Solutions To Tutorial Sheet No - 4

Q.1 (i) The given equation is y00 − 5y0 + 6y = 2ex . The characteristic equation is:

λ2 − 5λ + 6 = 0

with roots 2, 3. Therefore, y1 = e2x , y2 = e3x are two independent solutions of the homo-
geneous part. The Wronskian is given by
 
e2x e3x  
W = = e5x
 
2e2x 3e3x
 
 

Therefore a particular solution of the ODE is:

yp = −y1 x yW2 r dt + y2 x yW1 r dt

R R

= −e2x x 2e−t dt + e3x x 2e−2t dt

R R

= 2ex − ex = ex

(ii) y 00 + y = tan x, 0 < x < π/2.

Here y1 = cos x, y2 = sin x, W = 1 and hence

yp = − cos x x sin t tan t dt + sin x

R Rx
cos t tan t dt
= − cos x(ln | sec x + tan x|)

(iii) y 00 + 4y 0 + 4y = x−2 e−2x , x > 0.

Here the characteristic equation has a double root λ = −2. Therefore y1 = e−2x , y2 = xe−2x
are two independent solutions of the homogeneous part. The Wronskian W = e−4x and
−2t −2 −2t R x e−2t t−2 e−2t
yp = −e−2x x te et−4te dt + xe−2x
R
e−4t
dt
= −e−2x ln x − e−2x

(We can ignore the term e−2x and take yp = −e−2x ln x.

(iv) y00 + 4y = 3cosec 2x, 0 < x < π/2.
y1 = cos 2x, y2 = sin 2x, W = 2.
3 1
yp = − cos 2x + ln sin 2x.
2 2
(v) x2 y 00 − 2xy 0 + 2y = 5x cos x. The homogeneous part is a Cauchy-equation whose auxiliary
equation is
m2 − 3m + 2 = 0
with roots m = 1, 2. Therefore, y1 = x, y2 = x2 , W = x2 .
2
yp = −x x t (5tt2cos t) dt + x2 x t(5ttcos t)
R R
2 dt
Rx 2 x
= −5x t cos tdt + 5x
R
cos tdt
= −5x cos x

1
 (vi) xy 00 − y 0 = x2 ex . (2)
2 00
Rewrite this as x y − xy = x3 ex (3)
so that it is an equation of Cauchy’s type. The auxiliary equation is m2 − 2m = 0 and
hence y1 = 1, y2 = x2 are solutions of the homogeneous part. The Wronskian is W = 2x.
Before applying the variation of parameters, we rewrite (3) as
0
y 00 − yx = xex (4)
and now R 2 t t
yp = − x t 2tte dt + x2 x te2t dt
R

= 12 [−(x2 − 2x + 2)ex + x2 ex ]
= (x − 1)ex .

Q. 2 [CLASSWORK]
   
y y  
 y1 y2 
W = 100 200  = 
0
= −pW.
 
0 0
−py1 − qy1 −py2 − qy2
 
y1 y2  
Rx
Therefore, W = k exp( p(t)dt). If W (x0 ) = 0 then k = 0 and hence W ∼ 0.

Q.3 [CLASSWORK] Given a solution y = y1 of y 00 + p(x)y 0 + q(x)y = 0 we try for another solution
of the form
y2 = uy1
But then, y20 = y10 u + y1 u0 and y200 = y1 u00 + 2y10 u0 + y100 u. Therefore

0 = y200 + py20 + q = y1 u00 + (2y10 + py1 )u0

since y1 is a solution. Let us set v = u0 to reduce the order of this equation:

y1 v 0 + (2y10 + py1 )v = 0

Assuming that y1 does not vanish in the given interval, we can divide out by y1 , separate the
variables, and solve for v :
c1 exp( x p(t)dt)
R
v=
y12
Integrating once more to get the required expression: u = c2 + c1 ψ where
Z x Rt
exp[−p(s)ds]
ψ(x) = dt.
y1 (t)2
The point is that the steps are reversible and u obtained in this way satisfies the requirement
that y2 = uy1 is a solution. This can be directly checked also.

Q.4 In each of the following problem, we use the theory, as in Q. 3. Remember to put the given
equation in the standard form first. In order to get a second solution y2 independent of y1 , we
can take c2 = 0, c1 = 1 in the general formula of ψ. Then we simply put y2 = y1 ψ. In what
follows, we therefore obtain the expressions for ψ in each case:

2
 sin x 1
(i) y1 = √ , p(x) = . Therefore
x x
exp(− t p(u)du)
Z x R Z x dt
ψ(x) = dt = = − cot x
y1 (t)2 sin2 t
2
(ii) y1 = ex ; and p(x) = −4x. Therefore
Z x 2t2
e
ψ(x) = dt = x.
e2t2
x 3
(iii) y1 = ; p = . Therefore,
(x − 1)2 x−1

ψ(x) = x exp(−3 ln(t−1))

R
t2 /(t−1)4
dt
R x (t−1)
= t2
dt
1
= ln x + x

1
(iv) y1 = cos x2 ; p = . Therefore
x
Z x
tdt 1
ψ(x) = 2 2
= tan x2 .
cos t 2
s
1 − x2 x
(v) y1 = ; p=− . Therefore,
x 1 − x2
Z x
(1 − t2 )−1/2 t
ψ(x) = dt
1 − t2
Z x
t 1
= 2 3/2
dt = .
(1 − t ) (1 − x2 )1/2
2 2 6x
(vi) y1 = 1 + x2 , p = 2
= − . Therefore
x(1 + 3x ) x 1 + 3x2
Rx 3x2 +1
ψ(x) = x2 (1+x2 )2
R x  1−t2 
= (1+t2 )2
+ t12 dt
x
= 1+x2
− x1 = − x(1+x1
2)

Q.5 Putting yi = eri x , we have,y j = rij yi for all i, j. Therefore W (y1 , . . . , yn ) = exp((
P
i ri )x)V
where V is the Vandermond determinant
 

 1 1 ··· 1 

· · · rn
 

 r1 r2 

2
r1 r22 · · · rn2
 
 
 

 ··· ··· ··· ··· 

n−1
r2n−1 · · · rnn−1
 
 r1 

3
 which is equal to i>j (ri −rj ). Since ri are distinct, it follows that W never vanishes. Therefore
Q

yi which are solutions of an nth order linear differential equation are independent.
Aliter Suppose there exist real numbers αi such that i αi yi ∼
P
= 0. By taking derivatives up to
(n − 1)-th order, this yields
c (α , . . . , α )t = (0, 0, . . . , 0)t
W 1 n

where W c is the Wronskian matrix (not the determinant). Since the Wronskian itself is not
zero, this matrix is invertible. Therefore

(α1 , . . . , αn )t = (0, 0 . . . , 0)t .

P
Q.6 [CLASSWORK] We set y = i ci y i where ci are some smooth functions whose derivatives
satisfy
(j)
c0i yi = 0 for all j = 0, 1, . . . , n − 2.
X
(1)
i
Under this condition, we check that y is a solution of the given equation iff

(n−1)
c0i yi
X
= r(x) (2)
i

If W
c is the Wronskian matrix, then condition (1) and (2) are together equivalent to say

c (c0 , . . . , c0 )t = (0, . . . , 0, r(x))t

W 1 n

By Cramer’s rule this is the same as saying that

Di
c0i =
W
Upon integration, this is the same as
Z x
Di (s)
ci (x) = ds.
W (s)

Q.7 We first observe that if y1 , y2 are two solutions of a non homogeneous linear equation, then
y1 − y2 is a solution of the homogeneous part. Applying this to the present situation we get
y1 −y2 , y2 −y3 are solutions of the homogeneous part. We then check that these two solutions are
independent (by computing the Wronskian, for instance or otherwise). Therefore, the general
solution of the given equation may be written as

c1 (y1 − y2 ) + c2 (y2 − y3 ) + y1 .

Q.8 Similar to Q.4, we first write down the expression for ψ in each case. The second solution of
the homogeneous equation (independent of the first one) is given by y2 = y1 ψ. We then apply
Q. 6 to obtain a particular solution Yp of the inhomogeneous equation and then the general
solution of the given equation can be written; y = c1 y1 + c2 y2 + yp .

4
 2x
(i) y1 = x; p = ; r = x.
1 + x2
Z x
1 + t2 1
ψ(x) = 2
dt = x − .
t x
y2 = x2 − 1, W = x2 + 1.
yp = −y1 x yW2 r dt + y2 x yW1 r dt
R R
2 2
= −x x (tt2−1)t dt + (x2 − 1) x t2t+1 dt
R R
+1
= x ln(x2 + 1) − (x2 − 1) tan−1 x + 12 x(x2 − 2).
1
(ii) y1 = ex ; p = − ; r = x.
x
Z x
t e−2x (2x + 1)
ψ(x) = dt = − .
e2t 4
y2 = e−x (2x + 1); W = 4x.
yp = − 14 (−ex x (2t + 1)e−t dt + (2x + 1)e−x
R Rx t
e dt)
= −(x + 1).
x+1 e2x
(iii) y1 = e2x ; p = −4 ; r= .
2x + 1 2x + 1
Z x 2t
e (2t + 1)
ψ(x) = dt = −e−2x (x + 1).
e4t
Therefore y2 = x + 1; W = −e2x (2x + 1).
e (t+1) 2t 2t 2t
yp = e2x x (2t+1)e2t (2t+1) dt − (x + 1)
x e e
R R
(2t+1)(2t+1)e2t
dt
x 2t
1
= 14 ln(2x + 1) − 4(2x+1) − (x + 1) (2t+1) e
R
2 dt

x2 + 2x − 2 2 1 (x − 1)ex
 
2
(iv) y1 = x ; p = − = − 1 + + ; r(x) = .
x2 − x x x−1 x
Z x t 2
e t (t − 1) ex
ψ(x) = dt = .
t4 x
Therefore y2 = xex ; W = x2 (x − 1)ex ; and
t t R x t2 (t−1)et
yp = −x2 x tet2 (t−1)e
(t−1)e x
R
t dt + xe t2 (t−1)et
dt
t .
= x2 ex − x2 x et2 dt
R

Q.9 In each case, substitute y = ux in the given ODE and observe that the u-term vanishes. Now
put u0 = v to get an ODE of lower order:

y = ux; y 0 = u0 x + u; y 00 = u00 x + 2u0 ; y 000 = u000 x + 3u00 .

(i) x4 v 00 + (3x3 − 3x2 )v 0 − x4 v = 0.
(ii) xv 00 + (x3 + x + 3)v 0 + (2 + 2x2 − 2x3 )v = 0.

5
Q.10 We shall solve this exercise using M.U.C.
(i) (D2 + 1)2 y = sin x. The auxiliary equation is (λ2 + 1)2 = 0 with roots ±i repeated twice.
Therefore the complementary function (i.e., the general solution of the homogeneous part) is:

yc = c1 cos x + c2 sin x + c3 x cos x + c4 x sin x.

Now we find a linear operator L such that L(sin x) = 0. The simplest one is L = D2 + 1. It
follows that if yp is a solution of (i), then

(D2 + 1)3 (yp ) = 0.

The fundamental system

{cos x, sin x, x cos x, x sin x}
for the operator (D2 + 1)2 is extended to a fundamental system

{cos x, sin x, x cos x, x sin x, x2 cos x, x2 sin x}

for (D2 + 1)3 . Therefore, yp can be chosen to be of the form

yp = Ax2 cos x + Bx2 sin x.

We now compute (D2 + 1)2 (yp ) and determine the values of A, B.

D(x sin x) = x cos x + sin x.

D(x cos x) = −x sin x + cos x
2
(D + 1)(x sin x) = 2 cos x
(D2 + 1)(x cos x) = −2 sin x
(D2 + 1)(x2 sin x) = 4x cos x + 2 sin x
(D2 + 1)2 (x2 sin x) = −8 sin x.
(D2 + 1)(x2 cos x) = 2 cos x − 4x sin x
(D2 + 1)2 (x2 cos x) = −8 cos x
1
Therefore sin x = −8(A cos x + B sin x) from which we get on comparing coefficients A = 0, B = −
8
1 2
and hence yp = − x sin x.
8

[Alternate Method(for computing A, B):

By operator method, a particular solution is found by
1 1
yp = r(x) = sin x
F (D) (D + 1)2
2

Since ı is a repeated root of the auxiliary equation, we directly try with yp = cx2 sin x; (the
computation above justifies this) and see that c = −1/8.]

6
 (ii) Here we have (D − 1)3 (D + 2)y = xex + 3e−2x . So, the complementary function is:

yc = c1 ex + c2 xex + c3 x2 ex + c4 e−2x .

To compute yp we use the following formula:

(D − a)j (xk eax ) = k(k − 1) · · · (k − j + 1)xk−j eax ; k ≥ j ≥ 1

(D − a)j (xk eax ) = 0; 0 ≤ k < j.
(D − a)j (ebx ) = (b − a)j ebx ; a 6= b.
(D − a)xk ebx = [kxk−1 + (b − a)]ebx ; a 6= b, k ≥ 1.

So, we try yp = αx4 ex + βex − γxe−2x . Direct computations yield: α = 1/72, β = −1/9 and
γ = −1/9.

Q.11 In the homogeneous part, we put y = xλ , get the auxiliary equation and solve it for λ to obtain
the complementary solutions. We then try yp = cx3 and determine c.
(i) F (D)y := (x2 D2 + 2xD + 1)y = x3 .
The auxiliary equation is
m2 + m + 1 = 0 with ω, ω̄ as roots.
Since D4 x3 = 0, the MUC suggests that yp = ax3 + bx2 + cx + d. However, the homogeneous
nature of the Cauchy operator F (D) shows that F (D)(xk ) = αk xk for all k. Hence we may
directly take yp = ax3 . Now F (D)(ax3 ) = 13x3 . Therefore yp = x3 /13.
(ii) Here we have
F (D) = x4 D4 + 8x3 D3 + 16x2 D2 + 8xD + 1.
The auxiliary equation is:
(m2 + m + 1)2 = 0
with ω, ω̄ as repeated roots. So, yh = k1 eωx + k2 eω̄x + k3 xeωx + k4 2eω̄x . As in the previous case,
we try yp = cx3 . Since F (D)(cx3 ) = 121x3 we take yp = x3 /121.

Q.12 (i) If yp (x) = c1 (x)y1 (x) + c2 (x)y2 (x), where y1 (x) and y2 (x) are solutions of the correspond-
ing homogeneous
 4
equation

then we have y1 = x3 , y2 = x−2 , c1 = −1 5
(x−1 + 1) ln x and
3
c2 = −15
x
4
ln x − x3 .
(ii) With notation as in (i), y1 = x−3 , y2 = x2 , c1 = 12x5 , c2 = −x2 .

Q.13 Routine.

Q.14 The Wronskian

R W of y1 = x2 ex and y2 = x3 ex is W = x4 ex . From (4.2) we know that
W = e− pdx , whence it follows that p(x) = −2 − 4/x. Substituting y1 in the homogeneous
equation given in (4.2) we see that q(x) = 1 + 6/x2 .
Any constant coefficient homogeneous differential equation with x2 ex as a solution must also
have ex and xex as solutions. Thus its general solution cannot have the form c1 x2 ex + c2 x3 ex .

7
 Solutions To Tut-Sheet 5

Q.1 (i) and (ii) A direct integration by parts would be tedious. Instead, differentiate with respect
d
to ω and invert the order of L and in
dω
s ω
L(cos ωt) = 2 2
, L(sin ωt) = 2 .
s +ω s + ω2
to get the formulae

s2 − ω 2 2sω
L(t cos ωt) = , L(t sin ωt) = .
(s2 + ω 2 )2 (s2 + ω 2 )2

(iii) Work in the half plane s > 0.

Z ∞
−t
F (s) = L(e 2
sin t) = e−t(s+1) sin2 tdt
0

Then we get
Z ∞
1 1 s
2F (s − 1) = e−ts (1 − cos 2t)dt = − L(cos 2t) = − 2
0 s s s +4
From which follows
1 s+1
L(e−t sin2 t) = − 2
2(s + 1) 2(s + 2s + 5)
(iv) Assume a to be real1 . Then putting t(s + a) = u in the defining integral,
Z ∞ Z ∞
1 Γ(3) 2!
L(t2 e−at ) = t2 e−t(s+a) dt = u 2 −u
e du = =
0 (s + a)3 0 (s + a)3 (s + a)3

More generally for complex a and b the following holds:

Γ(a + 1)
L(ta e−bt ) = , (∗)
(s + b)a+1

Γ(a + 1)
(v)-(viii) Use the result L(ta e−bt ) = , by writing the trigonometric and hyperbolic
(s + b)a+1
functions in terms of exponentials.
Z ∞
(ix) L(t2 e−at sin bt) = t2 sin bte−(a+s)t dtF (a + s) where F (s) is the Laplace transform of
0
2 2b(3s2 − b2 )
t sin bt. But then (see note below) F (s) = so that
(s2 + b2 )2

2b(3(s + a)2 − b2 )
L(t2 e−at sin bt) =
(s + a)2 + b2
1
The result is true for complex a but the proof given does not go through. One must use Cauchy’s integral formula
to deform the path of the integral into the complex domain. The result however may be used even for complex a

8
 Note: To obtain F (s) differentiate twice with respect to the parameter b in the formula for
L(sin bt).
(xi) Starting with the defining formula for hyperbolic cosine,
1 
L(cosh at cos at) = L(eat cos at) + L(eat sin at)
2
s−a
If F (s) = L(cos at) then F (s−a) = L(eat cos at), so that L(eat cos at) = . Similarly
(s − a)2 + a2
compute L(eat sin at)

Q.2 (i) Partial Fraction decomposition gives

s2 − ω 2 A1 A2 B1 B2
2 2 2
= + 2
+ +
(s + ω ) s + iω (s + iω) S − iω (s − iω)2
A calculation shows that A1 = B1 = 0 and A2 = B2 = 1/2. Hence
 s2 − ω 2  1 h −1  1   1 i 1
L−1 2 2 2
= L 2
+ L −1
2
= (xe−iωx + xeiωx ) = x cos ωx
(s + ω ) 2 (s + iω) (s − iω) 2
(iii) The same procedure gives
1 A1 A2 B1 B2
= + + +
(s2 + ω 2 )2 s + iω (s + iω)2 S − iω (s − iω)2
−1 i
with B2 = A2 = and −B1 = A1 = whereby
4ω 4ω 3
 1  i x sin ωx x cos ωx
L−1 2 2 2
= 3
(e−iωx − eiωx ) − 2 (2 cos ωx) = −
(s + ω ) 4ω 4ω 2ω 3 2ω 2
(iv) Proceeding as before,

s3 1h 1 1 1 1 i
= + + +
s4 + 4a4 4 s + (a − ia) s + (a + ia) s − (a − ia) s − (a + ia)
whereby

−1
 s3  1 h −ax+iax −ax−iax ax−iax ax+iax i cos ax −ax ax
L = e +e +e +e = (e +e ) = cos ax cosh ax
s4 + 4a4 4 2
(v) In this case the decomposition is
s−2 1h1 1 1 3 i
= − − −
s2 (s + 4)2 8 s s2 s + 4 (s + 4)2
and the inverse Laplace transform is given by
 s − 2  1h i
L−1 = 1 − x − e −4x
− 3xe −4x
s2 (s + 4)2 8

9
 (vi) In this case the decomposition is
1 −1 1 1 1 1
= − − − +
s4 − 2s3 8s s2 4s2 2s3 8(s − 2)

and the inverse Laplace transform is

 1  −1 x x2 e2x
L−1 = − − +
s4 − 2s3 8 4 4 8
x3 x sin πx
(vii) Ans: 2
− 2+
6π π π5
(viii) The partial fraction decomposition is

s 2 + a2 1h 1 1 i
= +
(s2 − a2 )2 2 (s − a)2 (s + a)2

so that
−1
 s2 + a2  x ax
L 2 2 2
= (e + e−ax ) = x cosh ax.
(s − a ) 2
(ix) The partial fraction decomposition is

s3 + 3s2 − s − 3 1h 1 1 i i i
= + + −
(s2 + 2s + 5)2 2 s + 1 − 2i s + 1 + 2i (s + 1 − 2i)2 (s + 1 + 2i)2

and so
 s3+ 3s2 − s − 3  1 h (−1+2i)x i
−1 (−1−2i)x (−1+2i)x (−1−2i)x
L = e + e + 2ixe − 2ixe
(s2 + 2s + 5)2 2
= ex (cos 2x − 2x sin 2x)

(x) Ans: ex (1 − x − xe−x )

Comments: The Laplace transform method is superior to other methods inasmuch as the initial
conditions are incorporated at the outset and the problem of determining the constants of integra-
tion is absent. This is particularly advantageous in the case of systems of differential equations
with constant coefficients. Moreover the convolution theorem gives an integral representation of the
solution.

Q.3 (i) Laplace tranforming the DE and using the initial conditions we get
3
(s2 + 1)Ly =
s2 +9
and so
−3i h 1 1 i i h 1 1 i
Ly = − + −
16 s − i s + i 16 s − 3i s + 3i

10
 On taking the inverse Laplace transforms,
3 1
y= sin x − sin 3x.
8 8
The procedure is similar for the remaining bits. We merely record the answers.
(ii) Ans: y = te−t − e−t + e−2t
(iii) Ans: y = 2e2t − e−4t
1 1
(iv) Ans: y(x) = 3e−x + 3xe−x + sin x − xe−x
2 2
x
(v) Ans: y(x) = e cos 2x + 2 sin x
1 1 5
(vi) y(x) = − e3x − e−x − e2x + e−2x
2 6 3
Q.4 The procedure is routine. Question 17 is similar and is to be done by the student entirely by
himself. Lower case letters refer to the unkowns and the corresponding upper case letter refers
to the Laplca transform. This notation is standard and used extensively in books. Thus Y (s)
is the Laplace transform of y(t). First Laplace transform the given system. Solve the resulting
algebraic system of linear equations and proceed to compute the inverse Laplace transform.
The initial conditions are incorporated in the solution procedure.
(i) sX − x(0) = X + Y, sY − y(0) = 4X + Y . WE take x(0) = a and y(0) = b and solve for X
and Y :
a(s − 1) + b 4a + (s − 1)b
X= , Y =
(s − 3)(s + 1) (s − 3)(s + 1)
A partial fraction decomposition gives
 2a + b  1  2a − b  1
X= +
4 s−3 4 s+1
Taking the inverse Laplace transform gives
 2a + b   2a − b 
x(t) = e3t + e−t
4 4
A simple computation shows
 2a + b   b − 2a 
y(t) = e3t + e−t
2 2
(ii), (iii) and (iv) are similar.
2s
(v) sY1 + Y2 − y1 (0) = 2 and Y1 + sY2 − y2 (0) = 0. Putting in the given initial conditions
s +1
s
and solving Y2 = 2 whence y2 (t) = cos t and y1 = −y20 = sin t.
s +1

11
 (vi) Laplace transforming the equations we get the pair:

−5s s3 + s2 − s + 4
s2 Y1 + Y2 − sy1 (0) − y10 (0) = , s2 Y1 + Y2 =
s2 + 4 s2 + 4
5s −s + s2 + s + 4
3
s2 Y2 + Y1 − sy2 (0) − y20 = 2 , 2
Y1 + s Y2 =
s +4 s2 + 4

Either solve directly for Y1 and Y2 or else in this case it is easier to add and subtract to get
2
Y1 + Y2 = , y1 (t) + y2 (t) = 2 sin t
+1 s2
2s
Y1 − Y2 = 2 , y1 (t) − y2 (t) = 2 cos 2t
s +4

and we get from these the solutions

y1 (t) = sin t + cos 2t, y2 (t) = sin t − cos 2t.

(vii) Laplace Transforming the system of ODEs we get

2Y1 − Y − 2 − Y3 = 0,
4 2
Y1 + Y2 = 3 + 2 ,
s s
2 2
sY1 + Y3 = 3 +
s s

On solving,
1 h6 2 2i 2 2
Y1 =
3
+ 2
+ , y1 (t) = + t2 − e−3t
s+3 s s s 3 3
0 0 −3t 0 2
Now y2 = 4t + 2 − y1 = 2t + 2 − 2e and finally y3 (t) = −y2 + t + 2 from which y3 (t) may
be computed.
(viii) Ans: y1 (t) = e2t + et , y2 (t) = e2t .
∞ ∞ ∞ j
 1   1  1 
−1 1
 X 1 X 1  X t
Q.5 (i) L−1 = L−1 1 = L = L −1
= = et
s−1 s 1− s s j=0 sj j=0 s j+1
j=0 j!
∞ ∞
 1  X (−1)j  X (−1)j t2j+1
(ii) L−1 = L−1
= = sin t
s2 + 1 j=0 s
2j+2
j=0 (2j + 1)!
1  ∞
X (−1)j bj  X ∞
(−1)j bj tj √
(iii) L−1 e−b/s = L−1 j+1
= = J0 (2 bt)
s j=0 j!s j=0 (j!)2

12
 √
(iv) We use the binomial theorem to expand s2 + a2 into an infinite series.
∞
 1 
−1 1
 X (−1)j (1 · 3 · 5 · · · (2j − 1))a2j 
L−1 √ 2 = L
s + a2 s j=0 2j j!s2j
∞
1 X (−1)j (2j)!a2j 
= L−1
s j=0 4j (j!)2 s2j
∞
X (−1)j (ax)2j
= = J0 (at)
j=0 4j j!2
 e−b/s  ∞
(−1)j bj −1  1 
(v) L−1 . Now using the formula L(xp ) = Γ(p + 1)/sp+1 we get
X
√ = L j+ 12
s j=0 j! s
1 1 1
−1
 1  xj− 2 2j xj− 2 4j j!xj− 2
L = = √ = √
Γ(j + 12 )
1
sj+ 2 (2j − 1)(2j − 3) . . . 3 · 1 π (2j)! π
Hence
 e−b/s  1 √
L−1 √ = √ cos(2 bx)
s πx
(vi) Recalling from the chapter on power series, the series development for the inverse tangent
function, we get
∞ ∞
  X (−1)j 1  X (−1)j x2j sin x
L−1 tan−1 (1/s) = L−1 2j+1
= =
j=0 2j + 1 s j=0 (2j + 1)! x

Q.6 (i) Either an L-type or an I-type problem. The purpose is to derive the formula
Z p
1
L(f ) = e−st f (t)dt (Π)
1 − e−ps 0
for a periodic function of period p. The importance of this is not so much the computation
of Laplace transforms but to note that, in conjunction with Fourier series, it provides the
classical partial fraction expansions (Mittag-Leffler formulae) for the functions such as coth x.
For instance, compute the Fourier series for the square wave defined by f (x) = sgn(x) on
[−π, π] and Laplace transform the resulting expansion. The following Mittag-Leffler identity
pops out (!):
∞
1 πs 4X 1
tanh =
s 2 π n=0 (2n + 1)2 + s2
Now replace s by 2is and we get (!!)
∞
X 8s
π tan πs = 2 2
n=0 (2n + 1) − 4s

Term by term integration in turn yields the factorization (!!!)

∞ 
Y 4s2 
cos πs = 1−
n=0 (2n + 1)2

13
 We stop here to keep the iterated factorials down to three.
(ii) Period for | sin ωt| equals p = π/ω.
Z p Z π/ω
−st ω
f (t)e dt = e−st sin ωtdt = 2 2
(1 + e−πs/ω )
0 0 ω +s
Now substitute in the formula (Π).
(iii) This is the example of the square wave discussed in (i) above. Period is 2π and
1 Z 2π
−st 1 (1 − e−πs )2 1  1 − e−πs  1 πs
L(f ) = −2πs
f (t)e dt = −2πs
= −πs
= tanh .
1−e 0 s 1−e s 1+e s 2
(iv) Period p = 2 and
1 Z 2
−st 1 (1 − e−s )2 1 s
L(f ) = −2s
f (t)e dt = −2s 2
= 2 coth
1−e 0 1−e s s 2
(v) Period p = 2π.
1 Z 2π
−st 1 1 + e−πs 1
L(f ) = −2πs
f (t)e dt = 2 −2πs
= 2
1−e 0 s +11−e (s + 1)(1 − e−πs )
Q.7 The defining integral has to be broken up into pieces:
∞ Z n ∞
X
−st es − 1 X −ns es (es − 1) es
L(f ) = ne dt = ne = =
n=1 n−1 s n=1 s(es − 1)2 s(es − 1)

Q.8 Define u(t − a) = 0 if t ≤ a and u(t − a) = 1 if t > a. Then L(u(t − a)) = e−as /s. Assume
a > 0 and let us solve
dv
= u(t − a), v(0) = 0.
dt
The solution is given by va (t) = 0 if t < a and va (t) = t − a if t ≥ a. Now
 dv 
a
L = L(u(t − a)) = e−as /s.
dt
sL(va ) = e−as /s
e−as
which gives us the requisite formula L(va ) = 2 . Now,
s
 e−s e −2s
e −3s
e −4s 
f (t) = L−1 2 − 2 − 2 + 2 = v−1 (t) − v−2 (t) − v−3 (t) + v−4 (t)
s s s s
Unraveling the right hand side gets us the answer:
f (t) = 0 if t < 1
= t−1 if 1 ≤ t < 2
= 1 if 2 ≤ t < 3
= 4−t if 3 ≤ t < 4
= 0 if t ≥ 4

14
 Q.9 (i) We have to compute the Laplce transform of u(t − π) sin t. Write sin t = − sin(t − π) and
use the shifting theorem.
L(u(t − π) sin t) = −e−πs L(sin t) = −e−πs /(s2 + 1)
(ii) Writing e−2t as e2 e−2(t−1) prepares us to apply the shifting theorem.
L(u(t − 1)) = e2 L(u(t − 1)e−2(t−1) ) = e2−s L(e−2t ) = e2−s /(s + 2).

Q.10 First observe that

∞
s + a  a − b X (−1)j  a − b j+1
ln ( = ln 1 + =
s+b s+b j=0 j + 1 s + b

The series converges for s > |a − b| + |b| and so

∞
s + a (−1)j (a − b)j+1
L(e−bx xj )
X
ln =
s+b j=0 (j + 1)!
Hence
∞
h  s + a i e−bx X (x(b − a))j+1 e−bx − e−ax
L−1 ln =− =
s+b x j=0 (j + 1)! x
Likewise  s + a  s + c i
h 1 −bx
L−1 ln (e − e−ax + e−dx − e−cx )
=
s+b s+d x
In this example a = −2 + i, c = −2 − i, b = −1 + 2i and d = −1 − 2i.

Q.11 The first part is the convolution theoerm for which the text of Kreyszig may be referred to.
For the second part, let
1 1
G(s) = 2 2
= L(e−at sin at)
(s + a) + a a
so that by the convolution theorem,
1 Z t −a(t−u)
L−1 (F (s)G(s)) = f (t) ∗ g(t) = e sin a(t − u)f (u)du
a 0
Q.12 In this problem we see an application to linear differential equations with variable coefficients.
We compute the Fourier transform of the Bessel’s function. Use formula (1) on p275 of Kreyszig
(Eighth edition) and we get
d 2
− (s Y − sy(0) − y 0 (0)) + sY − k − Y 0 = 0
ds
which gives the first order ODE for Y (s):
(s2 + 1)Y 0 + sY = 0
which integrates to √
Y = C/ s2 + 1
√
The value of the constant is seen to be one and Y = 1/ s2 + 1.

15
Q.13 Note that the convolution is given by
Z ∞
G(λ) = ta−1 u(t)(λ − t)b−1 u(λ − t)dt
−∞

The presence of the factor u(t) reduces the integral to

Z ∞
G(λ) = ta−1 (λ − t)b−1 u(λ − t)dt
0

It is clear the convolution is zero if λ < 0 so that G(λ) = u(λ)G(λ). For t > λ the integrand
in the last integral vanishes and
Z t Z 1
G(λ) = u(λ) ta−1 (λ − t)b−1 dt = u(λ)λa+b−1 ta−1 (1 − t)b−1 dt = λa+b−1 u(λ)B(a, b).
0 0

In other words,
u(t)ta−1 ∗ u(t)tb−1 = u(t)ta+b−1 B(a, b) (III)
Let us compute the Laplace transform of u(t)tc−1 :
Z ∞
Γ(c)
L(u(t)t c−1
)= tc−1 e−st dt =
0 sc
Applying the convolution theorem to (III) we see that
Γ(a) Γ(b) Γ(a + b)
a b
= B(a, b)
s s sa+b
from which we get the fundamental relation

Γ(a)Γ(b) = Γ(a + b)B(a, b)

Q.14 Denoting the Laplace transform of f by F (s) we see that

1
L(f ) · L(f ) = (F (s))2 =
s2 + 1
By convolution theorem,
1
L(f ∗ f ) = (F (s))2 =
s2 +1
so that f ∗ f = sin x.

Q.15 Define f (t) = 0 for t < 0 and taking the Laplace transform of f (t) we get
Z ∞ sin tx  Z ∞ 1 Z ∞
dx 1π
F (s) = L dx = L(sin tx)dx = 2 2
=
0 x 0 x 0 s +x s2
π
so that f (t) = 2
giving the formula
Z ∞
sin tx π
dx = , t > 0.
0 x 2

16
 The case t < 0 can be reduced to the above case.
(ii) The value of the integral is independent of the sign of t and define f (t) = 0 for t < 0 and
for t > 0 take f (t) to be the integral to be computed. Then as before,
Z ∞ Z ∞
dx sdx πs n1 1o π
F (s) = 2 2
L(cos tx) = 2 2 2 2
= 2 2
− =
0 x +a 0 (x + a )(s + x ) 2(s − a ) a s 2a(s + a)

Taking the inverse Laplace transform we get

Z ∞
cos xtdx π
2 2
= e−at
0 x +a 2a

Q.16 Solving Abel’s integral equation:

√
Note that the integral is the convolution of y(t)u(t) and (1/ t)u(t). By the convolution
theorem,
A q
= Y (s) π/s
s
which gives
A A
Y (s) = √ , y(t) = √
πs π t
Observe that this is really a special case of Q.13.

Q.17 See the comments in Q.4.

Q.18-21 Cancelled.

17