General non-realizability certificates for spheres with linear programming

TL;DR

This paper introduces a linear programming-based method to generate certificates of non-realizability for abstract polytopal spheres, simplifying previous algebraic approaches and producing competitive results.

Contribution

It presents a straightforward linear programming technique to find positive polynomials in the ideal, offering a new, simpler way to prove non-realizability of polytopal spheres.

Findings

Method is competitive with existing techniques

Able to derive new non-realizable spheres

Simplifies the process of proving non-realizability

Abstract

In this paper we present a simple technique to derive certificates of non-realizability for an abstract polytopal sphere. Our approach uses a variant of the classical algebraic certificates introduced by Bokowski and Sturmfels in [Computational Synthetic Geometry, 1989], the final polynomials. More specifically we reduce the problem of finding a realization to that of finding a positive point in a variety and try to find a polynomial with positive coefficients in the generating ideal (a positive polynomial), showing that such point does not exist. Many, if not most, of the techniques for proving non-realizability developed in the last three decades can be seen as following this framework, using more or less elaborate ways of constructing such positive polynomials. Our proposal is more straightforward as we simply use linear programming to exhaustively search for such positive polynomials in the ideal restricted to some linear subspace. Somewhat surprisingly, this elementary strategy yields results that are competitive with more elaborate alternatives, and allows us to derive new examples of non-realizable abstract polytopal spheres.