Chapter 1. The Genesis of Fourier Analysis

Exercises

1. If z=x+iyz = x + iy is a complex number with x,yRx, y \in \mathbb{R}, we define

z=(x2+y2)1/2|z| = (x^2 + y^2)^{1/2}

and call this quantity the modulus or absolute value of zz.
(a)\quad \text{(a)} What is the geometric interpretation of z|z|?
Solution
The geometric interpretation of z|z| is the length of the vector zz.


(b)\quad \text{(b)} Show that if z=0|z| = 0, then z=0z = 0.
Solution
If z=0|z| = 0 then by definition 0=(x2+y2)1/2    0=x2+y2    x=y=00 = (x^2 + y^2)^{1/2} \implies 0 = x^2 + y^2 \implies x = y = 0. \square


(c)\quad \text{(c)} Show that if λR\lambda \in \R, then λz=λz|\lambda z| = |\lambda||z|, where λ|\lambda| denotes the standard absolute value of a real number.
Solution
Calculate that λz=λx+iλy\lambda z = \lambda x + i \lambda y, so λz=((λx)2+(λy)2)1/2=λz|\lambda z| = ((\lambda x)^2 + (\lambda y)^2 )^{1/2} = |\lambda||z|. \square


(d)\quad \text{(d)} If z1z_1 and z2z_2 are two complex numbers, prove that

z1z2=z1z2andz1+z2z1+z2.|z_1z_2| = |z_1||z_2| \quad \text{and} \quad |z_1 + z_2| \leq |z_1| + |z_2|.

Solution
Let z1=x1+iy1z_1 = x_1 + iy_1 and z2=x2+iy2z_2 = x_2 + iy_2. Calculate that z1z2=(x1+iy1)(x2+iy2)=x1x2+ix1y2+iy1x2y1y2=(x1x2y1y2)+i(x1y2+y1x2)z_1 \cdot z_2 = (x_1 + iy_1) \cdot (x_2 + iy_2) = x_1x_2 + ix_1y_2 + iy_1x_2 - y_1y_2 = (x_1x_2 - y_1y_2) + i(x_1y_2 + y_1x_2). So z1z22=(x1x2y1y2)+i(x1y2+y1x2)2=(x1x2y1y2)2+(x1y2+y1x2)2=x12x222x1x2y1y2+y12y22+x12y22+2x1x2y1y2+y12x22=x12x22+y12y22+x12y22+y12x22|z_1z_2|^2 = |(x_1x_2 - y_1y_2) + i(x_1y_2 + y_1x_2)|^2 = (x_1x_2-y_1y_2)^2 + (x_1y_2 + y_1x_2)^2 = x_1^2x_2^2 - 2x_1x_2y_1y_2 + y_1^2y_2^2 + x_1^2y_2^2 + 2x_1x_2y_1y_2 + y_1^2x_2^2 = x_1^2x_2^2 + y_1^2y_2^2 + x_1^2y_2^2 + y_1^2x_2^2.
Calculate separately that (z1z2)2=((x12+y12)1/2(x22+y22)1/2)2=(x12+y12)(x22+y22)=x12x22+x12y22+y12x22+y12y22(|z_1||z_2|)^2 = ((x_1^2 + y_1^2)^{1/2} \cdot (x_2^2+y_2^2)^{1/2})^2 = (x_1^2 + y_1^2)(x_2^2+y_2^2) = x_1^2x_2^2 + x_1^2y_2^2 + y_1^2x_2^2 + y_1^2y_2^2.
Therefore z1z22=(z1z2)2|z_1z_2|^2 = (|z_1||z_2|)^2. Since both z1z2|z_1z_2| and z1z2|z_1||z_2| are non-negative we can conclude thatz1z2=z1z2|z_1z_2| = |z_1||z_2|.
For the inequality we need to show that

z1+z2z1+z2    (x1+x2)+i(y1+y2)x1+iy1+x2+iy2    (x1+x2)2+(y1+y2)2x12+y12+x22+y22    (x1+x2)2+(y1+y2)2(x12+y12)+2(x12+y12)(x22+y22)+(x22+y22)    2x1x2+2y1y22(x12+y12)(x22+y22)    x1x2+y1y2(x12+y12)(x22+y22)    x12x22+2x1x2y1y2+y12y22(x12+y12)(x22+y22)    x12x22+2x1x2y1y2+y12y22x12x22+x12y22+y12x22+y12y22    2x1x2y1y2x12y22+y12x22    0x12y222x1x2y1y2+y12x22    0(x1y2y1x2)2|z_1 + z_2| \leq |z_1| + |z_2| \\ \iff |(x_1 + x_2) + i(y_1+y_2)| \leq |x_1 + iy_1| + |x_2 + iy_2| \\ \iff \sqrt{(x_1 + x_2)^2 + (y_1 + y_2)^2 } \leq \sqrt{x_1 ^ 2 + y_1^2} + \sqrt{x_2^2 + y_2^2} \\ \iff (x_1 + x_2)^2 + (y_1+y_2)^2 \leq (x_1^2 + y_1^2) + 2\sqrt{(x_1^2+y_1^2)(x_2^2+y_2^2)} + (x_2^2 + y_2^2) \\ \iff 2x_1x_2 + 2y_1y_2 \leq 2\sqrt{(x_1^2+y_1^2)(x_2^2+y_2^2)} \\ \iff x_1x_2 + y_1y_2 \leq \sqrt{(x_1^2+y_1^2)(x_2^2+y_2^2)} \\ \iff x_1^2x_2^2 + 2x_1x_2y_1y_2 +y_1^2y_2^2 \leq (x_1^2+y_1^2)(x_2^2+y_2^2) \\ \iff x_1^2x_2^2 + 2x_1x_2y_1y_2 +y_1^2y_2^2 \leq x_1^2x_2^2 + x_1^2y_2^2 + y_1^2x_2^2 + y_1^2y_2^2 \\ \iff 2x_1x_2y_1y_2 \leq x_1^2y_2^2 + y_1^2x_2^2 \\ \iff 0 \leq x_1^2y_2^2 - 2x_1x_2y_1y_2 + y_1^2x_2^2\\ \iff 0 \leq (x_1y_2 - y_1x_2)^2 \\ \square


(e)\quad \text{(e)} Show that if z0z \neq 0, then 1/z=1/z|1/z| = 1/|z|.
Solution
Using the previous result

1/zz=1/zz=1=1|1/z||z| = |1/z \cdot z| = |1| = 1

Dividing both sides by z|z| (which we know isn't zero thanks to (b)\text{(b)} ) gives the result. \square


2. If z=x+iyz = x + iy is a complex number with x,yRx, y \in \R, we define the complex conjugate of zz by

z=xiy.\overline{z} = x - iy.

(a)\quad \text{(a)} What is the geometric interpretation of z\overline{z}?
Solution
The geometric interpetation of z\overline{z} is a reflection of the vector zz over the xx-axis.


(b)\quad \text{(b)} Show that z2=zz|z|^2 = z\overline{z}.
Solution

zz=(x+iy)(xiy)=x2+y2=z2z\overline{z} = (x + iy)(x - iy) = x^2 + y^2 = |z|^2 \quad \square


(c)\quad \text{(c)} Prove that if zz belongs to the unit circle, then 1/z=z1/z = \overline{z}.
Solution
If zz belongs to the unit circle that means that z=1|z| = 1. Using part (b)\text{(b)} we get that zz=z2=1z\overline{z} = |z|^2 = 1. Diving both sides by zz gives the result. \square


3. A sequence of complex numbers {wn}n=1\{w_n\}_{n=1}^{\infty} is said to converge if there exists wCw \in \mathbb{C} such that

limnwnw=0,\lim_{n \to \infty} |w_n - w| = 0,

and we say that ww is a limit of the sequence.
(a)\quad \text{(a)} Show that a converging sequence of complex numbers has a unique limit.
Solution
Let x,yx, y be limits of {wn}n=1\{w_n\}_{n=1}^\infty

0=limnwnx=limnwny0 = \lim_{n \to \infty}|w_n - x| = \lim_{n \to \infty}|w_n - y| \\

So

xy=limnxy=limnxwn+wnylimnxwn+wny(using 1 (d))=limnwnx+wny=limnwnx+limnwny=0+0=0\begin{aligned} |x - y| &= \lim_{n \to \infty} |x - y| = \lim_{n \to \infty} |x - w_n + w_n - y| \\ &\leq \lim_{n \to \infty} |x - w_n| + |w_n - y| \quad \text{(using 1 (d))} \\ &= \lim_{n \to \infty} |w_n - x| + |w_n - y| \\ &= \lim_{n \to \infty}|w_n - x| + \lim_{n \to \infty}|w_n - y| \\ &= 0 + 0 = 0 \\ \end{aligned}

This shows that xy0|x - y| \leq 0, we also have xy0|x - y| \geq 0 by the definition of |\cdot| so we have xy=0|x - y| = 0. Using 1 (b)\text{1 (b)} we get xy=0    x=yx - y = 0 \implies x = y so any two limits of a converging sequence of complex numbers are equal, hence this limit is unique. \square


The sequence {wn}n=1\{w_n\}_{n=1}^\infty is said to be a Cauchy sequence if for every ϵ>0\epsilon > 0 there exists a positive integer NN such that

wnwm<ϵwhenever n,m>N.|w_n - w_m| < \epsilon \quad \text{whenever } n, m > N.

(b)\quad \text{(b)} Prove that a sequence of complex numbers converges if and only if it is a Cauchy sequence. [Hint: A similar theorem exists for the convergence of a sequence of real numbers. Why does it carry over to sequences of complex numbers?]
Solution

limnwnw=0    limnxn+iyn(x+iy)=0    limnxnx+i(yn+y)=0    limnxnx=0 and limnyny=0    xnaxnb<ϵ whenever a,b>N1 and ynaynb<ϵ whenever a,b>N2\lim_{n \to \infty} |w_n - w| = 0 \iff \lim_{n \to \infty} |x_n + iy_n - (x + iy)| = 0 \\ \iff \lim_{n \to \infty} |x_n - x + i(y_n + y)| = 0 \\ \iff \lim_{n \to \infty} |x_n - x| = 0 \text{ and } \lim_{n \to \infty} |y_n - y| = 0 \\ \iff |{x_n}_a - {x_n}_b| < \epsilon \quad \text{ whenever } a, b > N_1 \text{ and } \\ |{y_n}_a - {y_n}_b| < \epsilon \quad \text{ whenever } a, b > N_2

wnwm<ϵ    xn+iyn(xm+iym)<ϵ    xnxm+i(ynym)<ϵ    xnxm<ϵ and ynym<ϵ|w_n - w_m| < \epsilon \iff |x_n + iy_n - (x_m + iy_m)| < \epsilon \\ \iff |x_n - x_m + i(y_n - y_m)| < \epsilon \\ \iff |x_n - x_m| < \epsilon \text{ and } |y_n - y_m| < \epsilon \\


A series n=1zn\sum_{n=1}^\infty z_n of complex numbers is said to converge if the sequence formed by the partial sums

SN=n=1NznS_N = \sum_{n=1}^N z_n

converges. Let {an}n=1\{a_n\}_{n=1}^\infty be a sequence of non-negative real numbers such that the series nan\sum_{n}a_n converges.
(c)\quad \text{(c)} Show that if {zn}n=1\{z_n\}_{n=1}^\infty is a sequence of complex numbers satisfying znan|z_n| \leq a_n for all nn, then the series nzn\sum_{n} z_n converges. [Hint: Use the Cauchy criterion.]
Solution
Start with the Cauchy criterion for nan\sum_{n}a_n. For any ϵ>0\epsilon > 0 there exists nn such that for all N,M>KN, M > K we have

ϵ>ANAM=n=1Nanm=1Mam=j=min(N,M)max(N,M)aj\epsilon > |A_N - A_M| = \left|\sum_{n=1}^N a_n - \sum_{m=1}^M a_m \right| = \left|\sum_{j=\text{min}(N, M)}^{\text{max}(N, M)} a_j\right|

Finish with the Cauchy criterion for nzn\sum_{n}z_n

SNSM=n=1Nznm=1Mzm=j=min(N,M)max(N,M)zjj=min(N,M)max(N,M)zjj=min(N,M)max(N,M)ajj=min(N,M)max(N,M)aj<ϵ|S_N - S_M| = \left|\sum_{n=1}^N z_n - \sum_{m=1}^M z_m\right| = \left|\sum_{j=\text{min}(N, M)}^{\text{max}(N, M)} z_j\right| \\ \leq \sum_{j=\text{min}(N, M)}^{\text{max}(N, M)} |z_j| \leq \sum_{j=\text{min}(N, M)}^{\text{max}(N, M)} a_j \leq \left|\sum_{j=\text{min}(N, M)}^{\text{max}(N, M)} a_j\right| < \epsilon \quad \square


4. For zCz \in \Complex, we define the complex exponential by

ez=n=0znn!.e^z = \sum_{n=0}^{\infty} \frac{z^n}{n!}.

(a)\quad \text{(a)} Prove that the above definition makes sense, by showing that the series converges for every complex number zz. Moreover, show that the convergence is uniform on every bounded subset of C\Complex.
Solution
First we use 3. (c)\text{(c)} to show that the series converges. Let zCz \in \Complex, then we have a sequence of non-negative real numbers znn!\frac{|z|^n}{n!} whose sum converges to eze^{|z|}. Since znn!znn!\left|\frac{z^n}{n!}\right| \leq \frac{|z|^n}{n!} we are done.
We go on to show that the convergence is uniform on bounded subsets. Let SCS \subset C be bounded with radius R>0R>0, so for each zSz \in S we have z<R|z| < R. Let ϵ>0\epsilon > 0. Regardless of what zz we choose from SS

n=0Nznn!ez=n=Nznn!n=Nznn!n=Nznn!n=NRnn!\left| \sum_{n=0}^N \frac{z^n}{n!}-e^z \right| = \left| \sum_{n=N}^\infty \frac{z^n}{n!} \right| \leq \sum_{n=N}^\infty \left|\frac{z^n}{n!}\right| \leq \sum_{n=N}^\infty \frac{|z|^n}{n!} \leq \sum_{n=N}^\infty \frac{R^n}{n!}

So choose NN large enough so that we have

ϵ>n=0NRnn!eR=n=NRnn!=n=NRnn!.\epsilon > \left|\sum_{n=0}^N \frac{R^n}{n!} - e^R\right| = \left| \sum_{n=N}^\infty \frac{R^n}{n!} \right| = \sum_{n=N}^\infty \frac{R^n}{n!}.

Note that the above is valid because of the convergence for the exponential series of positive reals. Combining inequalities yields the desired result

n=0Nznn!ez<ϵ.\left| \sum_{n=0}^N \frac{z^n}{n!}-e^z \right| < \epsilon. \quad \square


(b)\quad \text{(b)} If z1,z2z_1, z_2 are two complex numbers, prove that ez1ez2=ez1+z2e^{z_1}e^{z_2} = e^{z_1 + z_2}. [Hint: Use the binomial theorem to expand (z1+z2)n(z_1 + z_2)^n, as well as the formula for the binomial coefficients.]
Solution
This problem had stumped me for a while. I ended up taking a break from this problem, learning some combinatorics, and coming back to it. Upon doing so I realized that there is a very easy combinatorical solution to this problem using the basic method of "double-counting". Take a look at the following infinite list of terms.

1aba22abb22a36a2b2ab22b36\begin{array}{ccccc} 1 \\[4pt] a & b \\[4pt] \frac{a^2}{2} & ab & \frac{b^2}{2} \\[4pt] \frac{a^3}{6} & \frac{a^2b}{2} & \frac{ab^2}{2} & \frac{b^3}{6} \\[4pt] \vdots & \vdots & \vdots & \vdots & \vdots \end{array}

Here the term at the n+1n + 1th row and the k+1k + 1th column is given by the formula ankbk(nk)!k!\frac{a^{n-k}b^{k}}{(n -k)!k!}, where we always have knk \leq n, so the sum of all of these terms is n=0k=0nankbk(nk)!k!\sum_{n=0}^{\infty}\sum_{k=0}^{n} \frac{a^{n-k}b^k}{(n-k)!k!}.

The key insight of the double-counting method is that there are two different ways of adding up all of the terms- you can add all the rows or add all of the columns- but the result is the same.

If we add all the rows we get

n=0k=0nankbk(nk)!k!=1+(a+b)+(a22+ab+b22)+(a36+a2b2+ab22+b36)+=(a+b)00!+(a+b)11!+(a+b)22!+(a+b)33!+=n=0(a+b)nn!=ea+b\sum_{n=0}^{\infty}\sum_{k=0}^{n} \frac{a^{n-k}b^k}{(n-k)!k!} \\ = 1 + (a + b) + \left(\frac{a^2}{2} + ab + \frac{b^2}{2}\right) + \left( \frac{a^3}{6} + \frac{a^2b}{2} + \frac{ab^2}{2} + \frac{b^3}{6} \right) + \cdots \\ = \frac{(a + b)^0}{0!} + \frac{(a + b)^1}{1!} + \frac{(a + b)^2}{2!} + \frac{(a + b)^3}{3!} + \cdots \\ = \sum_{n=0}^{\infty} \frac{(a + b)^n}{n!} = e^{a + b}

If we add all the columns we get

n=0k=0nankbk(nk)!k!=(1+a+a22+a36+)+(b+ab+a2b2+)+(b22+ab22+)+(b36+)+=n=0ann!+bn=0ann!+b22n=0ann!+b36n=0ann!+=n=0ann!(1+b+b22+b36+)=n=0ann!k=0bkk!=eaeb\begin{aligned} &\sum_{n=0}^{\infty}\sum_{k=0}^{n} \frac{a^{n-k}b^k}{(n-k)!k!} \\ &= \left(1 + a + \frac{a^2}{2} + \frac{a^3}{6} + \cdots \right) + \left( b + ab + \frac{a^2b}{2} + \cdots \right) \\ &+ \left( \frac{b^2}{2} + \frac{ab^2}{2} + \cdots \right) + \left(\frac{b^3}{6} + \cdots\right) + \cdots \\ &= \sum_{n=0}^{\infty} \frac{a^n}{n!} + b\sum_{n=0}^{\infty}\frac{a^n}{n!} + \frac{b^2}{2}\sum_{n=0}^{\infty}\frac{a^n}{n!} + \frac{b^3}{6}\sum_{n=0}^{\infty} \frac{a^n}{n!} + \cdots \\ &= \sum_{n=0}^{\infty}\frac{a^n}{n!} \left( 1 + b + \frac{b^2}{2} + \frac{b^3}{6} + \cdots \right) \\ &= \sum_{n=0}^{\infty}\frac{a^n}{n!} \cdot \sum_{k=0}^{\infty}\frac{b^k}{k!} = e^a \cdot e^b \end{aligned}

This is the heart and soul of the proof, but it is a little bit hand-wavey. I hope that the above gives the reader a clear intuition behind the proof. In order to turn this intuition into more rigorous mathematics, we use the binomial expansion

(a+b)n=k=0nn!ankbk(nk)!k!(a + b)^n = \sum_{k=0}^{n} \frac{n!a^{n-k}b^k}{(n-k)!k!}

to calculate that

n=0k=0nankbk(nk)!k!=n=0(a+b)nn!=ea+b\sum_{n=0}^{\infty}\sum_{k=0}^{n} \frac{a^{n-k}b^k}{(n-k)!k!} = \sum_{n=0}^{\infty} \frac{(a+b)^n}{n!} = e^{a + b}

and combinatorically rearrange the series to switch from adding the rows to adding the columns via the identity

n=0k=0nf(n,k)=k=0n=kf(n,k)\sum_{n=0}^{\infty}\sum_{k=0}^{n} f(n,k) = \sum_{k=0}^{\infty}\sum_{n=k}^{\infty} f(n,k)

which holds because both the LHS and RHS are summing over each k,nNk, n \in \N such that knk \leq n. Plugging in f(n,k)=ankbk(nk)!k!f(n,k)=\frac{a^{n-k}b^{k}}{(n-k)!k!} to the identity yields

n=0k=0nankbk(nk)!k!=k=0n=kankbk(nk)!k!=k=0bkk!n=kank(nk)!=k=0bkk!n=0ann!=eaeb\sum_{n=0}^{\infty}\sum_{k=0}^{n} \frac{a^{n-k}b^k}{(n-k)!k!} = \sum_{k=0}^{\infty}\sum_{n=k}^{\infty} \frac{a^{n-k}b^k}{(n-k)!k!} \\ = \sum_{k=0}^{\infty} \frac{b^k}{k!} \sum_{n=k}^{\infty} \frac{a^{n-k}}{(n-k)!} = \sum_{k=0}^{\infty} \frac{b^k}{k!} \sum_{n=0}^{\infty} \frac{a^n}{n!} = e^a \cdot e^b \qquad \square


(c)\quad \text{(c)} Show that if zz is purely imaginary, that is, z=iyz = iy with yRy \in \R, then

eiy=cosy+isiny.e^{iy} = \cos y + i \sin y.

This is Euler's identity. [Hint: Use power series.]
Solution
The general idea is

eiy=n=0(iy)nn!=1+(iy)+(iy)22+(iy)36+=1+iyy22iy36+=(1y22+)+(yy36+)i=cosy+isinye^{iy} = \sum_{n=0}^{\infty}\frac{(iy)^n}{n!} = 1 + (iy) + \frac{(iy)^2}{2} + \frac{(iy)^3}{6} + \cdots \\ = 1 + iy - \frac{y^2}{2} -\frac{iy^3}{6} + \cdots \\ = (1 - \frac{y^2}{2} + \cdots) + (y - \frac{y^3}{6} + \cdots)i \\ = \cos y + i \sin y

More rigorously,

eiy=n=0(iy)nn!=n=0n is even(iy)nn!+n=0n is odd(iy)nn!=n=0(iy)2n(2n)!+n=0(iy)2n+1(2n+1)!=n=0i2n(y)2n(2n)!+n=0i2n+1(y)2n+1(2n+1)!=n=0i2n(y)2n(2n)!+in=0i2n(y)2n+1(2n+1)!=n=0(1)n(y)2n(2n)!+in=0(1)n(y)2n+1(2n+1)!=cosy+isinye^{iy} = \sum_{n=0}^{\infty}\frac{(iy)^n}{n!} = \sum_{n=0 \text{, } n \text { is even}}^{\infty}\frac{(iy)^n}{n!} + \sum_{n=0 \text{, } n \text{ is odd}}^{\infty}\frac{(iy)^n}{n!} \\ = \sum_{n=0}^{\infty} \frac{(iy)^{2n}}{(2n)!} + \sum_{n=0}^{\infty} \frac{(iy)^{2n+1}}{(2n+1)!} \\ = \sum_{n=0}^{\infty} \frac{i^{2n}(y)^{2n}}{(2n)!} + \sum_{n=0}^{\infty} \frac{i^{2n+1}(y)^{2n+1}}{(2n+1)!} \\ = \sum_{n=0}^{\infty} \frac{i^{2n}(y)^{2n}}{(2n)!} + i\sum_{n=0}^{\infty} \frac{i^{2n}(y)^{2n+1}}{(2n+1)!} \\ = \sum_{n=0}^{\infty} \frac{(-1)^{n}(y)^{2n}}{(2n)!} + i\sum_{n=0}^{\infty} \frac{(-1)^{n}(y)^{2n+1}}{(2n+1)!} \\ = \cos y + i \sin y \quad \square


(d)\quad \text{(d)} More generally,

ex+iy=ex(cosy+isiny)e^{x + iy} = e^x(\cos y + i \sin y)

whenever x,yRx, y \in \R, and show that

ex+iy=ex\left| e^{x + iy} \right| = e^{x}

Solution
Combining parts (b)\text{(b)} and (c)\text{(c)} gives

ex+iy=exeiy=ex(cosy+isiny)e^{x + iy} = e^x \cdot e^{iy} = e^x ( \cos y + i \sin y)

and we can calculate that

ex+iy=ex(cosy+isiny)=excosy+siny=ex|e^{x + iy}| = |e^x (\cos y + i \sin y)| = |e^x| |\cos y + \sin y| = e^x \quad \square


(e)\quad \text{(e)} Prove that ez=1e^z = 1 if and only if z=2πkiz = 2 \pi k i for some integer kk.
Solution
Let z=x+iyz = x + iy with x,yRx, y \in \R

1=ez=ex+iy=ex(cosy+isiny)    excosy=1 and sinyex=01 = e^z = e^{x+iy} = e^x(\cos y + i \sin y) \\ \iff e^x \cos y = 1 \text{ and } \sin y \cdot e^x = 0

Remember that xx and yy are real so

sinyex=0    siny=0    y=2πk, for some integer k\sin y \cdot e^x = 0 \iff \sin y = 0 \iff y = 2 \pi k \text{, for some integer } k

We can now plug in yy to the other equation to find xx

excosy=1    excos2πk=1    ex=1    x=0e^x \cos y = 1 \iff e^x \cos 2\pi k = 1 \iff e^x = 1 \iff x = 0

So in conclusion

ez=1    y=2πk and x=0    z=2πkie^z = 1 \iff y = 2 \pi k \text{ and } x = 0 \iff z = 2\pi ki \quad \square


(f)\quad \text{(f)} Show that every complex number z=x+iyz = x + iy can be written in the form

z=reiθ,z = r e^{i \theta},

where rr is unique and in the range 0r<0 \leq r < \infty, and θR\theta \in \R is unique up to an integer multiple of 2π2 \pi. Check that

r=zandθ=arctan(y/x)r = |z| \quad \text{and} \quad \theta = \arctan(y/x)

whenever these formulas make sense.
Solution

Start by dividing out the radius r=z=x2+y2r = |z| = \sqrt{x^2 + y^2} from zz,

z=x+iyzr=xr+iyrz = x + iy \\ \frac{z}{r} = \frac{x}{r} + i \frac{y}{r} \\

Multiply by rr but don't cancel on the right hand side,
z=r(xr+iyr)z = r \left( \frac{x}{r} + i \frac{y}{r} \right)

Next, consider the following right triangle.

x y r θ

By the diagram and basic trig we have cosθ=xr\cos \theta = \frac{x}{r}, sinθ =yr\sin \theta \ = \frac{y}{r}, and θ=arctan(y/x)\theta = \arctan (y/x). Plugging in,

z=r(cosθ+isinθ)=reiθz = r \cdot ( \cos \theta + i \sin \theta) = re^{i\theta} \quad \square


(g)\quad \text{(g)} In particular, i=eiπ/2i = e^{i \pi / 2}. What is the geometric meaning of multiplying a complex number by ii? Or by eiθe^{i \theta} for any θR\theta \in \R
Solution
Multiplying by ii is rotating π/2\pi / 2 radians, in general multipling by eiθe^{i \theta} is rotating by θ\theta radians.


(h)\quad \text{(h)} Given θR\theta \in \R, show that

cosθ=eiθ+eiθ2andsinθ=eiθeiθ2i.\cos \theta = \frac{e^{i\theta} + e^{-i\theta}}{2} \quad \text{and} \quad \sin \theta = \frac{e^{i\theta} - e^{-i\theta}}{2i}.

There are also called Euler's indentities.
Solution

eiθ=cosθ+isinθeiθ=ei(θ)=cos(θ)+isin(θ)=cosθisinθeiθ+eiθ=(cosθ+isinθ)+(cosθisinθ)=2cosθcosθ=eiθ+eiθ2eiθeiθ=(cosθ+isinθ)(cosθisinθ)=2isinθsinθ=eiθeiθ2ie^{i \theta} = \cos \theta + i \sin \theta \\ e^{-i \theta} = e^{i \cdot (-\theta)} = \cos (-\theta) + i \sin (-\theta) = \cos \theta - i \sin \theta \\ e^{i \theta} + e^{-i\theta} = (\cos \theta + i \sin \theta) + (\cos \theta - i \sin \theta) = 2 \cos \theta \\ \cos \theta = \frac{e^{i \theta} + e^{-i \theta}}{2} \\ e^{i \theta} - e^{-i \theta} = (\cos \theta + i \sin \theta) - (\cos \theta - i \sin \theta) = 2i \sin \theta \\ \sin \theta = \frac{e^{i \theta} - e^{-i\theta}}{2i} \quad \square


(i)\quad \text{(i)} Use the complex exponential to derive trigonometric identities such as

cos(θ+ϕ)=cosθcosϕsinθsinϕ,\cos (\theta + \phi) = \cos \theta \cos \phi - \sin \theta \sin \phi,

and then show that

2sinθsinϕ=cos(θϕ)cos(θ+ϕ),2sinθcosϕ=sin(θ+ϕ)+sin(θϕ).2 \sin \theta \sin \phi = \cos(\theta - \phi) - \cos(\theta + \phi), \\ 2 \sin \theta \cos \phi = \sin(\theta + \phi) + \sin(\theta - \phi).

This calculation connects the solution given by d'Alember in terms of traveling waves and the solution in terms of superposition of standing waves.
Solution
First calculate that

2cos(θ+ϕ)=ei(θ+ϕ)+ei(θ+ϕ)=eiθeiϕ+eiθeiϕ=(cosθ+isinθ)(cosϕ+isinϕ)+(cosθisinθ)(cosϕisinϕ)=(cosθcosϕ+icosθsinϕ+isinθcosϕsinθsinϕ)+(cosθcosϕicosθsinϕisinθcosϕsinθsinϕ)=2cosθcosϕ2sinθsinϕ\begin{aligned} &2 \cos(\theta + \phi)\\ &= e^{i(\theta + \phi)} + e^{-i(\theta + \phi)} = e^{i\theta}e^{i\phi} + e^{-i\theta}e^{-i\phi} \\ &= (\cos \theta + i \sin \theta) (\cos \phi + i \sin \phi) + (\cos \theta - i \sin \theta)(\cos \phi - i \sin \phi) \\ &= (\cos\theta \cos\phi + i \cos\theta \sin\phi + i \sin \theta \cos \phi - \sin \theta \sin \phi) \\ &+ (\cos \theta \cos \phi - i \cos \theta \sin \phi - i \sin \theta \cos \phi - \sin \theta \sin \phi) \\ &= 2 \cos \theta \cos \phi - 2 \sin \theta \sin \phi \quad \square \end{aligned}

Using this formula

cos(θϕ)cos(θ+ϕ)=cos(θ+(ϕ))cos(θ+ϕ)=(cosθcos(ϕ)sinθsin(ϕ))(cosθcosϕsinθsinϕ)=cosθcosϕ+sinθsinϕcosθcosϕ+sinθsinϕ=2sinθsinϕ\begin{aligned} &\cos(\theta - \phi) - \cos(\theta + \phi) \\ &= \cos(\theta + (-\phi)) - \cos(\theta + \phi) \\ &= (\cos \theta \cos (- \phi) - \sin \theta \sin (-\phi)) - (\cos \theta \cos \phi - \sin \theta \sin \phi) \\ &= \cos \theta \cos \phi + \sin \theta \sin \phi - \cos \theta \cos \phi + \sin \theta \sin \phi \\ &= 2 \sin \theta \sin \phi \quad \square \end{aligned}

Now let's calculate sine's angle addition formula

2isin(θ+ϕ)=ei(θ+ϕ)ei(θ+ϕ)=eiθeiϕei(θ)ei(ϕ)=(cosθcosϕ+icosθsinϕ+isinθcosϕsinθsinϕ)(cosθcosϕicosθsinϕisinθcosϕsinθsinϕ)=2icosθsinϕ+2isinθcosϕsin(θ+ϕ)=cosθsinϕ+sinθcosϕ\begin{aligned} &2i \sin(\theta + \phi) \\ &= e^{i (\theta + \phi)} - e^{-i (\theta + \phi)} = e^{i\theta}e^{i\phi} - e^{i(-\theta)}e^{i(-\phi)} \\ &= (\cos \theta \cos \phi + i \cos \theta \sin \phi + i \sin \theta \cos \phi - \sin \theta \sin \phi) \\ &- (\cos \theta \cos \phi - i \cos \theta \sin \phi - i \sin \theta \cos \phi - \sin \theta \sin \phi) \\ &= 2i \cos \theta \sin \phi + 2i\sin \theta \cos \phi \\ \end{aligned} \\ \sin(\theta + \phi) = \cos \theta \sin \phi + \sin \theta \cos \phi

We can now solve the last identity

sin(θ+ϕ)+sin(θϕ)=(cosθsinϕ+sinθcosϕ)+(cosθsin(ϕ)+sinθcos(ϕ))=(cosθsinϕ+sinθcosϕ)+(cosθsinϕ+sinθcosϕ)=2sinθcosϕ\begin{aligned} &\sin(\theta + \phi) + \sin(\theta - \phi) \\ &= (\cos \theta \sin \phi + \sin \theta \cos \phi) + (\cos \theta \sin (-\phi) + \sin \theta \cos (-\phi)) \\ &= (\cos \theta \sin \phi + \sin \theta \cos \phi) + (-\cos \theta \sin \phi + \sin \theta \cos \phi) \\ &= 2\sin \theta \cos \phi \quad \square \end{aligned}


5. Verify that f(x)=einxf(x) = e^{inx} is periodic with period 2π2 \pi and that

12πππeinxdx={1if n=0,0if n0.\frac{1}{2\pi} \int_{-\pi}^{\pi} e^{inx} \, dx = \begin{cases} 1 \quad \text{if } n = 0, \\ 0 \quad \text{if } n \neq 0. \end{cases}

Use this fact to prove that if n,m1n, m \geq 1 we have

1πππcosnxcosmxdx={0if nm,1if n=m,\frac{1}{\pi} \int_{-\pi}^{\pi} \cos nx \cos mx \, dx = \begin{cases} 0 \quad \text{if } n \neq m, \\ 1 \quad \text{if } n = m, \end{cases}

and similarly

1πππsinnxsinmxdx={0if nm,1if n=m.\frac{1}{\pi} \int_{-\pi}^{\pi} \sin nx \sin mx \, dx = \begin{cases} 0 \quad \text{if } n \neq m, \\ 1 \quad \text{if } n = m. \end{cases}

Finally, show that

ππsinnxcosmxdx=0for any n,m.\int_{-\pi}^{\pi} \sin nx \cos mx \, dx = 0 \quad \text{for any } n, m.

[Hint: Calculate einxeimx+einxeimxe^{inx}e^{-imx} + e^{inx}e^{imx} and einxeimxeinxeimxe^{inx}e^{-imx} - e^{inx}e^{imx}.]

Solution

First we need to show that f(x)=einxf(x) = e^{inx} is periodic with period 2π2 \pi. Plug in and use 4(b)4(\text{b}) to get

f(x+2π)=ein(x+2π)=einx+2πin=f(x)e2πin.(1)\tag{1} f(x + 2\pi) = e^{in(x + 2\pi)} = e^{inx + 2\pi in} = f(x)e^{2\pi in}.

By Euler's identity (4(c))\left(4 (\text{c})\right)

e2πin=cos(2πn)+isin(2πn)=1,e^{2\pi i n} = \cos(2\pi n) + i \sin(2\pi n) = 1,

plugging this into (1)(1) gives

f(x+2π)=f(x),f(x + 2 \pi) = f(x),

so by definition ff is periodic with period 2π2 \pi.

Next we need to show that

12πππeinxdx={1if n=0,0if n0.(2)\tag{2} \frac{1}{2\pi} \int_{-\pi}^{\pi} e^{inx} \, dx = \begin{cases} 1 \quad \text{if } n = 0, \\ 0 \quad \text{if } n \neq 0. \end{cases}

Calculate that

ππeinxdx=ππcos(nx)+isin(nx)dx=ππcos(nx)dx+iππsin(nx)dx\int_{-\pi}^{\pi} e^{inx} \, dx = \int_{-\pi}^{\pi} \cos(nx) + i \sin(nx) \, dx \\ = \int_{-\pi}^{\pi} \cos(nx) \, dx + i \int_{-\pi}^{\pi} \sin(nx) \, dx

If n0n \neq 0 then

=sin(nx)nππicos(nx)nππ=sin(nπ)sin(nπ)nicos(nπ)cos(nπ)n=0,= \left. \frac{\sin(nx)}{n} \right|_{-\pi}^{\pi} - i \left. \frac{\cos(nx)}{n} \right|_{-\pi}^{\pi} \\[4px] = \frac{\sin(n\pi) - \sin(-n\pi)}{n} - i \frac{\cos(n\pi) - \cos(-n\pi)}{n} = 0,

the last equality following from sin(nπ)=0\sin(n\pi) = 0 and cos(x)=cos(x)\cos(x) = \cos(-x).
The case of n=0n=0 is trivial because einx=1e^{inx} = 1.

Next we need to show that if n,m1n, m \geq 1 then

1πππcosnxcosmxdx={0if nm1if n=m.\frac{1}{\pi} \int_{-\pi}^{\pi} \cos nx \cos mx \, dx = \begin{cases} 0 \quad \text{if } n \neq m \\ 1 \quad \text{if } n = m. \end{cases}

Using the hint,

einxeimx+einxeimx=(cos(nx)+isin(nx))(cos(mx)+isin(mx))+(cos(nx)+isin(nx))(cos(mx)+isin(mx))=(cosnxcosmxicosnxsinmx+isinnxcosmx+sinnxsinmx)+(cosnxcosmx+icosnxsinmx+isinnxcosmxsinnxsinmx)=2cosnxcosmx+2isinnxcosmx,e^{inx}e^{-imx} + e^{inx}e^{imx} \\ = (\cos(nx) + i \sin(nx))(\cos(-mx) + i \sin(-mx)) \\ + (\cos(nx) + i\sin(nx))(\cos(mx) + i \sin(mx)) \\ = (\cos nx \cos mx - i\cos nx \sin mx + i\sin nx\cos mx + \sin nx \sin mx ) \\ + ( \cos nx \cos mx + i \cos nx \sin mx + i \sin nx \cos mx - \sin nx \sin mx ) \\ = 2\cos nx \cos mx + 2i \sin nx \cos mx,

and similarly

einxeimxeinxeimx=2sinnxsinmx2icosnxsinmx.e^{inx}e^{-imx} - e^{inx}e^{imx} = 2 \sin nx \sin mx - 2i \cos nx \sin mx.

Integrate both sides to get

ππei(nm)xdx+ππei(n+m)xdx=2ππcosnxcosmxdx+2iππsinnxcosmxdx,\int_{-\pi}^{\pi} e^{i(n-m)x} \, dx + \int_{-\pi}^{\pi} e^{i(n+m)x} \, dx \\ = 2\int_{-\pi}^{\pi} \cos nx \cos mx \, dx + 2i\int_{-\pi}^{\pi} \sin nx \cos mx \, dx,

and similarly

ππei(nm)xdxππei(n+m)xdx=2ππsinnxsinmxdx2iππcosnxsinmxdx.\int_{-\pi}^{\pi} e^{i(n-m)x} \, dx - \int_{-\pi}^{\pi} e^{i(n+m)x} \, dx \\ = 2\int_{-\pi}^{\pi} \sin nx \sin mx \, dx - 2i\int_{-\pi}^{\pi} \cos nx \sin mx \, dx.

If n=mn = m then of course nm=0n - m = 0, so we can plug into (2)(2) to get

2π=2ππcosnxcosmxdx+2iππsinnxcosmxdx2 \pi = 2\int_{-\pi}^{\pi} \cos nx \cos mx \, dx + 2i\int_{-\pi}^{\pi} \sin nx \cos mx \, dx

and similarly

2π=2ππsinnxsinmxdx2iππcosnxsinmxdx.2 \pi = 2\int_{-\pi}^{\pi} \sin nx \sin mx \, dx - 2i\int_{-\pi}^{\pi} \cos nx \sin mx \, dx.

If nmn \neq m then we get

0=2ππcosnxcosmxdx+2iππsinnxcosmxdx,0 = 2\int_{-\pi}^{\pi} \cos nx \cos mx \, dx + 2i\int_{-\pi}^{\pi} \sin nx \cos mx \, dx,

and similarly

0=2ππsinnxsinmxdx2iππcosnxsinmxdx.0 = 2\int_{-\pi}^{\pi} \sin nx \sin mx \, dx - 2i\int_{-\pi}^{\pi} \cos nx \sin mx \, dx.

The result follows from equating real and imaginary parts. \quad \square


6. Prove that if ff is a twice continuously differentiable function on R\R which is a solution of the equation

f(t)+c2f(t)=0,f''(t) + c^2f(t) = 0,

then there exist constants aa and bb such that

f(t)=acosct+bsinct.f(t) = a \cos ct + b \sin ct.

This can be done by differentiating the two functions g(t)=f(t)cosctc1f(t)sinctg(t) = f(t) \cos ct - c^{-1}f'(t)\sin ct and h(t)=f(t)sinct+c1f(t)coscth(t) = f(t) \sin ct + c^{-1}f'(t)\cos ct.

Solution

Let's start by following the hint. First we differentiate g(t)g(t),

g(t)=ddt[f(t)cosct]c1ddt[f(t)sinct]=(f(t)cosct+f(t)csinct)c1(f(t)sinct+f(t)ccosct)=(f(t)cosctcf(t)sinct)c1(c2f(t)sinct+cf(t)cosct)=f(t)cosctcf(t)sinct+cf(t)sinctf(t)cosct=0.g'(t) = \frac{d}{dt} \left[ f(t)\cos ct \right] - c^{-1}\frac{d}{dt} \left[ f'(t) \sin ct \right] \\[4px] = \left(f'(t)\cos ct + f(t) \cdot -c\sin ct\right) - c^{-1}\left(f''(t)\sin ct + f'(t) \cdot c \cos ct\right) \\ = \left(f'(t)\cos ct - c f(t) \sin ct \right) - c^{-1} \left( -c^2 f(t) \sin ct + cf'(t) \cos ct \right) \\ = f'(t) \cos ct - c f(t) \sin ct + c f(t) \sin ct -f'(t) \cos ct = 0.

Then we differentiate h(t)h(t),

h(t)=ddt[f(t)sinct]+c1ddt[f(t)cosct]=(f(t)sinct+f(t)ccosct)+c1(f(t)cosct+f(t)(c)sinct)=(f(t)sinct+f(t)ccosct)+c1(c2f(t)cosct+f(t)(c)sinct)=f(t)sinct+f(t)ccosctcf(t)cosctf(t)sinct=0.h'(t) = \frac{d}{dt} \left[ f(t) \sin ct \right] + c^{-1} \frac{d}{dt} \left[ f'(t) \cos ct \right] \\ = \left(f'(t) \sin ct + f(t) \cdot c \cdot \cos ct \right) + c^{-1}\left( f''(t) \cos ct + f'(t) \cdot (-c) \cdot \sin ct \right) \\ = \left(f'(t) \sin ct + f(t) \cdot c \cdot \cos ct \right) + c^{-1}\left( -c^2 f(t) \cos ct + f'(t) \cdot (-c) \cdot \sin ct \right) \\ = f'(t) \sin ct + f(t) \cdot c \cdot \cos ct - c f(t) \cos ct - f'(t) \cdot \sin ct = 0. \\

Therefore there exist constants aa and bb such that

g(t)=a,h(t)=bf(t)cosctc1f(t)sinct=a,f(t)sinct+c1f(t)cosct=bc1f(t)sinct=f(t)coscta,f(t)=bc1f(t)cosctsinctg(t) = a, \quad h(t) = b \\ f(t) \cos ct - c^{-1}f'(t)\sin ct = a, \quad f(t) \sin ct + c^{-1}f'(t)\cos ct = b \\[4px] c^{-1}f'(t) \sin ct = f(t) \cos ct - a, \quad f(t) = \frac{b - c^{-1}f'(t)\cos ct}{\sin ct}

Plug the right into the left

c1f(t)sinct=bc1f(t)cosctsinctcosctac1f(t)sin2ct=bcosctc1f(t)cos2ctasinctc1f(t)sin2ct+c1f(t)cos2ct=bcosctasinctc1f(t)(sin2ct+cos2ct)=bcosctasinctc1f(t)=bcosctasinctf(t)=cbcosctcasinct0tf(τ)dτ=b0tccoscτdτa0tcsincτdτf(t)f(0)=bsincτ0t+acoscτ0tf(t)f(0)=bsinct+acosctac^{-1}f'(t)\sin ct = \frac{b - c^{-1}f'(t)\cos ct}{\sin ct}\cos ct - a \\[4px] c^{-1}f'(t)\sin^2 ct = b\cos ct - c^{-1}f'(t) \cos^2 ct - a \sin ct \\ c^{-1}f'(t)\sin^2 ct + c^{-1}f'(t) \cos^2 ct = b\cos ct - a \sin ct \\ c^{-1}f'(t) (\sin^2 ct + \cos^2 ct) = b\cos ct - a \sin ct \\ c^{-1}f'(t) = b\cos ct - a \sin ct \\ f'(t) = cb\cos ct - ca \sin ct \\ \int_{0}^{t} f'(\tau) \, d\tau = b \int_{0}^{t}c \cos c\tau \, d\tau - a \int_{0}^{t} c \sin c \tau \, d\tau \\ f(t) - f(0) = b\left.\sin c\tau \right|_{0}^{t} + \left.a \cos c \tau \right|_{0}^{t} \\ f(t) - f(0) = b \sin ct + a \cos ct - a

It suffices to show that

f(0)=af(0) = a

We know that

f(π2c)=bc1f(t)cosπ2sinπ2=bf\left(\frac{\pi}{2c}\right) = \frac{b - c^{-1}f'(t)\cos \frac{\pi}{2}}{\sin \frac{\pi}{2}} = b

We can plug this in to get

f(π2c)f(0)=bsinπ2+acosπ2a=baf(0)=af\left(\frac{\pi}{2c}\right) - f(0) = b \sin \frac{\pi}{2} + a \cos \frac{\pi}{2} - a = b-a \\ f(0) = a \quad \square


7. Show that if aa and bb are real, then one can write

acosct+bsinct=Acos(ctϕ),a \cos ct + b \sin ct = A \cos (ct - \phi),

where A=a2+b2A = \sqrt{a^2 + b^2}, and ϕ\phi is chosen so that

cosϕ=aa2+b2andsinϕ=ba2+b2.\cos \phi = \frac{a}{\sqrt{a^2+b^2}} \quad \text{and} \quad \sin \phi = \frac{b}{\sqrt{a^2+b^2}}.

Solution

First use the cosine angle addition formula from 4(i)4(i) to calculate that

cos(ctϕ)=cos(ct)cos(ϕ)sin(ct)sin(ϕ)=cosctcosϕ+sinctsinϕ.\begin{aligned} \cos (ct - \phi) &= \cos(ct) \cos (-\phi) - \sin(ct) \sin(-\phi) \\ &= \cos ct \cos \phi + \sin ct \sin \phi. \end{aligned}

Using the following right triangle

a b a 2 + b 2 φ

we can choose ϕ\phi such that cosϕ=aa2+b2\cos \phi = \frac{a}{\sqrt{a^2+b^2}} and sinϕ=ba2+b2\sin \phi = \frac{b}{\sqrt{a^2+b^2}}. Plugging in,

cos(ctϕ)=aa2+b2cosct+ba2+b2sinct\cos(ct - \phi) = \frac{a}{\sqrt{a^2+b^2}} \cos ct + \frac{b}{\sqrt{a^2+b^2}} \sin ct

Multiply both sides by A=a2+b2A = \sqrt{a^2+b^2} to get the desired result,

Acos(ctϕ)=acosct+bsinct.A\cos(ct - \phi) = a \cos ct + b \sin ct. \quad \square


8. Suppose FF is a function on (a,b)(a,b) with two continuous derivatives. Show that whenever xx and x+hx + h belong to (a,b)(a,b), one may write

F(x+h)=F(x)+hF(x)+h22F(x)+h2φ(h),F(x+h) = F(x) + hF'(x) + \frac{h^2}{2}F''(x) + h^2\varphi (h),

where φ(h)0\varphi(h) \to 0 as h0h \to 0.
Deduce that

F(x+h)+F(xh)2F(x)h2F(x)as h0.\frac{F(x+h) + F(x-h) - 2F(x)}{h^2} \to F''(x) \quad \text{as } h \to 0.

Hint: This is simply a Taylor expansion. It may be obtained by noting that

F(x+h)F(x)=xx+hF(y)dy,F(x+h) - F(x) = \int_{x}^{x+h} F'(y) \, dy,

and then writing F(y)=F(x)+(yx)F(x)+(yx)ψ(yx)F'(y) = F'(x) + (y-x)F''(x) + (y-x)\psi(y-x), where ψ(h)0\psi(h) \to 0 as h0h \to 0.


9. In the case of the plucked string, use the formula for the Fourier sine coefficients to show that

Am=2hm2sinmpp(πp).A_m = \frac{2h}{m^2} \frac{\sin mp}{p(\pi - p)}.

For what position of pp are the second, fourth, \dots harmonics missing? For what position of pp are the third, sixth, \dots harmonics missing?


10. Show that the expression of the Laplacian

=2x2+2y2\triangle = \frac{\partial^2}{\partial x^2} + \frac{\partial^2}{\partial y^2}

is given in polar coordinates by the formula

=2r2+1rr+1r22θ2.\triangle = \frac{\partial^2}{\partial r^2} + \frac{1}{r}\frac{\partial}{\partial r} + \frac{1}{r^2}\frac{\partial^2}{\partial \theta^2}.

Also, prove that

ux2+uy2=ur2+1r2uθ2.\left|\frac{\partial u}{\partial x}\right|^2 + \left| \frac{\partial u}{\partial y} \right|^2 = \left| \frac{\partial u}{\partial r} \right|^2 + \frac{1}{r^2} \left| \frac{\partial u}{\partial \theta} \right|^2 .


11. Show that if nZn \in \Z the only solutions of the differential equation

r2F(r)+rF(r)n2F(r)=0,r^2 F''(r) + rF'(r) - n^2F(r) = 0,

which are twice differentiable when r>0r > 0, are given by linear combinations of rnr^n and rnr^{-n} when n0n \neq 0, and 11 and logr\log r when n=0n = 0.
Hint: If FF solves the equation, write F(r)=g(r)rnF(r) = g(r)r^n, find the equation satisfied by gg, and conclude that rg(r)+2ng(r)=crg'(r) + 2ng(r) = c where cc is a constant.


0 u =0 u = f 0 u =0 1 u = f 1 u =0 π

Problem

1. Consider the Dirichlet problem illustrated above.
More precisely, we look for a solution of the steady-state heat equation u=0\triangle u = 0 in the rectangle R={(x,y):0xπ,0y1}R = \{(x,y): 0 \leq x \leq \pi, 0 \leq y \leq 1 \} that vanishes on the vertical sides of RR, and so that

u(x,0)=f0(x)andu(x,1)=f1(x),u(x,0) = f_0(x) \quad \text{and} \quad u(x,1) = f_1(x),

where f0f_0 and f1f_1 are the initial data which fix the temperature distribution on the horizontal sides of the rectangle.

Use separation of variables to show that if f0f_0 and f1f_1 have Fourier expansions

f0(x)=k=1Aksinkxandf1(x)=k=1Bksinkx,f_0(x) = \sum_{k=1}^{\infty} A_k \sin kx \quad \text{and} \quad f_1(x) = \sum_{k=1}^{\infty} B_k \sin kx,

then

u(x,y)=k=1(sinhk(1y)sinhkAk+sinhkysinhkBk)sinkx.u(x,y) = \sum_{k=1}^{\infty} \left( \frac{\sinh k(1-y)}{\sinh k}A_k + \frac{\sinh ky}{\sinh k}B_k \right) \sin kx.

We recall the definitions of the hyperbolic sine and cosine functions:

sinhx=exex2andcoshx=ex+ex2.\sinh x = \frac{e^x - e^{-x}}{2} \quad \text{and} \quad \cosh x = \frac{e^x + e^{-x}}{2}.

Compare this result with the solution of the Dirichlet problem in the strip obtained in Problem 3, Chapter 5.

\\

Chapter 1. The Genesis of Fourier Analysis