Absolute continuity of positive linear functionals (Q2341438)

Let us first describe the interesting setting from which the paper under review starts (cf.\ [\textit{S. Hassi} et al., J. Funct. Anal. 257, No. 12, 3858--3894 (2009; Zbl 1185.47002)] and preceding papers of the author). Since (at least) the 1970s, Lebesgue-type decompositions have been investigated in the surprisingly general setting of forms (short for `nonnegative quadratic forms' which are associated, as usual, to sesquilinear mappings by polarization) on a complex vector space \(E\). Two natural examples of how forms appear are: (a) a measure \(\lambda\) on a \(\sigma\)-algebra \(\Sigma\) defines a form on the vector space \(E\) spanned by the characteristic functions of the elements of \(\Sigma\) by \(\varphi\mapsto\int|\varphi|^2\text{d}\lambda\) or more generally, (a') a form is given by \(a\mapsto{\mathbf f}[a]:=f(a^*a)\) if \(f\) is a positive linear functional \(f\) on a \(^*\)-algebra; (b) a Hilbert space \(H\) with inner product \((f,g)\mapsto\langle f,g\rangle\) and a bounded nonnegative operator \(T\) on \(H\) give rise to the form \(f\mapsto\langle Tf,f\rangle\). A key notion in this context is the parallel sum \(({\mathbf t}:{\mathbf w})\) defined by \[ ({\mathbf t}:{\mathbf w})[x]:=\inf\{{\mathbf t}[x-z]+{\mathbf w}[z]; z\in E\}\quad (x\in E) \] for two forms \({\mathbf t},{\mathbf w}\) on the complex vector space \(E\). And a key fact is that this parallel sum is a form on \(E\). (Other properties are \(({\mathbf t}:{\mathbf w})=({\mathbf w}:{\mathbf t})\) and \(({\mathbf t}:{\mathbf w})\leq{\mathbf t}\) where \(\leq\) is naturally understood pointwise which must be distinguished from `\({\mathbf t}\) is dominated by \({\mathbf w}\)' meaning \({\mathbf t}\leq\alpha{\mathbf w}\) for an appropriate \(\alpha\geq0\).) By definition, two forms \({\mathbf t},{\mathbf w}\) are singular if \(({\mathbf t}:{\mathbf w})=0\). A form \({\mathbf t}\) is said to be absolutely continuous with respect to \({\mathbf w}\) if there is an increasing sequence \(({\mathbf t}_n)\) where each \({\mathbf t}_n\) is dominated by \({\mathbf w}\) such that \({\mathbf t}=\sup {\mathbf t}_n\). Now the Lebesgue decomposition of \({\mathbf t}\) with respect to \({\mathbf w}\) reads \[ {\mathbf t}={\mathbf t}_{reg}+{\mathbf t}_{sing} \text{ where } {\mathbf t}_{reg}=\sup({\mathbf t}:(n{\mathbf w})) \text{ and } {\mathbf t}_{sing}={\mathbf t}-{\mathbf t}_{reg} \] and where \({\mathbf t}_{reg}\) is absolutely continuous w.r.t.\ \({\mathbf w}\) while \({\mathbf t}_{sing}\) and \({\mathbf wt}\) are singular. Note that in the case (a) above this corresponds precisely to the classical Lebesgue decomposition of a measure \(\lambda\) (giving rise to \({\mathbf t}\)) with respect to a measure \(\mu\) (giving rise to \({\mathbf w}\)), but note also that in general the Lebesgue decomposition of \({\mathbf t}\) need not be unique. The paper under review deals, amongst other things, with the characterization of absolute continuity of forms that are induced by representable positive linear functionals \(f\) (in which case domination, absolute continuity and singularity are defined via the associated forms) on \(^*\)-algebras \(A\). Representable means that there exists a cyclic \(^*\)-representation \(\pi:A\rightarrow B(H)\) on a Hilbert space \(H\) and \(f\) is associated to the cyclic vector. Positive functionals on unital Banach \(^*\)-algebras are representable. A~main tool consists in the fact that if two forms on a \(^*\)-algebra \(A\) come from representable positive functionals \(f,g\), then so does their parallel sum. This leads to the characterization that \(f\) is absolutely continuous w.r.t.\ \(g\) if and only if there is an increasing sequence \((f_n)\) of representable positive functionals on \(A\) dominated by \(g\) such that \(f=\sup f_n\). Another characterization of absolute continuity concerns forms on \(C^*\)-algebras for which absolute continuity is characterized in terms of unital commutative \(C^*\)-algebras (namely the continuous functions on the compact state space). Still another characterization refers to the convolution Banach \(^*\)-algebra \(L^1(G,\beta)\) (\(G\) a compact group with Haar measure \(\beta\)): given two positive linear (automatically representable) functionals \(f,g\) on \(L^1(G,\beta)\), \(f\) is absolutely continuous w.r.t.\ \(g\) if and only if the kernel of the semi-norm induced by \(f\) contains the kernel of the semi-norm induced by \(g\). Furthermore, while in general the Lebesgue decomposition of representable positive functionals need not be unique, it is unique in the cases of \(L^1(G,\beta)\) and of commutative \(^*\)-algebras which answers a question of \textit{S. P. Gudder} [Pac. J. Math. 80, 141--149 (1979; Zbl 0406.46055)]. (Note that it can happen that the functional decomposition is unique while the decomposition of the corresponding forms is not.) As an application it is shown that a positive \(g\) on \(L^1(G,\beta)\) is faithful if and only if every positive \(f\) on \(L^1(G,\beta)\) is absolutely continuous w.r.t.\ \(g\).