2.16.3 Tensor Infinite Expansion Context

Tensor Infinite Expansion Context is the framework for writing a vector or tensor in an infinite-dimensional space as a genuinely infinite sum of basis contributions, a operation that requires a topology, typically induced by a norm or inner product, in order for the sum to have a well-defined meaning as a limit, in contrast to the finite linear combinations that suffice under a purely algebraic Hamel basis. Where the algebraic coordinate system guarantees only finite support and never needs to ask whether a sum converges, the expansion context is specifically about series that do not terminate and about the precise sense in which they nonetheless add up to a definite vector.

From Finite Sums to Convergent Series

The Limiting Procedure

An infinite expansion of a vector v relative to a countable set {e_1, e_2, ...} is a formal series Σ c_k e_k, and this series is said to converge to v if the sequence of partial sums converges to v in whatever topology V carries, most commonly the topology induced by a norm.

meaning the norm of the difference between v and the N-th partial sum tends to zero as N grows without bound. This is fundamentally a statement about a topology, not merely about algebra, and it has no counterpart in the strictly finite-support Hamel coordinate framework.

Orthogonal Expansions in Inner Product Spaces

When V is a Hilbert space with a countable orthonormal set {e_k}, every vector v admits a canonical candidate expansion given by its Fourier coefficients, c_k = ⟨v, e_k⟩, and the resulting series converges to v precisely when the orthonormal set is complete, meaning no nonzero vector is orthogonal to every e_k.

with ⟨·,·⟩ denoting the inner product, and convergence understood in the norm induced by that inner product.

Parseval's Relation as a Consistency Check

Norm Preservation Under Expansion

Completeness of the orthonormal expansion is accompanied by Parseval's relation, which states that the squared norm of v equals the sum of the squared magnitudes of its expansion coefficients, providing a quantitative way to confirm that no "mass" of v is missing from the expansion.

This relation converts an abstract convergence statement into a concrete, checkable numerical identity, and it is the mechanism by which one verifies, for instance, that a Fourier series expansion accounts for the entirety of a given function.

Extending Expansion to Tensors

Expanding Tensors Built From Orthonormal Sets

If V and W are Hilbert spaces with orthonormal sets {e_k} and {f_l}, an element T of the completed (Hilbert) tensor product V ⊗̂ W admits an expansion as a doubly-indexed infinite sum, T = Σ_{k,l} t_{kl} e_k ⊗ f_l, convergent in the tensor product norm, with the coefficients t_{kl} satisfying a two-index analogue of square-summability, Σ_{k,l} |t_{kl}|² < ∞. This is the natural extension of vector expansion to the tensor setting, but it requires working in the topological completion rather than the purely algebraic tensor product, since the algebraic tensor product only ever contains finite sums.

Simple Tensors Versus General Expanded Tensors

A crucial distinction in the expansion context is that a general element T of the completed tensor product is not, in general, expressible as a single simple tensor v ⊗ w; it is instead an infinite superposition of simple tensors, and the number of terms genuinely needed in any such expansion (after diagonalizing, via the singular value decomposition in the Hilbert space setting) measures how far T is from being simple.

Diagram of Partial Sums Converging

Why the Expansion Context Is Distinct From the Basis Context

Two Different Roles for "Infinitely Many Terms"

The Hamel basis context guarantees a basis exists and that coordinates relative to it always have finite support, deliberately avoiding any question of convergence. The expansion context addresses the opposite situation: constructing genuinely infinite sums that converge to a target vector or tensor in a specified topology, using a countable basis-like set that is typically not a Hamel basis of the space (a complete orthonormal set spans only in the closure, topological sense). Keeping these two notions separate prevents the common error of treating a convergent orthonormal expansion as though it were an algebraic Hamel-basis decomposition, when the two rely on fundamentally different notions of "spanning."