Multivariable Calculus Notes ©: David A. Santos

Multivariable Calculus Notes ©
David A. Santos
Philadelphia, PA October 22, 2003

ii
Contents
1 Brief Review of Linear Algebra 1

1.1 Rn . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Mm×n(R) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.3 Determinants . . . . . . . . . . . . . . . . . . . . . . . . . . 13
1.4 Linear Transformations . . . . . . . . . . . . . . . . . . . . . 16
1.5 Inner Products . . . . . . . . . . . . . . . . . . . . . . . . . 22
1.6 Cross product in R3 . . . . . . . . . . . . . . . . . . . . . . 27
1.7 Lines and Planes in R3 . . . . . . . . . . . . . . . . . . . . . 31
1.8 Topology of Rn . . . . . . . . . . . . . . . . . . . . . . . . . 37
1.9 Quadratic Forms . . . . . . . . . . . . . . . . . . . . . . . . 39
1.10 Quadratic Surfaces . . . . . . . . . . . . . . . . . . . . . . . 41
1.11 Canonical Surfaces in R3 . . . . . . . . . . . . . . . . . . . 44
1.12 Parametric Curves and Surfaces . . . . . . . . . . . . . . . 49
1.13 Frenet-Serret Formulæ . . . . . . . . . . . . . . . . . . . . . 54
1.14 Limits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
2 Differentiation 59
2.1 Local Study of Functions . . . . . . . . . . . . . . . . . . . . 59
2.2 Definition of the Derivative . . . . . . . . . . . . . . . . . . . 67
2.3 The Jacobi Matrix . . . . . . . . . . . . . . . . . . . . . . . 73
2.4 Gradients and Directional Derivatives . . . . . . . . . . . . 81
2.5 Extrema . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
2.6 Lagrange Multipliers . . . . . . . . . . . . . . . . . . . . . . 93
2.7 Arithmetic Mean-Geometric Mean Inequality . . . . . . . . . 97
iii
iv CONTENTS
3 Integration 105
3.1 Differential Forms . . . . . . . . . . . . . . . . . . . . . . . . 105
3.2 Integrating in ∧0(Rn) . . . . . . . . . . . . . . . . . . . . . 111
3.3 Integrating in ∧1(Rn) . . . . . . . . . . . . . . . . . . . . . 112
3.4 Closed and Exact Forms . . . . . . . . . . . . . . . . . . . . 122
3.5 Integrating in ∧2(R2) . . . . . . . . . . . . . . . . . . . . . . 127
3.6 Change of Variables in ∧2(R2) . . . . . . . . . . . . . . . . 141
3.7 Change to Polar Co-ördinates . . . . . . . . . . . . . . . . . 149
3.8 Integrating in ∧3(R3) . . . . . . . . . . . . . . . . . . . . . . 157
3.9 Change of Variables in ∧3(R3) . . . . . . . . . . . . . . . . 160
3.10 Integration in ∧2(R3) . . . . . . . . . . . . . . . . . . . . . . 165
3.11 Green’s, Stokes’, and Gauß’ Theorems . . . . . . . . . . . 171

Chapter 1
Brief Review of Linear Algebra
1.1 Rn

a1
a=
a2
Figure 1.1: A point in R2 . Figure 1.2: A bi-point in R2 .
1 Definition Rn is the set of real n-tuples

   

 x 1 

 
n
 x 2

R = x:x=  , xk ∈ R .
. . .

 

 
xn
These n-tuples are called points.
1
2 Chapter 1
Thus R2 is the collection of points on the plane (see figure 1.1) and R3 is
the collection of points in three-dimensional space.
Points on their own are very boring entities. They are devoid of arith-
metical properties (you cannot “add” two points), they are simply a list of
real numbers.
Suppose now that we are given two points x and y in Rn. Starting
from x we move on a straight line towards y, and we denote this ordered
movement by the notation [x, y] (read the “bi-point x, y”). This movement
involves a displacement of each of the n co-ördinates, the k-th co-ördinate
being displaced yk −xk units. If we let ak = yk −xk record the displacement
of the k-th co-ördinate, then we say that [x, y] is a representative of the
vector  
a1
 a2 
−
→  
a =  ..  .
 . 
an
Notice that there are infinitely many bi-points representing the same vector.
Thus in figure 1.2 we see a representative bi-point of a vector in R2. The
same vector would represent any parallel displacement of this bi-point.
2 Example Consider the points

1 3 3 5
x1 = , y1 = , x2 = , y2 = .
2 −4 5 −1
Though the bi-points [x1 , y1 ] and [x2 , y2 ] are in different locations on the
plane, they represent the same vector

−
→ 3−1 5−3
a = = .
−4 − 2 −1 − 5
These two bi-points are parallel and have the same length, moreover, they
are pointing in the same direction.
3 Definition A vector  
a1
→  a2 
−
 
a =  ..  ∈ Rn
 . 
an
Rn 3
is an equivalence class denoting all those bi-points whose displacement in

the k-th co-ördinate is ak.
! We write points of R n
using bold-face letters and parentheses, as in
 
x1
 x2 
x =  . .
 
 .. 
xn
We write vectors in Rn using bold-face letters with arrows on top and square
brackets, as in  
x1
 x2 
−
→  
x =  . .
 .. 
xn
−
→ −
→ − −
→ →
4 Definition If a and b are two vectors in Rn their vector sum a + b is
defined by the co-ördinatewise addition
 
a1 + b1
→ −
− →   a2 + b2 

a + b = .. . (1.1)
 . 
an + bn
5 Definition A real number α ∈ R will be called a scalar. If α ∈ R and

−
→
a ∈ Rn we define scalar multiplication of a vector and a scalar by the
co-ördinatewise multiplication
 
αa1
→  αa2 
−
 
α a =  ..  . (1.2)
 . 
αan
6 Theorem The operations of vector addition 1.1 and scalar multiplication

1.2 make Rn a vector space. That is, these operations satisfy
→ −
− → − →
∀( a , b , c ) ∈ (Rn)3, ∀(α, β) ∈ R2,
4 Chapter 1
➊ Closure under vector addition:

− −
→ →
a + b ∈ Rn (1.3)
➋ Closure under scalar multiplication:

−
→
α a ∈ Rn (1.4)
➌ Commutativity of addition:
− −
→ → − → − →
a + b = b + a (1.5)
➍ Associativity:
→ −
− → → −
− → −
→ − →
(a + b)+ c = a +(b + c) (1.6)
➎ Existence of additive identity:

−
→ → −
− → − → −→ − →
∃ 0 ∈ Rn : a + 0 = 0 + a = a (1.7)
➏ Existence of additive inverses:

−
→ −
→ −
→ → −
− → − →
∃ − a ∈ Rn : a + (− a ) = − a + a = 0 (1.8)
➐ Distributive Law:
→ −
− → −
→ −
→
α( a + b ) = α a + α b (1.9)
➑ Distributive Law:
−
→ −
→ −
→
(α + β) a = α a + β a (1.10)
➒
→ −
− →
1a = a (1.11)
➓
−
→ −
→
(αβ) a = α(β a ) (1.12)
Rn 5
−
→
7 Definition A set of vectors a k ∈ Rn, 1 ≤ k ≤ l is said to be linearly
independent if
l
X −
→ −
→
αk a k = 0 =⇒ α1 = α2 = · · · = αl = 0.
k=1
8 Definition An ordered set

→ −
− → −
→
A = { v 1, v 2, . . . , v n}
of n linearly independent vectors in Rn is called an ordered basis for Rn.

This means that every vector in Rn can be uniquely written as a linear
−
→ −
→
combination of the v k, that is, if u ∈ Rn then there exist unique scalars
−
→
αk (called the co-ördinates of v under A ) such that
 
α1
X n  α2 
−
→ −
→
u = αk v k :=  ..  .
 
k=1
 . 
αn A
→ −
− → − →
9 Example The family A = { i , j , k } with
     
− 1 − 0 0
→ → −
→
i = 0 , j = 1 , k = 0
0 0 1
forms an ordered basis for R3 (these is called the natural basis for R3). Any
−
→
vector u can be written uniquely as a linear combination of these vectors,
for example  
2 − −
→ → −
→
−3 = 2 i − 3 j + 4 k .
4 A
−
→ −→ − →
The family B = { i , k , j } forms a different ordered basis for R3. In this
case  
2 − −
→ → −
→
−3 = 2 i + 4 j − 3 k .
4 B
6 Chapter 1
! In most cases we will be using the standard ordered basis

→ −
− → −
→
A = { e 1, e 2, . . . , e n, }
with  
0
 .. 
.
−
→  
ek=
1

 .. 
.
0
(a 1 in the k slot and 0’s everywhere else). In such cases we will write


α1
n
X  α2 
−
→
αk e k =  . 
 
 .. 
k=1
αn
(without the A subscript) rather than



α1
n
X  α2 
−
→
αk e k =  .  .
 
 .. 
k=1
αn A
→ −
− → − →
10 Example Prove that the family C = { b 1, b 2, b 3} with
     
1 1 1
−
→ −
→ −
→
b 1 = 0 , b 2 = 1 , b 3 = 1
0 0 1
forms an ordered basis for R3 and find the co-ördinates of

 
2
−3
4
under C .
Mm×n(R) 7
Solution: We have
 
0 α1 + α2 + α3 = 0
−
→ −
→ −
→
α1 b 1 + α2 b 2 + α3 b 3 = 0 ⇐⇒ α2 + α3 = 0
0 α3 = 0.
Solving this triangular system gives α1 = α2 = α3 = 0, whence C is a
linearly independent family of 3 vectors and hence an ordered basis for
R3. It is easy to verify that
         
2 1 1 1 5
−
→ −
→ −
→
−3 = 5 0 − 7 1 + 4 1 = 5 b 1 − 7 b 2 + 4 b 3 = −7 .
4 0 0 1 4 C
1.2 Mm×n(R)
11 Definition An m × n (m by n) matrix A with m rows and n columns is a
rectangular array of the form
 
a11 a12 · · · a1n
 a21 a22 · · · a2n 
A =  .. ..  ,
 
..
 . . ··· . 
am1 am2 · · · amn
where ∀(i, j) ∈ {1, 2, . . . , m} × {1, 2, . . . , n}, aij ∈ R.
! As a shortcut, we often use the notation A = [a ] to denote the matrix

ij
A with entries aij. Notice that when we refer to the matrix we put square
brackets (as in “[aij]”), and when we refer to a specific entry we do not use
the surrounding parentheses (as in “aij”).
12 Example
0 −1 1
A=
1 2 3
is a 2 × 3 matrix and  
−2 1
B =  1 2
0 3
is a 3 × 2 matrix
8 Chapter 1
13 Definition We denote by Mm×n(R) the set of all m × n matrices with

real number entries. If m = n we use the abbreviated notation Mn(R) =
Mn×n(R). Mn(R) is thus the set of all square matrices of size n with real
entries.
14 Definition The n × n zero matrix 0n ∈ Mn(R) is the matrix with 0’s

everywhere,  
0 0 0 ··· 0
0 0 0 · · · 0
 
0n = 0 0 0 · · · 0 .

 .. .. .. .. 
. . . · · · .
0 0 0 ··· 0
15 Definition The n × n identity matrix In ∈ Mn(R) is the matrix with 1’s on

the main diagonal and 0’s everywhere else,
 
1 0 0 ··· 0
0 1 0 · · · 0
 
In = 0 0 1 · · · 0 .
 
 .. .. .. .
 . . . · · · .. 
0 0 0 ··· 1
16 Definition The main diagonal of a matrix matrix A = [aij] ∈ Mm×n(R) is

the set {aii|i ≤ min(m, n)}. A matrix is diagonal if every entry off its main
diagonal is 0.
17 Definition A ∈ Mm×n(R) is said to be upper triangular if
(∀(i, j) ∈ {1, 2, · · · , n}2), (i > j, aij = 0),
that is, every element below the main diagonal is 0. Similarly, A is lower
triangular if
(∀(i, j) ∈ {1, 2, · · · , n}2), (i < j, aij = 0),
that is, every element above the main diagonal is 0.
Mm×n(R) 9
18 Example The matrix A ∈ M3×4(R) shewn is upper triangular and B ∈

M4(R) is lower triangular.
 
  1 0 0 0
1 a b c 1 a 0 0
A = 0 2 3 0 B = 
0

2 3 0
0 0 0 1
1 1 t 1
19 Definition Let A = [aij] ∈ Mm×n(R), B = [bij] ∈ Mm×n(R) and α ∈ R.

The matrix A + αB is the matrix C = [cij] ∈ Mm×n(R) with entries cij =
aij + αbij.
20 Example Let
   
a −2a c 1 2a c
M= 0 −a b  , N =  a b − a −b .
a+b 0 −1 a−b 0 −1
Then
   
a+1 0 2c 2a −4a 2c
M+N= a b − 2a 0  , 2M =  0 −2a 2b  .
2a 0 −2 2a + 2b 0 −2
21 Definition Let A = [aij] ∈ Mm×n(R) and B = [bij] ∈ Mn×p(F). Then

the matrix product AB is defined as the matrix C = [cij] with entries cij =
Xn
ailblj.
l=1
! In order to obtain the ij-th entry of the matrix AB we multiply elemen-

twise the i-th row of A by the j-th column of B. Observe that AB is a m × p
matrix.
! Observe that we use juxtaposition rather than a special symbol to de-

note matrix multiplication. This will simplify notation.
10 Chapter 1
22 Example Let
 
1 2a
a −2a c
M= , N = a − b 0 
0 −a b
1 2
be matrices over R.Then
−2a − 2a2 c + 2ab
 
2
a
a − 2a(a − b) + c 2a + 2c
MN = , NM = a(a − b) −2a(a − b) (a − b)c .
−a(a − b) + b 2b
a −4a c + 2b
! Matrix multiplication is not necessarily commutative.

23 Example
Solution: We have
1 1 1 2 1 1
  
 3 −3 −3
 
3 3 3 0 0 0
1 1 1  1 2
  1 
− −  = 0 0 0 ,
  
3 3 3  3 3
  3
0 0 0
 
1 1 1  1 1 2 
− −
3 3 3 3 3 3
over R.
! Observe then that the product of two non-zero matrices may be the
zero matrix.
24 Theorem If (A, B, C) ∈ Mm×n(F) × Mn×r(F) × Mr×s(F) we have

(AB)C = A(BC),
i.e., matrix multiplication is associative.
Proof To shew this we only need to consider the ij-th entry of each side.
Both are equal to
Xn Xr
aikbkk0 ck0 j.
k=1 k0 =1
Mm×n(R) 11
! Even though matrix multiplication is not necessarily commutative, it is

associative.
! By virtue of associativity, a square matrix commutes with its powers,

that is, if A ∈ Mn(R), and (r, s) ∈ N2, then (Ar)(As) = (As)(Ar) = Ar+s.
25 Definition Let A = [aij] ∈ Mn(R). Then the trace of A, denoted by tr (A)

is the sum of the diagonal elements of A, that is
n
X
tr (A) = akk.
k=1
26 Theorem Let A = [aij] ∈ Mn(R), B = [bij] ∈ Mn(R), α ∈ R. Then
tr (αA) = αtr (A) , (1.13)
tr (A + B) = tr (A) + tr (B) , (1.14)
tr (AB) = tr (BA) . (1.15)
Proof The first assertion is trivial. To prove the second, observe that AB =
Xn n
X
[ aikbkj] and BA = [ bikakj]. Then
k=1 k=1
n X
X n n X
X n
tr (AB) = aikbki = bkiaik = tr (BA) ,
i=1 k=1 k=1 i=1
whence the theorem follows. ❑
27 Example Write  
1 2 3
A = 2 3 1 ∈ M3(R)
3 1 2
as the sum of two 3 × 3 matrices E1, E2, with tr (E2) = 10.
12 Chapter 1
Solution: There are infinitely many solutions. Here is one:

     
1 2 3 −9 2 3 10 0 0
A = 2 3 1 =  2 3 1 +  0 0 0 .
3 1 2 3 1 2 0 0 0
28 Example Given a square matrix A ∈ M4(R) such that tr A2 = −4, and

(A − I4)2 = 3I4,
find tr (A).
Solution:
tr (A − I4)2 = tr A2− 2A + I4

= tr A2 − 2tr (A) + tr (I4)

= −4 − 2tr (A) + 4
= −2tr (A) ,
and tr (3I4) = 12. Hence −2tr (A) = 12 or tr (A) = −6.
29 Definition The transpose of a matrix of a matrix A = [aij] ∈ Mm×n(R) is

the matrix AT = B = [bij] ∈ Mn×m(F), where bij = aji.
30 Example We have
   
a −2a c a 0 a+b
M= 0 −a b  , MT = −2a −a 0 .
a+b 0 −1 c b −1
31 Theorem Let
A = [aij] ∈ Mm×n(R), B = [bij] ∈ Mm×n(R), C = [cij] ∈ Mn×r(R), α ∈ R.
Then
ATT = A, (1.16)
(A + αB)T = AT + αBT , (1.17)
(AC)T = CT AT . (1.18)
Determinants 13
Proof The first two assertions are obvious. To prove the third put AT =
(αij), αij = aji, CT = (γij), γij = cji, AC = (uij) and CT AT = (vij). Then
n
X n
X n
X
uij = aikckj = αkiγjk = γjkαki = vji,
k=1 k=1 k=1
32 Definition A matrix A ∈ Mn(R) is symmetric if AT = A. A matrix B ∈

Mn(R) is skew-symmetric if BT = −B.
33 Example If    
1 2 3 0 −2 3
A = 2 4 5 , B =  2 0 5
3 5 6 −3 −5 0
then A is symmetric and B is skew-symmetric.
34 Theorem Any matrix A ∈ Mn(R) can be written as the sum of a sym-

metric and a skew-symmetric matrix.
Proof Observe that

(A + AT )T = AT + ATT = AT + A,
and so A + AT is symmetric. Also,
(A − AT )T = AT − ATT = −(A − AT ),
and so A − AT is skew-symmetric. We only need to write A as
1 1
A = ( )(A + AT ) + ( )(A − AT )
2 2
to prove the assertion. ❑
1.3 Determinants
We now define the notion of determinant of a matrix A ∈ Mn(R). We will
use an inductive definition, so that we can effect calculations of determi-
nants quickly.
14 Chapter 1
35 Definition The determinant of a square matrix A ∈ Mn(R), denoted by

det A is defined inductively as follows.
1. If n = 1, A = [a], then det A = a.

a c
2. If n = 2, A = , then det A = ad − bc.
b d
3. If n ≥ 3, A = [aij], let Aij ∈ Mn−1(R) denote the matrix obtained by
deleting the i-th row and the j-th column from A. Then
n
X
det A = aij(−1)i+j det Aij,
j=1
(the development along the i-th row) or, alternatively,

n
X
det A = aij(−1)i+j det Aij.
i=1
(the development along the j-th column).
! We have assumed that no matter which row or column we choose,

we always obtain the same determinant. This seems like an arbitrary as-
sumption, but in linear algebra courses we do see that this is indeed the
case. The result is independent of our choice. It is therefore advantageous
to choose that row, or column, with a maximal number of 0’s. It also follows
that det A = det AT , and that the determinant of a triangular matrix is the
product of its diagonal elements.
36 Example Find  
1 2 3
det 4 5 6
7 8 9
by expanding along the first row.
Solution: We have

1+15 6 1+2 4 6 1+3 4 5
det A = 1(−1) det + 2(−1) det + 3(−1) det
8 9 7 9 7 8
= 1(45 − 48) − 2(36 − 42) + 3(32 − 35) = 0.
Determinants 15
37 Example Find  
1 0 −1 1
 2 0 0 1 
det  
666 −3 −1 1000000
1 0 0 1
Solution: Since the second column has three 0’s, it is advantageous to

expand along it, and thus we are reduced to calculate
 
1 −1 1
−3(−1)3+2 det 2 0 1
1 0 1
Expanding this last determinant along the second column, the original de-
terminant is thus

3+2 1+2 2 1
−3(−1) (−1)(−1) det = −3(−1)(−1)(−1)(1) = 3.
1 1
−
→ −
→
38 Definition Let A ∈ Mn(R). A vector v 6= 0 is an eigenvector and a
scalar λ is an eigenvalue if
−
→ −
→
Av =λv.
To find the eigenvalues of a matrix A it is necessary and sufficient to solve
the equation (called the characteristic equation)
det(λIn − A) = 0.
39 Example Since
 
λ−1 1 1
det  −1 λ − 3 −1  = λ3 − 3λ2 − 4λ + 12 = (λ − 2)(λ + 2)(λ − 3),
3 −1 λ + 1
the matrix  
1 −1 −1
1 3 1
−3 1 −1
has eigenvalues −2, 2, 3.
16 Chapter 1
40 Theorem The eigenvalues of a real symmetric matrix are all real.
→ −
− → −
→
41 Definition Let v 1, v 2, . . . , v k be k vectors in Rn. The k-parallelotope
−
→
spanned by the v i is the set
k
X → −
ti v i : ti ∈ [0; 1] .
i=1
If
→ −
− → −
→
A = [ v 1, v 2, . . . , v k]
is the n × k matrix having the k vectors as columns, then
√
det AT A
→ −
− → −
→
is the k-dimensional volume of the k-parallelotope spanned by the v 1, v 2, . . . , v k.
! We will se later that det A A ≥ 0 so the square root of it is defined.

T
42 Theorem The signed area of a triangle in R2 spanned by the vectors

→ −
− →
v 1, v 2 is
1 → −
− →
det[ v 1, v 2].
2
1.4 Linear Transformations

43 Definition Let V, W be two vector spaces over the field of real numbers.
A function
V → W
L: −
→ → ,
−
a 7→ L( a )
is called a linear transformation if it is
→ −
− → −
→ −
→
• Linear: L( a + b ) = L( a ) + L( b ),
−
→ −
→
• Homogeneous: L(α a ) = αL( a ), for α ∈ R.
Linear Transformations 17
! These two properties can be condensed by saying that a linear trans-

formation is a function L : V → W such that
−
→ −
→ −
→ −
→
L( a + α b ) = L( a ) + αL( b ).
44 Example The trace map
Mn(R) → R
tr (·) :
A 7→ tr (A)
is linear, since in view of Theorem 26,
tr (A + B) = tr (A) + tr (B) , tr (αA) = αtr (A) .
45 Example The transpose map
Mn(R) → Mn(R)
L:
A 7→ AT
is linear, since in view of Theorem 31,
L(A + αB) = (A + αB)T = AT + (αB)T = AT + αBT = L(A) + αL(B).
46 Example Let X ∈ Mn(R) be a fixed square matrix. Prove that the map
Mn(R) → Mn(R)
L:
A 7→ XAX
is linear.
Solution: Let A, B be matrices in Mn(R) and let α be a scalar. We have
L(A + αB) = X(A + αB)X

= XAX + X(αB)X
= XAX + αXBX
= L(A) + αL(B),
whence the claim follows.

18 Chapter 1
47 Example Prove that the map
R2 →  R3 
x + 2y
f: x
7→  2x 
y
−y
is linear.
Solution: Put
−
→ x1 −→ x
v1= , v2= 2
y1 y2
and let α ∈ R. Then

−
→ −
→ x1 x
L( v 1 + α v 2) = L +α 2
y
1 y
2
x1 + αx2
= L
 y1 + αy2 
(x1 + αx2) + 2(y1 + αy2)
=  2(x1 + αx2) 
 −(y
 1 + αy
 2) 
x1 + 2y1 x2 + 2y2
=  2x1  + α  2x2 
−y1 −y2
−
→ −
→
= L( v 1) + αL( v 2),
whence L is linear.
n
X
−
→ −
→ −
→
If { v i}i∈[1;n] is an ordered basis for Rn, a = αi v i, and L is linear,
i=1
then !
n
X n
X
−
→ −
→ −
→
L( a ) = L αi v i = αiL( v i),
i=1 i=1
−
→
meaning that the action of L on an arbitrary vector a ∈ Rn is completely
determined by the action L has on the given ordered basis of Rn. This in
turn gives the following.
48 Theorem Let
Rn → Rm
L: →
− −
→
a 7→ L( a )
be a linear transformation. Then there is a unique matrix AL ∈ Mm×n(R)
−
→
such that for all a ∈ Rn,
−
→ −
→
L( a ) = AL a .
49 Example Find the matrix representation of the linear map in example

47 if
➊ both R2 and R3 have as ordered bases their standard ordered bases.
➋ R2 has the ordered basis

1 1
,
0 1
and R3 has the standard basis as ordered basis.
Solution:
➊ We have    
1 2
1 0
L = 2 , L
  = 0 ,

0 1
0 −1
and hence the desired matrix is
 
1 2
2 0  .
0 −1
➋ We have    
1 3
1 1
L = 2 , L =  2 ,
0 1
0 −1
and hence the desired matrix is
 
1 3
2 2  .
0 −1
20 Chapter 1
50 Example Consider L : R3 → R3, with

   
x x−y−z
L y = x + y + z .
z z
➊ Prove that L is a linear transformation.
➋ Find the matrix corresponding to L under the standard basis.
➌ Find the matrix corresponding to L under the ordered basis

     
 1 1 1 
B= 0 , 1 , 0 ,
   
 
0 0 1
for both the domain and the image of L.
Solution:
   
x a
−
→ −
→
➊ Let α ∈ R. Put u 1 = y , u 2 = b . Then
  
z c
 
x + αa
−
→ −
→
L( u + α u 2) = L y + αb
 z + αc 
(x + αa) − (y + αb) − (z + αc)
= (x + αa) + (y + αb) + (z + αc)
  z +αc 
x−y−z a−b−c
= x + y + z + α a + b + c
 z  c
x a
= L   y  + αL  b
z c
−
→ −
→
= L( u ) + αL( u 2)
proving that L is a linear transformation.

           
1 1 0 −1 0 −1
➋ We have L 0 = 1, L 1 =  1 , and L 0 =  1 , whence
0 0 0 0 1 1
the desired matrix is  
1 −1 −1
1 1 1 .
0 0 1
➌ We have
           
1 1 1 1 1 0
L 0 = 1 = 0 0 + 1 1 + 0 0 = 1 ,
          
0 0 0 0 1 0 B
           
1 0 1 1 1 −2
L 1 = 2 = −2 0 + 2 1 + 0 0 = 2  ,
          
0 0 0 0 1 0 B
and            
1 0 1 1 1 −3
L 0 = 2 = −3 0 + 2 1 + 1 0 = 2  ,
          
1 1 0 0 1 1 B
whence the desired matrix is
 
0 −2 −3
1 2 2 .
0 0 1 B
We will also use the following result.
51 Theorem Let
Rn → Rm Rm → Rl
L1 : →
− → , L2 : −
− → −
→
a 7→ L1( a ) a 7→ L2( a )
be linear transformations with matrix representations AL1 ∈ Mm×n(R) and
BL2 ∈ Ml×m(R) respectively. Then the composition map
Rn → Rl
L2 ◦ L1 : →
− −
→
a 7→ (L2 ◦ L1)( a )
has as matrix representation the product of matrices BL2 AL1 ∈ Ml×n(R).
22 Chapter 1
1.5 Inner Products

52 Definition Let V be a vector space over R. An inner product • is a
function
V ×V → R
• : → −
− → →•−
− →
( x , y ) 7→ x y
satisfying
→ −
− → − → − →− → − →− →
➊ ( x + y )• z = x • z + y • z
→ −
− → − → − → →−
− →
➋ (α x )• y = x •(α y ) = α( x • y ), α ∈ R.
→−
− → − →−→
➌ x•y = y•x
→−
− →
➍ x•x ≥ 0
→−
− → → −
− →
➎ x•x =0⇔ x = 0
53
−Definition Given a vector space V over R with inner product •, the norm
→
−
→
a of a vector a is

− p− →•−
→
a =
→
a a.
54 Definition Given a vector space V over R with inner product • and norm
→ −
− → −
→ −
→
||·||, the distance d a , b between a and b is
−
−
→ → → −
− →

d a , b = a − b .
−
→ −
→
55 Example Given a and b in Rn, the usual inner product is the dot prod-
uct defined by
n
→•−
− → X
a b = akbk. (1.19)
k=1
56 Example Let    
1 1
→   −
− →
a = 2 , b = 4 
3 −3
Inner Products 23
be vectors in R3. Their dot product is

− •−
→ →
a b = (1)(1) + (2)(4) + (3)(−3) = 0,
their norms are

− √ − p
→ √
→
p
a = (1)2 + (2)2 + (3)2 = 14, b = (1)2 + (4)2 + (−3)2 = 26,
and their distance is

 
0
−
−
→ →

−

→ −→
p √
d a , b = a − b = −2 = (0)2 + (−2)2 + (6)2 = 2 10.
6
! We took for granted the fact that the dot product in R n

does define an
inner product. This does require (a very easy) proof.
57 Example Let A, B in Mn(R). Prove that the operation
A•B = tr BT A

is an inner product. This is the standard inner product for matrices with
real number entries.
Solution: Observe that
tr BT (A1 + αA2)

(A1 + αA2)•B =
= tr BT A1+ BT (αA2)
= tr BT A1 + αtr BT A2
= A1•B + αA2•B,
and so (1) and (2) in definition 52 are verified. Also (3) follows from Theo-
rem 26. To check (4) and (5) we only need to observe that
n X
X n
A•A = a2ij.
i=1 j=1
24 Chapter 1
58 Example Given
1 1 −1 0
A= ,B =
0 −1 0 1
their standard inner product is
T !
−1 0 1 1 −1 −1
A•B = tr = tr = −2,
0 1 0 −1 0 −1
their standard norms are

√

1 1 p
||A|| = = (1)2 + (1)2 + (0)2 + (−1)2 = 3,
0 −1
and
√

−1 0 p
||B|| =
= (−1)2 + (0)2 + (0)2 + (1)2 = 2.
0 1
59 Theorem (Cauchy-Bunyakovsky-Schwarz Inequality) Let V be a vector

space over R having inner product • and corresponding norm ||·||. Then for
−
→ −
→
any two vectors x and y we have
→−
− → −
→− →
| x • y | ≤ x y .
Proof Since the norm of any vector is non-negative, we have

− − − → −
− −
→ →
x + t y ≥ 0 ⇐⇒
→ → →
( x + t y )•( x + t y ) ≥ 0
→•−
− → →−
− → →−
− →
⇐⇒ x x + 2t x • y + t2 y • y ≥ 0
− →−
− −
→ 2 → →2

⇐⇒ x + 2t x • y + t2 y ≥ 0.
This last expression is a quadratic polynomial in t which is always non-

negative. As such its discriminant must be non-positive, that is,
→−
− → −
→2 − →2 − −
→ → −
→− →
(2 x • y )2 − 4( x )( y ) ≤ 0 ⇐⇒ | x • y | ≤ x y ,
giving the theorem. ❑

Inner Products 25
60 Corollary (Triangle Inequality) Let V be a vector space over R having

−
→
inner product • and corresponding norm ||·||. Then for any two vectors a
−
→
and b we have
−→ − → −
→ −
→
a + b ≤ a + b .

Proof
→ −
− → → −
− → − → − →
|| a + b ||2 = ( a + b )•( a + b )
→•−
− → →−
− → − →− →
= a a +2a•b + b•b
−
→ → −
− → −
→
≤ || a ||2 + 2|| a |||| b || + || b ||2
−
→ −
→
= (|| a || + || b ||)2,
from where the desired result follows. ❑
−
→ −
→
61 Definition Let x and y be two non-zero vectors in a vector space over
→ −
−
the real numbers. Then the angle ( \
→
x , y ) between them is given by the
relation →•−
− →
→ −
− x y
cos ( \
→
x , y ) = − → .
−
→

x y
This expression agrees with the geometry in the case of the dot product
for R2 and R3.
→ −
− →
62 Example Let u , v be vectors in a vector space V over R with inner
product •. Prove the polarisation identity:
→ −
− → 1 − → − → → −
− →
u • v = || u + v ||2 − || u − v ||2 .
4
Solution: We have
→ −
− → → −
− → → −
− → − → − → → −
− → − → − →
|| u + v ||2 − || u − v ||2 = ( u + v )•( u + v ) − ( u − v )•( u − v )
−
→→ − →−
− → − →− → →−
− → →−
− → − →−
→
= u•u +2u•v + v•v −(u•u −2u•v + v•v)
→•−
− →
= 4u v,
giving the result.
→ −
− →
63 Example Let a , b be fixed vectors in R2. Prove that if
−
→ → −
− → − → − →
∀ v ∈ R2, v • a = v • b ,
→ −
− →
then a = b .
26 Chapter 1
−
→ → −
− → − →
Solution: We have ∀ v ∈ R2, v • ( a − b ) = 0. In particular, choosing
→ −
− → − →
v = a − b , we gather
→ −
− → − → − → → −
− →
( a − b )•( a − b ) = || a − b ||2 = 0.
−
→
But the norm of a vector is 0 if and only if the vector is the 0 vector.
→ −
− → − → → −
− →
Therefore a − b = 0 , i.e., a = b .
! The Cauchy-Bunyakovsky-Schwarz (CBS) Inequality applied to the

dot product in Rn gives
n n
!1/2 n
!1/2
X X X
xkyk ≤ x2k y2k . (1.20)

k=1 k=1 k=1
64 Example Assume that ak, bk, ck, k = 1, . . . , n, are positive real num-
bers. Shew that
n
!4 n
! n ! n !2
X X X X
akbkck ≤ a4k b4k c2k .
k=1 k=1 k=1 k=1
n
X
Solution: Using CBS on (akbk)ck once we obtain
k=1
n n
!1/2 n
!1/2
X X X
akbkck ≤ a2kb2k c2k .
k=1 k=1 k=1
n
!1/2
X
Using CBS again on a2kb2k we obtain
k=1
n n
!1/2 n
!1/2
X X X
akbkck ≤ a2kb2k c2k
k=1 k=1 k=1
n
!1/4 n
!1/4 n
!1/2
X X X
≤ a4k b4k c2k ,
k=1 k=1 k=1
which gives the required inequality.

Cross product in R3 27
1.6 Cross product in R3

We now define the standard cross product in R3 as a product satisfying
the following properties.
→ −
− → − →
65 Definition Let ( x , y , z , α) ∈ R3 × R3 × R3 × R. The cross product
× : R3 × R3 → R3 is a closed binary operation satisfying
→ −
− → → −
− →
➊ Anti-commutativity: x × y = −( y × x )
➋ Bilinearity:
→ −
− → − → − → −→ − → − → −
→ → −
− → → −
− → − → − →
( x + z )× y = x × y + z × y and x ×( z + y ) = x × z + x × y
−
→ → −
− → −
→ → −
− →
➌ Scalar homogeneity: (α x ) × y = x × (α y ) = α( x × y )
→ −
− → − →
➍ x × x = 0
➎ Right-hand Rule:
−
→ −
→ − → →− → −
− → → − → −
− →
i × j = k, j × k = i , k × i = j .
−
→ −
→
k k
−
→ −
→
j j
−
→ −
→
i i
Figure 1.3: Right-handed sys- Figure 1.4: Left-handed system.
tem.
To study points in space we must first agree on the orientation that we

will give our co-ördinate system. We will use, unless otherwise noted, a
right-handed orientation, as in figure 1.3.
28 Chapter 1
66 Example Find    
1 0
 0  × 1 .
−3 2
Solution: We have
−
→ −
→ −
→ −
→ −
→ −
→ −
→ − → → −
− → → −
− →
( i −3k)×( j +2k) = i × j +2 i × k −3k × j −6k × k
−
→ −
→ −
→ −
→
= k −2 j +3 i +60
−
→ −
→ → −
= 3 i −2 j + k.
Hence      
1 0 3
 0  × 1 = −2 .
−3 2 1
Operating as in example 66 we obtain
   
x1 y1
−
→   −
→  
67 Theorem Let x = x2 and y = y2 be vectors in R3. Then
x3 y3
→ −
− → −
→ −
→ −
→
x × y = (x2y3 − x3y2) i + (x3y1 − x1y3) j + (x1y2 − x2y1) k .
−
→ → −
− → −
→ → −
− →
68 Theorem x ⊥ ( x × y ) and y ⊥ ( x × y ).
Proof We will only check the first assertion, the second verification is anal-
ogous.
→• −
− → − → −
→ −
→ −
→ −
→
x ( x × y ) = (x1 i + x2 j + x3 k )•((x2y3 − x3y2) i
−
→ −
→
+(x3y1 − x1y3) j + (x1y2 − x2y1) k )
= x1x2y3 − x1x3y2 + x2x3y1 − x2x1y3 + x3x1y2 − x3x2y1
= 0,
completing the proof. ❑
69 Example Leta ∈R. Find a vector

  of unit length simultaneously perpen-
0 1
−
→   −
→  
dicular to v = −a and w = a .
a 0
Cross product in R3 29
−
→ −
→ → −
− →
v ×w v ×w
Solution: Either of −→ − → or − −→ − → will do. Now
|| v × w || || v × w ||
−
→ −
→ −
→ −
→ −
→ −
→
v ×w = (−a j + a k ) × ( i + a j )
→ −
− → → −
− → → −
− → → −
− →
= −a( j × i ) − a2( j × j ) + a( k × i ) + a2( k × j )
−
→ −
→ 2−
→
= ak + a
 2 j − a i
−a
=  a ,
a
→ −
− → p p
and || v × w || = a4 + a2 + a2 = 2a2 + a4. Hence we may take either
 2
−a
1
√  a 
2a2 + a4 a
or
 2
−a
1
−√  a .
2a2 + a4 a
! The cross product of vectors in R

3
is not associative, since
−
→ → −
− → −
→ − → −
→
i ×( i × j )= i × k =− j
but
→ −
− → −
→ − → −→ − →
( i × i )× j = 0 × j = 0.
We have, however, the following theorem.
70 Theorem
−
→ −
→ − → →−
− →−→ →−
− →−→
a ×(b × c) = (a•c)b −(a•b)c.
30 Chapter 1
Proof
−
→ −
→ −→ −
→ −
→ −
→ −
→
a ×(b × c) = (a1 i + a2 j + a3 k ) × ((b2c3 − b3c2) i +
−
→ −
→
+(b3c1 − b1c3) j + (b1c2 − b2c1) k )
−
→ −
→ −
→
= a1(b3c1 − b1c3) k − a1(b1c2 − b2c1) j − a2(b2c3 − b3c2) k
−
→ −
→ −
→
+a2(b1c2 − b2c1) i + a3(b2c3 − b3c2) j − a3(b3c1 − b1c3) i
−
→ −
→ −
→
= (a1c1 + a2c2 + a3c3)(b1 i + b2 j + b3 i )+
−
→ −
→ −
→
(−a1b1 − a2b2 − a3b3)(c1 i + c2 j + c3 i )
→−
− →− → →−
− →− →
= = (a•c)b −(a•b)c,
! Permuting the vectors in the above theorem and adding, we obtain

Jacobi’s Identity:
−
→ −
→ −
→ −
→ → −
− → −
→ → −
− → −
→
a ×(b × c)+ b ×(c × a)+ c ×(a × b) = 0.
→ −
−
71 Theorem Let ( \
→
x , y ) ∈ [0; π] be the convex angle between two vectors
−
→ −
→
x and y . Then
→ −
− → −
− → −
−
|| x × y || = || x |||| y || sin ( \
→ → →
x , y ).
Proof We have
→ −
− →
|| x × y ||2 = (x2y3 − x3y2)2 + (x3y1 − x1y3)2 + (x1y2 − x2y1)2
= x22y23 − 2x2y3x3y2 + x23y22 + x23y21 − 2x3y1x1y3+
+x21y23 + x21y22 − 2x1y2x2y1 + x22y21
= 2 2
(x1 + x2 + x23)(y21 + y22 + y23) − (x1y1 + x2y2 + x3y3)2
→ −
− → →−
− →
= || x ||2|| y ||2 − ( x • y )2
→ −
− → −
− → −
−
|| x ||2|| y ||2 − || x ||2|| y ||2 cos2 ( \
→ → →
= x, y)
→ −
− → −
−
|| x ||2|| y ||2 sin2 ( \
→ →
= x , y ),
The following corollaries are now obvious.

Lines and Planes in R3 31
→ −
− → → −
− → − →
72 Corollary Two non-zero vectors x , y satisfy x × y = 0 if and only if
they are parallel.
73 Corollary (Lagrange’s Identity)

→ −
− → →−
− →
|| x × y ||2 = ||x||2||y||2 − ( x • y )2.
−
→
74 Example Let x ∈ R3, ||x|| = 1. Find
→ −
− → → −
− → → −
− →
|| x × i ||2 + || x × j ||2 + || x × k ||2.
Solution: By Lagrange’s Identity,

→ − →2− →− →−
→ 2 − →2 → →
− − −
|| x × i || = x i − ( x • i )2 = 1 − ( x • i )2,

− →2− →− →−
→ 2 − →2 → →
−
→ − −
|| x × k || = x j − ( x • j )2 = 1 − ( x • j )2,

→ − →2− →− →−
→ →2 → →
− − − −
|| x × j ||2 = x k − ( x • k )2 = 1 − ( x • k )2,

→−
− → →−
− → →−
− → −
→2
and since ( x • i )2 + ( x • j )2 + ( x • k )2 = x = 1, the desired sum
equals 3 − 1 = 2.
1.7 Lines and Planes in R3

 
x
3 −
→ 3 −
→ −
→
75 Definition Let a ∈ R and v ∈ R \ { 0 }. Put r = y . The line
z
−
→
passing through a in the direction of v is the set
→ −
− → − → −
→
{ r : r = a + t v , t ∈ R}.
 
1
76 Example Find the equation of the line passing through 2 in the di-
  3
−2
rection of −1.

0
32 Chapter 1
Solution: The desired equation is

     
x 1 −2
y = 2 + t −1 .
z 3 0
   
1 −2
77 Example Find the equation of the line passing through 2 and −1.
3 0
Solution: The line follows the direction

   
1 − (−2) 3
2 − (−1) = 3 .
3−0 3
The desired equation is

     
x 1 3
y = 2 + t 3 .
z 3 3
? Why does      
x −2 3
y = −1 + t 3
z 0 3
represent the same line as
     
x 1 3
y = 2 + t 3 ?
z 3 3
! Given two lines in space, one of the following three situations might
arise: (i) the lines intersect at a point, (ii) the lines are parallel, (iii) the lines
are skew (one over the other, without intersecting).
78 Definition The set of points x ∈ Rn satisfying an equation of the form

−
→•−→
a x =c
−
→
( a ∈ Rn and c ∈ R fixed) is called a hyperplane in Rn.
In three-dimensions a hyperplane is simply a plane.

 To obtain the equation
n1
of a plane in R3 simply notice that if the vector n2 is normal (perpen-
 n 3
a1
dicular) to the plane passing through the point a2 then form any other
    a3
x x − a1
point y on the plane, the vector y − a2 will be perpendicular to the
  
z  z − a3
n1
vector n2 and so

n3    
n1 x − a1
n2 • y − a2 = 0
n3 z − a3
or equivalently,
n1(x − a1) + n2(y − a2) + n3(z − a3) = 0, (1.21)
gives the equation of a plane in R3.
 plane passing through the point (1, −1, 2)

79 Example The equationof the
−3
and normal to the vector 2  is

4
−3(x − 1) + 2(y + 1) + 4(z − 2) = 0.
! From n (x − a ) + n (y − a ) + n (z − a ) = 0 we obtain that the

1 1 2 2 3 3
equation of a plane is of the form
n1x + n2y + n3z = d,

34 Chapter 1
 
n1
where n2 is perpendicular to the plane, and d ∈ R.
n3
80 Example Find the equation of plane containing the point (1, 1, 1) and
perpendicular to the line x = 1 + t, y = −2t, z = 1 − t.
Solution: The vectorial form of the equation of the line is

   
1 1
−
→  
r = 0 + t −2 .
1 −1
   
1 1
Since the line follows the direction of −2 , this means that −2 is nor-
  
−1 −1
mal to the plane, and thus the equation of the desired plane is
(x − 1) − 2(y − 1) − (z − 1) = 0.
81 Example Find the equation of plane containing the point (1, −1, −1) and
containing the line x = 2y = 3z.
Solution: Observe that (0, 0, 0) (as 0 = 2(0) = 3(0)) is on the line, and
hence on the plane. Thus the vector
   
1−0 1
−1 − 0 = −1
−1 − 0 −1
lies on the plane. Now, if x = 2y = 3z = t, then x = t, y = t/2, z = t/3.

Hence, the vectorial form of the equation of the line is
     
0 1 1
−
→  
r = 0 + t 1/2 = t 1/2 .
0 1/3 1/3


1
This means that 1/2 also lies on the plane, and thus
1/3
     
1 1 1/6
−1 × 1/2 = −4/3
−1 1/3 3/2
is normal to the plane. The desired equation is thus

1 4 3
x − y + z = 0.
6 3 2
! Given three planes in space, they may (i) be parallel (which allows for
some of them to coincide), (ii) two may be parallel and the third intersect
each of the other two at a line, (iii) intersect at a line, (iv) intersect at a point.
82 Example Find the equation of the plane passing through the points
(a, 0, a), (−a, 1, 0), and (0, 1, 2a) in R3.
The vectors    
a − (−a) 2a
 0 − 1  = −1
a−0 a
and    
0 − (−a) a
 1−1 = 0 
2a − 0 2a
lie on the plane. A vector normal to the plane is
     
2a a −2a
−1 ∧  0  = −3a2 .
a 2a a
The equation of the plane is thus given by

   
−2a x−a
−3a2 • y − 0 = 0,
a z−a
36 Chapter 1
that is,
2ax + 3a2y − az = a2.
83 Example Find the equation of the line perpendicular to the plane ax +

a2y + a3z = 0, a 6= 0 and passing through the point (0, 0, 1).
 
a
Solution: A vector normal to the plane is a2. The line sought has the
a2
same direction as this vector, thus the equation of the line is
     
x 0 a
y = 0 + t a2 , t ∈ R.
z 1 a2
84 Example Find the equation of the plane perpendicular to the line ax =

by = cz, abc 6= 0 and passing through the point (1, 1, 1) in R3.
Solution: Put ax = by = cz = t, so x = t/a; y = t/b; z = t/c. The

parametric equation of the line is
   
x 1/a
y = t 1/b , t ∈ R.
z 1/c
 
1/a
Thus the vector 1/b is perpendicular to the plane. Therefore, the equa-
1/c
tion of the plane is      
1/a x−1 0
1/b • y − 1 = 0 ,
1/c z−1 0
or
x y z 1 1 1
+ + = + + .
a b c a b c
We may also write this as
bcx + cay + abz = ab + bc + ca.

Topology of Rn 37
85 Example (Putnam Exam 1980) Let S be the solid in three-dimensional

space consisting of all points (x, y, z) satisfying the following system of six
conditions:
x ≥ 0, y ≥ 0, z ≥ 0,
x + y + z ≤ 11,
2x + 4y + 3z ≤ 36,
2x + 3z ≤ 24.
Determine the number of vertices and the number of edges of S.
Solution: There are 7 vertices (V0 = (0, 0, 0), V1 = (11, 0, 0), V2 = (0, 9, 0), V3 =
(0, 0, 8), V4 = (0, 3, 8), V5 = (9, 0, 2), V6 = (4, 7, 0)) and 11 edges (V0V1,
V0V2, V0V3, V1V5, V1V6, V2V4, V3V4, V3V5, V4V5, and V4V6).
1.8 Topology of Rn
−
→ −
→
86 Definition Let a ∈ Rn and let ε > 0. An open ball centred at a of
radius ε is the set
−
→ → −
− →
Bε( a ) = {x ∈ Rn : d x , a < ε}.
87 Example An open ball in R is an open interval, an open ball in R2 is an

open disk and an open ball in R3 is an open sphere.
88 Definition A set S ⊆ Rn is said to be open if for every point belonging

to it we can surround the point by a sufficiently small open ball so that this
−
→
balls lies completely within the set. That is, ∀ a ∈ S ∃ε > 0 such that
−
→
Bε( a ) ⊆ S.
On the real line, an open ball is an interval of length 2r centred at a

−
→
point p, on the plane, an open ball is an open disk centred about p , in
3-dimensional space, an open ball is a sphere excluding its boundary and
−
→
centred at p .
−
→
89 Definition A set O ⊆ Rn is said to be open in Rn if ∀ x ∈ O ∃r > 0 such
−
→
that Br( x ) ⊆ O.
38 Chapter 1
That is, a set is open if for all its elements, there exist open balls centred
at the elements and totally contained in the set.
90 Example The open interval ] − 1; 1[ is open in R. The interval ] − 1; 1] is

not open, however, as no interval centred at 1 is totally contained in ]−1; 1].
91 Example The region ] − 1; 1[×]0; +∞[ is open in R2.
92 Example The ellipsoidal region {(x, y) ∈ R2 : x2 + 4y2 < 4} is open in

R2.
The reader will recognise that open boxes, open ellipsoids and their unions
and finite intersections are open sets in Rn.
93 Definition A set F ⊆ Rn is said to be closed in Rn if its complement

Rn \ F is open.
94 Example The closed interval [−1; 1] is closed in R, as its complement,

R \ [−1; 1] =] − ∞; −1[∪]1; +∞[ is open in R. The interval ] − 1; 1] is neither
open nor closed, however.
95 Example The region [−1; 1] × [0; +∞[×[0; 2] is closed in R3.
96 Example (Putnam Exam 1969) Let p(x, y) be a polynomial with real co-
efficients in the real variables x and y, defined over the entire plane R2.
What are the possibilities for the image (range) of p(x, y)?
Solution: Since polynomials are continuous functions and the image of a

connected set is connected for a continuous function, the image must be
an interval of some sort. If the image were a finite interval, then f(x, kx)
would be bounded for every constant k, and so the image would just be the
point f(0, 0). The possibilities are thus (i) a single point (take for example,
p(x, y) = 0), (ii) a semi-infinite interval with an endpoint (take for example
p(x, y) = x2 whose image is [0; +∞[), (iii) a semi-infinite interval with no
endpoint (take for example p(x, y) = (xy−1)2+x2 whose image is ]0; +∞[),
(iv) all real numbers (take for example p(x, y) = x).
Quadratic Forms 39
97 Example (Putnam Exam 1984) Let A be a solid a × b × c rectangular

brick in three dimensions, where a > 0, b > 0, c > 0. Let B be the set of all
points which are at distance at most 1 from some point of A (in particular,
A ⊂ B). Express the volume of B as a polynomial in a, b, c.
Solution: The set B can be decomposed into the following subsets:
➊ The set A itself, of volume abc.
➋ Two a × b × 1 bricks, two b × c × 1 bricks, and two c × a × 1 bricks,
➌ Four quarter-cylinders of length a and radius 1, four quarter-cylinders

of length b and radius 1, and four quarter-cylinders of length c and
radius 1,
➍ Eight eighth-of-spheres of radius 1.
Thus the required formula for the volume is
4π
abc + 2(ab + bc + ca) + π(a + b + c) + .
3
1.9 Quadratic Forms

−
→
98 Definition A matrix A ∈ Mn(R) is called positive definite if ∀ x ∈ Rn \
−
→
{ 0 },
→T −
− →
x A x > 0.
−
→ → −
− →
It is positive semi-definite if ∀ x ∈ Rn, x T A x ≥ 0. It is negative definite if
−
→ −
→
∀ x ∈ Rn \ { 0 },
→T −
− →
x A x < 0.
−
→ → −
− →
It is negative semi-definite if ∀ x ∈ Rn, x T A x ≤ 0. A matrix is indefinite if
it is neither positive nor negative definite (semi-definite).
99 Example The matrix  

1 0 0
A = 0 2 0
0 0 3
40 Chapter 1
is positive definite, since

  
1 0 0 x
x y z 0 2 0 y = x2 + 2y2 + 3z2 > 0,

  
0 0 3 z
for (x, y, z) 6= (0, 0, 0), being a sum of squares.
100 Example The matrix

 
1 0 0
A = 0 −2 0
0 0 3
is indefinite, since
     
1 0 0 1 1 0 0 0
1 0 0 0 2 0 0 = 1 > 0, 0 1 0 0 2 0 1 = −2 < 0.
0 0 3 0 0 0 3 0
101 Example The matrix

 
−1 0 0
A =  0 −2 0
0 0 0
is negative semi-definite, since
     
−1 0 0 0 −1 0 0 x
0 −2 0 y = −(x2+2y2) < 0.

0 0 1  0 −2 0 0 = 0, x y z
     
0 0 0 1 0 0 0 z
102 Definition The n principal minors of a matrix A = [aij] are
∆1 = a11,

a11 a12
∆2 = det ,
a21 a22
 
a11 a12 a13
∆3 = det a21 a22 a23 ,
a31 a32 a33
Quadratic Surfaces 41
..
.
 
a11 a12 a13 · · · a1k
a21 a22 a23 · · · a2k
∆k = det  .. ..  ,
 
.. .. ..
 . . . . . 
ak1 ak2 ak3 · · · akk
..
.
∆n = det A.
103 Theorem (Sylvester’s Criterion) A matrix A ∈ Mn(R) is positive definite

if and only if all its principal minors are positive. It is positive semi-definite
if all its principal minors are non-negative. It is negative definite if all of its
odd order minors are negative and all of its even order minors are positive.
It is negative semi-definite if all of its odd order minors are non-positive
and all of its even order minors are non-negative. It is indefinite if none of
the above cases occur.
1.10 Quadratic Surfaces

104 Definition A quadric (or quadratic) surface in R3 is a surface whose
cartesian equation is a polynomial of degree 2. Thus the general form of a
quadric surface is
Ax2 + 2Bxy + 2Cxz + Dy2 + 2Eyz + Fz2 + 2Gx + 2Hy + 2Iz + J = 0
where all the coefficients are real. This can be written in the matrix form
  
A B C G x
B D E H y
 
x y z 1   = 0, (1.22)
C E F I  z
G H I J 1
where we have identified a 1 × 1 matrix with a real number.
! The square matrix of coefficients is symmetric, and as such all its

eigenvalues are real by virtue of Theorem 40.
42 Chapter 1
105 Example Write the quadric surface
2x2 − 3y2 + 10xy + 5 + (z − 2)2 = 0
in matrix form.
Solution: First observe that
2x2 − 3y2 + 10xy + (z − 2)2 = 2x2 − 3y2 + 10xy + 5 + z2 − 4z + 9.
The desired form is

  
2 5 0 0 x
5 −3 0 0  y
 
x y z 1   = 0.
0 0 1 −2  z 
0 0 −2 9 1
In the case when there are no cross terms, things greatly simplify, and,
up to permutations of the letters, we have the following result.
106 Theorem If
x2 y2 z2
p + q + r = d,
a2 b2 c2
then this equation represents
• an ellipsoid: p = q = r = d = 1, a, b, c are the lengths of the semi
axes.
• a single-sheet hyperboloid: p = q = d = 1, r = −1.
• a double-sheet hyperboloid: r = d = 1, p = q = −1.
• Cone: p = q = 1, r = −1, d = 0.
If
x2 y2 z
p2
+ q 2
+ r 2 = d,
a b c
• an elliptic paraboloid: p = q = 1, r = −1, d = 0.
• a hyperbolic paraboloid: p = r = −1, q = 1, d = 0.

Quadratic Surfaces 43
• an elliptic cylinder: p = q = −1, r = d = 0.
• a hyperbolic cylinder: p = d = 1, q = −1, r = 0.
• a pair of planes: p = 1, q = −1, d = 0.
If
py2 + qx = d
• a parabolic cylinder: p, q > 0.
• a pair of parallel planes: d > 0, q = 0, p 6= 0.
• two coinciding planes: p 6= 0, q = d = 0.
Figure 1.5: Example 107.
107 Example Demonstrate that the surface in R3
S : 4x2 + y2 − 4 = 0
is a cylinder and draw its graph.

44 Chapter 1
Solution: The variable z is missing. Notice that on the plane, 4x2+y2−4 = 0

defines an ellipse. Thus the surface is an elliptic cylinder, with the z-axis
as directrix. Its graph appears in figure 1.5.
108 Example The surface in R3 given by

S : x2 − z2 = 1
is a cylinder, as the y-variable is missing. It is a cylindrical hyperboloid.
109 Example The surface in R3 given by

S : y − .05z2 = 1
is a cylinder, as the x-variable is missing. It is a cylindrical paraboloid.
110 Example (Putnam Exam 1970) Determine, with proof, the radius of the
largest circle which can lie on the ellipsoid
x2 y2 z2
+ + = 1, a > b > c > 0.
a2 b2 c2
Solution: The largest circle has radius b. Parallel cross sections of the
ellipsoid are similar ellipses, hence we may increase the size of these by
moving towards the centre of the ellipse. Every plane through (0, 0, 0)
which makes a circular cross section must intersect the yz-plane, and the
diameter of any such cross section must be a diameter of the ellipse x =
y2 z2
0, 2 + 2 = 1. Therefore, the radius of the circle is at most b. Arguing
b c
similarly on the xy-plane shews that the radius of the circle is at least b.
To shew that circular cross section of radius b actually exist, one may verify
that the two planes given by a2(b2−c2)z2 = c2(a2−b2)x2 give circular cross
sections of radius b.
1.11 Canonical Surfaces in R3

111 Definition A surface S consisting of all lines parallel to a given line ∆
and passing through a given curve Γ is called a cylinder. The line ∆ is
called the directrix of the cylinder.
Canonical Surfaces in R3 45
! To recognise whether a given surface is a cylinder we look at its Carte-

sian equation. If it is of the form f(A, B) = 0, where A, B are secant planes,
then the curve is a cylinder. Under these conditions, the lines generating S
will be parallel to the line of equation A = 0, B = 0. In practice, if one of the
variables x, y, or z is missing, then the surface is a cylinder, whose directrix
will be the axis of the missing coordinate.
112 Example Demonstrate that the surface in R3

2 +y2 +z2
S : ex − (x + z)e−2xz = 0,
implicitly defined, is a cylinder.
Solution: The planes A : x + z = 0 and B : y = 0 are secant. The surface

2 2
has equation of the form f(A, B) = eA +B − A = 0, and it is thus a cylinder.
−
→ − →
The directrix has direction i − k .
113 Example Shew that the surface S in R3 given implicitly by the equation
1 1 1
+ + =1
x−y y−z z−x
is a cylinder and find the direction of its directrix.
Solution: Considering the planes A : x − y = 0, B : y − z = 0, the equation

takes the form
1 1 1
f(A, B) = + − − 1 = 0,
A B A+B
thus the equation represents a cylinder. To find its directrix,
  we  find
 the
x 1
intersection of the planes x = y and y = z. This gives y = t 1. The
z 1
→ −
− → − →
direction vector is thus i + j + k .
114 Definition Given a point Ω ∈ R3 (called the apex) and a curve Γ (called
the generating curve), the surface S obtained by drawing rays from Ω and
passing through Γ is called a cone.
46 Chapter 1
! InApractice,
B
if the Cartesian equation of a surface can be put into the
form f( , ) = 0, where A, B, C, are planes secant at exactly one point,
C C
then the surface is a cone, and its apex is given by A = 0, B = 0, C = 0.
115 Example Demonstrate that the surface in R3 given implicitly by
z2 − xy = 2z − 1
is a cone
Solution: After rearranging, we obtain
(z − 1)2 − xy = 0,
or
x y
− + 1 = 0.
z−1z−1
Considering the planes
A : x = 0, B : y = 0, C : z = 1,
 
0
we see that our surface is a cone, with apex at 0.

1
116 Example The surface in R3 implicitly by
z2 = x2 + y2
x y
is a cone, as its equation can be put in the form ( )2 + ( )2 − 1 = 0.
z z  
0
Considering the planes x = 0, y = 0, z = 0, the apex is located at 0.
0
117 Definition A surface S obtained by making a curve Γ turn around a

line ∆ is called a surface of revolution. We then say that ∆ is the axis of
revolution. The intersection of S with a half-plane bounded by ∆ is called a
meridian.
Canonical Surfaces in R3 47
! If the Cartesian equation of S can be put in the form f(A, Σ) = 0, where

A is a plane and Σ is a sphere, then the surface is of revolution. The axis of
S is the line passing through the centre of Σ and perpendicular to the plane
A.
118 Example Shew that the surface S in R3 implicitly defined as
xy + yz + zx + x + y + z + 1 = 0
is of revolution and find its axis.
Solution: Rearranging,
(x + y + z)2 − (x2 + y2 + z2) + 2(x + y + z) + 2 = 0,
so we may take A : x + y + z = 0, Σ : x2 + y2 + z2 = 0 as our plane and

→ −
− → − →
sphere. The axis of revolution is then in the direction of i + j + k .
119 Example Shew that the surface in R3 implicitly defined by
x4 + y4 + z4 − 4xyz(x + y + z) = 1
is a surface of revolution, and find its axis of revolution.
Solution: Rearranging,
1
(x2 + y2 + z2)2 − ((x + y + z)2 − (x2 + y2 + z2)) − 1 = 0,
2
and so we may take A : x + y + z = 0, Σ : x2 + y2 + z2 = 0, shewing that
→ −
− → − →
the surface is of revolution. Its axis is the line in the direction i + j + k .
120 Example Find the equation of the surface of revolution generated by

revolving the hyperbola x2 − 4z2 = 1 about the z-axis.
Let (x, y, z) be a point on S. If this point were on the xz plane, it would

be on the hyperbola, and its distance to the axis of rotation would be |x| =
p
1 + 4z2. Anywhere else, the distance of (x, y, z) to the axis of rotation is
48 Chapter 1
p
the same as the distance of (x, y, z) to (0, 0, z), that is x2 + y2. We must
have p p
x2 + y2 = 1 + 4z2,
which is to say
x2 + y2 − 4z2 = 1.
121 Example Find the equation of the surface of revolution generated by

revolving the line 3x + 4y = 1 about the y-axis.
Solution: Let (x, y, z) be a point on S. If this point were on the xy plane,

it would be on the line, and its distance to the axis of rotation would be
1
|x| = |1 − 4y|. Anywhere else, the distance of (x, y, z) to the axis of
3 p
rotation is the same as the distance of (x, y, z) to (0, y, 0), that is x2 + z2.
We must have p 1
x2 + z2 = |1 − 4y|,
3
which is to say
9x2 + 9z2 − 16y2 + 8y − 1 = 0.
122 Example Find the equation of the surface of revolution S generated by

revolving the ellipse 4x2 + z2 = 1 about the z-axis.
Solution: Let (x, y, z) be a point on S. If this point were on the xz plane,

it would be on the ellipse, and its distance to the axis of rotation would be
1p
|x| = 1 − z2. Anywhere else, the distance from (x, y, z) to the z-axis
2 p
is the distance of this point to the point (0, 0, z) : x2 + y2. This distance
is the same as the length of the segment on the xz-plane going from the
z-axis. We thus have
p 1p
x2 + y2 = 1 − z2,
2
or
4x2 + 4y2 + z2 = 1.
Parametric Curves and Surfaces 49
123 Example The circle (y − a)2 + z2 = r2, (a, r) ∈ (R∗ )2 on the yz plane
is revolved around the z-axis, forming a torus T . Find the equation of this
torus.
Solution: Let (x, y, z) be a point on T . If this point were on the yz plane, it

would be on the circle, and
p the of the distance to the axis of rotation would
be y = a + sgn(y − a) r2 − z2, where sgn(t) (with sgn(t) = −1 if t < 0,
sgn(t) = 1 if t > 0, and sgn(0) = 0) is the sign of t. Anywhere else, the
p (x, y, z) to the z-axis is the distance of this point to the point
distance from
(0, 0, z) : x2 + y2. We must have
p p
x2 + y2 = (a + sgn(y − a) r2 − z2)2 = a2 + 2asgn(y − a) r2 − z2 + r2 − z2.
Rearranging
p
x2 + y2 + z2 − a2 − r2 = 2asgn(y − a) r2 − z2,
or
(x2 + y2 + z2 − (a2 + r2))2 = 4a2r2 − 4a2z2
since (sgn(y − a))2 = 1, (it could not be 0, why?). Rearranging again,
(x2 + y2 + z2)2 − 2(a2 + r2)(x2 + y2) + 2(a2 − r2)z2 + (a2 − r2)2 = 0.
The equation of the torus thus, is of fourth degree.
1.12 Parametric Curves and Surfaces

124 Definition Let [a; b] ⊆ R. A parametric curve representation r of a
−
→
curve Γ is a function r : [a; b] → Rn, with
 
x1(t)
 x2(t) 
−
→
r (t) =  ..  ,
 
 . 
xn(t)
−
→
and such that r ([a; b]) = Γ . r(a) is the initial point of the curve and r(b) its
terminal point. A curve is closed if its initial point and its final point coincide.
50 Chapter 1
−
→ −
→
The trace of the curve r is the set of all images of r , that is, Γ . If there
−
→ −
→ −
→
exist t1 6= t2 such that r (t1) = r (t2) = p, then p is a multiple point of
the curve. The curve is simple if its has no multiple points. A closed curve
whose only multiple points are its endpoints is called a Jordan curve.
It is important to realise that a curve Γ might have different parametric

representations. For example
[0; 2π] → R2
−
→
r : cos t
t 7→
sin t
and
[0; 2π] → R2
−
→
s : sin 2t
t 7→
cos 2t
are two parametrisations for for the unit circle Γ : {(x, y) ∈ R2 : x2 + y2 =
1}. Notice that r travels the unit circle once starting at (1, 0) and going
counterclockwise, and so r is a simple curve. One the other hand, s starts
at (0, 1) and travels the unit circle twice going clockwise.
We define
lim x1(t)
 
 t→a
lim x2(t) 
−
→  t→a 
lim r (t) = 
..
,
t→a
.

 
lim xn(t)
t→a
provided each of the limits on the right hand side exists and
 0 
x1(t)
 x 0 (t) 
−
→0  2 
r (t) =  ..  ,
 . 
xn0 (t)
provided each of the entries on the right hand side be differentiable at t. In
−
→
particular, we define the differential of r as
 
dx1
→  dx2 
−
 
d r =  ..  .
 . 
dxn
Parametric Curves and Surfaces 51
We will keep our practice of writing

−
→ x(t)
r (t) =
y(t)
in R2 and  
x(t)
−
→
r (t) = y(t)
z(t)
in R3.
125 Example The trace of

−
→ −
→ −
→ −
→
r (t) = i cos t + j sin t + k t
is known as a cylindrical helix.
126 Example Find a parametric representation for the curve resulting by

the intersection of the plane 3x + y + z = 1 and the cylinder x2 + 2y2 = 1 in
R3.
Solution: The projection of the intersection of the plane 3x + y + z and the

cylinder is the ellipse x2 + 2y2 = 1, on the xy-plane. This ellipse can be
parametrised as
√
2
x = cos t, y = sin t, 0 ≤ t ≤ 2π.
2
From the equation of the plane,
√
2
z = 1 − 3x − y = 1 − 3 cos t − sin t.
2
Thus we may take the parametrisation
√cos t
 
 
x(t) 2
−
 
→ sin t
r (t) = y(t) =  .
 
2 √
 
z(t) 2
 
1 − 3 cos t − sin t
2
52 Chapter 1
127 Example Let P be a point at a distance d from the centre of a circle of

radius r. The curve traced out by P as the circle rolls along a straight line
is called a trochoid. Find a parametrisation of the trochoid.
Solution: Let θ be the angle (in radians) of rotation of the circle, and let C
be the centre of the circle. At θ = 0 the centre of the circle is at (0, r), and
P = (0, r − d). Suppose the circle is displaced towards the right, making
the point P to rotate an angle of θ = θ0 radians. Then the centre of the
circle has displaced rθ0 units horizontally, and so is now located at (rθ, r).
The polar co-ördinates of the point P are (d sin θ0, d cos θ0), in relation to
the centre of the circle (notice that the circle moves clockwise). The point
P has moved X = rθ − d sin θ0 horizontal units and Y = r − d cos θ0 units.
This is the desired parametrisation.
128 Example A hypocycloid is a curve traced out by a fixed point P on a

circle C of radius r as C rolls on the inside of a circle with centre at O and
radius R. If the initial position of P is (R, 0), and θ is the angle, measured
counterclockwise, that a ray starting at O and passing through the centre
of C makes with the x-axis, shew that a parametrisation of the hypocycloid
is
(R − r)θ
x = (R − r) cos θ + r cos ,
r

(R − r)θ
y = (R − r) sin θ − r sin .
r
Solution: Suppose that starting from θ = 0, the centre O 0 of the small

circle moves counterclockwise inside the larger circle by an angle θ, and
the point P = (x, y) moves clockwise an angle φ. The arc length travelled
by the centre of the small circle is (R − r)θ radians. At the same time the
point P has rotated rφ radians, and so (R − r)θ = rφ. See figure 1, where
O 0 B is parallel to the x-axis.
Let A be the projection of P on the x-axis. From figure 2, ∠OAP =

π π
∠OPO 0 = , ∠OO 0 P = π − φ − θ, ∠POA = − φ, and OP = (R −
2 2
O θ
0

Parametric CurvesOand Surfaces
P
φ
B
O
O0
A
53
Figure 1.6: Hypocycloid Figure 1.7: Hypocycloid
r) sin(π − φ − θ). Hence

π
x = (OP) cos ∠POA = (R − r) sin(π − φ − θ) cos( − φ),
2
π
y = (R − r) sin(π − φ − θ) sin( − φ).
2
Now
π
x = (R − r) sin(π − φ − θ) cos( − φ)
2
= (R − r) sin(φ + θ) sin φ
(R − r)
= (cos θ − cos(2φ + θ))
2
(R − r)
= (R − r) cos θ − (cos θ + cos(2φ + θ))
2
= (R − r) cos θ − (R − r)(cos(θ + φ) cos φ).
r r
Now, cos(θ + φ) = − cos(π − θ − φ) = − 0
= − and cos φ =
OO R−r
(R − r)θ
cos and so
r

(R − r)θ
x = (R−r) cos θ−(R−r)(cos(θ+φ) cos φ = (R−r) cos θ+r cos ,
r
as required. The identity for y is proved similarly.
129 Definition A parametric representation r of a surface S in R3 is a func-

tion of the form
−
→ R2 → R3
r : ,
(u, v) 7→ r(u, v)
54 Chapter 1
where  
x(u, v)
−
→
r (u, v) = y(u, v) ,
z(u, v)
and whose image is S.
130 Example The surface of a sphere S : {(x, y, z) ∈ R3 : x2 + y2 + z2 = R2}

centred at the origin and of fixed radius R > 0 can be parametrised by
using spherical co-ördinates, with 0 ≤ θ ≤ 2π and 0 ≤ φ ≤ π and
   
x(θ, φ) R cos θ sin φ
−
→
r (θ, φ) = y(θ, φ) =  R sin θ sin φ  .
z(θ, φ) R cos φ
For
x2 + y2 + z2 = R2(cos2 θ sin2 θ + sin2 θ sin2 φ + cos2 φ) = R2.
131 Example A parametrisation for the torus generated by revolving the

circle (y − a)2 + z2 = r2 around the z-axis is
 
a cos θ + r cos θ cos α
−
→
r (θ, α) =  a sin θ + r sin θ cos α  ,
r sin α
with (θ, α) ∈ [−π; π]2.
1.13 Frenet-Serret Formulæ

−
→
132 Definition A curve r : [a; b] → R3 is said to be smooth if it is differen-
−
→
tiable on [a; b] and r 0 is continuous there. The unit tangent vector T of a
smooth curve at t is
−
→ r 0 (t)
T (t) = 0 .
||r (t)||
−
→
The unit normal vector N at t is
−
→0
−
→ T (t)
N (t) = − →0 ,
|| T (t)||
Frenet-Serret Formulæ 55
and the unit binormal vector at t is

−
→ −
→ −
→
B (t) = T (t) × N (t).
→ −
− → − →
The three vectors { T , N , B } are called the Frenet-Serret frame of the
curve r.
133 Example Compute the Frenet-Serret frame at t for the cylindrical helix
r(t) = i cos t + j sin t + kt.
Solution: We have
r 0 (t) = −i sin t + j cos t + k,
√
and ||r 0 (t)|| = 2. Thus
 
− sin t
−
→ 1
T (t) = √  cos t 
2 1
Also,
 
− cos t
−
→0 1
T (t) = √  − sin t  ,
2 0
−
→ 1
and || T 0 (t)|| = √ . Hence
2
 
− cos t
−
→
N =  − sin t  .
0
Finally,
   
− sin t − cos t
−
→ 1 1
B (t) = √  cos t  ×  − sin t  = √ (i(sin t) − j(cos t) + k).
2 1 0 2
56 Chapter 1
1.14 Limits
134 Definition A function f : Rn → Rm is said to have a limit L ∈ Rm at
a ∈ Rn if ∀ > 0∃δ > 0 such that
0 < ||x − a|| < δ =⇒ ||f(x) − L|| < .
In such a case we write,
lim f(x) = L.
x→a
The notions of infinite limits, limits at infinity, and continuity at a point, are
analogously defined. Limits in more than one dimension are perhaps trick-
ier to find, as one must approach the test point from infinitely many direc-
tions.
x2y
135 Example Find lim .
(x,y)→(0,0) x2 + y2
Solution: We use the sandwich theorem. Observe that 0 ≤ x2 ≤ x2 + y2,

x2
and so 0 ≤ 2 ≤ 1. Thus
x + y2
2
x y
lim 0 ≤ lim 2 ≤ lim |y|,
(x,y)→(0,0) (x,y)→(0,0) x + y2 (x,y)→(0,0)
and hence
x2y
lim = 0.
(x,y)→(0,0) x2 + y2
x5y3
(x,y)→(0,0) x6 + y4
Solution: Observe that if |x| ≤ |y| ≤ 0, then

5 3
x y y8 4
x6 + y4 ≤ y4 = y .

If |y| ≤ |x| ≤ 0, then 5 3

x y x8 2
x6 + y4 ≤ x6 = x .

Limits 57
Thus 5 3
x y 4 2 4 2
x6 + y4 ≤ max(y , x ) ≤ y + x −→ 0,

as (x, y) → (0, 0).

Aliter: Let X = x3, Y = y2.
5 3
x y X5/3Y 3/2
x6 + y4 = X2 + Y 2 .

Passing to polar co-ördinates X = ρ cos θ, Y = ρ sin θ, we obtain

5 3
x y X5/3Y 3/2 5/3+3/2−2
x6 + y4 = X2 + Y 2 = ρ
| cos θ|5/3| sin θ|3/2 ≤ ρ7/6 → 0,
as (x, y) → (0, 0).
1+x+y
(x,y)→(0,0) x2 − y2
Solution: When y = 0,
1+x
→ +∞,
x2
as x → 0. When x = 0,
1+y
→ −∞,
−y2
as y → 0. The limit does not exist.
xy6
(x,y)→(0,0) x6 + y8
Solution: Putting x = t4, y = t3, we find
xy6 1
6 8
= 2 → +∞,
x +y 2t
as t → 0. But when y = 0, the function is 0. Thus the limit does not exist.
((x − 1)2 + y2) loge((x − 1)2 + y2)

(x,y)→(0,0) |x| + |y|
58 Chapter 1
Solution: When y = 0 we have
2(x − 1)2 ln(|1 − x|) 2x

∼− ,
|x| |x|
and so the function does not have a limit at (0, 0).
sin(x4) + sin(y4)
140 Example Find lim p .
(x,y)→(0,0) x4 + y4
Solution: sin(x4) + sin(y4) ≤ x4 + y4 and so

sin(x4) + sin(y4) p
≤ x4 + y4 → 0,

p
4
x +y 4
as (x, y) → (0, 0).
sin x − y
(x,y)→(0,0) x − sin y
Solution: When y = 0 we obtain

sin x
→ 1,
x
as x → 0. When y = x the function is identically −1. Thus the limit does
not exist.
Chapter 2
Differentiation
2.1 Local Study of Functions

142 Definition A function α : R → R is said to be infinitesimal as x → a if
lim α(x) = 0.
x→a
143 Example sin : R → [−1; 1] is infinitesimal as x → 0, since lim sin x = 0.

x→0
1
144 Example f : R∗ → R, x 7→ is infinitesimal as x → +∞, since
x
1
lim = 0.
x→+∞ x
145 Definition We say that a function α : R → R is asymptotic to a function

β : R → R as x → a, and we write α ∼ β, if
α(x)
lim = 1.
x→a β(x)
sin x
146 Example We have sin x ∼ x as x → 0, since lim = 1.
x→0 x
2 x2 + x
147 Example We have x + x ∼ x as x → 0, since lim = 1.
x→0 x
59
60 Chapter 2
x2 + x
148 Example We have x2 + x ∼ x2 as x → +∞, since lim = 1.
x→+∞ x2
149 Definition (Du-Bois Raymond-Landau “small oh” notation) We say that

α : R → R is negligible in relation to β : R → R as x → a or that β is pre-
ponderant in relation to α as x → a, if
α(x)
lim = 0.
x→a β(x)
o
We express the condition above with the notation α(x) = (β(x))
x→a
(read “α of x is small oh of β of x as x tends to a”)
! When the limit to which the independent variable tends is understood,

o
we abbreviate (α(x)) as o (α(x))
x→a
In particular, if α is infinitesimal as x → a, then α(x) = o (1) as x → a.
150 Example We have x2 = o (x) as x → 0 since

x2
= lim x = 0.
lim
x→0 x x→0
151 Example We have x = o x2 as x → +∞ since

x 1
lim 2
= lim = 0.
x→+∞ x x→0 x
We write α(x) = γ(x) + o (β(x)) as x → a if α(x) − γ(x) = o (β(x)) as

x → a.
152 Example We have sin x = x + o (x) as x → 0 since

sin x − x sin x
lim = lim − lim 1 = 1 − 1 = 0.
x→0 x x→0 x x→0
153 Theorem Let α and β be infinitesimal functions as x → a. Then the

following hold.
Local Study of Functions 61
• The sum of two infinitesimals is an infinitesimal:
o (β(x)) + o (β(x)) = o (β(x)) .
• The difference of two infinitesimals is an infinitesimal:
o (β(x)) − o (β(x)) = o (β(x)) .
• ∀c ∈ R \ {0}, o (cβ(x)) = o (β(x)) .
• ∀n ∈ N, n ≥ 2, 1 ≤ k ≤ n − 1, o ((β(x))n) = o (β(x))k .

• o (o (β(x))) = o (β(x)).
• ∀n ∈ N, n ≥ 1, (β(x))no (β(x)) = o (β(x))n+1 .

o ((β(x))n)
= o (β(x))n−1 .

• ∀n ∈ N, n ≥ 2,
β(x)
o (β(x))
• = o (1).
β(x)
n
!
X
• If ck are real numbers, then o ck(β(x))k = o (β(x)).
k=1
• (αβ)(x) = o (α(x)) and (αβ)(x) = o (β(x)).
• If α ∼ β, then (α − β)(x) = o (α(x)) and (α − β)(x) = o (β(x)).
154 Example In view of theorem 153, we have
o −2x3 + 8x2 = o (x) ,

as x → 0.
−2x3 + 8x2 + x ∼ x,
as x → 0.
62 Chapter 2
o −2x3 + 8x2 = o x4 ,

as x → +∞.
−2x3 + 8x2 + x ∼ −2x3,
as x → +∞.
The following result follows easily from the Maclaurin expansion of the
given function.
158 Theorem Let x → 0. Then
x3 x5 x2n+1
− · · · + (−1)n + o x2n+2 .

• sin x = x − +
3! 5!
2
(2n + 1)!
In particular, sin x = x + o x .
x2 x4 x2n
− · · · + (−1)n + o x2n+1 .

• cos x = 1 − +
2! 4! (2n)!
In particular, cos x = 1 + o (x).
x3 2x5
+ o x5 .

• tan x = x + +
3 15
x2 x3 xn
• ex = 1 + x + + + ··· + + o (xn)
2! 3! n!
x2 x3 xn
• loge(1 + x) = x − + − · · · + (−1)n+1 + o (xn).
2 3 n
τ(τ − 1) 2 τ(τ − 1)(τ − 2)(τ − 3) · · · (τ − n + 1) n

• (1+x)τ = 1+τx+ x +· · ·+ x +
2 n!
o (xn).
159 Example Since tan x = x + o (x), we have

x2 x2
2
x x2
+ o x2 ,

tan = +o =
2 2 2 2
as x → 0. Also,
(tan x)3 = (x + o (x))3 = x3 + 3x2o (x) + 3xo x2 + (o (x))3 = x3 + o x3 .

x2
+ o x2 , we have

160 Example Since cos x = 1 −
2!
9x4
cos 3x2 = 1 − + o x4 .

2
161 Example Find an asymptotic expansion of cot2 x of type o x−2 as

x → 0.
Solution: Since tan x ∼ x we have

1
cot2 x ∼ .
x2

2 1 1
We can write this as cot x = 2 + o .
x x2
162 Example Calculate

x2
sin sin tan
lim 2 .
x→0 loge cos 3x
Solution: We use theorems 158 and 153.

x2
2
x 2

sin sin tan = sin sin +o x
2 2 2 2
x 2
x 2

= sin +o x +o +o x
22 2
x
+ o x2

= sin
2
x2
+ o x2 ,

=
2
64 Chapter 2
and
9x2

2

loge cos 3x = loge 1 − +o x
2
9x2 9x2

2 2

= − +o x +o − +o x
2 2!
9x2
+ o x2

= −
2
The limit is thus equal to
x2 1
+ o x2 + o (1) 1
lim 2 2 = lim 2 =− .
x→0 9x x→0 9 9
− + o (x2) − + o (1)
2 2
2
163 Example Find lim (cos x)(cot x)
.
x→0

1 2 1
Solution: By example 161, we have cot x = 2 + o . Also,
x x2
x2 x2

+ o x2 + o x2 .

loge cos x = loge 1− =−
2 2
Hence
2 x 2
(cos x)cot

= exp (cot x)
2log e cos x
1 1 x 2

= exp +o − +o x
x2 x2 2
1
= exp(− + o (1))
2
1
−
→ e 2,
as x → 0.
164 Example Find an

4
asymptotic development of loge(2 cos x+sin x) around
x = 0 of order o x .
Solution: By theorem 158,

x2 x4 x3

5 4

2 cos x + sin x = 2 1 − + +o x + x− +o x
2 24 6
x3 x4
2 + x − x2 − + o x4

= +
6 12
x x2 x3 x4

4

= 2 1+ − − + +o x ,
2 2 12 24
and so,
x x2 x3 x4

4

loge(2 cos x + sin x) = loge 2 1 + − − + +o x
2 2 12 24
x x2 x3 x4

4

= loge 2 + loge 1 + − − + +o x
2 2 12 24
x x2 x3 x4

+ o x4

= loge 2 + − − +
2 2 12 24
2 3
x4 2

1 x x x 4
− − − + +o x
2 2 2 12 24
1 x x2 x3 x4 3

4
+ − − + +o x
3 2 2 12 24
1 x x2 x3 x4 4

4
+ o x4

− − − + +o x
4 2 2 12 24
x x2 x3 x4 1 x2 x3 x4

= loge 2 + − − + − − +
2 2 12 24 2 4 2 6
1 x3 3x4 1 x4

+ o x4

+ − − ·
3 8 8 4 16
x 5x2 5x3 35x4
+ o x4

= loge 2 + − + −
2 8 24 192
as x → 0.
Our main concern, as far as asymptotic developments is concerned,
−
→
will be expansions when the vector tends to 0 . Let f, g : Rn → Rm. By
−
→ −
→ −
→ −
→ −
→
f( v ) = g( v ) + o v , as v → 0 , we mean that
−
→ −
→
||f( v ) − g( v )||
lim− − = 0.
→
→

−
→ v
v→0
In the direction of local behaviour of vector functions, we start with the

following theorem.
66 Chapter 2
−
→
165 Theorem Let L : Rn → Rm be a linear transformation. Then L( 0 ) =
−
→
0 . Moreover, there exists a constant M > 0 such that
−
→ −
→

||L( h )|| ≤ M h .
Proof By linearity,
−
→ → −
− → −
→ −
→
L( 0 ) = L( 0 + 0 ) = L( 0 ) + L( 0 ),
which gives the first part.
Now, let
X −n
−
→ →
h = hk e k,
k=1
−
→
where the e k are the standard basis for Rn. Then
n
X
−
→ −
→
L( h ) = hkL( e k),
k=1
and hence by the triangle inequality, and by the Cauchy-Bunyakovsky-

Schwarz inequality,
n
X
−
→ −
→
||L( h )|| ≤ |hk|||L( e k)||
k=1
Xn n
X
2 1/2 −
→
≤ ( |hk| ) ( ||L( e k)||2)1/2
k=1 k=1
n
→ X
−
−
→
= h ( ||L( e k)||2)1/2,
k=1
Xn
−
→
and the theorem follows upon putting M = ( ||L( e k)||2)1/2. ❑
k=1
166 Corollary With M as in theorem 165,

−
→ −
→ −
→2
−
→

|| h × L( h )|| ≤ M h = o h ,
−
→ −
→
as h → 0 .
Definition of the Derivative 67
2.2 Definition of the Derivative

We now define the derivative in a multidimensional vector space. Recall
that in one variable, a function g : R → R is said to be differentiable at
x = a if the limit
g(x) − g(a)
lim = g 0 (a)
x→a x−a
exists. The limit condition above is equivalent to saying that
g(x) − g(a) − g 0 (a)(x − a)

lim = 0,
x→a x−a
or equivalently,
g(a + h) − g(a) − g 0 (a)(h)

lim = 0.
h→0 h
This last condition we may write in small oh notation as
g(a + h) − g(a) = g 0 (a)(h) + o (h) ,
as h → 0. The above analysis provides an analogue definition for the

higher-dimensional case. Observe that since we may not divide by vec-
tors, the corresponding definition in higher dimensions involves quotients
of norms.
167 Definition Let A ⊆ Rn. A function f : A → Rm is said to be differen-

−
→
tiable at a ∈ A if there is a linear transformation, called the derivative of f
−
→ → n m
at a , D−a (f) : R → R such that
−
→ −
→ → −
− →
||f( x ) − f( a ) − D−→ (f)( x − a )||
a
lim→
→ −
− → −
− → = 0.
x→a || x − a ||
−
→ → (f)
Equivalently, f is differentiable at a if there is a linear transformation D−
a
such that
→ − − −
− → − → →

→
f( a + h ) − f( a ) = D− → (f)( h ) + o h ,
a
−
→ −
→
as h → 0 .
68 Chapter 2
! The condition for differentiability at −→a is equivalent to

−
→ −
→ → −
− → → −
− →
f( x ) − f( a ) = D−
→ (f)( x − a ) + o || x − a || ,
a
−
→ −
→
as x → a .
→ (f) is uniquely deter-

168 Theorem If A is an open set in definition 167, D−
a
mined.
Proof Let L : Rn → Rm be another linear transformation satisfying defini-

−
→ −
→ −
→
tion 167. We must prove that ∀ v ∈ Rn,L( v ) = D− → (f)( v ). Since A is
a
→ −
− → −
→ −
→ −
→
open, a + h ∈ A for sufficiently small h . By definition, as h → 0 ,
we have
→ − − −
− → − → →

→
f( a + h ) − f( a ) = D−
→ (f)( h ) + o h .
a
and
→ −
− → −
→ −
→
f( a + h ) − f( a ) = L( h ) + o (||vectorh||) .
Now, observe that
−
→ −
→ −
→ → −
− → −
→ → −
− → −
→ −
→
→ (f)( v )−L( v = D−
D− → (f)( h )−f( a + h )+f( a )+f( a + h )−f( a )−L( h ).
a a
By the triangle inequality,

−
→ −
→ −
→ → −
− →
→ (f)( v ) − L( v || ≤
||D− → (f)( h ) − f( a + h ) + f(a)||
||D−
a a
→ −
− → −
→
+||f(
a− + h ) −
f(a )− L(h)||
→ −
→

= o h + o h
−
→

= o h ,
−
→ −
→
as h → 0 This means that
−
→ −
→
||L( v ) − D−
→ (f)( v )|| → 0,
a
−
→ −
→
i.e., L( v ) = D−
→ (f)( v ), completing the proof.
a ❑
!
→ −
− →
−
→
If A = { a }, a singleton, then D (f) is not uniquely determined. For
−
→ −
→
−
→
a
−
→ −
→
|| x − a || < δ holds only for x = a , and so f( x ) = f( a ). Any linear
→ −
− → −
→ −
→
transformation T will satisfy the definition, as T ( x − a ) = T ( 0 ) = 0 , and
−
→ −
→ → −
− → −
→
||f( x ) − f( a ) − T ( x − a )|| = || 0 || = 0,
identically.
169 Example If L : Rn → Rm is a linear transformation, then D−

→ (L) = L,
a
n
for any a ∈ R .
Solution: Since Rn is an open set, we know that D− → (L) uniquely deter-

a
mined. Thus if L satisfies definition 167, then the claim is established. But
by linearity
−
→ −
→ → −
− → −
→ −
→ −
→ −
→
||L( x ) − L( a ) − L( x − a )|| = ||L( x ) − L( a ) − L( x ) + L( a )|| = ||0|| = 0,
whence the claim follows.

−
→
170 Example Let f : Rn → R, n ≥ 1, f( x ) = ||x|| be the usual norm in Rn,
−
→ 2 −
→− →
with x = x • x . Prove that
→•−
− →
−
→ x v
→ (f)( v ) =
D− ,
x
||x||
→ −
− → −
→
for x 6= 0 , but that f is not differentiable at 0 .
→ −
− → t
Solution: Assume that x 6= 0 . We use the fact that (1+t)1/2 = 1+ +o (t)
2
as t → 0, from theorem 153. We have
→ −
− → −
→ → −
− →
f( x + h ) − f( x ) = || x + h || − ||x||
− − → −
q
→ → − →
= ( x + h )•( x + h ) − ||x||
r
−
→
2 →−
− → − →2 −→
= x + 2 x • h + h − x
→−
− → −→2
2 x • h + h
= r 2 .
− −
→ −→
x + 2 x • h + h + −
→ 2 −
→ →
x

−
→ −
→
As h → 0 ,
r
−
→
2 →−
− → −→2 −→ −
→
x + 2 x • h + h + x → 2 x .
70 Chapter 2
− − − −
→2 → → →

Since h = o h as h → 0 , we have
→−
− → −→2
→•−
− →
2 x • h + h x h −
→

r → −
→
+ o h ,
−
→ 2 −
→ −
→ − → 2
−
→
x + 2 x • h + h + x
h
proving the first assertion.

To prove the second assertion, assume that there is a linear transfor-
mation D−→ (f) = L, L : Rn → R such that
0
→ −
− → −
→ −
→ −
→

||f( 0 + h ) − f( 0 ) − L( h )|| = o h ,
− − −
→ → →

as h → 0. Recall that by theorem 165, L( 0 ) = 0 , and so by example
−
→
−
→ −
→ −
→ L( h ) −
→ −
→
169, D− → (L)( 0 ) = L( 0 ) = 0 . This implies that → (L)( 0 ) = 0 ,
0 → → D−
− 0
h
−
→ −
→ −
→ −
→
as h → 0. Since f( 0 ) = ||0|| = 0, f( h ) = h this would imply that

− − −
→ → →

h − L( h ) = o h ,
or
−

→

1 − L( h )
→ = o (1) .
−
h
−
→ −
→ −
→ −
→
But the sinistral side → 1 as h → 0 , and the dextral side → 0 as h → 0 .
This is a contradiction, and so, such linear transformation L does not exist
−
→
at the point 0 .
171 Example Let L : R3 → R3 be a linear transformation and
R3 → R3
F: →
− −
→ → .
−
x 7→ x × L( x )
Shew that F is differentiable and that

−
→ −
→ −
→ −
→ −
→
D−→ (F)( h ) = x × L( h ) + h × L( x ).
x
Solution: We have
→ −
− → −
→ → −
− → → −
− → −
→ −
→
F( x + h ) − F( x ) = ( x + h ) × L( x + h ) − x × L( x )
→ −
− → −
→ −
→ −
→ −
→
= ( x + h ) × (L( x ) + L( h )) − x × L( x )
−
→ −
→ −
→ −
→ −
→ −
→
= x × L( h ) + h × L( x ) + h × L( h )
−
→ −
→ −
→
−
→ −
→
Now, || h × L( h )|| = o h as h → 0 by virtue of corollary 166. This
yields the desired result.
The following examples consider derivatives of functions with domains
other than Rn.
172 Example Let

R3 × R3 → R
f: → −
− → − •−
→ →
( x , y ) 7→ x y
be the usual dot product in R3. Shew that f is differentiable and that
→ −
− → →•−
− → − →−→
x , y )f( h , k ) = x k + h y .
D(−
→− → •
Solution: We have
→ −
− → −→ − → → −
− → → −
− → − → − → →−
− →
f( x + h , y + k ) − f( x , y ) = ( x + h )•( y + k ) − x • y
− −
→ → − →−→ − →− → − →− → − →−
→
= x•y + x•k + h•y + h•k − x•y
→•−
− → − →− → − →−→
= x k + h•y + h•k.
→ −
− → → −
− →
As ( h , k ) → ( 0 ,0 ), we haveby the Cauchy-Buniakovskii-Schwarz in-
→−
− → −
→ −→
equality, | h • k | ≤ h ||k|| = o h , which proves the assertion.

173 Example Shew that
Mn(R) → Mn(R)
f:
X 7→ X2
is differentiable and that
DXf(H) = XH + HX.
72 Chapter 2
Solution: We have
f(X + H) − f(X) = (X + H)2 − X2
= X2 + XH + HX + H2 − X2
= XH + HX + H2
= XH + HX + o (H) ,
proving the assertion.
174 Example Let

M2(R) → R
f: .
A 7→ det A
Given H ∈ M2(R), find an expression for DA(f)(H)
Solution: Put
a1 a2 h1 h2
A= , H= .
a3 a4 h3 h4
Then

a1 + h1 a2 + h2 a1 a2
f(A + H) − f(A) = det −
a3 + h3 a4 + h4 a3 a4
= (a1 + h1)(a4 + h4) − (a2 + h2)(a3 + h3) − a1a4 + a2a3
= a1a4 − a2a3 + a1h4 + a4h1 − a2h3 − a3h2
+h1h4 − h2h3 − a1a4 + a2a3
= h4 + a4h1 −
a1 a2h3 − a3
h2 + h1h4 − h2h3
h1 h2 a4 −a2
= tr + h1h4 − h2h3
h3 h4 −a3 a1
= tr (Hadj (A)) + h1h4 − h2h3,
where adj (A) is the (classical) adjoint matrix of A. Now, the Arithmetic-
Mean-Geometric-Mean inequality for real numbers a, b says that |a||b| ≤
a2 + b2
and so
2
h2 + h22 + h23 + h24 1
|h1h4 − h2h3| ≤ |h1||h4| + |h2||h3| ≤ 1 = ||H||2 = o (H) ,
2 2
as H → 02. This implies that
DA(f)(H) = tr (Hadj (A)) .
Just like in the one variable case, differentiability at a point, implies
continuity at that point.
The Jacobi Matrix 73
175 Theorem Suppose A ⊆ Rn is open and f : A → Rn is differentiable on

A. Then f is continuous on A.
−
→
Proof Given a ∈ A, we must shew that
−
→ −
→
→
lim f( x ) = f( a ).
− −
→
x→a
−
→
Since f is differentiable at a we have
−
→ −
→ → −
− → → −
− →
f( x ) − f( a ) = D−
→ (f)( x − a ) + o || x − a || ,
a
and so −
−
→ −
→ →
f( x ) − f( a ) → 0 ,
−
→ −
→
as x → a , proving the theorem. ❑
2.3 The Jacobi Matrix

We now establish a way which simplifies the process of finding the deriva-
tive of a function at a given point.
176 Definition Let A ⊆ Rn, f : A → Rm, and put

 
f1(x1, x2, . . . , xn)
 f2(x1, x2, . . . , xn) 
−
→
f( x ) =  .
 
..
 . 
fm(x1, x2, . . . , xn)
∂fi
Here fi : Rn → R. The partial derivative (x) is defined as
∂xj
∂fi fi(x1, x2, . . . , xj + h, . . . , xn) − fi(x1, x2, . . . , xj, . . . , xn)
(x) = lim ,
∂xj h→0 h
whenever this limit exists.
To find partial derivatives with respect to the j-th variable, we simply

keep the other variables fixed and differentiate with respect to the j-th vari-
able.
74 Chapter 2
 
x
−
→
177 Example If f : R3 → R, and f( r ) = f y = x + y2 + z3 + 3xy2z3 then
z
−
→
D1(f)( r ) = 1 + 3y2z3,
−
→
D2(f)( r ) = 2y + 6xyz3,
and
−
→
D3(f)( r ) = 3z2 + 9xy2z2.
! Sometimes we will use the equivalent notations

−
→ ∂f
D1(f)( r ) = (x, y, z),
∂x
−
→ ∂f
D2(f)( r ) = (x, y, z),
∂y
etc..
178 Example Suppose g : R → R is continuous and a ∈ R is a constant.

Find the partial derivatives with respect to x and y of
Z x2 y
2 x
f : R → R, f = g(t) dt.
y a
Solution: We have
∂f
(x, y, z) = 2xyg(x2y),
∂x
and
∂f
(x, y, z) = x2g(x2y).
∂y
Since the derivative of a function f : Rn → Rm is a linear transformation,
it can be represented by aid of matrices. The following theorem will allow
us to determine the matrix representation for D− → (f) under the standard
a
n m
bases of R and R .
179 Theorem Let  

f1(x1, x2, . . . , xn)
 f2(x1, x2, . . . , xn) 
−
→
f( x ) =  .
 
..
 . 
fm(x1, x2, . . . , xn)
Suppose A ⊆ Rn is an open set and f : A → Rm is differentiable. Then
∂fi
each partial derivative (x) exists, and the matrix representation of D−
→ (f)
x
∂xj
n m
with respect to the standard bases of R and R is the Jacobi matrix
 
∂f1 ∂f1 ∂f1
 ∂x1 (x) ∂x2 (x) · · · ∂xn (x)
 ∂f2 ∂f2 ∂f2
 
(x) (x) · · · (x)


J−→f =
 ∂x1. ∂x2 ∂xn
 .
x
 .. .
.. .
.. .
..


 
 ∂fn ∂fn ∂fn 
(x) (x) · · · (x)
∂x1 ∂x2 ∂xn
−
→
Proof Let e j, 1 ≤ j ≤ n, be the standard basis for Rn. To obtain the Jacobi
−
→
→ (f)( e j), which will give us the j-th column of
matrix, we must compute D− x
the Jacobi matrix. Let J− → f = (Jij), and observe that
x
 
J1j
 J2j 
−
→
→ (f)( e j) =  .  .
D−
 
x
 .. 
Jnj
→ −
− → −
→
and put y = x + ε e j, ε ∈ R. Notice that
−
→ −
→ → −
− →
||f( y ) − f( x ) − D− → (f)( y − x )||
x
→ −
− →
|| y − x ||
−
→
||f(x1, x2, . . . , xj + h, . . . , xn) − f(x1, x2, . . . , xj, . . . , xn) − εD−
→ (f)( e j)||
x
= .
|ε|
Since the sinistral side → 0 as ε → 0, the so does the i-th component of
the numerator, and so,
|fi(x1, x2, . . . , xj + h, . . . , xn) − fi(x1, x2, . . . , xj, . . . , xn) − εJij|
→ 0.
|ε|
76 Chapter 2
This entails that

fi(x1, x2, . . . , xj + ε, . . . , xn) − fi(x1, x2, . . . , xj, . . . , xn) ∂fi
Jij = lim = (x) .
ε→0 ε ∂xj
This finishes the proof. ❑
180 Example Let f : R3 → R2 be given by

−
→ xy + yz
f( r ) = .
loge xy
Compute the Jacobi matrix of f under the usual basis for R3 and R2. Also,
compute the matrix representation for D− → (f) under the the usual basis for
r  
4
3 1 1 2
R and the basis , in R . Finally, find D(1,2,3)f 5.

0 1
6
Solution: The Jacobi matrix is the 2 × 3 matrix
→ ∂f1 −
∂f1 − → ∂f1 −
  
→ 
(r) (r) (r) y x+z y
 ∂x
→f = 
∂y ∂z  1 1
J−r ∂f2 − ∂f ∂f = .
→ 2 −
→ 2 −
→ 0
(r) (r) (r) x y
∂x ∂y ∂z
Each of the column vectors of J− → f is expressed in the standard basis of

r
2 1 1
R . To express them in the basis , of R2 we simply write
0 1
" #
y 1 1

1 1
1 = (y − ) + ,
x 0 x 1
x
 
x+z 1

1

 1  = (x + z − ) 1 1
+ ,
y 0 y 1
y

y 1 1
=y +0 .
0 0 1
The desired matrix representation is
1 1
 
y− x+z− y
A− = 1 x y
.
→
 
r 1
0
x y
 
4
To compute D(1,2,3)f 5, we will use the Jacobi matrix, which will give us
6
the result in the standard basis of R2.
  " #  " #
4 2 4 2 4 40
J(1,2,3) 5 = 1 5 = 7 .
1 0
6 2 6 2
181 Example Let f(ρ, θ, z) = (ρ cos θ, ρ sin θ, z) be the function which changes
from cylindrical co-ördinates to Cartesian co-ördinates. We have
 
cos θ −ρ sin θ 0
J− −−−→ f =  sin θ ρ cos θ 0 .
(ρ,θ,z)
0 0 1
182 Example Let f(ρ, φ, θ) = (ρ cos θ sin φ, ρ sin θ sin φ, ρ cos φ) be the
function which changes from spherical co-ördinates to Cartesian co-ördinates.
We have
 
cos θ sin φ ρ cos θ cos φ −ρ sin φ sin θ
J− −−−→ f =  sin θ sin φ ρ sin θ cos φ ρ cos θ sin φ  .
(ρ,φ,θ)
cos φ −ρ sin φ 0
The Jacobi matrix provides a convenient computational tool to compute

the derivative of a function at a point. Thus differentiability at a point implies
that the partial derivatives of the function exist at the point. The converse,
however, is not true.
78 Chapter 2
183 Example Let f : R2 → R be given by


 y if x = 0,
f(x, y) = x if y = 0,

1 if xy 6= 0.
Observe that f is not continuous at (0, 0) (f(0, 0) = 0 but f(x, y) = 1 for val-
ues arbitrarily close to (0, 0)), and hence, it is not differentiable there. We
∂f ∂f
have however, (0) = (0) = 1. Thus even if both partial derivatives
∂x1 ∂x2
exist at (0, 0) is no guarantee that the function will be differentiable at (0, 0).
You should also notice that both partial derivatives are not continuous at
(0, 0).
We have, however, the following.
184 Theorem
  Let A ⊆ Rn be an open set, and let f : Rn → Rm. Put
f1
 f2 
f= . . .. If each of the partial derivatives Djfi exists and is continuous

fm
on A, then f is differentiable on A.
Just like in the one-variable case, we have the following rules of differen-
tiation. Let A ⊆ Rn, B ⊆ Rm be open sets f, g : A → Rm, α ∈ R, be
differentiable on A, h : B → Rl be differentiable on B, and f(A) ⊆ B. Then
we have
• Addition Rule: D−→ ((f + αg)) = D− → (f) + αD−→ (g).

x x x

• Chain Rule: D− → ((h ◦ f)) = D−−→ (h) ◦ (D− → (f)).
x f (x) x
Since composition of linear mappings expressed as matrices is matrix mul-

tiplication, the Chain Rule takes the alternative form when applied to the
Jacobi matrix.
→ (h ◦ f) = (J−−→ h)(J−
J− → f). (2.1)
x f (x) x
185 Example Let

uev
 

u
f = u + v ,
v
uv
2
−
→ x +y
h( r ) = .
y+z
→ (f ◦ h).
Find J−
r
Solution: We have  v
e uev

J−−−→ f =  1 1 ,
(u,v )
v u
and
2x 1 0
→h =
J− .
r 0 1 1
Observe also that
ey+z (x2 + y)ey+z

 
J−−→ f =  1 1 .
h(r)
2
y+z x +y
Hence
→ (f ◦ h) = (J−−→ f)(J−
J− → h)
r r
 h(r) 
y+z 2 y+z 
e (x + y)e  
  2x 1 0
= 
 1 1 
 
  0 1 1
y+z x2 + y
 
y+z 2 y+z 2 y+z
 2xe (1 + x + y)e (x + y)e 
 
= 
 2x 2 1 .

 
2xy + 2xz x2 + 2y + z x2 + y
80 Chapter 2
186 Example Let

2 u
f : R → R, f = u2 + ev,
v
−
→ −
→
u, v : R3 → R u( r ) = xz, v( r ) = y + z.

−
→ u(x, y, z) ∂h
Put h( r ) = f . Find (r) for i = 1, 2, 3.
v(x, y, z) ∂xi
−→
3 −
→ 2 u( r ) xz
Solution: Put g : R → R , g( r ) = → = y + z . Observe that
−
v( r )
h = f ◦ g. Now,
z 0 x
J−→g = ,
r 0 1 1
−−→ f = 2u ev ,

J− (u,v )
−→ f = 2xz ey+z .

J−
g(r)
Thus

∂h ∂h ∂h
(r) (r) (r) = J−
→h
r
∂x1 ∂x2 ∂x3
= (J−−→ f)(J−
g(r)
→ g)
r
 

z 0 x .
= 2xz ey+z  
0 1 1

= 2xz e
2 y+z 2
2x z + ey+z
Equating components, we obtain
∂h
(r) = 2xz2,
∂x1
∂h
(r) = ey+z,
∂x2
∂h
(r) = 2x2z + ey+z.
∂x3
Gradients and Directional Derivatives 81
2.4 Gradients and Directional Derivatives

A function
Rn → Rm
f: →
− −
→
x 7→ f( x )
is called a vector field. If m = 1, it is called a scalar field.
187 Definition Let

Rn → R
f: →
− −
→
x 7→ f( x )
be a scalar field. The gradient of f is the vector defined and denoted by
 
∂f
 ∂x1 (x) 
 ∂f
 
(x)

−
→ 
→ f)T =  ∂x2

∇f( x ) = (J− x
.

 .
..


 
 ∂f 
(x)
∂xn
The gradient operator is the operator

 
D1
 D2 
∇ =  ..  .
 
 . 
Dn
188 Theorem Let A ⊆ Rn be open and let f : A → R be a scalar field, and

−
→
assume that f is differentiable in A. Let K ∈ R be a constant. Then ∇f( x )
−
→
is orthogonal to the surface implicitly defined by f( x ) = K.
Proof Let
n
− R → R
→
c : −
→
t 7→ c (t)
−
→ −
→
be a curve lying on this surface. Choose t0 so that c (t0) = x . Then
−
→ −
→
(f ◦ c )(t0) = f( c (t)) = K,
82 Chapter 2
and using the chain rule

−
→
(J− → c ) = 0,
−→ f)(J−
c(t ) t0
0
which translates to
→ −
− →
(∇f( x ))•( c 0 (t0)) = 0.
−
→ −
→
Since c 0 (t0) is tangent to the surface and its dot product with ∇f( x ) is 0,
−
→
we conclude that ∇f( x ) is normal to the surface. ❑
! Let θ be the angle between ∇f(−→x ) and −→c (t ). Since 0

0
−
→ −
→ → −
− →
|(∇f( x ))•( c 0 (t0))| = ||∇f( x )|||| c 0 (t0)|| cos θ,
−
→
∇f( x ) is the direction in which f is changing the fastest.
189 Example Find a unit vector normal to the surface x3 + y3 + z = 4 at the

point (1, 1, 2).
Solution: Here f(x, y, z) = x3 + y3 + z − 4 has gradient

 2
3x
−
→
∇f( r ) = 3y2
1
 
3
which at (1, 1, 2) is 3. Normalising this vector we obtain
1
3
 
√
 19 
 3 
√ 
 .
 19 
 1 
√
19
190 Example Find the direction of the greatest rate of increase of f(x, y, z) =
xyez at the point (2, 1, 2).
Gradients and Directional Derivatives 83
Solution: The direction is that of the gradient vector. Here

 z
ye
−
→
∇f( r ) = xez 

xyez
 2
e
which at (2, 1, 2) becomes 2e2 . Normalising this vector we obtain

2e2
 
1
1
√ 2 .
5 2

 
x
f  y  = x + y2 − z2.
z
Find the equation of the tangent plane to f at (1, 2, 3).
Solution: A vector normal to the plane is ∇f(1, 2, 3). Now

 
1
−
→
∇f( r ) =  2y 
−2z
which is  
1
4
−6
at (1, 2, 3). The equation of the tangent plane is thus
1(x − 1) + 4(y − 2) − 6(z − 3) = 0,
or
x + 4y − 6z = −9.
84 Chapter 2
192 Definition Let

Rn → Rn
f: →
− −
→
x 7→ f( x )
be a vector field with −
→ 

f1( x )
 f (−
→ 
−
→  2 x )
f( x ) =  .  .
 .. 
−
→
fn( x )
The divergence of f is defined and denoted by
−
→ ∂f1 ∂f2 ∂fn
divf(x) = ∇•f( x ) = (x) + (x) + · · · + (x) .
∂x1 ∂x2 ∂xn
2
193 Example If f(x, y, z) = (x2, y2, yez ) then
2
divf(x) = 2x + 2y + 2yzez .
194 Definition Let gk : Rn → Rn, 1 ≤ k ≤ n − 2 be vector fields with

gi = (gi1, gi2, . . . , gin). Then the curl of (g1, g2, . . . , gn−2)
 −
→ −
→ −
→ 
e1 e2 ··· en
 D1 D2 ··· Dn 
 −
→ −
→ −
→ 
 g11( x ) g 12 ( x ) · · · g 1n( x ) 
curl(g1, g2, . . . , gn−2)(x) = det  g (− → −
→ −
→  .

 21 x ) g22( x ) ... g2n( x )  
 .. .. .. .. 
 . . . . 
−
→ −
→ −
→
g(n−2)1( x ) g(n−2)2( x ) . . . g(n−2)n( x )
2
195 Example If f(x, y, z) = (x2, y2, yez ) then
−→ − → − → 
i j k
2 −
→
curlf(r) = det D1 D2 D3  = (ez ) i .
 
2
x2 y2 yez
196 Example If f(x, y, z, w) = (exyz, 0, 0, w2), g(x, y, z, w) = (0, 0, z, 0) then

  −→ − → − → − → 
x e1 e2 e3 e4
y
 = det  D
 1 D2 D3 D4  2 xyz −
→
curl(f, g)  2  = (xz e ) e 4.

z exyz 0 0 w
w 0 0 z 0
Extrema 85
197 Definition Let A ⊆ Rn be open and let f : A → R be a scalar field,

−
→ −
→
and assume that f is differentiable in A. Let v ∈ Rn \ { 0 } be such that
−
→ −
→
x + t v ∈ A for sufficiently small t ∈ R. Then the directional derivative of
−
→ −
→
f in the direction of v at the point x is defined and denoted by
−
→ −
→ −
→
∂f f( x + t v ) − f( x )
(x) = lim .
∂x−
→
v t→0 t
! Some authors require that the vector −→v in definition 197 be a unit
vector.
198 Theorem Let A ⊆ Rn be open and let f : A → R be a scalar field,

−
→ −
→
and assume that f is differentiable in A. Let v ∈ Rn \ { 0 } be such that
−
→ −
→
x + t v ∈ A for sufficiently small t ∈ R. Then the directional derivative of
−
→ −
→
f in the direction of v at the point x is given by
→ −
− →
∇f( x )• v .
199 Example Find the  directional

 derivative of f(x, y, z) = x3 + y3 − z2 in
1
−
→
the direction of v = 2.
3
Solution: We have
3x2
 
−
→
∇f( r ) =  3y2 
−2z
and so
−
→ − →
(∇f( r ))• v = 3x2 + 6y2 − 6z.
2.5 Extrema
We now turn to the problem of finding maxima and minima for vector func-
tions. As in the one-variable case, the derivative will provide us with infor-
mation about the extrema, and the “second derivative” will provide us with
information about the nature of these extreme points.
86 Chapter 2
To define an analogue for the second derivative, let us consider the

following. Let A ⊂ Rn and f : A → Rm be differentiable on A. We know
−
→ → (f) is a linear transformation from Rn to Rm. This
that for fixed x 0 ∈ A, D−
x0
means that we have a function
A → L (Rn, Rm)
T: →
− ,
x 7→ D−→ (f)
x
where L (Rn, Rm) denotes the space of linear transformations from Rn to

−
→
Rm. Hence, if we differentiate T at x 0 again, we obtain a linear transfor-
2 n n m
mation D−→ (T ) = D−
x0 x0
→ (f)) = D−
→ (D−
x0 → (f) from R to L (R , R ). Hence,
x0
−
→ −
→
given x1 ∈ Rn, we have D− 2 n m
→ (f)(x1 ) ∈ L (R , R ). Again, this means that
x0
−
→ −
→ −
→
given x 2 ∈ Rn, D−2 m
→ (f)(x1 ))( x 2) ∈ R . Thus the function
x0
Rn × Rn → L (Rn, Rm)
→ :
B− → −
− → 2 → −
− →
x0 (x1 , x2 ) 7→ D−
→ (f)(x1 , x2 )
x0
−
→ −
→
is well defined, and linear in each variable x1 and x2 , that is, it is a bilinear
function.
Just as the Jacobi matrix was a handy tool for finding a matrix repre-
sentation of D−→ (f) in the natural bases, when f maps into R, we have the
x
following analogue representation of the second derivative.
200 Theorem Let A ⊆ Rn be an open set, and f : A → R be twice differen-

2 n n
tiable on A. Then the matrix of D−→ (f) : R × R → R with respect to the
x
standard basis is given by the Hessian matrix:
∂2f − ∂2f − ∂2f −
 
→ → →
 ∂x1∂x1 ( x ) ∂x1∂x2 ( x ) · · · ∂x1∂xn ( x ) 
 ∂2f − ∂2f − ∂2f −
 
→ → → 
 (x) ( x ) ··· ( x )
H−→ f =  ∂x2∂x1 ∂x 2∂x 2 ∂x 2∂x n

x 
 .
.. .
.. .
.. .
..


 
 ∂2f 2
∂ f − 2
∂ f −
−
→ → →

(x) ( x ) ··· (x)
∂xn∂x1 ∂xn∂x2 ∂xn∂xn

 
x
−
→
f( r ) = f y = xy2z3.

z
Extrema 87
Then
2yz3 3y2z2
 
0
→ f =  2yz3
H− 2xz3 6xyz2
r
3y z 6xyz2 6xy2z
2 2
From the preceding example, we notice that the mixed partial deriva-
∂2f ∂2f ∂2f ∂2f ∂2f ∂2f
tives = , = , and = , are
∂x1∂x2 ∂x2∂x1 ∂x1∂x3 ∂x3∂x1 ∂x2∂x3 ∂x3∂x2
equal. This is no coincidence, as guaranteed by the following theorem.
202 Theorem Let A ⊆ Rn be an open set, and f : A → R be twice differ-

2 2
entiable on A. If D−→ (f) is continuous, then D−
→ (f) is symmetric, that is,
x0 x0
→ −
− →
∀(x1 , x2 ) ∈ Rn × Rn we have
2 → −
− → 2 → −
− →
→ (f)(x1 , x2 ) = D−
D− → (f)(x2 , x1 ).
x0 x0
We are now ready to study extrema in several variables. The basic

theorems resemble those of one-variable calculus. First, we make some
analogous definitions.
203 Definition Let A ⊆ Rn be an open set, and f : A → R. If there is some

−
→ −
→ −
→ −
→
open ball Bx0 (r) on which ∀ x ∈ Bx0 (r), f( x 0) ≥ f( x ), we say that f( x 0)
is a local maximum of f. Similarly, if there is some open ball Bx1 (r) on which
−
→ −
→ −
→ −
→
∀ x ∈ Bx0 (r 0 ), f( x 1) ≤ f( x ), we say that f( x 1) is a local maximum of
f. A point is called an extreme point if it is either a local minimum or local
−
→ −
→
maximum. A point t is called a critical point if f is differentiable at t and
D−→ (f) = 0. A critical point which is neither a maxima nor a minima is called
t
a saddle point.
204 Theorem Let A ⊆ Rn be an open set, and f : A → R be differen-

−
→ −
→
→ (f) = 0, that is, x 0 is a
tiable on A. If x 0 is an extreme point, then D− x0
critical point. Moreover, if f is twice-differentiable with continuous second
−
→
derivative and x0 is a critical point such that H− → f is negative definite, then
x0
−
→
f has a local maximum at x 0. If H− → f is positive definite, then f has a local
x0
−
→
minimum at x 0. If H− → f is indefinite, then f has a saddle point. If H−
x0
→ f is
x0
semi-definite (positive or negative), the test is inconclusive.
88 Chapter 2
205 Example Find the critical points of
R2 → R
f:
(x, y) 7→ x2 + xy + y2 + 2x + 3y
and investigate their nature.
Solution: We have

→ f = 2x + y + 2 x + 2y + 3 ,
J−
r
and so to find the critical points we solve
2x + y + 2 = 0,
x + 2y + 3 = 0,
1 4
which yields x = − , y = − . Now,
3 3

2 1
→f =
H− ,
r 1 2

2 1
which is positive definite, since ∆1 = 2 > 0 and ∆2 = det = 3 > 0.
1 2
1 4
Thus x0 = (− , − ) is a relative minimum and we have
3 3
7 1 4
− = f(− , − ) ≤ f(x, y) = x2 + xy + y2 + 2x + 3y.
3 3 3
206 Example Find the extrema of
R3 → R
f: 2 2 2 .
(x, y, z) 7→ x + y + 3z − xy + 2xz + yz
Solution: We have

→ f = 2x − y + 2z 2y − x + z 6z + 2x + y ,
J−
r
Extrema 89
which vanishes when x = y = z = 0. Now,

 
2 −1 2
H−→ f = −1 2 1 ,
r
2 1 6

2 −1
which is positive definite, since ∆1 = 2 > 0, ∆2 = det = 3 > 0,
  −1 2
2 −1 2
and ∆3 = det −1 2 1 = 4 > 0. Thus f has a relative minimum at
2 1 6
(0, 0, 0) and
0 = f(0, 0, 0) ≤ f(x, y, z) = x2 + y2 + 3z2 − xy + 2xz + yz.
R3 → R
207 Example Find the extrema of f : .
(x, y, z) 7→ x + y + z2 + xyz
2 2
Solution: We have

→ f = 2x + yz 2y + xz 2z + xy ,
J−
r
and  
2 z y
→ f =  z 2 x .
H−r
y x 2

2 z
We see that ∆1(x, y, z) = 2, ∆2(x, y, z) = det = 4−z2 and ∆3(x, y, z) =
z 2
det H−→ f = 8 − 2x2 − 2y2 − 2z2 + 2xyz.
r
If J−→ f = (0 0 0) then we must have
r
2x = −yz,
2y = −xz,
2z = −xy,
and upon multiplication of the three equations,
8xyz = −x2y2z2,
90 Chapter 2
that is,
xyz(xyz + 8) = 0.
Clearly, if xyz = 0, then we must have at least one of the variables equalling
0, in which case, by virtue of the original three equations, all equal 0. Thus
(0, 0, 0) is a critical point. If xyz = −8, then none of the variables is 0,
8
and solving for x, say, we must have x = − , and substituting this into
yz
2
2x + yz = 0 we gather (yz) = 16, meaning that either yz = 4, in which
case x = −2, or zy = −4, in which case x = 2. It is easy to see then that
either exactly one of the variables is negative, or all three are negative.
The other critical points are therefore (−2, 2, 2), (2, −2, 2), (2, 2, −2), and
(−2, −2, −2).
At (0, 0, 0), ∆1(0, 0, 0) = 2 > 0, ∆2(0, 0, 0) = 4 > 0, ∆1(0, 0, 0) = 8 > 0,
and thus (0, 0, 0) is a minimum point. If x2 = y2 = z2 = 4, xyz = −8, then
∆2(x, y, z) = 0, ∆3 = −32, so these points are saddle points.
R3 → R
208 Example Find the extrema of f : 2 2 .
(x, y, z) 7→ x y + y z + 2x − z
Solution: We have
→ f = 2xy + 2 x2 + 2yz y2 − 1 ,

J−
r
and  
2y 2x 0
H−→ f =  2x 2z 2y .
r
0 2y 0

2y 2x
We see that ∆1(x, y, z) = 2y, ∆2(x, y, z) = det = 4yz − 4x2 and
2x 2z
→ f = −8y3.
∆3(x, y, z) = det H−
r
If J−→ f = (0 0 0) then we must have
r
xy = −1,
x2 = −2yz,
y = ±1,
Extrema 91
1 1 1
and hence (1, −1, ), and (−1, 1, − ) are the critical points. Now, ∆1(1, −1, ) =
2 2 2
1 1 1
−2, ∆2(1, −1, ) = −6, and ∆3(1, −1, ) = 8. Thus (1, −1, ) is a saddle
2 2 2
1 1 1
point. Similarly, ∆1(−1, 1, − ) = 2, ∆2(−1, 1, − ) = −6, and ∆3(−1, 1, − ) =
2 2 2
1
−8, shewing that (1, −1, ) is also a saddle point.
2
209 Example Determine the nature of the critical points of
f(x, y, z) = 4xyz − x4 − y4 − z4.
Solution: We find
4yz − 4x3
 
∇f(x, y, z) = 4xz − 4y3 .

4xy − 4z3
Assume ∇f(x, y, z) = 0. Then
4yz = 4x3, 4xz = 4y3, 4xy = 4z3 =⇒ xyz = x4 = y4 = z4.
Thus xyz ≥ 0, and if one of the variables is 0 so are the other two. Thus
(0, 0, 0) is the only critical point with at least one of the variables 0. Assume
now that xyz 6= 0. Then
1
(xyz)3 = x4y4z4 = (xyz)4 =⇒ xyz = 1 =⇒ yz = =⇒ x4 = 1 =⇒ x = ±1.
x
Similarly, y = ±1, z = ±1. Since xyz = 1, exactly two of the variables can
be negative. Thus we find the following critical points:
(0, 0, 0), (1, 1, 1), (−1, −1, 1), (−1, 1, −1), (1, −1, −1).
The Hessian is
−12x2
 
4z 4y
→ f =  4z
H−
x −12y2 4x  .
4y 4x −12z2
If 1 = xyz = x2 = y2 = z2, we have ∆1 = −12x2 = −12 < 0, ∆2 =
144x2y2 − 16z2 = 144 − 16 = 128 > 0, and
∆3 = −1728x2y2z2 + 192x4 + 192z4 + 128zyx + 192y4
= −1728 + 192 + 192 + 128 + 192
= −1024
< 0.
92 Chapter 2
This means that for xyz 6= 0 the Hessian is negative definite and the func-
tion has a local maximum at each of the four points (1, 1, 1), (−1, −1, 1),
(−1, 1, −1), (1, −1, −1). Observe that at these critical points f = 1. Now
f(0, 0, 0) = 0 and f(−1, 1, 1) = −7.
210 Example Determine the nature of the critical points of

2 −y2 −z2
g(x, y, z) = xyze−x .
2 2 2
Solution: To facilitate differentiation observe that g(x, y, z) = (xe−x )(ye−y )(ze−z ).
Now  
2 2 2
(1 − 2x2)(yz)(e−x )(e−y )(e−z )
2 2 2 
∇g(x, y, z) = (1 − 2y2)(xz)(e−x )(e−y )(e−z ) .

2 2 2
(1 − 2z2)(xy)(e−x )(e−y )(e−z )
The function is 0 if any of the variables is 0. Since the function clearly
assumes positive and negative values, we can discard any point with a 0.
1 1 1
If ∇(x, y, z) = 0, then x = ± √ ; y = ± √ z = ± √ . We find
2 2 2
(4x3 − 6x)(yz) (1 − 2x2)(1 − 2y2)z (1 − 2x2)(1 − 2z2)y
 
H− → g = t(x, y, z) (1 − 2y2)(1 − 2x2)z (4y3 − 6y)(xz) (1 − 2y2)(1 − 2z2)x ,

x
(1 − 2z2)(1 − 2x2)y (1 − 2z2)(1 − 2y2)x (4z3 − 6z)(xy)
2 2 2
with t(x, y, z) = (e−x )(e−y )(e−z ). Since at the critical points we have
1 − 2x2 = 1 − 2y2 = 1 − 2z2 = 0, the Hessian reduces to
(4x3 − 6x)(yz)
 
0 0
→ g = (e−3/2) 
H− 0 (4y3 − 6y)(xz) 0 .
x
3
0 0 (4z − 6z)(xy)
We have
∆1 = (4x3 − 6x)(yz)
∆2 = (4x3 − 6x)(4y3 − 6y)(xyz2)
∆3 = (4x3 − 6x)(4y3 − 6y)(4z3 − 6z)(x2y2z2).
Also,
√ √
3 3
1 1 1 1
4 √ −6 √ = −2 2 < 0, 4 −√ − 6 −√ = 2 2 > 0.
2 2 2 2
Lagrange Multipliers 93
This means that if an even number of the variables is negative (0 or 2),

then we the Hessian is negative definite, and if an odd numbers of the
variables is positive (1 or 3), the Hessian is positive definite. We conclude
that we have local maxima at
1 1 1 1 1 1 1 1 1 1 1 1
( √ , √ , √ ), (− √ , − √ , √ ), (− √ , √ , − √ ), ( √ , − √ , − √ )
2 2 2 2 2 2 2 2 2 2 2 2
and local minima at
1 1 1 1 1 1 1 1 1 1 1 1
(− √ , − √ , − √ ), (− √ , √ , √ ), ( √ , − √ , √ ), ( √ , √ , − √ ).
2 2 2 2 2 2 2 2 2 2 2 2
2.6 Lagrange Multipliers

In some situations we wish to optimise a function given a set of constraints.
For such cases, we have the following.
211 Theorem Let A ⊆ Rn and let f : A → R, g : A → R be functions

whose respective derivatives are continuous. Let g(x0) = c0 and let S =
g−1(c0) be the level set for g with value c0, and assume ∇g(x0) 6= 0. If the
restriction of f to S has an extreme point at x0, then ∃λ ∈ R such that
∇f(x0) = λ∇g(x0).
! The above theorem only locates extrema, it does not say anything
concerning the nature of the critical points found.
212 Example Optimise f : R2 → R, f(x, y) = x2 − y2 given that x2 + y2 = 1.
Solution: Let g(x, y) = x2 + y2 − 1. We solve

x x
∇f = λ∇g
y y
for x, y, λ. This requires

2x 2xλ
= .
−2y 2yλ
94 Chapter 2
From 2x = 2xλ we get either x = 0 or λ = 1. If x = 0 then y = ±1 and

λ = −1. If λ = 1, then y = 0, x = ±1. Thus the potential critical points are
(±1, 0) and (0, ±1). If x2 + y2 = 1 then
f(x, y) = x2 − (1 − x2) = 2x2 − 1 ≥ −1,
and
f(x, y) = 1 − y2 − y2 = 1 − 2y2 ≤ 1.
Thus (±1, 0) are maximum points and (0, ±1) are minimum points.
213 Example Find the maximum and the minimum points of f(x, y) = 4x +
3y, subject to the constraint x2 + 4y2 = 4, using Lagrange multipliers.
Solution: Putting g(x, y) = x2 + 4y2 − 4 we have

4 2x
∇f(x, y) = λ∇g(x, y) =⇒ =λ .
3 8y
x 16
Thus 4 = 2λx, 3 = 8λy. Clearly then λ 6= 0. Upon division we find = .
y 3
Hence
256 2 3 16
x2 + 4y2 = 4 =⇒ y + 4y2 = 4 =⇒ y = ± √ , x = ± √
9 73 73.
The maximum is clearly then
√

16 3
4 √ +3 √ = 73,
73 73
√
and the minimum is − 73.
214 Example Let a > 0, b > 0, c > 0. Determine the maximum and mini-
x y z x2 y2 z2
mum values of f(x, y, z) = + + on the ellipsoid 2 + 2 + 2 = 1.
a b c a b c
x2 y2 z2
Solution: We use Lagrange multipliers. Put g(x, y, z) = 2 + 2 + 2 − 1.
a b c
Then
Lagrange Multipliers 95
2x/a2
   
1/a
∇f(x, y, z) = λ∇g(x, y, z) ⇐⇒ 1/b = λ 2y/b2 .
1/c 2z/c2
a b c x2 y2 z2
It follows that λ 6= 0. Hence x = ,y = ,z = . Since 2 + 2 + 2 = 1,
√2λ 2λ 2λ a b c
3 3
we deduce 2 = 1 or λ = ± . Since a, b, c are positive, f will have a
4λ 2
maximum when all x, y, z are positive and a minimum when all x, y, z are
negative. Thus the maximum is when
a b c
x = √ ,y = √ ,z = √ ,
3 3 3
and
3
f(x, y, z) ≤ √
3
and the minimum is when
a b c
x = −√ , y = −√ , z = −√ ,
3 3 3
and
3
f(x, y, z) ≥ − √ .
3
215 Example Let a > 0, b > 0, p > 1. Maximise f(x, y) = ax + by subject

to the constraint xp + yp = 1.
Solution: Put g(x, y) = xp + yp − 1. We need a = pλxp−1 and b = pλyp−1.

Clearly then , λ 6= 0. We then have
1/(p−1) 1/(p−1)
a b
x= , y= .
λp λp
Thus
p/(p−1) p/(p−1)
p p a b
1=x +y = + ,
λp λp
96 Chapter 2
which gives
p/(p−1) p/(p−1)!(p−1)/p
a b
λ= + .
p p
This gives
a1/(p−1) b1/(p−1)
x= , y = .
(a1/(p−1) + b1/(p−1))1/p (a1/(p−1) + b1/(p−1))1/p
Since f is non-negative, these points define a maximum for f and so
ap/(p−1) bp/(p−1)
ax + by ≤ + .
(a1/(p−1) + b1/(p−1))1/p (a1/(p−1) + b1/(p−1))1/p
216 Example Find the extrema of f(x, y, z) = x2 + y2 + z2 subject to the

constraint (x − 1)2 + (y − 2)2 + (z − 3)2 = 4.
Solution: Let g(x, y, z) = (x − 1)2 + (y − 2)2 + (z − 3)2 − 4. We solve

   
x x
∇f y = λ∇g y
  
z z
for x, y, λ. This requires

   
2x 2(x − 1)λ
2y = 2(y − 2)λ .
2z 2(z − 3)λ
−λ −2λ −3λ
Clearly, λ 6= 1. This gives x = ,y= , and z = . Substitut-
1−λ 1−λ 1−λ
ing into (x − 1)2 + (y − 2)2 + (z − 3)2 = 4, we gather that
2 2 2
−λ −2λ −3λ
−1 + −2 + − 3 = 4,
1−λ 1−λ 1−λ
from where √
14
λ=1± .
2
Arithmetic Mean-Geometric Mean Inequality 97
This gives the two points

2 4 6
(x, y, z) = 1 + √ ,2 + √ ,3 + √
14 14 14
and
2 4 6
(x, y, z) = 1 − √ , 2 − √ , 3 − √ .
14 14 14
√
12 14
The first point gives an absolute maximum of 18 + and the second
√ 7
12 14
an absolute minimum of 18 − .
7
217 Example Optimise f(x, y, z) = x + y + z subject to x2 + y2 = 2, and

x + z = 1.
Solution: Put g(x, y, z) = x2 + y2 − 2, h(x, y, z) = x + z − 1. We must find

λ, δ such that
∇f(x, y, z) = λ∇g(x, y, z) + δ∇h(x, y, z),
which translates into
1 = 2λx + δ,
1 = 2λy,
1 = δ,
and
x2 + y2 = 1,
x + z = 1.
√ √
We deduce that x = 0, y =√± 2, z = 1. We may shew that (0, 2, 1) yields
a maximum and that (0, − 2, 1) yields a minimum.
2.7 Arithmetic Mean-Geometric Mean Inequality

218 Definition Let a1, a2, · · · an be n non-negative real numbers. Their
arithmetic mean or average is
a1 + a2 + · · · + a n
.
n
98 Chapter 2
Their geometric mean is

(a1a2 · · · an)1/n.
The Arithmetic-Mean-Geometric-Mean Inequality (AMGM) asserts that
a1 + a2 + · · · + a n
(a1a2 · · · an)1/n ≤ ,
n
with equality if and only if
a1 = a2 = · · · = an.
We will see that this simple inequality allows us to solve many extremum
problems without using calculus. We will give two proofs of AMGM, one
using Lagrange Multipliers, and one using one-variable calculus due to
George Pólya.
219 Theorem Let a1, a2, · · · an be n non-negative real numbers. Then
a1 + a2 + · · · + a n
(a1a2 · · · an)1/n ≤ ,
n
with equality if and only if
a1 = a2 = · · · = an.
Proof (Using Lagrange Multipliers) Let
f(x) = f(x1, x2, . . . , xn) = x1x2 · · · xn,
g(x) = g(x1, x2, . . . , xn) = x1 + x2 + · · · + xn,

with x1 ≥ 0, x2 ≥ 0, . . . , xn ≥ 0 and x1 + x2 + · · · + xn = S. Observe that if
any of the xi = 0, then f vanishes. So we may assume that no xi vanishes.
Observe that
f(x)
∀i 1 ≤ i ≤ n, ∇f(x) = λ∇g(x) =⇒ = x1x2 · · · xi−1xi+1 · · · xn = λ.
xi
f(x)
Since f(x) 6= 0, the equality = λ imposes x1 = x2 = · · · = xn. This
xi
S
means that xi = . Since f is non-negative on the stipulated conditions,
n
S
the extrema at xi = is a maximum, and so
n
n n
S x1 + x2 + · · · + xn
x1x2 · · · xn ≤ = ,
n n
which gives Arithmetic-Mean-Geometric-Mean Inequality. ❑
For Pólya’s proof we need the following lemma.
220 Lemma For all real numbers x it is verified that
x ≤ ex−1,
with equality only when x = 1.
Proof Put f : R → R, f(x) = ex−1 − x. Clearly f(1) = e0 − 1 = 0. Now,
f 0 (x) = ex−1 − 1,
f 00 (x) = ex−1.
If f 0 (x) = 0 then ex−1 = 1 implying that x = 1. Thus f has a single minimum
point at x = 1. Thus for all real numbers x
0 = f(1) ≤ f(x) = ex−1 − x,
which gives the desired result. ❑
Proof (George Pólya) Put

nak
Ak = ,
a1 + a2 + · · · + a n
and Gn = a1a2 · · · an. Observe that
nnGn
A1A2 · · · An = ,
(a1 + a2 + · · · + an)n
100 Chapter 2
and that
A1 + A2 + · · · + An = n.
By the preceding lemma, we have
A1 ≤ exp(A1 − 1),
A2 ≤ exp(A2 − 1),
..
.
An ≤ exp(An − 1).
Since all the quantities involved are non-negative, we may multiply all
these inequalities together, to obtain,
A1A2 · · · An ≤ exp(A1 + A2 + · · · + An − n).
In view of the observations above, the preceding inequality is equivalent to
nnGn
≤ exp(n − n) = e0 = 1.
(a1 + a2 + · · · + an)n
We deduce that n
a1 + a2 + · · · + a n
Gn ≤ ,
n
which is equivalent to
a1 + a2 + · · · + a n
(a1a2 · · · an)1/n ≤ .
n
Now, for equality to occur, we need each of the inequalities Ak ≤ exp(Ak −

1) to hold. This occurs, in view of the preceding lemma, if and only if
Ak = 1, ∀k, which translates into a1 = a2 = . . . = an. This completes the
proof. ❑
221 Example Let f(x) = (a + x)5(a − x)3, x ∈ [−a; a]. Find the maximum
value of f by means of AMGM.
Solution: If x ∈ [−a; a], then a + x ≥ 0 and a − x ≥ 0 thus we may use

a+x a−x
AMGM with n = 8, a1 = a2 = · · · = a5 = and a6 = a7 = a8 = .
5 3
We then deduce
 8
a+x a−x
5 3
5 +3
a+x a−x 5 3  a 8
≤
 = 4 ,

5 3 8
whence
5533a8
f(x) ≤
,
48
a+x a−x
and equality holds if and only if = .
5 3
222 Example For any positive integer n > 1 we have
1 · 3 · 5 · · · · (2n − 1) < nn.
For, by AMGM,
n n
n2

1 + 3 + 5 + · · · + (2n − 1)
1 · 3 · 5 · · · · (2n − 1) < = = nn.
n n
Notice that since the factors are unequal we have strict inequality.
n
1
223 Example The sequence xn = 1+ , n = 1, 2, . . . is increasing.
n
For the set of n + 1 numbers
1 1 1
1, 1 + ,1 + ,...,1 + ,
n n n
has arithmetic mean
1
1+
n+1
and geometric mean
n/(n+1)
1
1+ .
n
102 Chapter 2
Therefore n/(n+1)
1 1
1+ > 1+ ,
n+1 n
that is to say
n+1 n
1 1
1+ > 1+ ,
n+1 n
or
xn+1 > xn,
giving the result.
224 Example Find the volume of the largest rectangular box with sides par-
allel to the coordinate planes which can be inscribed in the ellipsoid
x 2 y 2 z 2
+ + = 1.
a b c
Here a > 0, b > 0, c > 0.
Solution: Let 2x, 2y, 2z be the dimensions of this box. We seek to maximise
x2 y2 z2
8xyz. Putting n = 3, x1 = 2 , x2 = 2 , x3 = 2 , we have
a b c
3
x2 y2 z2

x1 + x2 + x3 1
2 2 2
= x1x2x3 ≤ = .
a b c 3 27
Thus
8abc
8xyz ≤ √
3 3
225 Definition Let a1 > 0, a2 > 0, . . . , an > 0. Their harmonic mean is

given by
n
.
1 1 1
+ + ··· +
a1 a2 an
As a corollary to AMGM we obtain

226 Corollary (Harmonic Mean-Geometric Mean Inequality) Let b1 > 0, b2 >

0, . . . , bn > 0. Then
n
≤ (b1b2 · · · bn)1/n.
1 1 1
+ + ··· +
b1 b2 bn
1
Proof This follows by putting ak = in Theorem 219. For then
bk
1 1 1
1/n + + ··· +
1 1 1 b b2 bn
··· ≤ 1 .
b1 b2 bn n
Combining Theorem 219 and Corollary 226, we deduce
227 Corollary (Harmonic Mean-Arithmetic Mean Inequality) Let b1 > 0, b2 >

0, . . . , bn > 0. Then
n b1 + b2 + · · · + bn
≤ .
1 1 1 n
+ + ··· +
b1 b2 bn
228 Example Let ak > 0, and s = a1 + a2 + · · · + an. Prove that

n
X s n2
≥
k=1
s − ak n−1
and
n
X ak n
≥ .
k=1
s − a k n − 1
s
Solution: Put bk = . Then
s − ak
Xn Xn
1 s − ak
= =n−1
k=1
bk
k=1
s
104 Chapter 2
and from Corollary 227,

Pn s
n k=1
s − ak
≤ ,
n−1 n
s ak
from where the first inequality is proved. Since −1 = ,we
s − ak s − ak
Xn Xn
ak s
= −1
k=1
s − ak k=1
s − ak
Xn
s
= −n
have k=1
s − a k
n2
≥ −n
n−1
n
= .
n−1
Chapter 3
Integration
3.1 Differential Forms

We will now consider integration in several variables. In order to smooth
our discussion, we need to consider the concept of differential forms.
229 Definition Consider n variables
x1, x2, . . . , xn
in n-dimensional space (used as the names of the axes), and let

 
a1j
 a2j 
aj =  ..  ∈ Rn, 1 ≤ j ≤ k,
 
 . 
anj
be k ≤ n vectors in Rn. Moreover, let {j1, j2, . . . , jk} ⊆ {1, 2, . . . , n} be

a collection of k sub-indices. An elementary k-differential form (k > 1)
acting on the vectors aj, 1 ≤ j ≤ k is defined and denoted by
 
aj1 1 aj1 2 · · · aj1 k
aj 1 aj 2 · · · aj k
 2 2 2 
dxj1 ∧ dxj2 ∧ · · · ∧ dxjk (a1, a2 , . . . , ak ) = det  .. .. ..  .
 . . ··· . 
ajk 1 ajk 2 · · · ajk k
105
106 Chapter 3
In other words, dxj1 ∧ dxj2 ∧ · · · ∧ dxjk (a1, a2 , . . . , ak ) is the xj1 xj2 . . . xjk
component of the signed k-volume of a k-parallelotope in Rn spanned by
a1, a2 , . . . , ak .
! By virtue of being a determinant, the wedge product ∧ of differential

forms has the following properties
➊ anti-commutativity: da ∧ db = −db ∧ da.
➋ linearity: d(a + b) = da + db.
➌ scalar homogeneity: if λ ∈ R, then dλa = λda.
➍ associativity: (da ∧ db) ∧ dc = da ∧ (db ∧ dc).1
! Anti-commutativity yields
da ∧ da = 0.
230 Example Consider  

1
a =  0  ∈ R3.
−1
Then
dx(a) = det(1) = 1,
dy(a) = det(0) = 0,
dz(a) = det(−1) = −1,
are the (signed) 1-volumes (that is, the length) of the projections of a onto
the co-ördinate axes.
231 Example Consider the vectors in R4

   
1 −1
2 0
a= 3 , b =  1  .
  
4 2
1
Notice that associativity does not hold for the wedge product of vectors.
Differential Forms 107
Then, for example,

 
1 −1
2 0  1 −1
3 1  = det 2 0 = 2,
dx1 ∧ dx2(a, b) = d  
4 2
 
1 −1
2 0  2 0
dx2 ∧ dx4(a, b) = d 
  = det = 4,
3 1 4 2
4 2
 
1 −1
2 0  3 1
dx3 ∧ dx1(a, b) = d 
  = det = −4.
3 1 1 −1
4 2
232 Example Consider the vectors in R4

     
1 −1 1
2 0 0
a= 3 , b =  1  , c = 0 .
    
4 2 0
Then, for example,

dx1 ∧ dx2(a, b, c)
is undefined, since for a 2-form we need two vectors. We have, however,
 
1 −1 1  
2 0 0 1 −1 1
dx1 ∧ dx2 ∧ dx4(a, b, c) = d 
3 1 0 = det 2 0 0 = 4.
  
4 2 0
4 2 0
233 Example In R3 we have dx ∧ dy ∧ dx = 0, since we have a repeated

variable.
234 Example In R3 we have
dx ∧ dz + 5dz ∧ dx + 4dx ∧ dy − dy ∧ dx + 12dx ∧ dx = −4dx ∧ dz + 5dx ∧ dy.

108 Chapter 3
! In order to avoid redundancy we will make the convention that if a sum

of two or more terms have the same differential form up to permutation of
the variables, we will simplify the summands and express the other differ-
ential forms in terms of the one differential form whose indices appear in
increasing order.
235 Example In R5 we have
dx5 ∧ dx4 ∧ dx2 = −dx4 ∧ dx5 ∧ dx2 = dx4 ∧ dx2 ∧ dx5 = −dx2 ∧ dx4 ∧ dx5,
dx4 ∧ dx2 ∧ dx5 = −dx2 ∧ dx4 ∧ dx5

dx2 ∧ dx5 ∧ dx4 = −dx2 ∧ dx4 ∧ dx5.
Hence we write
dx2 ∧ dx5 ∧ dx4 + 5dx4 ∧ dx2 ∧ dx5 + 2dx5 ∧ dx4 ∧ dx2
as
−8dx2 ∧ dx4 ∧ dx5.
236 Definition A 0-differential form in Rn is simply a differentiable function

in Rn.
237 Definition A k-differential form field in Rn is an expression of the form

X
ω= aj1 j2 ...jk dxj1 ∧ dxj2 ∧ · · · dxjk ,
1≤j1 ≤j2 ≤···≤jk ≤n
where the aj1 j2 ...jk are differentiable functions in Rn.
238 Example
g(x, y, z, w) = x + y2 + z3 + w4
is a 0-form in R4.
239 Example An example of a 1-form field in R3 is
ω = xdx + y2dy + xyz3dz.

Differential Forms 109

ω = x2dx ∧ dy + y2dy ∧ dz + dz ∧ dx.

ω = (x + y + z)dx ∧ dy ∧ dz.
242 Theorem If ω is an l-form and φ is a k-form then ω ∧ φ is a l + k-form

and
ω ∧ φ = (−1)lkφ ∧ ω.
243 Example The product of the 1-form fields in R3

ω1 = ydx + xdy,
ω2 = −2xdx + 2ydy,
is
ω1 ∧ ω2 = (2x2 + 2y2)dx ∧ dy.
244 Theorem Let 0 ≤ k ≤ n, and let ∧k(Rn) denote the space of k-

differential forms in Rn. Then ∧k(Rn) is a vector space of dimension

n n!
= .
k k!(n − k)!

n
A basis for this space is given by the elementary forms
k
dxj1 ∧ dxj2 ∧ · · · dxjn , 1 ≤ j1 < j2 < · · · < jk ≤ n.

3 1 3 3
245 Example In R , ∧ (R ) has dimension = 3, and a basis for ∧1(R3)
1
is
{dx, dy, dz}.

3 2 3 3
2
is
{dx ∧ dy, dx ∧ dz, dy ∧ dz}.
110 Chapter 3

3 3 3 3
3
is
{dx ∧ dy ∧ dz}.

4 2 4 4
2
is
{dx1 ∧ dx2, dx1 ∧ dx3, dx1 ∧ dx4, dx2 ∧ dx3, dx2 ∧ dx4, dx3 ∧ dx4}.
249 Definition Let f(x1, x2, . . . , xn) be a 0-form in Rn. The exterior deriva-
tive df of f is
Xn
∂f
df = dxi.
i=1
∂x i
Furthermore, if
ω = f(x1, x2, . . . , xn)dxj1 ∧ dxj2 ∧ · · · ∧ dxjk
is a k-form in Rn, the exterior derivative dω of ω is the (k + 1)-form
dω = df(x1, x2, . . . , xn) ∧ dxj1 ∧ dxj2 ∧ · · · ∧ dxjk .
250 Example If in R2, ω = x3y4, then
d(x3y4) = 3x2y4dx + 4x3y3dy.
251 Example If in R2, ω = x2ydx + x3y4dy then
dω = d(x2ydx + x3y4dy)
= (2xydx + x2dy) ∧ dx + (3x2y4dx + 4x3y3dy) ∧ dy
= x2dy ∧ dx + 3x2y4dx ∧ dy
= (3x2y4 − x2)dx ∧ dy
252 Example Consider the change of variables x = u + v, y = uv. Then
dx = du + dv,
Integrating in ∧0(Rn) 111
dy = vdu + udv,
whence
dx ∧ dy = (u − v)du ∧ dv.
253 Example Consider the transformation of co-ördinates xyz into uvw co-
ördinates given by
z y+z
u = x + y + z, v = , w= .
y+z x+y+z
Then
du = dx + dy + dz,
z y
dv = − 2
dy + dz,
(y + z) (y + z)2
y+z x x
dw = − 2
dx + 2
dy + dz.
(x + y + z) (x + y + z) (x + y + z)2
Multiplication gives

zx y(y + z)
du ∧ dv ∧ dw = − −
(y + z)2(x + y + z)2 (y + z)2(x + y + z)2
z(y + z) xy
+ 2 2
− dx ∧ dy ∧ dz
(y + z) (x + y + z) (y + z) (x + y + z)2
2
z2 − y2 − zx − xy
= dx ∧ dy ∧ dz.
(y + z)2(x + y + z)2
3.2 Integrating in ∧0(Rn)

254 Definition A 0-dimensional oriented manifold of Rn is simply a point
x ∈ Rn, with a choice of the + or − sign. A general oriented 0-manifold is
a union of oriented points.
255 Definition Let M = +{b} ∪ −{a} be an oriented 0-manifold, and let ω

be a 0-form. Then Z
ω = ω(b) − ω(a).
M
112 Chapter 3
! −x has opposite orientation to +x and

Z Z
ω=− ω.
−x +x
256 Example Let M = −{(1, 0, 0)} ∪ +{(1, 2, 3)} ∪ −{(0, −2, 0)}2 be an ori-
ented 0-manifold, and let ω = x + 2y + z2. Then
Z
ω = −ω((1, 0, 0)) + ω(1, 2, 3) − ω(0, 0, 3) = −(1) + (14) − (−4) = 17.
M
3.3 Integrating in ∧1(Rn)

257 Definition A 1-dimensional oriented manifold of Rn is simply an ori-
ented smooth curve Γ ∈ Rn, with a choice of a + orientation if the curve
traverses in the direction of increasing t, or with a choice of a − sign if
the curve traverses in the direction of decreasing t. A general oriented
1-manifold is a union of oriented curves.
! The curve −Γ has opposite orientation to Γ and

Z Z
ω=− ω.
−Γ Γ
We now turn to the problem of integrating 1-forms.
258 Example Calculate

Z
xydx + (x + y)dy
Γ
where Γ is the parabola y = x2, x ∈ [−1; 2] oriented in the positive direc-

tion.
2
Do not confuse, say, −{(1, 0, 0)} with −(1, 0, 0) = (−1, 0, 0). The first one means that
the point (1, 0, 0) is given negative orientation, the second means that (−1, 0, 0) is the
additive inverse of (1, 0, 0).
Solution: We parametrise the curve as x = t, y = t2. Then

xydx + (x + y)dy = t3dt + (t + t2)dt2 = (3t3 + 2t2)dt,
whence Z Z2
ω = (3t3 + 2t2)dt
Γ −1 2
2 3 3 4
= t + t
3 4 −1
69
= .
4
What would happen if we had given the curve above a different parametri-
sation? First observe that the curve travels from (−1, 1) to (2, 4) on the
parabola
√ y = x2√. These conditions are met with the parametrisation
x = t − 1, y = ( t − 1)2, t ∈ [0; 9]. Then
√ √ √ √ √
xydx + (x + y)dy = ( t√− 1)3d( t −√1) + (( t −√1) + ( t − 1)2)d( t − 1)2
= (3( t − 1)3 + 2( t − 1)2)d( t − 1)
1 √ √
= √ (3( t − 1)3 + 2( t − 1)2)dt,
2 t
whence Z Z9
1 √ √
ω = √ (3( t − 1)3 + 2( t − 1)2)dt
Γ
0 22 t 3/2 9
3t 7t 5t √
= − + − t
4 3 2 0
69
= ,
4
as before.
! It turns out that if two different parametrisations of the same curve

have the same orientation, then their integrals are equal. Hence, we only
need to worry about finding a suitable parametrisation.
259 Example Calculate the line integral

Z
y sin xdx + x cos ydy,
Γ
where Γ is the line segment from (0, 0) to (1, 1) in the positive direction.
114 Chapter 3
Solution: This line has equation y = x, so we choose the parametrisation

x = y = t. The integral is thus
Z Z1
y sin xdx + x cos ydy = (t sin t + t cos t)dt
Γ 0 Z1
1
= [t(sin x − cos t)]0 − (sin t − cos t)dt
0
= 2 sin 1 − 1,
upon integrating by parts.
260 Example Calculate the path integral

Z
x+y x−y
2 2
dy + 2 dx
Γ x +y x + y2
around the closed square Γ = ABCD with A = (1, 1), B = (−1, 1), C =
(−1, −1), and D = (1, −1) in the direction ABCDA.
Solution: On AB, y = 1, dy = 0, on BC, x = −1, dx = 0, on CD, y =

−1, dy = 0, and on DA, x = 1, dx = 0. The integral is thus
Z Z Z Z Z
ω = ω+ ω+ ω+ ω
Γ ZAB BC CD
Z −1 DA Z1 Z1
−1
x−1 y−1 x+1 y+1
= 2
dx + 2
dy + 2
dx + 2
dy
1 x +1 1 y +1 −1 x + 1 −1 y + 1
Z1
1
= 4 2
dx
−1 x + 1
= 4 arctan x|1−1
= 2π.
! When the integral is along a closed Ipath, like in theZ preceding ex-
ample, it is customary to use the symbol rather than . The positive
Γ Γ
direction of integration is that sense that when traversing the path, the area
enclosed by the curve is to the left of the curve.

I
x2dy + y2dx,
Γ
where Γ is the ellipse 9x2 + 4y2 = 36 traversed once in the positive sense.
Solution: Parametrise the ellipse as x = 2 cos t, y = 3 sin t, t ∈ [0; 2π].

Observe that when traversing this closed curve, the area of the ellipse
is on the left hand side of the path, so this parametrisation traverses the
curve in the positive sense. We have
I Z 2π
ω = ((4 cos2 t)(3 cos t) + (9 sin t)(−2 sin t))dt
Γ Z02π
= (12 cos3 t − 18 sin3 t)dt
0
= 0.
I
262 Example Find zdx + xdy + ydz where Γ is the intersection of the
Γ
sphere x2 + y2 + z2 = 1 and the plane x + y = 1, traversed in the positive
direction.
Solution: The curve lies on the sphere, and to parametrise this curve,
we dispose of one of the variables, y say, from where y = 1 − x and
x2 + y2 + z2 = 1 give
x2 + (1 − x)2 + z2 = 1 =⇒ 2x2 − 2x + z2 = 0
2
1 1
=⇒ 2 x − + z2 =
2 2
2
1
=⇒ 4 x − + 2z2 = 1.
2
So we now put
1 cos t sin t 1 cos t
x= + , z= √ , y=1−x= − .
2 2 2 2 2
116 Chapter 3
We must integrate on the side of the plane

  that can be viewed from the
1
 
 
point (1, 1, 0) (observe that the vector  1 is normal to the plane). On

 
 
0
2
1
the zx-plane, 4 x − + 2z2 = 1 is an ellipse. To obtain a positive
2
parametrisation we must integrate from t = 2π to t = 0 (this is because
when you look at the ellipse from the point (1, 1, 0) the positive x-axis is to
your left, and not your right). Thus
I Z0
sin t 1 cos t
zdx + xdy + ydz = √ d +
2 2
Γ
Z 0 2
2π
1 cos t 1 cos t
+ + d −
2π 2 2 2 2
Z0
1 cos t sin t
+ − d √
2 2 2
Z 0 2π

sin t cos t cos t sin t 1
= + √ + − √ dt
2π 4 2 2 4 2 2
π
= √ .
2
Given a curve Γ how can we find its length? The idea, as seen in figure
3.1 is to consider the projections dxi, 1 ≤ i ≤ n at each point. The length
of the vector  
 dx1 
 
 
 dx 
 2
dx = 
 . 

 .. 
 
 
 
dxn
is p
||dx|| = (dx1)2 + (dx2)2 + · · · + (dxn)2.
Hence the length of Γ is given by

Z Z p
||dx|| = (dx1)2 + (dx2)2 + · · · + (dxn)2. (3.1)
Γ Γ
Similarly, suppose that Γ is a simple closed curve in R2. How do we

find the (oriented) area of the region it encloses? The idea, borrowed from
finding areas of polygons, is to split the region into triangles, each of area
 
1  x dx  1
det   = (xdy − ydx),
2   2
y dy
and to sum over the closed curve, obtaining a total oriented area of
 
I I
1  x dx  1
=
2 Γ
det 
  2 (xdy − ydx). (3.2)
Γ
y dy
I
Here denotes integration around the closed curve.
Γ
dx
Figure 3.2: Area enclosed by a

Figure 3.1: Length of a curve.
simple closed curve
118 Chapter 3
263 Example Let (A, B) ∈ R2, A > 0, B > 0. Find a parametrisation of the
ellipse
x2 y2
Γ : {(x, y) ∈ R2 : 2 + 2 = 1}.
A B
Furthermore, find an integral expression for the perimeter of this ellipse
and find the area it encloses.
Solution: Consider the parametrisation Γ : [0; 2π] → R2, with

   
 x  A cos t
 = .
   
y B sin t
This is a parametrisation of the ellipse, for

x2 y2 A2 cos2 t B2 sin2 t
2
+ 2
= 2
+ 2
= cos2 t + sin2 t = 1.
A B A B
Notice that this parametrisation goes around once the ellipse counterclock-
wise. The perimeter of the ellipse is given by
Z Z 2π p
||dx|| = A2 sin2 t + B2 cos2 t dt.
Γ 0
The above integral is an elliptic integral and quite hard to evaluate. We will
have better luck with the area of the ellipse, which is given by
I I
1 1
(xdy − ydx) = (A cos t d(B sin t) − B sin t d(A cos t))
2 Γ 2 Z
1 2π
= (AB cos2 t + AB sin2 t)) dt
2 0 Z
1 2π
= AB dt
2 0
= πAB.
264 Example Find a parametric representation for the astroid
Γ : {(x, y) ∈ R2 : x2/3 + y2/3 = 1},
in figure 3.3. Find the perimeter of the astroid and the area it encloses.
Solution: Take
   
3
 x  cos t
 = 
   
3
y sin t
with t ∈ [0; 2π] → R2. with Then
x2/3 + y2/3 = cos2 t + sin2 t = 1.
The perimeter of the astroid is
Z Z 2π p
||dx|| = 9 cos4 t sin2 t + 9 sin4 t cos2 t dt
Γ 0 Z 2π
= 3| sin t cos t| dt
0 Z
3 2π
= | sin 2t| dt
2 Z0
π/2
= 6 sin 2t dt
0
= 6.

120 Chapter 3
The area of the astroid is given by

I I
1 1
(xdy − ydx) = (cos3 t d(sin3 t) − sin3 t d(cos3 t))
2 Γ 2Z
1 2π
= (3 cos4 t sin2 t + 3 sin4 t cos2 t)) dt
2 0 Z
3 2π
= (sin t cos t)2 dt
2 0Z
3 2π
= (sin 2t)2 dt
8Z 0
3 2π
= (1 − cos 4t) dt
16 0
3π
= .
8
265 Example Find the area enclosed by the curve x = sin3 t, y = (cos t)(1+
sin2 t).
Solution: Observe that the parametrisation traverses the curve once clock-
wise if t ∈ [0; 2π]. The area is given by
 
I I
1  x dx  1
det   = 2 xdy − ydx
 
2 Γ
y dy
Z
4 0
= (sin3 t(− sin t(1 + sin2 t) + 2 sin t cos2 t)
2 π/2
− cos t(1 + sin2 t)(3 sin2 t cos t))dt

Z0
= 2 (−3 sin2 t + sin4 t)dt
Zπ/2
0
9 1
= 2 − + cos 2t + cos 4t dt
π/2 8 8
9π
= .
8


266 Definition Let Γ be a smooth curve. The integral
Z
f(x)||dx||
Γ
is called the path integral of f along Γ .
-3 -2 -1 1 2 3
-1
-2
-3
Z
267 Example Find x||dx|| where Γ is the triangle starting at A : (−1, −1)
Γ
to B : (2, −2), and ending in C : (1, 2).
Solution: The lines passing through the given points have equations LAB :
−x − 4
y= , and LBC : y = −4x + 6. On LAB
3
s 2 √
p 1 x 10dx
x||dx|| = x (dx)2 + (dy)2 = x 1 + − dx = ,
3 3
and on LBC
q √
x||dx|| = x (dx) + (dy) = x( 1 + (−4)2)dx = x 17dx.
p
2 2
122 Chapter 3
Hence
Z Z Z
x||dx|| = x||dx|| + x||dx||
Γ LAB LBC
Z2 √ Z1 √
x 10dx
= + x 17dx
√−1 3√ 2
10 3 17
= − .
2 2
3.4 Closed and Exact Forms

268 Lemma (Poincaré Lemma) If ω is a p-differential form of continuously
differentiable functions in Rn then
ddω = 0.
Proof We will prove this by induction on p. For p = 0 if
ω = f(x1, x2, . . . , xn)
then
Xn
∂f
dω = dxk
k=1
∂xk
and
Xn
∂f
ddω = d ∧ dxk
k=1
∂xk
n n
!
X X ∂2f
= ∧ dxj ∧ dxk
k=1 j=1
∂xj∂xk
Xn 2
∂2f

∂ f
= − dxj ∧ dxk
1≤j≤k≤n
∂x j∂xk ∂x k∂xj
= 0,
since ω is continuously differentiable and so the mixed partial derivatives

are equal. Consider now an arbitrary p-form, p > 0. Since such a form
Closed and Exact Forms 123
can be written as
X
ω= aj1 j2 ...jp dxj1 ∧ dxj2 ∧ · · · dxjp ,
1≤j1 ≤j2 ≤···≤jp ≤n
where the aj1 j2 ...jp are continuous differentiable functions in Rn, we have
X
dω = daj1 j2 ...jp ∧ dxj1 ∧ dxj2 ∧ · · · dxjp
1≤j1 ≤j2 ≤···≤jp ≤n !
X X n
∂aj1 j2 ...jp
= dxi ∧ dxj1 ∧ dxj2 ∧ · · · dxjp ,
1≤j ≤j ≤···≤j ≤n i=1
∂xi
1 2 p
it is enough to prove that for each summand

d da ∧ dxj1 ∧ dxj2 ∧ · · · dxjp = 0.
But

d da ∧ dxj1 ∧ dxj2 ∧ · · · dxjp = dda ∧ dxj1 ∧ dxj2 ∧ · · · dxjp

+da ∧ d dxj1 ∧ dxj2 ∧ · · · dxjp

= da ∧ d dxj1 ∧ dxj2 ∧ · · · dxjp ,
since dda = 0 from the case p = 0. But an independent induction argu-

ment proves that
d dxj1 ∧ dxj2 ∧ · · · dxjp = 0,
269 Definition A differential form ω is said to be exact if there is a continu-

ously differentiable function F such that
dF = ω.
270 Example The differential form

xdx + ydy
is exact, since
1 2 2
xdx + ydy = d (x + y ) .
2
124 Chapter 3
ydx + xdy
is exact, since
ydx + xdy = d (xy) .

x y
dx + 2 dy
x2 +y2 x + y2
is exact, since

x y 1 2 2
dx + dy = d loge(x + y ) .
x2 + y2 x2 + y2 2
! Let ω = dF be an exact form. By the Poincaré Lemma Theorem 268,

dω = ddF = 0. A result of Poincaré says that for certain domains (called
star-shaped domains) the converse is also true, that is, if dω = 0 on a
star-shaped domain then ω is exact.
273 Example Determine whether the differential form
2x(1 − ey) ey
ω= dx + dy
(1 + x2)2 1 + x2
is exact.
Solution: Assume there is a function F such that
dF = ω.
By the Chain Rule

∂F ∂F
dF = dx + dy.
∂x ∂y
This demands that
∂F 2x(1 − ey)
= ,
∂x (1 + x2)2
Closed and Exact Forms 125
∂F ey
= .
∂y 1 + x2
We have a choice here of integrating either the first, or the second expres-
sion. Since integrating the second expression (with respect to y) is easier,
we find
ey
F(x, y) = + φ(x),
1 + x2
where φ(x) is a function depending only on x. To find it, we differentiate
the obtained expression for F with respect to x and find
∂F 2xey
=− + φ 0 (x).
∂x (1 + x2)2
∂F
Comparing this with our first expression for , we find
∂x
2x
φ 0 (x) = ,
(1 + x2)2
that is
1
φ(x) = − + c,
1 + x2
where c is a constant. We then take
ey − 1
F(x, y) = + c.
1 + x2
274 Example Is there a continuously differentiable function such that
dF = ω = y2z3dx + 2xyz3dy + 3xy2z2dz ?
Solution: We have
dω = (2yz3dy + 3y2z2dz) ∧ dx
+(2yz3dx + 2xz3dy + 6xyz2dz) ∧ dy
+(3y2z2dx + 6xyz2dy + 6xy2zdz) ∧ dz
= 0,
126 Chapter 3
so this form is exact in a star-shaped domain. So put

∂F ∂F ∂F
dF = dx + dy + dz = y2z3dx + 2xyz3dy + 3xy2z2dz.
∂x ∂y ∂z
Then
∂F
= y2z3 =⇒ F = xy2z3 + a(y, z),
∂x
∂F
= 2xyz3 =⇒ F = xy2z3 + b(x, z),
∂y
∂F
= 3xy2z2 =⇒ F = xy2z3 + c(x, y),
∂z
Comparing these three expressions for F, we obtain F(x, y, z) = xy2z3.
We have the following equivalent of the Fundamental Theorem of Cal-
culus.
275 Theorem Let U ⊆ Rn be an open set. Assume ω = dF is an exact

form, and Γ a path in U with starting point A and endpoint B. Then
Z ZB
ω= dF = F(B) − F(A).
Γ A
In particular, if Γ is a simple closed path, then

I
ω = 0.
Γ
276 Example Evaluate the integral

I
2x 2y
2 2
dx + 2 dy
x +y x + y2
Γ
where Γ is the closed polygon with vertices at A = (0, 0), B = (5, 0), C =
(7, 2), D = (3, 2), E = (1, 1), traversed in the order ABCDEA.

2x 2y 4xy 4xy
d 2 2
dx + 2 2
dy = − 2 2 2
dy∧dx− 2 dx∧dy = 0,
x +y x +y (x + y ) (x + y2)2
and so the form is exact in a start-shaped domain. By virtue of Theorem
275, the integral is 0.
Integrating in ∧2(R2) 127

I
(x2 − y)dx + (y2 − x)dy,
Γ
where Γ is a loop of x3 + y3 − 2xy = 0 traversed once in the positive sense.
Solution: Since
∂ 2 ∂ 2
(x − y) = −1 = (y − x),
∂y ∂x
the form is exact, and since this is a closed simple path, the integral is 0.
278 Theorem If α is a p-form and β is a q-form on n-dimensional space,

then
d(α ∧ β) = dα ∧ β + (−1)pα ∧ dβ.
3.5 Integrating in ∧2(R2)

279 Definition A 2-dimensional oriented manifold of R2 is simply an open
set (region) D ∈ R2, where the + orientation is counter-clockwise and the
− orientation is clockwise. A general oriented 2-manifold is a union of open
sets.
! The region −D has opposite orientation to D and

Z Z
ω=− ω.
−D D
! In this section, unless otherwise noticed, we will choose the positive

orientation for the regions considered. This corresponds to using the area
form dx ∧ dy.
Let D ⊆ R2. Given a function f : D → R, the integral

ZZ
f(x, y)dx ∧ dy
D
128 Chapter 3
is the sum of all the values of f restricted to D. In particular,
ZZ
dx ∧ dy
D
is the area of D.
In order to evaluate double integrals, we need the following.
280 Theorem (Fubini’s Theorem) Let D = [a; b] × [c; d], and let f : A → R
be continuous. Then
ZZ Z b Z d Z d Z b
f(x, y)dx ∧ dy = f(x, y)dy dx = f(x, y)dx dy
a c c a
D
Fubini’s Theorem allows us to convert the double integral into iterated (sin-
gle) integrals.
281 Example
ZZ Z 1 Z 3
xydx ∧ dy = xydy dx
[0;1]×[2;3] 0 2
Z 1 2 3!
xy
= dx
0 2 2
Z1
9x
= − 2x dx
2
0 2 1
5x
=
4 0
5
= .
4
Notice that if we had integrated first with respect to x we would have ob-
tained the same result:

Z 3 Z 1 Z3 1!
x2y

xydx dy = dy
2 0 2 2 0
Z3
y
= dx
2 2
2 3
y
=
4 2
5
= .
4
Also, this integral is “factorable into x and y pieces” meaning that

ZZ Z 1 Z 3
xydx ∧ dy = xdx ydy
[0;1]×[2;3] 0 2

1 5
=
2 2
5
=
4
282 Example We have

Z4 Z1 Z4 Z1
(x + 2y)(2x + y) dx ∧ dy = (2x2 + 5xy + 2y2) dx ∧ dy
3 0 Z34 0
2 5 2
= + y + 2y dy
3 3 2
409
= .
12
283 Example Find

ZZ
(x + y)(sin x)(sin y)dx ∧ dy
D
where D = [0; π]2.

130 Chapter 3

Solution: The integral equals
ZZ ZZ Z π Z π
x sin x sin ydx ∧ dy + y sin x sin ydx ∧ dy = 2 y sin y dy sin x dx
0 0
D D
= 4π.
3 3
2 2
1 1
-3 -2 -1 1 2 3 -3 -2 -1 1 2 3
-1 -1
-2 -2
-3 -3
Figure 3.5: Example 284. Inte- Figure 3.6: Example 284. Inte-
gration order dydx. gration order dxdy.
In the cases when the domain of integration is not a rectangle, we

decompose so that, one variable is kept constant.
ZZ
284 Example Find xy dxdy in the triangle with vertices A : (−1, −1),
D
B : (2, −2), C : (1, 2).
The lines passing through the given points have equations LAB : y =
−x − 4 3x + 1
, LBC : y = −4x + 6, LCA : y = . Now, we draw the re-
3 2
gion carefully. If we integrate first with respect to y, we must divide the
region as in figure 3.5, because there are two upper lines which the upper
value of y might be. The lower point of the dashed line is (1, −5/3). The
integral is thus
Z1 Z (3x+1)/2 !Z 2 Z −4x+6
11
x y dy dx + x y dy dx = − .
−1 (−x−4)/3 1 (−x−4)/3 8

If we integrate first with respect to x, we must divide the region as in figure

3.6, because there are two left-most lines which the left value of x might
be. The right point of the dashed line is (7/4, −1). The integral is thus
Z −1 Z (6−y)/4 ! Z2 Z (6−y)/4 !
11
y x dx dy + y x dx dy = − .
−2 −4−3y −1 (2y−1)/3 8
285 Example Consider the region inside the parallelogram P with vertices
at A : (6, 3), B : (8, 4), C : (9, 6), D : (7, 5), as in figure 3.7. Find
ZZ
1
xy dx ∧ dy.
3
P
6
5
4
3
1 2 3 4 5 6 7 8 9 10
Solution: The lines joining the points have equations

x
LAB : y = ,
2
LBC : y = 2x − 12,
x 3
LCD : y = + ,
2 2
132 Chapter 3
LDA : y = 2x − 9.
The integral is thus

Z 4 Z 2y Z 5 Z (y+12)/2 Z 5 Z (y+12)/2
1 1 1 409
xy dx∧dy+ xy dx∧dy+ xy dx∧dy = .
3 3 (y+9)/2 3 4 (y+9)/2 3 4 2y−3 12
ZZ
286 Example Find (2x+3y+1) dxdy, where D is the triangle with vertices
D
at A(−1, −1), B(2, −4), and C(1, 3).
Solution: The line joining A, and B has equation y = −x − 2, line joining B,

and C has equation y = −7x + 10, and line joining A, and C has equation
y = 2x + 1. We split the triangle along the vertical line x = 1, and integrate
first with respect to y. The desired integral is then
ZZ Z 1 Z 2x+1
(2x + 3y + 1) dxdy = (2x + 3y + 1)dy dx
−1
D
Z 2−x−2
Z −7x+10
+ (2x + 3y + 1)dy dx
1 −x−2
Z1
21 2 3
= x + 9x − dx
−1 2 2
Z2
60x2 − 198x + 156 dx

+
1
= 4−1
= 3.
287 Example Find

ZZ
(xy(x + y))dx ∧ dy
D
where
D = {(x, y) ∈ R2|x ≥ 0, y ≥ 0, x + y ≤ 1}.
Solution: The domain of integration is a triangle. The integral equals

ZZ Z 1 Z 1−x
xy(x + y)dx ∧ dy = xy(x + y) dy dx
0 0
D
Z1 2 1−x
y y3
= x x + dx
2 3 0
Z01
x(1 − x)2 (1 − x)3

= x + dx
0 2 3
1
= .
30
288 Example Find ZZ

1
dx ∧ dy
(x + y)4
D
where
D = {(x, y) ∈ R2|x ≥ 1, y ≥ 1, x + y ≤ 4}.

ZZ Z 3 Z 4−x
1 dy
dx ∧ dy = dy dx
(x + y)4 1 1 (x + y)4
D
Z3 4−x
1 −3
= − (x + y) dx
1 3 1
Z
1 3

1 1
= − dx
3 1 (1 + x)3 64
1
= .
48
289 Example Find ZZ

xdx ∧ dy
D
where
D = {(x, y) ∈ R2|y ≥ 0, x − y + 1 ≥ 0, x + 2y − 4 ≤ 0}.

134 Chapter 3
ZZ Z 2/3 Z x+1 Z4 Z 2− x !
2
xdx ∧ dy = dy x dx + dy x dx
−1 0 2/3 0
D
Z 2/3 Z4
x
= x(x + 1) dx + x 2− dx
−1 2/3 2
275
= .
54
290 Example Find

ZZ
xydx ∧ dy
D
where
D = {(x, y) ∈ R2|y ≥ x2, x ≥ y2}.
ZZ Z1 Z √x !
xydx ∧ dy = x y dy dx
0 x2
D
Z1
1
= x(x − x4) dx
0 2
1
= .
12
291 Example Find

ZZ
loge(1 + x + y)dx ∧ dy
D
where
D = {(x, y) ∈ R2|x ≥ 0, y ≥ 0, x + y ≤ 1}.
Solution: Integrating by parts,

ZZ Z 1 Z 1−x
loge(1 + x + y)dx ∧ dy = loge(1 + x + y) dy dx
0 0
D
Z1
= [(1 + x + y) loge(1 + x + y) − (1 + x + y)]1−x
0 dx
Z01
= (2 loge(2) − 1 − loge(1 + x) − x loge(1 + x) + x) dx
0
1
= .
4
292 Example Find ZZ

xydx ∧ dy
D
where
D = {(x, y) ∈ R2|x ≥ 0, y ≥ 0, xy + y + x ≤ 1}.
Z1−x
 
ZZ Z1
 1+x
xydx ∧ dy = xy dy dx


0 0
D
Z1 2!
1 1−x
= x dx
0 2 1+x
Z2
(t − 1)(t − 2)2
= dt
1 t2
11
= 4 loge 2 − .
4
293 Example Find ZZ

loge(1 + x2 + y)dx ∧ dy
D
where
D = {(x, y) ∈ R2|x ≥ 0, y ≥ 0, x2 + y ≤ 1}.
136 Chapter 3
Solution: Using integration by parts,

ZZ Z 1 Z 1−x2 !
loge(1 + x2 + y)dx ∧ dy = loge(1 + x2 + y) dy dx
0 0
D
Z1
= (2 loge(2) − 1 − loge(1 + x2))dx
0Z
1
+ (−x2 loge(1 + x2) + x2) dx
0
2 8 π
= loge 2 + − .
3 9 3
294 Example Find ZZ

y
dx ∧ dy
x2 +1
D
where
D = {(x, y) ∈ R2|x ≥ 0, x2 + y2 ≤ 1}.
Solution: The integral is 0. Observe that if (x, y) ∈ D then (x, −y) ∈ D.

Also, f(x, −y) = −f(x, y).
295 Example Find ZZ

2x(x2 + y2)dx ∧ dy
D
where
D = {(x, y) ∈ R2|x4 + y4 + x2 − y2 ≤ 1}.
Solution: The integral is 0. Observe that if (x, y) ∈ D then (−x, y) ∈ D.

Also, f(−x, y) = −f(x, y).
296 Example Find ZZ

|x − y|dx ∧ dy
D
where
D = {(x, y) ∈ R2||x| ≤ 1, |y| ≤ 1}.
Solution: Let
D1 = {(x, y) ∈ R2| − 1 ≤ x ≤ 1, x ≤ y},
D2 = {(x, y) ∈ R2| − 1 ≤ x ≤ 1, x > y}.
Then D = D1 ∪ D2, D1 ∩ D2 = ∅ and so

ZZ ZZ ZZ
f(x, y)dx ∧ dy = f(x, y)dx ∧ dy + f(x, y)dx ∧ dy.
D1 D2
D
By symmetry,
ZZ ZZ
f(x, y)dx ∧ dy = f(x, y)dx ∧ dy,
D1 D2
and so
ZZ ZZ
f(x, y)dx ∧ dy = 2 f(x, y)dx ∧ dy
D
Z 1D1Z 1
= 2 (y − x) dy dx
Z 1 −1 x
= (1 − 2x + x2) dx
−1
8
= .
3
297 Example Find

Z 4 Z √y !
y/x
e dx dy.
0 y/2
Solution: We have
y √
0 ≤ y ≤ 4, ≤ x ≤ y =⇒ 0 ≤ x ≤ 2, x2 ≤ y ≤ 2x.
2
138 Chapter 3
We then have
Z 4 Z √y ! Z 2 Z 2x
y/x y/x
e dx dy = e dy dx
0 y/2 0 x2
Z2

= xey/x 2x
x2 dx
Z02
= (xe2 − xex) dx
0
= 2e2 − (2e2 − e2 + 1)
= e2 − 1
298 Example Find
Z 2 Z x Z 4 Z 2
πx πx
√
sin dy dx + √
sin dy dx.
1 x 2y 2 x 2y
Solution: Upon splitting the domain of integration, we find that the integral
equals
Z 2 Z y2 ! Z2 y2
πx 2y πx
sin dx dy = − cos dy
1 y 2y 1 π 2y y
Z2
2y πy
= − − cos dy
1 π 2
4(π + 2)
= ,
π3
upon integrating by parts.
299 Example Find the area of the region
√ √ √ p
R = {(x, y) ∈ R2 : x+ y ≥ 1, 1−x+ 1 − y ≥ 1}.
Solution: The area is given by
ZZ Z 1 Z 1−(1−√1−x)2 !
dx ∧ dy = √
dy dx
D 0 (1− x)2
Z1
√ √
= 2 ( 1 − x + x − 1)dx
0
2
= .
3
ZZ
√
300 Example Find xy dx ∧ dy, where
D
D = {(x, y) ∈ R2 : y ≥ 0, (x + y)2 ≤ 2x}.
1
Solution: Observe that x ≥ (x + y)2 ≥ 0. Hence we may take the positive
√ 2 √
square root giving y ≤ 2x − x. Since y ≥ 0, we must have 2x − x ≥ 0
which means that x ≤ 2. The integral equals
Z 2 Z √2x−x ! Z2
√ 2 √ √
xydy dx = x( 2x − x)3/2dx
0 0 3 0
Z √2
√
4
= u2(u 2 − u2)3/2du
3 0
Z
1 1
= (1 − v2)3/2(1 + v)2dv
6 −1
Z
1 π/2
= cos4 θ(1 + sin2 θ)dθ
6 −π/2
7π
= .
96
301 Example A rectangle R on the plane is the disjoint union R = ∪N k=1Rk of

rectangles Rk. It is known that at least one side of each of the rectangles
Rk is an integer. Shew that at least one side of R is an integer.
140 Chapter 3


Za 

 0 if a is an integer
sin 2πx dx =
 1 (1 − cos 2πa) if a is not an integer
0 

2π
Thus Za
sin 2πx dx = 0 ⇐⇒ a is an integer.
0
Now
N ZZ
X
sin 2πx sin 2πy dx ∧ dy = 0
k=1 Rk
since at least one of the sides of each Rk is an integer. Since
ZZ N ZZ
X
sin 2πx sin 2πy dx ∧ dy = sin 2πx sin 2πy dx ∧ dy,
R k=1 Rk
we deduce that at least one of the sides of R is an integer, finishing the

proof.
302 Example Evaluate

Z1 Z1 Z1
· · · (x1x2 · · · xn)dx1 ∧ dx2 ∧ . . . ∧ dxn.
0 0 0
Solution: We have
Z1 Z1 Z1 Yn Z 1 Y n
1 1
· · · (x1x2 · · · xn)dx1 ∧dx2 ∧. . .∧dxn = xkdxk = = n.
0 0 0 k=1 0 k=1
2 2

Z1 Z1 Z1
··· (x1 + x2 + · · · + xn) dx1 ∧ dx2 ∧ . . . ∧ dxn.
0 0 0
Change of Variables in ∧2(R2) 141
Solution: This is
Z1 Z1 Z1 X n
! n Z1 Z1
X Z1
··· xk dx1 ∧ dx2 ∧ . . . ∧ dxn = · · · xkdx1 ∧ dx2 ∧ . . . ∧ dxn
0 0 0 k=1 k=1 0 0 0
n
X 1
=
k=1
2
n
= .
2
304 Example (Putnam Exam 1965) Evaluate

Z1 Z1 Z1 π
2
lim · · · cos (x1 + x2 + · · · + xn) dx1 dx2 . . . dxn.
n→+∞ 0 0 0 2n
Solution: Make the change of variables xk = 1 − yk. Then

Z1 Z1 Z1 π
2
I= · · · cos (x1 + x2 + · · · + xn) dx1 dx2 . . . dxn
0 0 0 2n
equals
Z1 Z1 Z1 π
2
· · · sin (y1 + y2 + · · · + yn) dy1 dy2 . . . dyn.
0 0 0 2n
1
Since sin2 t + cos2 t = 1, we have 2I = 1, and so I = .
2
3.6 Change of Variables in ∧2(R2)

We now perform a multidimensional analogue of the change of variables
theorem in one variable.
305 Theorem Let (D, ∆) ∈ (Rn)2 be open, bounded sets in Rn with volume
and let g : ∆ → D be a continuously differentiable bijective mapping such
1
that det J−→ g 6= 0, and both | det J−→ g|, are bounded on ∆. For
u u
| det J−→ g|
u
f : D → R bounded and integrable, f ◦ g| det J− → g| is integrable on ∆ and
u
Z Z Z Z
· · · f = · · · f ◦ g| det J− → g|,
u
D ∆

142 Chapter 3
that is Z Z
··· f(x1, x2, . . . , xn)dx1 ∧ dx2 ∧ . . . ∧ dxn
D
Z Z
= ··· f(g(u1, u2, . . . , un))| det J−
→ g|du1 ∧du2 ∧. . .∧dun.
u
∆
7 7
6 6
5 5
4 4
3 3
2 2
1 1
1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10
Figure 3.8: Example 306. Figure 3.9: Example 306.

xy-plane. uv-plane.
One normally chooses changes of variables that map into rectangular re-
gions, or that simplify the integrand. Let us start with a rather trivial exam-
ple.
306 Example Evaluate the integral

Z4 Z1
(x + 2y)(2x + y)dx ∧ dy.
3 0
Solution: Observe that we have already computed this integral in example

282. Put
u = x + 2y =⇒ du = dx + 2dy,
v = 2x + y =⇒ dv = 2dx + dy,
giving
du ∧ dv = −3dx ∧ dy.
Now,     
u 1 2  x 
 =  
    
v 2 1 y
is a linear transformation, and hence it maps quadrilaterals into quadrilat-
erals. The corners of the rectangle in the area of integration in the xy-plane
are (0, 3), (1, 3), (1, 4), and (0, 4), (traversed counter-clockwise) and they
map into (6, 3), (7, 5), (9, 6), and (8, 4), respectively, in the uv-plane (see
figure 3.9). The form dx ∧ dy has opposite orientation to du ∧ dv so we
use
dv ∧ du = 3dx ∧ dy
instead. The integral sought is
ZZ
1 409
uv dv ∧ du = ,
3 12
P
from example 285.
307 Example The integral

ZZ Z1
4 4 1 4
(x − y )dx ∧ dy = − y dy = 0.
[0;1]2 0 5
Evaluate it using the change of variables u = x2 − y2, v = 2xy.
Solution: First we find

du = 2xdx − 2ydy,
dv = 2ydx + 2xdy,
and so
du ∧ dv = (4x2 + 4y2)dx ∧ dy.
We now determine the region ∆ into which the square D = [0; 1]2 is
mapped. We use the fact that boundaries will be mapped into boundaries.
Put
AB = {(x, 0) : 0 ≤ x ≤ 1},
144 Chapter 3
BC = {(1, y) : 0 ≤ y ≤ 1},
CD = {(1 − x, 1) : 0 ≤ x ≤ 1},
DA = {(0, 1 − y) : 0 ≤ y ≤ 1}.
On AB we have u = x, v = 0. Since 0 ≤ x ≤ 1, AB is thus mapped into

the line segment 0 ≤ u ≤ 1, v = 0.
v2
On BC we have u = 1 − y2, v = 2y. Thus u = 1 − . Hence BC is
4
v2
mapped to the portion of the parabola u = 1 − , 0 ≤ v ≤ 2.
4
On CD we have u = (1 − x)2 − 1, v = 2(1 − x). This means that u =
v2
− 1, 0 ≤ v ≤ 2.
4
Finally, on DA, we have u = −(1 − y)2, v = 0. Since 0 ≤ y ≤ 1, DA is
mapped into the line segment −1 ≤ u ≤ 0, v = 0. The region ∆ is thus the
v2 v2
area in the uv plane enclosed by the parabolas u ≤ − 1, u ≤ 1 − with
4 4
−1 ≤ u ≤ 1, 0 ≤ v ≤ 2.
We deduce that
ZZ ZZ
4 4 1
(x − y )dx ∧ dy = (x4 − y4) du ∧ dv
[0;1]2 ∆ 4(x + y2)
2
ZZ
1
= (x2 − y2)du ∧ dv
4 ∆
ZZ
1
= udu ∧ dv
4 ∆
Z Z 2 !
1 2 1−v /4
= udu dv
4 0 v2 /4−1
= 0,
as before.
308 Example Find ZZ

3 +y3 )/xy
e(x dx ∧ dy
D
where
D = {(x, y) ∈ R2|y2 − 2px ≤ 0, x2 − 2py ≤ 0, p ∈]0; +∞[ fixed},
using the change of variables x = u2v, y = uv2.
Solution: We have
dx = 2uvdu + u2dv,
dy = v2du + 2uvdv,
dx ∧ dy = 3u2v2du ∧ dv.
The region transforms into
∆ = {(u, v) ∈ R2|0 ≤ u ≤ (2p)1/3, 0 ≤ v ≤ (2p)1/3}.
The integral becomes

ZZ ZZ
u6v3 + u3v6

f(x, y)dx ∧ dy = exp (3u2v2) du ∧ dv
u3v3
D ∆
ZZ
3 3
= 3 eu ev u2v2 du ∧ dv
∆
Z (2p)1/3 !2
1 3
= 3u2eu du
3 0
1 2p
= (e − 1)2.
3
As an exercise, you may try the (more natural) substitution x3 = u2v, y3 =

v2u and verify that the same result is obtained.
ZZ
309 Example Find f(x, y)dx ∧ dy where
D
D = {(x, y) ∈ R2|a ≤ xy ≤ b, y ≥ x ≥ 0, y2 −x2 ≤ 1, (a, b) ∈ R2, 0 < a < b}
and f(x, y) = y4 − x4 by using the change of variables u = xy, v = y2 − x2.

146 Chapter 3
Solution: Here we argue that
du = ydx + xdy,
dv = −2xdx + 2ydy.
Taking the wedge product of differential forms,
du ∧ dv = 2(y2 + x2)dx ∧ dy.
Hence
1
f(x, y)dx ∧ dy = (y4 − x4) du ∧ dv
2(y2 + x2)
1 2
= (y − x2)du ∧ dv
2
v
= du ∧ dv
2
The region transforms into
∆ = [a; b] × [0; 1].

ZZ ZZ
f(x, y)dx ∧ dy = v du ∧ dv
D ∆
Z b Z 1
1
= du v dv
2 a 0
b−a
= .
4
310 Example Let D 0 = {(u, v) ∈ R2 : u ≤ 1, −u ≤ v ≤ u}. Consider
R2 → R2
Φ: .
u+v u−v
(u, v) 7→ ,
2 2
➊ Find the image of Φ on D 0 , that is, find D = Φ(D 0 ).

➋ Find ZZ
2 −y2
(x + y)2ex dx ∧ dy.
D
Solution:
u+v u−v
➊ Put x = and y = . Then x + y = u and x − y. Observe
2 2
that D 0 is the triangle in the uv plane bounded by the lines u = 0, u =
1, v = u, v = −u. Its image under Φ is the triangle bounded by the
equations x = 0, y = 0, x + y = 1. Clearly also
1
dx ∧ dy = du ∧ dv.
2
➋ From the above

ZZ ZZ
2 x2 −y2 1
(x + y) e dx ∧ dy = u2euvdu ∧ dv
D 2
Z DZ
0
1 1 u 2 uv
= u e dudv
2 0 −u
Z
1 1 2 2
= u(eu − e−u )du
2 0
1
= (e + e−1 − 2).
4
311 Example Use the following steps (due to Tom Apostol) in order to
prove that
X∞
1 π2
= .
n=1
n2 6
➊ Use the series expansion

1
= 1 + t + t2 + t3 + · · · |t| < 1,
1−t
in order to prove (formally) that
Z1 Z1 X∞
dxdy 1
= .
0 0 1 − xy n=1
n2
148 Chapter 3
➋ Use the change of variables u = x + y, v = x − y to shew that

Z1 Z1 Z 1 Z u Z 2 Z 2−u
dxdy dv dv
=2 2 2
du+2 2 2
du.
0 0 1 − xy 0 −u 4 − u + v 1 u−2 4 − u + v
➌ Shew that the above integral reduces to

Z1 Z2
2 u 2 2−u
2 √ arctan √ du + 2 √ arctan √ du.
0 4 − u2 4 − u2 1 4 − u2 4 − u2
π2
➍ Finally, prove that the above integral is by using the substitution
6
u
θ = arcsin .
2
Solution:
➊ Formally,
Z1 Z1 Z1 Z1
dxdy
= (1 + xy + x2y2 + x3y3 + · · · )dxdy
0 0 1 − xy
Z01 0 1
xy2 x2y3 x3y4
= y+ + + + ··· dx
0 2 3 4 0
Z1
x x2 x3
= (1 + + + + · · · )dx
0 2 3 4
1 1 1
= 1 + 2 + 2 + 2 + ···
2 3 4
➋ This change of variables transforms the square [0; 1] × [0; 1] in the xy

plane into the square with vertices at (0, 0), (1, 1), (2, 0), and (1, −1)
in the uv plane. We will split this region of integration into two disjoint
triangles: T1 with vertices at (0, 0), (1, 1), (1, −1), and T2 with vertices
at (1, −1), (1, 1), (2, 0). Observe that
1
dx ∧ dy = du ∧ dv,
2
Change to Polar Co-ördinates 149
and that u + v = 2x, u − v = 2y and so 4xy = u2 − v2. The integral

becomes
Z1 Z1 ZZ
dxdy 1 du ∧ dv
= 2 2
0 0 1 − xy 2 1 − u −v
4
T1 ∪T2
Z 1 Z u Z 2 Z 2−u
dv dv
= 2 2 2
du + 2 2 2
du,
0 −u 4 − u + v 1 u−2 4 − u + v
as desired.
➌ This follows by using the identity

Zt
dΩ
2
= arctan t.
0 1+Ω
➍ This is straightforward but tedious!
3.7 Change to Polar Co-ördinates

One of the most common changes of variable is the passage to polar co-
ördinates where
x = ρ cos θ =⇒ dx = cos θdρ − ρ sin θdθ,
y = ρ sin θ =⇒ dy = sin θdρ + ρ cos θdθ,

whence
dx ∧ dy = (ρ cos2 θ + ρ sin2 θ)dρ ∧ dθ = ρdρ ∧ dθ.
312 Example Find ZZ p

xy x2 + y2dx ∧ dy
D
where
D = {(x, y) ∈ R2|x ≥ 0, y ≥ 0, y ≤ x, x2 + y2 ≤ 1}.
150 Chapter 3
We use polar co-ördinates. The region D transforms into the region

h πi
∆ = [0; 1] × 0; .
4
Therefore the integral becomes
ZZ Z π/4 ! Z
1
4 4
ρ cos θ sin θ dρ ∧ dθ = cos θ sin θ dθ ρ dρ
0 0
∆
1
= .
20
313 Example Find ZZ p

x2 + y2dx ∧ dy
D
where
D = {(x, y) ∈ R2|x ≥ 0, y ≥ 0, x2 + y2 ≤ 1, x2 + y2 − 2y ≥ 0}.
Using polar co-ördinates,

ZZ Z π/6 Z 1
2
f(x, y)dx ∧ dy = ρ dρ dθ
0 2sin θ
D
Z
1 π/6
= (1 − 8 sin3 θ) dθ
3 0
π 16 √
= − + 3.
18 9
314 Example Find ZZ

(x2 − y2)dx ∧ dy
D
where
D = {(x, y) ∈ R2|(x − 1)2 + y2 ≤ 1}.
Solution: Using polar co-ördinates,

ZZ Z π/2 Z 2cos θ
2 2
x − y dx ∧ dy = ρ dρ (cos2 θ − sin2 θ)dθ
3
−π/2 0
D
Z π/2
= 8 cos4 θ(cos2 θ − sin2 θ) dθ
0
= π.
315 Example Find ZZ

√
xydx ∧ dy
D
where
D = {(x, y) ∈ R2|(x2 + y2)2 ≤ 2xy}.
Solution: Using polar co-ördinates,

ZZ Z π/4 Z √sin 2θ p !
√
xydx ∧ dy = 4 ρ ρ2 cos θ sin θ dρ dθ
0 0
D
Z
4 π/4 √ √
= ( sin 2θ)3 cos θ sin θ dθ
3 0
Z
4 π/4 2
= √ sin 2θ dθ
3 √2 0
π 2
= .
12
ZZ
D
D = {(x, y) ∈ R2 : b2x2 + a2y2 = a2b2, (a, b) ∈]0; +∞[ fixed}

and f(x, y) = x3 + y3.
Solution: Using x = aρ cos θ, y = bρ sin θ, the integral becomes

Z 2π Z 1
3 3 3 3 2
(ab) a cos θ + b sin θdθ ρ dρ = (ab)(a3 + b3).
4
0 0 15
152 Chapter 3
ZZ
D
D = {(x, y) ∈ R2|y ≥ 0, x2 + y2 − 2x ≤ 0}
and f(x, y) = x2y.
Solution: Using polar co-ördinates the integral becomes

Z π/2 Z 2cos θ
4
ρ dρ cos2 θ sin θdθ = .
4
0 0 5
ZZ
D
D = {(x, y) ∈ R2|x ≥ 1, x2 + y2 − 2x ≤ 0}
1
and f(x, y) = .
(x2 + y2)2
Solution: Using polar co-ördinates the integral becomes

Z π/4 Z 2cos θ Z π/4
sec2 θ

1 2 π
3
dρ dθ = cos θ − dθ = .
−π/4 1/cos θ ρ 0 4 8
ZZ
2 −xy−y2
319 Example Find e−x dxdy, where
D
D = {(x, y) ∈ R2 : x2 + xy + y2 ≤ 1}.
Solution: Completing squares

√ !2
y 2 3y
x2 + xy + y2 = x + + .
2 2
√
y 3y
Put U = x + , V = . The integral becomes
2 2
ZZ ZZ
−x2 −xy−y2 2 2 2
e dxdy = √ e−(U +V )dUdV.
{x2 +xy+y2 ≤1} 3 {U2 +V2 ≤1}
Passing to polar co-ördinates, the above equals

Z Z
2 2π 1 −ρ2 2π
√ ρe dρdθ = √ (1 − e−1).
3 0 0 3
320 Example Let

D = {(x, y) ∈ R2 : x2 + y2 − y ≤ 0, x2 + y2 − x ≤ 0}.
Find the integral ZZ
(x + y)2dx ∧ dy.
D
Solution: Put
D 0 = {(x, y) ∈ R2 : y ≥ x, x2 + y2 − y ≤ 0, x2 + y2 − x ≤ 0}.
Then the integral equals
ZZ
2 (x + y)2dx ∧ dy.
D0
Using polar co-ördinates the integral equals

Z π/2 Z cos θ Z
1 π/2

2 3
2 (cos θ + sin θ) ρ dρ dθ = cos4 θ(1 + 2 sin θ cos θ)dθ
π/4 0 2 π/4
3π 5
= − .
64 48
321 Example Let D = {(x, y) ∈ R2|y ≤ x2 + y2 ≤ 1}. Compute

ZZ
dx ∧ dy
2 2 2
.
D (1 + x + y )
Solution: Observe that D = D2 \ D1 where D2 is the disk limited by the

equation x2 + y2 = 1 and D1 is the disk limited by the equation x2 + y2 = y.
Hence
ZZ ZZ ZZ
dx ∧ dy dx ∧ dy dx ∧ dy
2 2 2
= 2 2 2
− 2 2 2
.
D (1 + x + y ) D2 (1 + x + y ) D1 (1 + x + y )
154 Chapter 3
Using polar co-ördinates we have

ZZ Z 2π Z 1
dx ∧ dy ρ π
2 2 2
= 2 2
dρdθ =
D2 (1 + x + y ) 0 0 (1 + ρ ) 2
and
ZZ Z π/2 Z sin θ Z π/2
dx ∧ dy ρ sin2 θdθ
= 2 dρdθ = 2
D1 (1 + x2 + y2)2 0 0 (1 + ρ2)2 0 √ 1 + sin θ
Z +∞
dt dt π π 2
= − = − .
0 t2 + 1 2t2 + 1 2 4
(We evaluated this last integral using t = tan θ) Finally, the integral equals
√ ! √
π π π 2 π 2
− − = .
2 2 4 4

ZZ p
x3y3 1 − x4 − y4 dx ∧ dy
{(x,y)∈R2 :x≥0,y≥0,x4 +y4 ≤1}
using x2 = ρ cos θ, y2 = ρ sin θ.
Solution: We have
2xdx = cos θdρ − ρ sin θdθ, 2ydy = sin θdρ + ρ cos θdθ,
whence
4xydx ∧ dy = ρdρ ∧ dθ.
It follows that
p 1 2 2 p
x3y3 1 − x4 − y4 dx ∧ dy = (x y )( 1 − x4 − y4)(4xy dx ∧ dy)
4
1 3 p
= (ρ cos θ sin θ 1 − ρ2)dρ ∧ dθ
4
Observe that
x4 + y4 ≤ 1 =⇒ ρ2 cos2 θ + ρ2 sin2 θ ≤ 1 =⇒ ρ ≤ 1.
Since the integration takes place on the first quadrant, we have 0 ≤ θ ≤

π/2. Hence the integral becomes
Z π/2 Z 1 Z π/2 ! Z
1
1 3 p
2
1 3
p
2
(ρ cos θ sin θ 1 − ρ )dρ ∧ dθ = cos θ sin θdθ ρ 1 − ρ dρ
0 0 4 4 0 0
1 1 2
= · ·
4 2 15
1
= .
60
323 Example William Thompson (Lord Kelvin) is credited to have said: “A

mathematician is someone to whom
Z +∞ √
−x2 π
e dx =
0 2
is as obvious as twice two is four to you. Liouville was a mathematician.”

Prove that Z +∞ √
−x2 π
e dx =
0 2
by following these steps.
➊ Let a > 0 be a real number and put Da = {(x, y) ∈ R2|x2 + y2 ≤ a2}.

Find ZZ
2 +y2 )
Ia = e−(x dx ∧ dy.
Da
➋ Let a > 0 be a real number and put ∆a = {(x, y) ∈ R2||x| ≤ a, |y| ≤ a}.
Let ZZ
2 +y2 )
Ja = e−(x dx ∧ dy.
∆a
Prove that
Ia ≤ Ja ≤ Ia√2.
➌ Deduce that Z +∞ √
−x2 π
e dx = .
0 2
156 Chapter 3
Solution:
➊ Using polar co-ördinates

Z 2π Z a
−ρ2 2
Ia = ρe dρ dθ = π(1 − e−a ).
0 0
➋ The domain of integration of Ja is a square of side 2a centred at the

origin. The respective domains of integration of Ia and Ia√2 are the
inscribed and the exscribed circles to the square.
➌ First observe that Z a 2

−x2
Ja = e dx .
−a
Since both Ia and Ia√2 tend to π as a → +∞, we deduce that Ja → π.

This gives the result.
Recall from formula 3.2 that the area enclosed by a simple closed curve Γ
is given by Z
1
xdy − ydx.
2 Γ
Using polar co-ördinates
xdy − ydx = (ρ cos θ)(sin θdρ + ρ cos θdθ) − (ρ sin θ)(cos θdρ − ρ sin θdθ)
= ρ2dθ.
This will be used in the next problem.
324 Example Prove that every closed convex region in the plane of area
≥ π has two points which are two units apart.
Solution: Parametrise the curve enclosing the region by polar co-ördinates

so that the region is tangent to the polar axis at the origin. Let the equation
of the curve be ρ = f(θ). The area of the region is then given by
Zπ Zπ Z
1 2 1 2 1 π/2
ρ dθ = (f(θ)) dθ = ((f(θ))2 + (f(θ + π/2))2)dθ.
2 0 2 0 2 0
By the Pythagorean Theorem, the integral above is the integral of the

square of the chord in question. If no two points are farther than 2 units,
their squares are no farther than 4 units, and so the area
Z
1 π/2
< 4dθ = π,
2 0
a contradiction.
325 Example (Putnam Exam 1976) In the xy-plane, if R is the set of points
inside and on a convex polygon, let D(x, y) be the distance from (x, y) to
the nearest point R. Show that
Z +∞ Z +∞
e−D(x,y) dx ∧ dy = 2π + L + A,
−∞ −∞
where L is the perimeter of R and A is the area of R.
Solution: Let I(S) denote the integral sought over a region S. Since D(x, y) =
0 inside R, I(R) = A. Let L be a side of R with length l and let S(L ) be
the half strip consisting of the points of the plane having a point on L as
nearest point of R. Set up co-ördinates uv so that u is measured parallel
to L and v is measured perpendicular to L. Then
Z l Z +∞
I(S(L )) = e−v du ∧ dv = l.
0 0
The sum of these integrals over all the sides of R is L.
If V is a vertex of R, the points that have V as nearest from R lie
inside an angle S(V ) bounded by the rays from V perpendicular to the
edges meeting at V . If α is the measure of that angle, then using polar
co-ördinates Z α Z +∞
I(S(V )) = ρe−ρ dρ ∧ dθ = α.
0 0
The sum of these integrals over all the vertices of R is 2π. Assembling all
these integrals we deduce the result.
3.8 Integrating in ∧3(R3)

326 Definition A 3-dimensional oriented manifold of R3 is simply an open
set (body) V ∈ R3, where the + orientation is in the direction of the outward
158 Chapter 3
pointing normal to the body, and the − orientation is in the direction of the
inward pointing normal to the body. A general oriented 3-manifold is a
union of open sets.
! The region −V has opposite

Z
orientation to D and
Z
ω=− ω.
−V V

orientation for the regions considered. This corresponds to using the vol-
ume form dx ∧ dy ∧ dz.
Let V ⊆ R3. Given a function f : V → R, the integral

ZZ
f(x, y, z)dx ∧ dy ∧ dz
V
is the sum of all the values of f restricted to V. In particular,

ZZ
dx ∧ dy ∧ dz
V
is the oriented volume of V.
327 Example Find ZZZ

x2yexyz dx ∧ dy ∧ dz.
[0;1]3
Solution: The integral is

Z 1 Z 1 Z 1 Z 1 Z 1
2 xyz xy
x ye dz dy dx = x(e − 1) dy dx
0 0 0
Z01 0
= (ex − x − 1)dx
0
5
= e− .
2
ZZZ
328 Example Find z dx ∧ dy ∧ dz if
R
√ √ √
R = {(x, y, z) ∈ R3|x ≥ 0, y ≥ 0, z ≥ 0, x + y + z ≤ 1}.

ZZZ Z1 Z (1−√z)2 Z (1−√z−√x)2 ! !
zdx ∧ dy ∧ dz = z dy dx dz
R 0 0 0
Z1 Z (1−√z)2 !
√ √
= z (1 − z− x)2dx dz
0 0
Z
1 1 √
= z(1 − z)4 dz
6 0
1
= .
840
329 Example Prove that

ZZZ
a2bc
xdxdydz = ,
24
V
where V is the tetrahedron

x y z
V = (x, y, z) ∈ R3 : x ≥ 0, y ≥ 0, z ≥ 0, + + ≤ 1 .
a b c
Solution: We have
ZZZ Z c Z b−bz/c Z a−ay/b−az/c
xdxdydz = xdxdydz
0 0 0
V
Z c Z b−bz/c
1 ay az 2
= a− − dydz
2 0 0 b c
Z
1 c a2 (−z + c)3 b
= dx
6 0 c3
a2bc
=
24
160 Chapter 3
3.9 Change of Variables in ∧3(R3)

330 Example Find
ZZZ
(x + y + z)(x + y − z)(x − y − z)dx ∧ dy ∧ dz,
R
where R is the tetrahedron bounded by the planes x+y+z = 0, x+y−z = 0,

x − y − z = 0, and 2x − z = 1.
Solution: We make the change of variables
u = x + y + z =⇒ du = dx + dy + dz,
v = x + y − z =⇒ dv = dx + dy − dz,
w = x − y − z =⇒ dw = dx − dy − dz.
This gives
du ∧ dv ∧ dw = −4dx ∧ dy ∧ dz.
These forms have opposite orientations, so we choose, say,
du ∧ dw ∧ dv = 4dx ∧ dy ∧ dz
which have the same orientation. Also,
2x − z = 1 =⇒ u + v + 2w = 2.
The tetrahedron in the xyz-co-ördinate frame is mapped into a tetrahedron

bounded by u = 0, v = 0, u + v + 2w = 1 in the uvw-co-ördinate frame.
Z Z Z
1 2 1−v/2 2−v−2w 1
uvw du ∧ dw ∧ dv = .
4 0 0 0 180
Consider a transformation to cylindrical co-ördinates
(x, y, z) = (ρ cos θ, ρ sin θ, z).
From what we know about polar co-ördinates
dx ∧ dy = ρdρ ∧ dθ.
Since the wedge product of forms is associative,
dx ∧ dy ∧ dz = ρdρ ∧ dθ ∧ dz.
ZZZ
331 Example Find z2dx ∧ dy ∧ dz if
R
R = {(x, y, z) ∈ R3|x2 + y2 ≤ 1, 0 ≤ z ≤ 1}.
Solution: The region of integration is mapped into
∆ = [0; 2π] × [0; 1] × [0; 1]
through a cylindrical co-ördinate change. The integral is therefore

ZZZ Z 2π Z 1 Z 1
2
f(x, y, z)dx ∧ dy ∧ dz = dθ ρ dρ z dz
R 0 0 0
π
= .
3
ZZZ
332 Example Evaluate (x2 + y2)dx ∧ dy ∧ dz over the first octant region
D
bounded by the cylinders x2 + y2 = 1 and x2 + y2 = 4 and the planes
z = 0, z = 1, x = 0, x = y.

Z 1 Z π/2 Z 2
15π
ρ3dρ ∧ dθ ∧ dz = .
0 π/4 1 16
333 Example Three long cylinders of radius R intersect at right angles.

Find the volume of their intersection.
Solution: Let V be the desired volume. By symmetry, V = 24V 0 , where

ZZZ
0
V = dx ∧ dy ∧ dz,
D0
162 Chapter 3
D 0 = {(x, y, z) ∈ R3 : 0 ≤ y ≤ x, 0 ≤ z, x2+y2 ≤ R2, y2+z2 ≤ R2, z2+x2 ≤ R2}.

In this case it is easier to integrate with respect to z first. Using cylindrical
co-ördinates

h πi p
0
∆ = (θ, ρ, z) ∈ 0; × [0; R] × [0; +∞[, 0 ≤ z ≤ R − ρ cos θ .
2 2 2
4
Now,
Z π/4 Z R Z √R2 −ρ2 cos2 θ ! !
V0 = dz ρdρ dθ
0 0 0
Z π/4 Z R p
= ρ R2 − ρ2 cos2 θdρ dθ
Z0π/4 0
1 2 2 2 3/2 R

= −2
(R − ρ cos θ) 0
dθ
0 Z 3 cos θ
R3 π/4 1 − sin3 θ
= dθ
3 0 cos2 θ √
Z 2 !
= R3 π/4 2 1 − u2
u = cos θ [tan θ]0 + du
3 1 u2
R3 √22

−1
= 1− u +u 1
3
√
2−1 3
= √ R .
2
Finally √
V = 16V 0 = 8(2 − 2)R3.
Consider now a change to spherical co-ördinates
x = ρ cos θ sin φ, y = ρ sin θ sin φ, z = ρ cos φ.
We have
dx = cos θ sin φdρ − ρ sin θ sin φdθ + ρ cos θ cos φdφ,
dy = sin θ sin φdρ + ρ cos θ sin φdθ + ρ sin θ cos φdφ,
dz = cos φdρ − ρ sin φdφ.

This gives
dx ∧ dy ∧ dz = −ρ2 sin θdρ ∧ dθ ∧ dφ.
From this derivation, the form dρ ∧ dθ ∧ dφ is negatively oriented, and so
we choose
dx ∧ dy ∧ dz = ρ2 sin θdρ ∧ dφ ∧ dθ
instead.
ZZZ
3
334 Example Let (a, b, c) ∈]0; +∞[ be fixed. Find xyzdx ∧ dy ∧ dz if
R
2

3 x y2 z2
R = (x, y, z) ∈ R : 2 + 2 + 2 ≤ 1, x ≥ 0, y ≥ 0, z ≥ 0 .
a b c
We use spherical co-ördinates, where
(x, y, z) = (aρ cos θ sin φ, bρ sin θ sin φ, cρ cos φ).
We have
dx ∧ dy ∧ dz = abcρ2 sin φdρ ∧ dφ ∧ dρ.
The integration region is mapped into
π π
∆ = [0; 1] × [0; ] × [0; ].
2 2
Z π/2 ! Z
1 Z π/2 !
(abc)2
(abc)2 cos θ sin θ dθ ρ5 dρ cos3 φ sin φ dφ = .
0 0 0 48
335 Example Let V = {(x, y, z) ∈ R3 : x2 + y2 + z2 ≤ 9, 1 ≤ z ≤ 2}. Then

ZZZ Z 2π Z π/2−arcsin 1/3 Z 2/cos φ
dx ∧ dy ∧ dz = ρ2 sin φ dρ ∧ dφ ∧ dθ
0 π/2−arcsin 2/3 1/cos φ
V
63π
= .
4
164 Chapter 3
336 Lemma Let m, n be non-negative integers. Then

Z1
m!n!
xm(1 − x)n dx = .
0 (m + n + 1)!
337 Example (Putnam Exam 1984) Find

ZZZ
x1y9z8(1 − x − y − z)4 dx ∧ dy ∧ dz,
R
where
R = {(x, y, z) ∈ R3 : x ≥ 0, y ≥ 0, z ≥ 0, x + y + z ≤ 1}.
Solution: We make the change of variables
u = x + y + z =⇒ du = dx + dy + dz,
uv = y + z =⇒ udv + vdu = dy + dz,

uvw = z =⇒ uvdw + uwdv + vwdu = dz.
This gives
x = u(1 − v),
y = uv(1 − w),
z = uvw,
u2v du ∧ dv ∧ dw = dx ∧ dy ∧ dz.
To find the limits of integration we observe that the limits of integration
using dx ∧ dy ∧ dz are
0 ≤ z ≤ 1, 0 ≤ y ≤ 1 − z, 0 ≤ x ≤ 1 − y − z.
This translates into
0 ≤ uvw ≤ 1, 0 ≤ uv − uvw ≤ 1 − uvw, 0 ≤ u − uv ≤ 1 − uv + uvw − uvw.
Thus
0 ≤ uvw ≤ 1, 0 ≤ uv ≤ 1, 0 ≤ u ≤ 1,
Integration in ∧2(R3) 165
which finally give
0 ≤ u ≤ 1, 0 ≤ v ≤ 1, 0 ≤ w ≤ 1.
The integral sought is then, using Lemma 336,

Z1 Z1 Z1
u20v18w8(1 − u)4(1 − v)(1 − w)9 du ∧ dv ∧ dw,
0 0 0
which in turn is
Z 1 Z 1 Z 1
20 4 18 8 9 1 1 1
u (1 − u) du v (1 − v)dv w (1 − w) dw = · ·
0 0 0 265650 380 437580
which is
1
= .
44172388260000
3.10 Integration in ∧2(R3)

338 Definition A 2-dimensional oriented manifold of R3 is simply a smooth
surface D ∈ R3, where the + orientation is in the direction of the outward
normal pointing away from the origin and the − orientation is in the direc-
tion of the inward normal pointing towards the origin. A general oriented
2-manifold in R3 is a union of surfaces.
! The surface −Σ has opposite orientation to Σ and

Z Z
ω=− ω.
−Σ Σ

orientation for the regions considered. This corresponds to using the or-
dered basis
{dy ∧ dz, dz ∧ dx, dx ∧ dy}.
166 Chapter 3
339 Definition Let f : R3 → R. The integral of f over the smooth surface Σ

(oriented in the positive sense) is given by the expression
ZZ

fd2x.
Σ
Here 2 p
d x = (dx ∧ dy)2 + (dz ∧ dx)2 + (dy ∧ dz)2
is the surface area element.

ZZ

340 Example Evaluate zd2x where Σ is the outer surface of the sec-
Σ
tion of the paraboloid z = x2 + y2, 0 ≤ z ≤ 1.
Solution: We parametrise the paraboloid as follows. Let x = u, y = v, z =

u2 + v2. Observe that the domain D of Σ is the unit disk u2 + v2 ≤ 1. We
see that
dy ∧ dz = −2udu ∧ dv,
dz ∧ dx = −2vdu ∧ dv,
and so 2 p
d x = 1 + 4u2 + 4v2du ∧ dv.
Now, ZZ ZZ
p
zd2x = (u2 + v2) 1 + 4u2 + 4v2du ∧ dv.
Σ D
To evaluate this last integral we use polar co-ördinates, and so

ZZ p Z 2π Z 1 p
2 2 2 2
(u + v ) 1 + 4u + 4v du ∧ dv = ρ3 1 + 4ρ2dρdθ
0 0
D
π √ 1
= (5 5 + ).
12 5
ZZ

341 Example Evaluate yd2x where Σ is the surface z = x + y2, 0 ≤
Σ
x ≤ 1, 0 ≤ y ≤ 2.
Solution: We parametrise the surface by letting x = u, y = v, z = u + v2.

Observe that the domain D of Σ is the square [0; 1] × [0; 2]. Observe that
dy ∧ dz = −du ∧ dv,
dz ∧ dx = −2vdu ∧ dv,
and so 2 p
d x = 2 + 4v2du ∧ dv.
ZZ Z2 Z1 p
2
y d x =
v 2 + 4v2du ∧ dv
0 0
Σ
Z 1 Z 2 p
= du y 2 + 4v2dv
√
0 0
13 2
= .
3
ZZ

342 Example Evaluate x2d2x where Σ is the surface of the unit sphere
Σ
x2 + y2 + z2 = 1.
Solution: We use spherical co-ördinates, (x, y, z) = (cos θ sin φ, sin θ sin φ, cos φ).
Here θ ∈ [0; 2π] is the latitude and φ ∈ [0; π] is the longitude. Observe that
dx ∧ dy = sin φ cos φdφ ∧ dθ,
dy ∧ dz = cos θ sin2 φdφ ∧ dθ,

dz ∧ dx = − sin θ sin2 φdφ ∧ dθ,
and so 2
d x = sin φdφ ∧ dθ.
168 Chapter 3

ZZ Z 2π Z π
cos2 θ sin3 φdφ ∧ dθ

2 2
x d x =
0 0
Σ
4π
= .
3
ZZ p

343 Example Evaluate zd2x over the conical surface z = x2 + y2
S
between z = 0 and z = 1.
Solution: Put x = u, y = v, z2 = u2 + v2. Then
dx = du, dy = dv, zdz = udu + vdv,
whence
u v
dx ∧ dy = du ∧ dv, dy ∧ dz = − du ∧ dv, dz ∧ dx = − du ∧ dv,
z z
and so
2 p
d x = (dx ∧ dy)2 + (dz ∧ dx)2 + (dy ∧ dz)2
r
u2 + v2
= 1+ du ∧ dv
z2
√
= 2 du ∧ dv.
Hence
ZZ ZZ √
p √ √ Z 2π Z 1 2 2π 2
zd2x = u2 + v2 2 dudv = 2 ρ dρ ∧ dθ = .
0 0 3
Σ u2 +v2 ≤1
344 Example Find the area of that part of the cylinder x2 + y2 = 2y lying
inside the sphere x2 + y2 + z2 = 4.
Solution: We have
x2 + y2 = 2y ⇐⇒ x2 + (y − 1)2 = 1.
We parametrise the cylinder by putting x = cos u, y − 1 = sin u, and z = v.

Hence
dx = − sin udu, dy = cos udu, dz = dv,
whence
dx ∧ dy = 0, dy ∧ dz = cos udu ∧ dv, dz ∧ dx = sin udu ∧ dv,
and so
2 p
d x = (dx ∧ dy)2 + (dz ∧ dx)2 + (dy ∧ dz)2
p
= cos2 u + sin2 u du ∧ dv
= du ∧ dv.
The cylinder and the sphere intersect when x2 + y2 = 2y and x2 + y2 + z2 =

4, that is, when z2 = 4 − 2y, i.e. v2 = 4 − 2(1 + sin u) = 2 − 2 sin u. Also
0 ≤ u ≤ 2π. The integral is thus
ZZ Z 2π Z √2−2sin u Z 2π √
2
d x = dvdu = 2 2 − 2 sin udu
√
0 − 2−2sin u 0
Σ
√ Z 2π √
= 2 2 1 − sin u du
0
= 16.

ZZ
xdy ∧ dz + (z2 − zx)dz ∧ dx − xydx ∧ dy,
Σ
where Σ is the top side of the triangle with vertices at (2, 0, 0), (0, 2, 0),
(0, 0, 4).
Solution: Observe that the plane passing through the three given points
has equation 2x + 2y + z = 4. We project this plane onto the co-ördinate
170 Chapter 3
axes obtaining
ZZ Z 4 Z 2−z/2
8
xdy ∧ dz = (2 − y − z/2)dy ∧ dz = ,
0 0 3
Σ
ZZ Z 2 Z 4−2x
2
(z − zx)dz ∧ dx = (z2 − zx)dz ∧ dx = 8,
0 0
Σ
ZZ Z 2 Z 2−y
2
− xydx ∧ dy = − xydx ∧ dy = − ,
0 0 3
Σ
and hence
ZZ
xdy ∧ dz + (z2 − zx)dz ∧ dx − xydx ∧ dy = 10.
Σ

ZZ
xydy ∧ dz − x2dz ∧ dx + (x + z)dx ∧ dy,
Σ
where Σ is the top of the triangular region of the plane 2x + 2y + z = 6

bounded by the first octant.
Solution: We project this plane onto the co-ördinate axes obtaining

ZZ Z 6 Z 3−z/2
27
xydy ∧ dz = (3 − y − z/2)ydy ∧ dz = ,
0 0 4
Σ
ZZ Z 3 Z 6−2x
2 27
− x dz ∧ dx = − x2dz ∧ dx = − ,
0 0 2
Σ
ZZ Z 3 Z 3−y
27
(x + z)dx ∧ dy = (6 − x − 2y)dx ∧ dy = ,
0 0 2
Σ
and hence
ZZ
27
xydy ∧ dz − x2dz ∧ dx + (x + z)dx ∧ dy = .
4
Σ
Green’s, Stokes’, and Gauß’ Theorems 171
3.11 Green’s, Stokes’, and Gauß’ Theorems

We are now in position to state the general Stoke’s Theorem.
347 Theorem (General Stoke’s Theorem) Let M be a smooth oriented man-

ifold, having boundary ∂M. If ω is a differential form, then
Z Z
ω= dω.
∂M M
In R2, if ω is a 2-form, this takes the name of Green’s Theorem.

I
348 Example Evaluate (x−y3)dx+x3dy where C is the circle x2 +y2 = 1.
C
Solution: We will first use Green’s Theorem and then evaluate the integral
directly. We have
dω = d(x − y3) ∧ dx + d(x3) ∧ dy
= (dx − 3y2dy) ∧ dx + (3x2dx) ∧ dy
= (3y2 + 3x2)dx ∧ dy.
The region M is the area enclosed by the circle x2 + y2 = 1. Thus by

Green’s Theorem, and using polar co-ördinates,
I Z
3 3
(x − y )dx + x dy = (3y2 + 3x2)dx ∧ dy
C ZM2π Z 1
= 3ρ2ρdρ ∧ dθ
0 0
3π
= .
2
Aliter: We can evaluate this integral directly, again resorting to polar co-
ördinates.
I Z 2π
3 3
(x − y )dx + x dy = (cos θ − sin3 θ)(− sin θ)dθ + (cos3 θ)(cos θ)dθ
C Z02π
= (sin4 θ + cos4 θ − sin θ cos θ)dθ.
0
172 Chapter 3
To evaluate the last integral, observe that 1 = (sin2 θ + cos2 θ)2 = sin4 θ +
2 sin2 θ cos2 θ + cos4 θ, whence the integral equals
Z 2π Z 2π
4 4
(sin θ + cos θ − sin θ cos θ)dθ = (1 − 2 sin2 θ cos2 θ − sin θ cos θ)dθ
0 0
3π
= .
2
I
349 Example Evaluate x3ydx+xydy where C is the square with vertices
C
at (0, 0), (2, 0), (2, 2) and (0, 2).
Evaluating this directly would result in evaluating four path integrals, one
for each side of the square. We will use Green’s Theorem. We have
dω = d(x3y) ∧ dx + d(xy) ∧ dy
= (3x2ydx + x3dy) ∧ dx + (ydx + xdy) ∧ dy
= (y − x3)dx ∧ dy.
The region M is the area enclosed by the square. The integral equals
I Z2 Z2
3
x ydx + xydy = (y − x3)dx ∧ dy
C 0 0
= −4.
350 Example Consider the triangle 4 with vertices A : (0, 0), B : (1, 1),
C : (−2, 2).
➊ If LPQ denotes the equation of the line joining P and Q find LAB, LAC,
and LBC.
➋ Evaluate I
y2dx + xdy.
4
➌ Find ZZ
(1 − 2y)dx ∧ dy
D
where D is the interior of 4.
Solution:
1 4
➊ LAB is y = x; LAC is y = −x, and LBC is clearly y = − x + .
3 3
➋ We have
Z Z1
2 5
y dx + xdy = (x2 + x)dx =
6
ZAB Z0−2 2 !
1 4 1 15
y2dx + xdy = − x+ − x dx = −
BC 1 3 3 3 2
Z Z0
14
y2dx + xdy = (x2 − x)dx =
CA −2 3
Adding these integrals we find
I
y2dx + xdy = −2.
4
➌ We have
ZZ Z0 Z −x/3+4/3 !
(1 − 2y)dx ∧ dy = (1 − 2y)dy dx
−2 −x
D
Z 1 Z −x/3+4/3 !
+ (1 − 2y)dy dx
0 x
44 10
= − −
27 27
= −2.
351 Example Use Green’s Theorem to prove that

Z
(x2 + 2y3)dy = 16π,
Γ
where Γ is the circle (x − 2)2 + y2 = 4. Also, prove this directly by using a

path integral.
174 Chapter 3

d(x2 + 2y3) ∧ dy = 2xdx ∧ dy.
Hence by the generalised Stokes’ Theorem the integral equals
ZZ Z π/2 Z 4cos θ
2xdx ∧ dy = 2ρ2 cos θdρ ∧ dθ = 16π.
−π/2 0
{(x−2)2 +y2 ≤4}
To do it directly, put x − 2 = 2 cos t, y = 2 sin t, 0 ≤ t ≤ 2π. Then the

integral becomes
Z 2π Z 2π
2 3
((2 + 2 cos t) + 16 sin t)d2 sin t = (8 cos t + 16 cos2 t
0 0
+8 cos3 t + 32 cos t sin3 t)dt
= 16π.
In R3, if ω is a 2-form, the above theorem takes the name of Gauß’ or

the Divergence Theorem.
ZZ
352 Example Evaluate (x − y)dy ∧ dz + zdz ∧ dx − ydx ∧ dy where S is
S
the surface of the sphere x2 + y2 + z2 = 9 and the positive direction is the
outward normal.
Solution: The region M is the interior of the sphere x2 + y2 + z2 = 9. Now,
dω = (dx − dy) ∧ dy ∧ dz + dz ∧ dz ∧ dx − dy ∧ dx ∧ dy
= dx ∧ dy ∧ dz.

ZZZ
4π
dx ∧ dy ∧ dz = (27)
3
M
= 36π.
Aliter: We could evaluate this integral directly. We have

ZZ ZZ
(x − y)dy ∧ dz = xdy ∧ dz,
Σ Σ
since (x, y, z) 7→ −y is an odd function of y and the domain of integration

is symmetric with respect to y. Now,
ZZ Z 3 Z 2π p
xdy ∧ dz = |ρ| 9 − ρ2dρdθ
Σ −3 0
= 36π.
Also ZZ
zdz ∧ dx = 0,
Σ
since (x, y, z) 7→ z is an odd function of z and the domain of integration is
symmetric with respect to z. Similarly
ZZ
−ydx ∧ dy = 0,
Σ
since (x, y, z) 7→ −y is an odd function of y and the domain of integration

is symmetric with respect to y.
The classical Stokes’ Theorem occurs when ω is a 1-form in R3.
I
353 Example Evaluate ydx + (2x − z)dy + (z − x)dz where C is the in-
C
tersection of the sphere x2 + y2 + z2 = 4 and the plane z = 1.
Solution: We have
dω = (dy) ∧ dx + (2dx − dz) ∧ dy + (dz − dx) ∧ dz
= −dx ∧ dy + 2dx ∧ dy + dy ∧ dz + dz ∧ dx
= dx ∧ dy + dy ∧ dz + dz ∧ dx.
Since on C, z = 1, the surface Σ on which we are integrating is the inside

of the circle x2 + y2 + 1 = 4, i.e., x2 + y2 = 3. Also, z = 1 implies dz = 0
176 Chapter 3
and so ZZ ZZ
dω = dx ∧ dy.
Σ Σ
Since this is just the area of the circular region x2 + y2 ≤ 3, the integral
evaluates to ZZ
dx ∧ dy = 3π.
Σ
354 Example Let Γ denote the curve of intersection of the plane x + y = 2

and the sphere x2 − 2x + y2 − 2y + z2 = 0, oriented clockwise when viewed
from the origin. Use Stoke’s Theorem to prove that
Z √
ydx + zdy + xdz = −2π 2.
Γ
Prove this directly by parametrising the boundary of the surface and eval-
uating the path integral.
Solution: At the intersection path
0 = x2+y2+z2−2(x+y) = (2−y)2+y2+z2−4 = 2y2−4y+z2 = 2(y−1)2+z2−2,
which describes an ellipse on the yz-plane. Similarly we get 2(x−1)2+z2 =

2 on the xz-plane. We have
d (ydx + zdy + xdz) = dy∧dx+dz∧dy+dx∧dz = −dx∧dy−dy∧dz−dz∧dx.
Since dx ∧ dy = 0, by Stokes’ Theorem the integral sought is

ZZ ZZ √
− dydz − dzdx = −2π( 2).
2(y−1)2 +z2 ≤2 2(x−1)2 +z2 ≤2
(To evaluate the integrals you may resort to the fact that the area of the
(x − x0)2 (y − y0)2
elliptical region + ≤ 1 is πab).
a2 b2
If we were to evaluate this integral directly, we would set
√
y = 1 + cos θ, z = 2 sin θ, x = 2 − y = 1 − cos θ.

Z 2π √ √
(1 + cos θ)d(1 − cos θ) + 2 sin θd(1 + cos θ) + (1 − cos θ)d( 2 sin θ)
0
which in turn
Z 2π √ √ √
= sin θ + sin θ cos θ − 2 + 2 cos θdθ = −2π 2.
0

Multivariable Calculus Notes ©: David A. Santos

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Multivariable Calculus Notes ©: David A. Santos

Uploaded by

Copyright:

Available Formats

Multivariable Calculus Notes ©

Philadelphia, PA October 22, 2003

1 Brief Review of Linear Algebra 1

Figure 1.1: A point in R2 . Figure 1.2: A bi-point in R2 .

1 Definition Rn is the set of real n-tuples

These n-tuples are called points.

2 Example Consider the points

is an equivalence class denoting all those bi-points whose displacement in

5 Definition A real number α ∈ R will be called a scalar. If α ∈ R and

6 Theorem The operations of vector addition 1.1 and scalar multiplication

➊ Closure under vector addition:

➋ Closure under scalar multiplication:

➎ Existence of additive identity:

➏ Existence of additive inverses:

8 Definition An ordered set

of n linearly independent vectors in Rn is called an ordered basis for Rn.

! In most cases we will be using the standard ordered basis

(without the A subscript) rather than

forms an ordered basis for R3 and find the co-ördinates of

! As a shortcut, we often use the notation A = [a ] to denote the matrix

13 Definition We denote by Mm×n(R) the set of all m × n matrices with

14 Definition The n × n zero matrix 0n ∈ Mn(R) is the matrix with 0’s

15 Definition The n × n identity matrix In ∈ Mn(R) is the matrix with 1’s on

16 Definition The main diagonal of a matrix matrix A = [aij] ∈ Mm×n(R) is

17 Definition A ∈ Mm×n(R) is said to be upper triangular if

(∀(i, j) ∈ {1, 2, · · · , n}2), (i > j, aij = 0),

18 Example The matrix A ∈ M3×4(R) shewn is upper triangular and B ∈

19 Definition Let A = [aij] ∈ Mm×n(R), B = [bij] ∈ Mm×n(R) and α ∈ R.

21 Definition Let A = [aij] ∈ Mm×n(R) and B = [bij] ∈ Mn×p(F). Then

! In order to obtain the ij-th entry of the matrix AB we multiply elemen-

! Observe that we use juxtaposition rather than a special symbol to de-

! Matrix multiplication is not necessarily commutative.

24 Theorem If (A, B, C) ∈ Mm×n(F) × Mn×r(F) × Mr×s(F) we have

! Even though matrix multiplication is not necessarily commutative, it is

! By virtue of associativity, a square matrix commutes with its powers,

25 Definition Let A = [aij] ∈ Mn(R). Then the trace of A, denoted by tr (A)

26 Theorem Let A = [aij] ∈ Mn(R), B = [bij] ∈ Mn(R), α ∈ R. Then

tr (αA) = αtr (A) , (1.13)

tr (A + B) = tr (A) + tr (B) , (1.14)

tr (AB) = tr (BA) . (1.15)

whence the theorem follows. ❑

Solution: There are infinitely many solutions. Here is one:

28 Example Given a square matrix A ∈ M4(R) such that tr A2 = −4, and

= tr A2 − 2tr (A) + tr (I4)

29 Definition The transpose of a matrix of a matrix A = [aij] ∈ Mm×n(R) is

(A + αB)T = AT + αBT , (1.17)

whence the theorem follows. ❑

32 Definition A matrix A ∈ Mn(R) is symmetric if AT = A. A matrix B ∈

34 Theorem Any matrix A ∈ Mn(R) can be written as the sum of a sym-

Proof Observe that

35 Definition The determinant of a square matrix A ∈ Mn(R), denoted by

(the development along the i-th row) or, alternatively,

(the development along the j-th column).

! We have assumed that no matter which row or column we choose,

Solution: Since the second column has three 0’s, it is advantageous to

40 Theorem The eigenvalues of a real symmetric matrix are all real.

! We will se later that det A A ≥ 0 so the square root of it is defined.

42 Theorem The signed area of a triangle in R2 spanned by the vectors

1.4 Linear Transformations

! These two properties can be condensed by saying that a linear trans-

44 Example The trace map

is linear, since in view of Theorem 26,

tr (A + B) = tr (A) + tr (B) , tr (αA) = αtr (A) .