Harmonic central binomial sums

edit · history · discussion · files · short url · special functions special values number theory
Numbers
$p$
$v$ 
$B_p(v)$
1
-1/4:
-0.1567304709511661789319947212437763772496834847639958717038958155117600104369787604667995721838564607
1
-1/16:
-0.05417393243910282148284663273364110367101419320320254994577496017017234904483284442730647624438677075
1
1/16:
0.07437783618339224842947007984581029486712328051375951000318706984597394146744333362651987460370799512
1
1/4:
1.831931188354438030109207029864768221548298748563344268532996239243526039552509538958713025852230212
2
-1/4:
-0.06207962938306453393699511241870068283960292701117628361469869732493985068397560403243773504652736288
2
-1/16:
-0.01909347118450404687302205609950144004670753911377461649421308918122732637161302571741890508410811962
2
1/16:
0.02307349537156204243812970484136793633021794915765693565508735006966768717046911558157258002054951458
2
1/4:
0.1692381702818480097272409133777823650458577840665683645687188536056576585563078745494087373897028326
3
-1/4:
-0.02308519119997816131757992968588935505577396138603349645290097455883399445155851041620520378234784749
3
-1/16:
-0.006585176155472691535816464380757369557179141584493372307705835371641806704153198170935807837037742480
3
1/16:
0.007374550290713563326551564881628403797855437630101234432535898217835585981031955706113523323806644477
3
1/4:
0.03892579785134552088162555088933642739233691432237822636675709546560088571471091815146733433510046093
4
-1/4:
-0.008248117212742477496419466354741915958858913449290661754892183933292298112137417934811514180283722953
4
-1/16:
-0.002241272625870465611691810676103822852400040424392877459457376771642917848582697663803634046636623393
4
1/16:
0.002398419764208375621504630158593493098668207355827657222036369580742898844153802964564878678112196677
4
1/4:
0.01105380477997816472654939492432581264334160678148439606890371264237230535345840815464022217761521413
5
-1/4:
-0.002872573169930109467721781479772606471990512344314080736443825368930751623050911244205023270800276756
5
-1/16:
-0.0007566465151032541351721410027098483414101730324672402244301069332596283412252033110146573131464346790
5
1/16:
0.0007879957748873718486241467779058184183031959142758543955663132089027808295152301031061390385874265192
5
1/4:
0.003403014694624110857998582939592092057266189047632338294909582168575573161984445491418815120571673161
6
-1/4:
-0.0009842190601338027839071446763648206733369194210761863834981519897087811871905582460698903751502479842
6
-1/16:
-0.0002541732317107180531988313431562890798927576250467319703156285760343275447355242752185134176136629434
6
1/16:
0.0002604342283362406189853422671261058040942466631895027594125140779749383249042687994642266582375469094
6
1/4:
0.001087384536383712828334211222955932402146321787813763808390628033751148735886011510148759929045056369
7
-1/4:
-0.0003337385155755454734052743424777719978657179013424621195800396530734868081702085963281169254504123357
7
-1/16:
-0.00008512283483466389326940944000663959094480145724970045588429491897021399660187454753574285257137660031
7
1/16:
0.00008637405421606753028387227777936492892611899924631338347953956151457732448823679911389092194770200516
7
1/4:
0.0003540758400909289597757633927363314229919813082737948263237774675833182990750546554177043109516636490
8
-1/4:
-0.0001124300906013625210700126874853650417554602137538847229548098058382792013151400555571968952323189941
8
-1/16:
-0.00002845497148896421941726978441608265049408006074476226826132666065573000484332273277621432990083704164
8
1/16:
0.00002870510678692544748403973532970314317580257778613200565850401716725734085106572541805410281582727213
8
1/4:
0.0001164665707862156789528168065994525083711487783470084038538081758325486495637489682578049877986917087
Definition
The harmonic central binomial sum is $B_p(v)=\sum_{n=1}^{\infty}\binom{2n}{n}H_{2n}v^n/(2n+1)^p$, where $H_m=\sum_{k=1}^{m}1/k$, for integers $p\geq1$ and rational $-1/4\leq v\leq1/4$ [1], Lemma 2.2. The endpoint values are ordinary absolutely convergent series.
Parameters
$p$
—   denominator power ($p\geq1$)
$v$
—   signed series argument ($-1/4\leq v\leq1/4$)
Formulas
(1)
Put $s=\sqrt{1-4q}$, taking the positive real square root. For real $|q|<1/4$, $G(q)=\sum_{n=1}^{\infty}\binom{2n}{n}H_{2n}q^n=\log((1+s)/(2s^2))/s$ [1], (2.8). All logarithms are natural.
(2)
$B_p(v)=\dfrac1{(p-1)!}\int_0^{\infty}t^{p-1}e^{-t}G(ve^{-2t})\,dt$, by the logarithmic-moment identity [1], (2.1), and the generating function.
(3)
If $U_p(v)=\sum_{n=0}^{\infty}\binom{2n}{n}v^n/(2n+1)^p$ is the unweighted central binomial sum, then $\sum_{n=0}^{\infty}\binom{2n}{n}H_{2n+1}v^n/(2n+1)^p=B_p(v)+U_{p+1}(v)$, since $H_{2n+1}=H_{2n}+1/(2n+1)$.
(4)
$B_1(1/4)=2\operatorname{Im}\operatorname{Li}_2(i)=2G_{\mathrm{Cat}}$, where $G_{\mathrm{Cat}}$ is Catalan's constant. The identities [1], (1.1)–(1.2), become $9B_3(1/16)+41U_4(1/16)=40\beta(4)+5\pi\zeta(3)/12$ and $5B_2(-1/16)+17U_3(-1/16)=14\zeta(3)$, with $\beta(s)=\sum_{n=0}^{\infty}(-1)^n/(2n+1)^s$. Here $\zeta$ is the Riemann zeta function.
Comments
(5)
Here $H_0=0$, and the harmonic index is $2n$. A negative $v$ supplies $(-1)^n$; no absolute value or normalization factor is applied. In [1], Lemma 2.2 uses $v=\pm z^2$ and writes $zB_p(\pm z^2)$.
(6)
The arguments $v=\pm1/16$ occur in the alternating and nonalternating Apéry-like identities [1], (1.1)–(1.2). The arguments $v=\pm1/4$ are the two real convergence boundaries. For $p\geq1$, their absolute summands are $O(\log(n)/n^{p+1/2})$. The zero-argument specialization is $B_p(0)=0$.
Programs
(P1)
Python
import mpmath as mp
mp.mp.dps = 130

def harmonic_central_binomial_sum(p, v):
    # Moment of the harmonic generating function; p >= 1, |v| <= 1/4.
    def integrand(x):
        if x == 0:
            return mp.mpf(0)
        t = x*x
        s = mp.sqrt(1 - 4*v + 4*v*(-mp.expm1(-2*t)))
        kernel = mp.log((1+s)/(2*s*s))/s
        return 2*x**(2*p-1)*mp.exp(-t)*kernel
    return mp.quad(integrand, [0, 1, 2, 4, mp.inf])/mp.factorial(p-1)

harmonic_central_binomial_sum(4, mp.mpf(1)/16)
References
[1]
S. Charlton, H. Gangl, L. Lai, C. Xu and J. Zhao, On two conjectures of Sun concerning Apéry-like series, Forum Mathematicum 35 (2023), 1533–1547. (arXiv) (doi)
Links
Similar tables
Central binomial sums —   These are the unweighted companions $U_p(v)$, including the initial term 1; the harmonic-index shift uses $U_{p+1}$.
Log-sine-cosine integrals —   For $v>0$, Lemma 2.2 of [1] expresses $B_p(v)$ as finite linear combinations of mixed logarithmic integrals at $\theta$ and $2\theta$, with $\theta=\arcsin(2\sqrt v)$.
Log-sinh-cosh integrals —   For $v<0$, the hyperbolic counterpart uses $\theta=\operatorname{arsinh}(2\sqrt{-v})$. At $v=-1/16$, these integral endpoints are $\log\varphi$ and $2\log\varphi$, with $\varphi=(1+\sqrt5)/2$.
Clausen function values —   Clausen values provide Catalan's constant and $\beta(4)$ in the evaluations displayed here.
Riemann zeta values —   The Sun identities involve $\zeta(3)$.
Harmonic inverse binomial sums —   Harmonic insertions arise as coefficients in hypergeometric parameter expansions in both families. Here the summand uses $\binom{2n}{n}H_{2n}/(2n+1)^p$; the inverse family uses $H_{n-1}^{(r)}/(n^s\binom{2n}{n})$. The harmonic lattices, binomial placement and convergence intervals distinguish the quantities.
Data properties
Entries are of type: real number
Table is complete: no (it holds every $1\leq p\leq8$ at $v=-1/4,-1/16,1/16,1/4$. The interior pair is selected by the Apéry-like identities and the outer pair consists of both real convergence boundaries. The noncanonical order cutoff covers successive moments and both parities before extending any one slice. The attached argument-selection.md explains every parameter group)
How they were obtained:

All 32 entries carry 100 significant digits. Lemma 2.2 of [1] reduces the sums to mixed logarithmic integrals. The generator uses Arb polynomial arithmetic with analytically integrated logarithmic endpoint powers, and adds a uniform bound for the entire omitted series tail. Reflection handles the positive endpoint. Exact rational arguments, radicals, logarithms and all subsequent arithmetic carry their uncertainty. At 100 requested digits, 557 working bits and degree 605 retain at least 528 accuracy bits after all linear combinations. The attached computation.md derives the tail bounds and records the adapted numerical components. Every entry is recomputed at 140 requested digits.

more

An independent 500-bit analytic quadrature of the Mellin integral, with explicit bounds for both infinite tails, checks every decimal interval. Exact-rational coefficient sums also check all 16 interior entries. Catalan's value and both Sun identities are checked, including on the read-back stored intervals; wrong-sign, last-place-perturbation, non-finite and vacuous-radius controls must fail. The generator and checker share Arb and the generating-function identity, but no numerical generating code. The attached checks.py and numerical evidence record the executed checks.