3.4 notes on notation!

partial derivative ∂Aa /∂xc = ∂c Aa = Aa,c (I won’t use the latter as its too
easy to lose a comma!)
covariant derivative ∇c Aa = Aa;c (I will use the semicolon notation because
physicists tend to freeze when they see del!!)
absolute derivative ∇s Aa = DAa /ds n o
Christoffel symbols (or connection coefficients) Γabc or bca or {a, bc}

3.5 Example: 2D flat space

The metric for flat space in cartesian coordinates gAB = diag(1, 1) DOES
NOT DEPEND ON POSITION. So the partial derivatives of the metric are
ZERO. so ΓABC = 0.
BUT this is not true if we’d done it in terms of polars, ds2 = dx2 + dy 2 =
dr 2 + r 2 dθ2 where gAB = diag(1, r 2 ). Here the derivaties are NOT zero.
So Christoffel symbols are like the metric - they do tell us about curvature
(what we are interested in) but they also tell us about what coordinate system
we have chosen (which isn’t at all fundamental). so if we are going to bother
calculating them, we may as well choose a real curved surface to play with.

3.6 Surface of a sphere

before we worked out the metric components for flat space
ds2 = dr 2 + r 2 dθ2 + r 2 sin2 θdφ2
we can describe a sphere just by saying we are at fixed radius r = a so dr = 0.
so
ds2 = a2 dθ2 + a2 sin2 θdφ2
i.e. gAB = diag(a2 , a2 sin2 θ)
Christoffel symbols are defined as
1
Γfca = g f b (∂c gab + ∂a gbc − ∂b gca )
2
so we need to also find the covariant metric components g AB from g AB gBC =
δCA . In general this leads to a set of simultaneous equations to solve
g AB gBC = g A1 g1C + g A2 g2C = δCA

1
BUT for diagonal metrics we don’t have to do a matrix inverse to solve these!
in a diagonal metric all cross terms are zero so we get g AB gBA = g AA gAA = 1
or g AA = 1/gAA ONLY FOR DIAGONAL METRICS!

g AB = diag[1/a2 , 1/(a2 sin2 θ)]

So in our 2D space we have

Γθθθ = 1/2g θB (∂θ gθB + ∂θ gBθ − ∂B gθθ )

as g θB = 0 except for B = θ then this collapses to

Γθθθ = 1/2g θθ (∂θ gθθ + ∂θ gθθ − ∂θ gθθ ) = 1/2 . 1/a2 . 0 = 0

Γθθφ = Γθφθ = 1/2g θB (∂φ gθB + ∂θ gBφ − ∂B gθφ )

again g θB is only nonzero for B = θ since the metric is diagonal so

Γθθφ = Γθθφ = 1/2g θθ (∂φ gθθ + ∂θ gθθ − ∂θ gθφ ) = 1/2 . 1/a2 . 0 = 0

Γθφφ = 1/2g θB (∂φ gφB + ∂φ gBφ − ∂B gφφ ) = 1/2 . 1/a2 (0 + 0 − ∂θ gφφ )

= 1/2 . 1/a2 (−2a2 sin θ cos θ) = − sin θ cos θ

Γφθθ = 1/2g φB (∂θ gθB + ∂θ gBθ − ∂B gθθ ) = 1/2g φφ(∂θ gθφ + ∂θ gθφ − ∂φ gθθ ) = 0
Γφθφ = Γφφθ = 1/2g φB (∂φ gθB + ∂θ gBφ − ∂B gθφ ) = 1/2g φφ(∂φ gθφ + ∂θ gφφ − ∂φ gθφ )
= 1/2 1/(a2 sin2 θ)∂θ gφφ = 1/2 1/(a2 sin2 θ)2a2 sin θ cos θ = cot θ
Γφφφ = 1/2g φB (∂φ gφφ + ∂φ gφφ − ∂φ gφφ ) = 0

3.7 Geodesic equations

And now we have everything we need to get the geodesic paths (inertial
frames). If we were differential geometers, then we would define a geodesic
as the shortest distance between two points. But we are phycisics so we can
use the alternative definition that these are inertial frames. and in an inertial
frame the velocity DOES NOT CHANGE - there are no forces to prodsce
an acceleration. So velocity v = v α eα = dxα /dτ eα If there is no change in

2
this then its derivative is zero but we also saw that there can be swings in
a vector which arise from curved space. In other words we have to cast this
in tensor language and say its the ABSOLUTE derivative which is zero in
order to simply parallel transport the velocity vector. So the components
can change like

dv α dxβ dxγ
+ Γαβγ =0
dτ dτ dτ
d dxα dxβ dxγ
+ Γαβγ =0
dτ dτ dτ dτ
d2 xα β
α dx dx
γ
+ Γ βγ =0
dτ 2 dτ dτ
or using notation where dot means derivative w.r.t. τ

ẍα + Γαβγ ẋβ ẋγ = 0

This works for any parameter u linearly related to path length s i.e. for
u = A + Bs. Then the affinely parameterized geodesic equation in an
N dimensional manifold is
d2 xa b
a dx dx
c
+ Γbc =0
ds2 ds ds
or in slighly cleaner notation, where dot denotes derivative w.r.t. s

ẍa + Γabc ẋb ẋc = 0

So now we know how to do geodesic paths, as well as derivatives.

3.8 Example: Geodesics in flat space

we saw that in flat space all the Γi jk = 0 so the geodesic equations becomes

ẍa = 0

were dot denotes derivative wrt path length s. integrate to get ẋa = A and
xa = As + B so geodsics in flat space have constant velocity and direction -
Newtonian inertial frame (constant motion, straight line!)

3
3.9 Example: Geodesics on a sphere
in general its too hard to do geodesics in full generality. so often we just
choose a path and see if its a geodesic! and a natural one to choose is a path
defined by only one of the parameters - its called a parameter curve.
so on our sphere, we know that for a sphere the only non-zero christoffel
symbols are Γθφφ = − sin θ cos θ, and Γφθφ = Γφφθ = cot θ.
Geodesic paths satisfy the equation

dxA B
A dx dx
C
+ ΓBC =0
ds2 ds ds
suppose the path s is just a change in θ then s = aθ so there is no dependance
on φ i.e. x2 = φconstant so dφ/ds = 0. While for θ we have dθ/ds = 1/a
and d2 θ/ds2 = 0. For θ then
B C
d2 θ θ dx dx d2 θ dφ dφ
+ ΓBC = + Γθφφ =0
ds 2 ds ds ds 2 ds ds
so this looks good. Check also that our condition for φ holds
B C
d2 φ φ dx dx d2 φ φ dθ dφ φ dφ dθ
+ ΓBC = + Γθφ + Γφθ =0
ds2 ds ds ds2 ds ds ds ds
so all such paths are geodesics. - these are GREAT CIRCLES.