129A Lecture Notes

Notes on Special Relativity

1 Why Relativity?
Particle Physics aims to study structure of space, time and matter at its
most fundamental level. It necessarily means that we study physics at the
shortest distance scales as possible. To probe short distances, Heisenberg’s
uncertainty principle ∆x∆p ≥ h̄/2 states that we need to provide large mo-
mentum. In other words, we are looking at scattering phenomena of particles
with large momenta and hence high energy. In practice, it often means the
use of particle accelerators that can bombard projectile particles (beams)
on target (or opposing beams) at very high momenta or energies. This is
why terms “particle physics” and “high-energy physics” had been often used
interchangingly.1 As we increase momentum of the particle, eventually it
exceeds the mass of the particle times the speed of light mc, and the particle
becomes relativistic. Therefore, special relativity is an integral part of par-
ticle physics and we need to understand it thoroughly. Even though special
relativity has been discussed already in 110AB, I’d like to review it in the
way that I can use it later on.
Relativity itself of course is one of the major revolutions in physics in 20th
century, together with quantum mechanics. The way Einstein discovered it
was based on theoretical consideration that Maxwell’s equation had unusual
invariance, Lorentz invariance. He boldly stated that it must be the invari-
ance of our space and time. It completely changed the way we view our space
and time, so intertwined that it is now called “spacetime,” leading to exotic
phenomena such as time dilation and Lorentz contraction. Its marriage with
quantum mechanics further led to a dramatic prediction of the existence of
anti-matter.
1
Of course, particle physics experiments that do not use accelerators had also been
quite successful as we will discuss in this course, such as neutrino physics and cosmology,
and sometimes people distinguish these two terms.

1
2 Gallilean space and time
In the pre-relativity view of the space and time, different reference frames
are related to each other by Gallilean transformations. The notion of “rela-
tivity” itself was already there. Namely, laws of physics remain the same in
any “inertial” reference frames that move relative to each other at constant
speed.2
Suppose the reference frame B is moving relative to the reference frame
A by velocity V~ . Their origins coincide at t = 0. If a particle moves at the
position ~x(t) in the frame A, its position is simply given by

~x0 (t) = ~x(t) − V~ t. (1)

This coordinate change is called Gallilean transformation. Just by taking

the time derivative of both sides, the velocities are related by
0
~v 0 ≡ ~x˙ = ~x˙ − V~ = ~v − V~ . (2)

Namely, velocities in two frames are related by a simple vector sum.

There is an implicit assumption in these equations that the time is com-
mon to both frames,
t0 = t (3)
It is normally not stated explicitly in discussions of mechanics, and is taken
for granted. No matter which frame you are in, time just flows the same,
right? This naive statement of course underwent big revision under relativity.

3 Michaelson–Morley experiment
Here, we follow a bottom-up approach to introduce relativity unlike the orig-
inal Einstein’s argument, namely starting with an experimental evidence.
Michaelson and Morley, using their sensitive interferometer mounted on
hills, demonstrated that the speed of light c is a fundamental constant inde-
pendent of the direction and the motion of the Earth itself.3 It implies that
2
Of course accelerating frames exhibit different laws of physics that include the cen-
trifugal force and Corioli’s force.
3
There were looking for the evidence of “ether”, a postulated medium that transmits
electromagenetic waves. People in those days could not accept the fact that waves can

2
change from one reference frame to another must be done in such a way that
the speed of light remains the same.
This experimental fact goes against Eq. (2), and hence the Gallilean trans-
formation itself. What we have to do is to change the relationship (coordinate
transformation) between two frames of reference. We try to figure out what
that is by requiring that the speed of light remains the same.
Consider small time and space intervals dt and d~x = (dx, dy, dz) in the
frame A. What we would like to figure out is what the corresponding intervals
are in the frame B, dt0 and d~x0 . For the light, the time and space intervals
are related by its speed,

(d~x)2 = (dx)2 + (dy)2 + (dz)2 = (cdt)2 . (4)

The fact that the speed of light is common to both frames means this is also
true in the frame B,

(d~x0 )2 = (dx0 )2 + (dy 0 )2 + (dz 0 )2 = (cdt0 )2 . (5)

In Gallilean transformation Eq. (1,3), the intervals in two reference frames

are given by

dt0 = dt, (6)

d~x 0
= d~x − V~ dt. (7)

If you insert these expressions into Eq. (4), you find

(d~x0 )2 = (d~x)2 − 2V~ · d~xdt + V~ 2 (dt)2 = (c2 + V~ 2 )(dt)2 − 2V~ · d~xdt. (8)

Therefore Eq. (5) is not satisfied.

propagate in “vacuum”. All other waves people were familiar with had some kind of
medium. Sound through air, ocean waves through water, etc. “Ether” was assumed to be
weightless and frictionless as not to disturb the planetary motion over billions of years.
If there existed “ether” as the medium of the electromagnetic wave, there must be its
“rest frame.” In any reference frame that moves relative to the rest frame, the speed of
light must be different. By measuring the speed of light along with and orthogonal to
the motion of the Earth, their experiment was sensitive enough to demonstrate this point.
This expectation was wonderfully crushed.

3
4 Lorentz Transformation
We are forced to look for coordinate transformations that preserve Eqs. (4,5).
The only assumption we make is that the transformation is linear , i.e., it is a
matrix multiplication on coordinates.4 Because it is awkward to deal with all
three spatial coordinates all the time, let us focus on z direction only for the
moment. We will generalize the result to the case of three spatial coodinates
later on. We use cdt and dz so that both time and space coordinates have
the same dimension. The matrix multiplication is simply
! ! ! !
cdt0 cdt Att (V ) Atz (V ) cdt
= A(V ) = . (9)
dz 0 dz Azt (V ) Azz (V ) dz
The Gallilean transformation Eq. (7) corresponds to the matrix
!
1 − Vc
A(V ) = . (10)
0 1
We start with requirements that do not concern with speed of light. First
obvious requirement is that an object at rest in frame A must appear to move
with velocity −V in frame B. In this case, dz = 0 in frame A because it is
not moving. In frame B, dz 0 /dt0 = −V . Therefore,
V dz 0 Azt cdt Azt
− = 0
= = . (11)
c cdt Att cdt Att
The next requirement is less obvious. If you go from frame A to B with
relative velocity V , and then combe back to A, which is another coordi-
nate transformation with relative velocity −V , you haven’t done anything.
Therefore, the matrices A(V ) and A(−V ) must be inverse of each other. On
the other hand, the coordinate transformation A(−V ) can be done by flip-
ping the z direction, doing the transformation A(V ), and then flipping the z
direction back. In other words,
A(−V ) = P A(V )P, (12)
where P flips the z coordinate,
!
1 0
P = . (13)
0 −1
4
Non-linear transformations appear in the general theory of relativity, Einstein’s theory
of gravity. Such transformations are called general coordinate transformations.

4
Therefore, the requirement is that A(−V )A(V ) = P A(V )P A(V ) = I, where
I is the unit matrix. By writing it out explicitly, we find
! !
A2tt − Atz Azt Atz (Att − Azz ) 1 0
P A(V )P A(V ) = = . (14)
Azt (−Att + Azz ) −Azt Atz + A2zz 0 1

The off-diagonal elements must vanish, and hence Att = Azz . Then the diag-
onal elements give the constraint A2tt − Atz Azt = 1. Together with Eq. (11),
we can solve it up to a single unknown parameter Att ,
 
A2tt −1
1 − Vc A2tt
A(V ) = Att (V )  V . (15)
−c 1

At this point, it is easy to see that Gallilean transformation Eq. (10) is a

special case of this general form with Att = 1. In fact, if you require that
“time” must be common to both frames, you need Att = 1, and Eq. (10)
is obtained as the unique choice under this requirement. Because “daily”
phenomena appear to support the idea that “time” is common no matter
how fast you go, people implicitly made this requirement and the Gallilean
transformation followed as the only possibility.
Now comes the crucial requirement that the speed of light is the same in
both frames. The requirement is that (cdt0 )2 −(dz 0 )2 = 0 if (cdt)2 −(dz)2 = 0.
We subsitute dt0 and dz 0 into Eq. (5) and find for light going right cdt = dz,

0 = (cdt0 )2 − (dz 0 )2
= (Att cdt + Atz dz)2 − (Azt cdt + Azz dz)2
= ((Att + Atz )2 − (Azt + Azz )2 )(dz)2 . (16)

This combination must vanish. Using the result obtained in Eq. (15), we find
 !2 
2
c A2tt − 1 V

A2tt  1 − − 1−  = 0. (17)
V A2tt c

Therefore,
A2tt − 1 V2
= , (18)
A2tt c2
and hence
1
Att = q . (19)
1 − V 2 /c2

5
We have now completely determined what coordinate transformation is con-
sistent with Michaelson–Morley experiment. As you have seen in the above
discussion, this is the unique choice. This is the Lorentz transformation that
allows us to take physical quantities from one reference frame to another.
It is conventional to use the notation
V
β = < 1, (20)
c
1
γ = √ > 1. (21)
1 − β2
Using this notation, the Lorentz transformation is written as
!
γ −γβ
A(V ) = . (22)
−γβ γ

This is the expression we will use a lot in special relativity.

When we consider all three spatial coordinates, the relative velocity be-
tween two frames is a vector V~ , qand we naturally define β~ = V~ /c. The
definition of γ is similar, γ = 1/ 1 − β~ 2 . Using the notation that β~ is a
column vector, the Lorentz transformation is given by

−γ β~ T
 
γ
~
A(V ) = ~ ~T ~ ~T  . (23)
−γ β~ (I − ββ~β2 ) + γ ββ~β2


The expression looks complicated, but what is means is very simple. The
lower 3 × 3 block has two terms. They decompose any vector into a piece
that is parallel to β~ by virtue of the projection β~ β~ T /β~ 2 , and another that is
orthogonal to β~ by its complement I − β~ β~ T /β~ 2 . The first term is not changed
because the vector orthogonal to the relative velocity V~ is not affected, while
the second term is multipled by γ.

5 Implications of Lorentz Transformations

There is one major difference between Gallilean transformation Eq. (10) and
Lorentz transformaion Eq. (22). It is that the “time” is different in two
frames after a Lorentz transformation.

6
 Suppose you consider a time interval dt in the rest frame of an object
(dz = 0). Then in a frame where the object is moving, you find
dt0 = γdt. (24)
This is the famous “time dilation” effect. For instance, a muon,5 a charged
particle similar to the electron, decays with a lifetime τ = (2.19703±0.00004)×
10−6 sec. At the velocity v = cβ, it appears that it travels only over distance
βcτ . However, because of the time dilation effect, it survives over much
longer time γτ , and hence the distance it can travel is γβcτ . This is why
muons, produced by the reaction of cosmic rays with atmosphere at an alti-
tude of 15–20 km, can reach the surface. You may have seen a demo where
a particle detector, often a spark chamber, literally keeps showing tracks of
muons going downward.
Another related point is the “Lorentz contraction.” In the rest frame
of the atmosphere of thickness dz 0 = L, the muon takes the time interval
dt0 = L/v = L/(cβ) to traverse it. In the rest frame of muon, the Lorentz
transformation gives
! ! ! !
cdt γ −γβ L/β L/(γβ)
= = . (25)
dz −γβ γ L 0
The atmosphere of certain thickness comes rushing on top of him with veloc-
ity V and it takes the time interval L/(γβc) to pass him. Therefore he would
conclude that the atmosphere had the thickness L/(γβc) × V = L/γ. It
appears “thinner” than it is in its rest frame. In the muon rest frame, he can
survive until the entire atmosphere passes him not because he lives longer
but because the atmosphere is squashed. In other words, Lorentz contraction
and time dilation are the same story told by different observers.
It is useful to check that Lorentz transformation reduces to Gallilean
transformation for small velocities V  c. For “daily” phenomena, typical
times intervals dt are measured in seconds, and typical distances dz in cen-
timeters. Given c = 3 × 1010 cm sec−1 , it follows that cdt  dz. Writing out
Eq. (22), we find
1 V
 
0
dt = q dt − 2 dz , (26)
1 − V 2 /c2 c
1
dz 0 = q (dz − V dt) . (27)
1 − V 2 /c2
5
We will discuss this particle a lot more later in this class.

7
In the first line, cV2 dz is completely negligible compared to dt, while both
dzqand V dt can be comparable in the second line. The prefactor γ =
1/ 1 − V 2 /c2 can be well approximated by 1 up to negligible corrections
of O(V 2 /c2 ). In the end the Lorentz transformation can be well approxi-
mated by
dt0 = dt, (28)
dz 0 = dz − V dt. (29)
This is nothing but the Gallilean transformation Eq. (7). In other words,
Lorentz transformation indeed does reduce to Gallilean transformation at
small velocities and.
Another important consequence of the Lorentz transformation is that the
combination
(cdτ )2 ≡ (cdt)2 − (dx)2 − (dy)2 − (dz)2 (30)
is the same in any reference frames. You can easily verify it using Eq. (22),
(cdτ 0 )2 = (cdt0 )2 − (dx0 )2 − (dy 0 )2 − (dz 0 )2
= (γcdt − γβdz)2 − (dx)2 − (dy)2 − (γdz − γβcdt)2
= (γ 2 − γ 2 β 2 )(cdt)2 − (dx)2 − (dy)2 − (γ 2 − γ 2 β 2 )(dz)2 = (cdτ )2 .(31)
A quantity that does not change from one reference frame to another is called
a “Lorentz invariant” or an “invariant” for short. In this case, dτ , a “proper
time,” is an invariant. Its physical meaning is the time interval in the rest
frame of an object.

6 Four-vector Notation
We have seen that time and space get “mixed up” under Lorentz transfor-
mations. It is therefore useful to consider time and space different compo-
nents of a single object, a four-component spacetime vector. The notation is
dxµ = (cdt, dx, dy, dz) where the Greek index µ runs from 0 (dx0 = cdt) to
3 (dx3 = dz). It is important that the index is a superscript. This is called
four-vector notation.
Lorentz transformation is given by the matrix Eq. (23) acting on the four-
vector dxµ written as a column vector. Instead of using matrices and column
vectors, the following notation is also used often,
(dx0 )µ = Aµ ν dxν . (32)

8
In this notation, Aµ ν corresponds to the matrix A whose (µ, ν) component
is Aµ ν . A repeated index (ν in this case) is always summed over from 0
to 3. This is often called “Einstein’s convention.”6 Note that the sum is
over a lower index and an upper index. This is also an important aspect
of the convention because of the reason we will see below. Any four-vector
that transforms the same way as the spacetime four-vector is said to be
“contravariant.” As we will see in the next section, energy and momentum
of an object are combined into a contravariant four-vector.
The proper time is an invariant. It is useful to introduce a “covariant”
vector that has a lower index,
dxµ = (cdt, −dx, −dy, −dz). (33)
Using this notation, the proper time can be written as
(cdτ )2 = dxµ dxµ . (34)
The point is that if you sum over a lower and an upper index, the indices
“cancel”, namely two indices combine to an invariant. In the case of Lorentz
transformation Eq. (32), the l.h.s. of the equation has only one index, while
the r.h.s. three indices. The point is that the lower index in Aµ ν and the
upper index in dxν are summed over and they “cancel.” Effectively the r.h.s.
also has only one index leftover and both sides of the equation are allowed
to be equated.
The way a contravariant and a covarint vector are related to each other
is given by a “metric tensor” gµν ,
1 0 0 0
 
 0 −1 0 0 
gµν =  
(35)
0 0 −1 0
 
 
0 0 0 −1
so that dxµ = gµν dxν . The inverse of the metric tensor has also the same
form
1 0 0 0
 
 0 −1 0 0 
g µν =  , (36)
 
 0 0 −1 0 
0 0 0 −1
6
Somebody told me that it wasn’t Einstein at all who invented this convention. I’m
not a historian to refute or verify this claim. Any volunteer?

9
so that dxµ = g µν dxν . The inverse relation can be written as gµν g νλ = δµλ
(Kronecker’s delta which is an identity matrix). Spacetime with the metric
of this form is called “Minkowsky space” as opposed to “Riemannian space”
whose metric is positive definite.

7 Energy and Momentum Four-Vector

The action of a system must be Lorentz invaraint. The only invariant for a
point particle is given by the proper time dτ . Then the natural candidate
for the action of a point particle is
Z
S=− mc2 dτ. (37)

You will see the reason for the factor −mc2 shortly below.
Because of its definition Eq. (30), the action can be written explicitly as
v
u !2
Z q Z u
2t 1 d~x
S=− mc (cdt)2 − (d~x)2 = − mc 1− 2 dt. (38)
c dt
The Lagrangian is given by the integrand
s
2 1 ˙2
L = −mc 1− ~x . (39)
c2
In the non-relativistic limit, we can Taylor-expand the Lagrangian in ~x˙ /c to
the second order and find
1 1 ˙2 1
 
2
L = −mc 1 − 2 ~x = −mc2 + m~x˙ .
2
(40)
2c 2
This is indeed the Lagrangian of a non-relativistic point particle up to a
constant term. The overall coefficient −mc2 in the action was chosen to
reproduce it.
Remember that the momentum is defined by
∂L
p~ = , (41)
∂ ~x˙
which gives m~x˙ in the non-relativistc case. In our relativistic case,
s
∂ 1 2 ~x˙
p~ = mc2 1 − 2 ~x˙ = m q ~
= mcγ β. (42)
∂ ~x˙ c ˙
1 − ~x /c
2
2

10
It is useful again to check the non-relativistic limit v  c. By Taylor-
expanding the expression for the momentum in v/c = |~x˙ |/c, the momentum
is indeed m~v up to corrections suppressed by ~v 2 /c2 .
The energy is given by the Hamiltonian E = p~ · ~x˙ − L. In the non-
relativistic case, it is

˙ 1 ˙2 1 ˙2 p~2
E = p~ · ~x − m~x = m~x = . (43)
2 2 2m
On the other hand in the relativistic case,
s
1 ˙2 mc2
E = p~ · ~x˙ + mc
2
1− 2
~
x = q
2
= mc2 γ. (44)
c 1 − ~x˙ /c2
2
Again by taking the non-relativistic limit, you find E = mc2 + 12 m~x˙ . The
first term is the rest energy of a particle E = mc2 .
By comparing the expressions of the energy and the momentum, it is easy
to see that  2
E
− p~2 = m2 c2 . (45)
c
No matter how fast the particle is moving, this combination is always the
same. In other words, it is an invariant. This observation suggests that we
can define a (contravariant) vector
E
pµ = ( , px , py , pz ). (46)
c
The above relationship can then be rewritten as

pµ pµ = m2 c2 . (47)

To see that this interpretation indeed makes sense, let us look at its
expression in the rest frame of the particle

pµ = (mc, 0, 0, 0). (48)

Then going to the frame moving with velocity −v along the z direction, the
Lorentz transformation gives

(p0 )µ = (mcγ, 0, 0, mcγβ). (49)

11
This precisely agrees E/c and p~ of the particle with what was obtained from
the Lagrangian.
In relativistic kinematics seen in particle physics, the velocity is often very
close to the speed of light, and it is not very useful to talk about it. Moreover,
the velocity itself is not a conserved quantity under collision processes, while
the energy and momentum are. Therefore, we almost exclusively talk about
energy and momentum, very rarely about the velocity. Especially when we
discuss massless particles (photon) or particles with tiny masses (electron,
neutrinos), the velocity is always (approximately) the speed of light, yet
energy can be any value. The formulae useful to remember are
q 1√ 2
E= c2 p~2 + m2 c4 , |~p| = E − m2 c4 . (50)
c
If you need the velocity, you can get it as

|~p|
v = c2 . (51)
E
It is amusing that the plane wave solution to the Schrödinger equation
p·~
x/h̄ −iEt/h̄ µ
ei~
e can be written as e−ipµ x /h̄ . Secretly, the Schrödinger equation
gave us a Lorentz-invariant expression for the wave function.7

8 Doppler Shift
As an application of Lorentz transformation and four-vectors, let us discuss
Doppler shift of light emitted from moving bodies, for example far-away stars
that are moving away from us because of expansion of Universe.8
The plane wave solution to Maxwell equation has time and space depen-
~
dence e−iωt+ik·~x . The exponent suggests that we can define contravariant
wave four-vector k µ = (ω/c, ~k). The plane wave solution then has manifestly
µ
Lorentz-invariant form e−ikµ x .9 This is enough information for us to figure
out the Doppler shift of light.
7
Of course, the non-relativistic Schrödinger equation does not give the relativistic re-
lation between the energy and the momentum.
8
To correctly calculate the redshift in expanding Universe, however, general relativistic
effects must also be taken into account.
9
If you compare it to the quantum mechanical plane wave solution in the previous
section, the relationship is obvious: kµ = pµ /h̄.

12
 The dispersion relation of light is ω = c|~k|. For light propagating along
the z-direction, therefore, the wave four-vector is simply k µ = (ω/c, ~k) =
(ω/c)(1, 0, 0, 1). Suppose ω is the frequency of light in the rest frame of the
emitter. If the light emitter is moving away from us with velocity v, all we
need to do is to Lorentz-transform this four-vector. Using by-now familiar
matrix Eq. (22), we find (k 0 )µ = (ω/c)(γ − γβ, 0, 0, γ − γβ). In other words,
the new (angular) frequency is related to the one in the rest frame of the
light emitter by
v
1 − v/c u 1 − v/c
u
ω 0 = ω(γ − γβ) = ω q = ωt . (52)
1 − v 2 /c2 1 + v/c

This must be a familiar formula from Physics 7C.

Because light emitted from atoms has definite spectrum that can be mea-
sured in the laboratory, and spectroscopy can be done very accurately, mea-
suring redshift of distant stars is a very powerful tool in astronomy.

9 Natural Unit
In the world of particle and nuclear physics where both relativity and quan-
tum mechanics play crucial roles, it is cumbersome to keep writing the speed
of light c and the reduced Planck constant h̄. It is customary to set c = h̄ = 1.
Why is this allowed?
Remember in the cgs system (also true in the MKS sytem), the funda-
mental units have dimensions of length (L), mass (M), and time (T). Other
units are derived from these three. For example, ampère is defined by the
force between two currents and it does not have to be a new unit. Indeed, in
esu unit in the cgs system, electric charge and current are defined in terms
of cm, g, and s only.
Given the three fundamental dimensions, we have the freedom of relating
them among each other. For example, the length can be measured in the
unit of time using the speed of light c which has dimension LT−1 . In daily
life, it is extremely inconvenient to talk about “length of 33ps” if you mean
just “a centimeter.” (Note ps is pico-second, 10−12 sec.) But in the world
of particle physics, it is not inconvenient at all. In fact, a typical size of a
particle detector 1–10m translates to nanoseconds, which immediately put
challenging demands on electronics. This way, L and T are now equivalent.

13
 Similarly, the Planck constant has dimension L2 MT−1 . It can be used to
relate the dimension of mass to other two. Most physicists use electronvolt
eV as the unit for energy. Using the constant h̄c = 197MeVfm (MeV is
106 eV, fm is femto-meter 10−15 m), we can relate the length and the energy.
Remember this constant! Any particle physicist remembers this number.
For example, we often refer to the mass of electron as me = 0.511 MeV/c2 .
For short, we don’t quote c2 . We just say the mass as 0.511 MeV. As long as it
is understood that we always use c and h̄ to relate different dimensions, there
is no source of confusion. When the value is quoted in MeV, in order to get
the dimension of mass, all you need to figure out is that you need a factor of
1/c2 (remember E = mc2 !). If you want, you can then work out the mass in SI
unit, by using conversion constans eV = 1.60×10−19 J and c = 3.00×108 m/s
as me = (0.511×106 eV)(1.60×10−19 J/eV)/(3.00×108 m/s)2 = 9.08×10−31 kg.
(Actually, it is 9.11×10−31 kg. We made this error because we kept only three
significant digits.)
If the electron is moving with energy 1 MeV, we work out the momentum
by √
p = E 2 − m2 = 0.860MeV. (53)
What we mean is that the momentum is 0.860 MeV/c, but we don’t write c.
Only when you want to convert the momentum to the SI unit, you need to
remember that there must be a factor of 1/c. Otherwise, you can consistently
drop c everywhere. This way, we quote values for energy, momentum, mass
all in eV unit.
If you want the wave length λ, we of course use de Broglie’s formula
λ = h/p = 2πh̄/p. But if you use natural unit, we quote p in eV, which is
the same thing as writing it as λ = 2πh̄c/cp. Then the only constant you
need to remember is h̄c. We immediately obtain λ = 1440 fm. This is the
“resolution” of an experiment where 1 MeV electron is scattered.
There is another fundamental constant which can be used to eliminate
any units: Newton’s constant GN . It has the dimension L3 M−1 T−2 . It can
2
be used to construct a unit for energy, EPlanck = h̄c3 /GN = (1.22×1019 GeV)2
called Planck energy. Correspondingly, the Planck length is `Planck = h̄c/EPlanck =
1.61×10−33 cm, an extremely short distance. Then any physical quantity can
be expressed in terms of pure numbers with no dimensions. However, this
energy is so enormous that it is not used as a unit even in particle physics.
On the other hand, this energy (length) scale signals where quantum effects
of gravity need to be considered. In fact, the candidate theory of quantum

14
gravity, string theory, is often discussed using this unit.10
Note that it is not only particle and nuclear physicists who choose the
unit system for convenience. Atomic physicists also use the same freedom to
fix all three dimensions, by setting e = me = h̄ = 1. It is called atomic unit.

10
Actually, what is used as the unit in string theory is not
√ precisely the Planck unit,
but a combination of string coupling constant and EPlanck / 8π is set to unity.

15