18.

022 Lecture notes

The course is divided into 6 parts:
Part 1 (Lectures I VII): Euclidean Spaces and Vector Algebra
Part 2 (VIII XIII): Differential Calculas for Scalar Fields and Functions of Several Real
Variables.
Part 3 (XIV XVII): Multiple Integrals
Part 4 (XVIII XXII): Line Integrals and Surface Integrals in Scalar and Vector Fields
Part 5 (XXIII XXVIII): Vector Integral Calculus in Two and Three Dimensions
Part 6 (XXIX XXXV): Linear Algebra in Multivariable Calculus
Lecture I
The Three-dimensional Space
We refer to the Euclidean two- and three-dimensional spaces as E
2
and E
3
, respectively.
Euclidean spaces have a measure of distance between points; for every two points P and
Q, we denote it by d(P, Q). This measure satisfies the following two laws:
i) For any two given points P and Q, d(P, Q) =0 if and only if P =Q.
ii) For any three given points P, Q and R, d (P, R) d (P,Q) +d (Q, R).
The following are some elementary facts in E
3
1) Given two distinct points P and Q, there is a unique line that contains both P and Q.
2) Given three non-collinear (not all contained by a single line) points P, Q, and R,
there is a unique plane that contains P, Q and R.
3) Given two distinct points P and Q on a plane M, then M contains the line
determined by P and Q.
4) Given two intersecting planes, their intersection is a line.
Any isometry is a mapping f from the points in E
3
onto the points in E
3
, such that, for
every pair of points P and Q, d(f(P), f(Q)) =d( P, Q). Two subsets of E
3
(subsets), are
said to congruent if there is an isometry which carries one figure onto the other.
Parallelism and Perpendicularity
Two lines are said to be parallel if they lie in common plane and they do not intersect.
Two lines are said to be skew if they do not lie in a common plane
Two planes that do not intersect are said to be parallel.
A line and a plane that do not intersect are said to be parallel
(1) For any line L and any point P not on L, there exists a unique line through P
parallel to L.
(2) For any plane M and any point P not on M, there exists a unique plane through P
parallel to M.

An angle is a figure formed by two rays (half-lines) with common vertex (end-point).
Two lines are perpendicular if they intersect and form right angles. A line L and a plane
M are perpendicular if they intersect at a single point P and L is perpendicular to every
line lying in M and going through P.
Projections
In E
2
, for a given line L and any point P, there exists a unique line L through P
perpendicular to L. The intersection point of L and L is called the projection of point P
on L. For any set S of points and any line L, the set of all projections on L of points in S
is called the projection of S on L.
In E
3
, for a given plane M and any point P, there exists a unique line going through P and
perpendicular to M. The intersection point of this line and M is called the projection of the
point P on the plane M. For any set S of points, the set of all projections on M of points in
S is called the projection of S on M.
For a given line L and any point P, there is a unique plane going through P and
perpendicular to L. The intersection point of L with this plane is called the projection of
the point P on the line L. For any set S of points, the set of all projections on L of points
in S is called the projection of S on L.
For more definitions and elementary facts in E
2
and E
3
, read the second part of Chapter 1
of the textbook.

|

|

|
Lecture II
Vectors and Vector Algebra
A set S of rays is called a direction if it satises the following laws:
(1) Any two rays in S have the same direction.
(2) Every ray that has the same direction as some member of S is in S.
A vector A consists of a non-negative real number, called the magnitude of
the

vector, and a direction. We denote the magnitude of

| A by A .
A vector

A such that A =1 is called a unit vector.

We will now describe four basic algebraic operations with vectors in E
3
:
1 Multiplication by a scalar
Let

Abythescalar Abyavector,andletcbyarealnumber. Multpying

c,weobtainavectordenotedbycA. Themagnitudeoftheresultisgiven
A

A if by cA =|c||

|. The direction of cA is the same as the direction of

c0, and opposite to the direction of

A if c<0.
2 Addition of vectors
The sum of two vectors A and B is denoted simply by A+

B. We can
denethissumgeometrically. Translate

B suchthatitsstartpointisthe
end-pointof

A+

A. Then

Bwillbethevectorhavingthesamestart-point
as A and the same end-point as

B.
3 Scalar product (Dot product)
The scalar product of two vectors A and

B is a scalar quantity denoted
B. Its value is

by

B=

B cos, where is the angle made by A |A||

|
B if we translate

and

B such that it has the same start-point as A. We

observe that B cos is the length of the projection of

B on A.
1

A
.
A
.

B
A
A+

B
Figure 1: Vector addition
The magnitude of A is equal to the square root of the dot product of A
with itself:
|A|=

A A
Hence

is a unit vector with the same direction as A.

|A|
4 Vector product (Cross product)
The cross product of two vectors

B is a vector denoted by

A and

AB.
The magnitude of the cross product is given by:

B = A B sin. |A

| |

|
Let A and B have the same start-point P and end-points Q
1
and Q
2
,
respectively. Let M be the plane of

B. The direction of

A and

AB is
normal to M in the manner established by the right-hand rule: if a right
handisplacedatP andthengersarecurlingfromPQ
1
toTQ
2
through
AB. theanglesmallerthan,thenthethumbindicatesthedirectionof

The following are some properties of these basic vector operations.
C =aC+b

1. (a+b)

C
C =a(b

2. (ab)

C)
B+

3. a(

C)=aB+aC
2
B
| | | | | | | |
|

A
.

A

.

B =B+

4. A+

A
B+

5. (

C)= B+ C

B

6. AB =

A
3

A
.

A
.

A
.


Lecture III
Vector Algebra in Cartesian Coordinates
Let us construct a Cartesian coordinates system in E
3
. First we choose a
point O, called the origin. Then we chose three mutually perpendicular rays
starting from O. These rays are called the positive x axis, positive y axis, and
positive z axis. Consider the lines containing these rays. For any of these lines,
everypointonitcanbeidentiedwitharealnumber: ifthepointisontheray,
the real number is the distance to O, if its not on the ray, the number is the
distance to O times 1. Let us denote these three lines by X,Y, and Z. Let P
beapointinspace. Considertheprojection-pointsofP onX,Y,andZ. These
pointsgivetheCartesiancoordinatesofP,denotedx
P
,y
P
,andz
P
. Anytriplet
ofrealnumbersformsthecoordinatesforsome pointP. Dierentpointshave
dierent coordinates.
The three unit vectors in the directions of the positive x,y, and z axes are
customarilydenotedby

j,and

AbeavectorinE
3
andletP bethepoint i,

k. Let

suchthatOP =A. Let(a
1
,a
2
,a
3
)bethecoordinatesofP. Considerthevectors

A
1
=a
1

i, A
2
=a
2
j, and A
3
=a
3
k. by vector addition and multiplication with
scalars, one obtains the following expression:

A
2
+

A=A
1
+

A
3
=a
1

i+a
2
j+a
3
k

Then a
1
,a
2
, and a
3
are called the scalar components of A, and a
1

i,a
2
j, and
a
3
k are called the vector components of

i,

k unit vectors, we

A. By using the

j,

obtain coordinate formulas for the four basic vector operations:

1. Multiplication by a scalar
cA=(ca
1
)

i+(ca
2
)

k,

j+(ca
3
)

for any scalar c.

1

2. Addition of vectors
If

B =b
1

i+b
2
j+b
3
k, then A=a
1

i+a
2
j+a
3
k and

A+

j+(a
3
+b
3
)

B =(a
1
+b
1
)

i+(a
2
+b
2
)

k.
3. Dot product

B = b
1

i+b
2
j+b
3
k, considering that

i = If A = a
1

i+a
2
j+a
3
k and



j

i=

i k=

i=

k j =0,wegetthat

k k=1and

i j =

k

jj =

jk=


B =a
1
b
1
+a
2
b
2
+a
3
b
3
.

4. Cross product
Let

B =b
1

i+b
2
j+b
3
k. To compute the cross A=a
1

i+a
2
j+a
3
k and

product, we use the following equalities:

i=

k

i jj =

k=0, and

ij =

i,

i=

ik=

k,

k=kj =

j. j

i=

j
We get the following formula:
AB =(a
2
b
3
a
3
b
2
)

i+(a
3
b
1
a
1
b
3
)

k.

j+(a
1
b
2
a
2
b
1
)

Using determinants one can easily remember the formula, since:

i j k

AB = a
1
a
2
a
3

b
1
b
2
b
3

A can be expressed as Considering that

i=

k k=1, we get that

jj =


A=( i)

j+( k)

i+( j)

k,
which is know as the frame identity.
Triple products
1. Scalar triple product
2

A
.

A
.
A
.

A
.

n

C = c
1

i+c
2
j+c
3
k. Let A = a
1

i+a
2
j+a
3
k, B = b
1

i+b
2
j+b
3
k,and

The triple product (

A B) is called a scalar triple product, since it is
a scalar quantity. Its value is given by:

a
1
a
2
a
3

B)
.

C = (

C)= b
1
b
2
b
3
.
c
1
c
2
c
3
Hencewecansimplywrite(

C]withoutspecifyingthe A B) as[A, ,

positions for the cross and dot signs. We will use the following equalities
in computing quadruple products:

B,

C,

A,

C,

B,

A,

[A,

C] = [B,

A] = [C,

B] =[A,

B] =[C,

A] =[B,

C].
2. Vector triple product
The cross product is not associative so we will give two formulas:
AB)C =(

A)

C
.
A, and (

C
.
B(

B)

A(

B( B)

BC)=( C)

C
Quadruple products
1. Scalar quadruple product
Theexpression(

B)
.
(

D)iscalledaquadruplescalarproduct,and A

C
by applying the formulas for triple products, we get the value:
(

B)
.
(

C
.
B
.
C
.
A

C D)=(

A)(

D)(

B)( D)
2. Vector quadruple product
The expression (

B)(

D) is called a quadruple vector product, A

C

and by applying the formulas for triple products, we get the value:
AB)(

D,

D,

(

C D)= [C,

A]B[C,

B]A
3

A

.

C
.

C
.

A
.
A
.

A
.

B
Lecture I
Analytic Geometry in E
2
and E
3
First we review some basic facts of analytic geometry in E
2
. Let us consider a
Cartesian coordinate system in E
2
. We denote by F[x, y] an algebraic formula
in the variables x and y. Any equation of the form F
1
[x, y] = F
2
[x, y] can
be reduced to an equation of the form F[x, y] = 0. We say that an equation
F[x, y] = 0 determines a set S if S is the set of all points in E
2
whose coordinates
satisfy the equation.
In E
2
, an equation is called linear if has the form Ax +By +C = 0. where
A, B, and C are real coecients, with A
2
+ B
2
> 0. Also, the following facts
are true:
(a) Every linear equation determines a unique straight line.
(b) Every straight line is determined by some linear equation.
Let

N = A

i +B

j. Then

N is normal to the line L determined by Ax+By +
C = 0. The vector

N =

N
|

N|
is a unit vector normal to the same line. Let O
be the origin of the Cartesian system, let P be a point position vector

R
P
.
Denote by d the distance from O to the line L. The following lemma holds:
Lemma 1 If P is on the line L then |

R
P


N| = d.
This result follows quickly from the equality

R
P


N = |

R
P
| cos , where is the
angle between

R
P
and

N.
We can use lemma 1 to nd the distance from O to the line L. Let R = (x, y)
be a point on L and let

R = x

i +y

j. Then |C| = |Ax +By| = |

R

N| = |

N|d =

A
2
+B
2
d. Hence we get the following result:
Theorem 1 The distance d between the origin O and the line determined by
Ax +By +C = 0 is given by d =
|C|

A
2
+B
2
.
1
V
with
In E
2
, an equation F[x, y] = is called a second-degree equation if F[x, y] =
Ax
2
+ By
2
+ Cxy + Dx + Ey + F, where A, B, C, D, E, and F are real and
A
2
+B
2
+C
2
> 0.
As we go from E
2
to E
3
, here is how the basic facts change:
An equation is called linear if it has the form Ax+By +Cz +D = 0, where
A, B, C, and D are real coecients, with A
2
+ B
2
+ C
2
> 0. In E
3
, a linear
equation determine a unique plane, not a line.
The vectors

N = A

i + B

j + C

k and

N =

N
|

N|
are normal to the plane M
determined by Ax + y+Cz +D = 0. Let d be the distance between the origin
O and the plane M. Consider a point P with position vector

R
P
. The following
properties hold:
Lemma 2 If P is in the plane M then |

R
P


N| = d.
Theorem 2 The distance d between the origin O and the plane M is given by:
d =
|D|

A
2
+B
2
+C
2
.
Let M and M

be two planes determined by the equations Ax+By+Cz+D = 0

and A

x +B

y +C

z +D

= 0. Then:
(a) If M and M

are parallel, then there exists a real c such that A = cA

,
B = cB

, and C = cC

.
(b) If M and M

arent parallel, the there exists a point P coordinates

(x, y, z) such that Ax +By +Cz +D = 0 and A

x +B

y +C

z +D

= 0.
In E
3
, a system of two linear equations for two non-parallel planes determines
a unique line. But more frequently parametric equations are used to determine
lines in E
3
.
The system of equations
x = a
1
+tb
1
, y = a
2
+tb
2
, z = a
3
+tb
3
with t in (, +) is called a system of linear scalar parametric equations
on (, +) in E
3
.
2
s
B
with
We say that a gure S is determined by a system of parametric equations
on (, +) if S is the set of all points in E
3
whose coordinates satisfy the
equations as the parameter t takes all real values. Then system of linear
scalar parametric equations in E
3
determines a unique line.
The vector interperpretation of this system of equations gives a clearer rea-
soning the property above. Let

R
0
be the position vector of the point P
0
coordinates (a
1
, a
2
, a
3
), and let

A = b
1

i +b
2

j +b
3

k. Then the set S of all points

R position vector

R = x

i + y

j + z

k such that

R =

R
0
+ t

A real t is a
line in E
3
.
3
each
for with
with for
Lecture V
Calculus of One-Variable Functions
Let us rst review some denitions in calculus on real numbers.
In order to dene limit on the real numbers we will use the concept of funnel
functions.
Denition 1 A function (t) on [0, d] is called a funnel function if it has the
following properties:
1. It is strictly increasing.
2. For every a > 0, there exists t, 0 < t d, such that 0 < (t) < a.
Denition 2 Let f(x) be a scalar function and let l be a real number. We say
that the limit of f at c is l and we denote this by
lim f(x) = l
x c
if for some d > 0 there exists a funnel function (t) on [0, d] such that for every
x with 0 < x c d it follows that f(x) l ( x c ). | | | | | | |
f (c +h)f (c )
For a real function f and c, h real numbers, the fraction
h
is called
dierence quotient and is denoted by
f
.
x
f (c +h)f (c )
Denition 3 For a real function f, if lim
h 0
h
exists and has a real

df
value, we call this value the derivative of f at c and denote it by f

(c) or
dt
|
c
.
We say that f is dierentiable at c.
Denition 4 A real function f is continuous at c if lim
x c
f(x) = f(c).

We say that f is continuous on [a, b] if f is continuous at all c [a, b]. We will

now see how the limit and derivative concepts act on the sum and product of
functions.
Let f and g be two real functions such that lim
x c
f(x) = l
1
and lim
x c
g(x) =

l
2
. Then :
1

1. lim
x c
(f + g)(x) = l
1
+ l
2
.

2. lim
x c
(fg)(x) = l
1
l
2
.

Let f and g be two real functions dierentiable at c. The two properties

below mean that the derivates exists and are equal with the given values.
1. (f + g)

(c) = f

(c) + g

(c).
2. (fg)

(c) = f(c)g

(c) + g(c)f

(c).
Denition 5 Let f(t) and F(t) be two real function. The function F(t) is
called the indenite integral of f(t) if F(t) is dierentiable and F

(t) = f(t) for

all t. We denote this by F(t) = f(t)dt + C. This is due to the fact that the
indenite integral of a function f is not unique, and the dierence between two
indenite integrals of the same function is always a constant function.
Let us now explain what the denite integral of f on [a, b] is. Take > 0
and divide [a, b] in subintervals [x
0
, x
1
], [x
1
, x
2
], . . . , [x
n1
, x
n
], smaller than ,
where x
0
= a and x
n
= b. Denote x
i
= x
i
x
i1
for all 0 < i n. Then
f(x
1
)x
1
+. . . +f(x
n
)x
n
is called a Riemann sum of mesh on [a, b] for the
functon f. If the limit of the Riemann sums exists as 0, then this limit is

b
called the denite integral of f on [a, b] and it is denoted by
a
f(t)dt.
We will now review how the denitions above extend to one-variable vector
functions.
Let

A(t) be a vector function on an interval [a, b].

Denition 6 For a vector L, if lim
tc
|A(t)

L|= 0, we say that the limit of

L and denote this by lim

t c
A(t) = L. A(t) as t goes to c is

Here are some basic facts about vector limit. Let

A(t) and B(t) be two

vector functions such that lim
t c
A(t) = L
1
and lim
t c
B(t) = L
2
. Then:

A(t) + B(t)) = L

1
+ L

2
. 1. lim
t c
(

2. If a(t) is a scalar function such that lim

t c
a(t) = l, then lim
t c
(a(t)

A(t)) =

lL

1
.
2


3. lim
t c
(

A(t) B(t)) = L

1
L

2
.

4. lim
t c
(

A(t) B(t)) = L
1
L
2
.

Denition 7 A vector function A(t) is continuous at c if lim

t c
A(t) = A(c).

A vector function

A(t) is said to be continuous on [a, b] if it is continuous

at all t [a, b]. If

A(t) and B(t) are continuous functions, then A(t) B(t) is a

continuous scalar function, and

B(t) and A(t)B(t) are continuous vector A(t)+

functions. Also, if a(t) is a continuous scalar function, a(t)

A(t) is a continuous
vector function.
Denition 8 Let

be the dierence quotient of a vector function

A
A(t). If
t

D is called the derivative of A(t) there exists D such that lim

tc
t
= D, then

at c. This is denoted by

A
A(t) is dierentiable at c. D =
d
dt

|
c
. We say that

A function

A(t) is called dierentiable on [a, b] if it is dierentiable at every

A(t) = a
1
(t)

i + a
2
(t)

+ a
3
(t)

c on [a, b]. Let

j k be a vector function. Then

a
1
(t), a
2
(t), and a
3
(t) are dierentiable on [a, b] if and only if

A(t) is dierentiable
on [a, b]. In this case,
d

A da
1
t

da
2
t

da
3
t

= i + j + k.
dt dt dt dt
Indenite and denite integrals for one-variable vector functions are dened
in the exact same manner as for scalar functions.
3
Lecture VI
Calculus of Vector Functions
R d
2

R
Recall that
d

denotes the rst-order derivative of

R(t), and that

dt
2
de-
dt
notes the second-order derivative of

R(t). We introduce new notations for these

R

R(t) = a
1
(t)

i + a
2
(t)

+ a
3
(t)

functions:
d

= R(t) and
d
2

= R(t). Let

j k.
dt dt
2
Then the following dierentiation rules stand:

j + a
3
(t)

1. R(t) = a
1
(t)

i + a
2
(t)

k.

a
1
(t)

i + j a
3
(t)

2. R(t) = a
2
(t)

+ k.
A(t) and

Let

B(t) be dierentiable vector functions, and let a(t) be a

dierentiable scalar function. The following dierentiation rules stand:
dt
(

3.
d
A(t) + B(t)) = A + B.
4.
d
A(t)) = a (t)

A(t) + a(t)A(t).
dt
(a(t)

dt
(

A B. 5.
d
A(t) B(t)) = A B +

6.
d
A(t)

B +

B(t)) = A

A B.
dt
(

R(t) be a unit vector function,

Theorem 1 (Unit-Vector Theorem) Let

R(t) =
u(t) for all t. Then:
u is perpendicular to that of 1. The direction of

u. The vector formula de-

scribing this is u(t) u

(t) = 0 for all t.

u(t)| equals the rate of change of the direction of 2. |

u(t) measured in radians

per time.
The proof for the rst part of the Unit-Vector Theorem is quite brief. Since
u(t)| = 1, u(t) u

(t) = | u(t) u(t) = 1. Using dierentiation rule 5, we get that 2

d1
= 0, i.e.

u(t). u(t) is perpendicular to

dt
1
Denition 1 Let

R(t) be a continuous vector function on an interval [a, b]. The

set C of all points having

R(t) as position vector for some t in [a, b] is called a

nite curve. The function

R(t) is called a path for C.

R
For a point P on a curve C, the vector
d

at P is called the velocity vector at

dt
R
P and is denoted (P). Similarly, the vector
d
2

at P is called the acceleration

dt
2
da R
(P). Let d

=
dt
=
d
3

vector at P and is denoted

dt
3
. We will now dene two
important scalar functions important for the curve C.
v(P )a(P )|
Denition 2 The curvature k is a function dened by k(P) =
|
|v(P )|
3
.
v(P ),

a(P ),d(P )]
Denition 3 The torsion is a function dened by (P) =
[
| a(P )|
2
.
v(P )
The curve C is uniquely determined by the functions k and .
2
v
a

Lecture VII
Paths and Curves
First we go through several basic notions about paths. Let

R(t) on [a, b] be a
given path.
Denition 1 R(t) is called elementary if for every pair (t
1
, t
2
), with t
1
and t
2
distinct in [a, b], R(t
1
) = R(t
2
).
Denition 2 R(t) is called simple if for every pair (t
1
, t
2
), with t
1
and t
2
distinct in [a, b], except possibly for the pair (a, b), R(t
1
) = R(t
2
).

R(a) = R(b). Denition 3 R(t) is called closed if

A closed and simple path is called a loop. A loop in E
2
is called a Jordan curve.
Given a Jordan curve C, E
2
can be divided into three regions, two bounded
and one unbounded. The bounded regions are C and R
i
, the interior of C. The
unbounded region is R
e
, the exterior of C.
Denition 4 A directed curve is a curve along which we have specied a di-
rection.
Given a curve C, nding a path for C lets us apply calculus techniques. If C is
R(t) = Rcos t

i+Rsin t

the circle of radius R and center O, then a path for C is

j.
Let C be the curve generated by a point P xed on a circle of radius a rolling
without slipping along the x axis. Suppose the curve begins with P at O and
ends when P again touches the x axis. The circle makes a full rotation, so it
would be advisable to take , the angle through which the circle rotates, as
the variable for a path for C. The coordinates of the center of the circle are
(a, a), so the coordinates of P are (a( sin ), a(1 cos )). Hence the path
R() = a( sin )

i + a(1 cos )

is

j, with from 0 to 2.
Let

R be a path for the curve C and let P be the point with position vector

R(t0)
=
R d

R(t
0
). Recall that lim
t 0
R(t0+t)

dt
| = v(P) is a tangent vector

t
t0
to C at P.
1

R
R(t
0
). If
d

Denition 5 Let P be a point with position vector

dt
| exists and
t0
is not equal to 0, then we denote by T

P
the unit tangent vector at P, the vector
t0
d

R d

R
T

P
=
dt dt t0

t0 d

R
Let s be the arc length of C, s(t
0
) =
a
|
dt
|dt. Then
d

R R ds R
v = = and T

=
d

, so
dt ds

dt ds
v = T

ds
= T

s
dt

Denition 6 For a point P on C, |

dT

at P is called the curvature of C at P

ds
|
and is denoted by k.
T
If
d

= 0 at P, then we can dene the unit normal vector to C Denition 7

ds

dT

dT

. at P, N
P
=
ds
|
ds
|
N =
dT

Clearly, k

ds
and N
P
is orthogonal to T

P
. Since v = sT

, we get that
a =
sT)

N
d(

= sT

+ sT. Since T

=
dT

=
dT

ds
= k

s, we have that
dt dt ds dt
a = sT

+ sT

= sT

+ ks
2

N
Denition 8 Let P be a point where T

P
and N
P
both exist. The vector

B
P
=

P
N
P
is called the unit binormal vector at P.

B
P
form a frame at P, that is called the moving frame The vectors T

P
, N
P
, and

at P.
2

| | | |

|
1
Lecture VIII
Scalar Fields
Cylindrical Coordinates
Scalar Fields
Denition 1 Let D be a subset of E
3
. A function f that associates each point
P in D to a real number f(P) is called a scalar eld. D is called the domain
of f.
Denition 2 Let f be a scalar eld on a domain D. Let u be a xed unit vector
and let P be a xed point. For any point Q such that PQ is parallel to

u, let

f(P)F(Q)
s(P, Q) = PQ u. Consider the limit lim
s 0
s(P,Q)
. If this limit exists,

then we denote it
df f
ds
|
u,P
= lim

s 0 s
and call it the directional derivative of f in the direction u at P.
Denition 3 Let f be a scalar eld on D, and let P(a, b) be a point in D.
The partial derivative of f with respect to x at P is the derivative at a of the
function f(x, b). It is denoted by
f
P
.
x
|
The partial derivative with respect to x equals the directional derivative in the
i :
f df
direction

x
|
P
=
ds
|
i,P
. Also, the partial derivative with respect to y equals

df
j :
f
the directional derivative in the direction

y
|
P
=
ds
|
j,P
. Since they are
dened on subsets of D, the partial derivatives can be viewed as scalar elds.

f
Let f(x, y) = x
2
+ cos xy. Then
x
= 2x y sin xy is dened on the same
domain as f.
Denition 4 Let f be a scalar eld on D and let P be a point in D. We say
the limit of f at P is l and denote this by lim
QP
f(Q) = l, if there exists a one-
variable function on a domain [0, d] such that (t) decreases to 0 as t decreases
to 0, and for all points Q = P such that |QP d, then f(Q) l ( QP ).
1
2 Cylindrical Coordinates
Let us dene a non-Cartesian coordinates system that will prove usefull in ap-
plying calculus techniques to certain functions: the cylindrical coordinates sys-
tem.The coordinates of a point P in this system are denoted r, and z. let us
rst take a Cartesian system in E
3
. To nd the cylindrical coordinates of a
point P, we rst let P

be the projection of P on the xy plane. Then r = |OP

|
and is the angle between Ox and OP

measured in the counterclockwise di-

rection starting from Ox. The third coordinate, z, has the same value as the
z coordinate from the Cartesian coordinate system. The dening equations of
the cylindrical coordinates system are
x = r cos , y = r sin , z = z.
For a point P of position vector

R, we write

j + zk R = x

i + yj + zk = (r cos )

i + (r sin )

2
Lecture IX
Linear Approximation
1 One-variable Functions
Let f be a one-variable function on a domain D. If f is dierentiable at x = c,
we say f has a linear approximation, which we dene in the following way.
Denition 1 If f is a function dierentiable at c, for any x dene f =
f(x) f(c) and x = x c. There exist a scalar A
c
and a function
c
(x) such
that
f = A
c
x +
c
(x)x and lim
c
(x) = 0.
x c
Then
f
A
c
= f

(c) and
c
(x) =
x
f

(c).
We say that f has a linear approximation at c and that the expression f
app
=
f

(c)x is a linear approximation formula for f at c. The function

c
is called
the relative error function for f at c.
2 Multivariable Functions
Denition 2 Let f be a multilinear function and let P be a point in its domain
D in E
2
. Dene f = f(a + x, b + y) f(a, b). We say that f has linear
approximation at P if there exist scalars A, B and functions
1
,
2
such that:
1. for all Q D,f = Ax + By +
1
(Q)x +
2
(Q)y.
2. lim
QP

1
(Q) = 0 and lim
QP

2
(Q) = 0.
Theorem 1 If f has a linear approximation at P, then f has partial derivatives
at P and
f
y
|
x
|
P
= A,
f
P
= B.
1
Theorem 2 If f has continuous partial derivatives on all the domain, then f
has a linear approximation.
For a multivariable function f that has a linear approximation at P, we
f
dene f
app
=
f
P
x +
y
|
P
y, the linear approximation formula for f at
x
|
P.
f f
For a three-variable function f, we say that f has a linear approximation
at P if there exist A, B, C scalars and
1
,
2
,
3
functions such that f =
Ax + By + Cz +
1
(Q)x +
2
(Q)y +
3
(Q)z and lim
QP

1
(Q) = 0,
lim
QP

2
(Q) = 0, lim
QP

3
(Q) = 0. If f has a linear approximation at P,
then f has partial derivatives at P, and A =
x
|
P
, B =
f
P
, C =
z
|
P
. The
y
|
f
P
x+
f
P
y+
f
linear approximation formula for f at P is f
app
=
x
|
y
|
z
|
P
z.
2

1
Lecture X
Linear Approximation
Chain Rule
Linear Approximation; Gradient
We say that a function has a linear approximation on a domain D if it has a
linear approximation at any point P D.
Theorem 1 If f has a linear approximation on a domain D, then f is contin-
uous on D.
Theorem 2 If f has continuous partial derivatives on D, then f has linear
approximation on D.
Denition 1 Let f be a function with partial derivatives at P. The vector

G
P
=

i +
f
j + k
x y z
is called the gradient at P and is usually denoted by

P
. f|
Theorem 3 (Gradient Theorem) Let f have domain D in E
3
and let P be
a point in D. If u is a xed unit vector, then
df
=
f

i +
f f

j +

k =

u f
P
|
ds u,P x P y P z P
Chain Rule
Physical variables, quantities that can be measured in a given physical system,
are related in various ways. For example, pressure p is a function of volume V
and temperature T, p = h(V, T). Also, T depends on V and internal energy U,
T = t(V, U). From these to equalities, we get that p is a function of volume V
1
2

and internal energy U, p = h(V, T) = h(V, t(V, U)) = g(V, U). So to nd the
partial derivative of p with respect to V , we have to speciy the function we
=
h
are referring to: h or g. We write (
p
=
g
and (
p
. Below we
V
)
U
V V
)
T
V
will see how such partial derivatives are related. By (
f
x
)
y
we mean the partial
derivative of f with respect to x when f is seen as a function of variables x and
y.
Theorem 4 (Chain Rule) Let w, u, v, x, y be physical variables such that there
exist dierentiable functions f, g, h such that w = f(u, v), u = g(x, y) and
v = h(x, y). Then

w
=
f g
+
f h
x
y
u x v x

w
=
f g
+
f h
y
x
u y v y
If we take w = f(u, v), u = g(x, y), and v = h(x), then applying the Chain Rule
we get
w f g
= ,
y
x
u y
h
since = 0.
y
2


Lecture XI
Chain Rule: Elimination Method
Let w = f(x, y) be a dierentiable function of x and y. The linear approximation
of f is given by
f
app
= f
x
(x, y)x + f
y
(x, y)y.
We introduce a new notation, the dierential notation for the increments f,
x, y, namely we write df, dx, dy instead: df = f
x
dx + f
y
dy. This expression
is called the dierential of f. For example, the dierential of
2
w = x
2
+ y 1 is dw = 2xdx + 2ydy.
For any function w = f(x, y), the equality
w w
dw = dx + dy
x
y
y
x
holds. By the elimination method, we can nd (
w
x
)
y
and (
w
if w is not given
y
)
x
directly as a function of x and y but can be reduced to such a function. We
ilustrate this in the following example. Consider the these two equalities:
e
xy
w = f(x, y, z) = xyz, z = g(x, y) = .
Then the dierentials of w and z are
dw = yzdx + xzdy + xydz and dz = ye
xy
dx + xe
xy
dy.
Substituting dz in the rst equality, we get
2 2
dw = (yz + xy e
xy
)dx + (xz + yx e
xy
)dy.
Then the derivative of w with respect to x when w is seen as a function of x
and y is precisely th term of dx in the equality above:
w
= yz + xy
2
e
xy
and
w
= xz + yx
2
e
xy
.
x
y
y
x
1
Lecture XII
Terminology for Point-Sets in Euclidean Spaces and
Minimum-Maximum Theorems
First let us take a short look at a problem that was on the exam. We are
given a level curve (in E
2
) or a surface(in E
3
) and a point P on that curve or
surface. How to nd a vector normal to that curve or surface at that point P?
Let us consider the case of E
3
. The answer lies in the gradient of the function
dening the surface. The graph of the surface is given by z = f(x, y), hence
z f(x, y) = g(x, y, z) = 0. The gradient of g at P is a vector normal to the
surface at P:
f f

P
=
P
j +

k. g|
x
|
P

i
y
|
Now let us take a look at some basic notions of point-set topology of Eu-
clidean spaces.
Denition 1 Given a point P, we dene a neighborhood of P in the following
manner:
(i) For the 1-dimensional space, a neighborhood is an interval of the form
[c r, c + r] for some r > 0, where c is the point P.
(ii) For the 2-dimensional space, a neighorhood is a disc of points with center
at P and radius r.
(iii) For the 3-dimensional space, a neighborhood is a solid ball of radius r and
center P.
In all three cases, a neighborhood is a set U for which there exists r R such
that U is the set of all points Q with PQ r.
Let D be a given set of points.
Denition 2 A point P is an interior point of D if there is some neighborhood
of P which is entirely contained in D.
1
Denition 3 A point P is a boundary point of D if every neighborhood of P
contains at least one point in D and at least one point not in D.
Denition 4 D is called open if every point in D is an interior point of D.
D is called closed if it contains all its boundary points.
Note: D is open if and only if D contains none of its boundary points.
Denition 5 D is bounded if there is some point P and some neighborhood of
P such that the neighborhood contains the set D.
Denition 6 D is compact if D is closed and bounded.
If f is a path for D, how do we nd points P in D where f has a maximum value
or a minimum value? We investigate this question, providing some methods of
solving it. In doing so, we will consider the input(i.e. the points we are working
with) to be of the form x, contained in the domain of f, and not of the form P,
contained in D.
Denition 7 Let f be dened on [(a, b)] and let x be a point in its denition
domain.
(i) x is a global maximum point if f(x) f(y) for all y in the denition
domain.
(ii) x is a global minimum point if f(x) f(y) for all y in the denition
domain.
(iii) x is a local maximum point if there exists a neighobrhood U of x such
that f(x) f(y) for all y in intersection of U and the denition domain.
(iv) x is a local minimum point if there exists a neighobrhood U of x such that
f(x) f(y) for all y in intersection of U and the denition domain.
All maximum and minimum points, global and local, are called extreme points
of f. The two following theorems are useful in nding extreme points.
Theorem 1 (Min-Max Existence Theorem) Let f be a continuous func-
tion, having D as its domain. Assume D is compact.Then f has at least one
global maximum point and at least one global minimum point in D.
2
Denition 8 Let f be a dierentiable function. If

P
= 0, then P is called f|
a critical point.
Theorem 2 (Critical Point Theorem) Let f be a dierentiable function with
a domain D and let P be an interior point of D. If P is an extreme point, then
P is a critical point.
3

Lecture XIII
Two-Variable Test
Constrained Maximum-Minimum Problems
1 The two-variable test
Recall that by the Critical Point Theorem, onlyif the gradient of a function f
at P is 0 (i.e. P is a critical point for f), can P be an extreme point of f.
Let f(x, y) = x
3
+ xy + y
3
. Then
f = (3x
2
+ y)

i + (x + 3y
2
)

j,
so if P = (x, y) is a critical point, then
2 2
x = 3y , y = 3x ,
1 1
hence x + 27x
4
= 0, so the only critical points are P
1
(0, 0) and P
2
(
3
,
3
).
Now we have to nd out if these points are extreme points. For this we can use
the following test.
Theorem 1 (The two-variable test) Let f be a two-variable C
2
function,
and let P be an interior critical point for f. We dene the functions H
1
and
H
2
, called Hessian functions in the following manner:
H
1
(P) = f
xx
(P),
H
2
(P) =
f
xx
(P) f
xy
(P)
f
yx
(P) f
yy
(P)
= f
xx
(P)f
yy
(P) f
2
xy
(P)
Then:
(a) If H
1
(P) > 0 and H
2
(P) > 0 then P is a local minimum point for f.
(b) If H
1
(P) < 0 and H
2
(P) > 0 then P is a local maximum point for f.
1
(c) If H
2
(P) < 0 then P is a saddle point for f.
Going back to the function f(x, y) = x
3
+ xy + y
3
, we nd that H
1
(x, y) = 6x
and H
2
(x, y) = 36xy 1. Hence H
1
(P
1
) = 0, H
2
(P
1
) = 1, so P
1
(0, 0) is a
1 1
saddle point. Also, H
1
(P
2
) = 2 and H
2
(P
2
) = 3, so P
2
(
3
,
3
) is a local
maximum point.
2 Constrained problems
Suppose f is a function dened on a region D in E
3
and C is a surface or a
curve in D. We can think of C as a new domain for f and try to nd extreme
point for f in C. Such a problem is a called a constrained problem.
Let f and g be C
1
and let D

be the set of all points P such that g(P) = c.
Assume

P
= 0 at all points P in D

. Let D

be the new restricted domain f|
of f. Such a case is called a one constraint case, since the restricted domain is
given by one function, g.
Denition 1 Let P be a point in D

. We call P a constrained critical point

f|
P
=
P

P
. for f on D

if there is a scalar
P
such that

g|
f|
P
= 0, then

Note that for a constrained critical point P, if

|
P
is normal
to D

at P.
Theorem 2 If P is a extreme point for f on D

, then P must be a constrained

critical point for f on D

.
The one-constraint Lagrange method of nding extreme points on D

uses the
theorem above in the following manner. First we dene a function h on D

,
h(x, y, z, ) = f(x, y, z) g(x, y, z). Then we form the system of equations
h
x
(x, y, z, ) = 0, h
y
(x, y, z, ) = 0, h
z
(x, y, z, ) = 0, g(x, y, z) = c
and solve it. Its solutions are the constrained critical points of f. Finally, among
these points we identify the extreme points for f on D

.
The case of two constraints can be described as follows. Let f, g
1
and g
2
be C
1
functions. Let D

be the set of all points P such that g
1
(P) = c
1
and
g
2
(P) = c
2
. Assume

g
2 P
= 0 for all points P in D

. g
1
|
P

|
2
Denition 2 Let P be a point in D

. We call P a constrained critical point for

f on D

if there exist scalars
P
and
P
such that

1
g
1 P
+
P

P
.
1 2
|
P
=
P

|
2
g
2
|
Theorem 3 If P is a extreme point for f on D

, then P must be a constrained

critical point for f on D

.
The Lagrange method for two constraints is similar to the method for one con-
straint. We dene h(x, y, z,
1
,
2
) = f(x, y, z)
1
g
1
(x, y, z)
2
g
2
(x, y, z).
We form the system of equations
h
x
(x, y, z,
1
,
2
) = 0, h
y
(x, y, z,
1
,
2
) = 0, h
z
(x, y, z,
1
,
2
) = 0,
g
1
(x, y, z) = c
1
g
2
(x, y, z) = c
2
and solve it. The solutions will be the constrained critical points. From among
these points, we identify the extreme points of f on D

.
3
Lecture XIV
Multiple Integrals
1 Integrals of one-variable functions
For a real-valued function f dened on an interval [a, b], the integral of f over

b
[a, b], denoted by
a
fdx, is dened as follows:
n

fdx = lim f(x

i
)x
i
,

n,max xi0
a
i=1
where a = x
1
< . . . < x
i
< x
i+1
< x
n
= b, x
i
= x
i
x
i1
and <
x

i
[x
i1
, x
i
] for all i 2, i n. We will now give a few properties and
theorems about integrals in one-variable calculus.
Theorem 1 (First Existence Theorem) Let f be a one-variable function.

b
If f is continuous on [a, b], then
a
fdx exists.

b
Let f and g be two functions dened on [a, b] such that
a
fdx and gdx exist.
a
Then the following properties hold:
1. Let c be such that a < c < b. Then the following integrals exist and the
equality holds:

b
fdx = fdx + fdx.
a a c
2. The integral in the left side of the equality exists and the equality holds:

b
(f + g)dx = fdx + gdx.
a a a
3. If c is a constant, the integral in the left side of the equality exists and the
equality holds:

b
(cf)dx = c fdx.
a a
1

2
4. If m
1
and m
2
are two constants such that m
1
f(x) m
2
for all x [a, b],
then

b
m
1
(b a) fdx m
2
(b a).
a
Scalar-valued integrals in E
2
and E
3
If f is a scalar-valued function dened on E
2
or E
3
, how do we dene the
integrals of f on a region R? We can only do this on a special kind of region,
called regular, which we will dene in section 5.
Denition 1 Let R be a regular region in E
2
and f a scalar eld on R. The
integral of f on R, denoted by
R
fdA, is dened as follows:

n
fdA = lim f(P
i

)A
i
,
R
nmax di0
i=1
where A
1
, . . . , A
n
form a subdivision of R into elementary regions, A
i
is the
area of A
i
, P
i

is a point in A
i
and d
i
is the diameter of A
i
, i.e. the longest
distance between two points in A
i
, for all i 1, i n.
Theorem 2 (Second Existence Theorem) If R is a regular region in E
2
and f is continuous on R, then
R
fdA exists.
For E
3
, the denition of the integral is very similar.
Denition 2 Let R be a regular region in E
3
and f a scalar eld on R. The
integral of f on R, denoted by
R
fdV , is dened as follows:

n
fdV = lim f(P
i

)V
i
,
R
nmax di0
i=1
where V
1
, . . . , V
n
form a subdivision of R into elementary regions, V
i
is the
volume of V
i
, P
i

is a point in V
i
and d
i
is the diameter of V
i
, for all i 1, i n.
Theorem 3 (Third Existence Theorem) If R is a regular region in E
3
and
f is continuous on R, then
R
fdV exists.
Remark: If we consider a constant function f(x, y) = 1 over a region R, then
fdA = dA is equal to the area of R.
R R
2

Sometimes it is useful not to use the Cartesian coordinate system. For
example, let R be the disc bounded by the curve x
2
+ y
2
= 4. Let f(x, y) =
2
x
2
+ y . We want to nd the integral
R
f(x, y)dA. We divide this disc into
n concentric rings, and further divide these rings by radial segments. For each
ring j , we choose a r
j

such that the area of the ring is 2r

r
j
, where r
j
j
is the dierence between the radii of rings j and j 1. Then we can choose a
point P

in each subregion of the ring j such that f(P

) = (r
j

)
2
. Hence the
m
Reimann sum for this choice of subdivision is 2(r
j

)
3
r
j
. Now we can
j=1
view this sum as the Reimann sum of a function of r. Hence

2
x
2
+ y
2
dA = 2r
3
dr = 8.
R 0
3 Properties of multiple integrals
1. Let R be a regular region in E
3
and let R
1
and R
2
be two regular disjoint
regions such that R = R
1
R
2
. If fdV exists, then the other
R
integrals in the following equality exists, and the equality holds:

fdV = fdV + fdV.
R R1 R2
The equivalent for E
2
holds as well.
Here is an application of this property. Let f(x, y) = x for all x, y, and let
R be the disc bounded by x
2
+y
2
= 1. Let R
1
be the part of R that is left
of Oy, and let R
2
be the part of R that is right of Oy. Then
R
fDA =


R1
fdA + fdA. But by symmetry,
R1
xdA =
R2
xdA, so
R2
fDA = 0.
R
2. Let R be a regular region in E
3
and f a function on R such that
R
fdV
exists. If m
1
and m
2
are two constants such that m
1
f(P) m
2
for all
points P in R, then the following inequalities hold:

m
1
V fdV =m
2
V,
R
where V is the volume of R. The equivalent for E
2
also holds.
As an application of this property, consider the region R to be the square
having (0, 0) and (2, 2) opposite vertices. Let f(x, y) = 2 +
xy
for all
100
3

(x, y) R. Then we can aproximate f by 2 f(x, y) 2.04 for all

(x, y) R. Hence by the property above, since the area of R is equal to
4, we have that 8 fdA 8.16 .
R
3. Let R be a regular region in E
3
and let f, g be two functions on R such

that fdV and
R
gdV exist. Then the integral on the left side
R
of the following equality exists and the equality holds:

(f + g)dV = fdV + gdV.
R R R
The equivalent property for E
2
holds as well.
2
Let us consider R to be the disc bounded by x
2
+y
2
= 4. Let f(x, y) = x
and g(x, y) = y
2
be two functions dened on R. By symmetry,
R
fdA =
R
gdA, so applying the property above, we get that
1
x
2
dA = (x
2
+ y
2
)dA.
2
R R
As weve seen in the example at the end of section 2,
R
(x
2
+y
2
)dA = 8,
hence

x
2
dA = 4.
4. Let R be a regular region in E
3
and f a function on R such that
R
fdV
exists. If c is a constant, the integral in the left side of the equality exists
and the equality holds:

(cf)dV = c fdV.
R R
4 Vector-valued integrals
Let

F(x, y, z) be a vector-valued function dened on E

3
, F(x, y, z) = L(x, y, z)

i+
M(x, y, z)

k. The integral of

j + N(x, y, z)

F exists if the integrals of L, M, and

N exist, and then the following equality holds:
F(x, y, z)dV =
R


LdV

i + MdV j + NdV k.
4

5
As an application, consider an object that ocupies a region R in space and has
density (x, y, z) at any point P(x, y, z). The center of mass of this object s
given by
1

C
M
= (x, y, z)OP(x, y, z)dV,
M
R
where M is the total mass of the object, i. e. M =
R
(x, y, z)dV . Hence

C
M
=
1
(x, y, z)(x

i + yj + zk)dV =
M
R
1
= x(x, y, z)dV

i+
M
R

1
j + k. + y(x, y, z)dV

1
z(x, y, z)dV

M
R
M
R
Further remarks
In the existence theorems, the condition that f be continuous can be replaced
with piecewise continuous.
Denition 3 A function f is said to be piecewise continuous on a regular region
R if it is dened on R and it is possible to divide R in a nite number of regular
subregions R
1
, . . . , R
n
such that for each subregion R
i
, the restiction of f to the
interior of R
i
can be extended to a function f
i
dened on the boundary of R
i
as
well, such that f
i
is continuous. It is not necessary that f
i
coincide with f on
the boundary of R
i
.
Now let us dene the concept of regular region. For this we will need to dene
rst what an elementary region is.
Denition 4 A region R is called elementary if there is a piecewise smooth,
simple, closed curve such that R is the union of the curve and its interior.
Denition 5 A region R is called regular if it can be divided into a nite num-
ber of elementary regions in the following fashion: the regions form a connected
whole and neighboring regions touch along a unique segment of common bound-
ary curve, that has positive length.
5
2
Let R be the region in E
2
dened by all points (x, y) such that x
2
+ y 1
and both x and y are irrational numbers. Then R is an example of a region
that is not regular.
6
Lecture XV
Iterated Integrals
In this lecture we look at methods to compute multiple integrals, introducing
iterated integrals. First, let us dene the type of regions for which it can be used.
These regions are called simple regions.
Denition 1 A region R in E
2
is called simple if it has one of the following
properties:
(i) There exists an interval [a, b] and there exist continuous functions g
1
(x)
and g
2
(x) dened on [a, b] such that g
1
(x) g
2
(x) for all x [a, b] and
the region R is the set of points (x, y) such that a x b and g
1
(x)
y g
2
(y). A region with this property is called y-simple.
(ii) There exists an interval [c, d] and there exist continuous functions h
1
(y)
and h
2
(y) dened on [c, d] such that h
1
(y) h
2
(y) for all x [c, d] and
the region R is the set of points (x, y) such that c y d and h
1
(y)
x h
2
(x). A region with this property is called x-simple.
Denition 2 Let R be an y-simple region in E
2
, with g
1
(x) and g
2
(x) its lower
and upper boundary functions, as in Denition 1(i). Let f be a continuous
function on R. For each x [a, b] we dene h(x) by

g2(x)
h(x) = f(x, y)dy
g1(x)

b
Then we compute the integral of h on [a, b],
a
h(x)dx. This two-step operation
is called a double iterated integral. The order of integration is rst y, then x.
Alternatively, we can denote the iterated integral by

g2(x)
f(x, y)dydx.
a g1(x)
In the same manner we can dene iterated integrals for x-simple regions. We
say then that the order of integration is rst x then y.
1

Let the region R be the set of all points (x, y) with 0 x 1, x
2
y x. We
dene f as the function on R such that f(x, y) = xy for all (x, y) R. Let us
compute the iterated integral of f:

1

1

x
x
3
x
5
1 1 1
xydydx = dx =
2

2 8

12
=
24
.
We can see that R is also x-simple, with 0 y 1 and h
1
(y) = y x

y =
h
2
(y). Hence we can compute the iterated integral of f in the order x, then y.

1

0 x
2
0
y
2
y
3
1 1 1
xydxdy = dy =
2

2 6

8
=
24
.
0 y 0
It is not by chance that the two integrals have the same value. In fact this can
be proved for any continuous function f on a x, y-simple region R.
Theorem 1 (Fubinis Theorem) Let R be a x, y-simple region and let f be
a continuous function on R. Then

b

g2(x)

h2(y)
fdA = f(x, y)dydx = f(x, y)dxdy,
R a g1(x) c h1(y)
where a, b, c, d and the functions g
1
, g
2
, h
1
, h
2
are as in Denition 1.
The notion of iterated integrals can be extended to E
3
. Consider a region R
in E
3
such that the projection R

of R into the xy plane is y-simple, and there

exist continuous functions h
1
(x, y) and h
2
(x, y) such that R is the set of all
points (x, y, z) with a x b, g
1
(x) y g
2
(x), and h
1
(x, y) z h
2
(x, y),
where a, b, g
1
, and g
2
are as in Deniton 1(i).Then the triple iterated integral
of a function f on R in the order z, y, x is

g2(x)

h2(x,y)
f(x, y, z)dz dy dx.
a g1(x) h1(x,y)
In E
3
, we can use two methods to compute triple integrals.
(i) Method of slices
Let f be a function on a region R. We dene R
z
to be the intersection
of R with the plane parallel to the xy plane, that intersects the z axis at

z. If the double integral
Rz
f(x, y, z)dA is a continuous function of f,
then

c2
fdV = f(x, y, z)dA dz,
R c1 Rz
where c
1
and c
2
are the minimum and maximum values for z in R.
2

(ii) Method of rods

Let be a function on a region R and let R

be the projection of R in
the xy plane. For each (x, y) R

we dene h
!
(x, y) and h
2
(x, y) to be
the minimal and maximal value that z takes when (x, y, z) R. If R
coincides wtih the set of all points (x, y, z) with (x, y) R

and h
1
(x, y)
z h
@
(x, y), then

h2(x,y)
fdV = f(x, y, z)dz dA.
R R

h1(x,y)
3

Lecture XVI
Integrals in Polar, Cylindrical, or Spherical Coordinates
Usually, we write functions in the Cartesian coordinate system. Hence we
write and compute multiple integrals in Cartesian coordinates. But there are
other coordinate systems that can help us compute iterated integrals faster. We
analyze bellow three such coordinate systems.
1 Polar coordinates
In E
2
, the polar coordinates system is often used along with a Cartesian system.
The polar coordinates of a point P(x, y) are r and where r is the distance from
the origin O and is the angle done by OP and Ox measured counter-clockwise
from Ox. The equations we use to switch from one system to the other are:
x = r cos , y = r sin , r = x
2
+ y
2
.
To transform an integral
R
f(x, y)dA from Cartesian coordinates to polar
coordinates, we take a look at Reimann sums. By consider a particular type of
subregions, we get that dA = rdrd, so we can write:

f(x, y)dA = f(r cos , r sin )rdrd,

R R
where

R is the region corresponding to R in a plane with Cartesian coordinates

where the axes are r and . This plane is called the r plane. Now if

R is
r-simple, we can compute the iterated integral in polar coordinates.
For example, let us compute the area of the region R enclosed by the curve
r = 1 + cos , where goes from 0 to 2. By looking at the graph of the curve,
we can see that the area is equal to twice the area enclosed when goes from
0 to . In polar coordinates, we can see that

R is a simple region, so we can

write:

1+cos
dA = 2 rdrd
R 0 0
1

2 1+cos
= (r | )d
0
0

= 1 + 2 cos + cos
2
d
0

= + cos
2
d
0
Knowing that for any n Z, the following equality holds

n
cos
2
d = sin
2
d,
0 0
and that cos
2
+ sin
2
= 1 for all , we get that

cos
2
d =
2
, so
0


3
dA = + cos
2
d =
2
.
R 0
2 Cylindrical coordinates
In E
3
, it is easier sometimes to use the cylindrical coordinates system. We can
view this system as the extension of the polar coordinates system in E
3
. Let
P(x, y, z) be a point in E
3
and let P

be its projection on the xy plane. The

cylindrical coordinates of P are r, , and z, where r and are the polar coordi-
nates of P

in the xy plane, and z is the same as in Cartesian coordinates. Hence

the equations that link the two systems are the same as for polar coordinates:
x = r cos , y = r sin , r = x
2
+ y
2
.
Hence for integrals the following equality holds:

f(x, y, z)dV = f(r cos , r sin , z)rdrddz,

R R

where R is the region corresponding to R in the rz space.

3 Spherical coordinates
In E
3
, we can also use the spherical coordinates system. Let x, y, z be the
Cartesian coordinates of a point P and let us denote its spherical coordinates
2



by , , and . Then is the distance from P to the origin O, is the angle
made by OP and Oz that is not greater than , and is dened exactly as in
the cylindrical coordinates system. The equations that link the spherical and
Cartesian systems are:
x = sin cos , y = sin sin , z = cos ,

=

x
2
+ y
2
+ z
2
, = arccos
z
.
x
2
+ y
2
+ z
2
To move integrals from the Cartesian system to the spherical system, we use
the following equality:
dV =
2
sin ddd.
Hence in terms of integrals, the following equality holds
f(x, y, z)dV =
R
= f( sin cos , sin sin , cos )
2
sin ddd,

where R is the region coresponding to R in the space.

3

Lecture XVII
Curvilinear Coordinates; Change of Variables
As we saw in lecture 16, in E
2
we can use the polar coordinates system.
In this system, we have a xed point O and a xed ray Ox. The coordinates
of a point P are given by r, the distance from P to O, and the angle made
by

Ox to OP. We can OP and Ox, as measured going counterclockwise from

change the system of coordiantes from polar to Cartesian through a system of
equations:
x = r cos , y = r sin .
This is called the dening system. To go from the Cartesian system to the polar
one, we use the inverse dening system:
r = x
2
+ y
2
, = arctan
y
,
x
for (x, y) in the rst quadrant. Generally, we can introduce new non-Cartesian
coordinates u, v by writing x, y as functions of these new coordiantes:
x = g(u, v), y = h(u, v).
These equations form the dening system for the new coordiantes. They can be
summarized by writting the position vector

j. R(u, v) = x

i + yj = g(u, v)

i + h(u, v)

In working with curvilinear coordinates, it is useful to introduce unit coordinate

u
P
, v
P
. To nd vectors u
P
, we take

R g h

P
=
P
j,
u
|
u
|
P

i +
u
|
and we form the unit vector

R
v
P
= u
P
=
R
P
/|

P
. Similarly, we dene
u
|
u
| |
R

R
v
|
P
/|
v
|
P
. For example, recall that for polar coordinates, the dening system |
is x = r cos , y = r sin . The position vector is:
R = r cos

i + r cos

j,
1

and the unit coordinate vectors are:

R
j = cos

i + sin

= r,
r

R
j

= sin

i + cos

= .

In the uv plane, consider the rectangle determined by the vectors vv uu and
with point P as their tail, where u and v are two small positive quantities.
Now consider the point P

correspoding to P and the region corresponding to

the rectangle in the xy plane. The area of this region is given by
P
v =

R R
u
P
|
P

v
| uv

R R
P
u
u
| |
v
This vector product can be easily computed, since

R g

R
=
g

i +
h
j and

= i + j.
u u u v v v
Hence

R R
u

v
= =
(x, y)
(u, v)
.
x x
u v
y y
u v
This expression is called the Jacobian of x and y with respect to u and v. For
polar coordinates, the Jacobian is
(x, y)
(u, v)
=
cos r sin
sin r cos
= r cos
2
+ r sin
2
= r.
When switching from the Cartesian system to a curvilinear one, u, v might be
given as functions of x, y, so the following equality is useful:
(u, v) (x, y)
= 1 .
(x, y) (u, v)
The notion of Jacobian can be extended to E
3
in the following manner. If x, y, z
are functions of u, v, w, the Jacobian is
(x, y, z)
(u, v, w)
=
x x x
u v w
y y y
u v w
z z z
u v w
2
.

For spherical coordinates , , , the Jacobian is:

(x, y, z)
= r
2
sin .
(, , )
Jacobians are particularly useful when we compute integrals, because we can
change variables in the following way:
f(x, y)dxdy = f(g(u, v), h(u, v))

R R
(x, y)
(u, v)
dudv.
3

1
Lecture XVIII
Change of Variables; Vector Fields
Change of Variables
Recall from lecture 17 that we change variables in integrals by the following
formula:

f(x, y)dxdy = f(x(u, v), y(u, v))

R R
x x
u v
y y
dudv,
u v
where R and

R are corresponding regions in the xy and uv planes.

Changing variables lets us easily compute integrals. For example, let us nd
the area in E
2
enclosed by the curves xy = 1, xy = 3, xy
2
= 1, and xy
2
= 2
in the rst quadrant. Denote the region enclosed by these curves by R. If we
choose new coordinates u = xy and v = xy
2
, the region

R corresponding to R in
the uv plane is the rectangle of vertices (1, 1), (3, 1), (3, 2), and (1, 2). Changing
variables to u, v, the area of R is given by
(x, y)
(u, v)
dudv =

2

3
dxdy =

2

3
(u, v)
(x, y)
1 dudv =
R 1 1 1 1

2

3

2

3
1 y x
= 1 dudv = dudv =
xy
2
y
2
2xy 1 1 1 1
=

2

3
1
dudv =

2
2
dv = 2 ln 2.
1 1
v
1
v
2 Vector Fields
Denition 1 A vector eld in E
2
is a function dened on a region R in E
2
that gives as outputs vectors in E
2
.
1

Clearly all vector operations apply to vector elds as well. We can write a vector
eld as following:
F(x, y) = f(x, y)

i + g(x, y)

j.
An important category of vector elds that we will be working with is the
category of gradient elds. We will see that not all vector elds are gradient
elds.
Denition 2 Let

F(x, y) be a vector eld in E

2
. If there exists a scalar eld f
such that
f
P
= F(P) for all P,
then F is called a gradient eld. Also, f is called a scalar potential for

F.
The following lemma, which helps us determine if a vector eld is a gradient
eld, is called the derivative test for a gradient eld.
Lemma 1 Let

F(x, y) = f(x, y)

i + g(x, y)

F be a vector eld,

j. F can be a
gradient eld only if f
y
= g
x
.
This test doesnt always tell us if F is a gradient led or not, it only states
that if the condition
f
=
g
isnt satised, then we can be sure that F isnt a
y x
gradient eld.
Proof of Lemma 1:
F

Let

= f

i + gj be a gradient eld. Then there exists f such that

F = f =

h
x

i + h
y
j.
Hence f = h
x
and g = h
y
, and since (h
x
)
y
= (h
y
)
x
, we get that f
y
= g
x
.
2

1
Lecture XIX
Visualizing Vector Fields; Line Integrals
Visualizing Vector Fields
Recall that a vector eld in E
2
is a function of the form
F(x, y) = f
1
(x, y)

i + f
2
(x, y)

j.
We dene two concepts that help us visualize vector elds.
Denition 1 Let C be a directed smooth curve in E
2
, and let

F be a vector
eld in E
2
. Then C is called an integral curve of F if, at any point P on C,
F(P) = 0 and

F(P) has the same direction

F(P) is tangent to the curve C, i.e.

as T

P
, the unit tangent vector to C at P .
Recall that C is the class of all continuous scalar functions and that C
1
is the
class of all dierentiable functions with continuous partial derivatives. We say
that the vector eld

F as dened above is in C
1
if f
1
, f
2
are in C
1
.
Lemma 1 For any point P such that F(P) = 0, there exists an integral curve
for F through P.
F

Let us take

= x

i +yj = rr. The integral curves of F are rays comming out of

1
the origin.

F passes the derivative test, and it is easy to see that f =

2
(x
2
+y
2
)
is a potential function for f, i.e.

r. f = x

i + yj = r
Denition 2 Let

F be a gradient eld with scalar potential f. The level curves

of f, i.e. the curves on which f is constant, are called equipotential curves for
F.
Lemma 2 Let

F be a gradient eld with scalar potential f. At any point P

where

F(P) = 0, the integral curve through P is normal to the equipotential

curve through P.
1

2 Line Integrals
Denition 3 Let C be a nite curve in E
2
and let f be a scalar eld dened
on C. We divide C into n pieces of arc-length s
1
, s
2
, . . . , s
n
and from each
n
piece we choose a point P
i

, forming the Riemann sum f(P

i

)s
i
. If the
i=1
limit
n
lim f(P
i

)s
i
n
max si0 i=1
exists, then it is called the scalar line integral of f on C. It is denoted by
C
fds.
ds R
Remember that s is the arc length of C and that
dt
= |
d

. We can evaluate
dt
|
line integrals by two methods.
1. Evaluation by denition
In this method, we simply use the denition of the integral. For example,
let C be the circle of radius 2 with center at the origin, and let f(x, y) =
x
2
+ y
2
. Then
fds = (x
2
+ y
2
)ds = 4ds = 4 4 = 16,
C C C
since the arc length of C is 4.
Let us extend the notion of line integral to vector elds. If

F(x, y) =
f
1
(x, y)

i + f
2
(x, y)

F on a curve C is j, the line integral of

Fds = f
1
ds

i + f
2
ds

j.
C C C
F

Let us take

= x

i + yj on the circle C of equation x

2
+ y
2
= 4. Since C
is symmetrical with respect to Ox and Oy,
Fds =

Fds, so Fds = 0.
C C C
2. Parametric evaluation
R(t) = x(t)

i +y(t)

Let us take for the nite curve C a path

j, with t going
from a to b. Then
=
d

R
dt
=
d

R dx

ds
j

i +
dy

i + j, x x
2
+ y
2
. = = y so
dt dt dt dt
2

Hence if f(x, y) is a function on C, we can evaluate the line integral of f:

d

R
dt
dt =

b
fds = f
C a
f(x(t), y(t)) x
2
(t) + y
2
(t)dt.
a
3

1
Lecture XX
Vector Line Integrals; Conservative Fields
Vector line integrals
Let

F be a vector eld of domain D. Let D be connected, i.e. for any two points
P, P

in D there is a curve C contained in D that goes from P to P

. Let C be
a nite directed curve contained in D that is rectiable, i.e. it has a length. We

F R as the limit of a Riemann sum: dene the vector line integral

C
n

d

F R = lim F(P
i

T(P
i

)s
i
,
C
n
max si0 i=1
where T

(P) is the unit tanget vector for C at P, and P

i

, s
i
are dened in the
same way as for scalar line integrals. Let us give two basic laws for vector line
integrals:
1. Let

F be a vector eld on a nite directed curve C, and let C be divided

into curves C
1
and C
2
, both having the same direction as C. Then:

d

d

d

F R = F R + F R.
C C1 C2
F and

2. Let

G be vector elds on a nite, directed curve C. Then:

d

G d

R. aF + b

R = a F R + b G d

C C C
There are two ways of evaluating vector line integrals. First, we can evaluate
using the denition of the integral as the limit of a Riemann sum. We can
also evaluate by parameter. Let R(t) be a path for C, with t from a to b,
d

R
j + z(t)

x(t)

i + y(t)

k. Since T

=
d

R
R(t) and
ds
dt
dR
, we get that = =
dt dt dt

b
d

d

F R = F T

ds = F dt
dt
C C a
1

2 Conservative Fields
Denition 1 Let D be a domain for a vector eld F. If for any points A, B D

and any curve C D from A to B, the integral F R depends only on A

C

and B and not on the choice of C, then

F is called a conservative eld.

An equivalent denition is the following:
Denition 2 Let D be a domain for a vector eld

F. If for any simple closed

curve C D, F R = 0, then

F is called a conservative eld.

C

Any conservative eld has a scalar potential.
2

1
C
2
Lecture XXI
Line Integrals; Conservative Fields
Line integrals
Let us recapitulate the basic notions reering to line integrals. For both scalar
and vector elds, we can dene line integrals. For a scalar eld f on a curve
fds.
C
For a vector eld

F, it is denoted C, the line integral is denoted by

F R. We can evaluate line integrals either by their denition as limits of

Riemann sums, or by parameters. Let us consider a path

R(t) for the curve C,

t going from a to b. Then we can evaluate the line integral using parameter t:
d

R
dt
dt,

b

b
d

d

F R = F fds = f dt.
dt
C C a a
For vector elds we can also dene other line integrals as well. If

F(x, y) =
R
Fds
M(x, y)

i + N(x, y)

j, then

i +

j. Also, is
C
R
Fds = Mds Nds
C C C ds
C
the average value of

F on C. We can also dene the line integral

R F d

=
C
R

dt.
C
F
dt
Conservative elds
F. We say that

Let D be the domain of a vector eld

F has independence of
F R = 0. If

d

F has independence of path, path on D if for any loop C in D,

C

F is called a conservative eld. Recall that

then

F is called a gradient eld if

f

there exists a scalar eld f such that

= F.
Theorem 1 (Conservative-eld Theorem) Let

F be a continuous vector
eld on a domain D. Then

F is conservative if and only if

F is a gradient
eld.
Proof:
1


Since

F is a gradient eld, it has a scalar potential f. Hence

b

d

R = R(t))dt = f(

R(a)). F R = f d

d
f(

R(b)) f(

dt
C C a
Let Q be a xed point in D. For any point P D, we dene f(P) =

F R, where C is a curve from Q to P. Clearly, since

F is conservative,
C

f is well dened. It follows that f is a scalar potential for

F is a F, so

gradient eld.
If we know that

F is conservative, there are two methods which we can use to

nd a scalar potential for

F.
1. The vector line integral method uses the conservative eld theorem. More

precisely, if we x a point Q D and let f(P) = F R, where C is a

C

curve from Q to P, then f is a scalar potential for

F.
2. The indenite integral method is often simpler than the line integral
method. Let

j F(x, y) = M(x, y)

i + N(x, y)

and f be the scalar potential

f
to be found. Then M =
f
and N = . Hence f = Mdx + C(y), so

x y
f
y
= N =
d
Mdx) + C

(y), and from this last equation we nd C(y),

dy
(
F j. hence nding f. For example, let us take

= (2x3y4)

i+(4y3x+2)

Then f = x
2
3xy 4x + C(y), so
f
= 3x + C

(y) = 4y 3x + 2.
y
Hence C

(y) = 4y + 2, so C(y) = 2y
2
+ 2y + c. It follows that f(x, y) =
x
2
3xy 4x + 2y
2
+ 2x + c is a scalar potential for

F.
2

Lecture XXII
Surfaces
Recall that an elementary region

D contained in E
2
is called convex if for
D, the line conecting them is contained in

any points in

D. Also recall that a

R on D is called injective if distinct points in

R. map

D give dierent values of

Using these notions we will give a denition for surfaces.

Denition 1 (Parametric expression for a surface) Let O be a xed point
in E
3
and let

D be a convex elementary region in E

2
, where E
2
is the Cartesian
R(u, v) be an injective continuous function on

plane of coordinates u, v. Let

D
with values in E
3
. The surface S is the set of all points that have

R(u, v) as
position vector for some (u, v)

D.
For example, in the uv plane we consider the rectangle 0 u , 0 v 2,

and the function

R(u, v) = sinu cos v

i +sin u sin vj +cos uk. Then the surface S

that has parametric representation given by

R is the sphere of radius 1 centered

D to be the disk of radius 1 centered at O, and

at O. Also, if we take

R(u, v) =

k dened on D, then the surface given by

ui +vj +

1 u
2
v
2
R is the sphere
of radius 1 centered at O.
Denition 2 Suppose

R(u, v) has continuous partial derivatives. Let P

be an

interior point of D. Then P) =

is called the parametric normal w(

u
|
P

v
|
P
vector given by P

. Taking P = R(P

), we can denote P) as w
P
and call it

w(

the parametric normal vector at P on S.

The parametric normal vector w
P
has the following properties:
1. If w
P
is normal to S at P. w
P
= 0, then
D to the surface area in 2. |w
P
| is an amplication factor from the area in

U is a region around a point P

in

S. More precisely, if

D, and U is the
corresponding region around P, then surface area (U) w
p
area (

| | U).

| 3. The surface area of S is given by the double integral

D
|w(u, v) dudv.
1

|
Lecture XXIII
Surface Integrals
Remember that the parametric expression for a curve is given by a vector func-

j + z(t)

tion R(t) = x(t)

i + y(t)

k. The parametric expression for a surface S

R(u, v) = x(u, v)

i + y(u, v)

is given by a vector function of two variables:

j +
z(u, v)

k. The following are two fundamental properties of the parametric normal

vector w(u, v):
1. If w
P
is normal to S at P. w
P
= 0, then
2. |w
P
| is an amplication factor from the area

U around P

to the surface
area U around P.
We can use the Jacobian to nd the paramteric nromal vector:
d

R d

R (y, z)

(x, z)

(x, y)
i + j +

k. w(u, v) = =
P

(u, v) (u, v) (u, v) du dv P
Consider the semicylinder S of radius a, with 0 z b and y 0. In cylindrical
coordinates, it is described by 0 , 0 z b, r = a. Clearly, since w
P
is normal to S,
w
P
= r. Since z and in the z plane correspond to z
|w
P
|
and a on the semicylinder, |w
P
= a. Hence w
P
= ar. We now give several
denitions concerning types of surfaces.
Denition 1 A surface is said to be elementary if its representation

R is a
one-to-one mapping.
Denition 2 A surface is nite if it can be divided into a nite number of
elementary surfaces, where any two of these surfaces intersect at most along
their common boundary, and no segment of boundary is contained in more than
two of these surfaces.
Denition 3 A nite surface S is one-sided if there exists a continuous loop
on S such that if we start from a point P and consider the parametric normal
1



vector on the loop, when arriving at P again the vector has direction opposite
of the one from the start. If such a loop does not exist, then we say the surface
is two-sided.
Denition 4 A nite surface that has no boundary is called closed.
Denition 5 Let D be a connected region in E
3
. D is simply-connected if for
every loop C contained in D, there exists a two-sided surface S such that C is
the boundary of S and S is contained in D.
Denition 6 Let D be a connected region in E
3
. D is two-connected if for
every closed surface S in D, the interior of S is also contained in D.
We will now dene surface integrals. There are scalar and vector surface inte-
grals.
Denition 7 Let f be a scalar eld dened on a nite surface S. The scalar
surface integral of f on S is denoted by
S
fd and is the limit of a Riemann
sum:

n

fd = lim f(P

i
)
i
,
n
S
max di0 i=1
where S
1
, . . . , S
n
form a subdivision of S intro elementary surfaces, P

is a
i
point in S
i
,
i
is the surface area of S
i
, and d
i
is the diameter of S
i
for
all i n.
Denition 8 A directed surface S is a two-sided surface with all its nonzero
normal vectors pointing away from the same side. The normal vectors are said
to give a direction to S.
Denition 9 Let

F be a vector eld dened on a nite directed surface S. The

F on S is denoted by
S
(

vector surface integral of

F n)d and is the limit of

a Riemann sum:

n
(

F n)d = lim [F(P

i

) n(P
i

)]
i
,
S
n
max di0 i=1
where S
1
, . . . , S
n
form a subdivision of S intro elementary surfaces, P

is
a point in S
i
,
i
is the surface area of S
i
, d
i
is the diameter of S
i
, and
n(P
i

) is the normal vector on S at P

for all i n.
i
2
i


|

We also denote the vector surface integral by
S
F d

. To evaluate surface
integrals, the following formula are useful:
fd = f(x(u, v), y(u, v), z(u, v))|w(u, v) dudv

S D
F d

= F w(u, v)dudv

S D
3
Lecture XXIV
Measures
Remember that an elementary region R in E
2
is a region that has a simple,
closed, piecewise smooth curve C as its boundary. A regular region R is a region
that is either regular or can be divided into nitely many regular regions. In the
second case, the boundary of R consists of two or more simple, closed, piecewise
smooth curves, of which one is the outer boundary curve, denoted by C, and the
others are interior boundary curves C
1
, . . . , C
n
. Then R consists of the points
in the interior of C, excluding those in the interior of C
1
, . . . , C
n
.
Similarly, in E
3
a region is regular if its boundary is a simple, closed, piece-
wise smooth surface S. An elementary region in E
3
is either regular or can
be divided into nitely many regular regions. The boundary of an elementary
region that is not regular is formed by one or more closed, but not simple piece-
wise smooth surfaces. In the case that the boundary is formed by two or more
surfaces, one of them is the outer boundary surface S, and the other are interior
boundary surfaces S
1
, . . . , S
n
. Then R consists of the points in the interior of
S, excluding those in the interior of S
1
, . . . , S
n
.
Denition 1 Let D be a region in E
2
or E
3
. Let be a real-valued function
that has as inputs the regular subregions of D. The function is called a nite
measure on D if for every regular subregion R and for every division of R into
regular subregions R
1
and R
2
we have
(R
1
) + (R
2
) = (R).
In E
2
, an example of a nite measure is the function that assigns to every
regular region its area. Similarly, in E
3
, the function that assignes to every
regular region its volume is a nite measure. More generally, in E
2
, if f is a
continuous scalar eld on a region D, for every regular subregion R of D we
dene

f
(R) = fdA.
R
1

Then
f
is a measure on D. Similarly, in E
3
,
f
(R) =
R
fdV is a measure
for every continuous scalar eld f.
Denition 2 Let

F be a continuous vector eld on a region D in E

2
. We
dene the nite measure

F
in the following manner:
1. if R is elementary,

=
C
F d

R, where C is the counterclockwise
F

directed boundary of R.
2. if R is not elementary,

R+
C1
F d

+
Cn
F d

R, where
F C
F d

R+
C is the counterclockwise directed outer boundary of R and C
1
, . . . , C
n
are
the clockwise directed inner boundaries of R.
It is easily veriable that

is a nite measure. It is called the circulation

F
measure on D given by

F.
Denition 3 Let

F be a continuous vector eld on D in E

3
. We dene the
nite measure
f
in the following way:
F
1. if R has a single boundary surface S, then
f
(R) =

S
F d

, where S is
F

directed outward, away from R.

2. if R has boundary surfaces S, S
1
, , S
n
, then
f
(R) =

S
F d +

F
F d + +

Sn
F d, where S is directed outward, and S
1
, . . . , S
n
S1

are directed inward.
It is easy to prove that
f
is a nite measure. It is called the ux measure on
F
D given by

F.
Denition 4 Let D be a region in E
3
. Let
s
be a real-valued function that
has as inputs the nite, piecewise smooth, directed surfaces in D. The function

s
is called a nite surface measure on D if for every nite, piecewise smooth,
directed surface S and for every division of S into nite, piecewise smooth,
directed surfaces S
1
and S
2
we have

s
(S
1
) +
s
(S
2
) =
s
(S).
Denition 5 Let

F be a continuous vector eld on D in E

3
. We dene the
c
nite surface measure in the following way:
F
2

1. if S is closed, then
c

F
= 0.
2. if S is not closed, let C
1
, . . . , C
n
be the piecewise smooth, simple, closed
boundary curves of S, with directions coherent with the direction of S.

F R +

F R. Then
c

F
=
C1
+
Cn

It is easily veriable that is a nite surface measure. It is called the circu-

c

F
lation measure on D given by

F.
3
















Lecture XXV
Greens Theorem
Let us dene a new type of derivative, called rotational derivative, applicable to
vector elds in E
2
. For such a vector eld

F on a domain D in E
2
, let us dene
the rotational derivative at interior points of D. Here, a point P in D is called
an interior point if there exists a circle of center P and radius a > 0 that has
its interior contained in D.
Denition 1 Let

F be a C
1
vector eld on D in E
2
. Let P be an interior point
of D and let C(P, a) be the circle of center P and radius a for all a > 0. The
rotational derivative, denoted by rot

P
, is given by the following formula: F|
rot

F
1

F R. = lim
P
a 0 a
2
C(P,a)
F| For boundary points Q of D, if lim
PQ,P
rot

P
= q exists, we say that rot

Q
F|
F is C
1
, then rot

|
P
exists at every point P on the exists and its value is q. If

F
domain D of

F. The rotational derivative satises the linearity law:

G) = a(rot

G). rot(aF + b

F) + b(rot

Theorem 1 (The parallel ow theorem in E

2
) Let

F be a vector eld on
D in E
2
for which there exists a scalar eld f and a unit vector w such that
F = f u be the unit vector

w. Such a vector eld is called a parallel ow. Let

w normal to w in E
2
, and let v = u. Then for any interior point of D, we
have that:
. rot

F
df
=
P
ds v,P
F j be a vector eld in E
2
with Cartesian coordinates. Using the Let

= L

i +M

linearity law and applying the parallel ow theorem for L

i and M

j, we get the
formula for rot

F at any interior point P of D:

L
rot

F
M
P
= M
x
(P) L
y
(P) =
P
.
x y P
1


F For example, if F = yi+xj, then rot

= M
x
L
y
= 1(1) = 2. The follow-
ing theorem links integrals on regions in E
2
with integrals on their boundaries,
with the help of rotational derivatives.
Theorem 2 (Greens theorem) Let

F be a vector eld on D in E
2
. Let R
be a regular region in D, with a counterclockwise directed outer boundary C and
clockwise directed interior boundaries C
1
, . . . , C
n
. The following equality holds:

d

d

d

rot

FdA = F R + F R + + F R.
R C C1 Cn
By Greens theorem, if R is a regular region in D, with C its counterclockwise
directed boundary, then

rot

FdA = F R.
R
F

Using this equality, if we let

= yi +xj, and let R be a disk of radius 1 with

boundary C, we get:

FdA = 2dA = 2 area(R) = 2 F R = rot

C R R
2

Lecture XXVI
The Divergence Theorem
In this lecture, we will dene a new type of derivative for vector elds on E
3
,
called divergence. Let

F be a vector eld dened on a domain D. Let us start

by dening the divergence of

F on interior points of D, i.e. points P such that

there exists a sphere of center P and radius a > 0 with its interior contained in
D.
Denition 1 Let

F be a continuous vector eld on D in E

3
. Let P be an
interior point of D, and let S(P, a) be the sphere of center P and radius a, for
all a > 0. The volume of S(P, a) is
4a
3
and the ux of

F through S(P, a) is
3
F d. Consider the limit
S(P,a)

lim F d.
a 0 4a
3

S(P,a)
If this limit exists, then it is called the divergence of

F at P, and it is denoted
by div

|
P
. F
Below are two important properties of divergence.
F is C
1
on D, then the divergence of

1. Existence: If

F exists at every
interior point of D.
2. Linearity: If

G are vector elds dened on D, for any two scalar F and

constants a and b the following equality holds:

G) = a(div

G). div(aF + b

F) + b(div

The following theorem helps us nd a formula for divergence in Cartesian coor-

dinates.
1

Theorem 1 (The parallel ow theorem in E

3
) Let

F be a vector eld on
D in E
3
such that there exists a scalar eld on D and a constant unit vector w
F w on D. Such a vector eld

for which

= f F is called a parallel ow. Suppose

f is C
1
. Then for any interior point P of D, the following equality holds:
div

F .
df
=
P
ds w,P
F j +N

Let

= L

i +M

k be a C
1
vector eld on a domain D in E
3
. By using the
linearity of divergence and applying the parallel ow theorem to L

i, M

j, and
N

k, we get the following formula:

div

F = L
x
(P) + M
y
(P) + N
z
(P) =
L
+
M
+
N
x P y P z P
P
Through the following theorem, we can use divergence to compute surface inte-
grals more easily.
Theorem 2 (The divergence theorem) Let

F be a C
1
vector eld on D in
E
3
. Let R be a regular region in D, with outward directed outer boundary surface
S and inward directed inner boundary surfaces S
1
, . . . , S
n
. Then
FdV = F d + F d div

+ + F d.
R
S S1 Sn
Let R be a regular region in D, and let S be its boundary. Then, by the
divergence theorem, we have that
div

FdV = F d.
R
S
2

Lecture XXVII
Stokess Theorem
In the previous lecture, we saw how Greens theorem deals with integrals on
elementary regions in E
2
and their boundaries. We also saw how the divergence
thoerem deals with elementary regions in E
3
and their boundaries. We will now
look at Stokess theorem, which applies to integrals on elementary surfaces in
E
3
and their boundaries.
First let us extend the concept of rotational derivatives to E
3
.
Denition 1 Let

F be a vector eld with a domain D in E

3
, and let P be an
interior point of D. Given a unit vector u in E
3
let M be the plane that goes
through P and is normal to u. Consider this plane directed, with the direction
given by u. For a > 0, let C
a
be the circle in M of radius a and center P.
Consider the limit

lim F R
a 0 a
2
Ca

If the limit exists, then it is called the rotational derivative of F in the direction
F

u at P and it is denoted by rot

|
u,P
.
F and F

If

u are given in Cartesian coordinates, then the value of rot

|
u,P
can

j + be easily found out from the given formulas. Let F = L(x, y, z)

i +M(x, y, z)

k and let

N(x, y, z)

u = u
1

i + u
2
j + u
3
k. Then
F|

rot

u,P
= (N
j
M
z
)u
1
+ (L
z
N
x
)u
2
+ (M
x
L
y
)u
3
.
Denition 2 Let

j k and let P be a
F. The vector

F at P and denoted

F
F = L(x, y, z)

i + M(x, y, z)

+ N(x, y, z)

point in the domain of

C
P
, called the curl of

by curl

|
P
, is dened by the formula

F| j + (M
x
L
y
)

curl

P
= (N
y
M
z
)

i + (L
z
N
x
)

k
1

The curl of

F can also be represented by a determinant:

F curl

=

i j k

x y z
L M N
The following equality results immediately from the formulas given above.
Theorem 1 (Curl theorem)
F

F| rot

|
u,P
= curl

P
u
From the equality, since u is a unit vector, it follows that the magnitude of
curl

|
P
is the maximum value of rot

u,P
as

F F|

u varies at P. The maximum is

F obtained when the direction of u coincides with the direction of curl

|
P
.
Theorem 2 (Stokess theorem) Let S be a two-sided directed surface whose
boundary consists of the simple, closed curves C
1
, , C
k
. Let

F be a vector
eld whose domain contains S. If we take the direction of each curve C
i
to be
coherent with the direction of S, then

d

d

F curl

d = F R + + F R
S C1 C
k
2

1
Lecture XXVIII
Measures; Irrotational elds
Circulation and ux measures
Let us rst see what the theorems from the past lectures say about the circula-
tion and ux measures. From Greens theorem, we get that, if

F is a C
1
vector
eld on D in E
2
, then
R
F
(R) = rot

FdA
for every regular region R of D, where

F
is the circulation measure given by
F. In E
3
, from the divergence theorem we have that
f

F
(R) = div

FdV,
R
f

F
is the ux measure for every regular region R of D, where D is in E
3
and
given by

F. From Stokess theorem, we have

F , (S) = curl

d
c

F

S
is the circulation measure given by

F. Remember that in E
2
, if where
c

F
F = L

i + M

j, then
L
rot

=
M
F
x

y
F j + N

In E
3
, if

= L

i + M

k, then
M N
div

=
L
+ + F
x y z

F curl

=

i j k

z

y

x
L M N
1

2 Irrotational elds
Denition 1 A vector eld F on D in E
2
such that rot

F = 0 everywhere is
called an irrotational eld on D.
If

F is an irrotational eld in E
2
and C
1
and C
2
are two simple, closed, piecewise
smooth curves having the same direction, such that C
2
is contained in C
1
, then
by Greens theorem we have

d

d

F R = F R.
C1 C2
Denition 2 A vector eld F on D in E
3
such that div

F = 0 everywhere is
called an divergenceless eld on D.
Let F be a divergenceless eld in E
3
. Let S
1
and S
2
be two directed nite
surfaces with identical boundaries, such that their directions are coherent with
the directions of their boundaries. Then by the divergence theorem,
F d = F d.
S1 S2
Denition 3 A vector eld

F F on D in E
3
such that curl

= 0 everywhere is
called an irrotational eld on D.
Let

F be an irrotational eld in E
3
. Let S be a piecewise C
2
smooth, two-sided
surface S with C and C

its boundary curves, both directed outward. Then by

Stokess theorem, we have

d

d

F R = F R.
C C

In E
3
, let us introduce the dierential operator del, denoted by

and dened
by

i +
d
j + k.
dx dy dz
F j + N

If

= L

i + M

k, then
dL dM dN

F = + + = div

dx dy dz
By Stokess theorem, if

F is a C
2
vector eld in E
3
, then

F curl

= 0.
2

Lecture XXIX
Mathematical Applications
1 Leibnitzs Rule
Leibnitzs Rule : Let f(x, t) be a C
1
function dened for a x b. Then

b
d f
f(x, t)dx = f
t
(x, t)dx, where f
t
(x, t) = .
dt t
a a

b
In other words, if we dene g(t) =
A
f(x, t)dx, then
dg
=

b
f
t
(x, t)dx.
dt a
Leibnitzs rule also works for functions that have more parameters. For
example, we have

b
d
f(x, y, z, t)dx = f
t
(x, y, z, t)dx.
dt
a a
The rule can be generalized and applied to gradients. Indeed, let f(x, y, z, t) be
a scalar eld dened for all a t b and (x, y, z) D. Then if P is a point in
D, we have that

P

fdt =
P
fdt.
a A
Let f(x, y, z, t) be a time-dependent mass-density function on a region D. Then,
by Leibnitzs rule, we have
d f
fdV = dV.
dt
D
t
D
This means that the rate of change of the mass( the left-hand side of the equality)
equals the integral of the rate of change of density( the right-hand side of the
equality).
2 Formulas in Spherical Coordinates
The following are the formulas for gradient, divergence, and curl in spherical
coordinates.
1

3
Let g(, , ) be a scalar eld and let G(, , ) = g
1
(, , ) +

+g
2
(, , )
g
3
(, , )

be a vector eld, both in spherical coordinates. We have

g =
g
+
1 g 1 g

sin
1 (
2
g
1
) 1 (g
2
sin )
G =

G = + +
1 g
3
div

2
sin sin
=
1

2
sin sin




g
1
g
2
g
3
sin
curl

G =

G =
1 (g
3
sin ) 1 g
2
+ =
sin sin
1 g
1
1 (g
3
) 1 (g
2
) 1 g
1

+ +
sin

The Laplacian
The Laplacian is a dierential operator on scalar and vector elds. It is denoted
by

2
. For a scalar eld f, the Laplacian is dened in the following manner:

2
f
2
f
2
f

2
f =

f = +
y
2
+ .

x
2
z
2
F j + N

For a vector eld

= L

i + M

k, the operator is dened by

F = (

2
L)

i + (

2
M)

k.

j + (

2
N)

Denition 1 A scalar eld f is said to be harmonic if it satises the equation

2
f = 0, called Laplaces equation.
While for scalar elds we have

2
f =

f, for vector elds we can compute

F using the formula

F = (

F)

2

F).

Theorem 1 Let

G be dierentiable vector elds on a simply connected do- F,

main D. Then

F =

G and

F =

G if and only if there exists a

harmonic scalar eld h such that F

G = h.
2

Theorem 1 can be used to prove the following result.

Theorem 2 (Helmholtzs Theorem) Let F be dierentiable on a simply con-
nected domain D. There exist

H such that: G and

(i) G is irrotational and

H is divergenceless;
(ii)

G +

H =

F; H) =

(iii)

G +

G =

F and H) =

G +

(iv) F (

H) = h for some harmonic scalar eld h.

3

1
Lecture XXX
n-Vectors and Matrices
n-Vectors
We dene E
n
to be the n-dimensional Euclidean space, and R
n
to be the set
of points in E
n
. Hence R
n
is the set of all ordered n-tuples of real numbers
(x
1
, . . . , x
n
). Ordered n-tuples allow repetitions, i.e. x
i
= x
j
is possible for i =
j, but if x
i
= x
j
, then (x
1
, . . . , x
i
, . . . , x
j
, . . . , x
n
) = (x
1
, . . . , x
j
, . . . , x
i
, . . . , x
n
).
These n-tuples can be viewed as vectors in the n-dimensional Euclidean space.
Hence we will call them n-vectors.
Denition 1 Let P = (p
1
, p
2
, . . . , p
n
) and Q(q
1
, q
2
, . . . , q
n
). We say that the
distance between P and Q is d(P, Q) = (p
1
q
1
)
2
+ (p
2
q
2
)
2
+ + (p
n
q
n
)
2
.
We dene

0 = (0, 0, . . . , 0). On n-vectors, we can dene operations analogous

to the ones on 3-vectors:
1. Multiplication of a vector by a scalar
A = (a
1
, a
2
, . . . , a
n
) and let k be a scalar. Then k

Let

A = (ka
1
, ka
2
, . . . , ka
n
).
2. Vector addition

A +

Let A = (a
1
, a
2
, . . . , a
n
) and B = (b
1
, b
2
, . . . , b
n
). Then

B = (a
1
+
b
1
, a
2
+ b
2
, . . . , a
n
+ b
n
).
3. Dot product

A

Let A = (a
1
, a
2
, . . . , a
n
) and B = (b
1
, b
2
, . . . , b
n
). Then

B = a
1
b
1
+
a
2
b
2
+ + a
n
b
n
.
4. Magnitude of a vector
Let A = (a
1
, a
2
, . . . , a
n
). We say the magnitude of

|

A is |A = A = A
2 2
a + a
2
+ + a
2
.
1 n

1

| |

|



| |

|
5. Unit vectors
A
1

Let

A =

0. Then

= A has magnitude 1, hence it is called unit vector.

|A|
Denition 2 We say that two nonzero vectors

A

A, B are orthogonal if

B = 0.
For example, in E
4
, e
1
= (1, 0, 0, 0), e
2
= (0, 1, 0, 0), e
3
= (0, 0, 1, 0) and e
4
=
(0, 0, 0, 1) play the role of

j, and

i,

k from E
3
. They are mutually orthogonal
and any 4-vector as a linear combination of these vectors.
Denition 3 Let

A
1
, A
2
, . . . , A
n
be mutually orthogonal n-vectors. We say that
they form a frame.
Theorem 1 (Cauchy-Schwartz inequality) For any given n-vectors A and

B, |A B 1. |
Theorem 2 (Triangle inequality) For any given n-vectors A and B, |A +
B| |A + B .
Proof of Theorem 1:

A A2

B B = 22

Since (

B) (

B) 0, we get that

A B+

AB 0,
hence

B 1. In the same manner, from (

B) (

B) 0, we get that A

A +

A B 1. Hence A B 1. |

|
Proof of Theorem 2:
From Theorem 1, we have that

A B
A B 1, hence
|A| |B|
1, so A A B|.

B |

Hence

A B +

|
2
= ( A + B

B|
2
= A A + 2

B B |A
2
+ 2 B + |B |A +

| |A||

| |

| |

|)
2
So |A +

B| |A + B .

| A and

Since |A B 1 and

B are orthogonal when A B = A B = 0, we

can dene the angle between A and

A

B to be cos
1
(

B).
2
2 Matrices
Denition 4 An mn matrix is a rectangular array of m rows and n columns
of numbers.
One-column matrices are called column-vectors and one-row matrices are called
row-vectors. Generally, we write an m n matrix A in the following way:

A =

a
11
a
12
a
1n
a
21
a
22
a
2n
. . .
. . .
. . .
a
m1
a
m2
a
mn

We can also write A = (a

ij
)
1im,1jn
. We dene the following operations on
matrices:
1. Multiplication with a scalar
Let A = (a
ij
) be an m n matrix. Then kA = (ka
ij
).
2. Matrix addition
Let A = (a
ij
) and B = (b
ij
) be m n matrices. Then A+B = (a
ij
+b
ij
).
Denition 5 Let A be an mn matrix. The transpose of A is an nm matrix
denoted by A
T
such that its i-th column reading from top to bottom is the i-th
row of A reading from left to right, for all i m.
For example,


1 3
if A =

, then A
T
=

1 3 4
3 1 0

3 1 .
4 0
The m n matrix all whose entries are 0 is denoted by 0
m,n
. An n n matrix
is called a square matrix. An n n matrix that has all elements on the main
diagonal (from upper left to lower right) equal to 1 and all other elements equal
to 0 is called an identity matrix, and is denoted by I
n
. Let us dene a new
operation on matrices, multiplication of a matrix with a column-vector :
3
Let A = (a
i,j
) be an m n matrix, and let B = (b
i
) be an n 1 matrix.
Then we dene the m 1 matrix AB as follows:
a
11
b
1
+ a
12
b
2
+ + a
1n
b
n

a
21
b
1
+ a
22
b
2
+ + a
2n
b
n

AB =

a
m1
b
1
+ a
m2
b
2
+ + a
mn
b
n

4

Lecture XXXI
Linear Equation Systems

As we saw in the previous lecture, we can multiply mn matrices by column
n-vectors. Consider the rows of an m n matrix A to be n-vectors:

r
1
c
1
r
1
C

r
2
.
.
.

, C =

c
2
.
.
.

then AC =

r
2

C
.
.
A =
.
C r
m
c
n
r
m

For example, if


7
1 2 1

27
22

, C = , then AC =

9 A =
4 0 3
2
We can use multiplication by a column vector to solve equation systems. An
m n equation system has the form
a
11
x
1
+ a
12
x
2
+ + a
1n
x
n
= d
1

a
21
x
1
+ a
22
x
2
+ + a
2n
x
n
= d
2


a
m1
x
1
+ a
m2
x
2
+ + a
mn
x
n
= d
m

This system has m equations and n unknowns: x
1
, x
2
, , x
n
. An example of a
2 3 system is
4x
1
+ 3x
2
x
3
= 1
x
1
x
2
+ x
3
= 6
This system has innitely many solutions.
An example of a 3 2 system is
x
1
+ x
2
= 4
1
X
2x
1
x
2
= 6
x
1
x
2
= 1
This system has no solutions.
An example of a 2 2 system is
2x
1
x
2
= 4
x
1
+ x
2
= 0
4 4
This system has an unique solution, namely x
1
=
3
, x
2
= .
3
A system that has all the coecients on the right-hand side equal to zero, i.e.
d
1
= d
2
= = d
m
= 0, is called a homogeneous system.
An m n system with the general form given above can by written as a matrix
equality. Let D be an m 1 column-vector whose entries are d
1
, d
2
, , d
m
from top to bottom. Let A = (a
ij
) be the m n matrix those entries are the
coecients on the left-hand side of the system. Let X be an n1 column-vector
whose entries are x
1
, x
2
, , x
n
from top to bottom. Then the left-hand side
of the system is given by the elements of AX, so the system can be written as
AX = D.

x
1
d
1
a
11
a
12
a
1n
=

x
2
.
.
.

, D =

d
2
.
.
.

, A =

a
21
a
22
a
2n
. . .
. . .
. . .

hence ,
x
n
d
m
a
m1
a
m2
a
mn

a
11
x
1
+ a
12
x
2
+ + a
1n
x
n

a
21
x
2
+ a
22
x
2
+ + a
2n
x
n

d
1
AX =

d
2
.
.
.
= D

a
m1
x
1
+ a
m2
x
2
+ + a
mn
x
n d
m
We dene three elementary operations on the equations of a system that do not
modify its solution-set:
() multiplying an equation by a non-zero scalar;
() adding to an equation some multiple of a dierent equation;
2
() interchanging two equations.
By using these operations through the method of row-reduction, we can simplify
the equation system without altering its solution-set. In the next lecture will
dene row-reduction, which is also known as Gauss-Jordan reduction.
3
1
Lecture XXXII
Row Reduction; Determinants
Row Reduction
Recall the 3 elementary operations we will use to solve the system of equations
AX = D:
() multiplying an equation by a non-zero scalar;
() adding to an equation some multiple of a dierent equation;
() interchanging two equations.
We call the matrix [A : D] the augmented matrix of the system AX = D. We
will use , , and to bring B = [A : D], or, in the case of homogeneous
systems, B = A to a row-reduced form.
Denition 1 The row-reduced form of a matrix B is a matrix obtained from
B by applying the elementary operations that has the following properties:
1. Every non-zero row has 1 as its rst non-zero element from left to right.
This element is called the pivot of the row.
2. Each pivot occurs in a column to the right of all pivots from upper rows.
3. All zero rows are at the bottom of the matrix.
4. For any pivot, all other elements in its column are equal to zero.
We obtain this form through the method of row-reduction, which is described
below:
1. Choose the rst non-zero column from left to right, then choose a row
having a non-zero element on this column.
1

2. Move this row on top of the matrix.

3. Multiply this row by
1
, where a is its rst non-zero element from left to
a
right.
4. For any row having b = 0 on the column on which the chosen row has its
rst non-zero element, add b times the chosen row to it.
5. Repeat the above steps, moving the selected row to the second position,
then the third position, and so on.
If, in the case of an nonhomogeneous system, the last column has a pivot, then
the system doesnt have solutions. If the last column doesnt contain any pivot
and there are other columns as well that dont contain pivots, then the system
has innitely many solutions. Suppose a nonhomogeneous system has the row-
reduced augmented matrix

1 0 7 0
0 1 3 2
0 0 0 0

Then we can write the equations of the system:
x
1
7x
3
= 0, hence x
1
= 7x
3
x
2
+ 3x
3
= 2, 2 3x
3
hence x
2
=

x
1
0 7

+ t

x
2
x
3
0 1

B, where Hence X = A + t

0 7
A =

and B =

0 1
In this case the solution set forms a line in 3-dimensional space. Generally, for
systems with 3 unknowns, the solution set is given by X = C + t
1
B
1
+ t
2
B
2
,
where C,

B
1
, B
2
are 3-vectors, possibly equal to

0. For homogeneous systems,

C =

0.
2

2 Determinants
Consider a square n n matrix A. The determinant of the matrix is a scalar
depending on A, denoted by A . For a 1 1 matrix, the determinant of the | |
matrix is simply the value of its element. Using the notion of minors, we can
dene a reccurent formula for determinants.
Denition 2 The minor of an element a
ij
of a square matrix is the determi-
nant of the matrix formed from the initial matrix excluding the row i and the
column j containing this element.
We denote by m
ij
the minor of a
ij
. Consider the nn matrix having as elements
the signs + and , in a checkerboard pattern:

+ +
+
+ +

We call the cofactor of an element a
ij
, the minor of a
ij
multiplied by +1 or 1
, depending on the position of the element in the above matrix. The cofactor
of a
ij
has the value (1)
i+j
m
ij
. For any column or row of A, the determinant
of A is given by the formula
A| = (1)
i+j
a
ij
m
ij
, |
where the sum is on the column or row, i.e. j is constant and i goes from 1 to
m, or i is constant and j goes from 1 to n.
3
1
Lecture XXXIII
Determinants; Matrix Algebra
Determinants
For a square matrix A, the determinant of A has the following properties:
1. Interchanging two rows of the matrix multiplies the value of the determi-
nant by -1.
2. If there exists two identical rows, then the value of the determinant is 0.
3. If we multiply a row by a scalar c, then the value of the determinant is
also multiplied by c.
4. Applying the operation to the matrix leaves the value of the determinant
unchanged.
5. If the determinant is 0, then there exists a row that can be written as a
linear combination of the other rows.
6. Properties (1) through (5) also hold for columns instead of rows.
In addition to the Laplace expansion method of nding determinants, there is
a second method that is faster in most cases. This is a simplied row-reduction
that brings the matrix in to a form called row-echelon matrix. In this method
we use operations and . The row-echelon matrix is dierent from the row-
reduced matrix in that:
1. pivots can have other values than 1, since we do not use operations;
2. in the column of a pivot, we only need to get 0s below the pivot.
1
|
2
If each column has a pivot, then the determinant of the inital matrix A is equal
to the product of the pivots times (1)
k
, where k is the number of operations
we applied to obtain the row-echelon form. If a column has no pivots, then there
will be a row containing only 0s, so the determinant will be 0. Let us dene
the rank of A as the number of pivots in the row-echelon form. The following
sentences hold:
1. If |A = 0, then rank(A) < n. In this case we say that A is a singular |
matrix.
2. If |A = 0, then rank(A) = n. In this case we say that A is a non-singular
matrix.
Matrix Algebra
We saw in the previous lecture that we can multiply matrices with column
vectors. Let us now dene in a larger sense matrix multiplication. In order to
dene the product AB of matrices A and B, A must have dimensions m n
and B must have dimensions n p. In other words, he number of columns
in A must equal the number of rows in B. Then multiplying A with the i-th
column of B, we obtain a column m-vector. This column-vector will be the i-th
column of AB. Hence AB is an m p matrix. It can be easily observed that
BA exists as well only if m = p. In this case AB is an m m matrix and BA
is an n n matrix. So even if AB and BA both exist, they may not have the
same size. Furthermore, even if AB and BA have the same size, it is possible
that AB = BA. Hence the multiplication of matrices is non-commutative. Let
A be an m n matrix, let B be an n p matrix, and let C be a p l matrix.
Then (AB)C and A(BC) both exist and furthermore
(AB)C = A(BC)
Hence, the multiplication of matrices is associative. The following properties
also hold:
1. AI
n
= A and I
n
A = A, if the products exists, where I
n
is the identity
n n matrix we dened in the previous lecture.
2

2. A0 = 0 and 0A = 0, if the products exist, where 0 a matrix with all its

entries equal to 0.
3. A(B + C) = AB + AC
4. (A + B)C = AC + BC
For square matrices A, B, if AB = I
n
, we say that A is inverse to B. The
following properties hold:
1. If AB = I, then BA = I, i.e. if A is inverse to B, then B is inverse to A.
2. If BA = I and CA = I, then B = C.
Hence the inverse of a square matrix A is unique, if it exists. We denote it by
A
1
.
Theorem 1 The inverse of matrix A exists if and only if A is non-singular.
We can use the inverse to solve systems of equations. Consider the system
AX = D, where A is a non-singular square matrix. Then X = A
1
D. One
method to nd the inverse of a matrix A is the following: consider the matrix of
1
cofactors of A, transpose it and multiply it by . The result is the inverse of A.
|A|
A second method envolves the n 2n matrix [A : I
n
]. If we apply row-reduction
to this matrix, we get a matrix [B : C], with B and C n n matrices. If B = I,
then A is singular, so A
1
does not exist. If B = I, then A is non-singular and
C = A
1
.
3
Lecture XXXIV
Subspaces
In the previous lecture we have seen that there are two methods for nding
the inverse of a square matrix A. In the rst method, we use the fact that if A
is non-singular, then the row-reduced form of [A : I] is [I : A
1
]. In the second
method, we take the matrix of cofactors A

, consider its transpose (A

)
t
, and
1 1
multiply it by
detA
to obtain the inverse of A:
det A
(A

)
t
. The folowing theorem
is an important determinant property.
Theorem 1 Let A and B be two square matrices of the same size. Then
AB|= . | |A||B|
Denition 1 A set V is a vector subspace of R
n
if it satises the following
properties:
(i) 0 is in V .
x, x + (ii) For any y in V , y is in V .
x in V and any scalar c in R, c (iii) For any x is in V .
Clearly, the set containing only the origin is a vector subspace. In R
3
, a vector
subspace must be either the set previously mentioned, a line through the origin,
a plane through the origin, or all of R
3
.
Denition 2 Let X be a vector in R
n
and let V be a vector subspace of R
n
.
Let us denote by A = X +V the set of all the vectors X +Y , with Y in V . We
say that A is an ane subspace of R
n
.
Let A be a given ane subspace that is not formed by a single point only. Let
X be a vector and let V be a vector space such that A = X + V . Then the
choice of V is unique, where X can be any vector in A.
1
Denition 3 Let c
1
, , c
n
be scalars and let Y
1
, , Y
n
be contained in a vector
subspace V . The vector c
1
Y
1
+ c
2
Y
2
+ +c
n
Y
n
is called a linear combination
of vectors Y
1
, , Y
n
.
Theorem 2 Any linear combination of vectors from a vector subspace V is
contained in V .
Denition 4 Let
n
be contained in a vector subspace V . We call the Y
1
, , Y
span of Y
1
, , Y
n
and denote by span(Y
1
, , Y
n
) the set of all linear combina-
tions of Y
1
, , Y
n
.
For any vectors Y
1
, , Y
n
in V , span(Y
1
, , Y
n
) is a vector subspace contained
in V .
Denition 5 Vectors Y
1
, , Y
k
are linearly independent if a linear combina-
tion c
1
Y
1
+ + c
k
Y
k
is equal to 0 if and only if c
1
= = c
k
= 0.
Denition 6 If Y
1
, , Y
k
V are linearly independent and the span of Y
1
, , Y
k
is V , then the set Y
1
, , Y
k
is called a basis for V . Any point in V can be ex-
pressed as a linear combination of Y
1
, , Y
k
in a unique way.
2