The algebraic reformulation of the Bessis–Moussa–Villani conjecture is equivalent to a family of dimension-free tracial inequalities involving positive semidefinite matrices. Sufficient conditions for these to hold in the form of algebraic identities involving polynomials in noncommuting variables have been given by Klep and Schweighofer [“Sums of Hermitian squares and the BMV conjecture,” J. Stat. Phys. 133, 739 (2008)]. Later the existence of these certificates has been settled for all but one case, which is resolved in this note.
To each semidefinite program (SDP) in primal form, we associate the matrix algebra generated by its constraint matrices. In this note, we show that this algebra is always a full matrix algebra for SDPs arising from (commutative or non-commutative) sum of squares (SOS) problems. For SDPs arising from non-commutative SOS and commutators problems, the situation is less clear. We identify an exceptional case, where the corresponding matrix algebra is not the full matrix algebra, and use it to reprove the Burgdorf–Klep non-commutative variant of Hilbert’s ternary quartics theorem: a bivariate non-commutative polynomial of degree at most is trace positive if it is a sum of four squares and commutators.
25 October 2012 in Computational Optimization and Applications