Upper multiplicity and bounded trace ideals in \(C^*\)-algebras (Q1369777): Difference between revisions

Let \(A\) be a \(C^*\)-algebra, let \(P(A)\) denote the set of pure states of \(A\), and let \(\widehat A\) denote the spectrum of \(A\). Let \({\mathcal N}\) be the weak\(^*\)-neighbourhood base at zero in the dual space \(A^*\) of \(A\) consisting of all open sets \(N\) of the form \[ N= \{y\in A^*:| y(a_j)|< \varepsilon,\;j=1,2,\dots, n\}. \] For each element \(x\) of \(P(A)\) and each element \(N\) in \({\mathcal N}\), let \[ V(x, N)= \theta((x+ N)\cap P(A)), \] where \(\theta\) is the continuous, open mapping from \(P(A)\) onto \(\widehat A\) defined by \(\theta(y)=\widehat\pi_y\), the equivalence class of irreducible representations of \(A\) unitarily equivalent to the GNS representation \(\pi_y\) corresponding to \(y\). Then \(V(x,N)\) is an open neighbourhood of \(\widehat\pi_x\). For \(\widehat\sigma\) in \(\widehat A\), let \[ \text{Vec}(\widehat\sigma, x,N)= \{\xi\in H_\sigma:\| \xi\|= 1,\;\langle\sigma(\cdot) \xi,\xi\rangle\in x+ N\}, \] and let \(d(\widehat\sigma, x,N)\) be the supremum of the cardinalities of finite orthonormal subsets of \(\text{Vec}(\widehat\sigma, x,N)\). Then, the upper and lower multiplicities of \(\widehat\pi_x\) are defined by \[ M_U(\widehat\pi_x)= \inf_{N\in{\mathcal N}} \Biggl(\limsup_{\widehat\sigma\to \widehat\pi_x} d(\widehat\sigma, x, N)\Biggr), \] and, if \(\widehat\pi_x\) is not open in \(\widehat A\), \[ M_L(\widehat\pi_x)= \inf_{N\in{\mathcal N}} \Biggl(\liminf_{\widehat\sigma\to \widehat\pi_x,\widehat\sigma\neq \widehat\pi_x} d(\widehat\sigma, x, N)\Biggr). \] This paper is concerned with the properties of generalizations of these multiplicity numbers. For each element \(x\) in \(P(A)\) and each net \(\Omega\) in \(\widehat A\), equal to \((\widehat\pi_j)_{j\in\Gamma}\) in \(\widehat A\), let \[ M_U(\widehat\pi_x, \Omega)= \inf_{N\in{\mathcal N}} \Biggl(\limsup_{j\in \Gamma} d(\widehat\pi_j, x, N)\Biggr). \] The generalized lower multiplicity \(M_L(\widehat\pi_x, \Omega)\) is similarly defined. Observe that \(M_U(\widehat\pi_x, \Omega)\) is non-zero if and only if \(\widehat\pi_x\) is a cluster point of \(\Omega\) and \(M_L(\widehat\pi_x, \Omega)\) is non-ero if and only if \(\Omega\) converges to \(\widehat\pi_x\). The authors show that, for a positive integer \(n\), \(M_U(\widehat\pi_x, \Omega)\) is majorized by \(n\) if and only if there exists a positive element \(a\) in \(A\) such that \(\pi_x(a)\) is non-zero projection and eventually the trace of \(\pi_j(a)\) is majorized by \(n\). Using this highly non-trivial result, it can be shown that the conditions that a \(C^*\)-algebra \(A\) is uniformly liminal, has bounded trace, and that, for all elements \(x\) in \(P(A)\), \(M_U(\widehat\pi_x)\) is finite, are all equivalent. Consequently, every \(C^*\)-algebra \(A\) contains a largest uniformly liminal ideal. It is shown that the upper multiplicity function is multiplicative on tensor products and that, for any positive element \(a\) in \(A\), and any net \(\Omega\), equal to \((\widehat\pi_n)_{j\in \Gamma}\), \[ \limsup_{j\in \Gamma}\text{ Tr}(\pi_j(a))\leq \sum_{\widehat\pi\in F} M_U(\widehat\pi, \Omega)\text{Tr}(\pi(a)), \] where \(F\) is the set of cluster points of \(\Omega\). In the final section, the authors go on to give a second characterization of the largest liminal ideal using nets \(\Omega\) which are properly convergent, and have the property that, for all elements \(\widehat\pi\) in \(\widehat A\), the upper and lower multiplicities \(M_U(\widehat\pi_x, \Omega)\) and \(M_L(\widehat\pi_x, \Omega)\) are equal.