Professional Documents
Culture Documents
1
X
bn sin nx (1)
nD1
If the numbers b1 ; b2 ; b3 ; : : : drop off quickly enough (we are foreshadowing the
importance of their decay rate) then the sum S.x/ will inherit all three properties:
Periodic S.x C 2/ D S.x/
Odd S. x/ D S.x/
S.0/ D S./ D 0
200 years ago, Fourier startled the mathematicians in France by suggesting that any odd
periodic function S.x/ could be expressed as an infinite series of sines. This idea started
an enormous development of Fourier series. Our first step is to find the number bk that
multiplies sin kx. The function S.x/ is transformed to a sequence of bs.
P
Suppose S.x/ D bn sin nx. Multiply both sides by sin kx. Integrate from 0 to :
Z
Z
Z
S.x/ sin kx dx D
b1 sin x sin kx dx C C
bk sin kx sin kx dx C (2)
0
On the right side, all integrals are zero except the highlighted one with n D k. This
property of orthogonality will dominate the whole chapter. For sines, integral D 0 is a
fact of calculus :
Z
sin nx sin kx dx D 0
cos mx dx D
sin nx sin kx D
sin mx
m
1
cos.n
2
D0
k/x
if n k :
(3)
0. So we use this:
1
cos.n C k/x :
2
(4)
Integrating cos .n k/x and cos .n C k/x gives zero, proving orthogonality of the sines.
The exception is when n D k. Then we are integrating .sin kx/2 D 21 12 cos 2kx:
Z
Z
Z
1
1
sin kx sin kx dx D
dx
cos 2kx dx D :
(5)
2
0
0 2
0 2
The highlighted term in equation (2) is .=2/bk. Multiply both sides by 2= to find bk .
435
Sine coefficients
S. x/ D
bk D
S.x/
2
S.x/ sin kx dx D
1
(6)
Notice that S.x/ sin kx is even (equal integrals from to 0 and from 0 to ).
I will go immediately to the most important example of a Fourier sine series.
S.x/ is an odd square wave with S W .x/ D 1 for 0 < x < . It is drawn in
Figure 8.1 as an odd function (with period 2) that vanishes at x D 0 and x D .
S W .x/ D 1
2
Figure 8.1: The odd square wave with S W .x C 2/ D S W .x/ D f1 or 0 or 1g.
Find the Fourier sine coefficients bk of the odd square wave S W .x/.
Example 1
Solution
bk D
2
sin kx dx D
2
cos kx
k
0
2
2 0 2 0 2 0
; ; ; ; ; ;:::
1 2 3 4 5 6
(7)
The even-numbered coefficients b2k are all zero because cos 2k D cos 0 D 1. The oddnumbered coefficients bk D 4 =k decrease at the rate 1=k. We will see that same 1=k
decay rate for all functions formed from smooth pieces and jumps.
Put those coefficients 4=k and zero into the Fourier sine series for S W .x/:
Square wave
SW.x/ D
4
sin x
sin 3x
sin 5x
sin 7x
C
C
C
C
1
3
5
7
(8)
Figure 8.2 graphs this sum after one term, then two terms, and then five terms. You can
see the all-important Gibbs phenomenon appearing as these partial sums include more
terms. Away from the jumps, we safely approach S W .x/ D 1 or 1. At x D =2, the
series gives a beautiful alternating formula for the number :
1 1
1 1
1 1
4 1 1
1D
C
C
so that D 4
C
C : (9)
1 3
5 7
1 3
5 7
The Gibbs phenomenon is the overshoot that moves closer and closer to the jumps.
Its height approaches 1:18 : : : and it does not decrease with more terms of the series.
This overshoot is the one greatest obstacle to calculation of all discontinuous functions
(like shock waves). We try hard to avoid Gibbs but sometimes we cant.
436
4 sin x
sin 3x
Solid curve
C
1
3
4 sin x
Dashed
1
4 sin x
sin 9x
5 terms:
CC
1
9
Gibbs overshoot !
SW D 1
x
2
Figure 8.2: The sums b1 sin x C C bN sin N x overshoot the square wave near jumps.
1
X
an cos nx:
(10)
nD1
Every cosine has period 2. Figure 8.3 shows two even functions, the repeating ramp
RR.x/ and the up-down train UD.x/ of delta functions. That sawtooth ramp RR is the
integral of the square wave. The delta functions in UD give the derivative of the square
wave. (For sines, the integral and derivative are cosines.) RR and UD will be valuable
examples, one smoother than S W , one less smooth.
First we find formulas for the cosine coefficients a0 and ak . The constant term a0 is
the average value of the function C.x/ :
Z
Z
1
1
a0 D average
a0 D
C.x/ dx D
C.x/ dx:
(11)
0
2
I just integrated every term in the cosine series (10) from 0 to . On the right side, the
integral of a0 is a0 (divide both sides by ). All other integrals are zero :
Z
sin nx
cos nx dx D
D 0 0 D 0:
(12)
n
0
0
In words, the constant function 1 is orthogonal to cos nx over the interval 0; .
The other cosine coefficients ak come from the orthogonality of cosines. As with sines,
we multiply both sides of (10) by cos kx and integrate from 0 to :
Z
Z
Z
Z
C.x/ cos kx dx D
a0 cos kx dxC
a1 cos x cos kx dxCC
ak .cos kx/2 dxC
0
You know what is coming. On the right side, only the highlighted term can be nonzero. For
k > 0, that bold nonzero term is ak =2. Multiply both sides by 2= to find ak :
Cosine coefficients
C. x/ D C.x/
2
ak D
0
1
C.x/ cos kx dx D
437
RR.x/ D jxj
2.x/
2
2.x
2/
Up-down UD.x/
2
2.x C /
2.x
/
Figure 8.3: The repeating ramp RR and the up-down UD (periodic spikes) are even.
The slope of RR is 1 then 1 : odd square wave S W . The next derivative is UD : 2.
Example 2
Find the cosine coefficients of the ramp RR.x/ and the up-down UD.x/.
The simplest way is to start with the sine series for the square wave :
4 sin x
sin 3x
sin 5x
sin 7x
S W .x/ D
C
C
C
C D slope of RR
1
3
5
7
Solution
Take the derivative of every term to produce cosines in the up-down delta function :
4
cos x C cos 3x C cos 5x C cos 7x C :
(14)
Those coefficients dont decay at all. The terms in the series dont approach zero, so
officially the series cannot converge. Nevertheless it is correct and important. At x D 0,
the cosines are all 1 and their sum is C1. At x D , the cosines are all 1. Then
their sum is 1. (The
R downward spike is 2.x /.) The true way to recognize .x/
is by the integral test .x/f .x/ dx D f .0/ and Example 3 will do this.
For the repeating ramp, we integrate the square wave series for S W .x/ and add a0 .
The average ramp height is a0 D =2, halfway from 0 to :
cos x
cos 3x
cos 5x
cos 7x
Ramp series RR.x/ D
C
C
C
C
: (15)
2
4
12
32
52
72
Up-down spikes
UD.x/ D
438
Find the (cosine) coefficients of the delta function .x/, made 2-periodic.
Solution
The spike in .x/ occurs at x D 0. All the integrals are 1, because the
cosine of 0 is 1. We divide by 2 for a0 and by for the other cosine coefficients ak .
Average a0 D
1
2
.x/ dx D
1
2
Cosines ak D
1
.x/ cos kx dx D
1
Then the series for the delta function has all cosines in equal amounts : No decay.
Delta function
.x/ D
1
1
C cos x C cos 2x C cos 3x C :
2
(16)
This series cannot truly converge (its terms dont approach zero). But we can graph the
sum after cos 5x and after cos 10x. Figure 8.4 shows how these partial sums are doing
their best to approach .x/. They oscillate faster while going higher.
There is a neat formula for the sum N that stops at cos N x. Start by writing each term
2 cos x as e ix C e ix . We get a geometric progression from e iN x up to e iN x .
N D
1 h
1 C e ix C e
2
ix
C C e iN x C e
iN x
1 sin.N C 12 /x
:
2
sin 12 x
(17)
height 21=2
5 .x/
height 11=2
height 1=2
height 1=2
Figure 8.4: The sums N .x/ D .1 C 2 cos x C C 2 cos N x/=2 try to approach .x/.
439
1
X
nD1
an cos nx C
1
X
bn sin nx :
(18)
nD1
On every 2 interval the sines and cosines are orthogonal. We find the Fourier
coefficients ak and bk in the usual way: Multiply (18) by 1 and cos kx and sin kx.
Then integrate both sides from to to get a0 and ak and bk .
a0 D
1
2
F .x/ dx ak D
1
F .x/ cos kx dx bk D
1
F .x/ sin kx dx
Orthogonality kills off infinitely many integrals and leaves only the one we want.
Another approach is to split F .x/ D C.x/ C S.x/ into an even part and an odd part.
Then we can use the earlier cosine and sine formulas. The two parts are
C.x/ D Feven .x/ D
F .x/ C F . x/
2
F .x/
F . x/
:
2
(19)
The even part gives the as and the odd part gives the bs. Test on a square pulse from
x D 0 to x D hthis one-sided thin box function is not odd or even.
1=h for 0 < x < h
Example 4 Find the as and bs if F.x/ D tall box D
0
for h < x < 2
The integrals for a0 and ak and bk stop at x D h where F .x/ drops to zero.
The coefficients decay like 1=k because of the jump at x D 0 and the drop at x D h :
Solution
ak D
1
h
h
0
cos kx dx D
a0 D
1
2
sin kh
kh
h
0
1= h dx D
bk D
1
h
1
D average
2
h
0
sin kx dx D
cos kh
:
kh
Important As h approaches zero, the box gets thinner and taller. Its width is h and its
height is 1= h and its area is 1. The box approaches a delta function ! And its Fourier
coefficients approach the coefficients of the delta function as h ! 0 :
a0 D
1
2
ak D
sin kh
1
approaches
kh
bk D
cos kh
approaches 0: (20)
kh
440
R
(21)
The energy in F .x/ equals the energy in the coefficients. The left side is like the length
squared of a vector, except the vector is a function. The right side comes from an infinitely long vector of as and bs. The lengths are equal, which says that thepFourier transp
form from function to vector is like an orthogonal matrix. Normalized by 2 and ,
sines and cosines are an orthonormal basis in function space.
1e
ix
C D
1
X
cn e inx
(22)
nD 1
If every cn D c n , we can combine e i nx with e i nx into 2 cos nx. Then (22) is the
cosine series for an even function. If every cn D c n , we use e i nx e i nx D 2i sin nx.
Then (22) is the sine series for an odd function and the cs are pure imaginary.
To find ck , multiply (22) by e ikx (not e i kx ) and integrate from to :
Z
F .x/e
i kx
dx D
Z
c0 e
i kx
dx C
Z
ix
c1 e e
i kx
dx C C
Z
ck e ikx e
ikx
dx C
The complex exponentials are orthogonal. Every integral on the right side is zero,
except for the highlighted term (when n D k and e i kx e i kx D 1). The integral of 1 is 2.
That surviving term gives the formula for ck :
Fourier coefficients
Z
F.x/e
ikx
dx D 2ck
for k D 0; 1; : : : l
(23)
441
Example 5
Example 6
For a delta function, all integrals are 1 and every ck is 1=2. Flat transform !
1 for s x s C h
Find ck for the 2-periodic shifted box F .x/ D
0 elsewhere in ;
Solution
iks
e i kh
2 i k
Notice above all the simple effect of the shift by s. It modulates each ck by e
The energy is unchanged, the integral of jF j2 just shifts, and je i ks j D 1.
Shift F.x/ to F.x
s/
! Multiply every ck
by e
iks
(24)
i ks
(25)
h
2
ck D e i kh=2
e i kh
1 sin.kh=2/
D
:
2 i k
2
k=2
Divide by h for a tall box. The ratio of sin.kh=2/ to kh=2 is called the sinc of kh=2.
1
1 X
kh
Fcentered
1= h for h=2 x h=2
D
sinc
e i kx D
Tall box
0
elsewhere in ;
h
2 1
2
1
That division by h produces area D 1. Every coefficient approaches 2
as h ! 0.
The Fourier series for the tall thin box again approaches the Fourier series for .x/.
ck e ikx is dF=dx D
ikck e ikx
The second derivative has coefficients .i k/2 ck D k2 ck . High frequencies are growing
stronger. And in the opposite direction (when we integrate), we divide by i k and high
frequencies get weaker. The solution becomes smoother. Please look at this example :
Response 1=.k2 C 1/
to frequency k
d 2y
e i kx
i kx
C
y
D
e
is
solved
by
y.x/
D
dx 2
k2 C 1
This was a typical problem in Chapter 2. The transfer function is 1=.k 2 C 1/. There we
learned : The forcing function e i kx is exponential so the solution is exponential.
442
All we are doing now is superposition. Allow all the exponentials at once !
X
d 2y
C
y
D
ck e i kx
dx 2
is solved by
y.x/ D
X ck e i kx
:
k2 C 1
(26)
1. Derivative rule dF =dx has Fourier coefficients ikck (energy moves to high k).
2. Shift rule F.x
iks
(27)
It remains to choose the constants ak and bk to make u D u0 on the boundary. For a circle,
and C 2 give the same point. This means that u0 ./ is periodic :
Set r D 1
(28)
This is exactly the Fourier series for u0 . The constants ak and bk must be the Fourier
coefficients of u0 ./. Thus Laplaces boundary value problem is completely solved, if
an infinite series (27) is acceptable as the solution.
Example 8 Point source u0 D ./. The boundary is held at u0 D 0, except for the
source at x D 1, y D 0 (where D 0). Find the temperature u.r; / inside the circle.
Delta function
u0 ./ D
1
1
1 X i n
1
C .cos C cos 2 C cos 3 C / D
e
2
2 1
u.r; / D
1
1
C .r cos C r 2 cos 2 C r 3 cos 3 C /
2
(29)
Poisson managed to sum this infinite series ! It involves a series of powers .re i /n .
His sum gives the response at every .r; / to the point source at r D 1, D 0:
Temperature inside circle
u.r; / D
1
1
2 1 C r 2
r2
2r cos
(30)
443
u0 ./ D 1 on the top half of the circle and u0 D 1 on the bottom half.
Example 9
The boundary values u0 are a square wave S W . We know its sine series :
4 sin
sin 3
sin 5
C
C
C
(31)
Square wave for u0 ./ S W ./ D
1
3
5
Solution
(32)
Laplaces equation has smooth solutions inside, even when u0 ./ is not smooth.
(a) To prove that cos nx is orthogonal to cos kx when k n, use the formula
1
.n C k/x C 12 cos .n k/x. Integrate from x D 0
.cos nx/ .cos kx/ D
R 2 cos
2
to x D . What is cos kx dx ?
(b) From 0 to , cos x is not orthogonal to sin x. The period has to be 2 :
Z
Z
Z2
Find .sin x/ .cos x/ dx and
.sin x/ .cos x/ dx and
.sin x/ .cos x/ dx:
(a) f1 .x/ D sin3 x, an odd function (sine series, only two terms)
(b) f2 .x/ D j sin xj, an even function (cosine series)
(c) f3 .x/ D x for x (sine series with jump at x D )
P
Find the complex Fourier series e x D
ck e i kx on the interval x .
1
The even part of a function is 2 .f .x/ C f . x//, so that feven .x/ D feven . x/. Find
the cosine series for feven and the sine series for fodd . Notice the jump at x D .
From the energy formula (21), the square wave sine coefficients satisfy
Z
Z
2
2
2
.b1 C b2 C / D
jS W .x/j dx D
1 dx D 2:
1
9
< jxj < ;
2
1
25
C /.
444
7
Plot the first three partial sums and the function x. x/ :
8 sin x
sin 3x
sin 5x
C
C
C ; 0 < x < :
x. x/ D
1
27
125
Why is 1=k 3 the decay rate for this function? What is its second derivative?
Sketch the 2-periodic half wave with f .x/ D sin x for 0 < x < and f .x/ D 0
for < x < 0. Find its Fourier series.
Suppose G.x/ has period 2L instead of 2. Then G.x C 2L/ D G.x/. Integrals
go from L to L or from 0 to 2L. The Fourier formulas change by a factor =L :
ZL
1
P
1
i kx=L
The coefficients in G.x/ D
C ke
are C k D
G.x/e i kx=L dx:
2L
1
L
Derive this formula for Ck : Multiply the first equation for G.x/ by
integrate both sides. Why is the integral on the right side equal to 2LCk ?
10
For Geven , use Problem 9 to find the cosine coefficient Ak from .Ck C C
Geven.x/ D
Geven is
1
P
0
kx
Ak cos
L
1
.G.x/
2
1
has Ak D
L
ZL
Geven.x/ cos
and
k /=2 :
kx
dx:
L
1
.ck C c k / on the usual interval from 0 to .
2
Find a similar formula for bk from
Pck and c k . In the reverse direction, find the
complex coefficient ck in F .x/ D ck e i kx from the real coefficients ak and bk .
11
12
Find the solution to Laplaces equation with u0 D on the boundary. Why is this
the imaginary part of 2.z z 2 =2 C z 3 =3 / D 2 log.1 C z/? Confirm that on the
unit circle z D e i , the imaginary part of 2 log.1 C z/ agrees with .
13
If the boundary condition for Laplaces equation is u0 D 1 for 0 < < and
u0 D 0 for < < 0, find the Fourier series solution u.r; / inside the unit
circle. What is u at the origin r D 0 ?
14
15
(a) Verify that the fraction in Poissons formula (30) satisfies Laplaces equation.
16
17
445
18
19
If f .x/ D 1 for jxj =2 and f .x/ D 0 for =2 < jxj < , find its cosine
coefficients. Can you graph and compute the Gibbs overshoot at the jumps ?
20
21
22
23