smooth locus:
(3.2) Au Al + A2, A4i D
, Ee, E7, E8.
We will abuse notation and call this the simply connected list.
We note that in general S is not determined by its singularities. However, for the simply
connected list, this is almost the case. See (3.10) below.
The /iT-negative MMP for Gorenstein surfaces is almost as simple as that for smooth surfaces.
3.3 Lemma. Let g : T W be a birational KT-contraction, of relative Picard number one,
with (irreducible) exceptional divisor E7 such that T is Gorenstein along E. Along E7 T is either
smooth, or has a unique singularity. In the latter case the singularity is an Ar singularity, with
i^T + E log terminal, and the singularity is removed by the contraction. Furthermore the induced

is an isomorphism.
Proof. On the minimal desingularisation, E is a 1-curve. Contractibility considerations give
the description of the possible singularities. The last claim follows from (7.3).
Gorenstein P r o r a t i o n s (of relative Picard number one) are also quite simple. Here we state
the possibilities. We will prove a more precise result in §11, see (11.5.4).
3.4 Lemma. Suppose n : T C is a
-fibration, of relative Picard number one, and G C T
is a fibre, contained in the Du Val locus ofT. One of the following holds:
(1) T is smooth along G.
(2) There are exactly two singularities, A\ points, along G. KT + G is It.
(3) There is a unique singularity, a Dn point along G.
Proof. Let T T be the minimal desingularisation, and let //, : T W be a relative minimal
model of T T, thus W C is smooth, and h is a composition of blow ups at smooth
points. Because TT has relative Picard number one, each blow up in h is at a point along the 1-
curve of the previous blow up. Now one checks easily that the proposition gives all possibilities
with Du Val singularities. For more details see (11.5).
3.5 Lemma. If T is a Gorenstein del Pezzo surface, then
Proof. Note if T' T is a degree n cover, etale in codimension one, then T' is again a
Gorenstein del Pezzo, and KT, nKT 8. Since KT is a positive integer, the result follows.
Recall that throughout the paper, a proper ^-negative curve on a normal surface T is called
a —1-curve, if C C T is a —1-curve in the usual sense. Note when T is Gorenstein, such a C is
a —1-curve iff C is rational, KT C 1 and C is smooth.
Previous Page Next Page