CHAPTER 3

“THE HAPPIEST THOUGHT” OF EINSTEIN’S LIFE

For an observer falling freely from

the roof of a house there exists -
at least in his immediate
surroundings - no gravitational
field [. . . ]. The observer therefore
has the right to interpret his state
as ’at rest’ (at least until he hits
the ground!).

A. Einstein (1920)

The Poisson equation for the gravitational field

∇2 Φ(t, xi ) = 4πGρ(t, xi )

is a linear partial differential equation of 2nd order which does not contain any explicit time depen-
dence. The gravitational potential responds instantaneously to the changes in the matter distribution!
This was awkward even for Newton
That one body may act upon another at a distance through a vacuum, without the mediation
of anything else, by and through which their action and force may be conveyed from one to
another, is to me so great an absurdity, that I believe no man, who has philosophical matters
a competent faculty of thinking, can ever fall into it (Principia, p. 643, Ref. 395).

and it is in clear contradiction with Special Relativity. The instinctive reaction of many physicist when
facing this consistency problem was to apply the recipes used when writing the covariant version of
Maxwell equations (promote the operator ∇ to 2, introduce some kind of vector potential Aµ for
the gravitational field, generalize the Newtonian force to some combinations of fields and 4-velocities,
get retarded potentials, etc . . . ). None of the attempts was sucessful1 . Einstein eventually concluded
that a new approach to the problem must be taken. The purpose of this chapter is to present you
Einstein’s new look on gravity. As you will see, the new look turned out to be a real old look that
went back to Galileo himself.
1 We will be back to this point in the future.
3.1 Inertial and gravitational masses 30

3.1 Inertial and gravitational masses

Two different masses enter in the Newton’s theory of mechanics and gravitation. According to New-
ton’s second law the acceleration ai = d2 xi /dt2 experienced by an object is proportional to the exerted
force divided by the inertial mass of the object

f i = mI ai , (3.1)

independently of the origin of the force. This mass mI measures the resistance of an object to ac-
celerations. On the other hand, we have the gravitational mass, which measures the strength of the
gravity (in the same way that the electric charge measures the strength of the electric force). The
force exerted on a gravitational mass mG close to the surface of the Earth is given by

f i = mG g i . (3.2)

Comparing the expressions (3.1) and (3.2), we conclude that the acceleration of gravity should depend
a priori on the ratio of the gravitational mass to the inertial mass.
 
mG
ai = gi . (3.3)
mI

Nevertheless, as verified by Galileo’s ramp experiments2 , all bodies fall with the same acceleration in
a gravitational field
ai = g i . (3.4)
This observation implies the equality of the quantity controlling inertia (mI ) and that measuring the
strength of gravity (mG )
mI = mG , (3.5)
for all materials, independently of its composition.

Exercise
Consider the magnitude of the electrostatic interaction at a distance r between two particles of
charges q1 , q2 and inertial masses m1i ,m2i
q1 q2
Fe = . (3.6)
4πr2
How does the magnitude of the acceleration felt by particle 2 depends on its properties?
2 Yes, ramps and a water clock. The image of Galileo dropping balls from the leaning power of Pisa is just a widespread

italian legend.
3.1 Inertial and gravitational masses 31

3:00 am

3:00 pm

The results of Galileo’s experiments were confirmed, among others, by Newton himself and by the
Baron Eötvös de Vásárosnamény, who used respectively pendula and a torsion balance with different
materials3 .

Exercise
How does the oscillation period of a simple pendulum depend on the ratio mI /mG ?

The difference in the acceleration experienced by the two bodies is encoded in the so-called Eötvös
parameter
E1I E2I
 
∆a 2|a1 − a2 | X I
η= = = η − , (3.7)
a |a1 + a2 | mI,1 c2 mI,2 c2
I
where in the last step we have made explicit the contribution of the various energy forms E I to the
difference between inertial and gravitational masses
X EI
mG − mI ≡ ηI 2 . (3.8)
c
I
The experimental results are summarized in the following table.

|mI −mG |
mI

Rest mass, proton and neutrons < 10−11

Rest mass, electrons < 2 × 10−8
Electric fields in nucleus < 4 × 10−10
Magnetic fields in nucleus < 2 × 10−7
Strong fields in nucleus < 5 × 10−10
Weak fields in nucleus < 10−2
Gravitational energy in Earth < 2 × 10−3

3 Eötvos located two test objects on the opposite ends of a dumbbell suspended from a forsion fiber. If the inertial

and gravitational masses of those objects were different the centripetal effects associated with the rotation of the Earth
would give rise to a torque (everywhere but at the poles) that could be measured with a delicate torsion balance pointing
west-east. For a detailed description of the Eötvös’ original experiments see for instance Weinberg’s book.
3.2 The Equivalence Principle 32

Figure 3.1: Lunar Laser Ranging (LLR) experiments

Note that the displayed cases do not include the contribution of the gravitational self-interaction
of the masses, which, for laboratory size experiments, is extremely small. Its contribution can be
however tested via the so-called Nördvedt effect. If the gravitational self-energy did not follow Galileo’s
equivalence principle, the Earth and the Moon would fall at different rates towards the Sun, elongating
the orbit of the Moon in the Sun direction. As shown by Lunar Laser Ranging experiments (LLR),
which use reflectors that were located in the surface in the Moon during the Apollo 11 mission, cf.
Fig. 3.1, the gravitational self-energy behaves as any other energy form, in perfect agreement with
Eötvos’ results. Indeed, LLR experiments provide the most accurate tests of the equivalence between
inertial and gravitational masses. The constrains are really impressive

2|aE − aM |
η= = (−1 ± 1.4) × 10−13 . (3.9)
(aE + aM )

Not only matter but also antimatter seems to follow the Galileo’s result. Important constraints of the
order 10−9 were obtained by the CPLEAR collaborations from neutral kaon systems4 .

3.2 The Equivalence Principle

The experimental equality of inertial and gravitational masses is a quite surprising and mysterious
property, relating two completely different concepts and not required at all for the consistency of New-
ton’s theory. For Galileo and Newton, this was just a coincidence. For Einstein, it would be the first
stone in the impressive geometrical edifice of General Relativity. Einstein’s theory will be constructed
on top of something so simple that even Galileo could have discovered: the relation between gravity
and inertia. To illustrate this equality let me consider one of the most famous Einstein’s Gedankenex-
perimente. Imagine yourself dropping a ball in the surface of the Earth. You will see the ball falling
with constant acceleration. “The effect of gravity”, you would say. Now imagine yourself performing
the same experiment inside a completely isolated rocket in outer space which moves with a constant
acceleration a = g. You will observe the same: the ball falling with constant acceleration. Without
knowing it, you would not be able to decide if you were in the true gravitational field of the Earth or in a
rocket! This apparently trivial observation is summarized in the so-called Weak Equivalence Principle:

4 Under the assumption of exact CPT symmetry.

3.2 The Equivalence Principle 33

g g

Figure 3.2: Einstein’s Gedankenexperiment

Einstein’s Equivalence Principle

The trajectories of particles in the gravitational field are locally indistinguishable from the
trajectories of free particles as viewed from an accelerated reference frame.

The gravitational interaction resembles also the pseudo forces resulting from the use of non-inertial
reference frames. For example, if there is a frame of reference rotating with angular velocity ω with
respect to an inertial reference frame, all bodies appear to accelerate spontaneously with the same
acceleration in that rotating frame. It seems that there is a universal force acting on all bodies with
a magnitude proportional to their inertial masses

F = −mI [ω̇ × r + 2 ω × ṙ + ω × (ω × r)] . (3.10)

Accelerated frames and local gravitational forces appear to be intimately related. Both of them act
in the same way on all bodies, are proportional to mass and can be transformed away by changing to
a suitable reference frame; a local free falling frame in the case of gravity.

Figure 3.3: A local inertial reference frame.

3.3 Life in the rocket: Rindler spacetime 34

Exercise
Einstein’s toy: A version of the following device was constructed as a birthday present for
Albert Einstein. The device consists of hollow broomstick with a cup at the top, together with
a metal ball and an elastic string. When the broomstick is held vertical, the ball can rest in
the cup. The ball is attached to one end of the elastic string, which passes through a hole in
the bottom of the cup, and down the hollow centre of the broomstick to the bottom, where its
other end is secured. You hold the broomstick vertical, with your hand at the bottom, the cup
at the top, and with the ball out of the cup, suspended on its elastic string. The tension in the
string is not enough to draw the ball back into the cup. The problem is to find an elegant way
to get the ball back into the cup. (Inelegant ways are: using your hands or shaking the stick up
and down).

3.3 Life in the rocket: Rindler spacetime

Since accelerated reference frames mimic the local effects of the gravitational field, the understanding
of their properties seems to be a first step towards the correct description of gravity. Let me start by
analyzing the rocket Gedankenexperiment presented above.

Note the slight change of notation below. From now on, we will reserve the first letters of the
Greek alphabet α, β, . . . for indices associated to inertial reference frames. Intermediate letters
of the Greek alphabet µ, ν, . . . will stand for general (non-inertial, accelerated) reference frames.

Consider the movement of the accelerated rocket from the point of view of an inertial observer {ξ α },
momentarily at rest with respect to the rocket’s trajectory. The orientation of the coordinate grid is
such that the rocket moves along the ξ 3 -direction. In that instantaneous inertial frame, the rocket is
seen to undergo a constant acceleration g

uµ = (1, 0, 0, 0) , aµ uµ = 0 → aµ = (0, 0, 0, g) , aµ aµ = g 2 . (3.11)

In order to determine the wordline ξ α (τ ) of the rocket at later times, let us look for a general solution
of the covariant equation uα uα = −1. Writing it explicitly

ηαβ uα uβ = −(u0 )2 + (u3 )2 = −1 , (3.12)

the solution becomes pretty obvious

u0 = cosh f (τ ) , u3 = sinh f (τ ) . (3.13)

The unknown function f (τ ) can be determined by taking the derivative of the last two equations
duα
aα = = f˙(τ ) (sinh f (τ ), 0, 0, cosh f (τ )) (3.14)
dτ
and imposing the covariant condition aµ aµ = g 2 . We get g 2 = f˙2 , f (τ ) = gτ and

uα = (cosh gτ, 0, 0, sinh gτ ) . (3.15)

The work is almost done. Integrating (3.15) with the initial condition ξ α (0) = (0, 0, 0, g −1 ), we get

ξ α (τ ) = g −1 (sinh gτ, 0, 0, cosh gτ ) . (3.16)

3.4 Beyond inertial observers 35

The “constantly accelerated” rocket describes an equilateral hyperbola with semi-major axis 1/g

(ξ 3 )2 − (ξ 0 )2 = g −2 . (3.17)

Let us now look at the problem from the point of view of an accelerated observer sitting in the rocket.
Since the transformation from inertial to accelerated frames is not a Lorentz transformation, we should
expect a change in the Minkowski line element ds2 . The natural coordinates for the accelerated
observer are those adapted to its trajectory. Let’s call them (x0 , x3 ) = (η, ρ). The transformation to
this frame takes the form5

ξ 0 (η, ρ) = ρ sinh η , ξ 3 (η, ρ) = ρ cosh η . (3.18)

In terms of the new coordinates η and ρ, the Minkowski line element ds2 = −(dξ 0 )2 + (dξ 3 )2 becomes
modified
ds2 = −ρ2 dη 2 + dρ2 ≡ gµν dxµ dxν . (3.19)
The metric gµν = diag(−ρ2 , 1) is now space-time dependent!

3.4 Beyond inertial observers

Let us formalize the concepts appearing in the previous example. The distance between two neigh-
boring points, as measured in a local inertial frame at rest with respect to the particle6 , is given by
the Minkowski metric ηαβ
ds2 = ηαβ dξ α dξ β . (3.20)
When performing a transformation to a general coordinate system7 xµ = xµ (ξ α ), the line element
becomes modified
∂ξ α µ ∂ξ β ν ∂ξ α ∂ξ β
ds2 = ηαβ dξ α dξ β = ηαβ µ
dx ν
dx = ηαβ µ ν dxµ dxν = gµν dxµ dxν . (3.21)
∂x ∂x ∂x ∂x
and distances are no longer determined by the Minkowski metric, but rather by a metric

∂ξ α ∂ξ β
gµν = ηαβ , (3.22)
∂xµ ∂xν
which generically depends on the spacetime coordinates. The inverse of the new metric is defined
through the relation g µν gνλ = δ µ λ and can be easily computed by taking into account the identity

∂xµ ∂ξ β
= δβ α . (3.23)
∂ξ α ∂xµ
We get
∂xµ ∂xν
g µν = η αβ . (3.24)
∂ξ α ∂ξ β

Exercise
Prove the similarity transformation (3.24).

5 The change of coordinates is just the Lorentzian analogue of polar coordinates. inspired on the wordline equation

(3.16).
6 In the context of a particle in the gravitational field these frames are called local free falling reference frames.
7 General means completely arbitrary. It can be a curvilinear coordinate system, an accelerated system, a rotating

system . . . whatever you want.

3.5 The geodesic equation 36

Note that the reference frame xµ is not at all privileged. We could perfectly move now into another
non-inertial reference frame x̄ρ (ξ α ) in which

∂xµ ρ ∂xν σ ∂xµ ∂xν ρ σ

ds2 = gµν dxµ dxν = gµν dx̄ dx̄ = gµν dx̄ dx̄ = ḡρσ dx̄ρ dx̄σ , (3.25)
∂ x̄ρ ∂ x̄σ ∂ x̄ρ ∂ x̄σ
with
∂xµ ∂xν
. ḡρσ = gµν (3.26)
∂ x̄ρ ∂ x̄σ
The results of this section are summarized in Figure 3.4.

(⇠ ↵
)
⌘↵

x̄
⇢
(⇠
xµ

↵
)
gµ⌫ ḡ⇢
x̄⇢ (xµ )

domingo, 15 de septiembre de 13

Figure 3.4

Exercise
Show that gµν must be symmetric, i.e. gµν = gνµ .

3.5 The geodesic equation

The equation of motion for a free particle in an accelerated reference frame can be obtained by applying
the transformation (3.22) to the Lagrangian of a free relativistic particle (2.51). We get

dxµ dxν
 
1
Z
S= dσ e−1 (σ)gµν − m2 e(σ) . (3.27)
2 dσ dσ

As in the Minkowski case, the action (3.27) is invariant under reparametrizations of the path σ →
σ = f (σ) provided that we let e(σ) transform in such a way that the quantity e(σ)dσ is left invariant.
Note also that the action (3.27) is invariant under general coordinate transformations8 , as can be
easily seen by taking into account the similarity relation (3.26).
∂L
The Euler-Lagrange equation ∂e = 0 for the non-dynamical variable e(σ)

dxµ dxν
gµν = −m2 e(σ)2 , (3.28)
dσ dσ
automatically incorporates the massive (e(σ) = 1/m) and massless (e(σ) = 1, m → 0) cases we are
interested in. In these two limits, the action (3.27) takes the form

dxµ dxν dxµ dxν

   
1 1
Z Z
Smassive = m dσ gµν −1 , Smassless = dσ gµν , (3.29)
2 dσ dσ 2 dσ dσ
8 This should be expected from pure Lagrangian mechanics.
3.5 The geodesic equation 37

with σ = τ for the massive case. Smassive and Smassless are very similar. Indeed, the computation of
the equations of motion is formally equivalent in both cases9 . Let us denote by a dot the derivative
with respect to σ and forget in what follows about the irrelevant factors m and m/2. The equations
of motion  
d ∂L ∂L
ρ
− =0 (3.30)
dσ ∂ ẋ ∂xρ
for the generalized coordinates xρ can be computed as follows. The simplest part is the variation of
1/2gµν ẋµ ẋν with respect to xρ . All the dependence on the coordinates is hidden in the metric
∂L 1 ∂gµν µ ν
= ẋ ẋ . (3.31)
∂xρ 2 ∂xρ
The variation with respect to ẋρ is slightly more involved, but can be however written in a very
compact way by taking into account the properties
∂ ẋµ
= δµ ρ gµν δ ν ρ = gµρ , gµν = gνµ , (3.32)
∂ ẋρ
together with some simple index relabeling. We get
∂ ẋµ ν ν
   
∂ 1 µ ν 1 µ ∂ ẋ 1
ρ
gµν ẋ ẋ = gµν ρ
ẋ + g µν ẋ ρ
= (gρν ẋν + gµρ ẋµ ) = gρν ẋν . (3.33)
∂ ẋ 2 2 ∂ ẋ ∂ ẋ 2
Collecting the two pieces, the Euler-Lagrange equations (3.30) become
d ∂L ∂L d 1 ∂gµν µ ν
− = (gρν ẋν ) − ẋ ẋ
dσ ∂ ẋρ ∂xρ dσ 2 ∂xρ
∂gρν σ ν 1 ∂gµν µ ν
= ẋ ẋ + gρν ẍν − ẋ ẋ
∂xσ  2 ∂xρ 
∂gρν 1 ∂gσν
= gρν ẍν + ẋν ẋσ −
∂xσ 2 ∂xρ
 
ν 1 ν σ ∂gρν ∂gρσ ∂gσν
= gρν ẍ + ẋ ẋ + −
2 ∂xσ ∂xν ∂xρ
= 0. (3.34)
The work is done. Multiplying by the inverse metric and relabeling indices we obtain the equation we
were looking for, the so-called geodesic equation
d2 xµ µ dxν dxρ 1 µσ
+ Γ νρ = 0, Γµ νρ = g (∂ρ gσν + ∂ν gσρ − ∂σ gνρ ) . (3.35)
dσ 2 dσ dσ 2

Exercise
• Consider a reparametrization σ → f (σ). Show that the geodesic equation (3.35) retains
its form only if f (σ) = aσ + b.
• Compute the geodesic equation associated to the Rindler metric (3.16).

The geodesic equation is automatically covariant since the Lagrangian from which it was derived was
invariant under general coordinate transformations. The transformation of the so-called Christoffel
symbols Γµ νλ is however non-homogeneous
0 0
0 ∂ x̄µ ∂xν ∂xρ ∂ x̄µ ∂ 2 xµ
Γ̄µ ν 0 ρ0 = Γµ νρ 0 0 + . (3.36)
µ ν
∂x ∂ x̄ ∂ x̄ ρ ∂xµ ∂ x̄ν 0 ∂ x̄ρ0
9 The only difference is the presence of a global factor m and a constant term m/2 which do not play any role in the

variation of the action

3.5 The geodesic equation 38

They are not a tensor.

Exercise
• Which is the form taken by Eq. (3.37) in a Cartesian coordinate system?

• Prove the transformation law (3.36) using the fact the the geodesic equation is covariant.
• How many independent components have the Christoffel symbols in four dimensions?

The Christoffel symbols encode the local aspects of the gravitational interaction as well as the
fictitious forces (centrifugal, Coriolis, etc)
d2 xµ dxν dxλ
Fµ ≡ 2
= −Γµ νλ (3.37)
dσ dσ dσ
arising when using non-inertial reference frames. This kind of forces can always be eliminated by
going to an inertial reference frame or to a free-falling frame. Note that this would not be the case if
the Christoffel symbols were tensors.

Local free-falling reference frames

The geodesic equation allows for a precise definition of local free-falling frames. According to
Equivalence Principle, in those frames the geodesic equation must become

d2 ξ α
= 0. (3.38)
dσ 2 P

A free-falling reference frame at P is therefore defined asa

gµν (P ) = ηµν , ∂σ gµν (P ) = 0 (3.39)

with the second condition being equivalent to the vanishing of the Christoffel symbols at that
point, i.e. Γµ νλ (P ) = 0.
a The existence of these frames is guaranteed by the so-called Local flatness theorem.

3.5.1 Massive particles don’t go on diet

It is easy to show that the geodesic equation (3.35) is equivalent to a conservation equation
d2 xµ dxν dxρ dxµ dxν
 
µ d
+ Γ νρ = 0 → gµν =0 (3.40)
dσ 2 dσ dσ dσ dσ dσ
associated to translations in the parameter σ, f (σ) = σ + c. The physical meaning of Eq. (3.40)
is quite obvious and should be expected; it simply states that massless/massive particles remain
massless/massive along the geodesic. In other words, given an initial condition
dxν dxλ

−1 for massive particles
gµν = , (3.41)
dσ dσ σ=0 0 for massless particles

it will by satisfied for all values of σ 10 . We will rediscover this equation in Chapter 5, when dealing
with the concept of parallel transport.

10 Note that for the massive case we can identify σ = τ

3.6 The Newtonian limit 39

Exercise
Prove the relation (3.40).

3.5.2 Conserved quantities

Note that if the metric coefficients are independent of one coordinate11 xν , the Lagrangian (3.27) will
be also independent of such a coordinate. In such a case, the covariant component ẋν is a conserved
quantity along affinity parametrized geodesics
∂L
pν = = gµν ẋµ = ẋν = constant . (3.42)
∂ ẋν
We will make use of this important property in Chapter 9, when dealing with the Schwarzschild
geometry.

3.6 The Newtonian limit

Let us see how the usual results of Newtonian gravity fit into the geometric picture. Of course, we
cannot expect to link the relativistic formulation presented above with a non-relativistic theory of
gravity without doing some assumptions. We will consider a massive particle moving at small velocity
(with respect to the speed of light)

dxi dxi dt
1 −→  (3.43)
dt dτ dτ
in a “weak”
gµν = ηµν + hµν , |hµν |  1 (3.44)
and stationary12 gravitational field
∂0 gµν = ∂0 hµν = 0 . (3.45)
The first two conditions (small velocities and weak fields) are quite natural from the point of view
of a non-relativistic description. On the other hand, the stationarity condition (3.45) is just a good
approximation for the particular cases we will be interested in in this section: the gravitational fields
of the Sun and the Earth.
At first order in the small perturbation hµν , the geodesic equation (3.35) takes the form
2
d2 xµ

dt
+ Γµ00 = 0, (3.46)
dτ 2 dτ

where the Christoffel symbols Γµ00 are completely determined by the perturbation hµν around the
Minkowsky metric13
 
µ 1 µρ ∂g0ρ ∂g0ρ ∂g00 1 ∂g00 1 ∂h00
Γ00 = g + + = − g µρ ρ = − η µρ . (3.47)
2 ∂x0 ∂x0 ∂xρ 2 ∂x 2 ∂xρ
11 The coordinate xν is then said to be cyclic.
12 Or varying sufficiently slow over the scale probed by the particle.
13 Note that, since we are interested only in first order terms, we can raise and lower indices with the Minkowski

metric ηµν . For example hµ ν = g µλ hλν ' η µλ hλν + O(h2µν ).

3.6 The Newtonian limit 40

Mass Size |Φ|/c2

Atome 10−26 Kg 10−10 m 10−43

Human 102 Kg 1m 10−25
Earth 1025 Kg 107 m 10−9
Sun 1030 Kg 109 m 10−6
Galaxy 1041 Kg 1021 m 10−7
White Dwarf 1030 Kg 107 m 10−4
Neutron Star 1030 Kg 104 m 0.1
Black Hole 1

Table 3.1: Gravitational self-energy: Orders of magnitude

Splitting the spatial and temporal components of Eq. (3.46) and using the stationarity condition
(3.45), we obtain14
d2 t d2 xi 1 ∂h00
= 0, = c2 , (3.48)
dτ 2 dτ 2 2 ∂xi
The first of these two equations allows us to identify the proper time τ with the coordinate time t and
write
d2 xi 1 ∂h00
= c2 . (3.49)
dt2 2 ∂xi
The value of the unknown function h00 can be determined by comparing Eq. (3.48) with the Newtonian
expression for a particle in a gravitational field

d2 xi ∂Φ
= −δ ij j . (3.50)
dt2 ∂x
The first true component of the gravitational metric tensor15 comes directly from Newton’s theory!
 
2Φ 2Φ
h00 = − 2 −→ g00 = − 1 + 2 . (3.51)
c c

Indeed. . . this is the first and the last component that we can expect to get from Newton . . . New-
tonian gravity involves just one scalar function: the gravitational potential Φ, nothing else. This
observation naturally raises the question of how to compute the remaining components of the metric.
Let’s forget about this problem for a while and enjoy our findings. As you will see, we can learn a
lot of new things without knowing the precise form of the other metric components. The correction
to the Minkowski metric is proportional to the so-called gravitational self-energy Φ/c2 . This quantity
can be understood as the ratio of the Newtonian potential energy to the relativistic energy. For an
object of mass M and typical size R we have

|Φ| GM 2 1 GM
= · = . (3.52)
c2 R M c2 Rc2
14 Note that we have restored the speed of light c for later convenience.
15 Note that we could in principle allow for an extra constant C in Eq. (3.51), i.e h00 = − 2Φ c2
+ C, which should
be fixed by requiring the metric to approach the flat Minkowski metric at infinity. For isolated mass distributions the
gravitational potential Φ vanishes at infinity and therefore C = 0.
3.7 The power of the equivalence principle 41

Figure 3.5: Gravitational redshift, Gedankenexperiment.

Some orders of magnitude for Φ can be found in Table 3.1. Note that, even for a white dwarf or a
galaxy, the gravitational self-energy is very small; the weak field approximation used in the derivation
of Eq. (3.51) is justified. The correction to the Minkowski metric is expected to be important only
for very compact object such as a neutron star or a black hole.

3.7 The power of the equivalence principle

Let us consider the direct consequences of the previous results. In order to get some intuition let me go
back for a moment to the constantly accelerated rocket and perform the following Gedankexperiment.
Imagine two observers in the rocket, one on the base and one on a platform close to the ceiling. The
observer at the bottom sends some light pulses with a frequency dictated by its proper time interval
ν1 = 1/∆τ1 . Due to the acceleration of the rocket, these pulses are received by the observer at the top
at a lower rate ν2 = 1/∆τ2 than the rate at which they were emitted16 . According to the Equivalence
Principle the same phenomenon should happen in the presence of gravity. Yes, gravity must affect
the flow of time!
Let us put our Gedankenexperiment into equations. Having a look to Eq. (3.51), we see that the
interval in proper time dτ at a fixed point in the vicinity of a massive object differs from the interval
in coordinate time dt r
p
µ ν
2Φ(r)
dτ = −gµν dx dx = 1 + dt . (3.53)
c2
A local measurement of this effect is nevertheless impossible since our measure instruments are affected
by gravity in the same way that the timing of the objects we want to measure. Observable effects on
the flow of time can only appear when we compare two different points in the gravitational potential
Φ, as we did in the rocket Gedankenexperiment.

3.7.1 Gravity and the flow of time

Consider two observers in the weak gravitational field of a spherically symmetric and stationary mass
distribution. Although the Newtonian limit developed in the Section 3.6 provides only information
16 By the time at which the light arrives to the top the ceiling is moving faster than when the light was emitted.
3.7 The power of the equivalence principle 42

about the g00 element, the large symmetry of the problem severely constrains the form of the metric
to be  
2Φ(r)
ds2 = − 1 + dt2 + grr (r)dr2 + r2 dθ2 + r2 sin2 θdφ2 , (3.54)
c2
with grr (r) an undetermined function of the radial distance, whose explicit form will not be needed in
what follows. In order to disentangle the effect of gravity from other velocity dependent Doppler-like
effects and to make the analysis as clear as possible, we will require the observers to be at rest in a
radial configuration with coordinates r1 and r2 . Imagine the observer at r1 sending pulses of light to
the observer at r2 . The period of emitted pulses is the interval in proper time of the emitter
Z p p Z p
∆τ1 = −g00 (r1 )dt = −g00 (r1 ) dt = −g00 (r1 )∆t1 . (3.55)

On the other hand, the period of received pulses is the interval in proper time of the receiver
Z p p Z p
∆τ2 = −g00 (r2 )dt = −g00 (r2 ) dt = −g00 (r2 )∆t2 . (3.56)

The coordinate interval elapsed between the emission of two pulses ∆t1 is equal to the coordinate
time interval elapsed between the reception on two pulses ∆t2 , as can be easily seen by noting that
the coordinate time interval needed to go from r1 to r2
Z r2  
2 2 2 −grr (r)
ds = −g00 dt + grr dr = 0 −→ ∆t = dr (3.57)
r1 g00 (r)

is independent of the coordinate time t. Taking the ratio of Eqs. (3.56) and (3.55), we get the first
prediction of the Equivalence Principle
s s
∆τ2 g00 (r2 ) 1 + 2Φ(r2 )/c2
= = . (3.58)
∆τ1 g00 (r1 ) 1 + 2Φ(r1 )/c2

For weak gravitational fields, the previous expression can be approximated by its binomial expansion17

∆τ2 Φ(r2 ) − Φ(r1 )

'1+ , (3.59)
∆τ1 c2
which is usually quoted in terms of the ratio

∆τ ∆τ2 − ∆τ1 Φ(r2 ) − Φ(r1 )

≡ = . (3.60)
τ ∆τ1 c2

Time goes by slower

Clocks slow down in those places where the gravitational potential is larger (in magnitude).
In particular, clocks at a distance rqfrom the surface of a massive spherical body of mass M
(Φ = − GMr ) slow down by a factor 1 − 2GM
rc2 with respect to clocks at r → ∞ (Φ = 0).

The dilation of time was tested by Hafele and Keating in 1972 using cesium-beam atomic clocks
transported on commercial flights around the Earth and compared on return to standard clocks in
the US Naval Observatory. The net effect on the reading of the on-flight clocks is a combination of
special relativistic effects and gravitational changes in the flow of time. The two contributions act in
17 (1 + x)1/2 = 1 + 12 x.
3.7 The power of the equivalence principle 43

Time of flight 41.2 h 48.6 h

∆τ (ns) Eastward Westward

∆τG 144 ± 14 179 ± 18
∆τSR −184 ± 18 96 ± 10
∆τtot −40 ± 23 275 ± 21
∆τobs −59 ± 10 273 ± 7

Table 3.2: Hafele Keating: Predictions and experimental results

Figure 3.6: The highs and lows: Redka and Pound at the top and bottom of the tower.

an opposite way. Special Relativity tends to decrease the rate of the clock in the plane with respect
to the standard clock in the surface of the Earth18 . On the other hand, gravity tends to speed up
the clock in the plane with respect to the clock in the stronger gravitational field of the Earth. The
experiment was performed twice, once flying towards the east and once flying towards the west. The
results and their comparison with the predictions are summarized in Table 3.2. As you can see, the
agreement between the theory and the theoretical prediction is notably good.

Exercise
How older are the theorists of the upper floor of the Cubotron with respect to the experimental-
ists in the lower floor at the end of their academic life? Should this effect be taken into account
by the Swiss pension system?
18 This is just a consequence of the well-known time dilation effect in Special Relativity.
3.7 The power of the equivalence principle 44

3.7.2 Gravitational shift of frequencies

An immediate consequence of the previous result is the gravitational frequency shift. Denoting by ν1
the frequency of the light emitted at r1 and by ν2 the frequency of the light received at r2 , Eq. (3.58)
can be rewritten as s s
ν1 g00 (r2 ) 1 + 2Φ(r2 )/c2
≡1+z = = , (3.61)
ν2 g00 (r1 ) 1 + 2Φ(r1 )/c2
where we have defined the so-called redshift parameter
∆ν ν1 − ν2
z= ≡ . (3.62)
ν ν2
It z > 0 the received light is said to be redshifted, while if z < 0 the light is said to be blueshifted. For
weak gravitational potentials, Eq. (3.61) can be approximated by
∆ν Φ(r2 ) − Φ(r1 )
z= ' . (3.63)
ν c2
To get an estimate of the order of magnitude of this frequency shift, consider for instance the light
from the Sun (r = r1 ) received on Earth (r = r2 )19 . Since r2 > r1 , |Φ(r1 )| > |Φ(r2 )| and therefore20
z > 0. As in our Gedankenexperiment, the light redshifts (ν1 > ν2 ) as it climbs upwards in the
gravitational potential!

The gravitational frequency shift is a test of the Equivalence Principle, not of the Einstein’s
theory of gravity in its full form. Note that the spatial part of the metric grr (r) did not played
any role in the previous developments.

Numerically, the gravitational redshift of the light emitted by the Sun is very small
∆ν
= 2.12 × 10−6 , (3.64)
ν
and indeed very difficult to detect due to the broadening of spectral lines and to Doppler shifts
associated to the convection currents in the solar atmosphere21 .
A more precise non-astronomical test of the gravitational frequency shift was performed by Pound
and Redka in 1960 using gamma rays produced in a 14.4 keV atomic transition in 57 Fe. These gamma
rays were emitted at the top of a tower of 22.6 meters in the Jefferson Physical Laboratory at Harvard
university and directed down towards a similar sample of 57 Fe located at the bottom of the tower. The
absorption of the gamma rays by the receiver is only efficient if the frequency at reception coincides
with the frequency at emission (Mössbauer effect). Due to the gravitational shift of frequencies this
was not the case. Pound and Rebka compensated the gravitational shift in a very clever way: a
Doppler shift induced by the vertical motion of the source at the top of the tower. By looking for
a resonance in the absorption they were able to obtain a direct measurement of the gravitational
redshift. The result was in excellent agreement with the Equivalence Principle’s prediction
(∆ν/ν)exp
= 1.05 ± 0.10 . (3.65)
(∆ν/ν)th
19 Φ(r ) = −GM /r and Φ(r ) = −GM /r are small (cf. Table 3.1). The binomial expansion is justified. The
1 1 2 2
gravitational field of the Earth is neglected.
20 Remember that the gravitational potential is negative.
21 The gravitational redshift (3.64) corresponds numerically to the Doppler shift associated to a velocity of 0.6 Km/h,

which is easily exceed by the hot gases in the surface of the Sun.
3.8 The weakness of the Equivalence Principle 45

Figure 3.7: Different tests of the gravitational redshift. The parameter α parametrizes the deviations
from the Equivalence Principle, ∆ν/ν = (1 + α)∆Φ/c2 .

g
g

Figure 3.8: Deflection of light, Gedankenexperiment.

Exercise
Determine the predicted value ∆ν/ν in the Pound-Rebka experiment. Are the gamma rays
traveling down the tower blueshifted or redshifted?

3.8 The weakness of the Equivalence Principle

A second (and incomplete) prediction of the Equivalence Principle is the deflection of light. This effect
can be easily understood with another Gedankenexperiment. Imagine an observer in the accelerated
rocket. A pulse of light is emitted by some device from one of the walls in the transverse direction to
the rocket motion. Due to the acceleration of the rocket, the pulse will hit the opposite wall at a height
below that of the emission. Since the uniform acceleration of the rocket is locally indistinguishable
from a gravitational field, we should expect the same deflection of light in a gravitational field.
3.8 The weakness of the Equivalence Principle 46

The bending of light in a gravitational field was considered by Newton himself, but he didn’t
performed any proper computation. The first known result about the deflection of light was presented
by the German astronomer Johann Georg von Soldner in 1804. Based on Newton’s corpuscular theory
of light, Soldner predicted a deflection angle of 0.8700 for a ray of light grazing the surface of the Sun.
Einstein, unaware of Soldner’s computations and based on the Equivalence Principle, obtained the
same number one hundred years later, in 1911. Let us reproduce his arguments and (wrong) results.

3.8.1 Einstein’s 1911 (wrong) treatment

In Einstein’s 1911 paper, the speed of light is considered as a scalar quantity which depends on the
value of the gravitational field

ds2 = −(1 + 2Φ)dt2 + dX 2 = 0 → c = c0 (1 + Φ) (3.66)

The Minkowski value c = c0 is only recovered at long distances (r → ∞), where the gravitational
potential is negligible22 . According to Huygens’ principle the position of a wavefront at a time t + ∆t
can be determined by considering each point of the wavefront at t as a source of spherical waves. The
wavefront at t + ∆t is then given by the envelope of the multiple spherical wavefronts originated at
t. Imagine a wave front in the vicinity of a matter distribution M . Consider two points P1 and P2
separated by a spatial distance δl at time t. The velocity of light at those points (c1 and c2 ) depends
of the value of the gravitational field. Having a look to Fig. ??, we conclude that in a time δt the
wavefront in refracted by an angle
(c1 − c2 ) δt δΦ
δα = = δt , (3.67)
δl δl
with δΦ/δl the component of the gravitational acceleration along the wavefront. This infinitesimal
refraction angle can be integrated along the full path to obtain the total deflection angle23
dΦ
Z Z
α = dα = dt . (3.68)
dl
Since the velocity of light along the path is nearly constant we can set dt = ds, with s measuring the
distance along the path. Evaluating the integral (3.68) for an impact parameter b, we get
Z π/2
dΦ GM 2GM
Z
α= ds = 2
cos θds = , (3.69)
dl −π/2 r b

which for the particular case of a photon grazing the surface of the Sun becomes24
2GM
α= ≈ 0.87500 . (3.70)
c2 R
22 In ”Relativity,The Special and General Theory”, Einstein wrote:
[. . . ] our result shows that, according to the general theory of relativity, the law of the
constancy of the velocity of light in vacuo, which constitutes one of the two fundamental
assumptions in the special theory of relativity and to which we have already frequently referred,
cannot claim any unlimited validity. A curvature of rays of light can only take place when the
velocity of propagation of light varies with position. Now we might think that as a consequence.
of this, the special theory of relativity and with it the whole theory of relativity would be laid
in the dust. But in reality this is not the case. We can only conclude that the special theory
of relativity cannot claim an unlimited domain of validity; its result hold only so long as we
are able to disregard the influences of gravitational fields on the phenomena (e.g. of light).
23 This is a good approximation for small deflections angles, as is the case of the deflection of light by the Sun.
24 Note that we have restored the powers of c.
3.8 The weakness of the Equivalence Principle 47

P2

P1

Figure 3.9: Deflection of light, Huygens’s principle.

As Einstein stated in the original paper, since ‘the fixed stars in the part of the sky near the sun
are visible during a total eclipse of the sun, this consequence of the theory may be compared to
experiment”. He indeed “urgently wishers astronomers to take up this question” and measure the
deflection of light during a solar eclipse. Fortunately for him . . . they didn’t do it on time. Einstein’s
1911 prediction based only in the equivalence principle was incomplete25 . No measurement of the
deflection angle was performed between 1911 and 1915, the moment at which he straightens out his
result to
2GM
α=2× 2 ≈ 2 × 0.87500 . (3.71)
c R
Although different expeditions to observe solar eclipses were organized, all of them were cancelled,
either for climatological or political reasons. One of the most interesting stories is that of the Ger-
man astronomer and mathematician Erwin Finlay Freundlich, which, interested on testing Einstein’s
prediction, convinced the german armament manufacturer Krupp to finance a trip to Crimea on
21st August 1914. Unfortunately for him, and fortunately for Einstein, the German astronomer was
arrested by the Russians as a suspected spy before being able to perform any measurement.

25 We will see why later on.