Séminaire de Géométrie complexe

Le problème de Kodaira demande si toute variété compacte kählérienne admet une déformation (arbitrairement petite) vers une variété projective. Nous présenterons des résultats positifs de ce problème pour certaines variétés de dimension 3 fibrées par des surfaces c_1-triviales. A un biméromorphisme près, ces variétés recouvrent les solides kählériens de dimension de Kodaira 1.

The quantum cohomology ring of a flag variety encodes the Gromov-Witten invariants that count curves meeting Schubert varieties in general position. When infinitely many such curves exist, the arithmetic genus of the corresponding family of curves is called a K-theoretic Gromov-Witten invariant. The quantum K-theory ring of
Givental is a generalization of the quantum cohomology ring that encodes the K-theoretic Gromov-Witten invariants. While little is known about the quantum K-theory of general flag varieties, the (small) quantum K-theory of cominuscule flag varieties has been studied in a series of papers with Chaput, Mihalcea, and Perrin. I will speak about geometric and combinatorial aspects of this work.

The « classical equals quantum » theorem states that any equivariant Gromov-Witten invariant (3 point, genus zero) of a Grassmann variety can be expressed as a triple intersection of Schubert classes on a two-step partial flag variety. An equivariant triple intersection on a two-step flag variety can in turn be expressed as a sum over puzzles that generalizes both Knutson and Tao’s puzzle rule for Grassmannians and the cohomological puzzle rule for two-step flag varieties. These results together give a Littlewood-Richardson rule for the equivariant quantum cohomology of Grassmannians. I will speak about geometric and combinatorial aspects of this story, which is based on papers with Kresch, Purbhoo, Mihalcea, and Tamvakis.

Gromov-Witten invariants are invariants of a smooth projective variety $X$ which count the number of curves of genus zero on $X$ meeting prescribed incidence conditions. The (small) quantum cohomology ring is a commutative graded $mathbb{Z}[q]$-algebra whose structure coefficients are given by three-point genus zero Gromov-Witten invariants. It is a deformation of the ordinary cohomology and depends on polynomial variables $q$ indexed by a basis of $H_2(X)$. In this talk, we will focus on the simplest case where $X=G/P$ is a homogeneous space. In this case, Fulton-Woodward gave a description of the minimal degrees $d$ such that $q^d$ occurs in the quantum product of two Schubert cycles. We will use this description and the theory of curve neighborhoods by Buch-Mihalcea to prove that there exists a unique minimal degree $d_X$ in the quantum product of two points. This degree $d_X$ can be completely understood in terms of Kostant’s cascade of strongly orthogonal roots. Moreover, it can be shown that any minimal degree in any quantum product of two Schubert cycles is bounded by $d_X$.

Une compactification lisse d’un quotient de domaine symétrique borné étant donnée, on souhaite étudier les notions de positivité usuelles de ses fibrés cotangents logarithmique et standard. Pour cela, on prouve un critère métrique de grosseur des fibrés cotangents, applicable à  toute paire logarithmique lisse. On peut ainsi montrer que le fibré cotangent logarithmique de la compactification précédente est toujours gros, ce qui redonne un résultat de Y. Brunebarbe.

Dans le cas d’un quotient de la boule, on s’intéresse aux revêtements ramifiés de la compactification, étales sur l’intérieur. On donne des ordres de ramification effectifs à  partir desquels toutes les sous-variétés d’un tel revêtement, non incluses dans le bord, ont leur fibré cotangent gros, ou nef.

The Littelmann paths model is a strong tool used to understand finite-dimensional representations of complex semi-simple Lie algebras. It has remarkable applications, such as a rule for the the decomposition into simple summands of the tensor product of two irreducible representations and of the restriction of a simple representation to a Levi sub algebra (those obtained by removing nodes from the Dynkin diagram). Such rules are called branching rules. We prove a conjecture of Naito-Sagaki about a branching rule for the restriction of irreducible representations of $mathfrak{sl}(2n,mathbb{C})$ to $mathfrak{sp}(2n,mathbb{C})$ in terms of Littelmann paths. The embedding is given by the folding of the type $A_{2n-1}$ Dynkin diagram, and is not of Levi type. So far, no non-Levi branching rules were known in terms of Littelmann paths. This is joint work with Bea Schumann.