*1*Def: A parameterized curve in R3 is a smooth function γ :I → R n, where I ⊂ R is an interval.

*1*

An interval means a nonempty connected subset of R which may takes one of these forms ( a , b )
, [ a , b ) , ( a , b ], [ a , b ], ( a , ∞ ) , , [ a , ∞ ) , (−∞, b ), and ¿. We will henceforth use the term curve as
an abbreviation for “parameterized curve”.

*1* def and prop: If γ :I → R n is a curve with components γ ( t ) =( x 1 ( t ) , x 2 ( t ) , x 3 ( t ) ), then its

derivative γ ' : I → R n is defined as γ ' ( t ) =( x 1 ' ( t ) , x 2 ' ( t ) , x 3 ' ( t ) ) given by
γ ( t+ h )−γ ( t )
γ ' ( t )=lim (see fig***). *1*
h →0 h

Zooming enough near γ ( t ) , the object’s path looks like a straight line, and γ ' ( t ) gives the
direction of the that straight line.

*1* DEF: The speed of a curve γ :I → R n at time t is |γ ' ( t )| where |.| is the Euclidean norm. the
t2

arc length between times t 1 and t 2 is ∫|γ ( t )| dt defies the total distance travelled.
'

t1

*1* Def: Let γ :I → R n a curve. It is called regular if its speed is always nonzero (
|γ ' ( t )|≠ 0 , ∀ t ∈ I ). It is called unit-speed or parameterized by arc-length if its speed is always
equal to one.

*ex* γ ( t ) =( t 3 , t 2 ) is an irregular planar curve as |γ ' ( 0 )|=0 at t=0 . This is the graph of the
equation y=x 2 /3 as shown in fig. the graph contains a sharp point at the origin. At this sharp
' γ ( h )−γ ( 0 )
point it is hard to find the straight line at this point through the limit γ ( 0 )=lim .
h→ 0 h

When a regular curve γ :I → R n represents the position function of a moving object, its image
(called the trace of γ ), γ ( I ) ={ γ ( t )∨t ∈ I } represents the path that object follows. Note that γ is
a function, while its trace is a subset in Rn. Trace contains no time information such that many
properties of curves depend on the position of the point on the trace at which we are interested
in not on the time parameterization. For example, when measuring the curvature at some point
of the trace of γ , it depends only on how sharply the trace bends at that point. In other words,
we want our curvature formula to be unchanged by a reparameterization which roughly means
a different association of the time parameter with the points of the trace “change in
parameter”.

*ex* Curves γ ( t ) =(t ,t 2) , t ∈ I =[ −1,1 ], γ ( λ )= ( 2 λ , ( 2 λ )2 ), λ ∈ J = [ −1/ 2,1/ 2 ] have the same

trace a shown in fig***. Set ϕ ( t ) : I → J . Then, γ ( t ) =γ ( ϕ ( t ) )=( γ ∘ ϕ )( t ) ⟹ γ =γ ∘ ϕ
We will refer to γ , γ as two parameterizations of the same path.

*1*def: Suppose that γ :I → R n is a regular curve. A reparameterization of γ is a function of the

form γ=γ ∘ ϕ : I → Rn where I is an interval and ϕ : I → I is a smooth bijection with nowhere-
vanishing derivative (ϕ ' (t )=0 ∀ t ∈ I ).

The requirement of nonzero derivative for ϕ ensures that γ itself is a regular curve. It can be
seen when applying the chain rule on γ=γ ∘ ϕ as follows

' ' ' ∂γ ∂γ ∂ϕ

γ ( t )=γ ( ϕ ( t ) ) ϕ ( t )∨ ( t )= ( ϕ ( t ) )
∂t ∂λ ∂t

Where λ=ϕ ( t ) .

*1* prop: Any regular curve γ :I → R n can be reparameterized by arc length (or a unit-speed
reparameterization of γ .

Proof: choose any t 0 ∈ I and the arc-length function s : I → R defined as

t
s ( t ) =∫ |ϕ ( u )| du
'

t0

Let I denote the image of s. Then s' (t )=|ϕ ' ( t )|≠ 0 . Thus, s is a smooth bijection onto I with
nowhere-vanishing derivative. Therefore, s has an inverse function ϕ : I → I , which is also a
smooth bijection with nowhere-vanishing derivative. Then applying chain rule results in

1
|γ '(t )|=|γ ' ( ϕ ( t ) ) ϕ ' (t )|=|γ ' ( ϕ ( t ) )||ϕ ' ( t )|= |γ ' ( ϕ ( t ) )|=1
s ( ϕ(t ))
'

*1* def: A rigid motion of Rn means a function f : R n → R n that preseves distances, such that
d ( f ( p ) , f ( q ) )=d ( p , q ) for all pairs p , q ∈ Rn . (where d ( p ,q )=‖ p−q‖ is the distance between
points p , q in the Euclidean norm).
n n
*1* def: Let T q ( p )=q+ p represent the translation and the linear transformation L A : R → R
denote the “left multiplication by A ” function, such that L A ( p )= A . p with A ∈ M n (the set of
all n × n real matrices) and p ∈ Rn can be thought as a n ×1 column matrix. In other words, A is
the matrix that represents the linear transformation with respect to the standard orthonormal
basis of Rn . Linear transformation L A is orthogonal if ‖L A ( p)‖=‖ p‖, ∀ p ∈ Rn. In this case,
A ∈O ( n ) , where O ( n ) is the set of orthogonal matrices

*1* prop: If f is a rigid motion of Rn , then it can be written as a composition of translation and
orthogonal linear transformation, f =T q ∘ L A for a unique choice of q ∈ Rn and A ∈O ( n ) .
Surfaces

Topology of subsets of Rn

The collection of all open subsets of Rn is called the topology of Rn . The notion of whether a
subset is closed is topological, while the distance between points of Rn is not topological.

An infinite sequence { p1 , p 2 , … } of points of Rn is said to converge to p ∈ Rn if

1) nlim dist ( p , p n )=0 ,or

→∞

2) For all open sets U , containing p, ∃ N ∈ N , such that pn ∈U ∀ n> N .

From the second definition, the notion of convergence is topological, although it may not
initially seems so at first. If a point belongs to an open set, then it is surrounded by elements of
this set and their sufficiently close neighbors also live in this set.

A point p ∈ Rn is called a limit point of a subset S ⊂ Rn if there exists an infinite sequence of

points of S that converges to p. The collection of limit points of S is the union of S and the
boundary of S. Set S is closed if it contains all of its limit points.

*1* DEF: Let V ⊂ S ⊂ R n be subsets. V is called open (closed) in S if there exist an open (closed)
of Rn whose intersection with S equals V .

Example: setting S2= {( x , y , z ) ∈ R3 ∨x2 + y 2 + z 2=1 }. Notice the hemisphere

{( x , y , z ) ∈ S 2∨z >0 } is open in S2 as it is the intersection of open set {( x , y , z ) ∈ R3∨z >0 }
with S2.

*1* PROP: Let V ⊂ S ⊂ R n be subsets

1) V is open in S iff (if and only iff) ∀ p ∈ V ,∃ r >0 such that { q ∈ S∨dist ( p , q )< r } ⊂ V .

2) V is closed in S iff ∀ p ∈ S that is a limit point of V is contained in V .

For example, the interval [ −1,1 ) is not open or close in R but is open in [ −1,2 ) and closed in
[ −2,1 ).

Def: let p ∈ S ⊂ Rn. A neighborhood of p in S means a subset of S that is open and contains p.

Example ( 1−ε ,1+ ε ) is a neighborhood of 1 in ( 0,2 ) ∀ ε ∈ ( 0,1 ]

[ 0 , ε ) is a neighborhood of 0 in [ 0,1 ] ∀ ε ∈ ( 0,1 ]

The collection of all subsets of S that are open in S is called the topology of S.

Continuity
 n1 n2
Assume subsets S1 ⊂ R and S2 ⊂R . A function f : S1 → S 2 is called continuous if it maps
nearby points to nearby points. A precious definition follows.

Def: A function f : S1 → S 2 is called continuous if for every infinite sequence { p1 , p 2 , … } of

points in S1 that converges to appoint p ∈ S1, the sequence { f ( p1 ) , f ( p2 ) , … } converges to f ( p )
in S2.

Example: the following function f : R → R is not continuous

{
f ( x )= 0 ,∧x ≤ 0
1 ,∧x >0

As {12 , 13 , 14 , …} in the domain of f converges to zero but the sequence of images

{( ) ( ) ( )
f
1
2
=1 , f
1
3
=1 , f
1
4 }
=1 , … converge to 1 rather than f ( 0 )=0.

Another discontinuous function is f : R 2 → R 2 defined as f ( x , y )=( x , x ( 1−| y|) sign ( y ) ), where

2

{
sign ( y )= −1 ,∧x <0
1 ,∧x ≥ 0

Prop: if f : S1 → S 2 is continuous, then for every open (closed) set in S2 , f −1 (U ) is open (closed)
in S1, where f −1 (U ) denotes the set { p ∈ S1 ∨f ( p ) ∈ U }

Example ** fails this continuity test

f
−1
((⏟ ,
)
2 2) ⏟
−1 1
= (−∞ , 0 ]
not open
open

From the last proposition, continuity is a topological concept since this definition involves only
topologies of S1 and S2.

What does it means S1 and S2 are topologically the same. There should be a bijection between
them that pairs of open sets with open sets. More precisely:

Def: a function f : S1 → S 2 is called a homeomorphism if f is bijective and continuous with

continuous inverse, f −1. If such function exist, then S1 and S2 are said to be homeomorphic.

Homeomorphic sets have the same “essential shape” such as the two subsets of R3 SHOWN IN
FIGURE****. The function f : [ 0,2 π ) → S 1 ⊂ R 2 defined as f ( t )=( cos t , sin t ) . It is easy to check
that f is comntinous and bijective, but f −1 is not continuous as an infinite sequence of points in

{( ))}
∞

S {p
1 ∞
}
n n=1 (
1
) (
= cos 2 π − ,sin 2 π −
n
1
n n=1
converges to p= (1,0 ) while sequence
 { }
∞
∞ 1
{f −1
( pn ) }n=1= 2 π−
n
converges to 2 π which not equal to f −1 ( p ) =0. Thus [ 0,2 π ) is not
n=1

homeomorphic to S . It will be seen later in the next section *** that the reason is S1 is
1

compact, while [ 0,2 π ) is not. (continuous function maps compact to compact).

Connected and path-connected

Compact sets

Def : A subset S ⊂ Rn is called bounded if S ⊂B ( p , r ) for some p ∈ Rn and some r >0 . Further,
S is called compact if it is closed and bounded.

Def: Let S ⊂ Rn. An open cover of S is a collection, 𝕆, of sets that are open in S and whose
union contains S. S is called compact if every open cover, 𝕆, of S has a finite subcover
(subcollection) { U 1 , … ,U n } ⊂O whose union contains S.

The last definition make the compactness is a topological notion. Even, boundedness is not
topological notion as it is not preserved by homeomorphisms since the bounded set ( 0,1 ) is
homeomorphic to the unbounded set R .

Examples****

Prop: If S ⊂ Rn compact set and f : S → R m be continuous, then the image f ( S ) is compact.

Proof:***

Prop: If S ⊂ Rn compact set and f : S → R be continuous, then f attains its supremum and
infimum ( f attains its supremum means that it supremum is finite and there is a point p ∈ S for
which f ( p) equals this supremum (because f (S ) is closed).

When henceforth writing f : U ⊂ Rm → R n, we mean that f is a function from the open set U of
Rm into Rn. In this case f can be considered as n separate component functions such that it can
be written as f =( f 1 , f 2 ,… , f n ) where f i : U ⊂ R → R ∀ i ∈ { 1,2, … , n } .
m

For example, function f : R 2 → R 3 defined as f ( x 1 , x 2 ) =(sin ( x 1 ) , e , ( x 1 ) + x 2 ) has

x 1+2 x 2 2

x 1+2 x 2 2
component functions f 1 ( x 1 , x 2 )=sin ( x1 ) , f 2 ( x 1 , x 2 )=e , and f 3 ( x 1 , x 2 ) =( x1 ) + x 2.
 ∂f
The partial derivative of f with respect to the input variable x i (denoted by or f x ) is defined
∂ xi i

∂f f ( p+t e i )−f ( p )
as ∀ p ∈ U f x ( p ) = ( p )=lim (if the limit exists), where e i ∈ Rm denotes
i
∂ xi t→0 t
the i th member of the standard orthonormal basis.

Note that partial derivatives can be computed componentwise as follows

∂f
∂ xi
( p)= (
∂f 1
∂ xi
∂f ∂f
( p ), 2 ( p) , … n ( p)
∂ xi ∂ xi )
Each component has the familiar meaning from multivariable calculus (the rate of change of
that component at p as x i increase while other input variable other than x i as constants).
Second order derivatives can be expressed as:

∂2 f
fx x = =f
i j
∂ xi ∂ x j x j
xi

Function f is called “C r on U ” if all r th-order partial derivative exist and are continuous on U . It
also called “smooth on U ” if f is C r on U for all positive integers r . A function is smooth if its
components are smooth.

*1*1 prop: If g : U ⊂ R l → Rm is smooth on U , f :V ⊂ Rm → R n is smooth on V and g ( U ) ⊂ V ,

then the composition f ∘ g: U ⊂ Rl → R n is smooth on U .

Example: Let f ( x , y )=( sin ( x ) , e x+2 y , x 2+ y ) is smooth as it could be considered as a

composition of two smooth functions (x , y )⟼ ( x , x+2 y , x 2 + y ) and
v
(u , v , w)⟼(sin u , e , w)

Definition of partial derivative in equation *** can be extended by replacing the coordinate
vector e i with an arbitrary vector v ∈ R m; the result is called a directional derivative.

*1* def: Assume f :U ⊂ Rm → R n, a point p ∈U and an arbitrary vector v ∈ R m. The directional

derivative of function f in the direction of v at p (denoted by d f p ( v )) is defined as

f ( p+t v )−f ( p )
d f p ( v )=lim = ( f ∘ γ )' ( t )
t→0 t

(if the limit exists) where γ ( t ) =p+ t v is a straight line in Rm passing p at t=0 with velocity v as
shown in the graph***.
If f is smooth on a neighborhood of p, then the directional derivative
f ( p+ t v ) −f ( p )
f p ( v )=lim exists for all directions v as from the chain rule
t →0 t
' ∂f dγ
d f p ( v )=( f ∘ γ ) ( t ) = ( γ ( t ) ) i ( t ). Where x i are the component functions of γ ( t ) and
∂ xi dt
∂f
γ ' ( t )=v . From smoothness of function f , ∂ x ( γ ( t ) ) exists and the derivative d f p can be
i
L
considered as a linear transformation A , where A is the n × m matrix of all first-order partial
derivatives of f at p:

d f p ( v )=L A ( v )= A . v

Where

[ ]
∂f1 ∂f1
( p) ⋯ (p)
∂ x1 ∂ xm
A= ⋮ ⋱ ⋮
∂f n ∂fn
( p) … (p)
∂ x1 ∂ xm

A is called the Jacobian matrix of f at p. Then function f can be approximated as

f ( p+t v ) ≈ f ( p ) +t f p ( v )=f ( p )+t A . v which can be considered as a good approximation of f
∂fi
near p. The j th column of A is ( p ) for i∈ { 1,2 ,… , n } represents the initial velocity vector of
∂x j
the image under f of p+t e j; a straight line in Rm passing p and pointing in the direction j th
coordinate vector (see fig*****).

Example: assume a smooth function f : R 2 → R 2 defined as f ( r , θ ) =( r cos θ , r sin θ ) where the

coordinate functions of f is x=r cos θ and y=r sin θ (the function express the conversion from

polar to rectangular coordinates). At point p= ( r ,θ )= 4 , ( π3 ), the Jacobian of f at this point

[ ][
∂x ∂x
(p) (p)
∂r
∂y
∂r
(p)
∂θ
∂y
∂θ
(p)
x (p)
= r
yr ( p) yθ ( p)][
xθ ( p )
=
cos θ −r sin θ
sin θ r cos θ p
=√ ] [
3/2 −2
1/2 2 √ 3 ]
The columns of the Jacobian represent the initial velocity vectors of the image under f of the
coordinate lines passing p in the rθ -plane as shown in fig***.

The straight line used in the definition of directional derivative can be replaced with any regular
curve as follow
1*1 prop 3.7: Assume a smooth function f : U ⊂ Rm → R n, a point p ∈U , a vector v ∈ R m, and a
'
regular curve γ with γ ( 0 )= p and γ ' ( 0 )=v , then d f p ( v )=( f ∘ γ ) ( t ) .

Example: Assume f : R 2 → R 3 a smooth function denoted as f ( u , v )=( x ( u , v ) , y (u , v ) , z ( u , v ) ).

Let γ :I → R be a plane curve in uv -plane with component functions denoted by
γ ( t ) =( u ( t ) , v ( t ) ) . Then ( f ∘ γ ) ( t )=d f γ (t ) ( γ ' ( t ) ) , ∀ t ∈ I which can be expanded using the chain
'

rule as follows:

' ∂f dγ
( f ∘ γ ) ( t )= ( γ (t )) i (t)
∂ xi dt

Where x i are the component functions of γ ( t ) and γ ( t )=( u ' ( t ) , v ' (t ) ), then
'

[ ][ ]
xu xv '
' u '
( f ∘ γ ) = yu y v ' =u f u + v ' f v
v
zu zv

*1* Prop: If g : U ⊂ R l → Rm is smooth on U , f :V ⊂ Rm → R n is smooth on V and g ( U ) ⊂V ,

then for all p ∈U , d ( f ∘ g ) p=d f g( p) ∘ d g p or d ( f ∘ g ) p ( w )=d f g ( p ) (v ) where
l
v=d g p ( w ) , ∀ w ∈ R .

The above proposition is a generalization of the chain rule for Euclidean spaces. The right side of
above equation, d f g( p) ∘ d g p , represents the composition of two linear transformation which
can be expressed as a product of two Jacobian matrices.

1*1 Prop: assume that a function f :U ⊂ Rm → R n is smooth on a neighborhood of q ∈ U and

invertible with smooth inverse, f −1, on a neighborhood of q ∈ U . Then its derivative is an
invertible matrix (map).

Proof: From the chain rule d ( f −1 ∘ f ) p=d ( f −1 )f ( p) ∘ d f p . As f −1 ∘ f is the identity function, whose
derivative at every point is the identity map. So ( f −1 )f ( p ) ∘ d f p is the identity linear
transformation, which means that d f p is an invertible linear transformation.

1*1 theorem (inverse function theorem): If If f :U ⊂ Rn → R n is smooth on a neighborhood of

p ∈ Rn and d f p is an invertible linear transformation, then there exist a (possibly smaller)
neighborhood U of p such that V =f ( U ) is a neighborhood of f ( p ) , and f :U → V is invertible
with smooth inverse.

We can see this theorem in the first-order approximation of f near p, f ( p+t v ) ≈ f ( p ) +t f p ( v ).

Theorem say that if this first-order approximation is invertible (between neighborhoods), then f
is invertible (between neighborhoods).
 (
f ( x , y )= ⏟
xy , x⏟
)
2
Example: Assume a smooth function f : R 2 → R 2 defined
+ y then the Jacobian
f1 f2

of function f at a point p= (1,1 ) is

[ ]
∂f1 ∂f 1
( p) ( p)
∂x
∂f2
( p)
∂y
∂f 2
(p)
[
=
y x
2x 1 ] =[21 11]
( 1,1 )
∂x ∂y

The determinant of the Jacobian f at ( 1,1 ) is nonzero, so the d f p is invertible. Using the inverse
function theorem implies that the restriction of f to a sufficiently small neighborhood of ( 1,1 ) is
an invertible function with smooth inverse between this neighborhood and its image (a
neighborhood of f ( 1,1 ) =( 1,2 )). In other words, instead of solving the system
{ f 1=xy , f 2 =x2 + y } for x and y in terms of f 1 and f 2 near x=1 , y=1 which in this problem is
length, we can use the inverse theorem to prove the existence of smooth inverse for f near
x=1 , y=1.

Def 3.12: if X ⊂ Rm is a (not necessary open) set, then f : X → Rn is called smooth if ∀ p ∈ X ,

there exists a neighborhood U of p in Rm and a smooth function f :U → Rn that agrees with f
on X ∩U .

*1* def: X ⊂ Rm and Y ⊂ R m are called diffeomorphic if there exist a smooth bijection
f : X → Y whose inverse is also smooth and f is called diffeomorphism.

Example

Figure 3.6 give three examples important

Regular surfaces

A regular surface means a subset of R3 that ”locally looks like” the plane R2. Sufficiently small
and nearsighted inhabitants of a regular surface would believe they live on a plane.

Def: A subset S ⊂ R3 is called a regular surface if each of its points has a neighborhood in S that
is diffeomorphic to an open set in R2. That is, for every p ∈ S, there exist a neighborhood V of
p in S, an open set U ⊂ R2, and a diffeomorphism σ :U →V . Such a diffeomorphism σ is called
a surface patch or a coordinate chart. A collection of surface patches that together cover all the
points of S is called atlas for S.

Definition of a surface is obtained by replacing diffeomorphism by homomorphism. Regular

surface means surface patches are smooth and have smooth inverses., so they are
diffeomorphisms rather than only homomorphisms. For example cone surface is not regular. It
is not differ as it is not diffeomorphic to an open set in R2 as it is not differentiable at origin.
Inhabitants of the cone living at the origin, no matter how small and nearsighted could observe
their world is different from the plane. Differential geometry involves doing calculus on surfaces,
so we will restrict our study on regular surfaces.

Assume the coordinate variables of R2 as { u , v } , so that our surface patch σ :U →V ⊂ S can be

written in components as σ ( u , v ) = ( x ( u , v ) , y ( u , v ) , z ( u , v ) ) . Its Jacobian matrix at
q=( u 0 , v 0) ∈ U is

[ ]
∂x ∂x
(q ) (q )
∂u ∂v
∂y ∂y
d σq= (q ) (q)
∂u ∂v
∂z ∂z
(q ) (q)
∂u ∂v

The Jacobian matrix columns represents the partial derivatives of σ with respect to u and v at q
in components form. Denoting them by σ u ( q ) and σ v ( q ) such that

σ u ( q )= ( ∂∂ux ( q ) , ∂∂ uy ( q ) , ∂∂ uz ( q )) and σ ( q )=( ∂∂ xv ( q) , ∂∂ vy ( q ) , ∂∂ vz ( q) ). As stated in the previous

v

section σ u ( q ) and σ v ( q ) are the initial tangent vectors to the images under σ of the coordinate
lines through q in R2 as shown in figure3.9. Then

σ u ( q )=d σ q ( e1 ) =( σ ∘ γ 1 )' ( 0 ) where γ 1 ( t )=q+t e 1=( u0 +t , v 0 ) ,

'
σ v ( q )=d σ q ( e2 ) =( σ ∘ γ 2) ( 0 ) where γ 2 ( t )=q+ t e 2=( u0 , v 0 +t ) .

The variables u and v are called local coordinates on V . Inhabitants of V could uniquely identify
their location from their u and v values. An Inhabitant living at σ ( q ) can move in the direction of
σ u ( q ) along the red coordinate curve or in the direction of σ v ( q ) along the green coordinate line.
The next proposition ensures that thses two possible directions are linearly independent. In
other words, it ensures that the local coordinate system does not degenerate anywhere on V .

Prop 3.21: if S is a regular surface, and σ :U ⊂ R2 → V ⊂ S is a surface patch, then ∀ q ∈U ,

2 3
the linear transformation d σ q : R → R has rank 2 or σ u ( q ) and σ v ( q ) are linearly independent.

Proof: Assume that σ −1 is the inverse function of σ restricted only on the neighborhood of p in
R3 which is a smooth function. From chain rule d ( σ ∘ σ )q =d ( σ ) σ ( q) ∘d σ q is the derivative of
−1 −1

an identity function which is the identity linear transformation on R2, which has rank 2. As the
rank of a composition of two linear transformations is less than or equal to the rank if the first,
we get d σ q has rank 2.
Example cylinder – sphere –

Def: Let U ⊂ R3 be open, f :U → R be smooth, and λ ∈ R be in the image of f . The set

f −1 ( λ )= { p ∈U ∨f ( p ) =λ } is called the preimage of λ . λ is called a regular value of f if
∀ p ∈ f −1 ( λ ), the Jacobian matrix for f (the gradient of function f ) at p is not zero matrix.

Example (temperature)

Theorem: If λ is aregular value of function f then the preimage f −1 ( λ ) is a regular surface.

The preimage f −1 ( λ ) is called a level set of f . In case that λ is a regular value of f , f −1 ( λ ) is

called level surface.

Example: consider the function f ( x , y , z )=x 2 + y 2−z 2. The Jacobian matrix of f at p= ( x , y , z )

is d f p =( 2 x , 2 y ,−2 z ) which vanishes only origin. The origin belongs toe the cone-shaped level
set f =0 which I not a regular surface. While other level sets with f ≠ 0 are regular surfaces. For
example the level surface f =1 is a hyperboloid of one sheet and the level surface f =−1 is a
hyperboloid of two sheets.

Def: a parameterized surface is a smooth surface σ :U → R 3 such that ∀ q ∈U , d σ q has rank 2.

Note that a regular surface is a set while a parameterized surface is a function. Every surface
patch for a regular surface is a parameterized surface. A parameterized surface need not to be
one-to-one.

Example: a smooth surface function σ : R2 → R3 defined as σ ( u , v ) =( cos u , sinu , v ) is a

parameterized surface that warps the plane infinitely many times around the cylinder. The
restriction of this function to a small domain (−π , π ) × R like used in example*** can be
considered as a surface patch.

Prop: if σ : U → R 3 is a parameterized surface, then ∀ q 0 ∈U , there exist an open set U ∈U

containing q 0 such that the image S=σ ( U ) is a regular surface covered by a single patch
represented by the restriction of σ to U .

Proof: let q 0 ∈ U . Choose any vector N ∈ R3 with N ∉ span {σ u ( q 0 ) , σ v ( q 0 ) } (normal to the

surface at σ ( q0 )). Consider the smooth function f : U × R → R3 defined as f ( q ,t )=σ ( q ) +t N .
3 3
Set p0=σ ( q 0 )=f (q 0 , 0). The derivative d f ( q , 0) :R → R is invertible because it sends the basis
0

{e 1 , e2 , e3 } to { σ u ( q0 ) , σ v ( q0 ) , N }. By the inverse function theorem, f restricts to a

diffeomorphism from a neighborhood, A , of ( q 0 , 0 ) in U × R ⊂ R2 × R, say U × (−ε , ε ) where
ε > 0 and U ⊂U is aneighborhoof of q 0, to a neighborhood, B, of p0 in R3 . Defining S=σ ( U ).
Since f is injective on 𝒜, σ is injective on U and it has an inverse σ −1 : S →U . To prove the
smoothness of σ −1, assume P : A →U denotes the natural projection that maps ( q ,t ) ⟼ q .
Then, function P ∘ f −1 : B →U is a smooth function that agree with σ −1 on B∩ S . Thus, σ −1 is
smooth.

From above proposition, the image of every injective parameterized surface is a regular surface
and it can be concluded that it does not matter whether to work with regular surfaces or
parameterized surfaces.

Tangent planes

Def: A regular curve in a regular surface S is defined as a regular curve in R3 whose trace is
contained in S. The tangent plane to S at the point p ∈ S is the set of all initial velocity vectors
of regular curves in S with initial position p defined as

T p S={ γ ' ( 0 )∨γ is aregular curve∈S with γ ( 0 ) =p }

You can consider the domain of such a curve γ to be (−ε , ε ) for ε > 0. A vector is called “tangent
to S at p” when it lies in T p S . The tangent plane to S at p, T p S can be characterized in terms
of a coordinate chart σ :U ⊂ R2 → V ⊂ S with p ∈V . Setting q=σ −1 ( p ) and { u , v } the
coordinate variables of U , we get T p S=span {σ u ( q ) , σ v ( q ) }. In particular T p S is a two-
dimensional subspace of R3.

Proof:

The previous definition the regular curve in S can be replaced with regular curve in V . As
σ : U →V is a diffeomorphism, the regular curves in V with initial position p are considered as
the compositions with σ of the regular curves in U with initial position q . From prop 3.7, we get

span { σ u ( q ) , σ v ( q ) }=image ( d σ q ) ={( σ ∘ β ) ( 0 )∨β is a regular curve∈U with β ( 0 )=q }= {γ ( 0 )∨γ is a regular curve
' '

Figures*****

Notice that T p S does not depend on the choice of the surface patch.

Def: Assume a smooth function f : S → R n where S is a regular surface. The derivative of f at

p ∈ S is the linear transformation d f p :T p S → R n defined so that
∀ v ∈T p S , d f p ( v )=( f ∘ γ )' ( 0 ), where γ is any regular curve in S with γ ( 0 )= p and γ ' ( 0 )=v.

As f is smooth, there is a neighborhood, O , of p in R3 and a smooth function f :O → R n that

agrees with f on (smoothness of f from the Def 3.12) (extending function from not Euclidean to
Euclidean) . For every v ∈ R 3, prop 3.7 (which apply only on Euclidean geometry) says that
'
d f p ( v )=( f ∘ γ ) ( 0 ) , where γ is any regular curve in R3 with γ ( 0 )= p and γ ' ( 0 )=v . If v ∈T p S ,
then there exist regular curves in S that satisfy these conditions. Since f restricted to a regular
' '
curve γ in S is identical to f (or f ∘ γ =f ∘ γ ), then d f p ( v )=( f ∘ γ ) ( 0 )= ( f ∘ γ ) ( 0 )=d f p ( v ). Thus,
d f p is identical to the well-defined linear transformation d f p : R3 → R n restricted to the domain
T p S.

Prop: if f : S1 → S 2 is a smooth function between a pair of regular surfaces, then for every p ∈ S1
, d f p :T p S1 →T f ( p ) S 2.

It means that the image of d f p lies in T f ( p) S 2, because f send curves in S1 to curves in S2.

Theorem (the inverse function theorem for surfaces): if f : S1 → S 2 is a smooth function between
a pair of regular surfaces. For p ∈ S1, if d f p :T p S1 →T f ( p ) S 2 is invertible, then f restricted to a
suffieciently small neighborhood of p in S1 is a diffeomorphism onto its image, and this image is
a neighborhood of f ( p) in S2

Proof: assume coordinate charts σ 1 :U 1 →V 1 and σ 2 :U 2 →V 2 where V 1 is a neighborhood of

p in S1, while V 2 is a neighborhood of f ( p ) in S2. They are sufficiently small such that
f ( V 1) ⊂V 2. Defining ψ :U 1 → U 2 as following composition

ψ=σ 2−1 ∘ f ∘ σ 1
−1 −1
f and ( σ 2 ) are smooth functions which is an extension to smooth functions f and ( σ 2 ) on
open neighborhoods in R3 of p and f ( p ) respectively (see def 3.12) such that f ∘ σ 1=f ∘ σ 1 and
−1 −1
( σ 2 ) ∘ f ∘ σ 1=σ 2 ∘ f ∘ σ 1.
−1 −1
Applying chain rule twice on ψ=σ 2 ∘ f ∘ σ 1 not σ 2 ∘ f ∘ σ 1 as chain rule can be applied on
open of Euclidean spaces, we get d ψ q=d ( σ 2 )f ( p ) ∘ d f p ∘ d ( σ 1 )q
−1

As stated before in the previous definition d f p ( v )=d f p (v ) and d ( σ 2 )f ( p) ( w )=d ( σ 2 ) f ( p) (w)

−1 −1

for all v ∈T p S 1 and w ∈ T f ( p ) S2. Then d ψ q=d ( σ 2 )f ( p ) ∘ d f p ∘ d ( σ 1 )q

−1

The following diagram shows the functions in the left side and the derivatives at the relevant
points in the right side.

f dfp
V 1 ⊂S 1 V 2 ⊂ S2 T p S1 T f ( p ) S2
⟶ ⟶
↑ d ( σ 1 )q ↑ d ( σ 2 )ψ (q) d ψ q
↑ σ1 ↑ σ2 ψ
¿ ⟶ U2 ¿ R
2
¿ ¿ ¿ ⟶
¿
Then, d ψ q is a composition if three linear transformations, and we argue that all three are
2
invertible. d ( σ 1 ) q : R →T p S 1 is invertible by prop 3.21. and d f p :T p S1 →T f ( p ) S 2 is invertible
−1
by hypothesis. d ( σ 2 )f ( p) : T f ( p) S 2 → R2 is invertible because it equals to ( d ( σ 2 )ψ (q ))
−1
, so it is the
inverse of an invertible linear transformation. Then d ψ q is invertible between two-dimensional
vector spaces. Applying the inverse function theorem after shrinking U 1 and U 2 to smaller
neighborhoods of q and ψ ( q ) , respectively, ψ is adiffeomorphism. And
−1
f =σ 2 ∘ψ ∘ σ 1 : V 1 → V 2 is a composition of three diffeomorphism, so it is a diffeomorphism.

Area distortion