The Kleiman-Piene conjecture and node polynomials for plane curves in \(\mathbb{P}^3\) (Q1618289)

A $\delta$-nodal curve is a reduced (not necessarily irreducible) projective curve having $\delta\geq 0$ nodes and no other singularity. Let $\mathcal{C}$ be a relative effective divisor on the total space of a threefold $q: \mathcal{S}\rightarrow B$ fibered in projective smooth surfaces, and consider the restriction of $q$ to $\mathcal{C}$. Let $[\overline{B(\delta)}]$ be the class of the locus in $B$ corresponding to $\delta$-nodal fibers. That there exists a universal polynomial in certain Chow classes of $B$, giving the number of $\delta$-nodal fibres for all $\delta\geq 0$, was a conjecture by \textit{S. L. Kleiman} and \textit{R. Piene} [Math. Nachr. 271, 69--90 (2004; Zbl 1066.14063)], who proved the statement only for $\delta$ ranging from $0$ to $8$, leaving open the case $\delta >8$. Good news. That conjecture is now a theorem, due to the author of the beautiful, very interesting and exceptionally well written paper under review. The result is proven by a masterly use of a large spectrum of mathematical tools, from the most classical (Schubert Calculus) to the more modern (BPS Calculus by Pandharipande and Thomas). To be more precise, let \[ \epsilon(a,b,c):=q_*(c_1(O_S(C))^ac_1(T_{S/B})^bc_2(T_{S/B})^c). \] Then Kleiman and Piene proved, under the assumption of some mild and reasonable hypotheses, that there exists a natural non negative cycle $U(\delta)$ supported on $\overline{B(\delta)}$, whose rational equivalence class $[U(\delta)]$ is given by a universal polynomial in $\epsilon(a,b,c)$ for all $0\leq \delta\leq 8$. The challenge they proposed was to prove, or disprove, the claim according which the universal polynomial should work for all $\delta \geq 0$, and not only in the limited range studied by them. The main result of the paper under review, Theorem A, shows that the KP conjecture holds. The proof is offered in Section 5 through a sequence of preliminary propositions, many of which are interesting in their own, such as Theorem 5.3 which is concerned with universal polynomials in Chern classes as well, in a situation that involves a finite sequence of vector bundles on the total space $\mathcal{S}$. \par The strategy of the author is to propose a bivariant class $\gamma(\mathcal{C})\in A^*(B)$, inspired by the BPS (as acronym of Bogomol'nyi-Prasad-Sommerfield) calculus introduced by \textit{R. Pandharipande} and \textit{R. P. Thomas} [J. Am. Math. Soc. 23, No. 1, 267--297 (2010; Zbl 1250.14035)], which shows that $\gamma(\mathcal{C})\cap [B]$ is a natural effective cycle supported on $\overline{B(\delta)}$. This can be universally expressed as a polynomial of degree $\delta$ in the classes $\epsilon(a,b,c)$. The author applies Theorem A to the enumerative geometry of plane curves in $\mathbb{P}^3$. A second main result of the paper is Theorem B (an application of Theorem A). It states that for all $\delta\geq 0$ the number of plane curves of degree $d$ in $\mathbb{P}^3$, intersecting $n:=d(d+3)/2+3-\delta$ general lines, is given by a polynomial function $N_\delta(d)$, of degree $\leq 9+2\delta$. These polynomials are explicitly computed in Appendix A. The author observes that on the basis of his computations, careful described in the paper, the degree of the polynomial function $N_\delta(d)$ actually seems to be exactly $9+2\delta$. In Section 8 some of the formula are checked in low degree. The paper ends with a few low degree checks using the classical methods provided by Schubert Calculus in the Grassmannian of lines $G(1, \mathbb{P}^3)$ (Section 8) and with two computational appendices, followed by a comprehensive reference list.