Positive rational series for reciprocal powers of Catalan’s constant and Dirichlet beta values
Narendra Bhandari · 2026
Read Positive rational series for reciprocal powers of Catalan’s constant and Dirichlet beta values by Narendra Bhandari (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2608-00429.txt.
Introduction
Bhandari builds positive series for inverse powers of Dirichlet beta values. Beta is the alternating sum over odd denominators. Its value at two is the famed Catalan constant G. He gives two families of positive series for beta to the minus r. The first stems from Euler's classic transform and is taken at one half.
That family holds for real s above zero. A second transform, valid for s at least two, gives a faster series at one third. Finite sums and integrals fix the base terms. A positive recurrence and a product rule cover all inverse powers. For whole-number s, every term is a ratio of whole numbers. Error bounds and exact root rates of convergence are derived. Uses include inverse powers of the Catalan constant and of pi.
Catalog record
| CREATOR | Narendra Bhandari |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 43 KB |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2608.00429
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Positive rational series for reciprocal powers of Catalan’s constant and Dirichlet beta values Narendra Bhandari Abstract . Let β ( s ) = ∑ k = 0 ∞ ( − 1 ) k / ( 2 k + 1 ) s \beta(s)=\sum_{k=0}^{\infty}{(-1)^{k}}/{(2k+1)^{s}} and G = β ( 2 ) , G=\beta(2), where G G is Catalan’s constant. We develop two families of positive series for reciprocal powers β ( s ) − r \beta(s)^{-r} , where r ≥ 1 r\geq 1 is an integer. The first comes from the classical Euler transformation and is evaluated at 1 / 2 1/2 ; it is valid for real s > 0 s>0 . A second transformation, valid for s ≥ 2 s\geq 2 , produces a faster series evaluated at 1 / 3 1/3 . We obtain finite-sum and integral formulas for the base coefficients, together with a positive recurrence and a composition formula for all positive integral reciprocal powers. When s s is an integer, all coefficients are rational. We also give explicit remainder estimates and determine the exact root-convergence rates of the two families. As applications, we obtain positive rational series for every reciprocal power of Catalan’s constant and for reciprocal powers of π \pi arising from odd values of the Dirichlet beta function. 2020 Mathematics Subject Classification Primary 11Y60, 40G05; Secondary 11M06, 33B30 1. Introduction 1.1. Background and purpose The Dirichlet beta function is defined by (1.1) \beta(s)=\sum_{n=0}^{\infty}\frac{(-1)^{n}}{(2n+1)^{s}},\qquad s>0,\qquad G=\beta(2). The value G is Catalan’s constant. It has a simple alternating-series definition, but its arithmetic nature is still not understood: in particular, it is not known whether G is irrational. Catalan studied transformations of related series and definite integrals [ 8 ] . The inverse tangent integral is discussed in [ 18 , 15 , 7 ] . Accelerated series and rational approximations for Catalan’s constant are studied in [ 6 ] , while continued-fraction representations appear in [ 5 , 22 ] . Important arithmetic results for Catalan’s constant and related beta values were obtained by Rivoal and Zudilin [ 20 ] , Rivoal [ 19 ] , Nesterenko [ 16 ] , and Krattenthaler and Zudilin [ 13 ] . Zudilin [ 23 , Thm. 1] proved that at least one of \beta(2),\beta(4),\ldots,\beta(12) is irrational. These results do not, however, settle the irrationality of G , and that is not the aim of this paper. Here we study reciprocal powers \beta(s)^{-r},\qquad r\geq 1, where r is an integer. The Euler product for L(s,\chi_{4})=\beta(s) already gives a Dirichlet-series expansion for these quantities. We take a different approach. Our goal is to obtain ordinary power series whose terms are positive and, when s is an integer, rational. The coefficients are given by finite formulas and can be generated efficiently by a simple recurrence. The first construction is based on Euler’s transformation of an alternating series; a classical treatment is given by Hardy [ 11 , Ch. VIII] . Related acceleration methods were studied by Wynn [ 21 ] and Cohen et al. [ 9 ] . The reciprocal of a power series has its own long history, including work of Kaluza [ 12 ] , Lamperti [ 14 ] , and Baricz et al. [ 4 ] . 1.2. Main results The idea is straightforward. We begin with a generating function for the beta series and apply two fractional-linear substitutions. The classical Euler substitution t=w/(1-w) leads to a positive series evaluated at w=1/2 for every real s>0 . A second substitution, (1.2) t=\frac{2u}{1-u}, leads to a positive series evaluated at u=1/3 when s\geq 2 . The resulting representations are \frac{1}{\beta(s)^{r}}=\sum_{n=0}^{\infty}\frac{b_{s,r,n}}{2^{n}},\qquad\frac{1}{\beta(s)^{r}}=\sum_{n=0}^{\infty}\frac{c_{s,r,n}}{3^{n}}, with positive coefficients. Their exact root-convergence rates are 1/2 and 1/3 , respectively. We also give explicit remainder bounds, finite-sum and integral formulas for the base coefficients, and a recurrence and composition formula for their reciprocal powers. Two applications are especially simple to state. For every integer r\geq 1 , Corollary 5.2 gives \displaystyle\frac{1}{G^{r}} \displaystyle=1+\sum_{n=1}^{\infty}\frac{1}{3^{n}}\sum_{j=1}^{n}\binom{r+j-1}{j}\sum_{\begin{subarray}{c}n_{1}+\cdots+n_{j}=n\\ n_{i}\geq 1\end{subarray}}\prod_{\ell=1}^{j}\left[\sum_{k=1}^{n_{\ell}}\frac{(-1)^{k+1}2^{k}}{(2k+1)^{2}}\binom{n_{\ell}-1}{k-1}\right], and Corollary 6.3 gives \displaystyle\frac{4^{r}}{\pi^{r}} \displaystyle=1+\sum_{n=1}^{\infty}\frac{1}{2^{n}}\sum_{j=1}^{n}\binom{r+j-1}{j}\sum_{\begin{subarray}{c}n_{1}+\cdots+n_{j}=n\\ n_{i}\geq 1\end{subarray}}\prod_{\ell=1}^{j}\left[\sum_{k=1}^{n_{\ell}}\frac{(-1)^{k+1}}{2k+1}\binom{n_{\ell}-1}{k-1}\right]. Every term after the initial 1 is a positive rational number. The first formula is proved in Section 5 , and the second in Section 6 . The method is Euler-type rather than modular or hypergeometric in the sense of Ramanujan-type series. 2. Preliminary representations of beta values 2.1. A classical Dirichlet series Before beginning the positive power-series construction, we recall a standard Dirichlet-series identity. Apostol gives the case r=1 explicitly [ 3 , §11.4, Example 3] , and the general positive-integral case appears in Ng and Toma [ 17 , Eq. (5.3)] . With \chi=\chi_{4} and s=2 , the identity becomes a series for G^{-r} . This classical representation will serve only as a point of comparison for the positive series developed below. Let \chi_{4} be the primitive character modulo 4 : (2.1) \chi_{4}(n)=\begin{cases}0,&2\mid n,\\ 1,&n\equiv 1\pmod{4},\\ -1,&n\equiv 3\pmod{4}.\end{cases} Then \beta(s)=L(s,\chi_{4})=\sum_{n=1}^{\infty}\frac{\chi_{4}(n)}{n^{s}}. Let \mu denote the Möbius function. For arithmetic functions f and g , their Dirichlet convolution is defined by (f*g)(n)=\sum_{d\mid n}f(d)g(n/d). Define the r -fold Dirichlet convolution (2.2) \mu_{r}=\underbrace{\mu*\mu*\cdots*\mu}_{r\ {\rm factors}}. Theorem 2.1 . For every integer r\geq 1 and every real s>1 , (2.3) \frac{1}{\beta(s)^{r}}=\sum_{n=1}^{\infty}\frac{\chi_{4}(n)\mu_{r}(n)}{n^{s}}. The series in ( 2.3 ) converges absolutely. Moreover, \mu_{r} is multiplicative and (2.4) \mu_{r}(p^{a})=\begin{cases}(-1)^{a}\binom{r}{a},&0\leq a\leq r,\\ 0,&a>r.\end{cases} A proof is given in Appendix A . Corollary 2.2 . For every integer r\geq 1 , (2.5) \frac{1}{G^{r}}=\sum_{n=1}^{\infty}\frac{\chi_{4}(n)\mu_{r}(n)}{n^{2}}. Proof. The result follows immediately by taking s=2 in ( 2.3 ) and using \beta(2)=G . ∎ For r=1 and r=2 , this gives the following explicit Dirichlet series for reciprocal powers of Catalan’s constant: (2.6) \displaystyle\sum_{n=1}^{\infty}\frac{\chi_{4}(n)\mu_{1}(n)}{n^{2}} \displaystyle=1+\frac{1}{3^{2}}-\frac{1}{5^{2}}+\frac{1}{7^{2}}+\frac{1}{11^{2}}-\frac{1}{13^{2}}-\frac{1}{15^{2}}-\cdots=\frac{1}{G}, (2.7) \displaystyle\sum_{n=1}^{\infty}\frac{\chi_{4}(n)\mu_{2}(n)}{n^{2}} \displaystyle=1+\frac{2}{3^{2}}-\frac{2}{5^{2}}+\frac{2}{7^{2}}+\frac{1}{9^{2}}+\frac{2}{11^{2}}-\frac{2}{13^{2}}-\frac{4}{15^{2}}-\cdots=\frac{1}{G^{2}}. These formulas provide a useful comparison. Their signs do not follow a fixed alternating pattern, and their terms decay algebraically rather than geometrically. The construction below replaces them with positive rational terms and geometric convergence. 2.2. An integral representation Both transformations rest on a simple integral representation. For real s>0 , define (2.8) F_{s}(t)=\sum_{k=0}^{\infty}\frac{(-1)^{k}t^{k}}{(2k+1)^{s}},\qquad|t|<1. Proposition 2.3 . For each fixed real s>0 , the function F_{s}(t) extends analytically to the half-plane \RePart t>-1 . Then (2.9) F_{s}(t)=\frac{1}{\Gamma(s)}\int_{0}^{1}\frac{(-\log x)^{s-1}}{1+tx^{2}}\,\mathrm{d}x. In particular, F_{s}(1)=\beta(s). Proof. After the change of variable y=-\log x , we note that (2.10) \frac{1}{(2k+1)^{s}}=\frac{1}{\Gamma(s)}\int_{0}^{1}x^{2k}(-\log x)^{s-1}\,\mathrm{d}x. For |t|<1 , substitute ( 2.10 ) into ( 2.8 ). Uniform convergence of the geometric series on 0\leq x\leq 1 allows integration term by term, and hence \displaystyle F_{s}(t) \displaystyle=\frac{1}{\Gamma(s)}\int_{0}^{1}(-\log x)^{s-1}\sum_{k=0}^{\infty}(-tx^{2})^{k}\,\mathrm{d}x=\frac{1}{\Gamma(s)}\int_{0}^{1}\frac{(-\log x)^{s-1}}{1+tx^{2}}\,\mathrm{d}x. The same integral extends the function to the larger half-plane. Let K be a compact subset of \RePart t>-1 . Compactness gives a constant \delta_{K}>0 for which \RePart(1+tx^{2})=1+x^{2}\RePart t\geq\delta_{K}\qquad(t\in K,\ 0\leq x\leq 1). Consequently, |1+tx^{2}|\geq\delta_{K} and hence \left|\frac{(-\log x)^{s-1}}{1+tx^{2}}\right|\leq\frac{(-\log x)^{s-1}}{\delta_{K}}. The function on the right is integrable because (2.11) \displaystyle\int_{0}^{1}(-\log x)^{s-1}\,\mathrm{d}x=\Gamma(s)<\infty. Thus, on every compact subset of \RePart t>-1 , the integrand is bounded by an integrable function. Because it is analytic in t for each 0<x<1 , the integral defines an analytic function there. On |t|<1 it agrees with the original power series, so it is the required analytic continuation. It remains to identify the value at t=1 . For 0<\rho<1 , the power series and the integral agree at t=\rho . Since the alternating series for \beta(s) converges for every s>0 , Abel’s