arXiv subjects
Publications and source records attributed to Frederic Campana.
An abelian fibration is a proper projective surjective map of complex varieties with general fiber an abelian variety. Consider a multiple fiber of an abelian fibration, and let $m_1, ..., m_k$ be the multiplicities of its irreducible components. We prove that the minimum of $m_i$ is equal to their greatest common divisor $gcd(m_1, ..., m_k)$
Contents 1. Algebraicity criterion: statement 2. Proof of the algebraicity criterion. 3. Pseudoeffectivity and movable classes. 4. Harder-Narasimhan filtrations and pseudo-effectivity. 5. Pseudo-effectivity of relative canonical bundles. 6. Rational curves and non-pseudoeffectivity of the canonical/cotangent bundles. 7. Birational stability of the cotangent bundle 8. Shafarevich-Viehweg conjecture. 9. Numerically trivial foliations. 10. The Beauville-Bogomolov-Yau decomposition 11. Some questions 11.1. Pseudoeffectivity of determinants. 11.2. Variations on Abundance. 11.3. Special manifolds. References
We show that any fibration of a 'special' compact K{ä}hler manifold X onto an Abelian variety has no multiple fibre in codimension one. This statement strengthens and extends previous results of Kawamata and Viehweg when $κ$(X) = 0. This also corrects the proof given in [2], 5.3 which was incomplete.
We give a simplified proof (in characteristic zero) of the decomposition theorem for complex projective varieties with klt singularities and numerically trivial canonical bundle. The proof rests in an essential way on most of the partial results of the previous proof obtained by many authors, but avoids those in positive characteristic by S. Druel. The single to some extent new contribution is an algebraicity and bimeromorphic splitting result for generically locally trivial fibrations with fibres without holomorphic vector fields. We give first the proof in the easier smooth case, following the same steps as in the general case, treated next.
We show the existence of metrically dense entire curves in rationally connected complex projective manifolds confirming for this case a conjecture according to which such entire curves on projective manifolds exist if and only if these are "special". We also show that such a dense entire curve may be chosen in such a way that it does not lift to any of its ramified covers, answering in this case a question of Corvaja and Zannier about the Nevanlinna analog of the `weak Hilbert property' of arithmetic geometry. We consider briefly the other test case of the conjecture, namely manifolds with $c_1=0$. Furthermore we discuss entire curves in normal rational surfaces avoiding the singular locus.
Small changes in sections 4 and 5, results not affected
We define, for smooth projective orbifold pairs $(X,D)$ notions of `slope Rational connectedness', and of orbifold `slope Rational quotient' . These notions extend to this larger context the classical notions of rationally connected manifold and `rational quotient' (sometimes called `MRC fibration'). Our notions and proofs work entirely in characteristic zero, and are based on the consideration of foliations with minimal positive slope with respect to some suitable movable class. The existence of covering or connecting families of `orbifold rational curves' is indeed presently unknown in the orbifold context, in situations analogous to the classical case $D=0$. By contrast, the notions we introduce here, are checkable in practice and can certainly be used to show general properties expected from the existence of connecting families of `orbifold rational curves'. The proofs given here in the orbifold context provide new proofs in the classical case where $D=0$, since the classical proofs did not seem to adapt, with the presently existing techniques, to this broader context.
We show that if $f:X\to B$ is a Lagrangian fibration from a compact connected Kähler hyperkähler manifold $X$ onto a projective normal variety $B$, then $f$ is locally projective. This answers a question raised by L. Kamenova and strengthens a former result (\cite{Ca05}, Proposition 2.1), according to which the smooth fibres of $f$ are projective. The proof below does not use the known additional results concerning fibres and base in this specific context).
In this article we consider log canonical pairs which are log-smooth. If the corresponding canonical bundle is pseudo-effective, then we show that any quotient of the orbifold cotangent bundle of the pair has a pseudo-effective determinant. One of the new ingredients in the proof is a generalization of the Bogomolov-McQuillan algebraicity criterion in the context of holomorphic foliations whose minimal slope with respect to a movable class is positive.
The notion of 'slope rational connectedness' is introduced in the context of smooth orbifold pairs. The main result parallels the characterization of the rational connectedness of projective manifolds in terms of either the non-existence of holomorphic covariant tensors, or of absence of fibrations onto manifolds with pseudo-effective canonical bundle, or of existence of movable classes for which the minimal slope of the tangent bundle is positive. We then use this result to construct the `rational quotient map' in orbifold category.
We prove two results about the natural representation of a group G of automorphisms of a normal projective threefold X on its second cohomology. We show that if X is minimal then G, modulo a normal subgroup of null entropy, is embedded as a Zariski-dense subset in a semi-simple real linear algebraic group of real rank < 3. Next, we show that X is a complex torus if the image of G is an almost abelian group of positive rank and the kernel is infinite, unless X is equivariantly non-trivially fibred.
The results known for Green-Lazarsfeld sets and solvable or nilpotent quotients of Kaehler groups are extended to the class of (compact Kaehler) geometric orbifolds with finite and integral multiplicities. The proofs are by reduction to the known case of compact Kaehler manifolds.
We define a birational version of the stability of cotangent sheaves for complex projective manifolds, and more generally for smooth orbifolds. We then show, using standard conjectures in birational classification, that these cotangent sheaves are birationally stable, unless the orbifold is uniruled.
We shall show how to decompose, by functorial and canonical fibrations, arbitrary $n$-dimensional complex projective {Although the geometric results apply to compact K\" ahler manifolds without change, we consider here for simplicity this special case only.} varieties $X$ into varieties (or rather ` geometric orbifolds\rq $)$ of one of the three \pure geometries determined by the `sign' (negative, zero, or positive) of the canonical bundle. These decompositions being birationally invariant, birational versions of these \pure geometries, based on the \canonical (or ` Kodaira\rq $)$ dimension will be considered, rather. A crucial feature of these decompositions is indeed that, in order to deal with multiple fibres of fibrations, they need to take place in the larger category of `geometric orbifolds' $(X| Δ)$. These are `virtual ramified covers' of varieties, which `virtually eliminate' multiple fibres of fibrations. Although formally the same as the `pairs' of the LMMP (see \cite{kmm}, \cite{KM}, \cite{BCHM} and the references there), they are here fully geometric objects equipped with the usual geometric invariants of varieties, such as sheaves of (symmetric) differential forms, fundamental group, Kobayashi pseudometric, integral points, morphisms and rational maps. It is intended to expose (with some addtional topics or developments), as briefly and simply as possible, and essentially skipping the proofs, the main content of arXiv:math/0110051 and arXiv:0705.0737 respectively published in ANN. Inst. Fourier (2004), and to appear in J.Inst. Math. Jussieu.
This is a sequel to [Ca01]=math.AG/0110051. We define the bimeromorphic {\it category} of geometric orbifolds. These interpolate between (compact K\" ahler) manifolds and such manifolds with logarithmic structure. These geometric orbifolds are considered from the point of view of their geometry, and thus equipped with the usual invariants of varieties: morphisms and bimeromorphic maps, differential forms, fundamental groups and universal covers, fields of definition and rational points. The most elementary properties, directly adapted from the case of varieties without orbifold structure, are established here. The arguments of [Ca01] can then be directly adapted to extend the main structure results to this orbifold category. We hope to come back to deeper aspects later. The motivation is that the natural frame for the theory of classification of compact K\" ahler (and complex projective) manifolds includes at least the category of orbifolds, as shown in [Ca01] by the fonctorial decomposition of {\it special} manifolds as tower of orbifolds with either $κ_+=-\infty$ or $κ=0$, and also, seemingly, by the minimal model program, in which most proofs work only after the adjunction of a "boundary". Also, fibrations enjoy in the bimeromorphic category of geometric orbifolds extension properties not satisfied in the category of varieties without orbifold structure, permitting to express invariants of the total space from those of the generic fibre and of the base. For example, the natural sequence of fundamental groups is exact there; also the total space is special if so are the generic fibre and the base. This makes this category suitable to lift properties from orbifolds having either $κ_+=-\infty$ or $κ=0$ to those which are special.
We study the algebraic dimension a(X) of a compact hyperkaehler manfold of dimension 2n. We show that a(X) is at most n unless X is projective. If a compact Kaehler manifold with algebraic dimension 0 and Kodaira dimension 0 has a minimal model, then only the values 0,n and 2n are possible. In case of middle dimension, the algebraic reduction is holomorphic Lagrangian. If n = 2, then - without any assumptions - the algebraic dimension only takes the values 0,2 and 4. The paper gives structure results for "generalised hyperkaehler" manifolds and studies nef lines bundles.
Consider a projective manifold X and suppose that some wedge power of the cotangent bundle contains a subsheaf whose determinant bundle has maximal Kodaira dimension. Then we prove that X is of general type. More generally we compute the Kodaira dimension if the determinant bundle has sufficiently large Kodaira dimension. This is based on the study of the determinant bundle of a quotient of the cotangent bundle of a non-uniruled manifold: this bundle is always pseudo-effective. We apply this to study the universal cover of a projective manifold. Finally we prove the following: if the canonical bundle is numerically equivalent to an effective Q-divisor, then the Kodaira dimension is non-negative.
We show a weak form of the function field version of Oesterle's abc conjecture. It asserts that, if $B$ is a complex projective connected curve, the number of intersection points, counted without multiplicities, of a fixed divisor $D$ of degree $d>0$ over $B$ with the graph $H$ of a section $h:B\to B\times \bP^1$ to the first projection is at least $(d-2)n-C(B,D)$, where $n$ is the degree of $H$ over $\bP^1$, and $C(D,B)$ a constant depending only on these two data. We show this number is at least $(d-2[\sqrt {d}]).n-C(D,B)$. The constant is ineffective.