Vector Spaces

⚡ Key Takeaways

A vector space is a set equipped with addition and scalar multiplication that satisfy eight axioms, generalizing the arithmetic of arrows in the plane to any setting — polynomials, functions, matrices, solutions of linear differential equations. Key result: every vector space has a basis, and its dimension (the size of any basis) is a well-defined invariant. Applies in: all of linear algebra, functional analysis, quantum mechanics, coding theory, computer graphics. Who/when: axiomatized by Giuseppe Peano (1888), building on Hamilton, Grassmann, and Cayley.

💡 The Core Insight

A vector space strips away everything specific about "arrows" and keeps only the rules of combination: you can add any two elements and scale any element, and these operations interact predictably (associativity, distributivity, identities, inverses). Once you accept those rules, every theorem about linear combinations, spans, independence, bases, and dimension follows — and applies identically to geometric vectors, polynomials, functions, and infinite sequences. Linearity is not a property of arrows; it is an abstract structure, and arrows are just one model of it.

DefinitionVector Space

Let $F$ be a field (typically $\mathbb{R}$ or $\mathbb{C}$ ). A vector space over $F$ is a set $V$ together with two operations,

1

\quad +: V \times V \to V, \qquad \cdot: F \times V \to V,

Derivation hint: these are the defining data; closure under both operations is built into the codomain $V$ . such that for all $\mathbf{u}, \mathbf{v}, \mathbf{w} \in V$ and all $a, b \in F$ the following eight axioms hold:

2

\quad (\mathbf{u}+\mathbf{v})+\mathbf{w} = \mathbf{u}+(\mathbf{v}+\mathbf{w}) \quad\text{(associativity of }+)

Derivation hint: grouping of repeated additions is irrelevant — inherited from the abelian-group structure.

3

\quad \mathbf{u}+\mathbf{v} = \mathbf{v}+\mathbf{u} \quad\text{(commutativity of }+)

Derivation hint: order of addition does not matter.

4

\quad \exists\, \mathbf{0}\in V:\ \mathbf{v}+\mathbf{0} = \mathbf{v} \quad\text{(additive identity)}

Derivation hint: the zero vector is a distinguished element fixed by all additions.

5

\quad \forall \mathbf{v}\,\exists (-\mathbf{v}):\ \mathbf{v}+(-\mathbf{v}) = \mathbf{0} \quad\text{(additive inverse)}

Derivation hint: every element has an opposite; together with [2]–[4] this makes $(V,+)$ an abelian group.

6

\quad a\cdot(\mathbf{u}+\mathbf{v}) = a\cdot\mathbf{u} + a\cdot\mathbf{v} \quad\text{(distributivity over vector addition)}

Derivation hint: scaling a sum scales each part.

7

\quad (a+b)\cdot\mathbf{v} = a\cdot\mathbf{v} + b\cdot\mathbf{v} \quad\text{(distributivity over field addition)}

Derivation hint: splitting the scalar splits the result.

8

\quad (ab)\cdot\mathbf{v} = a\cdot(b\cdot\mathbf{v}), \qquad 1\cdot\mathbf{v} = \mathbf{v} \quad\text{(compatibility & unit)}

Derivation hint: scalar multiplication respects field multiplication, and the field's unit acts trivially. Elements of $V$ are vectors; elements of $F$ are scalars.

DefinitionSubspace, Span, Linear Independence, Basis

A subspace $W \subseteq V$ is a nonempty subset closed under addition and scalar multiplication (hence itself a vector space). Given vectors $\mathbf{v}_1,\dots,\mathbf{v}_n$ , a linear combination is

9

\quad \sum_{i=1}^{n} a_i \mathbf{v}_i, \qquad a_i \in F,

Derivation hint: closure axioms guarantee any finite linear combination lands back in $V$ . Their span is the set of all such combinations, the smallest subspace containing them. The set is linearly independent if

10

\quad \sum_{i=1}^{n} a_i \mathbf{v}_i = \mathbf{0} \ \implies\ a_1=\cdots=a_n=0.

Derivation hint: the only "way to make zero" is the trivial way; otherwise one vector is redundant. A basis is a linearly independent spanning set; the dimension $\dim V$ is the cardinality of any basis.

TheoremExistence and Invariance of Dimension

Every vector space has a basis (assuming the Axiom of Choice for infinite-dimensional spaces). Moreover, any two bases of the same vector space have the same cardinality. Hence dimension is a well-defined invariant, and two vector spaces over the same field are isomorphic if and only if they have the same dimension.

ProofTwo finite bases have equal size (Steinitz Exchange)

We prove the finite-dimensional case: if $B = \{\mathbf{b}_1,\dots,\mathbf{b}_m\}$ spans $V$ and $L = \{\mathbf{l}_1,\dots,\mathbf{l}_n\}$ is linearly independent, then $n \le m$ . 1. Since $B$ spans, write $\mathbf{l}_1 = \sum_i a_i \mathbf{b}_i$ . Independence of $L$ forces $\mathbf{l}_1 \ne \mathbf{0}$ , so some coefficient $a_k \ne 0$ . Solve for $\mathbf{b}_k$ in terms of $\mathbf{l}_1$ and the other $\mathbf{b}_i$ . 2. Exchange step: replacing $\mathbf{b}_k$ by $\mathbf{l}_1$ yields a new set $B_1$ that still spans $V$ (anything expressible via $\mathbf{b}_k$ is now expressible via $\mathbf{l}_1$ and the rest). 3. Repeat: express $\mathbf{l}_2$ in terms of $B_1$ . The coefficients of the $\mathbf{l}$ 's alone cannot vanish (that would contradict independence of $L$ ), so some remaining $\mathbf{b}$ -coefficient is nonzero, and we exchange another $\mathbf{b}$ for $\mathbf{l}_2$ . 4. Each exchange swaps one $\mathbf{l}_j$ into the spanning set in place of a $\mathbf{b}$ . If $n > m$ we would run out of $\mathbf{b}$ 's before placing all the $\mathbf{l}$ 's, forcing a leftover $\mathbf{l}$ to be a combination of the already-placed ones — contradicting independence. Hence $n \le m$ . 5. Now apply the result both ways: if $B$ and $B'$ are both bases, each is independent and each spans, so $|B| \le |B'|$ and $|B'| \le |B|$ . Therefore $|B| = |B'|$ . $\blacksquare$

∎

mathematician · Giuseppe Peano | 1858–1932

Peano gave the first fully axiomatic definition of a vector space (he called it a "linear system") in his 1888 book Calcolo geometrico, listing essentially the modern axioms — including the abstract treatment of dimension and linear operators — decades before the concept became standard. Better known for his axioms of arithmetic and his space-filling curve, Peano also pioneered the formal, symbol-driven style of mathematical writing. His vector-space axioms were largely ignored at the time; the abstraction only took hold in the 1920s–30s through Banach, Weyl, and van der Waerden.

🧠 Intuition

Think of a vector space as a set of things you can mix. Two operations are allowed: pour two together (addition) and amplify one (scalar multiplication). The axioms simply say mixing behaves sanely — pouring is order-independent, there is an "empty" element (zero) that changes nothing, every element can be undone, and amplification distributes over mixing. A basis is a minimal set of "pure ingredients" from which every possible mixture can be recreated in exactly one way; the dimension counts those ingredients. The deep point is that the geometry of arrows is not required — polynomials, audio signals, and quantum states all "mix" by the same rules, so all of linear-algebra's machinery transfers to them unchanged.

Figure

A 2D vector space: basis vectors spanning the plane, one vector as a unique linear combination

ExampleVerifying a subspace and finding its dimension

Given: In $\mathbb{R}^3$ , let $W = \{(x,y,z) : x + 2y - z = 0\}$ . Is $W$ a subspace, and what is $\dim W$ ?

Steps: 1. Contains zero: $(0,0,0)$ satisfies $0+0-0=0$ . ✓ 2. Closed under addition: if $(x_1,y_1,z_1)$ and $(x_2,y_2,z_2)$ both satisfy the equation, add them: $(x_1+x_2)+2(y_1+y_2)-(z_1+z_2) = 0+0 = 0$ . ✓ 3. Closed under scaling: $a x + 2(ay) - (az) = a(x+2y-z) = a\cdot 0 = 0$ . ✓ So $W$ is a subspace. 4. Find a basis: solve $z = x + 2y$ , with $x,y$ free. Then

11

\quad (x,y,z) = x(1,0,1) + y(0,1,2).

Derivation hint: substitute $z=x+2y$ and split the free parameters. 5. The two vectors $(1,0,1)$ and $(0,1,2)$ span $W$ and are independent (neither is a scalar multiple of the other).

Answer: $\boxed{W \text{ is a subspace with } \dim W = 2}$ (a plane through the origin).

ExamplePolynomials as a vector space — dimension by counting

Given: $P_3 = \{a_0 + a_1 t + a_2 t^2 + a_3 t^3 : a_i \in \mathbb{R}\}$ , polynomials of degree $\le 3$ . Show it is a vector space and find $\dim P_3$ . Then express $p(t) = 2 - t + 4t^3$ in the standard basis.

Steps: 1. Sum of two such polynomials is again degree $\le 3$ ; scalar multiple likewise. The eight axioms hold because polynomial addition and scaling are just coefficient-wise operations on $\mathbb{R}^4$ . 2. The set $\{1, t, t^2, t^3\}$ spans (by definition) and is independent: a polynomial is the zero function only when all coefficients vanish. So it is a basis.

12

\quad \dim P_3 = 4.

Derivation hint: count the basis monomials $t^0,t^1,t^2,t^3$ . 3. Coordinate vector of $p$ relative to $\{1,t,t^2,t^3\}$ :

13

\quad [p] = (2,\,-1,\,0,\,4).

Derivation hint: read off the coefficients in order, inserting $0$ for the missing $t^2$ term.

Answer: $\boxed{\dim P_3 = 4,\quad [p]_{\{1,t,t^2,t^3\}} = (2,-1,0,4)}$ . Note $P_3 \cong \mathbb{R}^4$ as vector spaces.

⚠️ Common mistake

"Vectors are arrows / lists of numbers." They need not be either. Functions, matrices, infinite sequences, and polynomials are vectors in their own spaces. The word "vector" means any element of a vector space. Conversely, not every set of tuples is a subspace: a subspace must contain $\mathbf{0}$ and be closed under both operations. A line not through the origin is not a subspace (it fails to contain zero), even though it "looks like" a line of vectors.

⚠️ Common mistake

" $\dim$ always equals the number of coordinates." The ambient $\mathbb{R}^n$ has dimension $n$ , but a subspace inside it can have any dimension from $0$ to $n$ . Also, the zero vector space $\{\mathbf{0}\}$ has dimension $0$ — its only basis is the empty set, not $\{\mathbf{0}\}$ (which is linearly dependent since $1\cdot\mathbf{0}=\mathbf{0}$ ).

Counterexample

A "vector space" that fails an axiom. Let $V = \mathbb{R}^2$ with usual addition but redefine scaling as $a \odot (x,y) = (ax, 0)$ . Then axiom [8]'s unit law fails: $1 \odot (x,y) = (x,0) \ne (x,y)$ . Despite "looking like" $\mathbb{R}^2$ , this structure is not a vector space — confirming that all eight axioms must be checked, not just closure and the obvious-looking ones.

connection

Vector spaces are the objects of Linear Algebra; the structure-preserving maps between them are Linear Transformations, represented in coordinates by Matrices. Adding a notion of length and angle yields Inner Product Spaces, and completing those metrically gives Hilbert Spaces and Banach Spaces, the home of Functional Analysis. The set of all linear maps $V \to F$ forms the Dual Space, and the rank–nullity theorem links the dimensions of Kernel and Image. Generalizing the scalar field $F$ to an arbitrary ring produces Modules, of which vector spaces are the special "field" case.

🔬 Where this appears

⏳ History

- 1827 — August Möbius introduces barycentric coordinates, an early hint of coordinatized affine/vector structure.

- 1843 — William Rowan Hamilton discovers quaternions, the first non-commutative algebra with vector-like addition.

- 1844 — Hermann Grassmann publishes Die lineale Ausdehnungslehre, developing

n

-dimensional spaces, linear independence, and the exterior product — far ahead of its time and largely unread.

- 1850s — Arthur Cayley and James Sylvester develop matrix theory, providing the operator side of linear algebra.

- 1888 — Giuseppe Peano gives the first axiomatic definition of an (abstract, possibly infinite-dimensional) vector space in Calcolo geometrico.

- 1918 — Hermann Weyl, in Raum, Zeit, Materie, uses the axiomatic vector-space framework to formulate general relativity's geometry.

- 1920s — Stefan Banach and Hans Hahn develop normed/complete vector spaces, founding functional analysis.

- 1930 — Bartel van der Waerden's Moderne Algebra canonizes the modern axiomatic treatment, making vector spaces standard undergraduate fare.

- 1932 — John von Neumann gives Hilbert spaces their definitive axioms, grounding quantum mechanics in vector-space language.

open-problem

For infinite-dimensional spaces the comfortable finite picture frays. The existence of a basis (a Hamel basis) for every vector space is equivalent to the Axiom of Choice — and such bases are non-constructive; no one can write down an explicit Hamel basis for $\mathbb{R}$ over $\mathbb{Q}$ . In functional analysis one therefore prefers topological (Schauder) bases, where infinite linear combinations are allowed via convergence, but these need not exist (Per Enflo's 1973 counterexample showed some Banach spaces have no Schauder basis at all, settling the basis problem negatively). Cardinal arithmetic of dimensions also raises delicate questions: two infinite-dimensional spaces are isomorphic iff their bases have equal cardinality, which can depend on set-theoretic assumptions. These issues mark the boundary where pure linear algebra hands off to functional analysis and set theory.

Formal Statement

Why It Holds / Derivation

Building Intuition

Worked Examples

Common Misconceptions

Connections

Real-World Applications

Historical Development

Subtleties & Limits

See Also