Iterated double covers and connected components of moduli spaces (Q2564635)

Let \(S\) be a smooth minimal surface of general type over the field of complex numbers and denote by \({\mathcal M} (S)\) the coarse moduli space of surface of general type homeomorphic to \(S\), \({\mathcal M} (S)\) is a quasi-projective variety by Gieseker's theorem. Since \(K^2_S>0\), the divisibility \(r(S)\) of the canonical class \(k_S= c_1(K_S) \in H^2 (S,\mathbb{Z})\) is well defined, i.e. \(r(S)= \max \{r\in\mathbb{N} \mid r^{-1} c_1(S) \in H^2 (S,\mathbb{Z})\}\). \(r(S)\) is a positive integer which is invariant under deformation and the set \[ {\mathcal M}_d (S)=\bigl\{[S']\in {\mathcal M} (S) \mid r(S') =r(S)\bigr\} \] is a subvariety of \({\mathcal M} (S)\) and the number of connected components of \({\mathcal M}_d(S)\) is bounded by a function \(\delta\) of the numerical invariants \(K^2_S\), \(\chi ({\mathcal O}_S)\). It is known that \(\delta\) is not bounded [see \textit{M. Manetti}, Compos. Math. 92, N. 3, 285-297 (1994; Zbl 0849.14016)]. Here we prove that ``in general'' \(\delta\) takes quite large values, more precisely we have Theorem A. For every real number \(4\leq \beta \leq 8\) there exists a sequence \(S_n\) of simply connected surfaces of general type such that: (a) \(y_n= K^2_{S_n}\), \(x_n= \chi ({\mathcal O}_{S_n}) \to\infty\) as \(n\to \infty\). (b) \(\lim_{n\to \infty} (y_n/x_n) =\beta\). (c) \(\delta (S_n) \geq y_n^{(1/5) \log y_n}\) (here \(\delta (S_n)\) is the number of connected components of \({\mathcal M}_d (S_n))\). Theorem A relies on the explicit description of the connected components in the moduli space of a wide class of surfaces of general type whose Chern numbers spread in all the region \({1\over 2} c_2 \leq c^2_1 \leq 2c_2 \). Definition B. A finite map between normal algebraic surfaces \(p: X\to Y\) is called a simple iterated double cover associated to a sequence of line bundles \(L_1, \dots, L_n \in\text{Pic} (Y)\) if the following conditions hold: (1) There exist \(n+1\) normal surfaces \(X=X_0, \dots, X_n=Y\) and \(n\) flat double covers \(\pi_i: X_{i-1} \to X_i\) such that \(p=\pi_n \circ \cdots \circ \pi_1\). (2) If \(p_i: X_i\to Y\) is the composition of \(\pi_j\)'s \(j>i\) then we have for every \(i=1, \dots, n\) the eigensheaves decomposition \(\pi_{i*} {\mathcal O}_{X_{i-1}}= {\mathcal O}_{X_i} \oplus p^*_i (-L_i)\). For any sequence \(L_1, \dots, L_n\in \text{Pic} (\mathbb{P}^1 \times \mathbb{P}^1)\) define \((N(L_1, \dots, L_n)\) as the image in the moduli space of the set of surfaces of general type whose canonical model is a simple iterated double cover of \(\mathbb{P}^1 \times \mathbb{P}^1\) associated to \(L_1,\dots,L_n\). The main theme of this paper is to determine sufficient conditions on the sequence \(L_1,\dots,L_n\) in such a way that the set \(N(L_1, \dots, L_n)\) has ``good'' properties; the conditions we find are summarized in the following definition: Definition C. A sequence \(L_1, \dots, L_n\), \(L_i= {\mathcal O}_{\mathbb{P}^1 \times \mathbb{P}^1} (a_i,b_i)\), \(n\geq 2\) of line bundles on \(\mathbb{P}^1 \times\mathbb{P}^1\) is called a good sequence if it satisfies the following conditions. (C1) \(a_i\), \(b_i\geq 3\) for every \(i=1, \dots, n\). (C2) \(\max_{j<i} \min (2a_i-a_j\), \(2b_i-b_j) <0\). (C3) \(a_n\geq b_n+2\), \(b_{n-1} \geq a_{n-1}+2\). (C4) \(a_i\), \(b_i\) are even for \(i=2, \dots, n\). (C5) For every \(i<n\), \(2a_i -a_{i+1} \geq 2\), \(2b_i- b_{i+1} \geq 2\). The main result we prove is: Theorem D. Let \(L_1,\dots,L_n\) be a good sequence in sense of definition C, then: (a) \(N(L_1, \dots, L_n)\) is a nonempty connected component of the moduli space. (b) \(N(L_1, \dots, L_n)\) is reduced, irreducible and unirational. (For (a) and (b) the condition C5 is not necessary.) (c) The generic \([\text{S}] \in N(L_1, \dots, L_n)\) has \(\Aut (S)= \mathbb{Z}/2 \mathbb{Z}\). (d) If \(M_1, \dots, M_m\) is another good sequence and \(N (L_1, \dots, L_n)= N(M_1, \dots, M_m)\) then \(n=m\) and \(L_i=M_i\) for every \(i=1, \dots, n\). Theorem D gives us some new interesting examples of homeomorphic but not deformation equivalent surfaces of general type.