Sharp bounds on the ranks of negativity of certain sums (Q949456)

The article is a serious investigation of the rank of negativity of some forms. First, the author describes the original motivation of the investigation starting with one result of \textit{C.~Berg} and \textit{Z.~Sasvári} [Monatsh. Math. 107, No.~1, 9--34 (1989; Zbl 0671.43008)]. Here, the reader could familiarize himself with the notion of rank of negativity. Then, the author considers the following generalization to inner product spaces instead of \(*\)-algebras: If \(M\) is a complex vector space, \(k\) is a nonnegative integer, \(\langle \cdot,\cdot\rangle \) is a Hermitian sesquilinear form on \(M\) of rank of negativity \(k\) (i.e., \(k\) is the maximal dimension of any linear subspace \(E\) of \(M\) satisfying \(\langle x,x\rangle <0\) for each nonzero \(x\) in \(E\)), if \(n\) is a positive integer and \(a_1,\dots,a_n\) are endomorphisms of \(M\), how great can be the rank of negativity of the Hermitian sesquilinear form \[ (x,y)\mapsto\sum_{\nu=1}^n\langle a_\nu x,a_\nu y\rangle \tag{1} \] on \(M\)? The article contains the answer: The sesquilinear form (1) has rank of negativity at most \(nk\). The bound \(nk\) cannot be improved in general (even if \(M\) is assumed to be finite-dimensional). Less trivial is the fact that the bound \(nk\) of the rank of negativity cannot be improved by making the following assumption: {\parindent6mm \begin{itemize}\item[(a)] the space \(M\) is the \(*\)-algebra \(A:=\mathbf{C}[[w_1,w_2]]\) of polynomials in two self-adjoint non-commuting indeterminates; there is a (necessarily Hermitian) linear form \(\varphi\) on \(A\) such that \(\langle x,y\rangle =\varphi(y^*x)\) (\(x,y\in A\)); and each \(a_\nu\) is just left multiplication by some element of \(A\) (which we may denote by `\(a_\nu\)' at no great risk of confusion). \end{itemize}} Then the author considers the case when each \(a_\nu\) has a formal adjoint \(a_\nu^*\), being an endomorphism of \(M\) such that \[ \langle a_\nu x,y\rangle =\langle x,a_\nu^* y\rangle \quad (x,y\in M) \] and the mappings \(a_1,\dots,a_n\), \(a_1^*,\dots,a_n^*\) commute pairwise. In this case, the bound \(nk\) can be replaced by \(k\) (regardless of how large \(n\) may be). This result cannot be improved in general, since it may happen that each \(a_\nu\) is a scalar multiple of the identical mapping of \(M\) into itself (not all \(a_\nu\) equal to \(0\)), in which case the form (1) is a positive multiple of \(\langle \cdot,\cdot\rangle \) itself. The article is divided into three parts. The first part is an introduction that contains some motivations and describes the contents of the article. The second and third parts deal with the non-commutative and the commutative case, respectively. The author investigates the problem itself, and no connections to other areas are mentioned. Thus, this solid article is addressed to a smaller group of specialists interested in this area.