5
$\begingroup$

I recall from complex analysis a while ago that the series $$\sum_{n=1}^\infty \frac{z^{n!}}{n}$$ has radius of convergence 1, diverges on roots of unity, and converges elsewhere on the circle. It's simple enough to see that it has to diverge on roots of unity: consider $z$ such that $z^k = 1$ for some $k \in \mathbb{N}.$ Plug this into the series and you see that past the $k$'th term, $z^{n!} = (z^k)^m = 1$ for some $m \in \mathbb{N}$; hence, the series becomes harmonic and must diverge.

But I don't recall how to show that the series converges for other values on the circle. If this is simple I'd like just a hint, if it is involved I would prefer a sparse outline of the proof. Thank you!

$\endgroup$
1
  • $\begingroup$ @M. Winter: OK, I have not well understood the OP question. I delete my answer. Thank you very much. $\endgroup$
    – Kelenner
    Mar 8, 2017 at 13:13

2 Answers 2

4
$\begingroup$

Convention

On the unit circle, we write $z=e^{2\pi i \theta}$ with $\theta\in\mathbb{R}$. Then the roots of unity are exactly when $\theta\in\mathbb{Q}$. Let $\{\alpha\}$ be the fractional part of $\alpha$. Since $z^{n!} = e^{2\pi i \theta n!}$, the distribution of $\{\theta n!\}$ becomes important to consider.

There is an irrational $\theta$ that the series diverges

If we take $\theta = e$, then the series diverges. This is because $\{en!\}$ converges to $0$.

The set of all $\theta$ that the series diverges is a $G_{\delta\sigma}$ set

This result is intensively discussed in this post in MO. This is because the set of all $\theta$ that the series converges is a $F_{\sigma\delta}$ set.

For almost all $\theta$ the series converges

This is due to a result by Lebeque. The function $g(\theta, n) = \theta n!$ satisfies the hypotheses of Theorem 1 in the paper. Therefore, for every $\epsilon>0$, the series $$ \sum_{n=1}^{\infty} \frac{ e^{2\pi i \theta n!} } {n^{1/2+ \epsilon}} $$ converges for almost all $\theta$. This shows that the series with $\epsilon=\frac12$ converges for almost all $\theta$. Hence the set of all $\theta$ that the series diverges is a $G_{\delta\sigma}$ set of Lebesgue measure zero which contains all rational numbers and $e$, and it does not contain $e/2$ by @user254665's answer.

$\endgroup$
2
  • $\begingroup$ When $z=\exp( \pi i e) $we have $z^n!/n =1/n +O(1/n^2)$ as $n\to \infty$ so the summation diverges.............+1 $\endgroup$ Mar 9, 2017 at 7:13
  • $\begingroup$ My example is $\exp(2\pi i e)$, not $\exp(\pi i e)$. $\endgroup$ Mar 9, 2017 at 16:48
1
$\begingroup$

A specific example.

For $n\geq 1$, the integer part of $n!e$ is odd when $n$ is even, and even when $n$ is odd.

So for $z=\exp (e\pi i )$ we have $z^{n!}/n=(-1)^{n+1}/n +O(1/n^2)$ as $n\to \infty$, and the series converges.

$\endgroup$
2
  • 1
    $\begingroup$ (+1) Very nice to see a specific value of $z$ that the series converges which I didn't have in my answer. I think we can make a lot more examples like this by using $e$. $\endgroup$ Mar 9, 2017 at 18:14
  • $\begingroup$ @i707107. I got the idea from your A. My first thought was to try for $\exp (\pi i z)$ where $z$ is of the form $\sum_n A_n/n1! $ for some suitable series $(A_n)_n$ rapidly converging to $0$, but that looked complicated. $\endgroup$ Mar 9, 2017 at 21:59

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .