Combinatorics and topology of proper toric maps (Q1992249)
From MaRDI portal
scientific article
Language | Label | Description | Also known as |
---|---|---|---|
English | Combinatorics and topology of proper toric maps |
scientific article |
Statements
Combinatorics and topology of proper toric maps (English)
0 references
5 November 2018
0 references
One of the earliest successes in the theory of toric varieties was the computation of the rational cohomology of a complete simplicial toric variety in terms of the $f$-vector of the corresponding polytope. More generally one can compute intersection cohomology combinatorially. This paper gives, among other things, a relative version of these results. The authors study the fibres of toric maps $f: X \to Y$, chiefly in the case where $X$ is simplicial, and prove a formula for the Betti numbers of the fibres in the case where $f$ is a fibration. \par They also prove a more complicated result, describing in combinatorial terms the decomposition theorem of Beilinson, Bernstein and Deligne in this toric context. They acknowledge some overlap in the combinatorial part with [\textit{E. Katz} and \textit{A. Stapledon}, Adv. Math. 286, 181--239 (2016; Zbl 1325.05192)], but it would probably be fair to say that the two papers complement each other, in a way that is explained here. \par The first step is to control the cohomology of $X$ in the case where $X$ admits a proper toric map to an affine variety having a torus-fixed point. In this case (Theorem A) the class map from $A_q(X)_{\mathbb{Q}}$ to (Borel-Moore) homology is an isomorphism and the mixed Hodge structure of cohomology (with or without compact supports) is of Hodge-Tate type (i.e.\ concentrated in $H^{p,p}$ of pure weight $q$. Hence (Theorem B) the same is true for the fibres of $f$. This implies a formula (Corollary C) for the Betti numbers: \[ \dim H^{2m}(f^{-1}(y),\mathbb{Q})= \sum_{l\ge m}(-1)^{l-m}\binom{l}{m}d_l(X/\tau), \] where $y$ belongs to the torus orbit $O(\tau)$ corresponding to the cone $\tau$ and $d_l(X/\tau)$ counts torus-invariant closed subsets mapping to $\overline{O(\tau)}$ of relative dimension $l$. \par Turning to the decomposition theorem, they obtain (for $f$ a toric fibration) a decomposition of $Rf_*(\text{IC}_X)$ as the direct sum of $(\text{IC}_{\overline{O(\tau)}})^{\oplus s_{\tau,b}}(-b)$ over all cones $\tau$ in the fan of $Y$ and all integers $b$. The integers $s_{\tau,b}$ satisfy various relations coming from parity (toric varieties have no odd cohomology), Poincaré duality and Hard Lefschetz, and have combinatorial interpretations. The main invariant is the sum $\delta_\tau=\sum_b s_{\tau,b}$. If $Y$ is also simplicial one has (Theorem E) $\delta_\tau=\sum_{\sigma\preceq \tau}(-1)^{\dim\tau-\dim\sigma}d_0(X/\sigma)$: a similar result, rather more complicated to state, holds for general $Y$. \par The proofs use a wide range of techniques from toric geometry and Hodge theory but the paper is written in such a way as to be more or less self-contained. Some of the proofs could, as the authors point out, have been shortened, at the cost of being less detailed, by an appeal to general results in Hodge theory, but much of the strength of toric geometry lies in the possibility of writing in a concrete way and they choose to do that.
0 references
toric geometry
0 references
intersection cohomology
0 references
decomposition theorem
0 references
$f$-vector
0 references