This introduction to special relativity is for anyone who is already

comfortable with the Newtonian concepts of velocity, acceleration,

momentum, and energy. It is intended as a compromise between the
extremely rushed treatments of relativity that appear in standard introductory
physics textbooks, and the more leisurely treatments that appear in books
dedicated to teaching just relativity. I’ve tried to favor depth over breadth,
addressing the most important conceptual issues and introducing the
conceptual tools (including spacetime diagrams) that show how the theory
fits together.
In my own introductory physics course I deliver these five lessons during
five 50-minute class sessions, and we spend two more class sessions
discussing the assigned homework problems. This is nearly double the class
time for relativity that a traditional introductory course would allocate, but in
my mind it’s the absolute minimum.
In preparing these lessons I’ve drawn heavily on the treatments in the
three special relativity texts listed at the end under Further Reading. Many
thanks to authors William L. Burke, Peter Scott, Thomas A. Moore, Edwin F.
Taylor, and John A. Wheeler!

Contents
1. Setting the Stage
Reference Frames; The Principle of Relativity; Electromagnetism; Clock
Synchronization; Three Kinds of Time; Spacetime Diagrams

2. The Metric Equation

The Firecracker Experiment; The Muon Experiment; The Bouncing Light Pulse
Experiment

3. Applications of the Metric Equation

The Twin Paradox; Length Contraction

4. Two-Observer Spacetime Diagrams

Length Contraction Revisited; Combining Velocities; The Cosmic Speed Limit
 5. Momentum and Energy
Momentum Conservation; Relativistic Momentum; The Time Component of
the Four-Momentum; Relativistic Energy
6. Further Reading
7. About this Document

Lesson 1: Setting the Stage

“Relativity” is short for what Albert Einstein called his “special theory of
relativity”, published in 1905. This “theory” is a revised framework for the
laws of mechanics. It tells us that Newton’s laws of motion are only
approximate, and become especially inaccurate when we apply them to
objects that move extremely fast. Relativity replaces Newton’s laws, and the
related principles of momentum and energy conservation, with new versions
that are accurate at all speeds. (Ten years later Einstein published what he
called his general theory of relativity, which is actually a revised theory
of gravity. That theory is beyond the scope of these brief lessons.)
The mathematics of special relativity is no more difficult than that of
Newtonian mechanics: basic algebra and calculus, using derivatives to define
velocity and acceleration. The main difference you’ll notice is that a lot of the
formulas now involve square roots.
What makes relativity difficult to learn is not the mathematics, but
the concepts. You see, the problem with Newton’s laws isn’t merely that the
equations aren’t quite right; it’s that the underlying concepts turn out to be
inadequate. In particular, Newtonian mechanics rests on an oversimplified
conception of time. Unlearning the Newtonian concept of time and replacing
it with the far richer version in relativity theory can be quite a challenge. In
order to master relativity, you’ll need to take on this challenge and wrestle
with the concepts until they fall into place.
Another thing you should know about special relativity is that it doesn’t
have many direct practical applications. You’ll need it if you ever write
software for satellite-based navigation systems (GPS and similar), which
require super-accurate timing on satellites orbiting at thousands of miles per
hour. And it’s fundamental to understanding a lot of astrophysics and nuclear
physics and elementary particle physics. But in everyday life its role is
mostly hidden. So you might wonder whether you really need to know
relativity. Why should you put in the effort to learn an esoteric branch of
physics you may never use?
The short answer is that relativity is just too much fun to skip over. But a
more serious version of this answer is that special relativity is the best subject
I know for teaching us to think in new ways. It forces us to jettison our
preconceptions about time and velocity and mass. Doing that takes courage.
But in return for taking that brave step, relativity rewards us with a universe
that is strikingly beautiful and far more fantastic than anything a science
fiction writer could have invented. This process of opening ourselves to new
ideas is the essence of science, and the essence of learning. How could we
possibly pass up such an opportunity?

Reference Frames
So where do we start? With the concept of a reference frame.
I hope you recall something about reference frames from your study of
Newtonian physics. Basically a reference frame is a coordinate
system: xx, yy, and zz axes, laid out with respect with some origin, along
with a consistent method of measuring time. You can picture a reference
frame as a three-dimensional lattice of meter sticks, with a clock located at
each lattice point, as in this vivid illustration from the book Spacetime
Physics:
Using this structure we can measure both the location and the time of any
localized event, such as the collision of two billiard balls or the flashing of a
strobe light. In practice we never actually set up such a cumbersome
structure, but we need to remember that whatever method we do use to
measure locations and times of events must be equivalent to this one, in the
sense that it yields all the same numbers for position and time measurements.
(Also, in practice, we can often get by with a system for measuring positions
and times in just a two-dimensional plane, or even just along a one-
dimensional line, where our events of interest take place.)
When we set up a reference frame, we must make several arbitrary
choices: the location of the origin (labeled the “reference clock” in the
illustration above), the orientations of the three spatial axes, the origin of time
(when the clocks read zero), and, crucially, the state of motion of our
reference frame. Different choices will give us different measurements for
the xx, yy, zz, and tt coordinates of any given event, so we say that these
positions and times are relative.
If you’re not yet completely comfortable with the concept of a reference
frame and the idea that measured quantities can be relative, I highly
recommend this classic (albeit corny) film on the subject:
My reference frame has its origin at the post office, xx axis pointing
east, yy axis pointing north, and t=0t=0 at high noon. Your reference frame
has its origin at the library (5 blocks due east from the post office), xx axis
pointing north, yy axis pointing west, and t=0t=0 at 9:00 a.m. (when the
library opens). Suddenly, a door slams (Event S). Some time later, a dog
barks (Event B). Using my reference frame, I determine that the coordinates
of Event S are x=3x=3 blocks, y=2y=2 blocks, t=1t=1 hour, while the
coordinates of Event B
are x=−1x=−1 block, y=−3y=−3 blocks, t=3t=3 hours. (a) What
coordinates do you measure for these events? (b) Use
my xx and yy coordinates and the Pythagorean theorem to calculate the
distance between them (just in space, ignoring time). Then calculate the
distance again using your coordinates, and comment on the result.

The Principle of Relativity

Not everything is relative! Even though position and time measurements are
relative to our choice of reference frame, the laws of physics themselves are
absolute! More precisely, the laws of physics are the same from the
perspective of every inertial reference frame, that is, every frame in which
the law of inertia holds. (Inertial reference frames can move with respect to
each other slowly or rapidly, in any direction, so long as they do not
accelerate.) This fact of nature is called the principle of relativity:
The laws of physics are the same in every inertial reference frame.
– or –
All inertial reference frames are equally valid, so there’s no physical
experiment that can determine who is “really” moving.
This principle dates back 400 years, to the time of Galileo. It is built into
Newton’s laws of motion, which say that forces cause changes in an object’s
motion, but do not cause motion itself.
 Imagine a tennis ball as it is slapped back and forth across the court, with a
highly variable velocity u⃗ (t)u→(t). At every instant the ball’s
acceleration du⃗ /dtdu→/dt, with respect to a reference frame anchored to the
ground, is determined by the various forces F⃗ F→ exerted on it (by the
rackets, the ground surface, the air, and Planet Earth), according to Newton’s
second law: du⃗ /dt=(∑F⃗ )/mdu→/dt=(∑F→)/m, where mm is the ball’s
mass. Now imagine watching the game from a bicycle that’s coasting
alongside the court at constant velocity v⃗ v→. According to the Galilean
velocity transformation law, you measure the ball’s velocity at any instant to
be u⃗ ′=u⃗ −v⃗ u→′=u→−v→. (For example, if your speed is 3 m/s and at a
certain instant the ball is moving parallel to you at 10 m/s with respect to the
ground, then you measure its speed to be 7 m/s. We’ll see in Lesson 4 that
this transformation law is not quite correct, but please assume for now that it
is.) Assuming that the forces and mass are the same in all frames of reference,
prove that Newton’s second law holds true in your frame of reference as well.
Also explain why your proof would not apply if your bicycle were
accelerating.
A 1-kg block glides in the xx direction at 10 m/s on frictionless ice until it
collides with a second 1-kg block, initially at rest. The two blocks stick
together (there’s Velcro on them) and continue gliding in the xx direction. (a)
What is the initial momentum of the two-block system? (b) Using the
ordinary Newtonian law of momentum conservation, predict the final velocity
of the two combined blocks. (c) Now imagine viewing this collision from a
reference frame that is moving (with respect to the ground) in the xx direction
at 5 m/s. Use the ordinary Galilean velocity transformation law (see the
previous exercise) to determine the initial velocities of both blocks in this
frame, as well as their joint final velocity. (d) Use the results of part (c) to
calculate the system’s initial and final momentum in the new reference frame.
Is momentum the same in all frames of reference? Is the law of momentum
conservation consistent with the principle of relativity? Explain briefly.

Electromagnetism
But by the late 1800s, physicists had convinced themselves that the principle
of relativity does not apply to the laws of electromagnetism. That’s because
Maxwell’s equations for the electric and magnetic fields predict that
electromagnetic waves (including visible light) travel at a speed of
1ϵ0μ0−−−−√=c=3.00×108 m/s.(1)(1)1ϵ0μ0=c=3.00×108 m/s.
Question: 300 million meters per second with respect to what? Between 1865
and 1905, everyone assumed the answer was: with respect to some preferred
frame of reference, in which the (hypothetical) medium that transmits
electromagnetic waves (then called the ether) is at rest.
But Einstein wasn’t so sure. He was familiar with a common
demonstration that you may have seen, in which we move a coil of wire
relative to a magnet, inducing a current measured by a galvanometer:

When we do this we get the same deflection of the needle regardless of

whether we move the coil with the magnet fixed or move the magnet with the
coil fixed. And yet we explain those two scenarios in completely different
ways! When the coil moves, we say the moving electrons feel
a magnetic force (qv⃗ ×B⃗ qv→×B→) that makes them circulate around the
coil. On the other hand, when the magnet moves (and the coil is fixed), we
know the stationary electrons can’t feel any magnetic force
(since v⃗ =0v→=0), but they do feel an electric force from the circulating
electric field E⃗ E→ that’s temporarily created by the changing magnetic field
according to Faraday’s law.
Einstein said this can’t be a coincidence. Somehow, despite these
differing explanations, the laws of electromagnetism must fundamentally
respect the principle of relativity, so they predict the same phenomena
regardless of whether we view things from the reference frame of the magnet
or the reference frame of the coil. (Einstein’s 1905 paper begins with exactly
this example; you can read an English translation of it here or here.)
 But then what about the speed of light? If the laws of electromagnetism
are equally valid in all inertial frames of reference, and if those laws predict
that light travels at 300 million meters per second, then light must travel at
that speed with respect to every inertial frame of reference—despite the fact
that one reference frame could be moving with respect to another at (say) half
the speed of light or faster! That seems impossible, but in fact it’s not. It’s
just contrary to the intuition that we’ve developed by watching slow-moving
objects.
The price we pay for accepting this seemingly impossible fact about light
is that we must discard some of our assumptions about time. We use time to
define speed, so perhaps if time is sufficiently screwy, speeds might also defy
our intuition.
(Electromagnetism review.) In the illustration above of the magnet and coil,
the rightward swing of the galvanometer needle away from vertical indicates
that a positive current is flowing into the terminal marked +. Look at the
direction in which the coil is wound, then answer the following: (a) First
assume that the magnet is fixed and the coil is moving horizontally, along its
axis. Use F⃗ =qv⃗ ×B⃗ F→=qv→×B→ and the right-hand rule for cross-
products to determine whether the coil is moving toward or away from the
magnet. Which component of the magnetic field—rightward along the axis or
outward away from the axis—causes the current to flow? (b) Now assume
instead that the coil is fixed and the magnet is moving horizontally. Use
Faraday’s law (or Lenz’s law) to explain which direction the magnet must
move to give a positive current as shown.

Clock Synchronization
Recall the picture of a reference frame above, with a clock at every grid
location. To measure the time when an event occurs, we look at the reading
of whatever clock lies nearest to the event. But to compare the times of
events that happen in different places, we need to make sure our clocks are all
synchronized with each other. How do we do that?
You might think we could pick up our reference clock and carry it
sequentially to the locations of all the other clocks, setting each of them to
match it. As we’ll later see, that method won’t work: accelerating the
reference clock into motion, then stopping it when it arrives near some other
clock, will affect its measurements. (You don’t need to actually believe this
yet—just accept that the accelerations could affect the measurements, and
agree not to use this method of clock synchronization.) Fortunately, there’s a
much better way to synchronize the clocks: Just look at them! But when you
do, be sure to take into account that the light from different clocks might have
taken different amounts of time to reach you. If you’re the same distance
from two different clocks (as measured by all those meter sticks!), then you
should always see them reading the same time. If you’re looking at a clock
that’s one light-second away (300 million meters, or about 3/4 the distance to
the moon), then you should see it running exactly one second behind the
clock at your location, because the light from the distant clock took that long
to reach you. Because the speed of light is the same with respect to all inertial
reference frames (and doesn’t depend on the direction the light travels), we
can be confident that this light-travel-time adjustment will always work as
expected.
Suppose you’re standing still (with respect to the ground) and looking at a
clock tower 200 meters away. If your wristwatch reads exactly 8:18 a.m. and
the clock on the tower is perfectly synchronized with it, what time should
you see on the tower’s clock face?

Three Kinds of Time

Now that we know how to synchronize our clocks, I’m ready to carefully
define time. Specifically, I want to define the time between two given events,
each of which is localized in both time and space. For instance, Event A
might be the lighting of the fuse on a firecracker, while Event B might be the
explosion of the firecracker, some time later. How should we define the time
interval between Event A and Event B? There are actually three conceptually
distinct ways we can do it:
 Coordinate time: Use an inertial reference frame, with properly
synchronized clocks, to measure the time of each event according to
whichever clock is located at each event. If the two events happen in
different locations, this measurement requires two different clocks.
Subtract these two clock readings to obtain what we call coordinate
time between events A and B, denoted ΔtABΔtAB. (Note: As we’ll later
see, this value will depend on which inertial reference frame we use.
 So there are actually many different coordinate times between the
same two events!)
 Proper time: Forget about reference frames! Instead, find
a single clock that’s present at both Event A and Event B. Whatever
time interval that clock measures between the two events is
called proper time between the two events,
denoted ΔτABΔτAB (that’s the Greek letter tau). (Note: As we’ll later
see, this value will depend on how the clock moves in between the
two events. So there are actually many different proper times
between the same two events!)
 Spacetime interval: This is the same as proper time, but we add the
requirement that the single clock we’re using to measure the interval,
which must be present at both events, does not accelerate. The clock
must be present at Event A and must move with whatever constant
velocity is needed to arrive at Event B just as that event happens. The
spacetime interval between Events A and B is unique, and is
denoted ΔsABΔsAB. (Since the clock does not accelerate, we can
attach an inertial reference frame to it. Therefore the spacetime
interval is also the same as the coordinate time as measured in that
special inertial frame of reference in which both events happen at the
same place—allowing us to measure the coordinate time
interval ΔtΔt with a single clock.)

You run a single lap around a track, while your coach, standing at your
starting (and ending) location, times your lap with a stopwatch. Event A is the
start of your lap while Event B is the finish. What kind(s) of time between
these events is/are measured by the stopwatch? What kind(s) of time between
these events is/are measured by your wristwatch? Explain carefully.

You are standing near a railroad track when a train rushes past you, moving at
constant velocity. Let Event A be the locomotive (at the front of the train)
passing you, and let Event B be the caboose (at the rear of the train) passing
you. You measure the time between these events with your wristwatch, while
the train’s engineer (in the locomotive) and conductor (in the caboose)
measure the time between these events using their pocket watches, which they
have carefully synchronized. What kind of time do you measure between the
two events? What kind of time does the train’s crew measure?
Spacetime Diagrams
To visualize the times and locations of various events, I now want to
introduce a tool called a spacetime diagram. It’s really just a plot of one-
dimensional position (x) and coordinate time (t), like we use for one-
dimensional motion when studying basic kinematics. But there are two new
twists: First, it’s conventional to plot x horizontally and t vertically. This
reversal may seem strange at first, but I think you’ll soon get used to it.
Second, we space the tick marks equally along both axes, such that our unit
of distance is whatever distance light travels in one unit of time. For instance,
you’ve probably heard of measuring distances between stars in light-years,
where a light-year is the distance that light travels in one year
(nearly 10161016 meters); this will be our unit of distance if we measure
time in years. If instead we measure time in seconds, then our distance unit is
one light-second, or 300 million meters (about 3/4 of the distance to the
moon). Or, for events occurring in a smaller laboratory, we can measure time
in nanoseconds and distance in light-nanoseconds (one light-nanosecond is
0.3 meters, or about a foot). The slow-moving objects we’re used to don’t
travel very far in a nanosecond, and would take a long time to travel a light-
second, let alone a light-year. But for the fast-moving objects that make
relativistic effects apparent, these unit choices will be very convenient.
Here is a spacetime diagram, calibrated in seconds and light-seconds, on
which I’ve plotted several events:

Notice that an event, localized in both space and time, is represented on the
diagram by a point. We plot each point on the diagram according to its
coordinates (x,t)(x,t) in some particular inertial reference frame; if we used a
different inertial frame then the appearance of the diagram would change (as
we’ll see in detail in Lesson 4). But at least from the perspective
of this inertial reference frame, the diagram shows us at a glance that Event A
(starship’s warning sirens sound) occurs at t=1t=1 second
and x=2x=2 light-seconds; that Event B (deflector shields are raised) occurs
at the same place, two seconds later; that Event C (enemy ship fires photon
torpedoes) occurs at the same time as B, four light-seconds away to our right;
and that Event D (science officer raises eyebrow) occurs three seconds later
still and at x=3x=3 light-seconds.
Now think about a sequence of events that all happen to a particular
object: perhaps the flashes of a strobe light, or the beats of a person’s heart. If
we plot these events on a spacetime diagram and connect the dots together,
we have a record of that object’s (or person’s) motion:

We refer to the line or curve connecting all events that happen to a particular
object as that object’s worldline—its line through the “world” of space and
time. Often we draw an upward-pointing arrow on a worldline, to remind us
that the object’s history flows from bottom to top.
Here are some more worldlines:
Object 1 is at rest, always at the same xx value. Object 2 is moving to the
right at a constant velocity of 1/3 the speed of light (one light-second of
distance in each three seconds of time), while object 3 is moving to the left
(in the −x−x direction) at 2/3 the speed of light. Notice that the faster an
object’s motion, the shallower the slope of its worldline. Object 4 is initially
moving to the right but then slows down, stops, and gradually begins moving
to the left. Object 5 is a light pulse, moving rightward at the speed of light:
one light-second per second. A light pulse worldline always lies at a 45-
degree angle on a conventionally calibrated spacetime diagram.
Again, each of these spacetime diagrams is plotted from the viewpoint of
one particular inertial reference frame; let’s call it the Home Frame. If instead
you measure events with respect to some Other Frame that’s moving to the
right at 1/3 the speed of light (with respect to the Home Frame), then your
spacetime diagram will show Object 2 at rest, with a vertical worldline, and
Object 1 moving to the left at 1/3 the speed of light. Motion is relative!
Lesson 4 explains in detail how to translate a spacetime diagram from one
inertial reference frame to another.
For a delightful animated explanation of reference frames and spacetime
diagrams, I recommend the Minute Physics video Spacetime Diagrams. The
video even shows how to add a yy axis to a spacetime diagram, to depict
motion in two spatial dimensions. But it doesn’t use the convention of
calibrating the space and time axes so that light signal worldlines are always
at 45 degrees.
At t=0t=0 an uncrewed rocket is launched from earth, traveling in
the +x+x direction at 4/5 the speed of light (with respect to earth). After 10
seconds, as measured in earth’s frame of reference, the rocket explodes. A
 burst of light from the explosion travels back toward earth, where authorities
detect the light some time later. Draw a calibrated spacetime diagram that
accurately shows these objects and events, as observed in earth’s reference
frame. Label the launch event, the explosion event, and the detection-of-light
event, as well as the worldlines of earth, the rocket, and the light burst.

Look again at the spacetime diagram above with events labeled A through D.
This diagram is drawn from the perspective of some Home reference frame.
What velocity would some Other reference frame need to have, with respect
to the Home frame, in order for observers in the Other frame to observe
Event B and Event D to occur at the same place? Explain carefully.

Lesson 2: The Metric Equation

In the previous lesson I defined three different ways of measuring the time
between any two events: coordinate time, proper time, and the spacetime
interval. My goal for this lesson is to show you how the coordinate time
between two events, in any inertial reference frame, is related to the
spacetime interval between those events. (I’ll discuss the proper time
measured by an accelerated clock in the following lesson.) To show this
relationship from different perspectives I’ll describe three experiments:
1. The firecracker experiment;
2. The muon experiment;
3. The bouncing light pulse experiment.

Experiments 1 and 3 are mere thought experiments, which would be

impractical to carry out in the way I’ll describe. Experiment 2 is a real
experiment that has actually been performed.

1. The Firecracker Experiment

Imagine that I have a fistful of firecrackers, all with identical 10-second
fuses. (I know the fuses are identical, and absolutely reliable, because they
come from the world’s most reputable firecracker factory, and because I’ve
already tested many randomly chosen firecrackers from the same batch and
found all their fuses to burn for exactly 10 seconds.) I’m standing with my
firecrackers at the origin (x=0x=0) of an inertial reference frame, with a
long measuring tape stretched out in both directions along my xx axis, and an
array of carefully synchronized clocks located along the measuring tape at
short intervals:

At t=0t=0 I use a match to light the fuses on all the firecrackers, and
simultaneously hurl them in both directions, at an assortment of speeds. Some
go fast, while others go slow. Some go in the +x+x direction, while others
go in the −x−x direction. I give one of the firecrackers a velocity of zero,
holding it in my hand for reference.
Eventually all the firecrackers explode, and I carefully record the places
and times of these explosions. How do I do that? I could station an assistant
at each clock, with instructions to record the location and clock reading when
an arriving firecracker explodes. Or I could just watch for the explosions,
using binoculars to view the tape label and clock reading at each explosion
event. Of course the light from these explosions will take time to reach me,
and that delay will be longer for the more distant explosion events, so I don’t
expect to actually see all the explosions at the same time. But if I didn’t know
anything about special relativity, I would still expect all the explosion events
to occur at the same time. Plotted on a spacetime diagram, the explosion
events should (I expect) all lie on a horizontal line:
On this diagram my own worldline coincides with the tt axis, because I’m
at x=0x=0 and not moving (with respect to my own reference frame). The
vertical red line is the worldline of the firecracker that I’m holding in my
hand, and its explosion event is plotted at x=0x=0 and t=10t=10 seconds. I
expect all the other explosion events to also occur at t=10t=10 seconds, as
shown.
But that’s not what actually happens.
The firecracker that I’m holding in my hand really does explode
at t=10t=10 seconds, as expected. But the other firecrackers
explode later than t=10t=10 seconds, by an amount that’s tiny if they’re
moving slowly but that grows quite large if they’re moving at nearly the
speed of light (with respect to my reference frame). Plotted on a spacetime
diagram, the explosion events actually lie on a hyperbola that’s defined by
the formula t=(10 s)2+x2−−−−−−−−−−√t=(10 s)2+x2 (with the
understanding that xx is measured in light-seconds):
As you can see, the hyperbola is quite flat near the middle of the diagram, so
for slow-moving firecrackers we could easily mistake it for a horizontal line.
Meanwhile, the explosion events for fast-moving firecrackers can occur at
arbitrarily large distances (in either direction), and at arbitrarily late times,
because the hyperbola extends infinitely far in both directions, asymptotically
approaching the 45-degree worldlines of the light flashes traveling outward
from the match that I used to light the fuses.
Does this shocking result mean that something was wrong with those
firecracker fuses after all? No! Each fuse truly measures exactly 10 seconds
between the fuse-lighting event (call it Event A) and the explosion event (call
it Event B). But there’s no logical necessity to our expectation that the time
between these two events as measured by the fuse must be the same as the
time between them as recorded in my inertial reference frame (and plotted on
my tt axis). For the fuse, which is present at both events and doesn’t
accelerate along the way, measures the spacetime
interval ΔsABΔsAB between the fuse-lighting event A and the explosion
event B, whereas the clocks in my inertial reference frame, no one of which
is present at both events, instead measure coordinate time ΔtABΔtAB.
 The mathematical relationship between coordinate time and the spacetime
interval is summarized in the equation for the hyperbola given above. For an
arbitrary pair of events A and B, the equation reads
ΔtAB=(ΔsAB)2+(ΔxAB)2−−−−−−−−−−−−−−−−√,(2)
(2)ΔtAB=(ΔsAB)2+(ΔxAB)2,
where ΔsAB=ΔsAB= 10 seconds in the firecracker example, and again with
the understanding that ΔxABΔxAB is measured in units of the distance that
light travels in one time unit (e.g., light-seconds if the times are in seconds).
This relationship is called the metric equation of special relativity, and is
summarized in this simple spacetime diagram:

We can write the metric equation in many other ways. For instance, if we
square both sides and move the ΔxΔx term to the left, we obtain
(ΔtAB)2−(ΔxAB)2=(ΔsAB)2.(3)(3)(ΔtAB)2−(ΔxAB)2=(ΔsAB)2.
I like this version because it puts the frame-dependent coordinate differences
(which of course must be measured in the same reference frame) on one side
and the unique spacetime interval on the other. There is a deep analogy
between the metric equation and the Pythagorean formula for calculating
distances in a two-dimensional plane—but the metric equation has a minus
sign where the Pythagorean formula has a plus.
If you’d rather measure ΔxΔx in more conventional units, then
the ΔxΔx term in the metric equation requires a conversion factor. For
instance, if ΔxΔx is in meters and the times are in seconds, then to
convert ΔxΔx to light-seconds we divide by the number of meters in a light-
second, 3×1083×108. More generally, to convert ΔxΔx to appropriate
light-travel units we must divide by the speed of light:
(ΔtAB)2−(ΔxABc)2=(ΔsAB)2.(4)(4)(ΔtAB)2−(ΔxABc)2=(ΔsAB)2.
Yet another variation is to notice that Δx/ΔtΔx/Δt is the velocity of the
clock (e.g., a firecracker’s fuse) that measures ΔsΔs, with respect to our
inertial reference frame. Denoting this velocity vv, we can then
insert Δx=vΔtΔx=vΔt to write the metric equation in terms of vv instead
of ΔxΔx:
(ΔtAB)2(1−v2c2)=(ΔsAB)2,orΔtAB=ΔsAB1−(v/c)2−−−−−−−−√.
(5)(5)(ΔtAB)2(1−v2c2)=(ΔsAB)2,orΔtAB=ΔsAB1−(v/c)2.
This is the form of the metric equation that’s most often written in
introductory textbooks, although these books usually call it the “time dilation
equation”, and instead of ΔsΔs they often use the notation ΔτΔτ or Δt′Δt′.
I’ve been assuming that we orient our reference frame’s xx axis so
Events A and B are separated only in the xx direction, not yy or zz. To drop
this assumption, just replace (Δx)2(Δx)2 in the metric equation with the
square of the spatial distance between the events, (Δx)2+(Δy)2+
(Δz)2(Δx)2+(Δy)2+(Δz)2. When we write the metric equation in terms
of vv instead of ΔxΔx, that means v2=v2x+v2y+v2zv2=vx2+vy2+vz2.
When applying the metric equation, the most common difficulty is
figuring out whose clocks measure ΔtΔt and whose clock measures ΔsΔs. I
find it helpful to keep referring back to the last figure above, which shows
that ΔtΔt is always longer than ΔsΔs, and that ΔsΔs is the time interval as
measured by the unique, nonaccelerated clock that is present at both events.
Suppose that one particular firecracker, among those described above, has a
velocity (with respect to your reference frame) of 2/3 the speed of light. Draw
a spacetime diagram showing its worldline, its explosion event, and the light
from the explosion traveling back to the origin. What are
the xx and tt coordinates of the explosion event? At what time do you
(standing at the origin) see the light from this explosion? Label your
spacetime diagram to show your answers.
Suppose that one particular firecracker, among those described above,
explodes at x=−12x=−12 light-seconds. Draw a spacetime diagram
showing its worldline, its explosion event, and the light from the explosion
traveling back to the origin. What is this firecracker’&s velocity? At what
 time do you (standing at the origin) see the light from this explosion? Label
your spacetime diagram to show your answers.
Suppose that one particular firecracker, among those described above, has a
velocity (with respect to your reference frame) of 200 kilometers per second
(slightly faster than NASA’s fastest-ever space probe). What are
the xx and tt coordinates of its explosion event?
For the events A, B, C, and D shown in the first spacetime diagram in the
previous lesson, use the metric equation to calculate the spacetime
intervals ΔsABΔsAB, ΔsBDΔsBD, ΔsADΔsAD, and ΔsCDΔsCD. How
does ΔsADΔsAD compare to the sum ΔsAB+ΔsBDΔsAB+ΔsBD? What
happens if you try to calculate ΔsACΔsAC or ΔsBCΔsBC? Can you
generalize your answers to these questions?

You wish to travel to the Vega star system, 25 light-years from earth. Being
impatient, you would rather not spend more than 15 years of your own time
on the journey. How fast must your spaceship travel? How long does your trip
take, according to observers on earth (or on Vega, which is more or less at
rest with respect to earth)? Draw an accurate spacetime diagram showing the
worldlines of earth, Vega, and your spaceship.

Repeat the previous exercise for a trip to Polaris (the North Star), which is
430 light-years distant. Assume again that the trip should take no more than
15 years of your own time. Sketch a spacetime diagram to convey the idea of
your calculation, but don’t worry about making it quantitatively accurate
(which would be difficult). Is there any limit to how far out into the universe
you can travel within a human lifetime?

People often describe the metric equation with the ambiguous phrase moving
clocks run slow. Explain why this phrase can be misleading, and give an
example in which it is the “stationary” clock that “runs slow”.

2. The Muon Experiment

Of course nobody has ever hurled a firecracker at nearly the speed of light
over a distance several times farther than the moon. So how do we know that
time really obeys the metric equation?
One of the most direct actual experiments to test the metric equation uses
elementary particles called muons, which have their own built-in “fuses”. We
know from studying muons at rest that they spontaneously decay (into an
electron and a pair of neutrinos) with a half-life of about 1.5 microseconds
(μsμs). This means that the time when any particular muon will decay is
random, such that it has a 50% chance of decaying during any 1.5 μsμs time
interval.
Conveniently, muons are constantly being created in earth’s upper
atmosphere by the collisions of incoming cosmic rays (mostly protons) with
air molecules. The muons (unlike the cosmic ray protons) penetrate our
atmosphere quite readily, and are constantly raining down on earth’s surface
at a rate of about one per square centimeter per minute. We can then detect
them with Geiger counters or other detectors used in nuclear physics
laboratories.
In one version of the muon experiment, scientists from MIT operated
their muon detector in two different locations: on the MIT campus (at
approximately sea level), and on the summit of New Hampshire’s Mt.
Washington, 6000 feet (or 6 light-microseconds) above sea level. They
designed their apparatus to detect muons only within a narrow range of
speeds, then counted how many muons with speeds in that range arrived per
unit time at each elevation, in order to test whether the muons’ internal
“clocks” obey the metric equation.

Consider, for instance, a muon that happens to be moving directly

downward toward the MIT campus. Let Event A be its crossing the 6000-foot
altitude level, and let Event B be its arrival in the detector. If the muon is
moving at nearly the speed of light, then the coordinate time interval between
these events, as measured in earth’s frame of reference,
is ΔtAB≈6 μsΔtAB≈6 μs. That’s four times the muons’ half-life, so if we
didn’t know anything about relativity we would predict that once the muon
makes it to the 6000-foot level, it has only a 1-in-16 chance
(12⋅12⋅12⋅1212⋅12⋅12⋅12) of making it down to sea level without decaying
first. On average, therefore, we would expect to detect only 1/16 as many
fast-moving muons at MIT as on the top of Mt. Washington.
But according to the metric equation, the muon’s internal clock
does not measure 6 seconds of time between Event A and Event B. Because
the muon is present at both events (and doesn’t accelerate significantly in
between), it measures the spacetime interval,
ΔsAB=ΔtAB1−(v/c)2−−−−−−−−√=(6 μs)1−(v/c)2−−−−−−
−−√,(6)(6)ΔsAB=ΔtAB1−(v/c)2=(6 μs)1−(v/c)2,
where vv is the muon’s speed with respect to the earth. If, for
instance, v/c=0.995v/c=0.995, then 1−(v/c)2−−−−−−−−√=0.11−
(v/c)2=0.1, and the time measured by the muon’s clock is only 0.6
microseconds, less than a single half-life. Relativity therefore predicts that
the number of muons with this speed detected at MIT should be not 1/16 as
many as on Mt. Washington, but well over half as many.
And that’s what the experiment found: The number of muons observed at
MIT, compared to Mt. Washington, was much larger than the naive
prediction without time dilation, and fully consistent with the prediction of
the metric equation.
The MIT experiment is documented in this 1963 film, which is worth
watching just to see the old apparatus and counting techniques. It’s also
written up in the American Journal of Physics 31, 342-355 (unfortunately
paywalled). If you watch the film or read the paper you’ll see that the
experiment involved a number of complications that I’ve glossed over. Most
importantly, the muons decelerate slightly as they descend through the
atmosphere, so the researchers had to account for some deceleration in their
quantitative check of the metric equation. But even without sophisticated
calculations, the data plainly show that the number of muons arriving at sea
level is far too high to account for without relativistic time dilation.
I also recommend this Minute Physics description of the muon
experiment.
 Another type of unstable subatomic particle is the charged pion, whose half-
life is just 18 nanoseconds. Imagine a beam of charged pions traveling down
the length of a long vacuum pipe at an accelerator laboratory at
speed 0.98c0.98c. (a) If time were absolute, so the spacetime interval were
the same as coordinate time, how far would these particles travel before half
of them decay? (b) How far do they actually travel before half of them decay,
taking the metric equation into account?

3. The Bouncing Light Pulse Experiment

Muons weren’t discovered until 1936, so you may still be wondering how
Einstein figured out the metric equation in 1905. One line of reasoning that
he described involves another thought experiment.
Imagine two horizontal mirrors, facing each other, separated by some
fixed vertical distance dd. Between the mirrors we set off a strobe that emits
a single brief pulse of light moving vertically. This light pulse then repeatedly
bounces up and down between the mirrors. To measure the time it takes to
bounce up and down, we affix a clock to the bottom mirror.

Let Event A be a particular bounce of the light pulse off the bottom
mirror, and let Event B be the next bounce off the bottom mirror, after a
single round trip. Because the light travels a total distance 2d2d in between
these events, and it moves at speed cc, we can immediately write
2d=cΔsAB,(7)(7)2d=cΔsAB,
where ΔsABΔsAB is the time between the two events as measured by our
clock. Our clock measures the spacetime interval because it is present at both
events (and we won’t allow the apparatus to accelerate).
Now let’s view these same events from an inertial reference frame in
which the whole apparatus is moving to the right at some constant speed. The
illustration below shows three successive images of the mirrors from this
perspective (at the times of the three light pulse bounces), along with the path
of the light pulse. Notice that in this frame of reference the light pulse is
traveling diagonally, so it travels farther. But the principle of relativity
(together with the laws of electromagnetism) requires that the measured
speed of light still has the same value, cc, in this new frame of reference.
Because the light travels a greater distance at the same speed, it must take
more time.

To measure the time between Event A and Event B in our new frame of
reference, we require a pair of previously synchronized clocks, at rest in this
frame, one present at each event. (These two clocks are not shown in the
illustration, which shows only the clock attached to the bottom mirror.) The
distance between these two clocks is ΔxABΔxAB, so each of the two
diagonal legs of the light-pulse path is the hypotenuse of a right triangle with
height dd and base ΔxAB/2ΔxAB/2. We can therefore write the total
distance traveled by the light pulse, using the Pythagorean theorem, as
 Total distance=2d2+(ΔxAB/2)2−−−−−−−−−−−−
−√=(2d)2+(ΔxAB)2−−−−−−−−−−−−−√.(8)(8)Total
distance=2d2+(ΔxAB/2)2=(2d)2+(ΔxAB)2.
But since the pulse travels at speed cc, this distance must
equal cΔtABcΔtAB. Meanwhile we have already seen
that 2d=cΔsAB2d=cΔsAB, so this equation becomes
cΔtAB=(cΔsAB)2+(ΔxAB)2−−−−−−−−−−−−−−−−−√,(9)
(9)cΔtAB=(cΔsAB)2+(ΔxAB)2,
and this is just an algebraic rearrangement of the metric equation as written
above (with the factor of cc explicit so we can express ΔxABΔxAB in
traditional distance units if we like).

Lesson 3: Applications of the Metric

Equation
In this lesson I’ll work out two further implications of the metric equation.
The first involves proper time measured by a clock that accelerates. The
second involves the relativity of distance measurements.

The Twin Paradox

Alice and Betty are identical twins, but their abilities and ambitions are not
quite identical. Alice grows up to become an astronomer, while Betty trains
to become an astronaut. Their interests align, however, when Alice discovers
evidence that one of the planets near Alpha Centauri, the nearest star system
to our own, has conditions suitable for life. Betty is chosen to fly on a
spaceship and visit the Alpha Centauri system, to investigate further.
The spaceship is launched on the twins’ 30th birthday, and soon Betty is
flying toward Alpha Centauri (which is 4 light-years away from earth) at a
speed of 4/5 the speed of light. When she finally arrives at the planet of
interest, however, she discovers that although it has conditions suitable for
life, no life actually exists on its barren surface. Having nothing further to do
there, Betty immediately gets back on her spaceship and returns to earth,
again at 4/5 the speed of light.
Although the mission’s outcome is a disappointment, both twins are eager
to be reunited. But when does their reunion occur? From Alice’s viewpoint,
Betty’s spaceship must travel 4 light-years outward at 4/5 the speed of light,
so it takes 5 years to reach Alpha Centauri, and another 5 years to return. Ten
years pass in total, so Alice will be 40 years old upon Betty’s return. We can
easily visualize these events on a spacetime diagram:

But how much time passes from Betty’s perspective?

To answer this question we need to break up Betty’s trip into two
segments, connecting three events. As shown on the diagram, Event D is
Betty’s departure from earth; Event E is her exploration and turning around at
Alpha Centauri; and Event F is her final return to earth. During her outbound
journey, Betty’s clocks (including her biological “clock”) measure the
spacetime interval ΔsDEΔsDE, because she is present at both Event D and
Event E, and she travels at constant velocity. (I’m assuming, quite
unrealistically, that her acceleration and deceleration happen too quickly to
show on the diagram or to affect the calculations.) And according to the
metric equation, this time interval is
ΔsDE=(ΔtDE)2−(ΔxDE)2−−−−−−−−−−−−−−−−√=(5 year
s)2−(4 years)2−−−−−−−−−−−−−−−−−−√=3 years.(10)
(10)ΔsDE=(ΔtDE)2−(ΔxDE)2=(5 years)2−(4 years)2=3 years.
Similarly, during the return journey, Betty’s clocks measure the spacetime
interval ΔsEFΔsEF, which also equals 3 years. The total time elapsed during
the journey, according to Betty’s clocks, is therefore only 6 years! When they
are reunited and Alice is turning 40, Betty is only turning 36.
This result is astounding, but there’s just no getting around it if you
accept the metric equation. What some people have trouble understanding,
though, is how to reconcile the asymmetry in the twins’ ages with the
principle of relativity. If motion is relative, shouldn’t it be equally valid to
analyze these same events from the viewpoint of Betty’s reference frame?
And in Betty’s frame, wouldn’t it be Alice who races away (along with the
earth) at 4/5 the speed of light, then turns around and returns at the same
speed, and who therefore ends up being the younger twin when they are
reunited at Event F?
No. The two sisters’ reference frames are not equally valid because
Alice’s frame is inertial (to a good approximation) and Betty’s is not. It is
Betty, not Alice, who experiences enormous accelerations (“g-forces”) as her
spaceship speeds up during launch, turns around at Alpha Centauri, and
finally comes to a screeching halt when she returns to earth. The principle of
relativity says not that all reference frames are equally valid, but that
all inertial reference frames are equally valid. The concepts and tools we’ve
developed in these lessons tell us nothing about how to analyze events from
the perspective of a non-inertial reference frame.
But we don’t need any new tools to analyze the motion of
accelerated objects, such as Betty and her spaceship, as viewed from any
inertial frame of reference. More specifically, this example shows us how to
calculate the proper time between two events as measured by an accelerated
clock: just break up the clock’s worldline into segments along which the
velocity is (approximately) constant, apply the metric equation to each of
these smaller segments, and add up the ΔsΔs values for each segment to get
the proper time, ΔτΔτ, measured by the accelerated clock. In Betty’s case it
suffices to break the worldline DEF into just two segments, DE and EF, but
in other cases we might need more than two, and for a smoothly accelerated
clock we would need to divide the curved worldline into a large number of
small, nearly straight segments.
Moreover, it isn’t hard to prove that for any given pair of events, the
proper time interval ΔτΔτ measured by any accelerated clock (present at
both events) will always be less than the spacetime interval ΔsΔs, that is, the
time measured by a non-accelerated clock (present at both events). Betty’s
younger age compared to Alice is just a special case of this general fact about
spacetime.
Of course nobody (at least here on earth) has access to spaceships that
actually travel at 4/5 the speed of light. You may therefore be wondering
what real-world experiments have been done to verify that an accelerated
clock measures less time, between the same two events, than a non-
accelerated clock. These experiments fall into two categories. First, you can
again use sub-atomic particles such as muons, traveling at high speeds. In one
experiment in the 1970s, scientists at the CERN laboratory in Geneva
measured the decay of muons while they were accelerating around a storage
ring at v/c≈0.9994v/c≈0.9994, and found that on average these muons
lasted 29 times longer (or “aged” 29 times slower) than non-accelerated
muons, just as relativity predicts. Second, you can avoid the need for
absurdly high speeds if you use sufficiently accurate clocks. In a famous
experiment performed in 1971, scientists flew state-of-the-art cesium beam
atomic clocks around the world on commercial aircraft, comparing the clock
readings to those of an identical clock that remained at the U.S. Naval
Observatory. Again, the results were fully consistent with the predictions of
relativity.
Cedric and Denzel are twins, and both wish to travel from earth to the Sirius
star system, 9 light-years away. Cedric departs on their 25th birthday, taking a
spaceship that travels at 3/4 the speed of light. Denzel procrastinates and
doesn’t depart until two years later, but is then able to take a newly developed
spaceship that travels at 9/10 the speed of light. (a) Draw an accurate
spacetime diagram showing the worldlines of earth, Sirius, and both of the
twins. Label each of the departure and arrival events. (b) How old is each of
 the twins when they simultaneously arrive at Sirius? Explain how your
answers illustrate that the proper time between two given events along an
accelerated worldline is always less than the spacetime interval.
A muon travels around a circular storage ring at constant speed vv. Let
Event A be its passing by some fixed point in the ring, and let Event B be its
next passing by that same point, after one trip around the ring. Prove that the
time between these events as measured by the muon’s clock is ΔτAB=1−
(v/c)2−−−−−−−−√ΔsABΔτAB=1−(v/c)2ΔsAB, where as
usual ΔsABΔsAB is the time between these two events as measured by a non-
accelerated clock. (Hint: Imagine dividing the muon’s circular trip into many
small segments that are each essentially straight.)

Length Contraction
Now let’s return to the cosmic-ray muon experiment described in the
previous lesson. The principle of relativity tells us that it is equally valid to
analyze this experiment from the reference frame of one of the muons, in
which it is at rest and the earth’s surface is rushing upward toward it at, say,
99.5% of the speed of light. But as I calculated above, the time interval
between Event A (summit of Mt. Washington rushes past muon) and Event B
(ground at sea level smashes into muon) is only 0.6 microseconds in this
frame of reference. How is that possible, if Mt. Washington’s height is 6
light-microseconds?
The answer is that in this frame of reference, the mountain is not 6 light-
microseconds high. Instead it is only 0.6 light-microseconds high (about 600
feet), because whenever we observe an object from a reference frame in
which it is moving at speed vv, it appears shorter, along the direction of
motion, by a factor of 1−(v/c)2−−−−−−−−√1−(v/c)2. In the muon’s
reference frame, the situation looks something like this:
With the mountain’s height contracted by a factor of 10, it passes the
stationary muon in 0.6 microseconds.
More generally, if we denote an object’s true length (in the frame in
which it is at rest) as L0L0 and its measured length (in the frame in which it’s
moving, along the direction of this length, at speed vv) as LL, then the
general formula for this relativistic length contraction effect is
L=L01−(v/c)2−−−−−−−−√.(11)(11)L=L01−(v/c)2.
So LL for a moving object is always less than L0L0, and the difference
between LL and L0L0 is negligible when v≪cv≪c.
By now you’ve surely noticed that the expression 1−(v/c)2−−−−−−
−−√1−(v/c)2 comes up a lot in relativity. For convenience we therefore
often use a standard abbreviation for it, or actually for its reciprocal:
γ=11−(v/c)2−−−−−−−−√.(12)(12)γ=11−(v/c)2.
The symbol is the Greek letter gamma, and this quantity is often called
the Lorentz factor, after the Dutch physicist H. A. Lorentz, who derived
many of the formulas of relativity several years before Einstein (but arguably
didn’t fully understand their meaning, as Einstein did). The Lorentz factor
equals 1 when v=0v=0, and increases to infinity as vv approaches the speed
of light. In terms of the Lorentz factor, the metric equation
reads ΔtAB=ΔsAB⋅γΔtAB=ΔsAB⋅γ and the length contraction formula
reads L=L0/γL=L0/γ. The main downside of using this abbreviation is that
in some situations there can be more than one relevant velocity, and then you
need to be clear about which velocity your γγ abbreviation depends on.
A 50-foot (50 light-nanosecond) log is lying on the ground. A bird flies past
the log, just above it and parallel to its length, at 3/5 the speed of light. Let
Event A be the bird passing the first end of the log, and let Event B be the
bird passing the other end of the log. (a) Draw an accurate spacetime diagram,
from the viewpoint of earth’s reference frame, showing the worldlines of both
ends of the log, the worldline of the bird, and Events A and B. (b) What is the
time between Events A and B, as measured by the squirrels sitting on the log?
(c) What is the time between Events A and B, as measured by the bird? (d)
From the bird’s point of view, the log is rushing past at 3/5 the speed of light.
How far does the log move, during the time between Events A and B,
according to the bird’s calculations? Explain carefully.
 While peacefully watching cloud formations in the desert, you suddenly see a
roadrunner zip by (beep, beep!) at half the speed of light, pursued by a coyote
running at the same speed. According to your measurements, the coyote is ten
meters behind the roadrunner. How far behind does the roadrunner think the
coyote is? (Hint: If the two creatures were holding a pole between them, in
whose reference frame would the pole be moving?)
By how much is the length of a 100-meter-long commuter train contracted in
a reference frame in which it is moving at 30 m/s? (Hint: You may find it
helpful to use the binomial approximation, (1+ϵ)n≈1+nϵ(1+ϵ)n≈1+nϵ,
which is accurate when |nϵ||nϵ| is much less than 1.)

Lesson 4: Two-Observer Spacetime

Diagrams
Let’s do another thought experiment. I’m standing at position zero in my
carefully constructed inertial reference frame, armed with a strobe lamp.
Some distance away in the +x+x direction I’ve placed a mirror, anchored to
my frame, facing me. At a certain time I flash my strobe lamp (Event F),
sending a single light pulse outward toward the mirror. The pulse bounces off
the mirror (Event B) and returns toward me so that I see it (Event S) some
time later.
Here is a spacetime diagram showing these events:
I’ve placed the origin event (Event O, when my wristwatch reads zero) half-
way between Events F and S. This arbitrary choice of when the time is
zero isn’t terribly important, but it creates a nice symmetry in the diagram.
The important feature is that whatever event is half-way between F and
S must occur at the same time as Event B. I know this because light always
travels at the same speed, and the light had to travel the same distance on its
outward and return trips, so those half-trips must have required equal
amounts of time. I’ve drawn the light signal worldlines at 45-degree angles,
because light always travels at exactly one light-second per second.
Perhaps you can guess what we’re going to do next: view all these events
from a different reference frame (“yours”), in which my entire laboratory
(including me, my strobe lamp, and the mirror) is moving—let’s say at half
the speed of light in the +x+x direction. And we’ll draw a second spacetime
diagram of the very same events, from your point of view.
On your spacetime diagram, my worldline runs diagonally up and to the
right, with a slope of 2 (seconds per light-second), since my velocity (in light-
seconds per second) is 1/2. Let’s label your time and space axes as tt and xx,
and (to distinguish them) label my axes as t′t′ and x′x′, as I’ve already done
on the diagram above. Notice that the t′t′ axis is the same as my worldline,
with a slope of 2 on your spacetime diagram. This axis is the line connecting
all events that happen at the spatial origin (x′=0x′=0) in my reference frame,
just as the tt axis connects all events happening at the spatial origin
(x=0x=0) in your reference frame. Because I’m moving with respect to you,
our time axes are different—and I hope you agree that there’s nothing
surprising about this fact.

For simplicity I’ve drawn this diagram so that the origin events of our two
reference frames coincide. I’ve drawn Events F and S at appropriate points on
my worldline, so that the origin event O is again half-way between them.
Question: Where on this diagram should I locate Event B?
To answer this question we use the startling fact that you also measure
the light pulse to move at exactly one light-second per second, despite the
fact that the strobe lamp that emitted the pulse is moving (with respect to
you) at half that speed. That seems impossible, right? But please suspend
your disbelief for a while, so we can work out the logical consequences. If
you really measure the light to be traveling at one light-second per second,
then we must draw the light-pulse worldlines at 45-degree angles even on
your spacetime diagram. I started them on the diagram above. Extrapolating
each of them to the right, we can locate Event B at the unique point where
these 45-degree lines intersect:

And what we find is that even though Events O and B are simultaneous
in my reference frame, they are not simultaneous in yours: You observe Event
B to occur after Event O, at some positive tt value. More generally, when
two events occurring in different places are observed to be simultaneous from
one frame of reference, they will be observed to occur at different times from
a frame of reference that is moving, with respect to the first, along the
direction that separates the events. This phenomenon is called relativity of
simultaneity.
To emphasize the relativity of simultaneity, I’ve added another element to
the diagram: the x′x′ axis. What do I mean by the x′x′ axis? It’s the line
connecting all events that happen at time zero in my reference frame, that is,
at t′=0t′=0. (Compare the xx axis, which connects all events that happen
at t=0t=0, and the tt and t′t′ axes, which are lines connecting all events that
happen at x=0x=0 and x′=0x′=0, respectively.) In our case, the x′x′ axis
must connect events O and B, because they both occur at t′=0t′=0.
Notice that the x′x′ axis is sloped upward from the xx axis by the same
amount that the t′t′ axis is sloped rightward from the tt axis. Or, equivalently,
the angle between the xx and x′x′ axes is the same as the angle between
the tt and t′t′ axes. In the present case, where my frame is moving with
respect to yours in the +x+x direction at v/c=1/2v/c=1/2, the t′t′ axis has
slope 2 while the x′x′ axis has slope 1/2. (I won’t present a rigorous proof
that the slopes are always related in this way, but the proof isn’t hard. If
you’d like to try it, draw another light-signal worldline passing through Event
O and then look for similar triangles.)
A spacetime diagram showing two sets of axes, for two different
reference frames, is called a two-observer spacetime diagram. To use the
diagram quantitatively, we can add gridlines for both coordinate systems:

For the unprimed frame (“yours”, shown in blue), each vertical gridline
connects all the events happening at a particular place (as measured in your
frame), while each horizontal gridline connects all the events happening at a
particular time (as measured in your frame). We can use whatever unit we
like for the interval between gridlines, with the understanding that the space
interval is however far light travels in one time interval.
For the primed frame (“mine”, shown in green), each mostly-vertical
gridline connects all the events happening at a particular place (as measured
in my frame), while each mostly-horizontal gridline connects all the events
happening at a particular time (as measured in my frame). Importantly, I’ve
spaced these gridlines so they correspond to the same time and space
intervals as the gridlines in your frame. How did I do this? Using the metric
equation! For instance, if Event P lies on the t′t′ axis where the t′=1t
′=1 gridline crosses this axis, then my wristwatch measures a spacetime
interval ΔsOP=1ΔsOP=1. But according to the metric equation, your clocks
should measure Event P to occur at ΔtOP=γ⋅ΔsOP=1/1−(1/2)2−−−−
−−−−√=1.155ΔtOP=γ⋅ΔsOP=1/1−(1/2)2=1.155, as shown in this
enlarged portion of the diagram:

This particular two-observer spacetime diagram is drawn for a primed

frame moving with respect to the unprimed frame in the +x+x direction at a
speed of 1/2 the speed of light (half a unit of distance in each unit of time).
You can download a printable page of “relativistic graph paper”, also drawn
for v/c=1/2v/c=1/2, at this link. To see what the gridlines would look like
for other choices of the relative velocity of the two reference frames, check
out this cool web app by Prof. Steven Sahyun of the University of Wisconsin
at Whitewater.
 Once we have our relativistic graph paper, we can plot events on it
according to one reference frame, then read off their coordinates in the other
reference frame:

Here I’ve plotted Event A at t=5t=5 and x=4x=4, as indicated by the blue
dashed lines. Then, using the green dashed lines, we can read off the
approximate primed-frame coordinates t′≈3.5t′≈3.5 and x′≈1.7x′≈1.7.
Similarly, I plotted Event B at t′=−3t′=−3 and x′=4x′=4, as indicated by
the second pair of dashed green lines. But as the second pair of dashed blue
lines shows, the approximate coordinates of this same event in the unprimed
frame are t≈−1.15t≈−1.15 and x≈2.9x≈2.9.
These coordinate transformations have the crucial property that if you
square the time and space coordinates and then subtract, you get the same
thing in either reference frame:
(t′)2−(x′)2=t2−x2.(13)(13)(t′)2−(x′)2=t2−x2.
For instance, for Event A, the right-hand side is 52−42=952−42=9, while
the left-hand side is approximately (3.5)2−(1.7)2≈9(3.5)2−(1.7)2≈9. But
computed in either frame, this quantity is just the square of the spacetime
interval between the origin event (call it O) and Event A, that
is, (ΔsOA)2(ΔsOA)2. Because the spacetime interval between two events is
the unique time between those events as measured by a single non-
accelerating clock that’s present at both events, we must obtain the same
spacetime interval when we compute it using the coordinates in either
reference frame. This “invariance” of the spacetime interval is analogous to
how you can calculate the squared distance between two points in ordinary
two-dimensional space as (Δx)2+(Δy)2(Δx)2+(Δy)2, and you’ll get the
same result no matter how you orient your xx and yy axes. (Note, however,
that the “Pythagorean theorem” for spacetime intervals has a minus sign
where the ordinary Pythagorean theorem for two-dimensional space has a
plus.)
As you might guess, there are also equations that you can use to carry out
these kinds of transformations between primed and unprimed spacetime
coordinates. They’re called the Lorentz transformation equations, and you
can find them in just about any textbook on relativity. They’re analogous to
the trigonometric equations that transform coordinates in ordinary two-
dimensional space when we rotate the axes (x,y)(x,y) by some angle. (The
Lorentz transformation equations can be written in terms of
the hyperbolic sine and cosine of the “angle” whose hyperbolic tangent
is v/cv/c.)
Event C occurs at t=1t=1 and x=−2x=−2 in the unprimed reference frame.
What are its coordinates in the primed reference frame represented in the
diagrams above, moving rightward at half the speed of light with respect to
the unprimed frame? Show your construction on a copy of the two-observer
spacetime diagram. Also check that (t′)2−(x′)2=t2−x2(t′)2−(x′)2=t2−x2.
Event D occurs at t′=1t′=1 and x′=4x′=4 in the primed reference frame
represented in the diagrams above. What are its coordinates in the unprimed
frame, which is moving leftward at half the speed of light with respect to the
primed frame? Show your construction on a copy of the two-observer
spacetime diagram. Also check that (t′)2−(x′)2=t2−x2(t′)2−(x′)2=t2−x2.
At high noon, a solar flare erupts on the surface of the sun. Half an hour later,
at 12:30, a comet crashes into Jupiter, 780 million km away from the sun.
(These data are as measured in earth’s reference frame, which is moving at
negligible speed with respect to the sun and Jupiter.) Meanwhile an alien
spaceship zips by at 0.8c0.8c, headed in the direction from the sun toward
Jupiter. (a) Draw a spacetime diagram, calibrated in minutes of time and
light-minutes of space, showing the sun, Jupiter, the flare eruption, and the
 comet crash. (b) Add t′t′ and x′x′ axes for the alien spaceship’s reference
frame, and determine which event (solar flare or comet crash) occurs first in
the aliens’ frame. (c) Which of the two events do the aliens see first? Draw
worldlines to represent the light from each event traveling toward the alien
spaceship, and show that the answer depends on where the spaceship is
located within its reference frame (which I haven’t specified).
Redraw the two-observer spacetime diagram in the text above from the
viewpoint of the primed reference frame, so the t′t′ axis points straight up and
the x′x′ axis points straight to the right. Since the unprimed frame is now
moving to the left at half the speed of light, this means that the tt axis will
point up and to the left, with a slope of −2−2. Plot Event A according to its
unprimed coordinates, and check that its primed coordinates are the same as
what I found above. Plot Event B according to its primed coordinates, and
check that its unprimed coordinates are the same as what I found above.
Draw a calibrated two-observer spacetime diagram for the case where the
primed frame moves at v=0.7cv=0.7c with respect to the unprimed frame.
What are the primed coordinates of an event that occurs
at t=5t=5 and x=6x=6? What are the unprimed coordinates of an event that
occurs at t′=−2t′=−2 and x′=4x′=4?
Draw a two-observer spacetime diagram for an imaginary universe in which
time is absolute, so t′=tt′=t for every event. Label the axes and include
gridlines.

Length Contraction Revisited

One handy use of a two-observer spacetime diagram is to give us a better
perspective on length contraction. Suppose there’s a four-unit-long stick
that’s at rest with respect to the primed frame, with one end at x′=0x′=0 and
the other end at x′=4x′=4:
I’ve drawn the worldlines of each end of the stick in red, and shaded the
spacetime region in between to highlight what we might call the stick’s
world-sheet. You can see that the stick is four units long as measured in the
primed frame, because events O and F′′, which are simultaneous in the
primed frame, are four grid-spacings apart along the x′x′ axis. But how long
is the stick in the unprimed frame? To answer that question we should find
two events, one at each end of the stick, that are simultaneous in
the unprimed frame. The most convenient such events are O and F, and as
you can also plainly see, they are only about 3.5 units apart along the xx axis.
The stick appears length-contracted from the unprimed frame, in which it is
moving.
To check for consistency, let’s look next at what happens if the stick is at
rest with respect to the unprimed frame:
Now to determine the length in the unprimed frame we can look at events O
and G, which are four units apart along the xx axis. But to determine the
length in the primed frame we should look (for instance) at events O and G′′,
which are simultaneous in the primed frame. And sure enough, these events
are only about 3.5 grid spacings apart, as measured along the x′x′ axis.
Length contraction is really best understood as a side-effect of the
relativity of simultaneity: Different observers disagree on which pairs of
events (one at each end of the stick) are simultaneous, and therefore they
disagree about how far apart these events are, which is what they mean by the
stick’s length. An analogous phenomenon in ordinary two-dimensional space
would be how the width of a road appears different, depending on whether
you cross it at a right angle or at some other angle. But whereas the road
appears wider when you cross it along a diagonal, the minus sign in the
metric equation makes a stick appear shorter when you view it from a
reference frame in which it is moving.
Use the length contraction formula to check that a stick with a rest length of 4
units should appear roughly 3.5 units long in a reference frame in which it is
moving at half the speed of light.
 You own a 10-foot ladder that you would like to store inside an 8-foot-long
shed. Having studied relativity, you figure you can do it as long as you run
fast enough, holding the ladder horizontally so its contracted length is just
8 feet. (How fast is that?) But before putting your plan into action, you
explain it to your spouse, who expresses skepticism: “In your frame of
reference, won’t the ladder still be 10 feet long, while the moving shed is
contracted to 80% of its usual length, that is, just 6.4 feet? A 10-foot ladder
won’t fit inside a 6.4-foot shed!” Now you’re both puzzled. How can the
ladder fit (even momentarily) inside the shed as viewed from one reference
frame, but not fit as viewed from another frame? Does this paradox prove that
relativity is illogical nonsense? To resolve the paradox, draw a two-observer
spacetime diagram showing the shed (at rest) and the ladder (in motion),
identifying the key events when/where the ends of the ladder pass the ends of
the shed. For the sake of safety, please assume that the shed has an open door
at each end. Explain carefully what happens from the perspective of each
frame of reference, and why there is no logical contradiction.

Combining Velocities
Now let’s move on to a completely new example. Suppose I’m at rest in the
primed frame, moving with respect to you in the positive direction at half the
speed of light, and I toss a baseball forward at half the speed of light with
respect to me. How fast is the baseball moving with respect to you?
If you didn’t know anything about relativity, you would probably answer
this question by adding one-half to one-half to obtain one, that is, one times
the speed of light. But by now you may be more wary of such simple
answers.
To answer this question I’ve carefully drawn the baseball’s worldline on a
two-observer spacetime diagram below. I started the worldline at the origin
and then, looking only at the diagonal green gridlines, measured one unit of
space (along the x′x′ axis) and two units of time (parallel to the t′t′ axis), to
find another event along the worldline, under the assumption that I measure
the baseball to be moving at half the speed of light. I then repeated this
process to extend the worldline further, and also extended it backward from
the origin. The events that I used to draw this line are highlighted with red
dots, and you’ll notice that they’re all at intersections of the green gridlines.
Amazingly, the baseball’s worldline is steeper than 45 degrees, indicating
that you measure the ball to be moving somewhat slower than the speed of
light. And to find its actual speed, you can just look at the blue gridlines! I’ve
conveniently chosen the numbers in this example so the baseball’s worldline
passes exactly through the blue grid point at t=5t=5 and x=4x=4 (check
this!), meaning that you measure the ball’s speed to be only 4/5 the speed of
light.
This example is a special case of the famous Einstein velocity
transformation formula. Before I write the formula in general, I need to
carefully define symbols for the three different velocities that we’re talking
about:
 uxux = velocity of the baseball with respect to you (in the unprimed
frame);
 u′xux′ = velocity of the baseball with respect to me (in the primed
frame);
 vxvx = velocity of me (the primed frame) with respect to you (the
unprimed frame).
(Of course the object doesn’t have to be a baseball, but “baseball” seems
easier to remember than “object”.) Let’s also agree that all three of these
velocities are to be measured as fractions of the speed of light. The general
formula is then:
ux=u′x+vx1+u′xvx.(14)(14)ux=ux′+vx1+ux′vx.
Notice that the numerator of this formula is what we would expect if we
didn’t know anything about relativity: just add the two velocities! But the
denominator contains a “correction” term that’s the product of the two
velocities, measured as fractions of the speed of light. At ordinary speeds
these fractions would be tiny and their product would be tinier still, so we
could simply neglect this correction term. But when u′x=vx=1/2ux
′=vx=1/2, we obtain
ux=12+121+12⋅12=11+14=45,(15)(15)ux=12+121+12⋅12=11+14=45,
just as we already saw from the diagram.
Applying the velocity transformation formula to other examples can be
tricky, because it’s not always obvious which reference frame should be the
primed frame, which should be the unprimed frame, and which object should
correspond to the baseball. There are always multiple correct ways to set up
these correspondences, but you need to be consistent. Any of the three
velocities is allowed to be negative, and you often need to pay special
attention to minus signs. The best advice I can give you is to draw a picture
showing which way things are going and which direction you’re calling
positive; then write out, in English, exactly what you mean by each of the
three symbols. As a check, remember that if you neglect the correction term
in the denominator, you should get the answer you would expect if you didn’t
know about relativity.
What if, instead of a baseball, I “toss” a light pulse? Assuming that the pulse
moves forward at the speed of light with respect to me, while I move at half
the speed of light with respect to you, how fast does the light pulse move with
respect to you? Answer this question using a two-observer spacetime
diagram, then answer it again using the Einstein velocity transformation
formula. Finally, do it again (using both methods) for a light pulse that I
“toss” in the backward direction.
You are fleeing from Planet Vogsphere at speed 0.99c0.99c (with respect to
the planet) when your spaceship’s antimatter drive malfunctions, making
 further acceleration impossible. Knowing the Vogons are in hot pursuit, you
climb into your escape pod and set it to be launched forward at the maximum
speed of 0.95c0.95c (with respect to your spaceship). Once the pod is
launched, how fast is it going with respect to Vogsphere?

A supersonic jet, moving with respect to the ground at 1000 m/s, fires a
supersonic missile in the forward direction at a speed of 1000 m/s with
respect to the jet. What is the missile’s speed with respect to the ground? By
what percentage does the answer differ from the naive prediction, 2000 m/s?
A distant quasar is moving away from earth at speed 0.35c0.35c. The quasar
emits a jet of plasma in the direction toward earth. Astronomers on earth
measure the jet to be approaching at speed 0.27c0.27c. What is the jet’s
velocity with respect to the quasar?

The Cosmic Speed Limit

As you may now suspect, the rules of relativistic velocity transformations
make it impossible to combine two velocities less than 1—or even equal to 1
—to obtain a velocity greater than 1. The speed of light (which equals 1 in
the units I’m using) seems to be some sort of limit, which you can’t cross by
building up to it in stages.
But it’s still fair to ask whether there might be some sort of object or
signal that inherently travels faster than the speed of light, much as
electromagnetic waves inherently travel at the speed of light. As a final
application of two-observer diagrams, let’s ask whether such a thing is
possible.
Suppose, for instance, that you have a device capable of sending signals
at three times the speed of light. You aim your device at a friend located six
light-seconds away in the +x+x direction, and press the button
at t=0t=0 (Event O) to send a secret message. Traveling at three times the
speed of light, the message should take two seconds to reach your friend;
let’s call the arrival of the message Event C. Plotted on a spacetime diagram
(from the perspective of your reference frame), the events and the signal’s
worldline would look like this:
And perhaps you can now see the problem. Although there’s nothing wrong
with this diagram from your perspective, it’s nonsense from my perspective,
if I’m moving at half the speed of light in the positive direction with respect
to you. Whereas you observe a signal traveling from Event O to Event C at
three times the speed of light, I observe a signal traveling from Event
C to Event O (also at tremendous speed), because in my reference frame
Event C occurs before Event O (slightly more than one second before,
according to the diagram).
More generally, if you draw any purported worldline representing a signal
traveling faster than the speed of light on a spacetime diagram, I can always
draw an x′x′ axis for some primed reference frame, moving at less than the
speed of light with respect to you, that’s steeper on the diagram than your
signal’s worldline. From the perspective of this primed reference frame your
signal arrives before it was sent, or equivalently, it goes backwards in time.
In short: you show me a signal that travels faster than the speed of light, and
I’ll show you a reference frame in which that signal is traveling backwards in
time.
 Now perhaps the idea of signals going backwards in time doesn’t bother
you. But they’re certainly contrary to all of our experience—except, of
course, in science fiction stories. Moreover, this idea seems to lead to all sorts
of logical paradoxes, such as the question of who really wrote the song
“Johnny B. Goode” (in the movie Back to the Future), if Marty McFly
learned it from Chuck Berry, but Chuck Berry learned it from a time-
traveling Marty McFly.
The easiest way out of this paradox is simply to assume that it’s
impossible to send any signals faster than the speed of light. Any signal
traveling slower than the speed of light, or even at the speed of light, has a
steep enough worldline that it travels forward in time with respect to all
inertial reference frames (which we also assume must travel slower than the
speed of light). In this context we refer to the speed of light as the cosmic
speed limit —a fundamental property of spacetime that affects all
measurements and all motion. From this perspective, the fact that
electromagnetic waves happen to travel right at the cosmic speed limit is
interesting, but the cosmic speed limit itself is more fundamental than
electromagnetism.
At exactly 7:00 am, a charge of dynamite explodes at a road construction site
in the Rocky Mountains. At 7:00:00.0005 am (half a millisecond later), a
mysterious shaking is felt at a diner in Denver (200 km east of the
construction site), causing a cup of coffee to fall off the counter. Could the
explosion have caused the shaking? To answer this question, draw a
spacetime diagram that accurately shows the space and time separations of
the two events. Then consider whether there exists a frame of reference in
which the shaking occurred before the explosion. If there is such a frame,
which way must it be moving with respect to the earth, and how fast? If there
is no such frame, how can you tell?

Lesson 5: Momentum and Energy

Two-observer spacetime diagrams invite us to think of tt and xx as two
components of a single spacetime vector, analogous to the position
vector (x,y)(x,y) in ordinary two-dimensional space. Just as the components
of a position vector change if we use a rotated coordinate system, so also the
components of a spacetime vector change if we “boost” to a different inertial
reference frame (moving with respect to the original frame). If we include all
three dimensions of space, then the coordinates of any event form what we
call a four-vector, (t,x,y,z)(t,x,y,z).
In this final lesson I want to tell you about another important four-vector.
But before I do, let’s back up and think about why we use vectors in the first
place, even in plain old three-dimensional space. The most apparent reason is
for brevity of notation: we can write a single equation like
F⃗ =ma⃗ orp⃗ final=p⃗ initial(16)(16)F→=ma→orp→final=p→initial

instead of writing out three separate equations for the xx, yy,
and zz components. But there’s a more important reason besides brevity.
When we write that one vector equals another, we’re making a statement
about the vectors themselves, independent of how we orient our coordinate
axes to define their xx, yy, and zz components. This means that if a vector
equation is true in one coordinate system, it must also be true in any rotated
coordinate system. Writing the laws of physics in terms of vectors doesn’t
ensure that these laws are correct, but at least it ensures that they’re
consistent with the principle that space doesn’t have any “preferred
directions”; our choice of coordinate axes is arbitrary.
In a completely analogous way, writing the laws of physics using four-
vectors will ensure that if these laws are true in one inertial frame of
reference, then they will be true in all inertial frames of reference. In other
words, using four-vectors ensures that the equations we write will be
consistent with the principle of relativity.
With this principle in mind, let’s now think about some of the laws of
physics.
In Newtonian mechanics, after you learned the kinematic concepts of
position, time, velocity, and acceleration, you went on to study dynamics:
force, mass, Newton’s second law, and the laws of conservation of
momentum and energy.
We could now revisit each of these concepts in the context of relativity,
but it turns out that the relativistic version of Newton’s second law isn’t
nearly as useful as we might have guessed. Instead it’s more efficient to skip
over the concept of force and go straight to the relativistic version of
momentum.
Momentum Conservation
Let’s consider a simple one-dimensional momentum conservation problem. A
1-kg block is gliding frictionlessly (or drifting through space) at exactly 20
m/s, toward an identical 1-kg block that’s initially at rest. The blocks then
collide and stick together, conserving momentum because they form an
isolated system:

This view of the collision is from what I’ll call the “Home” reference frame.
Because the initial momentum of the system is 20 kg m/s and momentum is
conserved, the final velocity of the combined 2-kg block must be exactly 10
m/s.
Now let’s view this same collision from what I’ll call the “Other”
reference frame, which is moving to the right at exactly 10 m/s with respect
to the Home frame. In the Other frame the final velocity of the combined
blocks is zero, while the initial velocity of Block 2 is exactly −10 m/s:

But what’s the initial velocity of Block 1? If we didn’t know about relativity
we would simply subtract 10 m/s (the Other frame’s velocity) from 20 m/s
(Block 1’s velocity in the Home frame) to obtain 10 m/s. But the Einstein
velocity transformation tells us that this isn’t exactly right. For if we work
backwards, combining the 10 m/s velocity of the block in the Other frame
with the 10 m/s velocity of the Other frame with respect to the Home frame,
we would get a value very slightly less than 20 m/s for the block’s velocity
back in the Home frame. In order for this velocity to come out to exactly 20
m/s, the velocity of Block 1 with respect to the Other frame must instead be
very slightly greater than 10 m/s.
And now we have a problem: As viewed from the Other frame, the final
momentum of this system is exactly zero but the initial momentum is not; in
fact it is slightly positive. By assuming that momentum was conserved in the
Home frame, I’ve proved that momentum is not conserved in the Other
frame. Momentum conservation is therefore incompatible with the principle
of relativity, which requires that the laws of physics are valid in all inertial
reference frames.
So what do we do? One option would be to simply give up, and conclude
that momentum conservation isn’t a law of physics after all. That’s
conceivable, but it would be a sad outcome and it wouldn’t explain why
momentum conservation works so well at low speeds.
To make the discrepancy more dramatic, consider the collision example
above but multiply all the speeds by 107107, so Block 1 is initially moving at
200,000,000 m/s (2/3 the speed of light), and the blocks’ final speed is
100,000,000 m/s (1/3 the speed of light). (Try to ignore the absurdity of two
“blocks” simply sticking together after such a violent collision!) If the Other
frame is again moving along with the blocks after the collision, what is the
initial speed of Block 1 in the Other frame, according to the Einstein velocity
transformation rule? What is the system’s initial (Newtonian) momentum in
the Other frame?

Relativistic Momentum
Fortunately, there’s another option: Modify the definition of momentum!
Perhaps the formula we’re using for momentum is only approximately
correct—accurate enough at low speeds, but inaccurate at higher speeds.
And what is our definition, exactly? Well, it’s mass times velocity, for
instance,
px=mvx=mdxdt(old definition).(17)(17)px=mvx=mdxdt(old
definition).
In Newtonian mechanics, this formula defines a perfectly good vector
component because xx itself is a valid vector component,
while mm and dtdt are scalars, that is, numbers that are the same in all
coordinate systems. (The quantity dxdx is basically the difference between
two xx values, final minus initial, but subtraction of vectors works
component-wise, so this subtraction doesn’t affect the status of the expression
as a valid vector component.)
And now, perhaps, you can see the issue: In four-dimensional relativistic
spacetime, the denominator dtdt is no longer a scalar because it is
a coordinate time interval, different in different reference frames. But there’s
a straightforward fix! Instead of putting the coordinate time difference in the
denominator, we can use the proper time difference dτdτ (which for
infinitesimal time intervals is the same as the spacetime interval dsds). This
is the time interval as measured by the object’s own clock, so it is a true
scalar quantity, independent of any reference frame. Our new definition of
momentum is therefore
px=mdxdτ=γmdxdt(new definition),(18)
(18)px=mdxdτ=γmdxdt(new definition),
where in the final expression I’ve used the metric equation to
relate dτdτ to dtdt:
dτ=dt1−(v/c)2−−−−−−−−√=dtγ.(19)(19)dτ=dt1−(v/c)2=dtγ.
The Lorentz factor γγ is very close to 1 at low speeds, which is why we never
notice that we need it in everyday situations. But at speeds close to the speed
of light, the extra factor of γγ in the definition of momentum makes a big
difference.
In three spatial dimensions, by the way, the Lorentz factor depends on all
three components of the velocity:
1γ=1−(v2x+v2y+v2z)/c2−−−−−−−−−−−−−−−−−√.(20)
(20)1γ=1−(vx2+vy2+vz2)/c2.
Surprisingly, this implies that the xx component of an object’s momentum
depends on all three components of its velocity! Meanwhile, the momentum
vector also has yy and zz components,
py=mdydτ=γmdydt,pz=mdzdτ=γmdzdt,(21)
(21)py=mdydτ=γmdydt,pz=mdzdτ=γmdzdt,
each of which also depends, through γγ, on all three velocity components.
 How fast would an object need to be moving for its relativistic momentum
(mdx/dτmdx/dτ) to exceed its Newtonian momentum (mdx/dtmdx/dt) by
one percent? Express your answer as a fraction of the speed of light and also
in meters per second.

The Time Component of the Four-Momentum

But wait a minute: Vectors in spacetime are supposed to have not just three
components but four. That’s because in spacetime we’re allowed not only to
rotate our xx, yy, and zz axes, mixing the three spatial components of any
vector, but also to boost to a different inertial reference frame, mixing
the time component of any vector with its space components. So if
momentum conservation is to be a true law of physics, valid in all inertial
reference frames, the momentum vector must also have a time component
and this time component must also be conserved. And what is that time
component? It must be related to dtdt in the same way that the space
components are related to dxdx, dydy, and dzdz:
pt=mdtdτ=γmdtdt=γm=m1−(v/c)2−−−−−−−−√.(22)
(22)pt=mdtdτ=γmdtdt=γm=m1−(v/c)2.
For an object at rest, this quantity is simply the mass mm. And we normally
think of mass as a conserved quantity, so that’s a good sign! For an object in
motion, the quantity ptpt is greater than the mass, by an amount that depends
on the object’s speed but not on its direction of motion. It’s a lot greater at
speeds close to cc, but at much lower speeds it’s only a little greater.
To better understand this quantity ptpt, I’d like to simplify its formula in
the familiar limit of low speeds, v≪cv≪c. In this limit we can use a handy
formula called the binomial approximation, which says that 1 plus something
small, all raised to some power, is equal to 1 plus the product of the power
times the small thing:
(1+ϵ)n≈1+nϵwhen |nϵ|≪1.(23)(23)(1+ϵ)n≈1+nϵwhen |nϵ|≪1.
(If you’re not already familiar with this approximation, please add it to your
mathematical toolbox. It’s incredibly useful not just in relativity but in many
branches of science and engineering.) To apply the binomial approximation
to the formula for ptpt, I want to identify ϵϵ as −(v/c)2−(v/c)2 and
identify nn as −1/2−1/2. Then for v≪cv≪c I can approximate
 γ=(1−(vc)2)−1/2≈1+(−12)(−v2c2)=1+v22c2(when v≪c).
(24)γ=(1−(vc)2)−1/2(24)≈1+(−12)(−v2c2)=1+v22c2(when v≪c).
Inserting this approximation into the definition of ptpt then gives

pt=γm≈m+mv22c2(when v≪c).(25)(25)pt=γm≈m+mv22c2(when
v≪c).
So at low speeds, the time component of the momentum four-vector is approximately equal to
the mass, plus a small correction term that’s starting to look a lot like kinetic energy (another
conserved quantity!). There’s an extra factor of c2c2 in the denominator, but an overall
constant factor won’t affect whether this quantity is conserved.
What we therefore do is multiply ptpt by c2c2, and refer to this quantity as simply EE,
the relativistic energy of the object:
E=ptc2=γmc2 (at any v)≈mc2+12mv2(when v≪c).(26)(27)
(26)E=ptc2=γmc2 (at any v)(27)≈mc2+12mv2(when v≪c).
For the special case of an object at rest, we have simply E=mc2E=mc2 (an equation that
you may have seen before); this quantity is called the rest energy of the object. At low speeds, an
object’s total energy is its rest energy plus the familiar Newtonian formula for kinetic energy (to
a good approximation). And at high speeds, we can compute its total energy from the exact
formula γmc2γmc2, and we define the kinetic energy to be the amount by which this exceeds
the rest energy:
Kinetic energy=E−mc2(at any v).(28)(28)Kinetic energy=E−mc2(at
any v).
In any case, the relativistic energy must be a conserved quantity—and it’s the total energy that’s
conserved, not the rest energy or kinetic energy separately.
Calculate γγ for v/cv/c equal to 0.1, 0.01, and 0.001. Compare each value to
what you get using the binomial approximation, and comment on the results.

Relativistic Energy
Let me now summarize this remarkable story. First I showed that the old Newtonian definition of
momentum is incompatible with the principle of relativity. To fix this problem, I suggested a
modification to the definition of momentum that seems to have the desired vector properties. But
if momentum conservation is to be a true law of physics, valid in all inertial reference frames,
then the momentum vector can’t have just three components; it must also have a time
component, which must also be conserved. Finally, I showed that this time component is closely
related to the familiar concepts of mass and energy. I inserted a factor of c2c2 to give it the
right units, and arrived at the relativistic version of energy conservation.
 The implications of relativistic energy conservation are plentiful and astounding.

In the case v=0v=0, the equation E=mc2E=mc2 tells us that the mass of any object
is effectively a measure of its total energy content when it is at rest; the factor of c2c2 merely
converts between mass units and energy units. But look at the numbers! A 1-kg object at rest has
a total energy of
E=(1 kg)(3×108 m/s)2=9×1016 J,(29)(29)E=(1 kg)(3×108 m/
s)2=9×1016 J,
or a little over 20 megatons (of TNT explosive equivalent). It’s probably a good thing that
nobody knows a quick and practical way to convert all this energy to other forms!

Speaking of explosives, a typical combustion reaction releases about 107107 joules of

energy for every kilogram of reacting chemicals. To pick a specific number, when hydrogen and
oxygen burn to form water, the energy released (usually as heat to the surroundings) is about 16
MJ per kilogram. This might sound like a lot of energy, but it’s nothing compared to the rest
energy, which is again nearly 10171017 joules. In principle we could weigh the hydrogen and
oxygen gases before the reaction, then weigh the cooled-off water after the heat has dissipated,
and the difference, times c2c2, should equal 16 MJ (for one kilogram of reactants). In practice,
nobody has ever made a balance accurate enough to measure such tiny mass differences (roughly
one part in 10101010).
Where we can measure these mass differences is in reactions of individual atomic nuclei
and subnuclear particles—because these reactions tend to convert much larger fractions of the
total energy from one form to another. The exercises below give several examples.
Many reactions in nuclear and particle physics involve high-energy particles of light,
called photons (also called gamma rays). Naturally these particles travel at the speed of light. But
you might be puzzled by the fact that the Lorentz factor γγ (coincidentally the same symbol to
denote a γγ ray) is infinite when v=cv=c, apparently implying that every photon
carries infinite momentum and energy. That’s not the case, and the only way to reconcile the
facts with the formula is if every photon has zero mass. Then the product γmγm becomes
ambiguous (infinity times zero) for photons in all the formulas above, and those formulas
become merely useless for photons, not contradictory. In fact, a photon can have any amount of
energy and any momentum vector, though these two quantities are related. To see how, consider
the ratio
pxE=γmvxγmc2=vxc2,(30)(30)pxE=γmvxγmc2=vxc2,
in which γγ and mm both cancel out. This relation works just as well for a photon as for any
other object. If the photon is moving in the +x+x direction then vx=cvx=c, so this ratio is
simply 1/c1/c, implying E=pxcE=pxc. (One can also derive this energy-momentum
relation for light using Maxwell’s equations, but that derivation is much more difficult.)
If, on the other hand, an object has a nonzero mass, then the
formula E=γmc2E=γmc2 implies that we would have to give it an infinite amount of
energy to accelerate it up to the speed of light. This is another way of understanding the cosmic
speed limit.

Several of the exercises below involve decays and other reactions of subatomic
particles, so I’ve gathered the relevant masses and rest energies into the following
table. Note that in atomic and nuclear physics it is common to measure masses in
atomic mass units, also called daltons (abbreviated u or Da), where 1 u is the
approximate mass of a proton or neutron, and is defined as exactly 1/12 the mass of a
carbon-12 atom. It is also customary to measure energies in electron-volts (eV), where
1 eV is the energy that a one-volt battery gives to each electron it pushes around a
circuit, 1.60×10−191.60×10−19 joules. The table gives rest energies in MeV,
where M (mega) is the metric prefix for 106106. For comparison I’ve also included
masses in kilograms.

Particle Masses and Rest Energies

Particle
Symb
Mass (kg) Mass (u)
mc2mc2 (M
ol eV)

Photon γγ 00 00 00
Neutrino νν ∼0∼0 ∼0∼0 ∼0∼0
Electron/ 9.11×10−319.11×1 0.000550.00
positron
e∓e∓ or β∓β∓ 0.510.51
0−31 055
1.88×10−281.88×1 0.113430.11
Muon μ±μ± 105.66105.66
0−28 343

2.41×10−282.41×1 0.144900.14
Pion (neutral) π0π0 134.98134.98
0−28 490

2.49×10−282.49×1 0.149830.14
Pion (charged) π±π± 139.57139.57
0−28 983

1.67×10−271.67×1 1.007281.00
Proton pp 938.27938.27
0−27 728
Neutron nn 1.67×10−271.67×1 1.008661.00 939.57939.57
 0−27 866

Alpha (44He 6.64×10−276.64×1 4.001514.00 3727.383727.

αα
nucleus) 0−27 151 38

Estimate the mass increase of a cup of tea as its temperature increases from
room temperature to the boiling point. (It takes 4.2 J of energy to raise each
gram of water by one degree Celsius.)

Although the metric equation is good news for space travelers who wish to
reach distant stars within a human lifetime, the energy requirements pose a
challenge. Suppose, for instance, that you wish to take a trip on a starship that
travels at 4/5 the speed of light. If you and your luggage have a combined
mass of 100 kg, how much kinetic energy must you (including your luggage)
acquire to reach cruising speed? Convert your answer to kilowatt-hours
(kWh), then look up the current per-kWh cost of electrical energy in your
area. Use these numbers to estimate the cost of a ticket on this starship.
Discuss the assumptions behind your estimate, and the practicality of
interstellar travel.
The sun has a mass of 2×10302×1030 kg and radiates energy at a rate of
about 4×10264×1026 watts. How much mass does the sun lose in each
second? At the rate it’s going, how long would it take the sun to lose one
percent of its mass?
Starting from Avogadro’s number 6.022×10236.022×1023, determine the
number of kilograms in one atomic mass unit (to four significant figures). Use
this conversion factor to check a couple of the entries in the Particle Masses
and Rest Energies table above. Then calculate the rest energy in MeV of a
hypothetical particle whose mass is exactly 1 u, and use this conversion factor
to check a couple of the entries in the table.
Technetium-99m (99mTc99mTc) is an excited, metastable state of the isotope
technetium-99 that is commonly used in diagnostic medical procedures. It
spontaneously decays to ordinary 99Tc99Tc with a half-life of 6 hours,
emitting a photon (gamma ray) with energy 0.140 MeV. By about what
fraction does the mass of the technetium nucleus (or atom) decrease when it
loses this energy? (Hint: The atomic mass number 99 tells you that a mole
of 99Tc99Tc has a mass of 99 grams, where a mole
equals 6×10236×1023 atoms. The kinetic energy of the recoiling Tc nucleus
is negligible, but it’s a nice exercise to verify this using momentum
conservation.)
 An alpha particle, or helium-4 nucleus, consists of two protons plus two
neutrons, held together by the strong nuclear force. Notice from the table
above that its mass is somewhat less than the total mass of its constituent
protons and neutrons. How much energy would you need to provide, in MeV,
in order to separate these constituents from each other? (This quantity is
called the binding energy of the nucleus.) What is the binding energy as a
fraction of the alpha particle’s rest energy?
A uranium-238 nucleus decays (into thorium-234, with a half-life of 4.5
billion years) by emitting an alpha particle. If the uranium nucleus is initially
at rest, then the alpha particle comes out at a speed
of 1.42×1071.42×107 m/s. (a) Calculate the kinetic energy of the emitted
alpha particle in MeV. How accurate is the nonrelativistic
formula K=12mv2K=12mv2? (b) To conserve momentum, the thorium
nucleus must recoil as the alpha particle is emitted. What is its recoil speed?
(c) What fraction of the rest energy of the uranium nucleus is converted into
kinetic energy by this decay?

A free neutron (not bound inside a nucleus) is unstable, decaying with a half-
life of about 10 minutes into a proton, an electron, and a neutrino (technically
an antineutrino, but the distinction doesn’t matter here):
n⟶p+e−+ν.n⟶p+e−+ν.
Referring to the table for the needed mass data, determine the total kinetic energy of the products of
this decay. What fraction of the neutron’s rest energy is this?

The nuclear fusion reaction that powers our sun combines four protons and
two electrons to form a helium nucleus, also called an alpha particle, along
with several photons and neutrinos:
4p+2e−⟶α+photons+neutrinos.4p+2e−⟶α+photons+neutrinos.
What is the net gain in kinetic energy during this reaction? What fraction of the reactants’ rest energy is
converted to kinetic energy? (Refer to the table for the needed mass data.)

The neutral pion has a very short half-life of less than 10−1610−16 seconds.
It normally decays into a pair of photons (gamma rays). What are the energies
of these photons, assuming that the pion is initially at rest? Why is it
impossible for a neutral pion to decay into a single photon?

The positron is the electron’s so-called antiparticle, with the same mass as an
electron but opposite electric charge. When a positron interacts with an
electron, they can annihilate into two or more photons:
 e++e−⟶photons.e++e−⟶photons.
(a) Prove that there must be at least two photons produced in this reaction: annihilation into a single
photon is not possible. (b) If the positron and electron are initially at rest and just two photons are
produced, what are the photons’ energies?

The LEP collider at the CERN laboratory in Geneva accelerated electrons and
positrons to a maximum energy of 104,000 MeV (104 GeV) each. How does
the energy of such a particle compare to its rest energy? What is the velocity
of such a particle, expressed as a fraction of cc? (Hint: Solve for v/cv/c in
terms of γγ and then use the binomial expansion to determine the amount by
which v/cv/c differs from 1.)

Often an object’s energy and momentum are of more interest to us than its
velocity, so we want an equation that relates energy to momentum and mass,
with velocity eliminated. Show that for any object, this relation is
E2−|p⃗ c|2=(mc2)2,E2−|p→c|2=(mc2)2,
where p⃗ p→ represents the three spatial components of the momentum. Like the metric equation,
this equation tells us to subtract the squares of the three spatial components of a four-vector from the
square of its time component, resulting in a quantity that is frame-independent (a so-called Lorentz
scalar). What does this equation tell us about a massless particle such as the photon?

A charged pion usually decays (with a half-life of 18 ns) into a muon plus a
neutrino. Use both energy and momentum conservation to determine the
neutrino’s energy and the muon’s kinetic energy. (Hint: Use the energy-
momentum-mass relation from the previous problem, rather than working
with the muon’s velocity as a variable.)
A gamma-ray photon collides with a free electron that is initially at rest. This
problem and the next explore some possible outcomes. (a) First use
momentum and energy conservation to prove that the electron cannot simply
absorb the photon (a reaction we would write as e−+γ⟶e−e−+γ⟶e−).
This means that the simplest possible reaction is e−+γ⟶e−+γe−+γ⟶e−
+γ, with a single photon in the final state; this reaction is called Compton
scattering. (b) Suppose that the Compton-scattered photon comes straight
back, opposite to the initial photon’s direction of motion, while the electron
recoils in the initial photon’s direction. Use relativistic momentum and energy
conservation to derive a formula for the final photon energy in terms of the
initial photon energy and the electron’s rest energy mc2mc2. (c) Discuss
what happens in the limiting cases where the initial photon’s energy is much
less than and much greater than the electron’s rest energy.
 Repeat part (b) of the previous problem for the general case in which the final
photon’s direction makes an angle θθ with the initial photon’s direction. Now
there are two nontrivial momentum components, which are separately
conserved. You should obtain the Compton formula,
1Ef=1Ei+1mc2(1−cosθ),1Ef=1Ei+1mc2(1−cos⁡θ),
where EiEi and EfEf are the initial and final photon energies, respectively. Discuss the predictions of
this formula for some interesting values of θθ and Ei/mc2Ei/mc2.

Let’s return to the thought experiment of the two colliding blocks, at

the beginning of this lesson. In that example I assumed not only that
momentum is given by the Newtonian formula mv⃗ mv→, but also that we
could simply add the masses of the two blocks to obtain the mass of the final,
combined block. Now we know better: The kinetic energy lost in the collision
is converted to thermal energy, which adds to the mass of the combined
block. If we take the final mass and the final velocity to be our two
unknowns, then we need two equations to determine these two unknowns.
(a) It’s easier to work algebraically at first, so let each initial block have
mass mm and let v0v0 be the initial velocity of Block 1, with Block 2
initially at rest. Assuming that the two blocks stick together and that all
energy remains in the combined block, write the two equations that express
momentum and energy conservation. Combine these equations to eliminate
the final mass and obtain an expression for the final velocity in terms of v0v0.
(b) Low speeds like 20 m/s make relativistic effects hard to discern. So
evaluate the final velocity numerically (to three or four decimal places)
for v0/c=2/3v0/c=2/3. (Try to ignore the absurdity of two “blocks” sticking
together in such a violent collision!) Also evaluate the final mass in this case,
as a numerical multiple of the initial mass. Comment on your results.
(c) Now imagine viewing this collision from the reference frame in which the
final velocity of the combined block is zero. Sketch the initial and final
configurations in this frame. What is the initial velocity of Block 2 in this
frame? If momentum is to be conserved in this frame, what must be the initial
velocity of Block 1? Finally, use the Einstein velocity transformation formula
to transform Block 1’s initial velocity back to the original reference frame, to
check that everything is consistent. Explain the significance of this
consistency in your own words.
Further Reading
This concludes this abbreviated introduction to special relativity. If you would like to learn more
about special relativity, I highly recommend the following three texts. All are at about the same
level as these lessons but provide more detail, and all make frequent use of spacetime diagrams,
as I have:
 Special Relativity Primer, by William L. Burke and Peter Scott (61 pages). This is the short,
informal booklet from which I first learned special relativity, back in 1981. It was never
commercially published and for a long time was hard to find, but Scott has now reformatted it as
a pdf that you can download.
 Six Ideas That Shaped Physics, Unit R, by Thomas A. Moore (208 pages). Moore was my
instructor when I took introductory physics in 1981. His book draws upon some of the best
features of Burke and Scott (above) and Taylor and Wheeler (below), enhanced by his many
years of teaching experience and his uniquely careful writing style. Unfortunately the book is
somewhat expensive for its length, but perhaps you can find it in a library.
 Spacetime Physics, by Edwin F. Taylor and John A. Wheeler (312 pages). This famous book has
influenced whole generations of physicists since the first edition was published in 1963. It’s less
methodical and a little more advanced than Moore’s book, but great fun to browse. For reasons
having more to do with the corporate textbook publishing business than the popularity of the
book, this book is now out of print—but you can download it for free!

If you would like to see special relativity applied and extended in the context of
electromagnetism or gravity, here are some treatments of those subjects that I recommend:
 Magnetism, Radiation, and Relativity is a brief set of online materials I created long ago
showing how magnetism is a consequence of length contraction, and electromagnetic radiation
is a consequence of the cosmic speed limit. These materials are at the level of a calculus-based
introductory physics course.
 Exploring Black Holes, by Edwin F. Taylor, John A. Wheeler, and Edmund Bertschinger. This
unpublished book (an earlier edition was published but is now out of print) picks up
where Spacetime Physics leaves off, introducing many of the applications of general relativity at
a level that uses calculus but no more advanced mathematics. You can freely download the
book for personal use.
 A General Relativity Workbook, by Thomas A. Moore. This is a textbook for a complete
undergraduate course in general relativity, carefully introducing the mathematics needed to
deal with arbitrarily curved spacetime. The workbook format is great for self-study, and the
price is reasonable.

About this Document

Copyright ©2022-2024 by Daniel V. Schroeder. This work is licensed under
a Creative Commons Attribution 4.0 International License. (This copyright and
license do not apply to illustrations attributed to another source.)
This document is also available as a printer-friendly pdf (46 pages, 4.4 MB). The
pdf version is identical in content, aside from putting the external links (and the
embedded video) into footnotes with visible URLs.
Click here to download a zip archive (4.3 MB) containing the LaTeX source for
the pdf version along with the original vector-graphic illustrations. I recommend
starting from the vector-graphic version if you wish to adapt or translate an
illustration.
Version history:
 August 2022: First public version.
 July 2024: Added the firecracker experiment setup illustration, made a few
minor tweaks, and created the pdf printer-friendly version.

Colophon
I typed this HTML document directly into a text editor (BBEdit). The source file is
human-readable, so you can use your browser’s View Source or Page Source
command to see the (minimal) styling code. MathJax is typesetting the equations. I
created the illustrations in Adobe Illustrator, then exported them in png format at
double the (nominal) displayed resolution for sharpness on high-resolution displays.
To label the illustrations I used LaTeXiT (included in the MacTeX distribution),
exporting the output as pdf with outlined fonts. For the pdf version I manually ported
the document to LaTeX.