Stable postulation and stable ideal generation: conjectures for fat points in the plane (Q1049902)

The authors are interested in studying the problems to computing \(h^0(X,\mathcal{O}_X(tF))\) when \(t>>0\), where \(F\) is a divisor on the blow up \(\pi:X \to \mathbb{P}^2\) at a finite set of distinct generic points \(P_1, P_2, \dots, P_n\) of \(\mathbb{P}^2\), and to determining the dimension of the cokernel of the map \(\mu_{tF}: H^0(X ,\mathcal{O}_X(tF)) \otimes H^0(X ,\mathcal{O}_X(L))\to H^0(X, \mathcal{O}_X(L+tF))\) for \(t>>0\). One of the main motivation for computing \(h^0(X,\mathcal{O}_X(F))\), for arbitrary \(F\) on \(X\), comes from fat points. If \(Z=m_1P_1 + \cdots + m_n\) is a fat point subscheme, with ideal \(I(Z)\), then, for each \(t\), we can associate to \(Z\) the divisor \(F_t(Z)=tL-m_1E_1-\cdots -m_nE_n\), where \(E_i=\pi^{-1}(P_i)\). Then it is well known that \(h^0(X,\mathcal{O}_X(F_T(Z)))\) is equal to the Hilbert function of \(X\) at level \(t\), i.e. \(h(t,I(Z))\). Moreover, the dimension of the cokernel \(\mu_{F_t(Z)}\) is the number of generators of degree \(t+1\) in any minimal set of homogeneous generators of \(I(Z)\). In fact, computing \(h^0(X,\mathcal{O}_X(F_T(Z)))\) and the dimension of the cokernel \(\mu_{F_t(Z)}\) for each \(t\) is equivalent to computing the graded Betti numbers of a graded minimal free resolution of \(I(Z)\) over the coordinate ring. Both problems are still open, whether approached from the point of view of fat points or from the point of view of complete linear systems on \(X\). In this paper the authors consider stable (i.e. asymptotic) versions of these problems. From the perspective of fat points, given any \(s\) and \(Z\) supported at the points \(P_i\), the stable version of the postulation problem is to find \(h(ts,I(tZ))\) for all \(t>>0\). The stable version of the ideal generation problem is to find the minimum number of homogeneous generators of \(I(tZ)\) in degree \(ts+1\) for all \(t>>0\). From the point of view of divisors on \(X\), the stable versions of the postulation and ideal generation problems, given an arbitrary \(F\), are to determine \(h^0(X \mathcal{O}_X(tF))\) and the dimension of the cokernel of \(\mu_{tF}\) for all \(t>>0\). The authors set these stable versions in a pure geometric way. In fact they show that the well known Segre-Harbourne-Gimigliano-Hirschowitz conjecture, which gives a complete conjectural solution to the postulational problem, implies that to solve the stable postulational problem it is enough to determine the integral curves \(C\) on \(X\) with \(C^2\leq 0\), and it implies that to solve the stable ideal generation problem it is enough to determine the dimension of the cokernel of \(\mu_F\), in the case that \(F=L+iE\) where \(E\) is a smooth rational curve with \(E^2=-1\) and where \(i=L \cdot E\). In particular the authors introduced explicit conjectures for the complete solution to both stable problems. More in detail consider divisor classes \(l=[L], e_1=[E1], \dots, e_n=[E_n]\) which give a free \(\mathbb{Z}-\)basis for Cl\((X)\) and consider the Weyl group \(W\) acting on Cl\((X)\), and generated by the operators \(s_x\) for \(x\in\{ r_0, \dots, r_{n-1}\}\), where \(s_x(F)=F+(x\cdot F)x\) for any \(F\in\)Cl\((X)\), with \(r_0=l-e_1-e_2-e_3\) and \(r_i=e_-e_{i+1}\). Let \(\Delta'\) be the submonoid of Cl\((X)\) of all classes \(F\) satisfying \(F\cdot r_i \geq 0\). It is clear that, for any class \(F\), either \(wF\in \Delta'\) for some \(w\in W\) or there is an element \(w \in W\) such that \(wF \cdot L < 0\). Since the points \(P_i\) are generic, one has \(h^0(X,F)=h^0(X,wF)\). Consider \(\Delta=\{F \in \Delta': F\cdot e_n \geq 0 \}\). It follow by a previous paper of the same authors that, for any given class \(F\) it is easy to determine if \(wF \in \Delta'\) for some \(w \in W\) and thence to find a class \(H\in \Delta\) corresponding to \(wF\). The question remains as to what is the value of \(h^0(X,F)\) and the main problems remain: given \(F\in \Delta\) and \(w \in W\), (a) determine \(h^0(X,F)\) and (b) determine the dimension of the cokernel \(\mu_{wF}: H^0(X,wF) \otimes H^0(X,L) \to H^0(X,L+wF)\). In terms of the stable versions the problems become: given \(F\in \Delta\) and \(w \in W\), for \(t>>0\), (a) determine \(h^0(X,tF)\) and (b) determine the dimension of the cokernel \(\mu_{wF}: H^0(X,twF) \otimes H^0(X,L) \to H^0(X,L+twF)\). The authors proposed the following interesting \vskip0.2cm Conjecture 1: Let \(F \in \Delta\). If \(h^0(X,tF)>0\) for some \(t>0\), then \(h^0(X,tF)=1+((tF)^2-tK_X\cdot F)/2\) for all \(t\) sufficiently large and either \(F^2>0\), or \(F^2=0\) and \(F\) is nonnegative multiple of either \(3l-e_1-\cdots - e_9\) or \(l-e_1\). \vskip0.2cm Moreover they proved that the Segre-Harbourne-Gimigliano-Hirschowitz conjecture implies Conjecture 1 and that Conjecture 1 is equivalent to the following \vskip0.2cm Conjecture 2: Let \(C\) be the class of a reduced irreducible divisor on \(X\). Then \(C^2\leq 0\) if and only if \(C\) is the class of an exceptional curve, or \(C=w(l-e_1)\) or \(C=w(3l-e_1-\cdots - e_9)\) for some \(w \in W\). \vskip0.2cm Finally, they show that Conjectures 2 and 3, if true, give a complete solution to part (b) of the stable version of the problem, where Conjecture 3 is contained in a previous paper [\textit{A. Gimigliano}, \textit{B. Harbourne} and \textit{M. Idà}, Trans. Am. Math. Soc. 361, No. 2, 1103--1127 (2009; Zbl 1170.14006)]: Conjecture 3: Let \(F=L+iE\), where \(E\) is an exceptional curve and \(0\leq i \leq L \cdot E\). Then dim\((cok \mu_F)={i-b_E \choose 2}+ {i-a_e \choose 2}\), where \((a_E,b_E)\) is the splitting type of \(E\).