45

Here is an exercise, on analysis which i am stuck.

  • How do I prove that if $F_{n}(x)=\sum\limits_{k=1}^{n} \frac{\sin{kx}}{k}$, then the sequence $\{F_{n}(x)\}$ is boundedly convergent on $\mathbb{R}$?
Arturo Magidin
  • 417,286

3 Answers3

62

First, let's note that for $x\in(0,2\pi)$ $$\sum\limits_{k=1}^{n}\frac{\sin kx}{k}=\int_{0}^{x}\sum\limits_{k=1}^{n}\cos kt\ dt= -\frac{x}{2}+\int_{0}^{x}\frac{\sin \frac{(2n+1)t}{2}}{2\sin{\frac{t}{2}}}\ dt$$ $$=-\frac{x}{2}+\int_{0}^{x}\left(\frac{1}{2\sin{\frac{t}{2}}}-\frac{1}{t}\right)\sin \frac{(2n+1)t}{2}\ dt +\int_{0}^{x}\frac{\sin \frac{(2n+1)t}{2}}{t}dt.$$ Now, the first integral at the right-hand side tends to zero by the Riemann-Lebesgue lemma. The second one is equal to (via a substitution $s=(2n+1)t/2$) the integral $$\int_{0}^{\frac{(2n+1)x}{2}}\frac{\sin s}{s}\ ds\to\int_{0}^{\infty}\frac{\sin s}{s}\ ds=\frac{\pi}{2}.$$ Therefore $$\lim\limits_{n\to\infty}\ \sum\limits_{k=1}^{n}\frac{\sin kx}{k}=\frac{\pi-x}{2}=f(x),\qquad x\in(0,2\pi).$$ The series converges on $\mathbb R$ to the periodic extension of $f(x)$.

  • if you take $x=0$ then the left side is a summation of zeros and the right isn't zero. – Ofir Dec 08 '10 at 06:51
  • The argument works for $x\neq 2\pi m$. – Andrey Rekalo Dec 08 '10 at 07:09
  • 1
    @Prometheus: At that point the series converges to $(\lim_{x\to0,x>0}f(x)+\lim_{x\to0,x<0}f(x))/2=$ $(\lim_{x\to0,x>0}f(x)+\lim_{x\to2\pi,x<2\pi}f(x))/2 =$ $(\frac{\pi}{2}+\frac{\pi-2\pi}{2})/2=0$. – AD - Stop Putin - Dec 08 '10 at 11:41
  • 1
    @Prometheus Read about Dirichlet's theorems on Fourier series. – Pedro Aug 16 '12 at 21:52
  • 1
    @AndreyRekalo can you explain more why that the first integral at the right-hand side tends to zero by the Riemann-Lebesgue lemma ? I read the article at wiki but i faild to connect between the integral and law of Riemann–Lebesgue lemma – mnsh Aug 15 '13 at 08:01
  • You went a little above and beyond – Simply Beautiful Art Nov 26 '16 at 15:39
3

For any $n\geq 1$, we know where the stationary points of $F_n(x)$ occur since $F_n'(x)$ has a simple closed form.
It follows that $$ \sup_{x\in\mathbb{R}} |F_n(x)| = \sum_{k=1}^{n}\frac{1}{k}\sin\left(\frac{2\pi k}{2n+1}\right) $$ and over $[0,\pi]$ we have $\sin(x)\leq \frac{4}{\pi^2}x(\pi-x)$ by concavity, therefore $$ \sup_{x\in\mathbb{R}} |F_n(x)| \leq \frac{8n^2}{(2n+1)^2} <2.$$ We may also prove that the sequence $$ A_n = \sum_{k=1}^{n}\frac{1}{k}\sin\left(\frac{2\pi k}{2n+1}\right) $$ is increasing and convergent to $$ \int_{0}^{\pi}\frac{\sin x}{x}\,dx = \text{Si}(\pi) \approx 1.85194. $$ Ultimately we may check that $\sum_{k\geq 1}\frac{\sin(kx)}{k}$ is the Fourier series of the sawtooth wave, i.e. the $2\pi$-periodic extension of $\frac{\pi-x}{2}$ defined over $(0,2\pi)$. This is enough to ensure convergence in $L^2$, plus we have a uniform bound for $|F_n(x)|$.

Jack D'Aurizio
  • 361,689
  • Your $$\sup_{x\in\mathbb{R}} |F_n(x)| = \sum_{k=1}^{n}\frac{1}{k}\sin\left(\frac{2\pi k}{2n+1}\right)$$ seems not ture: for $n=1$, $F_1(x)=\sin x$, the left is $1$,but the right is $\sin\frac{2\pi}{3}=\frac{\sqrt{3}}{2}<1$. The right hand side maybe $$\sum_{k=1}^n\frac{1}{k}\sin\left(\frac{k\pi}{n+1}\right).$$ – Riemann Jul 08 '22 at 01:45
0

Here is an intuitive and easy-to-think method: $$ \begin{aligned} f(x):=\sum_{n=1}^{+\infty} \frac{\sin n x}{n} & =\sum_{n=1}^{+\infty} \frac{\exp (\mathrm{i} n x)-\exp (-\mathrm{i} n x)}{2 \mathrm{i} n} \\ & =\frac{1}{2 \mathrm{i}} \sum_{n=1}^{+\infty} \frac{\left(\mathrm{e}^{\mathrm{i} x}\right)^n-\left(\mathrm{e}^{-\mathrm{i} x}\right)^n}{n} \\ & =\frac{1}{2 \mathrm{i}}\left(-\ln \left(1-\mathrm{e}^{\mathrm{i} x}\right)+\ln \left(1-\mathrm{e}^{-\mathrm{i} x}\right)\right) \\ & =\frac{1}{2 \mathrm{i}} \ln \frac{1-\mathrm{e}^{-\mathrm{i} x}}{1-\mathrm{e}^{\mathrm{i} x}} \\ & =\frac{1}{2 \mathrm{i}} \ln \left(-\mathrm{e}^{-\mathrm{i} x}\right)=\frac{1}{2 \mathrm{i}} \ln \left(\mathrm{e}^{\mathrm{i}(\pi-x)}\right) \\ & =\frac{\pi-x}{2} \end{aligned} $$ The last step here should involve the knowledge of the multi-valued property of complex functions. I may not be strict in writing it this way, but I can get the correct result. One of my imagined solutions is to determine its single value based on $f(\pi/2)=1-1/3+1/5+\cdots=\pi/4$.

Soriak
  • 299
  • f(0) should be zero, but in your solution it is not. Also I think you need some justification in replacing a sum with two sums when you introduce logarithms - this can only be done if we already know the series is convergent. – lisyarus Jun 25 '24 at 06:22
  • @lisyaru, $f(0)$ cannot be calculated by $f(x)=(\pi/2-x)/2$ because in my 3rd line of caculation, $x+x^2/2+x^3/3+\cdots=-\ln(1-x)$ when $|x|<1$, but when $x=0$, $exp(\mathrm ix)=1$. – Soriak Jun 25 '24 at 06:35