Sum of squares conjecture: the monomial case in \(\mathbb{C}^3\) (Q2231156)

A long-standing and difficult problem in several complex variables is to classify CR mappings of spheres in different dimensions (or proper mappings of balls). It is known by the work of \textit{F. Forstnerič} [Invent. Math. 95, No. 1, 31--62 (1989; Zbl 0633.32017)] that if such maps are smooth enough, they are rational. Hence the problem reduces to solutions to a certain polynomial equality. A particularly simple version of such an equality comes about from studying the CR Gauss equation. \textit{P. Ebenfelt} [Contemp. Math. 681, 125--135 (2017; Zbl 1377.32010)] proved that understanding the possible dimensions of the mapping \(h\) in the equation \[ \lVert h(z) \rVert^2 = \lVert z\rVert^2 q(z,\bar{z}) \] in \({\mathbb{C}}^n\) solves the so-called gap conjecture of Huang-Ji-Yin on which dimensions admit CR sphere maps that do not map into a hyperplane (and are not simply a lower-dimensional map affinely embedded into a higher-dimensional sphere). The version of the conjecture about the polynomial equation above is called the Sum of Squares (SOS) conjecture. Besides solving the gap conjecture it has connections to the Hermitian analogue of Hilbert's 17th problem. In the paper under review, the authors prove the SOS conjecture in the case when \(n=3\) and the matrix of coefficients of \(\lVert h(z) \rVert^2\) is diagonal, in other words, when \(\lVert h(z) \rVert^2\) can be represented by sum of squares of monomials. The technique is to apply tools from commutative algebra, rather than the more combinatorial approach applied to such problems previously. In the process, the authors reprove a degree bound result of the reviewer and \textit{H. Peters} [Mosc. Math. J. 11, No. 2, 285--315 (2011; Zbl 1258.14068)], which can be interpreted as a degree bound on monomial maps between hyperquadrics under a certain extra hypothesis, and which was the basis of the proof of degree bounds for monomial sphere maps.