Uniqueness of certain polynomials constant on a line (Q979017)

Let \(\mathcal{H}(2,d)\) denote the set of polynomials \(p(x,y)\) of degree \(d\) with nonnegative coefficients such that \(p(x,y) = 1\) whenever \(x+y=1\). Let \(N = N(p)\) denote the number of distinct monomials for a given \(p\) in \(\mathcal{H}(2,d)\). \textit{J. D'Angelo}, \textit{S. Kos} and \textit{E. Riehl} [J. Geom. Anal. 13, No. 4, 581--593 (2003; Zbl 1052.26016)] showed that \(d \leq 2N-3\). Further, this inequality is sharp. For each odd \(d\), let \(f_d\) in \(\mathcal{H}(2,d)\) be defined as \[ f_d(x,y) := \left(\frac{x+\sqrt{x^2+4y}}{2}\right)^d + \left(\frac{x-\sqrt{x^2+4y}}{2}\right)^d + (-1)^{d+1}y^d. \] It is a fact that \(d = 2N(f_d) -3\). These family of polynomials have many other interesting properties. In the context of CR geometry, the CR maps that arise from \(f_d\) are one of the only two possible classes of group invariant maps of balls. The polynomials \(f_d\) are also related to Chebyshev polynomials, arise in denesting radicals, and have connections to number theory, for example, \(f_d(x,y) = x^d + y^d \pmod{d}\) if and only if \(d\) is an odd prime. If \(p \in \mathcal{H}(2,d)\) minimizes \(N(p)\) for a fixed \(d\), then \(p\) is called a \textit{sharp} polynomial. It can be proved that for a fixed degree only finitely many sharp polynomials exist. It is natural to attempt a classification of all sharp polynomials of a given degree. A related problem consists in finding sufficient and necessary conditions for \(f_d\) to be the \textit{unique} sharp polynomial in \(\mathcal{H}(2,d)\) up to swapping of variables. Using computational methods, the authors give a complete classification of sharp polynomials for degrees up to \(d = 17\) extending previous work on the topic. Along the way, the authors provide new theoretical results about the form of sharp polynomials that are of independent interest. The authors describe two different methods to find sharp polynomials; first by computing the nullspace of certain matrices and second by translating the problem into a mixed-integer programming problem. Finally, the authors also describe degrees in which uniqueness definitely fails.