by
Igor Dolgachev
Contents
0. i.
Introduction Weighted projective i.i. 1.2. 1.3. 1.4. 1.5. Notations Interpretations The first properties Cohomology Pathologies of 0F(n) space
2.
3.
complete intersections
Quasicones Complete intersections The dualizing The Poincare Examples sheaf series
4.
The Hodge structure on cohomology of weighted hypersurfaces. 4.1. 4.2. 4.3. 4.4. A resolution of ~i X The Griffiths theorem calculation
Explicit
O.
space) and deal with it as it would be a nonsingular space. A generalization of this approach
is the technique
projective
85
spaces,
Most of the results from this section Can be found in [7]. the Bott theorem on the cohomology of twisted spaces. Another proof
of the same result can be found in [23] and a similar result for torical spaces is discussed in [5]. the notion of a quasismooth subvariety of a weighted
For this we define the affine quasicone over a subvariety and is smooth outside its vertex. We show that quasismooth
have many properties of ordinary smooth complete The work of Mori [19] contains a similar re-
space.
conditions.
intersections.
of Catanese and Todorov show that the local Torelli theorem fails for some quasismooth weighted complete intersections In section 4 we generalize (see [4, 24]). the
present paper is partially based on my talks at a seminar on the Hodge-Deligne theory at Moscow State University participants in 1975/76. It is a pleasure to thank all of its
i. i.i.
Weishted projective
space.
Notations integers;
IQI = qo +-.. + q r ; algebra k[T 0 .... ,T r] over a field i= O,...,r ; ~(Q) = Proj(S(Q))
k,
~r" +i - {0}
we suppose in the sequel that the characteristic though many results are valid without (q0 ..... qr ) = i. The last assumption
this assumption.
is not essential
Lemma
Let
Then
~(Q)
Really,
([12], 2.4.D
isomorphism
= Proj(S(Q))
= Proj(S(Q)(a))
36
We refer to 1.3 for more general For any graded module the graded By 2.5.2) M A-module we denote M
0Proj(A)-Module , associated
with
M(n) k = M n + k Recall M.
([12];
= M(f)
fk : m e Mkd}
, Proj(A) .
D+(f) = Spec(A(f))
1.2. 1.2.1.
Interpretations. It is well known that a ~-graduation algebraic torus of a commutative G m ring is equivalent In our case
to an action of a 1-dimensional G
m
on its spectrum.
acts on
~r+l
= Spec(S(Q))
as follows
qi
, .
i = O,...,r
= Spec(k[X,x-l])
k'm k
is given
k' * k,r+l k' r+l (t,(a 0 ..... ar)) (a0 tq0 ..... ar tqr) The open set U = ~r+l_ quotient {0} u/G is invariant with respect to this action and the uni-
versal geometric If k = ~
exists and coincides with ~(Q) . m is the field of complex numbers then the analytic space is a complex analytic by the formulas quotient space ~r+l _ { 0 } / ~ *
~an where
asso-
(t,(z 0 .... ,Zr)) (Zo tqO,...,zr tqr) In view of this interpretation to a n o n e f f e c t i v e action of G the space m . ~(Q') from the lemma in i.i corresponds
1.2.2.
integer
we denote by subgroup of G m
This is a closed
~Q =
~q0 x ... ~ qr
on
which
in the coordinate
of rings
S(Q) s , T i -~ T i
$7
S(Q) = s~Q.
trums is well defined and gives an isomorphism P(Q) In case k = ~(Q)an = ~ r ( ~ ) / where ~Q(~) ~Q(C) = Proj(S Q) ~ ~r/ ~Q"
(g,(z 0 ..... Zr)) ~ (z0g 0 ..... Zrg r) g = (go'"''gr) 1.2.3. ' gi = exp(2~ibi/qi) ' easily gives, space ~(Q) 0 ~ b i < qi" for instance, that for Q =
{l,l,...,l,n}
~(l,l,n) , n # 1
n
is obtained by the blowing down the exceptional (when n= 2 it is ~ ordinary quadratic cone).
1.2.4.
For
Q = {l,ql,...,q r} ~r .
the spaces
~(Q)
of the af-
fine space
D+(To)
to the spectrum of
Its complement
~(ql,...,qr ) .
1.2.5.
(qo,...,qr )
is isomorphic to the polyhedral space ~r+l {(x 0 .... ,x r) : Eqix i = q0...qr } The first properties For different Q and Q' the corresponding
of [6], where
1.3. 1.3.1.
spaces
~(Q)
and
~(Q')
can
be isomorphic. Let d i = (q0,...,qi_l,qi+l,...,qr) a i = .c.m.(do,...,di_l,di+l,...,dr) a Note that = .c.m.(d 0 .... ,dr). (ai,d i) = I and aod. = a .
ii
ai[qi,
Proposition.
(Delorme
[7].)
Let
Q' = {qo/a,...,qr/ar} .
natural isomorphism
~(Q)
For the proof we consider the graded subringai note that S' = k[X0,...,Xr] , where Xi = Ti
is of degree
aqi/a i .
But then
38
S(Q')
Proj(S(Q'))
= Proj(S')
an isomorphism
Corollary. O,...,r .
Each
~(Q)
= ~ ( Q ' ) , where
for
i =
Corollary.
Assume
For example,
that qi = a. for i = 0,...,r. Then ~(Q) = ]pr. 1 it is so if all numbers .c.m.(q0,...,qr)/qi are coprime. fact was independently r = 1 discovered by M. Reid.
In
this case the previous Note that in case ~(q0,ql) 1.2.2. = ]pl for any
we can use the previous corollary and obtain that This fact however follows also from interpretation
q0,q I .
1.3.2.
Remarks.
i.
of
m o r p h i s m of sheaves
It can be shown that the isomorphism ~(Q) ~- ]P(Q') of 1.3.1 induces an isor 0~(n) = 0 ~ , ( ( n - ~ bi(n)qi)/a) , where h.(n)1 are uniquely i=O determined by the property n = bi(n)qi + ci(n)d i _ , O_<bl. <dl" "
1.3.2.
Let
G V
of a finite-dimensional
vector space
is called a pseudoreflection
such that
g(x) = x + fg(X)eg
for every
x V .
Lemma.
([3], ch.V,
5, th.4.)
Let
be
and in
BG ,
the subalgebra
of G-invariant assertions
elements.
is invertible
k.
are equivalent:
Example.
pQ
acts on
by pseudoreflections.
j # i,
i = O,...,n.
irreducible
projective
algebraic
variety;
39
(ii)
of
]P(Q)
are cyclic
quotients
singularities
(in particu-
(iii)
a nonsingular
group and use interpretation 1.2.2. To see (ii), we use inr terpretation 1.2.1. Let ~ = u U. be the canonical covering of , where U. = 1 1 i=0 D+(T i) . Consider the closed subvariety V i = Spec(S/(Ti-l)) of ~r+l = Spec(S). The action of tifying V.I G m with Spec(k[T0,...,Ti_ l,Ti+ l,...,Tr]) j = 0,...,i-l,i+l,...,r on Ar+l induces the action of ~ qi on V. l which, after iden-
U. = V./~ i i qi
and, since
Vo = ~r, 1
we have property
(ii)
For the proof of (iii) we use 1.3.1 and the previous we may assume the action of qi = i. that ~ (q0'''''qi-l'qi+l '''''qr ) = i . on V. is generated
construction.
By 1.3.1
by pseudoreflections
qi It remains
to apply 1.3.2.
Cohomology Recall
of that
0~(n) 0~(n)
S(Q)(n) .
which defines
a natural
Theorem. (i) (ii) (iii) ~n : Sn + H0(~' 0 F (n)) Hi(~,0~(n)) Hr(~,0~(n)) = 0 for is bijective for any neZg ;
i ~ 0,r, n ~ ~ ;
= S_n_IQI .
Proof.
According
to general
properties
of projective
spectrums
spectrum
0F-Algebra
projection Since p
coincides
morphism
U U/G
is an affine morphism
Hi(U,0u)
= Hi(p, p,(0U))
= Hi(p,
theory
([13]).
S H0(U,0u ) H~m}(S)
40
H~m}(S)
-- Hi-I(U,0u ) '
It is easy to see that the homomorphism Serre homomorphism have H~m}(S) = O, (i), i ~ r+l . en:Sn -~ H0(~,0]p(n))
is a Cohen-Macaulay
For the proof of (iii) we have to use the explicit We have r+l H{m}(S) = ~ m = i~ m where m (Tm) = (T O . . . . Let V V Extr+l(s/(T0 ..... Tr)m,s) Extr+l(s/(r_m),s) =
and consider
on its ex-
deg(eilA'''Aei P
The Koszul
r+l A (V) 0
e.A...Ae. Z(-I)kT TM e. A...A~. A...Ae. iI ip k ik iI ik Ip defines a resolution of graded S-modules for S/(T 2) and hence we have an isomor-
0 < a. _<m i
tm a 0, - ,a r
form a basis of
In this notation
. (t m = t m+s m,mts a0,...,ar ) a0+s,...,ar+S be the image of em a0,.-.,a r in lim I . ~> m Module r+l H{m}(S ) is a
Let
e a0,-.-a r
41
e a 0 ,..
We have r Z i=0aiqi
deg(eq0 ..... ar) = deg(eam 0 ..... ar) = ( m - a 0 > +... + ( m - a r) - m I Q I = Thus we obtain that e with a 0, .. ,a r r n = - E aiq i (a i > 0) i=0 Hr+l {m}(S)n as a k-space. Since dimkS_n_Q = #{(b 0 ..... b r) e ~ r + l :_n_IQ I = r E biq i} = i=0
generate
= #{(a0,...,a r) c IWr+l+ :-n = we have Hr+l -. {m}(S)n -- S_n_IQI. It remains to notice that Hr+l -. e H r ( ~ , O ~ ( n ) ) { m}(S)n 1.4.2. Let integers Ps(t) = Then as a c o r o l l a r y a n
r 2 aiq i} , i=0
a dimkHi(F,0F(n)) = 0
i=
an = dimkS(Q) n . 1.5. Pathologies If (i) (ii) (iii) ~ = ~r for any then the following properties are well known. ne ~ 0F(n) is an invertible sheaf; 0F(n) is ample; S(n) 8 l~(m) + S ( n + m ) ; n e induces the isomor-
an invertible sheaf
(iv)
~(Q) .
1.5.1.
Let
Q = {1,1,2}.
The restriction of
0 F (i)
to
D+(T 2)
is given by the
42
S(T2)-mdule S(1)(T2) = {_~: T2 It is clear that one. This is a counterexample to property (i). a ( S2k_l} . T1 S (T2)T22 is not a free S(T2~module of rang
TO S(1)(T2) = S ( T 2 )+ ~L
1.5.2. ble.
~ (qo,ql) to and
0~ (n)
are inverti-
is associated n = kqi-pqj
~ (q0,ql) _- ~i Moreover, b n if
F(~,01p(~)) # 0 , then
Thus,
(n)
is ample if (1.4.1).
r(~,0~(n))
= 0
This is a counterexample
to property
(ii).
1.5.3. b q0+ql +I
ql = q0 + I ,
qo > 1 .
Then
b q0
This is a counterexample
to property (iii).
1.5.4. that
M = S(m) , note
1.5.5.
We refer to the paper of Delorme ([7]) for more details concerning proper0~(n) . For example, one can find there a generalization of
We remark also that according to Mori ([19]) everything is well in the open set V = n D k, k>l where Dk = u D+(Ti) . kxqi Namely, V is the maximal open subVmc m .
0~(1) IV
is invertible and
2. 2.1.
43
2.1.1.
Let
i be the S-module of k-differentials of S . This is a free module ~S i i dT0,...,dT r . Denote by ~S its exterior i th power A(~) (as ~S 0 = S). . This is a free S-module with the basis Grade i ~S by the condition
{dTslA...AdT
si
d e g ( d T s l A ' " A d T s i) = q s l + ... + q s i We have an isomorphism of graded S(Q)-modules i ~ ^< < ~S u-s I ...<siNr For i = r we obtain ~r+l S Let I d:S a S S(_IQI ) universal differentiation. By definition of the S(-qs I ) .
....
qs i
partial derivatives
a S .
i ' w ~S
w' g~S
d(d(w)) = 0 ,
2.1.2.
V a ~ S(Q) n
Using the linearity of both sides of this identity we may verify this formula So .Tsr . But in this case it can be a is a monomial T O .. r done without any difficulties. only in the case when
2.1.3.
of grades S-modules
i ~ i
A(dTslA...^dT s ) = Z (-1)k+lqskTskdTslA...^dTskA...AdT
si
Lemma.
(i) (il)
(iii)
A(w^w')
WEas,
w'
A(da) = n a , A(dw) + d ( A ( w ) )
44
Using
the linearity
of
we may verify
w = dTslA...AdTsi,
formula.
(iii)
to consider
w = adTslA'''AdTs''l
a ~ Sk .
A(dw) = ~(daAdTslA...AdT s ) = ~(da)AdTslA...AdTs. = kadT sI d(A(w)) = d(aA(dT sI A...AdT s. 1 A...AdT si - daAA(dT SI ) = daAA(dT sI A...AdT si A...AdT si )
- da^A(dTslA...AdT s ) =
) +ad(A(dT sI ^
^...^dT si
) =
= (k+
i E )w = nw. =I qs~
2.1.4.
the sequence
r + .... + ~i + S aS S for the regular sequence. sequence (qoT0 .... ,qrTr ) Thus, we ob-
complex
A--+~si-l) = Im(a~ + l ~ a ~ )
2.1.5.
Define --i ~
the sheaf
--i ~ M
on
~(Q) denotes
--i = ~ , where
Also we put --iaF (n) = _~s(n) , Since MM is an exact n ~ 7z . functor we have exact -> ~iIp-l(n) ~ 0, , ~ : S(n(n) = 0, sequences of sheaves on ~(Q) :
(n) : ~ + l < n )
[qJ) : 0 ~ (n-
[q]).
45
2.1.6.
char(k) = 0
property
"'"
A r + l + O. + ~S
2.2.
Justifications. In this section we try to convince the reader that the sheaves % introduced
in the previous section are good substitutes for the sheaves of germs of differentials ~i ~r on the usual projective space ~r.
2.2.1.
Let
I~ = I~r .
is the complement to the zero section of the tautological line bundle on ]pr and the canonical morphism p : U + I~r is smooth.
0 -~ p ~
+ ~U ~ U / ~
~ 0
i-i Op.~ + O.
0 p ~ The homomorphism i A : aS S
-~ a U ~ U / ~
A:~ U
_~Ou
It is easy to verify that A(p*~ I) = 0 and
~: au/~
Since i ~U/l~
0 U.
~ is in fact an isomorphism.
and applying
p. 0
@ ~s(n)+
neff
@ ~-l(n)-~ neff
It is easy to see that in this way we obtain exact sequence 2.1.5 of the definition of g$(n) .
4B
2.2.2.
Lemma.
is generated by
pseudoreflections and its order is invertible in phism i i G ~A/k (~/k) is an isomorphism of A-modules.
Proof.
Since
i ~B/k
i (reap. ~A/k )
is a free B-module (reap. A-module). Since B is a free A-module ([3],ch.5, 5, i i i th.5), ~B/k is a free A-module. Let ~A/k ~ ~B/k be the canonical homomorphism of A-modules (the inverse image of a differential form). It is injective (because i ~A/k is free and it is injective over a dense open subset of Spec A). Let T be its cokernel and 0 ~ i i ~A/k * ~B/k + T 0 i m ~ ~-~ ~g(m)
be the corresponding exact sequence. Now, for every G-B-module M, the homomorphism is a projector
onto a direct summand (here we use the assumption that #G is invertible in k ) , G thus the functor ( ) is exact. Applying this functor to the above exact sequence, we get an exact sequence i i G TG 0 ~A/k ~ (~B/k) ~ ~ 0 where that Let Q i G , being a direct summand of a free A-module, is a projective A-module. (~B/k) and, hence, depth (TG)p = 0 QnA (TG) e d i m B - i. P of A This implies of height i. TG = 0 if its localization B Q, for any prime = P and
be a prime ideal of
GQ = {g c G: g(Q) = Q} be
equal to the ramification index of the extension B ~ B' ([3],ch.V, 5, n5). This Q i Q G!/G . Q Q~ i i Thus, it suffices shows that ~ / k @Ap A' Q = ~B~ and, hence, ~i Ap = (~B.)
Q
to show that i
~B~
= (~
.Q)G~
Passing to the completions, we may assume that generator unity Then i = ~B~ dt. A...^dt dT e ~B 4 J1 J i-i ~. g of
v GQ
and a
acts on
BQ
by multiplying
Let
tl,...,tn_ 1
47
= ~BQ dtj A...Adtj ^dT i i-i G' i Q i A direct computation shows that (aBQ/k) = ~B~/k" Q
2.2.3.
Let
p r ~ ~r/~Q = ~(Q)
from 1.2.2.
) ,
is the functor of invariant direct image ([ii],5.1). on ~r is induced by one on S . Since SG is a polynomial
The action of
algebra, this latter action is generated by pseudoreflections and hence, by lemma 2.2.2, we have an isomorphism of S(Q)-modules i = (a~)G aS(Q) and, hence, an isomorphism of sheaves ~(Q)
Applying a~
G'ai"
= a,t~S)
to the exact
0 a i a i a i-I ~r --S ~r 0 and using the exactness of a~ (p is an affine morphism and ( )G is an exact
functor) we obtain an exact sequence: i 0 + a~(a I r ) + aS(Q) Since G a,(0 r) = 0F Gt~i-l~ 0 Gi --i a,(a r ) = a ~ .
2.2.4.
~p~i
introduced
f o r any
V-variety
in [23].
~i a~
. ai = 3,( w )
Recall that
where
j :W +
In notations
j'
~ ~r
-+~
action of
and j' is the natural immersion. Since W is smooth, the -i (W) is generated locally by pseudoreflections. Then, by
48 ,G ~i
a ,(~ a -i (w) ) "
Since
codim(]P-W,~)
_>2
(~
is a normal ~r
scheme)
and
~r
is smooth,
~i
G, , ~i
aG(~r
2.3. 2.3.1. im } H{
Cohomology
of
~$(n)
the graded
for a S-module
(cf. 1.4).
k,
j = i+l
# r+l
Proof.
sequences i --i-i
which, mology
. . . .
after applying
the functor
H{m}(aS) H{m}(a S )
i = S(-n) ~S for some n ~~
j-i
j-l---i-l-
"
(~)
H~m }
Since if
is a Cohen-Macaulay
j # r+ I .
)
By induction,
= H{m}(~ S
-l-i-1
for
j # r + i.
we obtain
{m}
Now, first terms of the Koszul
m +
(ns)"
complex
0 ,
from
(2.1.4)
give an exart
sequence
which easily
implies
that
H~m}(~) = { 0,
k , This proves the proposition.
1 # 2, r + l
i = 2 # r+ i .
Corollary.
--i~s is a Cohen-Macaulay
S-module
if and only if
i = r.
49
2.3.2.
denote by Put
Notice that
Z jeJ an = dimeS(Q) n .
[Qj[
the sum
qj.
Theorem.
Let
h(0,i;n) = #J=iZ an_[Qj] - h(0, i-l;n) , h(j,i;n) = 0 , h(i,i;0) = i, h(i,i;n) = 0, h(r,i;n) = if j # 0,i,r , i = 0,...,r n # 0, i # r,0 n 2Z
i>0 - '
nc2Z
Proof.
Using the same arguments as in the proof of theorem 1.4.1 we obtain the
exact sequence
0 H +
n 7z
(n))
and an isomorphism
H{m}(as)
n~m Hi-1 ( ~ , ~ ( n ) )
@ S(-IQjI)
#J=i
is surjective
except the last one. Consider exact sequence 2.1.5 0 ~(n) _~s(n) ~ - l ( n ) 0
and the corresponding cohomology sequence Hr-l(~(n)) Since ~(n) -~ Hr-l(~-l(n)) -~ H r ( ~ (n)) Hr(_~s(n) ) H r ( ~ - l ( n ) ) 0.
0~ (n-[QjI) #J=i
r i dimkH (is(n)) = #J=r+l-i a-n- 1QJ [ Using this sequence and preceeding results we obtain the last equality.
2.3.3.
Corollary.
50
H0(~,~(n)) Hr(~,~(n))
= 0, = O,
if if
n <min{IQj[: n >-min{IQjl:
#J = i} #J = r+l-i} .
2.3.4.
Corollary
(Bott-Steenbrink). ~ 0 n >min{IQjl:
If
n >0
then
#J = i} .
2.3.5.
Corollary. h(0,i;n) = i ~ (-i)/+i E /=0 #J=~ an-IQJ 1 . follows from 2.3.2 and to verify the second
immediately
one we have to consider the identity h(r,i;n) - h(0,r-i;-n) = h(0,r+l;-n) = dimkH0(~,~--r+l (-n)) = 0 .
2.3.6.
Corollary.
If
k = K , then , i even
0 , i, ~P'q(]p) = 0,
odd
p = q
p # q P,q E1
=
This follows from the degeneracy of the spectral sequence HP+q(x,K) proven by Steenbrink [23].
Hq(~,~--P) ---->
3. 3.1. 3.1.1.
Weighted complete
intersections.
Quasicones. Let X be a closed subscheme of a weighted projective be the canonical projection. in A r+l is called the affine quasicone over C X. I of CX in S is equal space ~(Q) and
p : U + ~(Q)
p-l(x)
to
51
ideal of the graded ring depthm(S/I) ~ i); (iii) The closed embedding projection (iv) I S S/I
This is an easy exercise in the theory of projective schemes, which we omit (it will not be used in the sequel).
3.1.3.
An affine variety
is called Nuasiconic
(or quasicone)
if there is an
G on V such that the intersection of the closures of all orm is a closed point. This point is called the vertex of a quasicone.
3.1.4.
Proposition.
Let
k.
The
following properties are equivalent: (i) (ii) (iii) V is a quasicone; has a nonnegative grading with j : V ~r+l k[V]0 = k ; j(V) is invariant with
k[V] = F(V,0v)
such that
(iv)
G on A r+l defined as in 1.2.1; m there is a closed embedding j : V + A r + l such that the ideal of
j(V)
is
The proof consists of standard arguments of the algebraic group theory (cf.
[8], [20]).
Corollary. Any affine quasicone is a quasicone. Conversely any quasicone without X c ~(Q) .
3.1.5. bedding
A closed subscheme
X c ~(Q)
3.1.6.
Theorem.
X c ~(Q)
is a V-variety.
Proof.
Let
CX
and
x~ X
be a closed point.
In
W i = V in C X.
y ~W. i TCx(Y)
is not contained
*
of
to
C X = C X - {0}
and
F = p'
(X)re d .
The fibre
52
point
Gm
restricted of
to
C X.
If F coincides with
(yo,...,Yi_l,l,Yi+l,...,~r)
y , then
Gm = Spec(k[t,t-l])
= Spec(S)
--
q0
--
qi -I
qi --
q i +I
--
q2 )
(To, .... Ti_l,Ti,Ti+ 1 .... ,Tr) (yo t The tangent line to the curve defined by the equation To-Yo F
,. . ,Yi_l . t
. ,t
. ,Yi+l t
,.
,yr t
= qrYr "
TF(Y) ~ TV (y) = V.1 and, since TF(Y ) c TCx(Y ) , we obtain that i y is anonsingular point of W.. l The end of the proof is the same as in the proof of 1.3.2: we obtain that Uk c X is locally isomorphic Gy c G m to the quotient of the nonsingular variety Yi " Wi by the
It is clear that
isotropy group
of the point
3.2. 3.2.1.
Weighted
complete
are the degrees of these elements then we say that of multidegree d = (dl,...,dk) (F) and X and denote
plete intersection
In case
is a principal of degree
ideal d
FE S(Q) d by
we say that
weighted hypersurface
and denote
Vd(Q) .
3.2.2. Let
In the sequel,
be t h e
CX
C X - {0} .
p : CX X
corresponding
X = Vd(Q) if
is quasismooth,
Then
dim X ~ 3 ;
*
any G -equivariant etale covering of C X ,m Zl(Cx) = 0 if k = C and dim X e 2 ; HI(c~,0C*> ~ X = 0, 0< i< dim X.
is trivial if dim X z 2 ;
Proof.
(i)
0Cx, O
is a complete
intersection
ring of dimen-
sion 4, regular outside its maximal This shows that Pic(C X) = Pic(Cx)
ideal, it is a factorial
group of classes of invertible divisorial trivial (ii) ([i0]). This proves (i).
A similar reference CX
etale covering of
is a restriction
53
the same is true for Gm-equivariant G m-equivariant integral k[Cx] c B . Then etale covering of and coverings. Then Let f : Y + CX be an irreducible B = n~ Bn is an
CX . f
Y = Spec B , where
of graded rings o e C X-
m B c B>o
k-algebra. B
is integral,
o the only point lying over the vertex of braically (ii)' Let closed, CX this implies that f
is a field.
o Thus, we obtain
is a subalgebra
of an integral algebra Y
that CX .
is a quasicone and its vertex is Because f is etale and This proves k is alge-
is an isomorphism. into ~n . CX .
(ii). ~+ of posi~+-
embedded
acts freely on
every
orbit with a sphere S 2n-I of small radius c with the center at the origin, we , a s2n_ 1 get a map C X + ~ + x K e , where K = n CX . It is easily verified that this , map is a diffeomorphism of C X onto ~+xK . Now, since the vertex of C X is a complete intersection isolated singularity, the space nI(Cx) K is (d-2)-connected if dim X e 2 .
[14,18]).
Thus,
= ~I(KE) = 0 theory.
Since
C X = Spec(S/I)
i * 0C~) = Hi+I.c = i+l H (Cx, {0}( X,OCx ) H{mo}(S/I ) . Since S/I, being a quotient of a regular ring by a regular sequence, is a Cohen-
Macaulay
ring,
H{m0}(S/I)
i+l
= 0 ,
if
i+l
# dlm(S/I)
= dim X + i .
3.2.3. example,
Remark.
If
For
n ~ i g ( A n _ {0}) # 0,
has nontrivial
of the lemma
= ~ , = 0 ,
Zl!X) = 0 , HI(X,Ox(n))
dim X e 2
= 0 ,
n ~ ~ ,
Proof.
Let
be an invertible as a
sheaf on
Since
Pic(C X) = 0
,
P (L) = 0 * CX
and is determined = ~
XL e H I ( G m ~ .
f : Pic(X)
= L' = p~(p*L'))
and, hence,
and
be a corresponding of C X* is a
0 X-
54
covering,
by lemma 3.2.2
is tri-
Gm-0c~-Algebra
i.e.
B = B 1 B2 ,
B. are nontrivial. Since G is connected, the Subalgebras B 1 and B 2 l m are invariant Subalgebras and we have a splitting of Gm-Algebras B = B1 B2 Gm G Applying p, , we obtain that A = p,dm(B) = p,m(B I) p, Gm (B2) splits. This shows that the covering X' splits and proves (ii).
wise connected. To prove (iii) we note that Hi(c~,0 ,) = H i (Cx, * 0 XS0~ 0 U) = Hi(X,0~Sp,0 U) = n ~ H i ( X , 0 x @ 0 ~ Cx But 0XS0~ 0 F (n) = Ox(n ) and we can apply 3.2.2 and obtain 0~ (n)) .
(iii).
3.2.5.
Remark.
Pic(X)
is generated by some is
Pic(~(l,...,l,n))
3.2.6.
3.3. 3.3.1.
The dualizin$ sheaf. Recall that according to Grothendieck variety X there is a sheaf n-i (X;F,~x)) F. * for any normal wX integral projective such that
Cohen-Macaulay
0x-MOdule
The sheaf
In other words,
= ~n
~0 X where j :Z + X
3.3.2.
Lemma.
Let
be a closed quasismooth
subscheme X,
of
, Cx
be its pro-
locus of
then
S5
Proof.
CX * + U
7:1 ~U/~
(its construction
~ r , the latter is used only to check that ~ is an isomorphism). It is easy to i see that the induced map ~ : ~C~/X 0C.*. is well defined and is surjective. x G m Since p, is exact, the map G G m,~l . p,m(~) : P* t c* ) OX
~x/x
is surjective. side sheaf to Z Thus, it is sufficient is an invertible is local. to show that the restriction of the left hand sheaf. Let xe Z and Z x be its neighbourhood C~ of codimension of the G
is a nonsingular G
subvariety of
i and
m is regularly embedded i
constructed in i
0 NWlCx
Since one) NW/Cx that is locally NW/Cx = 0W .
~c~iXOOcOW+ ~WlZx
free of
o .
(replacing W by smaller
pT(~lc,/x00
X
C~
0W)
G i acts by pseudoreflections and hence p,(~ ) = 0 G W/Zx m Applying p, to the above sequence we obtain G 0
: 0Zx = p*(W)
c*IxJlZxx
3.3.3.
Proof.
Since
is a normal Cohen-Macaulay
variety
by 3.3.1 it is sufficient to show that G m,~n+l, P* t C~ ) Iz ~ an "z Consider the exact sequence 0 ~ p*~ + ~I[p-I(z) ~Ip-l(x) 0.
56
Applying
G p,m
~z
~Z~OzOz
P* < c~ )
IZ
q.e.d.
3.3.4.
Theorem.
Let
X = Vd(Q)
multidegree
where
1~I = dl+'''+d s
Proof.
Let
quasicone over
and
B = S/I.
1/12 = B(-dl) + ... + B(-d s) . The exact sequence i->0 0 1/12 ~ IS @S B -~ ~B gives the homomorphism s o r +i -s i r+l-s (-Idl) = A(I/I~)@ B n (~B) -> 2 + AI ( ~ ) 8 s B f : ~B Since C~ is smooth, the restriction of f to C~ = B(-IQI) . Hence
is an isomorphism.
~r+l-s
c~
= B([_d] - ]Q[) .
3.4. 3.4.1.
The Poincare series. Let A = n~0 An be a graded k-algebra of finite type. Then its Poincare
series is defined by PA(t) = If then Xo,...,x r PA(t) E (dimkAn)tn n=0 . A and qo,...,qr are its degrees,
are homogeneous
generators of
is a rational function of the form r eA(t) = F(t)/i~ 0 (i- t qi) is a polynomial ([2], Ii.i).
where
F(t)
3.4.2. 5, nl)
Assume that
A = S(Q)
57
PS(Q)(t) = I/i~o(l- t qi) . Let and fl,...,fs be a regular sequence of homogeneous elements of the ring A = S(Q)/(fl,...,f s) . Then
s d.
S(Q)
d l,...,dr
PA(t) = i~l(1-t l)/ ~ (i-tqi) . i=0 This formula follows from the above formula. Then Put A 0 = S (Q) , A i = S(Q)/(f I .... fi ).
Ai = A i - 1 / ( f i )
d.
and obviously
i =
l,...~S
we put
Px(t) = n~o(dimkHO(X,Ox(n))tn
Lemma.
Let
m = n~0 An A
A .
Assume that
depthm(A) ~ 2
(for example,
is normal).
PA(t) = Px(t) The same argument as in the proof of 3.2.4 (iii) and 1.4.2(i) shows that the Serre homomorphism of graded algebras A is bijective.
3.4.4.
Theorem.
Let
X = Vd(Q)
Px(t) = n~ 0 an tn
Corollary.
Define
= 0x(Id I -IQI)
3.5. 3.5.1.
Examples. We shall say that two closed subvarieties X c]P and X' c I~' are affine
58
isomorphi 9 if their affine quasicones are isomorphic and projectively isomorphic if their quasicones are Gm-iSomorphic. implications projectively isomorphic => affine isomorphic projectively isomorphic => isomorphic between these three notions. It is clear that in general there are only two
3.5.2.
A quasismooth hypersurface
X = Vd(Q)
in a weighted
projective plane
is a smooth projective curve. i coincides with the canonical sheaf ~X and we have (3.3.4): ~xl = 0x(d_q0_ql_q2) Its genus is calculated by the formula g = coefficient at td-IQl in the formal series
(i- td)/i~O (i- tqi) . The affine quasicone of such a curve is given by a weighted-homogeneous equation f(x0,xl,x 2) = 0 with an isolated singularity at the origin. Such singularities
m< 0
Equation
Name
1 2k 4k+2 2k-2 12 18 3O
x0 =
= 0 = 0
A2k_ I , A2k Dk E6 E7 E8 ,
The equations of corresponding projecting quasicones are well known two-dimensional singularities, which are called the platonic singularities, D__U_U Val singularities,
Klein singularities, ADE singularities, double rational singularities, simple singularities, O-modal singularities. Note that any curve which is affine isomorphic to a curve of type Dk or E
59
is projectively
isomorphic
following curves: d Equation 3 3 3 x 0+x l+x 2 +ax0xlx 2 = 0 , 4 4 2 2 2 x0+x l+x 2+axOx I x~+ 3 2 22 xl+x 2+axlx 2 = 0 , Name
(qo,ql,q 2)
3 4 6
a3+27 a 2 -4#0
# 0
36 37 38
or or or
P8 X9 Jl0
= 0 , 4a 3 + 27 # 0
It can be shown (V. I. Arnold) number of collections curve Vd(q0,ql,q2) . For m= 1 (q0,ql,q2;d)
affine quasiThere is
in ~ 3 .
a natural correspondence
of the Fuchsian groups of the first kind with compact quotient bra of automorphic Of course, plane curve. forms is generated by three elements
3.5.3.
Surfaces.
f(x0,xl,x 2) = 0
Vd(Q).
Then the equation f(x0,xl,x 2) + x ~ = 0 defines a quasismooth For curves with hypersurface m = 0 Vd(qO,ql,q2,1) . [15] of de-
we obtain in this way dell Pezzo surfaces (M. Reid). we obtain simply-connected Resolving projective
gree 3, 2 and 1 respectively For curves with the dualizing sheaf m = 1 ~X ~ 0X"
surfaces with
its singularities
tional points) we get minimal models of nonsingular such a surface is the following Klein surface: V42(6,14,21,I) : x7+ x3+
K3-surfaces.
2 42 x2 + x3 = 0
This surface has 3 singular points (i,-I,0,0) (0,-I,i,0) (-i,0,i,0) of type of type of type A1 A6 A2 .
60
For any such surface the complement affine surface with an equation f(x0,xl,x 2) + i = 0 which is diffeomorphic ([18]).
to the curve
x3 = 0
is isomorphic
to the
F@
This fact can be used for the explanation of some observations surfaces (see [21]).
3.5.4.
Multiple
spaces.
Let m
X ~r-i
group of order
f(x0,...,Xr_ I) = 0 is isomorphic
Assume that
(d,char(k))
to a weighted quasismooth
hypersurface
Vd(Q) : f(x0 ..... Xr-l) + xmr = 0 where Q = {i ..... l,d/m}. It is easy to see that such varieties are simply-connected k = ~) (cf. [22]). Moreover, Px(t) if X is smooth. re 3 (i.e. = ~ if From 3.2.4 we obtain that all such ~ig(x) r e4 . = 0 or ~I(X) = 0 if
Pic(X)
The Poincare
series
Px(t) = (i- td)/(l- t)r(l- t d/m) = (i+ t d / m + " ' ' + t d ( m - l ) / m ) / ( l In particular, pg(X) = the coefficient at td_r_d/m = m-i s$0 I d(m-l-S)m r-i For example m = 2, m = 2, r = 2 r = 3, (hyperelliptic d = 6 curve) pg = d/2 - i pg
=
(K3-surface)
i . geometry to conspaces.
It is very useful for the construction sider also weighted multiple planes, For example,
problems
in algebraic
4. 4.1.
The Hodge structure on the cohomology of weighted hypersurfaces. ~i X" be a quasismooth weighted hypersurface, C X, f ~ S(Q) N its generator,
CX
cone,
I c S(Q)
the ideal of C X, m0
A = S(Q)/I C X,
C~ = C X - {0} .
is a V-variety
its cohomology has (in case k = C) a pure Hodge numbers are calculated by the formula
8~
~X "
4.1.1.
Define a i df : m S
setting for homogeneous elements of the S-module df(w) = fdw + (-l)i+l[~ 1 wAdf , where [w[ denotes for brevity the degree of w.
i ~S
(2. i. i)
n emNa.
(i)
(ii) (iii)
(iv) (v)
i WE ~S'
w' 4 ;
Vwe ~i
S;
d(df(w)) = (i + df(dw) =
)dfAdw,
Ywe ~S ;
" dfAdw,
Yw m S ;
This is directly verified. Let us show that df induces a linear map of S-modules
~S = Ker(~s---+eS
) (2.1.4).
(continuation). df(&(w)) = -A(df(w)) , d f ( ~ ) c =i+l uS . It is clear that (vii) follows from (vi). Let us prove (vi). Recalling proi w ~S ;
(2.1.3), we obtain
A(df(w)) = 8(fdw+ (-i) i+l l~[ wAdf) = f&(dw) + (-i) i+l = -fd(A(w)) + f[w]w+ (-i) i+l [~[ A(w)Adf-[w]fw = -fd(A(w)) + (-I) i ] A ( ~ A(w)Adf = -df(A(w)) N
I~I
=
&(wAdf) =
4.1.2.
Properties R[ 1
(ii) and (v) of the lemma make possible to introduce the follow-
ing complex
of graded S-modules:
Rki = ~((k-i)N) d k = (-l)kdf : R~l R~+ll Property of complexes A : R; Ri_l[-i ] . Property (vii) shows that (vi) implies that the homomorphisms k + ~S k-i A : ~S determine morphisms
62
is a subcomplex
of
R~
1
such that
of complexes + 0,
0 + -- R i R i -- Ri_l[-l]
4.1.3.
by
defines
the inclusion
of graded
S-modules
which
induces
the inclusion
--
The exact
sequence
of complexes
from 4.1.2
induces
the exact
sequence
of com-
0 + Ki + K_ i Ki_l[-l]
+ 0
4.1.4. sequence
Lemma
(De Rham). of
Let A,
be a commutative
ring, Then
w c A r+l wAh = 0
of elements h = wAB.
h e ~(A r+l) , p ~ r .
such that
of the Koszul
complex
for
of
We shall use this lemma in the following situation: A CX, A r+l = ~s/f~s i i , w is the image of df in ~s/f~s. i i Since * CX is smooth, the jacobian ideal
is the coordinate
ring
of the complex
Corollary: H q(K~.) = 0 , q # r+ i , Vi ~ ZZ .
68
4.1.5.
Proof.
The above corollary and exact sequence 4.1.3 imply that Hq(Ki) = Hq-I(K;_I[-I]) = Hq-2(Ki_l ) , qNr .
Since for
q< 0
and
q >r
Hq(Ki) = 0
4.1.6.
Then we deduce from 4.1.5 that the sequence of graded A-modules 0 + "~A ~(N)/~ is a resolution of ~i" X = Proj(A) , we obtain the resolution of the
+ ~+I(2N)/~+I(N)
.....
~s((r-I)N)/~s((r-i-I)N)
+ 0
..... ~ ( ( r - l ) N ) / ~ ( ( r - i - l ) N )
0 .
4.1.7.
such that
0~x/~l z d
where NX/]pIZ = 0z(-N)
~1
80 V 0Z
~z1
0 ZI ~ .
0 +
i +
0z d_<
z+l(N)
which can be extended to the right to obtain the resolution 0 ~iz ~i~ (N) 80~ 0Z ~2m-i+l(2N)@ 0 P 0Z -~ "'" " Since ~ + k ( ( l + k)N) @ 0]p 0Z -~ ~ +k((l + k ) N ) / ~ +k(kN) l Z we see that this resolution is the resolution of Hence i and we obtain that --i ~7 --A (4.1.6) restricted on Z .
64
r ( ( r - i + l ) N ) / ~ ((r-l)N) 0 . ~i~
4~2.
This theorem generalizes for weighted hypersurfaces a result of [II] and allows i Hi to calculate the cohomology H (X,~ X) as certain quotient spaces of differential forms on with poles on X. th
4.2.1.
Denote by
Kp
the
~iax f r o m 4 . 1 . 7 :
Kp ~ 0
and the theorem of Bott-Steenbrink (2.3.4) we obtain that Hq(X,K p) = H q ( ~ , K p) = 0 , q> 0, p >0 = { k, q=i-i 0, q#i-I
Then we have the exact sequence of sheaves 0 Hi K0 ~X L 0 which gives the exact cohomology sequence . . . . Hq-I(x,L) H q ( x , ~ ) The sequence 0 L K1
is an acyclic resolution Hq(X,L) K2 of L. Kr - i - 1 T h u s we h a v e = 0, q > r-i-2 r-i-2) =
Hq(X,K 0) Hq(X,L) . . . .
= Hq(F(X,K'))
Hr-i-2(X,L
) = r(X,Kr-i-1)/imr(X,K
r (~,~
((r-i)N) / r ( P , g ~ ( ( r - i - 1 ) N ) ) .
I m r ( ~ - , ~ -1 ((r-i-1)N)
4.2.2.
The homomorphism
65
Proof. of
Follows .
cohomology
sequence
Hq(X,L)
Corollary. tures
Assume
that
k = ~ .
of the Hodge
struc-
Hn(x,~) if n # r-i
= Hn+l(~,~)
and an epimorphism Hr-I(x,K ) Hr(~,~) . follows from the theorem. For n < r-i we use the
For Poincare
n er-i duality
homology
Since
of
only cohomology
are interesting.
4.2.3.
Theorem
4.3. 4.3. i.
Explicit Let
calculation.
~f ~f Of = ( -~0 ..... ~--Tr ) be t h e j a c o b i a n affine quasicone ideal with respect Cx of a weighted to a generator f c S(Q) N of the i d e a l of the
quasismooth hypersurface X c ID(Q) . ~f By the Euler formula 2.1.2 each ~ . is a homogeneous element of S(Q) l degree N- qi" Hence t h e i d e a l Of i s homogeneous and t h e q u o t i e n t s p a c e S(Q)/0f has a natural gradation. Since ef is m0-primary this quotient
of
space
is finite d i m e n s i o n a l .
66
4.3.2.
Theorem
(Steenbrink).
Assume that
char(k) = 0.
Then
hi,r-i-l(x 0
= dimk(S(Q)/0f)(r_i)N_iQ I
Proof.
S(Q)-modules:
Since (see the proof of theorem 2.3.2) F(~,~(a)) and the differential df = (~)a, Ya~ ~ corresponds to the operator
F(~,~
from 4.1.1, we can reformulate theorem 4.2.3 in the following form: 0"i'r-i-l(X) = dimk(~s/df~ n S (-N) + f~s(_N))(r_i)N --r ~--r-i
Thus it remains to construct an isomorphism of graded S-modules ~s/df~s -I(-N) + f~s (-N) ~ ~r+l" r ~ S/~s(-N)'Ad-i . By property (iii) of lemma 2.1.3, we obtain that the k-linear map d: ~S ~ + i is in fact an isomorphism of graded S-modules (here we use that By property (iii) of lemma 4.1.1, we set d(d?~-l(-N))
In fact, we have here an
char(k) = 0!).
c a~(-N)^df
Since
equality.
is
S-linear
it
is
sufficient
to
show
that all forms dXilA'''AdX'l ^df ~ d(d?~-l(-N))) r But r-i s+l A~Xi d(df(A(dXil...dx i ))) = d(df(s~l(-l) x i dXilA... ^...^dx.l )) = r s s r ^ = d(rfdx. ^...^dx. + (-l)ra(z(-l)S+Ix. dx. A...Adx. A...Adx )^df) =
11
= C dx. ^...Adx.
1r ^df
11
ir
are some rational numbers which we are too lazy to write down explicitly. induces an isomorphism r+l r ~s/df~s -I(-N) ~ ~S /~s(-N) df .
We have
But, since
f E 8f
f_r+l e ~ A d f ~S
and hence
d(f~s) c dfA~s .
87
4.3.3. function
Define a
'
a = (a0, "" .,ar) e 2 z r->0 +l S(Q) whose residues mod ef generate Then
4.4. 4.4.1.
f(T0'''''Tr) If of
(Tb)bcj , are basic nomomials for g(T 0 .... ,Tr~ I) considered as elements b br {T T : b E J' , 0 ~ b ~N-2} is the set of basic
r r
monomials
for
f.
case.
4.4.2.
More generally,
if
hi'r-i-l(x) 0
For example, tain i'r-l-i(x) h0 where if g
is homogeneous
0_<b. _<N-2} 3
]b I = b 0 + --" + br_ I
(r-i-1)N-m
Z c s
in
(1 + . . .
+ tN~2) r
68
4.4.3. nomial
Let
g(T0,...,Tr)
hi,r-i-2(y) = #{b~ J' : (b) = r-i-l} 0 Assume now that k = K. The exact sequence of Hodge structures
. . . . Hi(x) + Hi(X-y) + Hi-l(y)(-l) ~ Hi+I(x) + ... (dual to the compact cohomology sequence) determines the morphisms of the Hodge structures: i, : Hi-I(Y) (-i) Hi+I(x) which are obviously induced by the analogous morphisms H i ( ~ (Q'))(-I) Hi+2(~ (Q)) Applying the Weak Lefschetz theorem (4.2.2) we have that if i # 0, r - i. Thus for U = X-Y i, is an isomorphism
GrW(Hr-I(u)) = 0, 1
GrW(Hr-I(U))r = Hr-2(Y)(-I)0 Gr~_I(Hr-I(u)) = Hr-I(H)0 , where Hr-l(x)0 = Coker(Hr-3(y)(-l) + Hr-I(x)) Hr-2(Y)0 = Ker(Hr-2(y)(-l) ----+Hr(x)) For the Hodge numbers hP'q(u) if
= h0
hP'q(u) = 0, hi'r-i-l(u)
h~,r-i(u) = . i-l,r-i-l~ h0 (Y) = #{b e J' :(b) = r-i-l} where we recall that r Tb J' = {b~ ~ e0 : are basic monomials for g(T0,...,Tr_l) }.
The calculations of 4.4.3 presents an interest since the open affine subis isomorphic to the nonsingular affine variety in ~ r with the equa-
g ( x 0 .... ,Xr_ I) = i .
This variety plays an important part in the theory of critical points of analytic functions. The cohomology space Hr-I(u) is isomorphic to the space of
the vanishing cohomology of the isolated critical point function t = g(ZO,...,Zr_ I) ([18]). Its dimension
0 ~ Cr
of the analytic
= dimcC[T 0 ..... Tr_l]/@g = #j' . It can be seen from above as follows: dimeHr-l(u) =
=
and so we have to show that for any basic monomial r-I deg(r b) < E ('N-qi) = rN - [Q'I i=0 Let ~k = #{b ~ J' : deg(T b) = k}
E ~k z Xg(Z) = k
Tb
l(b) < r
or, equivalently,
Then ([i]) r-i Xg(Z) = i=0 It is clear that g(Z) zN-qi_l zqi-I is of degree n = rN-21Q' I and hence for k>n ~k = 0.
This proves the assertion above. Note that the Hodge numbers of follows hi'r-i-l(u) hi'r-i(g) = E(r-i-I)N-IQ ] < k < (r-i)N-IQ 1 ~k Hr-I(u) can be expressed in terms of ~k as
4.4.5.
Let
X = Vn(Q)
be a quasismooth
surface
(r=3).
We know that
h0'2(X) = h2'0(X) = #{a e J: (a) = I} = HN_jQ j hl'l(x) = #{a~ J :l(a) = 2} = ~2N_jQ 1 B2(X) = 2h2,0(X) + NI'I(x) = 2HN_jQ j + ~2N_IQ I , where E~ k z k = It is clear that r n (z N-qi -l)/(z qi -i) i=0 In case ID (Q) = ~3 we
70
[l]
V.I. AP~OL'D:
[2]
[3]
N. BOURBAKI:
[4]
F CATANESE:
Surfaces with
k 2 = pg = 1
Copenhagen Sum. Meeting on Algo Geom., Lect. Notes in Math. vol. 732, 1-29, Springer-Verlag, 1979.
[5]
V. DANILOV:
Uspehi Mat. Nauk 33, No. 2 (1978) Russian Math. Surveys, 33, No. 2
[6]
V. DANILOV:
Priloz 13, No. 2 (1979), 32-47 (in Russian; Engl. Transl. Anal. Appl. 13, No. 2 (1979), 103-114)
[7]
C. DELORME:
1975, 203-223
[8]
I. DOLGACHEV:
paration).
[9]
I. DOLGACHEV:
[i0]
R. FOSSUM:
The divisor class group of a Krull domain, Ergeb. Math. Bd. 74, 1973.
Springer-Verlag.
[Zl]
P. GRIFFITHS:
(1969), 460-495.
71
[12]
A. GROTHENDIECK:
[13]
A. GROTHENDIECK:
[14]
H. HAMM:
Die Topologie isolierten Singularltaten von vollstandigen D~rchschnitten koplexer Hyperfl~cher Dissertation.
V
Bonn. 1969.
[15]
A. HOVANSKII:
Funkt. Anal.: Priloz. 12, No 1 (1978), 51-61 (in Russian; Engl. Transl.: Funct. Anal. Appl., ]2, No 1 (1978), 38-46).
[16]
E. KUNZ:
Holomorphen Differentialformen auf algebraischen Variet~ten mit Singularit~ten. Manuscripta Math., 15 (1975), 91-108.
[17]
YU.
I. MANIN:
Engl. Transl. by
[18]
J. MILNOR:
[19]
S. MORI:
On a generalization of complete intersections, Journ. Math. Kyoto Univ., vol. 15, 3 (1975), 619-646.
[20]
P. ORLIK, P. WAGREICH:
93 (1971), 205-228.
[21]
H. PINKHAM:
Singularit~s
exceptionneles,
la dualite
6trange d'Arnold et
les surfaces K3, C.R. Acad. Sci. Paris, 284 (1977), 615-617.
[22]
H. POPP:
Lect. Notes
[23]
J. STEENBRINK:
[24]
A. TODOROV:
[25]
P. WAGREICH:
Trans. AMS,