1 Relativistic momentum

The momentum of a body in an inertial frame should depend only on its rest
mass and velocity in this frame. This, of course, is not enough to define p⃗
uniquely, so we add the following axioms:

1. A restriction on direction of p⃗: it must be the same as the direction of ⃗v .

2. Due to isotropy of space, the absolute value of p⃗ should only depend on
the mass and the absolute value of the velocity.
3. Consider two objects having the same velocity and location and masses
m1 and m2 . We would very much like our physics to be linear, so the
momentum of the system of these two objects should equal the sum of the
two momenta, so we postulate p⃗ to be proportional to m.

We thus infer that p⃗(m, ⃗v ) is of form m⃗v · ϕ(v) (notice that the fact that the
momentum of a body at rest is zero is a theorem).
Additionally, conservation of net momentum of a closed system should apply.
We will not apply any particular mechanics for this restriction; instead, we will
require a conditional: if momentum of a system is conserved in one inertial
frame, it is also conserved in any other inertial frame.
Now, what form does ϕ(v) take? To figure that out, consider two balls of
mass m with velocities ⃗u and −⃗u (in K) respectively colliding elastically in plane
Oxy:
⃗u −⃗u

1 2

Due to symmetry (I am not saying ”due to conservation of net momen-

tum”, because we don’t know if this conservation exists yet), the velocities after
⃗ and −U
collision are opposite. Let them be U ⃗:
⃗
U −U⃗

1 2

Now, what is the resulting velocity U ⃗ ? This is the first problem we run into.
The problem is we don’t know if there is such a thing as conservation of energy,
or that (in)elastic collisions are possible for high velocities.
What we can do is derive the formula for ϕ in assumption that the balls
collide in such a way that Ux = ux and Uy = −uy (as if they ”bounced off” each
other), and then show that this formula also works for other U ⃗ s.
′
For an inertial frame K moving with the speed v = ux along the X axis wrt.
K, we expect the conservation of momentum to hold as well.
The formulae for the net momentum in K ′ before and after collision are:

1
p⃗′ = m(u⃗′1 ϕ(u′1 ) + u⃗′2 ϕ(u′2 )),
P⃗ ′ = m(U⃗ ′ ϕ(U ′ ) + U⃗ ′ ϕ(U ′ )).
1 1 2 2

If p⃗′ is to be conserved, then p′x = Px′ and p′y = Py′ . It is easy to check that
the former holds for all ϕ, and after substituting velocities to the latter and
using some symmetries we get:

ϕ(u′1 ) u′2y
= .
ϕ(u′2 ) u′1y
The ratio of u′y can be computed using the velocity addition formula:

r r
v2 uy u2 uy
u′1y = 1− 2 · 1 − 2x ·
= = γ(uy ) · uy ,
c 1 − vux /c2 c 1 − u2x /c2
r r
v2 −uy u2 −uy
u′2y = 1− 2 · 2
= 1 − 2x · .
c 1 + vux /c c 1 + u2x /c2
ux can be calculated from u2x as follows:

−ux − v −2ux
u′2x = 2
= =⇒
1 + vux /c 1 + u2x /c2
−u′2x
ux = p .
1 − (u′2x )2 /c2 + 1

Thus

r
ϕ(u′1 ) 1 − u2x /c2 2 u′2x (u′2x )2
′ = = −1 + = −1 − = 1− .
ϕ(u2 ) 1 + u2x /c2 1 + u2x /c2 ux c2
Now, the right-hand side does not depend on the vertical velocity of the
balls. This implies that the left-hand side does not depend on uy . In particular,
for infinitesimal uy : u′1y and u′2y approach zero, and u′1x = 0, hence u′1 = 0 and
u′2 = u′2x and thus:

ϕ(u′1 ) γ(u′1 )
= .
ϕ(u′2 ) γ(u′2 )
Now, this statement is coordinate-independent and thus holds for all col-
lisions similar to the one used. Also notice that this formula uses u′1 and u′2
instead of u1 and u2 , and while the latter two are opposite velocities, there is
no requirement on the former. It is in fact possible to construct a collision for
any pair of u1 and u2 , and thus this statement holds for absolutely arbitrary
velocities.

2
Thus, ϕ(v) is of form αγ(v), where α is a universal constant. There is no
way to narrow down this any further, because all functions of such form satisfy
any reasonable dimensionless requirements. However, it seems reasonable for
the well-known Newtonian momentum to approximate relativistic momentum
for small velocities, and this yields α = 1.
The relativistic momentum of a body of mass m and velocity ⃗v is:

p⃗ = γ(v) · m⃗v .
It can be shown algebraically that if net momentum of a system of bodies is
conserved in any one inertial frame, it is also conserved in other inertial frames.
Whether this is indeed the only possible well-defined formula for ϕ is up in
the air and apparently depends on the mechanics used.

2 Relativistic energy
Let F⃗ (⃗q, ⃗v , t) be a vector field indicating the force applied to a particular object
at coordinate ⃗q having velocity ⃗v at time t in inertial frame K.
The force satisfies F⃗ (⃗q, ⃗v , t) = d⃗p
dt .
We define the kinetic energy of a body as
Z B
∆K = F⃗ d⃗q.
A
That is, the increase in the kinetic energy equals the work on the body. This
requirement allows infinitely many definitions of kinetic energy, only differing
in the integration constant. Hence we also require the kinetic energy of a body
at rest to be zero.
The kinetic energy of a body of mass m is thus

Z Z Z Z Z
⃗ d⃗
p d⃗q
K = F · d⃗q = · d⃗q = d⃗p· = ⃗v · d⃗
p = ⃗v · p⃗ − p⃗ · d⃗v =
dt dt
Z Z
1
= γmv 2 − m γ⃗v · d⃗v = γmv 2 − m γ · d((⃗v )2 ) =
2
2
mc2
Z
1 v
= γmv 2 + mc2 γ · d 1 − 2 = γmv 2 + + C = γmc2 + C.
2 c γ

Where C is the constant of integration. For ⃗v = 0, we want K = 0 = mc2 +C,

hence C = −mc2 and

K = (γ − 1)mc2
is the kinetic energy of a body.
Now, most textbooks bring the constant of integration to attention and say
that, in fact, if we take C = 0 (and what is more natural than a zero?), we get
another formula for energy:

3
E = γmc2
Which is full energy, composed of kinetic energy and rest energy:

E0 = mc2
K = (γ − 1)mc2
E = γmc2

The rest energy is said to be the energy stored within matter, that is, the
binding energy.
This hand-waving brings up several questions:

1. The constant of integration is absolutely arbitrary, and so is rest energy.

In fact, the only reasonable limitation on rest energy is its linearity with
respect to mass; that is, E0 = αmc2 where α is a universal dimensionless
constant. Notice that this c2 coefficient does not come from Maxwell equa-
tions or Lorentz transformations: it is just the simplest value of dimension
energy
mass . Where does α = 1 come from?

2. Energy is defined as the largest amount of work a body can exert on its
surroundings. Can any body really exert exactly mc2 of energy? Is there
a body that can perform more work?
3. Mass having a role in rest energy seems somewhat arbitrary: Newton’s
second law for a body at rest does not include the mass of said body, so
why wouldn’t the rest energy depend on, say, electric charge as well?

Einstein tackles the first question as follows:

Consider a body emitting two symmetric electromagnetic plane waves of
energy L/2 each. Due to symmetry, the velocity of the emitting body won’t
change, but due to law of conservation of energy, the internal energy of the
body determined by mass will decrease.
In a frame K where the body is at rest, the full energies of the body before
(E) and after (H) the emission are connected as follows:

E−H =L
In a frame K ′ moving with speed v perpendicular to the plane waves, the
amount of energy emitted increases in a way similar to that of kinetic energy
(look up on transformation of electromagnetic fields between inertial frames for
more information), and so:

E ′ − H ′ = L′ = γL
Einstein notes that E and E ′ are the energies of the same body in different
frames. The only difference between them is the velocity v of K ′ with respect
to K. Since velocity only affects kinetic energy, we acquire:

4
E ′ − E = EK
′
− EK = (γ − 1)m0 c2
Where m0 is the original mass of the body. Similarly,

H ′ − H = HK
′
− HK = (γ − 1)m1 c2
Where m1 is the mass after the emission. Combining the four equations
yields:

(m0 − m1 )c2 = L
Hence, a decrease of ∆m in mess results in emission of ∆mc2 of energy, and
vice versa. This implies that a mass-less body cannot emit energy, and thus

E0 = mc2
is the correct formula for energy at rest.
One oft arising question about this proof is that Einstein implicitly postu-
lates a single body as seen from two inertial frames being identical to two bodies
only differing in velocity, and not in mass, charge, or otherwise.
The logic in lack of difference in mass is apparent. We think of mass as a
property intrinsic to matter. When we derived the expressions for momentum
and energy, we were thinking of mass as a constant, not changing upon accel-
eration or switch of frames. So it’s really no wonder rest energy, denoting an
intrinsic property of matter, is somehow connected to mass.
The absence of any other influencing factors is due to the definition of matter.
When we compute the rest energy of a body, we actually compute the rest
energy of its matter, and not potential energy with regards to fields, be they
electromagnetic or otherwise. Mass has a unique property of influencing the
behavior of a particle in the now, and not via fields.
After seeing this derivation, other questions raise up:

1. Einstein requires velocity to be perpendicular to the wave front for L′ /L

to be a constant. What would happen for other angles?
2. This derivation assumes that a body can emit light at all. Hydrogen at
room temperature cannot emit light–does this mean that hydrogen has
zero rest energy for low temperatures?

3. Bodies do not actually emit light in plane waves. Would the results differ
for other kinds of waves? And what if energy of a different kind was
emitted?

The first question is easy to answer. Einstein used ϕ = 0 purely for opti-
mization. A more complete expression for transformation of energy gives
v
L′ = γL 1 − cos ϕ .
c

5
When two plane waves are emitted at symmetric angles ϕ and π + ϕ, the net
energy is
L v L v
L′ = γ 1 − cos ϕ + γ 1 − cos(π + ϕ) = γL,
2 c 2 c
which yields the same result.
The answer to the second question is that we assume that full energy can be
separated into two parts: rest energy that is a function of the internal state of a
body and does not depend on velocity, and kinetic energy that only depends on
velocity and mass, and not on the internal state. This is an arguable distinction:
Planck has shown in 1907 that such a separation is impossible for black cavity
radiation. If, however, we believe that this is possible for matter, then rest
energy does not depend on temperature and therefore hydrogen indeed has the
same rest energy at all temperatures.
As for the third question, this is an implication of law of conservation of
energy. Consider body A emitting a plane wave of energy L which leads to a
decrease of ∆m1 = L/c2 in mass, and body B emitting a different kind of energy
of the same magnitude that leads to a different decrease in mass ∆m2 < ∆m1 .
Then let the second body lose additional ∆m = ∆m1 − ∆m2 mass as plane
wave radiation. The masses of the two bodies are now identical, but the second
body has emitted additional ∆mc2 amount of energy, which contradicts law of
conservation of energy–that is, assuming rest energy is a function of mass, of
course.
So yes, E0 = mc2 is in fact a conjecture or an implication of several laws
we took as-is from Newtonian mechanics. That it holds in reality according to
experiments is a nice bonus. Binding energy, for instance, almost sums up rest
energy, and the behavior of mass-less and massive elementary particles upon
conversion also confirms the theory.
There is also artificial significance in rest energy being exactly mc2 : it helps
link energy to momentum.
If E0 = mc2 and K = (γ − 1)mc2 , then total energy E = γmc2 . After
performing some algebraic transformations we get

2
2 2 4 Ev
E =m c + =⇒
c
E 2 = m2 c4 + (γmv)2 c2 =⇒
E 2 = m2 c4 + p2 c2 .
This is one of the more well-known identities of special relativity. After some
more reordering we get

(E0 /c)2 = (E/c)2 − p2 ,

and this can be interpreted as E0 /c being the absolute value of a vector in
4-space, called 4-momentum:

6

E0 E
= ∥p∥ = , px , py , pz ,
c c

where the metric signature is chosen to be (+, −, −, −). As E0 /c is the same

across inertial frames, it is Lorentz-invariant just like space-time interval.
Moreover, 4-momentum behaves the same way as 4-position under Lorentz
transformation, that is, it is Lorentz-covariant.
The Lorentz transformation for 4-position (ct, x, y, z) is:

v
ct′ = γ ct − x ,
c
v
x′ = γ x − ct ,
c
′
y = y,
z ′ = z.

And the Lorentz transformation for 4-momentum (E/c, px , py , pz ) is:

E′

E v
=γ − px ,
c c c

Ev
p′x = γ px − ,
c c
p′y = py ,
p′z = pz .

This is incredibly obvious for a body that is at rest in frame K, that is, for
p⃗ = ⃗0:

E′ E γmc2
=γ = ,
c c c
Ev
p′x = −γ = −γmv,
c c
p′y = 0,
p′z = 0.