You are on page 1of 103

Giovanni Leoni, Vector Analysis, Spring 2013, Carnegie Mellon University

Monday, January 14, 2013

Real Numbers

There are two ways to introduce the real numbers. The rst is to give them
in an axiomatic way, the second is to construct them starting from the natural
numbers. We will use the rst method.
The real numbers are a set R with two binary operations, addition
+:R

R!R

(x; y) 7! x + y
and multiplication
:R

R!R

(x; y) 7! x y
and a relation

such that (R; +; :; ) is an ordered eld, that is,

(A) (R; +) is an commutative group, that is,


(A1 ) for every a; b 2 R, a + b = b + a,

(A2 ) for every a; b; c 2 R, (a + b) + c = a + (b + c),

(A3 ) there exists a unique element in R, called zero and denoted 0, such
that 0 + a = a + 0 = a for every a 2 R,

(A4 ) for every a 2 R there exists a unique element in R, called the opposite
of a and denoted a, such that ( a) + a = a + ( a) = 0,
(M )
(M1 ) for every a; b 2 R, a b = b a,

(M2 ) for every a; b; c 2 R, (a b) c = a (b c),

(M3 ) there exists a unique element in R, called one and denoted 1, such
that 1 6= 0 and 1 a = a 1 = a for every a 2 R with a 6= 0,

(M4 ) for every a 2 R with a 6= 0 there exists a unique element in R, called


the inverse of a and denoted a 1 , such that a 1 a = a a 1 = 1,

(O)

is a total order relation, that is,


(O1 ) for every a; b 2 R either a
(O2 ) for every a; b; c 2 R if a
(O3 ) for every a; b 2 R if a

b or b
b and b

b and b

(O4 ) for every a 2 R we have a

a,

a,
c, then a
a, then a = b,

c,

(AM ) for every a; b; c 2 R, a (b + c) = (a b) + (a c),


(AO) for every a; b; c 2 R if a
(M O) for every a; b 2 R if 0

b, a + c
a and 0

b + c,
b, then 0

a b,

(S) (supremum property)


Remark 1 Properties (A), (M ), (O), (AM ), (AO), (M O), and (S) completely
characterize the real numbers in the sense that if (R0 ; ; ; 4) satises the same
properties, then there exists a bijection T : R ! R0 such that T is an isomorphism between the two elds, that is,
T (a + b) = T (a)

T (b) ;

T (a b) = T (a)

T (b)

for all a; b 2 R, and a b if and only if T (a) 4 T (b). Hence, for all practical
purposes, we cannot distinguish R from R0 .
If a

b and a 6= b, we write a < b.

Exercise 2 Using only the axioms (A), (M ), (O), (AO), (AM ) and (M O) of
R, prove the following properties of R:
(i) if a b = 0 then either a = 0 or b = 0,
(ii) if a
(iii) if a

0 then

0,

b and c < 0 then ac

(iv) for every a 2 R we have a2

bc,
0,

(v) 1 > 0.
Denition 3 Let E

R be a nonempty set.

(i) An element L 2 R is called an upper bound of E if x

L for all x 2 E;

(ii) E is said to be bounded from above if it has at least an upper bound;


(iii) if E is bounded from above, the least of all its upper bounds, if it exists, is
called the supremum of E and is denoted sup E.
(iv) E has a maximum if there exists L 2 E such that x
We write L = max E.

L for all x 2 E.

We are now ready to state the supremum property.


(S) (supremum property) every nonempty set E
has a supremum in R.

R bounded from above

The supremum property says that in R the supremum of a nonempty set


bounded from above always exists in R.
2

Remark 4 (i) Note that if a set has a maximum L, then L is also the supremum of the set.
(ii) If E
R is a set bounded from below, to prove that a number L 2 R is
the supremum of E, we need to show that L is an upper bound of E, that
is, that x L for every x 2 E, and that any number s < L cannot be an
upper bound of E, that is, that there exists x 2 E such that s < x.
Denition 5 Let E

R be a nonempty set.

(i) An element ` 2 R is called a lower bound of E if `

x for all x 2 E;

(ii) E is said to be bounded from below if it has at least an lower bound;


(iii) if E is bounded from above, the greatest of all its lower bounds, if it exists,
is called the inmum of E and is denoted inf E;
(iv) E has a minimum if there exists ` 2 E such that `
write ` = min E.

x for all x 2 E. We

Remark 6 (i) Note that if a set has a minimum `, then ` is also the inmum
of the set.
(ii) If E
R is a set bounded from above, to prove that a number ` 2 R is
the inmum of E, we need to show that ` is a lower bound of E, that is,
that ` x for every x 2 E, and that any number ` < s cannot be a lower
bound of E, that is, that there exists x 2 E such that x < s.
Example 7 Consider the set
E = y 2 R : y = arctan x2
arctan t
Since 2
the minimum of E is

for all t 2 R, the set E is bounded. Lets prove that


inf E = arctan 1 =

Since

1 +1 ; x2R :
1
:
4

d
1
(arctan t) = 2
> 0;
dt
t +1

the function arctan t is strictly increasing. If x 6= 1, we have that x2 1 +1 >


1, and so arctan x2 1 + 1 > arctan 1, while if x = 1, y = arctan (0 + 1) =
1
4 . Hence, min E = 4 . To nd the supremum note that since arctan t
2 for
all t 2 R, we have that 2 is an upper bound of the set E. To prove that it is the
supremum, we need to show that any s < 2 is not an upper bound of E. Since
lim arctan x2

x!1

1 +1 =

for every " > 0 we can nd M > 0 such that


arctan x2

1 +1
3

<"

for all x > M , that is,


" < arctan x2 1 + 1 < + "
2
2
for all x > M . Taking 0 < " < 2 s, we have that 2 " > s, and so
" < arctan x2 1 + 1
2
for all x > M , which shows that s is not an upper bound of E. Hence,
least upper bound of E. This implies that
s<

sup E =

is the

:
Wednesday, January 16, 2013

Example 8 Consider the set


E=

xy
; x; y 2 R; x < y :
x2 + y 2

t2R: t=

xy
1
The set E is bounded since 21
x2 +y 2
2 for all x; y 2 R, with x < y.
Moreover, by taking x = 1 and y = 1, we get t = 21 , so

inf E = min E =

1
:
2

Lets prove that


1
:
2
1
We need to show that any s < 21 is not an upper bound for the set E. If s
2,
then we can take x = 1 and y = 0, so that s < t = 0. Thus assume that
1
1
2 < s < 2 . Take the y = 1. Since
sup E =

lim

x!1

for every " > 0 we can nd

> 0 such that


x2

for all 1

1
<"
2

x
+1

< x < 1, that is,


1
2

for all 1

1
x
= ;
x2 + 1
2

"<

x
1
< +"
x2 + 1
2

< x < 1. Taking 0 < " <


s<

1
2

1
2

s, we have that

"<

x
x2 + 1

1
2

" > s, and so

for all 1
< x < 1, which shows that s is not an upper bound of E. Thus, 12
1
is the supremum of the set. Note that t = x2xy
+y 2 = 2 only if x = y, which is not
allowed, so the set does not have a maximum.
4

The Euclidean Space

Denition 9 A vector space, or linear space, over R is a nonempty set X,


whose elements are called vectors, together with two operations, addition and
multiplication by scalars,
X X!X
(x; y) 7! x + y

and

R X!X
(t; x) 7! tx

with the properties that


(i) (X; +) is a commutative group, that is,
(a) x + y = y + x for all x; y 2 X (commutative property),

(b) x + (y + z) = (x + y) + z for all x; y; z 2 X (associative property),

(c) there is a vector 0 2 X, called zero, such that x + 0 = 0 + x for all


x 2 X,

(d) for every x 2 X there exists a vector in X, called the opposite of x


and denoted x, such that x + ( x) = 0,
(ii) for all x; y 2 X and s; t 2 R,
(a) s (tx) = (st) x,
(b) 1x = x,
(c) s (x + y) = (sx) + (sy),
(d) (s + t)x = (sx) + (tx).
Remark 10 Instead of using real numbers, one can use C or a eld F . For
most or our purposes the real numbers will su ce. From now on, whenever we
dont specify, it is understood that a vector space is over R.
Example 11 Some important examples of vector spaces over R are the following.
(i) The Euclidean space RN is the space of all N -tuples x = (x1 ; : : : ; xN )
of real numbers. The elements of RN are called vectors or points. The
Euclidean space is a vector space with the following operations
x + y := (x1 + y1 ; : : : ; xN + yN ) ;

tx := (tx1 ; : : : ; txN )

for every t 2 R and x = (x1 ; : : : ; xN ) and y = (y1 ; : : : ; yN ) in RN .


(ii) The collection of all polynomials p : R ! R.
(iii) The space of continuous functions f : [a; b] ! R.

Denition 12 An inner product, or scalar product, on a vector space X is a


function
(; ):X X!R
such that
(i) (x; x)

0 for every x 2 X, (x; x) = 0 if and only if x = 0 (positivity);

(ii) (x; y) = (y; x) for all x; y 2 X (symmetry);


(iii) (sx + ty; z) = s (x; z)+t (y; z) for all x; y; z 2 X and s; t 2 R (bilinearity).
Remark 13 If X is a vector space over C, then an inner product is a function
( ; ) : X X ! C satisfying properties (i),
(ii)0 (x; y) = (y; x) for all x; y 2 X (skew-symmetry), where given z 2 C, the
number z is the complex conjugate;
(iii) (sx + ty; z) = s (x; z)+t (y; z) for all x; y; z 2 X and s; t 2 C (bilinearity).
Example 14 Some important examples of inner products are the following.
(i) Consider the Euclidean space RN , then
x y := x1 y1 +

+ xN yN ;

where x = (x1 ; : : : ; xN ) and y = (y1 ; : : : ; yN ), is an inner product.


(ii) Consider the space of continuous functions f : [a; b] ! R. Then
(f; g)L2 ([a;b]) :=

f (x) g (x) dx

is an inner product.
Friday, January 18, 2013
Solutions of the test.
Monday, January 21, 2013
Martin Luther Kings Day. No class
Wednesday, January 23, 2013
Denition 15 A norm on a vector space X is a map
k k : X ! [0; 1)
such that
(i) kxk = 0 implies x = 0;
(ii) ktxk = jtj kxk for all x 2 X and t 2 R;
(iii) kx + yk

kxk + kyk for all x; y 2 X.


6

A normed space (X; k k) is a vector space X endowed with a norm k k. For


simplicity, we often say that X is a normed space.
Given an inner product ( ; ) : X X ! R on a vector space X, it turns out
that the function
p
(1)
kxk := (x; x); x 2 X;
is a norm. This follows from the following result.

Proposition 16 (CauchySchwarzs inequality) Given an inner product ( ; ) :


X X ! R on a vector space X,
j(x; y)j

kxk kyk

for all x; y 2 X.
Proof. If y = 0, then both sides of the previous inequality are zeros, and so
there is nothing to prove. Thus, assume that y 6= 0 and let t 2 R. By properties
(i)-(iii),
2
2
0 (x + ty; x + ty) = kxk + t2 kyk + 2t (x; y) :
(2)
Taking
(x; y)

t :=

kyk

in the previous inequality gives


2

kxk +

(x; y)

or, equivalently,
(x; y)

kyk
2

kyk
2

(x; y)
2

kyk

kxk kyk :

It now su ces to take the square root on both sides.


Remark 17 It follows from the proof that equality holds in the CauchySchwarz
inequality if and only you have equality in (2), that is, if x + ty = 0 for some
t 2 R or y = 0.
Corollary 18 Given a scalar product ( ; ) : X X ! R on a vector space X,
the function
p
kxk := (x; x); x 2 X;
is a norm.

Proof. By property (i), k k is well-dened and kxk = 0 if and only if x = 0.


Taking t = 1 in (2) and using the CauchySchwarz inequality gives
0

kx + yk = kxk + kyk + 2 (x; y)


2

kxk + kyk + 2 kxk kyk = (kxk + kyk) ;


7

which is the triangle inequality for the norm. Moreover, by properties (ii) and
(iii) for every t 2 R,
p
p
p
p
ktxk = (tx; tx) = t (x; tx) = t (tx; x) = t2 (x; x) = jtj kxk :
Thus k k is a norm.

Example 19 Other important norms that one can put in RN are


kxk`1 := max fjx1 j ; : : : ; jxN jg ;
kxk`1 := jx1 j +

+ jxN j ;

kxk`p := (jx1 j +
for x = (x1 ; : : : ; xN ) 2 RN and where 1

p 1=p

+ jxN j )

p < 1.

The proof of the following theorem is left as an exercise.


Theorem 20 Let (X; k k) be a normed p
space. Then there exists an inner product ( ; ) : X X ! R such that kxk = (x; x) for all x 2 X if and only if k k
satises the parallelogram law
2

kx + yk + kx

yk = 2 kxk + 2 kyk

for all x; y 2 X.
Example 21 Using the previous theorem, we can prove that some normed
spaces are not inner product spaces. The space RN with the norm k k`p for
p 6= 2 is not an inner product . Take x = (1; 1; 0; : : :), y = (1; 1; 0; : : :). Then
x + y = (2; 0; : : :), x y = (0; 2; 0; : : :). Hence,
2

kx + yk`p + kx

yk`p = (2p ) p + (2p ) p = 8


2

6= 2 kxk`p + 2 kyk`p
2

= 2 (1p + 1p ) p + 2 (1p + 1p ) p = 22+ p :


Denition 22 Given a set E and a function f : E ! R, we say that f is
bounded from above if the set
f (E) := fy 2 R : y = f (x) ; x 2 Eg
is bounded from above. We say that f is bounded from below if the set f (E)
is bounded from below. Finally, we say that f is bounded if the set f (E) is
bounded.
Given a set E and a function f : E ! R, we write
sup f := sup f (E) ;

inf f := inf f (E) :


E

Exercise 23 Given a set E, consider the vector space X := ff : E ! R boundedg.


For f 2 X, dene
kf k := sup jf j :
E

Prove that k k is a norm.


Denition 24 Given a set E and functions fn : E ! R, n 2 N, we say that
the sequence of functions ffn g converges pointwise in a set F
E to some
function f : F ! R if
lim fn (x) = f (x)
n!1

for all x 2 F . We say that the sequence of functions ffn g converges uniformly
in a set F E to some function f : F ! R if
lim sup jfn

f j = 0:

n!1 F

Example 25 Let E = [0; 1] and let fn (x) = xn , x 2 [0; 1]. Then


lim xn =

n!1

0
1

if 0 x < 1;
if x = 1;

and so the sequence ffn g converges pointwise in [0; 1] to the function f dened
by
0 if 0 x < 1;
f (x) =
1 if x = 1:
On the other hand,
jfn (x)
Fix n 2 N. We claim that

xn
0

f (x)j =

sup jfn

if 0 x < 1;
if x = 1:

f j = 1:

[0;1]

To see this, note that


jfn f j ([0; 1]) = [0; 1) :
p
Indeed, if 0 y < 1, let x := n y. Then xn = y.
Lets prove that
sup [0; 1) = 1:
Note that 1 is an upper bound for the set [0; 1), and for any s < 1, if s < 0
then s is not an upper bound of [0; 1), while if 0 s < 1, then the middle point
1+s
2 belongs to [0; 1) and is greater than s. Thus, s is not an upper bound of
jfn f j ([0; 1]) = [0; 1), which proves that
sup jfn

f j = 1:

[0;1]

In turn,
lim sup jfn

n!1 [0;1]

f j = lim 1 = 1;
n!1

which shows that the sequence does not converges uniformly in [0; 1].
9

Friday, January 24, 2013


Example 26 (Continued) Now lets try to nd a smaller set of [0; 1] where
there is uniform convergence. Since the problem is at 1, lets remove a small
set near 1. Consider F = [0; a], where 0 < a < 1. Since fn f = fn 0 is
increasing in [0; a],
sup jfn

f j = max jfn j
[0;a]

[0;a]

= fn (a) = an ! 0:
Hence, there is uniform convergence in F .
Example 27 Consider the sequence of functions
fn (x) =
Let x 2 R. We have that

nx
;
1 + n4 x4

x 2 R:

fn (0) = 0 ! 0:

while if x 6= 0, we have that


lim

n!1

nx
n
= lim 4
n!1 n
1 + n4 x4

1
n4

x
1
= lim 3
n!1 n
+ x4

1
n4

x
x
=0
= 0:
0 + x4
+ x4

Thus, there is pointwise convergence to the function f = 0 in R.


To study uniform convergence, for each xed n we sketch the graph of the
function fn f in E. Since f = 0, we need to sketch the graph of fn . Note that
fn ( x) =

fn (x) ;

which means that the function fn is odd. So it is enough to study the function
nx
0 for x 0. Moreover,
for x 0. We have that fn (x) = 1+n
4 x4
lim

x!1

nx
x
= lim 4
x!1 x
1 + n4 x4

1
x4

n
1
= lim 3
x!1 x
+ n4

1
x4

n
n
=0
= 0:
0 + n4
+ n4

Hence the supremum of jfn j will be in the interior. To nd it, lets calculate
fn0 (x) =

n1 1 + n4 x4

nx 0 + 4n4 x3
2

(1 + n4 x4 )
5 4
n+n x
4n5 x4
n 3n5 x4
=
=
2
2
(1 + n4 x4 )
(1 + n4 x4 )

for n

3n5 x4

0. Thus fn0 (x)

Hence, fn is increasing in 0

0 for x4
n

1
p
4

1
3n4 ,

that is, for 0

and decreasing for x >

10

1
p
4

x
3

1
p
4

. It follows

that fn has a maximum at the point x =


sup jfn (x)

x2R

1
p
4

= fn

1
p
4

. It follows that

1
p
4
3
1
3

0 = max
x2R

nx
1 + n4 x4

1
nnp
4
3

n 3

1+

nx
1 + n4 x4

f (x)j = sup

x2R

1 + n4

1
p
4

4
3

9 0:

Hence, there is no uniform convergence in R.


1
Since x = n p
! 0, the problem is near zero. However if we remove only
4
3
0, we have that
sup jfn (x)

f (x)j = max

x2Rnf0g

x2Rnf0g

= fn
=

1
p
4

n 3

1
p
4
3

1+

nx
1 + n4 x4

1
3

9 0:

So in R n f0g, we still do not have uniform convergence. Lets remove a small


1
set near 0. Consider F = R n ( a; a). Since a > 0 and n p
! 0, there exists
4
3
n1 such that 0 <
x a. In turn,

1
p
4

< a for all n

sup jfn (x)

n1 . It follows that fn is decreasing for

f (x)j = max
x2F

x2F

nx
1 + n4 x4

= fn (a)
na
=
1 + n4 a 4
1
a
= 4 1
! 0:
n n4 + a4
This shows that there is uniform convergence in F .
Denition 28 A metric on a set X is a map d : X

X ! [0; 1) such that

(i) d (x; y) = 0 if and only if x = y,


(ii) d (x; y) = d (y; x) for all x; y 2 X (symmetry),
(iii) d (x; y)

d (x; z) + d (z; y) for all x; y; z 2 X (triangle inequality).

A metric space (X; d) is a set X endowed with a metric d. When there is


no possibility of confusion, we abbreviate by saying that X is a metric space.
11

Proposition 29 Let (X; k k) be a normed space. Then


d (x; y) := kx

yk

is a metric.
Proof. By property (i) in Denition 15, we have that 0 = d (x; y) = kx
and only if x y = 0, that is, x = y.
By property (ii) in Denition 15, we obtain that
d (y; x) = ky

xk = k 1 (x

y)k = j 1j kx

yk = kx

yk if

yk = d (x; y) :

Finally, by property (ii) in Denition 15,


d (x; y) = kx

yk = kx

z+z

yk

kx

zk + kz

yk = d (x; z) + d (z; y) :

Exercise 30 Prove that in R the function


d1 (x; y) :=

x
1 + jxj

y
1 + jyj

(3)

is a metric.
Monday, January 28, 2013

Topological Properties of the Euclidean Space

Denition 31 Given a point x0 2 RN and r > 0, the ball centered at x0 and


of radius r is the set
B (x0 ; r) := x 2 RN : kx

x0 k < r :

Remark 32 We can give a similar denition in a metric space (X; d), precisely,
given a point x0 2 X and r > 0, the ball centered at x0 and of radius r is the
set
B (x0 ; r) := fx 2 X : d (x; x0 ) < rg :
Denition 33 Given a set E RN , a point x 2 E is called an interior point
of E if there exists r > 0 such that B (x; r)
E. The interior E of a set
E
RN is the union of all its interior points. A subset U
RN is open if
every x 2 U is an interior point of U .
Example 34 The ball B (x0 ; r) is open. To see this, let x 2 B (x0 ; r). Then
B (x; r kx x0 k) is contained in B (x0 ; r). Indeed, if y 2 B (x; r kx x0 k),
then
ky

x0 k

ky

xk + kx

x0 k < r

and so y 2 B (x0 ; r).


12

kx

x0 k + kx

x0 k = r;

Example 35 Some simple examples of sets that are open and of some that are
not.
(i) The set (a; 1) = fx 2 R : x > ag is open. Indeed, if x > a, take r :=
x a > 0. Then B (x; r) (a; 1). Similarly, the set ( 1; a) is open.
(ii) The set (a; b) = fx 2 R : a < x < bg is open. Indeed, given a < x < b,
take r := min fb x; x ag > 0. Then B (x; r) (a; b).
(iii) The set (a; b] = fx 2 R : a < x bg is not open, since b belongs to the set
but there is no ball B (b; r) contained in (a; b].
Example 36 Consider the set
U = R n f0g [

1
: n2N
n

Lets prove that U is open. If x < 0, take r = x > 0, then B (x; r) = ( 2x; 0)
1
< x < n1 ,
U . If x > 1, take r = x 1, then B (x; r) = (1; 2x 1) U . If n+1
n
o
1
1
take r = min n1 x; x n+1
, then B (x; r) U . Hence, U is open.
= n+1

Example 37 Consider the set

1
: n2N
n

E =Rn

Lets prove that E is not open. The point x = 0 belongs to E, but for every
r > 0, by the Archimedean principle we can nd n 2 N such that n > 1r , and
so 0 < n1 < r, which shows that n1 2 ( r; r). Since n1 does not belong to E, the
ball ( r; r) is not contained in E for any r > 0. Hence, E is not open.
The main properties of open sets are given in the next proposition.
In what follows by an arbitrary family of sets of RN we mean that there
exists a set I and a function
f : I ! P RN

2 I 7! f ( ) = U

We write fU g or fU gI or fU g

2I

to denote the set ff ( ) :

2 Ig.

Proposition 38 The following properties hold:


(i) ; and RN are open.
RN , i = 1; : : : ; n, is a nite family of open sets of RN , then
\ Un is open.
S
(iii) If fU g is an arbitrary collection of open sets of RN , then
U is open.
(ii) If Ui
U1 \

13

Proof. To prove (ii), let x 2 U1 \ \ UM . Then x 2 Ui for every i = 1; : : : ; n,


and since Ui is open, there exists ri > 0 such that B (x; ri )
Ui . Take r :=
min fr1 ; : : : ; rn g > 0. Then
B (x; r)

U1 \

\ Un ;

which shows that U1 \


\ Un S
is open.
To prove (iii), let x 2 U :=
U . Then there is such that x 2 U and
since U is open, there exists r > 0 such that B (x; r) U
U . This shows
that U is open.
Remark 39 The same proof continues to hold for a metric space.
Properties (i)(iii) are used to dene topological spaces.
Denition 40 Let X be a nonempty set and let be a family of sets of X. The
pair (X; ) is called a topological space if the following hold.
(i) ;; X 2 .
(ii) If Ui 2

for i = 1; : : : ; M , then U1 \ : : : \ UM 2 .

(iii) If fU g is an arbitrary collection of elements of , then


The elements of the family

are called open sets.

U 2 .

Remark 41 The intersection of innitely many open sets is not open in gen1 1
eral. Take Un :=
n ; n for n 2 N. Then
1
\

n=1

1 1
;
n n

= f0g ;

but f0g is not open. Indeed, for every r > 0, the ball ( r; r) is not contained in
f0g.
Remark 42 Proposition 38 shows that the family of open sets in RN dened
in Denition 33 is a a topology, called the Euclidean topology. Unless specied,
in RN we will always consider the Euclidean topology.
Example 43 Given a nonempty set X, there are always at least two topologies
on X, namely,
1 = f;; Xg
(so according to

1,

the only open sets are the empty set and X) and
2

(so according to

= fall subsets of Xg

every set E

X is open).

14

Exercise 44 Prove that in RN the norms


kxk`1 := max fjx1 j ; : : : ; jxN jg ;
kxk`1 := jx1 j +

+ jxN j ;

kxk`p := (jx1 j +

p 1=p

+ jxN j )

generate the same topology.


Wednesday, January 30, 2013
Remark 45 For a topological space (X; ), given a point x 2 X, a neighborhood of x is an open set containing x. Neighborhoods play the role of balls in
metric spaces. Thus, given a set E
X, a point x 2 E is called an interior
point of E if there exists a neighborhood U of x such that U E.
A subset C RN is closed if its complement RN n C.
The main properties of closed sets are given in the next proposition.
Proposition 46 The following properties hold:
(i) ; and RN are closed.
RN , i = 1; : : : ; n, is a nite family of closed sets of RN , then
[ Cn is closed.
T
(iii) If fC g is an arbitrary collection of closed sets of RN , then
C is
closed.
(ii) If Ci
C1 [

The proof follows from Proposition 38 and De Morgans laws. If fE g is


an arbitrary collection of subsets of a set RN , then De Morgans laws are
!
[
\
N
R n
E
=
RN n E ;
N

R n

RN n E

Denition 47 Given a set E RN , a point x 2 RN is an accumulation point,


or cluster point of E if for every r > 0 the ball B (x; r) contains at least one
point of E di erent from x.
Note that x does not necessarily belong to the set E.
Remark 48 An interior point of E

RN is an accumulation point of E.

15

Example 49 Consider the set


1
n

E :=

n2N

[ 1+

1
n

:
n2N

We want to prove that 0 and 1 are accumulation points of E. Note that 0 2


= E,
while 1 2 E (so accumulation points may or may not be in the set E). For
r > 0, by taking a natural number n > 1r , we have that 0 < n1 < r, and so
1
1
n 2 B (0; r) \ E (of course n 6= 0). This shows that 0 is an accumulation points
of E.
Similarly, for r > 0, by taking a natural number n > 1r , we have that 0 <
1
1
1
n < r, and so 1 < 1 + n < 1 + r, which shows that 1 + n 2 B (1; r) \ E (of
1
course 1 + n 6= 1). This shows that 1 is an accumulation points of E. Next we
show that there are no other accumulation points of E.
Indeed, if x < 0, take r = x > 0, then B (x; r) = ( 2x; 0) does not intersect
E. If x > 2, take r = x 1, then Bn (x; r) = (1; 2x o 1) does not intersect E.
If

1
n+1

<x<

1
n,

take r = min

not intersect E. If x =

1
n,

If 1 +

< x < 1+

x; x

1
n+1

with n > 1, take r = min

B (x; r) intersects E only in


1
n+1

1
n

1
n,

1
n.

1
n+1 ,

1
n

Hence, U n
is open.
take r = min 1 + n1 x; x

then B (x; r) does


o
1
1
1
n+1 ; n 1
n . Then

then B (x; r) does not intersect E. If x = 1+ n1 , take r = min


Then B (x; r) intersects E only in 1 + n1 .

1+
n
1
n

1
n+1

1
1
n+1 ; n 1

1
n+1 ,
1
n

o
.

The set of all accumulation points of E is denoted acc E.


Remark 50 Note take if x 2 RN is an accumulation point of E, then by taking
r = n1 , n 2 N, there exists a sequence fxn g
E with xn 6= x for all n 2 N
such that kxn xk < n1 ! 0. Thus fxn g converges to x. Conversely, if there
exists fxn g E with xn 6= x for all n 2 N such that kxn xk ! 0, then x is
an accumulation point of E.
Exercise 51 Prove that a set E
accumulation points.

RN is closed if and only if it contains all its

Denition 52 Given a set E


RN , a point x 2 RN is a boundary point of
E if for every r > 0 the ball B (x; r) contains at least one point of E and one
point of RN n E. The set of boundary points of E is denoted @E.
Exercise 53 Prove that a set E
boundary points.

RN is closed if and only if it contains all its


Friday, February 1, 2013

16

Functions

Consider a function f : E ! RM , where E


RN . The set E is called the
domain of f . If E is not specied, then E should be taken to be the largest set
of x for which f (x) makes sense. This means that:
If there are even roots, their arguments should be nonnegative. If there are
logarithms, their arguments should be strictly positive. Denominators should
be dierent from zero. If a function is raised to an irrational number, then the
function should be nonnegative.
Given a set F
E, the set f (F ) = y 2 RM : y = f (x) for some x 2 F
is called the image of F through f . The function f is said to be bounded from
above in F , bounded from below in F , bounded in F if the set f (F ) is bounded
from above, bounded from below, bounded, respectively.
Given a set G
R, the set f 1 (G) = fx 2 E : f (x) 2 Gg is called the
inverse image of F through f . It has NOTHING to do with the inverse function.
It is just one of those unfortunate cases in which we use the same symbol for
two dierent objects.
The graph of a function is the set of RN RM dened by
gr f = f(x; f (x)) : x 2 Eg :
A function f is said to be
one-to-one or injective if f (x) 6= f (z) for all x; z 2 E with x 6= z.
If f : E ! F , where E; F

R, then f is said to be

onto or surjective if f (E) = F ,


bijective or invertible if it is one-to-one and onto. The function f 1 : F !
E, which assigns to each y 2 F = f (E) the unique x 2 E such that
f (x) = y, is called the inverse function of f .
If N = M = 1 a function f : E ! R is said to be
increasing if f (x)

f (y) for all x; y 2 E with x < y,

strictly increasing if f (x) < f (y) for all x; y 2 E with x < y,


decreasing if f (x)

f (y) for all x; y 2 E with x < y,

strictly decreasing if f (x) > f (y) for all x; y 2 E with x < y,


monotone if one of the four property above holds.

17

Limits of Functions

Denition 54 Let E RN , let x0 2 RN be an accumulation point of E, and


let f : E ! RN . We say that a number ` 2 RM is the limit of f (x) as x
approaches x0 if for every " > 0 there exists a real number = ("; x0 ) > 0
with the property that
kf (x) `k < "
for all x 2 E with 0 < kx

x0 k < . We write

lim f (x) = `

x!x0

or

f (x) ! ` as x ! x0 :

Note that even when x0 2 E, we cannot take x = x0 since in the denition


we require 0 < kx x0 k.
Remark 55 If (X; dX ) and (Y; dY ) are two metric spaces, E X, x0 2 E is an
accumulation point of E and f : E ! Y , we say that ` 2 Y is the limit of f (x)
as x approaches x0 if for every " > 0 there exists a real number = ("; x0 ) > 0
with the property that
dY (f (x) ; `) < "
for all x 2 E with 0 < dX (x; x0 ) < . We write
lim f (x) = `:

x!x0

If (X; X ) and (Y; Y ) are two topological spaces, E


X, x0 2 E is an accumulation point of E and f : E ! Y , we say that ` 2 Y is the limit of f (x) as
x approaches x0 if for every neighborhood V of ` there exists a neighborhood U
of x0 with the property that
f (x) 2 V
for all x 2 E with x 2 U n fx0 g. We write
lim f (x) = `:

x!x0

Note that unless the space Y is Hausdor , the limit may not be unique.
Remark 56 Let E RN , let x0 2 RN be an accumulation point of E, and let
f : E ! RN . Assume that there exists
lim f (x) = `:

x!x0

Then for every " > 0 there exists a real number


property that
kf (x) `k < "

("; x0 ) > 0 with the


(4)
N

for all x 2 E with 0 < kx x0 k < . if F E is a subset such that x0 2 R be


an accumulation point of E and we consider the resctriction of f to F , denoted

18

f jF , for every " > 0 let


(4) we have that

> 0 be the number given in (4). Then by restricting


kf (x)

for all x 2 F with 0 < kx

`k < "

x0 k < . Hence, there exists


lim f jF (x) = `:

x!x0

It follows that if we can nd two sets F


E and G E such that x0 2 acc F
and x0 2 acc G
lim f jF (x) = `1 6= `2 = lim f jG (x) ;
x!x0

x!x0

then the limit over E cannot exist.


Example 57 Lets study the limit
xm y
;
(x;y)!(0;0) x2 + y 2
lim

where m 2 N. In this case f (x; y) = xx2 +yy2 and the domain is R2 n f(0; 0)g.
m
2, we have that the limit is 0. Indeed, using the fact that jxj =
p For p
x2
x2 + y 2 , we have
xm y
x2 + y 2

m=2

x2 + y 2

jxj jyj
0 = 2
x + y2

x2 + y 2
x2 + y 2

1=2

= x2 + y 2

(m 1)=2

!0

as (x; y) ! (0; 0). On the other hand, if m = 1, taking y = x, we have that


lim f (x; x) = lim

x!0 x2

x!0

x2
1
= ;
+ x2
2

while taking y = 0, we have that


f (x; 0) =

x2

0
= 0 ! 0;
+0

and so the limit does not exist.


Remark 58 Note that the degree of the numerator is m + 1 and the degree
of the numerator is 2, so that in this particular example the limit exists if the
degree of the numerator is higher than the degree of the numerator, that is, if
m + 1 > 2.
The next example shows that checking the limit on every line passing through
x0 is not enough to gaurantee the existence of the limit.
Example 59 Let
f (x; y) :=

1
0

if y = x2 ; x 6= 0;
otherwise.
19

Given the line y = mx, the line intersects the parabola y = x2 only in 0 and in
at most one point. Hence, if x is very small,
f (x; mx) = 0 ! 0
as x ! 0. However, since f x; x2 = 1 ! 1 as x ! 0, the limit does not exists.
Example 60 Study the limit
xm y
;
y2

lim

(x;y)!(0;0) x2

where m 2 N. In this case f (x; y) =


Taking y = 0, we have that
f (x; 0) =

xm y
x2 y 2

and the domain is R2 nf(x; x) : x 2 Rg.

0
x2

= 0 ! 0:

Let us take y = x + xa , where a > 1. Then


f (x; x + xa ) =

xm (x + xa )

2
x2
x2 (x + xa )
xm+1 + xm+a
=
:
2xa+1 + x2a

xm+1 + xm+a
x2 2xa+1 x2a

Take a = m. Then
f (x; x + xm ) =
=

xm+1 + x2m
=
2xm+1 + x2m
1 + xm 1
9 0:
2 + xm 1

xm+1 1 + xm
xm+1 (2 + xm

1
1)

Hence the limit does not exist.


Remark 61 Note that the degree of the numerator is m + 1 and the degree of
the numerator is 2, but in this case the limit never exists no matter how high is
the degree of the numerator.
Monday, February 4, 2013
Denition 62 Let E RN , let x0 2 RN be an accumulation point of E, and
let f : E ! R. We say that
1 is the limit of f (x) as x approaches x0 if for every L > 0 there exists
a real number = (L; x0 ) > 0 with the property that
f (x) > L
for all x 2 E with 0 < kx

x0 k < . We write

lim f (x) = 1

x!x0

or

20

f (x) ! 1 as x ! x0 ;

1 is the limit of f (x) as x approaches x0 if for every L > 0 there exists


a real number = (L; x0 ) > 0 with the property that
f (x) <
for all x 2 E with 0 < kx

x0 k

lim f (x) =

x!x0

. We write
or

f (x) !

1 as x ! x0 :

Theorem 63 Let E
RN and let x0 2 RN be an accumulation point of E.
Given a function f : E ! RM , if the limit
lim f (x)

x!x0

exists, it is unique.
Proof. Assume by contradiction that there exist
lim f (x) = ` and

lim f (x) = L

x!x0

x!x0

with ` 6= L. Then k` Lk > 0. Fix 0 < " <


there exists 1 > 0 with the property that
kf (x)
for all x 2 E with 0 < kx x0 k <
exists 2 > 0 with the property that
kf (x)

1
2

k`

Lk. Since limx!x0 f (x) = `,

`k < "
1,

while, since limx!x0 f (x) = L, there


Lk < "

for all x 2 E with 0 < kx x0 k < 2 .


Take = min f 1 ; 2 g > 0 and take x 2 E with 0 < kx x0 k < . Note that
such x exists because x0 is an accumulation point of E. Then by the properties
of the norm,
k`

Lk = k`

f (x) + f (x)

< " + " < k`

which implies that k`

Lk < k`

Lk

Lk ;

k`

f (x)k + kf (x)

Lk

Lk. This contradiction proves the theorem.

Remark 64 This proof continues to hold in metric spaces, since we only used
the fact that the balls B (`; ") and B (L; ") are disjoint whenever 0 < " <
1
Lk. For topological spaces in general the limit is not unique. Given
2 k`
(X; X ) and (Y; Y ) are two topological spaces, E X, x0 2 E is an accumulation point of E and f : E ! Y , it can be shown that the limit is unique if the
space Y is Hausdor . A topological space Y is a Hausdor space, if for every
x and y 2 Y , with x 6= y, there exist disjoint neighborhoods of x and y.
21

We now list some important operations for limits.


Theorem 65 Let E
RN and let x0 2 RN be an accumulation point of E.
Given two functions f ; g : E ! RM , assume that there exist
lim g (x) = `2 2 RM :

lim f (x) = `1 2 RM ;

x!x0

x!x0

Then
(i) there exists lim (f + g) (x) = `1 + `2 ,
x!x0

(ii) there exists lim (f g) (x) = `1 `2 ,


x!x0

(iii) if M = 1, `2 6= 0 and F := fx 2 E : g (x) 6= 0g, then x0 is an accumulaf


`1
tion point of F and there exists lim
(x) = .
x!x0
g F
`2
Proof. Parts (i) and (ii) are left as an exercise. We prove part (iii). The fact
that x0 is an accumulation point of F is left as an exercise. Write
f (x)
g (x)

`1
f (x) `2 g (x) `1
f (x) `2 `1 `2 g (x) `1
=
=
`2
g (x) `2
g (x) `2
1
=
j`2 (f (x) `1 ) + `1 (g (x) `2 )j
jg (x)j j`2 j
1 j`1 j
1
jf (x) `1 j +
jg (x) `2 j :
jg (x)j
jg (x)j j`2 j
1
jg(x)j

Thus we need to bound

from above, or, equivalently, we need jg (x)j to

stay away from zero. Since `2 6= 0, taking " = j`22 j > 0, there exist
that
j`2 j
jg (x) `2 j
2
for all x 2 E with 0 < kx x0 k < 1 . Hence,
jg (x)j = jg (x)

`2 j

for all x 2 E with 0 < kx


f (x)
g (x)

`1
`2

Given " > 0 there exist

j`2 j

x0 k <

1.

jg (x)

`2 j

j`2 j

j`2 j
j`2 j
=
2
2

It follows that

1
1 j`1 j
jf (x) `1 j +
jg (x) `2 j
jg (x)j
jg (x)j j`2 j
2
2 j`1 j
jf (x) `1 j +
jg (x) `2 j :
j`2 j
j`2 j j`2 j
2

> 0 such that


jf (x)

`1 j
22

" j`2 j
4

> 0 such

for all x 2 E with 0 < kx

x0 k <

and

> 0 such that


2

jg (x)

" j`2 j
4 (1 + j`1 j)

`2 j

for all x 2 E with 0 < kx x0 k < 3 . Then for for all x 2 E with 0 <
kx x0 k < = min f 1 ; 2 ; 3 g, we have that
f (x)
g (x)

2
2 j`1 j
jf (x) `1 j +
jg (x)
j`2 j
j`2 j j`2 j
" " j`1 j
" "
+
+ 1 = ":
2 2 1 + j`1 j
2 2

`1
`2

`2 j

This completes the proof.


Remark 66 The previous theorem continues to hold if `1 ; `1 2 [ 1; 1], provided we avoid the cases 1 1, 01, 00 , 1
1.
Theorem 67 (Squeeze Theorem) Let E RN and let x0 2 RN be an accumulation point of E. Given three functions f; g; h : E ! R, assume that there
exist
lim f (x) = lim g (x) = `:
x!x0

x!x0

and that f (x)

h (x)

g (x) for every x 2 E. Then there exists lim h (x) = `.


x!x0

Proof. We prove the case in which ` 2 R and leave the cases ` = 1 and
` = 1 as an exercise. Given " > 0 there exist 1 > 0 such that
jf (x)
for all x 2 E with 0 < kx

x0 k <

`j

and

jg (x)

"
2

`j

> 0 such that


"

for all x 2 E with 0 < kx x0 k < 2 . Then for for all x 2 E with 0 <
kx x0 k < = min f 1 ; 2 g, we have that
`

"

f (x)

g (x)

h (x)

` + ":

Hence,
for all x 2 E with 0 < kx

jh (x)

`j

"

x0 k < , which shows that lim h (x) = `.


x!x0

Theorem 68 Let E
RN and let x0 2 R be an accumulation point of E.
Given two functions f ; g : E ! RM , assume that there exists
lim f (x) = 0;

x!x0

and that g is bounded, that is, kg (x)k


Then there exists lim (f g) (x) = 0.
x!x0

23

L for all x 2 E and for some L > 0.

Proof. Given " > 0 there existd

> 0 such that


"
kf (x) 0k <
1+L

for all x 2 E with 0 < kx


j(f g) (x)

x0 k < . Hence, by the CauchySchwarzs inequality

0j = j(f g) (x)j

for all x 2 E with 0 < kx

kf (x)k kg (x)k <

"
L<"
1+L

x0 k < .

Example 69 The previous theorem can be used for example to show that for
a>0
1
lim xa sin = 0:
x!0
x
Wednesday, February 6, 2013
We next study the limit of composite functions.
Theorem 70 Let E RN , F
RM and let x0 2 R be an accumulation point
of E. Given two functions f : E ! F and g : F ! RP , assume that there exist
lim f (x) = ` 2 RM ;

x!x0

that ` is an accumulation point of F , and that there exists


lim g (y) = L 2 RP :

y!`

Assume that either there exists 1 > 0 such that f (x) 6= ` for all x 2 E with 0 <
jx x0 j
1 , or that ` 2 F and g (`) = L. Then there exists lim g (f (x)) = L.
x!x0

Proof. Fix " > 0 and nd

= ("; `) > 0 such that


kg (y)

Lk < "

for all y 2 F with 0 < ky `k < .


Since limx!x0 f (x) = `, there exists
kf (x)

(5)
(x0 ; ) > 0 such that

`k <

for all x 2 E with 0 < kx x0 k < 1 .


We now distinguish two cases.
Case 1: Assume that f (x) 6= ` for all x 2 E with 0 < kx x0 k < 1 . Then
taking = min f 1 ; 2 g, we have that for all x 2 E with 0 < kx x0 k < ,
0 < kf (x)

`k < :

Hence, taking y = f (x), by (5), it follows that


kg (f (x))
24

Lk < "

for all x 2 E with 0 < kx


L.

1.

x0 k < . This shows that there exists lim g (f (x)) =


x!x0

Case 2: Assume that ` 2 F and g (`) = L. Let x 2 E with 0 < kx


If f (x) = `, then g (f (x)) = L, and so
kg (f (x))

x0 k <

Lk = 0 < ";

while if f (x) 6= `, then taking y = f (x), by (5), it follows that


kg (f (x))

Lk < ":

Example 71 Lets prove that the previous theorem fails without the hypotheses
that either f (x) 6= ` for all x 2 E near x0 or ` 2 F , L 2 R and g (`) = L.
Consider the function
1 if y 6= 0;
g (y) :=
2 if y = 0:
Then there exists
lim g (y) = 1:

y!0

So L = 1. Consider the function f (x) := 0 for all x 2 R. Then for every


x0 2 R, we have that
lim f (x) = 0:
x!x0

So ` = 0. However, g (f (x)) = g (0) = 2 for all x 2 R. Hence,


lim g (f (x)) = lim 2 = 2 6= 1;

x!x0

x!x0

which shows that the conclusion of the theorem is violated.


Example 72 We list below some important limits.
sin x
1 cos x
1
log (1 + x)
= 1;
lim
= ;
lim
= 1;
2
x!0
x!0
x
x
2
x
a
(1 + x)
1
ex 1
lim
= a for a 2 R;
lim
= 1:
x!0
x!0
x
x
lim

x!0

Note that the previous theorem can be used to change variables in limits.
Example 73 Lets try to calculate
lim

x!0

log (1 + sin x)
:
x

For sin x 6= 0, we have


log (1 + sin x)
log (1 + sin x) sin x
log (1 + sin x) sin x
=
=
:
x
x
sin x
sin x
x
25

Since limx!0

sin x
x

= 1, it remains to study
lim

x!0

log (1 + sin x)
:
sin x

and the function f (x) = sin x. As x ! 0,


Consider the function g (y) = log(1+y)
y
we have that sin x ! 0 = `, while
lim

y!0

log (1 + y)
= 1:
y
2; 2

Moreover sin x 6= 0 for all x 2 E :=


previous theorem to conclude that

n f0g. Hence, we can apply the

lim

log (1 + sin x)
= 1:
sin x

lim

log (1 + sin x)
= 1:
x

x!0

In turn,
x!0

Continuity

Denition 74 Let E RN . A point x0 2 E is called an isolated point of E if


there exists > 0 such that
B (x0 ; ) \ E = fx0 g :
It is clear that if a point of the set E is not an isolated point of E then it is
an accumulation point of E.
Denition 75 Let E
RN and let x0 2 E. Given a function f : E ! RM
we say that f is continuous at x0 if for every " > 0 there exists a real number
= ("; x0 ) > 0 such that for all x 2 E with kx x0 k < we have
kf (x)

f (x0 )k < ":

If f is continuous at every point of E we say that f is continuous on E and we


write f 2 C (E) or f 2 C 0 (E).
Remark 76 If (X; dX ) and (Y; dY ) are two metric spaces, E
X, x0 2 E,
and f : E ! Y , we say that f is continuous at x0 if for every " > 0 there exists
a real number = ("; x0 ) > 0 with the property that
dY (f (x) ; f (x0 )) < "
for all x 2 E with dX (x; x0 ) < .
If (X; X ) and (Y; Y ) are two topological spaces, E
X, x0 2 E, and
f : E ! Y , we say that f is continuous at x0 if for every neighborhood V of
f (x0 ) there exists a neighborhood U of x0 with the property that
f (x) 2 V
for all x 2 E with x 2 U .
26

RN and let x0 2 E. Given a function f : E ! RM ,

Theorem 77 Let E

(i) if x0 is an isolated point of E then f is continuous at x0 ;


(ii) if x0 is an accumulation point of E then f is continuous at x0 if and only
if there exists lim f (x) = f (x0 ).
x!x0

Proof of part (i). If x0 is an isolated point of E then there exists


that
B (x0 ; 0 ) \ E = fx0 g :

> 0 such

Fix " > 0 and take := 0 in the denition of continuity. Clearly if x 2 E and
kx x0 k < then necessarily x = x0 so that we have kf (x) f (x0 )k = 0.
Exercise 78 Prove that the functions sin x, cos x, xn , where n 2 N, are continuous.
The following theorems follows from the analogous results for limits.
Theorem 79 Let E RN and let x0 2 E. Given two functions f , g : E ! R
assume that f and g are continuous at x0 . Then
(i) f + g and f g are continuous at x0 ;
(ii) if g (x0 ) 6= 0 then
continuous at x0 .

f
restricted to the set F := fx 2 E : g (x) 6= 0g is
g

Example 80 In view of Exercise 78 and the previous theorem, the functions


sin x
cos x
tan x = cos
x and cot x = sin x are continuous in their domain of denition.
Theorem 81 Let E
RN , F
RM and let x0 2 E. Given two functions
P
f : E ! F and g : F ! R assume that f is continuous at x0 and that g is
continuous at f (x0 ). Then g f : E ! RP is continuous at x0 .
We now discuss the continuity of inverse functions and of composite functions. If a continuous function f is invertible its inverse function f 1 may not
be continuous.
Example 82 Let
f (x) :=
Then f

x
x

if 0 x
if 2 < x

1;
3:

: [0; 2] ! R is given by
f

(x) :=

x
x+1

which is not continuous at x = 1.

27

if 0 x
if 1 < x

1;
2;

We will see that this cannot happen if E is an interval or a compact set.


Theorem 83 Let K
RN be a closed and bounded set and let f : K ! RN
be one-to-one and continuous. Then the inverse function f 1 : f (K) ! RN is
continuous.
Theorem 84 Let I
R be an interval and let f : I ! R be one-to-one and
continuous. Then the inverse function f 1 : f (I) ! R is continuous.
Friday, February 8, 2013

Directional Derivatives and Dierentiability

Let E RN , let f : E ! R and let x0 2 E. Given a direction v 2 RN , let L be


the line through x0 in the direction v, that is,
L := x 2 RN : x = x0 + tv; t 2 R ;
and assume that x0 is an accumulation point of the set E \ L. The directional
derivative of f at x0 in the direction v is dened as
f (x0 + tv)
@f
(x0 ) := lim
t!0
@v
t

f (x0 )

provided the limit exists in R. In the special case in which v = ei , the directional
@f
derivative @e
(x0 ), if it exists, is called the partial derivative of f with respect
i
@f
to xi and is denoted @x
(x0 ) or fxi (x0 ) or Di f (x0 ).
i
Remark 85 The previous denition continues to hold if in place of RN one
takes a normed space V , so that f : E ! R where E V .
Remark 86 If N = 1, taking the direction v = 1, we have that
@f
f (x0 + t)
(x0 ) = lim
t!0
@1
t
which is the denition of derivative

df
dx

f (x0 )

(x0 ) or f 0 (x0 ).

Next we show that even if the directional derivatives at x0 exist and are
nite in every direction, then f does not have to be continuous at x0 .
Example 87 Let
f (x; y) :=

x2 y
x4 +y 2

if (x; y) 6= (0; 0) ;
if (x; y) = (0; 0) :

Lets nd the directional derivatives of f at 0. Given a direction v = (v1 ; v2 ),


with v12 + v22 = 1, we have
f (0 + tv1 ; 0 + tv2 ) = 0:
28

It follows that
f (0 + tv1 ; 0 + tv2 )
t

f (0; 0)

f (0 + tv1 ; 0 + tv2 )
t

f (0; 0)

(tv1 )2 tv2
(tv1 )4 +(tv2 )2

t
t3 v12 v2
= 5 4
:
t v1 + t3 v22

If v2 = 0 then

as t ! 0, so

@f
@x

0
t5 v14

+0

=0!0

(0; 0) = 0. If v2 6= 0, then,

f (0 + tv1 ; 0 + tv2 )
t
so

In particular,
We have

f (0; 0)

v12 v2
v12 v2
v12
!
=
;
t2 v14 + v22
0 + v22
v2

v2
@f
(0; 0) = 1 :
@v
v2
@f
@y

(0; 0) =

0
1

= 0. Now lets prove that f is not continuous at 0.

f (x; 0) =
as x ! 0, while

0
=0!0
0 + y2

x2 x2
1
1
= !
4
4
x +x
2
2
as x ! 0. Hence, the limit lim(x;y)!(0;0) f (x; y) does not exists and so f is not
continuous at (0; 0). Note that f is continuous at all other points (x; y) 6= (0; 0)
by Theorem 79, since h (x; y) = x and g (x; y) = y are continuous functions in
R2 .
f x; x2 =

The previous examples show that in dimension N 2 partial derivatives do


not give the same kind of results as in the case N = 1. To solve this problem, we
introduce a stronger notion of derivative, namely, the notion of dierentiability.
We recall that a function T : RN ! R is linear if
T (x + y) = T (x) + T (y)
for all x; y 2 RN and

T (sx) = sT (x)
PN
for all s 2 R and x 2 R . Write x = i=1 xi ei . Then by the linearity of T ,
!
N
N
X
X
T (x) = T
xi e i =
xi T (ei ) :
N

i=1

i=1

Dene b := (T (e1 ) ; : : : ; T (eN )) 2 R . Then the previous calculation shows


that
T (x) = b x
for all x 2 RN :
29

Denition 88 Let E RN , let f : E ! R, and let x0 2 E be an accumulation


point of E. The function f is dierentiable at x0 if there exists a linear function
T : RN ! R (depending on f and x0 ) such that
lim

f (x)

x!x0

f (x0 ) T (x
kx x0 k

x0 )

= 0:

provided the limit exists. The function T , if it exists, is called the dierential of
f at x0 and is denoted df (x0 ) or dfx0 .
Exercise 89 Prove that if N = 1, then f is di erentiable at x0 if and only
there exists the derivative f 0 (x0 ) 2 R.
Remark 90 The previous denition continues to hold if in place of RN one
takes a normed space V , so that f : E ! R where E
V . In this case,
however, we require T : V ! R to be linear and continuous. Note that in RN
every linear function is continuous.
The next theorem shows that dierentiability in dimension N
same role of the derivative in dimension N = 1.

2 plays the

Theorem 91 Let E RN , let f : E ! R, and let x0 2 E be an accumulation


point of E. If f is di erentiable at x0 , then f is continuous at x0 .
Proof. Let T be the dierential of f at x0 . Write T (x) = b x. We have
f (x)

f (x0 ) = f (x)
=

f (x0 )

f (x)

b (x

f (x0 ) b (x
kx x0 k

x0 ) + b (x
x0 )

kx

x0 )

x0 k + b (x

x0 ) :

Hence, by Cauchys inequality for x 2 E, x 6= x0 ,


0

jf (x)
f (x)

f (x0 )j

f (x)

f (x0 ) b (x
kx x0 k

f (x0 ) b (x x0 )
kx x0 k + jb (x
kx x0 k
x0 )
kx x0 k + kbk kx x0 k

x0 )j

! j0j 0 + kbk 0 = 0

as x ! x0 . It follows by the squeeze theorem that f is continuous at x0 .


Monday, February 11, 2013
Next we study the relation between directional derivatives and dierentiability. The next theorem gives a formula for the vector b used in the previous
proof and hence determines T . Here we need x0 to be an interior point of E.
Theorem 92 Let E
RN , let f : E ! R be di erentiable at some point
x0 2 E and T be the di erential of f at x0 . Then

30

(i) all the directional derivatives of f at x0 exist and


@f
(x0 ) = T (v) ;
@v
(ii) for every direction v,
N

X @f
@f
(x0 ) =
(x0 ) vi :
@v
@xi
i=1

(6)

Proof. Since x0 is an interior point, there exists B (x0 ; r) E. Let v 2 RN be


a direction and x = x0 + tv. Note that for jtj < r, we have that
kx

x0 k = kx0 + tv

and so x0 + tv 2 B (x0 ; r)
dierentiable at x0 ,

x0 k = ktvk = jtj kvk = jtj 1 < r

E. Moreover, x ! x0 as t ! 0 and so, since f is

f (x0 + tv) f (x0 ) T (x0 + tv


f (x0 ) T (x x0 )
= lim
x!x0
t!0
kx x0 k
kx0 + tv x0 k
f (x0 + tv) f (x0 ) tT (v)
= lim
:
t!0
jtj

0 = lim

Since
lim

jtj
t

f (x)

is bounded by one, it follows that

f (x0 + tv)

t!0

f (x0 )
t

tT (v)

= lim

t!0

jtj f (x0 + tv) f (x0 )


t
jtj

tT (v)

= 0:

But then
0 = lim

t!0

f (x0 + tv)

f (x0 )
t

tT (v)

f (x0 + tv)
t!0
t

= lim

f (x0 )

T (v) ;

@f
(x0 ) = T (v).
which shows that there exists @v
PN
Part (ii) follows from the linearity of T . Indeed, writing v = i=1 vi ei , by
the linearity of T ,
!
N
N
N
X
X
X
@f
@f
(x0 ) = T (v) = T
vi ei =
vi T (ei ) =
(x0 ) vi :
@v
@x
i
i=1
i=1
i=1

Remark 93 If in the previous theorem x0 is not an interior point but for some
direction v 2 RN , the point x0 is an accumulation point of the set E \ L, where
L is the line through x0 in the direction v, then as in the rst part of the proof
@f
(x0 ) and
we can show that there exists the directional derivative @v
@f
(x0 ) = T (v) :
@v
31

x0 )

If all the partial derivatives of f at x0 exist, the vector


@f
@f
(x0 ) ; : : : ;
(x0 )
@x1
@xN

2 RN

is called the gradient of f at x0 and is denoted by rf (x0 ) or grad f (x0 ) or


Df (x0 ). Note that part (ii) of the previous theorem shows that
N
X
@f
dfx0 (v) = T (v) = rf (x0 ) v =
(x0 ) vi :
@x
i
i=1

(7)

for all directions v. Hence, only at interior points of E, to check dierentiability


it is enough to prove that
lim

f (x)

f (x0 ) rf (x0 ) (x
kx x0 k

x!x0

x0 )

= 0:

(8)

The next theorem gives an important su cient condition for dierentiability at


a point x0 .
Theorem 94 Let E RN , let f : E ! R, let x0 2 E . Assume that there ex@f
ists r > 0 such that B (x0 ; r) E and the partial derivatives @x
, j = 1; : : : ; N ,
j
exist for every x 2 B (x0 ; r) and are continuous at x0 . Then f is di erentiable
at x0 .
Wednesday, February 13, 2013
Example 95 Let
f (x; y) :=

x2 jyj
x2 +y 2

if (x; y) 6= (0; 0) ;
if (x; y) = (0; 0) :

Lets study continuity, partial derivatives and di erentiability. For (x; y) 6=


(0; 0), we have that f is continuous by Theorem 79, while for (x; y) = (0; 0), we
need to check that
lim
f (x; y) = f (0; 0) :
(x;y)!(0;0)

We have
0

jf (x; y)

f (0; 0)j =

x2 jyj
x2 + y 2

0 =

x2 jyj
x2 + y 2

x2 + y 2 jyj
= jyj ! 0
x2 + y 2

as (x; y) ! (0; 0). Hence, f is continuous at (0; 0).


Next, lets study partial derivatives. For (x; y) 6= (0; 0), by the quotient rule,
we have
2x jyj x2 + y 2
x2 jyj (2x + 0)
@f
(x; y) =
;
(9)
2
@x
(x2 + y 2 )
32

while for (x; y) = (0; 0),


@f
f (0 + t1; 0 + t0)
(0; 0) = lim
t!0
@x
t

f (0; 0)

t2 j0j
t2 +0

= lim

= lim

t!0 t3

t!0

= 0:

For y 6= 0, by the quotient rule, we have


y
x2 jyj
x2 + y 2
x2 jyj 0 + y 2
@f
(x; y) =
;
2
@y
(x2 + y 2 )

(10)

while at a point (x0 ; 0),


@f
f (x0 + t0; 0 + t1)
(x0 ; 0) = lim
t!0
@y
t
x2

0
If x0 = 0, then jtj
t x20 +t2 =
x0 6= 0, we have

jtj 0
t 0+t2

f (x0 ; 0)

x20 jtj
x20 +t2

= lim

t!0

@f
@y

= 0 ! 0 as t ! 0, so

jtj x20
:
2
t!0 t x2
0+t

= lim

(0; 0) = 0, while if

jtj x20
t x20
x20
x20
=
lim
=
lim
=
= 1;
2
2
t x20 + t2
x20 + 0
t!0
t!0+ t x0 + t2
t!0+ x0 + t2
jtj x20
t x20
x20
x20
lim
=
lim
=
lim
=
=
2
2
t x20 + t2
x20 + 0
t!0
t!0+ t x0 + t2
t!0+ x0 + t2
lim+

1:

Hence, @f
@y (x0 ; 0) does not exist at (x0 ; 0) for x0 6= 0, and so by Theorem 92, f
is not di erentiable at (x0 ; 0) for x0 6= 0.
@f
On the other hand, at points (x; y) with y 6= 0, we have that @f
@x and @y exist
in a small ball centered at (x; y) (see (9) and (10)) and they are continuous by
Theorem 79. Hence, we can apply Theorem 94 to conclude that f is di erentiable
at all points (x; y) with y 6= 0.
It remains to study di erentiability at (0; 0). By 8, we need to prove that
lim

f (x; y)

(x;y)!(0;0)

f (0; 0) rf (0; 0) ((x; y)


k(x; y) (0; 0)k

(0; 0))

= 0:

We have
f (x; y)

f (0; 0) rf (0; 0) ((x; y)


k(x; y) (0; 0)k

(0; 0))

=
=

x2 jyj
x2 +y 2

x2 jyj

3=2

(x2 + y 2 )

(0; 0) ((x; y)
p
x2 + y 2
:

Taking y = x, with x > 0, we get

(x2

x2 jxj
+

3=2
x2 )

x2 x3
(x2

3=2
x2 )

Hence, f is not di erentiable at (0; 0).


33

1
3=2

(2)

9 0:

(0; 0))

Friday, February 15, 2013


The next exercise shows that the conditions in Theorem 94 are su cient but
not necessary for dierentiability.
Example 96 Let
f (x; y) :=

1
x2 + y 2 sin x+y
0

if (x; y) 6= (0; 0) or x + y 6= 0;
otherwise.

Lets study continuity, partial derivatives and di erentiability. For x+y 6= 0, we


have that f is continuous by Theorem 79 and Theorem 81. For (x; y) = (0; 0),
we need to check that
lim

(x;y)!(0;0)

f (x; y) = f (0; 0) :

We have
0

jf (x; y)

f (0; 0)j =

x2 + y 2 sin

1
x+y

x2 + y 2 1 ! 0

as (x; y) ! (0; 0). Hence, f is continuous at (0; 0). At a point (x0 ; x0 ) with
x0 6= 0, taking x = x0 and y = x0 + t, where t ! 0. We have
2

f (x0 ; x0 + t) = x20 + ( x0 + t)
=
Take t =
2

1
+2n

x0 ; x0 +
2

x20

+ ( x0 + t)

sin

x0
1
sin :
t

1
x0 + t

. Then
1
+ 2n

=
=

x20
x20

+
+

1
+ 2n

1
+ 2n

x0 +
x0 +

!
!

sin

+ 2n

1 ! x20 + ( x0 + 0)

as n ! 1. Hence f is not continuous at (x0 ; x0 ) with x0 6= 0, and so by


Theorem 91, f is not di erentiable at (x0 ; x0 ) with x0 6= 0. Next, lets study
partial derivatives. If x + y 6= 0,
!
@f
1
1
1
2
2
(x; y) = (2x + 0) sin
+ x +y
cos
;
2
@x
x+y
x+y
(x + y)
!
1
1
1
@f
(x; y) = (0 + 2y) sin
+ x2 + y 2 cos
:
2
@y
x+y
x+y
(x + y)
@f
Since for every (x; y), with x + y 6= 0, @f
@x and @y exist in a small ball centered
at (x; y) and they are continuous by Theorem 79 and Theorem 81, we can apply
Theorem 94 to conclude that f is di erentiable at all points (x; y) with x+y 6= 0.

34

= 2x20 6= 0

Next, lets study the partial derivatives at (0; 0). We have


f (x; 0)
x

1
x2 + 0 sin x+0
f (0; 0)
=
0
x 0

as x ! 0, since sin x1 is bounded. Hence,


For (x0 ; x0 ), with x0 6= 0,

@f
@x

1
!0
x

= x sin

(0; 0) = 0. Similarly,

@f
@y

(0; 0) = 0.

f (x0 + t0; x0 + t1)


@f
(x0 ; x0 ) = lim
t!0
@x
t
2

= lim

x20 + ( x0 + t)

x20 +

1
+2n

x0 +

and t =

1
2 +2n

1
2n

sin

sin x0

1
x0 +t

0
:

shows that this limit does not exist. Indeed,

sin
2

1
+2n

1
+2n

= lim

x20 + ( x0 + t)

t!0

1
t

t!0

Taking t =

f (x0 ; x0 )

x20 +

1
2 +2n

x0 +

=
2

1
!

1
+2n

2x20
=1
0+

as n ! 1, while
x20 +

1
2n
1
2n

x0 +

sin

1
1
2n

x20 +

x0 +

1
2n

0
=

1
2n

0
1
2n

=0!0

as n ! 1. Similarly, @f
@y (x0 ; x0 ) does not exist.
Lets prove that f is di erentiable at (0; 0) at (0; 0). We have,
8 2 2
sin 1
0) @f
0) < (x +y
f (x; 0) f (0; 0) @f
p 2) 2x+y if x + y 6= 0;
@x (0; 0) (x
@y (0; 0) (y
q
x +y
=
: 0
2
2
otherwise.
(x 0) + (y 0)
p
1
x2 + y 2 sin x+y
if x + y 6= 0;
=
0
otherwise.
1
as (x; y) ! (0; 0) since sin x+y
is bounded and

di erentiable at (0; 0), even if nearby

@f
@x

and

@f
@y

!0
x2 + y 2 ! 0. Hence, f is

do not always exist.

Higher Order Derivatives

Let E
RN , let f : E ! R and let x0 2 E. Let i 2 f1; : : : ; N g and assume
@f
that there exists the partial derivatives @x
(x) for all x 2 E. If j 2 f1; : : : ; N g
i
and x0 is an accumulation point of E \ L, where L is the line through x0 in
35

the direction ej , then we can consider the partial derivative of the function
with respect to xj , that is,
@
@xj

@f
@xi

=:

@f
@xi

@2f
:
@xj @xi

Note that in general the order in which we take derivatives is important.


Example 97 Let
f (x; y) :=

y 2 arctan xy
0

if y 6= 0;
if y = 0:

x
y

If y 6= 0, then
@f
@
(x; y) =
@x
@x
=

y 2 arctan

= y2
1+

x
y

@
@x

x
y

y3
;
x2 + y 2

and
@f
@
(x; y) =
@y
@y

x
y

y 2 arctan

= 2y arctan

= 2y arctan

x
+ y2
y

1
1+

x
y

@
@y

x
y

xy 2
;
x2 + y 2

x
y

while at points (x0 ; 0) we have:


f (x0 + t; 0) f (x0 ; 0)
0 0
@f
(x0 ; 0) = lim
= lim
= 0;
t!0
t!0
@x
t
t
t2 arctan xt0
@f
f (x0 ; 0 + t) f (x0 ; 0)
(x0 ; 0) = lim
= lim
t!0
t!0
@y
t
t
x0
= lim t arctan
= 0;
t!0
t

where we have used the fact that arctan xt0 is bounded and t ! 0. Thus,
(
(
2
y3
@f
@f
if
y
=
6
0;
2y arctan xy x2xy
if y 6= 0;
2
2
x +y
+y 2
(x; y) =
(x; y) =
@x
@y
0
if y = 0;
0
if y = 0:
To nd

@2f
@y@x

(0; 0), we calculate

@2f
@
(0; 0) =
@y@x
@y
= lim

t!0

@f
@x
t3
0+t2

(0; 0) = lim

@f
@x

t!0

= lim 1 = 1;
t!0

36

(0; 0 + t)
t

@f
@x

(0; 0)

while
@2f
@
(0; 0) =
@x@y
@x

@f
(0; 0) = lim
t!0
@y
0 0
= lim
= lim 0 = 0:
t!0
t!0
t

Hence,

@2f
@x@y

(0; 0) 6=

@2f
@y@x

@f
@y

(0 + t; 0)

@f
@y

(0; 0)

(0; 0).

Exercise 98 Let
f (x; y) :=
Prove that

@2f
@x@y

(0; 0) 6=

@2f
@y@x

x3 y xy 3
x2 +y 2

if (x; y) 6= (0; 0) ;
if (x; y) = (0; 0) :

(0; 0).

The next important theorem shows that at "good points" the order in which
we take derivatives does not matter.
Theorem 99 (Schwartz) Let E
RN , let f : E ! R, let x0 2 E , and let
i; j 2 f1; : : : ; N g. Assume that there exists r > 0 such that B (x0 ; r) E and for
2
@f
f
@f
(x), @x
(x), and @x@j @x
(x) exist.
all x 2 B (x0 ; r), the partial derivatives @x
i
j
i
Assume also that

@2f
@xj @xi

is continuous at x0 . Then there exists

@2f
@xi @xj

(x0 ) and

@2f
@2f
(x0 ) =
(x0 ) :
@xi @xj
@xj @xi
Monday, February 18, 2013
Next we prove Taylors formula in higher dimensions. We set N0 := N [ f0g.
N
A multi-index
is a vector = ( 1 ; : : : ; N ) 2 (N0 ) . The length of a multiindex is dened as
j j := 1 +
+ N:
Given a multi-index

@
@x

, the partial derivative


@
@x

where x = (x1 ; : : : ; xN ). If

:=

@j
@x1 1

= 0, we set

Example 100 If N = 3 and

@xNN
@0f
@x0

is dened as
;

:= f .

= (2; 1; 0), then


@ (2;1;0)

@ (x; y; z)

(2;1;0)

37

@3
:
@x2 @y

and x 2 RN , we set

Given a multi-index
! :=
If

1!

N !;

x := x1 1

xNN :

= 0, we set x0 := 1.
Using this notation, we can extend the binomial theorem.

Theorem 101 (Multinomial Theorem) Let x = (x1 ; : : : ; xN ) 2 RN and let


n 2 N. Then
X

(x1 +

+ xN ) =

m ulti-in dex, j j=n

n!
x :
!

Proof. Exercise.
Given an open set U
RN , for every nonnegative integer m 2 N0 , we
m
denote by C (U ) the space of all functions that are continuous together with
1
T
their partial derivatives up to order m. We set C 1 (U ) :=
C m (U ).
m=0

Theorem 102 (Taylors Formula) Let U


RN be an open set, let f 2
m
C (U ), m 2 N, and let x0 2 U . Then for every x 2 U ,
X

f (x) =

m ulti-in dex, 0 j j m

where
lim

x!x0

We write Rm (x) = o (kx


as x ! x0 .

1 @ f
(x0 ) (x
! @x

x0 ) + Rm (x) ;

Rm (x)
m = 0:
kx x0 k

x0 k ) and we say that Rm is a little o of kx

x0 k

Example 103 Lets calculate the limit


y

(1 + x)
1
p
:
2
2
(x;y)!(0;0)
x +y
lim

By substituting we get 00 . Consider the function


y

f (x; y) = (1 + x)

1 = elog(1+x)

1 = ey log(1+x)

1;

which is dened in the set U := (x; y) 2 R2 : 1 + x > 0 . The function f is of


class C 1 . Lets use Taylors formula of order m = 1 at (0; 0),
f (x; y) = f (0; 0) +

@f
(0; 0) (x
@x

0) +

38

@f
(0; 0) (y
@y

0) + o

p
x2 + y 2 :

We have
@f
(x; y) =
@x
@f
(x; y) =
@y

@
ey log(1+x)
@x
@
ey log(1+x)
@y

1 = ey log(1+x) y

1 = ey log(1+x) y log (1 + x) ;

and so
f (x; y) = 0 + 0 (x

1
;
1+x

0) + 0 (y

0) + o

which means that

p
x2 + y 2 ;

p
y
x2 + y 2
o
(1 + x)
1
p
p
=
lim
lim
(x;y)!(0;0)
(x;y)!(0;0)
x2 + y 2
x2 + y 2

= 0:

Note that if we had to calculate the limit

(1 + x)
1
;
2
2
x +y
(x;y)!(0;0)
lim

then we would need Taylors formula of order m = 2 at (0; 0),


@f
(0; 0) (x
@x
1 @2f
+
(0; 0) (x
(2; 0)! @x2
1 @2f
(0; 0) (y
+
(0; 2)! @y 2

f (x; y) = f (0; 0) +

@f
(0; 0) (y 0)
@y
1
@2f
2
0) +
(0; 0) (x
(1; 1)! @x@y
0) +

0) (y

0)

0) + o x2 + y 2 :

Another simpler method would be to use the Taylors formulas for et and log (1 + s).
Wednesday, February 20, 2013
First midterm
Friday, February 22, 2013
Solutions First midterm.
Example 104 (Example 103, continued) Second method: If either x =
0 or y = 0, we get
y
(1 + x)
1
0
p
=p
= 0:
2
2
2
x +y
x + y2

If x 6= 0 and y 6= 0, then
y

(1 + x)
ey log(1+x) 1 log (1 + x)
xy
1
p
p
=
:
2
2
y log (1 + x)
x
x +y
x2 + y 2

Now, using the limits limt!0

et 1
t

= 1 and limt!0

ey log(1+x) 1
= 1;
(x;y)!(0;0) y log (1 + x)
lim

39

log(1+t)
t

= 1, we have

log (1 + x)
;
x

while
0

xy

=p

x2 + y 2

as (x; y) ! (0; 0).

x2 jyj

x2 + y 2 jyj
p
= jyj ! 0
x2 + y 2

x2 + y 2

Monday, February 25, 2013


Example 105 Lets calculate
log 1 + sin2 (xy)

lim

(x2 + y 2 )

(x;y)!(0;0)

x2 y 2

Taylors formula of sin t of order one is given by


sin t = t + o t2
and so
2

sin2 t = t + o t2

= t2 + o t2

+ 2to t2

= t2 + o t3
where we have used the properties of the little o. Hence
log 1 + sin2 t = log 1 + t2 + o t3

Lets use now Taylors formula


log (1 + s) = s + o (s) ;
where for us s = sin2 t = t2 + o t3 . We get
log 1 + sin2 t = log 1 + t2 + o t3
= t2 + o t3

+ o t2 + o t3

= t2 + o t2 :

Hence,
log 1 + sin2 (xy)
(x2 + y 2 )

x2 y 2

=
=

if x 6= 0 and y 6= 0. Now
0
and so

x2 y 2 + o x2 y 2
(x2 + y 2 )
o x2 y 2
(x2 +

2
y2 )

x2 y 2
(x2

2
y2 )

40

x2 y 2
(x2 +

1
;
2

o x2 y 2
!0
2
(x2 + y 2 ) x2 y 2
x2 y 2

x2 y 2

2
y2 )

o x2 y 2
x2 y 2

by Theorem while if either x = 0 or y = 0, we get


log 1 + sin2 (xy)
(x2

x2 y 2

2
y2 )

0
(x2

+ y2 )

= 0:

Hence,
log 1 + sin2 (xy)

lim

(x2 +

(x;y)!(0;0)

x2 y 2

2
y2 )

=0

Exercise 106 Calculate the limit


lim

(x;y)!(0;0)

log 1 + sin2 (xy)


(x2 +

x2 y 2

4
y2 )

Wednesday, February 27, 2013

Local Minima and Maxima

Denition 107 Let f : E ! R, where E

RN , and let x0 2 E. We say that

(i) f attains a local minimum at x0 if there exists r > 0 such that f (x)
f (x0 ) for all x 2 E \ B (x0 ; r),
(ii) f attains a global minimum at x0 if f (x)

f (x0 ) for all x 2 E,

(iii) f attains a local maximum at x0 if there exists r > 0 such that f (x)
f (x0 ) for all x 2 E \ B (x0 ; r),
(iv) f attains a global maximum at x0 if f (x)

f (x0 ) for all x 2 E.

Theorem 108 Let f : E ! R, where E RN . Assume that f attains a local


minimum (or maximum) at some point x0 2 E. If there exists a direction v
and > 0 such that the set
fx0 + tv : t 2 (
@f
@v

; )g

(11)

@f
@v

and if there exists


(x0 ), then necessarily,
(x0 ) = 0. In particular, if x0 is
an interior point of E and f is di erentiable at x0 , then all partial derivatives
and directional derivatives of f at x0 are zero.
Proof. Exercise.
Remark 109 In view of the previous theorem, when looking for local minima
and maxima, we have to search among the following:
Interior points at which f is di erentiable and rf (x) = 0, these are
called critical points. Note that if rf (x0 ) = 0, the function f may not
attain a local minimum or maximum at x0 . Indeed, consider the function
f (x) = x3 . Then f 0 (0) = 0, but f is strictly increasing, and so f does
not attain a local minimum or maximum at 0.
41

Interior points at which f is not di erentiable. The function f (x) = jxj


attains a global minimum at x = 0, but f is not di erentiable at x = 0.
Boundary points.
To nd su cient conditions for a critical point to be a point of local minimum
or local maximu, we study the second order derivatives of f .
RN , and let x0 2 E. The Hessian

Denition 110 Let f : E ! R, where E


matrix of f at x0 is the N N matrix
0

@2f
@x21

B .
Hf (x0 ) : = B
@ ..

@2f
@x1 @xN

@2f
@xN @x1

(x0 )

(x0 )

..
.
@2f
@x2N

(x0 )

@2f
(x0 )
@xi @xj

(x0 )

1
C
C
A

;
i;j=1

whenever it is dened.
Remark 111 If the hypotheses of Schwartzs theorem are satised for all i; j =
1; : : : ; N , then
@2f
@2f
(x0 ) =
(x0 ) ;
@xi @xj
@xj @xi
which means that the Hessian matrix Hf (x0 ) is symmetric.
Given an N
nomial

N matrix H, the characteristic polynomial of H is the polyp (t) := det (tIN

H) ;

t 2 R:

Theorem 112 Let H be an N N matrix. If H is symmetric, then all roots


of the characteristic polynomial are real.
Theorem 113 Given a polynomial of the form
p (t) = tN + aN

N
1t

+ aN

N
2t

+ a1 t + a0 ;

where the coe cients ai are real for every i = 0; : : : ; N


of p are real. Then

t 2 R;

1, assume that all roots

(i) all roots of p are positive if and only if the coe cients alternate sign, that
is, aN 1 < 0, aN 2 > 0, aN 3 < 0, etc.
(ii) all roots of p are negative if and only ai > 0 for every i = 0; : : : ; N

1.

The next theorem gives necessary and su cient conditions for a point to be
of local minimum or maximum.
42

Theorem 114 Let U


RN be open, let f : U ! R be of class C 2 (U ) and let
x0 2 U be a critical point of f .
(i) If Hf (x0 ) is positive denite, then f attains a local minimum at x0 ,
(ii) if f attains a local minimum at x0 , then Hf (x0 ) is positive semidenite,
(iii) if Hf (x0 ) is negative denite, then f attains a local maximum at x0 ,
(iv) if f attains a local maximum at x0 , then Hf (x0 ) is negative semidenite.
Remark 115 Note that in view of the previous theorem, if at a critical point x0
the characteristic polynomial of Hf (x0 ) has one positive root and one negative
root, then f does not admit a local minimum or a local maximum at x0 .
Theorem 114 shows that if Hf (x0 ) is positive denite, then f admits a local
minimum at x0 . The following exercise shows that we cannot weakened this
hypothesis to Hf (x0 ) positive semidenite.
Example 116 Let f (x; y) := x2
have
@f
(x; y) =
@x
@f
(x; y) =
@y

y 4 . Lets nd the critical points of f . We

@
x2
@x
@
x2
@y

y 4 = 2x

0 = 0;

y4 = 0

4y 3 = 0:

Hence, (0; 0) is the only critical point. Lets nd the Hessian matrix at these
points. Note that the function is of class C 1 , so we can apply Schwartzs theorem. We have
@2f
(x; y) =
@x2
@2f
(x; y) =
@y 2
@2f
(x; y) =
@y@x

@
(2x ) = 2;
@x
@
4y 3 = 12y 2 ;
@y
@
(2x) = 0:
@y

Hence,
2
0

Hf (0; 0) =

0
0

so that
1
0

0 = det
= det
=(

2
0

0
1

Hf (0; 0)
0
0

2) ;
43

=(

2) (

0)

and so the roots are = 2 or = 0. Hence, at (0; 0) we cannot have a local


maximum. But it could be a local minimum. However, taking
y4 ;

f (0; y) =

which has a strict maximum at y = 0. This shows that f does not admit a local
minimum or a local maximum at (0; 0).Theorem 114 shows that if Hf (x0 ) is
positive denite, then f admits a local minimum at x0 . The following exercise
shows that we cannot weakened this hypothesis to Hf (x0 ) positive semidenite.
A critical point at which f does not admit a local minimum or a local
maximum is called a saddle point.
Friday, March 1, 2013
Example 117 Consider the function
f (x; y) := (x

y) e

x2 y 2

dened in the set


E := B ((0; 0) ; 1):
Lets nd the critical points of f . We have
@f
(x; y) =
@x
@f
(x; y) =
@y

@
(x
@x
@
(x
@y

x2 y 2

y) e

x2 y 2

y) e

=e
=

x2 y 2

2x2 + 2yx + 1 = 0;

x2 y 2

2y 2 + 2xy + 1 = 0:

By subtracting the two equations, we get


x2 + y 2 = 0;
2y 2 + 2xy + 1 = 0:
If x = y, then 2x2 + 2x2 + 1 = 0, which has no solutions, while if x = y,
1 1
then 4y 2 + 1 = 0. Hence, the critical points are 21 ; 12 and
2 ; 2 . Lets
nd the Hessian matrix at these points. Note that the function is of class C 1 ,
so we can apply Schwartzs theorem. We have
@2f
(x; y) =
@x2
@2f
(x; y) =
@y 2
@2f
(x; y) =
@y@x

2
2
2
@
e x y
2x2 + 2yx + 1 = 2e x
@x
2
2
@
e x y
2y 2 + 2xy + 1 = 2e
@y
2
2
2
@
e x y
2x2 + 2yx + 1 = 2e x
@y

y2

x2 y 2

y2

Hence,
Hf

1
;
2

1
2

3e 2
1
e 2
44

e 2
1
3e 2

2x3

2yx2

2y 3

2x2 y

3x + y ;

2xy 2

3y + x ;

2xy 2 + x

y :

so that
1
0

0 = det

+ 3e
1
e 2

= det
=

0
1

1
2

+ 6e

1
;
2

Hf
1

1
2

+ 8e

1
2

e 2
+ 3e
1

1
2

;
1
2;

and so the eigenvalues are both negative. Hence, at


maximum. On the other hand
Hf

1
2

+ 3e

1 1
;
2 2

0
1

Hf

3e
e

1
2

e
3e

1
2

1
2

1
2

we have a local

1
2

so that
1
0

0 = det

3e
1
e 2

= det
=

6e

1
2

1
2

+ 8e

1 1
;+
2 2

e 2
3e
1

1
2

3e

1
2

and so the eigenvalues are all positive by Theorem 113. Hence, at 21 ; 12 we


have a local minimum.
It remains to study the boundary. Using polar coordinates, we have x = cos ,
y = sin , so that
g ( ) := f (cos ; sin ) = (cos

sin ) e

2 [0; 2 ] :

We have
g 0 ( ) = ( sin
3
4

. Thus the point cos 34 ; sin 34

of local minimum and the point cos 74 ; sin 74


of local maximum. Are they?

for

7
4

cos ) e

0
p

p
2
2
2 ; 2
p
p
2
2
2 ;
2

could be a point
could be a point

Monday, March 4, 2013


Example 118 First method: To see if they are consider the restriction y =
x. Let
" p p #
2
2
2x2
h (x) = f (x; x) = 2xe
; x2
;
2
2
Then
h0 (x) = 2e

2x2

+ 2xe

2x2

( 4x) = 2e

45

2x2

4x2 > 0

1
2

for

< x < 21 . Hence, h decreases for x <


p

2
2

and decreases for x > 12 . Since h0

1
2,

1
1
2p< x < 2 ,
2
2
is not
2 ; 2
p
p
that 22 ; 22

increases for

< 0, we have that


p

a point of local minimum, while, since h0 22 < 0, we have


is not a point of local maximum.
Second method: Another way to look at this was to study the partial and
directional derivatives. We have
p p !
2
2
@f
= e 1 ( 1 1 + 1) = e 1 ;
;
@x
2
2
p p !
@f
2
2
;
= e 1 ( 1 1 + 1) = e 1 :
@y
2
2
Since

@f
@x

have that f

p
p
2
2
;
2
2
p
x; 22 ; y

< 0, for points

for all 0 < jt


p

2
2
2 ; 2

2
2
2 ; 2

<f

2
2

p !
f
2
2
;
= lim
t!0
2
2

@f
@x

2
2

2
2

"<

+ t;

2
2

2
2

we

+ t;

2
2

p
2
2
;
2
2

2
2

p
2
2
2 ; 2

f
t

<

+"<0
p

2
2

0j < . In particular, if 0 < t < , then f


p

> 0 such that

2
2
2 ; 2

, which shows that

close to

, we can nd
p

with x >

. To see this, note that

and so taking 0 < " < e

x;

+ t;

2
2

<

is not a point of local minimum.

Similarly,
@f
@x

p !
2
=e
2

2
;
2

( 1

and so reasoning as before we can nd

for all 0 < jt


f

2
2 ;

2
2

f
"<
0j <

2
2

+ t;

2
2

1 + 1) =

In particular, if

, which shows that

2
2 ;

2
2 ;

2
2

46

< 0;

> 0 such that

t
1.

2
2

<

< t < 0, then f

+"<0
p

2
2

+ t;

2
2

>

is not a point of local maximum.

Example 119 Let f (x; y; z) := x2 + y 4 + y 2 + z 3 2xz. We have


8
8
8
@f
>
x z=0
< @x = 2x 2z = 0
<
< x z=0
@f
3
2
=
4y
+
2y
=
0
y=0
y
2y
+
1
=
0
()
()
@y
>
:
:
: @f = 3z 2 2x = 0
3z 2 2z = 0
3z 2 2x = 0
@z
8
< x z=0
y=0
()
:
z (3z 2) = 0

and so the critical points are (0; 0; 0) and 32 ; 0; 32 . Note that (0; 0; 0) is not a
point of local minimum or maximum, since f (0; 0; z) = z 3 which changes sign
near 0. Lets study the point 23 ; 0; 23 . We have
0

B
Hf = @

@2f
@x2
@2f
@x@y
@2f
@x@z

@2f
@y@x
@2f
@y 2
@2f
@y@z

@2f
@z@x
@2f
@z@y
@2f
@z 2

and so
Hf

2
2
; 0;
3
3

We have

2
C @
0
=
A
2
0

2
=@ 0
2

1
1 0 0
0 = det @t @ 0 1 0 A
0 0 1
0
t 2
0
t 2
= det @ 0
2
0
t

0 0

2
0

12y 2 + 2
0
1
2
0 A:
6

0
2
0

2
@ 0
2
1

1
2
0 A
6z

0
2
0

A = t3

11
2
0 AA
4
8t2 + 16t

The eigenvalues are all positive by Theorem 113. Hence, at


local minimum.

8:

2
2
3 ; 0; 3

we have a

Theorem 120 (Weierstrass) Let K RN be closed and bounded and let f :


K ! R be a continuous function. Then there exists x0 ; x1 2 K such that
f (x0 ) = min f (x) ;

f (x1 ) = max f (x) :

x2K

x2K

Denition 121 Given a set E RN and a function f : E ! RM , the Jacobian


matrix of f = (f1 ; : : : ; fM ) at some point x0 2 E, whenever it exists, is the
M N matrix
1
0
1 0 @f1
@f1
rf1 (x0 )
@x1 (x0 )
@xN (x0 )
B
C
C B ..
..
..
Jf (x0 ) := @
A:
A=@ .
.
.
@fM
@x1

rfM (x0 )

47

(x0 )

@fM
@xN

(x0 )

It is also denoted

@ (f1 ; : : : ; fM )
(x0 ) :
@ (x1 ; : : : ; xN )

When M = N , Jf (x0 ) is an N N square matrix and its determinant is called


the Jacobian determinant of f at x0 . Thus,
det Jf (x0 ) = det

10

@fj
(x0 )
@xi

:
i;j=1;:::;N

Lagrange Multipliers

In Section 9 (see Theorem 114) we have seen how to nd points of local minima
and maxima of a function f : E ! R in the interior E of E. Now we are ready
to nd points of local minima and maxima of a function f : E ! R on the
boundary @E of E. We assume that the boundary of E has a special form, that
is, it is given by a set of the form
x 2 RN : g (x) = 0 :
Denition 122 Let f : E ! R, where E
We say that

RN , let F

E and let x0 2 F .

(i) f attains a constrained local minimum at x0 if there exists r > 0 such


that f (x) f (x0 ) for all x 2 F \ B (x0 ; r),
(ii) f attains a constrained local maximum at x0 if there exists r > 0 such
that f (x) f (x0 ) for all x 2 F \ B (x0 ; r).
The set F is called the constraint.
Wednesday, March 6, 2013
Theorem 123 (Lagrange Multipliers) Let U
RN be an open set, let f :
1
M
U ! R be a function of class C and let g : U ! R be a class of function C 1 ,
where M < N , and let
F := fx 2 U : g (x) = 0g :
Let x0 2 F and assume that f attains a constrained local minimum (or maximum) at x0 . If Jg (x0 ) has maximum rank M , then there exist 1 , . . . , M 2 R
such that
rf (x0 ) = 1 rg1 (x0 ) +
+ M rgM (x0 ) :
Example 124 We want to nd points of local minima and maxima of the function f (x; y; z) := x y + 2z over the set
E = (x; y; z) 2 R3 : x2 + y 2 + 2z 2

48

2 :

Lets nd critical points of f in the interior. Since @f


@x (x; y; z) = 1 6= 0, there
are no critical points in the interior. Thus, points of local minima and maxima,
if they exist, must be found on the boundary of E, that is,
@E = (x; y; z) 2 R3 : x2 + y 2 + 2z 2 = 2 :
In other words, we are looking at a constrained problem. The constraint is given
by x2 + y 2 + 2z 2 = 2. Dene g (x; y; z) := x2 + y 2 + 2z 2 2. In this case the
Jacobian matrix of g coincides with the gradient, that is,
Jg (x; y; z) = rg (x; y; z) = (2x; 2y; 4z) :
In this case the maximum rank is one, so it is enough to check that (2x; 2y; 4z) 6=
(0; 0; 0) at points in the constraint @E. But (2x; 2y; 4z) = (0; 0; 0) only at the
origin and this point does not belong to the constraint.
By the theorem on Lagrange multipliers, we need to nd (x; y; z) and such
that
8
@g
@f
@
y + 2z
x2 + y 2 + 2z 2 2 = 1 2x ;
>
@x (x; y; z)
@x (x; y; z) = @x x
> 0 = @f
<
@g
@
0 = @y (x; y; z)
y + 2z
x2 + y 2 + 2z 2 2 = 1 2y ;
@y (x; y; z) = @y x
@g
@f
@
>
y; z) = @z x y + 2z
x2 + y 2 + 2z 2 2 = 2 4z ;
>
: 0 = @z (x; y; z) 2 @z (x;
2
2
0 = g (x; y; z) = x + y + 2z
2:

If

= 0, we have no solution, while if 6= 0, we get


8
x = 21 ;
>
>
<
y = 21 ;
z= 1;
>
>
: 2 2 2
x + y + 2z 2 2 = 0:

Substituting gives
and so
p

2
2 ;

=
p
p
2
2
2 ; 2

p1 .
2

1
2

1
2

+2

1
2

2 = 0, that is, 2
p

2
2
2 ; 2 ;

Thus, the only two possible points are

1 = 0,

2
2

and

. Now, since E is closed and bounded (why?), it is compact.


Since f is continuous, it follows by the Weierstrass theorempthatp there
exist
p
2
2
2
;
;
=
the minimum and the maximum of f over E. Hence, f
2
2
2
p

2
2

2
2

2
the maximum.

2
2

is the minimum and f

2
2 ;

p
2
2
2 ; 2

2
2

2
2

+2

2
2

is

Example 125 Given the set


E = (x; y; z) 2 R3 : x2

xy + y 2

z 2 = 1; x2 + y 2 = 1 ;

we want to nd the points of E having minimal distance from the origin. So we


need to nd
q
2
2
2
min
(x 0) + (y 0) + (z 0) : (x; y; z) 2 E :
49

To avoid dealing with square roots, we consider the function f (x; y; z) = x2 +


y 2 + z 2 . We want to nd the minimum of f over the st E.
First method. Let g : R2 ! R3 be the function
g (x; y; z) = x2

xy + y 2

z2

1; x2 + y 2

1 :

By the theorem on Lagrange multipliers we need to nd (x; y; z) and 1 ,


that
8
@g1
@g1
@
2
2
2
2
0 = @f
xy + y 2 z 2
1 @x
2 @x = @x x + y + z
1 x
>
@x
>
>
@g
@g
@f
@
2
2
2
2
1
1
>
xy + y 2 z 2
< 0 = @y
1 @y
2 @y = @y x + y + z
1 x
@g1
@g1
@
2
2
2
2
0 = @f
xy + y 2 z 2
1 @z
2 @z = @z x + y + z
1 x
@z
>
>
2
2
2
>
+y
z
1 = 0;
>
: 0 = g1 (x; y; z) = x2 xy
0 = g2 (x; y; z) = x + y 2 1 = 0:
We have

8
0 = 2x
y)
>
1 (2x
>
>
>
< 0 = 2y
1 ( x + 2y)
0 = 2z
1 ( 2z) ;
>
>
> 0 = x2 xy + y 2 z 2
>
:
0 = x2 + y 2 1:

such

1
1
1

2
2
2

2 2x;
2 2y;

1;

Adding the rst two equations and substituting the last in the fourth one, we get
8
0 = (x + y) (2
2 2) ;
>
1
>
>
>
(
x
+
2y)
< 0 = 2y
1
2 2y;
0 = z (1 + 1 ) ;
>
>
0 = xy z 2 ;
>
>
:
0 = x2 + y 2 1:
From the third equation 0 = z (1 + 1 ) we have that either z = 0 or
If z = 0 we get
8
0 = (x + y) (2
2 2) ;
>
1
>
>
>
0
=
2y
(
x
+
2y)
<
1
2 2y;
0 = z;
>
>
0 = xy;
>
>
:
0 = x2 + y 2 1:

1.

Hence, either x = 0 or y = 0. If x = 0, then from the last last equation y =


so that
8
8
8
2 2;
0=1
1 = 2;
>
>
>
1
2;
> 0=2
>
>
>
>
>
>
>
>
< 0=1
< 0=1
< 0 = 1;
1
2;
1
2;
0 = z;
0 = z;
0 = z;
()
()
>
>
>
>
>
>
0 = x;
0 = x;
> 0 = x;
>
>
>
>
>
:
:
:
1 = y:
1 = y:
1 = y:

1,

50

x2 + y 2
x2 + y 2
x2 + y 2

1
1
1

;
;
;

If y = 0 then from the last last equation x = 1, so that


8
8
0 = (2
2 2) ;
1 = 2;
>
>
1
>
>
>
>
>
>
< 0 = 1;
< 0 = 1;
0 = z;
0 = z;
()
>
>
>
>
0
=
y;
0 = y;
>
>
>
>
:
:
1 = x:
1 = x:
If

1, then

8
0 = (x + y) (3 2 2 ) ;
>
>
>
>
< 0 = 4y x
2 2y;
1 = 1;
>
>
0 = xy z 2 ;
>
>
:
0 = x2 + y 2 1:

In the rst equation either x + y = 0 or 3 2 2 = 0. In the rst case, we get


8
8
x = y;
x = y;
>
>
>
>
>
>
5
>
>
< 2 = 2;
< 0 = x ( 5 + 2 2) ;
1 = 1;
1 = 1;
()
>
>
2
2
>
>
z
= x;
0
=
x
z
;
>
>
>
>
: p1 = x;
:
1 = 2x2 ;
2

while in the second, 23 = 2 , and so


8 3
8 3
= 2;
>
>
2
2 = 2;
>
>
>
>
>
>
< 0 = 4y x 3y;
< x = y;
1 = 1;
1 = 1;
()
>
>
2
>
>
0
=
xy
z
;
0
= x2 z 2 ;
>
>
>
>
:
:
2
2
0=x +y
1:
0 = 2x2 1:

The fourth and fth equation are incompatible, so there are no solutions in this
case.
In conclusion, we have found the points (0; 1; 0) with 1 = 0 and 2 = 1,
( 1; 0; 0) with

= 0 and

1,

p1 ;
2

p1 ;
2

p1
2

and

p1 ;
2

p1 ;
2

with 1 = 1 and 2 = 52 .
Lets prove that E is bounded. We have that x2 + y 2 = 1, so jxj
jyj 1, while
z 2 = 1 x2 + xy y 2 xy 1:

p1
2

1 and

Hence, E is bounded and closed. It follows by the Weierstrass theorem that f


has a mimimum over E. Since
f

1
p ;
2

1
p ;
2

1
p
2

=f

1
p ;
2

1
p ;
2

1
p
2

3
> 1 = f ( 1; 0; 0) = f (0; 1; 0) ;
2

it follows that ( 1; 0; 0) and (0; 1; 0) are the two points in E of minimal distance from the origin.
Second method. Exercise. Find a simpler way to solve this problem.
51

Friday, March 8, 2013


Midsemester break, no classes
Monday, March 11, 2013
Spring break, no classes
Wednesday, March 13, 2013
Spring break, no classes
Friday, March 15, 2013
Spring break, no classes
Monday, March 18, 2013

11

Chain Rule

Denition 126 Let E RN , let f : E ! RM , and let x0 2 E be an accumulation point of E. The function f is dierentiable at x0 if there exists a linear
function T : RN ! RM (depending on f and x0 ) such that
lim

x!x0

f (x)

f (x0 ) T (x
kx x0 kN

x0 )

= 0:

(12)

provided the limit exists. The function T , if it exists, is called the dierential of
f at x0 and is denoted df (x0 ) or dfx0 .
Theorem 127 Let E
RN , let f : E ! RM , and let x0 2 E be an accumulation point of E. Then f = (f1 ; : : : ; fM ) is di erentiable at x0 if and
only if all its components fj , j = 1; : : : ; M , are di erentiable at x0 . Moreover,
dfx0 = d (f1 )x0 ; : : : ; d (fM )x0 .
Proof. Exercise.
We study the dierentiability of composite functions.
Theorem 128 (Chain Rule) Let F
RM , E
RN , let g : F ! E, g =
(g1 ; : : : ; gN ), and let f : E ! R. Assume that at some point y0 2 F there
@gN
1
exist the directional derivatives @g
@v (y0 ), . . . , @v (y0 ) for some direction v and
f is di erentiable at the point g (y0 ). Then the composite function f g admits
a directional derivative at y0 in the direction v and
N

X @f
@ (f g)
@gi
(y0 ) =
(y0 )
(g (y0 ))
@v
@x
@v
i
i=1
= rf (g (y0 ))

@g
(y0 ) :
@v

Moreover, if g is di erentiable at y0 and f is di erentiable at g (y0 ), then f g


is di erentiable at y0 .
Example 129 Consider the function
g (x) := f (kxk) = f

q
52

x21 + x22 +

+ x2N

where f : [0; 1) ! R is di erentiable. Then for all x 6= 0,


q
@g
2xi
(x) = f 0
x21 + x22 +
+ x2N
p
@xi
2
x21 + x22 +
Example 130 Consider the function
Z
h (x; y) =

g(x;y)

t2

+ x2N

dt;

f (x;y)

where f : R2 ! R and g : R2 ! R are di erentiable. To apply the chain rule,


consider the function
Z z
2
k (z; w) =
e t dt;
w

and observe that

h (x; y) = k (g (x; y) ; f (x; y)) :


Hence, by the chain rule,
@h
(x; y) =
@x
@h
(x; y) =
@y

@k
@g
@k
@f
(g (x; y) ; f (x; y))
(x; y) +
(g (x; y) ; f (x; y))
(x; y) ;
@z
@x
@w
@x
@k
@g
@k
@f
(g (x; y) ; f (x; y))
(x; y) +
(g (x; y) ; f (x; y))
(x; y) :
@z
@y
@w
@y

Now

@k
(z; w) = e
@z

while

@k
(z; w) =
@w

z2

z2

and so
@h
(x; y) = e
@x
@h
(x; y) = e
@y

@g
(x; y)
@x
(g(x;y))2 @g
(x; y)
@y
(g(x;y))2

@f
(x; y) ;
@x
2 @f
e (f (x;y))
(x; y) :
@y
e

(f (x;y))2

Example 131 Consider the functions f : R2 ! R and g : R2 ! R2 dened by


f (x1 ; x2 ) = x1 x2

1;

g (y1 ; y2 ) = (y1 y2

ey1 ; y1 sin (y1 y2 )) :

Then
@f
@f
(x1 ; x2 ) = 1x2 0;
(x1 ; x2 ) = x1 1 0;
@x1
@x2
@g1
@g1
(y1 ; y2 ) = 1y2 ey1 ;
(y1 ; y2 ) = y1 1 0;
@y1
@y2
@g2
@g2
(y1 ; y2 ) = 1 sin (y1 y2 ) + y1 cos (y1 y2 ) (1y2 ) ;
(y1 ; y2 ) = y1 cos (y1 y2 ) (y1 1) :
@y1
@y2
53

First method: Consider the composition


(f

g) (y1 ; y2 ) = f (g1 (y1 ; y2 ) ; g2 (y1 ; y2 )) :

Then by the chain rule,


@ (f g)
@f
@g1
@f
@g2
(y1 ; y2 ) =
(g1 (y1 ; y2 ) ; g2 (y1 ; y2 ))
(y1 ; y2 ) +
(g1 (y1 ; y2 ) ; g2 (y1 ; y2 ))
(y1 ; y2 )
@y1
@x1
@y1
@x2
@y1
= y1 sin (y1 y2 ) (1y2 ey1 ) + (y1 y2 ey1 ) (1 sin (y1 y2 ) + y1 cos (y1 y2 ) (1y2 )) ;
while
@ (f g)
@f
@g1
@f
@g2
(y1 ; y2 ) =
(g1 (y1 ; y2 ) ; g2 (y1 ; y2 ))
(y1 ; y2 ) +
(g1 (y1 ; y2 ) ; g2 (y1 ; y2 ))
(y1 ; y2 )
@y2
@x1
@y2
@x2
@y2
= y1 sin (y1 y2 ) (y1 1 0) + (y1 y2 ey1 ) (y1 cos (y1 y2 ) (y1 1)) :
Second method: Write the explicit formula for (f
(f

g), that is,

g) (y1 ; y2 ) = f (g1 (y1 ; y2 ) ; g2 (y1 ; y2 ))


ey1 ) (y1 sin (y1 y2 ))

= (y1 y2

1:

Then
@
@ (f g)
(y1 ; y2 ) =
[(y1 y2 ey1 ) (y1 sin (y1 y2 )) 1]
@y1
@y1
= (1y2 ey1 ) (y1 sin (y1 y2 )) + (y1 y2 ey1 ) (1 sin (y1 y2 ) + y1 cos (y1 y2 ) (1y2 ))
and
@ (f g)
@
(y1 ; y2 ) =
[(y1 y2 ey1 ) (y1 sin (y1 y2 ))
@y2
@y2
= (y1 1 0) (y1 sin (y1 y2 )) + (y1 y2

1]
ey1 ) (y1 cos (y1 y2 ) (y1 1))

As a corollary of Theorem 128, we have the following result.


Corollary 132 Let F
RM , E RN , let g : F ! E, g = (g1 ; : : : ; gN ), and
let f : E ! RP . Assume that g is di erentiable at some point y0 2 F and that
f is di erentiable at the point g (y0 ) and that g (y0 ) 2 E . Then the composite
function f g is di erentiable at y0 and
Jf

(y0 ) = Jf (g (y0 )) Jg (x0 ) :


Wednesday, March 20, 2013

12

Implicit and Inverse Function

Given a function f of two variables (x; y) 2 R2 , consider the equation


f (x; y) = 0:
54

0:

We want to solve for y, that is, we are interested in nding a function y = g (x)
such that
f (x; g (x)) = 0:
We will see under which conditions we can do this. The result is going to be
local.
In what follows given x 2 RN and y 2 RM and f (x; y), we write
0

@f1
@x1

@f
B
(x; y) := @ ...
@x
@f

@f1
@y1

B .
@f
(x; y) := B
@ ..
@y
@f

@y1

(x; y)

@fM
@xN

(x; y)

@f1
@yM

(x; y)

@fM
@yM

(x; y)

..
.

@x1

and

@f1
@xN

(x; y)
(x; y)

(x; y)

..
.
(x; y)

1
C
A

C
C:
A

Theorem 133 (Implicit Function) Let U RN RM be open, let f : U !


RM , and let (a; b) 2 U . Assume that f 2 C m (U ) for some m 2 N, that
f (a; b) = 0

and

det

@f
(a; b) 6= 0:
@y

Then there exist BN (a; r0 )


RN and BM (b; r1 )
BM (b; r1 ) U , and a unique function

RM , with BN (a; r0 )

g : BN (a; r0 ) ! BM (b; r1 )
of class C m such that f (x; g (x)) = 0 for all x 2 BN (a; r0 ) and g (a) = b.
@f
(a; b) = 0, then anything can
The next examples show that when det @y
happen.

Example 134 In all these examples N = M = 1 and

@f
@y

(x0 ; y0 ) = 0.

(i) Consider the function


f (x; y) := (y
Then f (0; 0) = 0,

@f
@y

x) :

(0; 0) = 0 and g (x) = x satises f (x; g (x)) = 0.

(ii) Consider the function


f (x; y) := x2 + y 2 :
Then f (0; 0) = 0, @f
@y (0; 0) = 0 but there is no function g dened near
x = 0 such that f (x; g (x)) = 0.

55

(iii) Consider the function


1) x2 + y 2 :

f (x; y) := (xy
Then f (0; 0) = 0,

@f
@y

(0; 0) = 0 but
0

g (x) =

if x = 0;
if x =
6 0;

1
x

which is discontinuous.
Next we give some examples on how to apply the implicit function theorem.
Example 135 Consider the function
f (x; y) := xey

y:

y
1 so that @f
1. By the implicit
Then f (0; 0) = 0, @f
@y (x; y) = xe
@y (0; 0) =
function theorem there exist r > 0 and a function g 2 C 1 (( r; r)) such that
g (0) = 0 and f (x; g (x)) = 0. To nd the behavior of g near x = 0, we can use
Taylors formula. Lets nd g 0 (0) and g 00 (0). We have

xeg(x)

g (x) = 0:

Hence, di erentiating twice


1eg(x) + xg 0 (x) eg(x)

g 0 (x) = 0;
2

eg(x) g 0 (x) + g 0 (x) eg(x) + xg 00 (x) eg(x) + x (g 0 (x)) eg(x)

g 00 (x) = 0:

Substituting x = 0 and using g (0) = 0 we get


1e0 + 0

g 0 (0) = 0;

e0 g 0 (0) + g 0 (0) e0 + 0 + 0

g 00 (0) = 0:

So that g 0 (0) = 1 and g 00 (0) = 2. Hence,


g (x) = g (0) + g 0 (0) (x
= 0 + 1 (x

1
0) + g 00 (0) (x
2

1
0) + 2 (x
2

0) + o (x

0) + o (x

0)

0)

Friday, March 22, 2013


Correction homework.
Monday, March 25, 2013
Example 136 Consider the function
x2 + 1; y sin (xz)

f (x; y; z) = y cos (xz)


56

x :

Lets prove that there exist r > 0 and g : (1 r; 1 + r) ! R2 of class C 1 such


that g (1) = 1; 2 and f (x; g (x)) = 0. Note that f is of class C 1 . Here the
point is 1; 1; 2 and
f 1; 1;

= 1 cos 1

1 + 1; 1 sin 1

1 = (0; 0) :

Moreover,
@f
(x; y; z) =
@ (y; z)

@f1
@y
@f2
@y

@f1
@z
@f1
@z

(x; y; z)
(x; y; z)

@
@y
@
@y

1 cos (xz)
1 sin (xz)

y cos (xz)
(y sin (xz)
0
0

(x; y; z)
(x; y; z)

x2 + 1
x)
0

!
@
@z
@
@z

x2 + 1
x)

y cos (xz)
(y sin (xz)

xy sin (xz) 0
xy cos (xz) 0

and so
det

@f
1; 1;
@ (y; z)
2

cos 1 2
sin 1 2

= det

1 sin 1 2
1 cos 1 2

= 1 6= 0:

Hence, by the implicit function theorem there exist r > 0 and g : (1 r; 1 + r) !


R2 of class C 1 such that g (1) = 1; 2 and f (x; g (x)) = 0 for all x 2
(1 r; 1 + r), that is,
g1 (x) cos (xg2 (x))
g1 (x) sin (xg2 (x))

x2 + 1 = 0;
x = 0:

Reasoning as before, we can use Taylors formula to nd the behavior of g1 and


g2 near x = 1, that is,
g1 (x) = g1 (1) + g10 (1) (x
g2 (x) = g2 (1) +

g20

(1) (x

1) + o ((x

1)) ;

1) + o ((x

1)) :

Lets di erentiate the two equations. We get


g10 (x) cos (xg2 (x)) g1 (x) (1g2 (x) + xg20 (x)) sin (xg2 (x))
g10 (x) sin (xg2 (x)) + g1 (x) (1g2 (x) + xg20 (x)) cos (xg2 (x))
Taking x = 1 and using the fact that g1 (1) = 1 and g2 (1) =
g10 (1) cos 1 2
1
g10 (1) sin 1 2 + 1

2
2

+ 1g20 (1) sin 1 2


+ 1g20 (1) cos 1 2

that is,
0 1 2 + g20 (1) 1 2 = 0;
g10 (1) 1 + 0 1 = 0;
57

2,

2x + 0 = 0;
1 = 0:
we obtain

2 = 0;
1 = 0;

and so g10 (1) = 1 and g20 (1) =

2.

g1 (x) = 1 + 1 (x
g2 (x) =

Hence,

1) + o ((x
2

(x

1)) ;
1) + o ((x

1)) :

Next we give the inverse function theorem.


Theorem 137 (Inverse Function) Let U
RN be open, let f : U ! RN ,
m
and let a 2 U . Assume that f 2 C (U ) for some m 2 N and that
det Jf (a) 6= 0:
Then there exists B (a; r0 )

U such that f (B (a; r0 )) is open, the function

f : B (a; r0 ) ! f (B (a; r0 ))
is invertible and f

2 C m (f (B (a; r0 ))). Moreover,


Jf

(y) = Jf f

(y)

Exercise 138 Consider the function f : R2 ! R2 dened by


0
x + 2x2 sin x1

f1 (x; y) =
f2 (x; y) = y:

if x = 0;
if x =
6 0;

Prove that f = (f1 ; f2 ) is di erentiable in (0; 0) and Jf (0; 0) = 1. Prove that f


is not one-to-one in any neighborhood of (0; 0).
The next example shows that the existence of a local inverse at every point
does not imply the existence of a global inverse.
Example 139 Consider the function f : R2 ! R2 dened by
f (x; y) = (ex cos y; ex sin y) :
The function f is not injective. Indeed, f (x; y) = f (x; y + 2 ) for all (x; y) 2 R2 .
Hence, f does not admit a global inverse. However,
!
@f1
@f1
ex cos y
ex sin y
@x
@y
Jf (x; y) =
=
;
x
@f2
@f2
e sin y ex cos y
@x
@y
and so
det Jf (x; y) = det

ex cos y
ex sin y

ex sin y
ex cos y

= cos2 y e2x + e2x sin2 y = e2x 6= 0

for all (x; y) 2 R2 . Hence, by the inverse function theorem f admits a local
inverse in a neighborhood of every point (x; y) 2 R2 .
Wednesday, March 27, 2013
58

13

Curves

Let I R be an interval and let ' : I ! RN be a function. As the parameter


t traverses I, ' (t) traverses a curve in RN . Rather than calling ' a curve, it
is better to regard any vector function g obtained from ' by a suitable change
of parameter as representing the same curve as '. Thus, one should dene a
curve as an equivalence class of equivalent parametric representations.
Denition 140 Given two intervals I; J
R and two functions ' : I ! RN
N
and
: J ! R , we say that they are equivalent if there exists a continuous,
bijective function h : I ! J such that
' (t) =

(h (t))

for all t 2 I. We write '


and we call ' and
and the function h a parameter change.
Note that in view of a theorem done in class, h

parametric representations
1

: J ! I is also continuous.

Exercise 141 Prove that

is an equivalence relation.

Denition 142 A curve


tions.

is an equivalence class of parametric representa-

Given a curve with parametric representation ' : I ! RN , where I


R
is an interval, the multiplicity of a point y 2 RN is the (possibly innite)
number of points t 2 I such that ' (t) = y. Since every parameter change
h : I ! J is bijective, the multiplicity of a point does not depend on the
particular parametric representation. The range of
is the set of points of
RN with positive multiplicity, that is, ' (I). A point in the range of with
multiplicity one is called a simple point. If every point of the range is simple,
then is called a simple arc.
If I = [a; b] and ' (a) = ' (b), then the curve
is called a closed curve.
A closed curve is called simple if every point of the range is simple, with the
exception of ' (a), which has multiplicity two.
Example 143 The two functions
' (t) := (cos t; sin t) ;

t 2 [0; 2 ] ;

(s) := (cos 2s; sin 2s) ;

s 2 [0; ] ;

are equivalent, take h (t) = 2t , while the function


p (r) := (cos r; sin r) ;

r 2 [0; 4 ] ;

is not equivalent to the previous two, since the rst curve is simple and the
second no. Note that the range is the same, the unit circle.

59

Example 144 Consider the curve with parametric representations ' : R ! R2 ,


given by
' (t) := (t (t 1) ; t (t 1) (2t 1)) ; t 2 R:

Lets try to sketch the range of the curve. Set x (t) := t (t 1) and y (t) :=
t (t 1) (2t 1). We have x (t)
0 for t (t 1)
0, that is, when t
1 or
t 0, while x0 (t) = 1 (t 1) + t (1 0) = 2t 1 0 for t 21 . Moreover,
lim x (t) = lim t (t

t! 1

1) =

t! 1

lim x (t) = lim t (t

t!1

t!1

On the other hand, y (t)


or t 1, while y 0 (t) = 6t2

1) = 1 (1

0 for t (t 1) (2t
6t + 1 0 for t

lim y (t) = lim t (t

t! 1

t! 1

lim y (t) = lim t (t

t!1

1( 1

t!1

1) = 1:

1)

p
3+ 3
6

1) (2t
1) (2t

1) = 1;

0, that is, when 0 t 21


p
or t 3 6 3 . Moreover,

1) =

1;

1) = 1:

To draw the range of the curve in the xy plane, note that x increasing means
that we are moving to the right, x decreasing to the left, y increasing means that
we are moving upwards, y decreasing downwards.
Next we start discussing the regularity of a curve.
Denition 145 The curve is said to be continuous if one (and so all) of its
parametric representations is continuous.
The next result shows that the class of continuous curves is somehow too
large for our intuitive idea of curve. Indeed, we construct a continuous curve
that lls the unit square in R2 . The rst example of this type was given by
Peano in 1890.
Theorem 146 (Peano) There exists a continuous function ' : [0; 1] ! R2
such that ' ([0; 1]) = [0; 1] [0; 1].
Friday, March 29, 2013
In view of the previous result, to recover the intuitive idea of a curve, we
restrict the class of continuous curves to those with nite length.
In what follows, given an interval I
R, a partition of I is a nite set
P := ft0 ; : : : ; tn g I, where
t0 < t1 <

< tn :

Denition 147 Let I


R be an interval and let ' : I ! RN be a function.
The pointwise variation of ' on the interval I is
( n
)
X
VarI ' := sup
k' (ti ) ' (ti 1 )k ;
(13)
i=1

where the supremum is taken over all partitions P := ft0 ; : : : ; tn g of I, n 2 N. A


function ' : I ! RN has nite or bounded pointwise variation if VarI ' < 1.
60

Exercise 148 Prove that if two parametric representations ' : I ! RN and


: J ! RN are equivalent, then VarI ' = VarJ .
We are now ready to dene the length of a curve.
Denition 149 Given a curve , let ' : I ! RN be a parametric representation of , where I R is an interval. We dene the length of as
L ( ) := VarI ':
We say that the curve

is rectiable if L ( ) < 1.

In view of the previous exercise, the denition makes sense, that is, the length
of the curve does not depend on the particular representation of the curve.
Peanos example shows that continuous curves in general are not rectiable,
that is, they may have innite length. Next we show that C 1 or piecewise C 1
curves are rectiable.
Denition 150 Given a function ' : [a; b] ! RN , we say that f is piecewise
C 1 if there exists a partition P = ft0 ; : : : ; tn g of [a; b] 2 I, with a = t0 < t1 <
< tn = b such that ' is of class C 1 in each interval [ti 1 ; ti ], i = 1; : : : ; n.
Note that at each ti , i = 1; : : : ; n 1, the function ' has a right and a left
derivative, but they might be dierent.
Example 151 Consider the curve with parametric representations ' : [ 1; 1] !
R2 , given by
' (t) := (t; jtj) ; t 2 [ 1; 1] ;
is not a curve of class C 1 since at t = 0 the second component of ' is not
di erentiable, but it is piecewise C 1 , since
' (t) = (t; t) ;
' (t) = (t; t) ;

t 2 [ 1; 0] ;

t 2 [0; 1] ;

are both C 1 functions. Note that the left derivative at 0 is '0 (0) = (1; 1)
while the right derivative is '0+ (0) = (1; 1). Hence, the curve has a corner at
the point ' (0) = (0; 0).
By requiring parametric representations, parameter changes and their inverses to be dierentiable, or Lipschitz, or of class C n , n 2 N0 , etc., or piecewise
C 1 , we may dene curves that are dierentiable, or Lipschitz, or of class C n ,
n 2 N0 , or piecewise C 1 , respectively.
Theorem 152 Let
be a curve of class C 1 , with parametric representation
N
' : [a; b] ! R . Then
Z b
k'0 (t)k dt = L ( ) :
a

61

Given a curve , with parametric representation ' : I ! RN , where I R


is an interval. If ' is dierentiable at a point t0 2 I and '0 (t0 ) 6= 0, the straight
line parametrized by
(t) := ' (t0 ) + '0 (t0 ) (t

t0 ) ;

t 2 R;

is called a tangent line to at the point ' (t0 ), and the vector '0 (t0 ) is called
a tangent vector to at the point ' (t0 ). Note that if is a simple arc, then
the tangent line at a point y of the range of , if it exists, is unique, while
if the curve is self-intersecting at some point y 2 , then there could be more
than one tangent line at y.
Given a function f : I ! R, where I
R is an interval, the graph of f is
the range of a curve parametrized by
' (t) = (t; f (t)) :
If f is dierentiable at some point t0 2 I, then the tangent vector is given by
'0 (t0 ) = (1; f 0 (t0 )) :
Denition 153 Given a rectiable curve , we say that is parametrized by
arclength if it admits a parametrization
: [0; L ( )] ! RN such that
is
0
( ) = 1 for all but
di erentiable for all but nitely many 2 [0; L ( )] with
nitely many .
The name arclength comes from the fact that if is piecewise C 1 then for
every 1 < 2 , by Theorem 152 we have that the length of the portion of the
curve parametrized by : [ 1 ; 2 ] ! RN is given by
Z 2
Z 2
0
1d = 2
( ) d =
1:
1

Denition 154 A curve is regular if is piecewise C 1 and it admits a parametric representation ' : [a; b] ! RN for which the left and right derivative are
di erent from zero for all t 2 [a; b].
Theorem 155 If

is a regular curve, then

can be parametrized by arclength.

Proof. Let ' : [a; b] ! RN be a piecewise C 1 parametric representation of


and dene the length function
Z t
s (t) :=
k'0 (r)k dr:
a

Note that by by Theorem 152, s (b) = L ( ), while s (a) = 0. Hence, s : [a; b] !


[0; L ( )] is onto. We claim that s is invertible. Indeed, if t1 < t2 , then since
k'0 (r)k > 0 for all but nitely many t in [t1 ; t2 ], we have that
Z t2
s (t2 ) s (t1 ) =
k'0 (r)k dr > 0
t1

62

(why?). Hence, s : [a; b] ! [0; L ( )] is invertible. By a theorem in Real Analysis,


s 1 : [0; L ( )] ! [a; b] is continuous. By the fundamental theorem of calculus,
s0 (t) = k'0 (t)k > 0 for all but nitely many t. Hence, s is piecewise C 1 . In
turn, by a theorem in the Real Analysis, for any such t,
d
d

(s (t)) =

s0

1
:
(t)

Hence, there exists


d
d

( )=

1
s0 (s

( ))

1
k'0 (s

(14)

( ))k

for all but nitely many . Since ' is piecewise C 1 and the left and right
derivatives of ' are always dierent from zero, it follows from (14) and the
continuity of s 1 that s 1 is piecewise C 1 . Thus, the function : [0; L ( )] !
RN , dened by
( ) := ' s 1 ( ) ;
2 [0; L ( )] ;
(15)
is equivalent to '. Moreover, by the chain rule and (14), for all but nitely
many ,
'0 s 1 ( )
d
0
s 1 ( )=
( ) = '0 s 1 ( )
d
k'0 (s 1 ( ))k
and so

( ) = 1.

Example 156 Consider the curve with parametric representations ' : [0; 2 ] !
R3 , given by
' (t) := (R cos t; R sin t; 2t) ; t 2 [0; 2 ] :
Since
'0 (t) = ( R sin t; R cot st; 2) 6= (0; 0; 0) ;
the curve is regular. Lets parametrize it using arclength. Set
Z t
Z tp
s (t) :=
k'0 (r)k dr =
R2 sin2 t + R2 cos2 t + 4 dr
0
0
p
= R2 + 4t:

p
2
In particular, s (2 ) = 2 R
p + 4 is the length of the curve. The inverse of s
s
is t (s) := pR2 +4 , s 2 0; 2 R2 + 4 . Hence,
(s) := ' (t (s)) =
where s 2 0; 2

R cos p

s
R2

+4

; R sin p

s
R2

+4

;p

2s
R2 + 4

R2 + 4 , is the parametrization by arclength.

63

14

Curve Integrals

In this section we dene the integral of a function along a curve.


Denition 157 Given a piecewise C 1 curve with parametric representation
' : [a; b] ! RN and a bounded function f : E ! R, where ' ([a; b])
E, we
dene the curve (or line) integral of f along the curve as the number
Z

f ds :=

f (' (t)) k'0 (t)k dt;

whenever the function t 2 [a; b] 7! f (' (t)) k'0 (t)k is Riemann integrable.
R
Note that
f ds is always dened if f is continuous.
Next we show that the curve integral does not depend on the particular
parametric representation.
Proposition 158 Let be a piecewise C 1 curve and let ' : [a; b] ! RN and
: [c; d] ! RN be two parametric representations. Given a continuous function
f : E ! R, where E contains the range of ,
Z

b
0

f (' (t)) k' (t)k dt =

f ( ( ))

( ) d :

Proof. Exercise
Remark 159 In particular, if is a regular curve and : [0; L ( )] ! RN is
0
the parametric representation obtained using arclength, then
( ) = 1 for
all but nitely many . Hence,
Z

f ds =

L( )

f ( ( )) d :

Monday, April 1, 2013


Next we introduce the notion of an oriented curve.
Denition 160 Given a curve with parametric representations ' : I ! RN
and
: J ! RN , we say that ' and
have the same orientation if the parameter change h : I ! J is increasing and opposite orientation if the parameter
change h : I ! J is decreasing. If ' and have the same orientation, we write
'
.
Exercise 161 Prove that

is an equivalence relation.

Denition 162 An oriented curve is an equivalence class of parametric representations with the same orientation.

64

Note that any curve gives rise to two oriented curves. Indeed, it is enough
to x a parametric representation ' : I ! RN and considering the equivalence
class + of parametric representations with the same orientation of ' and the
equivalence class
of parametric representations with the opposite orientation
of '.
Let E RN and let g : E ! RN be a continuous function. Given a piecewise
1
C oriented curve with parametric representation ' : [a; b] ! RN such that
' ([a; b]) E, we dene
Z

g :=

N
bX

gi (' (t)) '0i (t) dt:

a i=1

Denition 163 Let U RN be an open set and let g : U ! RN . We say that


g is conservative vector eld if there exists a di erentiable function f : U ! R
such that
rf (x) = g (x)
for all x 2 U . The function f is called a scalar potential for g.
Denition 164 Given E

RN , we say that

(i) E is disconnected if there exist two nonempty open sets U; V


that
E

U [ V;

E \ U 6= ;;

E \ V 6= ;;

RN such

E \ U \ V = ;:

(ii) E is connected if it is not disconnected.


Theorem 165 Let U RN be an open connected set and let g : U ! RN be a
continuous function. Then the following conditions are equivalent.
(i) g is a conservative vector eld,
(ii) for every x, y 2 U and for every two piecewise C 1 oriented curves 1 and
N
and '2 : [c; d] ! RN ,
2 with parametric representations '1 : [a; b] ! R
respectively, such that '1 (b) = '2 (d) = x, '1 (a) = '2 (c) = y, and
'1 ([a; b]) ; '2 ([c; d]) U ,
Z
Z
g=
g:
1

(iii) for every piecewise C 1 closed oriented curve


Z
g = 0:

with range contained in U ,

Next we give a simple necessary condition for a eld g to be conservative.

65

Denition 166 Let U


RN be an open set and let g : U ! RN be di erentiable. We say that g is an irrotational vector eld or a curl-free vector eld
if
@gi
@gj
(x) =
(x)
@xj
@xi
for all i; j = 1; : : : ; N and all x 2 U .
Theorem 167 Let U RN be an open set and let g : U ! RN be a conservative vector eld of class C 1 . Then g is irrotational.
Proof. Since g is a conservative vector eld, there exists a a scalar potential
f : U ! R with rf = g in U . But since g is of class C 1 , we have that f is of
class C 2 . Hence, we are in a position to apply the Schwartz theorem to conclude
that
@2f
@2f
@gj
@gi
(x) =
(x) =
(x) =
(x)
@xj
@xj @xi
@xi @xj
@xi
for all i; j = 1; : : : ; N and all x 2 U .
The next example shows that there exist irrotational vector elds that are
not conservative.
Example 168 Let U := R2 n f(0; 0)g and consider the function
g (x; y) :=

y
x
;
x2 + y 2 x2 + y 2

Note that
@
@y

x2

y
+ y2

@
@x

x
x2 + y 2

x2 + y 2

2;

(x2 + y 2 )
x2 + y 2
(x2 + y 2 )

2;

and so g is irrotational. However, if we consider the closed curve of parametric


representation ' (t) = (cos t; sin t), t 2 [0; 2 ], we have that
Z
Z 2
sin t
cos t
0
0
g=
dt
2 (cos t) +
2 (sin t)
2
2
cos t + sin t
cos t + sin t
0
Z 2
=
sin2 t + cos2 t dt = 2 6= 0;
0

and so g cannot be conservative.


The problem here is the fact that the domain has a hole.
Wednesday, April 3, 2013
Denition 169 A set E RN is starshaped with respect to a point x0 2 E if
for every x 2 E, the segment joining x and x0 is contained in E.
66

Theorem 170 (Poincars Lemma) Let U


RN be an open set starshaped
N
with respect to a point x0 and let g : U ! R be an irrotational vector eld of
class C 1 . Then g is a conservative vector eld.
Example 171 Consider the function g : R2 ! R2 dened by
g (x; y) := (yex ; ex

cos y) :

We have,
@
(yex ) = 1ex ;
@y
@ x
(e
@x

cos y) = ex

0;

and so g is irrotational. Since R2 is convex, it is starshaped with respect to


every point. Hence, by the previous theorem, g is conservative. To nd a scalar
potential, note that
@f
(x; y) = yex ;
@x

@f
(x; y) = ex
@y

cos y

and so
f (x; y)

@f
(s; y) ds =
0 @x
= y (ex 1) :

f (0; y) =

yes ds

Hence,
f (x; y) = f (0; y) + y (ex

1) :

Di erentiating with respect to y, we get


ex
and so

cos y =

@f
@f
(x; y) =
(0; y) + ex
@y
@y

@f
(0; y) =
@y

cos y + 1:

Hence,
f (0; y)

f (0; 0) =

=y

@f
(0; t) dt =
@y
sin y

( cos t + 1) dt

and so
f (x; y) = f (0; y) + y (ex
= f (0; 0) + y

1) =

sin y + yex
67

y = f (0; 0)

sin y + yex :

The Poincar lemma continues to hold in a very important class of sets,


namely simply connected sets.
Denition 172 Given a set E
RN , two continuous closed curves, with
parametric representations ' : [a; b] ! RN and
: [a; b] ! RN such that
' ([a; b]) E and ([a; b]) E, are homotopic in E if there exists a continuous function h : [a; b] [0; 1] ! RN such that h ([a; b] [0; 1]) E,
h (t; 0) = ' (t) for all t 2 [a; b] ;

h (t; 1) =

h (a; s) = h (b; s) for all s 2 [0; 1] :

(t) for all t 2 [a; b] ;

The function h is called a homotopy in E between the two curves.


Roughly speaking, two curves are homotopic in E if it is possible to deform
the rst continuously until it becomes the second without leaving the set E.
Denition 173 A set E
RN is pathwise connected if for every two points
x; y 2 E there exists a continuous curve joining x and y and with range contained in E.
Denition 174 A set E RN is simply connected if it is pathwise connected
and if every continuous closed curve with range in E is homotopic in E to a point
in E (that is, to a curve with parametric representation a constant function).
Example 175 A star-shaped set is simply connected. Indeed, let E
RN be
star-shaped with respect to some point x0 2 E and consider a continuous closed
curve with parametric representation ' : [a; b] ! RN such that ' ([a; b]) E.
Then the function
h (t; s) := s' (t) + (1 s) x0
is an homotopy between

and the point x0 .

Remark 176 It can be shown that R2 n a point is not simply connected, while
R3 n a point is. On the other hand, R3 n a line is not simply connected.
Friday, April 5, 2013
Given an open set U
RN and a dierentiable function g : U ! RN , a
dierentiable function h : U ! R is called an integrating factor for g if the
function
hg = (hg1 ; : : : ; hgN )
is irrotational.
Exercise 177 Assume that N = 2 and that g : U ! R2 is of class C 2 . Prove
that if
@g1
@g2
(x; y)
(x; y) = a (x) g2 (x; y) b (y) g1 (x; y)
@y
@x
for some continuous functions a : I ! R and b : J ! R, where I and J are
intervals, then an integrating factor is given by
h (x; y) = e

Rx

x0

a(t) dt

Ry

y0

where x0 2 I and y0 2 J are some xed points.


68

b(s) ds

Remark 178 Conservative elds are useful is solving di erential equations.


Consider the di erential equation
y 0 (x) =

g1 (x; y (x))
;
g2 (x; y (x))

where
g (x; y) = (g1 (x; y) ; g2 (x; y))
1

is of class C . If g conservative, then there exists a function f : U ! R such


that rf = g, so that
@f
@x (x; y (x))
;
y 0 (x) = @f
@y (x; y (x))
which implies that
@f
@f
(x; y (x)) +
(x; y (x)) y 0 (x) = 0:
@x
@y
Consider the function
h (x) := f (x; y (x)) :
By the chain rule, we have that
h0 (x) =

@f
@f
(x; y (x)) +
(x; y (x)) y 0 (x) = 0:
@x
@y

Hence, h is constant in each connected domain of its domain. Assume that h is


dened in some interval I, then
f (x; y (x)) = constant
for all x 2 I. Thus, we have solved implicitly our di erential equation.
If g is not irrotational, but it admits an integrating factor h 6= 0, then
y 0 (x) =

g1 (x; y (x))
=
g2 (x; y (x))

h (x; y (x)) g1 (x; y (x))


;
h (x; y (x)) g2 (x; y (x))

and if the function hg is conservative with rf = hg, then we can conclude as


before that
f (x; y (x)) = constant
for all x 2 I, so we have solved implicitly our di erential equation.
Example 179 We want to solve the di erential equations
y0 =

2 log (xy) + 1
x
y

Consider the function


g (x; y) =

2 log (xy) + 1;

69

x
y

Lets see if g is irrotational. We have


@
2
(2 log (xy) + 1) = 6=
@y
y

1
@
=
y
@x

x
y

: Hence, g is not irrotational, and so it cannot be conservative. Lets try to nd


an integrating factor h. We have
@g1
(x; y)
@y

Take a (x) =

3
x

@g2
2 1
(x; y) = + = a (x) g2 (x; y) b (y) g1 (x; y)
@x
y y
x
= a (x)
b (y) (2 log (xy) + 1) :
y

and b (y) = 0. Then

h (x; y) = e

Rx

x0

3
x

dt

Ry

y0

0 ds

3 logjxj+3 log x0

=e

= celogjxj

c
3

jxj

is an integrating factor. Indeed,


@
@y

1
(2 log (xy) + 1)
x3

2
@
=
3
x y
@x

1
x2 y

Hence,
h (x; y) g (x; y) =

1
jxj

(2 log (xy) + 1) ;

jxj y

is irrotational. Its domain is the set U = (x; y) 2 R2 : xy > 0 , which is given


by the disjoint union of two convex sets. In each of these convex sets hg is
conservative by the the Poincar lemma. Lets nd a potential f . We have
@f
1
(x; y) =
3 (2 log (xy) + 1) ;
@x
jxj

@f
(x; y) =
@y

x
3

jxj y

and so
f (x; y) =
=

Z
@f
x
(x; y) dy =
3 dy
@y
jxj y
x
3 log jyj + c (x) :
jxj

Hence,
f (x; y) =

x
jxj

log jyj + c (x) :

If x > 0 and y > 0 di erentiating with respect to x, we get


1
@f
@
(2 log (xy) + 1) =
(x; y) =
3
x
@x
@x

1
log y + c (x)
x2
70

2
log y + c0 (x)
x3

and so

2
1
2
2
log x + 3 log y + 3 = 3 log y + c0 (x) :
x3
x
x
x

It follows that
c0 (x) =

1
2
log x + 3 :
x3
x

Hence,
c (x) =

c0 (x) dx =

2
1
log x + 3
x3
x

dx =

1
(log x + 1) + c1 :
x2

: If x < 0 and y < 0 di erentiating with respect to x, we get


1
@
@f
(x; y) =
(2 log jxyj + 1) =
x3
@x
@x
and so

1
log jyj + c (x)
x2

2
1
log jyj + 3 =
x3
x

2
log jxj
x3

2
log jyj+c0 (x)
x3

2
log jyj + c0 (x) :
x3

It follows that
c0 (x) =

2
1
log jxj + 3 :
x3
x

Hence,
c (x) =

c0 (x) dx =

In conclusion the potential is


(
x
jxj3
x
jxj3

f (x; y) =

2
1
log ( x) + 3
3
x
x

log jyj
log jyj +

1
x2
1
x2

dx =

(log x + 1) + c1
log ( x) + c2

1
log ( x) + c2 :
x2

if x > 0;
if x < 0:

Monday, April 8, 2013

15

Integration
R, a rectangle in RN is a set of the

Given N bounded intervals I1 ; : : : ; IN


form
R := I1

IN :

The elementary measure of a rectangle is given by


meas R := length I1

length IN ;

where if In has endpoints an bn , then we set length In := bn an . To highlight


the dependence on N , we will use the notation measN .
Remark 180 Note that the intersection of two rectangles is still a rectangle.
We will use this fact a lot in what follows.
71

Given a rectangle R, by a partition P of R we mean a nite set of rectangles


R1 , . . . , Rn such that Ri \ Rj = ; if i 6= j and
n
[

R=

Ri :

i=1

Exercise 181 Let R


RN be a rectangle and let P = fR1 ; : : : ; Rn g be a
partition of R. Prove that
meas R =

n
X

meas Ri :

i=1

Hint: Use induction on N .


Given a rectangle R, consider a bounded function f : R ! R and a partition
P of R. We dene the lower and upper sums of f for the partition P respectively
by
L (f; P) :=
U (f; P) :=

n
X
i=1
n
X

meas Ri inf f (x) ;


x2Ri

meas Ri sup f (x) :


x2Ri

i=1

Since f is bounded, using the previous exercise, we have that


meas R inf f (x) =
x2R

n
X

meas Ri inf f (x)

i=1

U (f; P)

x2R

n
X

L (f; P)

meas Ri sup f (x)

(16)
meas R sup f (x) :

x2R

i=1

x2R

The lower and upper integrals of f over R are dened respectively by


Z
f (x) dx := sup fL (f; P) : P partition of Rg ;
R

f (x) dx := inf fU (f; P) : P partition of Rg :

Given a rectangle R and a bounded function f : R ! R, we say that f is


Riemann integrable over R if the lower and upper integral coincide. RWe call the
common value the Riemann integral of f over R and we denote it by R f (x) dx.
Thus, for a Riemann integrable function,
Z
Z
Z
f (x) dx := f (x) dx = f (x) dx:
R

72

Remark 182 Note that the function f = 1 is Riemann integrable over a rectangle R and
Z
1 dx = meas R:
R

To determine when a function is Riemann integrable we need to introduce


the notion of sets of Lebesgue measure zero.
Denition 183 A set E
RN has Lebesgue measure zero if for every " > 0
there exists a countable family of rectangles Rn such that
[
X
E
Rn and
meas Rn ":
n

Example 184 Lets see some examples.


(i) A singleton E = fcg has Lebesgue measure zero. Given " > 0, take R to
be the cube centered at c and of side-length "1=N .
(ii) If a set E contains an open rectangle R, then it cannot have Lebesgue
measure zero. Indeed, for any countable family of rectangles Rn , we have
[
R E
Rn ;
n

and so

meas R

meas Rn :

Taking " < meas R, we obtain a contradiction.


(iii) A countable set E = fxn gn has Lebesgue measure zero. Given " > 0, take
Rn to be the cube centered at xn and of side-length
X

meas Rn = "

X 1
2n
n

" 1=N
.
2n

Then

":

In particular, N, Q, and Z all have Lebesgue measure zero.


(iv) If fEk gk is a countable family of sets, each with Lebesgue measure zero,
then their union
[
E :=
Ek
k

has Lebesgue measure zero. Indeed, given " > 0 x k. Since Ek has
(k)
Lebesgue measure zero, there exists a countable family of rectangles Rn
such that
[
X
"
Ek
Rn(k) and
meas Rn(k)
:
k
2
n
n

73

Consider the family

n
o
(k)
Rn

n;k

E=

. It is still countable,

Ek

and

meas Rn(k) =

n;k

[[
k

XX

Rn(k)

meas Rn(k)

X "
2k

":

(v) There are uncountable sets that have Lebesgue measure zero. One such
example is given by the Cantor set.
Exercise 185 Prove that the boundary of a rectangle R
measure zero.

RN has Lebesgue

Exercise 186 Prove that in Denition 183, the rectangles Rn can be assumed
to be open.
The following theorem characterizes Riemann integrable functions.
Theorem 187 Given a rectangle R, a bounded function f : R ! R is Riemann
integrable if and only if the set of its discontinuity points has Lebesgue measure
zero.
Wednesday, April 10, 2013
Second midterm
Friday, April 5, 2013
The next theorem is very important in exercises. It allows to calculate triple,
double, etc.. integrals by integrating one variable at a time.
Theorem 188 (Repeated Integration) Let S
RN and T
RM be rectangles, let f : S T ! R be Riemann integrable and assume that for every
x 2 S, theR function y 2 T 7! f (x; y) is Riemann integrable. Then the function
x 2 S 7! T f (x; y) dy is Riemann integrable and
Z

f (x; y) d (x; y) =

S T

f (x; y) dy

dx:

(17)

Similarly, if for every y 2 T , the function


x 2 S 7! f (x; y) is Riemann inteR
grable, then the function y 2 T 7! S f (x; y) dx is Riemann integrable and
Z

f (x; y) d (x; y) =

S T

Example 189 Lets calculate

[0;1] [0;1]

f (x; y) dx

f (x; y) d (x; y), where

f (x; y) = x sin (x + y) :

74

dy:

Since f is continuous and bounded, we can apply the previous theorem. We have
Z
Z 1 Z 1
x sin (x + y) dy dx
x sin (x + y) d (x; y) =
[0;1] [0;1]

dx

y=1

x [ cos (x + y)]y=0 dx

sin (x + y) dy

x [ cos (x + 1) + cos (x + 0)] dx:

We use integration by parts with h (x) = x so that h0 (x) = 1 and g 0 (x) =


cos (x + 1) + cos x, so that g (x) = sin (x + 1) + sin x. We have
Z

x[ cos (x + 1) + cos x] dx
h

"

g0

#x=1

= x[ sin (x + 1) + sin x]
h

= sin 1

sin 2

0+

1 [ sin (x + 1) + sin x] dx

0
0 h

x=0

[sin (x + 1)

sin x] dx:

We can also dene the Riemann integral over bounded sets E. Given a
bounded set E RN , let R be a rectangle containing E. We say that function
f : E ! R is Riemann integrable over E if the function
g (x) :=

f (x) if x 2 R
0
if x 2 R n E

is Riemann integrable over R and we dene the Riemann integral of f over E


to be
Z
Z
f (x) dx :=
g (x) dx:
E

Exercise 190 Prove the previous denition does not depend on the choice of
the particular rectangle R containing E.
Given a bounded set E
RN , let R be a rectangle containing E. We say
that E is PeanoJordan measurable if the function E is Riemann integrable
over R. In this case the PeanoJordan measure of E is given by
Z
meas E :=
E (x) dx:
R

In dimension N = 2, the PeanoJordan of a set is called area of the set and in


dimension N = 3 it is called volume of a set.
75

Exercise 191 Prove that the previous denition does not depend on the choice
of the rectangle R containing E.
Remark 192 Note that a rectangle is PeanoJordan measurable and its Peano
Jordan measure coincides with its elementary measure (see Remark 182).
Theorem 193 A bounded set E RN is PeanoJordan measurable if and only
if its boundary is PeanoJordan measurable and it has PeanoJordan measure
zero.
Denition 194 A set E RN +1 is called a normal domain if it can be written
as
E = f(x; y) 2 R R : (x) y
(x)g ;

where R RN is a rectangle, and : R ! R and : R ! R are two Riemann


integrable functions, with (x)
(x) for all x 2 R.
Theorem 195 A normal domain E
Z
measN +1 E =

RN +1 is PeanoJordan measurable and


!
Z (x)
1 dy dx:
(x)

Theorem 196 Let E


RN +1 be a normal domain and let f : E ! R be a
bounded continuous function. Then f is Riemann integrable over E and
!
Z
Z
Z
(x)

f (x; y) d (x; y) =

f (x; y) dy

dx:

(x)

p
Example 197 Calculate the volume of the set E R3 bounded by z = x2 + y 2
and x2 + y 2 + z 2 z = 0. Note that the rst is a cone, while the second can be
2
written as x2 + y 2 + z 12 = 14 and so it is a sphere of center 0; 0; 12 and
1
radius 2 . To see where these surfaces intersect, lets solve the system
p
z 2 = x2 + y 2
z 2 = x2 + y 2
z = x2 + y 2
,
,
2
2
2
2
2
2
x +y +z
z=0
z2 + z2 z = 0
x +y +z
z=0
z 2 = x2 + y 2
,
z (2z 1) = 0

z 2 = x2 + y 2
z = 0 or z =

1
2

and so they intersect at the origin or at the plane z = 21 . Note that E is a


normal domain, since we can write it as
(
)
r
p
1
1
1
3
2
2
2
2
2
2
E = (x; y; z) 2 R : x + y
z
+
(x + y ); x + y
:
2
4
4
We have
meas3 E =

[ 1;1] [ 1;1] [0;1]

76

(x) dx:

Lets apply Theorem 188,


meas3 E =

[0;1]

((x; y; z)) d (x; y)

[ 1;1] [ 1;1]

dz:

So we are xing z0 2 [0; 1] and we are cutting the set E with the plane z = z0
and calculating in R2 the area of the section
Ez0 = (x; y) 2 R2 : (x; y; z0 ) 2 E
=

(x; y) 2 R2 : x2 + y 2
(x; y) 2 R2 : x2 + y 2

if z0 2 21 ; 1 ;
if z0 2 0; 21 :

z02

z0
z02

Hence,
!

dz
E (x; y; z) d (x; y)
[ 1;1] [ 1;1]
[0; 12 ]
!
Z
Z
+
dz
E (x; y; z) d (x; y)
[ 1;1] [ 1;1]
[ 12 ;1]
Z
Z
=
meas2 Ez dz +
meas2 Ez dz
[0; 12 ]
[ 12 ;1]
Z
Z
2
=
z dz +
z z 2 dz:
[0; 12 ]
[ 21 ;1]

meas3 E =

Another possibility would have been to use


meas3 E =

[ 1;1] [ 1;1]

(x; y; z) dz

[0;1]

d (x; y) :

Mondy, April 15, 2013


Exercise 198 Calculate the integral
Z
log (xy) dxdy;
E

where E is the set of R2 bounded by the curves xy = 1, x = y 2 , and x = 3.


Exercise 199 Calculate the integral
Z
x log 1 + y 2 dxdy;
E

n
where E = (x; y) 2 R2 :

x
4

y2
2

1; x2

y 2 = 1; x

o
0 .

Exercise 200 Find the volume of the set E of R3 enclosed by z =


p
and z = 4 x2 y 2 .
77

1
3

x2 + y 2

Next we discuss some properties of Riemann integration.


Proposition 201 Given a rectangle R, let f; g : R ! R be Riemann integrable.
(i) If

2 R, then f is Riemann integrable and


Z
Z
f (x) dx =
f (x) dx:
R

(ii) The functions f + g and f g are Riemann integrable and


Z
Z
Z
(f (x) + g (x)) dx =
f (x) dx +
g (x) dx:
R

(iii) If f

g, then

(18)

f (x) dx

(19)

g (x) dx:

(iv) The function jf j is Riemann integrable and


Z
Z
jf (x)j dx:
f (x) dx
R

Exercise 202 Give an example of a bounded function f : R ! R such that jf j


is Riemann integrable over R, but f is not.
Remark 203 If E; F
RN are two PeanoJordan measurable sets, with E
F , then E
F , and so by Proposition 201,
meas E

meas F:

Exercise 204 Prove that if E


RN and F
RN are PeanoJordan measurable and R is a rectangle containing E, then E [ F , E \ F , R n E are
PeanoJordan measurable.
Exercise 205 Prove that if E1 ; : : : ; En
RN are PeanoJordan measurable
and pairwise disjoint, then
!
n
n
[
X
meas
Ei =
meas Ei :
i=1

i=1

ExerciseS206 Prove that if E1 ; : : : ; En


RN are PeanoJordan measurable
n
and E
i=1 Ei is PeanoJordan measurable, then
meas E

n
X
i=1

78

meas Ei :

Theorem 207 Given a bounded set E RN and a Riemann integrable function


f : E ! R, if F E is PeanoJordan measurable, then f is integrable over F .
Theorem 208 If E RN is given by the disjoint union of a nite number of
PeanoJordan measurable,
N
[
E=
En ;
n=1

and if f : E ! R is f is Riemann integrable over E, then


Z

f dx =

Corollary 209 If E
normal domains,

N Z
X

f dx:

En

n=1

RN is given by the disjoint union of a nite number of


E=

N
[

En ;

N
X

measN En

n=1

then
measN E =

n=1

and if f : E ! R is a bounded continuous function, then f is Riemann integrable


over E and
Z
N Z
X
f dx:
f dx =
E

Theorem 210 Let E


that is,

En

n=1

RN be symmetric with respect to the hyperplane xi = 0,

x = (x1 ; : : : ; xi ; : : : ; xN ) 2 E ) x = (x1 ; : : : ; xi ; : : : ; xN ) 2 E
and let f : E ! R is Riemann integrable over E. If f is odd with respect to xi ,
that is,
f (x1 ; : : : ; xi ; : : : ; xN ) = f (x1 ; : : : ; xi ; : : : ; xN ) ;
then

f dx = 0;

while if is even with respect to xi , that is,


f (x1 ; : : : ; xi ; : : : ; xN ) = f (x1 ; : : : ; xi ; : : : ; xN ) ;
then

f dx = 2

where E1 = fx 2 E : xi

E1

0g.
79

f dx;

Example 211 Lets compute the area of a circle of radius r centered at (0; 0).
We have
E = (x; y) 2 R2 : x2 + y 2 r2
and so

meas2 E =

1 dxdy:

Since f = 1 is even in x and in y and since E is symmetric with respect to both


axes, we have that
Z
meas2 E = 4
1 dxdy;
E1

where

E1 = (x; y) 2 R2 : x2 + y 2 r2 ; x 0; y 0
o
n
p
r 2 x2 ; 0 x r :
= (x; y) 2 R2 : 0 y

Since E1 is normal,

meas2 E = 4

r 2 x2

1 dy

=4

hp
r2

x2

dx = 4

i
0 dx:

p
y= r 2 x2

[y]y=0

dx

Using the change of variables x = r cos , where 0


2 , we get dx =
p
p
p
2
2
2
2
2
2
2
x = r
r cos = r sin = r sin , so that
r sin d and r
Z r hp
Z =2 p
i
meas2 E = 4
r2 x2 0 dx = 4
r2 r2 cos2 ( r sin ) d
0

= 4r2

=2

sin2

16

1
d = 4r2 :
4

Change of Variables

Theorem 212 Let U


RN be an open set and let g : U ! RM be a function
1
of class C , with N M . Let E RN be a PeanoJordan measurable set with
E U . Moreover, if N = M , assume that E has measure zero. Then g (E) is
PeanoJordan measurable with measure zero.
Theorem 213 (Change of variables for multiple integrals) Let U RN
be an open set and let g : U ! RN be a one-to-one function of class C 1 such
that det Jg (x) 6= 0 for all x 2 U . Let E
RN be PeanoJordan measurable
with E
U and let f : g (E) ! R be Riemann integrable. Then the function
x 2 E 7! f (g (x)) jdet Jg (x)j is Riemann integrable and
Z
Z
f (y) dy =
f (g (x)) jdet Jg (x)j dx:
g(E)

80

Example 214 Lets calculate the integral


ZZ
(x + y) dxdy;
F

where
F := (x; y) 2 R2 : 0 < x < y < 2x; 1 < xy < 2 :

Consider the change of variables u = xy and v = xy . Lets solve for x and y.


p
p
We have x = uv and y = uv. Dene g : U ! R2 as follows
g (u; v) :=

u p
; uv ;
v

where
U := (u; v) 2 R2 : u > 0; v > 0 :

Then g is one-to-one and its inverse is given by g 1 (x; y) := xy; xy . (Note


that in general it is much simpler to check that g is one-to-one than nding the
inverse). Moreover,
@g1
@u
@g2
@u

det Jg (u; v) = det

p1
2p uv
pv
2 u

= det

@g1
@v
@g2
@v
p
u
3=2
p2v
pu
2 v

(u; v)
(u; v)

(u; v)
(u; v)
!
=

1
:
2v

Set E = (1; 2) (1; 2) and note g (E) = F . By the change of variables theorem,
ZZ
ZZ
(x + y) dxdy =
(g1 (u; v) + g2 (u; v)) jdet Jg (u; v)j dudv
F
E
r
ZZ
1
u p
=
+ uv
dudv
v
2v
E
p
p
ZZ
u
u
1
+
dudv:
=
3=2
2
v
v
E
It follows that
ZZ

1
(x + y) dxdy =
2
=

1
2

v 3=2

u
+
v

u log 2 du =

dv

1
du =
2

p
1
(log 2) 2 2
3

v 1=2

v=2

u log v

1 :

Wednesday, April 17, 2013


The previous theorem unfortunately does not work for polar coordinates.
The problem is that polar coordinates are not one-to-one.

81

v=1

du

Corollary 215 Let E RN be a PeanoJordan measurable set, let g : E ! RN


@gi
be such that there exist @x
in E for all i; j = 1; : : : ; N , and that they are bounded
j
and continuous. Assume that there exists a PeanoJordan measurable set E1
E such that meas (E1 ) = meas (g (E1 )) = 0, E n E1 is open, det Jg (x) 6= 0
for all x 2 E n E1 and g : E n E1 ! RN is one-to-one. Let f : g (E) ! R
be Riemann integrable. Then the function x 2 E 7! f (g (x)) jdet Jg (x)j is
Riemann integrable and
Z
Z
f (y) dy =
f (g (x)) jdet Jg (x)j dx:
g(E)

The most important applications of the previous corollary is given by polar


coordinates in R2 . Let P = (x; y) 2 R2 n f(0; 0)g, let O = (0; 0), and let be
the angle that the OP forms with theppositive x axis. Then 0
2 and
x = r cos and y = r sin , where r = x2 + y 2 . Hence,
g (r; ) := (r cos ; r sin ) ;

(r; ) 2 [0; 1)

[0; 2 ) :

Note that partial derivatives of g of any order exist in R2 and are continuous.
The Jacobian of g is given by
det Jg (r; ) = det

@g1
@r
@g2
@r

= det

cos
sin

(r; )
(r; )

@g1
@
@g2
@

r sin
r cos

(r; )
(r; )
= r cos2 + r sin2 = r;

which is zero at r = 0. We observe that g is onto but g is one-to-one only in


the set where (0; 1) [0; 2 ). However, this is not open. Consider instead, the
open sets U := (0; 1) (0; 2 ) and V := R2 n f(x; 0) : x 0g and lets prove
that g : U ! V is invertible. Note that we cannot apply Theorem 213 to g
but we can apply Corollary 215, since given a PeanoJordan measurable set
E [0; 1) [0; 2 ) the badset E1 is given by the intersection of E with the
two lines r = 0 and = 0, precisely,
E1 := E n (f(r; 0) : r

0g [ f(0; ) :

2 [0; 2 )g) :

Moreover g (E1 ) is a bounded subset of the x-axis. Hence, meas (E1 ) = meas (g (E1 )) =
0.
Example 216 Lets calculate the integral
ZZ p 2
x + y2
dxdy;
x
F
where
F := (x; y) 2 R2 : x2 + y 2 > 1; x2 + y 2

82

2x < 0 :

Note that the domain is symmetric with respect to the x-axis and f is even in
y. Hence,
ZZ p 2
ZZ p 2
x + y2
x + y2
dxdy = 2
dxdy;
x
x
F
F1
where
F1 = (x; y) 2 R2 : x2 + y 2 > 1; x2 + y 2

2x < 0; y

0 :

Using polar coordinates we have r2 cos2 + r2 sin2 > 1, that is, r > 1, and
r2 cos2 + r2 sin2
2r cos < 0, that is, r (r 2 cos ) < 0, which gives r <
2 cos . So 1 < r < 2 cos and 0
2 . However, for the rst inequality to
make sense we need 1 < 2 cos , that is, 12 < cos , so 0
< 3 . It follows that
F1 is the image of the set
o
n
E1 = (r; ) 2 R2 : 0
< ; 1 < r < 2 cos ;
3
which is a normal domain. Using polar coordinates, we get
ZZ p 2
ZZ p 2
x + y2
r cos2 + r2 sin2
2
dxdy = 2
jrj d dr
x
r cos
F1
E1
!
Z =3 Z 2 cos
Z =3
r
1
=2
dr d =
cos
cos
0
1
0
Z =3
Z =3
1
=
4 cos2
1 d =
4 cos
cos
0
0

r2

r=2 cos
r=1

1
cos

For the second integral


Z

=3

1
d =
cos

=3

cos
cos2

d =

=3

cos
1 sin2

and we can now make the change of variables t = sin , so that dt = cos dt. It
follows that
Z

=3

cos
1 sin2

d =

3=2

1
1

t2

dt

Z p3=2
1
1
1
=
+
dt
2 0
1 t 1+t
p
1
t= 3=2
= [log j1 tj + log j1 + tj]t=0
2
1
1p
1
1p
= log 1
3 + log
3+1 :
2
2
2
2

Now lets introduce spherical coordinates in R3 . Let P = (x; y; z) 2 R3 n


f(0; 0; 0)g, let O = (0; 0; 0), and let be the angle that the OP forms with the
83

d
d :

positive z
Project P
OQ forms
where 0

p
axis. Then 0
2 and z = r cos , where r = x2 + y 2 + z 2 .
in the plane z = 0 to get a point Q and let ' be the angle that the
with the positive x-axis. Then x = r sin cos ' and y = r sin sin ',
' 2 . Then

g (r; ; ') := (r sin cos '; r sin sin '; r cos ) ;

(r; ; ') 2 [0; 1) [0; 2 ] [0; 2 ] :

Note that partial derivatives of g of any order exist in R3 and are continuous.
The Jacobian of g is given by
1
0 @g
@g1
@g1
1
@r (r; ; ')
@ (r; ; ')
@' (r; ; ')
C
B 2
@g2
@g2
det Jg (r; ; ') = det @ @g
@r (r; ; ')
@ (r; ; ')
@' (r; ; ') A
@g3
@g3
@g3
@r (r; ; ')
@ (r; ; ')
@' (r; ; ')
1
0
sin cos ' r cos cos '
r sin sin '
= det @ sin sin ' r cos sin ' r sin cos ' A = r2 sin > 0
cos
r sin
0

for r > 0 and, say, 0 <

< .

Example 217 Lets calculate the integral


ZZZ
1
p
dxdydz;
2 + y2 + z2
x
F
where

F := (x; y; z) 2 R3 : 1

x2 + y 2 + z 2 < 4; z

0 :

Since z 0, we have 2 0; 2 , while 1 < r < 2, 0 ' 2 . Hence, F is the


image of the set
n
o
E = (r; ; ') 2 R3 : 1 r < 2; 2 0;
;0 ' 2
2
and so
ZZZ
p
F

1
x2

y2

z2

dxdydz =
=

ZZZ

1 2
r sin
E r
Z 2
Z
r dr
1

r2
2

drd d'
=2

sin d

! Z

1 d'

r=2

[ cos ]

= =2
=0

2 =

r=1

1
2

(0

1) 2 :

Other important coordinates are cylindrical coordinates in R3 . We project


P = (x; y; z) into the plane z = 0 and in the plane xy we use polar coordinates.
Then
g ( ; '; z) := ( cos '; sin '; z) ;
84

( ; '; z) 2 [0; 1)

[0; 2 ]

R;

p
where = x2 + y 2 . Note that the partial derivatives of g of any order exist
in R3 and are continuous. The Jacobian of g is given by
1
0 @g
@g1
@g1
1
@ ( ; '; z)
@' ( ; '; z)
@z ( ; '; z)
C
B 2
@g2
@g2
det Jg ( ; '; z) = det @ @g
@ ( ; '; z)
@' ( ; '; z)
@z ( ; '; z) A
@g3
@g3
@g3
@ ( ; '; z)
@' ( ; '; z)
@z ( ; '; z)
0
1
cos '
sin ' 0
cos '
0 A= >0
= det @ sin '
0
0
1
for

> 0.

Example 218
p Lets calculate the volume of the bounded set F included between
the cone z = x2 + y 2 and the paraboloid z = x2 + y 2 . Hence,
n
o
p
F := (x; y; z) 2 R3 : x2 + y 2 z
x2 + y 2 ; x 2 + y 2 1 :
We have 0

'

2 ,

,0

E = ( ; '; z) 2 R3 :

1. Hence, F is the image of the set

;0

Note that E is a normal domain. Hence,


ZZZ
ZZZ
1 dxdydz =
1 j j d d'dz
F

E
1

=2

=2

dz
2

1; 0

dz

d =2

d =2

1 d'
Z

'

z=

[z]z=

=1

=0

1
=
:
6

Friday, April 19, 2013


Carnival, no classes
Monday, April 22, 2013

17

Dierential Surfaces

Denition 219 Given 1


k < N , a nonempty set M
RN is called a kdimensional dierential surface or manifold if for every x0 2 M there exist an
open set U containing x0 and a di erentiable function ' : W ! RN , where
W Rk is an open set such that
85

(i) ' : W ! M \ U is a homeomorphism, that is, it is invertible and continuous together with its inverse ' 1 : M \ U ! W ,
(ii) J' (y) has maximum rank k for all y 2 W .
The function ' is called a local chart or a system of local coordinates or
a local parametrization around x0 . We say that M is of class C m , m 2 N,
(respectively, C 1 ) if all local charts are of class C m (respectively, C 1 ).
Roughly speaking a set M
RN is a k-dimensional dierential surface if
for every point x0 2 M we can cut a piece of M around x0 and deform
it/atten it in a smooth way to get, say, a ball of Rk . Another way to say
this is that locally M looks like Rk . Thus the range of a simple dierentiable
curve is a 1-dimensional surface since locally it looks like R, while the range of a
self-intersecting curve is not, because at self-intersection point the range of the
curve does not look like R. Similarly, a sphere in R3 is a 2-dimensional surface
because locally it looks like R2 , while a cone is not because near the tip it does
not look like R2 .
A simple way to construct k-dimensional dierential surface is to start with
a set of Rk and then deform it in a smooth way.
Remark 220 The di erence between curves and surfaces, its that a surface is
a set of RN , while a curve is an equivalence class of functions. Also for surfaces
we do not allow self-intersections. Also, a curve can be described by only one
parametrization, while a surface usually needs more than one.
Remark 221 Note that Theorem 212 implies that if a bounded k-dimensional
di erential surface is PeanoJordan measurable, then its measure must be zero.
Example 222 Consider the hyperbola
M := (x; y) 2 R2 : x2

y2 = 1 :

Note that the set M is not the range of a curve (why?). To cover M we need at
least two local charts, precisely, we can take the open sets
V := (x; y) 2 R2 : x > 0 ;

W := (x; y) 2 R2 : x < 0 ;

and the functions ' : R ! M \ V and : R ! M \ W dened by


p
p
1 + t2 ; t ;
(t) :=
1 + t2 ; t ; t 2 R:
' (t) :=

Note that both ' and

are of class C 1 (the argument inside the square roots

is never zero). Moreover, '0 (t) =

p t
;1
1+t2

and

(t) :=

so the rank of J' (t) and of J (t) is one. Finally, '


1
: M \ W ! R are given by
'

(x; y) = y;

p t
;1
1+t2

: M \ V ! R and

(x; y) = y;

which are continuous. Thus, M is a 1-dimensional surface of class C 1 .


86

, and

Rk and a di erential function f : V !

Example 223 Given an open set V


R, consider the graph of f ,
Gr h := f(y; t) 2 V

R : t = f (y)g :

A chart is given by the function ' : V ! Rk+1 given by ' (y) := (y; f (y)).
Then,
Ik
J' (y) =
;
rf (y)
which has rank N . Note that ' is one-to-one and that ' (V ) = Gr h. Hence,
there exists ' 1 : Gr f ! V . Moreover, ' 1 is continuous, since the projection
: Rk+1 ! Rk
(y; t) 7! y

is of class C 1 and ' 1 is given by the restriction of


an k-dimensional di erential surface.

to Gr f . Hence, Gr f is

The next proposition gives an equivalent denition of surfaces, which is very


useful for examples.
Proposition 224 Given 1
k < N , a nonempty set M
then the following are equivalent:

RN ,and m 2 N,

(i) M is a k-dimensional surface of class C m .


(ii) For every x0 2 M there exist an open set U
RN containing x0 and a
N k
m
function g : U ! R
of class C , such that
M \ U = fx 2 U : g (x) = 0g
and Jg (x) has maximum rank N

(20)

k for all x 2 M \ U .
Wednesday, April 24, 2013

Example 225 (Torus) A torus is a 2-dimensional surface M obtained from a


rectangle of R2 by identifying opposite sides. Consider the chart ' : (0; 2 )
(0; 2 ) ! R2 given by
' (u; v) := ((r cos u + a) cos v; (r cos u + a) sin v; r sin u) ;

(21)

where a > 0 and r > 0. Dene M := ' ((0; 2 ) (0; 2 )) R3 .


First method. Lets prove that M is a 2-dimensional surface. Note that ' is
of class C 1 and
0
1
r sin u cos v
(r cos u + a) sin v
(r cos u + a) cos v A :
J' (u; v) = @ r sin u sin v
r cos u
0
87

Observe that if r cos u + a = 0, then


0
J' (u; v) = @

r sin u cos v
r sin u sin v
r cos u

1
0
0 A
0

and so J' (u; v) cannot have rank 2. Thus, for M to be a 2-dimensional surface
we need to assume that r cos u + a 6= 0 for all u 2 (0; 2 ). In what follows we
assume that
a > r:
r sin u cos v
(r cos u + a) sin v
has determinant r sin u (r cos u + a).
r sin u sin v
(r cos u + a) cos v
Then r cos u + a > 0. If sin u 6= 0, then the determinant is di erent from zero
and so so J' (u; v) has rank 2. On the other hand, when sin u = 0, that is, when
u = , then
1
0
0
( r + a) sin v
( r + a) cos v A :
J' ( ; v) = @ 0
r
0
The submatrix

For any v 2 (0; 2 ), either cos v 6= 0 or sin v 6= 0 (or both) and so either the
0 ( r + a) cos v
0
( r + a) sin v
submatrix
or
have determir
0
r
0
nant di erent from zero. Thus, we have shown that when a > r, J' (u; v) has
rank 2 for all (u; v) 2 (0; 2 ) (0; 2 ).
To see that ' is one-to-one, consider (x; y; z) 2 M . We want to nd a
unique pair (u; v) such that ' (u; v) = (x; y; z). Note that to nd u we can use
sin u = zr . The problem is that there are two values of u, one in 0; 2 [ 32 ; 2
and one in 2 ; 32 . If x2 + y 2 a2 , then
2

x2 + y 2 = (r cos u + a) cos2 v + (r cos u + a) sin2 v = (r cos u + a)

a2 ;

3
that is cos u (r cos u + 2a)
0, so that cos u
0, that is, 2
u
2 . So in
this case this determines uniquely u in terms of (x; y; z). On the other hand, if
x2 + y 2 > a2 , then cos u > 0, and so either 0 < u < 2 or 32
u < 2 . Again,
this determines uniquely u in terms of (x; y; z). In turn,

cos v =

x
;
(r cos u + a)

sin v =

y
;
(r cos u + a)

which determine uniquely v. Thus, ' is one-to-one. We leave as an exercise to


check that ' 1 is continuous.
2
Second method: Lets write down M explicitly. We have x2 +y 2 = (r cos u + a) ,
2
p
p
and so x2 + y 2 = r cos u + a > 0 (since a > r). In turn,
x2 + y 2 a =
r2 cos2 u. Adding r2 sin2 u = z 2 to both sides gives
M=

(x; y; z) 2 R3 : z 2 = r2
88

p
x2 + y 2

Consider the function


g (x; y; z) := z 2 +

x2 + y 2

r2 :

Note that g is not of class C 1 in R3 , since at points (0; 0; z), we have problems.
However, these points do not belong to M . Indeed,
g (0; 0; z) = z 2 + a2

r2

a2

r2 > 0:

Hence, we can take U = (x; y; z) 2 R3 : x2 + y 2 > 0 . Then g is of class C 1


in U since
Jg (x; y; z) = rg (x; y; z)
=

x2

y2

2x

p
;2
2 x2 + y 2

x2

y2

2y

p
; 2z
x2 + y 2

is continuous in U . To prove that Jg (x; y; z) has rank one in U , note that if


2
2
z 6= 0, then @g
@z = 2z 6= 0, while if z = 0, then x + y > 0, so x and y cannot be
@g
@g
both zero, so @x (x; y; z) or @y (x; y; z) are di erent from zero. Thus, Jg (x; y; z)
has maximum rank 1. This shows that M is a 2-dimensional surface of class
C 1.
Note that M is obtained by taking the circle of radius r and center (0; a; 0),
namely,
2
(y a) + z 2 = r2
and revolving it about the z.
The Klein bottle is obtained by starting from a rectangle of R2 , reecting
one of its sides across its center and then performing the identication.
Exercise 226 (Klein Bottle) Consider the function ' : (0; 2 ) (0; 2 ) ! R4
given by
u
u
' (u; v) := (r cos v + a) cos u; (r cos v + a) sin u; r sin v cos ; r sin v sin
:
2
2
Then Klein bottle is the set M := ' ((0; 2 )
2-dimensional surface of class C 1 .

R4 . Prove that M is a

(0; 2 ))

Example 227 Consider the unit sphere in RN , N

2,

M := x 2 RN : kxk = 1 :
2

Given x0 2 M consider the function g (x) := kxk


1 (here N k = 1 and so
k = N 1). Then Jg (x) = rg (x) = 2x. By taking a ball B (x0 ; r) so small that
it does not intersect the origin, we have that 2x 6= 0 for all x 2 M \ B (x0 ; r)
and so M is a (N 1)-dimensional surface of class C 1 .
89

Example 228 Given an open set U


RN and a function f : U ! R of class
1
C , for every c 2 R consider the level set
Bc := fx 2 U : f (x) = cg :
Given x0 2 Bc consider the function g (x) := f (x) c (here N k = 1 and so
k = N 1). Then Jg (x) = rf (x). Hence, we have a problem if Bc contains
some critical points. Thus, let
M := fx 2 U : f (x) = cg n fx 2 U : rf (x) = 0g :
Given x0 2 M , since rf (x0 ) 6= 0 and f is of class C 1 , by taking a small ball
B (x0 ; r) we have that rf (x) 6= 0 for all x 2 M \ B (x0 ; r) and so M is an
(N 1)-dimensional surface of class C 1 .
Friday, April 26, 2013
Denition 229 Given a k-dimensional surface M and a point x0 2 M , a
vector t is a tangent vector to M at x0 if there exists a function : ( r; r) ! M
such that (0) = x0 and
is di erentiable at t = 0 with 0 (0) = t. The set
of all tangent vectors to M at x0 is called the tangent space to M at x0 and is
denoted T (x0 ).
We will see that T (x0 ) is a vector space of dimension k.
Denition 230 Given a k-dimensional surface M and a point x0 2 M , a
vector n is a normal vector to M at x0 if n t = 0 for all t 2 T (x0 ).
Theorem 231 Given a k-dimensional surface M of class C 1 and a point x0 2
@'
(y0 ),
M , if ' : W ! RN is a local chart with ' (y 0 ) = x0 then the vectors @y
1

...,

@'
@yk

(y0 ) form a basis for the tangent space T (x0 ). Hence, a vector n is a

normal vector to M at x0 if n

@'
@yi

(y0 ) = 0 for all i = 1; : : : ; k.

Theorem 232 Given a k-dimensional surface M of class C 1 and a point x0 2


M , if M is represented as in (20), then the tangent space T (x0 ) is given by
the kernel (or null space) of the matrix Jg (x0 ). The vectors rg1 (x0 ), . . . ,
rgN k (x0 ) forms a basis for the normal space.

18

Surface Integrals

To dene the integral of a function over a surface, we use local charts. Given
1 k < N we dene
N;k

:=

2 Nk : 1

<

<

<

N :

Given a k-dimensional surface M of class C m , m 2 N, consider a local chart


' : V ! M , let E
' (V ) be such that ' 1 (E) is PeanoJordan measurable
90

and let f : E ! R be a bounded function. The surface integral of f over E is


dened as
v
u X
Z
Z
2
u
@ (' 1 ; : : : ; ' k )
f dHk :=
f (' (y)) t
det
(y) dy; (22)
@ (y1 ; : : : ; yk )
E
' 1 (E)
2

N;k

P
@ (' 1 ;:::;' k )
(E) 7! f (' (y))
2 N;k det @(y1 ;:::;yk ) (y)
R
is Riemann integrable. Note that E f dHk exists if f is bounded and continuous. In particular, if f = 1, the number
v
u X
Z
Z
2
u
@ (' 1 ; : : : ; ' k )
k
t
(y) dy
H (E) :=
det
1d =
@ (y1 ; : : : ; yk )
E
' 1 (E)

provided the function y 2 '

N;k

is called the k-dimensional surface measure of E.

Remark 233 To understand the expression under the square root, observe that
the Jacobian matrix of ',
0
1
@'1
@'1
@y1 (y)
@yk (y)
B
C
..
..
C
J' (y) = B
.
.
@
A
@'N
@'N
(y)
(y)
@y1
@yk

is an N k matrix, where 1
k < N . The expression under square root is
obtained by considering the sum of the squares of the determinant of all k k
submatrices of J' (y).
R
Remark 234 It can be shown that the number E f dHk does not depend on
the particular local chart '.
Example 235 Consider the set
M := (x; y; z; w) 2 R4 : x2 + y 2 = 1; z 2 + w2 = 1; x; z > 0 :
Lets prove that it is a 2-dimensional surface. A (local) chart ' :
2 ; 2 ! M is given by
' (u; v) := (cos u; sin u; cos v; sin v) ;
so that

sin u
B cos u
J' (u; v) = B
@
0
0
91

1
0
C
0
C:
sin v A
cos v

2; 2

cos u
0
has determinant cos u cos v <
0
cos v
0, and so J' (u; v) has rank 2. Moreover ' is one-to-one with continuous inverse
and of class C 1 . Hence,

Since cos u; cos v > 0, the submatrix

X
2

det

@ (' 1 ; ' 2 )
(u; v)
@ (u; v)

sin u 0
cos u 0

det

sin u
0

+ det

0
sin v

4;2

sin u
0

+ det
+ det

0
cos v

cos u
0
0
cos v

cos u
0

+ det
2

+ det

0
0

sin v
sin v

0
sin v

= 0 + sin2 u sin2 v + sin2 u cos2 v + cos2 u sin2 v + cos2 u cos2 v + 0


= 1:
Lets now calculate the surface integral
Z

x2 + z 2 dH2 =

Z
2

x2 + z 2 dH2 . We have

cos2 u + cos2 v

cos u du +
2

1 dudv

cos2 v dv =

Monday, April 29, 2013

19

Divergence Theorem

Denition 236 An open and bounded set U RN is regular if for every x0 2


@U there exist an open set V
RN containing x0 and a function g : V ! R ,
with rg (x) 6= 0 for all x 2 V , such that
U \ V = fx 2 V : g (x) < 0g ;

V \ @U = fx 2 V : g (x) = 0g :
It follows that @U is an (N

1)-dimensional surface of class C 1 .

Denition 237 Given an open, bounded, regular set U


RN , a unit normal
vector to @U at x0 is called a unit outward normal to U at x0 if there exists
> 0 such that x0 t 2 U and x0 + t 2 RN n U for all 0 < t < .
Exercise 238 Prove that if U RN is an open, bounded, regular set, and g is
as in Denition 236, then for every x0 2 @U , the unit outward normal to U at
x0 is the unit vector
rg (x0 )
(x0 ) =
:
krg (x0 )k
Hence,

: @U ! RN is a continuous function.
92

We are ready to state the divergence theorem.


Theorem 239 (Divergence Theorem) Let U
RN be an open, bounded,
N
regular set and let f : U ! R be such that f is bounded and continuous in
U and there exist the partial derivatives of f in RN at all x 2 U and they are
continuous and bounded. Then
Z
Z
div f (x) dx =
f (x)
(x) dHN 1 (x) ;
U

@U

where
div f :=

N
X
@fi
:
@xi
i=1

R
Remark 240 In physics @U f (x) (x) dHN 1 (x) represents the outward ux
of a vector eld f across the boundary of a region U .
We recall that given two vectors a = (a1 ; a2 ; a3 ) and b = (b1 ; b2 ; b3 ) of R3 ,
their cross-product or vector-product is dened by
b := (a2 b3
0

a3 b2 ; a3 b1

e1
= det @ a1
b1

e2
a2
b2

a1 b3 ; a1 b2
1

a2 b1 )

e3
a3 A :
b3

Remark 241 For a 2-dimensional manifold M in R3 , given a local chart ' :


W ! R2 , where W
R2 is an open set, we have that a basis for the tangent
@'
space at a point ' (u; v) 2 M is given by @'
@u (u; v) and @v (u; v) and the two
unit normals are given by
@'
@u
@'
@u

(u; v)
(u; v)

@'
@v
@'
@v

(u; v)

It follows by direct computation that @'


@u (u; v)
s
2
P
@ (' 1 ;' 2 )
det
(u;
v)
. Hence,
2 3;2
@(u;v)
Z

f (x; y; z)

f (' (u; v))

@'
@v

(u; v) is given exactly by

(x; y; z) dH2 (x; y; z)

'(W )

(u; v)

f (' (u; v))

@'
@u
@'
@u

(u; v)

(u; v)

@'
(u; v)
@u

v
(u; v) u
u X
@ (' 1 ; ' 2 )
t
det
(u; v)
@'
@ (u; v)
(u;
v)
2
3;2
@v
@'
@v

@'
(u; v)
@v

where the sign has to be chosen appropriately.


93

dudv;

dudv

If E RN and f : E ! RN is dierentiable, then f is called a divergence-free


eld or solenoidal eld if
div f = 0:
Thus for a smooth solenoidal eld, the outward ux across the boundary of a
regular set U is zero.
Corollary 242 The theorem continues to hold if U RN is open, bounded, and
its boundary consists of two sets E1 and E2 , where E1 is a closed set contained
in the nite union of closed and bounded manifolds of class C 1 and dimension
less than N
1, while for every x0 2 E2 there exist a ball B (x0 ; r) and a
function g 2 C 1 (B (x0 ; r)) such that with rg (x) 6= 0 for all x 2 B (x0 ; r) \ @U ,
such that
B (x0 ; r) \ U = fx 2 B (x0 ; r) : g (x) < 0g ;

B (x0 ; r) n U = fx 2 B (x0 ; r) : g (x) > 0g ;

B (x0 ; r) \ @U = fx 2 B (x0 ; r) : g (x) = 0g :


Remark 243 In particular, the previous corollary works if U
RN is open,
bounded, and @U is given by a nite union of manifolds of dimension N 1 and
a nite union of manifolds of dimension less than N 1.
Tuesday, April 30, 2013
Recitation.
Example 244 Lets calculate the outward ux of the function
f (x; y; z) := x; y; z 2
across the boundary of the region
U := (x; y; z) 2 R3 :

1<z<

x2

y2 :

First method. Lets use the divergence theorem. Note that U is not a regular
open set (why?) but it satises the hypotheses of the previous corollary (why?).
We have
div f (x; y; z) =

@
@
@
(x) +
(y) +
z 2 = 1 + 1 + 2z;
@x
@y
@z

and so by the divergence theorem


Z
Z
f
dH2 = 2
(1 + z) dxdydz:
@U

94

Since U is a normal domain, we have


2

(1 + z) dxdydz = 2

=2

=2

x2 y 2

(1 + z) dz
1

dxdy

z= x2 y 2

z2
z+
2

dxdy
z= 1

1 2
x + y2
2

x2

y2 +

1
2

dxdy;

where E = (x; y) 2 R2 : x2 + y 2 < 1 . Now using polar coordinates,


Z

1 2
x + y2
2

1
y +
2
2

dxdy =

=2

=2

1 4
r
2

1
2

1
r3 + r
2

1 5
r
2

1 6
r
12

r2 +

r4
r2
+
4
4

r dr

dr

r=1
r=0

1
=
6
Second method. Lets use the denition of outward ux. The boundary of U is
given by the union of two surfaces and by a curve (which is a lower dimensional
manifold). Precisely,
@U = M1 [ M2 [ M3 ;
where
M1 := (x; y; z) 2 R3 : z =

1; x2 + y 2 < 1 ;

M3 := (x; y; z) 2 R3 : z =

1; x2 + y 2 = 1 :

M2 := (x; y; z) 2 R3 : z =

x2

y 2 ; x2 + y 2 < 1 ;

Note that M1 and M2 are 2-dimensional surfaces, since M1 is the graph of


the function g1 (x; y) = 1 over E = (x; y) 2 R2 : x2 + y 2 < 1 and M2 is the
graph of the function g2 (x; y) = x2 y 2 over E. On theR other hand, M3
is a one-dimensional manifold. Hence, we need to compute M1 f
dH2 and
R
2
f
dH .
M2
A chart for M1 is given by
' (x; y) = (x; y; 1) ;
We have

(x; y) 2 E:

@'
(x; y) = (0; 1; 0) :
@y

@'
(x; y) = (1; 0; 0) ;
@x

95

Hence, to nd the normal, we need a vector orthogonal to (1; 0; 0) and (0; 1; 0),
namely, (0; 0; 1). The outward normal is (0; 0; 1). Hence, using Remark 241
Z
Z
2
f
dH =
(x; y; 1) (0; 0; 1) dxdy
M1
E
Z
=
1 dxdy =
:
E

A chart for M2 is given by


' (x; y) = x; y; x2
We have

y2 ;

(x; y) 2 E:

@'
(x; y) = (0; 1; 2y) :
@y

@'
(x; y) = (1; 0; 2x) ;
@x

Hence, to nd the normal, we need


@'
(x; y)
@x

e1
@'
(x; y) = det @ 1
@y
0

1
e3
2x A
2y

e2
0
1

= (2x; 2y; 1) :

Hence, the normals vectors are (2x; 2y; 1). The outward normal is (2x; 2y; 1).
Hence, using Remark 241
Z
Z
2
2
x; y; x2 y 2
(2x; 2y; 1) dxdy
f
dH =
M1
ZE
2
2x2 + 2y 2 + x2 + y 2
dxdy:
=
E

Now using polar coordinates,


Z
2x2 + 2y 2 + x2 + y 2

dxdy =

=2

2r2 + r4 r dr

2r3 + r5 dr

=2

1 4 r6
r +
2
6

r=1
r=0

4
=
:
3
Example 245 Lets calculate the outward ux of the function
f (x; y; z) := (0; yz; x)
across the boundary of the region
U := (x; y; z) 2 R3 : x2 + y 2 < z 2 ; x2 + y 2 + z 2 < 2y; z > 0 :
96

Note that U is not a regular open set (why?) but it satises the hypotheses of
the previous corollary (why?).
We have
div f (x; y; z) =

@
@
@
(0) +
(yz) +
(x) = 0 + 1z + 0;
@x
@y
@z

and so by the divergence theorem


Z
Z
2
f
dH =
z dxdydz:
@U

Using cylindrical coordinates x = r cos , y = r sin , z = z, we get r2 cos2 +


r2 sin2 < z 2 , that is, r2 < z 2 , r2 cos2 + r2 sin2 + z 2 < 2r sin , that is,
r2 + z 2 < 2r sin , and z > 0, hence, r2 < z 2 < 2r sin
r2 , which implies that
2
2
r < 2r sin
r , or equivalently, r < sin . In turn, sin should be positive,
and so 2 (0; ).
o
n
p
r2 :
W := (r; ; z) 2 R3 : 0 < < ; r < sin ; r < z < 2r sin

Hence, by changing variables


Z
Z
Z sin
z dxdydz =
U

sin

r2

2r sin

z dz

z2
2

p
z= 2r sin

r dr

r2

r dr
z=r

sin
2

r sin

dr d

r=sin

r3
r4
sin
3
4 r=0
1
1
sin4 d =
:
12
32

d =

1
sin4
3

1
sin4
4

Wednesday, May 1, 2013


Corollary 246 (Integration by Parts) Let U
RN be an open, bounded,
regular set and let f : U ! R and g : U ! R be such that f and g are bounded
and continuous in U and there exist the partial derivatives of f and g in R at
all x 2 U and they are continuous and bounded. Then for every i = 1; : : : ; N ,
Z
Z
Z
@g
@f
f (x)
(x) dx =
g (x)
(x) dx +
f (x) g (x) i (x) dHN 1 (x) :
@xi
@xi
U
U
@U
Proof. Fix i 2 f1; : : : ; N g. We apply the divergence theorem to the function
f : U ! RN dened by
fj (x) :=

f (x) g (x) if j = i;
0
if j 6= i:
97

Then
div f =
and so
Z
@f
@g
+g
f
@x
@x
i
i
U

N
X
@ (f g)
@g
@f
@fj
=
=f
+g
;
@xj
@xi
@xi
@xi
j=1

dx =

div f dx =

dHN

@U

fg

dHN

@U

Another important application is given by the areas formulas in R2 . Let


U R2 be an open, bounded set and assume that its boundary @U is the range
of a closed, simple, regular curve with parametric representation ' : [a; b] !
R2 . Then the hypotheses of Corollary 242 are satised (why?). Given t 2 [a; b],
the vector '0 (t) is tangent to the curve at the point ' (t) 2 @U . Hence, if
' (t) = (x (t) ; y (t)), the outer normal to @U at the point ' (t) is either
( y 0 (t) ; x0 (t))
q
2
2
(x0 (t)) + (y 0 (t))

or

( y 0 (t) ; x0 (t))
q
;
2
2
(x0 (t)) + (y 0 (t))

depending on the orientation of the curve. We say that has a positive orientation for @U if as we traverse the curve starting from t = a we nd U on our
left. In this case, the outward normal at the point ' (t) is is given by
(y 0 (t) ; x0 (t))
:
(' (t)) = q
2
2
(x0 (t)) + (y 0 (t))
Hence, if f; g : U ! R are bounded and continuous in U and there exist
the partial derivatives of f and g at all (x; y) 2 U and they are continuous
and bounded, then applying the divergence theorem to the function f (x; y) :=
(f (x; y) ; g (x; y))
Z
Z
@f
@g
(x; y) +
(x; y) dxdy =
(f; g)
dH1
@x
@y
U
@U
Z b
q
y 0 (t)
2
2
=
f (x (t) ; y (t)) q
(x0 (t)) + (y 0 (t)) dt
2
2
a
0
0
(x (t)) + (y (t))
Z b
q
x0 (t)
2
2
(x0 (t)) + (y 0 (t)) dt;
+
g (x (t) ; y (t)) q
2
2
a
0
0
(x (t)) + (y (t))
Z b
=
[f (x (t) ; y (t)) y 0 (t) g (x (t) ; y (t)) x0 (t)] dt
Za
Z
=:
f dy
g dx:
@U

@U

98

Taking g = 0 or f = 0 we get the Gauss-Green formulas


Z
Z
@f
(x; y) dxdy =
f dy:
@x
@U
Z
ZU
@g
(x; y) dxdy =
g dx:
@y
@U
U

(23)
(24)

Taking f (x; y) = x in (23), we get the second areas formula


Z
Z
Z b
x (t) y 0 (t) dt =
x dy;
meas U =
1 dxdy =
U

@U

while taking g (x; y) = y in (24) we get the second areas formula


Z
Z b
Z
y dx:
y (t) x0 (t) dt =
1 dxdy =
meas U =
U

@U

By adding these two identities, we get the third areas formula


Z
Z
1
1
meas U =
[ y (t) x0 (t) + x (t) y 0 (t)] dt =
x dy
2 @U
2 @U

y dx :

@U

Example 247 Lets nd the area of the region U enclosed by the curve of parametric representation
x ( ) = (1
y ( ) = (1

cos ) cos ;
cos ) sin ;

2 [0; 2 ] :

To see that it is simple, note that for 0 < < 2 , (1 cos ) > 0. The curve
starts from the origin and goes around clockwise only once. Note that
x0 ( ) =

sin + 2 sin cos = sin ( 1 + 2 cos ) ;

y ( ) = cos + sin2
0

cos2 =

Thus, x0 ( ) = 0 only for sin = 0, that is


for = 13 ; 53 . At those values,
y 0 (0) = y 0 (2 ) =
y0
y0

1
3
5
3

=
=

2 cos2 + cos + 1:
= 0; ; 2 ,and cos

2 + 1 + 1 = 0;

y0 ( ) =

= 21 , that is,

1 + 1;

1
1
+ cos
+ 1 = 1;
3
3
5
5
2 cos2
+ cos
+ 1 = 1:
3
3
2 cos2

Hence, the curve is not regular, but since these are only a nite number of bad
points, we can apply Corollary 242. Using the third area formula, we get
Z
1 2
(1 cos ) sin [sin cos
(1 cos ) sin ] d
meas U =
2 0
Z
1 2
+
(1 cos ) cos [sin sin + (1 cos ) cos ] d
2 0
Z
Z
1 2
1 2
3
2
2
=
(1 cos ) cos2 + sin2 d =
(1 cos ) d =
:
2 0
2 0
2
99

Friday, May 3, 2013

20

StokesTheorem

To state Stokes theorem, we need to introduce the notion of a manifold with


boundary.
Denition 248 Given a set E
Rk , a set F
E is relatively open in E if
k
there exists an open set V
R such that F = E \ V .
Denote Hk := y 2 Rk : xk

0 . Note that Hk is a closed upper half-space.

Denition 249 Given 1


k < N , a nonempty set M
RN is called a kdimensional dierential manifold with boundary if for every x0 2 M there exist
an open set U containing x0 and a di erentiable function ' : W ! RN , where
W Rk is either an open set or a relatively open set of Hk such that
(i) ' : W ! M \ U is a homeomorphism, that is, it is invertible and continuous together with its inverse ' 1 : M \ U ! W ,
(ii) J' (y) has maximum rank k for all y 2 W .
The function ' is called a local chart or a system of local coordinates or
a local parametrization around x0 . We say that M is of class C m , m 2 N,
(respectively, C 1 ) if all local charts are of class C m (respectively, C 1 ).
A point x0 2 M is an interior point of M if there is a local chart ' : W !
RN , where W Rk is open set, and is a boundary point of M , otherwise. The
set of boundary points is called the boundary of M and is denoted @M .
Theorem 250 Let M
RN be a k-dimensional di erentiable manifold with
boundary. Then @M is a (k 1)-dimensional di erentiable manifold.
Next we introduce the notion of orientation for a manifold with boundary.
Let M
RN be a k-dimensional manifold with boundary of class C 1 . Given
x0 2 M , the tangent space T (x0 ) is a vector space of dimension k. We say that
two bases B = fe1 ; : : : ; ek g and B1 = ff 1 ; : : : ; f k g of T (x0 ) are equivalent if the
matrix that transforms one basis into the other has positive determinant and we
write B B1 . An orientation at x0 is an equivalence class [B]. Note that there
are only two possible orientations at x0 . A (possibly discontinuous) orientation
O on M is a mapping that assigns to each x 2 M an orientation Ox . We say that
an orientation O is continuous if there exist continuous functions e1 : M ! RN ,
. . . , ek : M ! RN such that for every x 2 M , Bx = fe1 (x) ; : : : ; ek (x)g is a
basis for T (x) and Ox = [Bx ]. We say that M is orientable if it admits a
continuous orientation.
Remark 251 If M
RN 1 is an (N 1)-dimensional manifold of class C 1 ,
then for each x 2 M , the tangent space T (x) is a vector space of dimension
100

N 1. In turn, there are only two unit normals at the point x 2 M . It can be
shown that M is orientable if and only if for every x 2 M we can choose a unit
normal vector n (x) to M at x in such a way that the function n : M ! RN is
continuous.
Theorem 252 Let M RN be a k-dimensional orientable manifold with boundary of class C 1 . Then a continuous orientation on M induces a continuous
orientation on @M .
The Klein bottle or the Mbius strip are a typical examples of manifolds that
not orientable. Indeed, the normal as a function of (x; y; z) is discontinuous (but
it is continuous if you look at it as a function of (u; v)). If you think of the normal
as a person walking on the manifold, after going around once, that person would
nd himself/herself upside down, which means that there is a discontinuity.
Example 253 (The Mbius strip) Let ' : [0; 2 ]
' (u; v) :=

v sin

u
sin u; 2
2

It can be shown that M = ' ([0; 2 ]


have
@'
(u; v) =
@u
@'
(u; v) =
@v

v sin

[ 1; 1] ! R3 be given by

u
u
cos u; v cos
:
2
2

[ 1; 1]) is a 2-dimensional manifold. We

1
1
1
1
(cos u) v sin u 2
v sin u cos u; (sin u) v sin u
2
2
2
2
1
1
1
sin u sin u; cos u sin u; cos u :
2
2
2

1
1
v cos u cos u;
2
2

1
1
v sin u
2
2

Hence, to nd the normal, we need


@'
(u; v)
@u

e1
e2
1
1
1
1
1
v
sin
u
cos
u
(sin
u)
v
sin
u
2
(cos u) v sin 12 u 2
2
2
2
2 v cos u cos 2 u
1
1
sin u sin 2 u
cos u sin 2 u

=
=

@'
(u; v)
@v

1
v sin u
2

2 sin u cos

u
2

1
1
v cos u; v sin u
2
2

2 cos u cos

Hence,
@'
@'
(0; 0)
(0; 0) = (0; 2; 0) ;
@u
@v
@'
@'
(2 ; 0)
(2 ; 0) = (0; 2; 0) :
@u
@v

101

u 1
+ v sin u;
2 2

e3
1
1
v
2 sin 2 u
1
cos 2 u

1
v sin u
2

2 sin

u
2

Let U R3 be an open set and let f : U ! R3 be a dierentiable function.


The curl of f is the function curl f : U ! R3 dened by
@f3
@y
0

curl f (x; y; z) :=

= det @

@f2 @f1
;
@z @z
e1

e2

e3

@
@x

@
@y

@
@z

f1

f2

f3

@f3 @f2
;
@x @x
1
A=r

@f1
@y
f:

Theorem 254 (StokesTheorem) Let U


R3 be an open set and let f :
3
1
U ! R be a function of class C . Let M U be a closed bounded 2-dimensional
orientable manifold with boundary of class C 2 . Then
Z
Z
2
curl f n dH =
f t dH1 ;
M

@M

where n : M ! R is a continuous unit normal to M and t : @M ! R3 is a


continuous unit tangent to @M (with the orientation induced by the orientation
of M ).
Theorem 255 Given a k-dimensional manifold M
RN with boundary of
1
N
k
class C , let ' : W ! R , where W
R is an open set. Consider a bounded
open set V with V
W and let M1 := ' V . Then M1 is a k-dimensional
orientable manifold with boundary of class C 1 . Moreover, @M1 = ' (@V ).
The previous theorem shows that every manifold is locally orientable. It will
be useful in the exercises.
Example 256 Given the surface M of parametric representation
v; u; u2 + v 2 ;

' (u; v) := u

(u; v) 2 V;

where V := (u; v) 2 R2 : u2 + v 2 < 1 , lets calculate


Z
curl f
dH2 ;
M

where
f (x; y; z) := x2 z; y; yz ;

(x; y; z) 2 R3 :

Note that ' is dened in R2 so we can take W = R2 . To see that ' is one-toone, we nd the inverse, we have x = u v, y = u, z = u2 + v 2 and so, u = z,
v = u x = z x. Thus,
'

(x; y; z) = (z; z

x) ;

which is continuous. Hence, ' is a homeomorphism. Moreover,


0
1
1
1
0 A:
J' (u; v) = @ 1
2u 2v
102

1
1
has determinant di erent from zero, and so J' (u; v)
1 0
has always maximal rank two. In view of the previous theorem, M is orientable.
Thus, we are in a position to apply Stokes theorem. The boundary of V is the
unit circle. Thus a parametric representation of is
: [0; 2 ] ! R2 , where
(t) := (cos t; sin t). In turn, a parametric representation for @M is given by
:= '
: [0; 2 ] ! R3 , where

The submatrix

sin t; cos t; cos2 t + sin2 t :

(t) = cos t
In turn,
0

(t) = ( sin t

cos t;

sin t; 0) :

Hence, by Stokes theorem


Z
Z
2
curl f
dH =
f t dH1
M

@M
2

h
(cos t

sin t) 1 ( sin t

cos2 t

sin2 t (cos t

i
cos t) + (cos t) ( sin t) + (cos t) (1) 0 dt
sin t) + cos t sin t dt

2 sin2 t cos t

1 + 2 cos2 t sin t + cos t sin t dt

0
t=2

2
1
2
sin3 t + cos t + cos3 t + sin2 t
= 0:
3
3
2
t=0
R
Exercise 257 In the previous example calculate M curl f dH2 directly, without using Stokess theorem.
=

sin t

103