Curves on surfaces: Geodesics

Review of the geodesic curvature of the curves

in the surface

We first review the geodesic curvature of the curves in the surface introduced in section
4.1. Let M be a surface and let be a curve in the surface M, parametrized by its
arc-length (.i.e. k k = 1). By the trick of the unit-vector, is perpendicular to
, so
(4.2.1)

= n n + g n .

The scalar g is called the geodesic curvature of .

We have
g = (n ).

Geodesics

Definition 4.2.1. A curve in the surface M is said to be geodesic if its geodesic

curvature is identically zero.
We now look at, geometrically, what a geodesic curve looks like. Let be a unitspeed curve. Then is geodesic if and only if g = 0, i.e. = n n. So a curve in
the surface M is geodesic if and only if the vector is either zero or perpendicular
to the surface at (t)(i.e. parallel to its unit normal), for all values of the parameter
t.
From this, we see that any (part of a) straight line is a geodesic, and all great
circles on a sphere are geodesics (Click here to see the picture of great circles). Also,
any curves C on the surface, which are obtained by the intersection of S with a plane
that are perpendicular to the surface, are geodesics.
Another description a geodesic is that a geodesic is a path of extremal (maximal
or minimal) length between two points. See Theorem 4.2.3. Hence, we can view the
geodesic curve as the analogy of the straight line in the (flat) space R3 (as we see
that every straight line in geodesic in R3 and it is the shortest curve connecting any
two points P and Q).

Geodesic Equations

In this section, well derive the Geodesic Equations for the curves in a parametrized
surfaces. Let x : U R3 be a parametrized surface(we call it M), and let be
a curve in M. Then, we can write(since the curve is in M, it should satisfy the
parametrization x of M),
(t) = x(u(t), v(t)),
where u(t), v(t) are two differentiable functions. Then
Theorem 4.2.1 A curve on a surface M is geodesic if and only if, for (t) =
x(u(t), v(t)) (assumed to be the unit-speed parameterization), the following two equations are satisfied:
(4.2.2)

1
(Eu + F v ) = (Eu u2 + 2Fu u v + Gu v 2 ),
2

(4.2.3)

1
(F u + Gv ) = (Ev u2 + 2Fv u v + Gv v 2 ),
2
2

where {E, F, G} is the first fundamental form of x.

Proof. By definition = n n + g n . The fact that is geodesic implies that
= n n, i.e. is perpendicular to TP (M) for every P M. Since {xu , xv }is a
basis of TP (M), is a geodesic if and only if is perpendicular to xu and xv . This
gives two equations xu = 0, xv = 0. We now show that these two equations
give (4.2.2) and (4.2.3). In fact, since (t) = x(u(t), v(t)), by the chain rule,
= xu u + xv v .
Hence the equation xu = 0 becomes
(xu u + xv v ) xu = 0.
Hence
(4.2.4) ((xu u + xv v ) xu ) = (xu u + xv v ) xu + (xu u + xv v ) xu = (xu u + xv v ) xu .
By the definition, xu xu = E, xv xu = F , xv xv = G, so (4.2.4) becomes
(4.2.5)

(Eu + F v ) = (xu u + xv v ) xu .

We now calculate the term on the right-hand side above. By the chain rule, xu =
xuu u + xuv v , so
(xu u + xv v ) xu = (xu u + xv v ) (xuu u + xuv v )
= (xu xuu )u2 + (xu xuv + xv xuu )u v + (xv xuv )v 2 .
From xu xu = E, by taking the partial, we obtain 2xuu xu = Eu . Similarly,
2xuv xv = Gu and xvu xu + xv xuu = Fu . Thereofore,
1
(xu u + xv v ) xu = (Eu u2 + 2Fu u v + Gu v 2 ).
2
This, together with (4.2.4) proves (4.2.2). Similarly, using the equation xv = 0,
we can derive (4.2.3). This finishes the proof.
Q.E.D.
There is another way to derive and express the geodesic equations in terms of the
Christoffel symbols. Let (t) = x(u(t), v(t)) be a unit-seppd curve. Similar to the
above calculation,
= xu u + xv v .
= u xu + u (xuu u + xuv v ) + v xv + v (xvu u + xvv v ).
3

Write
xuu = 111 xu + 211 xv + en
xuv = 112 xu + 212 xv + f n
xvv = 122 xu + 222 xv + gn
and plug them into the above formula for , to get
= (u(t) + 111 u(t)2 + 2112 u(t)v (t) + 122 v (t)2 )xu
+ (v(t) + 211 u (t)2 + 2212 u (t)v (t) + 222 v (t)2 )xv + ()n.
Now is geodesic if and only if its tangential component is zero, i.e.
(u(t) + 111 u(t)2 + 2112 u (t)v (t) + 122 v (t)2 )xu
+ (v(t) + 211 u (t)2 + 2212 u (t)v (t) + 222 v (t)2 )xv = 0.
Since xu , xv is linearly independent, it means that
u(t) + 111 u (t)2 + 2112 u (t)v (t) + 122 v (t)2 = 0
v(t) + 211 u (t)2 + 2212 u (t)v (t) + 222 v (t)2 = 0.
This is another expression of the geodesic equations using the Chirstoffel symbol.
They are the same as (4.2.2) and (4.2.3).
The geodesic equations are non-linear differential equations, and are usually difficult or impossible to solve explicitly. However, form the theory of ordinary differential
equations, we can solve if (theoretically) if everything is smooth. So
Theorem (existence). Let P be a point of s surface M, and let v Tp M. Then
there exists a unique unit-speed geodesic on M passing through P and has tangent
vector v there.
Example 4.2.1. Let (t) be a part of great circle in the unit sphere S 2 . Then it is
geodesic.
Proof. Let
x(u, v) = (sin u cos v, sin u sin v, cos u),

0 < u < , 0 < v2

be a parametrization of the unit sphere S 2 . Without loss of generality (by doing

a rotation and translation), we can assume that (t) is the intersection of S 2 with
the xy-plane (u = /2), i.e. (t) = (cos t, sin t, 0). We now verify (4.2.2) and
4

(4.2.3). From section 3.3, we see that the First Fundamental Form of x(u, v) is is:
E = xu xu = 1, F = xu xv = 0, G = xv xv = sin2 u. Hence the equations (4.2.2)
and (4.2.3) become
(4.2.6)

1
u (t) = 2 sin u(t) cos u(t)(v (t))2 ,
2

(4.2.7)

(sin2 u(t)v (t)) = 0.

Now, since (t) = (cos t, sin t, 0), when we write (t) = x(u(t), v(t)) we see that
u(t) = /2, v(t) = t. Hence, it is easy to see that (4.2.6) and (4.2.7) are satisfied.
This shows that every (part of) great circle in the unit sphere S 2 is geodesic.
Q.E.D.
We now show that every geodesics in in the unit sphere S 2 must be a (part of)
great circle.
Example 4.2.2. Every geodesics in in the unit sphere S 2 must be a (part of) great
circle.
Proof.

Let
x(u, v) = (sin u cos v, sin u sin v, cos u),

0 < u < , 0 < v < 2

be a parametrization of the unit sphere S 2 and let

(t) = x(u(t), v(t)) = (sin u(t) cos v(t), sin u(t) sin v(t), cos u(t))
be a unit-speed curve on S 2 . As in Example 4.2.2, the geodesic equations become
(4.2.8)

1
u (t) = 2 sin u(t) cos u(t)(v (t))2 ,
2

(4.2.9)

(sin2 u(t)v (t)) = 0.

We now determine u(t) and v(t) by solving (4.2.8) and (4.2.9) jointly. In addition to
(4.2.8) and (4.2.9), we also need to use the condition that (t) is a unit-speed curve,
i.e. k (t)k = 1. To calculate k (t)k, we need to use the first fundamental form
(see Section 3.3, page 2). Since, from Example 4.2.1, E = 1, F = 0, G = sin2 u, we
have that k (t)k2 = E(u (t))2 + 2F u(t)v (t) + G(v (t))2 = (u(t))2 + (v (t))2 sin2 u(t).
Hence the condition that (t) is a unit-speed curve becomes
(4.2.10)

(u (t))2 + (v (t))2 sin2 u(t) = 1.

5

We now use equations (4.2.8), (4.2.9) and (4.2.10) to solve for u(t) and v(t).
From (4.2.9), sin2 u(t)v (t) = C, where C is a constant.
Case 1: If C = 0, then v (t) = 0, so v(t) is constant, and is part of a meridian,
hence is (part of) a great circle.
Case 2: C 6= 0. Then v (t) = C/ sin2 u(t) = C csc2 u(t). Substituting this into
(4.2.10) gives
C2
sin2 u(t) = 1,
(u (t))2 +
4
sin u(t)
i.e.
C2
(u(t))2 = 1
= C 2 csc u(t).
2
sin u(t)
So along the curve (we regard u is a parameter and v is a function of u), we have
dv
du

!2

Hence

dv/dt
du/dt

!2

v (t)
u (t)

!2

csc4 u
C 2 csc4 u
=
.
1 C 2 csc2 u
C 2 csc2 u

csc2 udu
csc2 udu
dv = 2
= 2
.
(C csc2 u)1/2
(C 1 cot2 u)1/2

Integrating it on both sides, we get

(v v0 ) =

csc2 udu
,
(C 2 1 cot2 u)1/2

where v0 is a constant. The integral can be evaluated by making the change of variable
= cot u, so
(v v0 ) =

d
.
= sin1
2
2
1/2
2
(C 1 )
(C 1)1/2

Hence, we have
cot u = (C 2 1)1/2 sin(v v0 ) = (C 2 1)1/2 (sin v cos v0 cos v sin v0 ),
which gives that
cos u = (C 2 1)1/2 (sin u sin v cos v0 ) sin u cos v sin v0 .
This implies that the coordinates (of in R3 ) x = sin u cos v, y = sin u sin v and
z = cos u satisfy the equation z = ax + by where a = ((C 2 1)1/2 sin v0 ), b =
6

(C 2 1)1/2 cos v0 . This shows that is contained in the intersection of S 2 with a

plane passing through the center of the sphere. Hence, it is (part of) a great circle.
Q.E.D.
Clairaut relation for geodesic on the unit sphere S 2 : Let be the angle
between the geodesic and the parallel (i.e u = u0 ), i.e. the angle between and
the tangent vector of the parallel. Note htat the parallel is x(u0 , v), so its tangent
vecotr is xv . Hence,
xv
.
cos =
k kkxv k

By the assumption, k k = 1, when write (t) = x(u(t), v(t)), then (t) = u (t)xu +
v (t)xv . In the unit sphere S 2 case, under its parametrization x(u, v) = (sin u cos v, sin u sin v, cos u),
we see from above E = 1, F = 0, G = sin2 u, so xu xv = 0, xv xv = G = sin2 u.
Hence, at the point x(u0 , v0 ) (or with (t0 ) = x(u0 , v0 )),
(t0 ) xv
k kkxv k
(u(t0 )xu + v (t0 )xv ) xv
=
kxv k

v (t0 )kxv k2
= v (t0 )kxv k = v (t0 ) sin u0.
=
kxv k

cos =

From (4.29) above, we have sin2 u(t)v (t) = C, where C is a constant, hence v (t0 ) =
C
. Thus,
sin2 u0
C
cos =
,
sin u0
i.e. sin u0 cos = C. Notice that sin u0 = raidus r of the parellel (latitude). This
gives the following Clairaut relation:
r cos = C,
where r = raidus of the parellel (latitude) at the intersection point, os the angle
between the geodesic and the parallel the intersection point, C is a constant.

Geodesics on Surfaces of Revolution

Let
x(u, v) = (f (u) cos v, f (u) sin v, g(u))

be a surface of revolution, where we assume that f > 0, and (f (u))2 + (g (u))2 = 1.

Then its first fundamental form is E = 1, F = 0 and G = f 2 (u). Let (t) =
x(u(t), v(t)) be a unit-speed curve in x. The the geodesic equations (see (4.2.2) and
(4.2.3) for are
(4.2.11)

u = f (u)f (u)(v )2 ,

d 2
(f (u)v ) = 0.
dt

In addition, from the condition that (t) is a unit speed, we have one more equation
(4.2.12)

u2 + f 2 (u)v 2 = 1.

From this, we can make the following easy deductions:

Theorem 4.2.2 On the surface of revolution
x(u, v) = (f (u) cos v, f (u) sin v, g(u)),
(i) every meridian is a geodesic;
(ii) a parallel u = u0 (say) is a geodesic if and only if df /du = 0 when u = u0 , i.e.
u0 is a stationary point of f .
Proof. On a meridian, we have v =constant, so the second equation in (4.2.11) is
obviously satisfied. Equation (4.2.12) gives that u = 1, so u is constant and the
first equation in (4.2.11) is also satisfied. So (i) is proved.
For (ii), note that if u = u0 is constant, then by (4.2.12), v = 1/f (u0) is nonzero, so the first equation holds in (4.2.11) holds only if df /du = 0. Conversely, if
df /du = 0 when u = u0 , the first equation in (4.2.11) obviously holds, and the second
holds because v = 1/f (u0 ) and f (u) = f (u0) are constant.
Q.E.D.
Theorem (Clairaut relation) Let
x(u, v) = (f (u) cos v, f (u) sin v, g(u))

()

be a surface of revolution, where we assume that f > 0, and (f (u))2 + (g (u))2 = 1

(i.e. the profil curve is unit-speed). Then the geodesics on a surfae of revolution
satisfy the equation
r cos C
where r is the distance from the axis of revolution and is the angle between the
geodesic and parallel. Conversely, any (constant speed) curve satisfying (*) that is
not parallel is a geodesic.
8

Proof. As we did abovem its first fundamental form is E = 1, F = 0 and G = f 2 (u).

Let (t) = x(u(t), v(t)) be a unit-speed curve in x. The the geodesic equations (see
(4.2.2) and (4.2.3) for are
(4.2.11).

u = f (u)f (u)(v )2 ,

(f 2 (u)v ) = 0

Notice that (f 2 (u)v ) = 0 f 2 (u)v = C Gv = C since G = f 2 (u).

Let be the angle between the geodesic and the parallel (i.e u = u0 ), i.e. the
angle between and xv . Then, using xu xv = F = 0, and k k = 1,
(u (t)xu + v (t)xv ) xv
v (t)xv xv
xv
=
=
= v (t)kxv k = v (t)G1/2 .
cos =

k kkxv k
kxv k
kxv k

Using Gv = C, we get G cos = C. Since G = f 2 (u), this gives

f (u) cos = C.
Notice that f (u) is the distance from the axis of revolution, this proves the Clairaut
relation:
r cos = C.
Conversely, assume that the Clairaut relation r cos = C, i.e. f (u) cos = C
is satisfied. For the above argument, this is equivalent to f 2 (u)v = C. So the
equation (f 2 (u)v ) = 0 is satiesfied. Thus we only need to verify the first equation,
i.e. u = f (u)f (u)(v )2 . To do so, we use the condition that kk is constant. From
kk2 = u2 + Gv 2 is constant, we have
u (t)2 + f (u(t))2v 2 (t) = 0.
We differntiate on both sides and obtain
u (t)u (t) + f (u(t))2 v (t)v (t) + f (u(t))f (u(t)u(t)v (t)2 = 0

substituting for v (t) using v + 2ff u v = 0 (which can be derived from (f 2 (u)v ) = 0),
we have
u (t)(u (t) f (u(t)))f (u(t)u(t)v (t)2 ) = 0.
In other words, provided u (t) 6= 0 (i.e. the curve is not parallel), a constant speed
curve is geodesic.
Q.E.D.

Geodesics as Shortest Paths

In this section, we shall state and prove a theorem indicate that a a geodesic is a path
of extremal (maximal or minimal) length between two points. To test whether has
smaller(or larger) length than any other paths in M passing through two fixed points
P and Q, we consider a smooth family of curves on M passing through two fixed
points P and Q. By such family, we mean a curve on M, for each (, ),
such that
(i) there is an > 0 such that (t) is defined for all t (, ) and all (, );
(ii) for some a, b with < a < b < , we have (a) = P, (b) = Q for all
(, ) (this means that every curve in this family passing through P at t = a
and Q at t = b);
(iii) 0 = (this means is among the middle of the family of the curves)
and is unit-speed.
The arc-length of the curve between P and Q (which depends on ), by the
formula, is
Z
( ) =

k (t)kdt.

We have the following theorem

Theorem 4.2.3 With the above notation, the unit-speed curve is a geodesic if and
only if
d
( ) = 0
when = 0
d
for all families of curves with 0 = . This means that, among all (family) of
curves connecting P, Q, 0 = is the curve with extremal (maximal or minimal)
arc-length, since the derivative of the arc-length function ( ) (w.r.t. ) is zero at
= 0.
Proof. Although is unit-speed, the curves may not be unit-speed, its arc-length
is
Z b
d
( ) =
k (t)kdt.
dt
a

Since we treat (t) as a (vector-valued) function of two variables of and t, we write

d

(t) as (t). Hence

dt
t
( ) =

k (t)kdt =
t

10

(t), (t)
t
t

+1/2

dt.

Differentiating both sides with respect to , we get

| =0 ( ) =
| =0
(0) =

(t), (t)
t
t

| =0
(t), (t)

t
t
1(
2

2
(t), t
(t)
t

dt

(t), t (t)
t
E1/2

(t)

| =0
t

| =0 +

(t),
t

+1/2

+1/2

| =0

dt

2
(t), t
(t) | =0
t
dt.
D
E1/2

a
(t), (t)

=0
t
t
b

Since is unit-speed, and (t)| =0 = (t), we have

*

(t), (t) | =0 = 1.
t
t

Hence

(0) =

2
(t), (t) | =0 dt.
t
t

Noticing that

(t), (t) =

2
2

(t), (t) +
(t), 2 (t) ,
t
t

i.e.
*

2
(t), (t) =
t
t
t

2

(t), (t)
(t), 2 (t) ,

hence
*

Z b

2

(0) =
(t), (t) | =0 dt
(t), 2 (t) | =0 dt

t
a t
a
+
+
*
Z b*
2




b
(t), (t) | =0 |a
(t), 2 (t) | =0 dt.
=

t
a

Now, since (a) = P, (b) = Q which are independent of , we have

(t)|t=a = 0,
(t)|t=b = 0.

11

Hence

(t), (t) | =0 |ba = 0,

which implies that

(0) =

2
(t), 2 (t) | =0 dt.

Again, use (t)| =0 = (t), we have

2
(t)| =0 = (t) = n n(t) + g (n(t) T(t)).
t2
Also, the cuvres (t) are on the surface M, hence
M (i.e. in the tangent space), so
*

Thus we have

(0) =

(t) are tangent to the surface

(t), n(t) | =0 = 0.

(t), g (n(t) T(t)) | =0 dt.

If is geodesic, then g = 0, so (0) = 0. This means that, among all (family)

of curves connecting P, Q, 0 = is the curve with extremal (maximal or minimal)
arc-length, since the derivative of the arc-length function ( ) (w.r.t. ) is zero at
= 0.
Conversely, by a standard method of variation, we can show that if is a curve
connecting P and Q such that among all (family) of curves connecting P, Q, 0 =
is the curve with extremal (maximal or minimal) arc-length, then it is geodesic(this
part of the proof is omitted).

Geodesic Coordinates

The existence of geodesics on a surface M allows us to construct a very useful local

parametrization for M. For this, let P be a point of M, and let , with parameter v
say, be a unit-speed geodesic on M with (0) = P . For any value v, there is a unique
v , with parameter u, say, such that
v (o) = (v) and which is
unit speed geodesic
v
(u).
perpendicular to at (v). We define x(u, v) =
12

Theorem With the above notation, there is an open subset U of R2 containing (0, 0)
such that x(u, v) : U R3 is a parametrization of M More over, E = 1, F = 1
and G = G(u, v) is a smooth function on U with G(0, v) = 1, Gu (0, v) = 0 whenever
(0, v) U.
Proof. Note first that, for any value of v,
xu (0, v) =

d v
d v
d
(u)|u=0, xv (0, v) =
(0) = (0),

du
dv
dv

and that these are perpendicular unit vectors by construction. We write

x(u, v) = (f (u, v), g(u, v), h(u, v)).
It follows that the jacobian matrix

fu fv

gu gv
hu hv

has rank 2 when u = v = 0. Hence at least one of the its three 2 2 submatrix is
invertibale at (0, 0), say
!
fu fv
.
gu gv
By the inverse function theorem, there is an open subset U of R2 containing (0, 0)
such that the map given by F (u, v) = (f (u, v), g(u, v)) is bijection from U to F (U)
and such that its inverse is form F (U) to U is also smooth. The matrix is then also
invertible for all (u, v) U, so xu , xv are linearly independent for (u, v) U. It
follows that x(u, v) : U R3 is a parametrization of M.
As foe th efirst fundamental form of x,
E = kxu k2 = k

d v
(u)k2 = 1,

du

v is unit-speed. Next, we apply the second geodesic equation to

v . The
because
unit-seppd parameter is u and v is constant, so we get Fu = 0. But when u = 0, e
have already seen that xu , xv is perpendicular, so F = 0. Hence F 0. Now
G(0, v) = kxv (0, v)k2 = k

d
k2 = 1
dv

because is unit-speed. Finally from the first geodesic equation (2) applied to ,
for which u = 0 and v is the unit-speed parameter, we get Gu (0, v) = 0. This proves
theorem.
13

The parametrization constructed above is called the geodesic parameterization,

and u, v are called geodesic coordinates.
Example. If P is point on the equator of the unit-sphere S 2 , take to be the equator
be the meridian parameterised by latitude
with parameter the longitude , and let
and passing through the point on the equator with longitude . The corresponding
parametrization is the usual latitude longitude patch, for which the first fundamental
form is d2 + cos2 d2 , i.e. E = 1, F = 0 and G = cos2 .

14