MA1506

MA1506 TUTORIAL 1
1. Solve the following dierential equations:

(a) x(x + 1)y
= 1 (b) (sec(x))y
= cos(5x)
(c) y
= e
(x3y)
(d) (1 + y)y
+ (1 2x)y
2
= 0
Use www.graphmatica.com to sketch the functions you found as solutions of [a]-[d], if y =
1/2 at x = 1. Graphmatica can also directly sketch the graph if you insert the equation itself; for
example, in [a] you just enter x(x + 1)dy = 1 {1, 1/2}. [curly brackets for the initial conditions,
with x followed by y.] Here dy represents y
. Use this to check that your answers were correct.

2. Experiments show that the rate of change of the temperature of a small iron ball is proportional
to the dierence between its temperature T(t) and that of its environment, T
env
(which is constant).
Write down a dierential equation describing this situation. Show that T = T
env
is a solution.
Does this make sense? The ball is heated to 300
F and then left to cool in a room at 75
F. Its
temperature falls to 200
F in half an hour. Show that its temperature will be 81.6
F after 3 hours
of cooling.
3. In very dry regions, the phenomenon called Virga is very important because it can endanger
aeroplanes. [See http://en.wikipedia.org/wiki/Virga ]. Virga is rain in air that is so dry that the
raindrops evaporate before they can reach the ground. Suppose that the volume of a raindrop is
proportional to the 3/2 power of its surface area. [Why is this reasonable? Note: raindrops are not
spherical, but lets assume that they always have the same shape, no matter what their size may
be.] Suppose that the rate of reduction of the volume of a raindrop is proportional to its surface
area. [Why is this reasonable?] Find a formula for the amount of time it takes for a virga raindrop
to evaporate completely, expressed in terms of the constants you introduced and the initial surface
area of a raindrop. Check that the units of your formula are correct. Suppose somebody suggests
that the rate of reduction of the volume of a raindrop is proportional to the square of the surface
area. Argue that this cannot be correct.
4. Moths navigate at night by keeping a xed angle between their velocity vector and the direction
of the Moon [or some bright star; see http://en.wikipedia.org/wiki/Moth]. A certain moth ies
near to a candle and mistakes it for the Moon. What will happen to the moth?
Hints: in polar coordinates (r, ), the formula for the angle between the radius vector
and the velocity vector is given by tan() = r
d
dr
. [If you want to derive this formula, remember
that the tangential component of a small displacement in polar coordinates (r,)(r + dr, +
d) is rd, and the radial component is just dr. Now use simple trigonometry.] Use the formula to
solve for r as a function of . Note that graphmatica can graph in polar coordinates if you spell
out [as theta!]
5. In Cosmology, the ratio of the sizes of the Universe at two dierent times is measured by a
function of time called the scale function, denoted a(t). What are the units of a(t)? The [rst]
1
Friedmann Equation [http://en.wikipedia.org/wiki/Friedmann equations] relates this function to
the energy density of the Universe and to its spatial curvature. In a particular cosmological model,
the Friedmann equation takes the form L
2
a
2
= a
2
2
a
2
+ 1, where L is a positive constant,
the dot denotes time dierentiation, and the initial condition is a(0) = 1. What are the units of
L? Show, without solving this equation, that the universe described by this model is never smaller
than a certain minimum size. Now solve the equation and describe the history of this universe.
6. Solve the following equations:
(a) y
=
12y4x
1+y+2x
(b) y
x+y+1
x+y+3
2
(c) x + y + 1 + (x + y 3)y
= 0
2
MA1506 Tutorial 1 Solutions

(1a)
c
x
x
y
x x x x
y +
+
=
+
=
+
=
1
ln
1
1 1
) 1 (
1
'

(1b)
| | c x x y x x x x y +
(
+ = + = = 4 sin
4
1
6 sin
6
1
2
1
4 cos 6 cos
2
1
5 cos cos '

(1c)
c e e dx e dy e e e
dx
dy
x y x y y x
+ = = =
3 3 3
3
1

(1d)
c x x
y
y dx x dy
y
y
+ = =
+
2
2
1
ln ) 1 2 (
1

(2)
) ( Tenv T k
dt
dT
= where k is a positive constant. (If k were a negative, then hot
objects would get hotter when left to cool. That doesnt happen if it did, we would
not be here to discuss it.) T = Tenv is obviously a solution since 0 =
dt
dTenv
and this
does make sense because objects do not spontaneously become hotter or colder.
Having settled this case, we can assume T Tenv = and so we can write
kdt
Tenv T
dT
=
so c kt Tenv T + = ln .

In the case at hand, T >Tenv so Tenv T Tenv T = and so T .
kt
e Tenv
+ = o
At t = 0, T = 300 so 300 225 75 = + = o o . At
2
1
= t , T = 200, so
1756 . 1
225
125
ln 2 = = k 225 75 200
2
+ =
e
k

Thus . 6 . 81 225 75 ) 3 (
3
~ + =
k
e T

(3)
The volume V is related to the area A by V = a A
(3/2)
where a is a positive constant
with no units; this is reasonable because volume has units of cubic metres and area
has units of square metres. [Of course, ``reasonable doesnt mean that its always
exactly true.] Then dV/dt = (3a/2)(dA/dt) A
(1/2)

The question tells us that dV/dt = - bA, where b is a positive constant with units of
metres/sec. This is reasonable because evaporation takes place at the surface of the
drop and so its rate can be expected to depend on the area. So we have dA/ A
(1/2)
=

- 2bdt/3a. Integrating from A
0
, the initial area, up to zero, we find that the time
taken for complete evaporation is 3aA
0
(1/2)
/b which does indeed have units of time
since the numerator has units of metres while the denominator has units of
metres/sec.

If, instead of dV/dt = - bA, we propose that dV/dt = - bA
2
, then we would have
obtained dA/ A
(3/2)
= - 2bdt/3a . When you try to integrate this from A
0
to zero, you
will get a divergent integral, meaning that the evaporation would take infinite time
and the rain would always reach the ground, contrary to the definition of Virga.

(4)
The moth flies in such a way that the angle remains constant at all times, so we
have a differential equation tan () = constant = rd/dr, hence dr/r = d / tan () ,
thus r = R exp( / tan ()), where we take it that = 0 when the moth first sees the
candle, and that her distance from the candle is R at that time. Remember that is
the angle between the radius vector of the moth [pointing outwards] and her velocity.
From the point of view of the moth looking towards a candle in front of her, would
be an angle greater than 90 degrees. Draw a diagram if this is not obvious! Thus tan
() will be negative and r will get steadily smaller as increases. Such a curve is
called a spiral. So the unfortunate moth will spiral into the candle with tragic
consequences. Of course if her first view of the candle is over her ``shoulder then tan
() will be positive and she will spiral outwards, something that would be a lot less
noticeable. Finally if is 90 degrees exactly, the moth will fly along a circle until it
drops dead from exhaustion or starvation, whichever comes first.

(5)
Since it is a ratio, a(t) has no units. The units of L are those of time, so that both sides
of the given equation have the same units [ie, none]. If a(t) were ever smaller than 1,
then 2/a >2, yet 1 + a
2
< 2, so the right side of the equation would become negative,
while the left side is never negative. Hence a(t) can never be less than 1. Next,
separate the variables in the obvious way and change the variable a to y = a
2
and you
will get
2
/ 2
2
+
=
y y
dy
L dt . Complete the square and make the substitution
) cosh(
2
3
2
1
x y = + [We know we can do this, because y is never less than 1 so the left
side is never less than 3/2. Remember that cosh is never less than 1.] Do the integral
and you get ) / 2 cosh(
2
3
2
1
2
L t + a
= . The right side is equal to 1 at t = 0, as it should

be. The graph of a starts out at a = 1, and then steadily increases at a faster and faster
rate. This describes a Universe that begins with non-zero size and then expands. The
expansion is not slowing down --- instead it is getting faster and faster.

(6)
These are examples of ODEs where a change of variable is needed.

(6a)
Let v so y x + = 2 ' 2 ' y v + =
c x y x y x
c x v v
v
v
v
v
v
+ = + + +
+ = +
+
=
+
=
3 ) 2 (
2
1
) 2 (
3
2
1
1
3
'
1
2 1
2 '
2
2

(6b)
y x v + =
2
3
1
1 ' ' |
.
|
\
|
+
+
= =
v
v
v y

( )
( ) c x y x y x y x
c x v v v
dx dv
v v
v
dx dv
v v
v v v
dx dv
v v
v
v
v v
v
v v
v
+ = + + + + + +
+ = + + +
=
|
.
|
\
|
+ +
+
+
=
+ +
+ + + +
=
+ +
+
+
+ +
=
+
+ +
+ =
2 5 4 4 ln
2 5 4 ln
2
5 4
4 2
1
2
5 4
4 2 5 4
2
5 4
) 3 (
) 3 (
10 8 2
) 3 (
1 2
1 '
2
2
2
2
2
2
2
2
2
2
2

(6c)
Here we need 2 changes of variables: first set o + = X x , | + = Y y so
) 1 ( 1 + + + + = + + | o Y X y x , ) 3 ( 3 + + + = + | o Y X y x so if we choose
0 3 1 = + = + + | o | o i.e. 2 = o , 1 = | then (i.e. 1 = y X Y , ) 2 + = x
0 ) ( = + + +
dX
dY
Y X Y X and this is homogeneous so set
c
dV
+
|
|
.
|
2
x
y
c
V
V
X
dX
\
|
|
.
|
\
|
+
+
+
+
=
2
2
1
1 ln
1
1
x
y
x
V X
XV V V V
XV V Y
X
Y
V
|
.
|
\
|
+
= +
=
+ + + + =
+ = =
2
1
arctan 2 ln
1 ln
2
1
arctan ln
)( 1 ( 1 0
' '
V
+
2
1
) '
2

since
2
1
+
= =
x
y
X
Y
V

MA1506 TUTORIAL 2
(a) xy
+ (1 +x)y = e
x
, x > 0
(b) y
(1 +
3
x
)y = x + 2, y(1) = e 1, x > 0
(c) y
+y +
x
y
= 0 (d) 2xyy
+ (x 1)y
2
= x
2
e
x
, x > 0
2. If a cable is held up at two ends at the same height, then it will sag in the middle, making a
U-shaped curve called a catenary. This is the shape seen in electricity cables suspended between
poles, in countries less advanced than Singapore, such as Japan and the US. It can be shown using
simple physics that if the shape is given by a function y(x), then this function satises
dy
dx
=

T
x
0
dy
dt
2
+ 1 dt,
where x = 0 at the lowest point of the catenary and y(0) = 0, where is the weight per unit length
of the cable, and where T is the horizontal component of its tension; this horizontal component is
a constant along the cable. Find a formula for the shape of the cable. [Hint: use the Fundamental
Theorem of Calculus, and think of the resulting equation as a rst-order ODE.]
3. Psychologists talk about something called a Performance Curve. Suppose an MA1506
student is solving mathematics problems. She starts with dierential equations. Let P(t) be a
non-negative function that measures her performance, that is, her success rate at solving DEs. Her
performance increases rapidly at rst, but then the rate of increase slows down as she becomes
more expert. Let M, a positive constant, be the best possible performance; then one can suppose
that P satises
dP
dt
= C[M P],
where C is a constant. What are the units of this constant? What does this constant measure?
Solve this equation assuming that she is completely incompetent at t = 0 [that is, P(0) = 0].
Now the student turns to another kind of problem, say in linear algebra. Again her per-
formance is low at rst but gets better in accordance with this equation. Now as the years go by,
her overall ability to solve mathematics problems gradually gets better, so C, instead of being a
constant, is really a slowly increasing function of time. Suppose that C(t) = K tanh(t/T), t 0,
where K and T are positive constants. [Is this reasonable? Why? What are the meanings of the
constants K and T?] Replacing C with C(t), solve for P, again assuming that P(0) = 0. Sketch
P(t), the performance curve. [Choose values of the constants for yourself, and use Graphmatica
[www.graphmatica.com] if necessary.]
4. A certain MA1506 student starts a rumour to the eect that one of the lecturers has been seen
dating Tang Wei. The number of students who have heard the rumour, R(t), is given by
dR
dt
= KR[1500 R],
1
where K is a positive constant, and 1500 is the number of students taking MA1506. What is the
meaning of K? Is this equation reasonable? [Hint: surely the rumour will spread slowly both when
hardly anyone has heard it yet, but also when nearly everyone has already heard it!] By regarding
this equation as a Bernoulli equation, nd R(t). Note that, in reality, R(t) is actually an integer
[whole number]. Comment on this fact in relation to your solution.
2

Question 1.

(1a)
x
e
x
y
x
y
= + +
1
)
1
1 ( '

Integrating factor is
x
xe x x
x
= + =
)
+ ) ln exp( )
1
1 ( exp (in general, + y x P y ) ( ' multiply
by ) ( ' exp
) ) )
= +
)
P P P
ye
dx
d
Pye e y P )

So )
1
( 1 ) (
x x x
e
x
xe yxe
dx
d
= =
x x x
e cx e y c x yxe

+ = + =
1

(1b)
x
x
x x x
x x
x
x x
x
e cx x y
c
x
e
c
x
e
x
e
x
e
c
x
e
x
e
e
x
y
e
x
x e
x
y
dx
d
e
x
x x
x
3
2
2 2 2
3 2 3
3 3
3
2
1
) 2 ( )
1
(
1
) ln 3 exp( )
3
1 ( exp
+ =
+
=
+
+ =
+ + =
+ =
= = +
) )
) )
)

since 1 1 1 ) 1 ( = + = = c ce e y

x
e x x y
3
+ =

(1c)
This kind is called a Bernoulli equation -- set
2
y z = ' 2 ' yy z =
y
z
y
2
'
' =

x x x
x x
ce x y c e x ze
xe z e
dx
d
x z z x z z
y
x
y
y
z
2 2 2 2
2 2
2
1
)
2
1
(
2 ) ( 2 2 ' 0 '
2
1
0
2
'
+
+ = + + =
= = + = + + = + +

(1d)
Since we define , )' ( ' 2
2
y yy =
2
y Y =
x
xe Y
x
Y = + )
1
1 ( ' ,
x
e
x x
1
)
1
1 ( exp =
)

x x
x x x x
cxe xe y
c e Y e
x
e Y e
x dx
d
+ =
+ = =
|
.
|
\
|
2
1
2
1 1 1
2
2 2

Question 2.
Define v = dy/dx. Then our equation is

( ) dt v v
x
1
T

0
2
)
+ =
.
Notice that v(0) = 0. Now by the fundamental theorem of calculus

1 '
2
+ = v
T
v

This is a separable differential equation with initial condition v(0) = 0. Separating the
variables and integrating [remember that ] we find that v = sinh(x/T).
Integrating, we get
1 sinh cosh
2 2
=

C x
T
T
y + = ) cosh(
,
where C must be T/ since y(0) = 0 and cosh(0) = 1. Thus we have

T
x
T
T
y = ) cosh( . The shape of the graph of this function is indeed U-shaped.
[Draw it!]

Question 3.

The constant C has units of 1/time. Solving the equation [either as linear first order or as
a separable equation] we get
Ct
Me M P
=

Clearly P will approach M more rapidly if C is large; that is, C measures how rapidly the
student is able to learn. Thus the equation indeed expresses the idea that the students
performance improves more slowly as she approaches her maximum possible
performance.

As the years go by and the student becomes more familiar with the methods of mastering
mathematics, her rate of learning new things might be expected to improve; but surely
there is an upper bound to how much she can improve. The tanh function is a simple way
of representing this since it always increases but is bounded above. [Remind the students
of the shape of tanh(x) if necessary.] Then K represents her maximum possible speed of
learning [since tanh is asymptotic to 1], and T measures the amount of time required for
her to realise her maximum potential. [Note that K has units of 1/time, while of course T
has units of time, so KT is dimensionless.] The equation can now be written as

0 ) )( / tanh( = + M P T t K
dt
dP
.

Its convenient now to define Q = P M, so the equation is

. 0 ) / tanh( = + Q T t K
dt
dQ

This is a first-order equation with integrating factor , so we have ) / ( cosh T t
KT

) / ( cosh T t Q
KT
= A,

where A is a constant. Since we are assuming that P(0) = 0, we have Q(0) = - M, so we
have A = - M, thus finally

| | ) / ( sech 1 T t M P
KT
= .

Students are encouraged to graph examples of such questions using [for example] the
software available free at http://www.graphmatica.com/

Question 4.

The constant K measures the rapidity with which the rumour will spread. It depends on
how interesting the rumour is, how much the students like to gossip, how gullible they
are, etc. The right hand side of the equation is designed to be small both near R = 1 and
near R = 1500, when indeed the rumour can be expected to spread slowly either because
not enough or too many people have heard it.

We have

2
1500 KR KR
dt
dR
= .

This is a Bernoulli equation as discussed in the notes. We solve it by defining Z = 1/R,
which transforms the equation into a linear one:

K KZ
dt
dZ
= +1500 ,

with solution

) 1500 exp(
1500
1 1
Kt C
R
+ = .

The problem says that this highly interesting rumour was started by one student, so R(0)
= 1. Thus C = 1499/1500. Hence

) 1500 exp(
1500
1499
1500
1 1
Kt
R
+ = .

Of course as t tends to infinity, R tends to 1500. The function R(t) is not really
continuous or differentiable, since it only takes integer values. However, this integer-
valued function can be well approximated by a smooth function defined to interpolate
between integral values.

OPTIONAL NOTE ON THE EQUATION IN QUESTION 2:
Consider a small element of the cable, with the left end located at a point (x,y), and
the right at (x+dx,y+dy). Let tan( ) u = dy/dx, and let U(x) denote the tension in the
cable. The horizontal forces acting are just Ucos( ) u and Ucos( ) u +d(Ucos( ) u ) at the
two ends of the small element, so if these forces are to balance we must have
Ucos( ) u = constant = T. The vertical forces are Usin( ) u and Usin( ) u +d(Usin( ) u )
and also ds, where ds is the length of the element. For these to balance we must
have d(Usin( ) u ) = ds, or d(Ttan( ) u ) = ds. Integrating this equation from x =
0 [where dy/dx = 0] we obtain the stated equation.
MA1506 TUTORIAL 3
(a) y
+ 6y
+ 9y = 0, y(0) = 1, y
(0) = 1
(b) y
2y
+ (1 + 4
2
)y = 0, y(0) = 2, y
(0) = 2(3 1)
2. Find particular solutions of the following:
(a) y
+ 2y
+ 10y = 25x
2
+ 3 (b) y
6y
+ 8y = x
2
e
3x
(c) y
y = 2xsin(x) (d) y
+ 4y = sin
2
(x)
3. Use the method of variation of parameters to nd particular solutions of
(a) y
+ 4y = sin
2
(x) (b) y
+ y = sec(x)
4. One of the most important kinds of nonlinear second-order odes takes the form y
= F(y).
[That is, y
can be expressed in terms of y only, with no y
.] These can be solved with the aid of

the following clever trick. Use the chain rule to prove that
d
2
y
dx
2
=
d
dy
[
1
2
y
2
].
Use this trick to show that any equation of the form y
= F(y) can be reduced to a separable

rst-order ode and therefore solved, because we know that we can always solve this kind of
equation.
Suppose that the Earth were to stop moving tomorrow. It would of course immediately
begin to fall towards the Sun. How long would we have before reaching the orbit of Venus [by
which time we would all have been fried]? Information: the radius of the earths orbit is about 150
billion m, that of Venus orbit is about 100 billion m [both orbits being approximately circular],
the acceleration due to gravity at a distance of r from a central object of mass M is GM/r
2
,
where M [the mass of the Sun] is about 2 10
30
kg, and G is Newtons constant = 6.67 10
11
in MKS units. So you have to solve the equation r = GM/r
2
. Use the above clever trick and
solve the resulting rst-order separable equation. You will get a nasty denite integral [which can
actually be done]; go to http://wims.unice.fr/wims/ if you would rather let a computer do it. [The
computer may not want to do it either [why?]; you may have to adjust your domain of integration
very slightly in order to persuade it to cooperate.]
5. Sometimes people wonder why Newtons second law has a second derivative why not a
third derivative? Lets investigate.
The method we used to solve second-order linear ODEs with constant coecients also works
for higher-order equations. Find the general solution of the generalized simple harmonic motion
equation
d
3
y
dt
3
= w
3
y,
1
where w is a positive constant [with what units?]. [Hint: there are three dierent numbers x
such that x
3
= -1; two of them are complex. The general solution should therefore contain three
terms and have three arbitrary constants.] This is what simple harmonic motion would look like
if Newtons Second Law involved third derivatives instead of second derivatives. Use matlab or
graphmatica to graph the solution of
d
3
y
dt
3
= 8y,
where y(0) = 0.0001, y(0) = 0, y(0) = 0. [Hint: d3y = -8y {0,0.0001,0,0}. You may have to zoom
in and out a few times to get graphmatica to work.]
2
(1a)
( )
x
x x x
e x y B A B y
A y
e Bx A Be y e Bx A y
y y y
3
3 3 3
2
t
) 2 1 ( 2 1 3 1 0 '
1 1 ) 0 (
) ( 3 ' ) (
3 0 9 6
e y Set 0 9 ' 6 "

+ = = = =
= =
+ = + =
= = + +
= = + +

(1b)
] 2 sin 3 2 cos 2 [ 3
2 2 ) 1 3 ( 2 ) 1 3 ( 2 ) 0 ( '
2 2 ) 0 (
] 2 cos 2 2 sin 2 [ '
] 2 sin 2 cos [
2 1 0 ) 4 1 ( 2
2 2
x x e y B
B y
A y
x B x A e y y
x B x A e y
i
x
x
x
t t
t t t
t t t t
t t
t t
+ = =
+ = =
= =
+ + =
+ =
= = + +
(2a)
Try C Bx Ax y + + =
2
x x
A
C B A B A A
x
C Bx Ax B Ax A
y y y
=
= = =
= + + = + =
+ =
+ + + + + =
+ +
2
2
2
2 5
3 10 2 2 0 10 4 25 10
3 25
10 2 2 2
10 2
2
5
y
0 C -1, B , /
, ,
) ( ) (
' "
(2b)
Try
x
e C Bx Ax y
3 2
) ( + + =
x
x x x x
x x
e x y
C C C B C B B A
B B B A B A A
A A A A
e C Bx Ax e B Ax e B Ax Ae y
e C Bx Ax e B Ax y
3 2
3 2 3 3 3
3 2 3
) 2 (
2 0 8 18 6 9 3 3 2
0 0 8 18 12 9 6 6
1 1 8 18 9
) ( 9 ) 2 ( 3 ) 2 ( 3 2 "
) ( 3 ) 2 ( '
=
= = + + + +
= = + + +
= = +
+ + + + + + + =
+ + + + =
(2c)
x x x y x x x z
x x x i x x x i x z
i B B i
A
x B Ax B Ax Ai z z
B Ax
e x y y
ix
sin cos sin cos Im
] sin cos [ sin cos e ) (
0 2 2
1
e 2 e ) 2 ( "
e ) ( iAe Aie z"
B)e i(Ax Ae z'
B)e (Ax z Try
above. the satisfies Imz then 2xe z"-z equation complex the solve can we if
part) imaginary (Im Im 2 "
ix
ix ix
ix ix ix
ix ix
ix
ix
= =
+ + = =
= =
=
= =
+ + =
+ + =
+ =
=
= =
(2d)
x x z y
x x i x ixe z
i B iB
A A
iBe Bxe A Bxe z z iBe e
e e
e x y y
ix
ix ix ix ix ix
ix ix
ix
2 sin
8
1
8
1
Re
)) 2 sin 2 cos ( 1 (
8
1
8
1
8
1
8
1
4
2
1
8
1
2
1
4
4 4 4 4 4 " 4 -4Bx z"
Bx A z Try
2
1
2
1
4z z" Solve
) Re(
2
1
2
1
) 2 cos 1 (
2
1
4 "
2
2 2 2 2 2
2 2
2
= =
+ = + =
= =
= =
+ + + = + + =
+ = = +
= = +
(3a)
W
x x
W
x x
x a
RHS RHS
2 cos ] ) 2 cos 1 (
2
1
[
B' ,
2 sin ] ) 2 cos 1 (
2
1
[ -
A'
satisfy chosen to are B(x) and A(x) if
) 2 cos 1 (
2
1
4y y" of solution is B(x)sin2x ) A(x)cos(2x Then
B(x). A(x), functions to B and A Promote
Bsin2x Acos2x y 0 4y y" solve first : parameters of Variation
) ( ) (

+
=
=
= + +
+ = = +
sin2x x constant) (arbitrary cos2x x constant) (arbitrary
solution general
the into absorbed be can cos - extra the and sin ) cos (
sin ) sin sin sin cos cos (cos
since (2d) in as same the is which
] sin ) sin (sin cos ) cos [(cos Bsin2x 2x Acos
is solution the so sin sin
cos cos
) (cos cos ) ( cos cos '
sin sin sin cos sin
4
1
A'
so
2 sin2x s2x)' in2x)'-(co (cos2x)x(s W where
+
=
+
+ = +
=
=
+ = =
+ = + =
= =
x x
x
x
x
x
x x x x x x
x x x x x x x
x
x x B
x x A
x x x x B
x x x x x
2
32
1
2
8
2
4
1
1
8
1
2
8
2 4
4
1
2 2 4
4
1
2
8
1
2 4
4
1
2 2 4
4
1
2
8
1
8
4
32
1
2
8
1
4
32
1
2
8
1
1 4
8
1
2
4
1
2
4
1
2
4
1
4
8
1
2
4
1
2 2
4
1
2
2 2
2
(3b)
xsinx | cosx | ln cosx y
x B |, cos | ln
cosx
sinx
- A
1 (sinx) (cosx)' - x)' (cosx)(sin w
cosx [sec(x)]
B'
sinx [sec(x)] -
A'
where ) sin( ) ( cos ) (
+ =
= = =
+ = =
+
= =
+
) x dx
w w
x x B x x A
(4)
] 2 / ) ' [( '
2
2
y
dy
d
y
dy
d
dx
dy
dx
y
= =
2
d
so the given equation can be written as
] 2 / ) ' [(
2
y
dy
d
= F(y). Just integrate both sides with respect to y, and you get a
separable first-order ODE.
Using this trick in the present case we have
] 2 / ) [(
2
.
r
dr
d
= - GM/
2
r [Sorry there should be a dot over the r on the left side, but
its not very clear.] From the problem statement it is clear that the Earths initial speed
is zero when r = R, where we use R to denote the original radius of the Earths orbit
[150 billion metres]. Using that to fix the arbitrary constant, we get upon integrating
both sides with respect to r:
R
GM
r
GM
r = 2 / ) (
2
.
.
Since the Earth will of course fall inward, the speed must always be negative after the
first instant, so when we take the square root we must be careful to take the negative
one. Hence
dt
R
GM
r
GM
dr
=
2 2
, where we have to integrate from r = R [Earth orbit] to r =2R/3
[Venus orbit radius, 100 billion metres]. Its convenient to change the variable from r
to x = r/R. You then get
1
1 2
1
3 / 2
2 / 3
=
)
x
dx
GM
R
t .
This is an improper integral, which is why the website recommended will sometimes
refuse to do it [though usually it does work!] If it is in a bad mood, just integrate from
2/3 up to 0.999 or something like that. With the given data you should get about 44.74
days. So thats how long it takes to reach the orbit of Venus. Remember that a day is
24 times 3600 seconds.
5.
Set y = exp ( t) as usual, and get
3 3
w =
So we need the cube roots of -1. Now
- 1 = ) exp( t i , ) 3 exp( t i , ) 5 exp( t i so the cube roots are
) 3 / exp( t i , ) 3 / 3 exp( t i , ) 3 / 5 exp( t i ,
which are
cos(t /3) + i sin(t /3) = (1/2) + i 2 / 3
cos(t ) + i sin(t ) = - 1
cos(5t /3) + i sin(5t /3) = (1/2) - i 2 / 3 .
You should verify that the cubes of these three numbers are all equal to 1.
Following the usual rule that real parts yield exponentials, and imaginary parts yield
cos and sin, we see that the general solution is
) 2 / 3 sin( ) 2 / 3 cos(
2 / 2 /
wt Ce wt Be Ae
wt wt wt
+ +
,
where A B C are the arbitrary constants. Notice that unless B and C are both exactly
zero, this is not a bounded function. In the case of the solution you are asked to graph
using graphmatica, you see that the solution behaves nicely for a while, but soon the
oscillations get totally out of control. So apart from very exceptional cases, the
solution will blow up --- even ordinary SHM would be unstable if Newtons 2
nd
law
had a third derivative instead of a second derivative! Then we would not be
here to talk about it.
MA1506 TUTORIAL 4
Question 1
Find the equilibrium points for the following equations. Determine whether these equi-
librium points are stable, and, if so, nd the approximate angular frequency of oscillation
around those equilibria.
(i) x = cosh(x) (ii) x = cos(x) (iii) x = tan(sin(x)).
Question 2
A certain electrical circuit has a resistor of resistance 5 ohms, an inductor of inductance
0.05 henries, and a capacitor of capacitance 4 10
4
farads. A varying voltage given by
V = 200 cos(100 t)
is applied. Find the current in this circuit if the initial current and the initial charge on
the capacitor are equal to zero.
[Answer: 2.35 e
50t
cos(50
19t) + 22.13 e
50t
sin(50
19t) + 2.35 cos(100t) 9.41 sin(100t)]

Question 3
Use Graphmatica to graph the solutions of
x + x = cos( t)
[with x(0) = 0, x(0) = 0] when = 1 and when = 0.9. In the second case, verify that
the two frequencies apparent in the graph are what you expect them to be. [Please get
the solutions from the lecture notes and then graph them dont use graphmatica to
solve the equations, except to check your answers!
Question 4
A fully loaded large oil tanker can be modelled as a solid object with perfectly vertical
sides and a perfectly horizontal bottom, so all horizontal cross-sections have the same
area, equal to A. Archimedes principle [http://en.wikipedia.org/wiki/Buoyancy] states
that the upward force exerted on a ship by the sea is equal to the weight of the water
pushed aside by the ship. Let be the mass density of seawater, and let M be the mass
of the ship, so that its weight is Mg, where g is 9.8 m/sec
2
. When the ship is at rest, nd
the distance d from sea level to the bottom of the ship. This is called the draught of the
ship.
Suppose now that the ship is not at rest; instead it is moving in the vertical direction.
Let d + x(t) be the distance from sea level to the bottom of the ship, where d is the
draught as above. Show that, if gravity and buoyancy are the only forces acting on the
ship, it will bob up and down with an angular frequency given by
Ag/M.
Next, suppose that in fact there is a small amount of friction between the sides of
the ship and the seawater, as the ship moves up and down in the sea. The frictional
force is equal to b x, where b is a constant and x is the downward speed of the ship.
Furthermore, waves from a storm strike the ship and exert a vertical force F
0
cos(t) on
the ship, where F
0
is the amplitude of the wave force and is the wave frequency. Find
the most dangerous [ = resonance] value of . Let H be the height of the deck of the ship
above sea level when the ship is at rest. For xed values of F
0
and b, show how to design
a ship that will be able to survive this storm, no matter what may be. That is, explain
how to choose safe values of the things you control, namely M and A.
2
MA1506 TUTORIAL 4 SOLUTIONS
Question 1
(i) x = cosh(x). An equilibrium solution of an ODE is just a solution that is identically
constant. That is not possible here because the cosh function never vanishes. So there is
no equilibrium for this ODE.
(ii) x = cos(x). Equilibria are at x = /2, 3/2, etc etc. Taylor expansion at /2 is
cos(x) = cos
(/2)[x /2] + ....

where we drop the higher order terms since we wish to consider a small perturbation. If
we dene y = x /2 then we have
y = y
so we have simple harmonic motion [ie, stable equilibrium] with angular frequency [ap-
proximately] 1.
Taylor expansion at 3/2 is
cos(x) = cos
(3/2)[x 3/2] + ....

we dene y = x 3/2 then we have
y +y
so this equilibrium is unstable. The other equilibria are like these two; they alternate as
we consider larger and smaller equilibrium values of x.
(iii) x = tan(sin(x)). Equilibria are at 0, , 2 etc etc etc. Taylor expansion at 0 is
tan(sin(x)) = cos(0)sec
2
(sin(0))[x 0] + ....
where we drop the higher order terms since we wish to consider a small perturbation. We
have
x +x
so we have an unstable equilibrium.
Taylor expansion at is
tan(sin(x)) = cos()sec
2
(sin())[x ] + ....
we dene y = x then we have
y y
so we have simple harmonic motion [ie, stable equilibrium] with angular frequency ap-
proximately 1. The other equilibria are like these two; they alternate as we consider larger
and smaller equilibrium values of x.
Question 2
The equation governing such a circuit is
Q + R

Q/L + Q/LC = V/L,
where Q is the charge on the capacitor, R is the resistance, L is the inductance, C is the
capacitance, and V is the applied voltage. So here we have
d
2
Q
dt
2
+ 100
dQ
dt
+ 50000Q = 4000 cos100t.
We can solve this in the usual way. The roots of the quadratic equation turn out to be
50 50 i
19, and using the method of undetermined coecients you will nd that a
particular solution is
16
170
cos100t +
4
170
sin100t
so the general solution of the equation is
Q = c
1
e
50t
cos50
19t + c
2
e
50t
sin50
19t +
16
170
cos100t +
4
170
sin100t.
We are told that Q(0) = 0 and
dQ
dt
{t = 0} = 0.
The rst condition gives
0 = q(0) = c
1
(1) + c
2
(0) +
16
170
but in order to use the second condition we need to dierentiate rst:
dQ
dt
= 0.0941 (50e
50t
cos50
19 t 50
19e
50t
sin50
19 t)
+c
2
(50e
50t
sin50
19 t + 50
19e
50t
cos50
19 t)
160
17
sin100t +
40
17
cos100t
So we get
0 =
dQ
dt
{t = 0} = 0.0941 (50) + c
2
(50
19) +
40
17
.
Solving these two simultaneous equations for c
1
and c
2
, we substitute them back into the
formula we found for dQ/dt [since that is the current] and we nd
2.35 e
50t
cos(50
19t) + 22.13 e
50t
sin(50
19t) + 2.35 cos(100t) 9.41 sin(100t)

Question 3
If = 1 then we have resonance [see the lecture notes, chapter 2]. So the relevant solution
is
x =
F
0
t
2m
sin(t).
In this case, F
0
, the coecient on the right side of the original ODE, is 1, and so are all
of the other coecients, so
x =
t
2
sin(t).
2
The graph [attached] shows the uncontrolled growth characteristic of resonance in the
absence of damping.
If is 0.9 then the relevant formula from the notes is
x =
2F
0
/m
2
sin
sin
+
2
.
Here we have
x =
2
(0.9)
2
1
sin
0.9 1
2
sin
0.9 + 1
2
.
The graph [attached] is clearly the result of combining two sine waves, one with angular
frequency 1/20 [and therefore with period 40] and the other with angular frequency 0.95
[and therefore period a little more than 2.] There are about 20 oscillations of one in
each complete oscillation of the other. [Note: a complete oscillation of the slow wave
encloses two of the lumps in the graph, so you have to be careful dont measure
the period by looking at the distance from one lump to the next. Confusingly, some
physicists do that, getting a beat frequency twice the one we use. Please ignore the
physicists.]
Question 4
When the ship is at rest, the part of it which is under sea level has a volume of Ad [that is,
the area of the base times the height]. Therefore, this is the volume of seawater that has
been pushed aside by the ship. If the density of seawater is , then the mass of seawater
pushed aside is Ad, and its weight is Adg. This upward force exactly balances the
weight of the ship, so we have
Ad g = Mg.
Thus
d = M/ A.
Now if the ship is moving and the distance from sea level to the bottom of the ship is d
+ x, where x is a function of time, we have to use Force = mass acceleration. Taking
the downwards direction to be positive, we nd that the buoyancy force is now A(d +
x)g, so we have
M x = Mg A(d + x)g,
which, using our formula for d, is just
x =
Ag
M
x
This represents simple harmonic motion with angular frequency
Ag/M, as claimed.
The ship will bob up and down at this frequency. Note the inverse dependence on M,
which is to be expected, but also that the frequency increases if A is large, which is not
so obvious.
Taking into account the friction and the force exerted by the waves, Force = mass
acceleration gives
M x = Mg A(d + x)g b x + F
0
cos(t)
3
or
M x + b x + Ag x = F
0
cos(t).
This is exactly the equation studied in the notes, except that k is replaced by Ag. We
assume that b is small, so that, in the absence of waves, the ship will undergo damped
harmonic motion. [The ship is sailing in seawater, not honey....]
So after the transient terms [the solution of the homogeneous equation, which decay
exponentially and so can be neglected] die out, we will have [see page 33 of the notes]
x(t) =
1
M
F
0
cos(t )
(
2
2
)
2
+
b
2
M
2
2
,
where is a constant and where denotes
Ag/M. So eventually the ship bobs up

and down at the same frequency as the waves, but the amplitude of x is given by
A() =
F
0
/M
(
2
2
)
2
+
b
2
M
2
2
Notice that this can be very large even if F
0
is quite small. So there is a danger that even
if the ship is safe for most values of , it might sink if takes a particular value. [The
ship will probably sink if the maximum possible value of x exceeds H, because that means
that the deck of the ship will be under water! ].
To see how large A() can be, look at the expression inside the square root in the
denominator and regard it as a function of , dened to be
2
. This expression is
f() =
2
+ [
b
2
M
2
2
2
] +
4
,
which is a quadratic with minimum at
2
b
2
/2M
2
; we may assume that b is so small
that this is positive. [If b is not so small, then actually the ship is in no danger left to
you as an exercise.] Thus the most dangerous value of is just
danger
=
2
(b
2
/2M
2
)
and simple algebra shows that [remembering the denition of ]
A(
danger
) = 2MF
0
/b
4MAg b
2
So to design a ship, we need a good estimate of the largest possible value of F
0
[see
http://en.wikipedia.org/wiki/Rogue wave], we need to measure b and , and then we
should choose A and M in such a way that
2MF
0
/b
4MAg b
2
< H,
because then, even in the worst possible case, the ship will never go so far down that the
sea comes over the deck.
4
MA1506 TUTORIAL 5
Question 1
Close examination of a can of Milo
TM
shows that, in fact, the bottom of the can is not
welded to the vertical part of the can; it is part of the same piece of metal. [If you dont
know what I mean, or even if you do, look at a real can of Milo
TM
.] Furthermore,
there is no seam down the side of the can: the whole thing, except for the top, is punched
out of one big piece of metal. The top of the can, however, is welded on. Assume that
the manufacturers of Milo
TM
optimise their costs, and measure a real can of Milo
TM
.
Compute the relative cost of 1 centimetre of welding compared to one square centimetre
of the aluminium alloy used to make the can. [Ignore wastage of aluminium.] Comment.
[Answer: about 7.5.]
Question 2
The bacteria in a certain culture number 10000 initially. Two and a half hours later there
are 11000 of them. Assuming a Malthus model, how many bacteria will there be 10 hours
after the start of the experiment? How long will it take for the number to reach 20000?
[Answers: About 14600; about 18.18 hours.]
Question 3
Read http://en.wikipedia.org/wiki/Overpopulation. Perhaps Malthus was not so wrong
after all? It has been suggested that the Earths population explosion problem can be
solved by sending excess population to colonise other planets. Assuming that a xed
number of colonists are sent out each year, and that the Malthus model would hold if
there were no emigration, set up an ODE to describe this plan. Solve it and analyse it.
[That is, consider various values of your parameters and make predictions, and try to say
something interesting about what you nd.] Next, modify the model by assuming that
the rate of emigration is proportional to time [that is, we send more and more people out
each year]. Analyse!
Question 4
On the island of Orpsengia, the human birth and death rates per capita are constant; and
the population of the island has been doubling every 20 years. However, one day several
pirate ships arrive. All of the island women under the age of 50, tired of being ordered
about by their mothers-in-law and ignored by their husbands, decide to elope with the
glamorous pirates, taking their children with them. After that, the remaining population
of Orpsengia declines by half over the next ten years. What was the original birth rate
per capita on Orpsengia? You will have to make several simplifying assumptions to solve
this problem; that is ok as long as you list your assumptions carefully! [Answer: about
10.4%.]
Question 5
The logistic model assumes that the death rate per capita increases with the population.
The neutrons ying around inside a nuclear reactor behave in the opposite waytheir
death rate [from being absorbed by uranium atoms] is fairly constant, but their birth
rate [from being released by decaying uranium atoms] increases with their number. Model
and analyse.
Question 6
You have 200 bugs in a bottle. Every day you supply them with food and count them.
After two days you have 360 bugs. It is known that the birth rate for this kind of bug is
150% per day. [Is this a sensible way of stating a birth rate per capita? Why?] Assuming
that the population is given by a logistic model, nd the number of bugs after 3 days.
Predict how many bugs you will have eventually. [Answers: about 372; about 376.]
Question 7
In the logistic model, we assume that the death rate per capita is given by D = sN.
But this does not really make sense for low populations, because there will always be
a minimum non-zero death rate per capita due to death from old age. So actually an
equation like D = D
0
+ sN, where D
0
is a positive constant, would make more sense,
since now D can never fall below D
0
. What dierential equation do you get now? Solve
and analyse.
2
ANSWERS TO MA1506 TUTORIAL 5
Question 1
The cost of making the can is a function of the radius r given by
C(r) = J(2r
2
+
2V
r
) +K 2r
C
(r) = J(4r
2V
r
2
) + 2K
= J(4r 2h) + 2K = 0
h
r
= 2 +
K/J
r
,
where J is the cost of aluminium per square cm, V is the volume of the can [a xed value
equal to the area of the base times the height, h], and K is the cost per cm of welding on
the top of the can. We have dierentiated and set the derivative equal to zero so as to nd
the minimum cost. [It clearly is a minimum.]
Measurement shows h 12.5 cm, r 2.5 cm, so K/J = 7.5cm, so each cm of welding
is about 7.5 times as expensive as 1 cm
2
of aluminium. So that is why the can has this
unusual shape: welding on the top is an expensive process. [If welding were cheap, h/r
would be about 2, not 5.]
Question 2
Following the standard equations for the Malthus Model [Chapter 3]:
N =

Ne
kt
; N(0) = 10000 =

N
N(2.5) = 10000e
2.5k
= 11000
e
2.5k
= 1.1 k =
1
2.5
n(1.1)
= 0.0381
N(10) = 10000e
10k
= 10000e
10(0.0381)
14600
20000 = 10000e
kt
t =
1
k
n(2)
= 18.18 hours
Question 3
Let c be the number of emigrants per year.
dN
dt
= kN c. You can solve this in the usual
way, as a linear ODE, or use a trick: let M = N
c
k
so
dM
dt
=
dN
dt
= k(N
c
k
)
= kM
1
So M = Ae
kt
i.e. N
c
k
= Ae
kt
. Let N(0) =

N, so
A =

N
c
k
i.e. N =
c
k
+ (

N
c
k
)e
kt
Three cases:
c
k
<

N, so

N
c
k
> 0
exponential growth there is no point in sending out the emigrants, since the Earths
population continues to grow exponentially.
Next case:
c
k
=

N, N =
c
k
=

N. This is the desired situation, with a constant population
on Earth.
Last case,
c
k
>

N, Earths population decreases to zero! Presumably not what you really
want.
Next, we assume an emigration rate proportional to t, so
dN
dt
= kN ct. This is a linear
ODE with an integrating factor e
kt
, [or you can use undetermined coecients] so the
solution is
e
kt
_
ct
e
kt
= ce
kt
_
1
k
te
kt
_
1
k
2
e
kt
_
+constant
_
integrate
by
parts
=
c
k
_
t +
1
k
_
+Ae
kt
.
Now

N N(0) = A+
c
k
2
N(t) =
_

N
c
k
2
_
e
kt
+
c
k
_
t +
1
k
3 cases:
N
c
k
2
> 0 population explosion on Earth;
N =
c
k
2
Earths population grows linearly; not quite an explosion but still not so
great;
N
c
k
2
< 0: population will grow for a while but eventually reach a maximum, followed
by a decline to zero [because the exponential function will eventually defeat a linear one;
note that we know that the population will grow initially because
dN
dt
(t = 0) =

Nk > 0.] So none of the outcomes is really satisfactory in this case.
Question 4
N = Ae
(BD)t
(constant B and D, so we have a Malthusian situation.)
Population doubles in 20 years, so
2 = e
(BD)20
B D =
n2
20
2
After the departure of the women, B is zero so N = ce
Dt
1
2
= e
D10
D =
n2
10

B =
n2
20
+
n2
10
B 0.10397 i.e. about 10.397% per year.
Assumptions:
1. men and old women have same death rate as young women, which is not true in
reality because men smoke, get into ghts etc while on the other hand old women are
indestructible;
2. the death rate of the remaining population is not changed by the departure of the
girls, a very questionable assumption since morale will be aected, etc.
Question 5
dN
dt
= (B D)N. We set B = sN, D= constant, opposite of the logistic model.
dN
(SN D)N
= dt =
s/D
SN D

1/D
N
t + constant =
1
D
n|SN D|
1
D
nN
|sN D|
N
= ke
Dt
Suppose rst that N < D/s so that |SN D| = D sN [remember that |x| = x for
negative x.] So equation

becomes
D sN
N
= ke
Dt
If N(0) =

N then
Ds

N
N
= K > 0 since we are assuming D sN > 0 always. Then
DsN
N
=
Ds

N
N
e
Dt
.
Solving for N we get
N(t) =
D
s +
Ds

N
N
e
Dt
Suppose instead that N > D/s, |sN D| = sN D
sND
N
= ke
Dt
N =
D
ske
Dt
[Note that setting t = 0 we have s
D
N
= k > 0.]
Clearly N= const =
D
s
is a solution.
So if

N < D/s the population of neutrons will die out. If

N > D/s it will explode (literally
!), tending to innity in a nite time given by
0 = s ke
Dt
, or t =
1
D
n
s
S
D
N
3
Question 6.
The logistic equation has 3 kinds of solution, one increasing, one constant, and one de-
creasing. Since the number of bugs in this problem clearly increases, the relevant solution
of the logistic equation is
N =
B
s +
_
B
N
s
_
e
Bt
=
N
1 +
_
N
N
1
_
e
Bt
Here

N = 200, B = 1.5, so at t = 2 we have
360 =
N
1 +
_
N
200
1
_
e
1.52
360 +
360
200
e
3
N
360e
3
= N
=
360(1 e
3
)
1
360
200
e
3
376
N(3) =
N
1 +
_
N
200
1
_
e
4.5
372
Question 7
Using the suggested form of D, we have
dN
dt
= BN DN = BN (D
0
+SN)N
= (B D
0
)N sN
2
= B
N sN
2
(B
BD
0
) so
dN
dt
= B
N sN
2
which is the logistic dierential equation. So nothing
new in that sense.
However, note that in the logistic model, B < 0 is obviously impossible, but here it IS
possible to have B
< 0. So we have a new case. Clearly if B
< 0 then
dN
dt
= B
N sN
2
< 0
so N always DECREASES. However, it cannot reach zero. To see this, note that N 0 is
a solution of this equation. According to the no-crossing principle, this means that no
other solution can reach zero. So the solutions of this equation just look like the solutions
of the standard logistic equation, but now there are new solutions which just decrease
monotonically, approaching zero asymptotically.
Note: the students may ask: how is it possible that D can be non-zero even when N is
zero? The answer is the usual one: this is just a model, which works in some circumstances
but not in all. For extremely low populations, the whole concept of a death rate per capita
loses its usefulness anyway.
4
MA1506 TUTORIAL 6
Question 1
In question 6 of Tutorial 5, let us assume that you are keeping the bugs not as a hobby,
but because you are developing a new insecticide. Suppose that you remove 80 bugs per
day from the bottle, and that all of these bugs die a painful but well-deserved death as a
result of being sprayed with this insecticide. What is the limiting population in this case?
What is the maximum number of bugs you can put to death per day without causing the
population to die out? [Answers: 312, 141.]
Question 2
The sandhill crane is a beautiful Canadian bird with an unfortunate liking for farm crops.
For many years the cranes were protected by law, and eventually they settled down to a
logistic equilibrium population of 194,600 with birth rate 9.866% per year. Eventually the
patience of the farmers was exhausted and they managed to have the hunting ban lifted.
The farmers happily shot 10000 cranes per year, which they argued was reasonable enough
since it only represents about 5% of the original population. Show that the sandhill crane
is doomed. How long will it take, from the legalisation of hunting, to exterminate them?
[If you dont want to do the integral yourself, you can get a computer to do it for you
here: http://wims.unice.fr/wims/ Or you could use Matlab.] [Answer: about 30 years.]
Question 3
Suppose that Peruvian shermen take a xed number of anchovies per year from an
anchovy stock which would otherwise behave logistically, apart from occasional natural
disasters. According to our lecture notes, any shing rate B
2
/4s will be disastrous.
Lets call this number E
. The shermen want to take as many anchovies as they safely

can, meaning that they want the sh to be able to bounce back from a natural disaster
that pushes their population down by 10%. Advise them. [The men, not the sh.] That
is, tell them the maximum number of sh they can take, expressed as a percentage of
E
. [Hint: assume that you start with the stable equilibrium population
2
, and compute
the value of E, the harvesting rate, such that
1
, the unstable equilibrium population,
becomes 90% of
2
.] [Answer: about 99.7%.]
Question 4
In the harvesting model we considered in the lectures, the population will rebound if all
harvesting is stopped. Unhappily, this is not always true: for some animals, if you drive
their population down too low, they will have trouble nding mates, or they will be forced
to breed with relatively close kin, which reduces genetic variability and hence their ability
to resist disease. For such animals [for example, certain rare species of tigers] extinction
will result if the population falls too low, even if all harvesting is forbidden. Biologists
call this depensation. Show that this situation can be modelled by the ODE
dN
dt
= N[B s(N u)
2
],
where N is the population and B, s, and u are constants such that u >
_
B/s. Find the
population below which extinction will occur. [Answer: u
_
B/s.]
Question 5
Ms Tan Ah Lian, the billionaire engineer who patented the idea of making NewImprovedWater
TM
by mixing hydrogen and oxygen in a Plug Flow Reactor, realises that she can save money
in the following way. The concentration of hydrogen in her PFR does not drop very much
near its end [because the exponential function is almost at there]. So why not make
the tube narrower there, and save on construction costs? Instead of having a constant
cross-sectional area, the new TALPFR has a cross-sectional area given by
A(x) = A
0
e
x
,
where A
0
is the cross-sectional area at the top of the reactor and is a positive constant
with units of 1/length [of course]. Show that the new TALPFR behaves like an ordinary
PFR, but with a smaller rate constant [see section 3.6 of the notes]: instead of k, it is
k (u/2),
where k is the true rate constant and u is the speed, taken to be constant, at which the
mixture moves through the reactor.
Question 6
Tan Ah Lian, billionaire engineer, also has another business which was awarded a huge
contract to retro-t a large balcony onto every HDB apartment in Singapore. The contract
xes the physical properties and the total weight W of the building materials to be used
in each balcony and also the length L of the balcony. Ms Tans competitor, Ah Huat
Contractor Services, proposed a cantilever design with a constant weight per unit length,
= W/L, but the ingenious Ah Lian defeated Ah Huat by proposing a cantilever with
the same total weight but with weight per unit length being given by
2[1 (x/L)],
where x is distance from the point of attachment. [Verify that the two designs have the
same total weight.] That is, the weight is concentrated near to the point where the balcony
is to be attached, and it tapers o towards the end. Ah Lian claims that her balcony
design is better than Ah Huats, because the [absolute value of the] bending moment at
the point of attachment is smaller so it will be cheaper to attach, since less reinforcement is
needed. This is how she won the contract, even without using her considerable personal
charm. How much smaller is the absolute value of the bending moment of Ah Lians
balcony compared to Ah Huats? [Please use Eulers equation from lectures, not your
knowledge of physics if any.] Which balcony dips less at the end? By what factor?
[Answer: the magnitude of ALs bending moment at x = 0 is 2/3 that of AH; ALs
balcony dips less at the end, by a factor of 8/15.]
You may enjoy playing with the parameter values at this site:
http://www.efunda.com/formulae/solid mechanics/beams/theory.cfm
2
Question1.
First compare 80 with
B
2
4s
.
From Tutorial 5 we know B = 1.5 and and N
= 376, so N
= B/s s =
1.5
376

B
2
4s
= 141. This is the maximum number we can kill without causing extinction.
Setting E = 80,
2
=
B
B
2
4Es
2s
=
64
312
.
Since the initial number of bugs was 200, which is between these two values, we see that
the limiting number is
2
= 312, since this is the stable equilibrium.
Question 2.
We have B
=
B
s
= 194600 so since B = 0.09866, s =
B
N
=
0.09866
194600
.
Maximum hunting rate is
B
2
4s
=
(0.09866)
2
4
0.09866
194600
= 4800
Since 10000 > 4800, birds are doomed.
Since
dN
dt
= N(B sN) E,
_
T
0
dt =
_
0
194600
dN
N
_
0.09866
0.09866N
194600
_
10000
We get 29.8 years if E = 10000.
Question 3.
For the sh to survive a 10% downward uctuation, we must have (in the extreme
case)
1
= 90%
2
i.e.
B
B
2
4Es
2s
= 0.9
_
B +
B
2
4Es
2s
_
B
= 0.9B + 0.9
0.1B = 1.9
0.01B
2
= 3.61(B
2
4Es) = 3.61B
2
14.44Es
14.44E = 3.6B
2
/s
E = 0.2493074
B
2
s
= 0.997
_
B
2
4s
_
So a less than 1% drop in the catch below E
will give a 10% margin of safety.

Question 4.
dN
dt
= N[Bs(N u)
2
] i.e. death rate per capita is s(N u)
2
which is large when
N is large BUT ALSO when N is small. This is what we want.
The expression on the right hand side is a cubic in N, and its graph has the typical
shape of a cubic.
Roots of this cubic:
N[B s(N u)
2
] = 0
N = 0, N = u
_
B/s, N = u +
_
B/s.
Since we are assuming u >
_
B/s, both non-zero roots are positive; the graph of the
cubic goes down to a minimum [after passing through the origin], then up through the
point N = u
_
B/s on the N axis, reaching a maximum and then going down again to
cut the N axis once more at N = u+
_
B/s. So there are 2 equilibrium populations, one
stable, one unstable, EVEN THOUGH THERE IS NO HARVESTING. The unstable
equilibrium is at N = u
_
B/s and we see that the tigers will become extinct if the
population ever falls below that value, EVEN IF WE DONT HUNT THEM! So we
have a good model of depensation.
Question 5
Follow the lecture notes [Chapter 3 section 6]. Again we have to use the fact
that the dierence between the rates at which hydrogen ows in and out of a plug [of
thickness x] is precisely the rate at which it is destroyed by the chemical reaction.
Hydrogen ows IN at the point x at a rate C
H
2
(x)A(x)u, where we note that A is now
a function of x [whereas it was a constant in the notes]. H
2
molecules are owing OUT
at the point x+x at a rate C
H
2
(x+x)A(x+x)u. H
2
molecules are being destroyed
by the reaction inside the small piece of tube, so their number is changing at a rate
2rA(x)x. [Here it doesnt matter whether we write A(x) or A(x+ x), because we
are going to drop nonlinear terms in small quantities anyway.] Since the total matter
content cannot increase or decrease,
C
H
2
(x)A(x)u C
H
2
(x +x)A(x +x)u 2rA(x)x = 0.
Putting C
H
2
C
H
2
(x +x) C
H
2
(x) and A A(x +x) A(x), and neglecting all
terms quadratic in small quantities, we have
C
H
2
Au C
H
2
Au 2rAx = 0
so taking the limit as x tends to zero we get [after dividing throughout by uAx]
dC
H
2
dx
=
dln(A)
dx
C
H
2

2r
u
=
dln(A)
dx
C
H
2

2k
u
C
H
2
,
where, as in the notes, we assume that the rate is a constant multiple of the concentra-
tion.
Now Ah Lians reactor has a particular shape, A(x) = A
0
e
x
, and substituting
this into the above equation we nd
dC
H
2
dx
= C
H
2

2k
u
C
H
2
= 2[
k u/2
u
]C
H
2
which is indeed exactly the equation we had in the lecture notes, except that k has been
replaced by k (u/2). Thus while the new design may save money on construction, it
slows down the reaction, and in fact the thing will not work at all if u/2 approaches k.
Question 6
If the weight per unit length is 2[1 (x/L)], then the total weight is obtained by
integrating this from 0 to L; the answer is L, which is indeed exactly the same weight
as Ah Huats balcony [which has constant weight per length .]
We have to solve Eulers equation
d
4
y
dx
4
= w(x)/EI
with exactly the same boundary conditions as in the lecture notes [section 7 of Chapter
3]. The only dierence is that instead of w(x) = constant = we now have w(x) =
2[1 (x/L)]. You just have to integrate four times! I leave the details to you. You
should nd along the way that
y
=
2
EI
[
x
2
2

x
3
6L

Lx
2
+
L
2
6
],
and that
y =
x
2
60EIL
[5Lx
2
x
3
10L
2
x + 10L
3
].
From the lecture notes, the bending moment at x = 0 is related to y
(0) by M(0) =
EIy
(0) and so we have Ah Lians bending moment at the point of attachment:

|M
AL
(0)| =
L
2
3
.
From the lecture notes you can nd that Ah Huats bending moment is [in absolute
value] quite a bit more than this:
|M
AH
(0)| =
L
2
2
.
So Ah Lians design is indeed better. Furthermore, substituting x = L into the above
formula for y, you will nd that the dip at the end of Ah Lians balcony is just
y
AL
(L) =
L
4
15EI
,
very much less [in absolute value] than Ah Huats dip
y
AH
(L) =
L
4
8EI
,
which is what we found in the lectures.
Remark: well-informed students may realise that the rst part of this question can
be answered simply by computing the total torque exerted by the weight of the balcony,
which is just the integral of x w(x) from 0 to L.
MA1506 TUTORIAL 7
1. Find the Laplace transforms of the following functions [where u denotes
the unit step function and the answers are given in brackets]:
(a) t
2
e
3t
. [
2
(s + 3)
3
]
(b) tu(t 2). [e
2s
{
1
s
2
+
2
s
}]
2. Find the inverse Laplace transforms of the following functions:
(a)
s
s
2
+ 10s + 26
. [e
5t
(cos t 5 sin t)]
(b) e
2s
1 + 2s
s
3
. [(
1
2
t
2
2)u(t 2)]
3. Solve the following initial value problems using Laplace transforms:
(a) y
= tu(t 2), y(0) = 4. [(

1
2
t
2
2)u(t 2) + 4]
(b) y
2y
= 4, y(0) = 1, y
(0) = 0. [e
2t
2t]
4. (i) Show, from the denition of the Laplace transform, that for any
function f(t),
L[tf(t)] = F
(s),
where F(s) is the Laplace transform of f(t). Hence nd the Laplace
transform of t sin(t).
(ii) Use Laplace transforms to solve the resonance equation
y + y = cos(t),
where y(0) = y(0) = 0. You should recognise the solution!
5. The oil tanker in Tutorial 4 is at rest in an almost calm sea. Suddenly,
at time t = T > 0, it is hit by a single rogue wave [http://en.wikipedia.org/
wiki/Rogue wave] which imparts to it a vertical [upward] momentum
P, doing so almost instantaneously. Neglecting friction, solve for x(t),
the downward displacement of the ship, and graph it. How far down
1
does the ship go [if it doesnt sink!]? [Hint: remember that, according
to Newtons second law, momentum is the integral of force. So to get
the force as a function of time in this problem, you have to nd a func-
tion which is zero except at t = T, and which has an integral equal to
P.] You may like to play with the Laplace solver at http://www.aw-
bc.com/ide/ [Note that it allows you to put a Dirac delta function into
the right hand side this is a hint, by the way.][Answer: the ship goes
down either to P/M or to Davy Jones http://en.wikipedia.org/wiki/
Davy Jones Locker]
6. In Question 5, suppose that you dont want to assume that the wave
hits instantaneously: you want to model the situation by assuming
that the momentum P is imparted to the ship over a short but non-
zero period of time , starting at t = T. Explain how you would do
this, using the step function. [Remember how to use the step function
to model situations where something is turned on, then turned o.]
Compute the Laplace transform of the displacement function. Show,
using LHopitals rule, that, in the limit 0, this Laplace transform
tends to the one you found in Question 5. That is reasonable, right?
7. When music is recorded digitally, for example for a CD, it can only
be sampled; that is, you record it only at discrete times, not contin-
uously. [The theory of this is called signal processing in engineering
http://en.wikipedia.org/wiki/Signal processing .] This sampling pro-
cess can be represented mathematically using an impulse train, also
known as Diracs comb, dened by
T
(t) =
k=
(t kT).
You can see that multiplying a continuous function by the comb essen-
tially throws away information about the function except at t = 0, t =
T, t = 2T, etc; this is what is meant by sampling. Note that the
comb is periodic with period T. Find the Fourier series of the comb.
If you graph the sum of the rst few terms [say, the rst 10] of this
series you will see why
T
(t) is called a comb. Play with the java
applet in the Fourier page at http://www.jhu.edu/signals/index.html
to understand how Fourier series perform these miracles.
2
Question 1.
(a) We shall use the following s-Shifting property:
L(f(t)) = F(s) L(e
ct
f(t)) = F(s c)

L(t
2
) =
2
s
3
use L(t
n
) =
n!
s
n+1

L(t
2
e
3t
) = L(e
3t
t
2
) =
2
(s + 3)
3
(b) Here u denotes the Unit Step Function given by
u(t a)
_
0 if t < a
1 if t > a
We shall use the following t-Shifting property:
L(f(t)) = F(s) L{f(t a)u(t a)} = e
as
F(s)
Letf(t 2) = t

f(t) = t + 2

L(f(t)) = L(t + 2) = L(t) + 2L(1)
=
1
s
2
+
2
s

L(tu(t 2)) = L{f(t 2)u(t 2)}
= e
2s
(
1
s
2
+
2
s
)
Question 2. (a)
s
s
2
+ 10s + 26
=
s
(s + 5)
2
+ 1
=
(s + 5) 5
(s + 5)
2
+ 1
LetF(s) =
s 5
s
2
+ 1

L
1
(
s
s
2
+ 10s + 26
) = L
1
(F(s + 5))
= L
1
(F(s (5))
= e
5t
L
1
(F(s)) use s-shifting
= e
5t
L
1
(
s
s
2
+ 1

5
s
2
+ 1
)
= e
5t
{L
1
(
s
s
2
+ 1
) 5L
1
(
1
s
2
+ 1
)}
= e
5t
(cost 5 sin t)
(b) Let F(s) =
1+2s
s
3
=
1
s
3
+
2
s
2

L
1
(F(s)) =
t
2
2
+ 2t (use L(t
n
) =
n!
s
n+1
)
Let f(t) =
t
2
2
+ 2t
Using t-shifting,
L
1
(e
2s
1 + 2s
s
3
) = L
1
(e
2s
F(s))
= f(t 2)u(t 2)
= {
(t 2)
2
2
+ 2(t 2)}u(t 2)
=
1
2
(t
2
4)u(t 2)
= (
1
2
t
2
2)u(t 2)
Question 3. (a)
Let L(y(t)) = Y (s)
We shall use L(y
(t)) = sY (s) y(0).

We have
L(y
) = L(tu(t 2))
sY (s) 4 = e
2s
(
1
s
2
+
2
s
)
Y (s) = e
2s
(
1 + 2s
s
3
) +
4
s

y(t) = L
1
(Y (s))
= L
1
{e
2s
(
1 + 2s
s
3
)} + 4L
1
(
1
s
)
= (
1
2
t
2
2)u(t 2) + 4
(by a previous question.)
(b) We shall use
L(y
) = s
2
Y sy(0) y
(0)
We have
L(y
2y
) = L(4)
s
2
Y sy(0) y
(0) 2{sY y(0)} =

4
s
s
2
Y s 2sY + 2 =
4
s
(s
2
2s)Y =
4
s
+ s 2 =
4 + s
2
2s
s
Y =
s
2
2(s 2)
s
2
(s 2)
=
1
s 2

2
s
2

y = L
1
(
1
s 2

2
s
2
)
= e
2t
2t
Question 4.
By denition,
F(s) =
_

0
e
st
f(t)dt.
Dierentiating both sides with respect to s [NOT t think of t as a constant in this
calculation, since s is doing the changing here] we get
F
(s) =
_

0
te
st
f(t)dt,
where the chain rule has been used. So
F
(s) =
_

0
e
st
[tf(t)]dt = L[tf(t)]
as required.
Now the Laplace transform of sin(t) is 1/(1 + s
2
). Hence the Laplace transform of
t sin(t) is minus the derivative of this, ie
L[tsin(t)] =
2s
(1 + s
2
)
2
.
For the resonance equation
y + y = cos(t),
take the Laplace transform of both sides; with the given initial data, we get
s
2
Y + Y =
s
(1 + s
2
)
,
and so
Y (s) =
s
(1 + s
2
)
2
.
Thus
y(t) =
1
2
tsin(t),
which is indeed exactly the solution we got in Chapter 2 for resonance in this case. [The
formula in the notes is x =
F
0
t
2m
sin(t); here F
0
= = m = 1.] Notice how the Laplace
method automatically takes care of the extra factor of t.
Question 5.
The original ODE describing this situation in the absence of friction was
M x = Mg A(d + x)g,
where the downward direction is positive. We need an extra term to account for the
sudden force exerted by the rogue wave. Since the force is exerted suddenly, this suggests
that we need a Dirac delta function, so the force will be proportional to (t T). Now
F = ma, Newtons law, can be written as F = [time derivative of mv], where v is the
velocity, so the change in the momentum is equal to the integral of the force. Recall
that
_

0
(t T)dt = 1, so clearly in our case F = P(t T), since P is the given
change in the momentum [integrate both sides to verify this, and remember that x(t)
here is the DOWNWARD displacement so the upward force of the wave [as stated in
the problem] is negative, like the buoyancy force]. So we have
M x = Mg A(d + x)g P(t T),
which, as in Tutorial 4, simplies to
x =
Ag
M
x
P
M
(t T).
Taking the Laplace transform of both sides, remembering that the ship is initially at
rest, we have
s
2
X =
Ag
M
X
P
M
e
Ts
,
or
X(s) =
P
M
e
Ts
s
2
+
2
=
P
M
e
Ts
s
2
+
2
,
where is the natural frequency of oscillation of the ship,
_
Ag/M. Using the t-
shifting theorem, we can nd the inverse Laplace transform:
x(t) =
P
M
sin[(t T)]u(t T).
The graph is at until t = T, then you get the usual Simple Harmonic Motion with
angular frequency . That makes sense the sudden impulse given by the wave should
trigger o SHM. The amplitude is P/M, so this is the maximum distance the ship goes
down if it doesnt sink.
Question 6
The force suddenly switches on at t = T and then switches o at t = T + . It
should therefore be proportional to u(t - T) - u(t - (T + )), since this function behaves
in just that way. Note that its integral from 0 to is . So we should set
F =
P
[u(t T) u(t (T + ))]

again you can check this by integrating both sides and remembering that the change
in momentum should be the integral of the force, and that up is negative in this problem.
So now the dierential equation in question 5 becomes
x =
Ag
M
x
P
M
[u(t T) u(t (T + ))].
Take the Laplace transform of both sides to get
s
2
X(s) =
Ag
M
X(s)
P
Ms
[e
Ts
e
(T+)s
],
and so
X(s) =

P
Ms
[e
Ts
e
(T+)s
]
s
2
+
2
,
where is dened as before. You can obtain the solution using the t-shifting theorem
as before.
Using LHopitals rule, one sees that
1
[e
Ts
e
(T+)s
] se
Ts
as 0. Hence our expression for X(s) does indeed tend to the same expression for
X(s) as in Question 5 if you let tend to zero. Thats as expected, because the impulse
should be like a delta function if is very short. So you can always think of a delta
function as a shorthand for a dierence of two step functions in this manner.
Question 7
Fourier coecients for any function f(t) with period T are given by
a
n
=
2
T
_
T/2
T/2
f(t)cos
_
2nt
T
_
dt,
b
n
=
2
T
_
T/2
T/2
f(t)sin
_
2nt
T
_
dt,
and the Fourier series is then
a
0
2
+
1
_
a
n
cos
_
2nt
T
_
+ b
n
sin
_
2nt
T
_
_
.
Now in the denition of Diracs comb, only one of the terms in the innite sum is non-
zero in the interval T/2 to T/2, namely the k = 0 term, so we only have to include
that term when we work out the integrals. We have
a
n
=
2
T
_
T/2
T/2
(t)cos
_
2nt
T
_
dt =
2
T
cos(0) =
2
T
,
b
n
=
2
T
_
T/2
T/2
(t)sin
_
2nt
T
_
dt =
2
T
sin(0) = 0,
using the property of the delta function explained in the notes. So the Fourier series of
the comb is just
1
T
+
2
T
1
cos
_
2nt
T
_
.
This may look ridiculous mathematically, but if you use graphmatica/matlab/whatever
to graph, for example,
y = 1/2+cos(x)+cos(2x)+cos(3x)+cos(4x)+cos(5x)+cos(6x)+cos(7x)+cos(8x)+cos(9x),
you can convince yourself that if you take enough terms you will indeed end up with
something that looks like Diracs comb. You are summing waves which interfere de-
structively everywhere except at regularly spaced points, where they suddenly interfere
constructively a bit like a rogue wave!
MA1506 TUTORIAL 8
Question 1
Despite her vast wealth, Tan Ah Lian continues to live with her mother. The latter does
TALs laundry [of course], including a large heavy jacket. TALs mother hangs the jacket
out to dry on a bamboo pole inserted into a socket outside the kitchen window. The pole
has negligible mass but the wet jacket has a mass of M kg. It is suspended at a point
exactly A metres along the pole, which is of length L metres. Assume that the usual
beam parameters E and I are given for this pole, and assume for simplicity that the pole
is perfectly horizontal before the jacket is attached. Use Laplace transforms to nd the
shape of the pole after it is loaded, assuming that it doesnt break. To do this, you need
some information as follows. Let y(x) represent the shape of the pole, with x = 0 being
the point where the pole is inserted in the socket. Then the third derivative at zero is
y
(0) = Mg/EI, because the shear force there has to balance the weight of the jacket.
Similarly y
(0) =
Mg
EI
A because the torque at x = 0 has to balance the torque exerted
by the jacket at x = A. [If you dont know the physics involved, just ignore this and take
these initial conditions as given.] The function w(x) giving the force per unit length can
be modelled by using a delta function, w(x) = Mg(x A). [Answer: for x A we
have y(x) =
Mg
EI
1
2
x
2
A
1
6
x
3
, for x A we have y(x) =

Mg
EI
1
2
xA
2
1
6
A
3
.]
Question 2
Tan Ah Lian attributes her enormous success to the fact that she never talked in class
when she was an Engineering student at NUS. Unfortunately, certain other students,
destined for badly paid minor administrative jobs if they manage to graduate at all, do
not share her virtues. One day in the lecture the prof announces that a certain gadget
contains a circuit with a resistance, capacitance, and inductance, which were all stated,
but TAL could not hear all of the numbers mentioned due to the incessant babbling of
the other students; all she could hear was that the resistance is 2 ohms. Undeterred, she
steals back into the room after class and quickly switches the gadget on and o at t =
0, thus ring a short burst of voltage into it, and observes that the resulting current at
t > 0 is I(t) = e
t
cos(t) e
t
sin(t) amperes. She then deduces what the prof must have
said about the inductance and the capacitance. Follow her good example and also deduce
these numbers. [See Chapter 2 and recall the formula for the Laplace transform of an
integral.][Answer: L = 1 and C = 1/2 in the appropriate units.]
Question 3
Find AB and BA if A =
1 2 3
4 5 6
, B =
7 0
8 1
. [Answer: BA =
7 14 21
4 11 18
,
and Im not telling you what AB is. If you feel the need for further routine exercises
about matrices, http://www.sosmath.com/matrix/matrix.html is recommended.]
Question 4
Express the matrix
1 2
3 4
as the sum of a symmetric matrix and an antisymmetric

matrix. [Answer:
1 5/2
5/2 4
0 1/2
1/2 0
.]
Question 5
Pick some 2 by 2 and 3 by 3 matrices and work out their determinants. Check your
answers at http://wims.unice.fr/wims/ or using matlab. If your answer is wrong, print
out what the computer claims, and take it to ask your tutor.
Question 6
In the discussion of shear matrices in the notes, we assumed that the shearing forces were
parallel to the x-axis. Find the matrix of a shear [shearing angle 30 degrees] when the
shearing forces are parallel to an axis which makes an angle of 45 degrees with the x axis.
[Hint: Rotate the axis down to the x-axis, do the shear there, and then rotate back up to
the original direction.] Check that your matrix has determinant equal to 1, as it should
[why?]. [Answer:
1
1
2
3
1
2
1
2
3
1 +
1
2
.]
Question 7
The exponential of a square matrix B is dened to be the innite series
e
B
= I + B +
B
2
2!
+
B
3
3!
+ .....,
where I is the identity matrix. Show that the rotation matrix
cos sin
sin cos
can be
expressed as e
A
, where A =
0 1
1 0
. [Hint: show that all powers of A are proportional

either to the identity matrix or to A itself.] Express the shearing matrix
1 tan
0 1
as
the exponential of a matrix.
2
MA1506 TUTORIAL 8 SOLUTIONS
Question 1
From the given hint, we see that we have to solve
d
4
y
dx
4
=
Mg
EI
(x A),
subject to the given boundary conditions. [Note that y(0) = y
(0) = 0 since the pole is

horizontal at the point where it joins the wall. In case you are wondering about the signs
of the second and third derivatives, the shearing force at that point has to be positive
because it has to balance the negative weight of the whole pole, and the torque has to be
anticlockwise [negative] to balance the clockwise torque exerted by gravity. If you dont
know enough physics to understand this then I suggest that you forget it.] Taking the
Laplace transform of both sides we get
s
4
Y (s) s
3
y(0) s
2
y
(0) sy
(0) y
(0) =
Mg
EI
e
sA
from which
Y (s) =
Mg
EI
s
4
e
sA
s
3
MgA
EI
+ s
4
Mg
EI
.
Taking the inverse we have
y(x) =
Mg
EI
1
6
(x A)
3
u(x A) +
1
2
x
2
A
1
6
x
3
,
where u(x - A) represents the unit step function as usual. So for x A we have
y(x) =
Mg
EI
1
2
x
2
A
1
6
x
3
,
but for x A we get
y(x) =
Mg
EI
1
6
(x A)
3
+
1
2
x
2
A
1
6
x
3
=
Mg
EI
1
2
xA
2
1
6
A
3
.
Question 2
From Chapter 2 we have, if V(t) is the applied voltage, I(t) is the current, L is the
inductance, R is the resistance, and C is the capacitance,
V (t) = RI + L

I +
1
C
t
I dt.
The problem here is that we dont actually know V(t); all we know is that it is some
multiple of (t). So we set V (t) = A(t) where A is some unknown constant. Thus we
have
A(t) = RI + L

I +
1
C
t
I dt.
Take the Laplace transform of both sides, and let (s) denote the transform of I(t); then
we have
A = R(s) + Ls(s) +
1
sC
(s),
recalling that the transform of the integral is given by (1/s) times the transform of the
integrand. Solving for (s) we get
(s) =
As
Ls
2
+ Rs + (1/C)
.
But the Laplace transform of the given current is
(s) =
s + 1
(s + 1)
2
+ 1

1
(s + 1)
2
+ 1
=
s
s
2
+ 2s + 2
=
As
As
2
+ 2As + 2A
,
from which we see immediately that R = 2A; since TAL knows that R = 2, we see that
A = 1. Clearly C must have been 1/2 and L must have been 1 in the appropriate units.
Question 3
B produces 2-dimensional vectors, but A can only eat 3-dimensional vectors, so AB is not
dened.
Question 4
In lectures we saw that given any square matrix A, A + A
T
is symmetric and A - A
T
is
antisymmetric. So if we write
A =
1
2
[A + A
T
] +
1
2
[A A
T
],
we see that any matrix can be expressed as a sum of a symmetric with an antisymmetric
matrix. The given answer is obtained by applying this idea to
1 2
3 4
.
Question 5
Note to students: you need to do this. And the more you think you dont, the more you
do.....
Question 6
As explained in the hint, the idea is to rotate the axis down to the x-axis, which is done
by means of the matrix
1/
2 1/
2
1/
2 1/
. We then shear by 30 degrees, which since

the shearing matrix is
1 tan
0 1
, corresponds to the matrix
1 1/
3
0 1
. Then we
rotate back up to the original axis, by means of
1/
2 1/
2
1/
2 1/
. So the matrix we
need is [note the order]
1/
2 1/
2
1/
2 1/
1 1/
3
0 1
1/
2 1/
2
1/
2 1/
.
If you multiply it out, you get the result stated. The determinant is 1, as it should be
because it is the very nature of a shear to preserve areas, and the determinant is the
2
measure of how much areas change under the action of a linear transformation. Note also
that each of the three matrices above has determinant equal to 1, so their product must
also have determinant equal to 1.
Question 7
Clearly A
2
= I, and A
3
= A, and A
4
= I, and A
5
= A and so on; all of the even
powers are equal to I, and all of the odd powers are A. So it makes sense to separate
them as follows:
e
A
= I + A +
1
2!
2
A
2
+
1
3!
3
A
3
+
1
4!
4
A
4
+ .....
= I
1
2!
2
I +
1
4!
4
I + .....
+A
1
3!
3
A +
1
5!
5
A ....
Taking out the common factors of I and A, we get
e
A
= I[1
1
2!
2
+
1
4!
4
+ .....]
+A[
1
3!
3
+
1
5!
5
....
But we recognise these Taylor series: we have
e
A
= Icos + Asin
=
cos 0
0 cos
0 sin
sin 0
which is the rotation matrix.

The case of
1 tan
0 1
is simpler: dene B =
0 1
0 0
and note that all powers of

B vanish. So we have
e
Btan
= I + Btan + 0 + 0 + ..... =
1 tan
0 1
.
3
MA1506 TUTORIAL 9
Question 1
One of the many virtues of Laplace transforms is that they turn calculus problems into
algebra problems. That works also for systems of simultaneous rst-order ordinary dier-
ential equations, as follows. Suppose you want to solve the following pair of simultaneous
equations:
dx
dt
= 4x 5y,
dy
dt
= 2x 2y
with x(0) = 0 and y(0) = 2. You can write this as
d
v
dt
= B
v,
where
v(t) is a column vector built from x(t) and y(t), and B is the matrix of coecients.
Now the Laplace transform of a vector function of time is dened in the obvious way
[treating each component separately] so if we let
V (s) be the Laplace transform of
v(t),
then we have
s
V (s)
v(0) = B
V (s).
This is now an algebra problem: by inverting a matrix you can nd
V (s), and then taking

the inverse Laplace transform you can nd
v(t). Do this and nd x(t), y(t). [Answer: x(t)

= 10sin(t)e
t
, y(t) = 2e
t
cos(t) 6e
t
sin(t). Compare with the example on pages 14-16 of
Chapter 7.]
Question 2
Billionaire engineer Tan Ah Lian believes that she can get even richer by gambling. To
this end, she goes to an Integrated Resort
1
and plays the following game [along with
several other players]. The players and the croupier each ip coins. If a players coin
matches that of the croupier [both heads or both tails] then the player pays $1. If they
do not match, the house pays the player $1. [This kind of game is designed to prevent
cheating by either party.] Initially TAL has $3. If at any point she loses all her money,
she will be violently thrown out of the den with probability 1, and the game goes on
without her. If at any point she wins a total of $2, then she will also be booted out
even more violently with probability 1, because the owner of the gambling den is her old
bitter enemy Lim Ah Huat, who doesnt allow anyone, especially Tan Ah Lian, to make
more than $2 from him. What is the probability that TAL will be broke by the time 5
rounds of this game have been played? What is the probability that she will have been
thrown out, by 5 rounds, for being too successful? Is Ah Huat a born loser?[Hint: over
the course of the game, there are six possible amounts of money that TAL can have. Set
up a 6 by 6 matrix to represent the Markov process of this game; that is, the rst column
represents the probabilities of reaching the six dierent possible amounts of money given
that TAL has $0, the second column represents the probabilities given that she has $1,
and so on. You can use the matrix calculator at http://wims.unice.fr/wims/ to work out
the necessary power of this matrix.][Answers: 22%, 38%.]
1
Gambling den
Question 3
The Leontief model can be applied to the economies of entire countries, as follows. The
economy of the Republic of Progensia consists of Agriculture, Manufacturing, and Energy,
and the corresponding technological matrix [in the order AME for both columns and
rows] is
0.30 0.00 0.00

0.10 0.20 0.20
0.05 0.01 0.02
, so for example this means that each Progensian tael of

Agricultural produce requires 0.30 taels worth of Agricultural produce, 0.10 taels worth
of manufactured goods, and 0.05 taels worth of energy [the tael being the Progensian
currency, 1 tael = 314.159 $US]. Progensias government hopes to export 140 million taels
worth of agricultural produce, 20 million taels worth of manufactured goods, and 2 million
taels worth of energy this year. Find out how much agricultural produce, manufactured
goods, and energy they have to produce in order to meet this target. [Answers: 200.2,
53.52, 12.04 million taels respectively.]
Question 4
Use the eigen-engine at the IDE website
[http://www.aw-bc.com/ide/ideles/media/JavaTools/eignengn.html] to nd the eigen-
values and eigenvectors of the following matrices:
2 1
0 2
1 1
1 1
1 4
1 1
2 1
1 2
0 0
0 1
Question 5
Find the eigenvectors and eigenvalues of the matrices in Question 2 by hand, that is, not
by using any computer other than the one inside your head.
Question 6
In our study of Markov processes, we used matrices of a special kind, called a [left]
stochastic matrix http://en.wikipedia.org/wiki/Stochastic matrix. In such a matrix, the
entries are numbers between 0 and 1, such that the total of the entries in each column
is 1. This is because the entries represent probabilities, and probabilities in any given
situation always have to total to 1. [Check this for our gambling game, above.] Prove
that a 22 matrix of this sort must have at least one eigenvalue equal to 1. Why is this
important?
Question 7
Use the wims website to nd all of the eigenvectors of
1 2 3
4 5 6
7 8 9
. You should nd that

this matrix has exactly two non-parallel eigenvectors with non-zero eigenvalues. The third
eigenvector is not parallel to either of these, but it has eigenvalue zero. Explain why this
implies that this matrix has rank 2.
The two non-parallel eigenvectors with non-zero eigenvalues dene a 2-dimensional space,
a plane [like any pair of non-parallel vectors in three-dimensional space]. Show that
2
the vector
1
2
4
does not lie in this plane. Hence explain why the system of equations
1 2 3
4 5 6
7 8 9
x
y
z
1
2
4
has no solutions. Show that
1
2
3
does lie in this plane, and

explain why this means that
1 2 3
4 5 6
7 8 9
x
y
z
1
2
3
has innitely many solutions.

3
MA1506 Tutorial 9 SOLUTIONS
Question 1. We have
V (s) = [sI B]
1
v(0),
or
V (s) =
s 4 5
2 s + 2
0
2
=
1
s
2
2s + 2
s + 2 5
2 s 4
0
2
and so
X(s) =
10
(s 1)
2
+ 1
, Y (s) =
2(s 1)
(s 1)
2
+ 1

6
(s 1)
2
+ 1
.
Computing the inverse Laplace transforms in the usual manner gives the stated answers.
Question 2. The thing to remember here is that there are six possible states in which
the cunning Miss Tan can nd herself: she can have $0, $1, $2....up to $5. She can never
have more than that because Ah Huat wont allow it. If at any point she has exactly $0,
then the probability that she will still have $0 after a round of the game is 1, because by
then she has been booted out into the street. So the rst column of the matrix is
1
0
0
0
0
0
.
If at any time she has $1, then there is a probability of 1/2 that she will have $0 after the
next round, and a probability 1/2 that she will have $2; all other probabilities are zero.
So the next column is
1/2
0
1/2
0
0
0
. Similarly for $2,3,4; in each case she can either win $1 or

lose it, each with probability 1/2, so there are exactly two non-zero entries in each of those
columns [the 3rd, 4th, and 5th]. Finally, if at any point she nds herself with $5, then the
probability that she will have $5 next time is 1, because again she will soon nd herself
outside in the gutter if she does that; her gambling career is over. So the nal column is
0
0
0
0
0
1
. Hence the full Markov matrix in this case is just
1 1/2 0 0 0 0
0 0 1/2 0 0 0
0 1/2 0 1/2 0 0
0 0 1/2 0 1/2 0
0 0 0 1/2 0 0
0 0 0 0 1/2 1
.
The fth power of this matrix can be computed by the wims website. We only care about
its fourth column, because that contains all of the probabilities given that Miss TAL started
with $3. That fourth column is
0.21875
0
0.25
0
0.15625
0.375
. The probability that she is thrown out into

the street with zero dollars is the top entry, 0.21875; the probability that she gets kicked
out, sore but triumphant, with the vast sum of $5 in her pocket is the last entry, 0.375.
Note that there is no chance that she has either $1 or $3 at this point; that is interesting,
though irrelevant. You might like to compute higher powers of this matrix. One nds, of
course, that it becomes more and more likely that TAL will be thrown out one way or the
other [that is, the sum of the rst and last entries approaches 1]. Note that Ah Huat is,
as always, the loser here, in that TAL is signicantly more likely to exit with a prot than
with a loss. [The ratio of probabilities is 0.375/0.21875 = 1.7143.]
Question 3. According to Chapter 5, we just have to work out
[I T]
1
140
20
2
where T is the technology matrix
0.30 0.00 0.00

0.10 0.20 0.20
0.05 0.01 0.02
. So the answers are given by
1.43 0.00 0.00

0.20 1.25 0.26
0.07 0.01 1.02
140
20
2
200.2
53.52
12.04
.
Question 4.
1
0
, 2
1
1
, 0,
1
1
, 2
1
1/2
, 3,
1
1/2
, 1
1
i
, 2 i,
1
i
, 2 + i
1
0
, 0,
0
1
, 1
Question 5.
2 1
0 2
= 0 (2 )
2
= 0
= 2 (only one eigenvalue)
2 1
0 2

1
0 1
0 0

1
0
0
= 0
1
0
1 1
1 1
= 0 (1 )
2
= 1
= 0, 2
= 0
1 1
1 1

1
= 0 = 1
1
1
= 2
1 1
1 1

1
= 0 = 1
1
1
1 4
1 1
= 0 (1 )
2
= 4
= 3, 1
= 3
2 4
1 2

1
= 0 =
1
2

1
1/2
= 1
2 4
1 2

1
= 0 = 1/2
1
1/2
2 1
1 2
= 0 (2 )
2
= 1
= 2 i, 2 + i
= 2 i
i 1
1 i

1
= 0 = i
1
i
= 2 + i
i 1
1 i

1
= 0 = i
1
i
0
0 1
= 0 ( 1) = 0
= 0, 1
= 0
0 0
0 1

1
= 0 = 0
1
0
= 1
1 0
0 0

= 0 = 0
0
1
Question 6 A 22 stochastic matrix has the form S =
a b
1 a 1 b
. For this to
have an eigenvalue equal to 1, we must have
det
a 1 b
1 a b
= 0,
which is obviously true. It is important because in studying Markov processes, we need
to take large powers of S. If both eigenvalues were less than 1 in absolute value, then S
n
would tend to zero as n becomes large; if one or both were greater than 1 in absolute value,
not all of the components of S
n
could remain bounded. Neither possibility makes sense,
since the entries have to be interpreted as probabilities.
Question 7. Eigenvectors are
1
16
(3
33 19)
1
8
(3
33 11)
1
1
16
(3
33 + 19)
1
8
(3
33 + 11)
1
2
1
Any vector with T
u = 0 is an eigenvector. There is only one eigenvector here with

eigenvalue zero (the last one). So the rank is 2.
To show that
1
2
4
does not lie in that plane, just work out the triple product
(rst eigenvector) (second eigenvector)
1
2
4
You will nd that the answer is 2.154211... which is not zero as it would have to be if
1
2
4
lay in that plane.

If you repeat this process (for example, get WIMS to work out the determinant) you will
nd that
1
2
3
does lie in that plane i.e.

(rst eigenvector) (second eigenvector)
1
2
3
= 0.
There will be innitely many solutions because if
x
y
z
is a solution, so is
x
y
z
+r
1
2
1
where r is any number.

MA1506 TUTORIAL 10
Question 1
Wherever possible, diagonalize the matrices in questions 4/5 of Tutorial 9. [That is, write
them in the form PDP
1
, after nding P and D, where D is diagonal.] Using this, work
out their 4th powers.
[Answers: The rst one cannot be diagonalized; the fourth powers of the others are
_
8 8
8 8
_
,
_
41 80
20 41
_
,
_
7 24
24 7
_
,
_
0 0
0 1
_
.]
Question 2
The sequence
0, 1, 1, 2, 3, 5, 8, 13, 21, 34, 55, 89, 144, .......
in which each term is the sum of the two preceding ones, is called the Fibonacci sequence
[http://en.wikipedia.org/wiki/Fibonacci number ]. These numbers are very popular with
mathematicians, also with believers in the Da Vinci Code, and other nutcases. Let x
k
denote the kth Fibonacci number, where x
0
is 0. Show that
_
x
k+1
x
k+2
_
=
_
0 1
1 1
__
x
k
x
k+1
_
.
Dene V
k
=
_
x
k
x
k+1
_
, so V
k+1
= AV
k
, where A =
_
0 1
1 1
_
. By diagonalizing the
matrix, express V
k
in terms of V
0
. Hence prove Binets formula,
x
k
=

k
+
5
,
where
=
1
2
[1
5].
Notice the mildly interesting fact that the two numbers
are irrational, and yet x

k
is a
whole number. Find lim
k
x
k+1
x
k
. This limiting ratio is called the golden ratio, see [but
read with a skeptical mind] http://en.wikipedia.org/wiki/Golden ratio.
Question 3
The great white shark [http://en.wikipedia.org/wiki/Great white shark] likes to eat seals.
Let
k
denote the number of sharks in a given area at year k, and let
k
denote the number
of seals in that area. In year k+1, the number of sharks depends on the number in year
k and also on the number of seals [the more seals there are in year k, the more food for
the sharks]. It is found that
k+1
=

k
2
+

k
100
.
Notice that this means that the sharks would die out if there were no seals for them to
eat, which makes sense, since seals are their main food. Similarly the number of seals in
year k+1 is determined by the number of sharks and by the number of seals in year k. It
is found that
k+1
=
5
k
4

50
k
4
.
Notice the minus sign: seals tend to disappear into the mouths of sharks. Notice too
that the seal population would grow uncontrollably if the sharks were not kind enough to
remove surplus seals. Suppose we start with 50 sharks and 1600 seals in year zero. How
many sharks and seals will we have in the long run? [Answer: 14 sharks and 700 seals.]
What is the answer if the 50/4 in the second equation is changed to 51/4? Comment.
Question 4
Let B be a diagonalizable matrix. Prove that det(e
B
) = e
Tr(B)
. Verify this for the 2 2
rotation matrix in question 7 of Tutorial 8.
Question 5
Romeo is fed up with Juliets ckle ways, and decides to dump her. He runs o to Venice
in disgust and takes up instead with Desdemona, a more straightforward kind of girl
whose love for Romeo grows when he likes her, and decreases when he doesnt. Romeo
remains as before. Consider the special case in which Desdemona has no feelings either
way for Romeo when the latter rst falls in love with her. Will Desdemonas love for
Romeo ever exceed his love for her? Can the two star-crossed lovers ever have a stable
relationship?
Question 6
Meanwhile Juliet, in a rage, leaves Verona [her home town, where she met Romeo] and
follows Romeo to Venice, where she discovers Romeo madly in love with Desdemona.
Heartbroken, she immediately falls in love with Othello, who is a ckle lover like Juliet:
that is, if a girl likes him, his opinion of her immediately drops, but his opinion be-
gins to improve the moment she shows signs of hating him
1
. Predict the outcome, as-
suming that Othello initially likes Juliet too. Draw some typical phase plane diagrams.
http://en.wikipedia.org/wiki/Othello
1
He loves not wisely, and maybe not too well either.
2
1. From the results of Tutorial 9,
_
1 1
1 1
_
=
_
1 1
1 1
__
0 0
0 2
__
1/2 1/2
1/2 1/2
_
_
1 1
1 1
_
4
=
_
1 1
1 1
__
0 0
0 16
__
1/2 1/2
1/2 1/2
_
=
_
8 8
8 8
_
_
1 4
1 1
_
=
_
1 1
1/2 1/2
__
3 0
0 1
__
1/2 1
1/2 1
_
_ _
4
=
_
41 80
20 41
_
_
2 1
1 2
_
=
_
1 1
i i
__
2 i 0
0 2 +i
__
1/2 i/2
1/2 i/2
_
_
2 1
1 2
_
4
=
_
1 1
i i
__
(2 i)
4
0
0 (2 +i)
4
__
1/2 i/2
1/2 i/2
_
=
_
7 24i 7 + 24i
24 + 7i 24 7i
__
1/2 i/2
1/2 i/2
_
=
_
7 24
24 7
_
_
0 0
0 1
_
=
_
1 0
0 1
__
0 0
0 1
__
1 0
0 1
_
_
0 0
0 1
_
4
=
_
1 0
0 1
__
0 0
0 1
_
4
_
1 0
0 1
_
=
_
0 0
0 1
_
2. By denition x
2+k
= x
1+k
+x
k
so
_
x
k+1
x
k+2
_
=
_
0 1
1 1
__
x
k
x
k+1
_
We need eigenvalues and eigenvectors of
_
0 1
1 1
_
1
1 1
= 0
2
1 = 0
=
1
2
_
+1
=
Eigenvectors are
_
1
+
__
1
_
1
So
_
0 1
1 1
_
=
1
5
_
1 1
__
+
0
0
__
+
1
_
Hence
_
0 1
1 1
_
k
=
1
5
_
1 1
__
k
+
0
0
k
__
+
1
_
=
1
5
_
1 1
__
k
+

k
+
_
=
1
5
_

+
k
+

k
+
k+1

k+1
+

k+1
+

k+1
_
Now V
k
= AV
k1
= A
2
V
k2
= A
3
V
k3
= etc etc etc = A
k
V
0
= A
k
_
0
1
_
So
V
k
=
_
x
k
x
k+1
_
=
_ __
0
1
_
=
1
k
+
k+1
+

k+1
x
k
=
1
5
_
k
+
_
x
k+1
=
1
5
_
k+1
+

k+1
x
k+1
x
k
=
_
k+1
+

k+1
_
/
_
k
+
_
=
+
_
k
1
_
+
_
k
But
5
1+
< 1 so lim
k
_
+
_
k
= 0, hence
x
k+1
x
k
+
.
3.
_
k+1
k+1
_
=
_
1/2 1/100
50/4 5/4
__
k
_
Usual procedure
_
1/2 1/100
50/4 5/4
_
=
_
1 1
25 50
__
3/4 0
0 1
__
2 1/25
1 1/25
_
Now clearly
_
k+1
k+1
_
=
_
1/2 1/100
50/4 5/4
_
k+1
_
0
_
2
Long run means k so since (3/4)
k
0 as k what we need is
_
1 1
25 50
__
0 0
0 1
__
2 1/25
1 1/25
__
50
1600
_
=
_
1 1/25
50 2
__
50
1600
_
=
_
14
700
_
.
If the 50/4 is changed to 51/4, then the eigenvalues change to 0.9896, 0.7604, both
of which are numbers less than 1 in absolute value; so when you take large powers
of them, they both tend to zero, so both seals and sharks die out. This shows that
this kind of model is not really satisfactory, since a small change in the parameter
values changes the conclusions in a drastic way.
4.
Since B is diagonalizable we can write it as B = PDP
1
for some diagonal matrix
D [with the eigenvalues of B down the diagonal] and some other matrix P. So
e
B
= I +B +
1
2
B
2
+
1
6
B
3
+...
gives us
e
B
= I +PDP
1
+
1
2
PD
2
P
1
+
1
6
PD
3
P
1
+...
and so
e
B
= P[I +D +
1
2
D
2
+
1
6
D
3
+...]P
1
= Pe
D
P
1
.
Thus det(e
B
) = det(e
D
). But since D is diagonal, it easy to see that its exponential
is just a diagonal matrix with e
1
, e
2
, e
3
etc down the diagonal, where
1
,
2
,
3
etc
are the eigenvalues of B. So
det(e
D
) = e
1
e
2
e
3
... = e
1
+
2
+
3
+...
= e
Trace(B)
since the trace of B is just the sum of its eigenvalues. [Actually B doesnt have to be
diagonalizable, the statement is true in general.]
For the 22 rotation matrix, the determinant is 1, and the matrix of which it is
the exponential has zero trace, which agrees since 1 = e
0
.
5. Let R(t) and D(t) be functions representing the feeling of Romeo and Desdemona.
We have
dR
dt
= aD,
dD
dt
= bR where a and b are BOTH positive. Also we know
that D(0) = 0, R(0) = > 0 from the wording of the question. Since D(0) = 0,
this
dR
dt
(0) = 0. Clearly

R = a

D = abR R = Ae
t
+ Be
t
, =

ab.
Initial conditions R = cosh(t). Also D =
1
a
R =

a
sinh(t). Now we have
an expression for D/R and we can see what happens as t tends to innity:
D
R
=

a
tanh(t)

a
;
3
that is,
D
R

_
b/a
Question is whether D can be > R i.e. whether
D
R
> 1. Clearly the answer is yes
if b > a.
Note that
R
2
a

D
2
b
=

2
a
cosh
2
t

2
2
a
2
b
sinh
2
t
=

2
a
(cosh
2
t sinh
2
t) =

2
a
= const Hyperbola.
If b is large then D will eventually exceed R.
4
But if b is small, that will never happen. Clearly no stable relationship is possible
under any circumstances: there is a point of equilibrium at D = R = 0, but it is
obviously unstable. With the given initial conditions, the whole relationship gets
totally out of control. But in a good way.
6. Let J(t) represent Juliet and Z(t) Othello. Notice that if we reverse the direction
of time, t t, then dt dt. So the equations here,
dJ
dt
= aZ,
dZ
dt
= bJ
are actually the SAME as in Question 5, but with the direction of time reversed.
So to get the phase plane diagrams here we just have to reverse the arrows.
We are told that J(0) > 0 and Z(0) > 0 so we start in the rst quadrant. From the
diagram [which I am afraid you have to draw for yourself its a typical saddle
diagram with the contracting axis in the rst and third quadrants we see that
there are two possible outcomes. In one case J while Z + while in
the other J + while Z . In both cases one of them is consumed by
hatred but the other one refuses to save him/herself by running away. In both cases
the result is tragic. [In principle, the initial conditions could be such that J and Z
move along the unique trajectory passing through the origin, in which case both
J and Z tend to zero as time tends to innity. This rather depressing possibility
would however require the initial conditions to be chosen with innite accuracy, so
it would never happen in reality.]
5
MA1506 TUTORIAL 11
Question 1
Classify the systems of ODEs with the following matrices. [That is, say whether they
represent a nodal source, spiral sink, etc etc etc.]
1 3
1 2
2 2
4 0
2 4
10 0
5 4
2 1
5 4
2 1
0 1
10 0
.
Go to http://www.aw-bc.com/ide/ideles/media/JavaTools/romejuli.html to check your
answers by looking at the phase diagrams. [You will have to divide all of these matrices
by 10 to use that website.] Make sure that you check a wide variety of initial conditions
[that is, starting points].
Question 2
Both Elves and Dwarves live in Rivendell, but there is a certain irrational resentment
between the two groups [placed in their hearts, no doubt, by Sauron]. The amount of
racial prejudice is not the same on both sides, Elves tending to be more intolerant than
Dwarves, but both groups are inclined to move out of Rivendell at a rate controlled by
the number of the other group. Apart from this, both groups reproduce as usual, but
Elves can only be killed by violence they never grow old so their overall death rate
per capita is lower than that of the Dwarves; also, the Elvish birth rate per capita is
higher since Dwarf women are scarce [ever seen one?]. Both groups have more births
than deaths. We can set up a model of this situation using a pair of simultaneous ODEs
with a matrix [where the rst row describes the rate of change of the Elf population]
5 4
1 2
. Explain why this matrix does represent the situation we described. At a

certain time, the number of Dwarves is slightly larger than the number of Elves. Predict
what will happen to the population of Elves. This pattern has in fact often been observed
in racially discriminatory societies.
Question 300
King Xerxes I of Persia has sent a million soldiers to conquer Greece. King Leonidas I of
Sparta decides to meet the Persians at Thermopylae. A typical Persian soldier can kill
one Spartan per hour, whereas a typical Spartan soldier can kill 11,111,111.1 Persians per
hour. Neither side suers from any disease
1
. How many soldiers does King Leonidas need
to take to Thermopylae if he wants to kill all of the Persians but expects no Spartans to
go home alive? [Answer: come on....]
Question 4
The highly dangerous Iranian nuclear programme consists of two 100-gallon tanks. Tank
A is initially lled with water in which 25 lbs of uranium hexauoride are dissolved;
tank B contains no uranium hexauoride initially. Pure water is poured into tank A at
1
They all die before they have time to get sick.
the constant rate of 4 gal/min. The well mixed solution from tank A is constantly being
pumped to tank B at a rate of 6 gal/min, and the solution in tank B is constantly pumped
to tank A at the rate of 2 gal/min. The solution in tank B also exits the tank at the rate
of 4 gal/min.
1. Write down equations for the amount of uranium hexauoride in the tanks. You
should get a pair of simultaneous rst-order linear equations. Write them in matrix
form and so nd the amount of uranium hexauoride in each tank at any time.
2. Go to http://www.aw-bc.com/ide/ideles/media/JavaTools/romejuli.html and use
the right-hand panel of the java window to draw graphs to show how the uranium
hexauoride level in each tank changes with respect to time. [You will have to
multiply your matrix by 10.]
3. Does the amount of uranium hexauoride in tank B ever exceed that in tank A?
4. What kind of phase plane diagram do we have here?
Ans: (1) x
A
(t) = 12.5(e
1
t
+e
2
t
), x
B
(t) = 12.5(
3e
1
t
3e
2
t
) with
1
=
6+2
3
100
and
2
=
62
3
100
; (3) Yes; (4) Nodal sink.
Question 5
Lions like to eat zebras. The lions have a constant death rate per capita of 20% per year,
but their [eective] birth rate per capita is not constant: it depends on the number of
zebras. [More zebras = more food for mother lion = more milk = more surviving baby
lions.]. In fact the eective birth rate per capita is B
L
= uZ, where Z is the number
of zebras and u is a constant equal to about 0.0008 [in certain units what sort of
units are they?]. Meanwhile, the zebras have a constant birth rate per capita of about
5%, but of course their death rate per capita is not constant: it depends on how many
lions there are around: we have D
Z
= sL, where s is a constant approximately equal to
0.004, and L is the number of lions. Write down a pair of simultaneous dierential equa-
tions governing the functions L and Z. These are the famous Lotka-Volterra equations,
http://en.wikipedia.org/wiki/Lotka volterra
Compute the equilibrium population of lions and zebras, assuming that neither pop-
ulation is zero. [Answer: 12.5 lions, 250 zebras.]
The dierential equations you obtained are not linear, so our theory does not apply
to them. However, suppose we consider small deviations away from equilibrium. That
is, we set
L = 12.5 + x
Z = 250 + y,
where x and y are functions of time which are initially small. Substitute these expressions
into your equations and drop all quadratic terms, so that you get a pair of linear equations.
Was your equilibrium point stable? Describe the behaviour of L and Z near equilibrium.
Play with http://www.aw-bc.com/ide/ideles/media/JavaTools/popltkvl.html to check
your answer.
2
1.
1 3
1 2
Tr = 1, det = 5 SADDLE
2 2
4 0
Tr = 2, det = 8, T
2
r
4 det = 28 < 0 SPIRAL SOURCE
2 4
10 0
Tr = 2, det = 40, T
2
r
4 det = 36 < 0 SPIRAL SINK
5 4
2 1
Tr = 4, det = 3, T
2
r
4 det = 4 > 0 NODAL SINK
5 4
2 1
Tr = 4, det = 3, T
2
r
4 det = 4 > 0 NODAL SOURCE
0 1
10 0
Tr = 0, det = 10, T
2
r
4 det = 40 < 0 CENTRE
2. Let E(t) be the number of elves, D(t) the number of Dwarves. Let B
E
, D
E
be the
birth and death rates per capita for Elves, similarly B
D
, D
D
for Dwarves. We are
told that B
E
> B
D
and D
E
< D
D
. So B
E
D
E
> B
D
D
D
> 0 (more births
than deaths).
Now
dE
dt
is controlled by (B
E
D
E
)E in the usual Malthus model, but here we are
told that
dE
dt
is reduced by the presence of Dwarves, so we propose
dE
dt
= (B
E
D
E
)E P
E
D
where P
E
is a constant that measures the prejudice of the elves. Similarly
dD
dt
= (B
D
D
D
)D P
D
E
Where P
D
represents the prejudice of the Dwarves. So
dE
dt
dD
dt
(B
E
D
E
) P
E
P
D
(B
D
D
D
)

E
D
So in a concrete case, the constants in the matrix should satisfy B

E
D
E
>
B
D
D
D
> 0, and we are told that P
E
> P
D
. Indeed
5 4
1 2
satises all of
these. Tr = 7, det=6, T
2
r
4 det = 49 24 = 25 > 0 so we have a nodal source.
Eigenvalues
(5 )(2 ) 4 = 0 = 1 or 6.
1
Eigenvectors are
1
1
1
1/4
.
So the phase plane diagram
(we only care about the
rst quadrant, since
D 0, E 0) is
bisected by the line D = E. All points ABOVE that line will move along trajectories
that eventually hit the D axis (that is, E = 0). So if at any time D > E, then the
Elf population may increase for a while, but eventually it will reach a maximum
and then collapse to zero. Rivendell is completely taken over by Dwarves, EVEN
THOUGH B
E
> B
D
AND D
E
< D
D
. The prejudice of the Elves cancels out their
other advantages and causes them to lose the competition.
3. Let S(t) be the number of Spartans, P(t) the number of Persians. We have
dS
dt
= P
dP
dt
= 11, 111, 111.1S
so
dS
dt
dP
dt
0 1
11, 111, 111.1 0

S
P
Eigenvalues
2
11, 111, 111.1 = 0 = 3333.3333
Eigenvectors
3333.3333 1
11, 111, 111.1 3333.3333

1
= 0
= 3333.3333.
Since we are interested only in the rst quadrant, take +. We have a SADDLE.
2
The trajectory that passes
through the origin is a straight
line parallel to the eigenvector
1
3333.3333
. That straight
line has equation P = 3333.3333S.
We want to start at the point on this straight line that has P = 1000000. So the
initial value of S has to be
1000000
3333.3333
= 300.
So Leonidas should take 299 soldiers, assuming that he intends to ght (and die)
himself.
4. (1) Use the rule:
Rate of change of the amount of UF
6
= (concentration in) (ow rate in) - (concentra-
tion out) (ow rate out), where concentration is the mass of uranium hexauoride
per unit volume of water you can understand this equation by looking at the
units:
lbs
sec
=
lbs
gallons

gallons
sec
Let x
A
be the mass of UF
6
in the rst tank, so the concentration in that tank is
x
A
/100. Similarly let x
B
be the mass of UF
6
in the second tank. Then from the
given data we have
x
A
=
2
100
x
B
6
100
x
A
and
x
B
=
6
100
x
A
6
100
x
B
.
Notice that the 4 gallons/min of pure water owing into the rst tank does not
appear here it contains no UF
6
; it is just there so that the amounts of water in
the tanks remain constant. The initial conditions are x
A
(0) = 25, x
B
(0) = 0.
In matrix form, we have
x
A
x
B
=
1
100
6 2
6 6

x
A
x
B
.
The eigenvalues of the coecient matrix can be found as
1
=
6 + 2
3
100
and
2
=
6 2
3
100
. The corresponding eigenvectors are
and
.
Hence

x
A
x
B
= c
1
e
1
t
+c
2
e
2
t
.
Substituting the initial values, we have c
1
= 12.5, c
2
= 12.5, so we get x
A
(t) =
12.5(e
1
t
+e
2
t
), x
B
(t) = 12.5(
3e
1
t
3e
2
t
).
3
b) In the java applet, x corresponds to x
A
and y to x
B
. You have to click on a point
on the x axis because we are told that x
B
[that is, y] is zero initially. You should
see the red curve going steadily down, but the green curve goes up at rst and then
down.
c) Yes. The two curves intersect when t
0
=
25
3
ln(2 +
3) 19, (which is in fact

the maximum point of x
B
). Before t
0
, the uranium hexauoride in tank A is more
than that in tank B; after that, its always less.
d) Here Trace = 12/100 and det = 24/100, so Trace
2
4 det = (144-96)/10000 which
is positive; hence we have a nodal sink. That is also clear from the java applet.
5.
dL
dt
= B
L
L D
L
L
(B
L
=birth rate per capita, D
L
=death rate per capita.)
But B
L
= uZ so
dL
dt
= uZL D
L
L. (Units of u =
1
year
.)
Similarly
dZ
dt
= B
Z
Z D
Z
Z
= B
Z
Z sLZ
At equilibrium,
dL
dt
=
dZ
dt
= 0, so
uZ DL = 0 Z =
D
L
u
B
Z
sL = 0 L =
B
Z
s
L =
0.05
0.004
= 12.5, Z =
0.20
0.0008
= 250.
Substitute L = 12.5 +x, Z = 250 +y, get
dL
dt
=
dx
dt
= 0.0008(250 +y)(12.5 +x) 0.2(12.5 +x)
= 0.01y + 0.0008xy
0.01y
dZ
dt
=
dy
dt
= 0.05(250 +y) 0.004(12.5 +x)(250 +y)
= x 0.004xy x
Hence

dx
dt
dy
dt
0 0.01
1 0

x
y
Tr = 0, det = 0.01
4
Tr
2
4 det < 0 centre.
So near equilibrium, the Lion and Zebra populations should uctuate up and down
but never vary much from their equilibrium values. The equilibrium is stable.
5

MA1506

Transféré par

Informations du document

Copyright

Formats disponibles

Partager ce document

Partager ou intégrer le document

Options de partage

Avez-vous trouvé ce document utile ?

Ce contenu est-il inapproprié ?

Droits d'auteur :

Formats disponibles

MA1506

Transféré par

Droits d'auteur :

Formats disponibles

MA1506 TUTORIAL 1

1. Solve the following dierential equations:

. Use this to check that your answers were correct.

F and then left to cool in a room at 75

F in half an hour. Show that its temperature will be 81.6

= . The right side is equal to 1 at t = 0, as it should

can be expressed in terms of y only, with no y

.] These can be solved with the aid of

= F(y) can be reduced to a separable

19t) + 2.35 cos(100t) 9.41 sin(100t)]

(/2)[x /2] + ....

(3/2)[x 3/2] + ....

19t) + 2.35 cos(100t) 9.41 sin(100t)

Ag/M. So eventually the ship bobs up

< 0. So we have a new case. Clearly if B

. The shermen want to take as many anchovies as they safely

will give a 10% margin of safety.

(0) and so we have Ah Lians bending moment at the point of attachment:

= tu(t 2), y(0) = 4. [(

(t)) = sY (s) y(0).

(0) 2{sY y(0)} =

[u(t T) u(t (T + ))]

, for x A we have y(x) =

as the sum of a symmetric matrix and an antisymmetric

. [Hint: show that all powers of A are proportional

(0) = 0 since the pole is

. We then shear by 30 degrees, which since

, corresponds to the matrix

which is the rotation matrix.

and note that all powers of

V (s) be the Laplace transform of

V (s), and then taking

v(t). Do this and nd x(t), y(t). [Answer: x(t)

0.30 0.00 0.00

, so for example this means that each Progensian tael of

. You should nd that

has no solutions. Show that

does lie in this plane, and

has innitely many solutions.

. Similarly for $2,3,4; in each case she can either win $1 or

. Hence the full Markov matrix in this case is just

. The probability that she is thrown out into

where T is the technology matrix

0.30 0.00 0.00

. So the answers are given by

1.43 0.00 0.00

Question 6 A 22 stochastic matrix has the form S =

Any vector with T

u = 0 is an eigenvector. There is only one eigenvector here with

lay in that plane.

does lie in that plane i.e.

where r is any number.

are irrational, and yet x

. Explain why this matrix does represent the situation we described. At a

So in a concrete case, the constants in the matrix should satisfy B

3) 19, (which is in fact

Vous aimerez peut-être aussi