C h a pter II

SPE C IAL R E L ATI V I T Y

Dvipsbugw

There are limitations on motion that are missed by the Galilean description. The first
limitation we discover is the existence of a maximal speed in nature. The maximum speed
implies many fascinating results: it leads to observer-varying time and length intervals,
to an intimate relation between mass and energy, and to the existence of event horizons.
We explore them now.

Motion Mountain – The Adventure of Physics available free of charge at [Link]

5. maximum speed, observers at rest, and motion of
light

“ ”
Fama nihil est celerius.*

L ight is indispensable for a precise description of motion. To check whether a

ine or a path of motion is straight, we must look along it. In other words, we use
ight to define straightness. How do we decide whether a plane is flat? We look across
it,** again using light. How do we measure length to high precision? With light. How do
we measure time to high precision? With light: once it was light from the Sun that was
Page 1154 used; nowadays it is light from caesium atoms.
In other words, light is important because it is the standard for undisturbed motion.
Physics would have evolved much more rapidly if, at some earlier time, light propagation
had been recognized as the ideal example of motion.
But is light really a phenomenon of motion? This was already known in ancient Greece,
from a simple daily phenomenon, the shadow. Shadows prove that light is a moving en-
tity, emanating from the light source, and moving in straight lines.*** The obvious con-
clusion that light takes a certain amount of time to travel from the source to the surface
Copyright © Christoph Schiller November 1997–May 2006

* ‘Nothing is faster than rumour.’ This common sentence is a simplified version of Virgil’s phrase: fama,
malum qua non aliud velocius ullum. ‘Rumour, the evil faster than all.’ From the Aeneid, book IV, verses 173
and 174.
** Note that looking along the plane from all sides is not sufficient for this: a surface that a light beam touches
right along its length in all directions does not need to be flat. Can you give an example? One needs other
Challenge 540 n methods to check flatness with light. Can you specify one?
*** Whenever a source produces shadows, the emitted entities are called rays or radiation. Apart from light,
other examples of radiation discovered through shadows were infrared rays and ultraviolet rays, which em-
anate from most light sources together with visible light, and cathode rays, which were found to be to the
motion of a new particle, the electron. Shadows also led to the discovery of X-rays, which again turned out
to be a version of light, with high frequency. Channel rays were also discovered via their shadows; they turn
out to be travelling ionized atoms. The three types of radioactivity, namely α-rays (helium nuclei), β-rays

Dvipsbugw
 276 ii special rel ativity • 5. speed, rest and light

Jupiter and Io
(second measurement)

Earth (second
measurement)

Dvipsbugw
Sun Earth (first Jupiter and Io
measurement) (first measurement)

F I G U R E 132 Rømer’s method of measuring the speed of light

Motion Mountain – The Adventure of Physics available free of charge at [Link]

Ref. 231 showing the shadow had already been reached by the Greek thinker Empedocles (c. 490
to c. 430 bce ).
We can confirm this result with a different, equally simple, but subtle argument. Speed
can be measured. Therefore the perfect speed, which is used as the implicit measurement
standard, must have a finite value. An infinite velocity standard would not allow meas-
Challenge 541 n urements at all. In nature, the lightest entities move with the highest speed. Light, which
is indeed light, is an obvious candidate for motion with perfect but finite speed. We will
confirm this in a minute.
A finite speed of light means that whatever we see is a message from the past. When
we see the stars, the Sun or a loved one, we always see an image of the past. In a sense,
nature prevents us from enjoying the present – we must therefore learn to enjoy the past.
The speed of light is high; therefore it was not measured until 1676, even though many,
including Galileo, had tried to do so earlier. The first measurement method was worked
out by the Danish astronomer Ole Rømer* when he was studying the orbits of Io and
the other moons of Jupiter. He obtained an incorrect value for the speed of light because
he used the wrong value for their distance from Earth. However, this was quickly correc-
ted by his peers, including Newton himself. You might try to deduce his method from
Challenge 542 n Figure 132. Since that time it has been known that light takes a bit more than 8 minutes
to travel from the Sun to the Earth. This was confirmed in a beautiful way fifty years later,
Page 101 in 1726, by the astronomer James Bradley. Being English, Bradley thought of the ‘rain
Copyright © Christoph Schiller November 1997–May 2006

Ref. 232 method’ to measure the speed of light.

How can we measure the speed of falling rain? We walk rapidly with an umbrella,
measure the angle α at which the rain appears to fall, and then measure our own velocity

(again electrons), and γ-rays (high-energy X-rays) also produce shadows. All these discoveries were made
between 1890 and 1910: those were the ‘ray days’ of physics.
* Ole (Olaf) Rømer (1644 Aarhus – 1710 Copenhagen), Danish astronomer. He was the teacher of the
Dauphin in Paris, at the time of Louis XIV. The idea of measuring the speed of light in this way was due
to the Italian astronomer Givanni Cassini, whose assistant Rømer had been. Rømer continued his measure-
ments until 1681, when Rømer had to leave France, like all protestants (such as Christiaan Huygens), so that
his work was interrupted. Back in Denmark, a fire destroyed all his measurement notes. As a result, he was
not able to continue improving the precision of his method. Later he became an important administrator
and reformer of the Danish state.

Dvipsbugw
 maximum speed, observers at rest, and motion of light 277

rain's perspective light's perspective

rain light

c
c
earth v
v
Sun

Dvipsbugw

human perspective human perspective

Motion Mountain – The Adventure of Physics available free of charge at [Link]

α v
Sun
v

F I G U R E 133 The rain method of measuring the speed of light

v. As shown in Figure 133, the speed c of the rain is then given by

c = v
tan α . (100)

The same measurement can be made for light; we just need to measure the angle at which
the light from a star above Earth’s orbit arrives at the Earth. Because the Earth is moving
relative to the Sun and thus to the star, the angle is not a right one. This effect is called
the aberration of light; the angle is found most easily by comparing measurements made
six months apart. The value of the angle is 20.5

; nowadays it can be measured with
a precision of five decimal digits. Given that the speed of the Earth around the Sun is
v = 2πR
T = 29.7 km
s, the speed of light must therefore be c = 3.00 ċ 108 m
s.* This is

* Umbrellas were not common in Britain in 1726; they became fashionable later, after being introduced from
Copyright © Christoph Schiller November 1997–May 2006

China. The umbrella part of the story is made up. In reality, Bradley had his idea while sailing on the Thames,
when he noted that on a moving ship the apparent wind has a different direction from that on land. He had
observed 50 stars for many years, notably Gamma Draconis, and during that time he had been puzzled by
the sign of the aberration, which was opposite to the effect he was looking for, namely the star parallax. Both
the parallax and the aberration for a star above the ecliptic make them describe a small ellipse in the course
Challenge 543 n of an Earth year, though with different rotation senses. Can you see why?
By the way, it follows from special relativity that the formula (100) is wrong, and that the correct formula
Challenge 544 n is c = v
sin α; can you see why?
To determine the speed of the Earth, we first have to determine its distance from the Sun. The simplest
method is the one by the Greek thinker Aristarchos of Samos (c. 310 to c. 230 bce ). We measure the angle
between the Moon and the Sun at the moment when the Moon is precisely half full. The cosine of that angle
gives the ratio between the distance to the Moon (determined, for example, by the methods of page 117) and
Challenge 545 n the distance to the Sun. The explanation is left as a puzzle for the reader.

Dvipsbugw
 278 ii special rel ativity • 5. speed, rest and light

half-silvered
mirror

mirror light
source

Dvipsbugw

Motion Mountain – The Adventure of Physics available free of charge at [Link]

F I G U R E 134 Fizeau’s set-up to measure the speed of light (© AG Didaktik
und Geschichte der Physik, Universität Oldenburg)

an astonishing value, especially when compared with the highest speed ever achieved by
a man-made object, namely the Voyager satellites, which travel at 52 Mm
h = 14 km
s,
with the growth of children, about 3 nm
s, or with the growth of stalagmites in caves,
about 0.3 pm
s. We begin to realize why measurement of the speed of light is a science
in its own right.
The first precise measurement of the speed of light was made in 1849 by the French
physicist Hippolyte Fizeau (1819–1896). His value was only 5 % greater than the modern
one. He sent a beam of light towards a distant mirror and measured the time the light
took to come back. How did Fizeau measure the time without any electric device? In fact,
Page 57 he used the same ideas that are used to measure bullet speeds; part of the answer is given
Challenge 547 n in Figure 134. (How far away does the mirror have to be?) A modern reconstruction of
Ref. 235 his experiment by Jan Frercks has achieved a precision of 2 %. Today, the experiment is
much simpler; in the chapter on electrodynamics we will discover how to measure the
Page 560 speed of light using two standard UNIX or Linux computers connected by a cable.
The speed of light is so high that it is even difficult to prove that it is finite. Perhaps the
most beautiful way to prove this is to photograph a light pulse flying across one’s field of
view, in the same way as one can photograph a car driving by or a bullet flying through
Copyright © Christoph Schiller November 1997–May 2006

The angle in question is almost a right angle (which would yield an infinite distance), and good instru-
Ref. 233 ments are needed to measure it with precision, as Hipparchos noted in an extensive discussion of the prob-
lem around 130 bce. Precise measurement of the angle became possible only in the late seventeenth century,
when it was found to be 89.86°, giving a distance ratio of about 400. Today, thanks to radar measurements
Page 1167 of planets, the distance to the Sun is known with the incredible precision of 30 metres. Moon distance vari-
Challenge 546 n ations can even be measured to the nearest centimetre; can you guess how this is achieved?
Ref. 234 Aristarchos also determined the radius of the Sun and of the Moon as multiples of those of the Earth.
Aristarchos was a remarkable thinker: he was the first to propose the heliocentric system, and perhaps the
first to propose that stars were other, faraway suns. For these ideas, several of his contemporaries proposed
that he should be condemned to death for impiety. When the Polish monk and astronomer Nicolaus Co-
pernicus (1473–1543) again proposed the heliocentric system two thousand years later, he did not mention
Aristarchus, even though he got the idea from him.

Dvipsbugw
 maximum speed, observers at rest, and motion of light 279

red
shutter
switch
beam

path of light pulse

Dvipsbugw
10 mm

F I G U R E 135 A photograph of a light pulse moving from right to left through a bottle
with milky water, marked in millimetres (© Tom Mattick)

Motion Mountain – The Adventure of Physics available free of charge at [Link]

F I G U R E 136 A consequence of the finiteness of the
speed of light

Ref. 236 the air. Figure 135 shows the first such photograph, produced in 1971 with a standard
off-the-shelf reflex camera, a very fast shutter invented by the photographers, and, most
noteworthy, not a single piece of electronic equipment. (How fast does such a shutter have
Challenge 548 n to be? How would you build such a shutter? And how would you make sure it opened at
the right instant?)
A finite speed of light also implies that a rapidly rotating light beam behaves as shown
Copyright © Christoph Schiller November 1997–May 2006

as in Figure 136. In everyday life, the high speed of light and the slow rotation of light-
houses make the effect barely noticeable.
In short, light moves extremely rapidly. It is much faster than lightning, as you might
Challenge 549 n like to check yourself. A century of increasingly precise measurements of the speed have
culminated in the modern value

c = 299 792 458 m
s. (101)

In fact, this value has now been fixed exactly, by definition, and the metre has been defined
in terms of c. Table 35 gives a summary of what is known today about the motion of light.
Two surprising properties were discovered in the late nineteenth century. They form the

Dvipsbugw
 280 ii special rel ativity • 5. speed, rest and light

TA B L E 35 Properties of the motion of light

O b s e r va t i o n s a b o u t l i g h t

Light can move through vacuum.

Light transports energy.
Light has momentum: it can hit bodies.
Light has angular momentum: it can rotate bodies.
Light moves across other light undisturbed. Dvipsbugw
Light in vacuum always moves faster than any material body does.
The speed of light, its true signal speed, is the forerunner speed. Page 579
In vacuum its value is 299 792 458 m
s.
The proper speed of light is infinite. Page 297
Shadows can move without any speed limit.
Light moves in a straight line when far from matter.

Motion Mountain – The Adventure of Physics available free of charge at [Link]

High-intensity light is a wave.
Light beams are approximations when the wavelength is neglected.
In matter, both the forerunner speed and the energy speed of light are lower than in vacuum.
In matter, the group velocity of light pulses can be zero, positive, negative or infinite.

Ref. 237 basis of special relativity.

Can one play tennis using a laser pulse as the ball and mirrors
as rackets?

“
Et nihil est celerius annis.*

”
Ovid, Metamorphoses.

We all know that in order to throw a stone as far as possible, we run as we throw it; we
know instinctively that in that case the stone’s speed with respect to the ground is higher.
However, to the initial astonishment of everybody, experiments show that light emitted
from a moving lamp has the same speed as light emitted from a resting one. Light (in
vacuum) is never faster than light; all light beams have the same speed. Many specially
Ref. 238 designed experiments have confirmed this result to high precision. The speed of light can
Copyright © Christoph Schiller November 1997–May 2006

be measured with a precision of better than 1 m
s; but even for lamp speeds of more than
Challenge 550 n 290 000 000 m
s no differences have been found. (Can you guess what lamps were used?)
In everyday life, we know that a stone arrives more rapidly if we run towards it. Again,
for light no difference has been measured. All experiments show that the velocity of light
has the same value for all observers, even if they are moving with respect to each other
or with respect to the light source. The speed of light is indeed the ideal, perfect measure-
ment standard.**

* ‘Nothing is faster than the years.’ Book X, verse 520.

** An equivalent alternative term for the speed of light is ‘radar speed’ or ‘radio speed’; we will see below
Page 563 why this is the case.
The speed of light is also not far from the speed of neutrinos. This was shown most spectacularly by the

Dvipsbugw
 maximum speed, observers at rest, and motion of light 281

Ref. 241 There is also a second set of experimental evidence for the
constancy of the speed of light. Every electromagnetic device,
such as an electric toothbrush, shows that the speed of light is
Page 536 constant. We will discover that magnetic fields would not res-
ult from electric currents, as they do every day in every motor
and in every loudspeaker, if the speed of light were not constant.
This was actually how the constancy was first deduced, by sev-
eral researchers. Only after understanding this, did the German–
Swiss physicist Albert Einstein* show that the constancy is also Dvipsbugw
in agreement with the motion of bodies, as we will do in this
section. The connection between electric toothbrushes and re-
Page 536 lativity will be described in the chapter on electrodynamics.** In Albert Einstein
simple terms, if the speed of light were not constant, observers
would be able to move at the speed of light. Since light is a wave, such observers would
see a wave standing still. However, electromagnetism forbids the such a phenomenon.

Motion Mountain – The Adventure of Physics available free of charge at [Link]

Therefore, observers cannot reach the speed of light.
In summary, the velocity v of any physical system in nature (i.e., any localized mass or
energy) is bound by
vc. (102)

This relation is the basis of special relativity; in fact, the full theory of special relativity
is contained in it. Einstein often regretted that the theory was called ‘Relativitätstheorie’
or ‘theory of relativity’; he preferred the name ‘Invarianztheorie’ or ‘theory of invariance’,
Ref. 244 but was not able to change the name.
observation of a supernova in 1987, when the flash and the neutrino pulse arrived a 12 seconds apart. (It
is not known whether the difference is due to speed differences or to a different starting point of the two
flashes.) What is the first digit for which the two speed values could differ, knowing that the supernova was
Challenge 551 n 1.7 ċ 105 light years away?
Experiments also show that the speed of light is the same in all directions of space, to at least 21 digits of
Ref. 239 precision. Other data, taken from gamma ray bursts, show that the speed of light is independent of frequency,
Ref. 240 to at least 20 digits of precision.
* Albert Einstein (b. 1879 Ulm, d. 1955 Princeton); one of the greatest physicists ever. He published three
important papers in 1905, one about Brownian motion, one about special relativity, and one about the idea
of light quanta. Each paper was worth a Nobel Prize, but he was awarded the prize only for the last one. Also
in 1905, he proved the famous formula E 0 = mc 2 (published in early 1906), possibly triggered by an idea of
Page 318 Olinto de Pretto. Although Einstein was one of the founders of quantum theory, he later turned against it.
Copyright © Christoph Schiller November 1997–May 2006

His famous discussions with his friend Niels Bohr nevertheless helped to clarify the field in its most counter-
intuitive aspects. He explained the Einstein–de Haas effect which proves that magnetism is due to motion
inside materials. In 1915 and 1916, he published his highest achievement: the general theory of relativity, one
of the most beautiful and remarkable works of science.
Being Jewish and famous, Einstein was a favourite target of attacks and discrimination by the National
Socialist movement; in 1933 he emigrated to the USA. He was not only a great physicist, but also a great
Ref. 242 thinker; his collection of thoughts about topics outside physics are worth reading.
Anyone interested in emulating Einstein should know that he published many papers, and that many
of them were wrong; he would then correct the results in subsequent papers, and then do so again. This
happened so frequently that he made fun of himself about it. Einstein realizes the famous definition of a
genius as a person who makes the largest possible number of mistakes in the shortest possible time.
** For information about the influences of relativity on machine design, see the interesting textbook by Van
Ref. 243 Bladel.

Dvipsbugw
 282 ii special rel ativity • 5. speed, rest and light

The constancy of the speed of light is in complete contrast with Galilean mechanics,
and proves that the latter is wrong at high velocities. At low velocities the description
remains good, because the error is small. But if we want a description valid at all velocities,
we have to discard Galilean mechanics. For example, when we play tennis we use the fact
that by hitting the ball in the right way, we can increase or decrease its speed. But with
light this is impossible. Even if we take an aeroplane and fly after a light beam, it still
moves away with the same speed. Light does not behave like cars. If we accelerate a bus
we are driving, the cars on the other side of the road pass by with higher and higher
speeds. For light, this is not so: light always passes by with the same speed.* Dvipsbugw
Why is this result almost unbelievable, even though the measurements show it un-
ambiguously? Take two observers O and Ω (pronounced ‘omega’) moving with relative
velocity v, such as two cars on opposite sides of the street. Imagine that at the moment
they pass each other, a light flash is emitted by a lamp in O. The light flash moves through
positions x(t) for O and through positions ξ(τ) (pronounced ‘xi of tau’) for Ω. Since the
speed of light is the same for both, we have

Motion Mountain – The Adventure of Physics available free of charge at [Link]

x ξ
=c= . (103)
t τ

However, in the situation described, we obviously have x  ξ. In other words, the con-
stancy of the speed of light implies that t  τ, i.e. that time is different for observers moving
Challenge 552 e relative to each other. Time is thus not unique. This surprising result, which has been con-
Ref. 245 firmed by many experiments, was first stated clearly in 1905 by Albert Einstein. Though
many others knew about the invariance of c, only the young Einstein had the courage to
say that time is observer-dependent, and to face the consequences. Let us do so as well.
Already in 1895, the discussion of viewpoint invariance had been called the theory of
relativity by Henri Poincaré.** Einstein called the description of motion without gravity
Ref. 241 the theory of special relativity, and the description of motion with gravity the theory of
general relativity. Both fields are full of fascinating and counter-intuitive results. In par-
ticular, they show that everyday Galilean physics is wrong at high speeds.
The speed of light is a limit speed. We stress that we are not talking of the situation
where a particle moves faster than the speed of light in matter, but still slower than the
speed of light in vacuum. Moving faster than the speed of light in matter is possible. If the
particle is charged, this situation gives rise to the so-called Čerenkov radiation. It corres-
ponds to the V-shaped wave created by a motor boat on the sea or the cone-shaped shock
Copyright © Christoph Schiller November 1997–May 2006

wave around an aeroplane moving faster than the speed of sound. Čerenkov radiation is
regularly observed; for example it is the cause of the blue glow of the water in nuclear re-
actors. Incidentally, the speed of light in matter can be quite low: in the centre of the Sun,

* Indeed, even with the current measurement precision of 2 ċ 10−13 , we cannot discern any changes of the
Ref. 239 speed of light with the speed of the observer.
** Henri Poincaré (1854–1912), important French mathematician and physicist. Poincaré was one of the
most productive men of his time, advancing relativity, quantum theory, and many parts of mathematics.
The most beautiful and simple introduction to relativity is still that given by Albert Einstein himself, for
example in Über die spezielle und allgemeine Relativitätstheorie, Vieweg, 1997, or in The Meaning of Relativity,
Methuen, London, 1951. It has taken a century for books almost as beautiful to appear, such as the text by
Ref. 246, Ref. 247 Taylor and Wheeler.

Dvipsbugw
 maximum speed, observers at rest, and motion of light 283

the speed of light is estimated to be only around 10 km
year, and even in the laboratory,
Ref. 248, Ref. 249 for some materials, it has been found to be as low as 0.3 m
s. In the following, when we
use the term ‘speed of light’, we mean the speed of light in vacuum. The speed of light in
air in smaller than that in vacuum only by a fraction of a percent, so that in most cases,
the difference can be neglected.

Special relativity in a few lines

The speed of light is constant for all observers. We Dvipsbugw
can thus deduce all relations between what two dif- first
t
Ref. 250 ferent observers measure with the help of Figure 137. observer second
or clock observer
It shows two observers moving with constant speed
or clock
against each other in space-time, with the first send-
ing a light flash to the second, from where it is reflec- k2 T
ted back to the first. Since light speed is constant, light light

Motion Mountain – The Adventure of Physics available free of charge at [Link]

is the only way to compare time and space coordin-
ates for two distant observers. Two distant clocks (like t1 = (k2+1)T/2 t2 = kT
two distant metre bars) can only be compared, or
synchronized, using light or radio flashes. Since light
speed is constant, light paths are parallel in such dia- T
grams.
A constant relative speed between two observers O
implies that a constant factor k relates the time co- x
Challenge 553 n ordinates of events. (Why is the relation linear?) If a
flash starts at a time T as measured for the first ob- F I G U R E 137 A drawing containing
server, it arrives at the second at time kT, and then most of special relativity
back again at the first at time k 2 T. The drawing shows
Challenge 554 n that 
c+v v k2 − 1
k= or = . (104)
c−v c k2 + 1

Page 284 This factor will appear again in the Doppler effect.*
The figure also shows that the time coordinate t 1 assigned by the first observer to the
moment in which the light is reflected is different from the coordinate t 2 assigned by the
second observer. Time is indeed different for two observers in relative motion. Figure 138
Copyright © Christoph Schiller November 1997–May 2006

illustrates the result.

The time dilation factor between the two time coordinates is found from Figure 137 by
comparing the values t 1 and t 2 ; it is given by

t1 1
= = γ(v) . (105)
t2 1− v2
c2

Time intervals for a moving observer are shorter by this factor γ; the time dilation factor
is always larger than 1. In other words, moving clocks go slower. For everyday speeds the

* The explanation of relativity using the factor k is often called k-calculus.

Dvipsbugw
 284 ii special rel ativity • 5. speed, rest and light

one moving watch

first second
time time

two fixed watches

F I G U R E 138 Moving clocks go slow

Dvipsbugw

Challenge 555 e effect is tiny. That is why we do not detect time differences in everyday life. Nevertheless,
Galilean physics is not correct for speeds near that of light. The same factor γ also appears
in the formula E = γmc 2 , which we will deduce below. Expression (104) or (105) is the
only piece of mathematics needed in special relativity: all other results derive from it.

Motion Mountain – The Adventure of Physics available free of charge at [Link]

If a light flash is sent forward starting from the second observer and reflected back,
he will make the same statement: for him, the first clock is moving, and also for him, the
moving clock goes slower. Each of the observers observes that the other clock goes slower.
The situation is similar to that of two men comparing the number of steps between two
identical ladders that are not parallel. A man on either ladder will always observe that the
steps of the other ladder are shorter. For another analogy, take two people moving away
from each other: each of them notes that the other gets smaller as their distance increases.
Naturally, many people have tried to find arguments to avoid the strange conclusion
that time differs from observer to observer. But none have succeeded, and experimental
results confirm this conclusion. Let us have a look at some of them.

Acceleration of light and the Doppler effect

Light can be accelerated. Every mirror does this! We will see in the chapter on electromag-
netism that matter also has the power to bend light, and thus to accelerate it. However, it
Page 567 will turn out that all these methods only change the direction of propagation; none has the
power to change the speed of light in a vacuum. In short, light is an example of a motion
Challenge 556 n that cannot be stopped. There are only a few other such examples. Can you name one?
What would happen if we could accelerate light to higher speeds? For this to be pos-
sible, light would have to be made of particles with non-vanishing mass. Physicists call
such particles massive particles. If light had mass, it would be necessary to distinguish
Copyright © Christoph Schiller November 1997–May 2006

the ‘massless energy speed’ c from the speed of light c L , which would be lower and would
depend on the kinetic energy of those massive particles. The speed of light would not be
constant, but the massless energy speed would still be so. Massive light particles could
be captured, stopped and stored in a box. Such boxes would make electric illumination
unnecessary; it would be sufficient to store some daylight in them and release the light,
slowly, during the following night, maybe after giving it a push to speed it up.*
Physicists have tested the possibility of massive light in quite some detail. Observations
Ref. 251, Ref. 252 now put any possible mass of light (particles) at less than 1.3 ċ 10−52 kg from terrestrial

* Incidentally, massive light would also have longitudinal polarization modes. This is in contrast to observa-
tions, which show that light is polarized exclusively transversally to the propagation direction.

Dvipsbugw
 maximum speed, observers at rest, and motion of light 285

sender receiver

Dvipsbugw

Motion Mountain – The Adventure of Physics available free of charge at [Link]

receiver θr
light x
signal

z
sender θs

v x

F I G U R E 139 The set-up for the observation of the Doppler effect

experiments, and at less than 4 ċ 10−62 kg from astrophysical arguments (which are a bit
less strict). In other words, light is not heavy, light is light.
But what happens when light hits a moving mirror? If the speed of light does not
change, something else must. The situation is akin to that of a light source moving with re-
spect to the receiver: the receiver will observe a different colour from that observed by the
sender. This is called the Doppler effect. Christian Doppler* was the first to study the fre-
quency shift in the case of sound waves – the well-known change in whistle tone between
Copyright © Christoph Schiller November 1997–May 2006

approaching and departing trains – and to extend the concept to the case of light waves.
As we will see later on, light is (also) a wave, and its colour is determined by its frequency,
or equivalently, by its wavelength λ. Like the tone change for moving trains, Doppler real-
ized that a moving light source produces a colour at the receiver that is different from the
colour at the source. Simple geometry, and the conservation of the number of maxima
Challenge 557 e and minima, leads to the result

* Christian Andreas Doppler (b. 1803 Salzburg, d. 1853 Venezia), Austrian physicist. Doppler studied the
effect named after him for sound and light. In 1842 he predicted (correctly) that one day we would be able
to use the effect to measure the motion of distant stars by looking at their colours.

Dvipsbugw
 286 ii special rel ativity • 5. speed, rest and light

λr 1 v v
= (1 − cos θ r ) = γ (1 − cos θ r ) . (106)
λs 1 − v 2
c 2 c c

The variables v and θ r in this expression are defined in Figure 139. Light from an approach-
ing source is thus blue-shifted, whereas light from a departing source is red-shifted. The
first observation of the Doppler effect for light was made by Johannes Stark* in 1905,
who studied the light emitted by moving atoms. All subsequent experiments confirmed
the calculated colour shift within measurement errors; the latest checks have found agree-
Dvipsbugw
Ref. 253 ment to within two parts per million. In contrast to sound waves, a colour change is also
found when the motion is transverse to the light signal. Thus, a yellow rod in rapid mo-
tion across the field of view will have a blue leading edge and a red trailing edge prior to
the closest approach to the observer. The colours result from a combination of the longit-
udinal (first-order) Doppler shift and the transverse (second-order) Doppler shift. At a
particular angle θ unshifted the colours will be the same. (How does the wavelength change
Challenge 558 n in the purely transverse case? What is the expression for θ unshifted in terms of v?)

Motion Mountain – The Adventure of Physics available free of charge at [Link]

The colour shift is used in many applications. Almost all solid bodies are mirrors for
radio waves. Many buildings have doors that open automatically when one approaches.
A little sensor above the door detects the approaching person. It usually does this by
measuring the Doppler effect of radio waves emitted by the sensor and reflected by the
Page 563 approaching person. (We will see later that radio waves and light are manifestations of
the same phenomenon.) So the doors open whenever something moves towards them.
Police radar also uses the Doppler effect, this time to measure the speed of cars.**
The Doppler effect also makes it possible to measure the velocity of light sources.
Indeed, it is commonly used to measure the speed of distant stars. In these cases, the
Doppler shift is often characterized by the red-shift number z, defined with the help of
wavelength λ or frequency F by

∆λ f S c+v
z= = −1= −1 . (107)
λ fR c−v

Challenge 560 n Can you imagine how the number z is determined? Typical values for z for light sources
in the sky range from −0.1 to 3.5, but higher values, up to more than 10, have also been
Challenge 561 n found. Can you determine the corresponding speeds? How can they be so high?
In summary, whenever one tries to change the speed of light, one only manages to
Copyright © Christoph Schiller November 1997–May 2006

change its colour. That is the Doppler effect.

Page 133 We know from classical physics that when light passes a large mass, such as a star, it is
Challenge 562 n deflected. Does this deflection lead to a Doppler shift?

* Johannes Stark (1874–1957), discovered in 1905 the optical Doppler effect in channel rays, and in 1913
the splitting of spectral lines in electrical fields, nowadays called the Stark effect. For these two discoveries
he received the 1919 Nobel Prize for physics. He left his professorship in 1922 and later turned into a full-
blown National Socialist. A member of the NSDAP from 1930 onwards, he became known for aggressively
criticizing other people’s statements about nature purely for ideological reasons; he became rightly despised
by the academic community all over the world.
Challenge 559 n ** At what speed does a red traffic light appear green?

Dvipsbugw
 maximum speed, observers at rest, and motion of light 287

The difference between light and sound

The Doppler effect for light is much more important than the Doppler effect for sound.
Even if the speed of light were not yet known to be constant, this effect alone would prove
that time is different for observers moving relative to each other. Why? Time is what we
read from our watch. In order to determine whether another watch is synchronized with
our own one, we look at both watches. In short, we need to use light signals to synchron-
Ref. 254 ize clocks. Now, any change in the colour of light moving from one observer to another
necessarily implies that their watches run differently, and thus that time is different for Dvipsbugw
the two of them. One way to see this is to note that also a light source is a clock – ‘ticking’
very rapidly. So if two observers see different colours from the same source, they meas-
ure different numbers of oscillations for the same clock. In other words, time is different
for observers moving against each other. Indeed, equation (104) implies that the whole
of relativity follows from the full Doppler effect for light. (Can you confirm that the con-
nection between observer-dependent frequencies and observer-dependent time breaks

Motion Mountain – The Adventure of Physics available free of charge at [Link]

Challenge 563 n down in the case of the Doppler effect for sound?)
Why does the behaviour of light imply special relativity, while that of sound in air does
not? The answer is that light is a limit for the motion of energy. Experience shows that
there are supersonic aeroplanes, but there are no superluminal rockets. In other words,
the limit v  c is valid only if c is the speed of light, not if c is the speed of sound in air.
However, there is at least one system in nature where the speed of sound is indeed a
limit speed for energy: the speed of sound is the limit speed for the motion of dislocations
Page 987 in crystalline solids. (We discuss this in detail later on.) As a result, the theory of special
relativity is also valid for such dislocations, provided that the speed of light is replaced
everywhere by the speed of sound! Dislocations obey the Lorentz transformations, show
Ref. 255 length contraction, and obey the famous energy formula E = γmc 2 . In all these effects
the speed of sound c plays the same role for dislocations as the speed of light plays for
general physical systems.
If special relativity is based on the statement that nothing can move faster than light,
this statement needs to be carefully checked.

Can one shoot faster than one’s shadow?

“ ”
Quid celerius umbra?*
Copyright © Christoph Schiller November 1997–May 2006

for Lucky Luke to achieve the feat shown in Figure 140, his bullet has to move faster than
Challenge 564 e the speed of light. (What about his hand?) In order to emulate Lucky Luke, we could take
the largest practical amount of energy available, taking it directly from an electrical power
station, and accelerate the lightest ‘bullets’ that can be handled, namely electrons. This ex-
periment is carried out daily in particle accelerators such as the Large Electron Positron
ring, the LEP, of 27 km circumference, located partly in France and partly in Switzerland,
near Geneva. There, 40 MW of electrical power (the same amount used by a small city)
accelerates electrons and positrons to energies of over 16 nJ (104.5 GeV) each, and their
Ref. 256 speed is measured. The result is shown in Figure 141: even with these impressive means

* ‘What is faster than the shadow?’ A motto often found on sundials.

Dvipsbugw
 288 ii special rel ativity • 5. speed, rest and light

Dvipsbugw

F I G U R E 140 Lucky Luke

Motion Mountain – The Adventure of Physics available free of charge at [Link]

it is impossible to make electrons move more rapidly than light. (Can you imagine a way
Challenge 565 e to measure energy and speed separately?) The speed–energy relation of Figure 141 is a
Page 313 consequence of the maximum speed, and is deduced below. These and many similar ob-
servations thus show that there is a limit to the velocity of objects. Bodies (and radiation)
cannot move at velocities higher than the speed of light.* The accuracy of Galilean mech-
anics was taken for granted for more than three centuries, so that nobody ever thought
of checking it; but when this was finally done, as in Figure 141, it was found to be wrong.
The people most unhappy with this
limit are computer engineers: if the speed 1
2 TGal = m v2
limit were higher, it would be possible v 2
to make faster microprocessors and thus 2
faster computers; this would allow, for c 1
example, more rapid progress towards T = m c2 ( – 1)
1 - v2/c 2
the construction of computers that un-
derstand and use language.
The existence of a limit speed runs T
counter to Galilean mechanics. In fact, F I G U R E 141 Experimental values (dots) for the
it means that for velocities near that of electron velocity v as function of their kinetic
light, say about 15 000 km
s or more, the energy T, compared with the prediction of Galilean
expression mv 2
2 is not equal to the kin- physics (blue) and that of special relativity (red)
Copyright © Christoph Schiller November 1997–May 2006

etic energy T of the particle. In fact, such

high speeds are rather common: many families have an example in their home. Just calcu-
late the speed of electrons inside a television, given that the transformer inside produces
Challenge 566 n 30 kV.

* There are still people who refuse to accept these results, as well as the ensuing theory of relativity. Every
physicist should enjoy the experience, at least once in his life, of conversing with one of these men. (Strangely,
no woman has yet been reported as belonging to this group of people.) This can be done, for example, via
Ref. 257 the internet, in the [Link] newsgroup. See also the [Link] website. Crackpots
are a fascinating lot, especially since they teach the importance of precision in language and in reasoning,
which they all, without exception, neglect. Encounters with several of them provided the inspiration for this
chapter.

Dvipsbugw
 maximum speed, observers at rest, and motion of light 289

t
second
first
observer third
observer
(e.g. train) observer
(e.g. Earth)
(e.g. stone)

kseT

kteT
Dvipsbugw
T

O
x

Motion Mountain – The Adventure of Physics available free of charge at [Link]

F I G U R E 142 How to deduce the
composition of velocities

The observation of speed of light as a limit speed for objects is easily seen to be a
consequence of its constancy. Bodies that can be at rest in one frame of reference obviously
move more slowly than the maximum velocity (light) in that frame. Now, if something
moves more slowly than something else for one observer, it does so for all other observers
Challenge 567 d as well. (Trying to imagine a world in which this would not be so is interesting: funny
things would happen, such as things interpenetrating each other.) Since the speed of light
is the same for all observers, no object can move faster than light, for every observer.
We follow that the maximum speed is the speed of massless entities. Electromagnetic
waves, including light, are the only known entities that can travel at the maximum speed.
Gravitational waves are also predicted to achieve maximum speed. Though the speed
of neutrinos cannot be distinguished experimentally from the maximum speed, recent
Ref. 258 experiments suggest that they do have a tiny mass.
Conversely, if a phenomenon exists whose speed is the limit speed for one observer,
Challenge 568 e then this limit speed must necessarily be the same for all observers. Is the connection
Challenge 569 r between limit property and observer invariance generally valid in nature? Copyright © Christoph Schiller November 1997–May 2006

The composition of velocities

If the speed of light is a limit, no attempt to exceed it can succeed. This implies that when
velocities are composed, as when one throws a stone while running, the values cannot
simply be added. If a train is travelling at velocity v te relative to the Earth, and somebody
throws a stone inside it with velocity v st relative to the train in the same direction, it is
usually assumed as evident that the velocity of the stone relative to the Earth is given by
v se = v st + v te . In fact, both reasoning and measurement show a different result.
The existence of a maximum speed, together with Figure 142, implies that the k-factors
must satisfy k se = k st k te .* Then we only have to insert the relation (104) between each

* By taking the (natural) logarithm of this equation, one can define a quantity, the rapidity, that measures

Dvipsbugw
 290 ii special rel ativity • 5. speed, rest and light

Challenge 570 e k-factor and the respective speed to get

v st + v te
v se = . (108)
1 + v st v te
c 2

Challenge 571 e This is called the velocity composition formula. The result is never larger than c and is
always smaller than the naive sum of the velocities.* Expression (108) has been confirmed
Page 313, page 536 by all of the millions of cases for which it has been checked. You may check that it reduces
Ref. 252 to the naive sum for everyday life values. Dvipsbugw

Observers and the principle of special relativity

Special relativity is built on a simple principle:
 The maximum speed of energy transport is the same for all observers.
Ref. 260 Or, as Hendrik Lorentz** liked to say, the equivalent:

Motion Mountain – The Adventure of Physics available free of charge at [Link]

 The speed v of a physical system is bound by

vc (109)

for all observers, where c is the speed of light.

This independence of the speed of light from the observer was checked with high preci-
Ref. 261 sion by Michelson and Morley*** in the years from 1887 onwards. It has been confirmed
in all subsequent experiments; the most precise to date, which achieved a precision of
Ref. 262 10−14 is shown in Figure 143.
In fact, special relativity is also confirmed by all the precision experiments that were
performed before it was formulated. You can even confirm it yourself at home. The way
Page 536 to do this is shown in the section on electrodynamics.
The existence of a limit speed has several interesting consequences. To explore them,
let us keep the rest of Galilean physics intact.**** The limit speed is the speed of light. It
is constant for all observers. This constancy implies:
the speed and is additive.
Ref. 259 * One can also deduce the Lorentz transformation directly from this expression.
** Hendrik Antoon Lorentz (b. 1853 Arnhem, d. 1928 Haarlem) was, together with Boltzmann and Kelvin,
one of the most important physicists of his time. He deduced the so-called Lorentz transformation and the
Lorentz contraction from Maxwell’s equation for the electrodynamic field. He was the first to understand,
Copyright © Christoph Schiller November 1997–May 2006

long before quantum theory confirmed the idea, that Maxwell’s equations for the vacuum also describe
matter and all its properties, as long as moving charged point particles – the electrons – are included. He
showed this in particular for the dispersion of light, for the Zeeman effect, for the Hall effect and for the
Faraday effect. He gave the correct description of the Lorentz force. In 1902, he received the physics Nobel
Prize, together with Pieter Zeeman. Outside physics, he was active in the internationalization of scientific
collaborations. He was also instrumental in the creation of the largest human-made structures on Earth: the
polders of the Zuyder Zee.
*** Albert Abraham Michelson (b. 1852 Strelno, d. 1931 Pasadena), Prussian–Polish–US-American physicist,
awarded the Nobel Prize in physics in 1907. Michelson called the set-up he devised an interferometer, a term
still in use today. Edward William Morley (1838–1923), US-American chemist, was Michelson’s friend and
long-time collaborator.
Page 86 **** This point is essential. For example, Galilean physics states that only relative motion is physical. Galilean
physics also excludes various mathematically possible ways to realize a constant light speed that would con-

Dvipsbugw
 maximum speed, observers at rest, and motion of light 291

AOM Power
30 driver servo

beat frequency change [Hz ]

Fiber
20 Frequency
FC counter
PZT AOM

angle/3 [deg]
Laser 1 PD
10 Nd: YAG FC
T °C
FC
0 T °C Laser 2 Res B PD
Nd: YAG Fiber BS
PZT AOM DBM
10 FC
∑ PD
PD
Frequency Local
20 servo oscillator PD Res A
0 100 200 300 400 500 600 Frequency Local
DBM
servo oscillator
time since begin of rotation [s]
∑
AOM Power
driver
servo Dvipsbugw

Motion Mountain – The Adventure of Physics available free of charge at [Link]

F I G U R E 143 The result, the schematics and the cryostat set-up for the most precise
Michelson–Morley experiment performed to date (© Stephan Schiller)

— In a closed free-floating room, there is no way to tell the speed of the room.
— There is no notion of absolute rest (or space): rest (like space) is an observer-dependent
concept.*
— Time depends on the observer; time is not absolute.
More interesting and specific conclusions can be drawn when two additional conditions
are assumed. First, we study situations where gravitation can be neglected. (If this not the
case, we need general relativity to describe the system.) Secondly, we also assume that the
data about the bodies under study – their speed, their position, etc. – can be gathered
Copyright © Christoph Schiller November 1997–May 2006

without disturbing them. (If this not the case, we need quantum theory to describe the
system.)
To deduce the precise way in which the different time intervals and lengths measured
by two observers are related to each other, we take an additional simplifying step. We start

tradict everyday life.

Einstein’s original 1905 paper starts from two principles: the constancy of the speed of light and the equi-
valence of all inertial observers. The latter principle had already been stated in 1632 by Galileo; only the
constancy of the speed of light was new. Despite this fact, the new theory was named – by Poincaré – after
Ref. 244 the old principle, instead of calling in ‘invariance theory’, as Einstein would have preferred.
Challenge 572 n * Can you give the argument leading to this deduction?

Dvipsbugw
 292 ii special rel ativity • 5. speed, rest and light

observer (greek)
v

light
c

observer (roman)

F I G U R E 144 Two inertial observers and a

beam of light

Galilean physics special relativity Dvipsbugw

t τ t τ

L L

Motion Mountain – The Adventure of Physics available free of charge at [Link]

O, Ω x, ξ O, Ω x

F I G U R E 145 Space-time diagrams for light seen from two

different observers using coordinates (t, x) and (τ, ξ)

with a situation where no interaction plays a role. In other words, we start with relativistic
kinematics of bodies moving without disturbance.
If an undisturbed body is observed to travel along a straight line with a constant ve-
locity (or to stay at rest), one calls the observer inertial, and the coordinates used by the
observer an inertial frame of reference. Every inertial observer is itself in undisturbed
motion. Examples of inertial observers (or frames) thus include – in two dimensions –
those moving on a frictionless ice surface or on the floor inside a smoothly running train
or ship; for a full example – in all three spatial dimensions – we can take a cosmonaut
travelling in a space-ship as long as the engine is switched off. Inertial observers in three
dimensions might also be called free-floating observers. They are thus not so common.
Challenge 573 e Non-inertial observers are much more common. Can you confirm this? Inertial observ-
ers are the most simple ones, and they form a special set:
— Any two inertial observers move with constant velocity relative to each other (as longs
as gravity plays no role, as assumed above).
Copyright © Christoph Schiller November 1997–May 2006

— All inertial observers are equivalent: they describe the world with the same equations.
Because it implies the lack of absolute space and time, this statement was called the
principle of relativity by Henri Poincaré. However, the essence of relativity is the exist-
ence of a limit speed.
To see how measured length and space intervals change from one observer to the other,
we assume two inertial observers, a Roman one using coordinates x, y, z and t, and a
Greek one using coordinates ξ, υ, ζ and τ,* that move with velocity v relative to each
other. The axes are chosen in such a way that the velocity points in the x-direction. The

* They are read as ‘xi’, ‘upsilon’, ‘zeta’ and ‘tau’. The names, correspondences and pronunciations of all Greek
letters are explained in Appendix A.

Dvipsbugw
 maximum speed, observers at rest, and motion of light 293

constancy of the speed of light in any direction for any two observers means that for the
motion of light the coordinate differentials are related by

0 = (cdt)2 − (dx)2 − (dy)2 − (dz)2 = (cdτ)2 − (dξ)2 − (dυ)2 − (dζ)2 . (110)

Assume also that a flash lamp at rest for the Greek observer, thus with dξ = 0, produces
two flashes separated by a time interval dτ. For the Roman observer, the flash lamp moves
with speed v, so that dx = vdt. Inserting this into the previous expression, and assuming
linearity and speed direction independence for the general case, we find that intervals are Dvipsbugw
Challenge 574 e related by

dτ + vdξ
c 2
dt = γ(dτ + vdξ
c 2 ) =  with v = dx
dt
1 − v 2
c 2
dξ + vdτ
dx = γ(dξ + vdτ) = 

Motion Mountain – The Adventure of Physics available free of charge at [Link]

1 − v 2
c 2
dy = dυ
dz = dζ . (111)

These expressions describe how length and time intervals measured by different observ-
ers are related. At relative speeds v that are small compared to the velocity of light, such
as occur in everyday life, the time intervals are essentially equal; the stretch factor or re-
lativistic correction or relativistic contraction γ is then equal to 1 for all practical purposes.
However, for velocities near that of light the measurements of the two observers give
different values. In these cases, space and time mix, as shown in Figure 145.
The expressions (111) are also strange in another respect. When two observers look
Challenge 575 n at each other, each of them claims to measure shorter intervals than the other. In other
words, special relativity shows that the grass on the other side of the fence is always shorter
– if one rides along beside the fence on a bicycle and if the grass is inclined. We explore
this bizarre result in more detail shortly.
The stretch factor γ is equal to 1 for most practical purposes in everyday life. The largest
value humans have ever achieved is about 2 ċ 105 ; the largest observed value in nature is
Challenge 576 n about 1012 . Can you imagine where they occur?
Once we know how space and time intervals change, we can easily deduce how coordin-
ates change. Figures 144 and 145 show that the x coordinate of an event L is the sum of
Copyright © Christoph Schiller November 1997–May 2006

two intervals: the ξ coordinate and the length of the distance between the two origins. In
other words, we have
dx
ξ = γ(x − vt) and v = . (112)
dt
Using the invariance of the space-time interval, we get

τ = γ(t − xv
c 2 ) . (113)

Henri Poincaré called these two relations the Lorentz transformations of space and time

Dvipsbugw
 294 ii special rel ativity • 5. speed, rest and light

after their discoverer, the Dutch physicist Hendrik Antoon Lorentz.* In one of the most
Ref. 263 beautiful discoveries of physics, in 1892 and 1904, Lorentz deduced these relations from
Page 546 the equations of electrodynamics, where they had been lying, waiting to be discovered,
since 1865.** In that year James Clerk Maxwell had published the equations in order to
describe everything electric and magnetic. However, it was Einstein who first understood
that t and τ, as well as x and ξ, are equally correct and thus equally valid descriptions of
space and time.
The Lorentz transformation describes the change of viewpoint from one inertial frame
to a second, moving one. This change of viewpoint is called a (Lorentz) boost. The formu- Dvipsbugw
lae (112) and (113) for the boost are central to the theories of relativity, both special and
general. In fact, the mathematics of special relativity will not get more difficult than that:
if you know what a square root is, you can study special relativity in all its beauty.
Many alternative formulae for boosts have been explored, such as expressions in which
Ref. 264 the relative acceleration of the two observers is included, as well as the relative velocity.
However, they had all to be discarded after comparing their predictions with experi-

Motion Mountain – The Adventure of Physics available free of charge at [Link]

mental results. Before we have a look at such experiments, we continue with a few logical
deductions from the boost relations.

What is space-time?

“
Von Stund’ an sollen Raum für sich und Zeit für
sich völlig zu Schatten herabsinken und nur
noch eine Art Union der beiden soll
Selbstständigkeit bewaren.***

”
Hermann Minkowski.

The Lorentz transformations tell us something important: that space and time are two
aspects of the same basic entity. They ‘mix’ in different ways for different observers. This
fact is commonly expressed by stating that time is the fourth dimension. This makes sense
because the common basic entity – called space-time – can be defined as the set of all
events, events being described by four coordinates in time and space, and because the set
Challenge 577 n of all events has the properties of a manifold.**** (Can you confirm this?)
In other words, the existence of a maximum speed in nature forces us to introduce
a space-time manifold for the description of nature. In the theory of special relativity,
the space-time manifold is characterized by a simple property: the space-time interval di
Ref. 265 between two nearby events, defined as
Copyright © Christoph Schiller November 1997–May 2006

v2
di 2 = c 2 dt 2 − dx 2 − dy 2 − dz 2 = c 2 dt 2 (1 − ), (114)
c2

* For information about Hendrik Antoon Lorentz, see page 290.

** The same discovery had been published first in 1887 by the German physicist Woldemar Voigt (1850–1919);
Voigt – pronounced ‘Fohgt’ – was also the discoverer of the Voigt effect and the Voigt tensor. Independently,
in 1889, the Irishman George F. Fitzgerald also found the result.
*** ‘Henceforth space by itself and time by itself shall completely fade into shadows and only a kind of union
of the two shall preserve autonomy.’ This famous statement was the starting sentence of Minkowski’s 1908
talk at the meeting of the Gesellschaft für Naturforscher und Ärzte.
Page 1214 **** The term ‘manifold’ is defined in Appendix D.

Dvipsbugw
 maximum speed, observers at rest, and motion of light 295

is independent of the (inertial) observer. Such a space-time is also called Minkowski

space-time, after Hermann Minkowski* the teacher of Albert Einstein; he was the first,
in 1904, to define the concept of space-time and to understand its usefulness and import-
ance.
The space-time interval di of equation (114) has a simple interpretation. It is the time
measured by an observer moving from event (t, x) to event (t + dt, x + dx), the so-called
proper time, multiplied by c. If we neglect the factor c, we could simply call it wristwatch
time.
We live in a Minkowski space-time, so to speak. Minkowski space-time exists inde- Dvipsbugw
pendently of things. And even though coordinate systems can be different from observer
to observer, the underlying entity, space-time, is still unique, even though space and time
by themselves are not.
How does Minkowski space-time differ from Galilean space-time, the combination of
everyday space and time? Both space-times are manifolds, i.e. continuum sets of points,
both have one temporal and three spatial dimensions, and both manifolds have the topo-

Motion Mountain – The Adventure of Physics available free of charge at [Link]

Challenge 578 n logy of the punctured sphere. (Can you confirm this?) Both manifolds are flat, i.e. free of
curvature. In both cases, space is what is measured with a metre rule or with a light ray,
and time is what is read from a clock. In both cases, space-time is fundamental; it is and
remains the background and the container of things and events.
The central difference, in fact the only one, is that Minkowski space-time, in contrast to
the Galilean case, mixes space and time, and in particular, does so differently for observers
with different speeds, as shown in Figure 145. That is why time is an observer-dependent
concept.
The maximum speed in nature thus forces us to describe motion with space-time. That
is interesting, because in space-time, speaking in simple terms, motion does not exist. Mo-
tion exists only in space. In space-time, nothing moves. For each point particle, space-
time contains a world-line. In other words, instead of asking why motion exists, we can
equivalently ask why space-time is criss-crossed by world-lines. At this point, we are still
far from answering either question. What we can do is to explore how motion takes place.

Can we travel to the past? – Time and causality

We know that time is different for different observers. Does time nevertheless order
events in sequences? The answer given by relativity is a clear ‘yes and no’. Certain sets
of events are not naturally ordered by time; others sets are. This is best seen in a space-
Copyright © Christoph Schiller November 1997–May 2006

time diagram.
Clearly, two events can be placed in sequence only if one event is the cause of the other.
But this connection can only apply if the events exchange energy (e.g. through a signal).
In other words, a relation of cause and effect between two events implies that energy or
signals can travel from one event to the other; therefore, the speed connecting the two
events must not be larger than the speed of light. Figure 146 shows that event E at the
origin of the coordinate system can only be influenced by events in quadrant IV (the past
light cone, when all space dimensions are included), and can itself influence only events

* Hermann Minkowski (1864–1909), German mathematician. He had developed similar ideas to Einstein,
but the latter was faster. Minkowski then developed the concept of space-time. Minkowski died suddenly at
the age of 44.

Dvipsbugw
 296 ii special rel ativity • 5. speed, rest and light

time t

ne

lig
II
lig

ne
T future T

co

ht
ht

co
future

ht

co
co

ht
lig

ne
ne

lig
III I
E elsewhere E elsewhere
elsewhere space y

Dvipsbugw
IV x

past
past

F I G U R E 146 A space-time diagram for a moving object T seen from an inertial observer O in the case
of one and two spatial dimensions

Motion Mountain – The Adventure of Physics available free of charge at [Link]

in quadrant II (the future light cone). Events in quadrants I and III neither influence nor
are influenced by event E. The light cone defines the boundary between events that can
be ordered with respect to their origin – namely those inside the cone – and those that
cannot – those outside the cones, happening elsewhere for all observers. (Some people
call all the events happening elsewhere the present.) So, time orders events only partially.
For example, for two events that are not causally connected, their temporal order (or their
simultaneity) depends on the observer!
In particular, the past light cone gives the complete set of events that can influence
what happens at the origin. One says that the origin is causally connected only to the past
light cone. This statement reflects the fact that any influence involves transport of energy,
and thus cannot travel faster than the speed of light. Note that causal connection is an
invariant concept: all observers agree on whether or not it applies to two given events.
Challenge 579 n Can you confirm this?
A vector inside the light cone is called timelike; one on the light cone is called lightlike
or null; and one outside the cone is called spacelike. For example, the world-line of an
observer, i.e. the set of all events that make up its past and future history, consists of time-
like events only. Time is the fourth dimension; it expands space to space-time and thus
Copyright © Christoph Schiller November 1997–May 2006

‘completes’ space-time. This is the relevance of the fourth dimension to special relativity,
no more and no less.
Special relativity thus teaches us that causality and time can be defined only because
light cones exist. If transport of energy at speeds faster than that of light did exist, time
could not be defined. Causality, i.e. the possibility of (partially) ordering events for all
observers, is due to the existence of a maximal speed.
If the speed of light could be surpassed in some way, the future could influence the past.
Challenge 580 n Can you confirm this? In such situations, one would observe acausal effects. However,
there is an everyday phenomenon which tells that the speed of light is indeed maximal:
our memory. If the future could influence the past, we would also be able to remember the
future. To put it in another way, if the future could influence the past, the second principle

Dvipsbugw
 curiosities of special rel ativity 297

of thermodynamics would not be valid and our memory would not work.* No other data
from everyday life or from experiments provide any evidence that the future can influence
the past. In other words, time travel to the past is impossible. How the situation changes in
quantum theory will be revealed later on. Interestingly, time travel to the future is possible,
as we will see shortly.

Curiosities of special rel ativity

Faster than light: how far can we travel? Dvipsbugw
How far away from Earth can we travel, given that the trip should not last more than
a lifetime, say 80 years, and given that we are allowed to use a rocket whose speed can
approach the speed of light as closely as desired? Given the time t we are prepared to
spend in a rocket, given the speed v of the rocket and assuming optimistically that it can
accelerate and decelerate in a negligible amount of time, the distance d we can move away
Challenge 581 e is given by

Motion Mountain – The Adventure of Physics available free of charge at [Link]

vt
d= . (115)
1 − v 2
c 2

The distance d is larger than ct already for v 0.71c, and, if v is chosen large enough, it
increases beyond all bounds! In other words, relativity does not limit the distance we can
travel in a lifetime, and not even the distance we can travel in a single second. We could,
in principle, roam the entire universe in less than a second. In situations such as these it
makes sense to introduce the concept of proper velocity w, defined as
v
w = d
t =  =γv. (116)
1 − v 2
c 2

As we have just seen, proper velocity is not limited by the speed of light; in fact the proper
velocity of light itself is infinite.**

Synchronization and time travel – can a mother stay younger

than her own daughter?
A maximum speed implies that time is is different for different observers moving relative
to each other. So we have to be careful about how we synchronize clocks that are far
apart, even if they are at rest with respect to each other in an inertial reference frame. For
Copyright © Christoph Schiller November 1997–May 2006

example, if we have two similar watches showing the same time, and if we carry one of
them for a walk and back, they will show different times afterwards. This experiment has

* Another related result is slowly becoming common knowledge. Even if space-time had a nontrivial shape,
such as a cylindrical topology with closed time-like curves, one still would not be able to travel into the
Ref. 266 past, in contrast to what many science fiction novels suggest. This is made clear by Stephen Blau in a recent
pedagogical paper.
** Using proper velocity, the relation given in equation (108) for the superposition of two velocities wa = γ a va
Challenge 582 e and wb = γ b vb simplifies to
w s
= γ a γ b (v a + v b
) and w s = w b , (117)
where the signs  and  designate the component in the direction of and the component perpendicular to
Ref. 267 va , respectively. One can in fact express all of special relativity in terms of ‘proper’ quantities.

Dvipsbugw
 298 ii special rel ativity • 5. speed, rest and light

first
twin

trip of
Earth second twin
time Dvipsbugw
time
comparison
and
first change of
twin rocket

Motion Mountain – The Adventure of Physics available free of charge at [Link]

F I G U R E 147 The twin paradox

Ref. 268, Ref. 269 actually been performed several times and has fully confirmed the prediction of special
relativity. The time difference for a person or a watch in an aeroplane travelling around the
Earth once, at about 900 km
h, is of the order of 100 ns – not very noticeable in everyday
life. In fact, the delay is easily calculated from the expression

t
=γ. (118)
t

Human bodies are clocks; they show the elapsed time, usually called age, by various
changes in their shape, weight, hair colour, etc. If a person goes on a long and fast trip, on
her return she will have aged less than a second person who stayed at her (inertial) home.
The most famous illustration of this is the famous twin paradox (or clock paradox).
An adventurous twin jumps on a relativistic rocket that leaves Earth and travels for many
years. Far from Earth, he jumps on another relativistic rocket going the other way and
returns to Earth. The trip is illustrated in Figure 147. At his arrival, he notes that his
twin brother on Earth is much older than himself. Can you explain this result, especially
the asymmetry between the two brothers? This result has also been confirmed in many
Copyright © Christoph Schiller November 1997–May 2006

Ref. 270 experiments.

Special relativity thus confirms, in a surprising fashion, the well-known observation
that those who travel a lot remain younger. The price of the retained youth is, however,
that everything around one changes very much more quickly than if one is at rest with
the environment.
The twin paradox can also be seen as a confirmation of the possibility of time travel
to the future. With the help of a fast rocket that comes back to its starting point, we can
arrive at local times that we would never have reached within our lifetime by staying
home. Alas, we can never return to the past.*
Ref. 271 * There are even special books on time travel, such as the well researched text by Nahin. Note that the concept

Dvipsbugw
 curiosities of special rel ativity 299

One of the simplest experiments confirming

the prolonged youth of fast travellers involves the higher atmosphere

counting of muons. Muons are particles that are

continuously formed in the upper atmosphere
Page 897 by cosmic radiation. Muons at rest (with respect
to the measuring clock) have a finite half-life of
2.2 µs (or, at the speed of light, 660 m). After this
amount of time, half of the muons have decayed. high
This half-life can be measured using simple muon counter Dvipsbugw
counters. In addition, there exist special counters
that only count muons travelling within a certain decays
speed range, say from 0.9950c to 0.9954c. One
can put one of these special counters on top of
a mountain and put another in the valley below, low
counter
as shown in Figure 148. The first time this exper-

Motion Mountain – The Adventure of Physics available free of charge at [Link]

iment was performed, the height difference was F I G U R E 148 More muons than expected
Ref. 272 1.9 km. Flying 1.9 km through the atmosphere at arrive at the ground because fast travel
the mentioned speed takes about 6.4 µs. With the keeps them young
half-life just given, a naive calculation finds that
Challenge 583 n only about 13% of the muons observed at the top should arrive at the lower site. However,
it is observed that about 82% of the muons arrive below. The reason for this result is the
relativistic time dilation. Indeed, at the mentioned speed, muons experience a time dif-
ference of only 0.62 µs during the travel from the mountain top to the valley. This shorter
time yields a much lower number of lost muons than would be the case without time
dilation; moreover, the measured percentage confirms the value of the predicted time
Challenge 584 n dilation factor γ within experimental errors, as you may want to check. A similar effect
is seen when relativistic muons are produced in accelerators.
Half-life dilation has also been found for many other decaying systems, such as pions,
hydrogen atoms, neon atoms and various nuclei, always confirming the predictions of
special relativity. Since all bodies in nature are made of particles, the ‘youth effect’ of high
speeds (usually called ‘time dilation’) applies to bodies of all sizes; indeed, it has not only
Ref. 252 been observed for particles, but also for lasers, radio transmitters and clocks.
If motion leads to time dilation, a clock on the Equator, constantly running around
the Earth, should go slower than one at the poles. However, this prediction, which was
Ref. 273 made by Einstein himself, is incorrect. The centrifugal acceleration leads to a reduction in
Copyright © Christoph Schiller November 1997–May 2006

gravitational acceleration that exactly cancels the increase due to the velocity. This story
serves as a reminder to be careful when applying special relativity in situations involving
gravity. Special relativity is only applicable when space-time is flat, not when gravity is
present.
In short, a mother can stay younger than her daughter. We can also conclude that we
cannot synchronize clocks at rest with respect to each other simply by walking, clock in
hand, from one place to another. The correct way to do so is to exchange light signals.
Challenge 585 n Can you describe how?

of time travel has to be clearly defined; otherwise one has no answer to the clerk who calls his office chair a
time machine, as sitting on it allows him to get to the future.

Dvipsbugw
 300 ii special rel ativity • 5. speed, rest and light

observations
observations by the pilot
by the farmer

pilot
time Dvipsbugw
farmer
time

plane ends
barn ends

Motion Mountain – The Adventure of Physics available free of charge at [Link]

F I G U R E 149 The observations of the pilot and the barn owner

A precise definition of synchronization allows us to call two distant events simultan-

eous. In addition, special relativity shows that simultaneity depends on the observer. This
is confirmed by all experiments performed so far.
However, the mother’s wish is not easy to fulfil. Let us imagine that a mother is acceler-
ated in a spaceship away from Earth at 10 m
s2 for ten years, then decelerates at 10 m
s2
for another ten years, then accelerates for ten additional years towards the Earth, and fi-
nally decelerates for ten final years in order to land safely back on our planet. The mother
has taken 40 years for the trip. She got as far as 22 000 light years from Earth. At her
return on Earth, 44 000 years have passed. All this seems fine, until we realize that the
necessary amount of fuel, even for the most efficient engine imaginable, is so large that
Challenge 586 e the mass returning from the trip is only one part in 2 ċ 1019 . The necessary amount of fuel
does not exist on Earth.

Length contraction
The length of an object measured by an observer attached to the object is called its proper
Copyright © Christoph Schiller November 1997–May 2006

length. According to special relativity, the length measured by an inertial observer passing
by is always smaller than the proper length. This result follows directly from the Lorentz
Challenge 587 e transformations.
For a Ferrari driving at 300 km
h or 83 m
s, the length is contracted by 0.15 pm: less
than the diameter of a proton. Seen from the Sun, the Earth moves at 30 km
s; this gives
a length contraction of 6 cm. Neither of these effects has ever been measured. But larger
effects could be. Let us explore some examples.
Imagine a pilot flying through a barn with two doors, one at each end. The plane
is slightly longer than the barn, but moves so rapidly that its relativistically contracted
length is shorter than the length of the barn. Can the farmer close the barn (at least for
a short time) with the plane completely inside? The answer is positive. But why can the

Dvipsbugw
 curiosities of special rel ativity 301

ski ski
h
trap trap

F I G U R E 150 The observations of the trap digger and of the snowboarder, as

(misleadingly) published in the literature

h
Dvipsbugw

v
B rope F

g < h v(t) v(t)

Motion Mountain – The Adventure of Physics available free of charge at [Link]

F I G U R E 151 Does the conducting glider keep the F I G U R E 152 What happens to
lamp lit at large speeds? the rope?

pilot not say the following: relative to him, the barn is contracted; therefore the plane
does not fit inside the barn? The answer is shown in Figure 149. For the farmer, the doors
close (and reopen) at the same time. For the pilot, they do not. For the farmer, the pilot is
in the dark for a short time; for the pilot, the barn is never dark. (That is not completely
Challenge 588 n true: can you work out the details?)
We now explore some variations of the general case. Can a rapid snowboarder fall into
a hole that is a bit shorter than his board? Imagine him boarding so fast that the length
contraction factor γ = d
d
is 4.* For an observer on the ground, the snowboard is four
times shorter, and when it passes over the hole, it will fall into it. However, for the boarder,
it is the hole which is four times shorter; it seems that the snowboard cannot fall into it.
Ref. 274 More careful analysis shows that, in contrast to the observation of the hole digger, the
snowboarder does not experience the board’s shape as fixed: while passing over the hole,
the boarder observes that the board takes on a parabolic shape and falls into the hole,
Challenge 590 e as shown in Figure 150. Can you confirm this? In other words, shape is not an observer-
invariant concept. (However, rigidity is observer-invariant, if defined properly; can you
Challenge 591 n confirm this?)
This explanation, though published, is not correct, as Harald van Lintel and Christian
Copyright © Christoph Schiller November 1997–May 2006

Ref. 275 Gruber have pointed out. One should not forget to estimate the size of the effect. At re-
lativistic speeds the time required for the hole to affect the full thickness of the board
cannot be neglected. The snowboarder only sees his board take on a parabolic shape if
it is extremely thin and flexible. For usual boards moving at relativistic speeds, the snow-
Challenge 592 ny boarder has no time to fall any appreciable height h or to bend into the hole before passing
it. Figure 150 is so exaggerated that it is incorrect. The snowboarder would simply speed
over the hole.
The paradoxes around length contraction become even more interesting in the case of
Ref. 276 a conductive glider that makes electrical contact between two rails, as shown in Figure 151.
Challenge 589 n * Even the Earth contracts in its direction of motion around the Sun. Is the value measurable?

Dvipsbugw
 302 ii special rel ativity • 5. speed, rest and light

The two rails are parallel, but one rail has a gap that is longer than the glider. Can you
work out whether a lamp connected in series stays lit when the glider moves along the
Challenge 593 n rails with relativistic speed? (Make the simplifying and not fully realistic assumption that
electrical current flows as long and as soon as the glider touches the rails.) Do you get
the same result for all observers? And what happens when the glider is longer than the
detour? (Warning: this problem gives rise to heated debates!) What is unrealistic in this
experiment?
Ref. 277 Another example of length contraction appears when two objects, say two cars, are
connected over a distance d by a straight rope, as shown in Figure 152 Imagine that both Dvipsbugw
are at rest at time t = 0 and are accelerated together in exactly the same way. The observer
at rest will maintain that the two cars remain
 the same distance apart. On the other hand,


the rope needs to span a distance d = d
1 − v 2
c 2 , and thus has to expand when the
two cars are accelerating. In other words, the rope will break. Is this prediction confirmed
Challenge 594 n by observers on each of the two cars?
A funny – but quite unrealistic – example of length contraction is that of a submarine

Motion Mountain – The Adventure of Physics available free of charge at [Link]

Ref. 278 moving horizontally. Imagine that the resting submarine has tuned its weight to float in
water without any tendency to sink or to rise. Now the submarine moves (possibly with
relativistic speed). The captain observes the water outside to be Lorentz contracted; thus
the water is denser and he concludes that the submarine will rise. A nearby fish sees the
submarine to be contracted, thus denser than water, and concludes that the submarine
Challenge 595 n will sink. Who is wrong, and what is the buoyancy force? Alternatively, answer the fol-
Challenge 596 n lowing question: why is it impossible for a submarine to move at relativistic speed?
In summary, length contraction is can almost never be realistically observed for mac-
roscopic bodies. However, it does play an important role for images.

Relativistic films – aberration and Doppler effect

We have encountered several ways in which observations change when an observer
moves at high speed. First of all, Lorentz contraction and aberration lead to distorted
images. Secondly, aberration increases the viewing angle beyond the roughly 180 degrees
that humans are used to in everyday life. A relativistic observer who looks in the direc-
tion of motion sees light that is invisible for a resting observer, because for the latter, it
comes from behind. Thirdly, the Doppler effect produces colour-shifted images. Fourthly,
the rapid motion changes the brightness and contrast of the image: the so-called search-
light effect. Each of these changes depends on the direction of sight; they are shown in
Copyright © Christoph Schiller November 1997–May 2006

Figure 154.
Modern computers enable us to simulate the observations made by rapid observers
with photographic quality, and even to produce simulated films.* The images of Figure 153
are particularly helpful in allowing us to understand image distortion. They show the
viewing angle, the circle which distinguish objects in front of the observer from those
behind the observer, the coordinates of the observer’s feet and the point on the horizon

* See for example images and films at [Link] by Anthony Searle, at

[Link] by Daniel Weiskopf, at [Link]
[Link]/~dragon/stonehenge/[Link] by Norbert Dragon and Nicolai Mokros, or at http://
[Link] by Hanns Ruder’s group.

Dvipsbugw
curiosities of special rel ativity 303

Dvipsbugw

Motion Mountain – The Adventure of Physics available free of charge at [Link]

Copyright © Christoph Schiller November 1997–May 2006

F I G U R E 153 Flying through twelve vertical columns (shown in the two uppermost images) with 0.9
times the speed of light as visualized by Nicolai Mokros and Norbert Dragon, showing the effect of
speed and position on distortions (© Nicolai Mokros)

Dvipsbugw
 304 ii special rel ativity • 5. speed, rest and light

F I G U R E 154 Flying through three straight and vertical columns with 0.9 times the speed of light as Dvipsbugw
visualized by Daniel Weiskopf: on the left with the original colours; in the middle including the
Doppler effect; and on the right including brightness effects, thus showing what an observer would
actually see (© Daniel Weiskopf )

Motion Mountain – The Adventure of Physics available free of charge at [Link]

F I G U R E 155 What a researcher standing and one running rapidly through a corridor observe
Copyright © Christoph Schiller November 1997–May 2006

(ignoring colour effects) (© Daniel Weiskopf )

toward which the observer is moving. Adding these markers in your head when watching
other pictures or films may help you to understand more clearly what they show.
We note that the shape of the image seen by a moving observer is a distorted version
of that seen by one at rest at the same point. A moving observer, however, does not see
different things than a resting one at the same point. Indeed, light cones are independent
of observer motion.
The Lorentz contraction is measurable; however, it cannot be photographed. This sur-
Ref. 279 prising distinction was discovered only in 1959. Measuring implies simultaneity at the ob-

Dvipsbugw
 curiosities of special rel ativity 305

ject’s position; photographing implies simultaneity at the observer’s position. On a photo-

graph, the Lorentz contraction is modified by the effects due to different light travel times
from the different parts of an object; the result is a change in shape that is reminiscent of,
but not exactly the same as, a rotation. The total deformation is an angle-dependent aber-
Page 277 ration. We discussed aberration at the beginning of this section. Aberration transforms
circles into circles: such transformations are called conformal.
The images of Figure 155, produced by Daniel Weiskopf, also include the Doppler effect
and the brightness changes. They show that these effects are at least as striking as the
distortion due to aberration. Dvipsbugw
This leads to the ‘pearl necklace paradox’. If the relativistic motion transforms spheres
into spheres, and rods into shorter rods, what happens to a pearl necklace moving along
Challenge 597 n its own long axis? Does it get shorter?
There is much more to be explored using relativistic films. For example, the author
Challenge 598 r predicts that films of rapidly rotating spheres in motion will reveal interesting effects.
Also in this case, optical observation and measurement results will differ. For certain

Motion Mountain – The Adventure of Physics available free of charge at [Link]

combinations of relativistic rotations and relativistic boosts, it is predicted* that the sense
of rotation (clockwise or anticlockwise) will differ for different observers. This effect will
play an interesting role in the discussion of unification.

Which is the best seat in a bus?

Ref. 277 Let us explore another surprise of special relativity. Imagine two twins inside two identic-
ally accelerated cars, one in front of the other, starting from standstill at time t = 0, as
described by an observer at rest with respect to both of them. (There is no connecting
rope now.) Both cars contain the same amount of fuel. We easily deduce that the accele-
ration of the two twins stops, when the fuel runs out, at the same time in the frame of
Challenge 599 e the outside observer. In addition, the distance between the cars has remained the same
all along for the outside observer, and the two cars continue rolling with an identical con-
stant velocity v, as long as friction is negligible. If we call the events at which the front car
and back car engines switch off f and b, their time coordinates in the outside frame are
related simply by t f = t b . By using the Lorentz transformations you can deduce for the
Challenge 600 ny frame of the freely rolling twins the relation

t b = γ∆x v
c 2 + t f , (119) Copyright © Christoph Schiller November 1997–May 2006

which means that the front twin has aged more than the back twin! Thus, in accelerated
systems, ageing is position-dependent.
For choosing a seat in a bus, though, this result does not help. It is true that the best
seat in an accelerating bus is the back one, but in a decelerating bus it is the front one. At
the end of a trip, the choice of seat does not matter.
Is it correct to deduce that people on high mountains age faster than people in valleys,
Challenge 601 n so that living in a valley helps postponing grey hair?

* In July 2005.

Dvipsbugw
 306 ii special rel ativity • 5. speed, rest and light

average speed: c/2

time time
average speed: c/3
t'

moving t'
judge

J light signal
moving Dvipsbugw
x'
judge
J
light signal x'

space space

Motion Mountain – The Adventure of Physics available free of charge at [Link]

F I G U R E 156 For the athlete on the left, the judge moving in the opposite
direction sees both feet off the ground at certain times, but not for the
athlete on the right

How fast can one walk?

To walk means to move the feet in such a way that at least one of them is on the ground at
any time. This is one of the rules athletes have to follow in Olympic walking competitions;
they are disqualified if they break it. A student athlete was thinking about the theoretical
maximum speed he could achieve in the Olympics. The ideal would be that each foot
accelerates instantly to (almost) the speed of light. The highest walking speed is achieved
by taking the second foot off the ground at exactly the same instant at which the first is put
down. By ‘same instant’, the student originally meant ‘as seen by a competition judge at
rest with respect to Earth’. The motion of the feet is shown in the left diagram of Figure 156;
it gives a limit speed for walking of half the speed of light. But then the student noticed
that a moving judge will see both feet off the ground and thus disqualify the athlete for
Ref. 280 running. To avoid disqualification by any judge, the second foot has to wait for a light
signal from the first. The limit speed for Olympic walking is thus only one third of the
Copyright © Christoph Schiller November 1997–May 2006

speed of light.

Is the speed of shadow greater than the speed of light?

Actually, motion faster than light does exist and is even rather common. Special relativity
only constrains the motion of mass and energy. However, non-material points or non-
energy-transporting features and images can move faster than light. There are several
Page 297 simple examples. To be clear, we are not talking about proper velocity, which in these
Challenge 602 n cases cannot be defined anyway. (Why?)
The following examples show speeds that are genuinely higher than the speed of light
in vacuum.
Consider the point marked X in Figure 157, the point at which scissors cut paper. If

Dvipsbugw
 curiosities of special rel ativity 307

The Beatles

The Beatles

v Dvipsbugw
X

The Beatles

F I G U R E 157 A simple example of motion that is F I G U R E 158 Another

Motion Mountain – The Adventure of Physics available free of charge at [Link]

faster than light example of faster-than-light
motion

the scissors are closed rapidly enough, the point moves faster than light. Similar examples
can also be found in every window frame, and in fact in any device that has twisting parts.
Another example of superluminal motion is a music record – an old-fashioned LP
– disappearing into its sleeve, as shown in Figure 158. The point where the edge of the
record meets the edge of the sleeve can travel faster than light.
Another example suggests itself when we remember that we live on a spherical planet.
Imagine you lie on the floor and stand up. Can you show that the initial speed with which
Challenge 603 n the horizon moves away from you can be larger than that of light?
Finally, a standard example is the motion of a spot of light produced by shining a laser
beam onto the Moon. If the laser is moved, the spot can easily move faster than light. The
same applies to the light spot on the screen of an oscilloscope when a signal of sufficiently
high frequency is fed to the input.
All these are typical examples of the speed of shadows, sometimes also called the speed
of darkness. Both shadows and darkness can indeed move faster than light. In fact, there
Copyright © Christoph Schiller November 1997–May 2006

Challenge 604 n is no limit to their speed. Can you find another example?
In addition, there is an ever-increasing number of experimental set-ups in which the
phase velocity or even the group velocity of light is higher than c. They regularly make
headlines in the newspapers, usually along the lines of ‘light moves faster than light’. We
Page 577 will discuss this surprising phenomenon in more detail later on. In fact, these cases can
also be seen – with some imagination – as special cases of the ‘speed of shadow’ phe-
nomenon.
For a different example, imagine we are standing at the exit of a tunnel of length l. We
see a car, whose speed we know to be v, entering the other end of the tunnel and driving
towards us. We know that it entered the tunnel because the car is no longer in the Sun or
because its headlights were switched on at that moment. At what time t, after we see it

Dvipsbugw
 308 ii special rel ativity • 5. speed, rest and light

time
observer
emitted or reflected light

tachyon

Dvipsbugw

light cone

Motion Mountain – The Adventure of Physics available free of charge at [Link]

space

F I G U R E 159 Hypothetical space-time diagram for

tachyon observation

entering the tunnel, does it drive past us? Simple reasoning shows that t is given by

t = l
v − l
c . (120)

In other words, the approaching car seems to have a velocity v appr of

l vc
v appr = = , (121)
t c−v

which is higher than c for any car velocity v higher than c
2. For cars this does not happen
too often, but astronomers know a type of bright object in the sky called a quasar (a
contraction of ‘quasi-stellar object’), which sometimes emits high-speed gas jets. If the
emission is in or near the direction of the Earth, its apparent speed – even the purely
transverse component – is higher than c. Such situations are now regularly observed with
Copyright © Christoph Schiller November 1997–May 2006

Ref. 281 telescopes.

Note that to a second observer at the entrance of the tunnel, the apparent speed of the
car moving away is given by
vc
v leav = , (122)
c+v

which is never higher than c
2. In other words, objects are never seen departing with
more than half the speed of light.
The story has a final twist. We have just seen that motion faster than light can be ob-
served in several ways. But could an object moving faster than light be observed at all?
Surprisingly, it could be observed only in rather unusual ways. First of all, since such
an imaginary object, usually called a tachyon, moves faster than light, we can never see

Dvipsbugw
 curiosities of special rel ativity 309

R v G
u
O w

F I G U R E 160 If O’s stick is parallel

to R’s and R’s is parallel to G’s, then Dvipsbugw
O’s stick and G’s stick are not

it approaching. If it can be seen at all, a tachyon can only be seen departing. Seeing a
tachyon would be similar to hearing a supersonic jet. Only after a tachyon has passed
nearby, assuming that it is visible in daylight, could we notice it. We would first see a

Motion Mountain – The Adventure of Physics available free of charge at [Link]

flash of light, corresponding to the bang of a plane passing with supersonic speed. Then
we would see two images of the tachyon, appearing somewhere in space and departing
in opposite directions, as can be deduced from Figure 159. Even if one of the two images
were approaching us, it would be getting fainter and smaller. This is, to say the least, rather
unusual behaviour. Moreover, if you wanted to look at a tachyon at night, illuminating
it with a torch, you would have to turn your head in the direction opposite to the arm
with the torch! This requirement also follows from the space-time diagram: can you see
Challenge 605 e why? Nobody has ever seen such phenomena. Tachyons, if they existed, would be strange
Ref. 282 objects: they would accelerate when they lose energy, a zero-energy tachyon would be the
Page 316 fastest of all, with infinite speed, and the direction of motion of a tachyon depends on the
motion of the observer. No object with these properties has ever been observed. Worse, as
we just saw, tachyons would seem to appear from nothing, defying laws of conservation;
and note that, just as tachyons cannot be seen in the usual sense, they cannot be touched
either, since both processes are due to electromagnetic interactions, as we will see later in
our ascent of Motion Mountain. Tachyons therefore cannot be objects in the usual sense.
In the second part of our adventure we will show that quantum theory actually rules out
the existence of (real) tachyons. However, quantum theory also requires the existence of
‘virtual’ tachyons, as we will discover.

Parallel to parallel is not parallel – Thomas rotation

Copyright © Christoph Schiller November 1997–May 2006

Relativity has strange consequences indeed. Any two observers can keep a stick parallel
to the other’s, even if they are in motion with respect to each other. But strangely, given a
a chain of sticks for which any two adjacent ones are parallel, the first and the last sticks
will not generally be parallel. In particular, they never will be if the motions of the various
observers are in different directions, as is the case when the velocity vectors form a loop.
The simplest set-up is shown in Figure 160. In special relativity, a general concatenation
Ref. 283 of pure boosts does not give a pure boost, but a boost plus a rotation. As a result, the
endpoints of chains of parallel sticks are usually not parallel.
An example of this effect appears in rotating motion. If we walk in a fast circle holding
a stick, always keeping the stick parallel to the direction it had just before, at the end of the
circle the stick will have an angle with respect to the original direction. Similarly, the axis

Dvipsbugw
 310 ii special rel ativity • 5. speed, rest and light

of a rotating body circling a second body will not be pointing in the same direction after
one turn. This effect is called Thomas precession, after Llewellyn Thomas, who discovered
it in 1925, a full 20 years after the birth of special relativity. It had escaped the attention of
dozens of other famous physicists. Thomas precession is important in the inner working
of atoms; we will return to it in a later section of our adventure. These surprising phenom-
ena are purely relativistic, and are thus measurable only in the case of speeds comparable
to that of light.

A never-ending story – temperature and relativity Dvipsbugw

The literature on temperature is confusing. Albert Einstein and Wolfgang Pauli agreed
on the following result: the temperature T seen by an observer moving with speed v is
related to the temperature T0 measured by the observer at rest with respect to the heat
bath via 
T = T0 1 − v 2
c 2 . (123)

Motion Mountain – The Adventure of Physics available free of charge at [Link]

A moving observer thus always measures lower values than a resting one.
In 1908, Max Planck used this expression, together with the corresponding transform-
ation for heat, to deduce that the entropy is invariant under Lorentz transformations.
Being the discoverer of the Boltzmann constant k, Planck proved in this way that the
constant is a relativistic invariant.
Not all researchers agree on the expression. Others maintain that T and T0 should
be interchanged in the temperature transformation. Also, powers other than the simple
Ref. 284 square root have been proposed. The origin of these discrepancies is simple: temperature
is only defined for equilibrium situations, i.e. for baths. But a bath for one observer is not
a bath for the other. For low speeds, a moving observer sees a situation that is almost a
heat bath; but at higher speeds the issue becomes tricky. Temperature is deduced from
the speed of matter particles, such as atoms or molecules. For moving observers, there
is no good way to measure temperature. The naively measured temperature value even
depends on the energy range of matter particles that is measured! In short, thermal equi-
librium is not an observer-invariant concept. Therefore, no temperature transformation
formula is correct. (With certain additional assumptions, Planck’s expression does seem
to hold, however.) In fact, there are not even any experimental observations that would al-
low such a formula to be checked. Realizing such a measurement is a challenge for future
experimenters – but not for relativity itself.
Copyright © Christoph Schiller November 1997–May 2006

Rel ativistic mechanics

Because the speed of light is constant and velocities do not add up, we need to rethink
the definitions of mass, momentum and energy. We thus need to recreate a theory of
mechanics from scratch.

Mass in relativity
Page 77 In Galilean physics, the mass ratio between two bodies was defined using collisions; it

Dvipsbugw
 rel ativistic mechanics 311

was given by the negative inverse of the velocity change ratio

m2 ∆v 1
=− . (124)
m1 ∆v 2

However, experiments show that the expression must be different for speeds near that of
Challenge 606 ny light. In fact, experiments are not needed: thinking alone can show this. Can you do so?
There is only one solution to this problem. The two Galilean conservation theorems
Ref. 285
i m i v i = const for momentum and
i m i = const for mass have to be changed into Dvipsbugw

 γ i m i v i = const (125)
i

and
 γ i m i = const . (126)

Motion Mountain – The Adventure of Physics available free of charge at [Link]

i

These expressions, which will remain valid throughout the rest of our ascent of Motion
Mountain, imply, among other things, that teleportation is not possible in nature. (Can
Challenge 607 n you confirm this?) Obviously, in order to recover Galilean physics, the relativistic correc-
tion (factors) γ i have to be almost equal to 1 for everyday velocities, that is, for velocities
nowhere near the speed of light.
Even if we do not know the value of the relativistic correction factor, we can deduce it
from the collision shown in Figure 161.
In the first frame of reference (A) we have
γv mv = γ V MV and γv m + m = γ V M. From the ob- Observer A
servations of the second frame of reference (B) we m m
Challenge 608 e deduce that V composed with V gives v, in other before: v
words, that after:
2V V
v= . (127) M
1 + V
c
2 2

When these equations are combined, the relativ- Observer B

istic correction γ is found to depend on the mag- before:
nitude of the velocity v through V V
m m
after:
Copyright © Christoph Schiller November 1997–May 2006

1 M
γv =  . (128)
F I G U R E 161 An inelastic collision of two
1 − v 2
c 2
identical particles seen from two
different inertial frames of reference
With this expression, and a generalization of
the situation of Galilean physics, the mass ratio
between two colliding particles is defined as the ratio

m1 ∆(γ 2 v 2 )
=− . (129)
m2 ∆(γ 1 v 1 )

(We do not give here the generalized mass definition, mentioned in the chapter on Ga-

Dvipsbugw
 312 ii special rel ativity • 5. speed, rest and light

before

pA
A B

after
θ
pA ϕ

rule: ϕ+θ = 90° Dvipsbugw

B

F I G U R E 162 A useful rule for playing

non-relativistic snooker

Page 79 lilean mechanics, that is based on acceleration ratios, because it contains some subtleties,

Motion Mountain – The Adventure of Physics available free of charge at [Link]

which we will discover shortly.) The correction factors γ i ensure that the mass defined
by this equation is the same as the one defined in Galilean mechanics, and that it is the
same for all types of collision a body may have.* In this way, mass remains a quantity
characterizing the difficulty of accelerating a body, and it can still be used for systems of
bodies as well.
Following the example of Galilean physics, we call the quantity

p = γmv (130)

the (linear) relativistic (three-) momentum of a particle. Again, the total momentum is a
conserved quantity for any system not subjected to external influences, and this conserva-
tion is a direct consequence of the way mass is defined.
For low speeds, or γ  1, relativistic momentum is the same as that of Galilean physics,
and is proportional to velocity. But for high speeds, momentum increases faster than
velocity, tending to infinity when approaching light speed.

Why relativistic snooker is more difficult

There is a well-known property of collisions between a moving sphere or particle and a
resting one of the same mass that is important when playing snooker, pool or billiards.
After such a collision, the two spheres will depart at a right angle from each other, as
Copyright © Christoph Schiller November 1997–May 2006

shown in Figure 162.

However, experiments show that the right angle rule does not apply to relativistic col-
lisions. Indeed, using the conservation of momentum and a bit of dexterity you can cal-
Challenge 610 ny culate that
2
tan θ tan φ = , (131)
γ+1

where the angles are defined in Figure 163. It follows that the sum φ + θ is smaller than
a right angle in the relativistic case. Relativistic speeds thus completely change the game

Challenge 609 e * The results below also show that γ = 1 + T
mc 2 , where T is the kinetic energy of a particle.

Dvipsbugw
 rel ativistic mechanics 313

θ
ϕ

accelerator beam target detector

Dvipsbugw
F I G U R E 163 The dimensions of detectors in particle accelerators are based on the relativistic
snooker angle rule

of snooker. Indeed, every accelerator physicist knows this: for electrons or protons, these
angles can easily be deduced from photographs taken in cloud chambers, which show

Motion Mountain – The Adventure of Physics available free of charge at [Link]

the tracks left by particles when they move through them. All such photographs confirm
Ref. 252 the above expression. In fact, the shapes of detectors are chosen according to expression
(131), as sketched in Figure 163. If the formula – and relativity – were wrong, most of these
detectors would not work, as they would miss most of the particles after the collision. In
fact, these experiments also prove the formula for the composition of velocities. Can you
Challenge 611 ny show this?

Mass is concentrated energy

Let us go back to the collinear and inelastic collision of Figure 161. What is the mass M
Challenge 612 n of the final system? Calculation shows that

M
m = 2(1 + γv ) 2 . (132)

In other words, the mass of the final system is larger than the sum of the two original
masses m. In contrast to Galilean mechanics, the sum of all masses in a system is not a
conserved quantity. Only the sum
i γ i m i of the corrected masses is conserved.
Relativity provides the solution to this puzzle. Everything falls into place if, for the
energy E of an object of mass m and velocity v, we use the expression Copyright © Christoph Schiller November 1997–May 2006

mc 2
E = γmc 2 =  , (133)
1 − v 2
c 2

applying it both to the total system and to each component. The conservation of the cor-
rected mass can then be read as the conservation of energy, simply without the factor c 2 .
In the example of the two identical masses sticking to each other, the two particles are
thus each described by mass and energy, and the resulting system has an energy E given
by the sum of the energies of the two particles. In particular, it follows that the energy E 0
of a body at rest and its mass m are related by

E 0 = mc 2 , (134)

Dvipsbugw
 314 ii special rel ativity • 5. speed, rest and light

which is perhaps the most beautiful and famous discovery of modern physics. Since c 2 is
so large, we can say that mass is concentrated energy. In other words, special relativity says
that every mass has energy, and that every form of energy in a system has mass. Increasing
the energy of a system increases its mass, and decreasing the energy content decreases the
mass. In short, if a bomb explodes inside a closed box, the mass, weight and momentum
of the box are the same before and after the explosion, but the combined mass of the
debris inside the box will be smaller than before. All bombs – not only nuclear ones –
thus take their energy from a reduction in mass. In addition, every action of a system –
such a caress, a smile or a look – takes its energy from a reduction in mass. Dvipsbugw
The kinetic energy T is thus given by

1 1 ċ 3 v4 1 ċ 3 ċ 5 v6
T = γmc 2 − mc 2 = mv 2 + m + + ... (135)
2 2 ċ 4 c2 2 ċ 4 ċ 6 c4
Challenge 613 e (using the binomial theorem) which reduces to the Galilean value only for low speeds.

Motion Mountain – The Adventure of Physics available free of charge at [Link]

The mass–energy equivalence E = γmc 2 implies that taking any energy from matter
results in a mass decrease. When a person plays the piano, thinks or runs, its mass de-
creases. When a cup of tea cools down or when a star shines, its mass decreases. The
mass–energy equivalence pervades all of nature.
By the way, we should be careful to distinguish the transformation of mass into energy
from the transformation of matter into energy. The latter is much more rare. Can you
Challenge 614 n give some examples?
The mass–energy relation (133) means the death of many science fiction fantasies. It
implies that there are no undiscovered sources of energy on or near Earth. If such sources
existed, they would be measurable through their mass. Many experiments have looked
for, and are still looking for, such effects with a negative result. There is no free energy in
nature.*
The mass–energy relation m = E 0
c 2 also implies that one needs about 90 thousand
million kJ (or 21 thousand million kcal) to increase one’s weight by one single gram. Of
course, dieticians have slightly different opinions on this matter! In fact, humans do get
their everyday energy from the material they eat, drink and breathe by reducing its com-
bined mass before expelling it again. However, this chemical mass defect appearing when
fuel is burned cannot yet be measured by weighing the materials before and after the
reaction: the difference is too small, because of the large conversion factor involved. In-
deed, for any chemical reaction, bond energies are about 1 aJ (6 eV) per bond; this gives
Copyright © Christoph Schiller November 1997–May 2006

a weight change of the order of one part in 1010 , too small to be measured by weigh-
ing people or determining mass differences between food and excrement. Therefore, for
everyday chemical processes mass can be taken to be constant, in accordance with Ga-
lilean physics.
Modern methods of mass measurement of single molecules have made it possible to
measure the chemical mass defect by comparing the mass of a single molecule with
that of its constituent atoms. David Pritchard’s group has developed so-called Penning

* There may be two extremely diluted, yet undiscovered, form of energy, called dark matter and (confusingly)
Page 441 dark energy, scattered throughout the universe. They are deduced from (quite difficult) mass measurements.
The issue has not yet been finally resolved.

Dvipsbugw
 rel ativistic mechanics 315

traps, which allow masses to be determined from the measurement of frequencies; the
attainable precision of these cyclotron resonance experiments is sufficient to confirm
Ref. 286 ∆E 0 = ∆mc 2 for chemical bonds. In the future, increased precision will even allow bond
energies to be determined in this way with precision. Since binding energy is often radi-
ated as light, we can say that these modern techniques make it possible to weigh light.
Thinking about light and its mass was the basis for Einstein’s first derivation of the
mass–energy relation. When an object emits two equal light beams in opposite directions,
its energy decreases by the emitted amount. Since the two light beams are equal in energy
and momentum, the body does not move. If we describe the same situation from the Dvipsbugw
Challenge 615 ny viewpoint of a moving observer, we see again that the rest energy of the object is

E 0 = mc 2 . (136)

In summary, all physical processes, including collisions, need relativistic treatment

whenever the energy involved is a sizeable fraction of the rest energy.

Motion Mountain – The Adventure of Physics available free of charge at [Link]

Every energy increase produces a mass increase. Therefore also heating a body makes
it heavier. However, this effect is so small that nobody has measured it up to this day. It
is a challenge for experiments of the future to do this one day.
How are energy and momentum related? The definitions of momentum (130) and en-
Challenge 616 e ergy (133) lead to two basic relations. First of all, their magnitudes are related by

m 2 c 4 = E 2 − p2 c 2 (137)

for all relativistic systems, be they objects or, as we will see below, radiation. For the mo-
mentum vector we get the other important relation

E
p= v, (138)
c2
Challenge 617 e which is equally valid for any type of moving energy, be it an object or a beam or a pulse of
radiation.* We will use both relations often in the rest of our ascent of Motion Mountain,
including the following discussion.

Collisions, virtual objects and tachyons

We have just seen that in relativistic collisions the conservation of total energy and mo-
Copyright © Christoph Schiller November 1997–May 2006

mentum are intrinsic consequences of the definition of mass. Let us now have a look at
collisions in more detail, using these new concepts. A collision is a process, i.e. a series of
events, for which
— the total momentum before the interaction and after the interaction is the same;
— the momentum is exchanged in a small region of space-time;
— for small velocities, the Galilean description is valid.
In everyday life an impact, i.e. a short-distance interaction, is the event at which both
objects change momentum. But the two colliding objects are located at different points

* In 4-vector notation, we can write v
c = P
P0 , where P0 = E
c.

Dvipsbugw
 316 ii special rel ativity • 5. speed, rest and light

time t τ

E'2 p'2
E'1 p'1
E
p
E2 p2
E1 p1

Dvipsbugw
object 1
object 2 object 1 object 2

space x ξ

F I G U R E 164 Space-time diagram of a collision for two observers

Motion Mountain – The Adventure of Physics available free of charge at [Link]

Ref. 287 when this happens. A collision is therefore described by a space-time diagram such as the
left-hand one in Figure 164, reminiscent of the Orion constellation. It is easy to check that
Challenge 618 e the process described by such a diagram is a collision according to the above definition.
The right-hand side of Figure 164 shows the same process seen from another, Greek,
frame of reference. The Greek observer says that the first object has changed its mo-
mentum before the second one. That would mean that there is a short interval when
momentum and energy are not conserved!
The only way to make sense of the situation is to assume that there is an exchange of a
third object, drawn with a dotted line. Let us find out what the properties of this object are.
If we give numerical subscripts to the masses, energies and momenta of the two bodies,
Challenge 619 e and give them a prime after the collision, the unknown mass m obeys

1 − v 1 v 1

m 2 c 4 = (E 1 − E 1
)2 − (p 1 − p
1 )2 c 2 = 2m 12 c 4 − 2E 1 E 1
( )<0. (139)
c2
This is a strange result, because it means that the unknown mass is an imaginary num-
ber!!* On top of that, we also see directly from the second graph that the exchanged object
moves faster than light. It is a tachyon, from the Greek ταχύς ‘rapid’. In other words, colli-
sions involve motion that is faster than light! We will see later that collisions are indeed the
Copyright © Christoph Schiller November 1997–May 2006

only processes where tachyons play a role in nature. Since the exchanged objects appear
only during collisions, never on their own, they are called virtual objects, to distinguish
them from the usual, real objects, which can move freely without restriction.** We will
study their properties later on, when we come to discuss quantum theory.

* It is usual to change
 the mass–energy and mass–momentum relation of tachyons to E = mc 2
v 2
c 2 − 1
and p = mv
v 2
c 2 − 1 ; this amounts to a redefinition of m. After the redefinition, tachyons have real
mass. The energy and momentum relations show that tachyons lose energy and momentum when they get
faster. (Provocatively, a single tachyon in a box could provide us with all the energy we need.) Both signs
for the energy and momentum relations must be retained, because otherwise the equivalence of all inertial
observers would not be generated. Tachyons thus do not have a minimum energy or a minimum momentum.
** More precisely, a virtual particle does not obey the relation m 2 c 4 = E 2 − p2 c 2 , valid for real particles.

Dvipsbugw
 rel ativistic mechanics 317

In nature, a tachyon is always a virtual object. Real objects are always bradyons – from
the Greek βραδύς ‘slow’ – or objects moving slower than light. Note that tachyons, despite
their high velocity, do not allow the transport of energy faster than light; and that they
do not violate causality if and only if they are emitted or absorbed with equal probability.
Challenge 620 ny Can you confirm all this?
When we study quantum theory, we will also discover that a general contact interac-
tion between objects is described not by the exchange of a single virtual object, but by
a continuous stream of virtual particles. For standard collisions of everyday objects, the
interaction turns out to be electromagnetic. In this case, the exchanged particles are vir- Dvipsbugw
tual photons. In other words, when one hand touches another, when it pushes a stone, or
when a mountain supports the trees on it, streams of virtual photons are continuously
Page 733 exchanged.
There is an additional secret hidden in collisions. In the right-hand side of Figure 164,
the tachyon is emitted by the first object and absorbed by the second one. However, it
Challenge 621 n is easy to imagine an observer for which the opposite happens. In short, the direction

Motion Mountain – The Adventure of Physics available free of charge at [Link]

of travel of a tachyon depends on the observer! In fact, this is a hint about antimatter.
In space-time diagrams, matter and antimatter travel in opposite directions. Also the
connection between relativity and antimatter will become more apparent in quantum
Page 759 theory.

Systems of particles – no centre of mass

Relativity also forces us to eliminate the cherished concept of centre of mass. We can see
this already in the simplest example possible: that of two equal objects colliding.
Figure 165 shows that from the viewpoint in which one of two colliding particles is at
rest, there are at least three different ways to define the centre of mass. In other words, the
Ref. 288 centre of mass is not an observer-invariant concept. We can deduce from the figure that
the concept only makes sense for systems whose components move with small velocities
relative to each other. For more general systems, centre of mass is not uniquely definable.
Will this hinder us in our ascent? No. We are more interested in the motion of single
particles than that of composite objects or systems.

Why is most motion so slow?

For most everyday systems, the time intervals measured by two different observers are
practically equal; only at large relative speeds, typically at more than a few per cent of the
Copyright © Christoph Schiller November 1997–May 2006

speed of light, is there a noticeable difference. Most such situations are microscopic. We
have already mentioned the electrons inside a television tube or inside a particle accel-
erator. The particles making up cosmic radiation are another example: their high energy
has produced many of the mutations that are the basis of evolution of animals and plants
on this planet. Later we will discover that the particles involved in radioactivity are also
relativistic.
But why don’t we observe any rapid macroscopic bodies? Moving bodies, including
observers, with relativistic velocities have a property not found in everyday life: when
they are involved in a collision, part of their energy is converted into new matter via
E = γmc 2 . In the history of the universe this has happened so many times that practically
all the bodies still in relativistic motion are microscopic particles.

Dvipsbugw
 318 ii special rel ativity • 5. speed, rest and light

A CM-0 B
v v

transformed CM
A CM-1 B

v=0 v 2v/(1+v2/c2 )

Dvipsbugw
geometrical CM
A CM-2 B

v=0 2 2 2v/(1+v2/c2 )
v/(1+v /c )

momentum CM

Motion Mountain – The Adventure of Physics available free of charge at [Link]

A CM-3 B

v=0 2 2 1/2 2v/(1+v2/c2 )

v/(1- v /c )

F I G U R E 165 There is no way to define a relativistic centre

of mass

A second reason for the disappearance of rapid relative motion is radiation damping.
Challenge 622 n Can you imagine what happens to charges during collisions, or in a bath of light?
In short, almost all matter in the universe moves with small velocity relative to other
matter. The few known counter-examples are either very old, such as the quasar jets men-
tioned above, or stop after a short time. The huge energies necessary for macroscopic
relativistic motion are still found in supernova explosions, but they cease to exist after
only a few weeks. In summary, the universe is mainly filled with slow motion because it
Page 450 is old. We will determine its age shortly.

The history of the mass–energy equivalence formula of de

Pretto and Einstein
Albert Einstein took several months after his first paper on special relativity to deduce
Copyright © Christoph Schiller November 1997–May 2006

the expression
E = γmc 2 (140)

which is often called the most famous formula of physics. He published it in a second,
Ref. 241 separate paper towards the end of 1905. Arguably, the formula could have been discovered
thirty years earlier, from the theory of electromagnetism.
In fact, at least one person did deduce the result before Einstein. In 1903 and 1904,
before Einstein’s first relativity paper, a little-known Italian engineer, Olinto De Pretto,
was the first to calculate, discuss and publish the formula E = mc 2 .* It might well be that

* Umberto Bartocci, mathematics professor of the University of Perugia in Italy, published the details of

Dvipsbugw
 rel ativistic mechanics 319

lig

ne
future T

ht

co
co

ht
ne

lig
E elsewhere y

Dvipsbugw
x

past

F I G U R E 166 The space-time diagram of a

moving object T

Motion Mountain – The Adventure of Physics available free of charge at [Link]

Einstein got the idea for the formula from De Pretto, possibly through his friend Michele
Besso or other Italian-speaking friends he met when he visited his parents, who were
living in Italy at the time. Of course, the value of Einstein’s efforts is not diminished by
this.
Ref. 289 In fact, a similar formula had also been deduced in 1904 by Friedrich Hasenöhrl and
published again in Annalen der Physik in 1905, before Einstein, though with an incorrect
numerical prefactor, due to a calculation mistake. The formula E = mc 2 is also part of
several expressions in two publications in 1900 by Henri Poincaré. The real hero in the
story might well be Tolver Preston, who discussed the equivalence of mass and energy
already in 1875, in his book Physics of the Ether. The mass-energy equivalence was thus
indeed floating in the air, only waiting to be discovered.
In the 1970s there was a similar story: a simple relation between the gravitational acce-
leration and the temperature of the vacuum was discovered. The result had been waiting
to be discovered for over 50 years. Indeed, a number of similar, anterior results were
found in the libraries. Could other simple relations be hidden in modern physics waiting
Challenge 623 n to be found?

4-vectors
Copyright © Christoph Schiller November 1997–May 2006

To describe motion consistently for all observers, we have to introduce some new quant-
ities. First of all, motion of particles is seen as a sequence of events. To describe events
with precision, we use event coordinates, also called 4-coordinates. These are written as

X = (ct, x) = (ct, x, y, z) = X i . (141)

In this way, an event is a point in four-dimensional space-time, and is described by four

coordinates. The coordinates are called the zeroth, namely time X 0 = ct, the first, usually

this surprising story in several papers. The full account is found in his book Umberto Bartocci, Albert
Einstein e Olinto De Pretto: la vera storia della formula più famosa del mondo, Ultreja, Padova, 1998.

Dvipsbugw
 320 ii special rel ativity • 5. speed, rest and light

called X 1 = x, the second, X 2 = y, and the third, X 3 = z. One can then define a distance d
between events as the length of the difference vector. In fact, one usually uses the square
of the length, to avoid those unwieldy square roots. In special relativity, the magnitude
(‘squared length’) of a vector is always defined through

XX = X 0 2 − X 1 2 − X 2 2 − X 3 2 = ct 2 − x 2 − y 2 − z 2 = X a X a = η ab X a X b = η ab X a X b .(142)

In this equation we have introduced for the first time two notations that are useful in
Dvipsbugw
relativity. First of all, we automatically sum over repeated indices. Thus, X a X a means the
sum of all products X a X a as a ranges over all indices. Secondly, for every 4-vector X
we distinguish two ways to write the coordinates, namely coordinates with superscripts
and coordinates with subscripts. (In three dimensions, we only use subscripts.) They are
related by the following general relation

X a = η ab X b = (ct, −x, −y, −z) , (143)

Motion Mountain – The Adventure of Physics available free of charge at [Link]

where we have introduced the so-called metric η ab , an abbreviation of the matrix*

1 0 0 0 
 0 −1 0 0 
η ab
= η ab =  . (144)
 0 0 −1 0 
 0 0 0 −1 

Don’t panic: this is all, and it won’t get more difficult! We now go back to physics.
The magnitude of a position or distance vector, also called the space-time interval, is
essentially the proper time times c. The proper time is the time shown by a clock moving
in a straight line and with constant velocity from the starting point to the end point in
space-time. The difference from the usual 3-vectors is that the magnitude of the interval
can be positive, negative or even zero. For example, if the start and end points in space-
time require motion with the speed of light, the proper time is zero (this is required for
Page 296 null vectors). If the motion is slower than the speed of light, the squared proper time is
positive and the distance is timelike. For negative intervals and thus imaginary proper
times, the distance is spacelike.** A simplified overview is given by Figure 166.
Now we are ready to calculate and measure motion in four dimensions. The meas-
urements are based on one central idea. We cannot define the velocity of a particle as
Copyright © Christoph Schiller November 1997–May 2006

the derivative of its coordinates with respect to time, since time and temporal sequences
depend on the observer. The solution is to define all observables with respect to the just-
mentioned proper time τ, which is defined as the time shown by a clock attached to the ob-
ject. In relativity, motion and change are always measured with respect to clocks attached
to the moving system. In particular, the relativistic velocity or 4-velocity U of a body is

* Note that 30 % of all physics textbooks use the negative of η as the metric, the so-called spacelike convention,
and thus have opposite signs in this definition. In this text, as in 70 % of all physics texts, we use the timelike
convention.
** In the latter case, the negative of the magnitude, which is a positive number, is called the squared proper
distance. The proper distance is the length measured by an odometer as the object moves along.

Dvipsbugw
 rel ativistic mechanics 321

thus defined as the rate of change of the event coordinates or 4-coordinates X = (ct, x)
with respect to proper time, i.e. as

U = dX
dτ . (145)

The coordinates X are measured in the coordinate system defined by the inertial observer
chosen. The value of the velocity U depends on the observer or coordinate system used;
so the velocity depends on the observer, as it does in everyday life. Using dt = γ dτ and
thus Dvipsbugw
dx dx dt dx 1
= =γ , where as usual γ =  , (146)
dτ dt dτ dt 1 − v 2
c 2

we get the relation with the 3-velocity v = dx
dt:

u 0 = γc , u i = γv i or U = (γc, γv) . (147)

Motion Mountain – The Adventure of Physics available free of charge at [Link]

For small velocities we have γ  1, and then the last three components of the 4-velocity
are those of the usual, Galilean 3-velocity. For the magnitude of the 4-velocity U we find
UU = U a U a = η ab U a U b = c 2 , which is therefore independent of the magnitude of the
3-velocity v and makes it a timelike vector, i.e. a vector inside the light cone.*
Note that the magnitude of a 4-vector can be zero even though all its components
are different from zero. Such a vector is called null. Which motions have a null velocity
Challenge 625 n vector?
Similarly, the relativistic acceleration or 4-acceleration B of a body is defined as

B = dU
dτ = d2 X
dτ 2 . (149)

Using dγ
dτ = γdγ
dt = γ 4 va
c 2 , we get the following relations between the four com-
Ref. 290 ponents of B and the 3-acceleration a = dv
dt:

va (va)v i
B0 = γ 4 , B i = γ2 a i + γ4 . (150)
c c2

The magnitude b of the 4-acceleration is easily found via BB = η cd B c B d = −γ 4 (a 2 +

γ 2 (va)2
c 2 ) = −γ 6 (a 2 − (v  a)2
c 2 ). Note that it does depend on the value of the
Copyright © Christoph Schiller November 1997–May 2006

* In general, a 4-vector is defined as a quantity (h 0 , h 1 , h 2 , h 3 ), which transforms as

h0 = γ V (h 0 − h 1 V
c)

h 1 = γ V (h 1 − h 0 V
c)
h2 = h 2

h3 = h3 (148)

when changing from one inertial observer to another moving with a relative velocity V in the x direction; the
corresponding generalizations for the other coordinates are understood. This relation allows one to deduce
the transformation laws for any 3-vector. Can you deduce the velocity composition formula (108) from this
Challenge 624 n definition, applying it to 4-velocity?

Dvipsbugw
 322 ii special rel ativity • 5. speed, rest and light

time
(E/c , p)

space

Dvipsbugw

F I G U R E 167 Energy–momentum is tangent to

the world line

Motion Mountain – The Adventure of Physics available free of charge at [Link]

3-acceleration a. The magnitude of the 4-acceleration is also called the proper accelera-
tion because B2 = −a 2 if v = 0. (What is the connection between 4-acceleration and
Challenge 626 n 3-acceleration for an observer moving with the same speed as the object?) We note
that 4-acceleration lies outside the light cone, i.e. that it is a spacelike vector, and that
BU = η cd B c U d = 0, which means that the 4-acceleration is always perpendicular to the
4-velocity.* We also note that accelerations, in contrast to velocities, cannot be called re-
lativistic: the difference between b i and a i , or between their two magnitudes, does not
depend on the value of a i , but only on the value of the speed v. In other words, acceler-
ations require relativistic treatment only when the involved velocities are relativistic. If
the velocities involved are low, even the highest accelerations can be treated with Galilean
methods.
When the acceleration a is parallel to the velocity v, we get B = γ 3 a; when a is perpen-
Page 330 dicular to v, as in circular motion, we get B = γ 2 a. We will use this result below.

4-momentum
To describe motion, we also need the concept of momentum. The 4-momentum is defined
as
P = mU (153)
Copyright © Christoph Schiller November 1997–May 2006

* Similarly, the relativistic jerk or 4-jerk J of a body is defined as

J = dB
dτ = d2 U
dτ 2 . (151)

Challenge 627 ny For the relation with the 3-jerk j = da
dt we then get

γ5 (va)2 γ5 (va)2 v i
J = (J 0 , J i ) =  (jv + a 2 + 4γ 2 2 ) , γ 3 j i + 2 ((jv)v i + a 2 v i + 4γ 2 + 3(va)a i )  (152)
c c c c2

Challenge 628 ny which we will use later on. Surprisingly, J does not vanish when j vanishes. Why not?

Dvipsbugw
 rel ativistic mechanics 323

and is therefore related to the 3-momentum p by

P = (γmc, γmv) = (E
c, p) . (154)

For this reason 4-momentum is also called the energy–momentum 4-vector. In short, the
4-momentum of a body is given by mass times 4-displacement per proper time. This is the
simplest possible definition of momentum and energy. The concept was introduced by
Max Planck in 1906. The energy–momentum 4-vector, also called momenergy, like the
4-velocity, is tangent to the world line of a particle. This connection, shown in Figure 167, Dvipsbugw
follows directly from the definition, since

(E
c, p) = (γmc, γmv) = m(γc, γv) = m(dt
dτ, dx
dτ) . (155)

The (square of the) length of momenergy, namely PP = η ab P a P b , is by definition the

same for all inertial observers; it is found to be

Motion Mountain – The Adventure of Physics available free of charge at [Link]

E 2
c 2 − p2 = m 2 c 2 , (156)

thus confirming a result given above. We have already mentioned that energies or situ-
ations are called relativistic if the kinetic energy T = E − E 0 is not negligible when com-
pared to the rest energy E 0 = mc 2 . A particle whose kinetic energy is much higher than
its rest mass is called ultrarelativistic. Particles in accelerators or in cosmic rays fall into
Challenge 629 n this category. (What is their energy–momentum relation?)
In contrast to Galilean mechanics, relativity implies an absolute zero for the energy.
One cannot extract more energy than mc 2 from a system of mass m. In particular, a zero
value for potential energy is fixed in this way. In short, relativity shows that energy is
bounded from below.
Note that by the term ‘mass’ m we always mean what is sometimes called the rest mass.
This name derives from the bad habit of many science fiction and secondary-school books
of calling the product γm the relativistic mass. Workers in the field usually (but not unan-
Ref. 291 imously) reject this concept, as did Einstein himself, and they also reject the often-heard
expression that ‘(relativistic) mass increases with velocity’. Relativistic mass and energy
would then be two words for the same concept: this way to talk is at the level of the tabloid
press.
Not all Galilean energy contributes to mass. Potential energy in an outside field does
Copyright © Christoph Schiller November 1997–May 2006

not. Relativity forces us into precise energy bookkeeping. ‘Potential energy’ in relativity
is an abbreviation for ‘energy reduction of the outside field’.
Can you show that for two particles with momenta P1 and P2 , one has P1 P2 = m 1 E 2 =
Challenge 630 n M 2 E 1 = c 2 γv 12 m 1 m 2 , where v 12 is their relative velocity?

4-force
The 4-force K is defined as
K = dP
dτ = mB . (157)

Dvipsbugw
 324 ii special rel ativity • 5. speed, rest and light

Therefore force remains equal to mass times acceleration in relativity. From the definition
Ref. 290, Ref. 292 of K we deduce the relation with 3-force f = dp
dt = md(γv)
dt, namely*

mva γ dE dp fv
K = (K 0 , K i ) = (γ 4 mva
c, γ 2 ma i + γ 4 v i 2
)=( , γ ) = (γ , γf) . (158)
c c dt dt c
Challenge 632 e The 4-force, like the 4-acceleration, is orthogonal to the 4-velocity. The meaning of the
zeroth component of the 4-force can easily be discerned: it is the power required to accel-
erate the object. One has KU = c 2 dm
dτ = γ 2 (dE
dt−fv): this is the proper rate at which Dvipsbugw
the internal energy of a system increases. The product KU vanishes only for rest-mass-
conserving forces. Particle collisions that lead to reactions do not belong to this class.
In everyday life, the rest mass is preserved, and then one gets the Galilean expression
fv = dE
dt.

Rotation in relativity

Motion Mountain – The Adventure of Physics available free of charge at [Link]

If at night we turn around our own axis while looking at the sky, the stars move with a
velocity much higher than that of light. Most stars are masses, not images. Their speed
should be limited by that of light. How does this fit with special relativity?
This example helps to clarify in an-
other way what the limit velocity ac- C
tually is. Physically speaking, a rotat-
ing sky does not allow superluminal A v
energy transport, and thus does not
v
contradict the concept of a limit v'
speed. Mathematically speaking, the B v'
speed of light limits relative velocities D
only between objects that come near F I G U R E 168 On the definition of relative velocity
to each other, as shown on the left of
Figure 168. To compare velocities of distant objects is only possible if all velocities in-
volved are constant in time; this is not the case in the present example. The differential
version of the Lorentz transformations make this point particularly clear. In many gen-
eral cases, relative velocities of distant objects can be higher than the speed of light. We
Page 307 encountered one example earlier, when discussing the car in the tunnel, and we will en-
Page 337 counter a few more examples shortly.
With this clarification, we can now briefly consider rotation in relativity. The first ques-
Copyright © Christoph Schiller November 1997–May 2006

tion is how lengths and times change in a rotating frame of reference. You may want
to check that an observer in a rotating frame agrees with a non-rotating colleague on
the radius of a rotating body; however, both find that the rotating body, even if it is ri-
Challenge 633 e gid, has a circumference different from the one it had before it started rotating. Sloppily
speaking, the value of π changes for rotating observers. The ratio between the circumfer-
Challenge 634 e ence c and the radius r turns out to be c
r = 2πγ: it increases with rotation speed. This

* Some authors define 3-force as dp
dτ; then K looks slightly different. In any case, it is important to note that
in relativity, 3-force f = dp
dt is indeed proportional to 3-acceleration a; however, force and acceleration are
Challenge 631 n not parallel to each other. In fact, for rest-mass-preserving forces one finds f = γma + (fv)v
c 2 . In contrast,
in relativity 3-momentum is not proportional to 3-velocity, although it is parallel to it.

Dvipsbugw
 rel ativistic mechanics 325

Ref. 293 counter-intuitive result is often called Ehrenfest’s paradox. Among other things, it shows
that space-time for an observer on a rotating disc is not the Minkowski space-time of
special relativity.
Rotating bodies behave strangely in many ways. For ex-
ample, one gets into trouble when one tries to synchronize O3 O2 O
1
clocks mounted on a rotating circle,as shown in Figure 169 On
If one starts synchronizing the clock at O2 with that at O1 , On–1
and so on, continuing up to clock On , one finds that the last
clock is not synchronized with the first. This result reflects Dvipsbugw
the change in circumference just mentioned. In fact, a care-
ful study shows that the measurements of length and time
intervals lead all observers Ok to conclude that they live in
a rotating space-time. Rotating discs can thus be used as an
introduction to general relativity, where this curvature and F I G U R E 169 Observers on a
its effects form the central topic. More about this in the next rotating object

Motion Mountain – The Adventure of Physics available free of charge at [Link]

chapter.
Is angular velocity limited? Yes: the tangential speed in an inertial frame of reference
cannot exceed that of light. The limit thus depends on the size of the body in question.
Challenge 635 ny That leads to a neat puzzle: can one see objects rotating very rapidly?
We mention that 4-angular momentum is defined naturally as

l ab = x a pb − x b p a . (159)

In other words, 4-angular momentum is a tensor, not a vector, as shown by its two indices.
Challenge 636 ny Angular momentum is conserved in special relativity. The moment of inertia is naturally
defined as the proportionality factor between angular velocity and angular momentum.
Obviously, for a rotating particle, the rotational energy is part of the rest mass. You may
Challenge 637 ny want to calculate the fraction for the Earth and the Sun. It is not large. By the way, how
Challenge 638 ny would you determine whether a microscopic particle, too small to be seen, is rotating?
In relativity, rotation and translation combine in strange ways. Imagine a cylinder in
uniform rotation along its axis, as seen by an observer at rest. As Max von Laue has dis-
cussed, the cylinder will appear twisted to an observer moving along the rotation axis.
Challenge 639 e Can you confirm this?
Here is a last puzzle about rotation. Velocity is relative; this means that the measured
Challenge 640 ny value depends on the observer. Is this the case also for angular velocity?
Copyright © Christoph Schiller November 1997–May 2006

Wave motion
In Galilean physics, a wave is described by a wave vector and a frequency. In special
relativity, the two are combined in the wave 4-vector, given by

1 ω
L= ( , n) , (160)
λ c

Dvipsbugw
 326 ii special rel ativity • 5. speed, rest and light

where λ is the wavelength, ω the wave velocity, and n the normed direction vector. Sup-
pose an observer with 4-velocity U finds that a wave L has frequency ν. Show that

ν = LU (161)

Challenge 641 ny must be obeyed. Interestingly, the wave velocity ω transforms in a different way than
Ref. 246 particle velocity except in the case ω = c. Also the aberration formula for wave motion
Challenge 642 ny differs from that for particles, except in the case ω = c.
Dvipsbugw
The action of a free particle – how do things move?
If we want to describe relativistic motion of a free particle in terms of an extremal prin-
Page 176 ciple, we need a definition of the action. We already know that physical action is a meas-
ure of the change occurring in a system. For an inertially moving or free particle, the only
change is the ticking of its proper clock. As a result, the action of a free particle will be

Motion Mountain – The Adventure of Physics available free of charge at [Link]

proportional to the elapsed proper time. In order to get the standard unit of energy times
time, or Js, for the action, the first guess for the action of a free particle is
τ2
S = −mc 2 ∫τ1
dτ , (162)

where τ is the proper time along its path. This is indeed the correct expression. It im-
plies conservation of (relativistic) energy and momentum, as the change in proper time
Challenge 643 ny is maximal for straight-line motion with constant velocity. Can you confirm this? Indeed,
in nature, all particles move in such a way that their proper time is maximal. In other
words, we again find that in nature things change as little as possible. Nature is like a
wise old man: its motions are as slow as possible. If you prefer, every change is maximally
effective. As we mentioned before, Bertrand Russell called this the law of cosmic laziness.
The expression (162) for the action is due to Max Planck. In 1906, by exploring it in
detail, he found that the quantum of action ħ, which he had discovered together with the
Boltzmann constant, is a relativistic invariant (like the Boltzmann constant k). Can you
Challenge 644 ny imagine how he did this?
The action can also be written in more complex, seemingly more frightening ways.
These equivalent ways to write it are particularly appropriate to prepare for general re-
lativity:
Copyright © Christoph Schiller November 1997–May 2006


t2 1 τ2  dx a dx bs2
S= ∫ L dt = −mc ∫2
t1 γ
dt = −mc
τ1
∫ u a u a dτ = −mc
s1
∫
ds ds
ds , η ab
(163)
where s is some arbitrary, but monotonically increasing, function of τ, such as τ itself. As
usual, the metric η α β of special relativity is

1 0 0 0 
 0 −1 0 0 
η ab
= η ab =  . (164)
 0 0 −1 0 
 0 0 0 −1 

Dvipsbugw
 rel ativistic mechanics 327

You can easily confirm the form of the action (163) by deducing the equation of motion
Challenge 645 ny in the usual way.
In short, nature is in not a hurry: every object moves in a such way that its own clock
shows the longest delay possible, compared with any alternative motion nearby.* This
general principle is also valid for particles under the influence of gravity, as we will see
in the section on general relativity, and for particles under the influence of electric or
magnetic interactions. In fact, it is valid in all cases of (macroscopic) motion found in
nature. For the moment, we just note that the longest proper time is realized when the
Challenge 647 ny difference between kinetic and potential energy is minimal. (Can you confirm this?) For Dvipsbugw
the Galilean case, the longest proper time thus implies the smallest average difference
between the two energy types. We thus recover the principle of least action in its Galilean
formulation.
Page 176 Earlier on, we saw that the action measures the change going on in a system. Special re-
lativity shows that nature minimizes change by maximizing proper time. In nature, proper
time is always maximal. In other words, things move along paths of maximal ageing. Can

Motion Mountain – The Adventure of Physics available free of charge at [Link]

Challenge 648 ny you explain why ‘maximal ageing’ and ‘cosmic laziness’ are equivalent?
We thus again find that nature is the opposite of a Hollywood movie: nature changes
in the most economical way possible. The deeper meaning of this result is left to your
personal thinking: enjoy it!

Conformal transformations – why is the speed of light constant?

The distinction between space and time in special relativity depends on the observer. On
the other hand, all inertial observers agree on the position, shape and orientation of the
light cone at a point. Thus, in the theory of relativity, the light cones are the basic physical
‘objects’. Given the importance of light cones, we might ask if inertial observers are the
only ones that observe the same light cones. Interestingly, it turns out that other observers
do as well.
The first such category of observers are those using units of measurement in which all
time and length intervals are multiplied by a scale factor λ. The transformations among
these points of view are given by
x a  λx a (165)

and are called dilations.

A second category of additional observers are found by applying the so-called special
Copyright © Christoph Schiller November 1997–May 2006

conformal transformations. These are compositions of an inversion

xa
xa  (166)
x2

with a translation by a vector b a , namely

xa  xa + ba , (167)

* If neutrinos were massless, the action (163) would not be applicable for them. Why? Can you find an
Challenge 646 ny alternative for this (admittedly academic) case?

Dvipsbugw
 328 ii special rel ativity • 5. speed, rest and light

and a second inversion. Thus the special conformal transformations are

xa + ba x 2 xa xa
xa  or  2 + ba . (168)
1 + 2b a x a + b 2 x 2 x 2 x

These transformations are called conformal because they do not change angles of (infin-
Challenge 649 ny itesimally) small shapes, as you may want to check. They therefore leave the form (of in-
finitesimally small objects) unchanged. For example, they transform infinitesimal circles
into infinitesimal circles. They are called special because the full conformal group includes Dvipsbugw
the dilations and the inhomogeneous Lorentz transformations as well.*
Note that the way in which special conformal transformations leave light cones invari-
Challenge 651 ny ant is rather subtle.
Since dilations do not commute with time translations, there is no conserved quantity
associated with this symmetry. (The same is true of Lorentz boosts.) In contrast, rotations
and spatial translations do commute with time translations and thus do lead to conserved

Motion Mountain – The Adventure of Physics available free of charge at [Link]

quantities.
In summary, vacuum is conformally invariant – in the special sense just mentioned
– and thus also dilation invariant. This is another way to say that vacuum alone is not
sufficient to define lengths, as it does not fix a scale factor. As we would expect, matter
is necessary to do so. Indeed, (special) conformal transformations are not symmetries of
situations containing matter. Only vacuum is conformally invariant; nature as a whole is
not.
However, conformal invariance, or the invariance of light cones, is sufficient to al-
low velocity measurements. Conformal invariance is also necessary for velocity measure-
Challenge 652 ny ments, as you might want to check.
We have seen that conformal invariance implies inversion symmetry: that is, that the
large and small scales of a vacuum are related. This suggest that the constancy of the speed
of light is related to the existence of inversion symmetry. This mysterious connection
gives us a glimpse of the adventures we will encounter in the third part of our ascent of
Motion Mountain. Conformal invariance turns out to be an important property that will
lead to some incredible insights.**

Challenge 650 ny * The set of all special conformal transformations forms a group with four parameters; adding dilations
and the inhomogeneous Lorentz transformations one gets fifteen parameters for the full conformal group.
The conformal group is locally isomorphic to SU(2,2) and to the simple group SO(4,2): these concepts are
Copyright © Christoph Schiller November 1997–May 2006

Page 1194 explained in Appendix D. Note that all this is true only for four space-time dimensions; in two dimensions
– the other important case, especially in string theory – the conformal group is isomorphic to the group of
arbitrary analytic coordinate transformations, and is thus infinite-dimensional.
** The conformal group does not appear only in the kinematics of special relativity: it is the symmetry
group of all physical interactions, such as electromagnetism, provided that all the particles involved have
zero mass, as is the case for the photon. A field that has mass cannot be conformally invariant; therefore
conformal invariance is not an exact symmetry of all of nature. Can you confirm that a mass term mφ 2 in a
Challenge 653 ny Lagrangian is not conformally invariant?
However, since all particles observed up to now have masses that are many orders of magnitude smaller
than the Planck mass, it can be said that they have almost vanishing mass; conformal symmetry can then
be seen as an approximate symmetry of nature. In this view, all massive particles should be seen as small
corrections, or perturbations, of massless, i.e. conformally invariant, fields. Therefore, for the construction
of a fundamental theory, conformally invariant Lagrangians are often assumed to provide a good starting

Dvipsbugw
 accelerating observers 329

Accelerating observers
So far, we have only studied what inertial, or free-flying, observers say to each other when
they talk about the same observation. For example, we saw that moving clocks always
run slow. The story gets even more interesting when one or both of the observers are
accelerating.
One sometimes hears that special relativity cannot be used to describe accelerating
observers. That is wrong, just as it is wrong to say that Galilean physics cannot be used for
accelerating observers. Special relativity’s only limitation is that it cannot be used in non- Dvipsbugw
flat, i.e. curved, space-time. Accelerating bodies do exist in flat space-times, and therefore
they can be discussed in special relativity.
As an appetizer, let us see what an acceler-
ating, Greek, observer says about the clock
Ref. 294 of an inertial, Roman, one, and vice versa. observer (Greek) v
Assume that the Greek observer, shown in
Figure 170, moves along the path x(t), as light

Motion Mountain – The Adventure of Physics available free of charge at [Link]

c
observed by the inertial Roman one. In gen-
eral, the Roman/Greek clock rate ratio is observer (Roman)
given by ∆τ
∆t = (τ 2 − τ 1 )
(t 2 − t 1 ). Here F I G U R E 170 The simplest situation for an
the Greek coordinates are constructed with inertial and an accelerated observer
a simple procedure: take the two sets of
events defined by t = t 1 and t = t 2 , and let τ 1 and τ 2 be the points where these sets
intersect the time axis of the Greek observer.* We assume that the Greek observer is iner-
tial and moving with velocity v as observed by the Roman one. The clock ratio of a Greek
observer is then given by

∆τ dτ  1
= = 1 − v 2
c 2 = , (169)
∆t dt γv

Challenge 654 ny a formula we are now used to. We find again that moving clocks run slow.
Ref. 294 For accelerated motions, the differential version of the above reasoning is necessary.
The Roman/Greek clock rate ratio is again dτ
dt, and τ and τ + dτ are calculated in the
same way from the times t and t + dt. Assume again that the Greek observer moves along
the path x(t), as measured by the Roman one. We find directly that

τ = t − x(t)v(t)
c 2
Copyright © Christoph Schiller November 1997–May 2006

(170)

and thus
τ + dτ = (t + dt) − [x(t) − dtv(t)][v(t) + dta(t)]
c 2 . (171)

Together, these equations yield

‘dτ
dt’ = γv (1 − vv
c 2 − xa
c 2 ) . (172)

approximation.
* These sets form what mathematicians call hypersurfaces.

Dvipsbugw
 330 ii special rel ativity • 5. speed, rest and light

This shows that accelerated clocks can run fast or slow, depending on their position x and
the sign of their acceleration a. There are quotes in the above equation because we can
see directly that the Greek observer notes

‘dt
dτ’ = γv , (173)

which is not the inverse of equation (172). This difference becomes most apparent in the
simple case of two clocks with the same velocity, one of which has a constant acceleration
д towards the origin, whereas the other moves inertially. We then have Dvipsbugw

‘dτ
dt’ = 1 + дx
c 2 (174)

and
‘dt
dτ’ = 1 . (175)

Motion Mountain – The Adventure of Physics available free of charge at [Link]

We will discuss this situation shortly. But first we must clarify the concept of acceleration.

Acceleration for inertial observers

Accelerations behave differently from velocities under change of viewpoint. Let us first
take the simple case in which the object and two inertial observers all move along the
x-axis. If the Roman inertial observer measures an acceleration a = dv
dt = d2 x
dt 2 ,
and the Greek observer, also inertial, measures an acceleration α = dω
dτ = d2 ξ
dτ 2 , we
Ref. 247 get
γv3 a = γ 3ω α . (176)

This relation shows that accelerations are not Lorentz invariant, unless the velocities are
small compared to the speed of light. This is in contrast to our everyday experience, where
accelerations are independent of the speed of the observer.
Expression (176) simplifies if the accelerations are measured at a time t at which ω
vanishes – i.e. if they are measured by the so-called comoving inertial observer. In that
case the acceleration relation is given by

a c = aγv3 (177)

and the acceleration a c = α is also called proper acceleration, as its value describes what
Copyright © Christoph Schiller November 1997–May 2006

the Greek, comoving observer feels: proper acceleration describes the experience of being
pushed into the back of the accelerating seat.
In general, the observer’s speed and the acceleration are not parallel. We can calculate
Ref. 295 how the value of 3-acceleration a measured by a general inertial observer is related to the
value ac measured by the comoving observer using expressions (150) and (148). We get
the generalization of (177):
vac = vaγv3 (178)

Dvipsbugw
 accelerating observers 331

and
1 (1 − γv )(vac )v γv (vac )v
a= 2
ac − −  . (179)
γv v2 c2

Squaring yields the relation

1 (ac v)2
a2 = a 2
−  (180)
γv4 c c2
Dvipsbugw
Page 321 which we know already in a slightly different form. It shows (again) that the comoving
or proper 3-acceleration is always larger than the 3-acceleration measured by an outside
inertial observer. The faster the outside inertial observer is moving, the smaller the accele-
Challenge 655 e ration he observes. Acceleration is not a relativistic invariant. The expression also shows
that whenever the speed is perpendicular to the acceleration, a boost yields a factor γv2 ,
whereas a speed parallel to the acceleration gives the already mentioned γv3 dependence.

Motion Mountain – The Adventure of Physics available free of charge at [Link]

We see that acceleration complicates many issues, and it requires a deeper investiga-
tion. To keep matters simple, from now on we only study constant accelerations. Interest-
Page 476 ingly, this situation serves also as a good introduction to black holes and, as we will see
shortly, to the universe as a whole.

Accelerating frames of reference

How do we check whether we live in an inertial frame of reference? Let us first define the
term. An inertial frame (of reference) has two defining properties. First, lengths and dis-
tances measured with a ruler are described by Euclidean geometry. In other words, rulers
behave as they do in daily life. In particular, distances found by counting how many rulers
(rods) have to be laid down end to end to reach from one point to another – the so-called
rod distances – behave as in everyday life. For example, they obey Pythagoras’ theorem in
the case of right-angled triangles. Secondly, the speed of light is constant. In other words,
any two observers in that frame, independent of their time and of the position, make the
following observation: the ratio c between twice the rod distance between two points and
the time taken by light to travel from one point to the other and back is always the same.
Equivalently, an inertial frame is one for which all clocks always remain synchronized
and whose geometry is Euclidean. In particular, in an inertial frame all observers at fixed
coordinates always remain at rest with respect to each other. This last condition is, how-
Copyright © Christoph Schiller November 1997–May 2006

ever, a more general one. There are other, non-inertial, situations where this is still the
case.
Non-inertial frames, or accelerating frames, are a useful concept in special relativity. In
fact, we all live in such a frame. We can use special relativity to describe it in the same
way that we used Galilean physics to describe it at the beginning of our journey.
A general frame of reference is a continuous set of observers remaining at rest with
respect to each other. Here, ‘at rest with respect to each other’ means that the time for a
light signal to go from one observer to another and back again is constant over time, or
equivalently, that the rod distance between the two observers is constant. Any frame of
reference can therefore also be called a rigid collection of observers. We therefore note
that a general frame of reference is not the same as a set of coordinates; the latter is usu-

Dvipsbugw
 332 ii special rel ativity • 5. speed, rest and light

τ
t

on
II

riz
ξ

ho
re
tu
fu
Ω

O
III c2/g x
I

pa
Dvipsbugw

st
ho
IV

riz
on
F I G U R E 171 The hyperbolic motion of an
rectilinearly, uniformly accelerating observer Ω

Motion Mountain – The Adventure of Physics available free of charge at [Link]

ally not rigid. If all the rigidly connected observers have constant coordinate values, we
speak of a rigid coordinate system. Obviously, these are the most useful when it comes to
describing accelerating frames of reference.*
Ref. 297 Note that if two observers both move with a velocity v, as measured in some inertial
frame, they observe that they are at rest with respect to each other only if this velocity
Challenge 656 ny is constant. Again we find, as above, that two people tied to each other by a rope, and at
a distance such that the rope is under tension, will see the rope break (or hang loose) if
they accelerate together to (or decelerate from) relativistic speeds in precisely the same
way. Relativistic acceleration requires careful thinking.
An observer who always feels the same force on his body is called uniformly accelerat-
ing. More precisely, a uniformly accelerating observer is an observer whose acceleration
at every moment, measured by the inertial frame with respect to which the observer is
at rest at that moment, always has the same value B. It is important to note that uniform
acceleration is not uniformly accelerating when always observed from the same inertial
frame. This is an important difference from the Galilean case.
For uniformly accelerated motion in the sense just defined, we need

B ċ B = −д 2 (181)
Copyright © Christoph Schiller November 1997–May 2006

Ref. 298 where д is a constant independent of t. The simplest case is uniformly accelerating motion
that is also rectilinear, i.e. for which the acceleration a is parallel to v at one instant of time
Challenge 657 ny and (therefore) for all other times as well. In this case we can write, using 3-vectors,

Ref. 296 * There are essentially only two other types of rigid coordinate frames, apart from the inertial frames:

— The frame ds 2 = dx 2 + dy 2 + dz 2 − c 2 dt 2 (1 + дk x k
c 2 )2 with arbitrary, but constant, acceleration of the

origin. The acceleration is a = −g(1 + gx
c 2 ).
— The uniformly rotating frame ds 2 = dx 2 + dy 2 + dz 2 + 2ω(−y dx + x dy)dt − (1 − r 2 ω 2
c 2 )dt. Here the
z-axis is the rotation axis, and r 2 = x 2 + y 2 .

Dvipsbugw
 accelerating observers 333

dγv
γ3a = g or =g. (182)
dt

Taking the direction we are talking about to be the x-axis, and solving for v(t), we get

дt
v= , (183)
д2 t2
1 + c2
Dvipsbugw
where it was assumed that v(0) = 0. We note that for small times we get v = дt and for
large times v = c, both as expected. The momentum of the accelerated observer increases
Challenge 658 ny linearly with time, again as expected. Integrating, we find that the accelerated observer
moves along the path 
c2 д2 t 2
x(t) = 1+ 2 , (184)
д c

Motion Mountain – The Adventure of Physics available free of charge at [Link]

where it is assumed that x(0) = c 2
д, in order to keep the expression simple. Because of
this result, visualized in Figure 171, a rectilinearly and uniformly accelerating observer is
said to undergo hyperbolic motion. For small times, the world-line reduces to the usual
x = дt 2
2 + x 0 , whereas for large times it is x = ct, as expected. The motion is thus
uniformly accelerated only for the moving body itself, not for an outside observer.
The proper time τ of the accelerated observer is related to the time t of the inertial
frame in the usual way by dt = γdτ. Using the expression for the velocity v(t) of equation
Ref. 298, Ref. 299 (183) we get*
c дτ c2 дτ
t = sinh and x = cosh (185)
д c д c

for the relationship between proper time τ and the time t and position x measured by
the external, inertial Roman observer. We will encounter this relation again during our
study of black holes.
Does all this sound boring? Just imagine accelerating on a motorbike at д = 10 m
s2
for the proper time τ of 25 years. That would bring you beyond the end of the known
universe! Isn’t that worth a try? Unfortunately, neither motorbikes nor missiles that ac-
Challenge 659 n celerate like this exist, as their fuel tanks would have to be enormous. Can you confirm
this?
Copyright © Christoph Schiller November 1997–May 2006

Ref. 300 * Use your favourite mathematical formula collection – every student should have one – to deduce this. The
hyperbolic sine and thehyperbolic cosine are defined by sinh y = (e
y
− e−y )
2 and cosh y = (e y + e−y )
2.
They imply that ∫ dy
y + a = arsinh y
a = Arsh y
a = ln(y + y 2 + a 2 ).
2 2

Dvipsbugw
 334 ii special rel ativity • 5. speed, rest and light

For uniform acceleration, the coordinates transform as

c ξ дτ
t = ( + ) sinh
д c c
c2 дτ
x = ( + ξ) cosh
д c
y=υ
z=ζ, (186) Dvipsbugw

where τ now is the time coordinate in the Greek frame. We note also that the space-time
interval dσ satisfies

dσ 2 = (1 + дξ
c 2 )2 c 2 dτ 2 − dξ 2 − dυ 2 − dζ 2 = c 2 dt 2 − dx 2 − dy 2 − dz 2 , (187)

Motion Mountain – The Adventure of Physics available free of charge at [Link]

and since for dτ = 0 distances are given by Pythagoras’ theorem, the Greek reference
Ref. 301 frame is indeed rigid.
After this forest of formulae, let’s tackle a simple question, shown in Figure 171. The
inertial, Roman observer O sees the Greek observer Ω departing with acceleration д,
moving further and further away, following equation (184). What does the Greek observer
say about his Roman colleague? With all the knowledge we have now, that is easy to
answer to answer. At each point of his trajectory Ω sees that O has the coordinate τ = 0
Challenge 660 e (can you confirm this?), which means that the distance to the Roman observer, as seen
by Greek one, is the same as the space-time interval OΩ. Using expression (184), we see
Ref. 302 that this is  
d OΩ = ξ 2 = x 2 − c 2 t 2 = c 2
д , (188)

which, surprisingly enough, is constant in time! In other words, the Greek observer will
observe that he stays at a constant distance from the Roman one, in complete contrast to
what the Roman observer says. Take your time to check this strange result in some other
way. We will need it again later on, to explain why the Earth does not explode. (Can you
Challenge 661 n guess how that is related to this result?)
The composition theorem for accelerations is more complex than for velocities. The
Ref. 303 best explanation of this was published by Mishra. If we call a nm the acceleration of sys-
tem n by observer m, we are seeking to express the object acceleration a 01 as function
Copyright © Christoph Schiller November 1997–May 2006

of the value a 02 measured by the other observer, the relative acceleration a 12 , and the
proper acceleration a 22 of the other observer: see Figure 172. Here we will only study
one-dimensional situations, where all observers and all objects move along one axis. (For
clarity, we also write v 11 = v and v 02 = u.) In Galilean physics we have the general connec-
Challenge 662 e tion
a 01 = a 02 − a 12 + a 22 (189)

Dvipsbugw
 accelerating observers 335

a11 : proper acceleration

v11 = 0

y
Observer 1
x
a22 : proper acceleration
v22 = 0
Dvipsbugw
v0n : object speed seen by observer n
Observer 2
x a0n : object acceleration
Object seen by observer n

F I G U R E 172 The definitions necessary to deduce the composition behaviour of

Motion Mountain – The Adventure of Physics available free of charge at [Link]

accelerations

because accelerations behave simply. In special relativity, one gets

(1 − v 2
c 2 )32 (1 − u 2
c 2 )(1 − v 2
c 2 )−12 (1 − u 2
c 2 )(1 − v 2
c 2 )32

a 01 = a 02 − a 12 + a 22
(1 − uv
c 2 )3 (1 − uv
c 2 )2 (1 − uv
c 2 )3
(190)
Challenge 663 ny and you might enjoy checking the expression.
Page 312 Can you state how the acceleration ratio enters into the definition of mass in special
Challenge 664 ny relativity?

Event horizons
There are many surprising properties of accelerated motion. Of special interest is the
trajectory, in the coordinates ξ and τ of the rigidly accelerated frame, of an object located
Challenge 665 ny at the departure point x = x 0 = c 2
д at all times t. One gets the two relations*

c2 дτ
ξ=− (1 − sech )
д c
дτ дτ
dξ
dτ = −c sech
Copyright © Christoph Schiller November 1997–May 2006

tanh . (192)
c c

These equations are strange. For large times τ the coordinate ξ approaches the limit value
−c 2
д and dξ
dτ approaches zero. The situation is similar to that of a car accelerating
away from a woman standing on a long road. Seen from the car, the woman moves away;

* The functions appearing above, the hyperbolic secant and the hyperbolic tangent, are defined using the
expressions from the footnote on page 333:

1 sinh y
sech y = and tanh y = . (191)
cosh y cosh y

Dvipsbugw
 336 ii special rel ativity • 5. speed, rest and light

τ
t

on
II

riz
ξ

ho
re
tu
fu
Ω

O
III c2/g x
I

pa
Dvipsbugw

st
ho
IV

riz
on
F I G U R E 173 Hyperbolic motion and event
horizons

Motion Mountain – The Adventure of Physics available free of charge at [Link]

however, after a while, the only thing one notices is that she is slowly approaching the
horizon. In Galilean physics, both the car driver and the woman on the road see the
other person approaching their horizon; in special relativity, only the accelerated observer
makes this observation.
A graph of the situation helps to clarify the result. In Figure 173 we can see that light
emitted from any event in regions II and III cannot reach the Greek observer. Those
events are hidden from him and cannot be observed. Strangely enough, however, light
from the Greek observer can reach region II. The boundary between the part of space-
time that can be observed and the part that cannot is called the event horizon. In relativity,
event horizons act like one-way gates for light and other signals. For completeness, the
Challenge 666 ny graph also shows the past event horizon. Can you confirm that event horizons are black?
So, not all events observed in an inertial frame of reference can be observed in a uni-
formly accelerating frame of reference. Uniformly accelerating frames of reference pro-
duce event horizons at a distance −c 2
д. For example, a person who is standing can never
see further than this distance below his feet.
By the way, is it true that a light beam cannot catch up with an observer in hyperbolic
Challenge 667 n motion, if the observer has a sufficient headstart?
Here is a more advanced challenge, which prepares us for general relativity. What is
Copyright © Christoph Schiller November 1997–May 2006

Challenge 668 n the shape of the horizon seen by a uniformly accelerated observer?

Acceleration changes colours

We saw earlier that a moving receiver sees different colours from the sender. So far, we
discussed this colour shift, or Doppler effect, for inertial motion only. For accelerating
frames the situation is even stranger: sender and receiver do not agree on colours even
Ref. 298, Ref. 304 if they are at rest with respect to each other. Indeed, if light is emitted in the direction of
the acceleration, the formula for the space-time interval gives

д0 x 2 2 2
dσ 2 = 1 +  c dt (193)
c2

Dvipsbugw
 accelerating observers 337

in which д0 is the proper acceleration of an observer located at x = 0. We can deduce in

Challenge 669 ny a straightforward way that
fr дr h 1
=1− 2 = (194)
fs c 1 + дs2h  c

where h is the rod distance between the source and the receiver, and where дs = д0
(1 +
д0 x s
c 2 ) and дr = д0
(1 + дo x r
c 2 ) are the proper accelerations measured at the source
and at the detector. In short, the frequency of light decreases when light moves in the Dvipsbugw
direction of acceleration. By the way, does this have an effect on the colour of trees along
Challenge 670 n their vertical extension?
The formula usually given, namely

fr дh
=1− 2 , (195)
fs c

Motion Mountain – The Adventure of Physics available free of charge at [Link]

is only correct to a first approximation. In accelerated frames of reference, we have to
be careful about the meaning of every quantity. For everyday accelerations, however, the
Challenge 671 ny differences between the two formulae are negligible. Can you confirm this?

Can light move faster than c?

What speed of light does an accelerating observer measure? Using expression (195) above,
an accelerated observer deduces that

дh
v light = c (1 + ) (196)
c2
which is higher than c for light moving in front of or ‘above’ him, and lower than c for
light moving behind or ‘below’ him. This strange result follows from a basic property of
any accelerating frame of reference. In such a frame, even though all observers are at rest
with respect to each other, clocks do not remain synchronized. This change of the speed
of light has also been confirmed by experiment.* Thus, the speed of light is only constant
when it is defined as c = dx
dt, and if dx and dt are measured with a ruler located at
a point inside the interval dx and a clock read off during the interval dt. If the speed
of light is defined as ∆x
∆t, or if the ruler defining distances or the clock measuring
Copyright © Christoph Schiller November 1997–May 2006

times is located away from the propagating light, the speed of light is different from c for
accelerating observers! This is the same effect you can experience when you turn around
your vertcial axis at night: the star velocities you observe are much higher than the speed
of light.
Note that this result does not imply that signals or energy can be moved faster than c.
Challenge 672 n You may want to check this for yourself.
In fact, all these effects are negligible for distances l that are much less than c 2
a. For
an acceleration of 9.5 m
s2 (about that of free fall), distances would have to be of the order

Page 410 * The propagation delays to be discussed in the chapter on general relativity can be seen as confirmations of
this effect.

Dvipsbugw
 338 ii special rel ativity • 5. speed, rest and light

time
clock 1 clock 2

t3

t2

Dvipsbugw

t1

space

F I G U R E 174 Clocks and the

Motion Mountain – The Adventure of Physics available free of charge at [Link]

measurement of the speed of
light as two-way velocity

of one light year, or 9.5 ċ 1012 km, in order for any sizable effects to be observed. In short,
c is the speed of light relative to nearby matter only.
By the way, everyday gravity is equivalent to a constant acceleration. So, why then
Challenge 673 n distant objects, such as stars, move faster than light, following expression (196)?

What is the speed of light?

We have seen that the speed of light, as usually defined, is given by c only if either the
observer is inertial or the observer measures the speed of light passing nearby (rather
than light passing at a distance). In short, the speed of light has to be measured locally.
But this condition does not eliminate all subtleties.
An additional point is often forgotten. Usually, length is measured by the time it takes
light to travel. In such a case the speed of light will obviously be constant. But how does
one check the constancy? One needs to eliminate length measurements. The simplest way
to do this is to reflect light from a mirror, as shown in Figure 174. The constancy of the
speed of light implies that if light goes up and down a short straight line, then the clocks
at the two ends measure times given by
Copyright © Christoph Schiller November 1997–May 2006

t 3 − t 1 = 2 (t 2 − t 1 ) . (197)

Here it is assumed that the clocks have been synchronised according to the prescription
on page 299. If the factor were not exactly two, the speed of light would not be constant.
In fact, all experiments so far have yielded a factor of two, within measurement errors.*

* The subtleties of the one-way and two-way speed of light will remain a point of discussion for a long time.
Many experiments are explained and discussed in Ref. 252. Zhang says in his summary on page 171, that the
one-way velocity of light is indeed independent of the light source; however, no experiment really shows that
Ref. 305 it is equal to the two-way velocity. Moreover, most so called ‘one-way’ experiments are in fact still ‘two-way’

Dvipsbugw
 accelerating observers 339

This result is sometimes expressed by saying that it is impossible to measure the one-
Challenge 674 n way velocity of light; only the two-way velocity of light is measurable. Do you agree?

Limits on the length of solid bodies

An everyday solid object breaks when some part of it moves with respect to some other
part with more than the speed of sound c of the material.* For example, when an object
hits the floor and its front end is stopped within a distance d, the object breaks at the
latest when Dvipsbugw
v 2 2d
 . (198)
c2 l
In this way, we see that we can avoid the breaking of fragile objects by packing them into
foam rubber – which increases the stopping distance – of roughly the same thickness as
the object’s size. This may explain why boxes containing presents are usually so much
larger than their contents!

Motion Mountain – The Adventure of Physics available free of charge at [Link]

The fracture limit can also be written in a different way. To avoid breaking, the accele-
ration a of a solid body with length l must obey

l a < c2 , (199)

where c is the speed of sound, which is the speed limit for the material parts of solids. Let
Ref. 306 us now repeat the argument in relativity, using the speed of light instead of that of sound.
Imagine accelerating the front of a solid body with some proper acceleration a. The back
end cannot move with an acceleration α equal or larger than infinity, or if one prefers, it
Challenge 675 n cannot move with more than the speed of light. A quick check shows that therefore the
length l of a solid body must obey

l α < c 2
2 , (200)

where c is now the speed of light. The speed of light thus limits the size of solid bodies. For
example, for 9.8 m
s2 , the acceleration of good motorbike, this expression gives a length
limit of 9.2 Pm, about a light year. Not a big restriction: most motorbikes are shorter.
However, there are other, more interesting situations. The highest accelerations achiev-
able today are produced in particle accelerators. Atomic nuclei have a size of a few femo-
Challenge 676 ny tometres. Can you deduce at which energies they break when smashed together in an
Copyright © Christoph Schiller November 1997–May 2006

accelerator? In fact, inside a nucleus, the nucleons move with accelerations of the order
of v 2
r  ħ 2
m 2 r 3  1031 m
s2 ; this is one of the highest values found in nature.
Note that Galilean physics and relativity produce a similar conclusion: a limiting speed,
be it that of sound or that of light, makes it impossible for solid bodies to be rigid. When
we push one end of a body, the other end always moves a little bit later.
What does this mean for the size of elementary particles? Take two electrons a distance
d apart, and call their size l. The acceleration due to electrostatic repulsion then leads to

experiments (see his page 150).

* The (longitudinal) speed of sound is about 5.9 km
s for glass, iron or steel; about 4.5 km
s for gold; and
about 2 km
s for lead. Other sound speeds are given on page 206.

Dvipsbugw
 340 ii special rel ativity • 5. speed, rest and light

Challenge 677 ny an upper limit for their size given by

4πε 0 c 2 d 2 m
l< . (201)
e2
The nearer electrons can get, the smaller they must be. The present experimental limit
gives a size smaller than 10−19 m. Can electrons be exactly point-like? We will come back
to this question during our study of general relativity and quantum theory.
Dvipsbugw
Special rel ativity in four sentences
This section of our ascent of Motion Mountain can be quickly summarized.
— All (free floating) observers find that there is a unique, perfect velocity in nature,
namely a common maximum energy velocity, which is realized by massless radiation
such as light or radio signals, but cannot be achieved by material systems.

Motion Mountain – The Adventure of Physics available free of charge at [Link]

— Therefore, even though space-time is the same for every observer, times and lengths
vary from one observer to another, as described by the Lorentz transformations (112)
and (113), and as confirmed by experiment.
— Collisions show that a maximum speed implies that mass is concentrated energy, and
that the total energy of a body is given by E = γmc 2 , as again confirmed by experiment.
— Applied to accelerated objects, these results lead to numerous counter-intuitive con-
sequences, such as the twin paradox, the appearance of event horizons and the appear-
ance of short-lived tachyons in collisions.
Special relativity shows that motion, though limited in speed, is relative, defined using
the propagation of light, conserved, reversible and deterministic.

Could the speed of light vary?

The speed of massless light is the limit speed. Assuming that all light is indeed massless,
could the speed of light still change from place to place, or as time goes by? This tricky
question still makes a fool out of many physicists. The first answer is usually a loud: ‘Yes,
of course! Just look at what happens when the value of c is changed in formulae.’ (In fact,
there have even been attempts to build ‘variable speed of light theories’.) However, this
often-heard statement is wrong.
Since the speed of light enters into our definition of time and space, it thus enters,
Copyright © Christoph Schiller November 1997–May 2006

even if we do not notice it, into the construction of all rulers, all measurement standards
and all measuring instruments. Therefore there is no way to detect whether the value
actually varies. No imaginable experiment could detect a variation of the limit speed, as
Challenge 678 n the limit speed is the basis for all measurements. ‘That is intellectual cruelty!’, you might
say. ‘All experiments show that the speed of light is invariant; we had to swallow one
counter-intuitive result after another to accept the constancy of the speed of light, and
now we are supposed to admit that there is no other choice?’ Yes, we are. That is the irony
of progress in physics. The observer-invariance of the speed of light is counter-intuitive
and astonishing when compared to the lack of observer-invariance at everyday, Galilean
speeds. But had we taken into account that every speed measurement is – whether we like
it or not – a comparison with the speed of light, we would not have been astonished by

Dvipsbugw
special rel ativity in four sentences 341

the constancy of the speed of light; rather, we would have been astonished by the strange
properties of small speeds.
In short, there is in principle no way to check the invariance of a standard. To put it
another way, the truly surprising aspect of relativity is not the invariance of c; it is the
disappearance of c from the formulae of everyday motion.

What happens near the speed of light?

As one approaches the speed of light, the quantities in the Lorentz transformation diverge. Dvipsbugw
A division by zero is impossible: indeed, neither masses nor observers can move at the
speed of light. However, this is only half the story.
No observable actually diverges in nature. Approaching the speed of light as nearly
as possible, even special relativity breaks down. At extremely large Lorentz contractions,
there is no way to ignore the curvature of space-time; indeed, gravitation has to be taken
into account in those cases. Near horizons, there is no way to ignore the fluctuations of

Motion Mountain – The Adventure of Physics available free of charge at [Link]

speed and position; quantum theory has to be taken into account there. The exploration
of these two limitations define the next two stages of our ascent of Motion Mountain.
At the start of our adventure, during our exploration of Galilean physics, once we had
defined the basic concepts of velocity, space and time, we turned our attention to gravit-
ation. The invariance of the speed of light has forced us to change these basic concepts.
We now return to the study of gravitation in the light of this invariance.

Copyright © Christoph Schiller November 1997–May 2006

Dvipsbugw
342 ii special rel ativity

Biblio graphy

231 Aristotle, On sense and the sensible, section 1, part 1, 350 bce. Cited in Jean-Paul
Dumont, Les écoles présocratiques, Folio Essais, Gallimard, p. 157, 1991. Cited on page 276.
232 The history of the measurement of the speed of light can be found in chapter 19 of the text
by Francis A. Jenkins & Harvey E. White, Fundamentals of Optics, McGraw-Hill,
New York, 1957. Cited on page 276.
233 On the way to perform such measurements, see Sydney G. Brewer, Do-it-yourself As- Dvipsbugw
tronomy, Edinburgh University Press, 1988. Kepler himself never measured the distances of
planets to the Sun, but only ratios of planetary distances. The parallax of the Sun from two
points of the Earth is at most 8.79

; it was first measured in the eighteenth century. Cited
on page 278.
234 Aristarchos, On the sizes and the distances of the Sun and the Moon, c. 280 bce, in
Michael J. Crowe, Theories of the World From Antiquity to the Copernican Revolution,

Motion Mountain – The Adventure of Physics available free of charge at [Link]

Dover, 1990. Cited on page 278.
235 J. Frercks, Creativity and technology in experimentation: Fizeau’s terrestrial determina-
tion of the speed of light, Centaurus 42, pp. 249–287, 2000. See also the beautiful website on
reconstrutions of historical science experiments at [Link]
forschung/nachbauten. Cited on page 278.
236 The way to make pictures of light pulses with an ordinary photographic camera, without
any electronics, is described by M.A. Duguay & A.T. Mattick, Ultrahigh speed pho-
tography of picosecond light pulses and echoes, Applied Optics 10, pp. 2162–2170, 1971. The
picture on page 279 is taken from it. Cited on page 279.
237 You can learn the basics of special relativity with the help of the web, without referring to
any book, using the [Link] web page
as a starting point. This page mentions most of the English-language relativity resources
available on the web. Links in other languages can be found with search engines. Cited on
page 280.
238 Observations of gamma ray bursts show that the speed of light does not depend on the
lamp speed to within one part in 1020 , as shown by K. Brecher, Bulletin of the American
Physical Society 45, 2000. He assumed that both sides of the burster emit light. The large
speed difference and the pulse sharpness then yield this result. See also his older paper K.
Brecher, Is the speed of light independent of the source?, Physics Letters 39, pp. 1051–
1054, Errata 1236, 1977. Measuring the light speed from rapidly moving stars is another way.
Some of these experiments are not completely watertight, however. There is a competing
Copyright © Christoph Schiller November 1997–May 2006

theory of electrodynamics, due to Ritz, which maintains that the speed of light is c only
when measured with respect to the source; The light from stars, however, passes through
the atmosphere, and its speed might thus be reduced to c.
The famous experiment with light emitted from rapid pions at CERN is not subject to this
criticism. It is described in T. Alväger, J.M. Bailey, F.J.M. Farley, J. Kjellman &
I. Wallin, Test of the second postulate of relativity in the GeV region, Physics Letters 12,
pp. 260–262, 1964. See also T. Alväger & al., Velocity of high-energy gamma rays, Arkiv
för Fysik 31, pp. 145–157, 1965.
Another precise experiment at extreme speeds is described by G.R. Kalbfleisch, N.
Baggett, E.C. Fowler & J. Alspector, Experimental comparison of neutrino, anti-
neutrino, and muon velocities, Physical Review Letters 43, pp. 1361–1364, 1979. Cited on
page 280.

Dvipsbugw
bibliography 343

239 See e.g. C. Will, Theory and Experiment in Gravitational Physics, Revised edition, Cam-
bridge University Press, 1993. Cited on pages 281 and 282.
240 B.E. Schaefer, Severe limits on variations of the speed of light with frequency, Physical
Review Letters 82, pp. 4964–4966, 21 June 1999. Cited on page 281.
241 The beginning of the modern theory of relativity is the famous paper by Albert Ein-
stein, Zur Elektrodynamik bewegter Körper, Annalen der Physik 17, pp. 891–921, 1905. It
still well worth reading, and every physicist should have done so. The same can be said of the
famous paper, probably written after he heard of Olinto De Pretto’s idea, found in Albert
Einstein, Ist die Trägheit eines Körpers von seinem Energieinhalt abhängig?, Annalen der
Dvipsbugw
Physik 18, pp. 639–641, 1905. See also the review Albert Einstein, Über das Relativität-
sprinzip und die aus demselben gezogenen Folgerungen, Jahrbuch der Radioaktivität und
Elektronik 4, pp. 411–462, 1907. These papers are now available in many languages. A later,
unpublished review is available in facsimile and with an English translation as Albert
Einstein, Hanoch Gutfreund, ed., Einstein’s 1912 Manuscript on the Theory of Relativity,
George Braziller, 2004. Cited on pages 281, 282, and 318.

Motion Mountain – The Adventure of Physics available free of charge at [Link]

242 Albert Einstein, Mein Weltbild, edited by Carl Selig, Ullstein Verlag, 1998. Cited
on page 281.
243 Jean van Bl adel, Relativity and Engineering, Springer, 1984. Cited on page 281.
244 Albrecht Fölsing, Albert Einstein – eine Biographie, Suhrkamp p. 237, 1993. Cited on
pages 281 and 291.
245 R.J. Kennedy & E.M. Thorndike, Experimental establishment of the relativity of time,
Physical Review 42, pp. 400–418, 1932. See also H.E. Ives & G.R. Stilwell, An experi-
mental study of the rate of a moving atomic clock, Journal of the Optical Society of America
28, pp. 215–226, 1938, and 31, pp. 369–374, 1941. For a modern, high-precision versions, see
C. Braxmeier, H. Müller, O. Pradl, J. Mlynek, A. Peters & S. Schiller, New
tests of relativity using a cryogenic optical resonator, Physical Review Letters 88, p. 010401,
2002. The newest result is in P. Antonini, M. Okhapkin, E. Göklü & S. Schiller,
Testing the constancy of the speed of light with rotating cryogenic optical resonators, Phys-
ical Review A 71, p. 050101, 2005, or [Link] Cited on page
282.
246 Edwin F. Taylor & John A. Wheeler, Spacetime Physics – Introduction to Special
Relativity, second edition, Freeman, 1992. See also Nick M.J. Woodhouse, Special Re-
lativity, Springer, 2003. Cited on pages 282 and 326.
247 Wolf gang R indler, Relativity – Special, General and Cosmological, Oxford University
Press, 2001. A beautiful book by one of the masters of the field. Cited on pages 282 and 330.
Copyright © Christoph Schiller November 1997–May 2006

248 The slowness of the speed of light inside stars is due to the frequent scattering of photons
by the star matter. The most common estimate for the Sun is an escape time of 40 000 to 1
million years, but estimates between 17 000 years and 50 million years can be found in the
literature. Cited on page 283.
249 L. Vestergaard Hau, S.E. Harris, Z. Du tton & C.H. Behroozi, Light speed
reduction to 17 meters per second in an ultracold atomic gas, Nature 397, pp. 594–598, 1999.
See also Ref. 25. Cited on page 283.
250 The method of explaining special relativity by drawing a few lines on paper is due to Her-
mann B ondi, Relativity and Common Sense: A New Approach to Einstein, Dover, New
York, 1980. See also Dierck-Ekkehard Liebscher, Relativitätstheorie mit Zirkel und
Lineal, Akademie-Verlag Berlin, 1991. Cited on page 283.

Dvipsbugw
344 ii special rel ativity

251 Rod S. L akes, Experimental limits on the photon mass and cosmic vector potential, Phys-
ical Review Letters 80, pp. 1826–1829, 1998. The speed of light is independent of frequency
within a factor of 6ċ10−21 , as was shown from gamma ray studies by B.E. Schaefer, Severe
limits on variations of the speed of light with frequency, Physical Review Letters 82, pp. 4964–
4966, 1999. Cited on page 284.
252 An overview of experimental results is given in Yuan Zhong Zhang, Special Relativity
and its Experimental Foundations, World Scientific, 1998. Cited on pages 284, 290, 299, 313,
338, and 345.
253 R.W. McGowan & D.M. Giltner, New measurement of the relativistic Doppler shift Dvipsbugw
in neon, Physical Review Letters 70, pp. 251–254, 1993. Cited on page 286.
254 The present record for clock synchronization seems to be 1 ps for two clocks distant 3 km
from each other. See A. Valencia, G. Scarcelli & Y. Shih, Distant clock synchron-
ization using entangled photon pairs, Applied Physics Letters 85, pp. 2655–2657, 2004, or
[Link] Cited on page 287.
255 J. Frenkel & T. Kontorowa, Über die Theorie der plastischen Verformung, Physikalis-

Motion Mountain – The Adventure of Physics available free of charge at [Link]

che Zeitschrift der Sowietunion 13, p. 1, 1938. F.C. Frank, On the equations of motion of
crystal dislocations, Proceedings of the Physical Society A 62, pp. 131–134, 1949. J. Eshelby,
Uniformly moving dislocations, Proceedings of the Physical Society A 62, pp. 307–314, 1949.
See also G. Leibfried & H. Dietze, Zeitschrift für Physik 126, p. 790, 1949. A general in-
troduction can be found in A. Seeger & P. Schiller, Kinks in dislocation lines and their
effects in internal friction in crystals, Physical Acoustics 3A, W.P. Mason, ed., Academic
Press, 1966. See also the textbooks by Frank R.N. Nabarro, Theory of Crystal Dislo-
cations, Oxford University Press, 1967, or J.P. Hirth & J. Lothe, Theory of Dislocations,
McGraw Hill, 1968. Cited on page 287.
256 This beautiful graph is taken from Z.G.T. Guiragossian, G.B. Rothbart, M.R.
Yearian, R. Gearhart & J.J. Murray, Relative velocity measurements of electrons
and gamma rays at 15 GeV, Physical Review Letters 34, pp. 335–338, 1975. Cited on page
287.
257 To find out more about the best-known crackpots, and their ideas, send an email to
majordomo@[Link] with the one-line body ‘subscribe psychoceramics’. Cited on page
288.
258 The speed of neutrinos is the same as that of light to 9 decimal digits. This is explained
by Leo Stod olsky, The speed of light and the speed of neutrinos, Physics Letters B 201,
p. 353, 1988. An observation of a small mass for the neutrino has been published by the Ja-
panese Super-Kamiokande collaboration, in Y. Fukuda & al., Evidence for oscillation of
atmospheric neutrinos, Physical Review Letters 81, pp. 1562–1567, 1998. The newer results
Copyright © Christoph Schiller November 1997–May 2006

published by the Canadian Sudbury Neutrino Observatory, as Q.R. Ahmad & al., Direct
evidence for neutrino flavor transformation from neutral-current interactions in the Sud-
bury Neutrino Observatory, Physical Review Letters 89, p. 011301, 2002, also confirm that
neutrinos have a mass in the 1 eV region. Cited on pages 289 and 1258.
259 B. Rothenstein & G. Eckstein, Lorentz transformations directly from the speed of
light, American Journal of Physics 63, p. 1150, 1995. See also the comment by E. Kapuścik,
Comment on “Lorentz transformations directly from the speed of light,” by B. Rothenstein
and G. Eckstein, American Journal of Physics 65, p. 1210, 1997. Cited on page 290.
260 See e.g. the 1922 lectures by Lorentz at Caltech, published as H.A. Lorentz, Problems of
Modern Physics, edited by H. Bateman, Ginn and Company, page 99, 1927. Cited on page
290.

Dvipsbugw
bibliography 345

261 A.A. Michelson & E.W. Morley, On the relative motion of the Earth and the lumini-
ferous ether, American Journal of Science (3rd series) 34, pp. 333–345, 1887. Michelson pub-
lished many other papers on the topic after this one. Cited on page 290.
262 S. Stephan, P. Antonini & M. Okhapkin, A precision test of the isotropy of
the speed of light using rotating cryogenic resonators, [Link]
0510169. Cited on page 290.
263 H.A. Lorentz, De relative beweging van de aarde en dem aether, Amst. Versl. 1, p. 74,
1892, and also H.A. Lorentz, Electromagnetic phenomena in a system moving with any
velocity smaller than that of light, Amst. Proc. 6, p. 809, 1904, or Amst. Versl. 12, p. 986, 1904. Dvipsbugw
Cited on page 294.
264 A general refutation of such proposals is discussed by S.R. Mainwaring & G.E. Sted-
man, Accelerated clock principles, Physical Review A 47, pp. 3611–3619, 1993. Experiments
on muons at CERN in 1968 showed that accelerations of up to 1020 m
s2 have no effect, as
explained by D.H. Perkins, Introduction to High Energy Physics, Addison-Wesley, 1972,
or by J. Bailey & al., Il Nuovo Cimento 9A, p. 369, 1972. Cited on page 294.

Motion Mountain – The Adventure of Physics available free of charge at [Link]

265 W. R indler, General relativity before special relativity: an unconventional overview of
relativity theory, American Journal of Physics 62, pp. 887–893, 1994. Cited on page 294.
266 Steven K. Bl au, Would a topology change allow Ms. Bright to travel backward in time?,
American Journal of Physics 66, pp. 179–185, 1998. Cited on page 297.
267 On the ‘proper’ formulation of relativity, see for example D. Hestenes, Proper particle
mechanics, Journal of Mathematical Physics 15, pp. 1768–1777, 1974. Cited on page 297.
268 The simple experiment to take a precise clock on a plane, fly it around the world and then
compare it with an identical one left in place was first performed by J.C. Hafele & R.E.
Keating, Around-the-world atomic clocks: predicted relativistic time gains, Science 177,
pp. 166–167, and Around-the-world atomic clocks: observed relativistic time gains, pp. 168–
170, 14 July 1972. See also Ref. 252. Cited on page 298.
269 A readable introduction to the change of time with observers, and to relativity in general,
is Roman U. Sexl & Herbert Kurt Schmidt, Raum-Zeit-Relativität, 2. Auflage,
Vieweg & Sohn, Braunschweig, 1991. Cited on page 298.
270 Most famous is the result that moving muons stay younger, as shown for example by D.H.
Frisch & J.B. Smith, Measurement of the relativistic time dilation using µ-mesons, Amer-
ican Journal of Physics 31, pp. 342–355, 1963. For a full pedagogical treatment of the twin
paradox, see E. Sheld on, Relativistic twins or sextuplets?, European Journal of Physics 24,
pp. 91–99, 2003. Cited on page 298.
271 Paul J. Nahin, Time Machines – Time Travel in Physics, Metaphysics and Science Fiction,
Copyright © Christoph Schiller November 1997–May 2006

Springer Verlag and AIP Press, second edition, 1999. Cited on page 298.
272 The first muon experiment was B. Rossi & D.B. Hall, Variation of the rate of decay of
mesotrons with momentum, Physical Review 59, pp. 223–228, 1941. ‘Mesotron’ was the old
name for muon. Cited on page 299.
273 A. Harvey & E. Schucking, A small puzzle from 1905, Physics Today, pp. 34–36, March
2005. Cited on page 299.
274 W. R indler, Length contraction paradox, American Journal of Physics 29, pp. 365–366,
1961. For a variation without gravity, see R. Shaw, Length contraction paradox, American
Journal of Physics 30, p. 72, 1962. Cited on page 301.
275 H. van Lintel & C. Gruber, The rod and hole paradox re-examined, European Journal
of Physics 26, pp. 19–23, 2005. Cited on page 301.

Dvipsbugw
346 ii special rel ativity

276 This situation is discussed by G.P. Sastry, Is length contraction paradoxical?, American
Journal of Physics 55, 1987, pp. 943–946. This paper also contains an extensive literature list
covering variants of length contraction paradoxes. Cited on page 301.
277 S.P. B oughn, The case of the identically accelerated twins, American Journal of Physics 57,
pp. 791–793, 1989. Cited on pages 302 and 305.
278 J.M. Supplee, Relativistic buoyancy, American Journal of Physics 57 1, pp. 75–77, January
1989. See also G.E.A. Matsas, Relativistic Arquimedes law for fast moving bodies and the
general-relativistic resolution of the ‘submarine paradox’, Physical Review D 68, p. 027701,
2003, or [Link] Cited on page 302. Dvipsbugw
279 The distinction was first published by J. Terrell, Invisibility of Lorentz contraction, Phys-
ical Review 116, pp. 1041–1045, 1959, and R. Penrose, The apparent shape of a relativistic-
ally moving sphere, Proceedings of the Cambridge Philosophical Society 55, pp. 137–139, 1959.
Cited on page 304.
280 G.R. Rybicki, Speed limit on walking, American Journal of Physics 59, pp. 368–369, 1991.
Cited on page 306.

Motion Mountain – The Adventure of Physics available free of charge at [Link]

281 The first examples of such astronomical observations were provided by A.R. Whitney &
al., Quasars revisited: rapid time variations observed via very-long-baseline interferometry,
Science 173, pp. 225–230, 1971, and by M.H. Cohen & al., The small-scale structure of
radio galaxies and quasi-stellar sources at 3.8 centimetres, Astrophysical Journal 170, pp. 207–
217, 1971. See also T.J. Pearson, S.C. Unwin, M.H. Cohen, R.P. Linfield, A.C.S.
R eadhead, G.A. Seielstad, R.S. Simon & R.C. Walker, Superluminal expansion
of quasar 3C 273, Nature 290, pp. 365–368, 1981. An overview is given in J.A. Zensus & T.J.
Pearson, editors, Superluminal radio sources, Cambridge University Press, 1987. Another
measurement, using very long baseline interferometry with radio waves, was shown on the
cover of Nature: I.F. Mirabel & L.F. Rodriguez, A superluminal source in the galaxy,
Nature 371, pp. 46–48, 1994. A more recent example was reported in Science News 152, p. 357,
6 December 1997.
Pedagogical explanations are given by D.C. Gabuzda, The use of quasars in teaching
introductory special relativity, American Journal of Physics 55, pp. 214–215, 1987, and by
Edwin F. Taylor & John A. Wheeler, Spacetime Physics – Introduction to Special
Relativity, second edition, Freeman, 1992, pages 89-92. This excellent book was mentioned
already in the text. Cited on page 308.
282 O.M. Bil aniuk & E.C. Sudarshan, Particles beyond the light barrier, Physics Today
22, pp. 43–51, 1969, and O.M.P. Bil anuk, V.K. Deshpande & E.C.G. Sudarshan,
‘Meta’ relativity, American Journal of Physics 30, pp. 718–723, 1962. See also E. R ecami,
editor, Tachyons, Monopoles and Related Topics, North-Holland, Amsterdam, 1978. Cited
Copyright © Christoph Schiller November 1997–May 2006

on page 309.
283 J.P. Costell a, B.H.J. McKell ar, A.A. R awlinson & G.J. Stephenson, The
Thomas rotation, American Journal of Physics 69, pp. 837–847, 2001. Cited on page 309.
284 See for example S.S. Costa & G.E.A. Matsas, Temperature and relativity, preprint avail-
able at [Link] Cited on page 310.
285 R.C. Tolman & G.N. Lewis, The principle of relativity and non-Newtonian mechanics,
Philosophical Magazine 18, pp. 510–523, 1909, and R.C. Tolman, Non-Newtonian mech-
anics: the mass of a moving body, Philosophical Magazine 23, pp. 375–380, 1912. Cited on
page 311.
286 This information is due to a private communication by Frank DiFilippo; part of the story
is given in F. DiFilippo, V. Natarajan, K.R. B oyce & D.E. Pritchard, Accurate

Dvipsbugw
bibliography 347

atomic masses for fundamental metrology, Physical Review Letters 73, pp. 1481–1484, 1994.
These measurements were performed with Penning traps; a review of the possibilities they
offer is given by R.C. Thompson, Precision measurement aspects of ion traps, Measure-
ment Science and Technology 1, pp. 93–105, 1990. The most important experimenters in the
field of single particle levitation were awarded the Nobel Prize in 1989. One of the Nobel
Prize lectures can be found in W. Paul, Electromagnetic traps for neutral and charged
particles, Reviews of Modern Physics 62, pp. 531–540, 1990. Cited on page 315.
287 J.L. Synge, Relativity: The Special Theory, North-Holland, 1956, pp. 208–213. More about
antiparticles in special relativity can be found in J.P. Costell a, B.H.J. McKell ar & Dvipsbugw
A.A. R awlinson, Classical antiparticles, American Journal of Physics 65, pp. 835–841,
1997. See also Ref. 302. Cited on page 316.
288 A. Papapetrou, Drehimpuls- und Schwerpunktsatz in der relativistischen Mechanik,
Praktika Acad. Athenes 14, p. 540, 1939, and A. Papapetrou, Drehimpuls- und Schwer-
punktsatz in der Diracschen Theorie, Praktika Acad. Athenes 15, p. 404, 1940. See also
M.H.L. Pryce, The mass-centre in the restricted theory of relativity and its connexion
with the quantum theory of elementary particles, Proceedings of the Royal Society in Lon-

Motion Mountain – The Adventure of Physics available free of charge at [Link]

don, A 195, pp. 62–81, 1948. Cited on page 317.
289 The references preceding Einstein’s E = mc 2 are: S. Tolver Preston, Physics of the Ether,
E. & F.N. Spon, 1875, J.H. Poincaré, La théorie de Lorentz et le principe de réaction,
Archives néerlandaises des sciences exactes et naturelles 5, pp. 252–278, 1900, O. De Pretto,
Ipotesi dell’etere nella vita dell’universo, Reale Istituto Veneto di Scienze, Lettere ed Arti tomo
LXIII, parte 2, pp. 439–500, Febbraio 1904, F. Hasenöhrl, Berichte der Wiener Akademie
113, p. 1039, 1904, F. Hasenöhrl, Zur Theorie der Strahlung in bewegten Körpern, An-
nalen der Physik 15, pp. 344–370, 1904, F. Hasenöhrl, Zur Theorie der Strahlung in be-
wegten Körpern – Berichtigung, Annalen der Physik 16, pp. 589–592, 1905. Hasenöhrl died
in 1915, De Pretto in 1921. All these publications were published before the famous paper
by Albert Einstein, Ist die Trägheit eines Körpers von seinem Energieinhalt abhängig?,
Annalen der Physik 18, pp. 639–641, 1905. Cited on page 319.
290 A jewel among the textbooks on special relativity is the booklet by Ulrich E. Schröder,
Spezielle Relativitätstheorie, Verlag Harri Deutsch, Thun, 1981. Cited on pages 321 and 324.
291 A readable article showing a photocopy of a letter by Einstein making this point is Lev B.
Okun, The concept of mass, Physics Today, pp. 31–36, June 1989. The topic is not without
controversy, as the letters by readers following that article show; they are found in Physics
Today, pp. 13–14 and pp. 115–117, May 1990. The topic is still a source of debates. Cited on
page 323.
292 Christian Møller, The Theory of Relativity, Clarendon Press, 1952, 1972. This standard
text has been translated in several languages. Cited on page 324.
Copyright © Christoph Schiller November 1997–May 2006

293 P. Ehrenfest, Gleichförmige Rotation starrer Körper und Relativitätstheorie, Physikalis-

che Zeitschrift 10, pp. 918–928, 1909. Ehrenfest (incorrectly) suggested that this meant that
relativity cannot be correct. A modern summary of the issue can be found in M.L. Rug-
giero, The relative space: space measurements on a rotating platform, [Link]
org/abs/gr-qc/0309020. Cited on page 325.
294 R.J. Low, When moving clocks run fast, European Journal of Physics 16, pp. 228–229, 1995.
Cited on page 329.
295 G. Stephenson & C.W. Kilmister, Special Relativity for Physicists, Longmans, Lon-
don, 1965. See also W.N. Matthews, Relativistic velocity and acceleration transforma-
tions from thought experiments, American Journal of Physics 73, pp. 45–51, 2005. Cited on
page 330.

Dvipsbugw
348 ii special rel ativity

296 The impossibility of defining rigid coordinate frames for non-uniformly accelerating observ-
ers is discussed by Charles Misner, Kip Thorne & John A. Wheeler, Gravitation,
Freeman, p. 168, 1973. Cited on page 332.
297 E.A. Desloge & R.J. Philpott, Uniformly accelerated reference frames in special re-
lativity, American Journal of Physics 55, pp. 252–261, 1987. Cited on page 332.
298 R.H. Good, Uniformly accelerated reference frame and twin paradox, American Journal
of Physics 50, pp. 232–238, 1982. Cited on pages 332, 333, and 336.
299 J. Dwayne Hamilton, The uniformly accelerated reference frame, American Journal of
Dvipsbugw
Physics 46, pp. 83–89, 1978. Cited on page 333.
300 The best and cheapest mathematical formula collection remains the one by K. Rottmann,
Mathematische Formelsammlung, BI Hochschultaschenbücher, 1960. Cited on page 333.
301 C.G. Adler & R.W. Brehme, Relativistic solutions to a falling body in a uniform gravit-
ation field, American Journal of Physics 59, pp. 209–213, 1991. Cited on page 334.
302 See for example the excellent lecture notes by D.J. R aymond, A radically modern ap-
proach to freshman physics, on the [Link]

Motion Mountain – The Adventure of Physics available free of charge at [Link]

website. Cited on pages 334 and 347.
303 L. Mishra, The relativistic acceleration addition theorem, Classical and Quantum Gravity
11, pp. L97–L102, 1994. Cited on page 334.
304 Edward A. Desloge, The gravitational red-shift in a uniform field, American Journal of
Physics 58, pp. 856–858, 1990. Cited on page 336.
305 One of the latest of these debatable experiments is T.P. Krisher, L. Maleki, G.F. Lu tes,
L.E. Primas, R.T. Logan, J.D. Anderson & C.M. Will, Test of the isotropy of the
one-way speed of light using hydrogen-maser frequency standards, Physical Review D 42,
pp. 731–734, 1990. Cited on page 338.
306 Edwin F. Taylor & A.P. French, Limitation on proper length in special relativity,
American Journal of Physics 51, pp. 889–893, 1983. Cited on page 339.

Copyright © Christoph Schiller November 1997–May 2006

Dvipsbugw