Advanced CAD/CAM

Differential Geometry: Curves and Surfaces1

Kwanghee Ko

The School of Mechanical Engineering

Gwangju Institute of Science and Technology

1
Part of the materials in this lecture note was permitted for use in class by Professor N. M. Patrikalakis,
Massachusetts Institute of Technology, Cambridge, MA 02139-4307, USA

1
Contents

1 Differential Geometry of Curves 3

1.1 Definition of curves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.1.1 Plane curves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.1.2 Space curves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.2 Arc length . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.3 Tangent vector . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.4 Normal vector and curvature . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.5 Binormal vector and torsion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
1.6 Serret-Frenet Formulae . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.7 Selected Topics of Global Properties of Plane Curves . . . . . . . . . . . . . . . 17

2 Differential Geometry of Surfaces 19

2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.2 Definition of surfaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.3 Curves on a surface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.4 First fundamental form (arc length) . . . . . . . . . . . . . . . . . . . . . . . . 22
2.5 Tangent plane . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
2.6 Normal vector . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
2.7 The Geometry of the Gauss Map . . . . . . . . . . . . . . . . . . . . . . . . . . 26
2.7.1 Definition of the Gauss Map and Its Fundamental Properties . . . . . . 26
2.8 Second fundamental form II (curvature) . . . . . . . . . . . . . . . . . . . . . . 27
2.9 Principal curvatures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30

Bibliography 33

2
Chapter 1

Differential Geometry of Curves

Differential geometry is a discipline in mathematics that studies various problems in geometry

using calculus, differential calculus, integral calculus and linear algebra. It presents the theory
of 2D and 3D curves and surfaces by dealing with the local geometric properties of curves and
surfaces. It is an essential part of geometric modeling, which is a building block of CAD/CAM.

1.1 Definition of curves

Two types of curves are defined: a curve on 2D planar space and a curve in 3D space. The
mathematical definitions of each type and their properties are provided as follows.

1.1.1 Plane curves

• An implicit curve in the xy plane is defined by a set of the points satisfying the equation
f (x, y) = 0. An example of an implicit curve is x2 + y 2 = a2 , which is a circle with the
center (0, 0) and the radius of a.

Suppose that we have an implicit curve f (x, y) = 0. The properties are summarized
as follows:

– It is difficult to trace the curve.

– It is easy to check if a point lies on the curve.
– It can have either multi-valued or closed form.
– It is easy to evaluate a tangent line to the curve even when it has a vertical or near
vertical tangent.
– It is axis dependent. Namely, it is complicated to transform the curve to a different
coordinate system.

Example: x3 + y 3 = 3xy. We have f (x, y) = x3 + y 3 − 3xy = 0.

Example: If we translate by (-2,-1) and rotate the axes by θ = −atan( 34 ), the hyperbola
x2 y2 2 2
4 − 2 = 1, will become 2x − 72xy + 23y + 140x − 20y + 50 = 0.

3
 Figure 1.1: (a) Descartes, (b) Hyperbola

• An explicit curve is defined by y = f (x). Namely, one of the variables is expressed in

terms of the other. An example of an explicit curve is y = x2 . The explicit curve has the
following properties.

– It is easy to trace an explicit curve.

– It is easy to check if a point lies on the curve.
– Multi-valued and closed curves can not be easily represented.
– It is difficult to evaluate tangent line to the curve when the curve has a vertical or
near vertical tangent.
– It is axis dependent.

Example: If a circle is √
represented by an explicit equation, it must
√ be divided into two
segments, with y = + r2 − x2 for the upper half and y = − r2 − x2 for the lower
half, see Figure 1.2. This kind of segmentation creates cases which are inconvenient in
geometric modeling.

y = + r 2 − x2

o x

y = − r2 − x2

Figure 1.2: Description of a circle with an explicit equation.

√
Note: The derivative of y = x at the origin x = 0 is infinite, see Figure 1.3.

4
 Figure 1.3: Vertical slopes for explicit curves involve non-polynomial functions.

• Parametric curve: A curve can be defined using a parameter. For example, we have (x, y),
each component of which is defined by x = x(t), y = y(t), t1 ≤ t ≤ t2 . 2D coordinates
(x, y) can be expressed as functions of a parameter t. An example of a parametric curve
is x = a cos(t), y = a sin(t), 0 ≤ t < 2π. The properties of the parametric curve
representation are as follows:

– It is easy to trace parametric curves.

– It is relatively difficult to check if a point lies on the curve.
– Closed and multi-valued curves are easy to represent.
– It is easy to evaluate tangent line to the curve when the curve has a vertical or near
vertical tangent.
– It is axis independent. Namely, it is easy to transform it to a different coordinate
system.

Example: Folium of Descartes, see Figure 1.1, can be expressed as

!
3t 3t2
r(t) = , − ∞ < t < ∞.
1 + t3 1 + t3

Given x(t) = x0 we solve for t, which is given to y(t) = y0 . We need to solve a nonlinear
√
equation to check if a point lies on the curve. Explicit curve y = x can be expressed as
x = t2 , y = t (t ≥ 0).

(2t, 1)
r = (t2 , t), ṙ = (2t, 1), t= √ , t = 0, t = (0, 1)
4t2 + 1

Therefore there is no problem representing a vertical tangent computationally.

5
1.1.2 Space curves
• Implicit curves
In 3D, a single equation generally represents a surface. For example x2 + y 2 + z 2 = a2 is
a sphere. Thus, the curve appears as the intersection of two surfaces.
F (x, y, z) = 0 ∩ G(x, y, z) = 0

Example: Intersection of the two quadric surfaces z = xy and y 2 = zx gives cubic

parabola. (These two surfaces intersect not only along the cubic parabola but also along
the x-axis.)
• Explicit curves
If the implicit equations can be solved for two of the variables in terms of the third, say
for y and z in terms of x, we get
y = y(x), z = z(x)
Each of the equations separately represents a cylinder projecting the curve onto one of
the coordinate planes. Therefore intersection of the two cylinders represents the curve.
Example: Intersection of the two cylinders y = x2 , z = x3 gives a cubic parabola.
• Parametric curves x = x(t), y = y(t), z = z(t), t1 ≤ t ≤ t2
The 3D coordinates (x, y, z) of the point can be expressed on functions of parameter t.
Here x(t), y(t), z(t) have continuous derivatives of the rth order, and the parameter t is
restricted to some interval called the parameter space (i.e., t1 ≤ t ≤ t2 ). In this case the
curve is said to be of class r, denoted as C r .
In vector notation:
r = r(t)
where r = (x, y, z), r(t) = (x(t), y(t), z(t))
Example: Cubic parabola
x = t, y = t2 , z = t3
Example: Circular helix, see Fig. 1.4.

x = a cos(t), y = a sin(t), z = bt, 0≤t≤π

Using v = tan 2t
s
t 1 − cos t 1 − cos t
v = tan = ⇒ v2 =
2 1 + cos t 1 + cos t
1 − v2 2v
⇒ cos t = 2
⇒ sin t =
1+v 1 + v2
Therefore the following parametrization will give the same circular helix.
!
1 − v2 2av
r= a , , 2b tan−1 v , 0≤v<∞
1 + v2 1 + v2

6
 >>
>> a= 2;
>> b = 3;
>> u = [0 : 6 * pi / 100 : 6 * pi];
>> plot3(a * cos(u), a * sin(u), b * u)
>> xlabel(’X’);
>> ylabel(’Y’);
>> zlabel(’Z’);
>> print(’circHelix.ps’)

60

50

40

30
Z

20

10

0
2
1 2
0 1
0
−1
−1
Y −2 −2
X

Figure 1.4: Circular helix plotted using MATLAB.

7
1.2 Arc length

Figure 1.5: A segment ∆r connecting two point p and q on a parametric curve r(t).

Consider a curve in 3D space with two points p and q as shown in Figure 1.5. First, let us
express the vector ∆r connecting two points p and q on an arc at parametric locations t and
t + ∆t, respectively, as
∆r = p − q = r(t + ∆t) − r(t).
As p and q become infinitesimally close, the length of the segment connecting the two points
approaches the arc length between the two points along the curve, r(t) and r(t + ∆t). Or using
Taylor’s expansion on the norm (length) of the segment ∆r and letting ∆t → 0, we can express
the differential arc length as
dr dr
∆s ' |∆r| = |r(t + ∆t) − r(t)| = | ∆t + O(∆t2 )| ' | |∆t.
dt dt
Thus as ∆t → 0
dr
ds = | |dt = |ṙ|dt.
dt
Definitions
d
≡ ˙
dt
d
≡ 0
ds
ds
Hence the rate of change dt of the arc length s with respect to the parameter t is
ds q 2
= ẋ (t) + ẏ 2 (t) + ż 2 (t) (1.1)
dt

8
 ds
dt is called the parametric speed. It is, by definition, non-negative (s being measured always
in the sense of increasing t).
If the parametric speed does not vary significantly, parameter values t0 , t1 , · · · , tN corre-
sponding to a uniform increment ∆t = tk − tk−1 , will be evenly distributed along the curve, as
illustrated in Figure 1.6.

ds y
dt t5

t2 t3 t4
Parameter Space t1

t0
t x
t0 t1 t2 t3 t4 t5

Figure 1.6: When parametric speed does not vary, parameter values are uniformly spaced along
a parametric curve.

The arc length of a segment of the curve between points r(to ) and r(t) can be obtained as
follows:
Z tq Z t√
s(t) = ẋ2 (t) + ẏ 2 (t) + ż 2 (t)dt = ṙ · ṙdt (1.2)
to to
Derivatives of arc length s w.r.t. parameter t and vice versa :
ds √
ṡ = = |ṙ| = ṙ · ṙ (1.3)
dt
dṡ ṙ · r̈
s̈ = =√ (1.4)
dt ṙ · ṙ
···
··· ds̈ (ṙ · ṙ)(ṙ· r +r̈ · r̈) − (ṙ · r̈)2
s = = 3 (1.5)
dt (ṙ · ṙ) 2
dt 1 1
t0 = = =√ (1.6)
ds |ṙ| ṙ · ṙ
dt 0 ṙ · r̈
t00 = =− (1.7)
ds (ṙ · ṙ)2
···
000 dt00 (r̈ · r̈ + ṙ· r)(ṙ · ṙ) − 4(ṙ · r̈)2
t = =− 7 (1.8)
ds (ṙ · ṙ) 2

1.3 Tangent vector

The vector r(t + ∆t) − r(t) indicates the direction from r(t) to r(t + ∆t). If we divide the vector
by ∆t and take the limit as ∆t → 0, then the vector will converge to the finite magnitude
vector ṙ(t).
ṙ(t) is called the tangent vector.
Magnitude of the tangent vector
ds
|ṙ| = (1.9)
dt

9
 Unit tangent vector
dr
ṙ dr
t= = dt
ds
= ≡ r0 (1.10)
|ṙ| dt
ds
Definition : A parametric curve is said to be regular if |ṙ(t)| =
6 0 for all t ∈ I. The points
where |ṙ(t)| = 0 are called irregular (singular) points.
Note that at irregular points the parametric speed is zero.
Example: semi-cubical parabola r(t) = (t2 , t3 ), see Figure 1.7
ṙ(t) = (2t, 3t2 )
p q
|ṙ(t)| = 4t2 + 9t4 = t2 (4 + 9t2 )

when t = 0, |ṙ(t)| = 0

t>0

x
0

t<0

Figure 1.7: A singular point occurs on a semi-cubical parabola in the form of a cusp.

Here are some useful formulae for computing the unit tangent vector:
• 3D Parametric curve r(t)
dr dr dt ṙ (ẋ, ẏ, ż)
t = r0 = = = =p 2
ds dt ds |ṙ| ẋ + ẏ 2 + ż 2

• 2D Implicit curve f (x, y) = 0

(fy , −fx )
t= q
fx2 + fy2

• 2D Explicit curve y = f (x)

(1, f˙)
t= q
1 + f˙2

10
 Example: For a circular helix r(t) = (a cos t, a sin t, bt)
• Parametric speed
ds q
= |ṙ(t)| = ẋ2 (t) + ẏ 2 (t) + ż 2 (t)
dt
ṙ(t) = (−a sin t, a cos t, b)
(
p The curve is regular and has
|ṙ(t)| = a2 + b2 = c = const ⇒
good parametrization.

• Unit tangent vector

ṙ a a b
t= = (− sin t, cos t, ) (1.11)
|ṙ| c c c
• Arc length Z t Z tp
s(t) = |ṙ|dt = a2 + b2 dt = ct (1.12)
0 0

• Arc length parametrization

s
t = (1.13)
c
s s bs
r(s) = (a cos , a sin , ) (1.14)
c c c
check (1.15)
dr a s a s b
= (− sin , cos , ) = t (1.16)
ds c c c c c

1.4 Normal vector and curvature

1
r’(s) r’(s+∆s)
1
∆s
r’(s) r’(s+∆s)−r’(s)=∆θ
n ∆θ
r’(s+∆s)
ρ

∆θ

center of
curvature

Figure 1.8: Derivation of the normal vector of a curve.

Let us consider the second derivative r00 (s), see Figure 1.8.
r0 (s + ∆s) − r0 (s)
r00 (s) = lim (1.17)
∆s→0 ∆s

11
 As ∆s → 0 r0 (s + ∆s) − r0 (s) becomes perpendicular to the tangent vector i.e. normal
direction.

Also |r0 (s + ∆s) − r0 (s)| = ∆θ · 1 = ∆θ as ∆s → 0.

Thus
∆θ
∆θ ρ 1
|r00 (s)| = lim = lim = ≡κ (1.18)
∆s→0 ∆s ∆s→0 ∆θ ρ
κ is called the curvature. It follows that

κ2 = r00 · r00 . (1.19)

Consequently

r00 (s) = t0 = κn (1.20)

Thus using equations (1.6) and (1.7), we obtain

d2 r dt d (ṙ · ṙ)r̈ − (ṙ · r̈)ṙ

κn = 2
= = (ṙt0 ) = r̈(t0 )2 + ṙt00 = (1.21)
ds ds ds (ṙ · ṙ)2
(ṙ · ṙ)r̈ − (ṙ · r̈)ṙ (ṙ · ṙ)r̈ − (ṙ · r̈)ṙ (ṙ × r̈) · (ṙ × r̈)
   
κ2 = (κn) · (κn) = 2
· 2
= (1.22)
(ṙ · ṙ) (ṙ · ṙ) (ṙ · ṙ)3

where the identity (a × b) · (a × b) = (a · a)(b · b) − (a · b)2 is used.

Here are some useful formulae for computing the normal vector and curvature:
• 2D parametric curve r(t), see Figure 1.9

z
n

t
ez κ<0
y n
κ=0
inflection t
point
κ>0
n
x
t

Figure 1.9: Normal and tangent vectors along a 2D curve.

(−ẏ, ẋ)
n = ez × t = p 2 , ez = (0, 0, 1) (1.23)
ẋ + ẏ 2
ẋÿ − ẏẍ
κ = 3 (1.24)
(ẋ2 + ẏ 2 ) 2

12
 • 2D implicit curve f (x, y) = 0

(fx , fy ) ∇f
n = ez × t = q = (1.25)
fx2 + fy2 |∇f |

fxx fy2 − 2fxy fx fy + fx2 fyy

κ = − 3 (1.26)
(fx2 + fy2 ) 2

• 2D Explicit curve y = f (x)

(−ẏ, 1)
n = ez × t = p (1.27)
1 + ẏ 2
ÿ
κ = 3 (1.28)
(1 + ẏ 2 ) 2

1.5 Binormal vector and torsion

b
t n t : osculating plane
n
b n : normal plane

y b t : rectifying plane

r(t)
x

Figure 1.10: The tangent, normal, and binormal vectors define an orthogonal coordinate system
along a space curve.

Let us define a unit binormal vector, see Figure 1.10

b=t×n (1.29)

We have

t·n=0 n·b=0 b·t=0

b=t×n t=n×b n=b×t

The osculating plane can be defined as the plane passing through three consecutive points
on the curve. The rate of change of the osculating plane is expressed by the vector
d dt dn
b0 = (t × n) = ×n+t× = t × n0 (1.30)
ds ds ds
where we used the fact that dt
ds = r00 = κn.

13
 From n · n = 1 → differentiate w.r.t. s → 2n0 · n = 0 → n0 ⊥ n
Thus n0 is parallel to the rectifying plane (b, t), and n0 can be expressed as a linear com-
bination of b and t.

n0 = µt + τ b (1.31)

Substitute (1.31) into (1.30)

b0 = t × (µt + τ b) = τ t × b = −τ b × t = −τ n (1.32)

τ is called the torsion.

Consequently
···
0 (r0 r00 r000 )
0 (ṙr̈ r)
τ = −n · b = −n · (t × n) = 00 00 = (1.33)
r ·r (ṙ × r̈) · (ṙ × r̈)
Triple scalar product

(abc) = a · (b × c) = (a × b) · c (1.34)
also

(abc) = (bca) = (cab) cyclic permutation (1.35)

Geometrically, (abc) equals to the volume of a parallelepiped having the edge vectors
a, b, c, as in Figure 1.11.

b ax ay az
a a (bxc)= bx by bz
c cx cy cz

Figure 1.11: The computation of the volume of a parallelepiped

1.6 Serret-Frenet Formulae

From equations (1.20) and (1.32), we found that

t0 = κn (1.36)
0
b = −τ n (1.37)

How about n0 ?

n0 = (b × t)0 = b0 × t + b × t0 = −τ n × t + b × (κn) = −κt + τ b (1.38)

In matrix form we can express the differential equations as

    
t0 0 κ(s) 0 t
 0  
 n  =  −κ(s) 0 τ (s)   n  (1.39)
 
b0 0 −τ (s) 0 b

14
 Thus, the curve is completely determined by its curvature and torsion as a function of
parameter s. The equations κ = κ(s), τ = τ (s) are called intrinsic equations. The for-
mulae 1.39 are known as the Serret-Frenet Formulae and describe the motion of moving a
trihedron (t, n, b) along the curve.
Intuitively, we can form a curve in R3 by bending and twisting a straight line. Here,
’bending’ creates a curvature and ’twisting’ a torsion. This way of construction of a curve leads
us to understand that the curvature and the torsion describe completely the local behavior of
the curve in R3 . Then we have the following fundamental theorem of the local theory of curves
[1].

Theorem 1.6.1 Given differentiable functions κ(s) > 0 and τ (s), s ∈ I, there exists a regular
parametrized curve α : I → R3 such that s is the arc length, κ(s) is the curvature, and τ (s) is
the torsion of α. Moreover, any other curve ᾱ satisfying the same conditions, differs from α by
a rigid motion; that is, there exists an orthogonal linear map ρ of R3 , with positive determinant,
and a vector c such that ᾱ = ρ · α + c.

Example: Determining the shape of a curve from curvature information and boundary
conditions only.
Given:
1
κ= = const
R
We find
dt n
= (1.40)
ds R
dn t
=− (1.41)
ds R
If we differentiate Equation 1.40 with respect to s,
d2 t 1 dn
2
= . (1.42)
ds R ds
Now, substitute Equation 1.42 into Equation 1.41
d2 t t
2
+ 2 = 0. (1.43)
ds R
dr
Recognizing that t = ds , we can change variables from t to r, transforming Equation 1.43
into
d3 r 1 dr
ds3
+ R2 ds
=0
! or ! !
d3 x(s) 1 d x(s) 0
ds3
+ R2 ds
= (1.44)
y(s) y(s) 0
The solution to Equation 1.44 is
! !
s s
x(s) = C1 + C2 cos + C3 sin (1.45)
R R
! !
s s
y(s) = C10 + C20 cos + C30 sin (1.46)
R R

15
 Assume we are given suitable initial conditions or boundary conditions. For this example,
we will use:
1
x(0) = R x0 (0) = 0 x00 (0) = − (1.47)
R
y(0) = 0 y 0 (0) = 1 y 00 (0) = 0 (1.48)

Solving for the constants in the general solution gives

C1 = C3 = 0 C2 = R (1.49)
C10 = C20 =0 C30 =R (1.50)

Thus, we find our solution is given by

s
x(s) = R cos (1.51)
R
s
y(s) = R sin (1.52)
R
which is precisely a circle of radius R satisfying the conditions (1.47) and (1.48).
Example: A circular helix r = (a cos sc , a sin sc , bs
c)

a s a s b
r0 (s) = (− sin , cos , )
c c c c c
a s a s
r00 (s) = (− cos , − 2 sin , 0)
c2 c c c
a s a s
r000 (s) = ( 3
sin , − 3 cos , 0)
c c c c

a2 2 s 2 s a2
κ2 = r00 · r00 = (cos + sin ) = = constant
c4 c c c4
(r0 r00 r000 ) (r0 r00 r000 )
τ = =
r00 · r00 κ2
 − a sin s a
cos sc b
 

 c c c c 
 
4
c  a
 
=  − 2 cos sc − ca2 sin sc 0 

a2  c 

 a
 3 sin s − ca3 cos sc

c c 0 

c4 b a2 s s
= ( (cos2 + sin2 ))
a2 c c5 c c
b
= = constant
c2
Note: when b > 0, it is a right-handed helix;
when b < 0, it is a left-handed helix.

16
1.7 Selected Topics of Global Properties of Plane Curves
• A differentiable function on a closed interval [a, b] is the restriction of a differentiable
function defined on an open interval containing [a, b].
• A closed plane curve is a regular parametrized curve α : [a, b] → R2 such that α and all
its derivatives agree at a and b; that is,
α(a) = α(b), α0 (a) = α0 (b), α00 (a) = α00 (b), · · · . (1.53)

• The curve α is simple if it has no further self-intersections; that is, if t1 , t2 ∈ [a, b), t1 6= t2 ,
then α(t1 ) 6= α(t2 ).
• For a simple closed plane curve we have the isoperimetric inequality property as follows:
Theorem 1.7.1 Let C be a simple closed plane curve with length l, and let A be the area
of the region bounded by C. Then
l2 − 4πA ≤ 0, (1.54)
and equality holds if and only if C is a circle.
• A vertex of a regular plane curve α : [a, b] → R2 is a point t ∈ [a, b] where κ0 = 0. Then
we have the four-vertex theorem.
Theorem 1.7.2 A simple closed convex curve has at least four vertices.
• Theorem 1.7.3 Let C be a regular plane curve with length l. The measure of the set of
straight lines (counted with multiplicities) which meet C is equal to 2l.
This theorem can be used to estimate lengths of curves. Consider a family of parallel
straight lines such that two consecutive lines are at a distance r. Rotate this family by
angles of π4 , 2π 3π
4 , 4 in order to obtain four families of straight lines. Let n be the number
of intersection points of a curve C with all these lines. Then the length of C can be
approximated by
1 π
length(C) = nr . (1.55)
2 4

Example [1]

Figure 1.12 is a drawing of an electron micrograph of a circular DNA molecule and we

want to estimate its length. The four families of straight lines at a distance of 7mm and
angles of π4 are drawn over the picture.
The number of intersection points is found to be 153. Thus,
1 π 1 3.14
n = 153 = 60. (1.56)
2 4 2 4
Since the reference line in the picture represents 1 micrometer and measures, in our scale,
25 millimeters, r = 25 7 , and thus the length of this DNA molecule, form our values, is
approximately
25
60 = 16.6mm. (1.57)
7
The actual value is 16.3mm.

17
Figure 1.12: Example figure

18
Chapter 2

Differential Geometry of Surfaces

2.1 Introduction
Differential geometry is concerned with those properties of surfaces which depend on their
behavior in a neighborhood of a point [1]. For this purpose, we need a mathematical definition
of a surface which is presented in the following section.

2.2 Definition of surfaces

Roughly speaking, a regular surface in R3 is obtained by taking pieces of a plane, deforming,
and arranging them in such a way that the resulting figure has no sharp points, edges, or
self-intersections and so that it makes sense to speak of a tangent plane at points of the figure
[1].
The correct mathematical definition of a surface is given by the following:

Definition 2.2.1 A subset S ⊂ R3 is a regular surface if, for each p ∈ S, there exists a
neighborhood V in R3 and a map x : U → V ∩ S of an open set U ⊂ R2 onto V ∩ S ⊂ R3
such that

1. x is differentiable.

2. x is a homeomorphism.

3. For each q ∈ U, the differential dxq : R2 → R3 is one-to-one.

∂x ∂x
Equivalently, Condition 3 means that the vector cross product ∂u × ∂u 6= 0.

• Implicit surfaces F (x, y, z) = 0

2 2 2
Example: xa2 + yb2 + zc2 = 1 Ellipsoid, see Figure 2.1.

• Explicit surfaces
If the implicit equation F (x, y, z) = 0 can be solved for one of the variables as a function
of the other two, we obtain an explicit surface, as shown in Figure 2.2. Example: z =
1 2 2
2 (αx + βy )

19
 Figure 2.1: Ellipsoid.

Figure 2.2: Explicit quadratic surfaces z = 12 (αx2 + βy 2 ). (a) Left: Hyperbolic paraboloid
(α = −3, β = 1). (b) Right: Elliptic paraboloid (α = 1, β = 3).

20
 • Parametric surfaces x = x(u, v), y = y(u, v), z = z(u, v)
Here functions x(u, v), y(u, v), z(u, v) have continuous partial derivatives of the rth order,
and the parameters u and v are restricted to some intervals (i.e., u1 ≤ u ≤ u2 , v1 ≤ v ≤ v2 )
leading to parametric surface patches. This rectangular domain D of u, v is called
parametric space and it is frequently the unit square, see Figure 2.3.
The rigorous mathematical definition of a parametrized surface is given as follows:

Definition 2.2.2 A parametrized surface x : U ⊂ R2 → R3 is a differentiable map x

from an open set U ⊂ R2 into R3 . The set x(U) ⊂ R3 is called the trace of x. x is
regular if the differential dxq : R2 → R3 is one-to-one for all q ∈ U. A point q ∈ U
where dxq is not one-to-one is called a singular point of x.

If derivatives of the surface are continuous up to the rth order, the surface is said to be
of class r, denoted C r .
In vector notation:

r = r(u, v)
where r = (x, y, z), r(u, v) = (x(u, v), y(u, v), z(u, v))

Example:

r = (u + v, u − v, u2 + v 2 )

x=u+v   1
y =u−v ⇒ eliminate u, v ⇒ z = (x2 + y 2 ) paraboloid
2
z = u2 + v 2 


2.3 Curves on a surface

Let r = r(u, v) be the equation of a surface, defined on a domain D (i.e., u1 ≤ u ≤ u2 ,
v1 ≤ v ≤ v2 ). Let β(t) = (u(t), v(t)) be a curve in the parameter plane. Then r = r(u(t), v(t))
is a curve lying on the surface, see Figure 2.3. A tangent vector of curve β(t) is given by

v z
r(u,v)
D

u y
r(u(t),v(t))
β(t)=(u(t),v(t))
x
Parametric Space D 3D Space

Figure 2.3: The mapping of a curve in 2D parametric space onto a 3D bi-parametric surface
.

21
β̇(t) = (u̇(t), v̇(t)) A tangent vector of a curve on a surface is given by:
dr(u(t), v(t))
(2.1)
dt
By using the chain rule:
dr(u(t), v(t)) ∂r du ∂r dv
= + = ru u̇(t) + rv v̇(t) (2.2)
dt ∂u dt ∂v dt

2.4 First fundamental form (arc length)

Consider a curve on a surface r = r(u(t), v(t)). The arc length of the curve on a surface is
given by
dr du dv
ds = | |dt = |ru + rv |dt
dt dt dt
q
= (ru u̇ + rv v̇) · (ru u̇ + rv v̇)dt
q
= (ru · ru )du2 + 2ru rv dudv + (rv · rv )dv 2

p
= Edu2 + 2F dudv + Gdv 2 (2.3)
where
E = ru · ru , F = ru · rv , G = rv · rv (2.4)
The first fundamental form is defined as
I = dr · dr = (ru du + rv dv) · (ru du + rv dv)
= Edu2 + 2F dudv + Gdv 2 (2.5)
E, F , G are called first fundamental form coefficients Note that E = ru · ru > 0 and G =
rv · rv > 0 if ru 6= 0 and rv 6= 0. The first fundamental form I is positive definite. That is I ≥ 0
and I = 0 if and only if du = 0 and dv = 0 since
1 EG − F 2 2
I= (E du + F dv)2 + dv and EG − F 2 = |ru × rv |2 > 0.
E E
I depends only on the surface and not on the parametrization.
The area of the surface can be derived as follows:

∂r
r(u0 , v0 + δv) − r(u0 , v0 ) ' δv
∂v
∂r
r(u0 + δu, v0 ) − r(u0 , v0 ) ' δu
∂u

δA = |ru δu × rv δv| = |ru × rv |δuδv

|ru × rv |2 = (ru × rv ) · (ru × rv )

22
 δA

r(u0,v0+δv) r(u0+δu,v0)

r(u0,v0+δv)−r(u0,v0) r(u0+δu,v0)−r(u0,v0)
r(u0,v0)

Figure 2.4: Area of an infinitesimal surface patch.

Using the vector identity (a × b) · (c × d) = (a · c)(b · d) − (a · d)(b · c), we get

|ru × rv |2 = (ru · ru )(rv · rv ) − (ru · rv )2 (2.6)

2
= EG − F (2.7)

p Z Z p
δA = EG − F 2 δuδv, A= EG − F 2 dudv (2.8)

Example: For the hyperbolic paraboloid r(u, v) = (u, v, u2 −v 2 ), let us derive an expression
for the area of a region of its surface corresponding to a the circle u2 + v 2 ≤ 1 in the parametric
domain D.
We begin by forming expressions for the derivatives of the position vector r and the first
fundamental form coefficients.

ru = (1, 0, 2u)
rv = (0, 1, −2v)
E = ru · ru = 1 + 4u2
F = ru · rv = −4uv
G = rv · rv = 1 + 4v 2

Using Equation (2.8), we find

EG − F 2 = (1 + 4u2 )(1 + 4v 2 ) − 16u2 v 2 = 1 + 4u2 + 4v 2 > 0

Z Z p
A = 1 + 4u2 + 4v 2 dudv
D

To compute the area, we need to evaluate the double integral over the unit disk u2 + v 2 ≤ 1
in the parametric domain D;
Z Z p
A= 1 + 4u2 + 4v 2 du dv.
u2 +v 2 ≤1

To perform the integration, let us change variables.

u = r cos(θ), v = r sin(θ), and du dv = r dr dθ

23
 Z Z p
A = 1 + 4r2 r dr dθ
r≤1
Z 2π Z 1 p
= 1 + 4r2 r dr dθ
0 0
π √
= (5 5 − 1)
6

2.5 Tangent plane

Condition 3 in the definition of a regular surface S guarantees that for every p ∈ S the set of
tangent vectors to the parametrized curves of S, passing through p, constitutes a plane [1]. By
a tangent vector to S, at a point p ∈ S, we mean the tangent vector α0 (0) of a differentiable
parametrized curve α : (−, ) → S with α(0) → p.
Therefore, a tangent plane at a point r(uo , vo ) is the union of tangent vectors of all curves
on the surface pass through r(uo , vo ), as shown in Figure 2.5. Since the tangent vector of a
curve on a parametric surface is given by dr du dv
dt = ru dt + rv dt , the tangent plane lies on the plane
of the vectors ru and rv . The equation of the tangent plane is

Tp (u, v) = r(u, v) + λru (u, v) + µrv (u, v) (2.9)

where λ and µ are real variables parameterizing the plane.

z
r=ruu+rvv

Tp
r(u0,v0)
y

Figure 2.5: The tangent plane at a point on a surface.

The tangent plane also allows us to speak of the angle of two intersecting surfaces at a point
of intersection [1]. Namely, the angle of two intersecting surfaces at an intersection point is the
angle of their tangent planes (or their normal lines) at the point.

2.6 Normal vector

The surface normal is the vector at point r(uo , vo ) perpendicular to the tangent plane, see
Figure 2.6. And therefore
ru × rv
N= (2.10)
|ru × rv |
Note that ru and rv are not necessarily perpendicular.
A regular (ordinary) point P on the surface is defined as one for which ru × rv 6= 0. A point
where ru × rv = 0 is called a singular point. The condition ru × rv 6= 0 requires that at that
point P the vectors ru and rv do not vanish and have different directions.

24
 z

rv
Tp ru
y

Figure 2.6: The normal to the point on a surface.

Example: Elliptic Paraboloid r(u, v) = (u + v, u − v, u2 + v 2 )

ru = (1, 1, 2u)
rv = (1, −1, 2v)
 
 e ey ez 
 x
ru × rv =  1 1 2u 
 
−1 2v 
 
 1

= 2(u + v)ex + 2(u − v)ey − 2ez 6= 0

q
|ru × rv | = 2 (u + v)2 + (u − v)2 + 1
p
= 2 2u2 + 2v 2 + 1 > 0 ⇒ Regular !

(2(u + v), 2(u − v), −2)

N = √
2 2u2 + 2v 2 + 1
(u + v, u − v, −1)
= √
2u2 + 2v 2 + 1

at (u, v) = (0, 0), N = (0, 0, −1)

Example: Circular Cone r(u, v) = (u sin α cos v, u sin α sin v, u cos α), see Figure 2.7

ru = (sin α cos v, sin α sin v, cos α)

rv = (−u sin αsinv, u sin αsinv, 0)
 

 ex ey ez 

ru × rv =  sin α cos v sin α sin v cos α
 

 −u sin α sin v
 
u sin α cos v 0 
= −u sin α cos α cos vex − u sin α cos α sin vey + u sin2 αez
At the origin n = 0,
ru × rv = 0

25
 z
usinα

p
α u
u
α

usinαsinv y
singular
v
x usinαcosv

Figure 2.7: Circular cone.

Therefore, the apex of the cone is a singular point.

2.7 The Geometry of the Gauss Map

For a curve, by considering the rate of change of the tangent line of a curve C, we define
an important geometric entity, i.e. the curvature of C. The same idea can be extended to
a regular surface to define the curvature of the surface. Conceptually, we can measure how
rapidly a surface S pulls away from the tangent plane Tp (S) in a neighborhood of a point
p ∈ S. This is equivalent to measuring the rate of change at p of a unit normal vector field N
on a neighborhood of p.

2.7.1 Definition of the Gauss Map and Its Fundamental Properties

Given a parametrization r : U ⊂ R2 → S of a regular surface S at a point p ∈ S, we have a
differentiable map N : r(U ) → R3
ru × rv
N(q) = (q), (2.11)
|ru × rv |
a unit normal vector on S at q. We shall say that a regular surface is orientable if it admits
a differentiable field of unit normal vectors defined on the whole surface. The choice of such a
field N is called an orientation of S.

Definition 2.7.1 Let S ⊂ R3 be a surface with an orientation N. The map N : S → R3 takes

its values in the unit sphere
n o
S2 = (x, y, z) ∈ R3 |x2 + y 2 + z 2 = 1 . (2.12)

The map N : S → S2 , thus defined, is called the Gauss map of S.

The differential dNp of N at p ∈ S is a linear map from Tp (S) to TN (p) (S2 ). Since Tp (S) and
TN (p) (S2 ) are parallel planes, dNp can be looked upon as a linear map on Tp (S). dNp measures
how N pulls away from N(q) in a neighborhood of p. In the case of curves, this measure is
given by a number, the curvature. In the case of surfaces, this measure is characterized by a
linear map.

26
2.8 Second fundamental form II (curvature)
We start with two definitions as follows:
Definition 2.8.1 The quadratic form II, defined in Tp (S) by IIp (v) = − < dNp (v), v >,
v ∈ Tp (S), is called the second fundamental form of S at p.

Definition 2.8.2 Let C be a regular curve in S passing through p ∈ S, κ the curvature of C

at p, and cos θ =< n, N >, where n is the normal vector to C and N is the normal vector to
S at p. The number κn = κ cos θ is then called the normal curvature of C ⊂ S at p.

S P
t
n
kn
C

kg k

Figure 2.8: Definition of normal curvature

As shown in Figure 2.8, the curvatures of a surface S can be quantified in the following
way: We consider a curve C on S which passes through point P as shown in Figure 2.8. t is
the unit tangent vector and n is the unit normal vector of the curve C at point P .
dt
= κn = kn + kg (2.13)
ds
kn = κn N (2.14)

where kn is the normal curvature vector normal to the surface, kg is the geodesic curvature
vector tangent to the surface, and k = κn is the curvature vector of the curve C at point P.
κn is called the normal curvature of the surface at P in the direction t.
From the definition, we have

II = −dr · dN = −(ru du + rv dv) · (Nu du + Nv dv)

= Ldu2 + 2M dudv + N dv 2 (2.15)

where

L = N · ruu , M = N · ruv , N = N · rvv (2.16)

The second fundamental form IIp can be interpreted in the following way: Consider a
regular curve C ⊂ S parametrized by α(s), where s is the arc length of C, and with α(0) = p.
We denote the restriction of the normal vector N to the curve α(s) by N(s). Then we have

< N(s), α00 (s) >= − < N0 (s), α0 (s) > . (2.17)

27
Therefore,

IIp (α0 (0)) = − < dNp (α0 (0)), α0 (0) >

= − < N0 (0), α0 (0) >=< N(s), α00 (s) >
= < N, κn >= κn (p)

Namely, the value of the second fundamental form IIp for a unit vector v ∈ Tp (S) is equal to
the normal curvature of a regular curve passing through p and tangent to v.
Meusnier’s Theorem : All curves lying on a surface S passing through a given point
p ∈ S with the same tangent line have the same normal curvature at this point.
Since N · t = 0, differentiate w.r.t. s
d
(N · t) = N0 · t + N · t0
ds
dt dN dr dN
· N = −t · =− · (2.18)
ds ds ds ds
Recognizing that ds · ds = dx2 + dy 2 + dz 2 = dr · dr, we can rewrite Equation 2.18 as:
dt dr · dN
·N = =−
ds dr · dr
dt
while · N = κn · N ≡ κn
ds

center of curvature N
P

P
center of curvature

(a) (b)

Figure 2.9: Definition of positive normal: (a) κn · N = κn ; (b) κn · N = −κn .

Therefore the normal curvature is given by

II L + 2M λ + N λ2
κn = = (2.19)
I E + 2F λ + Gλ2
dv
where λ = du .
Suppose P is a point on a surface and Q is a point in the neighborhood of P , as in
Figure 2.10. Taylor’s expansion gives
1
r(u + du, v + dv) = r(u, v) + ru du + rv dv + (ruu du2 + 2ruv dudv + rvv dv 2 ) + H.O.T. (2.20)
2

28
 N r=r(u,v)

d
P
Tp

Figure 2.10: Geometrical illustration of the second fundamental form.

Therefore
1
PQ = r(u + du, v + dv) − r(u, v) = ru du + rv dv + (ruu du2 + 2ruv dudv + rvv dv 2 ) + H.O.T.
2
Thus, the projection of PQ onto N
1
d = PQ · N = (ru du + rv dv) · N + II
2
and since ru · N = rv · N = 0, we get
1 1
d = II = (Ldu2 + 2M dudv + N dv 2 )
2 2
We want to observe in which situation d is positive and negative. When d = 0

Ldu2 + 2M dudv + N dv 2 = 0

Solve for du
p √
−M ± (M dv)2 − LN dv 2 −M ± M 2 − LN
du = = dv (2.21)
L L

N N
N

P Tp
Tp P P

Tp

Figure 2.11: (a) Elliptic point; (b) Parabolic point; (c) Hyperbolic point.

• If M 2 −LN < 0, there is no real root. That means there is no intersection between the surface
and its tangent plane except at point P . P is called elliptic point (Figure 2.11(a)).

• If M 2 − LN = 0, there is a double root. The surface intersects its tangent plane with one line
du = − M
L dv, which passes through point P . P is called parabolic point (Figure 2.11(b)).

• If M 2 − LN > 0, there
√ are two roots. The surface intersects its tangent plane with two
−M ± M 2 −LN
lines du = L dv, which intersect at point P . P is called hyperbolic point
(Figure 2.11(c)).

29
2.9 Principal curvatures
For each p ∈ S, there exists an orthonormal basis {e1 , e2 } of Tp (S) such that dNp (e1 ) = −κ1 e1 ,
dNp (e2 ) = −κ2 e2 . Moreover, κ1 and κ2 (κ1 ≥ κ2 ) are the maximum and minimum of the second
fundamental form IIp restricted to the unit circle of Tp (S). That is, they are the extreme values
of the normal curvature at p.
Definition 2.9.1 The maximum normal curvature κ1 and the minimum normal curvature κ2
are called the principal curvatures at p; the corresponding directions, that is, the directions
given by the eigenvectors e1 , e2 , are called principal directions at p.
From this we have the following definition.
Definition 2.9.2 If a regular connected curve C on S is such that for all p ∈ C the tangent
line of C is a principal direction at p, then C is said to be a line of curvature of S.
dκn
The extreme values of κn can be obtained by evaluating dλ = 0 of Equation 2.19, which
gives:
(E + 2F λ + Gλ2 )(N λ + M ) − (L + 2M λ + N λ2 )(Gλ + F ) = 0 (2.22)
Since
E + 2F λ + Gλ2 = (E + F λ) + λ(F + Gλ),
L + 2M λ + N λ2 = (L + M λ) + λ(M + N λ)
equation (2.22) can be reduced to
(E + F λ)(M + N λ) = (L + M λ)(F + Gλ) (2.23)
Thus
L + 2M λ + N λ2 M + Nλ L + Mλ
κn = 2
= = (2.24)
E + 2F λ + Gλ F + Gλ E + Fλ
Therefore κn satisfies the two simultaneous equations
(L − κn E)du + (M − κn F )dv = 0
(M − κn F )du + (N − κn G)dv = 0 (2.25)
These equations can be simultaneously satisfied if and only if
 
 L−κ E M − κn F 
n
=0 (2.26)
 
 M − κn F N − κn G



where | | denotes the determinant of a matrix. Expanding and defining K and H as

LN − M 2
K = (2.27)
EG − F 2
EN + GL − 2F M
H = (2.28)
2(EG − F 2 )
we obtain a quadratic equation for κn as follows:
κ2n − 2Hκn + K = 0 (2.29)
The values K and H are called Gauss (Gaussian) and mean curvature respectively.

30
Definition 2.9.3 Let p ∈ S and let dNp : Tp (S) → Tp (S) be the differential of the Gauss
map. The determinant of dNp is the Gaussian curvature K of S at p. The negative of half of
the trace of dNp is called the mean curvature H of S at p.

The discriminant D can be expressed as follows:

D = H2 − K
(EN + GL − 2F M )2 − 4(EG − F 2 )(LN − M 2 )
=
4(EG − F 2 )2

The denominator is always positive, so we only need to investigate the numerator. The numer-
ator can be written as:

(EN + GL − 2F M )2 − 4(EG − F 2 )(LN − M 2 )

!
EG − F 2 2F
=4 2
(EM − F L)2 + [EN − GL − (EM − F L)]2 ≥ 0
E E

Thus, D ≥ 0.
Upon solving Equation (2.29) for the extreme values of curvature, we have:
p
κmax = H + H2 − K (2.30)
p
κmin = H − H2 −K (2.31)

From Equations (2.30), (2.31), it is readily seen that

K = κmax κmin (2.32)

κmax + κmin
H = (2.33)
2
From Equation (2.27) (since EG − F 2 > 0, see Equation 2.6).

K > 0 ⇒ LN > M 2 ⇒ Elliptic point

K = 0 ⇒ LN = M 2 ⇒ Parabolic point
K < 0 ⇒ LN < M 2 ⇒ Hyperbolic point

Definition 2.9.4 If at p ∈ S, κ1 = κ2 , then p is called an umbilical point of S; in particular,

the planar points are umbilical points.

From this definition, we can conclude that if all points of a connected surface S are umbilical
points, then S is either contained in a sphere or in a plane.

31
 K<0
K=0

K>0

Figure 2.12: Curvature map of a torus showing elliptic, parabolic, and hyperbolic regions.

32
Bibliography

[1] P. M. do Carmo. Differential Geometry of Curves and Surfaces. Prentice-Hall, Inc., Engle-
wood Cliffs, NJ, 1976.

[2] E. Kreyszig. Differential Geometry. University of Toronto Press, Toronto, 1959.

[3] M. M. Lipschutz. Theory and Problems of Differential Geometry. Schaum’s Outline Series:
McGraw-Hill, 1969.

[4] D. J. Struik. Lectures on Classical Differential Geometry. Addison-Wesley, Cambridge,

MA, 1950.

[5] T. J. Willmore. An Introduction to Differential Geometry. Clarendon Press, Oxford, 1959.

33