betaFeedback

WorksMathematics

Discrete Hyperbolic Cosecant Distributions

Abstract

We introduce three families of discrete hyperbolic cosecant probability distributions, supported either bilaterally on Z\mathbb{Z} or unilaterally on N0\mathbb{N}_0, inspired by series evaluations of Srinivasa Ramanujan. We determine their exact normalization constants in terms of singular values of complete elliptic integrals and the Gaußian constant G=ϖ/πG=\varpi/π with the lemniscate constant Γ2(1/4)/(22π)Γ^2(1/4)/(2\sqrt{2π}) by explicitly calculating some series, e.,g. ∑κ∈Z2κcosech⁡(κπ)=G2+1π. \sum_{κ\in \mathbb{Z}} 2κ\operatorname{cosech}(κπ) = G^2 + \frac{1}π. Furthermore, by connecting these distributions to Fourier transforms of powers of the hyperbolic secant and applying the Poisson summation formula, we obtain closed-form evaluations for even moments and hyperbolic secant lattice sums, e.,g. ∑k∈Zsech4(kπ)=23(G4+G2+1π). \sum_{k \in \mathbb{Z}} \mathrm{sech}^4(k π) = \frac{2}{3} \left(G^4 + G^2 + \frac{1}π\right). This machinery proves conjectures on lattice sums arising in electrostatics proposed by Vigren and Dieckmann (Symmetry, 2020).

Introduction

We introduce three types of discrete hyperbolic cosecant probability distributions, denoted either by Cosech in the bilateral case supported on Z\mathbb{Z}, or by Cosech in the unilateral case supported on the non-negative natural numbers N0\mathbb{N}_0; this family of distributions is inspired by some series identities provided by Srinivasa Ramanujan. These distributions offer a view of the classical results through the lenses of probability theory and lemniscatic geometry.

Ramanujan’s formulae in his second notebook have a very general setup. They contain terms of Jacobi’s complete elliptic integral of the first and second kind, KK and EE, viz.

K(k):=∫0π/2dϑ1−k2sin⁡2(ϑ),E(k):=∫0π/21−k2sin⁡2(ϑ) dϑ.K(k):=\int_{0}^{\pi/2}\frac{\mathrm{d}\vartheta}{\sqrt{1-k^2\sin^2(\vartheta)}},\qquad E(k):=\int_{0}^{\pi/2}\sqrt{1-k^2\sin^2(\vartheta)}\,\mathrm{d}\vartheta.

Here, k∈(0,1)k \in(0,1) is called the elliptic modulus with complementary modulus k′k', such that k2+k′2=1k^2 + k'^2 = 1, cf. [7 p. 31]. For r>0r > 0, let kr∈(0,1)k_r \in(0,1) be the unique elliptic modulus with kr′:=1−kr2k'_r := \sqrt{1-k_r^2}, Kr:=K(kr)K_r := K(k_r), and Kr′=K(kr′)K'_r = K(k'_r), such that Kr′/Kr=rK'_r/K_r = \sqrt{r} holds. The elliptic lambda-star function then yields λ∗(r)=kr\lambda^*(r) = k_r. According to [4], in case of r∈Q+r \in\mathbb{Q}^{+}, the corresponding elliptic integrals KrK_r and Kr′K'_r can be expressed in closed form using a finite number of values of the gamma function. For r∈Nr \in\mathbb{N}, krk_r is called an elliptic singular modulus and the corresponding elliptic integrals KrK_r can often be calculated explicitly. Cf. [2 p. 298] for the values of KrK_r for the singular elliptic moduli k1,…,k16k_1,\ldots,k_{16}; numerous further values of singular elliptic moduli are stated in the exercises in the same book. For the calculation of Er:=E(kr)E_r := E(k_r), we use the elliptic alpha function α(r)\alpha(r) which is implicitly defined via Equation (5.1.2) in [2 p. 152]; it yields

Er=π4Krr+(1−α(r)r)Kr(1)E_r = \frac{\pi}{4K_r\sqrt{r}}+\left(1-\frac{\alpha(r)}{\sqrt{r}}\right)K_r \tag*{(1)}

and the Tables 5.2a., 5.2b., and 5.3. on p. 172–173 in the same book provide many special values for α(r)\alpha(r) for r∈Q+r \in\mathbb{Q}^{+}. For elliptic singular values krk_r, Ramanujan’s notation in the Chapters 17 and 18 of his second notebook can be translated into the common language via x=kr2x=k_r^2, y=rπy=\sqrt{r}\pi, and z=2Kr/πz=2K_r/\pi, cf. Entry 6 of Chapter 17 in [9 p. 102].

Let ϖ:=Γ2(1/4)/(4π)\varpi:= \Gamma^2(1/4)/(4\sqrt{\pi}) denote the lemniscate constant. A particularly important elliptic singular modulus occurs in the lemniscatic case where r=1r=1. Then, we get K1′=K1=ϖ/2K'_1=K_1=\varpi/\sqrt{2} and λ∗(1)=k1=k1′=1/2\lambda^*(1)=k_1=k'_1=1/\sqrt{2}. Hence, in the lemniscatic case, we may substitute x=1/2x=1/2, y=πy=\pi and z=2Gz=\sqrt{2}G into Ramanujan’s equations. Finally, Gauß’ constant G:=ϖ/πG:=\varpi/\pi yields particularly simple expressions for KrK_r, e.g. for its values at singular elliptic moduli krk_r for r=s2r=s^2 with s∈{1,…,5}s \in\{1,\ldots,5\}. In this case, also the normalization constants for our discrete type I and type III hyperbolic cosecant distribution are given in notably simple closed forms.

The Discrete Hyperbolic Cosecant Distribution – Type I

We introduce the discrete type I hyperbolic cosecant distribution of the first, second, and third kind which are mainly based on the hyperbolic cosecant series identities given in [9] and [3].

We begin with the basic case, viz. the bilateral lemniscatic Cosech distribution. Let w(x):=xcosech⁡(xπ/2)w(x):=x\operatorname{cosech}(x\pi/2) be the underlying continuous real weight function; we replace its removable singularity in 00 by setting w(0):=lim⁡x→0w(x)=2/πw(0):=\lim_{x\to0}w(x)=2/\pi. The discrete weight functions wX1w_{X_1}, wY1w_{Y_1}, and wZ1w_{Z_1} are obtained by sampling the continuous weight function ww on the corresponding even, odd, and integer lattices, respectively. Specifically, for κ∈Z\kappa\in\mathbb{Z}, define wX1(κ):=w(2κ)w_{X_1}(\kappa):=w(2\kappa), wY1(κ):=w(2κ+1)w_{Y_1}(\kappa):=w(2\kappa+1), and wZ1(κ):=w(κ)w_{Z_1}(\kappa):=w(\kappa) for the type I Cosech distributions of the first, second, and third kind, respectively. The proofs for the normalization constants of the lemniscatic type I Cosech distribution are based on quasi-modularity and a duplication formula for the Eisenstein series E2E_2 for the first kind; we cite Ramanujan’s formula (xiii) in Entry 15 in Chapter 17 of his second notebook for the second kind, cf. [9 p. 133]; finally, a simple parity split yields the normalization constant for the third kind. For the type I Cosech distribution of the first kind, the underlying weight function is

wX1(x):=w(2x)w_{X_1}(x):=w(2x). Its simple Fourier transform, viz.

F[wX1](ξ):=∫−∞∞2xcosech⁡(πx)e−2πixξ dx=sech⁡2(πξ),\mathcal{F}[w_{X_1}](\xi):=\int_{-\infty}^{\infty}2x\operatorname{cosech}(\pi x)e^{-2\pi i x\xi}\,\mathrm{d}x=\operatorname{sech}^{2}(\pi\xi),

will be particularly useful in the context of Poisson summation. We therefore call the CosechX1_{X_1} distribution simply the discrete hyperbolic cosecant distribution Cosech.

Proposition and Definition 2.1. [Lemniscatic Type I Cosech Distributions.] Let κ∈Z\kappa\in\mathbb{Z}. For the rows 2–4 of Table 1, let CAC_A denote the normalization constant of the listed CosechA_A distribution with weight function wA(κ)w_A(\kappa). Then the associated sequence (pA(κ))κ∈Z:=(CA⋅wA(κ))κ∈Z(p_A(\kappa))_{\kappa\in\mathbb{Z}}:=(C_A\cdot w_A(\kappa))_{\kappa\in\mathbb{Z}} is a probability mass function. We name CosechX1_{X_1}, CosechY1_{Y_1}, and CosechZ1_{Z_1} the discrete lemniscatic type I hyperbolic cosecant distribution of the first, second, and third kind, respectively. We omit adjectives when the context is clear. We call the CosechX1_{X_1} distribution simply the discrete hyperbolic cosecant distribution Cosech.

TypeDistributionNormalizing ConstantWeight Function
ICosech == CosechX1_{X_1}ππG2+1\frac{\pi}{\pi G^2+1}2κcosech⁡(κπ)2\kappa\operatorname{cosech}(\kappa\pi)
ICosechY1_{Y_1}12G2\frac{1}{\sqrt{2}G^2}(2κ+1)cosech⁡((2κ+1)π2)(2\kappa+1)\operatorname{cosech}\left((2\kappa+1)\frac{\pi}{2}\right)
ICosechZ1_{Z_1}ππG2+1+2πG2\frac{\pi}{\pi G^2+1+\sqrt{2}\pi G^2}κcosech⁡(κπ2)\kappa\operatorname{cosech}\left(\kappa\frac{\pi}{2}\right)
ICosechXr_{X_r}rπ2π+4α(r)Kr2\frac{\sqrt{r}\pi^2}{\pi+4\alpha(r)K_r^2}2κcosech⁡(rκπ)2\kappa\operatorname{cosech}\left(\sqrt{r}\kappa\pi\right)
ICosechYr_{Y_r}π24krKr2\frac{\pi^2}{4k_rK_r^2}(2κ+1)cosech⁡(r(2κ+1)π2)(2\kappa+1)\operatorname{cosech}\left(\sqrt{r}(2\kappa+1)\frac{\pi}{2}\right)
ICosechZr_{Z_r}rπ2π+4α(r)Kr2+4rkrKr2\frac{\sqrt{r}\pi^2}{\pi+4\alpha(r)K_r^2+4\sqrt{r}k_rK_r^2}κcosech⁡(rκπ2)\kappa\operatorname{cosech}\left(\sqrt{r}\kappa\frac{\pi}{2}\right)
ICosechX1_{X_1}2ππG2+3\frac{2\pi}{\pi G^2+3}(2n+2)cosech⁡((2n+2)π2)(2n+2)\operatorname{cosech}\left((2n+2)\frac{\pi}{2}\right)
ICosechY1_{Y_1}2G2\frac{\sqrt{2}}{G^2}(2n+1)cosech⁡((2n+1)π2)(2n+1)\operatorname{cosech}\left((2n+1)\frac{\pi}{2}\right)
ICosechZ1_{Z_1}2π(1+2)πG2+3\frac{2\pi}{(1+\sqrt{2})\pi G^2+3}(n+1)cosech⁡((n+1)π2)(n+1)\operatorname{cosech}\left((n+1)\frac{\pi}{2}\right)
ICosechXr_{X_r}2rπ24α(r)Kr2+3π\frac{2\sqrt{r}\pi^2}{4\alpha(r)K_r^2+3\pi}(2n+2)cosech⁡(r(2n+2)π)(2n+2)\operatorname{cosech}\left(\sqrt{r}(2n+2)\pi\right)
ICosechYr_{Y_r}π22krKr2\frac{\pi^2}{2k_rK_r^2}(2n+1)cosech⁡(r(2n+1)π2)(2n+1)\operatorname{cosech}\left(\sqrt{r}(2n+1)\frac{\pi}{2}\right)
ICosechZr_{Z_r}2rπ24α(r)Kr2+4rkrKr2+3π\frac{2\sqrt{r}\pi^2}{4\alpha(r)K_r^2+4\sqrt{r}k_rK_r^2+3\pi}(n+1)cosech⁡(r(n+1)π2)(n+1)\operatorname{cosech}\left(\sqrt{r}(n+1)\frac{\pi}{2}\right)

Table 1. Discrete bilateral (lemniscatic) type I Cosech distributions of the first (XX), second (YY), and third (ZZ) kind for κ∈Z\kappa\in\mathbb{Z} and r>0r>0 (r=1r=1); further the discrete unilateral (lemniscatic) type I Cosech distributions of the first (XX), second (YY), and third (ZZ) kind for n∈N0n \in\mathbb{N}_0 and r>0r>0 (r=1r=1).

Proof. The lemniscatic normalization constants could easily be deduced from the general elliptic formulae presented in Proposition 2.3, by setting Kr′/Kr=r=1K'_r/K_r=\sqrt{r}=1. But we want to examine some connections of the Cosech distribution to Eisenstein series and Jacobi theta functions, and further to Ramanujan’s work; hence, now we give separate proofs for the lemniscatic case.

Ad CosechX1_{X_1}: Unfortunately, the normalization constant CX1C_{X_1} is not provided by Entry 15 in Chapter 17 of Ramanujan’s second notebook, cf. [9 p. 132–133]. Instead, Ramanujan states here formulae for odd moments ≥3\geq3 of the Cosech distribution. We calculate the normalization constant CX1C_{X_1} as follows: First, we express wX1(n)w_{X_1}(n) in terms of two Eisenstein series, viz. E2(i)E_2(i) and E2(i/2)E_2(i/2). Due to the quasi-modularity of E2E_2, we can directly evaluate E2(i)=3/πE_2(i)=3/\pi and write E2(i/2)E_2(i/2) in terms of E2(2i)E_2(2i). Finally, we determine the value of E2(2i)E_2(2i) by showing a duplication formula. This includes Jacobi theta functions whose values we can directly compute. In detail: For every n∈Nn\in\mathbb{N}, by definition we have

sinh⁡(nπ)=eπn−e−πn2=1−e−2πn2e−πn.\sinh(n\pi)=\frac{e^{\pi n}-e^{-\pi n}}{2}=\frac{1-e^{-2\pi n}}{2e^{-\pi n}}.

Using the identity

x1+x=x1−x−2x21−x2(2)\frac{x}{1+x}=\frac{x}{1-x}-\frac{2x^2}{1-x^2} \tag*{(2)}

we can write

∑n=1∞2ncosech⁡(nπ)=4∑n=1∞ne−πn1−e−2πn=4(∑n=1∞ne−πn1−e−πn−∑n=1∞ne−2πn1−e−2πn).(3)\sum_{n=1}^{\infty}2n\operatorname{cosech}(n\pi) =4\sum_{n=1}^{\infty}\frac{ne^{-\pi n}}{1-e^{-2\pi n}} =4\left(\sum_{n=1}^{\infty}\frac{ne^{-\pi n}}{1-e^{-\pi n}}-\sum_{n=1}^{\infty}\frac{ne^{-2\pi n}}{1-e^{-2\pi n}}\right). \tag*{(3)}

Let τ∈H={z∈C:ℑ(z)>0}\tau\in\mathbb{H} = \{z \in\mathbb{C} : \Im(z) > 0\} lie in the upper half-plane and define the nome qτ=eiπτq_{\tau} = e^{i\pi\tau}. The Eisenstein series E2E_{2} has the Lambert-series expansion

E2(τ)=1−24∑n=1∞nqτ2n1−qτ2n.E_{2}(\tau) = 1 - 24 \sum_{n=1}^{\infty} \frac{nq_{\tau}^{2n}}{1-q_{\tau}^{2n}}.

Hence,

∑n=1∞2ncosech⁡(nπ)=16(E2(i)−E2(i2)).\sum_{n=1}^{\infty} 2n\operatorname{cosech}(n\pi) = \frac{1}{6}\left(E_{2}(i)-E_{2}\left(\frac{i}{2}\right)\right).

Let G2(τ)=2ζ(2)E2(τ)G_{2}(\tau) = 2\zeta(2)E_{2}(\tau) be a scaled version of E2E_{2} by the factor ζ(2)=π2/6\zeta(2) = \pi^{2}/6 of the Riemann zeta function. The quasi-modular transformation formula

G2(−1τ)=τ2G2(τ)−2πiτ⟺E2(−1τ)=τ2E2(τ)+6τπiG_{2}\left(-\frac{1}{\tau}\right) = \tau^{2}G_{2}(\tau)-2\pi i\tau\qquad\Longleftrightarrow\qquad E_{2}\left(-\frac{1}{\tau}\right) = \tau^{2}E_{2}(\tau)+\frac{6\tau}{\pi i}

is given by Equation (50) on p. 69 in [1]. This formula directly yields the values E2(i)=3/πE_{2}(i) = 3/\pi, and E2(i/2)=12/π−4E2(2i)E_{2}(i/2) = 12/\pi- 4E_{2}(2i). Hence, we get

∑n=1∞2ncosech⁡(nπ)=23E2(2i)−32π.\sum_{n=1}^{\infty} 2n\operatorname{cosech}(n\pi) = \frac{2}{3}E_{2}(2i)-\frac{3}{2\pi}.

Now, we represent the Eisenstein series E2(2τ)E_{2}(2\tau) by a duplication formula which includes E2(τ)E_{2}(\tau) and the Jacobi theta functions ϑ3\vartheta_{3} and ϑ4\vartheta_{4}. Those are defined by

ϑ3(qτ)=∑k∈Zqτk2,ϑ4(qτ)=∑k∈Z(−1)kqτk2,\vartheta_{3}(q_{\tau}) = \sum_{k\in\mathbb{Z}}q_{\tau}^{k^{2}}, \qquad\vartheta_{4}(q_{\tau}) = \sum_{k\in\mathbb{Z}}(-1)^{k}q_{\tau}^{k^{2}},

cf. [11 p. 464]. The duplication formula reads as follows.

Lemma 2.2 (Duplication formula). Let τ∈H={z∈C:ℑ(z)>0}\tau\in\mathbb{H} = \{z \in\mathbb{C} : \Im(z) > 0\} and define the nome qτ=eiπτq_{\tau} = e^{i\pi\tau}. Then

E2(2τ)=12E2(τ)+14(ϑ34(qτ)+ϑ44(qτ))E_{2}(2\tau) = \frac{1}{2}E_{2}(\tau)+\frac{1}{4}\left(\vartheta_{3}^{4}(q_{\tau})+\vartheta_{4}^{4}(q_{\tau})\right)

holds.

Proof. According to [2 Proposition 3.2, pp. 70–71], the fourth powers of the Jacobi theta functions admit the Lambert-series expansions

ϑ34(qτ)=1+8∑n=1∞2nqτ2n1+qτ2n+8∑n=1∞(2n−1)qτ2n−11−qτ2n−1\vartheta_{3}^{4}(q_{\tau}) = 1 + 8\sum_{n=1}^{\infty}\frac{2nq_{\tau}^{2n}}{1+q_{\tau}^{2n}} + 8\sum_{n=1}^{\infty}\frac{(2n-1)q_{\tau}^{2n-1}}{1-q_{\tau}^{2n-1}}

and

ϑ44(qτ)=1+8∑n=1∞2nqτ2n1+qτ2n−8∑n=1∞(2n−1)qτ2n−11+qτ2n−1.\vartheta_{4}^{4}(q_{\tau}) = 1 + 8\sum_{n=1}^{\infty}\frac{2nq_{\tau}^{2n}}{1+q_{\tau}^{2n}} - 8\sum_{n=1}^{\infty}\frac{(2n-1)q_{\tau}^{2n-1}}{1+q_{\tau}^{2n-1}}.

Adding both identities yields

ϑ34(qτ)+ϑ44(qτ)=2+32∑n=1∞nqτ2n1+qτ2n+16∑n=1∞(2n−1)qτ2(2n−1)1−qτ2(2n−1).\vartheta_3^4(q_\tau)+\vartheta_4^4(q_\tau)=2+32\sum_{n=1}^{\infty}\frac{nq_\tau^{2n}}{1+q_\tau^{2n}}+16\sum_{n=1}^{\infty}\frac{(2n-1)q_\tau^{2(2n-1)}}{1-q_\tau^{2(2n-1)}}.

Applying equation (2) to the first series and splitting the second series into odd and even indices gives

ϑ34(qτ)+ϑ44(qτ)=2+48∑n=1∞nqτ2n1−qτ2n−96∑n=1∞nqτ4n1−qτ4n.\vartheta_3^4(q_\tau)+\vartheta_4^4(q_\tau)=2+48\sum_{n=1}^{\infty}\frac{nq_\tau^{2n}}{1-q_\tau^{2n}}-96\sum_{n=1}^{\infty}\frac{nq_\tau^{4n}}{1-q_\tau^{4n}}.

Since E2(τ)=1−24∑n=1∞nqτ2n1−qτ2nE_2(\tau)=1-24\sum_{n=1}^{\infty}\frac{nq_\tau^{2n}}{1-q_\tau^{2n}} and E2(2τ)=1−24∑n=1∞nqτ4n1−qτ4nE_2(2\tau)=1-24\sum_{n=1}^{\infty}\frac{nq_\tau^{4n}}{1-q_\tau^{4n}}, we conclude

ϑ34(qτ)+ϑ44(qτ)=4E2(2τ)−2E2(τ),\vartheta_3^4(q_\tau)+\vartheta_4^4(q_\tau)=4E_2(2\tau)-2E_2(\tau),

and the claim follows.

Now, we use the well-known identities ϑ32(qτ)=2K(k)/π\vartheta_3^2(q_\tau)=2K(k)/\pi and ϑ42(qτ)=2k′K(k)/π\vartheta_4^2(q_\tau)=2k'K(k)/\pi, e. g. see Equation (5) in [Zucker, 1984]. In the lemniscatic case, where λ∗(1)=k1=1/2\lambda^*(1)=k_1=1/\sqrt{2} and K(1/2)=ϖ/2K(1/\sqrt{2})=\varpi/\sqrt{2}, this results in ϑ34(qi)=2G2\vartheta_3^4(q_i)=2G^2 and ϑ44(qi)=G2\vartheta_4^4(q_i)=G^2. Together with the Eisenstein value E2(i)=3/πE_2(i)=3/\pi from above, we get E2(2i)=3/(2π)+3G2/4E_2(2i)=3/(2\pi)+3G^2/4. Hence,

∑n=1∞2ncosech⁡(nπ)=πG2−12π.\sum_{n=1}^{\infty}2n\operatorname{cosech}(n\pi)=\frac{\pi G^2-1}{2\pi}.

Since by definition, we have wX1(0)=2/πw_{X_1}(0)=2/\pi, we finally deduce

∑κ∈Z2κcosech⁡(κπ)=2π+2∑n=1∞2ncosech⁡(nπ)=G2+1π⟺CX1=ππG2+1,(4)\sum_{\kappa\in\mathbb{Z}}2\kappa\operatorname{cosech}(\kappa\pi)=\frac{2}{\pi}+2\sum_{n=1}^{\infty}2n\operatorname{cosech}(n\pi)=G^2+\frac{1}{\pi}\quad\Longleftrightarrow\quad C_{X_1}=\frac{\pi}{\pi G^2+1}, \tag*{(4)}

as claimed.

Ad CY1C_{Y_1}: In the lemniscatic case, we translate Ramanujan’s formulae in Ramanujan’s Entry 15 in Chapter 17 of his second notebook, cf. [9 p. 132–133] into the common language via x=1/2x=1/2, y=πy=\pi, and z=2Gz=\sqrt{2}G. Now, Equation (xiii) in this Entry reads as

∑n=0∞2n+1sinh⁡(12(2n+1)y)=12z2x=∑n=0∞(2n+1)cosech⁡((2n+1)π2)=G22.\sum_{n=0}^{\infty}\frac{2n+1}{\sinh\left(\frac{1}{2}(2n+1)y\right)}=\frac{1}{2}z^2\sqrt{x}=\sum_{n=0}^{\infty}(2n+1)\operatorname{cosech}\left(\frac{(2n+1)\pi}{2}\right)=\frac{G^2}{\sqrt{2}}.

Finally,

∑κ∈Z(2κ−1)cosech⁡((2κ−1)π2)=2∑n=0∞(2n+1)cosech⁡((2n+1)π2)=2G2\sum_{\kappa\in\mathbb{Z}}(2\kappa-1)\operatorname{cosech}\left(\frac{(2\kappa-1)\pi}{2}\right)=2\sum_{n=0}^{\infty}(2n+1)\operatorname{cosech}\left(\frac{(2n+1)\pi}{2}\right)=\sqrt{2}G^2

yields the desired normalization constant CY1=1/(2G2)C_{Y_1}=1/(\sqrt{2}G^2).

Ad CZ1C_{Z_1}: Pairing even and odd integers yields

∑κ∈Zκcosech⁡(κπ2)=∑κ∈Z2κcosech⁡(2κπ2)+∑κ∈Z(2κ−1)cosech⁡((2κ−1)π2)=(1+2)G2+1π,\begin{aligned} \sum_{\kappa\in\mathbb{Z}} \kappa\operatorname{cosech}\left(\frac{\kappa\pi}{2}\right) = \sum_{\kappa\in\mathbb{Z}} 2\kappa\operatorname{cosech}\left(\frac{2\kappa\pi}{2}\right) + \sum_{\kappa\in\mathbb{Z}} (2\kappa-1)\operatorname{cosech}\left(\frac{(2\kappa-1)\pi}{2}\right) \\ &= \left(1+\sqrt{2}\right)G^2+\frac{1}{\pi}, \end{aligned}

due to absolute convergence of each series. Hence, we get

CZ1=π(1+2)πG2+1C_{Z_1}=\frac{\pi}{(1+\sqrt{2})\pi G^2+1}

as stated.

Next, we observe the discrete type I Cosech distribution in the general elliptic setting.

Proposition and Definition 2.3. [Type I Cosech Distributions.] Let κ∈Z\kappa\in\mathbb{Z} and r>0r>0. For the rows 5–7 in Table 1, let CAC_A denote the normalization constant of the listed Cosech⁡A\operatorname{Cosech}_A distribution with weight function wA(κ)w_A(\kappa). Then the associated sequence (pA(κ))κ∈Z:=(CA⋅wA(κ))κ∈Z(p_A(\kappa))_{\kappa\in\mathbb{Z}} := (C_A\cdot w_A(\kappa))_{\kappa\in\mathbb{Z}} is a probability mass function.

We name Cosech⁡Xr\operatorname{Cosech}_{X_r}, Cosech⁡Yr\operatorname{Cosech}_{Y_r}, and Cosech⁡Zr\operatorname{Cosech}_{Z_r} the discrete type I hyperbolic cosecant distribution of the first, second, and third kind, respectively. We omit adjectives when the context is clear.

Proof. Ad CXrC_{X_r}: For the calculation of the normalization constant CXrC_{X_r}, we follow the method in [3]. We use the Jacobi elliptic function dn⁡\operatorname{dn} and the Lambert-series expansion of dn⁡2\operatorname{dn}^2 for the nome e−πK′/Ke^{-\pi K'/K}, viz.

dn⁡2(u,k)=EK+2π2K2∑n=1∞nqn1−q2ncos⁡(uKπn),\operatorname{dn}^2(u,k)=\frac{E}{K}+\frac{2\pi^2}{K^2}\sum_{n=1}^{\infty}\frac{nq^n}{1-q^{2n}}\cos\left(\frac{u}{K}\pi n\right),

cf. Equation (14) in [3 p. 294]. Since dn⁡(0,k)=1\operatorname{dn}(0,k)=1, a modification of equation (3) yields the identities

4∑n=1∞nqn1−q2n=∑n=1∞2ncosech⁡(K′Knπ)=2K(K−E)π2.(5)4\sum_{n=1}^{\infty}\frac{nq^n}{1-q^{2n}}=\sum_{n=1}^{\infty}2n\operatorname{cosech}\left(\frac{K'}{K}n\pi\right)=\frac{2K(K-E)}{\pi^2}. \tag*{(5)}

cf. Equation (25) in [3 p. 294]. We choose the elliptic lambda-star function λ∗(r)=kr\lambda^*(r)=k_r such that K(kr′)/K(kr)=rK(k'_r)/K(k_r)=\sqrt{r}. By definition of the weight function in 0 and due to Equation (1), we get

∑κ∈Z2κcosech⁡(rκπ)=2rπ+2∑n=1∞2ncosech⁡(rnπ)=2π+4rKr(Kr−Er)rπ2=π+4α(r)Kr2rπ2.\begin{aligned} \sum_{\kappa\in\mathbb{Z}}2\kappa\operatorname{cosech}\left(\sqrt{r}\kappa\pi\right) = \frac{2}{\sqrt{r}\pi}+2\sum_{n=1}^{\infty}2n\operatorname{cosech}\left(\sqrt{r}n\pi\right) \\ &= \frac{2\pi+4\sqrt{r}K_r(K_r-E_r)}{\sqrt{r}\pi^2} = \frac{\pi+4\alpha(r)K_r^2}{\sqrt{r}\pi^2}. \end{aligned}

This provides the desired normalization constant

CXr=rπ2π+4α(r)Kr2.C_{X_r} = \frac{\sqrt{r}\pi^2}{\pi+ 4\alpha(r)K_r^2}.

Ad CYrC_{Y_r}: Equation (xiii) of Ramanujan’s Entry 15 in Chapter 17 of his second notebook, cf. [9 p. 133], can be translated into the common language via x=kr2x = k_r^2, y=πry = \pi\sqrt{r}, and z=2Kr/πz = 2K_r/\pi, cf. [9 p. 102], viz.

∑κ=0∞2κ+1sinh⁡(12(2κ+1)y)=12z2x=∑n=0∞(2n+1)cosech⁡(r(2n+1)π2)=2krKr2π2.\sum_{\kappa=0}^{\infty}\frac{2\kappa+1}{\sinh\left(\frac{1}{2}(2\kappa+1)y\right)} = \frac{1}{2}z^2\sqrt{x} = \sum_{n=0}^{\infty}(2n+1)\operatorname{cosech}\left(\sqrt{r}(2n+1)\frac{\pi}{2}\right) = \frac{2k_rK_r^2}{\pi^2}.

Now,

∑κ∈Z(2κ+1)cosech⁡(r(2κ+1)π2)=4krKr2π2\sum_{\kappa\in\mathbb{Z}}(2\kappa+1)\operatorname{cosech}\left(\sqrt{r}(2\kappa+1)\frac{\pi}{2}\right) = \frac{4k_rK_r^2}{\pi^2}

yields the normalization constant

CYr=π24krKr2,C_{Y_r} = \frac{\pi^2}{4k_rK_r^2},

as claimed.

Ad CZrC_{Z_r}: Pairing odd and even integers yields

∑κ∈Zκcosech⁡(rκπ2)=∑κ∈Z2κcosech⁡(r(2κ)π2)+∑κ∈Z(2κ+1)cosech⁡(r(2κ+1)π2)=π+4α(r)Kr2+4rkrKr2rπ2.\begin{aligned} \sum_{\kappa\in\mathbb{Z}}\kappa\operatorname{cosech}\left(\sqrt{r}\kappa\frac{\pi}{2}\right) = \sum_{\kappa\in\mathbb{Z}}2\kappa\operatorname{cosech}\left(\sqrt{r}(2\kappa)\frac{\pi}{2}\right) + \sum_{\kappa\in\mathbb{Z}}(2\kappa+1)\operatorname{cosech}\left(\sqrt{r}(2\kappa+1)\frac{\pi}{2}\right) \\ &= \frac{\pi+4\alpha(r)K_r^2+4\sqrt{r}k_rK_r^2}{\sqrt{r}\pi^2}. \end{aligned}

We deduce the normalization constant

CZr=rπ2π+4α(r)Kr2+4rkrKr2,C_{Z_r} = \frac{\sqrt{r}\pi^2}{\pi+4\alpha(r)K_r^2+4\sqrt{r}k_rK_r^2},

as stated.

In [9] and [3], the hyperbolic series are formulated as a summation over N0\mathbb{N}_0. Hence, the proof of Proposition 2.3 directly yields a unilateral version Cosech of our type I Cosech distribution. For the lemniscatic case, we simply insert r=1r=1, α(1)=1/2\alpha(1)=1/2, k1=1/2k_1=1/\sqrt{2}, and K1=ϖ/2K_1=\varpi/\sqrt{2} into the general elliptic formulae.

Corollary and Definition 2.4. [(Lemniscatic) Type I Cosech Distribution.] Let n∈N0n\in\mathbb{N}_0 and r>0r>0 (r=1r=1). For the rows 8–14 of Table 1, let CAC_A denote the normalization constants of the listed CosechA_A distribution with weight function wA(n)w_A(n). Then the associated sequence (pA(n))n∈N0:=(CA⋅wA(n))n∈N0(p_A(n))_{n\in\mathbb{N}_0}:=(C_A\cdot w_A(n))_{n\in\mathbb{N}_0} is a probability mass function.

We name Cosech‾Xr\underline{\mathrm{Cosech}}_{X_r}, Cosech‾Yr\underline{\mathrm{Cosech}}_{Y_r}, and Cosech‾Zr\underline{\mathrm{Cosech}}_{Z_r} the discrete unilateral (lemniscatic for r=1r=1) type I hyperbolic cosecant distribution of the first, second, and third kind, respectively. We omit adjectives when the context is clear.

The Discrete Hyperbolic Cosecant Distribution – Type II

Now, we introduce the discrete type II hyperbolic cosecant distribution, building on the foundational work of [9], [3], and [Zucker, 1984].

For the type I Cosech distribution, the underlying continuous weight function w(x)=xcosech⁡(xπ/2)w(x)=x\operatorname{cosech}(x\pi/2) has a removable singularity at 00. The xx factor is necessary to make ww positive and even on its domain Z\mathbb{Z}. In contrast, for the type II Cosech distribution, we use the underlying continuous real weight function wII(x):=cosech⁡(xπ/2)/xw_{\mathrm{II}}(x):=\operatorname{cosech}(x\pi/2)/x which has a non-removable double pole at 00. Hence, we first consider the unilateral type II Cosech distributions on N0\mathbb{N}_0 after index shift. We prove the general elliptic formulae and for the lemniscatic case where r=1r=1, we insert the elliptic singular modulus k1=1/2k_1=1/\sqrt{2} into the formulae.

Proposition and Definition 3.1. [(Lemniscatic) Type II Cosech Distributions.] Let n∈N0n \in\mathbb{N}_0 and r>0r > 0 (r=1r = 1). For the rows 2–7 of Table 2, let CAC_A denote the normalization constant of the listed CosechA_A distribution with weight function wA(n)w_A(n). Then the associated sequence (pA(n))n∈N0:=(CA⋅wA(n))n∈N0(p_A(n))_{n\in\mathbb{N}_0} := (C_A \cdot w_A(n))_{n\in\mathbb{N}_0} is a probability mass function.

We name CosechUr_{U_r}, CosechVr_{V_r}, and CosechWr_{W_r} the discrete (lemniscatic for r=1r = 1) unilateral type II hyperbolic cosecant distribution of the first, second, and third kind, respectively. We omit adjectives when the context is clear.

TypeDistributionNormalizing ConstantWeight Function
IICosech‾U1\underline{\mathrm{Cosech}}_{U_1}24π−ln⁡(8)\frac{24}{\pi-\ln(8)}cosech⁡((n+1)π)/(2n+2)\operatorname{cosech}((n+1)\pi)/(2n+2)
IICosech‾V1\underline{\mathrm{Cosech}}_{V_1}2ln⁡(1+2)=2arsinh⁡(1)\frac{2}{\ln(1+\sqrt{2})}=\frac{2}{\operatorname{arsinh}(1)}cosech⁡((2n+1)π2)/(2n+1)\operatorname{cosech}((2n+1)\frac{\pi}{2})/(2n+1)
IICosech‾W1\underline{\mathrm{Cosech}}_{W_1}24π−ln⁡(8)+12ln⁡(1+2)\frac{24}{\pi-\ln(8)+12\ln(1+\sqrt{2})}cosech⁡((n+1)π2)/(n+1)\operatorname{cosech}((n+1)\frac{\pi}{2})/(n+1)
IICosech‾Ur\underline{\mathrm{Cosech}}_{U_r}24rπ+2ln⁡(kr)−2ln⁡(2+2kr)−2ln⁡(2−2kr)\frac{24}{\sqrt{r}\pi+2\ln(k_r)-2\ln(2+2k_r)-2\ln(2-2k_r)}cosech⁡(r(n+1)π)/(2n+2)\operatorname{cosech}(\sqrt{r}(n+1)\pi)/(2n+2)
IICosech‾Vr\underline{\mathrm{Cosech}}_{V_r}2artanh⁡(kr)\frac{2}{\operatorname{artanh}(k_r)}cosech⁡(r(2n+1)π2)/(2n+1)\operatorname{cosech}(\sqrt{r}(2n+1)\frac{\pi}{2})/(2n+1)
IICosech‾Wr\underline{\mathrm{Cosech}}_{W_r}24rπ+2ln⁡(4kr)+4ln⁡(1+kr)−8ln⁡(2−2kr)\frac{24}{\sqrt{r}\pi+2\ln(4k_r)+4\ln(1+k_r)-8\ln(2-2k_r)}cosech⁡(r(n+1)π2)/(n+1)\operatorname{cosech}(\sqrt{r}(n+1)\frac{\pi}{2})/(n+1)
IICosechV1\mathrm{Cosech}_{V_1}1arsinh⁡(1)\frac{1}{\operatorname{arsinh}(1)}cosech⁡((2κ+1)π2)/(2κ+1)\operatorname{cosech}((2\kappa+1)\frac{\pi}{2})/(2\kappa+1)
IICosechVr\mathrm{Cosech}_{V_r}2ln⁡(1+kr)−ln⁡(1−kr)\frac{2}{\ln(1+k_r)-\ln(1-k_r)}cosech⁡(r(2κ+1)π2)/(2κ+1)\operatorname{cosech}(\sqrt{r}(2\kappa+1)\frac{\pi}{2})/(2\kappa+1)

Table 2. Discrete unilateral (lemniscatic) type II Cosech distributions of the first (UU), second (VV), and third (WW) kind for n∈N0n\in\mathbb{N}_0 and r>0r>0 (r=1r=1); further the discrete bilateral (lemniscatic) type II Cosech distributions of the second (VV) kind for κ∈Z\kappa\in\mathbb{Z} and r>0r>0 (r=1r=1).

Proof. Ad CUrC_{U_r}: We follow the technique in [Zucker, 1984]. Define the nome q=e−πK′/Kq = e^{-\pi K'/K} with the complete elliptic integral of the first kind KK and further the infinite products

Q0:=∏n=1∞(1−q2n),Q3:=∏n=1∞(1−q2n−1).Q_0 := \prod_{n=1}^{\infty}(1-q^{2n}), \qquad Q_3 := \prod_{n=1}^{\infty}(1-q^{2n-1}).

Equation (5) in [Zucker, 1984, p. 406] yields the equations

Q02Q34=2k′Kπ,2q1/4Q03=8K3kk′π3.Q_0^2Q_3^4 = \frac{2k'K}{\pi}, \qquad2q^{1/4}Q_0^3 = \sqrt{\frac{8K^3kk'}{\pi^3}}.

On the one hand, from these identities we deduce

Q3=q1/24(2k′)1/6k1/12⟺−ln⁡(Q3)=K′Kπ+2ln⁡(k)−4ln⁡(2k′)24.Q_3 = q^{1/24}\frac{(2k')^{1/6}}{k^{1/12}} \quad\Longleftrightarrow\quad-\ln(Q_3) = \frac{\frac{K'}{K}\pi+ 2\ln(k) - 4\ln(2k')}{24}.

On the other hand, the logarithmic series ln⁡(1−x)=−∑m=1∞xm/m\ln(1-x)=-\sum_{m=1}^{\infty}x^m/m and the geometric series results in

−ln⁡(Q3)=−∑m=1∞ln⁡(1−q2m−1)=∑m=1∞∑n=1∞q(2m−1)nn=∑n=1∞qnn(1−q2n)=12∑n=1∞1n(eπnK′/K−e−πnK′/K)=∑n=1∞cosech⁡(K′Knπ)2n.\begin{aligned} -\ln(Q_3)=-\sum_{m=1}^{\infty}\ln(1-q^{2m-1})=\sum_{m=1}^{\infty}\sum_{n=1}^{\infty}\frac{q^{(2m-1)n}}{n}=\sum_{n=1}^{\infty}\frac{q^n}{n(1-q^{2n})} \\ &=\frac{1}{2}\sum_{n=1}^{\infty}\frac{1}{n\left(e^{\pi nK'/K}-e^{-\pi nK'/K}\right)}=\sum_{n=1}^{\infty}\frac{\operatorname{cosech}\left(\frac{K'}{K}n\pi\right)}{2n}. \end{aligned}

Hence,

∑n=0∞cosech⁡(K′K(2n+2)π2)2n+2=K′Kπ+2ln⁡(k)−4ln⁡(2k′)24.(6)\sum_{n=0}^{\infty}\frac{\operatorname{cosech}\left(\frac{K'}{K}(2n+2)\frac{\pi}{2}\right)}{2n+2}=\frac{\frac{K'}{K}\pi+2\ln(k)-4\ln(2k')}{24}. \tag*{(6)}

We choose the elliptic lambda-star function λ∗(r)=kr\lambda^*(r)=k_r such that Kr′/Kr=rK'_r/K_r=\sqrt{r}. This yields

∑n=0∞cosech⁡(r(n+1)π)2n+2=rπ+2ln⁡(kr)−4ln⁡(2kr′)24.\sum_{n=0}^{\infty}\frac{\operatorname{cosech}\left(\sqrt{r}(n+1)\pi\right)}{2n+2}=\frac{\sqrt{r}\pi+2\ln(k_r)-4\ln(2k'_r)}{24}.

This results in the normalization constant

CUr=24rπ+2ln⁡(kr)−2ln⁡(2+2kr)−2ln⁡(2−2kr),C_{U_r}=\frac{24}{\sqrt{r}\pi+2\ln(k_r)-2\ln(2+2k_r)-2\ln(2-2k_r)},

as claimed.

Ad CVrC_{V_r}: The suitable formula is given by the Example after Entry 7 of Chapter 18 in Ramanujan’s second notebook, cf. [9 p. 156]. With x=k2x = k^2 and y=πK′/Ky = \pi K'/K, Ramanujan states

∑n=0∞1(2n+1)sinh⁡(12(2n+1)y)=14log⁡(1+x1−x)=∑n=0∞cosech⁡(K′K(2n+1)π2)2n+1=14ln⁡(1+k1−k).\begin{aligned} \sum_{n=0}^{\infty} \frac{1}{(2n+1)\sinh\left(\frac{1}{2}(2n+1)y\right)} = \frac{1}{4}\log\left(\frac{1+\sqrt{x}}{1-\sqrt{x}}\right) \\ &= \sum_{n=0}^{\infty} \frac{\operatorname{cosech}\left(\frac{K'}{K}(2n+1)\frac{\pi}{2}\right)}{2n+1} = \frac{1}{4}\ln\left(\frac{1+k}{1-k}\right). \end{aligned}

This is e.g. proven by Berndt in [9 p. 156–157]. We use Equation (17) in [3 p. 294] with k∈(0,1)k \in(0,1) and q=e−πK′/Kq = e^{-\pi K'/K}, viz.

−116ln⁡(1−k2)=∑n=0∞q2n+1(2n+1)(1−q4n+2).-\frac{1}{16}\ln(1-k^2) = \sum_{n=0}^{\infty} \frac{q^{2n+1}}{(2n+1)(1-q^{4n+2})}.

For each n∈N0n \in\mathbb{N}_0,

cosech⁡(K′K(2n+1)π)=2q2n+11−q4n+2.\operatorname{cosech}\left(\frac{K'}{K}(2n+1)\pi\right) = \frac{2q^{2n+1}}{1-q^{4n+2}}.

Thus

∑n=0∞cosech⁡(K′K(2n+1)π)2n+1=2∑n=0∞q2n+1(2n+1)(1−q4n+2)=−14ln⁡(k′).\sum_{n=0}^{\infty} \frac{\operatorname{cosech}\left(\frac{K'}{K}(2n+1)\pi\right)}{2n+1} = 2\sum_{n=0}^{\infty} \frac{q^{2n+1}}{(2n+1)(1-q^{4n+2})} = -\frac{1}{4}\ln(k').

The series converges because 0<q<10 < q < 1. In [Zucker, 1984, p. 409], the following transformations for changing the nome from qq to q1/2q^{1/2} are stated:

k→2k1+k,k′→1−k1+k,K′K→K′2K.(7)k \to\frac{2\sqrt{k}}{1+k}, \qquad k' \to\frac{1-k}{1+k}, \qquad\frac{K'}{K} \to\frac{K'}{2K}. \tag*{(7)}

We again choose the elliptic lambda-star function λ∗(r)=kr\lambda^*(r) = k_r such that Kr′/Kr=rK'_r/K_r = \sqrt{r}. This yields

∑n=0∞cosech⁡(r(2n+1)π2)2n+1=14ln⁡(1+kr1−kr)=artanh⁡(kr)2.\sum_{n=0}^{\infty} \frac{\operatorname{cosech}\left(\sqrt{r}(2n+1)\frac{\pi}{2}\right)}{2n+1} = \frac{1}{4}\ln\left(\frac{1+k_r}{1-k_r}\right) = \frac{\operatorname{artanh}(k_r)}{2}.

Therefore, we get the normalization constant

CVr=4ln⁡(1+kr)−ln⁡(1−kr)=2artanh⁡(kr),C_{V_r} = \frac{4}{\ln(1+k_r)-\ln(1-k_r)} = \frac{2}{\operatorname{artanh}(k_r)},

as claimed.

Ad CWrC_{W_r}: We use the transformations stated in Equation (7) from [Zucker, 1984, p. 409]. We choose the elliptic lambda-star function λ∗(r)=kr\lambda^{*}(r)=k_r such that Kr′/Kr=rK'_r/K_r=\sqrt{r}. Hence, Equation (6) yields

∑n=0∞cosech⁡(r(n+1)π2)n+1=rπ+2ln⁡(4kr)+4ln⁡(1+kr)−8ln⁡(2−2kr)24.\sum_{n=0}^{\infty}\frac{\operatorname{cosech}\left(\sqrt{r}(n+1)\frac{\pi}{2}\right)}{n+1}=\frac{\sqrt{r}\pi+2\ln(4k_r)+4\ln(1+k_r)-8\ln(2-2k_r)}{24}.

Thus, we finally get the normalization constant

CWr=24rπ+2ln⁡(4kr)+4ln⁡(1+kr)−8ln⁡(2−2kr).C_{W_r}=\frac{24}{\sqrt{r}\pi+2\ln(4k_r)+4\ln(1+k_r)-8\ln(2-2k_r)}.

as stated. □

Since it leaves the 0 out, we can also define a bilateral type II Cosech of the second kind. By symmetry, we just have to take the normalization constant CVr=CVr′/2C_{V_r}=C'_{V_r}/2.

Corollary and Definition 3.2. [Type II Cosech Distributions.] Let κ∈Z\kappa\in\mathbb{Z}. For the rows 8–9 of Table 2, let CAC_A denote the normalization constant of the listed CosechA_A distribution with weight function wA(κ)w_A(\kappa). Then the associated sequence (pA(κ))κ∈Z:=(CA⋅wA(κ))κ∈Z(p_A(\kappa))_{\kappa\in\mathbb{Z}}:=(C_A\cdot w_A(\kappa))_{\kappa\in\mathbb{Z}} is a probability mass function.

We name CosechVr_{V_r} and CosechV1_{V_1} the discrete bilateral (lemniscatic for r=1r=1) type II hyperbolic cosecant distribution of the second kind, respectively. We omit adjectives when the context is clear.

The Discrete Hyperbolic Cosecant Distribution – Type III

Next, we introduce a further version of the Cosech distribution: The type III hyperbolic cosecant distribution. Like for the type II Cosech distribution, the underlying continuous weight function wIII(x):=cosech⁡2(xπ)w_{\mathrm{III}}(x):=\operatorname{cosech}^{2}(x\pi) has a non-removable double pole at 0. Therefore, we again first concentrate on the unilateral type III Cosech distributions on N0\mathbb{N}_0 after index shift. We use techniques from [9], [2], and [12] to prove the general elliptic case for r>0r>0. For the lemniscatic case, we insert r=1r=1, k1=1/2k_1=1/\sqrt{2}, K1=ϖ/2K_1=\varpi/\sqrt{2}, and α(1)=1/2\alpha(1)=1/2 into the general formulae.

Proposition and Definition 4.1. [(Lemniscatic) Type III Cosech Distributions.] Let n∈N0n\in\mathbb{N}_0 and r>0r>0 (r=1r=1). For the rows 2–6 of Table 3, let CAC_A denote the normalization constant of the listed CosechA_A distribution with weight function wA(n)w_A(n). Then the associated sequence (pA(n))n∈N0:=(CA⋅wA(n))n∈N0(p_A(n))_{n\in\mathbb{N}_0}:=(C_A\cdot w_A(n))_{n\in\mathbb{N}_0} is a probability mass function.

TypeDistributionNormalizing ConstantWeight Function
IIICosech⁡R1\operatorname{Cosech}_{R_1}6ππ−3\frac{6\pi}{\pi-3}cosech⁡2((2n+2)π2)\operatorname{cosech}^{2}\left((2n+2)\frac{\pi}{2}\right)
IIICosech⁡S1\operatorname{Cosech}_{S_1}2ππG2−1\frac{2\pi}{\pi G^{2}-1}cosech⁡2((2n+1)π2)\operatorname{cosech}^{2}\left((2n+1)\frac{\pi}{2}\right)
IIICosech⁡T1\operatorname{Cosech}_{T_1}6ππ−6+3πG2\frac{6\pi}{\pi-6+3\pi G^{2}}cosech⁡2((n+1)π2)\operatorname{cosech}^{2}\left((n+1)\frac{\pi}{2}\right)
IIICosech⁡Rr\operatorname{Cosech}_{R_r}6rπ2rπ2−3π+4Kr2(3α(r)−(1+kr2)r)\frac{6\sqrt{r}\pi^{2}}{\sqrt{r\pi^{2}-3\pi+4K_{r}^{2}\left(3\alpha(r)-(1+k_{r}^{2})\sqrt{r}\right)}}cosech⁡2(r(2n+2)π2)\operatorname{cosech}^{2}\left(\sqrt{r}(2n+2)\frac{\pi}{2}\right)
IIICosech⁡Sr\operatorname{Cosech}_{S_r}2rπ24Kr2α(r)−π\frac{2\sqrt{r}\pi^{2}}{4K_{r}^{2}\alpha(r)-\pi}cosech⁡2(r(2n+1)π2)\operatorname{cosech}^{2}\left(\sqrt{r}(2n+1)\frac{\pi}{2}\right)
IIICosech⁡Tr\operatorname{Cosech}_{T_r}6rπ2rπ2−6π+4Kr2(6α(r)−(1+kr2)r)\frac{6\sqrt{r}\pi^{2}}{\sqrt{r\pi^{2}-6\pi+4K_{r}^{2}\left(6\alpha(r)-(1+k_{r}^{2})\sqrt{r}\right)}}cosech⁡2(r(n+1)π2)\operatorname{cosech}^{2}\left(\sqrt{r}(n+1)\frac{\pi}{2}\right)
IIICosech⁡S1\operatorname{Cosech}_{S_1}ππG2−1\frac{\pi}{\pi G^{2}-1}cosech⁡2((2κ+1)π2)\operatorname{cosech}^{2}\left((2\kappa+1)\frac{\pi}{2}\right)
IIICosech⁡Sr\operatorname{Cosech}_{S_r}rπ24Kr2α(r)−π\frac{\sqrt{r}\pi^{2}}{4K_{r}^{2}\alpha(r)-\pi}cosech⁡2(r(2κ+1)π2)\operatorname{cosech}^{2}\left(\sqrt{r}(2\kappa+1)\frac{\pi}{2}\right)

Table 3. Discrete unilateral (lemniscatic) type III Cosech distributions of the first (RR), second (SS), and third (TT) kind for n∈Nn \in\mathbb{N} and r>0r > 0 (r=1r=1); further the discrete bilateral (lemniscatic) type III Cosech distributions of the second (SS) kind for κ∈Z\kappa\in\mathbb{Z} and r>0r > 0 (r=1r=1).

We name CosechRr_{R_r}, CosechSr_{S_r}, and CosechTr_{T_r} the discrete (lemniscatic for r=1r=1) unilateral type III hyperbolic cosecant distribution of the first, second, and third kind, respectively. We omit adjectives when the context is clear.

Proof. Ad CRrC_{R_r}: For the calculation of the normalizing constant CRrC_{R_r} for the CosechRr_{R_r} distribution we can use the technique in [12 p. 197] or complete Exercise 7 of chapter 3.7 in [2 p. 96]. We do the latter. Let c:=Kr′/Krc := K'_r/K_r with nome q=e−πcq=e^{-\pi c}, where k=k(c)k=k(c) is determined implicitly. Further, define

Q0:=∏n=1∞(1−q2n).Q_0 := \prod_{n=1}^{\infty}(1-q^{2n}).

Equation (3.2.28i) in [2 p. 74] claims

F(c):=−ln⁡(Q0)=12∑n=1∞coth⁡(cnπ)−1n=cπ12−16ln⁡(2kk′K3π3).F(c):=-\ln(Q_0)=\frac{1}{2}\sum_{n=1}^{\infty}\frac{\coth(cn\pi)-1}{n}=\frac{c\pi}{12}-\frac{1}{6}\ln\left(\frac{2kk'K^{3}}{\pi^{3}}\right).

For c>0c>0, the series and its termwise derivative converge uniformly on every compact subinterval of (0,∞)(0,\infty). Hence, the derivative of the series yields

F′(c)=−π2∑n=1∞cosech⁡2(cnπ).F'(c)=-\frac{\pi}{2}\sum_{n=1}^{\infty}\operatorname{cosech}^{2}(cn\pi).

Differentiating the right-hand side with respect to cc results in

F′(c)=−π12−16ddcln⁡(2kk′K3).F'(c)=-\frac{\pi}{12}-\frac{1}{6}\frac{\mathrm{d}}{\mathrm{d}c}\ln(2kk'K^{3}).

Equating the two expressions for F′(c)F'(c), we obtain

∑n=1∞cosech⁡2(cnπ)=16+13πddcln⁡(2kk′K3).\sum_{n=1}^{\infty} \operatorname{cosech}^{2}(cn\pi) = \frac{1}{6} + \frac{1}{3\pi}\frac{\mathrm{d}}{\mathrm{d}c}\ln(2kk'K^{3}).

We now compute the derivative on the right-hand side. Since KK is a function of kk, and kk is a function of cc, the expression f(K,k)f(K,k) depends on cc through k(c)k(c). Thus, by the chain rule,

ddcf(K(k(c)),k(c))=(∂f∂KdKdk+∂f∂k)dkdc.\frac{\mathrm{d}}{\mathrm{d}c}f(K(k(c)),k(c)) = \left(\frac{\partial f}{\partial K}\frac{\mathrm{d}K}{\mathrm{d}k} + \frac{\partial f}{\partial k}\right)\frac{\mathrm{d}k}{\mathrm{d}c}.

The expression in parentheses is the total derivative with respect to kk. Hence,

ddcln⁡(2kk′K3)=ddkln⁡(2kk′K3)dkdc.\frac{\mathrm{d}}{\mathrm{d}c}\ln(2kk'K^{3}) = \frac{\mathrm{d}}{\mathrm{d}k}\ln(2kk'K^{3})\frac{\mathrm{d}k}{\mathrm{d}c}.

The standard derivative formula, given in equation 19.4.1 of [8], is

dKdk=E−k′2Kkk′2.\frac{\mathrm{d}K}{\mathrm{d}k} = \frac{E-k'^{2}K}{kk'^{2}}.

This implies

ddkln⁡(2kk′K3)=31KdKdk+1k+1k′dk′dk=3Ekk′2K−2k−kk′2.\frac{\mathrm{d}}{\mathrm{d}k}\ln(2kk'K^{3}) = 3\frac{1}{K}\frac{\mathrm{d}K}{\mathrm{d}k}+\frac{1}{k}+\frac{1}{k'}\frac{\mathrm{d}k'}{\mathrm{d}k} = \frac{3E}{kk'^{2}K}-\frac{2}{k}-\frac{k}{k'^{2}}.

Since c=K(k′)/K(k)c=K(k')/K(k), differentiating this quotient with respect to kk and using the derivative formula for KK gives

dcdk=−KE′+K′E−KK′kk′2K2=−π2kk′2K2,\frac{\mathrm{d}c}{\mathrm{d}k}=-\frac{KE'+K'E-KK'}{kk'^{2}K^{2}}=-\frac{\pi}{2kk'^{2}K^{2}},

where the last equality follows from Legendre’s relation

EK′+E′K−KK′=π2.EK'+E'K-KK'=\frac{\pi}{2}.

Therefore, inversion yields

dkdc=−2kk′2K2π.\frac{\mathrm{d}k}{\mathrm{d}c}=-\frac{2kk'^{2}K^{2}}{\pi}.

Altogether,

ddcln⁡(2kk′K3)=(3Ekk′2K−2k−kk′2)(−2kk′2K2π)=(4−2k2)K2−6KEπ.\frac{\mathrm{d}}{\mathrm{d}c}\ln(2kk'K^{3})=\left(\frac{3E}{kk'^{2}K}-\frac{2}{k}-\frac{k}{k'^{2}}\right)\left(-\frac{2kk'^{2}K^{2}}{\pi}\right)=\frac{(4-2k^{2})K^{2}-6KE}{\pi}.

We choose again the elliptic lambda-star function λ∗(r)=kr\lambda^{*}(r)=k_{r} such that c=Kr′/Kr=rc=K'_{r}/K_{r}=\sqrt{r}. Next, we use the formula

Er=π4rKr+(1−α(r)r)Kr.E_{r}=\frac{\pi}{4\sqrt{r}K_{r}}+\left(1-\frac{\alpha(r)}{\sqrt{r}}\right)K_{r}.

with the elliptic alpha function α(r)\alpha(r). This results in

∑n=0∞cosech⁡2(r(2n+2)π2)=π2−12KrEr+(8−4kr2)Kr26π2=rπ2−3π+4Kr2(3α(r)−(1+kr2)r)6rπ2.\sum_{n=0}^{\infty} \operatorname{cosech}^{2}\left(\sqrt{r}(2n+2)\frac{\pi}{2}\right)=\frac{\pi^{2}-12K_{r}E_{r}+(8-4k_{r}^{2})K_{r}^{2}}{6\pi^{2}} =\frac{\sqrt{r}\pi^{2}-3\pi+4K_{r}^{2}\left(3\alpha(r)-(1+k_{r}^{2})\sqrt{r}\right)}{6\sqrt{r}\pi^{2}}.

Therefore,

CBr=6rπ2rπ2−3π+4Kr2(3α(r)−(1+kr2)r),C_{B_{r}}=\frac{6\sqrt{r}\pi^{2}}{\sqrt{r}\pi^{2}-3\pi+4K_{r}^{2}\left(3\alpha(r)-(1+k_{r}^{2})\sqrt{r}\right)},

as claimed.

Ad CSrC_{S_{r}}: Berndt provides the eq:14.15 in the proof of Entry 14 in Chapter 18 of Ramanujan’s second notebook, cf. [9 p. 172], viz.

∑n=−∞∞cosech⁡{12(2n+1)y}=4∑n=1∞nsinh⁡(ny)=∑n∈Zcosech⁡2(K′K(2n+1)π2)=4∑n=1∞ncosech⁡(K′Knπ).\begin{aligned} \sum_{n=-\infty}^{\infty}\operatorname{cosech}\left\{\frac{1}{2}(2n+1)y\right\}=4\sum_{n=1}^{\infty}\frac{n}{\sinh(ny)} \\ &=\sum_{n\in\mathbb{Z}}\operatorname{cosech}^{2}\left(\frac{K'}{K}(2n+1)\frac{\pi}{2}\right)=4\sum_{n=1}^{\infty}n\operatorname{cosech}\left(\frac{K'}{K}n\pi\right). \end{aligned}

By the symmetry n↦−n−1n\mapsto-n-1, the left-hand side equals

2∑n=0∞cosech⁡2(K′K(2n+1)π2).2\sum_{n=0}^{\infty}\operatorname{cosech}^{2}\left(\frac{K'}{K}(2n+1)\frac{\pi}{2}\right).

By (5), the right-hand side equals 4K(K−E)/π24K(K-E)/\pi^{2}. We choose the elliptic lambda-star function λ∗(r)=kr\lambda^{*}(r)=k_{r} such that Kr′/Kr=rK'_{r}/K_{r}=\sqrt{r}. Now, (1) yields

∑n=0∞cosech⁡2(r(2n+1)π2)=2Kr(Kr−Er)π2=4α(r)Kr2−π2rπ2.\sum_{n=0}^{\infty}\operatorname{cosech}^{2}\left(\sqrt{r}(2n+1)\frac{\pi}{2}\right)=\frac{2K_{r}(K_{r}-E_{r})}{\pi^{2}}=\frac{4\alpha(r)K_{r}^{2}-\pi}{2\sqrt{r}\pi^{2}}.

with the elliptic alpha function α(r)\alpha(r). Therefore, we get the normalization constant

CSr=2rπ24α(r)Kr2−π,C_{S_{r}}=\frac{2\sqrt{r}\pi^{2}}{4\alpha(r)K_{r}^{2}-\pi},

as claimed.

Ad CTrC_{T_{r}}: We split the sum over n≥1n\geq1 into even and odd indices. Then

∑n=0∞cosech⁡2(r(n+1)π2)=∑n=0∞cosech⁡2(r(2n+2)π2)+∑n=0∞cosech⁡2(r(2n+1)π2).\sum_{n=0}^{\infty}\operatorname{cosech}^{2}\left(\sqrt{r}(n+1)\frac{\pi}{2}\right)=\sum_{n=0}^{\infty}\operatorname{cosech}^{2}\left(\sqrt{r}(2n+2)\frac{\pi}{2}\right)+\sum_{n=0}^{\infty}\operatorname{cosech}^{2}\left(\sqrt{r}(2n+1)\frac{\pi}{2}\right).

By the proofs for CRrC_{R_r} and CSrC_{S_r}, this yields

∑n=0∞cosech⁡2(r(n+1)π2)=rπ2−3π+4Kr2(3α(r)−(1+kr2)r)6rπ2+4α(r)Kr2−π2rπ2=rπ2−6π+4Kr2(6α(r)−(1+kr2)r)6rπ2.\begin{aligned} \sum_{n=0}^{\infty}\operatorname{cosech}^{2}\left(\sqrt{r}(n+1)\frac{\pi}{2}\right)=\frac{\sqrt{r\pi^{2}-3\pi+4K_{r}^{2}}\left(3\alpha(r)-(1+k_{r}^{2})\sqrt{r}\right)}{6\sqrt{r\pi^{2}}}+\frac{4\alpha(r)K_{r}^{2}-\pi}{2\sqrt{r\pi^{2}}} \\ &=\frac{\sqrt{r\pi^{2}-6\pi+4K_{r}^{2}}\left(6\alpha(r)-(1+k_{r}^{2})\sqrt{r}\right)}{6\sqrt{r\pi^{2}}}. \end{aligned}

Therefore,

CTr=6rπ2rπ2−6π+4Kr2(6α(r)−(1+kr2)r).C_{T_r}=\frac{6\sqrt{r\pi^{2}}}{\sqrt{r\pi^{2}-6\pi+4K_{r}^{2}}\left(6\alpha(r)-(1+k_{r}^{2})\sqrt{r}\right)}.

as stated. □\square

Since its support excludes 0, we can also define the bilateral type III Cosech of the second kind. By symmetry, we just have to take the normalization constant CSr=CSr/2C_{S_r}=C_{S_r}/2.

Corollary and Definition 4.2. [Type III Cosech Distributions.] Let κ∈Z\kappa\in\mathbb{Z}. For the rows 8–9 of Table 3, let CAC_A denote the normalization constant of the listed Cosech⁡A\operatorname{Cosech}_A distribution with weight function wA(κ)w_A(\kappa). Then the associated sequence (pA(κ))κ∈Z:=(CA⋅wA(κ))κ∈Z(p_A(\kappa))_{\kappa\in\mathbb{Z}}:=(C_A\cdot w_A(\kappa))_{\kappa\in\mathbb{Z}} is a probability mass function.

We name Cosech⁡Sr\operatorname{Cosech}_{S_r} and Cosech⁡S1\operatorname{Cosech}_{S_1} the discrete bilateral (lemniscatic for r=1r=1) type III hyperbolic cosecant distribution of the second kind, respectively. We omit adjectives when the context is clear.

Moments of the Discrete Hyperbolic Cosecant Distribution

We want to determine some moments of the discrete hyperbolic cosecant distribution Cosech. According to Table 1, its probability mass function is given by

(p(κ))κ∈Z=ππG2+12κcosech⁡(κπ).(p(\kappa))_{\kappa\in\mathbb{Z}}=\frac{\pi}{\pi G^{2}+1}2^{\kappa}\operatorname{cosech}(\kappa\pi).

Ramanujan’s series identities directly yield the even moments up to the eighth order.

Proposition 5.1. Let X∼Cosech⁡X\sim\operatorname{Cosech} be a discrete random variable which is distributed according to the hyperbolic cosecant distribution. The odd moments vanish and we have

(i)

E[X2]=Var⁡(X)=πG4πG2+1≈0.4781246,\mathbb{E}\left[X^{2}\right]=\operatorname{Var}(X)=\frac{\pi G^{4}}{\pi G^{2}+1}\approx0.4781246,

(ii)

E[X4]=3πG6πG2+1≈0.9991876,\mathbb{E}\left[X^{4}\right]=\frac{3\pi G^{6}}{\pi G^{2}+1}\approx0.9991876,

(iii)

E[X6]=18πG8πG2+1≈4.1762161,\mathbb{E}\left[X^{6}\right]=\frac{18\pi G^{8}}{\pi G^{2}+1}\approx4.1762161,

(iv)

E[X8]=189πG10πG2+1≈30.5461835,\mathbb{E}\left[X^{8}\right]=\frac{189\pi G^{10}}{\pi G^{2}+1}\approx30.5461835,

with the Gaußian constant G=ω/πG = \omega/\pi.

Proof. Let C=π/(πG2+1)C = \pi/(\pi G^2 + 1). We get E[Xj]=0\mathbb{E}[X^j] = 0 for odd j∈Nj \in\mathbb{N} due to symmetry. The other formulae directly follow by Ramanujan’s Entry 15 (i)-(iv) in Chapter 17 of his second notebook, cf. [9 pp. 132-133]: For j∈{2,4,6,8}j \in\{2,4,6,8\} we write

E[Xj]=C∑κ∈Zκj⋅2κcosech⁡(κπ)=4ππG2+1∑n=1∞nj+1sinh⁡(πn).\mathbb{E}[X^j] = C \sum_{\kappa\in\mathbb{Z}} \kappa^j \cdot2\kappa\operatorname{cosech}(\kappa\pi) = \frac{4\pi}{\pi G^2+1}\sum_{n=1}^{\infty}\frac{n^{j+1}}{\sinh(\pi n)}.

In the lemniscatic case, we can translate Ramanujan’s notation via x=1/2x=1/2, y=πy=\pi, and z=2Gz=\sqrt{2}G. We get

∑n=1∞n3sinh⁡(yn)=18z4x⇒E[X2]=πG4πG2+1,\sum_{n=1}^{\infty}\frac{n^3}{\sinh(yn)}=\frac{1}{8}z^4x \Rightarrow\mathbb{E}[X^2]=\frac{\pi G^4}{\pi G^2+1},
∑n=1∞n5sinh⁡(yn)=18z6x(1+x)⇒E[X4]=3πG6πG2+1,\sum_{n=1}^{\infty}\frac{n^5}{\sinh(yn)}=\frac{1}{8}z^6x(1+x) \Rightarrow\mathbb{E}[X^4]=\frac{3\pi G^6}{\pi G^2+1},
∑n=1∞n7sinh⁡(yn)=18z8x(1+132x+x2)⇒E[X6]=18πG8πG2+1,\sum_{n=1}^{\infty}\frac{n^7}{\sinh(yn)}=\frac{1}{8}z^8x\left(1+\frac{13}{2}x+x^2\right) \Rightarrow\mathbb{E}[X^6]=\frac{18\pi G^8}{\pi G^2+1},
∑n=1∞n9sinh⁡(yn)=18z10x(1+x)(1+29x+x2)⇒E[X8]=189πG10πG2+1,\sum_{n=1}^{\infty}\frac{n^9}{\sinh(yn)}=\frac{1}{8}z^{10}x(1+x)(1+29x+x^2) \Rightarrow\mathbb{E}[X^8]=\frac{189\pi G^{10}}{\pi G^2+1},

as claimed.

We want to evaluate some exact expressions for series of even powers of the hyperbolic secant via the Poisson summation. For that purpose we first prove a formula for their Fourier transform.

Lemma 5.2. Let m∈Nm \in\mathbb{N} be a positive integer and let ξ∈R\xi\in\mathbb{R}. Then the Fourier transformation formula

F[x↦sech⁡2m(xπ)](ξ)=∫−∞∞sech⁡2m(xπ)e−2πixξ dx=22m−1π(2m−1)!∣Γ(m+iξ)∣2\mathcal{F}\left[x \mapsto\operatorname{sech}^{2m}(x\pi)\right](\xi)=\int_{-\infty}^{\infty}\operatorname{sech}^{2m}(x\pi)e^{-2\pi i x\xi}\,dx=\frac{2^{2m-1}}{\pi(2m-1)!}\left|\Gamma(m+i\xi)\right|^2

holds for every mm. Further, the identity

∣Γ(m+iξ)∣2=ξπcosech⁡(ξπ)∏j=1m−1(j2+ξ2)\left|\Gamma(m+i\xi)\right|^2=\xi\pi\operatorname{cosech}(\xi\pi)\prod_{j=1}^{m-1}(j^2+\xi^2)

holds.

Proof. We use the hyperbolic substitution u=u(x):=(1+tanh⁡(xπ))/2u=u(x):=(1+\tanh(x\pi))/2, and hence 2u−1=tanh⁡(xπ)2u-1=\tanh(x\pi), which implies x=ln⁡(u/(1−u))/(2π)x=\ln(u/(1-u))/(2\pi). Now, 4u(1−u)=1−tanh⁡2(xπ)=sech⁡2(xπ)4u(1-u)=1-\tanh^2(x\pi)=\operatorname{sech}^2(x\pi) leads to

sech⁡2m(xπ)=4mum(1−u)m.\operatorname{sech}^{2m}(x\pi)=4^m u^m(1-u)^m.

For the boundaries, we get lim⁡x→−∞u(x)=0\lim_{x\to-\infty} u(x)=0 and lim⁡x→∞u(x)=1\lim_{x\to\infty} u(x)=1. Hence, by continuity, u∈(0,1)u\in(0,1) and thus, u/(1−u)∈(0,∞)u/(1-u)\in(0,\infty). Therefore, we can use the unique, real natural logarithm to get

e−2πixξ=exp⁡(−iξln⁡(u1−u))=u−iξ(1−u)iξ.e^{-2\pi i x \xi}=\exp\left(-i\xi\ln\left(\frac{u}{1-u}\right)\right)=u^{-i\xi}(1-u)^{i\xi}.

Further, u′(x)=πsech⁡2(xπ)/2u'(x)=\pi\operatorname{sech}^{2}(x\pi)/2 results in

dx=2duπsech⁡2(xπ)=du2πu(1−u).\mathrm{d}x=\frac{2\mathrm{d}u}{\pi\operatorname{sech}^{2}(x\pi)}=\frac{\mathrm{d}u}{2\pi u(1-u)}.

We use Euler’s beta function B:{(a,b)∈C2:ℜ(a),ℜ(b)>0}→CB:\{(a,b)\in\mathbb{C}^{2}:\mathop{\Re}(a),\mathop{\Re}(b)>0\}\to\mathbb{C} with its key property

B(a,b)=∫01ta−1(1−t)b−1 dt=Γ(a)⋅Γ(b)Γ(a+b).B(a,b)=\int_{0}^{1}t^{a-1}(1-t)^{b-1}\,\mathrm{d}t=\frac{\Gamma(a)\cdot\Gamma(b)}{\Gamma(a+b)}.

Altogether, we get

∫−∞∞sech⁡2m(xπ)e−2πixξ dx=∫0122mum(1−u)m⏟sech⁡2m(xπ)⋅u−iξ(1−u)iξ⏟e−2πixξdu2πu(1−u)=22m−1π∫01um−iξ−1(1−u)m+iξ−1 du=22m−1πB(m−iξ,m+iξ)=22m−1πΓ(m−iξ)⋅Γ(m+iξ)Γ(2m)=22m−1π(2m−1)!∣Γ(m+iξ)∣2,\begin{aligned} \int_{-\infty}^{\infty}\operatorname{sech}^{2m}(x\pi)e^{-2\pi i x\xi}\,\mathrm{d}x &=\int_{0}^{1}\underbrace{2^{2m}u^{m}(1-u)^{m}}_{\operatorname{sech}^{2m}(x\pi)}\cdot\underbrace{u^{-i\xi}(1-u)^{i\xi}}_{e^{-2\pi i x\xi}}\frac{\mathrm{d}u}{2\pi u(1-u)}\\ &=\frac{2^{2m-1}}{\pi}\int_{0}^{1}u^{m-i\xi-1}(1-u)^{m+i\xi-1}\,\mathrm{d}u\\ &=\frac{2^{2m-1}}{\pi}B(m-i\xi,m+i\xi)\\ &=\frac{2^{2m-1}}{\pi}\frac{\Gamma(m-i\xi)\cdot\Gamma(m+i\xi)}{\Gamma(2m)}\\ &=\frac{2^{2m-1}}{\pi(2m-1)!}\left|\Gamma(m+i\xi)\right|^{2}, \end{aligned}

as claimed.

The second formula follows from induction:

m=1: Euler’s reflection formula yields

Γ(iξ)⋅Γ(1−iξ)=πsin⁡(ξπ)=πisinh⁡(ξπ).\Gamma(i\xi)\cdot\Gamma(1-i\xi)=\frac{\pi}{\sin(\xi\pi)}=\frac{\pi}{i\sinh(\xi\pi)}.

Since Γ(1−iξ)=−iξΓ(−iξ)=−iξΓ(iξ)‾\Gamma(1-i\xi)=-i\xi\Gamma(-i\xi)=-i\xi\overline{\Gamma(i\xi)}, we get ∣Γ(iξ)∣2=πcosech⁡(ξπ)/ξ\lvert\Gamma(i\xi)\rvert^{2}=\pi\operatorname{cosech}(\xi\pi)/\xi. This results in

∣Γ(1+iξ)∣2=ξ2∣Γ(iξ)∣2=ξπcosech⁡(ξπ),\left|\Gamma(1+i\xi)\right|^{2}=\xi^{2}\left|\Gamma(i\xi)\right|^{2}=\xi\pi\operatorname{cosech}(\xi\pi),

as claimed by defining the empty product as 1. The case ξ=0\xi=0 follows from continuity.

m→m+1m \to m+1: Let

∣Γ(m+iξ)∣2=ξπcosech⁡(ξπ)∏j=1m−1(j2+ξ2)|\Gamma(m+i\xi)|^2=\xi\pi\operatorname{cosech}(\xi\pi)\prod_{j=1}^{m-1}(j^2+\xi^2)

hold for mm. Then we get

∣Γ(m+1+iξ)∣2=(m2+ξ2)ξπcosech⁡(ξπ)∏j=1m−1(j2+ξ2)=ξπcosech⁡(ξπ)∏j=1m(j2+ξ2),\begin{aligned} |\Gamma(m+1+i\xi)|^2=(m^2+\xi^2)\xi\pi\operatorname{cosech}(\xi\pi)\prod_{j=1}^{m-1}(j^2+\xi^2) \\ &=\xi\pi\operatorname{cosech}(\xi\pi)\prod_{j=1}^{m}(j^2+\xi^2), \end{aligned}

and the formula is proven. □

Now apply the Poisson summation formula in order to get the desired series values.

Theorem 5.3. The following identities hold:

(i)∑k∈Zsech⁡2(kπ)=G2+1π,\text{(i)}\qquad\sum_{k\in\mathbb{Z}}\operatorname{sech}^2(k\pi)=G^2+\frac{1}{\pi},
(ii)∑k∈Zsech⁡4(kπ)=23(G4+G2+1π)\text{(ii)}\qquad\sum_{k\in\mathbb{Z}}\operatorname{sech}^4(k\pi)=\frac{2}{3}\left(G^4+G^2+\frac{1}{\pi}\right)
(iii)∑k∈Zsech⁡6(kπ)=25G6+23G4+815(G2+1π)\text{(iii)}\qquad\sum_{k\in\mathbb{Z}}\operatorname{sech}^6(k\pi)=\frac{2}{5}G^6+\frac{2}{3}G^4+\frac{8}{15}\left(G^2+\frac{1}{\pi}\right)
(iv)∑k∈Zsech⁡8(kπ)=835G8+815G6+2845G4+1635(G2+1π)\text{(iv)}\qquad\sum_{k\in\mathbb{Z}}\operatorname{sech}^8(k\pi)=\frac{8}{35}G^8+\frac{8}{15}G^6+\frac{28}{45}G^4+\frac{16}{35}\left(G^2+\frac{1}{\pi}\right)

with the Gaußian constant G=ϖ/πG=\varpi/\pi.

Proof. Ad (i): For the application of the Poisson summation formula, we first calculate the Fourier transformation of sech⁡2(kπ)\operatorname{sech}^2(k\pi). According to Lemma 5.2, one has

F[x↦sech⁡2(xπ)](ξ)=2π∣Γ(1+iξ)∣2=2ξcosech⁡(ξπ).\mathcal{F}\left[x\mapsto\operatorname{sech}^2(x\pi)\right](\xi)=\frac{2}{\pi}|\Gamma(1+i\xi)|^2=2\xi\operatorname{cosech}(\xi\pi).

Now, the Poisson summation formula and Equation (4) in the proof of Proposition 2.1 yield

∑k∈Zsech⁡2(kπ)=∑k∈ZF[x↦sech⁡2(xπ)](ξ)∣ξ=k=∑k∈Z2ξcosech⁡(ξπ)∣ξ=k=G2+1π,\sum_{k\in\mathbb{Z}}\operatorname{sech}^2(k\pi)=\sum_{k\in\mathbb{Z}}\mathcal{F}\left[x\mapsto\operatorname{sech}^2(x\pi)\right](\xi)\big|_{\xi=k}=\sum_{k\in\mathbb{Z}}2\xi\operatorname{cosech}(\xi\pi)\big|_{\xi=k}=G^2+\frac{1}{\pi},

as claimed.

Ad (ii) [Probabilistic proof]: Let X∼Cosech⁡X \sim\operatorname{Cosech} be a discrete random variable which is distributed according to the hyperbolic cosecant distribution with normalization constant C=π/(πG2+1)C = \pi/(\pi G^{2} + 1). Therefore, by using the duality of the Fourier transformation from (i), the Poisson summation formula yields

E[X2]=ππG2+1∑κ∈Zκ2⋅2κcosech⁡(κπ)=ππG2+1∑κ∈ZF[x↦x2⋅2xcosech⁡(xπ)](ξ)∣ξ=κ=ππG2+1∑κ∈Z(i2π)2d2dξ2F[x↦2xcosech⁡(xπ)](ξ)∣ξ=κ=ππG2+1∑κ∈Z−14π2d2dξ2sech⁡2(ξπ)∣ξ=κ=ππG2+1(32∑κ∈Zsech⁡4(κπ)−∑κ∈Zsech⁡2(κπ)).\begin{aligned} \mathbb{E}\left[X^{2}\right] = \frac{\pi}{\pi G^{2}+1}\sum_{\kappa\in\mathbb{Z}}\kappa^{2}\cdot2\kappa\operatorname{cosech}(\kappa\pi) \\ &= \frac{\pi}{\pi G^{2}+1}\sum_{\kappa\in\mathbb{Z}}\mathcal{F}\left[x\mapsto x^{2}\cdot2x\operatorname{cosech}(x\pi)\right](\xi)\bigg|_{\xi=\kappa} \\ &= \frac{\pi}{\pi G^{2}+1}\sum_{\kappa\in\mathbb{Z}}\left(\frac{i}{2\pi}\right)^{2}\frac{\mathrm{d}^{2}}{\mathrm{d}\xi^{2}}\mathcal{F}\left[x\mapsto2x\operatorname{cosech}(x\pi)\right](\xi)\bigg|_{\xi=\kappa} \\ &= \frac{\pi}{\pi G^{2}+1}\sum_{\kappa\in\mathbb{Z}}\frac{-1}{4\pi^{2}}\frac{\mathrm{d}^{2}}{\mathrm{d}\xi^{2}}\operatorname{sech}^{2}(\xi\pi)\bigg|_{\xi=\kappa} \\ &= \frac{\pi}{\pi G^{2}+1}\left(\frac{3}{2}\sum_{\kappa\in\mathbb{Z}}\operatorname{sech}^{4}(\kappa\pi)-\sum_{\kappa\in\mathbb{Z}}\operatorname{sech}^{2}(\kappa\pi)\right). \end{aligned}

But Proposition 5.1 yields E[X2]=Var⁡(X)=πG4/(πG2+1)\mathbb{E}[X^{2}]=\operatorname{Var}(X)=\pi G^{4}/(\pi G^{2}+1) and from (i) we get ∑κ∈Zsech⁡2(κπ)=G2+1/π\sum_{\kappa\in\mathbb{Z}}\operatorname{sech}^{2}(\kappa\pi)=G^{2}+1/\pi. Thus, the equations above hold if and only if

∑κ∈Zsech⁡4(κπ)=23(G4+G2+1π),\sum_{\kappa\in\mathbb{Z}}\operatorname{sech}^{4}(\kappa\pi)=\frac{2}{3}\left(G^{4}+G^{2}+\frac{1}{\pi}\right),

as claimed.

Ad (ii) [Analytical proof]: First, we compute the Fourier transformation of sech⁡4(κπ)\operatorname{sech}^{4}(\kappa\pi). Lemma 5.2 yields

F[x↦sech⁡4(xπ)](ξ)=43π∣(1+iξ)Γ(1+iξ)∣2=23(1+ξ2)⋅2ξcosech⁡(ξπ).\mathcal{F}\left[x\mapsto\operatorname{sech}^{4}(x\pi)\right](\xi)=\frac{4}{3\pi}\left|(1+i\xi)\Gamma(1+i\xi)\right|^{2}=\frac{2}{3}(1+\xi^{2})\cdot2\xi\operatorname{cosech}(\xi\pi).

Now, our (4) in the proof of Proposition 2.1 provides

∑κ∈Z2κcosech⁡(κπ)=G2+1π.\sum_{\kappa\in\mathbb{Z}}2\kappa\operatorname{cosech}(\kappa\pi)=G^{2}+\frac{1}{\pi}.

Further, Ramanujan’s series identity (i) of Entry 15 in Chapter 17 of his second notebook, cf. [9 p. 132], which we already used in the proof of Proposition 5.1 with x=1/2x=1/2, y=πy=\pi, and z=2Gz=\sqrt{2G}, yields

∑n=1∞n3sinh⁡(yn)=18z4x=12∑κ∈Zκ3cosech⁡(κπ)=G44.\sum_{n=1}^{\infty}\frac{n^{3}}{\sinh(yn)}=\frac{1}{8}z^{4}x=\frac{1}{2}\sum_{\kappa\in\mathbb{Z}}\kappa^{3}\operatorname{cosech}(\kappa\pi)=\frac{G^{4}}{4}.

These formulae and the Poisson summation formula finally yield

∑κ∈Zsech⁡4(κπ)=∑κ∈ZF[x↦sech⁡4(xπ)](ξ)∣ξ=κ=23(∑κ∈Z2ξcosech⁡(ξπ)∣ξ=κ+2∑κ∈Zξ3cosech⁡(ξπ)∣ξ=κ)=23(G4+G2+1π),\begin{aligned} \sum_{\kappa\in\mathbb{Z}}\operatorname{sech}^{4}(\kappa\pi)=\sum_{\kappa\in\mathbb{Z}}\mathcal{F}\left[x\mapsto\operatorname{sech}^{4}(x\pi)\right](\xi)\big|_{\xi=\kappa} \\ &=\frac{2}{3}\left(\sum_{\kappa\in\mathbb{Z}}2\xi\operatorname{cosech}(\xi\pi)\big|_{\xi=\kappa}+2\sum_{\kappa\in\mathbb{Z}}\xi^{3}\operatorname{cosech}(\xi\pi)\big|_{\xi=\kappa}\right) \\ &=\frac{2}{3}\left(G^{4}+G^{2}+\frac{1}{\pi}\right), \end{aligned}

as claimed.

Ad (iii): Recall that for X∼Cosech⁡X\sim\operatorname{Cosech} with normalization constant C=π/(πG2+1)C=\pi/(\pi G^{2}+1), we have

∑κ∈Z2κcosech⁡(πκ)=1C=G2+1π,\sum_{\kappa\in\mathbb{Z}}2\kappa\operatorname{cosech}(\pi\kappa)=\frac{1}{C}=G^{2}+\frac{1}{\pi},

and for the even moments (j∈{2,4,6}j\in\{2,4,6\}):

∑κ∈Zκj⋅2κcosech⁡(πκ)=1CE[Xj].\sum_{\kappa\in\mathbb{Z}}\kappa^{j}\cdot2\kappa\operatorname{cosech}(\pi\kappa)=\frac{1}{C}\mathbb{E}\left[X^{j}\right].

For m=3m=3, the product in Lemma 5.2 evaluates to

∏j=12(j2+ξ2)=(1+ξ2)(4+ξ2)=4+5ξ2+ξ4,\prod_{j=1}^{2}(j^{2}+\xi^{2})=(1+\xi^{2})(4+\xi^{2})=4+5\xi^{2}+\xi^{4},

with the prefactor 24/5!=2/152^{4}/5!=2/15. Thus,

F[x↦sech⁡6(πx)](ξ)=215(4+5ξ2+ξ4)⋅2ξcosech⁡(πξ).\mathcal{F}\left[x\mapsto\operatorname{sech}^{6}(\pi x)\right](\xi)=\frac{2}{15}(4+5\xi^{2}+\xi^{4})\cdot2\xi\operatorname{cosech}(\pi\xi).

Applying the Poisson summation formula and substituting E[X2]/C=G4\mathbb{E}[X^{2}]/C=G^{4} and E[X4]/C=3G6\mathbb{E}[X^{4}]/C=3G^{6} yield

∑κ∈Zsech⁡6(κπ)=215∑κ∈Z(4+5κ2+κ4)⋅2κcosech⁡(πκ)=215⋅1C(4+5E[X2]+E[X4])=215(4(G2+1π)+5G4+3G6)=25G6+23G4+815(G2+1π),\begin{aligned} \sum_{\kappa\in\mathbb{Z}}\operatorname{sech}^{6}(\kappa\pi) &=\frac{2}{15}\sum_{\kappa\in\mathbb{Z}}(4+5\kappa^{2}+\kappa^{4})\cdot2\kappa\operatorname{cosech}(\pi\kappa)\\ &=\frac{2}{15}\cdot\frac{1}{C}\left(4+5\mathbb{E}\left[X^{2}\right]+\mathbb{E}\left[X^{4}\right]\right)\\ &=\frac{2}{15}\left(4\left(G^{2}+\frac{1}{\pi}\right)+5G^{4}+3G^{6}\right)\\ &=\frac{2}{5}G^{6}+\frac{2}{3}G^{4}+\frac{8}{15}\left(G^{2}+\frac{1}{\pi}\right), \end{aligned}

as claimed.

Ad (iv): For m=4m=4, the product in Lemma 5.2 expands to

∏j=13(j2+ξ2)=(1+ξ2)(4+ξ2)(9+ξ2)=36+49ξ2+14ξ4+ξ6,\prod_{j=1}^{3}(j^2+\xi^2)=(1+\xi^2)(4+\xi^2)(9+\xi^2)=36+49\xi^2+14\xi^4+\xi^6,

with the prefactor 26/7!=4/3152^6/7!=4/315. Hence,

F[x↦sech⁡8(πx)](ξ)=4315(36+49ξ2+14ξ4+ξ6)⋅2ξcosech⁡(πξ).\mathcal{F}\left[x\mapsto\operatorname{sech}^8(\pi x)\right](\xi)=\frac{4}{315}\left(36+49\xi^2+14\xi^4+\xi^6\right)\cdot2\xi\operatorname{cosech}(\pi\xi).

Applying Poisson summation along with E[X6]/C=18G8\mathbb{E}[X^6]/C=18G^8 results in

∑k∈Zsech⁡8(kπ)=4315⋅1C(36+49E[X2]+14E[X4]+E[X6])=4315(36(G2+1π)+49G4+14⋅3G6+18G8)=4315(18G8+42G6+49G4+36(G2+1π))=835G8+815G6+2845G4+1635(G2+1π),\begin{aligned} \sum_{k\in\mathbb{Z}}\operatorname{sech}^8(k\pi) &=\frac{4}{315}\cdot\frac{1}{C}\left(36+49\mathbb{E}\left[X^2\right]+14\mathbb{E}\left[X^4\right]+\mathbb{E}\left[X^6\right]\right)\\ &=\frac{4}{315}\left(36\left(G^2+\frac{1}{\pi}\right)+49G^4+14\cdot3G^6+18G^8\right)\\ &=\frac{4}{315}\left(18G^8+42G^6+49G^4+36\left(G^2+\frac{1}{\pi}\right)\right)\\ &=\frac{8}{35}G^8+\frac{8}{15}G^6+\frac{28}{45}G^4+\frac{16}{35}\left(G^2+\frac{1}{\pi}\right), \end{aligned}

which finishes the proof. □

Remark 5.4. Identity (ii) in Theorem 5.3 provides an analytic proof for a conjecture on electrostatic lattice sums proposed by Vigren and Dieckmann [Vigren and Dieckmann, [10] via experimental mathematics. Further identities and physical applications in this direction will be investigated in subsequent work.

Leonard Jobst Eberhard Pleschberger, Neptunstraße 18, Düsseldorf D-40223 & Sechshauser Straße 24/1/15, A-1150 Vienna.

Email address: [email protected]

References

  1. [1][Apostol, 1990] Apostol, Tom M.: Modular Functions and Dirichlet Series in Number Theory (2nd edition). Springer.
  2. [2][Borwein and Borwein, 1987] Borwein, Jonathan M.; Borwein, Peter B.: Pi and the AGM. A Study in Analytic Number Theory and Computational Complexity. Wiley.
  3. [3][Bruckman, 1977] Bruckman, P. S.: On the evaluation of certain infinite series by elliptic functions. Fibonacci Quart. 15 (1977), 293–310.DOI
  4. [4][Chowla and Selberg, 1967] Chowla, S., and Selberg, A. (1967). On Epstein’s Zeta-function. Journal für die reine und angewandte Mathematik, 227, 86–110.DOI
  5. [7][Jacobi, 1829] Jacobi, C. G. J. (1829). Fundamenta nova theoriae functionum ellipticarum. Publisher Sumtibus fratrum Borntreager
  6. [8][NIST, 2026] National Institute of Standards and Technology (NIST). Digital Library of Mathematical Functions (DLMF). Version 1.2.8; Release date: September 2026.
  7. [9][Ramanujan and Berndt, 1991] Ramanujan, Srinivasa; Berndt, Bruce (1991). Ramanujan’s Notebooks. Part III. Springer.
  8. [10][Vigren and Dieckmann, 2020] Vigren, Erik; Dieckmann, Andreas (2020). Simple Solutions of Lattice Sums for Electric Fields Due to Infinitely Many Parallel Line Charges. Symmetry (2020), 12(6), 1040.DOI
  9. [11][Whittaker and Watson, 1920] Whittaker, Edmund T.; Watson, George N. (1920). A Course of Modern Analysis; an Introduction to the General Theory of Infinite Processes and of Analytic Functions; with an Account of the Principal Transcendental Functions (3rd Edition). Cambridge University Press.
  10. [12][Zucker, 1979] Zucker, I. J.: The Summation of Series of Hyperbolic Functions. SIAM J. Math. Anal. 10(1) (1979), 192–206.DOI

Paper details

vlnc:W_agbzztd6
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

Not recorded for this paper

View statement
Looking for related papers…
Contents