Solution 10

1.D For any piecewise continuous function f (x), the Legendre expansion is

2n + 1 1
X Z
f (x) = an Pn (x), an = f (x)Pn (x) dx; x ∈ [−1, 1]
n=0
2 −1

(i) We can use the above formula. Alternately, using

1 = P0 , x = P1 , x3 = (2P3 + 3P1 )/5,

we get
f (x) = (5p0 + 8p1 + 2p3 )/5

(ii) Using the above formula,

a0 = 1/4, a1 = 1/2, a2 = 5/16.

Thus,
f (x) = P0 /4 + P1 /2 + 5P2 /16 + · · ·

2.T (i) x = 1 regular, x = 0 irregular (ii) x = 0, x = −1/3 regular

 3. (i) λ ≤ 0 leads to trivial solution. Thus, let λ = p 2 > 0. Then y = c1 cos px + c2 sin px.
Using the BCs c1 = 0 and sin p + p cos p = 0 or p + tan p = 0. This has infinite number
of roots (plot the curves y = −x and y = tan x). Thus, the eigen values are the roots
of the above equation and the eigen functions are y p = sin px.
(ii) This is Euler-Cauchy equation. Using the transformation x = e t , we get y �� + λy =
0, y(0) = y �(1) = 0. Again for λ ≤ 0 trivial solution. Thus, λ = p 2 > 0 and
y = c1 cos pt+c2 sin pt. Using BCs, we get c1 = 0 and p = (n+1/2)π, n = 0, 1, 2, 3, · · · .
Thus, λn = [(2n + 1)π/2]2 , n = 0, 1, 2, 3, · · · and y n = sin[(n + 1/2)π log x].
4 . Multiplying by y and using integration by parts, we get
� b � b � b
λ 2
ry dx = 2
qy dx + py �2 dx − [pyy �]ba
a a a

(i) p(a) = p(b) = 0 =⇒ [pyy �]ba = 0 (ii) p(a) = p(b) with y(b) = y(a), y � (b) =
y �(a) =⇒ [pyy �]ba = 0 (iii) y(a) − ky �(a) = y(b) + my �(b) = 0, k, m > 0, =⇒ [pyy �]ba =
−mp(b)y � (b)2 − kp(a)y �(a)2 .
Thus, in (i) & (iii) [pyy �]ba = 0 and in (ii) [pyy �]ba ≤ 0. Thus, λ is positive.

5. The BC y(a) �= y(b) =⇒ either y(a) �= 0 or y(b) �= 0 and y �(a) �= y � (b) =⇒ either
y �(a) �= 0 or y � (b) �= 0. Also y(a) = y �(a) = 0 is not possible since then we get trivial
solution only. Similarly y(b) = y � (b) = 0 is not possible. Thus, we can write the BCs
as
c1 y(a) + c2 y � (a) = 0 and d1 y(b) + d2 y �(b) = 0
where c1 or c2 not equal to zero and d 1 or d2 not equal to zero.
Let u and v are eigen functions corrsponding to an eigen value λ. Then (pu � )� + qu +
λru = 0 and (pv � )� + qv + λrv = 0. Multiplying the 1st by v and the second by u and
subtracting we get [pW (u, v)] � = 0 where W is the Wronskian. Since u and v satisfy
the above BCs, W (u, v) = 0 at x = a and x = b. Thus, pW (u, v) ≡ 0 or W (u, v) ≡ 0.
Hence u and v are linealy dependent.
 Laplace Transformation
MTH203:L

6. � ∞ � ∞
−st 1
L (f (at)) = e f (at) dt = e−(s/a)τ f (τ ) dτ = F (s/a)/a
0 a 0

7. (a)
� 2 � 3 � 4
−st −st
L([t]) = e dt + 2 e dt + 3 e−st dt + · · ·
1 2 3

e−s 1
= (1 + e−s + e−2s + e−3s + · · · ) =
s s(es − 1)

(b)

m! 1
L(tm ) = m+1
=⇒ L(tm cosh bt) = L(ebt tm + e−bt tm )
s � 2 �
(m! 1 1
= +
2 (s − b)m+1 (s + b)m+1

(c)
a a
L(sin at) = =⇒ L(et sin at) =
s2 +a 2 (s − 1)2 + a2
�∞
(d) Use L[f (t)/t] = s
F (s) ds. Now
� ∞
a 1
L(sin at) = =⇒ L(sin at/t) = a = tan−1 (a/s)
s2 + a 2 s s2 + a 2
� �
t −1 a
=⇒ L(e sin at/t) = tan
s−1

(e)
� � � � ��
−1 1 −1 1 −1 1
L(sin t/t) = tan (1/s) =⇒ L(cosh t sin t/t) = tan + tan
2 s−1 s+1

(f) � π
3(1 + e−πs )
L[f (t)] = e−st sin 3t dt =
0 s2 + 9

8. (a) Consider g(t) = u(t)−u(t−π)+u(t−2π) cos t = u(t)−u(t−π)+u(t−2π) cos(t−2π)

1 e−πs se−2πs
L[f (t)] = L[g(t)] = − + 2
s s s +1
(b) Consider g(t) = [u(t−1)−u(t−2)] cos(πt) = −u(t−1) cos π(t−1)−u(t−2) cos π(t−2)

s(e−s + e−2s )
L[f (t)] = L[g(t)] = −
s2 + π 2
9. (a) Use L(−tf (t)) = F � (s). Thus,
a sin at
F � (s) = − =⇒ L−1 [F � (s)] = − sin at =⇒ f (t) =
s2 + a 2 t
(b)

2s 2 2(e−t − cos t)
F � (s) = − =⇒ L −1
[F �
(s)] = 2(cos t − e −t
) =⇒ f (t) =
s2 + 1 s + 1 t
(c) � �
s+2 1 1 1
F (s) = 2 2
= 2
−
(s + 4s − 5) 12 (s − 1) (s + 5)2
Thus,
et − e−5t
f (t) = t
12
(d)
se−πs
= e−πs L(cos 2t) = L(u(t − π) cos[2(t − π)])
s2 + 4
Thus, � �
−1 se−πs
L = u(t − π) cos 2t
s2 + 4
(e)
(1 − e−2s )(1 − 3e−2s ) 1 4e−2s 3e−4s
= − + 2
s2 s2 s2 s
Thus,
f (t) = t − 4u(t − 2)(t − 2) + 3(t − 4)u(t − 4)
10(a) Taking Laplace Transform on both sides and simplifying (Y(s)= L[y(t)])

Y (s) = s/(s2 + 4)2 + 1/(s2 + 4)

Using convolution [or any other technique]

�
1 t sin 2t
y(t) = sin(2τ ) cos(2(t − τ ))dτ +
2 0 2
t sin 2t sin 2t
= +
4 2
(b) Let r(t) = 4u(t − 0)t + 4u(t − 1)(1 − (t − 1)). Taking Laplace Transform on both
sides of the ODE, we get
4 s−1
(s2 + 3s + 2)Y (s) = R(s) =⇒ Y (s) = + e−s 2
s2 (s + 1)(s + 2) s (s + 1)(s + 2)
Using partial fraction and solving we get
� �
y(t) = −3 − e−2t + 4e−t + 2t + u(t − 1) 5 + 3e−2(t−1) − 8e−(t−1) − 2(t − 1)

(c) Let r(t) = 8(u(t − 0) − u(t − π)) sin t = 8u(t − 0) sin t + u(t − π) sin(t − π). Taking
Laplace Transform on both sides of the ODE, we get
4 R(s)
(s2 + 9)Y (s) = R(s) + 4 =⇒ Y (s) = + 2
s2 +9 s +9
We can explicitly write R(s) and then use partial fraction technique. Otherwise, use
convolution as follows
�
4 1 t
y(t) = sin 3t + r(τ ) sin 3(t − τ ) dτ
3 3 0
Thus for 0 < t < π, we get
�
4 8 t 4 1
y(t) = sin 3t + sin τ sin 3(t − τ ) dτ = sin 3t + sin t − sin 3t = sin 3t + sin t
3 3 0 3 3
and for t > π, we get [since r(t) = 0]
� �
4 8 π 1 t 4
y(t) = sin 3t + sin τ sin 3(t − τ ) dτ + 0 sin 3(t − τ ) dτ = sin 3t
3 3 0 3 π 3