2
$\begingroup$

I'm struggeling with the proof of a theorem in [BO08]. The first part before the line is what I think I understood. The part after that I don't understand at all.

Let $\Gamma$ be a discrete group and $\mathbb{C}[\Gamma]$ the group ring of $\Gamma$. Let $C_\lambda^\ast(\Gamma)$ be the reduced group $C^\ast$-algebra, i.e. the completion of $\mathbb{C}[\Gamma]$ wrt. the norm $\|x\|=\|\lambda(x)\|$ where λ is the left regular representation of $\Gamma$ on $\mathbb{B}(\ell^2(\Gamma)$. Let $$L(\Gamma) = C_\lambda^\ast(\Gamma)'' \subset \mathcal{B}(\ell^2(\Gamma))$$ be the so called group von Neumann algebra. The full group $C^\ast$-algebra is defined as $C^*(\Gamma) = \overline{\mathbb{C}(\Gamma)}^{\|\cdot\|_u}$ with $\| x\|_u = \sup\{ \|\pi(x)\| \; | \; \pi \text{ is a unitary representation of } \Gamma\}$.

For two $C^\ast$-algebras $A$ and $B$ we say that a map $\varphi_A\to B$ is u.c.p if $\varphi$ is unital and $\varphi_n:M_n(A)→M_n(B)$ defined by $\varphi_n([a_{i,j}])=[\varphi(a_{i,j})]$ maps positive matrices to positive matrices for all $n\in \mathbb{N}$.

$\varphi$ is called normal if it is continuous wrt. ultra-weak topology.

A function $\varphi:\Gamma\to \mathbb{C}$ is called positive definite if for every finite subset $F:=\{s_1, \dots, s_n\}\subseteq \Gamma$ the matrix $$[\varphi(s_i^{-1}s_j)]_{i,j}\in M_n(\mathbb{C}) $$ is positive semidefinite.

For a function $\varphi:\Gamma\to\mathbb{C}$ set \begin{equation*} \begin{split} w_\varphi: \mathbb{C}[\Gamma] &\to \mathbb{C}\\ \sum_{t\in\Gamma}\alpha_t t &\mapsto \sum_{t\in\Gamma} \alpha_t \varphi(t) \end{split} \end{equation*} and \begin{equation*} \begin{split} m_\varphi: \mathbb{C}[\Gamma] &\to \mathbb{C}[\Gamma]\\ \sum_{t\in\Gamma}\alpha_t t &\mapsto \sum_{t\in\Gamma} \alpha_t \varphi(t)t. \end{split} \end{equation*}

The map $m_\varphi$ is also called the Schur-multiplier.

Theorem(2.5.11, [BO08])Let $\varphi:\Gamma\to\mathbb{C}$ be a function with $\varphi(e) =1$. The following conditions are equivalent:

  • The functional $w_\varphi$ extends to a positive functional on $C^\ast(\Gamma)$.
  • The Schur multiplier $m_\varphi$ extends to a u.c.p. map on the two group $C^\ast$-algebras $C^\ast(\Gamma)$ and $C^\ast_\lambda (\Gamma)$ and to a normal u.c.p map on $L(\Gamma)$.

Proof:Since $w_\varphi$ extends to a functional on $C^\ast(\Gamma)$ it follows that $w_\varphi$ extends to a state on $C^\ast(\Gamma)$, which will also be denoted by $w_\varphi$. Embedding $C^\ast(\Gamma)\subseteq \mathcal{B}(H_u)$ where $H_u$ is the Hilbert space of the universal representation $(\pi_u, H_u)$ of $C^\ast(\Gamma)$, we can regard $w_\varphi$ as a state on $\mathcal{B}(H_u)$. We define a map \begin{equation*} \begin{split} \phi: C_\lambda^\ast(\Gamma)\otimes 1 \cong C_\lambda^\ast(\Gamma) &\to C_\lambda^\ast(\Gamma)\otimes C^\ast(\Gamma) \subseteq C_\lambda^\ast(\Gamma)\otimes \mathcal{B}(H_u)\\ \sum_{t\in\Gamma}\alpha_t\lambda(t)\otimes 1 &\mapsto \sum_{t\in\Gamma}\alpha_t(\lambda(t)\otimes t) \end{split} \end{equation*} which is a $\ast$-homomorphism: By Fell's absorption there is a unitary operator $U$ such that $U^\ast(\lambda \otimes \pi_u)(x)U = (\lambda\otimes 1)(x)$ for $x\in \mathbb{C}[\Gamma]$ with \begin{equation*} \begin{split} \|\phi(\sum_{t\in \Gamma} \alpha_t (\lambda(t)\otimes 1))\| &= \|\sum_{t\in \Gamma} \alpha_t (\lambda(t)\otimes t)\| = \|(\lambda\otimes \pi_u)(\sum_{t\in \Gamma} \alpha_t t)\|\\ &=\|U^\ast (\lambda\otimes1)(\sum_{t\in\Gamma} \alpha_t t)U\| = \|\sum_{t\in\Gamma} \alpha_t(\lambda(t)\otimes 1)\|. \end{split} \end{equation*}


We conclude that $\phi$ extends to a normal $\ast$-homomorphism $\phi:L(\Gamma) \to L(\Gamma)\overline{\otimes} B(H_u)$(Why?).\ Checking that $(id_{L[\Gamma)} \otimes w_\varphi)\circ \phi$ coincides with $m_\varphi$ (identifying $L(\Gamma)\overline{\otimes} 1 \cong L(\Gamma)$) concludes the group von Neumann algebra case. The reduced group $C^\ast$-algebra case follows directly.(Why?)

I would be very thankful if someone could give me a hint or some explanations. Thank you very much in advance!


[BO08] Brown and Ozawa, "C∗-Algebras and Fnite-Dimensional Approximations"

$\endgroup$
2
$\begingroup$

It says in [BO08] that "Fell's principle is spatially implemented". That is, $$ \phi:\sum_t\alpha_t\lambda_t\longmapsto \sum_t\alpha_t(\lambda_t\otimes1)=U^*\left(\sum_t\alpha_t(\lambda_t\otimes \pi_u)\right)\,U\longmapsto \sum_t\alpha_t(\lambda_t\otimes \pi_u). $$ As mentioned in the book, both maps in the composition above are spatial, so they are sot continuous; so in particular $\phi$ is normal.

The argument given in book works for the reduced C$^*$-algebra case if you just don't extend to $L(\Gamma)$.

$\endgroup$
  • $\begingroup$ Sorry to bother you again after such a long time. But could you shortly explain how it follows that this map is SOT-continous? I'm still confused by that. To my understanding today, it directly follows that it is normal, since for a bounded, upward directed set of self-adjoint operators $a_i$, the operators $Ua_iU^{\ast} $ also form a upward directed set of selfadjoint operators that is bounded by $UaU ^{\ast} $ where $a$ denotes the upper bound of the original set of operators, $vand $U$ is a unitary. $\endgroup$ – Opalgal May 5 at 22:41
  • $\begingroup$ Well, that's basically it. If $a_j\to a$ sot, then $Ua_jU^*\to UaU^*$ sot. Is that not clear to you? I can try to explain more if needed. $\endgroup$ – Martin Argerami May 5 at 22:58
  • $\begingroup$ I am stressing myself out right now and even less is clear to me than it usually. I thought what you wrote was clear to me but when I tried to check it by calculation I just didn't know where I was going with it. So an elaboration would be helpful. Sorry. $\endgroup$ – Opalgal May 5 at 23:03
  • $\begingroup$ No worries. If $a_j\to a$ sot, then $$\|Ua_jU^*x-U^*aUx\|=\|U(a_j-a)(U^*x)\|=\|(a_j-a)(U^*x)\|\to0.$$ So $Ua_jU^*\to UaU^*$. Is that what you are asking? $\endgroup$ – Martin Argerami May 5 at 23:12
  • $\begingroup$ Ok, well, that was really stupid of me. For whatever reason my panicked self was trying to check that $Ua_jU^\ast$ strongly converges to $a$. Thank you very much for your time. $\endgroup$ – Opalgal May 5 at 23:19

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.