Christoffel Symbols

PHYS 471: Introduction to Relativity and Cosmology

1 The Metric and Coordinate Basis

Let’s recap some properties of the metric, before deriving the Christoffel symbols.
First, we know the metric is a tensor gµν , which can be written as a 4 × 4 diago-
nal matrix. The line element ds2 is the length-squared of an infinitesimal vector in
spacetime:

ds2 = gµν dxµ dxν (1)

The components of the metric will depend on the coordinate system, however. For
flat spacetime in Cartesian coordinates, this is the familiar Minkowski metric

ds2 = dt2 − dx2 − dy 2 − dz 2 , gµν = ηµν = diag(+1, −1, −1, −1)

But if we were to change coordinates to, say, spherical ones, the line element would
look different:
ds2 = dt2 − dr2 − r2 dθ2 − r2 sin2 θdφ2
These metric coefficients can be “encoded” in what’s called a “basis vector” êµ (more
technically a vierbein!). Formally, we define the metric as the product

gµν = êµ · êν (2)

Great, but what is this basis vector thing?? You’ve actually seen them before. When
you first start talking about vectors, we define the unit vectors x̂, ŷ, ẑ that point in
those orthogonal directions. A general vector is then written

~s = Ax x̂ + Ay ŷ + Az ẑ

When we change coordinates to spherical ones, we want to re-express these coefficients

as some other set
~s = Ar r̂ + Aθ θ̂ + Aφ φ̂
So, consider a vector that has infinitesimal length for all its components: Ax =
dx, Ay = dy, Az = dz. Then,
~ = dx x̂ + dy ŷ + dz ẑ
ds

If we switch to spherical coordinates, the appropriate transformation (using the Ja-

cobian) gives
~ = dr r̂ + rdθ θ̂ + r sin θˆdφφ
ds
But what we really want to do is express any such vector in a way that “looks like”
the Cartesian version:
~ = dx̂1 + dx̂2 + dx̂3
ds
In order to do so, we define the coordinate basis vector component êi to be the
thing that allows us to write
~ = dx̂1 + dx̂2 + dx̂3 = dx1 ê1 + dx2 ê2 + dx3 ê3
ds

Note that the “basis” quality has been taken away from dxi , which is now only a
number. So, for the Cartesian system, dx̂1 = dx êx , dx̂2 = dy êy , dx̂3 = dz êz , which
tells us that the components of the coordinate basis vector are êx = êy = êz = 1.

Using this formalism, the line element is

~ · ds
ds2 = ds ~ = (ê1 )2 dx2 + (ê2 )2 dx2 + (ê3 )2 dx2
1 2 3

Comparing the above to Equations ?? and ??, we see that g11 = ê1 · ê1 = (ê1 )2 , and
so forth. If we did the same thing in spherical coordinates, we’d find
~ = dr̂ + dθ̂ + dφ̂ = dr êr + dθ êθ + dφ êφ
ds

=⇒ êr = 1 , êθ = r , êφ = r sin θ

=⇒ grr = (êr )2 = 1 , gθθ = (êθ )2 = r2 , gφφ = (êφ )2 = r2 sin2 θ
Note that we’re used to seeing these as negative, which they still are in spacetime!
The above discussion was only for spatial components.

2 Properties of the Christoffel Symbols

The coordinate basis vectors have an obvious coordinate dependence in some systems,
so we must account for this when we compute things like derivatives. The Christoffel
symbols encapsulate this dependence when you take the derivative of a coordinate
basis vector. We define this as

∂µ êν = Γλµν êλ = Γ0µν ê0 + Γ1µν ê1 + Γ2µν ê2 + Γ3µν ê3

That is, the derivative of the basis vector is a linear combination of the four com-
ponents of the basis vector! This accounts for any motion the basis vector might
undergo as it moves around a non-trivial geometry.

Since we know gµν = êµ êν , it’s straightforward to show that the derivative of the
metric is
∂α gµν = (∂α êµ )êν + êµ (∂α êν ) = Γλαµ êλ êν + Γλαν êλ êµ
 =⇒ ∂α gµν = Γλαµ gλν + Γλαν gλµ (3)
From this, we can derive an expression for the Chirstoffel symbols in terms of first
derivatives of the metric. When we cycle the order of the indices α, µ, ν, we get
two more equations:
=⇒ ∂µ gνα = Γλµν gλα + Γλµα gλν (4)
=⇒ ∂ν gαµ = Γλνα gλµ + Γλνµ gλα (5)
As luck would have it (or for geometric reasons, if you prefer), the Christoffel symbols
are symmetric in their lower indices, which reveals that many of these terms are
the same. For example, Γλµν gλα in Equation ?? is the same as Γλνµ gλα in Equation ??,
because we can write Γλµν = Γλνµ .

Re-writing the above equations to make this obvious, we get

(3) becomes −→ ∂ν gαµ = Γλµν gλα + Γλαν gλµ (6)
(1) becomes −→ ∂α gµν = Γλαµ gλν + Γλαν gλµ (7)
(2) becomes −→ ∂µ gνα = Γλαµ gλν + Γλµν gλα (8)
Note each term in Equation ?? (formerly Eq 1) appears in Equation ?? and ??, re-
spectively. So, we combine these three equations and solve for the Christoffel symbols:

Equation 4 + Equation 6 − Equation 4

= ∂ν gαµ + ∂µ gνα − ∂α gµν
= Γλµν gλα + Γλµν gλα
= 2Γλµν gλα (9)
Success! All we need to do is isolate the Γ, and we do so by multiplying both sides
by the inverse metric g λα :
2Γλµν gλα = ∂ν gαµ + ∂µ gνα − ∂α gµν

1
Γαµν = g αλ (∂ν gµλ + ∂µ gλν − ∂λ gµν ) (10)
2

OK, so now that we have a way to calculate these coefficients, what good are they?
As mentioned before, these tell us something about how basis vectors change their
direction as a vector moves around some space(time) with curvature. They also tell
us something about the nature of lines in the curved spacetime, specifically straight
ones (which are kinda sorta important to General Relativity).
3 Uses of the Christoffel Symbols
There are varied applications of the Christoffel symbols in General Relativity, and
more broadly in differential geometry. We’ll encounter a few key ones in our discus-
sions, however, so let’s review three fundamental ones right now.

3.1 The Covariant Derivative

Christoffel symbols are defined in terms of the derivative of the coordinate basis
vector components. Their first and most important application is in the computation
of derivatives in non-flat spacetimes. If we take the derivative of a vector A = Aµ êµ ,
we get
∂µ A = ∂µ (Aν êν ) = (∂µ Aν )êν + Aν (∂µ êν )
The second term in the derivative is the Christoffel symbol, so

∂µ (Aν êν ) = ∂µ Aν êν + Aν Γαµν êα

Since the ν index in the first term is paired, we can arbitrarily replace it with α to
find
∂µ Aα êα + Aν Γαµν êα = ∂µ Aα + Aν Γαµν êα

The term in parenthesis shows how the derivative of the vector A also depends on
the coordinate basis terms. We call this the covariant derivative of Aν :

=⇒ Dµ Aν = ∂µ Aν + Γνµα Aα

Note the lower in the Christoffel symbol combine the derivative index with the vector
index (hence, “mixing” the coordinates due to derivatives).
In General Relativity, we also deal with objects with lower indices, though, so some
extra care must be taken. It turns out that the derivative of a covariant vector
Aµ looks almost the same, except the sign in front of the Christoffel symbol is negative:

=⇒ Dµ Aν = ∂µ Aν − Γαµν Aα

So the rule is: upstairs indices get a positive Christoffel symbol, downstairs
indices get a negative one. You can remember this easily, since up is positive,
down is negative.

Based on these rules, we can take the covariant derivative of a mixed tensor of any
type, as long as we have a Christoffel symbol for each index. For example, the
covariant derivative of Rµν , Rµν , and Rµν are

Positive for upper indices =⇒ Dα Rµν = ∂α Rµν + Γαµβ Rβν + Γαβν Rµβ
 Negative for lower indices =⇒ Dα Rµν = ∂α Rµν − Γβαµ Rβν − Γβαν Rµβ
+ and − for mixed indices =⇒ Dα Rµν = ∂α Rµν + Γµαβ Rβν − Γβαν Rµβ

3.2 The Christoffel Symbols and the Metric

Taking a derivative is “moving” something in a space, or spacetime. If I move in the
∂
x̂ direction, I think “ ∂x ”! This is great for flat spaces, but as the above discussion
suggests, moving in curved spaces (or spacetimes) means we have to think covariant
derivative. So – what happens if we take the covariant derivative of the metric
itself? (Whoa!)

The answer is: NOTHING! Since the metric defines the intrinsic geometry, it
shouldn’t change as we move from place to place. A curved spacetime has the same
curvature properties everywhere. So, that means
Dα gµν = 0 =⇒ Metric is the same everywhere!
Writing this out in terms of Christoffel symbols, we get
∂α gµν − Γβ αµ gβν − Γβ αν gµβ = 0
because we need a negative sign when we take the covariant derivative of things with
covariant indices. We can rewrite this as
∂α gµν = Γβ αµ gβν + Γβ αν gµβ
A

If we cycle through the indices as we did before (i.e. cyclically swap α, µ, and ν) we
get
∂ν gαµ = Γβ να gβµ + Γβ νµ gαβ
B
∂µ gνα = Γβ µν gβα + Γβ µα gνβ
C
I’ve called these
A , ,
B C . As before, we note that because of the symmetry of the
metric under exchanges of indices, terms with gαµ are the same as those with gµα , and
so forth. We can therefore whittle this expression down to solve for the Christoffel
symbol by writing:

C + B − A
=⇒ ∂µ gνα + ∂ν gαµ − ∂α gµν = 2Γβ µν gαβ
So, multiplying by the inverse of the metric on the RHS g αβ , we obtain the same
expression as before:
1
Γβ µν = g αβ (∂µ gνα + ∂ν gαµ − ∂α gµν )
2
But this time we derived it by understanding that the metric and geometry
doesn’t change when we move around through spacetime.
3.3 The Geodesic Equation: Straight Lines in Curved Space-
time
First, what defines a straight line? From a Newtonian perspective, it’s the path of
a free particle. How do we describe a free particle in relativity? Simple: the same
we do classically. A free particle is one whose acceleration is zero:
du
=0 , u = uµ êµ
dτ
where the explicit dependence on the basis vector is included. Expanding this out,
we get
du duµ dêµ
= êµ + uµ =0
dτ dτ dτ
µ
Furthermore, since uµ = dx
dτ
, this becomes

d 2 xµ dxµ dêµ
ê µ + =0
dτ 2 dτ dτ
What to do with the derivative of êµ ? Why, did you say “Christoffel symbol”? Great
idea! Except, the Christoffel symbol is defined in terms of the coordinate derivative
(∂µ ), and not dτd . But never fear – as calculus has taught us, we know that we can
address this implicit dependence as follows:
dêµ ∂êµ dxν
=
dτ ∂xν dτ
and since
∂êµ
= Γαµν êα (Note : I flipped µ and ν by symmetry)
∂xν
we get
∂êµ dxν α dxν
= Γ ê
µν α
∂xν dτ dτ
The entire expression is then

d 2 xµ dxµ dxν α
ê µ + Γ êα
dτ 2 dτ dτ µν
Again, µ and α are paired indices here, so we can substitute one for the other. We’ll
replace the first one in the double derivative (µ → α), since we don’t want too many
other µs in the second term:

d 2 xα dxµ dxν α
0= êα + Γ êα
dτ 2 dτ dτ µν
and factoring out the êα term gives us
 d2 x α µ
α dx dx
ν
+ Γ µν =0 (11)
dτ 2 dτ dτ

This is called the geodesic equation, and is the equation that describes a straight
path (or a free particle) in curved spacetime. Note that in flat spacetime, we
2 µ
would just have the equation ddτx2 = 0, which tells us that the Christoffel Symbols
are zero in flat spacetime.

3.4 The Riemann Curvature Tensor

The next application of the Christoffel symbols is the important one. Like, the im-
portant one! As I showed you earlier in the class, moving a vector around a closed
path on a spherical surface rotates the direction it points. Furthermore, rotating
it half-way around the closed path along two different routes rotates it by different
amounts. How this vector changes are we move it along a particular direction can be
mathematically quantified by considering derivatives.

Specifically, since we’re asking the question “How does moving it along θ̂ first and φ̂
second compare to moving it along φ̂ first and θ̂ second?” In flat space, these two
conditions would translate to:

~ along θ̂ first, φ̂ second ∂ ∂ ∂2

A =⇒ Ã = Ã
∂φ ∂θ ∂φ ∂θ

~ along φ̂ first, θ̂ second ∂ ∂ ∂2

A =⇒ Ã = Ã
∂θ ∂φ ∂θ ∂φ
As calculus has taught us, though, these two things are equal, because of the conve-
nient fact of equality of mixed partials. So, each would give the same answer –
both paths are the same... but only in flat space!

Of course, the situation is different when we consider curved spaces – but the premise
is the same. To calculate how the two paths differ, and hence determine how they
affect the vector, we consider the covariant derivative along each path. We’ll
change the notation a bit, so that instead of “θ̂” and “φ̂” (those specifically refer to
angular coordinates), we’ll talk êµ and êν . In this case:

Aα along êµ first, êν second =⇒ Dν Dµ Aα

Aα along êν first, êµ second =⇒ Dµ Dν Aα
But this time, these derivatives aren’t equal! In fact, how they differ is explicitly
related to the background curvature:

Dµ Dν Aα − Dν Dµ Aα = Something related to Aα with three free indices!

Taking inspiration from how we defined the Christoffel symbols, we say that this
“thing” relating Aα with three free indices is the curvature tensor:

=⇒ Dµ Dν Aα − Dν Dµ Aα = Rαβµν Aβ

and since it needs a proper name, we’ll call it.... uh... the Riemann curvature
tensor (because I’m sure that, like with me, it’s the first name that popped into your
head).

Again, this result is some variety of linear combination of the components of the vec-
tor A. You can verify for yourself that it is consistent, from the index perspective.

How do we calculate the Riemann curvature tensor? Write it out! Remember, from
before, the covariant derivative of a vector is Dµ Aα = ∂µ Aα + Γαµλ Aλ , and for a
co-vector it’s Dµ Aα = ∂µ Aα − Γλµα Aλ . This holds for the partial derivative, too! In
fact, we get
Dµ ∂ν = ∂µ ∂ν − Γλµν ∂λ

Term 1
Dµ Dν Aα = Dµ ∂ν Aα + Γανλ Aλ

= ∂µ ∂ν Aα + ∂µ Γανσ Aσ + Γανσ ∂µ Aσ + Γαµσ ∂ν Aσ + Γαµσ Γσνβ Aβ

−Γσµν ∂σ Aα − Γσµν Γασβ Aβ

By defining Term 2 in a similar fashion, we can show that the difference results in
something that looks like this:

Dµ Dν Aα − Dν Dµ Aα = Rαβµν Aβ
as we saw above, where the Riemann curvature tensor is

Rαβµν = ∂µ Γαβν − ∂ν Γαβµ + Γαµσ Γσβν − Γανσ Γσβµ (12)

The biggest thing to note about this tensor is that it depends on first and second
derivatives of the metric only! Aha! “General relativity is a rank-2 tensor theory
of up to second order derivatives of the metric!” Higher-order derivatives don’t figure
into any condition in which we consider curvature. This will be important for the
Einstein equations.

We can also define this in terms of what’s called geodesic deviation, which is
essentially a measure of whether or not parallel lines will ever meet! In flat spacetime,
they don’t (curvature is zero). In spaces of positive curvature, they converge! And in
spaces of negative curvature, they diverge!