You are on page 1of 110

CHAPTER 3: The Fundamental Forms on Surfaces

Mathematics 246

Institute of Mathematics
University of the Philippines-Diliman

1 / 110
Overview

1 The First Fundamental Form

2 The Gauss Map

3 The Second Fundamental Form

4 Normal and Principal Curvatures

5 Gaussian and Mean Curvatures

6 Surfaces of Revolution

7 Ruled Surfaces

8 Minimal Surfaces

2 / 110
3.1 The First Fundamental Form

In this section, our goal is to measure length, angle, and area on a surface
patch without having to re-write everything in the ambient space R3 .

Recall: In R3 , measurements are done using the inner product of vectors.


Let~v = 〈v1 , v2 , v3 〉 and ~
w = 〈w1 , w2 , w3 〉 be vectors in R3 . We define the inner
product of~v and ~ w as
~v · ~
w = v1 w1 + v2 w2 + v3 w3 .
Using this, we can compute p
length of a vector: k~vk = ~v ·~v
angle between vectors: If θ is the angle between~v and ~
w, then
~v · ~
w
cos θ =
k~vkk~ wk

area of parallelogram formed by ~u and~v :


q
¢2
A = k~v × ~ wk2 − ~v · ~
¡
wk = k~vk2 k~ w
3 / 110
3.1 The First Fundamental Form

Definition 3.1.1. The First Fundamental Form


Let S be a regular surface and p ∈ S. The first fundamental form Ip (·, ·) is the
restriction of the usual dot product in R3 to the tangent space of S at p.
That is, for~a,~b ∈ Tp S, Ip (~a,~b) =~a ·~b.

For each point p ∈ S, the first fundamental form Ip (·, ·) is defined only
on the tangent space Tp S. That is, Ip (·, ·) : Tp S × Tp S → R.
There exists a unique matrix that represents Ip (·, ·) with respect to the
standard basis on Tp S. This matrix has entries which are functions that
depend on the point p ∈ S.
If around p, the surface S can be parametrized by X with p = X (q), then
a natural basis for Tp S is {Xu (q), Xv (q)}. Hence, we can express the first
fundamental form in terms of these basis vectors.

4 / 110
3.1 The First Fundamental Form

Suppose that around p, the regular surface S can be parametrized by X with


p = X (q). Let~a and ~b be in Tp S. Then we can write

~a = a1 Xu (q) + a2 Xv (q)
~b = b1 Xu (q) + b2 Xv (q)

for some a1 , a2 , b1 , b2 ∈ R. The dot product is

~a ·~b =
¡ ¢ ¡ ¢
a1 Xu (q) + a2 Xv (q) · b1 Xu (q) + b2 Xv (q)
= a1 b1 Xu (q) · Xu (q) + a1 b2 Xu (q) · Xv (q)
+a2 b1 Xv (q) · Xu (q) + a2 b2 Xv (q) · Xv (q)
à !à !
Xu (q) · Xu (q) Xu (q) · Xv (q) b1
= (a1 a2 ) .
Xv (q) · Xu (q) Xv (q) · Xv (q) b2

This proves the following proposition.

5 / 110
3.1 The First Fundamental Form

Proposition 3.1.2
Let Ip (·, ·) : Tp S × Tp S → R be the first fundamental form at a point P on a
regular surface S. Given a regular parametrization X : U → R3 of a
neighborhood of p, the matrix associated with the first fundamental form
with respect to the basis {Xu , Xv } is
µ ¶
g11 g12
g= ,
g21 g22
where
g11 = Xu · Xu and g12 = Xu · Xv
g21 = Xv · Xu and g22 = Xv · Xv

If p = X (q), then the first fundamental form Ip (·, ·) at the point p can be
expressed as the bilinear form
Ip (~a,~b) =~aT g(q)~b ,
for all~a,~b ∈ Tp S.
6 / 110
3.1 The First Fundamental Form

The matrix g of the first fundamental form is symmetric since for any
differentiable function X : U → R3 ,
g12 = Xu · Xv = Xv · Xu = g21 .

Some classical differential geometry texts use the notations


E = g11 , F = g12 = g21 , G = g22 ,
and call g the metric tensor.
Example 3.1.3. [The xy-Plane] The xy-plane is a regular surface that has a
natural pamaretrization
X (u, v) = (u, v, 0), (u, v) ∈ R2 .
Since Xu (u, v) = (1, 0, 0) and Xv (u, v) = (0, 1, 0), the matrix associated to the
first fundamental form at any point on the plane is
à ! à !
Xu · Xu Xu · Xv 1 0
g= = .
Xv · Xu Xv · Xv 0 1
7 / 110
3.1 The First Fundamental Form

Example 3.1.4. [Circular Cylinder] Consider the circular cylinder given by

X (u, v) = (cos u, sin u, v) , with (u, v) ∈ (0, 2π) × R.

The partial derivatives of X are

Xu (u, v) = (− sin u, cos u, 0) and Xv (u, v) = (0, 0, 1) ,

and thus, the matrix of the first fundamental form is


à ! à !
Xu · Xu Xu · Xv 1 0
g= = .
Xv · Xu Xv · Xv 0 1

Observation: We see that the first fundamental forms of the xy-plane and
the unit circular cylinder are equal although they are not the same surfaces.
We shall see later (in the next sections) that because of this, the cylinder
and the plane share several geometric properties.
8 / 110
3.1 The First Fundamental Form

Example 3.1.5. [Spheres] Consider the following parametrization on the


sphere X (u, v) = (a cos u sin v, a sin u sin v, a cos v) , with (u, v) ∈ (0, 2π) × (0, π).
The partial derivatives of X are
Xu (u, v) = (−a sin u sin v, a cos u sin v, 0)
Xv (u, v) = (a cos u cos v, a sin u cos v, −a sin v) .
For this parametrization, we have
! Ã 2
a sin2 v
à !
Xu · Xu Xu · Xv 0
g= = .
Xv · Xu Xv · Xv 0 a2

Example 3.1.6. [Torus] Let 0 < b < a. Consider the torus parametrized by
X (u, v) = ((a + b cos u) cos v, (a + b cos u) sin v, b sin u) ,
where (u, v) ∈ (0, 2π) × (0, 2π). The matrix of the first fundamental form is
à ! à 2 !
Xu · Xu Xu · Xv b 0
g= = .
Xv · Xu Xv · Xv 0 (a + b cos u)2
9 / 110
3.1 The First Fundamental Form

Example 3.1.7. [Catenoid] The first fundamental form of the catenoid


parametrized by
X (u, v) = (a cosh u cos v, a cosh u sin v, au) ,
where (u, v) ∈ R × (0, 2π), is described by the matrix
a cosh2 u
à ! à 2 !
Xu · Xu Xu · Xv 0
g= = .
Xv · Xu Xv · Xv 0 a2 cosh2 u

Example 3.1.8. [Helicoid] The first fundamental form of the helicoid


parametrized by
X (u, v) = (u cos v, u sin v, bv) ,
where (u, v) ∈ R+ × R, is given by the matrix
à ! à !
Xu · Xu Xu · Xv 1 0
g= = .
Xv · Xu Xv · Xv 0 u2 + b2
10 / 110
3.1 The First Fundamental Form

The first fundamental form Ip (·, ·) : Tp S × Tp S → R, as a bilinear form on the


tangent space, is independent of parametrization. However, the specific
functions gij of its matrix depend on the parametrization.

Theorem 3.1.9
Let U and U 0 be open subsets of R2 . Suppose S is a regular surface that is
parametrized by X : U → R3 whose first fundamental form at p = X (q) has
associated matrix g. Let F : U 0 → U be a diffeomorphism so that Y = X ◦ F is
a reparametrization of X with p = Y (q 0 ). Then the matrix g̃ of the first
fundamental form at the point p ∈ S with respect to Y is
¡ ¢T ¡ ¢
g̃ = dFq 0 g dFq 0 .

Proof. It can be shown that with respect to any parametrization X of a


regular surface, the matrix of the first fundamental form at p = X (q) satisfies
¡ ¢T ¡ ¢
g = dXq dXq .

11 / 110
3.1 The First Fundamental Form

(continuation of proof ) Thus, with respect to the reparametrization Y ,


¡ ¢T ¡ ¢
g̃ = dYq 0 dYq 0
¡ ¢T ¡ ¢
= dXq ◦ dFq 0 dXq ◦ dFq 0
¡ ¢T ¡ ¢T ¡ ¢¡ ¢
= dFq 0 dXq dXq dFq 0
¡ ¢T ¡ ¢
= dFq 0 g dFq 0 .

2
What is the geometric relevance of the first fundamental form ?
The first fundamental form allows us to do geometry on a regular surface
without having to step out of the surface. In particular, with the knowledge
of the first fundamental form of a surface at any point on the surface, we
can compute
lengths of curves on the surface,
angles between curves on the surface, and
areas of patches.
12 / 110
3.1 The First Fundamental Form

Proposition 3.1.10
Let S be a regular surface with a coordinate neighborhood parametrized by
X : U → R3 , where U is an open subset of R2 . Consider a curve on S given by
γ(t) = X (α(t)), where α(t) = (u(t), v(t)) is a differentiable parametrized plane
curve in U. The length of the arc on γ on the interval [a, b] is given by
Z bq
s= g11 (α(t))u 0 (t)2 + 2g12 (α(t))u 0 (t)v 0 (t) + g22 (α(t))v 0 (t)2 dt.
a

Example 3.1.11. Consider the helicoid parametrized by

X (u, v) = (u cos v, u sin v, 2v) ,

where (u, v) ∈ U = R+ × R. Let α(t) = t 2 , t , with t ∈ [0, 3], be a curve in the


¡ ¢

domain U so that γ(t) = X (α(t)) is an arc on the helicoid. Find the length of
the arc γ.
13 / 110
3.1 The First Fundamental Form

Solution. With b = 2 in Example 3.1.8, the first fundamental form has


components
g11 = 1, g12 = g21 = 0, g22 = u2 + 4.
Since on γ, we have u(t) = t 2 and v(t) = t, then
Z 3p
s = (2t)2 + (t 4 + 4) dt
0
Z 3p
= (t 4 + 4t 2 + 4) dt
0
Z 3
= (t 2 + 2) dt
0
µ 3
t
¶¯
¯3
= + 2t ¯
3 0
= 15.

14 / 110
3.1 The First Fundamental Form

Proposition 3.1.12
Let S be a regular surface with a coordinate neighborhood parametrized by
X : U → R3 , where U is an open subset of R2 . Consider two curves that lie on
S given by γ1 (t) = X (α1 (t)) and γ2 (t) = X (α2 (t)), where α1 (t) = (u1 (t), v1 (t))
and α2 (t) = (u2 (t), v2 (t)) are differentiable parametrized plane curves in U.
If the two curves intersect at the point p = X (q) = X (α1 (t1 )) = X (α2 (t2 )), then
angle θ at which they intersect satisfies
A
cos θ = ,
BC
where
A = g11 u01 (t1 )u02 (t2 ) + g12 u01 (t1 )v20 (t2 ) + u02 (t1 )v10 (t2 ) + g22 v10 (t1 )v20 (t2 )
¡ ¢
q
B = g11 (u01 (t1 ))2 + 2g12 u01 (t1 )v10 (t1 ) + g22 (v10 (t1 ))2
q
C = g11 (u02 (t2 ))2 + 2g12 u02 (t2 )v20 (t2 ) + g22 (v20 (t2 ))2

15 / 110
3.1 The First Fundamental Form

Example 3.1.13. Consider the hyperbolic paraboloid parametrized by


X (u, v) = (u, v, uv) .
Determine the angle between the u-coordinate line with u = 1 and the
v-coordinate line with v = 1.

Proposition 3.1.14
Let S be a regular surface parametrized by X : U → R3 . The u-coordinate
lines and v-coordinate lines are orthogonal if and only if g12 = 0 for all
(u, v) ∈ U. In such a case, we call X an orthogonal parametrization.

Corollary 3.1.15
Let f : U → R be a differentiable function for some open subset U of R2 . The
coordinate lines of the graph of f are orthogonal if and only if fu = 0 or fv = 0.

16 / 110
3.1 The First Fundamental Form

Definition 3.1.16. Surface Integral


Let X : U → S be a parametrization for a coordinate neighborhood of a
regular surface S whose first fundamental form has matrix g and f : R3 → R
be a function defined on X (U). Suppose Q is a compact subset of U and
R = X (Q) is a region on S. The expression
Ï Ï
¡ ¢p
f dA := f ◦ X (U) detg dudv ,
X (Q) Q

is well-defined and is called the surface integral of f .

If f = 1, the surface integral gives the area of the patch R. That is,
Ï p
A(R) = detg dudv .
Q
We can see that detg > 0, which means that g is always a positive
definite matrix. That is, Ip (~a,~b) > 0 for any nonzero~a,~b ∈ Tp S. 17 / 110
3.1 The First Fundamental Form

Example 3.1.17. Let a, b ∈ R+ with b < a. Find the area of the torus
X (u, v) = ((a + b cos u) cos v, (a + b cos u) sin v, b sin u) ,
where (u, v) ∈ [0, 2π] × [0, 2π].
Solution. We have seen from Example 3.1.6 that the first fundamental form
of this torus is given by the matrix
à 2 !
b 0
g= .
0 (a + b cos u)2
So, p
detg = b |a + b cos u|
= b (a + b cos u), since a + b cos u ≥ a − b > 0.
Therefore, the area of the torus is
Z 2π Z 2π
A(R) = b (a + b cos u) du dv = 4π2 ab.
0 0

18 / 110
3.2 The Gauss Map

Recall: Let S be a regular surface parametrized by X : U → R3 .


The tangent space Tp S at a point p ∈ S is a two-dimensional subspace
of R3 . Hence, its complement, which is the set of vectors that are
normal to S at p form a one-dimensional subspace of R3 .
At any point p = X (q), there exist exactly two possible choices for a unit
normal vector:
Xu (q) × Xv (q)
N(q) = and − N(q).
kXu (q) × Xv (q)k

If S is orientable, we can specify its orientation using a continuous


function n : S → S2 , where S2 refers to the unit sphere in R3 .
If the parametrization X induces a positive orientation for S then the
functions n and N are related by N = n ◦ X .

Definition 3.2.1. The Gauss Map


Let S be an oriented regular surface in R3 with an orientation n : S → S2 . The
function n is called the Gauss map.
19 / 110
3.2 The Gauss Map

Example 3.2.2. At every point on a plane, the tangent plane is the plane
itself. Thus, an orientation of a plane can be defined by a constant normal
vector field n. Hence the Gauss map sends all the points on the plane to a
fixed point on S2 .
Example 3.2.3. Consider a sphere S in R3 centered at c = (c1 , c2 , c3 ) of radius
a and equipped with an orientation n, with unit normal vectors pointing
away from the center of the sphere. Let~x = (x, y, z). Then the sphere can be
viewed as the set of solutions to the equation
k~x − ck = a2 .
If γ(t) is any curve on the sphere, then
γ 0 (t) · γ(t) − c = 0.
¡ ¢

The tangent plane to the sphere S at a point p consists of all vectors γ0 (t0 ),
where γ(t) is any curve on S with γ(t0 ) = p. Therefore, there are two options
for unit normal vectors:
p−c p−c
or − .
kp − ck kp − ck
20 / 110
3.2 The Gauss Map

However, it is the former that provides an outward-pointing orientation n.


Thus, the Gauss map for the sphere is explicitly
p−c
n(p) = .
kp − ck

Furthermore, if S is the unit sphere centered at the origin, then the Gauss
map is the identity function.

Proposition 3.2.4
Let S be a regular surface in R3 , and let X : U → R3 be a parametrization of a
coordinate patch V = X (U), where U is an open subset of R2 . Define the
vector function
Xu × Xv
N= .
kXu × Xv k
If X is of class C r , then N is of class C r−1 .

21 / 110
3.2 The Gauss Map

Let S be a regular oriented surface with orientation supplied by n, and let


X : U → R3 be a regular positively oriented parametrization of a coordinate
patch X (U) of S. Suppose further that S is of class C 2 , which implies that N
is of class C 1 . Since kNk = N · N = 1 (constant) for all (u, v) ∈ U, we have
N · Nu = 0 and N · Nv = 0,
for all (u, v) ∈ U.
The partial derivatives Nu and Nv are vector functions such that Nu (q)
and Nv (q) are in Tp S when X (q) = p.
The vector function N : U → R3 is itself a parametrized surface whose
image lies in the unit sphere, though not necessarily giving a regular
parametrization of S2 .
If N admits a tangent space at q ∈ U, then
Tp S = TN(q) (S2 ).
In other words, the set of vectors {Xu , Xv } and {Nu , Nv } span the same
subspace Tp S.
22 / 110
3.2 The Gauss Map

The differential of the Gauss map at a point p on S is a linear


transformation dnp : Tp S → Tn(p) (S2 ). But by the previous result, we
can write it as a linear transformation
dnp : Tp S → Tp S.

Definition 3.2.5. The Differential of the Gauss Map


Let S be a regular oriented surface with orientation supplied by n and let
X : U → R3 be a regular positively oriented parametrization of a coordinate
patch X (U) of S. For all~v ∈ Tp S, we define dnp (~v) = ~
w ∈ Tp S if there exists a
curve α : I → U such that

α(0) = q, with X (q) = p






d

 ¯
(X ◦ α) (t)¯ =~v
¯
 dt t=0

 d (N ◦ α) (t)¯¯ = ~

 ¯
w.

dt t=0

23 / 110
3.2 The Gauss Map

In other words, if we write X (t) = X ◦ α(t) and N(t) = N ◦ α(t), the


differential of the Gauss map satisfies dnp X 0 (0) = N 0 (0).
¡ ¢

Using coordinate lines X (t, v0 ) and X (u0 , t) for the curve α in Definition
3.2.5, it can be seen that
dnp (Xu ) = Nu and dnp (Xv ) = Nv .

The linear map −dnp is called the shape operator or Weingarten map
of S at p.
Let S be a regular oriented surface of class C 2 with orientation n, and let
X : U → R3 be a positively oriented parametrization of a neigborhood V of a
point p ∈ S. Since N satisfies
N · Xi = 0 for i = 1, 2,
then by differentiating with respect to another variable, we have
Nj · Xi + N · Xij = 0 for i, j = 1, 2.
But since Xij = Xji , we conclude that
Ni · Xj = −N · Xij = Nj · Xi . (?) 24 / 110
3.2 The Gauss Map

Proposition 3.2.6
The linear map dnp is a self-adjoint operator with respect to the first
fundamental form.

w ∈ Tp S and write~v = v1 Xu + v2 Xv and ~


Proof. Let~v,~ w = w1 Xu + w2 Xv .
Recall that dnp (Xu ) = Nu and dnp (Xv ) = Nv . Thus,
¡ ¢ ¡ ¢
Ip dnp (~v),~
w = Ip v1 Nu + v2 Nv ,~ w
= (v1 Nu + v2 Nv ) · (w1 Xu + w2 Xv )
X2
= vi wj Ni · Xj
i,j=1
2
(?) X
= vi wj Nj · Xi
i,j=1
= (v1 Xu + v2 Xv ) · (w1 Nu + w2 Nv )
Ip ~v, dnp (~
¡ ¢
= w) .
25 / 110
3.3 The Second Fundamental Form

Definition 3.3.1. The Second Fundamental Form


Let S be an oriented regular surface of class C 2 with orientation n, and let p
be a point of S. The second fundamental form is the quadratic form on Tp S
defined by
¡ ¢
IIp (~v) = −dnp (~v) ·~v = −Ip dnp (~v),~v .

The second fundamental form provides a measure of how much the normal
vector changes if one travels away from p in a particular direction~v ∈ Tp S.

Recall from linear algebra that every quadratic form Q on a vector space V
of dimension n is of the form Q(~v) =~vT M~v, for some n × n matrix M. Define
the functions Lij : U → R by
à !
T
L11 (q) L12 (q)
IIp (~v) =~v ~v,
L21 (q) L22 (q)

where p = X (q), for all~v ∈ TP S.


26 / 110
3.3 The Second Fundamental Form
µ ¶
a
By writing~v = in the coordinate basis {Xu , Xv }, we get
b

IIp (~v) = − (aN1 + bN2 ) · (aX1 + bX2 )


= − a2 N1 · X1 + ab N1 · X2 + ab N2 · X1 + b2 N2 · X2
¡ ¢
à !à !
¡ ¢ −N1 · X1 −N2 · X1 a
= a b
−N1 · X2 −N2 · X2 b

Thus, Lij = −Nj · Xi , for i, j = 1, 2. Using property (?), we deduce that

Lij = N · Xij .

We have constructed a convenient way to calculate the Lij functions and,


hence, the second fundamental form. Moreover, it is obvious that the
¡ ¢
matrix Lij is a symmetric matrix, i.e., L12 = L21 .

27 / 110
3.3 The Second Fundamental Form

Example 3.3.2. [The xy-Plane] The xy-plane is a regular orientable surface


with parametrization X (u, v) = (u, v, 0), (u, v) ∈ R2 . Since Xu (u, v) = (1, 0, 0)
and Xv (u, v) = (0, 1, 0), then N = (0, 0, 1). Moreover, Xij =~0, for all i, j = 1, 2.
Hence, the second fundamental form of the plane has matrix (Lij ) = 02×2 .
Example 3.3.3. [Circular Cylinder] Consider the circular cylinder given by
X (u, v) = (cos u, sin u, v) , with (u, v) ∈ (0, 2π) × R.
The partial derivatives of X are
Xu (u, v) = (− sin u, cos u, 0) and Xv (u, v) = (0, 0, 1) ,
from which we get N = (cos u, sin u, 0). Moreover,
Xuu = (− cos u, − sin u, 0) and Xuv = Xvv =~0.
Thus, the second fundamental form is given by the matrix
à !
¡ ¢ −1 0
Lij = .
0 0
Observation: Although the plane and the unit circular cylinder have the
same first fundamental forms, their second fundamental forms are not
equal. 28 / 110
3.3 The Second Fundamental Form

Example 3.3.4. [Spheres] Let S be the sphere of radius a with outward


orientation and consider a coordinate patch parametrized by
X (u, v) = (a cos u sin v, a sin u sin v, a cos v) .
The unit normal vector is
N = (cos u sin v, sin u sin v, cos v) ,
while the second order partial derivatives are
Xuu = (−a cos u sin v, −a sin u sin v, 0)
Xuv = Xvu = (−a sin u cos v, a cos u cos v, 0)
Xvv = (−a cos u sin v, −a sin u sin v, −a cos v) .

Thus. the matrix for the second fundamental form is

−a sin2 v 0
à !
¡ ¢
Lij = .
0 −a

29 / 110
3.3 The Second Fundamental Form

Example 3.3.5. [Catenoid] Let S be the catenoid parametrized by


X (u, v) = (a cosh u cos v, a cosh u sin v, au) .
The unit normal vector is
N = (− cos v sech u, − sin v sech u, tanh u) ,
while the second order partial derivatives are
Xuu = (a cosh u cos v, a cosh u sin v, 0)
Xuv = Xvu = (−a sinh u sin v, a sinh u cos v, 0)
Xvv = (−a cosh u cos v, −a cosh u sin v, 0) .

Thus, the matrix for the second fundamental form is


à !
¡ ¢ −a 0
Lij = .
0 a

30 / 110
3.3 The Second Fundamental Form

Example 3.3.6. [Helicoid] Let S be the helicoid parametrized by


X (u, v) = (u cos v, u sin v, bv) .
The unit normal vector is
1
N=p (b sin v, −b cos v, u) ,
b2 + u2
while the second order partial derivatives are
Xuu = ~0
Xuv = Xvu = (− sin v, cos v, 0)
Xvv = (−u cos v, −u sin v, 0) .
Thus, the matrix for the second fundamental form is
 
b
0 −p
¡ ¢  b2 + u2 

Lij =  .

 b 
−p 0
b2 + u2
31 / 110
3.3 The Second Fundamental Form

Recall: If S is a regular surface, the tangent plane to S at p approximates


points on S near p. This is called the first-order approximation of S near p.
Definition 3.3.7. Osculating Paraboloid
At a point p ∈ S in a coordinate neighborhood parametrized positively by
X (u, v), if X (q) = p, the function

f (s, t) = L11 (q)s2 + 2L12 (q)st + L22 (q)t 2
¢
2
is called the osculating paraboloid of S at p.

This provides a second-order approximation of the surface near p in


reference to the normal vector N(q) in the following sense:
Let q = (u0 , v0 ). The second order Taylor approximation of X at q is
X (u, v) ≈ X (q) + Xu (q)(u − u0 ) + Xv (q)(v − v0 )

+ Xuu (q)(u − u0 )2 + Xuv (q)(u − u0 )(v − v0 )
2
+ Xvv (q)(v − v0 )2 .
¤
32 / 110
3.3 The Second Fundamental Form

This means

X (u, v) − X (q) ≈ Xu (q)(u − u0 ) + Xv (q)(v − v0 )



+ Xuu (q)(u − u0 )2 + Xuv (q)(u − u0 )(v − v0 )
2
+ Xvv (q)(v − v0 )2 .
¤

Since N is orthogonal both to Xu and Xv , the dot product of each side with
N(q) gives

X (u, v) − X (q) · N(q) ≈ N(q) · Xuu (q)(u − u0 )2 + N(q) · Xuv (q)(u − u0 )(v − v0 )
¡ ¢
2
+ N(q) · Xvv (q)(v − v0 )2 .
¤

Setting s = u − u0 and t = v − v0 , we get

(X (s + u0 , t + v0 ) − X (u0 , v0 )) · N(u0 , v0 ) ≈ f (s, t).

33 / 110
3.3 The Second Fundamental Form

Furthermore, we can see that the second derivatives of f are


f11 = L11 , f12 = f21 = L12 , f22 = L22 .
Applying the second derivatives test to the function f at (s, t) = (0, 0),
2
f attains a local extremum if L11 L22 − L12 >0
2
f has a saddle point if L11 L22 − L12 < 0.

This leads to the following definition.

Definition 3.3.8
Let S be a regular orientable surface of class C 2 . A point p ∈ S is called
¡ ¢
elliptic if det Lij > 0;
¡ ¢
hyperbolic if det Lij < 0;
¡ ¢
parabolic if det Lij = 0 but not all Lij = 0; and,
planar if Lij = 0, for all i, j.

34 / 110
3.3 The Second Fundamental Form

Remark 3.3.9.
1 Every point on a plane is planar.
2 Every point on a circular cylinder is parabolic.
3 Every point on a sphere is elliptic.
4 Every point on a catenoid is hyperbolic.
5 Every point on a helicoid is hyperbolic.
35 / 110
3.3 The Second Fundamental Form

Example 3.3.10. [Torus] Consider the torus parametrized by


X (u, v) = ((a + b cos u) cos v, (a + b cos u) sin v, b sin u) .
The partial derivatives are
X1 = (−b sin u cos v, −b sin u sin v, b cos u)
X2 = (−(a + b cos u) sin v, (a + b cos u) cos v, 0) .
The cross product of these vectors and its norm are
X1 × X2 = (a + b cos u) (−b cos u cos v, −b cos u sin v, −b sin u)
kX1 × X2 k = b (a + b cos u) .
The unit normal vector and the second order partial derivatives of X are
N = (− cos u cos v, − cos u sin v, − sin u)
X11 = (−b cos u cos v, −b cos u sin v, −b sin u)
X12 = (b sin u sin v, −b sin u cos v, 0)
X22 = (−(a + b cos u) cos v, −(a + b cos u) sin v, 0) .
36 / 110
3.3 The Second Fundamental Form

With Lij = N · Xij , we have

L11 = b, L12 = 0, and L22 = (a + b cos u) cos u.

We can see that


2
¡ ¢
det Lij = L11 L22 − L12 = b (a + b cos u) cos u.

Since 0 < b < a, then 0 < a − b ≤ a + b cos u ≤ a + b.


¡ ¢
Thus, the sign of det Lij is the same as the sign of cos u. Hence,
det Lij > 0 if − π2 < u < π2 ,
¡ ¢

det Lij = 0 if u = − π2 or u = π2 , and


¡ ¢

π 3π
¡ ¢
det Lij < 0 if 2 <u< 2 .

37 / 110
3.3 The Second Fundamental Form

Therefore,
the elliptic points on the torus are points where − π2 < u < π2 ,
the parabolic points are where u = − π2 or u = π2 , and
π 3π
the hyperbolic points are where 2 <u< 2 .

38 / 110
3.3 The Second Fundamental Form

Remark 3.3.11. The only quadratic surface possessing at least one planar
point is a plane. However, on a general surface S, existence of a planar point
does not imply that S is a plane. Existence of a planar point merely implies
that the third-order behavior, as opposed to the second-order, governs the
local geometry of the surface with respect to the normal vector.
Example 3.3.12. As an illustration, consider the monkey saddle which is
parametrized by
X (u, v) = u, v, u3 − 3uv2 .
¡ ¢

The second-order partial derivatives of X are


Xuu = (0, 0, 6u) , Xuv = (0, 0, −6v) , and Xvv = (0, 0, −6u) .
We can see that even without calculating the unit normal vector function,
we deduce that Lij (0, 0) = 0, for all i, j. Consequently, the origin (0, 0, 0) is a
planar point. This means that the best approximating quadratic to the
surface is a plane, though, obviously such an approximation describes the
surface poorly.
39 / 110
3.3 The Second Fundamental Form

40 / 110
3.3 The Second Fundamental Form

41 / 110
3.3 The Second Fundamental Form

Proposition 3.3.13
Let S be a regular surface of class C 2 .
1 If p ∈ S is an elliptic point, then there exists a neighborhood V of p
such that Tp S does not intersect V − {p}.
2 If p ∈ S is a hyperbolic point, then Tp S intersect every neighborhood
V − {p} of p.

42 / 110
3.3 The Second Fundamental Form

So far, we have seen that the first fundamental form (metric tensor) is given
by the matrix à !
Xu · Xu Xu · Xv
g=
Xv · Xu Xv · Xv
while the second fundamentalÃform has matrix!
N · Xuu N · Xuv
L= .
N · Xvu N · Xvv
It is now possible to calculate explicitly the matrix representation of the
differential of the Gauss map , dnp : Tp S → Tp S, in terms of the basis {Xu , Xv }.
Since Nu and Nv are in Tp S, there exist scalar functions aji (u, v) such that
Nu = a11 Xu + a12 Xv
Nv = a21 Xu + a22 Xv .
In terms of numerical indices, we have
2
aji Xi .
X
Nj =
i=1 43 / 110
3.3 The Second Fundamental Form

For any vector ~


w = w1 X1 + w2 X2 in Tp S, the differential of the Gauss map
satisfies

dnp (~
w) = w1 N1 + w2 N2
= a11 w1 + a21 w2 X1 + a12 w1 + a22 w2 X2
¡ ¢ ¡ ¢

à 1
a1 a21 w1
!Ã !
= .
a12 a22 w2

However, using (?),

−Lij = Ni · Xj
à !
2
aik Xk · Xj
X
=
k=1
2
aik gkj .
X
=
k=1

44 / 110
3.3 The Second Fundamental Form

In matrix notation, this means


! Ã 1
a21
à !à !
L11 L12 a1 g11 g12
− = 2 .
L21 L22 a1 a22 g21 g22

Since g is positive definite at any point on a regular surface, it is invertible


and so multiplying both sides by the inverse of g, we get
!−1
a11 a21
à ! à !Ã
L11 L12 g11 g12
=− .
a12 a22 L21 L22 g21 g22

Recall that g and L are symmetric matrices. Moreover, dnp is a self-adjoint


operator which means that its matrix is also symmetric. Therefore,
!−1 Ã
a11 a21
à ! à !
g11 g12 L11 L12
=− .
a12 a22 g21 g22 L21 L22

45 / 110
3.3 The Second Fundamental Form

Remark 3.3.14. Let S be an oriented regular surface of class C 2 oriented by


n. Suppose that at a point p on S, the first and second fundamental forms
are given by the matrices g and L , respectively. Then, the differential dnp is
given by the matrix
¢ ³ ´
dnp = aji = −g −1 L.
¡
¡ ¢
Note that detg > 0 and det L = det (−L) since L is 2 × 2. Hence, det dnp has
the same sign as det L. Thus, Definition 3.3.8 can be restated as follows.
Proposition 3.3.15
Let S be an oriented regular surface of class C 2 with orientation n. Then, a
point p on S is
¡ ¢
elliptic if det dnp > 0;
¡ ¢
hyperbolic if det dnp < 0;
¡ ¢
parabolic if det dnp = 0 but not all dnp 6= 0; and,
planar if dnp = 0, for all i, j.
46 / 110
3.4 Normal and Principal Curvatures

We have seen from the previous section that one way to analyze the shape
of an orientable regular surface S near a point p is to consider the best
quadratic surface approximation of S near p and classify the point p as
either elliptic, hyperbolic, parabolic, or planar.
In this section, we analyze the shape of S near a point p by considering
curves on S through p and taking the normal component of their principal
normal vector.
Definition 3.4.1. Normal Curvature
Let S be a regular surface of class C 2 , and let γ : I → R3 be a parametrization
of class C 2 for a curve that lies on S. The normal curvature of S along C is
the function
1
κn (t) = 0 T 0 · N = κ (P · N) = κ cos θ,
s
where T and P are, respectively, the unit tangent and principal normal
vector to γ ; N is the unit normal vector of S; and θ is the angle between P
and N.
47 / 110
3.4 Normal and Principal Curvatures

Though the curvature of a space curve γ, in general, depends on the second


derivative of γ, once one has the second fundamental form for a surface,
the normal curvature at a point depends only on the direction.
Theorem 3.4.2
Let p be a point on a regular surface S of class C 2 and suppose that γ is a
parametrization of a curve on S such that γ(0) = p and T (0) = ~w ∈ Tp S. Then
at p, we have
κn (0) = IIp (~
w).

Proof. Let X : U → R3 be a parametrization of a coordinate neighborhood V


of p. Suppose that γ = X ◦ α, where α(t) = (u(t), v(t)) is a curve in U. The
normal vector of S along the curve is given by the function N(t) = N (α(t)).
Since T (t) · N(t) = 0, for all t, then
T 0 · N = −T · N 0 ,
and hence, 1
κn (t) = − T (t) · N 0 (t).
s 0 (t) 48 / 110
3.4 Normal and Principal Curvatures

Since N(t) = N (u(t), v(t)), then N 0 (t) = Nu (α(t)) u 0 (t) + Nv (α(t)) v 0 (t).
In particular,
N 0 (0) = Nu (α(0)) u 0 (0) + Nv (α(0)) v 0 (0)
= dnp Xu (α(0)) u 0 (0) + Xv (α(0)) v 0 (0)
¡ ¢

= dnp (X ◦ α)0 (0)


¡ ¢

= dnp γ 0 (0)
¡ ¢

= dnp s 0 (0)T (0) .


¡ ¢

Therefore, at the point p, the normal curvature of S along C is


1
κn (0) = − 0 T (0) · N 0 (0)
s (0)
1
= − 0 T (0) · dnp s 0 (0)T (0)
¡ ¢
s (0)
w · dnp (~
= −~ w)
= IIp (~
w).
2
49 / 110
3.4 Normal and Principal Curvatures

Corollary 3.4.3
The normal curvature of S at a point p is the same along all curves on S
through p with unit tangent vector ~
w ∈ Tp S.

Example 3.4.4. Let p be the origin. Consider the surface parametrized by


X (u, v) = u, v, u2 + 2v2 .
¡ ¢
1 Determine the maximum and minimum normal curvatures at p.
¡ ¢
2 Find the eigenvalues of dnp .
3 Compare the values obtained in (1) and (2) and make a conjecture.
Solution.
1 Clearly, at the origin, (u, v) = (0, 0). The first and second-order partial
derivatives of X at p are
Xu (u, v) = (1, 0, 2u) ⇒ Xu (0, 0) = (1, 0, 0)
Xv (u, v) = (0, 1, 4v) ⇒ Xv (0, 0) = (0, 1, 0)
Xuu = (0, 0, 2) , Xvv = (0, 0, 4) , Xuv = (0, 0, 0).
50 / 110
3.4 Normal and Principal Curvatures

Moreover, the unit normal vector to S at p is N = (0, 0, 1). Thus, the matrix of
the second fundamental form at p is
à !
2 0
L= .
0 4
Let ~
w = aXu (0, 0) + bXv (0, 0) ∈ Tp S, with k~wk = 1. We can see that with the
computed partial derivatives at (0, 0), a + b2 = 1. Then by Theorem 3.4.2,
2

the normal curvature of S at p along curves with unit tangent vector ~ w is

κn = IIp (~
w)
à !à !
2 0 a
= (a b)
0 4 b

= 2a2 + 4b2 .

We want to optimize κn = 2a2 + 4b2 subject to the constraint a2 + b2 = 1.

51 / 110
3.4 Normal and Principal Curvatures

By the method of Lagrange multiplier, these optimal values may occur only
at solutions of the system 
4a = 2λa


8b = 2λb

a2 + b2 = 1,

for some real number λ. It can be shown that the system has four solutions:
à ! à ! à ! à !
0 0 1 −1
, , , and .
1 −1 0 0
We now conclude that, at the origin,
the maximum
à ! normal
à ! curvature is 4 which occurs along the tangent
0 0
vectors and ; and
1 −1
the minimum à ! curvature is 2 which occurs along the tangent
à ! normal
1 −1
vectors and .
0 0 52 / 110
3.4 Normal and Principal Curvatures

2 We have already found in (1) the matrix L for the second fundamental
form at the origin. With the first derivatives of X at the origin, we can
see that the metric tensor is given by g = Id2×2 . Hence,
à !
¡ ¢ −1
−2 0
dnp = −g L = −L = ,
0 −4

whose eigenvalues are −2 and −4 (since the matrix is diagonal).


3 Conjecture: The maximum and minimum normal curvatures of S at a
¡ ¢
point p are the negatives of the eigenvalues of dnp .

Proposition 3.4.5
Let S be a regular surface and let p be a point on S. The maximum and
minimum normal curvatures κ1 and κ2 of S at p are the negatives of the
¡ ¢
eigenvalues of dnp . Furthermore, there exists an orthonormal basis
{e1 , e2 } of Tp S such that dnp (e1 ) = −κ1 e1 and dnp (e2 ) = −κ2 e2 .
53 / 110
3.4 Normal and Principal Curvatures

Proof. Set ~
w = aXu + bXv ∈ Tp S with k~
wk = 1. We optimize

w) = L11 a2 + 2L12 ab + L22 b2 ,


IIp (~

which is a function of a and b, subject to the costraint

Ip (~ w) = g11 a2 + 2g12 ab + g22 b2 = 1.


w,~

Using the method of Lagrange multiplers, there exists λ ∈ R such that the
optimization problem is solved when
(
L11 a + L12 b = λ g11 a + g12 b
¡ ¢
¢ , (??)
L21 a + L22 b = λ g21 a + g22 b
¡

or in matrix form
à !à ! à !à !
L11 L12 a g11 g12 a
=λ .
L21 L22 b g21 g22 b

54 / 110
3.4 Normal and Principal Curvatures
à 0! à !à ! à 0! à 0!
a g11 g12 a a a
Setting = , we have Lg −1
=λ 0 .
b0 g21 g22 b b0 b

Recall that dnp = −Lg −1 . Thus, −λ is an eigenvalue of dnp .


¡ ¢ ¡ ¢

Therefore, the optimal values of the normal curvature are


à ! à 0!
a −1
a
IIp (~
w) = (a b) L = (a b) Lg
b b0
à 0! à !
a a
= (a b) λ 0 = λ (a b) g
b b

= λIp ~ w = λ.
¡ ¢
w,~

The second statement follows from the Spectral Theorem.


2

55 / 110
3.4 Normal and Principal Curvatures

Definition 3.4.6. Principal Curvatures


Let S be a regular surface and let p be a point on S. The maximum and
minimum normal curvatures κ1 and κ2 of S at p are called the principal
curvatures of S at p. The corresponding directions, i.e., eigenvectors e1 and
e2 with dnp (ei ) = −κi ei , are called principal directions at p.

Remark 3.4.7
1 If the principal curvatures are not equal, the principal directions are
perpendicular.
2 The principal curvatures depend on the orientation of the surface.
Thus, if under a parametrization X with orientation n the principal
curvatures are κ1 and κ2 , then under a reparametrization X̃ with
orientation −n, the principal curvatures are −κ1 and −κ2 .
3 All normal curvatures, and hence the principal curvatures, to a plane
are zero.
56 / 110
3.4 Normal and Principal Curvatures

4 At every point on a sphere, the normal curvature in every direction is


the same. Indeed, a sphere of radius a has first and second
fundamental forms (See Examples 3.1.5 and 3.3.4)
à 2
a sin2 v 0 −a sin2 v 0
! Ã !
g= and L = .
0 a2 0 −a
The matrix of the Gauss map is then
Ã1 !
a 0
dnp = −g −1 L =
¡ ¢
1
,
0 a
from which we immediately deduce that the principal curvatures are
1
κ1 (u, v) = κ2 (u, v) = − .
a

5 Let S be a regular oriented surface. If at every point on S, the principal


curvatures are equal, then S is contained in either a plane or a sphere.
57 / 110
3.4 Normal and Principal Curvatures
¡ ¢ ¡ ¢
Note that the product of the eigenvalues of dnp is equal to det dnp .
Thus, Definition 3.3.8 can be reformulated using principal curvatures.

Proposition 3.4.8
Let S be an oriented regular surface of class C 2 . Then, a point p on S is
elliptic if κ1 and κ2 have the same sign;
hyperbolic if κ1 and κ2 have opposite signs;
parabolic if exactly one of κ1 and κ2 is zero; and,
planar if κ1 = κ2 = 0.

Definition 3.4.9. Curvature Line


Le S be an oriented regular surface such that a coordinate neighborhood is
parametrized by X : U → R3 . A curve on S given by γ(t) = X (u(t), v(t)) such
that at γ(t0 ), its tangent vector γ 0 (t0 ) is a principal direction of S, is called a
curvature line of S.
58 / 110
3.4 Normal and Principal Curvatures

Remark 3.4.10.
1 At every point on a curvature line, the following equation holds:
¢ ¡ ¢2 ¡ ¢ ¡ ¢2
L11 g21 − L21 g11 u 0 + L11 g22 − L22 g11 u 0 v 0 + L12 g22 − L22 g12 v 0 = 0.
¡ ¢ ¡

2 The curvature lines form an orthogonal family of curves on the surface.

However, a regular oriented surface S can be covered by curvature lines


only at points where the first and second fundamental forms are not scalar
multiples of each other.
59 / 110
3.4 Normal and Principal Curvatures

Definition 3.4.11. Umbilical Point


Let S be a regular oriented surface of class C 2 . A point p on S is called an
umbilical point if, given a parametrization X of a neighborhood of p, the
corresponding matrices of the first and second fundamental forms are
scalar multiples of each other.

Proposition 3.4.12
Let S be a regular oriented surface. A point p on S is an umbilical point if
and only if the principal curvatures at p are equal.

Proof. We first recall that Ip (·, ·) is positive definite and dnp is self-adjoint
¡ ¢
with respect to Ip . By Spectral Theorem, the matrix dnp is diagonalizable.
¡ ¢
At a point p on S, the principal curvatures are equal if and only if dnp has
equal eigenvalues λ = −κ1 = −κ2 . Equivalently, there exists an invertible
matrix B such that
λ 0 −1
à !
¡ ¢
dnp = B B ,
0 λ 60 / 110
3.4 Normal and Principal Curvatures

and so,
−g −1 L = dnp = B (λI) B−1 = λBB−1 = λI.
¡ ¢

That is, L = −λg.


2
Remark 3.4.13. Every point on a surface S is an umbilical point if and only
if S is contained in either a plane or a sphere.

Recall: In general, the u-coordinate lines and v-coordinate lines may not be
orthogonal. The u-coordinate lines and v-coordinate lines are orthogonal if
and only if g12 = 0 (see Proposition 3.1.14).
In relation to this, the following proposition provides a necessary and
sufficient condition for coordinate lines to be curvature lines.
Proposition 3.4.14
The coordinate lines of a parametrization X of a surface are curvature lines
if and only if g12 = L12 = 0.
61 / 110
3.4 Normal and Principal Curvatures

Proof. Suppose that coordinate lines are curvature lines. Since any two
curvature lines intersect orthogonally, then g12 = 0.
The coordinate lines are given by γ1 (t) = X (u0 , t) or γ2 (t) = X (t, v0 ).
On these coordinate lines, either
u 0 (t) = 0 and v 0 (t) = 1 or u 0 (t) = 1 and v 0 (t) = 0.
By item (1) of Remark 3.4.10, the following must hold:
L11 g21 − L21 g11 = 0 and L12 g22 − L22 g12 = 0.
Since g12 = g21 = 0 then L21 g11 = 0 and L12 g22 = 0. However, since
2
det g = g11 g22 − g12 > 0,
the functions g11 and g22 cannot be both 0. Hence, L12 = L21 = 0.
Conversely, if g12 = L12 = 0, then both matrices g and L are diagonal.
Thus, the matrix (dnp ) is also diagonal, and hence, the standard basis
vectors Xu and Xv of Tp S are eigenvectors of dnp . That is, Xu and Xv are
principal directions. But Xu (u0 , v0 ) = γ2 0 (u0 ) and Xv (u0 , v0 ) = γ1 0 (v0 ), for
any (u0 , v0 ). Hence, coordinate lines are curvarture lines.
2
62 / 110
3.5 Gaussian and Mean Curvatures

We now define two fundamental geometric invariants that encapsulate a


considerable amount of useful information about the local shape of a
surface. In this section, we assume that S is a regular surface of class C 2 .
Defiiniton 3.5.1. Gaussian and Mean Curvatures
Let κ1 and κ2 be the principal curvatures of S at a point p.
1 The Gaussian curvature K of S at p is the product of the principal
curvatures, i.e., K = κ1 κ2 .
2 The mean curvature of S at p is the average of the principal curvatures,
i.e., H = 12 (κ1 + κ2 ).
¡ ¢
Since the dnp is diagonalizable, then
1 1 ¡
K = κ1 κ2 = det dnp and H = (κ1 + κ2 ) = − Tr dnp .
¡ ¢ ¢
2 2

Tha Gaussian and mean curvatures are geometric invariants. That is,
they do not depend on the position of S in space and on any particular
63 / 110
3.5 Gaussian and Mean Curvatures

Remark 3.5.2. The Gaussian curvature may also be computed as


2
det L L11 L22 − L12
K= = 2
.
det g g11 g22 − g12
This follows from the fact that dnp = −g −1 L, and det (AB) = det(A) det(B)
¡ ¢

and det A−1 = (det A)−1 , for any square matrices A and B.

Example 3.5.3. [Planes] Since the principal curvatures at any point on a


plane are zero, the Gaussian and mean curvatures at any point are zero.

Example 3.4.4. [Circular Cylinders] Consider a circular cylinder of radius


a given by
X (u, v) = (a cos u, a sin u, v) , with (u, v) ∈ [0, 2π] × R.
It can be verified that the first and second fundamental forms are given by
the matrices à 2 ! à !
a 0 −a 0
g= and L = .
0 1 0 0
64 / 110
3.5 Gaussian and Mean Curvatures

We can easily deduce that


det L
K= = 0.
det g
Moreover, Ã1 !
a 0
dnp = −g −1 L =
¡ ¢
.
0 0
Therefore,
1 ¡ ¢ 1
H = − Tr dnp = − .
2 2a

Example 3.5.5. [Spheres] We have seen from item (4) of Remark 3.4.7 that
the principal curvatures of the sphere
X (u, v) = (a cos u sin v, a sin u sin v, a cos v) , (u, v) ∈ [0, 2π] × [0, π],
are κ1 (u, v) = κ2 (u, v) = − a1 . Hence,
1 1
K= and H = − .
a2 a 65 / 110
3.5 Gaussian and Mean Curvatures

Example 3.5.6. [Catenoid] Consider the catenoid parametrized by

X (u, v) = (a cosh u cos v, a cosh u sin v, au) .

The first and second fundamental forms (See Examples 3.1.7 and 3.3.5) are
given by
a cosh2 u
à 2 ! à !
0 −a 0
g= and L = .
0 a2 cosh2 u 0 a
Therefore, the Gaussian curvature is
det L 1
K (u, v) = =− .
det g a cosh4 u
2

Furthermore,  1 
0
a cosh2 u
dnp = −g −1 L = 
¡ ¢
.
1
0 −
a cosh2 u

Hence, H = − 21 Tr dnp = 0.
¡ ¢

66 / 110
3.5 Gaussian and Mean Curvatures

Example 3.5.7. [Helicoid] Consider the catenoid parametrized by


X (u, v) = (u cos v, u sin v, bv) .
The first and second fundamental forms (See Examples 3.1.8 and 3.3.6) are
given by à ! à !
1 0 b 0 1
g= and L = − p .
0 u2 + b2 u2 + b2 1 0
Therefore, the Gaussian curvature is
det L b2
K (u, v) = = −¡ ¢2 .
det g u2 + b2
Furthermore,
 
0 1
b
dnp = −g −1 L = p
¡ ¢
.
 
2 2
 1
u +b ¡ 2 ¢ 0
u + b2
Hence, H = − 21 Tr dnp = 0.
¡ ¢
67 / 110
3.5 Gaussian and Mean Curvatures

Example 3.5.8. [Torus] Let a, b ∈ R+ with b < a. The torus


X (u, v) = ((a + b cos u) cos v, (a + b cos u) sin v, b sin u) ,
where (u, v) ∈ [0, 2π] × [0, 2π], has first and second fundamental forms (see
Examples 3.1.6 and 3.3.10) given by
à 2 ! à !
b 0 b 0
g= and L = .
0 (a + b cos u)2 0 (a + b cos u) cos u
Therefore, the Gaussian curvature is
det L cos u
K (u, v) = = .
det g b(a + b cos u)
Furthermore,  1
0

−b
dnp = −g −1 L = 
¡ ¢
cos u  .
0 −
a + b cos u
1 1 1 cos u
¡ ¢ ¡ ¢
Hence, H(u, v) = − 2 Tr dnp = 2 b + a+b cos u .
68 / 110
3.5 Gaussian and Mean Curvatures

Using Gaussian and mean curvatures, Definition 3.3.8 can be rerformulated


as follows.
Proposition 3.5.9
Let S be an oriented regular surface of class C 2 . Then, a point p on S is
elliptic if K > 0;
hyperbolic if K < 0;
parabolic if K = 0 but H 6= 0; and,
planar if K = H = 0.

Remark 3.5.10. The principal curvatures at a point can be recovered given


the Gaussian and mean curvatures at the point. That is,
p p
κ1 = H + H 2 − K and κ2 = H − H2 − K .

69 / 110
3.5 Gaussian and Mean Curvatures

The Gaussian curvature can be characterized using the basis vectors Xu and
Xv of Tp S and their images Nu and Nv under the Gauss map.

Proposition 3.5.11
Let S be a regular surface of class C 2 , and let V be a neighborhood on S
Xu × Xv
parametrized by X : R2 ⊃ U → R3 . Define the unit vector as N = .
kXu × Xv k
Then for all (u, v) ∈ U, the Gaussian curvature is the unique function K (u, v)
satisfying
Nu × Nv = K (u, v) (Xu × Xv ) .

Proof. By writing Nu = a11 Xu + a12 Xv and Nv = a21 Xu + a22 Xv , we have


Nu × Nv = a11 Xu + a12 Xv × a21 Xu + a22 Xv
¡ ¢ ¡ ¢

= a11 a22 − a21 a12 (Xu × Xv )


¡ ¢
¡ ¢
= det dnp (Xu × Xv )
= K (u, v) (Xu × Xv ) .
2
70 / 110
3.5 Gaussian and Mean Curvatures

With this interpretation, we can also show that the Gaussian curvature can
be viewed as follows.
Theorem 3.5.12
Let S be a regular oriented surface of class C 2 with Gauss map n : S → S2 ,
and let p be a point on S. Denote by K (p) the curvature of S at p and
suppose that K (p) 6= 0. Let B² be the ball of radius ² around p and define
V² = B² ∩ S. Then
Area (n(V² ))
|K (p)| = lim .
²→0 Area (V² )

71 / 110
3.5 Gaussian and Mean Curvatures

Proof. Let X : R2 ⊃ U → R3 be a regular parametrization of a neighborhood


of p = X (u0 , v0 ). Then V² ⊂ X (U) for all ² small enough. Since we assumed
that K (p) 6= 0, by the continuity of the Gaussian curvature function, we
deduce that K does not change sign for ² chosen small enough.
Define U² = X −1 (V² ). Since X is bijective, (u0 , v0 ) is the unique pre-image of
p in all U² . By the surface are formula,
Ï Ï
Area(V² ) = dA = kXu × Xv k du dv.
V² U²

Similarly, on the unit sphere,


Ï
Area(n(V² )) = kNu × Nv k du dv

Ï
= |K (u, v)| kXu × Xv k du dv.

72 / 110
3.5 Gaussian and Mean Curvatures

By the Mean Value Theorem for double integrals, for every ² > 0, there exist
points (u² , v² ) and (ũ² , ṽ² ) in U² such that
Ï
kXu × Xv k du dv = kXu (u² , v² ) × Xv (u² , v² )k Area(U² ) and

Ï
|K (u, v)| kXu × Xv k du dv = |K (ũ² , ṽ² )| kXu (ũ² , ṽ² ) × Xv (ũ² , ṽ² )k Area(U² ).

Since lim(u² , v² ) = lim(ũ² , ṽ² ) = (u0 , v0 ), we have


²→0 ²→0

Area (n(V² )) |K (ũ² , ṽ² )| kXu (ũ² , ṽ² ) × Xv (ũ² , ṽ² )k Area(U² )
lim = lim
²→0 Area (V² ) ²→0 kXu (u² , v² ) × Xv (u² , v² )k Area(U² )
|K (u0 , v0 )| kXu (u0 , v0 ) × Xv (u0 , v0 )k
=
kXu (u0 , v0 ) × Xv (u0 , v0 )k
= |K (p)|.
2
73 / 110
3.6 Surfaces of Revolution

Let us now consider a surface generated by revolving a curve about a line.


Definition 3.6.1. Surface of Revolution
A surface of revolution is obtained by revolving a regular plane curve, called
profile curve or meridian curve, about a line that is contained on the plane
of the curve. In particular, the surface generated by revolving a regular
profile curve γ : u 7→ f (u), g(u) on the xz-plane about the z-axis can be
¡ ¢

parametrized by
¡ ¢
X (u, v) = f (u) cos v, f (u) sin v, g(u) ,

where (u, v) ∈ I × [0, 2π).

In forming a surface of revolution, one can take f (u) ≥ 0, for all u ∈ I.


The partial derivatives of X are
Xu (u, v) = f 0 (u) cos v, f 0 (u) sin v, g 0 (u)
¡ ¢
¡ ¢
and Xv (u, v) = −f (u) sin v, f (u) cos v, 0 .
74 / 110
3.6 Surfaces of Revolution

At points where S is regular, f (u) 6= 0 and the unit normal vector is

1 ¡ 0 0 0
¢
N = q¡ ¢2 ¡ ¢2 −g (u) cos v, −g (u) sin v, f (u) .
f 0 (u) + g 0 (u)

Moreover, the second derivatives of X are

Xuu (u, v) = f 00 (u) cos v, f 00 (u) sin v, g 00 (u)


¡ ¢
¡ ¢
Xvv (u, v) = −f (u) cos v, −f (u) sin v, 0
Xuv (u, v) = −f 0 (u) sin v, f 0 (u) cos v, 0 .
¡ ¢

Therefore, the first and second fundamental forms are given by the matrices
q¡ ¢2 ¡ ¢2 
f 0 (u) + g 0 (u) 0
g= 
¡ ¢2

0 f (u)

75 / 110
3.6 Surfaces of Revolution

f 0 (u)g 00 (u) − f 00 (u)g 0 (u)


 
1 0
L = q¡ ¢2 ¡ ¢2
 .
0 0
f (u) + g (u) 0 f (u)g 0 (u)

In particular, suppose γ is arclength parametrized; that is,


¡ 0 ¢2 ¡ 0 ¢2
f (u) + g (u) = 1.
Then
f 0 (u)g 00 (u) − f 00 (u)g 0 (u)
à ! à !
1 0 0
g= ¢2 and L = .
f (u)g 0 (u)
¡
0 f (u) 0
Furthermore,
f 00 (u)g 0 (u) − f 0 (u)g 00 (u)
 
0
¡ ¢ 
dnp =  g 0 (u)  .

0 −
f (u)
76 / 110
3.6 Surfaces of Revolution
¢2 ¡ ¢2
Since f 0 (u) + g 0 (u) = 1, then f 0 (u)f 00 (u) = −g 0 (u)g 00 (u).
¡

We obtain the following.


principal curvatures
g 00 (u) f 00 (u)
κ1 = f 0 (u)g 00 (u) − f 00 (u)g 0 (u) = = −
f 0 (u) g 0 (u)
g 0 (u)
κ2 =
f (u)

Gaussian curvature
f 00 (u)
K = κ1 κ2 = −
f (u)

mean curvature
¢0
f (u)g 0 (u)
µ 00 ¶ ¡
1 f (u) g 0 (u)
H= − 0 + = ¡ ¢0
2 g (u) f (u) f (u)2
77 / 110
3.6 SOR with Constant Gaussian Curvature

Remark 3.6.2. For a surface of revolution, every condition on K and H


leads to an ordinary differential equation in f . In particular, with c ∈ R,
K = c ⇐⇒ f 00 + cf = 0
H = c ⇐⇒ (fg 0 )0 = c(f 2 )0 .

Let us identify surfaces of rotation with constant Gaussian curvature. We


look for a parametrization (f (u), g(u)) of the profile curve, where u is the
arclength parameter so that (f 0 )2 + (g 0 )2 = 1 satisfying the differential
equation
f 00 + Kf = 0.
The general solution is given by
 p p
a cos( K u) + b sin( K u) , if K > 0,


f (u) = au + b, with |a| ≤ 1 , if K = 0,
a cosh(p−K u) + b sinh(p−K u) , if K < 0.


78 / 110
3.6 SoR with vanishing Gaussian curvature

For K = 0, f (u) = au + b, with |a| ≤ 1.

79 / 110
3.6 SoR with vanishing Gaussian curvature

1 If a = 0, we get a cylinder of radius f = b.

80 / 110
3.6 SoR with vanishing Gaussian curvature

2 If a = 1 or a = −1, we get a plane orthogonal to the axis of rotation.

81 / 110
3.6 SoR with vanishing Gaussian curvature

3 If 0 < |a| < 1, we get a circular cone.

82 / 110
3.6 SoR with constant positive Gaussian curvature
p p
For K > 0, f (u) = a cos( K u) + b sin( K u). We can set b = 0 by a translation
of parameters. For (f 0 )2 + (g 0 )2 = 1 to have a real solution g, it is necessary
p
that 0 < a2 K sin2 ( K u) ≤ 1 and consequently,
Z tq p
g(u) = 1 − a2 K sin2 ( K x) dx.
0

83 / 110
3.6 SoR with constant positive Gaussian curvature

1 If a2 K = 1, we get a sphere.

84 / 110
3.6 SoR with constant positive Gaussian curvature

2 If 0 < a2 K < 1, we get an elongated sphere.

85 / 110
3.6 SoR with constant positive Gaussian curvature

3 If a2 K > 1, we get an oblate sphere.

86 / 110
3.6 SoR with constant negative Gaussian curvature
p p
For K < 0, f (u) = a cosh( −K u) + b sinh( −K u).
1 If a2 < b2 , we get a conic type.

87 / 110
3.6 SoR with constant negative Gaussian curvature

2 If a2 > b2 , we get a hyperboloid type.

88 / 110
3.6 SoR with constant negative Gaussian curvature

3 If a = b and K = −1, we get a pseudo-sphere (also known as the


Beltramis surface/tractroid/bugle surface) with
Z 1p
f (u) = aeu and g(u) = 1 − a2 e2x dx.
0

89 / 110
3.7 Ruled Surfaces

In this section we consider ruled surfaces.


A ruled surface is the union of a differentiable one-parameter family of
lines in R3 .
Each line can be specified by a point α(u) and a direction γ(u).
Being "differentiable" means that both α and γ are differentiable
vector functions over some interval.
Definition 3.7.1 Ruled Surface
A ruled surface is a surface in R3 which can be parametrized by

X (u, v) = α(u) + v~
γ(u),

for some vector functions α and γ differentiable on an interval I.

The lines Lt through α(u) with direction γ(u) are called rulings, and the
curve α is called the directrix of the surface.
Ruled surfaces may not be regular. The definition allows singular
points, that is, points where Xu × Xv =~0. 90 / 110
3.7 Ruled Surfaces

Example 3.7.2. [Cylinders] A cylinder is formed by a one-parameter family


of lines whose directrix α is planar and the ruling direction γ is constant.
Example 3.7.3. [Cones] A cone is formed by a one-parameter family of
lines whose directrix α lies on a plane π, and the rulings Lt all pass through
a common point p ∉ π.
Example 3.7.4. [Hyperboloid of One Sheet] The hyperboloid of one sheet
with equation x2 + y 2 − z2 = 1, is a ruled surface. We may consider the circle

α(u) = (cos u, sin u, 0)

as the directrix, and the vectors

γ(u) = α 0 (u) + k̂ = (− sin u, cos u, 1)

as ruling directions . We obtain the parametrized surface

X (u, v) = α(u) + v γ(u) = (cos u − v sin u, sin u + v cos u, v) .


91 / 110
3.7 Ruled Surfaces

Alternatively, we may consider the same directrix with γ̃(u) = α 0 (u) − k̂ as


ruling directions. In this case, we get the parametrized surface
X̃ (u, v) = (cos u − v sin u, sin u + v cos u, −v) .
It is very easy to verify that both X and X̃ satisfy the the Cartesian equation
of the "doubly-ruled" hyperboloid of one sheet.

92 / 110
3.7 Ruled Surfaces

Example 3.7.5. [Hyperbolic Paraboloid] Hyperbolic paraboloids are a


ruled surfaces as they can be parametrized following Definition 3.7.1.
In particular, we may take
α(u) = au, 0, u2
¡ ¢

as directrix and
γ(u) = (a, b, 2u)
as ruling directions. The parametrization would be
X (u, v) = au + av, bv, u2 + 2uv .
¡ ¢

Alternatively, we can use the same directrix and consider the ruling
directions
γ̃(u) = (a, −b, 2u).
In this case, the parametrization would be
X̃ (u, v) = au + av, −bv, u2 + 2uv .
¡ ¢

x2 y 2
It is easy to verify that both X and X̃ satisfy z = − .
a2 b2
93 / 110
3.7 Ruled Surfaces

Proposition 3.7.6
The Gaussian curvature of a ruled surface is everywhere nonpositive.

Proof. Consider a ruled surface X (u, v) = α(u) + v γ(u). Clearly, Xvv (u, v) =~0.
Hence, L22 = 0, and consequently,
det L (L12 )2
K= =− ≤0
det g det g
since g is positive-definite. 2
94 / 110
3.7 Ruled Surfaces

Remark 3.7.7. The only surfaces of revolution which are also ruled surfaces
are planes, circular cylinders, cones, and hyperboloids (of one sheet).

A surface with Gaussian curvature K = 0 is called flat or developable.

Examples of flat surfaces: plane, cylinders, cones


"developable" - comes from the fact that these surfaces can be easily
formed by bending a sheet of paper leaving the Gaussian curvature
unchanged and identically zero, as we shall see in the next chapter.

Corollary 3.7.8
A ruled surface is flat if and only if L12 = 0.

We will prove that there three classes of flat ruled surfaces: generalized
cylinders, generalized cones, and the least obvious but most interesting is
the class of tangent developables.
95 / 110
3.7 Ruled Surfaces

Definition 3.7.9. Generalized Cylinder and Generalized Cone


A ruled surface S is said to be
1 a generalized cylinder over a curve α : I → R3 if S can be parametrized
as
X (u, v) = α(u) + v~q ,
where~q ∈ R3 is fixed;
2 a generalized cone over a curve α : I → R3 if S can be parametrized as

p + v α(u) ,
X (u, v) =~

p ∈ R3 is fixed (called vertex of the cone).


where~

96 / 110
3.7 Ruled Surfaces

a cylinder over a figure eight a cone over a figure eight

97 / 110
3.7 Ruled Surfaces

Definition 3.7.9. Tangent Developable


A ruled surface S is said to be the tangent developable of a curve α : I → R3 if
S can be parametrized as
X (u, v) = α(u) + v α 0 (u).

tangent developable of a helix


98 / 110
3.7 Ruled Surfaces

Proposition 3.7.10
If S is a generalized cylinder, a generalized cone, or a tangent developable,
then S is flat.

Proof.
Consider a generalized cylinder S parametrized by X (u, v) = α(u) + v~q.
Clearly, Xuv (u, v) = 0, and thus L12 = 0. Hence, S is flat.
p + v α(u). Then
Consider generalized cone S parametrized by X (u, v) =~

Xu (u, v) = v α 0 (u) , Xv (u, v) = α(u) , and Xuv = α 0 (u).

Therefore,
v α 0 (u) × α(u)
µ ¶
L12 = · α 0 (u) = 0,
kv α 0 (u) × α(u)k
and hence, S is flat.

99 / 110
3.7 Ruled Surfaces

Consider a tangent developable S given by X (u, v) = α(u) + v α 0 (u).


Then

Xu (u, v) = α 0 (u) + v α 00 (u) , Xv (u, v) = α 0 (u) , and Xuv = α 00 (u).

We can see that


Xu × Xv = v α 00 (u) × α 0 (u),
and thus,
v α 00 (u) × α 0 (u)
µ ¶
L12 = · α 00 (u) = 0.
kv α 00 (u) × α 0 (u)k
Hence, S is flat.
2

100 / 110
3.7 Ruled Surfaces

For calculation purposes, we make simplifications on the parametrization


of a noncylindrical ruled surface X (u, v) = α(u) + v~ γ(u).
We impose kγ(u)k = 1 and this does not change the image of the
parametrization.
We choose the directrix to be some curve β that satisfies
β 0 (u) · γ 0 (u) = 0.
This curve is called the striction curve of the ruled surface.
Since β(u) lies on the surface,
β(u) = α(u) + v(u)γ(u).
Then
β 0 (u) = α 0 (u) + v 0 (u)γ(u) + v(u)γ 0 (u).
Taking the dot product of each side of the equation with γ 0 , we get
α 0 (u) · γ 0 (u) + v(u)kγ 0 (u)k2 = 0.
Thus,
α 0 (u) · γ 0 (u)
v(u) = − .
kγ 0 (u)k2
101 / 110
3.7 Ruled Surfaces

Remark 3.7.11. The striction curve of a noncylindrical ruled surface does


not depend on the choice of directrix. In other words, the striction curve is
unique.
Proof. Let S be a noncylindrical ruled surface. Suppose α and α̃ are distinct
directrices, with corresponding striction curves β and β̃. Then
α(u) + v γ(u) = α̃(u) + w γ(u)
for some vector function γ with kγk = 1, and w dependent on v. Since β and
β̃ lie on S, they satisfy
α 0 (u) · γ 0 (u)
β(u) = α(u) − γ(u)
kγ 0 (u)k2
˜ = α̃(u) − α̃ (u) · γ (u) γ(u).
0 0
and β(u)
kγ 0 (u)k2
Then á !
α 0 − α̃ 0 · γ 0
¢
β − β̃ = (α − α̃) − γ.
kγ 0 k2
102 / 110
3.7 Ruled Surfaces

On the other hand,


α − α̃ = (w − v)γ,
which further implies that

α 0 − α̃ 0 = (w − v)γ 0 .

Hence,
(w − v)γ 0 · γ 0
µ ¶
β − β̃ = (w − v)γ − γ = 0.
kγ 0 k2
2
Theorem 3.7.12
Let α : I → R3 be any directrix of a noncylindrical ruled surface S with unit
ruling directions γ(u). Then there exists a unique function β with β 0 · γ 0 = 0
such that S has reparametrization of the form
X (u, v) = β(u) + v γ(u).
103 / 110
3.7 Ruled Surfaces

Theorem 3.7.13
The curvature of a noncylindrical ruled surface parametrized using unit
ruling directions γ along its striction curve β is

kγ 0 k4 β 0 × γ · γ 0
¡ ¢
K = −¡ ¢2 .
β 0 × γ · γ 0 + v2 kγ 0 k4

Proof. left as an exercise

Proposition 3.7.14
Let X (u, v) = β(u) + v γ(u), with kγ(u)k = 1, parametrize a flat ruled surface S.

1 If γ 0 (u) = 0, then S is a generalized cylinder.


2 If β 0 (u) = 0, then S is a generalized cone.
3 If both β 0 and γ 0 are nonvanishing, then S is a tangent developable of
its striction curve.
104 / 110
3.7 Ruled Surfaces

Proof. Parts (1) and (2) immediately follow from Definition 3.7.9, so we just
need to prove (3). Assume β is a unit-speed striction curve of S; that is,

β 0 · γ 0 = 0.

Since K = 0, it follows from Theorem 3.7.13 that

β 0 × γ · γ 0 = 0.

Therefore, β 0 and γ are parallel. This means S is the tangent developable of


its striction curve.
2

105 / 110
3.8 Minimal Surfaces

Definition 3.8.1
A minimal surface is a parametrized surface of class C 2 that satisfies the
regularity conditon and for which the mean curvature is H = 0.

The principal curvatures of a minimal surface are negatives of each


other.
The Gaussian curvature of a minimal surface is K ≤ 0.
Geometrically, minimal surface locally minimizes surface area.
Physical models of area-minimizing minimal surfaces can be made by
dipping a wire frame into a soap solution, forming a soap film, which
is a minimal surface whose boundary is the wire frame.
Example 3.8.2. The helicoid X (u, v) = (v cos u, v sin u, c u), for some nonzero
constant c, is a minimal surface.
Example 3.8.2. The catenoid X (u, v) = (a cosh v cos u, a cosh v sin u, a v), for
some positive constant a, is a minimal surface.
106 / 110
3.8 Minimal Surfaces

107 / 110
3.8 Minimal Surfaces

108 / 110
3.8 Minimal Surfaces

109 / 110
3.8 Minimal Surfaces

110 / 110

You might also like