A homework on Engineering Analysis 10

By AMRIT ARYAL
Nov,2020

Problem 1: Inhomogeneous SL ODE

Given, The two point boundary value problem

y 00 + λy = x 0≤x≤1 (1)
with boundary conditions y’(0) = 0 and y(1) = 0.

1a: Using an appropriate eigen function expansion.

Solution
Solution of homogeneous form of above function is,
√ √
y = C1 sin λx + C2 cos λx
Applying boundary conditions
√ √ √ √
y 0 = λC1 cos λx − λC2 sin λx
At, y’(0)=0,
√
0 = λC1 ;C1 = 0 √
Then y is reduced to y = C2 cos λx
At y(1)=0,
√
0 = C2 cos λ

Here, √ √
either C2 = 0; it gives trivial solution and y=0 . or cos λ = 0 = cos 2n+1
2 π; λ =
2n+1
2 π

λ = ( 2n+1 2 2
2 ) π
Eigen value, λn = ( 2n+1 2 2
2 ) π
Eigen Function, φn = cos 2n+1
2 πx
P
Let y = An φn is the solution of inhomogeneous equation.

1
 fn
An = λ−λn
R1
f (x) cos 2n+1 πxdx
fn = 0R
1
cos2 2n+1
2
πxdx
0 2
Integration by parts in Numerator
1
x cos 2n+1
R
πxdx
fn = R01 2
cos2 2n+1 πxdx
0 2

here,

Integration by parts in Numerator

R1      
2n+1 4 π(2n+1) π(2n+1) π(2n+1)
0
x cos 2 πxdx = 2
π (2n+1) 2 2 sin 2 + cos 2 − 1 =
4 π(2n+1)
π 2 (2n+1)2 [ 2 (−1)n − 1]

Integration by parts in denominator

R1 2n+1
R1 2n+1
R1 2n+1
dx = 12 + 2π(2n+1)
1


0
cos2 2 πxdx = 0
cos2 2 πxdx = 0
cos2 2 πx sin (π (2n + 1)) =
1
2
Therefore;
8 π(2n+1)
fn = π2 (2n+1)2[ 2 (−1)n − 1]
Now,
π(2n+1)
fn fn
8
π 2 (2n+1)2
[
2 (−1)n −1]
An = λ−λn = λ−λn = 2n+1 2 2
λ−( 2 ) π P
∴, The solution of given inhomogeneous problem isy = An φ n =
π(2n+1)
P∞ 2
8
2[ 2 (−1)n −1]
y = n=0 π (2n+1)
λ−( 2n+1 )2 π 2
cos 2n+1
2 πx
2

1ab: Are there solutions for all values of λ

There are solutions for all values of λ expect λ = λn . There is no solution if
value of λ is equal to eigen value.

1ac. For what value of B will there be a solution

Given problem is,
y 00 + ( 3π 2 2
2 ) y = x + Bx 0 ≤ x ≤ 1.
with boundary condition y 0 (0) = 0 and y(1) = 0. Solution

The solution of above function can be written as,

y = C1 sin 3π 3π
2 x + C2 cos 2 x

Applying boundary condition,

y 0 = 3π 3π
2 C1 cos 2 x −
3π
2 C2 sin 3π
2 x
0
At y (0) = 0,

2
C1 = 0
y(x) is reduced to y = C2 cos 3π
2 x ,
Eigen value is λn = ( 3π
2 ) 2

and, Eigen function is φn = cos 3π2 x

RiemannR Lebesgue lemma gives condition for fn is convergence.

1
limn→∞ 0 f (x)φn dx → 0

Z 1
3π 2 3π 1 3π 3π
− (x + Bx) sin x| + (2x + B) sin xdx = 0
2 2 0 0 2 2

Z 1
3π 3π 3π 3π 1 3π 3π 3π
− (1 + B)(−1) + (2x + B) cos x|0 − 2 cos xdx = 0
2 2 2 2 2 2 0 2

3π 3π 3π 3π 3π 3π 3π 1
(1 + B) + B(−1) − 2 sin x| = 0
2 2 2 2 2 2 2 0

3π 3π 9π 2 27π 3
+B −B − =0
2 2 4 4

3π 3π 3π 9π
B(1 − )= ( − 1)
2 2 2 2

9π − 2
B=
2 − 3π

Problem 2: WKB Approximation

p0 (x) 0 λr(x) − q(x)
y 00 + y + y=0 (2)
p(x) p(x)

2a. Determine A(x) and B(x)

Given substitution is y(x) = α(x)u(x)
First derivative is,

y 0 = α0 (x)u(x) + α(x)u0 (x)

Second derivative is,

3
y 00 = α00 (x)u(x) + 2α0 (x)u0 (x) + α(x)u00 (x)
Substitute value of y’ and y” in eq(2)

p0 (x) 0 λr(x) − q(x)

y 00 + y + y=0
p(x) p(x)

p0 (x) 0 λr(x) − q(x)

α00 (x)u(x)+2α0 (x)u0 (x)+α(x)u00 (x)+ {α (x)u(x)+α(x)u0 (x)}+ α(x)u(x) = 0
p(x) p(x)

p0 (x) p0 (x) 0 λr(x) − q(x)

α(x)u00 (x)+u0 (x){2α0 (x)+ α(x)}+u(x){α00 (x)+ α (x)+ α(x)} = 0
p(x) p(x) p(x)

Dividing by α(x)

1 p0 (x) 1 p0 (x) 0 λr(x) − q(x)

u00 (x)+u0 (x) {2α0 (x)+ α(x)}+u(x) {α00 (x)+ α (x)+ α(x)} = 0
α(x) p(x) α(x) p(x) p(x)
(3)
compare with Given final ODE

u00 + A(x)u + λB(x)u = 0

Then,
1 0 p0 (x)
α(x) {2α (x) + p(x) α(x)} =0
0 p0 (x)
α (x) = − 2p(x) α(x)

Put value of α0 (x) in u(x) coefficient.

1 p0 (x) 0 λr(x) − q(x)

{α00 (x) + α (x) + α(x)}
α(x) p(x) p(x)

α00 (x) p0 (x) p0 (x) λr(x) − q(x)

− α(x) + α(x)}
α(x) p(x)α(x) 2p(x) α(x)p(x)

α00 (x) p0 (x)2 q(x) λr(x)

− 2
− +
α(x) 2p(x) p(x) p(x)

Where,

α00 (x) p0 (x)2 q(x)

A(x) = − −
α(x) 2p(x)2 p(x)

4
 r(x)
B(x) =
p(x)

2b.
u00 A(x)
+ u + B(x)u = 0 (4)
λ λ
Given Substitution is,

u(x) = exp(S(x))

Now, derivatives are,

u0 = exp(S(x))S 0 (x)

u” = exp(S(x))(S 0 (x))2 + exp(S(x))S 00 (x)

Substitute value of u” and u in eq(4).

u00 A(x)
+ u + B(x)u = 0
λ λ

exp(S(x))(S 0 (x))2 + exp(S(x))S 00 (x) A(x)

+ exp(S(x)) + B(x) exp(S(x)) = 0
λ λ

(S 0 (x))2 + S 00 (x) A(x)

exp(S(x))[ + + B(x)] = 0
λ λ

here, exp(S(x)) 6= 0
then,
(S 0 (x))2 + S 00 (x) A(x)
+ + B(x) = 0
λ λ

Is proved.

2c.
(S 0 (x))2 + S 00 (x) A(x)
+ + B(x) = 0 (5)
λ λ
For solve λ from above equation.
00
lets assume. S λ(x) ≈ −B(x), . Then, above ODE(5) is no longer second order

5
non linear ODE.

Next, Lets assume,

(S 0 (x))2
λ ≈ −B(x) Then ODE(5) is reduced to simpler form: S 00 (x) = A(x),
(S 0 (x))2 +S 00 (x)
∴ Only consistent method to solve above ODE is λ + A(x)
λ = −B(x)

2d.
(S 0 (x))2
≈ −B(x)
λ

s
p r(x)
S 0 (x) = ±i |λ|
p(x)

Let u(x) and v(x) be two functions of Sturm-Liouville ODE.

d dy
L(y) = dx p(x) dx + q(x)y(x) = −r(x)λy(x) a ≤ x ≤ b.

w(x) is auxiliary function such that

dv
w(x) = u dx − v du
dx

Satisfying boundary condition and SL ODE property.

uL(v) ≥ vL(u) and p(a)w(a) ≥ 0 ≥ p(b)w(b)

Then,

w(x) = 0

dv
i.e u dx − v du
dx = 0

dv
u dx = v du
dx

uv 0 = vu0

u0 u
= (6)
v0 v
0
Now,Integrate both side of S (x) equation.
Z s
0
p Z r(x)
S (x)dx = ±i |λ| dx
p(x)

6
Let a=0 and b=x
s
Z x Z x
0
p r(x)
S (x)dx = ±i |λ| dx
0 0 p(x)

s
p Z x
r(x)
S(x) = ±i |λ| dx
0 p(x)

From eq(6) we can rewrite this expression as

s
p Z x r(x0 ) 0
S(x) = ±i |λ| dx
0 p(x0 )

Final solution

y = c1 y1 + c2 y2

Calculate y from S(x)

u = exp(x) and y = αu.

y = α exp S(x)

y = c1 α exp [S1 (x)] + c2 α exp [S2 (x)]

" s # " s #
x p Z x r(x0 ) 0
p Z r(x0 ) 0
y = c1 α exp +i |λ| dx + c2 α exp −i |λ| dx
0 p(x0 ) 0 p(x0 )

is the general solution of above ODE.

2e.
Satisfying boundary condition is y(0)=0 and y(1)=0.

" s # " s #
0 p Z 0 r(x0 ) 0
p Z r(x0 ) 0
y(0) = 0 = c1 α exp +i |λ| dx +c2 α exp −i |λ| dx
0 p(x0 ) 0 p(x0 )

7
i.e c2 = −c1
" s # " s #
1 p Z 1 r(x0 ) 0
p Z r(x0 ) 0
y(1) = 0 = c1 α exp +i |λ| dx −c1 α exp −i |λ| dx
0 p(x0 ) 0 p(x0 )

" s #
1
p Z r(x0 ) 0
y(1) = 0 = 2ic1 α sin |λ| dx
0 p(x0 )

eix −e−ix
∵ sin x = 2i

∵ 2ic1 α 6= 0
" s #
1
p Z r(x0 ) 0
sin |λ| dx = 0
0 p(x0 )

" s #
1
r(x0 ) 0
p Z
sin |λ| dx = sin nπ
0 p(x0 )

Squaring both side

" s #2
1
r(x0 ) 0
Z
2
p
|λ| dx = [nπ]
0 p(x0 )

"Z s #2
1
r(x0 ) 0
λ dx = n2 π 2
0 p(x0 )
∴
n2 π 2
λ= h q i2
R 1 r(x0 )
0 p(x0 ) dx0

for n= 1,2,3,....

The general solution becomes.

" s #
x
p Z r(x0 ) 0
y = 2ic1 α sin |λ| dx
0 p(x0 )

8
Substitute value of λ here.
 
s
x
r(x0 )
Z
nπ
y = 2ic1 α sin  hR q 0 dx0 
 
p(x0 )
i
1 r(x ) 0
p(x0 ) dx
0
0

 R x q r(x0 ) 
0 p(x0 ) dx0
y = Cα sin nπ R q 
1 r(x0 ) 0
0 p(x0 ) dx

2f. eigenfunctions look like in this case for large

λ
The eigen function for this solution is
 R x q r(x0 ) 0

0 p(x0 ) dx
φn = sin nπ R q 
1 r(x0 ) 0
0 p(x0 ) dx

Let assume,

 R x q r(x0 ) 
0 p(x0 ) dx0
a = nπ R q 
1 r(x0 ) 0
0 p(x0 ) dx

’a’ arbitary constant.

then, Eigen function is reduced to

φn = sin anπ

For large value of λ

Eigen functions seem in sines wave , As λ is increases, distance between two
adjacent si zeros are closer to π .