5.25.3 Tensor Product Map Boundary

Tensor Product Map Boundary is the recognition that maps of the form f ⊗ g, built from a single linear map on each factor, occupy only a small and special subset of the full space of linear maps Hom(V ⊗ W, V′ ⊗ W′), and the delineation of exactly which linear maps between tensor product spaces do and do not arise as a tensor product of maps in this way. Just as simple tensors are a special subset of all tensors in V ⊗ W, tensor product maps f ⊗ g are a special, "decomposable" subset of all linear maps between tensor product spaces, and this boundary marks where the elementary tensor product of maps operation stops being sufficient to describe a general linear transformation.

The Space of All Maps Is Much Larger

Dimension Comparison

The space Hom(V ⊗ W, V′ ⊗ W′) has dimension dim(V)·dim(W)·dim(V′)·dim(W′), matching the dimension of Hom(V,V′) ⊗ Hom(W,W′) (in finite dimensions, these two spaces are canonically isomorphic); however, the maps actually expressible as a single f ⊗ g correspond only to simple tensors within Hom(V,V′) ⊗ Hom(W,W′), exactly mirroring the relationship between simple tensors and general tensors one level up.

Most Linear Maps Are Not of Product Form

A general element of Hom(V,V′) ⊗ Hom(W,W′) is a sum Σ fₖ ⊗ gₖ of several product maps, and — exactly as with simple versus general tensors — most such elements cannot be rewritten as a single f ⊗ g; the boundary of tensor product map notation is precisely the boundary between rank-one and higher-rank elements of this Hom-tensor-product space.

Diagram of the Map Boundary

Matrix Perspective: Kronecker Product versus General Matrix

The Kronecker Product Occupies a Thin Slice

In finite dimensions, f ⊗ g corresponds to the Kronecker product matrix A ⊗ B; but an arbitrary matrix of the same overall size representing some map V ⊗ W → V′ ⊗ W′ need not be expressible as any Kronecker product A ⊗ B at all, since Kronecker product matrices have a highly constrained block structure (each block a scalar multiple of a single fixed matrix B) that a generic matrix does not share.

Testing Whether a Given Map Has Product Form

Determining whether a specific linear map between tensor product spaces happens to equal some f ⊗ g is a nontrivial recognition problem, generally requiring checking that the matrix's block structure matches the required scalar-multiple pattern exactly — a condition satisfied only on a very restrictive subset of all possible matrices of the corresponding shape.

Physical and Applied Manifestation: Entangling Maps

Product Operators versus Entangling Operators

In settings such as quantum mechanics, where composite systems are modeled by tensor product spaces, an operator of the form f ⊗ g acts independently on each subsystem and is called a (local or product) operator, while a general operator not of this form can correlate or "entangle" the two subsystems in ways no combination of independent local actions on each factor could replicate.

The Boundary as a Meaningful Physical Distinction

This is precisely the tensor product map boundary manifesting in an applied setting: whether a physically realizable transformation is a product map f ⊗ g or a genuinely non-decomposable map on V ⊗ W corresponds directly to whether the transformation can or cannot be implemented by acting on each subsystem completely independently.

Closure Properties Across the Boundary

Sums and Compositions of Product Maps

While a single map f ⊗ g lies within the special "product" subset, sums of several such maps, and compositions of several such maps, generally do not remain within that subset (a sum typically does not, though a composition (f₁ ⊗ g₁) ∘ (f₂ ⊗ g₂) = (f₁f₂) ⊗ (g₁g₂) does remain a product map); recognizing which operations preserve product form and which do not is essential to correctly reasoning about maps near this boundary.

Significance of the Map Boundary

Extending the Simple-versus-General Distinction to Maps

The map boundary generalizes, to the level of linear transformations, the same fundamental distinction already seen for tensor elements: a small, well-structured subset (simple tensors, product maps) versus the much larger, generically unstructured full space (general tensors, arbitrary linear maps) that properly contains it.

Marking the Limits of Elementary Tensor Product of Maps Theory

Recognizing this boundary clarifies that the tensor product of linear maps operation f ⊗ g, however important as a building block, does not by itself describe every linear transformation one might wish to apply to a tensor product space, and that additional theory (sums of product maps, general Hom-space elements, entanglement structure) is needed to describe the full range of such transformations.