2

Just started learning Fourier series and am having a rough time trying to wrap my head around the textbook. Below are two excerpts from the book each followed up a question. Many thanks in advance!

"The theory of Fourier series deals with the representation of an "arbitrary" $2\pi$ periodic function $u$ as an (infinite) sum of multiples of these functions.

$$u(x)= \sum_{n=0}^\infty (a_n \cos{nx}+b_n \sin{nx}) \tag 1$$

can be rewritten using Euler's formulae as

$$u(x)= \sum_{n=-\infty}^\infty c_n e^{inx} \tag 2."$$

Question 1

How do I get from equation (1) to equation (2)? My attempt gives me-

$$(a_n \cos{nx}+b_n \sin{nx})=e^{inx}\bigg( \frac{a_n}{2}+\frac{b_n}{2i}\bigg)+e^{-inx}\bigg( \frac{a_n}{2}-\frac{b_n}{2i}\bigg) \text{???}$$

Also why do the limits of summation change from $\sum_{n=0}^\infty$ to $\sum_{n=-\infty}^{\infty}$?

Further the textbook goes on to explain-

"Now suppose that (2) is uniformly convergent. Then the sum $u(x)$ is continuous. Multiply both sides by $e^{-imx}$ and integrate from $0$ to $2\pi$. Since the convergence is uniform the right hand side can be integrated term-wise. This yields

$$\int\limits_{0}^{2\pi}u(x)e^{-imx}dx=\sum_{n=-\infty}^{\infty} c_n \int\limits_{0}^{2\pi} e^{i(n-m)x}dx=2\pi c_m$$

Where all the integrals in the sum are zero except the one with $m=n$ which equals $2\pi$."

Question 2

What is the reason for multiplying by $e^{-imx}$? Also I do not understand the final statement about all the integrals summing to zero? If we sum between two finite numbers e.g. between $-3$ and $3$ then we get-

$$c_{-3}(-2\pi)+c_{-2}(2\pi)+c_{-1}(-2\pi)+c_{0}(2\pi)+c_{1}(-2\pi)+c_{2}(2\pi)+c_{3}(-2\pi)$$

How can we possibly say that this sums to zero if we do not know the values of the $c_{n}$'s?

Eiraus
  • 349
  • 1
  • 3
  • 15
  • Sometimes my reason for answering is to try to make the answer as simple as possible. Whether I have done so you may try to judge below. – Michael Hardy Sep 15 '17 at 23:28

4 Answers4

1

Note that we can write

$$\begin{align} \sum_{n=0}^\infty(a_n\cos(nx)+b_n\sin(nx))&=\sum_{n=0}^\infty \left(a_n\frac{e^{inx}+e^{-inx}}{2}+b_n\frac{e^{inx}-e^{-inx}}{i2}\right)\\\\ &=\sum_{n=0}^\infty \left(\frac{a_n-ib_n}{2}\,e^{inx}\right)+\sum_{n=0}^\infty \left(\frac{a_n+ib_n}{2}\,e^{-inx}\right)\\\\ &=\sum_{n=0}^\infty \left(\frac{a_n-ib_n}{2}\,e^{inx}\right)+\sum_{n=-\infty}^0 \left(\frac{a_{-n}+ib_{-n}}{2}\,e^{inx}\right)\\\\ &=\sum_{n=-\infty}^\infty c_n e^{inx} \end{align}$$

where we have

$$c_n=\begin{cases}\frac{a_n-ib_n}{2}&,n>0\\\\ a_0&,n=0\\\\\frac{a_{-n}+ib_{-n}}{2}&,n<0\end{cases}$$

Mark Viola
  • 179,405
  • Thank you. I like the solution but haven't been able to figure out how you got from i2 in the denominator to ibn in the numerator on the second line? It is certainly obvious but I am just not seeing it. Thanks again! – Eiraus Sep 15 '17 at 19:38
  • You're welcome. My pleasure. Recall that $\frac1i=-i$. Then, $a_n+b_n/i=a_n-ib_n$. – Mark Viola Sep 15 '17 at 19:40
1

You got $$(a_n \cos{nx}+b_n \sin{nx})=e^{inx}\bigg( \frac{a_n}{2}+\frac{b_n}{2i}\bigg)+e^{-inx}\bigg( \frac{a_n}{2}-\frac{b_n}{2i}\bigg)=e^{inx}c_n+e^{-inx}d_n$$ so $$\sum_{n=0}^{\infty} (a_n \cos{nx}+b_n \sin{nx})=\sum_{n=0}^{\infty} e^{inx}c_n+\sum_{n=0}^{\infty}e^{-inx}d_n=\sum_{n=0}^{\infty} e^{inx}c_n+\sum_{-n=0}^{\infty}e^{inx}d_{-n}$$ since $c_n=\overline{d_n}$ then this will be $$=\sum_{n=0}^{\infty} e^{inx}c_n+\sum_{-n=0}^{\infty}e^{inx}\overline{c_n}=\sum_{n=-\infty}^{\infty} e^{inx}c_n$$

Note that for $n\neq m$ with $k=n-m\neq0$ $$\int\limits_{0}^{2\pi} e^{i(n-m)x}dx=\int\limits_{0}^{2\pi} e^{ikx}dx=\dfrac{e^{ikx}}{ik}\Big|_{0}^{2\pi}=\dfrac{e^{ik2\pi}-1}{ik}=0$$ and for $n=m$ $$\int\limits_{0}^{2\pi} e^{i(0)x}dx=\int\limits_{0}^{2\pi} dx=x\Big|_{0}^{2\pi}=2\pi$$

Nosrati
  • 29,995
  • Thanks for the clarification! Clears things up a bit. I made an error on my original post which I think you copy and pasted. I have changed it on my post but was unable to change it on your answer. In the first equation there should be a minus sign instead of plus sign in the set of brackets after the e^(-1nx). Many thanks again! – Eiraus Sep 15 '17 at 19:30
  • yes, I've copied the first assertion by you, will edit, Thanks. – Nosrati Sep 15 '17 at 19:33
0

Question 1

The summation goes from $-\infty$ to $+\infty$ to take into account the $\mathrm e^{-inx}$ terms.

Question 2

This is related to the periodicity of $\mathrm e^{ikx}$: $$\int_0^{2\pi}\mathrm e^{ikx}\mathrm dx =\begin{cases}0&\text{if }\; k\ne 0,\\2\pi&\text{if }\; k= 0.\end{cases}$$

Bernard
  • 175,478
0

\begin{align} a\cos(nx) + b\sin(nx) & = a \frac {e^{inx} + e^{-inx}} 2 + b \frac{e^{inx} - e^{-inx}}{2i} \\[12pt] & = \frac {a-bi} 2 e^{inx} + \frac{a+bi} 2 e^{-inx} \\[12pt] & = c_n e^{inx} + c_{-n} e^{-inx}. \end{align} (Here we used the fact that $\dfrac b i = -bi$ and $\dfrac {-b} i = bi.$)