1

PHYS2614: Oscillations, Waves and Optics

DEPARTMENT OF PHYSICS

UNIVERSITY OF THE FREE STATE

2023
 2

1. Mathematical Background

1.1. Introduction

The reader should already be proficient in calculus, both in taking derivatives and integrating functions. We
briefly review some mathematical ideas that will be of importance for the study of oscillations and waves.
Many of these concepts may already be familiar to the reader, while some topics, e.g. parity, could be new.

1.2. Trigonometry

The study of oscillations and waves in physics requires a good knowledge of oscillating mathematical
functions i.e. sinusoidal functions.

(a) (b)
Figure 1 (a) Sides of a right angled triangle relative to an angle 𝜽. (b) Position of a point on a circle.

In terms of a right-angled triangle having an angle 𝜃 at one vertex, such as vertex C in Figure 1(a), the
trigonometric ratios are given by:
opposite hypotenuse opposite
sin 𝜃 = ; sec 𝜃 = ; tan 𝜃 = .
hypotenuse adjacent adjacent

The co-functions are obtained by interchanging the opposite and adjacent sides, because the comple-
mentary angle to 𝜃 (namely 90° − 𝜃 ) is in the opposite corner of 𝜃 of a right-angled triangle i.e. at vertex B
in Figure 1(a):

adjacent hypotenuse adjacent

cos 𝜃 = = sin(90° − 𝜃) ; csc 𝜃 = = sec(90° − 𝜃) ; cot 𝜃 = = tan(90° − 𝜃).
hypotenuse opposite opposite

One can then see that certain functions are reciprocals of others, namely

1 1 1
sin 𝜃 = csc 𝜃 ; sec 𝜃 = cos 𝜃 ; tan 𝜃 = cot 𝜃 .
 3

From the theorem of Pythagoras [square of hypotenuse is sum of squares of other two sides] one also has
opposite 2 adjacent 2 opposite2 + adjacent 2
sin2 𝜃 + cos 2 𝜃 = ( ) +( ) = =1
hypotenuse hypotenuse hypotenuse2

and similarly
hypotenuse 2 opposite 2 hypotenuse2 − opposite2
sec 2 𝜃 − tan2 𝜃 = ( ) −( ) = = 1,
adjacent adjacent adjacent 2

hypotenuse 2 adjacent 2 hypotenuse2 − adjacent 2

csc 2 𝜃 − cot 2 𝜃 = ( ) −( ) = = 1.
opposite opposite opposite2

The trigonometric functions are also called circular functions and this new perspective will allow us to extend
the angle 𝜃 beyond 90°. Consider a circle of radius 𝑟 around the origin, as in Figure 1(b). If a point is rotated
anticlockwise from a point on the positive 𝑥 axis through an angle 𝜃 along this circle, it reaches a point (𝑥; 𝑦).
If a line is drawn vertically from this point to the 𝑥 axis, a right angled triangle is formed with the angle 𝜃 at
the origin. Its base (adjacent to 𝜃) has length 𝑥 while the vertical side (opposite to 𝜃) has length 𝑦 and the
hypotenuse corresponds to the circle radius, with 𝑥 2 + 𝑦 2 = 𝑟 2 . Then

𝑦 𝑥 𝑦
sin 𝜃 = 𝑟 ; cos 𝜃 = 𝑟 ; tan 𝜃 = 𝑥 .

The horizontal displacement of the point is therefore 𝑥 = 𝑟 cos 𝜃 while the vertical displacement is
𝑦 = 𝑟 sin 𝜃. Whereas in a right angled triangle 𝜃 cannot exceed 90°, for the circle there is no limitation and
𝜃 can be any angle (even negative ones for clockwise instead of anticlockwise rotations). Figure 2 shows
graphs of the trigonometric functions. Because of the nature of a circle, the functions repeat after 360°. This
makes them useful for representing oscillations and waves mathematically.

Figure 2

Graphs of the
trigonometric functions.
Note that the x-axis is in
radians.
 4

The CAST diagram (Figure 3) is useful to remember which trigonometric ratios are positive and which are
negative in each quadrant. When using a calculator to solve problems, two factors are vital to consider:

Figure 3

S A CAST diagram.
Sin is positive, Cos and Tan are not
All are positive Cosine is positive in quadrant IV

All ratios are positive in quadrant I

Sine is positive in quadrant II

T C Tangent is positive in quadrant III
Tan is positive, Sin and Cos are not Cos is positive, Sin and Tan are not

 Calculators generally allow one to set the unit for angles between degrees, radians and gradians. A
degree is defined so that the angle around the centre of a full circle is 360°. An angle measured in radians
from the centre of a circle is defined as the ratio of the arc length along the circle to its radius. Because
the circumference of a full circle is 2𝜋𝑟, the angle of a full circle is 2𝜋 radians, explaining the well-known
conversion formula 𝜋 radians = 180°. (Gradians are seldom used, but are motivated by the decimal
𝜋
system: a right angle (90° or 2 rad) is defined as 100 grad, so 400 grad make a full circle.) We generally
choose to work in radians. There is an excellent reason for this: formulae for derivatives and integrals of
𝑑
trigonometric functions, like 𝑑𝑥 (sin 𝑥) = cos 𝑥 and ∫ sin 𝑥 𝑑𝑥 = − cos 𝑥 + 𝐶, are only valid for angles
measured in radians. Take care that your calculator is set to the correct unit. A radian is actually a
strange unit, because an angle in radians is by definition a length (arc length) divided by a length (radius),
which is unitless. One could describe radians as a ‘unitless unit’. If an angle has no unit it is assumed to
be in radians and we mainly write the unit rad to avoid the reader thinking that the angle may be in
different unit, such as degrees. Do not leave out the degree sign if you mean degrees!
 While sin, cos and tan are functions (with a unique output for each input), their inverses are not true
𝜋
functions and can have different output values for each input. For example, consider the angle 30° or 6
𝜋 1 1 𝜋
radians: then sin ( 6 ) = 2, but what is the value of arcsin (2)? One answer is, of course, 6 but there are
others. From the cast diagram one knows that the sine function is positive in quadrant II and so for some
1
angle 𝜃 in this quadrant sin 𝜃 = 2. Your calculator does not give this angle but you can figure it out for
𝜋 5𝜋
yourself: its value is 𝜋 − 6 = 6
rad or 180 – 30 = 150° as you can easily verify with your calculator.
Because the trigonometric functions repeat with rotations of the angle in a circle, to each of these
1 𝜋 5𝜋
answers can be added any integer multiple of 2𝜋 rad (i.e. 360°). So then arcsin (2) = 6 + 2𝜋𝑛 or 6
+
2𝜋𝑛 for any integer 𝑛, giving an infinite number of answers. You cannot always just use the angle given
by your calculator when calculating inverse trigonometric ratios. The answer your calculator gives is
just one of infinite possibilities and may not be the suitable answer for your question.

3𝜋 3𝜋
Example: (a) Calculate tan ( 4 ). (b) Then find all angles 𝜃 such that tan 𝜃 = tan ( 4 ).
 5

3𝜋
Solution: (a) If your calculator is set to degrees, we convert 4
rad to 135°, and then tan (135°) = -1.
3𝜋
Alternatively, when your calculator is set to radians, you can calculate directly that tan ( 4 ) = −1. Your
calculator should have a button for 𝜋, so don’t use approximate values like 3.14 or 22/7 or your answer will
not be accurate.

𝜋
(b) We must solve tan 𝜃 = −1. My calculator tells me 𝜃 = arctan(−1) = -45° (or − rad). To this I can add
4
7𝜋
360° to get the non-negative answer of 315° (or 4
rad), lying in quadrant IV. The CAST diagram reminds me
3𝜋
that the tangent function is also negative in quadrant II. The appropriate angle is 180 – 45 = 135° or 4
rad,
3𝜋
which is no surprise since in the first part of the problem we found that tan ( 4 ) = −1. The complete answer
3𝜋 7𝜋
is that tan 𝜃 = −1 for 𝜃 = 4
+ 2𝜋𝑛 or 4
+ 2𝜋𝑛 for any integer 𝑛.

1
Example: Find all angles 𝜃 such that cos 𝜃 = − 2.

1 2𝜋
Solution: Your calculator will quickly give you one answer: mine gives arccos (− 2) = 120° or 3
, which is in
quadrant II. From the CAST diagram we also know that there should also be an angle in quadrant III which
1
has cos 𝜃 = − 2. But how do we find this angle. If one is not careful, one can easily make a mistake. The
following procedure is recommended when calculating inverse trigonometric functions: (1) Ignore the
negative sign, if there is one. (2) Use you calculator to find an angle in the first quadrant – beware, this is not
yet the angle you are looking for: let us call it 𝛼. (3) Check with the CAST diagram in which quadrants you
expect the angles you are looking for to occur, and calculate them using just 𝛼 for quadrant I, or 𝜋 − 𝛼 for
quadrant II, or 𝜋 + 𝛼 for quadrant III or 2𝜋 − 𝛼 for quadrant IV.

1
To apply this method for cos 𝜃 = − , we ignore the minus sign and calculate in the first quadrant 𝛼 =
2
1 𝜋
arccos (2) = 3 on our calculator. Cosine is negative in quadrants II and III, so the angles we seek are 𝜃 = 𝜋 −
𝜋 2𝜋 𝜋 4𝜋
= and 𝜃 = 𝜋 + = . Of course, to either of these we may add 2𝜋𝑛 for any integer 𝑛.
3 3 3 3

There are many trigonometric identities which you should be aware of and be able to apply. We emphasize
the sine and cosine functions as these are most important when studying oscillations and waves.

Addition and subtraction formulae Double and half angle formulae

sin(𝑢 + 𝑣) = sin 𝑢 cos 𝑣 + cos 𝑢 sin 𝑣 sin 2𝑢 = 2 sin 𝑢 cos 𝑢

cos 2𝑢 = cos 2 𝑢 − sin2 𝑢 = 1 − 2 sin2 𝑢 = 2 cos2 𝑢 − 1

cos(𝑢 + 𝑣) = cos 𝑢 cos 𝑣 − sin 𝑢 sin 𝑣
1
sin(𝑢 − 𝑣) = sin 𝑢 cos 𝑣 − cos 𝑢 sin 𝑣 sin 𝑣2 = √2(1 − cos 𝑣)

cos(𝑢 − 𝑣) = cos 𝑢 cos 𝑣 + sin 𝑢 sin 𝑣 1

cos 𝑣2 = √2(1 + cos 𝑣)
 6

Other useful formulae

1
sin 𝑢 sin 𝑣 = 2{cos(𝑢 − 𝑣) − cos(𝑢 + 𝑣)}

1
cos 𝑢 cos 𝑣 = 2{cos(𝑢 − 𝑣) + cos(𝑢 + 𝑣)}

1
sin 𝑢 cos 𝑣 = 2{sin(𝑢 − 𝑣) + sin(𝑢 + 𝑣)}

𝑢−𝑣
sin 𝑢 + sin 𝑣 = 2 sin (𝑢+𝑣
2
) cos (
2
)

𝑢+𝑣
sin 𝑢 − sin 𝑣 = 2 sin (𝑢−𝑣
2
) cos (
2
)

𝑢−𝑣
cos 𝑢 + cos 𝑣 = 2 cos (𝑢+𝑣
2
) cos (
2
)

𝑢−𝑣
cos 𝑢 − cos 𝑣 = 2 sin (𝑢+𝑣
2
) sin (
2
)

𝑁−1
sin(𝑁𝑣2) 𝑣
∑ sin(𝑢 + 𝑗𝑣) = sin (𝑢 + {𝑁 − 1}2)
sin(𝑣2)
𝑗=0

𝑁−1
sin(𝑁𝑣2) 𝑣
∑ cos(𝑢 + 𝑗𝑣) = cos (𝑢 + {𝑁 − 1}2)
sin(𝑣2)
𝑗=0

1.3. Manipulating graphs

Relationships between variables in physics are generally expressed as equations, but may also be visualized
using graphs. It is important to be able to interpret graphs correctly, and understand the connection of
mathematical equations and graphs.

1.3.1. Shifting graphs

If the variables 𝑥 and 𝑦 are related by the function 𝑦 = 𝑥 2 , this can be visualized as a graph of a parabola
with a minimum at the origin. But suppose that we change 𝑥 → 𝑥 + 1. The new function will be 𝑦 =
(𝑥 + 1)2 . This is also a parabola, similar to the first, but with its minimum at the point (-1; 0) as shown in the
left-most graph of Figure 4. Changing 𝑥 → 𝑥 + 1 therefore shifts the graph of the parabola one unit to the
left. We can generalize this result and state (assuming 𝑎 is a positive number) that:
 Changing 𝑥 → 𝑥 + 𝑎 transforms the graph of a function by shifting it 𝑎 units to the left.
 7

𝑦 = (𝑥 + 1)2 𝑦 = (𝑥 − 1)2 𝑦 + 1 = 𝑥2 𝑦 − 1 = 𝑥2

Figure 4. Shifting a parabola graph. The original parabola 𝒚 = 𝒙𝟐 is shown in each figure together with the shifted one.

It is then not difficult to also see the following:

 Changing 𝑥 → 𝑥 − 𝑎 transforms the graph of a function by shifting it 𝑎 units to the right.

 Changing 𝑦 → 𝑦 + 𝑎 transforms the graph of a function by shifting it 𝑎 units downwards.
 Changing 𝑦 → 𝑦 − 𝑎 transforms the graph of a function by shifting it 𝑎 units upwards.

Example: Show that the sine function is merely a shifted cosine function.

Solution: Consider the graphs of the cosine and sine functions.

y=cos(x)

y=sin(x)

  3 2
2 2

It is clear that if the cosine function is shifted to the right by 𝜋2 rad (i.e. 90°) then it will match the sine function
perfectly. Mathematically, we can express this as cos(𝑥 − 𝜋2) = sin 𝑥.

Alternatively, we can shift the cosine function to the left by an amount by 3𝜋

2
rad (i.e. 270°) and again it will
match the sine function perfectly. Mathematically, we can express this as cos(𝑥 + 3𝜋
2
) = sin 𝑥.

Example: Simplify the expression sin(𝑥 + 𝜋) by considering the shift of the sine graph.

Solution: Looking at the graph of the sine function in the previous example, one can see that shifting sin 𝑥 a
distance 𝜋 to the left would give the graph − sin 𝑥, so sin(𝑥 + 𝜋) = − sin 𝑥.

This can also be verified with the identity sin(𝑢 + 𝑣) = sin 𝑢 cos 𝑣 + cos 𝑢 sin 𝑣. Taking 𝑢 = 𝑥 and = 𝜋, one
gets sin(𝑥 + 𝜋) = sin 𝑥 cos 𝜋 + cos 𝑥 sin 𝜋 = (sin 𝑥)(−1) + (cos 𝑥)(0) = − sin 𝑥 as expected.
 8

1.3.2. Distorting graphs

If the variables 𝑥 and 𝑦 are related by 𝑥 2 + 𝑦 2 = 4, this can be visualized as a graph of a circle around the
origin (radius 2). But suppose that we change 𝑥 → 2𝑥. The new function will be (2𝑥)2 + 𝑦 2 = 4. This is no
longer a circle. If we set 𝑥 = 0 we can find the y-intercepts from 𝑦 2 = 4, so 𝑦 = ±2. These points correspond
to where the original circle cut the y-axis. To find the x-intercepts of (2𝑥)2 + 𝑦 2 = 4 we set 𝑦 = 0, so
(2𝑥)2 = 4 or 𝑥 = ±1. Changing 𝑥 → 2𝑥 distorts the original circle into an ellipse by compressing the graph
horizontally by a factor of 2, as shown in the left-most graph of Figure 5. We can generalize this result and
state (assuming 𝑎 is a positive number greater than one) that:

 Changing 𝑥 → 𝑎𝑥 transforms the graph of a function by compressing it horizontally by a factor 𝑎.

(2𝑥)2 + 𝑦 2 = 4 (𝑥/2)2 + 𝑦 2 = 4 𝑥 2 + (2𝑦)2 = 4 𝑥 2 + (𝑦/2)2 = 4

Figure 5. Distorting a circle. The original circle 𝒙𝟐 + 𝒚𝟐 = 𝟒 (radius 2) is shown in each figure with the distorted one.

It is then not difficult to also see the following:

 Changing 𝑥 → 𝑥/𝑎 transforms the graph of a function by stretching it horizontally by a factor 𝑎.

 Changing 𝑦 → 𝑎𝑦 transforms the graph of a function by compressing it vertically by a factor 𝑎.
 Changing 𝑦 → 𝑦/𝑎 transforms the graph of a function by stretching it vertically by a factor 𝑎.

Example: A picture of a circle with a radius of one unit is being copied using a photocopier. However, there
is a problem with the paper feeder and the page being printed moves at only 80% of the speed it should, so
that the circle is distorted into an ellipse. Give a formula for the ellipse, assuming the paper feeds along the
𝑦-direction. Assume the centre remains at the origin.

Solution: The original circle has formula 𝑥 2 + 𝑦 2 = 1. Due to the feeder problems, the circle is compressed
in the 𝑦-direction to form an ellipse. Since the printed page moves at only 80% of its usual speed, the
compression factor is 100/80 = 1.25 times. Thus the ellipse has equation 𝑥 2 + (1.25𝑦)2 = 1.

Example: Bob and Joe stand at the same place 30 m from a train line, and watch a train come past. Both
independently measure the distance to the front of the passing train as a function of time. Bob finds the
distance is given by 𝐷𝐵 = √900 + 𝑡 2 , while Joe finds that the distance is given by 𝐷𝐽 = √1000 − 20𝑡 + 𝑡 2 .
Although the results apparently disagree, show they are in fact the same except that Bob and Joe started
their stopwatches at different times. What was the time difference, and who started their watch first?
 9

Solution: Bob’s formula only has a 𝑡 2 term, while Joe's formula has both a 𝑡 2 term and a 𝑡 term. But the
distance measured by Joe can be rewritten as

𝐷𝐽 = √1000 − 20𝑡 + 𝑡 2 = √900 + (100 − 20𝑡 + 𝑡 2 ) = √900 + (𝑡 − 10)2 .

This now looks the same as that for Bob (𝐷𝐵 = √900 + 𝑡 2 ) except 𝑡 is replaced with 𝑡 − 10. When plotted,
the data will be identical but Joe’s data will be shifted 10 time units (seconds) later than Bob’s, showing he
started his stopwatch 10 s earlier.

1.3.3. Periodicity

A cosine function is simply a shifted sine function, so we classify both sine and cosine functions under the
general name of sinusoidal functions. All stretched and shifted versions of sine and cosine functions are still
called sinusoidal functions.

An important characteristic of sinusoidal functions is that they are periodic i.e. they repeat themselves. For
example, both sin 𝑥 and cos 𝑥 have a period 𝑇 = 2𝜋 rad, which is as a result of these functions being defined
using a point moving in a circle (Figure 1(b)).

Suppose some function is periodic and repeats itself. Then, imagine looking at its graph and lifting a copy of
the graph up from the original. If this copy is shifted horizontally to the right by exactly the period of the
graph then the original graph and the copy will coincide perfectly. Mathematically, one can say that a function
𝑓[𝑥] is periodic with period 𝑇 if 𝑓[𝑥 − 𝑇] = 𝑓[𝑥].

Note the following:

 There is no particular reason to shift the copy of the function to the left or right – either will do. If
the copy is shifted to the left the condition for a function to be periodic is 𝑓[𝑥 + 𝑇] = 𝑓[𝑥].
 Suppose a function has a period 𝑇, but we shift the copy a distance 2𝑇 or 3𝑇 to the right. Since the
function repeats every distance 𝑇, the copy shifted a distance 2𝑇 or 3𝑇 will also coincide perfectly
with the original. In general, if a function has a period 𝑇 it can also be said to be periodic with any
value 𝑛𝑇 where 𝑛 = 2, 3, 4 … When we talk of the period of a function, the smallest period is meant.

Example: What is the period of the following functions, if 𝑏 and 𝑐 are constants?

 cos(𝑥 + 𝑏): This is just a cosine functions shifted to the left by a distance 𝑏. The horizontal shift does not
change the repeat distance, so the period is the same as for cos 𝑥, namely 2𝜋 rad.

 𝑏 sin(𝑐𝑥): This is a sine function compressed horizontally by a factor 𝑐, and stretched vertically by a factor
𝑏. The vertical stretch does not affect the period, while the horizontal compression reduces the period
from 2𝜋 rad for sin 𝑥 to 2𝜋/𝑐 rad for the given function.

 cos(𝑥/𝑏 + 𝑐): This is a cosine functions stretched horizontally by a factor 𝑏, and the result shifted to the
left by a distance 𝑐. The horizontal shift does not affect the period, but the horizontal stretch increases
it from 2𝜋 rad for cos 𝑥, to 2𝜋𝑏 rad for the given function.
 10

Example: Consider the graph of the function tan 𝑥, and hence find its period.

Solution: From the graph, one sees that the period is 𝜋 rad.

1.3.4. Reflections and Parity

Suppose we have some function 𝑦 = 𝑓[𝑥]. What would happen to the function if we change 𝑥 → −𝑥? The
point on the graph for 𝑥 = 0 (i.e. the 𝑦 intercept) would be unaffected, but every other point would move
from one side of the 𝑦 axis to a corresponding point equally far on the other side of the 𝑦 axis.

Figure 6

Reflection through the y axis

when changing 𝑥 → −𝑥 for
the exponential function.

The new function with 𝑥 → −𝑥 would be a mirror image of the original function, reflected through the 𝑦 axis,
as illustrated for the exponential function in Figure 6.

Changing 𝑦 → −𝑦 would create a new graph which is a mirror image of the original function reflected
through the 𝑥 axis.

Most functions, when reflected, look different to the original function. That is certainly the case for the
exponential function in Figure 6. However, some functions have special reflection properties:

 If a function 𝑦 = 𝑓[𝑥] is such that its reflection in the 𝑦 axis (when 𝑥 → −𝑥) is the same as itself, we call
it an even function, or say that the function has even parity. Such a function satisfies

𝑓[−𝑥] = 𝑓[𝑥].
 11

 If a function 𝑦 = 𝑓[𝑥] is such that projecting each point through the origin, the same distance that the
point lies away from the origin, leaves the function unchanged then we call it an odd function, or say that
the function has odd parity. The origin is then an inversion pointy or centre of inversion. Projecting a
point through the origin is the same as reflecting it first in the 𝑦 axis (when 𝑥 → −𝑥) and then in the 𝑥
axis (when 𝑦 → −𝑦), so an odd function satisfies
−𝑓[−𝑥] = 𝑓[𝑥].
An alternative way of thinking about odd parity, which is often easier to use, is the equivalent
mathematical condition 𝑓[−𝑥] = −𝑓[𝑥], meaning a function 𝑓[𝑥] has odd parity if reflecting it in the 𝑦
axis produces the same result as reflecting it in the 𝑥 axis (although neither of these reflections are the
same as the original function).

an even function
Figure 7

Examples of functions

an odd function with even and odd parity.

Figure 7 shows graphs of functions having even and odd parity. Most functions do not have the special
symmetry properties necessary to make them even or odd, for example the exponential function of Figure
6. However, we can give some specific examples of even and odd functions. Consider the graph of the cosine
function (draw one if you do not have a picture handy). Looking at the graph of the cosine function, it is clear
it is a reflection of itself in the 𝑦 axis, so it is an even function. On the other hand, the sine function is an odd
function (draw it and convince yourself of this). These graphical observations can also be confirmed
mathematically:

 From trigonometry we have the identity cos(𝑢 − 𝑣) = cos 𝑢 cos 𝑣 + sin 𝑢 sin 𝑣. If we take 𝑢 = 0 and
𝑣 = 𝑥 , this gives cos(0 − 𝑥) = cos 0 cos 𝑥 + sin 0 sin 𝑥, so that cos(−𝑥) = cos 𝑥. Changing 𝑥 → −𝑥
leaves the cosine function unchanged, and it is an even function.

 We also have from trigonometry that sin(𝑢 − 𝑣) = sin 𝑢 cos 𝑣 − cos 𝑢 sin 𝑣. If we take 𝑢 = 0 and 𝑣 =
𝑥, this gives sin(0 − 𝑥) = sin 0 cos 𝑥 − cos 0 sin 𝑥, so that sin(−𝑥) = − sin 𝑥 which is the mathematical
condition for an odd function.

One may wonder why the parity of a function is called even or odd, and whether this is related to odd and
even numbers. The answer comes from the family of functions 𝑦 = 𝑥 𝑛 where 𝑛 is an integer. Specific
examples are the quadratic function (i.e. parabola) 𝑦 = 𝑥 2 and cubic function 𝑦 = 𝑥 3 shown in Figure 8.
 12

Figure 8

Quadratic and cubic

functions.

Looking at these functions and considering what their reflections would look like, you should easily convince
yourself that the quadratic function is even, while the cubic function is odd. We can easily prove
mathematically the general case that for 𝑓[𝑥] = 𝑥 𝑛 the function is even when 𝑛 is an even number and odd
when 𝑛 is an odd number. Consider 𝑓[−𝑥] = (−𝑥)𝑛 = (−1)𝑛 𝑥 𝑛 : when 𝑛 is an even number then (−1)𝑛 =
1 and 𝑓[−𝑥] = 𝑥 𝑛 = 𝑓[𝑥], so the function is even. But if 𝑛 is an odd number then (−1)𝑛 = −1 and 𝑓[−𝑥] =
−𝑥 𝑛 = −𝑓[𝑥], so the function is odd. So although the names of even and odd make some sense, one much
be careful since the properties of even and odd functions are often different what we are used to for even
and odd numbers.

Example: Two odd numbers added together give an even number e.g. 1 + 5 = 6. However, the sum of two
odd functions is a new odd function. As an example, consider the odd functions 𝑓[𝑥] = 𝑥 and 𝑔[𝑥] = 𝑥 5 .
Then let the sum of these functions be ℎ[𝑥] = 𝑥 + 𝑥 5 . Now ℎ[𝑥] is an odd function because ℎ[−𝑥] =
(−𝑥) + (−𝑥)5 = −𝑥 − 𝑥 5 = −(𝑥 + 𝑥 5 ) = −ℎ[𝑥].

While the product of an odd and an even number is an even number, the product of an odd and even function
is a new odd function. Consider ℎ[𝑥] above: we can write it as ℎ[𝑥] = 𝑥(1 + 𝑥 4 ) = 𝑓[𝑥]𝑚[𝑥] where 𝑚[𝑥] =
1 + 𝑥 4 . It is not hard to verify that 𝑚[𝑥] is an even function, so the odd function ℎ[𝑥] is the product of an
odd function 𝑓[𝑥] and an even function 𝑚[𝑥].

Example: Show that the function 𝑓[𝑥] = − sin(𝑥 2 ) is even.

Solution: We calculate 𝑓[−𝑥] = − sin({−𝑥}2 ) = − sin(𝑥 2 ) = 𝑓[𝑥]. Therefore the function is even.

Example: Show that the function 𝑓[𝑥] = tan 𝑥 is odd.

sin(−𝑥) −sin(𝑥)
Solution: We calculate 𝑓[−𝑥] = tan(−𝑥) = cos(−𝑥) = cos(𝑥)
= −tan(𝑥) = −𝑓[𝑥]. Therefore the function is
odd. The fact that the tangent function is odd should also be clear from its graph.
 13

Note the following useful properties of odd and even functions:

 The parity of a function is not changed is it is stretched or compressed either horizontally or vertically.
However, shifting the function can affect its parity.
 The product of two functions with the same parity (even and even or odd and odd) produces an even
function, whereas the product of two functions with different parity produces an odd function.
 If a function has a certain parity, then taking its derivative changes its parity.
 The integral of an even function over the region [−𝑎; 𝑎] is twice the integral over the positive region
[0; 𝑎], while integral of an odd function over the region [−𝑎; 𝑎] is always zero. These properties become
obvious when one considers Figure 7 and recalls that an integral gives the area under a curve.

1.4. Taylor series

The key idea for Taylor series is that any function can be approximated by a polynomial.

Taylor series polynomials for some functions

𝑥2 𝑥3 𝑥4
𝑒𝑥 = 1 + 𝑥 + 2!
+ 3!
+ 4!
+⋯

𝑥2 𝑥3 𝑥4
ln(1 + 𝑥) = 𝑥 − 2
+ 3
− 4
+⋯

𝑥3 𝑥5 𝑥7
sin 𝑥 = 𝑥 − 3!
+ 5!
− 7!
+⋯

𝑥2 𝑥4 𝑥6
cos 𝑥 = 1 − 2!
+ 4!
− 6!
+⋯

𝑥3 2𝑥 5 17𝑥 7
tan 𝑥 = 𝑥 + 3
+ 15
+ 315
+⋯

How does one find such Taylor series? Consider an 𝑛th degree polynomial

𝑝𝑛 = 𝑐0 + 𝑐1 𝑥 + 𝑐2 𝑥 2 + 𝑐3 𝑥 3 + ⋯ + 𝑐𝑛 𝑥 𝑛 .

It has 𝑛 + 1 constants, so we need to specify 𝑛 + 1 conditions to find these constants. For a Taylor series,
these conditions are chosen to be that the value of the polynomial and its first 𝑛 derivatives, when computed
at 𝑥 = 0, all match the corresponding values of the function being approximated.
𝑑𝑝𝑛
The first derivative of the polynomial is 𝑑𝑥
= 𝑐1 + 2𝑐2 𝑥 + 3𝑐3 𝑥 2 + ⋯ + 𝑛𝑐𝑛 𝑥 𝑛−1 , which has a value 𝑐1
𝑑 2 𝑝𝑛
when 𝑥 = 0. The second derivative is 𝑑𝑥 2
= 2𝑐2 + (3)(2)𝑐3 𝑥 + ⋯ + (𝑛)(𝑛 − 1)𝑐𝑛 𝑥 𝑛−2 , which has a value
2𝑐2 when 𝑥 = 0. In the general case the 𝑘 th derivative of the polynomial, computed for 𝑥 = 0, has a value
 14

𝑘! 𝑐𝑘 where 𝑘! = 𝑘(𝑘 − 1)(𝑘 − 2) … (2)(1) is the factorial function. Setting this equal to the corresponding
value for the function 𝑓 one wishes to approximate gives

𝑑𝑘 𝑓 1 𝑑𝑘 𝑓
𝑘! 𝑐𝑘 = | or 𝑐𝑘 = | ,
𝑑𝑥 𝑘 𝑥=0 𝑘! 𝑑𝑥 𝑘 𝑥=0

which is the formula for obtaining the Taylor polynomial constants.

Example: Find the quadratic Taylor series polynomial for approximating the function 𝑓[𝑥] = (1 + 𝑥)𝑏 where
𝑏 is a constant.

Solution: We wish to approximate the function by a polynomial 𝑝 = 𝑐0 + 𝑐1 𝑥 + 𝑐2 𝑥 2 , by matching their

values and the values of their derivatives when 𝑥 = 0.

 When 𝑥 = 0 the function 𝑓[𝑥] has a value (1 + 0)𝑏 = 1, while the polynomial 𝑝 with 𝑥 = 0 has a value
𝑐0 . For them to match we must choose 𝑐0 = 1.
 The derivative of the function is 𝑏(1 + 𝑥)𝑏−1 which has a value of 𝑏 when 𝑥 = 0. The first derivative of
the polynomial 𝑝 is 𝑐1 + 2𝑐2 𝑥, which has a value 𝑐1 when 𝑥 = 0. For a match we require 𝑐0 = 𝑏.
 The second derivative of the function is 𝑏(𝑏 − 1)(1 + 𝑥)𝑏−2 which has a value of 𝑏(𝑏 − 1) when 𝑥 = 0.
The second derivative of the polynomial 𝑝 is 2𝑐2 . For a match we require 𝑐2 = 𝑏(𝑏 − 1)/2.

The quadratic Taylor series polynomial approximating the function 𝑓(𝑥) = (1 + 𝑥)𝑏 is therefore 𝑝 = 1 +
𝑏𝑥 + 𝑏(𝑏−1)
2
𝑥 2 . We can write this as

𝑏(𝑏−1) 2
(1 + 𝑥)𝑏 ≈ 1 + 𝑏𝑥 + 𝑥 .
2

If 𝑏 = 1 or 𝑏 = 2 the match is perfect, while other values of 𝑏 give useful approximations, for example:

If 𝑏 = 3: (1 + 𝑥)3 ≈ 1 + 3𝑥 + 3𝑥 2
1 1 1
If 𝑏 = 2: √1 + 𝑥 ≈ 1 + 2𝑥 − 8𝑥 2
1
If 𝑏 = −1: ≈ 1 − 𝑥 + 𝑥2
1+𝑥

2.5
Example: Use the Taylor series of the exponential y = ex

function to obtain linear and quadratic approximations. 2

y = 1+ x
Solution: The Taylor series of the exponential function is 1.5
2 3 4
𝑒 𝑥 = 1 + 𝑥 + 𝑥2! + 𝑥3! + 𝑥4! + ⋯. A linear approximation is y = 1+ x +
x2
1 2
therefore 𝑒 𝑥 ≈ 1 + 𝑥, while a quadratic approximation is
𝑒 𝑥 ≈ 1 + 𝑥 + 12𝑥 2 . 0.5

0
These are good approximations near 𝑥 = 0, but less good
further away. -0.5

-1
-2 -1.5 -1 -0.5 0 0.5 1 1.5 2
 15

3 5 7
The Taylor series for the sine function is sin 𝑥 = 𝑥 − 𝑥3! + 𝑥5! − 𝑥7! + ⋯. In many cases, as long as 𝑥 is small
enough, the approximation sin 𝑥 ≈ 𝑥 is adequate. Check this on your calculator, noting that it is only true
1
for 𝒙 in radians. If greater accuracy is required one may use sin 𝑥 ≈ 𝑥 − 6 𝑥 3 . These approximations are
shown in the top part of Figure 9.

2
y= x
1.5

1 y = sinx

0.5

-0.5
x3
y= x -
-1 6

-1.5 Figure 9
-2
-4 -3 -2 -1 0 1 2 3 4

Taylor series
2
approximations for sine
1.5 2 4
x x and cosine functions.
y= 1 - +
1 2 24

0.5

-0.5

-1 x2
y= 1-
2
-1.5 y = cosx

-2
-4 -3 -2 -1 0 1 2 3 4

2 4 6
The Taylor series for the cosine function is cos 𝑥 = 1 − 𝑥2! + 𝑥4! − 𝑥6! + ⋯. For small 𝑥 the approximation
cos 𝑥 ≈ 1 is often adequate. This is more accurate that one might initially suspect, because there is no linear
term in the Taylor expansion of cos 𝑥. If greater accuracy is required one may use the parabola cos 𝑥 ≈ 1 −
1 2 1 1
2
𝑥 , or for more accuracy still cos 𝑥 ≈ 1 − 2 𝑥 2 + 24 𝑥 4 . These approximations are shown in the bottom part
of Figure 9.

The small angle approximations 𝐬𝐢𝐧 𝒙 ≈ 𝒙, 𝐜𝐨𝐬 𝒙 ≈ 𝟏 and 𝐭𝐚𝐧 𝒙 ≈ 𝒙 are often used in physics, together
with the useful result (𝟏 ± 𝒙)𝒏 ≈ 𝟏 ± 𝒏𝒙 for small 𝒙. Note also that 𝐬𝐢𝐧 𝒙 ≈ 𝐭𝐚𝐧 𝒙 for small 𝒙.

Notice how the Taylor expansion for sin 𝑥 only consists of 𝑥 𝑛 terms where 𝑛 is odd, while the Taylor
expansion for cos 𝑥 only consists of 𝑥 𝑛 terms where 𝑛 is even. It makes sense that an odd function should be
approximated by a polynomial with odd parity terms while an even function can be expressed as a polynomial
with only even terms. See that the Taylor series for the tan function only has odd terms, and recall that we
 16

earlier showed that the tan function has odd parity. This idea leads to something interesting, namely that
any function can be expressed as the sum of an even function and an odd function. To motivate this, consider
the following: any function can be written as its Taylor series, which in general will have 𝑥 𝑛 terms where 𝑛 is
both even and odd. But we can group all the terms with 𝑛 even together (making an even function) and also
group all the remaining terms with 𝑛 odd together (making an odd function). The original function is the sum
of the two groups, or the sum of an even and odd function.

Example: Consider the exponential function 𝑒 𝑥 , which is neither even nor odd. Its Taylor series is

𝑥2 𝑥3 𝑥4 𝑥5
𝑒𝑥 = 1 + 𝑥 + 2!
+ 3!
+ 4!
+ 5!
+⋯

2 4 3 5
We can group the even and odd terms, writing 𝑒 𝑥 = (1 + 𝑥2! + 𝑥4! + ⋯ ) + (𝑥 + 𝑥3! + 𝑥5! + ⋯ ). The first
bracket defines a new even function which is called the “cosh” function (or hyperbolic cosine function), while
the second bracket defines a new odd function called the “sinh” function:

𝑥2 𝑥4
cosh 𝑥 = 1 + 2!
+ 4!
+⋯

𝑥3 𝑥5
sinh 𝑥 = 𝑥 + 3!
+ 5!
+⋯

These functions should be available on your calculator. We have therefore shown that 𝑒 𝑥 = cosh 𝑥 + sinh 𝑥,
and, while doing so, we have expressed a function having no parity as the sum of even and odd functions.

An alternative way of obtaining the hyperbolic functions is now easily obtained: since 𝑒 𝑥 = cosh 𝑥 + sinh 𝑥,
it follows that 𝑒 −𝑥 = cosh(−𝑥) + sinh(−𝑥) = cosh 𝑥 − sinh 𝑥, the last step coming from their parity. This
means 𝑒 𝑥 + 𝑒 −𝑥 = (cosh 𝑥 + sinh 𝑥) + (cosh 𝑥 − sinh 𝑥) = 2 cosh 𝑥, so

𝑒 𝑥 + 𝑒 −𝑥
cosh 𝑥 = .
2
𝑒 𝑥 −𝑒 −𝑥
The interested reader should also be able to show in a similar way that sinh 𝑥 = 2
. Other hyperbolic
sinh 𝑥
functions can be defined analogously to trigonometric functions, e.g. tanh 𝑥 = cosh 𝑥 , which is an odd
𝑑 𝑑
function. Many relationships are similar, but with differences in sign e.g. 𝑑𝑥
sinh 𝑥 = cosh 𝑥 and 𝑑𝑥 cosh 𝑥 =
sinh 𝑥, without a negative sign as one can prove from their Taylor series. In addition cosh2 𝑥 − sinh2 𝑥 = 1
(note the negative sign). Hyperbolic functions can always just be rewritten using exponential functions, but
are sometimes encountered in physics problems and can be more useful because of their parity.

1.5. Complex numbers and the complex exponential function

1.5.1. Complex numbers and the complex conjugate

2
We represent √−1 by the letter 𝑖, so that 𝑖 2 = (√−1) = −1. Also 𝑖 3 = (𝑖 2 )𝑖 = −𝑖 and 𝑖 4 = (𝑖 2 )(𝑖 2 ) = 1.
 17

One can then write a complex number as 𝐴̃ = 𝐴𝑅 + 𝑖𝐴𝐼 . It can be represented by the point (𝐴𝑅 ; 𝐴𝐼 ) on the
complex (or Argand) plane, where the value 𝐴𝑅 is taken for the x- (or real) coordinate and the value of 𝐴𝐼 is
taken as the y- (or imaginary) coordinate. Both 𝐴𝑅 and 𝐴𝐼 are real numbers.

The complex conjugate of a complex number just has the imaginary part negated (sign reversed) and is
represented by an asterisk, i.e. 𝐴̃∗ = 𝐴𝑅 − 𝑖𝐴𝐼 . For example, the complex conjugate of 2 + 4𝑖 is 2 − 4𝑖,
while (−3 − 4𝑖)∗ = −3 + 4𝑖. It is interesting to note that when we add any complex number to its own
complex conjugate the result is always a real number, because 𝐴̃ + 𝐴̃∗ = (𝐴𝑅 + 𝑖𝐴𝐼 ) + (𝐴𝑅 − 𝑖𝐴𝐼 ) = 2𝐴𝑅 .

1.5.2. The complex exponential function

2 3 4 5
Recall the Taylor series of the exponential function, 𝑒 𝑥 = 1 + 𝑥 + 𝑥2! + 𝑥3! + 𝑥4! + 𝑥5! + ⋯. What would happen
if, instead of the real number 𝑥, one used an imaginary number 𝑖𝑢 (where 𝑢 alone is just an ordinary real
number)? If one believes the Taylor series can also be applied for complex numbers, then

(𝑖𝑢)2 (𝑖𝑢)3 (𝑖𝑢)4 (𝑖𝑢)5

𝑒 𝑖𝑢 = 1 + (𝑖𝑢) + 2!
+ 3!
+ 4!
+ 5!
+ ⋯.

In fact, one can use this as a definition of the exponential function extended to complex numbers. Now

𝑖 2 𝑢2 𝑖 3 𝑢3 𝑖 4 𝑢4 𝑖 5 𝑢5 𝑢2 𝑖𝑢3 𝑢4 𝑖𝑢5
𝑒 𝑖𝑢 = 1 + 𝑖𝑢 + 2!
+ 3!
+ 4!
+ 5!
+ ⋯ = 1 + 𝑖𝑢 − 2!
− 3!
+ 4!
+ 5!
+⋯.

We can split this into its real and imaginary parts as follows:

𝑢2 𝑢4 𝑢3 𝑢5
𝑒 𝑖𝑢 = {1 − 2!
+ 4!
− ⋯ } + 𝑖 {𝑢 − 3!
+ 5!
− ⋯} .

But the terms in brackets are exactly the Taylor series for cos 𝑢 and sin 𝑢, so we get a fantastic result:

𝑒 𝑖𝑢 = cos 𝑢 + 𝑖 sin 𝑢.

Therefore the complex exponential function is related to the trigonometric functions, which are central to
the study of oscillations and waves in physics. Although the size of physical quantities such as position,
velocity, etc. must be real numbers, we often use complex numbers to represent them. This is more abstract
but often leads to more compact, brief and elegant mathematics. For instance, one significant simplification
is that taking derivatives and integrals of exponential functions is easier than for any other functions, e.g. the
trigonometric functions.
∗
From 𝑒 𝑖𝑢 = cos 𝑢 + 𝑖 sin 𝑢 , it follows that (𝑒 𝑖𝑢 ) = 𝑒 −𝑖𝑢 = cos(−𝑢) + 𝑖 sin(−𝑢) = cos 𝑢 − 𝑖 sin 𝑢 . So
𝑒 𝑖𝑢 +𝑒 −𝑖𝑢
𝑒 𝑖𝑢 + 𝑒 −𝑖𝑢 = (cos 𝑢 + 𝑖 sin 𝑢) + (cos 𝑢 − 𝑖 sin 𝑢) = 2 cos 𝑢, so cos 𝑢 = 2
. The reader should also be
𝑒 𝑖𝑢 −𝑒 −𝑖𝑢
able to show (by subtracting rather than adding) that sin 𝑢 = 2𝑖
. Theoretically, the mathematician
could eliminate trigonometric functions and work only with (complex) exponential functions!

𝑒 𝑖𝑢 −𝑒 −𝑖𝑢 𝑒 𝑖𝑢 +𝑒 −𝑖𝑢
Example: Show, using the complex exponential forms sin 𝑢 = 2𝑖
and cos 𝑢 = 2
, that
2 2
sin 𝜃 + cos 𝜃 = 1.

2 2
𝑒𝑖𝜃 −𝑒−𝑖𝜃 𝑒𝑖𝜃 +𝑒−𝑖𝜃 𝑒𝑖2𝜃 −2+𝑒−𝑖2𝜃 𝑒𝑖2𝜃 +2+𝑒−𝑖2𝜃
Solution: Consider sin2 𝜃 + cos 2 𝜃 = ( 2𝑖
) +( 2
) = −4
+ 4
.

−𝑒𝑖2𝜃 + 2 −𝑒−𝑖2𝜃 + 𝑒𝑖2𝜃 + 2 + 𝑒−𝑖2𝜃 4

So sin2 𝜃 + cos 2 𝜃 = 4
= 4 = 1.
 18

1.5.3. Polar (vector) form of complex numbers

We have seen that a complex number 𝐴̃ = 𝐴𝑅 + 𝑖𝐴𝐼 can be represented by the point (𝐴𝑅 ; 𝐴𝐼 ) on the
complex plane. A second way which can be used to represent the same complex number 𝐴̃ = 𝐴𝑅 + 𝑖𝐴𝐼 is as
a vector from the origin to this point, as shown in Figure 10.

Figure 11

Representations of
complex number.

Recall that a vector has both magnitude and direction. From the theorem of Pythagoras, using Figure 12, the
magnitude is √𝐴2𝑅 + 𝐴2𝐼 which we write as |𝐴̃|. If the vector makes and angle 𝜃 with the horizontal (real)
𝐴
axis, then 𝐴𝑅 = |𝐴̃| cos 𝜃 and 𝐴𝐼 = |𝐴̃| sin 𝜃, so dividing these gives tan 𝜃 = 𝐴 𝐼 . Now consider:
𝑅

𝐴̃ = 𝐴𝑅 + 𝑖𝐴𝐼 = |𝐴̃| cos 𝜃 + 𝑖|𝐴̃| sin 𝜃 = |𝐴̃|{cos 𝜃 + 𝑖 sin 𝜃} = |𝐴̃|𝑒 𝑖𝜃 .

Hence a complex number can be expressed in terms of its size |𝐴̃| and its direction 𝜃 in the useful polar
(vector, or complex exponential) form 𝐴̃ = |𝐴̃|𝑒 𝑖𝜃 .

Example: Earlier we showed that the sum of a complex number and its complex conjugate is twice the real
part, i.e. 𝐴̃ + 𝐴̃∗ = (𝐴𝑅 + 𝑖𝐴𝐼 ) + (𝐴𝑅 − 𝑖𝐴𝐼 ) = 2𝐴𝑅 . Show this again, but now using the complex exponential
form of a complex number.

Solution: Consider the complex number 𝐴̃ represented by |𝐴̃|𝑒 𝑖𝜃 . For its vector representation |𝐴̃| is its size
and 𝜃 gives its direction. Its complex conjugate is 𝐴̃∗ = |𝐴̃∗ |𝑒 −𝑖𝜃 . But |𝐴̃∗ | = √𝐴2𝑅 + (−𝐴𝐼 )2 = √𝐴2𝑅 + 𝐴2𝐼 =
|𝐴̃|, meaning that a complex number and its complex conjugate always have the same size.

Now |𝐴̃|𝑒 𝑖𝜃 + |𝐴̃∗ |𝑒 −𝑖𝜃 = |𝐴̃|(𝑒 𝑖𝜃 + 𝑒 −𝑖𝜃 ) = |𝐴̃|(cos 𝜃 + 𝑖 sin 𝜃 + cos 𝜃 − 𝑖 sin 𝜃) = 2|𝐴̃| cos 𝜃 , which is
2𝐴𝑅 , twice the real part of 𝐴̃, hence we have the same result as found before.

Example: The previous example showed that the sum of a complex number and its complex conjugate is a
real number. Remarkably, the product of a complex number and its complex conjugate is also a real number!

Show that |𝐴̃| = √𝐴̃(𝐴̃∗ ). NOTE: this is not the same as |𝐴̃| = √𝐴̃2 , which gives the ‘absolute value’ of real numbers.

2
Solution 1: 𝐴̃(𝐴̃∗ ) = (𝐴𝑅 + 𝑖𝐴𝐼 )(𝐴𝑅 − 𝑖𝐴𝐼 ) = 𝐴2𝑅 + 𝐴2𝐼 = |𝐴̃| . Taking the square root gives the result.

2
Solution 2: 𝐴̃(𝐴̃∗ ) = (|𝐴̃|𝑒 𝑖𝜃 )(|𝐴̃∗ |𝑒 −𝑖𝜃 ), but |𝐴̃∗ | = |𝐴̃|, so 𝐴̃(𝐴̃∗ ) = |𝐴̃| , which gives the required result.
 19

1.6. Partial derivatives

Recall that an ordinary derivative of a function is defined by

𝑓[𝑥 + ℎ] − 𝑓[𝑥]
𝑓 ′ [𝑥] = lim .
ℎ→0 ℎ
𝑑𝑓
It represents the rate of change of the function, or its slope. This is why the alternative notation 𝑑𝑥 for the
derivative is so useful. We can find the maxima or minima of the function when the slope (derivative) is zero.

The reader should already be able to take ordinary derivatives, for example:

𝑑 𝑑 sin 𝑥 𝑑
(tan 𝑥) = ( )= ({sin 𝑥}{cos 𝑥}−1 )
𝑑𝑥 𝑑𝑥 cos 𝑥 𝑑𝑥

𝑑
∴ (tan 𝑥) = {sin 𝑥}((−1){cos 𝑥}−2 (− sin 𝑥)) + {cos 𝑥}−1 (cos 𝑥) = tan2 𝑥 + 1 = sec 2 𝑥
𝑑𝑥

This will be sufficient for the study of oscillations. However, for the study of waves one requires partial
derivatives.

Partial derivatives are defined for a functions of more than one variable e.g. suppose the function 𝑓 depends
on both 𝑥 and 𝑦, which we write as 𝑓[𝑥, 𝑦]. Then it is possible to consider the slope of the function as we
𝑓[𝑥+ℎ,𝑦]−𝑓[𝑥,𝑦]
change 𝑥 (but leave 𝑦 unchanged) which is lim ℎ
, but also the slope of the function as we change
ℎ→0
𝑓[𝑥,𝑦+ℎ]−𝑓[𝑥,𝑦]
𝑦 (and leave 𝑥 unchanged) which is lim ℎ
. This concept is illustrated in Figure 13.
ℎ→0

Figure 13

Partial derivatives of a
function of two
variables.
 20

We therefore have two types of partial derivatives for a function of two variables. The notation 𝑓′(𝑥, 𝑦) is
not useful since one cannot tell which variable is being changed. Rather we use the alternative notation but
with 𝜕 instead of 𝑑 to indicate a partial derivative. Therefore

𝜕𝑓 𝑓[𝑥 + ℎ, 𝑦] − 𝑓[𝑥, 𝑦]
= lim
𝜕𝑥 ℎ→0 ℎ

𝜕𝑓 𝑓[𝑥, 𝑦 + ℎ] − 𝑓[𝑥, 𝑦]
= lim
𝜕𝑦 ℎ→0 ℎ

We do not have to calculate partial derivatives from first principles using these formulae. In fact, we calculate
them just like we do ordinary derivatives, with the following important consideration:

𝜕𝑓
 when we calculate 𝜕𝑥 we regard 𝑦 as a constant, i.e. we treat it as a constant when differentiating
𝜕𝑓
 when we calculate 𝜕𝑦 we regard 𝑥 as a constant

Partial derivatives are calculated mathematically using the same rules as ordinary derivatives, except that
all variables except besides the one being differentiated are treated as constants.

Some examples should make this clear.

Example: Suppose 𝑓[𝑥, 𝑦] = 𝑥 2 𝑦 3 + 𝑥 5 cos 𝑦. Find all its possible partial derivatives up to the second order.

𝜕𝑓
Solution: The first order derivatives are = 2𝑥𝑦 3 + 5𝑥 4 cos 𝑦 (where we treated 𝑦 as if it were a constant)
𝜕𝑥
𝜕𝑓
and 𝜕𝑦 = 3𝑥 2 𝑦 2 − 𝑥 5 sin 𝑦 (where we treated 𝑥 as if it were a constant).

𝜕𝑓
 Taking 𝜕𝑥 = 2𝑥𝑦 3 + 5𝑥 4 cos 𝑦, we can differentiate again, either with respect to 𝑥 or with respect to 𝑦.
𝜕 𝜕𝑓 𝜕2 𝑓 𝜕 𝜕𝑓 𝜕2 𝑓
One gets 𝜕𝑥 (𝜕𝑥) = 𝜕𝑥 2 = 2𝑦 3 + 20𝑥 3 cos 𝑦 and 𝜕𝑦 (𝜕𝑥) = 𝜕𝑦𝜕𝑥 = 6𝑥𝑦 2 − 5𝑥 4 sin 𝑦.
𝜕𝑓
 Likewise we can take 𝜕𝑦 = 3𝑥 2 𝑦 2 − 𝑥 5 sin 𝑦 and differentiate it again with respect to 𝑥 or 𝑦. This gives
𝜕 𝜕𝑓 𝜕2 𝑓 𝜕 𝜕𝑓 𝜕2 𝑓
( ) = = 6𝑥𝑦 2 − 5𝑥 4 sin 𝑦 and ( ) = = 6𝑥 2 𝑦 − 𝑥 5 cos 𝑦.
𝜕𝑥 𝜕𝑦 𝜕𝑥𝜕𝑦 𝜕𝑦 𝜕𝑦 𝜕𝑦 2

𝜕2 𝑓 𝜕2 𝑓
Notice that in this example 𝜕𝑦𝜕𝑥
= 𝜕𝑥𝜕𝑦 although they were calculated in different ways. This is not a
coincidence, but a general result, that the order of performing partial differentiation does not affect the
result. The reader may test it for other examples if they would like.
 21

Example: Consider a function of three variables i.e. 𝑓[𝑥, 𝑦, 𝑧] = 𝑥𝑦 2 𝑧 3 . Find all the first order partial
derivatives.

Solution: There are three variables – when differentiating with respect to one of them, we treat the other
two as constants:

𝜕𝑓
 = 𝑦 2 𝑧 3 (treating 𝑥 as the variable and both 𝑦 and 𝑧 as constants)
𝜕𝑥
𝜕𝑓
 = 2𝑥𝑦𝑧 3 (treating 𝑦 as the variable and both 𝑥 and 𝑧 as constants)
𝜕𝑦
𝜕𝑓
 = 3𝑥𝑦 2 𝑧 2 (treating 𝑧 as the variable and both 𝑥 and 𝑦 as constants)
𝜕𝑧

1.7. Problems

1. A banner is hung over the entrance of a hall where a mathematics lecture is being held. It is supposed to
display the function 𝑦 = 10 sin2 3𝑥 , but is stretched horizontally by a factor 1.1. What function is
represented instead?

2. The function sin 𝑥 is printed on a physics T-shirt. (a) When the T-shirt is worn for the first time, it stretches
horizontally (in the 𝑥 direction) by a factor of 1.4. What function is now on the T-shirt? (b) After being worn
for the first time the T-shirt is washed in hot water and shrinks in all directions by the same factor 1.4. What
function is now on the T-shirt?

3. (a) By drawing the graph of 𝑦 = |cos 𝑥|, which is never negative, determine its period by inspection. Then
𝑥
give the periods of (b) 𝑦 = |cos 6𝑥| and (c) 𝑦 = |cos | without calculation.
10

4. What is the (a) period and (b) parity of the function 𝑦 = |tan 𝑥|?

5. A mass 𝑚1 is placed at the origin. It will attract a second mass 𝑚2 placed anywhere at 𝑥 on the x-axis. (a)
Consider how this attractive force 𝐹⃗ [𝑥] changes when we move the second mass from 𝑥 → −𝑥, and hence
give the parity of the gravitational attractive force. (b) In some textbooks one finds Newton’s universal law
𝑚1 𝑚2
of gravitation given as 𝐹[𝑥] = −𝐺 , which is clearly an even function. Can this be reconciled with your
𝑥2

answer for part (a)?

𝜕
6. Calculate 𝜕𝑦 (sin 2𝑥 + cos 3𝑦 + 4𝑥 2 𝑦 3 + 5).

7. Suppose 𝑌[𝑥, 𝑡] = 𝑥 + 3𝑡 + 𝑒 𝑥 + cos 𝑡 + 𝑥 3 𝑡 2. Find all the possible partial derivatives up to the second
order. Does the order of performing the second order mixed partial derivatives affect the result?
 22

2. Equations of motion

2.1. Introduction

The reader should be very familiar with the so-called equations of motion, namely:

𝑣 = 𝑣0 + 𝑎𝑡, (1)

1
𝑠 = 𝑣0 𝑡 + 𝑎𝑡 2 , (2)
2
1
𝑠 = (𝑣0 + 𝑣)𝑡, (3)
2

𝑣 2 = 𝑣02 + 2𝑎𝑠, (4)

1
𝑠 = 𝑣𝑡 − 𝑎𝑡 2 . (5)
2

In these equations 𝑡 represents time, 𝑠 represents displacement, 𝑣 represents velocity (with 𝑣0 being the
initial velocity) and 𝑎 represents acceleration. These equations, although useful in many problems, are
limited because they are only valid if the acceleration is constant.

In this chapter we will show how to derive these equations from some basic definitions and assumptions.
The main aim is then to show how these same basic definitions and different assumptions can be used to
derive different equations of motion when the acceleration is not constant. By doing this, you should realize
that the equations of motion given above are of limited use, while the ability to use the basic definitions is
much more useful in general.

2.2. Equations of motion for constant acceleration

We now show how one may derive the equations of motion given above, using only the definitions of
acceleration and velocity as well as the assumption that the acceleration is constant.

By definition, the acceleration is given by

𝑑𝑣
𝑎= . (6)
𝑑𝑡

Treating the derivative as a fraction, this can be rearranged as 𝑑𝑣 = 𝑎 𝑑𝑡. We can then integrate both sides
to get ∫ 𝑑𝑣 = ∫ 𝑎 𝑑𝑡. Since the acceleration 𝑎 is a constant, the integration yields

𝑣 = 𝑎𝑡 + 𝐶
 23

where 𝐶 is an integration constant. Now initially (at 𝑡 = 0) the velocity was 𝑣 = 𝑣0 . Substituting in this initial
condition gives 𝑣0 = 𝐶. Therefore 𝑣 = 𝑎𝑡 + 𝑣0 and we have derived equation (1).

Also, by definition, the velocity is given by

𝑑𝑥
𝑣= (7)
𝑑𝑡

where 𝑥 represents the position. Treating the derivative as a fraction, this can be written as 𝑑𝑥 = 𝑣 𝑑𝑡. We
can then integrate both sides to get ∫ 𝑑𝑥 = ∫ 𝑣 𝑑𝑡. Since we have already derived that the velocity is given
by equation (1), we can substitute it in on the right hand side to get

1
𝑥 = ∫ 𝑣0 + 𝑎𝑡 𝑑𝑡 = 𝑣0 𝑡 + 𝑎𝑡 2 + 𝐷
2

where 𝐷 is another integration constant. Now initially (at 𝑡 = 0) the position was 𝑥 = 𝑥0 . Substituting this in
1
gives 𝑥0 = 𝐷. Therefore we get 𝑠 = 𝑥 − 𝑥0 = 𝑣0 𝑡 + 2 𝑎𝑡 2 , which is equation (2).

𝑣−𝑣0
If one solves equation (1) for the acceleration one gets 𝑎 = . If one then substitutes this into equation
𝑡
(2) one gets

1 𝑣 − 𝑣0 2 1 1
𝑠 = 𝑣0 𝑡 + ( ) 𝑡 = 𝑣0 𝑡 + (𝑣 − 𝑣0 )𝑡 = (𝑣 + 𝑣0 )𝑡
2 𝑡 2 2

which is equation (3). Similarly, solving equation (1) for either 𝑡 or 𝑣0 and then substituting the result into
equation (2) gives equations (4) and (5) respectively - the reader should check this.

In this manner, we have derived all the equations of motion using only the definitions of acceleration and
velocity and the assumption that the acceleration is constant. These equations of motion are useful in many
problems.

Example: Consider an object of mass 𝑚 thrown horizontally across a floor with initial velocity 𝑣0 . The
coefficient of kinetic friction is 𝜇. Determine the motion of the object, as well as how far the object slides
before coming to a stop.

Solution: Suppose the object is moving towards the right, which we choose as the positive direction. Then
horizontally the only force is the kinetic friction 𝑓𝑘 = −𝜇𝑁. The negative sign is because the friction acts to
the left. In the vertical direction the net force is zero, so the normal force must equal the object's weight i.e.
𝑁 = 𝑚𝑔 . Therefore horizontally ∑ 𝐹 = 𝑓𝑘 = −𝜇𝑚𝑔 . Applying Newton's second law ( ∑ 𝐹 = 𝑚𝑎 ), the
horizontal acceleration of the object is

∑𝐹
𝑎= = −𝜇𝑔.
𝑚

This acceleration is constant and valid while the object is still moving. As soon as it stops, the frictional force
disappears and the acceleration changes to zero. We could now integrate the acceleration to get the velocity,
and then integrate again to get the position. However, this as a case of constant acceleration and so the work
has already been done: equations (1) to (5) need only be applied with 𝑎 = −𝜇𝑔. Therefore the velocity at
any time, until the object stops, is

𝑣 = 𝑣0 + 𝑎𝑡 = 𝑣0 − 𝜇𝑔𝑡.
 24

Thereafter the velocity is, of course, zero. The displacement at any time, until the object stops, is

1 1
𝑠 = 𝑣0 𝑡 + 𝑎𝑡 2 = 𝑣0 𝑡 − 𝜇𝑔𝑡 2 .
2 2
𝑣
The object comes to rest when 𝑣 = 0, which happens at time 𝑡 = 𝜇𝑔0 . After this the object remains stationary,
0 𝑣2
with its displacement being 𝑠 = 2𝜇𝑔 .

2.3. Equations of motion for time dependant accelerations

If the acceleration changes, we can introduce a new definition. Just as velocity is the rate of change of position
𝑑𝑥
with time i.e. 𝑣 = 𝑑𝑡
, and acceleration is the rate of change of velocity with time i.e.
𝑑𝑣
𝑎= 𝑑𝑡
, so we define the jerk to be the rate of change of acceleration with time i.e.

𝑑𝑎
𝑗= . (8)
𝑑𝑡

Note that for the case of constant acceleration the jerk is zero. For any motion with non-zero jerk (i.e. non-
constant acceleration) the original equations of motion given by equations (1) to (5) are no longer relevant.
However, we can apply the basic principles (definitions) to calculate new ones.

2.3.1. Equations of motion for constant jerk

A simple case of changing acceleration is when the jerk is a constant. This does not have to be the case, but
𝑑𝑎
if it is we can derive the equations of motion. By definition 𝑗 = 𝑑𝑡
and treating the derivative as a fraction
this can be written as 𝑑𝑎 = 𝑗 𝑑𝑡. We can then integrate both sides to get ∫ 𝑑𝑎 = ∫ 𝑗 𝑑𝑡. Since the jerk 𝑗 is a
constant (remember, here we are assuming constant jerk), we get

𝑎 = 𝑗𝑡 + 𝐶

where 𝐶 is an integration constant. Now if initially (at 𝑡 = 0) the acceleration was 𝑎 = 𝑎0 , then substituting
this in gives 𝑎0 = 𝐶, so that we get:

𝑎 = 𝑎0 + 𝑗𝑡. (9)

This means that for constant jerk, the acceleration increases linearly with time. This situation might exist
while pulling away in your car as you press down on the accelerator when you feel yourself pressed
𝑑𝑣
backwards onto your seat. We can take the problem further: by definition 𝑎 = 𝑑𝑡
and treating the derivative
as a fraction, this can be written as 𝑑𝑣 = 𝑎 𝑑𝑡. We can then integrate both sides to get ∫ 𝑑𝑣 = ∫ 𝑎 𝑑𝑡. Since
we have already derived that the acceleration is given by equation (9), we can substitute it in to get

1
𝑣 = ∫ 𝑎0 + 𝑗𝑡 𝑑𝑡 = 𝑎0 𝑡 + 𝑗𝑡 2 + 𝐷
2

where 𝐷 is another integration constant. Now initially (at 𝑡 = 0) the velocity was 𝑣 = 𝑣0 . Substituting this
gives 𝑣0 = 𝐷, so that:
 25

1
𝑣 = 𝑣0 + 𝑎0 𝑡 + 𝑗𝑡 2 . (10)
2
𝑑𝑥
Taking this still further, by definition 𝑣 = 𝑑𝑡
and treating the derivative as a fraction, this can be written as
𝑑𝑥 = 𝑣 𝑑𝑡. We can then integrate both sides to get ∫ 𝑑𝑥 = ∫ 𝑣 𝑑𝑡. Since we have already derived that the
velocity is given by equation (10), we can substitute it to get

1 1 1
𝑥 = ∫ 𝑣0 + 𝑎0 𝑡 + 𝑗𝑡 2 𝑑𝑡 = 𝑣0 𝑡 + 𝑎0 𝑡 2 + 𝑗𝑡 3 + 𝐸
2 2 6

where 𝐸 is yet another integration constant. Now initially (at 𝑡 = 0) the position was 𝑥 = 𝑥0 . Substituting
this in gives 𝑥0 = 𝐸, so

1 1
𝑠 = 𝑥 − 𝑥0 = 𝑣0 𝑡 + 𝑎0 𝑡 2 + 𝑗𝑡 3 . (11)
2 6

These new equations of motion (equations (9) to (11)) are only valid if the jerk is constant.

Example: In a drag race, car A is equipped with a jet engine which provides a constant acceleration of 5 m/s2,
while car B is equipped with a jet engine providing an initial acceleration of 3 m/s2 as well as a constant jerk
of 1 m/s3. Both cars begin the race from rest.
a) After what time will the cars have equal acceleration?
b) After what time will the cars have the same speed?
c) What must be the minimum distance of the race if car B is to win?
d) Suppose the cars tie the race. How much faster is car B travelling than car A as they cross the finish line?

Solution: We assume 𝑥0 = 0 for both cars. Also, since the cars start from rest, 𝑣0 = 0 for both cars.
 Car A moves with constant acceleration with 𝑎 = 5. So from equation (1) we get 𝑣 = 5𝑡 and from
5
equation (2) we get 𝑠 = 2 𝑡 2 .
 Car B moves with constant jerk, so it has 𝑗 = 1 and 𝑎0 = 3. From equation (9) we get 𝑎 = 𝑡 + 3, from
1 3 1
equation (10) we get 𝑣 = 3𝑡 + 2 𝑡 2 , and finally from equation (11) we get 𝑠 = 2 𝑡 2 + 6 𝑡 3 .

a) Setting the accelerations equal gives 5 = 3 + 𝑡, meaning the cars will have the same acceleration after
2 s.
1
b) Setting the velocities equal gives 5𝑡 = 3𝑡 + 2 𝑡 2 . Solving this gives 𝑡 = 0 s or 𝑡 = 4 s. Of course the cars
have the same speed initially (they both start from rest), but it is actually the time of 4 s we are interested
in.
c) Car A pulls ahead initially, but car B later catches up and passes it. We can find the time when the two
5 3 1
cars pass each other by equating their displacements: 2 𝑡 2 = 2 𝑡 2 + 6 𝑡 3 . Solving this gives 𝑡 = 0 s or 𝑡 =
6 s. Therefore car B will pass car A after exactly 6 s. During this time, both cars would have moved exactly
90 m. For car B to win, the race must therefore be more than 90 m long.
d) If the cars tie the race, they do so after 6 s. Car A then has a speed of 30 m/s, while car B then has a speed
of 36 m/s. Car B is travelling 6 m/s faster than car A as they cross the finish line.
 26

2.3.2. Other time dependent accelerations

For such problems we cannot apply any equations of motion, and each problem has to be solved using basic
principles. We give some examples.

Example: Consider a particle that has an acceleration 𝑎 = 𝑒 −𝑡 . Calculate its jerk, as well as its position at any
time. Also find an equation giving its position for any velocity.

Solution: To get the jerk, we apply its definition:

𝑑𝑎 𝑑
𝑗= = (𝑒 −𝑡 ) = −𝑒 −𝑡 .
𝑑𝑡 𝑑𝑡
𝑑𝑣
To get the position, we must first find the velocity. This can be done using 𝑎 = 𝑑𝑡
. This means 𝑑𝑣 = 𝑎 𝑑𝑡, so

𝑣 = ∫ 𝑎 𝑑𝑡 = ∫ 𝑒 −𝑡 𝑑𝑡 = −𝑒 −𝑡 + 𝐶.

With initial condition 𝑣 = 𝑣0 when 𝑡 = 0, we obtain 𝐶 = 𝑣0 + 1 and the equation becomes

𝑣 = 𝑣0 + 1 − 𝑒 −𝑡 .

𝑑𝑥
We can now apply the definition of velocity i.e. 𝑣 = 𝑑𝑡
. This means 𝑑𝑥 = 𝑣 𝑑𝑡, so

𝑥 = ∫ 𝑣 𝑑𝑡 = ∫ 𝑣0 + 1 − 𝑒 −𝑡 𝑑𝑡 = (𝑣0 + 1)𝑡 + 𝑒 −𝑡 + 𝐷.

With initial conditions 𝑥 = 𝑥0 when 𝑡 = 0, we obtain 𝐷 = 𝑥0 − 1 and so the equation becomes

𝑠 = 𝑥 − 𝑥0 = (𝑣0 + 1)𝑡 + 𝑒 −𝑡 − 1.

To get the position at any velocity, we use the relationship between velocity and time derived earlier, namely
𝑣 = 𝑣0 + 1 − 𝑒 −𝑡 . From this we get 𝑒 −𝑡 = 𝑣0 + 1 − 𝑣, as well as 𝑡 = − ln(𝑣0 + 1 − 𝑣). Substituting these
gives:

𝑠 = (𝑣0 + 1)[− ln(𝑣0 + 1 − 𝑣)] + [𝑣0 + 1 − 𝑣] − 1 = −(𝑣0 + 1) ln(𝑣0 + 1 − 𝑣) + 𝑣0 − 𝑣.

Note: Examining this solution, it should disturb you that the units give a problem – for instance, we have
displacement on the left side, but some velocity terms on the right side. This is a serious problem that will
recur in some other examples in this chapter. However, it need not worry us too much: it originates from the
original expression given for the acceleration, namely 𝑎 = 𝑒 −𝑡 . Clearly this does not ensure the acceleration
has the correct units! A better expression is 𝑎 = 𝑘𝑒 −𝑡/𝜏 where 𝑘 is a constant with units of acceleration and
𝜏 is a constant with units of time (and of course we could set 𝑘 = 𝜏 = 1, but with their proper units, to get
back our original problem). The serious reader is challenged to redo the problem with this expression and
verify that the units now work out correctly.
 27

Example: Consider a rocket launched from rest from the surface of the Earth. A rocket expels gas to generate
thrust, but while doing so it loses mass. Newton's second law does not apply directly to variable mass
systems. However, it can be applied to the total (fixed mass) system including the rocket and its exhaust gas
to show that ∑ 𝐹 = 𝑚𝑎 applies to the rocket1 if: (1) ∑ 𝐹 is the sum of any external forces plus the thrust,
given by kU where k is the rate at which the rocket’s mass decreases and 𝑈 is the speed of the ejected gas
relative to the rocket, and (2) the mass 𝑚 represents the decreasing mass of the rocket.

Assume k and U are constants and determine the velocity of a rocket, launched from rest, as a function of
time. Ignore air friction but take gravity (i.e. the weight of the rocket) into account. You can assume the
gravitational acceleration (g) is constant, i.e. the rocket does not go high enough that you need to consider
how the value of g decreases as one moves further from the centre of the Earth.

Solution: We are given that the rocket expels gas at a constant rate 𝑘, so if its initial mass is 𝑚0 then its mass
at any time must be 𝑚[𝑡] = 𝑚0 − 𝑘𝑡 until the fuel is used up. The acceleration is given by

∑𝐹 𝑘𝑈 − 𝑚[𝑡]𝑔 𝑘𝑈 𝑘𝑈
𝑎= = = −𝑔 = − 𝑔.
𝑚[𝑡] 𝑚[𝑡] 𝑚[𝑡] 𝑚0 − 𝑘𝑡

Therefore

𝑘𝑈 1
𝑣 = ∫ 𝑎 𝑑𝑡 = ∫ − 𝑔 𝑑𝑡 = 𝑘𝑈 ∫ 𝑑𝑡 − ∫ 𝑔 𝑑𝑡 = − 𝑈 ln(𝑚0 − 𝑘𝑡) − 𝑔𝑡 + 𝐶.
𝑚0 − 𝑘𝑡 𝑚0 − 𝑘𝑡

If 𝑣 = 0 when 𝑡 = 0 then 𝐶 = 𝑈 ln(𝑚0 ) and so

𝑚0 𝑚0
𝑣 = −𝑈 ln(𝑚0 − 𝑘𝑡) − 𝑔𝑡 + 𝑈 ln(𝑚0 ) = 𝑈 ln ( ) − 𝑔𝑡 = 𝑈 ln ( ) − 𝑔𝑡.
𝑚0 − 𝑘𝑡 𝑚[𝑡]

2.4. Equations of motion for position dependant accelerations

Suppose the acceleration is not constant, but it is known as a function of position rather than time. Physical
examples of such situations might be an object attached to a spring, or a marble rolling in a curved bowl, or
a satellite a certain distance from the earth.

We then have a position dependent acceleration, which we can write as 𝑎 = 𝑎[𝑥] . The definition of
𝑑𝑣
acceleration states that 𝑎 = 𝑑𝑡
, so that

𝑣 = ∫ 𝑎[𝑥] 𝑑𝑡.

While this result is not incorrect, it is not useful. We cannot perform this integral, since it is an integral with
respect to time (it ends with 𝑑𝑡, so time is the variable to be integrated), yet the acceleration 𝑎[𝑥] is a
function of position. We may not assume that since the acceleration is given in terms of position, it is
independent of time (i.e. constant). The acceleration will change with time, but in some unknown way. We
need to find a new way to deal with such problems.

1 See e.g. ‘Mathematical analysis of a model rocket trajectory’ by Nelson and Wilson, The Physics Teacher, March 1976, p.150.
 28

𝑑𝑣
Returning to the differential equation 𝑎[𝑥] = 𝑑𝑡
, one may notice that three variables are involved, namely
position, velocity and time – note that we do not count the acceleration as a separate variable because it is
known or given in terms of the position. An equation containing three variables cannot be solved directly,
because an equation only has two sides and hence one cannot separate all the variables. Fortunately, we can
derive a new physics result to help us.

Starting with the definition of acceleration, we apply the chain rule:

𝑑𝑣 𝑑𝑣 𝑑𝑥 𝑑𝑣
𝑎= = =𝑣 . (12)
𝑑𝑡 𝑑𝑥 𝑑𝑡 𝑑𝑥

This is a useful new law that we can use to solve any problem when a position dependent acceleration is
given. You do not have to remember it – just derive it from the definitions of acceleration and velocity when
required.

Again returning to our position dependent acceleration 𝑎 = 𝑎(𝑥) and applying this result, we can write

𝑑𝑣
𝑎[𝑥] = 𝑣 .
𝑑𝑥

Notice that there are now only two variables in the problem: position and velocity. Again we do not count
the acceleration, since it is known in terms of the position. Such an equation is solvable, but before we can
integrate it is vital to first separate the variables, placing all position dependent terms on one side and all
velocity dependent terms on the other as follows:

𝑎[𝑥]𝑑𝑥 = 𝑣𝑑𝑣.

Note that all the “differentials” (i.e. 𝑑𝑥 and 𝑑𝑣) must always be put above the line. Now we can integrate
both sides, as the following example illustrates.

Example: Suppose an object rolls from the top of a hill, and for a while the hill is shaped so that its horizontal
acceleration is given by 𝑎 = 6√𝑥. Find the horizontal velocity as a function of the position. Assume the object
starts from position 𝑥0 = 1 with initial velocity 𝑣0 = √8.

𝑑𝑣 𝑑𝑣 𝑑𝑥 𝑑𝑣
Solution: We have a position dependent acceleration 𝑎 = 6√𝑥. Thus we use 𝑎 = 𝑑𝑡
= 𝑑𝑥 𝑑𝑡 = 𝑣 𝑑𝑥, getting
𝑑𝑣
6√𝑥 = 𝑣 .
𝑑𝑥

We then separate the variables, keeping the differentials above the line:

6√𝑥 𝑑𝑥 = 𝑣 𝑑𝑣,

and then integrate both sides to get

1
4𝑥 3/2 = 𝑣 2 + 𝐶.
2

Applying the initial condition gives 𝐶 = 0, so after simplifying one gets

𝑣 = √8𝑥 3/4 .
 29

Notice that the method above gives us the velocity as a function of position. Suppose we need the velocity
as a function of time, or the position as a function of time?

 We will tackle the second issue first: suppose we have 𝑣 = 𝑣[𝑥], and we require the position as a function
of time.

𝑑𝑥
From the definition of velocity 𝑣[𝑥] = 𝑑𝑡
one can get

𝑥 = ∫ 𝑣[𝑥]𝑑𝑡.

We are faced with a similar problem to before – we cannot perform a time integral for a function of
𝑑𝑥
position. But returning to 𝑣[𝑥] = 𝑑𝑡
, we see that there are only two variables involved, namely position
and time, and so will be able to separate the variables and then integrate. To illustrate this, in the
previous example we obtained the position-dependent velocity 𝑣 = √8𝑥 3/4 for an object rolling off a
hill. To find the position as a function of time for this object, we apply the definition of velocity to get

𝑑𝑥
= √8𝑥 3/4 .
𝑑𝑡

Separating the position and time parts, while keeping the differentials above the line, yields

𝑥 −3/4 𝑑𝑥 = √8 𝑑𝑡.

Integrating both sides gives

4𝑥 1/4 = √8 𝑡 + 𝐷.

When 𝑡 = 0 the position is 𝑥 = 1 (see the initial conditions given in the previous example), so the
integration constant is 𝐷 = 4. Therefore

4
√8
𝑥=( 𝑡 + 1) .
4

 Now returning to the first issue, which is how to find the velocity as a function of time. There are two
possible ways of doing this:

(1) One can differentiate the expression just found for 𝑥, giving:

4 3 3
𝑑𝑥 𝑑 √8 √8 √8 √8
𝑣= = ( 𝑡 + 1) = 4 ( 𝑡 + 1) = √8 ( 𝑡 + 1) .
𝑑𝑡 𝑑𝑡 4 4 4 4

(2) One can substitute the expression found for 𝑥 into the equation for the position-dependent velocity:

4 3/4 3
3/4 √8 √8
𝑣 = √8𝑥 = √8 [( 𝑡 + 1) ] = √8 ( 𝑡 + 1) .
4 4
 30

2.5. Equations of motion for velocity dependant accelerations

Suppose the acceleration of a particle is known as a function of velocity i.e. 𝑎 = 𝑎[𝑣]. A physical example of
such a situation might be an object moving through the air and experiencing air resistance. For such a
problem we can proceed in two different ways:

𝑑𝑣
 Applying the definition of acceleration, we have 𝑎[𝑣] = 𝑑𝑡
which contains only velocity and time and
may be separated, then integrated to give the velocity as a function of time.

𝑑𝑣 𝑑𝑣 𝑑𝑥 𝑑𝑣
 But we have seen that the acceleration can also be expressed as 𝑎 = 𝑑𝑡
= 𝑑𝑥 𝑑𝑡 = 𝑣 𝑑𝑥. Using this form,
𝑑𝑣
we have 𝑎[𝑣] = 𝑣 𝑑𝑥, which contains only velocity and position and may be separated, then integrated
to give the velocity as a function of position.

Therefore given a velocity dependent acceleration, one can calculate the velocity either as a function of time
or as a function of position, depending on the method used. From either of these and the definition of velocity
the position as a function of time can be obtained.

Example: Suppose 𝑎 = −𝑣 2 . Find the position as a function of time, using both of the possible methods
discussed above.

Solution: We consider each of the two possible methods:

𝑑𝑣 𝑑𝑣
 Using 𝑎 = 𝑑𝑡
one has 𝑑𝑡 = −𝑣 2 . The separated equation is 𝑣 −2 𝑑𝑣 = −𝑑𝑡, which can be integrated to
−1 −1
get = −𝑡 + 𝐶. Assuming that 𝑣 = 𝑣0 when 𝑡 = 0 gives the integration constant 𝐶 = , and so after
𝑣 𝑣0
some rearranging

1
𝑣= .
𝑡 + 1/𝑣0

This velocity as a function of time can be integrated to get the position as a function of time: 𝑥 = ∫ 𝑣 𝑑𝑡 =
1
∫ 𝑡+1/𝑣 𝑑𝑡 = ln(𝑡 + 1/𝑣0 ) + 𝐷. Applying the initial condition 𝑥 = 𝑥0 at 𝑡 = 0 gives 𝑥0 = ln(1/𝑣0 ) + 𝐷, so
0
𝐷 = 𝑥0 + ln(𝑣0 ), and

𝑥 = ln(𝑡 + 1/𝑣0 ) + 𝑥0 + ln(𝑣0 ) = 𝑥0 + ln(𝑣0 𝑡 + 1).

𝑑𝑣 𝑑𝑣 𝑑𝑥 𝑑𝑣 𝑑𝑣 𝑑𝑣
 Using 𝑎 = = =𝑣 , one has 𝑣 = −𝑣 2. The separated equation is = −𝑑𝑥, which can be
𝑑𝑡 𝑑𝑥 𝑑𝑡 𝑑𝑥 𝑑𝑥 𝑣
integrated to get ln 𝑣 = −𝑥 + 𝐶 . Assuming that 𝑣 = 𝑣0 when 𝑥 = 𝑥0 (both when 𝑡 = 0 ) gives the
𝑣
integration constant 𝐶 = 𝑥0 + ln 𝑣0 , and so ln 𝑣 = −𝑥 + 𝑥0 + ln 𝑣0 , or ln 𝑣 = 𝑥0 − 𝑥. Therefore
0

𝑣 = 𝑣0 𝑒 𝑥0 −𝑥 .

𝑑𝑥 𝑑𝑥
To get the position as a function of time, we use the definition of velocity 𝑣 = 𝑑𝑡
, which means 𝑑𝑡 = 𝑣0 𝑒 𝑥0 −𝑥 .
This can be separated to get 𝑒 𝑥−𝑥0 𝑑𝑥 = 𝑣0 𝑑𝑡, which after integration yields 𝑒 𝑥−𝑥0 = 𝑣0 𝑡 + 𝐷. Applying the
initial condition 𝑥 = 𝑥0 at 𝑡 = 0 gives 𝐷 = 1, and so 𝑒 𝑥−𝑥0 = 𝑣0 𝑡 + 1, or

𝑥 = 𝑥0 + ln(𝑣0 𝑡 + 1).
 31

2.6. Problems

1. Consider a particle which has jerk 𝑗 = 𝑒 −𝑡 . Find the position of the particle at any time. Assume that at
𝑡 = 0 the motion is 𝑎 = 𝑎0 , 𝑣 = 𝑣0 and 𝑥 = 𝑥0 .

2. Suppose 𝑎 = 1 − 𝑡.
(a) Calculate the jerk.
(b) Integrate twice and find the position as a function of the time if the initial position is 𝑥0 and the initial
velocity is 𝑣0 .
(c) Draw graphs of 𝑎 versus 𝑡, 𝑣 versus 𝑡 and 𝑥 versus 𝑡, assuming that 𝑥0 = 0 and 𝑣0 = 0.

3. The force on an electron as a result of electromagnetic radiation is 𝐹[𝑡] = −𝑒𝐸0 sin(𝜔𝑡 + 𝜑). In this
equation 𝑒, 𝐸0 , 𝜔 and 𝜑 are all constants.
(a) Give an expression for the resulting acceleration of the electron.
(b) If the electron is initially at rest at the origin, find expressions for its velocity and position as a function of
time.
(c) One might expect that the displacement of a particle would be in the direction of the force applied to it.
Is this true in the present case (motivate your answer)?

4. A block of mass 𝑚 initially travels over a frictionless surface with constant velocity 𝑣0 . It suddenly (at 𝑡 =
0) begins to encounter a gust of wind, causing a retarding force 𝐹 = −𝐹0 𝑒 −𝜆𝑡 . Find its velocity and position
at any time. Also find a condition for 𝐹0 in terms of 𝑚, 𝜆 and 𝑣0 so that the wind changes the direction the
block is moving, and the time when it changes direction.

5. Find the position as a function of time, and differentiate your result to obtain the velocity as a function of
time, in each of the following cases:
(a) 𝑎 = 6𝑥, assuming 𝑥0 = 1 and 𝑣0 = √6.
(b) 𝑎 = 6𝑥 2 , assuming 𝑥0 = 1 and 𝑣0 = 2.

6. The gravitational force of the earth on an object having mass 𝑚, which is far from the earth's surface, is
𝑚𝑀
given by 𝐹 = −𝐺 𝑥2
.
(a) What is the resulting acceleration of the object?
(b) Calculate the velocity of the object as a function of its position, if it has initial velocity 𝑣0 at initial position
𝑥0 .
(c) Use your solution to calculate the minimum initial speed with which an object must be launched from the
earth's surface, if it is never to return (i.e. it carries on going to infinity). Note that this calculation ignores air
resistance.

7. An object of mass 𝑚 is thrown with initial velocity 𝑣0 along a horizontal surface so that the retarding force
is proportional to the square root of the velocity at any moment.
(a) Obtain expressions for the velocity and position of the object at any time.
(b) How far it the object travel before coming to rest?

8. An object of mass 𝑚 is thrown with initial velocity 𝑣0 along a horizontal surface so that the retarding force
(frictional force) is proportional to the velocity at any moment. Find the position of the object at any time.

9. Consider an acceleration given by 𝑎 = √1 − 𝑣 2 .

 32

𝑑𝑣
(a) Use 𝑎 = 𝑣 𝑑𝑥 and find the position as a function of time. (You will first have to obtain the velocity as a
function of position.)
𝑑𝑣
(b) Now use 𝑎 = 𝑑𝑡
and find the position as a function of time. (You will first have to obtain the velocity as a
function of time.)
(c) Are the two solutions equivalent?
 33

3. Case study: Free fall with drag

3.1. Drag forces

Drag is the name given to the resistive force acting on a solid body as it moves through a fluid (i.e. a liquid or
gas). We consider two simple models for the drag on a sphere (ball) of radius 𝑅 moving with speed 𝑣, and
work out the equations of motion associated with each.

Afterwards we consider under which circumstances these models are valid and give a more general
discussion of drag.

3.2. Motion with viscous drag

3.2.1. Viscous drag

Suppose the ball is small and moving slowly, so that the fluid can slide around the ball as it moves. This sliding
of the fluid around the ball is called laminar flow. The drag increases (a) if the fluid becomes more thick
(viscous) and flows less easily, (b) if the radius of the ball becomes larger so that the fluid must flow further
around the ball, and (c) if the ball moves faster. Since the drag depends on the viscosity 𝜂 of the fluid, it is
called viscous drag and it is given by Stokes law2

|𝑓𝑣 | = 6𝜋𝜂𝑅𝑣. (13)

3.2.2. Vertical motion with viscous drag

fv

Figure 14
v
Force diagram for an object falling and experiencing a
drag force.

mg

Consider a small sphere moving vertically and slowly through a fluid, so that viscous drag occurs. We choose
the coordinate axes so that the positive 𝑌 direction is downwards. Then the gravitational force (weight)

2Stokes, G.G. (1851) On the effect of the internal friction of fluids on the motion of pendulums. Cambridge Philosophical Society,
Transactions, v. 9, no. 8, p. 287.
 34

acting on the particle is 𝑚𝑔, while the drag (which is always in the opposite direction of the velocity) is given
by 𝑓𝑣 = −𝐾𝑣 where 𝐾 = 6𝜋𝜂𝑅 is a positive constant. We assume that the drag force remains in this form
(although the constant will change) even if the object is not a sphere. The force diagram is shown in Figure
14 for an object moving downwards. If the object was moving upwards the drag force would be in the
opposite direction (downwards) i.e. in the same direction as the weight.

Applying Newton's second law to the particle gives 𝑚𝑎 = 𝑚𝑔 − 𝐾𝑣, or

𝐾
𝑎=𝑔− 𝑣. (14)
𝑚

Figure 15

𝑉 Acceleration versus speed for viscous drag.

𝑣

Figure 15 shows the graph of 𝑎 versus 𝑣. The graph is a straight line with a negative slope. If the velocity is
zero, the acceleration is 𝑔 (since there is no drag while the object is not moving). If 𝑣 is negative (i.e. the
particle is moving upwards), then the drag force acts downwards and the total downward force on the
particle exceeds that from gravity alone. When 𝑣 is positive (i.e. the particle is falling), then the drag force is
upwards and acts in opposition to the weight of the object.3

Considering equation (14), if one solves for the situation where the acceleration of the particle is zero, this
𝑚𝑔
occurs when its velocity is given by 𝑣 = 𝐾
. This is defined as the terminal velocity

𝑚𝑔
𝑉= (15)
𝐾

and is called the terminal velocity because the acceleration is zero, so the velocity will not change. The
terminal velocity is a positive quantity corresponding to a downward velocity. We can eliminate 𝐾 from
equation (14) using equation (15), which gives

𝑣 𝑔
𝑎 = 𝑔 (1 − ) = (𝑉 − 𝑣). (16)
𝑉 𝑉

This equation of motion gives the acceleration as a function of velocity. In order to solve the equation, we
𝑑𝑣
could replace 𝑎 by 𝑑𝑡 to get a separable differential equation containing only 𝑣 and 𝑡, or alternatively we
𝑑𝑣
could replace 𝑎 by 𝑣 𝑑𝑦 to get a separable differential equation containing only 𝑣 and 𝑦 (where we use 𝑦

3 In fact, if the object has a very large positive velocity (downwards), it is possible that the drag force can exceed the weight, so that
the object has a negative acceleration (upwards)! This does not mean that the particle moves upwards. Remember that it has a very
large positive velocity, so the negative acceleration only means that the velocity decreases.
 35

instead of 𝑥 because we are dealing with vertical motion). We choose here the first option as it will give the
velocity as a function of time (while the second option would give the relationship between the velocity and
𝑑𝑣
position). If we substitute 𝑎 = 𝑑𝑡
in equation (16), separate variables and integrate, we get

𝑑𝑣 𝑔
∫ = ∫ − 𝑑𝑡,
𝑣−𝑉 𝑉
𝑔
ln(𝑣 − 𝑉) = − 𝑡 + 𝐶
𝑉

where 𝐶 is an integration constant. If 𝑣 = 𝑣0 when 𝑡 = 0, then 𝐶 = ln(𝑣0 − 𝑉 ), so

𝑔
ln(𝑣 − 𝑉) = − 𝑡 + ln(𝑣0 − 𝑉 ).
𝑉

Solving for the velocity gives

𝑔
𝑣 = 𝑉 + (𝑣0 − 𝑉 )𝑒 −𝑉𝑡 . (17)

𝑑𝑦
To determine the position as a function of time, we rewrite the velocity as 𝑣 = 𝑑𝑡
and integrate:

𝑔
∫ 𝑑𝑦 = ∫ 𝑉 + (𝑣0 − 𝑉 )𝑒 −𝑉𝑡 𝑑𝑡,

𝑉 𝑔
𝑦 = 𝑉𝑡 − (𝑣0 − 𝑉 )𝑒 −𝑉𝑡 + 𝐷
𝑔

𝑉
where 𝐷 is an integration constant. If 𝑦 = 𝑦0 when 𝑡 = 0, then 𝐷 = 𝑦0 + 𝑔 (𝑣0 − 𝑉 ), so

𝑉 𝑔
𝑠 = 𝑦 − 𝑦0 = 𝑉𝑡 + (𝑣0 − 𝑉 ) (1 − 𝑒 −𝑉𝑡 ). (18)
𝑔

Example: Find the maximum height reached by a particle thrown upwards, assuming viscous drag.

Solution: At the maximum height the object travelling upwards will come momentarily to rest (stop) i.e. 𝑣 =
0. Substituting this into equation (17) gives the time when this occurs:

𝑉 −𝑉 𝑉 𝑉 − 𝑣0
𝑡 = − ln = ln .
𝑔 𝑣0 − 𝑉 𝑔 𝑉

Note that if the object is thrown upwards, the value of 𝑣0 should be negative (recall that we chose
downwards as positive). We can substitute this into equation (18) to get the maximum height:

𝑉 𝑔 𝑉 𝑉 − 𝑣0 𝑉 −𝑉
𝑠 = 𝑉𝑡 + (𝑣0 − 𝑉 ) (1 − 𝑒 −𝑉𝑡 ) = 𝑉 ln + (𝑣0 − 𝑉 ) (1 − ),
𝑔 𝑔 𝑉 𝑔 𝑣0 − 𝑉

𝑉 2 𝑉 − 𝑣0 𝑉 𝑣0 − 𝑉 + 𝑉 𝑉 𝑉 − 𝑣0
𝑠= ln + (𝑣0 − 𝑉 ) ( ) = (𝑉 ln + 𝑣0 ).
𝑔 𝑉 𝑔 𝑣0 − 𝑉 𝑔 𝑉

Again, the value of 𝑠 should be negative, since we chose downwards as positive.

 36

3.3. Motion with turbulent drag

3.3.1. Turbulent drag

Suppose the ball is large and moving quickly, so that the fluid is forced out the way and cannot slide smoothly
around the ball. There is a turbulent flow of the fluid behind the ball where a low pressure region is created.
In this case the viscosity of the fluid is not so important, but rather the momentum that must be transferred
from the ball to the fluid in order to force the fluid out of the way. In a time 𝑡, the ball moving with velocity
𝑣 and having cross sectional area 𝐴 = 𝜋𝑅 2 sweeps out a volume 𝜋𝑅 2 𝑣𝑡, and must therefore remove fluid
with a total mass of 𝜌𝜋𝑅 2 𝑣𝑡 from its path, where 𝜌 is the density of the fluid. Making the rough assumption
that this fluid is given the same speed as the ball, the momentum transferred to the fluid is Δ𝑝 = 𝜌𝜋𝑅 2 𝑣 2 𝑡.
But the impulse-momentum theorem states that the impulse (i.e. force times time) equals the change of
momentum, therefore the turbulent drag force is given by 𝜌𝜋𝑅 2 𝑣 2. A more detailed analysis, considering
more carefully the speed transferred from the ball to the fluid, gives the equation

𝐶𝑑
|𝑓𝑡 | = 𝜌𝜋𝑅 2 𝑣 2 (19)
2
1
where 𝐶𝑑 is a factor called the drag coefficient, which is approximately for a ball/sphere. This formula can
2
be adapted to different shapes shown in Figure 16 by using the appropriate drag coefficient and replacing
𝜋𝑅 2 with the cross-sectional area.

Figure 16

Drag coefficients (𝐶𝑑 ) for different shapes.

3.3.2. Vertical motion with turbulent drag

Consider a large object moving quickly through a fluid. We will again assume the motion is vertical, with our
coordinate axes so that the positive Y direction is downwards. Then the gravitational force (weight) acting on
the particle is 𝑚𝑔, while the drag is proportional to the square of the velocity and in the opposite direction
𝐶𝑑
i.e. 𝑓𝑡 = ±𝐾𝑣 2 where 𝐾 = 2
𝜌𝜋𝑅 2 is a positive constant. If the object is falling downwards, then the drag is
upwards and the negative sign applies, but if the object is moving upwards, then the drag is downwards and
the positive sign applies. We can write this as
 37

+𝐾𝑣 2 if 𝑣 < 0 ;
𝑓𝑡 = {
−𝐾𝑣 2 if 𝑣 > 0.

Applying Newton's second law to the particle gives

𝑚𝑔 + 𝐾𝑣 2 if 𝑣 < 0 ;
𝑚𝑎 = {
𝑚𝑔 − 𝐾𝑣 2 if 𝑣 > 0,

or after dividing throughout by 𝑚:

𝐾 2
𝑔+ 𝑣 if 𝑣 < 0 ;
𝑚
𝑎= (20)
𝐾
𝑔 − 𝑣 2 if 𝑣 > 0.
{ 𝑚

Figure 17 shows the graph of 𝑎 versus 𝑣 . The graph consists of two parabolas, on either side of the
acceleration axis. If the velocity is zero, the acceleration is 𝑔 (since there is no drag while the object is not
moving).

𝑔 Figure 17

Acceleration versus speed for turbulent drag.

𝑉
𝑣

Considering equation (20), if one solves for the situation where the acceleration of the particle is zero, this
𝑚𝑔
occurs when its velocity is given by 𝑣 = √ 𝐾
, which is defined as the terminal velocity in this case:

𝑚𝑔
𝑉=√ . (21)
𝐾

Note that our definition of the terminal velocity depends on the type of drag we are considering (and likewise,
so does the meaning of 𝐾). Using our definition of the terminal velocity, we can eliminate 𝐾 from the
equation of motion (equation (20)) and obtain
 38

𝑣2
𝑔 (1 + ) if 𝑣 < 0 ;
𝑉2
𝑎= (22)
2
𝑣
𝑔 (1 − ) if 𝑣 > 0.
{ 𝑉2

Since we have two possible forms for the acceleration, we will solve each one separately, starting with an
object thrown upwards (i.e. with negative initial velocity).

3.3.3. Particle thrown upwards

For a particle moving upwards 𝑣 < 0, since we have chosen downwards as the positive direction. Therefore
the top half of equation (22) gives us the acceleration as a function of velocity. In order to solve the equation,
𝑑𝑣
we could replace 𝑎 by 𝑑𝑡 to get a separable differential equation containing only 𝑣 and 𝑡, or alternatively we
𝑑𝑣
could replace 𝑎 by 𝑣 𝑑𝑦 to get a separable differential equation containing only 𝑣 and 𝑦. Since we would
prefer to know the velocity as a function of time, rather than as a function of position, we shall use the first
option. Therefore

𝑑𝑣 𝑣2
= 𝑔 (1 + 2 ).
𝑑𝑡 𝑉

We first separate and then integrate this as follows:

𝑑𝑣
∫ = ∫ 𝑔𝑑𝑡,
𝑣2
1+ 2
𝑉
𝑣
𝑉 arctan = 𝑔𝑡 + 𝐶
𝑉
𝑣0
where 𝐶 is an integration constant. If 𝑣 = 𝑣0 when 𝑡 = 0, then 𝐶 = 𝑉 arctan 𝑉
and solving for the velocity
gives

𝑔𝑡 𝑣0
𝑣 = 𝑉 tan ( + arctan ). (23)
𝑉 𝑉

Remember to have your calculator set to work in radians, not degrees, when working with this formula, and
also that for an object initially thrown upwards 𝑣0 is negative.

To get the position as a function of time, we integrate again:

𝑔𝑡 𝑣0
∫ 𝑑𝑦 = ∫ 𝑉 tan ( + arctan ) 𝑑𝑡
𝑉 𝑉

Since ∫ tan 𝑢 𝑑𝑢 = − ln cos 𝑢 + 𝐷, we get

𝑉2 𝑔𝑡 𝑣0
𝑦=− ln cos ( + arctan ) + 𝐷.
𝑔 𝑉 𝑉
 39

𝑉2 𝑣
If 𝑦 = 𝑦0 when 𝑡 = 0, then 𝐷 = 𝑦0 + 𝑔
ln cos (arctan 𝑉0 ). This can be simplified because cos(arctan 𝑢) =
−1/2
𝑉2 𝑣0 2 𝑉2 𝑣 2
(1 + 𝑢2 )−1/2 , so 𝐷 = 𝑦0 + ln (1 + ( ) ) = 𝑦0 − 2𝑔 ln (1 + ( 𝑉0 ) ). Therefore
𝑔 𝑉

𝑉2 𝑔𝑡 𝑣0 𝑉2 𝑣0 2
𝑠 = 𝑦 − 𝑦0 = − ln cos ( + arctan ) − ln (1 + ( ) ). (24)
𝑔 𝑉 𝑉 2𝑔 𝑉

Example: Find the maximum height reached by a particle thrown upwards, assuming turbulent drag.

Solution: At the maximum height the object travelling upwards will come momentarily to rest (stop) i.e. 𝑣 =
0. Substituting this into equation (23) gives the time when this occurs:

𝑉 𝑣0
𝑡 = − arctan .
𝑔 𝑉

Note that if the object is thrown upwards, the value of 𝑣0 should be negative (recall that we chose
downwards as positive), so 𝑡 will be positive as expected. We can substitute this into equation (24) to get the
maximum height:

𝑉2 𝑔 𝑉 𝑣0 𝑣0 𝑉2 𝑣0 2 𝑉2 𝑣0 2
𝑠=− ln cos ( (− arctan ) + arctan ) − ln (1 + ( ) ) = − ln (1 + ( ) ),
𝑔 𝑉 𝑔 𝑉 𝑉 2𝑔 𝑉 2𝑔 𝑉

since the first term is zero. Again, the value of 𝑠 should be negative (upwards direction).

3.3.4. Particle dropped from rest

We will not solve the general case of a particle thrown downwards, but only the special case for a particle
dropped from rest i.e. 𝑣0 = 0. This simplifies the mathematics somewhat. The second part of equation (22)
gives us the acceleration as a function of velocity for a particle falling downwards. Therefore

𝑑𝑣 𝑣2
= 𝑔 (1 − 2 ).
𝑑𝑡 𝑉

We first separate and then integrate this as follows:

𝑑𝑣
∫ = ∫ 𝑔𝑑𝑡,
𝑣2
1− 2
𝑉
𝑣
𝑉 arctanh = 𝑔𝑡 + 𝐶
𝑉
0
where a hyperbolic function occurs in the integral. If 𝑣 = 0 when 𝑡 = 0, then 𝐶 = 𝑉 arctanh 𝑉 = 0 and
solving for the velocity gives

𝑔𝑡
𝑣 = 𝑉 tanh ( ). (25)
𝑉

To get the position as a function of time, we integrate again:

 40

𝑔𝑡
∫ 𝑑𝑦 = ∫ 𝑉 tanh ( ) 𝑑𝑡.
𝑉

Since ∫ tanh 𝑢 𝑑𝑢 = ln cosh 𝑢 + 𝐷, we get

𝑉2 𝑔𝑡
𝑦= ln cosh ( ) + 𝐷.
𝑔 𝑉

Since cosh 0 = 1, and ln 1 = 0, one finds from the initial conditions that 𝐷 = 𝑦0 , so that

𝑉2 𝑔𝑡
𝑠 = 𝑦 − 𝑦0 = ln cosh ( ). (26)
𝑔 𝑉

3.4. More about drag

In the previous sections we have assumed two simple models for the drag force. For small particles moving
slowly the viscous drag is proportional to the speed, while for large objects moving fast the turbulent drag is
proportional to the square of the speed. One of these models will probably suffice for most problems from
an engineering perspective, but one should know which one to use for a particular problem. In addition,
neither gives the complete story. In fact the drag force, even for the ideal shape of a sphere moving through
a fluid, is actually a very complicated problem which is still worked on by physicists today, although mainly
from an academic perspective. Here we give a bit more detail of a more complete model. Books on
aerodynamics and hydrodynamics gives much more detail, where applications include such interesting and
important applications as aircraft and pipeline design.

3.4.1. Reynolds number

In all drag problems, one of the important parameters is the so-called Reynolds number, defined (for a
sphere4 moving in a fluid) as

𝜌
Re = 2𝑅𝑣 . (27)
𝜂

Note that the Reynolds number is unitless. If we compare the Reynolds number to the ratio of the turbulent
to viscous drag forces, we get

𝐶𝑑
𝑓𝑡 2 𝜌𝜋𝑅 2 𝑣 2 𝐶𝑑 𝜌 𝐶𝑑 Re
= = 𝑅𝑣 = Re =
𝑓𝑣 6𝜋𝜂𝑅𝑣 12 𝜂 24 48

1
where we have used 𝐶𝑑 = 2 for a sphere. This means that the Reynolds number gives an indication of the
relative sizes of these two types of drag forces: if Re ≈ 48 they are similar in size. For smaller Reynolds
numbers (small objects, moving slowly) the viscous drag is more important, while for larger Reynolds
numbers (large objects, moving fast) the turbulent drag dominates.

4
Its definition depends on the geometry of the problem e.g. if we had a cube instead of a sphere, we could not use a radius.
 41

3.4.2. Predicting what type of drag occurs

Therefore to predict the type of drag that dominates, one must calculate the Reynolds number for the
situation. For this one needs to know the viscosity 𝜂 and the density 𝜌 of the fluid. Values for air and water
under normal conditions are as follows:

Air Water Units

Viscosity (𝜂) 1.71 x 10-5 10-3 kg.m-1.s-1
Density (𝜌) 1.293 103 kg.m-3

Example: A cricket ball is hit through the air. Predict whether the viscous or turbulent drag force is more
important. Estimate or look up the necessary parameters for your calculation.

Solution: The radius of a cricket ball is about 3.6 cm. We will assume it travels through the air at 10 m/s. Then
1.293
from equation (27), using data from the table above for air, one gets Re = 2(0.036)(10) 1.71 × 10−5 ≈
50 000. This is three orders of magnitude greater than Re ≈ 48 where both types of drag forces are similar
in size, and turbulent drag is definitely more important in this problem. Even of the speed was chosen ten or
a hundred times less, this would still be the conclusion. Most everyday drag problems involving wind
resistance (e.g. streamlining of cars, or designing buildings to withstand strong winds, parachutists falling
from aeroplane) should be solved using the model of turbulent drag.

3.4.3. A unified model

𝐶𝑑
In the turbulent drag model 𝑓𝑡 = 2
𝜌𝜋𝑅 2 𝑣 2 (equation (19)) the drag coefficient 𝐶𝑑 is regarded as a constant
(just depending on the shape of the object, not its size or speed). If we are willing to regard 𝐶𝑑 as a variable,
and adjust it correctly, then we can get this formula to also predict viscous drag. Suppose we take

𝐶𝑑
𝑓= 𝜌𝜋𝑅 2 𝑣 2 (28)
2

to be the general drag formula. With 𝐶𝑑 constant, this will by definition predict turbulent drag. If we wish it
𝐶𝑑
to also work for viscous drag, then we require 2
𝜌𝜋𝑅 2 𝑣 2 = 6𝜋𝜂𝑅𝑣 (the latter being Stokes law, equation
(13)), so that

12𝜂 24
𝐶𝑑 = = .
𝜌𝑅𝑣 Re

24
Using the single formula of equation (28), with 𝐶𝑑 = Re when Re ≪ 48 (viscous drag region), and with 𝐶𝑑 =
1
when Re ≫ 48 (turbulent drag region) therefore combines the two models into a single model for the drag
2
force of a sphere. The price we pay for this unification is that the drag coefficient 𝐶𝑑 is no longer a constant
(and so one should beware when using the values in Figure 16). We can take equation (28) as actually defining
the drag coefficient, which is then in general a function of the Reynolds number. Many experiments have
been done to measure the drag coefficient of a sphere. The results are given in Figure 18, where the thin
solid line represents the predictions of the two simple models already considered.
 42

1E+4

1E+3
Figure 18
Drag coefficient of a sphere

1E+2
Drag coefficient versus Reynolds
number for a sphere. The circles are
some experimental results, while the
1E+1 line is the prediction of the unified
24
model with 𝐶𝑑 = for viscous drag
Re

1E+0
(small Reynolds numbers) and 𝐶𝑑 =
1
for turbulent drag (large Reynolds
2
numbers). Note the break-down of
1E-1
the turbulent drag model for very
high Reynolds numbers (around 3.8 x
1E-2
105.
1E-2 1E-1 1E+0 1E+1 1E+2 1E+3 1E+4 1E+5 1E+6 1E+7
Reynolds number

The viscous drag model works well for Re ≤ 1 while the turbulent drag model works well for 1000 ≤ Re ≤
3 × 105 . At higher Reynolds numbers there is a sudden drop in the drag coefficient – this interesting
phenomenon is discussed further in the next section.

3.4.4. Flow of a fluid past a sphere

Figure 19

The flow pattern of a fluid past a sphere, from top to bottom

for increasing Reynolds number. For very small Reynolds
numbers the flow pattern is symmetric around the sphere
with no vortices (loops), as shown in the top diagram. This
flow pattern does not change even when the viscous drag
model begins to deviate from experimental results near Re =
1.

Near Re = 25 the flow pattern changes to the second image,

where closed loop vortices form in the region behind the
sphere. These attached vortices grow in size as Re increases.

Near Re = 250 these vortices become periodically detached

from the region behind the sphere, forming a “Karman vortex
street” behind the sphere.

Near Re = 1000 the detaching vortices lose their pattern

resulting in a more turbulent wake. For this pattern the
turbulent drag model predicts the drag force well.

Finally near Re = 3.8 × 105 the turbulent wake behind the

sphere suddenly decreases in size. This is due to the onset of
turbulence in the thin boundary layer of fluid surrounding the
sphere. The smaller wake actually causes less drag, and so for
a while the drag actually decreases with increasing speed.
This is called the ‘drag crisis’.
 43

Figure 19 shows the flow patterns of fluid past a sphere for various ranges of Reynolds numbers. The diagram,
taken from the Feynman Lectures on Physics vol. 2, is actually for flow over a cylinder but the numbers have
been adapted here for a sphere and are only approximate. The unified model is a reasonable fit to the
experimental data but does not work well in the intermediate range of Reynolds numbers 1 < Re < 1000.
It is also clearly not accurate for Re > 3.8 × 105 where a sudden decrease of the drag coefficient occurs
during the ‘drag crisis’. Although not shown, there may also be a deviation from experiment for very small
Reynolds numbers as well. For tiny objects when the size of the object becomes comparable to the mean
free path of the molecules of the fluid, then the fluid cannot be considered a continuous medium. This is
taken into account by modifying (reducing) the value of the viscosity and hence changing the Reynolds
number (the so-called Cunningham correction). An example of where this is necessary is during the analysis
of Millikan's famous oil drop experiment, which uses tiny charged oil droplets to find the charge of an
electron.

Figure 20

The onset of the drag crisis occurs at a

lower Reynolds number for a rough
sphere than a smooth one. This means
that for a small range of Reynolds
numbers (speeds) a rough ball
experiences less drag than a smooth
one.

The drag crisis is a very interesting and counter-intuitive phenomenon, since few would have expected that
the drag force could ever decrease as the speed increases. Since it is caused by the onset of turbulence in the
thin boundary layer of fluid surrounding the sphere, it occurs more readily for a rough sphere for a smooth
one (Figure 20). For this reason a fast moving rough ball may actually experience less drag than a smooth
one. This is applied in the game of golf, where the dimples on the golf ball actually allow it to travel about
twice as far than a similarly struck smooth ball.

Finally, it is appropriate to mention that the drag forces discussed here assume that the object has constant
velocity (i.e. it is not accelerating) and is also not spinning. In either case the prediction of the drag becomes
more complicated, and spinning balls may also experience lift, which is important in many sports e.g. soccer,
tennis and ping-pong. Lastly, if the speed of the object moving through a fluid begins to approach the speed
of sound in the fluid, new complications occur.
 44

3.5. Problems

1. Assume that air resistance is proportional to speed. A parachutist falls from rest. If he did not open his
parachute, he would reach a terminal velocity of 90 m/s. With an open parachute, he reaches a terminal
velocity of 10 m/s. He opens his parachute 20 s after he started falling. Calculate (a) The distance already
fallen at the moment when he opens his parachute; (b) The total distance fallen 40 s after he started falling.

2. Assume that air resistance is proportional to speed. A particle is thrown directly downwards with initial
speed equal to twice its terminal velocity. Obtain a formula for the time it takes for its speed to reach three
quarters of its initial value.

3. Assume that air resistance is proportional to the square of the speed, and that an object is initially thrown
upwards with the magnitude of its terminal velocity. (a) Show that the height it reaches is ln 2 times the
height it would have reached if there was no air resistance. (b) Find the time that the object takes to fall from
its maximum height back to its initial height, and then show that its velocity as it arrives back at the initial
𝑉
height is .
√2

4. Generalize the result of the previous problem, showing that in the general case an object thrown upwards
with initial velocity 𝑣0 lands with velocity

𝑉
𝑣= .
2
√1 + (𝑣0 )
𝑉

√𝑢2 −1
Hint: You may use the result that tan arcosh 𝑢 = 𝑢
.

5. Consider projectile motion with air resistance taken into account. If the resistance is proportional to the
velocity, and the initial velocity had vertical and horizontal components 𝑣0𝑦 and 𝑣0𝑥 respectively, show that
the trajectory is given by

𝑉2 𝑔𝑥 𝑣0𝑦 + 𝑉
𝑦(𝑥) = ln (1 − )+( ) 𝑥.
𝑔 𝑣0𝑥 𝑉 𝑣0𝑥
 45

4. Hooke's law and Simple Harmonic Motion

4.1. Hooke's law

It is common knowledge that a spring pulls back when it is extended (stretched). It also pushes back when it
is compressed. Therefore the spring exerts a force when its length is changed from its natural length. The
model we shall use here is Hooke's law, which can be applied to most elastic materials as long as they are
not stretched or compressed too far.

Fs F0

Figure 21

Hooke’s law.

F0 Fs

Consider a spring that lies on a horizontal table (Figure 21). It is fixed at one end and has natural length ℓ.
Choose the origin at the loose end of the spring, with the X-axis horizontal and positive to the right. If we
apply an external force 𝐹0 to the right on the loose end of the spring, the spring stretches and the loose end
of the spring is displaced by a distance 𝑥 to the right. We call 𝑥 the extension of the spring. The extension is
directly proportional to the applied force i.e. 𝐹0 = 𝑘𝑥. The proportionality constant 𝑘 is called the spring
constant and depends on the type of material of which the spring is made and also its dimensions.

The loose end of the stretched spring can be held in equilibrium, and so the net force on it must be zero. This
means the spring itself exerts an equal but opposite force to the externally applied force. We call this the
spring force 𝐹𝑠 and

𝐹𝑠 = −𝐹0 = −𝑘𝑥. (29)

This is Hooke's law: the force exerted by a spring is proportional to its extension, but in the opposite direction.
The spring force is an example of a restoring force.
 46

4.2. Simple harmonic motion

An object attached to the free end of a spring (which is fixed at the other end) undergoes to-and-fro motion.
It is said that the object oscillates. If the spring obeys Hooke's law, the motion is called simple harmonic
motion, abbreviated by SHM.

Fs

Figure 22

Force diagram for

SHM.

Consider an object having mass 𝑚 connected to the loose end of a spring (with spring constant 𝑘) lying on a
smooth (frictionless) horizontal table. Take the origin to be at the equilibrium position of the object (when
the spring has its natural length). Now consider when the object is displaced from its equilibrium position
and released with initial velocity 𝑣0 from initial position 𝑥0 .

The only (horizontal) force acting on the object is the spring force 𝐹𝑠 = −𝑘𝑥. Therefore Newton's second law
applied to this object gives 𝑚𝑎 = −𝑘𝑥 or

𝑎 = −𝜔2 𝑥 (30)

where we define

𝑘
𝜔=√ . (31)
𝑚

Note that 𝜔 is always positive by definition.

4.2.1. Quick solution

We could simply state that the sinusoidal function

𝑥 = 𝐴 cos(𝜔𝑡 − 𝜑) (32)

is a solution to the problem. To check that this is indeed so, we can differentiate it once to get the velocity
𝑣 = −𝐴𝜔 sin(𝜔𝑡 − 𝜑) and then again to get the acceleration 𝑎 = −𝐴𝜔2 cos(𝜔𝑡 − 𝜑) . Clearly then
𝑎 = −𝜔2 𝑥 as required.

Figure 23 shows graphs of 𝑥 and 𝑣 as functions of 𝑡. Simple harmonic motion therefore is a sinusoidal to-and-
fro type of motion. It oscillates between 𝐴 and −𝐴. The amplitude, defined as the furthest displacement from
the equilibrium position, is therefore given by 𝐴. Note that occasionally books refer to a “peak-to-peak
amplitude” which is given by 2𝐴, but we will not use this. The argument or input of the cosine function i.e.
(𝜔𝑡 − 𝜑) is called the phase. The time independent part of the phase, in this case −𝜑, is called the phase
constant. The part that multiplies the time, in this case 𝜔, is called the angular frequency.
 47

𝑥
𝑥 = 𝐴 cos(𝜔𝑡 − 𝜑)
𝐴

Figure 23

Position and
𝑡 velocity as a
function of time.

𝑣 = −𝐴𝜔 sin(𝜔𝑡 − 𝜑)

We can substitute in initial conditions 𝑥 = 𝑥0 and 𝑣 = 𝑣0 when 𝑡 = 0 to get

𝑥0 = 𝐴 cos(−𝜑) = 𝐴 cos 𝜑 and 𝑣0 = −𝐴𝜔 sin(−𝜑) = 𝐴𝜔 sin 𝜑.

Working from these equations, we can find the amplitude and phase constant as follows:

𝑥 2 𝑣 2
 To get 𝐴, we use the fact that cos 2 𝜑 + sin2 𝜑 = ( 𝐴0 ) + (𝜔𝐴
0
) = 1, which we can solve for 𝐴 to get

𝑣0 2
𝐴 = √𝑥02 + ( ) . (33)
𝜔

 Dividing the second expression with the first gives

𝑣0
tan 𝜑 = . (34)
𝜔𝑥0

allowing one to find 𝜑. But note that this equation does not uniquely define 𝜑, since it always has two
possible solutions for 0 ≤ 𝜑 < 2𝜋. Only one of these is correct, and in the example that follows we shall
show how to select the correct one.

Example: An object of mass 𝑚 = 1 kg is attached to a spring with spring constant 𝑘 = 100 N.m-1. The spring
is compressed by 10 cm and then the object is released with an initial velocity towards the equilibrium
position of 1 m.s-1. Find an equation giving the displacement of the object at any time.

Solution: We know the displacement as a function of time has the form

𝑥 = 𝐴 cos(𝜔𝑡 − 𝜑), but we must find the amplitude, angular frequency and phase constant.

𝑘 100
 To get the angular frequency we use 𝜔 = √ = √ = 10 Hz.
𝑚 1

𝑣 2 1 2
 To get the amplitude we use 𝐴 = √𝑥02 + ( 𝜔0 ) = √(−0.1)2 + (10) = 0.14 m.
 48

𝑣 1
 To get the phase constant we use tan 𝜑 = 𝜔𝑥0 = (10)(−0.1) = −1. Now the tangent function is negative
0
3 7
in quadrants II and IV, so 𝜑 = 4 𝜋 or 𝜑 = 4 𝜋. But which of these is the correct one? To answer this, we
must consider one of the original equations we got by substituting in the initial conditions (the equations
just above equation (33)), say 𝑥0 = 𝐴 cos 𝜑. Since 𝑥0 is negative and 𝐴 is always positive, cos 𝜑 must be
negative. That is only possible in quadrants II and III, meaning that 𝜑 must lie in quadrant II. Therefore
3
𝜑 = 4 𝜋.

3
The displacement as a function of time is therefore 𝑥 = 0.14 cos (10𝑡 − 4 𝜋. ).

4.2.2. Long solution

The question remains of how one originally got to the solution 𝑥 = 𝐴 cos(𝜔𝑡 − 𝜑). Was it just a lucky guess?
No - we can solve the problem using our techniques of mechanics. Doing this hopefully gives more insight
into the solution. Equation (30) shows that we have a position dependent acceleration. For this type of
𝑑𝑣
problem we must express the acceleration as 𝑎 = 𝑣 𝑑𝑥, so

𝑑𝑣
𝑣 = −𝜔2 𝑥.
𝑑𝑥

Separation of variables yields 𝑣 𝑑𝑣 = −𝜔2 𝑥 𝑑𝑥. Integration then gives

1 2 𝜔2 2
𝑣 =− 𝑥 + 𝐶,
2 2

or 𝑣 2 = 2𝐶 − 𝜔2 𝑥 2 where 𝐶 is an integration constant. Since the object has velocity 𝑣 = 𝑣0 at position

𝑥 = 𝑥0 (when time 𝑡 = 0), substitution gives 𝑣02 = 2𝐶 − 𝜔2 𝑥02 , so that

𝑣0 2
𝑣 = ±𝜔√( ) + 𝑥02 − 𝑥 2 .
𝜔

This is a big and complicated expression, but we can neaten it up by noting that the first two terms under the
𝑣 2
square root are just constants. Suppose we define 𝐴 = √( 0 ) + 𝑥02 (which is identical to equation (33) from
𝜔
the previous section), then

𝑣 = ±𝜔√𝐴2 − 𝑥 2 . (35)

The velocity as a function of the position of the object, given in Figure 24, is an ellipse intersecting the position
axis at ±𝐴 and the velocity axis at ±𝜔𝐴. From this graph, it is clear that the position must always lie between
±𝐴 , so 𝐴 represents the amplitude, while the velocity must always lie between ±𝜔𝐴 . Note that the
relationship is not a true “function” in the mathematical sense since there are two (equal but opposite)
velocities corresponding to every position. These velocities represent the particle moving either forwards or
backwards through the point as it oscillates. As time passes the point on the graph corresponding to the state
of the particle rotates clockwise around the ellipse. To see this, suppose 𝑥 > 0: then the acceleration of the
 49

particle is negative and the velocity must therefore be decreasing. If 𝑥 < 0 then the acceleration is positive
and the velocity increases.

𝑣
𝜔𝐴

Figure 24
𝑥
𝐴 Velocity as a function of position.

𝑑𝑥
To find the position as a function of time, we can substitute 𝑣 = 𝑑𝑡
in equation (35) and separate the
𝑑𝑥
variables to get ± = 𝜔𝑑𝑡. If the plus sign is used, integration yields
√𝐴2 −𝑥 2

𝑥
arcsin = 𝜔𝑡 + 𝜑 or 𝑥 = 𝐴 sin(𝜔𝑡 + 𝜑).
𝐴

Here 𝜑 is the integration constant. But if the minus sign is used, integration yields

𝑥
arccos = 𝜔𝑡 + 𝜑 or 𝑥 = 𝐴 cos(𝜔𝑡 + 𝜑).
𝐴
𝑥 𝑥
The reader is likely to be surprised that the integral on the left is not given as −arcsin 𝐴, but rather arccos 𝐴.
𝜋
That would not be wrong. However, arcsin 𝑢 + arccos 𝑢 = 2
and so these two functions only differ by a
constant and either can be used. The point is that the sine and cosine functions give exactly the same family
of curves when the phase constant 𝜑, which is the integration constant, can take on any value,5 Therefore
the result of integration is exactly the same whether we used the plus or minus sign, and either version found
can be used as the solution. For consistency we will in future always use the cosine function (but with a minus
sign in front of the phase constant), as follows:

𝑥 = 𝐴 cos(𝜔𝑡 − 𝜑) with 𝐴 > 0 and 0 ≤ 𝜑 < 2𝜋. (36)

This is the same as equation (32) given originally.

5
The mathematical reader may wish to note that there are actually even more possible (equivalent) ways of writing the solution,
obtained by taking the ±-sign to the right hand side before integrating. We will not complicate matters by considering these.
 50

4.3. Physical quantities in SHM

4.3.1. Period and frequency

𝑥 𝜑
𝜔

𝑇
𝐴

𝑥 = 𝐴 cos(𝜔𝑡 − 𝜑)
−𝐴

Figure 25 Phase constant.

Figure 25 shows the position as a function of time in more detail. The cosine function has a period of 2𝜋. The
motion 𝑥 = 𝐴 cos(𝜔𝑡 − 𝜑) consists of a cosine function that has been compressed by a factor 𝜔 and then
shifted to the right by an angle 𝜑. Only the compression affects the period, which is given by

2𝜋
𝑇= .
𝜔
𝜑
Looking at the graph, the reader may be surprised to see that the shift to the right is by an amount 𝜔 and not
𝜑
𝜑. This is because the shift of 𝜑 applies to 𝜔𝑡, making the shift in 𝑡 only 𝜔.

The frequency is the number of oscillations executed per unit time and is given by

1 𝜔
𝑓= = .
𝑇 2𝜋

From this it follows that the angular frequency is just 2𝜋 times the (normal) frequency i.e. 𝜔 = 2𝜋𝑓. Whereas
the frequency gives the number of oscillations per unit time, the angular frequency gives the angle (in
radians) by which the phase increases per unit time. The factor of 2𝜋 between angular frequency and
(normal) frequency comes from the fact that one full oscillation corresponds to an angle of 2𝜋 radians. The
units of these quantities are also interesting:

 The unit for frequency is “oscillations (or cycles) per second”. This is generally shortened to “per
second” i.e. s-1 or Hertz (Hz)
 The unit for angular frequency is “radians per second”. But an angle in radians is a ratio of two lengths
(arc length to radius) and is therefore actually unitless! So the unit for angular frequency may also
shortened to “per second” i.e. s-1 or Hertz (Hz).
 51

Although these two frequencies have the same units, they measure different things and it is important to
always be clear which one is meant. This should be clear from the symbol used i.e. 𝜔 for angular frequency
and 𝑓 for (normal) frequency. By convention we will use units of rad/s for angular frequencies and Hz for
normal frequencies.

4.3.2. Energy

The total energy of a simple harmonic oscillator at any moment is the sum of its kinetic and potential energy.
Its kinetic energy is given by

1 1 1
𝐸𝑘 = 𝑚𝑣 2 = 𝑚{−𝐴𝜔 sin(𝜔𝑡 − 𝜑)}2 = 𝑚𝜔2 𝐴2 sin2(𝜔𝑡 − 𝜑).
2 2 2

Its potential energy is given by

1 1 1
𝐸𝑝 = 𝑘𝑥 2 = 𝑘{𝐴 cos(𝜔𝑡 − 𝜑)}2 = 𝑘𝐴2 cos2(𝜔𝑡 − 𝜑).
2 2 2

𝑘 1
But recall that by definition 𝜔 = √ , so we can write the kinetic energy as 𝐸𝑘 = 𝑘𝐴2 sin2 (𝜔𝑡 − 𝜑) and
𝑚 2
therefore the total energy is

1 1
𝐸 = 𝐸𝑘 + 𝐸𝑝 = 𝑘𝐴2 {sin2 (𝜔𝑡 − 𝜑) + cos 2(𝜔𝑡 − 𝜑)} = 𝑘𝐴2
2 2

which is independent of the time i.e. constant, and it depends only on its spring constant and amplitude.

4.4. A mass hung on a spring

At the beginning of this chapter, we considered an object attached to a spring and sliding horizontally over a
frictionless surface. The only horizontal force acting on the object was the spring force, and the weight of the
object was of no importance. But suppose the object is hung from a spring in a vertical orientation. Then
both the spring force and the weight act vertically on the object. Suppose the position is given by 𝑦 and 𝑦 =
0 when the spring has its natural length, with the downwards direction chosen as positive. Then applying
Newton's second law gives

𝑚𝑎 = −𝑘𝑦 + 𝑚𝑔

where the two forces are the spring force and the weight respectively. Dividing throughout by 𝑚 gives

𝑎 = −𝜔2 𝑦 + 𝑔

which is similar to 𝑎 = −𝜔2 𝑥 for the horizontal case, but it has the additional term 𝑔 on the right. One could
proceed to solve this differential equation mathematically, but a good guess can save a lot of effort. In the
horizontal case the solution was 𝑥 = 𝐴 cos(𝜔𝑡 − 𝜑), so we try a solution for the vertical case of 𝑦 =
𝐴 cos(𝜔𝑡 − 𝜑) + 𝑐 where 𝑐 is some constant. Then 𝑣 = −𝐴𝜔 sin(𝜔𝑡 − 𝜑) and 𝑎 = −𝐴 𝜔2 cos(𝜔𝑡 − 𝜑).
Substituting these gives
 52

−𝐴 𝜔2 cos(𝜔𝑡 − 𝜑) = −𝜔2 {𝐴 cos(𝜔𝑡 − 𝜑) + 𝑐} + 𝑔,

𝑔 𝑚𝑔
or 𝑐 = 𝜔2 = 𝑘
, which is precisely how much the spring stretches under the weight. Therefore the vertical
motion is given by

𝑚𝑔
𝑦 = 𝐴 cos(𝜔𝑡 − 𝜑) +
𝑘

where the oscillations above and below the equilibrium point of the stretched spring. The amplitude and
angular frequency are identical whether the situation is vertical or horizontal.

4.5. Effect of the mass of the spring

Up to now we have ignored the mass of the spring, or assumed it to be negligible. But during oscillations the
coils of the spring move to-and-fro and this requires energy. If the entire spring moved along with the mass,
we could just add the mass of the spring to the mass of the object and get the total mass of the system. But
not all parts of the spring move the same amount. The coils near the fixed end hardly move at all, while the
coils near the end attached to the object move just like the object. One might guess that only half the mass
of the spring needs to be taken into account, but we will show that actually one third of the mass of the
spring should be added to the attached mass in order to get the effective mass of the system.6

Figure 26 Motion of the spring together with the attached mass.

Suppose that the fixed end of the spring is at the origin, while the other end of the spring has been displaced
by a distance 𝑥 from the equilibrium position to give it a total extended length 𝐿 as shown in Figure 27. Also
assume that this other end of the spring has velocity 𝑣 to match the speed of the attached mass. Now
𝑑𝑢
consider a short piece of the spring 𝑑𝑢 at a distance 𝑢 from the origin. This has mass ( ) 𝑚𝑠 where 𝑚𝑠 is
𝐿
the mass of the entire spring. If we assume that the stretching is uniform, then 𝑑𝑢 has been shifted
𝑢 𝑢
(displaced) from its equilibrium position by a distance ( 𝐿 ) 𝑥 and will have velocity ( 𝐿 ) 𝑣. The piece 𝑑𝑢 of the

6
The problem of the effective mass of a spring is actually rather complicated and the reader may consult “An improved calculation
of the mass for the resonant spring pendulum” by Christensen (American Journal of Physics, vol 72 (2004) 818) and the references
therein, as well as “Effective mass of an oscillating spring” by Rodrigues and Gesnouin (Physics Teacher, vol 45 (2007) 100) for further
information.
 53

1 𝑑𝑢 𝑢 2 1𝑚
spring therefore has kinetic energy 𝑑𝐸𝑘 = 2 {( 𝐿 ) 𝑚𝑠 } {( 𝐿 ) 𝑣} = 2 𝐿3𝑠 𝑣 2 𝑢2 𝑑𝑢 and the total kinetic energy
of the spring is

1 𝑚𝑠 2 𝐿 2 1 𝑚𝑠 2 𝐿3 1 1
𝐸𝑘 = 3
𝑣 ∫ 𝑢 𝑑𝑢 = 3
𝑣 ( ) = ( 𝑚𝑠 𝑣 2 ).
2𝐿 0 2𝐿 3 3 2

1
If the whole spring was moving at speed 𝑣, it would have had kinetic energy 2 𝑚𝑠 𝑣 2 , but because one end is
fixed it only has one third of this kinetic energy. This suggests that even though each part of the spring is
moving at a different speed, we can treat it as if one third of its mass is moving at the same speed as the
𝑚𝑠
attached mass and the remaining two thirds as stationary. Hence only one third of the spring's mass, ,
3
should to be added to the attached mass, 𝑚, to get the total effective mass of the system.

Unless stated otherwise, you may assume in problems that the mass of the spring is negligible.

4.6. Rotational systems

Consider an object suspended from a spring. Now suppose the object is twisted and released (without
changing its height) – then it will then oscillate in a rotational manner. This rotational simple harmonic motion
is very similar to what we have encountered up to know, but there are some important differences. Instead
of forces, we must deal with torques (twisting forces). Newton's second law, in rotational form, is

𝜏 = 𝐼𝛼 (37)

where 𝜏 is the torque, 𝐼 is the moment of inertia and 𝛼 is the angular acceleration. Note that the torque takes
the place of force and the moment of inertia takes the place of mass in a rotating system. Hooke's law for
rotations is that if an object is twisted then the restoring torque is proportional to the angle twisted, but in
the opposite direction:

𝜏 = −𝜅𝜃. (38)

Note that we have used 𝜅 instead of 𝑘 for the angular spring constant of a rotational system. For our twisted
system, we can combine Newton's second law with Hooke's law to get 𝐼𝛼 = −𝜅𝜃 or

𝛼 = −𝜔2 𝜃 (39)

𝜅
where we define 𝜔 = √ 𝐼 . The reader should check that 𝜔 has the correct units of rad/s. Equation (39) has
exactly the same form as equation (30), and therefore has the analogous solution

𝜃 = 𝜃𝑚 cos(𝜔𝑡 − 𝜑) (40)

where now the amplitude 𝜃𝑚 is an angle.

In the case of linear oscillations of a mass on a spring we could add one third of the mass of the spring to get
the total effective mass. Similarly one third of the moment of inertia of the spring can be added to the
moment of inertia of the object for a rotational system.
 54

4.7. Problems

1. Use Hooke's law to derive the expression for the potential energy stored in a spring with spring constant
𝑘 in terms of the extension 𝑥. Remember that potential energy due to a position-dependent force 𝐹[𝑥] is
𝑥
given by 𝐸𝑝 = ∫0 𝐹[𝑥]𝑑𝑥.

2. Suppose 𝑦 = 10 cos(5𝑡 + 1). What is the amplitude, phase, phase constant, angular frequency, period
and frequency?

3. A mass is attached to a spring, and released from rest at position 𝑥0 . (a) Does the amplitude of the motion
depend on any property of the spring? (b) Now suppose the mass is released with initial velocity 𝑣0 from
position 𝑥0 . Does the amplitude now depend on the spring's properties – motivate your answer. (c) The same
mass is released just as in the previous question, except that the initial velocity is now in the opposite
direction. Will the amplitude be greater or smaller (or the same)?

4. A spring with spring constant 𝑘 = 28 N/m has a mass of 7 kg attached to it. The mass is released at 𝑥0 =
12 cm with 𝑣0 = −10 cm/s. (a) What will be the angular frequency of the resulting simple harmonic motion?
(b) Calculate the period and frequency. (c) Taking 𝑥 = 𝐴 cos(𝜔𝑡 − 𝜑), differentiate it to get an expression
for the velocity and then substitute the initial conditions to get the amplitude 𝐴 and phase constant 𝜑. (d)
Calculate the energy of the system.

5. An object of mass 4 kg is attached to a spring. The spring is stretched 12 cm and released from rest. The
object oscillates with a period of 1.26 s. (a) Calculate the angular frequency and the spring constant. (b)
Calculate the energy of the system. (c) Using the law of conservation of energy, calculate the speed of the
object as it passes the equilibrium position. (d) Obtain an expression for the position as a function of time.
Then calculate how long it takes the object, after being released from rest, to move from its initial position
𝑥0 = 12 cm to 𝑥 = 4 cm.

6. A spring is suspended in a vertical position. With a mass of 20 g attached to the lower end of the spring,
its length is 12.0 cm. When an additional 20 g is added, its length becomes 12.8 cm. (a) Find the natural length
of the spring, the spring constant and calculate the oscillation frequency of the system when only 20 g is
attached to the spring. Take 𝑔 = 9.8 m/s2. (b) What will the frequency of this system be on the Moon, where
𝑔 = 1.6 m/s2? (c) We want the system to oscillate with a frequency of 6 Hz. What mass must be attached to
the spring?

21
7. Prove that the total energy of an object rotating with simple harmonic motion is given by 𝐸 = 𝜅𝜃𝑚 . You
2
𝜃
can derive an equation for the potential energy using 𝐸𝑝 = ∫0 𝜏[𝜃]𝑑𝜃 and the rotational kinetic energy is
1 𝑑𝜃 2
given by 𝐸𝑘 = 2 𝐼 ( 𝑑𝑡 ) .
 55

5. Examples of oscillating systems

5.1. Introduction

In the previous chapter we looked at the theory of simple harmonic motion for rectilinear and rotational
motion. The vital aspect was that when a system was disturbed from its equilibrium position, then a linear
restoring force (given by Hooke's law) acted to try and return the system to its equilibrium position. We will
now look at some other examples of oscillating systems. All these systems have only one degree of freedom
i.e. only one quantity or variable is necessary to describe the state of the system. In all these systems we will
ignore friction – of course, this will cause the systems to lose energy and the oscillations to become smaller
until they cease, but that will be dealt with in another chapter.

One may wonder why so many different systems can be described using Hooke's law i.e. why is it so effective
and important? The potential energy of an system with one degree of freedom is a function of its state i.e.
𝐸𝑝 = 𝐸𝑝 (𝑥) and we can expand this as a Taylor series

𝑑𝐸𝑝 𝑑2 𝐸𝑝 𝑥 2 𝑑3 𝐸𝑝 𝑥3
𝐸𝑝 (𝑥) = 𝐸𝑝 (0) + | 𝑥+ | + | + ⋯.
𝑑𝑥 𝑥=0 𝑑𝑥 2 𝑥=0 2! 𝑑𝑥 3 𝑥=0 3!

Now suppose that position 𝑥 = 0 corresponds to an equilibrium position of the system i.e. a position of
𝑑𝐸𝑝
minimum potential energy. Then from calculus | = 0, and the second term is zero. In addition, we are
𝑑𝑥 𝑥=0
free to choose a reference level for the potential energy, so we can choose 𝐸𝑝 (0) = 0. This leaves

𝑑2 𝐸𝑝 𝑥 2 𝑑3 𝐸𝑝 𝑥3
𝐸𝑝 (𝑥) = | + | + ⋯.
𝑑𝑥 2 𝑥=0 2! 𝑑𝑥 3 𝑥=0 3!

1
As long as 𝑥 is very small (i.e. for small displacements) the first term dominates and 𝐸𝑝 (𝑥) = 2 𝑘𝑥 2 where
𝑑 2 𝐸𝑝
𝑘= 𝑑𝑥 2 𝑥=0
| . But this is precisely the potential energy of a spring obeying Hooke's law! Therefore all systems,
when disturbed only a little from their equilibrium position, experience a restoring force which has the form
of Hooke’s law.

Of course, if 𝑥 becomes large then the other terms may become important, and then we say the system is
𝑑 2 𝐸𝑝
non-linear. Also if a system is engineered so that 𝑘 = 𝑑𝑥 2 𝑥=0
| ≈ 0, it will also be non-linear.

5.2. The simple pendulum

Galileo is said to have been watching the swinging chandeliers in church, and keeping time using his pulse or
heartbeat, when he noticed that the period of the oscillations did not depend on the amplitude. As we shall
 56

see that is only approximately true, but given Galileo's primitive apparatus and his probable excitement, he
should be forgiven for the small inaccuracy.7

ℓ
ℓ
Figure 28
Tension 𝑇
Simple pendulum.

Weight 𝑊 = 𝑚𝑔
𝑥 = ℓ𝜃

A simple pendulum consists of a mass 𝑚 attached to the end of a string (which cannot stretch and has
negligible mass). The string is tied to a fixed point and the mass hangs below it. If it is disturbed, it can swing
in the arc of a circle having a radius equal to the length of the string, say ℓ. Figure 28 shows the setup of a
simple pendulum with the string making an angle 𝜃 with the vertical. The distance that the “bob” or mass
has moved along the arc from its equilibrium position is 𝑥. This system is interesting in that we can analyze
it as a rectilinear or rotational system, depending on the variable 𝑥 or 𝜃 we choose to represent the state of
the system. These two variables are related to one another through 𝑥 = ℓ𝜃 i.e. they are directly proportional
to one another.

Consider a pendulum with the bob displaced a distance 𝑥 along the arc from the lowest position. Then two
forces act on the bob, namely its weight and the tension in the string. The tension acts perpendicular to the
arc, but the weight has a tangential component – 𝑚𝑔 sin 𝜃 which acts as a restoring force. If the angle is small
𝑥 𝑥
then we may approximate sin 𝜃 ≈ 𝜃 = ℓ and so the restoring force is 𝐹 ≈– 𝑚𝑔 ℓ . Comparing this to Hooke’s
𝑚𝑔
law 𝐹 = −𝑘𝑥 one gets that 𝑘 ≈ ℓ
and

𝑚𝑔
𝑘 √ ℓ 𝑔
𝜔=√ ≈ =√ .
𝑚 𝑚 ℓ

There is not really a spring in the system, but it acts as if the other forces in the system could be replaced by
a spring force.

7
Galileo's observations were done around 1600. In 1656 Huygens patented the pendulum clock based on this principle. Pendulum
clocks provided the most accurate measurement of time until the 1920s when the first quartz watches were introduced. Today these
have been superseded by atomic clocks, and even more accurate nuclear clocks have been proposed. The measurement of time
continues to be an extremely important part of the study of physics.
 57

Example: Derive the same result as above, but considering the pendulum as an angular rotation system.

Solution: Consider a pendulum with the string making an angle 𝜃 with the vertical. The moment of inertia of
the bob is given by 𝐼 = 𝑚ℓ2 . Then two forces act on the bob, namely its weight and the tension in the string.
The tension produces no torque, but the weight produces a restoring torque – 𝑚𝑔ℓ sin 𝜃. If the angle is small
then we may approximate sin 𝜃 ≈ 𝜃 and so the restoring torque is 𝜏 ≈ −𝑚𝑔ℓ𝜃. Comparing this to 𝜏 = −𝜅𝜃
one gets that 𝜅 ≈ 𝑚𝑔ℓ and

𝜅 𝑚𝑔ℓ 𝑔
𝜔=√ ≈√ 2 =√ .
𝐼 𝑚ℓ ℓ

The period of the pendulum is therefore

2𝜋 ℓ
𝑇= ≈ 2𝜋√ .
𝜔 𝑔

Notice that (as Galileo observed) the period depends only on the length of the string, and not on the
amplitude of the swing nor the mass attached. But of course we have used the small angle approximation
1
sin 𝜃 ≈ 𝜃. For large angles a better approximation is sin 𝜃 ≈ 𝜃 − 6 𝜃 3 and as a result of the additional 𝜃 3
term the restoring force (or torque) is no longer linear, making the motion not perfectly sinusoidal.8

5.3. The compound pendulum

Figure 29

Compound pendulum.

A compound pendulum (Figure 29) consists of any rigid object hanging freely, pivoted at a fixed point 𝑃.
Suppose the centre of mass 𝐶 is a distance 𝑑 from the pivot point 𝑃. The compound pendulum is most easily
modelled as a rotational system. In the equilibrium position 𝐶 will lie directly below 𝑃, but suppose the

8
Despite the nonlinearity, the motion is still periodic! A better approximation of the period than we have given was found by
ℓ 2
𝜃𝑚
Bernoulli in 1749 to be 𝑇 ≈ 2𝜋√ (1 + ) which shows that it increases with increasing angular amplitude. Exact expressions for
𝑔 16
the oscillations and their period (as a function of the amplitude) have been found but are mathematically complicated and we will
not take the matter further here. The interested reader can consult “Exact solution for the nonlinear pendulum” by Belendez et al.
(Revista Brasileira de Ensino de Fisika vol. 29 (2007) 645) and “An accurate formula for the period of a simple pendulum oscillating
beyond the small angle regime” by Lima and Arum (American Journal of Physics vol. 74 (2006) 892).
 58

system has been tilted through an angle 𝜃. The reaction force 𝑅 acts at the pivot and can therefore not cause
any torque around the pivot point. The restoring torque caused by the weight acting at the centre of mass is
given by 𝜏 =– 𝑚𝑔𝑑 sin 𝜃. If the angle is small then we may approximate sin 𝜃 ≈ 𝜃 and so the restoring
torque is 𝜏 ≈– 𝑚𝑔𝑑𝜃. Comparing this to 𝜏 = −𝜅𝜃 one gets that 𝜅 ≈ 𝑚𝑔𝑑 and

𝜅 𝑚𝑔𝑑
𝜔=√ ≈√ .
𝐼 𝐼

Example: A straight rod of uniform density is pivoted at one end and allowed to swing with small oscillations.
What would the angular frequency be? Use your result to find the period of a metre stick pivoted at one end.

1
Solution: The distance 𝑑 from the pivot point to the centre of mass is half the length of the rod i.e. 𝑑 = 2 ℓ.
1
The moment of inertia of a rod rotating about its end is 𝐼 = 𝑚ℓ2. Therefore the angular frequency is
3

1
𝑚𝑔𝑑 𝑚𝑔 ( ℓ)
𝜔≈√ =√ 2 = √3𝑔.
𝐼 1 2 2ℓ
3 𝑚ℓ

For a metre stick 𝜔 = 3.834 rad/s and so 𝑇 = 1.639 s.

5.4. Combinations of springs

If a system contains an object attached to more than one spring, then the effective spring constant of the
combination can be calculated as follows: when a force 𝐹 is applied to the object, it will be displaced from its
equilibrium position by an amount 𝑥, and the effective or resultant spring constant is given by

𝐹
𝑘eff = . (41)
𝑥

5.4.1. Springs in series

Figure 30

Springs connected in series.

Suppose two springs with spring constants 𝑘1 and 𝑘2 are connected in series, as in Figure 30. Now any force
𝐹 𝐹
𝐹 that is applied will act on both springs. Their extensions will be 𝑘 and 𝑘 respectively, so the total extension
1 2
𝐹 𝐹
will be 𝑥 = 𝑘 + 𝑘 . Substituting this into equation (41) and simplifying gives
1 2

1 1 1
= + . (42)
𝑘eff 𝑘1 𝑘2
 59

If there are more than two springs in series, further similar terms can be added on the right hand side. Note
that the effective spring constant is smaller than any of the original springs, as the combination can be more
easily stretched.

5.4.2. Springs in parallel

Figure 31

Springs connected in parallel.

Suppose two springs with spring constants 𝑘1 and 𝑘2 are connected in parallel, as in Figure 31. Now any force
𝐹 that is applied will be shared between the two springs, which will both have the same extension (say 𝑥).
The force applied to the first spring will be 𝐹1 = 𝑘1 𝑥 and the force applied to the second spring will be 𝐹2 =
𝑘2 𝑥. Together these will equal the applied force, so 𝐹 = 𝑘1 𝑥 +𝑘2 𝑥. Substituting this into equation (41) and
simplifying gives

𝑘eff = 𝑘1 + 𝑘2 . (43)

If there are more than two springs in parallel, further similar terms can be added on the right hand side. Note
that the effective spring constant is greater than any of the original springs, as the combination is more
difficult to stretch.

5.5. Elongation of a rod

Consider a mass hung not from a spring, but attached to the bottom of a rod of length ℓ having cross-
sectional area 𝐴 suspended securely from the ceiling. If the rod is stretched by pulling its end downwards
from its equilibrium position there will be a restoring force upwards, and if the rod is compressed by pushing
its end upwards there will be a restoring force downwards. What we wish to find out is how this restoring
force depends on the properties of the rod.

When Einstein was considering the problem of the heat capacity of a solid, he made the simplifying
assumption that the solid can be modelled as tiny atoms each connected to their nearest neighbours by
springs to represent the interatomic forces. We can justify this by saying that all restoring forces, including
those of the chemical bonds between atoms, obey Hooke’s law if the displacement is small. We can use the
same model now to consider the elongation of a rod.

Firstly, consider a chain of atoms running along the rod from the ceiling to its free end below. In our model
this chain consists of a huge number of identical springs joining the atoms which are connected in series. For
1 1 1 1
many springs in series = + + + ⋯, and so for 𝑛 identical springs of spring constant 𝑘 in series,
𝑘eff 𝑘1 𝑘2 𝑘3
 60

1 1 𝑘
= 𝑛 ( ) so that 𝑘eff = . Therefore the effective spring constant of a row or chain of atoms along the
𝑘eff 𝑘 𝑛
length of the rod will decrease in proportion to the number of atoms in the row i.e. with the length of the
rod.

But there is not just one chain of atoms running along the rod, but many. The number of such chains will be
directly proportional to the cross-sectional area of the rod. Each chain will act in parallel with the others, and
since for many springs in parallel 𝑘eff = 𝑘1 + 𝑘2 + 𝑘3 + ⋯ the effective spring constant will be proportional
to the area of the rod.

Taking these results together, one finds that for any solid rod (with constant cross-sectional shape, e.g. a
cylindrical or square rod) our model predicts that the effective spring constant has the form

𝐴
𝑘=𝐸 (44)
ℓ

where 𝐴 is its cross-sectional area and ℓ is its length, while 𝐸 is a constant that depends only on the type
material itself, but does not depend on its size or shape. It is called Young's modulus or the elastic modulus
and represents how much the material resists being stretched or compressed i.e. how difficult is it to change
the distance between the atoms. Values of Young's modulus have been tabulated for many materials in
reference books.

Example: Suppose a mass of 20 kg is suspended from a square aluminium bar having a length of 10 m and
edges 5 mm wide. If the elastic modulus of aluminium is 50 GPa, calculate the frequency when the mass is
struck vertically and oscillates.

𝐴
Solution: The effective spring constant is 𝑘 = 𝐸 = (50 x 109)(5 x 10-3)2/10 = 125 000 N/m. Therefore 𝜔 =
ℓ
𝑘 125 000 𝜔
√ =√ = 79 rad/s, giving a frequency 𝑓 = 2𝜋 = 13 Hz. Our solution neglects the mass of the
𝑚 20
aluminium bar itself: one third of this mass could be added for a more accurate answer.

5.6. Twisting of a rod

If a rod is attached to the ceiling and it is twisted at the base, there will be a restoring torque proportional to
the angle twisted (at least for small angles), so one expects that 𝜏 = −𝜅𝜃. What we wish to know is how the
angular spring constant 𝜅 depends on the properties of the rod. This is not as easily derived as the result 𝑘 =
𝐴
𝐸 for the stretching of a rod in section 5.5, but the result is similar, namely
ℓ

𝐽
𝜅=𝐺 . (45)
ℓ

𝐽 is called the torsion constant and depends on the cross-section of the rod, while ℓ is the beam length and
𝐺 is a constant that depends only on the material itself, but does not depend on its size or shape. It is called
the shear modulus and indicates how much the material resists being twisted. Values for many materials
have been tabulated in reference books. In fact, for an isotropic material the shear modulus 𝐺 and Young's
𝐸
modulus 𝐸 of a material are related to one another by 𝐺 = 2(1+𝜈) where 𝜈 is Poisson's ratio. For most
1 3
materials 𝜈 ≈ and so 𝐺 ≈ 𝐸.
3 8
 61

For a circular rod the torsion constant 𝐽 is equal to the polar moment 𝐼𝑝 defined by 𝐼𝑝 = ∫ 𝑟 2 𝑑𝑎 where the
distance 𝑟 from the rotation axis is integrated over the cross-sectional area of the rod, and having the value
𝜋𝑅4
2
where 𝑅 is the radius. For other shapes the polar moment 𝐼𝑝 does not give the exact value of the torsion
constant 𝐽, which is generally smaller due to warping of a cross-sectional plane as the rod is twisted. For a
rod of rectangular cross-section (longer side 𝑎 and shorter side 𝑏) a good approximation is:

1 𝑎 21 7 𝑏 4
𝐽 = 𝑏4 { ( ) − + ( ) }. (46)
3 𝑏 100 400 𝑎

Example: Suppose a circular plate with mass 5 kg and radius 15 cm has a steel rod welded to its centre point
perpendicular to the surface. If the rod is 2 m long and has a diameter of 10 mm, and is attached firmly to
the ceiling, what would the period of oscillations be when the circular plate is rotated slightly and released?
The shear modulus of steel is 75 GPa.

𝜅
Solution: If we can calculate the angular frequency from 𝜔 = √ , it will be easy to get the period. We need
𝐼
to determine the angular spring constant of the rod and the moment of inertia of the plate.

4
𝜋𝑅4 𝜋(5×10−3 )
 The angular spring constant is 𝜅 = 𝐺 2ℓ
= (75 × 109 ) 2(2)
= 36.8 N/m.
 The moment of inertia of a circular plate rotating around an axis perpendicular to its surface and through
1 1
its midpoint is given by 𝐼 = 2 𝑀𝑅 2 = 2 (5)(0.15)2 = 0.05625 kg.m2.

36.8 2𝜋
Using these results 𝜔 = √ = 25.7 rad/s. This corresponds to a period 𝑇 = = 0.25 s. Our solution
0.05625 𝜔
neglects the moment of inertia of the steel rod itself – one third of this could be added for a more accurate
answer, but due to the small radius of the rod its mass is probably negligible.

5.7. Real springs

Real springs are actually rather complicated. When a standard coil spring is stretched, the wire from which it
is made is not stretched, but instead the wire of the spring uncoils as it twists slightly. Therefore stretching
of the spring is associated with a twisting deformation, and so it is the shear modulus rather than the Young's
modulus of the spring material which influences its spring constant. The spring constant of a normal helical
spring (as in, made from wire with circular cross section) is

𝑑4
𝑘=𝐺 (47)
8𝑁𝐷 3

where 𝐺 is the shear modulus, 𝑑 is the diameter of the wire, 𝐷 is the diameter of the spring and 𝑁 is the
number of turns in the spring. Note that the spring constant is inversely proportional to its length.
 62

Figure 32

Helical spring.

Although we normally associate a spring with stretching and compression, springs are also often used for
rotational systems. Common examples are the spring in a clothes peg or a mousetrap or in wind-up toys (and
old mechanical wristwatches). Such springs are called torsional springs. If one end of a spring is fixed, and
the other end is rotated (twisted), then the wire from which it is made must stretch if the spring is not to
deform in shape. Actually, the wire of the spring usually does deform rather than stretch and so it is best
analysed similar to a bending beam. For this reason Young's modulus (𝐸) rather than the shear modulus (𝐺)
occurs in the expression for the angular spring constant, namely

𝐼𝑏
𝜅=𝐸 (48)
ℓ

where 𝐼𝑏 is the beam moment of the wire (see section 5.8.1) and ℓ is the total length of wire in the spring.

5.8. Bending of beams

A beam can be supported in a variety of ways. If it is held completely fast at one end (e.g. built into a wall)
but left free at the other end it is called a cantilever. We will consider this case in detail and then look at some
other ways the beam might be supported and how the results change. Our results are applicable for beams
lying horizontally or vertically (or any other orientation), although for horizontal beams we will ignore the
initial sagging of the beam due to gravity. We will also assume the oscillations are relatively small. For large
vibrations the results will be inaccurate due to non-linear effects.

5.8.1. The cantilever

Consider a beam built into a wall and projecting horizontally out of it. The vertical force required to deflect
the end of the beam upwards or downwards by a distance 𝑦 from its equilibrium position can be shown to
3𝐸𝐼𝑏
be 𝐹 = ℓ3
𝑦 where 𝐸 is Young’s modulus, ℓ is the length of the beam and 𝐼𝑏 is called the moment of the
beam. This is different from the moment of inertia 𝐼 or polar moment 𝐼𝑝 and is given by the integral 𝐼𝑏 =
∫ 𝑦 2 𝑑𝑎 over the cross-section of the beam where 𝑦 is the distance of 𝑑𝑎 from the neutral axis i.e. the middle
part of the bent beam which does experience either compression or dilation:

𝑤𝑡 3
 For a rectangular beam 𝐼𝑏 = 12
where 𝑤 is the width and 𝑡 is the thickness.
 63

𝜋𝑑 4
 For a circular beam the moment is given by 𝐼𝑏 = 64
where 𝑑 is the diameter.

The effective spring constant for the beam is therefore

𝐹 3𝐸𝐼𝑏
𝑘= = 3 .
𝑦 ℓ

Example: A diver with mass 60 kg stands at the end of a diving board fixed at the other end. The diving board
is 4 m long, 50 cm wide and 5 cm thick. It is made from wood with Young’s modulus 10 GPa. If the diver sits
down at the end of the board, causing it to oscillate, is the frequency of the oscillations?

𝑤𝑡 3 (0.5)(0.05)3
Solution: The rectangular diving board has beam moment 𝐼𝑏 = = = 5.2 × 10−6 m4. The
12 12
3𝐸𝐼𝑏 3(10×109 )(5.2×10−6 ) 𝑘
effective spring constant is then given by 𝑘 = ℓ3
= (4)3
= 2.4 × 103 N/m. Now 𝜔 = √𝑚 =

2.4×103 2𝜋
√ = 6.3 rad/s, so the frequency is 𝑓 = ≈ 1 Hz.
60 𝜔

In this and other examples involving beams, the mass of the beam is seldom negligible compared to the mass
placed on it, and ignoring the mass of the beam can produce a large error in the result. In fact, we are
sometimes just interested in the oscillations of a beam without any masses placed on it! Clearly (just as for a
spring) only a part of the mass of the beam has to be added, because not all parts of the beam move at the
same speed as the mass – for instance, the part of the cantilever near the wall hardly moves at all. But the
motion of a spring and beam are not the same, and so it is not a third of the beam's mass that should be
33
added, but rather the fraction 140 or just less than a quarter.9 Let us continue with the example, but now
including the mass of the beam.

Solution (continued): We need the mass of the beam, which is not given. We can estimate it by assuming
the wood has a density of 800 kg/m3. The volume of the beam is 𝑉 = 𝑤ℓ𝑡 = 0.1 m3, so its mass is 80 kg. Its
33 𝑘 2.4×103
effective mass to be added to the mass of the diver is 140 (80) = 19 kg. Now 𝜔 = √𝑚 = √ 60+19 = 5.5
rad/s, which is slightly less than before.

It is interesting to consider how fast the diving board would vibrate with no-one on it. Then the entire mass
𝑘 2.4×103
is just the effective mass of the board, and so 𝜔 = √𝑚 = √ 19
= 11.2 rad/s.

Example: One possible source of renewable energy is wind energy. Suppose the cylindrical tower of a wind
turbine is 54 m high. It is made of steel (𝐸 = 200 GPa, 𝜌 = 8000 kg/m3) and has a diameter of 50 cm. It is
hollow inside and the steel is 10 cm thick. (a) If the actual generator and fins to be mounted in the top of the
tower have a mass of 15 tons, with what frequency will the tower oscillate when displaced? (b) If the
oscillations have an amplitude of 50 cm, also calculate the maximum acceleration experienced at the top of
the tower.

9
The reader interested in this result and its accuracy may consult “On the representation of a cantilevered beam carrying a tip mass
by an equivalent spring-mass system” by Gürgöze (Journal of Sound and Vibration vol. 282 (2005) 538).
 64

Solution: (a) We can regard the tower as a vertical round beam fixed at the bottom and free at the top i.e. a
3𝐸𝐼𝑏 𝜋𝑑 4
cantilever. Therefore 𝑘 = ℓ3
. The beam moment for a circular beam is given by 𝐼𝑏 = 64
, but that is for a
𝜋(𝑑𝑜4 −𝑑𝑖4 )
solid beam. The tower is hollow, so we must subtract away the missing central part giving 𝐼𝑏 = 64
where 𝑑𝑜 is the outer diameter (50 cm) and 𝑑𝑖 is the inner diameter (30 cm). Using these values 𝐼𝑏 =
0.00267 m4, and so 𝑘 = 10175 N/m. To the 15 ton mass of the generator and fins we must add the effective
𝜋(𝑑𝑜2 −𝑑𝑖2 )ℓ
mass of the tower. The tower's volume is 𝑉 = 4
(where we divide by 4 because we are using the
diameter, not the radius), which is 𝑉 = 6.8 m3, hence its mass is 54.4 tons of which we only add the fraction
33 𝑘
(54.4) = 12.8 tons. The total effective mass is therefore 15 + 12.8 = 27.8 tons. Now 𝜔 = √ =
140 𝑚

10175
√ = 0.6 rad/s, which corresponds to a period of about 10 s. (b) If the top of the tower sways, its motion
27800
is given by simple harmonic motion 𝑥 = 𝐴 cos(𝜔𝑡 − 𝜑). Taking the derivative twice gives the acceleration
𝑎 = −𝐴𝜔2 cos(𝜔𝑡 − 𝜑) so the maximum acceleration is given by 𝐴𝜔2 . If the amplitude of the oscillations is
50 cm then the maximum acceleration experienced at the top of the tower is 0.18 m/s2, which is unlikely to
cause severe problems.

5.8.2. Beams supported in other ways

We list some further possibilities:

 For a beam supported (but not fixed) on both ends the vertical force required to deflect the midpoint of
the beam upwards or downwards by a distance 𝑦 from its equilibrium position can be shown to be 𝐹 =
48𝐸𝐼𝑏 48𝐸𝐼𝑏 17
𝑦 and the effective spring constant is therefore given by 𝑘 = . The fraction or about 49% of
ℓ3 ℓ3 35
the mass of the beam should be added to the mass placed at the midpoint.
 For a beam supported on one end and fixed on the other side (built into the wall) the vertical force
required to deflect the midpoint of the beam upwards or downwards by a distance 𝑦 from its equilibrium
768𝐸𝐼𝑏
position can be shown to be 𝐹 = 7ℓ3
𝑦 and the effective spring constant is therefore given by 𝑘 ≈
110𝐸𝐼𝑏 764
ℓ3
. One should add 1715 or about 45% of the mass of the beam to the mass placed at the midpoint.
 For a beam fixed on both ends the vertical force required to deflect the midpoint of the beam upwards
192𝐸𝐼𝑏
or downwards by a distance 𝑦 from its equilibrium position can be shown to be 𝐹 = ℓ3
𝑦 and the
192𝐸𝐼𝑏 13
effective spring constant is therefore given by 𝑘 = ℓ3
. The fraction 35 or about 37% of the mass of
the beam should be added to the mass placed at the midpoint.

5.9. Vibration of a diatomic molecule

Consider two masses 𝑚1 and 𝑚2 which are attached to one another by a spring with spring constant 𝑘. This
provides a simple model for a diatomic molecule – note that for small displacements the chemical bonding
force will have the form of Hooke’s law, allowing us think of the atoms as connected by a tiny (imaginary)
spring.

Let us start with the simple case when the two masses are equal i.e. 𝑚1 = 𝑚2 = 𝑚, e.g. the molecules H2,
N2, O2, F2 and Cl2. Then the two atoms of the molecule will vibrate symmetrically to either side, and the point
 65

on the spring halfway between atoms (at the centre of mass) will be a stationary point. Then we only need
to consider half the picture: we have a spring half the length of the original one joining the two atoms, which
is fixed at one end (the point mid-way between the atoms) and connected to one of the atoms at the other
end. Since this spring has only half the length original one joining the atoms it has double the spring constant,
𝑘eff 2𝑘
so 𝑘eff = 2𝑘 and the angular frequency of vibration will be 𝜔 = √ = √ . Of course, the other half of the
𝑚 𝑚
molecule is performing the mirror image vibration, so this is the vibration frequency of the whole molecule.

Now consider the general case when 𝑚1 ≠ 𝑚2 . Since we are not considering any external forces on the
molecule, its centre of mass will not accelerate and we can choose this point as the origin. Suppose the atoms
lie at distances ℓ1 and ℓ2 from the origin along the X-axis (with ℓ1 on the negative X-axis). The centre of mass
∑ 𝑚𝑖 𝑥𝑖
is given by 𝑥com = ∑ 𝑚𝑖
and since we chose the origin to lie at this point it follows that 𝑚1 (−ℓ1 ) + 𝑚2 ℓ2 =
0. Also ℓ1 + ℓ2 = ℓ which is the total separation of the atoms, or the total length of the imaginary spring
joining them. Eliminating ℓ2 from these equations gives

𝑚2
ℓ1 = ℓ.
𝑚1 + 𝑚2
𝑚2
The total spring is therefore fixed at the origin, and a fraction 𝑚 of its length joins it to the atom with
1 +𝑚2
mass 𝑚1 . Since the effective spring constant of a spring is inversely proportional to its length, this piece of
𝑚1 +𝑚2
the spring has effective spring constant 𝑘1 = 𝑚2
𝑘 and so the angular frequency of vibration will be 𝜔 =
𝑚1 +𝑚2
𝑘 𝑘 𝑘 1 1 1
=√
𝑚2
√𝑚1 𝑚1
= √𝑚 where 𝑚 = 𝑚 + 𝑚 . The other atom has a different effective spring constant
1 eff eff 1 2

and mass, but vibrates with exactly the same frequency, as the reader may verify. Therefore this represent
the vibration frequency of the whole molecule.

Example: Suppose a molecule of HCl has a vibration angular frequency 𝜔 = 5.7 x 1014 rad/s. What is the
spring constant of the bond between the atoms?

𝑘
Solution: Since 𝜔 = √𝑚 , we have 𝑘 = 𝑚eff 𝜔2. From a periodic table, the atomic masses of H and Cl are
eff

1.01 and 35.45 atomic mass units respectively, so the effective mass of the molecule can be calculated using
1 1 1
𝑚eff
= 𝑚 + 𝑚 as 0.98 atomic mass units, or 1.63 x 10-27 kg. This gives 𝑘 = 530 N/m, which is a strong bond.
1 2

5.10. Vibration of a ring

Consider a rod bent into a circular ring. The radius of the ring can be increased by stretching the material. If
the radius is increased, and then the ring is suddenly let free, the radius would oscillate. This is called the
radial breathing vibration, because the changing area inside the ring is similar to the changing volume in our
lungs as we breathe. This would be a rather strange vibration to encounter in a mechanical system, but it has
an important application: carbon nanotubes are carbon molecules in the shape of long thin tubes, which
show promise of many exciting applications in nanotechnology. The radii of the nanotubes can vary
depending on how they are made. It is important to be able to measure the radii, because they affect the
 66

properties of the nanotubes. It is possible to do so using a transmission electron microscope, but can be done
much more easily another way. When light of a certain frequency is shone on the nanotubes, some of the
scattered light has either an increased or decreased frequency (due to the so-called Raman effect) and the
change corresponds to the frequency of the radial breathing vibration. We will show that this frequency is
related to the radius of the nanotubes.

Suppose the ring has been stretched so that its original radius 𝑅 increases by 𝑑𝑅. Then treating the ring like
𝐴 𝐴
a curved rod which is stretched ( 𝑘 = 𝐸 ℓ , equation (44)) the tension in the ring will be = 𝐸 ℓ 𝑑ℓ =
𝐴 𝑑𝑅
𝐸 2𝜋𝑅 (2𝜋𝑑𝑅) = 𝐸𝐴 𝑅
. If we now consider a small section subtended by a small angle 𝑑𝜃 then the radial
force due to the tension will be

𝑑𝜃 𝑑𝑅
𝑑𝐹 = 2𝑇 sin ( ) ≈ 𝑇𝑑𝜃 = 𝐸𝐴 𝑑𝜃.
2 𝑅
𝑑𝜃
Comparing this to 𝑑𝐹 = 𝑘 𝑑𝑅, one gets the effective spring constant 𝑘 = 𝐸𝐴 𝑅
. The mass of the small piece
of material is given by 𝑚 = 𝜌𝐴𝑅𝑑𝜃 and so

𝑑𝜃
𝑘 √ 𝐸𝐴 𝑅 1 𝐸
𝜔=√ = = √ .
𝑚 𝜌𝐴𝑅𝑑𝜃 𝑅 𝜌

Thus the radial breathing frequency is inversely proportional to the radius of the ring.

Example: During a light scattering (Raman) experiment, light of 500 nm is shone on some carbon nanotubes,
𝐸
and some of the scattered light is found to have changed in frequency by 1%. If √ = 21 km/s for carbon
𝜌

nanotubes, estimate their radius.

𝑐 3×108
Solution: The incident light has a frequency 𝑓 = 𝜆 = 500×10−9 = 6 × 1014 Hz. The frequency of the radial
breathing vibration of the carbon nanotubes corresponds to the shift in frequency of the scattered light which
is 1% of this, or 6 x 1012 Hz, which corresponds to an angular frequency 𝜔 = 2𝜋𝑓 = 3 × 1013 rad/s. Therefore
the radius of the carbon nanotubes is

1 𝐸 1
𝑅= √ = (21 000) = 0.7 nm.
𝜔 𝜌 3 × 1013

The mechanical vibrations of a metal ring, if struck or dropped, are not radial breathing vibrations. Instead,
the ring is initially squashed into an elliptical shape, and then oscillates between the shapes of two ellipses
with perpendicular major and minor axes. In this type of vibration the length does not change, but the ring
is bent. This is rather more difficult to analyse, but for interest we give the result here. The angular frequency
is

6𝑝 𝐸
𝜔= √
√5𝑅 2 𝜌

where 𝑝 is half the radius of the cross-section of the ring if it is circular, or if it has a rectangular cross-section
then 𝑝 is the thickness of the ring divided by √12.
 67

Example: A gold metal wedding ring is dropped. It has a radius of 1 cm and a rectangular cross-section with
a thickness of 1.5 mm. Estimate the frequency of the sound emitted. The density of gold is 19 300 kg/m3 and
its elastic modulus is 78 GPa.

1.5×10−3
Solution: For this ring 𝑝 = = 4.33 × 10−4 m. Then the angular frequency of vibration will be 𝜔 =
√12
6(4.33×10−4 ) 78×109 𝜔
√ = 23 kHz. The audible frequency will be 𝑓 = 2𝜋 = 3.7 kHz.
√5(10−2 )2 19 300

5.11. Bobbing of a floating cylinder

Figure 33

Bobbing cylinder.

Consider a cylinder floating in a liquid as shown in Figure 33. When the cylinder (mass 𝑚) is pressed a distance
𝑦 down into the liquid from its equilibrium height, Archimedes' principle states that there is a buoyancy force
equal to the weight of the displaced liquid.

If the cross sectional area is 𝐴, the volume of displaced water is 𝐴𝑦 and its weight is (𝜌𝐴𝑦)𝑔. Therefore the
buoyancy force can be expressed as 𝐹 = −𝜌𝐴𝑔 𝑦. Comparing this to Hooke’s law 𝐹 = −𝑘𝑦 one gets that
𝑘 = 𝜌𝐴𝑔 and

𝑘 𝜌𝐴𝑔
𝜔=√ =√ .
𝑚 𝑚

There is not really a spring in the system, but the buoyancy force behaves like a spring force and results in
simple harmonic motion.
 68

5.12. Liquid in a U-tube (manometer)

Figure 34

Manometer.

Consider a liquid of density 𝜌 filling a U-shaped glass tube of constant cross-sectional area 𝐴. This device is
called a manometer and can be used to measure pressure. If initially both ends are open to the atmosphere,
the liquid in both sides will be at the same height. Then suppose the left hand side is closed and extra liquid
is added on the right. The liquid level on the right hand side would be higher than on the left, and the
difference in height would be an indication of the pressure (above atmospheric pressure) of the trapped air
on the left hand side. Consider what would happen if the left hand side is then suddenly opened. Then the
liquid will flow from the right hand side to the left and, if friction is neglected, oscillations would occur. But
what would be the period of these oscillations?

Suppose at some instant the water was displaced a distance 𝑥 above its equilibrium level on the right side,
meaning it must also be 𝑥 below its equilibrium level on the left side. The force then pushing on the water is
the weight of the unmatched part of the column, which has height 2𝑥. This force is therefore 𝐹 = −𝜌𝐴(2𝑥)𝑔
and the effective spring constant of the system is 𝑘 = 2𝜌𝐴𝑔. The angular frequency of the oscillations is
𝑘 2𝜌𝐴𝑔
therefore given by 𝜔 = √ = √ where 𝑚 is the mass of all the liquid in the tube. We can write this is a
𝑚 𝑚
nicer form if we define the total length of the liquid in the U-tube to be 𝐿. Then the mass of all the water is
𝑚 = 𝜌𝐴𝐿 and

2𝜌𝐴𝑔 2𝑔
𝜔=√ =√ .
𝜌𝐴𝐿 𝐿

𝑔
If the horizontal part of the U-tube is negligibly small we can write this as 𝜔 = √ ℓ where ℓ is the height of
the liquid column. Then the liquid in the U-tube oscillates at the same rate as does a simple pendulum with
length equal to the height of the liquid columns!

5.13. The springiness of a gas and the Helmholtz resonator

Consider a large gas-filled container with a narrow cylindrical neck in which a cylinder of mass 𝑚 fits perfectly.
The cylinder acts like a plug by blocking any gas from entering or leaving the container, but it can move up
along the neck without friction. If the cylinder is pulled out, say a distance 𝑑𝑥, then the gas inside the
container decrease in pressure, resulting in a restoring force 𝑑𝐹 = 𝐴 𝑑𝑃 where 𝑑𝑃 is the change in pressure
 69

of the gas (a negative quantity) when its volume is increased by 𝑑𝑉 = 𝐴 𝑑𝑥. We need to find the relationship
between 𝑑𝑉 and 𝑑𝑃.

One may consider using the ideal gas law 𝑃𝑉 = 𝑛𝑅𝑇, but it is applicable only for a gas in equilibrium. To
reach equilibrium, heat will have to move out of the compressed gas, and this probably take longer than the
period of the oscillations. Instead, we must use the fact that 𝑃𝑉 𝛾 is a constant, which is true for an adiabatic
process (one in which no heat flow occurs). Here 𝛾 represents the ratio of the heat capacity of the gas at
7
constant pressure to constant volume, and is 5 for diatomic gases such as the nitrogen and oxygen which
make up most of the atmosphere. Taking the derivative gives 𝑑𝑃 𝑉 𝛾 + 𝑃𝛾𝑉 𝛾−1 𝑑𝑉 = 0, or

𝑃
𝑑𝑃 = −𝛾 ( ) 𝑑𝑉.
𝑉
𝑃 𝑃
Therefore the restoring force is given by 𝑑𝐹 = −𝐴𝛾 (𝑉) 𝑑𝑉 = −𝐴2 𝛾 (𝑉) 𝑑𝑥 and so the effective spring
constant is

𝑃
𝑘 = 𝐴2 𝛾 ( ).
𝑉

𝑘 𝐴2 𝛾𝑃
Therefore the angular frequency of the oscillations is 𝜔 = √𝑚 = √ 𝑚𝑉 where 𝑚 is the mass of the
cylindrical plug in the neck of the container and 𝑃 and 𝑉 are the equilibrium pressure and volume of the gas.

The above example may seem rather contrived (unrealistic), but suppose we regard the small amount of air
in the neck of the container as the plug. This plug then has mass 𝑚 = 𝜌𝐴ℓ where 𝜌 is the density of the gas
and ℓ is the length of the neck. Then the oscillations would have angular frequency

𝐴2 𝛾𝑃 𝐴𝛾𝑃
𝜔=√ =√ .
(𝜌𝐴ℓ)𝑉 𝜌ℓ𝑉

These oscillations can occur, for instance, if we blow over the opening of a bottle. Such a setup is called a
Helmholtz resonator.

Example: A container has a volume of one litre and is at atmospheric pressure. If its neck has an area of 2
cm2 and is 3 cm long, estimate the frequency of the sound that could be produced by bowing over the neck.
The density of atmospheric gas is roughly 1 kg/m3.

7
Solution: Atmospheric pressure is about 105 Pa, one litre is 10-3 m3 and 𝛾 = = 1.4 for atmospheric gases.
5
𝐴𝛾𝑃 (2×10−4 )(1.4)(105 )
Therefore 𝜔 = √𝜌ℓ𝑉 = √ (1)(3×10−2 )(10−3 ) = 966 rad/s. This corresponds to a frequency 𝑓 = 154 Hz.

5.14. The vibration of an electron

There are many models for the atom. In the Bohr model of the hydrogen atom one imagines the electron
moving around the nucleus in a similar way to a planet moving around the sun, but with the electric force
taking the place of gravity. From the quantum mechanical model of the hydrogen atom, however, one gets
 70

a picture of the electron spread out into a spherical electron cloud with the nucleus at its centre. It is this
latter model we will use here.

If the positively charged nucleus is displaced from the centre of the electron cloud, then it will immediately
have less of the electron cloud in front of it and more behind, and since the electron cloud is negatively
charged the nucleus will experience a restoring force back towards the centre. Therefore the nucleus will
vibrate back and forth through its equilibrium position at the centre of the electron cloud. Actually, since the
electron (cloud) is about a thousand times lighter than the nucleus, it is the nucleus that will remain almost
stationary and the electron cloud that will oscillate!10 If one assumes the electron cloud does not distort but
remains spherical (with radius 𝑅) and that the electron's charge is spread out uniformly over this spherical
cloud, then the spring constant 𝑘 for an atom with atomic number 𝑍 can be calculated as

𝑍2𝑒 2
𝑘=
4𝜋𝜀0 𝑅3

where 𝜀0 is the permittivity of free space. This is only an estimate of 𝑘, because of the assumptions made.

Example: Calculate the vibration frequency of a helium atom based on the model above. The atomic radius
of helium is 58% as large as for hydrogen (which has the Bohr radius, 𝑎0 = 53 pm).

Solution: Based on the data given, the atomic radius for helium is 𝑅 = 31 pm. Also 𝑍 = 2 and so

(2)2 (1.6 × 10−19 )2

𝑘= = 3.1 × 104 N/m.
4𝜋(8.85 × 10−12 )(31 × 10−12 )3

The nucleus of helium consists of two protons and two neutrons, while the electron cloud consists of two
electrons. Although it would be most accurate to calculate the reduced mass of the nucleus and electron
cloud, because the nucleus is so much heavier we can assume it is stationary while the electron cloud
oscillates. We then use the mass of the two electrons, and obtain

𝑘 3.1 × 104
𝜔=√ =√ = 1.3 × 1017 rad/s.
𝑚 2(9.1 × 10−31 )

5.15. Problems

1. A simple pendulum has a length of 1 meter. (a) If it swings in Bloemfontein where 𝑔 = 9.788 m/s2, what
is the period for small oscillations and how many oscillations does it complete in one day? (b) What is its
period in Cape Town, where 𝑔 = 9.796 m/s2 and how many oscillations does it complete in one day? (c) Give
a brief but plausible reason why the gravitational acceleration is less in Bloemfontein than in Cape Town. (d)
Comment on the feasibility of using such a pendulum to measure the difference in 𝑔 between these two
cities.

10 1 1 1
Since both the masses are actually moving, we should really use their reduced mass = + . Since the electron cloud is
𝑚eff 𝑚1 𝑚2
much lighter than the nucleus, the reduced mass is very close to that of the electron, which we shall use.
 71

2. A mass 𝑚 is hung on a vertical spring having spring constant 𝑘, and stretches a distance ℎ. If the mass is
ℎ
now set in motion, show that the period is given by 𝑇 = 2𝜋√𝑔. (Note: this means that the mass-spring system
has the same period as a swinging pendulum with length ℎ!)

3. Consider a pendulum clock having a uniform rod as its pendulum. (a) If no extra mass is attached to the
end of the rod, what length ℓ should it be so that its period is precisely 1 s? (Take 𝑔 = 9.8 m/s2.) (b) When a
material is heated, it generally expands. Explain what effect this would have on the accuracy of a pendulum
clock with a metal pendulum. (Note: interesting compensation techniques have been devised for this
problem – you may wish to try find out more.) (c) Returning to the rod pendulum, suppose a mass equal to
the mass of the rod is added at its free end. Show that the angular frequency of the oscillations is then 𝜔 =
9𝑔
√ .
8ℓ

Figure 35

Rotating semi-circular disk

pivoted at midpoint.

4. A semi-circular disk of radius 𝑟 is pivoted at its midpoint (Figure 35). Show that the period of this compound
3𝜋𝑟 4
pendulum is given by 𝑇 = 2𝜋√ 8𝑔 . You may use the fact that the centre of mass lies a distance 𝑅 = 3𝜋 𝑟 from
the pivot.

5. Show that for a diatomic molecule with two atoms of the same type, the reduced mass of the molecule is
half the mass of either atom.

6. Estimate the vibration frequency of a hydrogen molecule if the bond has an effective spring constant 𝑘 ≈
1000 N/m.

7. Silicon is used to make computer chips (integrated circuits). During manufacturing the silicon may be
exposed to hydrogen plasma consisting of individual hydrogen atoms which can enter the silicon. Researchers
in the 1990s found evidence of vibrations from the processed silicon and suspected that they were due to
hydrogen molecules trapped in the silicon. Yet sceptics were unconvinced because the hydrogen in the
plasma consists of free atoms, not molecules, and they felt there were no spaces big enough inside the silicon
 72

for the hydrogen atoms to bond, forming molecules. In addition, the measured vibration frequency did not
match that of pure hydrogen gas. The researchers claimed that the frequency was shifted because the
molecules were trapped in the silicon instead of being free as in a gas, but to prove their point they did
another experiment. They mixed hydrogen with deuterium and used this to process the silicon. Deuterium is
an isotope of hydrogen having a neutron as well as a proton in the nucleus and thus double the mass,
although it has the same chemical properties.

In the new samples, the researchers found evidence of the same vibration frequency as before, but also two
new frequencies. Explain why this convinced the sceptics that hydrogen molecules were indeed forming
inside the silicon, especially since the newly observed frequencies were 86% and 71% of the original
frequency.

8. You have three springs, with spring constants of 5, 8 and 20 N/m. How can you connect the springs to give
you a system with the following spring constants: (a) 13 N/m (b) 33 N/m (c) 5.7 N/m (d) 7.9 N/m. You will
have to use trial and error and do not necessarily have to use all the springs.

Figure 36

Spring divided into two half-

springs.

9. Consider a mass hung on a spring with spring constant 𝑘, and having some angular frequency of oscillation
𝜔 (as in Figure 36(a)). Suppose the spring is cut in half, and the mass is hung from half while the other half is
inserted between the mass and floor (as in Figure 30(b)). (a) What will be the spring constants of the half-
springs, in terms of 𝑘? (b) Do the two half-springs act in series or in parallel? This requires careful thought!
(c) Using your results above, show that the frequency of oscillations in the new configuration is double the
original value.

10. Although we can model a solid as a set of atoms linked by springs, this model is only an approximation.
The temperature of a material is a measure of the energy of the atoms, which increases as the temperature
increases. By considering the fact that most materials expand when heated, do you think the “spring-like”
potential energy of an atom in a solid increases more quickly with distance from its equilibrium position as it
moves towards, or away from, another atom in the solid?
 73

Figure 37

Microscope image of atomic

force microscope cantilever.

11. The atomic force microscope improves upon on the scanning tunnelling microscope for which Binnig and
Rohrer won the Nobel Prize in 1986. In consists of a very small cantilever with a probe at its tip, which
oscillates near the surface of a material. It can measure and be used to manipulate individual atoms or strands
of DNA. From the microscope image of the AFM cantilever (Figure 37), one can estimate it is 40 μm wide and
3 μm thick. If it is 200 μm long, firmly secured at the back and made of silicon (for which you can look up the
density), and the mass at the tip is only 10% of the mass of the cantilever, then estimate its frequency of
vibration when displaced. Take 𝐸 = 150 GPa (but note, this is only an estimate: silicon is anisotropic i.e. its
elastic properties depend on the direction in the crystal.)

Figure 38

Parts of a flower.

12. In the American Journal of Botany vol. 82 (1995) p.1407 King and Buchmann state that “Although
Rhododendron spp. anthers have apical pores and should be expected to be buzz pollinated, bees do not
normally sonicate them to release pollen.” An important consideration during this study was the frequency
at which the filaments (holding the pollen-containing anthers) in the flower would vibrate. The authors
modelled this by taking the filament as a circular solid beam, fixed at one end and free at the other, with a
mass (the anther) at the tip. If a filament has a diameter of 0.6 mm, a length of 28 mm, an elastic modulus of
70 MPa and the density of water, and the anther at its free tip has a mass of 0.86 mg, show that the system
vibrates at an angular frequency of about 150 rad/s when displaced. [Note: Take care with the units – the SI
unit for mass is the kilogram, not gram.]
 74

Figure 39

Tuning fork modelled

as a cantilever. [Source:
Ong, Phys. Educ. 37
(2002) p.540]

13. Tuning forks (Figure 39(a)) are often used to produce sound of a certain frequency. The tuning fork has
two prongs, which vibrate out of phase to one another i.e. they both move outwards or inwards at the same
time. This “antiphase” motion produces equal but opposite forces on the handle, meaning that it does not
vibrate. This keeps the energy in the prongs and prevents it being transferred via the handle to the hand,
which would not be possible with just one prong. Actually, either prong can be considered as a cantilever
(Figure 39(b)) with the side near the handle fixed and the tip free. Estimate the frequency emitted by a tuning
fork made of aluminium 𝐸 = 50 GPa, 𝜌 = 2.7 g/cm3) where the thickness, width and length of the prongs
are 0.5 cm, 1 cm and 10 cm respectively.

Figure 40

Mass at centre of a
supported beam.

14. Consider a mass 𝑚 placed at the centre of a beam of length 𝐿 which is supported (but not fixed) at both
ends (Figure 40). If the beam is made from wood (𝐸 = 10 GPa, 𝜌 = 800 kg/m3, width 30 cm, thickness 10
cm, length 8 m) and the mass is 100 kg, estimate the frequency of vibrations when it is set in motion. What
is the percentage error when the mass of the beam is not taken into account?

15. Show that for a spring made of wire having a circular cross-section 𝜅 = 𝑘𝐷 2 .
 75

Figure 41

Mass on a tensioned
wire.

16 Consider a mass 𝑚 attached to a piece of wire of length 𝐿 which is under constant tension 𝑇 (Figure 41).
(a) If the mass of the wire is negligible, and the oscillations are small, show that when displaced the mass
𝑇𝐿
oscillates with angular frequency 𝜔 = √𝑚𝑎(𝐿−𝑎) where 𝑎 is the distance from the edge of the wire where the
mass is attached. (b) Where should the mass be attached if it is to have the lowest angular frequency, and
what is this angular frequency?
 76

6. Working with Simple Harmonic Motion

6.1. Different mathematical forms to express SHM

6.1.1. Amplitude-phase form

From the preceding chapters it should be clear that oscillations and simple harmonic motion (SHM) are very
important in many systems. The motion is given in equation (32) simply as 𝑥 = 𝐴 cos(𝜔𝑡 − 𝜑). However, in
section 4.2.2 we noted that one could just as well have used 𝑥 = 𝐴 sin(𝜔𝑡 + 𝜑). Both of these forms satisfy
the differential equation 𝑎 = −𝜔2 𝑥 for simple harmonic motion, which we get from Newton’s second law
where the only force on the object obeys Hooke’s law. Both give precisely the same family of curves when
𝐴 and 𝜑 are allowed to vary arbitrarily. Different books use different forms, so as a scientist it is important
to be able to work with any of them and even convert from one form to another.

Example: A physics book describes simple harmonic motion by means of the formula 𝑥 = 𝐴 sin(𝜔𝑡 − 𝜑),
𝜋
and for a certain problem finds that 𝐴 = 5 and 𝜑 = 2 . If instead we represent simple harmonic motion by
the formula 𝑥 = 𝐴 cos(𝜔𝑡 − 𝜑), what values must we use for 𝐴 and 𝜑 to get the same motion?

Solution: Using the expansion formula sin(𝑢 − 𝑣) = sin 𝑢 cos 𝑣 − cos 𝑢 sin 𝑣 one finds that

𝜋 𝜋 𝜋
𝑥 = 5 sin (𝜔𝑡 − ) = 5 (sin 𝜔𝑡 cos − cos 𝜔𝑡 sin ) = −5 cos 𝜔𝑡.
2 2 2

Therefore the motion is equivalent to the different form 𝑥 = 𝐴 cos(𝜔𝑡 − 𝜑) with 𝐴 = −5 and 𝜑 = 0.

But this answer is not unique: we could also have taken 𝐴 = −5 and 𝜑 = 2𝜋, or 𝐴 = −5 and 𝜑 = −6𝜋, since
the cosine function is periodic with period 2𝜋.

Another different option is 𝐴 = 5 and 𝜑 = 𝜋. Although all these possibilities are mathematically equivalent,
we generally prefer the last one as we like to work with 𝐴 > 0 and 0 ≤ 𝜑 < 2𝜋.

6.1.2. Form with no phase constant (cos+sin form)

Simple harmonic motion written as 𝑥 = 𝐴 cos(𝜔𝑡 − 𝜑), or one of its variations, is said to be in the amplitude-
phase form. One can express the motion with no phase constant as follows:

𝑥 = 𝐴 cos(𝜔𝑡 − 𝜑) = 𝐴{cos 𝜔𝑡 cos 𝜑 + sin 𝜔𝑡 sin 𝜑} = {𝐴 cos 𝜑} cos 𝜔𝑡 + {𝐴 sin 𝜑} sin 𝜔𝑡.

Defining two new constants 𝑃 = 𝐴 cos 𝜑 and 𝑄 = 𝐴 sin 𝜑 one gets

𝑥 = 𝑃 cos 𝜔𝑡 + 𝑄 sin 𝜔𝑡.

 77

Now the motion is described by both a cosine and a sine function, each with its own “amplitude” (which is
not the same as the amplitude of the motion) and with no phase constants. Some people prefer to work with
this form, which we shall call the cos+sin form.

Example: A physics book describes simple harmonic motion by means of the formula 𝑥 = 𝐴 cos(𝜔𝑡 + 𝜑),
and for a certain problem finds that 𝐴 = 7.3 and 𝜑 = 2.1. If instead we represent simple harmonic motion
by the form 𝑥 = 𝑃 cos 𝜔𝑡 − 𝑄 sin 𝜔𝑡, what values must we use for 𝑃 and 𝑄 to get the same motion?

Solution: Using the expansion formula cos(𝑢 + 𝑣) = cos 𝑢 cos 𝑣 − sin 𝑢 sin 𝑣 one finds that

𝑥 = 7.3 cos(𝜔𝑡 + 2.1) = 7.3(cos 𝜔𝑡 cos 2.1 − sin 𝜔𝑡 sin 2.1) = −3.7 cos 𝜔𝑡 − 6.3 sin 𝜔𝑡.

Therefore the motion is equivalent to the different form 𝑥 = 𝑃 cos 𝜔𝑡 − 𝑄 sin 𝜔𝑡 with 𝑃 = −3.7 and 𝑄 =
6.3.

Example: A mass attached to a spring oscillates according to 𝑥 = 6 cos 𝜔𝑡 − 4 sin 𝜔𝑡. Express this motion in
the amplitude-phase form 𝑥 = 𝐴 cos(𝜔𝑡 − 𝜑).

Solution: Using the expansion formula cos(𝑢 − 𝑣) = cos 𝑢 cos 𝑣 + sin 𝑢 sin 𝑣 one finds that

𝑥 = 𝐴 cos(𝜔𝑡 − 𝜑) = 𝐴{cos 𝜔𝑡 cos 𝜑 + sin 𝜔𝑡 sin 𝜑} = {𝐴 cos 𝜑} cos 𝜔𝑡 + {𝐴 sin 𝜑} sin 𝜔𝑡.

Therefore we require 𝐴 cos 𝜑 = 6 and 𝐴 sin 𝜑 = −4. We need to solve these two equations simultaneously.

 To get the amplitude we square the equations and add them, giving (𝐴 cos 𝜑)2 + (𝐴 sin 𝜑)2 =
𝐴2 [cos2 𝜑 + sin2 𝜑] = 𝐴2 = 36 + 16 = 52, so 𝐴 = √52.
𝐴 sin 𝜑 −4
 To get the phase constant we can divide the equations, giving 𝐴 cos 𝜑 = tan 𝜑 = . This has two solutions
6
with 0 ≤ 𝜑 < 2𝜋, namely in quadrants II (𝜑 = 2.55) and IV (𝜑 = 5.70). But only one is correct. From the
original equations 𝐴 cos 𝜑 = 6 and 𝐴 sin 𝜑 = −4 we can see that cosine must be positive while sine
must be negative, so the correct quadrant is IV and 𝜑 = 5.70.

Therefore 𝑥 = √52 cos(𝜔𝑡 − 5.70) is equivalent to 𝑥 = 6 cos 𝜔𝑡 − 4 sin 𝜔𝑡.

Example: Write the following in the amplitude-phase form: √8 sin 𝜔𝑡 + cos 𝜔𝑡.

Solution: We wish to convert the function √8 sin 𝜔𝑡 + cos 𝜔𝑡 into the amplitude-phase form 𝑥 =
𝐴 cos(𝜔𝑡 − 𝜑). Expanding this as in the previous example, we have by comparison that 𝐴 cos 𝜑 = 1 and
𝐴 sin 𝜑 = √8. Squaring and adding these equations gives 𝐴 = 3. Dividing the two equations gives tan 𝜑 =
√8. The tangent function is positive in quadrants I and III, but the fact that sin 𝜑 and cos 𝜑 are both positive
means the solution is in quadrant I and 𝜑 = 1.23 rad. Therefore

√8 sin 𝜔𝑡 + cos 𝜔𝑡 = 3 cos(𝜔𝑡 − 1.23 ).

 78

6.1.3. Complex exponential form

𝑑2 𝑥
If we consider the differential equation 𝑎 = −𝜔2 𝑥 for simple harmonic motion, we can write it as +
𝑑𝑡 2
2
𝜔 𝑥 = 0. But consider the following mathematical theorem:

A differential equation of the form

𝑑𝑛 𝑥 𝑑𝑛−1 𝑥 𝑑2 𝑥 𝑑𝑥
𝑐𝑛 𝑛 + 𝑐𝑛−1 𝑛−1 + ⋯ + 𝑐2 2 + 𝑐1 + 𝑐0 𝑥 = 0
𝑑𝑡 𝑑𝑡 𝑑𝑡 𝑑𝑡

always has a solution of form 𝑥 = 𝑒 𝑟𝑡 for some value of 𝑟. It may be possible to find up to a maximum of 𝑛
different possible value for 𝑟, and if this is so then the general solution is

𝑥 = 𝐶1 𝑒 𝑟1 𝑡 + 𝐶2 𝑒 𝑟2 𝑡 + ⋯ +𝐶𝑛 𝑒 𝑟𝑛 𝑡 .

𝑑2 𝑥 𝑑𝑥
Example: Find the general solution of 2 𝑑𝑡 2 + 3 𝑑𝑡 + 𝑥 = 0.

𝑑𝑥 𝑑2 𝑥
Solution: From the theorem 𝑥 = 𝑒 𝑟𝑡 must be a solution. Differentiating this gives = 𝑟𝑒 𝑟𝑡 and =
𝑑𝑡 𝑑𝑡 2
𝑟 2 𝑒 𝑟𝑡 , so by substitution 2(𝑟 2 𝑒 𝑟𝑡 ) + 3(𝑟𝑒 𝑟𝑡 ) + 𝑒 𝑟𝑡 = 0 . After dividing by 𝑒 𝑟𝑡 one gets the simple problem
1
2𝑟 2 + 3𝑟 + 1 = 0, with solutions 𝑟 = −1 and 𝑟 = − . Notice that we have found both possible value for
2
our problem (having 𝑛 = 2) so we can form the general solution

1
𝑥 = 𝑀𝑒 −𝑡 + 𝑁𝑒 −2𝑡

where 𝑀 and 𝑁 are two integration constants.

𝑑2 𝑥
We can now apply this theorem to the problem of SHM i.e. 𝑑𝑡 2 + 𝜔2 𝑥 = 0. Taking 𝑥 = 𝑒 𝑟𝑡 and substituting
it and its second derivative into the equation gives 𝑟 2 𝑒 𝑟𝑡 + 𝜔2 (𝑒 𝑟𝑡 ) = 0, or 𝑟 = ±𝑖𝜔 where 𝑖 = √−1. Since
we have two values of 𝑟 (and 𝑛 = 2) we can form the general solution

̃ 𝑒 𝑖𝜔𝑡 + 𝑁
𝑥=𝑀 ̃𝑒 −𝑖𝜔𝑡 .

This is the complex exponential form for simple harmonic motion, and we have placed tildes over the
integration constants 𝑀̃ and 𝑁̃ to emphasize that they are not normal (real) numbers, but complex numbers.
This equation needs more careful interpretation than the other forms of SHM, because although
mathematically 𝑥 might be allowed to be a complex number, in reality 𝑥 must actually be a real number in
order to represents the position of an object. If 𝑥 = 𝑀̃ 𝑒 𝑖𝜔𝑡 + 𝑁
̃𝑒 −𝑖𝜔𝑡 is to be a real number, the imaginary
̃ 𝑒 𝑖𝜔𝑡 and 𝑁
parts of 𝑀 ̃𝑒 −𝑖𝜔𝑡 must cancel out. This requires that 𝑁
̃ is the complex conjugate of 𝑀
̃ , i.e. 𝑁
̃=𝑀̃ ∗,
because then

̃ 𝑒 𝑖𝜔𝑡 + 𝑁
𝑥=𝑀 ̃𝑒 −𝑖𝜔𝑡 = 𝑀
̃ 𝑒 𝑖𝜔𝑡 + 𝑀
̃ ∗ 𝑒 −𝑖𝜔𝑡 = 𝑀 ̃ 𝑒 𝑖𝜔𝑡 )∗ = 2ℝ[𝑀
̃ 𝑒 𝑖𝜔𝑡 + (𝑀 ̃ 𝑒 𝑖𝜔𝑡 ]

where ℝ represents the real part of a complex number, and the imaginary parts have cancelled out.
 79

Strictly, the complex exponential form for simple harmonic motion where the position must be a real number
is therefore 𝑥 = 𝑀 ̃ 𝑒 𝑖𝜔𝑡 + 𝑀̃ ∗ 𝑒 −𝑖𝜔𝑡 , which has two integration constants (the real and imaginary parts of 𝑀
̃ ).
However, physicists very seldom actually use this form because the second term which must be added to
ensure that 𝑥 remains real is considered a “hassle” which makes mathematical expressions long. Instead
they simply use 𝒙 = 𝑪 ̃ 𝒆𝒊𝝎𝒕 without a second term, where 𝑪 ̃ is a complex constant. Immediately one realizes
that this will result in the problem that 𝑥 can take on complex (unphysical) values, but this is resolved with
the convention of just ignoring the imaginary part and using only the real part i.e. one actually uses 𝑥 =
ℝ[𝐶̃ 𝑒 𝑖𝜔𝑡 ], but the “ℝ[ ]” to show we consider only the real part is not explicitly written down – it is only
implied. All physicists know about this convention and if one speaks of the complex exponential form, then
one means 𝑥 = 𝐶̃ 𝑒 𝑖𝜔𝑡 with the convention of ignoring the imaginary part.

Example: Show that for the complex exponential form of SHM the complex integration constant 𝐶̃ is given in
𝑣
terms of the initial conditions by 𝐶̃ = 𝑥0 − 0 𝑖.
𝜔

Solution: We take 𝑥 = 𝐶̃ 𝑒 𝑖𝜔𝑡 (and it is implied that we only use the real part). If 𝑥 = 𝑥0 when 𝑡 = 0 then
𝑥0 = ℝ[𝐶̃ 𝑒 𝑖𝜔(0) ] = 𝐶𝑅 , where 𝐶̃ = 𝐶𝑅 + 𝑖𝐶𝐼 and 𝐶𝑅 is the real part.

By differentiating 𝑥 = 𝐶̃ 𝑒 𝑖𝜔𝑡 we get the velocity 𝑣 = 𝑖𝜔𝐶̃ 𝑒 𝑖𝜔𝑡 . If 𝑣 = 𝑣0 when 𝑡 = 0 then 𝑣0 =

𝑣 𝑣
ℝ[𝑖𝜔𝐶̃ 𝑒 𝑖𝜔(0) ] = ℝ[𝑖𝜔(𝐶𝑅 + 𝑖𝐶𝐼 )] = −𝜔𝐶𝐼 , so 𝐶𝐼 = − 0 . Therefore 𝐶̃ = 𝐶𝑅 + 𝑖𝐶𝐼 = 𝑥0 − 𝑖 0 .
𝜔 𝜔

Example: Convert (3 − 𝑖)𝑒 𝑖𝜔𝑡 to the cos+sin form.

Solution: We wish to write the function (3 − 𝑖)𝑒 𝑖𝜔𝑡 to the cos+sin form, namely 𝑃 cos 𝜔𝑡 + 𝑄 sin 𝜔𝑡. Now

(3 − 𝑖)𝑒 𝑖𝜔𝑡 = (3 − 𝑖){cos(𝜔𝑡) + 𝑖 sin(𝜔𝑡)} = 3 cos(𝜔𝑡) + sin(𝜔𝑡)

where we have ignored the imaginary part. Clearly then 𝑃 = 3 and 𝑄 = 1.

Example: Convert 2 sin 𝜔𝑡 − 5 cos 𝜔𝑡 to the complex exponential form.

Solution: We wish to write the function 2 sin 𝜔𝑡 − 5 cos 𝜔𝑡 in the complex exponential form, namely 𝑥 =
𝐶̃ 𝑒 𝑖𝜔𝑡 . Now

𝑥 = 𝐶̃ 𝑒 𝑖𝜔𝑡 = (𝐶𝑅 + 𝑖𝐶𝐼 ){cos(𝜔𝑡) + 𝑖 sin(𝜔𝑡)} = 𝐶𝑅 cos(𝜔𝑡) − 𝐶𝐼 sin(𝜔𝑡)

where we have ignored the imaginary part. By comparison 𝐶𝑅 = −5 and 𝐶𝐼 = −2. The final solution is that

2 sin 𝜔𝑡 − 5 cos 𝜔𝑡 = (−5 − 2𝑖)𝑒 𝑖𝜔𝑡 .

 80

Example: Convert 6 cos(𝜔𝑡 − 𝜋) to the complex exponential form.

Solution: We wish to write the function 6 cos(𝜔𝑡 − 𝜋) in the complex exponential form 𝐶̃ 𝑒 𝑖𝜔𝑡 . The method
is to convert both of these to the cos+sin form to compare them:

 Firstly 6 cos(𝜔𝑡 − 𝜋) = 6{cos(𝜔𝑡) cos 𝜋 + sin(𝜔𝑡) sin 𝜋} = −6 cos(𝜔𝑡).

 But 𝑥 = 𝐶̃ 𝑒 𝑖𝜔𝑡 = (𝐶𝑅 + 𝑖𝐶𝐼 ){cos(𝜔𝑡) + 𝑖 sin(𝜔𝑡)} = 𝐶𝑅 cos(𝜔𝑡) − 𝐶𝐼 sin(𝜔𝑡).

So by comparison 𝐶𝑅 = −6 and 𝐶𝐼 = 0. Therefore 6 cos(𝜔𝑡 − 𝜋) = −6𝑒 𝑖𝜔𝑡 .

6.2. Adding oscillations

6.2.1. Adding oscillations of the same frequency

If 𝑦1 = 𝐴1 cos(𝜔𝑡 − 𝜑1 ) and 𝑦2 = 𝐴2 cos(𝜔𝑡 − 𝜑2 ), then

𝑦1 + 𝑦2 ≠ {𝐴1 + 𝐴2 } cos(𝜔𝑡 − {𝜑1 + 𝜑2 }).

Before we tackle this problem in general, we shall consider two special cases:

 Suppose the two phases are equal (i.e. 𝜑2 = 𝜑1 ). Then the two oscillations are completely “in phase”
and constructive interference occurs. The added oscillations are then

𝑦1 + 𝑦2 = {𝐴1 + 𝐴2 } cos(𝜔𝑡 − 𝜑1 ).

 Suppose the two phases differ by 𝜋 (i.e. 𝜑2 = 𝜑1 ± 𝜋). Then the two oscillations are completely “out of
phase” and destructive interference occurs. To see this mathematically, consider the second oscillation:
𝑦2 = 𝐴2 cos(𝜔𝑡 − 𝜑2 ) = 𝐴2 cos(𝜔𝑡 − 𝜑1 ∓ 𝜋) = −𝐴2 cos(𝜔𝑡 − 𝜑1 ) . The last step can be proven
mathematically, but follows from the fact that shifting a cosine function to the left or right by 𝜋 is
equivalent to turning it upside down. Now adding the oscillations gives

𝑦1 + 𝑦2 = {𝐴1 − 𝐴2 } cos(𝜔𝑡 − 𝜑1 ).

An interesting situation will arise if 𝐴2 > 𝐴1 , since subtracting the amplitudes will then give a negative
quantity. Although we generally choose an amplitude to be positive, negative values may result during
calculations. This is not a serious problem, and if desired the amplitude can be made positive again by
shifting the phase constant by 𝜋.

We now return to the problem in general. Note that if the two oscillations are in the cos+sin form then adding
them is easy i.e. if 𝑦1 = 𝑃1 cos 𝜔𝑡 + 𝑄1 sin 𝜔𝑡 and 𝑦2 = 𝑃2 cos 𝜔𝑡 + 𝑄2 sin 𝜔𝑡, then

𝑦1 + 𝑦2 = {𝑃1 + 𝑃2 } cos 𝜔𝑡 + {𝑄1 + 𝑄2 } sin 𝜔𝑡.

 81

The easiest way to add two oscillations in the amplitude-phase form is to first convert them into the cos+sin
form, add them, and then convert the result back to the amplitude-phase form if required. This is best
illustrated with an example.

Example: Subtract the following oscillations: 3 cos(𝜔𝑡 + 0.5) − 6 cos(𝜔𝑡 − 1.2).

Solution: First we expand each of the functions into the cos+sin form:

3 cos(𝜔𝑡 + 0.5) = 3{cos 𝜔𝑡 cos 0.5 − sin 𝜔𝑡 sin 0.5} = 2.6327 cos 𝜔𝑡 − 1.4383 sin 𝜔𝑡.

6 cos(𝜔𝑡 − 1.2) = 6{cos 𝜔𝑡 cos 1.2 + sin 𝜔𝑡 sin 1.2} = 2.1741 cos 𝜔𝑡 + 5.5922 sin 𝜔𝑡.

The second is easily subtracted from the first to give 0.4586 cos 𝜔𝑡 − 7.0305 sin 𝜔𝑡. After converting back
to the amplitude-phase form the final result is 7.0454 cos(𝜔𝑡 − 4.7775).

6.2.2. Adding oscillations of different frequencies

Suppose the oscillations to be added are 𝑦1 = 𝐴 cos(𝜔1 𝑡 − 𝜑) and 𝑦2 = 𝐴 cos(𝜔2 𝑡 − 𝜑). Then

𝑦 = 𝑦1 + 𝑦2 = 𝐴[cos(𝜔1 𝑡 − 𝜑) + cos(𝜔2 𝑡 − 𝜑)]

(𝜔1 𝑡 − 𝜑) + (𝜔2 𝑡 − 𝜑) (𝜔1 𝑡 − 𝜑) − (𝜔2 𝑡 − 𝜑)

∴ 𝑦 = 𝐴 {2 cos cos }
2 2

𝜔1 + 𝜔2 𝜔1 − 𝜔2
∴ 𝑦 = 2𝐴 cos ({ } 𝑡 − 𝜑) cos ({ } 𝑡).
2 2
𝜔1 +𝜔2 𝜔1 −𝜔2
By defining 𝜔𝑐𝑎𝑟 = 2
and 𝜔𝑒𝑛𝑣 = 2
, we can express this more compactly as

𝑦 = 2𝐴 cos(𝜔𝑐𝑎𝑟 𝑡 − 𝜑) cos 𝜔𝑒𝑛𝑣 𝑡. (49)

Note that an interesting situation may arise if 𝜔2 > 𝜔1 since then 𝜔𝑒𝑛𝑣 will be negative. We prefer to work
with only positive frequencies. Since the cosine function is even, the sign of 𝜔𝑒𝑛𝑣 in cos 𝜔𝑒𝑛𝑣 𝑡 is unimportant,
𝜔1 −𝜔2
so we can rather define 𝜔𝑒𝑛𝑣 = | |, avoiding the possibility of a negative frequency. Also note that
2
because of their definitions 𝜔𝑒𝑛𝑣 < 𝜔𝑐𝑎𝑟 , and 𝜔𝑐𝑎𝑟 represents the average angular frequency of the
oscillations that were added.

The bottom of Figure 42 shows a graph of this combination, which is called a beat pattern. By rearranging
equation (49) one can write it as 𝑦 = {2𝐴 cos 𝜔𝑒𝑛𝑣 𝑡} cos(𝜔𝑐𝑎𝑟 𝑡 − 𝜑). This is the product of a first oscillation
2𝐴 cos 𝜔𝑒𝑛𝑣 𝑡 called the “envelope” which varies slowly (since 𝜔𝑒𝑛𝑣 < 𝜔𝑐𝑎𝑟 ) and a second oscillation
cos(𝜔𝑐𝑎𝑟 𝑡 − 𝜑) called the “carrier” having the average angular frequency of the original two oscillations.
The sum of two original oscillations of different frequencies is equal to the product of two new oscillations,
namely the envelope and carrier. Notice from the figure how the ‘carrier fits into the envelope’ – in fact, the
envelope controls the maximum size of the beat pattern.
 82

Physically what happens is that because the original two oscillations to be added have slightly different
angular frequencies, they are constantly changing from being in phase with one another, to a bit later being
out of phase with one another, and then later being in phase once again, with this cycle repeating. When the
waves are close to in phase one gets constructive interference of the two original waves and the resultant
envelope reaches its maximum value of 2𝐴, while when the waves are close to out of phase destructive
interference results in an envelope near zero. The envelope can be interpreted as a time varying amplitude
of the carrier oscillation and we say that the envelope ‘modulates’ the carrier. In this particular case the
envelope is also an oscillation, splitting the overall oscillation in a series of ‘wave packets’ or ‘beats’. Later we
will also encounter other types of envelope functions.

Figure 42

Adding oscillations with

different frequencies.

The top image shows the two

oscillations to be added. The
result is
𝑦 = 2𝐴 cos(𝜔𝑐𝑎𝑟 𝑡 − 𝜑) cos 𝜔𝑒𝑛𝑣 𝑡.

The second image shows

𝐴 cos(𝜔𝑐𝑎𝑟 𝑡 − 𝜑) having the
average frequency.

The third image shows the

second one with 2𝐴 cos 𝜔𝑒𝑛𝑣 𝑡
superimposed. Since 𝜔𝑒𝑛𝑣 < 𝜔𝑐𝑎𝑟
the envelope period is much
longer.

The last image shows the full

𝑦=2𝐴cos(𝜔𝑐𝑎𝑟 𝑡 − 𝜑) cos 𝜔𝑒𝑛𝑣 𝑡
obtained as the product of the
two cosine functions.

2𝜋 2𝜋 4𝜋
The envelope 2𝐴 cos 𝜔𝑒𝑛𝑣 𝑡 has a period 𝑇𝑒𝑛𝑣 = = 𝜔 −𝜔 =| |. Notice from Figure 42 that one
𝜔𝑒𝑛𝑣 | 1 2| 𝜔1 −𝜔2
2
period of the envelope corresponds to two ‘beats’ in the added oscillation pattern. Therefore the so-called
beat period is half the envelope period. Note that 𝑇𝑒𝑛𝑣 is not the beat period, but the envelope period.

6.2.3. Adding perpendicular oscillations (Lissajous figures)

A Lissajous figure is the orbit traced out by a particle in the xy-plane when both its position coordinates, 𝑥[𝑡]
and 𝑦[𝑡], are simple harmonic motions. In general the amplitudes, angular frequencies and phases of the
two coordinates may be different. However, special figures are obtained when the frequencies have a
rational ratio, e.g. 1:1, 2:3, or 3:5, etc.

(a) We first consider the simplest Lissajous figures, which have equal vertical and horizontal frequencies
i.e. a frequency ratio 1:1. The position coordinates are then 𝑥 = 𝐴1 cos(𝜔𝑡 − 𝜑1 ) and 𝑦 =
𝐴2 cos(𝜔𝑡 − 𝜑2 ). Elimination of 𝜔𝑡 from these equations gives
 83

𝑥 2 2 cos(𝜑2 − 𝜑1 ) 𝑦2
− 𝑥𝑦 + = sin2 (𝜑2 − 𝜑1 )
𝐴12 𝐴1 𝐴2 𝐴22

which is the equation of an inclined ellipse.

𝑦
Proof: From 𝑦 = 𝐴2 cos(𝜔𝑡 − 𝜑2 ), we can write 𝐴 = cos({𝜔𝑡 − 𝜑1 } − {𝜑2 − 𝜑1 }). Therefore
2

𝑦
𝐴2
= cos(𝜔𝑡 − 𝜑1 ) cos(𝜑2 − 𝜑1 ) + sin(𝜔𝑡 − 𝜑1 ) sin(𝜑2 − 𝜑1 ).

𝑦 𝑥
This means that − cos(𝜑2 − 𝜑1 ) = sin(𝜔𝑡 − 𝜑1 ) sin(𝜑2 − 𝜑1 ). Squaring both sides gives
𝐴2 𝐴1

𝑦2 𝑥𝑦 𝑥2
−2 cos(𝜑2 − 𝜑1 ) + cos2 (𝜑2 − 𝜑1 ) = sin2(𝜔𝑡 − 𝜑1 ) sin2 (𝜑2 − 𝜑1 ).
𝐴22 𝐴1 𝐴2 𝐴21

𝑥2
But sin2 (𝜔𝑡 − 𝜑1 ) = 1 − cos2 (𝜔𝑡 − 𝜑1 ) = 1 − 𝐴2. Substituting this gives
1

𝑦2 𝑥𝑦 𝑥2 𝑥2
−2 cos(𝜑2 − 𝜑1 ) + cos2 (𝜑2 − 𝜑1 ) = sin2(𝜑2 − 𝜑1 ) − sin2 (𝜑2 − 𝜑1 ),
𝐴22 𝐴1 𝐴2 𝐴21 𝐴21

𝑥2 2 cos(𝜑2 −𝜑1 ) 𝑦2
from which 𝐴2 − 𝐴1 𝐴2
𝑥𝑦 + 𝐴2 = sin2 (𝜑2 − 𝜑1 ) as required.
1 2

Such motion is exhibited by the conical pendulum, which is a simple pendulum that does not swing
in a plane, but is free to swing in any direction. The motion of a conical pendulum (for small
amplitudes) can be considered as the superposition of two simple harmonic motions in perpendicular
directions. In the x-direction 𝑥 = 𝐴1 cos(𝜔𝑡 − 𝜑1 ) while in the y-direction 𝑦 = 𝐴2 cos(𝜔𝑡 − 𝜑2 ).
The angular frequencies in both directions are the same and so the bob of the pendulum traces the
path of a 1:1 Lissajous figure which is in general an ellipse.

𝜋
Example: Determine the shape of a 1:1: Lissajous figure having 𝐴1 = 𝐴2 = 𝐴 and 𝜑2 − 𝜑1 = 2 .

𝜋
𝑥2 2 cos 𝑦2 𝜋
Solution: The equation becomes 𝐴2 − 𝐴2
2
𝑥𝑦 + 𝐴2 = sin2 2 or 𝑥 2 + 𝑦 2 = 𝐴2 , which is a circle of
radius 𝐴. Of course a circle is just a special case of an ellipse.

Example: Determine the shape of a 1:1: Lissajous figure having 𝜑2 − 𝜑1 = 0.

𝑥2 2 cos 0 𝑦2 𝑥2 2𝑥𝑦 𝑦2
Solution: The equation becomes 𝐴2 − 𝐴1 𝐴2
𝑥𝑦 + 𝐴2 = sin2 0 or 𝐴2 − 𝐴 + 𝐴2 = 0. The left hand
1 2 1 1 𝐴2 2
𝑥 𝑦 2 𝐴2
side is a perfect square (not the shape!), so one has ( − ) = 0 or 𝑦 = 𝑥 which is a straight
𝐴1 𝐴2 𝐴1
line through the origin. A line is also just a special case of an ellipse.

These and other 1:1 Lissajous figures are shown in Figure 43.
 84

𝜋 𝜋
∆𝜑 = 0 ∆𝜑 = ∆𝜑 = Figure 43
4 2

1:1 Lissajous figures.

For these figures it has been

5𝜋
assumes that the vertical and
3𝜋 ∆𝜑 = 𝜋
∆𝜑 = ∆𝜑 = horizontal amplitudes are
4 4
equal, and the value of ∆𝜑 =
𝜑2 − 𝜑1 is given next to each
figure.

3𝜋 7𝜋 ∆𝜑 = 2𝜋
∆𝜑 = ∆𝜑 =
2 4

(b) For the case of 1:2 Lissajous figures, we consider two examples.

Example: Determine the form of a 1:2 Lissajous figure formed by the motion 𝑥 = 𝐴1 cos(𝜔𝑡) and
𝑦 = 𝐴2 cos(2𝜔𝑡).

Solution: To get the form of the Lissajous figure, we should try to relate 𝑥 and 𝑦 to each other, by
eliminating 𝜔𝑡. This is easy if one recalls that cos 2𝑢 = 2 cos2 𝑢 − 1. Then

𝑦 𝑥 2
= cos(2𝜔𝑡) = 2 cos 2 𝜔𝑡 − 1 = 2 ( ) − 1.
𝐴2 𝐴1

This is the equation of a parabola. Note that since −𝐴1 ≤ 𝑥 ≤ 𝐴1 and similarly −𝐴2 ≤ 𝑦 ≤ 𝐴2 , the
parabola occurs within a rectangular box in which the Lissajous figure must lie.

Example: Determine the form of a 1:2 Lissajous figure formed by the motion 𝑥 = 𝐴1 cos(𝜔𝑡) and
𝜋
𝑦 = 𝐴2 cos (2𝜔𝑡 − ).
2

Solution: We use a similar strategy as in the first example. Now

𝑦 𝜋
= cos (2𝜔𝑡 − ) = sin(2𝜔𝑡) = 2 sin 𝜔𝑡 cos 𝜔𝑡.
𝐴2 2

𝑦 2
We can square both sides to get ( ) = 4 sin2 𝜔𝑡 cos2 𝜔𝑡 = 4{1 − cos 2 𝜔𝑡} cos 2 𝜔𝑡, or
𝐴2

𝑦 2 𝑥 2 𝑥 2
(𝐴 ) = 4 {1 − (𝐴 ) } (𝐴 ) .
2 1 1

This is not a simple shape, but the figure of a bow-tie.

 85

(c) General Lissajous figures with arbitrary amplitudes, frequencies and phase constants can be very
complicated. There is little value trying to find the formulae for such patterns. However, they are
useful for comparing the relative frequencies of two signals. This was done originally by Lissajous for
sound waves and finds application today to test the frequency of high speed electronic signals e.g. in
computers and telecommunications networks. Suppose the oscillations are given by 𝑥 =
𝐴1 cos(𝜔𝑥 𝑡 − 𝜑1 ) and 𝑦 = 𝐴2 cos(𝜔𝑦 𝑡 − 𝜑2 ).

The rate at which the point (𝑥[𝑡]; 𝑦[𝑡]) forming the Lissajous figure will cut the y-axis is proportional
to the horizontal frequency 𝜔𝑥 , while the rate it will cut the x-axis is proportional to 𝜔𝑦 . If we
𝑛
therefore count the number of intercepts on the y-axis (𝑛𝑦 ) and on the x-axis (𝑛𝑥 ), one has that 𝜔𝑦 =
𝑥
𝑛𝑥
, so
𝜔𝑦

𝑛𝑥 𝜔𝑥 = 𝑛𝑦 𝜔𝑦 . (50)

Instead of using the number of intercepts though the y-axis and x-axis for 𝑛𝑦 and 𝑛𝑥 , one may instead
use the number of times the Lissajous figure touches the vertical and horizontal sides of the box
−𝐴1 ≤ 𝑥 ≤ 𝐴1 and −𝐴2 ≤ 𝑦 ≤ 𝐴2 .

𝑛𝑦 1 𝑛𝑦 1 𝑛𝑦 1
= = =
𝑛𝑥 2 𝑛𝑥 3 𝑛𝑥 4

Figure 44

General Lissajous figures.

𝑛𝑦 2 𝑛𝑦 3 𝑛𝑦 5
= = =
𝑛𝑥 1 𝑛𝑥 1 𝑛𝑥 1

𝑛𝑦 2 𝑛𝑦 2 𝑛𝑦 4
= = =
𝑛𝑥 3 𝑛𝑥 5 𝑛𝑥 5
 86

6.3. Problems

1. Convert 𝑥 = 5 cos(𝜔𝑡 − 0.55) to the cos+sin form.

2. Convert 𝑥 = 10 sin 𝜔𝑡 − 7 cos 𝜔𝑡 to the amplitude-phase form.

3. Consider the complex exponential form 𝑥 = 𝐶̃ 𝑒 𝑖𝜔𝑡 . Show that the size of the complex number 𝐶̃ gives the
amplitude of the oscillations.

4. A radio transmitter emits an electromagnetic wave that causes an oscillating electric field 𝐸 = 10 cos 𝜔𝑡.
A second radio transmitter nearby emits a wave causing an oscillation 𝐸 = 7 cos(𝜔𝑡 + 0.8). If the oscillations
are added in the same direction, what is the resulting amplitude and phase constant?

5. A pianist simultaneously hits notes A and C with frequencies 440 Hz and 523 Hz respectively. (a) What are
the angular frequencies of these sounds? (b) Assuming the waves are in phase (i.e. both have zero phase
constant) and have the same amplitude 𝐴, obtain an expression for the resulting waveform. (c) What is the
carrier frequency? (d) What are the envelope angular frequency and the envelope period? (e) What are the
beat period and frequency?

6. Consider the waveform above, created by adding two oscillations of equal amplitude and similar but
different angular frequency. (a) Obtain the period of the carrier from the graph. (b) Obtain the period of the
envelope from the graph. (c) What are the corresponding angular frequencies? (d) What were the angular
frequencies of the two oscillations that were originally added? (e) What were the amplitudes of the two
oscillations that were originally added?
 87

7. Damped harmonic motion

7.1. Damping

In this section we shall once again consider the motion of a particle attached to a spring, which is free to
move horizontally. However, we now assume that the table on which the particle lies is no longer smooth.
One could consider a variety of models for the damping i.e. the friction or resistance force. Here we will
simply assume for our model that the resistance is proportional to the magnitude of the velocity. Let 𝛽 be
the proportionality constant: then the resistance force is given by

𝑅 = −𝛽𝑣.

𝑅
Figure 45
𝐹𝑠
Damped oscillator.
ℓ 𝑥

There are two forces acting on the particle: the elastic force in the spring 𝐹𝑠 = −𝑘𝑥 and the resistance force
𝑅 = −𝛽𝑣. Newton's second law for this particle is
𝑚𝑎 = −𝛽𝑣 − 𝑘𝑥.

𝑘
If we divide through by 𝑚, define 𝜔 = √𝑚 as before and also define11

𝛽
2𝛾 = ,
𝑚

then this differential equation becomes

𝑎 + 2𝛾𝑣 + 𝜔2 𝑥 = 0. (51)

In this equation the acceleration depends on both the position and the velocity. When solving problems with
𝑑𝑣
a position dependent acceleration we made the substitution 𝑎 = 𝑣 , and this same substitution also
𝑑𝑥
worked for velocity dependent accelerations, so it is a good idea to try it here. It gives

11
The factor of 2 used with 𝛾 here will improve the form of some equations later on.
 88

𝑑𝑣
𝑣 + 2𝛾𝑣 + 𝜔2 𝑥 = 0
𝑑𝑥

for which, try as one might, one cannot separate the variables! Unfortunately the technique does not work.
But fortunately we also had the following theorem:

A differential equation of the form

𝑑𝑛 𝑥 𝑑𝑛−1 𝑥 𝑑2 𝑥 𝑑𝑥
𝑐𝑛 𝑛 + 𝑐𝑛−1 𝑛−1 + ⋯ + 𝑐2 2 + 𝑐1 + 𝑐0 𝑥 = 0
𝑑𝑡 𝑑𝑡 𝑑𝑡 𝑑𝑡

always has a solution of form 𝑥 = 𝑒 𝑟𝑡 for some value of 𝑟. It may be possible to find up to a maximum of 𝑛
different possible value for 𝑟, and if this is so then the general solution is

𝑥 = 𝐶1 𝑒 𝑟1 𝑡 + 𝐶2 𝑒 𝑟2 𝑡 + ⋯ +𝐶𝑛 𝑒 𝑟𝑛 𝑡 .

Our differential equation has the necessary form and we can find the solution by substituting in the trial
solution 𝑥 = 𝑒 𝑟𝑡 . This gives the polynomial

𝑟 2 + 2𝛾𝑟 + 𝜔2 = 0

with solution

𝑟 = −𝛾 ± √𝛾 2 − 𝜔 2 . (52)

We now distinguish between three cases:

 Undercritical damping when 𝛾 < 𝜔.

 Critical damping when 𝛾 = 𝜔.
 Overcritical damping when 𝛾 > 𝜔.

Notice that 𝛾 is a measure of the strength of the resistance, while 𝜔 is associated with the strength of the
spring (recall its definition). Roughly speaking, then, undercritical damping therefore occurs when the
resistance is small compared to the force in the spring while overcritical damping occurs when the resistance
is large compared to the force in the spring. In the following sections each of the three cases will be treated
separately.

7.2. Undercritical damping (𝜸 < 𝝎)

7.2.1. Motion of an undercritically doped oscillator

From Equation (52) we have that 𝑟 = −𝛾 ± √𝛾 2 − 𝜔 2. In this case 𝛾 < 𝜔 and so the part under the square
root will be negative, making 𝑟̃ a complex number. We can write 𝑟̃ = −𝛾 ± √(−1)(𝜔 2 − 𝛾 2 ) and define the
ordinary real number

𝜔0 = √𝜔 2 − 𝛾 2 , (53)
 89

so that 𝑟̃ = −𝛾 ± 𝑖𝜔0 , and the general solution is

𝑥 = 𝐶𝑒 (−𝛾+𝑖𝜔0 )𝑡 + 𝐷𝑒 (−𝛾−𝑖𝜔0 )𝑡 = 𝑒 −𝛾𝑡 {𝐶𝑒 𝑖𝜔0 𝑡 + 𝐷𝑒 −𝑖𝜔0 𝑡 }.

This is just 𝑒 −𝛾𝑡 times simple harmonic motion (and with the angular frequency 𝜔 replaced by 𝜔0 ). Since we
can write simple harmonic motion in several forms, so we can also write the motion for damped harmonic
motion in different forms, namely:

𝑥 = 𝑒 −𝛾𝑡 {𝐴 cos(𝜔0 𝑡 − 𝜑)} = (𝐴𝑒 −𝛾𝑡 ) cos(𝜔0 𝑡 − 𝜑) ;

𝑥 = 𝑒 −𝛾𝑡 {𝑃 cos 𝜔0 𝑡 + 𝑄 sin 𝜔0 𝑡};

𝑥 = 𝑒 −𝛾𝑡 {𝐶̃ 𝑒 𝑖𝜔0 𝑡 } = 𝐶̃ 𝑒 (−𝛾+𝑖𝜔0 )𝑡

where in the last expression the use of only the real part is implied. Notice from the first of these forms that
undercritically damped harmonic motion is equivalent to simple harmonic motion but with an exponentially
decreasing amplitude. This is shown in Figure 46. Defining the amplitude envelope 𝐴𝑒𝑛𝑣 [𝑡] = 𝐴𝑒 −𝛾𝑡 , one
gets 𝑥 = 𝐴𝑒𝑛𝑣 cos(𝜔0 𝑡 − 𝜑). In addition to the exponentially decaying amplitude, this differs from simple
harmonic motion in a second important respect: the angular frequency is now 𝜔0 , which is less than 𝜔 =
𝑘
√ as can be seen from equation (53).
𝑚

𝑥
𝑥(𝑡) = 𝐴𝑒 −𝛾𝑡 cos(𝜔0 𝑡 − 𝜑)

𝐴𝑒 −𝛾𝑡

Figure 46

𝑡 Undercritical damped
harmonic motion.

−𝐴𝑒 −𝛾𝑡

The velocity can be obtained by differentiation as follows:

𝑑𝑥 𝑑
𝑣= = 𝐴𝑒 −𝛾𝑡 cos(𝜔0 𝑡 − 𝜑) = 𝐴𝑒 −𝛾𝑡 (−𝜔0 sin(𝜔0 𝑡 − 𝜑)) + (−𝛾𝐴𝑒 −𝛾𝑡 ) cos(𝜔0 𝑡 − 𝜑)
𝑑𝑡 𝑑𝑡

∴ 𝑣 = −𝐴𝑒 −𝛾𝑡 (𝜔0 sin(𝜔0 𝑡 − 𝜑) + 𝛾 cos(𝜔0 𝑡 − 𝜑)).

7.2.2. Energy of an undercritically damped oscillator

Recall that the total energy stored in an undamped oscillator having angular frequency 𝜔 and amplitude 𝐴 is
1
given by 𝐸 = 2 𝑘𝐴2 . But the amplitude of a damped oscillator decays over time and therefore there must be
a loss of energy from the system to the surroundings over time. For an undercritically damped harmonic
oscillator the potential energy is
 90

1 1 1
𝐸𝑝 = 𝑘𝑥 2 = 𝑘{𝐴𝑒 −𝛾𝑡 cos(𝜔0 𝑡 − 𝜑)}2 = 𝑘𝐴2 𝑒 −2𝛾𝑡 cos2(𝜔0 𝑡 − 𝜑)
2 2 2

while the kinetic energy is

1 1
𝐸𝑘 = 𝑚𝑣 2 = 𝑚{−𝐴𝑒 −𝛾𝑡 (𝜔0 sin(𝜔0 𝑡 − 𝜑) + 𝛾 cos(𝜔0 𝑡 − 𝜑))}2 .
2 2

Clearly the kinetic energy expression is rather complicated, and we will not show all the details here, but
after some effort one can obtain the following expression for the total energy:

1 1
𝐸 = 𝑘𝐴2 𝑒 −2𝛾𝑡 + 𝑚𝐴2 𝑒 −2𝛾𝑡 {𝛾 2 cos(2(𝜔0 𝑡 − 𝜑)) + 𝜔0 𝛾 sin(2(𝜔0 𝑡 − 𝜑))}.
2 2

The first term in the expression is a decaying function of the time, while the second term both oscillates and
decays. The reason why the energy has an oscillating part is that energy loss is due to friction, which in our
model is proportional to the velocity. Since the velocity oscillates, so does the rate of energy loss. If we ignore
the second (oscillating) term we obtain the average energy as a function of time as simply 𝐸𝑎𝑣𝑒 =
1 1 1
2
𝑘𝐴2 𝑒 −2𝛾𝑡 = 2 𝑘(𝐴𝑒 −𝛾𝑡 )2 = 2 𝑘𝐴2𝑒𝑛𝑣 . Of course the energy is actually more complicated than this, but 𝐸𝑎𝑣𝑒
gives the trend of what is happening and can generally be used instead.

Example: Use the expression for the average energy of an undercritically damped harmonic oscillator to find
an expression for how long it takes before the average energy of the system is halved.

Solution: If 𝑡1/2 is the time required for the average energy to become half its original value, then it follows
𝐸𝑎𝑣𝑒 [𝑡] 1
that 𝐸𝑎𝑣𝑒 [𝑡 + 𝑡1/2 ] = 2
. Therefore since 𝐸𝑎𝑣𝑒 [𝑡] = 2 𝑘𝐴2 𝑒 −2𝛾𝑡 , we require

1 2 −2𝛾𝑡
1 2 −2𝛾(𝑡+𝑡 ) 2 𝑘𝐴 𝑒
𝑘𝐴 𝑒 1/2 = .
2 2

𝑒 −2𝛾𝑡 𝑒 −2𝛾𝑡 1
Therefore 𝑒 −2𝛾(𝑡+𝑡1/2 ) = , so 𝑒 −2𝛾𝑡 𝑒 −2𝛾𝑡1/2 = and 𝑒 −2𝛾𝑡1/2 = . Taking the logarithm on both
2 2 2
ln 2
sides, one gets −2𝛾𝑡1/2 = − ln 2, so that finally 𝑡1/2 = 2𝛾 .

7.2.3. Different quantities describing the damping

𝛽
The quantity 𝛽 gives an indication of the strength of the damping force, while the related quantity 𝛾 = 2𝑚
gives an indication of the associated acceleration. However, it is difficult to measure either 𝛽 or 𝛾 directly
from a graph of the motion. We shall now give some alternative ways to describe the damping.

(a) The relaxation time (𝝉)

For an undercritically damped oscillator the amplitude envelope decays exponentially according to
𝐴𝑒𝑛𝑣 = 𝐴𝑒 −𝛾𝑡 and the rate at which it decays gives an indication of the damping. The relaxation time
 91

1
𝜏 is defined as the time taken for the amplitude envelope to decay to = 𝑒 −1 ≈ 37% of its original
𝑒
value.12 The larger the relaxation time 𝜏, the weaker the damping. From a graph of the amplitude
envelope as a function of time one can obtain the relaxation time as in Figure 47.

𝐴 𝐴𝑒𝑛𝑣 [𝑡] = 𝐴𝑒 −𝛾𝑡

Figure 47

𝐴𝑒 −1 Exponentially decaying
amplitude envelope.

𝜏 t

𝐴𝑒𝑛𝑣 [𝑡]
If 𝜏 is the relaxation time then by its definition 𝐴𝑒𝑛𝑣 [𝑡 + 𝜏] = 𝑒
. Using 𝐴𝑒𝑛𝑣 = 𝐴𝑒 −𝛾𝑡 we get

𝐴𝑒 −𝛾𝑡
𝐴𝑒 −𝛾(𝑡+𝜏) =
𝑒
1
which gives 𝜏 = 𝛾, so the relaxation time is actually just the inverse of the damping constant.

(b) The logarithmic decrement (𝜹)

The logarithmic decrement (𝛿) is defined as the natural logarithm of the ratio of the amplitude
envelope of a damped oscillator at any given time and then one (damped) period later.
Mathematically this definition can be expressed as

𝐴𝑒𝑛𝑣 [𝑡]
𝛿 = ln .
𝐴𝑒𝑛𝑣 [𝑡 + 𝑇0 ]

This provides a practical way to measure 𝛿 from experimental data. The larger the logarithmic
decrement, the stronger is the damping. Using 𝐴𝑒𝑛𝑣 = 𝐴𝑒 −𝛾𝑡 we get

𝐴𝑒 −𝛾𝑡 1 𝑇0
𝛿 = ln −𝛾(𝑡+𝑇 )
= ln −𝛾𝑇 = 𝛾𝑇0 = .
𝐴𝑒 0 𝑒 0 𝜏

Therefore the logarithmic decrement gives the ratio of the (damped) period to the relaxation time.
If this ratio is small i.e. 𝑇0 < 𝜏 then many oscillations occur during one relaxation time i.e. the
damping is weak.

(c) The quality factor (𝑸)

1
For a damped oscillator the average energy decays exponentially according to 𝐸𝑎𝑣𝑒 = 2 𝑘𝐴2 𝑒 −2𝛾𝑡
and the rate at which it decays also gives an indication of the damping. The quality factor 𝑄 is defined

12
For radioactive decay this fraction is chosen as 50% and the time taken is called the half life.
 92

1
as the phase angle (in radians) required for the average energy to decay to = 𝑒 −1 ≈ 37% of its
𝑒
original value. The larger the quality factor 𝑄, the less strong is the damping. If 𝑡𝑄 is the time (not
𝐸𝑎𝑣𝑒 [𝑡]
phase angle) for the required decay then it follows that 𝐸𝑎𝑣𝑒 [𝑡 + 𝑡𝑄 ] = 𝑒
. Therefore

1 2 −2𝛾𝑡
1 2 −2𝛾(𝑡+𝑡 ) 2 𝑘𝐴 𝑒
𝑘𝐴 𝑒 𝑄 =
2 𝑒
1 𝜏 2𝜋 𝜏 𝜋
which gives 𝑡𝑄 = 2𝛾 = 2. During this time the increase in the phase angle is 𝑄 = 𝜔0 𝑡𝑄 = 𝑇0 2
= 𝛿.

(d) The fraction of energy lost per oscillation

At some moment 𝑡 the average energy of a damped oscillator is given by 𝐸𝑎𝑣𝑒 [𝑡]. After one cycle or
oscillation the average energy has reduced to 𝐸𝑎𝑣𝑒 [𝑡 + 𝑇0 ]. The ratio of the energy stored initially to
𝐸𝑎𝑣𝑒 [𝑡] 1
that lost in the next cycle is 𝐸 [𝑡]−𝐸
. Using 𝐸𝑎𝑣𝑒 [𝑡] = 2 𝑘𝐴2 𝑒 −2𝛾𝑡 , this ratio is given by
𝑎𝑣𝑒 𝑎𝑣𝑒 [𝑡+𝑇0 ]

1 2 −2𝛾𝑡
𝑘𝐴 𝑒 𝑒 −2𝛾𝑡 1 1 1
2 = −2𝛾𝑡 = = =
1 2 −2𝛾𝑡 1 2 −2𝛾(𝑡+𝑇0 ) 𝑒 − 𝑒 −2𝛾𝑡−2𝛾𝑇0 1 − 𝑒 −2𝛾𝑇0 1 − 𝑒 −2𝛿 2𝜋 .
−
2 𝑘𝐴 𝑒 − 2 𝑘𝐴 𝑒 1 − 𝑒 𝑄

1 𝑄
But 𝑒 𝑢 ≈ 1 + 𝑢 for small 𝑢, so if 𝑄 ≫ 2𝜋 then the ratio becomes 2𝜋 ≈ 2𝜋 and so
1−(1− )
𝑄

Energy stored
𝑄 ≈ 2𝜋 ( ).
Energy lost in next cycle

7.3. Overcritical damping (𝜸 > 𝝎)

In this case we consider what happens when the damping is large. From equation (52) we have that 𝑟 =
−𝛾 ± √𝛾 2 − 𝜔 2. Since for overcritical damping 𝛾 > 𝜔, the part under the square root will be positive and
the values of 𝑟 will be normal (real) numbers. Define

𝛼 = √𝛾 2 − 𝜔 2 , (54)

Then 𝑟 = −𝛾 ± 𝛼 and the general solution for the motion is

𝑥 = 𝐶𝑒 (−𝛾+𝛼)𝑡 + 𝐷𝑒 (−𝛾−𝛼)𝑡

where the constants 𝐶 and 𝐷 are obtained from the initial conditions. Since 𝛾 > 𝛼 both terms decay
exponentially and there are no oscillations. By differentiating we can get the velocity

𝑑𝑥
𝑣= = (−𝛾 + 𝛼)𝐶𝑒 (−𝛾+𝛼)𝑡 − (𝛾 + 𝛼)𝐷𝑒 (−𝛾−𝛼)𝑡 .
𝑑𝑡
 93

7.4. Critical damping (𝜸 = 𝝎)

From equation (52) we have that 𝑟 = −𝛾 ± √𝛾 2 − 𝜔 2. Since for critical damping 𝛾 = 𝜔 the part under the
square root will be zero and 𝑟 = −𝛾. Since we have only one, not two, values of 𝑟 the solution from the
theorem in section 7.1, namely 𝑥 = 𝐶𝑒 −𝛾𝑡 , is not the general solution. One can show that 𝑥 = 𝐷𝑡𝑒 −𝛾𝑡 is also
a solution to equation (51) for critical damping, and so the general solution is

𝑥 = 𝐶𝑒 −𝛾𝑡 + 𝐷𝑡𝑒 −𝛾𝑡 = (𝐶 + 𝐷𝑡)𝑒 −𝛾𝑡 .

By differentiating we can get the velocity

𝑑𝑥
𝑣= = (𝐶 + 𝐷𝑡)(−𝛾𝑒 −𝛾𝑡 ) + (𝐷)𝑒 −𝛾𝑡 = (𝐷 − 𝛾𝐶 − 𝛾𝐷𝑡)𝑒 −𝛾𝑡 .
𝑑𝑡

When a system is critically damped, it does not undergo oscillations, which is similar to the case of overcritical
damping. But for critical damping the resistance is less than for overcritical damping, which can have
important implications.

Example: Explain why the suspension of a car should be set for critical rather than undercritical or overcritical
damping.

Solution: The suspension system consists of springs, as well as shock absorbers which provide the damping
force. Undercritical damping occurs if the shock absorbers are worn out, and the car will oscillate when set
in motion which is undesirable. Overcritical damping will occur if the shock absorbers are very strong, offering
too much resistance. In this case any bump in the road will be transferred to the car and passengers, which
is also undesirable. The optimum setting of the suspension is critical damping.

Example: The front door of a building is used by a large number of people. Since the building has air-
conditioning, the door must be kept closed, and so it is fitted with a spring. Explain why it should also be
fitted with a damping device and what type of damping is the best.

Solution: If the door is only fitted with a spring but no damping device, then when it is released it will slam
shut. This will make a noise, someone may get hurt and the door may be damaged, so clearly it is better if
the door closes gently. This can be accomplished by fitting a damping device. If the damping is undercritical
then the door would tend to oscillate (or if this was prevented by the door frame, the door would still slam
shut). Therefore the damping should be either critical or overcritical. But overcritical damping is not desirable
because the excessive resistance would make the door harder to open, and once opened the excessive
resistance would mean it would take a long time to close. That could annoy people and also increase the air-
conditioning costs. Critical damping would prevent oscillations, while still keeping the door easiest to open
and fastest to close, and is therefore the optimum damping.
 94

7.5. Comparison of types of damping

Figure 48

Types of
damping

Top left:
No damping

Top right:
Undercritical

Bottom left:
Critical

Bottom right:
Overcritical

Motion graphs for the different types of damping are shown in Figure 49. Compared to the undamped case,
the undercritically damped system also shows oscillations but the amplitude decreases and the damped
period is longer than the undamped period. Neither the critically damped or the overdamped systems show
oscillations, but the overcritically damped system takes longer to return to the equilibrium position.

7.6. Problems

1. A particle with mass 𝑚 hangs from a spring with natural length ℓ, spring constant 𝑘 and elongation 𝑦.
When the particle moves with velocity 𝑣, it experiences a resistive force with magnitude 𝛽𝑣. (a) Show that
the motion of the particle is governed by the equation 𝑎 + 2𝛾𝑣 + 𝜔2 𝑦 − 𝑔 = 0. You must give definitions
for 𝛾 and 𝜔 in terms of 𝑘, 𝛽 and 𝑚. (b) Substitute the trial solution 𝑦[𝑡] = 𝑢[𝑡] + 𝑐 where 𝑐 is a constant,
𝑑2𝑢 𝑑𝑢
and find the value of 𝑐 that will produce the new equation + 2𝛾 + 𝜔2 𝑢 = 0. (c) Now consider the new
𝑑𝑡 2 𝑑𝑡
𝑑2 𝑢 𝑑𝑢
equation + 2𝛾 + 𝜔2 𝑢 = 0 and substitute the trial solution 𝑢(𝑡) = 𝑒 𝑟𝑡 to show that it has the general
𝑑𝑡 2 𝑑𝑡
(−𝛾+√𝛾 2 −𝜔2 )𝑡 (−𝛾−√𝛾 2 −𝜔2 )𝑡
solution 𝑢(𝑡) = 𝐶𝑒 + 𝐷𝑒 . (d) If 𝛾 = 5 and 𝜔 = 4, find the height as a function of
time and state the type of damping.

𝛽2
2. (a) State the condition for undercritical damping and show that it leads to 𝑚 > 4𝑘. (b) Suppose a system is
critically damped. If its mass is increased while leaving everything else unchanged, will it become
undercritically or overcritically damped? Explain your answer.

𝑚𝑔
3. A particle hanging from a spring and set in motion has position 𝑦 = 𝑘
+ 𝐴𝑒 −𝛾𝑡 cos(𝜔0 𝑡 − 𝜑) if the
system is undercritically damped. Obtain the velocity of the particle at any time by differentiating and then
1 𝛾
show the particle comes momentarily to rest after a time 𝑡 = 𝜔 (arctan (− 𝜔 ) + 𝜑).
0 0
 95

4. The motion of an undercritically damped system is given by 𝑥 = 𝑒 −𝛾𝑡 (𝑃 cos 𝜔0 𝑡 + 𝑄 sin 𝜔0 𝑡). The mass
𝛾𝑥0
is released from rest with 𝑥 = 𝑥0 . Show that 𝑃 = 𝑥0 and 𝑄 = . Convert the form of your solution to the
𝜔0

𝛾 2
amplitude-phase form and show that 𝐴 = 𝑥0 √1 + (𝜔 ) .
0

5. A particle with mass 60 g hangs from a spring with natural length 30 cm. The spring stretches 1.5 cm under
the weight of the particle. (a) Find the spring constant. (b) When the particle is set in motion, it experiences
a resistive force 𝑅 = −𝛽𝑣 and oscillates with damped angular frequency 𝜔0 = 24 rad/s. Calculate the
damping constant 𝛾 and also the value of 𝛽 (including units).

6. A 5 kg mass is attached to a spring with spring constant 𝑘 = 100 N/m and moves along a horizontal surface
which produces a resistive force 𝑅 = −𝛽𝑣 where 𝛽 = 8 Nm-1s. (a) Calculate what the angular frequency 𝜔
would be if there were no resistive force/damping. Is the damped angular frequency greater or less than this?
(b) Calculate the damping constant 𝛾 and state what type of damped motion (undercritical, critical or
overcritical) occurs. (c) Calculate the actual (damped) angular frequency 𝜔0 of the motion. (d) The position
of the object at any time is 𝑥 = 𝑒 −𝛾𝑡 {𝑃 cos 𝜔0 𝑡 + 𝑄 sin 𝜔0 𝑡}. If the object is released from rest at initial
position 𝑥0 = 0.1 m, find the values of 𝑃 and 𝑄. (e) Convert your solution to the form 𝑥 = 𝐴𝑒 −𝛾𝑡 cos(𝜔0 𝑡 −
𝜑) and calculate how long it takes the mass to reach the equilibrium point (𝑥 = 0) for the first time. (f)
Differentiate your solution in the new form to get the velocity and calculate how long the mass takes to reach
its furthest distance from the starting point as well as how far this maximum point is from the initial position
[Hint: 𝑣 = 0 at the furthest point.]

7. An undercritically damped object moves according to 𝑥 = 10𝑒 −0.5𝑡 cos(6𝑡 + 1). (a) How long does the
object take to reach the equilibrium position? (b) Differentiate this to find the velocity and determine the
time for the object to reach its furthest point from the initial position.

8. The graph above shows the position of an undercritically damped oscillator as a function of time. Also
shown are the amplitude envelope and a dashed line indicating its initial value divided by 𝑒. (a) Measure the
damped period of oscillation 𝑇0 and calculate the damped angular frequency 𝜔0 . (b) Obtain the relaxation
 96

time from the graph, giving your answer to one decimal place. Use it to find the damping constant and hence
the spring constant if the mass is 0.4 kg.

9. A bank has a security door at the entrance, which is fitted with a spring providing a torque proportional to
the angle the door is open, so that the door closes by itself. The manager wants the door to close with the
shortest time, so that people do not have to wait longer than necessary. However, he is aware that he cannot
make the spring force too strong, or else the door will slam shut on somebody's fingers, and they will sue the
bank. For this reason, he wants the door to come to a close slowly, and not to pass (over-shoot) its equilibrium
condition (the equilibrium position is when the door is closed). The manager has the door fitted with a
damping device which provides a damping torque proportional to the door's angular velocity, but in the
opposite direction. (a) Explain why the damping needs to be critical, rather than undercritical or overcritical.
(b) Suppose the door's spring provides torque 𝜏 = −𝜅𝜃, where 𝜃 is the angle how far the door is open, and
𝑑𝜃
the damping device provides torque 𝜏 = −𝛽 𝑑𝑡 . Use Newton's second law in its rotational form to set up the
differential equation for the door's motion and find suitable definitions for 𝜔 and 𝛾 so that it takes the form
𝑑2 𝜃 𝑑𝜃
𝑑𝑡 2
+ 2𝛾 𝑑𝑡 + 𝜔2 𝜃 = 0. (c) Find an expression for the required value of 𝛽 to obtain critical damping, in terms
of the spring constant and moment of inertia of the door. (d) The general solution for the angular position of
the door if it is critically damped is 𝜃 = (𝐶 + 𝐷𝑡)𝑒 −𝛾𝑡 . Differentiate this to get the angular velocity at any
5𝜋
time, and find the specific solution if 𝛾 = 2.5 rad/s and the door is released from rest from an initial angle
12
rad.

10. A fisherman casts out 100 m of line. His sinker unfortunately gets caught on a rock at the bottom of the
sea. He decides to reel in, in the hope that the line pulls free. As he does so, the line stretches, and after a
while he has reeled in 10 m of line, leaving what would be 90 m of unstretched line, but now stretched across
the 100 m between him and where the sinker is caught. The fishing line is elastic and provides a restoring
force proportional to the distance it has been stretched (but unlike a spring, the fishing line cannot provide
a restoring force when it is compressed). Suddenly the sinker breaks off and the hook is pulled back by the
restoring force of the fishing line. However, the hook is entangled with seaweed which is dragged with it. The
hook with seaweed has a mass of 300 g and experiences a resistive force proportional to its velocity. This
resistive force is given by 𝑅 = −𝛽𝑣 where 𝛽 = 5.1 Nm-1s. Tests show that the fishing line has a spring
𝜆
constant 𝑘 = ℓ where ℓ is its natural length and 𝜆 = 200 N. The fisherman does not reel in any more after
the sinker breaks off. (a) Show that the spring constant of the fisherman's line (after reeling in) is 𝑘 = 2.2
N/m. (b) Calculate the damping constant due to the water resistance as the hook with seaweed is pulled back
towards the fisherman. (c) Calculate the value of 𝜔 for this situation and show that over-critical damping
occurs. (d) The general solution for overcritical damping is 𝑥 = 𝐶𝑒 (−𝛾+𝛼)𝑡 + 𝐷𝑒 (−𝛾−𝛼)𝑡 . Find the value of 𝛼.
(e) Taking 𝑥 = 0 at the equilibrium position of the fishing line (90 m from the fisherman), and 𝑡 = 0 at the
moment the sinker breaks off, give the initial position and initial velocity. (f) Differentiate the equation for
the position to obtain the velocity at any time, and substitute the initial conditions to obtain the integration
constants. (g) Calculate the position of the hook after 1 s.
 97

8. Sinusoidally driven harmonic motion

8.1. Introduction

Driven harmonic motion is the motion of a damped harmonic oscillator which is driven (or forced) by an
external force. Consider the motion of a damped harmonic oscillator with mass 𝑚 attached to a spring with
spring constant 𝑘, causing a restoring force 𝐹𝑠 = −𝑘𝑥. There is also resistance of the form 𝑅 = −𝛽𝑣 and in
addition to these two forces, we assume that there is an extra time dependent force 𝐹[𝑡] acting on the
object. Newton's second law then gives

𝑚𝑎 = 𝐹𝑠 + 𝑅 + 𝐹[𝑡] = −𝑘𝑥 − 𝛽𝑣 + 𝐹(𝑡).

𝑘 𝛽
As before we shall divide through by 𝑚 and use the definitions 𝜔 = √𝑚 and 2𝛾 = 𝑚, giving:

𝐹[𝑡]
𝑎 + 2𝛾𝑣 + 𝜔2 𝑥 = . (55)
𝑚

The simplest time dependent force which is often studied is a force whose magnitude changes sinusoidally.
In mechanical systems this force can be generated by a rotating wheel with a peg that is mounted off-centre
and a rod 𝑃𝐴 from the peg and to the end of spring as shown in Figure 50.13

𝑟 𝐴 𝑅
𝐹(𝑡)
𝜔𝑑 𝐹𝑠

Figure 50 Driven harmonic oscillator.

The rod pushes the left hand side of the spring to-and-fro as the wheel rotates with angular frequency 𝜔𝑑 .
The right hand side of the rod (𝐴) to which the spring is attached moves horizontally and its position can be
approximated by 𝑥𝐴 ≈ 𝑟 cos(𝜔𝑑 𝑡 − 𝜑𝑑 ).14 The phase constant 𝜑𝑑 depends on the initial angular position of
the peg 𝑃.

13
Although we derive the theory here for this mechanical system, the theory is general and applies to any damped harmonic
oscillator that is driven by an external oscillating force. For example, molecules which vibrate may be considered as damped harmonic
oscillators and when they are excited (driven) by electromagnetic radiation then the theory derived here can be used to describe
their behaviour.

14
If one does not wish to rely on this approximation, the so-called “Scotch
yoke” configuration can be used. In this case the motion of the horizontal arm
is perfectly sinusoidal.
 98

When the left hand side of the spring is moved a distance 𝑥𝐴 and the right hand side (attached to the mass)
is displaced a distance 𝑥 then the total extension of the spring is given by 𝑥 − 𝑥𝐴 , and so the spring force is
−𝑘(𝑥 − 𝑥𝐴 ) = −𝑘𝑥 + 𝑘𝑥𝐴 = −𝑘𝑥 + 𝑘𝑟 cos(𝜔𝑑 𝑡 − 𝜑𝑑 ). Thus, in addition to the “usual” spring force −𝑘𝑥
there is also the additional sinusoidal force 𝐹[𝑡] = 𝑘𝑟 cos(𝜔𝑑 𝑡 − 𝜑𝑑 ). Although this force is in fact applied
through the spring, it originates as a result of the rotating wheel and we can think of it separately as an
additional time dependent driving force.15 Newton’s second law for this system becomes

𝑘𝑟
𝑎 + 2𝛾𝑣 + 𝜔2 𝑥 = cos(𝜔𝑑 𝑡 − 𝜑𝑑 ) = 𝜔2 𝑟 cos(𝜔𝑑 𝑡 − 𝜑𝑑 ). (56)
𝑚

We wish to solve this equation to find the motion of the sinusoidally driven harmonic oscillator. In the next
sections we will present two methods to achieve this. We call the result the “steady state solution” because
we will see much later that an even more general solution exists.

8.2. The steady state solution

Now when a system is driven at some angular frequency 𝜔𝑑 one might guess that it should oscillate at this
same angular frequency. Therefore we seek a solution 𝑥[𝑡] that is an oscillation having the same angular
frequency as the driving angular frequency. Suppose it has the form

𝑥[𝑡] = 𝑃 cos(𝜔𝑑 𝑡 − 𝜑𝑑 ) + 𝑄 sin(𝜔𝑑 𝑡 − 𝜑𝑑 )

where we have used the cos+sin form of simple harmonic motion and matched the phase to that of the
driving force 𝐹[𝑡] = 𝑘𝑟 cos(𝜔𝑑 𝑡 − 𝜑𝑑 ). The velocity would then be

𝑣 = −𝜔𝑑 𝑃 sin(𝜔𝑑 𝑡 − 𝜑𝑑 ) + 𝑄𝜔𝑑 cos(𝜔𝑑 𝑡 − 𝜑𝑑 )

and the acceleration would be

𝑎 = −𝜔𝑑2 𝑃 cos(𝜔𝑑 𝑡 − 𝜑𝑑 ) − 𝜔𝑑2 𝑄 sin(𝜔𝑑 𝑡 − 𝜑𝑑 ).

We substitute these into equation (56) to obtain

−𝜔𝑑2 𝑃 cos(𝜔𝑑 𝑡 − 𝜑𝑑 ) − 𝜔𝑑2 𝑄 sin(𝜔𝑑 𝑡 − 𝜑𝑑 ) + 2𝛾{−𝜔𝑑 𝑃 sin(𝜔𝑑 𝑡 − 𝜑𝑑 ) + 𝑄𝜔𝑑 cos(𝜔𝑑 𝑡 − 𝜑𝑑 )} +

𝜔2 {𝑃 cos(𝜔𝑑 𝑡 − 𝜑𝑑 ) + 𝑄 sin(𝜔𝑑 𝑡 − 𝜑𝑑 )} = 𝜔2 𝑟 cos(𝜔𝑑 𝑡 − 𝜑𝑑 ).

Considering the cosine terms one gets −𝜔𝑑2 𝑃 + 2𝛾𝑄𝜔𝑑 + 𝜔2 𝑃 = 𝜔2 𝑟, while considering the sine terms one
gets −𝜔𝑑2 𝑄 − 2𝛾𝜔𝑑 𝑃 + 𝜔2 𝑄 = 0. These equations can be solved simultaneously to yield

𝜔2 𝑟(𝜔2 − 𝜔𝑑2 ) 𝜔2 𝑟(2𝛾𝜔𝑑 )

𝑃= and 𝑄 = .
(𝜔 2 − 𝜔𝑑2 )2 + (2𝛾𝜔𝑑 )2 (𝜔 2 − 𝜔𝑑2 )2 + (2𝛾𝜔𝑑 )2

The fact that we have managed to find 𝑃 and 𝑄 shows that our trial solution has worked.

15
Note that it is possible to construct a system where the applied force is truly external, and not applied “through” the spring force
as in this case. An example would be a magnetic particle placed in an oscillating magnetic field – in this case the spring could be
attached at a fixed position, giving a spring force −𝑘𝑥 and the magnetic force on the particle would give a truly external time-
dependent force. However, such an experiment is more difficult to set up than the simple mechanical system presented here.
 99

However, it is rather complicated, and a nice simplification occurs if we convert to the amplitude phase form
as follows: suppose 𝑥(𝑡) = 𝑃 cos(𝜔𝑑 𝑡 − 𝜑𝑑 ) + 𝑄 sin(𝜔𝑑 𝑡 − 𝜑𝑑 ) is the same as

𝑥(𝑡) = 𝐴 cos(𝜔𝑑 𝑡 − 𝜑𝑑 − 𝜑) = 𝐴 cos(𝜔𝑑 𝑡 − 𝜑𝑑 ) cos 𝜑 + 𝐴 sin(𝜔𝑑 𝑡 − 𝜑𝑑 ) sin 𝜑.

We require for the conversion that 𝐴 cos 𝜑 = 𝑃 and 𝐴 sin 𝜑 = 𝑄.

𝜔2 𝑟 𝜔2 𝑟
 Therefore 𝐴 = √𝑃2 + 𝑄 2 = 2 2
√(𝜔 2 − 𝜔𝑑2 )2 + (2𝛾𝜔𝑑 )2 = .
(𝜔2 −𝜔𝑑 ) +(2𝛾𝜔𝑑 )2 2 2 +(2𝛾𝜔 )2
√(𝜔2 −𝜔𝑑 ) 𝑑
𝑄 2𝛾𝜔
 Also tan 𝜑 = 𝑃 = 𝜔2 −𝜔𝑑2 . Since 𝐴 sin 𝜑 = 𝑄 and 𝑄 is always positive, 𝜑 is always in quadrants I or II. If
𝑑
𝜔𝑑 < 𝜔 (slow driving force) then tan 𝜑 is positive and the angle will lie in the first quadrant, otherwise
2𝛾𝜔𝑑
when 𝜔𝑑 > 𝜔 (fast driving force) the angle will lie in the second quadrant. If 𝜔𝑑 = 𝜔 then tan 𝜑 = 0
𝜋
which is not defined, and so 𝜑 = 2 .

8.3. An alternative way to obtain the steady state solution

We show a second way to solve equation (56), this time using complex numbers. No new results are obtained,
but the technique may be of interest. Let us write the right hand side of equation (56) in the complex form
i.e. let 𝜔2 𝑟 cos(𝜔𝑑 𝑡 − 𝜑𝑑 ) = 𝐹̃ 𝑒 𝑖𝜔𝑑 𝑡 and let us take a trial solution 𝑥 = 𝐶̃ 𝑒 𝑖𝜔𝑑 𝑡 with matching frequency.
Our aim is to find the value of 𝐶̃ . Substituting this into equation (56) gives

−𝜔𝑑2 𝐶̃ 𝑒 𝑖𝜔𝑑 𝑡 + 2𝛾𝑖𝜔𝑑 𝐶̃ 𝑒 𝑖𝜔𝑑 𝑡 + 𝜔2 𝐶̃ 𝑒 𝑖𝜔𝑑 𝑡 = 𝐹̃ 𝑒 𝑖𝜔𝑑 𝑡 ,

or, after dividing through by the exponential part, we get (𝜔2 − 𝜔𝑑2 + 2𝛾𝑖𝜔𝑑 )𝐶̃ = 𝐹̃ . The part in brackets is
a complex number with a positive imaginary part 2𝛾𝜔𝑑 and a real part 𝜔2 − 𝜔𝑑2 . Its magnitude (from the
2𝛾𝜔
theorem of Pythagoras) is √(𝜔 2 − 𝜔𝑑2 )2 + (2𝛾𝜔𝑑 )2 while its angle is given by tan 𝜑 = 𝜔2 −𝜔𝑑2 . Therefore
𝑑

𝐹̃ 𝑒 −𝑖𝜑
(√(𝜔 2 − 𝜔𝑑2 )2 + (2𝛾𝜔𝑑 )2 𝑒 𝑖𝜑 ) 𝐶̃ = 𝐹̃ , so 𝐶̃ = .
√(𝜔 2 − 𝜔𝑑2 )2 + (2𝛾𝜔𝑑 )2

Therefore the motion of the driven harmonic oscillator is

𝐹̃ 𝑒 −𝑖𝜑 𝐹̃ 𝑒 𝑖(𝜔𝑑 𝑡−𝜑)

𝑥 = 𝐶̃ 𝑒 𝑖𝜔𝑑 𝑡 = 𝑒 𝑖𝜔𝑑 𝑡 = .
√(𝜔 2 − 𝜔𝑑2 )2 + (2𝛾𝜔𝑑 )2 √(𝜔 2 − 𝜔𝑑2 )2 + (2𝛾𝜔𝑑 )2

But recall that we defined 𝐹̃ 𝑒 𝑖𝜔𝑑 𝑡 = 𝜔2 𝑟 cos(𝜔𝑑 𝑡 − 𝜑𝑑 ), so 𝐹̃ 𝑒 𝑖(𝜔𝑑 𝑡−𝜑) = 𝜔2 𝑟 cos(𝜔𝑑 𝑡 − 𝜑𝑑 − 𝜑) and
the solution is then

𝜔2 𝑟
𝑥= cos(𝜔𝑑 𝑡 − 𝜑𝑑 − 𝜑).
√(𝜔 2 − 𝜔𝑑2 )2 + (2𝛾𝜔𝑑 )2

This is the same result as obtained in the previous section.

 100

8.4. Effect of driving angular frequency

We have found that when a damped oscillator is driven sinusoidally, the steady state motion has amplitude

𝜔2 𝑟
𝐴[𝜔𝑑 ] =
(57)
√(𝜔 2 − 𝜔𝑑2 )2 + 4𝛾 2 𝜔𝑑2

and lags behind the driving force by a phase constant

2𝛾𝜔𝑑
𝜑[𝜔𝑑 ] = arctan (58)
𝜔 2 − 𝜔𝑑2

which lies in the first or second quadrants i.e. 0 ≤ 𝜑 < 𝜋. Both the amplitude and phase lag depend on the
angular driving frequency i.e. they vary depend on how fast the system is being driven.

𝛾 = 0.1 Hz Figure 51
𝛾 = 0.2 Hz
𝜋 𝛾 = 0.3 Hz Phase angle as a
2 function of driving
𝛾 = 0.4 Hz
angular frequency for
𝛾 = 0.5 Hz various damping
factors.
𝜔 = 1.0 Hz

0 𝜔 𝜔𝑑

Let us first consider the phase lag 𝜑. Figure 51 shows 𝜑 as a function of 𝜔𝑑 for different values of 𝛾. In all
cases as 𝜔𝑑 increasing from 0 to ∞, 𝜑 increases from 0 to 𝜋. This means that for small driving angular
frequencies the motion is almost in step with the driving force, but as the driving angular frequency increases
the motion becomes more out of step with the driving force, until at very high driving frequencies they are
2𝛾𝜔𝑑 𝜋
completely out of phase. When 𝜔𝑑 = 𝜔 we have tan 𝜑 = 0
which is not defined, and so 𝜑 = 2 . The rise
from 0 to 𝜋 is sharp when the damping 𝛾 is small, but more gradual when 𝛾 is large.

Next we consider the amplitude 𝐴. Figure 52 shows 𝐴 as a function of 𝜔𝑑 for different values of 𝛾.

 When the driving angular frequency is very slow (𝜔𝑑 → 0) the amplitude 𝐴 → 𝑟 which is the radius of
the driving wheel.
 When the driving angular frequency is very fast (𝜔𝑑 → ∞) the amplitude 𝐴 → 0. What this means
physically is that when one tries to drive a system extremely fast, it hardly has time to begin moving in
 101

one direction before the driving force has switched direction and as a result the amplitude of the motion
is very small.

𝜔 = 1.0 Hz

𝛾 = 0.1 Hz Figure 52
𝛾 = 0.2 Hz
𝛾 = 0.3 Hz Amplitude as a function
𝛾 = 0.4 Hz of driving angular
frequency for different
𝛾 = 0.5 Hz damping factors.

0
𝜔𝑑

At some intermediate driving angular frequency the amplitude may reach a maximum value. We say that
resonance occurs and this special driving angular frequency is called the resonance angular frequency (𝜔𝑅 ).
𝑑𝐴
It can be found by setting 𝑑𝜔 = 0 i.e.
𝑑

𝑑 𝜔2 𝑟
= 0.
𝑑𝜔𝑑
√(𝜔 2 − 𝜔𝑑2 )2 + 4𝛾 2 𝜔𝑑2
( )

However, it is not necessary to compute the whole derivative since the part above the line is constant. We
𝑑 2
need only consider 𝑑𝜔 ((𝜔2 − 𝜔𝑑2 ) + 4𝛾 2 𝜔𝑑2 ) = 2(𝜔2 − 𝜔𝑑2 )(−2𝜔𝑑 ) + 8 𝛾 2 𝜔𝑑 = 0 which, when solved
𝑑
for 𝜔𝑑 , yields the resonance angular frequency

𝜔𝑅 = √𝜔 2 − 2𝛾 2 . (59)
𝜔
Note that the part under the square root becomes negative if 𝛾 > . If this is the case then the damping is
√2
too large for resonance to occur. Otherwise, the corresponding maximum amplitude 𝐴𝑅 at resonance is
obtained by substituting equation (59) into equation (57) to give

𝜔2 𝑟 𝜔2 𝑟
𝐴𝑅 = 𝐴[𝜔𝑅 ] = = . (60)
√(𝜔 2 − {𝜔 2 − 2𝛾 2 })2 + 4𝛾 2 {𝜔 2 − 2𝛾 2 } 2𝛾√𝜔 2 − 𝛾 2

Figure 52 shows that both the resonance angular frequency and resonance amplitude decrease as the
damping increases. We can also substitute equation (59) into equation (58) to get the resonance phase lag

2𝛾√𝜔 2 − 2𝛾 2 √𝜔 2 − 2𝛾 2
𝜑𝑅 = 𝜑[𝜔𝑅 ] = arctan = arctan . (61)
𝜔 2 − {𝜔 2 − 2𝛾 2 } 𝛾
 102

8.5. The quality factor as an amplification factor

The ratio of the amplitude at resonance to the amplitude when the driving is very slow is

𝐴𝑅 𝜔2 𝜔
= ≈
𝑟 2𝛾√𝜔 2 − 𝛾 2 2𝛾

𝜋 𝑇0 2𝜋𝛾
for small damping. Earlier in section 7.2.3 we found the quality factor was given by 𝑄 = 𝛿 . Since 𝛿 = 𝜏
= 𝜔0
𝜋 𝜔0 𝜔
we have that 𝑄 = 2𝜋𝛾 = 2𝛾
≈ 2𝛾 for small damping. Therefore, for lightly damped systems the quality factor
𝜔0
𝐴𝑅
gives a good indication of the resonance amplification factor 𝑟
.

8.6. A complete solution

This chapter has dealt with the solution of equation (56) and until now we have referred to our solution as
the “steady state solution”. The fact is that this only a partial solution to the equation. One may see this by
realising that it contains no integration constants and does not depend on the initial conditions of the system.
Yet they must have a role to play. Equation (56) is a second order differential equation and its general solution
must therefore have two arbitrary constants. We shall now show where these appear by expanding our
solution to include a transient part.

If we represent our steady state solution of equation (56) by 𝑥ss then by definition

𝑑2 𝑥ss 𝑑𝑥ss
2
+ 2𝛾 + 𝜔2 𝑥ss = 𝜔2 𝑟 cos(𝜔𝑑 𝑡 − 𝜑𝑑 ).
𝑑𝑡 𝑑𝑡

Suppose we have another function 𝑥𝑡𝑟 which represents the solution of the same system but without driving:
then

𝑑2 𝑥tr 𝑑𝑥tr
2
+ 2𝛾 + 𝜔2 𝑥tr = 0.
𝑑𝑡 𝑑𝑡

We have solved undriven systems in the previous chapter, and 𝑥𝑡𝑟 represents such a motion. If we now add
these equations we obtain

𝑑2 𝑑
2
(𝑥ss + 𝑥tr ) + 2𝛾 (𝑥ss + 𝑥tr ) + 𝜔2 (𝑥ss + 𝑥tr ) = 𝜔2 𝑟 cos(𝜔𝑑 𝑡 − 𝜑𝑑 ),
𝑑𝑡 𝑑𝑡

which means that the function 𝑥ss + 𝑥tr is a solution to equation (56) i.e. the driven system. So the complete
solution for a driven harmonic oscillator is the steady state solution we have used before plus a solution 𝑥𝑡𝑟
of the same system without driving. This transient solution, which could correspond to undercritical, critical
or overcritical damping, contains the two necessary integration constants and always dies out after some
time (therefore it is called “transient”). After a long time only the steady state solution remains, so it is
generally of greater importance.
 103

𝑥tr [𝑡]

𝑥ss [𝑡]
Figure 53
𝑡
Amplitude as a function
of driving angular
frequency for different
𝑥[𝑡] = 𝑥tr [𝑡] + 𝑥ss [𝑡] damping factors.

beat beat

To see the effect of the transient part of the solution can have, assume that the system is undercritically
damped. Then from the previous chapter 𝑥tr = 𝐵𝑒 −𝛾𝑡 cos(𝜔0 𝑡 − 𝜓) , where we have used 𝐵 for the
amplitude and 𝜓 for the phase constant to avoid confusing symbols. This is the motion the system would
have if it was not driven and is shown at the top of Figure 53. When this system is driven we obtain the
steady-state solution which has the form 𝑥ss = 𝐴 cos(𝜔𝑑 𝑡 − 𝜑𝑑 − 𝜑). This is shown in the middle of Figure
53 and has an amplitude that remains constant. The complete solution of the driven system is then not just
the steady state solution, but the sum of it and the transient solution. This is shown at the bottom of Figure
53 and reveals a complicated motion near the beginning which often contains beats because the transient
and steady state solutions do not have the same frequency. Later the transient solution decays to zero,
leaving only the steady state solution.

8.7. Velocity resonance

The steady state motion of a sinusoidally driven oscillator was shown to be 𝑥ss = 𝐴 cos(𝜔𝑑 𝑡 − 𝜑𝑑 − 𝜑)
where 𝐴 and 𝜑 are constants which depend on the driving frequency. The steady state velocity is

𝑑𝑥ss
𝑣= = −𝐴𝜔𝑑 sin(𝜔𝑑 𝑡 − 𝜑𝑑 − 𝜑)
𝑑𝑡

𝜔2 𝑟𝜔𝑑
which has amplitude 𝐴𝜔𝑑 = where we have made use of equation (57). The question
2 2 +4𝛾2 𝜔2
√(𝜔2 −𝜔𝑑 ) 𝑑

arises: at which driving frequency is this velocity amplitude a maximum? We can rewrite the velocity
 104

𝜔2 𝑟
amplitude as 2
and then it is clear that the part in the square root is a minimum (and hence the
𝜔2 −𝜔2𝑑 +4𝛾 2
√( 2 )
𝜔𝑑

velocity amplitude is a maximum) when 𝜔𝑑 = 𝜔. For this driving angular frequency the velocity amplitude
𝜔2 𝑟 𝜋
becomes 2𝛾
and the phase lag angle from equation (58) is 𝜑 = 2 . Actually, this phase angle needs careful
interpretation since it represents the lag between the position and the driving force. At velocity resonance

𝜔2 𝑟 𝜋 𝜔2 𝑟
𝑣=− sin (𝜔𝑑 𝑡 − 𝜑𝑑 − ) = cos(𝜔𝑑 𝑡 − 𝜑𝑑 )
2𝛾 2 2𝛾

and the velocity is actually perfectly in phase with the driving force.

8.8. Power supplied by the driving system

A driven system loses energy due to damping (friction), and so the driving component must supply energy to
keep the system in steady state motion. The power required is given by the product of the driving force and
the velocity, and is given by

𝑃 = 𝐹𝑣 = {𝑘𝑟 cos(𝜔𝑑 𝑡 − 𝜑𝑑 )}{−𝐴𝜔𝑑 sin(𝜔𝑑 𝑡 − 𝜑𝑑 − 𝜑)}.

Expanding the sine term, one gets

𝑃 = −𝐴𝑘𝑟𝜔𝑑 {cos(𝜔𝑑 𝑡 − 𝜑𝑑 )}{sin(𝜔𝑑 𝑡 − 𝜑𝑑 ) cos 𝜑 − cos(𝜔𝑑 𝑡 − 𝜑𝑑 ) sin 𝜑}.

When multiplied out the first term averages to zero, while in the second term there occurs cos2(𝜔𝑑 𝑡 − 𝜑𝑑 )
which has an average value of a half. Therefore the average power required is

1
𝑃𝑎𝑣𝑒 = 𝐴𝑘𝑟𝜔𝑑 sin 𝜑.
2

𝜔2 𝑟(2𝛾𝜔𝑑 )
But from section 8.2 we know that 𝐴 sin 𝜑 = 2
2 +(2𝛾𝜔 )2
, so
(𝜔2 −𝜔𝑑 ) 𝑑

𝑘𝑟 2 𝛾𝜔2 𝜔𝑑2 𝑘𝑟 2 𝛾𝜔2

𝑃𝑎𝑣𝑒 = = .
(𝜔 2 − 𝜔𝑑2 )2 + (2𝛾𝜔𝑑 )2 𝜔 2 − 𝜔𝑑2
2
( 𝜔 ) + 4𝛾 2 (62)
𝑑

The maximum average power taken in by the system from the driving component (and then lost via friction)
occurs when the denominator is a minimum i.e. when the driving angular frequency is 𝜔𝑑 = 𝜔, which
corresponds to velocity resonance. It is left for the exercises to show that full width at half maximum of the
𝑃𝑎𝑣𝑒 is 2𝛾. Therefore the smaller the damping, the narrower the power absorption peak.

The power absorption curve has interesting applications in physics. These deal with light or electromagnetic
waves that are incident on a material: the electric field of the electromagnetic waves act as a sinusoidally
 105

driving force on any charged particles in the material, which oscillate. If the incident light contains many
frequencies, then according to the power absorption curve only the frequencies near the natural frequency
of the charged particles will be strongly absorbed, while other frequencies will pass through. Applications of
this include:

 Light from the sun generally represents a black-body spectrum. However, in 1802 Wollaston observed
that the intensity is much weaker at certain frequencies. This is due to the absorption of the light waves
of these frequencies which correspond to the natural vibration frequencies of the electrons inside atoms
of various elements. By measuring similar data from the stars, one can find out what elements they
contain. In principle, a similar effect gives rise to colours of glass and to the heating of food in a microwave
oven, as well as playing an important role in the efficiency of solar cells.

 When a material is placed in a magnetic field, the nuclei of the atoms in the material act as a quantum
oscillators, with a certain natural frequency (which depends on the strength of the magnetic field). Radio
or microwaves of corresponding frequency are absorbed. This is called nuclear magnetic resonance
(NMR), for which Bloch and Purcell shared the 1952 Nobel Prize. The natural frequency of hydrogen
nuclei (protons) in the human body change slightly depending on their position in molecules e.g. it is
different in water and in fat. This forms the basis for medical NMR imaging.

8.9. Problems

1. A damped harmonic oscillator consists of a mass attached to a spring in a damping medium. When it is set
5
in motion, it oscillates with a frequency (not angular frequency) of Hz. (a) Explain (without mathematics)
𝜋
why we can be sure that this damping is undercritical. (b) What is the value of the damped angular frequency
𝜔0 ? (c) When the system is driven by a sinusoidal force the steady-state amplitude is a maximum when 𝜔𝑑 =
8 rad/s. Calculate the damping constant 𝛾. (d) If the damping force could be eliminated and the system was
not driven, what would be its angular frequency when set in motion? (e) What would be the phase lag
between the driving force and the motion at resonance i.e. the value of 𝜑𝑅 ? (f) Calculate the spring constant
if the mass is 125 g. (g) If the radius of the driving wheel was 5 cm, calculate the resonance amplitude. (h) For
what two values of driving angular frequency will the amplitude be 90% of this maximum value?

2. The general solution for a driven harmonic oscillator with a sinusoidal driving force consists of two parts,
namely the transient and steady-state solutions. (a) Give a reason why the steady-state solution can be
regarded as the more important of the two. (b) Can resonance ever occur in an overdamped or critically
damped system? Motivate your answer. (c) Could resonance occur in all undercritically damped systems?
Motivate your answer by giving the maximum possible damping constant for undercritical damping and also
for resonance to occur. (d) If resonance occurs in a system, give expressions for the angular frequency of the
steady-state solution and the transient solution in terms of the system's spring constant, mass and damping
constant.

3. Just as one gets position resonance and velocity resonance, one also gets acceleration resonance. Show
that the acceleration amplitude is a maximum when the driving angular frequency is given by
 106

𝜔2
𝜔𝑑 = .
√𝜔 2 − 2𝛾 2

Also show that at very high driving frequencies the acceleration amplitude does not tend to zero, but rather
to 𝜔2 𝑟 which is the driving acceleration amplitude.

𝑚𝑟 2 𝜔4
4. Show that the maximum of the average power curve is given by 𝑃𝑎𝑣𝑒,𝑚𝑎𝑥 = 4𝛾
and its FWHM (full
width at half maximum) is 2𝛾.

5. When setting up the sinusoidally driven harmonic oscillator problem, we took the resistive force to be
proportional to the velocity. This velocity is relative to the ground. But in some applications, e.g. a table
standing on springs with the ground vibrating, this velocity must be taken relative to the driving system i.e.
it must be the relative velocity. In such cases the force on the mass is

𝐹 = −𝑘(𝑥 − 𝑥′) − 𝛽(𝑣 − 𝑣′)

where the unprimed values refer to the position and velocity of the object, while the primed values refer to
the position and velocity of the driving system. (a) Show that this leads to the equation of motion 𝑎 + 2𝛾𝑣 +
𝜔2 𝑥 = 𝜔2 𝑥 ′ + 2𝛾𝑣′. (b) If 𝑥 ′ [𝑡] = 𝑟 cos(𝜔𝑑 𝑡), show that the driving term on the right is 𝜔2 𝑟 cos(𝜔𝑑 𝑡) −
2𝛾𝑟𝜔𝑑 sin(𝜔𝑑 𝑡). (c) The first term in the previous part is what we used in this chapter, while the second
term is new. Together they cause a driving oscillation with amplitude √(𝜔 2 𝑟)2 + (2𝛾𝑟𝜔𝑑 )2 , so that in such
a case the amplitude of the steady state motion is

√(𝜔 2 𝑟)2 + (2𝛾𝑟𝜔𝑑 )2 𝜔 4 + 4𝛾 2 𝜔𝑑2

𝐴[𝜔𝑑 ] = = 𝑟√ 2 .
(𝜔 − 𝜔𝑑2 )2 + 4𝛾 2 𝜔𝑑2
√(𝜔 2 − 𝜔𝑑2 )2 + 4𝛾 2 𝜔𝑑2

By taking the derivative of the part inside the square root and setting it equal to zero, show that the maximum
occurs when 2𝛾 2 𝜔𝑑4 + 𝜔4 𝜔𝑑2 − 𝜔6 = 0, or

𝜔2 8𝛾 2
𝜔𝑑 = √√1 + − 1.
2𝛾 𝜔2
 107

9. NON-Sinusoidally driven harmonic motion

using Fourier analysis

9.1. Introduction

In the previous chapter we solved the problem of a sinusoidally driven harmonic oscillator. But what would
be the motion of a harmonic oscillator driven by a non-sinusoidal periodic function such as the square,
triangular or saw-tooth waves (Figure 54)?

Function description and Fourier series Graph

Square wave:

−1 if – 𝜋 < 𝑡 < 0
𝑓[𝑡] = {
1 if 0 < 𝑡 < 𝜋

4 sin(𝑡) sin(3𝑡) sin(5𝑡)

𝑓[𝑡] = ( + + + ⋯)
𝜋 1 3 5
Sawtooth wave:

𝑓[𝑡] = 𝑡 for – 𝜋 < 𝑡 < 𝜋

sin(𝑡) sin(2𝑡) sin(3𝑡)

𝑓[𝑡] = 2 ( − + + ⋯)
1 2 3
Triangular wave:

𝜋
+ 𝑡 if – 𝜋 < 𝑡 < 0
𝑓[𝑡] = {2𝜋
− 𝑡 if 0 < 𝑡 < 𝜋
2

4 cos(𝑡) cos(3𝑡) cos(5𝑡)

𝑓[𝑡] = ( 2 + + +⋯)
𝜋 1 32 52
Half-rectified wave:

0 if – 𝜋 < 𝑡 < 0
𝑓[𝑡] = {
sin 𝑡 if 0 < 𝑡 < 𝜋
1 1
𝑓[𝑡] = + sin(𝑡) −
𝜋 2

2 cos(2𝑡) cos(4𝑡) cos(6𝑡)

( + + + ⋯)
𝜋 1∙3 3∙5 5∙7

Figure 54 Fourier series of the periodic continuations of some functions defined over −𝝅 to 𝝅.
 108

The prediction of such motion is made possible through the mathematical theory of Fourier analysis, which
allows one to decompose any periodic function into sine and cosine functions. Figure 54 shows several
examples of this, and the method to obtain these Fourier series will be discussed at the end of the chapter.
To see how they can be used to solve the problem of a non-sinusoidally driven harmonic oscillator, we must
first consider a harmonic oscillator driven by two sinusoidal functions.

9.2. Double sinusoidally driven harmonic oscillator

𝜔𝑑2 𝑟2

𝑃
𝑟1 𝐴
𝐹(𝑡) 𝑅
𝐹𝑠
𝜔𝑑1

Figure 55 Doubly driven harmonic oscillator.

Consider Figure 55 which shows a more complicated driving system than that in Figure 50. In this case the
driving rod is attached near the perimeter of a smaller rotating wheel (having angular frequency 𝜔𝑑2 ) which
is attached near the perimeter of the original rotating wheel (having angular frequency 𝜔𝑑1 ). This will result
in second driving force superimposed upon the first, and the system can be described by the differential
equation

𝑎 + 2𝛾𝑣 + 𝜔2 𝑥 = 𝜔2 𝑟1 cos(𝜔𝑑1 𝑡 − 𝜑𝑑1 ) + 𝜔2 𝑟2 cos(𝜔𝑑2 𝑡 − 𝜑𝑑2 ). (63)

The radii, angular frequencies and phase constants of the two driving terms are completely independent of
one another. Although this may seem complicated, we will show that the steady state solution of this
system is just the sum of the steady state solutions if each driving force is considered alone.

Suppose we only had the first driving force, so that 𝑎 + 2𝛾𝑣 + 𝜔2 𝑥 = 𝜔2 𝑟1 cos(𝜔𝑑1 𝑡 − 𝜑𝑑1 ) and let 𝑥1 (𝑡)
be its steady state solution, so that

𝑎1 + 2𝛾𝑣1 + 𝜔2 𝑥1 = 𝜔2 𝑟1 cos(𝜔𝑑1 𝑡 − 𝜑𝑑1 ).

But suppose we only had the second driving force, so that 𝑎 + 2𝛾𝑣 + 𝜔2 𝑥 = 𝜔2 𝑟2 cos(𝜔𝑑2 𝑡 − 𝜑𝑑2 ) and let
𝑥2 (𝑡) be its steady state solution, so that

𝑎2 + 2𝛾𝑣2 + 𝜔2 𝑥2 = 𝜔2 𝑟2 cos(𝜔𝑑2 𝑡 − 𝜑𝑑2 ).

Adding these together we get

(𝑎1 + 𝑎2 ) + 2𝛾(𝑣1 + 𝑣2 ) + 𝜔2 (𝑥1 + 𝑥2 ) = 𝜔2 𝑟1 cos(𝜔𝑑1 𝑡 − 𝜑𝑑1 ) + 𝜔2 𝑟2 cos(𝜔𝑑2 𝑡 − 𝜑𝑑2 )

and so clearly 𝑥(𝑡) = 𝑥1 (𝑡) + 𝑥2 (𝑡) is the steady state solution to the doubly driven harmonic oscillator. So
the steady state motion is just the sum of the steady state motions that would result if each driving force was
allowed to act alone. This result is easily generalized into three or more sinusoidal driving forces.
 109

9.3. Non-sinusoidally driven harmonic motion

Using Fourier analysis, a non-sinusoidal periodic driving force can be expressed as a sum of sinusoidal
functions. The steady-state motion of a harmonic oscillator due to each of these sinusoidal driving forces
acting alone can be calculated (as in the previous chapter) and, from the result just shown in the previous
section, these can be added together to obtain the overall steady-state solution. Of course, a transient
solution could be added as well if desired, but we shall not do so.

9.3.1. Example: triangle wave

Consider a harmonic oscillator driven by a triangle-wave periodic force. The differential equation for its
𝐹(𝑡)
motion is given by equation (55), namely 𝑎 + 2𝛾𝑣 + 𝜔2 𝑥 = 𝑚
where in this case 𝐹[𝑡] is the driving
triangle-wave force. Now from Fourier analysis (see Figure 54), a triangular wave can be written as

𝜋
+ 𝑡 if – 𝜋 < 𝑡 < 0 4 cos(𝑡) cos(3𝑡) cos(5𝑡)
𝑓[𝑡] = {2𝜋 = ( 2 + + + ⋯ ).
− 𝑡 if 0 < 𝑡 < 𝜋 𝜋 1 32 52
2

We adjust this as follows:

𝜋 2𝐹0
 This wave has amplitude : to get a driving force of amplitude 𝐹0 we multiply by .
2 𝜋
 This wave has period 2𝜋. To adjust it to a period 𝑇, we first compress it by 2𝜋 and then stretch it by 𝑇,
2𝜋
which is achieved by replacing 𝑡 with 𝑇
𝑡.

2𝜋 2𝜋 2𝜋
8𝐹0 cos( 𝑇 𝑡) cos(3 𝑡)
𝑇
cos(5 𝑡)
𝑇
This gives the force 𝐹(𝑡) = 𝜋2
( 12 + 32
+ 52
+ ⋯ ). Substituting it into equation (55) and
8𝐹0
defining Ω = 𝜋2 𝑚 gives

2𝜋 2𝜋 2𝜋
cos ( 𝑇 𝑡) cos (3 𝑇 𝑡) cos (5 𝑇 𝑡)
𝑎 + 2𝛾𝑣 + 𝜔2 𝑥 = Ω ( + + + ⋯ ).
12 32 52

We now take each driving term on the right hand side separately and calculate the steady-state motion as in
the previous chapter. Afterwards we add each of these solutions together to get the complete steady state
solution for our problem. Recall from the previous chapter that a sinusoidally driven harmonic oscillator 𝑎 +
2𝛾𝑣 + 𝜔2 𝑥 = 𝜔2 𝑟 cos(𝜔𝑑 𝑡 − 𝜑𝑑 ) has the steady-state solution
𝜔2 𝑟 𝑑𝛾𝜔
𝑥(𝑡) = 𝐴 cos(𝜔𝑑 𝑡 − 𝜑𝑑 − 𝜑) where 𝐴 = and 𝜑 = arctan 𝜔2 −𝜔 2 .From this we have that:
2 2 +4𝛾2 𝜔2
√(𝜔2 −𝜔𝑑 𝑑
) 𝑑

cos
The driven harmonic oscillator 𝑎 + 2𝛾𝑣 + 𝜔2 𝑥 = 𝑎0 { }(𝜔𝑑 𝑡) has steady-state solution
sin
cos 𝑎0 2𝛾𝜔
𝑥(𝑡) = 𝐴 { }(𝜔𝑑 𝑡 − 𝜑) where 𝐴 = and 𝜑 = arctan 𝜔2 −𝜔𝑑2 .
sin 2 2
√(𝜔 −𝜔 ) +4𝛾 𝜔
2 2 2 𝑑
𝑑 𝑑
 110

Suppose we have a system with 𝑘 = 100 Nm-1, 𝑚 = 1 kg (so 𝜔 = 10 Hz), 𝛾 = 1 Hz, and it is driven by a
triangular force oscillation with amplitude 𝐹0 = 5 N (so Ω = 4.053 ms-2) and period 𝑇 = 2 s:

 Taking just the first driving term gives 𝑎 + 2𝛾𝑣 + 𝜔2 𝑥 = Ω cos(𝜋𝑡), meaning 𝑎0 = Ω and 𝜔𝑑 = 𝜋. This
has solution 𝑥 = 0.011068 Ω cos(𝜋𝑡 − 0.070) = 0.045 cos(𝜋𝑡 − 0.070).
Ω Ω
 Taking just the second driving term gives 𝑎 + 2𝛾𝑣 + 𝜔2 𝑥 = 9 cos(3𝜋𝑡), meaning 𝑎0 = 9
and 𝜔𝑑 = 3𝜋.
This has solution 𝑥 = 0.005071 Ω cos(3𝜋𝑡 − 1.036) = 0.021 cos(3𝜋𝑡 − 1.036).
Ω Ω
 Taking just the third driving term gives 𝑎 + 2𝛾𝑣 + 𝜔2 𝑥 = 25 cos(5𝜋𝑡), meaning 𝑎0 = 25 and 𝜔𝑑 = 5𝜋.
This has solution 𝑥 = 0.000267 Ω cos(5𝜋𝑡 + 0.211) = 0.001 cos(5𝜋𝑡 + 0.211).

More terms can be computed if necessary. A reasonable cut-off occurs when the 𝜔𝑑 value to be used
becomes much greater than the natural frequency of the system (𝜔). In this example the next driving term
has 𝜔𝑑 = 7𝜋 ≈ 22 which is much greater than 𝜔 = 10, so no more terms are actually necessary. Due to the
large number of calculations involved, it is much easier to do the computations in a spreadsheet. The overall
solution is the sum of these individual solutions, namely

𝑥 = 0.045 cos(𝜋𝑡 − 0.070) + 0.021 cos(𝜋𝑡 − 1.036) + 0.001 cos(5𝜋𝑡 + 0.211) + ⋯

9.3.2. Example: half-rectified wave

Consider a harmonic oscillator driven by a force in the form of a half-rectified wave. The differential equation
𝐹(𝑡)
for its motion is given by equation (55), namely 𝑎 + 2𝛾𝑣 + 𝜔2 𝑥 = 𝑚
where in this case 𝐹[𝑡] is the driving
half-rectified wave force. Now from Fourier analysis (see Figure 54), a half-rectified wave can be written as

0 if – 𝜋 < 𝑡 < 0 2 1 𝜋 cos(2𝑡) cos(4𝑡) cos(6𝑡)

𝑓[𝑡] = { = [ + sin(𝑡) − ( + + + ⋯ )].
sin 𝑡 if 0 < 𝑡 < 𝜋 𝜋 2 4 1∙3 3∙5 5∙7

We adjust this as follows:

 This wave has unit amplitude: we multiply by 𝐹0 to get a driving force of amplitude 𝐹0 .
 This wave has period 2𝜋. To adjust it to a period 𝑇, we first compress it by 2𝜋 and then stretch it by 𝑇,
2𝜋
which is achieved by replacing 𝑡 with 𝑇
𝑡.

2𝜋 2𝜋 2𝜋
2𝐹0 1 𝜋 2𝜋 cos(2 𝑡) cos(4 𝑡) cos(6 𝑡)
𝑇 𝑇 𝑇
This gives the force 𝐹[𝑡] = 𝜋
(2 + 4 sin ( 𝑇 𝑡) − 1∙3
− 3∙5
− 5∙7
− ⋯ ). Substituting it into
2𝐹
equation (55) and defining Ω = 𝜋𝑚0 gives

4𝜋 8𝜋 12𝜋
1 𝜋 2𝜋 cos ( 𝑇 𝑡) cos ( 𝑇 𝑡) cos ( 𝑇 𝑡)
𝑎 + 2𝛾𝑣 + 𝜔2 𝑥 = Ω ( + sin ( 𝑡) − − − − ⋯ ).
2 4 𝑇 1∙3 3∙5 5∙7

We proceed to deal with each term on the right hand side separately. Suppose we use the same system as
in the previous example and it is driven by a half-rectified wave amplitude 𝐹0 = 5 N (so Ω = 3.183 ms-2) and
period 𝑇 = 2 s:

Ω Ω
 Taking just the first driving term gives 𝑎 + 2𝛾𝑣 + 𝜔2 𝑥 = 2 cos(0𝑡), meaning 𝑎0 = 2
and 𝜔𝑑 = 0. This
has solution 𝑥 = 0.005Ω cos(0𝑡 − 0) = 0.016.
 111

πΩ πΩ
 Taking just the second driving term gives 𝑎 + 2𝛾𝑣 + 𝜔2 𝑥 = 4
sin(𝜋𝑡), meaning 𝑎0 = 4
and 𝜔𝑑 = π.
This has solution 𝑥 = 0.008693 Ω sin(𝜋𝑡 − 0.070) = 0.028 sin(𝜋𝑡 − 0.070).
Ω Ω
 Taking just the third driving term gives 𝑎 + 2𝛾𝑣 + 𝜔2 𝑥 = − 3 cos(2𝜋𝑡), meaning 𝑎0 = − 3 and 𝜔𝑑 =
2𝜋. This has solution 𝑥 = −0.005393 Ω cos(2𝜋𝑡 − 0.205) = −0.017 cos(2𝜋𝑡 − 0.205).
Ω Ω
 Taking just the fourth driving term gives 𝑎 + 2𝛾𝑣 + 𝜔2 𝑥 = − 15 cos(4𝜋𝑡), meaning 𝑎0 = − 15 and 𝜔𝑑 =
4𝜋. This has solution 𝑥 = −0.001056 Ω cos(4𝜋𝑡 + 0.409) = −0.003 cos(4𝜋𝑡 + 0.409).
Ω Ω
 Taking just the fifth driving term gives 𝑎 + 2𝛾𝑣 + 𝜔2 𝑥 = − 35 cos(6𝜋𝑡), meaning 𝑎0 = − 35 and 𝜔𝑑 =
6𝜋. This has solution 𝑥 = −0.000111 Ω cos(6𝜋𝑡 + 0.147) which is negligible.

The overall solution is the sum of these individual solutions, namely

𝑥 = 0.016 + 0.028 sin(𝜋𝑡 − 0.070) − 0.017 cos(2𝜋𝑡 − 0.205) − 0.003 cos(4𝜋𝑡 + 0.409).

9.4. Fourier analysis

We now consider how to obtain Fourier series. To describe a periodic function 𝑓[𝑡] we need only describe it
over one period. It then repeats this pattern to the left and right – this is called the periodic continuation of
the function. We will only consider functions with a period of 2𝜋, described between −𝜋 and 𝜋, such as in
Figure 54. As we saw in the previous section, it is elementary to stretch or compress it to give any period we
desire. Now suppose we wish to write a function in terms of sine and cosine functions as follows:

𝑓[𝑡] = 𝑐 + 𝑎1 cos(𝑡) + 𝑎2 cos(2𝑡) + ⋯ + 𝑏1 sin(𝑡) + 𝑏2 sin(2𝑡) + ⋯ (64)

Note that a constant 𝑐 has been included at the beginning. Since the average value of all sine and cosine
functions is zero, if a constant were not included it would be impossible to approximate a function with a
non-zero average value.

The question is: what would be the values of 𝑐 and the coefficients 𝑎1 , 𝑏1 , 𝑎2 , 𝑏2 etc. The answer was
discovered by Fourier, and we consider three cases:

To find 𝒄, we integrate both sides of equation (64) between −𝜋 and 𝜋. This gives

𝜋 𝜋 𝜋 𝜋 𝜋 𝜋

∫ 𝑓[𝑡] 𝑑𝑡 = 𝑐 ∫ 𝑑𝑡 + 𝑎1 ∫ cos(𝑡) 𝑑𝑡 + 𝑎2 ∫ cos(2𝑡) 𝑑𝑡 + ⋯ + 𝑏1 ∫ sin(𝑡) 𝑑𝑡 + 𝑏2 ∫ sin(2𝑡) 𝑑𝑡 + ⋯

−𝜋 −𝜋 −𝜋 −𝜋 −𝜋 −𝜋

𝜋
Now ∫−𝜋 𝑑𝑡 = 2𝜋, while all the other integrals on the right hand side are easily shown to be zero. Therefore

𝜋
1
𝑐= ∫ 𝑓(𝑡) 𝑑𝑡.
2𝜋
−𝜋

This formula shows that 𝑐 represents the area under the function from −𝜋 to 𝜋, divided by the length of the
interval i.e. 2𝜋, giving the average height of the function.
 112

To find 𝒂𝒌 , where 𝑘 = 1,2,3 ⋯ we multiply both sides of equation (64) with cos(𝑘𝑡) and then integrate
between −𝜋 and 𝜋. This gives

𝜋 𝜋 𝜋 𝜋

∫ 𝑓[𝑡] cos(𝑘𝑡) 𝑑𝑡 = 𝑐 ∫ cos(𝑘𝑡) 𝑑𝑡 + 𝑎1 ∫ cos(𝑡) cos(𝑘𝑡) 𝑑𝑡 + 𝑎2 ∫ cos(2𝑡) cos(𝑘𝑡) 𝑑𝑡 + ⋯

−𝜋 −𝜋 −𝜋 −𝜋

𝜋 𝜋

+𝑏1 ∫ sin(𝑡) cos(𝑘𝑡) 𝑑𝑡 + 𝑏2 ∫ sin(2𝑡) cos(𝑘𝑡) 𝑑𝑡 + ⋯

−𝜋 −𝜋

The first integral on the right hand side is zero, as well as all those of the bottom row as they are odd
𝜋
functions. Consider the rest of the integrals, which have the form ∫−𝜋 cos(𝑚𝑡) cos(𝑘𝑡) 𝑑𝑡 .

1
These are easily evaluated if we use the fact that cos(𝑚𝑡) cos(𝑘𝑡) = {cos(𝑚𝑡 − 𝑘𝑡) + cos(𝑚𝑡 + 𝑘𝑡)}.
2
𝜋 𝜋 1
Now if 𝑚 = 𝑘 then the integral becomes ∫−𝜋 cos2 (𝑘𝑡) 𝑑𝑡 = ∫−𝜋 2 {1 + cos(2𝑘𝑡)}𝑑𝑡 = 𝜋, while if 𝑚 ≠ 𝑘 the
𝜋 1 𝜋
integrals become ∫−𝜋 2 {cos(𝑚𝑡 − 𝑘𝑡) + cos(𝑚𝑡 + 𝑘𝑡)}𝑑𝑡 = 0. Therefore ∫−𝜋 𝑓[𝑡] cos(𝑘𝑡) 𝑑𝑡 = 𝑎𝑘 (𝜋) and
so

𝜋
1
𝑎𝑘 = ∫ 𝑓[𝑡] cos(𝑘𝑡) 𝑑𝑡.
𝜋
−𝜋

To find 𝒃𝒌 , where 𝑘 = 1,2,3 ⋯ we multiply both sides of equation (64) with sin(𝑘𝑡) and then integrate
between −𝜋 and 𝜋. This gives

𝜋 𝜋 𝜋 𝜋

∫ 𝑓[𝑡] sin(𝑘𝑡) 𝑑𝑡 = 𝑐 ∫ sin(𝑘𝑡) 𝑑𝑡 + 𝑎1 ∫ cos(𝑡) sin(𝑘𝑡) 𝑑𝑡 + 𝑎2 ∫ cos(2𝑡) sin(𝑘𝑡) 𝑑𝑡 + ⋯

−𝜋 −𝜋 −𝜋 −𝜋

𝜋 𝜋

+𝑏1 ∫ sin(𝑡) sin(𝑘𝑡) 𝑑𝑡 + 𝑏2 ∫ sin(2𝑡) sin(𝑘𝑡) 𝑑𝑡 + ⋯

−𝜋 −𝜋

The first integral on the right hand side is zero, as well as the rest on the top right row as they are odd
𝜋
functions. Consider the rest of the integrals, which have the form ∫−𝜋 sin(𝑚𝑡) sin(𝑘𝑡) 𝑑𝑡.

1
These are easily evaluated if we use the fact that sin(𝑚𝑡) sin(𝑘𝑡) = 2 {cos(𝑚𝑡 − 𝑘𝑡) − cos(𝑚𝑡 + 𝑘𝑡)}. Now
𝜋 𝜋 1
if 𝑚 = 𝑘 then the integral becomes ∫−𝜋 sin2(𝑘𝑡) 𝑑𝑡 = ∫−𝜋 2 {1 − cos(2𝑘𝑡)}𝑑𝑡 = 𝜋, while if 𝑚 ≠ 𝑘 the
𝜋 1 𝜋
integrals become ∫−𝜋 {cos(𝑚𝑡 − 𝑘𝑡) − cos(𝑚𝑡 + 𝑘𝑡)}𝑑𝑡 = 0. Therefore ∫−𝜋 𝑓[𝑡] sin(𝑘𝑡) 𝑑𝑡 = 𝑏𝑘 (𝜋) and
2
so

𝜋
1
𝑏𝑘 = ∫ 𝑓[𝑡] sin(𝑘𝑡) 𝑑𝑡.
𝜋
−𝜋

We can summarize by stating:

 113

For a given function 𝑓[𝑡] defined on the interval −𝜋 to 𝜋, we may express 𝑓[𝑡] as

∞ ∞

𝑓[𝑡] = 𝑐 + ∑ 𝑎𝑘 cos(𝑘𝑡) + ∑ 𝑏𝑘 sin(𝑘𝑡)

𝑘=1 𝑘=1

where the coefficients are given by

𝜋 𝜋 𝜋
1 1 1
𝑐= ∫ 𝑓[𝑡] 𝑑𝑡 ; 𝑎𝑘 = ∫ 𝑓[𝑡] cos(𝑘𝑡) 𝑑𝑡 ; 𝑏𝑘 = ∫ 𝑓[𝑡] sin(𝑘𝑡) 𝑑𝑡 .
2𝜋 𝜋 𝜋
−𝜋 −𝜋 −𝜋

Considerable work in evaluating these coefficients can be saved if the following is noted:

 The value of 𝑐 represents the average value of the function. Often it is clear this average value is zero
and then no calculation is required. Odd functions defined from −𝜋 to 𝜋 always have an average value
of zero, but even functions may also have an average value of zero. Even if 𝑓(𝑡) does not have an average
value of zero, its average value can often be easily calculated using the area under the curve instead of
having to integrate.
 If 𝑓(𝑡) is an odd function, the integral for 𝑎𝑘 is of an odd function and therefore 𝑎𝑘 = 0. This means all
the cosine terms will disappear from the Fourier series. This makes sense, since an odd function is best
represented by sine functions, which are odd. The integral for 𝑏𝑘 will be of an even function and the
𝜋 𝜋
calculation is often made easier by using the fact that ∫−𝜋 even function 𝑑𝑡 = 2 ∫0 even function 𝑑𝑡.
 If 𝑓(𝑡) is an even function, the integral for 𝑏𝑘 is of an odd function and therefore 𝑏𝑘 = 0. This means all
the sine terms will disappear from the Fourier series. This makes sense, since an even function is best
represented by cosine functions, which are even. The integral for 𝑎𝑘 will be of an even function and the
calculation is often made easier by using the formula already given in the previous point.

9.5. Examples of Fourier analysis

9.5.1. Square wave

−1 if – 𝜋 < 𝑡 < 0
The square wave 𝑓[𝑡] = { is shown at the top of Figure 56. Its average value is zero and so
1 if 0 < 𝑡 < 𝜋
𝑐 = 0. In addition, it is an odd function, therefore 𝑎𝑘 = 0. It only remains to calculate

𝜋
1
𝑏𝑘 = ∫ 𝑓[𝑡] sin(𝑘𝑡) 𝑑𝑡
𝜋
−𝜋

which is the integral of an even function. Therefore

𝜋 𝜋 𝜋
2 2 2 cos(𝑘𝑡) 2
𝑏𝑘 = ∫ 𝑓[𝑡] sin(𝑘𝑡) 𝑑𝑡 = ∫{1} sin(𝑘𝑡) 𝑑𝑡 = − | = {1 − cos(𝑘𝜋)}.
𝜋 𝜋 𝜋 𝑘 0
𝜋𝑘
0 0

From a graph of the cosine function it is easy to see that cos(𝑘𝜋) = (−1)𝑘 , so that
 114

2 0 if 𝑘 is even;
𝑘
{1 − (−1) } = { 4
𝑏𝑘 =
𝜋𝑘 if 𝑘 is odd.
𝜋𝑘
4 4 4
Therefore 𝑏1 = 𝜋 , 𝑏2 = 0, 𝑏3 = 3𝜋 , 𝑏4 = 0, 𝑏5 = 5𝜋 , etc. for our square wave. We can list the Fourier
coefficients to form a table, as for the square wave below:

Harmonic (𝒌) Amplitude of cosine term (𝒂𝒌 ) Amplitude of sine term (𝒃𝒌 )
1 0 4/
2 0 0
3 0 4/(3)
4 0 0
5 0 4/(5)
6 0 0
7 0 4/(7)
etc...

Note that the value of 𝑘 is often referred to as the harmonic. We can substitute these coefficients into the
original Fourier expansion (equation (64)) to get

4 sin(𝑡) sin(3𝑡) sin(5𝑡) sin(7𝑡)

𝑓(𝑡) = ( + + + + ⋯)
𝜋 1 3 5 7

All the other terms have coefficients of zero.

𝑓(𝑡)

Figure 56
𝑓1 (𝑡) Square wave
approximations.
𝑓3 (𝑡)
The top figure shows the
square wave, and the
𝑓5 (𝑡) graphs below show
Fourier series
approximations with
𝑓11 (𝑡)
increasing number of
harmonics.
𝑓23 (𝑡)

sin(𝑡)
Each of the terms in the Fourier series is also called a harmonic.16 For the square wave, the first term 1
is
sin(3𝑡) sin(5𝑡)
called the first (or fundamental) harmonic. The next term 3
is the third harmonic and 5
is the fifth
harmonic, etc. The square wave therefore has only odd harmonics. Figure 56 shows the Fourier series of the

16
The Fourier series of a function with no parity has both sine and cosine terms. In such cases the two terms 𝑎𝑘 cos(𝑘𝑡) + 𝑏𝑘 sin(𝑘𝑡)
together form the 𝑘 th harmonic.
 115

square wave with increasing number of harmonics taken into account. Note how the Fourier series finds it
difficult to approximate the step discontinuities at 𝑘𝜋 for integer 𝑘. This “overshoot” in the Fourier series at
the places where the function has discontinuities is called the Gibbs phenomenon.

9.5.2. Triangle wave

𝑓(𝑡)

Figure 57

𝑓1 (𝑡) Triangular wave

approximations.

The top figure shows the

𝑓3 (𝑡) triangular wave, and the
graphs below show
Fourier series
approximations with
𝑓5 (𝑡) increasing number of
harmonics.

𝑓7 (𝑡)

The triangle wave 𝑓[𝑡] = |𝑡| on the interval −𝜋 to 𝜋 is shown at the top of Figure 57 and is a different form
of the triangle wave than the one in Figure 54. Do not confuse the two! It is clearly an even function and so
1 𝜋
𝑏𝑘 = 0. Its average value is not zero. We can calculate 𝑐 = 2𝜋 ∫−𝜋 𝑓[𝑡] 𝑑𝑡 which is the integral of an even
function. Therefore
𝜋 𝜋
2 1 𝜋
𝑐= ∫ 𝑓[𝑡] 𝑑𝑡 = ∫ 𝑡 𝑑𝑡 = .
2𝜋 𝜋 2
0 0

That was the long way of doing it! The same result is easily obtained without integrating as follows: each of
the triangles has base 2𝜋 and perpendicular height 𝜋, and hence an area 𝜋 2 . The average height is this area
𝜋2 𝜋
divided by the base length of 2𝜋, so 𝑐 = 2𝜋 = 2 as before.

1 𝜋
We must calculate 𝑎𝑘 = 𝜋 ∫−𝜋 𝑓[𝑡] cos(𝑘𝑡) 𝑑𝑡 , which is the integral of an even function. Therefore

𝜋 𝜋 𝜋 𝜋 𝜋
2 2 2 sin(𝑘𝑡) sin(𝑘𝑡) 2 cos(𝑘𝑡)
𝑎𝑘 = ∫ 𝑓[𝑡] cos(𝑘𝑡) 𝑑𝑡 = ∫ 𝑡 cos(𝑘𝑡) 𝑑𝑡 = {𝑡 | − ∫(1) ( ) 𝑑𝑡} = 2
| .
𝜋 𝜋 𝜋 𝑘 0
𝑘 𝜋𝑘 0
0 0 0

From a graph of the cosine function it is easy to see that cos(𝑘𝜋) = (−1)𝑘 , so that 𝑎𝑘 = 0 if 𝑘 is even, while
−4
𝑎𝑘 = 𝜋𝑘 2 if 𝑘 is odd.
 116

We can substitute these coefficients into the original Fourier expansion (equation (64)) to get

𝜋 4 cos(𝑡) cos(3𝑡) cos(5𝑡) cos(7𝑡)

𝑓(𝑡) = − ( + + + + ⋯)
2 𝜋 (1)2 (3)2 (5)2 (7)2

All the other terms have coefficients of zero. Figure 57 shows the Fourier series of 𝑓(𝑡) with increasing
number of harmonics taken into account.

9.6. Problems
𝜋
−1 if – 𝜋 < 𝑡 <– 2
𝜋 𝜋
1. Draw a graph of the periodic continuation of 𝑓[𝑡] = 1 if – 2 < 𝑡 < 2
and find the Fourier coefficients
𝜋
{ −1 if 2
<𝑡<𝜋
of the function. Also write down the Fourier series.

2. (a) Draw a graph of the periodic continuation of 𝑓[𝑡] = 𝑡 for – 𝜋 < 𝑡 < 𝜋. (b) Show that the Fourier
2
coefficients are given by 𝑐 = 0 ; 𝑎𝑘 = 0 ; 𝑏𝑘 = − 𝑘 (−1)𝑘 . (c) Write down the Fourier series for 𝑓[𝑡]. (d)
𝜋 1 1 1 1
Calculate 𝑓[𝑡] as well as its Fourier series at 𝑡 = 2 to show that 𝜋 = 4 (1 − 3 + 5 − 7 + 9 ⋯ ).
 117

10. CASE Study: oscillating electrical

circuits

10.1. The 𝑳𝑪-circuit

An 𝐿𝐶-circuit consists of a capacitor and an inductor in series and gets its name because the symbol for
inductance is 𝐿 while that for capacitance is 𝐶. Of course there must be some resistance, even if it is just that
of the connecting wires, but this is assumed to be negligible. There is also no battery in the circuit. Suppose
𝑑𝑄
a current 𝐼 in the circuit flows into the capacitor and the capacitor plate has a charge 𝑄 , so that = 𝐼. The
𝑑𝑡
𝑄
other capacitor plate has charge 𝑄 and the potential over the capacitor is − . The current then flows through
𝐶
𝑑𝐼
the inductor and the potential difference is −𝐿 . Kirchhoff's law states that the sum of potential differences
𝑑𝑡
around the circuit must be zero, so

𝑄 𝑑2 𝑄
+ 𝐿 2 = 0.
𝐶 𝑑𝑡

Now recall a mass attached to a spring: if there is no damping then Newton's second law gave

𝑑2 𝑥
𝑘𝑥 + 𝑚 = 0.
𝑑𝑡 2

A comparison of this equation to the one for the 𝐿𝐶-circuit shows they are basically the same (both being for
simple harmonic motion), but:

 The charge 𝑄 plays the role of the position 𝑥;

 The current 𝐼 plays the role of the velocity 𝑣;
1
 The inverse of capacitance 𝐶 plays the role of the spring constant 𝑘;
 The inductance 𝐿 plays the role of the mass 𝑚.

There is therefore no need to solve the equation for the 𝐿𝐶-circuit, since we can just use all our results for
simple harmonic motion. We had that
𝑘
𝑥 = 𝐴 cos(𝜔𝑡 − 𝜑) where 𝜔 = √𝑚.

Therefore for the circuit

1
𝑄 = 𝐴 cos(𝜔𝑡 − 𝜑) where 𝜔 = √𝐿𝐶.

Note that in the second case 𝐴 and 𝜑 are constants that depend on the initial charge 𝑄0 (similar to the initial
position 𝑥0 ) and initial current 𝐼0 (similar to the initial velocity 𝑣0 ).
 118

Example: Suppose an 𝐿𝐶-circuit has a capacitor with capacitance 𝐶 = 1 nF and an inductor with inductance
𝐿 = 100 mH. If there is initially a charge of 5 nC on the capacitor and a current of 1 mA in the circuit, find the
charge and current at any later time.

1 1
Solution: The charge will be given by 𝑄 = 𝐴 cos(𝜔𝑡 − 𝜑), with angular frequency 𝜔 = √𝐿𝐶 = √(0.1)(10−9 ).

𝑣 2
Therefore 𝜔 = 105 rad/s. Instead of 𝐴 = √𝑥02 + ( 𝜔0 ) we use

2
𝐼0 2 10−3
𝐴 = √𝑄02 + ( ) = √(5 × 10−9 )2 + ( 5 ) = 11.18 nC.
𝜔 10

𝑄0
Since 𝑄 = 𝐴 cos(𝜔𝑡 − 𝜑), initially 𝑄0 = 𝐴 cos 𝜑 and so cos 𝜑 = = 0.447.
𝐴

0 𝐼
Also 𝐼 = −𝐴𝜔 sin(𝜔𝑡 − 𝜑), therefore initially 𝐼0 = 𝐴𝜔 sin 𝜑 and so sin 𝜑 = 𝐴𝜔 = 0.894.

Since both sine and cosine are positive, the phase constant lies in the first quadrant and is 1.107 rad.
Therefore

 the charge on the capacitor varies according to 𝑄 = 11.18 cos(105 𝑡 − 1.107) [nC];
 the current in the circuit varies according to 𝐼 = −1.118 sin(105 𝑡 − 1.107) [mA].

It is not easy to measure the charge on the capacitor, but instead we can measure the voltage over the
capacitor. This is given by

𝑄 11.18 cos(105 𝑡 − 1.107) [nC]

𝑉𝑐 = = = 11.18 cos(105 𝑡 − 1.107) [𝑉].
𝐶 1 [nF]

10.2. The 𝑹𝑳𝑪-circuit

If we add a resistor to an 𝐿𝐶-circuit, we produce an 𝑅𝐿𝐶-circuit. The potential drop over the resistor is −𝑅𝐼
and so Kirchhoff's rule gives

𝑄 𝑑𝑄 𝑑2 𝑄
+𝑅 + 𝐿 2 = 0,
𝐶 𝑑𝑡 𝑑𝑡

which is just like the equation for a mass-spring system with resistance proportional to the velocity, for which
Newton's second law gives

𝑑𝑥 𝑑2 𝑥
𝑘𝑥 + 𝛽 + 𝑚 2 = 0.
𝑑𝑡 𝑑𝑡
𝛽 𝑅
The resistance 𝑅 plays the role of the parameter 𝛽 and instead of 2𝛾 = 𝑚 we must use 2𝛾 = 𝐿 . As in the case
for the mechanical mass-spring system, we can have undercritical, critical or overcritical damping.
𝑅 1
Oscillations will only occur when the damping is undercritical, for which 𝛾 < 𝜔 or 2𝐿 < √𝐿𝐶, giving
 119

4𝐿
𝐶< .
𝑅2

For such a system, the charge on the capacitor is given by 𝑄 = 𝐴𝑒 −𝛾𝑡 cos(𝜔𝑡 − 𝜑) where 𝐴 and 𝜑 depend
on the initial conditions and 𝜔0 = √𝜔 2 − 𝛾 2 as before.

Example: Find an expression for the quality factor 𝑄 (not to be confused with charge!) of an 𝑅𝐿𝐶-circuit in
terms of the resistance, inductance and capacitance.

𝜔0
Solution: From the chapter on damped systems we find that the quality factor can be written as 𝑄 = 2𝛾
.
Adapting this, one gets

1 𝑅 2
√𝜔 2 − 𝛾 2 √𝐿𝐶 − (2𝐿) 𝐿 1
𝑄= = =√ 2 − .
2𝛾 𝑅 𝑅 𝐶 4
𝐿
100
So, for example, if 𝐿 = 1 mH, 𝐶 = 1 nF and 𝑅 = 100 Ω, the quality factor is 𝑄 = 100, which means 2𝜋
≈ 16
1
oscillations will occur before the average energy decays to of its initial value.
𝑒

10.3. The sinusoidally driven 𝑹𝑳𝑪-circuit

If we add an alternating voltage source 𝑉 = 𝑉0 cos(𝜔𝑑 𝑡 − 𝜑𝑑 ) we get a sinusoidally driven 𝑅𝐿𝐶-circuit and
Kirchhoff's law gives

𝑄 𝑑𝑄 𝑑2 𝑄
+𝑅 + 𝐿 2 = 𝑉0 cos(𝜔𝑑 𝑡 − 𝜑𝑑 ),
𝐶 𝑑𝑡 𝑑𝑡

which is just like the equation for a driven mass-spring system for which Newton's second law gives

𝑑𝑥 𝑑2 𝑥
𝑘𝑥 + 𝛽 + 𝑚 2 = 𝑘𝑟 cos(𝜔𝑑 𝑡 − 𝜑𝑑 ).
𝑑𝑡 𝑑𝑡

The driving voltage amplitude 𝑉0 plays the role of 𝑘𝑟. This driven system has a steady-state and a transient
solution. The steady-state solution has the same frequency as the driving voltage and is 𝑄 =
𝐴 cos(𝜔𝑑 𝑡 − 𝜑𝑑 − 𝜑) where 𝐴 and 𝜑 are constants which do not depend on the initial conditions, but do
depend on the driving frequency. In particular, the amplitude for the charge on the capacitor is given by

𝑉0 /𝐿 𝜔2 𝐶𝑉0
𝐴= = .
√(𝜔 2 − 𝜔𝑑2 )2 + 4𝛾 2 𝜔𝑑2 √(𝜔 2 − 𝜔𝑑2 )2 + 4𝛾 2 𝜔𝑑2

As long as the damping is small enough, resonance occurs at a driving angular frequency 𝜔𝑅 = √𝜔 2 − 2𝛾 2 .

The resonance above represents the maximum amplitude for the varying charge on the capacitor. For the
maximum current to occur in the circuit, which corresponds to velocity resonance in the mechanical system,
one must apply an alternating voltage with angular frequency equal to 𝜔. Then the current will be perfectly
 120

in phase with the driving voltage. The power being transferred from the driving component to the system is
given by the product of the voltage and current, 𝑃 = 𝑉𝐼, and this is then a maximum.

10.4. The non-sinusoidally driven 𝑹𝑳𝑪-circuit

Suppose the voltage generator for a driven 𝑅𝐿𝐶-circuit did not give a sinusoidal voltage, but perhaps a square
or triangular or some other periodic waveform. If the driving wave is divided up into sinusoidal functions
using Fourier analysis, the effect of each sinusoidal driving part can be calculated separately and then added
to give the total result. We illustrate this process using a square wave.

Consider an 𝑅𝐿𝐶-circuit driven by a square-wave periodic voltage. It has the differential equation

𝑄 𝑑𝑄 𝑑2 𝑄
+𝑅 + 𝐿 2 = 𝑉[𝑡],
𝐶 𝑑𝑡 𝑑𝑡

where 𝑉[𝑡] is the driving square-wave voltage. Dividing by 𝐿 gives

𝑑2 𝑄 𝑑𝑄 1
2
+ 2𝛾 + 𝜔2 𝑄 = 𝑉[𝑡]
𝑑𝑡 𝑑𝑡 𝐿

Now from Fourier analysis (see Figure 54) a square wave can be written as

−1 if – 𝜋 < 𝑡 < 0 4 sin(𝑡) sin(3𝑡) sin(5𝑡)

𝑓[𝑡] = { = ( + + + ⋯ ).
1 if 0 < 𝑡 < 𝜋 𝜋 1 3 5

We adjust this as follows:

 This wave has unit amplitude: we multiply by 𝑉0 to get a driving voltage of amplitude 𝑉0 .
 This wave has period 2𝜋. To adjust it to a period 𝑇, we first compress it by 2𝜋 and then stretch it by 𝑇,
2𝜋
which is achieved by replacing 𝑡 with 𝑇
𝑡.

2𝜋 2𝜋 2𝜋
4𝑉0 sin( 𝑇 𝑡) sin(3 𝑡)
𝑇
sin(5 𝑡)
𝑇 4𝑉0
This gives the voltage 𝑉(𝑡) = 𝜋
( 1 + 3
+ 5
+ ⋯ ). After defining Ω = 𝜋𝐿
we get

2𝜋 6𝜋 10𝜋
𝑑2 𝑄 𝑑𝑄 sin ( 𝑇 𝑡) sin ( 𝑇 𝑡) sin ( 𝑇 𝑡)
+ 2𝛾 + 𝜔2 𝑄 = Ω ( + + + ⋯ ).
𝑑𝑡 2 𝑑𝑡 1 3 5

We proceed to deal with each term on the right hand side separately, using the fact that a driven harmonic
𝑑2 𝑄 𝑑𝑄 cos cos
oscillator 𝑑𝑡 2 + 2𝛾 𝑑𝑡 + 𝜔2 𝑄 = 𝑎0 { }(𝜔𝑑 𝑡) has steady-state solution 𝑄(𝑡) = 𝐴 { }(𝜔𝑑 𝑡 − 𝜑) where
sin sin
𝑎0 2𝛾𝜔𝑑
𝐴= and 𝜑 = arctan 𝜔2 −𝜔2 .
2 2 +4𝛾2 𝜔2
√(𝜔2 −𝜔𝑑 𝑑
) 𝑑

Suppose we have a system with 𝐶 = 1 nF, 𝐿 = 10 H (so 𝜔 = 10 kHz), 𝛾 = 1 kHz, and it is driven by a square-
wave voltage with amplitude 𝑉0 = 5 V (so Ω = 0.6366 Cs-2) and period 𝑇 = 2 ms:
 121

𝑑2 𝑄 𝑑𝑄
 Taking just the first driving term: + 2𝛾 + 𝜔2 𝑄 = Ω sin(1000𝜋𝑡), so 𝑎0 = Ω and 𝜔𝑑 = 1000𝜋,
𝑑𝑡 2 𝑑𝑡
with solution 𝑄 = 11.1 × 10−9 Ω sin(1000𝜋𝑡 − 0.070) = 7.05 sin(1000𝜋𝑡 − 0.070) [nC].
𝑑2 𝑄 𝑑𝑄 Ω Ω
 Taking just the second driving term: + 2𝛾 + 𝜔2 𝑄 = sin(3000𝜋𝑡), so 𝑎0 = and 𝜔𝑑 = 3000𝜋,
𝑑𝑡 2 𝑑𝑡 3 3
with solution 𝑄 = 15.2 × 10−9 Ω sin(3000𝜋𝑡 − 1.036) = 9.68 sin(3000𝜋𝑡 − 1.036) [nC].
𝑑2 𝑄 𝑑𝑄 Ω Ω
 Taking just the third driving term: 𝑑𝑡 2
+ 2𝛾 𝑑𝑡 + 𝜔2 𝑄 = 5 sin(5000𝜋𝑡), so 𝑎0 = 5
and 𝜔𝑑 = 5000𝜋,
−9
with solution 𝑄 = 1.33 × 10 Ω sin(5000𝜋𝑡 + 0.211) = 0.848 sin(5000𝜋𝑡 + 0.211) [nC].

More terms could be computed if necessary. The overall solution is the sum of these individual solutions,
namely

𝑄 = 7.05 sin(1000𝜋𝑡 − 0.070) + 9.68 sin(3000𝜋𝑡 − 1.036) + 0.848 sin(5000𝜋𝑡 + 0.211) [nC].

This can be differentiated to get the current, or divided by the capacitance to get the voltage over the
capacitor.
 122

11. Waves on a string

11.1. Introduction

Everybody is familiar with waves, and waves are important phenomena in physics. If one throws a stone into
still water, waves spread out from the point of impact. Other important examples are sound and light. These
are examples of waves, but what exactly is a wave? It is difficult to give a definition, but one way to think
about a wave is as a vibration of a spread-out (distributed) system rather than a point object.

In our study of vibrations we investigated the motion of a mass (point object) when it was acted on by various
forces i.e. the spring force, resistance force and driving force. But suppose we add to this original mass a
second spring and second mass. This system of these two coupled (joined) oscillators will have a complicated
motion. If we add a third spring and mass the motion will get more complicated still. But suppose we have
hundreds, or even thousands, of masses joined together by springs. Then if one side of this distributed system
is disturbed, the disturbance will travel (propagate) through it as a wave. Thus waves occur in systems with
many coupled oscillators.

The motion of a system with just a few coupled oscillators is something between simple harmonic motion
and wave motion. It is interesting and important (e.g. it determines the vibrations of molecules and machines
with a few large parts), but we shall rather continue directly with studying wave motion rather than giving it
any further attention. One way to start would be with the system of many masses joined by springs. If we
take into account that the springs themselves have mass, we can dispense with the masses and just consider
the motion of joined springs, or even one large spring. Indeed, anyone who has played with a “slinky” spring
is aware of the waves that can be set up in it.

Figure 58 Waves on a spring (a) longitudinal waves (b) transverse waves.

By pushing the end of the spring back and forth one can set up longitudinal waves, while by pushing it up and
down one can set up transverse waves. These are illustrated in Figure 58 Waves on a spring (a) longitudinal
waves (b) transverse waves.. One can also rapidly twist the end of the spring, and so-called torsional waves
 123

will travel along it. For each type of wave the type of displacement of the parts of the spring as the wave
passes differs: for longitudinal waves the displacement is parallel to the direction that the waves are moving
while for transverse waves the displacement is perpendicular to the direction that the waves are moving. For
torsional waves the displacement is through an angle around the direction the waves are travelling, which is
still perpendicular to the direction the wave is moving and so these waves may be considered a special case
of transverse waves. All these types of waves could occur simultaneously, but that would be very
complicated, so we usually consider them one at a time. The different types of waves can have different
speeds of propagation along the spring. Also notice that single pulses, instead of continuous waves, can be
made to propagate down the spring.

From the time of the ancient Greeks, wave motion has been studied by starting with waves on a stretched
string. This is similar to a spring, but if a string is compressed or twisted there is no restoring force, and so
the string cannot support longitudinal or torsional waves. Nevertheless, if a string is stretched then transverse
waves can be created. The reason for studying waves on a string was that music was considered very
important. Many musical instruments create sound from stretched strings, and the study of the relationship
between the properties of the stretched string and the sounds produced constitutes some of the earliest
science of mankind. Although we shall therefore initially concentrate on waves on a stretched string, all the
concepts will be applicable to other types of waves as well.

11.2. The stretched string

𝑚𝑔
Figure 59 Vibrating string.

We consider a string of length ℓ stretched between two fixed points (the tops of the triangles) and attached
via a pulley to a block of mass 𝑚 (Figure 59). Suppose the mass per unit length of the string is

𝑚string
𝜇=
ℓ

and the tension in the string is 𝑇. If the block is stationary then the sum of all the forces acting on it must be
zero, and so the tension in the string is 𝑇 = 𝑚𝑔 where 𝑚 is the mass of the block. Place a system of axes
over the string such that the origin corresponds to the left stationary end of the string and the right stationary
end of the string is at position 𝑥 = ℓ. Any point on the string may then be identified by its horizontal position
coordinate 𝑥 . Transverse displacements of the string are denoted by 𝑌 , and since the transverse
displacement depends on both the position and time we write 𝑌[𝑥, 𝑡].
 124

𝑌 𝑇
𝜃2
𝑑𝑠
𝜃
𝜃1
𝑑𝑥
𝑇

𝑥 𝑥 + 𝑑𝑥 𝑥
Figure 60 Element of vibrating string.

In order to find the equation of motion of the string, we consider a small section of the string from 𝑥 to 𝑥 +
𝑑𝑥 at some moment in time 𝑡 (see Figure 60). It is customary to refer to such a section as an “element”. At 𝑥
the string is displaced transversely by an amount 𝑌[𝑥, 𝑡] and at 𝑥 + 𝑑𝑥 it is displaced by an amount
𝑌[𝑥 + 𝑑𝑥, 𝑡]. Let the length of the element be 𝑑𝑠. If 𝜃 is the average angle of the element with respect to the
𝑑𝑥
horizontal axis then 𝑑𝑠 cos 𝜃 = 𝑑𝑠
, or

𝑑𝑥
𝑑𝑠 = . (65)
cos 𝜃

Since the mass per unit length of the string is 𝜇, the mass of the element is 𝜇 𝑑𝑠.

We shall assume that the magnitude of the force (i.e. the tension) in the string is constant everywhere. The
direction in which the tension acts, however, always lies along (i.e. parallel) to the string. Therefore the
tension acting on the right hand side of the element (at 𝑥 + 𝑑𝑥) has a different direction to the tension acting
on the left hand side of the element (at 𝑥). If the angles between these forces and the horizontal axis are
given by 𝜃1 and 𝜃2 , then the total transverse force on the element is given by 𝑇 sin 𝜃2 − 𝑇 sin 𝜃1 and
Newton's second law applied in the transverse direction is

𝜕2𝑌
𝑇 sin 𝜃2 − 𝑇 sin 𝜃1 = 𝜇 𝑑𝑠 (66)
𝜕𝑡 2

𝜕2 𝑌
where 𝜕𝑡 2 is the transverse acceleration of the element.17

We now introduce a very important assumption – we assume that the angles 𝜃, 𝜃1 and 𝜃2 are so small that
for any of them we may use the approximations18

17
The total horizontal force on the element is given by 𝑇 cos 𝜃2 − 𝑇 cos 𝜃1 , and so the element does experience a force, and
therefore an acceleration, in the X-direction as well. This will be ignored. We are also neglecting the fact that the tension will not
actually be constant throughout the string, and that its mass per unit length will change as it stretches. This analysis should be
regarded as a first approximation.

18
The following example may help to make this assumption more credible. A typical guitar string has length 650 mm and when
plucked may deviate as much as 4 mm from the equilibrium position.
 125

sin 𝜃 ≈ tan 𝜃 ≈ 𝜃

cos 𝜃 ≈ 1.

If we substitute cos 𝜃 ≈ 1 into equation (65), we obtain 𝑑𝑠 ≈ 𝑑𝑥 and therefore the mass of the element is
given by 𝜇 𝑑𝑥. If we substitute this, as well as sin 𝜃1 ≈ tan 𝜃1 and sin 𝜃2 ≈ tan 𝜃2 , into equation (66) we
obtain

𝜕2𝑌
𝑇 tan 𝜃2 − 𝑇 tan 𝜃1 = 𝜇 𝑑𝑥 (67)
𝜕𝑡 2

Note that tan 𝜃1 is the slope of the function 𝑌[𝑥, 𝑡] relative to the horizontal direction at 𝑥 , therefore
𝜕𝑌 𝜕𝑌
tan 𝜃1 = [𝑥, 𝑡]. Similarly tan 𝜃2 = [𝑥 + 𝑑𝑥, 𝑡]. Substituting these into equation (67) gives
𝜕𝑥 𝜕𝑥

𝜕2𝑌 𝜕𝑌 𝜕𝑌
𝜇 𝑑𝑥 2
= 𝑇 ( [𝑥 + 𝑑𝑥, 𝑡] − [𝑥, 𝑡]).
𝜕𝑡 𝜕𝑥 𝜕𝑥

Dividing by 𝜇 𝑑𝑥 yields

𝜕𝑌 𝜕𝑌
𝜕 2 𝑌 𝑇 𝜕𝑥 [𝑥 + 𝑑𝑥, 𝑡] − 𝜕𝑥 [𝑥, 𝑡]
= ( ).
𝜕𝑡 2 𝜇 𝑑𝑥

Since the element has a very small length and the accuracy of the derivation increases as this length is made
smaller, we take the limit as 𝑑𝑥 → 0, which gives

𝜕2𝑌 𝑇 𝜕2𝑌
= . (68)
𝜕𝑡 2 𝜇 𝜕𝑥 2

This is the equation of motion for the string. It is an example of the well-known linear wave equation, namely

𝜕2𝑌 2
𝜕2𝑌
= 𝑣 .
𝜕𝑡 2 𝜕𝑥 2

By comparison we find that

𝑣 = √𝑇/𝜇, (69)

and we will show that this corresponds to the speed of propagation of the wave along the string. The reader
should verify at this stage that the units are correct.

4 mm
The angle 𝜃 shown then has the value arctan 1 = 0.0123 rad ≈ 0.7˚.
(650 mm)
2
 126

11.3. Energy carried by a wave on a string

A wave possesses both kinetic and potential energy. We calculate these using the wave on the string as an
example.

𝜕𝑌
 A short piece of the string 𝑑𝑥 will have mass 𝜇 𝑑𝑥 and transverse velocity . The kinetic energy of
𝜕𝑡
1 𝜕𝑌 2
this piece is 𝑑𝐸𝐾 = 2
𝜇 ( 𝜕𝑡
) 𝑑𝑥.

 The short piece of string will also store potential energy when it is stretched. If the string is perfectly
horizontal, it will have a length 𝑑𝑥, but if it makes an angle to the horizontal it will have a longer
length 𝑑𝑠. We neglected this difference when deriving the wave equation (and for the kinetic energy
above!) but here it is vital. The change in transverse displacement over the distance 𝑑𝑥 is given by
𝜕𝑌 𝜕𝑌 2
𝑑𝑌 = 𝑑𝑥 , so using the theorem of Pythagoras one gets 𝑑𝑠 = √(𝑑𝑥)2 + ( 𝑑𝑥) = 𝑑𝑥 {1 +
𝜕𝑥 𝜕𝑥

𝜕𝑌 2 1/2 1 𝜕𝑌 2
( ) } ≈ 𝑑𝑥 {1 + ( ) }, assuming the slope is small and applying (1 + 𝑢)𝑛 ≈ 1 + 𝑛𝑢 for small
𝜕𝑥 2 𝜕𝑥
1 𝜕𝑌 2
𝑢 . The extension is therefore 𝑑𝑠 − 𝑑𝑥 = 𝑑𝑥 ( )
2 𝜕𝑥
and occurs at constant tension 𝑇 , so the
1 𝜕𝑌 2
potential energy of this piece is 𝑑𝐸𝑃 = 2 𝑇 (𝜕𝑥) 𝑑𝑥.

The total energy stored in the wave is therefore

1 𝜕𝑌 2 𝜕𝑌 2
𝐸 = ∫ 𝜇 ( ) + 𝑇 ( ) 𝑑𝑥. (70)
2 𝜕𝑡 𝜕𝑥

11.4. Solution of the wave equation

11.4.1. General solution

𝜕2 𝑌 𝜕2 𝑌
The general solution to the wave equation 𝜕𝑡 2 = 𝑣 2 𝜕𝑥 2 is

𝑌[𝑥, 𝑡] = 𝑓[𝑥 − 𝑣𝑡] + 𝑔[𝑥 + 𝑣𝑡] (71)

where 𝑓 and 𝑔 are any functions (provided their second order partial derivatives exist).

𝜕𝑌
This solution is easy to verify. If 𝑌[𝑥, 𝑡] = 𝑓[𝑥 − 𝑣𝑡] + 𝑔[𝑥 + 𝑣𝑡] then = −𝑣𝑓′[𝑥 − 𝑣𝑡] + 𝑣𝑔′[𝑥 + 𝑣𝑡] and
𝜕𝑡
𝜕2 𝑌 𝜕𝑌 𝜕2 𝑌
𝜕𝑡 2
= 𝑣 2 (𝑓′′[𝑥 − 𝑣𝑡] + 𝑔′′[𝑥 + 𝑣𝑡]) . Similarly 𝜕𝑥
= 𝑓′[𝑥 − 𝑣𝑡] + 𝑔′[𝑥 + 𝑣𝑡] and 𝜕𝑥 2
= 𝑓′′[𝑥 − 𝑣𝑡] +
𝜕2 𝑌 𝜕2 𝑌
𝑔′′[𝑥 + 𝑣𝑡]. It is then clear that 𝜕𝑡 2 = 𝑣 2 𝜕𝑥 2 , and so 𝑌[𝑥, 𝑡] = 𝑓[𝑥 − 𝑣𝑡] + 𝑔[𝑥 + 𝑣𝑡] is indeed a solution to
the wave equation. We motivate that it is the general solution by the fact that it contains two general
functions i.e. 𝑓 and 𝑔. These play a role similar to the integration constants in ordinary differential equations.
 127

11.4.2. Interpretation of the solution

𝜕2 𝑌 𝜕2 𝑌
We have just shown that the wave equation 𝜕𝑡 2 = 𝑣 2 𝜕𝑥 2 has solution 𝑌[𝑥, 𝑡] = 𝑓[𝑥 − 𝑣𝑡] + 𝑔[𝑥 + 𝑣𝑡]. But
how can we interpret this solution physically? For the moment, let us just concentrate on the first term
𝑓[𝑥 − 𝑣𝑡]. Initially, at time 𝑡 = 0, this term is just 𝑓[𝑥] which a function of arbitrary shape.

 At time 𝑡 = 1, this term becomes 𝑓[𝑥 − 𝑣]. It has shifted to the right by a distance 𝑣.
 At time 𝑡 = 2, this term becomes 𝑓[𝑥 − 2𝑣]. It has now shifted to the right by a distance 2𝑣.
 At time 𝑡 = 3, this term becomes 𝑓[𝑥 − 3𝑣] and is shifted to the right by a distance 3𝑣.

Clearly, the term 𝑓[𝑥 − 𝑣𝑡] represents a disturbance or pulse of shape 𝑓[𝑥] which keeps its original shape
but moves to the right at a speed 𝑣. A similar analysis applied to the second term in the solution of the wave
equation, namely 𝑔[𝑥 + 𝑣𝑡], shows that it represents a pulse which keeps its original (but possibly different)
shape and moves to the left at the same speed 𝑣.

The solution to the wave equation therefore consists of two moving pulses, one moving to the left and the
other to the right, but each having the same speed 𝑣. The pulses can be of arbitrary shape and do not have
to be of the sinusoidal form generally associated with waves.

1
Example: Consider the function 𝑌[𝑥, 𝑡] = . Is it a solution to the wave equation, and if so then
1+(𝑥−𝑣𝑡)2
describe the wave.

𝜕2 𝑌 𝜕2 𝑌
Solution: One way to check if this is a solution to the wave equation 𝜕𝑡 2 = 𝑣 2 𝜕𝑥 2 is by calculating the partial
derivatives and checking that the wave equation is satisfied. But such a direct method can be time consuming,
tedious and mathematically challenging.

Let us first consider how the function changes with time:

Y
𝑌[𝑥, 0] 𝑌[𝑥, 1] 𝑌[𝑥, 2]

𝑣 2𝑣 x

1
 Initially (at 𝑡 = 0) this function is 𝑌[𝑥, 0] = 1+(𝑥)2
1
 At time 𝑡 = 1 this function becomes 𝑌[𝑥, 1] = . It has shifted to the right by a distance 𝑣.
1+(𝑥−𝑣)2
1
 At time 𝑡 = 2 it becomes 𝑌[𝑥, 2] = 1+(𝑥−2𝑣)2 and has then shifted to the right by a distance 2𝑣.

It is clear that 𝑌[𝑥, 𝑡] is a function that maintains its shape but shifts to the right at speed 𝑣. It is therefore a
solution to the wave equation. Although the solution to the wave equation can have pulses (of different
shape) moving both to the left and to the right simultaneously, they don’t both have to be present. However,
if both are present they have to have the same speed. We can explain this more formally and mathematically
 128

as follows: since the wave equation has general solution 𝑌[𝑥, 𝑡] = 𝑓[𝑥 − 𝑣𝑡] + 𝑔[𝑥 + 𝑣𝑡], any function of
1
this form is a solution to the wave equation. Specifically, consider if we make the functions 𝑓[𝑢] = and
1+𝑢2
1 1
𝑔[𝑢] = 0. Then 𝑓[𝑥 − 𝑣𝑡] + 𝑔[𝑥 + 𝑣𝑡] = 1+(𝑥−𝑣𝑡)2 + 0 = 1+(𝑥−𝑣𝑡)2 , showing that this is indeed a solution
to the wave equation (with the part corresponding to the pulse travelling to the left being zero). In general
any function that we can show has the form 𝑓[𝑥 − 𝑣𝑡] + 𝑔[𝑥 + 𝑣𝑡] with suitable choices for the functions 𝑓
and 𝑔 is then a solution to the wave equation.

It may seem strange that the solution to the wave equation consists of pulses of arbitrary shape, instead of
what we would generally call waves. What we generally call waves are actually sinusoidal or harmonic waves.
They are just a special shape of pulse! Note that a pulse (i.e. a wave in general) is not periodic and does not
have a wavelength or frequency – those are properties of sinusoidal (harmonic) waves. Harmonic waves are
very important and will be discussed in more detail in the next chapter.

Figure 61 Moving wave on a string, with parts moving to the left and right.

Figure 61 shows a more complicated wave, namely

1 1
𝑌[𝑥, 𝑡] = 2
+
1 + (𝑥 − 𝑣𝑡 − 1) 1 + (𝑥 + 𝑣𝑡 − 3)2

which has parts moving both to the right and left. The wave is moving along a string, not over a surface! Note
that the one axis is time, while the other is position, and what is plotted is 𝑌(𝑥, 𝑡). If you choose any fixed
time, the figure shows you the displacement of the string at any position i.e. the shape of the wave. If you
chose a fixed position, the figure shows how the displacement of the string at that point varies with time. If
you look carefully at the figure, for increasing time values, you should be able to see how the pulses (moving
in opposite direction, but with the same speed) collide and how they reappear unaltered after the “collision”.
It is a basic property of wave pulses that they superimpose while overlapping but retain their original shape
as they propagate apart once again.
 129

11.5. Reflection and transmission of waves

Up to now, we have only considered waves travelling in a uniform (unchanging) medium with no boundary.
Suppose the medium suddenly changes. Then we have two media – the initial one and the new one. This may
happen, for example, if two strings of unequal thickness are joined together and a wave moves along the
string. As the incident wave reaches the interface (boundary) between the strings, it is partly reflected back
and partly transmitted.

Let us place our coordinate system so that the interface is at the origin (𝑥 = 0). Suppose the incident wave
is a pulse moving to the right with speed 𝑣1 , which we may represent by the function 𝑌[𝑥, 𝑡] = 𝐼[𝑥 − 𝑣1 𝑡]
for 𝑥 < 0. The transmitted wave also moves to the right, but since it is in a different medium it will have a
different speed. Let us represent it by the function 𝑌[𝑥, 𝑡] = 𝐽[𝑥 − 𝑣2 𝑡] for 𝑥 > 0. The reflected wave moves
to the left, but in the original medium, and can therefore be represented by 𝑌[𝑥, 𝑡] = 𝑅[𝑥 + 𝑣1 𝑡] for 𝑥 < 0.
Note that in the original medium the incident and reflected waves occur together and superimpose, and so
the total wave in the first medium is actually 𝐼[𝑥 − 𝑣1 𝑡] + 𝑅[𝑥 + 𝑣1 𝑡], while in the new medium there is only
the transmitted wave 𝐽[𝑥 − 𝑣2 𝑡].

At the interface (𝑥 = 0) the transverse displacement of the string can be expressed from the perspective of
the initial medium (i.e. 𝑥 → 0 from the left) as 𝐼[0 − 𝑣1 𝑡] + 𝑅[0 + 𝑣1 𝑡] = 𝐼[−𝑣1 𝑡] + 𝑅[𝑣1 𝑡], but from the
perspective of the new medium (i.e. 𝑥 → 0 from the right) it is 𝐽[−𝑣2 𝑡]. These must of course agree (since
there can only be one value for the displacement of the string at the interface) and so 𝐼[−𝑣1 𝑡] + 𝑅[𝑣1 𝑡] =
𝐽[−𝑣2 𝑡]. Letting 𝑢 = −𝑣1 𝑡, this can be written as

𝑣2
𝐼[𝑢] + 𝑅[−𝑢] = 𝐽 [ 𝑢]. (72)
𝑣1

Although we defined 𝑢 in a specific way, it can take on any value (as time changes) and so this equation is a
general relationship between the functions 𝐼, 𝑅 and 𝐽 that must always be satisfied.

If we assume the incident wave (𝐼) is known, we still have two unknowns (𝑅 and 𝐽) and so we require a second
condition in order to determine them. The second condition is that the vertical force on the string at the
interface considered from both sides must match or agree. Looking back at the derivation of the wave
equation on a string, we saw that the vertical force component at any position in the string is given by 𝐹𝑦 =
𝜕𝑌
𝑇 sin 𝜃 ≈ 𝑇 tan 𝜃 = 𝑇 where 𝑇 is the tension in the string. 19 In the initial medium the total wave is
𝜕𝑥
𝜕𝑌
𝐼[𝑥 − 𝑣1 𝑡] + 𝑅[𝑥 + 𝑣1 𝑡] and so the vertical force component 𝑇 𝜕𝑥 becomes 𝑇1 (𝐼′[𝑥 − 𝑣1 𝑡] + 𝑅′[𝑥 + 𝑣1 𝑡]).
At the interface (i.e. 𝑥 → 0 from the left) this is 𝑇1 (𝐼′[−𝑣1 𝑡] + 𝑅′[𝑣1 𝑡]). In the new medium the wave is just
𝐽[𝑥 − 𝑣2 𝑡] so the vertical force component is 𝑇2 𝐽′[𝑥 − 𝑣2 𝑡], which at the interface (i.e. 𝑥 → 0 from the right)
is 𝑇2 𝐽′[−𝑣2 𝑡]. Therefore we require that 𝑇1 (𝐼′[−𝑣1 𝑡] + 𝑅′[𝑣1 𝑡]) = 𝑇2 𝐽′[−𝑣2 𝑡]. After integrating both sides
with respect to 𝑡 one gets

𝐼[−𝑣1 𝑡] 𝑅[𝑣1 𝑡] 𝐽[−𝑣2 𝑡]

𝑇1 ( + ) = 𝑇2 ( ) + 𝑐.
−𝑣1 𝑣1 −𝑣2

19
When deriving the equation for a wave on a string we assumed the tension was constant, and so we may be tempted to simplify
matters and just say that the slope of the string must be the same on both sides of the interface. That would not be wrong. However,
suppose the two strings were tied to opposite sides of a small ring that could slide along a rod without friction, but could not move
horizontally. Then the tension on either side of the interface could be made different. This may sound contrived and rather unrealistic
practically. But for other types of wave motion the changing “tension” occurs naturally, and we want our result to be applicable to
all waves in general.
 130

where 𝑐 is an integration constant. One can expect the physically the incident, transmitted and reflected
waves (pulses) must all tend to zero at infinity – if this is so then the functions 𝐼, 𝑅 and 𝐽 must all tend to zero
for large inputs, and then for the equation above to hold it means 𝑐 = 0. If one now also introduces the
impedance for a wave on a string as
𝑇
𝑍= (73)
𝑣

then this condition becomes −𝑍1 (𝐼[−𝑣1 𝑡] − 𝑅[𝑣1 𝑡]) = −𝑍2 𝐽[−𝑣2 𝑡], or

𝑣2
𝑍1 (𝐼[𝑢] − 𝑅[−𝑢]) = 𝑍2 𝐽 [ 𝑢]. (74)
𝑣1

11.5.1. Transmitted wave

To find the relationship between the transmitted and incident waves, we eliminate 𝑅 between equations
(72) and (74) by first dividing equation (74) by 𝑍1 and then adding them to get

𝑣2 𝑍2 𝑣2 𝑍1 + 𝑍2 𝑣2
2𝐼[𝑢] = 𝐽 [ 𝑢] + 𝐽 [ 𝑢] = 𝐽 [ 𝑢]
𝑣1 𝑍1 𝑣1 𝑍1 𝑣1

𝑣 2𝑍1
so that 𝐽 [𝑣2 𝑢] = 𝑍 𝐼[𝑢]. But the transmitted wave is given by 𝐽[𝑥 − 𝑣2 𝑡]. To get the left hand side into
1 1 +𝑍2
𝑣
this form we must set 𝑢 = 𝑣1 𝑥 − 𝑣1 𝑡, yielding
2

2𝑍1 𝑣1
𝐽[𝑥 − 𝑣2 𝑡] = 𝐼 [ 𝑥 − 𝑣1 𝑡]. (75)
𝑍1 + 𝑍2 𝑣2
2𝑍1
This shows the size of the transmitted wave is the fraction 𝑡 = 𝑍 of the size of the incident wave, which
1 +𝑍2
we call the transmission amplitude ratio (not to be confused with time). Note that the function of the right
𝑣
is not exactly the incident wave, which is actually 𝐼[𝑥 − 𝑣1 𝑡] rather than 𝐼 [ 1 𝑥 − 𝑣1 𝑡]. This means the shape
𝑣2
𝑣
(or form) of the transmitted wave is only similar to the shape of the incident wave, but the change 𝑥 → 𝑣1 𝑥
2
means that the transmitted wave is compressed by a factor 𝑣1 and then stretched by a factor 𝑣2 compared
to the incident wave. This is, of course, due to the different wave speeds in the two media.

11.5.2. Reflected wave

To find the relationship between the reflected and incident waves, we can eliminate 𝐽 between equations
(72) and (74) by first dividing equation (74) by 𝑍2 and then subtracting to get

𝑍1
𝐼[𝑢] + 𝑅[−𝑢] − (𝐼[𝑢] − 𝑅[−𝑢]) = 0,
𝑍2

𝑍1 − 𝑍2
𝑅[−𝑢] = 𝐼[𝑢].
𝑍1 + 𝑍2

But the reflected wave is given by 𝑅[𝑥 + 𝑣1 𝑡]. To get the left hand side into this form we must take 𝑢 = – 𝑥 −
𝑣1 𝑡, giving
𝑍1 − 𝑍2
𝑅[𝑥 + 𝑣1 𝑡] = 𝐼[– 𝑥 − 𝑣1 𝑡]. (76)
𝑍1 + 𝑍2
 131

𝑍1 −𝑍2
This shows the size of the reflected wave is the fraction 𝑟 = of the size of the incident wave, which we
𝑍1 +𝑍2
call the reflection amplitude ratio. If 𝑍2 > 𝑍1 (which, if the tension is constant, means the second string is
thicker than the first) then 𝑟 is negative and the reflected wave is turned upside down! Actually, the function
on the right is not exactly the incident wave 𝐼[𝑥 − 𝑣1 𝑡] because of the minus sign in front of the 𝑥. A change
of 𝑥 to −𝑥 represents a reflection in the Y-axis, and so the shape of the reflected wave is a mirror reflection
of the incident wave.

11.5.3. Transmission and reflection of the force wave

We defined 𝑌[𝑥, 𝑡] to be the transverse deflection of the string. Also, at each point and at any time on the
𝜕𝑌
string there is the transverse force given by 𝐹𝑦 [𝑥, 𝑡] = 𝑇 𝜕𝑥 . Therefore we can consider that there is a force
wave that travels together with the deflection wave in the string. Of course there is only one wave, but one
can consider both of these aspects of it. When viewing the wave on a string with one’s eyes, one will see the
displacement 𝑌. But if you close your eyes and place your fingers on the string, you will feel the wave pass:
what you will perceive if the force produced by the wave. We will refer to 𝑌[𝑥, 𝑡] as the displacement wave,
and to 𝐹𝑦 [𝑥, 𝑡] as the corresponding force wave.

NB: if one just says ‘wave’, you can assume the displacement wave is meant.

Sometimes the force wave is more important e.g. when considering sound: instead of viewing the movement
of the air molecules, one’s ear feels the changes in pressure. It is important to note that the transmission and
reflection amplitude ratios derived above for the (normal) wave cannot also be applied to the force wave.

If the incident wave is 𝐼[𝑥 − 𝑣1 𝑡] then the corresponding incident force wave is 𝑇1 𝐼′[𝑥 − 𝑣1 𝑡].

 Likewise, if the transmitted wave is 𝐽[𝑥 − 𝑣2 𝑡] then the corresponding transmitted force wave is
𝑇2 𝐽′[𝑥 − 𝑣2 𝑡]. We have already derived (see equation (75)) that the transmitted wave is given by
2𝑍1 𝑣 2𝑍1 𝑣 𝑣
𝐽[𝑥 − 𝑣2 𝑡] = 𝑍 𝐼 [𝑣1 𝑥 − 𝑣1 𝑡] and so its corresponding force wave is 𝑇2 𝑍 𝐼′ [𝑣1 𝑥 − 𝑣1 𝑡] 𝑣1 .
1 +𝑍2 2 1 +𝑍2 2 2
The ratio of the size of the transmitted to incident force waves is then

2𝑍 𝑣 𝑇2 2𝑍1 2𝑍
𝑇2 𝑍 +1𝑍 𝑣1 𝑣2 𝑍1 + 𝑍2 𝑍2 𝑍 +1𝑍 2𝑍2
1 2 2 1 2
𝓉 = = = = .
𝑇1 𝑇1 /𝑣1 𝑍1 𝑍1 + 𝑍2

 If the reflected wave is 𝑅[𝑥 + 𝑣1 𝑡] then the corresponding reflected force wave is 𝑇1 𝑅′[𝑥 + 𝑣1 𝑡].
𝑍 −𝑍
Equation (76) gives the reflected wave as 𝑅[𝑥 + 𝑣1 𝑡] = 𝑍1 +𝑍2 𝐼[– 𝑥 − 𝑣1 𝑡] and so the corresponding
1 2
𝑍1 −𝑍2
reflected force wave is 𝑇1 𝐼′[– 𝑥 − 𝑣1 𝑡](−1). Hence the ratio of the sizes of the reflected and
𝑍1 +𝑍2
incident force waves is
𝑍 −𝑍
𝑇1 𝑍1 + 𝑍2 (−1) 𝑍2 − 𝑍1
1 2
𝓇 = = .
𝑇1 𝑍1 + 𝑍2

Note that although the transmission and reflection amplitude ratios of the force wave, 𝓉 and 𝓇, are different
from those of the displacement wave (𝑡 and 𝑟), one just has to swap the values of 𝑍1 and 𝑍2 to change
between them. Also note that so 𝓇 = −𝑟 so these reflection amplitude ratios have the same size but
opposite sign.
 132

11.6. Impedance
𝑇
Earlier we gave equation (73) for impedance, namely = 𝑣 . There are many alternative ways to express the
impedance. Since 𝑣 = √𝑇/𝜇, one can also write it as
𝑇
𝑍= = √𝑇𝜇,
√𝑇/𝜇

or alternatively as
𝜇𝑣 2
𝑍= = 𝜇𝑣.
𝑣

Actually, the impedance can be defined more generally for a wave moving in a particular direction as the
ratio of the transverse force in the string to the transverse velocity (i.e. how fast the string moves up or
down). This is because
𝜕𝑌
𝐹𝑦 𝑇 𝑇
𝑍 = | | = | 𝜕𝑥 | =
𝑣𝑦 𝜕𝑌 𝑣
𝜕𝑡

where the last step, for a wave moving to the right, follows after taking the two different types of partial
derivatives of a general wave 𝑌[𝑥, 𝑡] = 𝑓[𝑥 − 𝑣𝑡]. Be sure to note that 𝑣𝑦 refers to the transverse velocity of
a piece of the string, while 𝑣 refers to the speed of the wave along the string.

Example: We have found that the transmission amplitude ratio (for the displacement wave) is given in terms
2𝑍1
of impedance as 𝑡 = . Using the different expressions for the impedance, this transmission amplitude
𝑍1 +𝑍2
ratio can be expressed in terms of other variables. Show that if the tension is constant, the transmission
2𝑣2
amplitude ratio is given by 𝑣 .
1 +𝑣2

2𝑍1
Solution: The transmission amplitude ratio is given by 𝑡 = , but we can express 𝑍 = 𝑇/𝑣. This means
𝑍1 +𝑍2
2𝑇1 /𝑣1 2/𝑣1 2𝑣2
that 𝑡 = and if 𝑇1 = 𝑇2 then 𝑡 = . Multiplying top and bottom by 𝑣1 𝑣2 gives 𝑡 = .
𝑇1 /𝑣1 +𝑇2 /𝑣2 1/𝑣1 +1/𝑣2 𝑣2 +𝑣1

11.7. Reflections of a wave from the end of a string

11.7.1. Fixed or loose ends

If a string ends, there can only be incident and reflected waves but no transmitted wave. One can do the
analysis in detail with only these two waves, but let us first consider a “short-cut” method.

 Suppose the end of the string is fixed. This is similar to the string being joined to a second infinitely
2𝑍1
thick string. Then 𝑍2 = √𝑇2 𝜇2 = ∞. This means the transmission amplitude ratio 𝑡 = 𝑍 = 0,
1 +𝑍2
 133

𝑍1 −𝑍2
corresponding to no wave being transmitted. Also the reflection amplitude ratio 𝑟 = = −1,
𝑍1 +𝑍2
meaning that the reflected wave has the same size as the incident wave but is turned upside down!
 Suppose the end of the string is loose. This is similar to the string being joined to a second infinitely
𝑍 −𝑍
thin string. Then 𝑍2 = √𝑇2 𝜇2 = 0. The reflection amplitude ratio is then 𝑟 = 𝑍1 +𝑍2 = +1, meaning
1 2
that the reflected wave has the same size as the incident wave (and is not upside down). But now
2𝑍1
the transmission amplitude ratio is 𝑡 = 𝑍 = 2, which is worrying as we do not want a transmitted
1 +𝑍2
wave. Since the second string is infinitely thin, however, we can ignore it.

We now do the same problem more formally. Suppose we have the wave reaching the end of the string.
Since there is only incident and reflected waves, the total wave on the string is 𝐼[𝑥 − 𝑣𝑡] + 𝑅[𝑥 + 𝑣𝑡].

 Suppose the end of the string (at 𝑥 = 0) is fixed. Then 𝐼[−𝑣𝑡] + 𝑅[+𝑣𝑡] = 0, or in general 𝑅[𝑢] =
−𝐼[−𝑢]. Since the reflected wave is actually 𝑅[𝑥 + 𝑣𝑡], we use 𝑢 = 𝑥 + 𝑣𝑡 and so obtain

𝑅[𝑥 + 𝑣𝑡] = −𝐼[−𝑥 − 𝑣𝑡 ].

The reflected wave is similar to the incident wave, but turned upside down and reflected in the Y-
axis. Clearly the reflection amplitude ratio is −1 as we had before.

 Suppose the end of the string is loose. Then the force at the end of the string (at 𝑥 = 0) must be zero.
𝜕𝑌
The force along the string is given by 𝑇 = 𝑇(𝐼′[𝑥 − 𝑣𝑡] + 𝑅′[𝑥 + 𝑣𝑡]) and at the string therefore
𝜕𝑥
′ [𝑢]
𝑇(𝐼′[−𝑣𝑡] + 𝑅′[𝑣𝑡]) = 0, so that 𝑅 = −𝐼′[−𝑢]. After integrating one gets 𝑅[𝑢] = 𝐼[−𝑢] where
we have left off the integration constant if both waves tend to zero at infinity. To get the correct
form for the reflected wave we use 𝑢 = 𝑥 + 𝑣𝑡 , giving
𝑅[𝑥 + 𝑣𝑡] = 𝐼[−𝑥 − 𝑣𝑡 ].

The reflected wave is similar to the incident wave and the same way up, but reflected in the Y-axis.
Clearly the reflection amplitude ratio is +1 as we had before.

11.7.2. End with resistance

Now suppose that instead of being completely fixed or completely loose at the end, the string was tied at its
end to a (massless) ring which slides up-and-down along a rod with friction −𝛽𝑣.20 The condition for the end
(𝑥 = 0) is then that there is no net force on the ring. Since there are only incident and reflected waves, the
total wave on the string is 𝐼[𝑥 − 𝑣𝑡] + 𝑅[𝑥 + 𝑣𝑡].

𝜕𝑌
The force that the string exerts on the ring at its end is −𝑇 𝜕𝑥 [0, 𝑡] = −𝑇(𝐼′[−𝑣𝑡] + 𝑅′[𝑣𝑡]), while the
𝜕𝑌
friction on the ring is −𝛽 𝜕𝑡 [0, 𝑡] = −𝛽(−𝑣𝐼′[−𝑣𝑡] + 𝑣𝑅′[𝑣𝑡]). We therefore require that

−𝑇(𝐼′[−𝑣𝑡] + 𝑅′[𝑣𝑡]) − 𝛽(−𝑣𝐼′[−𝑣𝑡] + 𝑣𝑅′[𝑣𝑡]) = 0

20
This may sound contrived for a wave on string, but later we will show it is applicable and important for electrical waves. We may
be able to achieve it for a string if we orientated it vertically and dipped the end into a fluid with the necessary viscosity.
 134

or
𝑍(𝐼′[−𝑢] + 𝑅′[𝑢]) = 𝛽(𝐼′[−𝑢] − 𝑅′[𝑢]).

After integrating (and dropping the integration constant as before), one obtains 𝑍(−𝐼[−𝑢] + 𝑅[𝑢]) =
𝑍−𝛽
𝛽(−𝐼[−𝑢] − 𝑅[𝑢]) which can be rewritten as 𝑅[𝑢] = 𝑍+𝛽 𝐼[−𝑢]. To get the reflected wave we use 𝑢 = 𝑥 +
𝑣𝑡, giving
𝑍−𝛽
𝑅[𝑥 + 𝑣𝑡] = 𝐼[−𝑥 − 𝑣𝑡].
𝑍+𝛽

This is identical to the reflection formula for two joined strings, but with 𝛽 taking the place of the impedance
𝑍2 of the second string.

 If 𝛽 = 𝑍 then there is no reflected wave - the incident wave is completely absorbed!

 If 𝛽 = 0 then there is no resistance and this must correspond to reflection from a loose end (the
reader should verify that indeed it does).
 If 𝛽 → ∞ then the very large resistance will prevent the end of the string from moving, and the result
must correspond to reflection from a fixed end (which the reader should again verify).

11.7.3. End with mass

Suppose that there is a mass 𝑚 at the end of the string (𝑥 = 0). Then the transverse force on this mass due
𝜕𝑌
to the end of the string, namely −𝑇 | = −𝑇(𝐼′[−𝑣𝑡] + 𝑅′[𝑣𝑡]), must by Newton's second law be equal
𝜕𝑥 𝑥=0
to the mass times its transverse acceleration. Since the transverse displacement of any point along the string
is 𝐼[𝑥 − 𝑣𝑡] + 𝑅[𝑥 + 𝑣𝑡], the transverse acceleration is obtained by differentiating twice with respect to time
and at the end of the string is 𝑣 2 (𝐼′′[−𝑣𝑡] + 𝑅′′[𝑣𝑡]). Therefore Newton's second law gives

−𝑇(𝐼′[−𝑣𝑡] + 𝑅′[𝑣𝑡]) = 𝑚𝑣 2 (𝐼′′[−𝑣𝑡] + 𝑅′′[𝑣𝑡])

or −𝑍(𝐼′[−𝑢] + 𝑅′[𝑢]) = 𝑚𝑣(𝐼′′[−𝑢] + 𝑅′′[𝑢]). We can integrate this (neglecting the constant) to get

𝑚𝑣
𝐼[−𝑢] − 𝑅[𝑢] = (−𝐼′[−𝑢] + 𝑅′[𝑢])
𝑍
but then is difficult to proceed because of the remaining derivatives. However, we can easily consider two
special cases:

 If 𝑚 = 0 (loose end): in this case the right hand side is zero, so 𝑅[𝑢] = 𝐼[−𝑢] and therefore
𝑅[𝑥 + 𝑣𝑡] = 𝐼[−𝑥 − 𝑣𝑡 ]. This is or earlier result and means the incident and reflected waves have
the same form but are mirror images of one another.
 If 𝑚 → ∞ (fixed end): in this case, to keep the right hand side finite we need the part in brackets to
tend to zero, and so 𝑅′[𝑢] = 𝐼′[−𝑢]. Integrating and neglecting the constant gives 𝑅[𝑢] = −𝐼[−𝑢]
so that the reflected part is 𝑅[𝑥 + 𝑣𝑡] = −𝐼[−𝑥 − 𝑣𝑡 ]. The reflected wave has the same form as the
incident wave but they are mirror images of one another and the reflected wave is upside down.

In general the reflected pulse will not be the same shape as the incident pulse.
 135

11.8. Problems

1. (a) What is the percentage error in making the approximations:

i. sin 𝜃 ≈ 𝜃
ii. tan 𝜃 ≈ 𝜃
iii. cos 𝜃 ≈ 1
when 𝜃 = 15°? [Note: the approximations are for angles in radians, not degrees.]

(b) By how much must the centre of a 1 m long string lying horizontally and fixed at both ends be lifted
from its equilibrium position so that each half makes an angle of 15˚ to the horizontal?

(c) Considering in general how far the strings of musical instruments move when played, do you think the
approximations above are justified when considering the sounds from musical instruments?

2. Consider the general solution to the wave equation, namely 𝑌[𝑥, 𝑡] = 𝑓[𝑥 − 𝑣𝑡] + 𝑔[𝑥 + 𝑣𝑡]. If an
infinitely long string is pulled into the shape of some function, say 𝜙[𝑥] and then released from this
𝜕𝑌
position with no initial velocity at each point, the initial conditions are 𝑌[𝑥, 0] = 𝜙[𝑥] and [𝑥, 0] = 0.
𝜕𝑡
𝜕𝑌
Calculate and substitute these initial conditions to show that
𝜕𝑡
1 1
𝑌[𝑥, 𝑡] = 𝜙[𝑥 − 𝑣𝑡] + 𝜙[𝑥 + 𝑣𝑡],
2 2
which means that the resulting wave consists of two parts with the same shape as the initial distortion
(but each having half the amplitude) travelling in opposite directions with the same speed away from the
distorted area.

3. If the tension in a string consisting of two joined pieces of unequal thickness is constant, show that the
𝑣 −𝑣
reflection amplitude ratio is given by 𝑟 = 𝑣2 +𝑣1.
2 1

4. If the tension in a string consisting of two joined pieces of unequal thickness is constant, find expressions
for the transmission and reflection amplitude ratios in terms of the linear densities 𝜇1 and 𝜇2 .

5. (a) A 10 m long string is made by joining two 5 m lengths. However, the second part of the string is slightly
thicker and has a linear density 80% greater than the first half. If the entire string is placed under tension
𝑇 and a wave moves through the first lighter part of the string with speed 𝑣1 , what would its speed 𝑣2
be in the second part of the string (in terms of 𝑣1 )?

(b) If the incident wave has an amplitude of 3 cm, what will be the amplitudes of the transmitted and
reflected waves caused by the change in linear density?
 136

12. Sinusoidal Waves

12.1. Properties of sinusoidal waves

𝑌[𝑥, 𝑡]
𝑡

Figure 62 Sinusoidal wave.

𝜕2 𝑌 𝜕2 𝑌
We have seen that the general solution to the wave equation = 𝑣2 is given by 𝑌[𝑥, 𝑡] = 𝑓[𝑥 − 𝑣𝑡] +
𝜕𝑡 2 𝜕𝑥 2
𝑔[𝑥 + 𝑣𝑡]. This consists of two pulses of arbitrary shape travelling to the left and to the right respectively at
the same speed 𝑣. Although the form of the pulses may be arbitrary, a very important case is when we take
them to be sinusoidal. For a wave travelling to the right we may take

𝑌[𝑥, 𝑡] = 𝐴 cos(𝜔𝑡 − 𝑘𝑥 − 𝜑). (77)

Figure 62 shows such a sinusoidal wave. Does it satisfy the wave equation? Direct substitution, as the reader
should verify, shows that it does as long as

𝜔
𝑣= . (78)
𝑘
𝜔
A shorter method is just to rewrite it as 𝑌[𝑥, 𝑡] = 𝐴 cos (−𝑘 (−𝑘 𝑡 + 𝑥) − 𝜑), where it is now clearly in the
𝜔
form 𝑓[𝑥 − 𝑣𝑡] with 𝑣 = 𝑘
and thus a solution to the wave equation travelling to the right.21 This wave has

21
A sinusoidal wave travelling to the left can be written as 𝑌(𝑥, 𝑡) = 𝐴 cos(𝜔𝑡 + 𝑘𝑥 − 𝜑). In general, a sinusoidal wave could, of
course, have parts travelling to the left and right simultaneously.
 137

amplitude 𝐴 , angular frequency 𝜔 and phase constant 𝜑 . It also has a new parameter we have not
encountered before, namely 𝑘.22 The quantity 𝑘 is measured in m-1 and is called the wave number.23

Why is this case of sinusoidal waves so important?

 Firstly, many waves in nature do have this sinusoidal form e.g. sound waves, electromagnetic waves
(including light) and water waves. Often this can be traced back to a vibrating source e.g. a tuning
fork for sound. Although we may say that waves are sinusoidal, one should be aware that unless the
wave is infinitely long (which is impossible) it is actually at best a truncated sinusoidal wave.
 Secondly, we have seen that using Fourier analysis we are able to create other functions by adding
sinusoidal functions. If we take a sinusoidal wave as a basic type, by adding many of them together
we can create pulses of arbitrary shape. So we can regard any wave as being composed of sinusoidal
waves.
 Thirdly, using sinusoidal waves will allow us to consider more general types of waves, namely those
that undergo dispersion.

Note that a sinusoidal wave may also be expressed in the complex exponential form as

𝑌[𝑥, 𝑡] = 𝐶̃ 𝑒 𝑖(𝜔𝑡±𝑘𝑥) (79)

𝜔
with 𝑣 = 𝑘 . Using only the real part is implied, and the plus or minus signs correspond to waves travelling to
the left or right respectively.

12.1.1. Periodicity of a sinusoidal wave

When a sinusoidal wave with angular frequency 𝜔 travels down a string, each part of the string (fixed 𝑥-
position) behaves as a simple harmonic oscillator with angular frequency 𝜔. In fact, a sinusoidal wave may
be generated in the string by moving a point in the string sinusoidally up and down. Sinusoidal waves display
periodicity in both the 𝑥 and 𝑡 variables i.e. in position and time:

 The period in time (𝑡) is just called the period and is denoted here by 𝑇.24 By the definition of a period
we must have 𝑌[𝑥, 𝑡 + 𝑇] = 𝑌[𝑥, 𝑡], but

𝑌[𝑥, 𝑡 + 𝑇] = 𝐴 cos(𝜔{𝑡 + 𝑇} ± 𝑘𝑥 − 𝜑) = 𝐴 cos(𝜔𝑡 + 𝜔𝑇 ± 𝑘𝑥 − 𝜑).

This is only equal to 𝑌[𝑥, 𝑡] if 𝜔𝑇 = 2𝜋 or25

2𝜋
𝑇= . (80)
𝜔

 The period in position (𝑥) is called the wavelength and is denoted by 𝜆. By the definition of period
we must have 𝑌[𝑥 + 𝜆, 𝑡] = 𝑌[𝑥, 𝑡], but

22
Do not confuse the symbol 𝑘 used previously for the spring constant with its present use for wavenumber.
23
Just as there is the angular frequency 𝜔 and the normal frequency 𝑓, so too is there an angular wavenumber and the normal
wavenumber. As given here, our wavenumber is actually the angular version: however, since we will only ever refer to it and never
the normal wavenumber, for simplicity we will just call it the wavenumber.
24
Do not confuse the symbol 𝑇 used previously for tension with its present use for period.
25
Mathematically one could say we require 𝜔𝑇 = 2𝜋𝑚 where 𝑚 is any integer. However, we take the period as the smallest shift
that reproduces the function.
 138

𝑌[𝑥 + 𝜆, 𝑡] = 𝐴 cos(𝜔𝑡 ± 𝑘{𝑥 + 𝜆} − 𝜑) = 𝐴 cos(𝜔𝑡 ± 𝑘𝑥 ± 𝑘𝜆 − 𝜑).

This is only equal to 𝑌[𝑥, 𝑡] if 𝑘𝜆 = 2𝜋 or

2𝜋
𝜆= . (81)
𝑘

The frequency is the inverse of the period, so

1 𝜔
𝑓= = . (82)
𝑇 2𝜋

The relationship of equation (78) may therefore be expressed in terms of 𝜆 and 𝑓 by using equations (81) and
(82) as

𝜔 2𝜋𝑓
𝑣= = = 𝑓𝜆. (83)
𝑘 2𝜋/𝜆

12.1.2. Reflection and transmission of a sinusoidal wave

v v

Figure 63 Reflection and transmission of waves.

We have already discussed the transmission and reflection of waves in general, and sinusoidal waves are just
a special case. Consider a string made up of two parts, as in Figure 63. Now consider an incident sinusoidal
wave travelling to the right with wave speed 𝑣1 . At the junction (interface) between the media two new
sinusoidal waves are generated, namely the transmitted wave travelling further towards the right with wave
speed 𝑣2 and the reflected wave travelling to the left with wave speed 𝑣1 . At the interface between two
 139

media, the point of the junction may be viewed as an oscillator which is set in motion by the incident wave.
It therefore oscillates with angular frequency 𝜔. This oscillator generates both the transmitted and the
reflected waves and therefore the angular frequency of both these waves is also 𝜔. This may be considered
as a general rule: as a sinusoidal wave crosses or is reflected at the interface between two media, its angular
frequency remains constant.

However, since the speeds in the different media differ, the wavelengths of the waves also differ. These
wavelengths are given by

2𝜋 2𝜋𝑣1 2𝜋 2𝜋𝑣2
𝜆1 = = and 𝜆2 = =
𝑘1 𝜔 𝑘2 𝜔

and because the angular frequency remains unchanged

𝜆1 𝜆2
= . (84)
𝑣1 𝑣2

Example: We have already seen that the transmission and reflection amplitude ratios of a wave at an
2𝑍1 𝑍 −𝑍
interface are given by 𝑡 = 𝑍 and 𝑟 = 𝑍1 +𝑍2 respectively. For a sinusoidal wave on a string we can write
1 +𝑍2 1 2
the impedance as
𝑇 𝑇𝑘
= .𝑍=
𝑣 𝜔
Since 𝜔 is the same for all the waves, if 𝑇 is unchanged at the interface then the impedance is only affected
by the wavenumber, and it is easy to see that the transmission and reflection amplitude ratios are given in
2𝑘1 𝑘 −𝑘
terms of the wavenumbers by 𝑡 = 𝑘 and 𝑟 = 𝑘1 +𝑘2 respectively.
1 +𝑘2 1 2

12.1.3. Energy of a sinusoidal wave

1 𝜕𝑌 2 𝜕𝑌 2
Equation (70) gives the energy for a wave on a string as 𝐸 = 2 ∫ 𝜇 ( 𝜕𝑡 ) + 𝑇 (𝜕𝑥) 𝑑𝑥. Taking 𝑌[𝑥, 𝑡] =
𝐴 cos(𝜔𝑡 − 𝑘𝑥 − 𝜑), we can substitute in its derivatives, as well the fact that 𝑇 = 𝜇𝑣 2, to get

1
𝐸 = 𝜇 ∫{−𝐴𝜔 sin(𝜔𝑡 − 𝑘𝑥 − 𝜑)}2 + 𝑣 2 {𝐴𝑘 sin(𝜔𝑡 − 𝑘𝑥 − 𝜑)}2 𝑑𝑥.
2
𝜔
Since 𝑣 = this becomes
𝑘

1
𝐸 = 𝜇𝐴2 𝜔2 ∫ 2 sin2 (𝜔𝑡 − 𝑘𝑥 − 𝜑) 𝑑𝑥.
2
1
The average value of sin2 𝑢 over one oscillation is 2, so the value of the integral above for one full oscillation
1
is equal to the wavelength 𝜆 and so the energy in one oscillation is 𝐸 = 2 𝜇𝐴2 𝜔2 𝜆. This takes one period to
pass a given point and since the wavelength divided by the period is the wave speed, the rate of energy flow
past a fixed point (the power carried by the wave) is

1 1
𝑃 = 𝜇𝐴2 𝜔2 𝑣 = 𝑍𝐴2 𝜔2 (85)
2 2
 140

since 𝑍 = 𝜇𝑣. The power carried by a sinusoidal wave is proportional to the square of the amplitude and the
square of the frequency, as well as the impedance.

Example: Show that when a wave is partially transmitted and reflected, the power is conserved.

Solution: The fraction of the incident power that is transmitted is

1 2 2
𝑃𝐽 2 𝑍2 𝐴𝐽 𝜔 𝑍2 𝐴𝐽 2 𝑍2 2𝑍1 2 4𝑍1 𝑍2
= = ( ) = ( ) = .
𝑃𝐼 1 𝑍 𝐴2 𝜔 2 𝑍1 𝐴𝐼 𝑍1 𝑍1 + 𝑍2 (𝑍1 + 𝑍2 )2
2 1 𝐼

The fraction of the incident power that is reflected is

1 2 2
𝑃𝑅 2 𝑍1 𝐴𝑅 𝜔 𝐴𝑅 2 𝑍1 − 𝑍2 2
= =( ) =( ) .
𝑃𝐼 1 𝑍 𝐴2 𝜔 2 𝐴𝐼 𝑍1 + 𝑍2
2 1 𝐼
The sum of these fractions is
𝑃𝐽 𝑃𝑅 4𝑍1 𝑍2 + (𝑍1 − 𝑍2 )2
+ = = 1,
𝑃𝐼 𝑃𝐼 (𝑍1 + 𝑍2 )2

therefore 𝑃𝐽 + 𝑃𝑅 = 𝑃𝐼 which means the power is conserved.

12.1.4. Power in terms of the force wave amplitude

We have derived that the power carried by a sinusoidal wave 𝑌[𝑥, 𝑡] = 𝐴 cos(𝜔𝑡 − 𝑘𝑥 − 𝜑) is given by
1
𝑃 = 𝑍𝐴2 𝜔2. Note that 𝐴 is the amplitude of the displacement wave. The associated force wave is 𝐹𝑦 =
2
𝜕𝑌 𝜔
𝑇 𝜕𝑥 = 𝑇𝐴𝑘 sin(𝜔𝑡 − 𝑘𝑥 − 𝜑), with amplitude 𝐵 = 𝑇𝐴𝑘 = 𝑇𝐴 𝑣 = 𝑍𝐴𝜔. Therefore the power carried by
the wave is given in terms of the amplitude of the force wave by

1 𝐵2
𝑃= .
2 𝑍

Example: Again show that when a wave is partially transmitted and reflected, the power is conserved.
However, this time use the force wave.

Solution: The fraction of the incident power that is transmitted is

2
1 𝐵𝐽
𝑃𝐽 2 𝑍2 𝑍1 𝐵𝐽 2 𝑍1 2𝑍2 2 4𝑍1 𝑍2
= 2 = ( ) = ( ) =
𝑃𝐼 1 𝐵𝐼 𝑍2 𝐵𝐼 𝑍2 𝑍1 + 𝑍2 (𝑍1 + 𝑍2 )2
2 𝑍1

as before. The fraction of the incident power that is reflected is

1 𝐵𝑅2
𝑃𝑅 2 𝑍1 𝐵𝑅 2 𝑍2 − 𝑍1 2
= = ( ) = ( ) .
𝑃𝐼 1 𝐵𝐼2 𝐵𝐼 𝑍1 + 𝑍2
2 𝑍1
which is also the same as before. The rest of the problem is identical to the previous one.
 141

12.1.5. Polarisation of a sinusoidal wave

The end of a string can be moved up and down (vertically) with simple harmonic motion to produce a
sinusoidal transverse wave on the string. Alternatively the end of the string can be moved side to side
(horizontally) to also produce a sinusoidal transverse wave on the string. Thus there are two possible
transverse sinusoidal waves that can be made to travel along the string.

Now suppose the end of the string is moved both up and down, and left to right, with simple harmonic
motions simultaneously. We have seen the result of superimposing orthogonal simple harmonic motions
before: it produces a Lissajous figure. If the frequencies of the vertical and horizontal motions are equal, then
we have a 1:1 Lissajous figure. In general this is an ellipse, but special cases include a line or a circle,
depending on the amplitude and phase differences between the vertical and horizontal vibrations. For
𝜋
instance, if the amplitudes are equal and the phase difference is 2 then the Lissajous figure will be a circle.
This corresponds to swinging the end of the string in a circle, which results in a helical displacement along
the string. This helical wave, however, is just made up of two transverse waves at right angles having equal
𝜋
amplitudes and phase difference.
2

The relationship between the two orthogonal transverse sinusoidal waves determines the polarization of the
wave:

𝜋
 If the phase difference is 2 and the amplitudes are equal then the wave is circularly polarised.
 If the phase difference is zero then 1:1 Lissajous figure will be a line and we then say the wave is
linearly polarised.
 In any other case the 1:1 Lissajous figure will be an ellipse and we say the wave is elliptically polarised.

Suppose the string is threaded through a vertical slit. Then if the end of the string is moved up and down, the
resulting wave can move through the slit, but a horizontal linearly polarized wave will not be able to pass
through the slit. The slit therefore acts as a filter which allows only the vertical linearly polarized component
of a wave through.

Now suppose that a linearly polarized wave of amplitude 𝐴 is created on the string, but at an angle 𝜃 from
the vertical. Any displacement 𝑟 of the string will then have vertical component 𝑟 cos 𝜃 and horizontal
component 𝑟 sin 𝜃 . Since only the vertical component will be able to pass through the slit (while the
horizontal component will be blocked) the wave passing through the slit will have vertical linear polarization
and amplitude 𝐴 cos 𝜃. Since the power carried by a wave is proportional to the square of its amplitude, the
fraction of power transmitted is

𝑃 𝜒(𝐴 cos 𝜃)2

= = cos2 𝜃
𝑃0 𝜒𝐴2

where 𝜒 is a proportionality constant. This is called Malus's law. It is applicable when the incoming wave is
already linearly polarized, but a polarizer changes the angle of linear polarization.

But suppose the incident wave is circularly polarised so that the angle 𝜃 is constantly increasing. The average
value of cos2 𝜃 is a half, and so under these circumstances

𝑃 1
= .
𝑃0 2
 142

And suppose the incident wave is unpolarised, meaning the angle 𝜃 is changing randomly. The average value
of cos2 𝜃 is still a half and the same result applies. In both these cases the transmitted wave is linearly
polarised in the direction of the slit and half the power is transmitted.

12.1.6. Interference and Coherence

 

 

 

 is random

Figure 64 Interference of two waves of equal amplitude and frequency, for various fixed phase differences and also
a randomly changing phase difference.

Suppose two sinusoidal waves which are identical, except for their phase constants, are both moving in the
same direction along a string and overlap one another. The resulting wave will be their superposition, namely
𝐴 cos(𝜔𝑡 − 𝑘𝑥 − 𝜑1 ) + 𝐴 cos(𝜔𝑡 − 𝑘𝑥 − 𝜑2 ). This becomes

(𝜔𝑡 − 𝑘𝑥 − 𝜑1 ) + (𝜔𝑡 − 𝑘𝑥 − 𝜑2 ) (𝜔𝑡 − 𝑘𝑥 − 𝜑1 ) − (𝜔𝑡 − 𝑘𝑥 − 𝜑2 )

2𝐴 cos ( ) cos ( )
2 2

𝜑1 +𝜑2 𝜑 −𝜑
which equals 2𝐴 cos (𝜔𝑡 − 𝑘𝑥 − ( 2
)) cos ( 2 2 1 ). Therefore the resultant wave has amplitude

∆𝜑
𝐴′ = 2𝐴 cos ( ) (86)
2
 143

where ∆𝜑 is the phase difference. Figure 64 shows the resulting wave for a variety of possible phase
differences. We have constructive interference (𝐴′ = 2𝐴) for ∆𝜑 = 0 and destructive interference (𝐴′ = 0)
for ∆𝜑 = 180°.

Suppose a person generating a wave on a string becomes tired, and asks someone to take over the job. The
new person would be careful to oscillate the end of the string at the same frequency, but during the switch
from one person to another there would likely be a change in the phase constant. One might consider this
random change in phase constant a defect in a sinusoidal wave, caused by an instability in the source. We
define the coherence time as the time between random changes in the phase constant, and the coherence
length is the wave speed times the coherence time. If the coherence length is very long, we speak of a
coherent wave, whereas if it is short we speak of an incoherent wave. In the above example of interference,
both waves were coherent as their phase constants were indeed constant.

But suppose one (or both) of the waves was incoherent, having a randomly changing phases constant.
Between the times when the phase constant changed, there would be interference which could range from
constructive, to destructive, to anywhere in between. The amplitude during this time would be
∆𝜑
𝐴′ = 2𝐴 cos ( 2 ). As soon as a random phase change occurs (in either wave) the value of ∆𝜑 will change,
and so will the resultant amplitude. It is possible for the waves to be contructively interfering one instant,
and destructively interfering the next. The average amplitude of the resultant wave is the average of |𝐴′| for
any values of ∆𝜑. We have to take the absolute value because mathematically 𝐴′ may be negative, but the
amplitude is always positive. Although ∆𝜑 could be anything, we need only consider values between 0 and
2𝜋 , after which the type interference pattern repeats. Therefore the average amplitude of interfering
incoherent waves is

2𝜋 2𝜋 ∆𝜑
∫0 |𝐴′| 𝑑(∆𝜑) ∫0 |2𝐴 cos ( 2 )| 𝑑(∆𝜑) 𝐴 2𝜋 ∆𝜑
|𝐴′|ave = 2𝜋 = = ∫ |cos ( )| 𝑑(∆𝜑).
∫0 𝑑(∆𝜑) 2𝜋 𝜋 0 2

It is left as a mathematical challenge to show that the value of this integral is 4. The average amplitude is
4𝐴
therefore 𝜋
≈ 1.27𝐴.

Generally, of more importance is the average power carried by the resultant wave. This is given by

2𝜋 2𝜋 ∆𝜑
∫0 𝜒𝐴′2 𝑑(∆𝜑) ∫0 4𝜒𝐴2 cos 2 ( 2 ) 𝑑(∆𝜑) 2𝜒𝐴2 2𝜋 ∆𝜑
′
𝑃ave = 2𝜋 = = ∫ cos2 ( ) 𝑑(∆𝜑).
∫0 𝑑(∆𝜑) 2𝜋 𝜋 0 2

In this case the integral has a value of 𝜋, and so the average power is 2𝜒𝐴2 . Note that since the power of
each of the original waves is 𝜒𝐴2 , this is just the sum of the powers of the original two waves. This may seem
unsurprising, but compare it to the case of interfering coherent waves: if destructive interference occurs the
power of the resultant wave is zero, while if constructive interference occurs the power of the resultant wave
is four times that of one of the original waves!26

26
The powers associated with coherent wave interference may trouble you - don't they violate the principle of conservation of
energy? The answer is of course no, but the reason is certainly not apparent at the moment. It comes down to the fact that one
cannot create a region of constructive interference without simultaneously creating a region of destructive interference as well. This
will be made clearer when we study double slit interference later.
 144

12.2. Standing waves

12.2.1. Standing waves due to superposition

Consider a sinusoidal wave travelling from left to right along a string and approaching its end. The incident
wave can be represented by 𝐴𝐼 cos(𝜔𝑡 − 𝑘𝑥 − 𝜑𝐼 ), while there will also be a reflected wave which can be
represented by 𝐴𝑅 cos(𝜔𝑡 + 𝑘𝑥 − 𝜑𝑅 ). In general the reflected wave will have the same amplitude as the
incident wave.27 The reflected and incident waves overlap but move in opposite directions. If they have the
same amplitude there will be no net energy flow! They have created a so-called standing wave, which can be
represented by

𝑌[𝑥, 𝑡] = 𝐴𝐼 cos(𝜔𝑡 − 𝑘𝑥 − 𝜑𝐼 ) + 𝐴𝑅 cos(𝜔𝑡 + 𝑘𝑥 − 𝜑𝑅 ). (87)

Let us consider a specific example: suppose both ends of a string are fixed. Place a system of axes in such a
way that the left end of the string is at 𝑥 = 0 and the right end at 𝑥 = ℓ (for a string of length ℓ). Then the
following boundary conditions apply:
𝑌[0, 𝑡] = 0, for all t (88)
and
𝑌[ℓ, 𝑡] = 0, for all t. (89)

We first impose the condition of equation (88):

𝑌[0, 𝑡] = 𝐴𝐼 cos(𝜔𝑡 − 𝜑𝐼 ) + 𝐴𝑅 cos(𝜔𝑡 − 𝜑𝑅 ) = 0.

This is satisfied if 𝜑𝐼 = 𝜑𝑅 and 𝐴𝐼 = −𝐴𝑅 . 28 Letting 𝜑 = 𝜑𝐼 = 𝜑𝑅 and 𝐴 = 𝐴𝐼 = −𝐴𝑅 , equation (87)

becomes

𝑌(𝑥, 𝑡) = 𝐴 cos(𝜔𝑡 − 𝑘𝑥 − 𝜑) − 𝐴 cos(𝜔𝑡 + 𝑘𝑥 − 𝜑),

which can be rewritten as

(𝜔𝑡 − 𝑘𝑥 − 𝜑) + (𝜔𝑡 + 𝑘𝑥 − 𝜑) (𝜔𝑡 − 𝑘𝑥 − 𝜑) − (𝜔𝑡 + 𝑘𝑥 − 𝜑)

𝑌[𝑥, 𝑡] = 2𝐴 (sin [ ]) sin ( ).
2 2

This simplifies to

𝑌[𝑥, 𝑡] = 2𝐴 sin(𝜔𝑡 − 𝜑) sin(−𝑘𝑥) = 𝐵 sin(𝑘𝑥) sin(𝜔𝑡 − 𝜑) (90)

where 𝐵 = −2𝐴. Note that the 𝑥 and 𝑡 variables are now in separate terms, and not together in the form
𝑥 ± 𝑣𝑡 . This always happens for standing waves, because they are actually the superposition of two
sinusoidal waves of equal size (amplitude) moving in opposite directions. As a result no energy is transferred
(transported) and the wave is stationary (standing). It has zero overall velocity, and so it does not contain
terms 𝑥 ± 𝑣𝑡, but rather the 𝑥 and 𝑡 variables separately.29 Looking at the expression for the standing wave,

27
This is true if the end is fixed, or loose, or even if it is attached to a mass or fixed via a spring. However, it assumes there are no
resistive forces at the end of the string.
28
This shows that the waves have the same amplitude and the reflected wave is turned upside down, as expected for reflection from
a fixed point. There are other possible ways of satisfying the condition, e.g. 𝐴𝐼 = 𝐴𝑅 and 𝜑𝐼 = 𝜑𝑅 ± 𝜋, and they lead to the same
result.
29
The question then arises: does a standing wave obey the wave equation, which has the general solution 𝑌(𝑥, 𝑡) = 𝑓(𝑥 − 𝑣𝑡) +
𝑔(𝑥 + 𝑣𝑡)? One may be tempted to answer no, since the equation for the standing wave does not consist of functions of 𝑥 − 𝑣𝑡 and
 145

one can interpret it as a simple harmonic motion sin(𝜔𝑡 − 𝜑) at each point with an amplitude 𝐵 sin(𝑘𝑥)
which depends on the position.

𝑗=4

𝑗=3

𝑗=2

𝑗=1

𝑥=0 𝑥=ℓ

Figure 65 Standing waves for a string fixed on both ends.

We now impose the condition of equation (89):

𝑌[ℓ, 𝑡] = 𝐵 sin(𝑘ℓ) sin(𝜔𝑡 − 𝜑) = 0.

If 𝐵 = 0 we would have no wave at all, and the term sin(𝜔𝑡 − 𝜑) oscillates and will not, in general, be zero.
Therefore to satisfy this condition we must have sin(𝑘ℓ) = 0. This is only true if 𝑘ℓ is a multiple of 𝜋 i.e. if 𝑘
takes on one of the possible values

𝑗𝜋
𝑘𝑗 =
ℓ

where 𝑗 is any natural number. 30 The allowed wavenumbers correspond to allowed angular frequencies
𝑗𝜋
according to 𝜔𝑗 = 𝑣𝑘𝑗 = 𝑣 and the possible standing waves are
ℓ

𝑗𝜋 𝑗𝜋𝑣
𝑌𝑗 [𝑥, 𝑡] = 𝐵 sin ( 𝑥) sin ( 𝑡 − 𝜑) . (91)
ℓ ℓ

Each value of 𝑗 is called a mode or an harmonic of the standing wave. The frequency of vibration of the 𝑗 th
mode is
𝜔𝑗 𝑗𝑣 𝑗 𝑇
𝑓𝑗 = = = √ .
2𝜋 2ℓ 2ℓ 𝜇

𝑥 + 𝑣𝑡. But the answer is actually yes, since although the 𝑥 and 𝑡 have become separated, we began with 𝑌(𝑥, 𝑡) consisting of the
sum of two sinusoidal waves. Therefore reversing our analysis, we could express the standing wave in the required form of travelling
waves. Of course there are two travelling waves of equal size and moving in opposite directions, which results in a standing wave.
30
Mathematically one could have 𝑗 = 0, but this makes no sense physically since we require 𝑘𝑗 ≠ 0 for a wave.
 146

Figure 65 shows the first four modes of vibration of a fixed string. The solution with 𝑗 = 1 is called the first
or fundamental mode. Note that there are certain points where the string does not vibrate at all. For the
fundamental mode this happens only at the ends of the string. For the second mode (also called the first
overtone) there is a point exactly in the middle of the string where it does not vibrate. These fixed points are
called nodes. There are also points where the amplitude of vibration is a maximum. These points are called
antinodes. The fundamental mode has only one antinode, while the second mode has two antinodes, etc.
When a string is struck, it initially vibrates with many modes all superimposed on one another. However, the
higher modes usually decay more quickly so that soon the string vibrates only in the fundamental mode. It is
this mode that is mostly responsible for the sound (note) that is heard.

12.2.2. Energy of a standing wave

1 𝜕𝑌 2
For a wave on a string we earlier already derived the kinetic energy is 𝐸𝐾 = 2 ∫ 𝜇 ( 𝜕𝑡 ) 𝑑𝑥 .
Taking 𝑌 = 𝐵 sin(𝑘𝑥) sin(𝜔𝑡 − 𝜑) as in equation (87) for a standing wave, we can substitute in its derivative
to get

1 1
𝐸𝐾 = ∫ 𝜇{𝐵𝜔 sin(𝑘𝑥) cos(𝜔𝑡 − 𝜑)}2 𝑑𝑥 = 𝜇𝐵2 𝜔2 cos2(𝜔𝑡 − 𝜑) ∫ sin2 (𝑘𝑥) 𝑑𝑥.
2 2
1
The average value of sin2 (𝑘𝑥) is , and the integral is over the length of the string, so
2

1 2 2 2 (𝜔𝑡
ℓ 𝜇𝐵2 𝜔2 ℓ
𝐸𝐾 = 𝜇𝐵 𝜔 cos − 𝜑) = cos2(𝜔𝑡 − 𝜑).
2 2 4

1 𝜕𝑌 2
The potential energy is 𝐸𝑃 = ∫ 𝑇 ( ) 𝑑𝑥. This gives
2 𝜕𝑥

1 1
𝐸𝑃 = ∫ 𝑇{𝐵𝑘 cos(𝑘𝑥) sin(𝜔𝑡 − 𝜑)}2 𝑑𝑥 = 𝑇𝐵2 𝑘 2 sin2 (𝜔𝑡 − 𝜑) ∫ cos2 (𝑘𝑥) 𝑑𝑥.
2 2
1
The average value of cos 2 (𝑘𝑥) is also and the integral is over the length of the string, so
2

1 ℓ
𝐸𝑃 = 𝑇𝐵2 𝑘 2 sin2 (𝜔𝑡 − 𝜑) .
2 2

𝜔 2
But 𝑇𝑘 2 = (𝜇𝑣 2 ) ( 𝑣 ) = 𝜇𝜔2 , and so

𝜇𝐵2 𝜔2 ℓ 2
𝐸𝑃 = sin (𝜔𝑡 − 𝜑).
4

The sum of the kinetic and potential energies then give

𝜇𝐵2 𝜔2 ℓ 𝑚string 2 2
𝐸𝐾 + 𝐸𝑃 = {cos2 (𝜔𝑡 − 𝜑) + sin2 (𝜔𝑡 − 𝜑)} = 𝐵 𝜔 .
4 4

The energy of a standing wave is proportional to the square of the amplitude and the square of the frequency.
 147

12.2.3. Driving of a stretched string

Consider a string of length ℓ lying on the X-axis, fixed at 𝑥 = 0 but subjected to a small sinusoidal motion at
its other end (𝑥 = ℓ). Because the string is fixed at the origin, we have already seen that the waves must have
the form of equation (87), namely 𝑌[𝑥, 𝑡] = 𝐵 sin(𝑘𝑥) sin(𝜔𝑡 − 𝜑) . If the driving motion at 𝑥 = ℓ is
𝑌[ℓ, 𝑡] = 𝐷 sin(𝜔𝑡 − 𝜑), then

𝐷
𝐵= .
sin(𝑘ℓ)
𝜔
The value of 𝑘 depends on the driving frequency because 𝑣 = 𝑘
and 𝑣 is constant. Therefore 𝑘 can be
𝑗𝜋
changed by varying the driving frequency. When 𝑘 → ℓ
then sin(𝑘ℓ) → 0 and 𝐵 → ∞, even if the driving
motion amplitude 𝐷 is small. But this corresponds to the allowed wavenumbers (and thus frequencies) for
the formation of standing waves on a string with both sides fixed. Therefore such a system has a strong
response when it is driven at any of its standing wave frequencies. This is a generalization of the idea of
resonance.

12.2.4. General approach for standing waves: separation of variables

We saw than for standing waves, the 𝑥 and 𝑡 variables in the equation representing the wave became
separated. Given any partial differential equation, a standard method to try and solve it is to use the method
of separation of variables. This gives the standing waves. We illustrate this technique with the wave equation
𝜕2 𝑌 𝜕2 𝑌
𝜕𝑡 2
= 𝑣 2 𝜕𝑥 2 . If the variables become separated, then we can write 𝑌[𝑥, 𝑡] = 𝑋[𝑥]𝑇[𝑡] where 𝑋 and 𝑇 are
some new functions to be found. Then
𝜕𝑌 𝜕2𝑌
[𝑥, 𝑡] = 𝑋′[𝑥]𝑇[𝑡] and [𝑥, 𝑡] = 𝑋′′[𝑥]𝑇[𝑡],
𝜕𝑥 𝜕𝑥 2
while
𝜕𝑌 𝜕2𝑌
[𝑥, 𝑡] = 𝑋[𝑥]𝑇′[𝑡] and [𝑥, 𝑡] = 𝑋[𝑥]𝑇′′[𝑡].
𝜕𝑡 𝜕𝑡 2

Substituting these into the wave equation gives 𝑋[𝑥]𝑇′′[𝑡] = 𝑣 2 𝑋′′[𝑥]𝑇[𝑡], and dividing throughout by
𝑋[𝑥]𝑇[𝑡] gives

𝑇′′ 𝑋′′
= 𝑣2 .
𝑇 𝑋

Let us call the quantity these are both equal to −𝜔2, which is known as the separation variable.

 Then from the first term 𝑇 ′′ + 𝜔2 𝑇 = 0. But this is just the equation for simple harmonic motion,
and it has solution 𝑇(𝑡) = 𝐴 𝑇 cos(𝜔𝑡 − 𝜑𝑇 ). Clearly 𝜔 represents the angular frequency, explaining
our rather strange choice for the separation variable.
𝜔 2
 Similarly from the second term 𝑋 ′′ + ( ) 𝑋 = 𝑋 ′′ + 𝑘 2 𝑋 = 0. But this is also like the equation for
𝑣
simple harmonic motion, and it has solution 𝑋(𝑥) = 𝐴𝑋 cos(𝑘𝑥 − 𝜑𝑋 ).

Therefore the standing wave solution to the wave equation is

𝑌[𝑥, 𝑡] = 𝑋[𝑥]𝑇[𝑡] = (𝐴𝑋 cos(𝑘𝑥 − 𝜑𝑋 ))(𝐴 𝑇 cos(𝜔𝑡 − 𝜑𝑇 )) = 𝐵 cos(𝑘𝑥 − 𝜑𝑋 ) cos(𝜔𝑡 − 𝜑𝑇 ).

 148

Although we have used cosine functions, we could just as well have used sine functions, and the form of this
equation is similar to what we had previously.

If we now apply the boundary condition 𝑌[0, 𝑡] = 0, then cos(−𝜑𝑋 ) = 0, which we can satisfy by choosing
𝜋
𝜑𝑋 = 2 . Then

𝜋
𝑌[𝑥, 𝑡] = 𝐵 cos (𝑘𝑥 − ) cos(𝜔𝑡 − 𝜑𝑇 ) = 𝐵 sin(𝑘𝑥) cos(𝜔𝑡 − 𝜑𝑇 ).
2

Applying the second boundary condition 𝑌[ℓ, 𝑡] = 0 means that sin(𝑘ℓ) = 0, which is exactly what we
obtained previously for standing waves and it leads to the allowed wavenumbers and frequencies.

12.3. Wave packets and dispersion

12.3.1. Wave packets

Consider two sinusoidal waves, given by

𝑌1 [𝑥, 𝑡] = 𝐴 cos(𝜔1 𝑡 − 𝑘1 𝑥 − 𝜑),

𝑌2 [𝑥, 𝑡] = 𝐴 cos(𝜔2 𝑡 − 𝑘2 𝑥 − 𝜑).

Suppose the waves are both travelling along the same string. Clearly they are moving in the same direction,
and are identical except for their frequencies and wavenumbers. They must have the same speed (since this
𝑇
depends only on the tension and linear density of the string through 𝑣 = √ ) and so
𝜇

𝜔1 𝜔2
𝑣= = . (92)
𝑘1 𝑘2

Recall that when we added two vibrations of different frequencies we got beats. We will now show that
adding these waves gives rise to wave packets. The resultant wave is 𝑌[𝑥, 𝑡] = 𝐴 cos(𝜔1 𝑡 − 𝑘1 𝑥 − 𝜑) +
𝐴 cos(𝜔2 𝑡 − 𝑘2 𝑥 − 𝜑), so

(𝜔1 𝑡 − 𝑘1 𝑥 − 𝜑) + (𝜔2 𝑡 − 𝑘2 𝑥 − 𝜑) (𝜔1 𝑡 − 𝑘1 𝑥 − 𝜑) − (𝜔2 𝑡 − 𝑘2 𝑥 − 𝜑)

𝑌[𝑥, 𝑡] = 2𝐴 cos ( ) cos ( ).
2 2

𝜔1 +𝜔2 𝑘1 +𝑘2 𝜔1 −𝜔2 𝑘1 −𝑘2

This simplifies to 𝑌[𝑥, 𝑡] = 2𝐴 cos ( 2
𝑡 − 2
𝑥 − 𝜑) cos ( 2
𝑡 − 2
𝑥). If one defines

𝜔1 + 𝜔2 𝑘1 + 𝑘2 𝜔1 − 𝜔2 𝑘1 − 𝑘2
𝜔𝑐𝑎𝑟 = , 𝑘𝑐𝑎𝑟 = , 𝜔𝑒𝑛𝑣 = and 𝑘𝑒𝑛𝑣 =
2 2 2 2

then 𝑌[𝑥, 𝑡] = 2𝐴 cos(𝜔𝑐𝑎𝑟 𝑡 − 𝑘𝑐𝑎𝑟 𝑥 − 𝜑) cos(𝜔𝑒𝑛𝑣 𝑡 − 𝑘𝑒𝑛𝑣 𝑥). The envelope values (subscript env) are
smaller than the corresponding carrier (average) values. The first cosine term represents a wave with the
average period and wavelength of the superimposed waves. It is modulated by the second cosine term
representing a wave with smaller frequency (longer period) and smaller wavenumber (longer wavelength).
This second term therefore divides the average wave (first term) into wave packets.
 149

Both the carrier wave and the envelope wave have the same speed as the original waves:

𝜔1 + 𝜔2
𝜔𝑐𝑎𝑟 2 𝑣𝑘1 + 𝑣𝑘2
𝑣𝑐𝑎𝑟 = = = = 𝑣,
𝑘𝑐𝑎𝑟 𝑘1 + 𝑘2 𝑘1 + 𝑘2
2
𝜔1 − 𝜔2
𝜔𝑒𝑛𝑣 2 𝑣𝑘1 − 𝑣𝑘2
𝑣𝑒𝑛𝑣 = = = = 𝑣,
𝑘𝑒𝑛𝑣 𝑘1 − 𝑘2 𝑘1 − 𝑘2
2

and the wave packets move along at the same speed as the original waves.

12.3.2. Wave pulses

If one superimposes not just two, but many waves of different frequencies/wavelengths and amplitudes
together, one can build up wave packets of more complicated forms. A wave pulse of arbitrary shape can be
created by the superposition of an infinite number of sinusoidal waves having a continuous spectrum of
frequencies/wavenumbers and amplitudes. The mathematics behind this is called Fourier transforms, and is
a generalization of discrete Fourier analysis we have dealt with before. Thus waves in general can be
considered to consist of superimposed sinusoidal waves.

The more detailed the pulse one is trying to create, the more waves of different frequencies/wavelengths
and amplitudes one has to add together. For instance, to make a Gaussian shaped wave pulse of time width
1
∆𝑡 requires one to add all sinusoidal waves within a frequency range of ∆𝑓 ≈ ∆𝑡. This is a general principle:
the smaller a pulse (in time) the more frequencies one requires, and the smaller the pulse (in space or length)
the more wavelengths one requires. In quantum mechanics (also called wave mechanics) this leads to the
famous “Heisenberg Uncertainty Principle”.

12.3.3. Dispersion

𝑇
For a wave on a string we found the speed was given by 𝑣 = √𝜇. Thus the speed of the wave depends only
on the properties of the string, and not on any properties of the wave itself. However, for some other types
of waves their speed can also be influenced by their wavelength (and therefore the wavenumber 𝑘), with
some wavelengths travelling faster and others slower. For such waves the wave equation is

𝜕2𝑌 2
𝜕2𝑌
= (𝑣[𝑘]) ,
𝜕𝑡 2 𝜕𝑥 2

where the dependence of the speed on the wavenumber has been made clear.

But a wave in general (a pulse) can be regarded as being made up of many sinusoidal waves of different
wavelengths. If each wavelength has a different speed, the speed cannot be regarded as a constant and our
solution to the wave equation, namely 𝑌[𝑥, 𝑡] = 𝑓[𝑥 − 𝑣𝑡] + 𝑔[𝑥 + 𝑣𝑡], is no longer valid. Rather, because
each wavelength moves with a different speed, the wave changes its shape as it propagates. Such waves are
 150

said to undergo dispersion.31 Most often the shorter wavelengths (i.e. higher wavenumbers) travel more
slowly. This is called ‘normal dispersion’. However, in some systems the opposite is true and this is called
‘anomalous dispersion’.

But if only a single wavenumber is involved (meaning we are dealing with a sinusoidal wave), then only one
speed is involved and our solution to the wave equation is still valid. Therefore sinusoidal waves (unlike waves
in general) are “immune” to dispersion and propagate without changing form even in a dispersive medium.
𝜔
But a sinusoidal wave, in order to obey the wave equation, must still have 𝑣 = 𝑘 . When the speed depends
on the wavenumber, then the angular frequency is no longer proportional to the wavenumber: the
relationship between the angular frequency and the wavenumber, i.e. the function 𝜔(𝑘), is called the
dispersion relationship. Of course

𝜔[𝑘] = 𝑘 𝑣[𝑘], (93)

and for a non-dispersive wave this is simply 𝜔[𝑘] = 𝑘 𝑣 with 𝑣 constant and not a function of 𝑘.

Consider again our superposition of two sinusoidal waves having different frequencies and wavenumbers,
𝜔1 𝜔2
but now suppose the two waves also have slightly different speeds, namely 𝑣1 = and 𝑣2 = . One can
𝑘1 𝑘2
still add the waves as before to get the wave packets. However, now the speeds of the carrier and envelope
require more careful analysis:

𝜔1 +𝜔2 𝑣1 𝑘1 +𝑣2 𝑘2
 First consider the carrier wave, which has speed 𝑣𝑐𝑎𝑟 = 𝑘1 +𝑘2
= 𝑘1 +𝑘2
. This speed lies somewhere
𝜔1 𝜔2
between 𝑣1 = 𝑘1
and 𝑣2 = 𝑘2
, and so the carrier wave speed lies between the speeds of the two original
sinusoidal waves.
𝜔1 −𝜔2 ∆𝜔
 Now consider the envelope wave, which has speed 𝑣𝑒𝑛𝑣 = 𝑘1 −𝑘2
= ∆𝑘
. This speed can be less than, or
greater than, the speeds of the original sinusoidal waves that were added. The wave packets as a whole
move with this speed, which is therefore physically more meaningful than the carrier speed.

For a wave in general, which is the superposition of waves with a continuous distribution of angular
∆𝜔
frequencies and wavenumbers, the envelope velocity 𝑣𝑒𝑛𝑣 = ∆𝑘
must be replaced with the so-called ‘group
velocity’, namely

𝑑𝜔
𝑣𝑔 = .
𝑑𝑘

This group velocity represents the speed with which the wave as a whole (and therefore energy or
𝜔
information) is propagated. In a dispersive medium it is a much more important quantity than 𝑣 = 𝑘
, which
is then referred to as the ‘phase velocity’. By taking the derivative of equation (93) one gets

31
Newton is famous for having used a prism to disperse the different wavelengths of white light, creating a rainbow. This effect is
possible because the refractive index of glass is different for different colours of light i.e. the speed of the light through the glass is a
function of the wavelength. One does not have to refract the light in order to separate the colours - if the light is sent down a long
optical fibre certain wavelengths travel faster and the colours become separated. This places a limitation on the speed of optical data
communication networks and the internet!
 151

𝑑𝑣
𝑣𝑔 = 𝑣 + 𝑘 .
𝑑𝑘
𝑑𝑣
 If 𝑑𝑘 < 0 then the group velocity will be less than the phase velocity, and this corresponds to normal
dispersion.
𝑑𝑣
 If 𝑑𝑘
> 0 the group velocity is greater than the phase velocity, and this corresponds to anomalous
dispersion.
𝑑𝑣
 If 𝑑𝑘 = 0 then the group velocity is equal to the phase velocity. This corresponds to no dispersion, with
𝑣 being independent of 𝑘.

12.4. Problems

1. Consider the travelling wave in one dimension given by 𝑌[𝑥, 𝑡] = sin(2𝜋{4𝑡 − 0.2𝑥}).
(a) What are its period, frequency, angular frequency, wave number, wavelength and speed?
𝑇
(b) Plot on the same graph, using a computer, the wave at 𝑡 = 0, 𝑡 = 10 and 𝑡 = 𝑇4. Use 𝑥-values from 0
to 15 with steps of 0.1.

2. (a) The (typical) human ear can perceive sounds from 30 Hz to 15 kHz. Given that the speed of sound in
air is about 340 m.s-1, calculate the shortest wavelength of sound wave that is audible. (b) The speed of
electromagnetic waves is vacuum is 3x108 m.s-1. Calculate the wavelength of a VHF [very high frequency]
TV signal of 100 MHz, and the frequency of visible blue-green light with wavelength 500 nm.

3. For a wave on a string under constant tension, show that the transmission and reflection amplitude ratios
2𝜆2 𝜆 −𝜆
of the displacement wave are given, in terms of the wavelengths, by 𝑡 = 𝜆 and 𝑟 = 𝜆2 +𝜆1
1 +𝜆2 1 2
respectively.

4. Consider a sinusoidal wave 𝑌𝐼 [𝑥, 𝑡] = 𝐴̃𝐼 𝑒 𝑖(𝜔𝑡−𝑘𝑥) travelling along a string. If a mass 𝑚 is attached to the
end of the string, assume that the reflected wave is also sinusoidal with amplitude 𝐴̃𝑅 and show that
𝐴̃𝑅 𝑖+𝑝 𝑚𝜔2 𝑖+𝑝 1
𝐴̃𝐼
= 𝑖−𝑝 where 𝑝 = 𝑇𝑘
. [Note: The ratio 𝑖−𝑝 = 𝑒 𝑖𝜃 where 𝜃 = 2 arctan 𝑝 − 𝜋 and it therefore has unit
magnitude, which means the reflected wave is equal in size to the incident wave but differs in phase.]
Hint: take the end of the string to be at 𝑥 = 0, so that the end condition is (by Newton's second law)
𝜕𝑌tot 𝜕 2 𝑌tot
−𝑇 [0, 𝑡] = 𝑚 [0, 𝑡]
𝜕𝑥 𝜕𝑡 2
where 𝑌tot is the sum of the incident and reflected wave.

5. Consider a sinusoidal wave 𝑌𝐼 [𝑥, 𝑡] = 𝐴̃𝐼 𝑒 𝑖(𝜔𝑡−𝑘𝑥) travelling along a string. If a mass 𝑚 is attached to the
string, which continues, assume that the transmitted and reflected waves are also sinusoidal with
amplitudes 𝐴̃𝐽 and 𝐴̃𝑅 respectively, and show that
 152

𝐴̃𝑅 𝑝 𝐴̃𝐽 𝑖
= and =
̃
𝐴𝐼 𝑖 − 𝑝 ̃
𝐴𝐼 𝑖 − 𝑝
𝑚𝜔2
where 𝑝 = 2𝑇𝑘
. Find the magnitude of these ratios and show that the sizes of the reflected and
𝑝 1
transmitted waves are the fractions and of the incident wave respectively. Hint: take the
√1+𝑝2 √1+𝑝2
point of the string where the mass is to be at 𝑥 = 0, so that the condition at the mass is (by Newton's
second law)
𝜕𝑌left 𝜕𝑌𝐽 𝜕 2 𝑌𝐽
−𝑇 [0, 𝑡] + 𝑇 [0, 𝑡] = 𝑚 2 [0, 𝑡]
𝜕𝑥 𝜕𝑥 𝜕𝑡
where 𝑌left is the sum of the incident and reflected waves.

6. Consider a sinusoidal wave 𝑌𝐼 [𝑥, 𝑡] = 𝐴̃𝐼 𝑒 𝑖(𝜔𝑡−𝑘𝑥) travelling along a string. At a point it encounters
resistance −𝛽𝑣 proportional to the transverse velocity of the wave. Assume it continues and that there
are also sinusoidal transmitted and reflected waves with amplitudes 𝐴̃𝐽 and 𝐴̃𝑅 respectively. Show that
𝐴̃𝑅 𝑝 𝐴̃𝐽 1
= and =
𝐴̃𝐼 1 − 𝑝 𝐴̃𝐼 1 − 𝑝
𝛽
where 𝑝 = 2𝑍. [Note: this means that when a wave encounters a resistance it not only loses energy due
to the resistance, but also generates a reflected wave so that not all the remaining energy is transmitted!]
Hint: take the point of the string where the resistance is to be at 𝑥 = 0, so that the condition at the mass
is (by Newton's second law)
𝜕𝑌left 𝜕𝑌𝐽 𝜕𝑌𝐽
−𝑇 [0, 𝑡] + 𝑇 [0, 𝑡] − 𝛽 [0, 𝑡] = 0
𝜕𝑥 𝜕𝑥 𝜕𝑡
where 𝑌left is the sum of the incident and reflected waves.

7. Consider the superposition of two coherent waves which only differ in their phase constant. Determine
the value of the resultant wave amplitude when Δ𝜑 = 90°, as well as the required phase difference so
that the resultant wave has exactly the same amplitude as the two original waves.

8. A 1.2 m length of wire is made of steel (density 7.5 g.cm-3) and has a diameter of 0.7 mm. (a) Calculate
its linear density (mass per unit length) 𝜇. (b) Calculate the tension it must be placed under if, when fixed
at both ends, the third harmonic (𝑗 = 3) of the standing waves has a frequency of 220 Hz. (c) Calculate
the wavelength of this wave.
 153

13. Examples of Waves

Studying one dimensional transverse waves on a stretched string introduces the important concepts
regarding waves. But in general waves can occur in two or three dimensions include longitudinal and torsional
waves.

A bulk solid is a three dimensional medium that supports all three types of waves, whereas only longitudinal
waves can occur in liquids and gases. In multi-dimensional media a new property of waves occurs that we
have not yet considered, namely refraction or bending when they pass non-perpendicularly through an
interface. Liquids and solids can also carry two dimensional surface waves e.g. water waves on a pond. These
waves are generally complicated and we will not discuss them in detail. All these waves are mechanical
waves, involving the displacement of matter.

Electrical waves on a transmission line are simple one dimensional waves. They involve the movement of
charge along an electrical conductor, instead of ordinary matter. Electromagnetic waves are three
dimensional transverse waves that can propagate without a medium in empty space (but also in a medium).
There is not physical displacement of any material, but rather changes in electric and magnetic fields. For an
electromagnetic wave in vacuum these fields are perpendicular to the direction of propagation and so the
waves are transverse and can display polarization effects. Although electromagnetic waves obey the wave
equation, they are more abstract than ordinary mechanical waves. Einstein’s general theory of relativity also
predicts gravitational waves as disturbances in the curvature of space-time: direct observation of these waves
is currently the goal of many physicists who are performing complex experiments in this regard.

Finally another important example of abstract waves must be mentioned, namely the matter waves of
quantum mechanics (wave mechanics). For instance, an electron or other particle may be represented by a
wave. These waves obey the so-called Schrodinger wave equation rather than the usual wave equation.
Nevertheless, familiarity with waves will help when studying the theory of quantum mechanics. Even more
advanced theories, such as string theory, make use of concepts from wave theory.

13.1. Waves in a bulk solid

When a solid is elastically deformed, there are forces that act to restore it to its original shape.32 Therefore
elastic waves occur in the solid. For simplicity, we shall assume the solid is a continuous body, although in
reality solids are not continuous but consist of atoms.33 We shall also assume the solid is isotropic i.e. its
elastic properties do not depend on direction. Because most solids are either amorphous or consist of

32
If a solid is deformed too much, it does not return to its original shape and is said to be plastically deformed.
33
This does have an influence on the waves and is important, for example, in the theory of the heat capacity of a solid, which you
will learn about in solid state physics.
 154

randomly oriented crystal grains, this is usually justified. For bulk material the effect of the surfaces is
ignored. This is applicable for wave moving inside the material far from the surfaces.

13.1.1. Transverse waves in a bulk solid

Consider a small piece (element) of the bulk material which, as one moves along the 𝑥 direction, has been
sheared perpendicularly from its original shape (Figure 66).

𝑑𝑌

𝑑𝑥

Figure 66 Element of material deformed by shear strain.

𝐹 𝑑𝑌
The shear modulus 𝐺 is defined by the ratio of the shear stress 𝐴 to the shear strain 𝑑𝑥, which means the
𝜕𝑌 𝜕𝑌
force is given by 𝐹 = 𝐴𝐺 𝜕𝑥. [This is similar to the expression 𝐹𝑦 = 𝑇 𝜕𝑥 for the force at some point on a piece
of string, but with the tension 𝑇 replaced by 𝐴𝐺. Therefore the derivation for the wave equation will be
𝜕𝐹 𝜕2 𝑌
similar.] The force changes with position at a rate 𝜕𝑥 = 𝐴𝐺 𝜕𝑥 2 , so that the change in (i.e. net or resultant)
𝜕2 𝑌
force on a segment from 𝑥 to 𝑥 + 𝑑𝑥 is given by 𝑑𝐹 = 𝐴𝐺 𝜕𝑥 2 𝑑𝑥.

𝜕2 𝑌
Using Newton's second law, we can equate this to the mass (𝜌𝐴𝑑𝑥) times the acceleration 𝜕𝑡 2 , which gives

𝜕2𝑌 𝜕2𝑌
𝐴𝐺 𝑑𝑥 = 𝜌𝐴𝑑𝑥 .
𝜕𝑥 2 𝜕𝑡 2

𝜕2 𝑌 𝜌 𝜕2 𝑌
This simplifies to the wave equation 𝜕𝑥 2 = 𝐺 𝜕𝑡 2 with wave speed

𝑣 = √𝐺/𝜌. (94)

Recall that the impedance of waves on a string is given by 𝑍 = 𝑇/𝑣. The impedance of transverse waves in a
𝐴𝐺
bulk material is given similarly by 𝑍 = . If the areas on both sides of the interface match, then one can
𝑣
𝑍 𝐺
rather use the simpler specific impedance given by 𝑍 ′ = 𝐴 = 𝑣 . The reader should be able to show, using the
wave speed, that equivalent expressions are 𝑍′ = √𝐺𝜌 = 𝜌𝑣.
 155

13.1.2. Longitudinal waves in a bulk solid

Consider a small piece (element) of the bulk material which, as one moves along the 𝑥 direction, has been
stretched parallel to its original shape. This is shown in Figure 67(a) and represents purely longitudinal
motion.

Figure 67(b) shows the case where the element has been stretched, but also becomes thinner. In that case
the motion of the medium is not purely longitudinal, but also has a transverse component. The ratio of the
transverse to longitudinal strain is called Poisson’s ratio ( 𝜈 ) and characterizes how much the element
becomes thinner as it is stretched.

Although Figure 67(a) is appropriate for longitudinal waves in a bulk solid, we must consider in Figure 67(b)
𝐹2
first for the definition of Young’s modulus. Young’s modulus 𝐸 is defined by the ratio of the tensile stress 𝐴
𝑑𝜂
to the longitudinal strain 𝑑𝑥 (allowing for the element to become thinner), which means the force is given by
𝜕𝜂
𝐹2 = 𝐴𝐸 𝜕𝑥.

(a) 𝐴 𝐹1

𝑑𝑥 𝑑𝜂

(b) 𝐴 𝐹2

𝑑𝜂
𝑑𝑥

Figure 67 Element of material deformed by tensile strain.

In Figure 67(a) the force 𝐹1 is greater than 𝐹2 because of the additional force required to stretch out the
element without it becoming thinner. This case of stretching without lateral contraction is an exercise in
elasticity theory, solved in Feynman’s Lectures in Physics (http://feynmanlectures.caltech.edu/II_38.html),
and the result is

1−𝜈 1−𝜈 𝜕𝜂 𝜕𝜂
𝐹1 = 𝐹2 = 𝐴 ( 𝐸) = 𝐴𝐸′
(1 + 𝜈)(1 − 2𝜈) (1 + 𝜈)(1 − 2𝜈) 𝜕𝑥 𝜕𝑥
 156

1−𝜈 𝜕𝐹 𝜕2 𝜂
where 𝐸 ′ = (1+𝜈)(1−2𝜈) 𝐸. The force changes with position at a rate = 𝐴𝐸′ , so that the change in (i.e.
𝜕𝑥 𝜕𝑥 2
2
𝜕 𝜂
net or resultant) force on a segment from 𝑥 to 𝑥 + 𝑑𝑥 is given by 𝑑𝐹 = 𝐴𝐸′ 2 𝑑𝑥.
𝜕𝑥

𝜕2 𝜂
Using Newton's second law, we can equate this to the mass (𝜌𝐴𝑑𝑥) times the acceleration 𝜕𝑡 2 , which gives

𝜕2𝜂 𝜕2𝜂
𝐴𝐸′ 𝑑𝑥 = 𝜌𝐴𝑑𝑥 .
𝜕𝑥 2 𝜕𝑡 2

𝜕2 𝜂 𝜌 𝜕2 𝜂
This simplifies to the wave equation 𝜕𝑥 2 = 𝐸′ 𝜕𝑡 2 with wave speed

𝐸′ 1−𝜈 𝐸
𝑣=√ =√ . (95)
𝜌 (1 + 𝜈)(1 − 2𝜈) 𝜌

1−𝜈
𝐴𝐸′ 𝐴((1+𝜈)(1−2𝜈)𝐸)
The impedance is given by 𝑍 = 𝑣
= 𝑣
.

13.2. Waves in a beam

13.2.1. Transverse waves in a beam

In a beam the surface plays an important role and cannot be ignored. Shearing forces applied to a beam
causes it to bend, leading to regions of compression and dilation. The resulting flexural waves do not obey
the standard wave equation, but instead must satisfy

𝜕4𝑌 𝜕2𝑌
−𝐸𝐼𝑏 = 𝜌𝐴
𝜕𝑥 4 𝜕𝑡 2

which contains a fourth order partial derivative. Calculating the static deflection of beams is an important
topic in mechanical engineering and we have used some of the results when approximating the vibrations of
beams modelled as springs in section 5.7. The standing wave patterns and frequencies can be obtained more
accurately by solving this equation. Such waves are dispersive (i.e. the wave speed depends on the
wavelength) and a pulse would not maintain its shape as it travels.

13.2.2. Longitudinal waves in a thin beam

The elongation of a beam when the surface is unconstrained was shown in Figure 67(b) and the force is given
𝜕𝜂
by 𝐹2 = 𝐴𝐸 𝜕𝑥, containing Young’s constant. The derivation of the wave equation then follows exactly the
same steps as for longitudinal waves in a bulk solid, but with 𝐸 instead of 𝐸′, and results in the wave equation
𝜕2 𝜂 𝜌 𝜕2 𝜂
= with wave speed
𝜕𝑥 2 𝐸 𝜕𝑡 2

𝑣 = √𝐸/𝜌. (96)
 157

For solids 𝐸 is in the order of ~1010 Pa and 𝜌 in the order of ~1000 kg/m3, so the velocity of these waves is in
𝐸
the order of ~3 km/s. The impedance of such waves is given by 𝑍 = 𝐴 and different ways to write the
𝑣
𝐸
specific impedance are 𝑍 ′ = 𝑣
= √𝐸𝜌 = 𝜌𝑣.

13.2.3. Longitudinal waves in a thick beam

The previous section did not consider the fact that the longitudinal waves on a thin beam are actually only
‘quasi-longitudinal’ and also have a transverse component as a result of the effect of Poisson’s ratio, which
causes the beam to become thicker in the compressed regions (see Figure 68).

Figure 68 Quasi-longitudinal wave in a thick rod causing transverse thickness variations.

For a thick beam, when this motion of the surface cannot be neglected, a complicated analysis34 results in
the Rayleigh-Love wave equation

𝜕2𝜂 𝜌 𝜕2𝜂 𝐼 𝜌 𝜕4𝜂

2 𝑝
= −𝜈 ,
𝜕𝑥 2 𝐸 𝜕𝑡 2 𝐴 𝐸 𝜕𝑡 2 𝜕𝑥 2

𝜋𝑅4
where the extra term on the right contains Poisson’s ratio 𝜈 and the polar moment 𝐼𝑝 = 2
divided by the
2
cross-sectional area 𝐴 = 𝜋𝑅 , assuming a circular cross-section. This is not the standard wave equation and
it can be shown that the wave speed is a function of the wavelength and varies according to

𝐸
𝑣=
√ 𝜋𝜈𝑅 2
𝜌 (1 + 2 ( ) )
𝜆

i.e. dispersion occurs. Note that if Poisson’s ratio is zero or the radius of the rod is much smaller than the
wavelength of the wave, this result simplifies to the result for a thin rod. We must consider any rod “thick” if
a very short wavelength (high frequency) wave passes through it, and even a thick rod can be considered
“thin” if the wavelength of the wave is long enough.35

Example: A metal rod carries a longitudinal wave having speed 3 km/s. If the wave has a frequency of 1 kHz,
estimate how large the radius of the rod can be made for it still to be considered “thin”.

Solution: Using 𝑣 = 𝑓𝜆, we obtain a wavelength of 3 m. Therefore, as long as the radius is much less than
this, the rod may be considered “thin”.

34
See ‘Wave Motion in Elastic Solids’ by Graff (Dover, 1991), section 2.5.3 (p. 116).
35
During a tensile test the rod is elongated very slowly and the wavelength can therefore be considered infinite. Hence any rod, no
matter how thick it is, must always be considered “thin” and one always measures Young's modulus.
 158

13.2.4. Torsional waves in a beam

Suppose the particles of a beam at position 𝑥 have been displaced by an angle 𝜃 from their normal position.
𝜕𝜃
Then the restoring torque is given by 𝜏 = 𝐽𝐺 where 𝐺 is the shear modulus and 𝐽 is the torsion constant.
𝜕𝑥
𝜕𝜏 𝜕2 𝜃
This torque is changing at a rate 𝜕𝑥 = 𝐽𝐺 𝜕𝑥 2 , so that the change in (i.e. net or resultant) torque on a segment
𝜕2 𝜃
from 𝑥 to 𝑥 + 𝑑𝑥 is given by 𝑑𝜏 = 𝐽𝐺 𝜕𝑥 2 𝑑𝑥. Using Newton's second law in rotational form, we equate this
to the moment of inertia times the angular acceleration. The moment of inertia for length 𝑑𝑥 of a beam
around its rotation axis is related to the polar moment by 𝐼 = ∫ 𝑟 2 𝑑𝑚 = ∫ 𝑟 2 (𝜌 𝑑𝑥 𝑑𝑎) = 𝜌 𝑑𝑥 ∫ 𝑟 2 𝑑𝑎 =
𝜌𝑑𝑥𝐼𝑝 , so

𝜕2𝜃 𝜕2𝜃
𝐽𝐺 𝑑𝑥 = 𝐼𝑝 𝜌𝑑𝑥 ,
𝜕𝑥 2 𝜕𝑡 2

𝜕2 𝜃 𝜌 𝐼𝑝 𝜕2 𝜃 𝐺𝐽
or 𝜕𝑥 2 = 𝐺 𝐽 𝜕𝑡 2
. This is the wave equation with 𝑣 = √𝜌𝐼 . For a circular rod 𝐽 = 𝐼𝑝 and this reduces to
𝑝

𝑣 = √𝐺/𝜌.

Hence these waves travel at the same speed as transverse waves in bulk material – considering that torsional
waves are a form of transverse waves, this is not merely a coincidence. Out-of-plane warping during the
twisting of non-circular beams generally causes the torsion constant 𝐽 to be less than the polar moment 𝐼𝑝 ,
and hence the speed of torsional waves in such beams will be slower. The impedance of torsional waves is
𝐺
given by 𝑍 = 𝐽 𝑣 = 𝐽√𝐺𝜌 = 𝐽𝜌𝑣.

13.2.5. Comparison of wave speeds in a solid

𝐸
Since for isotropic materials the shear modulus 𝐺 is related to Young's modulus 𝐸 through 𝐺 = 2(1+𝜈) where
𝜈 is Poisson’s ratio, we can compare the speeds of some waves encountered so far.

Transverse wave in bulk, and

Type Longitudinal wave in thin rod Longitudinal wave in bulk
torsional wave in circular rod
𝐺 𝐸 𝐸 (1 − 𝜈)𝐸
Speed √ =√ √ √
𝜌 2(1 + 𝜈)𝜌 𝜌 (1 + 𝜈)(1 − 2𝜈)𝜌

𝑬 1 (1 − 𝜈)
Speed relative to √ √ 1 √
𝝆 2(1 + 𝜈) (1 + 𝜈)(1 − 2𝜈)

𝟏 3 3
Relative speed if 𝝂 = √ ≈ 61% 100% √ ≈ 122%
𝟑
8 2

Note that the relative speed of a longitudinal wave is about double that of a transverse wave in bulk material.

Example: After an earthquake, the longitudinal waves (called the primary or P-waves) arrive first. The
transverse waves (also called the secondary or S-waves) travel slower and arrive a bit later. These are
generally more destructive because of their transverse nature. Note that there are also various surface waves
which travel even slower, arriving after the P- and S-waves. In practice the waves from an earthquake can be
very complicated and provide a lot of data for seismologists.
 159

13.3. Waves in a fluid

Unlike for rigid solids, when a fluid (liquid or gas) is deformed there are not always forces that act to restore
it to its original state. Three forces can however be considered, namely those that result from changes in
pressure, from gravity and from surface tension.

13.3.1. Pressure wave in a fluid

Although a fluid will not keep its shape, it will keep its volume. If its volume is changed there will be a restoring
force to try and return it to the original volume. The restoring force is due to pressure and, since pressure
cannot act transversely, the type of wave that is produced is a longitudinal wave.

The pressure waves in a fluid are very similar to the longitudinal waves in a thin solid rod. Consider an
imaginary cylinder of the fluid, where the particles in the fluid at position 𝑥 have been displaced by an
amount 𝜂(𝑥) from their normal position along (i.e. parallel to) the cylinder. The bulk modulus of a fluid is
𝜕𝑃
defined by the ratio of applied pressure to the fractional change in volume it produces i.e. 𝐾 = 𝜕𝑉, where
−( )
𝑉
the minus sign is necessary since a positive pressure will reduce the volume and we want 𝐾 to be positive.
Still considering an imaginary cylinder of fluid (with cross-section 𝐴 and deformation by an amount 𝜂 at
𝐹/𝐴 𝜕𝜂
position 𝑥) and ignoring the minus sign, this can be written as 𝐾 = 𝜕(𝐴𝜂) , or 𝐹 = 𝐴𝐾 . This is exactly like the
𝜕𝑥
𝐴𝑥
expression for the force at some point in bulk solid experiencing longitudinal waves, but with 𝐸′ replaced by
the bulk modulus 𝐾. We therefore skip the rest of the derivation and just give the result, namely

𝜕2𝜂 𝜌 𝜕2𝜂
=
𝜕𝑥 2 𝐾 𝜕𝑡 2

which is the wave equation with speed

𝑣 = √𝐾/𝜌. (97)

𝐾
The impedance of these waves is given by 𝑍 = 𝐴 , but generally the areas match and it is easier to use the
𝑣
𝐾
specific impedance given by 𝑍 ′ = = √𝐾𝜌 = 𝜌𝑣. This is also often called the specific acoustic impedance,
𝑣
because sound is an example of such a pressure wave in the atmosphere. Just as the wave on a stretched
string is associated with a force wave that travels along with it, similarly a sound wave is associated with a
pressure wave that travels along with it. This is not an absolute pressure, but the pressure change relative to
𝐹 𝜕𝜂
that in the fluid (called the sound pressure), given by the restoring force per unit area 𝑝 = 𝐴 = 𝐾 𝜕𝑥. This
acoustical pressure wave is transmitted and reflected similarly to the force wave in a string.

13.3.2. Speed of sound waves in a gas

The bulk modulus of different liquids can be looked up in tables. However, from the definition of the bulk
𝜕𝑃
modulus (𝐾 = 𝜕𝑉) we can use gas laws to determine the bulk modulus of a gas and hence the speed of
−( )
𝑉
sound (pressure) waves in a gas.
 160

 One might initially consider applying the ideal gas law 𝑃𝑉 = 𝑛𝑅𝑇 (where 𝑛 is the number of moles of gas
𝜕𝑃 −𝑛𝑅𝑇
and 𝑅 = 8.31 J/(mol.K) is the universal gas constant). Then 𝐾 = −𝑉 = −𝑉 ( ) = 𝑃 and one finds
𝜕𝑉 𝑉2
that the bulk modulus of a gas is the same as its pressure.

This is a good approximation, but the universal gas law holds for systems in equilibrium, while a wave passing
through a medium is a dynamic process. In regions of compression the gas may get warmer and in regions of
rarefaction it gets cooler, so the process is not isothermal. We consider another gas law, for adiabatic
situations i.e. where no heat flow occurs. This is more suitable for applying to sound waves in a gas, since the
frequency is generally high enough that there is not sufficient time available for any significant amount of
heat to flow.

 For an adiabatic process, in which no heat flows into or out of a system during the process, the gas law
is that 𝑃𝑉 𝛾 = 𝐶 where 𝐶 is a constant and 𝛾 is the ratio of the heat capacities of the gas at constant
pressure and at constant volume. For diatomic gas molecules (e.g. N2 and O2 which form most of the
𝜕𝑃
atmosphere) 𝛾 = 75 . Then 𝐾 = −𝑉 𝜕𝑉 = −𝑉(−𝛾𝐶𝑉 −𝛾−1 ) = 𝛾𝑃, which is larger by a factor 𝛾 from the
initial calculation.

𝛾𝑃
Therefore the speed of sound (longitudinal waves in a gas) is given by 𝑣 = √ 𝜌 . Although we could not use
𝑃
the universal gas law to determine the speed of sound waves accurately, it can be used to evaluate in the
𝜌
formula (since this has an equilibrium value). One then finds

𝑛𝑅𝑇
𝛾𝑃 𝛾( 𝑉 ) 𝑅𝑇
𝑣=√ =√ = √𝛾 .
𝜌 𝑛𝑀 𝑀
𝑉

This means that for a certain type of gas, the speed of sound varies only on its temperature (and not, for
instance, with its pressure if the temperature is unchanged!), and increases with the square root of the
absolute temperature.

Example: Calculate the speed of sound in the air at 15˚C.

Solution: The air of the atmosphere consists of about 80% N2 and 20% O2, with molar masses of 28 and 32
g/mol respectively. The weighted average molar mass is 0.0288 kg/mol. These gases both consist of diatomic
(8.31)(288)
molecules and so 𝛾 = 75 = 1.4. Also 15˚C = 273+15 K = 288 K. Therefore 𝑣 = √1.4 0.0288
which gives 341
m/s.

13.3.3. Water waves

Although sound waves definitely occur in water due to its bulk modulus (allowing the whales to sing and
dolphins to navigate with SONAR), other restoring forces can also occur in water giving rise to other types of
waves. These forces are gravity and the surface tension. Most of the waves that result from these restoring
forces are complicated and do not obey the simple wave equation. An exception are so-called tidal waves
which we discuss first.
 161

Tidal waves

These waves occur in a liquid which has a depth ℎ much smaller than the wavelength of the waves, or
alternatively when the wavelength of the waves is much larger than the depth of the liquid. In such a case it
can be shown that if 𝑌(𝑥, 𝑡) represents the height of the liquid above its natural position, then the waves
obey the wave equation and have speed

𝑣 = √𝑔ℎ

where 𝑔 = 9.8 m/s2 is the acceleration due to gravity. Tidal waves (also known as bores) can occur along rivers
and canals, and may be caused by incoming tides (hence the name tidal waves). Another important example
are so-called tsunamis, which are waves generated by shifts in the ocean floor caused by earthquakes under
the ocean. These waves, in the deep ocean, have long wavelengths but small amplitude and move extremely
quickly. For instance if ℎ = 5 km, then 𝑣 = 220 m/s (or about 800 km/h). The waves can move from their point
of origin through the great oceans and as they near the shallower regions they slow down (ℎ decreases).
Their amplitude therefore increases in shallower water and they sweep onto the coastline causing massive
destruction.

Surface waves

For surface waves the disturbance does not extend down to the bottom of the liquid, but is restricted to the
region near the surface. Most water waves fall in this category. Two forces play an important role in restoring
the water to a flat surface after it has been disturbed, namely gravity and surface tension. The resulting waves
do not obey the standard wave equation, but it can be shown that they have a speed given by

𝑔𝜆 2𝜋𝜎
𝑣=√ +
2𝜋 𝜆𝜌

where 𝜆 is the wavelength, 𝜎 is the surface tension and 𝜌 is the density.

 For very long wavelengths the first term due to gravity dominates and the wave speed is large. This
corresponds, for example, to ocean waves.
 For very short wavelengths the second term due to surface tension dominates and the wave speed is
again large. This corresponds, for example, to the ripples in a swimming pool caused by insects or the
wind.
 At some intermediate critical wavelength the there exists a minimum wave speed. It is left for an exercise
𝜎
to show that this occurs for a wavelength 𝜆crit = 2𝜋√ and the corresponding wave speed is 𝑣min =
𝑔𝜌

4 4𝑔𝜎
√ 𝜌
. The surface tension of water varies with temperature, but is approximately 𝜎 = 7 x 10-2 N/m.
Therefore the critical wavelength is about 17 mm and the corresponding minimum wave speed is 0.23
m/s.

One also gets surface waves in a solid e.g. the destructive Rayleigh waves from an earthquake. These differ
fundamentally from the surface waves in a liquid, due to the different forces at work.
 162

13.4. Wavefronts, Huygens' principle and Refraction

13.4.1. Wavefronts and Huygens' principle

A wavefront consists of all positions or points of a wave having the same phase. For a one-dimensional wave
this is a meaningless concept because every position along the line has a different phase. But consider a
water wave, which is two-dimensional: the wave crests and troughs are examples of wavefronts. A point
source (e.g. a stone dropped in water) will give rise to circular wavefronts, while a long line source (e.g. a
long plank dropped in the water) will give rise to linear wavefronts.36 Note that when a circular wavefront
has expanded greatly, its curvature will become negligible and any small part of it will approximate a line
wave. In a three dimensional medium a point source will give rise to spherical wavefronts, a line source will
give rise to cylindrical wavefronts and a surface source will give rise to plane waves. Both spherical and
cylindrical waves, after expanding greatly, approximate plane waves.

Huygens' principle gives a method of establishing how a wavefront moves.37 It states that every point on an
existing wavefront can be regarded as a source of a new secondary wavelet (circular for a 2-D wave and
spherical for a 3-D wave) which travel at the same speed as the original wave in the medium, as illustrated
for an arbitrary wavefront in Figure 69, and after a short time the new wavefront of the original wave lies
along the tangent to the wavefronts of the secondary wavelets.

Figure 69 Huygens’ construction.

The reader should confirm that Huygens' construction means that circular and linear 2-D waves will retain
their circular and linear wavefronts as they expand. Because the length of the circular wavefront increases in
proportion to the radius, the energy per unit length along the wavefront is inversely proportional to the

36
Another example of almost linear wavefronts occurs for waves arriving at the beach.
37
Huygens published this principle in 1678 in a treatise called Traite de la Lumiere on the wave theory of light. There is a lot of
interesting discussion on this principle at the website http://www.mathpages.com/home/kmath242/kmath242.htm.
 163

𝐴0
radius. This means the amplitude must decrease according to 𝐴 = , because the energy is proportional to
√𝑟
the amplitude squared. Note that the length of the wavefront of a linear wave remains constant as it
propagates and so its amplitude does not decrease. Similarly in 3-D, although harder to visualize, spherical,
cylindrical and plane waves all retain their shapes as they propagate.

 The area of a plane wave does not change as it propagates, so its amplitude does not decrease.
 The area of a cylindrical wavefront increases with its radius, so the amplitude decreases according to 𝐴 =
𝐴0
.
√𝑟
 The area of a spherical wavefront increases with its radius squared, so the amplitude decreases according
𝐴0
to 𝐴 = 𝑟
.

These are geometrical effects and are not due to damping of the wave, which may cause a faster decrease in
amplitude.

13.4.2. Huygens' principle applied to refraction and reflection

Consider a line or plane wave approaching an interface at an angle 𝜃1 from the normal. The wavefronts are
perpendicular to the direction of propagation. At point A a part of the wavefront reaches the interface, while
the rest of the wavefront is still in the first medium. Suppose the wavefront takes time 𝑡 for the wavefront
to reach point B. During that time the secondary circular wavelet travels a distance 𝑣2 𝑡 in the new medium.
The new wavefront must include point B and be tangent to the circular secondary wavelet.

Figure 70 Huygens' construction applied to refraction.

𝑣2 𝑡 𝑣1 𝑡
Using triangle ABC one has sin 𝜃2 = AB
, while using triangle ABD one has sin 𝜃1 = AB
. Therefore

𝑣1 𝑡
sin 𝜃1 𝑣1
= 𝑣AB𝑡 = . (98)
sin 𝜃2 2 𝑣2
AB
 164
𝑣
This means sin 𝜃2 = 𝑣2 sin 𝜃1 . It is called Snell's law (well-known from optics, but expressed in terms if
1
refractive indices) and determines the direction of the refracted wave.

Figure 71 Refraction.

 If 𝑣2 = 𝑣1 as in Figure 71(a), then 𝜃2 = 𝜃1 and refraction does not bend the wave. Stated better,
refraction does not occur.
 If 𝑣2 < 𝑣1 as in Figure 71(b), then 𝜃2 < 𝜃1 and refraction bends the wave towards the normal.
 If 𝑣2 > 𝑣1 as in Figure 71(c), then 𝜃2 > 𝜃1 and refraction bends the wave away from the normal. If
𝑣
𝑣2 = sin1𝜃 then sin 𝜃2 = 1, so 𝜃2 = 90° and the refracted wave will lie along the interface. The angle
1
𝑣1
of incidence for which this occurs is called the critical angle and is given by sin 𝜃𝑐 = . If the angle
𝑣2

of incidence exceeds this critical angle then the refracted wave cannot occur.

Interestingly, if the new medium can support different types of waves with different speeds, there can be
different refracted waves at different angles, one for each wave speed. Each type of refracted wave will have
its own critical angle.

Snell's law is equally valid for reflected waves as it is for refracted waves. For a wave incident on an interface,
the same type of wave can always be reflected, and because the speeds of the incident and reflected waves
are identical the angle of reflection is equal to the angle of incidence. But if the medium of the incident wave
supports more than one type of wave, and these have different speeds, it is possible for there to be additional
reflected waves at different angles to the angle of incidence!
 165

Figure 72 Acoustic energy partitioning at water/steel interface (ASM Handbook vol. 17, p.237).

Example: Figure 72 contains information on the reflection and refraction of acoustic energy when a
longitudinal wave in water strikes a piece of steel. In the steel both a longitudinal and a transverse wave are
created. The inset gives information on the relevant densities and wave speeds. As expected from our
discussion of waves in solids, the longitudinal wave in the steel travels about double as fast as the transverse
wave. Suppose the angle of incidence is 𝜃1 = 10˚. Then we can calculate the refracted angles as follows:

𝑣 5900
 For the longitudinal wave in steel sin 𝜃2 = 𝑣2 sin 𝜃1 = 1490 sin 10° = 0.688, therefore 𝜃2 = 43°.
1
3230
 For the transverse wave in steel sin 𝜃2 = 1490 sin 10° = 0.376, therefore 𝜃2 = 22°.
 166

The scales on the top left corner of the diagram confirm these calculations. These waves are both refracted
away from the normal, which occurs if the wave speed in the new medium is greater than in the original
medium.

The main diagram shows the relative intensities of the different waves as a function of the angle of incidence.
The calculation of this data is complicated as it depends on the boundary conditions for the waves at the
interface, and is made more difficult by the fact that there are two types of transmitted waves. However, it
is interesting to see that in general the reflected intensity fraction is high (above 80%) and that for low angles
of incidence (below 15˚) the transmitted wave is mainly longitudinal while for angles of incidence between
15˚ and 27˚ only the transverse wave is transmitted. Clearly the angles of 15˚ and 27˚ are critical angles for
the longitudinal and transverse waves respectively, and this can be confirmed by calculation:

𝑣 1490
 For the longitudinal wave in steel sin 𝜃𝑐 = 𝑣1 = 5900, therefore 𝜃𝑐 = 14.6°.
2

1490
 For the transverse wave in steel sin 𝜃𝑐 = 3230, therefore 𝜃𝑐 = 27.5°.

For normal incidence (0˚) of water to steel there is no transverse wave created in the steel for this angle of
incidence, which is expected because the force due to the incident wave will be completely perpendicular at
the interface. Also the reflected wave intensity fraction is about 88%, with the remaining 12% in the
transmitted longitudinal wave. While we will not attempt to reproduce all the necessary calculations to
create this diagram, we can test the values just quoted for normal incidence. The specific impedance of the
water is 𝑍1′ = 𝜌1 𝑣1 which is 1.49 x 106 in SI units, while the specific impedance of the steel (for longitudinal
waves) is 𝑍2′ = 𝜌2 𝑣2,𝑙 which is 45.43 x 106 in SI units. Now

𝑃𝐽 4𝑍1′ 𝑍2′
 The transmitted power (and therefore intensity) fraction is = 2 which is 0.123 or 12.3%.
𝑃𝐼 (𝑍1′ +𝑍2′ )
2
𝑃𝑅 (𝑍1′ −𝑍2′ )
 The reflected intensity fraction is = 2 which is 0.877 or 87.7%.
𝑃𝐼 (𝑍1′ +𝑍2′ )

Notice that we have used the specific impedance (impedance per unit area) – the areas of all the waves are
matched and would cancel out in the intensity calculations if included.

The water, being a liquid, supports only longitudinal waves. Therefore for any wave incident from the water
to the steel there is only one reflected wave and the reflected angle equals the incident angle. However, if a
wave were incident from the steel to the water, because the steel can support both longitudinal and
transverse waves there can be two different reflected waves at different angles.
 167

13.5. Electrical waves on transmission lines

Electrical signals are not transmitted instantaneously over a long cable (transmission line), but propagate as
waves. Suppose a wire has capacitance per unit length 𝐶 ′ , and let 𝑄[𝑥, 𝑡] represent the net amount of charge
that has passed the point 𝑥 by time 𝑡. Consider a tiny piece of the wire between 𝑥 and 𝑥 + 𝑑𝑥, which then
has capacitance 𝐶 = 𝐶 ′ 𝑑𝑥 and contains net charge 𝑄[𝑥, 𝑡] − 𝑄[𝑥 + 𝑑𝑥, 𝑡] . From the definition of
𝑄
capacitance (𝐶 = 𝑉 ) this piece of the wire will be at potential
𝑄[𝑥, 𝑡] − 𝑄[𝑥 + 𝑑𝑥, 𝑡] 1 𝜕𝑄
𝑉(𝑥, 𝑡) = ′
=− ′
𝐶 𝑑𝑥 𝐶 𝜕𝑥

as 𝑑𝑥 → 0.

Now consider a longer piece of the wire, between 𝑥 and 𝑥 + ∆𝑥. The potential difference across it is
𝜕𝑄 𝜕𝑄
1 𝜕𝑄 𝜕𝑄 1 𝜕𝑥 [𝑥 + ∆𝑥, 𝑡] − 𝜕𝑥 [𝑥, 𝑡]
∆𝑉 = 𝑉[𝑥 + ∆𝑥, 𝑡] − 𝑉[𝑥, 𝑡] = − ′ [ [𝑥 + ∆𝑥, 𝑡] − [𝑥, 𝑡]] = − ′ ∆𝑥.
𝐶 𝜕𝑥 𝜕𝑥 𝐶 ∆𝑥

If ∆𝑥 is made small then

1 𝜕2𝑄
∆𝑉 = − ∆𝑥.
𝐶 ′ 𝜕𝑥 2

But there is a second way to write this potential difference. Inductance is related to the rate of change of
𝜕𝐼
the current by ∆𝑉 = −𝐿 𝜕𝑡. If the inductance per unit length of the wire is 𝐿′ , then the inductance of our
𝜕𝑄
piece is 𝐿′ ∆𝑥. But 𝐼 = and so
𝜕𝑡
𝜕2𝑄
′
∆𝑉 = −𝐿 ∆𝑥.
𝜕𝑡 2

By comparing these two expressions for the potential difference, we get

𝜕2𝑄 ′ ′
𝜕2𝑄
=𝐿𝐶 .
𝜕𝑥 2 𝜕𝑡 2

This is the equation of a one dimensional wave, with velocity

1
𝑣= .
√𝐿′ 𝐶 ′

Typical values for a coaxial cable are 𝐿′ = 7 × 10−7 H/m and 𝐶 ′ = 6 × 10−11 F/m, giving a speed of 1.5 ×
108 m/s. This is half the speed of light!

1 𝜕𝑄 𝜕𝑄
The potential given by 𝑉 = − 𝐶 ′ 𝜕𝑥 and the current given by 𝑖 = 𝜕𝑡
also propagate as waves with the same
speed. Because the signals travel so fast, they can be considered as instantaneous for small systems. But for
large systems the wave model is vital. How large must the system first be? The answer depends on the
frequency at which we wish to work.

 Electricity in South Africa is an alternating current with a frequency of 50 Hz. If we assume that the
𝑣
signals propagate along the cables at half the speed of light, then the wavelength is 𝜆 = 𝑓 =
1.5×108
50
= 3 × 106 m or 3000 km. Over the scale of the whole country it might be possible that wave
effects will have to be taken into account, but within one city or municipality the distances are very
small in comparison and the electricity can probably be regarded as instantaneous.
 168

1.5×108
 A computer network runs at 100 MHz. The wavelength is 𝜆 = = 1.5 m. Even small computer
100 ×106
networks are much larger than this, and so the wave properties of the electrical signals is always of
importance! For example, if the signal reaches the end of the cable then it is reflected like a wave.
The incident and reflected signals can interfere to form standing waves. At the nodes the signals will
cancel out, and a computer attached there will receive no signal! To prevent this, we must correctly
terminate the cable to eliminate the reflection. To achieve this we must consider the impedance of
electrical waves, which we do shortly.

Electrical waves on transmission lines and how they propagate was essential physics required for developing
and improving the telegraph, and later telephone transmissions. Research funding from the
telecommunications companies was vital for this work. The table below list some analogies in the concepts
and quantities between mechanical waves on a string and electrical waves on a transmission line.

Mechanical wave on a string Electrical wave on transmission line

Transverse displacement (𝑌) Charge (𝑄)

𝜕𝑌 𝜕𝑄
Transverse velocity (𝑣𝑦 = ) Current (𝐼 = )
𝜕𝑡 𝜕𝑡

𝜕𝑌 1 𝜕𝑄
Transverse force (𝐹𝑦 = 𝑇 ) Potential (𝑉 = − )
𝜕𝑥 𝐶′ 𝜕𝑥

Tension (𝑇) Inverse of capacitance per unit length (1/𝐶 ′)

Linear density (𝜇) Inductance per unit length (𝐿′ )

𝑇 1
𝑣=√ 𝑣=√ ′ ′
𝜇 𝐿𝐶

𝑇 𝐹𝑦 1 𝐿′ 𝑉
𝑍= = √𝑇𝜇 = 𝜇𝑣 = 𝑍= = √ = 𝐿′ 𝑣 =
𝑣 𝑣𝑦 ′
𝐶𝑣 𝐶 ′ 𝐼

If we compare the derivation for electrical waves on a transmission line to that for mechanical waves on a
string, there is a correlation between the electrical and mechanical quantities:

 Here we are working with the charge 𝑄 passing a point, rather than the transverse displacement 𝑌
of the string.
 The potential difference ∆𝑉 over a piece of the wire is similar to the net transverse force on a piece
of the string.
1 𝜕𝑄 𝜕𝑌
 The potential 𝑉 = − at a point along the wire is similar to the transverse force 𝐹𝑦 = 𝑇 at a
𝐶 ′ 𝜕𝑥 𝜕𝑥
point along the string. Except for the change in sign, the inverse of the capacitance per unit length
plays the same role as the tension in the string.38
 The inductance per unit length 𝐿′ of the wire plays the same role as the mass per unit length of the
string.

38It is also interesting to note that the small angle approximation is not needed at all for the electrical case. This means that the
electrical waves do not have to be small like mechanical waves in order to obey the wave equation.
 169

𝑇
The impedance for a wave on a string is 𝑍 = , and similarly the impedance for an electrical wave is
𝑣
1
𝑍=
𝐶′𝑣

𝐿′
with equivalent expressions being 𝑍 = √𝐶 ′ = 𝐿′ 𝑣. This impedance has units of electrical resistance (Ω).39

Using the typical values of 𝐿′ and 𝐶 ′ given earlier, one gets 𝑍 = 108 Ω. If, for example, two coaxial cables are
joined, then the transmission and reflection amplitude ratios of a voltage wave can be computed similarly
to a force wave on a string, namely with
2𝑍2 𝑍2 − 𝑍1
(Voltage) Transmission: 𝓉 = Reflection: 𝓇 = .
𝑍1 + 𝑍2 𝑍1 + 𝑍2

The transmission and reflection amplitude ratios of the current wave must be calculated similarly to
transverse velocity wave (which are identical to those for a transverse displacement wave), namely
2𝑍1 𝑍1 − 𝑍2
(Current) Transmission: 𝑡 = Reflection: 𝑟 = .
𝑍1 + 𝑍2 𝑍1 + 𝑍2

To avoid a reflection at the joint, the cables that are being joined must have the same impedance. For
example, the antenna for a television has a certain impedance, and it must be joined to the television with a
coaxial cable having the same impedance, otherwise a part of the signal from the antenna will not be
transmitted to the television, and the image will be of poor quality.

A transmission line of impedance 𝑍 ended with the two wires connected via a resistor 𝑅 is similar to a
stretched string ended by tying it to a (massless) ring which can slide vertically but experiences friction
proportional to its speed. Adapting the results for the reflection amplitude ratio from the case of the string
gives

𝑅−𝑍 𝑍−𝑅
Reflection (Voltage) 𝓇 = (Current) 𝑟 = .
𝑅+𝑍 𝑍+𝑅

If the terminating 𝑅 is chosen to be equal to the impedance of the transmission line 𝑍, then there will be no
reflected signal! Therefore a computer network cable that is not being used must be terminated with a
resistance that is equal to the impedance of the cable. If a different resistance is used then the wave will be
partially reflected and partially absorbed and the amplitude of reflected wave will be smaller than the
incident wave. Something similar to a standing wave is created, but at the nodes the reflected wave will now
not be able to completely cancel the incident wave a small oscillation with amplitude equal to the difference
in the wave amplitudes will occur. Of course, at the antinodes the oscillation amplitude will be the sum of
the wave amplitudes. The ratio of these quantities is called the “standing wave ratio”
𝐴𝐼 + |𝐴𝑅 |
𝑆𝑊𝑅 =
𝐴𝐼 − |𝐴𝑅 |

and is an important quantity in electrical engineering.

39 The reader should check this. But we can also motivate it as follows: the general definition for the impedance of a wave on a string
is the ratio of transverse force to transverse velocity. For electrical waves the potential plays the role of the transverse force while
the current plays the role of the transverse velocity. Therefore 𝑍 = 𝑉/𝐼, which clearly has units of resistance.
 170

If a transmission line comes to an end, but it is not terminated, then this is equivalent to terminating it
through a resistor 𝑅 = ∞. A voltage wave which is reflected will not be turned upside down, but a current
wave will be. This is similar to a string which has its end fixed. But if the two wires in the cable (i.e.
transmission line) are joined at the end then 𝑅 = 0 and a voltage wave which is reflected will be turned
upside down, but not a current wave. This is similar to a string which is free at the end (i.e. attached to a
massless ring that can slide up and down without resistance).

13.6. Electromagnetic waves

13.6.1. Introduction

The nature of light is one of the great questions of physics. Here we will consider light as consisting of
electromagnetic waves, being only part of the electromagnetic spectrum (which includes radio waves,
microwaves, infrared light, ultraviolet light, X-rays and gamma rays). But the history of scientific thought on
the nature of light is very interesting and will show that this interpretation of light is only part of the story.

Figure 73 The complex nature of the shadow of light illuminating a razor blade.

Everyday experience with light rays and shadows shows that light seems to travel in straight lines, like a
stream of particles. On the contrary, Young's famous double slit interference experiment (around the year
1800) clearly demonstrated the wave nature of light, and careful investigation of shadows shows that they
are not really sharp, as Figure 73 clearly illustrates. That these effects are not obvious, but only visible under
careful investigation, can be attributed to the very short wavelengths of visible light (between 400 and 700
nm) as well as the fact that most sources of light are incoherent, which masks interference effects.

Maxwell first produced the modern accepted theory of classical electromagnetism. After calculating the
speed of electromagnetic waves (around the year 1862) and finding it was equal to the measured speed of
light, he wrote the following:
 171

We can scarcely avoid the conclusion that light consists in the transverse undulations of the
same medium which is the cause of electric and magnetic phenomena.

Maxwell therefore realised that light was an electromagnetic wave, and also that it was a transverse wave.40
But notice how Maxwell referred to a medium in which the light travels, a medium which would be distorted
as the wave travelled. This medium, which was called the aether, would have to be very stiff to give light its
great speed, yet still allow matter to pass freely through it. Attempts were made to measure the speed of
the earth through the aether (the most famous is that of Michelson and Morley) by trying to measure small
changes in the speed of light as the earth moved in opposite directions in its orbit around the sun. These
experiments failed to find any changes in the speed of light, and later Einstein postulated in his special theory
of relativity that the speed of light (in vacuum) is a universal constant, irrespective of the motion of the
observer. As a consequence, the special theory of relativity implied that the aether did not exist.
Electromagnetic waves do not require a medium through which to propagate.

Maxwell's theory of electromagnetism solved, in part, the mystery of the nature of light. However, there are
some effects it cannot explain, particularly where light (or in general, electromagnetic radiation) is created
or destroyed at the atomic scale. Specific examples are the relative intensities of different wavelengths
emitted by a hot glowing body (black body radiation) and the emission of electrons by metals when
illuminated by a short wavelength light source (photoelectric effect). Planck developed a theory for black
body radiation by postulating that the radiation consists of “quanta” and Einstein developed a theory for the
photoelectric effect by postulating that the radiation consists of “photons”. These theories are based on ideas
in quantum mechanics, from which it follows that light can also have a particle nature. Just as Einstein's
special theory of relativity has refined Newton's classical mechanics, quantum electrodynamics has refined
Maxwell's classical electromagnetism. Quantum electrodynamics can be used to correctly predict the
quantum optical effects occurring when light interacts with matter, and is necessary to understand the
production of blackbody radiation and the photoelectric effect, as well as how light is produced by a laser.
Despite these interesting advances, it is best to consider light to have a dual nature of both waves and
particles. The aspects of the electromagnetic wave nature of light are considered in the rest of the section.

13.6.2. Electromagnetic waves in vacuum

Using Maxwell's equations in vacuum, it can be shown that the electric and magnetic field vectors both obey
the wave equation, and both propagate with speed
1
𝑐= .
√𝜀0 𝜇0
The quantity 𝜀0 = 8.854 pF/m (used in the calculation of electric fields) is called the permittivity of free
space, while the quantity 𝜇0 = 400𝜋 nH/m (used in the calculation of magnetic fields) is called the
permeability of free space. These values give 𝑐 = 2.998 × 108 m/s, which is the speed of light.

The derivation of the wave equation of electromagnetic waves follows a different pattern from those we
have seen so far, because there is no physical quantity (matter or charge) which is being displaced. It is
therefore not obvious how to relate quantities we have encountered for other waves to electromagnetic
quantities. One way we can do this is by considering the equation for the energy density 𝑢 of an
electromagnetic field in vacuum, as given in electromagnetism textbooks, namely

40 Earlier scientists who had accepted the wave model believed light would be a longitudinal wave, similar to sound waves.
 172

1 1
𝑢 = 𝜇0 𝐻 2 + 𝜀0 𝐸 2 .
2 2
𝐵
In this equation 𝐸 is the electric field and 𝐻 = 𝜇 represents the magnetic field.41 Now recall that for a wave
0
1 𝜕𝑌 2 𝜕𝑌 2
on a string the energy is given by 2
∫ 𝜇 ( 𝜕𝑡
) + 𝑇 (𝜕𝑥) 𝑑𝑥, so the (linear) energy density is

1 𝜕𝑌 2 1 1 𝜕𝑌 2 1 2 1 1 2
𝑢 = 𝜇( ) + (𝑇 ) = 𝜇𝑣𝑦 + 𝐹 ,
2 𝜕𝑡 2 𝑇 𝜕𝑥 2 2𝑇 𝑦

𝜕𝑌 𝜕𝑌
where in the first term 𝑣𝑦 = represents the transverse velocity and in the second term 𝐹𝑦 = 𝑇 represents
𝜕𝑡 𝜕𝑥
the transverse force. By comparison of the formulae for an electromagnetic wave and wave on a string, the
following correspondence (analogy) can be found:

Wave on a string Electromagnetic wave in vacuum

Transverse displacement (𝑌) [ none ]

𝜕𝑌
Transverse velocity (𝑣𝑦 = ) Magnetic field (𝐻)
𝜕𝑡

𝜕𝑌
Transverse force (𝐹𝑦 = 𝑇 ) Electric field (𝐸)
𝜕𝑥

Tension (𝑇) Inverse of permittivity of free space (1/𝜀0 )

Linear density (𝜇) Permeability of free space (𝜇0 )

𝑇 1
𝑣=√ 𝑐=√
𝜇 𝜇0 𝜀0

𝑇 𝐹𝑦 1 𝜇0 𝐸
𝑍= = √𝑇𝜇 = 𝜇𝑣 = 𝑍0 = = √ = 𝜇0 𝑐 =
𝑣 𝑣𝑦 𝜀0 𝑐 𝜀0 𝐻

Note that 𝜀0 has the units of capacitance per unit length, just like 𝐶 ′ in the similar table for electrical
transmission waves in section 13.5, while 𝜇0 has the units of inductance per unit length, just like 𝐿′ . The
quantity 𝑍0 is called the impedance of free space and has a value of 377 Ω. Recall that the impedance of a
wave on a string is the ratio of transverse force to transverse velocity. Since in this scheme for
electromagnetic waves the electric field plays a role analogous to the transverse force, while the magnetic
𝐸
field plays a role analogous to the transverse velocity, it follows that 𝑍0 = 𝐻. This is equivalent to the result
𝐸 𝐸
𝜇0 𝑐 = , or = 𝑐 for an electromagnetic wave in vacuum.
𝐵/𝜇0 𝐵

41 A moving charge experiences a force in a magnetic field and this is given by 𝐹⃗ = 𝑞𝑣⃗ × 𝐵
⃗⃗ where 𝐵⃗⃗ is the magnetic field. What then
⃗⃗
𝐵
⃗⃗ ? In a vacuum 𝐻
is 𝐻 ⃗⃗ = , so it is a just scaled version of the magnetic field. Unfortunately calling both 𝐵 ⃗⃗ and 𝐻
⃗⃗ the magnetic field
𝜇0
creates confusion. There is debate which one should get the title and it might be easiest to just call them the 𝐵 ⃗⃗-field and 𝐻
⃗⃗ -field,
⃗⃗ is generally the ‘magnetic field’ first introduced in general physics textbooks when
respectively, as is often done. Although 𝐵
magnetism is discussed, in the SI system 𝐻⃗⃗ is called the magnetic field strength while 𝐵 ⃗⃗ is called the magnetic flux density. For
⃗⃗ ⃗⃗
electromagnetic waves, considering 𝐻 rather than the well-known 𝐵 makes an analogy for reflection and transmission with other
⃗⃗ (not 𝐵
types of waves easiest, because the boundary condition at an interface of 𝐻 ⃗⃗) is similar to that of the earlier discussed waves.
 173

13.6.3. Electromagnetic waves in insulators

In a material (matter) containing no free charges and currents (i.e. insulators) the electromagnetic fields still
obey the wave equation, but with the following changes:

 The permittivity is changed from 𝜀0 to 𝜀 = 𝜀𝑟 𝜀0 . The factor 𝜀𝑟 is called the relative permittivity or the
dielectric constant. It is unitless.
 The permeability is changed from 𝜇0 to 𝜇 = 𝜇𝑟 𝜇0 , not be confused with the linear density for a
𝐵
string. Therefore 𝐻 = 𝜇. The factor 𝜇𝑟 is called the relative permeability and is also unitless.
1
 The wave speed is changed from 𝑐 to 𝑣 = .
√𝜀𝜇
1 𝜇 𝐸 𝐸 𝐸
 The impedance is changed from 𝑍0 to 𝑍 = 𝜀𝑣 = √ 𝜀 = 𝜇𝑣 = 𝐻. Hence 𝜇𝑣 = 𝐵/𝜇, so 𝐵 = 𝑣.

The refractive index, 𝑛, is defined as the ratio of the speed of light in vacuum to that in a material, and so
1
𝑐 √
𝜇0 𝜀0
𝑛= = = √𝜇𝑟 𝜀𝑟 .
𝑣 1
√
𝜀𝑟 𝜀0 𝜇𝑟 𝜇0

There is no name given for the similar ratio of the impedances, namely
𝜇
𝑍0 √ 𝜀0 𝜀𝑟 𝜀𝑟 𝑛
0
= =√ = = .
𝑍 𝜇 𝜇 𝜇𝑟 𝑛 𝜇𝑟
√ 𝜀𝑟 𝜀 0
𝑟 0

Since insulators generally show only very weak magnetic effects, the approximation 𝝁𝒓 ≈ 𝟏 is excellent
𝒁𝟎 𝒁𝟎
and usually assumed. Under this assumption, both 𝒏 ≈ √𝜺𝒓 and ≈ √𝜺𝒓 , so that ≈ 𝒏.
𝒁 𝒁

sin 𝜃1 sin 𝜃2 sin 𝜃1 sin 𝜃2

Snell's law for refraction, namely 𝑣1
= 𝑣2
, can be written as 𝑐/𝑛1
= 𝑐/𝑛2
or 𝑛1 sin 𝜃1 = 𝑛2 sin 𝜃2 ,
while the critical angle is given by sin 𝜃𝑐 = 𝑛2 /𝑛1. When light is incident on an interface at an angle greater
than the critical angle, all the light is reflected and one calls this total internal reflection. This has application
in glass optical fibres: these have a cylindrical shape and are made with two types of glass. The glass near
the axis (the core) has a higher index of refraction than the glass around the outside (the cladding), and light
in the core undergoes repeated total internal reflection off the interface between the core and cladding, thus
travelling down the optical fibre. Most telephone and internet data is transmitted far distances (up to
thousands of kilometres) in this way using laser light. Optical fibre transmission of data is generally more cost
effective than using electrical signals or satellite transmissions.

𝑐 𝑐
The equation for the wave speed 𝑣 = 𝑓𝜆 can be written as 𝑛 = 𝑓𝜆 or 𝑓 = 𝑛𝜆. If the wave passes from one
medium to another the frequency is not altered and 𝑐 is fixed, so 𝑛1 𝜆1 = 𝑛2 𝜆2 .

Unlike 𝜀0 and 𝜇0 , which are universal constants, 𝜀 and 𝜇 differ from one material to another and are also
functions of the wavelength and frequency. Therefore electromagnetic waves in a material are dispersive
and a wave pulse may change its shape as it travels. This change of shape can make defining the speed of a
 174

wave tricky, which is behind some claims of waves being able to travel faster than the speed of light.42 The
dispersion of light in a glass prism, with the different wavelengths travelling at different speeds, allows the
colours to be separated, while the dispersion of light in spherical water droplets is responsible for rainbows.

Empty space is isotropic, i.e. the same in all directions. So are liquids and glasses, but some crystals (e.g.
calcite and quartz) have different properties in different directions. Then the speed of light in such crystals is
different in different directions. Such crystals are called birefringent and their permittivity and permeability
are not scalars, but tensors!

Note: It is interesting that electromagnetic waves also obey the wave equation in conductors (i.e. metals),
but only if the permittivity is considered as a complex number of which the imaginary part depends on the
conductivity. This means the refractive index and the wave speed are also complex quantities and the
electromagnetic waves do not propagate through metals simply as they do through insulators. Instead they
are partly absorbed and strongly reflected.

13.6.4. Polarization of electromagnetic waves

The theory of electromagnetism can be used to show that the electric and magnetic fields, in addition to
travelling as a wave, must also obey certain further constraints. For example, the electric and magnetic fields
must both be perpendicular to the direction of propagation, making the wave transverse and therefore
polarisable. As well as both the electric and magnetic fields being perpendicular to the direction of
propagation, they must also be perpendicular to one another. By convention, the direction of polarisation
of the electromagnetic wave is chosen as the direction of the electric field. Such an electromagnetic wave
is shown in Figure 74.

Figure 74 Electromagnetic wave.

Light from common sources such as incandescent and fluorescent light bulbs and the sun is produced by
many excited atoms all undergoing independent motion. This means that the light is generally incoherent

42For example, see https://phys.org/news/2007-01-mach-scientists-faster.html commenting on the article “Sound beyond the speed
of light: Measurement of negative group velocity in an acoustic loop filter.” Applied Physics Letters 90, 014102 (2007).
 175

and unpolarised as well as not being monochromatic. One might consider this “poor quality” light. In contrast,
the light produced by a laser is created by the process of stimulated emission, where an excited electron does
not decay randomly, but rather under the influence of a photon already existing in the laser cavity. The new
photon emitted is then dependent on, and matched to, the existing photon so that the light produced by the
laser is coherent, monochromatic and generally polarised.

Polarisation of light passing through a birefringent crystal

A birefringent crystal is one which is not optically isotropic. Rather, the speed of electromagnetic waves
depends on the direction they are travelling as well as the direction of polarisation. So if a beam of light
enters a birefringent crystal, it splits up into two beams of opposite polarisation which are refracted at
different angles because of their different speeds. This is illustrated in Figure 75.

Figure 75 Double refraction by a birefringent crystal.

The first report of this effect was made by Erasmus Bartholinus (University of Copenhagen) in 1669, who
discovered double images when light passed through a calcite crystal. In 1690 Huygens found that the two
beams were polarised.43 For the next 100 years there was little scientific progress on polarised light. In 1828
Nicol began using birefringent crystals to make the polarising filters, called Nicol prisms.

Polarisation of light by reflection

In 1808 Malus (in Paris) discovered that reflected light is (partially) polarised. He did this by observing sunlight
reflected from a glass window through a crystal of calcite, and found that the transmitted intensity changed
𝑃
as the calcite crystal was rotated. From this he developed Malus's law (𝑃 = cos2 𝜃) which has previously
0
been discussed in section 12.1.5.

43It is interesting that he had to use ellipses rather than circles for his secondary wavelets because the speed of the light was
dependant on the direction.
 176

Figure 76 Polarization by reflection and

the Brewster angle.

In 1812 Brewster discovered that the reflected light is completely polarised (sin a direction perpendicular to
the plane of incidence) when the reflected and refracted rays happen to be at right angles to one another.
This is called Brewster's law. Mathematically this occurs when the incident and refracted angles 𝜃 and 𝜑 are
complementary i.e. 𝜃 = 90° − 𝜑. Combining this with Snell's law 𝑛1 sin 𝜃 = 𝑛2 sin 𝜑, one gets

𝑛2
tan 𝜃𝐵 =
𝑛1

for the angle of incidence 𝜃𝐵 (Brewster's angle) necessary for complete polarisation of the reflected light.

Polarisation of light by scattering

Figure 77 Light which is scattered from directly above is polarised.

 177

The light scattered (absorbed and re-emitted) by atoms is also polarised. For example, if a ray is travelling in
a certain direction, and is scattered into a new direction perpendicular to its original path, the scattered ray
will be polarised in the direction perpendicular to the plane containing its original and new direction. Sunlight
is scattered by the gas atoms in the atmosphere.44 Suppose you are facing the sun (which is not directly
overhead) and look directly upwards (Figure 77): the light entering your eyes was travelling from the sun
directly over you and was then scattered downwards, so its polarization direction will lie in the direction
joining your shoulders. This effect was discovered by Arago in 1809, and in 1949 Von Frisch showed that bees
can sense the direction of polarisation of light in the sky and use it (like a compass) to navigate.

Polarisation of light by selective absorption

Some transparent materials contain long molecules which electrons can move along relatively freely. If these
molecules are aligned and an electromagnetic wave passes through the material, then the component of the
electric field wave parallel to the molecules will cause the electrons to oscillate along the molecules. During
this process energy will be transferred from the electromagnetic wave to the oscillating electrons (i.e.
absorbed) and so the component of the electric field parallel to the molecules will decrease. However, the
component of the electric field perpendicular to the molecules will pass through and so the electromagnetic
wave that exits will have half the intensity and be polarised perpendicular to the molecules. 45 Materials
having selective absorption depending on the polarisation direction are called dichroic.

Figure 78 Vertically oriented polaroid in sunglasses reduces reflected light (glare) which is horizontally polarised.

In 1852 Herapath created a synthetic crystal which showed very strong dichroism and could be used for
polarising light. In 1928 Edwin Land had the idea of embedding large molecules in plastic and then stretching
the plastic to create aligned long molecules. This led to the development of inexpensive polaroid sheets to
polarise light, so that it could be used in everyday applications. One such application is in sunglasses (Figure
78): reflected light (glare) is partially horizontally polarised, and so the glare is reduced if the light must pass
through the vertically oriented polaroid of the sunglasses.46

44 This is called Rayleigh scattering and occurs more readily for shorter wavelengths, which is why the sky is blue in the day (you are
looking at the scattered blue short wavelength light) and red around the sun in the evening (when you are looking at light from the
sun which has had the short blue wavelengths scattered out of it, leaving the red).
45 This process is very similar to how one can polarize a wave on a stretched string by passing it through a slit. However, note that for

the string the polarisation direction will be parallel to the slit while for electromagnetic waves the polarisation direction is
perpendicular to the long molecules.
46 Land also wished to use polaroid to solve the problem of having to look at the bright headlights of oncoming traffic at night. A

demonstration of how this would work was reported in “Time magazine” in 1936. The simplest scheme would involve every car
 178

13.6.5. Transmission and reflection of electromagnetic waves

(a) Normal incidence

Consider a light wave travelling and reaching an interface with a new material. It will be partly transmitted
and partly reflected. These effects have been discussed for waves on a string and equations for the
transmission and reflection amplitude ratios derived in terms in the impedances. Recall that on a string the
𝜕𝑌
transmission and reflection of the displacement wave (𝑌) and the associated force wave (𝐹𝑦 = 𝑇 𝜕𝑥) have
different equations. Since the electric field aspect of the electromagnetic wave acts analogously to a force
wave on a string, its transmission and reflection amplitude ratios are given by
2𝑍2 𝑍2 − 𝑍1
Transmission: 𝓉 = Reflection: 𝓇 = .
𝑍1 + 𝑍2 𝑍1 + 𝑍2

These formulae were derived for a one dimensional wave on a stretched string and are therefore only
applicable to electromagnetic plane waves incident normally (i.e. perpendicularly) on an interface. These can
𝑍0 𝑛
expressed in terms of the refractive index. If we apply 𝑍
= 𝜇 then
𝑟
𝜇 𝜇𝑟,2 𝜇
2𝑍0 𝑟,2 2𝑛1 𝜇𝑟,2 𝑍0 − 𝑍0 𝑟,1 𝑛 𝜇 − 𝑛 𝜇
𝑛2 𝑛 𝑛1 1 𝑟,2 2 𝑟,1
𝓉= 𝜇 𝜇 = 𝓇= 𝜇2 𝜇𝑟,2 = 𝑛 𝜇 + 𝑛 𝜇 .
𝑍0 𝑟,1 + 𝑍0 𝑟,2 𝑛2 𝜇𝑟,1 + 𝑛1 𝜇𝑟,2 𝑍0 𝑟,1
+ 𝑍0 2 𝑟,1 1 𝑟,2
𝑛1 𝑛2 𝑛1 𝑛2

Since generally 𝜇𝑟 ≈ 1 (except for strongly magnetic materials, which are never transparent anyway), the
commonly applied equations are
2𝑛1 𝑛1 − 𝑛2
𝓉≈ 𝓇≈ .
𝑛2 + 𝑛1 𝑛2 + 𝑛1

The first equation shows the ratio of the transmitted to incident electric field amplitude and is always positive
(it can vary between 0 and 2). The second equation gives the ratio of the reflected to incident electric field
amplitude. This will be negative if 𝑛2 > 𝑛1. Therefore the electric field aspect of the electromagnetic wave
will be turned ‘upside-down’ when reflection occurs from a material with a greater refractive index.

Example: Compare the reflection amplitude ratio and reflectivity for (a) light incident normally from air to
glass, and (b) from glass to air. Air has a refractive index of about 1, while that of glass can be taken as 3/2.

1−3/2
Solution: (a) For light travelling from air into glass, the reflection amplitude ratio is 𝓇 = 1+3/2 = −0.2. The
reflected electric field wave has one fifth of the amplitude of the incident electric field wave and is reversed.
Since the intensity is proportional to the amplitude squared, the reflectivity is given by 𝓇 2 = 4%.
3/2−1
(b) For light travelling from glass into air, the reflection amplitude ratio is 𝓇 = = 0.2. Again the
3/2+1
reflected electric field wave has size one fifth of the amplitude of the incident electric field wave, but in this
case it is not reversed. Despite this difference, the reflectivity is again 𝓇 2 = 4%. Although a reflectivity of
4% is not large, it is easy to see one’s reflection in an ordinary pane of glass when the background behind it
is dark, or the reflection of a bright window on a television screen. For an optical component like a lens, the

headlamp being covered with polaroid to polarize the light at 45° (between vertical and horizontal) and every driver to wear glasses
with polaroid of the same orientation. Any driver would then be able to observe light from their own headlamps, but light from the
headlamps of any oncoming car would be polarised perpendicularly to the orientation on one glasses and its intensity would be
greatly reduced. The idea was, however, not taken up by car manufacturers. You can read more about this and many other interesting
aspects of polarisation at the website www.polarization.com.
 179

reflection from both sides adds to about 8% and if an optical system contains several lenses, the reflection
losses can be substantial.

𝐵
Note: the magnetic field (𝐻 = 𝜇) aspect of an electromagnetic wave acts analogously to the transverse
velocity wave for a wave on a string, which has the same transmission and reflection amplitude ratios as the
displacement wave, namely
2𝑍1 𝑍1 − 𝑍2
(Magnetic field) Transmission: 𝑡 = Reflection: 𝑟 = .
𝑍1 + 𝑍2 𝑍1 + 𝑍2
These equations are similar to those for the electric field but with 𝑍1 and 𝑍2 interchanged. Therefore the
inversion of the reflected magnetic field at an interface is always opposite that of the electric field. By
convention, when we speak about a light wave or an electromagnetic wave, its ELECTRIC field aspect is
generally meant. This is true for both the direction of polarisation and reflected amplitude inversions.

(b) Incidence at an angle: Fresnel's equations

Figure 79 Electromagnetic wave incident on an interface at an angle (a) Electric field in the plane of incidence (b)
Electric field parallel to the interface.

If the light is not incident perpendicularly (so the electric and magnetic fields are no longer parallel to the
interface) then the situation is more complicated. We distinguish between two possibilities, given in Figure
79. Note the incident angle is 𝜃 and the refracted angle is 𝜑.

 In Figure 79(a) the electromagnetic wave is polarised in the plane of incidence. If we are only
interested in the components parallel to the interface then 𝐸𝑖 → 𝐸𝑖 cos 𝜃, but 𝐻𝑖 , 𝐻𝑡 and 𝐻𝑟 are
𝐸 𝐸
unchanged, and so 𝑍1 = 𝐻𝑖 → 𝑍1 cos 𝜃. For the transmitted wave 𝐸𝑡 → 𝐸𝑡 cos 𝜑 and so 𝑍2 = 𝐻𝑡 →
𝑖 𝑡
𝐸𝑡 2𝑍2 𝐸𝑡 cos 𝜑 2𝑍2 cos 𝜑
𝑍2 cos 𝜑. Therefore 𝐸𝑖
= 𝑍1 +𝑍2
→ 𝐸𝑖 cos 𝜃
= 𝑍1 cos 𝜃+𝑍2 cos 𝜑
and the expression on the right
hand side gives a transmission amplitude ratio

𝐸𝑡 2𝑍2 cos 𝜃
𝓉= = .
𝐸𝑖 𝑍1 cos 𝜃 + 𝑍2 cos 𝜑
 180

𝐸𝑟 𝑍2 −𝑍1 𝐸𝑟 cos 𝜃 𝑍2 cos 𝜑−𝑍1 cos 𝜃

For the reflected wave 𝐸𝑟 → 𝐸𝑟 cos 𝜃 and therefore = → =
𝐸𝑖 𝑍1 +𝑍2 𝐸𝑖 cos 𝜃 𝑍1 cos 𝜃+𝑍2 cos 𝜑
with the expression on the right hand side giving a reflection amplitude ratio

𝐸𝑟 𝑍2 cos 𝜑 − 𝑍1 cos 𝜃
𝓇= = .
𝐸𝑖 𝑍1 cos 𝜃 + 𝑍2 cos 𝜑

 In Figure 79(b) the electromagnetic field is polarised perpendicular to the plane of incidence. If we
are only interested in the components parallel to the interface then 𝐻𝑖 → 𝐻𝑖 cos 𝜃, but 𝐸𝑖 , 𝐸𝑡 and 𝐸𝑟
𝐸 𝑍 𝐸
are unchanged, and so 𝑍1 = 𝐻𝑖 → cos1 𝜃. For the transmitted wave 𝐻𝑡 → 𝐻𝑖 cos 𝜑 and so 𝑍2 = 𝐻𝑡 →
𝑖 𝑡
𝑍2 𝐸𝑡 2𝑍2 𝐸𝑡 2𝑍2 / cos 𝜑
cos 𝜑
. Therefore 𝐸𝑖
= 𝑍1 +𝑍2
→ 𝐸𝑖
= 𝑍1 / cos 𝜃+𝑍2 / cos 𝜑
and the expression on the right hand
side gives an transmission amplitude ratio

𝐸𝑡 2𝑍2 cos 𝜃
𝓉= = .
𝐸𝑖 𝑍1 cos 𝜑 + 𝑍2 cos 𝜃

𝐸𝑟 𝑍 −𝑍 𝐸𝑟 𝑍 / cos 𝜑−𝑍 / cos 𝜃

For the reflected wave 𝐻𝑟 → 𝐻𝑟 cos 𝜃 and therefore 𝐸𝑖
= 𝑍2 +𝑍1 → 𝐸𝑖
= 𝑍2 / cos 𝜃+𝑍 1/ cos 𝜑
1 2 1 2
with the expression on the right hand side giving a reflection amplitude ratio

𝐸𝑟 𝑍2 cos 𝜃 − 𝑍1 cos 𝜑
𝓇= =
𝐸𝑖 𝑍1 cos 𝜑 + 𝑍2 cos 𝜃

These are called Fresnel's equations.

Example: One can derive Brewster's angle from Fresnel's equations for the reflection of light incident on an
interface at an angle. If the reflected light is completely polarised perpendicular to the plane of incidence,
then the reflection amplitude ratio for the light parallel to the plane of incidence must be zero. This reflection
𝑍 cos 𝜑−𝑍 cos 𝜃 𝑍0 𝑛
amplitude ratio is 𝓇 = 𝑍2 cos 𝜃+𝑍 1 cos 𝜑 and so we require 𝑍2 cos 𝜑 = 𝑍1 cos 𝜃. If one uses 𝑍
= 𝜇 and accepts
1 2 𝑟

that 𝜇𝑟 ≈ 1 for both materials then

𝑛1 cos 𝜑 = 𝑛2 cos 𝜃.
But Snell's law 𝑛1 sin 𝜃 = 𝑛2 sin 𝜑 must also hold. We can eliminate 𝜑 from these two equations. By
𝑛1 sin 𝜃 2 𝑛2 cos 𝜃. 2
squaring sin2 𝜑 + cos2 𝜑 = ( ) +( ) = 1.
𝑛2 𝑛1

𝑛 4 𝑛 2 𝑛 4 𝑛 2
Then sin2 𝜃 + (𝑛2 ) cos 2 𝜃 = (𝑛2 ) , so 1 − cos2 𝜃 + (𝑛2 ) cos 2 𝜃 = (𝑛2 ) .
1 1 1 1

𝑛 4 𝑛 2 𝑛 4 𝑛 2 𝑛 2
Therefore {(𝑛2 ) − 1} cos 2 𝜃 = (𝑛2 ) − 1. Since we can factorize {(𝑛2 ) − 1} = {(𝑛2 ) − 1} {(𝑛2 ) + 1}
1 1 1 1 1

1
the equation simplifies to cos 𝜃 = , so that
𝑛 2
√( 2 ) +1
𝑛1

𝑛2
tan 𝜃 =
𝑛1
which is Brewster's law.
 181

13.7. Problems

1. Show that in a bulk medium the speed of a longitudinal wave is greater than the speed of a transverse
2−2𝜈
wave by a factor√ .
1−2𝜈

𝜎
2. Show that for surface waves in a liquid the minimum velocity occurs for a wavelength 𝜆crit = 2𝜋√𝑔𝜌 and

4 4𝑔𝜎
that the corresponding wave speed is 𝑣min = √ 𝜌
.

3. An optical fibre used for telecommunication is made up of two layers of silica glass (SiO2), one
surrounding the other. The core glass in the centre part of the fibre is doped with Ge which increases its
refractive index to 𝑛 = 1.488, while the pure cladding glass around the outside has refractive index 𝑛 =
1.475. Calculate the critical angle for light to experience total internal reflection and travels along the
inside (core) of the fibre.

4. A beam of white light moving through air strikes a piece of glass with an angle of incidence of 65˚. The
speed of green light (wavelength 550 nm) in the glass is 2.1 x 108 m/s. (a) What is the refractive index of
the glass for the green light? (b) What is the angle of refraction of the green light? (c) If the refractive
index decreases with increasing wavelength, is the angle of refraction of red light (wavelength 700 nm)
larger or smaller than for green light?

1 𝐿′
5. Prove that the quantity has units of speed and that the quantity √𝐶 ′ has units of resistance.
√𝐿′ 𝐶 ′

6. Light that is linearly polarised in the vertical direction is passed through a polarizer with its polarization
angle at 20˚ to vertical. (a) What percentage of the light intensity passes through the polarizer? (b) The
polarizer is then rotated until 40% of the light intensity passes through. What angle does its polarization
direction now make with the vertical?

7. Unpolarised light with intensity 𝐼0 passes through two polarizing sheets, where the angle between the
polarizing direction of the sheets is 𝜃 . What is the intensity of the light passing through both the
polarizing sheets, in terms of 𝐼0 ?

8. You are asked to test a new high intensity unpolarised light source. You have a light intensity meter that
can read a maximum light intensity of 0.5 W/m2, but when you try to measure the intensity the reading
is over the scale (i.e. it is too large to be measured). (a) You shine the light through a polarizing sheet
(polaroid), but the intensity is still off the scale. What is the minimum possible intensity of the light
source? (b) You place a second polarizing sheet behind the first, so that the angle between their polarizing
directions is 60˚. Your light intensity meter now reads 0.3 W/m2. What is the intensity of the light source?
 182

14. interference from thin films

14.1. Introduction

Figure 80 Different intensities and

colours from a thin soapy water film.

The layer's thickness increases from top

to bottom due to gravity.

When light meets an interface between one material with refractive index 𝑛1 and another with refractive
index 𝑛2 , some light is always reflected. The different colours and intensities observed when sunlight falls on
soap bubbles (Figure 80) or oil slicks are caused by the interference of two light waves, namely the one
reflected from the front surface and that reflected from the back surface of the transparent thin film,
respectively.

14.2. Reflection of light from a single interface

Treating light as a sinusoidal electromagnetic wave, it can be shown that for light incident perpendicularly on
a single interface the reflection amplitude ratio of the electric field is

𝑛1 − 𝑛2
𝓇= . (99)
𝑛1 + 𝑛2

It should be clear that:

 If light is reflected from a material having a higher refractive index compared to that of the material
in which it is travelling, then the reflected electric field amplitude will be reversed. This corresponds
to a phase change in the electric field wave of 180° or 𝜋 rad.
 183

 If the light is reflected from a material having a lower refractive index compared to that of the
material in which it is travelling, then there will be no phase change of the reflected wave.

14.3. Reflection from a thin film

Consider a thin non-absorbing (transparent) film of uniform thickness 𝐿 and index of refraction 𝑛2 , which is
illuminated by an incident light wave 𝑖 of wavelength 𝜆1 from the bottom left (Figure 81). The material in
front of the thin film (generally air) has refractive index 𝑛1 , and the material behind the thin film has
refractive index 𝑛3 .

Figure 81

Model of a thin film.

The incident light strikes the front surface of the film at position 𝑎 and undergoes reflection and refraction.
The reflected ray 𝑟1 moves towards the observer's eye, while the refracted ray crosses the film to position 𝑏
on the back surface where it undergoes reflection and refraction. We ignore the light refracted at 𝑏 which
passing through the thin film, but the light which is reflected at 𝑏 crosses back through the film to position 𝑐
where it again undergoes reflection and refraction.

Example: Consider a thin glass film in air. Estimate the percentage of the incident light intensity which is (a)
reflected into ray 𝑟1 , (b) reflected back from point 𝑏, (c) reflected at 𝑐, (d) transmitted at point 𝑐 into ray 𝑟2 .
𝑛 −𝑛
(a) At point 𝑎, 𝓇 = 𝑛1 +𝑛2 = −0.2, so 4% of the light intensity is reflected into ray 𝑟1 while the remaining 96%
1 2
is transmitted into the thin film.
𝑛2 −𝑛3
(b) At point 𝑏, 𝓇 = = 0.2, so 4% of the transmitted 96% i.e. 3.84% is reflected back from 𝑏 towards
𝑛2 +𝑛3
𝑐.
𝑛 −𝑛
(c) At point 𝑐, 𝓇 = 𝑛2 +𝑛1 = 0.2, so 4% of the 3.84% i.e. 0.15% is reflected at 𝑐.
2 1

(d) At point 𝑐 the remaining 3.84% − 0.15% = 3.69% is transmitted into ray 𝑟2

From this example, we make two approximations:

 The reflected light at point 𝑐 is negligible. We ignore it.

 184

 Although the intensity of the reflected ray 𝑟2 is always less than reflected ray 𝑟1 , they are
approximately equal. We ignore the small difference and treat them as being equal.

What is then the total intensity of rays 𝑟1 and 𝑟2 travelling to the observer’s eye? You might be inclined to
add their intensities, getting a value of about 8%, but this would be incorrect! That would be the situation
for stream of particles, but light is a wave and therefore the reflected rays 𝑟1 and 𝑟2 interfere. Now if the light
waves of rays 𝑟1 and 𝑟2 are exactly in phase with one another they will interfere constructively, yielding a
bright reflection, whereas if they are exactly out of phase, they will interfere destructively and the region will
appear dark. Any intermediate phase difference will yield a brightness somewhere in between.
Consequently, what will be observed by the observer is determined by the phase difference between rays 𝑟1
and 𝑟2 . The problem is therefore to determine this phase difference.

Before we do so, we note the following very strange properties of thin film interference:

 In the case when 𝑟1 and 𝑟2 are totally out of phase, if we assume their intensities and therefore
amplitudes have negligible difference, they cancel. If not light is then reflected to the observer, and
the thin film is not absorbing, what happens to the light energy? The answer is that it passes through
the thin film into the medium with refractive index 𝑛3 i.e. all of the light is transmitted. From the
example above you might argue that our calculation showed only about 92% of the light passed
through, and you would be correct. However, that calculation did not take into account the
interference of rays 𝑟1 and 𝑟2 and is therefore wrong, as confirmed by experiments!
 What then of the case when 𝑟1 and 𝑟2 are completely in phase and interfere constructively? Taking
each ray to have an intensity of about 4% of the incident light (before we consider interference),
they each have a fraction 0.2 of the incident wave’s amplitude. If they interfere constructively, the
resultant amplitude of the light ray passing to the observer has amplitude relative to the incident
light of 0.4, giving a reflection of 16%. This is incredible, considering that if we ignore interference
each ray only has an intensity of 4%. It is a consequence of the wave nature of light, and the fact
that the intensity is proportional to the amplitude squared. If 16% of the incident light is reflected,
only 84% can be transmitted despite the calculations in our previous example. This is confirmed
experimentally.

14.4. Conditions for constructive or destructive interference

If a thin film is engineered correctly, it can act as an antireflective coating, eliminating any reflection! This has
many technological applications e.g. anti-glare glass or a layer to place over solar panels to ensure the
maximum light enters to generate electricity. A thin film can also be designed to increase the reflectivity,
which has been used to enhance the glitter of costume jewellery. We wish to calculate the conditions for
such phenomena. As stated before, the key variable is the phase difference between rays 𝑟1 and 𝑟2 .

Although we will consider only a particular wavelength, we do not assume that the incident light is
monochromatic: it may contain many wavelengths, and we could calculate our results for each possible
wavelength.

Two effects contribute to the phase difference between rays 𝑟1 and 𝑟2 , namely the different paths of the light
producing the two rays, and the possible different nature of reflections they undergo.
 185

 First consider the phase difference due to their different paths: we will prove later that the phase of
2𝐿𝑛2 cos 𝜃′
ray 𝑟2 exceeds that of ray 𝑟1 by an amount Δ𝜑 = (2𝜋). Here 𝜃′ is the refracted angle inside
𝑛 1 𝜆1
the thin film.
 Second consider the reflections: ray 𝑟1 is reflected at point 𝑎 while ray 𝑟2 is reflected at point 𝑏.
Depending on the values of the refractive indices 𝑛1 , 𝑛2 and 𝑛3 either of these, or maybe both, could
cause the reflected ray to be “turned upside down”, which is equivalent to a 𝜋 phase shift.

Both of these effects must be considered in any problem. Before looking at examples, let us derive the
formula for phase difference due to the different paths of 𝑟1 and 𝑟2 , using Figure 82.

n1 n2 n3

 z x
'
L
'
b
d Figure 82
y z x
  ' Ray paths in a thin film.
  a

The incident wave splits into two parts at point 𝑎. Since rays 𝑟1 and 𝑟2 travel equal distances in the same
medium towards the observer from points 𝑑 and 𝑐 respectively, we only need to calculate their phase
difference up to these points. So we determine the phase difference of ray 𝑟2 at point 𝑐 relative to ray 𝑟1 at
point 𝑑.

 Beginning with ray 𝑟1 : its path length from point 𝑎 to point 𝑑 is the distance 𝑦. If the angle of
𝑦
incidence is 𝜃, then angle 𝑑𝑐𝑎 is also 𝜃 and sin 𝜃 = 2𝑧. Also if 𝜃′ is the refracted angle then tan 𝜃′ =
𝑧
𝐿
. Combining these gives 𝑦 = 2𝑧 sin 𝜃 = 2𝐿 tan 𝜃′ sin 𝜃.
It is inconvenient to have a mixture of the angles of incidence and refraction in the same formula,
but we can use Snell’s law 𝑛1 sin 𝜃 = 𝑛2 sin 𝜃′ to write it only in terms of the refracted angle. Then
𝑛 2𝐿𝑛2 sin2 𝜃′
𝑦 = 2𝐿 tan 𝜃′ (𝑛2 sin 𝜃′) = 𝑛1 cos 𝜃′
. This path is followed in the medium of refractive index 𝑛1
1

2𝐿𝑛2 sin2 𝜃′
where the wavelength is 𝜆1 , so the number of full wavelengths along the path is 𝑁1 = 𝑛1 𝜆1 cos 𝜃′
.

 Now for ray 𝑟2 : its path length from point 𝑎 to point 𝑐 is the distance 2𝑥. If the angle of refraction is
𝐿 2𝐿
𝜃′, then cos 𝜃′ = 𝑥 and so 2𝑥 = cos 𝜃′. This path is followed inside the thin film of refractive index 𝑛2
 186

𝑛 1 𝜆1 2𝐿
where the wavelength is 𝜆2 = , so the number of full wavelengths along this path is 𝑁2 = ÷
𝑛2 cos 𝜃′
𝑛 1 𝜆1 2𝐿𝑛2
𝑛2
=𝑛 .
1 𝜆1 cos 𝜃′

2𝐿𝑛2 2𝐿𝑛2 cos 𝜃′

The difference in the number of full wavelengths is 𝑁2 − 𝑁1 = 𝑛 (1 − sin2 𝜃′) = . Each
1 𝜆1 cos 𝜃′ 𝑛 1 𝜆1
wavelength corresponds to a phase of 2𝜋, so the phase difference due to the different paths is

2𝐿𝑛2 cos 𝜃′
Δ𝜑 = (2𝜋). (100)
𝑛1 𝜆1

Be sure to remember this is not the only source of phase difference: there might also be phase changes due
to the reflections of the rays that need to be taken into account. These must be considered to get Δ𝜑tot .

Example: Soap bubbles

Blowing soap bubbles can be very entertaining, and their properties are very interesting. Here we consider
whether the bubble will appear bright or dark for reflected light and what is the source of the colours in the
bubble. First note that we only consider one side of the bubble as a thin film. We also ignore the fact that it
is curved and treat it as a flat thin film of soapy water between two layers of air. Therefore our model is
better suited to the soap film shown in Figure 80 than a real bubble. We assume the soap does not affect the
refractive index of the water much, and take this to be 𝑛2 = 4/3. Since there is air both outside and inside
the bubble, we have 𝑛1 = 1 and 𝑛3 = 1.
Because 𝑛2 > 𝑛1, ray 𝑟1 will experience a 𝜋 phase shift when undergoing reflection. But since 𝑛3 < 𝑛2 ray 𝑟2
will not undergo any phase shift as a result of reflection. Since we are computing the phase difference of ray
𝑟2 relative to ray 𝑟1 we always add the reflection phase shift of ray 𝑟2 and subtract that of ray 𝑟1 , giving
2𝐿𝑛2 cos 𝜃′ 2𝐿𝑛2 cos 𝜃′ 1
Δ𝜑tot = (2𝜋) + 0 − 𝜋 = ( − ) (2𝜋).
𝑛1 𝜆1 𝑛1 𝜆1 2
All that remains is to interpret this result. To simplify matters, assume the incident light falls perpendicularly
onto the interface, i.e. the incident and refracted angles are zero. Then cos 𝜃′ = 1 and also substituting the
8𝐿 1
refractive indices gives Δ𝜑tot = (3𝜆 − 2) (2𝜋).
1

 For a very thin bubble 𝐿 → 0 and Δ𝜑tot → −𝜋, which corresponds to destructive interference, and the
bubble appears dark for reflected light, irrespective of the wavelength.
8𝐿 1 3𝜆1
 To obtain constructive interference, we can set Δ𝜑tot = 0, giving 3𝜆 = 2 or 𝐿 = 16
. If you were given
1
the wavelength 𝜆1 , you could calculate the thickness for which the bubble would strongly reflect the
light.
8𝐿 1 3𝜆1
 A second dark region will occur when Δ𝜑tot = 𝜋, or (3𝜆 − 2) (2𝜋) = 𝜋 which gives 𝐿 = 8
.
1

Using the expression for Δ𝜑tot one can calculate whatever is required, remember only that bright reflections
correspond to constructive interference for which Δ𝜑tot = 0, 2𝜋, 4𝜋, etc (any even multiple of 𝜋) while dark
 187

(or no) reflection corresponds to destructive interference for which Δ𝜑tot = 𝜋, 3𝜋, 5𝜋, etc (any odd multiple
of 𝜋, maybe even −𝜋)
Before leaving this example, note that the thicknesses for constructive or destructive interference depend
on the wavelength. If white light falls on the bubble, depending on the wavelength some colours may have
destructive interference while others have constructive interference for a given bubble thickness. This gives
rise to the different colours seen on the bubble. The actual colour seen depends on the relative amounts of
blue, green and red light reflected.

Example: Antireflective coatings

Some light is always reflected from a single interface, the amount being given by equation (99) if we assume
normal incidence. In some instances, we wish to transmit as much of the light as possible i.e. eliminate
reflection. In order to do this, we deposit a thin transparent film onto the surface of the material in order to
create thin film interference. If the film is chosen carefully we can reduce the reflectivity, having thus created
an antireflective coating.
In such a system we have light incident from air (𝑛1 = 1) into the antireflective thin film (𝑛2 ) and then into
the material of interest (𝑛3 > 1). As a definite example, suppose we wished to deposit an antireflective
coating on a glass lens: we are only considering one side of the lens, which is initially an air-glass interface.
After we deposit an antireflective coating on the glass it will become an air-coating-glass structure, with 𝑛3 =
3/2. (The back surface of the lens will also need its own separate antireflective coating, which we don’t
discuss further here.) We need to specify the thickness of the thin film coating as well as its refractive index
𝑛2 .
(a) The refractive index: For an antireflective coating to work well, it is important that the rays 𝑟1 and 𝑟2
interfere destructively, requiring a total phase difference of an odd multiple of 𝜋, but also equal amplitudes.
If the amplitudes are not equal, some reflected wave will remain even of the phase difference is chosen
correctly. Earlier in the chapter we gave an example suggesting the rays 𝑟1 and 𝑟2 had similar amplitude, but
did not prove it in general. In fact, their amplitudes can sometimes vary considerably, so we derive a condition
for the value of antireflective thin film refractive index (𝑛2 ) to ensure the amplitudes of these rays are close
to one another:
𝑛1 −𝑛2
The reflection amplitude ratio for ray 𝑟1 reflected from the front surface is 𝑛1 +𝑛2
, while the reflection
𝑛 −𝑛 𝑛 −𝑛
amplitude ratio for ray 𝑟2 reflected from the back surface is 𝑛2 +𝑛3. If these are set equal one gets 𝑛1 +𝑛2 =
2 3 1 2
𝑛2 −𝑛3
𝑛2 +𝑛3
which leads to the condition 𝑛2 = √𝑛1 𝑛3. In most cases 𝑛1 = 1 and this simplifies to 𝑛2 = √𝑛3 . Hence
for glass with 𝑛3 = 3/2 , the ideal antireflective coating material would have 𝑛2 = √3/2 ≈ 1.225.
Unfortunately there are no practical coating materials with this refractive index, and MgF2 with refractive
index 1.38 is often used.
(b) The required thickness: For a air-coating-glass antireflective coating one has 𝑛1 < 𝑛2 < 𝑛3 , and under
these conditions there will be phase changes for the reflection of both ray 𝑟1 on the front surface and ray 𝑟2
on the back surface, so the total phase difference will be
2𝐿𝑛2 cos 𝜃′ 2𝐿𝑛2 cos 𝜃′
Δ𝜑tot = (2𝜋) + 𝜋 − 𝜋 = ( ) (2𝜋).
𝑛1 𝜆1 𝑛1 𝜆1
 188

2𝐿𝑛2
Assuming the light is incident normally (perpendicularly), this reduces to Δ𝜑tot = (𝑛 ) (2𝜋) . For an
1 𝜆1
antireflective coating we must have destructive interference, so the thinnest possible coating layer will
2𝐿𝑛2 𝑛 1 𝜆1 𝜆2
correspond to Δ𝜑tot = 𝜋. Then ( ) (2𝜋) = 𝜋 or 𝐿 = = . A coating layer of thickness equal to one
𝑛 1 𝜆1 4𝑛2 4
quarter of the wavelength of the light inside the coating is therefore appropriate for an antireflective coating.
Thicker layers could also work, e.g. corresponding to Δ𝜑tot = 3𝜋, 5𝜋 etc, but would use more material and
take longer to produce.
The thickness of the required antireflective coating depends on the wavelength, and so the reflection of only
one wavelength can be completely eliminated, while the reflection of nearby wavelengths will only be
reduced. For practical purposes, we generally choose to eliminate the wavelength of light in the centre of
the visible spectrum, which is the yellow-green portion with 𝜆1 = 550 nm. Then the thinnest antireflective
𝑛 1 𝜆1 𝜆2
coating 𝐿 = 4𝑛2
= 4
is about 100 nm. With such a coating the reflection of the yellow-green light is
practically eliminated, but that of the longer wavelength red light and the shorter wavelength blue light is
only reduced. The small amount of reflected red+blue light give such coatings the purple colour often seen
reflected from camera lenses and people's glasses. Complicated multi-layer antireflective coatings using
several layers have been designed to work well over a range of wavelengths.

14.5. Why “thin” films – the importance of coherence

This title of this chapter refers to “thin” films, but why must they be thin? Nowhere in the discussion have
we seen any reason why the thickness 𝐿 cannot take on any value.

However, we do not see colours in a window pane (thickness of an few mm) whereas we do observe them in
a soap bubble, so the films must indeed be thin to make these observations. The reason for this has to do
with the quality or coherence length of the light. In the chapter on sinusoidal waves Figure 64 was used to
illustrate the concept of coherence: we can consider the coherence length of the light to be the distance the
wave travels before there is a random change in its phase constant. Now for rays 𝑟1 and 𝑟2 reflected from
the front and back of a film to interfere in a consistent non-random way, the difference in their paths should
be less than the coherence length of the light being reflected. Because ray 𝑟2 goes back and forth through
the film, a rough mathematical criterion to allow non-random interference is that the film cannot be thicker
than half the coherence length of the light.

A detailed discussion of coherence is beyond the scope of this work, but we note that the coherence length
of a light source can be estimated by
𝑐 𝜆2ave
𝐿𝑐 = = (101)
∆𝑓 Δ𝜆
where ∆𝑓 and Δ𝜆 are the full-width-at-half-maximum or FWHM values of the frequency or wavelength
distributions. The coherence length of direct unfiltered sunlight is a fraction of a micrometre (roughly equal
to the wavelength of the light), while that that from gas discharge lamps is in the range of millimetres to
centimetres, and that from lasers can be many kilometres. A “thin” film is then one having a thickness less
than half the coherence length of the light which illuminates it, and films have to be very thin to display
interference effects with sunlight or everyday lightbulbs.
 189

14.6. Problems

1. Light of wavelength 585 nm is incident perpendicularly on the thin soapy water film of a soap bubble,
which has index of refraction 𝑛 = 4/3 and thickness 1.21 m. (a) Will there be constructive or destructive
interference of the light reflected from the outer and inner surfaces of the soap film? (b) Will the bubble
appear bright or dark at this point?

Figure 83

Antireflective coating.

2. As illustrated in Figure 83, a glass lens with a refractive index of 1.50 is coated with a thin layer of
magnesium fluoride (MgF2), which has refractive index of 1.38, to act as an anti-reflective coating. (a) What
is the thinnest coating that can be applied so that when blue light of wavelength 400 nm strikes the MgF2
coated lens perpendicularly, the light reflected back from the MgF2 surface and from the MgF2-glass interface
interfere destructively? (b) What colour will the lens appear in reflected white light? (Hint: you may have to
look up about complementary colours.)

3. The rhinestones used in costume jewellery are glass with an index of refraction 1.50. To make them more
reflective, they are often coated with a layer of silicon monoxide of refractive index 2.00. What is the
minimum coating thickness required to ensure that light of wavelength 560 nm and of perpendicular
incidence will be reflected from the two surfaces of the coating with fully constructive interference?

4. In the chapter an antireflective coating for the front surface of a lens was discussed. Consider the situation
for the back surface of the lens and show that although some of the physics is different, the same
antireflective coating is required i.e. both surfaces of the lens are coated in the same way.

5. In the text, when there was a reflection phase change we took it as +𝜋. This could also have been taken
as – 𝜋, and although it is not recommended, one can even switch the signs within different parts of the same
problem. Explain why this is possible.

6. Calculate the coherence lengths of the following light sources: (a) The green light from a mercury discharge
lamp, having central wavelength at 546 nm and a line width of 0.025 nm. (b) The orange light from a krypton
86 isotope gas discharge lamp, having central wavelength at 606 nm and a line width of only 0.00047 nm. (c)
A CO2 laser emitting infrared light of wavelength 10.6 micrometres with line width of 10-5 nm.
 190

Figure 84

The spectrum of
sunlight.

6. Consider Figure 84 and estimate the coherence length of sunlight.

7. The red light of wavelength 643.8 nm from a cadmium discharge lamp has a coherence length of 30 cm.
Calculate its linewidth.

8. Two light filters are used to transmit yellow light centred at the wavelength of 590 nm. One filter has a
“broad” transmission width of 100 nm, while the other has a “narrow” pass band of 10 nm. Which filter would
be best to use in an interference experiment? Motivate your answer by calculating the coherence length of
the light from each filter.

9. Consider a light source which emits a (small) range of frequencies Δ𝑓. By taking the derivative of the
𝑐 𝑐
equation 𝑓 = , show that Δ𝑓 = Δ𝜆 and hence that both forms of equation (101) for the coherence length
𝜆 𝜆2
are equivalent.
 191

15. The Michelson interferometer

15.1. Introduction

An interferometer is a device that can be used to measure distances and wavelengths with great accuracy by
means of interference effects. In any interference experiment, light waves must originate from a point and
travel different paths before recombining at a point. At the recombination point, there are two possibilities:

 The path difference has been greater than the coherence length of the light. Then the interference
is random and the average intensity observed is just the intensity of the source.

 The path difference is less that the coherence length of the light. Then the two waves interfere in a
systematic way and depending on the phase difference we can get constructive interference or
destructive interference, or usually something between these extremes.

There are over a dozen different named types of interferometers47 but the most famous is the one first
introduced by Michelson in 1881. It was used to search for small changes for the speed of light, and the
unsuccessful results later contributed to Einstein's special theory of relativity that the speed of light (in
vacuum) is constant and independent of the motion the source or observer. It was also used to measure the
hyperfine energy structure of atoms (i.e. the fine structure of atomic spectra) which was used to test the
theory of quantum mechanics. It is currently being used (with other techniques) to measure polarization
mode dispersion (PMD) in optical fibres that carry the signals for global telecommunications and the internet
and is also the basis for the technique used to detect gravitational waves with LIGO (Laser Interferometer
Gravitational-wave Observatory).

15.2. Michelson's Interferometer

A schematic view of a Michelson interferometer is shown in Figure 85. It consists of an extended coherent
light source (e.g. an illuminated ground glass plate), a beam-splitter 48 and two mirrors, one of which is
movable. The light from the source which is reflected at the beam-splitter moves to mirror 𝑀2 and back
again, and light from the source which was transmitted at the beam-splitter moves to mirror 𝑀1 and back
again. Each of these waves is then again divided at the beam-splitter, and have half of their intensities moving

47
The index of Born and Wolf's Principles of Optics lists: Bates, Dowell, Dyson, Fabry-Perot, Fizeau, Hanbury Brown-
Twiss, Jamin, Kösters, Lummer-Gehrke, Mach-Zehnder, Michelson, Rayleigh, Sirks-Pringsheim, Twyman-Green.
48
A beam-splitter divides a wave by transmitting some of its intensity and reflecting the remaining part. This process is
commonly known as amplitude division, and unless otherwise stated it may be assumed that 50% reflection and 50%
transmission occurs. A beam-splitter is usually made by coating a glass plate with a very thin metal layer, too thin to be
completely reflecting.
 192

towards the detector (which may be for instance a camera, or a telescope and the observer's eye). These
waves interfere and an interference pattern can be observed.

Figure 85

Schematic view of a
Michelson
interferometer.

Note that the reflecting side of the beam-splitter was placed on the side away from the source. Thus the light
moving to the detector from the source after having been reflected from mirror 𝑀2 passes through the glass
of the beam-splitter three times, whereas the light reflected from mirror 𝑀1 only passes through the beam-
splitter once. To compensate for this, a compensating plate of glass 𝐶 having the same thickness as the beam-
splitter is placed in front of mirror 𝑀1 .

Now, if the two mirrors are precisely equidistant from the reflecting surface of the beam-splitter, the two
waves travel exactly the same distance apart before recombining. One might therefore expect them to be in
phase: however, the phase shifts due to reflections have not yet been considered. Each of the waves arriving
at the detector has undergone a pair of reflections. One of each pair occurs at a mirror, which is an interface
from air to glass, and so causes a phase shift of 𝜋 - there is no resulting relative phase shift. The other
reflections occur at the beam-splitter: the light from the source being reflected towards 𝑀2 is reflected from
an interface from glass to air and thus has no phase shift. But the light from mirror 𝑀1 being reflected
towards the detector is reflected from an interface of air to glass and thus undergoes a 𝜋 phase shift. The net
result is that there is a 𝜋 phase shift between the waves when they recombine due to reflection effects.49 So
if the mirrors are precisely equidistant from the reflecting surface of the beam-splitter, the two waves
recombine completely out of phase and then due to destructive interference the intensity at any point on
the detector is zero.

49
If the compensating plate were fused to the beam-splitter, so that the half reflecting layer was in the centre of a thick
piece of glass, the situation would be completely symmetrical and there would be no relative phase shift of the waves.
 193

If the mirror distances are not equal, the observer sees a ring pattern of dark and light fringes. For
monochromatic light the number of bright rings will be equal to twice the difference in the mirror distances
(from the beam-splitter) divided by the wavelength of the light, which we shall derive. To explain both the
ring patterns and the number of rings, we show that the Michelson interferometer can be interpreted as a
virtual thin film interference phenomenon. Consider the observer looking into a Michelson interferometer
along the direction of the detector (i.e. the eye is being used as a detector.) The mirror 𝑀2 will be seen
through the beam-splitter, and also a virtual image of the mirror 𝑀1 due to reflection in the beam-splitter. If
mirror 𝑀2 has been moved backwards from the position of equidistance by a distance 𝐿, the result is as
shown in Figure 86. Note also that in this picture, the light source appears to be on the same side as the
detector. This is a linearized picture of what is happening in the Michelson interferometer (Figure 85), with
the source and mirror 𝑀1 rotated anticlockwise around point O (the reflection point in the beam-splitter) by
90°.

M2
𝐿

Image of M1



Figure 86

Michelson
S interferometer analysed
as a thin air film
between mirrors.
Lens of
detecting system

Image of S

Light from the source, making an angle 𝜃 with the linearized axis of the interferometer is reflected at both
mirrors. The picture we have is identical to that of thin film interference, with the “thin film” being an air gap
between two mirrors, and with no refraction of the light occurring as it enters the gap. The beam-splitter and
compensator do not occur in this picture: the whole idea is to ignore the changes in direction that the beam-
 194

splitter introduces and draw everything along a straight line. Although the beam-splitter has been ignored,
we must remember about the 𝜋 phase shift due to reflection effects.

From the theory of thin films, the phase difference of the two reflected waves due to the different paths
lengths is given by equation (100) with 𝑛2 and 𝑛1 both being the refractive index of air and 𝜃 ′ = 𝜃: then the
2𝐿 cos 𝜃
equation reduces to Δ𝜑 = (2𝜋). This, together with the 𝜋 phase shift due to reflections, gives a total
𝜆air
phase difference of

2𝐿 cos 𝜃 1
Δ𝜑tot = ( + ) (2𝜋) (102)
𝜆air 2

and this can be used to explain the ring pattern. Remember that 𝐿 represents the distance that the movable
mirror 𝑀2 has been moved away from the position when the mirrors are equidistant from the beam-splitter.
Then:

 For a given value of 𝜃, there is a fixed value of Δ𝜑tot and thus a fixed intensity. As the angle changes
the intensity may change. This leads to a ring pattern.

 Very far from the centre of the pattern, 𝜃 → 90° and so Δ𝜑tot → 𝜋. This means that at large angles
the total phase difference of 𝜋 results in destructive interference and this outer region always
remains dark.

 The maximum total phase difference occurs at the centre of the image (for 𝜃 = 0). It is given by
2𝐿 1 2𝐿 1
( + ) 2𝜋 and if one defines 𝑁 = it becomes (𝑁 + ) 2𝜋 . If 𝑁 is any integer this total phase
𝜆air 2 𝜆air 2
difference at the centre will be an odd multiple of 𝜋 and so intensity at the centre will be dark.
 If 𝑁 = 1, the phase difference will decrease from 3𝜋 at the centre to 𝜋 at large angles. It will pass
through Δ𝜑tot = 2𝜋 which corresponds to a single bright ring.
 If 𝑁 = 2, the phase difference will decrease from 5𝜋 at the centre to 𝜋 at large angles. It will pass
through Δ𝜑tot = 4𝜋 and 2𝜋 which corresponds to two bright rings.
 If 𝑁 = 3, the phase difference will decrease from 7𝜋 at the centre to 𝜋 at large angles. It will pass
through Δ𝜑tot = 6𝜋 and 4𝜋 and 2𝜋 which corresponds to three bright rings.

2𝐿
 Clearly the value of 𝑁 gives the number of bright rings. Notice that we defined 𝑁 = 𝜆 i.e. twice the
air
mirror displacement divided by the wavelength of the light, so we have derived the result stated
earlier. Also note that the number of rings increases as the mirror displacement increases.

 Note that as the mirror displacement increases, new rings grow outwards from the centre. If 𝑁
increases by one, the new image formed is similar to what it was before, except that each ring has
moved outwards one position, with an extra ring on the outside and a new one replacing the centre
𝜆air
one. This corresponds to ∆𝑁 = 1 and thus ∆𝐿 = 2
. Be aware that a circular fringe pattern is only
observed if the Michelson interferometer is set up very accurately. If it is misaligned, the fringes are
still seen, but appear as straight or slightly curved lines. These lines shift position as the mirror is
𝜆air
displaced, and a mirror displacement of ∆𝐿 = 2
corresponds to a shift in the image of exactly one
fringe.
 195

15.3. Calculating the refractive index of a material

A shift in the fringe pattern can also be caused by the insertion of a thin transparent material into the optical
path of one the mirrors (while both are kept fixed in position). If a material with thickness 𝑑 and index of
refraction 𝑛 is inserted in the optical path associated with mirror 𝑀2 , then the number of full wavelengths
along this path will change. Since the light passé through the material twice on its way to 𝑀2 and back, the
2𝑑
number of full wavelengths inside the material will be where 𝜆mat is the wavelength of the light inside
𝜆mat
𝑛 2𝑑
the material. Since 𝑛air 𝜆air = 𝑛mat 𝜆mat , the number of full wavelengths inside the material is 𝑛 mat𝜆 .
air air

2𝑑
Originally, before the material was inserted, the number of full wavelengths in the same space was just 𝜆 ,
air
𝑛mat 2𝑑 2𝑑 2𝑑 𝑛mat
so the change in the number of full wavelengths is 𝑛air 𝜆air
− 𝜆air
= (
𝜆air 𝑛air
− 1). Each full wavelength
change corresponds to a phase change of 2𝜋 or the shift of one fringe in the interferometer image. Therefore
if Δ𝑁 is the number of fringe shifted then

2𝑑 𝑛mat
Δ𝑁 = ( − 1). (103)
𝜆air 𝑛air

15.4. Investigating the fine structure of spectral lines

Consider a Michelson interferometer illuminated by two different monochromatic light sources

simultaneously. There will therefore be two ring patterns, one for each wavelength. At certain angles the
bright rings from each of the two wavelengths will coincide with each other (and so will the dark rings), so
that the fringes remain easily visible. But at other angles the bright ring of one wavelength will correspond
to the dark ring of the other wavelength, and the fringes will thus be difficult to see. Figure 87 illustrates this
idea.

Figure 87

Superimposed ring
patterns.
 196

Each colour (i.e. wavelength) has its own total phase difference at a point on the image. For the first
2𝐿 cos 𝜃 1 2𝐿 cos 𝜃 1
wavelength this is ( 𝜆1
+ 2) (2𝜋) and for the second wavelength ( 𝜆2
+ 2) (2𝜋), with 𝜆1 and 𝜆2
being the two wavelengths of the light in air. Therefore the difference between these total phase differences
for the two colours is

1 1 ∆𝜆
Δ(Δ𝜑tot ) = 2𝐿 cos 𝜃 { − } (2𝜋) ≈ 2𝐿 cos 𝜃 { 2 } (2𝜋). (104)
𝜆1 𝜆2 𝜆ave

 If this difference between the total phase differences is zero, or 2𝜋, or 4𝜋 (or any even number times 𝜋)
then the type of interference happening for each wavelength will be matched i.e. both will be having the
same type of interference, be it constructive, or destructive or whatever. Since the images match, the
fringe pattern will be clear.
 If this difference between the total phase differences is 𝜋, or 3𝜋 (or any odd number times 𝜋) then the
type of interference happening for each wavelength will be opposite i.e. when one undergoes
constructive interference at a point, the other will undergo destructive interference at the same point
(and vice-versa). Since the constructive interference positions for one wavelength always lie on the
destructive interference positions of the other wavelength, the fringe patterns will not be visible and the
image will be uniform. Note: the image will not be dark, and there is not only destructive interference
between the two wavelengths.

∆𝜆
For the interference bands to be clearly visible 2𝐿 cos 𝜃 {
𝜆2ave
} (2𝜋) = 𝑚(2𝜋) where 𝑚 is an integer, so

∆𝜆
2𝐿 cos 𝜃 { }=𝑚
𝜆2ave

From one situation where the bands are clearly visible, to the next, ∆𝑚 = 1. If we keep the mirror positions
fixed and consider a change in angle, then the angular spacing from one region where the fringes are clearly
𝜆2
visible to the next is ∆(cos 𝜃) = 2𝐿ave
∆𝜆
. But if we only look at the centre (cos 𝜃 = 1) and shift a mirror, then
the mirror shift necessary to move from one position where the bands are clearly visible to the next position
𝜆2ave
where the bands are clearly visible is ∆𝐿 = .
2 ∆𝜆
 197

15.5. Problems

1. A Michelson interferometer is used in an experiment. The wavelength of the light is 550 nm. (a) By what
distance has the movable mirror been displaced if there is a shift of 80 fringes in the interference pattern?
(b) A new, different light source is now used. When the movable mirror is displaced by 0.233 mm, a shift of
792 fringes occurs. What is the wavelength of the new light source?

Figure 88

Use of a Michelson
interferometer to
measure the refractive
index of air.

2. Figure 88 shows an airtight chamber 5.0 cm long, which is placed in one path of a Michelson interferometer
using light of wavelength 500 nm. As the air is pumped out of the chamber (leaving a vacuum) the
interference pattern shifts with 60 fringes. Using this information, calculate the refractive index of air. (Hint:
The refractive index of vacuum is exactly 1, and for air it is slightly larger than 1. The refractive index cannot
be negative, so you may have to adapt the equation derived in this chapter slightly.)
 198

16. Two-point interference

16.1. Introduction

In many textbooks, the chapter discussing this work is headed “Double slit interference”, and in it we shall
indeed discuss Young's famous double slit experiment. However, it has been named differently to emphasize
that the interference pattern we are discussing is just that of two coherent point sources. An example for
water waves is the ripple tank photograph shown in Figure 89.

Figure 89 Two point interference of water waves in a ripple tank.

Consider two coherent wave sources of equal amplitude, in phase with each other, a distance 𝑑 apart. For
any chosen point around the sources, there is a path length difference from the two sources. If this path
 199

length difference is equal to an integral number of wavelengths, then the waves from each source arrive at
that point in phase and interfere constructively. At other points this path length difference is equal to an
“integer and a half” number of wavelengths, and then the waves from each source arrive at that point
completely out of phase and interfere destructively. A simple plot of the wavefronts (Figure 90) from each
source at some instant shows that linear regions having destructive and constructive interference radiate out
from the region between the sources: these are the so called nodal and antinodal lines. Although these
regions are linear far from the sources, they bend near the sources and we shall show that they are in fact
hyperbola.

Figure 90

Interference of waves from two

point sources.
 200

16.2. Mathematical analysis

Consider two point sources separated by a distance 𝑑, and place the axes as shown in Figure 91. For any point
𝑃, the difference in path lengths from the sources to 𝑃 is given by 𝛤 = 𝑆2 𝑃 − 𝑆1 𝑃. Using the theorem of
Pythagoras gives

𝑑 2 𝑑 2
𝛤 = √𝑥 2 + (𝑦 + ) − √𝑥 2 + (𝑦 − ) .
2 2

Those that enjoy mathematical challenges can show that this can be rewritten in the form

𝑦2 𝑥2 1
2
− 2 2
= .
𝛤 𝑑 −𝛤 4

For any choice of 𝛤 (up to the maximum possible path length difference 𝑑) this is the equation of a hyperbola,
and some examples are shown in Figure 92. This means that all points of equal path length difference,
including both nodes and antinodes, lie on hyperbolas.

Y P(x,y)

S1
d/2
O X
d/2
S2 

Figure 91 Path length difference from two point sources.

 201

Figure 92

Hyperbolas corresponding to
particular path length differences.
Both the nodal and antinodal
curves are hyperbolas.

1
When 𝑥 and 𝑦 become large compared to 𝑑 (i.e. far from the wave sources), the term 4 on the right hand
side of the hyperbola becomes negligible, and the equation can then be rewritten as

𝑦2 𝑥2 𝛤
= or 𝑦 = ± 𝑥.
𝛤 2 𝑑2 − 𝛤 2 √𝑑2 − 𝛤 2

These represent lines of equal path difference. Except near the origin these lines are an excellent
approximation of the hyperbolas of equal path difference.

P(𝑥, 𝑦)
Y

S1 y

𝑑/2
O 𝜃
X
𝑥
𝑑/2
S2

Figure 93 Geometry used in analysis of two point interference pattern.

 202

𝛤
Notice that the lines pass through the origin and that they have slope . But the slope of any line is also
√𝑑 2 −𝛤2
given by tan 𝜃 where 𝜃 is the angle the line makes with the horizontal axis (Figure 93). Therefore tan 𝜃 =
𝛤
, and so
√𝑑 2 −𝛤2

𝛤
sin 𝜃 = or 𝛤 = 𝑑 sin 𝜃.
𝑑

Thus, for any point 𝑃 far from the origin, the path difference between the waves from the two sources
arriving at 𝑃 is given by 𝛤 = 𝑑 sin 𝜃 , where the angle 𝜃 is the angle that the line 𝑂𝑃 makes with the
(horizontal) x-axis.

Using this expression for the path length difference, we may conclude the following:

 A nodal line of destructive interference occurs whenever the path length difference equals an
“integer and a half” number of wavelengths. Thus the angles 𝜃 at which the nodal lines occur are
1
given by 𝑑 sin 𝜃 = (𝑚 + ) 𝜆 where 𝑚 is an integer.
2

 An antinodal line of constructive interference occurs whenever the path length difference equals an
integral number of wavelengths. Thus the angles 𝜃 at which the antinodal lines occur are given by
𝑑 sin 𝜃 = 𝑚𝜆 where 𝑚 is an integer.

16.3. Young's double slit interference experiment

Despite some prominent scientists such as Hooke, Huygens and Euler suggesting a wave theory for light, most
scientists during the 18th century preferred the particle (or corpuscle) theory of Newton. In 1803 Young
presented his experiment to the Royal Society,50 showing the interference of light. Even then, some scientists
still were not convinced of the wave nature of light until the experimental verification of the diffraction
theory of Fresnel about 15 years later. Young’s original experiment was not done with slits, but by dividing a
beam of sunlight passing through a pin-hole by a thin obstacle. Generally, however, it is described in terms
of two slits and we shall do the same.

Figure 94

Slit producing
cylindrical waves.

50
Thomas Young, Experimental Demonstration of the General Law of the Interference of Light, Philosophical Transactions of the
Royal Society of London vol. 94 (1804).
 203

We regard a thin slit as a source of cylindrical waves, as shown in Figure 94. A cross-section through these
cylindrical waves gives circular wave fronts, similar to those for water waves for a point source. A top view
(cross-section) of Young’s double slit experiment is shown in Figure 95.

Figure 95 Young’s double slit experiment.

In the experiment, incoherent light is incident on a single very narrow slit 𝑆0 , and then this illuminates two
more very narrow slits 𝑆1 and 𝑆2 equidistant from the first one. These two slits produce coherent in-phase
waves, which interfere in the region behind the slits in a similar way to water waves from two point sources
in a ripple tank. If a screen is held behind the double slit, a pattern of bright and dark fringes appears. The
bright fringes correspond to where constructive interference is occurring, and the dark fringes correspond to
where the interference is destructive. Note that if the original source of light is coherent, e.g. a laser, then
the initial single slit is unnecessary.

From the previous section we have already determined that the angles where bright bands will occur is given
1
by 𝑑 sin 𝜃 = 𝑚𝜆, while dark bands will occur for angles given by 𝑑 sin 𝜃 = (𝑚 + 2) 𝜆. Now we derive an
expression for the intensity (𝐼) of the fringe pattern on the screen including the positions between the
maxima and minima. If the setup is symmetrical, then the two interfering waves will have the same initial
amplitude (say 𝐴) and the two sources will be in phase. We shall neglect any possible decrease in the
amplitudes of the waves as they propagate and represent them by 𝑌[𝑥, 𝑡] = 𝐴 cos(𝑘𝑥 − 𝜔𝑡). At some point
𝑃 the disturbance due to the circular wave from source 𝑆1 will be 𝑌1 [𝑡] = 𝐴 cos(𝑘{𝑆1 𝑃} − 𝜔𝑡) while that
from source 𝑆2 will be 𝑌2 [𝑡] = 𝐴 cos(𝑘{𝑆2 𝑃} − 𝜔𝑡). Therefore the resultant disturbance at 𝑃 is

𝑌𝑃 [𝑡] = 𝑌1 [𝑡] + 𝑌2 [𝑡] = 𝐴(cos(𝑘{𝑆1 𝑃} − 𝜔𝑡) + cos(𝑘{𝑆2 𝑃} − 𝜔𝑡)).

𝑢−𝑣
Using the identity cos 𝑢 + cos 𝑣 = 2 cos (𝑢+𝑣
2
) cos(
2
) gives

𝑆1 𝑃 + 𝑆2 𝑃 𝑆1 𝑃 − 𝑆2 𝑃
𝑌𝑃 (𝑡) = 2𝐴 cos (𝑘 − 𝜔𝑡) cos (𝑘 ).
2 2
 204

The first cosine function oscillates with time, but the second is time-independent and forms part of the
amplitude. The quantity 𝑆1 𝑃 − 𝑆2 𝑃 is just the path length difference and so the amplitude at point 𝑃 is 𝐴𝑃 =
𝛤 2𝜋 𝜋𝑑 sin 𝜃
2𝐴 cos (𝑘 2 ). Substituting 𝑘 = 𝜆
and 𝛤 = 𝑑 sin 𝜃 gives 𝐴𝑃 = 2𝐴 cos ( 𝜆
), so if we define

𝜋𝑑 sin 𝜃
𝛽= (105)
𝜆

then 𝐴𝑃 = 2𝐴 cos 𝛽. The light intensity is proportional to the square of the amplitude, so 𝐼 ∝ 4𝐴2 cos2 𝛽.
But the amplitude of each of the original wave sources was 𝐴, so their intensity 𝐼0 ∝ 𝐴2 and

𝐼 = 4𝐼0 cos 2 𝛽 (106)

represents the intensity on the screen. Note that 𝛽 is only a defined quantity and not an physical angle that
occurs somewhere in the geometry of the problem. Using this expression for the intensity, we may conclude
the following:

 Maximum intensity (bright regions) occur when cos 𝛽 = ±1 which happens when 𝛽 = 𝑚𝜋 for
𝜋𝑑 sin 𝜃
integral 𝑚. Since 𝛽 = 𝜆
= 𝑚𝜋, these maxima occur when 𝑑 sin 𝜃 = 𝑚𝜆, which is what we
found before.
1
 Minimum (zero) intensity (dark regions) occur when cos 𝛽 = 0 which happens when 𝛽 = (𝑚 + 2) 𝜋
𝜋𝑑 sin 𝜃 1 1
for integral 𝑚. Then 𝛽 = 𝜆
= (𝑚 + 2) 𝜋 so these minima occur when 𝑑 sin 𝜃 = (𝑚 + 2) 𝜆 as
found before.

16.4. Problems

Band B (dark)
Central bright band (𝜃 = 0) Band A (bright)
90Experimental fringe pattern for Young's double slit experiment.
Figure 96

1. Figure 96 is a photograph of the light and dark bands produced by Young's double slit experiment using
monochromatic light from a sodium lamp (𝜆 = 590 nm). The distance between the double slits was 0.1 mm.
Find the angle 𝜃 corresponding to bands A and B as indicated.

2. In an experiment like Young's experiment, monochromatic and coherent green light of wavelength 550 nm
is used to illuminate two very narrow slits 7.70 m apart. The centre bright band (𝑚 = 0) is called the zero
order bright band. Calculate the deviation angle 𝜃 for the third order (𝑚 = 3) bright band.
 205

3. In a double slit arrangement the slits are separated by a distance equal to 100 times the wavelength of the
light passing through the slits. (a) What is the angular separation between the central maximum and the
adjacent maximum? (b) What is the distance between these maxima on a screen 50 cm away from the slits?

4. Light of wavelength 600 nm is incident normally on two parallel narrow slits separated by 0.60 mm. Use a
computer to plot the intensity pattern observed on a distant screen as a function of the angle 𝜃 for the range
of values 0 < 𝜃 < 0.0040 rad. Do the calculation in steps of 0.0001 rad.

5. Consider the central intensity peak from a Young's double slit experiment projected on a screen. The
intensity is a maximum when 𝜃 = 0. Show that the intensity drops to half its maximum value at an angle
𝜆
given by sin 𝜃 = 4𝑑. What is the angular FWHM of the central intensity peak?
 206

17. N-Point interference

17.1. Introduction

In most textbooks, this chapter is called “diffraction gratings”, and in it we shall indeed discuss these very
important pieces of optical apparatus. However, we have named it differently to emphasize that the
interference pattern we are discussing is just that of 𝑁 coherent point sources that are equally spaced. Thus,
a better name for these objects would have been “interference gratings”. It is rather unfortunate that the
word ‘diffraction’ is used in their name, since the concept of diffraction is not important in understanding
their main properties and will only be studied in the next chapter. We will just refer to them as grating.
Gratings were invented by the American astronomer David Rittenhouse around 1785. However, they were
independently developed and made well known by Joseph von Fraunhofer some years later. Gratings are
used to disperse light into its colours (as Newton did with a prism) and to measure the wavelengths present.

Figure 97

Transmission grating with 5 slits.

A transmission grating is a device with a very large number (𝑁) of slits which are all equally spaced, having a
distance 𝑑 from one other. One also gets reflective gratings, which have reflective surfaces separated by non-
reflective regions instead of slits. The theory in this chapter is applicable to both types. Gratings can have up
to several thousand slits or lines per millimetre, but in Figure 97 a grating with only 5 slits is shown. It is
similar to the double slit problem dealt with in the previous chapter, but with additional slits. Again we
assume the slits act as sources of cylindrical waves. Viewed as a cross-section from above, the interference
pattern of a grating can be interpreted as arising from 𝑁 equally spaced point sources. We shall also assume
the light from all the sources is coherent and in phase at their origins, and represented by waves all having
the same amplitude 𝐴 which does not decay as the waves propagate outwards.
 207

17.2. Mathematical analysis of a grating

Y
P(x,y )

S1
d


S2
d
S3
(N-1)d d
x X
d

d
S
N

Figure 98 𝑵 equally spaced point sources, representing a grating. 𝑵 = 6 in this figure.

Consider 𝑁 point sources each separated by a distance 𝑑, and place the axes as shown in Figure 98. We
represent the wave from each source by an equation of the form 𝑌[𝑥, 𝑡] = 𝐴 cos(𝑘𝑥 − 𝜔𝑡). At some point
𝑃 the disturbance due to the circular wave from source 𝑆𝑖 will be 𝑌𝑖 [𝑡] = 𝐴 cos(𝑘{𝑆𝑖 𝑃} − 𝜔𝑡) and the
resultant wave at 𝑃 is due to the waves from all the sources is

𝑁 𝑁

𝑌𝑃 [𝑡] = ∑ 𝑌𝑖 [𝑡] = ∑ 𝐴 cos(𝑘{𝑆𝑖 𝑃} − 𝜔𝑡).

𝑖=1 𝑖=1

Since the path length difference between each successive source is 𝛤 = 𝑑 sin 𝜃, the path length 𝑆𝑖 𝑃 can be
written as

𝑆𝑖 𝑃 = 𝑆1 𝑃 + (𝑖 − 1)𝑑 sin 𝜃

where 𝑆1 𝑃 is the shortest path length and 𝑑 is the distance between the slits. Therefore

𝑁 𝑁−1

𝑌𝑃 (𝑡) = 𝐴 ∑ cos(𝑘{𝑆1 𝑃 + (𝑖 − 1)𝑑 sin 𝜃} − 𝜔𝑡) = 𝐴 ∑ cos({𝑘{𝑆1 𝑃} − 𝜔𝑡} + 𝑗{𝑘𝑑 sin 𝜃}).
𝑖=1 𝑗=0
 208

To make the summation we use the identity

𝑁−1
sin(𝑁𝑣2) 𝑣
∑ cos(𝑢 + 𝑗𝑣) = cos (𝑢 + (𝑁 − 1)2).
sin(𝑣2)
𝑗=0

By comparison 𝑢 = 𝑘{𝑆1 𝑃} − 𝜔𝑡 and 𝑣 = 𝑘𝑑 sin 𝜃, so

sin(𝑁𝑘𝑑 sin
2
𝜃
) 𝑘𝑑 sin 𝜃
𝑌𝑃 (𝑡) = 𝐴 cos (𝑘{𝑆1 𝑃} − 𝜔𝑡 + (𝑁 − 1) ).
sin(𝑘𝑑 sin
2
𝜃
) 2

The only time dependence occurs in the cosine term, which therefore oscillates between +1 and -1 as time
goes by. The remaining part of the expression represents the amplitude of the resultant oscillation at point
𝑃, which is

sin(𝑁𝑘𝑑 sin
2
𝜃
) sin(𝑁𝛽)
𝐴𝑃 = 𝐴 =𝐴
sin(𝑘𝑑 sin
2
𝜃
) sin 𝛽

𝑘𝑑 sin 𝜃 2𝜋 𝑑 sin 𝜃 𝜋𝑑 sin 𝜃

since 2
= 𝜆 2
= 𝜆
which we already defined as 𝛽. The light intensity is proportional to the
sin2(𝑁𝛽)
square of the amplitude, so 𝐼 ∝ 𝐴2 sin2 𝛽
. But the amplitude of each of the original wave sources was 𝐴, so
their intensity 𝐼0 ∝ 𝐴2 and
sin2(𝑁𝛽)
𝐼 = 𝐼0 (107)
sin2 𝛽

represents the intensity on the screen. Figure 99 shows a plot of the intensity given by this formula as a
function of 𝛽 for a few different values of 𝑁. Remember that 𝛽 is just a defined quantity, not any particular
angle related to the geometry of the problem.

Figure 99 Intensity as a function of 𝜷 for 𝑵-slit interference. The case 𝑵 = 𝟐 corresponds to a double slit.
 209

Irrespective of the quantity 𝑁, there are always maxima at the positions 𝛽 = 𝑚𝜋, which correspond to
𝜋𝑑 sin 𝜃
= 𝑚𝜋 or 𝑑 sin 𝜃 = 𝑚𝜆 (the maxima for a double slit). But as 𝑁 increases, there are additional minima
𝜆
and maxima between these positions. From the graphs we see that between 𝛽 = 0 and 𝛽 = 𝜋 there are 𝑁 −
1 minima, which cause the intensity to remain small and makes the width of the maxima peaks at 𝛽 = 0 and
𝛽 = 𝜋 narrower as 𝑁 increases. Since gratings generally have very large numbers of slits, the intensity is for
practical purposes zero everywhere, except at the maxima corresponding to 𝑑 sin 𝜃 = 𝑚𝜆. Thus the light
waves from a grating undergo destructive interference in almost every direction, except at certain angles for
which constructive interference occurs. The intensity pattern is thus that of several bright lines at

𝒅 𝐬𝐢𝐧 𝜽 = 𝒎𝝀 (108)

on a dark background.

Zeroes in the intensity pattern should occur whenever sin 𝑁𝛽 = 0. This occurs when 𝑁𝛽 = 𝑚𝜋, where 𝑚 is
𝑚
an integer i.e. for 𝛽 = 𝜋. This shows that the 𝛽-values corresponding to minima are evenly spaced and
𝑁
explains why are 𝑁 − 1 of them between 𝛽 = 0 and 𝛽 = 𝜋. An interesting thing happens when 𝑚 = 𝑁: one
would expect another minimum, but the graphs show a maximum! This is because when 𝑚 = 𝑁 we have
𝛽 = 𝜋 and the denominator in the intensity equation is zero. Mathematically one may not divide by zero,
but using L’Hopital’s rule it can be shown that

sin2 (𝑁𝛽)
lim 𝐼 = lim 𝐼0 = 𝑁 2 𝐼0 .
𝛽→𝑚𝜋 𝛽→𝑚𝜋 sin2 𝛽

Such points, instead of being minima, turn out to be the main maxima. Note that all of the plots in Figure 99
are shown to have the same maximum intensity values, but the vertical scales are in fact different, and the
maximum intensity for 𝑁 slits has a value of 𝑁 2 𝐼0.

17.3. Dispersion of a grating

Up to now we have only considered monochromatic light incident on a diffraction grating. But consider if the
incident light contained different wavelengths. Equation (108) shows that the angle at which a grating creates
maxima is wavelength dependent. If light consisting of various wavelengths is incident simultaneously on a
grating, each wavelength (colour) will have its maximum at a different angle i.e. dispersion will occur. This is
shown in Figure 100.

Figure 100

Dispersion of white light by a

transmission diffraction grating.
The transmitted central beam
remains white (no dispersion), but
the beams above and below are
dispersed into rainbows (colours
varying from blue to red with
increasing angle).
 210

Mathematically, dispersion is defined by

𝑑𝜃
𝐷= (109)
𝑑𝜆

and is a measure of the angular spread for light of different wavelengths. We can obtain an expression for
the dispersion of a grating by first solving equation (108) for 𝜃 and then differentiating:

𝑑 𝑚𝜆 1 𝑚
𝐷= (arcsin )= .
𝑑𝜆 𝑑 𝑚𝜆 2 𝑑 (110)
√1 − ( )
𝑑

The dispersion increases as the order 𝑚 increases or the slit separation decreases, but is independent of the
number of slits.

17.4. Resolving power of a grating

Consider light consisting of two wavelengths, which have been dispersed by a grating. The resolving power
of the grating is a measure of how close the two wavelengths can be, and yet still have their peaks separate
or distinguishable. Mathematically, the resolving power is defined by
𝜆ave
𝑅=
(Δ𝜆)min (111)

where 𝜆ave is the average of the wavelengths and (Δ𝜆)min is the minimum wavelength difference for which
the two peaks can still be distinguished as separate. Note that the resolving power is unitless.

As a result of dispersion, the angle for the maxima of two different wavelengths will always be different, but
despite this it is not always possible to distinguish between the peaks from two very close wavelengths of
light. The resolving power is limited by the fact that the peaks have a definite width. If the width of the peaks
is too large then, although the angles of the maxima are different, the curves themselves overlap too much
to be distinguished. This is illustrated in Figure 101. Clearly the resolving power of a grating is determined by
both its dispersion and the width of the peaks (maxima) it creates.

Figure 101 The effect of peak separation and peak width on peak resolution. In the top left there are two peaks and
their unresolved sum. The peaks may be resolved by maintaining their widths but increasing their angular
separation (top right), or maintaining their angular separation but narrowing their widths (bottom left).
 211

The question of when two peaks can be resolved (distinguished) is subjective. Consider Figure 102. The top
diagram shows two peaks clearly resolved, while the bottom shows two peaks unclearly unresolved. But
when, between these extremes, can we just distinguish or resolve two peaks? One suggestion by Sparrow is
illustrated second from the bottom, and suggests that if the sum of the peaks is flat at the top, this is
unnatural and it must contain two peaks. That is reasonable, but an older and more commonly accepted
criterion is that of Rayleigh, who suggested that two peaks can be resolved only if the maximum of the first
peak is no closer to the second peak than the second peak’s first minimum. This results in a slight dip in
intensity between the two peaks and is illustrated in the diagram second from the top. We will use Rayleigh’s
criterion to calculate the resolving power of a grating.

Clearly
resolved

Figure 102
Rayleigh

Clearly resolved peaks (above) and

unresolved peaks (below). The
assessment of when peaks are resolved is
Sparrow
subjective: the Rayleigh and Sparrow
limits are illustrated).

Not resolved

𝜆
Taking into account Rayleigh’s criterion, we can calculate the resolving power with 𝑅 = Δ𝜆 for a peak that
occurs for the wavelength 𝜆 and where Δ𝜆 is the change in wavelength necessary to shift the peak to the
position of its first minimum. The peak must satisfy 𝑑 sin 𝜃 = 𝑚𝜆. To find the change in wavelength that
would shift the peak to the position of its first maximum, we first consider the situation in terms of the
𝜋
parameter 𝛽. Figure 99 shows that for any maximum a shift of Δ𝛽 = 𝑁 will take one from the maximum to
𝜋𝑑 sin 𝜃 𝑑𝛽 𝜋𝑑 sin 𝜃 𝜋𝑚
the first side minimum of that peak. Since 𝛽 = 𝜆
, one has 𝑑𝜃 = − 𝜆2 = − 𝜆 . The sign shows that
Δ𝛽 𝜋𝑚
𝛽 decreases when 𝜆 increases, but in size Δ𝜆 ≈ 𝜆
. Therefore the resolving power is

𝜆 𝜋𝑚 (112)
𝑅= ≈ = 𝑁𝑚.
Δ𝜆 Δ𝛽
The resolving power increases as the order 𝑚 increases or the number of slits increases, but is independent
of the separation between the slits.
 212

17.5. Comparison of dispersion and resolving power

Dispersion is the ability to spread out the peak maxima corresponding to different wavelengths to different
angles, whereas resolving power is ability to separate the peaks themselves, and depends on the dispersion
as well as the peak/line widths.

Figure 103

Dispersion and resolving power of a

grating.

Figure 103 illustrates this: grating 𝐵 has the same dispersion, but a higher resolving power that grating 𝐴
(because the peak positions are the same, but the peaks for grating 𝐵 are narrower), while grating 𝐶 has a
higher dispersion than grating 𝐴 (because the two peaks have a greater angular separation).

Example: A diffraction grating has 12600 rulings evenly spaced over a distance of 1 inch (25.4 mm). It is
illuminated at normal incidence with the yellow light from a sodium lamp. This light contains two very closely
spaced lines with wavelengths 589.00 and 589.59 nm (called the sodium doublet). (a) At what angle does the
first order maximum occur for the shorter wavelength? (b) Using the dispersion of the grating, determine the
angular separation of the two lines for the first order. (c) What is the minimum number of rulings required
for resolving the sodium doublet in the second order?
25.4×10−3
Solution: (a) The slit spacing is 𝑑 = 12 600−1
= 2.016 m. (Subtracting 1 in the denominator makes no
difference but see Figure 98.) From 𝑑 sin 𝜃 = 𝑚𝜆 with 𝑚 = 1 and 𝜆 = 589.00 nm one gets 𝜃 = 16.99°.
1 𝑚 𝑑𝜃
(b) The dispersion is 𝐷 = 2 𝑑
= 5.19 × 105 m-1 using the average wavelength. Since 𝐷 = 𝑑𝜆 , the
√1−(𝑚𝜆)
𝑑

angular separation will be ∆𝜃 ≈ 𝐷∆𝜆 = 3.06 × 10−4 for ∆𝜆 = 0.59 nm. The answer is unitless, meaning it
must be in radians (not degrees). It is about 0.02°.
𝜆 589.295 nm
(c) The required resolving power is at least 𝑅 = (Δ𝜆)ave = 0.59 nm
= 999. A grating has resolving power
min
𝑅 = 𝑁𝑚, so if 𝑚 = 2 then at least 𝑁 = 500 slits are required. Note that we cannot get fractions of a slit.
 213

17.6. Problems

1. A grating 20.0 mm wide has 6000 slits. (a) Calculate the distance between adjacent slits. (b) At what angles
will intensity maxima occur if the light shone on the grating has a wavelength of 589 nm?

2. White light containing all the wavelengths from 400 nm (blue) to 700 nm (red) is shone through a
transmission grating, which has a spacing between the slits of 2 m. (a) Calculate the angles at which the first
order maxima occur for both the blue and the red light. Hence obtain the angle of spread of the white light
for first order diffraction. (b) Will the second order “rainbow” overlap the first? Motivate your answer with a
calculation.

𝑚
3. Show that equation (110) for the dispersion can be written more simply as 𝐷 = . [Hint: substitute
𝑑 cos 𝜃
𝑑 sin 𝜃 = 𝑚𝜆.]

4. A sodium lamp gives out light at two very close wavelengths, namely 589.00 and 589.59 nm. (a) What is
the minimum resolving power of a diffraction grating that is necessary to distinguish the diffraction maxima-
peaks from these two wavelengths? (b) If the third order (𝑚 = 3) diffraction maxima are observed for these
two wavelengths, what is the minimum number of slits that the diffraction grating must have in order to
distinguish the diffraction maxima-peaks from these two wavelengths?

5. A reflection grating may be “blazed” when the lines on the surface are not symmetrical but more shaped
like a saw-tooth pattern. These asymmetrical gratings were considered by Lord Rayleigh in 1888, but only
successfully manufactured by William Wood in 1910. Consult a reliable source and explain an advantage such
gratings.

6. In 1912 Max von Laue suggested that the planes of atoms in crystals might act almost like a three
dimensional grating. It was soon confirmed that maxima occurred when X-rays are passed through a crystal,
due to interference. von Laue won the Nobel prize in 1914 and the next year it was won by William Henry
Bragg and William Lawrence Bragg (father and son). Search the internet for “Bragg’s law” and comment if it
is identical or slightly different from that for the formula giving maxima for a grating.
 214

18. -Point interference

18.1. Introduction

Many books would call this chapter “diffraction” and that is indeed what will be discussed. We call it ∞-point
interference to emphasize the link to the previous chapters dealing with 2-point and 𝑁-point interference.
What is diffraction, and how does it relate to interference? On a simple level, diffraction refers to the bending
of light after it passes by next to a barrier e.g. the edge of an opaque object or an opening. On a deeper level,
diffraction can be interpreted as an interference effect of an infinite number of virtual waves originating on
different parts of a single opening, which can lead to maxima and minima – this idea will become clearer
from the mathematics later in the chapter.

In the past two chapters we considered light passing through openings (slits) and interfering. We assumed
that the slits acted as line sources of cylindrical waves (or in cross-section, as point sources of circular waves).
In this chapter we shall examine real slits with finite size, and derive the effects of diffraction as light passes
through them. We then use these results to refine our equations of the last two chapters, as well as
considering diffraction through rectangular and circular apertures.

18.2. Simple concept of diffraction

Figure 104 Diffraction of a wave with some constant wavelength by slits of different sizes.

When a plane wave encounters an obstacle that has an opening of about the same size as the wavelength of
the wave, then the part of the wave that passes through the opening will flare out (diffract) into the region
beyond the obstacle. Diffraction phenomena occur for waves of all types, including light waves. The relation
between the wavelength (𝜆) of an incident plane wave and the width of a slit (𝑎) is depicted in Figure 104.
The degree of diffraction depends on the wavelength of the wave as well as the size of the opening:
 215

 If the wavelength is small compared to the opening then the diffraction will be weak (not much flaring
out), but if the wavelength is large compared to the opening then the diffraction will be strong.

 Looking at it from another point of view, if the opening is large compared to the wavelength of the
wave then the diffraction will be weak (not much flaring out), but if the opening is small compared
to the wavelength then the diffraction will be strong.

Figure 105

Diffraction of waves with

different wavelengths by a slit
of constant size.

Figure 105 illustrates these results with experiment images. In fact, it will be shown later that the amount of
diffraction does not depend on the opening size 𝑎 or the wavelength 𝜆 independently, but only on their ratio
𝜆
𝑎
. The greater this ratio, the stronger the diffraction effect will be. In geometrical optics light is considered to
travel only in straight lines (like a stream of particles) and does not take account of diffraction: geometrical
𝜆
optics will therefore only be accurate provided the ratio 𝑎 is small and diffraction effects are negligible. This
will be the case if the slits (or other apertures/openings) have dimensions very much larger than the
wavelength of light – since the wavelength of visible light is very small (between 400 and 700 nm) this is often
the case.
 216

18.3. Huygens’ principle and diffraction

Figure 106

Propagation of a plane and

circular wave predicted by
Huygens’ principle.

Huygens’ principle is a method to predict how a wave propagates. It states that we can regard each point on
a wavefront as a source of secondary circular wavelets which propagate with the same characteristics as the
wave in the medium. Then, at a later time, the new wavefront is a tangent to the wavefronts of these
secondary wavelets. Figure 106 illustrates Huygens’ principle applied to the propagation of a plane wave and
a circular wave.

Figure 107

𝑎
Huygens’ principle predicts
diffraction of a wave passing
through an opening.

Huygens’ principle clearly predicts that a wave passing through an opening will spread out behind the
opening, as shown in Figure 107.
 217

18.4. The intensity transmitted by a finite sized slit

Although Huygens’ principle explains the phenomenon qualitatively, we require a quantitative

(mathematical) model for the intensity of the wave passing through an opening. For this we extend the idea
of Huygens’ secondary wavelets. As before, we regard each point on a wavefront as a source of secondary
circular wavelets which propagate with the same characteristics as the wave in the medium. But now (instead
of just considering the tangent to their wavefronts) we rather add all these wavelets together to obtain the
new wave at a later time. The new wave may therefore be regarded as the interference pattern of the infinite
secondary wavelets.

Y
P(x,y )

a /2 N
S1
a/N


S2
a/N
S3
a a/N
x X
a/N

a/N
S
a /2 N N

Figure 108 𝑵 equally spaced secondary wavelet point sources placed inside a slit of width 𝒂. In this figure 𝑵 = 6 but
for the theory we consider when each point in the wavefront acts as a source of secondary wavelets i.e. as 𝑵 → ∞.

When a plane wave reaches an opening, we regard each point in the opening as a source of secondary circular
wavelets. We add these together and their interference pattern then gives the diffracted wave behind the
opening. To take each point in the opening as the source of a secondary wavelet, we consider 𝑁 equally
spaced sources and allow 𝑁 → ∞. As shown in Figure 108, if the slit opening seen from above has a width 𝑎
𝑎
and contains 𝑁 wavelet sources (shown as open circles), the separation of these sources is 𝑁. (Note that gaps
𝑎
of have been included at the edges of the slit – this arrangement is convenient but if the gaps are removed
2𝑁
 218

or changed in size to rather be equal to the spacing of all the other sources, the mathematics of the derivation
is slightly different but the final result is the same.)

𝐴
We represent the wavelet from each source by an equation of the form 𝑌[𝑥, 𝑡] = 𝑁 cos(𝑘𝑥 − 𝜔𝑡). The
amplitude is divided by 𝑁 because the more wavelets consider, the smaller each one’s amplitude should
become. Adding an infinite number of sources without reducing their amplitude in this way would result in
an infinite disturbance. At some point 𝑃 the disturbance due to the circular wave from source 𝑆𝑖 will be
𝐴
𝑌𝑖 [𝑡] = 𝑁 cos(𝑘{𝑆𝑖 𝑃} − 𝜔𝑡) and the resultant wave at 𝑃 is due to the waves from all the (infinite) sources is

𝑁 𝑁
𝐴
𝑌𝑃 [𝑡] = lim ∑ 𝑌𝑖 [𝑡] = lim ∑ cos(𝑘{𝑆𝑖 𝑃} − 𝜔𝑡).
𝑁→∞ 𝑁→∞ 𝑁
𝑖=1 𝑖=1

This is very similar to our derivation for the interference pattern of a grating in the previous chapter, except
that we are considering an infinite number wavelets instead of a finite number of ordinary waves. But now
there is an important difference: the path length difference between each successive source is not 𝛤 =
𝑎
𝑑 sin 𝜃 as for a grating where 𝑑 is the source separation, but instead 𝛤 = 𝑁 sin 𝜃 because the wavelet
𝑎
sources are separated by . The path length 𝑆𝑖 𝑃 can therefore be written in terms of the shortest path length
𝑁
𝑆1 𝑃 as

𝑎
𝑆𝑖 𝑃 = 𝑆1 𝑃 + (𝑖 − 1) sin 𝜃.
𝑁

Therefore

𝑁 𝑁−1
𝐴 𝑎 𝐴 𝑎
𝑌𝑃 [𝑡] = lim ∑ cos (𝑘 (𝑆1 𝑃 + (𝑖 − 1) sin 𝜃) − 𝜔𝑡) = lim ∑ cos ({𝑘{𝑆1 𝑃} − 𝜔𝑡} + 𝑗 {𝑘 sin 𝜃}) .
𝑁→∞ 𝑁 𝑁 𝑁→∞ 𝑁 𝑁
𝑖=1 𝑗=0

To make the summation we use the identity

𝑁−1
sin(𝑁𝑣2) 𝑣
∑ cos(𝑢 + 𝑗𝑣) = cos (𝑢 + (𝑁 − 1)2).
sin(𝑣2)
𝑗=0

𝑎
By comparison 𝑢 = 𝑘{𝑆1 𝑃} − 𝜔𝑡 and 𝑣 = 𝑘 𝑁 sin 𝜃, so

𝑎
𝐴 sin (𝑁 𝑘 2𝑁 sin 𝜃) 𝑎
𝑌𝑃 (𝑡) = lim 𝑎 cos (𝑘{𝑆1 𝑃} − 𝜔𝑡 + (𝑁 − 1)𝑘 sin 𝜃)
𝑁→∞ 𝑁 2𝑁
sin (𝑘 2𝑁 sin 𝜃)

The only time dependence occurs in the cosine term, which therefore oscillates between +1 and -1 as time
goes by. The remaining part of the expression represents the amplitude of the resultant oscillation at point
𝑃, which is

𝑘𝑎 sin 𝜃
𝐴 sin ( 2
)
𝐴𝑃 = lim .
𝑁→∞ 𝑁 𝑘𝑎 sin 𝜃
sin ( 2𝑁 )

Defining
 219

𝑘𝑎 sin 𝜃 2𝜋 𝑎 sin 𝜃 𝜋𝑎 sin 𝜃

𝛼= = = (113)
2 𝜆 2 𝜆

allows us to write this more simply as

𝐴 sin 𝛼
𝐴𝑃 = lim .
𝑁→∞ 𝑁 sin 𝛼
𝑁
𝛼 𝛼 𝛼
But since 𝑁 → ∞, the value of 𝑁 will be small and sin 𝑁 ≈ 𝑁. Therefore

𝐴 sin 𝛼 sin 𝛼
𝐴𝑃 = lim 𝛼 =𝐴 .
𝑁→∞ 𝑁 𝛼
𝑁

sin2 𝛼
The light intensity is proportional to the square of the amplitude, so 𝐼 ∝ 𝐴2 𝛼2
. But the amplitude of each
𝐴
of the wavelets was 𝑁
and that of the original wave entering the slit was 𝐴, so the original wave intensity was
2
𝐼0 ∝ 𝐴 and
sin2 𝛼
𝐼 = 𝐼0 (114)
𝛼2

represents the diffracted intensity through the slit. Remember that the parameter 𝛼 is just a defined quantity
and not any particular angle related to the geometry of the problem.

Figure 109 Intensity distribution as a function of 𝜶 for single slit diffraction.

Figure 109 shows a plot of this intensity as a function of 𝛼. There is a large central maximum and smaller side
maxima. Between the maxima there are minima of zero. Our previous qualitative discussion about diffraction
did not reveal these details. If one now knows what to look for, hints of these characteristics are visible in
 220

the bottom experimental photograph of Figure 105, although they are not included in the schematic
representation of diffraction in Figure 104.

The minima (zeroes) in the diffracted intensity pattern occur whenever sin 𝛼 = 0. This happens whenever
𝜋𝑎 sin 𝜃
𝛼 = 𝑚𝜋 for integral 𝑚, and using the definition of 𝛼 this means that 𝑚𝜋 = 𝜆
or

𝒂 𝐬𝐢𝐧 𝜽 = 𝒎𝝀. (115)

But note that this predicts, for 𝑚 = 0, a minimum at 𝜃 = 0 which is on the screen opposite the centre of the
slit where the main central maximum actually occurs. This anomaly occurs because for 𝜃 = 0 one also has
0
𝛼 = 0 and the denominator in equation (114) is zero, giving an intensity 0 . Applying the small angle
approximation as 𝛼 → 0 shows the intensity tends to 𝐼0 and is not really zero in this case. Thus equation (115)
for the minima is valid for any integer 𝑚 besides zero.

A rough estimation from Figure 109 suggests the side maxima lie about half way between the zeroes at 𝛼 =
3𝜋 5𝜋 7𝜋
𝑚𝜋, i.e. at 𝛼 = ± ,± ,± etc. But a careful examination shows the side peaks are asymmetrical and
2 2 2
lie at ±1.43𝜋, ±2.46𝜋, ±3.47𝜋, etc. This can be confirmed by finding the peak positions using calculus. The
derivative of equation (114) gives

𝑑𝐼 2𝐼0
= sin 𝛼 (𝛼 cos 𝛼 − sin 𝛼)
𝑑𝛼 𝛼 3

and setting this to zero gives sin 𝛼 = 0 (the minima already discussed) or 𝛼 cos 𝛼 − sin 𝛼 = 0. The latter
simplifies to 𝛼 = tan 𝛼. This can be solved numerically or graphically (Figure 110) and confirms the positions
of the maxima.

Figure 110

Graphs of 𝒚 = 𝒙 and 𝒚 =
𝐭𝐚𝐧 𝒙 on the same axes, to
find values where 𝒙 = 𝐭𝐚𝐧 𝒙.
 221

Equation (114) and Figure 109 show the intensity for diffraction through a single slit as a function 𝛼. But 𝛼 is
just a defined parameter, and of more practical significance is the intensity as a function of the angle 𝜃, which
we can write out explicitly as

𝜋𝑎 sin 𝜃
sin2 ( )
𝐼 = 𝐼0 𝜆
2
𝜋𝑎 sin 𝜃
( )
𝜆

It is clear that the diffraction effect depends on the ratio 𝑎/𝜆 rather than the slit width or wavelength
independently. The relative intensity 𝐼/𝐼0 is plotted as a function of 𝜃 for various slit sizes (relative to the
wavelength) in Figure 111.

Figure 111

Single slit diffracted

intensities as a function of the
angle 𝜽 for various slit to
wavelength ratios.

The larger the ratio 𝑎/𝜆 is made, the more the diffracted wave is directed in the forwards direction i.e. the
less it flares out behind the slit. When this ratio is small, the wave flares out significantly. These diagrams
correspond qualitatively to the pattern in Figure 109, despite the horizontal axis being the deviation angle 𝜃
rather than the parameter 𝛼.
 222

18.5. Double slit interference including diffraction effects

In an earlier chapter we derived equation (106) for double slit interference, namely 𝐼 = 4𝐼0 cos2 𝛽. As part
of the derivation we assumed that each slit acted as a line source giving out cylindrical waves. These waves
were assumed to travel outwards in circles in all directions. If the slits are very narrow, this as a good
assumption since the diffraction is very strong and the waves spread out almost equally in all directions
behind the slits (somewhat like in Figure 111(a)). But if the slits have non-negligible width then the wave
behind them does not spread out in a circular fashion but according to equation (114) for single slit diffraction
sin2 𝛼
i.e. 𝐼 = 𝐼0 𝛼2
.

For double slit interference with diffraction, we assume that both of the slits have the same width and hence
diffraction pattern. We assume this pattern is wide compared to the separation of the slits, so that the
sin2 𝛼
diffraction patterns almost coincide and the single formula 𝛼2
describes the change in intensity due to
diffraction for both slits together. Then the intensity due to interference including diffraction is given by

sin2 𝛼 sin2 𝛼
𝐼 = (4𝐼0 cos 2 𝛽) ( ) = 𝐼max ( 2 ) (cos2 𝛽) (116)
𝛼2 𝛼

where 𝐼max = 4𝐼0 is the maximum intensity in the centre of the pattern.

Figure 112

Double slit interference with

diffraction.

(a) Double slit interference

pattern for slits acting as
sources of cylindrical waves.

(b) Diffraction pattern for each

of the slits. The slits are
assumed to lie close enough
together that this pattern is
also used for the combination
of both slits.

(c) The interference pattern

within the diffraction
envelope.
 223

18.6. Grating interference including diffraction effects

sin2(𝑁𝛽)
In an earlier chapter we derived equation (106) for grating interference, namely 𝐼 = 𝐼0 sin2 𝛽
. As part of
the derivation we assumed that each slit acted as a line source giving out cylindrical waves, travelling
outwards in circles equally in all directions. For narrow slits exhibiting strong diffraction this is a good
assumption, but if the slits have non-negligible width the wave behind them does not spread out in a circular
sin2 𝛼
fashion but according to equation (114) for single slit diffraction i.e. 𝐼 = 𝐼0 .
𝛼2

For grating interference with diffraction (similar to the double slit in the previous section) we just include the
sin2 𝛼
effect of the diffraction amplitude by multiplying the interference pattern by 𝛼2
, giving

sin2 (𝑁𝛽) sin2 𝛼

𝐼 = (𝐼0 ) ( 2 ). (117)
sin2 𝛽 𝛼

Figure 113 (a) Calculated 𝑵-slit interference intensity, including effect of finite slit widths (𝑵 = 𝟖 and 𝒅 = 𝟑𝒂). (b)
Experimental 𝑵-slit interference patterns with diffraction effects. The slit system is shown on the left of each
photograph, and the top one corresponds to just diffraction from a single slit.
 224

18.7. Diffraction through a rectangular opening

A slit is very long and narrow: what about diffraction through a rectangular opening where the light
encounters a barrier not only on the sides (width), but also above and below (height)? One may make such
a rectangular opening by taking a slit, rotating it 90° and placing it over another one which has not been
rotated. This is illustrated in Figure 114.

Now there will be diffraction both horizontally and vertically. A mathematical derivation is not required to
predict the diffraction pattern of the rectangular opening: rather we combine the formula for single
diffraction with a second formula for the perpendicular direction, obtaining

sin2 𝛼𝑦 sin2 𝛼𝑧
𝐼 = 𝐼0 ( )( ) (118)
𝛼𝑦2 𝛼𝑧2

𝜋𝑎𝑦 sin 𝜃 𝜋𝑎𝑧 sin 𝜃

where 𝛼𝑦 = 𝜆
and 𝛼𝑧 = 𝜆
, for the width of the opening 𝑎𝑦 and its height being 𝑎𝑧 .

Figure 114

Diffraction through a
rectangular aperture is just
like that through two
perpendicular slits. The
narrower the slit, the wider
the diffraction pattern.
 225

18.8. Diffraction from a circular opening

We now discuss diffraction by a circular aperture. This is very important in optics where many components
are circular e.g. the iris of an eye, a camera stop or a circular lens. When a distant point source of light (e.g.
a star) is viewed through a circular opening (or its image is formed by a lens) one gets the result shown in
Figure 115(a). This is the diffraction pattern of a circular aperture and consists of a central disk (called the
Airy disk) surrounded by several progressively fainter rings.

Figure 115 (a) Fraunhofer diffraction pattern of a circular aperture. (b) Bessel function 𝑱𝟏 (𝜸).
(c) Intensity profile across the diffraction pattern.

The intensity along a line through the centre of the diffraction pattern is shown in Figure 115(c) and
mathematically it can be shown to be described by
2
2𝐽1 (𝛾)
𝐼 = 𝐼0 ( ) (119)
𝜸
𝜋𝐷 sin 𝜃
where 𝜸 = 𝜆
and 𝐷 is the diameter of the circular opening. The function 𝑱𝟏 is called the Bessel function
of the first kind of order 1, and is rather similar to the sine function as shown by its graph in Figure 115(b),
but not quite periodic and with a decaying amplitude.51

The minima of the diffraction pattern occur when 𝐽1 (𝛾) = 0. The zeroes of this Bessel function can be looked
up in tables and occur for 𝛾 = 3.8317 (i.e. 1.22𝜋), 7.0156 (i.e. 2.23𝜋), 10.1735 (i.e. 3.24𝜋), etc. For the first
𝜋𝐷 sin 𝜃
minimum 𝛾 = 1.22𝜋 = 𝜆
, so

𝑫 𝐬𝐢𝐧 𝜽 = 𝟏. 𝟐𝟐𝝀. (120)

Besides the factor of 1.22, this is similar the equation for the first minimum of a single slit diffraction pattern,
namely 𝑎 sin 𝜃 = 𝑚𝜆 with 𝑚 = 1 and 𝑎 being the slit width, instead of the opening diameter 𝐷.

51
In general the Bessel function of the first kind of order 𝑛 can be expressed as a Taylor series
∞
(−1)𝑘 𝑥 𝑛+2𝑘
𝐽𝑛 (𝑥) = ∑ ( ) .
𝑘! (𝑛 + 𝑘)! 2
𝑘−0
 226

Example: The spreading of a laser beam.

Laser beams are highly collimated i.e. they travel along a line. But a laser beam does spread out due to
diffraction. This can be important if the laser light will travel a far distance (e.g. in experiments where laser
light is bounced off reflectors placed on the moon during the Apollo moon missions in order to very accurately
measure the distance to the moon). We can get some idea of this effect by supposing the laser beam is
emitted through an opening at the front of the laser with diameter 𝐷 and the size of the laser spot is taken
as the Airy disk of its diffraction pattern (i.e. for angles 𝛾 up to the first minimum). [Note that this Airy disk
contains 84% of the total light intensity, while the first ring contains another 7%.]
The first minimum occurs when 𝐷 sin 𝜃 = 1.22𝜆. The spreading out is small so we may use the small angle
𝜆
approximation sin 𝜃 ≈ 𝜃 and so the angle of spread is given by 𝜃 ≈ 1.22 𝐷. The light spreads out in a cone
shape and the diameter of a spot shone on a surface far away subtends and angle 2𝜃. If this surface is a
distance 𝑥 away and 2𝜃 is in radians, the diameter of the Airy disk on the surface will therefore be
2.44𝜆𝑥
𝒟 ≈ (2𝜃)𝑥 ≈ .
𝐷
For example, a green laser (𝜆 = 550 nm) with a 5 mm opening projected onto a surface 1 km away will have
2.44(550×10−9 )(1000)
a Airy spot diameter of about = 0.27 m (i.e. 27 cm).
5×10−3

Example: Resolving power.

The fact that images formed by lenses are actually diffraction patterns is of importance when we wish to
resolve (i.e. distinguish between) two point objects having a small angular separation. Examples may be
resolving double stars with a telescope or distinguishing very small nanoparticles in a microscope. The top
part of the figure above represents the visual appearance of two small objects observed as a function of
separation. We do not see the objects themselves, but instead their diffraction patterns. Below each picture
is a representation of the intensity of the diffraction patterns. In figure (a) the intensity peaks overlap to such
an extent that it is hard to resolve them, while in figure (c) the peaks are well separated and easily resolved.
Recall that the Rayleigh criterion suggests we will just be able to resolve two peaks when the maximum of
one coincides with the first side minimum of the other (see the section of the resolving power of gratings).
This situation is represented in figure (b). For a diffraction pattern from a circular opening, the central
maximum is at 𝜃 = 0 and the first side minimum occurs for 𝐷 sin 𝜃 = 1.22𝜆, so the minimum angular
1.22𝜆 1.22𝜆
separation necessary to distinguish two objects is 𝜃min = arcsin ( 𝐷
) ≈ 𝐷
using the small angle
approximation if 𝜆 ≪ 𝐷. To reduce 𝜃min so that one can resolve more closely separated objects, one can
reduce 𝜆 or increase 𝐷. Both techniques are used e.g. short wavelength ultraviolet light is used in some
microscopes and also projection equipment used to create the very fine details of integrated circuits for
computer chips, while astronomers make telescopes with larger and larger diameters.
 227

18.9. The fine structure of shadows

Figure 116 Fringes in the shadow cast by the hand of a person, illuminated by laser light.

The fact that objects cast shadows shows that light travels in straight lines and suggests that light acts like a
stream of particles emanating from a luminous object. However, shadows are not actually that simple. Firstly,
the edge of a shadow is blurred when the light source is not a point but an extended object like the sun.
When one uses a point source of monochromatic coherent light, fringes can be observed at the edge of
shadows as seen in Figure 116. These fringes are caused by diffraction of the light passing the boundary.
However, their analysis is beyond the scope of this work.

18.10. Problems

1. Consider the theory of single slit diffraction. Does the separation angle between adjacent minima increase
or decrease when a longer wavelength is used? Motivate your answer.

2. Light of wavelength 633 nm is incident on a narrow slit. The angle between the first diffraction minimum
on the one side of the central maximum and the first minimum on the other side is 1.20°. What is the width
of the slit?

3. A 0.10 mm wide slit is illuminated by light of wavelength 589 nm. Consider a point 𝑃 on a viewing screen
on which the diffraction pattern of the slit is viewed; the point is 30° from the central axis of the slit. What is
the phase difference between the Huygens wavelets arriving at point 𝑃 from the top and midpoint of the
slit?
 228

4. Monochromatic light with wavelength 538 nm is incident on a slit with width 0.025 mm. The distance from
the slit to a screen is 3.5 m. Consider a point on the screen 1.1 cm from the central maximum. (a) Calculate
𝜃 for this point. (b) Calculate 𝛼 for this point. (c) Calculate the ratio between the intensity at this point and
the intensity at the central maximum.

5. A coherent beam of laser light (with wavelength 600 nm) is shone through a narrow slit 0.1 mm wide, and
the single slit diffraction pattern projected onto a wall 10 m away from the slit. (a) Determine the angle at
which the first minimum occurs in the diffraction pattern. (b) Determine the distance on the wall between
the centre maximum and the first minimum positions.

6. Both blue light (wavelength 400 nm) and red light (wavelength 700 nm) is shone simultaneously through
a small slit in a single slit diffraction experiment. State for which colour light the angle of the first diffraction
minimum is smaller, and also which colour will be seen by an observer looking at the slit from this angle.

7. Sound waves with frequency 3000 Hz and speed 343 m/s

diffract through the rectangular opening of a speaker cabinet
and into a large auditorium. The opening, which has a width
of 30.0 cm, faces a wall 100 m away. Assume the height of
the opening is much larger than its width. Where along that
wall would a listener be at the first diffraction minimum and
thus possible have difficulty hearing the sound? (Ignore
reflections.)

8. Suppose that the central diffraction envelope of a double slit interference/diffraction pattern contains 11
bright fringes, and the first diffraction minima are coincident with interference bright fringes. How many
bright fringes lie between the first and second minima of the diffraction envelope?

9.(a) How many bright fringes appear between the first diffraction-envelope minima to either side of the
central maximum in a double slit interference pattern if 𝑑 = 0.150 mm and 𝑎 = 30.0 m? (b) What is the ratio
of the intensity of the third bright fringe to the intensity of the central fringe?

10. The two headlights of an approaching car are 1.2 m apart. Regard the headlamps as point sources of
light, with average wavelength 𝜆 = 550 nm. Also assume that diffraction effects alone limit the resolution so
that Rayleigh's criterion can be applied. (a) What is the minimum angle between the two headlamps, as
viewed by an observer, so that they will be resolved i.e. seen as separate light sources? Assume the iris of
the observer's eye is a round opening 3.0 mm in diameter. (b) How close must the oncoming car be before
the observer can tell if it has two lights, or just one light (e.g. it might also be a motor bike).
 229

19. CASE Study: Rainbows

As a sign of this everlasting covenant which I am making with you and with all living beings, I
am putting my bow in the clouds. It will be the sign of my covenant with the world. Whenever I
cover the sky with clouds and the rainbow appears, I will remember my promise to you and to
all the animals that a flood will never again destroy all living beings. Genesis 9:12-15

19.1. Introduction

When sunlight is scattered by raindrops, why is it that colourful arcs appear in certain regions of the sky?52
The scientific description of the rainbow is often supposed to be a simple problem in geometrical optics that
was solved long ago: however, a satisfactory quantitative theory of the rainbow was only developed in the
latter part of the 20th century. Moreover, the theory involves much more than geometric optics: allowance
must be made for the wave-like properties like interference, diffraction and polarization, as well as the
particle-like properties such as the momentum carried by a beam of light.

The first person known to try explaining the rainbow was Aristotle,53 who ascribed it to an unusual type of
reflection of sunlight from clouds, occurring at a fixed angle. The fixed angle of refection correctly predicted
the circular shape of the bow, and Aristotle understood that the rainbow is not a material object, but rather
a perceived virtual object.

In 1266 Roger Bacon accurately measured the angle made between the sun's rays and the primary bow to
be 42°. This means that the sunlight is turned through 180° - 42° = 138° before coming to the observer, which
is called the (primary) rainbow angle.

In 1304 the German monk Theordoric of Freiburg rejected Aristotle's hypothesis that the rainbow is formed
by the collective reflection by the raindrops in a cloud, and suggested that each drop is individually capable
of producing a rainbow. He experimented making a rainbow using a spherical flask filled with water.
Theodoric's work remained unknown for three centuries until the same discoveries were made again
independently by Descartes,54 who did similar experiments. Both these men determined that the primary
bow is made up from light rays entering a droplet and reflecting from its inner surface once before re-
emerging. The secondary bow is formed by light which experiences two internal reflections. These internal
reflections are not total internal reflections, and in fact a large fraction of the light is also transmitted out of
the droplet during the reflections, and hence does not contribute to the rainbow. This explains both the
rainbow's dimness, and why the secondary bow is dimmer than the primary one.

52 This account of the theory of rainbows closely follows that given by H. Moyses Nussenzveig in the Scientific American article “The
theory of the rainbow” of April 1977.
53 Aristotle lived from 384 to 322 BC.
54 Descartes was also the first person to write Snell's law in its present mathematical form (in 1637). The law was stated by Snell (in

a different form) in 1621, and was known to Thomas Harriot by 1602 and already published by Abu Said al-Ala Ibn Sahl in 984. (Physics
World, April 2002, p.64.)
 230

Figure 117 (a) A rainbow. (b) Different colours are scattered at different angles. Each drop scatters a different colour
to the observer and thus appears a different colour. Red light is scattered through a slightly smaller angle and
therefore appears closer to the sun (in an angular sense), and so further from the antisolar point (at the rainbow's
“centre”).

Although an individual drop of water is capable of producing an entire rainbow, each colour (wavelength) of
light is most strongly scattered at a slightly different angle. Therefore, an observer at a certain position will
only see the droplet as having a single colour, determined by his viewing angle. Should his position change,
the same droplet would appear to be a different colour. This is illustrated by Figure 117(b). So, in actual fact,
an observer can only see one colour of the rainbow from an individual droplet. Many droplets must be
observed, each at a different angle (and hence appearing a different colour), to see a complete rainbow. All
droplets viewed at the same angle, relative to the sun, appear the same colour, giving the rainbow its circular
(bow) shape.

19.2. Features of the rainbow

Figure 118 Features of a rainbow.

 231

The main features of the rainbow, obtained by careful observation and shown in Figure 118, are as follows:

 The primary bow: this arc makes an angle of about 42° about the antisolar point (the point in the sky
directly opposite the sun from the observer) and has a width of about 2°. The colours are arranged
from violet on the inside to red on the outside.
 The secondary bow: this makes an angle of about 50° about the antisolar point. The colours are
arranged from red on the inside to violet on the outside.
 Alexander's dark band:55 this is the region between the primary and secondary bow, and is darker
than the rest of the sky. The region inside the primary bow is also noticeably lighter.
 Supernumerary arcs: These are a series of faint bands, usually pink and green alternately, which are
sometimes present on the inner side of the primary bow. Rarely, they may also appear on the outer
side of the secondary bow. The supernumerary arcs are usually seen most clearly near the top of the
bow.

19.3. Analysis using geometric optics (Descartes model of the

rainbow)

Figure 119 (a) Ray incident on a water droplet. The impact parameter is the distance from the line through the droplet
centre for a droplet with unit radius. (b) Trace of rays with various impact parameters. The bounding curve to all the
rays is called a caustic, meaning burning and referring to the concentration of rays there.

As a result of surface tension, we assume that the water droplets in clouds are spherical.56 We also assume
that all the incident rays from the sun are parallel.57 Due to high degree of symmetry, there is only one

55 Named after the Greek philosopher Alexander of Aphrodisias who first described it in about 200 AD.
56 Air resistance as a raindrop falls can cause it to become distorted, especially if the drop is big. Deviations from a spherical shape
have been measured by suspending water droplets in the air stream of a vertical wind tunnel. This showed that drops with a radius
less than 0.14 mm remained essentially spherical, but for droplets with a radius of 1.4 mm the height/width ratio was 0.85. The
asymmetrical shape of the water droplets can cause the top of the rainbow to be less bright than the sides.
57 A correction for this approximation is considered later.
 232

significant variable: the displacement of the incident ray of light from the axis passing parallel to the light ray
and through the centre of the droplet. Consider a ray striking a water droplet with some impact parameter
𝑏, which is defined as the fraction of the distance of the incident ray from the central axis, to the radius of
the droplet. Figure 119(a) shows that the impact parameter equals in the sine of the angle of incidence:
𝑦
𝑏 = = sin 𝜃𝑖 .
𝑅

Then the following rays result:

 At the surface of the droplet, some of the light is reflected. This reflected ray is referred to as the Class 1
ray. The remaining light is refracted into the droplet, changing direction according to Snell's law, and then
striking the back surface of the droplet. Since the triangle formed by the first and second points of impact
and the centre of the circle is an isosceles triangle, the new angle of incidence is equal to the initial angle
of refraction.

Exercises

1. Show that the refracted angle of a ray with impact parameter 𝑏 in a droplet with refractive index 𝑛 is
𝑏
given by 𝜃𝑟 = arcsin .
𝑛

2. Assume the impact parameter is positive when the incoming ray strikes the upper half of the droplet.
Also consider clockwise rotations to be positive. Show that the Class 1 ray for a positive impact parameter
𝑏 is rotated through an angle given by ∆1 = 2 arcsin 𝑏 − 180°, which is a negative quantity.

3. (Difficult) The position where the light strikes the far side of the droplet depends on the impact
parameter, but for a water droplet it is always in the same side (top or bottom) of the central axis of the
droplet as the impact parameter. This would not be the case if water had a higher index of refraction, or
if the droplet was made from a different liquid. What is the minimum refractive index which would result
at least some of the refracted light crossing the central axis?

 A large fraction of the light is now refracted out of the droplet. This is called the Class 2 ray. But some of
the light is reflected58 inside the droplet (with angle of reflection equal to the angle of incidence), and
strikes the inner surface again.

Exercises

1. Prove that the angle (relative to the normal) that the light exits from the drop is the same as its original
angle of incidence.

𝑏
2. Prove that the Class 2 ray is rotated through an angle of ∆2 = 2 (arcsin 𝑏 − arcsin 𝑛).

3. Prove that a light ray entering a droplet from outside can never undergo total internal reflection inside
i.e. it is impossible for the light to get “trapped” inside the droplet. (An exception is light striking the
droplet tangentially.)

58 Note that this is not a total internal reflection, but merely an internal reflection.
 233

 At the new strike point, a large fraction of the light is again refracted out of the droplet. This is called the
Class 3 ray, and it is responsible for the primary bow. But, once again, some of the light is reflected inside
the droplet, and goes on to strike the inner surface again.

Exercise

𝑏
Prove that the Class 3 ray is rotated through an angle of ∆3 = 180° + 2 arcsin 𝑏 − 4 arcsin 𝑛.

 Once again most of the remaining light escapes via refraction to form the class 4 ray (responsible for the
secondary bow), while a small fraction is reflected inside the drop. Therefore higher and higher class rays
are formed. Although these may theoretically give rise to higher order bows, these are not ordinarily
visible.

Exercises

𝑏
1. Prove that the Class 4 ray is rotated through an angle of ∆4 = 360° + 2 arcsin 𝑏 − 6 arcsin 𝑛.

2. Give a general formula for the angle ∆𝑚 rotated for a Class 𝑚 ray, where 𝑚 > 2.

Sunlight illuminates the droplets at all possible impact parameters, and the directions of the Class 3 and Class
4 rays are not fixed, but vary greatly depending on the impact parameter (see Figure 119(b)). Therefore light
is scattered in all directions. Why then is the scattered intensity strongest at the rainbow angle, leading to
the formation of the rainbow?

Figure 120 Class 3 and 4 rays scattering angles (∆𝟑 and ∆𝟒 ) as a function of impact parameter.
 234

The answer to this question was first provided by Descartes. Consider the Class 3 rays. If the impact
parameter is zero, these are reflected straight backwards, turning through 180°. As the impact parameter
increases, this scattering angle decreases, it reaches a minimum value of 138°, after which it increases again.
Since the scattering angle changes direction at 138°, it is in this region at where the scattering angle changes
most slowly with changing impact parameter, and hence the angle at which the most light is scattered. Stated
another way, we assume that the reflected intensity (brightness) at a certain angle increases as the rate with
which the scattering angle changes with impact parameter decreases. This explains the value of the rainbow
angle, despite the fact that some light is scattered by the droplets in other directions as well. The ray
corresponding to the minimum scattering angle is called the Descartes ray or the rainbow ray.

Exercises

1. In a previous exercise an expression was derived for the angle rotated by the Class 3 ray, namely ∆3 . Plot
this function using a computer for 𝑏 between 0 and 1 in steps of 0.01. Your plot should correspond to Figure
120.

2. Calculate the derivative of this function.

3. Hence find the impact parameter for which the scattering angle is a minimum.

4. Obtain the formula for the minimum scattering angle (the rainbow angle), namely ∆3,min = 180° +
16−4𝑛2 16−4𝑛2
2 arcsin √ 12
− 4 arcsin √ 12𝑛2
.

A similar analysis can be made for the Class 4 rays. If the impact parameter is zero, these rays are reflected
twice, turning through 360°, and hence passing undeflected through the droplet. As the impact parameter
increases, this scattering angle decreases, until it reaches a minimum value of 230°, after which it increases
again. This corresponds to the secondary rainbow angle.

Exercises

1. In a previous exercise an expression was derived for the angle rotated by the Class 4 ray, namely ∆4 . Plot
this function using a computer for 𝑏 between 0 and 1 in steps of 0.01. Your plot should correspond to Figure
120.

2. Find the derivative of this function, the impact parameter for which the scattering angle is a minimum, and
obtain a formula for the minimum scattering angle (secondary rainbow angle).

This analysis only considered positive impact parameters. If the negative impact parameters are analysed,
the scattering angles are just reversed. It is important to realize that since the Class 3 rays have a rainbow
angle less than 180°, whereas the Class 4 rays have a rainbow angle of more than 180°, for us to see both
types of rainbow in the same region of sky (i.e. on the same side of a line through the antisolar point), the
Class 4 rays must originate from incident light with an impact parameter which is opposite in sign to those
for the Class 3 rays, and be reflected around the droplet in the opposite sense.
 235

Figure 121 Droplet showing the range of possible scattering angles for Class 3 and 4 rays. The angles where no
scattering occurs corresponds to the Alexander dark band.

Taking into account all scattered Class 3 and Class 4 rays, for all possible impact parameters, it is interesting
to note that none of these rays are scattered between 130° and 138° (and also 222° to 230°). This is illustrated
in Figure 121 and means that no light is scattered back to the observer for angles between the primary and
secondary rainbow angles, which accounts for Alexander's dark band.

19.4. The colours of the rainbow

1.40

1.39
Refractive index

1.38

1.37

1.36

1.35

1.34

1.33

1.32
200 300 400 500 600 700 800 900 1000
Wavelength (nm)

Figure 122 Refractive index of water at 0 °C as a function of wavelength, from Handbook of optical constants, M.J.
Weber (Ed.), CRC Press, 2003. p.381.

The theory given so far explains why light is scattered from water droplets predominantly at the rainbow
angle, but does not account for the rainbow's most striking feature: its bands of colour. These were first
explained by Newton following his prism experiments in 1666. The colours of the rainbow are due to the fact
that the refractive index of water is not constant, but rather a function of the colour (wavelength) of the light.
This fact is called dispersion. Figure 122 shows the refractive index of water as a function of wavelength for
 236

the region around the visible spectrum. Note that the visible region is not well defined, but we shall take it
from 400 nm (blue) to 700 nm (red).

Newton discovered that white is not a true colour, but a combination of the colours of the rainbow (visible
spectrum). Since the rainbow angle depends on the refractive index, each colour has a slightly different
rainbow angle. Therefore the different colours are most intense at slightly different scattered angles, causing
bands of colour to be observed.

Exercises

1. (a) Plot the primary rainbow angle ∆3,min as a function of 𝑛, for 𝑛 between 1.33 and 1.35 in steps of 0.0025.
(b) Is the primary rainbow angle larger for red or blue light? (c) Use this to explain the order of the colours of
the primary bow. (d) Estimate the angular width of the visible primary rainbow. (Note: Newton obtained
values of 137° 58' for red light and 139° 43' min for violet light. Your values may differ slightly.) (e) The sun's
rays are not all parallel as originally assumed, but make a small range of angles, equal to the angular size of
the sun in the sky. If the sun has a diameter of 1.392 million km, and is 150 million km away from earth,
calculate the angle it subtends in the sky, and hence obtain a better estimate of the primary rainbow's width.
(Note: Newton estimated the rainbow width with this correction to be 2° 15', in good agreement with his
observations.

2. (a) Now plot the secondary rainbow angle as a function of 𝑛, for 𝑛 between 1.33 and 1.35 in steps of
0.0025. (b) Is the secondary rainbow angle larger for red or blue light? (c) Explain why the order of colours in
the secondary bow are reversed, even though the primary and secondary rainbow angles change in the same
way with refractive index. (d) Is the secondary rainbow wider or narrower than the primary rainbow?

19.5. Beyond geometrical optics: explaining the supernumerary arcs

Figure 123 Rays having impact parameters (for a droplet of unit radius) above and below the impact parameter
corresponding to the rainbow angle exit the droplet parallel to one another after having traversed different paths
and path lengths. These rays may interfere constructively or destructively, depending on the phase difference. Unlike
other features discussed previously, this effect is dependent on the size of the droplet.
 237

Since the rainbow angle represents the minimum scattering angle of light incident on a droplet for all impact
parameters, it follows that light incident on the droplet just below, and just above the value corresponding
to the rainbow angle will leave the drop (at different points) parallel to one another, after having travelled
different paths inside the drop. This is illustrated in Figure 123. According to the laws of geometrical optics,
the intensities of these rays should simply be added. However, Thomas Young showed with his double slit
experiments (in 1801) that light has a wave-like nature and undergoes interference. He also pointed out (in
1803) that this could account for the supernumerary arcs. This is because the ray intensities cannot simply
be added, but rather account must be taken of their phase difference as well.

Figure 124 Supernumeraries are as a result of interference effects of light.

The two rays considered earlier can interfere with each other either constructively or destructively,
depending on the difference in their path lengths. The path length difference increases as the impact
parameter changes away from the value that corresponds to the rainbow angle. Since the path length
difference must increase to a value corresponding to a phase difference of 180° before the rays interfere
destructively for the first time, this occurs for some scattering angle which is a little larger than the rainbow
angle. Hence, supernumerary arcs occur on the edge of the rainbow, on the brighter side (opposite that of
Alexander's dark band) which corresponds to larger scattering angles. The formation of the supernumeraries
as an interference phenomenon is illustrated by Figure 124.

The details of the supernumeraries of a rainbow are dependent on the droplet size: the larger the droplets,
the faster the phase difference increases for a given change in impact parameter, and the narrower are the
supernumerary arcs. For droplets larger than about 1 mm diameter, the supernumerary arcs are difficult to
detect. Supernumerary arcs are most often seen near the top of a rainbow, higher up in the clouds where
the water droplets are smaller.
 238

19.6. Light (insight) on the Alexander dark band

Both the Descartes/Newton and the improved Young model of the rainbow predict that no light is scattered
into the Alexander dark band, at least as far as Class 3 and 4 rays are concerned. However, this region is not
completely dark, and the solution to this problem does not lie in considering higher class rays. The sharp
transition of the brightest area at the rainbow angle, to total darkness in the Alexander dark band cannot
occur, since light does not cast sharp shadows. Light confined to an opening in a barrier, or even just confined
to a certain region without a barrier, always spreads out: this is called diffraction and is another consequence
of the wave-like nature of light. The treatment of diffraction is a subtle and difficult problem in mathematical
physics, and the subsequent developments of the theory of the rainbow were stimulated mainly by efforts
to solve it.

In 1838 George Airy applied the Huygens-Fresnel theory of diffraction to the problem. This theory regards
every point on a wavefront to be a source of secondary wavelets which travel outwards at the same speed
of the wave, and after some time the new wavefront of the wave is given by the tangent to all the wavefronts
of the wavelets. To apply this method requires knowing the amplitude at each point on the wavefront
initially, which is difficult to determine. The initial wavefront used by Airy for the primary rainbow was a
surface inside the droplet normal to the Class 3 scattered rays, and with an inflection point at the rainbow
angle. The wave amplitudes along this wavefront were estimated through standard assumptions in the
theory of diffraction, and Airy was then able to express the intensity of the scattered light in the rainbow
region in terms of a new mathematical function. This function was initially called the rainbow integral, but is
today better known as the Airy function. The intensity distribution predicted by the Airy function is analogous
to the diffracted intensity appearing in the shadow of a sharp straight edge: on the lighted side of the primary
bow there are oscillations in intensity that correspond to the supernumerary arcs. Although both Young's
and Airy's models both predict supernumerary arcs, they differ with regard to the positions and widths.

Figure 125 Intensity as a function of scattering angle, predicted by the Descartes, Young and Airy models of the
rainbow.
 239

Furthermore, Airy's model predicts that the maximum scattered intensity occurs at an angle slightly larger
than the rainbow angle of Descartes and Young. Those two theories predict an infinite brightness at the sharp
interface between the rainbow angle and Alexander dark band, whereas Airy's diffraction model predicts
that as the scattering angle is decreased i.e. as we move outwards from the antisolar point, the scattered
intensity maximizes (at an angle slightly larger than the rainbow angle) and then tapers off smoothly to zero,
penetrating the Alexander dark band to some extent. Figure 125 illustrates these results.

Airy's theory, with the scattered intensity distribution described above, applied to a single wavelength
(colour) of light. The rainbow itself consists of a superposition of all the colours. The extent to which the
colours overlap, and how they are perceived requires some knowledge of colour vision. Suffice it to state
that rainbows formed from larger droplets (diameters of a few mm) have less colour overlap and appear
most pure, whereas tiny droplets (diameters of about 0.01 mm) produce a rainbow with colour overlap to
the extent that the rainbow appears almost white.

19.7. A new direction (polarization)

Figure 126 Predicted internal and external reflection percentages for the glass-air interface 𝒏𝒈𝒍𝒂𝒔𝒔 ≈ 𝟑/𝟐 and 𝒏𝒂𝒊𝒓 ≈
𝟏. TE (transverse electric) represents light polarized perpendicularly to the plane of incidence, while TM (transverse
magnetic) represents light polarized parallel to the plane of incidence. Note that the Brewster angle (denoted by 𝜽𝒑 )
differs for internal and external reflections, and that a critical angle for total internal reflection only exists for internal
reflections. (From Introduction to optics 2nd Ed., Pedrotti and Pedrotti, Prentice Hall, 1993. p. 413).

Another aspect of light to be considered is its state of polarization. Recall that light is a transverse wave which
has oscillations perpendicular to its direction of propagation. Direct sunlight is unpolarized, and the direction
of the electric field oscillations in the plane perpendicular to its propagation direction is random. The intensity
of light reflected and refracted at an interface, and the states of polarization of the reflected and refracted
 240

light, depend on the polarization of the light as determined by the Fresnel equations. These equations predict
the reflected and refracted light intensities when light is incident on an interface, taking into account the
possible polarization of the incident light. Consider light in a material with refractive index 𝑛1 incident on a
𝑛
boundary with a material with refractive index 𝑛2 . We define 𝑛 = 𝑛2. Then for light polarized perpendicular
1
to the plane of incidence, the fraction of intensity reflected is

2
cos 𝜃𝑖 − √𝑛2 − sin2 𝜃𝑖
𝑅⊥ = ( )
cos 𝜃𝑖 + √𝑛2 − sin2 𝜃𝑖

while for light polarized parallel to the plane of incidence, the fraction of intensity reflected is

2
𝑛2 cos 𝜃𝑖 − √𝑛2 − sin2 𝜃𝑖
𝑅∥ = ( ) .
𝑛2 cos 𝜃𝑖 + √𝑛2 − sin2 𝜃𝑖

The percentage of light reflected for a glass-air interface, both internally (with the incident light in the glass,
𝑛 < 1) and externally (with the incident light in the air, 𝑛 > 1) predicted using these equations is given in
Figure 126.

Consider light travelling inside the droplet, and striking the water-air interface. Then 𝑛 < 1. For small angles
of incidence, the reflectivity is (approximately) independent of the polarization, and has a value of about 2%.
For an angle of incidence greater than or equal to the so-called critical angle, defined as the incident angle
for which the refracted angle is 90°, all the light intensity is reflected irrespective of its polarization. Between
these two extremes, the different polarizations are reflected different amounts, and at the so called Brewster
angle59 none of the light polarized in the plane of incidence is reflected, resulting in the reflected light being
fully polarized perpendicular to the plane of incidence. By co-incidence, the Brewster angle for the water-air
interface is close to the angle with which a ray corresponding to the rainbow angle is internally incident on
the droplet surface. The component of light polarized in the plane of incidence is almost completely
transmitted during the internal reflections, whereas the component polarized perpendicularly is also largely
transmitted, but also reflected. The rainbow is formed from this internally reflected light (after the ray again
strikes the droplet surface and is then mostly refracted out of the droplet) and is thus composed of light
which is polarized perpendicularly to the plane of incidence. Light from the top of the rainbow is polarized
horizontally, and should not be visible through polaroid sunglasses which have a vertical polarizing direction.

Exercises

1. Derive the reflectance formulae for the two polarization states given in this section from the Fresnel
equations of section 13.6.5.

2. (a) Consider a reflection of light incident from a material with lower index of refraction. Take for example
a reflection from air (𝑛1 ≈ 1) into water (𝑛1 ≈ 4/3), which is an external reflection since 𝑛 > 1. Plot the
reflectance as a function of the angle of incidence. Make the plot for both cases of polarization on the same
set of axes. (b) Use your graph to estimate the Brewster angle. (c) Use your graph to estimate what
percentage of the light intensity is reflected when light first strikes a droplet at a impact parameter

59 Named after David Brewster who discussed its significance in 1815. It is defined as the angle for which the incident and refracted
light are perpendicular to each other.
 241

corresponding to the rainbow angle. (d) Now consider a reflection of light incident from a material with
higher index of refraction. Take for example a reflection from water (𝑛1 ≈ 4/3) into air (𝑛2 ≈ 1), which is an
internal reflection since 𝑛 < 1. Plot the reflectance as a function of the angle of incidence. Make the plot for
both cases of polarization on the same set of axes. Note that the reflection is undefined for both cases for
angles of incidence greater than the critical angle. (e) Use your graph to estimate the Brewster angle. (f) Show
1−𝑛 2
that in all possible cases if the light is incident perpendicularly then 𝑅 = (1+𝑛) , and evaluate this for the air-
water and water-air interfaces.

3. (a) Obtain the angle of incidence for the first internal reflection in a water droplet, for an impact parameter
corresponding to the rainbow angle. (b) Calculate the Brewster angle for light moving from water into air. (c)
How do the values in (a) and (b) compare? (d) Calculate the intensity (as a percentage of the incident
intensity) of a Class 3 ray which is initially polarized perpendicularly to the plane of incidence. Do the
calculation in 3 steps, corresponding to the three times the ray is refracted/reflected. Pay careful attention
to whether the reflection is internal or external, and remember that any light which is not reflected is
transmitted. (e) Calculate the intensity (as a percentage of the incident intensity) of a Class 3 ray which is
initially polarized parallel to the plane of incidence.

19.8. Particle rainbows

Figure 127 (a) Scattering of a light atom by a heavy atom, due to electromagnetic forces. (b) Intensity as a function
of scattering angle, showing an increased intensity at the maximum scattering angle (rainbow angle), as well as
supernumerary arcs resulting from the interference of the wave-aspect of the scattered atoms.

Quantum mechanics suggests that particles have a wave nature. Therefore, it may be expected that the
theory developed for rainbows may have some application in analysing particles being scattered. This turns
out to be the case. Particles scattering occurs at some optimum angle (the rainbow angle) and many of the
other features of the rainbow are present as well. A quantum mechanical treatment for atomic and nuclear
scattering was developed in 1959 by Kenneth Ford and John Wheeler. An atomic rainbow was first observed
by Hundhausen and Pauly at the University of Bonn in 1964, in the scattering of sodium atoms by mercury
atoms. Their graph of scattering intensity shows a maximum (at the rainbow angle) and two
 242

supernumeraries. The rainbows measured in these experiments can be used to obtain information about the
interatomic forces (in a similar way to which one could determine the refractive index of water from a normal
rainbow).

19.9. The exact solution

Recall that Airy had to estimate the amplitude of the wave along the initial wavefront for his calculations. His
estimation is valid only for larger droplets, which have a circumference more than about 5000 times the
light's wavelength. For the droplets in fog and mist, this ratio is only about 100.

However, when Maxwell published “A Treatise on Electricity and Magnetism” in 1873, the problem was (at
least in principle) solved, since the solution to the rainbow is just that of a electromagnetic plane wave
scattered by a sphere. In 1908, Gustuv Mie and Peter Debye solved this problem, giving a solution in the form
of an infinite series. This series could converges very slowly, and it was not until 1975 that Vijay Khare
obtained an “exact” solution by numerically summing the truncated infinites series using a computer.

Figure 128 Intensity as a function of scattering angle, predicted by the Airy model, complex angular momentum
approximation solution and numerically derived exact solution of the rainbow. The solution is for parallel-polarized
light scattered by droplets with a circumference equal to 1500 wavelengths of the light.

Instead of numerically summing the truncated infinite series, one can use a method originally developed by
Henri Poincaré and G. Watson to transform this type of solution into a more rapidly converging series. This
technique is called the Watson transform, or the complex angular momentum method. Although photons do
not have a rest mass, they do carry momentum (and energy). In the process of being scattered by a water
droplet, clearly the concept of angular momentum should play a role. What is much more abstract is why
one should also consider complex values of the angular momentum. 60 After the Watson transformation,

60A similar abstraction occurs when one allows the refractive index 𝑛 of a material to take on complex values in order to model
absorption of light in that material.
 243

instead of working with the infinite series, one can consider only a few special points, called saddle points
and poles, in the complex angular momentum plane. Each real saddle point corresponds to an ordinary ray
of light as we have been considering them up to now. Complex saddle points correspond to “complex rays”
scattered beyond the rainbow angle i.e. those attributed by Airy to diffraction. Physically, this is attributed
to the effects resulting from evanescent (quickly decaying) waves created at points where the light strikes
the water-air interface. The poles are attributed to surface waves that result when light strikes the droplet
tangentially i.e. at 90°. This wave travels along the water-air interface, decaying and hence radiating out
energy. It is even possible for this wave to enter the droplet (with a refracted angle equal to the critical angle
for the water-air interface), move along a chord and then be refracted again along a tangent to the droplet.
This corresponds to the surface wave taking a “short cut” through the droplet! This method was applied to
rainbows in 1937 by Van der Pol and Bremmer, and later in 1969 with much more success by Moysés
Nussenzveig.

Airy's solution, the complex angular momentum approximate solution and Vijay Khare's numerically derived
“exact” solution are compared in Figure 128. Part of the failure of Airy's model was the approximation to the
initial wavefront amplitude, but further problems include that account is not taken of the 180° phase change
of internally reflected waves and due to his theory ignoring higher class rays than Class 3 and 4. The complex
angular momentum solution is a good approximation to the exact solution, and of a much more simple form.

19.10. Possible websites for more information

 About rainbows

http://my.unidata.ucar.edu/content/staff/blynds/rnbw.html

 Circles of Light: The Mathematics of Rainbows

http://www.geom.uiuc.edu/education/calc-init/rainbow/

 Patterns in Nature: Light and optics: Rainbows

http://acept.la.asu.edu/PiN/rdg/rainbow/rainbow.shtml

 Museum Victoria: Rainbows

http://www.museum.vic.gov.au/scidiscovery/rainbows/index.asp

 Supernumerary Rainbows above Boulder, Colorado

http://www.jal.cc.il.us/~mikolajsawicki/rainbows.htm