Let $\Gamma$ be a countable discrete group and $\lambda$ the left regular representation of $\Gamma$ on $l^2(\Gamma)$. Let $\rho:\Gamma\rightarrow U(H)$ be a unitary representation of $\Gamma$ on some separable infinite-dimensional Hilbert space $H$, and consider the representation $\lambda\otimes\rho$ of $\Gamma$ on $l^2(\Gamma)\otimes H$.

Now let $H_0$ be the same Hilbert space as $H$ but equipped with the trivial representation of $\Gamma$; then one can show that, as $\Gamma$-representations,

$$l^2(\Gamma)\otimes H\cong l^2(\Gamma)\otimes H_0,$$

where $\Gamma$ still acts by the left regular representation on the first factor on the right. For example one isomorphism is the map $\delta_g\otimes u\mapsto\delta_g\otimes(\rho(g)\cdot u)$, where $\delta_g:\Gamma\rightarrow\mathbb{C}$ is the function taking value $1$ at $g\in\Gamma$ and $0$ elsewhere.

I would like to know whether this can be generalised to the case of $G$ a Lie group instead of $\Gamma$. More specifically:

Question 1: Given a unitary representation $\rho:G\rightarrow U(H)$, is there an isomorphism $$L^2(G)\otimes H\cong L^2(G)\otimes H_0,$$ where $G$ acts on $L^2(G)$ by the left regular representation and $H_0$ is the same space as $H$ but with the trivial representation?

I'm not sure if the left-regular representation of $G$ on $L^2(G)$ is the right representation to use here, since the proof in the discrete case relies on the fact that the $\delta_g$ form a basis for $l^2(\Gamma)$, which doesn't seem to have an analogue in the continuous case. If this is hard to establish, I would also like to ask:

Question 2: Can one replace $L^2(G)$ by another representation in question 1 so that the conclusion is true?


1 Answer 1


Fell absorption works for all locally compact groups and all strongly continuous unitary representations; the intertwining map that you give in the discrete case can be generalized to

$$ W: L^2(G)\otimes_2 H \to L^2(G)\otimes_2 H $$

given by

$$ \langle W(\xi\otimes f), (\eta\otimes g) \rangle := \int_G f(s)\overline{g(s)} \langle \pi(s)\xi,\eta\rangle_H $$

(my inner products being linear in the first variable.)

If you are worried about this being well-defined on the whole of $L^2(G)\otimes_2 H$: first show that it is well-defined for $\xi,\eta \in C_c(G)$; then show by direct calculation that $$ \Vert W(\xi\otimes f) \Vert^2 = \langle W(\xi\otimes f), W(\xi\otimes f) \rangle = \Vert \xi\otimes f\Vert^2$$ for all $\xi\in C_c(G)$ and all $f\in H$; so we have a linear map defined on a dense subspace which is isometric.


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.