The minimal regular model of a Fermat curve of odd squarefree exponent and its dualizing sheaf (Q2175454)

The paper under review is partly based upon an earlier paper by \textit{W. G. McCallum} [Prog. Math. 26, 57--70 (1982; Zbl 0542.14019)]. In this paper, McCallum analyzed the case when the degree of the Fermat curve is an odd integer \(p\) over \(K=\mathbb{Q}_{p}(\xi)\), where \(\xi=\exp(\frac{2\pi}{p})\) is the \(p\)-th root of unity. In this situation, the Fermat curve \(F_{K}:x^{p}+y^{p}=1\) has bad reduction at the prime \((\pi)=1-\xi\) of \(\mathcal{O}_{K}=\mathbb{Z}_{p}(\xi)\) lying over \(p\). To analyze this situation carefully McCallum introduced an auxillary curve \(C_{s,K}\): \[ y^{p}=x^{s}(1-x),\ s=1, 2,\dots, p-2 \] The analysis for \(F_{K}\) can thus be recovered from the analysis of \(C_{s,K}\) via the map from \(C_{s,K}\) to \(F_{K}\): \[ f_{s, K}: (x,y)\rightarrow (x^p, x^s y) \] Much of the analysis relies on an auxillary polynomial \(\phi(x,y)=((x+y)^{p}-x^p-y^p)/p\). The final result can be summarized as following: Theorem (page~59 of [loc. cit.]). The curve \(F_{K}\) has a regular model \(F\) whose geometric special fibre \(\overline{F}_{0}\) has configuration in diagram~3 (p.~69 of [loc. cit.]). The vertical components are parametrized by the roots of \(\phi(x,-y)\) as follows: \begin{itemize} \item The components labelled \(0, 1, \infty\) correspond to the factors \(x,x-y, y\) of \(\phi(x,-y)\). \item The components labelled \(\alpha\) correspond to double factors \((x+\alpha y)^2\) of \(\phi(x,-y)\), \(\alpha\in k\). \item The components labelled \(\beta\) correspond to simple factors \((x+\beta y)\) of \(\phi(x,-y)\), \(\beta\not \in k\). \end{itemize} The method involved in the proof is rather standard by working with the projective normalization of the curve \(C_{s,K}\) and \(F_{K}\) and repeatedly blowing up the singular points. McCallum apparently had further plans to study the Jacobian associated to \(F_{K}\) itself, as the map \(f_{s,K}\) induces an isogeny on the Jacobians. This leads to his later paper on the Tate-Shafarevich group of Fermat curves. The present paper continues the investigation of the Fermat curve when \(N\) is a square-free odd integer. While the methodologies are similar by utilizing the map \(\phi\), the analysis has become much more involved. For example to obtain the normalization, the standard blow-up of \(\mathcal{X}\) has become much more complicated (see the proof of Proposition~3.7, for example). The main result is summarized in Theorem~3.13, which gives a description of the geometric special fibres associated to the local minimal model. The global minimal model is then ``glued'' from the local minimal model. The main interest of current paper stems from the following result: Theorem. Let \(N>0\) be an odd square-free integer with at least two prime factors, and let \(\mathfrak{F}^{\min}_{N}\) be the minimal regular model of the Fermat curve \(F_{N}\) over \(\textrm{Spec}(\mathbb{Z}[\xi_{N}])\). Then the arithmetic self-intersection number of its dualizing sheaf equipped with the Arakelov metric satisfies \[ \omega^2_{\mathfrak{F}^{\min}_{N}}\le (2g-2)[\mathbb Q(\xi_{N}):\mathbb Q](k_1 \log(N)+k_2)+(2g-2)\sum_{p|N}\frac{\phi(N)}{\phi(p)}(\cdots) \log(p) \] where the expression in \((\cdots)\) is a function of order \(O(\log\log(N))\). This is a great improvement over McCallum's result as it connects the work on local minimal models to (global) arithmetic intersection theory. However, the reader should be aware that the authors' claim that ``[\ldots]\ While no bounds in complete families are known to date\ldots'' held only true at the time of submission (June 2016). By 1994--1995 \textit{C. Soulé} [Invent. Math. 116, No.~1--3, 577--599 (1994; Zbl 0834.14013)] had shown that such estimates are possible if one can estimate norm of sections of \(H^{1}(X, L^{-1})\), and by 2017 based upon previous work of K.~Yamaki, \textit{R.~Wilms} [``The delta invariant in Arakelov geometry'', \url{https://hdl.handle.net/20.500.11811/6757}; Invent. Math. 209, No.~2, 481--539 (2017; Zbl 1390.14073)] derived effective upper bounds for the case of hyper-elliptic curves (thus resolving the case for \(g=2\) completely). The authors' result thus provides effective upper bounds for Fermat curves (the first author and \textit{U.~Kühn} [``On the arithmetic self-intersection numbers of the dualizing sheaf for Fermat curves of prime exponent'', Preprint, \url{arXiv:0906.3891}] provided similar results for the case when \(N=p\)). The main questions now left are the following: \begin{itemize} \item For Fermat curve of low genus (\(g=3\), for example) -- how does the current result on lower/upper bounds expand our current knowledge of arithmetic intersection theory? Can we say more about other curves in \(M_{g}\)? \item It was announced by \textit{M.~van Frankenhuysen} [J. Number Theory 95, No.~2, 289--302 (2002; Zbl 1083.11042)] that providing similar upper bounds for the self-intersection number would suffice to prove Vojta's height inequality (Corollary~2.3). \end{itemize} I have followed up with a personal email to Prof. Vojta, who confirmed this should be correct. In the same paper Machiel van Frankenhuysen claimed that a proof of Vojta's height inequality will imply the ABC conjecture. Thus, the main issue is whether the current upper bounds suffice to prove Vojta's height inequality via a comparison of heights by translating \(\omega^2\) to the degree associated to a section of a line bundle over the curve. Traditionally this is done via Noether's formula in Faltings-Riemann-Roch. If this is the case and we can ``fill in details'' of each step, then this may lead to potentially interesting results.