Noncommutative convexity arises from linear matrix inequalities (Q860776): Difference between revisions

The authors study polynomials in \(g\) indeterminate noncommutative variables \(x=(x_1,\dots,x_g)\) which in most applications are represented by matrices. The analysis is extended to noncommutative ``rational expressions'' with real coefficients which are formally symmetric and ``analytic near zero''. A central concept is a ``linear (affine) pencil'' \(M(x)=M_0+L_M(x)\) in \(x\), where \(M_0\) is an \(m\times d\) matrix and \(L_M (x) = M_1 x_1 + \dots + M_g x_g\), where \(M=(M_1,\dots,M_g)\) is a \(g\)-tuple of \(m\times d\) matrices. The notation \(M_i x_i\) is short hand for the algebraic tensor product \(M_i\otimes x_i\). The paper gives a complete classification of matrix convex rational expressions in terms of symmetric linear pencils. In another direction, the paper gives a determinantal representation of a symmetric polynomial \(p\) in \(g\) noncommutative symmetric variables of the form \(\det p(x) = \det(A_0 - L_A (x))\), where \(A_0\) for some fixed \(d\) is a symmetric invertible \(d\times d\) matrix, and \(A=(A_1,\dots,A_g)\) is a \(g\)-tuple of symmetric \(d\times d\) matrices. The expression holds for each \(g\)-tuple \(x\) of symmetric \(n\times n\) matrices.