Solutions to Dynamical Systems 2010 exam. Each question is worth 25 marks.

1. [Unseen]Consider the following 1st order differential equation:

dy
= X(y) = −y(y 2 − 4). (1)
dt
(a) Find and classify all the fixed points of (1). Hence draw the phase diagram associated to (1).
We just need to find the solutions to X(y) = 0, which are y = 0, ±2. In order to classify the fixed
points we have to either look at the sign of the r.h.s. of (1) in the various regions between fixed points
or use the criterium coming from linearisation which means looking at the sign of the derivative of
the r.h.s. of (1) at each of the fixed points. Following this last criterium we find that,
X 0 (y) = −(y 2 − 4) − y(2y) = −3y 2 + 4.
So we find that:
• X 0 (0) = 4 > 0, therefore y = 0 is a repellor,
• X 0 (−2) = −8 > 0, therefore y = −2 is an attractor,
• X 0 (2) = −8 < 0,therefore y = 2 is also an attractor.
[3 marks]
Hence, the phase diagram is

[1 mark]
(b) Solve (1) for each of the following initial conditions:
• y(0) = 3,
• y(0) = −1,
indicating the range of values of t for which each solution is defined
The equation can be solved by separating variables in the usual way:
Z t Z y Z yµ ¶
dy 1 1 1
dt = t − t0 = − =− − +
t0 y y(y − 2)(y + 2) y0 8 (−2 + y) 4 y 8 (2 + y)
¯ 2 0 ¯
1 ¯ 2¯
(y − 4)y0 ¯
= − log ¯¯ 2 ,
8 (y0 − 4)y 2 ¯
with y0 = y(t0 ) the initial condition. This can be solved to
2e4(t−t0 )
y(t) = ± q .
4
e8(t−t0 ) − 1 + y02

The ± sign will be + for y0 > 0 and − for y0 < 0 in order to satisfy the initial condition.
[5 marks]
Let us now solve for each initial condition:
• for y(0) = 3 we have
2e4t
y(t) = q .
5
e8t − 9
5
This solution is well defined only when e8t − 9 > 0. This gives the condition
1 5
t> log = −0.0734733.
8 9
Thus the range of values for which the solution is defined is t ∈ ( 18 log 95 , ∞).
 [4 marks]
• For y(0) = −1 the solution becomes,

2e4t
y(t) = − √ ,
e8t + 3
where we had to introduce a minus sign to guarantee that the initial condition is fulfilled. In
this case, the denominator is never zero and e8t + 3 can never become negative, so the solution
is valid for all values of t. The interval of definition is t ∈ (−∞, ∞).
[4 marks]
(c) Provide a sketch of your solutions against the variable t, clearly showing the behaviour of the various
functions as t → ±∞.
In order to carry out the sketch it is important to study the behaviour of both solutions as t → ±∞.
The solution
2e4t
y(t) = q , y(0) = 3,
5
e8t − 9

is clearly always positive (for those values of t for which it is defined) and as t → ∞ it tends to the
value 2, which is the fixed point we identified before. This was to be expected, as the point is an
attractor. In this case it does not make sense to consider t → −∞ as the solution is only defined for
t ∈ ( 81 log 59 , ∞). Finally, since the denominator vanishes exactly at t = 18 log 95 the function should
diverge as it gets closer to that value.

Figure 1: The red figure corresponds to the initial condition y(0) = 3. As we can see it diverges at a small
negative value of t which was computed above. The blue graph is the solution with y(0) = −1. The green lines
are the fixed points y = ±2.

Concerning the other solution

2e4t
y(t) = − √ , y(0) = −1,
e8t + 3
it is clearly a negative function for all values of t. As t → ∞ it tends to the value -2 which is the
fixed point identified before. We should have expected that, since -2 is an attractor. As t → −∞
the solution tends to the value 0, which is also compatible with our fixed point analysis, as zero was
identified as a repellor. A plot of the two solutions is given above.
[5 marks]
(d) Write down the linearized version of equation (1) about a generic fixed point y = a.
At a fixed point X(a) = 0 and therefore a Taylor expansion of X(y) to first order in y gives

dy
= X 0 (a)(y − a) = (−3a2 + 4)(y − a).
dt

Page 2
 [1 mark]
(e) Use the result of (d) to find the solution of the linearized equation about the smallest of the fixed points
of (1), if the initial condition is y(0) = −3. We already found in (a) that X 0 (−2) = −8 therefore the
linearized equation is
dz
= −8z,
dt
for z = y + 2. The solution to this equation is z = Ae−8t where A is a constant which will be fixed
by the initial condition. The final answer is:

y(t) = −2 − e−8t .

[2 marks]

2. [Seen]Consider the following system of linear differential equations:

dx1 dx2
= ẋ1 = αx1 + γx2 + c1 , = ẋ2 = βx1 + αx2 + c2 , (2)
dt dt
where α, β, γ, c1 , c2 are real constants.
(a) Write the equations in the standard form ẋ = Ax + b. Write down the vector a which is a fixed point
of (2). µ ¶ µ ¶µ ¶ µ ¶
ẋ1 α γ x1 c1
= +
ẋ2 β α x2 c2
[1 mark]
The fixed point is the solution to the vector equation

Ax + b = 0,

that is µ ¶µ ¶ à !
αc1 −γc2
−1 1 α −γ c1 α2 −βγ
x = −A b=− 2 =− −βc1 +αc2
α − βγ −β α c2 α2 −βγ

[4 marks]
(b) Compute the eigenvalues and eigenvectors of the matrix A. The eigenvalues of A are obtained by
finding the zeroes of the characteristic polynomial of A:

|A − λI| = (α − λ)2 − βγ = (α2 − βγ) + λ2 − 2αλ = 0,

the solutions to this equation are

p
2α + 4α2 − 4(α2 − βγ) p
λ1 = = α + βγ,
2
and p
2α − 4α2 − 4(α2 − βγ) p
λ2 = = α − βγ.
2
[3 marks]
The eigenvectors corresponding to these eigenvalues are obtained by solving:
µ ¶µ ¶ µ ¶
α γ a p a
= (α ± βγ) .
β α b b

In components this gives two equations for each eigenvector:

p p
aα + bγ = (α ± βγ)a, aβ + bα = (α ± βγ)b,

Page 3
 simplifying p p
bγ = ± βγa, aβ = ± βγb.
q
β
The two equations are not independent and setting a = 1 this gives b = ± γ. Therefore the
eigenvectors would be: Ã ! Ã !
q 1 1
q
E1 = β , E2 =
γ − βγ

[5 marks]
(c) Define the vector z in terms of which the equation can be re-written in the form ż = Az.
We just need to change variables to
z = x − a,
where a is the fixed point obtained in section (a).
[1 mark]
(d) Indicate the conditions (if any) that α, β and γ must satisfy so that the fixed point is:
• an improper node,
• a focus,
• a star node.
An improper node arises when the eigenvalues of the matrix A are equal and real. In our case the
eigenvalues can only be equal if p
βγ = 0,
which corresponds to β = 0 or γ = 0.
[1 mark]
A focus occurs when the eigenvalues are complex conjugated to each other. This will happen only if
β > 0 and γ < 0 or viceversa.
[1 mark]
Finally, a star node occurs when the matrix A is proportional to the identity matrix. This will only
happen if β = γ = 0.
[1 mark]
(e) If α = 1, β =µ −2 ¶and γ = 2 classify the fixed point. Find the values of c1 and c2 so that the fixed
1
point is a = .
1
We found the fixed point in section (a)
à ! µ c1 −2c2 ¶ µ ¶
αc1 −γc2
α2 −βγ 5 1
a=− −βc1 +αc2 =− 2c1 +c2 = .
α2 −βγ 5
1

This is solved by c1 = −3 and c2 = 1.

[2 marks]
(f ) For α = 1, β = −2 and γ = 2 show that the matrix A is already in its Jordan Normal form. For
these values of α, β, γ and the values of c1 , c2 obtained in (e) find the general solution of equations
(2)
For the values of α, β, γ given in (e), the eigenvalues of A are λ1 = 1 + 2i and λ2 = 1 − 2i. The
Jordan form is therefore: µ ¶
1 2
J= .
−2 1
which coincides with the matrix A itself.
[1 mark]

Page 4
 We will use the change of variables of section (c) so that the equation becomes of the form

ż = Az,

where
z = x − a,
and µ ¶
1
a= ,
1
from section (e). The equation in z can be written in components as:

ż1 = z1 + 2z2 , ż2 = −2z1 + z2 ,

and can be solved by using the standard change of variables

z1 = r cos θ, z2 = r sin θ.

This leads to the equations

ṙ = r, ⇒ r = Aet ,
where A is a constant, and
θ̇ = −2, θ = −2t + B,
where B is an integration constant. Therefore, the general solutions are

z1 = Aet cos(−2t + B), z2 = Aet sin(−2t + B).

Therefore, for the original variables x1 , x2 the general solutions are

x1 = 1 + Aet cos(−2t + B), x2 = 1 + Aet sin(−2t + B).

or, in vector form

µ ¶ µ ¶ µ ¶
1 t 1 t 0
x= + Ae cos(−2t + B) + Ae sin(−2t + B) .
1 0 1

[5 marks]
3. [Seen] Consider the following system of first order nonlinear differential equations:

ẋ1 = x1 (1 − x1 − x2 ), ẋ2 = x2 (1 − x1 − x2 ). (3)

(a) Show that the fixed points of the system are either at the origin or on a line.
In order to find the fixed points we have to solve the simultaneous equations:

x1 (1 − x1 − x2 ) = x2 (1 − x1 − x2 ) = 0,

which give solutions (x1 , x2 ) = (0, 0) and x1 = 1 − x2 which is indeed a line.

[2 marks]
(b) Calculate the Jacobian matrix for a general point (x1 , x2 ) and deduce that the fixed point at the origin
is simple, whereas the fixed points on the line are not.
The Jacobian matrix is obtained as usual by computing the derivatives
à ! µ ¶
∂X1 ∂X1
∂x1 ∂x2 1 − 2x1 − x2 −x1
A(x1 ,x2 ) = ∂X2 ∂X2 =
∂x ∂x
−x2 1 − 2x2 − x1
1 2

with X1 (x1 , x2 ) = x1 (1 − x1 − x2 ) and X2 (x1 , x2 ) = x2 (1 − x1 − x2 ).

[3 marks]

Page 5
 Evaluating this matrix at the origin we obtain
µ ¶
1 0
A(0,0) = ,
0 1

which is the identity matrix. The determinant of this matrix is obviously 1 and therefore non-
vanishing. So the origin is a simple fixed point.
[2 marks]
For any point (α, 1 − α) lying on the line x1 = 1 − x2 we find that the Jacobian matrix is
µ ¶
−α −α
A(α,1−α) = ,
−1 + α −1 + α

which has determinant det(A) = 0, therefore the line of fixed points x1 = 1 − x2 contains only
non-simple fixed points.
[3 marks]
(c) By writing down the linearization about the fixed point at the origin classify its nature. Solve the
linearized equation.
We have already almost answered this question above when we computed the Jacobian matrix at the
origin. The linearized equation about the origin is:

x = A(0,0) x,
µ ¶
x1
which x = and since A(0,0) is the identity matrix, the fixed point at the origin is an
x2
unstable star node.
[2 marks]
The equation is very easy to solve. In components we have that

ẋ1 = x1 , ẋ2 = x2 ,

which is solved by
x1 = C1 et , x 2 = C2 e t ,
with C1 , C2 arbitrary constants. Since x1 /x2 = C1 /C2 , phase space trajectories near the fixed point
will be straight lines going by the origin.
[2 marks]
(d) By dividing the two equations in (3) obtain a new equation of the form dx
dx2 = X(x1 , x2 ). Find general
1

solution to this equation. Explain why your solution is what one would expect given the nature of the
fixed point at the origin.
The resulting equation is
dx1 x1 (1 − x1 − x2 ) x1
= = ,
dx2 x2 (1 − x1 − x2 ) x2
which only is well defined when x1 6= 1 − x2 and x2 6= 0. The equation can be solved by separation
of variables, Z Z
dx1 dx2
= ,
x1 x2
which gives
ln x1 = ln x2 + C,
or
x1 = eC x2 ,
where C is an integration constant. Therefore, in general phase space trajectories are straight lines
going by the origin.

Page 6
 [3 marks]
This is what one expects from the fact that the origin is an unstable star node which is precisely
characterized by trajectories which are straight lines with various slopes, depending on the initial
condition.
[2 marks]
(e) Use the results of (c) and (d) to sketch the phase space diagram associated to (3). Pay special
attention to the behaviour of phase space trajectories near the line of fixed points found in (a). Does
the direction of the trajectories change at this line? Why?
The phase space diagram is given in the figure below.

Figure 2: The phase space diagram of equation (3). The blue line is the line of fixed points x2 = 1 − x1 .

[3 marks]
The phase space trajectories change direction exactly at the line x1 = 1 − x2 this is because the sign
of the derivatives ẋ1 and ẋ2 changes at this line. Below the line, that is for x1 < 1 − x2 we have that
(1 − x1 − x2 ) > 0, therefore

ẋ1 = x1 (1 − x1 − x2 ) > 0, ẋ2 = x2 (1 − x1 − x2 ) > 0,

if both x1 , x2 > 0. This is compatible with trajectories moving away from the origin, in the direction
of increasing x1 and x2 . However for x1 > 1 − x2 , that is above the line of fixed points we have
exactly the opposite behaviour, that is

ẋ1 = x1 (1 − x1 − x2 ) < 0, ẋ2 = x2 (1 − x1 − x2 ) < 0,

if both x1 , x2 > 0. Therefore, here the trajectories have to move towards the origin at time grows,
that is in the direction of decreasing values of x1 and x1 .
[3 marks]
4. [Unseen] Consider the following second order linear differential equation:

ẍ − 4ẋ − 5x = 0, (4)
d2 x dx
where ẍ = dt2 and ẋ = dt .

(a) Consider now the following two first order linear differential equations,

ẋ1 = ax1 + bx2 + c, ẋ2 = dx1 + ex2 + f, (5)

Page 7
 where x1 = x above. Find the values of the constants a, b, c, d, e and f so that the system (5) is
equivalent to (4).
By defining x1 = x and x2 = ẋ = ẋ1 , we can rewrite (4) as:

ẋ1 = x2 , ẋ2 = 4x2 + 5x1 ,

so a = c = f = 0 and b = 1, d = 5, e = 4.
[2 marks]
(b) Using the result of (a) write equation (4) in the standard matrix form ẋ = Ax + b.
µ ¶ µ ¶µ ¶
ẋ1 0 1 x1
= ,
ẋ2 5 4 x2
with b = 0 in this case.
[1 mark]
(c) Find the eigenvalues and eigenvectors of the matrix A. Hence construct the Jordan normal form of
A and the matrix P which relates A to its Jordan form J as A = P JP −1 . The eigenvalues of A are
obtained by finding the zeroes of the characteristic polynomial of A:

|A − λI| = −λ(4 − λ) − 5 = 0 ⇒ λ1 = −1, λ2 = 5.

[2 marks]
The eigenvectors corresponding to these eigenvalues are obtained by solving:
µ ¶µ ¶ µ ¶
0 1 a a
= (−1) .
5 4 b b
In components this gives two equations:

b = −a, 5a + 4b = −b,

which are not independent. Taking a = 1 we find that the eigenvector associated to λ1 = −1 is
µ ¶
1
E1 = ,
−1

[2 marks]
and for λ2 = 5 we obtain the equation
µ ¶µ ¶ µ ¶
0 1 a a
=5 .
5 4 b b
In components this gives two equations:

b = 5a, 5a + 4b = 5b,

which are not independent. Taking a = 1 we find that the second eigenvector is
µ ¶
1
E2 = .
5

[2 marks]
The Jordan normal form of A is just the diagonal matrix
µ ¶
−1 0
J= ,
0 5
and the matrix P is built just by setting its columns to the eigenvectors above:
µ ¶
1 1
P = .
−1 5

Page 8
 [2 marks]
(d) Find the fixed point of the system of equations and classify its nature.
In this case the fixed point is just at the origin.
[1 mark]
Because the eigenvalues are real and one eigenvalue is negative and the other positive this would be
a saddle.
[3 marks]
(e) Find the general solution of the system of equations ẏ = Jy. Hence find the general solution for x.
In this case the canonical system of equations is simply:

ẏ1 = −y1 , ẏ2 = 5y2 ,

which is solved by y1 = C1 e−t and y2 = C2 e5t , where C1 and C2 are arbitrary constants.
[3 marks]
Now we need to take the solutions from part (e) and multiply them by the matrix P to obtain x:
µ ¶µ ¶ µ ¶ µ ¶ µ ¶
1 1 y1 y1 + y2 C1 e−t + C2 e5t x1
x = Py = = = = .
−1 5 y2 −y1 + 5y2 −C1 e−t + 5C2 e5t x2

This can be rewritten as

x = C1 e−t E 1 + C2 e5t E 2 .
[3 marks]
(f ) Sketch the phase diagram both in the y1 − y2 plane and in the x1 − x2 plane.

Figure 3: The y1 − y2 diagram is a typical saddle. As expected, the x1 − x2 diagram would be a rotated and
slightly stretched version of the first diagram, where the symmetry axes are now the directions determined by
the eigenvectors E 1 , E 2 .

[4 marks]
5. [Unseen] Consider the following first order nonlinear differential equations
x1 x2
ẋ1 = (1 − − x2 )x1 , ẋ2 = −(1 + − x1 )x2 , (6)
2 2
which can be interpreted as modelling the populations of two competing species.

Page 9
(a) Find the four fixed points of the system.
To find the fixed points we have to solve the simultaneous equations:
x1 x2
(1 − − x2 )x1 = 0, (1 + − x1 )x2 = 0.
2 2
An obvious solution is (x1 , x2 ) = (0, 0). For x1 = 0 we can also have x2 = −2 and for x2 = 0 we can
also have x1 = 2. Finally, if both x1 and x2 are not zero there is also the solution (x1 , x2 ) = ( 56 , 25 )
is also a fixed point. Therefore, we have the four points:
6 2
(0, 0), (2, 0), (0, −2), ( , ).
5 5
[2 marks]
(b) Find the general form of the Jacobian matrix of the system and evaluate it at each fixed point. Classify
the fixed points according to their linearization.
The Jacobian matrix is obtained by computing the derivatives of the r.h.s. of the equations (6) w.r.t.
x1 and x2 . Defining X1 (x1 , x2 ) = (1 − x1 /2 − x2 )x1 and X2 (x1 , x2 ) = −(1 + x2 /2 − x1 )x2 we obtain,
à ! µ ¶
∂X1 ∂X1
∂x ∂x 1 − x1 − x2 −x1
A= ∂X2
1
∂X2
2 =
∂x ∂x
x2 −1 − x2 + x1
1 2

[2 marks]
At the various fixed points it becomes:
µ ¶
1 0
A(0,0) =
0 −1
which is diagonal with real eigenvalues of different signs. This means that the origin is a saddle.
[1 mark]
µ ¶
−1 −2
A(2,0) = .
0 1
In this case we need to find the eigenvalues of A(2,0) in order to classify the fixed point. They are
the solution to the equation
(−1 − λ)(1 − λ) = 0,
that is λ = 1 and λ = −1. Therefore, as before this would also be a saddle.
[2 marks]
µ ¶
3 0
A(0,−2) = .
−2 1
The eigenvalues of A(0,−2) are the solutions to the equation
(3 − λ)(1 − λ) = 0,
that is λ = 1 and λ = 3. Therefore both eigenvalues are real and positive which corresponds to an
unstable node.
[2 marks]
Finally, µ ¶
− 53 − 65
A(6/5,2/5) = 2 .
5 − 51
The eigenvalues of A(6/5,2/5) are the solutions to the equation
3 1 12 4λ 3
(− − λ)(− − λ) + = λ2 + + = 0,
5 5 25 5 5
√
that is λ = − 25 ± i 511 . These are complex conjugated eigenvalues with negative real part which
would describe a stable focus.

Page 10
 [2 marks]
(c) For only those fixed points lying in the region −1.5 ≤ x1 ≤ 1.5 and −1.5 ≤ x2 ≤ 1.5 solve the
linearized equations.
The only fixed points lying on the indicated region are (0, 0) and (6/5, 2/5) = (1.2, 0.4). The first
fixed point is a saddle and the linearized equation is already in the canonical form. Therefore, the
general solution to this linearized equation is:
µ ¶ µ ¶
t 1 −t 0
x = C1 e + C1 e ,
0 1

where C1 , C2 are arbitrary constants.

[1 mark]
The fixed point (6/5, 2/5) = (1.2, 0.4) is a stable focus. In order to solve the linearized equation
about the fixed point we need to carry out the standard transformations. First we change variables
to z = x − a, where a is the fixed point so that we bring the fixed point to the origin. Second we
rewrite the resulting linearized equation ż = Az in the canonical form and solve this equation:

ẏ = Jy,

where J is the Jordan form of A and z = P y, where P is a nonsingular matrix that relates A and J
as
A = P JP −1 .
[1 mark]
√
11
In order to construct the matrix P we need the eigenvectors of A. The eigenvalues λ = − 52 ± i 5
mean that the Jordan form of the matrix A is
µ √ ¶
11
− 52
J= 5 .
− 11
5 − 25

The eigenvectors of A are obtained by solving the equations

µ 3 ¶µ ¶ à √ !µ ¶
− 5 − 65 a 2 11 a
2 = − ±i ,
5 − 15 b 5 5 b

which can be written as,

à √ ! à √ !
3a 6b 2 11 2a b 2 11
− − = − ±i a, − = − ±i b.
5 5 5 5 5 5 5 5

As usual,
³ the√ two ´ equations for each eigenvalue are not independent. Taking a = 1 we obtain
b = − 16 ± i 611 , therefore the eigenvectors and eigenvalues are
µ ¶ √
1 √ 2 11
E1 = , λ1 = − + i ,
− 16 − i 611 5 5

and µ ¶ √
1 √ 2 11
E2 = , λ2 = − − i .
− 16 + i 611 5 5
[3 marks]
The matrix P that relates A to its Jacobian form is obtained by taking the real and imaginary parts
of the first eigenvector, and setting them as columns of J,
µ ¶
1 √
0
P = ,
− 16 − 611

and we can check that indeed A = P JP −1 .

Page 11
 [1 mark]
The equation ẏ = Jy can be solved in the usual way, for J of the form above (see section (g) of
question 2). The solution takes the form
√ µ ¶ √ µ ¶
− 2t 11 1 − 2t 11 0
y = C1 e 5 cos( t + C2 ) + C1 e 5 sin( t + C2 ) ,
5 0 5 1

where C1 , C2 are constants.

[2 marks]
In order to get the solutions for the variable z we need to multiply the vector above by the matrix
P and finally, to get x we have to add the fixed point
µ 6 ¶ √ µ ¶ √ µ ¶
5 − 2t 11 1 − 2t 11 0
x = a + Py = 2 + C1 e 5 cos( t + C2 )P + C1 e 5 sin( t + C2 )P
5 5 0 5 1
µ 6 ¶ √ µ ¶ √ µ ¶
5 − 2t 11 1 − 2t 11 0
√
= + C1 e 5 cos( t + C2 ) + C1 e 5 sin( t + C2 ) .
2
5 5 − 16 5 − 611
In components: √
6 − 2t 11
x1 = + C1 e 5 cos( t + C2 )
5 5
√ √ √
2 C1 − 2t 11 11C1 − 2t 11
x2 = − e 5 cos( t + C2 ) − e 5 sin( t + C2 ).
5 6 5 6 5
[3 marks]
(d) In the same region of the x1 − x2 plane considered in section (c) and using the solutions to the
linearized equations, sketch the phase diagram associated to the nonlinear system of equations. You
may employ the vector field diagram below as a guiding tool in order to produce your sketch.

Figure 4: The sketch here is not entirely accurate, as I have ignored the presence of the other fixed points at
(0,-2) and (2,0). Specially (0,-2) would affect the shape of the trajectories on the lower right hand corner of the
picture, as one can see from the vector field diagram. The students were not expected to notice this.

[3 marks]

Page 12