2.15 Tensor Finite Dimensional Context

Tensor Finite Dimensional Context is the standing assumption, adopted throughout elementary and intermediate tensor algebra, that the vector space V underlying a tensor construction has finite dimension n. This assumption is not a mere simplification of notation; it is the structural condition that makes many of the identifications and counting arguments used in tensor algebra valid. Under finite dimensionality, a vector space becomes naturally isomorphic to the dual of its dual, tensor product spaces acquire dimensions that are simply the product of the dimensions of their factors, and every tensor can be written as a finite sum of basis tensor products with definite components. None of these facts hold without qualification once the finite-dimensional restriction is dropped.

Why Finite Dimensionality Matters

The Double Dual Isomorphism

For a finite-dimensional vector space V over a field F, the dual space V*, consisting of all linear functionals V → F, has the same dimension as V. Applying the dual construction a second time produces V**, the dual of the dual, and finite dimensionality guarantees that the canonical evaluation map

defined by σ(v)(f) = f(v) for v in V and f in V*, is a linear isomorphism. This map exists for every vector space, finite-dimensional or not, but it is only bijective when V is finite-dimensional. It is this isomorphism, often written V ≅ V**, that licenses the practice of treating vectors and "double dual vectors" as interchangeable, which in turn allows tensors to be described equally well as multilinear maps or as elements of a tensor product.

Basis Existence and Counting

Every finite-dimensional vector space has a finite basis, and any two bases have the same cardinality, equal to the dimension n. This gives a concrete, countable scaffold: a basis e_1, ⋯, e_n of V together with the dual basis e^1, ⋯, e^n of V*, satisfying e^i(e_j) = δ^i_j, where δ^i_j is the Kronecker delta. In an infinite-dimensional space, a basis in the linear-algebraic sense (a Hamel basis) still exists by the axiom of choice, but it is uncountable for most spaces of interest, has no natural dual basis, and cannot be used for the kind of finite index summation that tensor components rely on.

Consequences for Tensor Construction

Tensor Product Dimension Formula

When V and W are finite-dimensional with dimensions m and n, the tensor product V ⊗ W has dimension m × n, and a basis is given explicitly by all products e_i ⊗ f_j of basis vectors from each factor:

This formula extends to iterated tensor products, so a type (p, q) tensor space T^p_q(V) built from p copies of V and q copies of V* has dimension n^(p+q). The finite-dimensional context is what turns this into a finite number rather than an infinite cardinal, and what guarantees that the space of tensors is itself finite-dimensional and hence amenable to matrix and array representations.

Basis of Tensor Powers and Component Arrays

Because V has a finite basis, every tensor in T^p_q(V) decomposes uniquely as a finite linear combination of basis tensor products e_{i_1} ⊗ ⋯ ⊗ e_{i_p} ⊗ e^{j_1} ⊗ ⋯ ⊗ e^{j_q}, with indices ranging over 1 through n. The coefficients of this decomposition are exactly the tensor's components, T^{i_1 ⋯ i_p}_{j_1 ⋯ j_q}, and the finiteness of the index range is what makes the Einstein summation convention, contraction, and explicit component-wise transformation laws well defined operations rather than formal or symbolic ones.

Comparison with Infinite-Dimensional Settings

Failure of the Natural Double Dual Isomorphism

When V is infinite-dimensional, the evaluation map σ : V → V** remains injective but is never surjective; the algebraic dual V* is strictly larger than V itself, and V** is larger still. This means a general element of V** cannot be identified with any vector in V, so constructions that freely swap V and V**, as is routine in the finite-dimensional context, become unavailable or require additional structure, such as a choice of topology and a restriction to continuous functionals, before any analogous identification can be recovered.

Algebraic Duals Versus Continuous Duals

In infinite-dimensional settings that arise in analysis, such as Hilbert or Banach spaces, the relevant dual is usually the continuous dual, consisting only of bounded linear functionals, rather than the full algebraic dual of all linear functionals. Even then, results such as reflexivity, where the continuous double dual coincides with the original space, hold only for special classes of spaces and require analytic hypotheses that have no counterpart in the purely algebraic, finite-dimensional theory of tensors.

Role in the Definition of Tensors

Multilinear Maps Perspective

Under the finite-dimensional context, a type (p, q) tensor can be defined as a multilinear map taking p covectors and q vectors to a scalar, and this definition is equivalent to the tensor product definition precisely because V ≅ V**. Without finite dimensionality, the multilinear-map definition and the tensor-product definition can diverge, since the space of multilinear maps may be strictly larger than the corresponding tensor product space.

Universal Property Perspective

The tensor product's defining universal property, that every bilinear map out of V × W factors uniquely through V ⊗ W, holds regardless of dimension. What the finite-dimensional context adds is the concrete, computable description of V ⊗ W as a space of explicit dimension m × n with an explicit basis, turning an abstract universal construction into an object that can be manipulated with coordinates, matrices, and index notation.

Diagrammatic Summary

The diagram contrasts the finite-dimensional case, where a finite basis exists and the space is naturally isomorphic to its double dual, with the infinite-dimensional case, where a Hamel basis still exists in principle but the double dual strictly contains the original space. This contrast is the essential content of the finite dimensional context that underlies coordinate-based tensor algebra.