Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

I tried to prove that $$(1-2x)^2=1/3+4/\pi^2\sum_1^\infty \cos(2n x \pi)/n^2$$ for $x \in [0,1)$ with Fourier analysis, but I just found a Fourier series which defines the function. I also found the fourier series of $\cos(2n x \pi)$.

I don't think these results are helpful.

Any suggestions on how to prove this equation?

share|cite|improve this question
I don't know how to do it... – tomerg Jan 7 '12 at 8:26
Do you know about cosine Fourier series? – Hans Lundmark Jan 7 '12 at 8:30
Briefly (......) – tomerg Jan 7 '12 at 8:36
@tomerg: The most important point is to put the mathematical equation in dollar signs,'$'. – Gigili Jan 7 '12 at 9:11
OK............. – tomerg Jan 7 '12 at 9:15
up vote 1 down vote accepted


(Changed since the previous was wrong -- the function is not even in it self..)

  1. Note first that $x\mapsto e(x)=(1-2x)^2$ is even on $[0,1]$ in the sense $e(x)= e(1-x)$ (either visualise it or by computation $(1-2(1-x))^2 = (1-2+2x)^2 = (-1+2x)^2)$).

  2. Now, let us look at the function $s$ on $I=[-1,1]$, which is the even extension of the function $x\mapsto (1-2x)^2$ on $[0,1]$ (so the graph looks a bit like $\omega$).

  3. Note that the sine Fourier coefficients are 0, while the cosine Fourier coefficients are given by $$a_n = 2\int_0^1s(x)\cos(n\pi x)dx.$$

  4. Next $a_n=0$ for odd $n$, this follows from 1 (try to see why without a calculation) or a simple calculation, and for even $n\ne0$ we have $$a_n= \frac{16}{n^2\pi^2}$$ while $a_0=2/3$.

  5. Why is $s$ equal to the Fourier series on $[-1,1]$?

  6. What is $s$ equal to in $[0,1]$?

Last edit

For even $n$ we have $n=2k$ for some $k$ and that $a_0$ should be divided by 2 in the expansion.

share|cite|improve this answer
I found the fourier series with respect to [-1,1] by using step 1 but it does not fit the series in the question... – tomerg Jan 7 '12 at 10:43
You are right, I will add more appropriate hints! – AD. Jan 7 '12 at 11:33
This is not quite right, since $s(-x)=(1+2x)^2 \neq s(x)$... – Hans Lundmark Jan 7 '12 at 11:58
@HansLundmark Yes, I changed it accordingly.. I did not visualise the function correctly... – AD. Jan 7 '12 at 12:03
what's an even extension? can you write function s? – tomerg Jan 7 '12 at 18:30

The series on the right hand has terms of the form $a_n\cos(2\pi n x)$ with $a_n\ne0$ for all $n$. This indicates a fundamental period of length $L=1$. When the sum of such a series is claimed to be equal to the function $f\!: x\mapsto (1-2x)^2$ on the interval $[0,1]$ then it has to be the Fourier series of the function $\tilde f$ obtained from $f$ by periodic extension with period $1$. As $\tilde f$ is even (check it!) its Fourier series contains only $\cos$-terms $a_n\cos(2\pi n x)$ to begin with. The $a_n$ are given by the formula $$ a_n={2\over L}\int_0^L(1-2x)^2\ \cos(2\pi n x)\ dx = 2\int_0^1(1-2x)^2\ \cos(2\pi n x)\ dx\ .$$ For $n\geq 1$ the integral can be evaluated by partial integration (2 times); the case $n=0$ is immediate. Doing the calculations you will see that you get exactly the $a_n$ in the title.

So we know now that the formal Fourier series of $\tilde f$ is actually the series in the title. In order to finish the case we have to invoke a fundamental theorem about such series: When $\tilde f$ is an $L$-periodic continuous function of bounded variation per period then its formal Fourier series converges uniformly on ${\mathbb R}$ to $\tilde f(x)$. Since the assumptions of the theorem are obviously satisfied in our case, the claim follows.

share|cite|improve this answer
My problem is verifying that the fourier series are what written in the title. – tomerg Jan 7 '12 at 11:04

Consider the equivalent problem using $y=x-\frac12$ on the interval $[-\frac12,\frac12]$: Prove that $$ 4y^2=\frac13+\frac{4}{\pi^2}\sum_{n=1}^\infty(-1)^n\frac{\cos(2\pi ny)}{n^2}\tag{1} $$ Since $\pi\csc(\pi z)$ has residue $(-1)^n$ at each integer, let's consider $f_y(z)=\pi\csc(\pi z)\frac{\cos(2\pi zy)}{z^2}$.

Let $\gamma_N$ be the rectangular path $$ (N{+}\!\tfrac12)-iN\to(N{+}\!\tfrac12)+iN\to-(N{+}\!\tfrac12)+iN\to-(N{+}\!\tfrac12)-iN\to(N{+}\!\tfrac12)-iN $$ It is not hard to see that for $|y|<\frac12$, $$ \lim_{N\to\infty}\;\;\oint_{\gamma_N}f_y(z)\;\mathrm{d}z=0\tag{2} $$ Equation $(2)$ relates the residue of $f_y$ at $0$ with the sum in $(1)$: $$ \operatorname{Res}(f_y,0)+ 2\sum_{n=1}^\infty(-1)^n\frac{\cos(2n\pi y)}{n^2}=0\tag{3} $$ Let's look at the Laurent series for $f_y$: $$ \begin{align} \pi\csc(\pi z)\frac{\cos(2\pi zy)}{z^2} &=\left(\frac1z+\frac{\pi^2}{6}z+\dots\right)\left(\frac{1}{z^2}-2\pi^2y^2+\dots\right)\\ &=\frac{1}{z^3}+\left(\frac{\pi^2}{6}-2\pi^2y^2\right)\frac{1}{z}+\dots\tag{4} \end{align} $$ Equation $(4)$ says that $\operatorname{Res}(f_y,0)=\frac{\pi^2}{6}-2\pi^2y^2$. Combining this with $(3)$ yields $$ \sum_{n=1}^\infty(-1)^n\frac{\cos(2n\pi y)}{n^2}=\pi^2\left(y^2-\frac{1}{12}\right)\tag{5} $$ which immediately verfies $(1)$.

share|cite|improve this answer

Your Answer


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.