5.12.1 Tensor Product Bilinear Input Pair

Tensor Product Bilinear Input Pair is the ordered pair (v, w), drawn from the Cartesian product V × W of two vector spaces, that serves as the raw input to the tensor product map, standing in direct contrast to the single combined tensor v ⊗ w that the map produces as output, and carrying its own distinct algebraic structure — that of a Cartesian product, not a tensor product — before the tensoring operation is applied.

The Pair as a Cartesian Product Element

For vector spaces V and W over a field F, the bilinear input pair (v, w) is simply an element of the set-theoretic (and vector-space) Cartesian product V × W, equipped with its own componentwise vector space structure:

This componentwise structure is the ordinary direct product (equivalently, in the finite case, direct sum) vector space structure on V × W, which is entirely different from, and must not be confused with, the tensor product structure that V ⊗ W carries once the pair is passed through the tensor product map.

Distinguishing the Pair from Its Image

The bilinear input pair and the tensor it produces are conceptually and structurally distinct objects, related only by the (non-linear, merely bilinear) tensor product map.

The Pair Lives in a Linear Space of Different Dimension

The space V × W has dimension dim(V) + dim(W) (additive), while V ⊗ W has dimension dim(V) · dim(W) (multiplicative); the pair (v, w) and the tensor v ⊗ w therefore live in ambient spaces of generically different size, underscoring that they are not the same kind of mathematical object merely written differently.

The Map from Pairs to Tensors Is Not Linear

The tensor product map τ(v, w) = v ⊗ w is bilinear but not linear on V × W as a whole: τ((v1, w1) + (v2, w2)) generally does not equal τ(v1, w1) + τ(v2, w2), since expanding the left side using bilinearity of τ produces the cross-terms v1 ⊗ w2 and v2 ⊗ w1 in addition to the two terms on the right. The bilinear input pair, considered as an element of the linear space V × W, is thus acted on by τ in a genuinely nonlinear way, even though τ is linear in each coordinate of the pair separately.

Role of the Pair in the Universal Property

The bilinear input pair is precisely the kind of object that a bilinear map, by definition, takes as its argument, making the Cartesian product V × W the natural domain against which the universal property of the tensor product is stated.

Bilinear Maps Take Pairs, Linear Maps Take Tensors

A bilinear map B : V × W → U takes a bilinear input pair (v, w) directly as its argument, while the corresponding linear map f : V ⊗ W → U guaranteed by the universal property instead takes the single combined tensor v ⊗ w; the identity B(v, w) = f(v ⊗ w) is precisely the equation connecting evaluation on the input pair to evaluation on its image under τ.

The Pair as the Unit of Bilinear Specification

Because a bilinear map is fully determined by its values on all bilinear input pairs (indeed, by its values on pairs of basis vectors, extended by bilinearity), specifying a bilinear map reduces to specifying a value for every input pair consistent with the bilinearity constraints — the input pair is thus the natural unit in terms of which bilinear maps are defined and verified.

Pairs Versus Tensors in Computation

Working directly with bilinear input pairs, rather than immediately passing to their tensor product images, is sometimes preferable computationally, since the pair retains the individual factor structure that the combined tensor discards.

Retaining Factor-Level Structure

An algorithm that needs to apply operations to v and w separately — for instance, normalizing each vector independently, or applying different linear transformations to each — naturally operates on the bilinear input pair (v, w) rather than on the combined tensor v ⊗ w, only forming the tensor product at the point where the combined object is actually required.

Pairs as the Interface to External Bilinear Maps

Any external bilinear map B, such as an inner product pairing or a cross product in three dimensions, is naturally described and evaluated directly on bilinear input pairs; the tensor product perspective becomes relevant only when one wants to factor B uniformly through a single universal linear map, as guaranteed by the universal property.

Illustrative Diagram

The bilinear input pair, an element of the additive Cartesian product V × W, passes through the tensor product map τ to become the combined tensor, an element of the multiplicatively larger space V ⊗ W.