Splitting of the positive set of a \(C^*\)-algebra (Q1192453): Difference between revisions

Let \(A\) be a \(C^*\)-algebra, \(B\subset A\) a \(C^*\)-subalgebra and \(H\subset A\) a supplement of \(B\) in \(A\), in the Banach space sense: \(A= B\oplus H\). Suppose \(H\) is closed under the \(*\)-operation in \(A\). Obviously \[ A^s= B^s\oplus H^s,\tag{1} \] where \(s\) indicates the selfadjoint elements in the corresponding set. If one takes the exponentials of the sets involved one gets \(A^\oplus\), \(B^\oplus\), \(\{e^h: h\in H^s\}= E\), where the \(^\oplus\) signs indicate the positive invertible elements of the corresponding set. Consider the map (suggested by (1)), \[ \Phi: B^\oplus\times E\to A^\oplus,\quad \Phi(b, e)= (be)^+,\tag{2} \] where \((be)^+\) is the ``positive part'' of \(be\), i.e. \((be)^+=\) positive square root of \((be)(be)^*\). In this paper, we prove that \[ \Phi \text{ is a diffeomorphism}\tag{3} \] in several situations, notably if \(A\) is finite-dimensional and \(H= B^\perp\) for a suitable inner product in \(A\) and in some infinite-dimensional cases.