u  u

Sturm-Liouville Theory

1
Sturm-Liouville Problem
1) Linear equations 2) Linear equations
Introduction:
brief review of some of the
ordinary differential equations General solutions General solutions

3) Linear equations 4) Cauchy-Euler equation

General solutions General solutions

5) Parametric Bessel equation 6) Legendre's equation

General solutions Particular solutions

Relevance: Sturm-Liouville Problem
• The standard series solution or other methods do not work when
boundary conditions are exist (instead of initial conditions)

General B.Cs

OR

• Applications in beam deflection, heat/mass diffusion, steady state

solutions, etc where physical boundaries exist.

Understand Interesting
Extend to PDEs
ODEs with B.Cs Special functions

Dirichlet, Neumann Legendre, Bessel, Heat, Wave, Laplace

and mixed conditions Chebyschev, functions equations
Eigenvalues and Eigenfunctions

Matrix Example 2 1 
A 
 1 2 
1
1  1, v1   
 1

1
Boundary-Value Problem 1  3, v1   
1
Eigenvalues and Eigenfunctions

1) Boundary-Value Problem Example

4 2 𝒚  sin( 2 x )
 2 L
L

Example
4
is not an eigenvalue

Remark: This BVP has infinite number of eigenvalues

n 2 2  sin( nL x )
 2
L
Eigenvalues and Eigenfunctions
1) Boundary-Value Problem 2) Boundary-Value Problem

n 2 2
 2
n 2 2 L
 2
L  cos( nL x )
 sin( nL x )
𝝀 = 0 is an eigenvalue with
y = 1 as an eigenfunction

is the orthogonal set of functions This set, which is orthogonal

on the interval [0, L] used as the on [O, L], is the basis for the
basis for the Fourier sine series. Fourier cosine series.
Regular Sturm-Liouville Problem
Remark:
The problems in BVP(1) and BVP(2) are special cases
of an important general boundary-value problem

Regular Sturm-Liouville problem

Solve:

Subject to:

BVP 1 BVP 2
Regular Sturm-Liouville Problem
Regular Sturm-Liouville problem

Solve:

Subject to:

Remark:
5) BVP is homogeneous if it consists of a
homogeneous linear differential equation
and homogeneous boundary conditions
6) The boundary conditions (2) and (3) are
3) The differential equation (1) separated (each involves only a single
is linear and homogeneous boundary point). Otherwise, mixed.
4) The boundary conditions in
7) Regular Sturm-Liouville BVP always
( 2) and ( 3) both linear and
possesses the trivial solution y = 0.
homogeneous
Regular Sturm-Liouville Problem
Regular Sturm-Liouville problem

Solve:
Subject to:

Theorem 12.5.1 Properties of the Regular Sturm-Liouville Problem

(a) There exist an infinite number of real eigenvalues that can be arranged in
increasing order
such that
(b) For each eigenvalue there is only one eigenfunction (except for nonzero
constant multiples).

(c) Eigenfunctions corresponding to different eigenvalues are linearly independent.

(d) The set of eigenfunctions corresponding to the set of eigenvalues is
orthogonal with respect to the weight function p(x) on the interval [a, b ].
Regular Sturm-Liouville Problem
Regular Sturm-Liouville problem Orthogonality Relation
Solve:
Subject to:

Theorem 12.5.1
Orthogonality Relation
(d) The set of eigenfunctions corresponding
to the set of eigenvalues is orthogonal with
respect to the weight function p(x) on the
interval [a, b ].
proof: The inner product w.r.t
weight p is zero

b
(m  n )  p( x) ym yn dx  r (b)[ ym (b) yn ' (b)  yn (b) ym ' (b)]
a
 r (a )[ ym (a ) yn ' (a )  yn (a ) ym ' ( a )]
Regular Sturm-Liouville Problem
Regular Sturm-Liouville problem
Solve:
Subject to:

Example
Solve the boundary-
value problem

Positive roots

Remark: sin(  n x)n1

  n2
is orthogonal set with respect to the weight
function p(x) = 1 on the interval [ 0, 1].

 sin(  n x) Orthogonality Relation

tan   
Singular Sturm-Liouville Problem
1)Singular boundary-value problem b
(m  n )  p( x) ym yn dx 
Solve:
a

Subject to:
r (b)[ ym (b) yn ' (b)  yn (b) ym ' (b)]
2)Singular boundary-value problem
 r (a )[ ym (a ) yn ' (a )  yn (a ) ym ' ( a )]
Solve:
Remark:
Subject to:  Observe that if, r(a) = 0,
 then x = a may be a singular point of
the differential equation,
3) periodic boundary-value problem  and consequently a solution of (ode)
may become unbounded as x a.
Solve:  However, we see from (*) that if r(a) =
0, then no boundary condition is
Subject to: required at x = a
 to prove orthogonality of the
the boundary conditions are periodic eigenfunctions provided these
solutions are bounded at that point.
 This latter requirement guarantees
4) no boundary condition the existence of the integrals involved.
Solve: solutions are
Subject to: no boundary condition bounded at that point
Self-Adjoint Form

Self-Adjoint Form

Not Self-Adjoint Form

Legendre’s DE
Rewrite in self-adjoint form

dividing the equation by the

coefficient of y’’ and then
multiplying by the integrating
factor
Self-Adjoint Form
Legendre’s DE

Legendre’s DE

self-adjoint form Legendre polynomials

r ( x)  1  x 2 , q( x)  0,
p( x)  1,   n(n  1)
orthogonal set
Observe that
r (1)  r (1)  0 Pn ( x)n0 is orthogonal with respect to the
weight function p(x) = 1 on [ -1, l].
The orthogonality relation
Parametric Bessel Equation

Parametric Bessel Equation Self-Adjoint Form

self-adjoint form Bessel Functions

n2 The two solutions J n ( x) Yn ( x)
r ( x)  x, q( x)  
x
, only J n ( x) is bounded at x = 0.
p ( x )  x,    2
orthogonal set
Observe that
r (0)  0 J n (i x)i1 isweight
orthogonal with respect to the
function p(x) = x on [0, b].

2 n 
The orthogonality relation

(1) n  x
J ( x)    
n 0 n!  (1    n) 2

cos J ( x)  J  ( x )
Y ( x ) 
sin( )
Exercises
 Second-Order Linear Differential
Equations (SOLDE)
A SOLDE has the form
d2y dy
p0 ( x ) 2  p1 ( x )  p2 ( x ) y  f ( x )
dx dx

If f ( x )  0, the equation is said to be "homogeneous."

The inhomogeneous equation can be solved once we know the solution to the
homogeneous equation using the method of Green's functions (discussed later).

Boundary conditions (BC) are usually of the form

y ( a )  y (b)  0 (Dirichlet )
y ( a )  y (b)  0 (Neumann)

17
 Sturm-Liouville Form
If we multiply the general differential equation
p0 ( x) y  p1 ( x) y   p2 ( x) y  f ( x)
x p 1 (t )
 p 0 (t )
dt
e
by the integrating factor w( x)   we have :
p0 ( x)
x p 1 (t ) x p 1 (t )
 dt
p ( x)  dt
e y  1 y  p2 ( x ) w( x) y  w( x) f ( x)
p 0 (t ) p 0 (t )
e
p0 ( x)

d   p 0 (t ) dt 
 x p 1 (t )

 e y  p2 ( x) w( x) y  w( x) f ( x)
dx  
 

Dividing this result by w( x) yields

1 d  dy ( x) 
 P ( x )  Q ( x) y ( x)  f ( x)
w( x) dx  dx 

x p 1 (t )
 dt
where P( x)  e p 0(t )
, Q ( x )  p2 ( x )
18
 Sturm-Liouville Operator
This is called the Sturm-Liouville or self-adjoint form of the
differential equation:

1 d  dy( x) 
  P( x)   Q( x ) y ( x )  f ( x)
w( x) dx  dx 

or (using u instead of y):

uf

Where L is the (self-adjoint*) “Sturm-Liouville” operator:

1 d  d * Discussed later
  P ( x )   Q( x )
w( x) dx  dx 

Note:
The operator L is assumed to be real here (w, P, Q are real). The solution u
does not have to be real (because f is allowed to be complex).
19
 Inner Product Definition

An inner product between two functions is defined:

We define an inner product as

b
 u , v    u ( x) v* ( x) w( x) dx
a

where w( x) is called a weight function.

Although the weight function is arbitrary, we will choose it to be

the same as the integrating function w( x) in the Sturm - Liouville equation.
This will give us the nice "self - adjoint" properties, as we will see.

20
 The Adjoint Problem
The adjoint operator L† is defined from

 u, v    u, †v 

For a Sturm - Liouville operator  u, v    u, v  so †

 .
(proof given next)

Hence, the Sturm-Liouville operator is said to be self-adjoint:

1 d  d
 †
  P( x )   Q ( x )
w( x ) dx  dx 

Note:
Self-adjoint operators have nice properties for eigenvalue problems,
which is discussed a little later.
21
 Proof of Self-Adjoint Property
Consider the inner product between the two functions u and v :
b
 u , v    v* ( x) w( x) u ( x)dx
a
b
 1 d  d  
  v ( x) w  x   
*
 P( x)   Q( x)  u ( x)dx
a  w   
x dx dx  
 d  d  
b
  v* ( x)    P( x)   Q( x) w  x   u ( x)dx
a  dx  dx  

The first term inside the square brackets is first integrated by parts, twice :
b
d  du ( x)   * du ( x)   du ( x)  dv* ( x)
b b
  v* ( x )  P ( x ) dx   v ( x ) P ( x )    P( x)  dx
a
dx  dx   dx  a a
dx  dx
b
 du ( x) 
b
 dv* ( x)  du ( x)
   v* ( x ) P ( x )    P( x)  dx
 dx  a a
dx  dx
b
 * du ( x) dv* ( x)  b
d  dv* ( x) 
  v ( x ) P ( x )  P( x) u ( x)    u ( x)  P( x)  dx
 dx dx  a a dx  dx 
22
Proof of Self-Adjoint Property (cont.)

Hence we have:
b
 * du ( x) dv* ( x)  b
d  dv* ( x) 
 u , v    v ( x ) P ( x )  P( x) u ( x)    u ( x)  P( x)  dx
 dx dx  a a dx  dx 
b
  v* ( x)Q( x)u ( x) w  x  dx
a
Multiply and divide by w(x)

or
b
 du( x ) dv ( x )*

 u, v     v * ( x ) P ( x )  P( x ) u( x ) 
 dx dx  a

  *
b
1 d  d 
 a u ( x ) w( x ) 
 w( x ) dx 
 
P ( x )
dx 

 Q ( x )  v ( x )dx


23
 Proof of Self-Adjoint Property (cont.)
b
 * du ( x) dv* ( x) 
 u , v    v ( x ) P ( x )  P( x) u ( x) 
 dx dx  a
b
 1 d  d   *
 a  w( x) dx  dx 
u ( x )  P ( x )  Q ( x )  v ( x) w( x)dx


This can be written as:

b
Note :
 u, v   J (u, v) a
  u, v  v*   v 
*

  real operator 
where
 * du( x ) dv* ( x ) 
J ( u, v )  P ( x )   v ( x )  u( x ) 
 dx dx 

From boundary conditions we have:

b
J ( u, v ) a  0
  u, v    u, v  proof complete 
24
 Eigenvalue Problems
We often encounter an eigenvalue problem of the form

u  u
(The operator L can be the Sturm-Liouville operator, or any other operator here.)

 The eigenvalue problem (with boundary conditions) is usually only satisfied

for specific eigenvalues:

  n , n  1, 2,

 For each distinct eigenvalue, there corresponds an eigenfunction u = un

satisfying the eigenvalue equation.

25
 Property of Eigenvalues
The eigenvalues corresponding to a self-adjoint operator are real.

Proof:

u  u
u  u
b b
 
a
u  u* w dx    u u* w dx
a
  u    *u*
*

 u, u   u, u
b b

 u  u w dx   *  u*u w dx
*

 u, †
u   u, u a a

 u, u   * u, u
 u, u   u, u

Hence:   *
26
 Orthogonality of Eigenfunctions
The eigenfunctions corresponding to a self-adjoint operator
equation are orthogonal if the eigenvalues are distinct.

Consider one solution of the eigenvalue problem :

um  mum , ( 1)

Next, consider another solution of the eigenvalue problem with a

different index n  m and distinct eigenvalue n  m :

un  n un , (2)

We multiply ( 1 ) by un* , the conjugate of ( 2 ) by um , subtract the second

from the first,weight the result by w( x ), and integrate on x  ( a, b) :

  
b b
um  un*  um  un  w dx   m  n*   umun* w dx ( 3 )
*

a a

Next, we consider the LHS of the above equation.

27
 Orthogonality of Eigenfunctions (cont.)
The LHS is:

  
b
um  un*  um  un  w dx   um , un    um , un 
*

  um , †
un    um , un  ( from the definition of adjoint )

  um , un    um , un  ( from the self - adjoint property )

0

Hence, for the RHS we have

b b

m 
*
n u u *
m n wdx  0   umun* wdx  0  orthogonality 
a a

since n*  n , m  n

28
 Summary of Properties
Assume an eigenvalue problem with a self-adjoint operator:

u  u,  †

 u, v    u, v 
Then we have:

 The eigenvalues are real.

The eigenfunctions corresponding to distinct eigenvalues are orthogonal.

That is,
a) m*  m
b
b) un , um   unum* w dx  0, m  n
a
29
 Example
Orthogonality of Bessel functions

Derive this identity:

 J  p x  J  p x  x dx  0,
0
n nm n nm m  m

pnm  mth root of J n  x 

30
 Example
Orthogonality of Bessel functions
b
Recall :  um*  x  un  x  w  x  dx  0 (for a Sturm - Liouville problem)
a

if um  a   ua  a   um  b   ua  b   0 (Dirchlet boundary conditions)

Consider : um  x   J n  pnm x  , pnm  mth root of J n  x 

Choose : a  0, b  1

J n  pnm 0  0, J n  pnm1  0

(The first equation is true for n  0. The case n  0 can be considered as a limiting case.)

What is w(x)?
We need to identify the appropriate DE that u(x) satisfies in Sturm-Liouville form.
31
 Example (cont.)
Bessel equation : t 2 y   ty    t 2  n 2  y  0, y  J n  t 

Use:
t  pnm x, dt  pnm dx
 1  y  J n  pnm x 
 
t pnm x

Note:
x y   xy    p x  n
2 2
nm
2 2
y0 The primes now mean
differentiating with
respect to x.

1  2 n2 
y  y   pnm  2  y  0
x  x 
32
 Example (cont.)
Rearrange to put into Sturm-Liouville eigenvalue form:

1  2 n2 
y  y   pnm  2  y  0
x  x 

1  n2 
 y  y   2  y  pnm
2
y
x x 

1  n2 
 y  y   2  y   y, y  J n  pnm x  ,   pnm
2

x x 

 1 d  d  n2 
  x  2  y  y
 x dx  dx  x  33
 Example (cont.)
Hence, we have:

 1 d  d  n2 
  x   2  y   y, y  J n  pnm x  ,   pnm
2

 x dx  dx  x 

Compare with our standard Sturm-Liouville form:

 1 d  d  
  w( x ) dx  P( x ) dx   Q ( x )  y   y
   

We can now see that the u(x) functions come from a Sturm-Liouville
problem, and we can identify:

n2
w  x   x, P  x   x, Q  x   2
x
34
 Example (cont.)

Hence we have:

 um  x  un
*
 x  w  x  dx  0
a

 J  p x  J  p x  x dx  0,
0
n nm n nm m  m

where J n  pnm   0  pnm  mth root of Bessel function J n 

35
 Adjoint in Linear Algebra

An inner product between two vectors is defined as:

a, b  a  b *   ai bi*
i

An adjoint of a square complex matrix [A] is defined from:

Au, v  u , A† v

Self-adjoint means:

A  A†

36
 Adjoint in Linear Algebra (cont.)
Theorem : For a complex matrix  A, the adjoint is given by :
 A†    At*  (i.e., the conjugate of the transpose)

Proof :
To show this, we need to show :
Au, v  u , At*v

where a, b  a  b *   ai bi*
i

To show this :

Au, v     Aij u j  vi*   u j Aij vi*

i j i j

u , At*v   ui  Aijt v*j    ui A ji v*j   u j Aij vi*  relabeling i and j 

i j i j i j

37
 Adjoint in Linear Algebra (cont.)
For a complex matrix we have established that

 A†    At* 

Therefore, if a complex matrix is self-adjoint, this means that the

matrix is Hermetian:

 A   At* 

Definition of a Hermetian matrix :  A   At*   Aij  A*ji 

Note: For a real matrix, self-adjoint means that the matrix is symmetric.

38
 Orthogonality in Linear Algebra

Because a Hermetian matrix is self-adjoint, we have the following properties:

 The eigenvalues of a Hermetian matrix are real.

 The eigenvectors of a Hermetian matrix corresponding to distinct

eigenvalues are orthogonal.

 The eigenvectors of a Hermetian matrix corresponding to the same

eigenvalue can be chosen to be orthogonal (proof omitted).

39
 Diagonalizing a Matrix
If the eigenvectors of an NN matrix [A] are linearly
independent, then it can be diagonalized as follows:

 A  e D  e
1
(proof on next slide)

1 0 0 0   e11 e12 e1N 

0  0 0  e e e2 N 
 D    0 02 0


e  e1  e2    eN   e21 e22 e3 N 
31 32
   
0 0 0 N  e41 e42 e4 N 

en   nth eigenvector (a column vector) corresponding to eigenvalue n

40
 Diagonalizing a Matrix (cont.)
Proof

e D   1 e1  2  e2  N  eN    from properties of a diagonal matrix 

 Ae   1 e1  2 e2  N eN    from properties of eigenvectors 

Hence, we have

 Ae  e D
Note:
The inverse will exist
so that since the columns of
the matrix [e] are
 A  e D e
1
linearly independent,
by assumption.

41
 Diagonalizing a Matrix (cont.)
If the matrix [A] is Hermetian, then:

 et*  e   I  identity matrix 

This follows from the orthogonality property of the eigenvectors.

Note: For the diagonal elements, the eigenvectors have been scaled so that en  en*  1

Hence for a Hermetian matrix we then have:

e  et*   the eigenvalue matrix is unitary 

1

Therefore, for a Hermetian (self-adjoint) matrix we have:

      
A  e D  e t*
A Hermetian matrix is always diagonalizable!

42