3
$\begingroup$

I know

\begin{align} \prod_{p~is~ prime} \frac{1}{1-\frac{1}{p^2}} = \zeta(2) = \frac{\pi^2}{6} \end{align} which has a convergent number.

actually I can even generalized this to

\begin{align} \prod_{p ~is~ prime} \frac{1}{1-\frac{1}{p^s}} = \zeta(s) \end{align} For $s>1$ [consider $s\in \mathbb{R}$] we know zeta function converges, so this has a convergent number.

How about generalization to arbitrary integers? [i.e., I want to replace $p$ with arbitrary integer $n$.]

For example $s=2$, we have

\begin{align} \prod_{n>1} \frac{1}{1-\frac{1}{n^2}} \end{align}

taking log we need to show \begin{align} - \sum_{n=2} \log\left(1-\frac{1}{n^2} \right) \end{align} is convergent or not.

simply by telescope method I can see this value converges to $\log(2)$, that means $ \prod_{n>1} \frac{1}{1-\frac{1}{n^2}} = 2$.

Now consider $s>1$.

\begin{align} -\sum_{n>1} \log\left(1-\frac{1}{n^s}\right) \end{align} This is convergent from comparison test.

Simply take $a_n = -\log(1-\frac{1}{n^s})$ and $b_n = \frac{1}{n^s}$, then \begin{align} \lim_{n\rightarrow \infty} \frac{a_n}{b_n} = \lim_{x\rightarrow 0} \frac{-\log(1-x)}{x} = 1 >0 \end{align} and since $\sum_{n=1}^{\infty} b_n = \zeta(s)$ is convergent for $s>1$, $\sum_{n=2}^{\infty} a_n$ also converges.

What I want to obtain is the value of such convergent series, first i tried telescope method, but it seems difficult even for $s=3$.

Is there a way to compute exact value of those products?

How and what is the values of those products?

$\endgroup$
  • 1
    $\begingroup$ As far as I'm concerned no closed form formula is known for $s>2$. By the way, when $s=2$ the product equals $2$, not $1/2$. $\endgroup$ – Klangen Jul 16 at 9:03
  • 2
    $\begingroup$ @Klangen, thanks, I found typos and corrected it! $ \sum_{n=2}^{\infty} \log(1-\frac{1}{n^2} ) = - \log(2)$. I am quite surprised that for $s>2$, no closed form formula is known! $\endgroup$ – phy_math Jul 16 at 9:13
  • 1
    $\begingroup$ math.stackexchange.com/q/2603561 $\endgroup$ – user514787 Jul 16 at 9:53
3
$\begingroup$

For $2k\ge 2$ $$\prod_{n=2}^\infty (1-\frac{1}{n^{2k}})=\prod_{m=1}^{2k} \prod_{n=2}^\infty (1-\frac{e^{2i \pi m/(2k)}}{n}) = \prod_{m=1}^{k}\prod_{n =-\infty, |n|\ge 2}^\infty (1-\frac{e^{2i \pi m/(2k)}}{n}) = \prod_{m=1}^{k} f(e^{2i \pi m/(2k)})$$ where $f(x) =\frac{\sin(\pi x)}{\pi x(1-x^2)} $ and in those products the order of summation is meant to be $\lim_{N \to \infty} \prod_{|n| \le N} $

For $k \ge 2$ $$\prod_{n=2}^\infty (1-\frac{1}{n^k})=\prod_{m=1}^kg(e^{2i \pi m/k}), \qquad g(x) =\frac{1}{(-x)(1-x)\Gamma(-x)}= \frac{1}{\Gamma(2-x)}$$

$\endgroup$
  • $\begingroup$ what is $g(x) = \frac{1}{\Gamma(-x)(-x)(1-x)}$?, you mean $\Gamma(-x)\Gamma(-x)\Gamma(1-x)$? $\endgroup$ – phy_math Jul 16 at 10:08
  • $\begingroup$ @phy_math: he means $\;g(x)=\dfrac{1}{x(x-1)\Gamma(-x)}\;$ which returns the correct result (notice that he is computing the multiplicative inverse of your formula). $\endgroup$ – Raymond Manzoni Jul 16 at 10:26
  • $\begingroup$ @Raymond Manzoni Thanks! I got it $\endgroup$ – phy_math Jul 16 at 10:38
5
$\begingroup$

Let $$P_s=\prod_{n=2}^\infty \frac{1}{1-n^{-s}}$$

Using a CAS, there are some nice expressions such as $$\left( \begin{array}{cc} s & P_s \\ 2 & 2 \\ 3 & 3 \pi \text{sech}\left(\frac{\sqrt{3} \pi }{2}\right) \\ 4 & 4 \pi \text{csch}(\pi ) \\ 6 & 6 \pi ^2 \text{sech}^2\left(\frac{\sqrt{3} \pi }{2}\right) \end{array} \right)$$ $P_5$ and the other ones are quite ugly.

Have a look here.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.