arXiv:2610.37850v1 Announce Type: new Abstract: We construct a natural spectrahedral relaxation of a rigidly convex set (rcs) that depends only on the cubic part and on the degree of the defining real zero polynomial. For quadratic polynomials, we prove exactness and for a product of linear polynomials, we show that a higher order version of our relaxation is exact. We also establish a way of tightening the relaxation by moving the origin inside the rcs, and that the rcs is recovered by sweeping the origin over its interior. To obtain all these results we use a linear form that is closely connected to the zeros of a given real zero polynomial. We also prove that any rcs can be written as an affine linear slice of the cone of non-negative polynomials. Using this, we can define a hierarchy of tightenings that are spectrahedral shadows. We can show that this tightening is exact if the given polynomial admits a determinantal representation. This tightening is similar to a tightening of a hyperbolicity cone due to Kummer, Plaumann, and Vinzant.
Continue to arXiv Mathematics →You are being forwarded in 6s — stay on this page