Class Notes for Modern Physics, Part 1

J. Gunion
U.C. Davis

9D, Spring Quarter

J. Gunion
 What is Modern Physics?

The study of Modern Physics is the study of the enormous revolution in our
view of the physical universe that began just prior to 1900.
At that time, most physicists believed that everything in physics was
completely understood. Normal intuition and all experiments fit into the
context of two basic theories:

1. Newtonian Mechanics for massive bodies;

2. Maxwell’s Theory for light (electromagnetic radiation).

Consistency of the two required that there be a propagating medium (and,

therefore, a preferred reference frame) for light.
However, even a little thought made it clear that there was trouble on
the horizon. And then came many new experimental results that made it
clear that the then-existing theoretical framework was woefully inadequate
to describe nature.
In a relatively short period of time, physicists were compelled to adopt:

J. Gunion 9D, Spring Quarter 1

1. the theory of special relativity based on the idea that there was no
propagating medium for light (so that light traveled with the same speed
regardless of the “frame” from which the light was viewed);
2. the theory of quantum mechanics, according to which the precise
position and precise momentum of a particle cannot both be determined
simultaneously.
In fact, one must think of particles not as particles, but as waves, much
like light.
3. At the same time, experiments made it clear that light comes in little
quantum particle-like packets called photons.
4. In short, both particles and light have both a particle-like and wave-like
nature.

It is useful to focus first on the inconsistencies of the “ether” picture

and of the above-outlined naive picture of space and time. This will lead us
to the theory of special relativity.
The latter inconsistencies are revealed by thinking carefully about Galilean
transformations between coordinate systems that underpinned the pre-
relativity view of space and time.

J. Gunion 9D, Spring Quarter 2

 Before proceeding, let me just emphasize that in this course we will
be embarking on an exploration that has been repeated in a certain sense
several times now.
Indeed, the business of looking for inconsistencies in existing theories
now has a long history of success, beginning with the development of special
relativity, general relativity, and quantum mechanics. We have learned not
to be arrogant, but rather to expect that the best theories of a given
moment are imperfect and to look for difficulties (perhaps subtle ones) or
extensions that are suggested by thought experiments that push the theories
into a new domain.
As an example, the development of the Standard Model of fundamental
interactions (that you may have heard of) began with the realization that
the theory that was developed to explain the weak interactions would violate
the laws of probability conservation when extended to high energies.
In fact, nowadays, we have many arguments that suggest that the
Standard Model is itself little more than an “effective” theory valid at the
energy scales that we have so far been able to probe. It has undesirable
features when we try to extend it to higher energies (e.g. from the scale of
the masses of the new W and Z bosons to the Planck mass scale that is
some 16 orders of magnitude larger).

J. Gunion 9D, Spring Quarter 3

 The ether picture for light propagation

• At the end of the 19th century, light waves were an accepted fact, but
all physicists were “certain” that there had to be a medium in which the
light propagated (analogous to water waves, waves on a string, etc.).

• However, the “ether” in which light propagated had to be quite unusual.

The speed of light was known to be very large (the precise value we
now know is c = 3.00 × 108 m/s). A medium that supported this high
speed had to be essentially incompressible (i.e. something vastly more
incompressible than water, and even more vastly incompressible than air).

• And yet, it was clear that light traveled over great distances from the
stars, implying that this ether extended throughout a large section of the
universe.
This means that the planets, stars, galaxies, . . . , were traveling through
this ether according to Newton’s laws without feeling any frictional,
viscosity, . . . , type of effects.

J. Gunion 9D, Spring Quarter 4

• Well, for anyone thinking about this nowadays, this is obviously ridiculous.
But, at the end of the 1900’s it was impossible for physicists to accept
the fact that there was no ether medium in which light traveled and it
was bizarre to imagine that light could travel through “vacuum”, despite
the fact that Maxwell’s equations were most easily understood in this
context.

• We will shortly turn to the Michelson-Morley (MM) experiment performed

in 1887 in which MM set out to demonstrate the existence of the ether.
We will learn that they failed. To show how entrenched thinking can
become, it should be noted that Michelson (who was quite a brilliant
guy) never believed the result of his experiment and spent the next 20
years trying to prove his original result was wrong. He failed, but provided
ever-increasing accuracy for the precise speed of light.

• It would be natural to presume that Einstein’s theory of special relativity

was a response to this experiment.
But, in fact, he stated that when he developed his theory he was
completely unaware of the MM result. He simply was thinking of

J. Gunion 9D, Spring Quarter 5

 Maxwell’s theory of light as a medium independent theory and asking
about its consequences.
This is not totally implausible given the fact that the MM experiment
was performed in the “back-woods”, frontier town of Cleveland Ohio
(some would say that the MM experiment is still the most important
thing, other than some baseball greats, to come out of Cleveland). And
communications were not so hot back in those days.

To understand the ideas behind the MM experiment and to set the stage
for how we discuss space and time in an “inertial” frame, we must consider
how to relate one frame to another one moving with constant velocity with
respect to the first frame.
The 1900’s view of this relationship is encoded in “Galilean transformations”.

J. Gunion 9D, Spring Quarter 6

 Galilean transformations
A coordinate system is to be thought of as:

1. a system of “meter” sticks laid down in the x, y, z directions throughout

all of space.

2. a single universal clock time that applies throughout all of space (i.e. is
the same no matter where in space you are).

An event is thus specified by its location in the (t, x, y, z) space.

But, now suppose that there is another person moving with constant
velocity in the original coordinate system in the positive x direction.
His coordinates will be related to (t, x, y, z) by:

x0 = x − vt
y0 = y
z0 = z
t0 = t, (1)

J. Gunion 9D, Spring Quarter 7

 where the last equation, according to which both observers can use the
time, or set of clocks, is particularly crucial.

This relation between frames is illustrated below.

Clicktoaddtitle

Fig. 1-2, p. 4

Fig. 1-1, p. 4

Figure 1: The Galilean Transformation.

Of course, if we are tracking an object moving in the x-direction in the

two coordinate systems, we may compare its velocity and acceleration as
viewed in the two coordinate systems by taking derivatives of the first

J. Gunion 9D, Spring Quarter 8

 equation above to obtain (using dt = dt0 from above)

dx0 dx
u0x ≡ = − v ≡ ux − v , and
dt0 dt
dux dux
a0x ≡ = ≡ ax, . (2)
dt dt

Note that the accelerations are the same, which is consistent with the
idea that the force causing the acceleration should be the same as viewed
by the two different observers (given that forces depend on separations
between objects which will be the same in the two different frames):

Fx = max = ma0x = m0a0x = Fx0 , (3)

where we assumed m = m0 is frame independent. The fact that

Fx = max and Fx0 = ma0x have the same form in the two different
reference systems is called covariance of Newton’s 2nd law.

But, there is already a problem with covariance in the case of light. Light
cannot be subject to the same covariance rules without conflicting with
the idea of an ether in which it propagates.

J. Gunion 9D, Spring Quarter 9

 In particular, imagine you are at rest in the ether and look at your
reflection in the mirror. No problem – it just takes a very short time for
the light to travel to the mirror and back.

However, now suppose you are on a rocket ship moving with velocity
v > c with respect to the ether. Then the light traveling with speed c in
the ether never makes it to the mirror!!

• Thus,

1. either we are forced to give up the general concept that motion with
constant velocity is indistinguishable from being at rest (i.e. there must
be a preferred rest frame), or
2. the Galilean transform equations eq. (1) are wrong.

It is the latter that is true.

These thoughts led to the Michelson-Morley experiment. They showed

that either there is no ether or that the earth is not moving through the
ether (a very geo-centric point of view by then, since it would make much
more sense if the earth was moving with its orbital velocity through an ether

J. Gunion 9D, Spring Quarter 10

that was a rest with respect to the galaxy or universe as a whole).
The Michelson Morley Experiment

The experimental arrangement for the MM experiment appears in the

diagram below.

Fig. 1-4, p. 8

Figure 2: The Michelson Morley experimental set-up.

In the pictured arrangement, the light (wave) is split by a half-silvered

mirror into two components, one traveling parallel to the earth’s motion,

J. Gunion 9D, Spring Quarter 11

 the other traveling perpendicular to the earth’s motion through the ether.
The time of travel for the horizontal light to and back from the mirror
will (Galilean assumed) be

L L
thorizontal = + . (4)
c+v c−v

The time of travel for the vertical light (which must actually be aimed
“up-stream” in order to return to the splitting mirror) is given by

2L
tvertical = √ . (5)
c2 − v2

From this, we find (for v = vearth ∼ 3 × 104 m/s and c ' 3 × 108 m/s)

v2
∆t = th − tv ' L for v  c or
c3
∆d ≡ c∆t ∼ 10−7 m for L = 10 m and (v/c)2 ∼ 10−8 . (6)

Although this is a very small number, an interferometer which measures

the interference between the vertical and horizontal light waves can be
sensitive to it.

J. Gunion 9D, Spring Quarter 12

 We now rotate the apparatus by 90◦ so that the roles of the two light
paths are interchanged. Since the waves are sensitive to the wave pattern
oscillation  
2π
sin (x − ct) (7)
λ
with t = th or t = tv , one sees a shift (relative to that if th = tv ) in
the constructive interference fringe corresponding to an angular amount
given by
2π 2π
(c2∆t) = 2∆d (8)
λ λ
(factor of 2 from fact that net time shift is twice the time difference
between wave arrival times in any one set-up). This would be noticeable
for λ ∼ f ew × 10−7m in a typical laboratory sized set-up using visible
light.

What did they see? No shift in the interference pattern.

The time was ripe for a new idea. Enter Einstein in 1905.

J. Gunion 9D, Spring Quarter 13

 Special Relativity

Postulates of special relativity:

1. The Principle of Relativity: All the laws of physics have the same form
in all inertial reference frames.
In other words, covariance applies to electromagnetism (there is no ether)
as well as to mechanics.

2. The Constancy of the Speed of Light: The speed of light in vacuum has
the same value, c = 3.00 × 108 m/s, in all inertial frames, regardless of
the velocity of the observer or the velocity of the source emitting the
light.
This postulate is in fact more or less required by the first postulate. If the
speed of light was different in different frames, the Maxwell equations
governing the propagation of light would have to be frame-dependent.
In fact, Einstein said he was completely unaware of the MM experiment
at the time he proposed his postulates. He was just thinking about the
theory of light as being absolute and frame independent.

J. Gunion 9D, Spring Quarter 14

These two apparently simple postulates imply dramatic changes in how we
must visualize length, time and simultaneity.

1. The distance between two points and the time interval between two events
both depend on the frame of reference in which they are measured.

2. Events at different locations that occur simultaneously in one frame are

not simultaneous in another frame moving uniformly with respect to the
first.

To see what exactly is true, we need to first think about how an

inertial reference frame is defined. We use a coordinate grid and a set of
synchronized clocks throughout all space.

As an aside, we should note that we are already questioning that such a

picture actually exists when looking at very tiny distance scales where
effects of quantum gravity are expected to enter.

An inertial reference frame is probably only an effective description that

is only valid up to the Planck mass scale.

J. Gunion 9D, Spring Quarter 15

 Clicktoaddtitle

Fig. 1-8, p. 13

Figure 3: Picture of an inertial reference frame.

Let’s return to the concept of time.

Example A

Suppose time were uniquely definable and the same in all frames.

Consider a (small) plane moving at speed v (and very close to ground)

in the +x direction relative to someone on the ground.

J. Gunion 9D, Spring Quarter 16

 x’=−D x’=0

The whole (P) plane picture above is moving with velocity v relative to the ground (G).

Figure 4: The frame for a plane moving relative to the earth.

When the plane is at x0 = 0 someone at x0 = −D flashes a light (these

are the plane’s coordinates). If time is universal then both P (plane) and
G (ground) agree that the light flashes at a certain time, say t = t0 = 0.
The time at which P thinks the light arrives at x0 = 0 (the plane never
moves from x0 = 0 – he is at rest in his coordinate system) is t0 = D/c
(assuming light travels with velocity c).
The time at which G thinks the light arrives at the plane would also be
t = t0 = D/c if time is universal. However, since the plane has moved

J. Gunion 9D, Spring Quarter 17

 by an amount
D
extra distance = v (9)
c
according to G while the light has been traveling, the G observer
concludes that the velocity of light is

distance D + vD
c
= D
= v + c. (10)
time c

Well, this contradicts Einstein’s postulates of relativity. It has to be that

the clocks in the G and P frames are not synchronized in the manner we
assumed or that distance scales are not the same in the two frames. In
fact, both apply.
Example B
• Consider 2 observers A and B that pass one another, with, say, B moving
with velocity v in the x direction relative to A who we envision is at rest
in “our” frame.
A burst of light is emitted as they pass one another. Each claims that the
light travels outward in spherical waves with velocity c, with the spheres
centered on themselves.

J. Gunion 9D, Spring Quarter 18

 A modern day application is that a terrorist dropping a bomb (that
immediately detonates) from a fast moving car might hope to quickly
leave behind the destruction and explosion. But, to the extent that
electromagnetic radiation was the only consideration, he would always be
at the center of the explosion no matter how fast he was moving in some
other frame.

• This is completely different from what one would conclude if light traveled
in a medium like water.

Consider two boats, one (A) at rest in a pond, the other (B) moving
rapidly (but without creating any wake) relative to the first boat.

B drops a rock in the pond as he passes A.

Because A is at rest in the pond, the ripples spread out in concentric

circles from his position and B, looking back, agrees. Indeed, he could
even go faster than the ripples, in which case they would never catch up
to him.

Putting Einstein’s visualization into mathematical language, we would say

J. Gunion 9D, Spring Quarter 19

 that the expanding spheres of electromagnetic radiation should obey

x2 + y 2 + z 2 = c2t2 for A; x02 + y 02 + z 02 = c2t02 for B . (11)

These are the equations defining how each sees the light fronts (in his
own frame) emanating from the initial flash.
Demanding that the transformation from the unprime to prime system
be a linear transformation1 with coefficients determined only by the
fundamental constant c and by the relative velocity of the two frames v,
and requiring that x0 = 0 must correspond to x = vt (see Fig. 1), there
is only one solution, the so-called Lorentz transformation:

x0 = γ(x − vt)
y0 = y
z0 = z v
t0 = γ(t − 2 x) , (12)
c
1
Very roughly, you don’t want anything else for time since otherwise you could get negative time corresponding to
positive time (for a quadratic relation) or even phases would be introduced if higher powers were employed. Once
linearity for the relation between t and t0 is chosen, the rest must also be linear. If you are a stickler for mathematical
precision, you could try looking at http://arxiv.org/pdf/physics/0110076.

J. Gunion 9D, Spring Quarter 20

 where
1
γ=q . (13)
v2
1− c2
Note that if v/c  1, then γ → 1 and we get back the Galilean
approximation.
But, if v/c → 1, then γ  1 and there are big changes.
Proof of above Lorentz transformation.
Write (since y 0 = y and z 0 = z in this situation, I ignore them)
x0 = γ(x − vt) , t0 = αt + βx , (14)
where α, β, γ are all to be determined and we have inserted in the first
equation the requirement that x = vt gives x0 = 0.
Now, by symmetry (i.e. who says which frame should be prime and
which unprime) it must also be true that the inverse transformations
are obtained by simply interchanging the roles of the prime and unprime
coordinates and changing the sign of v. Thus, we should also have

x = γ(x0 + vt0) . (15)

J. Gunion 9D, Spring Quarter 21

 (We will come back to the time relation.)
So, Einstein now says that for a light pulse starting at x = 0, t = 0 along
the +x axis we must have x = ct. However, since we have chosen the
prime frame position and clock so that this corresponds to x0 = 0, t0 = 0,
then we must also have x0 = ct0.
So, substitute these two statements into x0 = γ(x − vt) to obtain

v
ct0 = γ(ct − vt) = γ(1 − )ct . (16)
c

Next, substitute these two statements into x = γ(x0 + vt0) to obtain

0 0 v 0 v v
ct = γ(ct + vt ) = γ(1 + )ct = γ(1 + )γ(1 − )ct , (17)
c c c

where the last step employed the equation just above. The left and final
right side of the above equation are equal only if

2 1
γ = v2
. (18)
1− c2

J. Gunion 9D, Spring Quarter 22

 Now return to the time equation. Let us take x0 = γ(x − vt) and
substitute this into x = γ(x0 + vt0). The result is

x = γ(γ(x − vt) + vt0) . (19)

We can easily solve this for t0:

0 γ 2vt + (1 − γ 2)x v v
t = = γt − γx = γ(t − x) , (20)
γv c2 c2

2 v2 2
as claimed. In the above, we used the simple algebra that 1−γ = − c2 γ .

An important consequence

Using the Lorentz transform equations, we can easily show that x2 − c2t2 =
x02 − c2t02 , for any choices of x, t and the corresponding values of x0, t0.

A Note on Four Vectors

We can place the quantities ct, x, y, z into an array called a 4-
vector: x4 = (ct, x, y, z). The square of such a 4-vector is defined as

J. Gunion 9D, Spring Quarter 23

x4 · x4 ≡ c2t2 − x2 − y 2 − z 2. In the prime frame, x04 = (ct0, x0, y 0, z 0) and
x04 · x04 ≡ c2t02 − x02 − y 02 − z 02.
We see that a restatement of the Lorentz transformation equations,
equivalently Einstein’s frame-independent for the velocity of light, is to say
that the square of a 4-vector is frame-independent.
Time Dilation

A particular application of the Lorentz equations is time dilation. I will first

give a mathematical derivation and then a more intuitive approach.

Mathematical approach.

Consider an observer S 0 moving with velocity v with respect to S.

Consider a clock at rest in the S 0 rest frame and located at x0 = 0,

corresponding to x = vt.

Suppose it ticks at t0 = 0 and t0 = T 0. Now use

0 v v v2 t
t = γ(t − 2
x) = γ(t − 2
vt) = γt(1 − 2
)= , (21)
c c c γ

J. Gunion 9D, Spring Quarter 24

 to conclude that the t0 = 0 tick occurs at t = 0, while the t0 = T 0 tick
occurs at t = T = γT 0.
So, the person at rest in S sees an interval between ticks that is larger
than the person who is moving with the clock sees.
This is what we call time dilation. The time interval between ticks seen
by observer S 0 who is a rest with respect to the clock is called proper
time. Observer S, who sees this clock moving past him swears that it is
ticking more slowly.
We could also have derived this result by using the inverse Lorentz
transform equations:

x = γ(x0 + vt0)
0 v 0
t = γ(t + 2 x ) , (22)
c

in particular the second one above. In this approach, we say that the
clock sits at some fixed x0 and ticks at t0 = t01 and then at t0 = t02. Using
the second equation above, we find

T = t1 − t2 = γ(t01 − t02) ≡ γT 0 . (23)

J. Gunion 9D, Spring Quarter 25

 Is there any more intuitive way to understand this bizarre result?
Intuitive Approach
Consider the situation depicted below. An observer sits at rest in a
freight car that moves with velocity v relative to the ground. He sends a
light signal to a mirror on the top of the car that bounces back to him.
The time between sending and receipt is defined as the tick of the clock.
He concludes that

0 0 distance traveled 2d
T = ∆t = = . (24)
speed of light c

Figure 5: The “mirror” clock.

Meanwhile, the observer on the ground sees the freight car move a
Fig. 1-10, p. 15
J. Gunion 9D, Spring Quarter 26
 distance 12 v∆t = 12 vT (where T is the full tick time) between the time
the signal is sent and the time the signal hits the top of the car. The
geometric picture then allows us to compute 12 vT as follows. The actual
q
distance traveled by the light for this half of the trip is ( 21 vT )2 + d2,
1
and this must equal the amount of distance 2
cT that light can travel
when moving with velocity c: i.e.

1 1
( vT )2 + d2 = ( cT )2 , (25)
2 2

from which we obtain

4d2 2d
2
T = , or T = γ = T 0γ . (26)
c2 −v 2 c

Note how we employed the fact that the light is moving with velocity c
according to both observers.

Now, you might ask if there is any experimental verification of this bizarre
result. The answer is many!

J. Gunion 9D, Spring Quarter 27

 An example: the decay of an unstable elementary particle called the muon
(denoted µ).

Clicktoaddtitle

Fig. 1-11, p. 17

Figure 6: Muon decay in the muon (a) and the earth (b) frame.

The µ has its own internal clock that determines how fast it decays.
On average, the lifetime of the µ is denoted τ (I stick to the book,
even though I would rather have called this τ 0 — so you have to switch
prime and unprime relative to above discussion.) Measurements of

J. Gunion 9D, Spring Quarter 28

 the exponential decay curve e−t/τ (t is the muon’s internal time) give
τ = 2.2 µs for muons at rest.

However, if the µ is moving with v = 0.99c (γ = 7.1) with respect to

the earth (as typical of µ’s contained in cosmic rays), the earth thinks
the muon clock is running much slower and has lifetime γτ = 16 µs.

Thus, on average its travel distance in the earth rest frame is given by
vγτ = 4700 m, far longer than without the time dilation factor — as
far as the muon observer is concerned, he only moves vτ = 650 m on
average before decaying.

Length Contraction

We imagine laying out a ruler in the moving frame S 0 from x0 = 0 to

x0 = L0. How long does this ruler appear to the stationary viewer S?

Mathematical Approach

We wish to determine the length as measured by S at some fixed time,

J. Gunion 9D, Spring Quarter 29

 say t = 0. So go to the Lorentz transform equation

x0 = γ(x − vt) (27)

and set t = 0. Then, the above equation implies that x0 = 0 corresponds

to x = 0, while x0 = L0 corresponds to L0 = γx ≡ γL. As a result, we
find that
L = L0/γ . (28)
Since γ > 1, the apparent length, L, of the ruler to observer S is smaller
than the length L0 as seen by observer S 0.
Thus, if a ruler at rest in S 0 is zooming past S with velocity v, S swears
that the ruler is shorter by the factor γ1 .
The quantity L0 measured by the observer S 0 who is at rest wrp to the
ruler is called the proper length, sometimes denoted Lp. It is defined just
like the proper time was defined as the time between ticks of a clock as
measured by someone at rest wrp to the clock.
An Intuitive Approach
This approach (which follows the book) will make use of the time dilation
result already derived.

J. Gunion 9D, Spring Quarter 30

 We consider two stars which are separated by a proper length Lp as
observed by someone on earth (which is essentially, i.e. for the purpose
of this discussion, at rest wrp to the stars).
Now we consider a moving space ship. According to the observer on
earth the space ship takes a time ∆t = Lp/v to travel between the stars.
Because of time dilation, the space traveler measures a smaller time of
travel between the two stars: ∆t0 = ∆t/γ. (Which is just another way
of writing ∆t = γ∆t0, the form written earlier when discussing time
dilation.) The space traveler interprets this time of travel (which, to
repeat, is the time he measures) as corresponding to a distance of travel

0 ∆t Lp
L = v∆t = v = . (29)
γ γ

Return to the µ example

Suppose we have a mountain at rest on the earth (as in the original
picture).
The height of this mountain as measured by someone on earth is the
mountain’s proper height, call it hp.

J. Gunion 9D, Spring Quarter 31

 Similarly, let us call the amount of time that the muon believes it takes
to reach the bottom of the mountain (when moving vertically) tp, the
proper time interval in the muon rest frame.
The amount of time according to the earth observer that the muon takes
to traverse the height hp (moving vertically) is hp/v and this must be
related to tp by the time dilation factor: i.e. γtp = hp/v or

1 hp
tp = . (30)
γ v

From the muon’s point of view has he reached the bottom of the
mountain after this time tp?
For this to be true, it must be that vtp is the height of the mountain as
seen by the µ. Computing, we have
 
1 hp hp
vtp = v = , (31)
γ v γ

which is precisely the contracted mountain height as viewed by the µ.

J. Gunion 9D, Spring Quarter 32

 So the two observers agree about when the µ has passed from the top
of the mountain to the bottom of the mountain!

Note that the two observers had to agree about their relative velocity for
this to work out.

Well, I suspect your head is swimming at this point with confusion about
when to use γ and when to use 1/γ. It takes practice, and so you must do
a bunch of problems to get the hang of this.
The crucial rule is to always remember that:

1. proper time refers to the time measured at a fixed location in some frame
(fixed location means the clock is at rest in that frame).

2. proper length refers to a length measured (at some given time) for an
object by an observer at rest with respect to that object.

Applications
There are many other interesting applications of all this. Here, I will
focus on the relativistic Doppler shift.

J. Gunion 9D, Spring Quarter 33

 This is a particularly important application as it is the Doppler “red”-shift
that we use to tell us that the universe is expanding from something like
an initial big bang.

You are all familiar with the usual Doppler shift in which the pitch of sound
for a whistle on a train headed towards you has a higher pitch than the
whistle sound when the train is moving away.

This is because successive waves emitted by a source moving towards you

are closer together than normal because of the advance of the source —
and since their separation is the wavelength of sound, the corresponding
frequency is higher.

The formula in the case of sound is probably something you have derived
in an earlier course. !
v
1+ c
f = f0 V
, (32)
1− c
where f0 is the frequency of the sound as measured by the source itself,
f is the frequency as measured by the observer, c is the speed of sound,
v is the speed of the observer (+ for motion toward source), and V is
the speed of the source (+ for motion toward the observer).

J. Gunion 9D, Spring Quarter 34

 This classical Doppler effect evidently varies depending upon whether the
source, the observer, or both are moving.
• This does not violate relativity because sound does travel in a medium
— unlike light.
• Since light does not travel in a medium, the light wave Doppler effect
will be different.
It can be derived using the concepts of time dilation. My derivation is
an alternative approach to that given in the book. Note, in particular,
that the book assumes that the source is moving towards the observer.

Imagine a light source as a clock that ticks f0 times per second and emits
a light wave peak at each tick. The proper time in the source rest frame
between ticks is t0 = 1/f0.
Consider a source at rest and an observer moving away from it with
velocity v. The interval between ticks as seen by this observer is given
by time dilation: t = γt0
As viewed by the observer, he travels the distance vt away from the
source between ticks.
Thus, each tick takes a time vt/c longer to reach him than the simple

J. Gunion 9D, Spring Quarter 35

 time t between ticks.
The total time between the arrival of successive peaks (successive ticks)
is then
vt v v
T = t+ = t(1 + ) = γt0(1 + )
c c
s c
1 + vc 1 + vc
= t0 q = t0 v . (33)
2
1 − vc2 1− c

The corresponding frequency of the ticks or wave peaks is just the inverse:
s s
v v
1 1 1− c
1− c
f (receding case) = = v = f0 v (34)
T t0 1+ c
1+ c

• Now, although I derived this for an observer moving away from the
source, the same result applies if the source moves away from the
observer.
• You should also note that the frequency shift depends only on the
relative velocity of the source and observer. One does not need to
reference any medium in which light travels.

J. Gunion 9D, Spring Quarter 36

 • Since wavelength and frequency are inversely related, f λ = c, the shift
in λ obeys the inverse formula.

Example:
Determining the speed of recession of the Galaxy Hydra.
A certain absorption line that would be at λ0 = 394 nm were Hydra at
rest, is shifted to λ = 475 nm according to observations on earth.
We use s
v
1+ c
λ= v λ0 (35)
1− c
to find that
λ2
v λ20
−1
= λ2
= 0.185 . (36)
c +1
λ20
Therefore, Hydra is receding from us with a velocity of v = 0.185 c =
5.54 × 107 m/s.
***Show tape #42 on space-time diagram starting at 20 min mark. ****

Lorentz Velocity Transformation

J. Gunion 9D, Spring Quarter 37

 This is obtained by taking derivatives of the Lorentz transform formulae of
eq. (12), in close to analogy to what we briefly described for the Galilean
case.
From eq. (12), we have

0 0 v
dx = γ(dx − vdt) dt = γ(dt − dx) (37)
c2
Thus,
dx
dx0 dx − vdt −v ux − v
u0x ≡ = v = dt
v dx
= vux . (38)
dt0 dt − c2
dx 1− c2 dt
1− c2

Similarly, if the object we are examining has some y direction velocity

(but the relative velocity of the frames is still along the x axis), we have

dy 0 = dy (39)

and the same formula for dt0 as above, from which we obtain
dy 0 dy uy
u0y ≡ = v

= vux ,

(40)
dt0 γ dt − c2 dx γ 1 − c2

J. Gunion 9D, Spring Quarter 38

 and a similar result for u0z .
Important notes:
1. These results reduce to Galilean forms if v  c.
2. If ux = c and uz = uy = 0, we find
0 c−v
ux = v = c, (41)
1− c
which is to say that if something (like light) moves with velocity c in
one frame, it also moves with c in the other frame.
This was obviously required for consistency of the Lorentz transforms
with Einstein’s original postulate.
Another example of this is to consider the case where uy = c, ux = 0
and uz = 0. Our velocity transform equations give
c
u0y = , u0x = −v , (42)
γ

from which we compute

2 2 c2 v2
u0x + u0y 2
=v + 2 2
= v + c (1 − ) = c 2
. (43)
γ2 c 2

J. Gunion 9D, Spring Quarter 39

 Again we get the same velocity for light, c, in both frames.

An Application: the relative velocity of 2 space ships.

Fig. 1-19, p. 30

Figure 7: Two spaces ships moving towards one another relative to a

stationary observer.

Consider two space ships A and B. A moves with velocity 0.75 c in the
+x direction relative to a stationary observer. B moves with velocity

J. Gunion 9D, Spring Quarter 40

 0.85 c in the −x direction relative to the same stationary observer. Find
the relative velocity of B with respect to A. The configuration is pictured
above.
One key phrase here is “Find the relative velocity of B with respect to
A.” What these words mean is that A looks at B from his (A’s) rest
frame. What this means for solving this problem is that we must lock
one frame, let us choose S 0 to A, i.e. we choose S 0 to be A’s rest frame.
For S we choose the stationary observer. So, in applying the formulas, we
identify the speed of the S 0 frame with the speed of A, i.e. v = 0.75 c
wrp to the earth.
Now, we apply the formulas imagining that both S and S 0 are looking at
B. We want the velocity of B in the rest frame of A. This is u0x in the
context of the formulas.
We have
ux − v −0.85c − 0.75c
u0x = ux v = (−0.85c)(0.75c)
= −0.9771c . (44)
1− c2 1− c2

Note that the naive Galilean result would be −1.6c, whereas the correct
result is smaller than c in magnitude.

J. Gunion 9D, Spring Quarter 41

 Space-Time and Causality

If we focus on just the x and ct coordinates, we can picture space-time

as below.

Fig. 1-23, p. 32

Figure 8: Two-dimensional picture of space-time.

The coordinates x and ct in frame S are drawn along perpendicular axes.

J. Gunion 9D, Spring Quarter 42

 An event E is characterized by its (x, ct) location. The trajectory of a
particle is characterized by its “world-line” in this space.

Fig. 1-25, p. 34

Figure 9: Causal division of space-time

Since the velocity ux of the particle is defined by

∆x c
ux = c = , (45)
∆ct slope

J. Gunion 9D, Spring Quarter 43

 and since ux ≤ c (nothing moves faster than light), we see that slope ≥ 1
is required, so that the blue “world-line” can at most be 45◦ from the
vertical ct axis. At light pulse with ux = c would travel along the 45◦
line. Nothing can ever travel into the section of the diagram below the
45◦ line. Thus, the picture gets divided up into the future, the past, and
unreachable regions denoted “elsewhere” in the picture above.
If you want to know how things in frame S appear to a moving observer,
S 0, you need to draw the manner in which his coordinate axes relate to
those of S. We will not go through the details, but the result appears
in Figs. 8 and 9. The larger the relative velocity, the more collapsed the
axes become. There is a cute Carl Sagan dvd that illustrates this. (DVD
#8, Chapter 4, starting about 20:35 with motorscooter.)
We can now discuss the concept of causality. Consider two events in
space time, as depicted in Fig. 10. One occurs at (ct1, x1) the other
occurs at (ct2, x2). (We will assume that y1 = y2 and z1 = z2.) Compute
the four-vector
∆x4 = (ct2 − ct1, x2 − x1) (46)
and its square:

(∆s)2 ≡ ∆x4 · ∆x4 = (ct2 − ct1)2 − (x2 − x1)2 . (47)

J. Gunion 9D, Spring Quarter 44

 Fig. 1-24, p. 33

Figure 10: Depiction of two events in space-time

We know that this quantity is actually independent of frame. That we

stated is a consequence of the Lorentz transformation that you can easily
check. Thus,

(∆s0)2 ≡ ∆x04 · ∆x04 = (ct02 − ct01)2 − (x02 − x01)2 = (∆s)2 . (48)

J. Gunion 9D, Spring Quarter 45

 This frame-independent quantity, (∆s)2, has a very important interpretation.
If (∆s)2 ≥ 0, then event E1 can influence E2 by virtue of the fact that
a light ray emitted from x1 at time t1 could reach x2 before or at time
t2. We say that “E2 is causally related to E1”.
If (∆s)2 > 0, we call the interval time-like. In this case, even something
traveling slower than light could reach E2 from E1 in time to influence
E2 .
If (∆s)2 = 0, we call the interval light-like. In this case, only light
emitted from E1 could influence E2.
If (∆s)2 < 0, the interval is called space-like. In this case, there is no
possibility for E1 to influence E2 since nothing can travel faster than
light.
Lorentz Transformations for the 4-vector components
We can rewrite the Lorentz transform equations of eq. (12) in a form
which displays a great symmetry between the roles of ct and x.
0 v
x = γ(x − ct) (49)
c
v
ct0 = γ(ct − x) . (50)
c
J. Gunion 9D, Spring Quarter 46
 Example of direct use of Lorentz Transformation:
Discuss an event as seen in two different frames.

Anna and Bob are in identical spaceships 100 m long each, with distances
from the back labeled along the sides. Prior to taking up space travel
in retirement, Bob and Anna owned a clock shop, and they glued the
leftover clocks all over the walls of their ships.
As Bob’s space ship is flying along, Anna passes him with relative speed
v = 0.8 c, headed in the same direction (say +x direction). Just as the
back of Anna’s ship passes the back of Bob’s ship, the clocks on both
ships read 0. At this same instant, Bob Jr., on board Bob’s space ship,
is aligned with the very front edge of Anna’s ship. He peers through a
window in Anna’s ship and looks at the clock.
(a) In relation to his own ship, where is Bob Jr., and (b) what does the
clock he sees read?

Solution:

J. Gunion 9D, Spring Quarter 47

 Let us choose Bob as the stationary observer (frame S). Anna will be in
frame S 0, which then has v = +0.8c relative to S (Bob).
The key here is “at the instant shown, Bob Jr. on Bob’s space ship”.
What this means is that Bob Jr. is doing something (looking at Anna’s
front end) at time t = 0 (the time on Bob’s space ship), which will not
be the same as the time on some clock opposite him on Anna’s space
ship. (Only the backs of the two passing space ships agree about the
time.) However, we do know where Bob Jr. is looking! He is looking at
x0 = 100 m (i.e. the front of Anna’s space ship).
(a) So where is Bob Jr as he is doing the looking? Use x0 = γ(x − vt) =
γx at t = 0 with γ = 1/0.6 to compute x = x0/γ = 60 m. This result
is the same as you would get just thinking about Lorentz contraction of
Anna’s space ship as seen by Bob’s space ship.
(b) What is the time on Anna’s clock opposite Bob? Use

0 v
t = γ(t − 2 x)
v c
= −γ 2 x since t = 0 as above
cv
= −γ 2 x0/γ using our result above for x
c
J. Gunion 9D, Spring Quarter 48
 0.8c
= − x0
c2
1
= −0.8 100 m
3× m/s 108
= −2.66 × 10−7 s (51)

Causality Example: The death ray problem.

At a certain instant, two identical rocket ships pass one another (headed
in opposite directions) so that at a certain instant the cockpit of one is
lined up with the tail of the other. (a) Can a tail gunner of number 1 fire
a death ray at pilot 2? (b) Could copilot 2 send a signal to tail gunner
2 at the velocity of light enabling tail gunner 2 to kill pilot 1? Ignore
separation between rockets and assume perpendicular firing of death rays.

We will examine in rest frame of 2 with 1 zipping by: i.e. 1=S 0 and 2=S.

(a) At t0 = 0 suppose tail of 1 is at x0 = 0 and pilot and copilot of 1 are

at x0 = d. Use x = γ(x0 + vt0) and t = γ(t0 + cv2 x0) to conclude that
x = 0 and t = 0 is where tail 1 is, and so tail of 1 can kill pilot 2.

J. Gunion 9D, Spring Quarter 49

 (b) First, we need to find out where pilot 1 is from perspective of the
tail gunner of ship 2. Pilot 1 is at x0 = d at t = 0 when pilot 2 is killed.
Using x0 = γ(x − vt), we find x = x0/γ = d/γ (length contraction).
So, it looks like there is hope. But, copilot 2 still has to get the signal
to tail gunner 2. He has a certain amount of time to do so. The signal
must reach tail gunner 2 by the time the front of space ship 1 passes the
tail gunner 2. The time to reach this configuration (in ship 2 in S frame)
is t = (d − d/γ)/v and this must be > d/c if the signal light ray from
copilot 2 is to reach the tail gunner in 2 early enough. This requirement
thus is:

d(1 − γ1 ) d
> or
v c
1 v
1− > or
γ c
s
v v2
1− > 1− 2 or
r c r c
v v
1− > 1+ , (52)
c c

which is impossible.

J. Gunion 9D, Spring Quarter 50

 This would have been obvious from the perspective of pilot 1 as he looks
at tail gunner 2. At t0 = 0, tail gunner 2 is at x = d (as always) and
using x = d = γ(x0 + vt0) (inverse LT) at t0 = 0 gives x0 = γd , equivalent
to Lorentz contraction. The physical picture is that the tail gunner 2 at
x0 = d/γ is already well behind the pilot of ship 1 located at x0 = d so
that there was clearly no hope to begin with!
Causality interval approach.
event 1: kill at x = x0 = t = t0 = 0.
event 2: passage of tail 2 past pilot 1.
This latter passage occurs when x0 = d and x = d, since the pilot of 1
is at x0 = d and the tail gunner of 2 is at x = d. Thus, we must have

d = γ(d − vt) or
d
tv = d − or
γ
d 1
t = (1 − ) . (53)
v γ
So now compute the causality interval from the tail gunner 2 point of
view (i.e. using the unprime frame coordinates for the two events, the

J. Gunion 9D, Spring Quarter 51

 2nd being hypothetical for the moment) as

c2d2 c2
 
1 1
c2(t2 − t1)2 − (x2 − x1)2 = (1 − )2 − d2 = d2 (1 − )2 − 1
v2 γ v2 γ
(54)
If this is negative, ⇒ no causal connection. We thus ask if

1 v2
(1 − )2 − 2 < 0? or
sγ c
v2 v
1 − 1 − 2 < ? or
sc c
v v2
1 − < 1 − 2 ? or
r c r c
v v
1 − < 1 + ? which is true. (55)
c c

Thus, no causal connection possible. The tail gunner of 2 cannot kill the
pilot of 1 in response to tail gunner of 1 killing the pilot of 2.

J. Gunion 9D, Spring Quarter 52

 Causality and Lorentz Transforms applied to the µ example
Let us define event 1 (E1) as the passage of the µ by the top of the
mountain. We will define earth coordinates S and muon coordinates S 0
so that this occurs at x = x0 = 0 and t = t0 = 0. (We measure both x
and x0 in the vertical downwards direction.)
Let us define event 2 (E2) as the passage of the µ by the bottom of the
mountain. In system S, we know that this happens at t = hp/v (hp
recall is the proper mountain height measured in the S frame). We also
know that it happens a x = hp.
What are the coordinates of E2 as viewed by the µ in his frame?
According to LT we have x0 = γ(x − vt) and t0 = γ(t − cv2 x). So for E2
we have
0 hp
x = γ(hp − v ) = 0 , (56)
v
(which makes sense since the µ does not change his location in his own
frame) and

0 hp v hp v2 hp
t = γ( − h )=γ
2 p
(1 − 2
)= , (57)
v c v c γv

J. Gunion 9D, Spring Quarter 53

 in agreement with the time dilation factor for S vs. S 0.
So, now let us compute (∆s)2 in the two different frames. (Recall, all
coordinates = 0 for E1.)
2
c2
  
hp
(∆s)2S = c2 − h2p = h2p −1 (58)
v v2
2
c2 v2 c2
    
hp
(∆s)2S 0 = c2 − 02 = h2
2 p
1− = h2p − 1 .(59)
γv v c2 v2

We get the same result in both frames and, further, (∆s)2 > 0 implies
that E1 and E2 are causally connected. Well, we already knew this since
the traveling µ itself was the means of establishing this causal connection!

J. Gunion 9D, Spring Quarter 54

 Relativity II: Mass, Energy and Momentum
Velocity transform equations and momentum conservation

Suppose we on earth see 2 rocket ships passing one another.

y
A (before) A (later) A (before) A (later)

B (later) B (before) B
(before and after)

x
Equal but opposite velocity frame B rest frame

Figure 11: Depiction of two rockets passing one another: a) in equal but
opposite velocity frame, each having velocity of magnitude v 0; b) in B rest
frame, where velocity of A appears to be v.

J. Gunion 9D, Spring Quarter 55

 Consider first the equal-but-opposite velocity frame (the left-hand picture).
Someone on rocket ship A throws a billiard ball of mass m downwards
(in the A rest frame) and someone in B throws an identical billiard ball
upwards (in the B rest frame) with exactly the same vertical velocity,
i.e. uyA = −uyB . (We will assume that these y direction velocities are
 c in magnitude.) The billiard balls collide elastically and go back to
the respective starting points with exactly reversed y direction velocities.
Obviously, we have
∆uyA = −∆uyB , (60)
and after multiplying by m

m∆uyA = −m∆uyB , (61)

or using the usual definition of momentum as m × velocity,

∆pyA = −∆pyB , ⇒ ∆(pA + pB )y = 0 . (62)

Momentum is conserved and Newton is happy! Note that we do not

care about the exact magnitude of the y velocity as viewed from the
equal-but-opposite rocket ship velocity frame. It will not be the same as
the y velocities discussed below in the B rest frame.

J. Gunion 9D, Spring Quarter 56

 But, if momentum is conserved in one frame then by covariance (i.e.
all physical laws the same in all frames), momentum should also be
conserved if we view this same process in the rest frame of rocket ship
B (2nd picture in Fig. 11). In this frame, we choose rocket ship A
to carry the prime coordinates and rocket ship B to carry the unprime
coordinates.
The situation described is that

0
uyA = −u , and uyB = u , (63)

which is to say that in their respective frames each thinks he has thrown
the ball with velocity of magnitude u. But, if we look entirely from B’s
rest frame (the unprime frame), we have (using time dilation and the
0
fact that yA = yA for relative frame velocity in the x direction)
   0   
y
 ∆y A
  ∆y A
 1 y 0
 y
|uA| = 
 = 
0
=  uA  6
= |u B| , (64)
∆t   γ∆t   γ
p
where γ = 1/ 1 − v 2/c2 with v being the velocity of rocket ship A
along the x direction in B’s rest frame.

J. Gunion 9D, Spring Quarter 57

 y0
uA
(Alternatively use uyA = vux 0
, the inverse velocity transform equation,
γ (1+ 2A )
c
0
with ux
A = 0 since A’s ball only has y velocity in A’s rest frame.)

As a result, we find that

1
∆uyA = 2 |uyA| = − ∆uyB . (65)
γ

and if we define momentum by pyA = muyA and pB = muyB we do not

get ∆pyA = −∆pyB . To keep momentum conservation, observer B must
claim that mA 6= mB when looking at A’s billiard ball in his B rest
frame. What we need is an extra factor of γ. Sometimes, we write

mA = γmB , (66)

where mB is the mass of the billiard ball when viewed from the B rest
frame and mA is the mass of an identical billiard ball that is nearly at
rest in the A rest frame but being viewed from the B rest frame.
u)y  c
In the above discussion, we implicitly kept our vertical velocity (~
so that in the B rest frame mB ' m0, where m0 is the mass of the

J. Gunion 9D, Spring Quarter 58

 billiard ball as measured in its own rest frame. This approximation is fine
if v is substantial (as implicitly assumed) so that γ = √ 1 2 2  1.
1−v /c
How do we generalize?
In general, we can retain momentum conservation provided we define

p)y
(~ = m(total velocity) × (velocity of mass in y direction)
= γ(u)m0(~u )y , (67)

and similarly for the x and z directions.

Here, the notation m0 is reserved for the proper mass of the billiard ball,
defined to be the mass of the billiard ball as measured in the billiard ball
rest frame.
Hopefully, this discussion is sufficient motivation for the general result
that
p
~ = γ(u)m0u ~ (68)
where
1
γ(u) = q , (69)
u2
1− c2

J. Gunion 9D, Spring Quarter 59

 with u = |~
u| being the absolute value of the total billiard ball velocity as
viewed from some frame. (In the above discussion, u ~ = v x̂ + (~ u)y ŷ '
v x̂.)
If we return to our original “equal-but-opposite” rocket ship velocity
frame, then the billiard balls have equal total velocities, say v 0,2 in the
observer’s rest frame and momentum will be defined (keeping the small
uy approximation) using m = γ(v 0)m0, with γ(v 0) = r 1 2 , since in
0
1− v 2
c
that frame each rocket ship (and the billiard ball each throws with small y
direction velocity) is moving with velocity v 0 with respect to the observer.
Thus, momentum will still be conserved in that frame as well.
Nowadays, we tend to back away from the use of the concept of a
“frame-dependent mass”, m = γ(u)m0 for total velocity u, and simply
say that momentum must include the extra factor of γ(u) in its definition.
The two points of view are algebraically equivalent.

Newton’s 2nd Law

Given the above definition of the momentum that is conserved in the

02
“ ”
2 0 0 2 0
p
2 2
Compute v from velocity addition: v = 2v /(1 + v /c ) ⇒ v = c c/v − c /v − 1 .

J. Gunion 9D, Spring Quarter 60

 absence of force, p
~ = γ(u)m0u~ , in the presence of force it is natural to
define force by
d~
p
~
F = . (70)
dt
This equation reduces to the non-relativistic Newton’s law when v  c
so that γ → 1.
Note that because γ(u)m0 increases with the object’s total velocity, u,
becoming ∞ at u = c, it is never possible to accelerate a particle past
the speed of light, or even to the speed of light.

Relativistic Energy

We start with (use one-dimensional derivation – force and motion along x

axis) Z x2 Z x2
dp
W = F dx = dx . (71)
x1 x1 dt
We can rewrite using

dp dp dx dp
dx = dt = dpu = duu . (72)
dt dt dt du
J. Gunion 9D, Spring Quarter 61
 Inserting into eq. (71), we get

uf
dp
Z
W = udu , (73)
ui du

where ui and uf are the initial and final velocities at locations x1 and
x2. Now,  
dp d  m0 u  m0
= q = 3 / 2
. (74)
du du 2

1−u c2 1− u2
c2

Inserting into eq. (73) and assuming ui = 0 gives

uf
m0 u m0 c2
Z
2
W =   3/ 2
du = r − m 0 c . (75)
2 2
0
1 − uc2 u
1 − c2f

At this point, it is conventional to drop the subscript f and simple write

W in terms of the particle velocity u. The book also drops the subscript
0 on m0 and so m will henceforth denote the intrinsic or proper mass.

J. Gunion 9D, Spring Quarter 62

 We then have,

2 2 1
W = γ(u)mc − mc , with γ(u) = q . (76)
u2
1− c2

Often, γ(u) is simply written as γ, but you must remember that it is the
particle’s velocity in your frame that goes into this form and not some
relative velocity of two different frames.

Interpretation of W

How should we interpret W ? In the non-relativistic case, W would be

the kinetic energy, K. We can check this correspondence for small u/c
by expanding:
1 1 u2
q '1+ 2
(77)
2
1 − uc2 2c

from which we obtain

1
W ≡K' mu2 (78)
2
in agreement with the usual kinetic energy formula.

J. Gunion 9D, Spring Quarter 63

 Since K is a change in energy, this formula suggests that mc2 and
γ(u)mc2 should be thought of as the rest frame energy and the energy
in motion, respectively.
Thus,

γ(u)mc2 = rest energy + kinetic energy = total energy ≡ E . (79)

If this is correct then mass is a form of energy. This is dramatically

confirmed in many ways, as we shall see.
An example
A proton has kinetic energy equal to half its rest mass energy. (a) What
is the proton’s speed? (b) What is its total energy? (c) Determine
the potential difference ∆V through which the proton would have to be
accelerated to attain this speed.
(a) K = (γ(u) − 1)mpc2 = 12 mpc2 ⇒ γ(u) = 1.5. For γ(u) ≡
√ 1 2
= 1.5, ⇒ u = 0.745c.
1−(u/c)

(b) E = γ(u)mpc2 = 1.5(1.67 × 10−27 kg)(3 × 108 m/s)2 = 2.25 ×

10−10 J = 1409 MeV.

J. Gunion 9D, Spring Quarter 64

 (c)
K = (γ(u) − 1)mpc2 = 0.5mpc2
= 0.5(1.67 × 10−27 kg)(3 × 108 m/s)2
= |q∆V | = (1.6 × 10−19 C)∆V (80)

⇒ ∆V = 470 M V .
Classically, much less ∆V would have sufficed (291 M V ), but that
increasing γ(u) factor implies more and more is needed as u gets closer
to c.

The relationship between energy, momentum and mass

We have E 2 = γ 2(u)m2c4 and p2c2 = γ 2(u)m2c2u2. Using these inputs

we find  2
u
E 2 − p2c2 = γ 2(u)m2c4 1 − 2 = m2c4, . (81)
c
Since m is an intrinsic property of the particle, the right-hand side is
always the same, no matter what frame we examine E and p in.
If we were to set m = 0, we get E = pc. This is in fact the relationship
between energy and momentum that applies to the particles of light
called photons and to other massless particles.

J. Gunion 9D, Spring Quarter 65

 In any case, the above invariance implies that

p4 = (E/c, px, py , pz ) (82)

is a 4-vector in the sense we defined earlier since its square

p4 · p4 = E 2/c2 − p ~ = m2 c2 ,
~·p (83)

is an invariant independent of frame, just like x4 · x4 = c2t2 − ~

x·~x
is. Based on this fact, a reasonable guess would be that E/c and p ~
transform from frame to frame using the same Lorentz transform type of
equations as apply to ct and ~
x! This is indeed the case.
Proof We need one basic identity that you can check:

0 1 uxv
γ(u ) ≡ q = (1 − )γ(v)γ(u) . (84)
u0x 2 +u0y 2 +u0z 2 c2
1− c2

This follows algebraically from the velocity transform equations, eqs. (38)
and (40). Using the above identity, we compute, again using the velocity

J. Gunion 9D, Spring Quarter 66

 transform equations, eqs. (38) and (40),
0 0 0
px ≡ γ (u )mux
ux − v
„ «
ux v
= 1− 2 γ (v )γ (u )m “ ”
c ux v
1 − c2
„ «
v
= γ (v ) γ (u)mux − γ (u)mc
c
„ «
vE
= γ (v ) px − , (85)
cc

which has the same form as eq. (49) with x replaced by px and ct
replaced by E/c! Similarly, we find
E0 γ (u0)mc2
≡
c „ c «
ux v
= 1− 2 γ (v )γ (u)mc
„ c «
v
= γ (v ) γ (u)mc − γ (u)mux
„ «c
E v
= γ (v ) − px (86)
c c

which has the same form as eq. (50) with x replaced by px and ct

J. Gunion 9D, Spring Quarter 67

 replaced by E/c! For the y and z component momenta, we expect no
change. Let us check.
0 0 0
py ≡ γ (u )muy
„ «
ux v uy
= 1− 2 γ (v )γ (u )m “ ”
c γ (v ) 1 − ux v
c2
= γ (u)muy
= py . (87)

Note: If E and p are constant in frame S in the absence of applied

forces, then the above transformation equations imply that they are also
constant in frame S 0. (Just take a time derivative of the two sides of
the equations.) The physical law that says “energy and momentum are
constant in the absence of applied force” is thus frame independent,
provided E and p are defined as we have done.
In other words, the fundamental postulate of relativity is obeyed by the
physical laws pertaining to momentum and energy.
For a group of particles, we would sum over their individual momenta and
energies, but that would not change the above statements. If the sum of
energies and the sum of momenta remain unchanged in one frame due
to the absence of force, these sums will remain unchanged in all frames.

J. Gunion 9D, Spring Quarter 68

 Energy Conservation

To repeat, it is natural that E as defined is the fundamental quantity,

as opposed to the kinetic energy, K, since it is E conservation that is
independent of frame. In contrast, K may be conserved in one frame
and not in another frame (or not conserved in any frame).

A really simple example of why we must deal with the total energy E is
provided in elementary particle physics. Experimentally, it is possible to
collide an electron and a positron (coming together with equal magnitude
but oppositely directed velocities — the center-of-mass frame) to make
a proton and an antiproton at rest:

e+e− → pp . (88)

In the initial state, most of the energy resides in the kinetic energies of
the e+ and e−, whereas in the final state all of the energy is contained in
the mass of the p and p. Kinetic energy is thus converted to rest mass
(energy). And, the opposite is also true! — think nuclear fission bomb.

J. Gunion 9D, Spring Quarter 69

 Another example is illustrated below.
v v v=0 for both blocks

BEFORE AFTER

Figure 12: Depiction of two blocks of equal mass with spring attached one,
colliding to create a single object with compressed (massless) spring and
motionless blocks. (Imagine there are little latches that catch and hold the
blocks together when the spring becomes compressed.)

Where did the kinetic energy go? Obviously, it went into the compression
of the spring. Einstein says that initially Ei = 2mc2 + 2K. If we view the
two boxes plus compressed spring as one total object, with zero kinetic
energy, then kinetic energy is clearly not conserved, but energy would be
conserved provided we simply define

Ef = M c 2 , (89)

J. Gunion 9D, Spring Quarter 70

 where M must account for not only the initial block masses, but also the
kinetic energy that was converted to spring compression energy. That is,
M > 2m. In fact,
2K
M = 2m + 2 (90)
c
is the appropriate definition.
Sometimes, the combined object can have mass smaller than the objects
of which it is composed. This happens if there is some force that attracts
the objects towards one another. We write
X
2
Mbound systemc + EB = mi c2 (91)
i

where EB is the binding energy. An example of this situation is provided

by the deuteron which is composed of a proton and a neutron. It is a
stable object precisely because EB > 0 which means that

Md < mp + mn (92)

and energy cannot be conserved by the process in which

d → p + n, (93)

J. Gunion 9D, Spring Quarter 71

 since kinetic energy can only be positive, whereas for the above process
we would need to have negative kinetic energies for the p and n.
The magnitude of the deuteron’s binding energy is (using the M eV type
units that you are all familiar with and writing mp and mn in these units,
where a factor of c2 is implicitly supplied)

EB (deuteron) = mp + mn − Md
= 938.27 MeV + 939.57 MeV − 1875.61 MeV
= 2.23 MeV . (94)

So, it is only sensible to suppose that E is conserved and that this

conservation is independent of frame. Then, for any process in any frame
we would have
X X
Ei = Ef . (95)
i=initial particles f =final particles

If we combine this with three-momentum conservation, and put all the

J. Gunion 9D, Spring Quarter 72

 energies together with their three momenta into 4-vectors, we get
X X
pi4 = pf4 . (96)
i=initial particles f =final particles

We have seen that going from one frame to another mixes up the
components of the p4 vectors (just like spatial rotation would mix up
the 3-vector components), but the equality between the two sides of the
equation would not be altered.
***At this point, please read the 4vector.pdf file on my web page and do
the extra problem at the end of this file.***

J. Gunion 9D, Spring Quarter 73

 The Quantum Theory of Light

Maxwell’s equations imply that E&M waves are possible:

Fig. 3−2, p. 67

Figure 13: Depiction of electromagnetic wave.

We know:
√
1. c = 1/ µ00, where µ0 and 0 were determined by magnetic induction
and static Gauss law experiments. ⇒ c = 3 × 108 m/s, which was
known to be speed of light from MM experiment.
⇒ light= E&M wave.

J. Gunion 9D, Spring Quarter 74

2. Wave nature of light verified by interference phenomena.

3. Maxwell Eqs. ⇒ same things should hold for E&M waves of much lower
f (larger λ since c = λf ).

4. Hertz verified interference and velocity c for radio waves. ⇒ confirmation.

Blackbody Radiation
Please read the book on this subject, the bottom line being the famous
formula due to Planck for ef , the power emitted by a perfect black body
per unit area, per unit frequency at temperature T :

3  
c 8πhf 1
ef = , (97)
4 c3 ehf /kB T −1

where Planck had to introduce a new fundamental constant h = 6.626 ×

10−34 J · s and kB is Boltzmann’s constant kB = 1.38 × 10−23 J/K.
A now-famous example of a blackbody spectrum is the cosmic microwave
background. Following the big-bang, the universe expanded very rapidly
and came to near thermal equilibrium, so that the spectrum of the residual

J. Gunion 9D, Spring Quarter 75

electromagnetic radiation should follow rather closely a perfect blackbody
spectrum. The spectrum, along with the Planck curve, is shown below.

Figure 14: The CMB spectrum compared to a perfect Planck blackbody

spectrum, assuming the current blackbody temperature of T = 2.725 K.

J. Gunion 9D, Spring Quarter 76

 The “residuals” contain important information. In particular, they tell
us about the early history of structure (e.g. galaxy cluster) formation and
the like. In particular, they tell us something about dark matter and dark
energy. But these residuals are extremely small perturbations about the
overall blackbody spectrum.
Planck was able to explain his formula by assuming:

1. The walls of the glowing cavity were composed of billions of “resonators”,

all vibrating at different frequencies.

2. However, Maxwellian theory says that an oscillator of frequency f could

have any value of energy and could change its amplitude continuously as
it radiated any fraction of its energy.
This would not have allowed Planck to derive his formula.
Instead, Planck had to assume that the total energy of a resonator
with mechanical frequency f could only be an integral multiple of hf ,
E = nhf (n = 1, 2, 3, . . .), where h was his new fundamental constant.
Further, the resonator had to only be able to change its energy by
dropping to the next lowest state, implying ∆E = hf always.

J. Gunion 9D, Spring Quarter 77

The derivation of the Planck formula is basically the computation of the
average energy of an oscillator using the standard statistical e−E/kB T
weighting for different energy values. According to Planck this is done
by assuming that E comes in units of hf so that one sums with above
weighting over E = nhf , n = 1, 2, . . . , ∞:
P∞ −nhf /kB T
n=0 (nhf )e hf
E= P∞ −nhf /k T = hf /k T . (98)
n=0 e
B e B −1

The other factors in Planck’s formula eq. (97) have a trivial origin that you
can read about.
How much energy for a resonator are we talking about? For green light,
we have

hc (6.63 × 10−34 J · s)(3 × 108 m/s)

hf = = = 3.68×10−19J = 2.3 eV .
λ 540 × 10−9 m
(99)
The Photoelectric Effect
The 1900 Planck formula and explanation would have remained an
obscure and esoteric idea except for the Photoelectric Effect, which was in

J. Gunion 9D, Spring Quarter 78

turn explained by (who else) Einstein. Einstein’s explanation required that
light come in little packets (photons) each with energy hf .
What is the photoelectric effect?

Fig. 3−14, p. 82

Figure 15: Photoelectric Effect apparatus.

Shine light on a metal surface and out come electrons.

How would you view this in Maxwell’s eyes?

J. Gunion 9D, Spring Quarter 79

 ~ and B
• E ~ come in and shake the e− and get it oscillating and it gets
out of metal.
~ and B
• In this picture, the bigger E ~ (higher intensity) the bigger the
shaking and the more energy you can give to the e−. In fact, the
energy density of an E&M wave is

Energy 1 1
u= = ~ 2+
0|E| ~ 2
|B| (100)
V olume 2 2µ0

which clearly gets bigger and bigger as E~ and B

~ get larger.
Conversely, by decreasing E~ and B ~ it should be possible to have the
light wave carry only a very tiny amount of energy.
Einstein looked at the experimental data and concluded this could not
be correct.
Instead, he proposed that light comes out in “photon” bundles of definite
energy E = hf .
An e− is very unlikely to be “hit” by more than one photon,
⇒ e− can only get so much energy — no more and no less (even for tiny
classical light intensity).

J. Gunion 9D, Spring Quarter 80

 In contrast, and to repeat, the classical expectation is that the maximum
kinetic energy of the ejected e−, Kmax, should be a function of intensity,
which means it should increase with increasing |E| ~ and |B|~ since more
energy would be deliverable to individual electrons.
Further, by decreasing the intensity, the classical expectation is that we
would eventually reach a point at which too little energy is delivered to
the e− to release it from the metal.
Neither is true experimentally.
We know for sure now that Einstein was right (of course). (This is what
he got his Nobel for and not for relativity!)
Back to the experiment.
Two classically unexpected results were seen:
e
1. Kmax (as measured by finding the “stopping” voltage, V = −Vs, that
repels the electron emitted from the emitter from reaching the collector
e
given by Kmax = eVs) depends on f and only f (i.e. not on intensity at
all).

2. For f < f0, where f0 depends upon the metal on which the light shines,
no e−’s come out.

J. Gunion 9D, Spring Quarter 81

 e
3. In fact, one finds that Kmax is directly proportional to f with slope given
by h.

Fig. 3−15, p. 82

Figure 16: Photoelectric Effect plots of (a) photocurrent vs. applied V for
e
two different light intensities and (b) of Kmax vs. f .

Enter Einstein
1. Assume that light comes in discrete photons, each carrying energy hf .

J. Gunion 9D, Spring Quarter 82

 2. Higher intensity is to be interpreted as more photons.
3. Only one photon light bundle is absorbed by any one single e− in the
metal.
4. If the energy of the photon can overcome the metal binding, then the
e− gets out. The picture is (our notation for a photon is γ)
Potential

 "#%$ &'



  ! 

 



Outside Metal Inside Metal

Figure 17: Einstein’s picture of photoelectric excitation of e− in a metal.

J. Gunion 9D, Spring Quarter 83

 The maximum energy of the ejected e− would then be Kmax e
=
hf − φ , where φ ≡ work function is the metal potential the e− must
overcome to get out — φ depends on the metal.

Fig. 3−17, p. 84

Figure 18: Dependence of Vs on the metal and on f .

e
Note that some electrons get energy < Kmax because the photon
does not give all its energy to the e− (energy imparted depends
on scattering angle, ....) and also the e− coming out may collide
with other stuff before trying to exit from the metal surface where it
encounters φ.

J. Gunion 9D, Spring Quarter 84

 5. This picture predicts that there is a certain minimum frequency f0 =
e
φ/h such that Kmax = 0, as observed experimentally.
An example:
e
Suppose for iron we observe that the frequency f0 for which Kmax =
eVs = 0 is f0 = 1.1 × 1015 Hz. What is the stopping potential for light
with λ = 250 nm?
Two steps are required:
e
I) Kmax = 0 ⇒ hf0 = φ or

φ = (4.14 × 10−15 eV · s)(1.1 × 1015 s−1) = 4.5 eV , (101)

where we have written h = 6.63 × 10−34 J · s in eV units using

1 eV = 1.602 × 10−19 J .
II) Given a λ, we know f = c/λ and then we also know that

e −15 3 × 108 m/s

Kmax = eVs = hf − φ = (4.14 × 10 eV · s) − 4.5 eV
250 × 10−9 m
= 4.96 eV − 4.5 eV = 0.46 eV . (102)

J. Gunion 9D, Spring Quarter 85

 Since electrons have charge e, we conclude that Vs = 0.46 V .

To summarize, our new picture of light is:

Fig. 3−16, p. 84

Figure 19: The classical Maxwell vs. Einstein picture of light.

In this picture, we can ask the following question:

How many photons per second emanate from a 10 mW , 633 nm laser?

J. Gunion 9D, Spring Quarter 86

 Answer: For each photon,

3 × 108 m/s
 
c
E = hf = h = (6.63×10−34 J ·s) = 3.14×10−19 J .
λ 633 × 10−9 m
(103)
To find the number of particle per unit time, we divide energy per unit
time by energy per particle:

number of particles 10 × 10−3J/s

= = 3.18×1016 photons/s .
time 3.14 × 10−19J/photon
(104)

X-rays

We use the term X-ray for E&M radiation with λ ∈ [10−2, 10] nm region
of the spectrum.

We want to use X-rays to demonstrate the particle nature of E&M

radiation.

J. Gunion 9D, Spring Quarter 87

 X-rays are produced by smashing high-speed electrons into a metal target.

Fig. 3−18, p. 87

Figure 20: Smashing an electron into a target to produce X-rays.

Any charged particle radiates E&M energy when it accelerates, and the
smashing gives violent acceleration so that much radiation is produced.

J. Gunion 9D, Spring Quarter 88

 Classically (a la Maxwell) one expects to see radiation produced over the
entire spectrum of wavelengths.

Experiment differs. There is always a cutoff wavelength, λmin below

which there is no radiation.

Fig. 3−21, p. 88

Figure 21: Typical x-ray spectrum showing λmin cutoff. (λ is determined

by using crystal diffraction to spread different λ’s onto different parts of a
film (in old days).

J. Gunion 9D, Spring Quarter 89

 The broad part of the spectrum agrees with Maxwellian theory, but the
cutoff requires the photon concept.

The explanation is as follows:

1. The minimum λmin arises when all of the energy of the incoming e−
is converted to energy of the outgoing x-ray photon.
2. In such a case, and assuming an accelerating voltage of V , we would
have
hc hc
eV = hf = , or λmin = . (105)
λmin eV
This is precisely what is observed. Of course, if the collision is less
than “perfect”, then less energetic, and therefore larger λ, photons
can come out. Also, the electron energy could be spread out over
many photons. In either case, the photon energy (or energies) will be
smaller and their wavelength longer. We can never do better than to
put all the incoming e− energy into a single outgoing photon.
3. In contrast, if there were no minimum energy (analogous to hf of the
photon) associated with E&M radiation of a given frequency, then
still large f ’s (and smaller λ’s) would have been possible by setting
eV = arbitrarily small number × f .

J. Gunion 9D, Spring Quarter 90

 The case was becoming stronger and stronger that this photon concept
was really correct.

Enter Compton (1922). (The above described X-ray wavelength studies

were by Bragg around 1912.)

The Compton Effect

Even more important than understanding X-rays themselves was what

Compton did with them.

He took energetic photons from a peak in the spectrum and collided

them with a stationary e−.

Classically (wave picture) we expect

1. e− oscillates with frequency of incoming E&M radiation and reradiates

with fout ≤ fin over broad range of fout.
2. The exact spectrum should depend on details of intensity and length
of exposure to the X-rays.

J. Gunion 9D, Spring Quarter 91

 Fig. 3−22, p. 90

Figure 22: Pictures of Compton scattering: wave point of view vs. particle
point of view.

J. Gunion 9D, Spring Quarter 92

 What Compton found was very different:
1. fout (equivalently, λout) depends only on fin (i.e. λin) and the angle
of scattering of the X-ray, θ.
fout does not depend on intensity or exposure time.
2. Further, fout < fin unless θ = 0.
The exact result found was (using λ for λin and λ0 for λout)

0 h
λ −λ= (1 − cos θ) , (106)
me c

where mhec = 0.00243 nm = 2.43 × 10−12 m is called the Compton

Wavelength. In a very real sense, it turns out to be the size of the e− as
seen by a X-ray photon.

This result can be understood in one and only one way:

1. Light must act as a particle with E = hf and p = E/c (as consistent
with zero mass limit of E 2 − p2c2 = m2c4 = 0).
2. The collision must be like relativistic billiard balls conserving energy
and momentum in the relativistic sense.

J. Gunion 9D, Spring Quarter 93

 This result of eq. (106) is a slight approximation in that it neglects the
binding work function energy of the target e− which is actually in a
graphite or similar target.
However, this is a good approximation for work functions in the eV range
given that X-rays have energies in the keV = 103 eV range.

Derivation of Compton Formula

We repeat the picture:

Fig. 3−24, p. 92

Figure 23: Compton scattering from the particle point of view.

J. Gunion 9D, Spring Quarter 94

 Before the collision, we have a photon with px = p and energy E = pc
~ = 0 and E = mec2.
hitting an electron at rest with p

electron with pye 0 = p0e sin φ, px

After the collision, we have an q 0 0
e = pe cos φ,
where p0e = |~
p0e|, and Ee0 = p0e2c2 + m2ec4.
We also have the scattered photon with pyγ 0 = −p0 sin θ, px
γ
0
= p0
cos θ
and E 0 = p0c. (I have defined p0 = |~
pγ |.)

Energy conservation thus gives:

E + mec2 = Ee0 + E 0 (107)

Momentum conservation in the x and y directions gives:

p = p0e cos φ + p0 cos θ (108)

0 = p0e sin φ − p0 sin θ , (109)
E hf h
and we need to keep in mind the energy relations: p = c
= c
= λ
,
0
q
p0 = hf
c
= h
λ0 and E 0
e = p0 2 c2 + m2 c4 .
e c

J. Gunion 9D, Spring Quarter 95

 Inputing the latter, and choosing θ to be some particular value defined
by the point at which I look at the scattered photon, we end up with 3
equations in 3 unknowns: φ, p0e and p0. We can solve these as follows:

1. Rewrite eq. (108) in the form

p − p0 cos θ = p0e cos φ . (110)

2. Rewrite eq. (109) in the form

p0 sin θ = p0e sin φ . (111)

3. Square these two equations and add ⇒ φ disappears and we end up

with
2
p0e = (p0)2 + p2 − 2pp0 cos θ . (112)
4. Using the photon energy relations, this is rewritten as

0 2 2
2h2f f 0
 
2 hf hf
p0e = + − cos θ (113)
c c c2

J. Gunion 9D, Spring Quarter 96

 while the energy conservation equation eq. (107) becomes

2
Ee0 = hf − hf 0 + mec2 ⇒ Ee0 = (hf − hf 0 + mec2)2. (114)

2 2 2
5. Now use Ee0 − p0e c2 = m2ec4 and the above expression for Ee0 to
solve (Problem 33, not assigned but would be good for you to do) and
obtain
0 h
λ −λ= 2
(1 − cos θ) . (115)
me c

Note the importance of using E = pc for the X-ray photon.

This is in fact something you know holds for Maxwell equation waves (I
hope you did this in the E&M course).

It is also what is required for the m = 0 limit of E 2 = p2c2 + m2c4.

Note that the reason that λ0 > λ (except at θ = 0 is that the e− takes
away some kinetic energy in this relativistic elastic collision so that the final
photon must have less energy and momentum and therefore smaller f 0,
implying larger λ0.

J. Gunion 9D, Spring Quarter 97

 We also note that p = λ h
and p = hfc
are equivalent for light. However,
only the first turns out to be correct when we consider the wave nature of
a particle with mass.

An example:
An X-ray photon of 0.0500 nm wavelength strikes a free, stationary
electron. The photon scatters at 90◦. Determine the momenta of the
incident photon, the scattered photon, and the electron.
Answer:
For the incident photon we have

h 6.63 × 10−34 J · s
p= = = 1.33 × 10−23 kg · m/s . (116)
λ 0.05 × 10−9 m

Inserting the other specified inputs, we solve eq. (115) for the scattered
photon’s wavelength:

0 (6.63×10−34 )
λ − 0.0500 × 10 m =9
(9.11×10−31 kg )(3×108 m/s)
(1 − cos 90◦)
0
⇒ λ = 0.0524 nm . (117)

J. Gunion 9D, Spring Quarter 98

 From this, we compute

h 6.63 × 10−34 J · s
p0 = = = 1.26 × 10−23 kg · m/s . (118)
λ0 0.0524 × 10−9 m

Finally, we may use eq. (112) to compute (in units of 10−23 kg · m/s)

(p0e)2 = (1.33)2 + (1.26)2 − 2(1.33)(1.26) cos 90◦ = 3.36 (119)

or p0e = 1.83 × 10−23 kg · m/s. The angle at which this electron is

traveling can then be obtained using eq. (111)

p0 sin 90◦ p0 1.26

sin φ = = = = 0.69 ⇒ φ = 0.76 rad = 43.55◦ .
p0e p0e 1.83
(120)

What is an elementary particle?

Consider the mechanical analogy below of two blocks held together by a

J. Gunion 9D, Spring Quarter 99

 very stiff spring.

Two Blocks with very stiff spring holding them together,

all at rest.
Figure 24: A composite system consisting of two blocks held together by
a very stiff spring.

If a billiard ball comes in and collides with this system in a fairly “soft” low
energy manner, then the two-block system will simply recoil as though it
were a single unified object (albeit, one with a non-spherically symmetric
nature). Energy and momentum conservation can be applied to the
process as if the two-block system had no internal structure.
However, if the billiard ball is sufficiently energetic, then the stiff spring
will start to oscillate in some way and the two-block system will not act as

J. Gunion 9D, Spring Quarter 100

 a single unit. Some of the incoming billiard ball energy will be absorbed
into the energy of the oscillation, which will amount to an increase in
mass of the two-block system. A Compton-like calculation would no
longer be applicable.
Similarly, the billiard ball itself could have internal structure (maybe it
has an interior that might heat up under the collision, . . . ).

In Compton scattering, no matter how energetic a photon we employ

(even photons with E ∼ 1011 eV ), there is no sign of any such internal
structure. The Compton formula always works.
We call particles that have no internal structure elementary particles. All
evidence to date is consistent with both the photon and the electron
being truly elementary (also termed “point-like”) particles.
In contrast, the proton we now know (by bombarding it with very
energetic photons, for example) is a composite object made of what are
called quarks. We will return to this picture late in the quarter.

Is E&M radiation a wave or a particle?

J. Gunion 9D, Spring Quarter 101

 Answer: it is both!
Which kind of nature it displays depends upon the experimental situation
considered, that is how the E&M wave or particle interacts with its
environment. This is partly determined by its own wavelength λ and
partly by the nature of the object it interacts with, in particular the size
D of the experimental apparatus or probe that “looks” at the photon.
We will find that a similar statement applies also to particles with mass
such as the electron. This will be the next subject we come to. But, for
now, let’s explore this issue in a bit more depth for E&M radiation.
An analogy
Consider a boat of length D in a lake with, let us say, 5 waves of
substantial amplitude but rather long wavelength, λ  D, moving
toward and eventually past the boat. The boat goes up one side of each
wave and rides down the other. Observers in the boat would easily know
that a wave consisting of several troughs and several peaks had passed
by, even if they were blindfolded.
Now suppose the situation is that we have 5 waves, but the wavelength
is such that λ  D. The observers in the boat would feel a sudden
impact from the peaks of the waves, but would certainly not be able to

J. Gunion 9D, Spring Quarter 102

 resolve the separation between the peaks. It would feel like some object
had collided with them. The boat responds to all parts of a λ  D
wave pattern at once, and not to different parts of the wave pattern at
different times.

Back to E&M radiation and photons

To see the particle nature of light, Compton scattering works very nicely,
but only if λ0 − λ is a significant fraction of λ is it experimentally easy
to see the change. In terms of the Compton size of the the electron
λe ≡ mhec , the shift from λ to λ0 would be an immeasurably tiny fraction
of the original λ if λ  λe. This is why Compton needed keV photons
to really reveal the particle nature of light — he needed large enough
energy so that λ ∼ λe. Thus, λe is the equivalent of the boat size D in
the earlier analogy.
If the change λ0 − λ ≡ ∆λ is very small compared to λ, ∆λ  λ,
then there is hardly any difference between the Compton result and the
classical result which says that λ would not change. This is an example of
the Correspondence Principle, according to which there is a limit of the
quantum theory in which it becomes indistinguishable from the classical
theory (in this case, Maxwell’s wave theory of light).

J. Gunion 9D, Spring Quarter 103

 Another aspect of taking λ large, is that as the wavelength increases, the
energy per photon decreases so that a given light intensity would have
to be produced by more and more photons, which would kind of become
continuous and everywhere in the wave. Intuitively, this implies that the
Maxwell wave description would become more and more accurate.

Back to wave vs. particle

Passing light through a single slit is a good way to reveal its wave nature.
But to see the diffraction, the wavelength has to be big compared to the
slit size, D. (If D  λ, then the wave, just like a particle, would just
pass straight through the slit without being modified significantly.)
In contrast, small λ compared to the electron Compton wavelength was
what Compton needed to reveal the particle nature. These two natures
of the light are not incompatible, but rather complementary.
To emphasize the wave limit, let us consider in a bit more detail light
passing through a single slit. We know that one gets a diffraction pattern
with a first minimum from the central peak at an angle of θ ∼ λ/D.
Below, you will see an example with λ  D so that the incoming plane
wave is completely diffracted to very large angles (so large that you don’t
even see the 1st diffraction minimum).

J. Gunion 9D, Spring Quarter 104

 In contrast, if D  λ, the wave behavior is indistinguishable from what
would be expected for a particle — as stated earlier both would just pass
straight through.
singleslit1

So, now you should ask yourself how this hugely wave-light diffraction

J. Gunion 9D, Spring Quarter 105

 phenomenon can be consistent with light traveling in little photon packets.
First, recall that there must be a lot of them to achieve a given intensity
of light in the large λ limit where each photon carries only a small amount
of energy.
What has to happen is that some of the photons go in one direction after
passing through the slit, and some in other directions.
The obvious question is how is this possible. You would be tempted to
say that each photon acts like a particle, goes through the slit and hits
the screen more or less opposite the slit.
This becomes even more confounding if you keep λ  D and decrease
the intensity of the light to the point where only one photon is traveling
through the slit and to the screen at a time. If you develop the film on
the screen after just a few photons have arrived, you would in fact find a
few dots on the screen, seemingly sort of randomly distributed.
If you continue operating at low intensity long enough, you would
eventually get the wide distribution (or if λ ∼ D, a diffraction pattern)
showing on the exposed film. This is not just some crazy thought
experiment, but is actually what happens.
Momentum Conservation?

J. Gunion 9D, Spring Quarter 106

 If you think a moment about the above photon-by-photon limit, you will
realize that each photon that is sent off in a direction perpendicular to the
original wave direction must have a non-zero momentum perpendicular
to that original direction.

Where is this momentum coming from? Well, momentum is still

conserved, it is just that the apparatus (the slit) was implicitly presumed
to be extremely massive and although it is recoiling in such a way
as to conserve overall momentum, that recoil is totally negligible and
unobservable.

What is important to note is that for small D, there is a big spread

in possible photon momenta in the direction ⊥ to the wave direction.
Conversely, if D  λ, then there is little spreading and redirection of
the incoming wave, corresponding to not very much ⊥ momentum.

This is an example of what we will call the uncertainty principle, according

to which: the better you try to define the location of a wave in a particular
direction, the more uncertain is the momentum (in the same direction)
of the particles that are actually contained in that wave

J. Gunion 9D, Spring Quarter 107

 The 2-slit case

We can push all this further by going to the case of two identical slits
separated by some distance S. (For this discussion, λ will be much
larger than the size of the slits, called D earlier, but not necessarily large
compared to the separation S.)
Using the above discussion, we would expect that if λ  S then the
result would be a very spread out interference pattern. In fact, in the
limit where the two slits lie on top of one another, we get (for slit width
D  λ) the same very spread out pattern shown earlier. Of course, for
λ ∼ S there should be more and more interference minima visible.
And, you should ask, what is the photon by photon picture in this case?
In fact, somehow the individual photons know that they should only go to
where the locations of constructive interference are and that they should
go with much less probability to the places where there is destructive
interference!

J. Gunion 9D, Spring Quarter 108

 Figure 25: Top: λ ∼ S. Bottom: λ < S.
J. Gunion 9D, Spring Quarter 109
 dots

J. Gunion 9D, Spring Quarter 110

 Probability and Wave-Particle Duality

The only way of mathematically realizing the above situation is in terms

of a probability amplitude.
The idea, to which we will return again and again, is that the particles
(photons) arrive at a given location on the screen (or other detector)
with a probability determined by the square of the wave amplitude (that
is the intensity of the wave) at that particular location!
To reemphasize, we write this in equation form as

Probability (density) of finding a particle ∝ |amplitude of wave|2 (121)

where we had to use the absolute square since a typical wave amplitude
that is the solution of the Maxwell wave propagation equations is a
complex exponential of some sort. (Be sure to review about complex
exponentials if you do not remember that mathematics. Also very useful
to remind yourself about Fourier transforms at this time.)
At this point, you might ask yourself the following question: can we
determine the slit through which a given photon passed?

J. Gunion 9D, Spring Quarter 111

 This is, in fact, not possible. In fact, if you temporarily closed one slit or
the other in between the infrequent photons of the low-luminosity limit,
the interference pattern would disappear! By changing the apparatus, we
alter the phenomena that depends upon both slits being simultaneously
present. We must “allow” each photon the opportunity to pass through
either slit, even though on a photon-by-photon basis it will always have
passed through one or the other. We just can’t say which.

An Example
Light of wavelength 633 nm is directed at a double slit, and the
interference pattern is viewed on a screen. The intensity at the center of
the screen is 4.0 W/m2.
(a) At what rate are photons detected at the center of the screen?
(b) At what rate are photons detected at the first interference minimum?
(c) At what rate are photons detected at a point on the screen where
the waves from the two sources are out of phase by 1/3 of a cycle?
(Note: you should recall from physical optics that the double-slit intensity
varies according to I = I0 cos2( 12 φ), where φ is the phase difference

J. Gunion 9D, Spring Quarter 112

 between the waves from the two slits and I0 is the intensity when φ = 0,
i.e. at the center of the screen.)
Answer:
(a) Each photon has energy

c 3 × 108 m/s
E=h = (6.63×10−34 J ·s) = 3.14×10−19 J . (122)
λ 633 × 10−9 m

We must therefore have

4.0 J/(s · m2)

= 1.27 × 1019 photons/(s · m2) . (123)
3.14 × 10−19 J/photon

(b) The first interference minimum is a point at which destructive

interference is complete by virtue of the phase difference between the
waves from the two slit sources being out of phase by π. This means 0
intensity and therefore no photons are arriving.
(c) For out of phase by φ = 13 (2π) (here, 2π corresponds to a complete

J. Gunion 9D, Spring Quarter 113

 cycle), one has
 
1 1
I = (4.0 W/m2) cos2 = 1.0 W/m2 .


3
2π (124)
2

Since this is one-fourth of the intensity at the center, the number of

photons per second per unit area is 1/4 of the value at the center of the
screen, or 3.18 × 1018 photons/(s · m2).
To repeat: Although it is photon hits that are being recorded on the
screen, the probability of their detection is governed by the behavior of
the associated electromagnetic wave.

J. Gunion 9D, Spring Quarter 114