635934719238979599198829670217597547173604320717355160702784561
2:
0.1519252445489181559657047361762750643923988379602425959456870544938409751554549815559079543189055428
3:
64.80041588353659912343117216585758668467849413865053465886868675684466552839215120026801270842119725
4:
2211888.592785118504035448507266102751987402780692829779355456764826854082564898272542977401388579568
5:
9062853781540.966021824573488312315652220395494778595305603650530437421994271761989699876270122195496
6:
6240170306410925830759.139741535789133774469072568986412298138989621594160251265949115233440964300505
7:
962676007368224583474270341214424.4004069864287506620985563560990633482131089143242818515182393539760
8:
42780211580729167189883729515626115585746637845.47954479816470962179519811112049220177741878797733037
9:
684504167963460116269705564169185532354136215037643969989792952.0711109844874647334203085823781285238
10:
4819593279984316978660275397661180724151894076224486818573813457981679644091212642.013258034866257969
11:
1.791872901906773130777547410769536195204062322165545224280344288297866967747090693370895656601512743e103
12:
4.157164961175818777809364206539739347441689210532729702640400559530308570641728545887737828102059031e127
13:
7.021139993728237279622881646991333320185079533083977333624881525926537555005990574575394895577306053e154
14:
9.960176531919017676106519499096498357607683899985781049443491084491896010229200995063045903846249033e184
15:
1.356278934995374674166945297018166596571021659770616546905042047025579825675967720459665463250623900e218
16:
2.009064536293805771533086796160151136485383368515900338229822853325377197247768194386449065649466280e254
17:
3.641984533010236111855778196352396688888287506117886464887750085995720390299242238101487660464721217e293
18:
9.029666757902694353557222774587504430736897394768929478185521745436398805611618907257914119606453131e335
19:
3.402039054708891485437137540402007994892751106398706827445966656659766623502216735663428529416920809e381
20:
2.152758565343971010512921856546770370622362828110323958834543870492075275263245620425757415770598533e430
Definition
The first-shifted Hankel determinants are $E_n=\det[\mu_{i+j+1}]_{i,j=0}^{n}$, where $\mu_k=\frac14\int_0^\infty e^{-t/2}\Psi(t)t^k\,dt$, $\Psi=-g$, and $g$ is Suzuki's Riemann zeta screw function [1], equation (1.13) and Theorem 1.8.
Parameters
$n$
—   Hankel index ($n\geq0$)
Formulas
(1)
Let $\xi(s)=\frac12s(s-1)\pi^{-s/2}\Gamma(s/2)\zeta(s)$, with removable singularities filled in, where $\Gamma$ and $\zeta$ are the gamma and Riemann zeta functions. Then $\xi(1)=1/2$ and $\displaystyle\mu_k=k![X^k]\frac{\xi'(1-X)}{(1-2X)^2\xi(1-X)}$, where $[X^k]$ denotes the Taylor coefficient at $X=0$. [1], equation (1.15).
(2)
$E_0=\mu_1=6\lambda_1-\lambda_2$ and $E_1=\mu_1\mu_3-\mu_2^2$, where $\lambda_j$ are Li coefficients, without division by $j$. [1], equation (8.1).
(3)
For $w_1(t)=4^{-1}t e^{-t/2}\Psi(t)$ and $\Psi=-g$ with the screw function $g$, $\displaystyle E_n=\frac{1}{(n+1)!}\int_{(0,\infty)^{n+1}}\prod_{0\leq a<b\leq n}(t_a-t_b)^2\prod_{a=0}^{n}w_1(t_a)\,dt_0\cdots dt_n$. This absolutely convergent identity follows by determinant expansion; it does not assume that $w_1$ is nonnegative.
Comments
(4)
The matrix has size $n+1$, so $E_0=\mu_1$ is not an empty determinant. The source denotes this matrix by $\Delta_n^{(1)}$; this table denotes its scalar determinant by $E_n$.
(5)
The moments $\mu_k$ are absolutely convergent unconditionally for every $k\geq0$, by [1], Theorem 1.1(3). Their weight uses $\Psi=-g$ with the screw function $g$. The entries are signed determinants of ordinary moments, with no absolute value, logarithm, moment rescaling or division by factorials.
(6)
[1], Theorem 1.8, states that the Riemann hypothesis is equivalent to $D_n\geq0$ and $E_n\geq0$ for every integer $n>0$, where $D_n=\det[\mu_{i+j}]_{i,j=0}^{n}$ is the unshifted determinant. This table contains only $E_n$; a finite positive prefix does not establish that all-order paired criterion.
Programs
(P1)
Sage
import numberdb.sage as numberdb
from math import factorial
from sage.rings.complex_arb import ComplexBallField
from sage.rings.polynomial.polynomial_ring_constructor import PolynomialRing
import sage.rings.polynomial.laurent_polynomial_ring
from sage.matrix.constructor import matrix

def E(n, bits=1024):
    if n < 0:
        raise ValueError("n must be nonnegative")
    C = ComplexBallField(bits)
    P = PolynomialRing(C, 'u')
    u = P.gen()
    s = 1 + u
    length = 2*n + 3
    z = 1 + u*s._zeta_series(length, deflate=True)
    h = (s*z*(s/2)._gamma_series(length)
         *(-s*C.pi().log()/2)._exp_series(length)/2)._log_series(length)
    mu = [factorial(k)*sum(((k-j+1)*2**(k-j)*(-1)**j*(j+1)*h[j+1]
                           for j in range(k+1)), C(0)) for k in range(2*n+2)]
    value = matrix(C, n+1, n+1, lambda i,j: mu[i+j+1]).det()
    assert value.real().is_finite() and value.imag().is_finite()
    assert value.imag().contains_zero()
    return value.real()

E(20)  # Increase bits if more precision is needed at larger orders.
References
[1]
Masatoshi Suzuki, Aspects of the screw function corresponding to the Riemann zeta-function, Journal of the London Mathematical Society 108 (2023), 1448–1487; equations and theorem numbers refer to arXiv version 4. (arXiv) (doi)
Links
Similar tables
Moments of the Riemann zeta screw function —   The matrix entries are the ordinary weighted moments $\mu_1,\ldots,\mu_{2n+1}$, without factorial normalization.
Hankel determinants of Riemann zeta screw-function moments —   The unshifted companion uses $\mu_{i+j}$ in place of $\mu_{i+j+1}$, with the same matrix size $n+1$.
Values of the Riemann zeta screw function —   The moment weight is $4^{-1}e^{-t/2}\Psi(t)$ with $\Psi=-g$; that table stores the signed function $g$.
Data properties
Entries are of type: real number
Table is complete: no (it contains every index $0\leq n\leq20$, hence every leading first-shifted matrix size from 1 through 21, including both parities. The bound is a noncanonical practical prefix limiting conditioning and independent-check cost; larger orders, unshifted determinants and other shifts or minors are omitted)
How they were obtained:

All 42 moment inputs through order 41 are recomputed from the pole-removed completed product using native Arb/FLINT Taylor series. The unscaled matrices are evaluated by the native complex-ball determinant. Working precision is 1024 bits; every determinant retains at least 830 relative-accuracy bits before conversion to 100 significant decimal digits, with one last-place unit of uncertainty. [3]

more

A separate implementation recomputes the moments by the bounded theta integral [2], equations 25.5.13–14, and the Li-coordinate transform [1], equation (8.3). Both omitted theta summands and the infinite integration tail have explicit bounds. At 1152 bits, a separately implemented Bareiss determinant uses factorial row and column scaling, reversed exactly. Every independent enclosure is checked strictly inside the corresponding stored decimal interval. Nonzero quadrature, initial-moment, factorial-moment and signed-measure controls pass. Wrong sign, factor, matrix size, offset, normalization, omitted rescaling, NaN and wide-ball controls are rejected. The methods share completed-xi theory and Arb/FLINT arithmetic. Executable checks, analytic bounds and checkpoints are attached.