Answers of the exam of November 1, 2013

Differential equations (wi2180LR)

1. Question:
Use the Laplace transform to solve the following initial value problem:

y + 4y = (t 10) + 1 , y(0) = 1, y (0) = 0.

Answer:
The transformed equation is
1
s2 Y (s) sy(0) y (0) + 4Y (s) = e10s + ,
s
which, after substitution of the initial conditions, gives
1
s2 Y (s) s + 4Y (s) = e10s + .
s
The transformed solution is therefore
e10s 1 s
Y (s) = + + .
s2 + 4 s(s2 + 4) s2 + 4

Apply the technique of partial fractions to simplify the second term:

1 a bs + c
= + 2 ,
(s2 + 4)s s s +4

and hence
a(s2 + 4) + (bs + c)s = 1 .
Collecting terms of like power gives the following four equations

a + b = 0, c=0,

4a = 1.
1
Solving these systems gives c = 0, a = 4, and b = 41 . The transformed
solution can now be written as
e10s 3 s 1
Y (s) = + + .
s2 + 4 4 s2 + 4 4s

Taking the inverse transform term by term gives

1 3 1
y(t) = u10 (t) sin(2(t 10)) + cos(2t) + .
2 4 4

2. Question:
Find the general solution of the following system of differential equations
   
2 1 0
x = x+ .
3 2 2et

Answer:
First solve the homogeneous equation. Put x = ert . Substitution in the
homogeneous equation gives the eigenvalue problem
 
2 1
= r .
3 2
The characteristic equation is

(2 r)(2 + r) + 3 = 0 r2 1 = 0

so
r1 = 1 and r2 = 1
with corresponding eigenvectors
   
1 1
1 = and 2 =
1 3
The general solution of the homogeneous equation reads
   
1 t 1
xh = c1 e + c2 et .
1 3
A particular solution of the inhomogeneous problem can be found using the
method of undetermined coefficients (alternative techniques are diagonali-
sation and variation of parameters). Put

xp = (a + bt) et .

Substitution in the system of differential equations gives

 
t t t 0
(a + bt) e + be = A (a + bt) e + et .
2
in which  
2 1
A= .
3 2
Collecting the terms of the form t0 et gives
 
0
Aa = a + b 2
1

and collecting the terms of the form t1 et yields

b = Ab .

This last equation states that b is an eigenvector with eigenvalue 1, hence

 
1
b= ,
1
with any nonzero realnumber. From the first equation we infer, after the
1
substitution b =
1
 

(A I)a = ,
2
where I denotes the 2 2 identity matrix. Thissystem only has solutions if
a1
= 1. In that case the solution for a = reads
a2
     
1 + a2 1 1
a= = a2 +
a2 1 0
We can choose any value for a2 , as we only need one particular solution. If
we choose a2 = 0, the particular solution xp (t) is
   
1 1
xp (t) = +t et .
0 1
 The general solution now becomes
       
1 1 1 1
x = xh + xp = c1 e t + c2 et + t et
1 3 0 1

3. Question:
Consider the following system of nonlinear equations
 dx
dt = y
dy
dt = sin x y
3 3


i) Find all critical points with x-coordinates in the interval 2 , 2
ii) Investigate type and stability of the critical points.
iii) Sketch the phase portrait in the neighbourhood of the critical points;
argue and indicate the direction of motion.

Answer:
i) The equilibria satisfy x = y = 0. Therefore we find y = 0 and sin x = 0,
which implies that the
3

equilibria are (n, 0) for n = , 2, 1, 0, 1, 2, . In
the interval 3
2 , 2 there are three equilibria: (0, 0), (, 0), (, 0).
ii) Define the vector function
   
F (x, y) y
=
G(x, y) sin x y

The Jacobian matrix of this vector field is:

 
0 1
J(x, y) = .
cos x 1

At the critical point (0,0), the coefficient matrix of the linearized system is
 
0 1
J(0, 0) =
1 1

with eigenvalues 21 i 21 3. As the real parts of the eigenvalues are negative
the equilibrium (0, 0) corresponds to an (asymptotically) stable spiral point.
At the critical points (, 0), the coefficient matrix of the linearized system is
 
0 1
J(, 0) =
1 1

with eigenvalues r1,2 = 12 12 5. As r1 > 0 and r2 < 0, the critical points
(, 0) are both (unstable) saddle points.
The phase portraits in the neighbourhood of the equilibria are shown in the
figure below

Figure 1: Phase portraits of the equilibria. One focus and two saddles.
4. Question:
Find the formal solution (using the method of separation of variables) of the
heat equation
ut = uxx + 2 , 0 < x < 1, t > 0 ,
with nonhomogeneous boundary conditions

u(0, t) = 0 and u(1, t) = 10, t > 0 ,

and with initial condition

u(x, 0) = 4 , 0x1.

i) Compute the steady state part of the nonhomogeneous solutions v(x)

ii) Calculate the transient part w(x, t). Integrals arsing in this calculation
need not to be computed, but their expressions should be written down.

Answer:
Set u(x, t) = w(x, t) + v(x), where v(x) is a stationary solution. If we substi-
tute this in the heat equation, we find

ut = uxx + 2. wt = wxx + vxx + 2.

As v(x) is per definition a solution of the nonhomogeneous heat equation in

which time dependence is absent, it obeys

vxx + 2 = 0.

Integrating this differential two times with respect to x gives

v(x) = x2 + Ax + B.

As v(x) satisfies the boundary conditions imposed on the nonhomogeneous

heat equation, that is,
v(0) = 0, v(1) = 10.
We find that A = 11 and B = 0. Moreover, this immediately implies that
w(0, t) = 0 and w(1, t) = 0, so the transient solution w(x, t) is a solution of
the homogeneous heat equation with homogeneous boundary conditions.
Writing w(x, t) = X(x)T (t), and substituting this in the homogeneous heat
equation we find
T X = X T,
which can be rewritten as
X T
= = .
X T
As the left-hand side of this equation depends only on x and the T /T term
only on t, both sides must equal a constant value, say .
This leads after introduction of the separation constant to the spatial
equation
X + X = 0
with boundary conditions

X(0) = 0 and X(1) = 0

and to the time equation

T + T = 0 .
 The spatial equation is an eigenvalue problem. We consider three cases:
< 0 , = 0 , and > 0. First < 0. With = 2 the equation
becomes
X 2 X = 0 .
The general solution of this equation is

X = c1 ex + c2 ex .

Substitution of the b.c. X(0) = 0 yields c1 = c2 . The b.c. X(1) = 0 gives

c1 e + c2 e = 0. With c1 = c2 , this equation can only be satisfied if = 0
(not allowed) or if c1 = c2 = 0 (trivial solution).
For = 0 we get the general solution X = ax + b. From both X(0) = 0
and X(1) = 0 follows that a = b = 0. So = 0 does only allow the trivial
solution.
The third case to consider is > 0. Putting = 2 , we get the equation

X + 2 X = 0

The general solution of this equation is

X = c1 cos(x) + c2 sin(x) .

From the b.c X(0) = 0 follows c1 = 0. From the b.c. X(1) = 0, it follows that
sin() = 0, so we choose = n, n = 1, 2, .
In conclusion, the eigenvalues are n = (n)2 , n = 1, and the correspond-
ing eigenfunctions are Xn = sin( nx
L ), n = 0, 1, .
Solving the equation for T for n yields
n 2
Tn = kn e( L ) t

Hence we get the following fundamental solutions

2
t)
wn (x, t) = e(n) sin(nx) , n = 1, 2,

In order to satisfy the initial condition we write down the general solution:

X 2
u(x, t) = v(x) + cn e(n) t sin(nx) .
n=1

Since u(x, 0) = 4, the cn s are such that

X
4 = x2 + 11x + cn sin(nx)
n=1

Using the orthogonality properties of the cosine functions on [0, 1] it follows

that
Z1
cn = 2 (4 + x2 11x) sin(nx)dx , n = 1, 2, .
0

5. Question:
Consider the following function

f (x) = x , 0x2.

Let g(x) be the even periodic extension of f (x) such that g(x + 4) = g(x).
i) Sketch the graph of g(x) on the interval (4, 4)
 ii) Find the Fourier series for g(x).

Answer:

i) The function g(x) is given by


x 2 x 0
g(x) = g(x + 4) = g(x).
x 0x2

The period is 4, hence L = 2. The function is continuous on R.

g(x)

4 2 0 2 4 x

Figure 2: Graph of g(x)

ii) g(x) is even, and therefore has the Fourier series

a0 X nx
g(x) = + an cos
2 n=1
L

with
ZL
2 nx
an = g(x) cos dx, n = 0, 1, 2,
L L
0

Hence
Z2  2
nx 2
an = x cos dx an = (cos n 1), n = 1, 2, .
2 n
0

For n = 0 we find a0 = 2.
Note that cos n = 1 for n even. Take n 2n 1. Then the Fourier
series for g(x) can be written as
 
X 8 (2n 1)x
g(x) = 1 2 (2n 1)2
cos .
n=1
2

6. Question:

i) Determine the normalized eigenfunctions of

y + y = 0 , y (0) = 0, y(1) = 0 .

ii) Determine the solution of the following boundary value problem

3x
y + ( )2 y = sin( ), y (0) = 0, y(1) = 0 .
2 2
or else show that it has no solution.
Answer:
is real valued. Examine the three cases < 0, = 0, and > 0. First < 0.
With = 2 the equation becomes

y 2 y = 0

The general solution of this equation is

y = c1 ex + c2 ex .

Substitution of the b.c. y (0) = 0 yields c1 = c2 . The b.c. y(1) = 0 gives

c1 e + c2 e = 0. With c1 = c2 we get c1 e2 = 0, which is only the case for
c1 = c2 = 0. Hence for < 0 there are no nontrivial solutions.
For = 0 we get the general solution y = c1 x + c2 . From y (0) = 0 follows
that c1 = 0 and from y(1) = 0 that c2 = 0. So also in this case there is no
nontrivial solution.
The third case to consider is > 0. Putting = 2 , we get the equation

y + 2 y = 0 .

The general solution of this equation is

y = c1 cos(x) + c2 sin(x) .

For the derivative we have

y = c1 sin(x) + c2 cos(x) .

From the b.c y (0) = 0 follows c2 = 0. From the b.c. y(1) = 0 follows
cos() = 0, which is true if = (n 12 ), n = 1, 2, . So the non-normalized
eigenfunctions are yn = An cos((n 12 )x), n = 1, 2, 3, .
The normalization constant An must be chosen such that
Z1
1
[An cos((n )x)]2 dx = 1
2
0

which yields An = 2.
So the normalized eigenfunctions n (x) are
1
n (x) = 2 cos((n )x), n = 1, 2, 3, .
2

ii) Write the solution as

X
y(x) = bn n (x)
n=1

and substitute in the equation. This gives

X 3x
bn n (x) + ( )2 bn (x) = sin( ).
n=1
2 2

The term sin( 3x

2 ) can also be expanded in eigenfunctions n (x), which gives
 
3x X
sin = cn n (x),
2 n=1
with cn given by

2
c2n =
(2n + 1)

2
c2n1 =
(2n 3)




X 1 2 2 2 1
2bn ((n ) + ( ) ) cn cos((n )x) = 0.
n=1
2 2 2

As the n (x) are linearly independent, the coefficients for each of the n (x)
have to vanish. For n = 1 the term multiplying b1 vanishes, whereas c1 6=0.
Therefore there is no solution of the boundary value problem.
An alternative (standard) technique to find the solution is to compute the
solution to the homogeneous equation yh and a particular solution yp which
gives the general solution y = yp + yh .
x x
yh = c1 sin + c2 cos .
2 2
Computing the particular solution gives

1 3x
yp = sin .
2 2 2
Imposing the boundary conditions gives

1 3
c1 + = 0, c1 =0
2 2 2 4
These two equations for c1 do not have a solution. Therefore the boundary
value problem does not have a solution either.