The moduli space of surfaces with \(K^2 = 6\) and \(p_g = 4\) (Q2509013)

\textit{E. Horikawa} [Invent. Math. 47, 209--248 (1978; Zbl 0409.14005)] showed that the moduli space \(\mathcal{M}_{6,4}\) of minimal surfaces of general type with \(K^2=6, \; p_g=4\) has exactly 4 irreducible components and at most 3 connected components. He obtained this result by making a stratification of \(\mathcal{M}_{6,4}\) in 11 non empty locally closed strata, \(I_a,I_b,II,III_a,III_b,IV_{a_1},IV_{a_2},IV_{b_1},IV_{b_2},V_1,V_2\), and then analysing some of their local deformations. The question of specializations among the three components remained open. The main result of the paper under review is that the moduli space \(\mathcal{M}_{6,4}\) has at most two connected components. More precisely, the authors prove that surfaces of type \(III_b\) deform to surfaces of type \(II\). In Horikawa's terminology, a surface of type \(II\) is one with canonical map of degree 3, and then the canonical image is necessarily a quadric cone, while a surface of type \(III_b\) is one which has no genus 2 pencil, and whose canonical system has a fixed component. In both cases, there is a decomposition of the canonical system of the form \(K_S=2D+G\), where \(G\) is a fundamental cycle and \(D\) is a genus 3 pencil with one simple base point. Considering the canonical model \(X\) of \(S\), one obtains that the canonical class of \(X\) is then 2-divisible (as a Weil divisor), since \(K_X=2L\), \(L\) being the image of \(D\) in \(X\). So for both type of surfaces one has a semicanonical ring \(\mathcal{B}=:\bigoplus{H^0(X,nL)}\). In case \(II\), \(X\) is hypersurface of degree 9 in \(\mathbb{P}(1,1,2,3)\) (as already remarked by \textit{F. Catanese} [Invent. Math. 98, No.~2, 293--310 (1989; Zbl 0701.14039)]), so \(\mathcal{B}\) is a Gorenstein ring of codimension 1; in case \(III_b\), the authors prove that \(X\) can be embedded in \(\mathbb{P}(1,1,2,3,4,5,6)\) and \(\mathcal{B}\) is Gorenstein of codimension 4. Then the strategy is to construct a flat family of rings over a small disk, whose central fibre is the ring of a surface of type \(III_b\), and whose general fibre is a ring of a surface of type \(II\). In order to describe the semicanonical ring and its deformations in case \(III_b\), the authors use formats of \(4 \times 4\) Pfaffians of antisymmetric, extrasymmetric \(6 \times 6\) matrices (for more on formats, see M. Reid: Surfaces with \(p_g=3,\; K^2=4\) according to [\textit{E. Horikawa} and \textit{D. Dicks}, in: Proceedings of Algebraic Geometry mini Symposium, Tokyo Univ. 1--22 (1989); \textit{M. Reid}, Graded rings and birational geometry. in: Proceedings of Algebraic Geometry Symposium, Kinosaki, 1--72 (2006) and \textit{I. Bauer, F. Catanese}, \textit{R. Pignatelli}, in: Complex geometry. Coll. papers ded. Hans Grauert's 70th birthday. 37--72 (2002; Zbl 1052.14043)]). The deformation trick consists in filling entries of homogeneous degree \(0\) in the matrix with a parameter \(t\) (and respecting the extrasymmetric format). For \(t\neq 0\) the outcome is that from the 9 Pfaffians one can eliminate the variables of respective degrees 4,5,6 and then one is left with variables of respective degrees 1,1,2,3 and with a single equation of degree 9, i.e. one has a semicanonical ring of type \(II\), and this concludes the proof.