On the Goulden-Jackson-Vakil conjecture for double Hurwitz numbers (Q2143650)

\textit{Hurwitz numbers} [\textit{A. Hurwitz}, Math. Ann. 39, 1--61 (1891; JFM 23.0429.01)] count branched covers of \(\mathbb{C}P^1\) subject to topological specifications such as the genus of the covering, the degree of the covering and/or the ramification structure. The ELSV formula [\textit{T. Ekedahl} et al., Invent. Math. 146, No. 2, 297--327 (2001; Zbl 1073.14041)] is central in the modern narrative of Hurwitz numbers, and expresses single Hurwitz numbers (Single in the sense that non-simple branching may occur only over \(\infty\in\mathbb{C}P^1\).) as an integral over \(\overline{\mathcal{M}}_{g,n}\) -- specifically, as sums of top intersection numbers on \(\overline{\mathcal{M}}_{g,n}\). The titular \textit{double Hurwitz numbers} \(h_{g;\mu;\nu}\) count the number of genus \(g\) ramified covers (weighted by automorphisms) of \(\mathbb{C}P^1\) with two ramification points -- set to be \(\infty\) and \(0\) without loss of generality, so that \begin{itemize} \item the partitions \(\mu\) and \(\nu\) respectively specify the branching over \(\infty\) and \(0\); \item the remaining ramifications are all simple and occur over chosen points on \(\mathbb{C}P^1-\{0,\infty\}\). \end{itemize} Given a partition \(\mu=(\mu_1,\ldots,\mu_n)\), if the partition \(\nu\) has a single part equal to \(|\mu|=\mu_1+\ldots+\mu_n\), then we refer to \(h_{g;\mu;\nu}\) as a \emph{one-part double Hurwitz number} \(h^{\text{one-part}}_{g;\mu}\). Goulden-Jackson-Vakil conjectured a ELSV-type formula which re-interprets one-part double Hurwitz number in terms of top intersection numbers over a hypothetical compactification of the \emph{universal Picard variety} \(\mathrm{Pic}_{g,n}\). To clarify, they conjecture the existence of the following objects (see [\textit{I. P. Goulden} et al., Adv. Math. 198, No. 1, 43--92 (2005; Zbl 1086.14022)] for precise details): \begin{itemize} \item \textbf{\(\overline{\mathrm{Pic}}_{g,n}\) moduli space.} The universal Picard variety \(\mathrm{Pic}_{g,n}\) is the moduli space of genus \(g\) smooth curves paired with degree \(0\) line bundles, and \(\overline{\mathrm{Pic}}_{g,n}\) is a conjectural compactification of \(\mathrm{Pic}_{g,n}\). The compactified universal Picard variety \(\overline{\mathrm{Pic}}_{g,n}\) is expected to carry a virtual fundamental class of complex dimension \(4g-3+n\). \item \textbf{\(\Psi_i\)-classes.} The \(\Psi_i\in H^2(\overline{\mathrm{Pic}}_{g,n})\) are first Chern classes of conjectural line bundles on \(\overline{\mathrm{Pic}}_{g,n}\). They fill the role of \(\psi\)-classes on \(\overline{\mathcal{M}}_{g,n}\) in the classical ELSV formula. \item \textbf{\(\Lambda_{2k}\)-classes.} The \(\Lambda_{2k}\in H^4(\overline{\mathrm{Pic}}_{g,n})\) are Chern classes of a conjectural rank \(2g\) vector bundle. This supplants the role of Chern classes of the dual of the Hodge vector bundle in the classical ELSV formula. \item \textbf{Morphisms.} As a part of the conjecture, the forgetful morphism \(\pi:\overline{\mathrm{Pic}}_{g,n+1}\to\overline{\mathrm{Pic}}_{g,n}\) and its natural sections \(\sigma_i:\overline{\mathrm{Pic}}_{g,n}\to\overline{\mathrm{Pic}}_{g,n+1}\) should satisfy certain relations involving the aforementioned fundamental classes and \(\Psi_i\) classes. \end{itemize} Their conjectural ELSV-type formula then takes the form: \[ h^{\text{one-part}}_{g;\mu}=(\mu_1+\cdots+\mu_n)\int_{\overline{\mathrm{Pic}}_{g,n}}\frac{\sum_{k=0}^{g}(-1)^k\Lambda_{2k}}{\prod_{i=1}^n(1-\mu_i\Psi_i)}. \] Motivated by the Goulden-Jackson-Vakil conjecture, Do and Lewańsky establish several ELSV-type avatars of the one-part double Hurwitz numbers as top intersection numbers of characteristic classes on various moduli spaces: \begin{itemize} \item[1.] \textbf{Hodge classes on moduli spaces of stable maps to the classifying space \(\mathcal{B}\mathbb{Z}_{\mu_1+\cdots+\mu_n}\):} for each partition \(\mu\), the one-part double Hurwitz number \(h^{\text{one-part}}_{g;\mu}\) may be expressed in terms of top intersection numbers over the moduli space \[ \overline{\mathcal{M}}_{g;-\mu_1,\ldots,-\mu_n}(\mathcal{B}\mathbb{Z}_{\mu_1+\cdots+\mu_n}) \] of principal \(\mathbb{Z}_{\mu_1+\cdots+\mu_n}\)-bundles over stable curves of genus \(g\) with \(n\) marked points with \(\mathbb{Z}_{\mu_1+\cdot+\mu_n}\) monodromy specified by \(\mu\) [\textit{P. Johnson} et al., Mich. Math. J. 60, No. 1, 171--198 (2011; Zbl 1222.14119)]. The role of \(\Lambda_{2k}\) is taken up by the Hodge classes \(\overline{\lambda}_k\). \item[2.] \textbf{Chiodo classes on moduli spaces of \(r\)-spin curves:} for each partition \(\mu\), \(h^{\text{one-part}}_{g;\mu}\) may be expressed in terms of top intersection numbers over the moduli space \[ \overline{\mathcal{M}}^{\mu_1+\cdots+\mu_n,\mu_1+\cdots+\mu_n}_{g,n;-\mu_1,\ldots,-\mu_n}(\mathcal{B}\mathbb{Z}_{\mu_1+\cdots+\mu_n}) \] of \(r\)-spin curves of genus \(g\), with \(n\) marked points [\textit{E. Witten}, in: Topological methods in modern mathematics. Proceedings of a symposium in honor of John Milnor's sixtieth birthday, held at the State University of New York at Stony Brook, USA, June 14-June 21, 1991. Houston, TX: Publish or Perish, Inc.. 235--269 (1993; Zbl 0812.14017)] and the role of \(\Lambda_{2k}\) is taken up by Chiodo classes [\textit{A. Chiodo}, Compos. Math. 144, No. 6, 1461--1496 (2008; Zbl 1166.14018)] in this context. \item[3.] \textbf{Tautological classes on moduli spaces of stable curves (version~1):} the third setting follows from the general theory of topological recursion. Specifically, one-part double Hurwitz numbers may be generated using the machinery of topological recursion via an infinite discrete family of spectral curves (parametrised by \(\mu_1+\cdots+\mu_n\)), and thus necessarily [\textit{B. Eynard}, Commun. Number Theory Phys. 8, No. 3, 541--588 (2014; Zbl 1310.14037)] arise as tautological intersection numbers on moduli spaces of stable curves. In particular, the formula that Do and Lewańsky give expresses \(h^{\text{one-part}}_{g;\mu}\) as the intersection numbers over \(\overline{\mathcal{M}}_{g,n}\), with the role of \(\Lambda_{2k}\) taken up by pushforwards of Chiodo classes. \item[4.] \textbf{Tautological classes on moduli spaces of stable curves (version~2):} all hitherto examples have been over moduli spaces of complex dimension \(3g-3+n\), and the authors give an argument (by the dilaton equation) whereby one adds additional marked points and hence beefs up the moduli space to \(\overline{\mathcal{M}}_{g,n+g}\). The specific construction given is non-canonical, but nevertheless realises one-part double Hurwitz numbers as top intersection numbers in a space which is ``a compactification of an open substack of the Picard stack''. \end{itemize} Although Do and Lewańsky's top intersection number interpretations of the one-part double Hurwitz numbers fall just shy of satisfying the requirements of the Goulden-Jackson-Vakil conjecture, the results they derive are spiritually in-keeping with one of the central motivations for these studies: the desire to better understand the geometry of the moduli space of curves.