A general look at local principles with special emphasis on the norm computation aspect (Q1109296): Difference between revisions

In order to extend Gelfand theory, \textit{I. Gohberg} and \textit{N. Ya. Krupnik} [Introduction to the theory of 1-dimensional integral operators (in Russian) (1973; Zbl 0271.47017)] introduced localising classes as the analogue, for a noncommutative Banach algebra \({\mathfrak A}\) with unit element, of the maximal ideals in Gelfand theory. Denote the set of invertible elements of \({\mathfrak A}\) by Inv \({\mathfrak A}\) and the maximal ideal space by M(\({\mathfrak A})\). A subset M of \({\mathfrak A}\) is called a localising class if \(0\not\in M\) and for each pair \(f_ 1,f_ 2\in M\) there exists an \(f\in M\) such that \(ff_ j=f_ jf=f\), \(i=1,2\). A family \((M_ t)_{t\in T}\) is called covering if, for any family \((f_ t)_{t\in T}\), \(f_ t\in M_ t\), there exists a finite set \(f_{t_ 1},...,f_{t_ m}\) such that \(f_{t_ 1}+...+f_{t_ m}\in Inv {\mathfrak A}\). Given a localising class M, elements \(a,b\in {\mathfrak A}\) are called M-equivalent if \(\inf_{f\in M}\| (a-b)f\| =\inf_{f\in M}\| f(a-b)\| =0\). Define \({\mathfrak M}=\cup_{t\in T}M_ t\), Com \({\mathfrak M}=\{a\in {\mathfrak A}:af=fa\) for all \(f\in {\mathfrak M}\}\), \(Z_ t\) to be the set of elements of Com \({\mathfrak M}\) which are \(M_ t\)- equivalent to 0 and let \(\tilde a\) t denote the coset \(a+Z_ t\). Gohberg and Krupnik proved that if \(M_ t\) is bounded in \({\mathfrak A}\) then \(a\in Inv {\mathfrak A}\) if and only if \(\tilde a\) \(t\in Inv(Com {\mathfrak M}/Z_ t)\) for all \(t\in T\). The authors topologise calling a family \((M_ t)_{t\in T}\) of localising classes overlapping if each \(M_ t\) is bounded, \({\mathfrak M}\) is commutative and T has a topology such that if \(f\in M_{t_ 0}\) then \(f\in M_ t\) for all t in a neighbourhood of \(t_ 0:\) then \(t\mapsto \tilde a\) t is upper semicontinuous and, if \({\mathfrak A}\) is a C *-algebra and each \(M_ t\) is self-adjoint, then \(\| a\| =\max_{t\in T}\| \tilde a\) \(t\|.\) Consider a closed subalgebra \({\mathfrak L}\) of the centre of \({\mathfrak A}\), containing the unit element and denote the Gelfand transform of \(b\in {\mathfrak L}\) by \(\hat b.\) Let \(N\in M({\mathfrak L})\) and denote by \(J_ N\) the smallest closed two-sided ideal of \({\mathfrak A}\) containing \(\{\) \(b\in {\mathfrak L}:\hat b(N)=0\}\) and by \(\tilde a\) N the coset \(a+J_ N\). Then \((M_ N)_{N\in M({\mathfrak L})}\), where \(M_ n=\{b\in {\mathfrak L}:\hat b\equiv 1\) in a neighbourhood of N and \(\hat b\equiv 0\) outside a neighbourhood of \(N\}\), form an overlapping covering family of localising classes. Noting that Com \({\mathfrak M}={\mathfrak A}\) and letting \(J_ N\) play the role of \(Z_ t\), the results of \textit{G. R. Allan} [Proc. Lond. Math. Soc., III. Ser. 18, 193-216 (1968; Zbl 0194.445)] and \textit{R. G. Douglas} [Banach algebra techniques in operator theory (1972; Zbl 0247.47001)] are a special case of the Gohberg-Krupnik theory. Denote the algebras of bounded and compact linear operators on L p(X), where X is compact Hausdorff, with non-negative Borel measure and \(1\leq p\leq \infty\), by \({\mathfrak L}\) and \({\mathfrak K}\) respectively. An element of \({\mathfrak L}\) is said to be of local type if Aa-aA\(\in K\) for all \(a\in C(X)\), where a also represents the operation of multiplication by a continuous function on X. When X is finite dimensional and p finite this definition is equivalent to that of \textit{I. B. Simonenko} [Izv. Akad. Nauk SSSR, Ser. Math. 28, 277-306 (1964; Zbl 0136.069)]. Denote the Banach algebra of local type operators by \(\Lambda\). The essential norm \(| A|\) of \(A\in {\mathfrak L}\) is defined to be \(\inf \{\| A+K\|:K\in {\mathfrak K}\}\) and let \(| A|_ x=\inf \{| aA|:a\in C(X)\), \(0\leq a\leq 1\), \(a(x)=1\}\). The authors' results on the norms of the generalised Gelfand transform apply also to these local essential norms, generalising Simonenko. However, they do not prove existence of a generalised Gelfand transform and in this way their tone is misleading. \({\mathfrak A}\) is called KMS with respect to a C *-subalgebra \({\mathfrak L}\), with unit element, if \(\| (b+c)a\| \leq \max (\| ba\|,\| ca\|)\) for all \(a\in {\mathfrak A}\), and \(b,c\in {\mathfrak L}\) with disjount supports. It is shown that \({\mathfrak A}\) is KMS with respect to \({\mathfrak L}\) if and only if \(\| a\| =\max \{\| a\) \(N\|:N\in M({\mathfrak L})\}\) for all \(a\in {\mathfrak A}\). Unital C *-algebras are KMS with respect to unital C *-subalgebras of their centres. The authors construct examples of KMS-algebras of local type operators. If \(A_ x\), \(x\in X\), is of local type and A is of local type such that \(| A-A_ x|_ x=0\) for all \(x\in X\), then A is called the envelope of the family \((A_ x).\) The authors show that a family of local type operators has an envelope if and only if the essential norms satisfy an upper semi-continuity condition, and in that case it is unique modulo compacts, and \(| A| \leq \max \{| A_ x|:x\in X\}\) [c.f. \textit{I. B. Simonenko} (loc. cit.)]. Finally the authors apply their results to express essential norms as maxima of essential local norms to obtain Simonenko's results on singular integral operators. They use Allan-Douglas theory to obtain similar results for Toeplitz operators, not based on L p-spaces.