Take the 2-minute tour ×
Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

Let $\alpha$ be a real number. I'm asked to discuss the convergence of the series $$ \sum_{k=1}^{\infty} \frac{\sin{(kx)}}{k^\alpha} $$ where $x \in [0,2\pi]$.

Well, I show you what I've done:

  • if $\alpha \le 0$ the series cannot converge (its general term does not converge to $0$ when $k \to +\infty$) unless $x=k\pi$ for $k=0,1,2$. In other words, if $\alpha \le 0$ there is pointwise convergence only in $x=0,\pi,2\pi$.

  • if $\alpha \gt 1$, I can use the Weierstrass M-test to conclude that the series is uniformly convergent hence pointwise convergent for every $x \in [0,2\pi]$. Moreover the sum is a continuous function in $[0,2\pi]$.

Would you please help me in studying what happens for $\alpha \in (0,1]$? Are there any useful criteria that I can use?

Does the series converge? And what kind of convergence is there? In case of non uniform but pointwise convergence, is the limit function continuous?

Thanks.

share|improve this question
    
Perhaps an analysis of the $\zeta(x)$ series may help reveal some answers. –  Argon Jun 7 '12 at 20:10
3  
try this en.wikipedia.org/wiki/Dirichlet's_test –  Norbert Jun 7 '12 at 20:12
    
Dirichlet's test seems a very beautiful tool! Thanks! Using Dirichlet, I have pointwise convergence for every $x \in [0,2\pi]$ also when $\alpha \in (0,1]$. What about the uniform convergence? –  Romeo Jun 7 '12 at 20:16

3 Answers 3

up vote 3 down vote accepted

The main problem (at least for me) is to prove that sum of the series $$ f_\alpha(x)=\sum\limits_{k=1}^\infty\frac{\sin kx}{k^\alpha},\quad x\in[0,2\pi] $$ is discontinuous for $\alpha\in(0,1]$.

Lemma 1. For $\alpha\in(0,1]$ we have $$ \sum\limits_{k=1}^\infty\int\limits_{k-1/2}^{k+1/2}\frac{\sin(xt)}{k^\alpha}dt= \int\limits_{1/2}^\infty\frac{\sin(xt)}{t^\alpha}dt+\varphi(x) $$ where $|\varphi(x)|\leq 2^{1-\alpha}$ for all $x\in[0,2\pi]$.

Proof. It is enough to show that difference between this sum and this integral is bounded by some constant. Now, we make estimation $$ \left|\sum\limits_{k=1}^\infty\int\limits_{k-1/2}^{k+1/2}\frac{\sin(xt)}{k^\alpha}dt- \int\limits_{1/2}^\infty\frac{\sin(xt)}{t^\alpha}dt\right|= \left|\sum\limits_{k=1}^\infty\int\limits_{k-1/2}^{k+1/2}\left(\frac{\sin(xt)}{k^\alpha}- \frac{\sin(xt)}{t^\alpha}\right)dt\right|\leq $$ $$ \sum\limits_{k=1}^\infty\int\limits_{k-1/2}^{k+1/2}|\sin(xt)|\left|\frac{1}{k^\alpha}-\frac{1}{t^\alpha}\right|dt\leq \sum\limits_{k=1}^\infty\int\limits_{k-1/2}^{k+1/2}\left|\frac{1}{k^\alpha}-\frac{1}{t^\alpha}\right|dt= $$ $$ \sum\limits_{k=1}^\infty\left(\int\limits_{k-1/2}^{k}\left(\frac{1}{t^\alpha}-\frac{1}{k^\alpha}\right)dt+\int\limits_{k}^{k+1/2}\left(\frac{1}{k^\alpha}-\frac{1}{t^\alpha}\right)dt\right)= \sum\limits_{k=1}^\infty\left(\int\limits_{k-1/2}^{k}\frac{1}{t^\alpha}dt-\int\limits_{k}^{k+1/2}\frac{1}{t^\alpha}dt\right)\leq $$ $$ \sum\limits_{k=1}^\infty\left(\int\limits_{k-1/2}^{k}\frac{1}{(k-1/2)^\alpha}dt-\int\limits_{k}^{k+1/2}\frac{1}{(k+1/2)^\alpha}dt\right)\leq \sum\limits_{k=1}^\infty\left(\frac{1}{2(k-1/2)^\alpha}-\frac{1}{2(k+1/2)^\alpha}\right)=2^{\alpha-1} $$

Lemma 2. For $\alpha\in(0,1]$ we have $$ \sum\limits_{k=1}^\infty\frac{\sin kx}{k^\alpha}= \frac{x^\alpha}{2\sin(x/2)}\int\limits_{x/2}^\infty\frac{\sin y}{y^\alpha}+\frac{x\varphi(x)}{2\sin(x/2)} $$ where $|\varphi(x)|\leq 2^{1-\alpha}$ for all $x\in[0,2\pi]$.

Proof. Note that $$ \int\limits_{k-1/2}^{k+1/2}\sin(xt)dt= -\frac{1}{x}\cos(xt)\biggl|_{k-1/2}^{k+1/2}= \frac{2\sin(kx)\sin(x/2)}{x} $$ so, $$ \sin(kx)=\frac{x}{2\sin(x/2)}\int\limits_{k-1/2}^{k+1/2}\sin(xt)dt $$ Hence from lemma 1 we conclude $$ \sum\limits_{k=1}^\infty\frac{\sin kx}{k^\alpha}= \sum\limits_{k=1}^\infty\frac{x}{2k^\alpha\sin(x/2)}\int\limits_{k-1/2}^{k+1/2}\sin(xt)dt= \frac{x}{2\sin(x/2)}\sum\limits_{k=1}^\infty\int\limits_{k-1/2}^{k+1/2}\frac{\sin(xt)}{k^\alpha}dt= $$ $$ \frac{x}{2\sin(x/2)}\left(\int\limits_{1/2}^\infty\frac{\sin(xt)}{t^\alpha}dt+\varphi(x) \right) $$ Making substitution $y=tx$ we get $$ \sum\limits_{k=1}^\infty\frac{\sin kx}{k^\alpha}= \frac{x}{2\sin(x/2)}\left(\frac{1}{x^{1-\alpha}}\int\limits_{x/2}^\infty\frac{\sin y}{y^\alpha}dy+\varphi(x) \right)= \frac{x^\alpha}{2\sin(x/2)}\int\limits_{x/2}^\infty\frac{\sin y}{y^\alpha}dy+\frac{x\varphi(x)}{2\sin(x/2)} $$

Corollary 3. For $\alpha\in(0,1]$ the function $f_\alpha$ is discontinuous at $0$.

Proof. Obviously $f_\alpha(0)=0$. Let $\alpha\in(0,1)$, then from fromula proved in lemma 2 we see that $$ \lim\limits_{x\to +0}f_\alpha(x)=\lim\limits_{x\to +0}\left(\frac{x^\alpha}{2\sin(x/2)}\int\limits_{x/2}^\infty\frac{\sin y}{y^\alpha}dy+\frac{x\varphi(x)}{2\sin(x/2)}\right) $$ Since $\varphi$ is bounded then the second term is bounded while the first tends to infinity. Hence the last limit is $\lim\limits_{x\to +0}f_\alpha(x)=+\infty$.

If $\alpha=1$, then $|\varphi(x)|\leq 1$ and since $$ \lim\limits_{x\to+0}\frac{x}{2\sin(x/2)}=1, $$ then $$ \left|\frac{x\varphi(x)}{2\sin(x/2)}\right|<\frac{\pi}{3} $$ for some $\delta_1>0$ and $x\in(0,\delta_1)$. Since $$ \lim\limits_{x\to+0}\frac{x}{2\sin(x/2)}\int\limits_{x/2}^\infty\frac{\sin y}{y}dy= \int\limits_{0}^\infty\frac{\sin y}{y}dy=\frac{\pi}{2} $$ then $$ \frac{x}{2\sin(x/2)}\int\limits_{x/2}^\infty\frac{\sin y}{y}dy>\frac{2\pi}{5} $$ for some $\delta_2>0$ and all $x\in(0,\delta_2)$. Thus for all $x\in(0,\min(\delta_1,\delta_2))$ we see that $$ f_\alpha(x)=\frac{x^\alpha}{2\sin(x/2)}\int\limits_{x/2}^\infty\frac{\sin y}{y^\alpha}dy+\frac{x\varphi(x)}{2\sin(x/2)}>\frac{2\pi}{5}-\frac{\pi}{3}>0 $$

In both cases $\lim\limits_{x\to+0}f_\alpha(x)\neq 0$. hence $f_\alpha$ is discontinuous at $0$.

Corollary 4. For $\alpha\in(0,1]$ the series $$ \sum\limits_{k=1}^\infty\frac{\sin kx}{k^\alpha} $$ doesn't converges uniformly on $[0,2\pi]$.

Proof. Assume that this series converges uniformly. Since this series is the sum of continuous functions and its converges uniformly, then its sum $f_\alpha$ must be continuous function on $[0,2\pi]$. This contradicts corallary 3, hence our series is not uniformly convergent on $[0,2\pi]$.

Remark 5. Despite the above, this series converges uniformly on $[\delta,2\pi-\delta]$ for all $\delta\in(0,\pi]$. You can use Dirichlet test to prove this.

share|improve this answer
    
Great, your proof is beautiful! Thanks! I have still one more question: what about the continuity of the limit function (for $\alpha \in (0,1]$)? –  Romeo Jun 8 '12 at 6:56
    
@Romeo, see edits to my answer. –  Norbert Jun 8 '12 at 12:18
    
Your proof doesn't work when $\alpha=1$: the RHS of Lemma 2 remains bounded. –  Generic Human Jun 8 '12 at 19:22
    
@GenericHuman, I've edit my answer. Now it covers both cases. –  Norbert Jun 8 '12 at 21:03

Note that $$\sum_{k=1}^{N} \sin(kx) = \dfrac{\sin(Nx/2)}{\sin(x/2)} \sin \left( \left(\dfrac{N+1}2 \right)x\right)$$ Hence, for each given $x$, the sum is bounded by $\dfrac1{\sin(x/2)}$.

Hence by generalized alternating series test (also known as Dirichlet's test) the sum converges.

share|improve this answer
    
I thank you all for your kind replies. The fact that the bound depends on $x$ suggests me that the convergence is not uniform when $\alpha \in (0,1]$... Am I right? How can I prove this formally? –  Romeo Jun 7 '12 at 20:23

There is an easier proof that $f_\alpha$ is discontinuous at 0.

Let $x=\pi/n$ for some even $n$. Then for $1\le i\le n$, group terms $a_{2nk+i}$ and $a_{2nk+i+n}$ together: $$\frac{\sin (2nk+i)x}{(2nk+i)^\alpha}+\frac{\sin (2nk+i+n)x}{(2nk+i+n)^\alpha}\ge\frac{n\alpha}{(2nk+i)(2nk+i+n)^\alpha}\sin \frac{i\pi}{n} \ge 0$$ We need only use $k=0$ and $i\le n/2$ for the lower bound: $$\begin{align} f_\alpha(x)\ge&\alpha \sum_{i=1}^{n/2} \frac{n}{i(i+n)^\alpha}\sin \frac{i\pi}{n}\\ \ge&\alpha\sum_{i=1}^{n/2} \frac{2}{(i+n)^\alpha}\qquad{\text{($\sin t\ge 2/\pi\cdot t$)}}\\ \ge&2\alpha\log\frac{3n/2+1}{n+1}\\ \rightarrow&2\alpha\log 3/2 \end{align}$$ So $f_\alpha$ is bounded from below by a positive number as $x\rightarrow 0$, for all $\alpha\in(0,1]$. Because $f_\alpha(0)=0$, the convergence cannot be uniform around 0.

share|improve this answer
    
Right, I was bounding $\cos t$ instead of $\sin t$. Thanks! Note that actually $f_1(x)=\frac{\pi-x}{2}$. –  Generic Human Jun 8 '12 at 20:52
    
It seems that your proof shows that $\lim\limits_{x\to+0} f_1(x)=+\infty$. But it is known that $f_1(x)=(\pi-x)/2$ for $x\in(0,\pi)$. –  Norbert Jun 8 '12 at 21:04
    
Ok, now no question left :) –  Norbert Jun 8 '12 at 21:06

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

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