Professional Documents
Culture Documents
Geodesics 1
1.7 Geodesics
Note. A curve α
~ (s) on a surface M can curve in two different ways. First, α
~ can
bend along with surface M (the “normal curvature” discussed above). Second, α
~
can bend within the surface M (the “geodesic curvature” to be defined).
~r
~ 00tan = (ur00 + Γrij ui0 uj0 )X
α
(29)
~
~ 00nor = (Lij ui0 uj0 )U
α
and the parameters on the right hand side are defined in Section 5. α ~ 00nor reflects
the curvature of α
~ due to the bending of M and α~ 00tan reflects the curvature of α ~
within M . Now
n o
α 00 ~ r00 r i0 j0 ~ ~ =0
~ tan · U = (u + Γij u u )Xr · U
~ ~
~ = X1 × X2 ) and
(recall U
~ ×X
kX ~ 2k
~ 00tan · α
α ~0 = α
~ 00tan · α
~0 + 0 = α ~ 00tan · α
~0 + α~ 00nor · α
~0
(recall α ~ i and X
~ 0 = ui0 X ~i · U
~ = 0)
α00tan + α
= (~ ~ 00nor ) · α
~0 = α~ 00 · α
~0 = 0
α0 k = k~
(recall k~ α0 (s)k = 1 and 0 = d/ds).
~ 00tan = kg w
α ~ ×α
~ = kg (U ~ 0 ). (30)
~ × B)
(A ~ ·C
~ = (B
~ × C)
~ ·A
~ = (C
~ × A)
~ · B.
~
kg w
~ ·w ~ 00tan · w
~ =α ~=α ~ ×α
~ 00tan · (U ~ 0)
or
α00tan + α
kg = (~ ~ ×α
~ 00nor ) · U ~0
(since α ~)
~ 00nor is parallel to U
= α ~ ×α
~ 00 · U ~ ·α
~0 = U ~0 × α
~ 00 .
Note. If α
~ is a geodesic on M if and only if
and
~ ·α
U ~0 × α
~ 00 = 0. (32b)
Proof. Suppose
~
X(u, v) = (f (u) cos v, f (u) sin v, g(u)).
Let m
~ (s) = (f (s) cos v, f (s) sin v, g(s)) be a meridian of the surface (where we
assume the curve has been parameterized in terms of arclength s). Then
Now
~1 × X
X ~ 2 = (−f (s)g 0(s) cos v, −f (s)g 0 (s) sin v, f 0 (s)f (s))
1.7. Geodesics 4
and so
~1 × X
X ~2 (−f (s)g 0 (s) cos v, −f (s)g 0 (s) sin v, f 0 (s)f (s))
~
U= = .
~1 × X
kX ~ 2k (f (s)g 0(s))2 + (f 0(s)f (s))2
Therefore
~ ·m 1
U ~0×m
~ 00 =
(f (s))2{(g 0(s))2 + (f 0(s))2}
×{(f 0 (s)g 00(s) − f 00 (s)g 0(s))(−f (s)g 0(s)) cos v sin v
+(−f 0(s)g 00 (s) + f 00 (s)g 0(s))(−f (s)g 0(s)) cos v sin v, 0)}
1
= (0) = 0.
(f (s))2{(g 0(s))2 + (f 0(s))2}
Therefore m
~ (s) is a geodesic (see equation (32b)).
~ 1 , u2 ) be a surface and let gij (see page 34) and Γr (see page
Definition. Let X(u ij
43) be as defined in Sections 4 and 5. The Christoffel symbols of the first kind are
for i, j, k = 1, 2.
Definition. The Γrij defined in Section 1.5 are the Christoffel symbols of the second
kind.
Note. Since Γrij = Γrji (see (17), page 43) then Γijk = Γjik ; that is, there is
symmetry in the first two indices of the Christoffel symbols of the first kind. Also,
since (gij )−1 = (g ij ), we have Γm
ij = Γijk g
km
.
1.7. Geodesics 5
~ ij · X
Γijk = X ~k (34)
1 ∂gik ∂gjk ∂gij
Γijk = + − k (36)
2 ∂uj ∂ui ∂u
and
1 ∂gik ∂gjk ∂gij
Γrij = g kr + − k . (37)
2 ∂uj ∂ui ∂u
Proof. Since X ~
~ ij = Γr X ~
ij r + Lij U (by definition, see page 43) then
~ ij · X
X ~ k = Γr X
~ ~ ~ ~ r
ij r · Xk + (Lij U ) · Xk = Γij grk + 0 = Γijk
~r · X
establishing the first identity (recall grk = X ~ k ). Next,
∂gik ∂ ~ ~ ~ ij · X
~k + X
~ kj · X
~ i = Γijk + Γkji.
= [Xi · Xk ] = X (35a)
∂uj ∂uj
Permuting the indices:
∂gji ∂gkj
= Γ jki + Γ ikj and = Γkij + Γjik . (35b and 35c)
∂uk ∂ui
Now
1 1
Γijk = (2Γijk ) = (Γijk + Γjik )
2 2
1
= (Γijk + Γkji − Γkji + Γkij − Γkij + Γjik ) since Γijk = Γjik by
2
symmetry in the first two indices
1
= {(Γijk + Γkji) + (Γkij + Γjik ) − (Γjki + Γikj )} since Γkji = Γjki and
2
Γkij = Γijk by symmetry in the first two indices
1 ∂gik ∂gjk ∂gij
= + − k
2 ∂uj ∂ui ∂u
1.7. Geodesics 6
and the second identity is established. Finally, multiplying this identity on both
sides by g kr , summing over k and using the definition of Γrij we have
1 ∂gik ∂gjk ∂gij
Γrij = Γijk g kr = g kr + − k
2 ∂uj ∂ui ∂u
(recall (g ij ) = (gij )−1 ), and the third identity is established.
Note. Since the Christoffel symbols depend only on the metric form (or First
Fundamental Form), they are part of the intrinsic geometry of the surface M .
Corollary 1.7.A. Let X(u ~ 1 , u2 ) be a surface and let gij and Γr be as defined in
ij
~ 1 and X
Sections 4 and 5. If X ~ 2 are orthogonal coordinates, then
1 ∂g ir ∂gjr ∂g ij
Γrij = + − r
2grr ∂uj ∂ui ∂u
(no sums over any of i, j, r).
Corollary 1.7.B. With the hypotheses of the previous corollary (with i, j, r = 1, 2),
when j = r
1 ∂grr 1 ∂
Γrir = = [ln grr ]
2grr ∂ui 2 ∂ui
1.7. Geodesics 7
and when i = j 6= r
1 ∂gii
Γrii = − r .
2grr ∂u
Note. By symmetry, Γrij = Γrji , and so the previous two corollaries cover all possible
cases of orthogonal coordinates when i, j, r ∈ {1, 2} (i.e., when we deal with two
dimensions). In dimensions 3 and greater (in particular, in the 4 dimensional
spacetime of Chapter III) we have a third case which we state now, and address in
detail later:
we have the First Fundamental Form (or metric form) ds2 = Edu2 + Gdv2 on
~
surface X(u, v). In this notation, the Christoffel symbols are then
Eu Gv
Γ111 = 2E
Γ222 = 2G
Ev
Γ112 = Γ121 = 2E Γ221 = Γ212 = G
2G
u (40)
Γ122 = − G
2E
u
Γ211 = − E
2G .
v
1.7. Geodesics 8
Example 17, page 62. In the Euclidean plane, ds2 = du2 + dv2 . Therefore
E = G = 1 and all the Christoffel symbols are 0. Therefore a geodesic α
~ satisfies
for r = 1, 2, or ur00 = 0 for r = 1, 2. That is, u100 = u00 = 0 and u200 = v00 = 0.
Therefore u(s) = as + b and v(s) = cs + d for some a, b, c, d. Therefore, geodesics
in the Euclidean plane are straight lines.
Note. We will show in Theorem I-9 that the shortest path on a surface joining two
points is a geodesic. This theorem, combined with the previous example PROVES
that the shortest distance between two points in a plane is a straight line. Oddly
enough, you’ve probably never seen this PROVED before!
Example 18, page 62. Consider a sphere of radius r with “geographic coordi-
nates” (like latitude and longitude) u and v. Then the sphere is given by
~
X(u, v) = (r cos u cos v, r sin u cos v, r sin v)
(see Example 7, page 23). The metric form is (see page 33) ds2 = r2 cos2 vdu2 +
r2 dv2 (since there is no du dv term, F = g12 = g21 = 0 and these coordinates are
orthogonal). Therefore E = r2 cos2 v and G = r2 (a constant). Then Eu = Gu =
Gv = 0 and the nonzero Christoffel symbols are
Ev −2r2 cos v sin v
Γ112 = Γ121 = = = − tan v
2E 2r2 cos2 v
−Ev 2r2 cos v sin v
Γ211 = = = cos v sin v.
2G 2r2
It is shown in Exercise 1.7.14 (at the end of this section) that this implies geodesics
are great circles.
1.7. Geodesics 9
Note. In Example 19 page 62, it is shown that the Euclidean plane when equipped
with polar coordinates (which are orthogonal coordinates) yields geodesics which
are lines (as expected).
Note. In general, to determine the geodesics for a surface, requires that one solve
differential equations. This can be difficult (sometimes impossible to do in terms
of elementary functions). In Chapter III we will compute some geodesics in 4-
dimensional spacetime (in fact, planets and light follow geodesics if 4-D spacetime).
Proof. Let α ~ 1 (s), u2 (s)). Consider the family of curves of the form
~ (s) = X(u
for i = 1, 2, s ∈ [a, b] where vi are smooth functions with vi (a) = vi (b) = 0 for
~ 1, U 2 ) still joins α
i = 1, 2 (so X(U ~ (a) and α ~ 1 , U 2 ) ⊂ M , but otherwise vi
~ (b)), X(U
are arbitrary. We take α ~ 1(s, ε), U 2 (s, ε)).
~ ε(s) = X(U
1.7. Geodesics 10
where 1/2
∂U i ∂U j
1 2
λ(s, ε) = gij (U , U )
∂s ∂s
(the square root of the metric form of M along α
~ ε ). Now L has a minimum at
ε = 0 so Z b b
d d ∂
Z
[L(ε)] = λ(s, ε) ds = [λ(s, ε)] ds
dε dε a a ∂ε
(since λ and ∂λ/∂ε are continuous) satisfies
Z b
∂
L0 (0) = [λ(s, 0)] ds = 0.
a ∂ε
Now
" #
i j 1/2
∂λ ∂ ∂U ∂U
= gij (U 1, U 2 )
∂ε ∂ε ∂s ∂s
i j
i
∂U ∂U j
1 −1 ∂ 1 2 ∂U ∂U 1 2 ∂
= (λ(s, ε)) [gij (U , U )] + gij (U , U )
2 ∂ε ∂s ∂s ∂ε ∂s ∂s
1.7. Geodesics 11
i
j
∂U ∂ ∂U
+gij (U 1, U 2 )
∂s ∂ε ∂s
1 2
∂U i ∂U j
1 ∂ 1 2 ∂U ∂ 1 2 ∂U
= [gij (U , U )] + [gij (U , U )]
2λ(s, ε) ∂U 1 ∂ε ∂U 2 ∂ε ∂s ∂s
i
∂U j
1 2 ∂U ∂
+2gij (U , U )
∂s ∂ε ∂s
k i j i 2 j
1 ∂ ∂U ∂U ∂U ∂U ∂ U
= k
[gij (U 1 , U 2 )] + 2gij (U 1, U 2 )
2λ(s, ε) ∂U ∂ε ∂s ∂s ∂s ∂ε ∂s
∂gij k ∂U i ∂U j ∂U i ∂ 2 U j
1
= v + 2gij (notice that we sum
2λ(s, ε) ∂U k ∂s ∂s ∂s ∂ε ∂s
over k here since we treat the partial derivative with respect to U k
as if it were a subscript)
∂U k ∂U j
since = vk . With ε = 0, = vj and λ(s, 0) = 1 (because λ(s, ε) ds|ε=0 =
∂ε ∂ε
kα0 k ds = 1 ds; s is arclength on α~ =α~ 0 ) we have
∂λ 1 ∂gij k i0 j0
(s, 0) = k
v U U + 2gik U i0 vk0
∂ε 2 ∂U
and since ε = 0 implies U i = ui , then
∂λ 1 ∂gij k i0 j0
(s, 0) = k
v u u + 2gik ui0 vk0
∂ε 2 ∂u
and so
1 b ∂gij i0 j0 k
Z
0 i0 k0
L (0) = u u v + 2gik u v ds = 0.
2 a ∂uk
Now by Integration by Parts
Z b
2gik ui0 vk0 ds Let u = 2gik ui0 and dv = vk0 ds.
a
∂
Z
Then du = [2gik u ]ds and v = vk0(s)ds = vk .
i0
∂s
b
∂
Z
i0 k i0 k
= 2gik u v − [2gik u ]v ds
∂s a
1.7. Geodesics 12
b
∂
Z
= 0− [2gik ui0 ]vk ds since vk (a) = vk (b) = 0.
a ∂s
Therefore
Z b
1 ∂gij ∂
L0(0) = ui0 uj0 vk − [2gik ui0 ]vk ds
2 a ∂uk ∂s
Z b
1 ∂gij i0 j0 ∂
= k
u u − [2gik ui0 ] vk ds
2 a ∂u ∂s
= 0.
Rb
Since the integral must be zero for all arbitrary vk (and since a f (s)g(s) ds = 0
for arbitrary g(s) implies f (s) = 0), then the remaining part of the integrand must
be zero:
1 ∂gij i0 j0 ∂
k
u u − [gik ui0 ] = 0
2 ∂u ∂s
for k = 1, 2. Now when ε = 0, U i = ui and
∂ ∂
[gik ui0 ] = [gik (u1 , u2 )ui0]
∂s ∂s
i0
∂gik du1 ∂gik du2
i0 1 2 du
= + u + gik (u , u )
∂u1 ds ∂u2 ds ds
∂gik 10 ∂gik 20 i0
= u + u u + gik (u1 , u2 )ui00
∂u 1 ∂u 2
∂gik j0 i0 m00 ∂gik j0 i0
= j
u u + gmk u = j
u u + gmk um00 .
∂u ∂u
Therefore
1 ∂gij i0 j0 ∂
k
u u − [gik ui0 ] = 0
2 ∂u ∂s
for k = 1, 2 implies
1 ∂gij i0 j0 ∂gik i0 j0
k
u u − j
u u − gmk um00 = 0
2 ∂u ∂u
for k = 1, 2, or using the notation of equations (35a, 35b, 35c)
1
(Γikj + Γjki) − (Γkji + Γijk ) ui0 uj0 − gmk um00 = 0 (∗)
2
1.7. Geodesics 13
for k = 1, 2. Since Γikj ui0 uj0 = Γjkiui0 uj0 (interchanging dummy variables i and j)
and Γkji = Γjki (symmetry in the first and second coordinates) then
1 1
(Γikj + Γjki) − Γkji ui0 uj0 = (Γjki + Γjki) − Γjki ui0 uj0 = 0
2 2
and the above equation (∗) becomes
Note. Again, Theorem I-9 along with Example 17 shows that the shortest distance
between two points in the Euclidean plane is a “straight line.” Theorem I-9 along
with Example 18 show that the shortest distance between two points on a sphere is
part of a great circle (explaining apparently unusual routes on international airline
flights).
Note. The converse of Theorem I-9 is not true. That is, there may be a geodesic
joining points which does not minimize distance. (Recall that we set L0 (0) ≡ 0,
but did not check L00 (0); we may have a maximum of L!) For example, we can
travel the six miles from Johnson City to Jonesborough (along a very small piece
of a geodesic), or we can travel in the opposite direction along a very large piece of
a geodesic (≈ 24, 000 miles) and travel around the world to get to Jonesborough
(NOT a minimum distance).
1.7. Geodesics 14
Note. Not all surfaces may allow one to create a geodesic joining arbitrary points.
For example, the Euclidean plane minus the origin does not admit a geodesic from
(1, 1) to (−1, −1).
Note. In the next theorem, we prove that for any point on a surface, there is a
unique (directed) geodesic through that point in any direction.
~ i(u1 , u2 ) · X
f (0) = gij (u10 , u20 )ui0(0)uj0 (0) = X ~ j (u1 , u2 )ui0(0)uj0 (0)
0 0 0 0
1.7. Geodesics 15
~ 1 2 ~ 1 2 i j i~ 1 2 j ~ 1 2
= Xi(u0 , u0 ) · Xj (u0 , u0 )v v = v Xi(u0 , u0 ) · v Xj (u0 , u0 )
= ~v · ~v = k~v k2 = 1.
Next,
∂gij k0 i0 j0
f 0(t) = k
u u u + gij ui00 uj0 + gij ui0 uj00 .
∂u
Since
∂gij
= Γikj + Γjki (equation (35b), page 60)
∂uk
= Γrik grj + Γrjk gri (equation (33), page 59)
= gjr Γrik + gir Γrjk (symmetry of gij )
then
f 0(t) = (gjr Γrik + gir Γrjk )ui0 uj0 uk0 + grj ur00 uj0 + gir ui0 ur00
= gir ui0 (ur00 + Γrjk uj0 uk0 ) + grj uj0 (ur00 + Γrik ui0 uk0 )
= 0 (from the first condition of the ODE).
2
ds
Therefore f (t) is a constant and f (t) = 1. Hence = f (t) = 1 and t = s
dt
(that is, t is arclength). Therefore α ~ 1 (s), u2(s)) is the desired geodesic.
~ (s) = X(u
Example (Exercise 1.7.14(a)). If M has metric form ds2 = Edu2 + Gdv2 with
Eu = Gu = 0, then a geodesic on M satisfies
√
du h G
=√ √
dv E E − h2
for some constant h (see Exercise 1.7.12). Use this above equation to show that a
geodesic on the geographic sphere
~
X(u, v) = (R cos u cos v, R sin u cos v, R sin v)
1.7. Geodesics 16
satisfies
du h sec2 v h sec2 v
= √ = √
dv R2 − h2 sec2 v R2 − h2 − h2 tan2 v
where h is a constant.
Solution. First,
~ 1 = (−R sin u cos v, R cos u cos v, 0)
X
~ 2 = (−R cos u sin v, −R sin u sin v, R cos v)
X
~1 · X
E = g11 = X ~ 1 = R2 cos2 v
~2 · X
G = g22 = X ~ 2 = R2 sin2 v + R2 cos2 v = R2 .
Then
√
du h R2
= √ √
dv R2 cos2 v R2 cos2 v − h2
hR
= √ since v ∈ (−π/2, π/2)
R cos v R2 cos2 v − h2
h sec v h sec2
= √ =p
cos v R2 − h2 sec2 v R2 − h2 (1 + tan2 v)
h sec2 v
= √ .
R2 − h2 − h2 tan2 v
Therefore
w h tan v
cos(u − u0 ) = √ =√ .
R 2 − h2 R 2 − h2
√
With γ = −h/ R2 − h2 we have
cos(u − u0 ) + γ tan v = 0.
Example (Exercise 1.7.14(c)). Show that the equation given in (b) when writ-
ten in Cartesian coordinates is a linear equation of the form αx + βy + γz = 0 and
so represents the intersection of the sphere with a plane passing through the origin
(and therefore the geodesic is a great circle).
Solution. Multiplying by R cos v on each side of the equation gives R cos v cos(u −
u0 ) + γR sin v = 0 or R cos v(cos u cos u0 + sin u sin u0 ) + γR sin v = 0 or