You are on page 1of 19

1

SIM2002
LINEAR ALGEBRA

Semester 1, Session 2022/2023

A Note to Students

Why Linear Algebra ? Linear algebra is the branch of mathematics concerned with the study
of linearity in algebraic systems, which is one of the most important subjects in mathematics due to its
widespread applications in statistics, natural science, engineering, computer science, data analysis, social
science, business, economics and etc. The basic notions of linear algebra are linear systems, matrices,
vectors, linear spaces, inner product spaces, linear transformations, determinants, eigenvalues, eigenvectors,
decompositions, tensors and etc. In linear algebra, the concepts are as important as the computations.
Solving systems of linear equations adepts matrix computations, and at the same time, it will simultaneously
develop abstract reasoning skill. Its balance of computation and mathematical abstraction makes linear
algebra unique among mathematics. For those who make use of pure and applied mathematics in their
professional careers, an understanding and appreciation of linear algebra is indispensable.
The major prerequisite of the mathematical background required for an understanding of linear algebra
is not a specific list of courses in mathematics, but rather the ability to reason abstractly, to proceed logically
from hypothesis to conclusion. Anyone who processes this quality is capable of understanding the material
presented in this course.
Chapter 1

Vector Spaces

Linear Algebra is a relatively recent mathematical development. Its roots extend back 100 years to the
work of Hamilton, Cayley and Sylvester, but it has attracted widespread interest only in the past two or
three decades. Today linear algebra and matrices are effective tools in computer sciences and data analysis;
in quantum theory as well as classical mechanics; in aeronautical, mechanical and electrical engineering; in
statistics, economics and linear programming and thus in all the social sciences which these theories serve.

The major prerequisite of the mathematical background required for an understanding of linear algebra
is not a specific list of courses in mathematics, but rather the ability to reason abstractly, to proceed logically
from hypothesis to conclusion. Anyone who needs more rigor exercises in mathematical writing and proving
technique, we refer them to [17, 18]).

1.1 Preliminaries
Together with the usual operations of addition and multiplication, the set Z of integers and the set R of real
numbers are examples of an algebraic structure called a ring. Basically, a ring is a set in which two binary
operations can be defined so that the sum and the product of any two elements in the set is an element of
the set. More precisely, a ring is defined as follows:

Definition 1.1.1 A nonempty set R with two operations called addition ’+’ and multiplication ’ · ’, is said
to be a ring if the following conditions hold true for all elements a, b, c ∈ R.
(i) a + b and a · b are unique elements in R;
(ii) a + b = b + a (commutative law);

11
12 CHAPTER 1. VECTOR SPACES

(iii) a + (b + c) = (a + b) + c and a · (b · c) = (a · b) · c (associative laws);


(iv) there exists a unique element in R, called zero and denoted by 0, such that a + 0 = 0 + a = a (existence
of identity with respect to addition);
(v) for each a ∈ R, there exists a unique element in R denoted by −a such that a + (−a) = (−a) + a = 0
(existence of inverse with respect to addition);
(vi) a · (b + c) = a · b + a · c, (a + b) · c = a · c + b · c (distributive laws).

A ring R is said to be commutative if ab = ba for every a, b ∈ R. If a ring R contains an element e


such that ae = ea = a for all a ∈ R, we say that R is a ring with identity. It is easy to verify that if a
ring has an identity, then the identity is unique (Suppose that e, f ∈ R are identities. Then e = ef = f ).
The identity e is usually denoted by 1. If 1 = 0, then for any a ∈ R, a = a · 1 = a · 0 = 0 (note that
a0 = a(0 + 0) = a0 + a0), so the ring consists of a single element. This is the trivial (or zero) ring.

Example 1.1.2
(i) Let “+” and “.” denote the usual addition and multiplication for numbers. Then Z, Q, R and C are
commutative rings with identity with respect to “+” and “.”.
(ii) The set all even integers 2Z = { 2k : k ∈ Z } is a commutative ring under the usual operations on Z,
but it has no identity.
(iii) Let n be a positive integer. The set Zn = {0, 1, . . . , n − 1} is a commutative ring under addition and
multiplication modulo n defined by
a ⊞ b = (a + b) mod n, and a ⊙ b = (a · b) mod n

for all a, b ∈ Zn . Therefore, (Zn , ⊞, ⊙) is a ring with 1 ∈ Zn the identity.


(iv) Let n be a positive integer. The set of all n × n matrices Mn (C) over C is a ring under matrix addition
and multiplication. Note that Mn (C) is not a commutative ring, and the identity matrix In is the
identity for Mn (C).

An element a in a ring R is called invertible or a unit if it has an inverse, i.e., there exists b ∈ R such
that ab = ba = 1. It can be verified that for each unit a ∈ R, the existence of an inverse for a is unique. For
the real ring R, every nonzero element a ∈ R is a unit because a1 ∈ R is an inverse of a. However, we note
1
that not every nonzero element of a ring is a unit. For example, 2 ∈ 2Z but it is not a unit since 2 ∈
/ 2Z.
Also, for the ring M2 (C), we see that
 
i 0
A= ∈ M2 (C).
0 0

But, A is not a unit because A is not an invertible matrix.

A nonzero commutative ring with identity such that each nonzero element is invertible is called a field
and is usually denoted by F. More precisely, we have

Definition 1.1.3 A field F is a ring for which the following conditions hold.
(i) a · b = b · a for all a, b ∈ F,
(ii) there exists a unique element 1 in F, called the identity of F, such that 1 · a = a · 1 = a for all a ∈ F,
and
(iii) for every nonzero element a ∈ F, there exists a unique element in F, denoted a−1 , such that a−1 · a =
a · a−1 = 1.
1.1. PRELIMINARIES 13

In view of Definitions 1.1.1 and 1.1.3, we say that a field is a nonempty set F with two operations called
addition ’+’ and multiplication ’ · ’, in which the following conditions hold.
(i) for each a, b ∈ F, a + b and a · b are unique elements in F,
(ii) a + b = b + a for every a, b ∈ F,
(iii) a · b = b · a for all a, b ∈ F,
(iv) a + (b + c) = (a + b) + c and a · (b · c) = (a · b) · c for all a, b ∈ F,
(v) there exists a unique element 0 in F, called the zero of F, such that a + 0 = 0 + a = a for all a ∈ F,
and
(vi) there exists a unique element 1 in F, called the identity of F, such that 1 · a = a · 1 = a for all a ∈ F,
and
(vii) for each a ∈ F, there exists a unique element in F denoted by −a such that a + (−a) = (−a) + a = 0
for all a ∈ F,
(viii) for each nonzero element a ∈ F, there exists a unique element in F, denoted a−1 , such that a−1 · a =
a · a−1 = 1, and
(ix) a · (b + c) = a · b + a · c, (a + b) · c = a · c + b · c for all a, b, c ∈ F.

Example 1.1.4
(i) The number systems Q, R and C are fields with respect to the usual addition and multiplication.
(ii) The number system Z is not a field since 2 ∈ Z and 2 ̸= 0 but there does not exist a ∈ Z such that
2 · a = 1.
(iii) Let Z2 = {0, 1} with the following two operations:

+ 0 1 × 0 1
0 0 1 and 0 0 0
1 1 0 1 0 1

It can be checked that (Z2 , +, ×) constitutes a field which is usually called the Galois field of two
elements. Note that 1 + 1 = 0; and a(a + 1) = 0 and a2 = a for any a ∈ Z2 .

(iv) Let Z3 = {0, 1, α} with the following two operations:

+ 0 1 α × 0 1 α
0 0 1 α 0 0 0 0
and
1 1 α 0 1 0 1 α
α α 0 1 α 0 α 1

It can be checked that (Z3 , +, ×) is the field of three elements. Note that 1 + 1 + 1 = α + 1 = 0. In
this case, α = −1.
(v) Let F be a field. Since 1 ∈ F, we have −1 ∈ F. Let a ∈ F. Thus (−1)a + a = (−1)a + 1a = ((−1) + 1)a =
0a = 0 (as 0a = a0 = 0). Hence (−1)a + a = 0. By the uniqueness of −a in F, we get (−1)a = −a.

Let R be a ring with identity. The smallest positive integer k such that k · 1 = 0 is called the charac-
teristic of R and is denoted by char R. For example, Z2 = {0, 1} has char Z2 = 2 because 1 + 1 = 0 in Z2 ;
and Z3 = {0, 1, α} has char Z3 = 3 because 1 + 1 + 1 = 0 in Z3 . If no such positive integer k exists, we say
that R has characteristic zero. We note that if char R = k, then k · a = 0 for every a ∈ R. This is because

k · a = (1 + · · · + 1) · a = (k · 1) · a = 0 · a = 0.
| {z }
k times

On the other hand, in Z, Q, R and C, the equation n · 1 = 0 implies n = 0, and hence, no such positive
integer exists. Consequently, we conclude that char R = 0 when R = Z, Q, R and C.
14 CHAPTER 1. VECTOR SPACES

1.2 Vector Spaces

Linear algebra is that branch of mathematics which treats the common properties of algebraic systems which
consist of a set, together with a reasonable notion of a “linear combination” of elements in the set. In this
section we shall define the mathematical object which experience has shown to be the most useful abstraction
of this type of algebraic system.
Let us begin with a definition of our principle object of study.

Definition 1.2.1 Let F be a field. A vector space (or linear space) over F is a nonempty set V together
with operations of addition : V × V → V, denoted by “+”; and scalar multiplication : F × V → V, denoted
by “ · ”, such that the following properties hold for every vectors u, v, w ∈ V and scalars α, β ∈ F:
(A1) u + v is a unique element in V (closure of addition);
(A2) u + v = v + u (commutativity of addition);
(A3) u + (v + w) = (u + v) + w (associativity of addition);
(A4) there exists an element 0 ∈ V, called the zero vector, such that u + 0 = 0 + u = u (additive identity);
(A5) for each u ∈ V, there exists a unique element −u ∈ V such that u + (−u) = 0 (additive inverse);
(M1) α · u is a unique element in V (closure of scalar multiplication);
(M2) α · (u + v) = α · u + α · v (distributive law);
(M3) (α + β) · u = α · u + β · u (distributive law);
(M4) α · (β · u) = (αβ) · u (associative law);
(M5) 1 · u = u (monoidal law).

In what follows, we will suppress the use of dot “·” of scalar multiplication and write αu instead of α · u.
We note that Axioms (A1) − (A5) of Definition 1.2.1 state that the vector space V is a commutative group
under addition +, thus it can be verified that the zero vector 0 of V is unique, and also, each element u of
V has a unique additive inverse in it, denoted −u, and is called the negative of u. In fact, we can show
that (−1)u = −u for every u ∈ V (see Theorem 1.2.11 (iii)). The subtraction of vectors is defined by

u − v = u + (−v) for all u, v ∈ V. (1.2.1)

We first consider a simple example of vector space R3 .


Example 1.2.2 Let R be the field of real numbers, or simply the real field. In the study of vector algebra
in R3 = R × R × R, we have been introduced that each vector u in R3 can be represented by
u = ai + bj + ck

where a, b, c ∈ R, and i = (1, 0, 0), j = (0, 1, 0) and k = (0, 0, 1) are unit vectors of R3 in the direction of
x-axis, y-axis and z-axis, respectively. Recall that a1 i + b1 j + c1 k = a2 i + b2 j + c2 k in R3 if and only if
a1 = a2 , b1 = b2 and c1 = c2 .
Let u = ai + bj + ck and v = si + tj + hk be vectors in R3 and let α be a scalar in R. The vector addition
and the scalar multiplication are, respectively, defined by
u + v = (a + s)i + (b + t)j + (c + h)k and αu = (αa)i + (αb)j + (αc)k. (1.2.2)

For simplicity of notation, we abbreviate ai + bj + ck to (a, b, c) in R3 . It follows from (1.2.2) that the vector
addition and the scalar multiplication are represented by
u + v = (a, b, c) + (s, t, h) = (a + s, b + t, c + h) and αu = α(a, b, c) = (αa, αb, αc). (1.2.3)
1.2. VECTOR SPACES 15

Let u = (a, b, c), v = (s, t, h), w = (p, q, r) ∈ R3 and let α, β ∈ R. In view of (1.2.3), it is not difficult to
check that u + v and αu are unique vectors in R3 which verify the closure of addition (A1) and the closure
of scalar multiplication (M1).
Note that
u + v = (a + s, b + t, c + h) = (s + a, t + b, h + c) = v + u
proves (A2), and
u + (v + w) = (a, b, c) + (s + p, t + q, h + r)
= (a + (s + p), b + (t + q), c + (h + r))
= ((a + s) + p, (b + t) + q, (c + h) + r) = (u + v) + w

proves (A3). We define 0 = (0, 0, 0) and −u = (−a, −b, −c). Clearly, 0 amd −u are elements of R3 . We see
that
u + 0 = (a + 0, b + 0, c + 0) = (0 + a, 0 + b, 0 + c) = 0 + u ,
u + (−u) = (a + (−a), b + (−b), c + (−c)) = (a − a, b − b, c − c) = (0, 0, 0) = 0.
Hence axioms (A4) and (A5) are claimed.
We further now to verify axioms for the scalar multiplication. Notice that

α(u + v) = α(a + s, b + t, c + h)
= (α(a + s), α(b + t), α(c + h))
= (αa + αs, αb + αt, αc + αh)
= (αa, αb, αc) + (αs, αt, αh) = αu + αv.

This proves (M2). Likewise, for (M3) and (M4), we see that

(α + β)u = ((α + β)a, (α + β)b, (α + β)c) = (αa + βa, αb + βb, αc + βc) = αu + βu,
α(βu) = α(βa, βb, βc) = (α(βa), α(βb), α(βc)) = ((αβ)a, (αβ)b, (αβ)c) = (αβ)u.
Lastly, we verify (M5). Let 1 ∈ R. Since 1x = x for any x ∈ R, it follows that 1u = (1a, 1b, 1c) = (a, b, c) = u
as desired.
This concludes that R3 constitutes a vector space over R under the operations of vector addition and
scalar multiplication defined in (1.2.3).

Example 1.2.3 Using a similar argument as in Example 1.2.2, we see that R is a vector space over R.
More generally, let F be a field. It can be checked that F is a vector space over itself under the usual addition
in F as the vector addition, and the usual multiplication in F as the scalar multiplication. In this example
vectors and scalars are indistinguishable.

Example 1.2.4 (Vector space of n-tuples) Let X be a nonempty set and let n be a positive integer. Then
the set
X n = { (x1 , . . . , xn ) : xi ∈ X, i = 1, . . . , n }
is called the n-fold Cartesian product of X. Let F be a field. If X = (X, + , · ) is a vector space over F,
then X n constitutes a vector space over F with respect to the following two operations:
(a1 , . . . , an ) + (b1 , . . . , bn ) := (a1 + b1 , . . . , an + bn )
α(a1 , . . . , an ) := (αa1 , . . . , αan )
for every (a1 , . . . , an ), (b1 , . . . , bn ) ∈ X n and α ∈ F. To see this, let u = (u1 , . . . , un ), v = (v1 , . . . , vn ) and
w = (w1 , . . . , wn ) be elements in X n and let α, β ∈ F. Note that
(A1): u + v = (u1 + v1 , . . . , un + vn ) is in X n since ui + vi ∈ X for each i = 1, . . . , n.
(A2): u + v = (u1 + v1 , . . . , un + vn ) = (v1 + u1 , . . . , vn + un ) = v + u since ui + vi = vi + ui in X for each
i = 1, . . . , n.
16 CHAPTER 1. VECTOR SPACES

(A3): u + (v + w) = (u1 , . . . , un ) + (v1 + w1 , . . . , vn + wn ) = (u1 + (v1 + w1 ), . . . , un + (vn + wn )) =


((u1 + v1 ) + w1 , . . . , (un + vn ) + wn ) = (u + v) + w because ui + (vi + wi ) = ui + vi + wi = (ui + vi ) + wi
in X for each i = 1, . . . , n.
(A4): We denote 0 := (0, . . . , 0) ∈ X n . Then 0 + u = (0 + u1 , . . . , 0 + un ) = (u1 , . . . , un ) = u.
(A5): We denote −u := (−u1 , . . . , −un ) ∈ X n . Then u+(−u) = (u1 +(−u1 ), . . . , un +(−un )) = (0, . . . , 0) =
0. Also, by (A2), we have −u + u = u + (−u) = 0.
(M1): αu = (αu1 , . . . , αun ) ∈ X n since αui ∈ X for each i = 1, . . . , n.
(M2): α(u + v) = α(u1 + v1 , . . . , un + vn ) = (α(u1 + v1 ), . . . , α(un + vn )) = (αu1 + αv1 , . . . , αun + αvn ) =
(αu1 , . . . , αun ) + (αv1 , . . . , αvn ) = αu + αv, as α(ui + vi ) = αui + αvi in X for each i = 1, . . . , n.
(M3): (α+β)u = ((α+β)u1 , . . . , (α+β)un ) = (αu1 +βu1 , . . . , αun +βun ) = (αu1 , . . . , αun )+(βu1 , . . . , βun ) =
αu + βu, as (α + β)ui = αui + βui in X for each i = 1, . . . , n.
(M4): α(βu) = α(βu1 , . . . , βun ) = (αβu1 , . . . , αβun ). On the other hand, (αβ)u = (αβu1 , . . . , αβun ).
Therefore, we have α(βu) = (αβ)u,
(M5): 1u = 1(u1 , . . . , un ) = (1u1 , . . . , 1un ) = (u1 , . . . , un ) = u. This is because 1ui = ui in X for each i.
We are done. As a side remark, if X = F, then Fn is called the n-tuple space; and if F = R, then Rn is
called the Euclidean n-space. When X = R and n = 3, it is the case of Example 1.2.2.

Example 1.2.5 (Vector space of matrices) Let F be a field and m, n be positive integers. Let Mm,n (F)
denote the set of all m × n matrices over F. Symbolically,

Mm,n (F) = {A = (aij ) : aij ∈ F, i = 1, . . . , m, j = 1, . . . , n} .

Then Mm,n (F) forms a vector space over F under the following matrix operations:

(aij ) + (bij ) = (aij + bij ), (1.2.4)


α(aij ) = (αaij ) (1.2.5)

for every matrices (aij ), (bij ) ∈ Mm,n (F). Note that operation (1.2.4) is called matrix addition and operation
(1.2.5) is called scalar multiplication. Recall that two matrices A = (aij ) and B = (bij ) in Mm,n (F) are
equal if and only if
aij = bij for all i = 1, . . . , m, j = 1, . . . , n. (1.2.6)

We now proceed to check (Mm,n (F), + , · ) forms a vector space over F.


Let A = (aij ), B = (bij ) and C = (cij ) be matrices in Mm,n (F), and α, β ∈ F.
(A1): Note that aij , bij ∈ F implies that aij + bij ∈ F for all i = 1, . . . , m and j = 1, . . . , n (in what
follows, we abbreviate it by ‘for all i, j’). By (1.2.4), we see that (aij ) + (bij ) = (aij + bij ) ∈ Mm,n (F). Then
A + B ∈ Mm,n (F).
(A2): Note that aij + bij = bij + aij for all i, j. So A + B = (aij + bij ) = (bij + aij ) = B + A by (1.2.6).
(A3): Note that aij + bij is the (i, j)-th entry of A + B by (1.2.4). So (aij + bij ) + cij is the (i, j)-th
entry of (A + B) + C. On the other hand, aij + (bij + cij ) is the (i, j)-th entry of A + (B + C). Since
aij + (bij + cij ) = aij + bij + cij = (aij + bij ) + cij for all i, j, we have A + (B + C) = (A + B) + C by (1.2.6).
(A4): Let 0m,n ∈ Mm,n (F) be the matrix such that all entries are zero scalars. Since aij +0 = aij = 0+aij
for all i, j, we obtain A + 0m,n = A = 0m,n + A. Usually, 0m,n is called the zero matrix of Mm,n (F).
(A5): Let −A = (−aij ) ∈ Mm,n (F). Since aij + −aij = aij − aij = 0 = −aij + aij for all i, j, it follows
that A + −A = 0m,n = −A + A. Usually, −A ∈ Mm,n (F) is called the negative matrix of A.
(M1): Since αaij ∈ F for every i, j, it follows from (1.2.5) that α(aij ) = (αaij ) ∈ Mm,n (F).
(M2): Since aij + bij is the (i, j)-th entry of A + B by (1.2.4), we get α(aij + bij ) is the (i, j)-th entry of
α(A + B) by (1.2.5). On the other hand, αaij and αbij are the (i, j)-th entry of αA and αB, respectively.
1.2. VECTOR SPACES 17

So αaij + αbij is the (i, j)-th entry of αA + αB by (1.2.4) and (1.2.5). Since α(aij + bij ) = αaij + αbij for
all i, j, it follows from (1.2.6) that α(A + B) = αA + αB.
(M3): By (1.2.5), we have (α + β)A = ((α + β)aij ). By (1.2.4) and (1.2.5), we see that αA + βA =
(αaij + βbij ). Since (α + β)aij = αaij + βaij for all i, j, it follows from (1.2.6) that (α + β)A = ((α + β)aij ) =
(αaij + βaij ) = αA + βA.
(M4): The (i, j)-th entry of (αβ)A is (αβ)aij by (1.2.5). Also, βaij is the (i, j)-th entries of βA, and so
α(βaij ) is the (i, j)-th entries of α(βA) by (1.2.5). Since (αβ)aij = α(βaij ) for all i, j. By (1.2.6), we have
(αβ)A = α(βA).
(M5): The (i, j)-th entry of 1A is 1aij by (1.2.5). Since 1aij = aij for all i, j, we get 1A = A by (1.2.6).

Consequently, (Mm,n (F), + , · ) forms a vector space over F by the matrix addition and scalar multiplication
defined in (1.2.4) and (1.2.5). We remark that, using the matrix addition defined in (1.2.4), the subtraction
of matrices is defined by
A − B = A + (−B) = (aij − bij )
for every A = (aij ), B = (bij ) ∈ Mm,n (F). when m = n, the n-square matrix space Mn,n (F) (Mn (F) for
short) has the identity In called the identity matrix of order n. Moreover, if m = 1 (respectively, n = 1),
then M1,n (F) (respectively, Mm,1 (F)) consists of all row vectors (respectively, column vectors) and is
called the n-tuple row space (respectively, n-tuple column space). The elements of Fn are identified
with the elements of M1,n (F) (or sometimes with the elements of Mn,1 (F)).

Example 1.2.6 Let V = {x} be a singleton set and let F be a field. The vector addition and the scalar
multiplication on V are defined by
x + x = x and αx = x
for x ∈ V and α ∈ F. It is easy to show that V forms a vector space over F, and it is called the trivial
vector space. The element x is usually denoted by the zero vector 0.

Example 1.2.7 (Vector space of real functions) Let RR denote the totality of real functions f : R → R.
Let f, g ∈ RR and c ∈ R. Define the addition and scalar multiplication operations as follows:

(f + g) (x) = f (x) + g(x) for all x ∈ R,


(cf ) (x) = c (f (x)) for all x ∈ R.

Then RR is a vector space over R (refer the detailed proof in [12] or verify yourself as an exercise).
As a side remark, the zero function 0 : R → R is defined by

0(x) = 0 for all x ∈ R.

Then 0 + f = f + 0 = f for all f ∈ RR , and so 0 is the zero vector in RR . For each f ∈ RR , we define the
function −f : R → R by
(−f )(x) = −f (x) for all x ∈ R.
It can be verified that (−f ) + f = f + (−f ) = 0.

Example 1.2.8 (Vector space of real polynomials) Let R[x] denote the set of all real polynomials, i.e.,
R[x] = { p(x) = a0 + a1 x + . . . + an xn : n ∈ Z+ ∪ {0}, ai ∈ R, i = 1, . . . , n }, where Z+ is the set of positive
integers. For any p, q ∈ R[x] and c ∈ R, we define two operations as follows:

(p + q) (x) = p(x) + q(x) for all x ∈ R,


(cp) (x) = c (p(x)) for all x ∈ R.

Then R[x] is a vector space over R (Exercise).


18 CHAPTER 1. VECTOR SPACES

Example 1.2.9 (Vector space of the solution set of a homogeneous linear system) Let A ∈ Mm,n (R). Let
V be the set of all solutions of the homogeneous linear system

AX = 0 with X = (xi ) ∈ Mn,1 (R)

of m linear equations in n variables with real coefficients, i.e., V = {X = (xi ) ∈ Mn,1 (R) : AX = 0 }. Then
V is a vector space under the usual matrix operations in Mn,1 (R) as defined in Example 1.2.5 (Exercise).

Example 1.2.10 Let B be a nonempty set and V be a vector space over a field F. A function ψ : B → V
is said to be of finite support provided that ψ(b) = 0 for all b ∈ B except possibly for finitely many b ∈ B.
Let (V B )0 denote the totality of all functions ψ : B → V with finite support. Then (V B )0 constitutes
a vector space over F with respect to the usual operations of addition functions and multiplication by a
constant defined as follows:
(ψ + ϕ)(b) = ψ(b) + ϕ(b) for all b ∈ B
(αψ)(b) = α(ψ(b)) for all α ∈ F and b ∈ B.
(Exercise).

For more examples of vector space, see, for example, [1, 3, 5, 6, 7, 8]. We shall end this section by proving
some straightforward by important consequences of the definition of vector space.

Theorem 1.2.11 Let V be a vector space over a field F. Let u, v ∈ V and α, β ∈ F. Then the following
assertions hold true.
(i) 0u = 0.
(ii) α0 = 0.
(iii) (−1)u = −u.

(iv) (−α)u = −(αu) = α(−u).


(v) −(−u) = u.
(vi) −(u + v) = −u − v.
(vii) α(u − v) = αu − αv.
(viii) (α − β)v = αv − βv.

Proof. (i) By (M3), we have 0u + 0u = (0 + 0)u = 0u. It follows from (A5), (A3) and (A4) that

0 = 0u + (−0u) = (0u + 0u) + (−0u) = 0u + (0u + (−0u)) = 0u + 0 = 0u.

Hence 0u = 0.
(ii) By (A4) and (M2), we have α0 = α(0 + 0) = α0 + α0. So

α0 = α0 + 0 = α0 + (α0 + (−α0)) = (α0 + α0) + (−α0) = α0 + (−α0) = 0

by (A4), (A5) and (A3).


(iii) Note that (1 + (−1))u = 0u = 0 by (i). It follows from (A4), (M3), (M5), (A3) and (A5) that −u =
−u+0 = −u+(1+(−1))u = −u+(1u+(−1)u) = −u+(u+(−1)u) = (−u+u)+(−1)u = 0+(−1)u = (−1)u.
(iv) In F, we note that −α = (−1)α. By (M4) and (iii), (−α)u = ((−1)α)u = (−1)(αu) = −(αu). By
the commutativity of multiplication in F, we have α(−1) = (−1)α. So α(−u) = α((−1)u) = (α(−1))u =
((−1)α)u = (−1)(αu) = −(αu) by (M4) and (iii).
(v), (vi), (vii) and (viii) left as exercise. □
1.3. SUBSPACES 19

Theorem 1.2.12 Let V be a vector space over a field F. Let u, v, w ∈ V and α ∈ F. Then the following
assertions hold true.
(i) If u + v = w, then u = w − v.
(ii) If u + v = 0, then u = −v.
(iii) If u + w = v + w, then u = v (cancellation law for vector addition).
(iv) If αv = 0, then either α = 0 or v = 0.

Proof. (i) Adding −v to both sides of u+v = w, we have (u+v)+(−v) = w+(−v). Thus u+(v+(−v)) = w−v
by (A3) and (1.2.1). It follows from (A5) and (A4) that u = w − v.
(ii) Letting w = 0 in (i), we obtain u = 0 − v = 0 + (−v) = −v.
(iii) Adding −w to both sides of u + w = v + w, we get (u + w) + (−w) = (v + w) + (−w). By the
associativity of addition, u + (w + (−w)) = v + (w + (−w)). So u + 0 = v + 0 by (A5). Hence u = v by (A4).
(iv) It suffices to show that if αv = 0 and α ̸= 0, then v = 0. To see this, since α−1 α = 1, together with
(M5), we have v = 1v = (α−1 α)v = α−1 (αv) = α−1 0 = 0 as desired. This completes our proof. □

1.3 Subspaces

It is possible for one vector space to be contained within another vector space. For example, all lines in the
Euclidean 2-space R2 through the origin (0, 0) are vector spaces that are contained within the larger vector
space R2 . In this section, the concept of subsystems of vector spaces will be introduced.

Definition 1.3.1 Let V be a vector space over a field F. A subset S is called a subspace of V, denoted
by S ⩽ V, if S itself is a vector space over F with respect to the operations of vector addition and scalar
multiplication defined in V.

If we go through the axioms of a vector space V carefully, it is not difficult to notice that a subset S of
V is a subspace if the following four conditions hold true:

(i) S is closed under addition, i.e., u + v ∈ S whenever u, v ∈ S;


(ii) S is closed under scalar multiplication, i.e., αu ∈ S whenever u ∈ S and α ∈ F;
(iii) the zero vector of V is also an element in S (and therefore S is nonempty);
(iv) each element s of S has the additive inverse −s in S.

We must also point out that the commutativity and the associativity of vector addition (i.e., (A2) and (A3)),
and the axioms (M2), (M3), (M4) and (M5) of scalar multiplication in Definition 1.2.1 do not need to be
verified since these are properties of the operations inherited from V. One can simplify things still further,
and we have the following easy criterion for a subset of a vector space to be a subspace.

Theorem 1.3.2 (Subspace Criterion) Let V be a vector space over a field F and S be a subset of V. Then
the following are equivalent:
(i) S is a subspace of V;
(ii) S contains the zero vector of V, and S is closed under vector addition and scalar multiplication of V,
i.e., u + v ∈ S and αu ∈ S whenever u, v ∈ S and α ∈ F.
20 CHAPTER 1. VECTOR SPACES

Proof. (i) ⇒ (ii) Since S is a vector space under the operations of vector addition and scalar multiplication
defined in V, it is clear that S is closed under vector addition and scalar multiplication by (A1) and (M1).
We now claim that the zero vector 0 of V is contained in S. Since S is a vector space, it follows from (A4)
that there is an element 0′ ∈ S such that s + 0′ = s for every s ∈ S. Clearly, 0′ ∈ V, and 0′ + s = s = 0 + s
for any s ∈ S. By the cancellation law for vector addition, we conclude that 0′ = 0. Hence 0 ∈ S.
(ii) ⇒ (i) By (ii), we note that S is nonempty, since it contains the zero vector 0 of V, and (A1) and
(M1) are also satisfied. We next see that vectors in S satisfy (A2), (A3) and (M2) − (M5). This is because
S ⊆ V, and hence vectors in S are in V. So they automatically satisfy (A2), (A3) and (M2) − (M5) as well.
We next claim (A4) and (A5) hold in S. Let s ∈ S. Then s + 0 ∈ S since 0, s ∈ S by (A1). Moreover,
since 0 is the zero vector of V, s + 0 = s. By (A2), s + 0 = 0 + s. It follows that s + 0 = s = 0 + s for every
s ∈ S, so (A4) is satisfied. We proceed to claim (A5). Firstly, in V, we see that −s = (−1)s by Theorem
1.2.11 (iii), and (−s) + s = 0 = s + (−s). Since (−1)s ∈ S by (M1), it follows that −s ∈ S for any s ∈ S.
Consequently, (A5) holds true in S.
Consequently, S is a vector space over F under the operations of vector addition and scalar multiplication
defined in V. This completes our proof. □

Some author prefers to use the property (ii) of Theorem 1.3.2 to define subspaces (see [5], for example).
Indeed, Theorem 1.3.2 provides us a simple method for determining whether or not a given subset of a vector
space is a subspace. As side remark, it is not difficult to verify that if S is a nonempty subset of a vector
space V over a field F, then
u + v, αu ∈ S ⇔ αu + v ∈ S
for every u, v ∈ S and α ∈ F (verify it).

Example 1.3.3
(i) Let V be a vector space. It can be proved that V and the subset {0}, the singleton consists only the
zero vector 0 of V, are always subspaces of V (verify it). These are called trivial subspaces. The
subspace {0} is also called the zero subspace of V. A subspace S is said to be nonzero if S ̸= {0},
and called a proper subspace of V if S is nonzero and S = ̸ V.
(ii) In the Euclidean 2-space R2 , for each m ∈ R, the set of 2-tuples (x, y) such that y = mx, i.e.,
Sm = {(x, y) : x, y ∈ R with y = mx} = {(x, mx) : x ∈ R}
2
is a subspace of R over R. Clearly, for any fixed m, Sm ̸= ∅ since (0, 0) ∈ Sm . Let u = (x1 , y1 )
and v = (x2 , y2 ) ∈ Sm and α ∈ R. Then αy1 + y2 = α(mx1 ) + mx2 = m(αx1 + x2 ), and so
αu + v = (αx1 + x2 , αy1 + y2 ) = (αx1 + x2 , m(αx1 + x2 )) ∈ Sm . Hence Sm is a subspace of R2 .
(iii) Let n be a positive integer and Rn [x] be the set of all real polynomials of degree ⩽ n. Then Rn [x] is a
vector space under the operations of usual addition of real functions and scalar multiplication (verify
it). So Rn [x] is a subspace of R[x]. Note also that R[x] is a subspace of RR .
(iv) Let n be a positive integer. In the matrix space Mn (F) over a field F, the set M0n of n-square matrices
A = (aij ) ∈ Mn (F) such that tr A = 0, where trA stands for the trace of A, is a subspace of Mn (F).
We first note that 0n ∈ M0n because tr 0n = 0. Let A, B ∈ M0n and α ∈ F. Then tr A = 0 and tr B = 0.
Then
tr (A + B) = tr A + tr B = 0 + 0 = 0 and tr (αA) = α(tr A) = α0 = 0.
Hence M0n is closed under matrix addition and scalar multiplication of Mn (F). So M0n is a subspace
of Mn (F) by Theorem 1.3.2.
(v) Let a, b, c ∈ R, and let S = {(x, y, z) ∈ R3 : ax + by + cz = 0 }. We first see that S ̸= ∅ since
(0, 0, 0) ∈ S. Let u = (u1 , u2 , u3 ), v = (v1 , v2 , v3 ) ∈ S and α ∈ R. Then au1 + bu2 + cu3 = 0 and
av1 + bv2 + cv3 = 0. Note that αu + v = (αu1 + v1 , αu2 + v2 , αu3 + v3 ), and

a(αu1 + v1 ) + b(αu2 + v2 ) + c(αu3 + v3 ) = α(au1 + bu2 + cu3 ) + (av1 + bv2 + cv3 ) = 0.

Thus αu + v ∈ S. By Theorem 1.3.2, S is a subspace of R3 .


1.4. NULLSPACES 21

(vi) Let n be a positive integer and C n (R) denote the set of all real functions f : R → R which have
continuous n-th derivatives. It can be shown that C n (R) is a subspace of RR . Let S be the set of all
real functions f in C 2 (R) such that f ′′ (x) + f (x) = 0 for all x ∈ R. Then S is a subspace of C 2 (R).
(vii) Let A = (aij ) ∈ Mm,n (R). The solution space defined in Example 1.2.9 is a subspace of Mn,1 (R).
(viii) For more examples, we refer the reader to [3].

Example 1.3.4 We now see some example of subsets which are not subspace.
(i) Let S = {(x, y) : y = x + 1 }. We note that S is not a subspace of R2 . This is because (0, 0) ∈
/ S.
y↑

→ x
−1

Figure 1.3.1
(ii) Let   
a b
S= a, b, c, d ∈ R and a ⩾ 0 .
c d

We see that S is not a subspace of M2 (R) because


     
1 0 1 0 −1 0
∈ S but (−1) = ∈
/ S.
0 0 0 0 0 0
(iii) Let F be a field. Let A ∈ Mm,n (F) and B ∈ Mm,1 (F) be nonzero matrices. Then the solution set
of the nonhomogeneous linear system AX = B, i.e., {X ∈ Mn,1 (F) : AX = B} is not a subspace of
Mn,1 (F). This is because the zero column vector 0n,1 is not a solution of the linear system.
(iv) For more examples, we refer the reader to [3].

1.4 Nullspaces

In this section we will study nullspaces associated with matrices. Certain subspaces are a rich source of
information about a behavior of a matrix. For a given matrix A, there are three useful vector spaces
associated with A, namely, the nullspace, the row space and the column space of A. We shall first explore
the properties of nullspaces. Row spaces and column spaces will be introduced once the required tool has
been established.

Definition 1.4.1 Let F be a field and m, n be positive integers. The nullspace of a matrix A ∈ Mm,n (F),
denoted N (A), is defined by
N (A) = { X ∈ Mn,1 (F) : AX = 0 } .

In the language of linear systems, N (A) consists of all solutions X in Mn,1 (F) of the homogeneous linear
system AX = 0, that means, N (A) is the solution set of the homogeneous linear system AX = 0. Indeed,
in the following result, we show that N (A) is a subspace of Mn,1 (F).
22 CHAPTER 1. VECTOR SPACES

Theorem 1.4.2 Let A ∈ Mm,n (F). Then N (A) is a subspace of Mn,1 (F).

Proof. Since A0n,1 = 0m,1 = 0, it follows that 0n,1 ∈ N (A). Let X1 , X2 ∈ N (A) and α ∈ F. Then
A(X1 + X2 ) = AX1 + AX2 = 0 + 0 = 0 and A(αX1 ) = α(AX1 ) = α0 = 0. Thus X1 + X2 and αX1 are in
N (A), and so N (A) is a subspace of Mn,1 (F) as desired. □

Remark 1.4.3
(i) If the linear system AX = B is not homogeneous, with A ∈ Mm,n (F) and 0 ̸= B ∈ Mm,1 (F), then
the solution set is never a subspace of Mn,1 (F). This is because 0 ∈ Mn,1 (F) is not a solution for the
non-homogeneous linear system AX = B.
(ii) Let A ∈ Mn (F). If A is invertible, then N (A) = {0}. This is because the system AX = 0 has only the
trivial solution when A is invertible. Does the converse hold true ? (see [5, Theorem 8])

r
Recall that two matrices of the same size A, B ∈ Mm,n (F) are row equivalent, denoted A ∼ B, if
there exists a finite sequence of elementary row operations that transforms one matrix into the other (see [5,
Chapter 1]). Also, if two matrices A and B are row equivalent, then the linear systems AX = 0 and BX = 0
r
have the same solution set (see [5, page 24]). It follows that A ∼ B infers N (A) = N (B). In particular, if
E is a row echelon form (or the reduced row echelon form) of A, then N (E) = N (A). By this observation,
solving a system of linear equations AX = 0, we may consider to reduce the matrix A to a row echelon form
E via Gaussian elimination (or to the reduced row echelon form via Gauss-Jordan elimination) and solve for
the linear system EX = 0.

 
1 1 1 0
Example 1.4.4 Determine N (A) if A = ∈ M2,4 (R).
2 1 0 1
Solution: We note that
  R2 → R2 − 2R1 ,  
R1 → R1 + R2
1 1 1 0 0 1 0 −1 1 0
−→
2 1 0 1 0 0 −1 −2 1 0
R2 → −R2
 
1 0 −1 1 0
−→
0 1 2 −1 0
which is the reduced row echelon form. Then
x1 − x3 + x4 = 0 ⇒ x1 = x3 − x4 ,
x2 + 2x3 − x4 = 0 ⇒ x2 = −2x3 + x4 .

There are two free variables x3 , x4 . Hence,


  

 s−t


−2s +t
  
N (A) =   s, t ∈ R .


 s 


t
 

Note that every vector in N (A) is of the form


   
1 −1
 −2   1 
s
 1  + t 0
  
 where s, t ∈ R.
0 1
1.5. INTERSECTION OF SUBSPACES 23

The fact that the nullspace of a real matrix allows us to prove the solutions of linear systems: they have
either no solution, a unique solution, or infinitely many solutions. More precisely, we have
Example 1.4.5 Let F be an infinite field and let A ∈ Mm,n (F). For any system of linear equation
AX = B, exactly one of the following is true.
(i) There is no solution.
(ii) There is a unique solution.
(iii) There are infinitely many solutions.

Solution: If the linear system AX = B has either no solutions or exactly one solution, then we are done.
Assume that there are at least two distinct solutions of AX = B, let say X1 , X2 ∈ Mn,1 (F). Therefore

AX1 = B and AX2 = B


with X1 ̸= X2 . It follows that A(X1 − X2 ) = B − B = 0. Letting X0 = X1 − X2 . Then X0 ̸= 0 and AX0 = 0.
Hence, the nullspace of A is nontrivial. Since N (A) is closed under scalar multiplication, αX0 ∈ N (A) for
every α ∈ F. Consequently, N (A) contains infinitely many vectors. We now consider the vectors of the form
X1 + αX0 with α ∈ F. Then we have
A(X1 + αX0 ) = AX1 + αAX0 = B + 0 = B.
Therefore there are infinitely many solutions of the linear system AX = B.

1.5 Intersection of Subspaces

Let V be a vector space, and let S and T be subspaces of V. The intersection of S and T , denoted S ∩ T ,
is defined by
S ∩ T = { v ∈ V : v ∈ S and v ∈ T } .
We will show that S ∩ T is the largest subspace of V that is contained in both S and T . More precisely, we
have the following result.

Theorem 1.5.1 Let V be a vector space, and let S and T be subspaces of V. Then
(i) S ∩ T is a subspace of V contained in S and T ;
(ii) if U is a subspace of V contained in S and T , then U is a subspace of S ∩ T .

Proof. (i) We first claim that S ∩ T is a subspace of V. Clearly, 0 ∈ S ∩ T since S and T are subspaces of
V. Let u, v ∈ S ∩ T and α ∈ F. Then u, v ∈ S and u, v ∈ T . Since S and T are subspaces, it follows that
u + v ∈ S, αu ∈ S, and u + v ∈ T , αu ∈ T . Therefore, u + v ∈ S ∩ T and αu ∈ S ∩ T . By Theorem 1.3.2,
S ∩ T is a subspace of V.
(ii) Since U is contained in S and T , and so U ⊆ S ∩ T . Moreover, U and S ∩ T are subspaces of V
implies that U is a subspace of S ∩ T . This is because U is a vector space with respective to the operations
of vector addition and scalar multiplication defined in V, and so in S ∩ T . We are done. □

Example 1.5.2 Let U and W be two planes in R3 , where


U = {(x, y, 0) : x, y ∈ R} and W = {(0, y, z) : y, z ∈ R}.

Clearly, U and W are respectively the xy-plane and the yz-plane of R3 . We now show that U and W are
subspaces of R3 . It is clear that U is nonempty as (0, 0, 0) ∈ U. Let u, v ∈ U and α ∈ R. So, u = (x1 , y1 , 0)
and v = (x2 , y2 , 0) for some x1 , x2 , y1 , y2 ∈ R. Then

αu + v = (αx1 , αy1 , 0) + (x2 , y2 , 0) = (αx1 + x2 , αy1 + y2 , 0)


24 CHAPTER 1. VECTOR SPACES

So αu + v ∈ U. Hence, U is a subspace of R3 . Likewise, we can show that W is a subspace of R3 . Evidently,


{(0, y, 0) : y ∈ R} ⊆ U ∩ W. Let w = (u1 , u2 , u3 ) ∈ U ∩ W. Then u1 = 0 since w ∈ W, and u3 = 0 as w ∈ U.
Therefore, w = (0, u2 , 0) ∈ {(0, y, 0) : y ∈ R}. Hence, U ∩ W = {(0, y, 0) : y ∈ R}, which is the y-axis of R3 ,
is a subspace of R3 .

Example 1.5.3 Let


S = {(x, y, z) ∈ R3 : x + 2y + 3z = 0},

T = {(x, y, z) ∈ R3 : 2x + 3y + 4z = 0}.

It can be shown that S and T are subspaces of R3 . We note (geometric interpretation of S and T ) that S
and T are plans in R3 with the equations x + 2y + 3z = 0 and 2x + 3y + 4z = 0, respectively. So, S ∩ T is
the intersection of planes S and T , which is the line L: x = −y/2 = z. To see this, in view of S and T , we
have
x + 2y + 3z = 0, (1.5.1)

2x + 3y + 4z = 0. (1.5.2)

2(1.5.1)-(1.5.2) yields y = −2z. Substituting into (1.5.1), we get z = x. Therefore,

S ∩ T = {(x, −2x, x) : x ∈ R}.

Figure 1.5.1

1.6 Sums and Direct Sums

Definition 1.6.1 Let V be a vector space over a field F, and let S and T be subspaces of V. Then the
sum of S and T , denoted by S + T , is defined by

S + T = { u + v : u ∈ S and v ∈ T }.

Example 1.6.2 Let L1 = {(x, 0) : x ∈ R} and L2 = {(0, y) : y ∈ R}. We see that

L1 + L2 = { (x, 0) + (0, y) : (x, 0) ∈ L1 and (0, y) ∈ L2 } = {(x, y) : x, y ∈ R} = R2 .

So, L1 + L2 = R2 .
1.6. SUMS AND DIRECT SUMS 25

Example 1.6.3 Let L1 = {(x, y) ∈ R2 : y = 2x} and L2 = {(x, y) ∈ R2 : y = −4x}. Clearly, L1 and L2
2
are subspaces of R , and they are straight lines passing through the origin (0, 0). What is L1 + L2 ?
Solution: Let (x, y) ∈ R2 . Suppose that (x, y) = (a, 2a) + (b, −4b) for some a, b ∈ R. Thus x = a + b and
y = 2a − 4b, and so
  R2 → R2 − 2R1
 
1 1 x 1 1 x
−→
2 −4 y 0 −6 y − 2x
R1 → 6R1       
R1 → R1 + R2
6 0 y + 4x 6 0 a y + 4x
−→ ⇒ =
0 −6 y − 2x 0 −6 b y − 2x.
Therefore, it has a unique solution
y + 4x 2x − y
a= and b = .
6 6

Consequently, (x, y) = ( y+4x y+4x 2x−y 2y−4x


6 , 3 )+( 6 , 3 ) = ( y+4x y+4x 2x−y 2x−y
6 , 2( 6 )) + ( 6 , −4( 6 )) ∈ L1 + L2 . So,
R2 ⊆ L1 + L2 , and hence, L1 + L2 = R2 .

In the following result, we prove that the sum of two subspaces of a vector space V is a subspace of V.

Theorem 1.6.4 Let V be a vector space over a field F, and let S and T be subspaces of V. Then S + T
is a subspace of V. Furthermore, S + T is the smallest subspace of V containing S and T .

Proof. Since S and T are subspaces of V, we have 0 ∈ S and 0 ∈ T . Thus, 0 = 0 + 0 ∈ S + T , and so,
S + T is nonempty. Let α ∈ F and u, v ∈ S + T . Then u = s1 + t1 and v = s2 + t2 for some s1 , s2 ∈ S
and t1 , t2 ∈ T . Then s1 + αs2 ∈ S and t1 + αt2 ∈ T . Consequently, u + αv = (s1 + t1 ) + α(s2 + t2 ) =
(s1 + αs2 ) + (t1 + αt2 ) ∈ S + T . Hence, S + T is a subspace of V.
We now claim that S ⊆ S + T . Let s be an arbitrary element in S. Thus s = s + 0 ∈ S + T since 0 ∈ T ,
so S ⊆ S + T . Similarly, T ⊆ S + T . Suppose W is a subspace of V such that S ⊆ W and T ⊆ W. Let
u ∈ S + T . Then u = s + t for some s ∈ S and t ∈ T . Since W ⩽ V, it follows that u = s + t ∈ W. Hence
S + T ⊆ W. This completes our proof. □

Figure 1.6.1

Remark 1.6.5
(i) Let L1 and L2 be two distinct straight lines in R3 passing through the origin (0, 0, 0). Then it is easy
to check that L1 and L2 are subspace of R3 , and L1 + L2 is the plane determined by L1 and L2 which
contains the lines L1 and L2 .
(ii) Let S, T be subspaces of a vector space V. Then S + T = T + S.
26 CHAPTER 1. VECTOR SPACES

Definition 1.6.6 Let S and T be subspaces of a vector space V. The sum S + T is said to be direct if
S ∩ T = {0}. In this case, we write S ⊕ T . Furthermore, we say that V has the direct sum decomposition
relative to S and T , or simply, V is the direct sum of S and T , denoted V = S ⊕ T , if V = S + T and
S ∩ T = {0}. In this case, T is called a complement subspace of S in V.

Example 1.6.7
(i) Let L = {(0, 0, z) : z ∈ R } and P = {(x, y, 0) : x, y ∈ R }. Then L is the z-axis and P is the xy-plane.
Clearly, L and P are subspaces of R3 , and R3 = L + P with L ∩ P = {0}. Therefore, R3 = L ⊕ P .
(ii) In view of Example 1.6.3, we see that L1 and L2 are subspaces of R2 . Let (x, y) ∈ L1 ∩ L2 . Then
2x = y = −4x, and so, 6x = 0. Therefore, x = 0 and y = 0, so L1 ∩ L2 = {0}. Since R2 = L1 + L2 , it
follows that R2 = L1 ⊕ L2 .
(iii) Let L = {(z, z, z) : z ∈ R } and P = {(x, y, 0) : x, y ∈ R }. Then L is a straight line passing through
the origin and P is the xy-plane. Is R3 = L ⊕ P ?
Solution : We first claim that L + P = R3 . Clearly, L + R ⊆ R3 . Let (a, b, c) ∈ R3 . Suppose that
(a, b, c) = (z, z, z) + (x, y, 0) for some x, y, z ∈ R. Then a = x + z, b = y + z and c = z. Thus x = a − c,
y = b − c and z = c. So (a, b, c) = (c, c, c) + (a − c, b − c, 0) ∈ L + P . We thus obtain R3 ⊆ L + P .
Since L + R ⊆ R3 , we have R3 = L + R. Next, we show that L ∩ P = {0}. Let (a, b, c) ∈ L ∩ P . Then
(a, b, c) ∈ L and (a, b, c) ∈ P . So, a = b = c and c = 0. Thus, (a, b, c) = (0, 0, 0), and so, L ∩ P = {0}.
Hence R3 = L ⊕ R.
(iv) More examples on sums and direct sums of subspaces can be found, for example, in [3, 10].

Remark 1.6.8 In general, if P is a plane in R3 passing through the origin, L is a straight line passing
through the origin and L ⊈ P , then R3 = L ⊕ P . We recall that the line L and the plane P in R3 can be
represented by:
L = {(x, y, z) : x = a1 t, y = b1 t, z = c1 t for t ∈ R} a1 , b1 , c1 ∈ R,
Remark 1.6.9
P = {(x, y, z) ∈ R3 : a2 x + b2 y + c2 z = 0} a2 , b2 , c2 ∈ R.
In the following table, we list all subspaces of R2 and R3 .

Subspaces of R2 Subspaces of R3

• {0} • {0}
• Straight lines passing through the origin • Straight lines passing through the origin
• R2 • Planes passing through the origin
• R3

Here are some important subclasses (subspaces) of the real function spaces. We recall that RR , C(R), C 1 (R),
C n (R), R[x] and Rn [x] denote the totalities of real functions, real continuous functions, real differentiable
functions, real n-differentiable functions, real polynomials and real polynomials of degree n, respectively.

Problem 1.6.10
Let V be a vector space over a field F. Let S be a subset of V. Show that the following statements are
equivalent:
(i) S contains the zero vector of V, and S is closed under vector addition (i.e., u+v ∈ S whenever u, v ∈ S),
and S is closed under scalar multiplication of V (i.e., αu ∈ S whenever u ∈ S and α ∈ F).
(ii) S contains the zero vector of V, and αu + v ∈ S whenever u, v ∈ S and α ∈ F.
1.6. SUMS AND DIRECT SUMS 27

Solution (i) ⇒ (ii) Let u, v ∈ S and α ∈ F. Then αu ∈ S, and so αu + v ∈ S.


(ii) ⇒ (i) Let u, v ∈ S and α ∈ F. Note that u = 1u (by M5). Then u + v = 1u + v ∈ S. Also, we see
that αu = αu + 0 ∈ S. We are done.
Consequently, in view of Subspace Criterion theorem, we see that a subset S of V is a subspace if and
only if S contains the zero vector of V, and αu + v ∈ S whenever u, v ∈ S and α ∈ F.

Problem 1.6.11
Let P1 = {(x, y, z) ∈ R3 : x + 2y − z = 0} and P2 = {(x, y, z) ∈ R3 : 2x − y + 4z = 0} be two planes in R3 .
Is it true that
(i) P1 + P2 = R3 ,
(ii) P1 ∩ P2 = {0},
(iii) R3 = P1 ⊕ P2 .

Solution (i) Notice that P1 = {(x, y, x + 2y) : x, y ∈ R} and P2 = {(s, 2s + 4t, t) : s, t ∈ R}. Firstly,
P1 + P2 ⊆ R3 . Let (a, b, c) ∈ R3 . Suppose that (a, b, c) = (x, y, x + 2y) + (s, 2s + 4t, t) for some x, y, s, t ∈ R.
Then x + s = a, y + 2s + 4t = b and x + 2y + t = c. We obtain
   
1 0 1 0 a 1 0 1 0 a
r
 0 1 2 4 b  ∼  0 1 2 4 b .
1 2 0 1 c 0 0 5 7 a + 2b − c

a+2b−c 4a−2b+c
Hence x + s = a, y + 2s + 4t = b and 5s + 7t = a + 2b − c. We choose t = 0. Then s = 5 , x= 5
−2a+b+2c
and y = 5 . Consequently, we have
   
4a − 2b + c −2a + b + 2c a + 2b − c 2a + 4b − 2c
(a, b, c) = , ,c + , , 0 ∈ P1 + P2 .
5 5 5 5

Hence P1 + P2 = R3 .
28 CHAPTER 1. VECTOR SPACES

(ii) Let (a, b, c) ∈ P1 ∩ P2 . Then (a, b, c) ∈ P1 yields c = a + 2b, and (a, b, c) ∈ P2 leads to b = 2a + 4c. So
6a + 7b = 0. We choose a = 7, so b = −6, and thus c = −5. Then (7, −6, −5) ∈ P1 ∩ P2 (WHY?). Therefore
P1 ∩ P2 ̸= {0}.
(iii) P1 + P2 = R3 but the sum is not direct.

You might also like