7

Let $$ \alpha = \prod_{k=1}^{\infty} \left(1- \frac{1}{e^{ \sqrt{2} \pi k}}\right) $$ and $$ \beta = \prod_{k=1}^{\infty} \left(1 + \frac{1}{e^{ \sqrt{2} \pi k}}\right) = \frac{\exp \left(\frac{\sqrt{2}\pi }{24}\right)}{\sqrt[4]{2}} $$

Then $$ \sum_{n=1}^{\infty} \text{arctanh} \left(\frac{1}{e^{ \sqrt{2} \pi n}} \right) = -\text{arctanh} \left( \frac{\alpha - \beta}{\alpha + \beta}\right) \approx 0.0119025137323901862 $$

Is there a way to solve for the product $\alpha$? I figure that it's likely given that $\beta$ can be.

  • That's $f(z)=\prod(1-z^i)$ for $z=e^{-\sqrt2}.$ I have a vague recollection that that is related to partition functions, and has a name, or at least is related to a function with a name. Specifically, $1/f(z)=\sum_{n=0}^{\infty} p_nz^n$ where $p_n$ is the number of partitions of $n.$ – Thomas Andrews Nov 05 '22 at 03:27
  • 1
    How did you get the closed formula for $\beta?$ – Thomas Andrews Nov 05 '22 at 03:30
  • That would be the Q-Pochhammer function, I think. I was hoping that it could be expanded like similar to the $\beta$ instead of that. – DecarbonatedOdes Nov 05 '22 at 03:30
  • @ThomasAndrews I used the relation: $$ x^\frac{1}{4} e^{\frac{-x}{12}} QPochhammer[e^{-2x},e^{-2x}] = y^\frac{1}{4} e^{\frac{-y}{12}} QPochhammer[e^{-2y},e^{-2y}] $$ for $x*y=\pi^2$ – DecarbonatedOdes Nov 05 '22 at 03:37
  • 1
    It belongs to the theory of modular forms. $j(i\sqrt 2)$ and $\Delta(2i \sqrt2)/\Delta(i \sqrt2)$ have a closed-form but $\Delta(i\sqrt 2)$ probably doesn't @ThomasAndrews – reuns Nov 05 '22 at 08:30
  • 1
    With Mathematica: $\prod _{k=1}^{\infty } \left(1-\frac{1}{\exp \left(\sqrt{2} \pi k\right)}\right)=\exp \left(\sum _{k=1}^{\infty } \ln \left(1-\frac{1}{\exp \left(\sqrt{2} \pi k\right)}\right)\right)=e^{\frac{\pi }{12 \sqrt{2}}} \eta \left(\frac{i}{\sqrt{2}}\right)$ where: $\eta \left(\frac{i}{\sqrt{2}}\right)$ is Dedekind eta modular elliptic function. – Mariusz Iwaniuk Nov 05 '22 at 12:26
  • 1
    Then we have: $\sum _{n=1}^{\infty } \tanh ^{-1}\left(\frac{1}{\exp \left(\sqrt{2} \pi n\right)}\right)=-\frac{1}{8} \ln \left(2 \eta \left(\frac{i}{\sqrt{2}}\right)^4\right)$ – Mariusz Iwaniuk Nov 05 '22 at 12:33

1 Answers1

6

Let $q=e^{-\pi\sqrt{2}}$ and then the product $\alpha$ in question is $$\alpha=\prod_{n=1}^{\infty}(1-q^n)\tag{1}$$ This is related to the famous Dedekind eta function given by $$\eta(q) =q^{1/24}\prod_{n=1}^{\infty}(1-q^n)\tag{2}$$ which in turn has a closed form in terms of elliptic modulus $k$ associated with nome $q$ and complete elliptic integral of first kind $K(k) $: $$\eta(q) =2^{-1/6}\sqrt{\frac{2K(k)}{\pi}}k^{1/12}k'^{1/3}\tag{3}$$ where \begin{align} k&=\frac{\vartheta_{2}^2(q)}{\vartheta_3^2(q)}\tag{4a}\\ k'&=\sqrt{1-k^2}\tag{4b}\\ K(k)&=\int_0^{\pi/2}\frac{dx}{\sqrt{1-k^2\sin^2x}}\tag{4c}\\ \vartheta_2(q)&=\sum_{n\in\mathbb{Z}}q^{(n+(1/2))^2}\tag{4d}\\ \vartheta_3(q)&=\sum_{n\in\mathbb{Z}}q^{n^2}\tag{4e} \end{align} Both Jacobi and Ramanujan established that if $p$ is a positive rational number and $q=e^{-\pi\sqrt{p}} $ then the modulus $k$ in $(4a)$ is an algebraic number. Chowla and Selberg further proved (in this paper) that for such $q$ and corresponding $k$ the elliptic integral $K(k) $ can be expressed in closed form containing $\pi$ and values of Gamma function at rational points.

The evaluation of $k$ for $p=2,q=e^{-\pi\sqrt{2}}$ is done via modular equation of degree $2$ and one can show that $$k=\sqrt{2}-1,k' =\sqrt{2(\sqrt{2}-1)}$$ The value of $K(k) $ is given by $$K(k) = \frac{\sqrt{\sqrt{2} +1} \Gamma (1/8)\Gamma (3/8)}{2^{13/4}\sqrt{\pi}} $$ and is evaluated in this answer. Using these values the value of $\eta(q) $ in $(2)$ is obtained and then $\alpha =q^{-1/24}\eta(q)$ gets evaluated in closed form.

The product $\beta$ is simpler to handle because we have $$\beta=q^{-1/24}\cdot\frac{\eta(q^2)}{\eta(q)}\tag{5}$$ and we have $$\eta(q^2)=2^{-1/3}\sqrt {\frac{2K(k)}{\pi}}(kk')^{1/6}\tag {6}$$ and thus $\beta$ is $e^{\pi\sqrt {2}/24}$ times an algebraic number (which turns out to be $2^{-1/4}$ using given values of $k, k'$).

  • Amazing. Thank you – DecarbonatedOdes Nov 28 '22 at 05:13
  • I don't understand how you relate $\Gamma(1/8)\Gamma(3/8)$ with $\int_0^{\pi/2} \frac{dx}{\sqrt{1-(\sqrt2-1)^2 \sin^2 x}}$ please avoid linking to very long answers linking to more long answers not giving much insights – reuns Nov 28 '22 at 11:09
  • @reuns: the linked answer gives complete proof of the evaluation of this integral, but it uses the expression of this integral as hypergeometric function and further identities related to such functions. The proof is complicated but self contained in that answer. – Paramanand Singh Nov 28 '22 at 11:11
  • Another option is to use Selberg Chowla paper on Epstein Zeta functions @reuns. – Paramanand Singh Nov 28 '22 at 11:12
  • This doesn't answer my question. The evaluation of $\Delta(i)$ is based on the beta function identity. How does it generalize, and why do we get 2 or 3 (or more) $\Gamma$ terms. – reuns Nov 28 '22 at 11:20
  • @reuns: the evaluation of $\Delta(i), \Delta(i\sqrt{2}),\Delta(i\sqrt{3})$ is classical and based on evaluating corresponding elliptic integrals. The generalisation to $\Delta(i\sqrt{n}) $ is using Selberg Chowla formula. They prove that the expression consists of product of gamma values and powers of $\pi$. Let me update my answer with a link to their paper. – Paramanand Singh Nov 28 '22 at 11:27