Determinants of even-power Bessel moments

edit · history · discussion · files · short url · integral moments period special values number theory
Numbers
$n$
normalization 
determinant in the selected normalization
1
bessel:
1.570796326794896619231321691639751442098584699687552910487472296153908203143104499314017412671058534
1
period:
-3.141592653589793238462643383279502884197169399375105820974944592307816406286208998628034825342117068
2
bessel:
2.467401100272339654708622749969037783828424851810197656603337344055011205604801310750443350929638058
2
period:
9.869604401089358618834490999876151135313699407240790626413349376220044822419205243001773403718552232
3
bessel:
2.983580364656519516507820469439664851149318378147098904891232524778265376237465756180213773082724342
3
period:
-i*149.9711064799948497135424659436318263907964677060810788500205744713991693429358471199854319123852286
4
bessel:
4.566051142218864245458140594783052089830980578407129141787555908904883849538774821411962526629046526
4
period:
-i*918.0590543558443597882239301303068194409021198842249716604773310587049764008420734573505014205153567
5
bessel:
6.619456597228324681732161922452355850921854217069344673258165527784230093110109544185492756936837221
5
period:
-i*105085.2877961714702608379914016951236339672788790089672128720280166480132719678266175237129481060498
6
bessel:
7.721786308651183274030416262944959144447574439812079013761467622938565187037252845080944276545341087
6
period:
i*980680.0599129686570189854733572822234252157371320098366938734721296999728341850658047706030190085203
7
bessel:
17.05257354885543905008457360244337527313763026170824989577443413939518636969801928530185410106354429
7
period:
-1074407665.337479846131281719459003941949325176376303935319637339610526536602251607077441063728962899
8
bessel:
12.85450512302640935492128966750568002504200912034580140002631840897762571651668965441987810594312395
8
period:
-12958493378.12259096810771910782834627944734750214781616955737229420887312881543726979249615147894966
9
bessel:
50.90368597175864418817251092344345548476162138100366025427026309670973585249035555080519706084862473
9
period:
159955010840332.0676167141019239204325118409171201853205709914417713972725430361812530307276989792876
10
bessel:
21.89715428768250071679674605768523317708565602814499070685812857990393359672532359353406822052170175
10
period:
-2201842627051815.540416474222985828838162423862275851433116242701253321249230479102819573084555105729
11
bessel:
175.9097565405617104041382185942695940961996280971756715951646468339723957472638058607327137014558678
11
period:
i*346431888948317993537.6866037411166067424274814294640926397280264905225645107145286593721085568859256
12
bessel:
39.08612889833238426594878537023116630491526050226985923783527200243018330385742992510652625347938539
12
period:
i*4926410171112293718942.028830089454851721350754685824822143454317156254198287600122774940963636288454
13
bessel:
703.3795234979928818872656602647672559692847313940551917482202983672249297065162934611902004398319163
13
period:
i*10909549954937735975298430114.71617019013487249051038156901429458910058260395620224527180709258668454
14
bessel:
74.28707586411577212429643582816824394663239430094774479528370336176869476614044098176936962026890298
14
period:
-i*147482451326948566383620192578.0284603095595950424793123605639984380205122813170548001721834642804686
15
bessel:
3253.230012621569484998494117543726466653667025486612587325966097197766635222442739868014738999121343
15
period:
4993797506373113634569190463362170543.202779843625300847332352555845119430696886032005961994806607758
16
bessel:
152.0769617145500211269946912413138776284206401341882309825469593410390661205285239177185566498559378
16
period:
59761233176126742036769919203697416354.36510232789966155046638178996864542426704250065340401361057680
Definition
For $n\geq1$ and $r=\lfloor(n+1)/2\rfloor$, the determinant $\Delta_n=\det[B_n(a,2j)]_{a,j=0}^{r-1}$ is given in two invertible conventions: bessel returns $\Delta_n$ and period returns $\Pi_n=\det P_n$ [1]. The moments $B_n$ and fixed-basis matrix $P_n$ are defined in Comments.
Parameters
$n$
—   total number of Bessel factors ($n\geq1$)
normalization
—   determinant normalization ($\mathrm{bessel}$ or $\mathrm{period}$)
Formulas
(1)
For $r=\lfloor(n+1)/2\rfloor$ and $s=r(r-1)/2$, $\Pi_n=(-1)^{nr-s}2^{nr-2s}(\pi i)^s\Delta_n$, by factoring the rows and columns [1], Remark 27. The constant $\pi$ is in T7.
(2)
For $r\geq1$, $\displaystyle\Delta_{2r-1}=(\sqrt\pi)^{r(r+1)}(\sqrt2)^{r(r-3)}\prod_{a=1}^{r-1}\frac{a^{r-a}}{(\sqrt{2a+1})^{2a+1}}$ [1], Corollary 39. Every square root is positive and an empty product is $1$.
(3)
For $r\geq1$, $\displaystyle\Delta_{2r}=\frac{(\sqrt\pi)^{(r+1)^2}}{(\sqrt2)^{r(r+3)}\Gamma((r+1)/2)}\prod_{a=1}^{r-1}\frac{(2a+1)^{r-a}}{(a+1)^{a+1}}$ [1], Corollary 39. Here $\Gamma$ is Euler’s gamma function and an empty product is $1$.
(4)
$\Delta_1=\pi/2$, $\Delta_2=\pi^2/4$ and $\Delta_3=\pi^3/(6\sqrt3)$; the corresponding period determinants are $\Pi_1=-\pi$, $\Pi_2=\pi^2$ and $\Pi_3=-i\,8\pi^4/(3\sqrt3)$ [1].
(5)
Let $D_n=\det[B_n(a,2b-1)]_{a,b=1}^{\lfloor(n-1)/2\rfloor}$ be the Broadhurst–Mellit determinant. For $r\geq2$, $\Delta_{2r-1}=(r-1)!\pi^r D_{2r-1}/2$. This follows by comparing [1], Corollary 39, with [2], equation (1.4).
(6)
For $r\geq2$, $\displaystyle\Delta_{2r}=\frac{(2r-1)!!\,\Gamma(r/2)\pi^{r+1/2}}{2^{r+1}\Gamma((r+1)/2)}D_{2r}$, where $D_n$ is the Broadhurst–Mellit determinant, $!!$ denotes the double factorial and $\Gamma$ is Euler’s gamma function. Compare [1], Corollary 39, with [2], equation (1.6).
Comments
(7)
Define $B_n(a,q)=\int_0^\infty I_0(t)^aK_0(t)^{n-a}t^q\,dt$ and $(P_n)_{a,j}=(-1)^{n-a}2^{n-2j}(\pi i)^aB_n(a,2j)$ for $0\leq a,j<r$. Both matrices have their rows and columns in increasing order [1], §6.2 and Corollary 26.
(8)
The functions $I_0$ and $K_0$ are the standard order-zero modified Bessel function of the first kind and modified Bessel function of the second kind, at common unit scale on the positive real axis. The positive-real principal branch of $K_0$ is used. Rows $a$ and columns $j$ increase from zero; no factorial or loop factor is absorbed. The moment $B_n(a,q)$ is $\operatorname{IKM}(a,n-a;q)$.
(9)
The period convention is the ordered pairing of the rapid-decay cycles $\delta_a$ with the de Rham classes $v_0^n z^j\,dz/z$ in [1], Corollary 26, with $t=2\sqrt z$. Here $i^2=-1$. The explicit matrix in the definition fixes its phase and basis scaling. It is an invertible convention for the same determinant, rather than a period defined only up to a rational factor.
(10)
Every entry in the defining matrix has $n-2a\geq1$. Its integrand is $O(t^{2j}(1+|\log t|)^{n-a})$ at zero and $O(t^{2j-n/2}e^{-(n-2a)t})$ at infinity, so these are ordinary convergent integrals [4] [5].
(11)
Chuang–Yu denote the matrices for $n=2r-1$ and $n=2r$ by $M_r$ and $N_r$, respectively. Those names denote different matrices from Zhou’s odd-power determinant matrices in T647. The raw determinants are positive; the period signs and real or imaginary phases are retained.
Programs
(P1)
Python
import numberdb.sage as numberdb
from sage.rings.rational_field import QQ
from sage.rings.real_arb import RealBallField
from sage.rings.complex_arb import ComplexBallField

def even_power_determinant(n, normalization='bessel', digits=100):
    n = int(n)
    if n < 1 or normalization not in ('bessel', 'period'):
        raise ValueError('positive n; normalization bessel or period')
    r = (n+1)//2
    R = RealBallField(numberdb.bits(digits, losing=96))
    if n % 2:
        v = R.pi().sqrt()**(r*(r+1)) * R(2).sqrt()**(r*(r-3))
        for a in range(1,r):
            v *= R(a)**(r-a) / R(2*a+1).sqrt()**(2*a+1)
    else:
        v = R.pi().sqrt()**((r+1)**2) / (R(2).sqrt()**(r*(r+3))*R(QQ(r+1)/2).gamma())
        for a in range(1,r):
            v *= R(QQ((2*a+1)**(r-a))/QQ((a+1)**(a+1)))
    C = ComplexBallField(R.precision())
    if normalization == 'bessel':
        return C(v)
    s = r*(r-1)//2
    return C(v*(-1)**(n*r-s)*QQ(2)**(n*r-2*s)*R.pi()**s)*C.gen(0)**s

# Default selection: all n=1..16, bessel then period.
even_power_determinant(17, 'period')
References
[1]
Ping-Hsun Chuang and Jeng-Daw Yu, On the Periods of Twisted Moments of the Kloosterman Connection, The Ramanujan Journal 65 (2024), 1227–1277. ArXiv v2 (24 September 2024), §5.1, Corollary 26, Remark 27, §6.2 and Corollary 39; equation (29) is journal equation (A6). (arXiv) (doi)
[2]
Yajun Zhou, Wrońskian factorizations and Broadhurst–Mellit determinant formulae, Communications in Number Theory and Physics 12 (2018), 355–407, equations (1.3)–(1.6) and their proofs in §4. (arXiv)
Links
Similar tables
Bessel moments —   Individual integrals constituting these determinant matrices.
Broadhurst–Mellit determinants of Bessel moments —   Full odd-power determinants with positive first-kind exponents; the conversion formulas compare their different matrices.
Broadhurst–Mellit vacuum determinants —   Selected vacuum-containing constructions with odd powers and a special column combination at eight factors.
Bessel period determinants from odd rows —   First-kind row-parity selections with odd integration powers, rather than consecutive rows and even powers.
Bessel period determinants from even rows —   Even first-kind exponents with odd integration powers; its row parity differs from integration-power parity here.
Data properties
Entries are of type: complex number
digits: 100 (Unknown key)
Table is complete: no (it holds every factor count $1\leq n\leq16$, with bessel followed by period, covering matrix orders one through eight for both parities. This noncanonical initial-prefix cutoff bounds the independent integral checks; the mathematical family continues for all $n\geq1$)
How they were obtained:

The proved products in (2) and (3) are evaluated using exact integer and rational factors, positive real square roots, and Sage/Arb pi and gamma balls. The 100-digit computation uses 429 working bits, including 96 guard bits. The weakest nonzero component retains 419 relative accuracy bits before serialization. Every written last-place interval contains the full recomputation at 180 requested decimal digits. The exactly zero component follows from the explicit algebraic phase, not from rounding.

more

A separate mpmath implementation at 190 working decimal digits uses the coupled recurrences of [1], equation (29), starting with $\Delta_1=\pi/2$ and $\Delta_2=\pi^2/4$. Every result in both normalizations lies inside its recorded 100-digit interval. Its period scaling is computed by separate products of row and column factors.

All 16 defining moment matrices, 408 entries per working precision, were separately integrated by direct-$t$ tanh-sinh quadrature at 70 and 100 working decimal digits. The largest relative determinant errors against the recurrence values were $3.8e-52$ and $5.2e-81$, respectively. The quadrature checks use Arb Bessel midpoints, with $K_0$ evaluated through Tricomi $U$ [6]; they are heuristic checks, not integration certificates. The rigorous digit bounds come from the proved products. The four nonzero moment controls $\int K_0=\pi/2$, $\int K_0^2=\pi^2/4$, $\int tK_0=1$ and $\int tK_0^2=1/2$ pass. Wrong sign, phase, a perturbation at relative size $10^{-90}$, NaN, infinity and vacuously wide intervals are rejected.

The attached generator, argument-selection note, executable checks, moment checkpoints, numerical evidence and matching archived skill manifest preserve the computation. Reproduction against the generator is checked separately from mathematical agreement.