2
$\begingroup$

Does there exist a $C^*$-algebra $A$ such that the center of $A$ is $0$ and $A$ also has a tracial state?

I know the fact that the center of $\mathcal{K}(H)$ is $0$, but $\mathcal{K}(H)$ has no tracial states.

$\endgroup$
2
  • $\begingroup$ But the center of $A$ is not 0,$(0,Id)$ is the center element. $\endgroup$
    – math112358
    Jan 12, 2019 at 12:38
  • $\begingroup$ Oops, you're right, I'll delete that erroneous claim $\endgroup$
    – Yemon Choi
    Jan 12, 2019 at 13:07

3 Answers 3

7
$\begingroup$

There are lots of AF simple nonunital C$^*$-algebras with (finite) traces (these have trivial centre). For example, begin with the usual $2^{\infty}$ UHF algebra, call it $B$, and take an infinite strictly increasing sequence of projections, $(p_n)$, and form $\cup_n p_n B p_n$; let $A$ be its closure. Then $A$ is simple (easy to check); it is nonunital (very easy) [and being simple, is thus centreless]; it has unique trace (given by the restriction of the unique trace on $B$) (easy). $A$ is of course a hereditary subalgebra of $B$, and there are hordes of similar examples.

If we set $\alpha$ to be the limit of the traces of the $p_n$, then there is a significant difference (certainly wrt classification) between the algebras we obtain when $\alpha$ is rational, and when $\alpha $ is irrational.

$\endgroup$
2
  • $\begingroup$ Since $B$ is simple, $\tau (aa*) >0$ for all $a \in B$ (since the set of $a \in B$ st $\tau (aa*) = 0$ is an ideal of $B$), hence for $a \in A$. $\endgroup$ Jan 12, 2019 at 19:16
  • 1
    $\begingroup$ @mathrookie Of course, I meant "for all nonzero $a \in B$"; I wish we could edit comments. $\endgroup$ Jan 12, 2019 at 19:27
4
$\begingroup$

I think you can obtain an example by modifying the classical Toeplitz algebra. $\newcommand{\H}{\mathbf{H}}$ $\newcommand{\bT}{\mathbf{T}}$ $\newcommand{\cT}{\mathcal T}$ $\newcommand{\KH}{{\mathcal K}(\H)}$ $\newcommand{\BH}{{\mathcal B}(\H)}$

Recall that ${\mathcal T} \subseteq \BH$ where $\H=\ell^2({\mathbf N}_0))$ and $\cT/\KH\cong C(\bT)$. Let $q: \cT \to C(\bT)$ be the corresponding quotient homomorphism, let $J=\{ f\in C(\bT) \colon f(1)=0\}$ and let $\cT_0= q^{-1}(J)$. More concretely, $\cT_0$ is the set of closed algebra generated by all Toeplitz operators $T_f$ for which the symbol function $f$ belongs to $J$.

Since $\cT_0$ quotients onto the commutative ${\rm C}^*$-algebra $J$ it has loads of tracial states (just pull back the states on $J$).

On the other hand, suppose $a\in Z(\cT)$. Then $[a,k]=0$ for all $k\in \KH$. But it is standard and not too difficult to show that if $b\in\BH$ satisfies $[k,b]=0$ for all $k\in\KH$, then $b$ is a scalar multiple of the identity. Since $q(I_\H)\notin J$ the only possibility is that $a=0$.

$\endgroup$
6
  • $\begingroup$ Nice answer. But the statement that $\mathcal{T}_0$ consists of Toeplitz operators whose symbol vanishes at $1$; you need to throw in the compact perturbations or allow products. $\endgroup$ Jan 12, 2019 at 14:46
  • $\begingroup$ Irrational rotation algebras (quantum tori) are another example. $\endgroup$
    – Nik Weaver
    Jan 12, 2019 at 14:56
  • $\begingroup$ Also, the reduced group C*-algebra of any discrete group always has a tracial state. $\endgroup$
    – Nik Weaver
    Jan 12, 2019 at 15:03
  • 1
    $\begingroup$ Or just $A = M_n$ ... $\endgroup$
    – Nik Weaver
    Jan 12, 2019 at 15:03
  • 2
    $\begingroup$ Thanks Mateusz. @NikWeaver: the OP wants $Z(A)=0$ not $\dim Z(A)=1$ $\endgroup$
    – Yemon Choi
    Jan 12, 2019 at 15:51
2
$\begingroup$

Jacelon's $\mathcal W$ (https://arxiv.org/abs/1006.5397) should also do.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.