Bessel period determinants from even rows

edit · history · discussion · files · short url · integral period special values number theory
Numbers
$n$ 
$E_n$
5:
0.6603448690186723578372668317059942638542419916968738583008035875538949586837994454109810720121753277
6:
1.093670186350183889271856823917256856644045856785500562365575315163594611211087327750765401655813100
7:
2.535350386125462350369009228497001507200656712330465008909477466497718326860200117994795034210763915
8:
7.345809609162108125830752329014780651525567410837081335185721782165027597765231431299512844388336461
9:
2.222698213626676789294987929320071185044712356063782185872619103596149161897108363485003099482451229
10:
9.058092849199337725300084303567341133027787277282426806905092565914564096051890549603861907921884583
11:
58.09247314932704325241961082578779502486004075533762267871987095067573727709163339129217178383078023
12:
514.1465241118917036778212212034732849655967211447421485097997771868660589159136259379677329593033611
13:
91.23101029713787135579257658649096555518329054590288138712273363510819286655921817971031205109609604
14:
1017.672201807256815249828302210223282998135852986380701154638966913413094555593229247794819936562971
15:
19146.49923945965211347944165349027700694665809354733607693330411532199636631684152806490129347111965
16:
529116.5234187940141274334010008344622800470353850044508671669304689589788517265492840265825340559356
Definition
The even-row Bessel period determinant is $E_n=\det\bigl(M(2i,n-2i;2j-1)\bigr)_{i,j=1}^{d_e(n)}$, with $M$ and $d_e$ specified in (4) [1].
Parameters
$n$
—   total number of Bessel factors ($n\geq5$)
Formulas
(1)
$E_5=\sqrt3\,\Gamma(1/15)\Gamma(2/15)\Gamma(4/15)\Gamma(8/15)/(480\pi)$ [2], equations (4.37)–(4.38). The factors occur in Gamma values and $\pi$.
(2)
For $n=2k+1$ with $k\geq2$, $E_{2k+1}=\dfrac{\sqrt{[(2k+1)!!]^{2-(-1)^k}}}{\pi^{\lfloor(k+1)/2\rfloor}2^{\lfloor k/2\rfloor}(k-1)!!(k!!)^{1-(-1)^k}}O_{2k+1}$, where $O_n$ is the odd-row Bessel period determinant and $!!$ denotes the double factorial [2], equation (1.21).
(3)
In particular, $E_5=\sqrt{15}\,O_5/(2\pi)$, $E_7=35\sqrt{105}\,O_7/(12\pi^2)$ and $E_9=\sqrt{105}\,O_9/(4\pi^2)$ [2]. Here $O_n$ denotes the odd-row determinant.
Comments
(4)
Write $M(a,b;r)=\int_0^\infty I_0(t)^aK_0(t)^b t^r\,dt$ and $d_e(n)=\lfloor n/4\rfloor-\mathbf{1}_{4\mid n}$. The functions $I_0$ and $K_0$ are the standard order-zero modified Bessel functions, and row and column indices increase from one.
(5)
The integration path is the positive real axis, both Bessel functions have unit scale, and $K_0$ has its positive-real principal value. Values are raw signed determinants, with no additional powers of $\pi$, alternating column signs or absolute values. Even refers to the exponent of $I_0$ in the rows; every column uses an odd power $t^{2j-1}$. Constituent moments are defined in Bessel moments.
(6)
For $4\nmid n$, $E_n=D_{n,\mathrm{even}}$ in [1], Notation 8.2. For $4\mid n$, $E_n=D^\prime_{n,\mathrm{even}}$: delete the final row and final column of the unreduced matrix before evaluating it. The zero-dimensional case $n=4$ is omitted. An unreduced matrix at $n=4m$ includes $M(2m,2m;2m-1)$, whose integral diverges logarithmically at infinity; no regularization is used here.
(7)
These specified representatives enter the critical-period formulas for symmetric-power Kloosterman motives [1], Theorem 8.4 and Table 1. An abstract Deligne period is defined only up to a nonzero rational factor; the integral normalization here specifies an actual real number.
(8)
Every defining entry has $b>a$ and $r=2j-1\geq1$. Its integrand is $O(t^r(1+|\log t|)^b)$ at zero and $O(t^{r-n/2}e^{-(b-a)t})$ at infinity, so the integrals are ordinary convergent integrals.
Programs
(P1)
Sage
# With generate.py attached to this table:
# sage -pip install numberdb
from generate import determinant
n = 9
E_n = determinant(n, digits=100)  # rigorous ball; n>=5
# Default table selection: list(range(5,17)).
# d_e(n) = n//4 - int(n%4 == 0).
References
[1]
J. Fresán, C. Sabbah and J.-D. Yu, Quadratic relations between Bessel moments, arXiv manuscript v2 (2021). Notation 8.2, equation (8.3), Theorem 8.4 and Table 1. (arXiv)
[2]
Y. Zhou, Wrońskian algebra and Broadhurst–Roberts quadratic relations, arXiv manuscript v3. Equations (1.21), (4.37)–(4.38) and Proposition 4.5. (arXiv)
Links
Similar tables
Bessel period determinants from odd rows —   Those minors use odd first-kind exponents. For odd total factor count the two raw quantities satisfy the reflection formula above; both use odd integration powers.
Bessel moments —   The entries of every defining matrix are $\operatorname{IKM}(a,b;r)$. The scalar cases $E_5=M(2,3;1)$ and $E_6=M(2,4;1)$ overlap its moment values.
Broadhurst–Mellit determinants of Bessel moments —   Those full determinants use consecutive positive first-kind exponents; the present determinants select even exponents and use the prescribed smaller size at multiples of four.
Broadhurst–Mellit vacuum determinants —   Those determinants include a vacuum row and source-specified columns, rather than the even-row minors here.
Data properties
Entries are of type: real number
digits: 100 (Unknown key)
Table is complete: no (it holds every factor count $n=5,6,\ldots,16$, with $d_e(n)=\lfloor n/4\rfloor-\mathbf{1}_{4\mid n}$. This practical prefix covers all four residue classes, all initial scalar cases, determinants through order three, and reduced orders one, two and three at $n=8,12,16$. The natural family continues beyond this noncanonical computational cutoff)
How they were obtained:

The defining moments use SageMath 10.9 complex Arb integration after $t=e^x$, at 525 working bits. The positive integral representations in [5] give the explicit origin and exponential tail bounds derived in the attached generator. Arb encloses each finite integration segment, both omitted endpoints, and the signed determinant [7]. The principal $K_0$ is evaluated through Tricomi $U$ [6]; the analytic callback rejects balls leaving the right half-plane of its argument. The weakest final enclosure retains 388 relative accuracy bits before serialization to 100 significant digits. Exact rational endpoint comparisons confirm containment of each complete computed ball in its stored last-place interval.

more

All 12 determinants were independently recomputed with mpmath at 130 decimal working digits, direct-$t$ tanh-sinh quadrature and a signed permutation expansion. Every result lies in its stored interval. This quadrature comparison is heuristic; the complete Arb enclosures support the stored precision. The five-factor Gamma evaluation and the odd/even reflection formula at every odd $5\leq n\leq15$ pass [2]. Both sides of these reflection checks were integrated independently of the sibling table. The nonzero controls $M(0,1;1)=1$ and $M(0,2;1)=1/2$ were also checked, at 50 working decimal digits. Wrong sign, missing pi, wrong dimension, divergent-boundary, NaN, infinite and vacuously wide interval controls are rejected.

The attached generator, exact binary-rational moment checkpoints, executable checks, result files, argument-selection note and matching archived skill manifest preserve the methods and evidence. Checkpoint replay with the same generator tests reproducibility separately from independent mathematical agreement.