Z H A N P E I FA N G

A B R I E F S K E T C H O F R E L AT I V I T Y

P R E PA R E D F O R T H E S U M M E R S C I E N C E P R O G R A M , J U LY 2 0 2 0
2 fang

0.1 Introduction

These lecture notes provide a glancing survey of the basic mathemati-

cal framework of special and general relativity. I will try to keep things
relatively light on the mathematical side for conceptual clarity, and
avoid too much historical detail, though the historical development of
the theory is of definite intellectual interest.
The sources I consulted can be found listed at the end of these
lecture notes. I have basically cribbed all of Carroll’s notes in "A
No-Nonsense Introduction to General Relativity", also referencing his
textbook Spacetime and Geometry, and I will be using his notational con-
ventions, such as letting the speed of light c = 1, and in particular his
sign convention for the metric. So this really cannot be considered
my own work. I have done my best to make the material palatable
and readily digestible to somebody with minimal prior exposure to
things such as tensor calculus; I am not sure if I’ve succeeded in this
regard. I would additionally encourage you to read Wikipedia’s non-
technical introduction to the subject, which can be found at https:
//en.wikipedia.org/wiki/Introduction_to_general_relativity.
1
Special relativity

1.1 Space and time

All of the strangeness of special relativity arises from the invariance of

the speed of light c in all reference frames. This leads to our consider-
ing the notion of a unified spacetime, for which we want to be able to
talk about space and time coordinates in a mathematical sense.
The spacetime interval is given by ∆s2 = −∆t2 + ∆x2 + ∆y2 + ∆z2 ,
and is what we use when we talk about the "distance" between two
"things"; it is a "distance" measure which also includes information
on how events are separated in time. Instead of using t’s and x, y, z’s
however, which can get kind of clunky later on, we can instead write
all of the space and time quantities (the coordinates of spacetime) in a
single vector1 . For Cartesian coordinates: 1
Yes, in this lecture you’re going to see
why index notation is extremely useful.
 Note that the superscripts here are in-


 x0 ≡ ct = t dices and not exponents.

 x1 = x

xµ (1.1)


 x2 = y

 x3 = z


The collection of all four coordinates is denoted as x µ ; since this vector

has four components we call it a 4-vector. So our usual space compo-
nents xi are given by:2 2
Convention has it that we use Latin
 indices as free/dummy indices when
1 we’re just talking about the space com-
x = x


ponents i ∈ {1, 2, 3}, and Greek let-
xi x2 = y (1.2) ters when we’re talking about both space
and time components, for example µ ∈

x3


=z {0, 1, 2, 3}.

Special relativity is set within the Minkowski spacetime, which com-

bines 3-dimensional space and time into a 4-dimensional manifold3 3
Informally, a manifold is a “possibly
and contains all information about the geometry of the manifold; we curved space which in small enough (in-
finitesimal) regions looks like flat space."
say that the x µ are coordinates on this manifold. The Minkowski met-
4 fang

ric is given as a matrix:

 
−1 0 0 0
 0 1 0 0
= (1.3)
 
ηµν 
 0 0 1 0
0 0 0 1

and so, moving away from the discrete form ∆s2 , the spacetime inter-
val can be succinctly written using the metric as4 4
An equation of this form is also, unfor-
tunately, often called "the metric", par-
ticularly in cosmology; for example the
ds2 = ηµν dx µ dx ν = −dt2 + dx2 + dy2 + dz2 (1.4)
Friedmann–Lemaître–Robertson–Walker
metric (or FLRW for short).
In this I have introduced our implementation of the dot product:

A · B ≡ ηµν Aµ Bν = − A0 B0 + A1 B1 + A2 B2 + A3 B3 (1.5)

The Minkowski metric has the Lorentzian signature (− + ++). This

is why the interval elapsed as a particle fixed in space moves forward
in time is actually negative: ds2 = −dt2 < 0. The proper time is the
time measured by a clock following a timelike worldline in the metric,
dτ 2 ≡ −ds2 , and the discrete proper time interval between any two
events is given by (∆τ )2 = −(∆s)2 = −ηµν ∆x µ ∆x ν . The proper time
elapsed along a trajectory through spacetime will be the actual time
measured by an observer on that trajectory.
Z r
dx µ dx ν
Z p
τ= −ds2 = −ηµν dλ (1.6)
dλ dλ
Finally, the 3D Cartesian space we typically consider in classical me-
chanics (and during SSP), we discard the time component and the in-
terval is given simply by ds2 = (dx )2 + (dy)2 + (dz)2 = dr2 + r2 dθ 2 +
r2 sin2 θdφ2 , where the second equality is for a spherical coordinate
system like the ones we’ve been considering throughout the program.

A little more on 4-vectors

Vectors in spacetime are always fixed at a particular event in our space-
time. A four-vector V µ is known as a timelike vector if it has negative
norm, null vector if the norm is zero, and spacelike if the norm is
positive.

µ
< 0 V timelike

 Figure 1.1: A lightcone as it appears on a
ηµν V µ V ν = 0 V µ lightlike or null (1.7) spacetime diagram: a single space coor-

 dinate x on the horizontal, time coordi-
 µ
> 0 V spacelike nate t on the vertical axis. [c/o Carroll]

Similarly, trajectories with ds < 0 are timelike, ds = 0 null, and ds > 0

spacelike. The set of null trajectories in/out of an event are called its
light cone (Fig. 1.1).
 a brief sketch of relativity 5

A path through spacetime is specified by giving the four space-

time coordinates as a function of some parameter λ, so x µ (λ). A
path is timelike/null/spacelike if its tangent vector dx µ /dλ is time-
like/null/spacelike. The tangent vector dx µ /dτ = U µ is the 4-velocity,
and the 4-momentum is pµ = mU µ , with m the mass of the particle.

1.2 Lorentz transformations

In an introductory course in special relativity,5 you might see the 5

I don’t know if they still do this, but
Lorentz transformation written as, for a boost in the + x-direction: when I was a freshman in the Physics
60 series at Stanford, we began with spe-
cial relativity and then did your typical
x 0 = γ( x − vt) classical mechanics with free-body dia-
grams, blocks on inclines, rolling with-
y0 = y
out slipping, etc. Talk about a trial by
z0 = z fire!

t0 = γ (t − vx )
√
where γ is the Lorentz factor6 γ = 1/ 1 − v2 . The Lorentz factor is 6
There’s a hidden speed-of-light c
the factor by which time, length and relativistic mass change for an in here—you’ll typically first see the
Lorentz
√ factor firstpwritten as γ =
object in motion. 1/ 1 − v2 /c2 = 1/ 1 − β2 . But again,
However, the above transformation can be more succinctly written I’m suppressing all c’s.
as:
0 µ0
xµ = Λ ν xν (1.8)
where the matrix
   
cosh φ − sinh φ 0 0 γ −γv 0 0
0
− sinh φ cosh φ 0 0 −γv γ 0 0
Λ ν=
µ
=  (1.9)
   
 0 0 1 0  0 0 1 0
0 0 0 1 0 0 0 1

for the + x-boost. Another example of a transformation matrix is rota-

tion in the xy plane:
 
1 0 0 0
0 cos θ sin θ 0
µ0
Λ = (1.10)
 
ν
− sin θ

0 cos θ 0
0 0 0 1

In general, Lorentz transformations are defined as those matrices Λ

which satisfy the equation η = Λ T ηΛ, or, writing all the indices out,
µ0 0 µ0 0
ηρσ = Λ ρ ηµ0 ν0 Λν σ = Λ ρ Λν σ ηµ0 ν0 . When we say that a quantity or
physical law is invariant under Lorentz transformation, we mean that
said thing does not change when Λ is applied to it.
And that’s probably all you need to know about special relativity
for right now.7 7
Apologies for not going deeper into the
typical thought experiments—clocks on
a spaceship for time dilation, pole-barn
paradox for length contraction, the twin
paradox and so on. Feel free to look
those up on your own time! They are
essential for getting an intuition for the
’weirdness’ of special relativity.
2
Some formalism

Some more proper mathematical formalisms before we get any further

into the mathematical weeds of GR, which is on some level an exercise
in applied differential geometry.

2.1 Vectors

Suppose that each tangent space we set up a basis of four vectors ê(µ) ,
µ ∈ {0, 1, 2, 3}. Then any abstract vector A can be written as a linear
combination of basis vectors:

A = Aµ ê(µ) (2.1)

The coefficients Aµ are the components of the vector A, and often we

refer to them interchangeably (ignoring the basis and saying things
like "the vector Aµ ".)
A standard example of a vector in spacetime is the tangent vector to
a curve; as mentioned earlier, a parameterized curve through space-
time is specified by the coordinates as a function of the parameter, for
example x µ (λ). The tangent vector V (λ) has components

dx µ
Vµ = (2.2)
dλ
In this case, the entire vector is formally understood as V = V µ ê(µ) .
Under a Lorentz transformation the coordinates x µ change according
to Eq. (1.8), while the parameterization λ doesn’t change; so the com-
0 µ0
ponents of the tangent vector change as V µ → V µ = Λ ν V ν . How-
ever, this is just for V’s components in some coordinate system; the
vector V itself is invariant under Lorentz transformation.
Basis vectors transform as follows. The old basis ê(µ) transforms
into the new one ê(ν0 ) by multiplying by the Lorentz transformation:

ê(ν0 ) = Λ
µ
ê
ν0 (µ)
(2.3)
 a brief sketch of relativity 7

Dual vectors (one-forms)

Once we have set up a tangent vector space Tp , there is an associated
space known as the dual vector space, usually denoted by an asterisk
as Tp∗ . Every dual vector (also known as a one-form) can be written in
terms of its components ω = ωµ θ̂ (µ) and transforms like ωµ0 = Λνµ0 ων .
Like with vectors, we typically write ωµ to represent the entire dual
vector.
Now is also probably a good time to define contravariant and co-
variant vectors, and explain the upper/lower index notation I’ve used
extremely liberally thus far. Elements of Tp ("vectors") are referred to
as contravariant and are the ones with upper indices, and elements of
Tp∗ ("dual vectors") are referred to as covariant and have their indices
down below as subscripts.
In spacetime the simplest example of a dual vector is the gradient
of a scalar function, the set of partial derivatives with respect to the
spacetime coordinates.
∂φ
dφ = µ θ̂ (µ) (2.4)
∂x
The components of dual vectors transform like

∂φ ∂x µ ∂φ µ ∂φ
0 = 0 = Λ µ0 µ (2.5)
∂x µ ∂x µ ∂x µ ∂x
Because the gradient is a dual vector, we can write the partial deriva-
tive in this shorthand, using commas:

∂φ
= ∂µ φ = φ,µ (2.6)
∂x µ
The gradient of a tangent vector to a curve is just the ordinary deriva-
tive of the function along the curve, as one might intuit.

∂x µ dφ
∂µ φ = (2.7)
∂λ dλ

2.2 Tensors

As we make our transition from flat to curved spacetime, we can no

longer use our trusty old Cartesian coordinate system; we are moti-
vated to make our equations coordinate-invariant. Tensors help us write
equations in a coordinate-invariant form; they can be understood as
like a vector, just with more indices. More abstractly, tensors can be
defined as those objects which transform under a change of coordinates
0
x µ → x µ as tensors. For example, this is a tensor:
0
µ0 ∂x µ ∂x ν ∂x ρ µ
S ν0 ρ0
= S νρ (2.8)
∂x µ ∂x ν0 ∂x ρ0
8 fang

The unprimed indices in the RHS are the dummy indices we’re sum-
ming over, so that we can convert everything to the primed coordinate
system.
GR has no preferred coordinate system: if an equation relating two
tensors holds in one coordinate system, it holds in all coordinate sys-
tems. As with vectors, the upper indices are contravariant, and the
lower indices are covariant.
Again, the index notation is really quite powerful. For instance, all
of classical electromagnetism can be summarized in just two lines:

∂µ F νµ = J ν (2.9)
∂[µ Fνλ] = ∂m uFµλ + ∂ν Fλµ + ∂λ Fµν = 0 (2.10)

where Fµν is the electromagnetic field strength tensor:

 
0 − E1 − E2 − E3
E
 1 0 B3 − B2 
Fµν =  = − Fµν (2.11)

 E2 − B3 0 B1 
E3 B2 − B1 0

and J ν = (ρ, J) is the four-current. This formulation of E&M turns out

to hold in curved space so long as we replace the partial derivatives ∂µ
by covariant derivatives ∇µ (defined later).

Manipulating tensors
An (n, m) tensor (that is, one with n upper and m lower indices) can be
contracted to form a (n − 1, m − 1) tensor by summing over one upper
and lower index:
µλ
Sµ = T λ (2.12)

The contraction of a 2-index tensor is called the trace. A tensor is

symmetric in two indices if we can interchange them without chang-
ing the tensor, S ... αβ ... = S ... βα ... , and antisymmetric if it changes
sign, S ... αβ ... = −S ... βα ... . For an arbitrary tensor (without any par-
ticular symmetry properties), we can pick out the antisymmetric and
symmetric pieces by taking appropriate linear combinations.1 1
Included for completeness; please don’t
worry about this too much.
1
T(µ1 µ2 ... µn ) = ( Tµ1 µ2 ... µn + sum over permutations of µ1 ...µn )
n!
1
T[µ1 µ2 ... µn ] = ( Tµ1 µ2 ... µn + alternating sum over permutations µ1 ...µn )
n!

The metric tensor

The most important tensor in GR is the metric tensor gµν , which gen-
eralizes the Minkowski metric ηµν we met before. As in Minkowski
 a brief sketch of relativity 9

space, we can use the metric to raise or lower indices:

Aµ ≡ gµν Aµ , Aµ = gµν Aν (2.13)

and take dot products:

A · B ≡ gµν Aµ Bν = Aν Bν (2.14)

We can define the inverse metric gµν (already spoiler-ed a few lines
above) as the matrix inverse of the metric tensor:
µ
gµν gνρ = δρ (2.15)
µ
where δρ is the Kronecker delta in this spacetime (!!).

2.3 Curvature

Covariant derivatives
The partial derivative (Eq. 2.6) is, unfortunately, not a tensor. Trans-
forming the partial derivative of a scalar returns a reasonable result (a
(0, 1) tensor):
∂x µ
∂µ φ → ∂µ0 φ = µ0 ∂µ φ (2.16)
∂x
0
∂x µ
but if we try the same move on a vector, using V µ → ∂x µ Vµ:
0
 µ  !
ν0 ∂x ∂x ν ν
∂µ V → ∂µ0 V =
ν
0 ∂µ V
∂x µ ∂x ν
0 0
∂x µ ∂x ν ∂x µ ∂2 x ν
= µ 0 ν
( ∂µ V ν ) + µ0 ν µ V µ
∂x ∂x ∂x ∂x ∂x
which has an extra second term we don’t want. Instead we need to
define a covariant derivative to be a partial derivative plus a correction
which is linear in the original tensor:

∇µ V ν = ∂µ V ν + Γνµλ V λ (2.17)

where the new symbol I’ve introduced Γνµλ stands for a collection of
numbers called connection coefficients, which transform not as ten-
sors but as
0 0
0 ∂x µ ∂x λ ∂x ν ν ∂x µ ∂x λ ∂2 x ν
Γνµ0 λ0 = µ 0 λ 0 ν
Γµλ − µ0 λ0 µ λ (2.18)
∂x ∂x ∂x ∂x ∂x ∂x ∂x
If we set the transform of Γνµλ as such, ∇µ V ν is guaranteed to transform
like a tensor. Covariant derivatives of tensors with lowered indices are
defined in a similar manner; for a 1-form,

∇µ ων = ∂µ ων − Γλµν ωλ (2.19)
10 fang

Let’s return to these connection coefficients I introduced. The connec-

tion coefficients of the Levi-Civita connection (or pseudo-Riemannian
connection) expressed in a coordinate basis are also called Christof-
fel symbols; feel free to use the terms interchangeably for now.2 The 2
The third name for these objects is the
Christoffel symbol is symmetric in its lower indices:3 Levi-Civita
3 connection
Luckily, there coefficients.
are ways to automate:) the
calculating of Christoffel symbols and
1 σρ other quantities; see Mathematica note-
Γσµν = g (∂µ gνρ + ∂ν gρµ − ∂ρ gµν ) (2.20) book.
2

Using these, we have that the covariant derivative of the metric and its
inverse are always zero, an incredibly useful behavior known as metric
compatibility:
∇σ gµν = 0, ∇σ gµν = 0 (2.21)

If there exists a metric-compatible and torsion-free (Γλµν = Γλ(µν) ) con-

nection, it must be of the form of a Christoffel symbol.

The Riemann curvature tensor

I briefly gave a definition of a manifold in the footnote on page 3, but

let’s talk about it more. One example of a manifold could be the Earth,
which is so large that we (puny) observers may reasonably assume
that it’s flat.4 Some other examples of manifolds are n-dimensional 4
[Insert flat-Earther joke of your choice.]
flat space Rn , or the n-dimensional sphere Sn . R1 is the real line, R2 is
a plane, etc.; S1 is a circle, S2 a sphere, etc.5 5
For example, if we use θ and φ as coor-
Manifolds can be (and often are) curved, which means we need to dinates for a sphere S2 with radius r = 1,
the metric is ds2 = dθ 2 + sin2 θdφ2 .
whip out some of the formalisms we’ve already laid out in the pre-
ceding pages for a non-Cartesian basis. As I said on page 3, all "infor-
mation" about the curvature of a manifold is contained in the metric.
How do we get this information back? Turns out, the information
about curvature is contained in a four-component tensor which we
call the Riemann curvature tensor:

≡ ∂µ Γνσ − ∂ν Γµσ + Γµλ Γλνσ − Γνλ Γλµσ

ρ ρ ρ ρ ρ
R σµν (2.22)

This tensor is built up from non-tensors (partial derivatives and Christof-

fel symbols) but miraculously transforms as a tensor. Importantly, the
Riemann tensor works well as a measure of curvature because all of
its components vanish iff the space is flat; "flat" meaning that ∃ a global
coordinate system in which the metric components are everywhere
constant. We can give names to two particularly useful contractions of
the Riemann tensor: the Ricci tensor is given by Rµν = Rλµλν , and the
µ
Ricci scalar by R = R µ = gµν Rµν .
While there are many possible components of the Riemann tensor
due to its many indices, it turns out that its symmetry properties sim-
 a brief sketch of relativity 11

plify calculations a fair bit. In particular:

Rρσµν = − Rσρµν = − Rρσνµ

Rρσµν = Rµνρσ
Rρ[σµν] = Rρσµν + Rρµνσ + Rρνσµ = 0
R[ρσµν] = 0

This means that the Ricci tensor is symmetric, Rµν = Rνµ . Additionally
the Riemann tensor behaves according to a differential identity called
the Bianchi identity:
∇[λ Rρσ]µν = 0 (2.23)
If we define a new tensor, called the Einstein tensor (’Einstein’ denot-
ing its importance for reasons which will soon become clear!),

1
Gµν ≡ Rµν − Rgµν (2.24)
2
the Bianchi identity implies that the divergence of this tensor vanishes.

1
∇µ Rρµ = ∇ρ R ⇐⇒ ∇µ Gµν = 0 (2.25)
2

Parallel transport and geodesics

To be brief and informal, a geodesic is "the shortest path between two
points."6 More formally, the geodesic extremizes the length functional 6
For example, Earth’s "great circles" are
the geodesics of the Earth.
R
ds. For a path x µ (λ), the infinitesimal distance along the curve is
given by s 
 dx µ dx ν 
ds =  gµν dλ (2.26)
dλ dλ 
R
which, if integrated, gives the full length of the curve L = ds. Find-
ing extrema7 of L requires calculus of variations, which I am going to 7
I say "extrema" and not "minima" be-
skip for very rational reasons, but the upshot is that x µ (λ) is a geodesic cause massive test particles move on
geodesics of maximum proper time. In
iff it satisfies the geodesic equation: the twin paradox, a pair of twins take
two paths through flat spacetime, one
d2 x µ ρ
µ dx dx
σ
along a geodesic (sitting on their couch
2
+ Γρσ =0 (2.27) at home) and the other traveling into
dλ dλ dλ
space and back. The couch-bound twin
This turns out to only hold iff λ is a affine parameter, i.e. if it can is older upon their reunion, because
geodesics maximize proper time.
be related to the proper time by λ = aτ + b.8 Geodesics are extremely 8
We nearly always use τ as the affine pa-
important because, in GR, a freely moving or falling body (a test body) rameter, when parameterizing timelike
always moves along a geodesic in a manner governed by the geodesic geodesics.

equation (2.27). If the body is massless, ds2 = 0 and the geodesic

is null, and if the body is massive then ds2 < 0 and the geodesic is
timelike. Upshot: in GR, particles move along geodesics unless acted
on by an external force, and the geodesic equation is "our" version of
Newton’s second law.
 3
General relativity

Finally, the promised land! The core of the theory of general relativity
can be given in two statements.

(1) Spacetime is a curved pseudo-Riemannian manifold with a metric

of Lorentzian signature (− + ++).

(2) The relationship between matter and the curvature of spacetime is

wholly contained in the Einstein equation:

1
Rµν − Rgµν = 8πGTµν (3.1)
2

The left-hand side of Eq. 3.1 is the Einstein tensor Gµν , defined in Eq.
(2.24).1 Tµν is a symmetric tensor called the stress-energy tensor2 , and 1
Be careful to not confuse it with G
contains all information about the energy and momentum of matter which is the familiar Newtonian gravita-
tional constant (and not the trace of Gµν ).
fields, which are the source of gravity. So Eq. 3.1 basically says: 2
Also known as the energy-momentum
tensor.
LHS ↔ curvature of spacetime;
RHS ↔ energy and momentum contained in the spacetime.

Eq. (2.24) functions as our "equation of motion" for the metric. One el-
egant and succinct formulation is "Spacetime tells matter how to move,
matter tells spacetime how to curve."3 3
Quote attributed to J.A. Wheeler.
What is the stress-energy tensor, exactly? The components of T µν
are given by the flux of the µth component of momentum pµ across
a surface of constant x ν (in the νth direction). But what does this
mean? Consider a perfect fluid which is isotropic4 in its rest frame. In 4
Isotropy = uniform in all orientations;
any frame, the stress-energy tensor of this perfect fluid can be speci- no viscosity or heat flow.

fied entirely in terms of its rest-frame energy density ρ and rest-frame

pressure p, which is isotropic and equal in all directions:

Tµν = (ρ + p)Uµ Uν + pgµν (3.2)

Normalizing by gµν Uµ Uν = −1 and raising the second index, this

 a brief sketch of relativity 13

becomes:  
−ρ 0 0 0
 0 p 0 0
Tµ ν = (3.3)
 
p

 0 0 0
0 0 0 p
We claim that T µν contains all information about energy and momen-
tum, so it must also contain the conservation laws. This is done by
asserting that the covariant divergence of T µν vanishes:

∇µ T µν = 0 (3.4)

Since the Bianchi identity (Eq 2.23) guarantees that the divergence of
the Einstein tensor vanishes, Eq. (3.1) guarantees energy-momentum
conservation.
Let us remind ourselves that Einstein’s formulation of GR should
be a natural extension of Newtonian gravity, and should approximate
Newton’s laws in some limiting case. For the Newtonian limit, we
assume that (1) the particles are moving slowly, (2) the gravitational
field is weak, and (3) that it is static. Doing the requisite math, and
using Poisson’s equation for the Newtonian potential ∇2 Φ = 4πGρ,
we will find that Φ = − GM r , recovering the Newtonian gravitational
law.

3.1 Schwarzschild solution

Unfortunately the TA lecture has probably gone on far too long by this
point, so I won’t be able to cover the Schwarzschild solution and black
holes in any depth, so consider this supplementary reading.
How do we even go about solving an equation like (3.1)? The
Schwarzschild metric is one idealized solution to the Einstein field
equations, which describes the gravitational field outside of a spher-
ically symmetric mass, assuming that the electric charge of the mass,
angular momentum of the mass, and the cosmological constant are all
zero.

2GM −1 2
   
2GM
ds2 = − 1 − dt2 + 1 − dr + r2 (dθ 2 + sin2 θdφ2 )
r r
(3.5)
where M is the mass of the body with radius r. Perhaps it would
be interest for you to see the nonzero Christoffels written out for this
metric:
Γrtt = GM
(r − 2GM) Γrrr = − GM Γttr = GM
r3 r (r −2GM) r (r −2GM)
Γrθ 1
Γrθθ = −(r − 2GM ) Γrφ = 1r
θ = φ
r
Γrφφ = −(r − 2GM ) sin2 θ Γθφφ = − sin θ cos θ Γθφ = cos
φ θ
sin θ
(3.6)
14 fang

As it turns out, the Schwarzschild metric is the unique solution to Ein-

stein’s equation in a vacuum with a spherically symmetric matter dis-
tribution.5 As such, the Schwarzschild solution is capable of describ- 5
This is known as Birkhoff’s theorem.
ing things as un-exotic as our Solar System (notably the precession of
Mercury’s orbit, a problem of definite historical interest), as well as
things like black holes.
It turns out that the metric components blow up (go to infinity) at
r = 0 and r = 2GM (can you see why?), two points which we clas-
sify as singularities. True singularities point to an actual breakdown
of the geometry, while coordinate singularities, which are just a side
effect of choosing a bad coordinate system. In this case, r = 0 is a
true singularity, while r = 2GM is a coordinate singularity which also
happens to specify the event horizon of a black hole.6 6
In order to show this, we can make
What happens at r = 2GM? To an external observer, a clock falling a coordinate transformation to what is
known as the Kruskal coordinates and
into a black hole will appear to move more and more slowly as it observe that nothing blows up at r =
approaches r, never actually crossing the surface; but for the observer 2GM in that basis.
who is actually falling into the black hole, time will progress at its
usual rate since their watch will be inertial in their frame. They will
readily pass through r = 2GM and fall straight into r = 0. This is
in fact inevitable, as r becomes a timelike coordinate for r > 2GM;
falling to the center of the black hole is identical and equivalent to
moving forwards in time. Assuming that you haven’t yet been torn to
unrecognizable strips of flesh by tidal forces, you would travel along a
geodesic which maximizes the proper time τ.
There are other metric solutions to slightly more complex black
holes, which build from the Schwarzschild black hole—such as the
Reissner-Nordstrom metric for charged black holes, or the Kerr metric
for rotating black holes.

3.2 Cosmology

In cosmology we make the simplifying assumption that the Universe is

homogeneous and isotropic,7 from which we can claim a "rest frame" 7
Whether it is either, both, or neither of
of the Universe and write down the Robertson-Walker metric in co- these things is a matter of considerable
active research today.
moving coordinates:

dr2
 
2 2 2
ds = −dt + a (t) + r2 (dθ 2 + sin2 θdφ2 ) (3.7)
1 − kr2

where a(t) is the scale factor, a measure of the relative expansion of

the Universe, and

−1 for an open universe;


k= 0 for a flat universe;



+1 for a closed universe.
 a brief sketch of relativity 15

This is the only possible homogeneous and isotropic metric, and we

just need to solve for a(t) using Einstein’s equation, in which case
Einstein’s equation becomes two differential equations known as the
Friedmann equations:
 2
ȧ 8πG k
= ρ− 2 (3.8)
a 3 a
ä 4πG
=− (ρ + 3ρ) (3.9)
a 3
These equations determine the evolution of RW metrics, in what we
call the Friedmann-Robertson-Walker cosmology. From solving the
Friedmann equations we can begin to do things such as examine the
behavior of different universes (matter-, radiation-, or vacuum-dominated),
characterize their evolution and their final fates.
4
Bibliography

S. M. Carroll. A No-Nonsense Introduction to General Rel-

ativity. URL https://www.astro.caltech.edu/~george/ay21/
readings/grtiny.pdf.

S. M. Carroll. Lecture Notes on General Relativity. arXiv preprint gr-

qc/9712019, 1997.

S. M. Carroll. Spacetime and Geometry: An Introduction to General Relativ-

ity. Cambridge University Press, 2019. doi: 10.1017/9781108770385.