Rational curves and ruled orders on surfaces (Q2344283)

On a smooth projective surface \(Z\), an \textit{order} is a torsion-free coherent sheaf of algebras \(A\) such that \(k(A)=A\otimes_Z k(Z)\) is a central simple \(k(Z)\)-algebra. Orders are subalgebras of a central simple algebra, and can be ordered by inclusion. The maximal orders correspond to normal surfaces, and orders are Azumaya on a dense open subset \(U\subseteq Z\), and the complement \(Z-U\) is the \textit{ramification locus}. This ramification locus is then a union of \textit{ramification curves}. A maximal order is called a \textit{terminal order} if the ramification locus is a normal crossing divisor, and if \(p\) is a node of the ramification locus lying in the intersection of ramification curves \(C_1,C_2\), then the ramification of \(A\) at \(C_i\) is itself totally ramified at \(p\) for at least one of the curves \(C_1\) or \(C_2\). In the commutative situation, a ruled surface has a corresponding Mori contraction. The noncommutative counterpart are the ruled orders: Start with a ruled surface \(f:Z\rightarrow C\) and let \(A\) be a terminal order on \(Z\) in a \(k(Z)\)-central division ring \(k(A)\). \(A\) is said to be a \textit{ruled order} if the ramificaton of \(A\) is the union of a bisection \(D\) of \(f\) with some fibres. Then the \textit{degree} \(e=\sqrt{\mathrm{rank}_ZA}\) of \(A\) is the ramification index of \(A\) on \(D\). A noncommutative scheme is a \(k\)-linear abelian category \(\mathcal C\) together with a distinguished object \(\mathcal O\) called the \textit{structure sheaf}. Given an order \(A\), a noncommutative scheme can be obtained by putting \(\mathcal C=A-\mathrm{Mod}\), the category of quasi-coherent \(A\)-modules. The authors change \(\mathcal O\) to a local pro-generator \(P\), which has the same effect as replacing the order \(A\) with the Morita equivalent \(A^\prime=\mathrm{End}_ZP\) and letting \(\mathcal O=A^\prime\). The authors use \(\mathrm{Spec}_ZA\) for the noncommutative scheme \((\mathcal C,\mathcal O)=(A-\mathrm{Mod},A)\). A ruled surface corresponds to the Mori contraction of a \(K\)-negative rational curve with self-intersection zero. For a terminal order \(A\) on a smooth projective surface \(Z\), a rational curve on \(\mathrm{Spec}_Z A\) is a cyclic (left) \(A\)-module \(M\) which is pure of dimension one with cohomology \(H^0(M)=k\) and \(H^1(M)=0\). The curve is \(K\)-non-effective if \(H^0(\omega_A\otimes_A M)=0\) where \(\omega_A\) is the \(A\)-bimodule \(\mathrm{Hom}_Z(A,\omega_Z)\), it is said to have self-intersection zero if the Euler form \(\chi(M,M)=\sum_{i=0}^2\dim_k\mathrm{Ext}^i_A(M,M)=-m^2=0.\) The component \(Y\) of the Hilbert scheme containing the point corresponding to a rational curve \(M\), with universal family \(\mathcal M\), is called the \textit{Hilbert scheme of deformations of} \(M\). The authors then prove that if \(\pi_Y:Y\times Z\rightarrow Y\) and \(\pi_Z:Y\times Z\rightarrow Z\) denote the projections maps, then the Fourier-Mukai type transform \(\pi_{Z\ast}(\mathcal M\otimes_{Y\times Z}\pi^\ast_Y(-)):QCoh(Y)\rightarrow A-\mathrm{Mod}\) is an exact functor with a right adjoint. Now, a morphism of noncommutative schemes is a pair of adjoint functors, so the Fourier-Mukai transform in ii) is the \textit{noncommutative Mori contraction contracting} \(M\). The article contains the proof that ruled orders give examples of non-commutative Mori contractions. To prove this, the following partial result is proved, and is very interesting: Let \(A\) be a ruled order on the ruled surface \(f:Z\rightarrow C\). Then the \(\mathcal O_C\)-algebra \(f_\ast A\) is commutative and is a sheaf of integral domains on \(C\). Orders on projective surfaces is an interesting research area. They provide examples of noncommutative surfaces which can be studied by the techniques of algebraic geometry. Their classification give a connection between the so called terminal orders without exceptional curves and del Pezzo orders, ruled orders, and minimal models. Del Pezzo orders have been studied in the generic case as they are examples of Sklyanin algebras, and some more special examples have been studied via the cyclic covering construction. The authors study orders \(A\) assuming it is terminal, a condition on ramification data that ensures that locally at any closed point \(A\) has dimension 2. If the ramification locus consists of a bisection \(D\) together with some fibres, \(A\) is called a ruled order. The main object of this article is the global structure of orders on projective surfaces. A motivating question is ``In what sense is a ruled order \(A\) ruled?'' This is because it often happens that equivalent concepts in commutative algebraic geometry are inequivalent in noncommuative algebraic geometry. In this article, a rational curve is a quotient \(A\rightarrow M\) with \(M\) an \(A\)-module supported on a smooth rational curve \(F\) such that as a sheaf on \(F\), \(M\) is locally free of rank \(e\) and \(h^0(M)=1,\;h^1(M)=0.\) In this article, it is proved that ruled orders have many rational curves, and their moduli is computed. The interest in these rational curves originally comes from a noncommuative Mori contraction theorem: Given a rational curve \(M\) as above, which is \(K\)-non-effective and has self-intersection zero. Then there is a morphism of noncommutative schemes \(\mathrm{Spec}_ZA\rightarrow Y\) where \(Y\) is the Hilbert scheme of deformations \(A\rightarrow M\), and \(Y\) is generically a smooth commutative curve. The authors consider the question whether or not a ruled order has a \(K\)-non-effective rational curve with self-intersection zero, and if so, what are the corresponding component of the Hilbert scheme and noncommutative Mori contraction. They also consider \(\overline A=A\otimes_Z\mathcal O_F\), look for a description of the fibres \(\overline A\) of \(A\), and look into the question if \(\overline A\) is rational in the sense that \(H^1(\overline A)=0\). if \(F\) is not a ramification curve, then \(\overline A\) is an order and embeds in a maximal order, which by Tsen's theorem has the form \(\mathrm{End}_F V\) for some vector bundle \(V\) on \(F\). Thus it is only needed to determine possibilities for \(V\) and sub-algebras of \(\mathrm{End}_F V\). It is challenging that the ramification data is Morita invariant, while the cohomology is not. The authors try to select the best model in the Morita equivalence class by asserting the minimality of the second Chern class \(c_2(A)\). A key tool is the Morita transforms to construct Morita equovalences, introduced by Artin and de Jong in the case of Azumaya algebras. The main results of the article is a description in matrix form of \(\overline A\) in the case when there exists a fibre \(F\subset Z\) which meets the ramification locus transversally in two distinct points. This in particular proves that \(H^1(\overline A)=0.\) Also, assume that every ramified fibre has ramification index equal to \(\deg A\). Then \(C^\prime=\mathrm{Spec}_X f_\ast A\) is a smooth projective curve, \(C^\prime\) is a component of the Hilbert scheme \(\mathrm{Hilb}A\), \(\Psi:A\otimes_Zf^\ast f_\ast A\rightarrow A\) is the universal rational curve, and the noncommutative Mori contraction is given by the algebra homomorphism \(f_\ast\rightarrow A\). The article is very interesting and contains a lot of good techniques and results.