[Paper Review] Characterization of circuits supporting polynomial systems with the maximal number of positive solutions

Boulos El Hilany|arXiv (Cornell University)|Mar 6, 2016

Polynomial and algebraic computation15 references4 citations

TL;DR

This paper fully characterizes circuits in $ℝ^n$ that support polynomial systems with the maximal number of $n+1$ non-degenerate positive solutions. Using Grothendieck’s real dessins d’enfant for restrictions and Viro’s combinatorial patchworking to construct systems, it proves a circuit supports such a system if and only if its affine relation satisfies specific alternating inequalities on partial sums of coefficients, depending on the parity of $n$. This provides a complete classification of extremal positive solution supports in the fewnomial setting.

ABSTRACT

A polynomial system with $n$ equations in $n$ variables supported on a set $\\mathcal{W}\\subset\\mathbb{R}^n$ of $n+2$ points has at most $n+1$ non-degenerate positive solutions. Moreover, if this bound is reached, then $\\mathcal{W}$ is minimally affinely dependent, in other words, it is a circuit in $\\mathbb{R}^n$. For any positive integer number $n$, we determine all circuits $\\mathcal{W}\\subset\\mathbb{R}^n$ which can support a polynomial system with $n+1$ non-degenerate positive solutions. Restrictions on such circuits $\\mathcal{W}$ are obtained using Grothendieck's real dessins d'enfant, while polynomial systems with $n+1$ non-degenerate positive solutions are constructed using Viro's combinatorial patchworking.

Motivation & Objective

• To determine all circuits in $ℝ^n$ that can support a polynomial system with $n+1$ non-degenerate positive solutions.
• To establish necessary and sufficient conditions on the affine relations of such circuits for the maximal solution count to be achieved.
• To unify and extend prior results on fewnomial bounds and maximally positive systems by providing a complete classification up to affine equivalence.
• To use Viro’s patchworking to explicitly construct systems achieving the $n+1$ solution bound.

Proposed method

• Apply Grothendieck’s real dessins d’enfant to derive restrictions on the coefficient structure of circuits supporting $n+1$ positive solutions.
• Use Viro’s combinatorial patchworking to construct polynomial systems with $n+1$ non-degenerate positive solutions from circuits satisfying specific coefficient inequalities.
• Represent the circuit’s affine relation as a signed sum with positive coefficients on two disjoint subsets of size $\lfloor(n+2)/2\rfloor$ and $\lceil(n+2)/2\rceil$, respectively.
• Define partial sums of coefficients on both sides of the affine relation and impose strict interleaving inequalities: $\sum_{i=1}^r \alpha_i < \sum_{i=s+1}^{s+r} \beta_i < \sum_{i=1}^{r+1} \alpha_i$ for $r=1,\dots,s-1$, with adjustment for odd $n$.
• Construct a Viro polynomial $G_t$ by gluing truncated polynomials $P_{i,t}$ with exponents determined by the partial sums $p_i$, ensuring the lower hull of the Newton polytope matches the required subdivision.
• Prove that each facial subpolynomial of $G_t$ is a binomial with opposite-sign coefficients, guaranteeing one simple positive root per segment, and apply Lemma 4.1 to conclude $n+1$ positive roots for small $t>0$.

Experimental results

Research questions

• RQ1 Which circuits in $\mathbb{R}^n$ can support a system of $n$ polynomials with $n+1$ non-degenerate positive solutions?
• RQ2 What structural conditions on the coefficients of the affine relation of a circuit are necessary and sufficient for achieving the maximal number of positive solutions?
• RQ3 How can Viro’s patchworking method be adapted to construct such systems explicitly?
• RQ4 What role do Grothendieck’s real dessins d’enfant play in restricting the possible coefficient configurations of such circuits?

Key findings

• A circuit $\mathcal{W} \subset \mathbb{R}^n$ supports a system with $n+1$ non-degenerate positive solutions if and only if its affine relation satisfies the specified interleaving inequalities on partial sums of positive coefficients.
• The construction via Viro’s patchworking ensures that for sufficiently small $t>0$, the polynomial $G_t$ has exactly $n+1$ non-degenerate positive roots, one per segment of the lower hull of the Newton polytope.
• The condition on the affine relation is both necessary and sufficient: if the inequalities fail, the system cannot achieve $n+1$ positive solutions.
• For even $n$, the partial sums of the coefficients on the two sides of the affine relation must satisfy $\sum_{i=1}^r \alpha_i < \sum_{i=s+1}^{s+r} \beta_i < \sum_{i=1}^{r+1} \alpha_i$ for $r=1,\dots,s-1$, where $s = (n+2)/2$.
• For odd $n$, the inequality chain is $\sum_{i=1}^r \alpha_i < \sum_{i=s+2}^{s+r+1} \beta_i < \sum_{i=1}^{r+1} \alpha_i$, reflecting a shifted indexing due to the odd number of points.
• The result generalizes and completes earlier work by Bihan and Sottile, providing a full classification of extremal circuits in the fewnomial theory of positive solutions.