10

$$\int_0^1 \prod_{n\geq 1} (1-x^n) \, dx = \frac{4\pi\sqrt{3}\sinh(\frac{\pi}{3}\sqrt{23})}{\sqrt{23}\cosh(\frac{\pi}{2}\sqrt{23})}$$

This monstrous expression is from Tolaso Network (tolaso.com.gr). I have no idea how to approach it - converting the product to a sum of logarithms does no good, and one cannot switch the order of product/integral either. The product in itself doesn't converge to anything nice either.

I am interested in seeing the proof of the above identity, as well as an explanation of how exactly $\sqrt{23}$ becomes involved in such a deceptive integral. Both real and complex analytic solutions are welcome. A proof without the pentagonal number theorem would be nice as well, since that somewhat trivializes the problem.

Teoc
  • 8,700

2 Answers2

10

Note that $$\prod_{n=1}^\infty(1-x^n)=\sum_{m=-\infty}^\infty(-1)^mx^{m(3m+1)/2} $$ (Euler's pentagonal number formula), so the integral equals \begin{align} I&=\sum_{m=-\infty}^\infty(-1)^m\frac2{3m^2+m+2}\\ &=\frac4{\sqrt{23}} \sum_{m=-\infty}^\infty(-1)^m\left(\frac{1}{m+(1-\sqrt{23})/6} -\frac{1}{m+(1+\sqrt{23})/6}\right). \end{align} Now you can attack this with identities for the digamma function.

Angina Seng
  • 158,341
7

As suggested by Pisco, use the Pentagonal number theorem, for $|x|<1$, $$\prod\limits_{n=1}^{+\infty}(1-x^n)=\sum\limits_{m=-\infty}^{+\infty}(-1)^m x^{\frac{3m^2-m}{2}}.$$ Then by integrating each term (here some details are needed), we get $$2\sum\limits_{m=-\infty}^{+\infty}\frac{(-1)^m }{3m^2-m+2}.$$ Now note that the discriminant of the quadratic polynomial at the denominator is the number $-23$.

Can you take it from here?

Robert Z
  • 145,942