Determine if the following series converges absolutely:

$$\sum_{k=1}^\infty \sin(kx)$$



I know how to deal with whether they converge. First one diverge by n-term test, second one converge by Dirichlet's Test. However, I do not know how to deal with whether they converge absolutely. Can someone teach me how to determine if both series converge absolutely. Thanks.

  • 1
    $\begingroup$ The first converges only if $x\in\pi\mathbb Z$, and then it also converges absolutely. $\endgroup$ – Hagen von Eitzen Apr 28 '14 at 15:46
  • $\begingroup$ what about other values of x? $\endgroup$ – user10024395 Apr 28 '14 at 15:51

The series $$\tag1 \sum_{k=1}^\infty \sin(kx)$$ converges absolutely for $x\in\pi \mathbb Z$. For all other cases, it doesn' even converge: Note that $$ \sin((k+1)x)=\sin(kx)\cos x+\cos(kx)\sin x$$ hence whenever $\sin(kx)\approx 0$ and hence $|\cos(kx)|\approx1$, then $|\sin((k+1)x)|\approx|\sin x|$. More precisely, if $\sin x\ne0$, we can find $a>0$ with $$ \sqrt{1-a^2}\cdot |\sin x|>2a.$$ Then $|\sin(kx)|<a$ implies $|\cos(kx)|>\sqrt{1-a^2}$ and hence $$ |\sin((k+1)x)\ge |\cos(kx)\sin x|-|\sin(kx)\cos x|> \sqrt{1-a^2}|\sin x|-a>a$$

We conclude that $\sin(kx)\not\to 0$ unless $\sin x=0$.

In summary, the series $(1)$

  • converges absolutely (trivially) for $x\in\pi\mathbb Z$
  • diverges for $x\notin\pi\mathbb Z$.

We have seen above that for $x\notin\pi\mathbb Z$, at least one of two successive summands is $>a$ in absolute value. Apply this to $$\tag2 \sum_{k=1}^\infty \left|\sin(kx)\sin(\tfrac1{kx})\right|$$ For $k$ sufficiently large we have $\sin\frac1{kx}\approx\frac1{kx}$, say $|\sin\frac1{kx}|\ge\frac1{2k|x|}$. Then for such $k$ we have $$\left|\sin(kx)\sin(\tfrac1{kx})\right|+\left|\sin((k+1)x)\sin(\tfrac1{(k+1)x})\right| \ge \frac{a}{2(k+1)|x|}.$$ We conclude by comparision with the harmonic series that $(2)$ diverges for $x\notin\pi\mathbb Z$. In summary, the series $$ \sum_{k=1}^\infty\sin(kx)\sin\frac1{kx}$$

  • is not defined for $x=0$
  • converges absolutely (trivially) for $x\in\pi\mathbb Z\setminus\{0\}$
  • converges (as you have shown, by the Dirichlet test), but not absolutely, for $x\in\mathbb R\setminus\pi\mathbb Z$.
  • $\begingroup$ +1 Nice argument, self-contained and does not require equidistribution theorem, unlike the one I proposed. $\endgroup$ – Marcin Łoś Apr 28 '14 at 16:11
  • $\begingroup$ i don't really understand this part: The "≈" of above can be made precise i nthe followin sense: Unless sinx=0, there exists a positive number a>0 such that |sin(kx)|<a implies |sin((k+1)x)>a. In other words: at least one of two successive summands is >a in absolute value. $\endgroup$ – user10024395 Apr 28 '14 at 16:25
  • $\begingroup$ can u explain more pls thanks $\endgroup$ – user10024395 Apr 28 '14 at 16:26
  • $\begingroup$ i am really sorry for asking you another time, but even after the edit, i still don't understand why we can find such a $\endgroup$ – user10024395 Apr 28 '14 at 23:56

The first one, as it diverges, cannot converge absolutely (except for trivial case $x\in \pi\mathbb{Z}$). As for the second one - it is not absolutely convergent for most values of $x$ (when $x$ is not a rational multpile of $\pi$), the reason being roughly that in such case "many" values of $sin\ kx$ are "quite large", greater in absolute value than some fixed $0<a<1$. As the second term - $\sin \frac{1}{kx}$ - can be approximated by $\frac{1}{kx}$, which is divergent, it follows that partial sums are unbounded. Of course, it's just a vague sketch of a proof. You can find precise formulation of the above ideas and relevant results here


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.