Symmetric semi-algebraic sets and non-negativity of symmetric polynomials (Q258134): Difference between revisions

After the pioneer work [\textit{G. Stengle}, Math. Ann. 207, 87--97 (1973; Zbl 0253.14001)] the search of algebraic certificates to decide either if a polynomial \(f\in\mathbb R[{\mathtt x}_1,\dots,{\mathtt x}_n]\) assumes only non-negative values on a given semialgebraic subset of \(\mathbb R^n\) or if a semialgebraic set described by polynomial equalities and inequalities is empty or not is a problem around which a lot of research has been developed. The author of the article under review approaches both problems in the particular case where just symmetric polynomials are involved. In this setting it is worthwhile mentioning two basic previous results; the so called half-degree principle and the degree principle, that were obtained in [\textit{V. Timofte}, J. Math. Anal. Appl. 284, No. 1, 174--190 (2003; Zbl 1031.05130)] and [\textit{C. Riener}, J. Pure Appl. Algebra 216, No. 4, 850--856 (2012; Zbl 1242.05272)]. To state them let us denote \(\pi(x)\) the number of distinct coordinates of a point \(x\in\mathbb R^n\) and, for each positive integer \(d\leq n\), let \(\text{A}_d\) be the set of those \(x\in\mathbb R^n\) such that \(\pi(x)\leq d\). The half-degree principle states that if \(f\in\mathbb R[{\mathtt x}_1,\dots,{\mathtt x}_n]\) is a symmetric polynomial of even degree \(2d\) such that \(f(x)\geq0\) for every point \(x\in\text{A}_d\) then \(f(x)\geq0\) for every point \(x\in\mathbb R^n\). The degree principle says that a semialgebraic set \(\text{S}\) can be described as a boolean combination of equalities and inequalities given by polynomials \(f_1,\dots,f_m\) where each \(f_i\) is symmetric with degree at most \(d\), is empty if and only if \(\text{S}\cap\text{A}_d=\varnothing\). For positive integers \(j,n\) denote \({\mathtt p}_j^{(n)}:=\sum_{k=1}^n{\mathtt x}_k^j\). Given a subset \(J:=\{j_1,\dots,j_d\}\subset\mathbb Z^+\), a symmetric polynomial \(f\in\mathbb R[{\mathtt x}_1,\dots,{\mathtt x}_n]\) is called \(J\)-sparse if there exists \(g\in\mathbb R[{\mathtt y}_1,\dots,{\mathtt y}_d]\) such that \(f({\mathtt x}_1,\dots,{\mathtt x}_n)=g\big({\mathtt p}_{j_1}^{(n)},\dots,{\mathtt p}_{j_d}^{(n)}\big)\). The paper, that contains a new proof of the half degree principle (Theorem 3.1), is mainly devoted to prove the following generalization of the degree principle: Theorem 2.6. Let \(J:=\{j_1,\dots,j_d\}\subset\mathbb Z^+\) with \(j_1<\cdots<j_d\) and let \(f_1,\dots,f_m\in\mathbb R[{\mathtt x}_1,\dots,{\mathtt x}_n]\) be symmetric \(J\)-sparse polynomials. Let \(\text{S}\) be a semialgebraic set described as a boolean combination of equalities and inequalities given by \(f_1,\dots,f_m\). Then, {\parindent=6mm \begin{itemize}\item[(i)] If each \(j_i\) is even the set \(\text{S}\) contains at least a point whose coordinates are non-negative and at most \(d\) of them are distinct and non-zero. \item [(ii)] If at least one of \(j_1,\dots,j_d\) is odd and \(\ell:=\min\{j_d,2d+1\}\) then \(\text{S}\cap\text{A}_{\ell}\neq\varnothing\). \end{itemize}} The proof of the main result and the auxiliary lemmas are elementary, clever and very clearly written. There exist some classical precedents of the study of positivity in the symmetric framework. Particularly deep are the papers [\textit{C. Procesi}, Adv. Math. 29, 219--225 (1978; Zbl 0383.12013)] and [\textit{C. Procesi} and \textit{G. Schwarz}, Invent. Math. 81, 539--554 (1985; Zbl 0578.14010)], and it seems interesting to study them again in connection with the paper under review.