Suppose that $A,B$ and $C$ are $C^*$-algebras. If $\pi: A \otimes B \to C$ is a $*$-homomorphism, then there exists a unique $*$-homomorphism from $A\otimes_{\text{max}}$B to $C$ which extends $\pi$. In particular, any pair of $*$-homomorphisms with commuting ranges $\pi_A: A \to C$ and $\pi_B:B \to C$ induces a unique $*$-homomorphism $$\pi_A \times \pi_B:A\otimes_{\text{max}}B \to C$$

Given $A$ and $B$, we define the maximal $C^*$-norm on $A\otimes B$ to be $$\|x\|=\sup\{\|\pi(x)\|: \pi:A \otimes B \to \mathcal{B}(H) \text{ a cyclic $*$-homomorphism}\}$$

For the second statement: Since the map from $A \times B \to C$ defined by sending $(a,b)$ to $\pi_A(a)\pi_B(b)$ is bilinear, by universality there exists a unique map between $A \otimes B$ to $C$ which send $a \otimes b$ to $\pi_A(a)\pi_B(b)$. This happens to be a $*$-homomorphism because of the commuting ranges. Hence by the first statement, this map extends to a map from $A\otimes_{\text{max}}B$ to $C$.

For the first Statement: Let $\pi_C$ be a faithful representation of $C$ inside $\mathcal{B}(H)$. Then for $x \in A \otimes_{\text{max}} B$, define $$\tilde{\pi}(x)=\lim_n\pi_c \circ \pi(x_n)$$ where $x_n \in A\otimes B$ is a sequence converging to $x$ in $\|.\|_{\text{max}}$. This definition makes sense because $\|\pi_c \circ \pi(x_n)-\pi_c \circ \pi(x_m)\| \le ||x_n-x_m||_{\text{max}}$ and $\{\pi_c \circ \pi(x_n)\}$ being a cauchy sequence in $\mathcal{B}(H)$ converges.

I am unable to show precisely why the map has to end up in $C$. I think the idea is to think of $C$ as already being faithfully represented in some Hilbert space. I am not able to show it mathematically though.

Thanks for the help!!


Let $\psi : C \hookrightarrow B(H)$ be a faithful non-degenerate representation of $C $ (e.g. the universal representation coming from the GNS construction). Then $$ \lVert \pi(x) \rVert_C = \lVert (\psi \circ \pi)(x)\rVert_{B(H)} \leq \lVert x \rVert_{\mathrm{max}}. $$

The first equality holds true, since $\pi$ is injective and hence isometric. The second inequality holds by definition of the maximal norm. Since each non-degenerate representation decomposes as a direct sum of cyclic representations, it is easy to check that $$ \lVert x \rVert_{\mathrm{max}} = \sup \{ \lVert \pi(x) \rVert : \pi \text{ is a non-degenerate $*$-rep. of } A \otimes B \}. $$

This shows that $\pi$ extends to the completion $A \otimes_{\mathrm{max}} B$.

  • $\begingroup$ S. What about my arguement?? $\endgroup$ – tattwamasi amrutam Jul 4 '17 at 2:08
  • $\begingroup$ You show how to extend $\pi$ to the closure. You should replace $\pi_C \circ \pi (x_n)$ by $\pi(x_n)$ everywhere, because (as you say) you want to map into $C$. Then use everywhere that $\pi_C$ is isometric. $\endgroup$ – user42761 Jul 4 '17 at 7:06

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.