6 Language for curved spacetime: Tensor Analysis

So far we have learnt that the EP is very powerful. It allows you trade uniform gravity with a uniformly
accelerated reference frame. We also learnt that in a gravitational field one can erect an inertial reference
frame where the laws of special theory of relativity holds. This is only possible locally in a sufficiently
small region of space, for example, by letting a reference frame fall freely in the gravitational field. Another
important learning was that the geometry of the spacetime is curved by the presence of gravity, and
the equation of motion of a free particles follows geodesics in the curved spacetime. According to GTR
gravitational e↵ect is just the consequence of motion in the curved spacetime. So in this chapter we will
develop the language of curved spacetime.

6.1 Non-relativistic particle in curved space

Consider a non-relativistic particle of mass in flat Euclidean space. The Lagrangian of the system in
Cartesian coordinate system xi = (x, y, z) is
1
L= m(ẋ2 + ẏ 2 + ż 2 ) . (73)
2
To find the equations of motion of the particle we make use of the Euler-Lagrange equation
✓ ◆
d @L @L
= 0, (74)
dt @ ẋi @xi
which gives
d2 xi
ẍi = = 0 , or ẍ = 0 , ÿ = 0 , z̈ = 0 .
dt2
These are noting but the equations of a free particle in ordinary space.
We want to generalize this to particle moving in a curved space, for example the surface of the sphere.
Any such curved space is defined by a metric, gij such that the infinitesimal distance between two points
xi and xi + dxi is given as
ds2 = gij (x)dxi dxj .
Few things to note: gij = gji is a symmetric 3 ⇥ 3 metric. We have made gij as a function of the coordinate
– it simply means that distance between two points depend on where you are. We will give specific example
of the metric gij . But for the moment let us study the motion in the presence of the metric. We can
generalize the Lagrangian (73) as
m
L = gij ẋi ẋj . (75)
2
It is easy to check that this generalization is true and reproduces Euclidean space Lagrangian (73) if gij = 1
for all i, j. Applying the EL EoM (74) to (75) yields
✓ ◆
@L m @gjk j k @L k d @L @gik
= ẋ ẋ , = mg ik ẋ , = m j ẋj ẋk + mgik ẍk ,
@xi 2 @xi @ ẋi dt @ ẋi @x
✓ ◆
@gik 1 @gjk j k
=) gik ẍk + ẋ ẋ = 0 . (76)
@xj 2 @xi

In the equation (76), the terms in brackets contract with ẋj ẋk . Hence inside the bracket terms must be
symmetric. So we can rewrite the above equation as
✓ ◆
1 @gik @gij @gjk j k
gik ẍk + + ẋ ẋ = 0 (77)
2 @xj @xk @xi

To put this equation in an identifiable form we multiply by inverse of the metric which we denote as g il ,
such that
g il gik = kl

29
Multiplying g il the first term becomes ẍl . After multiplying g il to the second term we make some changes
to the label in this order: first interchange j and k index, then interchange the i and the l index. With
this we get ✓ ◆
k g il @glj @glk @gjk j k
ẍ + + ẋ ẋ = 0 , ẍi + ijk ẋj ẋk . (78)
2 @xk @xj @xl
Now consider the surface of the sphere of radius r = R: an appropriate choice of coordinates is
x = R sin ✓ cos , y = R sin ✓ sin , z = R cos ✓ .
An infinitesimal distance on the surface is ds2 = R2 d✓2 + R2 sin2 ✓d 2 . So the metric gij and its inverse
g ij are ✓ 2 ◆ ✓ 2 ◆
r 0 ij R 0
gij = , g = (79)
0 r2 sin2 ✓ 0 (R2 sin2 ✓) 1
Hence the affine connections are (only the non-vanishing ones)
✓
= sin ✓ cos ✓ , ✓ = ✓ .
Substituting the affine connections in (78) we obtain the following EoMs
✓¨ = sin ✓ cos ✓ ˙ 2 , ¨ sin ✓ + ✓˙ ˙ cos ✓ = 0 . (80)
These are the equation of motions of the particle on the surface of the sphere. This exercise demonstrates
that information of the curvature of a surface is contained in the metric of the surface. And the equation
of motion can be written by calculating the affine connection from the metric components.
Important comment: non-vanishing affine connection does not mean the space is curved. If the motion
was not restricted to the surface of the sphere, i.e., r 6= R, we still would have got non-vanishing affine
connections. Therefore, the definition of curvature need to be properly defined.
The example was that of a non-relativistic particle. We will now make relativistic generalization. For
that we need language of tensors.

6.2 Vectors and Tensors(reloaded)

In curved spacetime, there is no universal coordinate frame. Locally you can have a coordinate system.
But as you move to another position you need to construct a new coordinate system. The reason behind
this drawback is that in a curved spacetime you can not extend a straight line to infinity. Note that this
problem is equivalent to saying that in a gravitational field one can have a locally interial frame. Since we
have to always jump from one coordinate system to the other we have to learn how quantities transform
due to coordinate transformation. Because we want to construct equations that are invariant under general
coordinate transformations. We have to know transformation of three types of quantities – scalars, vectors,
and tensors.
The simplest of the transformation rules is that of scalars. Scalars do not change under a coordinate
transformation. Example includes the proper time d⌧ . As we have seen, the metric gµ⌫ transforms in a
specific way to keep the proper time invariant.
The next simplest transformations is that of a vector. Under a coordinate transformation xµ ! x0µ ,
contravariant and a covariant vector transform as
@x0µ ⌫ @x⌫
V 0µ = V , Vµ0 = V⌫ , (81)
@x⌫ @x0µ
respectively. An example of contravariant vector is the di↵erential dxµ . It transforms as
@x0µ ⌫
dx0µ = dx .
@x⌫
The gradiant of a scalar transforms as
@ @x⌫ @
= . (82)
@x0µ @x0µ @x⌫

30
 Next comes tensors. If a physical quantity requires more than one Lorentz index for its description then
it is a tensor, provided that it transform in a certain way under coordinate transformation. A tensor can
any number of contravariant and covarint indices. The transformation rules for the them are
@x0µ @x⇢ @x0 
T 0µ ⌫ = T ⇢ (83)
@x @x0⌫ @x
In a tensor, if all the indices are up then it is contravariant tensor. If all the indices are down it is called
a covariant tensror, and if indices are both up and down then it is called a mixed tensor. An important
tensor is the metric tensor that transforms under a coordinate transformation xµ ! xµ as
0 @x↵ @x
gµ⌫ = g↵ . (84)
@x0µ @x0⌫
The inverse of gµ⌫ is g µ⌫ ,
µ
g gµ⌫ = ⌫ , (85)
µ
and g transforms as
@x0 @x0µ
g0 µ
= g↵
. (86)
@x↵ @x
Finally, the Kronecker delta symbol is a mixed tensor
0⇢
µ @x @x⌫ @x0⇢ @xµ ⇢
⌫ = = . (87)
@x @x0
µ @xµ @x0

6.2.1 Tensor Algebra

The following tensor algebra properties will be used the subsequent sections
• Linear Combination: Linear combination of two tensors with same indices is also a tensor. If
T⌫µ = aAµ⌫ + bB⌫µ , where a and b are scalars, then the tensor T⌫µ transform as
@x0µ @x @x0µ @x
T⌫0µ = a 0⌫
A⇢ + b ⇢ B⇢ ,
⇢
@x @x @x @x0⌫
@x0µ @x ⇢
= T . (88)
@x⇢ @x0⌫
• Direct Product: The direct product of tensors is a tensors. For instance, if Aµ ⌫ and B ⇢ are tensors
then T µ ⌫ ⇢ = Aµ ⌫ A⇢ is also a tensor, because it transforms as
⇢
T 0µ ⌫ = A0µ
⌫ B
0⇢

0µ
@x @x @x0
= A B ,
@x @x0⌫  @x
@x0µ @x @x0
= T  ,
@x @x0⌫ @x
• Contraction: Setting one upper and one lower index the same and summing over all four values
yields a new tensor with those two indices absent. For instance T µ⇢ = T µ ⌫ ⇢⌫ . It transforms as
@x0µ @x @x0⇢ @x0⌫  ⌘⌧
T 0µ⇢ = T ,
@x @x0⌫ @x⌘ @x⌧
@x0µ @x0⇢  ⌘
= T ,
@x @x⌘
@x0µ @x0⇢ ⌘
= T ,
@x @x⌘
• Raising and Lowering indices: The metric tensor gµ⌫ can be used to raise and lower indices in
a tensor. For instance
gµ⌫ T µ⇢ = S⌫ ⇢ , g µ⌫ S⌫ ⇢ = T µ⇢ ,

31
6.2.2 Transformation of volume element
The determinant of gµ⌫
g⌘ Det(gµ⌫ ) ,
is a non-tensor. To find its transformation, we recall the transformation of the metric tensor
0 @x⇢ @x
gµ⌫ = g⇢ . (89)
@x0µ @x0⌫
Taking the determinant
2 2
@xµ @x0µ
g 0 (x0µ ) = 0µ
g(xµ ) = g(xµ ) . (90)
@x @xµ
Here |@x0 /@x| is the Jacobian of a coordinate transformation x ! x0 . So, g transforms like a scalar, expect
for the extra factor of the Jacobian. The determinant of metric is therefore called the scalar density.
Similarly, a quantity that transforms like a tesnor except for a Jacobian factor is called the tensor density.
The factor of |@x0 /@x| is called the weight, so g is scalar density of weight -2.
The need of tensor density arise in transformation of volume. As is known from the fundamental
theorem of calculus that under a coordinate transformation xµ ! x0µ the volume transforms as
@x0 4
d4 x0 = d x.
@x
p
Hence the quantity that is invariant is not the d4 x but the invariant volume element gd4 x.
Just to give an example, consider the surface of a three-dimensional sphere. A line element on the
surface of the sphere is ds2 = r2 d✓2 + r2 sin ✓2 d 2 . Hence, the metric on the surface is
✓ 2 ◆
r 0
gµ⌫ = .
0 r2 sin2 ✓
p
Hence, g = r4 sin2 ✓, and the volume element is gdrd✓d = r2 sin ✓drd✓d .

6.2.3 Transformation of the Affine Connection

The affine connection
@x @ 2 ⇠ ↵
µ⌫ = ,
@⇠ ↵ @xµ @x⌫
is not a tensor as it transforms like

0 @x0 @x⌧ @x ⇢ @x0 @ 2 x⇢

µ⌫ = ⌧ + . (91)
@x⇢ @x0µ @x0⌫ @x @x0µ @x0⌫
⇢

If only the first term was there, then it would have been a tensor. But it is the second term that spoils the
tensorial property. We will write the second term in a slightly di↵erent form. We di↵erentiate the identity
@x0 @x⇢
= ⌫ , (92)
@x⇢ @x0⌫
with respect to x0µ
@x0 @ 2 x⇢ @ 2 x0 @x @x⇢
+ = 0, (93)
@x⇢ @x0µ @x0⌫ @x⇢ @x @x0µ @x0⌫
where in the last term we have used transformation of derivative eq. (82). Therefore, we can write the
transformation of the affine connection as
0 @x0 @x⌧ @x ⇢ @ 2 x0 @x @x⇢
µ⌫ = ⌧ (94)
@x⇢ @x0µ @x0⌫ @x⇢ @x @x0µ @x0⌫
As we discuss in the next section, the fact that affine connection does not transform like a tensor makes
life easier.
Exercise: Show that the geodesic equation is covariant.

32