Lecture 0: Differential Geometry, General Relativity, Classical Yang-Mills Theory
Lecture 0: Differential Geometry, General Relativity, Classical Yang-Mills Theory
Lecture 0: Differential Geometry, General Relativity, Classical Yang-Mills Theory
Let’s start with a very condensed review of some of the elements of Differ-
ential Geometry, General Relativity, and Yang-Mills Theory. This should not
be confused with an introduction to these subjects. If you plan to be a theo-
retical physicist, it is a good idea to eventually learn about these topics from
a somewhat more mathematically sophisticated perspective than the one pre-
sented here: I recommend the relevant sections of Wald’s “General Relativity”
and Arnold’s “Mathematical Methods of Classical Mechanics.”
Let me start with a few notational remarks. If you find any of my nota-
tion confusing, please let me know – I have probably made a mistake, forgotten
to explain something, or in any case will be happy to explain it again! I will
use the notation ∂v α /∂xβ = ∂β v α = v α,β for ordinary partial derivatives, and
∇β v α = v α;β for covariant derivatives. I will follow the Einstein summation
convention: whenever a term contains a repeated index, with one up and one
down, a sum over that index is implied. I will take the “timelike” and “space-
like” components of the metric to have negative and positive signs, respectively:
e.g. ds2 = −dt2 + d~x2 . More generally, I will use GR sign conventions that
match those in the textbooks by Misner-Thorne-Wheeler (MTW), Wald, and
Weinberg; these conventions are nicely summarized on the inside cover of MTW.
1 Differential Geometry
1.1 Manifolds and tensors.
Informally, an n-dimensional manifold M is a space that “looks like” Rn in the
neighborhood of every point. For example, the plane, the surface of a ball, and
the surface of a donut are all examples of 2-dimensional manifolds: a sufficiently
small patch surrounding any point looks like a patch of R2 .
We are free to choose many different coordinate systems to describe a given
part of a manifold; let xα and x̃α be two such coordinate systems. A tensor
of rank (m, n) has m upper (“contravariant”) indices and n lower (“covariant”)
indices; if the components in the x-coordinate system are Tβα11...β
...αm
n
, then the
components in the x̃-coordinate system are
1
Rank (1, 0) and (0, 1) tensors are called “contravariant vectors” and “covariant
vectors,” respectively.
∇µ Tβα11...β
...αm
n
= ∂µ Tβα11...β
...αm
n
ν...αm α1 ...ν
+Γα αm
µν Tβ1 ...βn + . . . + Γµν Tβ1 ...βn
1
α1 ...αm
−Γνµβ1 Tν...β n
− . . . − Γνµβn Tβα11...ν
...αm
. (3)
This does transform like a tensor. (Check this.) We will assume that the
connection has vanishing torsion: Γα α
βγ = Γγβ .
Now consider a curve x (τ ) with tangent vector uµ (τ ) = dxµ /dτ . A tensor
µ
α1 ...αm
Tβ1 ...βn (τ ) is parallel transported if it evolves along the curve as
d Tβα11...β
...αm h
ν...αm αm α1 ...ν
n
+ u µ Γα
µν Tβ1 ...βn + . . . + Γµν Tβ1 ...βn
1
dτ i
α1 ...αm α1 ...αm
−Γνµβ1 Tν...β n
− . . . − Γ ν
T
µβn β1 ...ν =0 (4)
or, in other words, if it is covariantly constant along the direction of the curve
uµ ∇µ Tβα11...β
...αm
n
= 0. (5)
duα
+ Γα µ ν
µν u u = 0 ⇔ uµ ∇µ uα = 0. (6)
dτ
This is called the geodesic equation.
Unlike ordinary partial derivatives, covariant derivatives do not commute.
The Riemann curvature tensor Rαβγδ may be defined as a measure of their
failure to commute
(∇γ ∇δ − ∇δ ∇γ )v α = Rαβγδ v β , (7)
from which we find (check this)
Rαβγδ = Γα α α σ α σ
βδ,γ − Γβγ,δ + Γγσ Γβδ − Γδσ Γβγ . (8)
2
Eq. (7) is called the Ricci identity. More generally, the commutator of two
covariant derivatives acting on a tensor is given by (check this):
(∇γ ∇δ − ∇δ ∇γ )Tβα11...β
...αm
n
= Rαµγδ
1
Tβµ...α m
1 ...βn
+ . . . + Rαµγδ
m
Tβα11...β
...µ
n
− Rµβ1 γδ Tµ...β
α1 ...αm
n
− . . . − Rµβn γδ Tβα11...µ
...αm
. (9)
The Riemann tensor may be contracted to give the Ricci curvature tensor:
R ≡ g αβ Rαβ . (11)
1 αδ
Γα
βγ = g [gβδ,γ + gγδ,β − gβγ,δ ]. (12)
2
(Derive this. Hint: start from gβδ;γ + gγδ;β − gβγ;δ = 0.)
We can define the length of a curve xµ (τ ), connecting two points A and B;
RB
it is A ds where the line element is
Now geodesics have another special property: they are paths from A to B whose
length is stationary under small variations of the path. This is reminiscent of
classical mechanics, in which classical trajectories are paths from A to B whose
action is stationary under small variations of the path. Indeed, this perspective
gives a different (and often more convenient) route to the geodesic equation, as
the Euler-Lagrange equations
∂L d ∂L 1
µ
= µ
L = gαβ ẋα ẋβ (14)
∂x dτ ∂ ẋ 2
where, in this equation, ẋµ = dxµ /dτ , where τ is an affine parameter along the
curve (which, for a massive particle, is proportional to its proper time). Check
that Eqs. (6) and (14) are equivalent.
3
Before moving on to General Relativity, note that the Riemann tensor sat-
isfies the Bianchi identity:
and Rαβγδ = gαµ Rµβγδ has a number of symmetries under index permutation:
2 General Relativity
General Relativity is defined by the action
4 √
R − 2Λ
Z
S = d x −g + Lm , (17)
16πGN
where g is the determinant of gµν , GN is Newton’s gravitational constant, and Λ
is the cosmological constant. The action is a functional of the gravitational field
gµν , as well as the various matter fields (which are all contained in Lm ). We
obtain the classical equations of motion for the matter fields Φ by demanding
that the action must be stationary (δS = 0) under an arbitrary variation δΦ
around a classical solution Φ.
We could obtain the Einstein equations (the classical equations of motion
for the metric) in a similar way, by demanding that δS = 0 under an arbitrary
variation δgµν around a classical solution gµν ; but this turns out to be rather
cumbersome, and there is an easier and better way: the Palatini method. In
the Palatini method, we start by forgetting the relationship (12) between the
metric and the connection, and regarding g µν and Γα βγ as independent fields to
be varied. Thus we write the action in the form
αβ
√
g Rαβ − 2Λ
Z
S = d4 x −g + Lm , (18)
16πGN
and note that Rαβ is independent of the metric (it only depends on Γα βγ ), while
everything else in the action (except for Rαβ ) is independent of Γα
βγ .
We start by varying Γα α
βγ . Note that δΓβγ is a tensor, even though Γβγ is
α
4
Before proceeding with the metric variation, we need the following two facts.
First, if we make a variation δgµν , what is the induced variation δg µν in the
inverse metric? The answer is obtained by varying the relation g αβ gβγ = δγα to
obtain gγβ δg βα +g αβ δgβγ = 0. Second, if we make a variation δgµν in the metric
(or in any invertible matrix, for that matter), what is the induced variation δg
in its determinant? The answer is given by Jacobi’s formula from linear algebra:
δg = gg αβ δgαβ ; also note that this result can be derived from another famous
fact in linear algebra: for a square matrix A, DetA = exp[Tr(ln A)]. With these
facts, we can apply a variation in δg µν to Eq. (18)
1
√ µν Rµν − 2 gµν R + Λgµν
1 ∂Lm
Z
4
δS = d x −g δg − gµν Lm + µν (20)
16πGN 2 ∂g
∂Lm
Tµν = gµν Lm − 2 . (23)
∂g µν
T αβ;β = 0. (24)
1
Tµν = ϕ,µ ϕ,ν − gµν [ g αβ ϕ,α ϕ,β + V (ϕ)]. (26)
2
A final point: in the absence of external (non-gravitational) forces, test
particles move along geodesics of the metric gµν .
5
3 Yang-Mill Theory
Let us briefly look at Yang-Mills theory. You can learn a lot by comparing its
structure to that of General Relativity.
We first introduce the gauge covariant derivative
Dµ ≡ ∂µ + igAµ . (27)
Then we use the Ricci identity to define the Yang-Mills field strength Fµν as
the curvature of the gauge connection Aµ :
This implies
Fµν = ∂µ Aν − ∂ν Aµ + ig[Aµ , Aν ]. (29)
Under a Yang-Mills gauge transformation U (x), we want the covariant derivative
operator to transform “nicely”:
Dµ → D̃µ = U Dµ U −1 (30)
i i
Aµ → õ = U Aµ U −1 − U (∂µ (U −1 )) = U Aµ U −1 + (∂µ U )U −1 . (31)
g g
From Eqs. (28, 30) we see that Fµν also transforms nicely:
Φ → Φ̃ = U Φ Ψ → Ψ̃ = U Ψ; (34)
then we can construct gauge invariant scalar and spinor kinetic terms as follows:
Aµ = Aaµ T a , a
Fµν = Fµν T a, [T a , T b ] = if abc T c (36)
6
which are chosen to be orthonormal in the following sense
Tr(T a T b ) = ξδ ab (37)
Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.
Alternative Proxies: