4.17 Tensor Universal Multilinear Property

Tensor Universal Multilinear Property is the characterization of the tensor product of vector spaces by its relationship to every possible multilinear map defined on those same spaces, stating that any multilinear map factors uniquely through the canonical tensor multilinear input map, followed by a single linear map out of the tensor product itself. It is the defining property that singles out the tensor product, among all conceivable vector spaces that might receive a multilinear map from V_1, ..., V_k, as the unique space through which every multilinear map universally and uniquely factors.

Statement of the Universal Property

The Precise Requirement

For finite-dimensional vector spaces V_1, ..., V_k, the tensor product V_1 ⊗ ⋯ ⊗ V_k, together with the tensor universal property input map ι, satisfies: for every vector space W and every multilinear map

there exists a unique linear map ℓ : V_1 ⊗ ⋯ ⊗ V_k → W such that M = ℓ ∘ ι. This single statement combines an existence claim, that some such ℓ can always be found, with a uniqueness claim, that only one such ℓ works.

The Property Determines the Tensor Product Up to Isomorphism

Any two vector spaces, each equipped with a multilinear map from V_1 × ⋯ × V_k, satisfying this same universal factorization property, must be isomorphic to each other via an isomorphism compatible with the respective input maps; the universal property therefore characterizes the tensor product completely, up to a canonical isomorphism, without reference to any particular construction of the tensor product as a quotient of a free vector space or as any other specific model.

Existence: Constructing the Linear Map

Defining ℓ on Elementary Tensors

Existence of ℓ is established by defining it first on elementary tensors, setting ℓ(v_1 ⊗ ⋯ ⊗ v_k) = M(v_1, ..., v_k), and then extending linearly to sums of elementary tensors, which span the tensor product space; this construction relies on tensor extension compatibility to confirm that the definition of ℓ does not depend on how a given element of the tensor product happens to be written as a sum of elementary tensors.

Compatibility Guaranteed by the Defining Relations

The relations used to construct the tensor product, such as (u + w) ⊗ v_2 ⊗ ⋯ = u ⊗ v_2 ⊗ ⋯ + w ⊗ v_2 ⊗ ⋯, are exactly the relations that the multilinearity of M already respects, so the compatibility condition required to define ℓ consistently is automatically satisfied by any multilinear M, which is precisely why the universal property holds for every multilinear map without exception.

Uniqueness: No Alternative Choice of ℓ

Forced Agreement on Elementary Tensors

Since M = ℓ ∘ ι requires ℓ(v_1 ⊗ ⋯ ⊗ v_k) = M(v_1, ..., v_k) for every tuple, the values of ℓ on elementary tensors are completely forced by M; because elementary tensors span the tensor product space, tensor map determination by basis values, applied here to the spanning set of elementary tensors rather than strictly to a basis, guarantees that no two linear maps agreeing on all elementary tensors can differ anywhere on the tensor product space.

Uniqueness as a Direct Consequence of Spanning

The uniqueness of ℓ therefore rests entirely on the fact that elementary tensors span the tensor product space, connecting the universal property directly to tensor extension uniqueness, which guarantees that any linear map agreeing with a prescribed assignment on a spanning set is the only such linear map possible.

Consequences of the Universal Property

Multilinear Maps Correspond to Linear Maps

The universal property establishes a natural, bijective correspondence between multilinear maps out of V_1 × ⋯ × V_k and linear maps out of V_1 ⊗ ⋯ ⊗ V_k, converting the generally more complicated study of multilinear maps into the simpler and more thoroughly understood study of ordinary linear maps, once the tensor product has been formed.

Basis for Defining Operations on Tensors

Many standard tensor operations, including tensor contraction and the identification underlying the tensor vector valued tensor role, are defined precisely by invoking the universal property: a natural multilinear map is written down informally, and the universal property is invoked to promote it automatically to a genuine, well-defined linear map on the appropriate tensor product space.

The Universal Property and Basis-Free Reasoning

Independence from Any Chosen Basis

The universal property is stated without reference to any basis of V_1, ..., V_k, or of W; this basis-free character is what allows conclusions drawn from the universal property, such as the existence and uniqueness of a corresponding linear map, to hold regardless of how any of the vector spaces involved happen to be coordinatized.

Recovering Component Formulas as a Special Case

When bases are introduced for computational convenience, the abstract linear map ℓ obtained from the universal property reduces to exactly the component evaluation formulas already developed for ordinary tensor multilinear component evaluation, confirming that the universal property and the earlier, more computational treatment of tensors describe the same underlying mathematical objects from two complementary points of view.

Diagrammatic Summary

The diagram shows the universal property as a triangle that always closes uniquely: every multilinear map M out of the product of the vector spaces factors as a unique linear map ℓ composed with the fixed input map ι into the tensor product.