Canonical: https://knowlpedia.com/functional-analysis/rigged-hilbert-space/

# Rigged Hilbert space

A dense nuclear test space inside a Hilbert space, together with its continuous anti-dual.

A **rigged Hilbert space**, or **Gelfand triple**, is a chain
\[
\Phi\hookrightarrow H\hookrightarrow\Phi^\times
\]
in which \(H\) is a [Hilbert space](https://knowlpedia.com/linear-algebra/hilbert-space/), \(\Phi\) is a Hausdorff [nuclear locally convex space](https://knowlpedia.com/functional-analysis/nuclear-space/) continuously and densely embedded in \(H\), and \(\Phi^\times\) is its continuous anti-dual with the [strong dual topology](https://knowlpedia.com/functional-analysis/strong-dual/). Taking the [inner product](https://knowlpedia.com/linear-algebra/inner-product/) on \(H\) linear in its first variable, the second embedding sends \(h\) to the conjugate-linear functional \(\phi\mapsto\langle h,\phi\rangle_H\). Density makes this embedding injective.

## Test vectors and generalized vectors

The space \(\Phi\) carries a topology finer than the Hilbert norm topology, encoding regularity or decay. Elements of \(\Phi^\times\) act as generalized vectors, including [distributions](https://knowlpedia.com/functional-analysis/distribution/) that do not belong to \(H\). The Hilbert space lies between these two scales and identifies ordinary vectors with continuous anti-linear functionals on the test space.

## Operators and spectral analysis

If a [densely defined operator](https://knowlpedia.com/functional-analysis/densely-defined-operator/) on \(H\) preserves \(\Phi\) and acts continuously on its locally convex topology, duality extends it to \(\Phi^\times\). Generalized eigenvectors may then live in \(\Phi^\times\) even when the operator has no eigenvectors in \(H\). Nuclearity supplies the compactness and kernel-theorem structure used in generalized spectral expansions.

## Example and convention

The canonical example is
\[
\mathcal S(\mathbb R^n)\hookrightarrow L^2(\mathbb R^n)\hookrightarrow\mathcal S(\mathbb R^n)^\times,
\]
with [Schwartz functions](https://knowlpedia.com/functional-analysis/schwartz-space/) as test vectors and the [anti-linear version](https://knowlpedia.com/functional-analysis/continuous-antidual/) of [tempered distributions](https://knowlpedia.com/functional-analysis/tempered-distribution/) as generalized vectors. Complex conjugation of functional values identifies this with the conventional linear distribution space \(\mathcal S'(\mathbb R^n)\) conjugate-linearly.

Some authors use **rigged Hilbert space** for any continuous dense embedding \(\Phi\hookrightarrow H\), without requiring \(\Phi\) to be nuclear, and reserve **nuclear Gelfand triple** for the definition above. The nuclear convention is used here. For the smooth vectors of a finite-dimensional Lie-group representation, this extra condition is equivalent to the representation being [trace class](https://knowlpedia.com/lie-groups/nuclear-smooth-vectors-trace-class/).

## References

1. I. M. Gel′fand and N. Ya. Vilenkin, *Generalized Functions, Volume 4: Applications of Harmonic Analysis*, Academic Press, 1964; AMS Chelsea reprint. [AMS publisher record](https://bookstore.ams.org/CHEL/380.H). Relevant: Chapter I on nuclear spaces, rigged Hilbert spaces, and spectral analysis.
2. Michael Reed and Barry Simon, *Methods of Modern Mathematical Physics, Volume I: Functional Analysis*, Academic Press, 1972. [Elsevier DOI record](https://doi.org/10.1016/B978-0-12-585001-8.X5001-6). Relevant: Chapter V on locally convex spaces and distributions.
