8
$\begingroup$

Conjecture: \begin{align} \lim_{n\to \infty } \, \frac{\left(\prod _{k=1}^n \phi (k)\right){}^{1/n}}{n}\sim 0.2059\text{...} \end{align}

The numerical result from 100000 terms is:

enter image description here

enter image description here

My questions are:

1) exists this limit ?

2) if yes, what is closed form of this constant ?

I am sure that this limit is NOT equal to 6/Pi^2 * exp(-1) = 0.2236438825...

$\endgroup$
  • 11
    $\begingroup$ Yes and it equals $\frac{1}{e}\prod_p (1-1/p)^{1/p} $. $\endgroup$ – Lucia Feb 5 '16 at 18:04
11
$\begingroup$

We may just write down $\varphi(k)=k\cdot \prod_{p|k} (1-1/p)$, $p$ runs over the primes which divide $k$, then $$ \frac{\left(\prod_{k\leqslant n}\varphi(k)\right)^{1/n}}{n}=\frac{\sqrt[n]{n!}}n\prod_{p} (1-1/p)^{\frac{[n/p]}{n}}. $$ First multiple tends to $1/e$ by weak version of Stirling's formula, and in the product we may replace each exponent $\frac{[n/p]}n$ to $1/p$ for $p\leqslant n$ with multiplicative error between 1 and $$ \prod_{p\leqslant n} (1-1/p)^{1/n}\geqslant \prod_{2\leqslant k\leqslant n}(1-1/k)^{1/n}=n^{-1/n}\rightarrow 1. $$ Hence the limit exists and equals $e^{-1}\prod_p (1-1/p)^{1/p}$ as Lucia and so-called friend Don have already said.

$\endgroup$
8
$\begingroup$

Yes, the limit exists, and is equal to $\exp(-1 + \sum_{p} \frac{1}{p}\log(1-\frac{1}{p}))$ (as Lucia said), where $p$ runs over primes. Hint: Reduce the problem to computing $\lim_{N\to \infty} \frac{1}{N} \sum_{n \le N} \log \frac{\phi(n)}{n}$. The arithmetic function $\log\frac{\phi(N)}{N}$ is strongly additive and (under favorable conditions which hold here) the mean value of a strongly additive function $g$ is given by $\sum_{p} g(p)/p$.

$\endgroup$

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.