5

It is known that every unital separable C*-algebra is a quotient of the full group C*-algebra $C^*(F_I)$, where $F_I$ is the free group generated by some index set $I$.

Can we drop the separability assumption here?

Any references would be appreciated.

Jorg
  • 691
  • Where did you found proof of this fact? – Norbert Apr 19 '13 at 10:04
  • 1
    It is stated in Blackadar's book Operator Algebras: Theory of C${}^\ast$-algebras and Von Neumann Algebras V.2.1.12 and very often used in certain papers without a warning. I would like to have a proof in full generality. It seems that $C^*(F_I)$ plays the same role as $\ell_1(I)$ spaces in the Banach space theory. – Jorg Apr 19 '13 at 10:06
  • Stated and not proved? – Norbert Apr 19 '13 at 10:07
  • Just stated. If I had a proof, I could probably figure out the non-separable case as well. – Jorg Apr 19 '13 at 10:08
  • 1
    so may be you can write here this proof and someone will suggest a way of generalisation – Norbert Apr 19 '13 at 10:19
  • 1
    Isn't it enough to observe that the closed convex hull of the unitary elements in $A$ is the unit ball of $A$? Take the free group $F(U)$ generated by the unitaries $U$ in $A$ and let $\pi \colon F(U) \to U$ be the quotient homomorphism. This quotient map should extend to a *-homomorphism $C^\ast(F(U)) \to A$ which is onto because the image of the unit ball in $C^\ast(F(U))$ is dense in the unit ball of $A$. – Martin Apr 19 '13 at 11:00
  • @Norbert: what proof are you talking about? usao said the fact was only stated without reference. It is only an aside after a corollary to a theorem attributed to this paper of Man-Duen Choi. – Martin Apr 19 '13 at 11:08
  • @Martin I missed "if" in usao's comment and thought that he have a proof – Norbert Apr 19 '13 at 21:23
  • @Norbert I found it interesting to observe that all of a sudden you found this question interesting after I contributed, but none of my contributions useful or worthy of crtiticism. – Martin Apr 20 '13 at 22:52
  • Are you angry with me that I have not upvoted your answer? I just forgot about that. Here it is. – Norbert Apr 21 '13 at 06:13

1 Answers1

6

I think the idea in my comment works in detail as follows. Probably one can simplify the argument by applying the appropriate universal properties of the group $C^\ast$-algebra on the free group $F(U)$ on the unitaries $U$ of $A$, but I'm not very familiar with them, so here's a pedestrian's description:

By the Gelfand-Naimark theorem we can assume that $A$ is isometrically contained as a unital $C^\ast$-subalgebra of $B(H)$ for some Hilbert space $H$. The unitaries $U$ of $A$ act as unitaries on $H$, so we have a unitary representation of $U$ on $H$. This unitary representation extends to a *-representation of the complex group algebra $\mathbb{C}[U]$. Notice that the image of $\mathbb{C}[U]$ under this representation is contained in $A$. By the definition of the group $C^\ast$-algebra $C^\ast(U)$ this representation extends uniquely to a *-homomorphism $\rho \colon C^\ast(U) \to B(H)$.

I claim that $\rho$ is a homomorphism of $C^\ast(U)$ onto $A$. Since $\rho$ is continuous and $\rho(\mathbb{C}[U]) \subseteq A$ and $\mathbb C[U]$ is dense in $C^\ast(U)$, we have $\rho(C^\ast(U)) \subseteq A$. On the other hand, the elements of $U$ are unitary in $C^\ast(U)$, so they have norm one. Since $\rho$ is linear and norm-decreasing, the image of the unit ball of $C^\ast(U)$ contains the convex hull of the unitaries of $A$. But this convex hull is dense in the unit ball of $A$ by the Russo-Dye-Gardner theorem and by the open mapping theorem it follows that $\rho$ is onto $A$, as claimed.

Next, recall that the group $C^\ast$-algebra is functorial with respect to quotient homomorphisms, and $U$ is a quotient of the free group $F(U)$ generated by the set underlying $U$, so we have a surjective *-homomorphism $C^\ast(F(U)) \to C^\ast(U)$ and composing this surjection with $\rho$ we get the desired surjection $C^\ast(F(U)) \to A$ from the group $C^\ast$-algebra of some free group.

If $A$ is separable, it suffices to take a countable dense subset of the unitaries on $A$, and the above argument shows that every separable $C^\ast$-algebra is a quotient of $C^\ast(F_\infty)$ where $F_\infty$ is the free group on countably many generators, so we see that $C^\ast(F_\infty)$ is universal for separable $C^\ast$-algebras in the same sense that $\ell_1$ is universal for separable Banach spaces.

Martin
  • 8,491