The functor of a toric variety with enough invariant effective Cartier divisors (Q1127025): Difference between revisions

One can regard the projective space \({\mathbb{P}}^n\) as the variety representing the functor taking \(Y\) to surjections \({\mathcal O}_Y^{n+1}\to{\mathcal L}\), \({\mathcal L}\) a line bundle on \(Y\), since a map \(Y\to{\mathbb{P}}^n\) is given by \(n+1\) sections of a suitable \({\mathcal L}\) that do not all vanish simultaneously. In view of the very close analogy between toric varieties over a field~\(k\) and projective spaces, established by \textit{D. A. Cox} [J. Algebr. Geom. 4, No. 1, 17-50 (1995; Zbl 0846.14032)], it is reasonable to expect a good description of morphisms into toric varieties more generally. Several results of this nature are indeed known. Hitherto the most general has been one due also to \textit{D. A. Cox} [Tôhoku Math. J., II. Ser. 47, No. 2, 251-262 (1995; Zbl 0828.14035)], in which the toric variety \(X_\Delta\) is assumed to be smooth. Similar theorems, giving more directly applicable results but with strong conditions on~\(Y\) as well as~\(X_\Delta\), have been proved by \textit{M. A. Guest} [Bull. Am. Math. Soc., New Ser. 31, No. 2, 191-196 (1994; Zbl 0832.55009)] and by Tadao Oda and the reviewer. [This last result, unpublished at the time the present paper was written, has since appeared as \S 2 of the following: \textit{G. K. Sankaran}, Math. Ann. 313, No. 3, 409-427 (1999; Zbl 0919.14009){}. The author of the paper under review has pointed out an error in \S 1 of that paper, but \S 2 is unaffected.] In this context one should also mention \textit{M. Audin}, in whose book [``The topology of torus actions on symplectic manifolds'' (Birkhäuser 1991; Zbl 0726.57029)] on the topology of torus actions some of these results are foreshadowed. All these results are unified and generalised in the present paper. There is no hope of obtaining a good description on morphisms \(Y\to X_\Delta\) by means of invertible sheaves in all cases, because there are pathological toric varieties when for instance \(\text{Pic } X_\Delta\) is trivial. Hence the need to impose the condition ``enough effective Cartier divisors'', which means that the complement of every \(T\)-affine open subset is the support of an effective Cartier divisor. This condition is quite weak: All quasi-projective or simplicial toric varieties fulfil it. Given such an \(X_\Delta\), the author first studies the homogeneous coordinate ring \(S_\Delta\), which is the monoid ring on the monoid of effective Cartier divisors. He shows, following Cox's paper in J. Algebr. Geom. [loc.cit.], that \(X_\Delta\) is naturally isomorphic to a geometric quotient of an open subscheme of \(\mathop{\text{Spec}} S_\Delta\). The ring \(S_\Delta\) is \(\Delta\)-graded (i.e. has a \(\mathop{\text{Pic}}X_\Delta\)-grading) but, as the author points out, it is not possible to identify \(X_\Delta\) with the homogeneous spectrum of \(S_\Delta \) in general because \(\mathop{\text{Pic}} X_\Delta\) may have torsion. The author gives a necessary and sufficient condition for this torsion to vanish, incidentally correcting a minor error in Fulton's book on toric varieties [\textit{W. Fulton}, ``Introduction to toric varieties'' (Princeton 1993; Zbl 0813.14039)]. The next step is to establish a correspondence between \(\Delta\)-graded \(S_\Delta\)-modules and quasi-coherent \({\mathcal O}_\Delta\)-modules, largely a matter of verifying that proofs of similar results for smooth varieties still work. This is enough to prove the main general result: If \(Z\subset X_\Delta\) is a closed subscheme given by a homogeneous ideal \(I\) in \(S_\Delta\), and \(Y\) is a \(k\)-scheme, then \(k\)-morphisms \(Y\to Z\) correspond to \(\Delta\)-graded \({\mathcal O}_Y\)-algebra homomorphisms \(\phi:{\mathcal O}_Y\otimes_k S/I\to {\mathcal L}\) satisfying a nondegeneracy condition at each point of~\(Y\). Here \({\mathcal L}\) is a \(\Delta\)-graded \({\mathcal O}_Y\)-algebra with invertible components, so the product map \({\mathcal L}_\alpha\otimes{\mathcal L}_\beta\to{\mathcal L}_{\alpha+\beta}\) is an isomorphism for every \(\alpha,\beta\in\mathop{\text{Pic}}(X_\Delta)_{\geq 0}\). In the last section of the paper the author applies this result in the case \(Z=X_\Delta\) and deduces the previous results mentioned above, in most cases in a stronger form than previously. He also gives some examples.