5.10.2 Tensor Product Factor Dimension Relation

Tensor Product Factor Dimension Relation is the collection of structural relationships linking the individual dimensions of the factor spaces in a tensor product to properties of the product space beyond its own total dimension — including monotonicity under subspace inclusion, the bound each factor dimension places on tensor rank, and the way shrinking or enlarging a single factor propagates through the whole construction.

Monotonicity Under Subspace Inclusion

If a factor space is replaced by a subspace, the tensor product changes in a predictable, dimension-controlled way rather than arbitrarily.

Embedding of Tensor Products of Subspaces

If V1′ is a subspace of V1, the tensor product V1′ ⊗ V2 ⊗ ⋯ ⊗ Vn embeds naturally as a subspace of V1 ⊗ V2 ⊗ ⋯ ⊗ Vn, with the embedding induced directly by the inclusion of V1′ into V1 combined with the identity on the remaining factors. This embedding respects dimension exactly:

so a strictly smaller factor dimension in any single position produces a strictly smaller, proportionally reduced tensor product dimension, holding the remaining factor dimensions fixed.

Consistency Across Multiple Substitutions

Because the embedding described above can be applied independently, and in either order, to more than one factor at a time, shrinking several factors to subspaces simultaneously produces an embedded tensor product whose dimension is the product of the reduced dimensions, matching what a direct application of the general dimension multiplication law to the smaller factor spaces would predict on its own.

Factor Dimension as a Bound on Tensor Rank

Each individual factor dimension imposes a hard limit on how large the tensor rank of an element in the tensor product can be, independent of the total ambient dimension.

The Minimal Factor Bound

For a two-factor tensor product V1 ⊗ V2, the rank of any element is bounded above by min(dim V1, dim V2), since the tensor corresponds to a matrix whose rank cannot exceed the smaller of its two dimensions. This is a direct, elementary consequence of ordinary matrix rank theory, expressed here in terms of the two factor dimensions rather than the total tensor product dimension d1 d2.

Extension to More Factors via Flattening

For n factors grouped into two blocks by any partition, the same bound applies to the corresponding flattening: tensor rank is bounded by the smaller of the two flattened dimensions, and taking the tightest such bound over all partitions of the factors into two blocks gives the sharpest rank bound obtainable purely from the individual factor dimensions, without further structural information about the specific tensor in question.

Effect of Enlarging a Factor

Just as shrinking a factor to a subspace produces a predictable embedding, enlarging a factor space produces a predictable, dimension-tracked extension of the tensor product.

Extension by a Larger Ambient Space

If V1 is replaced by a larger space V1″ containing it, the original tensor product V1 ⊗ V2 ⊗ ⋯ embeds into V1″ ⊗ V2 ⊗ ⋯, and the dimension of the enlarged tensor product grows by exactly the factor dim(V1″)/dim(V1) relative to the original, again following the multiplicative dimension law applied to the single changed factor while the others remain fixed.

Padding and Zero Extension

A common practical instance of this relation occurs when a factor space is enlarged by adjoining additional basis directions in which the original data has no support (zero-padding); the tensor product dimension grows accordingly, but every original tensor embeds unchanged, with zero coefficients on every newly introduced basis tensor involving the padded directions, illustrating that the factor dimension relation governs not just abstract dimension counts but the concrete coordinate structure of embedded tensors.

Isomorphism Invariance of the Relation

Because the tensor product dimension formula depends only on the dimensions of the factors, not on any other property of the spaces themselves, isomorphic factor spaces of the same dimension always produce isomorphic tensor products, and the factor dimension relation is entirely blind to any distinction between two factor spaces beyond their dimension.

Basis-Free Statement

This isomorphism invariance means that the factor dimension relation, unlike statements about specific bases or specific subspace embeddings, is a purely numerical relation among the integers d1, …, dn; any two families of factor spaces sharing the same dimension sequence produce tensor products related by a canonical isomorphism once bases are chosen compatibly, with the dimension relation as the only numerical data required to describe the outcome.

Illustrative Diagram

The solid rectangle marks the tensor product of a half-dimensional subspace V1′ with V2, embedded inside the full dashed tensor product region, its area — dimension — shrinking in direct proportion to the shrinkage of the single altered factor.