CHAPTER 2.

MORE APPLICATIONS OF
ODEs
2.1. THE HARMONIC OSCILLATOR
Consider the pendulum shown.
The small object, mass m,
at the end of the pendulum,
is moving on a circle of radius
L, so the component of its velocity
tangential to the circle is L
Hence its tangential acceleration is

L and so by F = M a we have

mL = mg sin .
1

An obvious solution is = 0. This is called an EQUILIBRIUM solution, meaning that is a CONSTANT

function. This means that if you set = 0 initially,
then will remain at 0 and the pendulum will not
move which of course we know is correct. There
is ANOTHER equilibrium solution, = . Again,
IN THEORY, if you set the pendulum EXACTLY
at = , then it will remain in that position forever. IN REALITY, of course, it wont! Because
the slightest pu of air will knock it over! So this
equilibrium is very dierent from the one at = 0.
This is a very important distinction!
Equilibrium is said to be STABLE if a SMALL push
2

away from equilibrium REMAINS small. If the small

push tends to grow large, then the equilibrium is
UNSTABLE. Obviously this is important for engineers! Especially you want vibrations of structures,
engines, etc to remain small.
Lets look at = . By Taylors theorem, near =
, we have
1
f () = f () + f ()( ) + f ()( )2 + ...
2

Now sin() = 0, sin () = cos() = 1, sin () =

sin() = 0 etc so
1
sin() = 0 ( ) 0 + ( )3 etc
6
3

For small deviations away from , is small,

( )3 is much smaller, etc, so we can approximate
sin() ( )
so our equation is approximately
M L = mg sin = mg( ).
and now
Let = , so = ,
g
= .
L
The general solution is

= Ae(
so = + = Ae(

g/L)t

g/L)t

+ Be(
+ Be(

g/L)t

g/L)t

+ .

As you know, the exponential function grows very

quickly; so even if is close to initially, it wont
stay near to it very long! Very soon, will arrive
either at = 0 or 2, far away from = . The
equilibrium is UNSTABLE! How long does it take
for things to get out of control? That is determined

by g/L or rather L/g, which has units of TIME.

Note that it takes longer to fall over if L is large.

EXAMPLE:
An eccentric professor likes to balance pendula
near their unstable equilibrium point. In a given
performance, the pendulum is initially slightly away
5

from that point, and is initially at rest. The profs

skill is such that he can stop the pendulum from
falling provided that the angular deviation from the
vertical angle does not double. If the shortest pendulum for which he can perform this trick is 9.8 centimetres long, estimate the speed of his reexes.

Solution: The problem is saying that the angle [the deviation from the vertical] is initially
very small, and its initial rate of change is zero. So

(0) = [some very small number] and (0)

= 0.
Dierentiating our solution for (t) above and substituting we get A + B = and A - B = 0, so in fact
6

A = B = /2 and so

= cosh((

g/L)t) = cosh(10t)

since from the given data g/L = 100/sec2 . Now

our objective is to calculate how long it takes for
to double, so we need to nd t such that 2 =
cosh(10t). Clearly t = cosh1 (2)/10 0.132 sec.

SUMMARY: The equation = + Lg is a symptom

of INSTABILITY. The system is at equilibrium, but
it will run away uncontrollably on a time scale xed

by L/g.

Now what about = 0? Here of course we use Taylors theorem around zero,
1
f () = f (0) + f (0) + f (0)2 + ...
2
1 3
sin() = 0 + 0 + ...
6
so sin() and we have approximately
mL = mg or
g

= = 2
L
with 2 = g/L. That minus sign is crucial!
General solution is C cos( t) + D sin( t) where C
and D are arbitrary constants.
Now using trigonometric identities you can show that
ANY expression of the form C cos(x) + D sin(x) can
8

be written as

C cos(x) + D sin(x) = C 2 + D2 cos(x )

where tan() = D/C. [You can see this easily by

[ ]
C
taking the scalar product of the vectors
and
D
]
[
cos(x)
.]
sin(x)
So now we can write our general solution as

= A cos(t )
[Check: this does satisfy = 2 and it does contain TWO arbitrary constants, A and ]. In this
case, is never larger than A, never smaller than
A, so IF WAS SMALL INITIALLY, it REMAINS
9

SMALL! [We call A the AMPLITUDE.] So the equilibrium in this case is STABLE. This is called SIMPLE HARMONIC MOTION. Clearly repeats its
values every time t increases by 2 [since cos is
periodic with period 2]. Now
t t + 2
means
2
tt+

So

= 2 L/g is the time taken for to return

to its initial value, the PERIOD. Again it takes a

long time if L is large. The number is called the
ANGULAR FREQUENCY.
10

SUMMARY: The equation = 2 is a symptom of STABILITY. The system oscillates, with a

constant amplitude, around equilibrium on a time

scale xed by L/g. The angular frequency is .

2.2. DAMPED, FORCED OSCILLATORS.

When an object moves fairly slowly through air, the

RESISTANCE DUE TO FRICTION is approximately
proportional to its speed, and of course in the OPPOSITE DIRECTION. So in the case of the pendu the DAMPlum, where the speed of the object is L,
11

ING FORCE is
SL
where S is some positive constant. We now have
mL = mg sin SL
mg SL
if is close to zero. Thus
mg

= 0,
m + S +
L
and this is the equation of DAMPED HARMONIC
MOTION. We can also attach a motor to the pendulum, that is, an external force F (t) which may
depend on time. Then
mL = mg sin SL + F (t)
12

and if remains small, we get

mg
1

m + S +
= F (t),
L
L
the equation of FORCED damped harmonic motion.
Lets look at the unforced but damped case rst.
The ODE will always have the form

m
x + bx + kx = 0.

With the usual trick, x = et , we get

m2 + b + k = 0.

Quadratic in , so 3 cases: both roots real, both

complex, both equal.
13

[a] BOTH REAL : OVERDAMPING

Example, x
+ 3x + 2x = 0
2 + 3 + 2 = 0 = 1, 2,
general solution B1 et + B2 e2t . The motion very
rapidly dies away to zero. Obviously we have too
much friction.
[b] BOTH COMPLEX: UNDERDAMPING
Example, x
+ 4x + 13x = 0
2 + 4 + 13 = 0 = 2 3i
general solution B1 e2t cos(3t)+B2 e2t sin(3t), which
can be written as

x = Ae2t cos(3t ).
14

The graph is obtained by multiplying together the

graphs of e2t and A cos(3t ):

This is like a simple harmonic oscillator such that

the amplitude is a function of time.
In general if we have an equation for an unforced,
damped harmonic oscillator, such that
m
x + bx + kx = 0,
m2 + b + k = 0,
then if the solutions for are COMPLEX, we say
15

that the system is UNDERDAMPED. We get

x(t) = Ae

where =

1
2m

bt
2m

cos(t )

4mk b2 . You can think of this as

SHM with frequency and amplitude Ae

bt
2m

Here is often called the QUASI-FREQUENCY and

2

is the QUASI-PERIOD. Notice that in this prob-

lem, UNLIKE in true SHM, there are actually TWO

independent time scales:
so does

2m
b .

has units of time but

This second time scale tells you how

quickly the amplitude dies out. [They are INDEPENDENT because given m and b you can work
out

2m
b

but not

[since you have not been given

16

k.]
2.3. FORCED OSCILLATIONS
Suppose you have a mass m which can move in a
horizontal line. It is attached to the end of a spring
which exerts a force

Fspring = kx

where x is the extension of the spring and k is a constant (called the spring constant). This is Hookes
Law. Now we attach an external MOTOR to the
mass m. This motor exerts a force F0 cos(t), where
F0 is the amplitude of the external force and is the
17

frequency. If F0 = 0 we just have, from Newton,

m
x = kx,

so we get x
= x, = k/m. Here is the fre2

quency that the system has if we leave it alone that

is, it is the NATURAL frequency. It has NOTHING
TO DO with of course we can choose to suit
ourselves.
If F0 = 0, then we have
m
x + kx = F0 cos t.

Let z be a complex function satisfying

m
z + kz = F0 eit .
18

Clearly the real part, Re z, satises the above equation, so we can solve for z and them take the real
part. We try
z = C eit
and get
mC(i)2 eit + C k eit = F0 eit
F0
F0 /m
C=
= 2
k m2
2
So

Re z =

F0 /m
2 2

cos(t)

and the general solution is

x = A cos(t ) +

F0 /m
cos(t).
2
2

Note that
F0 /m
x = A sin(t ) 2
sin(t).
2
19

The arbitrary constants A and are xed by giving

x(0) and x(0)

as usual. For example, suppose x(0) =

x(0)

= 0, then
F0 /m
0 = A cos() + 2
2
0 = A sin().
Assuming F0 = 0, we cannot have A = 0, = 0.
/m
So A = F20
2,

x=

F0 /m
[cos(t) cos(t)] .
2
2

Using the trigonometric identity

(
cos A cos B = 2 sin
20

AB
2

(
sin

A+B
2

we nd

x=

2F0 /m
sin
2
2

[(

) ]
[(
) ]

+
t sin
t
2
2

Now remember that is under our control, is not.

Suppose we adjust to be very close to , but not
equal to it. Then we can think of this solution in the
following way:
{
x=

2F0 /m
sin
2
2

[(

) ]}
[(
) ]
+
t sin
t
2

will

be very SMALL, so LOW-FREQUENCY, while

+
2

The reason for splitting it like this is that

is much larger, high frequency. So if we write

[(
x = A(t) sin
21

+
2

) ]
t

then we have a sine function with an amplitude which

is a (MUCH LOWER-FREQUENCY) sine function!
We say that the system has a BEAT, and

is

called the BEAT FREQUENCY.

Notice that the amplitude function

A(t) =

2F0 /m
sin
2
2

[(

) ]

t
2



 2F0 /m 
has a maximum value  2 2  which becomes very
large when is very close to .
22

What happens if we let ? We have

[ ]
2F0 /m sin 2 t
A(t) =

F0
t
F0 t

=
m 2
2m
by LHopitals rule. So in this limit

x=

F0 t
sin(t)
2m

and we see that the oscillations

go completely out of control.
This situation is called
RESONANCE. We see that IF A
SYSTEM IS FORCED IN A WAY
THAT AGREES WITH ITS OWN
23

NATURAL FREQUENCY, IT CAN OSCILLATE

UNCONTROLLABLY.
This can be very dangerous! In reality resonance
does not get completely out of control, because we
cannot really ignore friction [or RESISTANCE in the
case of an electrical circuit]. So we should really solve
m
x + bx + kx = F0 cos t
m
z + bz + kz = F0 eit
Set z = ceit and take the real part at the end.

F0
F0 [k m2 ib]
c=
=
2
k m + ib
(k m2 )2 + b2 2
(Using the identity

1
A+iB

=
24

AiB
(A+iB)(AiB)

AiB
A2 +B 2 .)

So we need the real part of this complex number

times eit = cos(t) + i sin(t). Now in general we
have

(C + iD)(E + iF ) = CE DF + i[CF + DE]

Similarly here
F0 [k m2 ib]
[cos(t) + i sin(t)]
(k m2 )2 + b2 2
has real part
F0 (k m2 ) cos(t) + F0 b sin(t)
x(t) =
.
2
2
2
2
(k m ) + b
To this we should add the general solution of m
x+
bx + kx = 0. But we know already what that looks
25

like whether overdamped or underdamped, the solution rapidly (exponentially) TENDS TO ZERO.
We call that part of the solution the TRANSIENT.
So after the transient dies o, we are left with this
expression.
We saw earlier that ANY expression of the form
C cos(x) + D sin(x) can be written as

C cos(x) + D sin(x) = C 2 + D2 cos(x )

where tan() = D/C.

So here we have

x(t) =

1
m F0

cos(t )

( 2 2 )2 +
26

b2 2
m2

So the system eventually settles down into a steady

oscillation, BUT AT FREQUENCY , NOT ! Also,
the AMPLITUDE of this oscillation is a FUNCTION
OF ,
F0 /m
A() =
( 2 2 )2 +

b2 2
m2

The graph of this is called the AMPLITUDE RESPONSE CURVE. Depending on the values of the
parameters, this curve might have a SHARP MAXIMUM, meaning that the system will suddenly respond strongly if is chosen to be the value that
gives that maximum. This is something to be avoided
in some cases [things might break] but welcomed in
27

others [you want your mobile phone to ignore all frequencies but one].

2.4. CONSERVATION.
Newtons 2nd law involves TIME derivatives, but
sometimes it can be expressed in terms of SPATIAL
derivatives, by means of the following trick:
(
)
dx
dx dx
d 1 2
x
= x
=
=x
(chain rule).
dx 2
dx
dt dx
For SHM we have
m
x = kx
28

so
d 1 2
m ( x ) = kx.
dx 2
But now we can integrate both sides:
1
1
mx 2 = kx2 + E
2
2
where E is a constant of integration. So we have

E=

As you know,

1
1
mx 2 + kx2 .
2
2

1
2
m
x

is called the KINETIC EN-

ERGY of the oscillator, and 21 kx2 is called the POTENTIAL ENERGY. We call E the TOTAL ENERGY. The fact that E is CONSTANT is called the
conservation of energy.
29

Actually, this idea of turning time derivatives

into space derivatives can be very useful when studying certain kinds of non-linear second-order dierential equations. For example, consider the following
problem:

+ sin() = 0, (0) = 1, (0)

= 1.

This is in fact the pendulum problem (with g/L =

1), but without assuming that is small. So we cant
solve it exactly. But we can still answer questions
like: what is the maximum value of ? We can do
this by using the above method to turn our equation
30

into
2 /2 cos() = C,
()
where C is a constant equal, by the initial conditions,
to (1/2) cos(1).
At the maximum point we have = 0, so we
have
(max) = arccos(cos(1) (1/2)) 1.53.
There is a good way of thinking about what we
are doing here. Look at the equation for and use
it to think of as a function of . If you graph
that function, you will see that the graph is a closed
curve. You can imagine that as time goes by, the
31

 moves around and around that closed

point (, )
curve. This means that the solution, whatever it is,
must be a periodic function of time; and that makes
sense, since the physical system is a pendulum. This
method is called the Phase Plane method.
This trick is useful for many other non-linear
second-order equations too. For example, you can
do similar things with the problem

y + tan(y) = 0, y(0) = 0, y(0)

= 3.

(Graphs of y(t) and of the phase plane for this equation (obtained by using a computer) are attached at
the end of this chapter.)
32

2.5. EULERs EQUATION.

A BEAM is a long, thin object used in buildings etc,

which bends when subjected to loads. Suppose you
have a beam which is only supported at one end,
which we take to be the origin (x, y) = (0, 0); this is
called a CANTILEVER We assume that there is a
LOAD on the beam this might just be the weight
of the beam itself, but it could be things we put on
top of the beam, or it could be other forces pushing
upward. Then of course the beam must bend. We
want to know the shape of the beam; that is, we want
to nd y [the distance through which the beam bends
33

down] as a function of x [the horizontal position].

We measure the load as FORCE PER UNIT LENGTH,

w(x). Note the units of w(x): newtons per metre.
Our convention is that w(x) is positive in the UPWARD direction.
Now of course the amount of bending of the
beam depends on several things. It will of course
depend on the load, but it will also depend on what
the beam is made of you would expect aluminium
to bend dierently from steel, even if the load is
34

the same. This stiness is measured by a constant

E called YOUNGS MODULUS, which has units of
pressure [force per unit area, newtons per square metre]. It could be a function of x.
The amount of bending also depends on the
SHAPE of the cross-section of the beam, measured
by a constant I is called the SECOND MOMENT
OF AREA of the beam. This has units of [length]4 .
It turns out that the deection of the beam, y =
y(x), is governed by the FOURTH-ORDER ODE,
EULERS EQUATION,
d2 [ d2 y ]
EI 2 = w(x)
2
dx
dx
35

[Please check that the left side has the right units,
newtons per metre!] Lets suppose that E and I
are both constant, and that w(x) is only due to the
weight of the beam itself; we assume
w(x) = constant = . (Remember that w is positive in the UPWARD direction.)
QUESTION: What is , the maximum deection,
in terms of L, the horizontal length of the beam?
Answer:
d4 y
w

=
=

4
dx
EI
EI
d3 y
x
=
+A
dx3
EI
by integration. Now for physics reasons
36

d3 y
dx3

=0

when evaluated at x = L [basically because there is

no force there]. Thus
d3 y
L
0 =
(L)
=

+ A.
3
dx
EI
Hence we have
A=

L
EI

So

d3 y
x
L
=

3
dx
EI
EI
d2 y
x2
Lx
=
+
+ B,
dx2
2EI
EI
where we have integrated again. But

d2 y
dx2

= 0 when

evaluated at x = L [basically because there is no

37

twisting at the end]. So

L2
d2 y
L2
0 = 2 (L) =
+
+B
dx
2EI
EI
B=

L2
2EI

L2
EI

L2
2EI

d2 y
x2
Lx L2
2 =
+

dx
2EI
EI
2EI
dy
x3
Lx2
L2 x

=
+

+ C.
dx
6EI
2EI
2EI
But if the beam is horizontal where it is supported,
dy
dx (0)

= 0 so
0 = 0 + 0 0 + C

C=0

Lx3
L2 x2
x4
+

+ D,
y=
24EI
6EI
4EI
where we have integrated yet again! Finally the
point of support is the origin, so y(0) = 0, so
D=0
38

Thus
[
]
L4
1 ( x )4 1 ( x )3 1 ( x )2
y=

2EI
12 L
3 L
2 L
So now nally we can work out the maximum deection of the beam: you can easily show that this
maximum occurs at the end [x = L] in agreement
with common sense, so all we have to do is to substitute x = L into y and get

[
]
L4
1
1
1
= y(L) =

2EI
12
3
2
and so we obtain the famous CANTILEVER DEFLECTION FORMULA
L4
=
.
8EI
39

The negative answer of course reects the fact

that the cantilever always bends DOWNWARDS.
Check that the answer MAKES SENSE: the deection is larger for larger loads , of course that makes
sense. The deection is smaller if EI is large, that
is, if the beam is very sti: that too makes sense.
Finally, the deection is greater if the beam is very
long [large L], which also makes sense. [Are you
in the habit of checking that your answers MAKE
SENSE??]
There is still a surprise here though: notice that
if you make your cantilever twice as long, the downwards deection does not double it increases
40

by a factor of SIXTEEN! So watch out if you want

to build balconies etc. [The power of 4 here is a relic
of the fact that we solved a FOURTH-order ODE.]

41

Figure 1: y + tan(y) = 0, y(0) = 0, y(0)

= 3.

Figure 2: y 2 + 2ln(sec(y)) = 9.