5.19.3 Tensor Product Vector Space Dimension Relation

The tensor product combines vector spaces, with dimension equal to the product of their original dimensions.

Tensor Product Vector Space Dimension Relation is the precise formula, dim(V ⊗ W) = dim(V) · dim(W), governing how the dimension of a tensor product of two finite-dimensional vector spaces is determined by the dimensions of its factors, together with the proof of this formula and its consequences for finite- and infinite-dimensional cases alike. This relation is the single most consequential numerical fact about the tensor product, distinguishing it sharply from the direct sum's additive dimension formula and underlying essentially every counting argument involving tensors.

Statement of the Dimension Relation

The Finite-Dimensional Formula

For finite-dimensional vector spaces V and W over a field F, with dim(V) = m and dim(W) = n:

dim (V \otimes W) = m \cdot n

Contrast with the Direct Sum

This multiplicative relation stands in direct contrast to the additive dimension formula for the direct sum:

dim (V \oplus W) = m + n

underscoring that the tensor product is a fundamentally different combination of two vector spaces than the direct sum, despite both being commonly encountered binary operations on vector spaces.

Proof of the Dimension Relation

Constructing a Basis of the Tensor Product

Let {e₁, ..., eₘ} be a basis of V and {f₁, ..., fₙ} be a basis of W. The claim is that the mn simple tensors {eᵢ ⊗ fⱼ : 1 ≤ i ≤ m, 1 ≤ j ≤ n} form a basis of V ⊗ W.

Spanning Argument

Every simple tensor v ⊗ w can be expanded using additivity and homogeneity by writing v = Σ aᵢeᵢ and w = Σ bⱼfⱼ:

v \otimes w = \sum_{i, j} a_{i} b_{j} (e_{i} \otimes f_{j})

Since simple tensors span V ⊗ W, this shows {eᵢ ⊗ fⱼ} also spans V ⊗ W.

Linear Independence Argument

Linear independence of {eᵢ ⊗ fⱼ} is established using the universal property: for any fixed pair (k, l), define a bilinear map β_{kl}: V × W → F by β_{kl}(v, w) = a_k b_l where v = Σ aᵢeᵢ, w = Σ bⱼfⱼ; this induces a linear functional f_{kl}: V ⊗ W → F with f_{kl}(eᵢ ⊗ fⱼ) = δ_{ik}δ_{jl}. Applying f_{kl} to a hypothetical linear dependence Σ c_{ij}(eᵢ ⊗ fⱼ) = 0 isolates c_{kl} = 0 for every pair (k,l), proving independence.

Diagram of the Basis Grid

Consequences of the Dimension Relation

Rapid Growth Under Repeated Tensoring

Because dimension multiplies rather than adds, the dimension of an n-fold tensor power V^{⊗n} = V ⊗ V ⊗ ... ⊗ V grows exponentially:

dim (V^{\otimes n}) = dim (V)^{n}

a fact with major consequences for the computational cost of algorithms and physical models built from repeated tensor products, such as multi-particle quantum state spaces.

Recovering Matrix Space Dimensions

Since V ⊗ W is identified, once bases are chosen, with the space of m × n matrices, the dimension relation mn is simply the familiar fact that the space of m × n matrices has dimension mn, connecting the abstract tensor product dimension formula to elementary linear algebra.

The Infinite-Dimensional Case

Dimension as Cardinal Arithmetic

When V or W is infinite-dimensional, the dimension relation still holds in the sense of cardinal arithmetic, dim(V ⊗ W) = dim(V) · dim(W) using cardinal multiplication, since a basis of V ⊗ W is still given by {eᵢ ⊗ fⱼ} indexed over the (now possibly infinite) index sets for the bases of V and W.

Practical Caveat for Infinite Tensor Products

The dimension relation as stated applies to the tensor product of two spaces, not to an infinite tensor product of infinitely many spaces, which requires a separate and more delicate construction (such as a restricted or Hilbert-space tensor product) outside the scope of the ordinary bilinear tensor product described here.

Broader Significance

A Diagnostic for Correct Tensor Product Identification

The multiplicative dimension relation serves as an immediate diagnostic check: any proposed construction claiming to be "the tensor product" of two vector spaces but failing to produce a space of dimension dim(V) · dim(W) cannot in fact satisfy the universal property, providing a fast, computable way to rule out incorrect candidate constructions.

Foundation for Counting Arguments Throughout Multilinear Algebra

The dimension relation underlies counting arguments used throughout multilinear algebra and its applications, from determining the number of independent components of a physical tensor of a given rank and dimension, to bounding the parameter count of tensor-based models used in areas such as machine learning and quantum information theory.