Special theory of relativity

Electrodynamics and Michelson-Morley experiment

At the end of the 19th century physics was solidly shaped by two great theories: Newton's mechanics
and Maxwell's electrodynamics. Maxwell's equation provided a complete description of all phenomena
known at the time associated with electricity and magnetism.

Maxwell's equations in vacuum (Gaussian units) are:

E = 0

B = 0
1 B
E=
c t
1 E
B=
c t
They determine the time evolution of electric E(r, t) and magnetic B(r, t) elds. By taking curls of
the last two equations one easily nds:

1 2
2 E E=0
c2 t2
1 2
2 B B=0
c2 t2
where 2 is Laplacian:
2f 2f 2f
2 f = + +
x2 y 2 z 2
These equations describe the propagation of waves. The plane-wave solution is:

E(r, t) = E0 cos(kr t)

B(r, t) = B0 cos(kr t)
where the vectors E0 and B0 are perpendicular to each other and to the wavevector k which determines
the direction of propagation and the wavelength = 2/|k|. The wave frequency is = /2 . By
substituting the plane-wave solutions in the wave equations one nds 2 c2 k 2 = 0, which implies
= ck or = c. In order to determine the velocity of wave propagation (phase velocity), we ask how
fast must r change in time to keep the phase xed, kr t = const. We nd
dr
= =c
dt k
Hence, all electromagnetic waves propagate with the velocity c 3 108 m/s, the speed of light.

Maxwell's equations have been extremely successful in describing electromagnetic phenomena, and were
spectacularly successful in predicting the existence of electromagnetic waves subsequently discovered
by Hertz. Therefore, their validity can be hardly doubted.

However, Maxwell's equations are written without regard to the frame of reference. Consequently, they
predict that the speed of light is the same in any reference frame, regardless of how fast one moves
with respect to the light source, or the source of any other electromagnetic radiation. Material objects
behave dierently. If a particle was kicked with relative velocity v from a source, and the source is
moving with velocity u relative to an observer, then the observer nds that the particle moves with
velocity v + u. In other words, velocities of objects are relative and depend on the reference frame in
which motion is measured.

1
 In order to explain the constancy of the speed of light without contradicting Newtonian mechanics, a
proposal was put forward that an invisible substance called ether lls space and acts as a medium for
electromagnetic waves, in a similar fashion to air being a medium for sound waves. Then, the speed
of light predicted by Maxwell's equation would be the velocity of radiation measured with respect to
ether, that is the velocity measured in the reference frame in which ether is at rest (doesn't ow). If
we were to think of ether as a uid lling the universe, it would have to be at rest everywhere in space
in some reference frame, otherwise we would see various optical distortions of distant stars, possibly
time dependent (just like looking through non-uniformly owing water). Therefore, there should be a
preferred reference frame in the universe, as far as electromagnetic radiation is concerned.

Michelson-Morley experiment

Michelson and Morley attempted to detect motion through ether based on the described physical
picture. They constructed the rst interferometer. A monochromatic collimated light beam was split
by semi-transparent mirrors into two beams which propagated at right angles. These two beams
were reected back by regular mirrors positioned at the same distance from the splitter. The reected
beams would fall back on the semi-transparent splitter and interfere. The observed interference fringes,
specically their locations, allowed measuring distances (or velocities) traveled by the two split light
beams with accuracy set by the light wavelength (which can be nanometers).

Since the two split beams propagated at right angles, they could have dierent velocities according to
the ether hypothesis. For example, one could rotate the equipment to align the path of one beam with
the velocity vector of Earth relative to ether. The other beam would be perpendicular to the relative
motion of Earth through ether, so it would propagate with dierent velocity than the rst beam. One
could then measure interference fringes while gradually rotating the equipment and look for changes.

Michelson and Morley found that interference fringes did not meet the expectations from the ether
hypothesis. Their experiment was more sensitive than required to see a clear signal if ether existed.
Many other subsequent experiments, up to modern times, conrmed this nding.

By now, it is overwhelmingly clear that there is no ether.

Einsteins theory

After the Michelson-Morley experiment, Einstein attempted to nd a dierent resolution to the appar-
ent conict between Maxwell's electromagnetism and Newtonian mechanics. He made a bold proposi-
tion that the Newtonian rule of adding relative velocities, known as Galilean transformations, applied
only when velocities were much smaller than the speed of light. This did not contradict any observation
made to date on the kinematics of material objects, because all tests of Newtonian mechanics had been
indeed conducted only at small velocities in comparison to c. However, it left open the question of
how the laws of kinematics should be modied at extremely large velocities in order to accommodate
the properties of electromagnetic radiation. Einstein constructed the minimal possible generalization
of Newtonian kinematics which accomplished this goal, and it turned out to be a spectacular success
despite some of its strange predictions.

Einstein put forward two postulates:

1. All reference frames are physically equivalent. There is no preferred frame of reference in the
universe which could be distinguished by some experiment.

2. The speed of light c is the same in all reference frames, irrespective of the relative velocity of the
source.

The rst postulate is very natural and holds true even in Newtonian mechanics. It is an implicit
statement that ether does not exist. The second postulate is a rearmation of Maxwell equations'
consistency with the rst postulate. The speed of light does not add up to the velocity of source.
Therefore, Galilean transformations cannot be absolutely correct.
Galilean transformations

A reference frame is dened by a coordinate system used to specify positions of objects, and a clock
used to measure time. Coordinates of an object express the distance in x, y and z directions of the
object from the coordinate system origin. The clock measures elapsed time since some initial moment
labeled t = 0. The position of the origin, orientation of the coordinate system axes and the moment
t=0 are arbitrary.

Any event can be specied by a point (x, y, z, t) in the space-time coordinate system of any reference
frame. The same event will have dierent coordinates in dierent reference frames. Since there is no
physically preferred reference frame, we need transformation rules for expressing the coordinates of the
same event between dierent reference frames.

Galilean transformations are such rules dened in Newtonian mechanics.

Consider two reference frames, A and B. In order to avoid mathematical complexity let us assume that
x, y, z axes of A and B are always parallel. Also, let us assume that at t = 0 measured in both reference
frames the two frames coincide. Otherwise, the origin of the frame A moves with constant velocity u
along x axis as observed in the frame B.

Now, consider an event (xA , yA , zA , tA ) observed in the frame A. When and where does this event
occur in the frame B? First, time is absolute in Newtonian mechanics and ows equally fast for all
possible observers. Therefore, tB = tA . Second, the two frames move relatively to each other only in
x direction. The y and z axes coincide all the time, so yB = yA and zB = zA . Only the x coordinates
need translation. The same event has the same x coordinates in the two reference frames only at t=0
when the two frames coincide. An object at rest in the frame A moves at velocity u in the frame B, so
its xB coordinate increases at the rate u measured in B: xB = xA + utB . (note tB ). Now, we use the
nding tA = tB and summarize Galilean transformations:

tB = tA
xB = xA + utA
yB = yA
zB = zA

Galilean transformations imply the rule for adding velocities. Velocity of an object v = (vx , vy , vz ) is
a vector with components vx = dx/dt, vy = dy/dt and vz = dz/dt. If an object has velocity vA in the
reference frame A, then its velocity in the frame B is:

dxB d
vBx = = (xA + utA ) = vAx + u
dtB dtA
dyB dyA
vBy = = = vAy
dtB dtA
dzB dzA
vBz = = = vAz
dtB dtA

Despite all of its intuitive appeal and accuracy in describing the kinematics of ordinary objects around
us, we see that these velocity addition rules do not work for light because c must be always the same
constant given by Maxwell's equations.

Lorentz transformations

How can we generalize Galilean transformations to obtain something consistent with Maxwell's equa-
tions and Einstein's second postulate? In order to answer this question we will set up the reference
frames A and B exactly the same way as in the above discussion. In the spirit of the Einstein's rst
postulate, the transformation rules must be the same for any relative velocity u of the two frames.
However, the modied rules must reduce to Galilean transformations in the limit of small velocities,
that is uc (Galilean transformations are obviously correct in this limit).

3
 Next, we note that Galilean transformations are linear. We expect the modied transformations to also
be linear. Otherwise uniform motion of an object in one reference frame could look like accelerated
motion in the other frame. This is explicitly forbidden by the rst Einstein's postulate. The laws
of physics must be the same in all reference frames: if a force acts on an object, then its eects
(acceleration) must be observed in all reference frames

Note that we made an implicit assumption that all reference frames we consider are inertial. A reference
frame is inertial if we can apply the rst Newton's law in it (if the net sum of all forces on an object
exerted by other bodies is zero, then the object will sit at rest or move along a straight line with a
constant velocity). Any reference frame moving uniformly with respect to an inertial frame is also
inertial. However, a frame which accelerates with respect to an inertial frame is not inertial, and one
can nd that objects accelerate in it without being aected by other objects.

Let us then begin by writing generic linear transformation laws. Since there is no relative motion of
the frames A and B along y and z axis, there is no reason to convert these coordinates: yB = yA ,
zB = zA . The linear transformation AB of x and t coordinates is:

tB = a11 xA + a12 tA
xB = a21 xA + a22 tA

The new thing is that we allow the time coordinate to be converted in a non-trivial way. We will soon
see why this is necessary.

Consider an object sitting at the origin in the frame A: xA = 0. In the frame B this must be perceived
as motion at velocity u:
dxB dtA a22 dtA
= a22 = =u
dtB dtB a12 dtA
where we rst used xB = a22 tA and then tB = a12 tA for xA = 0. Similarly, an object sitting at the
origin in the frame B must be perceived as moving with the velocity u in the frame A:
dxA a22
= = u
dtA a21
where we have used just the second transformation rule with xB = 0. By comparing these two equations
we immediately nd a22 = ua12 and a12 = a21 .
Now consider an object moving with velocity c in the frame A. According to the second postulate, this
must be perceived as motion at velocity c in the frame B as well. Therefore:

dxB a21 dxA + a22 dtA a21 (dxA /dtA ) + a22 a21 c + a22
= = = =c
dtB a11 dxA + a12 dtA a11 (dxA /dtA ) + a12 a11 c + a12

This was calculated by considering how small increments of coordinates and times are related in the
two reference frames. It follows that

a21 c + a22 = a12 c + a11 c2 = a22 = a11 c2

since we found earlier that a12 = a21 . One important observation is that we used to have a22 = u
in the case of Galilean transformations, but now this requires a11 to be non-zero so that time itself
becomes dependent on the reference frame. Such non-Newtonian relativity of time is necessary in order
to formulate coordinate transformations which do not contradict Maxwell's electrodynamics.

Let us summarize the ndings so far and rewrite the transformation rules in a more compact form.
First we introduce a single symbol to capture the equality between a12 and a21 : = a12 = a21 .
Similarly, the relationship between a22 = a11 c2 can be expressed by introducing a symbol and writing
a11 = /c and a22 = c. This particular form is convenient because the remaining relationship
a22 = ua12 we found implies c = u , that is = u/c.

4
 It will be convenient to express time variables as products ct because these products have the same
physical dimension as the x coordinates. The transformation rules for AB can now be written as:

ctB = (xA + ctA )

The only remaining unknown is and in order to nd it we need to invert the above equations and nd
the transformation rules B A. According to the rst Einstein's postulate, there are no preferred
reference frames so that the B A transformations must have the same mathematical form as the
A B transformations. The only dierence is the sign of u: if the frame A moves with velocity u
relative to B, then B moves with velocity u relative to A. Let us obtain the inverted transformation
1
by eliminating xA from the second equation, xA = xB ctA and substituting this into the rst
equation:
xB + ctB
ctB = 1 xB 2 ctA + ctA = ctA =
(1 2 )
We see that the same mathematical form as in the rst AB equation is obtained if (1 2 ) = 1/ 2 .
The sign of the term in the numerator is changed, but this is just as expected because = u/c and
u must appear with the opposite sign in the BA transformation.

The full set of transformations is:

ctB = (xA + ctA ) ctA = (xB + ctB )

yA = yB zA = zB
where
u 1
= =p
c 1 2
These equations are called Lorentz transformations. Clearly, they reduce to Galilean ones whenever
we can neglect as a small number.

Relativistic phenomena

Lorentz transformations predict a few unusual phenomena. The strangeness comes from the funda-
mental relativity of time which we are not used to in our everyday experience. For example, two events
which occur simultaneously in one reference frame may appear non-simultaneous in other reference
frames. Furthermore, time intervals and spatial distances between events can be perceived dierently
in dierent reference frames.

Consider two events which occur at the same location in the frame A, say xA1 = xA2 = 0, but at
dierent times tA1 = 0 and tA2 . The perceived time interval between these events in the frame A is
tA = tA2 tA1 = tA2 . According to Lorentz transformations, the events are seen to occur in the
frame B at:
xB1 = 0 , tB1 = 0
xB2 = utA2 , tB2 = tA2
Therefore, the interval between the two events is perceived in the frame B as:

tB = tB2 tB1 = tA

This is known as time dilatation, because >1 for any nite u. If we put a clock in front of a camera
in a fast space ship, and broadcast the captured video to an observer on Earth, the observer will see
a slower passage of time on the clock in the space ship than on his/her own clock. Interestingly, if the
pilot of the space ship received a similar video transmission of the clock on Earth, he/she would also
have an illusion that time on Earth goes more slowly than inside his vessel.

5
 Now, consider two events which occur simultaneously in the frame A, say at tA1 = tA2 = 0, but at
dierent locations xA1 = 0 and xA2 = xA . This setup can represent a measurement of the length of
some object in the frame A. Using Lorentz transformations we get:

xB1 = 0 , tB1 = 0
u
xB2 = xA2 , tB2 = xA2
c2
The distance between events (the object length) is perceived in the frame B as:

xB = xB2 xB1 = xA

This is called the contraction of lengths based on the premise that an observer ideally measures the
length of a moving object in one instant of time. We have placed such an observer in the frame A
in this analysis, so xA = 1 xB < xB . If we took an instant photograph of the fast space ship
from the previous example, it would look shortened in the direction of its motion. The pilot could also
take a photograph of Earth, and while he wouldn't notice anything unusual about his own appearance,
Earth would look to him like a Rugby ball.

In our everyday experience we do not see such strange things, but then 1 because u  c. If the
velocity u becomes comparable with the speed of light, time dilatation and length contraction become
noticeable. We have indirect evidence of these phenomena. For example, certain unstable elementary
particles created in accelerator experiments have a very short lifetime before decaying into more stable
particles. Yet, we nd such particles in cosmic radiation, created far away and traveling toward Earth
for much longer than their proper lifetime. Relativity reconciles this because the short lifetime in a
particle's own frame of reference appears signicantly longer to us when the particle moves close to
the speed of light.

The theory of relativity implies that nothing can move faster than the speed of light. In order for
Lorentz transformations to be physical, u must be strictly smaller than c, otherwise is not a nite
real number. The correct interpretation of this is that no physical inuence of any kind can propagate
faster than light. One event cannot be a cause of another if the time interval between them is shorter
than it would take light to travel the distance between them. It is causality which is limited by the
speed of light. One can reect light from a powerful collimated lamp on clouds in the sky and make
the reected spot move faster than light by rotating the lamp very fast. This does not violate the
relativistic causality.

Velocity transformation

Here we derive the relativistic rules for adding velocities. Let an object move with velocity vA =
(vAx , vAy , vAz ) in the reference frame A. Its velocity in the frame B appears to be:

dxA
dxB dxA + udtA dtA + u vAx + u
vBx = = u = u dxA
=
dtB c2 dxA + dtA c2 dtA + 1
1 + uvAx /c2

dyA
dyB dyA vAy
vBy = = u = 1 u dxdtAA = 1 2
dtB c2 dxA + dtA c2 dt + 1 1 + uvAx /c
A

dzA
dzB dzA vAz
vBz = = u = 1 u dxdtAA = 1
dtB c2 dxA+ dtA c2 dtA + 1 1 + uvAx /c2