You are on page 1of 81

Notes of General Relativity

Gaspari Andrea
January 30, 2021

Abstract
Just some notes of General Relativity taken during the course of Roberto Balbinot in 20-21
and later reviewed.

Contents
1 Introduction (22-09-2020) 3

2 Special Relativity (23-09-2020) 4


2.1 Explicit form for parallel space axes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.2 Matrix form . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.3 The light cones . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.3.1 Accelerated motion in SR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

3 Tensor formalism (29-09-2020) 9


3.1 Tensor Algebra . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

4 Electromagnetism (6-10-2020) 11
4.1 Covariance in Maxwell’s equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
4.2 Tensor formalism in Electromagnetism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
4.3 Solution to Maxwell’s equations in tensor formalism . . . . . . . . . . . . . . . . . . . . . 15
4.4 Scalars of the electromagnetic field (07-10-2020) . . . . . . . . . . . . . . . . . . . . . . . . . 16
4.5 Electromagnetic waves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
4.6 Helicity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
4.7 Solution in presence of sources (10-12-2020) . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
4.8 Systems of charged particles (12-10-2020) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
4.9 Energy-momentum tensor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

5 Perfect fluids (13-10-2020) 24


5.1 Equations of motion of fluids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

6 General Relativity (19-10-2020) 26


6.1 Equivalence principle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
6.2 Gravitational forces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
6.2.1 Generalization for massless particles . . . . . . . . . . . . . . . . . . . . . . . . . . 29
6.3 Discussions over the metric tensor (20-10-2020) . . . . . . . . . . . . . . . . . . . . . . . . . 30
6.3.1 Derivation of the Affine connection . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
6.4 Construction of a local inertial frame . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

1
7 Curved space-time (21-10-2020) 32
7.1 Brief some-up . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
7.2 Principle of General Relativity: introduction of tensor elements . . . . . . . . . . . . . . . 33
7.2.1 Tensor densities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34

8 Tensor analysis (27-10-2020) 35


8.1 Differentiation of tensors: the covariant derivative . . . . . . . . . . . . . . . . . . . . . . 35
8.1.1 Covariant derivative of the metric tensor . . . . . . . . . . . . . . . . . . . . . . . . 37
8.1.2 Particular cases of covariant derivatives . . . . . . . . . . . . . . . . . . . . . . . . 37
8.2 Covariant differentiation along a curve (28-10-2020) . . . . . . . . . . . . . . . . . . . . . . . 38

9 Interaction with gravity 40


9.1 Principle of General Covariance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
9.2 Point particle in a gravitational field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
9.3 Electromagnetic charges in a gravitational field . . . . . . . . . . . . . . . . . . . . . . . . 40
9.4 Energy-momentum tensor in presence of a gravitational field . . . . . . . . . . . . . . . . 42
9.5 Newtonian limit of geodesic equations (03-11-2020) . . . . . . . . . . . . . . . . . . . . . . . 42
9.6 Perfect fluids coupled with gravitational field . . . . . . . . . . . . . . . . . . . . . . . . . 44

10 Curvature of space-time (04-11-2020) 46


10.1 The Riemann Tensor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
10.2 Geodesic deviation & relative acceleration . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
10.3 Algebraic properties of the Riemann tensor (10-11-2020) . . . . . . . . . . . . . . . . . . . . 50
10.3.1 The Bianchi identities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51

11 Einstein Field Equations 51


11.1 Algebraic analysis of the Einstein’s equation (11-11-2020) . . . . . . . . . . . . . . . . . . . . 54
11.2 The Cosmological constant . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55

12 The Schwarzschild problem 56


12.1 Solution for stationary space-time (17-11-2020) . . . . . . . . . . . . . . . . . . . . . . . . . 58
12.2 Solution for non-stationary space-time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
12.3 Space and time coordinates in Schwarzschild solution . . . . . . . . . . . . . . . . . . . . . 60
12.4 Geodesic equations in Schwarzschild space-time . . . . . . . . . . . . . . . . . . . . . . . . 61

13 Discussions over Einstein’s tests for GR (18-11-2020) 62


13.1 Planets’ orbits in Schwarzschild space-time . . . . . . . . . . . . . . . . . . . . . . . . . . 62
13.2 Light rays’ deflection in Schwarzschild space-time . . . . . . . . . . . . . . . . . . . . . . . 65
13.3 Gravitational redshift (24-11-2020) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
13.3.1 Gravitational redshift from Special Relativity . . . . . . . . . . . . . . . . . . . . . 69

14 Gravitational waves (25-11-2020) 70


14.1 Helicity for Gravitational waves (01-12-2020) . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
14.2 The TT gauge . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
14.3 Effects of Gravitational waves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75

15 Gravitational radiation (02-12-2020) 77


15.1 Solution of Einstein’s equation in presence of sources . . . . . . . . . . . . . . . . . . . . . 77
15.2 Multiple expansion of a field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80

2
1 Introduction (22-09-2020)

Special Relativity (1905 by Einstein) Framework for physical system only governed by E.M., weak, strong
interactions but with no gravity. (For all non gravitational interactions) Observers associated to inertial
frames in special relativity (if we move to accelerated frames we are no longer inside Special Relativity
Theory ) Special relativity is based on two postulates:
- equivalence of physics in all inertial frames
- speed of light c is constant
(Lorentz transformations and covariant formalism(tensors), we will describe Electromagnetic Theory and
Maxwell equations in covariant form). Gauge invariance and solutions of Maxwell equations (starting
point of spacial theories) Solutions of Maxwell equations are of two kinds:
- homogeneous → Electromagnetic waves (no sources of field) (role of gauge invariance)
- inhomogeneous → More complicated solutions due to the presence of sources (charges and corrents),
introduction of retarded potentials and use of Green functions
Understand SR is a fundamental step to able to do some computation in GR Theoretical need of introduc-
ing gravity in SR (but not experimental) Newton theory → Celestial Mechanics Anomalous precession of
the planet Mercury, Newton Theory was almost able to describe all the influences of other planets but the
computation was not perfect (perturbation of a further planet?) (no experimental evidence that Newton
Theory was wrong, the need was just theoretical) Newton Theory is based on Poisson’s Equation:

∇2 φ = 4πGρ (1)

this equation represent an action at distance, in fact, if we change the mass density ρ, then immedi-
ately all the potential change is form, which is clearly incompatible with special relativity People
attempt to modify Poisson equation in order to make it compatible with Special Relativity principles.
Nevertheless, Einstein noticed that in presence of a gravitational fields there are no inertial systems,
hence SR was not enough to describe also Gravity → Need or General Relativity.
Construction of General Relativity
The first step in the construction of GR was the Einstein Equivalence Principle (1907) which states
that Gravitational Forces are equivalent to Inertial Forces and so their description as to be given in the
same way. (Inertial forces are just related to the transformation of the systems of reference). Then, the
final form of GR came out in 1915 with the Einstein Field equation. In this time Einstein developed
the revolution idea that space-time is no longer a passive arena but it is dynamical and it changes with
the physical system. It changes according to Einstein’s equation: Gravitational Field → Curvature of
space-time → Differential Geometry Now to do physics we have to be able to solve differential equation
in a curved manifold (Differential geometry framework to construct this theory) Another fundamen-
tal aspect was the introduction of 10 Gravitational potentials instead of 1 as theorized by Newton (In
Maxwell Electromagnetism we have just four potentials) and unlike Newton and Maxwell theories these
potentials are non linear in the first derivative of the potential (but linear in the second). The source
of gravitational fields are energy and momentum and in some way those also gravitates: self-interacting
potentials. Electromagnetic field has no charges, reason why it does not self interact, while on the other
hand, gravitational field has, the field carries energy and momentum which are also the sources of the
field and so we have self-interaction (self-interaction which is enclosed mathematically in the non linearity
of the equations). Even though this makes equations much more difficult, mostly in the researches for
symmetries, by imposing spherical symmetry Schwarzschild has been able in 1916 to find an exact solu-
tion to Einstein’s equations for a gravitational field outside a spherical body. In 1963, this solution found
a generalization in the case the body is rotating (axial symmetry). The Kerr solution, which is basically
the solution for a gravitational field generated by a black hole, a Kerr hole (?), while it gives a asymptotic
solution in the case the source is a star. In GR if we have no symmetry we cannot have an exact solution
and we must give up with just numerical solution; the two-body problem itself has no solution, we have to
solve it numerically (solvable in Newton framework). Three proofs from Einstein: Gravitational Redshift,
Mercury precession and also Deflection of light rays → causal structure of space-time determined by
gravity, which determines if two events can be correlated or not)
In the last part we will also talk of Gravitational Waves: how this waves come out of field equation,
their gauge invariance and their degrees of freedom in a gravitation way, the phenomenum of polarization

3
(linked to helicity), and the fact that these have a quadrupole nature. Without a quadrupole moment
we cannot generate a gravitational wave. General Relativity is an incredibly precise theory to describe
the gravity we have and has many experimental proofs, nevertheless it is probably not the ultimate grav-
itation theory: its quantization does not lead to a consistent quantum theory, it gives a non normalizable
theory, many infinity comes from it and we cannot gives sense to them (Loop quantum theory, GR as
low energy configuration of String Theory...?) At galaxy and cosmological scale we have two problems: a
mismatch between the predicted behaviour of galaxies and the ones actually observed, and the fact that
GR also does not explain the expansion of the universe in the way we are attending it. If a differential
equation does not fit exactly the observations we can either modify the left-hand side, which usually cor-
responds to the operator (this way we are actually modifying the theory), or the right-hand side, hence
the sources of the field. In this sense we can save GR in the classical limit by introducing dark energy
and dark matter, however, this would imply accepting the fact that the largest part of the universe is
unknown to us.
Bibliography General Relativity by Weinberg (arrives very rapidly to Einstein Field equation, )

2 Special Relativity (23-09-2020)

Special relativity is fundamental to understand GR, if we want to understand how a system evolve in
presence of gravity, first we must know how it behaves in its assence and that is described by SR Main
postulates:
• Equivalence of inertial frames of reference

• Speed of light is constant


SR is a framework which describes physics of observers associated to inertial frames, physic described
by observers not in inertial frames is not described within the work of SR. The equivalence of inertial
frames means that the law of physics should be invariant in form (covariant) under the transformations
that map one inertial frame in one other. According to SR these transformations are the Lorentz
Transformation (Galilean Transformation does not fit the SR postulates):

z z0

x0
~v
o0
o x
y y0

xα → x0α = Λαβ xβ (2)


with: x ≡ ct, x ≡ x, x ≡ y and x ≡ z
0 1 2 3

which can be represented by a matrix with the elements:

Λ00 = γ
Λ0i = Λi0 = −βi γ
(γ − 1)
Λij = δ ij + βi βj
β2
~v 1
β = and γ = p
c 1 − β2

plus a rotation component R ∈ O(3) in the case the inertial frame we chose are not parallel. The fact
that the speed of light is constant has the immediate consequence that time is no longer absolute,
events can be simultaneous in one FoR but not in another one, moreover, due to its definition, β also
implies that the the speed of light is the maximum achievable velocity (only massless particles

4
can reach that speed). Lorentz transformations form a group under which Maxwell equations are left
invariant (Symmetry group). A Lorentz transformation L leaves invariant the quadratic form ds2 :
ds2 = c2 dt2 − dx2 − dy 2 − dz 2 = ηαβ dxα dxβ Not positive definite (3)

η00 = 1

ηαβ = 0 if α 6= β
ηii = −1


L
→ ds02 = c2 dt02 − dx02 − dy 02 − dz 02 = ηαβ dx0α dx0β ≡ ds2
ds2 − (4)
The invariance of this quadratic form is a direct consequence of the principles of SR. Lorentz transforma-
tions L are linear transformations (allows to see for both O and O0 a motion with uniform velocity)
that satisfy the relation:
ηαβ Λαµ Λβγ = ηµγ (5)
Using this relation we can verify immediately the equation (4):
ds02 = ηαβ dx0α dx0β = ηαβ Λαµ dxµ Λβγ dxγ
(6)
= ηαβ Λαµ Λβγ dxµ dxγ = ηµγ dxµ dxγ = ds2
Lorentz transformations are the only non-singular coordinate transformations (x → x0 ) such that ds2 is
invariant.

2.1 Explicit form for parallel space axes


(
~x0 = ~x + (γ − 1)( ~vv·~2x ) − γ~v t
(7)
t0 = γ (t − ~vc·~
2 )
x

(In the case we also have axes rotation we should substitute ~x = R~x where R is a 3x3 rotation matrix)
If ~v is parallel to one specific axis (for example ~v k x) we talk about Lorentz Boost along an axis, in
this case we have:

z z0

~v

o x o0 x0
0
y y

= γ(x − vt)
 0

 x
y 0 =y

(8)


 z0 =z
= γ(t − cv2 x)
0
t
And from this equations we can clearly see that in SR space and time are mixed → introduction of the
concept of space-time.

2.2 Matrix form


Recalling the relation (5) and the definitions of Λ and η, we can rewrite everything in terms of matrices:
Λ 0 Λ01 Λ02 Λ03
 0 
Λ10 Λ11 Λ12 Λ13 
Λ ≡ Λαβ =  Λ20 Λ21 Λ22 Λ23 
 (9)
Λ30 Λ31 Λ32 Λ33
1 0 0 0
 
0 −1 0 0
η ≡ ηαβ =  0 0 −1 0 
 (10)
0 0 0 −1

5
and, due to the fact that Λαβ = ΛTβ α , we have:

ηαβ Λαµ Λβγ = ηµγ


(ΛT )µα ηαβ Λβγ = ηµγ (11)
ΛT ηΛ = η.
So, we can think at a Lorentz transformation as a linear transformation on space-time coordinates gen-
erated by a 4x4 matrix that satisfies the relation (11). At a first look then, one may think Λ has to
satisfy 16 constraints but actually, due to the fact that η is symmetric we only have 10 equations (4 from
diagonal terms and only 6 from off-diagonal), so 6 parameters are left free:
• 3 related to boost velocity ~v
• 3 related to rotation angles (we can think of Euler angles)
An interesting observation is that if η was 1, then this relation would have just been the orthogonality
relation for matrices OT O = 1, this suggests that Lorentz transformations are not simple rotations in a
4-dimensional space. Subsequently, if we analyze the determinant we can find that:
det(Λt ηΛ) = det η
det(ΛT ) det(η) det(Λ) = det(η) but det(ΛT ) = det(Λ) (12)
det(Λ) = 1 =⇒ det(Λ) = ±1
2

Usually we refer to Lorentz transformation Λ as:


• Proper transformation if det(Λ) = 1
• Improper transformation if det(Λ) = −1
Further, by considering the case for µ = 0 and γ = 0, we find that:
ηαβ Λα0 Λβ0 = η00 ≡ 1
Then we have to sum over the free indices α and β, but we know that ηαβ 6= 0 only if α = β, thus we get:

η00 (Λ00 )2 + ηii (Λi0 )2 = 1


3
X
(Λ00 )2 = 1 + (Λi0 )2 ≥ 1
i=1
=⇒ Λ00 ≥ 1 (orthochronous) or Λ00 ≤ −1

In this course we will consider mainly Lorentz transformations with det(Λ) = 1 and Λ00 ≥ 1, the so
called proper orthochronous Lorentz transformations, which are those that we can get out from a
sequence of infinitesimal transformations starting from the identity times infinitesimal transformations.
Observation. Assume we are in R3 , in this space the Euclidean distance:
dl2 = dx2 + dy 2 + dz 2
is invariant under rotations R, and is positive definite dl2 ≥ 0.
(x, y, z) ←→ (x + dx, y + dy, z + dz)
with the two points coinciding if dl2 = 0.
Assume a xy plane and a rotation matrix R = cos θ − sin θ
. The vector ~x = (1, 0) has a module |~x|2 =

sin θ cos θ
x + y = 1 and its transform:
2 2

~x0 = R~x = (cos θ, sin θ)


also has a module |~x0 |2 = cos2 (θ) + sin2 (θ) = 1, nevertheless the two points do not coincide, in fact:
dl2 = (1 − cos θ)2 + (0 − sin θ)2
= 1 − 2 cos θ + cos2 (theta) + sin2 (θ) =
= 2 − 2 cos θ
which is not 0 unless θ = 2nπ, in this case we have full rotations in the plane.

6
In SR we can introduce the notion of events:

xα = (ct, ~x) (13)

which can be thought as points in a 4-dimensional manifolds, the Minkowski space-time M, where ct
indicates the time a specific event has occurred and ~x its position. In this framework then

ds2 = ηαβ dxα dxβ = c2 dt2 − d~x2 (14)

takes the role of "distance" in M invariant under Lorentz transformations. This is a not euclidean
distance cause is not positive definite, it can be ≶ 0 and even 0 for distinct points. Using this distance
we can somehow classify two different events:
- if ds2 > 0: events time-like separated, we can for example imagine the motion of a massive body,
ds2 > 0 → c2 dt2 > d~x2 = ~v 2 dt2 and so we have ~v < c
- if ds2 = 0: events light-like separated, events associated to light propagation, c2 dt2 = d~x2
- if ds2 < 0: there cannot be physical motion associated to this events cause it requires something to
move faster than light, events space-like separated.

2.3 The light cones


The fact we have a maximum limit in physical velocities for motion/propagation allows to introduce a
causal structure in the Minkowski space which can be represented by the light cone:

ct
x = ct
Absolute future
p c
r
~x
O

p0
Absolute past

Figure 1: Light cone

In particular the two red lines (x = ct and x = −ct) represent the propagation of a light signal and also
delimit 4 regions. Those on top and bottom are the regions containing the time-like events (ds2po > 0),
on the light cone we have the null-separated events (ds2co = 0) and finally, in left and right regions we
have the space-like events which consist in all those events that have no causal connection with the origin
(ds2ro < 0). If we now consider two different events A = (ct1 , x1 , y1 , z1 ) and B = (ct2 , x2 , y2 , z2 ) in the
FoR K we have a distance:

ds2AB = c2 (t1 − t2 )2 − [(x1 − x2 )2 + (y1 − y2 )2 + (z1 − z2 )2 ] = c2 dt2 − dl2

if we then move to another FoR K 0 we can have, for example, A = (ct01 , x01 , y10 , z10 ) and B = (ct02 , x02 , y20 , z20 )
characterized by a Minkowski distance:

ds0AB
2
= c2 (t01 − t02 )2 − [(x01 − x02 )2 + (y10 − y20 )2 + (z10 − z20 )2 ] = c2 dt0 2 − dl0 2

but we know that ds2 is invariant, so we have:

ds2AB = ds0AB
2

c2 dt2 − dl2 = c2 dt0 2 − dl0 2

Now we notice that, if A and B are time-like separated (ds2AB > 0), it cannot exist a FoR K 0 such that
dt0 = 0 (in which A and B are simultaneous) −→ The events A and B must be separated in time

7
in every reference frame. For this reason we can name the upper region as the Absolute future and
the lower region Absolute past (p will always be in the future of O and p0 always in the past). Another
important aspect to discuss is that, given an observer in space-time, it can only have information at t = 0
of the events inside the past light cone, all the events outside hypothetically require super-luminar speed
to reach the observer. Only the events in the past cone can affect the observer.

ct ct
x = ct x = ct

t̄ O
~x ~x
O
r r

(a) No information regarding the event r (b) Acknowledgment of the event r

Figure 2: If an event is out the past cone, the observer must wait to be further in time to get information
about it.

In Minkowski space-time we have just to wait and information will eventually reach us.

2.3.1 Accelerated motion in SR


According to Newton’s law, an accelerated FoR can be described by:
1 2
x= at + v0 t
2 (15)
v = at + v0 −−−→ ∞
t→∞

and of course, it can surpass the speed of light, but in SR an accelerated system is described by different
equations:
c4
x2 − c2 t2 =
a2
at (16)
v =q −−−→ c
t→∞
1+ a2 t2
c2

in particular the first equation describes a parabola in the Minkoski space which is asymptotic with the
line x = ct, this means that a non inertial frame would never know information about certain
region of space (the region on the other side of the asymptote).

ct
x = ct

O
~x

Figure 3: Line x = ct is like a horizon of events for the non-inertial observer

8
SR does not discuss the physic described by an accelerated observer (SR is only intended for inertial
Observer) but of course we can describe accelerated motion as seen by an inertial observer. (We want to
describe physic with the same laws, Maxwell equations are equal only in inertial frames).

3 Tensor formalism (29-09-2020)

In SR we study transformations in the form:

xα → x0α = Λαβ xβ + aα Poincarè Transformation


(17)
with: Λ ηΛ = η and the invariant Minkowski distance ds2
T

In the case of Poincarè transformations we have in total 10 free parameters (6 from Lorentz transforma-
tions and 4 from space-time translations), but in particular we will focus on Lorentz transformations with
determinant 1 and the element Λ00 ≥ 1: Proper orthochronous Lorentz transformations. What we
want is a framework in which, given a physics law, this is invariant in form in different FoR (under
Lorentz transformations, principle of equivalence). Covariant physics laws. That’s what Lorentz did
with Maxwell’s equations, but go through such analysis is very difficult using vectorial formalism, that’s
why we prefer tensor formalism.
Example. In the case of rotations in R3 we have matrices R ∈ O(3) such that RRT = 1 and R−1 = RT
(3x3 orthogonal matrices) which transform vectors according to the rule:

xi −→ x0i = Rij xj
∂x0i
Rij =
∂xj
and subsequently:

~v → ~v q0 = R~v (vectorial form)


v →v =
i 0i
Rij v j (tensorial form)

then we can define scalars as:


Φ(x) −−−−→ Φ0 (x0 ) = Φ(x)
x→Rx

objects invariant under rotations, so we have that:



∇Φ = Φ(x) = ∂i Φ(x) −−−→ ∂i0 Φ0 (x0 )
∂xi R
∂ ∂Φ(x) ∂Φ(x) ∂xj
with ∂i0 Φ0 (x0 ) = Φ (x
0 0
) = =
∂x0i ∂x0i ∂xj ∂x0i
and now, we can notice that since:

x0i = Rij xj −→ xj = (R−1 )ji x0i then


j
∂x
= (R−1 )ji ≡ (RT )ji
∂x0i
so we can write the equation:
∂Φ(x)
∂i0 Φ0 (x0 ) = (R−1 )ji
∂xj
in the form:
∂Φ(x) ∂Φ(x)
∂i0 Φ0 (x0 ) = (RT )ji = [(RT )T ]i j = Ri j ∂j Φ(x)
∂xj ∂xj
hence, we can conclude that the transformation of the gradient is proportional to the inverse matrix, but,
for rotations (and in general for orthogonal matrices), it can be written as a function of the original
matrix R.
Let’s now consider the case of Lorentz transformation L in the Minkowski space M:

xα −−→ x0α = Λαβ xβ with: ΛT ηΛ = η (18)


L

In this framework we then define:

9
- scalars or (0, 0) tensors (no indices), as quantities that do not change under the action of L:
L
Φ(x) −−−−→ Φ0 (x0 ) = Φ(x) definition similar for R ∈ O(3)
x→Λx

and, in particular, we have already seen the scalar ds2 (14). (In particle physics scalar fields are
typically used to describe spin-0 particles)
- controvariant four vectors or a (1, 0) tensors (index up), which transform exactly like the 4-
coordinates:

Aα = (A0 , A1 , A2 , A3 ) = (A0 , A)
~
L (19)
Aµ (x) −−−−→ A0µ (x0 ) = Λµν Aν (x)
x→Λx

A very common example of a 4-vector is the 4-momentum


E
pµ = ( , p~ ) with E = γmc2 and p~ = γm~v
c
and in particular, if pµ is conserved in one frame, then it must be conserved in any frame (Con-
servation laws of energy and momentum).
- covariant four vectors or a (0, 1) tensors (index down), which under L transform according to
the rule:
L
Vµ (x) −−−−→ Vµ0 (x0 ) = Λµν Vν with Λµν = [(Λ−1 )T ]µν
x→Λx (20)
in fact: Λµα Λµβ = δαβ

and an example of a 4-covariant vector is the 4-gradient of Φ:

L ∂Φ0 (x0) ∂Φ(x) ∂xν


∂µ Φ(x) −−−−→ ∂µ0 Φ0 (x0 ) = = = (Λ−1 )νµ ∂ν Φ(x) =
x→Λx ∂x0µ ∂xν ∂x0µ
= [(Λ−1 )T ]µν ∂ν Φ(x) = Λµν ∂ν Φ(x)

To be clearer, their transformation law uses the inverse matrix, while for controvariant 4-vectors
we have the original matrix. (In particle physics 4-vectors fields are used for spin-1 particles)
- (µ, ν) tensors which are a generalization of tensors with µ upper indices and ν lower, so µ con-
trovariant and ν covariant components. A basic example can be the (2, 1) tensor T αβγ :

L
T αβγ −−−−→ T 0 αβγ = Λαµ Λβν Λγλ T µνλ
x→Λx

In SR the difference between covariant and controvariant vectors is minimal, in GR no. Further, tensors
can also be:
- symmetric if: Tµν = Tνµ
- or antisymmetric: Tµν = −Tνµ

Starting from this entities we can define a scalar product:

given: Aα = (A0 , A1 , A2 , A3 ) and Vα = (V0 , V1 , V2 , V3 )


(21)
Aα Vα = A0 V0 + A1 V1 + A2 V2 + A3 V3

which is a scalar, in fact if we consider:

A0α V 0α = Λαµ Aµ Λαν Vν = Λαµ Λαν Aµ Vν = δµν Aµ Vν = Aµ Vµ ≡ Aα Vα

This scalar product has no definition (it can be positive, negative or even zero) and if two vectors V α , Wα
are such that V α Wα = 0, we say these are orthogonal (not in the Euclidean sense clearly).

10
3.1 Tensor Algebra
Basic operations in tensor algebra:
a) Linear combinations :
(µ, ν) ⊕ (µ, ν) = (µ, ν) like T µν = αAµν + βB µν

b) Direct products :
(µ, ν) ⊗ (µ0 , ν 0 ) = (µ + µ0 , ν + ν 0 ) like T µνλ = Aµ Bνλ

c) Contraction :
T µµνλ = Rνλ reduction from (1, 3) → (0, 2) µ was not a free index

d) Differentiation :
∂T βγ (x)
T βγ (x) −→ is this a tensor?
∂xα
∂T βγ (x) ∂T 0 βγ (x0 ) ∂ ∂ ∂T µν (x) ∂xρ
−−→ 0
= 0
(Λβµ Λγν T µν (x)) = Λβµ Λγν 0 α T µν (x) = Λβµ Λγν
∂xα L ∂x α ∂x α ∂x ∂xρ ∂x0 α
∂T (x) ∂x
µν ρ
∂T (x)
µν
and as we have already seen: Λβµ Λγν ρ 0 α
= Λβµ Λγν Λαρ
∂x ∂x ∂xρ
So the answer is yes, we find a tensor with an additional lower index:
(µ, ν) → (µ, ν + 1)
however, this whole discussion is possible only because L is a linear transformation in SR and so Λ
is not a function of space-time coordinates (For non-linear transformation, as we will encounter in
GR, this won’t be true anymore).
Observation. Note how ~v · ~v is a scalar under rotations but not under Lorentz transformations.
As we have already seen, the main scalar for Lorentz transformations is ds2 (14), and now we can compute
that:
ds2 = ηµν dxµ dxν = ds02 can be thought as:
(0, 2) ⊗ (1, 0) ⊗ (1, 0) = (2, 2) −−−−−−−−−−−−→ (0, 0) a scalar
double contraction

with ηµν called Minkoski tensor or metric of M (because it actually defines how to measure distances
in M). An important relation about the Minkowski tensor is the following:
η αµ ηµβ = δ αβ Kronecker tensor (22)
A fundamental operation we can perform using the metric is the lowering or rising of indices:
Aα = (A0 , A)
~ (controvariant vector) −→ Aβ ηαβ = Aα = (A0 , −A)
~ (covariant vector)
Bα = (B0 , B)
~ (covariant vector) −→ Bβ η αβ = B α = (B0 , −B)
~ (controvariant vector)

One last important tensor that will be employed for the study of SR and GR is the Levi-Civita symbol
(generalized for M):
αβγ → µνλρ (a pseudotensor) (23)

4 Electromagnetism (6-10-2020)

To discuss Electromagnetism, first thing first we must write Maxwell’s equations and in particular it
is useful to recall them using Gaussian unit convention ([E] = [B]):
~ x) = 4πρ(~x)
~ · E(~
∇ Gauss’ Law (24)

~ x) − 1 ∂ E(~x) = 4π J(~
~
~ × B(~
∇ ~ x) Ampere-Maxwell equation (25)
c ∂t c
~ x) = 0
~ · B(~
∇ No magnetic mono-poles (26)

~ x) + 1 ∂ B(~x) = 0
~
~ × E(~
∇ Faraday law (27)
c ∂t

11
(the first two are the in-homogeneous Maxwell’s equations while the last two the homogeneous ones) and
the equation of Lorentz force:
∂~
p ~ + ~v × B)
= e(E ~ (~ p is the relativistic momentum) (28)
∂t c
∂E
= F~L · ~v = eE
~ · ~v (29)
∂t
Some calculations.
1 d~p d~
p
F = m~a = m =
m dt dt
∂E ∂(F~ · d~s)
W = F~ · d~s → = = F~ · ~v
∂t ∂t
Deep look. In SR we have the problem of the non-universality of time, for this reason, for example,
an observer who measures the interval between two events can obtain a different result if compared with
one in another FoR. For this reason is very useful to define the proper time, which is a function of the
scalar ds2 . The proper time is the time measured by a clock solidal with the system itself.
ds2 = c2 dt2 − dx2 − dy 2 − dz 2 = c2 dτ 2
1√ 2
r r
dx v2 (dx2 + dy 2 + dz 2 )
→ dτ = γ(dt − β ) = γdt(1 − β ) = γdtγ = dt 1 − 2 = dt2 −
2 −2
= ds
c c c2 c

ct


~x
O

Figure 4: Time measured on the world-line.

Subsequently, we can define another vector, the velocity quadri-vector:


dxµ  cdt d~x     c ~v 
uµ = = , = γc, γ~v = p ,p (30)
dτ dτ dτ 1 − β2 1 − β2
from which we derives the most important relativistic quadri-momentum:
E 
pµ = muµ = (mγc, mγ~v ) = (m0 c, m0~v ) = , p~ (31)
c
where we have introduced the inertial mass m0 = γm.
On this theory, Einstein based his work. Maxwell’s equations are already relativistic, there were no
need to introduce a "relativistic Electromagnetism" (as we have introduced, for example, "relativistic
mechanics"), this means that on the transformation of a FoR to another, Maxwell’s equations are already
invariant in form.

4.1 Covariance in Maxwell’s equations


If we assume to have two FoRs K and K 0 such that x −
→ x0 we can verify that the equations of the fields
L
~ and B
E ~ are invariant in form in the new FoR under Lorentz transformations:
 invariant
~ x, t)), B(~
~ x, t) − ~ 0 (x~0 , t0 ), B
~ 0 (x~0 , t0 )

E(~ −−−−−→ E
under L

12
and so, in K 0 , we will have:
~ 0 (~x0 ) = 4πρ0 (x0 )
~0 · E

~ 0 (~x0 ) − 1 ∂ E (~x ) = 4π J~0 (~x0 )


~0 0
~0 × B

c ∂t c
~ 0 (~x0 ) = 0
~0 · B

~ 0 (~x0 ) + 1 ∂ B (~x ) = 0
~0 0
~0 × E

c ∂t
On the other hand, this fact is not true under Galilean transformations:

x~0 = ~x + ~v t
(32)
t0 = t

After this mathematical result, Einstein has been able to come up with the SR theory.
Some calculations. A general solution to the homogeneous Maxwell’s equations can be given by intro-
ducing a scalar potential Φ and a vector potential A
~ such that:

~ =∇
B ~ ×A
~
(33)
~ − 1 ∂A
~
~ = −∇Φ
E
c ∂t
in fact, we can verify that:

∇ ~ =∇
~ ·B ~ ·∇
~ ×A ~ = 0 (Properties of the ∇ operator)

~ + 1 ∂ B = −∇ ~ + 1 ∂ A ) + 1 ∂ B = −∇
~ × 1 ∂A + 1 ∂B = 1 ( ∂B − ∂∇ × A ) = 0
~ ~ ~ ~ ~ ~ ~ ~
~ ×E
∇ ~ × (∇Φ
c ∂t c ∂t c ∂t c ∂t c ∂t c ∂t ∂t
subsequently, if we substitute in the in-homogeneous equations we get:

~ − 1 ∂ A ) = −(∇2 Φ + 1 ∇
~ ~
~ =∇
~ ·E
∇ ~ · (−∇Φ ~ · ∂ A ) = 4πρ
c ∂t c ∂t
1 ∂ ~
A
from which: ∇2 Φ + ∇ ~ · = −4πρ (34)
c ∂t
~ − 1 ∂E = ∇ ~ − 1 ∂ − ∇Φ ~ − 1 ∂A
~  ~
~ ×B
∇ ~ × (∇~ × A)
c ∂t c ∂t c ∂t
~ + 1 ∂ ∇Φ + 1 ∂ A = 4π J~
~ 2~
= ∇(
~ ∇~ · A)
~ − ∇2 A (35)
c ∂t c ∂t2
2 c
now we can notice that if we consider a different vector and scalar potential:
~
A(x) ~ 0 (x) = A
−−−−−−−−−−→ A ~ + ∇ζ(x)
~ (36)
gauge transform
1 ∂ζ
Φ(x) −−−−−−−−−−→ Φ0 (x) = Φ(x) − (37)
gauge transform c ∂t

~ and E
the equations of B ~ remain the same:

~ 0 (x) = ∇
~ →B
B ~ ×A ~0 = ∇
~ ×A~+∇ ~ × (∇ζ(x))
~ =∇
~ ×A ~=B ~

~ Φ(x) − 1 ∂ζ − 1 ∂ (A 1 ∂A
  ~
E ~ 0 (x) = −∇
~ →E ~ + ∇ζ(x))
~ = −∇Φ(x)
~ − = E(x)
~
c ∂t c ∂t c ∂t

so both B~ and E ~ are gauge invariant. However, the in-homogeneous Maxwell’s equations generate a
different solution for every gauge transformation.

13
4.2 Tensor formalism in Electromagnetism
In SR we can rephrase electromagnetic covariance using tensor formalism, in fact, with the three compo-
nents of E
~ and B
~ we can define the antisymmetric Maxwell tensor F αβ :

0 −Ex −Ey −Ez


 
Ex 0 −Bz By 
F αβ =  with F 0i = −Ei , F ij = −ijk Bk (38)
0

Ey Bz −Bx 
Ez −By Bx 0

then, we can also define the (Hodge) dual tensor F ∗ ≡ F ? as:


1
F ∗µν ≡ µναβ Fαβ = −F ∗νµ with: F ∗0i = −Bi , F ∗ij = ijk Ek (39)
2
0 0 −Bx −By −Bz
  
Ex Ey Ez
−Ex 0 −Bz By   Bx 0 Ez −Ey 
with: Fαβ =  =⇒ F ∗αβ
=
0 0
   
−Ey Bz −Bx  By −Ez Ex 
−Ez −By Bx 0 Bz Ey −Ex 0
actually the duality transformation F αβ → F ∗αβ can be synthetized as: (E, ~ → (B,
~ B) ~ −E)
~

and we can immediately write the transformation law using tensor properties:

F αβ (x) −−→ F 0αβ (x0 ) = Λαµ Λβν F µν (40)


L

with:
γ−1 ~
~ 0 (x0 ) = γ(E(x)
E ~ + β~ × B(x))
~ − ~ β~
(E(x) · β) (41)
β2
~ 0 (x0 ) = γ(B(x) ~ × E(x)) γ−1 ~ ~ β~
B ~ −β ~ − (B(x) · β) (42)
β2

clearly the extended version is much more complicated, but what emerges is both fields E ~ and B~ are
frame-dependent in the sense that if we change FoR the new fields E ~ 0 and B
~ 0 are functions of the
previous (for this reason we can pass from a FoR where we have only a pure electric field (rest frame of a
particle) to another where both electric and magnetic appear and are orthogonal (one where the particle
is moving)).
Maxwell’s equations (24) using tensor formalism become:
4π ν
∂µ F µν = J with: J µ = (cρ, J)~ (in-homogeneous equations) (43)
c
∂µ F ∗µν ←→ ∂µ Fνρ + ∂ρ Fµν + ∂ν Fρµ = 0 (homogeneous equations) (44)

(Note we have c in the source quadrivector J~ in order to not have it in F αβ ).


Some calculations. Verification of tensor Maxwell’s laws:
∂Ei ∂Bi
∂µ F µ0 = =∇ ~ = 4πρ
~ ·E ∂µ F ∗µ0 = =∇ ~ ·B~ =0
∂xi ∂xi
1 ∂Ex ∂Bz ∂By 4π 1 ∂Bx ∂Ez ∂Ey
∂µ F µ1 =− + − = Jx ∂µ F ∗µ1 = − − + =0
c ∂t ∂y ∂z c c ∂t ∂y ∂z
1 ∂Ey ∂Bz ∂Bx 4π 1 ∂By ∂Ez ∂Ex
∂µ F µ2 =− − + = Jy ∂µ F ∗µ2 = − + − =0
c ∂t ∂x ∂z c c ∂t ∂x ∂z
1 ∂Ez ∂By ∂Bx 4π 1 ∂Bz ∂Ey ∂Ex
∂µ F µ3 =− + − = Jz ∂µ F ∗µ3 = − − + =0
c ∂t ∂x ∂y c c ∂t ∂x ∂y
~ − 1 ∂ E = 4π J~ ~ + 1 ∂B = 0
~ ~
~ ×B
⇒∇ ⇒∇ ~ ×E
c ∂t c c ∂t
(Factor 1c comes from the time derivation)
Or, if we assume for example to take µ = 0, ν = 1 and ρ = 2, using Bianchi identity:

1 ∂Bz ∂Ec ∂Ey 


~ + 1 ∂B
~
− + − = ∇~ ×E =0
c ∂t ∂y ∂x c ∂t z

14
To conclude, we have also to translate the formula of the Lorentz Force which becomes:
dpα e dxα
= F αβ uβ with: uβ = ηαβ = γ(c, −~v ) (45)
dτ c dτ
where uβ is the velocity quadri-vector (30).

4.3 Solution to Maxwell’s equations in tensor formalism


Recalling the general solutions of Maxwell’s equations as functions of Φ and A
~ (33) we can re-write them
using by introducing the 4-vector A = (Φ, A)
µ ~ and using the tensor F as: (actually its definition should
µν

come from here)


F µν = ∂ µ Aν − ∂ ν Aµ (46)
We can verify, for example, the case of:
1 ∂ ~
F i0 = ∂ i A0 − ∂ 0 Ai = ∂ i Φ − ∂ 0 A
~ = −∇
~ iΦ − A
c ∂t
where we obtain exactly the definition of all the three components of E.
~ (Note that ∂ µ = (∂t , −∇)
~ due
to the metric we have chosen).
Deep look. Actually we usually provide the definition of the electromagnetic tensor starting from the
one with the lower indices Fµν by introducing the potential Aµ = (Φ, −A),
~ from which we derive the
definition of an anti-symmetric tensor (6 equations):

Fµν = ∂µ Aν − ∂ν Aµ (47)

with ∂µ = (∂t , ∇).


~ This way we define compactly all the components of both the electric and magnetic
field within one in equation.
As a consequence, F µν immediately satisfies the homogeneous Maxwell’s equations (43). What we do
next, is to substitute F µν inside the in-homogeneous equations:
4π ν
∂µ F µν = ∂µ (∂ µ Aν − ∂ ν Aµ ) = Aν − ∂µ ∂ ν Aµ = J (48)
c
Nevertheless, again we can perform a gauge transformation like:

Aµ (x) −−−−−−−−−−→ A0µ (x) = Aµ − ∂ µ ζ(x) (49)


gauge transform

that does not affect the tensor F µν , in fact:

F 0µν = ∂ µ A0ν − ∂ ν A0µ = ∂ µ Aν − ∂ ν Aµ = F µν

But, as we have noticed previously, the in-homogeneous equations (4 equations) have infinite solutions,
in fact, given a Aµ we can always perform a gauge transformation to obtain a different solution. This is
why Maxwell’s equations are said to have a gauge symmetry.
Another interesting result comes by performing the 4-divergence of the in-homogeneous Maxwell’s equa-
tions:

∂ν ∂µ F µν = ∂ν J ν
c
but we know that F µν is anti-symmetric while ∂µ ∂ν is symmetric, so their combination is zero and we
just obtain:

∂ν J ν = ρ + ∇ ~ · J~ = 0 Continuity equation (50)
∂t
Due to the fact that we have introduced a potential ζ(x) without any constraint, we can try to simplify
the computation of the in-homogeneous Maxwell’s equations by performing a so called gauge fixing,
in particular, we can look for a potential that verify the condition ∂µ Aµ = 0 (this condition actually
diagonalize my system). If we are able to do that we obtain the system:
(
∂µ Aµ = 0 Lorentz gauge
(51)
Aν = 4πc J ν

15
However, first we must prove how to find such ζ(x).
So, let’s assume we have a 4-vector Aµ with ∂µ Aµ = f (x) and we perform a gauge transformation:

Aν −→ A0ν = Aν − ∂ ν ζ(x) such that:


∂ν A 0ν
= ∂ν (A − ∂ ζ(x)) = 0
ν ν
and so we just find the condition:
f (x) − ζ(x) = 0 −−−→ ζ(x) = f (x) (52)

(In literature we also have different gauge: the Coulomb gauge ∇ ~ = 0,the temporal gauge Φ = 0...).
~ ·A
Now, with this condition, my problem is uniquely defined? The answer is still no, in fact:

Aν −−−→ A0ν = Aν − ∂ ν θ
gauge

∂µ A0µ = ∂µ Aµ − θ = 0 −−−→ θ = 0
4π ν
Proof: A0ν = (Aν − ∂ ν θ) = Aν − ∂ ν (θ) = Aν + ∂ ν 0 = J
c
so, up to a harmonic functions we still have a sort of residual gauge freedom.
And so, we started with four degrees of freedom but we end up with just two as we have imposed two
additional conditions during the computation, however this is a very well known fact as electromagnetic
waves have only two physical components: physical quantities are gauge invariant.
Observation. The importance of a formulation based on potentials comes from a quantistic point of view:
in quantum mechanics we quantize potentials, not fields. Further, a formulation based on gauge theory
allows us to build a renormalizable quantum theory which is a extremely strong result (Renormalization
is fundamental for a quantum theory to make sense).

4.4 Scalars of the electromagnetic field (07-10-2020)

When looking for scalars we can finally take advantage of all this rephrasing of the electromagnetic theory,
in fact, identify scalars became much easier by using the electromagnetic tensor F µν . The first intuitive
but trivial scalar is its trace F µµ which unfortunately is zero, and so it is not very interesting. One more
scalar we might probably think of is and which is a very important Lorentz invariant is:

F µν Fµν = −2(|E| ~ 2 ) → F 0µν F 0 = F αβ Fαβ = −2(|E


~ 2 − |B|
µν
~ 0 |2 )
~ 0 |2 − |B (53)

so, for example, if the module of the two fields are equals in one FoR, they will always be equals in every
other FoR cause F µν Fµν has to be constantly 0.
Some calculations.
3 
X 
F µν Fµν = F µ0 Fµ0 + F µ1 Fµ1 + F µ2 Fµ2 + F µ3 Fµ3 =
µ=0

= −(Ex2 + Ey2 + Ez2 ) + (−Ex2 + Bz2 + By2 ) + (−Ey2 + Bz2 + Bx2 ) + (−Ez2 + By2 + Bx2 ) =
= −2|E|
~ 2 + 2|B|
~ 2 = −2(|E| ~ 2)
~ 2 − |B|

We can write another important invariant:

F ∗µν Fµν = −4(E ~ → F 0∗µν F 0µν = −4(E


~ · B) ~ 0)
~0 · B (54)

which basically tells us that, if for example E


~ and B
~ are orthogonal in one FoR, they will be orthogonal
in all FoRs. Let’s suppose we have:

K : E⊥
~ B~ with |E|
~ > |B|
~ −−→ K0 : E ~ 0 are 0
~ 0 or B
L

thanks to the second invariant, we can immediately tell that is |B|


~ the one which has to be zero and vice
versa if we start with |B| > |E|.
~ ~

16
4.5 Electromagnetic waves
Maxwell’s equations in Lorentz gauge in the case of no sources.
(
∂2
Aµ = 0 with  = c12 ∂t 2 − ∇
2
(D’Alambert operator)
(55)
∂µ A = 0
µ

The solutions to this problem are planes waves in the form:


α ω
Aν (x) = ν eikα x + c.c. with ν = (0 ,~) and k α = ( , ~k) (56)
c
α ~
where µ is a constant 4-vector and in which we can expand eikα x = ei(ωt−k·~x) (c.c. stands for complex
conjugates).
If we now substitute we find (remember  = ∂mu ∂ µ ):
∂ α α α
∂µ Aν = (ν eikα x ) = ν eikα x (ikα δ αµ ) = ν (ikµ )eikα x
∂xµ
α α
Aν = ν (ikµ )eikα x (ik µ ) = −ν (kµ k µ )eikα x = 0 and so:
ω2
kµ k µ = 0 −→ − |~k|2 = 0 −→ ω = c|~k| Dispersion relation
c2
and from this we learn a very important property of the EM waves which is that:

vg = =c (57)
d|~k|
Electromagnetic waves travel at the speed of light. But in order for Aν to be a solution, it must
also satisfy the Lorentz gauge condition:
α
∂µ Aµ = µ (ikµ )eikα x = 0 −→ kµ µ = 0
so the four components of µ are not completely independent, on the contrary:
~k · ~
k0 0 − ~k · ~ = 0 −→ 0 =
ω/c
In addition, we also know that the solution we are generating is not unique as we can always perform a
gauge transformation to add an harmonic component:
Aν −→ A0ν = Aν − ∂ ν θ
( ( α
A0ν = 0 A0ν = 0ν eikα x
−→ (58)
∂µ A0µ = 0 kµ 0µ = 0
in particular, to know the relation between Aν and A0ν we have to define the additional term (we can
chose it to be a simple plane waves too):
α
θ(x) = θ0 eikα x + c.c.
and by doing so we obtain the following relation:
α α α
0ν eikα x = A0ν = Aν − ∂ ν θ = ν eikα x − ik ν θ0 ei(kα x )

−→ 0ν = ν − iθ0 k ν
If we now consider the case for ν = 0 we find one first constraint for the general solution:
ω 0
00 = 0 − iθ0 with θ0 = −→ 00 = 0 (59)
c iω/c
(θ0 can be whatever we what cause we have no constraint on the amplitude of the function θ(x)). After
that, if we now implement the Lorentz gauge (58) we find the second constraint:
~ 0 · ~k
kµ 0µ = 0 −→ 00 = =0 −→ ~ 0 · ~k = 0 (60)
ω/c
which forces ~0 and ~k to be orthogonal. At the end then, we can conclude that by performing a gauge
transformation we obtain a solution A0ν with just two degrees of freedom instead of the appearing starting
four (we have got rid of the "time-component" (or scalar) and the longitudinal component).

17
Electromagnetic wave on the z-axis
Assume ~k k z, and so k α = (k 0 , 0, 0, k 3 ), since:

ω2 ω
kα k α = 0 −→ k0 k 0 − |~k|2 = 2
− k32 = 0 ←→ k 0 = k 3 = (From Maxwell’s equations)
c c
~0 · ~k
kα 0α = 0 −→ 00 = ⇐⇒ k0 00 − k3 03 −→ 00 = 03 (Lorentz gauge)
ω/c

but, as we have seen in equation (59) we can always find a gauge transformation such that 00 = 0 which
implies for both the amplitudes 00 = 03 to be zero, leaving only 01 and 02 free. Further, recalling the
equation:
0ν = ν − iθ0 k ν =⇒ 01 = 1 and 02 = 2
we see how the physical components are gauge invariant.

4.6 Helicity
An additional way to characterize a electromagnetic wave is by defining its helicity h, which is a quantity
that essentially describes how the wave is propagating:

Ψ −−−−−→ Ψ̃ = eihθ Ψ (61)


Rotation

If we consider again the example of an electromagnetic wave we have:

1 0 0 0
 
α 0 cos θ − sin θ 0
A0µ (x) = 0µ eikα x −−−−−−→ Ã0µ (x0 ) = Λµν A0ν (x) with: Λµν =
0 sin θ cos θ 0

Rotation
0 0 0 1
0± ≡ 01 ∓ i02 with: 0± −→ ˜0± = e±iθ 0± (Helicity h = ±1)
00 −→ 00 and 03 −→ 03 (zero helicity components)

Basically we passed from the linear to the circular polarization where it is immediate to see if our
electromagnetic wave has a positive or negative helicity (clockwise or anti-clockwise propagation).

Figure 5: Circular polarizations.

4.7 Solution in presence of sources (10-12-2020)

General solution to Maxwell’s equations with Lorentz gauge:


(
Aα = 4πc J
α
(62)
∂α A = 0
α

To solve this problem we will use a Green function D(t, ~x, t0 , ~x0 ) such that:

x D(t, ~x, t0 , ~x0 ) = δ 4 (x − x0 ) with δ 4 (x − x0 ) = δ(t − t0 )δ 3 (~x − ~x0 )


(63)
Z
remember that, by definition: d4 x f (x) δ 4 (x − x0 ) = f (x0 )

which is a bi-scalar (cause it depends on two points), but since the box operator and the sigma4 operator
are both scalars, then also D has necessarily to be a scalar. Moreover, due to the fact that the Minkowski
space-time is translation invariant, also D is translation invariant, in particular according to the only

18
possible relation D(t, ~x, t0 , ~x0 ) = D(t − t0 , ~x − ~x0 ).
The solution of this problem is in the form homogeneous term + particular term. The homogeneous
part is related to the propagation of electromagnetic waves as we have already seen, so we will focus on
the particular solution. By looking at the problem we can immediately write a particular solution in the
form:

Z
A (t, ~x) =
α
d4 x0 D(t − t0 , ~x − ~x 0 ) J α (t0 , ~x 0 ) from which we obtain:
c
(64)
4π 4π 4π α
Z Z
⇒ x Aα (t, ~x) = d4 x0 x D(x − x0 ) J α (t0 , ~x 0 ) = d4 x0 δ 4 (x − x0 ) J α (x0 ) = J (t, ~x)
c c c
So the point now is just to solve the Green function and to do that we need to use 4-d Fourier transform:
1
Z
α 0α
D(x − x ) =
0
d4 ke−ikα (x −x ) D̃(k) with D̃(k) Fourier transform of D(x − x0 )
(2π)4
1
Z
α 0α
δ 4 (x − x0 ) = d4 ke−ikα (x −x ) with k α = (k 0 , ~k) (Fourier representation)
(2π)4

Then, by replacing the Fourier representations in the starting equation we get:

x D(x − x0 ) = δ 4 (x − x0 )
h 1 Z i 1
Z
−ikα (xα −x0α ) α 0α
x 4
d ke D̃(k) = d4 ke−ikα (x −x )
(2π)4 (2π)4
1 1
Z Z
µ −ikα (xα −x0α ) α 0α
d 4
k(−k µ k )e D̃(k) = d4 ke−ikα (x −x )
(2π)4 (2π)4
1
− kµ k µ D̃(k) = 1 =⇒ D̃(k) = −
k − ~k 2
2
0

and so, by using Fourier transformation we have been able to reduce a differential equation into an
algebraic equation that we are able to solve and from which we obtain:
0
1 1 1 +∞
e−ik0 (ct−ct )
Z Z Z
−ikα (xα −0α ) 3~ i~ x0 )
D(x − x ) =
0 4
d ke = d ke x−~
k·(~
dk0 (65)
(2π)4 ~k 2 − k 2
0
(2π) 4
−∞ k 2 − k02

Once we arrive at this point we note that this function has two simple poles (k0 = ±k) and further, we
know that many Green functions can be solution of the problem up to an homogeneous term, but due to
the causality structure of SR the D we are looking for has to take the role of a propagator (basically,
it has to make the source in (t0 , ~x0 ) to generate a field that only exist when t > t0 and that permeate the
space at the speed of light). And so we can impose:

D(t − t0 , ~x − ~x0 ) = 0 for t < t0 (Retarded Green function) (66)

Im k0

Re k0

−k − i k − i

Figure 6: Poles representations in complex plane.

To solve this integrals then we shift our points a little bit lower then the actual poles:

(k 2 − k02 ) = (k − k0 )(k + k0 ) = lim+ (k − i − k0 )(k + i + k0 )


→0

19
and we obtain the limit:
+∞ 0
e−ick0 (t−t )
Z
Jretarded ≡ lim+ dk0 (67)
→0 −∞ (k − i − k0 )(k + i + k0 )

Finally we can proceed to solve the integral and in particular for t < t0 we close the circle in the positive
imaginary part (So D(x − x0 ) = 0) while we close it in the negative imaginary part when t > t0 , from
which we obtain:
 e−ick(t−t0 ) 0
eick(t−t ) 
0
 eick(t−t ) − e−ick(t−t )  i 0

Jret = −2iπ + = −2iπ · θ(ct − ct0 ) =


−2k 2k 2k i
sin(ck(t − t0 ))
= 2π θ(ct − ct0 )
k
where we have added the θ function just as reminder that we are considering the case of t > t0 (from now
on it will be implicit). Once performed the dk0 integration we can complete the computation with the
space variables:

1 1 d3~k i~k(~x−~x 0 )
Z Z
~ 0
Dretarded (x − x0 ) = d3~keik·(~x−~x ) Jret (k) = e sin(ck(t − t0 ))
(2π)4 (2π)3 k (68)
~k · (~x − ~x ) = k|~x − ~x | cos θ, and if we chose: (~x − ~x ) k z ⇒ θ becomes the azimut angle of ~k
0 0 0

and by defining d3~k = k 2 dk sin(θ)dθdφ we can solve the integrals in spherical coordinates:

1 dk k 2 sin(θ) dθdφ ikR cos θ


Z
Dref (x − x0 ) = e sin(ck(t − t0 )) with R = |~x − x~0 |
(2π) 3 k

Z +∞ Z π
= dk k sin(ck(t − t )) 0
eikR cos(θ) sin(θ)dθ = ξ = cos θ → dξ = − sin θdθ
(2π)3 0 0
Z −1
1
Z +∞
= dk k sin(ck(t − t0 )) −eikRξ dξ =
(2π)2 0 1
1
Z +∞ Z 1
= dk k sin(ck(t − t 0
)) eikRξ dξ =
(2π)2 0 −1
1 1 2 sin(kR)
Z +∞ h eikRξ i1 Z +∞
= dk k sin(ck(t − t 0
)) = dk k sin(ck(t − t0 )) =
(2π)2 0 ikR −1 (2π)2 0 kR
2 1
Z +∞ Z +∞
= dk sin(ck(t − t 0
)) sin(kR) = dk sin(ck(t − t0 )) sin(kR) =
(2π)2 R 0 (2π)2 R −∞
0 0
1
Z +∞
eick(t−t ) − e−ick(t−t ) eikR − e−ikR
= dk =
(2π)2 R −∞ 2i 2i
1
Z +∞
0 0
= dk (e−ick(t−t ) − eick(t−t ) )(eikR − e−ikR ) =
(2π)2 4R −∞
1
Z +∞
0 0 0 0
= dk e−ik(c(t−t )−R) − e−ik(c(t−t )+R) − eik(c(t−t )+R) + eik(c(t−t )−R)
(2π)2 4R −∞

Now, since t > t0 → c(t − t0 ) + R > 0 always, so, recalling the definition of the delta function:

1 +∞
Z
0
δ(x − x0 ) = dk eik(x−x )
2π −∞

the fact that δ(x) = δ(−x) and, finally, that δ(· · · ) only has support when (· · · ) = 0 we can conclude
that: Z +∞
0
dk e±ik(c(t−t )+R) −→ 0 always
−∞

and we are left with:


1 1
Dret (x − x0 ) = 2(2π)δ(−[c(t − t0 ) − R]) = δ(c(t − t0 ) − R) (69)
4R(2π)2 4Rπ

20
At the end, we can conclude with the particular solution

Z
Aαret (t, ~
x ) = d4 x0 Dret (x − x0 ) J α (t0 , ~x 0 ) =
c
1 − 0
) −
Z
δ(c(t t R) α 0 0
= d4 x0 0
J (t , ~x )
c |~x − ~x |
R
but d4 x0 = cdt0 d3 ~x 0 and the time integration is not null only if: c(t − t0 ) = R → t0 = t −
c
1 − 0
) −
Z Z
δ(c(t t R)
= d3 ~x 0 cdt0 J α (t0 , ~x 0 ) =
c |~x − ~x0 |
J α (t − Rc , ~x 0 )
Z
= d3 ~x (70)
|~x − ~x0 |

which is the formula of a potential, generated by a source, that propagates at the speed of light (the
potential at time t is the consequence of a source generating it at a time t0 (retarded time) such that
t − t0 = Rc where R is the distance between the point where I measure the potential and the source).
An easy way to identify the retarded time graphically, is by drawing the past light cone in the chosen
point. Nevertheless, to make this solution representing an electromagnetic potential it must also verify
the Lorentz gauge, which we still have not discussed.

ct
Aα (t, ~x)
J α (tj , ~xj )

t0
~x

Figure 7: Retarded time identification on a moving source.

The proof for the Lorentz gauge condition is pretty complicated and can be performed in many ways, in
our case we will have a very physical approach.
Assume two points in space-time xα and x0α and consider a delta function of the scalar product of their
difference:
δ[(x − x0 )α (x − x0 )α ] = δ[(x0 − x00 )2 − (~x − ~x 0 )2 ] = δ[(x0 − x00 )2 − R2 ] =
= δ{[(x0 − x00 ) + R][(x0 − x00 ) − R]}

if we then consider a function g(x) with simple zeroes for x = xi we can write the formula:
X δ(x − xi )
δ[g(x)] =
i
|g 0 (x)|xi

and in our case, we obtain (xi = ±R + x00 with g 0 (xi ) = 2xi − 2x00 ):
1
δ[(x − x0 )α (x − x0 )α ] = {δ[(x0 − x00 ) − R] + δ[(x0 − x00 ) + R]}
2R
but since t > t0 then x0 − x00 + R could never be zero, and so we can ignore it to just re-obtain the Green
function (69) up to a constant:

1 δ[c(t − t0 ) − R]
Dret (x − x0 ) = θ(ct − ct0 )
4π R
1
Dret (x − x0 ) = δ[(xα − x0α )(xα − x0α )]θ(x0 − x00 )

but written in this form its covariance is very explicit cause the first argument is a scalar and the second
argument only involves time variables, but we have already seen that due to the causality structure a

21
Lorentz transformation cannot reverse time. Further we have that Dret (x − x0 ) 6= 0 only if x − x0 is a null
vector, but we have also said that t > t0 , and so we can conclude that the support of Dret leans exactly
on the light cone. And finally we can verify the Lorentz gauge ∂µ Aµret = 0:

Z
µ ∂F ∂F
∂µ Aret = ∂µ Dret (x − x0 )J µ (x0 )d4 x0 but since for F (x − y) → =−
c ∂x ∂y

Z
=− ∂µ0 Dret (x − x0 )J µ (x0 )d4 x0 (4-divergence acts only on Dret )
c
4π 4π
Z Z
=− ∂µ [Dret (x − x )J (x )]d x +
0 0 µ 0 4 0
Dret (x − x0 )∂µ J µ (x0 )d4 x0
c c
but ∂µ0 J µ is just zero, is exactly the continuity equation (50) →

Z
=− ∂µ0 [Dret (x − x0 )J µ (x0 )]d4 x0 (Gauss theorem in 4 dimensions)
c

Z
=− Dret (x − x0 )J µ (x0 )dΣµ = 0 with dΣµ = nµ dΣ
c Σ
where Σ is the boundary of space-time. To explicate this result we must first realize the fact that in M
we have two types of infinity: a space infinity and a time infinity. At space infinity J µ is just zero
cause the sources are localized while at time infinity it is the Green function that results equal to zero
cause at the edge of time, two points will always be space-like separated (they will be on a line at t = ∞)
but we have seen that Dret 6= 0 only for two points null separated.

4.8 Systems of charged particles (12-10-2020)

We will now study the case of systems of charged particles which will be always described by a 4-vector
J α = (cρ, J)~ with ρ and J~ that, in this case, will assume the form:
X X d~xn
ρ(t, ~x) = en δ 3 (~x − ~xn (t)) and J(t, ~ ~x) = en δ 3 (~x − ~xn (t)) (t) (71)
n n
dt
X dxα (t)
−→ J α ≡ en δ 3 (~x − ~xn (t)) n = with x0 = ct ↔ J 0 = cρ
n
dt
dxα (t0 )
Z X
= dt0 δ(t − t0 ) en δ 3 (~x − ~xn (t0 )) n 0 =
n
dt
δ[x]
then, considering the relation δ[ax] = we can write: δ(t − t0 ) = cδ(ct − ct0 ) from which
|a|
0 dxn (t )
Z X α 0
= c dt0 en δ 4 (~xα − ~xα
n (t )) =
n
dt0
dt0 X dxα n (s) ds
Z
=c ds en δ 4 (~xα − ~xαn (s)) 0
= where ds2 = ηµν dxµ dxν =
ds n
ds dt
dxαn (s)
Z X
J α = c ds en δ 4 (~xα − ~xαn (s)) (72)
n
ds

which clearly is a 4-vector cause ds and δ 4 (· · · ) are both scalars while xα


n (s) is a 4-vector, so it will
transform following the rule J α → J 0α = Λαβ J β . If we now consider:
3 X
X ∂ 3 dxiµ (t)
~ · J~ =
∇ en [ δ (~
x − ~
x n (t))]
i=1 n
∂xi dt
X ∂ 3 dxiµ (t)
=− en [ δ (~
x − ~
x n (t))] (Remember the contraction)
n
∂xin dt
∂ 3 (73)
X
=− en δ (~x − ~xn (t)) =
n
∂t
∂ X ∂
=− en δ 3 (~x − ~xn (t)) = − ρ
∂t n ∂t

−→ ρ+∇
~ · J~ = 0 ←→ ∂µ Aµ = 0 (Continuity equation)
∂t

22
And if a charge distribution satisfies such a relation, we can also write:
Z Z
Q = d3 ~x J 0 (x) = d3 ~x ρ(x)
Z Z Z
∂Q 3 ∂ 0
= d ~x J (x) = d ~x (−∇ · J) = − d3 ~x (∇
3 ~ ~ ~ · J)
~
∂t ∂t

4.9 Energy-momentum tensor


Now, similarly to what we have done for a charge distribution, we will discuss a distribution of energies
and momenta. To be able to do that we need to introduce the concepts of density and current of
energy and momentum. Their formulas are basically the same, we just have to substitute the charges
which now are:
mn c2 mn v~n (t)
En = q and p~n = q (74)
2 (t)
vn v 2 (t)
1 − c2 1 − nc2
and so we have:
X
T α0 ≡ c n (t)δ (~
pα x − ~xn (t)) Density of n momentum pα (75)
3

n
X dxin (t)
T αi ≡ n (t)δ (~
pα x − ~xn (t)) Current of n momentum pα (76)
3

n
dt

which can be contracted as:


X dxβn (t) 3
T αβ = pα
n δ (~x − ~xn (t)) Energy-momentum tensor (77)
n
dt

Some calculations. To show that this is a tensor we have to follow the same proof employed for J α :

0 dxn (t )
Z X β 0
T αβ = dt0 n (t )
pα 0
δ(t − t0 )δ 3 (~x − ~xn ) =
n
dt
dxβn (s) 4
Z X
= c ds pα
n (s) δ (x − xn (s))
n
ds

An interesting property comes from the relation:


E En dxα
p~ = ~v −→ n =
pα n
c2 c2 dt
in fact, if we now replace it in the equation (77), we find that T αβ is symmetric:
X En dxα dxβ (t)
T αβ = n n
δ 3 (~x − ~xn (t)) = T βα
n
c2 dt dt

The next step is to study its divergence:

∂ αi X α ∂ 3 dxin
T = p n δ (~
x − ~
x n (t)) =
∂xi n
∂xi dt
X ∂ 3 dxi
=− pα
n i
δ (~x − ~xn (t)) n =
n
∂xn dt
X ∂ 3
=− pα
n δ (~x − ~xn (t)) =
n
∂t
∂ X α 3 X ∂pα
=− pn δ (~x − ~xn (t)) + δ (~x − ~xn (t)) =
n 3
∂t n n
∂t
1 ∂ α0 X ∂pα
=− T + δ (~x − ~xn (t))
n 3
c ∂t n
∂t

23
and in particular, if we introduce:
X dpα
Gα ≡ δ (~x − ~xn (t)) = F α
n 3
(Force density)
n
dt

we can write:
∂ αβ
T = Gα (78)
∂xβ
(for a free particle Gα is zero).
To conclude, we have the definition of the total 4-momentum P α , which corresponds to the integration
over the volume of the density of momenta:
Z
P = d3 ~x T α0
α
(79)

this quantity is conserved in absence of external forces. Nevertheless, in our case all particles are
dpα
electrically charged and so we have ∂tn 6= 0 due to the Lorentz force (45):

dpα en αβ dxβn
n
= F unβ with: uβn = (ds = dτ )
ds c dτ
so we can write:
X dpα dpα
n ds
X
Gα ≡ n 3
δ (~x − ~xn (t)) = δ 3 (~x − ~xn (t)) =
n
dt n
ds dt
X en ds
= δ 3 (~x − ~xn (t)) F αβ unβ =
n
c dt
X en α β ds
= δ 3 (~x − ~xn (t)) F β un =
n
c dt
X en α dxβn F αβ β
= δ 3 (~x − ~xn (t)) F β = J
n
c dt c
αβ
To conclude, we can say that generally we have ∂α Tpart 6= 0, but if we also consider the energy-momentum
tensor of the electromagnetic field (for this example) Te.m.
αβ
we can write:
α Z
αβ αβ dPtot
∂α (Tpart + Te.m.
αβ
) = ∂α Ttot = 0 ←−→ = 0 with Ptotα
= d3 ~x Ttotα0
dt
and say that the total energy and momentum of both particles and the electromagnetic field is conserved.

5 Perfect fluids (13-10-2020)

(Hints on the fact that energy-momentum tensor can be also extracted from Lagrangian dentity as a
Noether current in QFT.)
In physic we have many different systems that can be approximated to fluids, however this is a rather
complicated model, so we will only discuss perfect fluids. Energy-momentum tensor for fluids usually has
this interpretation:
- T 00 Energy density
- T i0
Density of i-component of momentum
- T 0i
Flux of energy along i
- T ij Stress tensor: in particular for i = j we have the pressures on the principal axes while for
i 6= j we have the shear forces
in particular we distinguish a perfect fluid as a system in which, considered the energy-momentum tensor
T̃ in the rest frame K̃ of a given particle, we have isotropic conditions and absence of frictions (no heat
conduction, no shear forces and no viscosity):
ρ 0 0 0
 
(
 0 p 0 0 T̃ 00 = ρ
T̃ = 
αβ
0 0 p 0 −−→ T̃ ij = pδ ij
 (80)
0 0 0 p

24
From the laboratory FoR then we can describe a perfect fluid with a tensor in the form:

T αβ = L(−~v )T̃ αβ = Λαµ (−~v ) Λβν (−~v ) T̃ µν with: (81)


(γ − 1)
with: Λ00 = γ, Λ0i = Λi0 = −βi γ and Λij = δij + βi β J (82)
β2

and in particular, if we compute the transformation (β changes sign cause of −~v ) we find:

T 00 = Λ0µ Λ0ν T̃ µν = Λ00 Λ00 T̃ 00 + Λ0i Λ0j T̃ ij =


= γ 2 ρ + γ 2 βi βj pδij =
= γ 2 (ρ + β 2 p) (83)
T 0i
= Λ0µ Λiν T̃ µν
= Λ00 Λi0 T̃ 00
+ Λ0j Λik T̃ jk
=
h (γ − 1) i
= γ 2 βi ρ + γβj δik + βi βk pδjk =
β2
h (γ − 1) i
= γ 2 βi ρ + γβj δij + 2
βi βj p =
β
= γ βi ρ + γβi p + γ(γ − 1)βi p =
2

= γ 2 βi ρ + γ 2 βi p =
= γ 2 βi (ρ + p) (84)
T ij
= Λiµ Λjν T̃ µν
= Λi0 Λj0 T̃ 00
+ Λik Λjl T̃ kl=
h (γ − 1) ih (γ − 1) i
= γ 2 βi βj ρ + δik + 2
βi βk δjl + 2
βj βl pδkl =
β β
h (γ − 1) ih (γ − 1) i
= γ 2 βi βj ρ + δik + β β
i k δ jk + β j k p=
β
β2 β2
h (γ − 1) (γ − 1) (γ − 1)2 i
= γ 2 βi βj ρ + δij + β β
j i + β β
i j + β i j p=
β
β2 β2 β2
h (γ − 1) γ 2 − 2γ + 1 i
= γ 2 βi βj ρ + δij + 2 β i β j + β i β j p=
β2 β2
h γ2 − 1 i
= γ 2 βi βj ρ + δij + β i β j p=
β2
h βi βj  1 i
= γ 2 βi βj ρ + δij + 2 − 1 p=
β 1 − β2
h β i βj β 2 i
= γ 2 βi βj ρ + δij + 2 p=
β 1 − β2
= γ 2 βi βj (ρ + p) + δij p (85)
dxµ
however, we can write them in a more compact form using the relation uµ = dτ = γ(1, ~v ) → ui = γβi
(usually the definition of uµ is γ(c, ~v )):

T ij = (ρ + p)ui uj − η ij p (86)
T 0i
= (ρ + p)u u
0 i
(87)
T 00
= (ρ + p)(u ) − η p
0 2 00
(88)

Equations that can be further simplified obtaining the energy-momentum tensor:

T µν = (ρ + p)uµ uν − η µν p (89)

which, by being a tensor, is valid in every FoR. In particular we can verify rapidly that:

T̃ µν = (ρ + p)uµ uν − η µν p with uα = γ(1, ~v ) = (1, 0) in K̃

One last element to complete the theory behind a perfect fluid is an equation of state. A very common
model is the Dust model in which we assume a pressureless condition p = 0 (model of pure gravitational
interaction).

25
5.1 Equations of motion of fluids
α

dt = 0 if ∂µ T
As already seen, for isolated systems we can prove that dp µν
= 0, from which we can extract
the equations of motion of fluids (relativistic version of Euler equations).
So let’s compute:
∂T αβ ∂p αβ ∂ h i
= − η + (ρ + p)u α β
u =
∂xβ ∂xβ ∂xβ
∂p ∂ h i
~
=− + (ρ + p)uα uβ = 0 for c = 1 then uλ = γ(1, β)
∂xα ∂xβ
∂T 0β ∂p ∂ h i
for α = 0 : = − + (ρ + p)u 0 β
u =
∂xβ ∂t ∂xβ
∂p ∂ h i ∂ h i
=− + (ρ + p)(u0 )2 + (ρ + p)u 0 i
u =
∂t ∂t ∂xi
∂p ∂ h i ∂ h i
=− + (ρ + p)γ 2 + (ρ + p)γ 2 v i = 0 (90)
∂t ∂t ∂xi
∂T iβ ∂p ∂ h i
for α = i : = − + (ρ + p)u i β
u =
∂xβ ∂xi ∂xβ
∂p ∂ h i
= i
+ (ρ + p)ui uβ =
∂x ∂xβ
∂p ∂h i ∂ h i
= + (ρ + p)u i 0
u + (ρ + p)u i j
u =
∂xi ∂t ∂xj
∂p ∂h i ∂ h i
= i
+ (ρ + p)γ 2 v i + j
(ρ + p)γ 2 v i v j =
∂x ∂t ∂x
∂p ∂v i h i
i ∂
h i ∂ h i
= + (ρ + p)γ 2
+ v (ρ + p)γ 2
+ (ρ + p)γ 2 i j
v v =
∂xi ∂t ∂t ∂xj
∂ h i ∂p ∂ h i
but from α = 0 we know: (ρ + p)γ 2 = − (ρ + p)γ 2 v i and we get
∂t ∂t ∂xi
∂p ∂v i h i n ∂p ∂ h io ∂ h i
= i
+ (ρ + p)γ 2 + v i − j
(ρ + p)γ 2 v j + j
(ρ + p)γ 2 v i v j = 0
∂x ∂t ∂t ∂x ∂x
∂p ∂v i h i ∂p ∂ h i
= + (ρ + p)γ 2 + v i − v i j (ρ + p)γ 2 v j +
∂xi ∂t ∂t ∂x
i ∂ j ∂
h i h i
+v (ρ + p)γ 2 j
v + v (ρ + p)γ 2 i
v =
∂xj ∂xj
∂p ∂v i h i ∂p ∂ h i
= i
+ (ρ + p)γ 2 + v i + v j j (ρ + p)γ 2 v i = 0
∂x ∂t ∂t ∂x
(1 − v 2 ) h ~ ∂p i ∂~v
=⇒ − ∇p + ~v = + (~v · ∇)~
~ v Relativistic Euler equation (91)
ρ+p ∂t ∂t
(in particular, if we re-introduce c, assuming it at its standard value, we obtain the classic Euler equations
for vc << 1 and ρ >> p).
Some calculations. We can verify that using the dust model equation of state (p = 0), from equation
(90) we get the continuity equation in the rest frame:

∂T 0β ∂p ∂h i ∂ h i
β
=− + (ρ + p)γ 2 + i
(ρ + p)γ 2 v i =
∂x ∂t ∂t ∂x
∂ h 2i ∂ h 2 ii
= ργ + ργ v =
∂t ∂xi
h ∂ρ
~ · (ρ~v ) = 0 −→ ∂ ρ + ∇
i
= γ2 +∇ ~ · J~ = 0
∂t ∂t

6 General Relativity (19-10-2020)

Attempts of introducing gravity in relativistic framework (just from a theoretical point of view, cause
observations were already almost perfect, there was only a residual precession angle of Mercury left
unexplained by Newton’s theory). Newtonian gravitation is based on the Poisson’s equation:

∇2 Φ = 4πGρ (92)

26
nonetheless this equation is clearly incompatible with SR due to the fact that a change in mass density
ρ, immediately cause a reshape of the entire gravitational field. In SR, when talking about forces and
fields we consider solutions in the form of retarded potentials that satisfy the equation:
4π µ
Aµ = J (93)
c
Einstein however, thought that the implementation of gravity in SR framework was possible by introduc-
ing an extension of the principle of Relativity: we have to consider also accelerated FoRs. (Gravity is a
very different interaction.)

6.1 Equivalence principle


The first important step in the introduction of GR was the formulation of the Equivalence principle.
Starting with Newtonian Physics, if we consider the system:

N
~g
K

we can describe it by the point of view of an inertial observer O using the equation:
2
(i) d ~xN (g)
X
mN 2
= mN ~g + F~ (~xN − ~xM ) (94)
dt
M

where m(i) is the inertial mass and m(g) the gravitational one. If we then perform a coordinate transfor-
mation (non-Newtonian transformation) in the form:
(
~x 0 = ~x − 21 ~g t2
(95)
t0 = t

d2 ~ x0
d2 ~
we move to a non-inertial FoR K 0 . In K 0 an observer O0 can rephrase the definition of dt2
x
= dt2 + ~g
and so, the equation of motion becomes:
2 0
(i) d ~
xN (i) (g)
X
mN + mN ~g = mN ~g + F~ (~xN0 − ~xM
0
) (96)
dt2
M

which, if we consider m(i) = m(g) , simplifies to:


2 0
(i) d ~
xN X
mN = F~ (~xN0 − ~xM
0
) (97)
dt2
M

So, the observer in K 0 describes a system without the influence of the gravitational field cause
the introduction of the inertial force miN ~g (related to the non-Newtonian transformation from an inertial
frame to one non-inertial), cancels the gravitational term. (Inertial forces are forces that directly depend
on the type of transformation, like the Coriolis force for rotating FoRs). Further, if we remove the
interaction between particles (non-interacting particles) we just obtain the equation of a uniform motion,
without accelerations at all:
2 0
(i) d ~
xN
mN =0 (98)
dt2
If we now think of K 0 from the point of view of O in K, we find that K 0 is just a free falling FoR under
the gravitational field in K, in fact:

1 d2 ~x
~x = ~x0 + ~g t2 but ~x 0 = cost in K 0 =⇒ = ~g (99)
2 dt2
If we now suppose to be in a different world such that:

27
~g
S0
p
S

we should consider, in passing from S to S 0 , the transformation:


(
~x 0 = ~x + 21 ~g t2
(100)
t0 = t

and so, one fixed point p in S (~xp = cost), in S 0 would have as equation of motion:

d2 ~x d2 ~xp0
m(i) =0 −−−−−−0→ m(i) = −m(g)~g (101)
dt2 K→K dt2
thus, we end up with two one-to-one relations between K ↔ S 0 and K 0 ↔ S. This means that the
description from an inertial frame on Earth can be linked to that of a non-inertial one and vice versa, so,
now, am I sure that the frame on Earth is inertial? To verify whether a frame is inertial or not we have
to be able to detect and remove all the forces, but this is not possible if a Gravitational Field is present
(we cannot remove it), so if I consider two FoRs, one on the Earth and one solidal with an object, and
I am in the one on the object I can say the object is in an inertial frame (it stands still in its FoR in
which seems that no forces are acting, not even gravity), besides it might seems untrue cause the object
is free-falling on Earth, id est it is accelerating with ~a = −~g . (In the previous examples we have that in
K 0 particles move freely, just like no gravitation fields are present, and so it might be said that it is the
inertial frame instead of the Earth’s surface K).
Observation. Just keep in mind that all these considerations are possible under the assumption of
~g = cost, so generally this will only true locally, where gravitational acceleration can be considered homo-
geneous.
From this reasoning, Einstein formulated the Equivalence principle:
in presence of a gravitational field, considered a given point, it is always possible to identify, locally,
an inertial FoR where the effects of gravity are null; so Physics is only ruled by the remaining funda-
mental interactions (weak, strong and electromagnetic) and sticks to SR. This basically implies the
equivalence between inertial and gravitational forces: gravity is just the result of a non-inertial
transformation that implies the introduction of an inertial force. Thus Gravity only appears in
non-inertial FoRs and for this reason it is not within SR, which only describes inertial FoRs −→
An extension of SR is needed to describe Gravity: General Relativity Theory.

6.2 Gravitational forces


Let’s now discuss the physics of a test particle in a gravitational field.

S0
~g {ξ α }
S

{xα }

where S 0 is a local inertial FoR solidal with the free falling particle. For an observer in S 0 we have:
dpα dξ α
= Fα = 0 with pα = muα and uα =
ds ds
d2 ξ α
and in particular = 0 with ds2 = ηµν dξ µ dξ ν (solidal FoR)
ds2

28
 
and then, to describe it in S we must know the transformation law S 0 ←→ S {ξ α } ←→ {xα } :

d  dξ α  d  dξ α dxµ  dξ α d2 xµ d2 ξ α dxµ dxν


= µ
= µ + =0
ds ds ds dx ds dx ds 2 ∂xµ ∂xν ds ds
∂xλ
and by multiplying it by we find that:
∂ξ α
d2 xλ ∂xλ ∂ 2 ξ α dxµ dxν
+ = 0 (Equations of motion)
ds2 ∂ξ α ∂xµ ∂xν ds ds
so we obtain 4 different equations of motion that are not equally zero, we actually have a force in this
coordinates, an inertial force that depends on the transformation performed to go in {xα } coordinates.
In particular we can define the Affine connection:

∂xλ ∂ 2 ξ α
Γλµν ≡ (102)
∂ξ α ∂xµ ∂xν

which is a (µ − ν) symmetric object. Using this definition we can extract the geodesic equation:

d2 xλ dxµ dxν
2
+ Γλµν =0 (103)
ds ds ds
the most natural
R trajectory we can have in a curved space-time (it can be also obtained by maximizing
the functional ds, in contrast with the least action principle, cause we are in a space with a different
signature).
Observation. Note how the problem gets more difficult compared with Newtonian theory in which we
just had 3 linear equations:
d2 xi ∂U
F~ = m~a −→ m 2 = F i = − i
dt ∂x
now, our equations of motion are four and are no longer linear (there is a double dependence on linear
variations):
d2 xλ λ dx dx
µ ν
+ Γ µν =0
ds2 ds ds
Further, if we write:
dξ α dξ β µ ν
ds2 = ηαβ dξ α dξ β = ηαβdx dx (104)
dxµ dxν
we can introduce the definition of the metric tensor gµν (which is actually symmetric):

dξ α dξ β
gµν (x) = ηαβ (105)
dxµ dxν
and finally synthetize the entire problem in the equations:
( 2 λ
λ dxµ dxν
ds2 + Γ µν ds ds = 0
d x
(106)
ds2 = gµν (x)dxµ dxν

An useful condition that we can attach to the problem is the normalization condition of the 4-velocity in
the non-inertial FoR which comes from:
dξ α dξ β
ηαβ = 1 cause ηαβ dξ α dξ β = ds2
ds ds
dξ α dξ β dxµ dxν dxµ dxν
−→ ηαβ µ ν = gµν =1 (107)
dx dx ds ds ds ds

6.2.1 Generalization for massless particles


If we consider the case of a photon, it comes from SR that in S 0 we have the equation of a straight line:

d2 ξ α dξ α dξ β
= 0 with η αβ =0 (108)
dσ 2 dσ dσ

29
Observation. For a light-speed particle it is not possible to parametrize its path in its rest frame cause
these particles have no rest frame [ds2 = c2 dτ 2 ], so we have to consider just the constraint ds = 0
(imagine the particle to move along the x-axis, then dx = cdt, so we have ds = (cdt)2 − (cdt)2 = 0 for
every time-parameter we chose.
but, if we explicit the variables, the problem in S takes the form:
( 2 λ
λ dxµ dxν
dσ 2 + Γ µν dσ dσ = 0
d x
µ
dxν
(109)
dσ dσ = 0
gµν dx

and so, in a FoR where the action of gravitational field is up, the solution of the problem is a curved
2 λ
trajectory ( ddσx2 6= 0).

At this point, one of the most interesting conclusions come by the comparison of the definitions of
ds2 in S 0 and S. In S 0 it has its common definition that comes from the SR, in particular we have that
ds2 represents a general distance in a flat plane that is related to the Minkowski metric ηµν . On the
other hand, in S the definition of ds2 is different and is related to the metric tensor gµν (xµ ) which is a
function of space coordinates (it depends from point to point, the space-time is now in-homogeneous), so
we are no longer in a flat plane but rather in a curved space-time (metric property of space-time). Is
this interpretation compatible with reality? Yes, in fact we can just reduce the principle of equivalence to
the Geometric Gauss Theorem which states that a curved surface can be locally approximated to a plane
surface (we have to think that in a 4-dimension manifold, which is not that easy...) and in such portion
of space-time gαβ → ηαβ and Γµαβ = 0. According to Einstein’s interpretation then, Gravitational Fields
and curved space-time are basically synonyms. In presence of a gravitational field inertial, FoRs no longer
exist as global entities but we can only recover them at a local level.

6.3 Discussions over the metric tensor (20-10-2020)

By looking at their definition, eqs.(105)-(102), it can be noticed that Γρµν is a function of the second
derivative of inertial coordinates while gαβ just of first derivative, this should suggest the possibility of
expressing the Affine Connection in term of derivatives of the metric tensor. Actually, this also allows
to make a parallel with Newtonian theory in which gravitational force was a derivative of gravitational
potential: in GR we actually have ten potentials represented in a tensorial form (the fact that gαβ is
a tensor will be proven later) and Γρµν represents all the gravitational and inertial forces coming from
them.

6.3.1 Derivation of the Affine connection


By performing a little of computing on the definition of the Affine connection:

∂xλ ∂ 2 ξ α ∂ξ β
Γλµν = multiply by
∂ξ α ∂xµ ∂xν ∂xλ
β
∂ξ ∂ξ β ∂xλ ∂ 2 ξ α ∂ξ ∂xλ
β
Γλµν λ = with = δαβ
∂x ∂xλ ∂ξ α ∂xµ ∂xν ∂xλ ∂ξ α
β
∂ξ ∂2ξβ
Γλµν λ = (110)
∂x ∂xµ ∂xν
∂gµν
and considering the formula for ∂xλ
, we find that:

∂gµν  ∂ 2 ξ α ∂ξ β ∂ξ α ∂ 2 ξ β 
= + ηαβ
∂xλ ∂xµ ∂xλ ∂xν ∂xµ ∂xν ∂xλ
α β α β
∂gµν 
ρ ∂ξ ∂ξ ρ ∂ξ ∂ξ
= Γ µλ + Γ νλ ηαβ
∂xλ ∂xρ ∂xν ∂xµ ∂xρ
∂gµν
= Γρµλ gρν + Γρνλ gρµ (111)
∂xλ

30
subsequently, by considering the combination:
∂gµν ∂gλν ∂gµλ
+ − = Γρµλ gρν + Γρνλ gρµ + Γρλµ gρν + Γρνµ gρλ − Γρµν gρλ − Γρλν gρµ
∂xλ ∂xµ ∂xν
∂gµν ∂gλν ∂gµλ
+ − = Γρµλ gρν + Γρνλ gρµ + Γρµλ gρν + Γρµν gρλ − Γρµν gρλ − Γρνλ gρµ
∂xλ ∂xµ ∂xν
∂gµν ∂gλν ∂gµλ
+ − = Γρµλ gρν + Γρµλ gρν = 2Γρµλ gρν (112)
∂xλ ∂xµ ∂xν
and introducing the inverse metric tensor:

∂ξ α ∂ξ β
gµν = ηαβ
∂xµ ∂xν
∂xµ ∂xν
g µρ : g µρ gρν = δ µν −→ g µν = g νµ = η αβ (113)
∂ξ β ∂ξ β
we finally obtain, by multiplying for g νσ :
1 νσ  ∂gµν ∂gλν ∂gµλ 
g νσ Γρµλ gρν = g + −
2 ∂xλ ∂xµ ∂xν
1  ∂gµν ∂gλν ∂gµλ 
Γρµλ δ σρ = g νσ + −
2 ∂x λ ∂x µ ∂xν
1  ∂gµν ∂gλν ∂gµλ 
Γσµλ = g νσ + − (114)
2 ∂x λ ∂x µ ∂xν
so the Affine connection can be expressed in terms of the elements of the metric tensor (10 potentials)
generated by the gravitational field. Therefore, what we have to find now are the relations between these
ten potentials and the sources of the gravitational field. (Just to recall, in electromagnetism we had the
Maxwell’s equations: Aµ − ∂ µ (∂ν Aν ) = 4π
c J that associate the four potentials in A to the sources in
µ µ

J ).
µ

6.4 Construction of a local inertial frame


Discussion to construct a local inertial frame {ξ α } from the metric tensor gαβ in the non-inertial FoR
{xα }.

X
~g {ξ α }

{xα }
gαβ

Starting from the formula:


∂2ξβ ∂ξ β
µ ν
= Γλµν λ (115)
∂x ∂x ∂x
we have to find the coordinates ξ α (xα ) around the point X we are considering (Basically we have to solve
this differential equation in the neighbourhood of X ). The solution to this problem can be written in the
form of a Taylor expansion like:
1 (β) λ
µ (x − X ) + b λ Γ µν
ξ(x)(β) = a(β) + b(β) µ µ
(xµ − X µ )(xν − X ν ) (116)
2 X

where:

a(β) = ξ (β) (x) (117)


X
∂ξ (β) (x)
µ =
b(β) (also named vierbein) (118)
∂xµ X

31
(β)
(pay attention to the fact in b µ the index (β) is between brackets to differentiate it from µ as the former
refers to the new local inertial frame coordinates while the latter to the coordinates of the initial FoR).
In particular a(β) are 4 constants that substantially set the origin of X (are related to translations, the
(β)
additional four parameters of the Poincarè group). The discussion around b µ is a bit less immediate
and, recalling the formula of gµν , we can prove that these are not 16 independent constants as one might
expect:

∂ξ α (x) ∂ξ β (x)
gµν (x) = ηαβ and for x = X :
∂xµ ∂xν
gµν (X ) = bαµ bβν ηαβ (119)

and so, not all components are independent cause gαβ is symmetric: 16 parameters with 10 equations
to satisfy implies that we are left with 6 degrees of freedom, which correspond exactly to those of
Lorentz transformations, in fact a local inertial frame remains indeed a local inertial frame under Lorentz
transformations (simplified example were the origin is not translated):

ξ (α) (x) −−→ ξ 0(α) (x) = Λαβ ξ (β) (x) with


L
1 (β)
ξ(x) 0(α)
= Λαβ b(β)µ (xµ − X µ ) + Λαβ b λ Γλµν (xµ − X µ )(xν − X ν )
2 X
β (λ) α β
0
and: gµν (x) = b0(α)
µ b ν ηαβ = Λ ρ b µ Λ λ b ν ηαβ but Λ ρ Λ λ ηαβ = ηρλ
0(β) α (ρ)
X

so 0
gµν (x) = b0(α)
µ b ν ηαβ = b µ b ν ηρλ = gµν (x)
0(β) (ρ) (λ)
X X

Substantially then, with the definition of bβµ we are choosing a specific FoR among the infinite possibilities
available using Lorentz transformations.

7 Curved space-time (21-10-2020)

7.1 Brief some-up


In GR space-time becomes a dynamical arena. Back in SR for example, thinking of electromagnetism, an
electric charge generates an electromagnetic potential that permeates all the space and which, of course,
given another test charge, exercise a force over it, but the Minkowski space-time M does not change
during this process. Instead, in the case of a mass that generates a gravitational field, given a test mass,
the situation maybe appears rather similar in term of a force exercised but how it comes is completely
different: the space-time is no longer flat and the test particle does not curve as a consequence of a
force, it just moves freely on a curved surface. Thus, the concept of gravitational space field is replaced
by curve space-time. Even light couples with gravitational fields (case of massless particle), so photons
follow bended trajectories too, nevertheless the equations are a little different cause of the nature of light:
( 2 µ ( 2 µ
µ dxα dxβ µ dxα dxβ
d x
ds2 + Γ αβ ds ds = 0 d x
dσ 2 + Γ αβ dσ dσ = 0
µ
dxν
−−−−−−−−−→ µ
dxν
ds ds = 1
gµν dx gµν dxdσ dσ = 0
mass-less case

Another important aspect is that GR inherit the causal structure coming from the maximum speed limit
c of SR but, in a curved space-time, how do light-cones look like? Due to the fact that these were actually
drawn from light particles in the Minkowski space-time, in GR the causal structure is determined
by gravity, which is able to curve photons and so, also to bend light cones:

ct ct
x = ct x = ct

~x ~x
O
O

(a) Light-cone in SR. (b) Example of a light-cone in


GR.

32
7.2 Principle of General Relativity: introduction of tensor elements
An important aspect to discuss is now related to non-inertial FoRs, in fact, it is easy to imagine that,
using specific transformations T , it is possible to identify much more than a single coordinates system
S 0 = {xα }, and, as we have seen in SR for inertial FoR, physics should now be the same in any non-inertial
FoR S 0 according to the principle of General Relativity (it still does not exist a preferable FoR). To
verify this invariance of physical equations, the most convenient way is again to use tensor formalism as
we had done in SR.

{ξ α }

{x0α }
α
{x } gαβ

So we start with the introduction of a generic transformation rule like:

∂x0α ∂xβ
xα −→ x0α = Λαβ xβ with T αβ = ←→ (T αβ )−1 = (120)
∂x β ∂x0α
(T αβ is not necessarily linear, Lorentz transformations were linear cause were made of constants, here
matrices are in general point-dependent). Subsequently we have to define mathematically all the physical
tools needed to describe phenomena, so scalars, tensors, derivatives and such things to perform differential
calculus in a curve manifold:

- scalars or (0, 0) tensors:


Φ(x) −−−→ Φ0 (x0 ) = Φ(x)
T

such the distance:


∂ξ α ∂ξ β
ds2 = gµν (x)dxµ dxν = ηαβ dxµ dxν which becomes:
∂xµ ∂xν
∂ξ α ∂ξ β
ds02 = g0µν (x0 )dx0µ dx0ν = ηαβ dx0µ dx0ν =
∂x0µ ∂x0ν
∂ξ α ∂xρ ∂ξ β ∂xσ ∂x0µ ∂x0ν
= ρ 0µ σ 0ν
ηαβ λ dxλ γ dxγ =
∂x ∂x ∂x ∂x ∂x ∂x
∂ξ α ∂ξ β
= ηαβ dxλ dxγ δ ρλ δ σγ =
∂xρ ∂xσ
∂ξ α ∂ξ β
= ηαβ dxλ dxγ = gλγ (x)dxλ dxγ = ds2
∂xλ ∂xγ

- controvariant vectors or (1, 0) tensors (transforms with T ):

∂x0µ ν
V µ (x) −−−→ V 0µ (x0 ) = V (x)
T ∂xν
an example of this is given by:
∂x0µ ν
dxµ −→ dx0µ = dx
∂xν
- covariant vectors or (0, 1) tensors (transforms with T −1 ):

∂xν
Uµ (x) −−−→ U 0µ (x0 ) = Uν (x)
T ∂x0µ
and as we did for SR, an example is the gradient:

∂Φ(x) ∂ 0 Φ0 (x0 ) ∂Φ(x) ∂xα



− −
→ =
∂xµ T ∂x0µ ∂xα ∂x0µ

33
- (µ, ν)tensors (just a generalization):
∂x0µ ∂xβ ∂x0γ α ρ
F µν ρ (x) −−−→ F 0µν ρ (x0 ) = F (x)
T ∂xα ∂x0ν ∂x0ρ β
Proof. Now we are going to prove that the metric gµν is a tensor.
Let’s consider two metrics in different FoR and do some computations:
∂ξ α ∂ξ β
gµν (x) = ηαβ
∂xµ ∂xν
∂ξ α ∂ξ β
gµν (x0 ) = ηαβ 0µ 0ν =
∂x ∂x
∂ξ α ∂xρ ∂ξ β ∂xσ ∂xρ ∂xσ
= ηαβ ρ 0µ σ 0ν = gρσ (x) 0µ 0ν
∂x ∂x ∂x ∂x ∂x ∂x
so it transform with two inverse matrices and so it is a (0, 2) tensor. Similarly it can be proven that
g µν (x) is a (2, 0) tensor −→ gµν g νρ = δ ρµ .
Recalling the fundamental operations of Tensor Algebra seen in section 3.1:
a) Linear combinations :
(µ, ν) ⊕ (µ, ν) = (µ, ν)
b) Direct products :
(µ, ν) ⊗ (µ0 , ν 0 ) = (µ + µ0 , ν + ν 0 )
c) Contraction :
(µ, ν) −→ (µ − 1, ν − 1)
what is different now are the lowering and rising of indices, which make use of the metric tensor gµν this
time:
gµν T µρσ (0, 2) ⊗ (2, 1) −−−−−→ (2, 3) −−−−−−−→ Rν ρσ (1, 2)
product contraction
and so, in the final object we can have completely different components and not just changes in signs as
we had with ηαβ ; as a consequence, also physical and mathematical properties can be completely different
(that’s why we also change the name of the tensor typically).

7.2.1 Tensor densities


0
Tensor densities are a mathematical tool which instead of just transforming like covariant ( ∂x ∂x ) or con-
trovariant tensors ∂x0 , are also divided by the determinant of the transformation matrix (the Jacobian).
∂x

For example, given:


∂xα ∂xβ
g 0µν (x0 ) = gαβ (x) 0µ 0ν (121)
∂x ∂x
we have that:
∂x ∂x0 −2
g 0 (x0 ) = | 0 |2 g(x) = | | g(x) with:
∂x ∂x
∂x
g = | det gµν (x)|, g 0 = | det g 0µν (x0 )| and | 0 | is the Jacobian of x0 −→ x
∂x
so we can see that the determinant of the metric tensor is not properly a scalar but a scalar density
(scalar density of weight −2). Similarly we have tensor densities, :
∂x0 ω ∂x0µ ∂x0σ ρ
T 0µν (x0 ) = | | T (122)
∂x ∂xρ ∂x0ν σ
in this case for example, this is a (1,1) tensor density of weight ω. As one might already intuit, the power
of the Jacobian defines the weight.
From this discussion we can extract an invariant after the transformation:
∂x0 4
d4 x −→ d4 x0 = | |d x (123)
∂x

in fact, we actually find that g d4 x is a scalar (d4 x transform with a weight of 1, g with a weight of −2,
√ 4
so the total weight of g d x is zero). So we have an invariant measure:
√ 4
Z
gd x (124)

34
8 Tensor analysis (27-10-2020)

Given two tensors equal in one FoR T µν = S µν , then, also their transforms will be equal T 0µν = S 0µν .
For this reason, it easy to note that tensors perfectly fulfill the principle of General Relativity (there is no
preferable FoR, equations written in tensor formalism remain valid in any FoR). As a corollario, we have
that, if a tensor is zero in one FoR, then it is zero in every FoR: T αβ = 0 −→ T 0αβ = 0. In this geometrical
interpretation of the theory, the basic objects we have seen are ds2 = gµν (x)dxµ dxν , which describes the
curvature of the space-time, and the Affine connection Γαβγ which describes the gravitational and inertial
forces. Although it might seems the opposite, actually Γ is not a tensor, in fact, it is sufficient to think
that it is zero in the inertial frame but it is inevitably not zero in all other FoRs (so it is not in accord
with the previous corollario).
Proof. Mathematical proof of Γ not being a tensor.
So if we think of two different Γ’s we have:

∂xλ ∂ 2 ξ α
Γλµν =
∂ξ α ∂xµ ∂xν
∂x0λ ∂ 2 ξ α
Γ0λµν = =
∂ξ α ∂x0µ ∂x0ν
∂x0λ ∂xρ ∂  ∂ξ α ∂xσ 
= =
∂xρ ∂ξ α ∂x0µ ∂xσ ∂x0ν
∂x0λ ∂xρ  ∂xσ ∂ 2 ξ α ∂xλ ∂ξ α ∂ 2 xσ 
= + =
∂xρ ∂ξ α ∂x0ν ∂xλ ∂xσ ∂x0µ ∂xσ ∂x0µ ∂x0ν
∂x0λ  ρ ∂xσ ∂xλ ∂ 2 xσ 
= Γ λσ + δ ρ
σ =
∂xρ ∂x0ν ∂x0µ ∂x0µ ∂x0ν
∂x0λ ∂xσ ∂xλ ρ ∂x0λ ∂ 2 xρ
= 0ν 0µ
Γ λσ +
ρ
∂x ∂x ∂x ∂xρ ∂x0µ ∂x0ν
and it clearly emerges an extra term (the first term actually corresponds to the tensor transformation
law). This additional term, also called in-homogeneous term, do disappear for linear transformation (due
to the second derivatives), but this is only a specific case between all the possible FoR transformations.
By performing some computation it is also possible to obtain another version of that transformation law,
and in particular, if we start from the relation:

∂x0λ ∂xρ
= δ λν
∂xρ ∂x0ν
∂  ∂x0λ ∂xρ 
=0
∂x0µ ∂xρ ∂x0ν
∂x0λ ∂ 2 xρ ∂xρ ∂ 2 x0λ ∂xσ
= −
∂xρ ∂x0µ ∂x0ν ∂x0ν ∂xσ ∂xρ ∂x0µ
we end up with an alternative expression of the in-homogeneous term, and so, for {xα } ↔ {x0α }, we
have:
∂x0λ ∂xσ ∂xλ ρ ∂x0λ ∂ 2 xρ
Γ0λµν = 0ν 0µ
Γ λσ + =
ρ
∂x ∂x ∂x ∂xρ ∂x0µ ∂x0ν (125)
∂x0λ ∂xσ ∂xλ ρ ∂xρ ∂ 2 x0λ ∂xσ
= Γ −
∂xρ ∂x0ν ∂x0µ λσ ∂x0ν ∂xσ ∂xρ ∂x0µ
(in particular if we chose {x0α } to be {ξ α }, the Affine connection would just be zero).

8.1 Differentiation of tensors: the covariant derivative


But if gµν is a tensor and Γ can be written as function of ∂gµν , how does it come that Γ is not a tensor?
This is due to how differentiation works in the framework of GR. If we consider a scalar:

Φ(x) −−−→
0
Φ0 (x0 ) = Φ(x)
x→x

then it’s immediate to prove that ∂µ Φ(x) is a (0, 1) tensor (controvariant vector).

35
Proof. Mathematical computation:
∂   ∂Φ(x) ∂xρ ∂xρ
∂µ Φ(x) −−−→ ∂ 0µ Φ0 (x0 ) = Φ(x) = = ∂ρ Φ(x)
0
x→x ∂x0µ ∂xρ ∂x0µ ∂x0µ
which is the transformation law of a controvariant vector.
But if we consider a covariant vector:
∂x0µ ν
V µ (x) −−−→ V 0µ (x0 ) = V (x)
0
x→x ∂xν
then we have:
∂V µ (x) ∂V 0µ (x0 ) ∂  ∂x0µ ν 
−−−→ = V (x) =
∂xλ x→x0 ∂x0λ ∂x0λ ∂xν
∂x0µ ∂V ν (x) ∂xρ ∂ 2 x0µ ∂xρ ν
= + V (x)
∂xν ∂xρ ∂x0λ ∂xρ ∂xν ∂x0λ
and so the derivative of a tensor is not a tensor cause, again, we have the presence of an additional
in-homogeneous term (and, in general, this is not zero). And this explain why Γ cannot be a tensor.
Therefore things get much more complicated cause the equations of physics are mostly differential equa-
tions, which, as we have now proven, are not invariant in form. To bypass this problem we have to think
of a way such that in transformation rules the in-homogeneous terms just cancel out. By comparing the
transformation laws of Γ and ∂λ V µ we can note a certain similarity between the in-homogeneous terms:
 ∂x0µ ∂xρ ∂xσ ∂ 2 x0µ ∂xρ ∂xσ  ∂x0β µ

Γµλβ V β −−−→ Γ λβ V 0β
= Γ ν
ρσ − V =
x→x0 ∂xν ∂x0λ ∂x0β ∂xρ ∂xσ ∂x0λ ∂x0β ∂xµ
 ∂x0µ ∂xρ 2 0µ
∂ x ρ
∂x σ 
= δ σµ Γνρσ − δ Vµ =
ν
∂x ∂x 0λ ∂x ∂x ∂x0λ µ
ρ σ

∂x0µ ∂xρ ν ∂ 2 x0µ ∂xρ σ


= 0λ
Γ ρσ V σ − V
ν
∂x ∂x ∂xρ ∂xσ ∂x0λ
in fact, it can be noticed that the in-homogeneous term of Γ0µλβ V 0β is just the opposite of that of ∂ 0λ V 0µ
(up to indices):
∂ 2 x0µ ∂xρ ν ∂ 2 x0µ ∂xρ σ
V (x) ←→ − V
∂xρ ∂xν ∂x0λ ∂xρ ∂xσ ∂x0λ
and if we combine them, their transformation law becomes:
∂V 0µ 0µ ∂x0µ ∂xρ  ∂V ν 
+ Γ λβ V 0β
= + Γ ν
ρσ V σ
(126)
∂x0λ ∂xν ∂x0λ ∂xρ

so, we can conclude that ∂V
∂x0λ
+ Γ0µλβ V 0β is a (1, 1) tensor. It follows that we have just defined the
covariant derivative, which is indicated as:
∂V µ ∂x0µ ∂xρ ν
V µ;λ ≡ + Γµλσ V σ −−−−−→ V 0µ;λ = V (127)
∂xλ ∂xν ∂x0λ ;ρ
Observation. It clearly is not an ordinary derivative, in fact the product Γµλσ V σ has to be added in
order to have a tensor but remember that working with tensors is fundamental to identify invariances,
and since the majority of physics equations are differential, being able to define differentials as tensors is
strictly necessary.
Further, we can verify that:
∂Vµ
Vµ;ν = − Γλµν Vλ (128)
∂xν
is a (0, 2) tensor, and so that the covariant derivative also works on controvariant objects.
Hence, we could have, for example:
∂T µσλ
T µσλ;ρ = + Γµρν T νσλ + Γσρν T νµλ − Γνλρ T µσν
∂xρ
so, basically, the covariant differentiation rule can be schematized as:
D(m, n) −→ (m, n + 1) (129)
Speaking about the properties of covariant differential operators we have that:

36
- Linearity:
(αAµν + βBνµ );λ = αAµν;λ + βB µν;λ
- Leibniz rule:
(Aµν B λ );ρ = Aµν;ρ B λ + Aµν B λ;ρ

8.1.1 Covariant derivative of the metric tensor


What we are looking for now are tensor equations able to describe a gravitational field in relation to its
sources. Recalling the differential equation of Newtonian theory, we should expect some tensor equations
with covariant derivatives of the potentials (gµν ) on one side and sources on the other. Nevertheless, if
we compute the covariant derivative of the metric tensor (the normal derivative we already show that
does not return a tensor when we have discussed Γµνλ ) we find:
∂gµν
gµν;λ = − Γρµλ gσν − Γρνλ gρµ =
∂xλ
∂gµν 1 ρσ   1 ρσ  
= − g g λσ,µ + gµσ,λ − g µλ,σ gρν − g gλσ,ν + gνσ,λ − gνλ,σ gρµ =
∂xλ 2 2
∂gµν 1  1 
= − gλσ,µ + gµσ,λ − gµλ,σ δ σν − gλσ,ν + gνσ,λ − gνλ,σ δ σµ =
∂x λ 2 2 (130)
∂gµν 1  1 
= − gλν,µ + gµν,λ − g µλ,ν − g λµ,ν + gνµ,λ − gνλ,µ =
∂xλ 2 2
∂gµν 1 1
= − gµν,λ − gνµ,λ =
∂xλ 2 2
∂gµν
= − gµν,λ = 0
∂xλ
so our idea of writing Field Equation as covariant derivatives of gµν has to be aborted, the problem is
much more complicated. (Similarly we can check that g µν;λ = 0). So, under covariant differentiation, the
metric tensor is just like a constant and this also implies, for example, that:
(g µν Vσ );λ −−−−−−−→ g µν Vσ;λ
Leibniz rule

As a consequence, the only way we have to write differential equations for the metric tensor is to combine
non-tensorial terms in a way that their transformation law has self-erasing in-homogeneous terms (Gauss
and Riemann has been able to solve this problem).

8.1.2 Particular cases of covariant derivatives


i) scalars, in case of a scalar S we just have:
∂S
S;µ =
∂xµ
and if we want to consider even a more difficult case:
 ∂V   ∂V λ 
λ
(Vλ V λ );µ = Vλ;µ V λ + Vλ V λ;µ = V λ − Γ σ
λµ V σ + V λ + Γ λ
σµ V σ
∂xµ ∂xµ
λ
∂Vλ ∂V ∂
= V λ µ + Vλ µ − V λ Γσλµ Vσ + Vλ Γλσµ V σ = (V λ Vλ )
∂x ∂x ∂xµ
ii)Vµ;ν − Vν;µ , in fact, due to Γσµν symmetry for µ − ν, we have:
∂Vµ  ∂V
ν
 ∂V
µ ∂Vν
Vµ;ν − Vν;µ = ν
− Γαµν Vα − λ
− Γβµν Vβ = −
∂x ∂x ∂xν ∂xλ
∂V
iii)covariant divergence, which correspond to V µ;µ = ∂xµµ + Γµµλ V λ , but if we perform some compu-
tation we find that:
1  ∂g
µρ ∂gρλ ∂gλµ   ∂g
ρλ ∂gλµ 
Γµµλ = g µρ + − = but − is anti-symmetric
2 ∂xλ ∂xµ ∂xρ ∂xµ ∂xρ
1 ∂gµρ ∂ ∂
= g µρ λ = and if we consider Tr[M −1 (x) λ M (x)] = ln (det M (x))
2 ∂x ∂x ∂xλ
1 ∂ ∂ 1 ∂ p
= ln |g| = ln |g| = p
p
|g|
2 ∂xλ ∂x λ
|g| ∂xλ

37
so we can write:
∂Vµ Vλ ∂ p 1 ∂ √
V µ;µ = + |g| = p (V µ g)
∂xµ λ µ
p
|g| ∂x |g| ∂x

iv)covariant divergence of (2, 0) tensors, like in the case of:


∂T µν
T µν;µ = + Γµµλ T λν + Γνµλ T µλ = (we use the previous result)
∂xµ
∂T µν
1  ∂ p 
= µ
+p |g| T λν + Γνµλ T µλ =
∂x |g| ∂xλ
1 ∂  µν p 
=p T |g| + Γνµλ T µλ
|g| ∂xµ

and in the case T µλ is anti-symmetric, then the last term just vanish cause Γν µλ is symmetric in
µ − λ. Thus, in the case of Aµν = −Aνµ we just have:
1 ∂ p 
Aµν;µ = p |g|A µν
|g| ∂xµ

v)particular case of an anti-symmetric tensor, if we have Aµν = −Aνµ and we look at:
∂Aµν
Aµν;λ = − Γρµλ Aρν − Γρνλ Aµρ
∂xλ
∂Aλµ
Aλµ;ν = − Γρλν Aρµ − Γρµν Aλρ
∂xν
∂Aνλ
Aνλ;µ = − Γρνµ Aρλ − Γρλµ Aνρ
∂xµ
we can find that:
∂Aµν ∂Aλµ ∂Aνλ
Aµν;λ + Aλµ;ν + Aνλ;µ = + +
∂xλ ∂xν ∂xµ
due to anti-symmetric property and cycling combination.

8.2 Covariant differentiation along a curve (28-10-2020)

Sometimes in physics we have to deal with tensor that are not defined throughout all the space-time but
only along a precise curve. That could be the case, for example, of the 4-momentum of a particle moving
along a trajectory parametrized with respect to its proper time:

pµ (xµ (s))

xµ (s)

Figure 8: Example of the 4-momentum tensor define only on a precise curve in space-time

So what we want is being able to study the variation of a generic vector Aµ (s) along a precise path that
satisfies the usual transformation law:
∂x0µ ν
Aµ (x(s)) −−−→ A0µ (x0 (s)) = A (x(s))
0
x→x ∂xν
dAµ (s)
The very natural modus operandi is to define the object ds , but we have to verify if this is a tensor:
dAµ (s) dA0µ (s) d  ∂x0µ ν 
−−−→ = A (s) =
ds x→x0 ds ds ∂xν
∂x0µ dAν (s) ∂ 2 x0µ dxλ ν
= + A (s)
∂xν ds ∂xν ∂xλ ds

38
But, as it can be noticed, this is not a tensor due to the presence of an additional term in the transfor-
mation law. Thus, again, we have to find a slightly different expression for the differential, that will be
named covariant derivative along a curve, able to cancel the in-homogeneous term of the previous
differential:
DAµ (s) dAµ (s) dxλ ν
≡ + Γµνλ A (131)
Ds ds ds
which transforms in:
DA0µ (s) ∂x0µ DAν
=
Ds ∂xν Ds
and, as a consequence, we can also write the definition for controvariant vectors:
DBµ (s) dBµ (s) dxν
≡ − Γλµν Bλ (132)
Ds ds ds
Observation. The definitions are very similar to that of covariant derivative for covariant and con-
ν
trovariant tensors but, in addition, here it does appear the tangent element of the curve: dx ds (just
remember we are differentiating in function of the proper time basically), in fact, if we perform some
computation we find that:
DAµ (s) dAµ (s) dxλ ν dAµ (s) dxλ dxλ ν
= + Γµνλ A = + Γ µ
νλ A =
Ds ds ds dxλ ds ds (133)
 dAµ (s)  dxλ dxλ
= λ
+ Γµνλ Aν ≡ Aµ;λ
dx ds ds
and similarly:
DT µν dxλ
= T µν;λ = T µν;λ uλ (134)
Ds ds
In the specific case in which the covariant derivative along a curve of a vector field is zero , we say that the
vector field is parallel transported along that curve which is described by the differential equation:
DAµ dAµ dxλ ν
= 0 −→ + Γµλν A =0
Ds ds ds
(this means that the vector field, say the vector, is equal in every point of the curve).

x(s)
A (x2 )
µ
Aµ (x1 )
Aµ (x0 )

Figure 9: Visualization of a vector field parallel transported along a curve.

Subsequently, another interesting object to study is the covariant derivative along the curve of the tangent
µ
vector uµ = dxds(s) , which corresponds to:

Duµ duµ dxλ ν duµ


= + Γµνλ u = + Γµνλ uλ uν
Ds ds ds ds
but, if we now recall the geodesic equations of a test particle, we find that those are equal to:
d2 xµ µ dxν dxλ Duµ
+ Γ νλ = 0 ←−−−−→ =0
ds2 ds ds Ds
so we can conclude that the geodesic is a curve whose tangent vector is parallel transported
(additional geometrical characterization of a geodesic). Further, in this way we have also proven that
geodesic equations satisfy perfectly the principle of General Relativity cause they form a tensor (which
2 µ
was indeed not an obvious fact cause both ∂dsx2 and Γµνλ are not tensors) so if we go in another FoR we
have:
Duµ D0 u0µ
= 0 ←−−−−−0→ =0 (135)
Ds x←→x Ds

39
9 Interaction with gravity
Finally, after describing all the maths elements we will need, we can head back talking of physics and we
can study the interactions between physical systems and gravitational fields. In particular, we want to
study how physical laws change when written in a FoR permeated by a gravitational fields.

9.1 Principle of General Covariance


A physical equation holds in a generic gravitational fields if:
• the equation holds in absence of gravity → holds in S.R.
• the equation is generally covariant, which basically means it has to be a tensor equation
So if a physical equation satisfies these two requirements, it is automatically valid in every FoR.
Observation. These two requirements substantially imply for an equation to be tensorial and to be valid
in the FoR in absence of gravity (the local inertial frame), in fact, if those two conditions are verified,
then by considering the Equivalence principle (from {ξ α } to {xα }) and the principle of General Relativity
(from {xα } to {x0α }) we end up with an equation valid in every FoR.

9.2 Point particle in a gravitational field


Suppose to study the simple system of a point particle in a gravitational field:

{ξ α }

{xα }
gµν

according to the Principle of General Covariance, first thing first we have to describe the behaviour of
the particle in the absence of gravity, and so in its local inertial frame where we have:
α
( α
ds = 0 with ũα = ds
dũ d dξ
ds (136)
ds2 = ηαβ dξ α dξ β −→ ηαβ ũα ũβ = 1

dxµ ∂xµ α
subsequently, we can introduce uµ = ds = ∂ξ α ũ in the FoR {xα }, where we have:
(
Duµ
=0
Ds (137)
ds = gµν dxµ dxν −→ gµν uµ uν = 1
2

This proposal satisfies both requirements: these are tensor equations and in absence of gravity gµν = ηµν ,
Γ → 0 and the covariant derivative just becomes the usual derivative Ds D
→ dsd
(we just obtain the SR’s
equations). So this set of equations fulfill the requirements of the Principle of General Covariance and
also highlights the only differences introduced by the gravitational field, which indeed are the metric
tensor and the covariant derivative (this kinda suggest the algorithm to describe a system in presence of
a gravitational field).

9.3 Electromagnetic charges in a gravitational field


Assume now we have to describe an electromagnetic field in presence of gravity, so basically we have to
rewrite the Maxwell’s equations coupled with a gravitational field. By following the previous approach,
first we just describe it as an usual E.M field generated by J α = (ρ, J)
~ in absence of gravity:
( αβ
∂ξ α = 4πJ
∂F β
(in-homogeneous equations)
(138)
Fβγ,α + Fγα,β + Fαβ,γ = 0 (homogeneous equations)

40
with F µν = −F νµ and Fµν = ηµα ηνβ F αβ .
Subsequently, when moving to a FoR permeated by a gravitational field {xα }, we just have to substitute
the metric and the derivatives: (
F αβ;α = 4πJ β
(139)
Fβγ;α + Fγα;β + Fαβ;γ = 0

with Fµν = gµα gνβ F αβ .


So this equations describe the interactions of our system with the gravitational field, interactions that
are hidden in the additional term proportional to Γ brought by the covariant derivatives. Now, recalling
that:
∂Aµν
Aµν;λ = − Γρµλ Aρν − Γρνλ Aµρ
∂xλ
and, to be more specific, the last particular covariant derivative we have analyzed, we can simplify the
homogeneous equations as:

Fβγ;α + Fγα;β + Fαβ;γ = 0 −→ Fβγ,α + Fγα,β + Fαβ,γ = 0 (140)

we obtain, once more, a combination of non-tensors that just transforms as a tensor. To conclude, we
can express the general solution as:
Fαβ = ∂α Aβ − ∂β Aα (141)
which remains the same also in curved space-time (pay attention to the indices cause now the operation
of lowering and rising is no longer pseudo-trivial, now we have to use gµν (xα ) which is a function of
space-time coordinates). So even in gravitational theory we can express the entire solution introducing a
4-potential and repeat the study of gauge invariances, not-uniqueness of the solution and so on. Further,
by working a little on the in-homogeneous equations:

∂F αβ
F αβ;α = + Γααµ F µβ + Γβαµ F αµ = but Γβαµ is symmetric and F αµ anti-symmetric
∂xα
1 ∂ √ αβ 
=√ gF = 4πJ β
g ∂xα

where we have also employed the identity of covariant divergences:


1 ∂ √
Γααµ = √ g
g ∂xµ

we obtain the equations in a very compact form (the gravity is present as g) in which we can also stick
the general solution, although it makes everything look much more complicated:
1 ∂ √ αβ  1 ∂ √ αµ βν
= g g g Fµν =

√ α
gF √ α
g ∂x g ∂x
(142)
1 ∂ √ αµ βν
=√ g g g (∂µ Aν − ∂ν Aµ ) = 4πJ β

g ∂xα

Nevertheless, even if finding the new equations coupled to the gravitational field is pretty straightforward,
it does not mean that solving them requires the same effort.
Observation. The description of an electromagnetic field coupled to gravity truly resembles the one of
a charged particle coupled to an electromagnetic field in which the interaction is typically introduced by
replacing ∂µ −→ ∂µ − iqAµ and so, for example, if we assume to have a scalar charged particle:

( + m2 )ϕ = 0 free field theory

we just have to replace  = ∂µ ∂ µ , and so we obtain:


 
(∂µ − iqAµ )η µν (∂ν − iqAν ) + m2 ϕ = 0

(where ϕ is the complex field that describes, for example, pions π + and π − ).

41
9.4 Energy-momentum tensor in presence of a gravitational field
As we have seen in sec.4.9, in SR framework the energy-momentum tensor T αβ satisfies the equations:

∂T αβ
= Gβ (we can get the EoM out of here)
∂xα
where Gβ represents the density of external forces acting on our system (so, in general, it is not a
αβ αβ
continuity equation) but if we define T αβ = Tpart + Text.field then our system is necessarily isolated and
G = 0, so, if we introduce the quantity p :
β α

dpα
Z
pα = d3 x T α0 −−−−−−−−→ =0
we have that dt
At this point, to couple our system to a gravitational field we just have to perform the usual substitutions,
so we find:
T αβ
T αβ;α = + Γααν T νβ + Γβαν T αν
∂xα
T αβ T νβ ∂ √
= + √ g + Γβαν T αν (143)
∂xα g ∂xν
1 ∂ √ αβ 
=√ gT + Γβαν T αν = 0
g ∂xα

from which we can extract the EoM of the system under the influence of the gravitational field. Moreover,
once more this is not a continuity equation due to the presence of the additional term Γβαν T αν and, as a
consequence, if we define some 4-vector:

Z
pα = d3 x gT α0

α
this is not conserved ∂p ∂t 6= 0 (this is not even a tensor actually). We might expect that this non-
αβ
conservation is caused by the non consideration of Tgrav. field but there is no such a thing as the energy-
momentum tensor for a gravitational field and the reason is very simple: if a tensor is zero in one frame
αβ
then it is zero in every frame, but Tgrav. field would be certainly zero in the local inertial frame, and so
it cannot be properly defined (Landau discuss the introduction of a pseudo-tensor, in addition we can
only define something resembling pα only if the space-time is flat at infinity, cause otherwise it is not
something local?).

9.5 Newtonian limit of geodesic equations (03-11-2020)

As said many times, GR was not necessary at an experimental level since Newtonian Theory was already
able to predict almost perfectly the orbits of planets up to a little discrepancy with the residual precession
of Mercury which, nevertheless, did not seemed to bother too much the scientific community. Anyway,
assuming that the theoretical Theory of General Relativity is correct, there must be a sort of Newtonian
limit such that the geodesic equations found in GR satisfy Newtonian Theory of Gravity. Therefore we
may hypothesize that for:
i) Typical velocities much smaller than the speed of light, v << c
ii) Weak gravitational fields, which imply: gµν = ηµν +hµν , where |hµν | << 1 is just a small deviation
iii) Stationary gravitational fields, so gravitational fields that do not depend on time ∂t gµν = 0 (this
condition is fundamental because Newton’s Theory is based on a Poisson’s potential while General
Relativity on retarded potentials, so fields variable in time would immediately cause discrepancies)
the two theories can be compared.
So assuming again we are studying the system of a test particle in a gravitational field. This is described
by the equations: ( 2 µ
µ dxα dxβ
ds2 + Γ αβ ds ds = 0
d x

ds2 = gµν dxµ dxν

42
but then, due to the definition of uµ = γ(c, ~v ) and the first assumption (v << c), in the upper equations
only terms with α and β equal zero are relevant, the others can be neglected, so we can write:
dxα dxβ  dx0 2 1  ∂g
0α ∂gα0 ∂g00 
Γµαβ ' Γµ00 with Γµ00 = g µα + −
ds ds ds 2 ∂x0 ∂x0 ∂xα
but partial derivatives with respect to ∂x0 are just time derivatives which also, according to the third
assumptions, can be ignored. Further, because of second assumption, the derivative of g00 reduces to be
the derivative of h00 (cause η is made of constants) and g µα simplify as η µα (we are just keeping first
order), so we end up with the equations:
1 ∂h00 d2 xµ  dx0 2
Γµ00 = − η µα α −−−−−→ + Γµ00 =0 (144)
2 ∂x ds 2 ds
So we can start solving them:
- for µ = 0:
d2 x0  dx0 2
+ Γ 0
00 =0
ds2 ds
but Γ0 00 is zero cause η is a diagonal matrix and if we consider the case of µ = α = 0 we obtain a
time derivative which, by hypothesis, has to be zero, so we end up with:
d2 x0
= 0 −−−−−→ x0 = s + const (145)
ds2
(which is a solution that makes sense as it implies we are parametrizing our particle’s motion as a
function of its proper time)
- for µ = i:
d2 xi  dx0 2
2
+ Γi 00 =0
ds ds
and again, Γi 00 has to be evaluated only for µ = α = i due to the presence of η, so, as η ii = −1,
we obtain:
d2 xi 1~  dx0 2
= − ∇ i h00
ds2 2 ds
Subsequently, for the spacial components, we have to change the parameter cause in Newtonian theory
it does not exist such a thing as the proper time, so:
dxi dxi cdt dxi dt d  dxi dt  d2 xi  dt 2 dxi d2 t d2 xi  dt 2
= = −→ = + =
ds cdt ds dt ds ds dt ds dt2 ds dt ds2 dt2 ds
(where in the last step we have just simplified the second derivatives d2 t/ds2 cause x0 = s + const) and
we eventually obtain:
d2 xi  dt 2 1~  dx0 2
= − ∇ i h00
dt2 ds 2 ds
d2 xi 1 2~
= − c ∇i h00 (146)
dt2 2
d2 ~x 1 ~
= − c2 ∇h 00 (vectorial form)
dt2 2
which has to be compared with the Newtonian equation:
d2 ~x 2
~ −−−−−−−−−−−→ c h00 = Φ + const
= − ∇Φ (147)
dt2 so we find that: 2
so we work out that h00 is related to the Newtonian potential Φ = −GM/r as:
2GM
h00 = − + const (148)
c2 r
Observation. By applying the principle of asymptotic flatness, we can also determine that const = 0.
In fact, at an asymptotic distance the effects of space-time curvature of a gravitational field should be null
(g00 = η00 ), so:
2GM
g00 = η00 − 2 + const −−−−−→ η00 + const = η00
c r r→∞

43
Nevertheless, before concluding, we must verify that |h00 | << 1 because if it is not our entire discussion
just has no sense; to do that we can compute the cases of the principal gravitational potentials:
(
−2M G rg ' 10−9 for the Earth
h00 = = =
2
c r r ' 10−6 for the Sun

(Although those values are really small, to develop a GPS system the space-time curvature caused by the
Earth has to be taken in consideration and so does the effect of Sun in Mercury’s orbit...)

9.6 Perfect fluids coupled with gravitational field


Given the definition of the energy-momentum tensor of a perfect fluid in the laboratory (for further
details sec.5):
T µν = −pη µν + (p + ρ)uµ uν
if it is an isolated system, we can extract its equations of motion by computing:

T µν,ν = 0 (continuity equations)

and that’s is basically all we need for a full description of our system (→ relativistic Euler’s equations of
motion). What we want to keep in account now is the self-gravity of the fluid. A common example
to consider when studying this effect are the stars in which self-gravity is in contrast with the pressure
of the fluid:

Figure 10: Schematization of gravity acting in a star.

Once more, by just applying the Principle of General Covariance we find the equations that reckon
gravity: (
T µν = −pg µν + (p + ρ)uµ uν
(149)
T µν;ν = 0
then, by first looking at the upper equations we can note that T µν is divided in two terms, and, when
comes the covariant derivative, we can compute it by splitting them as:
∂p
(pg µν );ν = g µν ν the derivative of g µν is zero and p is a scalar
  ∂x
(p + ρ)uµ uν = S µν;ν = S µν,ν + Γµνλ S λν + Γν νλ S µλ =

S µλ ∂ √
= S µν,ν + √ g + Γµνλ S λν =
g ∂xλ
1 ∂ √ µν 
=√ gS + Γµνλ S λν
g ∂xν

so we obtain:
∂p 1 ∂ √ µν 
T µν;ν = −g µν + √ gS + Γµνλ S λν =
∂xν g ∂xν
(150)
∂p 1 ∂ √ 
= −g µν ν + √ g(p + ρ)uµ ν
u + Γµνλ (p + ρ)uλ uν = 0
∂x g ∂xν

At this point, assuming a condition of hydrostatic equilibrium, we have that ui = 0, so the normal-
ization condition:
1 1
gµν uµ uν = 1 −→ g00 (u0 )2 = 1 in other words u0 = (g00 )− 2 = √
g00

44
with g00 greater than zero to make sense. Further, the equations reduce to:
∂p 1 ∂ √ 
T µν;ν = −g µν ν
+√ ν
g(p + ρ)uµ uν + Γµνλ (p + ρ)uλ uν =
∂x g ∂x
∂p 1 ∂ √ 
= g µν ν + √ 0
g(p + ρ)uµ u0 + Γµ00 (p + ρ)u0 u0 =
∂x g ∂x
∂p
= g µν ν + Γµ00 (p + ρ)g00
−1
=0
∂x
due to the simplification of the time derivative and all we have to compute is:
1  
Γµ00 = g µν g0ν,0 + gν0,0 − g00,ν
2
1
= − g µν g00,ν (cause the other term are time derivatives again)
2
so, we get:
∂p 1 −1
g µν − g µν g00,ν (p + ρ)g00 =0
∂xν 2
and finally, by multiplying for gµλ :
 ∂p 1 −1

gµλ g µν ν − g µν g00,ν (p + ρ)g00 =0
∂x 2
∂p 1 −1

δ νλ ν − δ νλ g00,ν (p + ρ)g00 =0
∂x 2
∂p 1 −1
− = g00,λ (p + ρ)g00
∂xλ 2
∂p ∂ √
− = (p + ρ) λ (ln g00 ) (151)
∂xλ ∂x
For λ = 0 the solution is trivial due to the presence, once more, of time-derivatives, while for λ = i we
have that:
∂p ∂ √
− i = (p + ρ) i (ln g00 )
∂x ∂x
subsequently, to solve these equations, it is necessary to introduce a state equation p = p(ρ) so we can
integrate the problem and ending up with the equation for hydrostatic equilibrium in GR:

Z
dp
= − ln g00 + const (152)
(p + ρ)

with dp = dρ
dp
dρ (to define the constant this time we have to consider boundary conditions, id est verify
that the gravitational field generated by the fluid itself match the gravitational field on its surface).
Observation. A similar result can be found in the Newtonian limit, in fact, for v << c the energy
densities E = mc2 + K are dominated by the matter energy and the pressure becomes irrelevant with
respect to ρ, further we have that g00 = η00 + h00 ' 1 with h00 = − 2M G
c2 r , but considering Poisson’s
equation:
Gm
∇2 Φ = 4πGρ −−→ Φ = − 2
c r
so we can rewrite h00 = 2Φ and find:
dp dp
−−−−→
p+ρ p<<ρ ρ
√ 1
− ln g00 = − ln (1 + 2Φ) −−−−−→ −Φ
2 2Φ<<1

from which we recover the equation for hydrostatic equilibrium in Newtonian Theory:

dp = −ρΦ = ρ~g dh

45
10 Curvature of space-time (04-11-2020)

As seen in the previous sections, a Gravitational field is described by ten potentials whose magnitude
defines a metric tensor gµν , moreover, thanks to the Equivalence Principle we have also said that the
effects of a Gravitational Field are equivalent to those of an inertial force, but then, how can we really
distinguish if we are actually in presence of a gravitational field or just in an accelerated FoR? Is it
sufficient to consider when the metric tensor gµν is different from the Minkowski metric ηµν to assure the
presence of gravity?

~g
gµν

(a) System in a gravitational field (b) Accelerated FoR

Example. Suppose to have a metric like:

1 0 0 0
 
0 −1 0 0
ds2 = dt2 − dr2 − r2 (dθ2 + sin2 θdϕ2 ) −→ gµν
0
=  6= ηµν
 
 0 0 −r2 0 
0 0 0 −r sin θ
2 2

so Γµνλ 6= 0 and a free particle would just follow the geodesic:

d2 xµ µ dxν dxλ
+ Γ νλ =0
ds2 ds ds
But this is clearly a bait, in fact these equations in spherical coordinates are just describing a straight line
in the Minkowski flat space-time:

x = r sin θ cos ϕ

y = r sin θ sin ϕ −→ ds2 = dt2 − dx2 − dy 2 − dz 2
z = r cos θ

0
and gµν is not describing any gravitational field. For how stupid this example was it however makes the
problem of recognizing if there is or not a Gravitational field very clear and serious.
Example. The Rindler coordinates (η, ξ, y, z).
Consider the Rindler metric:
e2aξ 0 0 0
 
 0 −e2aξ 0 0 
ds2 = e2aξ (dt2 − dξ 2 ) − dy 2 − dz 2 −→ gµν
0
=
 0
=6 ηµν (153)
0 −1 0 
0 0 0 −1

from which we derive an Affine connection Γµνλ 6= 0. Does this gµν represent a gravitational field? Again,
the answer is no, in fact this metric is obtained from the transformations:

t = a−1 eaξ sinh aµ





x = a−1 eaξ cosh aµ

−→ ds2 = dt2 − dx2 − dy 2 − dz 2


 y = y
z=z

so we are not dealing with a gravitational field but just with some unusual coordinates in which gµν just
describes the inertial forces. In particular, considering:

x2 − t2 = a−2 e2aξ

it can be easily noted that if ξ is a constant, then we are on a hyperbole in the Minkowski space-time:

46
t x=t
ξ = const

~x

which we know represents a uniform accelerated motion in SR (of acceleration ae−aξ ).


The solution to this problem is hidden in the Equivalence Principle itself, which states that only locally
it is possible to identify an inertial FoR, from which we eventually obtain a metric tensor gµν after
coordinates transformations ξ α → xα :

∂ξ α ∂ξ β
gµν (xα ) = ηαβ
∂xµ ∂xν
this detail will allow us to understand if we are in presence of a Gravitational field or if we are just in an
accelerated FoR in unusual coordinates. In addition, the actual entities that describes a field are not its
potentials, id est gµν , but the forces that it generates, so in the gravitational case we will have to consider
the Affine Connection Γλµν .
Observation. If we consider an E.M. field, this is not described by Aµ but by the Maxwell tensor Fµν
in fact, given a 4-potential Aµ (x) 6= const, we immediately compute:

Fµν = ∂µ Aν − ∂ν Aµ

cause the components of the electromagnetic field E~ and B


~ are in there. For the sake of clarity, assume
a potential Aµ = ∂µ Φ, in this case we would just have:

Fµν = ∂µ ∂ν Φ − ∂ν ∂µ Φ = 0

which means no electromagnetic field at all. This is a consequence of the non uniqueness of the solutions
of Maxwell’s equations, in fact if we consider a gauge transformation:

Aµ −→ A0 µ = Aµ − ∂µ ζ

~ and B
both the potentials generate the same electromagnetic field cause E ~ are gauge invariant.

10.1 The Riemann Tensor


Then, at this point, what actually assure me that, given a certain gµν (xα ), do not exist transformations
such:
∂xρ ∂xλ
gµν (xα ) −−−−−→ ηµν (ξ α ) = gρλ (xα )
α α
x −→ξ ∂ξ µ ∂ξ ν
µ ν
ρλ α ∂ξ ∂ξ
g µν (xα ) −− −−−→ η (ξ
µν α
) = g (x )
xα −→ξ α ∂xρ ∂xλ
The necessary and sufficient conditions for gµν (x) ≡ ηµν everywhere in the space-time are:
- the Riemann tensor has to be zero everywhere
- the matrix g µν must have 1 positive and 3 negative eigenvalues at a certain point X
where we have introduced the Riemann tensor, defined as:

∂Γλµν ∂Γλµσ
Rλµνσ = − + − Γη µν Γλση + Γη µσ Γλνη (154)
∂xσ ∂xν
If our system satisfies these conditions, then we are just in an accelerated FoR in flat space-time, so we
have to deal only with inertial forces. Thanks to this enunciate then, the problem of recognizing whether
a potential describes or not a gravitational field, is basically reduced to the computation of the Riemann
tensor.

47
Observation. Note how the Riemann tensor is a function of g, ∂g and ∂ 2 g: R = R(g, ∂g, ∂ 2 g) . This
is actually an essential requisite cause we know that surely exist a FoR where g = η and in such frame
∂g = 0 (cause the covariant derivative of η is zero), so we must compute also the second derivative. To
be more specific, first derivatives appear quadratically as (∂g)2 and this is the only tensor that can be
constructed considering g and its first and second derivative, linear in second derivative.
An important property coming from the definition of the Riemann tensor is related to the non commu-
tation relation of covariant derivatives, in fact:

Vµ;ν;λ − Vµ;λ;ν = Vα Rαµνλ


(155)
V µ;ν;λ − V µ;λ;ν = −V α Rµανλ

so, if we are in a flat space-time these relations are just zero cause of the Riemann tensor, but in other
cases we have that covariant derivatives do not commute. Recalling covariant derivatives along a curve,
there is another property related to the Riemann tensor, to be specific related to parallel transported
vector field. As we have seen in section 8.2, a vector field sµ (τ ) is called parallel transported if:

DSµ dSµ dxν


= 0 −→ − Γλµν Sλ = 0
Dτ dτ dτ
and if we integrate this function along a close curve, so in between two proper time instants τ0 , τ1 such
that xµ (τ0 ) = xµ (τ1 ), we obtain, in general, the equation:

1
Z
Sµ (τ1 ) − Sµ(τ0 ) = Rσµρν Sσ xρ dxν (156)
2
so we notice that in a curved space-time a parallel transported vector field does not necessarily return
equal after a complete cycle along a curve, thing which is instead always true in a flat structure where
Rσµρν = 0.

Figure 11: Example of a vector field parallel transported in a curved space-time.

10.2 Geodesic deviation & relative acceleration


Another very interesting aspect to distinguish if we are in a gravitational field or not at large scales, is
to consider the relative acceleration between two objects following their respective geodesic. In fact,
given that the sources of gravitational fields are point-like, then it is impossible (we would need an infinite
source) to have an homogeneous field, thus, any two points in space experience a relative acceleration to
each other, like in Fig12.

O0
O ~g

Figure 12: In-homogeneous effects of a gravitational field that cause a relative acceleration between objects
observable at large scales.

48
To discuss this phenomenum the Riemann tensor has again a central role but it is also useful to introduce
the concept of geodesic congruence. As we have seen in the previous sections, geodesics in GR are
described with the equations:
Duα
= uα;β uβ = 0
Ds
and given a geodesic it is possible to identify in space-time infinite identical copies of it, as shown in
fig.13; copies that can be parametrized as functions of the proper time s and an index vi : xµ (s, v) with
α
uα = ∂x ∂s (partial derivative cause now x is also a function of vi ).
α

∂xα
ηα uα = ∂s

s=0
xµ (s, v)
v1 v1 + δv v2 v3

Figure 13: Congruence in GR.

at this point it possible to define a motion vector of a geodesic:


∂xα ∂xα
ηα = δv = v α δv with v α = (157)
∂v ∂v
and basically, what we are going to do is computing the acceleration between two geodesics, in simple
words the covariant derivative of η.
In this set up the following two lemmas are verified:
∂uα ∂  ∂xα  ∂  ∂xα  ∂v α
i) = = =
∂v ∂v ∂s ∂s ∂v ∂s
ii) uα;β v β = v α;β uβ

Proof. Proof of the second lemma:


 ∂uα  ∂uα β
uα;β v β = + Γ α
βν u ν
v β
= v + Γαβν uν v β
∂xβ ∂xβ
 ∂v α  ∂v α β
v α;β uβ = + Γ α
βν v ν
u β
= u + Γαβν v ν uβ
∂xβ ∂xβ
but the last terms simplify and we just have to verify that:

∂uα β ∂uα ∂xβ ∂uα ∂v α β ∂v α ∂xβ ∂v α


v = = u = =
∂xβ ∂xβ ∂v ∂v ∂xβ ∂xβ ∂s ∂s
are equal, which is a direct consequence of the first lemma.
Then, as anticipated, we have to compute:
D2 ηα D2 (v α δv) D2 vα
= = δv =
Ds2 Ds2 Ds2
α
D Dv
 D  α β  
= δv = δv v ;β u = δv v α;β uβ uµ =
Ds Ds Ds ;µ
     
= δv u ;β v
α β
u = δv u ;β;µ v + u ;β v ;µ u = δv uα;β;µ v β uµ + uα;β v β;µ uµ =
µ α β α β µ

 
= δv u ;µ;β v u − uν Rανβµ v β uµ + uα;β (v β;µ uµ ) =
α β µ

 
= δv uα;µ;β v β uµ − uν Rανβµ v β uµ + uα;β (uβ;µ v µ ) =
 
= δv − uν Rανβµ v β uµ + uα;µ;β v β uµ + uα;µ uµ;β v β =
 
= δv − uν Rανβµ v β uµ + v β (uα;µ;β uµ + uα;µ uµ;β ) =
 
= δv − uν Rανβµ v β uµ + v β (uα;µ uµ );β

49
but in the small brackets we have the geodesic equation, that is equal to zero, so we just end up with:
D2 ηα
= −uν δv Rανβµ v β uµ = −uν Rανβµ η β uµ Geodesic deviation equations (158)
Ds2
Observation. The fact that in this result the Riemann tensor show up should not be a surprise cause it
consider second derivatives of g, in addition, we can note how this phenomenum only affects gravitational
fields, in fact in accelerated FoR we do not experience relative accelerations.

10.3 Algebraic properties of the Riemann tensor (10-11-2020)

Besides the Riemann tensor is characterized by 4 indices, thanks to its algebraic properties it is not a
256-parameter entity but it only counts 20 independent components. To discuss such properties and how
the components are taken down to 20, it is more useful to consider:
Rλµνσ = gλα Rαµνσ =
1  ∂ 2 gλν ∂ 2 gµν ∂ 2 gλσ ∂ 2 gµσ  
η η
 (159)
=− − − + − gηα Γ Γ α
− Γ Γα
2 ∂xσ ∂xµ ∂xσ ∂xλ ∂xµ ∂xν ∂xν ∂xλ νλ µσ σλ µν

from which we can check:


- Symmetry between λµ and νσ
Rλµνσ = Rνσλµ
- Anti-symmetry between (λ, µ) and (ν, σ)
Rλµνσ = −Rµλνσ = −Rλµσν = Rµλσν

- Cyclicity between the last three indices


Rλµνσ + Rλσµν + Rλνσµ = 0
if two indices are equal then the relation becomes trivial
further we have also to consider the following possible contraction of Rν µλσ :
- 1-3 contraction
Rν µνσ = g λν Rλµνσ = Rµσ Ricci tensor
with g λν
Rλµνσ = g νλ
Rνσλµ = Rσµ −→ The Ricci tensor is symmetric

- 1-4 contraction
Rσµνσ = g λσ Rλµνσ = −g λσ Rλµσν = −Rµν

- 1-2 contraction
Rµµνσ = g λµ Rλµνσ = 0
cause g is λµ symmetric while R is λµ anti-symmetric
- Ricci tensor contraction
Rµµ = g µα Rµα ≡ R Ricci scalar

[The Ricci tensor and scalar will have a central role in Einstein’s Theory]
To perform the estimation of the independents components of R we use the Potreov notation in which
basically the couples of indices (λ, µ) and (ν, σ) are replaced by A and B. By doing that the symmetry
relation can be reduced to RAB = RBA and we can do a first guess of k(k+1) 2 independent components,
where k is the number of possible combinations for A, which, assuming the most generic case of a N-
dimensional space-time, is k = N ×N . Subsequently, recalling that the indices in A and B are respectively
anti-symmetric, k reduces to k = N (N2−1) and to conclude, by considering the last property, cylicity, which
implies the subtraction of N (N −1)(N4!−2)(N −3) independent relation, the final relation is:
N (N −1) N (N −1)
( 2 + 1)
N (N − 1)(N − 2)(N − 3)
CN = 2

2 4!
1 2 N (N − 1)(N − 2)(N − 3) N 2 (N 2 − 1)
= (N − N )(N − N + 2) −
2
=
8 4! 12
and, since in our case N = 4 −→ CN = 20.

50
10.3.1 The Bianchi identities
Bianchi identities are differential relations defined as:

Rλµνρ;η + Rλµην;ρ + Rλµρη;ν = 0 (160)

(note the similarity with the Bianchi identities for the electromagnetic field Fµν;λ + Fνλ;µ + Fλµ;ν = 0)
which become very important for Einstein’s Theory when expressed in their contracted form, achievable
by multiplying for g λν :  
g λν Rλµνρ;η + Rλµην;ρ + Rλµρη;ν = 0

but then, remembering that the covariant derivative of g is zero, this relation can be rewritten as:
     
g λν Rλµνρ + g λν Rλµην + g λν Rλµρη =0
;η ;ρ ;ν

and now, employing the contraction rule, we get:

Rµρ;η − Rµη;ρ + Rν µρη;ν = 0


 
g µρ Rµρ;η − Rµη;ρ + Rν µρη;ν = 0 (second contraction)
     
g µρ Rµρ − g µρ Rµη + g µρ Rν µρη =0
;η ;ρ ;ν
R;η − Rρ η;ρ + g µρ Rν µρη;ν = 0

but:
 
g µρ Rν µρη;ν = g µρ g να Rαµρη;ν = −g µρ g να Rµαρη;ν = − g µρ g να Rµαρη =

 
= − g να Rαη = −g να Rαη;ν = −Rν η;ν = −Rρ η;ρ

so:
R;η − Rρ η;ρ + g µρ Rν µρη;ν = R;η − Rρ η;ρ − Rρ η;ρ = R;η − 2Rρ η;ρ = 0
and finally, we can simplify it in:
1 1  1 
Rρ η;ρ − R;η = Rρ η;ρ − δ ρη R;ρ = Rρ η − δ ρη R =0
2 2 2 ;ρ
 1 
g νη Rρ η − δ ρη R =0 (third contraction)
2 ;ρ
 1 
Rρν − g ρν R =0 contracted Bianchi identities (161)
2 ;ρ

with Rµν − 12 g µν R also called Einstein tensor due to its importance later in the Theory.

11 Einstein Field Equations


By field equations are intended those equations that, given the sources, describe a precise field.
Example. Again, an intuitive example are the Maxwell’s equations in Electromagnetism:
4π ν
Aν − ∂ ν (∂µAµ ) = J
c
(second order equations of the potentials).
The elements we should consider for an intuitive discussion are that also those equations have to be of
second order with respect to the gravitational potentials, id est gµν , and the fact that, in an appropriate
limit, they must reduce to the Poisson’s equation of Newtonian Theory :

∇2 Φ = 4πGρ with ρ = mass density

51
further, as mass dominates on any other forms of energy in non relativistic scenarios, the energy density
basically reduces to T00 = ρ and as in the Newtonian limit g00 ' 1 + 2Φ (sec.9.5), we can rewrite this
equation as:
∇2 g00 = 8πG T00 (162)
At this point, the Principle of General Relativity or, in other words, the fact that physical laws have to
be written in tensor formalism to be invariant in every FoR, suggests us to consider all forms of energy
Tαβ and not only the T00 component, but of course, this also implies to have a tensor also on the left
hand side:
Gαβ = 8πG Tαβ with G = G(g, ∂g, ∂ 2 g) (163)
Observation. In principle, G can be a function of higher orders of derivatives, but in theoretical physics
this is a very dangerous path cause it could lead to non-physical solution. An example is given by the
perturbative study of a charged particle emitting radiations while moving in an Electromagnetic field which
can be reduced to the equation:
...
 x (t) + ẍ(t) = F (x)
however its solutions are proportional to 1 which is a term not perturbative at all. So it is better to avoid
the employment of higher orders of derivatives and stop just at second order like in Maxwell’s Theory, in
addition, as this is our first guess, we can also assume just a linear contribution for ∂ 2 g.
To summarize then, Gαβ has to satisfy the following points:
- is a (0, 2) tensor
- is symmetric, cause Tαβ is symmetric too
- G αβ;β = 0, always cause Tαβ satisfies the same relation T αβ;β = 0
- G00 = ∇2 g00 = 8πGT00 in Newtonian limit
But as we have seen in the previous sections the only tensor proportional to g, ∂g and, linearly, to ∂ 2 g,
is the Riemann tensor, which, however, is a 4-index object, so it must be contracted −→ Gαβ must be
a combination of the Ricci tensor Rαβ and the Ricci scalar R:

c1 Rαβ + c2 gαβ R = 8πGTαβ

(Gαβ is not only proportional to Rαβ cause it won’t satisfy the third condition but, on the other hand, we
must construct a (0, 2) symmetric tensor, so R has to be multiplied for gαβ ). To define the two constants
we have to compute when:  
c1 Rµν + c2 g µν R =0

which is where the Bianchi identity enters in play, in fact, as we might recall:
 1  1
Rµν − g µν R = 0 −→ Rµν;ν = g µν R;ν
2 ;ν 2
so we obtain:
1
c1 g µν R;ν + c2 g µν R;ν = 0
2 (
c1 R;ν = 0
( + c2 )R;ν = 0 −→ (164)
2 c2 = − c21

But the first solution is not acceptable, in fact by considering:


 
g µν c1 Rµν + c2 gµν R = g µν 8πGTµν
c1 R + c2 4R = 8πGT νν

we find that R is directly related to T νν , so, imposing R;µ = 0 implies for T νν to be constant too, which
is generally not true (it is rather rare to have homogeneous energy densities). As a consequence, our
equations become:
1
c1 (Rµν − gµν R) = 8πGTµν (165)
2

52
where we can already see the appearance of the Einstein’s tensor. At this point there is only one step left
which is the evaluation of c1 by studying the Newtonian limit (weak and stationary gravitational
fields):
Gαβ = ∇2 g00 (166)
As anticipated, when we discuss a system in non relativistic conditions all the components of the energy
density Tµν are negligible with respect to T00 =⇒ |Tij | << T00 which, roughly, implies also that Gij ' 0,
so, in this approximation:
1
Rij ' gij R
2
but in this limit we also have that gµν = ηµν + hµν with |hµν | << 1, so the Ricci scalar can be expressed
as (we just keep the first order in the metric):
X X
R = Rµµ = R0 0 + Ri i = R00 − Rii
i i

then, by considering the case of i = j:


1 1
Rii = ηii R = − R
2 2
(in this precise step we are not using Einstein notation, these are only the diagonal elements)
we derives that: X1 3
R = R00 + R = R00 + R −→ R = −2R00
i
2 2
and subsequently, substituting this result, we find:
 1   1 
G00 = c1 R00 − g00 R = c1 R00 − R = 2c1 R00 =
2 X 2
= 2c1 R 0α0 = 2c1 (R0000 −
α
Ri0i0 )
i

now, we can compute R0000 and Ri0j0 by taking in account the linear part of the definition of the Riemann
tensor (as the terms with two Γ are proportional to |h00 |2 in Newtonian limit):
1  ∂ 2 gλν ∂ 2 gµν ∂ 2 gλσ ∂ 2 gµσ   
Rλµνσ = − − − + − gηα Γη νλ Γαµσ − Γησλ Γα
µν =
2 ∂x ∂x
σ µ σ
∂x ∂x λ µ
∂x ∂x ν ν
∂x ∂x λ

1  ∂ 2 gλν ∂ 2 gµν ∂ 2 gλσ ∂ 2 gµσ 


=− − − +
2 ∂xσ ∂xµ ∂xσ ∂xλ ∂xµ ∂xν ∂xν ∂xλ
and we obtain trivially that R0000 = 0 cause every term as its anti-symmetric, while, instead, to compute
Ri0j0 we must recall that in Newtonian limit we only consider stationary fields so all time derivatives are
just null:
1  ∂ 2 gij ∂ 2 g0j ∂ 2 gi0 ∂ 2 g00 
Ri0j0 = − − − + =
2 ∂x0 ∂x0 ∂x0 ∂xi ∂x0 ∂xj ∂xi ∂xj
1 ∂ 2 g00
=−
2 ∂xi ∂xj
and we end up with:
X X ∂ 2 g00
G00 = 2c1 (− Ri0i0 ) = c1 i ∂xi
= c1 ∇2 g00 −→ c1 = 1 (167)
i i
∂x

in Cartesian coordinates.
Finally, we can conclude with the definitions of the Einstein’s field equations:
1
Rαβ − gαβ R = 8πG Tαβ (168)
2
which is a system of ten coupled non-linear differential equations (non linearity is hidden in the contrac-
tions of the Riemann tensor, which contains terms proportional to ΓΓ −→ (∂g)2 ). In empty spaces the
energy density tensor is null (there are no sources of gravity) and Einstein’s equations become:
1
Rαβ − gαβ R = 0 (169)
2

53
if we then reduce the relation by multiplying for g αβ we find that:
 1 
g αβ Rαβ − gαβ R = 0
2
1
R − 4R = 0 −→ R = 0
2
so in the vacuum the Ricci scalar is zero and, as a consequence, also Rαβ = 0 (consider equation (168)),
however this do not implies for the Riemann tensor to be also zero (Rαβ = 0 is a set of ten equations
while the Riemann tensor counts a total of twenty independent components).
Observation. If, instead of a space of dimension N = 4, we consider N = 3 this last clarification is not
true and Rαβ = 0 (6 equations) −→ Rλµνρ = 0 (CN = N 2 (N 2 − 1)/12) = 6 independent components).

11.1 Algebraic analysis of the Einstein’s equation (11-11-2020)

Given the expressions of the Einstein’s equations, it seems possible to define uniquely the metric tensor
of a gravitational field but actually that’s no true. In fact, by noting that these equations are functions of
the Einstein tensor, we have also to consider the additional four differential relations given by the Bianchi
identities, so the first ten algebraic equations are not functionally independent, and so the total number
of independent algebraic equations is reduced to six. This implies that 4 dofs are not bounded, which
is, actually, a necessary requisite cause given a metric tensor gµν (x) we can always perform a coordinate
transformation (4 functions) to obtain a new metric g 0 µν (x0 ) completely equivalent to the first. The
theory is invariant under coordinate transformation so there must be a sort of gauge symmetry
in the theory which is re-parametrization invariant.
Example. Consider the case of vacuum in which we have gµν = ηµν (there is not a gravitational field),
so:
ds2 = c2 dt2 − dx2 − dy 2 − dz 2
if we move in spherical coordinates, the same metric becomes:

ds2 = c2 dt2 − dr2 − r2 (dθ2 + sin2 θdϕ2 )

and we can also invent unusual coordinates:


v−u
ds2 = dudv − ( )(dθ2 + sin2 θdϕ2 )
2
we are always describing the same space-time, id est the trivial case of no gravity, only in different
coordinates.
Example. This coordinates invariance was also true for Maxwell’s equations:
4π µ
Aµ − ∂ µ (∂ν Aν ) = J
c
and if we consider the divergence of this equations (just like the Bianchi identities):

∂µ (Aµ − ∂ µ (∂ν Aν )) = 0

once more the total number of independent equations is reduced, passing from 4 to 3; in fact, as we well
know, we cannot define uniquely Aµ cause we can always perform a transformation like:

Aα −→ Aα − ∂ α (ζ)

we can only define Aµ up to some scalar fields. For this reason we introduce gauge fixing to extract a
precise solution, in the particular case of Electromagnetism we consider ∂µ Aµ = 0.

The same thing happens for the Einstein’s equations for which we must fix the FoR to find a unique
solution, basically breaking the re-parametrization invariance. That was the case of Newtonian limit
when we have fixed gµν = ηµν + hµν .

54
11.2 The Cosmological constant
So Einstein’s equations were able to describe the gravitational field inside the source itself, where Tµν 6= 0,
and even in its proximity. When Einstein formulated the theory of General Relativity he also proposed
three tests to verify it: the precession of Mercury perihelion, the deflection of light rays in a gravitational
field and the gravitational redshift. Further, he even tried to apply its theory to the whole Universe, he
was convinced that the whole Universe was ruled by GR−→ Cosmological model. At that time the
main conception about Cosmology was that of a static Universe but considering Einstein’s equations
it is not possible to find static solutions; for this reason he tried to modify it in order to get appropriate
results. In particular he tried to change the left-handed side Gµν , in according to the requisites we already
know, it had to be a (0, 2) tensor, symmetric, with G µν;ν = 0 and in agreement with Newton’s law in
Newtonian limit. The only way to do that was to add a term proportional to the metric, the so called,
Cosmological term:
1
Rµν − gµν R − Λgµν = 8πG Tµν (170)
2
which he thought not to be a term with local effects (the save Newtonian limit) but only affects Cosmo-
logical scale, in fact, by adding this term he was indeed able to find a static solution. Nevertheless, the
introduction of this term was not in agreement with the idea that Gravitational fields were generated by
matter, and this can be noted if the two versions are compared in vacuum:
1
Rµν − gµν R = 0 −→ gµν = ηµν
2
1
Rµν − gµν R − Λgµν = 0 −→ gµν = ηµν is no longer solution
2
in the first case the equations become trivial (if gµν = ηµν the Riemann tensor is zero, and so are the
Ricci tensor and scalar) but in the second case if gµν = ηµν we obtain −Λgµν = 0 which is not true, so
we must have a curved space-time without sources. And in particular, if we consider its trace:
 1 
g µν Rµν − gµν R − Λgµν = 0
2
1
R − 4R − 4Λ = 0
2
R = −4Λ ←→ Rµν = −Λgµν de Sitter space-time

the curvature in absence of matter is even clearer. Only later in time, with the discoveries of Hubble,
the science community has been convinced of the non-static nature of the Universe, time at which, the
implementation of the Cosmological constant in Einstein’s equations lost its physical interest, replaced by
pure mathematical studies. Nonetheless the theory was not complete at all, in fact, Hubble’s observations
described an expanding Universe (non-static), but the most relevant detail was that the expansion is
accelerated, while, according to Einstein’s equations, the expansion is decelerated due to self-gravity, so
this equations are unable to describe experimental observations. In order to give an explanation to this
phenomenum there still must be something missing (hypothesis of non linearity in the second derivatives,
...) but, in particular, we might assume there is some energy-momentum content we are not taking
into account, so the problem can be in the right-handed side. The most convenient way to develop this
hypothesis is to re-introduce our Cosmological constant, but this time on the right side:
1
Rµν − gµν R = 8πG Tµν + Λgµν (171)
2
and interpret it as a sort of energy-momentum tensor of the vacuum, the Dark Energy cause we do
not know what it is (also from Quantum Mechanics we have hints vacuum energy).
Example. Consider the case of a perfect fluid:

Tµν = (p + ρ)uµ uν − pgµν


with: p = p(ρ) −→ p = −ρ = const (vacuum state equation)

then we just obtain a sort of energy-momentum tensor of the vacuum:

Tµν = ρ gµν

55
Observation. Besides considerations about what set of equations is the best from a Cosmological point of
view, id est if we prefer to introduce something we don’t know like Dark energy or to modify the left hand
side, Einstein’s Theory is actually the best theory at astrophysical scale, which can be discussed regardless
Cosmological elements.
One last aspect we are going to discuss is the following: Einstein’s equations describes the curvature of
space-time caused by matter, but, considering Bianchi identities, these also tells us how matter moves
1
Rµν − gµν R = 8πG Tµν
2
1 µν
(R − g R);ν = 0 ←→ (T µν );ν = 0
µν
2
Example. Assuming again the case of a perfect fluid:

T µν = (ρ + p)uµ uν − pg µν with p = 0 (dust model)


T µν = ρ uµ uν

in the dust case our equations of motion are:

T µν;ν = (ρ uµ uν ) ν = [(ρ uν )uµ ] ν =


= uµ (ρ uν );ν + ρ uν uµ;ν = 0
uµ T µν;ν = uµ uµ (ρ uν );ν + ρ uν uµ uµ;ν = 0 (we contract the equations)

but, remembering that:

ds2 = gµν dxµ dxν −→ gµν uµ uν = uµ uµ = 1 and:


(uµ uµ );ν = uµ;ν uµ + uµ uµ;ν = uµ;ν uµ + g µρ uρ (gµλ uλ )ν = uµ;ν uµ + g µρ uρ gµλ uλ;ν =
= uµ;ν uµ + δ ρλ uρ uλ;ν = 2uµ uµν = 0 −→ uµ uµν = 0

so our equations become:


uµ T µν;ν = (ρ uν );ν = 0
but if this is zero, then:
Duµ
T µν;ν = uµ (ρ uν );ν + ρ uν uµ;ν = ρ uν uµ;ν = 0 −→ uµ;ν uν = =0
Ds
so our fluid is moving along the geodesic.
To conclude then, in Einstein’s equations matter tells space-time how to curve and space-time
tells matter how to move.

12 The Schwarzschild problem


The problem regards the definition of the gravitational field outside a symmetrical spherical source (a
star), so where Tαβ and Rαβ are both zero:

Tαβ = 0
Rαβ = 0

Rsource

Figure 14: Schwarzschild’s problem set up.

To study this problem we cannot ignore the symmetry offered by the problem itself when choosing the
coordinates, which, of course, will be the spherical ones. What we expect is an homogeneous gravitational
field over the surface of a given sphere centered in the origin of the source, a surface where the gravitational
field is invariant under rotations and so described by a metric tensor invariant too. Naively, by thinking

56
at the elements invariant under rotations in Cartesian coordinates we can obtain their counterparts in
spherical ones:

~x · ~x −→ r2
~x · d~x −→ rdr
d~x · d~x −→ dr2 + r2 (dθ2 + sin2 θdϕ2 )

from which we can extract the most general definition of a line element invariant under rotations expressed
in terms of four unknown functions:

ds2 (t, x, y, z) = gµν dxµ dxν from which:


ds (t, r, θ, ϕ) =
2
g 0µν dx0µ dx0ν = g 0µν (dt, dr, rdθ, r sin θdϕ)(dt, dr, rdθ, r sin θdϕ)
−→ ds = −α(r, t)dr − β(r, t)(dθ + sin θdϕ ) + γ(r, t)dt + δ(r, t)drdt
2 2 2 2 2 2
(172)

Note how in the generic definition of the invariant under rotations do not appear all the mixed guys
drdθ, dtdϕ and so on, cause these are certainly not invariant, basically we have to consider only distances
on the surface of a sphere (r2 dθ2 + r2 sin2 θdϕ2 ) or in the directions perpendicular to it. In this setup we
can still perform certain coordinate transformations cause we have not completely set the FoR, but keep
in mind we must leave ds2 invariant:
- (θ, ϕ) −→ (θ̃, ϕ̃) such that dθ2 + r2 sin2 θdϕ2 is invariant
- (r, t) −→ (r̃, t̃)
and we can take advantage of this freedom to simplify the metric.
In particular, considering (for one moment we just ignore the signs which are a consequence of mostly-
negative metric):
ds2 = β(r, t)(dθ2 + sin2 θdϕ2 ) = ĝab dxa dxb
t,r=const

where a and b can only assume two values (dx1 = dθ and dx2 = dϕ) we basically obtain the line element
of a 2-sphere multiplied for a constant and the area of this surface can be computed using the formula:
Z p Z
A= ĝ dθdϕ = β(r, t) sin θdθdϕ = 4πβ(r, t) (173)

At this point, by performing the transformation:


(      1  −1
r̃2 = β(r, t) dr̃ = 1 1 ∂β dr + ∂β dt
dr = 2β 2 dr̃ − ∂β
∂t dt̃
∂β

∂r ∂t ∂r
⇒ 2β 2 and inverses:
t̃ = t dt̃ = dt dt = dt̃

(the easiest solution is to consider β = const as dr would be null, but this is actually a non physical
condition because we expect space-time flatness for r −→ ∞, so, asymptotically, β −→ r2 ) we move to a
FoR where:
ds2 = −α̃(r̃, t̃)dr̃2 + γ̃(r̃, t̃)dt̃2 + δ̃(r̃, t̃)dr̃dt̃ − r̃2 (dθ2 + sin2 θdϕ2 ) (174)
q
and the invariant 2-sphere for t̃ and r̃ = const, has A = 4πr̃2 with r̃ = 4π . (Substantially, we are
A

progressively fixing our FoR). Now, how do we chose the time coordinate? One choice, which has
roots purely mathematical, is to define it in a way such that the off-diagonal element is cancelled and our
metric tensor is represented by a completely diagonal matrix:

dr̃ = dr0
( (
r0 = r̃ dr0 = dr̃
⇒ and inverses:   −1
t0 = Φ(r̃, t̃) dt0 = ∂Φ ∂ r̃ dr̃ + ∂Φ
∂ t̃
d t̃  d t̃ = dt 0
− ∂Φ
∂ r̃ dr̃ ∂Φ
∂ t̃
 ∂Φ −2 h ∂Φ ∂Φ i
−→ ds2 = [· · · ] + δ 0 (r0 , t0 ) − 2 γ 0 (r0 , t0 ) dr0 dt0
∂ t̃ ∂ t̃ ∂ r̃
and given some δ 0 and γ 0 we can chose a Φ such that the content of the square brackets is null. By doing
that, we have basically proved that a generic spherical metric can be always simplified to a more compact
form with just two unknown functions:

ds2 = eν(r,t) dt2 − eµ(r,t) dr2 − r2 (dθ2 + sin2 θdϕ2 ) (175)

57
(for mathematical simplicity the unknown functions have been re-parametrized as exponentials) and as a
consequence, Einstein’s equations will only be differential equations of two unknown functions. This kind
of coordinates in which the line element is completely diagonal, are called Schwarzschild coordinates.
Once we get to this point we can eventually solve the entire problem by computing:

gµν −→ Γαµν −→ Rα βγρ −→ Rµν = 0 (176)

Starting from the definition of the elements of the metric tensor:

gtt = g00 = eν grr = g11 = −eµ gθθ = g22 = −r2 gϕϕ = g33 = −r2 sin2 θ

we can define the the elements of the Affine connection considering the formula:
 
Γαµν = g αρ gµρ,ν − gνµ,ρ + gρν,µ

(usually the technique is to fix α and take advantage of the symmetric property for µν) and subsequently:
∂Γαµν ∂Γαµσ
Rλ µνσ = − − Γη µν Γαση + Γη µσ Γαµη
∂xσ ∂xν

12.1 Solution for stationary space-time (17-11-2020)

To solve this problem, Schwarzschild also made an ansatz and assumed that ν = ν(r) and µ = µ(r) (this
assumption of stationarity only concerns space-time, we are not saying anything about the source which
can be, for example, a pulsating star), according to this hypothesis it is possible to work out the following
non trivial relations for the Ricci scalar:
 ν 00 ν0 v0 
R00 = eν−µ + − (µ0 − ν 0 )
2 r 4
ν 00 µ0 ν0 0
R11 = − + − (ν − µ0 )
2 r 4
R33  r 0 
R22 = = −e −µ
1 − e µ
+ (ν − µ0
)
sin2 θ 2
(the 0 s stand for derivatives with respect to r) and to solve the problem we have to find the functions
µ(r) and ν(r) such that these are equal to zero as predicted by the vacuum Einstein’s equations:
 00
ν0 v0
 2 + r0 − 4 (µ − ν ) = 0
 ν 0 0
00
ν0
2 − r − 4 (µ − ν ) = 0
ν µ 0 0

1 − eµ + 2r (ν 0 − µ0 ) = 0

Observation. We got three equations for two unknowns which should suggest some kind of dependence
between the relations: such dependence is actually described by the Bianchi identities, as one might
remember. In fact, it can be checked that out of the four differential equations only one is non-trivial,
that one defines the dependence in this last set of relations.
0 0
To simplify this system we consider eq1 − eq2 from which we obtain νr + µr = 0 −→ ν 0 + µ0 = 0 (we can
get rid of r−1 cause we are considering the gravitational field outside a source, so surely r > Rsource ) and
we can immediately integrate it, finding that ν + µ = λ, at this point we can rewrite the system like:
(
ν+µ=λ
1 − eµ − rµ0 = 0

then we have to solve the differential equation:

1 − eµ − rµ0 = 0 multiply for e−µ


e−µ − 1 − rµ0 e−µ = 0
e−µ − rµ0 e−µ = 1
(re−µ )0 = 1 now we can integrate
2m
re−µ = r + const −→ e−µ = 1 −
r

58
where m is a length. Once µ is found we can explicit ν as:
2m
eν+µ = eλ −→ eν = eλ e−µ = eλ (1 − )
r
and replace these expression in ds2 :
2m 2 2m −1 2
ds2 = eλ (1 − )dt − (1 − ) dr − r2 (dθ2 + sin2 θdϕ2 )
r r
λ
Now, the variable λ can be absorbed by rescaling time as t −→ t̃ = e 2 t , which basically means we can
set it to zero without loss of generality, and so we finally get the so called Schwarzschield solution:
2m 2 2m −1 2
ds2 = (1 − )dt − (1 − ) dr − r2 (dθ2 + sin2 θdϕ2 ) (177)
r r
The Schwarzschild solution is probably one of the most known results obtained in General Relativity.

12.2 Solution for non-stationary space-time


We will now go through the solution of the problem relaxing the requisite of a static space-time, so,
basically, going back to ν = ν(t, r) and µ = µ(t, r) with the non-static metric:

ds2 = eν(r,t) dt2 − eµ(r,t) dr2 − r2 (dθ2 + sin2 θdϕ2 )

Working out the various steps (computations of gµν , Γλµν and Rλ µνσ ) leads to the following non-zero
relations for the Ricci scalar:
1 1 1 
R00 = (· · · ) − µ̈ + µ̇2 − µ̇ν̇ = 0
2 2 2
1 1 2 1 
R11 = (· · · ) + µ̈ + µ̇ − µ̇ν̇ = 0
2 2 2
R33
R22 = = (· · · )+ =0
sin2 θ
µ̇
R01 = = 0
r
(the parenthesis with dots are substitutes for the previous results while dotted functions are just time
derivatives) in which we can note and additional non trivial term, R01 , that immediately tells us µ is
constant in time µ(t, r) −→ µ(r). As a consequence, all the extra terms entering in those relations just
vanish and considering eq1 − eq2 , we get ν 0 + µ0 = 0 −→ ν + µ = λ(t) (cause, in principle, ν is still a
function of time) and so, a very similar system:
(
ν + µ = λ(t)
1 − eµ − rµ0 = 0

At this point, plugging the solution of the differential equation e−µ = 1 − 2m


r and the relation:
2m
eν+µ = eλ(t) −→ eν(t) = eλ(t) (1 − )
r
in the expression of the line element we find:
2m 2 2m −1 2
ds2 = eλ(t) (1 − )dt − (1 − ) dr − r2 (dθ2 + sin2 θdϕ2 )
r r
λ(t)
but, as before, we can perform a coordinate transformation introducing a new time t −→ t̂ = dte 2 ,
R

and so, again, without loss of generality, we can set λ = 0 and obtain again the Schwarzschild solution:
2m 2 2m −1 2
ds2 = (1 − )dt − (1 − ) dr − r2 (dθ2 + sin2 θdϕ2 ) (178)
r r
Observation. This result is actually very interesting cause, although the assumption of a stationary
space-time has been relaxed, we find the Schwarzschild solution, which means that, even in presence of a
pulsating, collapsing or expanding source, the gravitational field is still stationary. Such conclusion
usually goes under the Birkhoff theorem.

59
The last point is to define the characteristic length m, which is usually done by imposing continuity
between the solutions of Einstein’s equations inside and outside the source, however, it does exist a
shortcut, in fact m is the last parameter left in our solution and so it must contain information about
the source. In particular, recalling the Newtonian limit of geodesic equation, we find:
2GM 2M G
g00 ' 1 + 2Φ = 1 − −→ 2m = = rS (179)
c2 r c2
where rS is the Schwarzschild radius. Using this we can get the final form of the Schwarzschild solution
valid outside the source, so for r > Rsource :
rS 2 rS
ds2 = (1 − )dt − (1 − )−1 dr2 − r2 (dθ2 + sin2 θdϕ2 ) (180)
r r
(inside the source we have to solve the Einstein’s equations for Tµν 6= 0).
Observation. Singularities of Schwarzschild solution
By looking at its final form it can be easily noted that the Schwarzschild solution suffers of three possible
singularities:
- r = 0, which has to be excluded for hypothesis: r > Rsource
- θ = 0, inherits from spherical coordinates
- r = rS
this last one is particularly problematic (our equation fails, argument of GR2) but it is a very rare case,
in fact we have, for example, that the Schwarzshild radius of the Sun is rS = 1.5Km which is infinitely
smaller than the actual radius of the Sun, outside which our solution is valid. Same thing for the Earth
which has a rS = 0.5cm.

12.3 Space and time coordinates in Schwarzschild solution


Remember that r is not the the distance between the origin but it stands for r̃, which was defined a little
differently. This aspect can be seen pretty easily by fixing t, θ and ϕ and considering two points r1 and
r2 :

r2
r1

Their distance, in our system of coordinates can by obtained from the Schwarzschild solution, and since
in our case it simplifies to:
2m −1 2
ds2 = (1 − ) dr
r
we get: Z r2
dr
l= q
r1 1 − 2m
r

so, except asymptotically for r → ∞, our distance does not correspond to r2 − r1 .


In the same way also time could use a deeper discussion. As we have pointed out at the beginning, for
r −→ ∞ we expect space-time flatness, in fact:
rS 2 rS
ds2 = (1 − )dt − (1 − )−1 dr2 − r2 (dθ2 + sin2 θdϕ2 ) −→ ds2 = dt2 − dr2 − r2 (dθ2 + sin2 θdϕ2 )
r r

60
so at an infinite distance from our gravitational source, SR is basically restored, and we are in an inertial
FoR described by the Minkowski metric (in this case in spherical coordinates). Now, if we get closer to
the source fixing r < ∞, θ and ϕ, the proper time in our FoR is described by:
2m 2
ds2 = (1 − )dt
r
where dt is the time measurement in an inertial FoR at r = ∞. This result basically tells us that a
gravitational field has direct effects on time flow (it slow down time), further discussions will clarify this
particular aspect and will show the importance of having a static metric, or, in other words, a metric
time-invariant. In particular, this last property of staticity is absolutely not obvious but emerges clearly
in spherical coordinates (in other coordinates it can be hidden in more complicated equations).

12.4 Geodesic equations in Schwarzschild space-time


As one might recall geodesics are usually described by the equations:
d2 xµ dxα dxβ Duµ dxµ
2
+ Γµ αβ = = uµ;ν uν = 0 with uµ =
ds ds ds Ds ds
so, in our case, to find the trajectory of a test particle we have to compute Γµ αβ in Schwarzschild space-
time and then plug it in these equations. Considering eν = (1 − 2m r ) and indicating with dotted functions
the derivatives with respect to s and with 0 s those with respect to r, the extended result is:

r̈ − 2 ν ṙ − re θ̇ − re sin θϕ̇ + 2 e ν ṫ = 0


1 0 2 1 2ν 0 2
 ν 2 ν 2 2

θ̈ + 2 ṙθ̇ − sin θ cos θϕ̇2 = 0

r (181)
ϕ̈ + 2r ṙϕ̇ + 2sin
 θ θ̇ ϕ̇ = 0
cos θ

ẗ + ν 0 ṙṫ = 0

which seems very challenging from an analytical point of view. Nevertheless, if we take the plane:
π
θ= (182)
2
and we consider a point P such that:

θ̇ =0
P

on this kind of trajectory, the second equation just reduces to


...
θ̈ = 0 −→ θ =0
P P

thus the entire trajectory relies on the plane θ = π/2 and the set of equations can be largely simplified
(thanks to rotational invariance we are actually not losing of generality, in fact, in spherical coordinates,
all planes are basically equivalent). At this point, starting from the equations:
uµ;β uβ = 0 multiply for gµα
gµα uµ;β uβ = (gµα uµ );β uβ = uα;β uβ = 0
we can get the geodesic equations in a completely equivalent form by computing the covariant derivatives:
1  
uα,β uβ = Γλαβ uλ uβ = g λσ gασ,β + gσβ,α − gβα,σ uλ uβ =
2
1 
= gασ,β + gσβ,α − gβα,σ uσ uβ =
2
1
= gσβ,α uσ uβ
2

61
from which we derives that:
∂uα dxβ duα 1
uα,β uβ = −−−−−→ = gσβ,α uσ uβ (183)
∂xβ ds ds 2
so, if we have a metric tensor independent from some coordinate xα :

∂gσβ (xµ )
gσβ,α = =0
∂xα
it follows, for the covariant component uα , to be constant. In our case, recalling the Schwarzschild
metric:
2m 2 2m −1 2
ds2 = (1 − )dt − (1 − ) dr − r2 (dθ2 + sin2 θdϕ2 )
r r
we immediately note it is not proportional to t and ϕ, so, by applying this last theorem, we find that
u0 = Ẽ and u3 = L̃ are constants for the geodesics in this space-time (in particular Ẽ is an a-dimensional
quantity that is conserved in time due to the stationarity of the metric tensor, the energy per unit mass,
while L̃ is the angular momentum). Out of this we immediately get two equations:

dx0 2m −1
ṫ = = u0 = g 0α uα −−−−−−−−−→ g 00 u0 = (1 − ) Ẽ (184)
ds g is diagonal r
dx3 L̃ L̃ π
ϕ̇ = = u3 = g 3α uα −−−−−−−−−→ g 33 u3 = − 2 2 = − 2 (cause θ = ) (185)
ds g is diagonal r sin θ r 2
Finally, the last differential equation to completely integrate the system can be found from the line
element, and in particular, from the fact that the tangent vector has to be normalized to one (which is a
totally generic requisite for massive bodies):

ds2 = gµν dxµ dxν −→ 1 = gµν uµ uν

by performing this summation we get (solution on the equatorial plane θ = π/2):

1 = g00 (u0 )2 + g11 (u1 )2 + g22 (u2 )2 + g33 (u3 )2


2m  2m −1 2 2m −1 2  L̃ 2
= (1 − ) (1 − ) Ẽ − (1 − ) ṙ − r2 sin2 θ − 2 =
r r r r
2m −1 2 2m −1 2 L̃2
= (1 − ) Ẽ − (1 − ) ṙ − 2
r r r
Once we solved this, it sufficient to plug r(s) in the expression of ṫ and ϕ̇ and integrate them to define
entirely the motion of our particle.
Observation. In the case of massless particles, it is necessary to re-parametrize the whole problem cause
we no longer have something like the proper time and solve this last differential equation equals to zero
instead of one.

13 Discussions over Einstein’s tests for GR (18-11-2020)

13.1 Planets’ orbits in Schwarzschild space-time


Recalling the three proofs Einstein proposed to verify its theory, it is possible to prove the one related
to Mercury’s orbit using the previous approach, in fact, as its mass is negligible compared with the mass
of the Sun, its motion can approximated to that of a test particle around a source and, by extracting
r = r(ϕ) from:
dr dr dϕ L̃ dr
ṙ = = =− 2
ds dϕ ds r dϕ
we can make a comparison with experimental data. According to this transformation, we can rewrite our
differential equation as:

2m −1 2 2m −1  L̃ 2  dr 2 L̃2
(1 − ) Ẽ − (1 − ) − 2 − 2 =1 (186)
r r r dϕ r

62
and, if we consider the relation:
d 1 1  dr   d 1 2 1  dr 2
=− 2 −→ = 4
dϕ r r dϕ dϕ r r dϕ
while performing some computations:
2m −1 2 2m −1 2 h 1  dr 2 1 2m i
(1 − ) Ẽ − (1 − ) L̃ + (1 − ) =1
r r r4 dϕ r2 r
h 1  dr 2 1 2m i 2m
Ẽ 2 − L̃2 4 + 2 (1 − ) = (1 − )
r dϕ r r r
2m h 1  dr 2 1 2m i
Ẽ 2 − (1 − ) = L̃2 4 + 2 (1 − )
r r dϕ r r
h 2m i 1  dr 2 1 2m
L̃−2 Ẽ 2 − (1 − ) = 4 + 2 (1 − )
r r dϕ r r
h 2m i  d 1 2 1 2m
L̃−2 Ẽ 2 − (1 − ) = + 2 (1 − ) but 2m = 2GM
r dϕ r r r
Ẽ 2 − 1 2GM 2GM  d 1 2 1
+ + 3
= + 2 differentiation with respect to ϕ
L̃2 2
L̃ r r dϕ r r
2GM  dr  2GM  dr   d 1  d2  1  1  dr 
− 2 2 −3 4 =2 − 2
L̃ r dϕ r dϕ dϕ r dϕ2 r r3 dϕ
2GM dr   2GM dr  1 dr d
  2  
1 1  dr 
+ 3 = 2 + 2
L̃2 r2 dϕ r4 dϕ r2 dϕ dϕ2 r r3 dϕ
we eventually get the differential equation for the orbit according to General Relativity:

d2  1  1 GM 3GM
2
+ = 2 + (187)
dϕ r r L̃ r2
which is a non-linear in-homogeneous differential equation of second order. Before going through its
solution, it is useful to compare it with the counterpart given by Newton’s theory (always in spherical
coordinate, obtained imposing conservation of energy and angular momentum):
1 h dr 2  dϕ 2 i Gm M
0
m0 + r2 − =E
2 dt dt r

m0 r2 =L
dt
At this point, if we look again for r = r(ϕ), we can work out the Newton-Binet equation:

d2  1  1 GM
+ = 2 (188)
dϕ2 r r L̃
which, this time, is a linear in-homogeneous differential equation that differs from the GR proposal for
the quadratic term 3GMr 2 . In GR then, we basically have a correcting term of the Newtonian potential
and its influence can be weighted computing:

3GM r−2 GM rS
' ' (189)
r−1 r r
and, in the particular case of Mercury’s orbit around the Sun, this term is proportional to 10−7 while,
for farther planets, this ratio obviously drops even more. As a consequence, we expect the most relevant
correction to pop out from the study of Mercury’s orbit and, due to its smallness, we can solve the GR
differential equation with a perturbative approach. Taken the solution of the Newton-Binet equation,
the Newtonian orbits’ equation, id est a combination of an homogeneous and a particular term for
r = L̃2 :
1 GM

1 GM 1 GM
= 2 (1 +  cos(ϕ + ϕ0 )) −−−→ = 2 (1 +  cos(ϕ)) (190)
rN (ϕ) L̃ ϕ0 =0 rN (ϕ) L̃
which, as we well know, define ellipses around the Sun whose bending depends on the eccentricity 
(in the case of Mercury, the orbit is almost a circle, id est  << 1). If we now plug this result in the

63
quadratic term, our GR non-linear equation become linear and, once more, we can solve it considering
an homogeneous and a particular term, further, since  << 1, we can neglect terms in  of orders higher
than one:
d2  1  1 GM 3GM d2  1  1 GM 3GM
+ = + −→ + = 2 + 2
dϕ2 r r L̃2 r 2 dϕ 2 r r L̃ rN

d2  1  1 GM  GM 2
+ = 2 + 3GM (1 +  cos(ϕ))
dϕ2 r r L̃ L̃2
d2  1  1 GM 3G M
3 3 
+ = 2 + 1 + 2 cos(ϕ) + o( 2
)
dϕ2 r r L̃ L̃4
Consequently, considering the ratio between the two constants:

3G3 M 3  GM −1 G2 M 2 GM rS
= −−−→ ' ' 10−7 (for Mercury)
L̃2 L̃2 L̃2 <<1 rN rN
we can remove the smaller term and simplify our differential equation to:

d2  1  1 GM 6G3 M 3
2
+ = 2 +  cos(ϕ)
dϕ r r L̃ L̃4
(non homogeneous differential equation with an harmonic term) then, worked out a particular solution:

1 3G3 M 3 GM
= ϕ sin ϕ + 2
r L̃ 4 L̃
we can construct the GR orbits’ equation in the form:

1 GM  3G2 M 2 
= 2 1 +  cos ϕ + ϕ sin ϕ (191)
r L̃ L̃2
which presents an additional term compared with the Newtonian proposal, the GR correction (this is no
longer an ellipse). At this point, remembering that  << 1 and G2 M 2 /L̃2 << 1, it can be re-written in
the form:
1 GM n h  3G2 M 2 io
= 2 1 +  cos ϕ 1 − = (192)
r(ϕ) L̃ L̃2
GM n  3G2 M 2   3G2 M 2 o
= 2 1 +  cos ϕ cos + sin ϕ sin ϕ
L̃ L̃2 L̃2
 3G2 M 2   3G2 M 2  3G2 M 2
with: cos −→ 1 and sin ϕ −→ ϕ
L̃2 L̃2 L̃2
Finally, by studying the perihelium of such orbit (shortest distance), given when the cosine is maximized:
h  3G2 M 2 i 2π 4π
cos ϕ 1 − = 1 −→ ϕ = 0, ϕ = , ϕ=
L̃2 1 − 3G M /L̃
2 2 2 1 − 3G2 M 2 /L̃2
we note that it does not correspond to periodic values of ϕ but it constantly changes position: our orbit
completes a revolution not after 2π but in:

2π  3G2 M 2  6πG2 M 2
−−− −− −−→ 2π 1 + = 2π +
1 − 3G M /L̃ 3GL̃2M <<1
2 2 2 2 2
L̃2 L̃2

so, according to GR, planets’ orbits are not closed around the Sun but experience a motion of precession,
a relativistic precession (it is not a precession induced by the other planets, we have not considered
L̃2
them in this discussion). In particular, by introducing the quantity r0 = GM (major semi-axis of the
unperturbed ellipse) we can rewrite and estimate the angle difference generated by Mercury’s precession
as:
6πGM
δϕ = −−−−−−−−→ 0.103800
r0 for Mercury

so, in a century, given that the period of revolution of Mercury is Trev ' 0.24years, we have a residual
precession of 4300 (perfect prediction regarding the anomalous precession).

64
13.2 Light rays’ deflection in Schwarzschild space-time
In the geometrical vision of Einstein’s Theory the space-time around a source is curved. As a consequence
light rays should not travel in straight lines in presence of a gravitational field because, along their path,
geodesics experience small deviation due to the curvature of the space-time. Consequently, starting by
recalling the Schwarzschild metric that describes our space:
 2m  2  2m −1 2  
ds2 = 1 − dt − 1 − dr − r2 dθ2 + sin2 θdϕ2
r r
to estimate the deflection for massless particle, like photons, we have to study null geodesycs in
Schwarzschild space-time:
Duα
=0 with gµν uµ uν = 0 (193)
Dp
where p is our parameter since we have no proper time anymore. Further, considering an analogous
discussion as the one in the previous section, we have u0 = Ẽ and u3 = L̃ (cause of gµν not depending
on t and ϕ) and, consequently:
Ẽ L̃
ṫ = u0 = ϕ̇ = u3 = −
1 − 2m/r r2
where the dotted functions now represent derivatives with respect to p. At this point, if we compute the
normalization condition:
g00 (u0 )2 + g11 (u1 )2 + g22 (u2 )2 + g33 (u3 )2 = 0
 2m  E 2  2m −1 2 L̃2
1− − 1− ṙ − r2 4 = 0
r 1 − 2m/r r r
Ẽ 2 ṙ2 L̃2
− − 2 =0
1 − 2m/r 1 − 2m/r r

while considering the transformation ṙ → r(ϕ):


dr dr dϕ dr L̃ dr
ṙ = = = ϕ̇ = − 2
dp dϕ dp dϕ r dϕ
and recalling the identity:
d 1 1 dr  d 1 2 1  dr 2
=− 2 −→ = 4
dϕ r r dϕ dϕ r r dϕ
we can work out the differential equation:
Ẽ 2 ṙ2 L̃2
− − 2 =0
1 − 2m/r 1 − 2m/r r
L̃ 2
2m 
Ẽ 2 − ṙ2 = 2 1 −
r r
L̃ 2
dr 2 L̃ 2
2m 
Ẽ 2 − 4 = 2 1−
r dϕ r r
 d 1 2 L̃ 2
2m 
Ẽ 2 − L̃2 = 2 1−
dϕ r r r
L̃ 2
2m   d 1 2
Ẽ 2 = 2 1 − + L̃2
r r dϕ r
Ẽ 2
1  2m   d 1 2
= 1 − + but 2m = 2GM
L̃2 r2 r dϕ r
Ẽ 2 1 2GM  d 1 2
= − + differentiate with respect to ϕ
L̃2 r2 r3 dϕ r
2GM  dr  1  dr   d 1  d2  1 
−3 4 =− 3 +2
r dϕ r dϕ dϕ r dϕ2 r
6GM 1 1 d 2  
1 d2  1  1 3GM
4
= 3
− 2 2 2
=⇒ − 2
+ = (194)
r r r dϕ r dϕ r r r2

65
which is a non-linear differential equation of second order and, by comparing it with the one obtained for
massive particles, it differs only by the constant GM/L̃2 . Finding the solution of this equation is not an
easy task but again we can work on it with a perturbative approach, in fact, the ratio:
3GM  1 −1 GM rS
' '
r2 r r r
its maximized when considering the photons passing near the surface of the Sun but, even in that case,
this term is proportional to 10−6 , hence it is basically a perturbation of the Newtonian potential.
Observation. Neglecting this term remarkably simplify the problem as the solution we get:
1 1
= cos(ϕ + ϕ0 )
rN (ϕ) r0

is just the equation of a straight line in polar coordinates:


rN  
1= cos ϕ cos ϕ0 − sin ϕ sin ϕ0
r0
(
x = rN cos ϕ x y
−→ 1 = cos ϕ0 − sin ϕ0
y = rN sin ϕ r0 r0
π  r0
and for ϕ0 6= 0 : y = x tan − ϕ0 −
2 sin ϕ0
while for ϕ0 = 0 : x = r0

what we expect then, is for the correcting term to bend the trajectory:

To find the GR solution, as anticipated, we can take advantage of a perturbative approach, hence, by
replacing the perturbative term with the Newtonian solution, our differential equation is no longer non-
linear and can be resolved much more easily:

d2  1  1 3GM d2  1  1 3GM
− + = −→ − + = 2 (195)
dϕ2 r r r2 dϕ2 r r rN

in particular, choosing ϕ0 = π/2 (no loss of generality), we find:

d2  1  1 3GM
− + = cos2 ϕ (196)
2
dϕ r r r02

(non-linear differential equation with an harmonic term). Then, given the solution of the homogeneous
rN (ϕ) and found a particular solution, like:
1

1 GM  
= 2 1 + sin2 ϕ
r r0

we can finally compose the general solution:


1 1 GM
= cos ϕ + 2 (1 + sin2 ϕ) (197)
r(ϕ) r0 r0

that describes the motion of photons in presence of a gravitational field according to General Relativity.
If we now compare the behaviour for r → ∞ of the unperturbed and perturbed solution, it emerges that:
1 π
lim = 0 implies ϕ −→
r→∞ rN (ϕ) 2
1 1 GM
lim = 0 implies cos ϕ + 2 (1 + sin2 ϕ) −→ 0
r→∞ r(ϕ) r0 r0

66
 
and, since we are dealing with very small particles, we can approximate ϕ = ± π2 + α with α << 1,
hence:
1 GM
0= cos ϕ + 2 (1 + sin2 ϕ) =
r0 r0
1 GM
= − sin α + 2 (1 + cos2 α) =
r0 r0
α GM 2GM
= − + 2 (1 + 1) −→ α =
r0 r0 r0
The deflection of massless particle in Schwarzschild space-time then can be computed as:
2GM
δ = 2α = 2 (198)
r0
assuming to be on Earth, if some light rays pass near the surface of the Sun (r = 7 · 108 m), these suffer
a deflection of 1.7500 .
π
π
2 +α 2

− π2 − α
− π2

Figure 15: Schematization of mass-less particle deflection in Schwarzschild space-time.

These effects has been verified by the observations of Eddington in 1919.

13.3 Gravitational redshift (24-11-2020)

Suppose a spherical source of gravitational field (stars, planets, ...) emitting a wave, we want to compare
the time intervals of a wavelength of two observers O1 and O2 at different distances from the source:

O1 O2

Figure 16: Redshift of a radiation climbing up the gravitational field.

In a flat space-time, waves are described by the wave equation:


1 ∂2
Φ = 0 −→ with:  = ∂µ ∂ µ = − ∇2
c2 ∂t2
but, in a curved space-time, we have to replace partial derivatives with covariant derivatives, so it becomes:
∂ν Φ √
∇µ ∇µ Φ = ∇µ (∂ µ Φ) = ∂µ ∂ µ Φ + Γµµν ∂ ν Φ = ∂µ ∂ µ Φ + √ ∂ν ( g) =
g
1 √ µ 1 √ µν (199)
= √ ∂µ ( g∂ Φ) = √ ∂µ ( gg ∂ν )Φ = 0
g g
√ µν
=⇒ ∂µ ( gg ∂ν )Φ = 0

67
In order to solve this equation we have to insert the metric tensor, which, in our case, since we are in a
Schwarzschild space-time, is:
 2m  2  2m −1 2 √
ds2 = 1 − dt − 1 − dr − r2 (dθ2 + sin2 θdϕ2 ) with: g = r2 sin θ
r r
At this point, without going precisely through the computation, what we can expect due to the fact that
gµν is symmetric and static (not proportional to time), is a solution in the form:

F e(r) ρ
Φ ∼ e−iωt Y m (θ, ϕ)
r
where the time is factorized out and spherical symmetry is represented by spherical harmonics, never-
theless this is still a very complicated differential equation that has no analytical solution and we would
need a numerical approach. Now, what can be immediately noted is that the solution is periodic with
∆t = T = 2π ω , so the time interval is constant, it is the same in every point of the space-time (con-
sequence of metric’s staticity) but to understand completely this result we have to think deeply at what
time we are actually referring to. In fact, as we have seen in sec.12.3, the time variable that appears in
the metric has the meaning of a measurable time only at infinity, id est in a inertial FoR where space-time
is flat and SR is restored. Hence, if we compute the time interval for O1 and O2 , assuming, for the sake
of simplicity, static observers waiting for the passage of a wavelength in their FoR, so basically fixing
(r1 , θ1 , ϕ1 ) and (r2 , θ2 , ϕ2 ), the clocks are now physically measuring their respective proper times:

2m  2 2m 
 r
∆t 1 ≡ ds1 = 1 −
2 2
dt −→ ∆t1 = 1− dt
r1 r1
2m  2 2m 
 r 
∆t2 2 ≡ ds22 = 1 − dt −→ ∆t2 = 1− dt
r2 r2
and since r1 6= r2 , these are clearly different. In particular, if we consider their ratio:
q
∆t1 1 − 2m
r1 ν2
=q =
∆t2 1− 2m ν1
r2

we can extract the formula for the frequency observed in O2 :


v
u 1 − 2m
u
r1
ν2 = ν1 t 2m −−−−−−−→ ν2 < ν1 Redshift phenomenum (200)
1 − r2 r1 <r2

So, our mathematical ∆t measured in an inertial, far away, frame, is not equal if measured in a FoR
immersed in the gravitational field; fact that can be also seen considering the limit for r2 → ∞:

1 2m
r
ν∞ = = ν1 1 −
∆t r1

The frequency change can be estimated considering ∆ν = ν2 − ν1 , or, eventually, the fractional difference:

∆ν 2m
r
ν∞ m GM
= −1= 1− − 1 −−−−−−−−−−−−−−−→ − ' − 2 = Φ(r1 )
ν1 ν1 r1 weak gravitational field r1 c r1
which, for a weak gravitational field, corresponds to the Newtonian potential. In the case of the
GM
Sun, the fractional change in frequencies we expect for a radiation coming from its surface is − c2 r =
−2.12 · 10−6 . Nonetheless, measuring this effect is very difficult cause, generally, the emitting source
does not stand still (atoms on the Sun’s surface shake very rapidly due to the high temperature) so we
have also so keep in account the Doppler’s effect. Roughly, considering the formula for kinetic energy
2 m < v >= 2 kb T , with T ' 3000K we can predict a Doppler’s redshift proportional to 3∆ν that
1 2 3

completely covers Gravitational redshift. Pound and Rebka managed to verify this effect on Earth scale
measuring a frequency difference proportional to 10−15 .

68
Deep look. As anticipated many times now, the stacity of Schwarzschild metric comes handy in many
occasions (it allows us to fix the energy mainly) but understand completely the time coordinate used to
define it, is not an easy task. An intuitive way to try to understand it is the following: assume two static
observers, O1 immersed in a gravitational field and O2 very far away, where the space-time is flat. Then,
imagine the observer O1 records two events, and for each of them, it emits a signal towards O2 . The time
relation between the intervals measured by the two observers is precisely:

2m
r
ds(O1 ) = 1 − dt(O2 ) (201)
r
so, basically, the time coordinate involved is a time measured in a stationary FoR at infinite distance from
the source where the effect of the gravitational field are none. Sometimes it is also called bookkeeping
time as it basically keeps track of all the events super partes. Gravitational fields do not affect only space,
also time flow is affected by gravity, they curve the entire space-time but the slowing down of clocks can
not be measured locally as the proper time itself is slowed down too.

13.3.1 Gravitational redshift from Special Relativity


Suppose to have an observer O0 moving on a trajectory with velocity uα that encounters a photon with
quadri-momentum pµ , with both uα and pµ measured with respect to an inertial FoR. From SR we know
that for photons:
dxα
pα = with pα pα = gαβ pβ pα = 0 and pα;µ pµ = 0

Further, we can express the energy of the photon seen by O0 considering that in its FoR u0α = (1, ~0) and
p0µ = (E 0 , p~ 0 ), so we have the scalar:

E 0 = p0α u0α −−−−−−−−→ pµ uµ


invariant

Subsequently, if we move the entire system in a Schwarzschild space-time, due to the staticity of the
Schwarzschild tensor, we know that p0 = const = E; in addition, considering O0 to be at rest in this new
setup, we also have that uα = (u0 , ~0) with:
1
gµν uµ uν = g00 (u0 )2 = 1 −→ u0 = q
1− 2m
r

and so:
E
E 0 = pµ uµ = p0 u0 = q = hν 0
1− 2m
r

where in the last the energy of the photon has being linked to the frequency observed by O0 (considering
the limit for r → ∞, out of this equation we can also found the energy at rest). Using this relation for
the system in fig.13.3, gives:
E
E1 = q = hν1
1− 2m
r1
E
E2 = q = hν2
1− 2m
r2

from which we can extract the very same equation obtained previously:
v
u 1 − 2m
u
r1
ν2 = ν1 t (202)
1 − 2m
r2

If instead of a static observer we now consider an observer moving along a geodesic:


dxα
uα = with gαβ uβ uα = 1 and uα;µ pµ = 0
ds

69
with both the observer and the photon having a radiant motion (such that θ and ϕ are constants),
considering the Schwarzschild metric we still have that p0 = E and u0 = Ẽ are constants. Consequently,
to work out the scalar E 0 = pµ uµ we have first to compute the relations:
 2m  0 Ẽ
u0 = g0µ u0 = g00 u0 = 1 − u −→ u0 =
r 1 − r(s)
2m

 2m  0 2  2m −1 r 2
gαβ uβ uα = 1 − (u ) − 1 − (u ) = 1
r r r
2m 2 0 2  2m  2m
r
−→ ur = ± 1− (u ) − 1 − = ± Ẽ 2 − (1 − )
r r r
 2m  0 E
p0 = g0µ p0 = g00 p0 = 1 − p −→ p0 =
r 1 − r(λ)
2m

 2m  0 2  2m −1 r 2
pα pα = gαβ pβ pα = 1 − (p ) − 1 − (p ) = 0
r r
 2m 0 E
−→ pr = ± 1 − p = ±E with pr = ∓
r 1 − r(λ)
2m

in which, to the avoid ambiguities in signs, we must how the observer and the photon are moving, or, in
simpler words, if these are free falling or escaping the gravitational field (ur < 0 if the observer is free
falling, pr > 0 if the photon is outgoing). Once the signs are established, we can eventually find a more
general solution:

2m
r
E Ẽ E
E = pµ u = p0 u + pr u =
0 µ 0 r
2m + Ẽ 2 − (1 − ) (203)
1 − r(s) 1 − r(λ)
2m r

which, in the limit of r → ∞ (Ẽ is 1 at infinity), as expected, returns the energy in a rest frame E 0 = E.

14 Gravitational waves (25-11-2020)

To introduce this section, first thing first we are going derive waves equations from vacuum Einstein’s
equations Rµν = 0 under the assumption of weak field → gµν = ηµν + hµν with |hµν | << 1 (small
perturbation of space-time).
Observation. This and further assumptions are only necessary for the sake of simplicity from an ana-
lytical point of view, waves equations are solution of Einstein’s equation in general. In addition, the first
assumption of weak field is not even that much nonphysical, the first time gravitational waves have been
detected on Earth, hµν was proportional to 10−21 , so the approximation |hµν | << 1 is perfectly legit.
To begin, we have to work out the Ricci tensor at first order in hµν , basically linearizing our problem
(This is not a negligible step, due to the self-interacting nature of gravitational field, solving exactly field
equations is actually very complicated):

∂Γλµλ ∂Γλµν
Rµν = Rλ µλν = − − Γη µλ Γλνη + Γη µν Γλλη
∂xν ∂xλ
but Γλµν applied to our metric can be generally expressed as (remember that η is made of constants):

1 λρ  ∂hµρ ∂hρν ∂hνµ 


Γλµν = g + −
2 ∂xν ∂xµ ∂xρ
so, to neglect all the terms o(h2 ) we have to replace g λρ directly with η λρ and compute Rµν considering
only the components linear in Γ:

1 ∂ 2 hλν ∂ 2 hλµ ∂ 2 hλλ 


(1)
Rµν =− hµν − − + Ricci tensor first order in hµν (204)
2 ∂xλ ∂xµ ∂xλ ∂xν ∂xµ ∂xν
and consequently vacuum Einstein’s equations assume the form:

∂ 2 hλν ∂ 2 hλµ ∂ 2 hλλ


∂µ ∂ µ hµν − − + =0 (205)
∂xλ ∂xµ ∂xλ ∂xν ∂xµ ∂xν

70
where all the indices have been lowered or raised using the Minkowski metric (consequence of linearity).
This is a second order linear system of ten equations which, however, are not functionally independent
due to the symmetry of re-parametrization invariance of the underlying theory, if we have a solution, its
coordinate transformation is also a solution completely equivalent to the former one: gauge symmetry
of General Relativity. When discussing Electromagnetism we experienced a very similar situation
with the equation:
Aν − ∂ ν (∂µ A) = 0
and like in that case, in which we couldn’t define uniquely the components of the four-potential because
of the gauge symmetry (Aν (x) → A0µ (x) = Aµ + ∂ µ ζ), now it is the same for hµν unless we consider
gauge fixing. To impose this re-parametrizetion symmetry, we will study the most general coordinate
transformation which leaves the field in the weak form:
∂µ
xµ −→ x0µ = xµ (xα ) + µ (xα ) with ∼ hµν
∂xν
0
such that gµν (x0 ) = ηµν + h µν (x )
0 0

from which we will discuss the relation:


∂x0µ ∂x0ν λρ
g 0µν (x0 ) = g (x)
∂xλ ∂xρ
where g µν = η µν − hµν is the inverse of the Minkowski metric. Then, by considering:

∂h0µν
g 0µν (x0 ) = ηµν + h0µν (x0 ) = ηµν + h0µν (x + ) = ηµν + h0µν (x) + =
∂x
' ηµν + h0µν (x)

we note that, at first order, h0µν (x0 ) ≡ h0µν (x), hence we find:
 ∂µ  ν ∂ν  λρ
g 0µν (x0 ) = η µν − h0µν (x0 ) = δ µλ + λ
δρ+ (η − hλρ (x))
∂x ∂xρ
∂ν ∂µ
η µν − h0µν (x) + o(h2 ) = δ µλ δ νρ η λρ − δ µλ δ νρ hλρ (x) + δ µλ ρ η λρ + δ νρ λ η λρ + o(h2 )
∂x ∂x
ν µ
∂ ∂
η µν − h0µν (x) = η µν − hµν (x) + η µρ + η λν
∂xρ ∂xλ
∂ν µρ ∂µ λν
h0µν (x) = hµν (x) − η − η
∂xρ ∂xλ
∂µ ∂ν
−→ h0µν (x) = hµν (x) − ν
− gauge transformation law
∂x ∂xµ
and we can check that, if hµν (x) is a solution of Einstein’s equation, then also h0µν (x) is a solution too.
Essentially we obtained the same result of Electromagnetism except for the fact that here we are not
handling a field variable in the form of a 4-vector potential but a symmetric tensor. The following step
is to introduce some auxiliary conditions, the gauge fixing conditions, which, in our case, will be the
harmonic gauge:
g µν Γλµν = 0 (206)
by doing that we are limiting the gauge freedom of our system, or, in simpler words, fixing the coordinates
of our system (four constraints on the gravitational potential → only 6 component left free). This gauge
fixing is not made of tensor equations, a detail that already breaks the re-parametrization invariance. By
expanding these equations, always at first order in hµν , we can express our gauge constraints as:

1  ∂h
ρµ ∂hρν ∂hµν 
η µν η λρ + − =0
2 ∂x ν ∂x µ ∂xρ
∂hλν ∂hλµ ∂hµµ
+ − =0
∂xν ∂xµ ∂xλ
∂hλµ 1 ∂hµµ
= multiply for ηλν
∂xµ 2 ∂xλ
∂hµν 1 ∂hµµ
=
∂xµ 2 ∂xν

71
and now we have to introduce these gauge conditions inside field equations:

∂ 2 hλν ∂ 2 hλµ ∂ 2 hλλ


∂µ ∂ µ hµν − − + =0
∂xλ ∂xµ ∂xλ ∂xν ∂xµ ∂xν
∂ 2 hλν ∂  ∂hλν  ∂  1 ∂hλλ  1 ∂ 2 hλλ
= = =
λ
∂x ∂x µ µ
∂x ∂x λ ∂xµ 2 ∂xν 2 ∂xµ ∂xν
2 λ
∂ hµ ∂ ∂h µ
 λ 
∂ 1 ∂hλλ
  1 ∂ 2 hλλ
= = =
λ
∂x ∂x ν ν
∂x ∂x λ ∂x 2 ∂x
ν µ 2 ∂xν ∂xµ
1 ∂ hλ
2 λ
1 ∂ hλ 2 λ 2 λ
∂ hλ
∂µ ∂ µ hµν − − + =0
2 ∂xµ ∂xν 2 ∂xν ∂xµ ∂xµ ∂xν
thus, our problem can be simplified to:
(
hµν = 0
∂hµν
µ
1 ∂h µ
(207)
∂xµ = 2 ∂xν

where it appears a waves equation. The initial perturbations of the Minkowski metric then propagates in
the space-time as a wave that we can imagine as ripples of curvature in a flat space-time. At this point we
have to verify if the harmonic gauge completely fix the gauge of our system (remove entirely the functional
dependence between equations) or there is a residual gauge, like in the case of Electromagnetism. So,
essentially, we have to check if, by performing a gauge transformation on a solution hµν (x):

∂θµ ∂θν
hµν (x) −→ h0µν (x) = hµν (x) − −
∂xν ∂xµ

also h0µν (x) is a solution (satisfy both the waves equation and the gauge fix). Of course for arbitrary θ(x)
this is not true, but are there any requirements to validate this hypothesis? Considering the harmonic
gauge:

∂h0µν 1 ∂h0µµ
=
∂xµ 2 ∂xν
∂  µ ∂θµ ∂θν  1 ∂  µ ∂θµ ∂θµ 
hν − − = hµ− −
∂xµ ∂xν ∂xµ 2 ∂xν ∂x µ ∂xµ
µ
∂h ν 2 µ
∂ θ 1 ∂h µ
µ 2
∂ θ µ
− − θν = −
∂xµ ∂xµ ∂xν 2 ∂xν ∂xν ∂xµ
θν = 0

we immediately find that θµ have to be four harmonic functions to ensure that h0µν is still an harmonic
gauge. Consequently, it can be easily verified that h0µν also satisfies the field equations, which implies
that the gauge was actually not completely fixed and we have a residual freedom to perform gauge
transformations generated by harmonic functions. The introduction of this four further constraints reduce
the independent parameters of hµν just to two. So, as we have done for Electromagnetism, we can
hypothesize a solution in terms of plane waves (ω is fixed to simplify the calculations):
α
hµν (x) = µν eikα x + c.c. with k α = (ω, ~k)

with µν symmetric tensor of constants; then, imposing it to satisfy the waves equations hµν = 0, we
derive the relations:
kα k α = 0 −→ ω 2 − c2 |~k|2 = 0
from which, we also find immediately that the gravitational waves we are studying propagate at the speed
of light:

vg = =c
d|~k|
Subsequently, taking into account the harmonic gauge conditions:

∂hµν 1 ∂hµµ 1
= −→ µν kµ = µµ kν
∂xµ 2 ∂x ν 2
we extract four relations related to the parameters of µν , which now are not completely independent.

72
Gravitational waves along the z-axis
At this point, assume for simplicity that our waves is propagating along ẑ (there is no loss of generality
here cause we can always perform a rotation of space-time) −→ k µ = (k 0 , 0, 0, k 3 ) and since k µ has to be
a null vector we also have that (k 0 )2 = (k 3 )2 −→ k 0 = k 3 ≡ k (we just consider the positive solution for
the moment), hence:
k µ = (k, 0, 0, k) kµ = (k, 0, 0 − k)
Using this 4-vectors we can expand the harmonic gauge relations for µν :
1 µ
µν kµ =  kν
2 µ
1
0ν k0 + 3ν k3 = kν (00 + 11 + 22 + 33 )
2
1
k( ν −  ν ) = kν (00 − 11 − 22 − 33 )
0 3
2
in particular we find that:
1
for ν = 0 : k(00 − 30 ) = k(00 − 11 − 22 − 33 )
2
1
00 + 30 = (00 − 11 − 22 − 33 )
2
1
for ν = 3 : k(03 − 33 ) = − k(00 − 11 − 22 − 33 )
2
1
03 + 33 = − (00 − 11 − 22 − 33 )
2
which can be simplified in (in the first we have summed them):
1
00 + 30 + 03 + 33 = 0 −→ 30 = − (00 + 33 )
2
1 1
(00 − 11 − 22 − 33 ) = 00 + 30 = 00 − (00 + 33 ) −→ 11 = −22
2 2
then, we can complete the set of gauge conditions considering the relations for ν = 1, 2:

for ν = 1 : k(01 − 31 ) = 0 −→ 01 = −31


for ν = 2 : k(02 − 32 ) = 0 −→ 02 = −32

Afterwards, we have to consider the residual gauge freedom, so, given:


∂θµ ∂θν
h0µν = hµν − − with: θµ (x) = 0
∂xν ∂xµ
and chosen an harmonic function in the form:
λ
θµ (x) = iθµ eikλ x + c.c. with: θµ = constants

(the residual gauge is still not fixed there cause the amplitudes θµ have not yet been determined) we
have:
  λ
h0µν (x) = hµν (x) + θµ kν + θν kµ eikλ x + c.c.
α
  λ
= µν eikα x + θµ kν + θν kµ eikλ x + c.c.
λ
= 0µν eikλ x + c.c. with 0µν = µν + θµ kν + θν kµ

which, again, has to satisfy the gauge harmonic, thus we inherit the four harmonic gauge conditions (the
very same computation to extract them):
1
030 = − (000 + 033 ) 001 = −013
2
022 = −011 002 = −023

73
Finally then, by computing the new amplitudes 0µν as functions of the six previous parameters left
independent (remember that µν is symmetric):
000 = 00 + 2kθ0 011 = 11 023 = 23 − kθ2
033 = 33 − 2kθ3 012 = 12 013 = 13 − kθ1
(with 11 and 12 that are gauge invariant) we can fix the residual gauge by picking the four θµ constants
that simplify our calculations more:
00 13
θ0 = − θ1 =
2k k
23 33
θ2 = θ3 =
k 2k
which essentially corresponds in setting all the non-invariant 0µν components equal to zero. By doing that,
all the harmonic gauge conditions we have identified previously are trivially verified. So, we can conclude
that, given a gravitational wave, it is always possible to perform a gauge transformation such that, out
of the ten starting parameters describing it we end up only with three components left non null and only
two of them completely independent (so only two parameters have absolute physical significance). In the
specific case of our gravitational wave oriented along ẑ, we end up with:
012 , 011 , 022 6= 0 with 022 = −011
which basically means that only the transverse components of the wave are the one left non null.

14.1 Helicity for Gravitational waves (01-12-2020)

So far the similarities with Electromagnetic field’s equations have been many, but when it comes the
computation of the helicity of a gravitational wave we have a discrepancy. Considering again a Gravita-
tional wave propagating along the z-axis, ~k k ẑ, after a rotation of angle θ in the xy plan, how do our
field’s equations change? Due to the fact we have employed a tensor formalism during its entire study
and since rotations are a subgroup of the Lorentz group, this transformation can be expressed as:
hµν −→ Λµα Λνβ hαβ so µν −→ Λµα Λνβ αβ
(the standard transformation law for 4-vectors only counts one Λ, this is the cause of the discrepancy
with Electromagnetism waves’ helicity) with Lorentz matrices in the form:
1 0 0 0
 
0 cos θ − sin θ 0
Λαβ = 
0 sin θ cos θ 0

0 0 0 1
Then, to recover the equation that defines helicity:
Φ −−−−−−−−→ Φ̃ = eihθ Φ
rotation of θ

we have to define (in EM we had to change polarization, from linear to circular):


± = 11 ∓ i12 = −22 ∓ i12
f± = 31 ∓ i32 = −01 ± i02
from which we obtain the following transformation law for the six independent components of hµν :
± −→ ˜± = e±2iθ ±
f± −→ f˜± = e±iθ f±
33 −→ ˜33 = 33
00 −→ ˜00 = 00
and so, to ± we have associated an helicity of h ± 2, for f± we have h ± 1 and finally, for 00 and 33 ,
h is just zero, but, by performing a precise gauge transformation and fixing the residual gauge freedom,
all the components except 12 and 11 can be equaled to zero, so the only relevant helicity is h = ±2
related to ± . This also implies that the physical components of a Gravitational waves are transverse to
the direction of propagation.

74
14.2 The TT gauge
As anticipated hµν is a symmetric tensor and, considered the harmonic gauge conditions and with the
residual gauge freedom fixed to minimize the non null components, it can be represented as a matrix in
the form:
0 0 0 0
 
0 hxx hxy 0
hµν = 0 hyx −hxx 0 −→ h µ = 0
 µ

0 0 0 0
so this gauge is also called the TT gauge, or Transverse Traceless gauge.
Deep look. At a quantum level we have been able to quantize the Electromagnetic Theory and we have
discovered that the Electromagnetic interaction is mediated by a massless particle, the photon, with spin
s = 1 and Sz = ±1 (massless cause EM waves travel at the speed of light, S = 1 cause their helicity can
only be h = ±1). In the case of Gravitational Theory we have not still being able to formulate a quantum
theory, but what we can hypothesize is for the Gravitational interaction to be mediated by a massless
particle too, which can be called graviton, of spin S = 2 and Sz = ±2 (in both cases we do not have
the usual degeneracy 2S + 1, typical of massive particles). Particle states are associated to irreducible
representations of the Poincare group (related to SU(2)) but for massless particles these representations
are completely different (there is not a rotations group algebra).

14.3 Effects of Gravitational waves


A fundamental aspect to being able to detect Gravitational waves is to know their effects. To study
them we will first consider a very simple system composed by a static test particle (uα t=0 = (1, 0)) in a
Minkowski space-time invested by a Gravitational wave. The particle considered, being free, moves along
a geodesic described by the equations:
Duα duα dxµ ν duα
= + Γαµν u = + Γαµν uµ uν = 0
Ds ds ds ds
and when the Gravitational wave arrives, the space-time is perturbed gµν −→ ηµν + hµν . In particular,
if we assume a TT gauge Gravitational wave, its acceleration can be computed considering the relation
(first order in hµν ):
duα
= −Γα00
ds t=0
with:
1  ∂h
β0 ∂hβ0 ∂h00 
Γα00 = η αβ + − −−−−−−−−−−→= 0
2 ∂x 0 ∂x0 ∂xβ in the TT gauge
and so we can conclude that a particle at rest, when invested by a Gravitational wave, stands static as
its acceleration is null. As a consequence we might suppose that Gravitational waves basically have no
effects and so are undetectable. Nonetheless, if we reflect deeper, what this equations are actually telling
us is just that a particle in some given position before the arrive of a wave remains at this fixed position
of coordinate; but, since we have chosen the FoR (necessary to fix the TT gauge) and due to fact that
this latter moves with the Gravitational wave, we cannot see its effect. To be able to do that we have to
find an invariant of the Gravitational field, id est some other observables, but one particle alone would
never be enough because we can always find a FoR which moves with it.
Thus, considered a system composed of two particles and, again, a TT gravitational wave parallel to
the z-axis, this time we do not focus on the motion of one of them but we will observe the proper
distance between the two and if it does change as the wave passes. First, by looking at the line element
in space-time, defined as:

ds2 = (ηµν + hµν )dxµ dxν with gµν = ηµν + hµν

it can be noted the two-fold significance of the metric tensor: on one side, it represents the gravitational
potentials, on the other, it describes the metric properties of the space-time. If a flat space-time is
invested by a Gravitational wave, its metric properties change and, as a consequence, the proper distance
between two points does change too (the line element changes). As said then, given a TT gauge wave
propagating along ẑ, hµν = hµν (t − z), we can witness the following phenomena:

75
- if hxy = 0, then our line element becomes:

ds2 = dt2 − (1 − hxx )dx2 − (1 + hxx )dy 2 − dz 2 (208)

so, if our two particles are positioned like:

y
(x0 , y0 + dy0 )

(x0 , y0 )

their distance is just dl2 = dy02 in a flat space-time, but when the wave passes, it becomes:

dl2 = (1 + hxx )dy 2

in particular, when hxx > 0 the proper distance increases, while, when hxx < 0, it reduces, but the
points are still static, they are not moving, it is the space itself that is stretched and shrinked under
the perturbation of the Gravitational wave (we are studying a wave so hxx ∝ sin ω(t − z), it oscillates
between positive and negative values).
Similarly, if our particles are positioned like:

y
(x0 , y0 ) (x0 + dx0 , y0 )

their unperturbed proper distance is dl2 = dx2 and becomes

dl2 = (1 − hxx )dx2

after the effect of the Gravitational wave. This time however, when hxx > 0 the distance shrinks while
when hxx < 0 it stretches. Their propert distance then is in anti-phase with respect to the one describing
an y-shift.
- if hxx = 0, then the line element is a little bit more complicated due to appearance of a mixed
term:
ds2 = dt2 − dx2 − dy 2 − dz 2 + 2hxy dxdy (209)
but essentially, the effects are identical to the former case, just in a rotated plane:
(
x −→ x̄ = √12 (x − y)
−→ ds2 = dt2 − (1 − h12 )dx̄2 − (1 + h12 )dȳ 2 − dz 2
y −→ ȳ = √12 (x + y)

Observation. Roughly, we can hypothesize how their detection has been performed, in fact, it is sufficient
to consider an interferometer with two equal light paths on the x and y axes that has a destructive
interference at the end. If this system is invested by a Gravitational wave, due to the stretch on one
axis and the shrink on the other, we should observe a de-phasing factor in the interference, which should
no longer perfectly destructive. Nevertheless it must be said that the difficulties are many and many
as hµν ∝ 10−21 . When these were first observed, the space perturbation we observed was proportional
to 10−18 , which can be compared with the fraction of an atom. In general, the most immediate way to
visualize the effects of a gravitational waves is by considering a ring of particles and its deformations.

76
Another way to obtain the same result is by considering Geodesic deviation, in particular, assuming
two particles moving along congruent geodesics we have the formula:
D2 δxµ
= Rµαβγ uα uβ δxγ
Ds2
but, if we study this system from the FoR of one of the two particles, we have that uα = (1, ~0 ) and s ≡ t
(the proper time is equivalent to the time coordinate), so, at first order in hT T , the equation simplifies
to:
D2 δxi 1 iµ ∂ 2 hT T d2 δxi 1 ik ∂ 2 hTkjT j
= η −−−−−→ = η δx (210)
Dt2 2 ∂t2 o(hT T )2 dt2 2 ∂t2

15 Gravitational radiation (02-12-2020)

15.1 Solution of Einstein’s equation in presence of sources


In order to describe how a source of Gravitational field emits a Gravitational radiation, first we have to
solve the full Einstein’s equations:
1
Rµν − gµν R = 8πG Tµν (211)
2
and, to simplify the problem, we will assume the weak field condition, gµν = ηµν + hµν , at first order
approximation o(h2 ). Further, as we may recall, we have to consider the Bianchi identities:
1 1
(Rµν − δ µν R);µ = 0 −−−−2−→ (R(1)µν − δ µν R(1) ),µ = 0
2 o(h ) 2
from which we extract that T µν,ν = 0 (continuity equation), and, given the gauge freedom of our theory,
we can impose the harmonic gauge condition (break of the re-parametrization invariance):
∂hµν 1 ∂hµµ
=
∂x µ 2 ∂xν
with the residual gauge that can be eventually fixed choosing the remaining four functions µ that appears
in the gauge transformation law:
hµν −→ h0µν = hµν − ∂µ ν − −∂ν µ
In particular, in the harmonic gauge the Ricci tensor at first order reduces to:
1
Rµν(1)
= − hµν
2
and, if we consider its trace, it is defined as:
 1  1
R(1) = g µν R(1)µν = η µν R(1)µν = η µν − hµν = − hµµ
2 2
Combining all these relations together, allows us to rewrite the initial Einstein’s equations as:
1 1  1 
− hµν − ηµν − hλλ = 8πG Tµν
2 2 2
1
hµν − ηµν h λ = −16πG Tµν
λ
2
and so, due to the fact we have also performed a gauge fixing, our entire problem can be expressed as:
(
hµν − 21 ηµν hλλ = −16πG Tµν
∂hµν 1 ∂h µ
µ (212)
∂xµ = 2 ∂xν

which is not in the usual form of wave equations. To restore such structure we have two ways, one of
them is by computing:
1
Rµν − gµν R = 8πG Tµν multiply for g µν
2
R − 2R = 8πGT λλ
−R = 8πGT λλ replaced in the first equation
1
Rµν = 8πG (Tµν − gµν T λλ )
2

77
so we find an expression of the Ricci tensor as a function of the energy-momentum tensor and its trace,
and if we now head back to the linearized problem, we obtain:
1
(1)
Rµν = − hµν
2
1 1
− hµν = 8πG (Tµν − gµν T λλ )
2 2
1
hµν = −16πG (Tµν − gµν T λλ )
2
which is a wave equation proportional to its sources. Nevertheless this actually is not the best way of
doing it because it is difficult to assign a physical interpretation to the left hand side that mixes the
energy-momentum tensor and its trace. The other approach starts by considering:
1
hµν − ηµν hλλ = −16πG Tµν
2
 1 
 hµν − ηµν hλλ = −16πG Tµν
2
and, if we introduce the new variable:
1
h̄µν ≡ hµν − ηµν hλλ
2
it becomes a wave equation in terms of this latter:

h̄µν = −16πG Tµν (213)

At this point, computing the trace of h̄µν we derives and interesting propertiy:
1
h̄µν = hµν − ηµν hλλ multiply by η µν
2
h̄µµ = h1µ
µ − 2h µ = −h µ
µ µ

from which also derives its name: trace reverse variable. After that, to reformulate our field equations
in terms of h̄, first we have to rewrite the harmonic gauge condition, so, considered the inverse relation:
1 1
hµν = h̄µν + ηµν hλλ = h̄µν − ηµν h̄λλ
2 2
we find:
∂hµν 1 ∂hµµ
=
∂xµ 2 ∂xν
∂  1  1 ∂  
h̄µν − δ µν hλλ = − h̄λλ
∂x µ 2 2 ∂xν

−→ ∂µ h̄µν = 0

and so we have: (
h̄µν = −16πG Tµν
(214)
∂µ h̄µν = 0
Now, to interpret physically the trace reverse variable we have to recall the TT gauge, which we have
defined for hνν = 0. When we have discussed the TT gauge we have seen that fixed some precise
conditions, the physical degrees of freedom of a Gravitational wave are completely explicit, hence, since
the two variables coincide in such conditions cause h̄νν = hνν = 0, we can conclude that also h̄ has two
physical degrees of freedom. A further analogy is given from the structure of our problem which is
analogous to that of Electromagnetic field:
( (
Aµ = 4π c J
µ
f (x) = s(x)
−−−−−−−−−→
∂µ A = 0
µ schematized as ∂µ f (x) = 0

To solve this type of problems we have to introduce the Green function for the d’Alambert operator
(remember that we are in a Minkowski space-time):

x D(t, ~x, t0 , ~x 0 ) = δ 4 (x − x0 )

78
and if we find a solution we can express our variables as:
Z
f (x) = d4 x0 D(x − x0 )s(x0 )

To being able to do that, we have to employ the Fourier transform of our Green function:
1
Z
α 0α
D(x − x ) =
0
d4 k e−ikα (x −x ) D̃(k)
(2π)4

extract an expression like D̃(k) = − k2 −k


1
2 by computing the wave equation and then replace it in the
0
first definition and integrate the Green function. Among all the infinite Green functions (cause once we
have a Green function we can always add a solution of the homogeneous to find another Green function)
the one we have to chose is the Retarded Green function defined for:

Dret (x − x0 ) = 0 for t < t0

(the source has to determine a solution in the future) and so:


0
1
Z +∞
e−ik0 (t−t )
Z
i~ x 0)
D(x − x ) =
0 3
d ke x−~
k·(~
dk0 2
(2π)4 −∞ k − k02

Im(k)
t < t0

-k k
Re(k)
− −
t > t0

where we clearly note that for t < t0 we have no singularities and so the integrals is just zero as requested,
while for t > t0 :
1 δ[t − t0 − |~x − ~x 0 |]
Dret (x − x0 ) = non null for t0 = t − |~x − ~x 0 |
4π |~x − ~x 0 |
and so the solution will be in the form:
Z
fret (x) = d3 x0 dt0 Dret (x − x0 )s(x0 ) (215)

In our case we starts with:


h̄µν = −16πG Tµν
so our solution is:
Tµν (t − |~x − ~x 0 |, ~x 0 )
Z
h̄µν (~x, t) = −4G d3 x0
|~x − ~x 0 |
which basically implies that the gravitational potential at a given point in space-time is the summation of
all the infinitesimal contributes of our source (out of our source Tµν = 0). Compute this integral is rather
complicated but, considering the far field approximation (we fix our observation point very far) we
i
can replace |~x| = r and |~x − ~x 0 | = r − ni x0i + o(r−1 ) with ni = xr , so |~x − ~x 0 | ' r, as a consequence our
integral becomes:
Tµν (t − r, ~x 0 )
Z
h̄µν (~x, t) = −4G d3 x0
r
4G
Z
=− d3 x0 Tµν (t − r, ~x 0 )
r
afterwards, recalling that ∂µ T µν = 0 (continuity equation from the Bianchi identities) we have:
- for ν = 0:
∂t T tt + ∂i T it = 0

79
- for ν = i:
∂t T ti + ∂j T ji = 0
and if we compute the derivative with respect to time of the first equation:
∂t (∂t T tt + ∂i T it ) = ∂t2 T tt + ∂t (∂i T ti ) =
= ∂t2 T tt + ∂i (∂t T ti ) =
= ∂t2 T tt + ∂i (−∂j T ij )
we derive that:
∂t2 T tt = ∂k ∂l T kl multiply for xi xj
(∂t2 T tt )xi xj = (∂k ∂l T kl )xi xj
∂t2 (T tt xi xj ) = ∂k ∂l (T kl xi xj ) + 2T ij − 2∂k (T ik xj + T kj xi )
from which we can extract the definition of T ij :
1 1
T ij = ∂t2 (T tt xi xj ) − ∂k ∂l (T kl xi xj ) + ∂k (T ik xj + T kj xi )
2 2
that can be employed to solve the integral for h̄ij :
4G
Z
h̄ij = − d3 x0 Tij (t − r, ~x 0 ) =
r
4G n h1 1
Z i Z h io
=− d3 x0 ∂t2 (T tt x0i x0j ) + d3 x0 ∂k0 (T ik x0j + T kj x0i ) − ∂k0 ∂l0 (T kl x0i x0j )
r 3 2 R3 2
ZR
4G n h1 i Z h io
=− d3 x0 ∂t2 (T tt x0i x0j ) + dΣ (T ik x0j + T kj x0i ) − (T kl x0i x0j )
r 3 2
Z R Σ
4G 3 0 1 2
h i
=− d x ∂t (T x x ) =
tt 0i 0j
r R3 2
2G 2
Z
=− ∂ d3 x0 T tt x0i x0j
r t R3
where we have use the Gauss theorem to simplify many calculus since T µν is always equally zero if
computed on a surface at infinity. Finally, assuming a non relativistic source, id est T tt = ρ, and
considering the second moment of mass distribution:
Z
Iij ≡ d3 x0 ρ(~x 0 , t)xi xj

we obtain the quadrupole formula:


2G d2
h̄ij = − Iij (216)
r dt2 ret

By looking at this formula we can conclude that in order for a source to generate a Gravitational wave
it must have ∂t2 Iij 6= 0 or, in simpler words, a time varying quadrupole momentum.
Observation. This result should now clarify why in the Schwarzschild’s solution, a time varying spherical
gravitational source, like a pulsating star, does not generate a Gravitational potential that varies in time:
because its quadrupole momentum is zero. To have Gravitational waves we must consider binary systems,
like two black holes spinning and so on.

15.2 Multiple expansion of a field


Given a field described by a function f = f (r, θ, ϕ), it is possible to expand its definition according to
the equation:
∞ X∞ X l
X Dm
f (r, θ, ϕ) = Yl (θ, ϕ)
ls m
s
(217)
s=1
r
l=0 m=−l

and in particular, there is a theorem which states that the expansion of l starts for l ≥ h (the helicity).
In the case of Electromagnetism we have seen that h = ±1, and so, it follows that the minimal source of
Electromagnetic waves are dipoles; similarly, for Gravitational waves we have seen that h = ±2, so we
must consider, at least, quadrupoles.

80
Observation. Actually gravity is the weakest of the fundamental forces, the only reason why it is so
relevant at large scales is because negative masses do not exist, so their effects keep adding.

81

You might also like