5. Singularities in Mori theory [02EL]
Original official author HTML, exact retained edition. Historical TeX conversion verdicts remain unchanged. Cited-edition alignment and mathematical self-containment are not assessed.
Complete original source context · Original author HTML
5. Singularities in Mori theory
The singular locus of the normal complex space of pure dimension is a codimension analytic subvariety denoted by . Let and be the natural open immersion.
5.1. Log terminal singularities
Since this material may not be familiar to complex analysts or differential geometers, we briefly recall some basic facts on some of the singularities encountered in the Minimal Model Program (MMP for short). See [KM] for a detailled account in the algebraic case, the analytic theory being also surveyed there in less detail.
The sheaf of holomorphic functions is the subsheaf of the sheaf of continuous functions on consisting of the functions whose restriction to is holomorphic. Actually, by Hartogs’ theorem, any holomorphic function on extends to , which means that .
Every meromorphic n-form on extends to , i.e. let be a resolution of singularities of , then the meromorphic -form defined on extends to a meromorphic -form on . Let be the canonical sheaf of the smooth variety . The sheaf is a coherent analytic sheaf on .
More generally every meromorphic pluricanonical form on extends to and , is a coherent analytic sheaf on .
Definition 5.1.
Say is 1-Gorenstein iff one of the following equivalent conditions holds:
- (1)
Every has an open neighborhood such that carries an holomorphic -form with an empty zero divisor.
- (2)
is a rank one locally free sheaf.
- (3)
Every has an open neighborhood such that is isomorphic to .
A local section of defining an holomorphic -form without zeroes on will be called a local generator of . If furthermore is Cohen-Macaulay, is said to be Gorenstein.
Say is -Gorenstein iff one of the following equivalent conditions is satisfied:
- (1)
Every has an open neighborhood such that carries an holomorphic pluricanonical form with an empty zero divisor.
- (2)
For every , there exists and an open neighborhood of such that is a rank one locally free sheaf.
- (3)
For every there is and an open neighborhood of such that is isomorphic to .
A local section of defining an holomorphic pluricanonical form without zeroes on will be called a local generator of .
For every , the smallest fulfilling condition 3 near is called the local index of at . The l.c.m. of all local indices, if finite, is called the index of .
Definition 5.2.
Say has only canonical singularities iff is -Gorenstein, of finite index and one of the following equivalent conditions is fulfilled:
- (1)
Let be a resolution. Let be a local generator of . The meromorphic pluricanonical form is holomorphic.
- (2)
Let be a resolution. For every .
- (3)
(Assuming is an algebraic variety) Let be a resolution. Then with where means numerical equivalence of -Cartier divisors and the sum runs over the exceptional divisors of .
Observe that it is enough to check the first two conditions for some resolution. In the third condition is the order of vanishing of along the divisor .
Definition 5.3.
Say has only log-terminal singularities iff is -Gorenstein, of finite index and the following holds: let be a log-resolution and let be a local generator of : then the pole along any component of of the meromorphic -canonical form on is of order .
When is algebraic an equivalent formulation is: let be a log-resolution. Then with .
The importance of the class of canonical singularities comes from a theorem due to M. Reid [R 1] (see also [Deb], p. 174):
Theorem 5.4.
Let be a projective algebraic manifold of general type whose canonical ring is of finite type. Then the canonical model of , has only canonical singularities. If then is ample.
The finiteness of the canonical ring for varieties of general type is known in dimension 3 [Ka]. In higher dimension, Y.Kawamata has proved that it is a consequence of the existence of minimal models. is a uniquely defined singular birational model of . The minimal models of in the sense of the MMP are crepant terminalizations of and do not enjoy the above unicity since they may be related by non trivial flops.
Examples 5.5.
Let be a normal algebraic surface. The following are equivalent:
- (1)
has only canonical singularities.
- (2)
is locally analytically isomorphic to , a finite subgroup.
- (3)
The exceptional divisors of the minimal resolution of , have simple normal crossings, their components are (-2) smooth rational curves, their incidence graphs are of type A-D-E (Du Val singularities).
The log terminal surface singularities are precisely the singularities of the form , a finite subgroup.
Examples 5.6.
In higher dimension, quotient singularities are still log terminal. Fix and let be a smooth degree hypersurface. The affine cone over has only canonical singularities iff .
In particular, the ordinary double point has only canonical singularities but it is not a quotient singularity.
The hypersurface singularities of type are canonical.
5.2. Normal Kähler spaces
Plurisubharmonic functions
Let be a normal analytic space of pure dimension . A plurisubharmonic (psh) function on is an upper semicontinuous function on with values in , which is not locally , and extends to a psh function in some local embedding . The function is strongly psh (resp. , resp. ) iff it extends to a strongly psh function (resp. , resp. ) in some local embedding. A continuous function is psh iff its restriction to is so [FN]. A bounded psh function on extends to .
A pluriharmonic function on is a real valued continuous function on on such that one of the following equivalent conditions holds:
- •
is locally the real part of a holomorphic function.
- •
Given a local embedding , extends locally to a pluriharmonic function on .
- •
is pluriharmonic.
Semi-Kähler currents
Definition 5.7.
A semi-Kähler, resp. Kähler, resp. smooth Kähler, potential on is a family where is an open covering of and a psh function, resp. a strongly psh function, resp. a -smooth strongly psh function, on such that is pluriharmonic on .
Define an equivalence relation on semi-kähler potentials requiring that iff is pluriharmonic on .
Definition 5.8.
A smooth Kähler metric on is a -equivalence class of smooth Kähler potentials. A semi-Kähler (resp. Kähler) current on is a -equivalence class of semi-Kähler (resp. Kähler) potentials.
A semi-Kähler current is said to have (resp. , resp. Hölder continuous) potentials iff each is (resp. , resp. Hölder continuous).
We will on occasion drop the requirement that the local potentials of are psh, replacing it by the requirement that they are locally the sum of a smooth and a psh function. The current will then be called a quasi positive closed current on .
If it has locally bounded potentials, is fully determined by the closed form on defined on by .
Let be a smooth Kähler metric on with Kähler potential . An upper semi-continuous function is said to be -psh iff is psh on . The semi-Kähler current whose potential is is denoted by .
Example 5.9.
Let . Let be the usual affine coordinates on , those on . The formulas realize as the closed subscheme of whose equation is . We have two ‘natural’Kähler metrics on , the first one is smooth with potential , induced by the euclidean Kähler metric of , the second one is the Kähler current whose potential is . On it is the quotient of the euclidean metric restricted to . Near , .
The metric is an example of an orbifold Kähler metric on . The results of [Y] extend without major modifications to Kähler orbifolds. For instance, in each Kähler class of a nodal K3 surface there is a unique Ricci flat orbifold metric.
Chern-Weil forms and hermitian metrics
Let be the sheaf of real-valued pluriharmonic functions on . By definition, a closed (1,1)-form on is a section of the sheaf . We have the exact sequence:
A class in will be called Kähler, if it is in the image of a smooth Kähler metric.
Remark 5.10.
Assume is smooth. A class in will be called numerically base point free iff there exists a proper surjective holomorphic mapping , normal, such that is the pull back of a Kähler class on . This is a stronger condition than being semi-Kähler.
In the non-big case (i.e.: ), it is straightforward to construct semi-Kähler classes that are not numerically base point free (e.g. on complex tori). On the other hand, it is still unknown whether there exists a smooth projective variety and a semi-Kähler form which is big without being numerically base point free.
Let be a holomorphic line bundle on . The notion of smooth hermitian metric on is defined as in the smooth case. Let be such a metric on .
Let be a nowhere zero local holomorphic section of (a local generator of ) defined over the open subset . Set , where is a -smooth function on . The current is a smooth closed (1,1)-form on which does not depend on ; it is a semi-Kähler current if is psh.
More generally, let be an open covering of and a local generator of . Let . The datum defines a smooth closed (1,1)-form on .
Definition 5.11.
The Chern-Weil form of (or of ) is the -equivalence class of the data constructed above. We will denote it by .
It is immediate that is independent of . Hence there is a linear map . The connection with the more widely known smooth case is made by the observation that, if is a compact Kähler manifold, .
Proposition 5.12.
Let a compact normal complex analytic variety.
The space is finite dimensional.
Let a holomorphic line bundle on . Every representative of in is the Chern-Weil form of a smooth hermitian on .
If there exists a smooth hermitian metric such that is Kähler, then is projective-algebraic and is ample.
Proof.
The most difficult task is to show that, in the last assertion, is Moishezon. This follows from Siu’s solution of the Grauert-Riemenschneider conjecture [Siu]. ∎
A singular metric on is an expression , being a locally smooth + psh function and a smooth hermitian metric. Its Chern-Weil form is the quasi-positive current .