The \(C^*\)-algebra of a function algebra (Q1416669)

The author considers the pair \((A,G)\) where \(G\) is a compact Hausdorff space and \(A\) is a function algebra on \(G\), i.e., \(A\) is a norm-closed (proper) subalgebra of \(C(G)\), the unital C*-algebra of continuous complex-functions on \(G\) separating the points of \(G\). The classical case is where \(A\) is the disc algebra and \(G\) is the unit circle \(\mathbb{T}\). Toeplitz matrices \((a_{j-k})_{j,k =1}^{\infty}\) define operators on \(\l^{2}\) which can be analysed using the `symbol' \(\sum_{-\infty}^{+\infty}{\lambda^{j}}\) for \(|\lambda| = 1\). A sequence \((a_{n})_{ -\infty}^{+\infty}\) defines a bounded operator on \(l^{2}(\mathbb{Z})\) or \(l^{2}(\mathbb{N}_{0})\) if and only if the \(a_{n}\) are Fourier coefficients of some \(\phi \in L^{\infty}(\mathbb{T})\). This leads to the consideration of Hardy spaces as the Fourier transform takes \(l^{2}(\mathbb{N}_{0})\) to \(H^{2}(\mathbb{T})\). A Toeplitz operator having symbol \(\phi \in L^{\infty}\) can be defined by \(T_{\phi}(f) = P(\phi f)\) where \(f \in H^{2}\) and \(P:L^{2} \to H^{2}\) is an orthogonal projection. The (generalised) Hardy spaces \(H^{2}(A,m)\) are Hilbert subspaces, generated by \(A\), in \(L^{2}(G,m)\); here \(m\) denotes a probability measure on \(G\), not necessarily a unique representing measure for a character but a measure which represents a linear functional \(\tau\) on \(A\), i.e., \(\tau(\phi) = \int{\phi \,dm}\). A Toeplitz representation \(V\) of \((A,G)\) in a unital C*-algebra is defined essentially by the properties that it is norm-decreasing, preserves adjoints and \(T_{\phi}T_{\psi} = T_{\phi \psi}\), \(\phi \in C(G)\), \(\psi \in L^{2}(A,m)\). The author shows that a norm-decreasing unital homomorphism \(A \to B(H)\) need not extend to a Toeplitz representation of \((A,G)\). He illustrates this on the disc algebra, i.e., the closed subalgebra of elements in \(C(\mathbb{T})\) which can be extended to analytic functions inside the disk. The mapping \(z \to {1 \over z}\) is not analytic so there is a character \(\tau\) on \(A\) for which \(\tau(z) = 0\) for all \(z\) inside the disc; this implies, by a simple contradiction, that \(\tau\) does not admit an extension to a Toeplitz representation \(C(T) \to \mathbb{C} = B(\mathbb{C})\). The author introduces a unique C*-algebra \(C^{*}(A,G)\) generated by the range of a Toeplitz representation of \((A,G)\). It is defined by a universal property, viz., for any Toeplitz representation \(V\), if \(T\) is any other Toeplitz representation of \((A,G)\) in a unital C*-algebra \(B\), there exists a unique C*-homomorphism \(\pi: C^{*}(A,G) \to B\) for which \(\pi(V_{\phi}) = T_{\phi}\). C*(\(A,G\)) is said to have the injective homomorphisms property if for each non-commutative Toeplitz representation of \((A,G)\) the induced homomorphism on \(C^*(A,G)\) is injective. Since it has been assumed that \(A \neq C(G)\) there are, for any non-Abelian \(A\), probability measures on \(G\) for which the corresponding Toeplitz representations are not multiplicative. He gives a couple of necessary and sufficient conditions, involving the commutator ideal of \(C^{*}(A,G)\), for that any non-multiplicative Toeplitz representation of \((A,G)\) the induced homomorphism be injective. A \(C^{*}(A,G)\) with injective homomorphisms property has trivial centre and this implies further that \(A\) is antisymmetric. Also, \(A\) is a maximal function algebra on \(G\), i.e., it is not contained in any other functional algebra on \(G\). If \(A\) is unimodular, maximal and antisymmetric, then \(C^{*}(A,G)\) has the injective homomorphism property; the author wonders whether this is true for more general \(A\).