Chapter 107. Infinite-Dimensional Vector Spaces

Infinite-dimensional vector spaces extend linear algebra beyond finite lists of coordinates. Their elements may be sequences, functions, distributions, operators, or formal series. Many ideas from finite-dimensional linear algebra survive, but they often require additional structure, especially topology, norm, inner product, or convergence.

The main distinction is this: algebraic linear algebra uses finite linear combinations, while analysis often needs infinite limiting processes. This difference separates Hamel bases from Schauder bases and separates vector-space dimension from analytic approximation. A Hamel basis expresses each vector as a finite linear combination, while a Schauder basis permits infinite series converging in the topology of the space.

107.1 Definition

A vector space $V$ over a field $F$ is infinite-dimensional if no finite set of vectors spans $V$ .

Equivalently, for every finite subset

\{v_1,\ldots,v_n\}\subseteq V,

there exists some vector

w\in V

that cannot be written as

w=c_1v_1+\cdots+c_nv_n.

Thus the space cannot be described by finitely many coordinates.

Examples include:

Space	Elements
$F[x]$	Polynomials over $F$
$C([0,1])$	Continuous functions on $[0,1]$
$\ell^2$	Square-summable sequences
$L^2([0,1])$	Square-integrable functions
$M_\infty(F)$	Infinite matrices, with suitable restrictions

Finite-dimensional linear algebra remains the local model, but infinite-dimensional spaces require stronger care.

107.2 Polynomial Spaces

The polynomial space

F[x]

is the set of all polynomials

a_0+a_1x+\cdots+a_nx^n

with coefficients in $F$ .

It is infinite-dimensional because no finite list of powers

1,x,x^2,\ldots,x^m

can span all polynomials.

The monomials

1,x,x^2,x^3,\ldots

form a basis in the algebraic sense.

Every polynomial is a finite linear combination of these monomials.

Thus

F[x]

is infinite-dimensional, but each individual vector still has finite coordinate support relative to this basis.

107.3 Sequence Spaces

A sequence space consists of infinite scalar sequences

x=(x_1,x_2,x_3,\ldots).

Common examples include:

Space	Condition
$c_{00}$	Only finitely many nonzero entries
$c_0$	Entries converge to $0$
$\ell^1$	(\sum_{n=1}^{\infty}
$\ell^2$	(\sum_{n=1}^{\infty}
$\ell^\infty$	(\sup_n

The standard unit vectors are

e_1=(1,0,0,\ldots),

e_2=(0,1,0,\ldots),

and so on.

In $c_{00}$ , every sequence is a finite linear combination of the $e_i$ . In $\ell^2$ , many sequences require infinite expansions.

For example,

\left(1,\frac12,\frac13,\frac14,\ldots\right)

does not lie in $\ell^2$ , but

\left(1,\frac12,\frac14,\frac18,\ldots\right)

does.

107.4 Function Spaces

Function spaces are among the most important infinite-dimensional vector spaces.

If $X$ is a set and $F$ is a field, the set of all functions

f:X\to F

forms a vector space under pointwise addition and scalar multiplication:

(f+g)(x)=f(x)+g(x),

(cf)(x)=c f(x).

Important subspaces include:

Space	Meaning
$C([a,b])$	Continuous functions
$C^k([a,b])$	$k$ -times continuously differentiable functions
$C^\infty([a,b])$	Smooth functions
$L^p$	$p$ -integrable functions
Polynomial functions	Finite-degree algebraic functions

Function spaces are usually studied with norms, inner products, or topologies, since convergence is central.

107.5 Algebraic Basis: Hamel Basis

A Hamel basis of a vector space $V$ is a set

\mathcal B\subseteq V

such that every vector $v\in V$ can be written uniquely as a finite linear combination

v=c_1b_1+\cdots+c_nb_n,

where

b_i\in\mathcal B.

The word finite is essential.

This definition is exactly the same as the finite-dimensional basis definition, except that the basis set may be infinite.

For example,

\{1,x,x^2,x^3,\ldots\}

is a Hamel basis for $F[x]$ .

Every polynomial uses only finitely many powers of $x$ .

107.6 Existence of Hamel Bases

Every vector space has a Hamel basis, assuming the usual form of the axiom of choice.

The proof commonly uses Zorn’s lemma. One considers the partially ordered set of all linearly independent subsets of $V$ , ordered by inclusion. A maximal linearly independent set is then shown to span $V$ . This gives a basis.

In finite-dimensional spaces, such bases can be constructed algorithmically by extending linearly independent lists. In infinite-dimensional spaces, existence may be nonconstructive.

For many analytic spaces, Hamel bases exist but are too large and irregular to be useful. In infinite-dimensional Banach spaces, Hamel bases must be very large, at least of continuum cardinality.

107.7 Dimension

The dimension of a vector space is the cardinality of a Hamel basis.

If $V$ has a finite basis with $n$ elements, then

\dim V=n.

If $V$ has an infinite basis, then

\dim V

is an infinite cardinal.

For example,

\dim F[x]=\aleph_0

because the monomials form a countable Hamel basis.

Other spaces, such as $C([0,1])$ over $\mathbb R$ , have much larger Hamel dimension.

Dimension remains a purely algebraic concept. It does not record convergence, approximation, continuity, or norm.

107.8 Linear Independence

A subset

S\subseteq V

is linearly independent if every finite relation

c_1v_1+\cdots+c_nv_n=0

with distinct vectors

v_1,\ldots,v_n\in S

forces

c_1=\cdots=c_n=0.

Even in infinite-dimensional spaces, linear independence is tested only by finite linear combinations.

For example,

1,x,x^2,x^3,\ldots

are linearly independent in $F[x]$ , because a polynomial that is identically zero has all coefficients zero.

107.9 Span

The span of a subset $S\subseteq V$ is the set of all finite linear combinations of elements of $S$ :

\operatorname{span}(S) = \left\{ c_1s_1+\cdots+c_ns_n: n\ge 0,\ c_i\in F,\ s_i\in S \right\}.

The definition uses finite sums.

This is important. In a plain vector space, infinite sums are not defined. To speak of infinite series, one must add a topology, norm, metric, or other notion of convergence.

Thus

\operatorname{span}\{e_1,e_2,e_3,\ldots\}

inside all sequences consists only of sequences with finite support, unless a topology is introduced.

107.10 Topological Vector Spaces

A topological vector space is a vector space with a topology in which addition and scalar multiplication are continuous.

This allows one to discuss limits:

v_n\to v.

It also allows infinite series:

\sum_{n=1}^{\infty} v_n.

A normed vector space is a topological vector space where the topology comes from a norm.

A Banach space is a complete normed vector space.

A Hilbert space is a complete inner product space.

These structures are central in infinite-dimensional linear algebra because many useful vectors appear as limits rather than finite combinations.

107.11 Schauder Basis

A Schauder basis is an ordered sequence

(b_1,b_2,b_3,\ldots)

in a topological vector space $V$ such that every vector $v\in V$ has a unique expansion

v=\sum_{n=1}^{\infty} a_n b_n,

where the series converges in the topology of $V$ .

This differs from a Hamel basis. A Hamel basis allows only finite sums. A Schauder basis allows infinite convergent sums. This makes Schauder bases more suitable for Banach spaces and other infinite-dimensional topological vector spaces.

The standard unit vectors form a Schauder basis for spaces such as $c_0$ and $\ell^p$ for $1\le p<\infty$ . In a separable Hilbert space, every countable orthonormal basis gives a Schauder basis.

107.12 Hamel Basis Versus Schauder Basis

The two notions of basis answer different questions.

Feature	Hamel basis	Schauder basis
Type of sum	Finite	Infinite convergent
Requires topology	No	Yes
Coordinates	Algebraic	Analytic
Common in finite dimension	Yes	Yes
Useful in Banach spaces	Often poor	Often useful
Always exists	Yes, with choice	No

In finite-dimensional spaces, the distinction disappears because all expansions are finite.

In infinite-dimensional analysis, the distinction becomes essential.

A Hamel basis of a Banach space is usually too large for computation or approximation. A Schauder basis is more compatible with limits, projections, and numerical approximation.

107.13 Dense Span

In a normed space, a set may fail to span the whole space algebraically but still have dense span.

A subset $S$ has dense span if every vector $v\in V$ can be approximated arbitrarily well by finite linear combinations of elements of $S$ .

That means for every $\varepsilon>0$ , there exists

w\in\operatorname{span}(S)

such that

\|v-w\|<\varepsilon.

Dense span is weaker than algebraic span.

For example, polynomials are dense in $C([a,b])$ with the supremum norm by the Weierstrass approximation theorem. But polynomials do not equal all continuous functions.

107.14 Hilbert Spaces

A Hilbert space is a complete inner product space.

The inner product allows one to define length:

\|v\|=\sqrt{\langle v,v\rangle}.

It also allows orthogonality:

\langle u,v\rangle=0.

Hilbert spaces are the infinite-dimensional setting closest to Euclidean geometry.

Examples include:

Hilbert space	Inner product
$\ell^2$	$\langle x,y\rangle=\sum x_n\overline{y_n}$
$L^2([a,b])$	$\langle f,g\rangle=\int_a^b f(x)\overline{g(x)}\,dx$

Orthogonal projection, Fourier expansion, and spectral theory all rely on Hilbert-space structure.

107.15 Orthonormal Bases

In a Hilbert space, an orthonormal basis is a set

\{e_i\}_{i\in I}

such that

\langle e_i,e_j\rangle=0

for $i\ne j$ , and

\|e_i\|=1

for every $i$ , with the additional condition that the closed linear span is the whole space.

The word closed is essential.

In a separable Hilbert space, an orthonormal basis may be countable:

e_1,e_2,e_3,\ldots.

Then every vector has an expansion

v=\sum_{n=1}^{\infty}\langle v,e_n\rangle e_n.

The convergence is norm convergence.

107.16 Fourier Series as Linear Algebra

Fourier series are infinite-dimensional coordinate expansions.

For suitable functions on $[-\pi,\pi]$ , one writes

f(x) \sim \sum_{n=-\infty}^{\infty} c_n e^{inx}.

The functions

e^{inx}

play the role of basis vectors.

The coefficients are inner products:

c_n= \frac{1}{2\pi} \int_{-\pi}^{\pi} f(x)e^{-inx}\,dx.

This is linear algebra in a Hilbert space. The difference from finite-dimensional coordinate systems is that convergence must be studied carefully.

107.17 Linear Operators

In infinite-dimensional spaces, linear maps are often called linear operators.

A linear operator is a map

T:V\to W

satisfying

T(u+v)=T(u)+T(v)

and

T(cv)=cT(v).

When $V$ and $W$ are normed spaces, one usually asks whether $T$ is continuous.

In finite-dimensional normed spaces, every linear map is continuous.

In infinite-dimensional spaces, discontinuous linear maps can exist. This is one of the major differences between finite-dimensional and infinite-dimensional theory.

107.18 Bounded Operators

A linear operator

T:V\to W

between normed spaces is bounded if there exists a constant $C\ge 0$ such that

\|Tv\|\le C\|v\|

for all $v\in V$ .

A linear operator between normed spaces is continuous if and only if it is bounded.

The operator norm is

\|T\| = \sup_{\|v\|\le 1}\|Tv\|.

Bounded operators form the natural class of linear maps in functional analysis.

107.19 Unbounded Operators

Many important operators are not bounded.

For example, differentiation

D(f)=f'

is not bounded on many common function spaces.

This reflects the fact that a function may be small in norm while its derivative is large.

Unbounded operators are essential in differential equations and quantum mechanics, but they require careful treatment of domains.

An unbounded operator is usually not defined on the whole space. Instead, it has a domain

\mathcal D(T)\subseteq V.

The domain is part of the operator.

107.20 Dual Spaces

The algebraic dual

V^*

is the space of all linear functionals

f:V\to F.

For infinite-dimensional normed spaces, one often studies the continuous dual instead:

V' = \{f:V\to F : f\text{ is linear and continuous}\}.

The continuous dual may be much smaller than the algebraic dual.

This distinction has no analogue in finite-dimensional normed spaces, where all linear functionals are continuous.

107.21 Quotient Spaces and Closed Subspaces

If $M\subseteq V$ is a subspace, the quotient space

V/M

can be formed algebraically.

In normed spaces, the quotient norm is well behaved when $M$ is closed.

If $M$ is not closed, then the quotient may fail to be Hausdorff as a normed space.

Thus infinite-dimensional analysis often distinguishes subspaces from closed subspaces.

In finite-dimensional spaces, every subspace is closed.

107.22 Compactness

Compactness behaves differently in infinite dimensions.

In finite-dimensional normed spaces, closed and bounded sets are compact.

In infinite-dimensional normed spaces, the closed unit ball is generally not compact.

This difference has major consequences. Many finite-dimensional arguments rely on compactness and fail in infinite dimensions.

For example, a bounded sequence need not have a convergent subsequence.

Weak topologies and compact operators are partly designed to recover usable compactness principles.

107.23 Compact Operators

A compact operator is a linear operator that sends bounded sets to relatively compact sets.

In Hilbert and Banach spaces, compact operators behave in some ways like matrices with singular values tending to zero.

Integral operators are common examples:

(Tf)(x)=\int_a^b K(x,y)f(y)\,dy.

Compact operators are central in spectral theory because their nonzero spectral behavior resembles finite-dimensional eigenvalue theory more closely than general bounded operators.

107.24 Finite-Rank Operators

A linear operator $T:V\to W$ has finite rank if

\dim \operatorname{im}(T)<\infty.

Finite-rank operators are the closest infinite-dimensional analogue of matrices with finite-dimensional output.

They are often used to approximate more complicated operators.

If a space has a Schauder basis, the partial-sum projections

P_n(v)=\sum_{k=1}^n a_k b_k

are finite-rank operators that approximate the identity on each vector.

This is one reason bases are useful in analysis.

107.25 Failure of Determinants

In finite-dimensional linear algebra, determinants are central.

For a square matrix $A$ , the determinant detects invertibility, volume scaling, eigenvalue products, and orientation.

In infinite-dimensional spaces, there is no determinant for arbitrary linear operators.

Special determinant theories exist for special classes of operators, such as trace-class perturbations of the identity. But a general bounded operator on an infinite-dimensional Banach space does not have an ordinary determinant.

Thus infinite-dimensional linear algebra cannot simply copy finite-dimensional matrix theory.

107.26 Failure of Eigenvalue Completeness

In finite dimensions, many operators can be studied through eigenvalues and eigenvectors.

In infinite dimensions, an operator may have no eigenvectors, or its eigenvectors may fail to span a useful subspace.

For example, the shift operator on sequence spaces moves coordinates:

S(x_1,x_2,x_3,\ldots) = (0,x_1,x_2,\ldots).

Such operators illustrate that spectral theory must include more than eigenvalues.

The spectrum of an operator becomes more important than the point spectrum alone.

107.27 Infinite Matrices

An infinite matrix is an array

A=(a_{ij})_{i,j\ge 1}.

It may define a linear operator on a sequence space by

(Ax)_i=\sum_{j=1}^{\infty}a_{ij}x_j.

But this formula is meaningful only when the series converge and the resulting sequence lies in the target space.

Thus infinite matrices require analytic conditions.

Unlike finite matrices, arbitrary infinite arrays do not automatically define linear operators on useful spaces.

107.28 Approximation

Approximation is the practical heart of infinite-dimensional linear algebra.

One often replaces an infinite-dimensional problem by a finite-dimensional one.

Examples include:

Infinite problem	Finite approximation
Function $f$	Polynomial approximation
Fourier series	Truncated Fourier series
Differential equation	Finite element method
Integral operator	Matrix discretization
Hilbert-space projection	Projection onto finite subspace

The central question is whether the approximations converge and how fast.

This is why norms, projections, completeness, and stability are essential.

107.29 Example: $\ell^2$

The space

\ell^2 = \left\{ x=(x_1,x_2,\ldots): \sum_{n=1}^{\infty}|x_n|^2<\infty \right\}

has inner product

\langle x,y\rangle = \sum_{n=1}^{\infty}x_n\overline{y_n}.

The standard unit vectors

e_1,e_2,e_3,\ldots

form an orthonormal basis.

Every $x\in\ell^2$ has expansion

x=\sum_{n=1}^{\infty}x_ne_n.

The convergence is in the $\ell^2$ -norm:

\left\| x-\sum_{n=1}^{N}x_ne_n \right\|_2 \to 0.

This is not a Hamel expansion because the sum may be infinite. It is an analytic expansion.

107.30 Summary

Infinite-dimensional vector spaces extend linear algebra to spaces too large to be spanned by finitely many vectors.

Concept	Finite-dimensional case	Infinite-dimensional case
Basis	Finite Hamel basis	Hamel basis may be infinite and nonconstructive
Coordinates	Finite tuples	Finite or infinite expansions
Linear maps	Always continuous in normed spaces	May be discontinuous
Subspaces	Always closed	May fail to be closed
Unit ball	Compact when closed and bounded	Usually not compact
Determinant	General square matrices	Only special operator classes
Eigenvectors	Often central	May be insufficient

The central lesson is that algebra alone is not enough for most infinite-dimensional problems. One must also study convergence, topology, continuity, completeness, and approximation. Infinite-dimensional linear algebra therefore becomes the entry point to functional analysis, operator theory, and modern analysis.

Chapter 107. Infinite-Dimensional Vector Spaces

107.1 Definition

107.2 Polynomial Spaces

107.3 Sequence Spaces

107.4 Function Spaces

107.5 Algebraic Basis: Hamel Basis

107.6 Existence of Hamel Bases

107.7 Dimension

107.8 Linear Independence

107.9 Span

107.10 Topological Vector Spaces

107.11 Schauder Basis

107.12 Hamel Basis Versus Schauder Basis

107.13 Dense Span

107.14 Hilbert Spaces

107.15 Orthonormal Bases

107.16 Fourier Series as Linear Algebra

107.17 Linear Operators

107.18 Bounded Operators

107.19 Unbounded Operators

107.20 Dual Spaces

107.21 Quotient Spaces and Closed Subspaces

107.22 Compactness

107.23 Compact Operators

107.24 Finite-Rank Operators

107.25 Failure of Determinants

107.26 Failure of Eigenvalue Completeness

107.27 Infinite Matrices

107.28 Approximation

107.29 Example: ℓ2\ell^2ℓ2

107.30 Summary

107.29 Example: $\ell^2$