1 Preliminaries

The linear spaces that we consider are real or complex vector spaces, that is, vector spaces over
the fields R or C, and throughout we use the symbol K to denote R or C. Sometimes we have to
distinguish between real and complex vector spaces, and in that case we always explicitly do so.
In such cases, it’s a good idea to make sure you understand whether we use special properties of
C, for example the fact that we can solve any polynomial.

1.1 Vector spaces

Let’s first recall some basic facts abut vector spaces and linear maps between them which you first
encountered in a Linear Algebra module.
Definition 1 (Vector space). A vector space V over the field K is a set V together with the operations of
vector addition, ` : V ˆ V Ñ V , and scalar multiplication, ˚ : K ˆ V Ñ V , such that:
• There exists an element 0 P V , the zero vector, such that x ` 0 “ x for every x P V .
• If 1 P K is the multiplicative identity, then 1 ˚ x “ x for every x P V .
• For every x P V , there exists an additive inverse ´x P V such that x ` p´xq “ 0.
• Vector addition is commutative, i.e. x ` y “ y ` x for every x, y P V .
• Vector addition is associative, i.e. px ` yq ` z “ x ` py ` zq for every x, y, z P V .
• For any vector x P V and any two α, β P K, we have α ˚ pβ ˚ xq “ pα ¨K βq ˚ x, where ¨K denotes
multiplication in the field K.
• Scalar multiplication distributes over vector addition, i.e. α˚px`yq “ α˚x`α˚y and pα`K βq˚x “
αx ` βx for every x, y P V and α, β P K, where `K denotes addition in the field K.
To keep things easy to read we will omit the symbol ˚, so we write αx instead of α˚x. Similarly,
we omit ¨K and the subscript in `K , e.g. αβ instead of α ¨K β, and α ` β instead of α `K β. We are
therefore using the symbol ` for summation and no symbol for multiplication in both V and K,
but it will always be clear from the context which one we mean; moreover, the axioms ensure that
it doesn’t really matter whether we are doing operations in the field or applying them to vectors.
Definition 2 (Linear subspace). Let V be a vector space over K and W Ď V be non-empty. Then W is a
linear subspace of V if it closed under the operation ` and scalar multiplication, i.e. for any x, y P W and
α P K, then it follows that x ` y P W and αx P W . In other words, a subspace is a subset of a vector space
that is itself a vector space using the same operations.
Recall from Linear Algebra that to show that a given subset W Ď V is a linear subspace of V ,
it is enough to verify that x ` αy P W for every x, y P W and α P K.
Definition 3 (Linear span). Let V be a vector space over K and E Ď V . The linear span of E, SpanpEq,
is the collection of all finite linear combinations of elements from E, i.e.
# +
ď ÿ n
SpanpEq “ αi vi such that vi P E and αi P K for all i “ 1, . . . , n .
nPN i“1

4
 If V “ SpanpEq, then weř say that the set E spans V . In other words, E spans V if every v P V
n
can be written as a sum v “ i“1 αi vi for some n P N, vi P E and αi P K. This representation need
not be unique. Note also that SpanpEq is a linear subspace of V : this is straightforward to check
from the appropriate definitions.
One other thing to note is that the vector space axioms in Definition 1 only allow us to talk
about finite linear combinations of vectors. To talk about infinite sums, we will need to first intro-
duce norms and convergence (see Section 2).
Definition 4 (Linear independence). A set E Ď V is called linearly independent if, for any finite
subset te1 , . . . , en u Ď E,

ÿ
n
αi ei “ 0 implies that αi “ 0 for all i “ 1, . . . , n.
i“1

A set of vectors is linearly independent if there are no ‘redundant’ vectors in the set, by which
we mean that there is no vector that can be expressed as a linear combination of other members of
the set. To see this, we negate the definition above: a set of vectors E is not linearly independent
if there exists te1 , . . . , en u Ď E, βi P K for all i “ 1, . . . , n and v P Ezte1 , . . . , en u such that

ÿ
n
βi ei “ v.
i“1

Combining the concept of linear independence and the notion of a spanning set allow us to
define a Hamel basis.
Definition 5 (Hamel basis). A subset E Ď V is a Hamel basis of V if it is both linearly independent
and spans V .

Note that a Hamel basis has the same as the definition of a plain old basis you saw in Linear
Algebra. The reason for describing this concept as a ‘Hamel’ basis is that we will see a different
sort of basis, a Schauder basis, later in the module. The latter involves infinite sums, and so we
again need to define many more things before we can talk about these!
The three following results (Proposition 1, Theorem 2 and Theorem 1) provide facts about
Hamel bases that allow us to talk about the dimension of a vector space. While the first two
statements are things you will probably have seen during your study of Linear Algebra, the last
uses requires the use of the axiom of choice. This is a set theoretic axiom which asserts that, given an
infinite collection of sets, we may construct a new set which takes one element from each set in the
collection. Rather than spending a lot of time to go into detail about such things, we will omit the
proof (which is not therefore examinable). To learn more about this sort of thing, consider taking
a Set Theory module!
Proposition 1. Let E be a Hamel basis of V . Then every non-zero element of V can be uniquely written
as a finite linear combination of elements from E.
Proof. From the definition of Hamel basis, Definition 5, we know that E spans V . Therefore we
only need to show that every v P V zt0u can be expressed uniquely as a finite linear combination of
elements from E.

5
 Suppose not; then if follows that for some v P V , we have

ÿ
n ÿ
ñ
αi ei “ v “ α̃j ẽj
i“1 j“1

with n, ñ P N, te1 , . . . , en u, tẽ1 , . . . , ẽñ u Ď E and αi , α̃j P K. Without loss of generality, we may
assume that n “ ñ and ei “ ẽi for every 1 ď i ď n. To convince yourself of this, note that we can
always expand the sets to the union te1 , . . . , en u Y tẽ1 , . . . , ẽñ u and set any relevant coefficients for
the new vectors to zero in the expressions above. This implies that
ÿ
n ÿ
n ÿ
n
0“v´v “ αi ei ´ α̃j ej “ pαi ´ α̃i qei .
i“1 j“1 i“1

Applying Definition 4, we deduce that αi “ α̃i , and so the representation is indeed unique.
Theorem 1. Every vector space V has a Hamel basis.
As with a number of results in mathematics that are fairly simple to state, this is a deep result
and is therefore a little tricky to prove in full generality. We won’t do this here, but if you are
interested, see Section 1.6 of James Robinson’s textbook.
Theorem 2. If V admits a finite Hamel basis of size n P N, then every Hamel basis of V is finite and has
size n.
We won’t prove this; look back to your first module on Linear Algebra for the details. Together,
these results allow us to make the following definition of the dimension of a vector space:
Definition 6 (Dimension). If V has a Hamel basis consisting of a finite number of elements then the
dimension of V is the number of elements in this basis. If V does not have a finite Hamel basis then we say
V is infinite-dimensional.

1.2 Examples of vector spaces

We will now provide a series of standard examples of vector spaces, many of which we will refer
back to throughout the module.
Below, we use boldface to denote both finite and infinite sequences of elements from K. For
example, x P Rn means we have x “ px1 , . . . , xn q. Similarly for infinite sequences we write
N
x “ px1 , x2 , . . . q “ pxi q8
i“1 . We denote the set of all infinite sequences of elements from K as K .

Example 1. The following are examples you are already know:

• For every n P N, Rn is a real vector space of dimension n.
• For every n P N, Cn is a complex vector space of dimension n.
• For every n P N, Cn is also a real vector space of dimension n (since the complex numbers contain the
reals as a subfield).
• Rn is not a complex vector space, since for any x P Rn zt0u, we have that ix R Rn , and so Rn is not
closed under scalar multiplication with complex numbers.

6
 The following meta-example is fundamental for this course, and provides a clear indication for
why this subject is referred functional analysis.
Example 2. Let X be a set and V be a vector space over K. Let FpX, V q be the set of all functions from X
to V ; then FpX, V q is a vector space over K.
In this case, the operations of vector addition and scalar multiplication are given as follows: given any
f, g P FpX, V q and α P K, then f ` g P FpX, V q and αf P FpX, V q are the functions which satisfy
pf ` gqpxq “ f pxq ` gpxq and pαf qpxq “ αf pxq for everyx P X.
The axioms in Definition 1 are easy to verified if we set 0 P FpX, V q to be the constant function which maps
all points in X to 0 P V .
The examples we considered previously are actually particular cases of this example. To see
this, note we can naturally identify Kn with Fpt1, . . . , nu, Kq and KN with FpN, Kq; more precisely,
these spaces are isomorphic, where the appropriate isomorphism is the identification of the vector
px1 , x2 , . . .q with the function f piq “ xi and vice versa. When vector spaces are isomorphic, we
will use the – sign to indicate this, so in particular we have
Kn – Fpt1, . . . , nu, Kq and KN – FpN, Kq.
In the case of the space of sequences, given x, y P KN and α P K, their sum and a scalar multipli-
cation
x ` y “ px1 ` y1 , x2 ` y2 , . . . q and αx “ pαx1 , αx2 , . . . q
are defined pointwise. All of the other examples we consider below will be linear subspaces of
spaces of the form FpX, V q defined using some additional properties of X and V .
Example 3. The set Cpr0, 1s, Kq of all continuous functions from the interval r0, 1s to K is a linear subspace
of Fpr0, 1s, Kq, hence it is a vector space. More specifically, given f, g P Cpr0, 1s, Kq and α P K we have
pf ` gqptq “ f ptq ` gptq and pαf qptq “ αf ptq
for every t P r0, 1s. In the case where K “ R, we often simply write Cpr0, 1sq. Analogous definitions can be
made with any other open or closed interval I Ď R, such as CpI, Kq or CpIq.
Example 4. The set of all real polynomials PpIq defined on an interval I Ď R is a real vector space.
In Exercise 1.1, you are asked to show that the infinite set of polynomials ttn u8 n“0 defined on
the interval r0, 1s is a Hamel basis for Ppr0, 1sq. With this result in hand, we see that Ppr0, 1sq
is infinite-dimensional; since Ppr0, 1sq is a linear subspace of Cpr0, 1sq, we can directly infer that
Cpr0, 1sq is infinite-dimensional too.
One the other hand, note that Ppr0, 1sq is a proper linear subspace of Cpr0, 1sq, i.e. it is strictly
contained within Cpr0, 1sq. The reason for this is that there are non-differentiable continuous func-
tions (for example x ÞÑ |x ´ 21 |, or the much more extreme case of the Weierstrass function). This
means that t1, t, t2 , . . . u is therefore not a Hamel basis for Cpr0, 1sq.
The following examples are vector subspaces of the bigger space of all sequences, KN .
Example 5 (`p spaces). For 1 ď p ă 8 we define the set `p pKq to be
# +
8
ÿ
p N p
` pKq “ x “ px1 , x2 , . . . q P K : |xi | ă 8 ,
i“1

7
i.e. the set of all sequences for which the pth powers are absolutely summable. When K “ R we will
sometimes write `p for short.
In the case where p “ 8, `8 pKq Ă KN denotes the the set of all bounded sequences in K. `8 pKq is
a linear subspace of KN – FpN, Kq. Indeed, the pointwise sum or multiplication by a scalar of a bounded
sequence is again bounded.
Proposition 2. For all 1 ď p ď 8, `p pKq is a subspace of KN , and each is therefore a vector space itself.
Proof. It is enough to show that x ` y P `p pKq whenever x, y P `p pKq as `p pKq is clearly closed
under scalar multiplication. For p P r1, 8q, we have
˜ ¸
ÿn n ´
ÿ ¯p n ´
ÿ ¯ 8
ÿ 8
ÿ
|xj ` yj |p ď 2 maxt|xj |, |yj |u ď 2p |xj |p ` |yj |p ď 2p |xj |p ` |yj |p .
j“1 j“1 j“1 j“1 j“1

The right-hand side of this chain of inequalities is a fixed constant, and the left-hand side is an
increasing sequence of partial sums, so sending n Ñ 8 implies that x ` y P `p pKq.
For the case where p “ 8, we see that

|xj ` yj | ď |xj | ` |yj | ď sup |xj | ` sup |yj | ă `8,

jPN jPN

and hence taking the supremum over j P N, we have that x ` y P `8 pKq.

Remark 1. Note that for 1 ď p ă 8 and x “ px1 , x2 , . . . q P `p pKq we must have xi Ñ 0. In Exercise 1.2,
you are asked to show that `p pKq is a proper linear subspace of `q pKq whenever 1 ď p ă q ď 8.
An important subset of all `p pKq, 1 ď p ď 8, is the set tep1q , ep2q , . . . u defined for i, j P N as
piq piq
ej “ 0 unless j “ i in which case ei “ 1

e.g. ep3q “ p0, 0, 1, 0, 0, . . . q P KN . It is easy to see that tepiq uiPN is a set of linearly independent vectors,
hence all the
` ´ispaces
˘ `p pKq are infinite-dimensional. Note that this set is not a Hamel basis: for example, the
sequence 2 iPN , which is in `p pKq for all 1 ď p ď 8 cannot be written as a finite linear combination of
elements from tepiq uiPN . The linear subspace Spanptepiq uiPN q, e.g. in `1 pKq, is denoted c00 pKq and is the
vector space of all sequences in K that are eventually 0, see Exercise 1.3. Note that c00 Ă `p pKq for every
1 ď p ď 8.
Example 6. Let L̃1 p0, 1q be the set of all real-valued continuous functions f defined on the open interval
p0, 1q which satisfy
ż1
|f ptq| dt ă 8.
0

We claim that L̃ p0, 1q is a linear subspace of Fpp0, 1q, Rq, and is therefore a vector space. Indeed, take
1

f, g P L̃1 p0, 1q and α P R; then it is enough to show that f ` αg P L̃1 p0, 1q. We can check that f ` αg is
continuous, and moreover we have
ż1 ż1 ż1 ż1
|pf ` αgqptq| dt “ |f ptq ` αgptq| dt ď |f ptq| dt ` |α||gptq| dt ă 8.
0 0 0 0

In Exercise 1.4, you are asked to show that Cp0, 1q (formally Cpp0, 1qq) the space of all continuous
functions defined on the open interval p0, 1q is not a subspace (subset) of L̃1 p0, 1q. On the other hand, we

8
can view Cpr0, 1sq as a subspace of L̃1 p0, 1q. Given f P Cpr0, 1sq its restriction to p0, 1q, that we also denote
by f , is bounded in absolute value by some M P R. Hence we have
ż1 ż1
|f ptq| dt ď M dt “ M ă 8
0 0

and f P L̃1 p0, 1q.

1.3 Linear maps

Along with vector spaces, another important concept will be that of a linear map, which we define
here.
Definition 7. Let U, V be vector spaces over K. An operator (or a map) T : U Ñ V is linear if

T pαx ` βyq “ αT pxq ` βT pyq

for every x, y P U and α, β P K.

We write LpU, V q for the set of all linear operators from U to V . This is a linear subspace of FpU, V q,
and therefore is a vector space with the usual notions of vector addition and scalar multiplication.
A linear operator T : U Ñ V that is bijective (injective and surjective) is called a vector space
isomorphism, or linear isomorphism. An isomorphism T has an inverse which is itself a linear map,
i.e. T ´1 P LpV, U q. In this case, U and V are called isomorphic, as they are identical as vector
spaces. Note that we implicitly assume that both vector spaces are defined over the same field K;
for example R2n as a vector space over R and Cn as a vector space over C are not isomorphic, but
R2n over R and Cn over R are.
In Exercise 1.5, you are asked to show that injective linear map between two vector spaces of
the same finite dimension is necessarily a vector space isomorphism. This is no longer true for
infinite-dimensional spaces, see Exercise 1.6.
Proposition 3. Any n-dimensional vector space V over K is isomorphic to Kn .
řn
Proof. Let tei uni“1 be a Hamel basis of V . Define a map pα1 , . . . , αn qřÞÑ i“1 αi ei . This map is
n
clearly linear, and is surjective since every v P V can be written as v “ i“1 αi ei for a unique vector
pαi qni“1 P Kn . Moreover, the uniqueness of the representation means that this map is injective,
hence V and Kn are isomorphic.
The following definition will be useful when we discuss operators in more detail in Section 5.
Definition 8. Let U, V be vector spaces over C. A map T : U Ñ V is called conjugate-linear, or anti-
linear, if
T px ` yq “ T pxq ` T pyq and T pαxq “ αx
for every x, y P U and α P K, where α is the conjugate of α in C.
To illustrate the definition consider the map T : C Ñ C defined as T pαq “ α. It is clear that
T px ` yq “ T pxq ` T pyq for every x, y P C. If we view C as a vector space over C, then T is not
linear but conjugate-linear. Indeed, T pαxq “ αx “ α x “ αT pxq. On the other hand, when we
consider C as a vector space over R, then T is a linear map.

9
1.4 Exercises
1.1 Show that the infinite sequence of polynomials t1, t, t2 , . . . u defined on the interval r0, 1s is
a Hamel basis in Ppr0, 1sq. Deduce from this that the space Cpr0, 1sq has infinite dimension.
As a starting point, you may wish to consider how differentiation operates here.
1.2 Show that `p pKq is a proper linear subspace of `q pKq whenever 1 ď p ă q ď 8.
1.3 Let c00 pKq denote the set of all sequences with only a finite number of non-zero terms. Show
that this is a vector space, and that tepiq uiPN is a Hamel basis of this space.

1.4 Show that there is f P Cp0, 1q (the space of all continuous functions defined on the open
interval p0, 1q) such that f R L̃1 p0, 1q. Thus Cp0, 1q is not a subspace of L̃1 p0, 1q.

1.5 Let U, V be vector spaces over K of the same finite dimension n P N. Suppose that T : U Ñ V
is a linear injective map and tei uni“1 is a Hamel basis in U . Show that tT pei quni“1 is a Hamel
basis in V . Use this to deduce that an injective (or surjective) linear map between two vector
spaces of the same finite dimension is necessarily a vector space isomorphism.
1.6 Find a linear map between infinite-dimensional spaces that is injective but not surjective,
and surjective but not injective. You may wish to use the result of Exercise 1.1.
1.7 Show that the collection of all real-valued smooth (infinitely differentiable) functions defined
on the interval r0, 1s is a vector space. This space is denoted as C 8 pr0, 1sq.

10