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 $f: \mathbb{R} \rightarrow \mathbb{R}$ be $2\pi$-periodic and integrable on $[-\pi, \pi]$. Assume that $f(x)\sim \frac{a_0}{2}+\sum_{n=1}^\infty (a_n \cos nx+b_n \sin nx )$ is its Fourier series.

How to prove that $$ \int_0^x \ldots \int_0^x f(x) dx\ldots dx=\int_0^x \ldots \int_0^x \frac{a_0}{2} dx\ldots dx+ \atop \sum_{n=1}^\infty \int_0^x \ldots \int_0^x (a_n \cos nx+ b_n \sin nx) dx \ldots dx, $$ where in both sides of equality are $m$-fold integrals.

For $m=1$ the proof is following. Let's consider the function $F(x)=\int_0^x (f(t)-\frac{a_0}{2})dt$ for $x\in \mathbb{R}$, where $a_0=\frac{1}{\pi} \int_{-\pi}^\pi f(t)dt$. Then $F$ is $2\pi$-periodic and absolutely continuous. Assume that $F(x) \sim \frac{A_0}{2}+\sum_{n=1}^\infty (A_n \cos nx+B_n \sin nx)$. Then by Dirichlet-Jordan theorem

$ (*) \ \ F(x) = \frac{A_0}{2}+\sum_{n=1}^\infty (A_n \cos nx+B_n \sin nx) \ \ for \ x\in \mathbb{R}. $ (even uniformly)

By definition of Fourier coefficients and formula of integration by parts we find $A_n=-\frac{b_n}{n}$, $B_n=\frac{a_n}{n}$ for $n \in \mathbb{N}$. Taking in $(*)$ $x=0$ we obtain $\frac{A_0}{2}=-\sum_{n=1}^\infty A_n=\sum_{n=1}^\infty \frac{b_n}{n}$. By $(*)$ we have now $\int_0^x f(t(dt=\int_0^x \frac{a_0}{2}dt +\sum_{n=1}^\infty \int_0^x (a_n \cos nt +b_n \sin nt)dt$.



By $m-fold$ integral $\int_0^x... \int_0^x f(x) dx...dx$ I mean

$\int_0^x... \int_0^x f(x) dx...dx:=\int_0^x[ \int_0^{x_m}...(\int_0^{x_3}(\int_0^{x_2} f(x_1) dx_1) dx_2)...dx_{m-1}]dx_m$.

It is equal, by Cauchy's formula,

$\int_0^x... \int_0^x f(x) dx...dx:=\frac{1}{(m-1)!} \int_0^x f(t) (x-t)^{m-1} dt$.

share|improve this question

1 Answer 1

up vote 1 down vote accepted

If $S_n(t)$ is the $n$'th partial sum of the Fourier series for $f(t)$, you want to show that for any $x \in [0, 2 \pi]$, $\int_0^{2\pi} g(t) S_n(t) \ dt \to \int_0^{2\pi} g(t) f(t) \ dt$ as $n \to \infty$, where $g(t) = (x-t)^{m-1}/(m-1)!$ for $0 < t < x$, 0 otherwise. Now if $T_n(t)$ is the $n$'th partial sum of the Fourier series for $g(t)$, $\int_0^{2 \pi} g(t) S_n(t) = \int_0^{2 \pi} T_n(t) f(t)\ dt$. $g(t)$ is a very nice function, smooth except for jumps at 0 (when identified with $2 \pi$) and $x$, so the partial sums of its Fourier series are bounded and converge to $g(t)$ away from the jumps. Break up $[0,2\pi]$ into small intervals near $0$, $x$ and $2 \pi$, small enough so the integral of $f$ over those intervals is small, and large intervals on which $T_n - g$ is uniformly small...

share|improve this answer
Thanks for answer, but there is a thing I dont understand. Does it suffice to show that $\int_0^{2\pi} g(t) S_n(t)dt \rightarrow \int_0^{2\pi} g(t)f(t)dt$ as $n\rightarrow \infty$ for $x \in [0,2 \pi$ ? The function $\int_0^x... \int_0^x f(x) dx...dx=\int_0^{2\pi} g(t)f(t)dt$ is generally not $2\pi$-periodic. –  Richard Oct 28 '11 at 6:34
If $x = 2 k \pi + s$ for positive integer $k$, do the change of variables $u = t - 2 j \pi$ on each interval $t \in [2 j \pi, 2 (j+1)\pi]$ for $j = 0, 1, \ldots k$. You'll get a more complicated $g(u)$, but still smooth except for jumps at 0 and $s$. –  Robert Israel Oct 28 '11 at 6:59
Thanks for answer. –  Richard Oct 28 '11 at 7:05

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.