2.

7 Geometric Interpretation

As you have seen in the introductory module vectors in R2 /R3 can be represented as points in 2 or 3
dimensional space. In this section we will look geometric interpretations of some of the things we have
seen so far.
A system of linear equations in n unknowns specifies a set in n-space.

Example 2.7.1.
Consider:
x1 + x2 + x3 = −1
2x1 + x3 = 1
3x1 + x2 = −4

Using row reduction we get x1 = −0.5, x2 = −2.5 x3 = 2, which specifies a point. Whereas:

x1 + x2 + x3 = −1
2x1 + x3 = 1

Using row reduction we get x1 = −2.5 − 0.5x3 and x2 = 1.5 − 0.5z giving the line
   
−2.5 −0.5
 1.5  + λ  −0.5  for λ ∈ R
0 1

We have seen that we can apply matrices to vectors via matrix multiplication. So we can see a matrix
A ∈ Mm×n (R) as a map:

A: Rn →7 Rm
A(v) = Av

We can represent many different operations using matrices.

Example 2.7.2.
 
5 0
Consider A =
0 5
   
x 5x
Then A = .
y 5y
This is a stretch by a factor of 5.

17
Definition 2.7.3. Let T be a function from Rn to Rm then we say T is a linear transformation if
for every v1 , v2 ∈ Rn and every α, β ∈ R we have:

T (αv1 + βv2 ) = αT (v1 ) + βT (v2 )

Proposition 2.7.4. Let A ∈ Mn×m (R) then seen as a map from Rn to Rm A is a linear transfor-
mation.
Proof:

A(αv1 + βv2 ) = A(αv1 ) + A(βv2 ) by distributivity of matrix multiplication

= αA(v1 ) + βA(v2 ) by exercise

Proposition 2.7.5. Let A ∈ Mn×n (R). The following are equivalent:

(i) A is invertible with inverse A−1 = AT

(ii) AT A = In = AAT .

(iii) A preserves inner products (i.e. for all x, y ∈ Rn (P x) · (P y) = x · y.

Proof:
(i) ⇔ (ii) is just by definition.
(ii) ⇔ (iii)First note that for x, y ∈ Rn x · y as defined in the intro to maths course is just xT y as
matrix multiplication. So A preserves inner products if and only if:

(P x) · (P y) = x · y ∀x, y ∈ Rn
⇐⇒ (P x)T (P y) = xT y ∀x, y ∈ Rn
⇐⇒ xT P T P y = xT In y ∀x, y ∈ Rn
⇐⇒ xT (P T P − In )y = 0 ∀x, y ∈ Rn

(ii) ⇒ (iii) now trivial.

 
0
 .. 
 . 
  th
(iii) ⇒ (ii) let xi =  1  i.e. column vector with 0’s everywhere except the i row where

 .. 
 . 
0
there is a 1. Then we know for each xi (xi )T (P T P − In )y = 0 so we can conclude that
 
0
T  .. 
(P P − In )y =  . 
0

Similarly taking yi to be the column vector with 0’s everywhere except the ith row where there is a
1 we get (P T P − In ) = 0 so P T P = In .

18
Definition 2.7.6. A matrix A ∈ Mn×n is called Orthogonal if it is such that A−1 = AT

Example 2.7.7.

1. Consider the matrix  

0 −1
A=
1 0
 
0 1
This matrix is orthogonal as A−1 =
−1 0
   
x −y π
If we apply it to we get . This is a rotation through 2 radians anti clockwise.
y x
2. Consider the matrix  
0 −1
B=
−1 0
 
0 −1
This matrix is orthogonal asA−1 =
−1 0
   
x −y
If we apply it to we get . This is a reflection through the line y = −x.
y −x

Exercise 2.7.8. Watch the linear Algebra video to help you.

1. Let Rθ be the anticlockwise rotation of R2 about the origin through θ radians. Using any
school geometry or trigonometry you like, find the matrix representing Rθ .
Assuming Rθ is linear (see lectures!) the vector e1 = 10 is rotated to cos θ 0




sin θ while e2 = 1
is rotated to −cos
sin θ


θ so the matrix is
 
cos θ − sin θ
Rθ = .
sin θ cos θ

19
2.8 Fields

So far, for both matrices and linear equations, we have only been using entries in R. However, we could
have taken entries from any field.
Every field has distinguished elements 0 (additive identity) and 1 (multiplicative identity).

Fact 2.8.1. Over any field F we can define:

1. The null matrix (i.e. the additive identity matrix) for Mn×m (F ) as
 
0 0 ... 0
 0 0 
 
 .. . . 
 . . 
0 0

2. The (multiplicative) identity matrix for Mn×n (F ) as

 
1 0 ... 0
 0 1 
 
 .. .. 
 . . 
0 0

Remark 2.8.2 It is important to know what field we are working in, and that we don’t say take scalars
from a different field to the one matrix entries are from. (e.g. the set of matrices Mn×m (Q) is not closed
under scalar multiplication by elements from R).

Being able to work over a general field allows us to use finite fields.

Theorem 2.8.3. Let Fp = {0, 1, ..., p − 1}, consider Fp with addition defined by addition modulo
p and multiplication as multiplication modulo p. Then the structure (Fp , + (mod p) , × (mod p) ) is a
field.
Proof:
A1-4 (Additive (commutative) group) obvious from properties of addition in Z.
M1-3 (mulitplicative semigp with 1)obvious from properties of addition in Z.
M4: inverses: obviously for 0 ≤ x < p we have gcd(x, p) = 1 by Intro to Uni Maths we have:
∃s, t ∈ Z such that 1 = sx + tp then take x−1 = s (mod p).
D1 (distributive law) obvious from properties of addition in Z.
Example 2.8.4. F6 defined as above is not a field. For example 3 6= 0 does not have an inverse.

20
3 Vector Spaces

3.1 Intro to Vector Spaces

Definition 3.1.1. Let F be a field. A vector space over F is a non-empty set V together with the
following maps:

1. Addition
⊕: V ×V → 7 V
(v1 , v2 ) → 7 v1 ⊕ v2

2. Scalar Multiplication
: F ×V 7→ V
(f, v2 ) 7→ f v2

⊕ and must satisfy the following Vector Space axioms:

For Vector Addition:

A1 Associative law: (u ⊕ v) ⊕ w = u ⊕ (v ⊕ w).

A2 Commutative law: v ⊕ w = w ⊕ v.

A3 Additive identity: 0V ⊕ v = v, where 0V is called the IDENTITY vector (or sometimes the
zero vector).

A4 Additive inverse: v ⊕ ( v) = 0V .

For scalar multiplication:

A5 Distributive law: r (v ⊕ w) = (r v) ⊕ (r w).

A6 Distributive law: (r + s) v = (r v) ⊕ (s v).

A7 Associative law: r (s v) = (rs) v.

A8 Identity for scalar mult: 1 v = v.

From now on we will drop the ⊕ and , and use + and · the point was to emphasise that these are not
the same as field addition and multiplication.

Definition 3.1.2. Let V be a vector space over F we call:

• Elements of V are called vectors.

• Elements of F are called scalars.

• We sometimes refer to V as an F -vector space.

21
Example 3.1.3. The following are examples of vector spaces over R:

• The canonical example is the set of vectors Rn over R, where ⊕ is normal vector addition
and is multiplication by a scalar. The additive inverse of v is simply −v

• The set Mmn of all m×n matrices. This is because addition of two m×n matrices produces an
m×n matrix and multiplication of an m×n matrix by a scalar also produces a m×n matrix.
The zero vector is the zero matrix, and for any matrix A, the matrix −A is the additive inverse.
Properties of matrix arithmetic covered in Chapter 1, show that all properties in Definition
3.1.2 required of a vector space are satisfied. We will see this later in the course in detail.

• Define RX to be the set of real valued functions on X (i.e. RX := {f : f a function, f :

X → R}). Then for f, g ∈ RX and α ∈ R define:

f ⊕ g :X → R (α f ) :X → R
(f ⊕ g)(x) = f (x) + g(x) (α f )(x) = α(f (x))

Exercise 3.1.4. Which of the following examples of vector spaces over R :

1. The set of vectors

  
a
V = : a, b ∈ Z with the usual vector addition and multiplication
b

√
   
1 1
No, because ∈ V, , but 2 6∈ V
1 1

2. The set of vectors:

  
a+1
V = : a ∈ R with the usual vector addition and multiplication
2

 
0
No, because 6∈ V
0

3. V = R2 with the following addition and scalar multiplication operations:

         
x a x+a x 0
⊕ = and r =
y b y+b y ry
yes

22
3.2 Subspaces

Definition 3.2.1. A subset W of a vector space V is a subspace of V if

S1 W is not empty (i.e. e ∈ W )

S2 for v, w ∈ W , then v ⊕ w ∈ W closed under vector addition

S3 v ∈ W and r ∈ R, then r v ∈ W closed under scalar multiplication.

N.B. Sometimes we use the notation U ≤ V to mean U is a subspace of V .

Remark 3.2.2 Note that V and the zero subspace, 0 are always subspaces of V . Any other subspace of
V is called a proper subspace of V .

Proposition 3.2.3. Every subspace of an F -vector space V must contain the zero vector.
Proof:
Claim: For an F -vector space V with 0 ∈ F the field additive identity we have 0v = 0V for all
v ∈ V . Proof of claim: Exercise (note enought to show that 0v is the vector space additive identity.
Let v ∈ V (V is non-empty) then 0V = 0 ⊕ v ∈ V as V is closed under scalar multiplication.

  
x
Example 3.2.4. Show that the set X = ;x ∈ R is a subspace of R2 .
0

Worked Answer
 
1
S1 The vector ∈ X, therefore X is non-empty.
0
   
x1 x2
S2 If v = and w = , then
0 0
   
x1 x2
v⊕w = ⊕
0 0
 
x1 + x2
= ∈X
0
 
x1
S3 If v = and r ∈ R, then
0
 
x1
r v = r
0
 
rx1
= ∈X
0
  
x
Therefore, X = ;x ∈ R is a subspace of R2 .
0

23
Exercise 3.2.5. All subspaces of a vector space over F are vector spaces over F in their own right.

Theorem 3.2.6. Let U, W be subspaces of V . Then U ∩ W is a subspace of V. In general, the

intersection of any set of subspaces of a vector space V is a subspace of V .

ProofLet C be a set of subspaces of V and T is their intersection. Then T 6= ∅ since every subspace
of V (and therefore every subspace in C) contains the zero vector, and so does T .
Suppose that x, y ∈ T . Since x and y belong to every subspace W in C, so does x⊕y, and therefore
x ⊕ y ∈ T.
If x ∈ T , then x belongs to every subspace W in the set C, and so does r x and so r x ∈ T.
Therefore T is a subspace of V .

Example 3.2.7. Note that in general if U and W are subspaces of V , then U ∪ W is not a subspace
of V . For example, let
     
x 0
U= : x ∈ R ,W = :y∈R and V = R2 .
0 y

Then    
1 0
, ∈U ∪W
0 1
but      
1 0 1
⊕ = ∈
/ U ∪ W.
0 1 1

24