Math 53M, Fall 2003 Professor Mariusz Wodzicki

Introduction to differential 2-forms

January 7, 2004

These notes should be studied in conjunction with lectures.1

1 Oriented area Consider two column-vectors

   
v11 v12
v1 = and v2 = (1)
v21 v22

anchored at a point x R2 . The determinant

 
v11 v12
(x; v1 , v2 ) det = v11 v22 v21 v12 (2)
v21 v22

equals, up to a sign, the area of the parallelogram spanned by v1 and v2 . We will denote

7
 two ways of ordering a pair of vectors v1

and v2 correspond to two ways of orient-
x (v1 , v2 ) 
ing parallelogram x (v1 , v2 ): if v1 comes
 v1


 first then one traverses the boundary of

 x (v1 , v2 ) by following the direction of v1 ;
 
v2 x if v2 comes first then one follows the di-
rection of v2 .
this parallelogram by x (v1 , v2 ) and call quantity (2) its oriented area.
Note the following properties of :
(a) Linearity in each of its two column-vector variables:

(x; au + bv, w) = a(x; u, w) + b(x; v, w) (3)

(x; u, av + bw) = a(x; u, v) + b(x; u, w) (4)
Abbreviations DCVF and LI stand for Differential Calculus of Vector Functions and Line Integrals, re-
1

spectively.

1
 Math 53M, Fall 2003 Professor Mariusz Wodzicki

(b) Antisymmetry: (x; v, u) = (x; u, v) ,

(u, v and w being column-vectors and a and b being scalars).

2 Differential 2-forms Any function : D Rm Rm R satisfying the above two

conditions will be called a differential 2-form on a set D Rm . By contrast, differential
forms of LI will be called from now on differential 1-forms.

3 Exterior product Given two differential 1-forms 1 and 2 on D, the formula

 
1 (x; v1 ) 1 (x; v2 )
(x; v1 , v2 ) det (5)
2 (x; v1 ) 2 (x; v2 )
gives us a differential 2-form. We denote it 1 2 and call it the exterior product of
1-forms 1 and 2 .
Note that

2 1 = 1 2 . (6)

Indeed,
 
2 (x; v1 ) 2 (x; v2 )
(2 1 )(x; v1 , v2 ) = det
1 (x; v1 ) 1 (x; v2 )
 
1 (x; v1 ) 1 (x; v2 )
= det = (1 2 )(x; v1 , v2 ) .
2 (x; v1 ) 2 (x; v2 )

In particular, for any 1-form one has

=0 . (7)

W Exercise 1 Verify that for any differential 1-forms , , and 2 scalars a and b, one has:

(a1 ) (a + b) = a + b ;
(a2 ) (a + b) = a + b .
2
Greek letter is called khee while letter is called ypsilon.

2
Math 53M, Fall 2003 Professor Mariusz Wodzicki

4 Example Let us calculate df1 df2 where f1 and f2 are two functions D R on a
subset of R2 . We have
f1 f1
df1 = dx1 + dx2
x1 x2
f2 f2
df2 = dx1 + dx2
x1 x2
(it is more instructive to use notation x1 and x2 instead of x and y), and
   
f1 f1 f2 f2
df1 df2 = dx1 + dx2 dx1 + dx2
x1 x2 x1 x2
f1 f2 f2 f1
= dx1 dx2 + dx2 dx1 (since dxi dxi = 0)
x1 x2 x1 x2
 
f1 f2 f1 f2
= dx1 dx2 (since dx2 dx1 = dx1 dx2 )
x1 x2 x2 x1

= (det Jf (x)) dx1 dx2 . (8)

 
f1
where f denotes the vector function D R2 having f1 and f2 as its components.
f2

5 dx dy Note that
 
v11 v12
dx dy (x; v1 , v2 ) = det (9)
v21 v22
which is the right-hand-side of (2) and, up to a sign, the area of parallelogram formed by
column-vestors v1 and v2 at point x R2 . We call the differential 2-form on R2 , dx dy,
the oriented-area element.

6 Basic differential forms dxi dxj Differential forms dxi dxj , i 6= j, on Rm are called
basic differential 2-forms. What is their meaning?

u1 v1
If u = ... and v = ... , then

um vm
 
ui vi
dxi dxj (x; u, v) = det (10)
uj vj
3
Math 53M, Fall 2003 Professor Mariusz Wodzicki

which is the (oriented) area of the parallelogram

u, v)
x ( (11)

where the column-vectors

   
ui vi

u and v , (12)
uj vj

and the point  

xi

x (13)
xj
are projections of column-vectors u and v, and point x, respectively, onto the plane R2xi xj
spanned by xi - and xj -axes.3

7 2-forms on R2 Let be any differential 2-form on a set D R2 . For a pair of

column-vectors

v1 = v11 i + v12 j (14)

v2 = v21 i + v22 j (15)

to calculate value (x; v1 , v2 ) we plug first (14) and use Property (a1 ) from Exercise 1:

(x; v1 , v2 ) = (x; v11 i + v12 j, v2 ) = v11 (x; i, v2 ) + v12 (x; j, v2 ) , (16)

and then plug (15) into the right-hand-side of (16) and use Property (a2 ) from the same
exercise:

= v11 (v21 (x; i, i) + v22 (x; i, j)) + v12 (v21 (x; j, i) + v22 (x; j, j))
= (v11 v22 v21 v12 ) (x; i, j)
= (x; i, j) (dx dy)(x; v1 , v2 )
= ((x; i, j) dx dy)(x; v1 , v2 ) . (17)
3
In other words, parallelogram x ( ) is obtained by projecting parallelogram x (u, v) onto xi xj -plane.
u, v

4
 Math 53M, Fall 2003 Professor Mariusz Wodzicki

In other words, any differential 2-form on a subset of R2 can be represented as a multiple

of the oriented-area element:

= f dx dy where f(x) (x; i, j) . (18)

The function-coefficient f in (18) is, for obvious reasons, denoted

. (19)
dx dy

8 2-forms on R3 A similar, completely straightforward, calculation shows that any 2-

form on a subset D R3 can be represented as

= f1 dy dz + f2 dz dx + f3 dx dy (20)

or,

= f1 dx2 dx3 + f2 dx3 dx1 + f3 dx1 dx2 (21)

if one uses notation x1 , x2 , x3 instead of x, y, z.

W Exercise 2 Verify that

f1 (x) = (x; j, k) , f2 (x) = (x; k, i) and f3 (x) = (x; i, j) . (22)

A very important observation follows from formulae (2021):

on subsets of R3 , and of R3 only, both differential 1-forms and differential

. (23)
2-forms are given in terms of three function-coefficients f1 , f2 and f3

5
Math 53M, Fall 2003 Professor Mariusz Wodzicki

9 Area element The function that associates with a pair of column-vectors v1 and v2
anchored at apoint x Rm , the area of parallelogram x (v1 , v2 ) will be called the area
element and denoted .
We already know that in R2 the area element coincides with the absolute value of basic
differential 2-form
= |dx1 dx2 | = |dx dy| . (24)
In general, for vectors in Euclidean space Rm , we have the formula

 p
(x; v1 , v2 ) = kv1 k kv2 k  sin ] vv21  = (kv1 k kv2 k)2 (1 cos2 (] vv21 ))


(kv1 k kv2 k)2 (v1 v2 )2

p
= (25)

Let us see what does this formula look like in R3 . We have:

(kv1 k kv2 k)2 = (v211 + v212 + v213 )(v221 + v222 + v223 ) (26)

and
(v1 v2 )2 = (v11 v21 + v12 v22 + v13 v23 )2 . (27)
After expanding the right-hand side of (27) and subtracting it from (26), we get the follow-
ing formula for the area element in R3 :
v
u  2   2   2
u v21 v32 v31 v12 v11 v22
(x; v1 , v2 ) = t det + det + det
v31 v22 v11 v32 v21 v12
(28)
Recognizing that the 2 2 determinants are just the values of basic forms dx2 dx3 ,
dx2 dx3 and dx2 dx3 , we can rewrite (28) in more legible (as well as more easily
memorizable!) form:

(dx2 dx3 )2 + (dx3 dx1 )2 + (dx1 dx2 )2

p
= . (29)

This is Pythagoras Theorem4 for the area function, since identity (29) can be expressed
4
PUAGORAS (6th Century BC), one of the most mysterious and influential figures in Greek, and there-
fore also our, intellectual history. He was born in Samos in the mid-6th century BC and migrated to Croton

6
Math 53M, Fall 2003 Professor Mariusz Wodzicki

also as saying:

The square of the area of a parallelogram is the sum

of the squares of areas of orthogonal projections of (30)
that parallelogram onto all coordinate planes.

As stated, Theorem (30) holds for any n. For n = 2, formula (29) reduces to formula (24).
For n = 3, the coordinate planes are R3x2 x3 , R3x3 x1 and R3x1 x2 , respectively.

10 Example: cross-product of vectors in R3 Let us calculate the exterior product of two

1-forms on R3
(a1 dx1 + a2 dx2 + a3 dx3 ) (b1 dx1 + b2 dx2 + b3 dx3 ) (31)
with constant coefficients a1 , a2 , a3 , b1 , b2 , b3 . The result is the sum of 3 3 = 9 forms
ai bj dxi dxj . However, three of them are zero, since dxi dxi = 0. For the remaining six,
one has ai bj dxi dxj = bj ai dxj dxi , so the final result is the following combination
of three basic 2-forms on R3 :
(a2 b3 a3 b2 )dx2 dx3 + (a3 b1 a1 b3 )dx3 dx1 + (a1 b2 a2 b1 )dx1 dx2 (32)
in around 530 BC. There he founded the sect or society that bore his name, and that seems to have played
an important role in the political life of Magna Graecia for several generations. Pythagoras himself is said to
have died as a refugee in Metapontum. Pythagorean political influence is attested well into the 4th century,
with Archytas of Tarentum.
The name of Pythagoras is connected with two parallel traditions, one religious and one scientific. Pythago-
ras is said to have introduced the doctrine of transmigration of souls into Greece, and his religious influence
is reflected in the cult organization of the Pythagorean society, with periods of initiation, secret doctrines and
passwords (akousmata and symbola), special dietary restrictions, and burial rites. Pythagoras seems to have
become a legendary figure in his own lifetime and was identified by some with the Hyperborean Apollo. His
supernatural status was confirmed by a golden thigh, the gift of bilocation, and the capacity to recall his pre-
vious incarnations. Classical authors imagine him studying in Egypt; in the later tradition he gains universal
wisdom by travels in the east. Pythagoras becomes the pattern of the divine man: at once a sage, a seer, a
teacher, and a benefactor of the human race.
The scientific tradition ascribes to Pythagoras a number of important discoveries, including the famous
geometric theorem that still bears his name. Even more significant for Pythagorean thought is the discovery
of the musical consonances: the ratios 2 : 1, 3 : 2, and 4 : 3 representing the length of strings corresponding
to the octave and the basic harmonies (the fifth and the fourth). These ratios are displayed in the tetractys,
an equilateral triangle composed of 10 dots; the Pythagoreans swear an oath by Pythagoras as author of the
tetractys. The same ratios are presumably reflected in the music of the spheres, which Pythagoras alone was
said to hear. (Quoted from The Oxford Classical Dictionary, 3rd edition, Oxford: 1996).

7
Math 53M, Fall 2003 Professor Mariusz Wodzicki

The column-vector made of coefficients of 2-form (32) is known under the name of cross-

a1 b1
product of column-vectors a = a2 and b = b2 :
a3 b3

a2 b3 a3 b2
a b a3 b1 a1 b3 . (33)
a1 b2 a2 b1

Exterior product of 1-forms is the best way to understand the peculiar character of cross-
product (unlike dot-product, it exists only in 3-dimensional Euclidean space), and its prop-
erties:
a) b a = a b ;
b) a b is orthogonal to both a and b (and thus also to the plane they span);
c) ka bk = Area((a, b)) where (a, b) denotes the parallelogram with sides a and
b (this follows from formulae (28) and (33)).

11 Differential The operation that associates with a function f : D R its differential,

f 7 df, defines a correspondence

d : {0-forms on D} {1-forms on D} (34)

(we think of functions as differential 0-forms).5

We can similarly produce differential 2-forms from 1-forms:

d : {1-forms on D} {2-forms on D} (35)

if = f1 dx1 + + fm dxm then its differential d is defined as the 2-form

d df1 dx1 + + dfm dxm . (36)

5
Note that (34) is a function from the set of 0-forms to the set of 1-forms.

8
Math 53M, Fall 2003 Professor Mariusz Wodzicki

fi fi
By plugging dfi = x 1
dx1 + + xn
dxm into the right-hand-side of formula (36) and
taking into account the properties of exterior product, we obtain the representation of d
in terms of basic 2-forms:

 
X fj fi
d = dxi dxj . (37)
16i<j6n
xi xj

This formula is particularly simple in R2 :

 
f2 f1
d(f1 dx + f2 dy) = dx dy . (38)
x y

or,  
f2 f1
d(f1 dx1 + f2 dx2 ) = dx1 dx2 . (39)
x1 x2
if one uses notation x1 , x2 instead of x, y.

12 Calculation: d d = 0 Calculation of (d d)f = d(df) is a simple exercise:

 
f f
d(df) = d dx1 + + dxm
x1 xn
   
f f
= d dx1 + + d dxm
x1 xn
X 2 f
= dxi dxj
16i,j6n
x i xj

2 f 2 f
X  
= dxi dxj (40)
16i<j6n
xi xj xj xi

The classical Clairauts Theorem, cf. Section 21 in DCVF, says that the mixed partial deriva-
tives
2 f 2 f
and
xi xj xj xi
9
 Math 53M, Fall 2003 Professor Mariusz Wodzicki

are equal if they are continuous. Thus, for functions f which have continuous second order
partial derivatives, one has the following fundamental identity

d(df) = 0 . (41)

13 Example: curl of a vector field In College textbooks of Multivariable Calculus, a

function
f1
F = f2 : D R3 (42)
f3
defined on a subset D R3 , is often called a vector field on D. Properly speaking, a


vector field on a set D Rm is a family of vectors, ab, one per each point a D. In this
case, the correspondence

point a 7 column-vector b a,

which sends each point a D to the column-vector b a, becomes a function D Rm .

In 3-dimensional space the formula for the differential of a 2-form, (37), acquires the form:

     
f3 f2 f1 f3 f2 f1
d(f1 dx + f2 dy + f3 dz) = dy dz + dz dx + dx dy
y z z x x y

(43)
or, preferably,
     
f3 f2 f1 f3 f2 f1
d(f1 dx1 + f2 dx2 + f3 dx3 ) = dx2 dx3 + dx3 dx1 + dx1 dx2 .
x2 x3 x3 x1 x1 x2
(44)
if one uses notation x1 , x2 , x3 instead of x, y and z.
The vector function made of function-coefficients of 2-form (43) is known under the name

10
 Math 53M, Fall 2003 Professor Mariusz Wodzicki

of curl6 of vector function F:

f3 f2 f3 f2



y z

x2 x3



f1 f3
f
1 f3
curl F = (45)

z x

x3 x1

f2 f1 f2 f1

x y x1 x2

Formula for the differential of a 2-form in R3 , (43), is the real reason why curl F is impor-
tant.
In the language that avoids mentioning differential forms, identity (41) becomes the follow-
ing statement:

curl f = 0 . (46)

14 Properties of the operation of differential For any function f : D R, differential

1-forms and on D, and scalars a and b, one has:
(a) d(a + b) = ad + bd ; Linearity

(b) d(f) = df + fd ; the Leibniz Rule

(c) d(f ) = f (d) . the Chain Rule

Here f : E Rm is a vector function sending its domain into D and the pullback of 2-forms
is defined in exactly the same manner as for 1-forms:

(f )(x; u, v) (f (x); fx0 (u), fx0 (v)) . (47)

W Exercise 3 Verify the following properties of pullback of 2-forms:

(a) f (a1 + b2 ) = a f 1 + b f 2 ; (a, b R)

6
The curl of F is also denoted F.

11
Math 53M, Fall 2003 Professor Mariusz Wodzicki

(b) f ( ) = f f ;
(c) (f g) = g (f ) . (Hint: Use the Chain Rule.)

15 Example Let D R2 and f : D R2 be a function. Pullback f (dx dy) is a 2-form

on a subset of R2 and, therefore, is a multiple of the basic form dx dy, see Section 7. We
shall find the ratio
f (dx dy)
. (48)
dx dy
One has

f (dx dy) = (f dx) (f dy) = df1 df2

   
f1 f1 f2 f2
= dx + dy dx + dy
x y x y
 
f1 f2 f1 f2
= dx dy (49)
x y y x

= (det Jf ) dx dy . (50)

Thus we obtain the following beautiful formula:

f (dx dy)
= det Jf . (51)
dx dy

For a function f : D Rm , defined on a subset of Rm , the determinant, det Jf (a), of the

Jacobi matrix of f at point a D plays a fundamental role in Multivariable Calculus. It is
usually referred to as the Jacobian of f at point a.

16 Riemann sums and integral of 2-forms in R2 For any plane rectangle

I = x R2 | a1 6 x 6 b1 , a2 6 y 6 b2

(52)

its diameter kb ak will be denoted |I|.

A partition of I consists of a finite family P = {J} of closed rectangles with non-overlapping
interiors, whose union equals I.

12
Math 53M, Fall 2003 Professor Mariusz Wodzicki

Tagging a partition is the same as choosing for each member J P a point xJ . Tagged
members of P will be called cells.
Each cell J defines two vectors uJ and vJ (see Figure 1).

rectangle I

vJ xJ
uJ

Figure 1: A tagged partition of a rectangle I; shown one cell.

With any differential 2-form on I and any tagged partition P we can associate the so
called Riemann sum: X
SP () (xJ ; uJ , vJ ) . (53)
JP

The Riemann integral of over I is defined as the limit of Riemann sums when the mesh
of partition P :
kPk max |J| (54)
JP

approaches zero: Z
lim SP () . (55)
I kPk0

Riemann integral over more general bounded sets D R2 is introduced as follows. Take

13
Math 53M, Fall 2003 Professor Mariusz Wodzicki

any closed rectangle I containing D and extend to a form on I by the formula

(x; u, v) if x D
(
u, v) =
(x; (56)
0 otherwise .

This is called extension by zero. Then, the Riemann itegral of form over D is defined as
over rectangle I:
the integral of
Z Z
. (57)
D I

The result does not depend on the choice of rectangle containing D.

17 Double integrals For any function f : D R on a rectangle I, its Riemann integral

ZZ
fdxdy , (58)
I

which is also denoted ZZ

f(x, y) dxdy , (59)
I

is defined as the limit of the Riemann sums:

X X
SP (f |dx dy|) = f(xJ ) |dx dy|(xJ ; uJ , vJ ) = f(xJ ) Area(J) . (60)
JP JP

Traditional notation (58)(59) explains why it is called the double integral of f. A more
accurate notation Z
f |dx dy|
I

is also used.
Double integral over more general bounded regions D R2 is defined as described above
in the case of differential forms.
From this definition it is clear that for plane regions
ZZ Z
f(x, y) dxdy = f dx dy . (61)
D
D

14
Math 53M, Fall 2003 Professor Mariusz Wodzicki

ZZ
When D is a rectangle I, like in (52), double integral f(x, y) dxdy should not be
I
confused with iterated intregrals
Z b2 Z b1  Z b1 Z b2 
f(x, y)dx dy and f(x, y)dy dx . (62)
a2 a1 a1 a2

That all three are in fact equal is a nontrivial fact.

18 Area of a bounded set in the plane By definition, the area of a bounded subset D R2
equals ZZ
Area(D) = dxdy . (63)
D

Integral (63) exists precisely when the characteristic function of set D

1 if x D
(
D (x) (64)
0 otherwise

is Riemann integrable. In order for characteristic function D not to be Riemann integrable

the boundary of D:

D x R2 | any neighborhood of x contains both points from D and not from D



(65)
must be massive.
We say that a subset B R2 has measure zero if, for any given number  > 0, there exists
a (possibly infinite) sequence of squares m such that
X
B 1 2 . . . and Area(n ) <  . (66)
n=1

Any countable set D = {x1 , x2 , . . . } and any rectifiable curve have measure zero; in fact,
most fractal curves in the plane (e.g. the curve in figure 2 in LI) have measure zero.7
7
A discipline of Mathematics called Measure Theory was developed in the first half of XX-th century;
the definition of integral based on it (Lebesgue integral) is today a standard anybody who is serios about
Mathematics must know. Modern Probability Theory and Mathematical Statistics are largely formulated in
the language of Measure Theory.

15
Math 53M, Fall 2003 Professor Mariusz Wodzicki

One can indeed show that

a bounded subset D R2 has a well defined area (i.e., integral

. (67)
(63) exists) if and only if boundary D has measure zero

19 An integral inequality A function f : D R is said to be bounded if there exists a L

number M > 0 such
|f(x)| 6 M for all x D . (68)
Any such number is then called a bound for f. For example, the function
 
x1
f = x1 sin x2
x2
is bounded on the rectangle
D = x R2 | 3 6 x1 6 2 , 4 6 x2 6 10 ,

(69)
since
|f(x)| = |x1 sin x2 | 6 |x1 | + | sin x2 | 6 3 + 1 = 4
for x D. In particular, any M > 4 is a bound for f.
For a bounded function f : D R which is integrable over D, the important inequality

Z Z 
 

 f(x, y) dxdy 6 M Area(D) (70)
D

follows directly from the following obvious inequality satisfied by Riemann sums (60):
X  X
 SP (f |dx dy|) =   f(xJ ) Area(J)
  
f(xJ ) Area(J) 6 (71)

JP JP
X
6 M Area(J) = M Area(I) . (72)
JP

In particular, for any bounded function on a set D R2 of zero area, one has
ZZ
f(x, y) dxdy = 0 . (73)
D

16
Math 53M, Fall 2003 Professor Mariusz Wodzicki

20 The so called Fubinis Theorem When the double integral exists, then it is equal to
either of the two iterated integrals
Z b2 Z b1  ZZ Z b1 Z b2 
f(x, y) dx dy = f(x, y) dxdy = f(x, y) dy dx . (74)
a2 a1 a1 a2
I

This theorem, in fact, has been proved in 1882 by Paul David Gustav du Bois-Reymond
(1831-1889).8 Italian Guido Fubini (1879-1943) in 1907 proved a much more general
theorem and his name has been often attached even to the special case we encounter here.
Fubinis theorem reduces computation of double integrals to integrals of Freshman Calculus.

21 Greens Theorem A loop : [a, b] Rm , cf. Section 36 of LI, is said to be simple if

it is one-to-one on interval [a, b). A curve C parametrized by simple loop will be called a
simple closed curve. A simple closed curve is a continuous one-to-one image of a circle.

Figure 2: An example of a simple plain region D: it has two

connected components and its boundary D consists of five
simple closed curves with induced orientation indicated by ar-
rows.
8
Ueber das Doppelintegral, Journal fr Mathematik (Crelle), 94 (1883), 273290.

17
Math 53M, Fall 2003 Professor Mariusz Wodzicki

Let D be a region in R2 whose boundary is the disjoint union of simple closed curves.
We shall call such sets simple (plane) regions. The orientation of the boundary, D, which
leaves set D leftwards when we traverse the boundary, will be called the induced orientation
(or, positive with respect to set D).
The following theorem, due to George Green (17931841), expresses a fundamental link
between the theories of integration of differential 1- and 2-forms. It could be called the
Fundamental Theorem of Calculus for 1-forms.

Let be a differential 1-form on a simple plane region D. If

the components of the boundary, D, of D are rectifiable simple
closed curves, then Z Z (75)
d = .
D D

One should understand Greens Theorem (75) as saying that the two integrals are equal
when they exist. Each component of the boundary of D is supposed to be positively ori-
ented (with respect to D).

22 Parametric surfaces A continuous function : D Rm , where D is a subset of R2 ,

will be called a parametric surface. Set D will be referred to as the parameter set of .
We shall say that parametric surface is contained in set E Rm if () E for any point
D.
Suppose is differential 2-form on E. Were we to define integral of along parametric
surface by analogy with our definition of path integrals in Section 1 of LI, we would have
encountered serious difficulties.9
Recall that path integrals reduce to Riemann integrals thanks to the Change
Z of Variables
Formula, see (47) and (49) in LI, we could have defined path integral of a 1-form

as the integral
Z Z b

= .
[a,b] a

9
See, e.g. an elementary article by Tibor Rad, What is the area of a surface?, Amer. Math. Monthly, 50
(1943), 139141.

18
Math 53M, Fall 2003 Professor Mariusz Wodzicki

Z
Guided by this, we will define integral as

Z ZZ
. (76)

D

The Change of Variables Formula:

Z Z
= f (77)
f

is then built-in into our definition.

23 Example: integrating a 2-form over a cone A slice of height r, cut out by planes z = 0
and z = r, of the cone in R3 given by the equation

z2 = x2 + y2 , (78)

admits a simple and nearly one-to-one parametrization

  t2 cos t1
t1
= t2 sin t1 (0 6 t1 6 2, 0 6 t2 6 r) . (79)
t2
t2

by the rectangle
 
t1
n  o
I= = R2  0 6 t1 6 2, 0 6 t2 6 r . (80)

t2

19
Math 53M, Fall 2003 Professor Mariusz Wodzicki

For any function f whose domain contains this slice of the cone surface, we have

Z Z Z t2 cos t1
f dx dy = (f dx dy) = f t2 sin t1 d1 d2
I I
t2

t2 cos t1
d1 d2
Z
= f t2 sin t1 dt1 dt2
I dt1 dt2
t2

1 1


Z r Z 2 t2 cos t1 t1
t2
dt1 dt2 . (81)
= f t2 sin t1 det
2
0

0 2
t2
t1 t2
Here
1 1

 
t1 t2
det = det cos t1 t2 sin t1 = t2 . (82)
2 2 sin t1 t2 cos t1
t1 t2
Thus,

Z Z r Z 2 t2 cos t1
f dx dy = f t2 sin t1 t2 dt1 dt2 . (83)
0 0
t2

For example,
Z Z r Z 2 
x z dx dy =
2 2
(t2 cos t1 ) t2 t2 dt1 dt2
0 0

2 r
r5
Z  Z 
2
= cos t1 dt1 t42 dt2 = . (84)
0 0 5

24 Area of a parametric surface The area of a parametric surface : D Rm is defined

as the integral over of the area element introduced in Section 9:10
Z Z
Area() = . (85)
D
10
The difficulties with introducing the area as the limit of the areas of polyhedral approximations, have
been signalled in a footnote on page 18.

20
Math 53M, Fall 2003 Professor Mariusz Wodzicki

For parametric surfaces in two or three dimensional space, we have explicit formulae for
the area element, namely (24) and (28), respectively. Thus,
ZZ
(in R2 )  
Area() = det J (t, u) dtdu (86)
D
ZZ q
(in R3 )
= (det J1 (t, u))2 + (det J2 (t, u))2 + (det J3 (t, u))2 dtdu (87)
D

where i : D R2 , i = 1, 2, 3, are the functions:

     
2 3 1
1 , 2 and 3 . (88)
3 1 2

25 Stokes Theorem (parametric form) Suppose that parameter set D satisfy hypothesis
of Greens Theorem, see Section 21.
Then we have
Z Z
d = d by Definition (76)
D
Z
= d( ) by Property 14(c)
D
Z
= by Greens Theorem (89)
D

We shall call the equality

Z Z
d = (90)
D

the parametric form of Stokes Theorem. We shall now give a remarkable application of
(90).

26 Homotopic paths Let E be a subset of Rm . Suppose we are given two paths in E,

0 : [a, b] E and 1 : [a, b] E, having the same starting point

0 (a) = 1 (a) = a

21
Math 53M, Fall 2003 Professor Mariusz Wodzicki

and the same endpoint

0 (b) = 1 (b) = b .
We say that these paths are homotopic (in set E) if there exists a continuous function

0
a b
E

Figure 3: An example of a homotopy between paths.

H : I E defined on the rectangle

n t   o
I [a, b] [0, 1] = R2  a 6 t 6 b , 0 6 u 6 1 (91)

u

such that    
t t
H = 0 (t) and H = 1 (t) (92)
0 1
for all t [a, b], and
   
a b
H =a and H =b (93)
u u

for all u [0, 1]. Function H in this case is called a homotopy between 0 and 1 . If the
paths are differentiable and homotopic, then one can always find a differentiable homotopy.

22
Math 53M, Fall 2003 Professor Mariusz Wodzicki

0
a b
E
1

Figure 4: An example of nonhomotopic paths:

there is no homotopy between these two path in
set E.

C1
1

Ca rectangle I Cb

0
a C0 b

Figure 5: The positively oriented boundary of rect-

angle I is the union of C0 , Cb , C1 and Ca .

A homotopy between paths is a parametric surface. Stokes Theorem (90) thus gives us, for

23
Math 53M, Fall 2003 Professor Mariusz Wodzicki

any 1-form on E, the equality:

Z Z Z Z Z Z

d = H = H + H + H + H
H I C0 Cb C1 Ca
Z Z Z Z
= H + H H H . (94)
C0 Cb C1 Ca

Obvious parametrizations of horizontal segments C0 and C1 are:

 
2 t
0 : [a, b] R , 0 (t) =
0

and  
2 t
1 : [a, b] R , 1 (t) =
1
respectively. Similarly, obvious parametrizations of vertical segments C0 and C1 are:
 
2 a
a : [0, 1] R , a (u) =
u

and  
2 b
b : [0, 1] R , b (u) =
u
respectively. Thus, the right-hand side of (94) equals
Z b Z 1 Z b Z 1

(H 0 ) + (H b ) (H 1 ) (H a ) . (95)
a 0 a 0

Now,
   
t t
(H 0 )(t) = H = 0 (t) and (H 1 )(t) = H = 1 (t) ,
0 0

while (H a ) and (H b ) are degenerate one-point paths:

   
a b
(H 0 )(u) = H =a and (H b )(u) = H =b.
u u

In particular, (H a ) = (H b ) = 0 for any form , and (95) equals

Z b Z b Z Z

(H 0 ) (H 1 ) = (96)
a a 0 1

24
Math 53M, Fall 2003 Professor Mariusz Wodzicki

by Change of Variables Formula (49) in LI.

By combining, (94), (95) and (96), we obtain the following important identity:

Z Z Z
= d . (97)
0 1 H

27 (Freely) homotopic loops A path : [a, b] Rm which ends where it starts:

(a) = (b) (98)

is called a loop. For loops, one can relax the definition of homotopy. For two loops
0 : [a, b] E and 1 : [a, b] E, a function H : I E defined on rectangle (91), which
satisfies condition (92) and the following condition
   
a b
H =H (for all u [0, 1]), (99)
u u

which is weaker11 than condition (93), will be called a free homotopy between loops 0
and 1 .
A loop : [a, b] E is said to be contractible if it is homotopic in E to a constant loop

0 (t) = a, (for all t [a, b]), (100)

i.e., a loop that stays at a point a E. Examples of noncontactible loops in subsets of R3

are provided by links, see Section 17 in 3F (and illustrations to Problems ?? and ???? in
Problembook).
Z 1 Z 1

When H is a free homotopy between two loops, integrals (H a ) and (H b )
0 0
in formula (95) do not necessarily vanish. Instead, they cancel each other, because now
a = b . In particular, identity (97) holds also for free homotopies between loops.
 
t
11
Condition (99) means that each path u : [a, b] Rm , u [0, 1], where u (t) H , is a loop.
u

25
Math 53M, Fall 2003 Professor Mariusz Wodzicki

28 Closed 1-forms A 1-form is said to be closed if d = 0. According to Theorem

(41), any exact form12 is closed.
For closed forms, identity (97) yields the following fundamental fact.

Let 0 and 1 be two homotopic paths in E Rm ,

or two (freely) homotopic loops. Then, for any closed
1-form on E,
Z Z
= .
0 1

Since the integral of any form over constant loop (100) (101)
is zero, we deduce that
Z
= 0

for any closed 1-form and any loop that is con-

tractible in the domain of .

Z
m
In other words, for any closed differential 1-form on a set E R , integral

depends only on the homotopy class of .

29 Winding number of an oriented closed curve Let a be a point belonging to the com-
plement of C:
R 2 \ C x R2 | x

/C (102)
of an oriented closed13 curve C in the plane. The number of times curve C winds itself
around point a is called the winding number of curve C about point a or, the index of
point a with respect to curve C. To determine it, draw a ray from point a to infinity
which is transversal to curve C. This means that is allowed to intersect C only at regular
points14 of C, and that is not tangent to C at any point of intersection. Two examples of
transversal rays are shown in Figure 6.
12
Recall from LI, Section 35, that a 1-form on set D is exact if = df for some function on D.
13
We shall be content here with the following working definition of a closed curve: a continuous image of
a finite number of circles. Orienting such a curve is the same as orienting each constituent loop.
14
A point x C is regular if in its vicinity C is a regular arc, see Section 30 of LI.

26
Math 53M, Fall 2003 Professor Mariusz Wodzicki

1
1
a 1
1 C
1

Figure 6: The winding number of an oriented closed curve C about a

point a R2 \ C does not depend on the choice of transversal ray: it
equals 1 + 1 = 2 along ray 1 and 1 + 1 + 1 + 1 = 2 along ray 2 .

Then, moving outwards along from point a, each time you cross C you add 1, if at that
point curve C moves leftwards, and you subtract 1 if curve C moves rightwards. The result
does not depend on the choice of ray (as illustrated by Figure 6). Thus determined number
will be denoted ind C (a).
Let us make some simple observations:
(a) The index is always an integer. Changing orientation of curve C reverses the sign of
index of all points in R2 \ C.
(b) Any integer can occur as index: for example, point a in Figure 7 has index 5.
(c) The index does not change if one replaces point a by points sufficiently close to it. This,
in turn, implies that all points belonging to the same connected component of complement
R2 \ C have the same index. I have illustrated this in Figure 8, by indicating, for every
connected component of R2 \ C, the index of points belonging to it.
(d) Among connected components of complement R2 \ C, there is exactly one which is
unbounded. For every point a in the unbounded component, ind C (a) = 0.

27
Math 53M, Fall 2003 Professor Mariusz Wodzicki

a.

Figure 7: Curve C winds itself 5 times in clockwise (i.e.,

negative) direction around point a, hence ind C (a) = 5.

1
1
0

1
C 0
2
1

Figure 8: Index ind C (a) is constant on each

connected component of complement R2 \ C .

30 Index Formula We shall now establish the following remarkable formula:

Z
1 (x1 a1 )dx2 (x2 a2 )dx1
ind C (a) = . (103)
2 C kx ak2

Let E R2 \ {a} be the plane with point a removed. It is sufficient to prove Index Formula
(103) for curves parametrized by a single loop : [0, 1] E. If ind C (a) = m, then is

28
Math 53M, Fall 2003 Professor Mariusz Wodzicki

(freely) homotopic in E to the path 1 : [0, 1] E which traverses m times the unit circle
with center at point a:  
cos(2mt)
1 (t) a + . (104)
sin(2mt)
Since the differential 1-form on set E:
1 (x1 a1 )dx2 (x2 a2 )dx1
(105)
2 kx ak2

is closed, we have
Z
1 (x1 a1 )dx2 (x2 a2 )dx1
2 C kx ak2
Z
1 (x1 a1 )dx2 (x2 a2 )dx1
= (in view of Theorem (101))
2 1 kx ak2
1
cos(2mt)(2m cos2 (2mt)) sin(2mt)(2m sin2 (2mt))
Z
1
= dt
2 0 cos2 (2mt) + sin2 (2mt)
Z 1
2m
= dt = m (106)
2 0
Z
(x1 a1 )dx2 (x2 a2 )dx1
which proves Index Formula (103) (compare the computation of
1 kx ak2
to computation (95) in LI).

31 Remark Let
1 x1 dx2 x2 dx1
1 (107)
2 x21 + x22

be a differential 1-form on R2 \ {0}. Except for the factor 1/2, this is differential form
(91) which we encountered in Section 37 of LI. Let us denote by C a the curve obtained
by translating curve C by column-vector a. If function parametrizes C then function
a, defined as ( a)(t) (t) a, parametrizes C a . Note that Index Formula
(103) can now be written as follows:

Z
1
ind C (a) = 1 . (108)
2 Ca

29
Math 53M, Fall 2003 Professor Mariusz Wodzicki

Note that the integrand in formula (108) depends neither on curve C nor on point a.

32 Change of Variables Formula for Double Integral A differentiable function h : D

D 0 is called a diffeomorphism if there exists a differentiable function g : D 0 D such that
g h is the identity function on D and h g is the identity function on D 0 .15
Diffeomorphisms form a very important class of functions.
Suppose that f is an integrable function on set D 0 R2 and h : D D 0 is a diffeo-
morphism. The following formula is a 2-dimensional analog of the Change of Variables
Formula from Freshman Calculus:

ZZ ZZ
 
f(u, v)dudv = (f h)(x, y) det Jh (x, y)  dxdy . (109)
D0 D

Formula (109) has many applications. For example, if 1 : D1 Rm and 2 : D2 Rm

are two parametrizations of one and the same surface S such that

1 = 2 h (110)

for some diffeomorphism h : D1 D2 of their respective parameter spaces, then

Z Z
= (111)
1 2

with:
plus sign: when det Jh > 0 everywhere on D1 , (111 + )
minus sign: when det Jh < 0 everywhere on D1 . (111 )
In the first case, we say that diffeomorphism h preserves orientation, in the second case we
say that h reverses orientation.
15
In other words, function h has an inverse and that inverse is differentaible too.

30
Math 53M, Fall 2003 Professor Mariusz Wodzicki

Formula (111) is nothing but a 2-dimensional version of formula (63) for path integrals,
see LI.
We can say, as we did in Section 29 of LI, that parametrizations 1 and 2 are equivalent
if equality (110) holds for a diffeomorphism satisfying condition (111 + ). Then, identity
(111) says that Z Z
= (112)
1 2

for equivalent parametrizations.

33 Integrating 2-forms over oriented patches Particularly important are one-to-one reg-
ular parametrizations by simple regions, see Section 21. For the sake of these notes, surfaces
admitting such parametrizations will be called regular patches.
Note that regular pathches are 2-dimensional versions of regular arcs, introduced in Section
30. All regular parametrizations of a connected patch S define one of the two possible
orientations of its boundary S, which is a closed curve in Rm . Choosing between these
two possible orientations will be refered to as orienting the patch. L
Z
Then, for an oriented patch S and a 2-form on S, we define the integral as
S

Z Z
(113)
S

where is any regular parametrization of S which is compatible with the given orientation
of S.
From formula (109) we know that the result does not depend on the particular choice of
parametrization .

34 Orienting regular patches in R3 When one looks at a small neighborhood of an in-

ternal point on a regular patch S embedded in R3 , then points not belonging to the patch
are on one or another side of the patch. Orienting a patch embedded in R3 is the same
as telling which side is in and which one is out: when one looks at the patch from
the side that is in, the positively oriented simple loops on S are the ones that are traversed L
counterclockwise.

31
Math 53M, Fall 2003 Professor Mariusz Wodzicki

The situation is entirely different for regular patches in Rm when dimension m > 3. The
concept of side does not make sense in that case.16

35 Orientable surfaces Suppose that S is a surface in Rm which is decomposed into a

number of nonoverlapping oriented regular patches S1 , . . . , Sk .
When we look at arcs of the common boundary between two adjacent patches, they receive
orientation from either of the two patches. When these two orientations are opposite to
each other, we say that the patches are compatibly oriented.

S3 S3

S1 S1

S2 S2

(a) Compatibly oriented patches (b) Incompatibly oriented

patches

Figure 9: On the left, all the patches have the same orien-
tation; on the right, patches S2 and S3 have the same, and
patch S1 has the opposite orientation.

A surface S that can be decomposed into compatibly oriented regular patches will be called
orientable. The choice of orientation of S is the same as the choice of compatible orientation
for all patches.
Not every surface is orientable. The simplest example is provided by Mbius surface,17 see
16
We shall explain this by analogy with arcs. An arc in R2 has two sides, and orienting an arc in R2 is the
same as telling which side is positive and which one is negative. The situation is entirely different for arcs
in Rm when dimension m > 2. Look, for example at the case of an arc in R3 .
17
August Ferdinand Mbius (17901868)

32
Math 53M, Fall 2003 Professor Mariusz Wodzicki

figure 10.

Figure 10: Mbiuss surface.

For an oriented surface S and a differential 2-form on S, we define the integral of

over S as the sum of integrals of over all patches:
Z Z Z
+ + (114)
S S1 Sk

The right-hand-side does not depend on the decomposition, but changes sign if we reverse
all patch orientations.

36 Remark Some oriented surfaces which are not patches, e.g. spheres, ellipsoids, tori,18
etc., admit a parametrization : D S by a simple region D of plane R2 which is regular
and one-to-one except for a subset Dbad D of area zero. One can show without effort
that in this situation Z Z
= (115)
S
18
A torus is the surface of a donut; tori is the plural.

33
Math 53M, Fall 2003 Professor Mariusz Wodzicki

for any differential 2-form .19

37 An application: computation of the Gauian integral For any real number R > 0, let
DR R2 denote the disk of radius R with center at the origin, and let R denote the square
[R, R] [R, R]. Consider the positive function on R2 :
2
x 7 ekxk . (116)

Since, DR R D2R , we have the double inequality:

ZZ ZZ ZZ
kxk2 kxk2 2
e dxdy < e dxdy < ekxk dxdy . (117)
DR R D2R

Disk DR has the polar-coordinates parametrization:

   
r r cos
h : [0, R] [0, 2] DR , h = (118)
r sin

which is a diffeomorphism (preserving orientation) except for the subset of [0, R] [0, 2] :
  
r
| r = 0 or = 0 or 2 (119)

which is of area zero. Thus, by Remark 36,

ZZ Z
kxk2 2
e dxdy = ekxk dx dy (120)
h
BR

ZZ  
(r cos2 +r sin2 ) cos r sin
= e det drd (121)
sin r cos
[0,R][0,2]

Z R Z 2
r2
= e rdr d (by Fubinis Theorem) (122)
0 0

1 2 2
= (1 eR ) 2 = (1 eR ) . (123)
2
Z Z
19
At least when the function-coefficients of are bounded; otherwise, integral may exist while
S
may not.

34
Math 53M, Fall 2003 Professor Mariusz Wodzicki

By Fubinis Theorem, also

ZZ ZZ Z R Z R Z R 2
kxk2 x2 y2 x2 y2 x2
e dxdy = e e dxdy = e dx e dy = e dx .
R R R
R R
(124)
Therefore, double inequality (117) yields the estimates
Z R
p 2 p
1 eR < 2
ex dx < 1 e2R2 , (125)
R

and, in the limit R ,

Z
2
6 ex dx 6 . (126)

It follows that the value of the Gauian integral is:

Z
2
ex dx = . (127)

This is a fundamentally important result (especially for Probability Theory and Physics).
You were told when learning Freshman Calculus that the anti-derivative of the function
2
x 7 ex cannot be expressed in terms of elementary functions of Calculus. Thus, what
you
Z have learned in Freshman Calculus did not allow you to compute Gauian integral
2
ex dx.

38 Stokes Theorem (general case) For any differential 1-form on an oriented surface
S, one has
Z Z
d = (128)
S S

Indeed, if S is a regular patch, then equality (128) follows from the respective definition of
both integrals in (128) and from the parametric form of Stokes Theorem, see (90).

35
Math 53M, Fall 2003 Professor Mariusz Wodzicki

In general case, if S decomposes into patches S1 , . . . , Sk , then

Z Z Z Z Z
d = d + + d = + + (129)
S S1 Sk S1 Sk

in view of Stokes Theorem applied to each patch Si separately.

Notice that for each arc C in the boundary of a given patch Si , C either belongs
Z also to
Z the
boundary of some other patch, say Sj , in which case its contributions to and
Si Sj
are with opposite signs (and therefore cancel each other!), or C is solitary, in which case
it is an arc of the boundary of surface S (see, e.g., Figure 9(a)). Therefore, the sum
Z Z
+ +
S1 Sk

coincides with the sum over all arcs of the boundary of S, i.e.,
Z Z Z
+ + = ,
S1 Sk S

and equality (128) has been proved.

36