M1GLA lecture notes

as lectured by Prof. Liebeck

Imperial College London

Mathematics 2006/2007

ii

CONTENTS

iii

Contents
1

Introduction

Geometry in two dimensions

2.1 Vector operations . . . .
2.2 Lines . . . . . . . . . . .
2.3 Triangles . . . . . . . . .
2.4 Distances . . . . . . . .
2.5 Dot product . . . . . . .
2.6 Equation of a line . . . .
2.7 Perpendicular distance .
2.8 Two famous inequalities

Linear inequalities

Conics
4.1 Transforming conics .
4.1.1 Translation .
4.1.2 Rotation . . .
4.2 The Theory . . . . .
4.3 Geometrical definition

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

3
3
4
6
7
7
8
9
10
13

. . . . . .
. . . . . .
. . . . . .
. . . . . .
of conics

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

15
18
18
19
20
22

Matrices and linear equations

25
5.1 Method Gaussian elimination . . . . . . . . . . . . . . . . . . . . . . 27

Some applications of matrix algebra

6.1 Linear equations . . . . . . . . . . .
6.2 Population Distribution . . . . . . .
6.3 Matrices and Geometry . . . . . . .
6.3.1 Rotation . . . . . . . . . . .
6.3.2 Reflection . . . . . . . . . .
6.3.3 Combining Transformations

.
.
.
.
.
.

37
37
38
41
41
41
41

Inverses
7.1 Relation to linear equations . . . . . . . . . . . . . . . . . . . . . . . .
7.2 Finding inverses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
7.3 Finding inverses in general . . . . . . . . . . . . . . . . . . . . . . . .

43
43
44
45

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

iv

CONTENTS

Determinants
47
8.1 Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
8.2 Effects of row operations on determinant . . . . . . . . . . . . . . . . 49
8.3 n n determinants . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51

Eigenvalues and eigenvectors

9.1 Eigenvectors . . . . . . . . .
9.2 How to find eigenvectors and
9.2.1 Back to Fibonacci .
9.3 Diagonalization . . . . . . .
9.3.1 Repeated eigenvalues

. . . . . . .
eigenvalues
. . . . . . .
. . . . . . .
. . . . . . .

10 Conics(again) and Quadrics

10.1 Reduction of conics . . . . . . .
10.1.1 Orthogonal matrices . .
10.2 Quadric surfaces . . . . . . . . .
10.3 Linear Geometry in 3 dimensions
10.4 Geometry of planes . . . . . . .

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

53
53
54
55
56
60

.
.
.
.
.

63
65
67
67
70
70

Chapter 1

Introduction
Some Greek highlights

Chapter 2

Geometry in two dimensions

This means geometry in the plane. Based on the real numbers. Think of real numbers
as points on the real line.
Definition 2.1. R2 is the set of all ordered pairs (x1 , x2 ) with x1 , x2 real numbers.
Note: (1, 3) 6= (3, 1)
Call (x1 , x2 ) a vector. Geometrically, vector (x1 , x2 ) will be used to represent several
things:
Point p in the plane with coordinates (x1 , x2 )

Position vector OP

line in direction OP , length OP

Any vector AB of the same direction as OP

Usually we write x = (x1 , x2 ).
The origin or zero vector is O = (0, 0).

2.1

Vector operations

Definition 2.2.
1) Addition
define the sum of two vectors by
(x1 , x2 ) + (y1 , y2 ) = (x1 + y1 , x2 + y2 )
2) Scalar multiplication
A scalar is a real number. If x = (x1 , x2 ) is a vector and R is a scalar
x = (x1 , x2 ) = (x1 , x2 )
Subtraction
x y = x + (y ) = (x1 y1 , x2 y2 )

2.2. LINES

Note. If , R and x , y R2
i) (x + y ) = x + y
ii) ( + )x = x + x
Proof.
(i)
(x + y ) = (x1 + y1 , x2 + y2 )
= ((x1 + y1 ), (x2 + y2 ))
= (x1 + y1 , x2 + y2 )
= (x1 , y1 ) + (y1 , y2 )
= x + y
(ii) excercise


2.2

Lines

In Euclid, points and lines are undefined. But we need to define points and lines. We
have already defined points.
Definition 2.3. Let a, u be vectors in R2 with u 6= O. The set of vectors {a + u |
R} is called a line.
Definition 2.4. Two lines {a + u | R} and {b + v | R} are parallel iff
v = u for some R.
Next we will prove the three famous propositions from Euclid.
Proposition 2.1. If L, L are parallel lines which have a point in common, then L = L .
Proposition 2.2. Through any two points, there is exactly one line.
Proposition 2.3. Two non parallel lines meet in exactly one point.
Proof. Proposition 2.1.
Let
L = {a + u | R}

L = {b + v | R}
As L, L are parallel, v = u for some scale R, so
L = {b + u | , R} = {b + u | R}

2.2. LINES

Given that L and L have x in common, we have

x

= a + 1 u = b + 2 u

b = a + 1 u 2 u

1 , 2 R

= a + (1 2 )u

b = a + 3 u

(3 = (1 2 ) 3 R)

Finally, we get
L = {b + u | R}

= {a + 3 u + u | R}

= {a + (3 + )u | R}
= {a + u | R}
= L

Proof. Proposition 2.2.

Let u, v R2 to be our two points. Let us take the line
L = {u + (v u) | R}

Suppose now that L is some line through u and v

L = {a + w | R}

Since points u and v line on L , we get

u = a + 1 w
v

= a + 2 w

v u = (2 1 )w

Therefore, lines L and L are parallel. As they also have a common point, they are (by
proposition 1) identical.


Example 2.1. Find common point of

L1 = {(0, 1) + (1, 1) | R}

L2 = {(4, 0) + (2, 1) | R}

By Proposition 3, there exists a unique common point, call it x .

x

= (0, 1) + 1 (1, 1)

= (4, 0) + 2 (2, 1)

Writing equations for individual coordinates, we get

0 + 1 = 4 + 22
1 + 1 = 0 + 2
1 = 4 + 2
2 = 5

1 = 4 + 2(5)

1 = 6

2.3. TRIANGLES

Therefore
x

= (0, 1) 6(1, 1) = (6, 5)

= (4, 0) 5(2, 1) = (6, 5)

2.3

Triangles

Definition 2.5. A triangle is a set of 3 non-colinear points {a, b, c | a, b, c R2 }.

Edges of a triangle are the line segments ab, ac, bc, where
ab = {a + (b a) | 0 1}
Midpoint of ab is the point
1
1
a + (b 1) = (a + b)
2
2
A median of a triangle is a line joining one of a, b, c with the midpoint of the oposite
side (line segment).
Proposition 2.4. The 3 medians of a triangle meet in a common point.
Proof. Let the three medians be Ma , Mb and Mc
1
Ma = {a + ( (b + c) a) | R}
2
1
Mb = {b + ( (a + c) b) | R}
2
1
Mc = {c + ( (a + b) c) | R}
2
We just show, that for = 23 , medians meet at the same point
1
Ma = a + b +
3
1
Mp = b + a +
3
1
Ma = c + a +
3

1
c
3
1
c
3
1
b
3

2
a=
3
2
b=
3
2
c=
3

1
(a + b + c)
3
1
(a + b + c)
3
1
(a + b + c)
3

Therefore, all the medians contain the point (sometimes called centroid) 13 (a + b + c).


Note. Other interesting properties of triangles are

3 altitudes meet at a point (sometimes called orthocentre)
3 perpendicular bisectors meet at a point (sometimes called circumcentre)
the 3 centres (orthocentre, circumcentre, centroid) are colinear, (they lie on Euler
line)

2.4. DISTANCES

2.4

Distances

Definition 2.6. The length of a vector x = (x1 , x2 ) (denoted kx k) is a real number

kx k =

x12 + x22

The distance between x and y (x , y R2 ) is the length of (x y ) called dist(x , y )

dist(x , y ) =

2.5

(x1 y1 )2 + (x2 y2 )2

Dot product

Definition 2.7. For x , y R2 , dot product (scalar product) of x and y is a real number
x .y = x1 y1 + x2 y2
Example 2.2. x = (1, 2), y = (1, 0)
kx k =

dist(x , y ) =
x .y
Proposition 2.5.

12 + 22 =

22 + 22 = 2 2

= 1 + 0 = 1

(i) kx k2 + ky k2 kx y k2 = 2x .y

(ii) If is the angle between x and y then x .y = kx kky k cos

(iii) The position vectors of x and y are at right angles iff x .y = 0
Proof.
(i)
LHS = x12 + x22 + y12 + y22 (y1 x1 )2 (y2 x2 )2
= (2x1 y1 ) (2x2 y2 )

= 2x .y
(ii) From cosine rule

kx k2 + ky k2 ky x k2
= cos
2kx kky k
Using part (i), we get
2x .y
x .y

= 2kx kky k cos

= kx kky k cos

2.6. EQUATION OF A LINE

2.6

Equation of a line

Consider a line L = {a + u | R}, u = (u1 , u2 ), define n = (u2 , u1 ), so that

n.u = 0.
For any x L, x = a + u
x .n = (a + u).n = a.n + u.n = a.n
x .n = a.n
Proposition 2.6. Let L = {a + y | R}, n = (u2 , u1 ). Then
(i) Every x in L satisfies x .n = a.n
(ii) Every solution of x .n = a.n lies on L
Proof.
(ii) Suppose x .n = a.n (we need to show that x = a + u, i.e. x a = u). Then
(x a).n = 0.
Let y = x a, and say y = (y1 , y2 ). As y .n = 0
y1 u2 + y2 u1 = 0

y2 u1 = y1 u2

So, y = uy11 (u1 , u2 ) if u1 6= 0, or y =

Therefore
y
x

y2
u2 (u1 , u2 )

if u2 6= 0.

= u = x a
= a + u L

Definition 2.8. For a line L = {a + u | R}, the vector n = (u2 , u1 ) (or any
scalar multiple of it) is called a normal to L.
The equation x .n = a.n is called the equation of L.
Proposition 2.7. The linear equation px1 + qx2 + r = 0 (1) is an equation of a line
with normal (p, q) and with direction (q, p).
Proof.
Suppose q 6= 0. Then the solution of (1)
r
p
x1 + x2 + = 0
q
q
Let x1 = , x2 =

r
q

p
q ,

so



p
r
(x1 , x2 ) =
,
q
q




p
r
+ 1,
=
0,
q
q


1
which is a line with direction vector 1, p
= q
(q, p). Normal is clearly
q
(p, q).

2.7. PERPENDICULAR DISTANCE

When q = 0. Equation (1) is px1 + r = 0, with solution



r
(x1 , x2 ) =
, 0 + (0, 1)
p
Which is clearly a line with equation 1p (0, p) = p1 (q, p).


Definition 2.9. Lines L1 = {a + u | R} and L2 = {b + v | R} are

perpendicular iff u.v = 0.
Proposition 2.8. Let lines L1 , L2 have equations
L1 : p1 x1 + q1 x2 + r1 = 0
L2 : p2 x1 + q2 x2 + r2 = 0

Then
(i) L1 , L2 are parallel iff (p1 , q1 ) = (p2 , q2 ) for some scalar
(ii) L1 , L2 are perpendicular iff (p1 , q1 ) (p2 , q2 ) = 0
Proof.
(i) By Proposition 2.7
L1 = {a + (q1 , p1 ) | R}

L2 = {b + (q2 , p2 ) | R}

These are parallel iff R such that (q2 , p2 ) = (q1 , p1 ), i.e. (p1 , q1 ) =
(p2 , q2 ).
(ii) excercise


2.7

Perpendicular distance

Definition 2.10. For a line L and point p, p

/ L, dist(p, L) is the length of the
perpendicular from p to L.
Proposition 2.9. If a is a point on the line and n is a normal to the line, then


 (p a).n 


dist(p, L) = 
knk 
Note. A unit vector is a vector of length 1. If w is any vector, then w =
vector. Therefore we can write
dist(p, l) = |(p a).
n|

w
kw k

is a unit

10

2.8. TWO FAMOUS INEQUALITIES

Proof.
dist(p, L) = |kp ak cos |


 kp akknk cos 


= 

knk


 (p a).n 

= 
knk 

2.8

Two famous inequalities

Proposition 2.10. (Cauchy - Schwartz inequality)

Let x , y R2 .
(i) |x .y | kx kky k
(ii) If |x .y | = kx kky k then one of x .y is a scalar multiple of the other
Proof.
(i)
|x .y |
2

(x1 y1 + x2 y2 )
x12 y12

+ 2x1 y1 x2 y2 +

x22 y22
0
2

(x1 y2 x2 y1 )

kx kky k

(x12 + x22 )(y12 + y22 )

x12 y12 + x12 y22 + x22 y12 + x22 y22

x12 y22 + x22 y12 2x1 y1 x2 y2

(ii) If |x .y | = kx kky k, above proof shows

0 = x1 y2 x2 y1
y1
x1
=
x2
y2
Therefore one of x and y is a scalar multiple of the other.
Note that x .x = (x12 + x22 ) = kx k2 .
Proposition 2.11. (Triangle inequality) If x , y R2 , then
kx + y k kx k + ky k

2.8. TWO FAMOUS INEQUALITIES

11

Proof.
kx + y k2 = (x + y )(x + y ) = x .x + y .y + 2x .y
= kx k2 + ky k2 + 2x .y

kx k2 + ky k2 + 2kx kky k

by Cauchy - Schwartz

(kx k + ky k)
Therefore kx + y k kx k + ky k.

12

2.8. TWO FAMOUS INEQUALITIES

13

Chapter 3

Linear inequalities

14

15

Chapter 4

Conics
In R2 , a linear equation px1 + qx2 + r = 0 defines a line. Now we define a curve in R2
by quadratic equation.
Example 4.1. Circle of center c, radius r . It contains all the points x such that
kx ck = r

kx ck2 = r 2

(x1 c1 )2 + (x2 c2 )2 = r 2

x12 + x22 2c2 x2 + c12 + c22 r 2 = 0

Definition 4.1. A conic section in R2 is the set of poitns x R2 , x = (x1 , x2 ),

satisfying a quadratic equation
ax12 + bx22 + cx1 x2 + dx1 + ex2 + f = 0
where not all of a, b and c are 0.
Here are some basic examples
(1)
x12 x22
+ 2 =1
a2
b
is an ellipse
b
a

(2)
x12 x22
2 =1
a2
b
is a hyperbola

(4.1)

16

(3)
x2 = ax12 + b
(a 6= 0) is a parabola.

(4)
x12 x22
2 =0
a2
b
is a pair of lines.

(5)
x12
=1
a2
is a pair of lines.

17

(6)
x12
=0
a2
is a double line
(7)
x12 x22
+ 2 =0
a2
b
is a point
(8)
x12 x22
+ 2 = 1
a2
b
is the empty set (empty conic)
We call the conics (6), (7) and (8) degenerate conics. Well see how to reduce an
arbitrary conic to one of these basic examples.

18

4.1. TRANSFORMING CONICS

Note. Why conic sections?

4.1

Transforming conics

Aim start with an arbitrary conic in the form (4.1), do some geometrical chages to
coordinates so that the equation becomes one of our standard equations (1) (8).
Two kinds of changes are allowed.

4.1.1

Translation

To change coordinates from (x1 , x2 ) to (y1 , y2 ) with new origin (, )

y1 = x1
y2 = x2
This is called a translation.
Example 4.2. What type is the following conic?
x12 + 2x22 2x1 + 4x2 10 = 0
We complete the square
(x1 1)2 + 2(x2 + 1)2 13 = 0
So the new coordinates are
y1 = x1 1

y2 = x2 + 1
Now, we can easily see that the our conic is an ellipse.
Remark If (4.1) has no x1 x2 term (c = 0), then we can find a translation which
reduces the equation to one of the standard ones (examples (1) (6)).

4.1. TRANSFORMING CONICS

4.1.2

19

Rotation

Rotate axes anticlockwise through . What happens to the coordinates of a general

point P ?
Say P has old coordinates (x1 , x2 ) and new coordinates (y1 , y2 ).

x2
b

y2
r

y1

x1

Now
y1 = r cos
y2 = r sin
and
x1 = r cos( + )
x2 = r sin( + )
Hence
x1 = r cos cos r sin sin
= y1 cos y2 sin

and
x2 = r sin cos + cos sin
= y2 cos + y1 sin
Summarizing change of coordinates when we do a rotation through angle is
x1 = y1 cos y2 sin

x2 = y1 sin + y2 cos

20

4.2. THE THEORY

Example 4.3. Rotation through

4.

x1 =
x2 =

1
(y1 y2 )
2
1
(y1 + y2 )
2

Example 4.4. What is the following conic (find the rotation and translation to change
coordinates and get standard equation)
x12 + x22 + 4x1 x2 = 1
(From the hat method) Lets rotate through
equation becomes

4.

By magic (see later). Then our

1
1
1
(y1 y2 )2 + (y1 + y2 )2 + 4 (y1 y2 )(y1 + y2 ) = 1 3y12 y22 = 1
2
2
2
This is a hyperbola.

4.2

The Theory

Start with general conic

ax12 + bx22 + cx1 x2 + dx1 + ex2 + f = 0
Aim
(i) find the rotation which gets rid of x1 x2
(ii) complete the square to find the translation which changes the equation to one
of our six standard equations
Part(i).
If c = 0, we dont need to rotate. So assume that c 6= 0. When we do general rotation
through we change coordinates to y1 , y2
x1 = y1 cos y2 sin
x2 = y1 sin + y2 cos

We aim to find so that the new equation has no y1 y2 term.

ax12 = a(y1 cos y2 sin )2

bx22 = b(y1 sin + y2 cos )2

cx1 x2 = c(y1 sin + y2 cos )(y1 cos y2 sin )
So the y1 y2 term when we change coordinates will be
2a sin cos + 2b sin cos + c(cos2 sin2 )
=

(b a) sin 2 + c cos 2

4.2. THE THEORY

21

So we want to choose to make this expression zero.

(a b) sin 2 = c cos 2
If a 6= b then
tan 2 =

c
ab

If a = b, we want to cos 2 = 0, so take = 4 .

Summary

Step 1 Rotation. If c 6= 0 in (4.1), we rotate through , where

=
4
when a = b or
c
tan 2 =
ab
when a 6= b.
Step 2 Translation. After Step 1, equation becomes

a y12 + b y22 + d y1 + e y2 + f = 0

Now we complete the square to find a translation which changes equation to one
of the standard ones.
Weve proved:
Theorem 4.1. Every conic in the form (4.1) can be changed by rotation and translation
of the axes to one of the standard equations (1) (8). Thus every conic is either an
ellipse, hyperbola, parabola, or one of the degenerate conics.
Example 4.5. Reduce conic





2x12 + 2 3x1 x2 + 3 3 1 x1 + 3 + 3x2 = 0

to standartd form by rotation and translation.

Step 1 Rotation Here a = 2, b = 0, c = 2 3. So rotate through , where

tan 2 = 3
Therefore 2 = 3 , so =

6.

So, our new coordinates are

x1 = y1 cos y2 sin

1
3
=
y1 y2
2
2
x2 = y1 sin + y2 cos

1
3
y1 +
y2
=
2
2
So, the equation becomes

2

3y1 y2 + 2 3 14
3y1 y2 (y1 + 3y2 )
2 41

+(3 3 1) 21 ( 3y1 y2 ) + (3 + 3) 21 (y1 + 3y2 ) = 0

Which is
3y12 y22 + 6y1 + 2y2 = 0

22

4.3. GEOMETRICAL DEFINITION OF CONICS

Step 2 Translation complete the square
3(y1 + 1)2 (y2 y1 )2 = 2
So put z1 = y1 + 1 , z2 = y2 1, and equation is now standard equation
3z12 z22 = 2
This is a hyperbola. Sketch
new originhas
y1 , y2 coordinates (1, 1), so has

1
x1 x2 coordinates 2 ( 3 + 1), 21 (1 + 3)

Note. Usually tan 2 =

use formula

c
ab

is not so convinient (cant write what 2 is). In general,

tan 2 =

to work out tan , hence sin and cos .

2 tan
1 tan2

Note. Standard parabola is x2 = ax12 + b. What about parabola x1 = ax22 + b?

We can rotate this through 2 to the standard equation.

4.3

Geometrical definition of conics

Ingredients of this definition

a line L
a point p not on L
a real number e > 0
Definition 4.2. Curve C is a set of all points x R2 , such that
kx pk = e dist(x , L)
i.e.



C = x R2 | kx pk = e dist(p, L)

Example 4.6. Let e = 1.

Let e = 21

Theorem 4.2. Curve C (from previous definition) is a conic. It is

a parabola if e = 1
an ellipse if e < 1
a hyperbola if e > 1
Proof. Do a rotation and translation to make p the origin and L vertical line x1 = s.
Then the equation defininig C is
kx pk = e dist(x , L)

4.3. GEOMETRICAL DEFINITION OF CONICS

23

i.e
kx k = e |x1 s|

i.e

kx k2 = e 2 (x1 s)2

i.e

x12 + x22 = e 2 (x12 2sx1 + s 2 )

i.e

x12 (1 e 2 ) + x22 + 2se 2 x1 s 2 e 2 = 0

If e = 1, the x12 term vanishes this is a parabola.

Now suppose e 6= 1.
Complete square

2
se 2
s 2e 4
(1 e 2 ) x1 +
=0
+ x22 s 2 e 2
2
1e
1 e2
So put y1 = x1 +

i.e

se 2
1e 2

and y2 = x2 . Then get standard equation



e2
(1 e 2 )y12 + y22 = s 2 e 2 1 +
1 e2

y22
s 2e 2
=
1 e2
(1 e 2 )2
If e < 1, this is an ellipse. If e > 1, this is hyperbola.
y12 +

(4.2)


Definition 4.3. The conic has focus p, directric L, excentricity e from the previous
proof.
Example 4.7. Find e, p and L for the ellipse
x12
+ x22 = 1
2
This is the standard equation. We compare it with (4.2)
s 2e 2
y22
=
(4.3)
1 e2
(1 e 2 )2
 2 
se 2
se
From the y coordinate picture, the focus is 1e
2 , 0 , directrix is x1 = s + 1e 2 .
Compare equations
y12 +

y22
s 2e 2
=
1 e2
(1 e 2 )2
2
2
x1 + 2x2 = 2

y12 +

So,

So e =

1 ,
2

1
1 e2
s 2e 2
(1 e 2 )2
s = 1. So the excentricity is

1 ,
2

= 2
= 2
focus (1, 0) and the directrix is x1 = 2.

Note. Ellipse has in fact two foci p, and two directrices L, L .

24

4.3. GEOMETRICAL DEFINITION OF CONICS

25

Chapter 5

Matrices and linear equations

Definition 5.1. The space Rn define
R1 = R
R2 = set of ordered pairs
R3 = set of all triples
R4 = set of all quadruples
In general, Rn is the set of all n-tuples (x1 , x2 , . . . , xn ) with x1 R. Call these n-tuples
vectors.
Note that Ri 6 Ri+1 .
The Rn has interesting structure
geometric points, lines, curves
algebraic
Definition 5.2. Addition of vectors
(x1 , x2 , . . . , xn ) + (y1 , y2 , . . . , yn ) = (x1 + y1 , . . . , xn + yn )
Scalar multiplication
(x1 , . . . , xn ) = (x1 , . . . , xn )
Definition 5.3. A linear equation in x1 , x2 , . . . xn is
a1 x 1 + a2 x 2 + + an x n = b
where coefficients ai , b R.
Definition 5.4. A solution to a linear equation is a vector (k1 , . . . , kn ) such that the
equation is satisfied when we put xi = ki .
Example 5.1. (1, 2, 3) is a solution to the linear equation x1 + x2 + x3 = 2.
Definition 5.5. A system of linear equations in x1 , . . . xn is a collection of one or more
linear equations in these variables.
Example 5.2.

26
1)
3x1 + 4x2 = 5
8x1 x2 = 2
is a system of 2 linear equations in x1 , x2 .
2)
x1 + x2 + x3 + x4 = 0
2x1 x2 + 5x4 = 2
x1 + x2 x4 = 3

is a system of 3 equations in x1 , x2 , x3 , x4 .
Definition 5.6. General system
a11 x1 + a12 x2 + + a1n xn

a21 x1 + a22 x2 + + a2n xn

am1 x1 + am2 x2 + + amn xn

b1

b2

bm

where x1 , . . . , xn are unknowns and aij , bbi are constants. A solution to this system is
a vector (k1 , . . . , kn ) which satisfies all the equations.
Aim is to find the method to find all solutions of any system of the form from previous
definition.
Example 5.3. System
x1 + x1 = 1
2x1 + 3x2 = 5
Eliminate x1 take twice first, add to second
5x2 = 7
One solution (x1 , x2 ) =

7
( 2
5 , 5 ).

Example 5.4.
2x1 + x2 = 2
6x1 3x2 = 1
These are just two parallel (nonidentical) lines the system has no solution.
Example 5.5. System
x1 + x2 = 0
x2 x3 = 0

x1 x2 + 2x3 = 0
Notice (cleverly), that third equation is equal to first minus two times second. So any
solution of system of first and second equation will automatically satisfy third.
So general solution is
(x1 , x2 , x3 ) = (a, a, a)
for any a R. This system has infinitely many solutions. Soon we will see that every
system has either no solution, one solution, or unlimited number of solution.

5.1. METHOD GAUSSIAN ELIMINATION

5.1

27

Method Gaussian elimination

Example 5.6. System

x1 + 2x2 + 3x3 = 9
4x1 + 5x2 + 6x3 = 24
3x1 + x2 2x3 = 4
Step 1 Eliminate x1 from second and third, using first. We get equivalent system
x1 + 2x2 + 3x3 = 9
3x2 6x3 = 12

5x2 11x2 = 23

(2) 4(1)

(3) 3(1)

This system has the same solutions as the original one as new equations are combinations of orignal ones, and vice versa.
Step 2 Eliminate x2 from the third equation using only second equation.
x1 + 2x2 + 3x3 = 9
3x2 6x3 = 12
x3 = 3

5
(3) (2)
3

Step 3 Solve!
By third, x3 = 3. By the second, 3x2 = 12 18, so x2 = 2. By the first, x1 = 4. So
the system has one solution (4, 2, 3).
For bigger systems, we need better notation. This is provided by matrices.
Definition 5.7. A matrix is a rectangular array of numbers. Eg


1 2 3
4 5 6
This is 2 3 matrix.

1
5

This is 3 1 matrix.
Call matrix m n if it has m rows, n collumns. We use matrices to encapsulate systems
of linear equations. System from previous example has the coefficients matrix

1 2
3
4 5
6
3 1 2
and augmented matrix

1 2
3 9
4 5
6 24
3 1 2 4

28

5.1. METHOD GAUSSIAN ELIMINATION

For general system

coeff matrix =

augmented matrix =

a11
am1
a11
am1

a12 . . .
...
am2 . . .
a12 . . .
...
am2 . . .

a1n
amn

a1n

b1

amn bm

A row of augmented matrix corresponds to an equation in the system.

The number in the i-th row and j-th collumn of a matrix is called the ij-entry.
Operation on the equations in a system are operations
matrix.

1 2
3 9
1
2
4 5
6 24 0 3
3 1 2 4
0 5

1
2
0 3
0
0

on the rows of the augmented

3
9
6 12
11 23

3
9
6 12
1 3

Definition 5.8. Elementary row operations are the following operations on the rows
of an (augmented) matrix
1) Add a scalar multiple of one row to another
r1 ri + rj , R
2) Swap two rows
ri rj
3) Multiply any row by a nonzero scalar
ri ri , 6= 0
Doing these elementary row operations to an augmented matrix does not change the
solution of the system.
Idea of Gaussian elimination
We start with a hard system. Then we do a couple of row operations and get an
easy system. What makes it easy are zeros under the main diagonal.
Definition 5.9. An m n matrix is in echelon form if
(i) The first non-zero number in each row occurs to the right of the first non-zero
number in any higher row
(ii) All rows (0, 0, . . . , 0) appear at the bottom.

5.1. METHOD GAUSSIAN ELIMINATION

29

Example 5.7. The following matrices

1
2
3
9
0 3 6 12
0
0 3 9


0 0 1
0 0 0

are in echelon form. This one is not

0 1 3
1 0 0
0 0 0
The point If a system has its augmented matrix in echelon form, the system is easy
to solve.
Example 5.8. Solve the system of equations with augmentated matrix

2 1 3 0
0
1 1 1
0
0 0 0
This system is
2x1 x2 + 3x3 = 0
x2 + x3 = 1

Solve from bottom up. Equation 3 tells us nothing. Let x3 = a (any a R. Then eq
2 implies that x2 is 1 a.
Equation 1 implies 2x1 = x2 3x2 , therefore x1 = 14a
2 .
1
(1

4a),
1 a, a). E.g when a = 0, we
Therefore the solutions
are
(x
,
x
,
x
)
=
(
1
2
3
2


get solution 21 , 1, 0 .

Example 5.9. Solve the system with augmentated matrix

2 1 3 0
0
1 1 0
0
0 0 2
System thus is
2x1 x2 + 3x3 = 0
x2 + x3 = 1

0 = 2
Third equation instantly implies no solution at all.

30

5.1. METHOD GAUSSIAN ELIMINATION

Example 5.10.

1 1 1 3
2 4 0
0 0 0 2 2 0 3
0 0 0 0
1 1 0

The system is

x1 + x2 + x3 + 3x4 + 2x5 + 4x6 = 0

2x4 x5 = 3
x5 + x6 = 0

Equation 3 sets x6 = a (any a). Then x5 = a. From 2 we get x4 = 21 (x5 + 3) =

1
2
2 (3 a). From 1 we get x3 = b and x2 = c. Then x1 = c b 3 (3 a) + 2a 4a
1
9
= 2 2 a b c. The general solution is
(x1 , x2 , x3 , x4 , x5 , x6 )
=
( 92

1
2a

b c, c, b, 12 (3 a), a, a)

for any a, b, c R.
In general, if augmentated matrix is in echelon form, solve the system by solving the
last equation, then the next last, and so on.
This method is called back substitution.
The variables we can put free equal to a, d, c etc are free variables. E.g. in previous
example, the free variables are x6 , x3 , x2 .
Theorem 5.1. (Gausian elimination theorem)
Any matrix can be reduced by elementary row operations to a matrix which is in echelon
form.
Method 5.1. (Gausian elimination method)
System of linear equations (augmented matrix). We put the augmented to echelon
form using elementary row operations. Solve the new system by back substitution.
Example 5.11. Solve the system with augmentation matrix

1
0 3
2 5 1
1
0 3
5 8 1

A=
1
1 4
5 7 0
1 1 2 1 3 3
Answer.

Step 1 clear first collumn using top left hand entry (i.e. equations 2, 3, 4 take
away equation 1)

1
0
3
2
5
1
0
0
0
0
3
3

A
0
1
1
3
2 1
0 1 1 3 2
1

5.1. METHOD GAUSSIAN ELIMINATION

Step 2 swap rows 2 and 4

1
0
3
2
5
1
0 1 1 3 2
1

0
1
1
3
2 1
0
0
0
3
3
0
Step 3 clear second collumn using the row 2

1
0
3
2
5 1
0 1 1 3 2 1

0
0
0
0
0 0
0
0
0
3
3 0
Step 4

31

1
0
3
2
5 1
0 1 1 3 2 1

0
0
0
3
3 0
0
0
0
0
0 0

Step 5 solve by back substitution Let x5 = a. Then x4 = a, x3 = b,

x2 = 1 b + a, x1 = 1 3b 3a.

Matrix Algebra
Definition 5.10. Matrix multiplication
Dot product in Rn


y1
x1
..
..
x =. , y =.
yn
xn
x .y = x1 y1 + + xn y n
It is convenient given a row vector z = (za , . . . , zn ) to also define z .y = z1 y1 + +zn yn .
Then one can define matrix multiplication by

a1

A = . . . , B = b1 . . . bn
an

c11 . . . c1p
...
AB = . . .
cm1 . . . cmp

where cij = ai bj

b1
..
Note. If A is 1 n (a1 . . . an ) = a, B is n 1 . = b, AB is a 1 1 matrix (a.b)
bn

32

5.1. METHOD GAUSSIAN ELIMINATION

Key properties of matrix multiplication

1) One can multiply an m n matrix A with an r s matrix B iff r = n. Then, AB
will be an m s matrix.
2) Commutativity fails, i.e. in general AB 6= BA.
In order for both AB and BA to be defined, they must be square matrices of
same size, i.e. n n for some n.
Example 5.12.
A=
Then




1 2
0 1
, B=
3 4
1 2



2 5
AB =
4 11


3 4
BA =
5 6

3) Matrix multiplication, like for on R, is associative.

Theorem 5.2. Given any matrices A m n, B n p, C p q
(AB)C = A(BC)
Example 5.13.
A=
Then






1 2
0 1
1 2
, B=
, C=
3 4
1 2
3 0




2 5
AB =
(AB)C =
4 11



3
0
BC =
, A(BC) =
5 2

13 4
29 8

13 4
29 8

Proof. First, we do special case, when C is p 1 matrix, i.e. column vector. Let
A is m n, B is n p, x is p 1. We show that
(AB)x = A(Bx )
Let y = Bx be a vector and z = Ay . So z is the right hand side. Claim is
z = (AB)x .
y
=
B
x
(n 1)
(n p) (p 1)


x1
y1
b11 . . . b1p
.. ..
.. ..
y =.= .
. .
xp
yn
bn1 . . . bnp

5.1. METHOD GAUSSIAN ELIMINATION

a11 . . .
z1
.. ..
z = . = .
bm1 . . .
zm

33

y1
a1n
.. ..
. .

bmn

yn

For i i n zi = ai1 y1 + + ain yn , yj = bji x1 + + bjp xp .

Then
zi
=
ai1 (b1i x1 + + b1p ) + ai2 (b2i x1 + + b2p )+
...

+ain (bni x1 + + bnp )

=
(ai1 b11 + ai2 b21 + + ain bn1 )x1 + (ai2 b12 + ai2 b22 + + ain bn2 )x2
+ + (ai2 b1p + ai2 b2p + + ain bnp )xp

Thus coefficient of xi in zi is
(ai1 bij + ai2 b2j + + ain bnj )
which is ai .bj , where

ai1
b1j
..
..
ai = . , bj = .
ain
bnj

which by definition is the (i, j) entry of AB!

Claim
(AB)C = A(BC)

..
..
.
.

cq
Write C = c1
Then
..
..
.
.
BC = Bc1 . . .

Bcq

Therefore
A(BC) = A(Bc1 ) . . .

A(Bcq )

4) Again, as for + and on R, distributivity holds

Proposition 5.3. Given matrices A m n, B, C, both n p
A(B + C) = AB + AC

34

5.1. METHOD GAUSSIAN ELIMINATION

Powers of matrices
Definition 5.11. A square matrix is one which is n n for some n.
If A is n n, define
A2

AA

(AA)A = A(AA)

A3 A

...
n

An1 A

A
Example 5.14.


1 2
0 1

2


 

1 2
1 2
1 4
=
0 2
0 2
0 1

Definition 5.12. The identity matrix is

2 2 I2
3 3 I3

n n In

1
=
0

0
=
0

1
0

=
0
0


0
1

0 0
1 0
0 1

0 ...
1 ...
.
0 ..
0

Proposition 5.4. If A is m n and B is n p then

AIn = A
In B = B
Link between addition and multiplication
Proposition 5.5.
(i) If A is m n and B, C are n p, then
A(B + C) = AB + AC
(ii) If D, E are m n and F is n p, then
(D + E)F = DF + EF
Proof.
(i) Let A = (aij ), B = (bij ), C = (cij ). The ij-entry of AB is

bij
..
(aj1 . . . ain ) . = ai1 bij + + ain bnj
bnj

5.1. METHOD GAUSSIAN ELIMINATION

35

So ij-entry of A(B + C) is
ai1 (b1j + c1j ) + + ain (bnj + cnj )
=

ai1 b1j + ai1 c1j + + ain bnj + ain cnj

=

(ai1 b1j + + ain bnj ) + (ai1 c1j + + ain cnj )

This is the ij entry of AB + AC.

(ii) Left to enthusiastic reader.

36

5.1. METHOD GAUSSIAN ELIMINATION

37

Chapter 6

Some applications of matrix

algebra
6.1

Linear equations

We know that a general system of linear equations

a11 x1 + + a1n xn = b1
..
.
am1 x1 + + amn xn = bn
can be expressed as a matrix product
Ax = b


x1
b1


..
..
where A = aij , x = . and b = .
xn
bm

Proposition 6.1. The number of solutions of a system Ax = b is 0, 1 or .

Proof. Suppose the number of solutions is not 0 or 1, i.e. there are at least two
solutions. We prove that this implies that there are infinitely many solutions.
Let p and q be two different solutions (p, q Rn ), so
Ap = b
Aq = b
and p 6= q. Then
A(p q) = Ap Aq = 0
For any scalar R
A(p q) = A(p q) = 0
So
A(p + (p q)) = Ap + A(p q) = Ap + 0 = b
So p + (p q) is a solution of the original system.
Since p 6= q, p q 6= 0 and so each different scalar gives a different solution. So
there are solutions.


38

6.2. POPULATION DISTRIBUTION

Structure of solutions
Proposition 6.2.
(i) System Ax = 0 either has one solution (which must be x = O) or it has an
infinite number of solutions.
(ii) Suppose p is a solution of a system Ax = b (i.e. p Rn , Ap = b). Then all
solutions of Ax = b take the form
p+h
where h is a solution of the system Ax = 0.

Example 6.1. System

 
x
1 0 3 1
2
x2 =
. The general solution is
0 1 1
1
x3


x = (2 3a, 1 a, a)
Particular solution p = (2, 1, 0). So general solution is
(2 3a, 1 a, a) = (2, 1, 0) + (3a, a, a) = p + h


 
1 0 3
0
Where h = (3a, a, a) is the general solution of
x=
.
0 1 1
0
Proof.
(i) clear from previous proposition
(ii) Let q be a solution of A. So Ap = b, Aq = b. So A(q p) = 0.
Put q p = h, a solution to Ax = 0. Then q = p + h.


6.2

Population Distribution

Example 6.2. Population classified in 3 income states

(1) Poor (P)
(2) Middle incom (M)
(3) Rich(R)
Over one generation (20 year period)
P

20%M, 5%R rest stay P

M 25%P, 10%R
R 5%P, 30%M

6.2. POPULATION DISTRIBUTION

39

Summarize this information in a matrix. Its 3 3 matrix T , T = (tij ), where

tij = proportion moving from state j state i
So

0.75 0.25 0.5

T = 0.20 0.65 0.30
0.05 0.10 0.65

This is called the transition matrix. All entries are 0, and column sums are 1.
Say we start with proportions p1 , p2 , p3 in states 1, 2, 3.

p1
p (0) = p2
p3
is the initial population vector.
After 1 generation

proportion in state 1

= t11 p1 + t12 p2 + t13 p3

proportion in state 2

= t21 p1 + t22 p2 + t23 p3

proportion in state 3

= t31 p1 + t32 p2 + t33 p3

So population vector after 1 generation is

p1
p (1) = T p2 = T p (0)
p3

Similarly, after 2 generations

p (2) = T p (1) = T 2 p (0)

Continuiing, see that after n generations, population vector is
p (n) = T n p (0)
This is an example of a Markov chain a population is divided into states 1, . . . , n, and
were given the proportion tij moving from state j to state i over a generation. The
transition matrix T = (tij ) is n n with properties
1) tij 0
2) all column sums are 1
Markov chain is regular if some power of T has no zero entries.
Example 6.3. The above example is regular.


0 1
T =
1 0
is not regular.

40

6.2. POPULATION DISTRIBUTION

Basic Fact
m (0)
In a regular Markov chain,
asn grows, the vector T p gets closer and closer to a
s1
s2

steady state vector s = .. where
.
sn

1) s1 + + sn = 1
2) T s = s
This is true, whatever the initial population vector p (0) . Proof is not hard.
Example 6.4. On a desert island, a veg-prone community dies according to
(1) no-one eats meat 2 days in a row
(2) if someone doesnt eat meat one day, they toss a coin: heads eat meat next day,
tails dont
What proportion can be expected to eat meat on a given day?
Answer Markov chain.
State 1: meat
State 2: no meat
generation: 1 day
Transition matrix
T =

Notice

T2

vector s =
So

=


1

2
1
2

1
4
3
4

0
1

1
2
1
2

. So it is regular. By our Basic Fact, we have a steady state

s1
, where s1 + s2 = 1 and T s = s.
s2
1
s2 = s1
2
1
s1 + s2 = s2
2
s1 + s2 = 1

So in long run,

1
3

of population will be eating meat on a given day.

6.3. MATRICES AND GEOMETRY

6.3
6.3.1

41

Matrices and Geometry

Rotation

Consider a rotation about the origin through angle . Then

y1 = r cos( + )
= r (cos cos sin sin )

= x1 cos x2 sin

y2 = r sin( + )

= r (sin cos + cos sin )

= x1 sin + x2 cos
So

  
 
cos sin
y1
x1
=
y2
sin cos
x2


cos sin
Call R =
, a rotation matrix.
sin cos

6.3.2

Reflection

Let s be the reflection in the x1 axis sending

  
 
 
1
0
x1
x1
x1
=

x=
x2
x2
0 1
x2


1
0
So matrix S =
represents the reflection s.
0 1

6.3.3

Combining Transformations

Two rotations, say through then

x
x

r (r (x ))

R (R (x ))
=

(R R )x =

R+

cos sin
sin
cos



cos sin
sin
cos

Reflection+Rotation
Now consider sr , sending x s(r (x )). This sends
x

S(R x )
=

=
=

(SR )x



1
0
cos sin
x
0 1
sin cos


cos sin
x
sin cos

42

6.3. MATRICES AND GEOMETRY

43

Chapter 7

Inverses
Definition 7.1. Let A be a square matrix (n n). Say another n n matrix B is an
inverse of A iff AB = BA = In .
Inverse of A is denoted A1 . If A has an inverse, A is invertible.


1 1
Example 7.1. A =
. Then
0 1







1 1
1 0
=
0
1
0 1





1 1
1 1
1 0
=
0
1
0 1
0 1

Example 7.2.

1 1
0 1


 

1 2
a b
a + 2c b + 2d
=
0 0
c d
0
0

which cannot equal to I.

Proposition 7.1. If A is invertible then its inverse is unique.
Proof. Suppose B, C are inverses of A. Then
AB = BA = I
AC = CA = I
B = BI = B(AC) = (BA)C = IC = C


7.1

Relation to linear equations

Proposition 7.2. Suppose A is invertible. Then any system Ax = b has a unique

solution
x = A1 b

44

7.2. FINDING INVERSES

Proof.
Ax
1

A1 b

(A1 A)x

A1 b

A1 b

(Ax )

Example 7.3. System

x1 + x2 = 2
x2 = 3
is

By previous proposition
x=

7.2


 
1 1
2
x=
0 1
3

1 1
0
1

   
2
1
=
3
3

Finding inverses

2 2 matrices
Let A =


a b
. Observe
c d


 
a b
d b
ad bc
=
c d
c
a
0

0
ad bc

d
c

= (ad bc)I

Using this, we can prove the following

Proposition 7.3.
1) If ad bc 6= 0 then A is invertible and
1

1
=
ad bc

b
a

2) if ad bc = 0 then A is not invertible.

Proof.
2) Suppose ad bc = 0. Then AB = 0 (B =


d b
). Assume A is invertible.
c a

A1 (AB) = A1 0 = 0
A1 (AB) = (A1 A)B = IB
Therefore IB = 0 and thus A = B = 0. But zero matrix is not invertible.


7.3. FINDING INVERSES IN GENERAL

7.3

45

Finding inverses in general

Let A = (aij ) is n n matrix. We want n n matrix X = xij such that AX = I. To

solve, we write one giant matrix

a1n 1 . . . 0
.. .. . .
.
. .. = (A|I)
. .
ann 0 . . . 1

a11 . . .

..

.
an1 . . .

Now use Gaussian elimination to reduce to Gaussian form. There are two possibilities

1) There is a row with zeros in the left side and non zero elements in the right
side. There is no inverse in this case.

2) We dont get into the first situation. Therefore we can put the left matrix into
Echelon form. Then, we can get rid of non-zero elements above the main diagonal
of the left matrix and get a matrix

(I|E)

So AE = I.
Start with (E|I) and reverse everythung, we end up with (I|A). So solving EX = I
gives X = A, completing proof that E = A1 .

Proposition 7.4.
1) If we can reduce (A|I) to (I|E) using elementary row operations, then E = A1 .
2) If we can reduce (A|I) to matrix with zeros in the left side of the last line and a
non-zero element in the right side of the last line, then A is not invertible.

Example 7.4. Find inverse of

1 3 2
2 5 3
3 2 4

46

7.3. FINDING INVERSES IN GENERAL

Augmented matrix is

1 3 2 1 0 0
2 5 3 0 1 0
3 2 4 0 0 1

1 0 0
1
3 2
0 1
1 2 1 0
3 0 1
0 11 10

1 0 0
1
3 2
0 1
1 2 1 0
0
0
1 19 11 1

1
0 0
1 3 2
0 1 1
2 11 0
11 1
0 0
1 19

1 3 0 37 22 2
0 1 0 17 10 1
0 0 1 19 11 1

14 8 1
1 0 0
0 1 0 17 10
1
1
0 0 1 19 11
Thus the inverse

14 8 1
17 10
1
19 11
1

A result linking inverses, linear equations and echelon forms

Proposition 7.5. Let A be a square matrix n n. The following four statements are
equivalent
(1) A is invertible
(2) Any system Ax = b has a unique solution
(3) The system Ax = 0 has the unique solution x = 0
(4) A can be reduced to the identity In using ERO.
Proof. We prove (1) (2), (2) (3), (3) (4), (4) (1).
(1) (2) Is done, x = A1 b
(2) (3) Is obvious
(3) (4) Needs to be proved.
(4) (1) Is proved earlier.


47

Chapter 8

Determinants
Recall that matrix

a b
c d

is invertible iff ad bc 6= 0.

Definition 8.1. The determinant of a 2 2 matrix is



a b
det
= ad bc
c d
Also write as



a b 


c d 

or |A|.
Example 8.1.

Recall



1 2


3 4 = 2


 cos sin 
=1

|R | = 
sin cos 

Proposition 8.1. For a 2 2 matrix A

|A| =
6 0 A is invertible Ax = b has unique solution.
Aim is to define det(A) for 3 3 and larger matrices in such a way that this result is
still true.
Definition 8.2. For a 3 3 matrix

a11 a12 a13

A = a21 a22 a23
a31 a32 a33

define the determinant of A to be







a21 a22 
a21 a23 
a22 a23 






+ a13 
a12 
det(A) = a11 
a31 a32 
a31 a33 
a32 a33 
= a11 (a22 a33 a23 a32 ) a12 (a21 a33 a23 a31 ) + a13 (a21 a32 a22 a31 )

48

8.1. PROPERTIES

Example 8.2.







1 2 3









4 5 6 = 1 5 6 2 4 6 + 3 4 5
 1 1
1 2

1 2

1 1 2
= 3
Example 8.3.

8.1



a x y 


0 b z  = abc


0 0 c 

Properties

Definition 8.3. If A is a 3 3 matrix, the ij-minor of A is the 2 2 matrix Aij obtained

by deleting ith row and jth collumn of A.
Then
|A| = a11 |A11 | a12 |A12 | + a13 |A13 |
This is called the expansion of |A| by the first row.
Example 8.4.



1 2 3
5
6
A = 4 5 6 , A11 =
1 2
1 1 2

We dont have to expand by the first row.

Proposition 8.2. Expansion by the second row

|A| = a21 |A21 | + a22 |A22 | a23 |A23 |
third row
|A| = a31 |A31 | a32 |A32 | + a33 |A33 |
Proof. The second row.
RHS = a21 (a12 a33 a13 a32 ) + a22 (a11 a33 a13 a31 ) a23 (a11 a32 a12 a31 )
= |A|

Check third row.

Proposition 8.3. If A (3 3) has two equal rows, then |A| = 0.

Proof. WLOG1 say that

a1 a2 a3
A = a1 a2 a3
c1 c2 c3

Expand by the third row to get |A| = 0.

1

Without Loss Of Generality

8.2. EFFECTS OF ROW OPERATIONS ON DETERMINANT

8.2

49

Effects of row operations on determinant

r1
Proposition 8.4. Let A = r2 be 3 3.
r3

1) Row operation ri ri + rj (i 6= j). does not change |A|.

2) Swapping two rows changes |A| to |A|.

3) Row operation ri ri changes |A| to |A|
Proof.
1) Say i = 2, j = 1, so the row. op. sends

a1 a2 a3
a1
a2
a3
A = b1 b2 b3 A = b1 + a1 b2 + a2 b3 + a3
c1 c2 c3
c1
c2
c3
Expand by the second row

a
|A | = (b1 + a1 )  2
c2

a1 a2

= |A| + a1 a2
c1 c2
= |A|






a1 a2 
a1 a3 
a3 

 (b3 + a3 ) 
+ (b2 + a2 ) 
c1 c2 
c1 c3 
c3 

a3 
a3 
c3 

2) Say we swap rows 1 and 2 to get



b1 b2 b3 






a1 a2 a3  = a1 b2 b3  + . . .


 c2 c3 
c1 c2 c3 

3) It is obvious when we do the expansion by the row we multiply.



Example 8.5.



 1 2 1 
 1
2



 2 3
 =  0
4
7



 5 9 4 
 0 1

 7 2
= 
1 1
= 9

1
2
1



+0+0


50

8.2. EFFECTS OF ROW OPERATIONS ON DETERMINANT

Example 8.6.

1 x

1 y

1 z


x 2 
y 2 x 2 
z 2 x 2


2 
1 x
x


= (y x )(z x ) 0 1 y + x 
0 1 z + x 


1 x
x 2 

= (y x )(z x ) 0 1 y + x 
0 0 z y 
= (y x )(z x )(z y )



1
x
x 2 

2


y  = 0 y x
0 z x
z 2

This is the 3 3 Vandermande determinant.

Proposition 8.5. Let A be 3 3 matrix, and let A be obtained from A by el. row.
ops. Then
|A| = 0 |A | = 0
Proof. Doing a row op. changes |A| to |A|, -|A| or |A| (where 6= 0).

Main result
Theorem 8.6. Let A be 3 3 matrix. Then
|A| 6= 0 A is invertible
(or the system Ax = 0 has unique solution, or A I3 by row ops.)
Proof.
Suppose |A| 6= 0. Reduce A to echelon form A by row
operations.
By 8.5,
1
|A | =
6 0. If A has a zero row, |A| = 0. Hence A = 0 1 so can be
0 0 1
reduced to I. So A is invertible.
Suppose A is invertible. A can be reduced to I by row ops. Since |I| = 1, so
|A| =
6 0.


Example 8.7. For which values of a is A invertible?

1 2 5
A = 1 3 7
1 4 a



1 2

5



2  = a 9
|A| = 0 1
0 2 a 5

So by 8.6, A is invertible iff a 6= 9.

8.3. N N DETERMINANTS

51

Important consequence
Corollary 8.7. Let A be 3 3 matrix. Suppose the system Ax = 0 has a non-zero
solution x 6= O. Then |A| = 0.

1 2 5
Example 8.8. Let A = 1 3 7. System Ax = 0 has augmented matrix
1 4 a

1 2 5 0
1 2
5
1 3 7 0 0 1
2
1 4 a 0
0 2 a5

1 2
5

0
1
2

0 0 a9

0
0
0

0
0
0

So if a 6= 9, only solution is x = 0. If a = 9, system has solutions. We saw that

|A| = 0 a = 9.

8.3

n n determinants

It is possible to define det(A) for any n n matrix A, and to prove that all the results
of this section are true for n n matrices. Wait for proofs until next year.
Definition 8.4. 4 4 determinant. If A = (aij ) is 4 4,
|A| = a11 |A11 | a12 |A12 | + a13 |A13 | a14 |A14 |
Similarly, define n n determinant in terms of (n 1) (n 1) determinants (recursive
definition).
Example 8.9.

 1
2

 1
0

 2
1

 0 1

3
1
0
1

4
5
2
3







 0 1 5 
 1 1 5




 =  1 0 2  2 2 0 2





 1 1 3 
 0 1 3





 + 3 + ...



52

8.3. N N DETERMINANTS

53

Chapter 9

Eigenvalues and eigenvectors

Fibonaccis Rabbits
Rabbits, born in pairs. Newly born pair produce no offspring in the firs month, but then
one pair in each following month. No deaths. Start with one pair. How many pairs
after n months?
Answer. Let Fn = number of pairs. Then
Fn = Fn1 + Fn2
So F0 , F1 , F2 , . . . = 0, 1, 1, 2, 3, 5, 8, 13, . . . is the Fibonacci sequence. In matrix form
 



1 1
Fn
Fn1
=
Fn1
1 0
Fn2
   
 
 
1
F1
F
F2
=
Now
= A 1 and so on.
, so
F0
0
F0
F1


 
Fn+1
n F1
=A
Fn
F0

9.1

Eigenvectors

v1
..
Definition 9.1. Let A be a n n matrix. Vector v = . R is an eigenvector of
vn
A if
(1) v 6= 0
(2) Av = v ( R)
The scalar is an eigenvalue of A.
Example 9.1.






1
1
= 1
2
2




1
3 2
So
is an eigenvector of A =
and 1 is its eigenvalue.
2
2 0
3 2
2 0

1
2

54

9.2. HOW TO FIND EIGENVECTORS AND EIGENVALUES

Example 9.2.

3 2
2 0



1
1

5
2

6=

1
1

 


1
3 2
So
is not an eigenvector of
.
1
2 0

9.2

How to find eigenvectors and eigenvalues

Let A be a n n matrix. Vector x is a non-zero solution of the system

Ax = x
Example 9.3.

3 2
2 0

 
 
x
x1
= 1
x2
x2

We can get
(3 )x1 + 2x2 = 0

2x1 x2 = 0

So the equation is



3 2 
=0

 2

2 3 4 = 0

Eigenvalues are 1 and 4.

For = 1, eigenvectors are non-zero solution of

So eigenvectors are
For = 4

So eigenvectors are

a
2a

= 0
= 0

(a R, a 6= 0)


A + Ix

4 2
x
2 1

1
2
2 4

x =0


2b
(b R,b 6= 0).
b

When does have the following equation a non-zero solution?

Ax x

(A I)x
Precisely when |A I| = 0 (by 8.6).

= 0
= 0

9.2. HOW TO FIND EIGENVECTORS AND EIGENVALUES

55

Proposition 9.1.
(1) If A is a 33 or 22, then the eigenvalues of A are the solutions of |AI| = 0.
(2) If is an eigenvalue, then the corresponding eigenvectors are the non-zero solutions of
(A I)x = 0
Definition 9.2. The equation |A I| = 0 is the characteristic equation of A, and
|A I| is the characteristic polynomial of A.

9.2.1

Back to Fibonacci

We had

Fn+1
Fn

= An

 


1
1 1
where A =
.
0
1 0

Strategy
(1) Find eigenvalues 1 , 2 and eigenvectors v1 and v2 of A. Observe
Av1 = 1 v1
Av2 = 2 v2
Then
A2 v1 = A(Av1 )
= A(1 v1 )
= 1 Av1
= 21 v1
Similarly
An v1 = n1 v1
An v2 = n2 v2
 
1
(2) Express
as a combination of v1 and v2 .
0
 
1
= v1 + v2 (, R)
0
Then
 
1
A
= An (v1 + v2 )
0
= An v1 + An v2
 
1
An
= n1 v1 + n2 v2
0
n

(9.1)

56

9.3. DIAGONALIZATION

Calculations
(1) Characteristic equation is |A I| = 0, i.e.


1 1 
=0

 1


i.e

2 1 = 0

1
1 =
(1 + 5)
2

1
2 =
(1 5)
2
Eigenvectors for 1

So v1 =

1
1


1 1
x =0
1

is an eigenvector.
 
2
Similarily v2 =
.
1
(2) We now find and such that
 
 
 
1
1
2
=
+
0
1
1
1
=
1 2
1
=
2 1
Putting all this into (9.1)


Fn+1
Fn

 
1
= A
0
n
= 1 v1 + n2 v2
  
  
 
 
 n 1
 n 2
1
1
1
1
=
1 5
1 5

1
1
5 2
5 2
n

To get formula
1
Fn =
5

9.3

n 
n !
1+ 5
1 5

2
2

Diagonalization
1

Want to investigate functions of matrices, e.g. An , A n , f (A).

9.3. DIAGONALIZATION

57

Definition 9.3. An n n matrix D is diagonal matrix if

1
O

..
D =

.
O
n
Example 9.4.



1 0
0 2

Its easy to find powers of diagonal matrices.

Proposition 9.2. Let

D =

Then

and

E =

1
..
O
1
O

1 1

DE =
O

Dk =

k1

..

.
n
O
..

.
n n
O

..

kn

Proof. DE given by definition of matrix multiplication.

Take E = D to get D 2 and repeat to get D k .

Aim - To relate an arbitrary square matrix A to a diagonal matrix, and exploit this to
find An , etc.
22
Let A be 2 2 and suppose A has eigenvalues 1 , 2 with eigenvectors v1 , v2 . Assume
1 6= 2 . We get
Av1 = 1 v1
Av2 = 2 v2
Cleverly define 2 2 matrix P (v1 , v2 are collumn vectors)
P = v1 v2

58

9.3. DIAGONALIZATION

Then
= A(v1 v2 )

AP

Av1 Av2

1 v 1 2 v 2



1 0
v1 v2
0 2

So if we write D =


1 0
, weve shown
0 2
AP

= PD

(9.2)

We claim that P is invertible. For if not, then |P | = 0, which means that v1 = v2 ,

which is false as
Av1 = 1 v1 = 1 v2
Av2 = 2 v2
and these are not equal as 1 6= 2 .
So from (9.2)
P 1 AP = P 1 P D = D
Summary
Proposition 9.3. Let A be 2 2 with distinct eigenvalues 1 , 2 , eigenvectors v1 , v2 .
Let


P = v1 v2
Then P is invertible and

AP = D =

1 0
0 2

Note. Also true for 3 3, . . . matrices.



0 1
Example 9.5. Let A =
.
2 3
1) Find P such that P 1 AP is diagonal.
2) Find the formula An .
3) Find a 5th root of A, i. e. find B such that
B5 = A
Answer.

9.3. DIAGONALIZATION

59

1) Characteristic equation of A is


1
|A I| = 
2 3

i. e.



 =0


2 3 + 2 = 0

( 1)( 2) = 0
So eigenvectors are
 
1
=1 a
, (a 6= 0).
1
 
1
=2 b
, (b 6= 0)
2


1 1
Let P =
. Then by Proposition 9.3,
1 2



1 0
P AP = D =
0 2




2 1
1 1
Note that many other P s work, e. g.
. Or if Q =
, then
2 2
2 1


2 0
1
Q AQ =
0 1
1

2) Find An . Know P 1 AP = D. By the 9.2, we know that



1 0
n
D =
0 2n
So


1 0
0 2n

= D n = P 1 AP

n

1
1
= P
. . P 1 AP P 1 AP}
| AP P AP .{z
n

= P 1 An P
So

P 1 An P

= Dn

P P 1 An P P 1 = P D n P 1
An = P D n P 1
So
n

1 1
1 2



1 0
0 2n



2 1
1 1

1 2n
1 2n+1



2 1
1
1

60

9.3. DIAGONALIZATION
3 Find B such that B 5 = A.
Well, if



1
0
C=
0 21/5

Then C 5 = D. So
(P CP 1 )5 = P CP 1 . . . P CP 1
= P C 5 P 1
= P DP 1 = A
So take
!


1
0
1
1
2 1
1
B = P CP =
1
1 2
1 1
0 25
!
1
1
2 25 25 1
=
6
6
2 25 25 1


Note. Usually a matrix has many square roots, fifth roots, etc. Eg, I has infinitely
many.
Note. Similarly can calculate polynomial functions
p(A) = an An + an1 An1 + + a1 A + a0 I

Summary
If a square matrix A has distinct eigenvalues, then it can be diagonalized, i. e. there
exists an invertible P such that P 1 AP is diagonal.

9.3.1

Repeated eigenvalues

If the characteristic polynomial of A has a repeated root , we call a repeated

eigenvalue of A.
Some As with the repeated eigenvalue can be diagonalised and some cant.


1 1
Example 9.6. Let A =
. Then the characteristic polynomial is
0 1


1
1 

= (1 )2
 0
1 

so 1 is a repeated eigenvalue. Claim A cannot be diagonalized (i.e. no invertible P

exists such that P 1 AP is diagonal).
Proof. Assume there exists an invertible P such that


1 0
1
P AP = D =
0 2

9.3. DIAGONALIZATION

61

Then
= PD

AP
Writing

A v1

P =


v2 =

v1 v2

v1 v2 D

1 v 1 2 v 2

Hence Av1 = 1 v1 , Av2 = 2 v2 .

So v1 , v2 are eigenvectors of A.

eigenvectors are a

0 1 0
0 0 0

 
1
, a 6= 0. Hence
0
P =

a b
0 0

But this is not invertible. Contradiction.

Point P must have evectors of A as collumns, but A does not have enough independent evectors to make invertible P .

1
0 0
2 0 . Then the characteristic polynomial is
Example 9.7. Let A = 1
1 1 1


 1
0
0 


1 2
0 


1
1 1 
= (1 )2 (2 )

So 1 is a repeated evalue.
Can A be diagonalized?
Lets try.
Evectors for = 2

0
eigenvectors a 1 , a 6= 0.
1
For = 1

1
0
0 0
1
0
0 0
1 1 1 0

0
0 0 0
1
1 0 0
1 1 0 0

62

9.3. DIAGONALIZATION

eigenvectors b .
c
So for collumns of P choose eigenvectors.

0
1
0
1 , 1 , 0
1
0
1
So take

0 1 0
P = 1 1 0
1 0 1

Then |P | = 1, so P is invertible, and from 9.2

2 0 0
P 1 AP = 0 1 0
0 0 1

Summary If A has enough independent vectors for the repeated eigenvalue, A can be
diagonalized. If not, it cant.

63

Chapter 10

Conics(again) and Quadrics

Recall equation of conic
ax12 + bx1 x2 + cx22 + dx1 + ex2 + f = 0
Well write this in matrix form
 


a b
x1
2
x1 x2
=
b
x2
2 c

ax1 +

b
2 x2

So in matrix form, the equation of conic is

 
x
where x = 1 , x T = x1
x2

b
2 x1

 

x1
= ax12 + bx1 x2 + cx22
+ cx2
x2

x T Ax + d e x + f = 0




a b2
x2 , A = b
2 c

Digression Transposes
Definition 10.1. If A = (aij ) is mn, the transpose of A is the mn matrix AT = (aji ).



1 2
1 3 5
Example 10.1. A = 3 4, AT =
2 4 6
5 6
 


x
x = 1 , x T = x1 x2
x2
Note.

(AT )T
x .y

= A
=

x1

 

y1
x2
= xT y
y2

Definition 10.2. A square matrix A is symmetric if A = AT .



a b2
Example 10.2. b
is symmetric.
2 c
Proposition 10.1. Let A be m n, B be n p. Then
(AB)T = B T AT

64
Proof. The ij-th entry of (AB)T is the ji-th entry of AB, which is the j-th row of A
multiplied by i-th column of B. And ij-th entry of B T AT is i-th row of B T multiplied
by j-th collumn of AT , i.e. i-th column of B multiplied by j-th row of A. These are
equal.


Back to conics
We formulatied the equation of a conic as
x T Ax + d

b
2 .
c

 

x1
a
where x =
,A= b
x2
2
Notice

xT

1 0
0 2

e x +f =0

(10.1)

= 1 x12 + 2 x22

 
y
So, aim is to find a rotation matrix P , such that the change of coordinates to y = 1 ,
y2
where x = P y changes the equation (10.1) to


(P y )T A(P y ) + d e P y + f = 0


y T P T AP y + d e y + f = 0
and P T AP is diagonal.
We can do this.

Theorem 10.2. Let A =

a
b
2

b
2

be symetrix, with b 6= 0. Then

1) A has two real eigenvalues 1 , 2 , with 1 6= 2 .

2) If v1 , v2 are eigenvectors of A corresponding to 1 and 2 , then
v1 .v2 = 0
3) We can choose unit eigenvectors v1 , v2 , such that


P = v1 v2

has determinant 1 and P is a rotation matrix. Moreover, P 1 = P T , and P T AP

is diagonal.

Proof.
1) Characteristic polynomial of A is


b 
a
b2
2  = 2 (a + c) + ac

b

c 
4
2

Roots are

 1
p
p
1
a + c (a + c)2 4(ac + b2 ) =
a + c (a c)2 + b2
2
2

10.1. REDUCTION OF CONICS

65

Since b 6= 0, (a c)2 + b2 > 0, so the roots are real and distinct. Call them 1
and 2 .
2) Let
Av1 = 1 v1
Av2 = 2 v2
Consider
v1T Av2
(a) This is v1T (2 v2 ) = 2 v1T v2 .
(b) It is also (because A is simetric, i.e. A = AT )
(AT v1 )T v2 = (Av1 )T v2 = 1 v1T v2
So
2 v1T v2 = 1 v1T v2
As 1 6= 2 , this forces v1T v2 = 0, i.e. v1 .v2 = 0.
3) Choose unit v1 , v2 (see picture). Then
v1 =
v2 =




cos
sin

sin
cos

So
P

v1 v2


cos sin
=
sin
cos
= R

Finally, by 9.3 ,
T

P AP = P

AP =

1 0
0 2




10.1

Reduction of conics

Start with equation (10.1).

1) Find the evalues and evectors of A


2) Find unit eigenvectors v1 , v2 , such that P = v1 v2 has determinant 1, so is a
rotation matrix.

66

10.1. REDUCTION OF CONICS

 
y
3) Change coordinates to y = 1 when x = P y , so equation (10.1) becomes
y2


y T (P T AP )y + d e y + f = 0

i.e.
y

i.e.




y1 0
y + d e y + f = 0
0 y2

1 y12 + 2 y22 + d y1 + e y2 + f = 0
This is an ellipse if 1 2 > 0, hyperbola if 1 2 < 0, parabola if 1 2 = 0 (or
possible degenerate cases).
Example 10.3. Reduce conic
5x12 + 4x1 x2 + 2x22 = 1
to standard form. This is


5 2
x
x
2 2
T

= 1

1) Characteristic polynomial


5
2 

= 2 7 + 6 = ( 1)( + 6)
 2
2 



1
For = 1, eigenvectors are a
.
2
 
2
For = 6, eigenvectors are b
.
1
2) Unit vectors are

15

Take

 

1
2
1

, 5
.
1
1



1
1 2
P =
5 2 1


cos sin
Then |P | = 1, and P = R =
, where cos =
sin cos
3) Let x = P y (i.e. x1 =

1 (y1 +2y2 ),
5

x2 =

1 (2y1 +y2 ).Then

5

y12 + 6y22 = 1
which is an ellipse.

1 ,
5

sin =

.
5

equation becomes

10.2. QUADRIC SURFACES

10.1.1

67

Orthogonal matrices

Recall, rotation matrix satisfies

RT = R
So does a reflection matrix


cos
sin
sin cos

Definition 10.3. A square matrix P is an orthogonal matrix if P T = P 1 , i. e.

P P T = I.
The key property:
Proposition 10.3. Orthogonal matrices preserve lengths, i. e.
kP v k = kv k
Proof.
kP v k2 = (P v ).(P v )

= (P v )T (P v )
= vT PT Pv
= v T Iv
= kv k2


10.2

Quadric surfaces

Definition 10.4. A quadric surface is a surface in R3 defined by a quadratic equation

ax12 + bx22 + cx32 + dx1 x2 + ex1 x3 + f x2 x3 + gx1 + hx2 + jx3 + k = 0

Standard examples
1) Ellipsoid
x12 x22 x32
+ 2 + 2 =1
a2
b
c
2) Hyperboloid of 1 sheet
x12 x22 x32
+ 2 2 =1
a2
b
c
3) Hyperboloid of 2 sheets
x12 x22 x32
+ 2 2 = 1
a2
b
c
4) Elliptic cone
x12 x22 x32
+ 2 2 =0
a2
b
c

(10.2)

68

10.2. QUADRIC SURFACES

5) Elliptic parabloid
x3 =

x12 x22
+ 2
a2
b

x3 =

x12 x22
2
a2
b

6) Hyperbolic parabloid

Some degenerate cases

7) Elliptic cylinder
x12 x22
+ 2 =1
a2
b
8) Parabolic cylinder
x12 ax2 = 0
Aim is to find a rotation and translation, which reduces (10.2) to one of the standard
examples. Here is the procedure.
1. Write (10.2) in matrix form

where

x T Ax + g h j x + k

x1

x2
x =
x3

a d2
A = 22 b
e
2

f
2

= 0

e
2
f
2

Notice that A is symmetric.

2. Find the eigenvalues 1 , 2 , 3 of A (theory these are real). Find corresponding
unit eigenvectors v1 , v2 , v3 which are perpendicular to each other (theory this
can be done). Choose a directions for the vi so that the matrix
P

v1 v2 v3

has determinant 1.
Then, P is a rotation matrix, and P 1 = P T . Then
P 1 AP

=
P T AP

1 0 0
= 0 2 0
0 0 3

10.2. QUADRIC SURFACES

69

y1

3. Change coordinates to y = y2 , where x = P y . Then (??) reduces to

y3


y T (P T AP )y + g h j P y + k = 0

i. e.

1 y12 + 2 y22 + 3 y32 + g y1 + h y2 + j y3 + k = 0

Finally, complete the square to find translation reducing to a standard equation.
Note. All the assumed bits of theory will be covered in Algebra II next year.
Example 10.4. Reduce the quadric
2x1 x2 + 2x1 x3 x2 1 = 0
to standard form.
Answer.
1. Equation is

0 1 1
x T 1 0 0 x
1 0 0

= x2 + 1

2. Characteristic polynomial is



 1
1


 1 0  = (2 2)


 1
0 
We get equal to 0,
Eigenvectors.

2, 2.

0
0
For = 0 a 1 , unit eigenvectors 12 1
1
01
1

2
2

For = 2 b 1 , unit eigenvectors 12

1
1
2

1
2
2

For = 2 b 1 , unit eigenvectors 12

1
1
2

Let

0
1
P =
2
1
2

1
2
1
2
1
2

12
1
2
1
2

70

10.3. LINEAR GEOMETRY IN 3 DIMENSIONS

Then |P | = 1 and P is a rotation.
Change of coordinates x = P y changes equation to

2y22

1
1
1
2y32 = y1 + y2 + y3 + 1
2
2
2

Finally, complete the square

1
2(y2 )2 2(y3 )2 = (y1 )
2
This is a hyperbolic paraboloid.

10.3

Linear Geometry in 3 dimensions

Definition 10.5. For two vectors x = (x1 , x2 , x3 ), y = (y1 , y2 , y3 ) in R3 , define

x + y = (x1 + y1 , x2 + y2 , x3 + y3 )
x = (x1 , x2 , x3 ), R
q
length of x , denoted kx k = x12 + x22 + x32

dot product x .y = x1 y1 + x2 y2 + x3 y3 (note that x .x = kx k2 )

Proposition 10.4. For x , y R3
x .y = kx kky k cos
Proof.
ky x k2 = kx k2 + ky k2 2kx kky k cos

2kx kky k cos = x .x + y .y (y x )(y x )

= 2x .y

10.4

Geometry of planes

Fact there is a unique plane through any 3 points in R3 provided thay arent colinear.
We want to describe planes in terms of vectors and equations.
A plane in R3 is specified by
1) a point A on the plane
2) a vector n normal to the plane
Then x (x a).n = n x .n = a.n.
Proposition 10.5. If there are points A, B, C R3 that are not collinear, there exists
a unique plane through A, B, C.

10.4. GEOMETRY OF PLANES

71

Proof. Because both A B and B C are both in the plane

n.(a b) = 0

n.(c a) = 0
i.e.
r1 n1 + r2 n2 + r3 n3 = 0
s1 n1 + s2 n2 + s3 n3 = 0
where (r1 , r2 , r3 ) = a b, (s1 , s2 , s3 ) = c a.
We want to show that all solutions are {n | R}. Because A, B, C are not colinear,
r and s are not parallel.

How to find dist(P, )?
Find foot of perpendicular, Q and arbitrary A in the plane.
dist(P, Q) = ka pk cos
But
(a p).n = ka pkknk cos
(a p).n
dist(P, Q) =
knk

72

10.4. GEOMETRY OF PLANES

Notes