Topological tensor product

The topological tensor product of two topological vector spaces X and Y is a topological vector space X ⊗_τ Y together with a continuous bilinear map from X × Y such that every continuous bilinear map from X × Y to a third space factors uniquely through it. For normed spaces, the algebraic tensor product can be endowed with many crossnorms; the two most important are the projective (or π) norm and the injective (or ε) norm, giving the projective and injective tensor products respectively.¹

For Hilbert spaces, the tensor product is unique up to isomorphism and is itself a Hilbert space with the inner product defined on elementary tensors. For Banach spaces, the projective tensor product is the largest crossnorm, while the injective is the smallest, and they coincide only in finite-dimensional cases.²

The topological tensor product is central to the theory of vector-valued functions and distributions. For example, the space of continuous functions on a product of compact spaces is isometrically isomorphic to the injective tensor product of the corresponding function spaces. Similarly, the space of distributions on a product domain is a topological tensor product of distribution spaces, enabling the rigorous treatment of partial differential equations.

In operator algebra theory, the tensor product of C*-algebras and von Neumann algebras relies on these constructions; the spatial and maximal tensor products are specific cases. The injective tensor product is used to define nuclear operators and nuclear spaces, which have applications in the theory of integral equations and in the structure theory of Banach spaces.³

Beyond the classical projective and injective norms, there exist uncountably many crossnorms on the algebraic tensor product of infinite-dimensional Banach spaces, a fact established by Grothendieck in his thesis. This work led to the concept of nuclear spaces and the approximation property, which is a subtle invariant: not every Banach space has it, and its failure was a major discovery.⁴

The topological tensor product also appears in the theory of locally convex spaces, where the inductive and projective topologies can be defined. In the context of Fréchet spaces, the tensor product is not always complete, leading to the use of completed tensor products. A notable edge case is the tensor product of the space of test functions with itself, which yields the space of test functions on the product domain, a result that is not trivial and relies on the nuclearity of the space.⁵

The concept was introduced by John von Neumann in the 1920s for Hilbert spaces, and later extended by Murray and von Neumann to von Neumann algebras. Alexander Grothendieck's 1955 thesis "Produits tensoriels topologiques et espaces nucléaires" laid the foundation for the modern theory, introducing the projective and injective tensor products and the notion of nuclear spaces.⁴

Today, topological tensor products are used in quantum information theory, where the tensor product of Hilbert spaces describes composite quantum systems, and in the theory of operator spaces, where the tensor product is refined to account for the operator space structure. The theory continues to evolve, with applications in noncommutative geometry and the study of Banach algebras.⁶

Crossnorm

A norm on the algebraic tensor product of normed spaces that satisfies the property ||x⊗y|| = ||x||·||y||.

Projective tensor product

The tensor product equipped with the largest crossnorm, often denoted by ⊗̂_π.

Injective tensor product

The tensor product equipped with the smallest crossnorm, often denoted by ⊗̂_ε.

Nuclear space

A locally convex space whose projective and injective tensor products with any Banach space coincide.