Infinite Addition

Big Idea

We can approximate any continuous T-periodic function (that is differentiable on (a, a+T) for some a \in \mathbb R) using its Fourier series given by

\displaystyle f(t) = \frac{a_0}{2} + \sum_{n=1}^\infty a_n \cos\left( \frac{2\pi n t}{T} \right) + b_n \cos\left( \frac{2\pi n t}{T} \right),

where

\displaystyle a_n = \frac{2}{T} \int_{-T/2}^{T/2} f(t)\cos\left( \frac{2\pi n t}{T} \right)\, \mathrm dt,\quad \displaystyle b_n = \frac{2}{T} \int_{-T/2}^{T/2} f(t)\sin\left( \frac{2\pi n t}{T} \right)\, \mathrm dt.

To simplify computations, it helps to exploit the symmetric properties of odd and even functions. Furthermore, we can use Fourier series to compute infinite sums.

A derivation of the Fourier series (and even a proof that it converges) will require us to venture into abstract linear algebra and even advanced real analysis, far beyond the scope of pre-university mathematics.

Questions

Question 1. Given f(t) = |t| for -2 \leq t \leq 2 with periodic extension f(t) = f(t+4), evaluate the Fourier series of f(t).

(Click for Solution)

Solution. We contextualise the Fourier series to T = 4:

\displaystyle f(t) = \frac{a_0}{2} + \sum_{n=1}^\infty \left(a_n \cos \left( \frac{\pi n t}{2} \right) + b_n \sin \left( \frac{\pi n t}{2} \right) \right),

where

\displaystyle a_n = \frac 12 \int_{-2}^2 f(t) \cos \left( \frac{\pi n t}{2} \right)\, \mathrm dt,\quad b_n = \frac 12 \int_{-2}^2 f(t) \sin \left( \frac{\pi n t}{2} \right)\, \mathrm dt.

We observe that f(t) = |t| = |{-t}| = f(-t) so that f is even on [-2, 2]. Since f(\cdot) \sin(\cdot) is odd on [-2, 2], b_n = 0. For the even term, since f(\cdot) \cos(\cdot) is even on [-2 ,2],

\displaystyle a_n = \frac 12 \int_{-2}^2 f(t) \cos \left( \frac{\pi n t}{2} \right)\, \mathrm dt = \int_0^2 t \cos \left( \frac{\pi n t}{2} \right)\, \mathrm dt.

For the case n = 0,

\displaystyle a_0 = \int_0^2 t \cos (0)\, \mathrm dt = \int_0^2 t\, \mathrm dt = \frac 12 \cdot 2 \cdot 2 = 2.

For the case n \neq 0, we first make the substitutions

\displaystyle u = (\pi n t)/2 \quad \iff \quad \mathrm dt = \frac 2{\pi n}\, \mathrm du

so that u(2) = n \pi and

\displaystyle \begin{aligned} a_n &= \int_0^2 t  \cos \left( \frac{\pi n t}{2} \right)\, \mathrm dt  \\ &= \frac 2{n \pi} \int_0^{n \pi} u  \cos (u) \cdot \frac 2{\pi n}\, \mathrm du \\ &= \frac 4{ n^2 \pi^2 } \int_0^{n \pi} u  \cos (u) \, \mathrm du. \end{aligned}

Integrating by parts,

\displaystyle \begin{aligned} \int u \cos(u)\, \mathrm du &= \underbrace{\sin(u)}_{\mathrm I} \underbrace{u}_{\mathrm S} - \int \underbrace{\sin(u)}_{\mathrm I} \underbrace{1}_{\mathrm D}\, \mathrm du \\ &= u \sin(u) - \int \sin(u)\, \mathrm du\\ &= u \sin(u) - (-\cos(u)) + C \\ &= u \sin(u) + \cos(u) + C. \end{aligned}

Therefore,

\begin{aligned} a_n &= \frac 4{ n^2 \pi^2 } \left[u \sin(u) + \cos(u) \right]_0^{n \pi} \\ &= \frac 4{ n^2 \pi^2 } ((0 + \cos(n \pi)) - (0 + 1)) \\ &= \frac 4{ n^2 \pi^2 } ( (-1)^n - 1) \\ &= \begin{cases} -\frac{8}{ n^2 \pi^2 }, & n\ \text{odd}, \\ 0, & n\ \text{even}. \end{cases} \end{aligned}

Substituting the values of a_n and b_n, the Fourier series of f(t) is given by

\displaystyle f(t) = 1 - \frac {8}{\pi^2} \sum_{k=1}^\infty \frac{1}{(2k-1)^2} \cos \left( \frac{\pi (2k-1) t}{2} \right).

Question 2. Given f(t) = |{ \sin(t) }|, use the Fourier series of f(t) to evaluate exactly the infinite sum

\displaystyle \frac 1{2^2 - 1} + \frac 1{4^2 - 1} + \frac 1{6^2 - 1} + \cdots.

(Click for Solution)

Solution. We contextualise the Fourier series to T = \pi:

\displaystyle f(t) = \frac{a_0}{2} + \sum_{n=1}^\infty \left(a_n \cos (2n t) + b_n \sin (2n t) \right),

where

\displaystyle a_n = \frac 2{\pi} \int_{-\pi/2}^{\pi/2} f(t) \cos (2nt)\, \mathrm dt,\quad b_n = \frac 2{\pi} \int_{-\pi/2}^{\pi/2} f(t) \sin (2nt)\, \mathrm dt.

We observe that f(t) = |{\sin(t)}| = |{\sin(-t)}| = f(-t) so that f is even on [-2, 2]. Since f(\cdot) \sin(\cdot) is odd on [-2, 2], b_n = 0. For the even term, since f(\cdot) \cos(\cdot) is even on [-2 ,2],

\displaystyle a_n =  \frac 2{\pi} \int_{-\pi/2}^{\pi/2} f(t) \cos (2nt)\, \mathrm dt =  \frac 4{\pi} \int_0^{\pi/2} f(t) \cos (2nt)\, \mathrm dt.

For the case n = 0,

\displaystyle a_0 = \frac 4{\pi} \int_0^{\pi/2} \sin(t)\, \mathrm dt = \frac 4{\pi} \left[-{\cos(t)}\right]_0^{\pi/2} = \frac 4{\pi}(0 - (-1)) = \frac 4{\pi}.

For the case n \neq 0, we integrate by parts to get

\displaystyle \begin{aligned} & \int  \sin(t) \cos (2nt)\, \mathrm dt \\ &= \underbrace{-\cos(t)}_{\mathrm I} \underbrace{\cos (2nt)}_{\mathrm S} - \int \underbrace{-\cos(t)}_{\mathrm I} \underbrace{-2n \sin (2nt)}_{\mathrm D}\, \mathrm dt \\ &= -\cos(t) \cos(2nt) - 2n \int \cos(t) \sin(2nt)\, \mathrm dt \\ &= -\cos(t) \cos(2nt) - 2n \left( \underbrace{\sin(t)}_{\mathrm I} \underbrace{\sin (2nt)}_{\mathrm S} - \int \underbrace{\sin(t)}_{\mathrm I} \underbrace{2n \cos (2nt)}_{\mathrm D}\, \mathrm dt \right) \\  &= -\cos(t) \cos(2nt) -2n \sin(t) \sin(2nt) + 4n^2 \int  \sin(t) \cos (2nt)\, \mathrm dt. \end{aligned}

By algebruh,

\displaystyle \int  \sin(t) \cos (2nt)\, \mathrm dt = \frac{ \cos(t) \cos(2nt) + 2n \sin(t) \sin(2nt)}{4n^2 - 1} + C

Therefore,

\begin{aligned} a_n &= \frac 4{\pi} \left[\frac{ \cos(t) \cos(2nt) + 2n \sin(t) \sin(2nt)}{4n^2 - 1} \right]_0^{\pi/2} \\ &= \frac 4{\pi} \left( 0 - \frac 1{4n^2 - 1}\right) = -\frac{4}{\pi(4n^2 - 1)}. \end{aligned}

Substituting the values of a_n and b_n, the Fourier series of f(t) is given by

\displaystyle f(t) = \frac 2{\pi} - \frac{4}{\pi} \sum_{k=1}^\infty \frac{1}{4n^2 - 1} \cos (2nt).

Setting t = 0 on both sides, since f(0) = 0,

\displaystyle 0 =\frac 2{\pi} - \frac{4}{\pi} \sum_{k=1}^\infty \frac{1}{4n^2 - 1}.

By algebruh,

\displaystyle \frac 1{2^2 - 1} + \frac 1{4^2 - 1} + \frac 1{6^2 - 1} + \cdots = \sum_{k=1}^\infty \frac{1}{4n^2 - 1} = \frac 12 .

Exercise. Try to obtain the same result without the use of Fourier series using a telescoping sum.

—Joel Kindiak, 19 Apr 25, 2240H

,

Published by


Leave a comment