Afﬁne cones over smooth cubic surfaces - arXiv · Afﬁne cones over smooth cubic surfaces 5 In...

arX

iv:1

303.

2648

v4 [

mat

h.A

G]

6 Ja

n 20

15

Affine cones over smooth cubic surfaces

Ivan Cheltsov Jihun Park Joonyeong Won

Abstract

We show that affine cones over smooth cubic surfaces do not admit non-trivialGa-actions.

Keywords. affine cone,α-invariant, anticanonical divisor, cylinder, del Pezzo sur-face,Ga-action, log canonical singularity.

Throughout this article, we assume that all considered varieties are algebraic and de-

fined over an algebraically closed field of characteristic0.

1 Introduction

One of the motivations for the present article originates from the articles of H. A. Schwartz

( [34]) and G. H. Halphen ( [17]) in the middle of 19th century,where they studied polyno-

mial solutions of Brieskorn-Pham polynomial equations in three variables after L. Euler

(1756), J. Liouville (1879) and so fourth ( [12]). Meanwhile, since the middle of 20th cen-

tury the study of rational singularities has witnessed great development ( [2], [5], [26]).

These two topics, one classic and the other modern, encounter each other in contemporary

mathematics. For instance, there is a strong connection between the existence of a rational

curve on a normal affine surface, i.e., a polynomial solutionto algebraic equations, and

rational singularities ( [15]).

As an additive analogue of toric geometry, unipotent group actions, speciallyGa-

actions, on varieties are very attractive objects to study.Indeed,Ga-actions have been

I. Cheltsov: School of Mathematics, The University of Edinburgh, James Clerk Maxwell Building, TheKing’s Buildings, Mayfield Road, Edinburgh EH9 3JZ, UK; e-mail: [email protected]

J. Park: Center for Geometry and Physics, Institute for Basic Science (IBS), 77 Cheongam-ro, Nam-gu,Pohang, Gyeongbuk, 790-784, Korea; Department of Mathematics, POSTECH, 77 Cheongam-ro, Nam-gu,Pohang, Gyeongbuk, 790-784, Korea; e-mail: [email protected]

J. Won: KIAS 85 Hoegiro Dongdaemun-gu, Seoul 130-722, Korea; e-mail: [email protected] Subject Classification (2010):Primary 14E15, 14J45, 14R20, 14R25; Secondary 14C20,

14E05, 14J17

1

http://arxiv.org/abs/1303.2648v4

2 Ivan Cheltsov, Jihun Park, Joonyeong Won

investigated for their own sake ( [3], [18], [29], [35], [40]). We also observe thatGa-

actions appear in the study of rational singularities. In particular, the article [15] shows

that a Brieskorn-Pham surface singularity is a cyclic quotient singularity if and only if the

surface admits a non-trivial regularGa-action. Considering its 3-dimensional analogue,

H. Flenner and M. Zaidenberg in 2003 proposed the following question ( [15, Ques-

tion 2.22]):

Does the affine Fermat cubic threefoldx3 + y3 + z3 + w3 = 0 in A4

admit a non-trivial regularGa-action?

Even though it is simple-looking, this problem stands open for 10 years. It turns out

that this problem is purely geometric and can be considered in a much wider setting

( [19], [20], [21], [22], [31]).

To see the problem from a wider view point, we letX be a smooth projective variety

with a polarisationH, whereH is an ample divisor onX. The generalized coneover

(X,H) is the affine variety defined by

X = Spec

(

⊕

n>0

H0 (X,OX (nH))

)

.

Remark 1.1. The affine varietyX is the usual cone overX embedded in a projective

space by the linear system|H| provided thatH is very ample and the image of the variety

X is projectively normal.

Let Sd be a smooth del Pezzo surface of degreed and letSd be the generalized cone

over(Sd,−KSd). For 3 6 d 6 9, the anticanonical divisor−KSd

is very ample and the

generalized coneSd is the affine cone inAd+1 over the smooth variety anticanonically

embedded inPd. In particular, ford = 3, the cubic surfaceS3 is defined by a cubic ho-

mogenous polynomial equationF (x, y, z, w) = 0 in P3, and hence the generalized cone

S3 is the affine hypersurface inA4 defined by the equationF (x, y, z, w) = 0. Ford = 2,

the generalized coneS2 is the affine cone inA4 over the smooth hypersurface in the

weighted projective spaceP(1, 1, 1, 2) defined by a quasi-homogeneous polynomial of

degree4. Ford = 1, the generalized coneS1 is the affine cone inA4 over the smooth hy-

persurface in the weighted projective spaceP(1, 1, 2, 3) defined by a quasi-homogeneous

polynomial of degree6 ( [16, Theorem 4.4]).

It is natural to ask whether the affine varietySd admits a non-trivialGa-action. The

problem at the beginning is just a special case of this.

T. Kishimoto, Yu. Prokhorov and M. Zaidenberg have been studying this generalized

problem and proved the following:

Theorem 1.2. If 4 6 d 6 9, then the generalized coneSd admits an effectiveGa-action.

Affine cones over smooth cubic surfaces 3

Proof. See [19, Theorem 3.19].

Theorem 1.3. If d 6 2, then the generalized coneSd does not admit a non-trivialGa-

action.

Proof. See [22, Theorem 1.1].

Their proofs make good use of a geometric property called cylindricity, which is

worthwhile to be studied for its own sake.

Definition 1.4 ( [19]). Let M be aQ-divisor on a smooth projective varietyX. An M-

polar cylinder inX is an open subset

U = X \ Supp(D)

defined by an effectiveQ-divisor D on X with D ∼Q M such thatU is isomorphic to

Z × A1 for some affine varietyZ.

They show that the existence of anH-polar cylinder onX is equivalent to the existence

of a non-trivialGa-action on the generalized cone over(X,H).

Lemma 1.5. LetH be an ample Cartier divisor on a smooth projective varietyX. Sup-

pose that the generalized coneX over (X,H) is normal. Then the generalized coneX

admits an effectiveGa-action if and only ifX contains anH-polar cylinder.

Proof. See [21, Corollary 3.2].

Remark 1.6. If X is a rational surface, then there always exists an ample Cartier divisor

H on X such thatX is normal andX contains anH-polar cylinder (see [19, Proposi-

tion 3.13]), which implies, in particular, thatX admits an effectiveGa-action.

Indeed, what T. Kishimoto, Yu. Prokhorov and M. Zaidenberg proved for their two

theorems is that the del Pezzo surfaceSd has a(−KSd)-polar cylinder if4 6 d 6 9 but

no (−KSd)-polar cylinder ifd 6 2.

The main result of the present article is

Theorem 1.7.A smooth cubic surfaceS3 in P3 does not contain any(−KS3)-polar cylin-

ders.

Together with Theorems 1.2 and 1.3, this makes us reach the following conclusion via

Lemma 1.5.

Corollary 1.8. Let Sd be a smooth del Pezzo surface of degreed. Thenthe generalized

cone over(Sd,−KSd) admits a non-trivial regularGa-action if and only ifd > 4.


In particular, we here present a long-expected answer to thequestion raised by H. Flen-

ner and M. Zaidenberg.

Corollary 1.9. The affine Fermat cubic threefoldx3 + y3 + z3 + w3 = 0 in A4 does not

admit a non-trivial regularGa-action.

The following lemma shows that having anticanonical cylinders on del Pezzo sur-

faces is strongly related to the log canonical thresholds oftheir effective anticanonical

Q-divisors3. It may also be one example that shows how important it is to study sin-

gularities of effective anticanonicalQ-divisors on Fano manifolds. Indeed, the proof of

Theorem 1.7 is substantially based on the lemma below.

Lemma 1.10. Let Sd be a smooth del Pezzo surface of degreed 6 4. Suppose thatSd

contains a(−KSd)-polar cylinder, i.e., there is an open affine subsetU ⊂ Sd and an

effective anticanonicalQ-divisorD such thatU = Sd \ Supp(D) andU ∼= Z × A1 for

some smooth rational affine curveZ. Then there exists a pointP onSd such that

• the log pair(Sd, D) is not log canonical at the pointP ;

• if there exists a unique divisorT in the anticanonical linear system|−KSd| such that

the log pair(Sd, T ) is not log canonical atP , then there is an effective anticanonical

Q-divisorD′ on the surfaceSd such that

– the log pair(Sd, D′) is not log canonical atP ;

– the support of the divisorT is not contained in the support ofD′.

Proof. This follows from [19, Lemma 4.11] and the proof of [19, Lemma4.14] (cf. the

proof of [22, Lemma 5.3]). Since the proof is presented dispersedly in [19] or [22], for the

convenience of the readers, we give a detailed and streamlined proof in Appendix A.

Applying Lemma 2.2below, we easily obtain

Corollary 1.11. LetS3 be a smooth del Pezzo surface of degree3. Suppose thatS3 con-

tains a(−KS3)-polar cylinder. Then there is an effective anticanonicalQ-divisorD on

S3 such that

• the log pair(S3, D) is not log canonical at some pointP onS3;

• the support ofD does not contain at least one irreducible component of the tangent

hyperplane sectionTP of S3 at the pointP .

3An anticanonicalQ-divisor on a varietyX is a Q-divisor Q-linearly equivalent to an anticanonicaldivisor ofX , meanwhile, an effective anticanonical divisor onX is a member of the anticanonical linearsystem| −KX |.


In order to prove Theorem 1.7, it suffices to show that there isno such a divisorD

described in Corollary 1.11 on a smooth del Pezzo surface of degree3. In this article, this

will be done in a bit wider setting. To be precise, we prove

Theorem 1.12. Let Sd be a smooth del Pezzo surface of degreed 6 3 and letD be

an effective anticanonicalQ-divisor onSd. Suppose that the log pair(Sd, D) is not log

canonical at a pointP . Then there exists a unique divisorT in the anticanonical linear

system|−KSd| such that the log pair(Sd, T ) is not log canonical at the pointP . Moreover,

the support ofD contains all the irreducible components ofSupp(T ).

Corollary 1.13. Let S3 be a smooth cubic surface inP3 and letD be an effective an-

ticanonical Q-divisor on S3. Suppose that the log pair(S3, D) is not log canonical

at a pointP . Then for the tangent hyperplane sectionTP at the pointP , the log pair

(S3, TP ) is not log canonical atP andSupp(D) contains all the irreducible components

of Supp(TP ).

Note that Corollary 1.13 contradicts the conclusion of Corollary 1.11. This simply

means that the hypothesis of Corollary 1.11 fails to be true.This shows that Theorem 1.12

implies Theorem 1.7. Moreover, we see that Theorem 1.12 recovers Theorem 1.3 through

Lemma 1.10 as well.

Remark 1.14. The conditiond 6 3 is crucial in Theorem 1.12. Indeed, ifd > 4, then

the assertion of Theorem 1.12 is no longer true (seethe proof of [19, Theorem 3.19]). For

example, consider the case whend = 4. There exists a birational morphismf : S4 → P2

such thatf is the blow up ofP2 at five points that lie on a unique irreducible conic. Denote

this conic byC. Let C be the proper transform of the conicC on the surfaceS4 and let

E1, . . . , E5 be the exceptional divisors of the morphismf . Put

D =3

2C +

5∑

i=1

1

2Ei.

It is an effective anticanonicalQ-divisor onS4 and the log pair(S4, D) is not log canonical

at any pointP on C. Moreover, for anyT ∈ | −KS4|, its support cannot be contained in

the support of the divisorD.

To our surprise, Theorem 1.12 has other applications that are interesting for their own

sake.

From here to the end of this section, letX be a projective variety with at worst Kawa-

mata log terminal singularities and letH be an ample divisor onX.

Definition 1.15. Theα-invariant of the log pair(X,H) is the number defined by

α (X,H) = sup

λ ∈ Q

∣

∣

∣

∣

∣

the log pair(X, λD) is log canonical for every

effectiveQ-divisorD onX with D ∼Q H.

.


The invariantα(X,H) has been studied intensively by manyauthorswho used dif-

ferent notations forα(X,H) ( [1], [6], [14], [4, § 3.4] [10, Definition 3.1.1], [11, Ap-

pendix A], [38, Appendix 2]). The notationα(X,H) is due to G. Tian who defined

α(X,H) in a different way ( [38, Appendix 2]). However, both the definitions coincide

by [11, Theorem A.3]. In the case whenX is a Fano variety, the invariantα(X,−KX)

is known as the famousα-invariant of Tian and it is denoted simply byα(X). Theα-

invariant of Tian plays a very important role in Kahler geometry due to the following.

Theorem 1.16( [13], [30], [36]). LetX be a Fano variety of dimensionn with at worst

quotient singularities. Ifα(X) > nn+1

, thenX admits an orbifold Kahler–Einstein metric.

The exact values of theα-invariants of smooth del Pezzo surfaces, as below, have been

obtained in [7, Theorem 1.7]. Those of del Pezzo surfaces defined over a field of positive

characteristic are presented in [28, Theorem 1.6] and thoseof del Pezzo surfaces with du

Val singularities in [8] and [33].

Theorem 1.17.LetSd be a smooth del Pezzo surface of degreed. Then

α(Sd) =

1/3 if d = 9, 7 or d = 8 andS8 = F1;

1/2 if d = 6, 5 or d = 8 andS8 = P1 × P1;

2/3 if d = 4;

α(S3) =

2/3 if S3 is a cubic surface in P3 with an Eckardt point;

3/4 if S3 is a cubic surface in P3 without Eckardt points;

α(S2) =

3/4 if | −KS2| has a tacnodal curve;

5/6 if | −KS2| has no tacnodal curves;

α(S1) =

5/6 if | −KS1| has a cuspidal curve;

1 if | −KS1| has no cuspidal curves.

Remark 1.18.Theorem 1.12 also provides the exact values of theα-invariants for smooth

del Pezzo surfaces of degrees6 3. We here show how to extract the values from Theo-

rem 1.12. Letν be the greatest number such that(Sd, νC) is log canonical for every

memberC in | − KSd|. The numberν can be easily obtained from [32, Section 3] and

checked to be the same as the value listed in Theorem 1.17 for theα-invariant ofSd. By

the definition ofν, there is an effective anticanonical divisorC on the surfaceSd such that

(Sd, νC) is log canonical but not Kawamata log terminal. This gives usα(Sd) 6 ν.

Suppose thatα(Sd) < ν. Then there are an effective anticanonicalQ-divisorD and a

positive rational numberλ < ν such that(Sd, λD) is not log canonical at some pointP

on Sd. Sinceλ < 1, the log pair(Sd, D) is not log canonical at the pointP either. By

Theorem 1.12, there exists a divisorT ∈ | −KSd| such that(Sd, T ) is not log canonical

atP . In addition,Supp(D) contains all the irreducible components ofSupp(T ).


The log pair(Sd, λT ) is log canonical sinceλ < ν. PutDǫ = (1+ ǫ)D− ǫT for every

non-negative rational numberǫ. ThenD0 = D andDǫ is effective for0 < ǫ ≪ 1 be-

causeSupp(D) contains all the irreducible components ofSupp(T ). Choose the greatest

ǫ such thatDǫ is still effective. ThenSupp(Dǫ) does not contain at least one irreducible

component ofSupp(T ).

Since(Sd, λT ) is log canonical atP and(Sd, λD) is not log canonical atP , the log pair

(Sd, λDǫ) is not log canonical atP (see Lemma 2.2). In particular, the log pair(Sd, Dǫ)

is not log canonical atP . However, this contradicts Theorem 1.12 sinceDǫ is an effective

anticanonicalQ-divisor. Therefore,α(Sd) = ν.

Corollary 1.19. LetSd be a smooth del Pezzo surface of degreed 6 3. If d = 3, suppose,

in addition, thatS3 does not contain an Eckardt point. ThenSd admits a Kahler–Einstein

metric.

The problem on the existence of Kahler–Einstein metrics onsmooth del Pezzo sur-

faces is completely solved by G. Tian and S.-T. Yau in [37] and[39]. In particular, Corol-

lary 1.19 follows from [37, Main Theorem].

The invariantα(X,H) has a global nature. It measures the singularities of effectiveQ-

divisors onX in a fixedQ-linear equivalence class. F. Ambro suggested in [1] a function

that encodes the local behavior ofα(X,H).

Definition 1.20 ( [1]). Theα-functionαHX of the log pair(X,H) is a function onX into

real numbers defined as follows: for a given pointP ∈ X,

αHX(P ) = sup

λ ∈ Q

∣

∣

∣

∣

∣

the log pair(X, λD) is log canonical atP for

every effectiveQ-divisorD onX with D ∼Q H.

.

Lemma 1.21.The identityα(X,H) = infP∈X αHX(P ) holds.

Proof. It is easy to check.

In the case whenX is a Fano variety, we denote theα-function of the log pair(X,−KX)

simply byαX .

Example 1.22.One can easily see thatαPn(P ) 61

n+1for every pointP on Pn. This

implies that theα-functionαPn is the constant function with the value1n+1

sinceα(Pn) =1

n+1.

Example 1.23. It is easy to seeαP1×P1(P ) 61

2for every pointP on P1 × P1. Since

α(P1 × P1) = 1

2by Theorem 1.17, theα-functionαP1×P1 is the constant function with

the value1

2by Lemma 1.21. Moreover, ifX is a Fano variety with at most Kawamata log

terminal singularities, then the proof of [11, Lemma 2.21] shows that

αX×P1(P ) = min

1

2, αX (pr1 (P ))


for every pointP onX×P1, wherepr1 : X×P1 → X is the projection on the first factor.

Usingthe same argument as thatin the proof of [11, Lemma 2.29], one can show that the

α-function of a product of Fano varieties with at most Gorenstein canonical singularities

is the point-wise minimum of the pull-backs of theα-functions on the factors.

As shown in Remark 1.18, the following can be obtained from Theorem 1.12 in a

similar manner.

Corollary 1.24. Let Sd be a smooth del Pezzo surface of degreed 6 3. Then theα-

function ofSd is as follows:

αS3(P ) =

2/3 if the pointP is an Eckardt point;

3/4 if the tangent hyperplane section atP has a tacnode atP ;

5/6 if the tangent hyperplane section atP has a cusp atP ;

1 otherwise;

αS2(P ) =

3/4 if there is an effective anticanonical divisor with a tacnode atP ;

5/6 if there is an effective anticanonical divisor with a cusp atP ;

1 otherwise;

αS1(P ) =

5/6 if there is an effective anticanonical divisor with a cusp atP ;

1 otherwise.

By Lemma 1.21, Corollary 1.24 implies that Theorem 1.17 holds for smooth del Pezzo

surfaces of degrees at most3. Thus, it is quite natural that we should extend Corollary 1.24

to all smooth del Pezzo surfaces in order to obtain a functional generalisation of Theo-

rem 1.17. This will be done in Section 6, where we prove

Theorem 1.25.LetSd be a smooth del Pezzo surface of degreed > 4. Then theα-function

of Sd is as follows:

αP2(P ) = 1/3;

αF1(P ) = 1/3; αP1×P1(P ) = 1/2;

αS7(P ) =

1/3if the pointP lies on the(−1)-curve that

intersects two other(−1)-curves;

1/2 otherwise;

αS6(P ) = 1/2;

αS5(P ) =

1/2 if there is(−1)-curve passing through the pointP ;

2/3 if there is no(−1)-curve passing thoughP ;

αS4(P ) =

2/3 if P is on a(−1)-curve;

3/4if there is an effective anticanonical divisor that

consists of two0-curves intersecting tangentially atP ;

5/6 otherwise.


The primary statement in this article is Theorem 1.12. As explained before, it imme-

diately implies the main result of the article, Theorem 1.7 and also recovers Theorem 1.3.

Theorem 1.12 will be proved in the following way.

In Section 2, we review the results that will be used in this article. As a warm-up, we

verify Theorem 1.12 for a smooth del Pezzo surface of degree1 (see Lemma 2.3). This is

very easy and instructive.

In Section 3, we establish two results aboutsingular del Pezzo surfaces of degree 2

that play a role in the proof of Theorems 1.12 for smooth cubicsurfaces. In addition, these

two results immediately yield Theorem 1.12 for a smooth del Pezzo surface of degree2

(see Lemma 3.4).

In Section 4, we prove Theorem 1.12 for a smooth cubic surface. This will be done

by a thorough case-by-case analysis of all possible types oftangent hyperplane sections

on a smooth cubic surface. Indeed, for a given pointP on the smooth cubic surface,

we show that every effective anticanonicalQ-divisor is log canonical atP if the tangent

hyperplane section atP is log canonical atP (Lemmas 4.7, 4.8 and 4.9), whereas we show

that its support contains the support of the tangent hyperplane section atP if an effective

anticanonicalQ divisor and the tangent hyperplane section atP are not log canonical atP

(see Lemmas 4.3, 4.5 and 4.6).

The proof of Lemma 4.8 deserves a separate section because itis the central and the

most beautiful part of the article and it is a bit lengthy. This will be presented in Section 5.

Appendix A will deal with Lemma 1.10 for the readers’ convenience.

2 Preliminaries

This section presents simple but essential tools for the article. Most of the described results

here are well-known and valid in much more general settings (cf. [23], [24] and [25]).

LetS be a projective surface with at most du Val singularities, let P be a smooth point

of the surfaceS and letD be an effectiveQ-divisor onS.

Lemma 2.1. If the log pair(S,D) is not log canonical at the smooth pointP , then

multP (D) > 1.

Proof. This is a well-known fact. See [25, Proposition 9.5.13], forinstance.

Write D =∑r

i=1aiDi, whereDi’s are distinct prime divisors on the surfaceS and

ai’s are positive rational numbers.

Lemma 2.2. LetT be an effectiveQ-divisor onS such that

• T ∼Q D butT 6= D;


• T =∑r

i=1biDi for some non-negative rational numbersb1, . . . , br.

For every non-negative rational numberǫ, putDǫ = (1 + ǫ)D − ǫT . Then

1. Dǫ ∼Q D for everyǫ > 0;

2. the setǫ ∈ Q>0 | Dǫ is effective attains the maximumµ;

3. the support of the divisorDµ does not contain at least one component ofSupp(T );

4. if (S, T ) is log canonical atP but (S,D) is not log canonical atP , then(S,Dµ) is

not log canonical atP .

Proof. The first assertion is obvious. For the rest we put

c = max

biai

∣

∣

∣i = 1, . . . , r

.

For some indexk we havec = bkak

.

Suppose thatc 6 1. Thenai > bi for everyi. This means that the divisorD − T =∑r

i=1(ai − bi)Di is effective. However, it is impossible sinceD − T is non-zero and

numerically trivial on a projective surface. Thus,c > 1, and hencebk > ak.

Putµ = 1

c−1. Thenµ = ak

bk−ak> 0 and

Dµ =bk

bk − akD −

akbk − ak

T =

r∑

i=1

bkai − akbibk − ak

Di,

wherebkai − akbi > 0 by the choice ofk. In particular, the divisorDµ is effective and

its support does not contain the curveDk. Moreover, for every positive rational numberǫ,

Dǫ =∑r

i=1(ai + ǫai − ǫbi)Di. If ǫ > µ, then

ǫ(bk − ak) > µ(bk − ak) =ak

bk − ak(bk − ak) = ak,

and henceDǫ is not effective. This proves the second and the third assertions.

If both (S, T ) and(S,Dµ) are log canonical atP , then(S,D) must be log canonical

atP becauseD = µ

1+µT + 1

1+µDµ and µ

1+µ+ 1

1+µ= 1.

Despite its naıve appearance, Lemma 2.2 is a very handy tool. To illustrate this, we

here verify Theorem 1.12 for a del Pezzo surface of degree1. This simple case also im-

mediately follows from the proof of [7, Lemma 3.1] or from theproof of [22, Proposi-

tion 5.1].

Lemma 2.3. Suppose thatS is a smooth del Pezzo surface of degree1 and D is an

effective anticanonicalQ-divisor onS. If the log pair(S,D) is not log canonical at the

pointP , then there exists a unique divisorT ∈ |−KS| such that(S, T ) is not log canonical

atP . Moreover, the support ofD contains all the irreducible components ofT .


Proof. Let T be a curve in| −KS| that passes through the pointP . Note thatT is irre-

ducible. If the log pair(S, T ) is log canonical atP , then it follows from Lemma 2.2 that

there exists an effective anticanonicalQ-divisorD′ on the surfaceS such that the log pair

(S,D′) is not log canonical atP andSupp(D′) does not contain the curveT . We then

obtain1 = T · D′ > multP (D′). This is impossible by Lemma 2.1. Thus, the log pair

(S, T ) is not log canonical atP .

Moreover,by Lemma 2.1the divisorT is singular at the pointP . Therefore, the point

P is not the base point of the pencil| −KS|. Consequently, such a divisorT is unique.

If the curveT is not contained inSupp(D), then we obtain an absurd inequality

1 = T · D > multP (D) > 1. Therefore, the curveT must be contained inSupp(D)

by Lemma 2.1.

The following is a ready-made Adjunction for our situation.

Lemma 2.4. Suppose that the log pair(S,D) is not log canonical at the smooth pointP .

If a componentDj with aj 6 1 is smooth atP , then

Dj ·

(

∑

i 6=j

aiDi

)

>∑

i 6=j

ai (Dj ·Di)P > 1,

where(Dj ·Di)P is the local intersection number ofCi andCj at P .

Proof. It immediately follows from [24, Theorem 5.50].

Let f : S → S be the blow up of the surfaceS at the pointP with the exceptional

divisorE and letD be the proper transform ofD by the blow upf . Then

KS + D + (multP (D)− 1)E = f ∗ (KS +D) .

The log pair(S,D) is log canonical atP if and only if the log pair(S, D+ (multP (D)−

1)E) is log canonical along the curveE.

Remark 2.5. If the log pair(S,D) is not log canonical atP , then there exists a pointQ

onE at which the log pair(S, D + (multP (D) − 1)E) is not log canonical. Lemma 2.1

then implies

multP (D) + multQ(D) > 2. (2.1)

If multP (D) 6 2, then the log pair(S, D+ (multP (D)− 1)E) is log canonical at every

point of the curveE other than the pointQ. Indeed, if the log pair(S, D + (multP (D)−

1)E) is not log canonical at another pointO onE, then Lemma 2.4 generates an absurd

inequality

2 > multP (D) = D · E > multQ(D) + multO(D) > 2.


Notation 2.6. From now on, when we have a birational morphism of a surface denoted by

a capital roman character with tilde onto a surface, in orderto denote the proper transform

of a divisor by this morphism, we will add tilde to the same character that denotes the

original divisor. For example, in the similar situation as the one preceding Remark 2.5,

we useD for the proper transform ofD by f without mentioning.

3 Del Pezzo surfaces of degree2

LetS be a del Pezzo surface of degree2 with at most two ordinary double points. Then the

linear system| −KS| is base-point-free and induces a double coverπ : S → P2 ramified

along a reduced quartic curveR ⊂ P2. Moreover, the curveR has at most two ordinary

double points. In particular, the quartic curveR is irreducible.

Lemma 3.1. For an effective anticanonicalQ-divisorD onS, the log pair(S,D) is log

canonical outside finitely many points onS.

Proof. Suppose the converse.Then we may writeD = a1C1 + Ω, whereC1 is an irre-

ducible reduced curve,a1 is a positive rational number strictly bigger than1 andΩ is an

effectiveQ-divisor whose support does not contain the curveC1. Since

2 = −KS ·D = −KS · (a1C1 + Ω) = −a1KS ·C1−KS ·Ω > −a1KS ·C1 > −KS ·C1,

we have−KS · C1 = 1. Thenπ(C1) is a line inP2. Thus, there exists an irreducible

reduced curveC2 onS such thatC1+C2 ∼ −KS andπ(C1) = π(C2). Note thatC1 = C2

if and only if the lineπ(C1) is an irreducible component of the branch curveR. SinceR

is irreducible, this is not the case. Thus, we haveC1 6= C2.

Note thatC21 = C2

2 becauseC1 andC2 are interchanged by the biregular involution of

S induced by the double coverπ. Thus, we have

2 = (−KS)2 = (C1 + C2)

2 = 2C2

1 + 2C · C2,

which implies thatC1 ·C2 = 1−C21 . SinceC1 andC2 are smooth rational curves, we can

easily obtainC21 = C2

2 = −1 + k2, wherek is the number of singular points ofS that lie

onC1.

Now we writeD = a1C1 + a2C2 +Γ, wherea2 is a non-negative rational number and

Γ is an effectiveQ-divisor whose support contains neitherC1 norC2. Then

1 = C1 · (a1C1 + a2C2 + Γ) = a1C2

1 + a2C1 · C2 + C1 · Γ

> a1C2

1 + a2C1 · C2 = a1C2

1 + a2(1− C2

1),

and hence1 > a1C21+a2(1−C2

1). Similarly, fromC2·D = 1, we obtain1 > a2C21+a1(1−

C21). The obtained two inequalities imply thata1 6 1 anda2 6 1 sinceC2

1 = −1 + k2,

k = 0, 1, 2. Sincea1 > 1 by our assumption,this is a contradiction.


The following two lemmas can be verified in a similar way as that of [7, Lemma 3.5].

Nevertheless we present their proofs since we should carefully deal with singular points

onS that have been considered neither in [7] nor in [22].

Lemma 3.2. For anyeffective anticanonicalQ-divisorD onS, the log pair(S,D) is log

canonical at every point outside the ramification divisor ofthe double coverπ.

Proof. Suppose that(S,D) is not log canonical at a pointP whose image byπ lies out-

sideR.

LetH be a general curve in| −KS| that passes through the pointP . Sinceπ(P ) 6∈ R,

the surfaceS is smooth at the pointP . Then

2 = H ·D > multP (H)multP (D) > multP (D),

and hencemultP (D) 6 2.

Let f : S → S be the blow up of the surfaceS atP . We have

KS + D + (multP (D)− 1)E = f ∗ (KS +D) ,

whereE is the exceptional curve of the blow upf . Then, Remark 2.5 gives a unique point

Q onE such that the log pair(S, D+(multP (D)− 1)E) is not log canonical atQ onE.

Sinceπ(P ) 6∈ R, there exists a unique reduced but possibly reducible curveC ∈

|−KS| such thatC passes throughP and its proper transformC passes through the point

Q. Note thatC is smooth atP . Since(S, C) is log canonical atP , Lemma 2.2 enables us

to assume that the support ofD does not contain at least one irreducible component ofC.

If the curveC is irreducible, then

2−multP (D) = 2−multP (C)multP (D) = C ·D > multQ(C)multQ(D) = multQ(D).

This contradicts (2.1). Thus, the curveC must be reducible.

We may then writeC = C1 +C2, whereC1 andC2 are irreducible smooth curves that

intersect at two points. Without loss of generality we may assume that the curveC1 is not

contained in the support ofD. The pointP must belong toC2: otherwise we would have

1 = D · C1 > multP (D) > 1.

We putD = aC2 + Ω, wherea is a non-negative rational number andΩ is an effective

Q-divisor whose support does not contain the curveC2. Then

1 = C1 ·D = (2−1

2k)a+ C1 · Ω > (2−

1

2k)a,

wherek is the number of singular points ofS on C1. On the other hand, the log pair

(S, aC2 + Ω + (multP (D) − 1)E) is not log canonical atQ, where we havea 6 1 by

Lemma 3.1. We then obtain

(2−1

2k)a = C2 · (Ω + (multP (D)− 1)E) > 1

from Lemma 2.4. This is a contradiction.


Lemma 3.3. For a smooth pointP of S with π(P ) ∈ R, let TP be the unique divisor

in | − KS| that is singular at the pointP . If the log pair(S, TP ) is log canonical atP ,

then foranyeffective anticanonicalQ-divisorD onS the log pair(S,D) is log canonical

atP .

Proof. Suppose that(S,D) is not log canonical at the pointP . Applying Lemma 2.2 to

the log pairs(S,D) and(S, TP ), we may assume thatSupp(D) does not contain at least

one irreducible component ofTP . Thus, if the divisorTP is irreducible, then Lemma 2.1

gives an absurd inequality

2 = TP ·D > multP (TP )multP (D) > 2multP (D) > 2

sinceTP is singular atP . Hence,TP must be reducible.

We may then writeTP = T1 + T2, whereT1 andT2 are smooth rational curves. Note

that the pointP is one of the intersection points ofT1 andT2. Without loss of generality,

we may assume that the curveT1 is not contained in the support ofD. Then

1 = T1 ·D > multP (T1)multP (D) = multP (D) > 1

by Lemma 2.1. The obtained contradiction completes the proof.

Lemmas 3.2 and 3.3 prove the following result.

Lemma 3.4. Suppose that the del Pezzo surfaceS of degree2 is smooth. LetD be an

effective anticanonicalQ-divisor onS. Suppose that the log pair(S,D) is not log canon-

ical at a pointP . Then there exists a unique divisorT ∈ | −KS| such that(S, T ) is not

log canonical atP . The support of the divisorD contains all the irreducible components

of T . The divisorT is either an irreducible rational curve with a cusp atP or a union of

two (−1)-curves meeting tangentially atP .

Proof. By Lemma 3.2, the pointπ(P ) must lie onR. Then there exists a unique curve

T ∈ | −KS| that is singular at the pointP . By Lemma 3.3, the log pair(S, T ) is not log

canonical atP .

Suppose that the support ofD does not contain an irreducible component ofT . Then

the proof of Lemma 3.3 works verbatim to derive a contradiction.

The last assertion immediately follows from [32, Proposition 3.2].

Lemma 3.4 shows that Theorem 1.12 holds for a smooth del Pezzosurface of degree2.

4 Cubic surfaces

In the present section we prove Theorem 1.12. Lemma 2.3 and Lemma 3.4 show that

Theorem 1.12 holds for del Pezzo surfaces of degrees1 and 2, respectively. Thus, to


complete the proof, we letS be a smooth cubic surface inP3 and letD be an effective

anticanonicalQ-divisor onS.

Lemma 4.1. The log pair(S,D) is log canonical outside finitely many points.

Proof. Suppose not. Then we may writeD = aC +Ω, whereC is an irreducible curve,a

is a positive rational number strictly bigger than1 andΩ is an effectiveQ-divisor whose

support does not contain the curveC. Then

3 = −KS · (aC + Ω) = −aKS · C −KS · Ω > −aKS · C > −KS · C.

This implies thatC is either a line or an irreducible conic.

Suppose thatC is a line. LetZ be a general irreducible conic onS such thatZ +C ∼

−KS. SinceZ is general, it is not contained in the support ofD. We then obtain

2 = Z ·D = Z · (aC + Ω) = 2a+ Z · Ω > 2a.

Thiscontradicts our assumption.

Suppose thatC is an irreducible conic. Then there exists a unique lineL on S such

thatL+ C ∼ −KS. WriteD = aC + bL+ Γ, whereb is a non-negative rational number

andΓ is an effectiveQ-divisor whose support contains neither the conicC nor the lineL.

Then

1 = L ·D = L · (aC + bL+ Γ) = 2a− b+ L · Γ > 2a− b.

On the other hand,

2 = C ·D = C · (aC + bL+ Γ) = 2b+ C · Γ > 2b.

Combining these two inequalities, we obtaina 6 1. This contradicts our assumption

too.

For a pointP on S, let TP be the tangent hyperplane section of the surfaceS at the

point P . This is the unique anticanonical divisor that is singular at P . The curveTP is

reduced but it may be reducible.

In order to prove Theorem 1.12 we must show that(S,D) is log canonical atP pro-

vided that one of the following two conditions is satisfied:

• the log pair(S, TP ) is log canonical atP ;

• the log pair(S, TP ) is not log canonical atP butSupp(D) does not contain at least

one irreducible component ofTP .


The log pair(S, TP ) is log canonical atP if and only if the pointP is an ordinary

double point ofTP (see [32, Proposition 3.2]). Thus,(S, TP ) is log canonical atP if and

only if TP is one of the following curves: an irreducible cubic curve with one ordinary

double point, a union of three coplanar lines that do not intersect at one point, a union of

a line and a conic that intersect transversally at two points.

Overall, we must consider the following cases:

(a) TP is a union of three lines that intersect atP (Eckardt point);

(b) TP is a union of a line and a conic that intersect tangentially atP ;

(c) TP is an irreducible cubic curve with a cusp atP ;

(d) TP is an irreducible cubic curve with one ordinary double point;

(e) TP is a union of three coplanar lines that do not intersect at onepoint;

(f) TP is a union of a line and a conic that intersect transversally at two points.

We consider these cases one by one in separate lemmas, i.e., Lemmas 4.3, 4.5, 4.6, 4.7,

4.8 and 4.9. We however present the detailed proof of Lemma 4.8 in Section 5 to improve

the readability of this section. These lemmas altogether imply Theorem 1.12.

For simplicity, putm = multP (D).

Lemma 4.2. If the log pair(S,D) is not log canonical at the pointP , then the support of

D contains all the lines onS passing throughP .

Proof. Let L be a line passing through the pointP that is not contained in the support of

D. Then the inequality1 = L ·D > m implies that the log pair(S,D) is log canonical at

P by Lemma 2.1.

Lemma 4.3( [19, Lemma 4.13]). Suppose that the tangent hyperplane sectionTP consists

of three lines intersecting at the pointP . If the support ofD does not contain at least one

of the three lines, then the log pair(S,D) is log canonical at the pointP .

Proof. It immediately follows from Lemma 4.2.

From now on, letf : S → S be the blow up of the cubic surfaceS at the pointP . In

addition, letE be the exceptional curve off . We then have

KS + D + (m− 1)E = f ∗ (KS +D) .

Note that the log pair(S,D) is log canonical atP if and only if the log pair

(S, D + (m− 1)E)

is log canonical along the exceptional divisorE.


Remark 4.4. If there is a line passing throughP , then the surfaceS is a weak del Pezzo

surface of degree2, i.e.,K2

S= 2 and−KS is nef and big. The proper transforms of the

lines passing thoughP will be (−2)-curves onS. All the (−2)-curves onS are disjoint

each other and they come from the lines passing throughP on S. By contracting these

(−2)-curves we obtain a birational morphismg : S → S. ThenS is a del Pezzo surface

of degree2 with ordinary double points. In particular, the linear system | −KS| induces

a double coverπ : S → P2 ramified along a quartic curveR ⊂ P2. The(−2)-curves onS

are contracted to the ordinary double points onS. Therefore, the number of the ordinary

double points onS is given by the number of lines passing throughP onS. Since we have

at most two lines passing thoughP , the surfaceS has at most two ordinary double points,

and hence the quartic curveR must be an irreducible curve with at most two ordinary

double points.

Lemma 4.5. Suppose that the tangent hyperplane sectionTP consists of a line and a

conic intersecting tangentially at the pointP . If the support ofD does not containboth

the line and the conic, then the log pair(S,D) is log canonical at the pointP .

Proof. Suppose that the log pair(S,D) is not log canonical at the pointP . LetL andC be

the line and the conic, respectively, such thatTP = L+C. By Lemma 4.2, we may assume

that the conicC is not contained but the lineL is contained in the support ofD. We write

D = aL+Ω, wherea is a positive rational number andΩ is an effectiveQ-divisor whose

support contains neither the lineL nor the conicC. We havem 6 C ·D = 2.

Note that the three curvesL, C andE meetat one point transversally. Sincem 6 2,

we have the unique pointQ onE defined in Remark 2.5. The pointQ does not belong to

C, and hence not toL either. Indeed,otherwise

2−m = C ·(

aL+ Ω)

> a +multQ(Ω) = multQ(D).

This contradicts (2.1).

Let g : S → S be the contraction defined in Remark 4.4. Note that the pointg(L) is

the ordinary double point of the surfaceS. Put Ω = g(Ω), E = g(E), C = g(C) and

Q = g(Q). Thenπ(E) = π(C) sinceE + C is an anticanonical divisor onS. The point

π(Q) lies outsideR because the pointQ lies outsideC. Since the divisorΩ +(

m− 1)

E

is Q-linearly equivalent to−KS by our construction, Lemma 3.2 shows that the log pair

(S, Ω + (m− 1)E) is log canonical atQ. However, it is not log canonical at the pointQ

sinceg is an isomorphism in a neighborhood of the pointQ. This is a contradiction.

Lemma 4.6.Suppose that the tangent hyperplane sectionTP is an irreducible cubic curve

with a cusp atP . If TP is not contained in the support ofD, then the log pair(S,D) is

log canonical atP .


Proof. Suppose that(S,D) is not log canonical atP . From the inequality

3 = TP ·D > m ·multP (TP ) = 2m,

we obtainm 63

2. Then, we have the unique pointQ onE defined in Remark 2.5.

The surfaceS is a smooth del Pezzo surface of degree2. The linear system| − KS|

induces a double coverπ : S → P2 ramified along a smooth quartic curveR ⊂ P2. Then

the integral divisorE + TP is linearly equivalent to−KS , and henceπ(E) = π(TP ) is a

line in P2. Moreover,TP tangentially meetE at a single point. Thus the pointπ(Q) lies

onR if and only if the pointQ is the intersection point ofE andTP .

Applying Lemma 3.2 to the log pair(S, D +(

m − 1)

E), we see that the pointπ(Q)

belongs toR because the log pair(S, D + (m− 1)E) is not log canonical at the pointQ

and the divisorD + (m − 1)E is Q-linearly equivalent to−KS. The pointQ therefore

lies on the curveTP . Then from (2.1) we obtain

3− 2m = TP · D > multQ(D) > 2−m.

This contradicts Lemma 2.1.

For the remaining three cases, we show that the hypothesis ofTheorem 1.12 is never

fulfilled, so that Theorem 1.12 is true.

Lemma 4.7. If the tangent hyperplane sectionTP is an irreducible cubic curve with a

node atP , then the log pair(S,D) is log canonical atP .

Proof. Suppose that(S,D) is not log canonical atP . The surfaceS is a smooth del Pezzo

surface of degree two. SinceD+(m−1)E ∼Q −KS and the log pair(S, D+(m−1)E)

is not log canonical at some pointQ on E, it follows from Lemma 3.4 that there must

be an anticanonical divisorH on the surfaceS that has either a tacnode or a cusp at the

pointQ.

If the divisorH has a tacnode atQ, then it consists of the exceptional divisorE and

another(−1)-curveL meetingE tangentially atQ. Then the divisorf(H) is an effective

anticanonical divisor onS such that it has a cusp atP and it is distinct from the divisor

TP . This is impossible.

If the divisorH has a cusp at the pointQ, then it must be irreducible. However, it is

impossible sinceH is singular atQ andE ·H = 1.

Lemma 4.8. Suppose that the tangent hyperplane sectionTP consists of three lines one

of which does not pass through the pointP . Then the log pair(S,D) is log canonical

atP .

Proof. The proof of this lemma is the central and the most beautiful part of the proof

of Theorem 1.12.Since it is a bit lengthy, it will be presented in a separate section. See

Section 5.


Lemma 4.9. Suppose that the tangent hyperplane sectionTP consists of a line and a

conic intersecting transversally. Then the log pair(S,D) is log canonical at the pointP .

Proof. We writeTP = L+C, whereL is a line andC is an irreducible conic that intersect

L transversally atP . Suppose that(S,D) is not log canonical atP .

By Lemmas 2.2 and 4.2, we may assume that the conicC is not contained but the line

L is contained in the support ofD. We writeD = aL+ Ω, wherea is a positive rational

number andΩ is an effectiveQ-divisor whose support contains neither the lineL nor the

conicC.

We have the unique pointQ onE defined in Remark 2.5 sincem 6 D · C = 2.

Suppose that the pointQ does not belong to the(−2)-curveL. Let g : S → S be the

contraction defined in Remark 4.4. ThenS is a del Pezzo surface of degree2 with only

one ordinary double point at the pointg(L). Put Ω = g(Ω), E = g(E), C = g(C) and

Q = g(Q). Thenπ(E) = π(C) sinceE + C is an anticanonical divisor onS. The point

π(Q) lies onR if and only if the pointQ lies onC. The log pair(S, Ω+ (m− 1)E) is not

log canonical atQ sinceg is an isomorphism in a neighborhood of the pointQ. Since the

divisor Ω +(

m− 1)

E is Q-linearly equivalent to−KS by our construction, Lemma 3.2

shows that the pointQ belongs toC.

Note thatC + E is the unique curve in| −KS| that is singular atQ. But the log pair

(S, C+E) is log canonical atQ. Hence, it follows from Lemma 3.3 that(S, Ω+(m−1)E)

is log canonical atQ. This is a contradiction. Therefore, the pointQ must belong to the

(−2)-curveL.

Now we can apply [8, Theorem 1.28] to the log pair(S, aL + (m − 1)E + Ω) at

the pointQ to obtain a contradiction immediately. Indeed, it is enoughto putM = 1,

A = 1, N = 0, B = 2, andα = β = 1 in [8, Theorem 1.28] and check that all the

conditions of [8, Theorem 1.28] are satisfied. However, there is a much simpler way to

obtain a contradiction. Let us take this simpler way.

There exists another lineM on the surfaceS that intersectsL at a point. The lineM

does not intersect the conicC since1 = TP ·M = (L + C) ·M = L ·M . In particular,

the pointP does not lie on the lineM . Let h : S → S be the contraction of the proper

transform of the lineM on the surfaceS. SinceM is a(−1)-curve and the pointP does

not lie onM , the surfaceS is a smooth cubic surface inP3.

Put Ω = h(Ω), E = h(E), L = h(L), C = h(C), P = h(Q) andD = h(D). Then

(S, D) is not log canonical at the pointP sinceh is an isomorphism in a neighborhood

of the pointQ. On the other hand, the divisorL + C + E is an anticanonical divisor of

the surfaceS. Since the pointP is the intersection point ofL andE and the divisorD is

Q-linearly equivalent to−KS, Lemma 4.8 implies that(S, D) is log canonical atP . This

is a contradiction.

As we already mentioned, Theorem 1.12 follows from Lemmas 4.3, 4.5, 4.6, 4.7, 4.8


and 4.9. Thus Theorem 1.12 has been proved under the assumption that Lemma 4.8 is

valid. The assumption will be confirmed in the following section.

5 The proof of Lemma 4.8

To prove Lemma 4.8, we keep the notations used in Section 4. WewriteTP = L+M+N ,

whereL, M , andN are three coplanar lines onS. We may assume that the pointP is the

intersection point of the two linesL andM , whereas it does not lie on the lineN . We

also writeD = a0L + b0M + c0N + Ω0, wherea0, b0, andc0 are non-negative rational

numbers andΩ0 is an effectiveQ-divisor onS whose support contains none of the lines

L, M andN . Putm0 = multP (Ω0).

Suppose that the log pair(S,D) is not log canonical at the pointP . Let us look for a

contradiction.

By Lemma 4.1, the log pair(S,D) is log canonical outside finitely many points. In

particular, we have0 6 a0, b0, c0 6 1. Also, Lemma 2.1 implies thatm0 + a0 + b0 > 1

and Lemma 4.2 implies thata0, b0 > 0.

Lemma 5.1. The inequalitym0 + a0 + b0 > c0 + 1 holds.

Proof. Since the log pair(S, a0L + b0M + Ω0) is not log canonicalat the pointP , it

follows from Lemma 2.4 that

1 + a0 − c0 = L · (D − a0L− c0N) = L · (b0M + Ω0) > 1,

which impliesa0 > c0. Similarly,b0 > c0.

The log pair(S, L + M + N) is log canonical. Since the log pair(S, a0L + b0M +

c0N +Ω0) is not log canonical atP , it follows from Lemma 2.2 and its proof that the log

pair(

S,1

1− c0D −

c01− c0

TP

)

is not log canonical atP . Then Lemma 2.1 showsthat

multP

(

1

1− c0D −

c01− c0

TP

)

= multP

(

a0 − c01− c0

L+b0 − c01− c0

M +1

1− c0Ω0

)

=a0 − c01− c0

+b0 − c01− c0

+m0

1− c0> 1.

Thisverifiesm0 + a0 + b0 > c0 + 1.

Since the rational numbersa0, b0, c0 aresmaller or equal to1 and the log pair(S, L+

M + N) is log canonical, the effectiveQ-divisorΩ0 cannot be the zero-divisor. Letr be


the number of the irreducible components of the support of the Q-divisor Ω0. Then we

write

Ω0 =

r∑

i=1

eiCi0,

whereei’s are positive rational numbers andCi0’s are irreducible reduced curves of de-

greesdi0 onS. We then see

3 = −KS ·

(

a0L+ b0M + c0N +r∑

i=1

eiCi0

)

= a0 + b0 + c0 +r∑

i=1

eidi0. (5.1)

We have

KS + a0L+ b0M + c0N + (a0 + b0 +m0 − 1)E +r∑

i=1

eiCi0 = f ∗ (KS +D) .

Recall thata0 + b0 +m0 = m.

Lemma 5.2. The inequalitym = a0 + b0 +m0 6 2 holds.

Proof. It immediately follows from the three inequalities

1 = L · (a0L+ b0M + c0N + Ω0) = −a0 + b0 + c0 + L · Ω0 > −a0 + b0 + c0 +m0,

1 = M · (a0L+ b0M + c0N + Ω0) = a0 − b0 + c0 +M · Ω0 > a0 − b0 + c0 +m0,

1 = N · (a0L+ b0M + c0N + Ω0) = a0 + b0 − c0 +N · Ω0 > a0 + b0 − c0.

The log pair(

S, a0L+ b0M + c0N + (a0 + b0 +m0 − 1)E +

r∑

i=1

eiCi0

)

(5.2)

is not log canonical at some pointQ on E. SincemultP (D) = a0 + b0 + m0 6 2, it

follows from Remark 2.5 thatQ is the only point onE where the log pair(5.2) fails to be

log canonical.

Let g : S → S be the contraction defined in Remark 4.4. ThenS is a del Pezzo surface

of degree2 with two ordinary double points at the pointsg(L) andg(M).

Lemma 5.3. The pointQ on the exceptional curveE belongs to eitherL or M .

Proof. Suppose that the pointQ lies on neitherL nor M . Put E = g(E), N = g(N)

andQ = g(Q). In addition, we putCi0 = g(Ci0) for eachi. Thenπ(E) = π(N). The

pointπ(Q) lies outside the quartic curveR sinceQ is a smooth point of the anticanonical

divisor E + N on S.


Sinceg is an isomorphism in a neighborhood of the pointQ, the log pair

(

S, c0N + (a0 + b0 +m0 − 1) E +

r∑

i=1

eiCi0

)

(5.3)

is not log canonical at the pointQ. The divisorc0N + (a0 + b0 +m0 − 1)E +∑r

i=1eiCi0

is an effective anticanonicalQ-divisor on the surfaceS. Hence, we are able to apply

Lemma 3.2 to the log pair(5.3) to obtain a contradiction.

From now on we may assume that the pointQ is the intersection point ofL andE

without loss of generality.

Let ρ : S 99K P2 be the linear projection from the pointP . Thenρ is a generically2-to-

1 rational map. Thus the mapρ induces a birational involutionτP of the cubic surfaceS.

The involutionτP is classically known as the Geiser involution associated tothe pointP

(see [27]).

Remark 5.4. By construction, the involutionτP is biregular outside the unionL∪M∪N .

In fact, one can show thatτP is biregular outside the pointP and the lineN . Moreover,

one can show thatτP (L) = L andτP (M) = M .

For eachi, putCi1 = τP (Ci0) and denote bydi1 the degree of the curveCi1. We then

employ new effectiveQ-divisors

Ω1 =

r∑

i=1

eiCi1;

D1 = a1L+ b1M + c1N + Ω1,

wherea1 = a0, b1 = b0 andc1 = a0 + b0 +m0 − 1. Note thata0 + b0 +m0 − 1 > 0 by

Lemma 2.1 (cf. Lemma 5.1).

Lemma 5.5. The divisorD1 is an effective anticanonicalQ-divisor on the surfaceS. The

log pair (S,D1) is not log canonical at the intersection point ofL andN .

Proof. Let h : S → S ′ be the contraction of the(−1)-curveN . ThenS ′ is a smooth cubic

surface inP3. PutE ′ = h(E), L′ = h(L), M ′ = h(M), Q′ = h(Q) andC ′i0 = h(Ci0)

for eachi. Then the integral divisorL′ + M ′ + E ′ is an anticanonical divisor ofS ′. In

particular, the curvesL′, M ′ andE ′ are coplanar lines onS ′. Moreover, the pointQ′ is

the intersection point ofL′ andE ′ by the assumption right after Lemma 5.3. It does not

lie on the lineM ′.

Let ιP be the biregular involution of the surfaceS induced by the double coverπ. Then

ιP induces a biregular involutionυP of the surfaceS since the surfaceS is the minimal


resolution of singularities of the surfaceS. Thus, we have a commutative diagram

S

f

g

%%

υP // S

f

g

xxrrrrrrrrrrrrr

S

ρ //

❲ ❩ ❪

S

π

ιP // S

π⑧⑧⑧⑧⑧⑧⑧⑧

S.

ρoo

s

♦

P2

This showsτP = f υP f−1. On the other hand, we haveυP (E) = N sinceπ g(E) =

πg(N). This means that there exists an isomorphismσ : S → S ′ that makes the diagram

S

h

υP // S

f

S ′ oo σS

commute. By construction,σ(L) = L′, σ(M) = M ′, σ(N) = E ′, andσ(Ci1) = C ′i0 for

everyi. Recall thatQ′ is the intersection point ofL′ andE ′.

Sinceh is an isomorphism locally aroundQ, the log pair

(

S ′, a0L′ + b0M

′ + (a0 + b0 +m0 − 1)E ′ +

r∑

i=1

eiC′i0

)

is not log canonical atQ′. Sincea0L+b0M+c0N+(a0 + b0 +m0 − 1)E+∑r

i=1eiCi0 ∼Q

−KS, we havea0L′+ b0M′+(a0 + b0 +m0 − 1)E ′+

∑r

i=1eiC

′i0 ∼Q −KS′ . Therefore,

it follows that

a0L+ b0M + (a0 + b0 +m0 − 1)N +r∑

i=1

eiCi1 ∼Q −KS,

and the log pair(S, a0L+ b0M +(a0+ b0+m0−1)N +∑r

i=1eiCi1) is not log canonical

at the intersection the point ofL andN .

Now we are able to replace the original effectiveQ-divisor D by the new effective

Q-divisorD1. By Lemma 5.5, both theQ-divisors have the same properties that we have

been using so far. However, the newQ-divisorΩ1 is slightly betterthan the original one

Ω0 in the sense of the following lemma.

Lemma 5.6. The degree of theQ-divisorΩ1 is strictly smaller than the degree ofΩ0, i.e.,

r∑

i=1

eidi1 <

r∑

i=1

eidi0.


Proof. SinceD1 ∼Q −KS by Lemma 5.5, we obtain

3 = −KS ·

(

a0L+ b0M + (a0 + b0 +m0 − 1)N +r∑

i=1

eiCi1

)

= 2a0 + 2b0 +m0 − 1 +

r∑

i=1

eidi1.

On the other hand, we havea0 + b0 + c0 +∑r

i=1eidi0 = 3 by (5.1). Thus, we obtain

r∑

i=1

eidi1 =r∑

i=1

eidi0 − (a0 + b0 +m0 − 1− c0) <r∑

i=1

eidi0

becausea0 + b0 +m0 − 1− c0 > 0 by Lemma 5.1.

Repeating this process, we can obtain a sequence of the effective anticanonicalQ-

divisors

Dk = akL+ bkM + ckN + Ωk

on the surfaceS such that each log pair(S,Dk) is not log canonical at one of the three

intersection pointsL ∩M , L ∩N andM ∩N . Note that

Ωk =

r∑

i=1

eiCik,

whereCik’s are irreducible reduced curves of degreesdik. We then obtain a strictly de-

creasing sequence of rational numbers

r∑

i=1

eidi0 >r∑

i=1

eidi1 > · · · >r∑

i=1

eidik > · · ·

by Lemma 5.6. This is a contradiction since the subset

r∑

i=1

eini

∣

∣

∣n1, n2, . . . , nr ∈ N

⊂ Q

is discrete and boundedfrom below.Thiscompletes the proof of Lemma 4.8.

6 α-functions on smooth del Pezzo surfaces

In this section, we prove Theorem 1.25. LetSd be a smooth del Pezzo surface of degreed.

Before we proceed, we here make a simple but useful observation.

Lemma 6.1. Let f : Sd → S be the blow down of a(−1)-curveE on the del Pezzo

surfaceSd. ThenS is a smooth del Pezzo surface andαSd(P ) > αS(f(P )) for a pointP

of Sd outside the curveE.


Proof. It is easy to check that−KS is ample. The second statement immediately follows

from the definition of theα-function.

We have already shownthat theα-functionαP2 of the projective plane is the constant

function with the value13

(see Example 1.22) and theα-functionαP1×P1 of the quadric

surface is the constant function with the value1

2(see Example 1.23).

Lemma 6.2. Theα-functionαF1on the blow-upF1 of P2 at one point is the constant

function with the value13.

Proof. Let P be a given point onF1. Let π : F1 → P1 be theP1-bundle morphism onto

P1. Let C be its section withC2 = −1 and letLP be the fiber of the morphismπ over

the pointπ(P ). Since2C + 3LP ∼ −KF1, we haveαF1

(P ) 61

3. But α(F1) = 1

3by

Theorem 1.17. Thus,αF1is the constant function with the value1

3by Lemma 1.21.

The surfaceS7 is the blow-up ofP2 at two distinct pointsQ1 andQ2. Let E be the

proper transform of the line passing throughQ1 andQ2 by the two-point blow upf :

S7 → P2 with the exceptional curvesE1 andE2.

Lemma 6.3. Theα-function on the del Pezzo surfaceS7 of degree7 has the following

values

αS7(P ) =

1/2 if P 6∈ E

1/3 if P ∈ E.

Proof. LetP be a point onS. ThenαS7(P ) > α(S) = 1

3by Theorem 1.17 and Lemma 1.21.

If the point P belongs toE, thenαS7(P ) 6

1

3since2E1 + 2E2 + 3E ∼ −KS.

Therefore,αS7(P ) = 1

3.

Suppose that the pointP lies outsideE. LetL be a line onP2 whose proper transform

by the blow upf passes throughP . Sincef ∗(2L)+E is an effective anticanonical divisor

passing throughP , we haveαS7(P ) 6 1

2.

Let g : S → P1 × P1 be the birational morphism obtained by contracting the(−1)-

curveE. Then this morphism is an isomorphism aroundP . ThenαS7(P ) > αP1×P1(g(P ))

by Lemma 6.1. SinceαP1×P1 is the constant function with the value12, we obtainαS7

(P ) =1

2.

Lemma 6.4. Theα-functionαS6on the del Pezzo surfaceS6 of degree6 is the constant

function with the value12.

Proof. Let P be a given point on the del Pezzo surfaceS6. One can easily checkthat

αS6(P ) 6 1

2. One the other hand, we have a birational morphismh : S6 → S7, whereS7

is a del Pezzo surface of degree7, such that the morphismh is an isomorphism around

the pointP and the pointh(P ) is not on the(−1)-curve ofS7 connected to two different

(−1)-curves. ThenαS6(P ) > 1

2by Lemmas 6.1 and 6.3.


Lemma 6.5. Theα-function on a del Pezzo surfaceS5 of degree5 has the following

values

αS5(P ) =

1/2 if there is a(−1)-curve passing throughP ;

2/3 if there is no(−1)-curve passing thoughP .

Proof. Let P be a point onS5. Suppose thatP lies on a(−1)-curve. Then there exists an

effective anticanonical divisor not reduced atP . Thus,αS5(P ) 6 1

2. Meanwhile, we have

1

2= α(S5) 6 αS5

(P ) by Lemma 1.21 and Theorem 1.17. Therefore,αS5(P ) = 1

2.

Suppose that the pointP is not contained in any(−1)-curve. Then there exist exactly

five irreducible smooth rational curvesC1, . . . , C5 passing through the pointP with−KS ·

Ci = 2 for eachi (cf. the proof of [7, Lemma 5.8]). Moreover, for everyCi, there are four

irreducible smooth rational curvesEi1, E

i2, E

i3 andEi

4 such that3Ci+Ei1+Ei

2+Ei3+Ei

4

belongs to the bi-anticanonical linear system| − 2KS5| (cf. Remark 1.14). Therefore,

αS5(P ) 6 2

3.

Suppose thatαS5(P ) < 2

3. Then there is an effective anticanonicalQ-divisorD such

that (S, λD) is not log canonical atP for some positive rational numberλ < 2

3. Then

multP (D) > 1

λby Lemma 2.1. Letf : S4 → S5 be the blow up ofS5 at P with the

exceptional curveE and letD be the proper transform of the divisorD onS4. Then the

surfaceS4 is a smooth del Pezzo surface of degree4. We have

KS4+ λD + (λmultP (D)− 1)E = f ∗ (KS5

+ λD) ,

which implies that the log pair(S4, λD + (λmultP (D)− 1)E) is not log canonical.

On the other hand, the log pair(S4, λD + λ(multP (D) − 1)E) is log canonical be-

cause the divisorD + (multP (D) − 1)E is an effective anticanonicalQ-divisor of S4

andα(S4) =2

3by Theorem 1.17. However,this is absurd becauseλ(multP (D) − 1) >

λmultP (D)− 1.

Lemma 6.6. Theα-function on a del Pezzo surfaceS4 of degree4 has the following

values

αS4(P ) =

2/3 if P is on a(−1)-curve;

3/4if there is an effective anticanonical divisor that consists of

two0-curvesmeetingtangentially atP ;

5/6 otherwise.

Proof. Let P be a point onS4. If the pointP lies on a(−1)-curveL, then there are

mutually disjoint five(−1)-curvesE1, . . . , E5 that intersectL. Let h : S4 → P2 be the

contraction of allEi’s. Sinceh(L) is a conic inP2, we see that3L+∑

16i65Ei is a member

in the linear system|−2KS4| (cf. Remark 1.14). This means thatαS4

(P ) 6 2

3. Therefore,

αS4(P ) = 2

3sinceα(S4) 6 αS4

(P ) by Lemma 1.21 andα(S4) =2

3by Theorem 1.17.


Suppose that the pointP does not lie on a(−1)-curve. Putω = 3

4in the case when

there is an effective anticanonical divisor that consists of two 0-curves intersecting tan-

gentially at the pointP and putω = 5

6otherwise.

One can easily find an effective anticanonical divisorF onS4 such that(S4, λF ) is not

log canonical atP for every positive rational numberλ > ω (see [32, Proposition 3.2]).

This shows thatαS4(P ) 6 ω. Moreover, it is easy to check that the log pair(S4, ωC) is

log canonical atP for eachC ∈ | −KS4|.

SupposeαS4(P ) < ω. Then there is an effective anticanonicalQ-divisorD such that

(S, ωD) is not log canonical atP . Note that there are only finitely many effective anti-

canonical divisorsC1, . . . , Ck such that each(S4, Ci) is not log canonical atP . Apply-

ing Lemma 2.2, we may assume that for eachi at least one irreducible component of

Supp(Ci) is not contained in the support ofD.

Let f : S3 → S4 be the blow up of the surfaceS4 at P with the exceptional curve

E and letD be the proper transform of the divisorD on S3. ThenS3 is a smooth cubic

surface inP3 andE is a line inS3. Moreover, the log pair(S3, D+ (multP (D)− 1)E) is

not log canonical at some pointQ onE because the log pair(S4, D) is not log canonical

atP .

Let TQ be the tangent hyperplane section of the cubic surfaceS3 atQ. Note that the

divisor TQ contains the lineE. SinceD + (multP (D) − 1)E is an effective anticanon-

ical Q-divisor onS3, it follows from Corollary 1.13 that the log pair(S3, TQ) is not log

canonical atQ and the support ofD contains all the irreducible components ofTQ. In

fact, it follows that the divisorTQ is either a union of three lines meeting atQ or a union

of a line and a conic intersecting tangentially atQ. The divisorf(TQ) is an effective anti-

canonical divisor onS4 such that the log pair(S4, f(TQ)) is not log canonical atP . This

contradicts our assumption since the support ofD contains all the irreducible components

of the divisorf(TQ).

Consequently, Theorem 1.25 follows from Examples 1.22 and 1.23, and Lemmas 6.2,

6.3, 6.4, 6.5 and 6.6.

A Appendix

This appendix is devoted to the proof of Lemma 1.10. The prooforiginates from [19]

and [22], where the proof is presented dispersedly. For the readers’ convenience, we give

a detailed and streamlined one here.

Let S be a smooth del Pezzo surface of degree at most4. Suppose thatS contains a

(−KS)-polar cylinder, i.e., there is an open affine subsetU ⊂ S and an effective anti-

canonicalQ-divisorD such thatU = S \ Supp(D) andU ∼= Z × A1 for some smooth


rational affine curveZ. PutD =∑r

i=1aiDi, where eachDi is an irreducible reduced

curve and eachai is a positive rational number.

Lemma A.1 ( [22, Lemma 4.4]). The number of the irreducible components of the divisor

D is not smaller than the rank of the Picard group ofS, i.e.,r > rkPic(S) = 10−K2S >

6.

To prove Lemma 1.10, we must show that there exists a pointP ∈ S such that

• the log pair(S,D) is not log canonical at the pointP ;

• if there exists auniquedivisorT in the anticanonical linear system|−KS| such that

the log pair(S, T ) is not log canonical atP , then there is an effective anticanonical

Q-divisorD′ on the surfaceS such that

– the log pair(S,D′) is not log canonical atP ;

– the support of the divisorT is not contained in the support ofD′.

The natural projectionU ∼= Z × A1 → Z induces a rational mapπ : S 99K P1 given

by a pencilL on the surfaceS. Then eitherL is base-point-free or its base locus consists

of a single point.

Lemma A.2 ( [22, Lemma 4.2]). The pencilL is not base-point-free.

Proof. Suppose that the pencilL is base-point-free. Thenπ is a morphism, which implies

that there exists exactly one irreducible component ofSupp(D) that does not lie in a fiber

of π. Moreover, this component is a section. Without loss of generality, we may assume

that this component isDr. LetL be a sufficiently general curve inL. Then

2 = −KS · L = D · L =

r∑

i=1

aiDi · L = arDr · L,

and hencear = 2. This implies thatα(S) 6 1

2. However,this contradicts Theorem 1.17

since the degree of the surfaceS is at most4.

Denote the unique base point of the pencilL by P . Let us show thatP is the point

we are looking for. Resolving the base locus of the pencilL, we obtain a commutative

diagram

Wf

⑧⑧⑧⑧⑧⑧⑧⑧ g

Sπ //❴❴❴❴❴❴❴ P1,

wheref is a composition of blow ups at smooth points overP andg is a morphism whose

general fiber is a smooth rational curve. Denote byE1, . . . , En the exceptional curves of


the birational morphismf . Then there exists exactly one curve among them that does not

lie in the fibers of the morphismg. Without loss of generality, we may assume that this

curve isEn. ThenEn is a section of the morphismg.

For everyDi, denote byDi its proper transform on the surfaceW . Then every curve

Di lies in a fiber of the morphismg.

The following lemma is a bit stronger version of [22, Lemma 4.6] even though its

proof is almost the same as that of [22, Lemma 4.6].

Lemma A.3 (cf. [22, Lemma 4.6]). For every effective anticanonicalQ-divisorH with

Supp(H) ⊆ Supp(D), the log pair(S,H) is not log canonical at the pointP .

Proof. The proof of [22, Lemma 4.6] works verbatim for this generalized version.

Applying Lemma A.3 to(S,D), we see that the log pair(S,D) is not log canonical

atP . Thus, if there exists no anticanonical divisorT such that(S, T ) is not log canonical

at P , then we are done. Hence, to complete the proof of Lemma 1.10,we assume that

there exists auniquedivisorT ∈ | −KS| such that(S, T ) is not log canonical atP . Then

Lemma 1.10 follows from the lemma below.

Lemma A.4. There exists an effectiveanticanonicalQ-divisor D′ on S such that the

log pair (S,D′) is not log canonical atP and Supp(D′) does not contain at least one

irreducible component ofSupp(T ).

Proof. If Supp(D) does not contain at least one irreducible component ofSupp(T ), then

we can simply putD = D′. Suppose that it is not the case, i.e., we haveSupp(T ) ⊆

Supp(D). ThenT 6= D. Indeed, the number of the irreducible components ofSupp(D)

is at least6 by Lemma A.1. On the other hand, the number of the irreduciblecomponents

of Supp(T ) is at most4 because−KS · T = K2S and−KS is ample.

SinceT 6= D, there exists a positive rational numberµ such that theQ-divisor (1 +

µ)D−µT is effective and its support does not contain at least one irreducible component

of Supp(T ). PutD′ = (1 + µ)D − µT . Note thatD′ is also an effective anticanonical

Q-divisor onS. By our construction,Supp(D′) ⊆ Supp(D). Thus, the log pair(S,D′) is

not log canonical atP by Lemma A.3. This completes the proof.

Remark A.5. Note thatU 6= S \ Supp(D′), which implies that the number of the ir-

reducible components ofSupp(D′) may be less thanrkPic(S). Because of this, we can

apply Lemma 2.2 only once here. This shows that we really needto use theuniqueness

of the divisorT in the anticanonical linear system| − KS| such that(S, T ) is not log

canonical atP in the proof of Lemma A.4. Indeed, if there is another divisorT ′ in |−KS|

such that(S, T ′) is not log canonical atP , then we would not be able to apply Lemma 2.2

since wemighthaveD′ = T ′.


Acknowledgments.The authors would like to express their thanks to the refereefor the careful reading andthe comments that improve this article. The first author was supported by AG Laboratory NRI-HSE, RFgovernment grant, 11.G34.31.0023. The second author was supported by the Institute for Basic Science(Grant No. CA1305-02).

References

[1] Ambro, F.: The minimal log discrepancy. In: Proceedingsof the workshop, Arc

spaces and multiplier ideal sheaves (Ed. Kei-ichi Watanabe), RIMS kokyuroku, No.

1550, 121–130 (2006)

[2] Artin, M.: On isolated rational singularities of surfaces. Amer. J. Math.88, 129–136

(1966)

[3] Bass, H.: A nontriangular action ofGa onA3. J. Pure Appl. Algebra33, 1–5 (1984)

[4] Berman, R.: A thermodynamical formalism for Monge–Amp`ere equations, Moser–

Trudinger inequalities and Kahler–Einstein metrics. Adv. Math. 248, 1254–1297

(2013)

[5] Brieskorn, E.: Rationale Singularitaten komplexer Flachen. Invent. Math.4, 336–

358 (1968)

[6] Cheltsov, I.: Log canonical thresholds on hypersurfaces. Sb. Math.192, 1241–1257

(2001)

[7] Cheltsov, I.: Log canonical thresholds of del Pezzo surfaces. Geom. Funct. Anal.11,

1118–1144 (2008)

[8] Cheltsov, I., Kosta, D.: Computingα-invariants of singular del Pezzo surfaces. To

appear in J. Geom. Anal. DOI: 10.1007/s12220-012-9357-6.

[9] Cheltsov, I., Park, J.: Global log-canonical thresholds and generalized Eckardt

points. Sb. Math.193, 779–789 (2002)

[10] Cheltsov, I., Park, J., Won, J.: Log canonical thresholds of certain Fano hypersur-

faces. To appear in Math. Z. DOI: 10.1007/s00209-013-1186-1.

[11] Cheltsov, I., Shramov, C.: Log canonical thresholds ofsmooth Fano threefolds (with

an appendix by Jean-Pierre Demailly). Russian Math. Surveys63, 73–180 (2008)

[12] Darmon, H., Granville, A.: On the equationszm = F (x, y) andAxp + Byq = Czr.

Bull. London Math. Soc.27, no. 6, 513–543 (1995)

[13] Demailly, J.-P., Kollar, J.: Semi-continuity of complex singularity exponents and

Kahler-Einstein metrics on Fano orbifolds. Ann. Sci.Ecole Norm. Sup.34, 525–556

(2001)


[14] de Fernex, T., Ein, L., Mustata, M.: Bounds for log canonical thresholds with appli-

cations to birational rigidity. Math. Res. Letters10, 219–236 (2003)

[15] Flenner, H., Zaidenberg, M.: Rational curves and rational singularities. Math. Z.

244, 549–575 (2003)

[16] Hidaka, F., Watanabe, K.: Normal Gorenstein surfaces with ample anti-canonical

divisor. Tokyo J. Math.4, no. 2, 319–330 (1981)

[17] Halphen, G. H.: Sur la reduction des equation differentielles lineaires aux formes

integrales. In: Memoires presentes par divers savantsa lAcademie des sciences de

lInstitut National de France, T. XXVIII, N 1 (Paris, F. Krantz,1883), Oeuvres.Vol.3, Paris, 1–260 (1921)

[18] Kaliman, S.: FreeC+-actions onC3 are translations. Invent. Math.156, 163–173

(2004)

[19] Kishimoto, T., Prokhorov, Yu., Zaidenberg, M.: Group actions on affine cones. In:

Affine algebraic geometry, CRM Proc. Lecture Notes54, Amer. Math. Soc., 123–

163 (2011)

[20] Kishimoto, T., Prokhorov, Yu., Zaidenberg, M.: Affine cones over Fano threefolds

and additive group actions. To appear in Osaka J. Math. arXiv:1106.1312.

[21] Kishimoto, T., Prokhorov, Yu., Zaidenberg, M.:Ga-actions on affine cones. Trans-

form. Groups18, 1137–1153 (2013)

[22] Kishimoto, T., Prokhorov, Yu., Zaidenberg, M.: Unipotent group actions on del

Pezzo cones. Algebraic Geometry1, 46–56 (2014)

[23] Kollar, J.: Singularity of pairs. In: Algebraic geometry (Santa Cruz, 1995) Part 1,

Proc. Sympos. Pure Math.62, Amer. Math. Soc., 221–287 (1997)

[24] Kollar, J., Mori, S.: Birational geometry of algebraic varieties. Cambridge Tracts in

Mathematics134, Cambridge University Press (1998)

[25] Lazarsfeld, R.: Positivity in algebraic geometry II. Ergeb. Math. Grenzgeb. (3)49,

Springer-Verlag (2004)

[26] Laufer, H. B.: On rational singularities. Amer. J. Math. 94, 597–608 (1972)

[27] Manin, Yu.: Rational surfaces over perfect fields, II. Mat. Sb.72, 161–192 (1967)

[28] Martinez-Garcia, J.: Log canonical thresholds of del Pezzo surfaces in characteristic

p. arXiv:1203.0995.

[29] Masuda, K., Miyanishi, M.: The additive group actions onQ-homology planes. Ann.

Inst. Fourier (Grenoble)53, 429–464 (2003)

http://arxiv.org/abs/1106.1312



[30] Nadel, A.: Multiplier ideal sheaves and Kahler–Einstein metrics of positive scalar

curvature. Ann. of Math.132, 549–596 (1990)

[31] Perepechko, A.: Flexibility of affine cones over del Pezzo surfaces of degree4 and

5. To appear in Funct. Anal. Appl. arXiv:1108.5841v1.

[32] Park, J.: Birational maps of del Pezzo fibrations. J. Reine Angew. Math.538, 213–

221 (2001)

[33] Park, J., Won, J.: Log-canonical thresholds on del Pezzo surfaces of degrees> 2.

Nagoya Math. J.200, 1–26 (2010)

[34] Schwartz, H. A.:Uber diejenigen Falle, in welchen die Gaußsche hypergeometrische

Reihe eine algebraische Funktion ihres vierten Elementes darstellt. J. Reine Angew.

Math.75, 292–335 (1873)

[35] Snow, D.: Unipotent actions on affine space. In: Topological methods in algebraic

transformation groups, Progress in Math.80, 165–176 (1989)

[36] Tian, G.: On Kahler–Einstein metrics on certain Kahler manifolds withc1(M) > 0.

Invent. Math.89, 225–246 (1987)

[37] Tian, G.: On Calabi’s conjecture for complex surfaces with positive first Chern class.

Invent. Math.101, 101–172 (1990)

[38] Tian, G.: Existence of Einstein metrics on Fano manifolds. In: Metric and differential

geometry, Progress in Math.297, 119–162 (2012)

[39] Tian, G., Yau. S.-T.: Kahler-Einstein metrics on complex surfaces withC1 > 0.

Comm. Math. Phys.112, 175–203 (1987)

[40] Winkelmann, J.: On free holomorphicC-actions onCn and homogeneous Stein

manifolds. Math. Ann.286, 593–612 (1990)


Afﬁne cones over smooth cubic surfaces - arXiv · Afﬁne cones over smooth cubic surfaces 5 In...

Documents

Transcript of Afﬁne cones over smooth cubic surfaces - arXiv · Afﬁne cones over smooth cubic surfaces 5 In...