no. 17 · Wednesday 18 November 2026
Towards the componentwise three-soliton conjecture: cones, collisions, soft momenta and graphs
Abstract
The componentwise three-soliton conjecture of Part~I states that if an even polynomial $P$ with $P(0)=0$ satisfies Hirota's three-soliton condition on a non-hyperplane component of multiplicity one of its dispersion variety, then it satisfies the $N$-soliton condition there for every $N$. We prove restrictions on components carrying this condition and classify the polynomials satisfying it on several families of components, with no degree bound. Restrictions: if the top homogeneous part of the equation of the component has a nonlinear irreducible factor, then $P$ is a multiple of a quadratic form; the order of $P$ at the origin is tied to the multiplicities of the component and of the doubled component; a component through the origin has odd multiplicity; the condition passes to all limits of the rescaled component, and the Newton polygon of a plane component has edges of only ten directions. Classifications: on the cubic graph $v=u^3$, where the four-soliton condition holds exactly in multiplicity one; on the graphs $v=q(u)$ with $°q\ge3$, where $P$ exists only if $q$ is an odd cubic or an odd quintic, so the conjecture holds for all graphs of polynomials; and for three families of Hietarinta's list, where each irreducible member is, up to a constant factor, the only such polynomial on its dispersion variety. All three satisfy the $N$-soliton condition for every $N$, so the conjecture holds for them: two belong to the BKP hierarchy, and for Hietarinta's quartic equation, known to satisfy it for $N\le4$, we prove it by fusing resonant momenta and counting zeros on its curve. For translated hyperbolas we exclude the one through the origin and all $P$ of degree at most eight. The solutions of multiplicity one on graphs are known bilinear equations; new is that there are no others. The conjecture remains open.
- fields
- Mathematical Physics