Vector space

• 1. Introduction

  ›

  Vector space is the mathematical structure that sits beneath nearly all of modern science, engineering, and computation. Forces in physics, solutions to differential equations, signals in engineering, even the wavefunctions of quantum mechanics: all of them live inside vector spaces. What makes this idea so powerful is deceptively simple. You take a collection of objects, you give them two operations, and eight rules, and suddenly the whole machinery of linear algebra opens up. The questions worth exploring are these: where did this concept come from, who shaped it, and why does a single abstract structure reach so far across mathematics and the physical world?

• 2. Eight Rules

  ›

  A vector space over a field F is a non-empty set V together with two operations: vector addition, which combines any two elements of V into a third, and scalar multiplication, which pairs any scalar from F with any vector to produce another vector. Both operations must satisfy eight axioms. Addition must be associative and commutative. There must be a zero vector that leaves any vector unchanged when added to it. Every vector must have an additive inverse. Scalar multiplication must be compatible with field multiplication, must respect the identity element of F, and must distribute over both vector addition and field addition. What is striking is that these eight rules are the minimum. Satisfy all eight and you automatically inherit a vast range of consequences, including the fact that the zero vector is unique and that multiplying any vector by the scalar zero always yields the zero vector.

• 3. Roots and Founders

  ›

  Around 1636, Rene Descartes and Pierre de Fermat founded analytic geometry by linking solutions of two-variable equations to curves in the plane. Their coordinate approach planted the seed. Bernard Bolzano, in 1804, introduced operations on points, lines, and planes that anticipated what vectors would later formalize. A further step came when an unnamed researcher introduced an equivalence relation on directed line segments sharing the same length and direction, calling this relation equipollence, so that a Euclidean vector became an equivalence class. Complex numbers, Argand's and Hamilton's presentations of them, and Hamilton's invention of quaternions pushed the idea further: these were elements in R2 and R4, and treating them through linear combinations goes back to Laguerre in 1867, who also defined systems of linear equations. In 1857, Cayley introduced matrix notation, which brought harmony and simplicity to the study of linear maps. Grassmann, working around the same time, studied barycentric calculus and envisaged sets of abstract objects with operations; his 1844 work already contained the concepts of linear independence, dimension, and scalar products. Italian mathematician Giuseppe Peano was the first to give the modern definition of vector spaces and linear maps, in 1888, although he called them linear systems. Peano's axiomatization allowed for infinite-dimensional vector spaces, but he did not develop that direction himself. In 1897, Salvatore Pincherle adopted Peano's axioms and made the first serious steps into the theory of infinite-dimensional spaces.

• 4. Dimension and Basis

  ›

  Every vector space has at least one basis: a subset whose elements are linearly independent and whose span fills the entire space. The remarkable fact, guaranteed by the dimension theorem for vector spaces, is that all bases of a given vector space share the same cardinality. That shared cardinality is the dimension. For a finite-dimensional space of dimension n over a field F, choosing a basis allows every vector to be written uniquely as a linear combination of the basis vectors. The scalars in that combination are the coordinates of the vector on that basis, and the map from vectors to their coordinate vectors is a vector space isomorphism. Two vector spaces over the same field are isomorphic if and only if they have the same dimension. This means that any n-dimensional vector space over F is isomorphic to Fn, the coordinate space of n-tuples. In the infinite-dimensional case, the existence of a basis, often called a Hamel basis, depends on the axiom of choice, and in general no such basis can be explicitly described. The real numbers, for example, form an infinite-dimensional vector space over the rational numbers for which no specific basis is known.

• 5. Arrows, Pairs, and Functions

  ›

  Arrows in a fixed plane, all starting from one fixed point, form the first canonical example of a vector space. Their addition is given by the parallelogram rule: the sum of two arrows is the diagonal of the parallelogram they span. Scaling an arrow stretches or shrinks it; a negative scalar reverses its direction. This example is the one used in physics to model forces and velocities. Pairs of real numbers form a second key example, with addition and scalar multiplication defined componentwise, and this example is isomorphic to the arrow space via Cartesian coordinates. Function spaces carry the concept much further: functions from any fixed set to a field form a vector space by performing addition and scalar multiplication pointwise. Many analytic properties, such as continuity, integrability, and differentiability, are preserved under these operations, so the sets of functions possessing those properties are themselves vector spaces. Systems of homogeneous linear equations connect directly to this picture: their solution sets are vector spaces, and a matrix condenses such a system into a single vector equation. The same structure appears for homogeneous linear differential equations, whose solution spaces are also vector spaces over the reals or complex numbers.

• 6. Infinite Dimensions and Completeness

  ›

  Henri Lebesgue's construction of function spaces was an important development in the theory of vector spaces, later formalized by Banach and Hilbert around 1920. The key issue in analysis is convergence: does a sequence of vectors approach a limit inside the space? A vector space is called complete if every Cauchy sequence converges to a limit within it. The vector space of polynomials on the unit interval, equipped with the topology of uniform convergence, is not complete, because continuous functions can be uniformly approximated by polynomials yet those limits need not themselves be polynomials. The space of all continuous functions on the same interval with that topology, by contrast, is complete. Banach spaces, introduced by Stefan Banach, are complete normed vector spaces. Lebesgue spaces, denoted Lp, consist of functions on a given domain whose p-th power is integrable; these spaces are complete, and this completeness is one justification for Lebesgue's integration theory over Riemann's. Hilbert spaces, named for David Hilbert, are complete inner product spaces. A key example has an inner product defined using the complex conjugate. In a Hilbert space, the Gram-Schmidt process constructs orthogonal bases, which are the Hilbert-space generalization of coordinate axes in finite-dimensional Euclidean space. The Stone-Weierstrass theorem and Fourier expansion are tools for approximating functions within Hilbert spaces, and the time-dependent Schrodinger equation in quantum mechanics has solutions, called wavefunctions, that live in a Hilbert space, with physical observables corresponding to eigenvalues of linear differential operators.

• 7. Structures Built on Top

  ›

  A vector space equipped with a bilinear multiplication between vectors becomes an algebra over a field. Polynomials form an algebra, the polynomial ring, because the product of two polynomials is again a polynomial; rings of polynomials and their quotients underpin algebraic geometry. Lie algebras are neither commutative nor associative, but their failure to be so is precisely controlled by anticommutativity and the Jacobi identity. The vector space of n-by-n matrices with the commutator as product is one example. The tensor product of two vector spaces V and W is a universal recipient of bilinear maps: any bilinear map from V times W into another vector space factors uniquely through it. This universal property is an instance of a method widely used in abstract algebra to define objects indirectly through the maps they receive. The tensor algebra on any vector space is constructed by concatenating tensor symbols and imposing distributivity; forcing two elements to be equal yields the symmetric algebra, while forcing a different relation yields the exterior algebra. Moving further, a vector bundle is a family of vector spaces parametrized continuously by a topological space, and properties of bundles encode topological information. The hairy ball theorem, for instance, follows from the fact that there is no nonzero tangent vector field on the 2-sphere S2. K-theory studies isomorphism classes of vector bundles and has purely algebraic consequences, including the classification of finite-dimensional real division algebras: R, C, the quaternions H, and the octonions O.