Introduction to General Relativity

Lectured by Jeong-Hyuck Park

Department of Physics, Sogang University, 35 Baekbeom-ro, Mapo-gu, Seoul 04107, Korea

Scribed by Hyunwoo Kwon

Department of Mathematics, Sogang University, 35 Baekbeom-ro, Mapo-gu, Seoul 04107, Korea

Abstract

These are lecture notes that I typed up for Professor Jeong-Hyuck Park’s course (PHY4010)
on General Relativity in Spring 2017. I should note that these notes are not polished and
hence might be riddled with errors. If you notice any typos or errors, please do contact me at
willkwon@sogang.ac.kr1

1 Revised: April 26, 2018

1
Contents
1 Lorentz group and Lorentz transformation 3
1.1 Lorentz transformation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2 The Lorentz group . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.3 The Einstein convention . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.4 Baker-Campbell-Hausdorff formula . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.5 Covariant transformation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

2 General Relativity 21
2.1 The geodesic equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.2 Tensor and covariant derivatives . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.3 Existence of Metric compatibility . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2.4 The Einstein field equation, Tensor density . . . . . . . . . . . . . . . . . . . . . 32
2.5 Riemann curvature and spacetime . . . . . . . . . . . . . . . . . . . . . . . . . . 39
2.6 Some solutions of Einstein’s equation . . . . . . . . . . . . . . . . . . . . . . . . 43

Bibliography 47

2
Lorentz group and Lorentz transformation
1.1 Lorentz transformation
1
For temporary, we introduce some notation. Let

R1+3 = {(t, x1 , . . . , x3 ) : t, xi ∈ R} .

We regard t as a time variable and (x1 , . . . , x3 ) as a spatial variable. In the case of vector field,
we use boldface notation. For E : R1+3 → R3 , we define

i
∑ ∂ k
(∇ × E) = εijk E ,
∂xj
j,k
∑ ∂E i
∇·E = ,
i
∂xi

the curl of E and the divergence of E. Here εijk is the Levi-Civita symbol defined by


1 (i, j, k) is a even permutation of (1, 2, 3)
εijk = −1 (i, j, k) is a odd permutation of (1, 2, 3)


0 otherwise.

We take curl and divergence in spatial variable only. We denote ∇ the gradient defined by
( )
∂f ∂f ∂f
∇f = , ,
∂x1 ∂x2 ∂x3

and ∇2 f the Laplacian defined by

∑
3
∂2f
∇2 f = .
i=1
∂x2i

Consider the following Maxwell equation in the differential form:



 ∇ · E = ερ0

∇ · B = 0

 ∇ × E = − ∂B

 (
∂t )
∇ × B = µ0 J + ε0 ∂E
∂t

Here E : R1+3 → R3 , B : R1+3 → R3 denote the electric field, magnetic field, respectively. ρ
denotes the electric charge density and J denotes the current density. We denote ε0 the permittivity
of free space and µ0 the permeability of free space.

Exercise 1.1. If ρ = 0 and J = 0, prove that E and B satisfies the equation □F = 0, where
( 2 )
∂ 1 2
□= − ∇
∂t2 c2

and c = √ 1
µ 0 ε0 .

We call □ the wave operator or the d’Alembertian on R1+3 .

3
1. Lorentz group and Lorentz transformation

Remark. Let us recall how to solve the wave equation in R1+1 :

∂2u 1 ∂2u
− = 0.
∂t2 c2 ∂x2
Now we consider the following new set of variables ξ = x+ct, η = x−ct. Define v (ξ, η) = u (x, t).
Then the change of variables formula shows that

∂2v
= 0.
∂ξ∂η
Integrating this twice gives
v (ξ, η) = F (ξ) + G (η) ,
i.e.,
u (x, t) = F (x + ct) + G (x − ct) .
For the case of higher dimension, see [1, 2].
In what follows, c denote the speed of light. Unfortunately, the Maxwell equation is not Galieli
invariance. We need some symmetry to study the Maxwell equation well. The Lorentz transform
gives a positive answer for this question. First, let us explain the Lorentz transform in physical way.
Consider two parallel mirrors and let L to be the distance between mirrors. Then light travels
once in ∆t = 2L c .
Now we consider a train moving with a constant speed v in the x-direction. Then if we look the
light outside of the train in the coordinate (t, x, y, z) with (t, x, 0, 0) we have
√ ( )2
2 L2 + v 2 ∆t 2
= c,
∆t
Here ∆t denotes the time which light travels between mirrors. Solving this equation, we obtain
2L
∆t = √ .
c2 − v2
If we see the light in the train in the moving coordinate (t′ , x′ , y ′ , z ′ ) with (t′ , 0, 0, 0), we have
2L
∆t′ = .
c
Since L is invariant, we obtain √
c∆t′ = c2 − v 2 (∆t) ,
i.e.,
∆t′
∆t = √ ,
1 − η2
where η = vc . We obtain the time-dillation. Write γ = √ 1
. Then
1−η 2

(t′ , 0, 0, 0) ⇐⇒ (γt′ , vγt′ , 0, 0) = (t, x, y, z) .

This is the special case of Lorentz transformation.

Now let’s move to more general case. Consider a fixed frame S and moving frame S ′ with
velocity v in the x-direction.
Consider an event A in (t′A , x′A , yA′ ′
, zA ) if we write it in S ′ frame and in (tA , xA , yA , zA ) if
we write it in S frame. Also consider an event B in (t′B , 0, 0, 0) if we write it in S ′ frame and
(tB , xB , 0, 0) if we write it in S frame.

4
 1.1. Lorentz transformation

Now consider a light starting from A to the B. Then the light travels
√
(x′A ) + (yA′ )2 + (z ′ )2
2
′ ′ A
tB = tA +
c
in S ′ and
√
2
(xA − vtB ) + yA
2 + z2
A
tB = tA + (1.1)
c
xB = vtB .

Then solve the equation (1.1) to get

√
2
c2 tA − vxA + (c2 tA − vxA ) − (c2 − v 2 ) (c2 t2A − x2A − yA
2 − z2 )
A
tB =
c2 − v 2
{ √ }
η 1 2 1 2
= γ 2 t A − xA + (xA − vtA ) + 2 (yA + zA 2) .
c c γ

Since tB = γt′B ,
√
{ √ } (x′A ) + (yA
′ ) + (z ′ )
2 2 2
η 1 2 1 2 2 ) = t′ + A
γ t A − xA + (xA − vtA ) + 2 (yA + zA A .
c c γ c
′ ′
Since yA = yA and zA = zA
√
[ √
(x′A ) + (yA ) + (zA ) η ] 1
2 2 2
′ 2
tA + = γ tA − xA + (xA − vtA ) + yA
2 + z2 .
A
c c c
The equation holds for all y, z. Hence we obtain
 ′ ( )

 t = γ t − ηc x
 ′
x = γ (x − vt)

 y′ = y
 ′
z = z.

In the matrix form, we write

 ′   
t γ − ηc 0 0 t
x′  − η γ 0 0  
 ′ =  c  x .
y   0 0 1 0  y 
z′ 0 0 0 1 z

We call this transform as the Lorentz transform.

Observe that γ 2 − (γη) = 1 since γ = √
2 1
. Then there is ϕ such that cosh ϕ = γ and
1−η 2
sinh ϕ = γη. So we can write
[ ] [ ][ ]
ct cosh ϕ sinh ϕ ct′
= .
x sinh ϕ cosh ϕ x′

It seems similar to the rotation matrix in R2 . In the next section, we analyze this kind of matrices
in detail.

5
1. Lorentz group and Lorentz transformation

1.2 The Lorentz group

Let us compare two matrices
[ ]
cosh ϕ sinh ϕ
A= ,
sinh ϕ cosh ϕ
[ ]
cos θ − sin θ
B= .
sin θ cos θ

In the case of B, B T B = I and [ ′] [ ]

x x
= B
y′ y
implies
(x′ ) + (y ′ ) = x2 + y 2 .
2 2

Indeed,
[ ]
[ ] x′
(x′ ) + (y ′ ) = x′
2 2
y′
y′
[ ]
[ ] x
= x y BT B
y
= x2 + y 2 .

We can regard B T B = I as B T IB = I, i.e., B stablizes I.

By considering cosh2 ϕ − sinh2 ϕ = 1, we obtain the following relation holds:
[ ] [ ]
T −1 0 −1 0
A A= .
0 1 0 1

Also by the similar reason as before, we obtain

− (ct) + x2 = − (c′ t′ ) + (x′ )

2 2 2

if [ ] [ ′]
ct ct
=A ′ .
x x
Write [ ]
cos θ − sin θ
B (θ) = .
sin θ cos θ
Then entries by entries differentiation with respect to θ gives
[ ]
0 −1
B ′ (θ) = B (θ) .
1 0

Inductively, we see [ ]n
dn 0 −1
B (θ) = B (θ) .
dθn 1 0
Evaluating at θ = 0, we get  [ ]n
dn  0 −1
B = .
dθn θ=0 1 0

6
 1.2. The Lorentz group

So we obtain the series expression of B (θ):

∞ [
∑ ]n
0 −1 1 n
B (θ) = θ
1 0 n!
n=0
∑∞ [ ]n
1 0 −θ
=
n=0
n! θ 0
([ ])
0 −θ
=: exp .
θ 0
[ ] [ ]
dB 0 −1 dB T 0 1
Note that = B (θ) and = B (θ). Hence
dθ 1 0 dθ −1 0

d ( T ) d ( T) dB
B IB = B B + BT = O.
dθ dθ dθ
Similarly, [ ]
dA 0 1
= A
dϕ 1 0
and inductively we have [ ]n
dn A 0 1
= A.
dϕn 1 0
[ ] [ ]
dA 0 1 dAT 0 1
Also dϕ = A and dϕ = AT . Hence
1 0 1 0
( [ ] ) ([ ][ ] [ ][ ])
d T −1 0 T 0 1 −1 0 −1 0 0 1
A A =A + A
dϕ 0 1 1 0 0 1 0 1 1 0
([ ] [ ])
0 1 0 −1
= AT + A
−1 0 1 0
= O.

Note that we have A (ϕ1 ) A (ϕ2 ) = A (ϕ1 + ϕ2 ) and B (θ1 ) B (θ2 ) = B (θ1 + θ2 ) by the
addition law of hyperbolic cosine, sine and cosine and sine, respectively.
Let us recall the definition of group.

Definition 1.2. A group (G, ·) is a set G with binary operation · : G × G → G satisfying

1. g1 · (g2 · g3 ) = (g1 · g2 ) · g3 for all g1 , g2 , g3 ∈ G;

2. there is e ∈ G such that g · e = e · g = g for all g ∈ G;

3. for any g ∈ G, there exists g ′ ∈ G such that g ′ · g = e.

There are many examples of groups. (Z, +), (R, +), (R\ {0} , ×) are groups.
We define { }
O (n) = A ∈ Mn×n (R) : AT A = In .
With matrix multiplication, it is a group and we call O (n) as the orthogonal group of order n.
There is a subgroup
SO (n) = {A ∈ O (n) : det (A) = 1}
of O (n) and we call this the special orthogonal group of order n.

7
1. Lorentz group and Lorentz transformation

We define O (p, q) the set of all (p + q) × (p + q) invertible matrices C satisfying

C T ηC = η
 
 
where η = diag −1, . . . , −1, 1, . . . , 1 . It is easy to check that O (p, q) is a group with matrix
| {z } | {z }
p−times q−times
multiplication. We call this group as indefinite-orthogonal group. Similarly, there is a subgroup

SO (p, q) = {A ∈ O (p, q) : det (A) = 1}

of O (p, q) and we call this the special indefinite-orthogonal group of order p and q.
In the special case SO (1, 3), we call this group as the Lorentz group. Usually, we write L as
an element of SO (1, 3). Under the relation
 ′  
ct ct
 x′  x
 ′ = L ,
y  y
z′ z

if we denote η = diag (−1, 1, 1, 1), then we have

= − (cdt′ ) + (dx′ ) + (dy ′ ) + (dz ′ )

2 2 2 2

2
= − (cdt) + dx2 + dy 2 + dz 2
2
In this sense, we denote ds2 = − (cdt) + dx2 + dy 2 + dz 2 and we regard this as the proper length
in our setting.

1.3 The Einstein convention

In order to describe the theory of general relativity, we need to introduce some notations and rules.
It might confuse in the first time.
From now on we write
( )
xµ = x0 , x1 , x2 , x3 , µ = 0, 1, 2, 3

and we call µ as the space-time indices. Here we regard x0 as ct. We call this coordinate as
Minkowskian 4-dimensional spacetime coordinate. We call
 
−1 0 0 0
 0 1 0 0
η=  0 0 1 0


0 0 0 1

as the Minkowski metric.

Now we study the Einstein conventions. There are four types of expressing matrices: Mµν ,
K µν , Lµν , Nµν . Multiplication is allowed if there is a contraction between indices. As an example,
M K is allowed since
λ
Mµν K νλ = (M L)µ
since ν is a contraction. Also M L is allowed since

Mµν Lνλ = (M L)µλ .

8
 1.3. The Einstein convention

However M M and M N are not allowed since between M and M , there is no contraction. Also
Mµν Nµν has no contraction.
If Mµν , K µν , Lµν , Nµν , then their inverses are

M −1 = M µν , K −1 = Kµν , L−1 = Lνµ , N −1 = N νµ

and their transpose are

( ) ( )µν ( )ν ( )µ
MT µν
KT LT µ
NT ν
.

From this
LT ηL = η ⇐⇒ ηµν = Lµλ ηλρ Lρν
We write
∂
= ∂µ .
∂xµ
Then note that
∂ (x′ )
µ
∂xµ ( −1 ) ν
= Lµν , ′ ν = L .
∂xν ∂ (x ) µ

Then by chain rule,

∂ ∂xν ( )ν
∂µ′ = ′ µ = ′µ
∂µ = L−1 µ ∂ν .
∂ (x ) ∂x
( )µν
For the case ηµν , η −1 ≡ η µν by definition. Metric interchanges the placement of indicies.
As an example
v µ ηµν = vν , η µν vν = v µ .
So dxν = ηµν dxν , ds2 = dxµ dxν ηµν ≡ dxµ dxµ .
If we can write some equality which can be written by Einstein convention, we write

ηµν = ηλρ Lµλ Lρν

instead of
ηµν = Lµλ ηλρ Lρν .
Note that
λ λ ρ ρ λ
(M1 M2 ) ν = (M1 ) ρ (M2 ) ν = (M2 ) ν (M1 ) ρ
λ λ ρ ρ λ
(M2 M1 ) ν = (M2 ) ρ (M1 ) ν = (M1 ) ν (M2 ) ρ .

So in general (M1 M2 ) ̸= (M2 M1 ).

Note that

δ µν = η µκ Lλν Lκρ ηλρ

= (ηλρ Lκρ η κµ ) Lλν
= Lµλ Lλν .
( )µ
Hence Lλµ = L−1 λ .
In this convention,
( )ν
∂µ′ = L−1 µ ∂ν = Lµν ∂ν
d (x′ ) = Lµν dxν .
µ

9
1. Lorentz group and Lorentz transformation

We define
⃗ · dx
dτ 2 = dt2 − dx ⃗
 2
and is called the proper time. In nature, dτ ≥ 0. So c ≥  dx
2 2 
dt .
dxµ dxµ
In some sense dt = dx0 is not inappropriate. From now on, we assume c ≡ 1. Note that
c

dxµ dxµ
= √ .
dτ dt 1 − ⃗v · ⃗v
So
d⃗x ⃗v
=√
dτ 1 − ⃗v 2
Note that
dt 1 1
=√ ≈ 1 + v2 .
dτ 1 − v2 2
Hence the momentum of
dxµ
Pµ = m ,
dτ
in particular,
1
P 0 = m + mv 2 + · · · .
2
We write the Maxwell equation in simple way. Let F = Fµν be a 2-form defined by
1
Ei = F0i , Bi = εijk F jk ,
2
where i, j, k runs {1, 2, 3}.
Then

∂t F0i − εijk ∂ j B k
1
= ∂0 F0i + εijk εiab ∂ j Fab
2
1
= ∂0 F0i + (δia δjb − δib δja )∂ j Fab
2
1
= ∂0 F0i + (∂ j Fij + ∂ j Fij )
2
= ∂0 F0i + ∂ j Fij
= −∂0 Fi0 + ∂ j Fij
= η µν ∂µ Fiν .
So η µν ∂λ µFiν = ji . Also ∇ · E = ρ implies ∂j F0i = ρ. So η µν ∂µ Fν = ρ. We write
J = (ρ, j1 , j2 , j3 ), then we get

η µν ∂µ Fλν = Jλ .
From ∇ · B = 0 and ∂t B + ∇ · ×E = 0, we obtain
∂λ Fνλ + ∂ν Fλµ + ∂λ Fµν = 0.
Hence, we rewrite the Maxwell equation in the following way:
{
η µν ∂µ Fλν = Jλ .
∂λ Fνλ + ∂ν Fλµ + ∂λ Fµν = 0.

10
 1.4. Baker-Campbell-Hausdorff formula

Recall

xµ 7→ (x′ ) = Lµν xν
µ

( )ν
∂µ 7→ ∂µ′ = L−1 µ ∂ν = Lµν ∂ν

under the Lorentz transformation. Hence

Fµν 7→ Lµλ Lνρ Fλρ

gives the Maxwell equation invariant under the Lorentz transformation.

Note that J µ 7→ J ′µ = Lµν J gives ∂λ F λµ = J µ to

∂λ′ F ′λµ = J µ

to ( )
Lλρ ∂ρ Lλκ Lµρ F κσ (x) = Lµν J ν .
On the LHS, this is equal to
δκρ ∂ρ F κσ Lµσ = Lµσ ∂ρ F ρσ .
So
Lµσ (∂ρ F ρσ − J σ ) = 0,
which reduces to the original Maxwell equation.

1.4 Baker-Campbell-Hausdorff formula

Recall
∑∞
An
eA = ,
n=0
n!
∞
∑ n
(−x)
ln (1 + x) = −
n=0
n!
( A
)
ln e = A.

We write
eA eB = eC .
If A and B commutes, then C = A + B. But this is not generally holds. However, we have the
following formula:

Theorem 1.3 (Baker-Campbell-Hausdorff formula).

∫ ∑∞
( ) 1
1( )n−1
ln e e A B
=B+ dt 1 − etadA eadB A
0 n=1
n
= A + B + 12 [A, B] + 12 1 1
[A[A, B]] + 12 1
[B, [B, A]] + + 24 [A, [[A, B], B]] + · · · .
(1.2)
which holds for an arbitrary pair of operators, A and B.

Proof. Introduce a real parameter, t ∈ R and define

( )
C(t) := ln etA eB , eC(t) = etA eB . (1.3)

11
1. Lorentz group and Lorentz transformation

We have [d ]
C(0) = B , d C(t)
dt e = AeC(t) , A= dt e
C(t)
e−C(t) . (1.4)
Further for an arbitrary, M ,

eadC(t) M = eC(t) M e−C(t) = etA eB M e−B e−tA = etadA eadB M , (1.5)

and hence, like (1.3),

eadC(t) = etadA eadB . (1.6)
Further, we set with one more real parameter, s ∈ R,
[ ]
∂ sC(t) −sC(t)
F (s, t) := e e . (1.7)
∂t

It is straightforward to see
F (0, t) = 0 , F (1, t) = A , (1.8)
and [ ] [ ]
∂ sC(t) dC(t) −sC(t) sadC(t) dC(t)
F (s, t) = e e =e . (1.9)
∂s dt dt
Hence,
dC(t) ∂
= e−sadC(t) F (s, t) , (1.10)
dt ∂s
and ∫ 1
∂
A = ds
F (s, t)
∫0 1 ∂s [ ]
dC(t)
= ds esadC(t)
dt
∫0 1 ∑ ∞ [ ]
sn n dC(t)
= ds [adC(t)] (1.11)
0 n! dt
n [ ]
n=0
∑∞
[adC(t)] dC(t)
=
(n + 1)! dt
n=0 [ ]
= G(adC(t)) dC(t)dt ,

where we put a function,

∞
ex − 1 ∑ xn
G(x) = = ,
x n=0
(n + 1)!
∑∞ ∞
x ln(ex ) ln[1 − (1 − ex )] ∑ (1 − ex )n−1
G(x)−1 = x =− xekx = x =− = .
e −1 e −1 1 − ex n=1
n
k=0
(1.12)
Therefore, we see, with (1.6),

dC(t) ∑ 1 ( )n−1
∞ ∑∞
1( )n−1
= 1 − eadC(t) A= 1 − etadA eadB A, (1.13)
dt n=1
n n=1
n

and finally,
∫ ∑∞
( ) 1
1( )n−1
ln eA eB = B + dt 1 − etadA eadB A. (1.14)
0 n=1
n
This completes the proof.

12
 1.4. Baker-Campbell-Hausdorff formula

In the quantum mechanical setting, since [x̂, p̂] = iℏ,

eax̂ ebp̂ = eax̂+bp̂+ 2 ℏab

i

If L is a Lorentz transformation, then L = eM , where M T η + ηM = 0. We will show it.

Note that
T T
(ηM1 ) = −ηM1 , (ηM2 ) = −ηM2 .

So
T T
(ηM1 M2 ) = M2T (ηM1 )
= −M2T ηM1
= ηM2 M1
̸= −ηM1 M2 .

However, we have
T
[η (M1 M2 − M2 M1 )] = −η (M1 M2 − M2 M1 )
= −η [M1 , M2 ] .
( [ ]) ( [ ])
0 1 0 1
In the case SO (2), we write exp θ and SO (1, 1), we write exp ϕ . We
[ ] −1 0 [ ] 1 0
0 1 0 1
call is a generator1 of SO (2). Similarly, is a generator of SO (1, 1).
−1 0 1 0
Let Ma , a = 1, . . . , n. We define

g = {θa Ma : θa ∈ R, a = 1, . . . , n}

c c
and [Ma , Mb ] = fab Mc for some fab . We call g as a Lie algebra and M1 , . . . , Mn are said to be
generator of g.
Roughly speaking, we define the Lie group as the exponential of Lie algebra. (For mathematical
completeness, see Appendix)
For L ∈ SO (p, q), we write L = exp (θM ). Note that LT ηL = η and dL dθ |θ=0 = M and
L |θ=0 = I. Taking derivative, we get

M T η + ηM = 0,
T
i.e., (ηM ) = −ηM .
If eθM ≡ L and M T η + ηM = 0, then d
dθ L = M L = LM . So

d ( T )
L ηL = LT M T ηL + LT ηM L
dθ ( )
= LT M T η + ηM L
= 0.

Hence LT ηL = η.

1 In the mathematical terminology, this is a basis

13
1. Lorentz group and Lorentz transformation

d(d−1)
Note that there are 2 matrices ηM which is anti-symmetric. In the case d = 4, 6 = 3 + 3.
Note that
 
0 −ϕ1 −ϕ2 −ϕ3
ϕ1 0 θ3 −θ2 
M = η −1 
ϕ2 −θ3

0 θ1 
ϕ3 θ2 −θ1 0
 
0 ϕ1 ϕ2 ϕ3
ϕ1 0 θ −θ 
=ϕ2 −θ3
3 2
.
0 θ1 
ϕ3 θ2 −θ1 0

Hence (θ1 , θ2 , θ3 ) represents a rotation and (ϕ1 , ϕ2 , ϕ3 ) represents boosts.

[ ] [ ] [ ]
1 0 1 2 0 −i 3 1 0
Remark. σ = ,σ = ,σ =
1 0 i 0 0 −1

iθa σ a

Recall SU (2), the set of all 2×2 matrices which entries are complex and U † U = 1 and |det U | = 1.
Write u = eθM . Taking derivatives, we have M † + M = 0. So
 θM 
e  = eθImM .

Roughly speaking, SU (2) and SO (3) are same.

1.5 Covariant transformation

Recall the definition of wave function. We say a complex-valued function ψ as a wave function if
∫
2
|ψ| dx = 1.

Write ψ = |ψ| eiφ . One may ask whether phase φ has physical meaning. Unfortunately, φ is
non-physical. To deal the theory of Quantum mechanics, complex-valued function is essential, not
a real-valued function.
For arbitrary function θ on spacetime ,consider a transformation ψ → ψ ′ = ψeiθ . Consider the
Schrödinger equation
1 2
iℏ∂t ψ = (−iℏ∇) ψ + V ψ.
2m
Write
1 2
Eψ = p⃗ ψ + V ψ
2m
and consider
∂
E → iℏ
∂t
p⃗ → −iℏ∇. (1.15)

Note that

Pµ = −iℏ∂µ = (−E, p⃗)

dxµ
Pµ = m = (E, p⃗) .
dτ

14
 1.5. Covariant transformation

So the above transformation (1.15) makes sense.

We want to claim that the transformation ψ 7→ ψ ′ does not change the law of physics. Note that
( )
∂µ ψ → ∂µ ψ ′ = ∂µ ψeiθ = (∂µ ψ + i∂µ θ) eiθ .

Then from this transformation, we can check that ψ ′ does not satisfies the Schrödinger’s equation
since ∂µ is not covariant. To overcome this difficulty, we introduce the “covariant derivative”

∇µ = ∂µ − iqAµ , q ≡ 1.

We choose A so that
∇µ ψ → ∇′µ ψ ′ = (∇µ ψ) eiθ .

Indeed,
( )
∇′µ ψ ′ = ∂µ − iA′µ ψ ′
( )
= ∂µ ψ ′ − iA′µ ψ ′
( )
= ∂µ eiθ ψ − iA′µ eiθ ψ
( )
= ∂µ ψ + i∂µ θ − iA′µ eiθ ψ.

So if we write
A′µ = Aµ − i∂µ θ,

∇µ behaves ‘covariant’ under the transformation

ψ 7→ ψeiθ
Aµ 7→ Aµ − ∂µ θ.

We call this kind of transform as U (1) gauge transform. Note that the transform does not change
the physics since |ψ| is invariant.
There are two kind of ‘symmetry’. One is local symmetry, that is, there exists a parameter
θ depending on spacetime point. This is however is non-physical symmetry. The other is global
symmetry, i.e., this is a parameter is constant. We call this symmetry as physical symmetry.
Although the gauge symmetry is not physical symmetry, the concept gauge symmetry is a
principle concept in 20th-21st physics.
So far we didn’t defined the terminology ‘symmetry’ in rigorous way. From now on, we define
‘symmetry’ and give the fundamental theorem due to Noether.
We say a transform is symmetry of the Lagrangian L if its action is invariant under the transfor-
mation. If the transform is infinitesimal, we say this symmetry as Noether symmetry.
Note that
∂L ∂L d
δL = δq + δ q̇ = K.
∂q ∂ q̇ dt

We state the Noether theorem informally.

Theorem 1.4. Let L be a Lagrangian. If there is a Noether symmetry, then there exists a conserved
charge
∂L
Q = a δq a − K.
∂ q̇

15
1. Lorentz group and Lorentz transformation

Note that
[ ]
dQ d ∂L a
= δq − K
dt dt ∂ q̇ a
[ ]
d ∂L ∂L dK
= δq a + a [δ q̇ a ] −
dt ∂ q̇ a ∂ q̇ dt
[ ( ) ] [ ]
d ∂L ∂L d
= − δq a
+ δL − K .
dt ∂ q̇ a ∂ q̇ a dt

Example 1.5. Consider q (t) 7→ q (t + a), time translation. Then

q (t + a) ≃ q (t) + aq̇ (t) .

So
∂L ∂L ∂L
δL = q̇ + q̈ , q̇ = .
∂q ∂ q̇ ∂t
Hence K = L and so
∂L a
Q= q̇ − L = H,
∂q a
the Hamiltonian. So by the Noether’s theorem, the Hamiltonian is conserved. So we call this kind
of symmetry as time symmetry.
Example 1.6. Let L = m2 ẋ , x 7→ x + c, spartial symmetry.
2

Q = mẋ. Then the momentum is conserved by the Noether’s theorem.

( 2 )
Example 1.7. L = m 2
2 ẋ + ẏ , consider rotation.
Note that
[ ] [ ]
x y
δ =
y −x
[ ] [ ]
ẋ ẏ
δ =
ẏ −ẋ

Then
δL = m (ẋδ ẋ + ẏδ ẏ) = 0.
So
∂L ∂L
Q= δ ẋ + δ ẏ
∂ ẋ ∂ ẏ
= mẋy − mẏx = px y − py x.

So by the Noether’s theorem, the angular momentum is conserved.

Observe the symmetry is global. There is no local dependence. Also, Noether charge satsifies

{Q, H} ≡ 0,

and
{Q, q} = δq, {Q, P } = δp, {Q, H} = 0.
Back to the original problem, to recover the covariance, we need
1 ( )2
iℏDt ψ = −iℏD
⃗ ψ,
2m

16
 1.5. Covariant transformation

i.e.,
ℏ2 ( ⃗ )2
iℏ (∂t ψ − iA0 ψ) = − ∇ − iA
⃗ ψ.
2m
So
ℏ2 ( ⃗ )2
iℏ∂t ψ = − ∇ − iA
⃗ ψ − ℏA0 ψ,
2m
where ( )
⃗
Aµ = ϕ, A

and ϕ is a columb potential and A ⃗ is magentic potential. We call this potential as electromagnetic
potential.
In Schrödinger’s equation, local symmetry, i.e., Gauge symmetry required. The solution
2
ψ (t, x, y, z) itself does not have physical meaning. However, |ψ| has a physical meaning, the
probability density. So we can transform ψ to e iθ(t,x,y,z)
ψ (t, x, y, z) ≡ ψ ′ (t, x, y, z), where
θ (t, x, y, z) is arbitrary function on spacetime. The θ depends on a point. So we can regard it as it
has a local symmetry. However, the classical Schrödinger equation does not transform in covariant
sense. So we replace ∂µ to Dµ = ∂µ − iAµ , where Aµ is vector potential (in the mathematical
way, gauge connection).satisfying

ψ 7→ ψ ′ = eiθ ψ
Aµ 7→ A′µ = Aµ + ∂µ θ

(gauge) transformation rule.

Is their any physical interpretation of Aµ ? No since the gauge transformation is non-physical
symmetry, i.e.,
η µν Aµ Aν ̸= η µν A′µ A′ν .
However, we can make a physical quantity from A,

Fµν := ∂µ Aν − ∂ν Aµ ,

field strength of A. It is gauge invariance since

′
Fµν = ∂µ A′ν − ∂ν A′µ
= ∂µ (Aν + ∂ν θ) − ∂ν (Aµ + ∂µ θ)
= ∂µ Aν − ∂ν Aµ + ∂µ ∂ν θ − ∂ν ∂µ θ
= Fµν .

It is physical quantity. Note that Fµν = −Fνµ . Actually, it is a electromagnetic field. Why we have
a light? Some people says it is due to gauge symmetry.
Note that eiθ ∈ U (1). We call this kind of symetry as U (1)-gauge symmetry.
Remark. We can generalize this concept to various groups. In the case of SU (3), it corresponds to
the strong force. In the case of SU (2), it corresponds to weak force. In the case of diffeomorphism,
it corresponds to general relativity.
SU (3) × SU (2) × U (1): Gauge symmetry of the standard model in particle physics, which is
quite accurate model in physics in theoratical way and experimental way. It is accepted as a theory
in nowadays, not just a model.
Remark. They are two kind of feature in quantum mechanics.
Schrödinger picture: |E (t) >, p̂, x̂, i.e., time variance on state, but operator does not change in
time.

17
1. Lorentz group and Lorentz transformation

Heisenberg picture: | E >, p̂ (t) , x̂ (t), i.e., operator change in time but the state does not
depend on time.

⟨ ⟩ ⟨ ⟩
Ψ | Â | E = E (t) | Â | E (t)
⟨ ⟩
= Ψ | eiℏĤ(t) Âe−iℏĤ(t) | Ψ
⟨ ⟩
= Ψ | Â (t) | Ψ

Hence in the sense of Heisenberg picture, we have gauge symmetry also.

The gauge symmetry is a principle in physics in nowadays although it is not physical.
Note that U (1) is abelian, however, SU (2), SU (3), diffeomorphism are not abelian. Actually,

Fµν := ∂µ Aν − ∂ν Aµ + [Aµ , Aν ]

in general.
We will figure out why this is natural. Consider a Gauge group G (ex. SU (3), SU (2), U (1),
etc)2 and transform
ψ (t, x, y, z) 7→ ψ ′ = gψ, g ∈ G
with Dµ ψ := ∂µ ψ + Aµ ψ. Note that Aµ is in general matrix. We require

Dµ′ ψ ′ = ∂µ ψ ′ + A′µ ψ ′ = gDµ ψ.

Then by computation

Dµ′ ψ ′ = ∂µ ψ ′ + A′µ ψ ′
= ∂µ gψ + g∂µ ψ + A′µ gψ
[ ( ) ]
= g ∂µ ψ + g −1 A′µ g + g −1 ∂µ g ψ .

So we require
Aµ = g −1 A′µ g + ∂ −1 ∂µ g.
So ( )
A′µ = g Aµ − g −1 ∂µ g g −1 .
Hence the transform must satisfy
( )
Aµ 7→ g Aµ − g −1 ∂µ g g −1 .

Under this transformation, it is easy to check by computation

′
Fµν = ∂µ Aν − ∂ν Aµ + [Aµ , Aν ] 7→ Fµν = gFµν g −1 .

So
Fµν ψ 7→ gFµν g −1 gψ = gFµν ψ.

2 In mathematical language, it is Lie group,

[ ]
∑
a
g (θ ) = exp a
θ Ta , {Ta } : Lie algebra,
a

We say it has a Gauge symmetry if θa (t, x, y, z): arbitrary / local parameter.

18
 1.5. Covariant transformation

Note that Dµ ψ 7→ gDµ ψ. So Dµ Dν ψ 7→ Dµ′ Dν′ = gDµ Dν ψ. Note that

Dµ Dν ψ = ∂µ (Dν ψ) + Aµ Dν ψ
= ∂µ (∂ν ψ + Aν ψ) + Aµ (∂ν ψ + Aν ψ)
= ∂µ ∂ν ψ + ∂µ Aν ψ + Aν ∂µ ψ + Aµ ∂ν ψ + Aµ Aν ψ.
So
[Dµ , Dν ] ψ = (∂µ Aν − ∂ν Aµ + Aµ Aν − Aν Aµ ) ψ = Fµν ψ.
Proposition 1.8. If A, B are in Lie algebra, then eA Be−A is in Lie algebra.
Proof. We claim that
eadA B = eA Be−A ,
where
1 1
eadA B = e[A,·] B = B + [A, B] + [A, [A, B]] + · · · + [A [A, · · · , [A, B]]] + · · · .
2 n!
Here
1 n 1
(adA) B = [A [A, · · · , [A, B]]] .
n! n!
Since the Lie algebra is closed under commutator, we are done. Hence it suffices to show the claim.
Note that
d ( tA −tA ) [ ]
e Be = A, etA Be−tA
dt
and
d ( tadA ) [ ]
e B = A, etadA B .
dt
Hence by the uniqueness of initial value problem, we are done.
For g ∈ G, write a
g (θ) = eθ Ta
.
Note that
a b a
g (θ1 ) g (θ2 ) = eθ1 Ta eθ2 Tb = ef (θ1 ,θ2 )Ta

= g (f (θ1 , θ2 )) .
−1
Here f a satisfies f a (θ, 0) = f a (0, θ) = θa . Since g (−θ) = e−θ
a
Ta
= g (θ) ,
f (θ, −θ) = f (−θ, θ) = 0.
a a

Then
∂µ gg −1 = ∂µ g (θ) g (−θ)
= ∂µ g (θ) g (−ϕ)|ϕ=θ
= ∂µ [g (θ) g (−ϕ)]|ϕ=θ
= ∂µ g (f (θ, −ϕ))|ϕ=θ

∂ 
= ∂µ f a (θ, −ϕ) g (f ) 
∂f a 
ϕ=θ

∂ 
= ∂µ f (θ, −ϕ)|ϕ=θ
a
g (f )
∂f a 
f =0
= ∂µ f a (θ, −ϕ)|ϕ=θ Ta ∈ Lie algebra.

19
1. Lorentz group and Lorentz transformation

So
d
g (θ) |θ=0 = Ta .
dθa
From this, Aµ 7→ A′µ ∈ G. So by previous proposition, Fµν ∈ G. Hence we can write
∑
Aµ = Aaµ Ta .
a

Remark. Note that

Fµν = ∂µ Aν − ∂ν Aµ
is Gauge invariant. However,

Fµν = ∂µ Aν − ∂ν Aµ + [Aµ , Aν ]

is Gauge covariant. Although it is in some sense is good, it is not observable by experiment.

However,
Tr (Fµν ) , Tr (Fµν F µν )
are Gauge invariant.
Yang-Mills first gives Gauge symmetry argument in the theory of standard model of particle
physics.
In the case of free particle, we write a Schrödinger equation
1 ( ( ))2
iℏ (∂t − iA0 ) ψ = −iℏ ∇ − iA
⃗ ψ,
2m
under Gauge transform consideration. We rewrite

ℏ2 ( )2
iℏ∂t ψ = − ∇ − iA
⃗ ψ − A0 ψ.
2m
Here A0 is a coulomb potential. This is so-called the magnetic Schrödinger equation.

20
General Relativity
2.1 The geodesic equation
2
In this chapter, we introduce the general relativity which is the Einstein’s theory of gravity. In the
previous chapter,

dθ2 = dxµ dxν ηµν

ηµν = diag (− + ++) .

Recall that freely falling frame is locally inertial frame since it has a tidal force effect. In this
setting, gravity disappears. Hence the special relativity works in this setting.
Let xµ denote a general coordinate system and consider y µ a coordinates of the locally inertial
frame at xµ = X µ . Then there is a local coordinate transform xµ (y), y µ (x), e.g., xµ (y ≡ 0) =
X µ . Then in the near y = 0,
ds2 = ηµν dy µ dy ν .
By considering the transformation, we have
 
2 ∂y µ  λ ∂y 
ν
ds = ηµν dx dxρ
∂xλ x=X ∂xρ x=X
(   )
∂y µ  ∂y ν 
= ηµν dxλ dxρ
∂xλ x=X ∂xρ x=X
= gµν (X) dxµ dxν .

Here 
∂y a ∂y b 
gµν (x) = ηab .
∂xλ ∂xρ y=0
Hence to describe the physics, we need to consider the above metric. Hence

ds2 = gµν (x) dxµ dxν .

Note that gµν is metric and this is the only geometric quantity. This describes the gravitational
effect in physics. Our aim is to study the PDE which gives a solution gµν , so called the Einstein
field equation.
If we consider a R2 which is in the standard coordinate system and polar coordinate system,
then one metric is constant but the other has variable. If one can transform the metric into Lorentz
metric, then the surface is flat. However, it is not. We will figure out in detail.
From now on, we denote
dy a
= eµa .
dxµ
We call this by vielbein(vierbein). We denote its inverse matrix by eaµ . Note that

eaµ eµb = δab

eνa eaµ = δνµ .

Then we can write

gµν = eµa eνa .
By using this convention, it is easy to see that

eaµ = g µν eνb ηba

21
2. General Relativity

holds. Indeed, ( )
g µν eνb ηba eλa = g µν gνλ = δ µλ .
dy µ
Under coordinate transformations, y µ → xµ (y), the velocity of a particle, ẏ µ = dτ , transforms
as
∂xµ
ẏ µ =⇒ ẋµ = ẏ ,
∂y ν
where τ denotes the proper time.
a) Show that the acceleration transforms to satisfy

∂xµ ν µ ∂xµ ∂ 2 y σ
ÿ = ẍ + ẋα ẋβ .
∂y ν ∂y σ ∂xα ∂xβ

From the invariance of the proper length,

ds2 = gµν (x)dxµ dxν = g̃µν (y)dy µ dy ν ,

the metric transforms as

∂y ρ ∂y σ
gµν = g̃ρσ .
∂xµ ∂xν
We may derive straightforwardly

∂λ gµν = ∂˜α g̃ρσ ∂λ y α ∂µ y ρ ∂ν y σ + g̃ρσ ∂λ ∂µ y ρ ∂ν y σ + g̃ρσ ∂µ y ρ ∂λ ∂ν y σ ,

where we put ∂˜α = ∂

∂y α .

b) Using the above result, derive

( )
∂µ gλν +∂ν gµλ −∂λ gµν = ∂˜α g̃γβ + ∂˜β g̃αγ − ∂˜γ g̃αβ ∂λ y γ ∂µ y α ∂ν y β +2g̃ρσ ∂µ ∂ν y ρ ∂λ y σ .

c) Using the above results, derive the transformation rule of the Christoffel symbol,
1 λρ
Γλµν = g (∂µ gρν + ∂ν gµρ − ∂ρ gµν ) ,
2
as
∂y α ∂y β ∂xλ γ ∂xλ ∂ 2 y σ
Γλµν = Γ̃ + .
∂xµ ∂xν ∂y γ αβ ∂y σ ∂xµ ∂xν
Combining the above results, we can show the ‘covariance:
∂xµ ( ν ν ρ σ
)
ÿ + Γ̃ ρσ ẏ ẏ = ẍµ + Γµρσ ẋρ ẋσ .
∂y ν
In local frame, when it is in free fall, the object moves in linear motion. So ÿ a = 0. From the
above equation, we have
ẍλ + Γλµν ẋµ ẋν = 0. (2.1)
We call this equation Geodesic equation . We derive the geodesic equation in geometrically. Recall

dτ 2 = −gµν dxµ dxν .

We consider the relativisitic point particle action:

∫ xf ∫ √
S= dτ = dσ −gµν (x) ẋµ ẋν .
xi

22
 2.1. The geodesic equation

Here
d µ
ẋµ = x .
dσ
Then ∫ √
dxµ dxν
S= dσ −gµν .
dσ dσ
′
√ is invariant x (σ) 7→ x (σ ). So it has a local symmetry. Hence we
µ µ
Note that the above action
can choose σ ≡ τ to −gµν ẋ ẋ = 1.
µ ν

Therefore, by the least action principle, we deduce the Euler-Lagrange equation. From this, we
will derive the geodesic equation.
Taking variation, we have
∫
√
0 = δS = dτ δ −gµν ẋµ ẋν
∫ ( )
1 δ (−gµν ẋµ ẋν )
= dτ √
2 −gλρ ẋλ ẋρ
∫
δxλ ∂λ gµν ẋµ ẋν + gµν δ ẋµ ẋν + gµν ẋµ δ ẋν
= − dτ √
2 −gλρ ẋλ ẋρ
∫ [ ]
δxλ (−∂λ gµν ẋµ ẋν ) µ d gµν ẋν
= dτ √ + δx √
2 −gλρ ẋλ ẋρ dτ −gλρ ẋλ ẋρ
∫
1 [ ]
= dτ − δxλ (∂λ gµν ẋµ ẋν ) + δxµ ẋλ ∂λ gµν ẋν + gµν ẍν
2
∫ [ ( ) ]
1
= dτ δxσ gσλ ẍλ + ∂µ gνσ − ∂σ gµν ẋµ ẋν
2
∫
[ ]
= dτ dxρ gρλ ẍλ + Γλµν ẋµ ẋν .

Hence we obtain the geodesic equation

ẍλ + Γλµν ẋµ ẋν = 0,
which we already derived in (2.1).
On the other hand, ∫ √
S= dτ −ηab ẏ a ẏ b .
The equation describes linear motion on a manifold. Since g is metric, it has geometrical
meaning. Hence(Christoffel symbol.
) Here note that the equation has no mass part.
∂ ∂
Write xµ = x0 = ct, xi . Then ∂x 0 = c∂t . For large c and slowing moving paricle, we have

∂x0 ≈ 0
∂
( )
cdt dxi
ẋµ = , ≈ (c, 0) .
dτ dτ
From this assumption, we get
∂µ = (0, ∂i )
and so
Γλµν ẋµ ẋν ≈ Γλ00 c2
1
= c2 g λρ (∂0 gρ0 + ∂0 g0ρ − ∂ρ g00 )
2
1
= − c2 g λi ∂i g00 .
2

23
2. General Relativity

Hence
1 2 λi
ẍλ ≈
c g ∂i g00 .
2
If we have a weak gravity, it goes almostly flat. Hence in the weak gravity limit, we have

gµν ≈ ηµν + hµν .

So
g µν ≈ η µν − η µλ hλρ η ρν
up to second-order approximation.
From this, we see
1 2 ( µi )
ẍλ ≈ c η − η µρ hρσ η σi ∂i (η00 + h00 )
2
1 2 λi
≈ c η ∂i h00 .
2
In spartial part,
1 2
ẍi ≈ c ∂i h00 .
2
Recall the theory of Newton:
mẍi ≡ −∂i Unewton ,
So
2 Unewton
h00 ≈ − .
c2 m
In particular, when we have a spherical symmetry, the potential is given by
mM G
Unewton (r) = − .
r
So we get
2M G
h00 ≈ .
c2 r
Hence
2M G/c2
g00 ≈ −1 + .
r2
Therefore the theory of gravity by newton is a special case of Einstein’s. Actually, the approximation
is equal. This was proved by Schwarzschild.
All physical object must obey the geodesic equation. Light also obey the equation. From this,
Einstein predicts the light should band. Eddington examine the Einstein’s prediction by observing
total solar eclipse.
If x is a solution of geodesic equation, then
d µ ν
[ẋ ẋ gµν (x)]
dτ
= (ẍµ ẋν + ẋµ ẍν ) gµν + ẋµ ẋν ẋλ ∂λ gµν
= 2ẍµ ẋν gµν + ẋµ ẋν ẋλ ∂λ gµν
= −2g µρ (∂κ gρσ + ∂σ gκρ − ∂ρ gκρ ) ẋκ ẋσ ẋν gµν + ẋµ ẋν ẋλ ∂λ gµν
= −ẋκ ẋσ ẋρ (∂κ gρσ + ∂σ gκρ − ∂ρ gκρ − ∂κ gσρ )
= 0.

24
 2.2. Tensor and covariant derivatives

From this, ẋµ ẋν gµν is constant with respect to τ . There are three cases. >, =, <. We say ẋµ
is space-like if the quantity is strictly positive. Similarly, ẋµ is null, time-like if the quantity is = 0,
strictly negative, respectively.
We can normalize so that

ẋµ ẋν gµν = +1 if it is space-like

ẋ ẋ gµν = −1
µ ν
if it is time-like.

In space-like, ds2 = gµν dxµ dxν . In time-like, −c2 dτ 2 = gµν dxµ dxν . In the case of null, we do
not call τ as a proper-time rather than affine parameter.
For ordinary massive particle, it is time-like. For massless particle, or photon, it is null. In the
space-like, there is no such particle since the particle must exceed the speed of light. But we call it
just Tachyon.
Write mẋµ = pµ . In the time-like, we have

pµ pν gµν = −m2 c2 .

From now on, we will derive an equation for gρν , the Einstein field equation.

2.2 Tensor and covariant derivatives

Consider the diffeomorphism

xµ 7→ x′µ (x)
x′µ → xµ (x′ ) .

Then
∂x′µ
dxµ 7→dx′µ = dxν
∂xν
∂xν
∂µ 7→ ∂ν .
∂x′µ

Note that ds2 must be invariant under diffeomorphism. From this assumption, we have

′ ∂x′µ ∂x′ν ρ σ
gµν dxµ dxν = gµν dx dx .
∂xρ ∂xσ
So
′ ∂xρ ∂xρ
gµν 7→ gµν (x′ ) = gσρ (x) .
∂x′µ ∂x′ν
Definition 2.1. We define (p, q)-Tensor(p, q = 0, 1, 2, . . . ) T if it satisfies

µ1 µ2 ···µp ∂x′µ1 ∂x′µp ∂x′σ1 ∂x′σq ρ1 ···ρp

T ν1 ν2 ···νq (x) 7→ · · · ρ
· · · T σ1 ···σq (x)
ρ
∂x 1 ∂x p ∂x 1ν ∂xνq
under diffeomorphism. We call this invariance as the covariant transformation rule.

Remark. All the physics laws must be expressed / expressible by tensors. This ensures that physics
law is invariant under coordinate transformations.

Example 2.2. dxµ , ∂µ , gµν are tensor. But xµ is not a tensor in general.

25
2. General Relativity

Example 2.3. Scalar is (0, 0)-tensor: ϕ (x) 7→ ϕ′ (x′ ) = ϕ (x).

′µ
Vector is (1, 0)-tensor: v µ (x) 7→ v ′µ (x′ ) = ∂x ν
∂xν v (x)
∂xν
Contravariant vector is(0, 1)-tensor: wµ (x) 7→ wµ (x′ ) = ∂x
′
′µ wν (x)

′ ∂xλ ∂xρ
Metric is (0, 2)-tensor: gµν (x) 7→ gµν (x′ ) = ∂x′µ ∂x′ν gλρ (x).
Physics law requires differentiation. Let us observe something.
If ϕ is scalar, i.e., (0, 0)-tensor, then
∂xν
∂µ ϕ (x) 7→ ∂µ′ ϕ′ (x′ ) = ∂ν ϕ (x) .
∂x′µ
So the dervative of scalar is (0, 1)-tensor.
If v is vector, i.e., (1, 0)-tensor, then
[ ′ν ]
∂xλ ∂x ρ
∂µ v ν (x) 7→ ∂µ′ v ′ν (x′ ) = ∂λ v (x)
∂x′µ ∂xρ
∂xλ ∂x′ν ∂xλ ∂ 2 x′ν ρ
= ′µ
∂λ v ρ (x) + v (x) .
∂x ∂x ρ ∂x′µ ∂xλ ∂xρ
Note that
∂xλ ∂x′ν
∂λ v ρ (x)
∂x′µ ∂xρ
is (1, 1)-tensor but
∂xλ ∂ 2 x′ν ρ
v (x)
∂x′µ ∂xλ ∂xρ
is not.
Similarly, if Aν is a contravariant vector, then
[ ρ ]
∂xλ ∂x
∂µ Aν (x) 7→ ∂µ′ A′ν (x′ ) = ∂ λ A ρ (x)
∂x′µ ∂x′ν
∂xλ ∂xλ ∂ 2 xρ
= ′µ ′ν
∂λ Aρ + Aρ (x) .
∂x ∂x ∂x′µ ∂x′ν
So first part is tensor but the second one is not. In all, the usual derivative of (p, q)-tensor is not
tensor except (p, q) = (0, 0).
To eliminate anomalous term, we define ∇σ by
µ1 ···µp µ1 ···µp
∇σ T ν1 ···νq (x) = ∂σ T ν1 ···νq (x)
∑p
µ ···µ ρµ ···µ
∑
q
µ1 ···µp
+ Γµσiρ T 1 i−1 i+1 p − T ρ
ν1 ···νj−1 ρνj+1 ···νq Γσ νj .
i=1 j=1

Note that this derivative ∇σ satisfies the Leibniz rule:

∇σ (V µ Wµ ) = (∇σ V µ ) Wµ + V µ (∇σ Wµ ) .

Also
∇σ δµν = 0.
Indeed,

∇σ V µ = ∂σ V µ + Γσµρ V ρ
∇σ Wµ = ∂σ Wµ − Wρ Γρσ µ

26
 2.2. Tensor and covariant derivatives

implies

(∇σ V µ ) Wµ + V µ (∇σ Wµ )
= ∂σ (V µ Wµ )
= ∇σ (V µ Wµ ) .

Also

∇σ δ µν = ∂σ δ µν + Γµσ ρ δ ρν − δ µρ Γρσ ν
= 0 + Γµσ ν − Γµσ ν = 0.

∇σ should send tensor to tensor under diffeomorphism. To find the condition, we calculate
µ1 ···µp µ1 ···µp
∇σ T ν1 ···νq (x) = ∇′σ (T ′ ) ν1 ···νq
∂x′µ1 ∂x′µp ∂xτ ∂xσ1
= · · ·
∂xλ1 ∂xλρ ∂x′σ ∂x′vq 
λ1 ···λρ
∑ ∑ λ1 ···λp
× ∂τ T κ1 ···κq + Γλτ ρi T ···ρ···κ1 ···κq − T ρ
···ρ··· Γτ κj

i j

∂x ′µ1
∂x ∂x ∂x′µp τ σ1 [ ]
λ ···λ
= λ
··· λ ′σ ∂x′vq
∂τ T 1 ρκ1 ···κq
∂x 1 ∂x
[ ′µ1
ρ ∂x ]
∂ ∂x ∂x′µp ∂xτ ∂xσ1 λ ···λ
+ ′σ · · · T 1 pκ1 ···κq
∂x ∂xλ1 ∂xλρ ∂x′σ ∂x′vq
∑ ···ρ···
∑ µ ···µ
(Γ′ )σiρ (T ′ ) (T ′ ) 1 p ···ρ··· (Γ′ )σ νj .
µ ρ
+ ν1 ···νq −
i j

Write
∂x′µ ∂xρ κ
(Γ′ )λν (x′ ) =
µ
Γ (x) + ∆µλν .
∂xκ ∂x′ν ρσ
Then
∂xα ∂x′µ1 ∂xγ1
0 = ∇′λ T ′µ1 ···ν1 ··· − · · · · · · ∇α T β1 ···γ1 ···
∂x′λ ∂xβ1 ∂x′ν1
∑ [[ ∂ ( ∂x′µi )] ∂xβi ] ∑ [ ∂ 2 xγj ∂x′ρ ]
µi ′···ρ ρ ′µ1 ···
= ′λ λi ′ρ
· · · + ∆ λρ T ν1 + ′λ ∂x′νj ∂xγj
− ∆ λ νj T ···ρ··· ..
i
∂x ∂x ∂x j
∂x

Now

∂ 2 xγ ∂x′ρ
∆ρλν ≡
∂x′λ ∂x′µ ∂xγ
( ′µ )
∂ ∂x ∂xβ
∆µλρ ≡ − ′λ
∂x ∂xβ ∂x′ρ
( ′µ )
∂ ∂x ∂xβ ∂x′µ ∂ 2 xβ
≡− ′ ′ρ
+
∂x β
∂x ∂x ∂xβ ∂x′λ ∂x′ρ
∂ 2 xβ ∂x′µ
=
∂x′λ ∂x′ρ ∂xβ
∂x′µ ∂xβ
( )
since ∂xβ ∂x′ρ
= δρµ and δM −1 = −M −1 δM M −1 (from M M −1 = I, δ M M −1 = 0.)

27
2. General Relativity

Hence we obtain the desired covariant derivative ∇µ = ∂µ + Γµ , where

∂x′µ1 ∂xβ1
T µ1 ···ν1 ··· (x) 7→T ′µ1 ···ν1 ··· (x′ ) = · · · · · · T α1 ···β1 ··· (x)
∂xα1 ∂x′ν1
∂x′µ ∂xβ ∂xγ α ∂ 2 xρ ∂x′µ
Γµλ ν (x) 7→Γ′µ ′
λν (x ) = ′λ ′ν
Γβγ (x) +
α
∂x ∂x ∂x ∂x′λ ∂x′ν ∂xρ
α ′µ1
∂x ∂x ∂xγ1
∇λ T µ1 ···ν1 ··· (x) 7→∇′λ T ′µ1 ···ν1 ··· (x′ ) = · · · ∇α T β1 ···γ1 ··· (x) .
∂x′λ ∂xβ1 ∂x′ν1
So ∑ ∑
∇λ T µ1 ···ν1 := ∂λ T µ1 ···ν1 ··· + Γµλiρ T ···ρ···ν1 ··· − T µ1 ······ρ··· Γρλνj .
i j

One can check the derivative behaves covariantly.

Example 2.4. Recall the Maxwell equation. We write
∂λ F λµ = J µ
∂λ Fµν + ∂µ Fνλ + ∂ν Fλµ = 0. (2.2)
In fact, we should write
∇λ F λµ = J µ (2.3)
∇λ Fµν + ∇µ Fνλ + ∇ν Fλµ = 0.
to be invariant under coordinate transform. Note that (2.2) is a special case of (2.3). In fact, if
Γνλµ ≡ 0(flat spacetime), then (2.3) becomes (2.2).
So we want to find the condition on flatness. In free falling, the metric gµν acting on a particle
satisfies
gµν |origin → ηµν
∂λ gµν |origin = 0.
So by considering the Taylor expansion, we have
1
gµν (x) = ηµν + xλ xρ ∂λ ∂ρ gµν (x) + · · ·
2
If ∂λ gµν |origin = 0 is of physics law, this should be ∇λ gµν (x) = 0. We call this as the metric
is covariantly constant (metric compatiblity condition). So the one of the axiom of the Einstein
theory is
• Metric is covariantly constant, i.e., ∇λ gµν (x) = 0.
From this,
0 = ∇λ gµν = ∂λ gµν − gρν Γρλµ − gµρ Γρλν .
Note that
∂x′µ ∂xβ ∂xγ α ∂ 2 xρ ∂x′µ ∂x′µ ∂xβ ∂xγ α ∂ 2 xρ ∂x′µ
Γµλν − Γµνλ 7→ Γ (x) + − Γ (x) −
∂xα ∂x′λ ∂x′ν βγ
∂x′λ ∂x′ν ∂xρ ∂xα ∂x′ν ∂x′λ βγ
∂x′ν ∂x′λ ∂xρ
′µ β γ ′µ β γ
∂x ∂x ∂x α ∂x ∂x ∂x α
= Γ (x) − Γ (x)
∂xα ∂x′λ ∂x′ν βγ ∂xα ∂x′ν ∂x′λ βγ
( )
∂x′µ ∂xβ ∂xγ ∂xβ ∂xγ
= − Γα
βγ (x)
∂xα ∂x′λ ∂x′ν ∂x′ν ∂x′λ
∂x′µ ∂xβ ∂xγ ( α )
= α ′λ ′ν
Γβγ (x) − Γαγβ (x) .
∂x ∂x ∂x

28
 2.2. Tensor and covariant derivatives

Hence Γµλν − Γµνλ is a tensor. We call this tensor as a Torsion tensor.

In the theory of Einstein’s gravitiy, we assume Γλµν = Γλνµ , i.e., Torsionless connection. From
this assumption,
0 = ∇λ gµν = ∂λ gµν − gρν Γρλµ − gµρ Γρλν
0 = ∇µ gλν = ∂µ gλν − gρν Γρµλ − gλρ Γρµν
0 = ∇ν gλµ = ∂ν gλµ − gρµ Γρνλ − gλρ Γρνµ .
From the above, we obtain
0 = ∂λ gµν + ∂µ gλν − ∂µ gλµ − 2gρν Γρλµ .
Hence
1
gρν Γρλµ = (∂λ gµν + ∂µ gλν − ∂µ gλµ ) .
2
Taking inverse g ρν , we have
1 ρν
Γρλµ = g (∂λ gµν + ∂µ gλν − ∂µ gλµ ) .
2
It is easy to see that Γρλµ = Γρµλ .
In summary, from ∇λ gµν = 0 and Γλµν = Γλνµ , we obtain
{λ } 1 λρ
µν = Γλµν = Γλνµ = g (∂µ gρν + ∂ν gµρ − ∂ρ gµν )
2
.and we call this the Christoffel symbol. Here note that we obtained the above formula from the
geodesic equation (2.1). We call ∇λ gµν = 0 as an equivalence principle, which is a generalization
of free falling or no gravitiy.
Remark. Note that in (x, y, z) coordinate,
ds2 = δij dxi dxj .
In (r, θ, φ) coordinate, we have
ds2 = dr2 + r2 dθ2 + r2 sin2 θdφ2 .
Note that
∇µ ∇µ ϕ = g µν ∇µ ∇ν ϕ
= g µν ∇µ (∂ν ϕ)
( )
= g µν ∂µ ∂ν ϕ − Γλµν ∂λ ϕ
= g µν ∂µ ∂ν ϕ − g µν Γλµν ∂λ ϕ
= ∂ µ ∂µ ϕ − g µν Γλµν ∂λ ϕ..
Note
1
g µν Γλµν = g µν g λρ (∂µ gρν + ∂ν gβµ − ∂ρ gµν )
2
1 λρ µν
= g (g ∂µ gρν + g µν ∂ν gβµ − g µν ∂ρ gµν )
2
1
= g λρ g µν ∂µ gρν − g λρ g µν ∂ρ gµν
2
1( λ )
= −∂µ g − µλ
∂ gµν g µν .
2

29
2. General Relativity

In (r, θ, φ) coordinate, grr = 1, gθθ = r2 , gφφ = r2 sin2 θ. So g rr = 1, g θθ = 1

r2 ,
φφ 1
g = r2 sin 2 θ . So

∂µ g µr = ∂r g rr + ∂θ g θr + ∂φ g φr = 0,
∂µ g µθ = ∂θ g θθ = 0
∂µ g µφ = ∂φ g φφ = 0.

Now

g µν ∂λ gµν = g rr ∂λ grr + g θθ ∂λ gθθ + g φφ ∂λ gφφ

1 ( ) 1 ( )
= 2 ∂λ r2 + 2 2 ∂λ r2 sin2 θ .
r r sin θ
So
4 4
g µν ∂r gµν = ⇒ g µν ∂ r gµν =
r r
2
g µν ∂θ gµν = 2 cot θ ⇒ g µν ∂ θ gµν = cot2 θ.
r2
g µν ∂φ gµν = 0.

Hence
( )
∂2 1 ∂2 1 ∂2 2 ∂ 1 ∂
∇µ ∇µ ϕ = 2
+ 2 2
+ 2 2 2
ϕ+ ϕ + 2 cot θ ϕ
∂r r ∂θ r sin θ ∂φ r ∂r r ∂θ
2
( ) ( )
1 ∂ ∂ϕ 1 ∂ ∂ 1 ∂2ϕ
= 2 2 r2 + 2 sin θ ϕ + 2 2 .
r ∂r ∂r r sin θ ∂θ ∂θ r sin θ ∂φ2
For the case of vector,
( )
∇µ ∇µ V ν ̸= ∂ µ ∂µ − g µρ Γλµρ ∂λ V ν .

In fact,

∇µ ∇µ V ν = g µλ ∇λ ∇µ V ν
( )
= g µν ∂λ ∇µ V ν + Γνλρ ∇µ V ρ − Γρλµ ∇ρ V ν
[ ( ) ( ) ( )]
= g µλ ∂λ ∂µ V ν + Γνµρ V ρ + Γνλρ ∂µ V ρ + Γρµκ V κ − Γρλµ ∂ρ V ν + Γνρκ V κ .

2.3 Existence of Metric compatibility

In this section, we show the existence of metric gµν satisfying

∇λ gµν = 0.

Recall the geodesic equation of motion

ẍλ + Γλµν ẋµ ẋν = 0. (2.4)

If xµ (τ ) is a solution to the geodesic motion, so is xµ (kτ ) for k ∈ R, i.e., if x′µ (τ ) ≡ xµ (kτ ),

then
ẍ′λ + Γλµν (x′ ) ẋ′µ ẋ′ν = 0.

30
 2.3. Existence of Metric compatibility

Indeed,
d ′µ
x (τ ) = k ẋµ (kτ )
dτ
d2 ′µ
x (τ ) = k 2 ẍµ (kτ )
dτ 2
gives [ ]
ẍ′λ + Γλµν (x′ ) ẋ′µ ẋ′ν = k 2 ẍλ + Γλµν ẋµ ẋν = 0.
Now we consider initial-value problem. If x is an analytic solution of (2.4), then
τ2 µ
xµ (τ ) = xµ0 + τ v µ − Γ (x0 ) v µ v ρ + · · · ,
2 νρ
where x0 denotes the initial position and v denotes the initial velocity.
Note that for k ∈ R,
( )
xµ k · 0, v λ = 0,

d µ( )
λ 
x kτ, v  = kv λ .
dτ τ =0

From this observation, we get ( τ )

( )
xµ τ, v λ = xµ τ̂ , v λ ,
τ̂
where τ̂ is an arbitrary fixed time.
If f is an analytic solution of (2.4), then we can choose a coordinate transform xµ 7→ y µ (x)
with xµ = xµ0 and y µ = 0 satisfying
( )
yλ
xµ = f µ τ̂ , .
τ̂
The idea is to choose a coordinate in terms of velocity vector in the unit time. This identity defines
a coordinate transformation from xµ to y µ .
Note that the geodesic equation transforms like a tensor under coordinate transforms. Indeed,
∂x′µ 2 ′µ
ν ρ ∂ x
ẍ′µ = ẍν + ẋ ẋ
∂xν ∂xν ∂xρ
and
∂x′µ κ ρ σ ∂ 2 xκ ∂x′ρ ∂xσ α β
Γ′µ ′ ′ν ′ρ
νρ (x ) ẋ ẋ = Γρσ ẋ ẋ + ẋ ẋ
∂x κ ∂x′ρ ∂x′σ ∂xα ∂xβ
∂x′µ κ ρ σ ∂ 2 x′µ α β
= Γρσ ẋ ẋ − ẋ ẋ .
∂xκ ∂xα ∂xβ
So
ẍ′λ + Γ′λ ′ ′µ ′ν
µν (x ) ẋ ẋ
∂x′λ ( κ )
= κ
ẍ + Γκρσ (x) ẋρ ẋσ .
∂x
This shows xµ is a solution of (2.4). Note that
( ) ( )
xµ τ, v λ = f µ τ, v λ
( )
τ vλ
= f µ τ̂ , .
τ̂

31
2. General Relativity

( )
From this, y µ τ, v λ = τ v µ . This describes a geodesic motion in y-coordinate system, i.e., this
satisfies
ÿ µ + Γµνρ (y) ẏ ν ẏ ρ = 0.
and this implies
Γµνρ (τ v) = v v v ρ = 0.
Especially at τ = 0, for the y-coordinate system

Γλµν (y = 0) v µ v ν = 0


for arbitrary v µ . So Γλµν (y = 0) = 0. Hence ∂y∂λ gµν (y) = 0. Thus, y µ is identified as the
y=0
locally inertial frame. We call this coordinate as a Riemann normal coordinate.
Usually, we write
D 2 xµ
:= ẍµ + Γµνρ ẋν ẋρ
Dτ 2
and we call this as a covariant acceleration

2.4 The Einstein field equation, Tensor density

In this section, we study the Einstein field equation:
1
Rµν − gµν R = 8πGTµν .
2
Here G is a Newton constant, Rµν and R will be defined later. Tµν is an energy-momentum tensor.
On the left hand side, this concerns geometry. On the right hand side, this concerns matter, source
of gravity or curved spacetime.
As an example, if we consider a point particle, one of example of energy-momentum tensor is
dxµ dxν
T µν ∼ m (τ ) (τ ) .
dτ dτ
There is an energy-momentum tensor conservation

∂µ T µν = 0,
∇µ T µν = 0.

There is some ambiguity on the equation since

∇µ (Λg µν ) = 0,

where Λ is a constant. So later Einstein add some terms on Einstein equation

1
Λgµν + Rµν − gµν R = 8πGTµν .
2
Here Λ is a cosmology constant. Einstein thought that universe is static. However, due to Hubble,
it is not. Einstein later mentioned that “adding that term is the one of my big mistakes!”
One might consider
1
Rµν − gµν R = 8πGTµν − Λgµν .
2
Later we will explain the meaning of Λterm.
From now on, we consider Λ = 0.

32
 2.4. The Einstein field equation, Tensor density

Consider a transformation

′ ∂xλ ∂xρ
gµν (x) 7→ gµν (x) = gλρ (x) .
∂x′µ ∂x′ν

   ′ −2
( )  ∂x 2  ∂x 
g = det gµν = ∥g∥µν → g = det ′ ′
gµν =  ′  g = 
   g.
∂x ∂x 

So
 
 ∂x 2

g 7→  ′  g.
∂x

We call this as scalar density with weight ω = 2. Based on this observation, one may define the
tensor density with weight ω

 ′ −ω
µ ···µp ′µ1 ···µp  dx  ∂x′µ1 ∂x′µp ∂xρ1 ∂xρq µ1 ···µp
Tω 1 ν1 ···νq (x) 7→ T ν1 ···νq (x′ ) =  
 · · · · · · Tω ν1 ···νq .
dx λ
∂x 1 λ ′ν
∂x q ∂x 1 ∂x′νq

We require
( µ ···µ
) µ ···µ
Tω 1 ν1p···νq ∇λ Tω 1 ν1p···νq µ ···µ (√ )−ω
∇λ √ ω = √ ω + Tω 1 ν1p···νq ∇λ −g .
( −g) ( −g)

By computation,
( µ ···µ
)
µ ···µ (√ )ω Tω 1 ν1p···νq (√ ) ω (√ )−ω µ1 ···µp
∇λ Tω 1 ν1p···νq = −g ∇λ √ ω − −g ∇λ −g Tω ν1 ···νq
( −g)
µ ···µp
∑ µ ∑ ρ µ ···µp
= ∂λ Tω 1 ν1 ···νq + Γλρi Tω···ρ···ν1 ···νq − Γλνj Tω 1 ···ρ···
i j
(√ )ω [ (√ )−ω (√ )−ω ] µ1 ···µp
+ −g ∂λ −g − ∇λ −g Tω ν1 ···νq .

Also by computation,

(√ )ω
(√ )ω−1 √
∇λ −g
= ω −g ∇λ −g
(√ ) ω ω ω
−1
∇λ −g = (−g) 2 ∇λ (−g) .
2

Later we show
( )
δ ln (det M ) = Tr M −1 δM .

To show this, recall

|M | = det M = εa1 a2 ···an M1a1 · · · Mnan ,

33
2. General Relativity

where εa1 ···an denotes the Levi-Civita symbol. Now taking variation to get

∑
n
δ |M | = εa1 a2 ···an δMiai M1a1 · · · Mi−1ai−1 Mi+1ai+1 · · · Mnan
i=1
∑
n
= εa1 ···an δMik δaki M1a1 · · · Mi−1ai−1 Mi+1ai+1 · · · Mnan
i=1
∑
n
( )kl
= εa1 ···an δMik · · · M −1 Mlai Ma1 · · · Mi−1ai−1 Mi+1ai+1 · · · Mnan
i=1
∑
n
( )kl
= δMik M −1 εa1 ···an M1a1 · · · Mi−1ai−1 Mlai Mi+1ai+1 · · · MN aN
i=1
∑
n
( )kl
= |M | δMik M −1 δli
i=1
( )
= |M | Tr M −1 δM .

So from this we get

∂λ g = g g µν ∂λ gνµ .
Now define

∇λ g : = ∂λ g − (g µν ∂λ gνµ ) g
= ∂λ g − 2Γµλµ g.

Hence
√ √ √
∇λ −g := ∂λ −g − Γµλµ −g ≡ 0.
Similarly, (√ )ω
∇λ −g ≡ 0.
Observe that
(√ ) ∂λ (−g)
∂λ −g = √
2 −g
−2Γµλµ g √
= √ = Γµλµ −g.
2 −g

From this observation, we define

µ ···µp µ ···µp
∑ ∑ µ ···µp µ1 ···µρ
∇λ Tω 1 ν1 ···νq = ∂λ Tω 1 ν1 ···νq + Γµλρi Tω···ρ···ν1 ···νq − Γρλνj Tω 1 ρ
···ρ··· − ωΓλρ Tω ν1 ···νq .
i j

Now if
 ′ −ω
µ ···µp ′µ1 ···µp  dx  ∂x′µ1 ∂x′µp ∂xρ1 ∂xρq µ1 ···µp
Tω 1 ν1 ···νq (x) 7→ T ν1 ···νq (x′ ) =  
 · · · · · · Tω ν1 ···νq ,
dx λ
∂x 1 λ ′ν
∂x q ∂x 1 ∂x′νq

then  ′ −ω
µ ···µ  ∂x  ∂xκ ∂x′µ ∂xσ
∇λ Tω 1 ν1p···νq →   · · · ∇κ Tωρ···σ··· .
∂x  ∂x′λ ∂xρ ∂x′ν
Thus the covariant derivative behaves like tensor under the coordinate transform.

34
 2.4. The Einstein field equation, Tensor density

Example 2.5. The Levi-Civita symbol is Tensor density weight 1 since

 ′ −1 ′µ
 ∂x  ∂x 1 ∂x′µ4
εµ1 ···µ4 =  
 ν
· · · .
∂x ∂x 1 ∂xν4
From this, we have
∑
∇λ εµ1 ···µ4 = ∂λ εµ1 ···µ4 + Γµλρi εµ1 ···ρ···µ4 − Γρλρ εµ1 ···µ4
i
= ∂λ εµ1 ···µ4 .

Example 2.6. Vector density with weight one, ω = 1 J µ

 
 ∂x  ∂x′µ ν
J (x) 7→ J (x ) =  ′ 
µ ′µ ′  J (x) .
∂x ∂xν
Then

∇µ J µ = ∂µ J µ + Γµµν J ν − Γρµρ J µ
= ∂µ J µ .

So the covariant divergence is the classical divergence.

For xµa (τ ), qa a charge,
∑∫
J µ (x) = dτ qa ẋµa (τ ) δ (x − xa (τ ))
a

We call this as a current vector density with weight ω = 1. Then

[ ]
∂ ∑∫
∂µ J µ = dτ qa ẋµ (τ ) δ (x − xa (τ ))
∂xµ a
∑∫ ∂
= dτ qa ẋµ (τ ) µ δ (x − xa (τ ))
a
∂x
∑∫ [ ]
∂
= dτ qa ẋ (τ ) − µ δ (x − xa (τ ))
µ

a
∂xa
∑∫ d
= dτ qa δ (x − xa (τ ))
a
dτ
∑ ∞
= qa δ (x − xa (τ ))|τ =−∞ = 0.
a

Since t (τ ) ∼ τ .
Let f be a scalar function, i.e., f (x) 7→ f (x′ ) = f (x). Then
∫ ∫
d4 xf (x) δ (x − y) dy = f (y) 7→ d4 x′ f ′ (x′ ) δ ′ (x′ − y ′ ) = f (x)
∫  ′

4  dx 

= d x f (x) δ ′ (x′ − y ′ )
dx 
∫
= d4 xf (x) δ (x − y)

= f (y) .

35
2. General Relativity

So J µ is a vector density weight ω = 1. So for the unweighted version, we have

µ
∑∫ δ (x − xa (q))
Jω=0 = dτ qa ẋµa √ .
a
−g

∑∫
Now we define
δ (x − xa (τ ))
T µν
(x) = dτ ma ẋµ ẋν √ .
a −g (x)
Then
∇µ T µν (x) = ∂µ T µν + Γµµρ T ρν + Γνµρ T µρ .
Note that
∑∫ [ ]
∂µ δ (x − xa (τ )) −1
∂µ T µν = dτ ma ẋµa (τ ) ẋνa (τ ) √ + δ (x − xa (τ )) ∂µ (−g) 2 .
a
−g

Observe that
∫ ∫
∂µ δ (x − xa (τ )) d
dτ δ (x − xa (τ ))
dτ ẋµa (τ ) ẋνa (τ ) √ = dτ (−ẋνa (τ )) √
−g (x) −g (x)
∫
δ (x − x a (τ ))
= dτ ẍνa (τ ) √ .
−g (x)
From this observation, we continue the calculation. Then
[ ( ρ )]
∑∫ δ (x − xa (τ )) −Γµρ
µν
∂µ T = ν
dτ ma ẍa (τ ) √ + ẋa (τ ) ẋa (τ ) δ (x − xa (τ )) × √
µ ν
.
a
−g −g

From
− 12 − 12 − 21
0 = ∇µ (−g) = ∂µ (−g) + Γρµρ (−g) ,
we get
∑∫ δ (x − xa (τ )) [ ν ]
∂µ T µν
= dτ ma √ ẍa (τ ) − ẋνa (τ ) ẋµa (τ ) Γρµρ .
a
−g
Hence
∑∫ δ (x − xa (τ )) [ ν ]
∇µ T µν (x) = dτ ma √ ẍa (τ ) − ẋνa (y) ẋµa (τ ) Γρµρ + Γµµρ ẋρa ẋνa + Γνµρ ẋµa ẋνa
a
−g
∑∫ δ (x − xa (τ )) [ ν ]
= dτ ma √ ẍa (τ ) + Γνµρ (xa (τ )) ẋµa ẋνa .
a
−g

In contrast to the current vector case,

∇µ T µν ≡ 0
up to geodesic motion. This shows energy-momentum
( dt ) is conserved if there is no exterior force.
In the case of static case, ẋµ (τ ) = dτ , 0, 0, 0 = (1, 0, 0, 0). So T 00 ̸= 0 otherwise T µν = 0.
In this case,
∑∫ δ (x − xa (τ ))
T 00 = dτ ma √ .
a −g (x)
Note that the source of gravity is mass. From this, the right-hand side of Einstein equation can be
regareded as a generalization of Newton’s mechanics since

∇2 Φ = 4πGρ

36
 2.4. The Einstein field equation, Tensor density

Let us explain the left hand side of the Einstein field equation. If we have a connection, we
consider its curvature. To find a curvature, it suffices to calculate [∇µ , ∇ν ]. For scalar function,

∇µ ∇ν ϕ = ∇µ ∂ν ϕ
= ∂µ ∂ν ϕ − Γρµν ∂ρ ϕ.

So [∇µ , ∇ν ] ϕ = 0.
Let us consider
( )
∇µ ∇ν V λ = ∇µ ∇ν V λ − Γρµν ∇ρ V λ + Γρµν ∇ν V ρ
( ) ( )
= ∂µ ∂ν V λ + Γλνρ V ρ − Γρµν ∂ρ V λ + Γλρσ V σ + Γλµρ (∂ν V ρ + Γρνσ V σ )
[ ]
= ∂µ ∂ν V λ − Γλνρ ∇ρ V λ + Γλν ρ ∂µ V ρ + Γλµ ρ ∂ν V ρ + ∂µ Γλν σ + Γλµ ρ Γρν σ V σ .

Similarly,

∇µ ∇ν Uλ = ∂µ (∇ν Uλ ) − Γρµν ∇ρ Uλ + Γρµν ∇ν Uρ

( )
= ∂µ (∂ν Uλ − Γσνλ Uσ ) − Γρµν ∇ρ Uλ − Γρµλ ∂ν Uρ − Γσνρ Uσ
[ ]
= ∂µ ∂ν Uλ − Γρµν ∇ρ Uλ − Γσµν ∂µ Uσ − Γσµν ∂ν Uσ − Uσ ∂µ Γσµν − Γσν ρ Γρµ λ .

Hence
[ ]
[∇µ , ∇ν ] V λ = ∂µ Γλν σ − ∂ν Γλµ σ + Γλµ ρ Γρν σ − Γλν ρ Γρν σ V σ ,
[ ]
[∇µ , ∇ν ] Uλ = −Uσ ∂µ Γσν λ − ∂ν Γσµ ρ + Γσµ ρ Γρν λ − Γλν ρ Γρµ λ ,

From this we define

Rκλµν = ∂µ Γν − ∂ν Γµ + Γκµ ρ Γρν λ − Γκν ρ Γρµ λ = −Rκλνµ

and we call this as a Riemann curvature.

It is better to memorize
{
R××µν = ∂µ Γν − ∂ν Γµ + [Γµ , Γν ]
[∇µ , ∇ν ] = [∂µ + Γµ , ∂ν + Γν ] = ∂µ Γν − ∂ν Γµ + [Γµ , Γν ] .

From the above calculation law,

κ ·κ
∑ κ ···ρ···κp
∑ κ1 ···κp ρ
[∇µ , ∇ν ] T 1 p λ1 ···λq = Rκρµν
i
T 1 − T λ1 ···ρ···λq R λj µν .
i j

Note

Rκλµν = gκρ Rρλµν

[ ]
= gκρ ∂µ Γρν λ − ∂ν Γρµ λ + Γρµ σ Γσν λ − Γρν σ Γσµ λ
( )
= ∂µ (gκρ Γρνλ ) − ∂ν gκρ Γρµν − ∂µ gκρ Γρνλ + ∂ν gκρ Γρµλ + gκρ Γρµσ Tνλ
σ
− gκρ Γρνσ Γσµλ
( ) ( )
= ∂µ (gκρ Γρνλ ) − ∂ν gκρ Γρµν − gκσ Γσµρ + gρσ Γσµκ Γρνλ
( )
+ gκσ Γσνρ + gρσ Γσνκ Γρµλ + gκρ Γρµσ Γσνλ − gκρ Tνσ ρ
Γσµλ
( ) ( )
= ∂µ (gκρ Γρνλ ) − ∂ν gκρ Γρµν + gµν Γρµλ Γσνκ − Γρνλ Γσµκ .

37
2. General Relativity

Here we used

0 = ∇µ gκρ = ∂µ gκρ − Γσµκ gσρ − Γσµρ gκσ

∂µ gκρ = gρσ Γσµκ + gκσ Γσµρ .

Now note that

1
gκρ Γρνλ = (∂ν gκλ + ∂λ gκν − ∂κ gνλ ) .
2
So
1
Rκλµν = [∂µ ∂ν gκλ + ∂µ ∂λ gκν − ∂µ ∂κ gνλ − ∂ν ∂µ gκλ − ∂ν ∂λ gκµ + ∂ν ∂κ gµλ ]
2 ( )
+gρσ Γρµλ Γσνκ − Γρνλ Γσµκ
1 ( )
= (∂µ ∂λ gκν + ∂ν ∂κ gµλ − ∂µ ∂κ gνλ − ∂ν ∂λ gκµ ) + gρσ Γρµλ Γσνκ − Γρνλ Γσµκ .
2
From the above identities, we can derive some properties of Riemann tensor.
• Rλρµν = −Rλρνµ .
• Rκλµν = −Rκλνµ .
• Rκλµν = −Rλκµν .
• Rκλµν = Rµνκλ .
• Rκλµν + Rκµνλ + Rκνλµ = 0.
Actually, by considering equilibrium principle, we can prove this theorem more easily.
Consider
′ ∂xα ∂xβ ∂xγ ∂xδ
Rκλµν (x) 7→ Rκλµν (x′ ) = Rαβγδ (x) .
∂x′κ ∂x′λ ∂x′µ ∂x′ν
Recall the Jacobi’s identity

[∇λ [∇µ , ∇ν ]] + [∇µ [∇ν , ∇λ ]] + [∇ν [∇λ , ∇µ ]] = 0.

Note that

[∇λ [∇µ , ∇ν ]] V κ = ∇λ [∇µ , ∇ν ] V κ − [∇µ , ∇ν ] ∇λ V κ

( )
= ∇λ Rλσµν V σ − Rκσµν ∇λ V σ + ∇σ V κ Rσλµν
= ∇λ Rκσµν V σ + Rκσµν ∇λ V σ − Rκσµν ∇λ V σ + ∇σ V κ Rσλµν
= ∇λ Rκσµν V σ + ∇σ V κ Rσλµν .

By considering metric, we have

[∇λ [∇µ , ∇ν ]] Vκ = −Vσ ∇λ Rσκµν + ∇σ Vκ Rσλµν .

Hence by considering cyclicity on indices and Jacobi identity, we have

( )
0 = ∇λ Rκσµν + ∇µ Rκσνλ + ∇ν Rκσλµ V σ
( )
+∇σ V κ Rσλµν + Rσµνλ + Rσνλµ .

By the last property of Riemann tensor and arbitrariness of V , we finally get

∇λ Rµνρσ + ∇µ Rνλρσ + ∇ν Rλµρσ = 0.

38
 2.5. Riemann curvature and spacetime

2.5 Riemann curvature and spacetime

Theorem 2.7 (Poincare’s Lemma). Let A be a 1-form on a simply connected domain. Then there
exists a function ϕ such that Aµ = ∂µ ϕ if and only if Fµν = 0, where Fµν = ∂µ Aν − ∂ν Aµ .

Proof. One direction is obvious. Define

xµ (s) = s (xµ − xµ0 ) + xµ0 =: xµs

where 0 ≤ s ≤ 1. with xµ (0) = xµ0 , xµ (1) = xµ .Then

d µ
x (s) = xµ − xµ0 .
ds
Define ∫ 1
dxµ
Φ (x) = dx Aµ (x (s)) .
0 ds
Then
∫ 1
∂ ∂ ( )
µ
Φ = µ
ds xλ − xλ0 Aλ (x (s))
∂x ∂x 0
∫ 1
( ) ∂xν ∂Aλ (xs )
= dsAµ (xs ) + xλ − xλ0 µ
(s)
0 ∂x ∂xνs
∫ 1
( ) ∂Aλ (xs )
= dsAµ (xs ) + xλ − xλ0
0 ∂xµs
∫ 1 [ ]
( λ ) ∂Aµ
= dsAµ (xs ) + s x − x0 Fµν (xs ) +
λ
(xs )
0 ∂xλs
∫ 1
dxλ ∂Aµ
= dsAµ (xs ) + s s (xs )
0 ds ∂xλs
∫ 1
d
= dsAµ (xs ) + s Aµ (xs )
0 ds
∫ 1
d
= ds (sAµ (xs ))
0 ds
= Aµ (x) .

This completes the proof of Poincaré’s lemma.

Theorem 2.8. Rκλµν = 0 if and only if spacetime is flat, i.e, there exists a coordinate system
∂µ gνρ = 0.
Hence the Riemann curvature determines whether the spacetime is flat.

Proof. (⇐): Obvious. ( )

µ
(⇒): Define Γ (s) a 4 × 4 matrix amd Γ (s) ν = − xλ − xλ0 Γµλ ν (xs ), where xµs =
µ µ
s (xµ − x0 ) + x0 . Now define w (s) a matrix
d
w (s) = Γ (s) w (s) , w (s = 0) = I.
ds
Note that such w (s) exists since
∫ s
w (s) = I + ds1 Γ (s1 ) w (s1 )
0

39
2. General Relativity

Taking iteration, then we get

∞ ∫
∑ s ∫ s1 ∫ sn−1
w (s) = I + ds1 ds2 · · · dsn Γ (s1 ) · · · Γ (sn ) .
n=1 0 0 0

[∫ ]
We denote this as P exp dsΓ (s) , where 0 ≤ sn ≤ sn−1 ≤ · · · ≤ s1 ≤ s.
The above w satisfies
w (sf , sm ) w (sm , si ) = w (sf , si )
for 0 ≤ si ≤ sm ≤ sf .
Now
∑ n ∫
∞ ∑ s ∫ s1 ∫ sn−1
∂µ w = ds1 ds2 · · · dsn Γ (s1 ) · · · ∂µ Γ (sj ) · · · Γ (sn ) . (2.5)
n=1 j=1 0 0 0

Note that
λ [ ]
∂µ Γ (s) ν = ∂µ − (xρ − xρ0 ) Γλρ ν (xs ) + xλ0
∂
= −Γλµ ν (xs ) − s (xρ − xρ0 ) µ Γλρν (xs ) .
∂xs
Since Rκλµν = 0, we have
[ ]
∂
∂µ Γ (s) = −Γµ (xs ) − s (x − ρ
xρ0 ) Γµ (xs ) − [Γµ , Γρ ]
∂xρs
d
=− (sΓµ (xs )) − [Γµ (xs ) , sΓ (xs )] .
ds
Plugin this relation to (2.5). Then we have

∑ n ∫
∞ ∑ 1 ∫ s1 ∫ sn−1
∂µ w = ds1 ds2 · · · dsn Γ (s1 ) · · · Γ (sj−1 )
n=1 j=1 0 0 0
[ ]
d ( ( )) ( ) ( ) ( ) ( )
× − Γµ xsj − sΓµ xsj Γ xsj + Γ xsj sΓµ xsj × Γ (xj+1 ) · · · Γ (xsn )
ds
∫ 1
d
=− ds′ ′ [w (s, s′ ) s′ Γµ (x′s ) w (s′ , 0)]
0 ds
= Γµ (x) w.

So

Γλµ ρ = ∂µ wλσ w−1σ

ρ

= −wλσ ∂µ w−1σ
ρ

Γλµρ = Γλρµ

yields ∂µ w−1σ
ρ = ∂ρ w−1σ −1σ
µ . Hence by Poincaré’s lemma, there exists f such that w µ = ∂µ f σ .
So
∂xλ ∂ 2 f σ
Γλµρ = − σ µ ρ .
∂f ∂x ∂x
Finally, consider xµ 7→ f µ (x) coordinate transform. Then we have Γλµν = 0, i.e., ∂λ gµν = 0.

40
 2.5. Riemann curvature and spacetime

Remark. (i) We call w as a Wilson line.

∂
w = Γ (s) w
∂s
∂
w = −wΓ (si ) .
∂si
(ii) For 0 ≤ sn ≤ sn−1 ≤ · · · ≤ s1 ≤ 1, by Fubini’s theorem, we have
∫ 1 ∫ s1 ∫ sn−1
ds1 ds2 · · · dsn F (s1 , . . . , sn )
0 0 0
∫ 1 ∫ 1 ∫ 1 ∫ 1
= ds1 dsn−1 dsn−2 · · · ds1 F (s1 , . . . , sn ) .
0 sn sn−1 s2

Substitute 0 7→ si and 1 7→ s, we can prove identities in (i).

Now we study some contraction on Riemann curvature. By antisymmetricity of Riemann
curvature, we have

g κλ Rκλµν = 0
g µν Rκλµν = 0.

Also
g κµ Rκλµν = −g κµ Rκλνµ = −g κµ Rλκµν = g κµ Rλκνµ .
We define
Rλν := g κµ Rκλµν .
Note that
Rλν = g κµ Rλκνµ = g κµ Rκνµλ = Rνλ .
So the Ricci curvature is symmetric.
By considering some contraction, we have

R = g µν Rµλ = g µν g κλ Rκµλν ,

and we call this as Scalar curvature.

In summary,

Rµν = Rκµκν = Rνµ

R = g µν Rµν = Rµµ = Rκλ
κλ .

From the Bianci’s identity

∇ρ Rκλµν + ∇µ Rκλνρ + ∇ν Rκλρµ = 0,

we have

0 = g µλ (∇ρ Rκλµν + ∇µ Rκλνρ + ∇ν Rκλρµ )

( ) ( )
= −∇ρ g µλ Rκλµν + ∇λ Rκλνρ + ∇ν g µλ Rκλρµ
= −∇ρ Rκν − ∇λ Rλκνρ + ∇ν Rκρ .

Hence we obtain
∇κ Rκλµν = ∇µ Rνλ − ∇ν Rµλ .

41
2. General Relativity

Contract λµ. Then

−∇κ Rκν = ∇λ Rλν − ∇ν R.
Hence
1
∇µ Rµν = ∇ν R
2
So ( )
1
0 = ∇µ Rµν − gµν R .
2
From this, we may define
1
Gµν = Rµν − gµν R
2
and call the Einstein curvature or tensor. Note that Gµν has conservation law ∇µ Gµν = 0.
Now we have to connect Gµν and Tµν . We will show that

Gµν = −8πGN Tµν ,

where GN is the Newton constant. Note that

( )
1 1
Gµµ = R − DR = 1 − D R.
2 2

So ( )
1
1 − D R = −8πGT µµ .
2
Hence
16πGT µµ
R= .
D−2
So
1
Rµν = −8πGTµν + gµν R
[ 2 ]
1
= −8πG Tµν + gµν T λλ .
D−2

Assume gravity is weak and the situation is static(large c limit). Write

gµν = ηµν + hµν .

Its inverse is
g µν ≈ η µν − η µν η νρ hλρ .
So

∂λ gµν = ∂λ hµν
2M G
gtt = g00 ≈ −1 +
r
Then

R00 = Rλ0λ0
= ∂λ Γλ0 0 − ∂0 Γ0λ0 + Γλλρ Γρ00 − Γλ0ρ Γρλ0
= ∂λ Γλ0 0 + Γλλρ Γρ00 − Γλ0ρ Γρλ0

42
 2.6. Some solutions of Einstein’s equation

since ∂0 Γ0λ0 = 0. Note that

1 λρ 1
Γλλν = g ∂ν gλρ ≈ η λρ ∂ν hλρ
2 2
1 λρ
Γλ0ν ≈ g (∂ν h0ρ − ∂ρ h0ν )
2
1 1
Γλ00 ≈ − g λρ ∂ρ h00 ≈ − η λρ ∂ρ h00 .
2 2
From these rule, we get R00 ≈ −∇2 h0 . From g00 ≈ −1 + 2M G
r . we have

R00 ≈ 2∇2 Φnewton = −2 (ρ4πG)

 
ρ 0 0 0
0 0 0 0

where ρ is the mass density. Since T ≈  , we have
0 0 0 0
0 0 0 0

1 D−3
T00 − g00 T λλ = ρ
D−2 D−2

2.6 Some solutions of Einstein’s equation

Recall the Einstein equation
1
Rµν − gµν R = 8πGN Tµν .
2
The left hand side describes the spacetime geometry and the right-hand side describes matters.

2.6.1 Weak curvature limit

If gµν ∼ ηµν + hµν , then g µν ∼ η µν − η µλ η νρ hλρ . From this, Einstein solved the problem on
mercury.

2.6.2 Exact solution: Schwarzschild solution

In spherical symmetric setting with vaccum matter and asymtopotically flat(gµν → ηµν as r → ∞),
one may guess ( )
ds2 = −Adt2 + Bdr2 + C dθ2 + sin2 θdϕ2
since in unit sphere we have a metric

dΩ2 = dθ2 + sin2 θdϕ2 .

Since we assumed spherical symmetry, A, B, C are functions of t and r. By using some

diffeomorphism, we may assume C (r) = r2 .
He solved Rµν − 12 gµν R = 0 if and only if Rµν = 0. Several calculations yield AB = 1 and
d d
dt A = dt B = 0. This result has a physical meaning when we consider collapsing a star which is
in spherical symmetric.
d2
2 (Ar) = 0. So Ar = c1 r + c2 , i.e., A = c1 + r . So if r → ∞, the space is flat. So
c2
Also, dr
c1 = 1. By considering the Newton’s potential, c2 = −2M G.
In all, the Schwarzschild solution is
( )
2M G dr2 ( )
ds2 = − 1 − dt2 + + r2 dθ2 + sin2 θdϕ2 .
r 1− r 2M G

43
2. General Relativity

As an example,
M+ G
= 109 .
R+
So the earth is flat in some sense.
We call rh = 2M G, radius of horizion. If r ≤ 2M G, then we say it is Black hole. From this,
Einstein predicts the existence of Black hole.
In spherical setting, if we flash a photon from r1 to r, then ds2 = 0 since the proper time of
photon is 0. So
∫ r
dr
∆t =
r1 1 −
2M G
r
[ ]
r − 2M G
= (r − r1 ) + 2M G ln .
r1 − 2M G

If r1 = rh , then △t = ∞. So the photon cannot escape the radius of horizon.

2.6.3 Spherical symmetric with Circular motion

π
If r is constant, and θ = 2, ϕ (t), then the orbital velocity is
√
MG
v = rϕ̇ = .
r

2.6.4 Gravitational time dilation

If dr = dθ = dϕ = 0, then
√
2M G
∆τ = 1− ∆t < ∆t.
r
If r1 < r2 , then
v
u ( )
∆t2 u1 − 2M G
1− MG
1 1
=t r1
≈ r1
≈ 1 + MG − < 1.
∆t1 1− 2M G
r2 1− MG
r2
r2 r1

Hence we obtain the gravitational time dilation. The stronger grativty we have, the slower time
flows. This effect is observed in experiment.

2.6.5 Kerr solution

In rotating, Kerr obtained another black-hole solution.

2.6.6 Cosmology
The theory of cosmology is to apply the Einstein equation to universe.

Fact (Cosmological principle). Space is isotropy and homogeneity.

Then the metric of universe is approximated to

( )
ds2 = −dt2 + a2 (t) dx2 + dy 2 + dz 2

44
 2.6. Some solutions of Einstein’s equation

and  
ρ 0 0 0
0 p 0 0

T = .
0 0 p 0
0 0 0 p
Put it into the Einstein’s equation to get some equation. We say a (t) is a scaler factor and ȧa the
Hubble ‘constant’
It was observed ȧ ̸= 0 and ä > 0. ȧ ̸= 0 implies universe expansion. It is conjecture that
ä > 0 in more theoratical way. It seems there is no bouncing force. Some people introduced some
cosmological constant
1
Rµν − ρµν = Tµν + Λgµν .
2
But this model is unsatisfactory since we must assume Λ is very small. This is called the Dark
Energy Problem.
There is another problem: Dark Matter Problem. The result is not satisfactory even in nowadays.

45
Bibliography
[1] L. C. Evans, Partial Differential Equations, AMS
[2] E. M. Stein and R. Sharkarchi, Fourier Analysis: An introduction
[3] K. Kim, Riemannian geometry
[4] B. E. Hall, Lie Groups, Lie Algebras and Representations
[5] http://www.damtp.cam.ac.uk/user/tong/qft/qft.pdf
[6] Goldstein
[7] http://www.maths.ed.ac.uk/~jmf/Teaching/Lectures/CoV.pdf

47