The \(abc\)-theorem, Davenport's inequality and elliptic surfaces (Q930128)

In this nice paper complex elliptic surfaces \(\pi: X\to \mathbb P^1\) with a section are studied. It is well-known that if such a surface has the additional property that all singular fibers are semi-stable then \(\pi\) has at least \(2\chi+2\) singular fibers. It is shown that for each \(\chi>0\) (i) there exists a semistable elliptic surface \(X\) over \(\mathbb P^1\) with a section and \(2\chi+2\) singular fibers. (ii) The number of isomorphism classes of such surfaces is finite. (iii) In each isomorphism class there exists a surface defined over some algebraic number field. (iv) Any such surface is extremal, i.e., the Mordell-Weil rank is zero and \(h^{1,1}(X)=\rho(X)\). The proof consists of a study of elliptic surfaces of the from \(y^2=x^3-3f(t)x-2g(t)\) such that \(\deg(f)=2(N_0(h)-1)\) and \(\deg(g)=3(N_0(h)-1)\), where \(N_0(h)\) is the number of distinct zeroes of \(h=f^3-g^2\). Such polynomials are `extremal' in the sense that from the \(abc\)-theorem for polynomials it follows that \(\deg(f)\leq 2(N_0(h)-1)\) and \(\deg(g) \leq 3(N_0(h)-1)\). This links the extreme case of the \(abc\)-theorem with extremal elliptic surfaces. Reviewer's remark: It should be remarked that these results are already present in the literature. These results are somewhat hidden and the methods differ notably from the paper under review: From, e.g., Theorem 6.3 in the reviewer's paper [Mich. Math. J. 52, No. 1, 141--161 (2004; Zbl 1076.14044)] it follows that a semistable elliptic surface over \(\mathbb P^1\) is extremal if and only if it has \(2\chi+2\) singular fibers. This implies (iv). The statements (ii) and (iii) are then immediate consequences of the main result of \textit{M. Nori} [Topology 24, 175--186 (1985; Zbl 0575.14034)] and of Theorem 6.3 in \textit{R. Kloosterman} [Mich. Math. J. 52, No. 1, 141--161 (2004; Zbl 1076.14044)]. The point (i) is implicitly present at several places in the literature, e.g., the first step in the proof of Theorem 5.1 in the reviewer's paper [J. Differ. Geom. 76, No. 2, 293--316 (2007; Zbl 1141.14019)] implies point (i).