C OMPOSITIO M ATHEMATICA
J OSEPH L IPMAN
Double point resolutions of deformations of rational singularities
Compositio Mathematica, tome 38, n
o1 (1979), p. 37-42
<http://www.numdam.org/item?id=CM_1979__38_1_37_0>
© Foundation Compositio Mathematica, 1979, tous droits réservés.
L’accès aux archives de la revue « Compositio Mathematica » (http:
//http://www.compositio.nl/) implique l’accord avec les conditions géné- rales d’utilisation (http://www.numdam.org/conditions). Toute utilisation commerciale ou impression systématique est constitutive d’une infrac- tion pénale. Toute copie ou impression de ce fichier doit contenir la présente mention de copyright.
Article numérisé dans le cadre du programme Numérisation de documents anciens mathématiques
http://www.numdam.org/
37
DOUBLE POINT RESOLUTIONS OF DEFORMATIONS OF RATIONAL SINGULARITIES
Joseph Lipman1
1
Let
Yo
be a normalalgebraic
surface over analgebraically
closedfield
k,
and letfo : Zo- Yo
be the minimal resolution ofsingularities.
There is a
unique
way to factorfo
asZo Xo - Yo
with a proper birational go and a normal surfaceXo,
such that a reduced irreduciblecurve C on
Zo
which blows down to apoint
onYo already
blowsdown on
Xo
iff C isnon-singular
and rational with self-intersectione2
= -2(cf. [1,
p.493, (2.7)];
or, forgreater generality, [7,
p.275, (27.1)]).
The
singularities
ofXo
are all Rational Double Points.Xo,
which isuniquely
determinedby Yo,
will be called the RDP-resolution ofYo.
Suppose
now thatYo
is affine and hasjust
onesingularity
y, ybeing
rational. There is a
conjecture
of Wahl[9],
andBurns-Rapoport [2, 7.4]
to the effect that the Artin
component
in the(formal
orhenselian)
versal deformation space of y is obtained from the deformation space ofZo by factoring
out a certain Coxeter group.According
to Wahl[10],
this wouldimply
that the Artincomponent
issmooth,
and is in factformally
identical with the deformation space of
Xo (at
least afterthings
aresuitably localised).
Wahl also shows how theconjecture
reduces to thestatement that under
"blowing down",
the set of first order infinitesimal deformations ofXo
mapsinjectively
into that ofYo.
Our purpose here isto outline a
proof
that thisinjectivity (and
so theconjecture)
does indeed hold.Wahl has
previously
established somespecial
cases of the con-jecture by showing
that a certaincohomology
group vanishes[10].
1am
grateful
to him for all the information and motivation he hasprovided.
’AMS(MOS) subject classifications (1970). Primary 14J 15, 32C45.
Research supported by the National Science Foundation.
Sijthoff & Noordhoff International Publishers - Alphen aan den Rijn
Printed in the Netherlands
38
2
Let g : Y - S be a flat map of finite
type
of noetherianschemes,
such that for each s E S the fibre
YS
is a normal surfacehaving only
rationalsingularities.
We assume - forsimplicity only - that
every closedpoint
of Y maps to a closedpoint
of S whose residue field isalgebraically
closed.By
an RDP-resolution of Y- S is meant aproper map
f :
X - Y such thatgo f:
X --> S isflat,
and for each s themap
fs : Xs --. YS
is the RDP-resolution ofYS (see above).
THEOREM: Given Y - S as
above,
there exists at most one(up
toY -isomorphism)
RDP-resolution.(For
S =Spec(k [t]lt2)
we get the above-indicatedinjectivity
state-ment.)
In fact we show the
following:
Let U C Y be the open set where g is smooth and let i : U - Y be the inclusion map. Set
(with f
the canonicalmap).
In other words X is the scheme-theoretic closureof
U in theprojective
bundleP(w) [4, II, (4. 1. 1)].
EXAMPLE
(Suggested by Riemenschneider).
LetYo
be the coneover a
non-singular
rational curve ofdegree
4 inp4 (say
over thecomplex
numbersC).
The versaldeformation
of the vertex has aone-dimensional non-Artin
component,
foundby
Pinkham: the cor-responding
deformation is Y - S =Spec(C[t]),
where Y CCS x
S isthe zero-set of the 2 x 2 minors of
Hère,
of course, no RDP-resolution can exist. Onecomputes
that bJn isisomorphic
toJn,
where J is thefractionary
idéalgenerated by
singular,
but not an RDP-resolution of Y.(The
fibre over t = 0 hastwo
components!)
3
After
outlining
theproof
of somepreliminary
facts fromduality theory (Lemma 1),
we prove the Theorem in Lemmas 2 and 3 below.Let
f :
X --> Y be an RDP-resolution of g : Y - S.LEMMA 1:
(cf. [5,
p.298]
or5’, Prop. 9) (i)
Since g : Y ---> S isflat,
withCohen-Macaulay fibres,
the relativedualizing complex g’Os
hasjust
onenon-zero
cohomology sheaf
bJy/s, and wyls isflat
over S. Since gof
isflat,
with Gorenstein
fibres,
thesimilarly defined Ox-module
bJx/s is invertible.(ii)
For any map S’-->S,
with S’noetherian, if
Y’ = Yx s S’
andPROOF: We may assume
homomorphic image
of apolynomial ring 383,
p.388]).
Then the firstpart
of(i)
states thatand
(ii)
says thatis an
isomorphism.
(i*)
and(ii*), together
withNakayama’s lemma,
reduce theproof
ofthe last
assertion
in(i)
to the well-known case where S is thespectrum
of a field[5,
p.296, Prop. 9.3].
We first show that B has
homological
dimension n - 2 over R. Forthis we may
replace R by
its localization at anarbitrary
maximal ideal3M,
and Aby
itslocalization
atm nA [8,
p.188,
Thm.11].
Let m bethe maximal ideal of
A,
letR
=RQ?)A(Alm)
andB
=BQ?)A(Alm),
sothat
B
is a normaltwo-dimensional homomorphic image
of theregular n-dimensional
localring R,
and soB
hashomological
dimen-40
sion n - 2 over
R. (What
matters here andsubsequently
is thatB
isCohen-Macaulay.)
Let K be the residue field of R. Since R and B areA-flat,
we have for anyR-projective (hence A-flat)
resolution P. of B that thehomology Hj(P. Q9A (Alm))
=Tort(B, Alm)
= 0for j
>0,
i.e.P.
Q9A (Alm)
is anR-projective
resolution ofB, whence,
for alli,
and our assertion follows from
[8,
p.193,
Thm.14].
Thus B has a
finitely generated R-projective (hence A-flat)
resolu- tionLet
Q.
be thecomplex
withFor any
A-algebra
A’ and any i we have(For
the secondequality
cf. theproof
of(#) above.)
Now
(i*)
results from thefollowing
Lemma.(We
assumeagain,
aswe may, that A and R are
local,
and let m be the maximal ideal ofA.)
LEMMA la: Let
Q.
be anA-flat complex of finitely-generated
R-modules,
withQi
= 0for
i0,
and such that thehomology H;(Q. ®A (A/ttt))
= 0for all j
> 0. ThenHo(Q.)
isA-flat,
andHj(Q.) =
0
forall j>0.
The
proof
is left to the reader.Finally, applying [4, III’, (7.3.1)(c)
and(7.3.7)]
to thehomological
functor T. of A-modules M
given by
(Q.
asabove)
we see that for every A-module M andevery i,
there isa natural
isomorphism
LEMMA 2: Let L be the invertible
Ox-module
toxls. Then :(i)
I£ is veryample for f.
(ii)
For every n >0,
the canonical mapjective.
PROOF: We first show that
R’f*(Cx)
= 0: for any closedpoint
y ofY,
let My be the maximal idealof O yy;
then for allr 0, MrOXIMr+,C
is an
Cf-i(y)-module generated by
itsglobal sections;
sinceY, (s
=g(y))
has rationalsingularities,
thereforeH’(Cf-,(y»
=0,
and sincef -’(y)
hasdimension 1,
weget H (, y r CXIM y r+l)
=0;
conclude with[4,
III, (4.2.1)].
Now
by [7,
p.220, (12.1)
and p.211, proof
of(7.4)]
it will beenough
to show for any reduced irreducible curve
C C f -’(y)
that(L. C) (the degree
of 0pulled
back toC)
is > 0. Lemma 1(ii)
reduces us to thecase S =
Spec(k), k
analgebraically
closed field. Let p : Z --> X be aminimal resolution of
singularities,
and letC
be thecomponent
ofp-l(C)
which maps onto C. Then(L. C) = (p-lL.C).
Butp-l T
is adualizing
sheaf on Z[1,
p.493, (2.7)];
since(by
definitior of RDP-resolution) C
is not anonsingular
rational curve withC2 >- -2,
there- fore(p-lL.C)
> 0.Q.E.D.
PROOF: Let U C Y be as
before,
andlet j : f -1( U) X
be theinclusion map.
f
induces a proper mapf -’(U) -->
U which is fibrewise(over S)
anisomorphism;
hencef -’(U) -->
U is anisomorphism.
Nowand 0n
is obtainedby applying f*
to the natural map0© - j*j*Yo".
For the
injectivity,
it suffices that the(X - f-l( U)) depth
of L be >1[3,
1.9 and3.8]. [4, IV, (11.3.8)]
and(ii)
of Lemma 1 above enable usto
verify
this fibrewise(over S);
in otherwords,
we needonly
check42
the
simple
case where S =Spec(k), k
a field.Similarly
we see that the( Y - U)-depth
of wYis is>2,
and conclude thatREMARK: For S =
Spec(k[t]lt2),
Wahl has aproof
of Lemma 3which avoids
duality theory [10, §2].
REFERENCES
[1] M. ARTIN: Some numerical criteria for contractibility of curves on algebraic
surfaces. Amer. J. Math. 84 (1962) 485-496.
[2] D. BURNS, M. RAPOPORT: On the Torelli problem for Kählerian K-3 surfaces.
Ann. Sci. Ecole Norm. Sup. 8 (1975) 235-274.
[3] A. GROTHENDIECK: Local Cohomology. Lecture Notes in Math. no. 41, Springer- Verlag, 1967.
[4] A. GROTHENDIECK, J. DIEUDONNÉ: Éléments de Géometrie Algébrique II, III, III’, IV; Publ. Math. IHES nos. 8, 11, 17, 28 (1961-1966).
[5] R. HARTSHORNE: Residues and Duality: Lecture Notes in Math. no. 20, Springer- Verlag, 1966.
[5’] S. KLEIMAN: Relative duality for quasi-coherent sheaves (Preprint).
[6] H. LAUFER: On rational singularities. Amer. J. Math. 94 (1972) 597-608.
[7] J. LIPMAN: Rational singularities. Publ. Math. IHES no. 36 (1969) 195-279.
[7’] J. LIPMAN: Desingularization of two-dimensional schemes. Annals of Math. 107 (1978) 151-207.
[8] D.G. NORTHCOTT: An introduction to homological algebra, Cambridge University Press, London, 1960.
[9] J. WAHL: Local cohomology groups for resolutions of singularities. Symposia in
Pure Math. vol. 30 (1977) 91-94.
[10] J. WAHL: Simultaneous resolution of rational singularities. Comp. Math. 38 (1979) 43-54.
(Oblatum 6-VI-1977 & 1-IX-1977) Joseph Lipman
Department of Mathematics Purdue University
W. Lafayette, Indiana 47907