Chapter 3

Vector spaces

In this Chapter, we provide an abstract framework which encompasses what we have

seen on Rn and for Mmn (R).

3.1 Abstract definition

Before introducing the abstract notion of a vector space, let us make the following
observation. For any X, Y, Z ∈ Rn , for any A, B, C ∈ Mmn (R) and for any λ, µ ∈ R one
has

(i) (X + Y ) + Z = X + (Y + Z), (i) (A + B) + C = A + (B + C),

(ii) X + Y = Y + X, (ii) A + B = B + A,
(iii) X + 0 = 0 + X = X, (iii) A + O = O + A = A,
(iv) X − X = 0, (iv) A − A = O,
(v) 1X = X, (v) 1A = A,
(vi) λ(X + Y ) = λX + λY , (vi) λ(A + B) = λA + λB,
(vii) (λ + µ)X = λX + µX, (vii) (λ + µ)A = λA + µA,
(viii) (λµ)X = λ(µX). (viii) (λµ)A = λ(µA).

Note that these properties are borrowed from Chapter 1 and 2 respectively. Another
example which would satisfy the same properties is provided by the set of real functions
defined on R, together with the addition of such functions and with the multiplication
by a scalar. In this case, the element 0 (or O) is simply the function which is equal to
0 at any point of R.

Question: Can one give an abstract definition for these rules ?

47
48 CHAPTER 3. VECTOR SPACES

In the first definition, we give a more general framework in which λ and µ live. You
can always think about R as the main example for the following definition.
Definition 3.1.1. A field (F, +, ·) is a set F endowed with two operations + and · such
that for any λ, µ, ν ∈ F one has
(i) λ + µ ∈ F and λ · µ ∈ F, (internal operations)

(ii) (λ + µ) + ν = λ + (µ + ν) and (λ · µ) · ν = λ · (µ · ν), (associativity)

(iii) λ + µ = µ + λ and λ · µ = µ · λ, (commutativity)

(iv) There exist 0, 1 ∈ F such that λ + 0 = λ and 1 · λ = λ, (existence of identity

elements)

(v) There exists −λ ∈ F such that λ + (−λ) = 0, and if λ ̸= 0 there exists λ−1 ∈ F
such that λ · λ−1 = 1, (existence of inverse elements)

(vi) λ · (µ + ν) = λ · µ + λ · ν. (distributivity)
Note that for simplicity, one usually writes λ − µ instead of λ + (−µ) and λ/µ
instead of λ · µ−1 .
Example 3.1.2. Some examples of fields are (R, +, ·) the set of real numbers together
with
{ the usual addition and} multiplication, (Q, +, ·) the set of fractional numbers Q =
a/b | a, b ∈ Z with b ̸= 0 together with the usual addition and multiplication, (C, +, ·)
the set of complex numbers together with its addition and multiplication (as we shall
see at the end of this course). Note that for simplicity, one usually writes R, Q, C, the
other two operations being implicit.
Let us provide a more general framework for the elements of X, Y, Z or A, B, C and
for the properties stated in the table above. However, you can always think about Rn
or Mmn (R) as the main examples for the following definition. Note that two slightly
different fonts are used for the two different multiplications and for the two different
additions.
Definition 3.1.3. A vector space over a field (F, +, ·) consists in a set V endowed with
two operations + : V × V → V and · : F × V → V such that if X, Y, Z ∈ V and λ, µ ∈ F
the following properties are satisfied:
(i) (X + Y ) + Z = X + (Y + Z),

(ii) X + Y = Y + X,

(iii) There exists (a unique) 0 ∈ V such that X + 0 = 0 + X = X,

(iv) For all X ∈ V there exists −X ∈ V such that X + (−X) = 0,

(v) λ · X ∈ V and 1 · X = X,
3.1. ABSTRACT DEFINITION 49

(vi) λ · (X + Y ) = λ · X + λ · Y ,

(vii) (λ + µ) · X = λ · X + µ · X,

(viii) (λ · µ) · X = λ · (µ · X).

Before providing some examples, let us just mention a consequence of the previous
conditions, namely 0 · X = 0 for any X ∈ V . Indeed, for any X ∈ V one has

X = 1 · X = (1 + 0) · X = 1 · X + 0 · X = X + 0 · X,

from which one infers that 0 · X = 0. Let us also note that whenever the field F consists
in R, one simply says a real vector space instead of a vector space over the field R.

Examples 3.1.4. (i) F = R and V = Rn with the addition and multiplication by a

scalar, as introduced in Chapter 1,

(ii) F = R and V = Mmn (R) with the addition and the multiplication by a scalar, as
introduced in Chapter 2. More generally, for any field F the set Mmn (F), defined
exactly as Mmn (R), is a vector space over F,

(iii) F = R and V is the set of real functions defined on R, with the addition of
functions and the multiplication by scalar,

(iv) F = R and V = {polynomial functions on R}

{ ∑
m
}
= f : R → R | f (x) = aj xj with aj ∈ R ,
j=0

(v) F = R and V = {f : R → R continuous },

(vi) F = R and V = {f : R → R differentiable }.

From now and for simplicity we shall no more use two different notations for the two
different multiplications and for the two different additions. This simplification should
not lead to any confusion. In addition, we shall simply write F for the field, instead of
(F, +, ·), and the multiplication will be denoted without a dot; the sign · will be kept
for the scalar product only.

Definition 3.1.5. Let V be a vector space over F, and let W be a (non-void) subset of
V . Then W is a subspace of V if the following conditions are satisfied:

(i) If X, Y ∈ W , then X + Y ∈ W ,

(ii) If X ∈ W and λ ∈ F, then λX ∈ W .

50 CHAPTER 3. VECTOR SPACES

In other words, a subspace of a vector space V is a subset W of V which is stable

for the two operations, i.e. the addition and the multiplication by a scalar. The next
statement will be very useful when checking that a certain set is a vector space. Its
proof will be provided in Exercise 3.5.
Lemma 3.1.6. Any subspace W of a vector space V over a field F is itself a vector
space over F.
{ }
Examples 3.1.7. (i) t (x1 , x2 , . . . , xn−1 , 0) | xj ∈ R for any j ∈ {1, . . . , n − 1} ⊂
Rn is a subspace of Rn ,
(ii) If P, N ∈ Rn with N ̸= 0, then HP,N is a subspace of Rn if and only if 0 ∈ HP,N .
In particular, H0,N is a subspace of Rn ,
(iii) The set of upper triangular n × n matrices is a subspace of Mn (R),
(iv) The set of n × n symmetric matrices is a subspace of Mn (R).
The following statement deals with the intersection of subspaces or with the sum
of subspaces of a vector space.
Lemma 3.1.8. Let V be a vector space over a field F, and let W1 , W2 be two subspaces
of V . Then
{ }
(i) W1 ∩ W2 = X ∈ W1 and X ∈ W2 is a subspace of V ,
{ }
(ii) W1 + W2 = X = X1 + X2 | X1 ∈ W1 and X2 ∈ W2 is a subspace of V .
Proof. The proof consists in checking that the two conditions of Definition 3.1.5 are
satisfied.
(i) If X, Y ∈ Wj for j ∈ {1, 2}, then X + Y ∈ Wj because Wj is a subspace. In
particular, this implies that if X, Y ∈ W1 ∩ W2 , then X + Y ∈ W1 ∩ W2 . Similarly, in
this case one also has λX ∈ W1 ∩ W2 , since W1 and W2 are stable for the multiplication
by a scalar.
(ii) If X = X1 + X2 and Y = Y1 + Y2 with Xj , Yj ∈ Wj , then X + Y = X1 + X2 +
Y1 + Y2 = (X1 + Y1 ) + (X2 + Y2 ) with (X1 + Y1 ) ∈ W1 and (X2 + Y2 ) ∈ W2 , which implies
that X +Y ∈ W1 +W2 . Similarly, in this case one also has λX = λX1 +λX2 ∈ W1 +W2 ,
since both W1 and W2 are stable for the multiplication by a scalar.

3.2 Linear combinations

Let us start with a definition:
Definition 3.2.1. Let V be a vector space over a field F, and let X1 , . . . , Xr ∈ V . One
sets { }
Vect(X1 , . . . , Xr ) := λ1 X1 + · · · + λr Xr | λj ∈ F for j ∈ {1, . . . , r} ,
and call this set the subspace of V generated by X1 , . . . , Xr .
3.2. LINEAR COMBINATIONS 51

Obviously, the first thing to do is to check that this set is indeed a subspace of V .

Lemma 3.2.2. In the above setting, Vect(X1 , . . . , Xr ) is a subspace of V .

Proof. The proof consists in checking that both conditions of Definition 3.1.5 are sat-
isfied. First of all, if X = λ1 X1 + · · · + λr Xr and X ′ = λ′1 X1 + · · · + λ′r Xr , then

X + X ′ = (λ1 + λ′1 ) X1 + · · · + (λr + λ′r ) Xr ∈ Vect(X1 , . . . , Xr ).

| {z } | {z }
∈F ∈F

Similarly, if X = λ1 X1 + · · · + λr Xr and λ ∈ F, then

λX = (λλ1 ) X1 + · · · + (λλr ) Xr ∈ Vect(X1 , . . . , Xr ).

| {z } | {z }
∈F ∈F

Since both conditions are checked, it is thus a subspace of V .

Since Vect(X1 , . . . , Xr ) is a subspace of V , it was legitimate to call it as we did.

Note that one also says that λ1 , . . . , λr are the coefficients of the linear combination
λ1 X1 + · · · + λr Xr .

Remark 3.2.3. If Vect(X1 , . . . , Xr ) = V , then one says that V is generated by X1 , . . . , Xr ,

or that {X1 , . . . , Xr } is a generating family.

The following three examples are related to real vector spaces, as it is the case in
most of the examples of these lecture notes.

Examples 3.2.4. (i) Recall that Ej = t (0, . . . , 1, . . . , 0) with the entry 1 at the posi-
tion j. Then {Ej }nj=1 ≡ {E1 , E2 , . . . , En } is a generating family for Rn .

(ii) If N ∈ Rn with N ̸= 0, then Vect(N ) is the line passing through 0 and having the
direction N , i.e. Vect(N ) = L0,N , with the L0,N defined in Definition 1.5.1.

(iii) If X, Y ∈ R3 with X ̸= 0, Y ̸= 0, and Y ̸= λX for any λ ∈ R, then Vect(X, Y )

defines a plane in R3 passing through 0. In fact, it corresponds to the plane passing
through the three points 0, X, Y , as seen in Exercise 1.18.

Remark 3.2.5. If F = R and if one considers X1 , . . . , Xr ∈ Rn , then one can set

{ }
Box(X1 , . . . , Xr ) := λ1 X1 + · · · + λr Xr | λj ∈ [0, 1] for j ∈ {1, . . . , r} .

This is a subset of Vect(X1 , . . . , Xr ), called the hyperbox or generalized box generated

by X1 , . . . Xr . Note that Box(X1 , . . . , Xr ) is not a subspace. It is also easily observed
that Box(X1 ) corresponds to the segment between 0 and X1 and that Box(X1 , X2 ) cor-
responds to the parallelogram generated by X1 and X2 , and with one apex at 0.
52 CHAPTER 3. VECTOR SPACES

3.3 Convex sets

In this section we consider only real vector spaces, i.e. the field F is equal to R for all
vector spaces.

Definition 3.3.1. Let S be a subset of a real vector space V . Then S is convex if for
any X, Y ∈ S and for any t ∈ [0, 1] one has

X + t(Y − X) ≡ (1 − t)X + tY ∈ S.

Examples 3.3.2. (i) A ball is convex, but a doughnut is not convex,

(ii) For any X1 , . . . , Xr in a real vector space V , Vect(X1 , . . . , Xr ) is convex. Indeed,

if X = λ1 X1 + · · · + λr Xr and X ′ = λ′1 X1 + · · · + λ′r Xr with λj , λ′j ∈ R then
( ) ( )
(1 − t)X + tX ′ = (1 − t)λ1 + tλ′1 X1 + (1 − t)λr + tλ′r Xr ∈ Vect(X1 , . . . , Xr ),
| {z } | {z }
∈R ∈R

(iii) For any X1 , . . . , Xr in a real vector space V , Box(X1 , . . . , Xr ) is convex. Indeed, in

the framework of the previous example, observe that if 0 ≤ λj ≤ 1 and 0 ≤ λ′j ≤ 1
then one has for any t ∈ [0, 1]

0 ≤ (1 − t)λj + tλ′j ≤ (1 − t)1 + t1 = 1.

As a consequence, one infers that

( ) ( )
(1 − t)X + tX ′ = (1 − t)λ1 + tλ′1 X1 + (1 − t)λr + tλ′r Xr ∈ Box(X1 , . . . , Xr ).
| {z } | {z }
∈ [0,1] ∈ [0,1]

Definition 3.3.3. Let V be a real vector space, and let X1 , . . . , Xr ∈ V . We set

{
CS(X1 , . . . , Xr ) := λ1 X1 + · · · + λr Xr | 0 ≤ λj ≤ 1 for j ∈ {1, . . . , r}
}
and λ1 + λ2 + · · · + λr = 1

and call is the convex set generated or spanned by X1 , . . . , Xr .

Note that by definition, the following inclusions always hold

CS(X1 , . . . , Xr ) ⊂ Box(X1 , . . . , Xr ) ⊂ Vect(X1 , . . . , Xr ).

Example 3.3.4. If V = Rn , then CS(X1 ) corresponds just to the point X1 , CS(X1 , X2 )

corresponds to the segment between X1 and X2 while CS(X1 , X2 , X3 ) corresponds to the
triangle of apexes X1 , X2 and X3 .

Obviously, one has to show immediately the following statement:

3.4. LINEAR INDEPENDENCE 53

Lemma 3.3.5. If V is a real vector space and X1 , . . . , Xr ∈ V , then CS(X1 , . . . , Xr )

is convex.

In fact, CS(X1 , . . . , Xr ) is the smallest convex set containing X1 , . . . , Xr .

Proof. Let X = λ1 X1 + · · · + λr Xr with 0 ≤ λj ≤ 1 and λ1 + λ2 + · · · + λr = 1, and

let X ′ = λ′1 X1 + · · · + λ′r Xr with 0 ≤ λ′j ≤ 1 and λ′1 + λ′2 + · · · + λ′r = 1. Then for any
t ∈ [0, 1] one has
∑r
( )
′
(1 − t)X + tX = (1 − t)λj + tλ′j Xj
j=1

with 0 ≤ (1 − t)λj + tλ′j ≤ (1 − t)1 + t1 = 1 and

∑
r
( ) ∑
r ∑
r
(1 − t)λj + tλ′j = (1 − t) λj + t λ′j = (1 − t)1 + t1 = 1.
j=1 j=1 j=1

As a consequence, (1 − t)X + tX ′ ∈ CS(X1 , . . . , Xr ), which means that CS(X1 , . . . , Xr )

is convex.

3.4 Linear independence

The following definition will be of importance in the sequel.

Definition 3.4.1. Let V be a vector space over a field F, and let X1 , . . . , Xr ∈ V . The
elements X1 , . . . , Xr are linearly dependent if there exist λ1 , . . . , λr ∈ F not all equal to
0 such that
λ1 X1 + · · · + λr Xr = 0. (3.4.1)
The elements X1 , . . . , Xr are said linearly independent if there do not exist such scalars
λ1 , . . . , λ r .

Note that alternatively, the vectors X1 , . . . , Xr are linearly independent if whenever

(3.4.1) is satisfied, then one must have λ1 = λ2 = · · · = λr = 0. In this case, one also
says that the family {X1 , . . . , Xr } is linearly independent.

Examples 3.4.2. (i) For V = Rn , the family {Ej }nj=1 is linearly independent,

(ii) For V = Mn (R), the family {Irs }nr,s=1 of elementary matrices introduced in Section
2.5 is linearly independent,

(iii) For V = R2 , the elements ( 10 ), ( 01 ), and ( 11 ) are linearly dependent since

( ) ( ) ( )
1 0 1
1 +1 −1 = 0.
0 1 1
54 CHAPTER 3. VECTOR SPACES

(iv) Let f1 , f2 be two continuous real functions on R. In this case f1 , f2 are linearly
dependent if there exists λ1 , λ2 ∈ R with (λ1 , λ2 ) ̸= (0, 0) such that λ1 f1 +λ2 f2 = 0,
or more precisely
λ1 f1 (x) + λ2 f2 (x) = 0 for all x ∈ R.
For example, if f1 (x) = cos(x) and f2 (x) = sin(x), then f1 and f2 are linearly
independent even if 0 cos(0) + λ sin(0) = 0 for arbitrary λ ∈ R.
Definition 3.4.3. Let V be a vector space over a field F, and let X1 , . . . , Xr ∈ V . If
Vect(X1 , . . . , Xr ) = V and if X1 , . . . , Xr are linearly independent, then {X1 , . . . , Xr } is
called a basis for V . Alternatively, one also says that the family {X1 , . . . , Xr } constitutes
or forms a basis for V .
Examples 3.4.4. (i) {Ej }nj=1 forms a basis for Rn ,
(ii) {Irs }nr,s=1 forms a basis for Mn (R),
(iii) {x 7→ xn }∞
n=0 forms a basis for the vector space of all polynomials on R.

Let us consider a special case of the previous definition in the case n = 2. The
content of the following lemma will be useful later on, and its proof will be provided in
Exercise 3.9.
Lemma 3.4.5. Let ( ab ), ( dc ) ∈ R2 , with a, b, c, d ∈ R.
(i) The two vectors are linearly independent if and only if ad − bc ̸= 0,
(ii) If the two vectors are linearly independent, they form a basis of R2 .
Given a basis of a vector space V , any point X can be expressed as a linear com-
binations of elements of this basis. More precisely, one sets:
Definition 3.4.6. Let {X1 , . . . , Xr } be a basis for a vector space V over F. Then, for
X = λ1 X1 + · · · + λr Xr the coefficients {λ1 , . . . , λr } are called the coordinates of X
with respect to the basis {X1 , . . . , Xr } of V .
In order to speak about “the” coordinates, the following lemma is necessary.
Lemma 3.4.7. The coordinates of a vector with respect to a basis are unique.
Proof. Let {X1 , . . . , Xr } be a basis, and assume that
X = λ1 X1 + · · · + λr Xr = λ′1 X1 + · · · + λ′r Xr .
It then follows that
X − X = 0 = (λ1 − λ′1 )X1 + · · · + (λr − λ′r )Xr .
By independence of X1 , . . . , Xr , it follows that (λj −λ′j ) = 0 for all j ∈ {1, . . . , r}, which
means that λj = λ′j . Thus, the coordinates of X with respect to a basis are unique.
3.5. DIMENSION 55

3.5 Dimension
Question: Can one find 3 linearly independent elements in R2 ? For instance, if
A = ( 12 ), B = ( −5 10
7 ) and C = ( 4 ), are they linearly independent vectors ? The answer
is no, and there is no need to do any computation for getting this(answer. ) Indeed, let
us consider the more general setting provided by A = ( aa12 ), B = bb12 and C = ( cc12 ).
Then one has
( ) ( ) ( ) {
a1 b1 c1 λ1 a 1 + λ2 b 1 + λ3 c 1 = 0
λ1 + λ2 + λ3 = 0 ⇐⇒ .
a2 b2 c2 λ1 a 2 + λ2 b 2 + λ3 c 2 = 0
Note that this corresponds to a system of two equations for the three unknowns λ1 , λ2
and λ3 . As seen in Theorem 2.3.4, such a homogeneous system of equations has al-
ways a non-trivial solution, which means that there exists a non trivial solution for
the corresponding equation 3.4.1. As a consequence, three vectors in R2 can never be
independent.
More generally, one has:
Theorem 3.5.1. Let {X1 , . . . , Xr } be a basis of a vector space V over F. Let Y1 , . . . , Ym ∈
V and assume that m > r. Then Y1 , . . . , Ym are linearly dependent.
Note that if m ≤ r, the statement does not imply that Y1 , . . . , Ym are linearly
independent.
Now, in order to give the proof in its full generality, we need to extend the definition
of Mmn (R) to Mmn (F), for an arbitrary field F. In fact, since elements in a field can be
added and multiplied, all definitions related to Mmn (R) can be translated directly into
the same definitions for Mmn (F). The only modification is that any entry aij of a matrix
A belongs to F instead of R, and the multiplication of a matrix by a scalar λ ∈ R is
now replaced by the multiplication by an element of F. Then, most of the statements of
Section 2 are valid (simply by replacing R by F), and in particular Theorem 2.3.4 can
be obtained in this more general context. This theorem is precisely the one required for
the proof of the above statement.
Proof. Since X1 , . . . , Xr generate V , there exists aij ∈ F such that
Yj = a1j X1 + · · · + arj Xr for any j ∈ {1, . . . , m}.
Then, let us consider λ1 , . . . , λm ∈ F and observe that
λ1 Y1 + · · · + λm Ym = 0
⇐⇒λ1 (a11 X1 + · · · + ar1 Xr ) + · · · + λm (a1m X1 + · · · + arm Xr ) = 0
⇐⇒(λ1 a11 + λ2 a12 + · · · + λm a1m )X1 + · · · + (λ1 ar1 + . . . λm arm )Xr = 0
which is still equivalent to the following expression, by linear independence of X1 , . . . , Xr :
    
a11 a12 . . . a1m λ1 0
a21 a22 . . . a2m   λ2  0
    
 .. .. . . ..   ..  =  ..  .
 . . . .   .  .
ar1 ar2 . . . arm λm 0
56 CHAPTER 3. VECTOR SPACES

Finally, since m > r, it follows from a simple adaptation of Theorem 2.3.4 to Mn (F)
that this system of equation always has a non-trivial solution. However, this means
precisely that the elements Y1 , . . . , Ym are linearly dependent.

Corollary 3.5.2. Let V be a vector space and suppose that {X1 , . . . , Xn } is a basis for
V . Then any other basis for V also contains n elements.

Proof. Let {Y1 , . . . , Ym } be a second basis for V . If m > n then Y1 , . . . , Ym can not be
linearly independent, by the previous theorem. Similarly, if m < n then X1 , . . . , Xn can
not be linearly independent, also by the previous theorem. Since any basis is made of
linearly independent vectors, one obtains that the only possibility is m = n.

We now define a notion which has been implicitly used from the beginning for Rn .

Definition 3.5.3. Let V be a vector space over a field F, and let {X1 , . . . , Xn } be a
basis for V . Then one says that V is of dimension n, since this number is independent
of the choice of a particular basis for V . The dimension of the vector space V is denoted
by dim(V ).

Remark 3.5.4. In these lecture notes except in a few exercises, all vector spaces are
of finite dimension. This fact is tacitly assumed in many statements later on, but note
that vector spaces of infinite dimensions often appear in physics or in mathematics.

Examples 3.5.5. (i) Rn is of dimension n, Fn is also of dimension n,

(ii) For any vector space V and any X ∈ V , the dimension of Vect(X) is 1,

(iii) Any plane in R3 (passing through the origin) is of dimension 2, while a line passing
through the origin is of dimension 1.

The following result is often useful, when the dimension of the vector space is already
known.

Lemma 3.5.6. Let V be a vector space of dimension n, and let X1 , . . . , Xn ∈ V be

linearly independent. Then {X1 , . . . , Xn } is a basis for V .

Proof. One only has to show that Vect(X1 , . . . , Xn ) = V . By contradiction, assume that
there exists Y ∈ V such that Y ̸∈ Vect(X1 , . . . , Xn ). Then, the vectors X1 , . . . , Xn , Y
are linearly independent, and {X1 , . . . , Xn , Y } would generate a basis for V of dimension
n + 1, which is impossible by the previous Corollary.

3.6 The rank of a matrix

Let A ∈ Mmn (F), and recall from Section 2.2 that the columns of A have been denoted
by A1 , . . . , An . Each column is an element of Fm (a column vector with m entries in
F) and the family {A1 , . . . , An } generates a subspace of Fm , which we have denoted by
3.6. THE RANK OF A MATRIX 57

Vect(A1 , . . . , An ). In this case this subspace is called the column space. Alternatively,
the rows of A generate the subspace Vect(tA1 , . . . , tAm ) of Fn , which is called the row
space. The dimension of the first subspace is called the column rank, while the dimension
of the second subspace is called the row rank.
By what we have seen in the previous sections, the column rank corresponds to the
maximal number of linearly independent columns, while the row rank corresponds to
the maximal number of linearly independent rows. Our aim in this section is to study
these numbers.

Lemma 3.6.1. Elementary row operations do not change the row rank of a matrix.

Proof. For this proof, it is sufficient to observe that

Vect(tA1 , . . . , tAj , . . . , tAk , . . . , tAm )

= Vect(tA1 , . . . , tAk , . . . , tAj , . . . , tAm )
= Vect(tA1 , . . . , ctAj , . . . , tAk , . . . , tAm )
= Vect(tA1 , . . . , tAj + ctAk , . . . , tAk , . . . , tAm )

for any c ̸= 0. Since these subspaces are the same, their dimension coincide.
Similarly, one can show that elementary row operations do not change the column
rank of a matrix. Note that this proof is less easy since elementary row operations have
been defined on rows, and not on columns.

Theorem 3.6.2. For any A ∈ Mmn (F), the column rank and the row rank of A are
equal.

Thanks to this statement, it is sufficient to speak about the rank of a matrix, denoted
by rank(A), there is no need to specify if it is the column rank or the row rank.
Proof. First of all, recall that the matrix A is row equivalent to a matrix B in the
standard form, see Corollary 2.4.10. Since elementary row operations do not change the
row rank or the column rank, the matrix B has the same row rank and column rank as
the original matrix A. Then, it is easily observed that the number of leading coefficients
of B is equal to the number of linearly independent rows, but also to the number of
linearly independent columns of B. Therefore, the number of leading coefficients of B
is equal to the row rank of B and to the column rank of B. It follows that these two
numbers are equal, and that the row rank of A and the column rank of A are also equal
to this number.
58 CHAPTER 3. VECTOR SPACES

3.7 Exercises
Exercise 3.1. Show that the following sets of elements of R3 form subspaces :
i) S1 := {t (x, y, z) ∈ R3 | x + y + z = 0},

ii) S2 := {t (x, y, z) ∈ R3 | x = y and 2y = z},

iii) S3 := {t (x, y, z) ∈ R3 | x + y = 3z}.

Exercise 3.2. Let V be a subspace of Rn , and let W be the set of all elements of Rn
which are perpendicular to all elements of V . Show that W itself is a subspace of Rn .
This subspace is often denoted by V ⊥ and called the orthogonal complement of V in Rn .
Exercise 3.3. Let A1 , . . . , Ar be generators of a subspace V of Rn . Let W be the set of
all elements in Rn which are perpendicular to A1 , . . . , Ar . Show that W = V ⊥ .
Exercise 3.4. Show that the set of all real polynomials is a subspace of the vector space
of all real and continuous functions on R. Exhibit a generating family for this subspace.
Exercise 3.5. Let V be a vector space over a field F. Show that any subspace of V is
itself a vector space.
Exercise 3.6. Let S be a convex set in a real vector space V .
i) For λ ∈ R, show that λS is a convex set in V , with λS = {λX | X ∈ S}.

ii) For Y ∈ V , show that S + Y is a convex set in V , with S + Y = {X + Y | X ∈ S}.

Exercise 3.7. Show that the intersection of two convex sets is still convex.
( ) ( ) ( )
1 0 1
Exercise 3.8. Show that the vectors 0 , 1 and 1 form a basis of R3 .
0 0 1

Exercise 3.9. Let ( ab ), ( dc ) ∈ R2 . Show that these two vectors are linearly independent
if and only if ad − bc ̸= 0.
Exercise 3.10. Express the coordinates of Y in the basis generated by X1 and X2 :
( ) ( ) ( )
1 1 0
i) Y = , X1 = and X2 = ,
0 1 1
( ) ( ) ( )
2 1 1
ii) Y = , X1 = and X2 = .
1 −1 1
Exercise 3.11. Let X1 , . . . , Xr be non-zero elements of Rn and assume that Xj ·Xk = 0
for each j ̸= k. Show that these elements are linearly independent.
Exercise 3.12. Determine the dimension of the following subspaces:
i) S1 := {t (x, y, z) ∈ R3 | x + y + z = 0},
3.7. EXERCISES 59

ii) S2 := {t (x, y, z) ∈ R3 | x = y and 2y = z},

iii) S3 := {t (x, y, z) ∈ R3 | x + y = 3z}.

Exercise 3.13. Determine the rank of the following matrices :

       
6 3 −4 1 0 2 0 1 3 −2 1 2 −1 2 1

a) −4 1 −6  
b) 2 −1 3 c) 2 1 −4 3  d) 2 4 1 −2 3
1 2 −5 4 1 8 2 3 2 −1 3 6 2 −6 5

Exercise 3.14. A doubly stochastic matrix is a n × n matrix A = (ajk ) such that

ajk ∈ [0, 1] and such that the sum of the elements of each line is equal to 1, as well as
the sum of the elements of each column.

(i) Show that the product of two doubly stochastic matrices is still a doubly stochastic
matrix,

(ii) Show that the set of all doubly stochastic matrices is a convex set.
60 CHAPTER 3. VECTOR SPACES