CHAPTER 12

ORTHOGONAL FUNCTIONS AND FOURIER SERIES

(12.1~12.6)

1
Chapter Contents
12.1 Orthogonal Functions

12.2 Fourier Series

12.3 Fourier Cosine and Sine Series

12.4 Complex Fourier Series

12.5 Strum-Liouville Problems

12.6 Bessel and Legendre Series

2
 12.1 Orthogonal Functions
For functions on an interval [a, b]
b
Inner Product of Functions: ( f1 , f 2 ) = ∫a f1 ( x) f 2 ( x)dx
b
=
Orthogonal Functions: ∫=
( f1 , f 2 )f ( x) f ( x) dx 0
a 1 2

2 b b
a function: φ ( x) ∫= φ dx and φ ( x) ∫ φn2 ( x)dx
2
Square norm of = n n n
a a

Orthogonal Set: {φ0(x), φ1(x), φ2(x), …} if

b
(φm ,=
φn ) ∫
a
φm ( x), φn ( x=
) dx 0, m ≠ n

Orthonormal set: {φn(x)} is an orthogonal set and ||φn(x)|| = 1

3
 {φn(x)}: orthogonal
Orthogonal Series Expansion ∞ set on [a, b].
 f (= x) c0φ0 ( x) + c1φ1 ( x) +  + cnφn ( x) +  or f ( x) = ∑ cnφn ( x),
n =0
b b b

∫a m ∫ m ∫
f ( x)φ ( x) dx = c φ ( x)φ ( x) dx + c φ ( x)φ ( x) dx + 
0 a 0 1 a 1 m

c0 (φ0 , φm ) + c1 (φ1 , φm ) +  + cn (φn , φm ) +  All zeros except m = n

b
Thus,
cn =
∫ a
f ( x)φn ( x) dx or
f ( x) = ∑
∞
( f , φn )
φ ( x)
2 n
|| φn ( x) || n =0 || φn ( x ) ||
2
b
Orthogonal w.r.t. a weight function w(x): w( x)φm ( x)φn ( x=
b ∫a )dx 0, m ≠ n

cn =
∫a
f ( x) w( x)φn ( x) dx
2
b
|| φn ( x) || = ∫ w( x)φn ( x) dx
2

|| φn ( x) ||2 a

Complete orthogonal set:

If the only function orthogonal to each term is the zero function.
4
 12.2 Fourier Series Use this:

 π 2π 3π π 2π 3π 
1, cos p x, cos p x, cos p x, , sin p x, sin p x, sin p x,  is orthogonal on [−p, p].
 
Function:
a0 ∞  nπ nπ 
f ( x) = + ∑  an cos x + bn sin x 1 p
2 n=1  p p  a0 = ∫ f ( x)dx
0 (orthogonality) 0
p −p
p a0 p ∞
 p nπ p nπ 
 𝑓𝑓 𝑥𝑥 , 1 = ∫− p f ( x )dx =
2 ∫ − p
dx + ∑ 
n =1 
an ∫− p cos
p
xdx + bn ∫− p sin
p
xdx  = pa0

𝑚𝑚𝜋𝜋 p mπ 1 nπ
 𝑓𝑓 𝑥𝑥 , cos 𝑥𝑥 = ∫− p f ( x )cos xdx  0, m ≠ n
p
an = ∫ f ( x)cos x dx
𝑝𝑝 p =
p − p p
0  p, m = n 0
a0 p mπ ∞
 p mπ nπ p mπ nπ 
2 ∫− p
= cos
p
xdx + ∑  n ∫− p
n =1 
a cos
p
x cos
p
x dx + bn ∫− p cos
p
x sin
p
x dx  = an p

𝑚𝑚𝜋𝜋 1 p nπ
Similarly, from 𝑓𝑓 𝑥𝑥 , sin 𝑥𝑥  bn = ∫ f ( x)sin xdx
𝑝𝑝 p −p p
5
 Fourier Series
a0 ∞ nπ nπ
 Fourier series of f(x) on (−p, p): f ( x) = + ∑ (an cos x + bn sin x)
2 n=1 p p
1 p 1 p nπ 1 p nπ
a0 = ∫ f ( x) dx an = ∫ f ( x) cos x dx bn = ∫ f ( x) sin x dx
p −p p −p p p −p p
EX1: Expand 0, −π < x < 0
f ( x) =  in a Fourier series.
π − x, 0≤ x <π
Solution:
1 π
a0 = ∫
π π
f ( x) dx
−
π
1 0 π
 1 x  π 2
=  ∫
π  −π
0dx + ∫ (π − x) dx  = π x −  =
0  π 2 0 2
6
 Example 1
1 π 1 π 1
an = ∫
π π
f ( x)cos nx dx
−
bn =∫ (π − x)sin nxdx =
π 0 n
1 0 π

π  ∫−π ∫0
= 0 dx + (π − x )cos nx dx  f ( x) =
π
1  sin nx 1 π 
= (π − x) + ∫ sin nx dx  π 1 − (−1) n
∞
1 
π n 0 n 0 + ∑ cos nx + sin nx 
 4 n=1  n π 2
n 
π
1 cos nx  − cos nπ + 1
=
− = ←cos nπ = (–1) n
nπ n  0 n 2π
1 − (−1) n
=
n 2π
7
 Convergence of a Fourier Series
At points of continuity, Fourier series of f converge to f(x).
At points of discontinuity, the Fourier series converge to f ( x + ) + f ( x −)
2
Periodic extension of the f in Example 1:
Partial Sums:
Series of 15 terms

converge to converge to
f (π −) + f (−π + ) f (0+ ) + f (0−) π
=0 =
2 2 2
Gibbs phenomenon: Spikes near the discontinuities;
remains constant even when N is large.
8
 12.3 Fourier Cosine and Sine Series
Even and Odd Functions
even odd
Properties: f(−x) = f(x) f(−x) = −f(x)
(a) even*even=even
(b) odd*odd=even
(c) even*odd=odd
(d) even +/- even = even f is even
(e) odd +/- odd = odd a a  Fourier series become…
(f) If f is even then a ∫− a f ( x)dx = 2 ∫0 f =( x)dx 1 p 2 p
(g) If f is odd then ∫− a f ( x)dx = 0
a0 =
p ∫− p
f ( x)dx
p ∫0
f ( x)dx

1 p nπ 2 p = nπ 1 p nπ
an
p ∫− p
= f ( x )cos
p
x dx
p ∫0
f ( x )cos
p
xdx bn =∫
p −p
f ( x)sin x dx 0
p
      
even odd
9
 Fourier Cosine and Sine Series
Fourier Cosine Series (even function f on (−p, p) )
a0 ∞ nπ 2 p 2 p nπ
f ( x=
) + ∑ an cos x a0 = ∫ f ( x) dx an = ∫ f ( x)cos x dx
2 n=1 p p 0 p 0 p
Fourier Sine Series (odd function f on (−p, p) )
∞
nπ 2 p nπ
f ( x) = ∑ bn sin x bn = ∫ f ( x)sin x dx
i =1 p p 0 p
Half-Range Expansions (arbitrary function defined on −L < x < 0)

Use Fourier Use Fourier Periodic

Cosine Sine extension
10
 Example 3: Expansion in Three Series
Expand f(x) = x2, 0 < x < L, (a) in a cosine series, (b) in a sine series (c) in a
Fourier series.
Solution:
2 L 2 2 2 L nπ 4 L2
( −1) n
(a) ∫ ∫
2
= a0 = = x 2
dx L , a n = x cos x dx
L 0 3 L 0 L n π
2 2

L2 4 L2 ∞ (−1) n nπ
f ( x=
) + 2 ∑ 2 cos x
3 π n=1 n L
(b) 2 L 2 nπ 2 L2 (−1) n+1 4 L2
∫
n
= bn x sin = x dx + [( −1) − 1]
L 0 L nπ nπ
3 3

2 L2 ∞ (−1) n+1 2  nπ
f ( x) = ∑ 
π n=1  n nπ
n
+ 3 2 [(−1) − 1] sin
L
x
 11
 Example 3
(c) With p = L/2, nπ/p = 2nπ/L

2 L 2 2 2
a0 =∫
L 0
x dx
3
L

2 L 2 2nπ L2
bn = ∫ x sin xdx = −
L 0 L nπ
2 L 2 2nπ L2
an =
L ∫0
x cos
L
xdx
n 2π 2

L2 L2 ∞  1 2nπ 1 2nπ 
f ( x) = + ∑  2 cos x − sin x
3 π n=1  n π L n L 
12
 12.4 Complex Fourier Series Euler’s Formula
a ∞
nπ nπ eix + e − ix
Fourier series f ( x) = 0
+ ∑ (an cos x + bn sin x) cos x =
2 n=1 p p 2
a0 ∞  einπ x/p + e − inπ x/p einπ x/p + e − inπ x/p  eix − e − ix
= + ∑  an + bn sin x =
2 n=1  2 2i  2i

a0 ∞  1 1 −inπ x/p 
= + ∑  (an − ibn )e inπ x/p
+ (an + ibn )e 
c0 2 n=1  2 cn 2 c-n
1 1 p 1
c0 = ． ∫ f ( x) dx c−n = (an + ibn )
2 p −p 2
1 1 p  nπ nπ 
cn = (an − ibn )
2 =e −inπ x / p =
2 p −p ∫ f ( x) cos
 p
x + i sin  dx
p
1 p  nπ nπ  1 p
= ∫
2 p −p
f ( x )

cos
p
x − i sin
p
x 

dx =
2 ∫− p
f ( x ) e inπ x / p
dx
13
 Complex Fourier Series
Complex Fourier Series - f defined on an interval (−p, p)
∞
1 p
∑ce ∫
− inπ x / p
f ( x) = n
inπ x / p
cn = f ( x ) e dx, n = 0, ± 1, ± 2, 
n =−∞ 2p −p
f ( x + ) + f ( x −)
Convergence: converges to f(x) at a point of continuity and to
at a point of discontinuity. ∞
2
Fundamental period: T = 2p. Fourier series  f ( x) = ∑ n
inω x
c e
n =−∞
a0 ∞
= + ∑ (an cos nω x + bn sin nω x)
2 n=1
Fundamental angular frequency: ω = 2π/T
If f is periodic and has fundamental period T, the plot of the points (nω,
|cn|) is called the frequency spectrum of f.
14
 Example 1&2: Complex Fourier Series
Expand f(x) = e–x, −π < x <π, in a complex Fourier series.
Solution: 1 π − x − inx 1 π − ( in+1) x 1
∫ ∫
− ( in +1)π ( in +1)π
= c
with p = π, n = e e dx e dx = − 
 e − e 

2π −π 2π −π 2π (in + 1)
π −π
n ( e − e ) n sinh π 1 − in π
=e −π
(cos nπ − i sin n π ) = ( −1) n −π
e
 cn = (−1) = (−1)
2(in + 1)π π n2 + 1 = eπ (cos nπ + i sin nπ ) = (−1) n eπ
sinh π ∞
1 − in inx (converges to the 2π-
 f ( x)
=
π
∑
n =−∞
n
(−1) 2 e
n +1 periodic extension of f )

Since ω = 1, nω = 0, ±1, ±2, … and frequency

sinh π 1 spectrum 
| cn | =
π n2 + 1 15
 Example 3: Frequency Spectrum
Find the spectrum of the wave shown in Fig12.4.2. The wave is the
periodic extension of the function f:
 0, − 12 < x < − 14

f ( x) = 1, − 14 < x < 14
 0, 1 < x< 1
 4 2
Solution:
T = 1 = 2p  p = ½. check that 14 1
= c0 ∫= dx
12 14
−1 4 2
cn ∫ f ( x)e =
= 2 inπ x
dx ∫ 1 ⋅ e dx 2 inπ x
−1 2 −1 4

2 inπ x 1/ 4
e  1 einπ / 2 − e − inπ / 2 1 nπ
=  = sin
2inπ  −1/ 4 nπ 2i nπ 2
16
 12.5 Sturm-Liouville Problem
Solve the homogeneous boundary-value problem
y′′ + λ=
y 0, y (0)
= 0, y ( L=) 0
Solution:
Case I: λ = 0 y = c1x + c2, y(0) = c2 = 0, y(L) = c1L = 0, c1 = 0
then y = 0, trivial solution.
Case II: λ < 0, λ = −α2, α > 0 Choose y = c1 cosh αx + c2 sinh αx,
y(0) = 0, c1 = 0; y(L) = 0, c2 = 0 then y = 0, trivial solution.
Case III: λ > 0, λ = α2, α > 0 Choose y = c1 cos αx + c2 sin αx
y(0) = 0, c1 = 0; y(L) = 0, c2 sin αL= 0, if c2 = 0, y = 0, trivial solution.
So c2 ≠ 0, sin αL = 0, αL = nπ, α = nπ/L y = c2 sin (nπx/L) are solutions.
Eigenvalues: λn = n2π2/L2, n = 1, 2, 3,….
Eigenfunctions: y = sin (nπx/L)
17
 Self-Adjoint Form
 L is a linear differential operator
 Adjoint differential operator: (Lu, v) = (u, 𝐿𝐿∗ v)
𝑏𝑏 𝑏𝑏
or ∫𝑎𝑎 𝐿𝐿𝐿𝐿 𝑣𝑣𝑣𝑣𝑣𝑣 = ∫𝑎𝑎 𝑢𝑢 𝐿𝐿∗ 𝑣𝑣 𝑑𝑑𝑑𝑑
 Self-adjoint differential operator: (Lu, v) = (u, Lv) (say, 𝐿𝐿 = 𝐿𝐿∗ )
 For 2nd-order differential equation, 𝑎𝑎 𝑥𝑥 𝑦𝑦 ′′ + 𝑏𝑏 𝑥𝑥 𝑦𝑦 ′ + 𝑐𝑐 𝑥𝑥 𝑦𝑦 = 0 or
𝑑𝑑 2 𝑑𝑑
𝐿𝐿𝑦𝑦 = 0 where 𝐿𝐿 = 𝑎𝑎 𝑥𝑥 + 𝑏𝑏 𝑥𝑥 + 𝑐𝑐 𝑥𝑥
𝑑𝑑𝑥𝑥 2 𝑑𝑑𝑑𝑑
𝑑𝑑 𝑑𝑑
 If 𝑏𝑏 𝑥𝑥 = 𝑎𝑎′ 𝑥𝑥 , 𝐿𝐿 = 𝑎𝑎 𝑥𝑥 + 𝑐𝑐 𝑥𝑥 which is a self-adjoint operator
𝑑𝑑𝑑𝑑 𝑑𝑑𝑑𝑑
𝑒𝑒 ∫ 𝑏𝑏(𝑥𝑥)/𝑎𝑎 𝑥𝑥 𝑑𝑑𝑑𝑑
 If L is not a self-adjoint operator, the integrating factor 𝐼𝐼(𝑥𝑥) =
𝑎𝑎 𝑥𝑥
can be always found to let 𝐼𝐼 𝑥𝑥 𝐿𝐿 becomes a self-adjoint operator.
18
 Sturm-Liouville Equation
Sturm-Liouville Equation: 𝐿𝐿𝑦𝑦 𝑥𝑥 = −λ𝑝𝑝 𝑥𝑥 𝑦𝑦(𝑥𝑥)
𝑑𝑑 𝑑𝑑
where 𝐿𝐿 = 𝑟𝑟 𝑥𝑥 + 𝑞𝑞 𝑥𝑥 is a self-adjoint operator; 𝑝𝑝 𝑥𝑥 is a weighting
𝑑𝑑𝑑𝑑 𝑑𝑑𝑑𝑑
function. Or, d
[r ( x) y′] + (q ( x) + λ p ( x)) y =
0
dx
 p, q, r and r′ are real-valued functions continuous on [a, b]
 r(x) > 0 and p(x) > 0 for every x in [a, b]
 λ is the eigenvalue; the solutions y are the eigenfunctions.
 If 𝑢𝑢 𝑥𝑥 , 𝑣𝑣 𝑥𝑥 are eigenfunctions and (Lu, v) = (u, Lv), the system is called
Sturm-Liouville boundary value problem.

19
 Sturm-Liouville Boundary Value Problem
Boundary Conditions:
 Regular A1 y (a ) + B1 y′(a ) =
0
A2 y (b) + B2 y′(b) = 0
 Periodic r(a) = r(b), and y(a) = y(b); y′(a) = y′(b)
 Singular (1) r(a) = 0 and no B.C. at x=a; A2 y (b) + B2 y′(b) = 0
(2) r(b) = 0 and no B.C. at x=b; A1 y ( a ) + B1 y′( a ) =
0
(3) r(a) = r(b) = 0 and no B.C. at x=a or b;
Properties:
(a) Infinite number of real eigenvalues (d) Eigenfunctions are
(b) One eigenfunction for each eigenvalue orthogonal through p(x) on [a, b].
b
(c) Eigenfunctions are linearly independent. ∫a p ( x) ym ( x) yn (=
x) dx 0, λm ≠ λn
20
 Proof of Orthogonality
Eigenfunctions: ym and yn d
[r ( x) ym′ ] + (q ( x) + λm p ( x)) ym = 0
Eigenvalues: λm and λn.  dx
d
[r ( x) yn′ ] + (q ( x) + λn p ( x)) yn = 0
dx
d d
Subtraction of equations  (λm − λn ) p ( x) ym yn = ym [ r ( x) yn ] − yn [ r ( x) ym′ ]
′
b dx dx
Integration (λm − λn ) ∫ p ( x) ym yn dx
a

= r (b)[ ym (b) yn′ (b) − yn (b) ym′ (b)] − r (a )[ ym (a ) yn′ (a ) − yn (a ) ym′ (a )]

A1 ym (a ) + B1 ym′ (a ) =0
B.C. for x=a since A1 and B1 not both zero
A1 yn (a ) + B1 yn′ (a ) =
0 ym (a ) yn′ (a ) − yn (a ) ym′ (a ) =
0
b
0  ∫ p ( x) ym ( x) yn (=
B.C. for x=b ym (b) yn′ (b) − yn (b) ym′ (b) = x) dx 0, λm ≠ λn
a
21
 Example 2: Regular S-L Problem
Solve y′′ + λ = = 0, y (1) + y′(1)
y 0, y (0) = 0
Solution:
λ = 0, λ < 0  trivial solution.
λ = α2 > 0, α > 0,  y = c1 cos αx + c2 sin αx.
B.C.: y(0) = 0  c1 = 0, thus y = c2 sin αx.
y(1) + y′(1) = 0  c2 sin α + c2α cosα = 0. Choosing c2 ≠ 0,
 tanα = −α
 Infinitely many solution for α > 0.
Eigenvalues: λn = αn2, n = 1, 2, 3, …
Eigenfunctions: yn = sin αnx.

22
 Integrating Factor
 Convert a ( x) y′′ + b( x) y′ + (c( x) + λ d ( x)) y =
0 into self-adjoint form
We want µ𝑏𝑏 𝑥𝑥 = µ𝑎𝑎 𝑥𝑥 ′
𝑏𝑏(𝑥𝑥) 𝑏𝑏(𝑥𝑥)
From a (x)y′ + b(x)y = 0, we need μ′ = μ so that µy′ + µ y=0
𝑎𝑎(𝑥𝑥) 𝑎𝑎(𝑥𝑥)
 integrating factor: µ = e∫ b(x)/a(x)dx
Let Y = y′  e ∫ b ( x )/ a ( x ) dxY ′ + b( x) e ∫ b= ( x )/ a ( x ) dx
Y + ...
d  ∫ b ( x )/ a ( x ) dx 
e Y + ...
a( x) dx  
∫ ( b / a ) dx b ( x ) ∫ ( b / a ) dx  c ( x ) ∫ ( b / a ) dx d ( x ) ∫ ( b / a ) dx 
 e y′′ + e y′ +  e +λ e  y= 0
a( x)  a( x) a( x) 
or d  ∫ ( b / a ) dx   c( x) ∫ ( b / a ) dx d ( x) ∫ ( b / a ) dx 
e y′ +  e +λ e  0
y=

dx  
  a( x) a( x) 
23
 Example 3: Parametric Bessel Equation
2
′′ ′
Parametric Bessel DE: x y + xy + (α x − n ) y = 0, n = 0, 1, 2, ...
2 2 2

=
general solution: y c1 J n (α x) + c2Yn (α x)
Integrating factor: e∫ (1/x) dx = eln x = x, x > 0
 2 n2  d  2 n2 
′′ ′
 xy + y +  α x −=  y 0 or [ xy ] +  α x −=
′  y 0 𝑌𝑌𝑛𝑛 0 = −∞
 x dx  x
where r = x, q = x, λ =
− n 2 /x , p = α2 eigenvalues
B.C. for singular S-L BVP: r(0) = 0, λi = αi2
no B.C. at x=a, but y still should be bounded at x=a  𝑐𝑐2 = 0
 set {Jn(αix)}, i = 1, 2, 3, …, is orthogonal through p(x) = x on [0, b].
 b xJ (α x) J (α x)=
∫0 n i n j dx 0, λ ≠ λ , B.C. at x=b: A J (αb) + B αJ ′(αb) = 0
i j 2 n 2 n
24
 Example 4: Legendre’s Equation
Re-write the Legendre’s differential equation (1 − x 2
) y′′ − 2 xy′ + n(n + 1) y =
0
as d
[(1 − x 2 ) y′] + n(n + 1) y =
0
dx

 q(x) = 0, p(x) = 1 and λ = n(n+1).

Legendre’s polynomials: Pn(x) when n = 0, 1, 2, …
 r(−1) = r(1) = 0
 Pn(x) are the only solutions bounded on [−1, 1],
 set {Pn(x)}, n = 0, 1, 2, …, is orthogonal w.r.t. the
weight function p(x) = 1 on [−1, 1].
Thus, 1
∫ Pm ( x) Pn ( x=
−1
)dx 0, m ≠ n
25
 12.6 Bessel and Legendre Series
Self-adjoint form
Parametric Bessel Equation d  2 n  2

 x y′′ + xy′ + (α x − n ) y = 0, n = 0, 1, 2, ... or dx [ xy ] +  α x − x  y =

′ 0
2 2 2 2
 
Set {Jn(αix)}, i = 1, 2, 3, … is orthogonal w.r.t. p(x) = x on [0, b]
B.C.: A2 J n (α b) + B2α J n′ (α b) =
0
Fourier-Bessel Series
Use {Jn(αix)} as a basis to expand a function
Generalized Fourier series of a function f on [0, b]
b
∞
f ( x) = ∑ ci J n (α i x)
∫ xJ n (α i x) f ( x) dx 2 b
J n (α i x) = ∫ xJ n2 (α i x)dx
0
ci = 2
i =1 J n (α i x) 0
coefficients square norm
26
 Square Norm
d 2 d
[ xy ] + (α x − n ) [ y ]2 =
2
′ 2 2
0
Self-adjoint form × 2xy′  dx dx
( )
b b
Integrating by parts on [0, b]  2α 2
∫ xy
0
2
dx = [ ]
xy ′
2
+ (α 2 2
x − n 2
) y 2
0
= 𝑏𝑏𝑏𝑏𝑏 2 + 𝛼𝛼 2 𝑏𝑏 2 − 𝑛𝑛2 𝑦𝑦 2 +𝑛𝑛2 𝑦𝑦 2
Substitute y = Jn(αx), y′ = αJn′(αx)
x=0: y(0)= Jn(0)=0 for n>0
b
 2α 2 xJ 2 (α
∫0 n = ′
x) dx α b [ J n (α b)] + (α b − n )[ J n (α b)] ,
2 2 2 2 2 2 2

2
𝑛𝑛
= J n (α i x) = 𝐽𝐽𝑛𝑛 𝛼𝛼𝑏𝑏 − 𝐽𝐽𝑛𝑛+1 𝛼𝛼𝑏𝑏
𝑥𝑥
*Using the recurrence relation:
xJ= ′
v ( x) vJ v ( x) − xJ v +1 ( x) (EX6 in Ch5.3.1)
𝑥𝑥𝐽𝐽′ ν 𝑥𝑥 = −ν𝐽𝐽ν 𝑥𝑥 + 𝑥𝑥𝐽𝐽ν−1 𝑥𝑥
(Prob27 in Ch5.3.1)
27
 Fourier-Bessel Series square norm
b 2

 Case I: Choose A2 = 1, B2 = 0, B.C. J n (α b) = 0  || J nα i ( x) ||2 = J n2+1 (α ib)

2
Infinite positive eigenvalues: λi = αi2 = xi2/b2.
No negative eigenvalues since Jn(−x) = (−1)nJn(x).
0 is not an eigenvalue for any n since Jn(0) = 0, n= 1, 2, 3, … and J0(0) = 1.
 Case II: Choose A2 = h > 0, B2 = b, B.C. hJ n (α b) + α bJ n′ (α b) = 0
square norm  || J (α x) ||2 = α i b − n + h J 2 (α b)
2 2 2 2

n i n i
Infinite positive eigenvalues: λi = αi2 = xi2/b2. 2α i2
0 is not an eigenvalue for n = 1, 2, 3, ….
 Case III: Choose A2 = 0, B2 = b, n = 0 B.C. J 0′ (α b) = 0
J1(αb) = -J0′(αb) = 0  0 is an eigenvalue 2
b b
∫
2
For αi = 0  J0(0) = 1 is a nontrivial function 2 square norm:= ||1|| = x dx
b 2 0 2
For αi > 0  square norm: || J 0 (α i x) || =
2
J 0 (α ib)
2 28
 Fourier-Bessel Series Converges to f(x) at continuities
f ( x + ) + f ( x −)
 Case I: αi are defined by Jn(αb) = 0 Converge to 2 at discontinuities
∞ 2 b
f ( x) = ∑ ci J n (α i x) ci = 2 2 ∫
b J n+1 (α ib) 0
xJ n (α i x) f ( x) dx
i =1

 Case II: αi are defined by hJn(αb) + αbJ’n(αb) = 0

∞
2α i2 b
f ( x) = ∑ ci J n (α i x) ci = 2 2 ∫ xJ n (α i x) f ( x) dx
i =1 (α i b − n + h ) J n (α ib) 0
2 2 2

 Case III: αi are defined by J′0(αb) = 0 2 b

c1 = 2 ∫ x f ( x) dx
∞ b 0
f ( x)= c1 + ∑ ci J 0 (α i x) 2 b
i =2
ci = 2 2 ∫
b J 0 (α ib) 0
xJ 0 (α i x) f ( x) dx
29
 Example: Vibrations of a circular
membrane
https://en.wikipedia.org/wiki/Vibrations_of_a_circular_membrane

30
 Example 1
Expand f(x) = x, 0 < x < 3, in a Fourier-Bessel series, using Bessel function
of order one satisfying the boundary condition J1(3α) = 0.
Solution: 2 3
∫ J1 (α i x)dx
2
c
Case I, b = 3: i = x
3 J 2 (3α i ) 0
2 2

Let t = αi x, dx = dt/αi, x2 = t2/αi2, and use 𝑡𝑡𝐽𝐽′ ν 𝑡𝑡 = −ν𝐽𝐽ν 𝑡𝑡 + 𝑡𝑡𝐽𝐽ν−1 𝑡𝑡

× 𝑡𝑡 ν−1 on both sides and let ν=2  d[t2J2(t)]/dt = t2J1(t):
2 3α i d 2
∫
2
ci = [t J 2 (t )]dt
9α i J 2 (3α i ) 0 dt
3 2
α i J 2 (3α i )
∞
2
 f ( x ) = 2∑ J1 (α i x)
i =1 α i J 2 (3α i )
31
 Example 2
If the αi in Example 1 are defined by
J1(3α) + αJ1′(3α) = 0
Solution:
The only thing changing is the square norm.
Case II, h = 3, b = 3 and n = 1. Thus
18α i J 2 (3α i )
ci =
(9α i2 + 8) J12 (3α i )

∞
α i J 2 (3α i )
f ( x) = 18∑ J1 (α i x)
i =1 (9α i + 8) J1 (3α i )
2 2

32
Fourier–Legendre Series Converges to f(x) at continuities
f ( x + ) + f ( x −)
Converge to 2 at discontinuities

Fourier-Legendre series of f on (–1, 1)

∞
2n + 1 1
f ( x) = ∑ cn Pn ( x) cn = ∫ f ( x) Pn ( x) dx
n =0 2 −1
Let x = cos θ : x = 1  θ = 0, x = −1  θ = π. Since dx = −sin θ dθ,
f(cos θ) F(θ):
∞
2n + 1 π
F (θ ) = ∑ cn Pn (cosθ ) cn = ∫ F (θ ) Pn (cosθ )sin θ dθ
n =0 2 0

33
 Example: Spherical Harmonics
https://en.wikipedia.org/wiki/Sph
erical_harmonics
https://www.theochem.ru.nl/~pw
ormer/Knowino/knowino.org/wik
i/Spherical_harmonics.html
https://www.youtube.com/watch?
v=2yprbHra3yU&ab_channel=Q
uantumVisions%28WWUM%C3
%BCnster%29

34
Solution: Example 3
1 1 1 1 1 Write out the first four nonzero
c0 = ∫
2 −1
f ( x) P0 ( x)dx = ∫
2 0
1 ⋅1dx =
2 terms in the Fourier-Legendre
3 1 3 1 3 expansion of
c1 = ∫ f ( x) P1 ( x)dx = ∫ 1 ⋅ xdx = 0, − 1 < x < 0
2 −1 2 0 4 f ( x) = 
5 1 5 1 1 2 1, 0 ≤ x < 1
c2 = ∫ f ( x) P2 ( x)dx = ∫ 1 ⋅ (3 x − 1)dx = 0
2 −1 2 0 2
7 1 7 1 1 3 7
c3 = ∫
2 −1
f ( x) P3 ( x)dx =∫ 1 ⋅ (5 x − 3 x)dx =
2 0 2
−
16
9 1 9 1 1
c4 = ∫ f ( x) P4 ( x)dx = ∫ 1 ⋅ (35 x 4 − 30 x 2 + 3)dx = 0
2 −1 2 0 8
11 1 11 1 1 11
c5 = ∫ f ( x) P5 ( x)dx = ∫ 1 ⋅ (63 x − 70 x + 15 x)dx =
5 3

2 −1 2 0 8 32 35
 Example 3
1 3 7 11 65
S5 ( x) = P0 ( x) + P1 ( x) − P3 ( x) + P5 ( x) − P7 ( x)
2 4 16 32 256
See Fig 12.6.2.

36
 Generalized Fourier Series
A function f on an interval can be expanded by a complete orthogonal set
(inner product w.r.t. the weighting function p(x))
𝑛𝑛𝜋𝜋𝑥𝑥 ∞
Fourier Cosine Series: Set 1, cos , interval (−p, p), 𝑝𝑝 𝑥𝑥 = 1
𝑝𝑝 𝑛𝑛=1
𝑛𝑛𝜋𝜋𝑥𝑥 ∞
Fourier Sine Series: Set sin , interval (−p, p), 𝑝𝑝 𝑥𝑥 = 1
𝑝𝑝 𝑛𝑛=1
𝑛𝑛𝜋𝜋𝑥𝑥 𝑛𝑛𝜋𝜋𝑥𝑥 ∞
Fourier series: Set 1, cos , sin , interval (−p, p), 𝑝𝑝 𝑥𝑥 = 1
𝑝𝑝 𝑝𝑝 𝑛𝑛=1
∞
Fourier-Bessel series: Set 𝐽𝐽𝑛𝑛 (𝛼𝛼𝑖𝑖 𝑥𝑥) 𝑖𝑖=1 , interval (0, b), 𝑝𝑝 𝑥𝑥 = 𝑥𝑥
∞
Fourier-Legendre series: Set 𝑃𝑃𝑛𝑛 (𝑥𝑥) 𝑛𝑛=0 , interval (-1, 1), 𝑝𝑝 𝑥𝑥 = 1

37