Numerical properties of exceptional divisors of birational morphisms of smooth surfaces (Q6161922)

In this nicely written paper, the authors provide a very detailed analysis of effective divisors whose support is contained in the exceptional locus of a birational morphism of smooth projective surfaces. In order to formulate main results (numerical descriptions), let us start with the following lemma. Lemma. Let \(p : T \rightarrow W\) be a birational morphism between smooth complex projective surfaces and one sets \(s:K_{W}^{2}-K_{T}^{2}\). Then there are effective non-zero divisors \(E_{1}, \dots, E_{s}\) of \(T\) such that \begin{itemize} \item[i)] \(K_{T} = p^{*}K_{W} + \sum_{i=1}^{s}E_{i}\) and \(p^{*}K_{W}\Gamma = 0\) for every component \(\Gamma\) of \(\sum_{i=1}^{s}E_{i}\); \item[ii)] the irreducible components of \(E_{i}\) are smooth rational curves and the dual graph of the support of \(E_{i}\) is a tree; \item[iii)] \(K_{T}E_{i}=-1\) for all \(i\) and \(E_{i}E_{j} = -\delta_{ij}\); \item[iv)] the intersection form on the components of \(\sum_{i=1}^{s}E_{i}\) is negative definite. \end{itemize} Recall that a curve \(D\) is \(m\)-connected if for every decomposition \(D=A+B\) as a sum of two curves \(A\) and \(B\) one has \(AB\geq m\). The first result, in the sense of a numerical description of exceptional divisors, is the following Proposition A. In the setting as in the above lemma, let \(\triangle\) be an effective divisor on \(T\) such that \(K_{T}\triangle = 0\) and \(\triangle^{2}=-2\). If \(\triangle\) is contracted by \(p\), then \begin{itemize} \item[i)] \(\triangle\) is \(1\)-connected; \item[ii)] if \(\triangle'\neq \triangle\) is also an effective divisor contracted by \(p\) and such that \(K_{T}\triangle'=0\), \((\triangle')^{2} = -2\), then \(-1 \leq \triangle \triangle'\leq 1\); \item[iii)] there are indices \(j\) and \(k\) such that \[\triangle E_{j}=1, \quad \triangle E_{k}=-1, \quad \triangle E_{i}=0 \quad \text{for} \quad i\neq j,k\] and \(E_{k} = E_{j} + \triangle\); \item[iv)] let \(E_{j}\) be such that \(\triangle E_{j} = 1\). If \(\triangle' \neq \triangle\) is also an effective divisor contracted by \(p\) and such that \(K_{T} \triangle ' =0\), \((\triangle ')^{2}=-2\) and \(\triangle' \triangle =0\), then \(\triangle' E_{j} = 0\). \end{itemize} Then the authors focus on \(ADE\)-configurations. Let us recall that an \(ADE\)-configuration of curves on a smooth complex projective surface \(T\) is a reduced effective divisor \(\triangle\) whose components are \((-2)\)-curves and whose dual graph is a Dynkin diagram of type \(A_{n}\), \(D_{n}\), or \(E_{m}\) with \(m \in \{6,7,8\}\). Observe that we have \(K_{T}\triangle =0\) and one can show that \(\triangle^{2}=-2\) by the induction argument with respect to the number \(n\) of components of \(\triangle\). Proposition B. Let \(p : T \rightarrow W\) be a birational morphism of smooth complex projective surfaces and \(\triangle, \triangle'\) be \(ADE\)-configurations of type \(A_{n}\) contracted by \(p\). Then \begin{itemize} \item[i)] if \(\triangle E_{j}=1\), then \(\triangle\) and \(E_{j}\) have no common components and so they intersect transversely in one point; \item[ii)] if \(\triangle E_{j}=1\), then the unique component \(\theta_{\triangle}\) of \(E_{j}\) such that \(\theta_{\triangle}\triangle =1\) is also a unique component of \(E_{j}\) such that \(\theta_{\triangle}E_{j}=-1\); \item[iii)] if \(\triangle E_{j}=1\) and \(\triangle \cap \triangle'=\emptyset\), then either \(\triangle '\) is disjoint from \(E_{j}\) or \(\triangle' < E_{j}\); \item[iv)] if \(\triangle \cap \triangle ' = \emptyset\), then \(\theta_{\triangle} \neq \theta_{\triangle'}\). \end{itemize} As a consequence we can observe that the number of \(k\)-disjoint irreducible \((-2)\)-curves contracted by \(p\) satisfies \(2k\leq s\), where \(s=K_{W}^{2}-K_{T}^{2}\) is the number of blowups from \(W\) to \(T\). The last part of the paper, the most interesting from my very subjective viewpoint, is devoted to surfaces with canonical singularities. Let \((Y,q)\) be a germ of canonical surface singularity and let \(\triangle_{q}\) be the exceptional divisor of the minimal resolution - \(\triangle_{q}\) is nothing but an \(ADE\)-configuration. Let \(G\) be the local fundamental group of \((Y,q)\), then we define the following numerical invariant \[\nu(q) := e(\triangle_{q}) = \frac{1}{|G|}.\] In that setting, the authors show the following interesting result. Theorem. Let \(X\) be a surface with only canonical singularities \(q_{1}, \dots, q_{h}\) and smooth elsewhere. Let \(f : T \rightarrow X\) be the minimal resolution and \(p : T \rightarrow W\) be the morphism to the minimal model. If \(\kappa(X) \geq 0\), then \[\sum_{i=1}^{h} \nu(q_{i}) \leq 12\chi(\mathcal{O}_{X}) - \frac{4}{3}K_{X}^{2} - \frac{s}{3},\] where \(s=K_{W}^{2}-K_{T}^{2}\) is the number of blowups from \(W\) to \(T\).