Definition 32.2.1 (Representation of $L^{1}(G; \complex)$).label Let $G$ be a locally compact group and $(H, \pi_{0})$ be a unitary representation of $G$, then

  1. (1)

    The mapping

    \[\pi: L^{1}(G; \complex) \to B(H) \quad \dpn{\pi(f)\xi, \eta}{H}= \int_{G} f(x)\dpn{\pi_0(x)\xi, \eta}{H}dx\]

    defines a non-degenerate *-representation of $L^{1}(G; \complex)$.

  2. (2)

    For any $x \in G$ and $f \in L^{1}(G; \complex)$,

    \[\pi_{0}(x)\pi(f) = \pi(L_{x}f) \quad \pi(f)\pi_{0}(x) = \Delta_{G}(x^{-1})\pi(R_{x^{-1}}f)\]

Conversely, if $\pi: L^{1}(G; \complex) \to B(H)$ is a non-degenerate *-representation of $L^{1}(G; \complex)$, then:

  1. (3)

    There exists a unique unitary representation $\pi_{0}: G \to B(H)$ such that

    \[\dpn{\pi(f)\xi, \eta}{H}= \int_{G} f(x)\dpn{\pi_0(x)\xi, \eta}{H}dx\]

    for all $f \in L^{1}(G; \complex)$ and $\xi, \eta \in H$. Moreover, for any approximate identity $\angles{\phi_\alpha}_{\alpha \in A}\subset L^{1}(G; \complex)$ and $x \in G$, $\pi_{0}(x) = \sotlim_{\alpha \in A}\pi(L_{x}\phi_{\alpha})$.

The mapping $\pi: L^{1}(G; \complex) \to B(H)$ is the representation of $L^{1}(G; \complex)$ determined by $\pi_{0}$.

Proof, [Theorem 3.9, Theorem 3.11, Fol16]. (1, Well-Defined): For each $x \in G$, $\pi_{0}(x)$ is unitary, so for any $f \in L^{1}(G; \complex)$ and $\xi, \eta \in H$,

\[\abs{\dpn{\pi(f)\xi, \eta}{H}}\le \int_{G} \abs{f(x)\dpn{\pi_0(x)\xi, \eta}{H}}dx \le \norm{\xi}_{H}\norm{\eta}_{H}\norm{f}_{L^1(G; \complex)}\]

and the representation is well-defined.

(1, Homomorphism): For any $f, g \in L^{1}(G; \complex)$ and $\xi, \eta \in H$, since $\pi_{0}$ is a unitary representation,

\begin{align*}\dpn{\pi(f * g)\xi, \eta}{H}&= \int_{G} (f * g)(x)\dpn{\pi_0(x)\xi, \eta}{H}dx \\&= \iint_{G \times G}f(y)g(y^{-1}x)\dpn{\pi_0(x)\xi, \eta}{H}dydx \\&= \int_{G}f(y)\int_{G}g(x)\dpn{\pi_0(y)\pi_0(x)\xi, \eta}{H}dxdy\\&= \int_{G}f(y)\int_{G}g(x)\dpn{\pi_0(x)\xi, \pi_0(y^{-1})\eta}{H}dxdy \\&= \int_{G} f(y) \dpn{\pi(g)\xi, \pi_0(y^{-1})\eta}{H}dy\\&= \int_{G} f(y) \dpn{\pi_0(y)\pi(g)\xi,\eta}{H}dy = \dpn{\pi(f)\pi(g)\xi, \eta}{H}\end{align*}

(1, *-Homomorphism): For any $f \in L^{1}(G; \complex)$ and $\xi, \eta \in H$, since $\pi_{0}$ is unitary,

\begin{align*}\dpn{\pi(f^*)\xi, \eta}{H}&= \int_{G} \Delta_{G}(x^{-1})\ol{f(x^{-1})}\dpn{\pi_0(x)\xi, \eta}{H}dx \\&= \int_{G} \ol{f(x)}\dpn{\pi_0(x^{-1})\xi, \eta}{H}dx = \int_{G} \ol{f(x)}\dpn{\xi, \pi_0(x)\eta}{H}dx \\&= \int_{G} \ol{f(x)}\cdot \ol{\dpn{\pi_0(x)\eta, \xi}{H}}dx = \overline{\int_G f(x)\dpn{\pi_0(x)\eta, \xi}{H} dx}\\&= \ol{\dpn{\pi(f)\eta, \xi}{H}}= \dpn{\xi, \pi(f)\eta}{H}\end{align*}

(1, Non-Degenerate): For any $\xi \in H \setminus \bracsn{0}$, since $\pi_{0}$ is strong-operator continuous, there exists $V \in \cn_{G}(1)$ such that $\norm{\pi_0(x)\xi - \xi}_{H} < \norm{\xi}_{H}$ for all $x \in V$. In which case, if $f = |V|^{-1}\one_{V}$[1], then for any $\eta \in H$,

\begin{align*}\dpn{\pi(f)\xi - \xi, \eta}{H}&= \frac{1}{|V|}\int_{V} \dpn{\pi_0(x)\xi - \xi, \eta}{H}dx \\ |\dpn{\pi(f)\xi - \xi, \eta}{H}|&\le \frac{1}{|V|}\int_{V} \norm{\pi_0(x)\xi - \xi}_{H}\norm{\eta}_{H}dx < \norm{\xi}_{H}\norm{\eta}_{H}\end{align*}

so $\norm{\pi(f)\xi - \xi}_{H} < \norm{\xi}_{H}$, and $\pi(f)\xi \ne 0$.

(3): Using Proposition 31.4.10, let $\angles{\phi_\alpha}_{\alpha \in A}$ be an approximate identity in $L^{1}(G; \complex)$. For each $x \in G$ and $f \in L^{1}(G; \complex)$, $L_{x}\phi_{\alpha} * f \to L_{x} f$ in $L^{1}(G; \complex)$ implies that $\angles{\pi(L_x\phi_\alpha)}_{\alpha \in A}$ converges pointwise on

\[K := \text{span}\bracsn{\pi(g)\xi|\xi \in H, g \in L^1(G; \complex)}\]

Given that $\pi$ is non-degenerate, $K$ is dense in $H$. As $\bracsn{\pi(L_x\phi_\alpha)|\alpha \in A}$ is equicontinuous and $H$ is complete, the Arzelà-Ascoli Theorem implies that there exists $\pi_{0}(x) \in \ol{B_{B(H)}(0, 1)}$ such that $\pi(L_{x}\phi_{\alpha}) \to \pi_{0}(x)$ in the strong operator topology. In particular, $\pi_{0}(1) = \text{Id}$.

For each $f \in L^{1}(G; \complex)$, $\xi \in H$, and $x, y \in G$,

\[\pi_{0}(xy)\pi(f)\xi = \pi(L_{xy}f)\xi = \pi(L_{x}L_{y}f)\xi = \pi_{0}(x)\pi_{0}(y)\pi(f)\xi\]

By density of $K$ in $H$, $\pi_{0}(xy) = \pi_{0}(x)\pi_{0}(y)$, so $\pi_{0}$ is a homomorphism. For any $x \in G$, $\text{Id}= \pi_{0}(1) = \pi_{0}(x)\pi_{0}(x^{-1})$, so $\pi_{0}(x)$ is isometric and invertible, and hence unitary.

Finally, for any $f \in L^{1}(G; \complex)$ and $\xi, \eta \in H$, by the Dominated Convergence Theorem,

\begin{align*}\dpn{\pi(f)\pi(\phi_\alpha)\xi, \eta}{H}&= \dpn{\pi(f * \phi_\alpha)\xi, \eta}{H}= \angles{\pi\paren{\int_G f(y)L_y\phi_\alpha dy}\xi, \eta}_{H}\\&= \int_{G} f(y) \dpn{\pi(L_y\phi_\alpha)\xi, \eta}{H}dy \\ \dpn{\pi(f)\xi, \eta}{H}&= \int_{G} f(y)\dpn{\pi_0(y)\xi, \eta}{H}dy\end{align*}

$\square$

  1. $|V|$ is the value of the Haar measure at $V$.keyboard_return

Post a Comment

Name:Email:
Please enter the tag of the current page (1GX) to post the comment.
Tag: