A problem in Makarov's Selected problems in real analysis asks to investigate the convergence of $\displaystyle \sum_n \frac{|\sin(n^2)|}{n}$

I'm clueless at the moment. I can't find any good property of the sequence $|\sin(n^2)|$.

$|\sin(n^2)|$ is small whenever $n\sim \sqrt{p\pi}$, and, as $p\to \infty$, the $\sqrt{p\pi}$ get closer to each other since $\sqrt{(p+1)\pi}-\sqrt{p\pi}\sim \frac 12 \sqrt{\frac{\pi}{p}}$.

Any hint is appreciated.

  • $\begingroup$ The letter $p$ doesn't mean "prime", does it? Just in case. $\endgroup$ – ajotatxe Aug 23 '16 at 20:18
  • $\begingroup$ @ajotatxe no, just any integer $\endgroup$ – Gabriel Romon Aug 23 '16 at 20:18
  • $\begingroup$ I suspect this diverges because it's similar to $\sum_{n=1}^\infty(\frac{1}{n})$ however, that sequence unfortunately bounds our sequence of interest from above rather than below (and hence isn't useful for proving divergence). $\endgroup$ – Justin Benfield Aug 23 '16 at 20:45
  • 1
    $\begingroup$ Your series diverges with the above absolute value, and converges without it. It's a tricky one. As a hint, try considering the average value of sine (with and without absolute value) along a single period. $\endgroup$ – Shlomi A Aug 23 '16 at 20:50
  • $\begingroup$ @ShlomiA is equidistribution what you were hinting at ? $\endgroup$ – Gabriel Romon Aug 23 '16 at 21:39

Weyl's equidistribution theorem, see e.g. Weyl-Tao, Corollary 6 implies that $n^2$ is asymptotically equidistributed mod $2\pi$. Thus as $N\rightarrow +\infty$: $$ M_N=\frac{1}{N} \sum_{n=1}^N |\sin(n^2)| \rightarrow \frac{1}{2\pi}\int_0^{2\pi} |\sin(t)|dt=\frac{2}{\pi}$$ Now, find a sequence $N_k$, $k\geq 1$, so that $10N_k\leq N_{k+1}$ and $M_{N_k}\geq \frac{1}{\pi}$. Then for every $k\geq 1$ $$ \sum_{n=N_k+1}^{N_{k+1}} \frac{|\sin n^2|}{n} \geq \frac{1}{N_{k+1}}\sum_{n=N_k+1}^{N_{k+1}} |\sin n^2| \geq M_{N_{k+1}} - \frac{N_k}{N_{k+1}} \geq \frac{1}{\pi}-\frac{1}{10}>0$$ Summing over $k$ we conclude the divergence of $\sum_{n\geq 1} |\sin(n^2)|/n$

One may note that the result (as well as the proof) goes through when replacing $n^2$ by any polynomial in $n$, $p(n)=a_d n^d+\cdots +a_0$ as long as the leading coefficient $a_d$ is rationally independent from $\pi$.

  • $\begingroup$ Ok, thanks for the clarification, great proof! $\endgroup$ – Gabriel Romon Aug 23 '16 at 21:30
  • $\begingroup$ @H-H-Rugh, nice! Can you demonstrate that it converges without the absolute value? $\endgroup$ – Shlomi A Aug 24 '16 at 18:52
  • $\begingroup$ @ShlomiA Nope, excellent question. I have thought about it but for that the equidistribution has to be sufficiently rapid and that is not guaranteed by Weyl (as far as I know). $\endgroup$ – H. H. Rugh Aug 24 '16 at 18:57
  • $\begingroup$ @ShlomiA PS: I have posed the question on MO (I don't think it has an easy answer). $\endgroup$ – H. H. Rugh Aug 24 '16 at 19:39
  • $\begingroup$ @H-H-Rugh, how about the Dirichlet convergence test? I think that oughtta do it. $\endgroup$ – Shlomi A Aug 24 '16 at 19:42

The series diverges. The sloppy reasoning is that $|\sin(n^2)|$ is roughly a random value between $0$ and $1$ with an average greater than or equal to $0.1$ and $\sum_n \frac{0.1}{n}$ certainly diverges.

To make this more rigorous, as @H.H.Rugh points out in his solution, as $n \to \infty$, $n^2$ is equidistributed mod $\pi$ (using Corollary 6 from here). Thus

$$|\sin(n^2)| > \frac{1}{2} \ \text{ whenever }\ n^2 \text{ mod }\pi \in \left(\frac{\pi}{6},\frac{5\pi}{6}\right)$$

which happens two thirds of the time (asymptotically). Therefore for large enough $N \in \mathbb{Z}$,

$$\sum_{n=N+1}^{2N} \frac{|\sin(n^2)|}{n}> \frac{1}{2N} \sum_{n=N+1}^{2N} |\sin(n^2)| > \frac{1}{2N} \frac{N}{2} \frac{1}{2} = \frac{1}{8}$$

since $|\sin(n^2)| > \frac{1}{2}$ for more than one half of the $n$ in $\{N+1,...,2N\}$

Therefore the series diverges since the tail of the series diverges:

$$\sum_{n=N+1}^{\infty} \frac{|\sin(n^2)|}{n} = \sum_{k=0}^\infty \left(\sum_{n=2^k N+1}^{2^{k+1}N} \frac{|\sin(n^2)|}{n} \right)> \sum_{k=0}^\infty \frac{1}{8} = \infty$$

  • $\begingroup$ it's Corollary* and thanks for the heuristics $\endgroup$ – Gabriel Romon Aug 24 '16 at 8:44

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.