Part I Special Relativity

G. W. Gibbons
D.A.M.T.P.,
Cambridge University,
Wilberforce Road,
Cambridge CB3 0WA,
U.K.
February 14, 2008
The views of space and time which I wish to lay before you have
sprung from the soil of experimental physics, and therein lies their
strength. They are radical. Henceforth space by itself, and time by
itself, are doomed to fade away into mere shadows, and only a kind
of union will preserve an independent reality. H Minkowski (1908).

Contents
1 The Schedule
1.1 Units . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

5
5

2 Einsteins Theory of Special Relativity

3 *Early ideas about light*

3.1 Maxwells equations . . . . . . . . . . . . . . . . . . . . . . . . .

6
10

4 *The Speed of Light*

11
4.1 *Roemers measurement of c* . . . . . . . . . . . . . . . . . . . . 11
4.2 *Fizeaus measurement of c* . . . . . . . . . . . . . . . . . . . . 12
4.3 *Foucaults rotating mirror* . . . . . . . . . . . . . . . . . . . . . 12
5 Absolute versus Relative motion

12

6 Velocity composition formulae

14

7 Galilean Principle of Relativity

14
7.1 Waves and Galilean Transformations . . . . . . . . . . . . . . . . 16

8 Spacetime
16
8.1 Example: uniform motion in 1+1 dimensions . . . . . . . . . . . 17
8.2 Example: uniform motion in 2+1 dimensions . . . . . . . . . . . 17
8.3 Example: non-uniform motion in 1+1 dimensions . . . . . . . . . 17
9 Minkowskis Spacetime viewpoint

17

10 Einsteins Principle of Relativity

10.1 Michelson-Morley Experiment . . . . . . . . . . . .
10.2 Derivation of the Lorentz Transformation formulae
10.3 Relativistic velocity composition law . . . . . . . .
10.4 *Observational for Einsteins second postulate* . .
10.5 Light in a medium: Fresnel Dragging . . . . . . . .
10.6 Composition of Lorentz Transformations . . . . . .
10.7 Velocity of light as an upper bound . . . . . . . . .
10.8 *Super-LuminalRadio sources* . . . . . . . . . . .
10.9 The two-dimensional Lorentz and Poincare groups
11 The
11.1
11.2
11.3

invariant interval
Timelike Separation . . . . . . . . . . . . . . .
Spacelike separation . . . . . . . . . . . . . . .
Time Dilation . . . . . . . . . . . . . . . . . . .
11.3.1 Muon Decay . . . . . . . . . . . . . . .
11.4 Length Contraction . . . . . . . . . . . . . . . .
11.5 The Twin Paradox: Reverse Triangle Inequality
11.5.1 *Hafele -Keating Experiment* . . . . .
11.6 Accelerating world lines . . . . . . . . . . . . .

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

18
18
20
22
22
23
24
25
25
26

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

26
27
28
28
28
29
29
30
30

12 Doppler shift in one space dimension

31
12.1 *Hubbles Law* . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
13 The Minkowski metric
33
13.1 Composition of Lorentz Transformations . . . . . . . . . . . . . . 34
14 Lorentz Transformations in 3 + 1 spacetime dimensions
34
14.1 The isotropy of space . . . . . . . . . . . . . . . . . . . . . . . . . 35
14.2 Some properties of Lorentz transformations . . . . . . . . . . . . 36
15 Composition of non-aligned velocities
15.1 Aberration of Light . . . . . . . . . . . . . . . . . .
15.2 * Aberration of Starlight* . . . . . . . . . . . . . .
15.3 Water filled telescopes . . . . . . . . . . . . . . . .
15.4 Headlight effect . . . . . . . . . . . . . . . . . . . .
15.5 Solid Angles . . . . . . . . . . . . . . . . . . . . . .
15.6 *Celestial Spheres and conformal transformations*
15.7 *The visual appearance of rapidly moving bodies*
15.8 Transverse Doppler effect . . . . . . . . . . . . . .
2

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

36
37
38
39
39
39
40
40
41

15.9 *The Cosmic Microwave Background* . . . . . . . . . . . . . . .

41

16 * Kinematic Relativity and the Milne Universe*

42
16.1 *The Foundations of Geometry* . . . . . . . . . . . . . . . . . . 43
16.2 The Milne metric and Hubbles Law . . . . . . . . . . . . . . . . 46
16.3 *Relativistic composition of velocities and trigonometry in Lobachevsky
space* . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
16.4 Parallax in Lobachevsky space . . . . . . . . . . . . . . . . . . . 49
17 *Rotating reference frames*
17.1 Transverse Doppler effect and time dilation . . . .
17.2 *The Sagnac Effect* . . . . . . . . . . . . . . . . .
17.3 Length Contraction . . . . . . . . . . . . . . . . . .
17.4 Machs Principle and the Rotation of the Universe

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

50
50
51
52
52

18 General 4-vectors and Lorentz-invariants

18.1 4-velocity and 4-momentum . . . . . . . .
18.2 4-velocity . . . . . . . . . . . . . . . . . .
18.3 4-momentum and Energy . . . . . . . . .
18.4 Non-relativistic limit . . . . . . . . . . . .
18.5 Justification for the name energy . . . . .
18.6 *Hamiltonian and Lagrangian* . . . . . .

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

53
53
54
54
55
55
56

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

19 Particles with vanishing rest mass

56
19.1 Equality of photon and neutrino speeds . . . . . . . . . . . . . . 57
20 Particle decays collisions and production
20.1 Radioactive Decays . . . . . . . . . . . . . . . . . . . .
20.2 Impossibility of Decay of massless particles . . . . . .
20.3 Some useful Inequalities . . . . . . . . . . . . . . . . .
20.4 Impossibility of emission without recoil . . . . . . . . .
20.5 Decay of a massive particle into one massive and one
particle . . . . . . . . . . . . . . . . . . . . . . . . . .
20.6 Decay of a massive particle into two massless particles

. . . . . .
. . . . . .
. . . . . .
. . . . . .
massless
. . . . . .
. . . . . .

21 Collisions, centre of mass

21.1 Compton scattering . . . . . .
21.2 Production of pions . . . . . . .
21.3 Creation of anti-protons . . . .
21.4 Head on collisions . . . . . . .
21.5 Example: Relativistic Billiards
21.6 Mandelstam Variables. . . . . .

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

57
57
58
59
60
60
60
61
62
63
63
64
66
66

22 Mirrors and Reflections

22.1 *The Fermi mechanism* . . . . . . .
22.2 *Relativistic Mirrors* . . . . . . . .
22.3 *Corner Reflectors on the Moon* . .
22.4 Time reversal . . . . . . . . . . . . .
22.5 Anti-particles and the CPT Theorem

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

67
67
68
69
70
70

23 4-acceleration and 4-force

70
23.1 Relativistic form of Newtons second law . . . . . . . . . . . . . . 70
23.2 Energy and work done . . . . . . . . . . . . . . . . . . . . . . . . 71
23.3 Example: relativistic rockets . . . . . . . . . . . . . . . . . . . . . 71
24 The Lorentz Force
72
24.1 Example: particle in a uniform magnetic field . . . . . . . . . . . 73
24.2 Uniform electro-magnetic field and uniform acceleration . . . . . 74
25 4-vectors, tensors and index notation
25.1 Contravariant vectors . . . . . . . . . . . . . . .
25.2 Covariant vectors . . . . . . . . . . . . . . . . . .
25.3 Example: Wave vectors and Doppler shift . . . .
25.4 Contravariant and covariant second rank tensors
25.5 The musical isomorphism . . . . . . . . . . . . .
25.6 De Broglies Wave Particle Duality . . . . . . . .
25.7 * Wave and Group Velocity: Legendre Duality* .
25.8 The Lorentz equation . . . . . . . . . . . . . . .

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

26 Uniformly Accelerating reference frames: Event Horizons

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

75
75
76
77
77
79
79
80
82
83

27 Causality and The Lorentz Group

84
27.1 Causal Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
27.2 The Alexandrov-Zeeman theorem . . . . . . . . . . . . . . . . . . 84
27.3 Minkowski Spacetime and Hermitian matrices . . . . . . . . . . . 85
28 Spinning Particles and Gyroscopes
86
28.1 Fermi-Walker Transport . . . . . . . . . . . . . . . . . . . . . . . 86
28.2 Spinning particles and Thomas precession . . . . . . . . . . . . . 86
28.3 Bargmann-Michel-Telegdi Equations . . . . . . . . . . . . . . . . 87

The Schedule

Read as follows:
INTRODUCTION TO SPECIAL RELATIVITY
8 lectures, Easter and Lent terms [Lecturers should use the signature convention
(+ ).]
Space and time The principle of relativity. Relativity and simultaneity. The
invariant interval. Lorentz transformations in (1 + 1)-dimensional spacetime.
Time dilation and muon decay. Length contraction. The Minkowski metric for
(1 + 1)-dimensional spacetime.[4]
4vectors Lorentz transformations in (3 + 1) dimensions. 4vectors and Lorentz
invariants. Proper time. 4velocity and 4momentum. Conservation of 4
momentum in radioactive decay.[4]
BOOKS
G.F.R. Ellis and R.M. Williams Flat and Curved Space-times Oxford University Press 2000 24.95 paperback
W. Rindler Introduction to Special Relativity Oxford University Press 1991
19.99 paperback
W. Rindler Relativity: special, general and cosmological OUP 2001 24.95
paperback
E.F. Taylor and J.A. Wheeler Spacetime Physics: introduction to special
relativity Freeman 1992 29.99 paperback

1.1

Units

When quoting the values of physical quantities, units in which c = and h=1,
will frequently be used. Thus, at times for example, distances may be expressed
in terms of light year. Astronomers frequently use parsecs which is the distance
at which is short for paralax second. It is the distance at which the radius of
the earth subtends one second of arc. One parsec works out to be 3.0 1013Km
or 3.3 light years. A frequently used unit of energy, momentum or mass is the
electron volt or eV which is the work or energy required to move an electron
through a potential difference of one Volt.
Physical units, masses and properties of elementary particles are tabulated
by the Particle Data Group and may be looked up at
http://pdg.lbl.gov
.
Although not necessary in order to follow the course, it is a frequently illuminating and often amusing exercise to go back to the original sources. Many
of the original papers quoted here may be consulted on line. For papers in the
Physical Review, back to its inception in the late nineteenth century go to
http://prola.aps.org
5

.
For many others, including Science and Philosophical Transactions of the
Royal Society (going back its beginning in the to seventeenth century) go to
http://uk.jstor.org
.

Einsteins Theory of Special Relativity

is concerned with the motion of bodies or particles whose relative velocities are
comparable with that of light
c = 299, 792, 458 ms1 .

(1)

In a nutshell, Newtons Second Law remains unchanged in the form

dp
=F
dt

(2)

where F the force acting on a particle of momentum p and mass m1 , but while
according to
Newton s Theory
p = mv
(3)
in
Einstein s

Theory

mv
.
p= q
2
1 vc2

(4)

If this were all there is in it, relativity would, perhaps, not be especially
interesting. What makes Relativity important is that it entails a radical revision
of our elementary ideas of space and time and in doing so leads to the even
more radical theory of General Relativity which comes into play when gravity
is important. In this course we shall ignore gravity and confine our attention to
Special Relativity. For matters gravitational the reader is directed to[34].
To see why relativity has such a profound impact on ideas about space and
time, note that we are asserting that there actually is such a thing as the velocity
of light.
For the benefit of those who have not studied Physics at A-level, or who did,
but have now forgotten all they ever knew, the next section contains a review
of the elementary physics of light.

*Early ideas about light*

Experiments with shadows and mirrors lead to the idea that light is a form of
energy that propagates along straight lines called light rays. On reflection at a
smooth surface S at rest, it is found that
1 properly

speaking rest-mass

(i) The incident ray, the reflected ray and the normal to the surface at the point
of reflection are co-planar
(ii) The incident and reflected rays make equal angles with the normal.
Hero of Alexander showed that these laws may be summarized by the statement that if A is a point on the incident ray, B on reflected ray and x S the
point at which the reflection takes place, then x is such that the distance
d(A, x) + d(x, B)

(5)

is extremized among all paths from AyB, y S to the surface and from the
surface to B.
When light is refracted at a smooth surface S it is found that
(i) The incident ray, the refracted ray and the normal to the surface at the point
of refraction are co-planar
(ii) The incident and refracted rays make angles i and r with the normal such
that
sin i
nr
=
(6)
sin r
ni
where the quantities ni and nr are characteristic of the medium and may depend
upon the colour of the light and are called its refractive index. By convention
one sets n = 1 for the vacuum.
Pierre Fermat showed that these laws, first clearly enunciated in about 1621
by the Leyden mathematician Willebrod Snellius or Snell in work which was
unpublished before his death in 1626, and later by Descartes, although probably
known earlier to Thomas Harriot, may be summarized by the statement that if
A is a point on the incident ray, B on refracted ray and x the point at which
the refraction takes place, then x is such that the optical distance
ni d(A, x) + nr d(x, B)

(7)

is extremized among all paths AyB, y S from A to surface and from the
surface to B. In other words the differential equations for light rays may be
obtained by varying the action functional
Z
nds
(8)

where ds is the element Euclidean distance.

By the time of Galilei its was widely thought that light had a finite speed, c
and attempts were made to measure it. Broadly speaking there were two views
about the significance of this speed.
The Emission or Ballistic Theory held by Isaac Newton and his followers according to which light consisted of very small particles or corpuscles with mass
m speed c and momentum p = mc, the speed varying depending upon the

medium. On this theory, Snells law is just conservation of momentum parallel

to the surface.
pi sin i = pr sin r ,
(9)
whence, assuming that the mass is independent of the medium
sin i
cr
= .
sin r
ci

(10)

The Wave Theory proposed by the dutch physicist Christian Huygens in 1678 ,
according to which light is a wave phenomenon having a speed c and such that
each point on the wave front gives rise to a secondary spherical wave of radius
ct whose forward envelope gives the wavefront at a time t later.
On this theory, Snells law arises because the wavelength i of the incident
wave and the wavelength of the refracted wave r differ. Applying Huygens
construction gives
i
sin i
=
.
(11)
sin r
r
Since, for any wave of frequency f , f = c and since the frequency of the wave
does not change on refraction, we have according to the wave theory:
sin i
ci
= .
sin r
cr

(12)

The two theories gave the opposite prediction for the speed of light in a
medium. Since refractive indices are never found to be less than unity, according
to the emission theory the speed of light in a medium is always greater than in
vacuo, while according to the wave theory it is always smaller than in vacuo.
One way to distinguish between the two theories was to measure the speed of
light in vacuo and in a medium. This was first done by Foucault in 1850, and
more accurately by Michelson in 1883 using the rotating mirror method of the
former, which will be described shortly. By interposing a tube filled with water
in the path of the light, they showed that the speed of light in water was slower
than in vacuo 2 . It follows that Hero and Fermats variational properties may
be summarized by the statement that the time taken for light to traverse the
physical path is extremized.
Another way to distinguish the theories is by their ability to account for the
diffraction of light by very small obstacles as observed by Grimaldi in 1665 or by
experiments on slits, such as were performed by the polymath Thomas Young
3
in 1801. Following a large number of subsequent experiments, notably by
Fresnel, by Foucaults time, some form of wave theory was accepted by almost
2 The argument is in fact slightly indirect since these experiments actually measure the
group velocity of light while refraction depends on the phase velocity. The distinction is
described later. Given one, and information about the dispersion, i.e. how the refractive
index varies with wavelength, one may calculate the other.
3 Young played an equal role with Champillon in the translation of the Egyptian hieroglyphics on the Rosetta stone.

all physicists. In its simplest form, this postulated that in vacuo, some quantity
satisfies the scalar wave equation
1 2
= 2 ,
c2 t2

(13)

which, if c is constant, is easily seen to admit wavelike solutions of the form

= A sin 2(
or more generally
2

x
f t) ,

= f (k t),

(14)
(15)

|k|

where f ( ) is an arbitrary C function of its argument and

= c. Since equation (13) is a linear equation, the Principle of Superposition holds and solutions
with arbitrray profiles, moving in arbitrary directions may be superposed. A
fact which not only explains many opetical phenomena but also led to the idea
of Fourier Analysis.Note that solutions (13) are non-dispersive, the speed c is
independent of the wavelength or frequency f .
Until Einsteins work, almost all physicists believed that wave propagation
required some form of material medium and that light was no exception. The
medium was called the luminiferous aether (or ether) and many remarkable
properties were ascribed to it. Many physicist , incorrectly as it turned out, believed ed that it was inextricably linked with the nature of gravitation. Others,
like Lord Kelvin, postulated that atoms could be thought of as knotted vortex
rings. This seemed to require that the ether was some sort of fluid. A key
question became: what is the speed of the earth relative to the aether?
The properties of the ether became even harder to understand when it was
established that light could be polarized. This was first noticed by Huygens
who was studying the refraction of light through a crystal of calcite also known
as Iceland spar. In 1808 Malus discovered that light could be polarized by in
the process of reflection.These observations led directly to the idea that light
due to some sort of motion transverse to the direction of propagation, and so
the quantity should be some sort of vector rather than a scalar. They also
suggested to many that the aether should be some sort of solid.
The realization that light was an electromagnetic phenomenon and the great
achievement of the Scottish physicist James Clerk Maxwell (1831-1879) in providing in 1873 a complete, unified and consistent set of equations to describe
electromagnetism, which moreover predicted the existence of electromagnetic
waves moving at the speed of light and the subsequent experimental verification
by the German physicist Heinrich Hertz (1857-1894) around 1887 4 did nothing
to dispel the wide-spread confusion about the aether. Elaborate mechanical
models of the aether were constructed and all the while, it and the earths motion through it, eluded all experimental attempts at detection. The general
frustration at this time is perhaps reflected in the words of the president of the
4 In fact it seems clear that Hertz had been anticipated by the English Electrician D E
Hughes in 1879, but the significance of his work was not appreciated until much later [49].

British Association, Lord Salisbury who is reported to have proposed, at one it

its meeting held at Oxford, a definition of the aether as the nominative of the
verb to undulate.
In a similar vein. discussing the various allegedly physical interpretations,
Hertz declared that
To the question What is Maxwells Theory, I know of no shorter
or more definite answer than the following: Maxwells theory is
Maxwells system of equations. Every theory which leads to the
same system of equations, and therefore comprises the same possible phenomena, I would consider a form of Maxwells theory.
Maxwells equations have many beautiful and remarkable properties, not
the least important of which is invariance not under Galilei transformations as
might have been expected if the aether theory was correct, but rather under
what we now call Lorentz transformations. This fact was noticed for the scalar
wave equation (13) in 1887, long before Einsteins paper of 1905, by Woldemar
Voigt(1850-1919) and both Lorentz and Poincare were aware of the Lorentz
invariance of Maxwells equations but they regarded this as a purely formal
property of the equations. As we shall see, Einsteins insight was in effect to
see that it is perhaps the single most important mathematical fact about the
equations. From it flows all of Special Relativity and much of General Relativity.

3.1

Maxwells equations

These split into two sets. The first set always holds, in vaccuo or in any material medium and independently of whether any electric charges or currents
are present. They deny the existence of magnetic monopoles and asserts the
validity of Michael Faradays law of induction.
div B = 0 ,

curl E =

B
t .

(16)

The second set describe the response of the fields to the pressence of electric
charges, charge density and currents, current density j. At the expense of
introducing two additional fields they may also be cast in a form which is always
correct. They assert the validity of Coulombs law, and Amperes law, provided
it is supplemented by the last, crucial, additional term, called the displacement
current due to Maxwell himself.
div D =

curl H = j +

D
.
t

(17)

It follows from the identity div curl = 0, that electric charge is conserved

+ divj = 0 .
t

10

(18)

In order to close the system one requires constitutive relations relating D

and H to E and B. In vacuo these are linear relations
D = 0 E ,

H=

1
B,
0

(19)

where 0 and 0 are two universal physical constants constants called respectively the permeability and permittivity of free space .Thus, in vacuuo, Maxwells
equations are linear and the principle of superposition holds for their solutions.
Thus, in vacuo
div E =

curl

E
1
B = j + 0
.
0
t

(20)

If there are no charges or currents present, use of the identity curl curl =
grad div 2 gives
2E
(21)
0 0 2 = 2 E ,
t
2B
0 0 2 = 2 B .
(22)
t
Thus each component of the electric and magnetic field travels non-dispersiveley
with velocity
1
.
c=
(23)
0 0
The divergence free conditions imply that solutions of the form E = E0 f (k
t), B = B0 f (k t), are transversely (plane) polarized
k.E = 0 ,

k.B = 0

(24)

E0 = B0 n,

(25)

and moreover since

B0 = n E 0 ,

k
with n = |k|
, the vectors (E0 , B0 , n) form a right handed normal but not orthonormal triad. Physically the direction of the polarization is usually taken to
be that of the electric field, since this is easier to detect. Thus for any given
propagation direction n there are two orthogonal polarization states. in the sense
that one may choose the solutions such that E1 .E2 . = 0 and thus B1 .B2 = 0.

4
4.1

*The Speed of Light*

*Roemers measurement of c*

That light does indeed have a finite speed was first demonstrated, and the speed
estimated by the Danish astronomer Olaus Roemer (1614-1710) in 1676 [1]. He
observed the phases of Io, the innermost of the four larger satellites or moons
11

of Jupiter (Io, Europa, Ganymede and Callisto in order outward) which had
been discovered in 1610 by Galileo Galilei (1564-1642) using the newly invented
telescope 5 and of Ios motion around Jupiter is about 1.77 days can be deduced
by observing its phases, when it is eclipsed by Jupiter whose orbital period is
11.86 years. In 1688 G D Cassini had published a set of predictions for these but
Roemer observed that they were inaccurate by about 15 minutes. The periods
are shorter when Jupiter is moving toward the earth than when it is moving
away from the earth. Roemer explained this and obtained a value for the speed
of light by arguing that when Jupiter is moving toward the earth Io the time
between eclipses is shorter than when Jupiter is moving away from us because
in the former case light the total distance light has to travel is shorter than
in the latter case. He obtained a value of 192,000 miles per second or roughly
310,000 Km per sec.
If we think anachronistically 6 , we might say if we think of Io as a clock, its
period is Doppler shifted.

4.2

*Fizeaus measurement of c*

The first accurate terrestrial measurement of the speed of light was by the
French physicist Fizeau who, in 1849 [6], passed a beam of light through a
rotating toothed wheel with 720 teeth, reflected it off a plane mirror 8.633 Km
away and sent the light back toward the toothed disc. For a rotation speed of
12.6 turns per second the light was eclipsed giving a speed of about 315,000 Km
per second.

4.3

*Foucaults rotating mirror*

In 1850 another French physicist, Foucault [7] reflected light off a mirror which
was rotating about an axis parallel to its plane. The reflected light was then
sent back in the same direction. If the rate of rotation of the mirror was chosen
suitably the light arrived back at its point of departure. From this Foucault
deduced a value for the velocity of light of 298, 000 Kms1 . As mentioned above,
he was also able to establish that the speed of light in water is less than in vacuo.

Absolute versus Relative motion

Newton based his theory on the assumption that space was uniform and described by the usual laws of Euclidean geometry. There then arose the issue of
whether motion with respect to that background was observable. If it was, then
one would have a notion of absolute as opposed to relative motion. Newton
argued, using the idea of a suspended bucket of water, that one does have an
5 The true inventor of the telescope is not known. It seems to have been known to the
English cosmologist Thomas Digges and the Oxford mathematician and explorer of Virginia,
Thomas Harriot(1560-1621
6 The Doppler effect was proposed by the Austrian physicist C.J. Doppler in 1842

12

idea of absolute rotational motion. However according to his laws of motion

there is no obvious dynamical way of detecting absolute translational motion.
Since his laws imply that the centre of mass of an isolated system of bodies
one could define an absolute frame of rest as that in which the centre of mass
of for, example the visible stars, is at rest. One later suggestion was that one
could take the centre of the Milky way. Lambert had suggested that it was the
location of a dark regent or massive body, a suggestion also made by M
adler.
Interestingly we now know that at the centre of the Milky Way there is a massive Black Hole of mass around a million times the mass of the sun, 106 M .
The mass of the Milky way is about 1012 M . However any such centre of mass
frame can only be determined by astronomical observations. It could not be
found using purely dynamical experiments beneath cloudy skies here on earth.
Later, physicists, like Ernst Mach[50], began to worry about the logical
foundations of Newtons laws. Exactly what was meant by the statement that
a particle continues in a state of uniform motion if unaffected by an external
force? Wasnt Newtons second law in effect a tautology? etc. L. Lange in 1885
[52] and others [50] had realized that an operational meaning could be given to
Newtons laws if one introduces the idea of an inertial frame of reference. This
Lange thought of as a coordinate system for R3 which could be determined by
the free, mutually non-parallel, motion of three particles. Then the first law
could be formulated as the non-trivial and empirically verifiable statement that
any fourth free particle would move in this frame with uniform motion. In
effect we are to use use the straight line motion of particles to build up what
is sometimes called an inertial coordinate system or inertial reference system.
In fact this construction closely resembles various constructions in projective
geometry, especially if one adds in time as an extra coordinate. We will discuss
this in more detail later.
Mach pointed out that even if one used astronomical observations to determine a fundamental inertial frame of reference which is at rest with respect to
the fixed stars, that is stars so distant that their proper motions are negligible,
this raises a puzzle. For example, in principle we can define a non-rotating frame
in two different ways,
(i) Using gyroscopes for example which, if they are subject to no external torque
will point in a constant direction in an inertial frame of reference, in other words
using what has come to be called the compass of inertia .
(ii) Using the fixed stars. Nowadays we use quasi stellar radio sources quasars .
It is then a remarkable coincidence that, as we shall see in detail later, to very
high accuracy these two definitions agree. Mach had some, not very specific,
suggestions about a possible explanation. Machs ideas strongly influenced those
of Einstein, especially when he was formulating his General Theory of Relativity.
They really cannot be pursued in detail without General Relativity and without
some understanding of Relativistic Cosmology.
It was against this background that the question of the aether became so
important. If it really existed, it would provide an alternative frame of reference, which might,or might not, coincide with the astronomically determined or
13

dynamically determined frames of reference. It could, for example, remove the

ambiguity about translational motion. The obvious guess was that it all three
frames agreed. But if this was true, then the earth should be moving through
the ether and this motion should be detectable.

Velocity composition formulae

Given that the speed of light is finite and presumably well defined on would ask,
on the basis of Newtonian theory, in what frame? If the there is such a thing as
the velocity of light, independent of reference frame then the standard velocity
addition formula in
Newtonian Theory

v v+u

(26)

cannot be right. In fact, as we shall see later, one has a velocity composition
(rather than addition) formula. In
Einstein s Theory
so that if v = c,
v

v+u
1+ uv
2

(27)

c+u
= c.
1 + cu
c2

Exercise Show that if u c and v c then

u+v
1+ uv
c2

(28)
c.

Exercise Using the formula dW

dt = F.v, for the rate of doing work W by a
force F acting on a point moving with velocity v, show that the work done in
2
accelerating a particle of mass m from rest to a final velocity v is pmc v2 mc2 .
1 c2

The theory is called the Theory of Relativity because it deals with relative
velocities and what is called the Principle of Relativity. This idea began, at
least in modern times, with Galileo and we shall begin with his version of it.

Galilean Principle of Relativity

Suppose a boat is moving with uniform velocity along a canal and we are looking
at it. We are asked the following
Question The lookout is in the crows nest and drops a heavy weight onto the
deck. Will it hit the captain below?
Answer Yes.
Reason We pass to a frame of reference S moving with the boat. The frame at
rest with respect to the canal is an inertial frame of reference. Galileo assumed
that

14

The laws of dynamics are the same in all frames of reference which are in
uniform motion with respect to an inertial frame of reference
Now if we drop something from rest in frame S it will fall vertically down,
Therefore if we drop something from rest in frame S it will fall vertically down,
The boxed statement is Galilean Relativity follows in Isaac Newtons (1642account of dynamics because
In frame

d2 x
= F(x, t).
dt2

(29)

But to transform to frame S we set

= x ut,
x

(30)

and hence

d2 x

= F(
x + ut)
in frame S.
(31)
dt2
Note that Galileo assumed that the passengers in the boat would use the same
coordinate t. In principle one might have thought that one would also have to
change the time coordinate to a new coordinate t for this equivalence to work
out but both Galileo and Newton agreed that
m

Time is an absolute coordinate

(32)

that is, it takes the same value in all inertial frames of reference
t = t

(33)

Formulae (30,33) constitute a

Gallean Transformation

t = t,

= x ut.
x

(34)

We have just shown that Newtons equations of motion are invariant under
Galilean Transformations. We shall now use Galilean transformations to deduce
the Non-relativistic Velocity Addition Formulae.
If a particle moves with respect to a frame S such that
=v
t + x
0
x

(35)

+x
0
x ut = v

(36)

)t + x
0 .
x = (u + v

(37)

,
v=u=v

(38)

then
Thus
and hence
gives the velocity with respect to S. Later we will imitate this simple calculation
to obtain the velocity addition formula in special relativity.
15

7.1

Waves and Galilean Transformations

If, in a frame S at rest with respect to the aether, we have a wave of the form


= sin k.x t
(39)

2
Its speed is c = |k|
, its wavelength = |k|
and frequency f =
If we submit it to a Galilei transformation it becomes


= sin k.
x ( u.k)t .

2 .

(40)

the wave has the same wavelength but the frequency f =

In the frame S,
is
changed
u
(41)
f = f (1 cos )
c
and the speed is c, where
c = c u cos .
(42)

The formula for velocity in the moving frame S is very much what one
expects on the basis of a particle viewpoint but note that the angle is the
k
angle between the direction of the wave n = |k|
and the relative velocity u of

the two frames S and S. Both frames agree on this as do they on the direction
of motion of the wave. In other words, Galileis transformation formulae predict
that there is no aberration.
Later, we will obtain the physically correct results using the same method
as above, but instead of a Galilei transformation we shall substitute using a
Lorentz transformation.

Spacetime

Before proceeding, we will pause to develop a way of thinking about kinematics

that in fact goes back to Lagrange and DAlembert. The latter wrote, in his
article on dimension in the Encylopedie ou Dictionaire raisonee des sciences,
des arts et des metiers in 1764
A clever acquaintance of mine believes that it is possible to think of
time as a fourth dimension, so that the product of time and solidity
would in some sense be the product of four dimensions; it seems to
me that this idea, while debatable, has certain merits-at least the
merit of novelty.
The German pioneer of psycho-physics Gustav Theodor Fechner (1801-1887)
wrote a popular article entitled Der Raum hat vier Dimensionwhich discusses
related ideas. By that time the study of extra spatial dimensions was quiet
advanced and the German Astronomer Johann C F Zollner (1834-1882) gained
notoriety for claiming that the alleged ability of self-claimed spiritualists to

16

untie knots sealed in glass jars was only explicable if they had been moved into
a fourth spatial dimension.
By 1880s the French railway engineer Ibry was using spacetime diagrams in
a practical way to construct railway time tables (see illustration on p 55 of [34])
The following examples illustrate the power of the view point in solving this
type of mundane problem.

8.1

Example: uniform motion in 1+1 dimensions

A commuter is usually picked up by his/her spouse who drives at constant speed

from their house to meet the commuter at 5 oclock. One day the commuter
arrives on an earlier train at 4 oclock and decides to walk. After a while the
commuter is picked up by his/her spouse who has driven to meet him as usual.
They arrive back at their house 10 minutes earlier than usual. For how long did
the commuter walk?

8.2

Example: uniform motion in 2+1 dimensions

Four ships, A, B, C, D are sailing in a fog with constant and different speeds
and constant and different courses. The five pairs A and B, B and C, C and A,
B and D, C and D have each had near collisions; call them collisions. Show
that A and D necessarily collide.
Hint Consider the triangle in the three-dimensional spacetime diagram formed
by the world-lines of A, B and C.

8.3

Example: non-uniform motion in 1+1 dimensions

A mountain hiker sets off at 8.00 am one morning and walks up to a hut where
he/she stays the night. Next morning he/she sets of at 8.00 am and walks back
down the same track. Show that, independently of how fast or slowly he/she
walks there is at least one time on the two days when he/she is at the same
point on the track.
For an interesting history of ideas of the fourth dimension before Einstein
in art and popular culture ,including H G Wellss ideas about time travel, one
may consult the interesting book [23].

Minkowskis Spacetime viewpoint

In what follows we shall initially be concerned with the simplified situation

in which all motion is restricted to one space dimensions. Thus the position
vectors x have just one component. In this case, it is convenient to adopt a
graphical representation, we draw a spacetime diagram consisting of points we
call events with spacetime coordinates (t, x). The two-dimensional space with
these coordinates is called spacetime.

17

Passing to another frame of reference is like using oblique coordinates in

spacetime. However, according to Galilei, all observersuse the same time coordinate. Geometrically while the lines of constant x have different slopes in
different frames, the lines of constant time are all parallel to each other. This
means that two events (t1 , x1 ) and (t2 , x2 ) which are simultaneous in frame S
that is
must be simultaneous in frame S,
(t1 , x1 ) = (t1 ,1 ut1 ),

(43)

2 ) = (t1 , x2 ut2 )
(t2 , x

(44)

t1 t2 = 0 t1 t2 = 0.

(45)

thus
We say that in Newtonian theory simultaneity is absolute, that frame independent.
Lets summarize
(i) The Laws of Newtonian dynamics are invariant under
Galilei transformations

t = t,

= x ut.
x

(46)

(ii)
velocities

add

+ u.
v=v

(47)

(iii) Time is absolute.

(iv) Simultaneity is absolute.

10

Einsteins Principle of Relativity

We have discovered that no purely dynamical experiment can determine our

absolute velocity. If we are in a closed railway carriage moving uniformly we
cannot tell, by dropping particles etc, how fast we are traveling.
The natural question to ask is whether we can tell using experiments involving light. If this has speed c relative to some privileged inertial frame S
, (identified before Einstein with the mysterious Aetheror Ether), it should,
according to Galileo, have speed c u relative to a frame S moving with respect
to the aether. By measuring this speed it should be possible to determine u.
This was tried in the

10.1

Michelson-Morley Experiment

This is described clearly and in detail in Michelsons own words in [2]. Therefore
the present description will be brief. The light travel times T and Tk of light
moving in directions restively perpendicular and parallel to the motion along

18

arms of an interferometer of lengths L and Lk are measured. It was argued

that in the parallel direction (working in frame S)

Tk = Lk

cu

2Lk 1
1 i
.
Tk =
cu
c 1 uc22

(48)

On the other hand (working in frame S) it was argued that the total distance
the perpendicularly moving light has to travel is, by Pythagoras,
r
2T
2L
uT 2
(49)
L2 + (
) =
T = q
2
2c
c
c 1 u
c2

Thus, for example, if L = L| and T 6= T| we should be able to measure

u. However in 1887 the experiment carried out by the American Physicist
MIchelson and Morley [3] revealed that T = Tk !
Einstein drew the conclusion that no experiment, including those using light,
should allow one to measure ones absolute velocity, that is he assumed.
The Invariance of the Speed of Light The velocity of light is the same in
all frames of reference which are in uniform motion with respect to an inertial
frame.

19

In Einsteins own words

the same laws of electrodynamics and optics will be valid for all
frames of reference for which the equations of mechanics hold good.
We will raise this conjecture (the purport of which will hereafter be
called the Principle of Relativity) to the status of a postulate and
also introduce another postulate, which is only apparently irreconcilable with the former, namely that light is always propagated in
empty space with a definite velocity c which is independent of the
state of motion of the emitting body.
If Einstein is correct, then Galileis transformations cannot be correct. We need
a new transformations called Lorentz Transformations. They turn out to be
(proof shortly)
Lorentz Transformations

x ut
,
x = q
2
1 uc2

t u2 x
t = q c
.
2
1 uc2

(50)

Note that
(i) the time t gets transformed to t as well as x to x.
(ii) Simultaneity is no longer absolute

and hence

u
t1 t2
2 (x1 x2 )
t1 t2 = q
c
u2
1 c2
t1 t2 ; t1 t2 ,

if

x1 6= x2 .

(51)

(52)

(iii) If we take the non-relativistic limit c in which the speed of light is

infinite we Lorentz transformations (50) we recover the Galilei transformations
(34).

10.2

Derivation of the Lorentz Transformation formulae

We assume
(i) (ct, x
) are linear functions of (ct, x)
(ii) c2 t2 x2 = c2 t2 x
2 and hence the speed of light is invariant because
x = ct x
= ct.
In this first look at the subject we assume (ii) but in more sophisticated
treatments one makes considerably weaker assumptions. A precise statement
will be made later. Even at this point it should be clear that we are ignoring
trivial dilations or homotheties x
= x, t = t, for 6= 0 which obviously leave
the speed of light invariant. However we do not usually include these in the
set of Galilei transformations. We shall also treat space and time translations
20

t t + t0 , x x + x0 as trivial Thus it is sufficient to consider light rays

through the origin of spacetime (t, x) = (0, 0) We shall also regard as trivial
space reversal x = x, t = t and time reversal x = x, t = t.
Clearly (50) satisfy (i) and (ii). The converse is obtained by setting
  
 
ct
A B
ct
=
,
(53)
x
C D
x
with A > 0, D > 0 because we are excluding time reversal and space reversal.
Substitution gives
(Act B)2 (Cct Dx)2 c2 t2 x2 = 0.

(54)

Thus equating coefficients of t2 and x2 to zero, we get

(i) A2 C 2 = 1

A = cosh 1 ,

C = sinh 1

(55)

(ii) D B = 1

D = cosh 2 ,

B = sinh 2

(56)

For some 1 and 2 . Now equating the coefficient of xt to zero gives

(iii) AB = CD

cosh 1 sinh 2 = cosh 2 sinh 1

1 = 2 . (57)

Thus


ct
x

cosh
sinh

sinh
cosh



ct
x

(58)

Setting x = 0 allows us to see that the origin of the S frame satisfies x cosh =
ct sinh . But if this is to agree with x = ut, where u is the relative velocity,
we must have
u
(59)
= tanh := ,
c
where is called the rapidity. It follows that

and

1
cosh = q
1

u2
c2

:=

(60)

= .

(61)

sinh = q c
1

u2
c2

The quantities and do not, as far as I am aware, have individual names,

and perhaps for that reason is often, rather inelegantly, referred to as the
relativistic gamma factor. The use of the symbols and is both traditional
and almost universal in the subject. A Lorentz transformation of the form (50)
is often called a boost which is analogous to a an ordinary rotation. The analogue
of the verb rotating is, unsurprisingly, boosting. A useful relation, particularly
in Tripos questions, is
2 (1 2 ) = 1.
(62)
21

10.3

Relativistic velocity composition law

Of course the point is that velocities dont add. Suppose that

In frame S

x
= vt x
0 ,

(63)

then using the lorentz transformations (50) we have that

In frame S

Thus

v(t ux
x uv
c2 )
q
= q
+ x0 .
2
u2
1 c2
1 uc2

uv
v + u)c + t + x
0
x(1 + 2 ) = (
c

u2
,
c2

(64)

(65)

and hence
Relativistic velocity composition law

v =

u + v
.
v
1 + u
c2

(66)

Thus, for example, v = c v = c, which is the invariance of the speed of light.

10.4

*Observational for Einsteins second postulate*

This is that the velocity of light is independent of the velocity of its source.
Many high precision experiments give indirect evidence for its validity. In
addition, direct observational support for this includes
(i) The light curves of binary stars. De-Sitter [9] pointed out that if, for example,
two stars are in orbit around each other with orbital period T , then if light
coming from that portion of the orbit when the star is moving toward us had a
larger speed than when it was moving away from us, then light from an earlier
part of the motion might even arrive more than half an orbital period before
light coming from the intermediate portion of the orbit when it is neither moving
toward us or away from us. This would lead to significant distortion of the plot
of luminosity or of velocity against time.
Consider, for example, the case when we are in the plane a circular orbit of
radius R and period P whose centre is a large distance L from us. The relation
between time of emission te and time of observation to expected on the basis of
Newtonian theory is, since R << L,
to = te +

e
L R sin 2t
P
.
e
c + v cos 2t
P

(67)

In this formula, the quantity v is the extra velocity supposed to be imparted

to the photons moving toward us. According to the Ballistic theory of light of
Newton and Galilei we would apply the usual rules for particles of speed c in

22

the rest frame of the moving stars. Thus we expect v = vorbital = 2R

P , but in
general it could be much smaller. Since L >> R and c >> v we have
to te +

2te
2te Lv
L R
sin
+ 2 sin
.
c
c
P
c
P

(68)

The observed Doppler shift is given by

dt0
v
2te
Lv2
2te
= 1 cos
+ 2 sin
.
dte
c
P
c P
P

(69)

If unless Lv2
P c2 is small there will be significant distortion of the light curves.
Indeed to may not be a monotonic function of te , in which case, te will not be a
unique function of to . In other words, pulses from different phases of the orbit
may arrive on earth at the same time to . Such effects have not be seen.
De Sitter himself considered the binary star -Aurigae.
One example sometimes quoted is, the binary star Castor C. It is 45 light
years away and has a period of .8 days. The stars have vorbital = 130Kms1.
The effect should be very large, but the light curves of the two stars are quite
normal [11]. Using pulsating X-ray sources in binary star systems, Brecher [24]
was able to conclude that Einsteins second postulate was true to better than 2
parts in a thousand million
v
< 2 109 .
vorbital

(70)

This is certainly an improvement on Zurhellen, who in (1914) obtained a limit

of 106 using ordinary binary stars [29].
(ii) The time of travel over equal distances of gamma rays emitted by a rapidly
moving positron annihilating with a stationary electron can be measured as they
are found to be equal [12].
(iii) A similar measurement can be done using the decay of a rapidly moving
neutral pion which decays into two gamma rays [14].

10.5

Light in a medium: Fresnel Dragging

In a medium, the velocity of light is reduced to nc , where n 1 is called the

refractive index. In general n may depend upon wavelength but here we will
neglect that effect. Fresnel proposed, in the 1820s, measuring the speed of light
in a stream of water moving with speed u relative to the laboratory. Naive
Newtonian theory would give a speed
c
+u
n

(71)

but experiments by Fizeau in 1851 using the toothed wheel method did not
agree with this. If we use the relativistic addition formula in the case that uc is
small we get instead
1
c
(72)
+ u(1 2 ) + . . .
n
n
23

which does agree with Fizeaus experiments. The factor (1 n12 ) is called Fresnels dragging coefficient and had in fact been proposed earlier by the French
physicist Fresnel around 1822 using an argument based on wave theory. The
experiment was repeated after Einstein had proposed his theory by the 1904?
Nobel prize winning Dutchman Zeeman ( ).

10.6

Composition of Lorentz Transformations

We could just multiply the matrices but there is a useful trick. We define
x = x ct.

(73)

Thus
x = 0 we have a right moving light ray
x+ = 0 we have a left moving light ray

Now Lorentz transformations (50) take the form

x+ = e x+
x = e

x+

x+ = e+ x
+ ,

x = e

x
+ .

(74)
(75)
(76)
(77)

We immediately deduce that the inverse Lorentz transformation is given by

setting u u, ,i.e. the inverse of (50 ) is
Inverse

Lorentz

x
+ ut
,
x= q
2
1 uc2

Transformations

t + u2 x
t= q c
.
2
1 uc2

(78)
such that we get from
Now consider three frames of reference S, S and S
by boosting with relative
S to S by boosting with velocity u1 and from S to S

velocity u2 . To get from S to S we have to boost with relative velocity u3 . If

1 , 2 , 3 are the associated rapidities, we have

Thus

x
= e1 x ,

(79)

= e
x

tx .

(80)

= e3 x ,
x

(81)

i.e.
rapidities add

3 = 1 + 2 .

(82)

Using a standard addition formula for hyperbolic functions

tanh 1 + tanh 2
.
1 + tanh 1 tanh 2
That is we re-obtain the velocity composition formula:
tanh 3 =

u1
+ u2
u3
= c u1 uc2 .
c
1 + c2

24

(83)

(84)

10.7

Velocity of light as an upper bound

Suppose that |u1 | < c and |u2 | < c, then |u3 | < c.
Proof Since the hyperbolic tangent function is a one to one map of the real line
onto the open interval (1, +1), we have
|u1 | < c < 1 < ,

(85)

|u2 | < c < 2 < .

(86)

< 1 + 2 < |u3 | < c.

(87)

Thus
Thus no matter how we try, we cannot exceed the velocity of light.

10.8

*Super-LuminalRadio sources*

An interesting apparent case of super-luminal velocities but which is perfectly

explicable without invoking the existence of anything moving faster than light,
has been discovered by radio astronomers. What are called quasars or quasistellar radio sources exhibit jets of matter symmetrically expelled from a dense
central region probably associated with a black hole. For the sake of a simple
first look we assume that we can use the geometry of Minkowski spacetime
despite the great distances and that the central quasar is located a distance L
away from us. We shall also assume that there is a frame in which both the
central quasar and ourselves are at rest
We assume, in the simplest case possible, that the matter in the jets are
expelled at right angles to our line of sight at time t = 0 and therefore at time
t = to the material in the jets has have travelled a distance vte . Light or radio
waves coming from the jets will arrive here at time
to = te +

1p 2
L + v 2 t2e .
c

(88)

The angle subtended is, for small angles

=

vte
.
L

(89)

The rate of change with respect to the observation time is

d
v dte
=
.
dt0
L dto
We have

Thus

1 
to +
te =
2
1 vc2

t2o + (

L2
v2

+ t20 )(1 2 .
2
c
c


c
dte
to
1 
q
1
+
.
=
2
dto
v t2 + L2 ( 1 1)
1 vc2
o
v2
25

(90)

(91)

(92)

For large to we get

=

v 1
L1

v
c

(93)

dte
Clearly if v is close to c, then dt
can be much bigger than unity. Thus the size
o
of the effect is much larger than ones naive Newtonian expectations. If the jet
makes an angle with the line of sight we obtain

1
v sin

L 1 v cos
c

(94)

The existence of such apparent superluminal motions was suggested by the

present Astronomer Royal in 1966 while a research student in DAMTP[22]. Just
over 4 years later, in 1971, the radio astronomers Irwin Shapiro amd Marshall
Cohen and Kenneth Kellerman astronomers found, using very long base line
interferometry (VLBI) such jets, changing in apparent size over a period of
months, in the quasars 3C273 and 3C279. Nowadays the observation of such
apparently super-luminal sources is commonplace.

10.9

The two-dimensional Lorentz and Poincar

e groups

Clearly Lorentz transformations, i.e. boosts in one space and one time dimension, satisfy the axioms for an abelian group (closure under composition,
associativity and existence of an inverse) which is isomorphic to the positive reals under multiplication (one multiplies e ) or all the reals under addition (one
adds ). This is completely analogous to the group of rotations, SO(2) in two
spatial dimensions. The standard notation for the group of boosts is SO(1, 1).
If we add in the abelian group of time and space translation translations
t t + t0 ,

x x + a,

(95)

we get the analogue of the Euclidean group plane, E(2) which is called the
Poincare group and which may be denoted E(1, 1).

11

The invariant interval

Consider two spacetime events (ct1 , x1 ) and (ct2 , x2 ) in spacetime. The invariant
interval between them is defined by
2 = (t1 t2 )2

(x1 x2 )
.
c2

(96)

The name is chosen because 2 is invariant under Lorentz transformations

(50). This is because of the linearity

 


ct1 ct2
cosh sinh
ct1 ct2
=
.
(97)
sinh cosh
x
1 x
2
x1 x2
26

Now there are three cases:

Timelike separation

2 > 0 |t1 t2 | >

|x1 x2 |
.
c

(98)

In this case a particle with v < c can move between the two events.
Lightlike separation

2 = 0 |t1 t2 | =

|x1 x2 |
.
c

(99)

In this case a light ray or particle with v = c can move between the two events.
|x1 x2 |
.
c
In this case no particle with v < c can move between the two events.
Spacelike separation

11.1

2 < 0 |t1 t2 | <

(100)

Timelike Separation

In this case there exists a frame S in which both events have the same spatial
position, x1 = x2 2 = (t1 t2 ) = |t1 t2 |, where we have fixed the
sign ambiguity to make positive.
Proof We need to solve for the equation



cosh sinh
ct1 ct2
= ( ct1 ct2
sinh cosh
x1 x2

0 ) tanh =

x1 x2
.
ct1 ct2
(101)

Clearly a real solution for exists.

Strictly speaking, this is all we can say purely mathematically. However we
can say more if we accept the physical clock postulate that a physical clock
at rest in frame S would measure an elapsed time t1 t2 . Then we can identify
with the time measured by a clock at rest in S and passing between the two
events. We call this the proper time between the two events.
At this stage it may be helpful to recall the definition of the second according
to the Bureau International des Poids et Mesures (BIPM) who are responsible
for defining and maintaining the International System of Units (SI units). Traditionally 1/86 400 of the mean solar day, it has been since 1960 had the definition
The second is the duration of 9 192 631 770 periods of the radiation
corresponding to the transition between two hyperfine levels of the
ground state of the caesium 133 atom.
The definition of the metre is formerly defined in 1960 of the wavelength of
krypton 86 radiation but in 1983 the BIPM declared that
The metre is the length of the path travelled by light in vacuum
during a time interval of 1/299 792 458 of a second.
Note that not only does the BIPM completely accept Einsteins Principle of the
invariance of light but also that the velocity is independent of wavelength.
27

11.2

Spacelike separation

In this case there exist a frame S in which

both events are simultaneous, t1 =
t2 2 = c12 (
x1 x
2 )2 |
x1 x
2 | = c2 2 .
Proof This runs along the same lines as above.

By analogy with the clock postulate, we assume that 2 is the distance

measured between the two events in the frame S in which are both simultaneous.

11.3

Time Dilation

Since

(x1 x2 )2
+ (t1 t2 )2 = 2 ,
(102)
c2
r
(x1 x2 )2
| |.
(103)
|t1 t2 | = 2 +
c2
Thus varying over all frames we see that is the least time between the two
events as measured in any frame. Moreover
u (
x1 x
2 )
t1 t2
+ q
|
|t1 t2 | = | q
2
c 1 u2
1 uc2
2
c

Thus the time between the two events in frame S is

t1 t2 = q
.
2
1 uc2

(104)

(105)

In other words, moving clocks appear to run more slowly than those at rest.
11.3.1

Muon Decay

It was first demonstrated by the physicists Rossi and Hall working in the USA
in 1941 [36] that one must use time dilation to account for the properties of
elementary particles called muons which arise in cosmic ray showers.
Cosmic rays, mainly protons, strike the earths upper atmosphere at a height
of about 16Km and create pions (mass m = 140 MeV). The pions rapidly
decay to muons (mass m = 105MeV) and anti-muon neutrinos (mass very
nearly zero)
+ + +
(106)
with lifetime = 2.6 108 s c = 7.8m. The muons then decay to positrons
(mass me = .5MeV) and electron and muon anti-neutrinos
+ e+ + e +

(107)

with lifetime = 2.1 106 s c = 658.654m.

In other words, in this time as measured in the laboratory, a muon, or indeed
any other particle, should be able to travel no more than about .66 Km.. How
28

then can it be detected on the earths surface 16 Km. away from where it was
produced?
The point is that because of the effect of time dilation the time of decay in
the rest frame of the earth is
1
2.2 106 s q
1

v2
c2

(108)

where v is the speed of the muon. It only needs

i.e.

1
q
1

v2
c2

> 24,

r
r
1
v
v
1
1+ <
c
c
24

To produce the necessary amount of time dilation. Since

(109)

(110)
v
c

1 this requires

v
1
1 1
(1 ) <
=
.
2
c
2 24
1152

11.4

(111)

Length Contraction

Now consider two spacelike separated events. Since

(x1 x2 )2
+ (t1 t2 )2 = 2 ,
c2
p
p
|x1 x2 | = c2 2 + (t1 t2 )2 c2 c2 2 .

(112)
(113)

Thus
the distance between two spacelike separated events is never less than

c2 2 . In fact if u is the relative velocity of S and S ,

thus

x
1 x
2
u (t1 t2 )
|x1 x2 | = | q
+ q
|,
2
c 1 u2
1 uc2
c2

c2 2
|x1 x2 | = q
.
2
1 uc2

We call

c2 2 the proper distance between the two events.

11.5

The Twin Paradox: Reverse Triangle Inequality

(114)

(115)

According to this old chestnut, timorous stay at home Jack remains at rest in
frame S for what he thinks is a propertime 3 , while his adventurous sister
Jill takes a trip at high but uniform speed u1 (with respect S) to the nearest
29

star, Alpha Centauri, (a distance R 4light years according to Jack) taking

propertime 1 and then heads a back, at speed u2 taking what she thinks is
proper time 2 . Their world lines form a triangle with timelike sides whose
proper times are 1 , 2 and 3 . Jack reckons that the two legs of Jills journey
take times
R
R
t1 =
,
t2 =
,
3 = t1 + t2 .
(116)
u1
u2
But

1
t1 = q
1

u21
c2

1
t2 = q
1

Evidently we have the

Reverse triangle inequality

u22
c2

3 > 1 + 2 .

(117)

(118)

In other words, by simply staying at home Jack has aged relative to Jill. There
is no paradox because the lives of the twins are not strictly symmetrical. This
might lead one to suspect that the accelerations suffered by Jill might be responsible for the effect. However this is simply not plausible because using identical
accelerating phases of her trip, she could have travelled twice as far. This would
give twice the amount of time gained.
11.5.1

*Hafele -Keating Experiment*

This effect was verified in (1972) [16] in what is called the Hafele-Keating experiment. Atomic clocks were flown around the world in opposite directions. On
their return they had lost, i.e. measured a shorter time, relative to an atomic
clock left at rest. The full interpretation of this result is complicated by the fact
that to work this out properly one must also take into account the Gravitational
Redshift effect due to General Relativity. When all is said and done however,
a fairly accurate verification of the time dilation effect was obtained. Before
this experiment, discussion of the twin paradox and assertions that it implied
that special relativity was flawed were quite common. Since The Hafele-Keating
experiment, and more recently the widespread use of GPS receivers, which depend on the both time dilation and the gravitational redshift, the dispute has
somewhat subsided. For an interesting account of the confusion that prevailed
in some quarters just before the experiment see [15]

11.6

Accelerating world lines

If one has a general particle motion x = x(t) with a non-uniform velocity v = dx

dt ,
we get a curve in spacetime with coordinates (ct, x(t)). We can work out the
proper time d elapsed for a short time interval dt by working infinitesimally
r
v2
1 2
2
2
d = dt 2 dx ,
i.e. d = dt 1 2 .
(119)
c
c

30

The total proper time measured by a clock moving along the world line is
Z
Z t2 r
Z t2
v2
dt 1 2
d =
dt = t2 t1 .
(120)
c
t1
t1
From this we deduce
Proposition Among all world lines beginning at t1 at a fixed spatial position x
and ending at t2 at the same spatial position x, none has shorter proper time
than the world line with x constant.

12

Doppler shift in one space dimension

It was first suggested by Christian Doppler(1803-1853) in 1842 that the colour of

light arriving on earth from binary star systems should change periodically with
time as a consequence of the high stellar velocities [26]. The radial velocities
of a star was first measured using this in the winter of 1867 by the English
astronomer William Huggins(1824-1910) at his private observatory at Tulse Hill,
at that time, outside London. Huggins compared the F line in the spectrum of
light coming from the Dog star, Sirius, and compared it with the spectrum of
Hydrogen in Tulse Hill. The light was shifted toward the red and he deduced
that Sirius was receding from Tulse Hill with a speed of 41.4 miles per second.
Taking into account the motion of the earth around the sun, he concluded that
the radial velocity of Sirius away from the solar system was 29.4 miles per
second.
Consider a wave travelling to the right with speed v in frame v. We can
represent the wave by function (t, x) describing some physical property of the
wave of amplitude A the form
x
(121)
= A sin(t kx) = A sin(2(f t )),

where
the frequency

f=

and the wavelength

2
.
k

(122)

The speed v of the wave is

v = f =

.
k

(123)

In frame S


= A sin q
1+

u2
c2

(t

u
x
(
x + ut) 
= A sin(
t kx)
) kq
2
2
c
1 u

(124)

c2

Thus the angular frequency and wave number in frame S are

k u2
k = q c .
2
1 uc2

ku
,

= q
2
1 uc2
31

(125)

The quantities and k are called the angular frequency and wave number
respectively.
Consider the special case of a light wave for which v = f = c = k . One
has
s
(1 uc )
1 uc
,
(126)

= q
=
2
1 + uc
1 u
c2

(1 uc )
k = k q
=
2
1 uc2

Thus

f = f

1
1+

u
c
u
c

1
1+
s

u
c
u
c

1+
1

(127)

u
c
u
c

(128)

Thus if the emitter and receiver recede from one another the wavelength is
increased and the frequency is decreased. One says that the signal is redshifted
because red light has a longer wavelength than blue light. If the emitter and
receiver approach one another the signal is blue-shifted. A quantitative measure
z called the red shift is given by

= 1 + z.

12.1

(129)

*Hubbles Law*

In 1929, the American Astronomer Edwin Hubble discovered that that the universe is in a state of expansion. The light coming from Galaxies, lying outside
our own Milky Way with distances L > .5Mega parsecs was found to be systematically red shifted rather than blue shifted, moreover the further way they are
the greater is their radial velocity vr . Quantitatively,
Hubble s law states that

z = vr = H0 L ,

(130)

where H0 is a constant of proportionality called the Hubble constant . Hubble

estimated that H0 was about 500 Km per second per Mega parsec. Currently
H0 is believed to be rather smaller, about 70 Km per second per Mega parsec.
By now, galaxies have been observed moving away from us with redshifts
greater than 4. If interpreted in terms of a relative velocity we have
1+
1

u
c
u
c

= (1 + z)2 1

u
1
2
=
=
.
2
c
1 + (1 + z)
13

(131)

Note that while a completely accurate account of Hubbles law can only be
given using General Relativity, for which see for example, [34], as long as one is
well inside the
c
7
(132)
Hubble radius
6000M pc
H0
7 Often

incorrectly referred to be astronomers as the horizon scale.

32

and considers times much shorter than the

1
10Giga years
H0

Hubble time

(133)

then spacetime is sufficiently flat that it is safe to use the standard geometrical
ideas of special relativity.

13

The Minkowski metric

In two spacetime dimensions

Thinking of

Lorentz transforms leave invariant c2 t2 x21 .

x=

ct
x1

(134)

as a position vector in a 2=1+1 dimensional spacetime

indefinite inner product

we can define an

x x = c2 t2 x21 = xt x

(135)

where what is called the

Minkowski metric

+1 0
0 1

= t .

(136)

is a symmetric matrix10 Under a Lorentz transformation

x x = x
with
=

cosh
sinh

sinh
cosh

(137)


(138)

Now under a Lorentz transformation, i.e. under (137)

x x xt t x = xt x,

x,

(139)

we must have11
t = .

(140)

Obviously is analogous to a rotation matrix but it is not the case that

t = 1 , as it would be for an orthogonal matrix, rather (in two dimensions)
since 1 () = () 6= t () = ().
8 From now on we shall use x for the space coordinate. This notation is consistent with the
1
fact that in higher spacetime dimensions there will be more than one spacetime coordinate.
9 From now on an x without a subscript or superscript should be thought of as a column
vector.
10 t denotes transpose and we denote this indefinite inner product with a so-called centred
do . You should distinguished this from a lowered dot . which will be used for the ordinary
dot product of ordinary vectors in three dimensions as in x.y.
11 Although in two spacetime dimensions = t , this is no longer true in higher dimensions.
For that reason we prefer to include the transpose symbol t on the first .

33

13.1

Composition of Lorentz Transformations

If we do two Lorentz transformations in succession we have

x (1 ) x (2 )(1 ) x = (3 ) x,

(141)

(3 ) = (2 )(1 ).

(142)

where
In two spacetime dimensions we have
3 = 1 + 2 = 2 + 1 ,

(143)

i.e. Lorentz transformations, like rotations, are commutative.

It is not true in general in higher dimensions that either Lorentz transformations or orthogonal transformations commute.

14

Lorentz Transformations in 3 + 1 spacetime

dimensions

We now consider 4-vectors

x=

ct
x

(144)

where an ordinary vector x is conveniently thought of a column vector

x1
x = x2 .
x3

(145)

The Minkowski indefinite inner product is

x x = c2 t2 x.x = c2 t2 x2 = c2 t2 |x|2 = xt x,
with

1
0
=
0
0

(146)

0
0
0
1 0
0
.
0 1 0
0
0 1

(147)

(148)

There are differing conventions for the Minkowski metric. Here we have chosen
the mainly minus convention for which timelike vectors have positive length
squred but equally popular for some circumstances is the mainly plus convention
obtained by changing .
A general Lorentz transformation satisfies (137) but in general 6= t .
Example
=

1
0

34

0
R

with R1 = Rt a three-dimensional rotation or orthogonal matrix.

The general Lorentz transformation is rather complicated; it may contain up
to 6 arbitrary constants. It simplifies if we dont rotate spatial axes. Roughly
speaking, three of them correspond to a general; rotation of the spatial axes
and the other three to the three components of the relative velocity of the two
frames S and S. Example

ct
cosh
x1 sinh
=
x2
0
x3
0

sinh
cosh
0
0

0
0
1
0

0
ct
0x
1
.
x
2
0
1
x
3

(149)

In this case S moves with respect to S along the x1 axis with velocity u = tanh .

14.1

The isotropy of space

The foregoing work implicitly assumes that space is isotropic and this assumption calls for some comment. The universe we see about us is certainly not
completely isotropic. At the level of the solar system and the galaxy, there
are gross departures from isotropy. At larger scales, for example the distribution of galaxies, quasars, radio galaxies etc there is certainly good evidence for
statistical isotropy but there are significant departures from complete spherical
symmetry about us. In particular, as we shall describe in greater detail later,
the most distant parts of the universe that we have direct optical access to, the
cosmic microwave background (CMB) is isotropic only to a part in one hundred
thousand or so.
This might lead one to postulate that the metric of spacetime should be
anisotropic. Physically such an anisotropy could manifest itself in at least two,
not completely unrelated, ways.
(i) The speed of light could depend upon direction
(ii) The dynamics of particles could be anisotropic, for example, rather than the
masses m of particles just being scalars, they could be tensors mij and Newtons
second law might read
d2 xi
mij 2 = Fi .
(150)
dt
One clearly has to be careful here that one cannot eliminate these effects
by redefinition of lengths or times. For example if the mass tensor of every
particle were proportional then we could eliminate any interesting effect as far
as particles were concerned by using linear transformations of the spatial coordinates to diagonalize mij . The same can be said for the motion of light. It
only makes sense to say it is anisotropic relative to some choice of clocks, for
example caesium clocks, otherwise we could always declare it to be isotropic by
choice of units.
We can say the above in a slightly different way and, in doing so, make
contact with some basic ideas in General Relativity. A key ingredient of Special
35

Relativity is that there is just one metric , and hence just one fundamental speed,
which gives a universal upper bound for the velocity of all types of matter. We
say that the Minkowski metric is universal. The ideas which have just been
described extend to situations where gravity is important and form the basis of
Einsteins Equivalence Principle12 .
Experimentalists have not be slow to test these assumptions and there exist
some extremely stringent bounds on departures from isotropy. For example,
using rotating interferometers Brillet and Hall [46] found that fractional length
changes
l
= (1.5 2.5) 1015 ,
(151)
l
completely consistent with isotropy.
Perhaps even more impressive are what are called Hughes-Drever experiments which make use of the earths daily rotation. Thy look for any twenty
four hour periodicity in the Zeeman effect.

14.2

Some properties of Lorentz transformations

It is clear from the definition (140) that the the composition 2 1 of two Lorentz
transformations is again a Lorentz transformation. Moreover taking determinants gives
det = 1,
(152)
Lorentz transformations are (up to a sign) uni-modular. Moreover the inverse
is given by13
1 = 1
(153)
It follows that Lorentz transformations form a group, called the Lorentz
group, written as O(3, 1) or O(1, 3). If we insist that det = 1 , we get the
special Lorentz group SO(3, 1) or SO(1, 3).

15

Composition of non-aligned velocities

A particle moves with respect to S with velocity v what are the velocities with
respect to S? We have

ct
ct
ct
ct(cosh + vc1 sinh
1 v1 t
x
x1 t(v1 cosh + c sinh )
.
=
=
v2 t
v2
x
2
x2
v3 t
x
3
tv3
x3
We read off
v1 =

v1 + u
.
1 + vc12u

12 Technically,

(154)

(155)

what we are referring to is the Weak Equivalence Principle

prefer not to use the fact that in a standard orthonormal basis = 1 because this
is a basis dependent statement.
13 We

36

v2
v2 =
=
cosh + vc1 sinh
v3 =

15.1

v3
=
cosh + vc1 sinh

v2

q
1

u2
c2

1
1 + uv
c2
q
2
v3 1 uc2

1+

uv1
c2

(156)

(157)

Aberration of Light

Suppose that we have light ray making an angle

with the x1 axis in frame S.
What angle does it make in frame S? By choice of axes we can suppose that
the ray moves in a x
3 = 0 plane. Thus we put v1 = c cos
, v2 = c sin
and
v3 = 0. We get
r
u2 c sin

c cos
+u
1

,
v
=
.
(158)
v1 =
2
u
u
2

1 + c cos
c 1 + c cos
We check that v12 + v22 = c2 as expected. Thus we may put v1 = c cos ,
v2 = c sin . Moreover
r
u2 sin

tan = 1 2
.
(159)
c cos
+ uc
Example show that (159) may be re-written as
r

cu
tan( ) .
tan( ) =
2
c+u
2

(160)

If = tan1 ( xx21 ) and = tan1 ( xx21 ), then we shall refer to the map S 2
given by (160) supplemented = as the aberration map.
S 2 : (, ) (
)
Stereographic coordinates are defined by
= ei

1
,
tan 2

(161)

so that

dd
(162)
2.
(1 + )
We may express the aberration map as a simple dilation of the complex plane
r
c+u

=
,
(163)
cu
d2 + sin2 d2 = 4

which is clearly conformal. It belongs to group of Moebius transformations of

the sphere, P SL(2C ), which become fractional linear transformations of the
complex plane
a + b
,
ab cd = 1,
(164)

c + d
with a, b, c, d C .
37

15.2

* Aberration of Starlight*

If
= then tan = uc . If u << c we get 2 u2 . This more or less what
one expects on the basis of Newtonian Theory.
For the case of the earth moving around the sun we choose S be a frame at
rest with respect to the sun and S to be one at rest with respect to the earth.
Thus u 30Kms1 . We deduce that the apparent positions of stars should
change over a 6 month period with an amplitude of about 104 radians. Now
there are 360 degrees in a full circle and 60 minutes of arc in each degree and
60 seconds of arc in each minute of arc. Thus, for example, there are 21, 600
arc minutes in a full circle. James Bradley, Savilian Professor of Astronomy
at Oxford 14 and successor in 1742 of Edmund Halley as Astronomer Royal
developed the technology to measure positions to better than an arc minute
and was thus able to prove for the first time to prove that the earth moves
round the sun. In fact he announced in 1728 [4] that the apparent position of
the star Eltanin (-Draconis) and all adjacent stars partake of an oscillatory
motion of amplitude 20.4 seconds of arc.
The plane of the earths orbit is called the plane of the ecliptic . Spherical
polar coordinates with respect to the normal of the ecliptic are called right
ascension (analogous to longitude) and declination, (analogous to latitude and
measured from the celestial equator).
For stars in the ecliptic, i.e. with zero declination the apparent motion due
to abberation is along a straight line. For stars whose direction is perpendicular
to the ecliptic, i.e with declination 2 , it is circular. At intermediate declinations
it is an ellipse.
Remark Bradleys observations do not contradict Einsteins Principle of Relativity since in effect he measured the velocity of the earth relative to what are
called the fixed stars.
In the early 1930s shortly after Edwin Hubbles demonstration of the expansion of the universe, the astronomers Str
omberg and Biesbroeck working in the
USA pointed out that, as expected according to Special Relativity, observations
of galaxies believed then to be 70 million light years away, in the constellation of Ursa Major , which, by virtue of Hubbles law, are moving away from
us at speeds of 11,500 Km per sec exhibit the same amount of annual aberration [30, 31] as do nearby stars. In fact, we would now assign these galaxies a
greater distance because they were using Hubbless value for his constant 500
Km per second per Mega parsec. This is almost a hundred times larger than
the currently accepted value.
Similar remarks have been made by Heckmann [32].
In fact, nowadays astronomers use not the fixed stars but rather about 500
distant stellar radio galaxies to provide a fundamental inertial frame of reference called the International Celestial Reference Frame (ICRF) centred on the
barycentre, that is centre of mass or centroid of the solar system.
14 At

an annual salary of 138 5s 9d

38

Remark Bradley was also able to establish that the earths axis nods or nutates.
Much later in the 1830s the German astronomer Bessel established that the
apparent positions of nearby stars alter because over a 6 month period because
we see them from two ends of a baseline given by the diameter of the earths
orbit around the sun. This effect is called stellar parallax.

15.3

Water filled telescopes

Bradleys explanation of aberration gave rise to various controversies, partly

because it is difficult to understand on the basis of the wave theory. In particular
Boscovich suggested that one would obtain a different result if the telescope tube
is filled with water, since the speed of light in water is different from that in
vacuo or what is almost the same, in air. Indeed Boscovich hoped to measure
the speed in water in this way.
Eventually, in 1871 the then Astronomer Royal George Biddell Airy put the
matter to the test [48]. During the spring and autumn of that year he observed
the same star Draconis as Bradley had. In March the correction for aberration
was -19.66 arc secs while in September it was 19.74 arc secs. These results agreed
with those of Bradley, within the errors and showed that the presence of the
water was irrelevant. From a modern point of view, this is obvious because
the entire effect is due to the passage to a frame moving with the earth. In
that frame light from distant stars enters the telescope tube at the aberrated
angle and, provided the surface of the water is perpendicular to the axis of the
telescope tube, it will suffer no refraction or further change of direction.

15.4

Headlight effect

We deduce from (159) that

lim = 0,
uc

6= .

(165)

Thus a photons emitted by a rapidly moving source a thrown forward and

occupy a very small cone around the direction of motion.
This effect has been invoked to explain why the two sources seen in many
quasars have such different apparent brightnesses. Typically the quasars seem
to emit from a dense central source, possibly a black hole, two blobs of plasma
(i.e.highly ionized gas) in moving n opposite directions. The idea is that one is
moving toward us and one away. Light coming from the latter is highly beamed
toward us and hence appears much brighter than the other, the light of which is
beamed away from us. Relativistic gamma factors as higher as 10 are quoted
in the astrophysical literature.

15.5

Solid Angles

In order to quantify the headlight effect, we note that the aberration map preserves angles but not areas. The infinitesimal area element on S 2 is the same
39

as the solid angle

d2 = sin dd.

(166)

Under the aberration map

2

1 vc2
d d =
d2 .
(1 vc cos )2
2

15.6

(167)

*Celestial Spheres and conformal transformations*

The set of light rays passing through the origin O of the frame of reference S
and any given time t constitutes the celestial sphere of an observer at O. The
celestial sphere may be coordinatized by a spherical polar coordinate system
(, ) symmetric with respect to the direction of motion, i.e. the x3 axis. Thus
exists for the observer sit = tan1 ( xx23 ). A similar coordinate system (
, )
of the frame of reference S.
The aberration formulae,
uated at the origin O
i.e.(159) or better (160) and = provide a map from one celestial sphere to
the other. This map allows one to relate the visual perceptions of one observer
to the other. A short calculation reveals that infinitesimally
d
d

=
,
sin
sin

d = d .

(168)

This implies that aberration map preserves angles. To see why, note that,
these would be calculated using the metrics or infinitesimal line elements
ds2 = d2 + sin2 d2 ,

and

Thus
ds2 = (

15.7

d
s2 = d
2 + sin2
d2 .

sin 2 2
) d
s .
sin

(169)

(170)

*The visual appearance of rapidly moving bodies*

Formula (168) has a striking consequence which was only noticed 55 years after
Einsteins paper of 1905 by Terrell[18] and by Roger Penrose independently[19].
Previously it had been believed that a something seen as a sphere or a cube
in frame S say would, because of length contraction, be seen as an ellipsoid
The truth is more complicated, because from (168) it
or a cuboid in frame S.
follows that the aberration map is conformal, it preserves angles.This implies
that the cube would appear rotated rather than merely squashed in the direction
of motion. It also implies that a sphere always appears as a sphere.
Nowadays, there are a number of simulations, using ray-tracing techniques,
of what would be seen for example by a relativistic tram passing in front of the
patent office in Berne. See for example
http://www.anu.edu.au/Physics/Searle

40

15.8

Transverse Doppler effect

We consider a wave moving at the speed of light of the form

= A sin t

x1 cos + x2 sin

.
c

(171)

Substitution of the Lorentz transformations gives expressions for the angles

in frame S which are equivalent to the aberration formulae derived earlier to
together with the relation

1 u cos
c
.

= q
2
1 uc2

(172)

If = 0, we recover our previous result. If = 2 we find, contrary to what

is predicted by Newtonian theory, there is a frequency change. To interpret
what is going on, we think of the a photon emitted in frame S with frequency
e =
and received in frame S with frequency o = in a direction (according
to S) perpendicular to the direction of motion. We have
r
u2
(173)
0 = e 1 2 .
c
The observed frequency 0 is smaller than the emitted frequency frequency
precisely be a time dilation factor.
This effect was verified experimentally with great precision by Ives and Stillwell in 1938 using the light emitted by moving atoms [10].

15.9

*The Cosmic Microwave Background*

One of the most striking applications of the transverse Doppler effect formula is
to the cosmic microwave background or CMB. This was first observed, using a
ground based radiometer, by the American physicists Arno Penzias and Robert
Wilson in 1965 [37] and led to their Nobel Prize in 1978. They had discovered
an almost perfectly isotropic background of microwave photons with a thermal
or Planckian spectrum with temperature T 3K. This bath of thermal photons
is believed to be spatially uniform and to fill the entire universe and to a relic
from an earlier and much hotter phase called the Hot Big Bang .
In a certain sense it defines an absolute frame of rest, reminiscent of the old
aether concept and, superficially, one might think that this contradicts Einsteins
Principle of Relativity. However, as with the fixed stars observed by Bradley,
this is not so.
Later observations, using satellite and aircraft and balloon borne radiometers
observations have shown that the solar system is in motion relative to the CMB.
At any given time, the temperature distribution observed To is not exactly

41

isotropic but varies with angle to the direction of motion as

To () =

q
1

u2
c2
u cos
c

Te ,

(174)

where Te is the temperature one would see at rest.

We will derive (174) shortly, in the mean time we note that since uc is small
we get a small dipole term
u cos
T =
Te .
(175)
c
The observations[28] exhibit a term of magnitude 3m K in the direction the
constellation of Hercules in the souther hemisphere, declination 7 deg, right
ascension 11 hours 12 min, corresponding to a velocity of about 365Kms1 . In
fact it is also possible to detect the earths annual motion around the sun which
at 30Kms1 is a factor of ten or so smaller.
We turn now to the justification of (174). If k is Boltzmanns constant, a
Black Body distribution of photons at temperature Te has
4fe2 hdfe d2 e dAdte
hfe
1 + exp( kT
)
e

(176)

photons of frequency fe crossing area dA in time dt in frequency interval dfe

and solid angle d2 e , where h = 2 is Plancks constant.
Now using the aberration and Doppler shift formulae derived earlier dA is
the area perpendicular to the motion and
fe d2 e = fo d2 0 ,

fe dte = f0 dt0 ,

(177)

we see that the moving observer sees a Planckian spectrum in direction

with temperature T0 such that
fo
fe
=
,
Te
To

(178)

which gives (174).

16

* Kinematic Relativity and the Milne Universe*

The Oxford astronomer Edward Arthur Milne (1896-1950) brother of the childrens writer Alan Alexander Milne (11882-19560 was dissatisfied with Einsteins
theory of gravity, General Relativity and the resultant cosmological models it
gives rise to and proposed a theory(Kinematic Relativity and a cosmological
model of his own now called the Milne universe. While nowadays his theory
has largely been rejected, his simple cosmological model still provides valuable
42

insights. For us it is interesting because it illustrates that many of the ideas

associated with the expanding universe are implicit in special relativity, and
moreover Milnes model ca be obtained as an approximation to the behaviour
of the exact, and highly complicated, equations of General Relativity in the
limit that the universe is expanding very fast and so that gravitational effects
can be ignored.
The essential points of Milnes ideas were that (i) Spacetime was the same as
Minkowski spacetime E 3,1 but a particular creation event, lets us pick it as the
arbitrary origin of Minkoswki spacetime, was the origin of violent explosion such
that the galaxies are now moving away from it in rapid motion with constant
speed 15 .
(ii) Astronomical and Atomic, i.e. and laboratory, time measurements need not
necessarily agree. In other words, he questioned the clock hypothesis.
In order to link these ideas we introduce a set of co-moving coordinates
, , , moving with the galaxies. The coordinate is just the proper time
along each galaxys world line. The coordinates , , label the individual
galaxies. Because the galaxies move no faster than light the are confined to the
interior of the future light cone of the origin:
t2 x2 = 0.
Each galaxy cuts a surface of constant proper time
p
= t2 x2 .

(179)

(180)

once and only once. The surfaces of constant are hyperboloids on which the
Minkowski metric induces a positive definite 3- metric which is clearly invariant
under the action of the Lorentz group O(3, 1). In fact the metrics on the surfaces
= constant are all proportional to a the fixed metric, say that with = 1.
The curved 3-dimensional space = 1
t2 x2 = 1,

(181)

is called hyperbolic space H 3 and is the analogue with negative curvature of the
unit 3-sphere S 3 which has positive curvature.

16.1

*The Foundations of Geometry*

Hyperbolic space first arose during the investigations of the Hungarian mathematician Bolyai and the Russian mathematician Lobachevsky into the foundations of geometry and in particular into Euclids fifth axiom about parallel lines
in Euclidean geometry. This states that
If a straight line falling on two straight lines make the interior angles
on the same side less than two right angles, the two straight lines,
if produced indefinitely meet on that side on which the angle is less
than two right angles.
15 i.e.

along a straight line in Minkowski spacetime

43

After many years of work by many people Bolyai, Lobachevksy and their followers, were finally able to show that Euclids axiom is genuinely independent
of the other axioms of Euclidean geometry and that there exist three consistent
congruence geometries, E 3 and S 3 and H 3 .
In important intermediate step was taken by the Swiss mathematician and
cosmologist Johann Heinrich Lambert (1728-1777).16 Lambert focused attention on a quadrilateral with three right angles, and realized that one could
make three hypotheses about the fourth angle. In self-explanatory terms, these
he called the hypothesis of the right angle, the obtuse angle and the acute angle.
Clearly the Lambert himself rejected the third hypothesis, according to which
the sum A + B + C of the interior angles of a triangle are less than 2, because
he realized that the deficit 2 A B C is proportional to the area S of the
triangle
(A + B + C 2) = KS ,
(182)

where K is now called the Guass-curvature of the space17 But this would mean
that we would have an absolute unit of length. Given that there was no natural
value to assign it, not surprisingly perhaps the French philosopher Auguste
Calinon suggested in 1893? that it might vary with time. Of course, this is
exactly what can happen according to Einsteins theory of General Relativity
as was first realized by Friedmann in 1922.
An interesting contribution to this debate was made by the German physiologist and physicist Hermann Helmholtz in 1870. His view of The Origin and
Meaning of Geometrical Axiomswas that they should be based on the idea of
free mobility of rigid bodies. In other words, thinking operationally, the properties space are what can measured using ideal rigid rods which can be translated
to any point in space, and, moreover, which remain rigid when rotated about a,
and hence every, point.
Consider, for example, a rigid body in ordinary Euclidean space E 3 . This
may be rotated about any point still keeping its shape and it can similarly be
translated to any point. The set of such motions 18 constitute what is called the
Euclidean group E(3) which may be identified with its configuration space. All
those configurations related by rotation about some point clearly correspond to
the same point in Euclidean space. By isotropy, the continuous rotations about
single point constitute the group SO(3) and so we recover ordinary space as a
coset
E 3 = E(3)/SO(3).
(183)
16 Lambert shares with Thomas Wright and Immanuel Kant the credit for first recognizing
that the Milky way is a roughly flat disc made up of stars in Keplerian orbits about some
central body. Our sun having a period about 250,000 years. The central body is now known
to be a black holes in the direction of Sagittarius of whose mass is about 1 million times that
of the sun.
17 Gauss actually checked, by surveying, the angle sum for the triangle of sides 69Km ,85km
and 107km whose vertices are the three peaks Inselberg, Brocken and Hohenhagen in Germany.
18 That is isometries.

44

Helmholtz thus raised the question: what 6-dimensional groups G exist ,

with an SO(3) subgroup such that we can regard space as
G/SO(3).

(184)

If we make various simplifications, we arrive at the three possibilities

E 3 = E(3)/SO(3)

S 3 = SO(4)/SO(3)

H 3 = SO(3, 1)/SO(3)/, .
(185)
From a modern perspective, Helmholtzs assumption of isotropy is well justified by experiments. Actually we can say more, if we are prepared to accept
local Lorentz invariance. We can re-run Helmholtzs reasoning, but replacing
space by spacetime. If E(3, 1) is the Poincare group, then Minkowski spacetime
E 3,1 is the co-set
E 3,1 = E(3, 1)/SO(3, 1).
(186)
There are three possible, so called maximally symmetric spacetimes. The
other two are called de-Sitter dS4 and Anti-de-Sitter spacetime AdS4 . Their
properties may be explored in detail using the methods of General Relativity.
For the present we note that corresponding to Helmholtzs list we have
E 3,1 = E(3, 1)/SO(3, 1)

dS4 = SO(4, 1)/SO(3, 1)

AdS4 = SO(3, 2)/SO(3, 1) .

(187)
Toward the end of the nineteenth century there was an increasing widespread
attitude, that which of these geometries is the correct one is a matter for astronomical observation. Various attempts were made to determine curvature
of space for example in Germany by Karl Friedrich Gauss and later astronomically by Karl Schwarzschild in 1900, at least 15 years before Einsteins Theory of
General relativity. In Schwarzschilds static universe the spatial geometry was
taken, like that of Einsteins Static Universe constructed 17 later, taken to have
positive curvature. However Schwarzschild differed from Einstein in making the
antipodal identification on S 3 turning it into real projective space 19 RP 3 . His
motivation for doing so was to avoid the feature, present in the case of S 3 , that
all light rays in his world, which correspond to all straight lines, passing through
a point on S 3 are refocused at the antipodal point. In projective spaces, distinct
straight lines intersect once and only once and in Schwarzschilds universe any
two light rays sent off by us in different directions would eventually return at
the same time in the future.
The English philosopher, mathematician and logician, and winer in 1950 of
the Nobel Prize for Literature, Bertrand Russell(1872-1970) wrote of this period
My first philosophical book,An Essay on the Foundations of Geometry, which was an elaboration of my Fellowship dissertation, seems
to me now somewhat foolish. I took up Kants question,how is geometry possibleand decided that it was possible, if and only if, space
was one of the three recognized varieties, one of them Euclidean, the
19 The

distinction is precisely the same as between SU (2) and SO(3).

45

other two non-Euclidean but having the property of preserving a

constant measure of curvature. Einsteins revolution swept away
everything resembling this point of view. The geometry of Einsteins
Theory of Relativity is such as I had declared to be impossible. The
theory of tensors, upon which Einstein based himself, would have
been useful to me, but I never heard about it until he used it. Apart
form details, I do not think there is anything valid in this early book.
Russell went on to write many more books, one of them a popular book was
on relativity.

16.2

The Milne metric and Hubbles Law

We can write down the Minkowski line element

ds2 = dt2 dx2

(188)

in co-moving coordinates , , , by setting

t = cosh , x1 = sinh sin cos , x2 = sinh sin cos , x3 = sinh cos .
(189)
On substitution, the Minkowski metric becomes


ds2 = d 2 2 d2 + sinh2 (d2 + sin2 d2 .20
(190)

The standard metric on Lobachevsky space is obtained by setting = 1 to

obtain the expression inside the brace.

Exercise Show that the geodesics of Lobachevsky space may be identified with
the intersections of the hyperboloid = 1 with of timelike two-planes through
the origin.(A timelike two plane is one containing one, and hence many, timelike
vectors).
In Milnes model, the galaxies move with the coordinates , , constant. It
follows that in Milnes universe the proper distance between two galaxies at the
same proper time from the origin increases in proportion to . This is Milnes
explanation for Hubbles law.
We shall now show that, using the redshift formula we derived earlier that
a photon emitted from one galaxy with frequency e at time e and received at
another galaxy with frequency o at time e satisfies
o
o
=1+z = .
e
e

(191)

. From our previous work on the redshift in one dimension

1 + z = exp

(192)

20 You should check that there is a unique value of , , for every event inside the future
light cone of the origin, except at the obvious coordinate singularities at = 0, .

46

where is the relative rapidity between the two frames.

From the embedding equations (189), taking one the emitting galaxy to have
= e , = e and the observing galaxy to e , = e and for both to have the
same angular coordinates, we have
xe xo = e o cosh(o e ) .

(193)

= |o e |.

(194)

Thus relative rapidity is

Now, from the metric form (190), a radial light ray satisfies

and so

d = d,

(195)

o
= exp = 1 + z .
e

(196)

In fact (196) is identical to what one would obtain for this metric using the
standard rules of General Relativity. For nearby sources, z s small and we get
the simple form of Hubbles Law
z = ,

o e = .

(197)

For large redshifts however there are substantial differences from this simple
linear relation. The observational data also exhibit departures from the linear
law at high redshifts. At present the consensus among astronomers is that this
departure is inconsistent with the predictions of the Milne model21 . However
it should be borne in mind that not many years ago that the consensus among
the same astronomers was that the observations did support the Milne model!

16.3

*Relativistic composition of velocities and trigonometry in Lobachevsky space*

It was pointed out by Varicak in 1911 [38] that the composition of velocity law,
which reads in vector notation

u
1 
u
+
v
(u

(u

v))
,
(198)
(v, u)
1 + u.v
c2 1 + u
c2
with u = p

1
2

1 u
c2

, which, considered as a map from B 3 B 3 B 3 , where B 3

is the ball of radius c in three dimensional Euclidean space, is in general neither

commutive nor associative, has an elegant geometrical description in terms of
the trigonometry of hyperbolic space.
Recall that, in conventional spherical trigonometry, one considers triangles
bounded by portions of great circles. The great circles can be thought of as intersections of planes through the centre of a sphere of unit radius. It is standard
21 Known in this context as a k = 1 Friedmann-Lemaitre-Robertson- Walker low density
universe.

47

notation to write a, b, c for the lengths of the sides and A, B, C for the three
angles, angle A being opposite side a etc. All relevant formulae can be derived
from the basic relations
cos a = cos b cos c + sin b sin c cos A ,

etc,

(199)

first apparently derived by the Arab prince and astronomer Mohammad ben
Gebir al Batani (830-929)22 known in Latin as Albategnius.
A concise derivation of Albategniuss formulae is provided by contemplating
the vector identity
(r s).(t u) = (r.t)(s.u) (r.u)(s.t) ,

(200)

which is equivalent to the tensor identity

irs ibu = rt su ru st .

(201)

Now let r = na , s = nb , t = nc , u = na , where na is a unit vector from the

origin to the vertex A etc.
We now return to relativity. Consider three frames of reference Sa ,Sb and
Sc . Associate with each a pseudo-orthonormal basis, also known as a tetrad or
vierbein or rep`ere mobile whose time like legs are a e0 etc. By the results of the
previous section these three timelike legs may be thought of the three vertices of
a triangle in the two-dimensional hypberbolic plane. The sides of the triangles
are geodesics corresponding for example to the two-plane spanned by a e0 and
b e0 . The lengths of the sides, labeled a, b, c of the triangle are related given by
the relative rapidities:
cosh a = b e0 c e0
etc.
(202)
Thus the formulae for composition of boost may be interpreted as giving the
lengths of the sides of a hyperbolic triangle.
In order to obtain the formula, we need only modify the proof given above.
In three dimensional Minkowski space we can still define a cross product:
(u v) = u v .

(203)

= .

(204)

We can also see that

Thus for Minkowski 3-vectors

(r s) (t u) = (r t)(s u) (r u)(s t) .

(205)

If we substitute r = t s = u and both r and s are timelike and r s is spacelike

we see that
|r s|2 = (r r)(s s)(cosh2 1)
(206)
22 All

dates in these notes are CE, i.e. Christian era

48

that is
|r s| = |r||s| sinh .

(207)

cosh a = cosh b cosh c sinh b sinh c cos A

(208)

Our desired formula is thus

Note that formally, one may obtain (208) from the Albategniuss formula
by analytic continuation a ia which may be interpreted as passing to the
case of imaginary radius. In fact, one may ask whether hyperbolic geometry,
like spherical geometry, can be obtained by considering a surface embedded
in ordinary Euclidean space E 3 . The answer turns out be no. However, as we
have seen there is no difficulty in obtaining it form a surface in three-dimensional
Minkowski spacetime E 2,1 .

16.4

Parallax in Lobachevsky space

Consider an equilateral hyperbolic triangle ABC whose sides AC and AB are

equal and whose side AB has length 2r.As the point C recedes to infinity the
angles CAB = CBA tend to a constant value (r) called the angle of paralleism
which depends on the distance r. The formulae of the previous section may be
used to derive the relation
1
sin =
.
(209)
cosh r
Thus for small r we obtain the Euclidean value 2 but in general the angle of
parallism is less than 2 .
Now suppose that A and B are the positions of the earth on its orbit around
the sun at times which differ by six months.
If a star is situated at S somewhere on the line OC, where O is the midpoint
of the side AB, and if at C there is some much more distant star, then the angles
CAS,SAB and CBS and SBA may be measured. In the Euclidean case, one
has

CAS + SAB = .
(210)
2

The angle 2 SAB = is called the parallax and it may be used to estimate the
distance of the star OS in terms of the radius of the earths orbit. This was
first done by Bessel in 1838 for the star 61 Cygni and he found a parallax of
0.45 seconds of arc. Astronomers say that 61 Cygni is situated at a distance of
.45 parsecs.
Lobachevsky, not believing that space was Euclidean, attempted unsuccsefully to measure the curvature of space by measuring the angle of parallism
= CAS + SAB

(211)

for the star Sirius some time before Bessel . Later Schwarschild repeated this
attempt for other stars and obtained the lower bound of 64 light years. In fact,
if = 2 and is small, then
r,
49

(212)

where r is the ratio of the earths radius to the radius of curvature of Lobachevsky
space. The International Celestial Reference System is accurate to no better
than .05 milli-arc-seconds. Thus one could expect to get a lower bound for the
curvature of space in this way using present day technology, no better than
about 105 parsecs.

17

*Rotating reference frames*

We know both from elementary experience and from Newtonian dynamics that
we can tell by means of local experiments if the reference frame we are using is
rotating with respect to an inertial frame of reference, for example that determined by the fixed stars. As Newton himself observed, the water in a rotating
bucket at rest on a rigidly rotating platform rises up the sides due to apparent
centrifugal forces. No trip to Paris is complete without a visit to the Pantheon
to view Foucaults pendulum demonstrating the rotation of the earth, even on
those days when the skies are covered by cloud and astronomical methods are
not available.
Thus we do not expect the Minkowski metric of flat spacetime, when written
in co-rotating coordinates t, z, , to take the same form
ds2 = dt2 dz 2 d2 d2 ,

(213)

as it does in non-rotating cylindrical polars t, z, , . For particles in uniform

rigid rotation about the z-axis, = + t, where is the rate of rotation in
, z labels the particles. Their velocity, relative to an
radians per sec, and ,
inertial frame is v() = .
Substitution and completion of a square gives the so-called Langevin form
of the flat Minkowski metric
ds2 = (1 2 2 ) dt

2 d
2
2
dz 2 d2
d2 .
2
2
1
1 2 2

(214)

Clearly the Langevin form of the metric breaks down at the velocity of light
cylinder = 1 . For > 1 the particles on the platform would have to travel
faster than light. Any physical platform must have smaller proper radius than
1 . Inside the velocity of light cylinder, = 1 , the metric is independent of
time. Thus all distances are independent of time. In other words, the system
really is in a state of rigid rotation.

17.1

Transverse Doppler effect and time dilation

We can read off immediately that for a particle at rest on the platform,
p
d = 1 2 2 dt.
(215)

Thus means that a signal of duration dt consisting of n pulses sent from = e

with frequency fe , so that fe = dne will be received at = e with frequency
50

fo =

n
do .

Thus
f0
=
fe

1 2 e
=
1 2 o

1 ve2
.
1 vo2

(216)

Thus, for example, photon emitted from somewhere on the platform and
absorbed at the centre will be redshifted. In fact this is just the transverse
Doppler effect in a different guise.
The effect was first demonstrated experimentally in 1960 by Hay, Schiffer,
Cranshaw and Egelstaff making use of the M
ossbauer effect for a 14-4 KeV
rays emitted by a Co57 source with an Fe57 absorber [33].

17.2

*The Sagnac Effect*

The Langevin form of the flat Minkowski metric is invariant neither under time
reversal nor reversal of the co-moving angle but it is invariant under simultaneous reversal of both. This gives rise to a difference between the behaviour
of light moving in the direction of rotation compared with that moving in the
opposite direction. The effect is really rather elementary but it has given rise
to considerable discussion.
A light ray, passing along a light pipe for example, satisfies
2 d
+ dl
1 2 2

(217)

2 d2
dr2 + dz 2
+
.
2
2
1
(1 2 2 )2

(218)

dt =
where
dl2 =

If the light ray executes a closed curve C , as judged on the platform in the
pro-grade (in the direction of rotation d > 0) it will take a longer t= than the
time taken t if it traverses the curve in the retro-grade sense (d < 0). The
time difference between these times is
I
Z Z
2 d
dd
t+ t = 2
=
2
(219)
2
2
2 2
C 1
D (1 )
with D = C. If the curve C is well inside the velocity of light cylinder and so
2 2 << 1, we find
t+ t = 4A,
(220)
where A is the area of the domain D enclosed by the curve C.
The difference between the two travel times really has nothing much to do
with Einsteins theory of Special Relativity; it may be ascribed to the simple
fact that light has to travel further in one direction than in the other. The effect
is now usually named after Sagnac. If C is taken to be around the equator, at
sea level t+ t = 414.8ns. This is substantial and must be taken into account
in the calibration of GPS receivers.
51

The Sagnac effect was first demonstrated using interferometry by Harres

in 1911 and Sagnac in 1913 [35]. The method was later used to measure the
absolute rotation rate of the earth by Michelson and Gale in 1925. Using an
optical loop 2/5 miles wide and 2/5 mile long, they verified the shift of 236/1000
of a fringe predicted. Current laser technology allows the measurements of
0.00001 deg h1 . Of the experiment, Michelson said
Well gentlemen, we will undertake this, although my conviction is
strong that we shall prove only that the earth rotates on its axis, a
conclusion which I think we may be said to be sure of already

17.3

Length Contraction

It is a striking mathematical fact is that if we use dl2 in (218) as a spatial line

element in the disc, i.e. set dz = 0 we find that


1  2
dl2 =
(221)
d + sinh2 d2 ,
4 2
where = tanh( 2 ). The curved metric in brackets in (221) is that of the
unit pseudo-sphere, i.e. two-dimensional hyperbolic space with constant Gauss
curvature = 1.
However the physical line element or proper distance on the rotating disc ,
is not (218) but rather ( from (214)
ds2 = d2 +

2 d2
.
1 2

(222)

The metric (222) is also curved. Note that radial directions, orthogonal to
the motion agree with those in an inertial frame, but, as expected, circumferential distances are increased, relative to those in an inertial frame by a factor
1
p
.
1 v 2 ()

(223)

Because of this increase, it is not possible to embed isometrically the surface

with coordinates , and metric (222) as a surface of revolution in ordinary
Euclidean 3-space E 3 .

17.4

Machs Principle and the Rotation of the Universe

The Austrian physicist and philosopher Ernst Mach was much exercised by
ideas of absolute motion and absolute rest. In effect he pointed out that inertial
frames of reference in which Newtons laws hold, and those at rest with the
fixed starsneed not necessarily coincide. In fact they do, to high accuracy, and
this requires some sort of explanation. Nowadays this can be provided using the
theory of Inflation. This requires developing some General Relativity, but it is
possible to use Special Relativity to quantify the extent of the agreement.
52

The English physicist Stephen William Hawking, 17th Lucasian professor

pointed out that if the universe were rotating about us, then distant light received here should suffer a direction dependent transverse Doppler shift . The
photons which have travelled furthest in their journey toward us are part of
the Cosmic Microwave Background (CMB). If the universe were rotating with
angular velocity we would expect a temperature variation with angle of the
form
T0
T =q
,
(224)
2 r 2 cos2
1
c2

where is the angle made by the line of sight with the axis of rotation. Roughly
speaking we may take rc as the age of the universe and so the angle turned
through in that time , in radians, is related to the maximum variation of temperature across the sky by
r
T
.
(225)
= 2
T

Given that the measured temperature fluctuations are about one part in 105 ,
the universe cannot have rotated by more than one hundredth of a turn since
its beginning.

18

General 4-vectors and Lorentz-invariants

In general we set
 
v0
v=
, etc, and
v

v u = v t u = ut v = v0 u0 v.u

(226)

One may check that v w is a Lorentz invariant using (137) but it may also
be seen from the following elementary but illuminating
Proposition If v and w are 4-vectors then v u = u v is a Lorentz invariant.
Proof Evidently
(v + u) (v + u) = v v + v u + u v + u u = v v + u u = 2u v.

(227)

The left hand side is Lorentz-invariant and the first two terms on the left hand
side are Lorentz. Thus u v is Lorentz-invariant. In other words we can a
calculate with the Minkowski inner product in the same way we would for any
quadratic form.

18.1

4-velocity and 4-momentum

The world line of a particle in spacetime is a curve f : x() and is specified

by giving its spacetime coordinates t = t() and x = x() as a function of
some parameter 23 along the curve. One might thing it more natural to
23 In

this section has nothing to do with wavelength.

53

describe the motion by giving x as a function of t, and indeed this is possible

for the particular choice = t but this is distinguishes the time coordinate from
the spatial coordinates but as we have discovered this is against the spirit of
Relativity. Moreover, as we shall see, there are advantages in not making that
choice.

18.2

4-velocity

We can then define the tangent 4-vector of the curve f by

 cdt 
dx
d
= dx
T () =
.
d
d

(228)

If we insist that under a Lorentz transformation the parameter is unchanged

, the T will transform under Lorentz transformations (137) as a 4-vector.
Note that this would not be true if we made the choice = t. Suppose that
T T > 0, then the curve f is said to be timelike and we can make the choice
= where is proper time along the world line. We then define the velocity
4-vector often called the 4-velocity by
 cdt 
dx
d
.
(229)
U=
= dx
d
d
It follows from the definition of proper time that
U U = c2 (

dt 2
dx
d
) ( )2 = c2 ( )2 ,
d
d
d

(230)

i.e.
U U = c2 .

(231)

In units in which c = 1, U is a unit timelike vector. We shall always assume

that t is a strictly monotonically increasing function of , i.e. T0 > 0, which in
the timelike case means that U0 > 0. Such a timelike tangent vector is called
future directed

18.3

4-momentum and Energy

This is defined by
p = m U,

(232)

where m is a positive constant called the rest mass of the particle. If U is future
directed then 4-momentum will also future directed, p0 > 0. We have
p p = m 2 c2 .
Now
p0 =

E
mc
mcdt
:= ,
= q
2
d
c
1 vc2
54

p=m

(233)
dx
mv
.
=q
2
d
1 vc2

(234)

In other words
p=

E
c

(235)

The quantity E is the energy of the particle, as will be justified shortly. Now
pp=

E2
p2 = m2 c2 ,
c2

(236)

thus
E=
and moreover

p
m2 c4 + p2 c2 .
p=

18.4

E
v.
c

(237)

(238)

Non-relativistic limit

For small

v2
c2

we have up to terms of O(v3 ),

p = mv + . . . ,

1
E = mc2 + mv2 + . . . .
2

(239)

We call mc2 the rest mass energy. Note that E is non-zero even if the particle
is at rest. Therefore it is reasonable to define the kinetic energy by
T = E mc2 .

18.5

(240)

Justification for the name energy

If we suppose a general equation of motion of the form

dp
=F
dt

(241)

then the rate of doing work W on the particle is

dp
d
dv
dW
= F.v = v.
= (v.p) p .
dt
dt
dt
dt

(242)

dW = d(v.p) p.dv.

(243)

That is
In the special case that we set

we find

mv
,
p= q
2
1 vc2
mv
dW
mv2

d
q
.v
q
=
2
2
dt
dt
1 vc2
1 vc2
55

(244)

(245)

That is

mv2
d
q
+ mc2
=
dt
v2
1 c2
F.v =

v2
dE
=
.
c2
dt

dE
.
dt

(246)

(247)

It is reasonable therefore to regard E or T = E mc2 as the energy of the particle. In fact we usually include the rest mass energy mc2 in the energy because
in energetic nuclear processes in which particles decay into other particles of
different rest masses, for instance, the rest mass term must be included in the
total energy budget.
That E really is the type of energy you might pay your electricity bill to
acquire gas been demonstrated by timing rapidly electrons to find their velocity
and then absorbing them into a calorimeter to measure their energy in calories.
Sadly for those who dream of perpetual motion, Einsteins formula (237) was
verified [13].

18.6

*Hamiltonian and Lagrangian*

Quite generally in dynamics, for example when considering excitations in condensed matter systems, we define what is called the Hamiltonian function H(p)
of a free particle whose spatial momentum is p by
dH = v.dp, .

(248)

If the system is to be conservative, H must be an exact differential and we have

H(p) = W.

(249)

We also define the Lagrangian function L(v) of a particle with spatial velocity
v as the Legendre transform of the Hamiltonian, i.e.,
L + H = p.v,

(250)

so that from (248, 243)

v=

H
,
p

p=

L
.
v

(251)

In the special case of a relativistic particle one has

p
H = m2 c4 + c2 p2 ,

19

L = mc

v2
.
c2

(252)

Particles with vanishing rest mass

Einsteins theory allows for the possibility of particles which cannot be treated in
Newtons mechanics, those whose speed is strictly constant.The constant value
56

of the speed can only be, by Einsteins Principle of Relativity, exactly that of
light. It turns out that by using only momentum p and energy E as the basic
variables the basic equations still make sense if we set m = 0.
We have in general v = Ec v and so if |v| = c, we have
E = c|p|
Now since
p=
we have

E

(253)

(254)

pp= 0 .

(255)

Particles of this type include the photon which is responsible, according to quantum electro-dynamics, for electromagnetic phenomena and more speculatively
the graviton which, according to quantum gravity, is responsible for gravitational
phenomena. In addition there are three types of neutrinos, e , , associated
with the electron, muon and tau particle respectively.

19.1

Equality of photon and neutrino speeds

At 7:35:40 UT24 on 23 February 1987 electron neutrinos e from the Large Magelllanic Cloud arrived in Japan and were detected using the KAMIOKANDE
neutrino telescope. By 10:38 UT the same day, the first optical brightening
of what is now known as the supernova SN1987A were seen. Thus the travel
time for neutrinos and photons (160,000 years) differed by less than 3 hours.
It follows that their speeds differ by less than two parts in an American billion
(109 ) [17].
For most purposes therefore one may regard neutrinos as being massless,like
the photon. Other experiments however based on neutrinos arriving here on
earth from the sun indicate that they do have a very small mass, of the order
104 eV.
Example If the neutrino actually has a mass m and energy E, and the SN1987
is at a distance L from us, then if T = Lc is the transit time of the photon and
T = Lv that of a neutrino, we have
s
(T T )(T + T )
mc2
=
.
(256)
E
T2

20
20.1

Particle decays collisions and production

Radioactive Decays

Perhaps the simplest process one may consider is the decay of a particle of rest
mass m1 into particles of rests mass m2 and m3 . To get a Lorentz-invariant law
24 Universal

time , i.e. almost exactly Greenwich Mean Time, GMT

57

of decay we express it in terms of 4-vectors. The simplest possibility would be

p1 = p2 + p3 .

(257)

The components of (257) give four equations: the conservation of

Energy

E1
c

E2
c

E3
c ,

and Momentum p1 = p2 + p3 .

(258)

Suppose the decaying particle is at rest in frame S. Then

p1 = (m1 c, 0) p2 + p3 = 0,

(259)

and hence |p2 | = |p3 | = p. The two particles produced move off with equal and
opposite momentum. We also have
q
q
E1 = m1 c2 = m22 c4 + p2 c2 + m23 c4 + p2 c2 m1 m2 + m3 .
(260)

Particle 1 can only decay into particle 2 plus particle 3 if its rest mass exceeds
the sum of the rest masses of the products. If this is true, then a solution for
p always exists. Put another way, the kinetic energy liberated T = T2 + T2 =
(m1 m2 m3 )c2 and this must be positive and this must come from the original
rest mass energy.
In general one expects that unless there is some reason, for example a conservation law like that of electric charge, that heavy particles will always be
able to decay into lighter particles. Only the particle with least rest mass can
be stable. This is the electron. It could, in principle decay into two photons,
but photons carry no electric charge and hence this is impossible.

20.2

Impossibility of Decay of massless particles

Suppose that
p1 = p2 + p3 ,

(261)

p1 p1 = 0.

(262)

p22 + p23 + 2p2 p3 = 0 = m22 c2 + m23 + 2p2 p3 .

(263)

with p1 massless, i.e.

It follows that

But this is impossible because if for p2 and p3 future directed timelike or lightlike
p2 p3 0.

(264)

To prove (264) note that

p2 p3 =

E2 E3
p2 .p3 =
c2

q
q
m21 c2 + p22 m23 c2 + p23 p2 .p3
58

(265)

but the left hand side is bounded below by

|p2 ||p3 | p2 .p3 0,

(266)

since, by the usual Schwarz inequality, v.u |v||u|, for any pair of 3-vectors v
and u.
Note that equality can only be attained if p2 and p3 are two parallel lightlike
vectors. However the decay of a photon for example into two collinear photons
should perhaps be better thought of as superposition. Moreover in quantum
mechanical terms it is a process with vanishing small phase space volume.

20.3

Some useful Inequalities

For a timelike or light 4-vector v we define |v| =

previous subsection can be re-arranged to show

v v. The working in the

Proposition(Reverse Schwarz Inequality) If p2 and p3 are two future directed

timelike or lightlike vectors, then
p2 p3 |p2 ||p3 |.

(267)

Similarly
Proposition(Reverse triangle Inequality) if p1 , p2 and p3 are all timelike or
light like sides of a triangle then
|p1 | |p2 | + |p3 |.

(268)

The reverse triangle inequality can be extended to give

Proposition If p 1, p2 , . . . pn+1 are future directed timelike or null and
p1 = p2 + . . . + pn+1

(269)

|p1 | |p1 | + . . . + |pn+1 |, .

(270)

then

A slight extension of the same working yields

PropositionIf p q 0 for all future directed timelike and lightlike 4-vectors p,
then q is future directed timelike or lightlike.
Finally
Proposition(Convexity of the future light cone) If p and q are future directed
and timelike then so is any positive linear combination ap + bq, a 0, b 0.

59

20.4

Impossibility of emission without recoil

A particle of rest mass m cannot emit one or more particles keeping its rest
mass constant. Thus free electron cannot emit one or more photons. To see
why not, suppose
p1 = p2 + p3 ,
(271)
with |p| = |p2 | and p3 future directed and timelike. We get
p21 = p22 + p23 + 2p p3 , p23 + 2p2 p3 = 0,

(272)

which is impossible. This has an application to so called bremsstrahlung radiation emitted by an accelerated electron. This can only occur if there is some
other body or particle to take up the recoil.

20.5

Decay of a massive particle into one massive and one

massless particle

For example one could consider pion( m = 140 Mev) decaying into a muon
(m = 105 Mev) and an anti-muon-neutrino.
+

(273)

p1 = p2 + p3 , (p1 p3 )2 = 0 = p21 + p23 2p1 p3 .

(274)

We have p22 = 0, thus

This gives

E1 E2
p1 .p3 ).
(275)
c2
Suppose we are in the rest frame of particle 1 so p1 = (m1 c, 0)t . We get
m21 c2 + m23 c2 = 2(

m21 c2 = m23 = 2m1 E3 , E3 =

m21 + m23 2
c .
2m1

(276)

The relativistic gamma factor of the third particle is given by

1
= q
1

20.6

v2
c2

E3
1 m1
m3
= (
+
).
m 3 c2
2 m3
m1

(277)

Decay of a massive particle into two massless particles

For example one might consider the decay of a neutral pion (mpi0 = 135MeV)
into two photons ( symbol ) with lifetime = 8.4 1017 s.
0 2
60

(278)

We have (because p22 = p23 = 0)

p1 = p2 + p3 , p21 = 2p2 p3 .
Thus
m21 c2 = 2(

E2 E3
p2 .p3 ).
c2

(279)

(280)

Now for a massless particle

E2
(281)
c
where n3 is a unit vector in the direction of motion. Thus if n2 .n3 = cos ,
where is the angle (not rapidity!) between the directions of the two massless
particles, we have
p2 = n2

m21 c4
= (1 n2 .n3 ) = (1 cos ) = 2 sin2 ( ).
2E2 E3
2

(282)

m21

.
sin2 ( ) =
2
4E2 E3

(283)

Thus

There are two simple cases

(i) Particle 1 decays from rest. We have E2 = E3 = 21 m21 c2 . This implies = .
(ii) Particle 1 is moving along say the 1-axis with relativistic gamma factor
= p 1 v2 and the two photons are emitted symmetrically making an angle 2
1 c2

to the 1-axis. We have E2 = E3 = 12 m1 c2 , whence

1

sin2 ( ) = 2 .
2

(284)

If particle 1 is moving very fast, >> 1 then will be very small. This is
the headlight effect.

21

Collisions, centre of mass

We have in general
p1 + p2 = p3 + p4 .

(285)

The 4-vector p+ p= mcom U , with U U = c2 defines the centre of mass

energy, that is mcom c2 is the total energy in the centre of mass frame, i.e. in
the frame in which p1 + p2 = 0.
Th relation
(p1 + p2 )2 = (p3 + p4 )2 = m2com c2
(286)
often brings simplifications to the algebra.

61

As an example, if a particle of mass m and 3-momentum p collides with a

particle of mass m which is at rest, then
r
E
mcom = 2 2 M + M 2 + m2 ,
(287)
c
p
where E = m2 c4 + p2 c2 is the energy of the incident particle.
The quantity mcom c2 is usually taken as a measure of the available energy in
1
such collisions. For large E it is much smaller than E, rising as E 2 rather than
linearly with energy. This is because so much energy must go into the recoil. An
extreme case is provided by Ultra-High Energy Cosmic rays. These hit protons
(mass 938MeV) in the upper atmosphere with energies up to 1021 eV = 109 TeV.
They are by far the fastest particles known to us. However energy available for
nuclear reactions is no more than about 103 TeV.

21.1

Compton scattering

In this process, for the discovery of which Compton was awarded the Nobel
prize in 1927, an X-ray photon is scattered of an electron which is initially at
rest. We have p21 = p23 = 0, p22 = p24 = m2e c2 , p2 = (me c, 0)t . Now
(p1 p3 ) = p4 p1 , (p1 p3 )2 = (p4 p2 )2 .

(288)

2p1 p3 = 2m2e c2 2p4 p2 .

(289)

This gives
This gives, using the method for photons used earlier
2

2E4 me c
E1 E3
(1 cos ) = 2m2e c2
c2
c2

(290)

But energy conservation gives

E4 = E1 + me c2 E3 .

(291)

Substitution and simplification gives

(1 cos ) = me c2

 1
1 
.

E3
E1

(292)

According to quantum theory E1 = hf1 = hc3 , where h is Plancks constant, f1

is the frequency and 1 the wavelength of the incident photon and f3 and 3
that of the scattered photon. We get
(1 cos ) =

me c
(3 1 ).
h

(293)

Clearly the wavelength of the scattered photon is longer than that of the incident
photon because kinetic energy has been imparted to the electron.

62

21.2

Production of pions

Protons in cosmic rays striking the upper atmosphere may produce either neutral ( 0 ) or positively charged ( + ) pions according to the reactions
p + p p + p + 0

or

p + p p + n + +

(294)

respectively, where n is the neutron. Since the mass of the proton is 1836.1
times the electron mass and that of the neutron 1838.6, which is why the latter
can decay to the former according to the reaction
n p + e + e

(295)

in about 13 minutes, we ignore the difference and call the common mass M .
Numerically it is about 938MeV. Despite the fact that the mass of the neutral
pion o is 264 times the electron mass and that the charged pions are
both 273.2 times the electron mass, the latter cannot decay into the former by
conservation of electric charge. In either case, we call the mass m. Its value
is about 140Mev. If T is the kinetic energy of the incident proton and p is its
momentum, then equating the invariant (p1 + p2 )2 = (p3 + p4 + p5 )2 and using
the inequality
|p3 + p4 + p5 | (m3 + m4 + m5 )c,
(296)
we get
(T + 2M c2 )2 c2 p2 (2M + m)2 c4 .

(297)

Using what is sometimes called the on-shell condition

(

E
+ M c)2 p2 = M 2 c2 ,
c

one finds that the T 2 terms cancel and one obtains the threshold
m
T 2mc2 (1 +
).
4M

(298)

(299)

This is about 290Mev in the present case.

21.3

Creation of anti-protons

If a proton p collides at sufficiently high speed against a stationary proton a

proton-anti-proton pair can be created as in the following reaction
p + p p + p + (p + p)

(300)

(
p denotes an anti-proton). One might have thought that the least kinetic energy
T required for this process is 2mp c2 , but this is not correct. Most of the incident
kinetic energy goes into the kinetic energy of the recoiling proton. In fact the
threshold , i.e., the minimum energy required, is 6mp c2 = 5.6MeV. To see why,
note that 4-momentum conservation gives
p1 + p2 = p1 + p2 + p3 + p4
63

(301)

thus
|p1 + p2 | = |p1 + p2 + p3 + p4 | |p1 | + |p2 | + |p3 | + |p4 | = 4mp c.

(302)

(we use the fact that mp = mp.) Now p1 = (mp c, 0) and p2 = ( Ec , p), where E
is the total energy, including rest mass energy of the incident proton and p is
3-momentum. Thus
r
E
(303)
|p1 + p2 | = ( + mp c)2 p2 4mp c2 .
c
But
p2 =

E2
m2p c2 .
c2

(304)

Simplifying gives
E 7mp c2 , T 6mp c2 .

(305)

The first production of anti-protons on earth was achieved by Chamberlain and Segre at the Berkley Bevatron in California. This linear accelerator
was built to be capable of accelerating protons up to energies of 6.6 BeV 25
Chamberlain and Segre received the Nobel Prize in 1959 for this work.

21.4

Head on collisions

In this case we have

p1 = (M c, 0)t

p2 = (

E
, p, 0.0)t
c

(306)

with
E 2 c2 p 2 = m 2 c4 .

(307)

By head onwe mean

p3 = (
with

E2
, pM , 0, 0)t
c

E3 2 c2 pm 2 = m2 c2 ,

p4 = (

and

E4
, pM , 0, 0)t
c

E4 2 c2 pM 2 = M 2 c2 .

The two conservation equations are

q
p
E + M = c2 p2m + c4 p2m + c2 p2M + c4 M 2 ,

p = pm + pM .

(308)

(309)

(310)

Our strategy is to eliminate pm and solve for pM in terms of E. This leads

to some moderately heavy algebra so we will go through all the steps, if only
to illustrate the superiority of 4-vector methods. We set c = 1 during the
intermediate stages of the calculation.
25 A

BeV is nowadays called a Gev.

64

Using momentum conservation, the energy conservation equation gives

q
p
E + M p2M + M 2 = (p pM )2 + m2 .
(311)

Squaring gives

q
(E + M )2 2(E + M ) p2M + M 2 + p2M + M 2 = (p pM )2 + m2 .

(312)

Thus

q
E 2 + 2M E + 2M 2 + p2M 2(E + M ) p2M + M 2 = p2 2pm p + m2 .

(313)

But E 2 = p2 + m2 , and hence,

q
2EM + 2ppM + 2M 2 = 2(E + M ) p2M + M 2 .

(314)

Dividing by two and squaring once more gives

M 2 (E + M )2 + 2M (M + E)ppM + p2 p2M = (E + M )2 M 2 + (E + M )2 p2M . (315)

Thus, taking out a factor of pM ,
2M (E + M )p = ((E + M )2 p2 )pM

(316)

Finally we get, restoring units,

pM =

2M p( cE2 + M )
,
2M cE2 + M 2 + m2

with
E=

pm =

p(m2 M 2 )
,
2M cE2 + M 2 + m2

p
m 2 c4 + c2 p 2 .

(317)

(318)

You should check that in the non-relativistic limit, c , one recovers the
usual Newtonian formulae. Just as in that case, if the incident particle is more
massive than the particle it hits (i.e.m > M ) it moves forward after the collision,
while if it is less massive it reverse its direction. If the two particles are perfectly
matched, (i.e. m2 = M 2 ) all the incident energy will be transferred to the target
particle. By contrast if, for example, the target is very massive, (i.e. M ),
the incident particle is reflected back with the same speed it arrives with.
If the incident energy is very large compared with the rest masses of both
itself and the target then
pM p,

pm

(m2 M 2 )c
.
2M

All the incident momentum is transferred to the target.

65

(319)

21.5

Example: Relativistic Billiards

It is well known to players of billiards that if ball is struck and the process is
elastic, i.e. no energy is lost, and the collision is not exactly head-on, then the
two balls move off at an angle = 90 deg between each others direction, as seen
in the frame of reference of the table.
In relativistic billiards this is not so. The angle depends on the ratio of
the energies imparted to the two balls and the incident kinetic energy T . If
the balls have mass m and each emerges form the collision carrying the same
energy, then one finds
4mc2

.
(320)
2 sin2 =
2
T + 4mc2
If T << mc2 we recover the non-relativistic result. By contrast, if T >> mc2
we find 0. This is another manifestation of the headlight effect.

21.6

Mandelstam Variables.

Consider a four-body scattering

a + b c + d,

(321)

with particles of masses ma , mb , mc , md . Conservation of 4-momentum gives

pa + pb = pc + pd ,

(322)

where pa , pb , pc , pd are all taken to be future directed. One has

pa pa = m2a ,

etc.

(323)

The energy momentum in the centre of mass frame is given by

pa + pb = pc + pd

(324)

and
thus the energy available for any reaction, i.e. the centre of mass energy is
s, where
s = (pa + pb )2 = (pc + pd )2 .
(325)
Because pa +pb is non-spacelike s is non-negative, s 0. The energy momentum
transferred from particle a to particle c
t = (pa pc )2 = (pb pd )2 .

(326)

By the reverse Schwartz inequality t (ma mc )2 . In particular if ma = mc ,

the momentum transfer pa pc is spacelike. Now the four vectors pa , pb , pc , pd
are not linearly dependent, they lie in a timelike three-plane. Their endpoints
thus define a tetrahedron. In a tetrahedron, the lengths of opposite sides are
equal and s and t give the lengths squared of two of the three possible pairs of
opposite sides. The remaining length squared is given by
u = (pa pd )2 = (pb pc )2 .
66

(327)

From three vectors linearly independent vectors with 12 components one expects to be able to form only 6 = 12 6 independent Lorentz scalars and hence
the seven quantities ma , mb mc , md , s, t, can not be independent. A simple calculation shows that
s + t + u = m2a + m2b + m2c + m2d .

(328)

If the masses ma , mb , mc , md are fixed s, t, u may be thought of as a set over

three over complete coordinates on a two-dimensional space of scattering states
since they are constrained by the relation (328).
Example all four masses equal.
In the centre of mass frame the ingoing 3-momenta are
pequal and opposite as
are the outgoing momenta. Their common magnitude is p2 and the scattering
angle is , one has
s = 4(m2 + p2 )

t = 2p2 (1 cos ) ,

u = 2p2 (1 + cos ) .

(329)

The allowed range of s is thus s m2 and of t 0 t 4p2 .

It is convenient to regard s, t, u as triangular coordinates in the plane. In
particular we regard them as giving the oriented perpendicular distances from
the sides of an equilateral triangle of height m2a + m2b + m2c + m2d . The sides of
the triangle are thus given by s = 0,t = 0 and u = 0. Not all points in the plane
correspond to physically allowed values of s, t, u. For example, in the case of
four equal masses, the physical region lies outside the equilateral triangle and
occupies an infinitely large 60 deg sector starting from the vertex s = 4m2 and
bounded by two half lines given by t = 0 and u = 0 obtained by producing the
two sides adjacent to that vertex.
It is possible to give physical meaning but this requires the idea of antiparticles.

22

Mirrors and Reflections

22.1

*The Fermi mechanism*

Fermi proposed a mechanism for accelerating cosmic rays. The details have
something in common with a well known thought experiment in which photons
are confined within a cylinder and work done on them by means of a slowly
(i.e. adiabatically) moving piston. What Fermi had in mind was a large cloud
of mass M moving slowly with velocity u << c. Incident on it is a relativistic
particle with momentum p and energy E = pc. The particle is scattered back
with momentum p and energy E = E + E = p . The velocity energy
of the cloud becomes u = u + du but it;s rest mass is unchanged. In this
approximation, energy and momentum conservation become
1
1
M u2 + E = M (u + u)2 + E ,
2
2

M u p = M (u + u) + p .
67

(330)

We have not include the rest mass energy M c2 because it cancels on both sides
of the energy conservation equation. One gets
M uu = E E,

M u = (p p) =

and

E + E
.
c

(331)

Thus
M u

2E
,
c

E = 2E

u
.
c

(332)

The first equation of (332) tells us by how much the cloud slows down, and the
second that the energy of the particles is multiplied by a factor (1 + 2 uc ) which
is greater than unity as long as u is positive. Note that this factor depends only
on u. It does not depend on the mass of the cloud.
Fermi imagined particles bouncing backwards and forward between two
clouds which were slowly approaching each another. The energy of the trapped
particle would go up like a geometrical progression and it would seem that very
high energies could be achieved. In practice, while it is easy to see why particles
might scatter off such clouds, it is not so easy to see how they would get trapped
between two clouds and so Fermis theory has fallen out of favour. However it is
interesting as an illustration of scattering. In fact the cloud behaves like a mirror
and the effect may be understood heuristically as a manifestation of the Doppler
effect. The incoming particle has energy E(1 + uc ) with respect to a rest frame
sitting on the cloud. on the cloud with frequency E(1 + uc ) In the rest frame of
the cloud it is re-emitted with this frequency in the opposite direction and this
is seen in the original rest frame as having energy (1 + uc )(1 + uc )E (1 + 2 uc )E.
In fact reflection problems of this type can also be solved by composing
Lorentz transformations.
Now if L is the distance between the two clouds, then the time between
bounces is 2L
c and in this time the distance has diminished by an amount L =
uL
c . Thus during an adiabatic change
E
L
=
,
L
E

EL = constant.

(333)

If, for example, we think of a photon with frequency f or wavelength find

from (333) f L1 or L.

22.2

*Relativistic Mirrors*

Suppose the mirror occupies the region of spacetime

e1 x > d ,

(334)

where e1 is a unit spacelike vector, corresponding geometrically to the normal

to the timelike hyperplane,
e1 x = d,
(335)
68

and d a constant giving the distance of the plane from the origin.
If a particle, for example a light ray, with 4-momentum p is incident on the
mirror and elastically or specularly reflected off the mirror the reflected particle,
has 4-momentum
R1 (p) = p + 2ei (e p) .
(336)
The possibly unfamiliar sign in (336) is because the normal satisfies
e1 e1 = 1

(337)

Note that the reflection operator R1 (p) leaves the rest mass unchanged since
R1 (p) R1 (p) = p p .

(338)

In other words R1 (p) is an isometry, it leaves Minkowski lengths unchanged.

In components, if
e1 = (, , 0, 0)t ,

p = (E, p1 , p2 , p3 )t ,

(339)

where and have there usual meanings in terms of the velocity v of the mirror,
and we take p1 to be positive so that the incident particle is moving from right
to left. Then reflected particle has momentum
R1 (p) = (E

1 + v2
1 + v2
2v
2v
+
p
,
+p
+
E, p2 , p3 )t .
1
1
1 v2
1 v2
1 v2
1 v2

(340)

As an example, consider the mirror at rest, v = 0.

R1 (p) = (E, p1 , p2 , p3 ) .

(341)

The energy E is unchanged and only the component of momentum perpendicular to the mirror is reversed. Another interesting case is of a light ray or photon
moving perpendicular to the mirror. Thus p = (E, E, 0, 0)t , and
r 1 + v 2
R1 (p) =
(E, E, 0, 0)t
(342)
1v
We see that the light ray is reflected backward with two factors of the Doppler
shift as described in the previous section.
Note that if the mirror is moving and the incident photon is not moving
exactly perpendicularly to the mirror then the angle of reflection will not equal
the angle of incidence.

22.3

*Corner Reflectors on the Moon*

During the first Apollo landing on the moon in 1969 a corner reflector was left
on the lunar surface. Within weeks laser photon pulses sent from the Lick Observatory in California reflected off the reflector and received back in California.
Over the past 30 years or so the number of reflectors and the precision has been
69

increased so that at any given time, the distance to the moon can be determined
to better than 1cm.
A corner reflector effect three successive reflections in there mutually perpendicular mirrors, the walls of an orthant in the rest frame of the reflector. If
the walls have spacelike normals e1 , e2 , e3 , then the effect of three reflections is
given by
R1 R2 R3 (p) = P (p) = e0 (p)e0 + e1 (e1 p) + e2 (e2 p) + e3 (e3 p).

(343)

The operator P (p) is called spatial parity and reverses the spatial components
of any 4-vector it acts on. Thus, according to an observer in its rest frame, a
the corner reflector send back a photo in precisely the direction it comes from.

22.4

Time reversal

22.5

Anti-particles and the CPT Theorem

23

4-acceleration and 4-force

Given a timelike curve x = x( ), where is propertime along the curve,, we

define its acceleration 4-vector by
a=

d2 x
dU
= 2.
d
d

(344)

But

dU
dU
dU
U +U
= 0, 2U
= 0.
d
d
d
Thus 4-acceleration and 4-velocity are orthogonal
U U = c2 ,

aU = 0.

(345)

(346)

Thus, since U is timelike, a must be spacelike. Geometrically, U is the unit

tangent vector of the world line and a is its curvature vector.

23.1

Relativistic form of Newtons second law

This may be written as

m

d2 U
= G,
d 2

or

ma = G ,

(347)

where the 4-force G is not an arbitrary 4-vector but must be orthogonal to U ,

a G = 0.

70

(348)

23.2

Energy and work done

We have
dp
1
p = mU,
= G, q
d
1

d
dt

v2
c2

E

v2
G.
c2

G0
G

(349)

This becomes

dE
=c
dt

v2
1 2 G0 ,
c

dp
=
dt

(350)

Thus
r

v2
G = F,
c2

where F is the old fashioned Newtonian force.

Now
G.v
c
q
= 0.
G U = G0 q
2
2
1 vc2
1 vc2

(351)

(352)

Thus

G0 =

1 F.v
q
c 1 v2

(353)

c2

which gives

dE
= F.v ,
dt

(354)

which is precisely the result we used earlier.

23.3

Example: relativistic rockets

These have variable rest-mass, m = m( ). The equation of motion is

d(mU )
= J,
d

(355)

where J is the rate of emission of 4-momentum of the ejecta. Physically J must

be timelike, J J > 0, which leads to the inequality
m

> |a|.
m

(356)

Thus to obtain a certain acceleration, as in the Twin-Paradox set-up over a

certain proper time requires a lower bound on the total mass of the fuel used
Z
mfinal
ln(
) < |a|d.
(357)
minitial
71

In two dimensional Minkowski spacetime E 1,1

U a = (cosh , sinh ) |aa | =

d
,
d

(358)

where is the rapidity.We find

r
r
1 + vinitial 1 vfinal
1
mfinal
<
=
minitial
1 vinitial 1 + vfinal
1+z

(359)

Consider for example two observers, one of whom is at rest and and engaged
in checking Goldbachs conjecture that every even number is the sum of two
primes using a computer. The second observer, initialy at rest with respect to
the first observer vinitial = 0, decides to use time dilation to find out faster by
accelerating toward the stationary observer thus acquiring a velocity vfinal and
blue shift factor 1+z. The increase in the rate of gain of information is bounded
by the energy or mass of the fuel expended.

24

The Lorentz Force

We will illustrate the general theory in the previous section by means of the
simplest way of solving the constraint G U = 0. We set
G = eF U,

(360)

where e is a constant and F is a 4 4 matrix. (the inclusion of is for later

convenience. Since is invertible, and in fact 2 = 1, we could absorb it into
the definition of F ). Now, using the fact that 2 = 1,
U G = ut F U = U t F U.

(361)

Thus we may satisfy our constraint by demanding that

F = F t .

(362)

We call F the Faraday tensor. The word tensor will not be explained here since
we wont need at this point. We give the components of F suggestive names.

E2
E3
E1
0
c
c
c
E1
0
B3 B2
c

.
F = E2
(363)

B3
0
B1
c

Ec3

B2

B1

E1
c

E2
c

0
B3
B2

B3
0
B1

Thus

1
0
F =
0
0

0
0
0
0
Ec1
1 0
0

0 1 0
Ec2
0
0 1
E3
c

72

E3
c

E1
B2
= c

E2
B1
c
E3
0
c

E1
c

0
B3
B2

E2
c

E3
c

B3 B2

0 B1
B1
0
(364)

G = eF U = e

0
E1
c
E2
c
E3
c

E1
c

E2
c

0
B3
B2

B3
0
B1

E3
c

c
ev. Ec
B2
v1 e(E1 + v2 B3 v3 B2 )

e(E2 + v3 B1 v1 B3 )
B1 v2
v3
e(E3 + v1 B2 v2 B1 )
0
(365)

or in 3-vector notation
dp
dt

= e(E + v B),

dE
dt

= eE.v .

(366)

These are just the Lorentz force equations for a particle carrying an electric
charge e in an electric field E and magnetic field B.

24.1

Example: particle in a uniform magnetic field

In a vanishing electric field, E = 0, the energy E and hence the speed v is

constant. Thus p = mv, where the relativistic gamma factor is constant.
If the magnetic field is uniform, independent of time, and aligned, for example,
along the x3 direction, we have p3 = constant x3 (t) = v3 t + x3 (0), where v3
is the constant component of the velocity in the 3-direction. Thus v12 + v22 :=
2
v
= constant. Now we have
v 1 =

e|B|
v2 ,
m

e|B|
v1
m

(367)

x 2 = v2 = v sin t,

(368)

v 2 =

Thus, with a choice of origin for time

x 1 = v1 = v cost,
with

L
,

(369)

e|B|
is the Larmor Frequency.
m

(370)

=
and
L =

The projection of the motion in the x1 , x2 plane is circular. Up to a translation

x1 = R sin t,
x2 = R cos t,
(371)
with

v = R.
p
2
Thus if p = mv , p = p1 + p23 ,

p = e|B|R .

(372)

(373)

This result is used by cosmic ray physicist, who measure the radius R of
particles the tracks of particles, to obtain their momentum. Numerically, (373)
p = 300|B|R,
73

(374)

with p in eV, |B| in Gauss and R in cm. The radius of the earth is 6, 400 Km
and its magnetic moment 8 1025 Gausscm2 . Thus only particles of 59.5 GeV
or more can be expected to reach the surface of the earth.
Example The relation (373) was used by Bucherer in 1909 [20] to check the relativistic formula relating energy and momentum. Bucherer produced electrons
of known kinetic energy by sending through a known potential difference V and
then sent them through known magnetic fields and measured the radii of their
orbits. He found agreement with the relation
p
eV = m2e c4 + e2 c2 B2 R2 me c2 .
(375)

24.2

Uniform electro-magnetic field and uniform acceleration

If F is a constant matrix we can integrate the equation of motion rather easily.

d
In this denote we shall denote d
by a dot. The equation of motion is 26
a = x =
Now

e 1
F x.

(376)

d
(a a) = 2a (a)
= 2at a .

(377)

But

e
e
F U = F a.
(378)
m
m
e t
e t
Thus at a = m
a 1 F a = m
a F a = 0, Because F is antisymmetric, F = F t .
Thus the magnitude of the acceleration is constant. Of course its direction
changes.
Now the first integral of (377) is
a =

x =

e 1
F x + U0 .
m

(379)

Lets set U0 = 0 and consider the case of a purely electric field along the x1
axis.

|E| 
0
1
c
.
(380)
F = |E|
0
c
The equations becomes, with
cdt
e|E|
=
x1 ,
d
mc

dx1
e|E|
=
t.
d
m

(381)

Thus, with a choice of origin of proper time ,

ct = A sinh

e|E|

,
mc

x1 = A cosh

e|E|

,
mc

(382)

26 The inverse is explicitly included to make contact with the index notation we will introduce
later

74

where A is a constant of integration. The world line is a hyperbola

c2 t2 x21 = A2 .

(383)

The magnitude of the acceleration is

|a| =

25

|AE|
.
mc

(384)

4-vectors, tensors and index notation

In advanced work, particularly when passing to Einsteins theory of General

Relativity, it is helpful to adopt a notation which is a natural extension of elementary Cartesian tensor analysis (see e.g. [21]). The notation is universally
used in physics and engineers throughout the world and despite the initial impression that it is rather complicated, rather more so than the matrix notation
we have been using so far, experience shows that when the basic conventions
have been absorbed, it provides both a very compact notation and one which
allows for very efficient calculations. All legal expressions are automatically
covariant, , i.e have well defined transformation rules under Lorentz transformations, and, from mathematical point of view, it allows one to write down
mathematically well defined formulae and make well defined mathematical constructions without needing expertise in abstract algebra or needing to be familiar with the complicated basis independent definitions introduced in books on
multi-linear algebra. What, for example pure mathematicians call functoriality
is almost guaranteed. In fact the notation was introduced and widely adopted
by pure mathematicians during the first half of the twentieth century and then
abandoned by them, in favour of coordinate free notations. Such notations have
many merits but they often require detailed explanations to unpack them. The
wise words of Arthur Cayley(1821-1895), inventor of matrices and explorer of
higher dimensions, invariants and co-variants speaking in a related context seem
appropriate:
My own view is that quaternions are merely a particular method, or
say a theory, in coordinates. I have the highest admiration for the
notion of a quaternion;but, . . . as I consider the full moon far more
beautiful than any moonlit view, so I regard the notion of a quaternion as far more beautiful than any of its applications. As another
illustration . . . I compare a quaternion formula as a pocket-map - a
capital thing to put in ones pocket, but for use must be unfolded:the
formula, to be understood, must be translated into coordinates.

25.1

Contravariant vectors

One labels the components of a 4-vector in some basis with indices which take
values 0, 1, 2, 4 and which are placed upstairs. In common with most modern
books these indices will be denoted by letters from the Greek alphabet. Instead
75

of 0 one sometimes uses 4. Thus the following notation should be thought of as

conveying the same information.
0
x
 0
x
x1

.
(385)
x
x
2
xi
x
x3

The usual time coordinate is given by x0 = ct but in advanced work we usually

set c = 1 so we will always think of x0 as the time coordinate. Of course, mathematically, the first is the abstract 4-vector, the second the set of its components
in a basis (i.e. in a particular frame of reference) and the third its representation
as a column vector. The real 4-dimensional space of 4- vectors will be called V .
The Lorentzian inner product is written as
x y = xt x = x y .

(386)

Evidently are the components of a the quadratic form in the basis.

Now, as with any vector space, Lorentz transformation may be be viewed
passively: as a change of basis or actively as a linear map : V V . In either
case we have
x x
= x .
(387)
Thus are the components of the linear map or endomorphism . The
first, upper, so-called contravariant index labels rows and the second, so called
covariant index labels the columns of the associated matrix. Note that the Einstein summation convention applies in the modified form that contractions are
allowed only between a covariant index and a contravariant index, i.e. between
an upstairs and a downstairs index.

25.2

Covariant vectors

Now what about row vectors, e.g z = y t , where y is a column vector?. We write
the components of z with indices downstairs and so all of the following should
convey the same information
z

( z0

z1

z2

x3 )

( z0

zi ) .

(388)

Thus
zx = z x = z0 x0 + z1 x1 + z2 x2 + z3 x3 = z0 x0 + zi xi .

(389)

We would like zx to be invariant Lorentz transformations and hence it must

transform like
z z = z1 ,

z = z (1 ) z = z .

(390)

Clearly column vectors and row vectors transform in the opposite way, one
with and the other with (1 )t . We say they transform contragrediently.
76

Alternatively we refer to column vectors as contravariant vectors and row vectors

as covariant vectors. Another way to say this is that x are the components
of an element of the four-dimensional vector space V of 4-vectors, and z , the
components of the four-dimensional dual vector space V , i.e. the space of linear
maps from V to R.

25.3

Example: Wave vectors and Doppler shift

If one looks back at our derivation of the Doppler effect, we wrote

x
= A sin(kx t) = A sin(k
t)

(391)

and deduced the transformation rules for the angular frequency and wave
vector k (125 using the invariance of the phase
x
k
t = k t.

(392)

In our present language we see that we can think of ct, x as a contravariant

vector x and , k as a covariant vector k . Thus
= A sin(k x ).

(393)

Now we see that our Lorentz transformation rule (50) is that for a contravariant
vector and our Doppler shift rule (125) is that for a covariant vector. The
invariance of the phase is the statement that
k x = k x .

(394)

Abbreviating the term covariant vector to covector and contravariant vectors

to contravectors , , we can say that a wave covector k belongs to the vector
space V dual to the vector space V of contravectors.
Geometrically, the surfaces of constant phase are hyperplanes in Minkowski
spacetime E 3,1
= k x = constant.
(395)
The wave covector k corresponds to the co-normal
k =

(396)

to the 3-dimensional hypersurfaces of of constant phase.

25.4

Contravariant and covariant second rank tensors

Now consider how a quadratic form given by

xt Qx = x Q y

(397)

changes under a Lorentz transformation

xx
= x,

i.e.

x = x ,
77

y = y .

(398)

If we define the transformed quadratic form by

y,
xt Qy = x
t Q

(399)

then
= (1 )t Q1 t Q
=Q
Q

or

= Q .
Q

(400)

The Lorentz invariance condition (137) reads

= .
(t ) =

and

(401)
= .

(402)

We say that are the components of a (symmetric) second rank covariant

tensor since they transform in the same fashion as the tensor product or outer
product x y of two covariant vectors x and y .
The components of the inverse of the metric are
( 1 ) = = ,

(403)

= ,

(404)

and satisfy
where is the Kronecker delta, i.e. the unit matrix, whose trace or contraction
is = 4. We say that are the components of a (symmetric) second rank
contravariant tensor since they transform in the same fashion as the tensor
product or outer product of two contravariant vectors x and y .
The Minkowski metric = 27 can be thought of as a symmetric second
tensor , i.e. mathematically speaking a symmetric bilinear map V V R. The
Faraday tensor F = F is an example of an antisymmetric second tensor ,
i.e. mathematically speaking an antisymmetric bilinear map V V R. Under
a Lorentz transformation its components change
F F s.t. F = F = F .

(405)

Note that the transformation rule is exactly the same as for the metric .
The same rule holds for an arbitrary second rank tensor Q , symmetric, antisymmetric or with no special symmetry. The components of an n-th rank
covariant tensor transform , i.e. a tensor with n indices downstairs or mathematically speaking, a multi-linear real valued map from the n-fold Cartesian
product V . . . R transform analogously. The symmetry or anti-symmetry
of a tensor is a Lorentz-invariant. In the case of rank the symmetric and antisymmetric parts
Q = Q() + Q[] ,
(406)
27 Note

that from now on we are will be indulging in the standard abuse of language which
refers to an object by its components.

78

with

1
1
(Q + Q ),
Q[] = (Q Q )
(407)
2
2
transform separately into themselves. The proof proceeds by index shuffling.
For example in the anti-symmetric case
Q() =

[] = Q [ ] = Q [ [ ] = Q[] .
Q

(408)

An identical argument with square brackets replacing round brackets applies in

the symmetric case.

25.5

The musical isomorphism

Note that if v transforms like a contravariant vector and Q is a second rank

covariant tensor the
Q x ,
and
Q x
(409)
are covariant vectors which coincide or coincide up to a sign of q is symmetric
or antisymmetric respectively.
Thus, in the case of Minkowski space, the distinction between contravariant
and contravariant vectors is more apparent than real, because one may pass
from one to the other by index lowering and index raising using the metric
or inverse metric respectively. We use a notation in which the same kernel
letter is used for vectors and tensors which are identified using index raising
or lowering.
Thus we write, for example
x = x x = x .

(410)

In other words the metric effects an isomorphism between the vector space
V of contravariant 4-vectors and its dual vectors space V of covariant 4-vectors.
Thus
p q = q p = p q = p q = p q .
(411)
Note that the order of indices is still important. F = F and Fn u =
F should be distinguished.
One sometimes uses the musical symbols and to denote index raising
and lowering respectively and so the isomorphism is referred to as the musical
isomorphism.

25.6

De Broglies Wave Particle Duality

Tn
In 1924 the French aristocrat Louis-Victor 7e duc de Broglie(1892-1987) proposed, in his doctoral dissertation that just as light, believed since the interference experiments of Thomas Young, to be a wave phenomenon, has, according
to Albert Einsteins photon hypothesis (for which he was awarded the Nobel

79

prize in 1922) some of the properties of particles, so should ordinary particles,

like electrons, and indeed all forms of matter of waves, according to the universal
scheme,
h
,

(412)
where h is Plancks constant. The American physicist Clinton Josephson Davidson(18811958) and the English physicist George Paget Thomson(1892-19750) were awarded
the Nobel prize in 1937 for the experimental demonstration of the diffraction
of electrons. George Thomson was the son of the 1906 Nobelist Joseph John
Thomson (1856-1940) who established the existence of the electron. It was said
of the pair that the father received the prize for proving that electrons are particles and the son for proving that they waves. No parallel case appears to
be known, and indeed may not be possible, in the case of the mathematicians
equivalent of the Nobel prize, the Fields medal.
Note that de Broglies proposal allows us to reconcile the two opposing theories of refraction, the emission and the wave theory described earlier. One may
indeed think of Snells law as expressing conservation of momentum parallel to
the refracting surface as long as one uses de Broglies relation p = h for the
momentum rather than Newtons formula p = mc.
An important part of de Broglies preposterous proposal, for which he was
awarded the Nobel prize in 1929, was that he could show that it is covariant with
respect to Lorentz transformations. With the formalism we have just developed
h
this is simple. Defining = 2
, his proposal becomes
Energy and frequency E = hf

Wavelength and momentum p =

p = k = k ,

(413)

In other words his wave-particle duality is equivalent to the musical isomorphism.

25.7

* Wave and Group Velocity: Legendre Duality*

I order to reconcile de Broglies proposal with our usual ideas it is necessary to

recall some facts about wave motion. In fact, these facts are also relevant for
some of the optical experiments mentioned earlier. In general, monochromatic
wave motion, that is waves of a single fixed wavelength have a single well defined
frequency and conversely 28 Thus the phase travels with the phase velocity
vp = f =

.
k

(414)

In general the motion is dispersive , which means that the phase velocity vp
depends on wavelength . For example, for light, we define the refractive index
by vp = nc and a little familiarity with prisms and the rainbow soon convinces
28 In some situations, such as in condensed matter physics, it may be that the frequency is
a multivalued function of wavelength. In what follows, we exclude this possibility.

80

one that refractive index depends upon wavelength, n = n(). In other words
the dispersion relation = (k) is not, in general = c|k|, but more general.
Now pure monochromatic waves never exist in practice. The best one can
arrange is a superposition of a group or wave packet of waves with almost the
same frequency
Z

(x, t) = A(k )ei k .x(k )t d3 k

(415)

where A(k ) is peaked near k = k.

We set
k = k + s
(k ) = (k) + vg .s + O(|s|2 )

(416)

where the
group velocity

vg =

..
k

(417)

One now performs a stationary phase or saddle point evaluation of the integral.
This amounts to assuming that
A(s) = e
One finds that
ei

k.x(k)t

a2
2

|s|2

(418)
2

e 2a |xvg t| .

(419)

One sees that the peak of the wave packet moves with the group velocity,
not the phase velocity.
Note that de Broglies proposal is compatible with Hamiltonian mechanics.
If we set
H = ,
p = k,
(420)
then (417) and (251 ) become identical.
Now lets turn to the special case of a relativistic particle. Using units in
which = c = 1, the dispersion relation is
p
(421)
= m2 + k2 .
Thus

k
.
vg =
2
m + k2

r
m
vp = 1 + ( )2 ,
|k|

(422)

The group velocity vg coincides with what we have been thinking of the
velocity v of the relativistic particle and is never greater in magnitude than the
speed of light c. By contrast the phase velocity is always greater than that of
light. If vg = |vg , then and we have the strikingly simple relation
vp vg = c2 .

(423)

In Hamiltonian mechanics, the passage between momentum and velocity is

via the Legendre transform. The Legendre transform is a duality or involution because the Legendre transform of a Legendre transform gets you back to
81

yourself. The musical isomorphism is also an involution. These facts are of

course related. We can consider an arbitrary a covariant Lagrangian L(v ) and
covariant or super Hamiltonian H(p ) such that
H + L = p v ,
and
p =

L
,
v

v =

(424)
H
.
p

(425)

To obtain the standard, experimentally well verified Lorentz-invariant relation between energy and momentum, we choose
L=

1
m
v v
p p ,
2
2m

(426)

we have

1
p .
(427)
m
It is illuminating to look at this from the Galilean perspective. 29
Unlike the case with the Lorentz group, Galilean boosts form a three-dimensional
invariant subgroup subgroup of the full Galilei group. Under its action, the four
quantities x, t transform linearly as
 

 
x
1 u
x

.
(428)
t
0 1
t
p = m v v =

which gives a reducible but not fully reducible representation since the subspaces t = constant are left invariant. The phase k.x t is left invariant
and so the wave vector k and frequency transform under the contragredient
representation (i.e. under the transpose of the inverse)
 

 
k
1
0
k

.
(429)

ut 1

These two representations are not equivalent, essentially because no nondegenerate metric is available to raise and lower indices. This is one way of
understanding the difference between the predictions about aberration made
according to the particle and wave viewpoint in Galilean physics.

25.8

The Lorentz equation

Having set up the notation, we are now in a position to write down the equation
of a relativistic particle of mass m and charge e moving in an electro-magnetic
field F = F
dx
d2 x
= eF
,.
(430)
m
d
d
29 In what follows we shall use some standard group-theoretic terminology which will not be
defined here. An understanding of the rest of this section is not necessary for the rest of the
lectures.

82

26

Uniformly Accelerating reference frames: Event

Horizons

Uniform translational motion is, according to Special Relativity, unobservable.

Uniformly accelerated motion however is observable. To reveal some of its effects, we may pass to an accelerated system of coordinates, often called Rindler
coordinates
x0 = sinh t ,
x3 = cosh t ,
(431)
which, if 0 < < , < t < , cover only one quarter of two-dimensional
Minkowski spacetime E 1,1 , the so-called Rindler wedge
x1 > |x0 |.

(432)

In this wedge the flat Minkowski metric takes the static form
ds2 = 2 dt2 d2 .

(433)

From our previous work, we see that The curves = constant have constant
acceleration a = 1 . We shall refer to these curves as Rindler observers. They
are in fact the orbits of a one parameter family of Lorentz boosts, t t + t0
x = x1 x0 et0 x . Note that the propertime Rindler along a Rindler
observer is given by Rindler = t.
The acceleration of the set of Rindler observers goes to infinity on the
boundary of the Rindler wedge, i.e. on the pair of null hypersurfaces surfaces
x0 = x1 . These surfaces are called the future and past horizons of the Rindler
observers. That is because the past, respectively future, light cones of all the
points on the worldline of a Rindler observer, and thus necessarily their interiors,
lie to the past, respectively future of these null hypersurfaces. In other words
the future horizon is the boundary of the set of events that can ever causally
influence a Rindler observer and the past horizon the boundary of the set of
events which a Rindler horizon may causally influence. Thus the nature of all
events for which X 0 > x1 can never be known to Rindler observers. On the
other hand, there is no boundary to the past of an inertial observer, i.e. a timelike geodesic. For example a timelike observer with say x1 = constant > 0 will
simply pass through the future horizon and out of the Rindler wedge in finite
propertime Inertial = x1 . A simple calculation shows that a light ray emitted
from the event (x0 , x1 ) will be received by a Rindler observer at a propertime

 1
x Inertial
1
.
(434)
Rindler = ln

According to the Rindler observer, the light coming from the Inertial observer is increasing redshifted. The motion appears to be slower and slower. So
much so, that the redshift becomes infinite as the Inertial observer is on the
point of passing through the future event horizon and according to the Rindler
observer the Inertial observer never actually passes through in finite time.
83

The rather counter-intuitive phenomena described above have a very precise

parallel in the behaviour of the event horizons of black holes. The fact that
they may occur in such as simple situation as that two-dimensional Minkowski
spacetime shows that although apparently paradoxical, there is nothing logically
inconsistent about them.

27

Causality and The Lorentz Group

27.1

Causal Structure

We may endow Minkowski spacetime with a causal structure , that is a partially

ordering, called a causal relation which is reflexive and transitive. In a general,
time orientable, spacetime, one says that x causally precedes y and writes
xy

(435)

if the event x can be joined to the event y by a future directed timelike or null
curve. Thus
(i)

xy

and y  z x  z

and (ii)

x  x.

(436)

There is an obvious dual relation, written as x  y in which past and future

are interchanged. A stronger relation, called chronology can also be introduced.
We say that x chronologically precedes y if there is a future directed timelike
curve joining x to y and write
x y.
(437)
In Minkowski spacetime the curves may be taken to be straight lines, i.e.
geodesics. We write
xy

27.2

x0 y 0

p
(x1 y 1 )2 + (x2 y 2 )2 + (x3 y 3 )2 .

(438)

The Alexandrov-Zeeman theorem

Our derivation of the Lorentz group earlier depended upon the assumption of
linearity. In fact this may be removed.
Alexandrov[54] and independently Zeeman[53] have shown that any continuous map of Minkowski spacetime into Minkowski spacetime, as long as it is
higher than 1+1 dimensional which preserves the light cone of the origin must in
fact be linear. It follows that such a transformation is the product of a dilation
and a Lorentz transformation. In other words, in four spacetime dimensions,
one may characterize the eleven dimensional group consisting of the Poincare
group semi-direct product dilatations as the automorphism group of the causal
structure of Minkowski spacetime. Since the proof entails special techniques we
will not give it here.
84

In 1 + 1 dimensions, things are very different. In light cone coordinates the

metric is
ds2 = dx+ dx .
(439)
The light cone and causal structure is clearly left unchanged under

x x
= f (x ) ,

(440)
1

where the two functions f are arbitrary monotonic C functions of their argument. Thus the group of causal automorphisms of two-dimensional Minkowski
spacetime is infinite dimensional. It is the product of two copies of the infinite
dimensional group Diff(R) of invertible and differentiable maps of the real line
into itself. This fact plays an important role in what is called String Theory.

27.3

Minkowski Spacetime and Hermitian matrices

We may identify four-dimensional Minkowski spacetime with the space of 2 2

Hermitian matrices X = X according to the scheme
 0

x + x3 x1 + ix2
X=
(441)
.
x1 ix2 x0 x3
Now if X and Y are lightlike separated, then
det (X Y ) = 0

(442)

One can say more,

XY

X Y

is non negative definite

(443)

The Minkowski metric may be written as

ds2 = det dX .

(444)

Now consider the group GL(2, C ) acts on Hermitian matrices by conjugation

,
X SXS = X

=X
.
X

(445)

Moreover if we insist that

det S = 1,

(446)

we obtain the group, SL(2, C ) of 2 2 uni-modular complex valued matrices,

which is 6 dimensional. We have exhibited a homomorphism from SL(2, C) to
SO(3, 1) 30 The kernel of this homomorphism is easily seen to be the group Z2
given by S = 1. Thus the homomorphism is a double covering, S and S give
the same element of the identity component SO0 (3, 1).
We could pursue this homomorphism further, but at this point we prefer to
return to the causal structure (443) on Hermitian matrices. Orderings of this
type were studied by Hua. They have other applications, including to providing
a natural ordering, corresponding to purity on density matrices in quantum
mechanics. A special case of Huas formalism is the case of 2 2 matrices. In
tis case he is able to re-obtain the Alexandrov-Zeeman result [55].
30 Strictly

speaking since SL(2, C ) is connected it is onto the connected component S0 (3, 1).

85

28
28.1

Spinning Particles and Gyroscopes

Fermi-Walker Transport

We begin by considering a timelike curve C with unit tangent vector u =

and a vector e defined along the curve satisfying
de
+ u(e u)
= 0,
d
where u =

du
d

dx
d

(447)

is the acceleration of the curve C. One has

d
(e u) = 0,
d

d
1
(e s) = (e u)(e u).

d
2

(448)

Thus if s satisfies (447) along C, we say that it is Fermi-Walker transported

along C.
From (448) it follows that if s is orthogonal to u, (u s) = 0, at one point on
the curve
C then it is orthogonal to u at all points of C. Moreover its length
|e| = e e will be constant along e. If e1 and e2 are Fermi-Walker transported
along C, then
d
(e1 e2 ) = (e1 u)(e2 u)
(e2 u)(e1 u).

(449)

Thus, if e1 and e2 are initially orthogonal to u and each other they will remain
orthogonal to u and to each other. Introducing a third vector e3 we can arrange
that e0 = u, e1 , e2 , e3 we may constrict in this way a pseudo-orthonormal frame
along the curve C.
Physically we can think of ea , a = 0, 1, 2, 3 as a locally non-rotating frame
defined along the accelerating worldline C.

28.2

Spinning particles and Thomas precession

Let s be the spin vector of a particle whose 4-velocity is u. In a local rest frame,
s should be purely spatial, so
s u = 0.
(450)
In the absence of an external torque, we postulate that its components are
constant in a Fermi-Walker transported frame, along the world line i.e.
ds
+ u(s u)
= 0.
d

(451)

Note that if the world line of the particle is accelerating, even in the absence of
an external torque, the spin, while staying constant in magnitude, will change
in direction. This is called Thomas precession. Its existence was pointed out in
1927[39]. If s u = 0, then we can write (451) as
ds
+ U s = 0,
d
86

(452)

with
U = u (u)
(u)
u .

(453)

One may think of U as an infinitesimal rotation. Thomas regards this as the

result of the commutator of two successive, non parallel boosts.
If an external torque H is applied the equation becomes
ds
+ u(s u)
=H,
d

28.3

where

uH = 0.

(454)

Bargmann-Michel-Telegdi Equations

In (1926) Goudsmit and Uhlenbeck, studying the fine structure of atomic spectral lines and their behaviour in external magnetic fields, the Zeeman effect,
realized that the electron at rest has both an intrinsic spin s of magnitude
|s| = 2 , and an intrinsic magnetic moment , so that immersed in a magnetic
field B the spin changes as
ds
= B.
(455)
dt
In fact Goudsmit and Uhlenbeck argued on the grounds of atomic spectra that
s=g

e
s,
2m

(456)

with the gyromagnetic ratio g = 2. Thus the spin precesses according to the
equation
ds
e
(457)
=g
s B.
dt
2m
The reason for the apparently odd normalization is that for ordinary orbital
motion for which the spin coincides with the orbital angular momentum, s = L,
g takes the value 1. In fact, a little later in 1927? Paul Adrien Maurice Dirac
(1902-1984), the 15th Lucasian professor) proposed that relativistic electrons
satisfy what we now call the Dirac equation, rather than its non-relativistic
approximation the Schr
odinger equation. Dirac showed that the value g = 2
follows naturally from his equation. His work was recognized by the ward of the
Nobel prize in (1933).
Later, in the 1940s advances in radio engineering allowed more precise measurements in atomic spectral lines, and revealed a level of hyper-fine structure
beyond that predicted from the Dirac equation. In particular, there is shift or
splitting in the lowest lines of hydrogen, due to a difference in the energy between
an orbiting electron spinning up or spinning down, relative to the direction of
the orbital angular momentum. The difference in energy, called the Lamb shift
after the man who measured it is extremely small and a transition between the
two levels gives rise to radio waves of 21cm wavelength. It was suggested in 1944
by the Dutch astronomer Henk van der Hulst that radiation of this wavelength
should be emitted by interstellar clouds of neutral Hydrogen and its detection
was achieved by various groups in 1951. Today radio-astronomy using the 21cm
87

line allows is an important area of research, not least because the precise frequency allows the measurement of the velocity of clouds of neutral Hydrogen
using the Doppler effect.
To account for the Lamb shift it is necessary to assign an anomalous gyromagnetic moment to the electron, in other words g 2 6== 0. The value of g 2
can be calculated using the relativistic quantum mechanical theory of photons
interacting with electrons known as quantum electrodynamics QED . At present
the agreement between theory and experiment is better than
To measure g 2 one accelerates electrons in known electromagnetic fields
and measures the precession of the spin. A relativistically covariant set of
equations describing this, properly taking into account the effects of Thomas
precession, was set up by Bargmann-Michel and Telegdi, then at Princeton.
Basically one needs a covariant expression for the torque which will reproduce
(??) in the rest frame of the electron. Ones first guess might be
e
e
F s,
H = g
F s ?
(458)
H =g
2m
2m
but this does not satisfy
H u = 0.
(459)
In order to remedy this defect we introduce a projection operator
h = u u .

(460)

which thought of as an endomorphism h projects an arbitrary vector orthogonal

to u. One may also think of h = h as the restriction of the spacetime
metric to a local 3-plane orthogonal to the tangent of the world line C. The
projection operator satisfies
h2 = h ,

h h = h

h u = 0.

Now using the projection operator we are able to propose

e
e
H=g
hF s = g
F s u(u F s)
2m
2m
or


e
e
F s u u F s .
h F s = g
H = g
2m
2m

(461)

(462)

(463)

Note that s H = (u s)(u F s) and so if u s = 0, the length |s| of the

spin-vector s is constant. In the presence of an electromagnetic field one has
e
(464)
u = F u ,
m
thus
e
e
ds
(465)
=g
F s + (2 g)
u(u F s) .
d
2m
2m
Clearly the case g = 2 is very special. The spin vector s and the 4-velocity
u obey the same equation, and thus they move rigidly together. By contrast, if
g 6= 2, this is not the case, the spin precesses in the moving frame, allowing a
measurement to be made.
88

References
[1] O Roemer, Phil Trans 12 (1677) 893
[2] A A Michelson Studies in Optics(1927) University of Chicago Press
[3] A A Michelson and Morley Phil Mag 24 (1887) 449
[4] Bradley Phil Trans 35(1728)637
[5] A Einstein et al. The Principle of Relativity (Dover)
[6] H Fizeau, Compte Rendu Aacadamie des Sciences de Paris 29 (1849) 90
[7] M L FoucaultCompte Rendu Aacadamie des Sciences de Paris 30 (1850)
551 Compte Rendu Aacadamie des Sciences de Paris 55 (1862) 501 Compte
Rendu Aacadamie des Sciences de Paris 55 (1862) 792
[8] D F Comstock Phys Rev 30 (1920) 267
[9] W De-Sitter,Proc Amsterdam Acad16 (1913) 395
[10] H E Ives and CR Stillwell J opt Soc Amer28(1939 215;31(1941)369
[11] A H Joy and R F Sanford Ap J 64 (1926) 250
[12] D Sadeh Phys Rev Lett Experimental Evidence for the constancy of the
velocity of of gamma rays using annihilation in flight10 (1963) 271-273
[13] W Bertozzi Amer J Phys 55(1964)1
[14] Alvager, Farley, Kjellamn and Wallin Phys Lett 12(1964) 260
[15] L Marder , Time and the space traveller (1971) George Allen and Unwin
[16] J C Hafele and R E Keating Science 177 (1972) 166-168,168-170
[17] M J Longo, Phys Rev36D (1987) 3276-3277
[18] J Terrell Phys Rev116(1959) 1041
[19] R Penrose, Proc Camb Phil Soc 55 (1959) 137
[20] A H Bucherer Ann Physik 28(1909) 513
[21] H Jeffreys Cartesian tensors
[22] M J Rees, Nature 211 (1966) 468- 470
[23] L Dalrymple Henderson The Fourth Dimension in Non-Euclidean geometry
in Modern Art Princeton University Press
[24] K Brecher Phys Rev Lett 39(1977) 1051

89

[25] W Pauli Theory of Relativity Pergamon

[26] C Doppler Uber

das farbige Licht der Doppelsterne (1842)
[27] Y Z Zhang Special Relativity and its Experimental Foundations World Scientific (1997)
[28] M Lachi`eze-Rey and E Gunzig The Cosmological Background Radiation
CUP
[29] W Zurhellen Astr. Nachr. 198(1914) 1
[30] G Str
omberg,Pubs Astr. Socs of the Pacific 43 (1931) 266
[31] G Van Biesebroeck, Ap J 75 (1932) 64
[32] O Heckmann Ann dAstrophysique 23 (1960) 410
[33] H J Hay, J P Schiffer, T E Cranshaw and P A Egelstaff Phys Rev Lett 4
(1960) 165
[34] J B Hartle Gravity
[35] G Sagnac, Compte Rendue de Academie des Sciences de Paris 157(1913)
708
[36] B Rossi and D B Hall Phys Rev 59(1941) 223
[37]
[38]
[39] L H Thomas Phil Mag3(1927)1-22
[40] R Bonola Non-Euclidean Geometry Dover re-print of 1912 translation of
1911 original. Contains translations of the original papers by Bolyai and
Lobachevksy
[41] B A Rosenfeld A History of Non-Euclidean geometry Springer
[42] R Toretti, Philosophy of Geometry from Riemann to Poincare D Reidel
[43] A Calinon Revue Philosophique 32(1891) 368-375
[44] J C F Zollner, Transcendental Physics
[45] R Goldblatt Orthogonality and Spacetime Geometry (Springer)
[46] A Brillet and J L Hall Phys Rev 42 (1979) 549-552
[47] H Helmholtz Mind 1 (1876) 301-321
[48] G B Airy Proc Roy Soc20(1871-1872) 35-39

90

[49] The Electrician May 5th (1899) 40-41

[50] E Mach, The Science of Mechanics
[51] M Jammer, Concepts of Space, Dover
[52] L Lange Ber der k
ongl. s
achs. Ges. der Wis. (1885)335-351
[53] J Zeeman, J Math Phys4 (1964) 490-493
[54] A D Alexandrov Can J Math 1119-1128
[55] L-L Hua, Proc Roy Scoc bf A 380 (1982)487-488

91

Index
4-force, 70
4-velocity, 54

Cosmic Microwave Background, 53

cosmic microwave background, 35
cosmic microwave background, 41
covariant, 7577
covector, 77
curvature vector, 70

aberration map, 37
absolute unit of length, 44
acceleration 4-vector, 70
adiabatically, 67
amplitude, 31
angle of paralellism, 49
angular frequency, 32
anomalous gyromagnetic moment, 88
antipodal identification, 45
available energy, 62

declination, 38
dilations, 20
Dirac equation, 87
dispersion, 80
dispersion relation, 81
displacement current, 10
duality, 81

Ballistic theory of light, 22

barycentre, 38
blue-shifted, 32
boost, 21
boosting, 21
bremsstrahlung radiation, 60
Bureau International des Poids et
Mesures , 27

Einsteins Static Universe, 45

Einsteins Equivalence Principle, 36
elastic, 66
elastic or specular reflection, 69
electron and muon anti-neutrinos, 28
electron volt, 5
Emission or Ballistic theory of light,
7
Encyclopedie, 16
events, 17

causal relation, 84
causal structure, 84
celestial equator, 38
celestial sphere, 40
centre of mass, 38
centre of mass energy, 61, 66
centre of mass frame, 61
centroid, 38
chronology, 84
clock hypothesis, 43
CMB, 41
CMB dipole, 42
co-moving coordinates, 43
co-rotating coordinates, 50
compass of inertia, 13
configuration space, 44
constitutive relations, 11
contragrediently, 76
contravariant, 76, 77
contravectors, 77

Faraday tensor, 72
Fermi-Walker transport, 86
fixed stars, 38
fixed stars, 13, 38
foundations of geometry, 43
Fourier Analysis, 9
fractional linear transformations, 37
frame of reference, 14
Fresnels dragging coefficient, 24
future directed, 54
future horizon, 83
Gauss-curvature, 44
General Relativity, 6
Gravitational Redshift, 30
graviton, 57
Greenwich Mean Time, 57

92

Hafele-Keating experiment, 30
Hamiltonian function, 56
homotheties, 20
horizon scale, 32
Hot Big Bang, 41
Hubble constant, 32
Hubble radius, 32
Hubble time, 33
Hubbles law, 32
Hyperbolic space, 43
hyperfine structure, 87
hyperplanes, 77
hypersurfaces, 77

magnetic moment, 74, 87

maximally symmetric spacetimes, 45
Milne universe, 42
minutes of arc, 38
Moebius transformations, 37
momentum transfer, 66
monochromatic, 80
muons, 28

index lowering, 79
index raising , 79
index shuffling, 79
inertial reference system, 13
inertial coordinate system, 13
inertial frame of reference, 13, 14
infinitesimal line elements, 40
infinitesimal area element, 39
Inflation, 52
International . Celestial Reference
Frame, 38
International System of Units (SI units),
27
intrinsic spin, 87
invariant interval, 26
involution, 81
isometry, 69
kernel letter, 79
Kinematic Relativity, 42
kinetic energy, 55
Kronecker delta, 78
Lagrangian function, 56
Lamb shift, 87
Langevin, 50
Larmor frequency, 73
Legendre transform, 56
light rays, 6
light years, 5
Lorentz group, 36
Lorentz Transformations, 20

n-th rank covariant tensor, 78

neutrinos, 57
non-dispersive, 9
non-relativistic limit, 20
normal, 68
nutates, 39
oblique coordinates in spacetime, 18
on-shell condition, 63
optical distance, 7
orthant, 70
orthogonal poariztaion states, 11
outer product, 78
past horizon, 83
permeability, 11
permittivity, 11
phase velocity, 80
photon, 57
pions, 28
Planckian spectrum, 41
plane of the ecliptic, 38
plasma, 39
Poincare group, 26
Principle of Relativity, 14
principle of superposition, 9
pro-grade, 51
projection operator, 88
proper distance, 29
proper motions, 13
proper time, 27
pseudo-orthonormal basis, 48
pseudo-sphere, 52
QED, 88
quantum electrodynamics, 88
quasars, 13, 25

93

quasi-stellar radio sources, 13, 25

uni-modular, 36
unit tangent vector, 70
Universal Time, 57

railway time tables, 17

rapidity, 21
red shift, 32
redshifted, 32
reflection operator, 69
refractive index, 23
relative velocities, 14
relativistic gamma factor, 21, 39
rep`ere mobile, 48
rest mass, 54
rest mass energy, 55
retro-grade, 51
right ascension, 38
Rindler coordinates, 83
Rindler observers, 83
Rindler wedge, 83
rotating bucket, 50

velocity 4-vector, 54
very long base line interferometry,
26
vierbein, 48
VLBI, 26
wave covector, 77
wave number, 32
wave packet, 81
Wave theory of light, 8

scalar wave equation, 9

Schrodinger equation, 87
Schwarzschilds static universe, 45
seconds of arc, 38
simultaneity is absolute, 18
space, 6
space reversal, 21
spacetime, 17
spacetime diagram, 17
spatial parity, 70
special Lorentz group, 36
stellar parallax, 39
stereographic coordinates, 37
symmetric second rank tensor, 78
tangent 4-vector, 54
tensor product, 78
tetrad, 48
thermal, 41
Thomas precession, 86
threshold, 63
time, 6
time reversal, 21
timelike, 54
transverse Doppler shift, 53
triangular coordinates, 67

94