Diagonalizing matrices over C(X) (Q759974)
From MaRDI portal
scientific article
Language | Label | Description | Also known as |
---|---|---|---|
English | Diagonalizing matrices over C(X) |
scientific article |
Statements
Diagonalizing matrices over C(X) (English)
0 references
1984
0 references
If A(x) is a normal \(n\times n\)-matrix, depending continuously on a parameter x in some compact Hausdorff space X, one may ask for the existence of a continuous unitary matrix function U(x), \(x\in X\), such that \(U(x)A(x)U(x)^*\) is diagonal for every x in X. If A(x) is multiplicity-free for each x, we show that A is diagonalizable when X is a 2-connected CW-complex. When multiplicity occur, the situation is much more complicated: A necessary condition for diagonalizability is (i) that the space X be sub-Stonean, i.e. every two disjoint, open \(\sigma\)- compact subsets of X have disjoint closures. An early reference to this obstruction is: \textit{L. Gillman} and \textit{M. Henriksen} [Trans. Am. Math. Soc. 82, 362-365 (1956; Zbl 0073.022)]. A survey of sub-Stonean spaces can be found in: \textit{K. Grove} and \textit{G. K. Pedersen} [J. Funct. Anal. 56, 124-143 (1984; Zbl 0539.54029)]. Other obstructions for diagonalizability arise if some closed subset \(X_ 0\) of X carries a non-trivial G-bundle, when G is a symmetric group or the circle group. The resulting necessary conditions are (ii) that the covering dimension for X be \(\leq 2\), (iii) that \(H^ 1(x_ 0,S_ m)=0\) for every \(X_ 0\subset X\) and all m; and (iv) that \(H^ 2(X_ 0,{\mathbb{Z}})=0\) for every \(X_ 0\subset X.\) We show that on a space X, satisfying the conditions, (i)-(iv), diagonalization is always possible. As an example of such a space we have \(X=\beta (Y)\setminus Y\), for some infinite, contractible graph Y with Stone-Čech compactification \(\beta\) (Y). A simpler example (with a much simpler proof) occurs if X is a Stonean space (or just a closed subset of a Rickart space). Here we can even triangularize an arbitrary (continuous) matrix function A. This paper is a reply to a question raised by \textit{R. V. Kadison} [Bull. Am. Math. Soc. 8, 84-86 (1983; Zbl 0518.46047)] who shows that if \({\mathcal M}\) is an arbitrary von Neumann algebra and \({\mathfrak A}\) is a commutative \({}^*\)-subalgebra of \({\mathcal M}\otimes {\mathbb{M}}_ n\)- the \(n\times n\)- matrices over \({\mathcal M}\)- then there is a unitary U in \({\mathcal M}\otimes {\mathbb{M}}_ n\) such that U\({\mathfrak A}U^*\) is a diagonal algebra. In this setting our theorem characterizes those commutative \(C^*\)-algebras C(X), for which Kadison's result is valid for a countably generated, commutative \({}^*\)-subalgebra \({\mathfrak A}\) of \(C(X)\otimes {\mathbb{M}}_ n\). If one insists on the diagonalizability of every commutative \({}^*\)-subalgebra, then we show that X must be a Stonean space. Thus Kadison's result can never be extended beyond \(AW^*\)-algebras.
0 references
2-connected CW-complex
0 references
sub-Stonean spaces
0 references
obstructions for diagonalizability
0 references
covering dimension
0 references
Stone-Čech compactification
0 references
0 references
0 references