Enriques surfaces I (Q7006915)

This is the second edition of a book that appeared in 1989 at Birkhäuser Verlag (Boston, Basel, Berlin) [Zbl 0665.14017]. The second author kept insisting that some day part 2 of the book would appear. So, now it finally appeared, after 36 years. As a consequence, part 1, the book under review, required a thorough revision. This was clearly necessary since many new results on Enriques surfaces have been obtained in the meantime. The authors have chosen to place all preliminaries (also those needed for volume 2) in this revised part 1. Now this chapter counts 198 pages instead of just 62 in the previous edition. There are other notable differences with the first edition. Apart from some rearrangements, the most prominent change is the addition of a new chapter on moduli spaces about which more will be said below.\N\NIn this book a surface always is a smooth and proper algebraic surface over an algebraically closed field \(k\). An \textit{Enriques surface} can be defined in several ways. Over the complex numbers one usually defines them as surfaces with \(b_1 \text{(the first Betti number)}=0\) and with \(K^{\otimes 2}\) (the square of the canonical bundle) trivial, while \(K\) itself is not trivial. One can then deduce that the Kodaira dimension is 0 and that \(b_2=10\). One can show that these two conditions are equivalent to the previous definition. This approach can be seen to imply that the universal cover is an étale degree 2 cover which has trivial canonical bundle. This is by definition a \textit{K3 surface}. To obtain examples one thus may start with a suitable \(K3\) surface. It turns out that the family of \(K3\) surfaces which are a complete intersection of 3 quadrics in \(\mathbb{P}^5\) contains a 10-dimensional subfamily, whose members are invariant under a fixed point free involution and so these give the searched for examples. In fact in a certain sense this gives almost all of them.\N\NSince \(k\) is not necessarily of characteristic 0, some care is needed to define Enriques surfaces in the setting of the book. One rather uses the given consequence of the definition over \(\mathbb{C}\): an Enriques surface is a surface of Kodaira dimension 0 with \(b_2=10\). If the characteristic is not equal to 2, there again is an étale cover which is a \(K3\) surface, but in characteristic 2 the situation changes dramatically. It is still the case that \(b_1=0\) and that \(K^{\otimes 2}\) is trivial, but \(K\) itself can also be trivial. Moreover, since the Hodge versus de Rham spectral sequence need not degenerate, one may have \(b_1 < h^{1,0}+h^{0,1}=2h^{0,1}\), which in this case comes down to \(h^{0,1}= h^{1,0}=1\). This is shown to be equivalent to \(K\) being trivial. These anomalies give rise to three types of Enriques surfaces: \textit{the classical ones} for which \(K\) is not trivial, \textit{ordinary ones}, if \(h^{0,1}= h^{1,0}=1\) and Frobenius is bijective on the 1-dimensional space \(H^{0,1}\), and the \textit{supersingular ones} if Frobenius is 0 on this 1-dimensional space. Of course the question then arises: do such surfaces exist? In \S 1.6 a variant of the above example of the intersection of 3 quadrics in \(\mathbb{P}^5\) invariant under a fixed point free involution is described, as found by E.~Bombieri, D.~Mumford and M.~Reid in 1977. The examples in this context are invariant under one of the 3 groups schemes of length 2 in characteristic 2 and do give the 3 possible types in that characteristic. Now also most Enriques surfaces arise in this way.\N\NLet me say a bit more about the contents of this book. Chapter 0 (Preliminaries) contains a lot of material such as on group schemes, cyclic covers, inseparable morphisms, Picard schemes and on the cohomology of algebraic surfaces. This is especially useful for someone working in a characteristic free setting. Not all of this material will be used in the book, as the authors state in the introduction, but it is certainly a good idea to have this material available at one place since it is widely scattered in the literature.\N\NChapter 1 contains the basic material concerning Enriques surfaces, their \(K3\)-covers, their cohomological invariants and the so-called Enriques lattice. Unlike for \(K3\) surfaces this lattice can be defined algebraically by using the intersection product on the lattice formed by the classes of numerically equivalent divisors. As in the complex case it is a unimodular lattice of signature \((1,9)\) and can be investigated by lattice-theoretical means.\N\NLinear systems are investigated in Chapter 2 using Riemann-Roch and Serre duality and this is used in the next chapter where various projective models that were hinted at in Chapter 1 are more fully treated. The oldest of those is due to Enriques himself, namely that of a sextic surface passing doubly through the edges of a tetrahedron (and having cusps at its vertices). In his thesis (1960), M.~Artin worked out most of the details (in all characteristics), and a modern treatment is given in this section. Next, other classical models using 2:1 covers to del Pezzo surfaces are reviewed, intimately related to the so-called Horikawa models. Finally, the Bombieri-Mumford-Reid construction is more fully explained.\N\NChapter 4 gives an extensive treatment of the genus 1 fibrations on Enriques surfaces and on related rational surfaces. Here the emphasis is on positive characteristic, where some new phenomena occur. For instance, one may have pencils of curves of arithmetic genus 1 none of whose members are smooth. The latter are called quasi-elliptic pencils and occur only in characteristics 2 and 3, but cause a lot of technically involved special arguments.\N\NFinally, Chapter 5 describes the moduli theory. Over the complex numbers this can be done using the period map of the doubly covering \(K3\)-surfaces, as initiated by E.~Horikawa and V.~Nikulin around 1978. Here the precise result appears as Theorem~5.3.9. The authors consider much more such as several classical compactifications of the moduli spaces. Its boundary can be described using semi-stable degenerations of families of Enriques surfaces, in particular the so-called Kulikov degenerations (these were also considered in U. Person's 1975 thesis). The chapter is carefully set up with broad attention for the modern view and techniques of moduli spaces and stacks, so that in section 5.11 the authors can explain the rather recent results on moduli in positive and mixed characteristic due to C. Liedtke (2015) following previous work by T. Ekedahl, N. Shepherd-Barron, and J. Hyland (2004,2012).\N\NAs should be clear from what has just been written, this is a technical book of encyclopedic nature. It has been carefully written and contains much to the delight of the specialist. Any one working on \(K3\) surfaces and Enriques surfaces should have it ready on the bookshelf.\N\NHowever, to understand all the details of the constructions requires, besides the necessary background knowledge, stamina and perseverance. As such it is barely suited for beginning researchers. Some suggestions to aid them now follow. To begin with, they need to be well acquainted with algebraic surfaces, for instance as explained in the book by \textit{A.~Beauville} [Complex algebraic surfaces. 2nd ed. Cambridge: Cambridge Univ. Press (1996; Zbl 0849.14014)]. Also instructive is some additional background on complex Enriques surfaces which for example can be found in Chapter VIII of \textit{W. P. Barth} et al. [Compact complex surfaces, 2nd enlarged ed. Berlin: Springer (2004; Zbl 1036.14016)]. Finally, it helps to know the basics involved in the classification of algebraic surfaces in positive characteristic. Undoubtedly, still today the best technical introduction consists of the three well written basic articles part I by \textit{D. Mumford} [in: Global Analysis, Papers in Honor of K. Kodaira 325--339 (1969; Zbl 0188.53201)], part II, by \textit{E. Bombieri} and \textit{D. Mumford} [in: Complex Anal. Algebr. Geom., Collect. Pap. dedic. K. Kodaira, 23--42 (1977; Zbl 0348.14021)], and part III, also by \textit{E. Bombieri} and \textit{D. Mumford} [Invent. Math. 35, 197--232 (1976; Zbl 0336.14010)].