3.11.4 Tensor Infinite Dual Dimension Caution

Tensor Infinite Dual Dimension Caution is a collection of practical warnings against carrying finite-dimensional intuitions about dual spaces uncritically into the infinite-dimensional setting, where the dimension mismatch between V and V* breaks several familiar constructions and can silently invalidate arguments that appear to generalize the finite case correctly. Recognizing these pitfalls in advance prevents common errors when tensor algebra techniques are applied to infinite-dimensional spaces such as sequence spaces or function spaces.

Pitfall: Assuming a Dual Basis Spans V*

The Naive Construction Fails to Span

Given an infinite basis e_1, e_2, ... of V, defining e^i(e_j) = δ^i_j for each i produces a linearly independent family in V*, but this family does not span the full algebraic dual space. These functionals can only ever act nontrivially on finitely many basis vectors at a time, whereas a general element of V* may assign nonzero values to infinitely many basis vectors simultaneously, placing it entirely outside the span of the naive dual family.

Correct Statement

The correct statement is that the naive family e^1, e^2, ... is a basis for a proper subspace of V*, sometimes called the restricted dual, consisting exactly of functionals with finite support relative to the basis, while the full algebraic dual V* is, by the Erdős–Kaplansky theorem, strictly larger.

Pitfall: Assuming V ≅ V*

Finite-Dimensional Isomorphism Does Not Generalize

In finite dimensions, V and V* are isomorphic simply because they share the same dimension. In infinite dimensions, this reasoning fails outright, since dim(V*) is always strictly greater than dim(V), so no isomorphism between them can exist at all, let alone a canonical one.

Consequence for Component Formulas

Formulas that implicitly rely on identifying vectors with covectors through matching components, valid and convenient in finite dimensions, do not carry over; any such identification in an infinite-dimensional setting must instead come from additional structure, such as an inner product on a Hilbert space, and even then applies only to a restricted class of functionals, not the entire algebraic dual.

Pitfall: Assuming V ≅ V** Directly Without Checking Reflexivity

Injectivity Without Surjectivity

The canonical map ι : V -> V** remains well-defined and injective in infinite dimensions, but it is not generally surjective. A vector space for which ι fails to be surjective is called non-reflexive, and most naturally occurring infinite-dimensional spaces of interest, including many common sequence and function spaces under the algebraic dual, are non-reflexive.

What Remains True

Even without surjectivity, every vector v in V still defines a legitimate functional ι(v) on V*, so evaluation of covectors on vectors, and the natural pairing itself, remain perfectly well-defined; only the stronger claim that this pairing exhausts all of V** fails.

Pitfall: Ignoring the Difference Between Algebraic and Topological Duals

Two Different Notions of Dual Space

In infinite-dimensional analysis, the term dual space more commonly refers to the topological dual, consisting only of continuous linear functionals with respect to some given topology, such as a norm, rather than the full algebraic dual consisting of all linear functionals regardless of continuity. These two notions generally differ substantially, and results proven about one do not automatically transfer to the other.

Better-Behaved Alternative

Many familiar finite-dimensional facts, including a form of dimension matching and, for Hilbert spaces specifically, a canonical identification with the dual via the Riesz representation theorem, do hold for the topological dual under suitable conditions, even though they fail for the full algebraic dual. When infinite-dimensional dual spaces must be used in applications, working with the topological dual under an appropriate topology is usually the more productive and better-behaved choice.

General Guidance

Verify Finiteness Before Applying Finite-Dimensional Identities

Before applying any identity that relies on dim(V*) = dim(V), the existence of a spanning dual basis, or a natural identification between V and V**, it is essential to confirm that V is genuinely finite-dimensional; none of these facts can be assumed to hold automatically once infinite dimension is in play.

Specify Which Dual Is Intended

When infinite-dimensional spaces are involved, always state explicitly whether the algebraic dual or the topological dual is meant, since the dimension structure, spanning properties, and reflexivity behavior differ sharply between the two, and much confusion in applied settings traces back to this distinction being left implicit.

Diagrammatic Summary

The diagram lists the principal cautionary pitfalls that arise when finite-dimensional dual space reasoning is applied without modification to infinite-dimensional vector spaces.