Professional Documents
Culture Documents
RGpart 1
RGpart 1
1.1 Manifolds
but for our purposes we can think of it as a blank canvas. We will add
1.2 Coordinates
1
1.3 Curves and surfaces
by
xµ = xµ (λ)
xµ = xµ (y 1 , y 2 , y 3 )
S(x0 , x1 , x2 , x3 ) = 0
This one equation is sufficient to specify the position of the space, i.e.
given any values for 3 of the xµ , we can solve it to find the fourth.
2
1.4 Tangent vectors
where δ~s is the infinitesimal separation vector between point P and some
3
Example: in 2D Euclidean space consider the curve
x = λ, y = λ2
At λ = 1
4
1.5 Tangent spaces
such that it tangential to the curved space, e.g. imagine a flat solid sheet
balanced on the surface of a ball. Any vector that lies in this plane is a
The tangent vector to any curve that passes through P can be drawn as
an arrow in TP .
where the geometry of space is close to the geometry of the tangent space
5
1.6 Vectors and vector fields
Now that we have tangent spaces, we can define vectors and vector fields
that exists in the tangent space TP and that obeys the rules of vector
addition and multiplication with any other vector in that same space.
As well as vector let’s give the definition of a vector field. A vector field
~v (xµ ) is the assignment of a vector to each point in the manifold such that
~v (xµ ) evaluated at any point P is a vector that exists in the tangent space
TP .
Note that vectors at two different points in a curved space exist in two
different tangent spaces, this means that they cannot be directly compared
6
1.7 Basis vectors
vectors” ~ea at each point P , if these basis vectors are linearly independent,
then we can use them to write any vector ~v in the following way
~v = v a~ea
basis vectors, whereas in the same space you can assign the same basis
fined by
δ~s ∂~s
~eµ ≡ lim µ
= µ
µ
δx →0 δx ∂x
where ∂~s is the vector displacement between p and a nearby point q, whose
connects p and q to vary in only one coordinate then we see that ~eµ is the
tangent vector to that curve. This isn’t the only choice of basis, however
7
it does make certain calculations easier, but is not necessarily the easiest
∂~s µ
d~s = µ
dx = ~eµ dxµ
∂x
vector between points p and q, we take the inner product of d~s with itself,
which gives
ds2 = d~s · d~s = (~eµ dxµ ) · (~eν dxν ) = (~eµ · ~eν )dxµ dxν
If we now promote these basis vectors to a set of basis vector fields then
Because the inner product is symmetric this means that gµν = gνµ . This
8
the infinitesimal separation between points p and q. Having said that
for any choice of coordinates and for any curve C. Note that any arbitrary
following way
~ea = ea µ~eµ
~eµ = eµ a~ea
Note that substituting one of these expression into the other gives
ea µ eµ b = δ a b
and
eµ a ea ν = δ µ ν .
9
1.8 The Metric
The metric gµν can be used to calculate the lengths of curves, as well as
the angles at which they meet. The first of these properties follows from
Z Z q
S= |d~s| = |gµν dxµ dxν |
C C
µ
~t ≡ ds = dx ~eµ = tµ~eµ
dλ dλ
Z q
S= |gµν tµ tν |dλ
C
Note that it is possible to choose this parameter λ such that gµν tµ tν = ±1.
If I’ve chosen λ such that this is true then λ measures distance along the
R
curve directly as S = C dλ. If λ satisfies this requirement, then it is
10
Example: what is the length of the curve with the following tangent
You may assume that the curve exists in flat Euclidean space.
Z
p
S= gµν tµ tν dλ
C
R1
so S = 0 dλ = 1, in this case.
tance”. These are the measures of time and distance that clocks and
To use the metric to infer angles between vectors we note that it acts
11
as the inner product between two vectors in the same tangent space
As the tangent space has a flat geometry, the angle between the vector ~v
~v · w
~ gµν v µ wν
~v · w
~ = |~v ||w|
~ cos θ ⇒ cos θ = =√ √
|~v ||w|
~ gµν v µ v ν gµν wµ wν
~ are both unit vectors that satisfy such that |~v | = |w|
If ~v and w ~ = 1 then
cos θ = gµν v ν wµ
For any set of vectors ~ea , we can define a second set by the equation
~ea · ~eb = δ a b
As before, we can expand any arbitrary vector in terms of this new dual
basis
~v = va~ea
12
and if these new vectors are dual to a coordinate basis, ~eµ , then we can
g µν = ~eµ · ~eν
Strictly speaking, the dual basis vectors live in what’s called a dual tan-
gent space, but this distinction is not important here. Let’s consider the
different ways we can express the inner product, in terms of ~eµ and ~eµ
gµν v µ wν = vν wν = v ν wν = g µν vµ wν
wν = gµν wν , wµ = g µν wν
13
as well as gµν g νρ = δµ ρ . The g µν and gµν are therefore inverses of each
other, and raise and lower coordinate indices. Finally, we can note that
that the gµν and g µν also relate ~eµ and ~eµ , as follows:
and
and
~eν = g µν ~eµ
The g µν and gµν and therefore also raise and lower indices on the basis
vectors themselves.
gµν g µσ = δν σ
14
Solution: to show this, let’s start with
and
⇒ g µν vµ = δµ ν v µ = v ν
also
and
⇒ gµν v ν = δ µ ν vµ = vν
v ν = g µν vµ = g µν gµσ v σ = δ ν σ v σ
⇒ δ ν σ = g µν gµσ
15
1.10 Coordinate transformations
require
µ ∂xµ 0ν
~e = 0ν ~e
∂x
∂x0σ ∂xµ ∂x0σ
∂xµ
0ρ
µ
⇒ ~e · ~eν = 0ρ
~e · ν
~e0σ = 0ρ ν δ ρ σ
∂x ∂x ∂x ∂x
∂xµ ∂x0ρ ∂xµ
= 0ρ ν = ν
= δµν
∂x ∂x ∂x
∂xµ
∂xν ∂xµ ∂xν
0ρ
g µν µ
≡ ~e · ~e =ν
0ρ
~e · 0σ
~e0σ = 0ρ 0σ g 0ρσ
∂x ∂x ∂x ∂x
Finally, let’s derive the transformation laws for the coordinate components
of ~v :
16
∂xµ 0ρ ∂xµ 0ρ 0σ 0 ∂xµ 0σ ρ ∂xµ 0ρ
= ~
e · ~
v = ~
e · (v ~
e σ ) = v δ σ = v
∂x0ρ ∂x0ρ ∂x0ρ ∂x0ρ
∂x0ν 0
Similarly, ~eν · ~v ⇒ vν = ∂xµ vν
∂xµ 0ν
µ
X = 0ν X
∂x
∂x0µ 0
Xν = X
∂xµ ν
Example: explain how considering the inner product ~eµ · ~v leads to the
transformation law
∂x0ν 0
vµ = v
∂xµ ν
and
17
now
∂~s ∂~s ∂x0ν ∂~s ∂x0ν 0
~e0µ ≡ 0µ , ~eµ ≡ µ = = ~e
∂x ∂x ∂xµ ∂x0ν ∂xµ ν
so
∂x0ν ∂x0ν 0 ∂x0ν 0
~eµ · ~v = ~e0
µ ν
· ~v = (~e · ~v ) = v
∂x ∂xµ ν ∂xµ µ
which gives
∂x0ν 0
⇒ ~eµ · ~v = vµ = v
∂xµ ν
We can choose to think of the inner product between two vectors as the
~u(~v ) ≡ ~u · ~v
of not one but two vectors? The answer is yes, and that we can con-
struct these new objects using a new operation called the outer product
18
Note that in general ~u ⊗ ~v 6= ~v ⊗ ~u, and that ⊗ should not be confused
where
so
as required; the proof that ~h(~p, α~q) = α~h0 (~p, ~q) proceeds in the same way.
19
1.12 Tensors
The object ~h = ~u ⊗ ~v , that takes two vectors in its two arguments and
an objects that is linear in each of its arguments, and that takes n vectors
~t(~u, ~v , w)
~ is a real number and
~t(~u + ~v , ~v , w)
~ = t(~u, ~v , w)
~ + t(~v , ~v , w)
~
~t(α~u, ~v , w)
~ = α~t(~u, ~v , w)
~
tions for physical laws we know we will end up with a set of equations that
are valid in all coordinate systems. This is exactly what is required from
our tensors. For a rank n tensor the appropriate basis is made from the
20
outer product of n basis vectors:
We could equally well have constructed our tensor basis from dual basis
vectors
= gµν uµ v ν = ~u · ~v
21
1.13 Tensor operations
From the definitions in the previous sections we can show that a rank-2
• gµν and g µν lower and raise the indices of tensors, just as with vectors.
• rank-n tensors with two indices contracted are the components of rank-
(n − 2) tensors.
Example: prove that the metric gµν and g µν lower and raise the indices
22
solution: consider a tensor ~t. If this is of rank 2 then
tab v b = ta b vb = ta b gbc v c
⇒ tab = ta c gbc
tab vb = ta b v a = ta b g bc vc
⇒ tab = ta c g bc
ta b ua = tab ua = tab g ac uc
⇒ ta b = tcb g ac
23
second and third
⇒ ta b = tcb gac
1.14 Connection
we will also be interested in how they change value from point to point.
For scalars this is not a problem, as the rate of change can simply be
∂φ
≡ ∂µ φ
∂xµ
For vectors it is more difficult. Two vectors, at two different points, lie in
how basis vectors in each space are related to each other. This is done by
24
supposing they change as:
∇~ea~eb = Γc ba~ec
of its argument ~eb in the direction indicated by the subscript (~ea ). The
Γc ba are known as the connection coefficients, and encode how the basis
interested in changes of coordinate basis, and for this we use the shorthand
∇µ ≡ ∇~eµ .
∂φ
∇µ φ =
∂xµ
which tells us how to connect dual basis vectors in different tangent spaces.
25
Exercise: Prove the above relation by assuming the connection obeys
With a given connection, we can work out how to calculate the change
relationships:
neighbouring tangent spaces, and does not apply solely to the basis vectors.
26
1.15 The Levi-Civita Connection
Γµ νρ = Γµ ρν
and
∇ν g = 0
The first of these says that the connection is “torsionless”, the second says
the Levi-Civita connection and its connection components are called the
Christoffel symbols. The two equations above, the Leibnitz rule, and the
1
Γµ νρ = g µσ (∂ν gσρ + ∂ρ gσν − ∂σ gνρ )
2
27
1.16 Covariant derivative
operator:
∇ν ~v ≡ (Dµ v ν )~eν
∇µ~v = ∇µ (v ν ~eν ) = (∇µ v ν )~eν +v ν (∇µ~eν ) = (∂µ v ν )~eν +v ν Γσ νµ~eσ = (∂µ v ν +Γν ρµ v ρ )~eν
⇒ Dµ v ν = ∂µ v ν + Γν ρµ v ρ
Dµ vν = ∂µ vν − Γρ µν vρ
Dµ tν1 ...νp ρ1 ...ρr = ∂µ tν1 ...νp ρ1 ...ρr + tσ...νp ρ1 ...ρr Γν1 µσ + . . . − tν1 ...νp σ...ρr Γσ µρ1 − . . .
28
where . . . denotes extra similar terms corresponding to each raised/lowered
In fact, using ∇~v~t = v µ ∇µ~t it can be seen immediately that Dµ tν1 ...νp ρ1 ...ρr
known way, then each index of Dµ tν1 ...νp ρ1 ...ρr must transform like a tensor,
if they didn’t then ∇~ν~t would change under some change of coordinates,
fields under consideration are fields over the whole manifold. However,
sometimes they are only defined along a single curve. In this case we may
For a vector:
d~v Dv µ
≡ ~eµ
dλ Dλ
where λ is the parameter along a curve xµ (λ). If the tangent vector to the
29
curve is ~u, then
d~v µ dxµ ν ν ρ
dv µ
µ dx
µ
= ∇~u~v = u ∇µ~v = (∂µ v + Γ µρ v )~eν = + Γ νρ v ρ ~eµ
dλ dλ dλ dλ
so
Dv µ dv µ µ dx ρ
ν
= + Γ νρ v
Dλ dλ dλ
ν
Dvµ dvµ ρ dx
= − Γ µν vρ
Dλ dλ dλ
30
There are many ways that we could do this, but one particularly use-
ful one is known as parallel transport. We say that ~v has been parallel
transported along C if
d~v
=0
dλ
dv µ µ ν dx
ρ
= −Γ νρ v
dλ dλ
If we specify v µ at any point on C, we can use this equation to work out its
value at any other. The result of this is a set of vectors that are “parallel”
tangent spaces.
A case where the interpretation is clear is in flat space. In this case the
31
the curve so that it points in the same direction in the higher dimensional
space, and then taking the components that lie in the tangent space to the
curved space at the new point. This is the closest thing to parallel that
If the tangent vector to a curve, ~u, obeys the parallel transport condition
d~u
≡ ∇~u~u = 0
dλ
true if
32
Example: Parallel transport in vector ~v along a curve with constant
dxν
+ vθ Γ + vφΓ = 0
dφ
cos θ
Γθ φφ = − sin θ cos θ and Γφ θφ = Γφ φθ =
sin θ
dv θ
⇒ = sin θ0 cos θ0 v φ
dφ
dv φ cos θ0 θ
=− v
dφ sin θ0
Differentiate the first of the above equations and sub into the second
d2 v θ dv φ cos θ
0
⇒ 2
= sin θ0 cos θ0 = sin θ0 cos θ0 − v θ = − cos2 θ0 v θ
dφ dφ sin θ0
⇒ v θ = A cos(αφ) + B sin(αφ)
similarly
v φ = C cos(αφ) + D sin(αθ)
33
1.19 Euler-Lagrange formulation of the geodesic equation
µ ν dxµ dxν ds 2
L ≡ gµν ẋ ẋ = gµν =
dλ dλ dλ
∂L d ∂L
− =0
∂xµ dλ ∂ ẋµ
∂L ∂gµν µ ν ∂L
= ẋ ẋ and = 2gσµ ẋµ
∂xσ ∂xσ ∂ ẋσ
or
34
Example: Find the geodesic equations for Schwarzschild geometry:
2
2Gm 2 2Gm −1 2
ds = − 1 − dt + 1 − dr + r2 (dθ2 + sin2 θdφ2 )
r r
π
Solution: choose a particle that lies in the plane θ = 2
2Gm 2 ṙ2
⇒L=− 1− ṫ + 2Gm
+ r2 φ̇2
r (1 − r )
t-equation:
∂L ∂L 2Gm
=0 and = −2 1 − ṫ
∂t ∂ ṫ r
so
" #
∂L d ∂L d 2Gm A
− =0 ⇒ 1− t =0 ⇒ t=
∂t dλ ∂ ṫ dλ r 1− 2Gm
r
φ-equation:
∂L ∂L
=0 and = 2r2 φ̇
∂φ ∂ φ̇
so
∂L d ∂ d 2 B
− =0 ⇒ [r φ̇] = 0 ⇒ φ̇ =
∂φ dλ ∂ φ̇ dλ r2
35