Mechanical Engineering Dept. Msc.

Course 21-22

Orthogonal Functions and the General Expansion Problems:

Definition 1:- If a sequence of real functions
( ) n  1,2,3,
which are defined over some interval (a,b), finite or infinite, has the property that
  0, mn
  m ( x ) n ( x )dx
b

a   0, mn

Then the functions are said to form an orthogonal set on that interval.
Definition 2:- If the functions of an orthogonal set  n ( x ) have the property that

  n ( x )dx = 1
b
2

For all values of n then the functions are said to be orthonormal on that interval (a,b).
Notes :
1- It is no specialization to assume that an orthogonal set of functions is also orthonormal.
2- Any set of orthogonal functions can easily be converted into an orthonormal set. In fact, if

the function of the set  n ( x ) are orthogonal and if kn is the value of   n ( x )dx , then the
b
2

1 ( x )  2 (x) 3 (x)
function , , , ··· (kn must be positive) are orthonormal.
k1 k2 k3
Definition 3:- If a sequence of real functions
 n ( x ) n  1,2,3,
has the property that over some interval (a,b), finite or infinite
  0, mn
 p( x ) m ( x ) n ( x )dx
b

a   0, mn

Then the functions are said to be orthogonal with respect to the weight function p(x) on that
interval.
Now, 1- any set of functions orthogonal with respect to a weight function p(x) can be converted
into a set of functions orthogonal in the first sense (Definition 1) simply by multiplying each
member of the set by p( x ) . ( ) > 0.

14
Mechanical Engineering Dept. Msc. Course 21-22

2- with respect to any set of functions  n ( x ) orthogonal over an interval (a,b), an arbitrary
function f(x) has a formal expansion analogous to a Fourier expansion,
f(x) = C11 (x)  C 2  2 ( x)  C33 ( x)    C n  n ( x)   (1)
then multiplying both sides of equation (1) by  n ( x ) and integrating formally between the
appropriate limits a and b, we have

 f ( x ) n ( x )dx    n (x )1 (x )dx +   n ( x ) 2 ( x )dx +   n ( x ) 3 ( x )dx

b b b b
+
a a a a

   n ( x ) n 1 ( x )dx
b b
+ n ( x ) n ( x )dx + + ······
a a

From the property of orthogonality, all integrals on the right are zero except   2n ( x )dx
b

 f ( x) n (x)dx
b

= a

  n ( x)dx
b
2

EXAMPLE:-
Show that the given set is orthogonal on the given interval I and determine the corresponding
orthonormal set  1 , cosx , cos2x , cos3x , ······ , cosnx , ····· , 0 2
Solution

in the   n ( x ) m ( x )dx
b
Let  n ( x )  cos nx  m ( x )  cos mx ,
a

 cos mx. cos nxdx  2 [cos( n  m) x  cos( n  m) x ]dx

2 2
1
=
0 0

1  sin(n  m ) x sin( n  m ) x 
2
=  ·········(1)
2  n  m n  m  0
Since n and m are integer  eq. (1) = 0 for n  m
Now, for n = m
the first term in eq. (1) = 0, but for second term we take limit as n  m as follow
lim sin( n  m ) x  2  lim sin(n  m).2
=
n  m n  m  0 nm nm

15
Mechanical Engineering Dept. Msc. Course 21-22

lim 2. cos(n  m).2

Taking limit using L'opitals rule  2
nm 1

 cos
2
1
2
nx .dx  .2  = π
0
2

Now we need checking the orthogonality condition for 1 with cosnx for n  1,2,3, 

 1. cos nxdx =
sin nx 
2 2
 0, n  1,2,3,
0
n  0

For n = 0, cos0 = 1

 1 .dx  2
2
2

 the given set is orthogonal on the interval 0 2

Then the corresponding orthonormal set is
1 cos x cos 2x cos 3x
, , , , ········
2   
Definition 4 :- A real function f(x) is said to be Null function on the interval (a,b) if

f
b
2
( x )dx  0
a

EXAMPLE:-

Show that the set  sinnx  are orthogonal at the interval (  ,  ) and then show that the
function g(x) = x2 cannot be represented on this interval by a series of the form
C1 sin x  C 2 sin 2x  C3 sin 3x    C n sin nx  
Solution

in the   n ( x ) m ( x )dx
b
Let  n ( x )  sin nx  m ( x )  sin mx ,
a

 sin mx. sin nxdx =  2 [ cos( n  m) x  cos( n  m) x ]dx

 
1
 

1  sin(n  m ) x sin(n  m ) x 

=  ·········(1)
2  n  m n  m   
Since n and m are integer  eq. (1) = 0 for n  m
Now, for n = m

16
Mechanical Engineering Dept. Msc. Course 21-22

the first term in eq. (1) = 0 , but for second term we take limit as n  m as follow

lim sin( n  m ) x   lim 2 sin(n  m).

= =π
n  m n  m    n  m nm
 the given set is orthogonal on the interval −
Then, let us find Cn

 f (x) n (x)dx x x
b  
2
sin nxdx 2
sin nxdx

Cn = a
= 
=

  n (x)dx  sin
b 
2 2
nxdx
a 

Using integrating by part

Note that cos (nπ) = cos (-nπ)
D I
x2 + sinnx
2x cos nx

- n
2 sin nx

+ n2
cos nx
0
n3
2
[cos n  cos(  n)]
Cn = n 3
0

Evidently, important as it is, orthogonality is not the whole story, and the functions in our
orthogonal systems must posse some further property before the expansion (1) can be used with
confidence. What is required is that the set of functions  n ( x ), in addition to being orthogonal,
should also posse the property of completeness.
Definition 5 :- A set of orthogonal functions  n ( x ) is said to be complete if the relation

 f ( x ) n ( x )dx  0 can hold for all values of n only if f(x) is a null function on (a,b).
b

Theorem :- If the formal expansion , C11 ( x)  C22 ( x)  C33 ( x)    Cnn ( x)   of a function

g(x) in terms of the members of a complete orthogonal set  n ( x ) converges and can be

17
Mechanical Engineering Dept. Msc. Course 21-22

integrated term by term, then the sum of the series differs from g(x) by at most a null function; i.e
the sum of the series cannot differ from g(x) over any interval of finite length.
Proof

 C  ( x) converges to some function, hence it is meaningful to


By hypothesis the series n n
n 1

consider the difference

 C n  n (x)

h(x) = g(x) -
n 1

If we can prove that h(x) is a null function, the assertion of the theorem will be established. To do
this consider

  m ( x ).h ( x )dx =   m ( x ).[g( x )   C n  n ( x )]dx

b b 

a a n 1

=   m ( x ).g ( x )dx    m ( x ).[ C n  n ( x )]dx

b b 

a a n 1

=   m ( x ).g ( x )dx   C n   n ( x ) m ( x )dx

b  b

a n 1 a

Strume – Liouville Problems I

Various important orthogonal sets of function arise as solutions of second order differential
equation of the form

r ( x )   q( x )  p( x )y  0

d  dy 
 [Strume – Liouville equation]
dx  dx 
On some interval a ≤ x ≤ b satisfying conditions of the form
a1y(a) - a2 y  (a) = 0 , b1y(b) - b2 y  (b) = 0
where a1,a2,b1 and b2 are given constants such that a1 and a2 are not both zero and b1 and b2 are
not both zero. r(x) , q(x) , p(x) are be real-valued and continuous on the interval a ≤ x ≤ b. If λ1 , λ2
, λ3 , ········ , λn are distinct values of the parameter λ for which there exist nontrivial solutions y1 ,
y2 , y3 , ········ , yn , then the functions y n ( x ) form a system orthogonal with respect to the
weight function p(x) over the interval (a,b)
Note:- y n ( x ) is called Eigenfunction and λn is called Eigenvalues

18
Mechanical Engineering Dept. Msc. Course 21-22

proof
Let ym and yn are Eigenfunction and λm and λn are corresponding Eigenvalues, then from Strume –
Liouville differential equation
[ ry m ]  [q   m p ]y m  0 ·········(1)
[ ry n ]  [q   n p ]y n  0 ········(2)
Now, multiply equation (1) by yn and equation (2) by ym
y n [ ry m ]  [q   m p]y n y m  0 ·········(3)

y m [ ry n ]  [q   n p]y m y n  0 ·········(4)
Then, subtracting equation (4) from equation (3) yield
d[ryn ] d[rym ]
( m   n )p(x ) y m y n  y m  yn ·········(5)
dx dx
Integrate both sides of equation (5) over x from a to b

d[ ry n ] d[ ry m ]
( m   n )  p( x ) y m ( x ) y n ( x )dx   y m dx   y n
b b b
dx ·········(6)
a a
dx a
dx

Using integration by part for right hand side

d[ ry n ] d[ ryn ]
 ym
b
dx  Let u = ym dv = dx 
dx dx

 
a

dx = y m ryn   ym ryn dx

d[ ry n ]
 ym
b

b

a
dx a

d[ rym ] d[rym ]
 yn
b
dx  Let u = yn dv = dx 
dx dx

 
a

dx = y n rym   yn ry m dx
d[ ry m ]
 yn
b

b

a
dx a

equation (6) reduced to

( m   n )  p( x ) y m ( x ) y n ( x )dx  ry m y n  ry n y m ab

b

= r ( b )y m ( b) y n ( b )  y n ( b) y m ( b ) - r (a )y m (a ) y n (a )  y n (a ) y m (a ) 
But from boundary conditions

19
Mechanical Engineering Dept. Msc. Course 21-22

a1
a1y(a) - a2 y  (a) = 0  y  (a) = y(a)
a2
b
b1y(b) - b2 y  (b) = 0  y  (b) = 1 y(b)
b2

( m   n )  p ( x ) y m ( x ) y n ( x )dx  0
b

a
since λm and λn were any two distinct values of λ  λm - λn  0

  p( x ) y m ( x ) y n ( x )dx  0
b

so y n ( x ) is orthogonal set with respect to weight function p(x) on the interval (a,b).
EXAMPLE:

Prove that the general linear second order differential equation Po (x ) y  P1 (x) y  P2 ( x) y  y
can be reduced to an equation of Strum – Liouville form by multiplying it by the factor
 x P (x) 
exp   1
1
dx 
Po ( x )  x o Po ( x ) 

Solution
the given D.E can e written as Po y  P1 y  (P2  ) y  0 ·········(1)

e h where h   1
1 x
P ( x) dh P1
multiply both sides of eq. (1) by the factor dx  
Po ( x ) x
Po ( x ) dx Po
o

P1 h P e h
e h y  e y  ( 2 e h   ) y  0 ·········(2)
Po Po Po
d h d (e h ) dh
Since 
(e y ) = y   e h y = y e h  e h y
dx dx dx
d h P1
(e y ) = y e h  e h y 
dx Po
d h P eh
 equation (2) can be written as (e y)  ( 2 e h   ) y  0
dx Po Po
Comparing with Strum – Liouville we see that

P2 h eh
r (x )  e h q( x )  e p( x )    
Po Po

20
Mechanical Engineering Dept. Msc. Course 21-22

EXAMPLE :
a- Verify that the system y  y  0 , y(0) = 0 and y(1) = 0 is Strum – Liouville system.
b- Find the eigenfunctions and eigenvalues of the system. c- Prove that the eigenfunctions are
orthogonal in 0 ≤ x ≤ 1. d- Find the corresponding set of normalized eigenfunctions. e- Expand
f(x)=1 as series of these orthonormal functions.

Solution
a-First compare the given D.E with Po (x ) y  P1 (x ) y  P2 (x ) y  y then,

Po ( x )  1 , P1 ( x )  0 , P2 ( x )  0 ,   

 Po (x ) dx  0 ,
P1 ( x )
x
h eh  1,  r ( x )  1 , q ( x )  0 , p( x )  1 ,   
xo

a = 0, b = 1 a1=1 , a2 = 0 , b1 = 1 , b2 = 0
b-now, to find n- nontrivial solution y n ( x ) corresponding to λ1 , λ2 , λ3 , ········ , λn, we check
for λ = 0 , λ<0 , λ>0
1- Let λ = 0 then from given D.E y  0 ,  y(x) = C1 + C2x
Applying the conditions y(0) = 0 and y(1) = 0  C1 = 0 , C2 = 0
2- Let λ < 0 let λ = -k2 where k>0 then the given D.E becomes to

y  k 2 y  0  characteristic equation is m 2  k 2  0 or m1, 2   k

y

coshx
sinhx

 y(x) = C1ekx + C2 e-kx = C1coshkx + C2 sinhkx 1

Applying the conditions y(0) = 0 and y(1) = 0  C1 = 0 , C2 = 0 x

3- Let λ > 0 let λ = k2 where k>0 then the given D.E becomes to

y  k 2 y  0  characteristic equation is m 2  k 2  0 or m1, 2  ik

y(x) = C1coskx + C2 sinkx

Applying the conditions y(0) = 0  C1 = 0 and y(1) = 0 , C2  0 , then sink = 0

21
Mechanical Engineering Dept. Msc. Course 21-22

or kn = nπ λn = n2π2 are Eigenvalues

yn(x) = Cn sinnπx Eigenfunctions
c-

 sin mx. sin nxdx =  2 [ cos( n  m)x  cos( n  m)x ]dx

1
1 1

0 0

1  sin( n  m ) x sin( n  m ) x 
1
=   ·········(1)
2  nm nm 0

Since n and m are integer  eq. (1) = 0 for n  m

Now, for n = m
the first term in eq. (1) = 0 , but for second term we take limit as n  m as follow

lim sin( n  m) x  1 lim sin( n  m ).x 1 x cos 0 1

 =  =π
nm nm 0 nm nm 0 1 0

 the given set is orthogonal on the interval 0 ≤ x ≤ 1, and

 sin
1
1
2
nx.dx 
0
2
1
d- the obtained eigenfunction can orthonormalized by dividing it by or multiplying it by
12

2
Thus the set yn(x) = 2 sinnπx is an orthonormal set
e- To expand f(x) = 1 by this orthonormalized

 f ( x) n (x )dx  
b 1 1
2 sin nxdx 2 sin nxdx
2
Cn = a
= 0
= 0
= (1  cos n)
n
  n (x )dx  2 sin nxdx
b
2
1
2
1
a 0

 n (1  cos n).

2
 1= 2 sin nx
n 0

22
Mechanical Engineering Dept. Msc. Course 21-22

Example:
Find the eigenvalues and eigenfunctions of the of y  y  0 subjected to the boundary
conditions y(0)  0 , and y()  y ()  0 , then use these eigenfunctions to expand f(x) =
x.
Solution
First write the given D.E as Strum – Liouville system by comparing the given D.E with general
form of second order Po (x ) y  P1 (x ) y  P2 (x) y  y then,

Po ( x)  1 , P1 ( x )  0 , P2 ( x )  0 , 

 Po (x) dx  0 ,
P1 ( x )
x
h e h  1,  r ( x )  1 , q ( x )  0 , p( x )  1 ,   
xo

d h P eh
(e y)  ( 2 e h   ) y  0
dx Po Po
a = 0, b = π a1=1 , a2 = 0 , b1 = 1 , b2 = 1
now, to find n- nontrivial solution y n ( x ) corresponding to λ1 , λ2 , λ3 , ········ , λn, we check for
λ=0 , λ<0 , λ>0
1- Let λ = 0 then from given D.E y  0 ,  y(x) = C1 + C2x
applying the conditions y(0) = 0  C1 = 0 , then y(x) = C2x  y  C2
applying the conditions y()  y ()  0 , C2 π- C2 =0  C2 (π- 1) = 0,  C2 = 0
2- Let λ < 0 let λ = -k2 where k >0 then the given D.E becomes to

y  k 2 y  0  characteristic equation is m 2  k 2  0 or m1, 2   k

 y(x) = C1ekx + C2 e-kx = C1coshkx + C2 sinhkx

applying the conditions y(0) = 0 ,  C1 = 0 y(x) = C2 sinhkx
y  C2 k coshkx
applying the conditions y()  y()  0 , C2 sinhkπ - C2 k coshkπ = 0
or C2 (sinhkπ - k coshkπ) = 0 since always coshkπ > sinhkπ
Then sinhkπ - k coshkπ  0  C2 =0
3- Let λ > 0 let λ = k2 where k >0 then the given D.E becomes to

y  k 2 y  0  characteristic equation is m 2  k 2  0 or m1, 2  ik

y(x) = C1coskx + C2 sinkx
applying the conditions y(0) = 0  C1 = 0 y(x) = C2 sinkx, y  C2 k coskx

23
Mechanical Engineering Dept. Msc. Course 21-22

applying the conditions y()  y()  0 , C2 sinkπ - C2 k coskπ = 0

C2 (sinkπ - k coskπ) = 0, let C2  0 then sinkπ - k coskπ = 0  sinkπ = k coskπ ·········(1)
sin k
or k  tan k
cos k
since k is any value grater than 0 k  tan k To find the n-th root of this
equation graph y = k and y = tankπ

k
1.29 2.37 3.4

k1 = 1.29 , k2 = 2.37 , k3 = 3.4 , k4 = 4.42 , ········· , kn

so the solution of y  y  0 subjected to the boundary conditions y(0)  0 , and

 C n sin k n x

y()  y()  0 is y(x) =
n 0

Now, to expand f(x) = x by  n ( x )  sin k n x

 f ( x ) n ( x )dx  x sin k n xdx

 2

2
b
1
sin k n   cos k n  sin k n   cos k n 
kn
2
kn kn
2
kn
Cn = a
= 
0
= =
1  1
  n (x )dx 
b
1  sin 2k n 
2
sin 2
k xdx    sin 2 k  
n
2  2 k n
n
 2k n
a 0

But sin2knπ = 2 sinknπ cosknπ , and from equation (1) sinknπ = kn cosknπ

24
Mechanical Engineering Dept. Msc. Course 21-22

cos k n 
2(1  )
kn
Cn = Note that Cn does not depends on x i.e Cn= constant
  cos 2 k n 
cos k n 

 x = 2(1  ) 

kn
where k1 = 1.29 , k2 = 2.37 , k3 = 3.4 , k4 = 4.42 , ········· , kn
n 0   cos k n 
2

H.W
Find the eigenvalues and eigenfunctions of the of y  y  y  0 subjected to the
boundary conditions y(0)  y(L)  0 .

H.WS
If the following differential equations satisfies the conditions
( )+ ( )=0 ( )+ ( )=0
where , , and are constants, and both ≠ 0, and both ≠ 0
Verify the orthogonality condition of each of the equations

−2 −2 =0 where n is a positive integer Hermit’ Equation

+ (1 − ) + =0 where n is a positive integer Laguerres’ Equation
(1 − ) −2 + ( + 1) = 0 Legendre’s Equation
+ +( + ) =0 Bessel’s Equation

25