Dynkin graphs
and
combinations
of
singularities
on some
algebraic
varieties
都立大理 卜部東介 (Tohsuke Urabe)
Inthis article Iwouldlike to explainmy recentresults
on
Dynkin graphs andglobal theory of singularitieson
algebraic varieties. Weassume
thatevery varietyis definedover
the complexfield $C$.
First, let
us
consider plane cubiccurves, thatis,curves
ofdegree 3 inthe 2-dimensional projective space. Itis easy to give thefollowing
classi-fication.
cuspidal
(normal definingpolynomial)
$y^{2}=x^{3}+a\kappa+b$ $y^{2}=x^{3}+x^{2}$ $y^{2}=x^{3}$
$(4a^{3}+27b^{2}\neq 0)$
$A_{2}$ $A_{1}$
$op$ $+\infty$
$2A_{1}$ $A_{3}$ $3A_{1}$
$O$
$D_{4}$
The first
one
is the smoothone.
Inspite thatwe
considercurves
over
the complexfield, above
we
havedrawnthepictures of real points. Becausea
smooth cubiccurve
over
thecomplex fieldisa
2-dimensional manifold ina
4-dimensional manifold $P^{2}(C)$, it is impossible to drawcorrectpictures. Anyway, wehave 9types of cubic
curves.
The last two havemultiple components, and theyare
not worthcalling cubiccurves, ifwe
treat themas
figures.Here Iexplain
an
importantconceptin singularity theory. We havea
series ofsingularitieswiththe
same
names as
Dynkin graphs withoutmultiple edges.
$A_{k}$
:
$x^{k+1}+y^{2}(+z^{2})=0$ $D_{l};x^{l-1}+xy^{2}(+z^{2})=0$$E_{6}:x^{3}+y^{4}(+z^{2})=0$ $E_{7}$
:
$x^{3}y+y^{3}(+z^{2})=0$ $E_{8};x^{3}+y^{5}(+z^{2})=0$When
we
consider singularities oncurves, we abbreviate the aboveterms$+z^{2}$
.
Whenwe
considersurface singularities, we add the above $terms+z^{2}$.
If
a
singularity is definedbyone
oftheseequations undera
suitablelocal coordinate,we
callthe singularityone
oftype$A,$$D$or
$E$.
Forexample, if
a curve
singularity isdefined bya
powerseries of2 variables $x,$$y$with cubic termsin thebeginning part, and ifthesum
ofcubic terms defines
a
homogeneouspolynomial withouta
multipleroot,then
we
can
make thispower
series into $x^{3}+xy^{2}$ bya
suitable coordinatechange. Therefore, thesingularity is of type $D_{4}$
.
The above 7-th cubic
curve
hasa
unique singularity andit is oftype$D_{4}$
.
We drawa
Dynkin gr.aphoftype $D_{4}$ beneath the 7-thcurve.
By thesame
methodwe can
associatea
Dynkingraph(possiblywithmultiplethe empty graph, $A_{1},$ $A_{2},2A_{1},$ $A_{3},3A_{1}$ and $D_{4}$
.
Now, perhaps you
can
notice that classification of cubiccurves
corre-spondsto subgraphs of $D_{4}.7$types of cubic
curves
have one-to-onecorrespondences with7 kinds of subgraphs of $D_{4}$
.
This is notan
acciden-talcoincide. Wecan
give theoretical explanation. Moreover,we
can
observe similarfactsfor
curves
andsurfaces oflow degree ina
projectivespace.
For suchobjects there existsa
common
law dominating possibleappearance
ofsingularities.By my study
so
farthe basic frame ofthecommon
theorycan
be explainedas
follows. Fora
given class of objects,a
basic Dynkin graphcan
be determined, anda
certainoperation by whichwe
can
makea new
Dynkin graph from
a
given Dynkin graph is defined. The set of allDynkin graphsmadeby the operationfrom the basic graphcoincideswith thesetof possible combinations ofsingularities.
Aclass ofobjects
basic Dynkingraph $arrow^{operation}$ {Dynkin graphsmade from the basicone}
I
1
possible combinationsofsingularities Needless to say, in the
case
of cubiccurves, thebasic graph $=D_{4}$, theoperation$=taking$
a
subgraph, andthe set ofgraphsmadefrom the basicone
consists of 7 graphs. $r$Bythe study
so
far, I know thatthe following geometrical objectsare
dominated by theory described by the above frame.
$l$ plane
curves
cubic, quartic, -, sextic6
space surfaces cubic, quartic$C$ deformationfibers ofasingularity
rational doublepoints $A_{k},$ $D_{l},$$E_{6},$ $E_{7},$$E_{8}$
simple elliptic singularities $P_{8},$$X_{9},$ $J_{10}$
9 of 14 triangle singularities
6quadrilateral singularities (singularitiesofrectangles)
Heretheconcept of
a
deformation fiber isto beexplained. Letthe origin. Let $g(x,y,z)$ be
an
arbitrarypowerseries. We consider the zero-locus defined ina
ballwitha
sufficiently smallradius $\epsilon>0$ with thecenter atthe originbythefollowing equality. $f(x,y,z)+tg(x,y,z)=0$
Here $t$ is
a
complex numberwith $0<|t|<\delta$, where $\delta$ isa
sufficientlysmallpositivenumber compared with $\epsilon$
.
This locus iscalleda
deformation fiberofthe singularity definedby $f$
.
Wecan
showthat thecombination ofsingularities onit is independentfrom the choice of $t$
.
Now,
we
havea
classification list ofhypersurface singularitiesdue toArnold. (Amold [1].) Itshouldbe remarked that the singularities in the above item “
Ct
deformation fibers...”appear
in the beginning partof his list.So far
we
considered singularities. Nextwe
would like to consider ellipticsurfaces,because theyare
also relatedto Dynkin graphs of type$A$,$D$
or
$E$.
A compactcomplex surface $X$ witha
morphism $\Phi$ toa curve
$C,$ $\Phi:Xarrow C$ is called
an
elliptic surface, if the inverse image $\Phi^{-1}(c)$ isaconnected smooth
curve
ofgenus 1 fora
general point $c\in C$.
Inthis article because of
a
technical reason,we
assume moreover
that there isa
section of $\Phi$, thatis,a
morphism $s:Carrow X$ such that thecomposition $\Phi s$ is the identity
on
$C$.
Aninverse image of
a
pointon
$C$ iscalleda
fiber inthiscase.
Possi-ble singular fibers inan
elliptic surfaceare
classifiedbyKodaira.(Kodaira [2].)
(1) An irreducible fiberis
one
ofthefollowing.(1.1)
an
ellipticcurve
$=a$curve
ofgenus 1(1.2)
a
rationalcurve
withan
$A_{1}$ singular point (It is isomorphictoa
plane nodalcubiccurve.)
(1.3)
a
rationalcurve
withan
$A_{2}$ singular point(It is isomorphic to aplane cuspidal cubic curve.)
(2)Areducible fiber is
a
union ofsmooth rationalcurves.
The graph of intersectionof these smooth rationalcurves
(Inthis graph the setofver-ticeshas one-to-one correspondence withthe set of smooth rational
curves.
If two rationalcurves are
disjoint, then the correspondingtwover-tices
are
notconnected. If theyhave intersection-number 1, then the correspondingtwo verticesare
connected by a single edge. If they have intersection-number2, then the corresponding verticesare
connectedbya
bold edge. Anintersection-number $\geq 3$
never
appears.) coincides withan
extendedDynkingraphoftype $A_{k},$ $D_{l},$ $E_{6},$$E_{7}$
or
$E_{8}$.
Therefore
we
can
associatea
Dynkin graphto eachsingular fiber ina
natural
manner.
(We associate theempty graphtoan
irreducible fiber.cor-respondingextendedDynkingraph.) Under this correspondence
we
can
apply the basic frame explained above also to describe possible combina-tions of singularfibers
on
elliptic surfaces. So farwe
have obtained results to thisdirectionfor rationalelliptic surfaces and K3 ellipticsur-faceswith
a
singular fiberof type $D_{4}$.
8
elliptic surfacesrational elliptic surfaces
K3 elliptic surfaces with
a
$D_{4}$-fiber.We divide the above mentioned objects into 3 gradesI, II, and III. Inthe firstgradethe operationis the simplest
one
–takinga
subgraph.Cubic
curves
and rational doublepoints fallinto this grade. Thebasic graph fora
rational double point coincides with the graph of thename
of the singularity.I. operation$=taking$
a
subgraphIn the second gradearathercomplicated operationcalled an elemen-tary transformation is introduced. In anelementary transformation, first
we
replace everycomponentto the associated extended Dynkin graph. In the second stepwe
choosea
propersubgraph.Anelementary transformation
1.
a
Dynkin$grapharrow an$ extendedDynkin graph.2. Choose
a
propersubgraph ofthe extended graph. In the second gradeas
the operationweuse
elementary transformations repeatedtwice:II. operation $=elementary$transformations repeatedtwice Quarticcurves, cubic surfaces, simple elliptic singularities and rational elliptic surfaces fallinto this grade. The basicgraphis
one
of type$E$.
Consider quartic
curves
as
anexample. Now, 4 lines intersecting atone
point is areducible quarticcurve., However, the singularityonit is notoftype$A,$ $D$
or
$E$.
We have toexclude thiscase.
Excluding only this case, the set of possible combinations of singularitiescoincides with the set of Dynkin graphs whichcan
be madefroma
basic graph $E_{7}$ byele-mentary transformationsrepeated twice.
{plane quartic
curves
without$\{*\}$
a
multiplecomponent}$\cup$
$E_{7}-$
{$possible$combinations ofsingularities}elementary transformation
In
an
elementary transformation starting from $E_{7}$we
can erase a
vertexas
in the figure below.
Under this choice
we
getthe graph $D_{6}+A_{1}$.
Wecan
repeatan
elementary transformationonce
more.
In the second transformation verticescan
be erasedas
in the figure below.The remaining classes fallinto the third grade. In the third gradethe thirdoperation called
a
tie transformationhas to beintroduced, and the number of the basic graphs is notnecessarily 1.III. operation $=elementary$ transformation&tie transformation repeated2 times
(Fourkinds ofcombinations, i.e., “elementary“ twice, “tie” twice,
“elememtary” after“tie”, and “tie” after“elementary”
are
allpermitted.)
basic graphs–possibly 2
or more
specialvertices (blackvertices) have tobe erased until the final stage.
For example forquartic surfaceswehave 9basic graphs. For sextic
curves
we
have 4basic graphs.Here I explainthe
case
ofone
of6 quadrilateral singularities called$J_{3,0}$
.
$J_{3,0}$ is definedby the followingequality.$J_{3,0}:x^{3}+ax^{2}y^{3}+y^{9}+bxy^{7}+z^{2}=0$ $(4a^{3}+27\neq 0)$
Weconsider only deformation fibers with$ADE$ singularities only. Inthis
The set ofallpossible combinations of singularitiesminus $3A_{3}+2A_{2}$
coincides with the setof Dynkin graphs with onlycomponentsof type$A$, $D$
or
$E$ made from $E_{8}+F_{4}$ by2 kinds of transformationsrepeated2 times.Here
we
need to givesome
explanationon
the Dynkin graph $F_{4}$,becauseithas
a
double edge andan arrow.
Now,every
vertex ina
Dynkin graph$COlTesponds$toa
vectorcalleda
rootinan
Euclideanspace. Ifthe graphhasan arrow
witha
multiple edge, it indicates that the length of roots atboth endsare
different. Thereforefor $F_{4}$ two vertices correspondto shorter roots. In fact the ratio oflengthis
as
in the following.Because length is different,
we can
replacethese vertices of shorterroots byblackvertices. Inour
theorywe use
the expressionincluding blackvertices forthe graph $F_{4}$
.
$F_{4}$ : $\ovalbox{\tt\small REJECT}$
HereIexplain the concept of tie transformations usingthis
case.
Atthe firststep ofa
tie transformationwe
makeevery
component to the corresponding extendedDynkingraph, andmoreover
weattachthe corresponding coefficient ofthemaximal root to each vertexof the extended graph.At the second step
we
choosetwo subsets $A$ and $B$ ofverticessatis-fying th$e$followingthree conditions$<a>,$ $<b>and<c>$
.
Thus the basic graph $E_{8}+F_{4}$ becomes the graphlike the following.
The conditions:
$<a>A\cap B=\emptyset$
$<b>$ Let $\overline{G}_{0}$ be
an
arbitrarycomponentofthe extendedgraph. Let $N$ bethe
sum
ofintegers attached to $B\cap\tilde{G}_{0}$. Let$n_{1},$ $n_{2},$$\cdots,$$n_{l}$.be all the
attached integers to $A\cap\tilde{G}_{0}$
.
Then, G.C.$D.(N, n_{1}, n_{2}, \cdots, n_{l})=1$.
$<c>7he$resulting graph afterthe third step isa Dynkingraph.togetherwith edgesissuing from them. Next
we
drawa new
white$ve$rtexcalled $\theta$ andconnecteach vertex in $B$to $\theta$by
an
edge.Under the choice of $A$ and $B$ in the above figure, we get the graph
$E_{7}+B_{6}$
.
This isan
example ofa
tie transformation.Now,
we
can
applya
transformationonce more.
Atthe second step of the secondtie transformationwe
getthefollowing graph undera
choice of $A$ and $B$.
Underthe above choice,
we
get 2$E_{7}$as
the resultofthe secondtietransformation. Note that $2E_{7}$ contains
no
black vertex.Here Iexplain the roleplayed by blackvertices. Ifwe startfrom
a
graph withablack vertex,we
can
make agraph witha
black vertex after the second transformation. However, obviously sucha
graph does not correspondtoa
combination ofsingularitieson
a
surface. Thus in the thirdgradewe
considerthe setof Dynkingraphs with only white vertex madefroma
basic graphby two transformation. This setcoincideswith theset of possible combination of singularitieson
surfaces inthegivenclass.
Therefore we
can
conclude that there isa
deformation fiber of $J_{3,0}$with $2E_{7}$ singularities.
{defornationfibers of $J_{3,0}$ with only$ADE$ singularities}
$u$
{possiblecombinations of singularities}-- $\{3A_{3}+2A_{2}\}$
$11$
{Dynkin$grap^{\text{ヒ}}hs$ made {Dynkingraphs with
$E_{8}+F_{4}arrow$ from $E_{8}+F_{4}$ } blackvertices} elementary transf.
2 times tie transf.
Here Iconclude the explanation ofmy theory. I do notexplain why these facts
can
hold.theory relatedtosimple algebraic
groups.
For the grade II and thegrade IIIwe
apply the theory of periods of algebraic surfaces. By periodswe
reduceour
problemto the theory of integral bilinear forms. With the aid of the theory ofreflectiongroupswe can
show theabovefacts.Itis slightly strange that forthe grade IIand III
no
Lie groups appearin thetheory at present, thoughthe grade I iscloselyrelatedto Lie groups. It is quite natural thatwe
expecta
relation betweenLiegroups
andthe grade II and III.I hope that
someone
inthe world–perhapsa
reader ofthis $article-$can
find outthe relation to Lie groups.References
[1] Arnold,V.:Localnormal forms of functions. Invent. Math. 35, 87-109 (1976) [2] Kodaira,K.:Oncompactanalytic surfaces II.Ann.of Math.77,563-626(1963) [3]Urabe,T.: On singularitiesondegenerate DelPezzo surfacesofdegree 1,2. Proc.
Symp. Pure Math. 40,(Part2)587-591 (1983)
[4]Urabe,T.: On quartic surfaces andsexticcurves withcertain singularities. Proc.
JapanAcad.,Ser. A 59,434-437(1983)
[5]Urabe, T.:On quartic surfaces andsexticcurves withsingularities oftype
$\tilde{E}_{8},$
$T_{2,3,7},$$E_{12}$.Publ.RIMS. Kyoto Univ. 20, 1185-1245 (1984)
[6]Urabe,T.: Dynkingraphs andcombinations of singularitiesonquartic surfaces. Proc. JapanAcad.,Ser.A61,266-269 (1985)
[7]Urabe,T.: Singulanities inacertainclassof quartic surfaces andsexticcurvesand
Dynkin graphs.Proc. 1984VancouverConf. Alg. Geom.,CMSConf. Proc. 6,
477-497 (1986)
[8]Urabe, T.:Classification of non-normal quartic surfaces. Tokyo J. Math. 9,265-295
(1986)
[9]Urabe,T.: Elementary
transformations.of
Dynkin graphs and singularities onquartic surfaces. Invent. Math. 87,549-572 (1987)
[10]Urabe, T.: Dynkin graphs and combinations of singularitiesonplanesextic
curves.In:Randell,R. (ed.) Singularities.Proceedings,Univ. Iowa1986
(Contemporary Math.,vol. 90,pp.295-316)Providence,RhodeIsland: Amer.
Math. Soc. 1989
[11]Urabe,T.: Tietransformations of Dynkin graphsandsingularitiesonquartic surfaces. Invent. Math. 100,207-230 (1990)