Integrating over Curves

(by Christian Gorski )

Definition 1. A curve is a continuous function γ : [a, b] → Rn defined on a closed interval
[a, b] ⊂ R.
• If the derivative γ 0 of γ exists and is continuous, then we say that γ is C 1 . We may
refer to γ 0 (t) as the tangent vector to γ at a the point γ(t).
• γ is piecewise C 1 if there exist finitely many parameters a = a0 < a1 < · · · < ak =
b ∈ R such that γ is C 1 on [ai−1 , ai ] for all 1 ≤ i ≤ k.
• γ is simple if (γ(s) = γ(t)) ⇒ (s = t or {s, t} = {a, b}) (that is, γ is one-to-one,
except that its endpoints may be equal).
• γ is closed if γ(a) = γ(b).
Definition 2. Let γ : [a, b] → Rn be a C 1 curve. We say γ̃ is a reparametrization of
γ if there exists a diffeomorphism Φ : [c, d] → [a, b] such that γ̃ = γ ◦ Φ. We say that Φ is
orientation-preserving if it is increasing, and we say that it is orientation reversing
if it is decreasing.
Now we introduce the first of two ways of integrating over curves: integration with respect
to arc-length.
Definition 3. Given a C 1 curve
γ : [a, b] → Rn
and a continuous function
f : Rn → R,
the integral of f with respect to arc-length on γ is the quantity
Z Z b
f ds := f (γ(t))kγ 0 (t)kdt.
γ a

Remark. Recall that

Z b Z
0
Length(γ) = kγ (t)kdt = 1ds.
a C

Theorem 4. Let γ : [a, b] → Rn be a C 1 curve and f : Rn → R be continuous. Let γ̃ be a

reparametrization of γ. Then Z Z
f ds = f ds
γ γ̃

Note that the arc-length integral is the same, regardless of whether the reparametrization
reverses or preserves orientation. This will not be the case for the next type of integration
we consider.
Before we introduce the second (and more important) type of integration, we must in-
troduce the concept of a covector.
Covectors
Definition 5. A covector on Rn is a linear function ω : Rn → R.

Suppose x = (x1 , . . . , xn ) ∈ Rn is a vector. Then for each 1 ≤ j ≤ n we have a covector

dxj : Rn → R given by
dxj (x1 , . . . , xn ) = xj
We will call dx1 , . . . , dxn the “standard basis covectors.”
The set of all covectors ω : Rn → R is denoted (Rn )∗ .

Proposition 6. Any covector ω : Rn → R may be written uniquely as

n
X
ω= ωi dxi
i=1

where each ωi ∈ R is a scalar.

Remark.

• Covectors are also called “linear forms” or “linear functionals.”

• For ω : Rn → R, x ∈ Rn one often writes hω|xi := ω(x).

ωi dxi : Rn → R as the 1 × n matrix

P
• We can alternatively think of a covector ω =
[ω1 , . . . , ωn ] (a “row vector”).

• Using the dot product can confuse a covector with a vector;

     
x1 ω1 x1
· · · ωn  ...  =  ..  ·  .. 
X  
ω(x) = ωi xi = ω1
 
.   . 
xn ωn xn

Definition 7. Let T : Rn → Rm be a linear transformation and ω : Rm → R be a covector.

Then the pullback of ω is the covector T ∗ : Rn → R given by

T ∗ (ω)(x) = ω(T (x))

for all x ∈ Rn .

Line Integrals
Definition 8. Let U ⊂ Rn be open. A differential 1-form is a function

ω : U → (Rn )∗ .
Remark. More concretely, ∀x ∈ U ,
X
ω(x) = ωi (x)dxi

where ωi : U → R are scalar-valued functions. We will often refer to differential 1-forms as

simply 1-forms. We will always assume ω is continuous(i.e. that each ωi is continuous), and
often assume that ω is C 1 (i.e. that each ωi is C 1 ).

Definition 9. Let U ⊂ Rn be open, ω be a continuous 1-form on U and γ : [a, b] → U be a

C 1 parametrized curve. Then the line integral of ω over γ is the quantity
Z Z b
ω := hω(γ(t)), γ 0 (t)idt
γ a

Theorem 10. Let γ : [a, b] → Rn be a C 1 curve and γ̃ : [c, d] → Rn be a reparametrization.

Then, for any continuous 1-form ω on Rn , we have
Z Z
ω= ω
γ̃ γ

if γ̃ preserves orientation, and Z Z

ω=− ω
γ̃ γ

if γ̃ reverses orientation.

Note that, unlike in the case of arc length integrals, the orientation of the parametrization
does matter to the value of the integral.

Closed and Exact 1-Forms

Definition 11. If f : U → R is a C 1 function on an open set U ⊂ Rn , then the 1-form
n
X ∂f
df := dxi
i=1
∂xi

is called the differential of f .

Definition 12. A 1-form ω : U → (Rn )∗ on an open set U ⊂ Rn is called exact if there

exists a C 1 function f : U → R such that ω = df . In this case, the function f is called a
potential for ω.

Note that potentials are not unique; in fact, given a potential function f for some 1-form
ω, we can obtain another potential f˜ for ω by simply adding a constant—i.e. by defining

f˜(x) = f (x) + c

for some constant c ∈ R.

Theorem 13. (The Fundamental Theorem of Calculus for Curves) Let U ⊂ Rn be an open
set, and let [a, b] ⊂ R be a compact interval. If γ : [a, b] → U is a C 1 curve and f : U → R
is a C 1 function, then
Z
df = f (γ(b)) − f (γ(a)).
γ

Note that this theorem implies that the line integral of a differential over a curve is
path-independent; that is, the integral depends only on the endpoints and orientation of the
curve, and not the path itself.
Note also that this theorem implies that the integral of an exact 1-form over a closed
curve is zero. In fact, we have the following set of equivalences:
Theorem 14. Let ω be a continuous 1-form defined on an open set U ⊂ Rn . Then the
following statements are equivalent:
1. ω is exact.
R
2. γ ω is path independent for any curve γ.
R
3. γ ω = 0 for any closed curve γ.
A related notion is that of a closed 1-form:
Definition 15. A C 1 1-form ω = ωi dxi defined on U ⊂ Rn is closed if
P

∂ωi ∂ωj
= for all 1 ≤ i, j ≤ n.
∂xj ∂xi
Note that, since partial derivatives commute, a C 1 and exact 1-form is always closed.
The next theorem gives a condition under which exactness is equivalent to being closed; to
state it, we need one more definition:
Definition 16. A subset U ⊂ Rn is star-shaped if there exists x ∈ U such that, for any
y ∈ U , the line segment from x to y is contained in U.
Theorem 17. Let U ⊂ Rn be star-shaped, and let ω be a C 1 1-form on U. Then ω is exact
if and only if ω is closed.
Rn , rectangles and convex sets are examples of star-shaped sets. However, the set R2 −{0}
is not star-shaped; this allows for the interesting 1-form utilized in the next definition:
Definition 18. Let γ : [a, b] → R2 − {0} be a C 1 closed curve, and set
−ydx + xdy
ω= .
x2 + y 2
1
R
Then the winding number (about the origin) of γ is the quantity 2π γ
ω.
Note that ω is closed, but not exact, and so can have a nonzero line integral over a closed
curve. As we saw in class, the 1-form ω measures the angle a curve turns through; the
winding number is an integer which gives the number of times the curve winds about the
origin.
Pullbacks of 1-forms
Definition
Pn 19. Let U ⊂ Rk , V ⊂ Rn be open. Given a C 1 function Φ : U → V and a 1-form
ω = i=1 ωi di on V, the pullback Φ∗ ω of ω by Φ is the 1-form on U given by
n
X
∗
Φ ω := (ωi ◦ Φ)dΦi
i=1

where Φi is the ith component of Φ.

Suppose that γ : [a, b] → Rn is a C 1 curve and ω = ωi is a C 1 1-form on Rn . Then

P

   
x1 γ1 (t)
x =  ...  = γ(t) =  ... 
   
xn γn (t)

and so X X
γ ∗ω = (ωi ◦ γ)dγi = (ωi ◦ γ)γi0 (t)dt = hω(γ(t)), γ 0 (t)idt
i i
Rb
γ ∗ ω.
R
and therefore γ
ω= a

Green’s Theorem
Definition 20. Let ω = ωx dx + ωy dy be a C 1 1-form on an open subset U of R2 . Then the
exterior derivative of ω is the expression
 
∂ωy ∂ωx
dω := − dxdy.
∂x ∂y

Theorem 21. (Green’s Theorem) Let Ω ⊂ R2 be a region such that ∂Ω is a finite union of
piecewise C 1 simple closed curves; let ω be a C 1 1-form on an open set U ⊃ Ω. Then
Z Z
ω= dω
∂Ω Ω

where ∂Ω is oriented positively relative to ω—i.e. each curve in ∂Ω is parametrized so that

Ω lies counterclockwise from (i.e. to the left of ) the parametrization’s tangent vector at each
point in the curve.

Cross Products
Now we turn to the task of defining the integral over a surface. Our first definition will be
analogous to the arc-length integral over a curve, and will be based on the idea of surface
area. However, to develop a concept of surface area, we must first define the cross product.
Definition 22. The cross product of two vectors x = (x1 , x2 , x3 ), y = (y1 , y2 , y3 ) ∈ R3 is
the vector in R3 given by

x2 y2 x 1 y1 x1 y 1
x × y := e1 − e2 + e3 .
x3 y3 x3 y 3 x2 y 2

Remark. Consider the formal determinant

e1 x1 y1
e2 x2 y2 .
e3 x3 y3

Although this is not the determinant of a real matrix, it is a helpful mnemonic device; if
we perform “cofactor expansion” on the first column, we obtain the formula for the cross
product of two vectors (x1 , x2 , x3 ), (y1 , y2 , y3 ) ∈ R3 .

Proposition 23. (Basic Properties of Cross Products) Given u, v, w ∈ R3 , c ∈ R, we have:

1. u · (v × w) = det[u v w]

2. v × w = −w × v

3. v × w ⊥ span{v, w}

4. v × v = 0

5. (cv) × w = c(v × w)

6. (u + v) × w = (u + v) × (u + w)

7. kv × wk2 + (v · w)2 = kvk2 kwk2

Remark. In the above proposition, (2) - (6) follow fairly directly from (1). (7) can be
shown by brute force calculation, but its geometrical motivation will be discussed shortly.
Recall that u · (v × w) = det[u v w] is the three-dimensional area of the parallelotope
determined by u, v, w. If u is a unit vector orthogonal to v and w, then the parallelotope
is a prism whose base is the 2-dimensional parallelogram determined by v and w and whose
height is 1. We would then expect the volume of this parallelotope to be the area of the
base parallelogram. But since v × w is also orthogonal to v and w, it is parallel to the unit
vector u, and we have

det[u v w] = u · (v × w) = kukkv × wk = kv × wk.

This motivates the following definition:

Definition 24. The two-dimensional area of the parallelogram P (v, w) determined by the
vectors v, w ∈ R3 is the quantity kv × wk.
 Remark. Recall that we defined the cosine of the angle θ between two vectors v and w
by the equation
v · w = kvkkwk cos θ.
Moreover, we know from geometry that the area of the parallelogram spanned by v, w should
be kvkkwk sin θ; combining with Definition 24, we should then have
kv × wk = kvkkwk sin θ.
Substituting these equations into (7) from Proposition 23, we get
kvk2 kwk2 sin2 θ + kvk2 kwk2 cos2 θ = kvk2 kwk2 ,
which is consistent with a familiar and fundamental trigonometric identity. This justifies
our defining sin θ and cos θ in this way.
Now that we have a convenient formula for the 2-dimensional area of a parallelogram in
3
R , we can define the first notion of integration over a surface.

Integrating over Surfaces

Definition 25. Let U ⊂ R2 be open and let G : U → Rn be a C 1 function such that the
following hold:
• G is one-to-one (i.e. G(a) = G(b) ⇒ a = b ∀a, b ∈ U ).
• For all a ∈ U , DG(a) has rank 2 (i.e. the columns of DG(a) are linearly independent).
Then the image G(U ) ⊂ Rn is called a (smooth) parametrized surface. If Ũ ∈ R2 is
another open set and Φ : Ũ → U is a diffeomorphism, then we call G̃ = G ◦ Φ : Ũ → R3 a
reparametrization of G(U ).
Remark. For the purposes of this class, we will only consider surfaces G(U ) ⊂ R3 .
Definition 26. Let U ⊂ R2 be open, and let G : U → R3 , (u, v) 7→ (x, y, z) be a parametrized
surface. If Ω ⊂ U is a region, we define the surface area of G(Ω) to be the quantity
Z Z
∂G ∂G
dA := × dV2 .
G(Ω) Ω ∂u ∂v
Moreover, if f : R3 → R is continuous, then we define the integral of f over Ω with
respect to surface area to be the quantity
Z Z
∂G ∂G
f dA := f ◦G × dV2 .
G(Ω) Ω ∂u ∂v
Theorem 27. Let U, Ũ ⊂ R2 be open, with a diffeomorphism Φ : Ũ → U . Let Ω ⊂ U ,
Ω̃ ⊂ Ũ be regions such that Ω = Φ(Ω̃). If G : U → R3 is a parametrized smooth surface and
G̃ = G ◦ Φ : Ũ → R3 is a reparametrization of G(U ), then
Z Z
∂ G̃ ∂ G̃ ∂G ∂G
× dV2 = × dV2 .
Ω̃ ∂ ũ ∂ṽ Ω ∂u ∂v
Note that Theorem 27 simply states that surface area is independent of parametrization,
as we would hope.