1 Vector Calculus

Lecture 1
1.1 Differentiation of vectors
Definition: A vector-valued function of a real variable is a rule that associates a vector f (t) with a real
number t, where t is in some subset D of R1 (called the domain of f ).

E.g f (t) = ti + t2 j + t3 k is a vector-valued function in R3 , defined for all real numbers t.

The derivative f(t) at a, denoted by f 0 (a) or dfdt (a), is the limit:

f (a + h) f (a)
f 0 (a) = lim
h0 h
if that limit exists. Equivalently f 0 (a) = (f10 (a), f20 (a), f30 (a)) if the component derivatives exists. We say
that f 0 (t) is differentiable at a if f 0 (a) exists. The derivative of a vector-valued function is a tangent vector
to the curve in space which the function represents.

Example:
Let f (t) = (cos t, sin t, t).

(a) Find f 0 (t);

(b) Find the tangent line L at t = 2.
Solution:
(a) f 0 (t) = ( sin t, cos t, 1)

(b) L = f (a) + sf 0 (a) = f (2) + sf 0 (2) = (1, 0, 2) + s(0, 1, 1)

z 6

r f 0 (a) L
-

f (a + h) f (a)
~ 

f (a) >
f (t)
f (a + h)

-
y

+x

Theorem: Let f(t) and g(t) be differentiable vector-valued functions, let u(t) be a differentiable scalar
function, let k be a scalar, and let c be a constant vector. Then
d
(a) dt c =0
d
(b) dt (kf ) = k df
dt

1
 d df dg
(c) dt (f g) = dt dt
d du
(d) dt (uf ) = dt f + u df
dt

d df dg
(e) (f g) = g+f
dt dt dt
d df dg
(f) (f g) = g+f
dt dt dt
Example: Suppose f (t) is differentiable. Find the derivative of ||f (t)||.
Solution:

d d
||f (t)||2 = 2||f (t)||||f (t)|| chain rule
dt dt
d d
but ||f (t)||2 = f (t) f (t), so ||f (t)||2 = (f (t) f (t))
dt dt
d
||f (t)||
2||f (t)|| = f 0 (t) f (t) + f (t) f 0 (t) = 2f 0 (t) f (t)
dt
d f 0 (t) f (t)
||f (t)|| = , if ||f (t)|| =
6 0
dt ||f (t)||

Note that if ||f (t)|| 6= 0, then ||f (t)|| is constant if and only if f(t) is perpendicular to f 0 (t) for all
t.

Just as in single variable calculus, higher order derivatives of vector-valued functions are found by repeated
differentiation of the first derivative of the function:
dn f d dn1 f
 
d d
f 00 (t) = f 0 (t), f 000 (t) = f 00 (t), . . . , n = , n = 2, 3, 4, . . .
dt dt dt dt dtn1
Velocity, acceleration, force, momentum, etc are physical quantities which can be represented by vector-
valued functions. Let t= time elapsed from some initial time t=0 and suppose that an object of mass m is
subjected to some force so that it moves in space, with its position (x,y,z) at time t. i.e x = x(t), y = y(t), z =
z(t). Let r(t) = (x(t), y(t), z(t)) be the position vector of the object. Position: r(t) = (x(t), y(t), z(t))
Velocity: v(t) = r0 (t) = (x0 (t), y 0 (t), z 0 (t))
2
Acceleration: a(t) = v0 (t) = ddt2r = (x00 (t), y 00 (t), z 00 (t))
Momentum: p(t) = mv(t)
Force: F(t) = p0 (t) = dp
dt

1.1.1 Partial Derivative

If the vector-valued function is given in terms of more than one variable, e.g (x,y,z), we can find its partial
derivative as follows:

Definition: The partial derivative of a vector-valued function f=f(x,y,z) with respect to x is given by :

f f (x + x, y, z) f (x, y, z)
= lim
x x0 x
f f
If the limit exists. In a similar way, and , are also found.
y z

1.1.2 Gradient
Definition: The gradient, of a scalar-valued function is defined by:

= i+ j+ k (1)
x y z

2
 The gradient is perpendicular to the level surface at each point and is in the direction of increasing
. The rate of change of in a given direction is the directional derivative. Given a unit vector u, the
directional derivative of in the direction of u
is

Du
= u (2)
Example:
Find the directional derivative of = x2 y 3z at the point (2,1,1) in the direction of the circle
x2 + y 2 + z 2 = 6.
Solution:
= 2xyi + x2 j 3k, the unit vector normal to the given circle is:

(x2 + y 2 + z 2 6) 2xi + 2yj + 2zk xi + yj + zk

u
= =p =
|(x2 + y 2 + z 2 6)| 4x2 + 4y 2 + 4z 2 6
2i + j + k
At (2,1,1), = 4i + 4j 3k and u= .The directional derivative in the direction of the
6
circle is:
2i + j + k 9
Du = (4i + 4j 3k) ( )=
6 6

1.1.3 Divergence
Definition: The divergence of a vector field F is given by:

F1 F2 F3
div F = F = + + (3)
x y z
A solenoidal vector field is one that is neither expanding nor contracting (i.e with div F=0).
Example:
Find the divergence of the vector field F = 2xyi + (x2 y 6z)j (3 sin z + 6y)k
Solution:

F1 F2 F3
F = + +
x y z
2
= (2xy) + (x y 6z) (3 sin z + 6y)
x y z
= 2y + x2 3 cos z

Exercise: Show that the vector field F = 2xyi + (x2 y 2 6z)j (3xey + 6)k is solenoidal.

1.1.4 Curl
The curl of a vector field denoted curl F or F is related to the rotation or twist of the field. The meaning
of curl given should not be taken to be technically strictly correct since there are some vector fields with
non-trivial curl but do not appear to be rotating.

Definition: The curl F or F of a vector field F is defined as:

i j k 
 

 
 
 
F =  (4)

x y z 


 
 
 F 1F 2 F 
3

3
 An irrotational field is one that is neither rotating nor twisting, one in which F = 0.

Example:

Find the curl of the vector field F = yzi + (3x z)j + 2xyzk
Solution

 i j k
 

 
 
 
F =

 x y z 
 
 
 
 yz 3x z 2xyz 
= (2xz + 1)i y(2z 1)j + (3 z)k

Exercise: Show that the vector field F = 2xyi + (x2 6z)j + (3 sin z + 6y)k is irrotational.

Lecture 2
1.2 Integration of Vector functions
1.2.1 Line Integrals
Let r(u) = x(u)i + y(u)j + z(u)k where r(u) is the position vector of (x, y, z). Define a curve C joining points
P1 and P2 where u = u1 and u = u2 respectively. Assume that C is composed of a finite number of curves
for each of which r(u) has a continuous derivative. Let A(x, y, z) = A1 i + A2 j + A3 k be a vector function of
position defined and continuous along C. Then the integral of the tangential component of A along C from
P1 to P2 , written as:
Z P2 Z Z
A dr = A dr = A1 dx + A2 dy + A3 dz (5)
P1 C C

is an example of a line integral. If C is a closed curve, then the integral around C is denoted by
I I
A dr = A1 dx + A2 dy + A3 dz (6)

In general, any integral which is to be evaluated along a curve is called a line integral. Suppose C is a
curve parameterized by x = f (t), y = g(t), a t b and A and B are the points (f(a), g(a)) and (f(b),g(b))
respectively.

(i) C is a smooth curve if f 0 and g 0 are continuous on [a, b] and not simultaneously zero on (a,b).
(ii) C is piecewise smooth if it can be expressed as a union of a finite number of smooth curves.
(iii) C is a closed curve if A = B

(iv) C is a simple closed curve if A=B and the curve does not cross itself.
(v) If C is not a closed curve, then the positive direction on C is the direction corresponding to increasing
values of t.

4
 B B A=B
r r
r

r =B
A

r p
A A (c) closed but not simple
(a) smooth curve
(b) piecewise smooth curve (d) simple closed curve

The region R, interior to a simple closed curve, is said to be simply connected. A simply connected region
does not have any holes in it.
C1


C2
R


C3


R is not simply connected. Let z = G(x, y) be a function of two variables x and y defined on a region
of the plane
R containing a smooth curve C, then the integral of G along C from A to B with respect to arc
length is C G(x, y)ds.
Method of Evaluation
If C is a smooth curve parameterized by x = f (t), y = g(t), a t b, then we have
Z Z b
G(x, y)dx = G(f (t), g(t))f 0 (t)dt, dx = f 0 (t)dt. (7)
C a
Z Z b
G(x, y)dy = G(f (t), g(t))g 0 (t)dt, dy = g 0 (t)dt. (8)
C a
Z Z b p
G(x, y)ds = G(f (t), g(t)) (f 0 (t))2 + (g 0 (t))2 dt (9)
C a

p explicit function y = f (x), a x b, we can use x as a parameter. In this

If the curve is defined by an
case dy = f 0 (x)dx and ds = 1 + [f 0 (x)]2 dx, we then make the appropriate substitutions in the equations
above.

Example: Evaluate C xydx + x2 dy where C is given by y = x3 , 1 x 2.

R

Solution

y = x3 dy = 3x2 dx, we use x as the parameter and hence we have

Z Z 2 Z 2
132
xydx + x2 dy = x(x3 )dx + x2 (3x2 dx) = 4x4 dx =
C 1 1 5

Line integrals along simple closed curves

5
 6
8 r (2,8)

-1
-
2
p
(1, 1)

For a simple closed curve C, its essential to specify the direction of intergration since the initial and fi-
nal points may not be given. The positive direction around a simple closed curve C is that direction a person
must walk on C in order to keep the region R bounded by C to the left. The positive and negative directions
correspond to the counterclockwise and clockwise directions respectively.

} -
' $ ' $
i
R R
R 6
W

& %
 & %
7 

(a) positive direction (b) positive direction (c) negative direction

y 2 dx x2 dy on the curve C shown in the figure below

H
Example: Evaluate C

y 6
(2,4)

c3

y = x2  6c2

- - -
0 c1 (2,0) x

Solution

6
 y 2 dx x2 dy =
H R R R R
= + + , we use x as a parameter. On C1 , y = 0, hence dy = 0 and
RC2 C1 C2 C3 C1

0
0dx x2 (0) = 0. On C2 , we use y as a parameter, x 2, dx = 0, hence
Z Z 4 Z 4
y 2 dx x2 dy = y 2 (0) 4dy = 4dy = 4y|40 = 16
C2 0 0

On C3 , we use x as a parameter, y = x2 , dy = 2xdx implying that

Z Z 0 Z 0
2 2 4 2 8
y dx x dy = x dx x (2xdx) = (x4 2x3 )dx =
C3 2 2 5

8 72
y 2 dx x2 dy = 0 16 +
H
Hence C
=
5 5
Definition: A line integral whose value is the same for every curve connecting A and B is said to be
independent of the path.

Theorem:
R Suppose a function (x, y) such that d = P dx + Qdy i.e P dx + Qdy is an exact differential.
Then C P dx + Qdy depends only on the endpoints A and B of the path C and
Z
P dx + Qdy = (B) (A) (10)
C
R
Note that an integral C P dx + Qdy which is independent of the path between the endpoints A and B is
RB
often written A P dx + Qdy
Theorem: Let P and Q be continuous and have continuous first partial derivatives in a simply connected
R P Q
region. Then C P dx + Qdy is independent of the path C if and only if = for all (x,y) in the region.
y x
Exercise: Show that the integral C (x2 2y 3 )dx + (x + 5y)dy is not independent of the path C.
R

Work done

Suppose that we are moving a particle in a vector field along a smooth path C. Let the force be F such that
F is continuous. In order to get the work done, partition C into N partitions r1 , r2 , . . . , rN , then the
work done in moving a partition ri is approximately Wi = F ri . Summing over all partitions, we get
N
X N
X
W Wi = F ri
i=1 i=1

Thus the work done is

Z N
X
W = F dr = lim Wi (11)
C N
i=1

If r is a position vector, r = xi + yj + zk and hence dr = dxi + dyj + dzk. Suppose F = F1 i + F2 j + F3 k then

Z Z
W = F dr = F1 dx + F2 dy + F3 dz. (12)
C C
I
For a closed path C, we have a closed integral W = F dr.

Example: If F = x sin yi + cos yj is a force field, calculate the work done in moving a particle along the
paths:
(a) A straight line from (0,0) to (1,0) and then to (1,1).
(b) From (0,0) to (1,1) along the curve y = 2x.

7
 (a) (b)

y 6 y 6
(1,1)
(1,2)

C C
c2

- -
0 c1 (1,0) x 0 x

Solution

(a)
Z Z Z
F dr = F dr + F dr
C C1 C2
Along C1
F dr = F1 dx + F2 dy = x sin ydx + cos ydy
but, along C1 , y = 0, hence dy = 0, and sin y = sin(0) = 0
Z
F dr = 0
C1
Along C2
x = 1, dx = 0 F dr = x sin ydx + cos ydy
Z Z 1
F dr = 0+ cos ydy = sin(1)
0
Z
F dr = 0 + sin(1) = sin(1)

(b) On y = 2x, dy = 2dx hence

Z Z
F dr = (x sin 2xi + cos 2xj) (dxi + 2dxj)
C
ZC
= (x sin 2x + 2 cos 2x)dx
C
1 1
Z
1
= x cos 2x 10 + cos 2xdx + sin 2x 10
 
2 2 0
1 5
= 1 cos 2 + sin 2
2 4
Z
Exercise: Evaluate the integral F dr where F = yi 5zj + (xz + z 2 )k along the curve x = 1 t, y = t2
C
and z = t + 1 from t = 1 to t = 2.

Conservative Vector I Fields

The closed integral F dr plays an important role in engineering and other applications. As an example,
C I
if the veclocity of a fluid is v, the integral v dr is called the Circulation of the fluid. A vector field

8
 I
is said to be conservative if F dr = 0. e.g in the gravitational field, the work done in moving a parti-
cle and returning it to the same point is zero, hence the gravitational field is an example of a conservative field.

If a vector field is conservative, then F = 0 and = 0, where F = and is some scalar field .

Example: Show that F = (ex sin y + 2 cos z)i + (ex cos y 2y)j 2x sin zk represents a conservative
field and find a scalar potential such that F = .
Solution
i j k
 
 
 
F =
 
x y z
 
 
 ex sin y + 2 cos z ex cos y 2y 2x sin z 
= 0i (2 sin z + 2 sin z)j + (ex cos y ex cos y)k
= 0i + 0j + 0k = 0.

Hence F is conservative. Now F = for some scalar , that is F = i+ j+ k. Equating
x y z

corresponding components, we find = ex sin y + 2 cos z. Integrating with respect to x gives
x
= ex sin y + 2x cos z + h(y, z), we differentiate with respect to y, to get

= ex cos y + h0 (y, z) comparing with F, gives
y
h0 (y, z) = 2y, integrating with respect to y gives
h(y, z) = y 2 + g(z)
Therefore = ex sin y + 2x cos z y 2 + g(z)
lastly, differentiating with respect to z yields

= 2x sin z + g 0 (z)
z
giving g 0 (z) = 0 g(z) = k
where k is some arbitrary constant
Therefore = ex sin y + 2x cos z y 2 + k

Lecture 3
1.2.2 Surface Integrals

~ru Si
~ru 
6

= ~n
- -
%
% ~rv
-
~rv

A surface integral S can be defined by a function r(u, v) where u and v are parameters. At any point P,
r r ru rv
ru = and rv = are tangential to the u and v coordinates. Then n = is a normal vector to
u v |ru rv |

9
S. Now considering an element Si formed by ru and rv , n is the normal to Si . We define the surface
integral as
Z Z N
X
F dS = lim F Si (13)
S Si 0
i=1

Note that ds = lim |rui ui rvi vi | = |ru rv |dudv. In the cartesian coordinate system, the area element
i
for the xy- plane is dxdy, for the yz-plane is dydz and the xz-plane is dxdz.
dxdy
The projection of dS into the xy-plane is . The surface integral is sometimes called the flux integral
|
n k| HH
as it represents the total flux out of the surface S. If the surface is closed, we have F dS. We may also
have the following forms of surface integrals
Z Z Z Z Z Z
dS, FdV, dS
S S S
I I
Example: Evaluate the integral Fn
dS over the region bounded by z = 0, z = 2 and the surface
S
x2 + y 2 = 4 in the first octant and F = xyi + y 2 j + 2yzk.
Solution

z 6

2 s2

*
s4
n

-
s3
s5

2 -
y

s1
3
2
x~

I I Z Z
On S1 : z = 0, n
= k, Fn
dS = 2 yzdxdy = 0
S C

I I Z Z Z 2 Z 4y2 Z 2 p
On S2 : , z = 2, n
= k, F
ndS = 2 yzdxdy = 4 ydxdy = 4 y 4 y 2 dy =
S 0 0 0
32
3

10
 I I Z Z
= j,
On S3 : y = 0, n Fn
dS = y 2 dxdz = 0
S C
I I Z Z
= i,
On S4 : x = 0, n Fn
dS = xydydz = 0
S C

(x2 + y 2 4) 2xi + 2yj xi + yj

On S5 : the normal to the surface is n
= 2 2
= p = . Project the
|(x + y 4)| 2
(2x) + (2y) 2 2
dxdz 2dxdz y
surface onto the xz-plane dS = = . Note that |n j| = | |. We now evaluate the integral:
|
n j| y 2
Z Z 2 Z 2Z 2 2 Z 2Z 2
x y + y 3 dxdz x y + y3
Z Z Z Z
dxdz
Fn dS = Fn = = dxdz = 4dxdz = 16
S S (y/2) 2 (y/2) 0 0 y 0 0
I I
32 80
When we add everything, we get Fn dS = + 16 = .
S 3 3

Application: The flow rate per unit time can be represented by a surface integral. As an example, if a fluid
is flowing through a pipe, the amount of fluid through a crossectional area per unit time can be determined
as a surface integral.

1.3 Integral theorems

1.3.1 Gauss (divergence) theorem
Theorem: Let F~ be a vector field that is continuously differentiable in a volume V, and S be the surface
forming the boundary of the volume V, and n be the unit normal vector to S, then:
Z Z Z I I
F~ dV = F~ n
dS (14)
V S

which can be stated as follows in rectangular coordinate system

Z Z Z   I I
F1 F2 F3
+ + dV = F1 dx + F2 dy + F3 dz (15)
V x y z
I I
Example: Use the divergence theorem to evaluate dS where F~ = xy~i + y 2~j + 2yz~k in the
F~ n
region bouned by z = 0, z = 2 and x2 + y 2 = 4.

Solution:
Using the divergence theorem, we have the following set up
I I Z Z Z Z Z Z
F~ n
dS = F~ dV = (y + 2y + 2y)dV
V
Z 2 Z 2 Z 4y2 Z 2Z 2 p
= 5ydxdydz = 5y 4 y 2 dydz
0 0 0 0 0
Z 2 Z 0 Z 2
5 1 5 80
= u 2 dudz = 8dz =
2 0 4 3 0 3

1.3.2 Stokes Theorem

Theorem: Let C be a closed curve which forms the boundary of a surface S. If the vector F~ is continuously
differentiable, then:
Z Z I
F~ n
dS = F~ d~r (16)
S C

where the direction of the line integral around C and the normal n
are orientated in the right-hand sense.

11
 I
Example: Use Stokes theorem to evaluate F~ d~r where F~ = z 2 x~i + (x 1)~j + 2~k and S is the upper
hemisphere x2 + y 2 + z 2 = 4.

Solution:

z 6

-2

2
2 q
x y

2x~i + 2y~j + 2z~k x~i + y~j + z~k

The unit normal to the hemisphere is n
= p = . Projecting the surface
2 x2 + y 2 + z 2 2
dxdy
onto the xy-plane gives dS = . We npw calculate F~ as follows:
(z/2)
 

 ~i ~j ~k 

~
  = 2xz~j + ~k

F = 
 x y z 

 z2x x1 2 

x~i + y~j + z~k 2xyz + z

F~ n
= (2xz~j + ~k) =
2 2
I Z 2 Z 4y2 Z 2 Z 4y2
2xyz + z dxdy
Hence F~ d~r = = (1 + 2xy)dxdy
C 2 4y 2 2 (z/2) 2 4y 2
Z 2 Z 2
 2  4y2 p
= x y + x 2 dy = [y(4 y 2 ) y(4 y 2 ) + 2 4 y 2 ]dy
2 4y 2
(let y = 2 sin y 2 = 4 sin2 , dy = 2 cos )
Z /2
8 /2
Z
= 8 cos2 = (1 + cos 2)d = 4
/2 2 /2

1.3.3 Greens Theorem

P
Theorem: Suppose that C is a picewise smooth simple closed curve bounding a region R. If P, Q, and
y
Q
are continuous on R then:
x
I Z Z  
Q P
P dx + Qdy = dA (17)
C R x y

12
 I
Example: Evaluate (x2 y 2 )dx + (2y x)dy where C consists of the boundary of the region in the
C
first quadrant that is bounded by the graphs of y = x2 and y = x3 .

Solution:
y 6

y = x2

^ R y = x3
9



-
0 1 x

I Z Z Z 1 Z x2
(x2 y 2 )dx + (2y x)dy = (1 + 2y)dA = (2y 1)dydx
C R 0 x3
Z 1 Z 1
x2 11
y + y 2 x3 dx = (x6 + x4 + x3 x2 )dx =

=
0 0 420

13
2 Linear Algebra
This section will discuss four major concepts: eigenvalues, eigenvectors, diagonalisation and applications of
eigenvalues and eigenvectors.

Lecture 4
2.1 Eigenvalues and eigenvectors
Characteristic polynomial:
If A is an nn matrix, the polynomial Pn () of degree n in the scalar defined as Pn () = det(AI)
is called the characteristic polynomial of A.
Eigenvalues:
The roots of the equation Pn () = 0 are called eigenvalues of A.
Eigenvectors:
The column vectors X1 , X2 , . . . , Xn satisfying the equation (A i I)Xi = 0 are the eigenvectors of A.
In general, we seek solutions to the homogeneous system of equations AX = X or (A I)X = 0. It
should be noted that a matrix with complex coefficients will have complex eigenvalues. It is however
possible for a matrix with real entries to have complex eigenvalues.
If an eigenvalue is repeated r times corresponding to the presence of a factor ( )r in Pn (),
the number r is called the algebraic multiplicity of .
Spectrum: is the set of all eigenvalues 1 , 2 , . . . , n .
Spectral radius of A: R = max(|1 |, |2 |, . . . , |n |)
theorem:
Linear independence of eigenvectors: the eigenvectors X1 , X2 , . . . , Xm corresponding to m distinct
eigenvalues of an n n matrix A are linearly indepndent.
When an eigenvalue with algebraic multiplicity r > 1 has s different eigenvectors associated with it,
where s < r then s is called the geometric multiplicity of the eigenvalue.
The set of all eigenvectors associated with an eigenvalue with geometric multiplicity s together with
the null vector 0 forms the eigenspace associated with the eigenvalue.
Example 2.1: Find the characteristic polynomial, the eigenvalues and the eigenvectors of the matrix

2 1 1
A= 3 2 3
3 1 2
Solution to Example 2.1

 
 2 1 1 
 
P3 () = |A I| =  3 2 3 

 3 1 2 
     
 2 3   3 3   3 2 
= (2 )  1  1 
1 2  3 2  3 1 
= 3 + 22 + 2

The characteristic equation is P3 () = 0, hence 3 + 22 + 2 = 0 with roots = 2, 1 and -1 as

eigenvalues.
Eigenvectors: we now find eigenvectors for each eigenvalue as follows:
Case 1 = 2

14

22 1 1 x1 0
(A I)X = 3 22 3 x2 = 0
3 1 2 2 x3 0

0 1 1 x1 0
3 0 3 x2 = 0
3 1 4 x3 0
x2 x3 = 0, 3x1 3x3 = 0, 3x1 + x2 4x3 = 0
x2 = x3 and x1 = x3 x1 = x2 = x3

Which also satisfies the last equation, thus the first eigenvector is given by:

k1 1 1
X1 = k1 = k1 1 = 1
k1 1 1
This is possible after setting x3 = k1 and then choosing k1 = 1.

Case 2 = 1

21 1 1 x1 0
3 21 3 x2 = 0
3 1 2 1 x3 0
x1 + x2 x3 = 0 (18)
3x1 + x2 3x3 = 0 (19)
3x1 + x2 3x3 = 0 (20)

From equation (18) x2 = x3 x1 and from (19) and (20), x2 = 3x3 3x1 , implying that x1 = x3 and
x2 = 0. The second eigenvector is thus found as follows:

k2 1 1
X2 = 0 = k2 0 = 0 , if if we let x3 = k2 and k2 = 1
k2 1 1

Case 3 = 1

3 1 1 x1 0
3 3 3 x2 = 0
3 1 1 x3 0
3x1 + x2 x3 = 0 (21)
3x1 + 3x2 3x3 = 0 (22)
3x1 + x2 x3 = 0 (23)

From (21), x2 = x3 3x1 and substitute in (22) to obtain the following:

3x1 + 3[x3 3x1 ] 3x3 = 3x1 9x1 + 3x3 3x3 = 0 implying that x1 = 0 and x2 = x3 = k3 , thus:

0 0
X3 = k3 = 1 if we let k3 = 1.
k3 1

Exercise
1:
1 2 1
Let A = 1 0 1 . Find the characteristic polynomial, eigenvalues and eigenvectors of A.
4 4 5

15
 solution:
P3 () = 3 62 + 11 6 and the eigenvalues are 1 = 1, 2 = 2 and 3 = 3. The associated
eigenvectors are as follows

1 2 1
X1 = 1 , X2 = 1 , X3 = 1
2 4 4

2.1.1 Normalisation of eigenvectors

a
Let X = b be an eigenvector, then the normalised eigenvector ]bfX is found by dividing X by
c
a


a2 +b2 +c2
= b
a2 + b2 + c2 , so that X . The normalised eigenvector will have unit length.

a2 +b2 +c2
c
a2 +b2 +c2

Theorem: Eigenvalues and eigenvectors of a symmetric matrix:

Let A be an n n real symmetric matrix. Then
(i) The eigenvalues of A are all real;
(ii) The eigenvectors of A corresponding to distinct eigenvalues are mutually orthogonal.
Theorem: Similar matrices have the same characteristic polynomial. (Proof: Kolman B, page 351).
Let A and B be similar, then B = P 1 AP for some nonsingular matrix P.

2.1.2 Properties of Eigenvalues and Eigenvectors

Property 1:
Let A be an n n matrix and 1 , 2 , . . . , n be the eigenvalues of A. Then:
(i) AT has the same eigenvalues 1 , 2 , . . . , n .
(ii) A1 (if it exists) has eigenvalues 1 1 1
1 , 2 , . . . , n .
(iii) The matrix A I has eigenvalues 1 , 2 , . . . , n .
(iv) For any non-negative integer k, the matrix Ak has eigenvalues k1 , k2 , . . . , kn
Property 2:
For any square matrix A, the sum of eigenvalues is equal to the sum of diagonal elements of
A(trace(A)).
Property 3:
For any square matrix A, the product of eigenvalues is equal to the determinant of A.
Property 4:
The eigenvectors of a square matrix A corresponding to distinct eigenvalues are linearly indepen-
dent,i.e One cannot be written as a linear combination of the other eigenvectors.

Exercises
1. For each of the following matrices A1 to A6 , find the characteristic polynomial (Pn ()), eigenvalues
and eigenvectors.

1 1 0 1 0 0 3 2 2
A1 = 1 1 0 A2 = 0 1 1 A3 = 6 4 6
0 0 1 0 1 1 2 1 3

0 1 2 2 1 3 4 1 1
A4 = 2 1 2 A5 = 1 1 1 A6 = 1 0 2
2 2 2 1 0 1 1 1 2

2. For each of the matrices given above, find eigenvalues of A1 and AT where possible.

16
 2.2 Diagonalisation of matrices
Lecture 5
We wish to diagonalise an n n matrix A so that it reflects the simple properties of a diagonal matrix.
Let D denote a general n n diagonal matrix:

1 0 0
0 2 0
D=


0 0 0 n

then the eigenvalues of D are 1 , 2 , . . . , n and the corresponding n linearly independent eigenvectors
are:
0
1 0 0 0
0 1 0
0
0 0
, , Xn1 = , Xn = 0

X1 = , X2 =










1
0 0 1
0

m

1 0 0
m
 
and D1 = diag 1 , 1 , . . . , 1 .
0 0
Note that Dm =
2
1 2 n
0 0 0 m
n

Theorem: Diagonalisation of an n n matrix: Let the nn matrix A have n eigenvalues 1 , 2 , . . . , n ,

not all of which need to be distinct and let there be n corresponding distinct eigenvectors X1 , X2 , . . . , Xn
so that
AXi = i Xi , i = 1, 2, . . . , n
Define the matrix P = [X1 X2 Xn ], then P 1 AP = D. Note that P is composed of eigenvectors
of A.
Remarks about diagonalisation:
(i) An n n matrix can be diagonalised provided it possesses n linearly independent eigenvectors.
(ii) A symmetric matrix can always be diagonalised.
(iii) The diagonalising matrix for a real n n matrix may contain complex elements.
(iv) A diagonalising matrix is not unique because its form depends on the order in which the eigen-
vectors of A are used to form columns.
Note that since D = P 1 AP , by simple matrix manipulation, it follows that A = P DP 1 . Multiples
of A can easily be calculated from this relationaship as follows:

A2 = (P DP 1 )(P DP 1 ) = P DP 1 P DP 1 = P D2 P 1
A3 = P D2 P 1 P DP 1 = P D3 P 1
.. ..
. .
Am = P Dm P 1

Example 2:

2 1 1
Diagonalise the matrix A = 3 2 3 and use the result to find A5 .
3 1 2
solution to Example 2:

17
 It has already been shown that the eigenvalues are 1 = 2, 2 = 1, 3 = 1 and the corresponding
eigenvectors are:
1 1 0
X1 = 1 , X2 = 0 , and X3 = 1
1 1 1

1 1 0
Hence the diagonalising matrix P is given by P = [X1 X2 X3 ] = 1 0 1
1 1 1

1 1 1 2 0 0
P 1 = 0 1 1 and D = P 1 AP = 0 1 0
1 0 1 0 0 1

25

1 1 0 0 0 1 1 1
A5 = P D5 P 1 = 1 0 1 0 15 0 0 1 1
1 1 1 0 0 (1)5 1 0 1

32 31 31
= 33 32 33
33 31 32

If the normalised eigenvectors of A are used to form the columns of P , then the resulting diagonal-
ising matrix will be an orthogonal matrix. However, if an eigenvalue is repeated, the corresponging
eigenvectors will not be orthogonal to the other eigenvectors.
Theorem:
An n n matrix A is diagonalisable if all the roots of Pn () are real and distinct.
Note that if the roots of Pn () are all real, then A can be diagonalised if and only if for each eigenvalue
of multiplicity k, we can find k linearly independent eigenvectors.
Example 3:

0 0 1
Let A = 0 1 2 . Show that A cannot be diagonalised.
0 0 1
Solution to example 3:

 
 0 1     
   1 2   0 1 
P3 () = |A I| =  0 1 2
 =  0
  + 1 
1  0 0 
 0 0 1 
= [(1 )2 ] = (1 )2 = 0
Eigenvalues are 1 = 0, 2 = 3 = 1. Its clear that 2 = 1 has multiplicity 2 and A is diagonalisable
if we can find two linearly independent eigenvectors corresponding to 2 = 1.
Case 2 = 1:

1 0 1 x1 0
0 0 2 x2 = 0 , giving the following equations: x1 + x3 = 0, x3 = 0 and x1 = 0.
0 0 0 x3 0
x2 is a free variable in this case leaving us with the following eigenvector

0 0 0
X2 = k2 = k2 1 = 1 , if we let k2 = 1
0 0 0
We cannot find another linearly independent vector for 2 = 1. Hence A cannot be diagonalised.

18
 2.2.1 Diagonalisation of symmetric matrices

Theorem: All the roots of the characteristic polynomial of a real symmetric matrix are real numbers.
If A is a symmetric n n matrix and all the eigenvalues of A are distinct, then A is diagonalisable.

Theorem: If A is a symmetric n n matrix, then eigenvectors that belong to distinct eigenvalues of A

are orthogonal.
Definition: A square matrix A is called orthogonal if A1 = AT OR if AT A = In .
Theorem: If A is a symmetric n n matrix, then there exists an orthogonal matrix P such that
P 1 AP = P T AP = D, a diagonal matrix. The eigenvalues of A lie on the main diagonal of D.

2.2.2 Application of eigenvalues and eigenvectors

Many systems arising in engineering applications can be represented as discrete models involving
matrices. Many of the applications involve the use of eigenvalues and eigenvectors in the process of
transforming a given matrix into a diagonal matrix.

The natural frequency of a bridge is the eigenvalue of smallest magnitude of a system that models the
bridge. Engineers use this knowledge to ensure the stability of their constructions.
Eigenvalue analysis is used in the design of car stereo systems where it helps to reduce the vibration
of the car due to the music.
In electrical engineering, the application of eigenvalues and eigenvectors is useful for decoupling three-
phase systems through symmetrical component transformation.
Eigenvalues can be used in solving systems of differential equations. We consider this application in
more detail by looking at one simple example.
Example 4:

Solve the following system of linear equations

dx1
x 1 = = 4x1 + 2x2
dt
x 2 = x1 + x2 (24)

with initial conditions x1 (0) = 1 and x2 (0) = 0.

Solution to example 4:

Express (24) in matrix form as X = AX, implying that we can express it as follows:
      
x 1 4 2 x1 4 2
= , A=
x 2 1 1 x2 1 1

In order to find the diagonalising matrix P , we find the eigenvalues and eigenvectors of A.
 
 4 2 
|A I| =  = (4 )(1 ) + 2
1 1 
2 5 + 6 = 0
( 2)( 3) = 0

giving eigenvalues 1 = 2 and 2 = 3.

Case:1 = 2:

19
     
2 2 x1 0
= . Which gives the following equations: x1 + x2 = 0 and x1 x2 = 0
1 1 x2 0  
1
suggesting that x1 = x2 , hence the eigenvector is X1 = .
1
Case: 2 = 3
    
1 2 x1 0
= , x1 + 2x2 = 0, x1 2x2 = 0.Thus x1 = 2 and x2 = 1, hence
1 2 x2 0
     
2 1 2 2 0
X2 = . Thus the diagonalising vector P = with D = .
1 1 1 0 3

But A = P DP 1 and X = AX = P DP 1 . We premultiply by P 1 to obtain

P 1 X = DP 1 X (25)

Let y = P 1 X, implies that (25) becomes

y = Dy (26)

C1 e2t
 
Solving, we get y(t) = . We then do a back-transformation to obtain the following:
C2 e3t

C1 e2t C1 e2t 2C2 e3t

    
1 2
X = Py = =
1 1 C2 e3t C1 e2t + C2 e3t
Hence x1 (t) = C1 e2t 2C2 e3t
x2 (t) = C1 e2t + C2 e3t

Using the initial conditions: x1 (0) = C1 2C2 = 1 and x2 (0) = C1 + C2 = 0 giving C1 = 1 and
C2 = 1 hence we have the following solution:
x1 (t) = 2e3t e2t and x2 (t) = e2t e3t .

20