3
$\begingroup$

Suppose $G$ is a sufficiently nice (maybe locally compact and abelian) group which acts on the separable Hilbert space $\mathcal{H}$ by unitary transformations. Is there a generalization of the spectral theorem to this context? Specifically, what ought the spectrum of such a group action be, and, do generalizations of spectral measures exist in this context?

This is motivated by the following line of thinking. A unitary operator $U$ naturally induces a $\mathbb{Z}$-action on $\mathcal{H}$ via $ k \cdot \psi = U^k \psi $. Then, one version of the spectral theorem produces Borel measures $\mu_\psi^U$ supported on $\mathbb{T}$ with the property that $$ \langle \psi, U^k \psi \rangle = \langle \psi, k \cdot \psi \rangle = \int_{\mathbb{T}} \! z^k \, d\mu_\psi^U(z). $$ for all $ \psi \in \mathcal{H}, k \in \mathbb{Z} $. There are a few observations that one can make. First, the spectrum of $U$ lives in the dual group $ \widehat{\mathbb{Z}} = \mathbb{T} $. Also, if $ E_k:\mathbb{T} \to \mathbb{T} $ is the map $E_k(z) = z^k$, then $ k \mapsto E_k $ is the usual Pontryagin isomorphism $ \mathbb{Z} \to \widehat{\mathbb{T}} $.

Similarly, a bounded self-adjoint operator $A$ acting on $\mathcal{H}$ induces a (continuous) unitary action of $\mathbb{R}$ on $\mathcal{H}$ via $ t \cdot \psi = e^{-itA} \psi $. Again, the spectral theorem produces measures $ \mu = \mu_\psi^A $ supported on $\mathbb{R}$ such that $$ \langle \psi, e^{-itA} \psi \rangle = \langle \psi, t \cdot \psi \rangle = \int_{\mathbb{R}} \! e^{-itx} \, d\mu_\psi^A(x), $$ and the spectrum lives in the dual group $ \widehat{\mathbb{R}} = \mathbb{R} $. This leads me to wonder the following: given a (sufficiently nice) action of a (sufficiently nice) group $G$ by unitary operators, do there exist measures $ \mu_\psi^G $ supported on $\widehat{G}$ such that $$ \langle \psi, g \cdot \psi \rangle = \int_{\widehat{G}} \! E_g(\chi) \, d\mu_\psi^G(\chi)? $$ As above, $E_g(\chi) = \chi(g)$.

This seems like a fairly reasonable question to ask, but I haven't been able to find a discussion of things like this in any of my texts on spectral theory (or any of my colleagues' texts on unitary representations).

Remark: In case this is nonstandard notation, $ \widehat{G} $ denotes the multiplicative group of continuous group homomorphisms $ G \to \mathbb{T} $, and $ \mathbb{T} = \{ z \in \mathbb{C} : |z| = 1\} $.

$\endgroup$
5
$\begingroup$

This is true for locally compact abelian $G$; you ought to be able to find it in any text on abstract harmonic analysis (a reference I have at hand is Theorem 4.44 in Folland's A Course in Abstract Harmonic Analysis).

Here's a bit of general perspective that may be helpful. Unitary representations of $G$ are the same as representations of the group C*-algebra $C^*(G)$ (essentially by definition of $C^*(G)$), and the spectrum of $C^*(G)$ can be identified with $\widehat{G}$. By the classification of commutative C*-algebras, this identifies $C^*(G)$ with $C_0(\widehat{G})$, the algebra of functions on $\widehat{G}$ vanishing at infinity. For any locally compact Hausdorff space $X$, there is a spectral theorem for representations of $C_0(X)$ whose proof is not much harder than the case of a single normal operator (which is just the case when $X$ is a compact subset of $\mathbb{C}$).

$\endgroup$
3
$\begingroup$

It sounds like you want the Arveson spectrum.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.