On the existence of covers of \(\mathbb{P}_1\) associated to certain permutations (Q1650242)

Let \(S\) be a closed Riemann surface, of genus \(g\), and let \(\delta:S \to \widehat{\mathbb C}\) be a non-constant holomorphic map, of degree \(d \geq 2\), whose branch values are contained in the set \(\{\infty,0,1\}\). In this case, \(S\) is called a Belyi curve and \(\delta\) a Belyi map. Associated to \(\delta\) is a transitive subgroup of the permutation group \({\mathscr S}_{d}\), generated by three permutations \(\sigma_{0}, \sigma_{1}, \sigma_{\infty}\) satisfying the relation \(\sigma_{0}\sigma_{1}\sigma_{\infty}=1\). These permutations are obtained as follows. The \(\delta\)-preimage of the closed interval \([0,1]\) induces a bipartite graph on \(S\) (the points in \(\delta^{-1}(0)\) are the black vertices and those in \(\delta^{-1}(1)\) the white ones) producing a \(2\)-cell decomposition of it (a dessin d'enfant as introduced by Grothendieck). Make a labelling of the \(d\) edges using numbers in \(\{1,\ldots,d\}\). The permutation \(\sigma_{0}\) (respectively, \(\sigma_{1}\)) is obtained as the product of the cycles of the edges following the counter-clockwise orientation at each of the black (respectively, white) vertices. The cycle structure of each of the three permutations is encoded by the tuple \((a_{1},\ldots,a_{\alpha};b_{1},\ldots,b_{\beta};c_{1},\ldots,c_{\gamma})\), called the passport of \(\delta\), where \(\alpha\) (respectively, \(\beta\) and \(\gamma\)) denotes the number of cycles of \(\sigma_{0}\) (respectively, \(\sigma_{1}\) and \(\sigma_{\infty}\)) and \(a_{j}\) (respectively, \(b_{i}\) and \(c_{k}\)) are the length of its cycles. It holds that \(a_{1}+\cdots+a_{\alpha}=b_{1}+\cdots+b_{\beta}=c_{1}+\cdots+c_{\gamma}=d\) and, by the Euler formula, that \(2-2g=\alpha+\beta+\gamma-d\). Now, one may consider any tuple \((a_{1},\ldots,a_{\alpha};b_{1},\ldots,b_{\beta};c_{1},\ldots,c_{\gamma})\) of integers \(a_{j},b_{i},c_{k} \geq 1\), satisfying \(a_{1}+\cdots+a_{\alpha}=b_{1}+\cdots+b_{\beta}=c_{1}+\cdots+c_{\gamma}=d\), and to ask if it is the passport of some Belyi map of degree \(d\). M. Zieve considered the tuples \((3^m;2,4,3^{m-2};3^m)\), \((2^{2m};3,5,4^{m-2};4^{m})\) and \((2^{3m};5,7,6^{m-2};3^{2m})\), where \(a^{r}\) means that the integer \(a\) appears \(r\) times, and experimentally evidencied the impossibility for them to be the passport of some Belyi map. In the paper under review, the authors provide a simple and short argument to see that they are in fact impossible passports. The main idea is that if one of them is the passport of a Belyi map \(\delta:S \to \widehat{\mathbb C}\), then \(S\) has genus \(g=1\). By working with the divisors of \(\delta\), \(1-\delta\) and the form \(d \delta\) (which happens to be also the divisor of a meromorphic map \(\eta:S \to \widehat{\mathbb C}\)), one gets the existence of a meromorphic map \(\rho:S \to \widehat{\mathbb C}\) of degree one, a contradiction (for instance, for the first tuple \(\rho=\delta^{2}(1-\delta)^{2}/\eta^{3}\)). In the paper it is also studied Belyi maps (necessarily of genus one) with passport \((3^m;3^m;3^m)\); these are obtained as quotients of the classical Fermat curve of degree three.