5.25.2 Tensor Product Space Boundary

Tensor Product Space Boundary is the precise delineation of which vector space actually qualifies as "the" tensor product V ⊗ W of two given spaces, as distinguished from several other naturally associated spaces — the direct sum, the space of bilinear forms, the Hom space, and the Cartesian product — that are easily confused with the tensor product but are not identical to it, and that in some cases coincide with it only under additional hypotheses such as finite-dimensionality. Marking this boundary prevents common misidentifications that arise from these spaces sharing superficial similarities with V ⊗ W.

Distinguishing V ⊗ W from the Direct Sum V ⊕ W

Different Dimension Formulas

The direct sum V ⊕ W has dimension dim(V) + dim(W), in contrast to the tensor product's dimension dim(V) · dim(W); whenever both dimensions exceed 1, these two spaces are of different sizes and cannot be identified, marking one of the clearest boundaries around what counts as the tensor product.

Different Universal Properties

The direct sum is characterized by a universal property for pairs of maps out of V and W separately (a coproduct), while the tensor product is characterized by a universal property for a single bilinear map from V × W; these are genuinely different categorical constructions, and no natural map identifies them except in trivial cases.

Distinguishing V ⊗ W from the Space of Bilinear Forms

Bil(V,W) versus V ⊗ W

The space Bil(V, W; F) of bilinear forms on V × W is naturally isomorphic not to V ⊗ W itself but to its dual, (V ⊗ W)* ≅ V* ⊗ W* (in finite dimensions), since a bilinear form is precisely a linear functional on the tensor product by the universal property; conflating V ⊗ W with the space of bilinear forms on it mistakes a space for its dual.

Why This Distinction Matters

Elements of V ⊗ W are tensors, while elements of Bil(V, W; F) are functionals that evaluate on tensors; only in finite dimensions, where a space and its double dual coincide, does this distinction become easy to blur, but the two spaces remain conceptually and constructively separate.

Diagram of Nearby but Distinct Spaces

Distinguishing V ⊗ W from Hom Spaces

The Finite-Dimensional Coincidence

In finite dimensions, there is a natural isomorphism V* ⊗ W ≅ Hom(V, W), sending φ ⊗ w to the rank-one (or sum of rank-one) linear map v ↦ φ(v)w; this isomorphism is genuine and useful, but it depends on V being finite-dimensional and specifically pairs the dual of one factor with the plain space of the other — it is not a statement that V ⊗ W in general "is" a Hom space.

Failure in Infinite Dimensions

When V is infinite-dimensional, V* ⊗ W still consists only of finite sums of simple tensors and is generally a proper subspace of Hom(V, W), which can contain linear maps not expressible as any finite sum of rank-one maps; the boundary between the tensor product and the Hom space becomes strict precisely in this infinite-dimensional regime.

Distinguishing V ⊗ W from the Cartesian Product V × W

Elements Are Fundamentally Different Kinds of Objects

An element of V × W is always a single ordered pair (v, w), with componentwise vector space operations and dimension dim(V) + dim(W). An element of V ⊗ W is, in general, a sum of several simple tensors, not a single pair; while (v₁+v₂) ⊗ w = v₁ ⊗ w + v₂ ⊗ w echoes the additive structure of the product in this special case, a general element of V ⊗ W need not reduce to any single simple tensor v ⊗ w at all, whereas every element of V × W is, by definition, exactly one pair.

The Bilinear Map Connecting Them

The canonical map ⊗ : V × W → V ⊗ W, (v,w) ↦ v ⊗ w, is the bilinear map whose existence and universal property define the tensor product in the first place; this map is neither injective nor surjective in general (once dimensions exceed 1), underscoring that V × W and V ⊗ W are related but distinct spaces, connected only through this specific bilinear map.

Significance of the Space Boundary

Avoiding Common Identification Errors

Explicitly marking the boundary between V ⊗ W and these nearby spaces — direct sum, bilinear forms, Hom spaces, Cartesian product — prevents a frequent source of error in which formulas or intuitions valid for one of these spaces are mistakenly applied to another, particularly given that some of these spaces do become isomorphic to V ⊗ W under specific finite-dimensional hypotheses.

Clarifying When Isomorphisms, Not Identities, Apply

Every genuine relationship between V ⊗ W and a nearby space (such as V* ⊗ W ≅ Hom(V,W) in finite dimensions) is an isomorphism holding under specific conditions, not a definitional identity; keeping this distinction sharp is essential to correctly tracking which properties transfer across the isomorphism and which do not.