Professional Documents
Culture Documents
SEUNG-JO JUNG
Abstract. A 3-fold terminal quotient singularity X = C3 /G admits the economic resolution Y X, which is close to being
crepant. This paper proves that the economic resolution Y is isomorphic to a distinguished component of a moduli space of certain
G-equivariant objects using the King stability condition introduced by Kedzierski
[11].
Contents
1. Introduction
2. G-bricks and moduli spaces of G-constellations
2.1. Moduli spaces of G-constellations
2.2. G-prebricks and local charts of M
2.3. G-bricks and the biratioanl component Y
3. Weighted blowups and economic resolutions
3.1. Terminal quotient singularities in dimension 3
3.2. Weighted blowups and round down functions
3.3. Economic resolutions
4. Moduli interpretations of economic resolutions
4.1. G-bricks and stability parameters for 1r (1, r 1, 1)
4.2. G-bricks for 1r (1, a, r a)
4.3. Stability parameters for S(r, a)
4.4. Main Theorem
5. Kedzierskis
GIT chamber
1
6. Example: type 12
(1, 7, 5)
References
1
4
4
6
11
12
12
14
17
18
18
19
25
27
28
32
37
1. Introduction
The motivation of this work stems from the philosophy of the McKay
correspondence, which says that if a finite group G acts on a variety M,
then a crepant resolution of the quotient M/G can be realised as a
moduli space of G-equivariant objects on M.
2010 Mathematics Subject Classification. 14E15, 14E16, 14L24.
Key words and phrases. the McKay correspondence, terminal quotient singularities, economic resolutions.
1
S.-J. JUNG
Acknowledgement. This article is part of my PhD thesis in the University of Warwick [8]. I am very grateful to my advisor Miles Reid
for sharing his views on this subject and many ideas. I would like to
thank Diane Maclagan, Alastair Craw, Hiraku Nakajima, and Timothy
Logvinenko for valuable conversations. I thank Andrew Chan and Tom
Ducat for their comments on earlier drafts.
S.-J. JUNG
Definition 2.1. A G-constellation on C3 is a G-equivariant coherent sheaf F on C3 with H0 (F ) isomorphic to the regular representation C[G] of G as a C[G]-module.
L
The representation ring R(G) of G is G Z . Define the GIT
stability parameter space
= HomZ (R(G), Q) (C[G]) = 0 .
xi yi+1 yi xi+2 ,
(2.3)
xi zi+1 zi xi+3 ,
y z
i i+2 zi yi+3 .
Q
Let = (1, . . . , 1) Zr0 . The group GL() := iG C = (C )r acts
on Rep G via change of basis. For a parameter , define the GIT
quotient with respect to
Rep G // GL() := Repss
G/ GL()
ss
parametrising closed GL()-orbits in Repss
G where Rep G denotes
the -semistable locus in Rep G.
y
y
C3 /G M0
(i) Y is a not-necessarily-normal toric variety which is birational
to the quotient variety C3 /G.
(ii) Y is projective over the quotient variety C3 /G.
Definition 2.6. The unique irreducible component Y in Theorem 2.5
is called the birational component of M .
Since Craw, Maclagan and Thomas [3] constructed Y as GIT quotient of a reduced affine scheme, it follows that Y is reduced.
Remark 2.7. Since T = (C )3 acts on C3 , the algebraic torus T acts
on the moduli space M naturally. Fixed points of the T-action play
a crucial role in the study of the moduli space M .
S.-J. JUNG
z
y
and wt1 (y 3 ) =
z2
.
y2
n m
wt (n m )
0,
m , n M 0 .
Observe that:
(i) M 0 M S(),
(ii) () + ,
(iii) S() () M .
for some
An arbitrary generator of S() is of the form wtnm
(nm )
n M 0 , m . We may assume that n m 6 . In particular,
n 6= 1. Since n has positive degree, there exists f {x, y, z} such
that f divides n, i.e. nf M 0 and deg( nf ) < deg(n). Let m denote
wt ( nf m ). Note that
n
wt (f m ) = wt (f m ) = wt (n m ).
f
Thus
n
m
f wt ( nf m )
n m
f
=
wt (n m )
wt ( nf m )
wt (n m )
n
m
f m
f
=
.
n
wt ( f m ) wt (f m )
By induction on the degree of monomial n, the assertion is proved.
The set B() in the lemma above is called the Border bases of . As
B() is finite, the semigroup S() is finitely generated as a semigroup.
Thus the semigroup S() defines an affine toric variety. Define two
affine toric varieties:
U() := Spec C[S()],
U () := Spec C[() M].
S.-J. JUNG
given by f 7 f 1.
x m = x wt (x m ),
y m = y wt (y m ),
z m = z wt (z m ),
x ywt(xm ) y xwt(ym ) ,
(2.18)
x zwt(xm ) z xwt(zm ) ,
y z
wt(ym ) z ywt(ym ) .
where I = the quadrics in (2.18), f 1 f () . By King [12],
the affine scheme D() is an open set of M containing the point
10
S.-J. JUNG
f
corresponding to C(). More precisely, consider an affine open set U
in Rep G, which is defined by f to be nonzero for all f (). Note
f is GL()-invariant and that U
f is in the -stable locus. Since
that U
ss
the quotient map Rep G M is a geometric quotient for generic ,
f ]GL() is an affine open subset of
from GIT [15], it follows that Spec C[U
M . On the other hand, setting f () to be 1 for all f () gives
f ]GL() .
a slice of the GL()-action. Thus D() is isomorphic to Spec C[U
Remark 2.20. The affine open subset D() of the moduli space M
parametrises G-constellations whose basis is .
D() M
where the vertical morphisms are closed embeddings.
Proof. Let us assume that the G-constellation C() is -stable. Let p
be an arbitrary point in U() and G a submodule of C()p . By the
definition of C()p , there is a submodule G of C() whose support is
the same as G. Since C() is -stable, (G) = (G ) > 0. Thus C()p
is -stable.
Since there is a C-algebra epimorphism from C[D()] to C[S()]
given by
f m
f 7
wt (f m )
for f {x , y, z }, it follows that U() is a closed subscheme of D().
As Craw, Maclagan, and Thomas [3] proved that the birational component Y is a unique irreducible component of M containing the
torus T = (C )3 /G, it follows that Y D() is a unique irreducible
component of D() containing the torus T .
The morphism U() D() M induces an isomorphism between the torus Spec C[M] and the torus T of Y by Lemma 2.16 (iv).
Note that U() is in the component Y D() because U() is a closed
subset of D() containing T . Since both U() and Y D() are reduced and of the same dimension, U() is equal to Y D(). Thus
there exists an open immersion from U() to Y .
11
[8], there is another proof using the language of the McKay quiver
representations.
12
S.-J. JUNG
(ii) The normalization of Y is isomorphic to the normal toric variety whose toric fan consists of the 3-dimensional cones ()
for S and their faces.
Proof. Let G be the group of type 1r (1 , 2 , 3 ). Consider the lattice
L = Z3 + Z 1r (1 , 2 , 3 ).
Let Y be the birational component of the moduli space of -stable
G-constellations and Y the normalization
of Y . Let Y denote the
S
not-necessarily-normal toric variety S U(). Define the fan in LR
whose maximal cones are () for S. Note that the corresponding
toric variety Y := X is the normalization of Y .
By Proposition 2.21, there is an open immersion : Y Y . From
Proposition 2.22, it follows that the image (Y ) contains all torus
fixed points of Y . The induced morphism : Y Y is an open
embedding of normal toric varieties with the same number of torus
fixed points. Thus the morphism should be an isomorphism. This
proves (ii).
To show (i), suppose that Y \(Y ) is nonempty so it contains a torus
orbit O of dimension d 1. Since the normalization morphism is torus
equivariant and surjective, there exists a torus orbit O in Y
= Y
of dimension d which is mapped to the torus orbit O. At the same
time, from the fact that Y is the normalization of Y and that the
normalization morphism is finite, it follows that the image of O is a
torus orbit of dimension d, so the image is O. Thus O is in (Y ), which
is a contradiction.
Corollary 2.25. With the notation as in Theorem 2.24, Y is a normal
toric variety if and only if S() = () M for all S.
3. Weighted blowups and economic resolutions
Let G GL3 (C) be the finite subgroup of type 1r (1, a, r a) with r
coprime to a, i.e.
G = hdiag(, a , ra ) r = 1i.
3.1. Terminal quotient singularities in dimension 3. In this section, we collect various facts from birational geometry. Most of these
are taken from [17].
Definition 3.1. Let X be a normal quasiprojective variety and KX
the canonical divisor on X. We say that X has terminal singularities
(resp. canonical singularities) if it satisfies the following conditions:
13
14
S.-J. JUNG
2 = Cone(e1 , v, e3 ),
3 = Cone(e1 , e2 , v).
3
v = 1r (1, a, r a)
e3
e2
2 := xy a ,
2 := y a ,
2 := y
ar
a
z.
3 := xz ra ,
3 := yz ra ,
3 := z ra .
15
1 := x r ,
1 := x r y,
1 := x
ra
r
z.
v6 = 71 (6, 4, 3)
v5 = 71 (5, 1, 6)
v3 = 71 (3, 2, 5)
v4 = 17 (4, 5, 2)
v2 = 71 (2, 6, 1)
3
v1 =
e3
1
(1, 3, 4)
7
e2
2 (xm1 y m2 z m3 ) = 2m1 2 r
16
S.-J. JUNG
(1, a, r a) is defined by
m3
1 m1 + ar m2 + ra
r
1 (xm1 y m2 z m3 ) = 1 r
7 k (),
17
Proof. Here we prove the assertion for the left round down function.
Let , , denote the eigencoordinates for the G2 -action. Let k be a
monomial in M2 and g {, , }.
Consider the case where g = . Since 2 is surjective, there exists
m = xm1 y m2 z m3 M such that 2 (m) = k. If k = 2 (z m), then
we are done.
Suppose k 6= 2 (z m). This means that
1
a
ra
ra
1
a
ra
m1 + m2 +
m3 +
m1 + m2 +
m3 + 1.
r
r
r
r
r
r
r
There is a positive integer l0 3 such that 2 ( m
) = k for all 0 l l0
yl
6= k. Since 2 (z yml0 ) = k, the assertion follows.
with 2 ylm
0 +1
For the other cases, we can prove the assertion similarly.
3.3. Economic resolutions. By the fact that the quotient variety
X = C3 /G has terminal singularities, X does not admit crepant resolutions. However X has a certain toric resolution introduced by
Danilov [5] (see also [17]).
Definition 3.11. Let G GL3 (C) be the group of type 1r (1, a, r a).
For each 1 i < r, let vi := 1r (i, ai, r ai) L where denotes the
residue modulo r. The economic resolution of C3 /G is the toric variety
obtained by the consecutive weighted blowups at v1 , v2 , . . . , vr1 from
C3 /G.
Proposition 3.12 (see [17]). Let : Y X = C3 /G be the economic
resolution of C3 /G. For each 1 i < r, let Ei denote the exceptional
divisor of corresponding to the lattice point vi .
(i) The toric variety Y is smooth and projective over X.
(ii) The morphism satisfies
X i
Ei .
KY = (KX ) +
r
1i<r
In particular, each discrepancy is 0 <
i
r
< 1.
From the fan of Y , we can see that Y can be covered by three open
sets U2 , U3 and U1 , which are the unions of the affine toric varieties
corresponding to the cones on the left side of, the right side of, and
below the vector v = 1r (1, a, r a), respectively. Note that U2 and U3
are isomorphic to the economic resolutions for the singularity of type
1
1
(1, r, r) and of type ra
(1, r, r), respectively.
a
3This
18
S.-J. JUNG
Example 3.14. Let G be the group of type 17 (1, 3, 4) as in Example 2.10. The fan of the economic resolution of the quotient variety
C3 /G is shown in Figure 3.3.
e1
1
2
v6
v5
v4
v3
v2
e3
v = 17 (1, 3, 4)
e2
resolution of C3 /G.
19
1
of type r (1, a, r a) with a = 1 or r 1. Then G-Hilb C3 is isomorphic
to the economic resolution of the quotient variety C3 /G.
For each 1 i < r, set vi = 1r (i, r i, i). Set v0 = e2 and vr = e3 .
The toric fan corresponding to G-Hilb C3 consists of the following 2r1
maximal cones and their faces:
i = Cone(e1 , vi1 , vi )
for 1 i r,
for 1 i r 1.
i = {1, y, y 2, . . . , y i1, z, z 2 , . . . , z ri }
r+i = {1, y, y 2, . . . , y i1, x, x2 , . . . , xri }
for 1 i r,
for 1 i r 1,
Example 4.4. Let G be the finite group of type 31 (1, 2, 1) with eigencoordinates , , . Set v1 = 31 (1, 2, 1) and v2 = 13 (2, 1, 2). Recall that the
economic resolution Y of X = C3 /G can be obtained by the sequence
of the weighted blowups:
2
Y Y1 X,
where 1 is the weighted blowup with weight (1, 2, 1) and 2 is the
toric morphism induced by the weighted blowup with weight (2, 1, 2).
The fan corresponding to Y consists of the following five 3-dimensional
cones and their faces:
1 = Cone(e1 , e2 , v1 ), 2 = Cone(e1 , v1 , v2 ), 3 = Cone(e1 , v2 , e3 ),
4 = Cone(e3 , e2 , v1 ), 5 = Cone(e3 , v1 , v2 ).
The following
1 = {1, , 2}, 2 = {1, , },
4 = {1, , 2}, 5 = {1, , }.
are their corresponding G-bricks.
3 = {1, , 2},
20
S.-J. JUNG
toric varieties corresponding to the cones on the left side of, the right
side of, and below the lattice point v = 1r (1, a, r a), respectively.
Proposition 4.5. For k = 1, 2, 3, let be a Gk -brick. Define
:= m M k (m) .
The set is a G-brick with S() = S( ).
7 k ,
k : xi
2 (xi )
i.e. xi 2k(x i ) is in . To show uniqueness, assume that two monomials
m, m of the same weight are mapped into . As k (m) and k (m ) are
of the same weight, we have k (m) = k (m ) . From Lemma 3.7,
m
m
k (m) = k m = k (m ) ,
m
m
k (n n m ) k (n m )
k (m ) ,
k (n m )
k (m )
)
it follows that k (n m ) = kk(nm
k (m ) is in because is a
(m )
Gk -prebrick. This proves that is a G-prebrick.
It remains to prove that S() = S( ). Note that S() is generated
by wtnm
for n M 0 and m . Let n be a genuine monomial in
(nm )
M 0 and m an element in . Let k denote k (m ) . Define k to
)
. From the definition of the round down functions, we know
be kk(nm
(m )
is G-invariant,
that k is a genuine monomial in , , . Since wtnm
(nm )
it follows that
k (nm )
k (m )
k (n m )
n m
k k
= k (m )
=
=
wt (n m )
wt (k k )
k wt (n m )
k wt (n m )
21
Proposition 4.8. Let G be the group of type 1r (1, a, r a) with r coprime to a. Let max be the set of maximal cones in the fan of the
economic resolution Y of X = C3 /G. Then there exists a set S(r, a)
of G-bricks such that there is a bijective map max S(r, a) sending
to satisfying S( ) = M. In particular, U( ) is isomorphic
to the smooth toric variety U = Spec C[ M] corresponding to .
Proof. From Section 4.1, the assertion holds when a = 1 or r 1. We
use induction on r and a.
Let be a 3-dimensional cone in the fan of the economic resolution
Y of X = C3 /G. For v = 1r (1, a, r a), we have three cases:
(1) the cone is below the vector v.
(2) the cone is on the left side of the vector v.
(3) the cone is on the right side of the vector v.
Case (1) the cone is below the vector v. Since there is a unique
3-dimensional cone below v, = Cone(v, e2 , e3 ). Consider the central
round down function 1 of the weighted blowup with weight (1, a, ra).
For m = xm1 y m2 z m3 M, note that
1 (m) = 1 if and only if m2 = m3 = 0 and 0
m1
r
< 1.
2
r1
The set := 1
} is a G-prebrick with the
1 (1) = {1, x, x , . . . , x
22
S.-J. JUNG
Case (2) the cone is on the left side of v. From the fan of the
economic resolution, it follows that U2 is isomorphic to the economic
resolution Y2 of a1 (1, r, r) with eigencoordinates , , . There exists
a unique 3-dimensional cone in the toric fan of Y2 corresponding to
. Let G2 be the group of type a1 (1, r, r). Note that a is strictly less
than r so that we can use induction.
Assume that there exists a G2 -brick with S( ) = ( ) M. By
Proposition 4.5, there is a G-brick with S() = S( ) = M.
Case (3) the cone is on the right side of v. The case where the
cone is on the right side of v can be proved similarly.
Definition 4.9. A G-brick in S(r, a) described above is called a
Danilov G-brick.
Proposition 4.10. With the notation as is in Proposition 4.5, we have
D( )
= D(). Moreover we have a commutative diagram
U( ) U()
y
y
D( ) D()
with the vertical morphisms closed embeddings. Therefore for a G-brick
S(r, a), we have U() = D()
= C3 .
Proof. Let be a G-brick and the corresponding Gk -brick. Let
, , denote the eigencoordinate for the Gk -action. From (2.19), the
coordinate rings of the affine schemes D(), D( ) are
C[D()] = C[x , y , z G ] I ,
C[D( )] = C[ , , Gk ] I
f () and the
where the ideal
I
is
the
quadrics
in
(2.18),
f
ideal I is the commutative relations, g 1 g ( ) .
By Lemma 3.10, we have an algebra epimorphism
: C[x , y , z G ] C[D( )] f 7 k()
23
yf
C m
yf
C wt (f m ) C wt (f m ),
it suffices to show that n wt (f m ) is a base, which implies that n
induces a linear map equal to 1 on wt (f m ). Since
wt (f m )
k (n wt (f m )) = k f m
f m
wt (f m )
= k (f m )
f m
wt (f m )
= k (f m )
= k wt (f m ) ,
f m
the monomial n wt (f m ) is in and is a base.
Now define the algebra morphism : C[ , , Gk ] C[D()]
by (g ) = f for g { , , } so that (f ) = g . Since the generators of I are in ker , induces : C[D( )] C[D()].
To show is surjective, we prove that generators f are in the image
m )
is of degree 1, f is in the image of
of . For f such that kk(f(m
)
by definition. By Lemma 4.11 below, it follows that is surjective.
Since and are the inverses of each other, D() is isomorphic to
D( ). Note that U() = D()
= C3 for S(r, 1) from Section 4.1.
Using induction, we get D() = C3 for S(r, a).
Lemma 4.11. In the situation as in Proposition 4.10, define
n
o
m )
S := f {x , y, z } kk(f(m
is
of
degree
1
.
)
m )
is of degree 2 for some f {x , y , z }, then f can be
If kk(f(m
)
written as a multiple of some elements in S modulo I .
Proof. We prove this for the left round down function 2 . Note that
k (ym )
is of degree 1 for all G . Thus y s are in S.
k (m )
m )
is of degree 2 with m = xm1 y m2 z m3 . Then
Suppose that kk(f(m
)
the monomial f is either x or z. In the case where f = z, this means
that
a
ra
1
a
ra
a
1
m1 + m2 +
m3 m1 + m2 +
m3 .
r
r
r
r
r
r
r
2 ( m
) = 2 (m ) for all 0 l l with 2 yml+1 6= 2 (m ). Note
y l
24
S.-J. JUNG
2 (f m )
commutation relations
C m
yf
m
.
yl
y l
C m
yf
y l
C wt (f m ) C wt (f m ),
since y l induces a linear map on C m set to be 1, we have
l
l
f f y(
) y() f
mod I .
Example 4.12. Let G be the group of type 71 (1, 3, 4) as in Example 2.10. The fan of the economic resolution of the quotient variety is
shown in Figure 3.3.
e1
e1
1 2
e1
1
(1, 2, 1)
3
e3
1
(1, 2, 1)
3
1
(1, 3, 1)
7
e2
e3
1
(1, 3, 1)
4
e2
1
(1, 3, 1)
7
e3
1
(1, 3, 1)
4
e2
Note that the left side of the fan corresponds to the economic resolution
for the quotient singularity of type 13 (1, 2, 1), which is G2 -Hilb C3 , where
G2 is of type 31 (1, 2, 1). Let , , denote the eigencoordinates. Let 1
be the cone in the fan of G2 -Hilb C3 which corresponds to 1 . Observe
that the corresponding G2 -brick is
1 = 1, , 2 .
Since the left round down function 2 is
1
2 (xm1 y m2 z m3 ) = m1 7 m1 + 7 m2 + 7 m3 m3 ,
25
On the other hand, the right side of the fan corresponds to the economic resolution of the quotient variety 14 (1, 3, 1) which is G3 -Hilb C3 ,
where G3 is of type 41 (1, 3, 1) with eigencoordinates , , . Let 2 be
the cone in the fan of G2 -Hilb C3 which corresponds to 2 . Observe
that the corresponding G3 -brick is
2 = 1, , 2, 3 .
Since the right round down function 3 is
3 (xm1 y m2 z m3 ) = m1 m2 7 m1 + 7 m2 + 7 m3 ,
the G-brick corresponding to 2 is
def
2 = xm1 y m2 z m3 M 2 (xm1 y m2 z m3 ) 2
2
3
3
= 1, z, y, y 2, yz , yz , yz 2 .
4.3. Stability parameters for S(r, a). Let G GL3 (C) be the finite
subgroup of type 1r (1, a, r a) with r coprime to a. We may assume
2a < r. Let G2 and G3 be the groups of type a1 (1, r, r) and of type
1
(1, r, r), respectively.
ra
Given stability conditions (2) for Danilov G2 -bricks and (3) for
Danilov G3 -bricks, take a GIT parameter P satisfying the following system of linear equations:
P
(2)
P ()
for all G2 ,
() =
2 ()=
P
(4.13)
(3)
P () for all G3 .
(
)
=
3 ()=
1 if 0 wt() < a ,
(4.14)
() = 1
if r a wt() < r,
0
otherwise.
P
Observe that
() = 0 for all Gk 4. For a sufficiently large
k ()=
:= P + m.
P
addition, if any satisfies that k ()= () = 0 for all G
k and
k = 2, 3, then must be a constant multiple of . This also explains the existence
of a solution P for (4.13).
4In
26
S.-J. JUNG
1
3
1
1
P (0 ) + P (3 ) + P (6 ),
P (1 ) + P (4 ),
P (2 ) + P (5 ),
P (0 ) + P (4 ),
P (1 ) + P (5 ),
P (2 ) + P (6 ),
P (3 ).
Take P = (1, 3, 3, 1, 2, 2, 2) as a partial solution. For the parameter = (1, 1, 1, 0, 1, 1, 1), define = P + m for large m.
Consider the following G-brick
n
o
2
2
= 1, y, y 2, z, yz , zy , yz 2 .
2
Let F be the submodule of C() with basis A = z, yz , zy . Note
that (F ) > 0 and
n z z2 z2 o
1
(A)
= z, y , y , y2 ) A.
2
2
Thus (F ) is positive for large enough m. More precisely,
(F ) = 3 m + (2 + m) + (2 + m) = m 1
is positive if m > 1.
Onthe other
hand, consider the submodule G of C() with basis
z z2
B = y , y . Note that (G) = 0 and 1
2 (B) = B. In this case,
2
27
(A)
, we have m A, which implies f m A
2
2
as A is a C-basis of G. Thus g 2 (m ) is in 2 (A).
Remark 4.18. At this moment, our stability parameter in (4.15) has
nothing to do with Kedzierskis
GIT chamber C(r, a) described in [11].
X.
28
S.-J. JUNG
5. Kedzierskis
GIT chamber
Kedzierski
[11] described his GIT cone in using a set of inequalities.
Using his lemma, we can prove further that the cone is actually a GIT
chamber C. In this section, we provide a description of C using the
Ar1 root system. Define
S(r, a)0 = { S(r, a) x 6 }.
Kedzierskis
lemma. By the same argument as in Lemma 6.7 of [11],
we can prove that it suffices to check the -stability for G-bricks not
containing x.
Lemma 5.1 (Kedzierskis
lemma [11]). For a parameter , the
29
Root system Ar1 . Identify I := Irr(G) with Z/rZ. Let i i I be
an orthonormal basis of Qr , i.e. hi , j i = ij . Define
:= {i j i, j I, i 6= j}.
The case of 1r (1, r 1, 1). From Theorem 4.1, we know that the economic resolution of X = C3 /G is isomorphic to G-Hilb C3 if G is of
type 1r (1, r 1, 1). Thus in this case, the G-bricks are just Nakamuras
G-graphs, which are -stable for + , where
+ := () > 0 for 6= 0 .
The corresponding set of simple roots is
= i i+1 i I, i 6= 0
= {1 2 , 2 3 , . . . , r1 0 } .
Example 5.2. For the group of type 31 (1, 2, 1), let Lj j = 0, 1, 2 be
the standard basis of Q3 . The corresponding set of simple roots is
L = L1 L2 , L2 L0 .
k = 0, 1, 2, 3 the
Similarly, for the group of type 14 (1, 3, 1) with R
k
standard basis of Q4 ,
R R
R R
R
R = {R
1 2 , 2 3 , 3 0 }
30
S.-J. JUNG
The case of 1r (1, a, r a). Let G be the group of type 1r (1, a, r a). Let
L and R denote the sets of simple roots for the types of a1 (1, r, r)
1
and of ra
(1, r, r), respectively. As in Section 5, let
L
l l = 0, 1, . . . , a 1}, {R
k k = 0, 1, . . . , r a 1
= L { r1 a r1 (ra)a } R .
a
ra
is called Kedzierskis
GIT chamber for G = 1r (1, a, r a).
ra
2 (i )=0
31
satisfies the system of equations (4.13) for some (2) C(a, r) and
(3) C(r a, r) by construction. Since in (4.14) is a ray of the
chamber C(r, a), it follows that C(r, a).
It remains to prove (iii). By considering G-constellations supported
on the hyperplane (x = 0) C3 , it follows that any facet of C(r, a) is
an actual GIT wall in . Therefore Kedzierskis
GIT chamber C(r, a)
32
S.-J. JUNG
e1
v5
12
11
10
v10
3
v3
v8
v1
6
2
v6
v11
7
v4
v9
8
v2
4
10
0
v7
e3
e2
Figure 6.1. Toric fan of the economic resolution for
1
0
0 1 0
1
0
0 0 0
0 1 0 0
1
1
1
1 0 1
1 1
0 0 1
1
0
0 0 1
1
(1, 7, 5)
12
0 0
0 1
1 1
1 1
1 0
0 0
with the dual basis {i } with respect to {i }. Observe that for any
C(r, a), (i ) is negative if and only if 0 i < 3.
6. Example: type
1
(1, 7, 5)
12
33
1
(1, 7, 5) with eigencoordinates
Let G be the finite group of type 12
1
3
(1, 7, 5). Let X denote the
x, y, z and L the lattice L = Z + Z 12
3
quotient variety C /G and Y the economic resolution of X. The toric
fan of Y is shown in Figure 6.1.
To use the recursion process in Section 4, first we need to investigate
the cases of type 71 (1, 2, 5) and of type 51 (1, 2, 3). Let G2 be the group
of type 17 (1, 2, 5) with eigencoordinates 2 , 2 , 2 and G3 be the group of
type 15 (1, 2, 3) with eigencoordinates 3 , 3 , 3 . Consider the toric fans
2 and 3 of the economic resolutions for the type 17 (1, 2, 5) and the
type 51 (1, 2, 3), respectively.
1
(1, 7, 5).
12
Observe that the cones 4 , 3 are on the left side of 3 . To use the
recursion, let G32 be the group of type 21 (1, 1, 1) with eigencoordinates
32 , 32 , 32 . Let 32 denote the fan of the economic resolution of the
quotient C3 /G32 . In 32 , there exist two cones 4 , 3 corresponding to
4 , 3 , respectively:
4 = Cone 21 (2, 0, 0), 21 (0, 2, 0), 12 (1, 1, 1) ,
3 = Cone 12 (0, 0, 2), 21 (0, 2, 0), 12 (1, 1, 1) .
e1
e1
4
3
3
e3
e2
1
(1, 1, 1)
2
e1
e3
e2
e3
1
(1, 2, 3)
5
e2
1
(1, 7, 5)
12
1
(1, 7, 5)
12
34
S.-J. JUNG
a
32 : 3a 3b 3c 7 32
32
4 = 1
= 1, 3 , 32 , 3 , 33 ,
32 4
def
3 = 1
= 1, 3 , 32 , 3 , 3 3 .
32 2
a+7b+5c
12
3 : xa y b z c 7 3a 3b 3
We get
o
n
def
y y2 y2
2 3 4 z4 z5 z6
,
,
,
,
z,
z
,
z
,
z
,
,
,
4 = 1
(
)
=
1,
y,
2
3
4
z z z
y y y
o
n
2
2
def
y y
y
xy
2
2
.
3 = 1
3 (2 ) = 1, x, xz, xz , xy, z , y, z , z , z 2 , z, z
,
1
1
1
7 = Cone 12
(2, 2, 10), 12
(9, 3, 9), 12
(4, 4, 8) .
9 = Cone
1
1
1
(12, 0, 0), 12
(9, 3, 9), 12
(4, 4, 8)
12
Observe that the cones 9 , 7 are on the left side of v. The cones in 2
corresponding to 9 , 7 are
9 = Cone 17 (12, 0, 0), 71 (5, 3, 4), 17 (2, 4, 3) ,
7 = Cone 17 (1, 2, 5), 17 (5, 3, 4), 71 (2, 4, 3) .
Note that the cones 9 , 7 are on the right side of the fan 2 and that
the right side is equal to the fan 3 of the economic resolution for
1
(1, 2, 3). Moreover, the cones in 3 corresponding to 9 , 7 are 4 , 3 ,
5
respectively, in (6.1). Thus the corresponding G23 -bricks 9 , 7 are:
2
9 = 1, 23 , 23
, 23 , 23
,
23
2
7 = 1, 23 , 23 23 , 23 , 23
,
where G23 denotes the group of type 51 (1, 2, 3) with eigencoordinates
23 , 23 , 23 . Using the right round down function 23 for 71 (1, 2, 5)
a+2b+5c
7
a b
23 : 2a 2b 2c 7 23
23 23
2
2 2 2
,
9 = 1
(
)
=
1,
,
2
2
23
9
2
2 2 2 ,
n
o
def
2
2
7 = 1
(
)
=
1,
2
2 2
2 2
2
2
23
7
2
2 .
35
1
(1, 7, 5)
12
c
a+7b+5c
12
2 ,
2 : xa y b z c 7 2a 2
Kedzierskis
GIT chamber. We calculate Kedzierskis
GIT chamber
1
for 12 (1, 7, 5). Since the economic resolution is G-Hilb for the group of
type 1r (1, r 1, 1), the sets of simple roots for 21 (1, 1, 1) and 13 (1, 2, 1)
are {1 0 }, {1 2 , 2 0 }, respectively. By the identification (5.3),
the set of simple roots for 51 (1, 2, 3) is
{3 4 , 4 1 , 1 2 , 2 0 },
where the underlined root is the added root as in (5.4). Similarly, the
admissible set of simple roots for 17 (1, 2, 5) is
{5 6 , 6 3 , 3 4 , 4 1 , 1 2 , 2 0 }.
1
Thus the corresponding set of simple roots for 12
(1, 7, 5) is
5 6 , 6 10 , 10 11 , 11 8 , 8 9 , 9 7 ,
.
7 3 , 3 4 , 4 1 , 1 2 , 2 0
1
1
1
1
1
1
0
0
0
0 0 0 1 0 0 0 0
0 1
0
0 0 0 1 0 1 0 0
1 1
0
0 0 0 1 1 1 0 0
1 1
0 1 0 0 1 1 1 0 1
1 1 1 1 0 0 1 1 1 1 1
1
0 1 1 0 0 0 1 1 1 1 .
1
0 1
0 0 0 0 1 0 1 1
1
0
0
0 0 0 0 0 0 1 1
1
0
0
0 0 1 0 0 0 1 0
1
0
0
0 1 1 0 0 0 0 0
0
0
0
0 1 0 0 0 0 0 0
36
S.-J. JUNG
Cone Generators
Coordinates
on U
G-brick
e1 , e2 , v11
1, z, z 2 , z 3 , z 4 , z 5 , z 6 , z 7 , z 8 , z 9 , z 10 , z 11
x
, y , z 12
z 5 z 11
e1 , v10 , v11
1, y, yz , z, z 2 , z 3 , z 4 , z 5 , z 6 , z 7 , z 8 , z 9
x y 2 z 11
, ,
z 5 z 10 y
e1 , v9 , v10
1, y, yz , yz , yz 2 , yz 3 , yz 4 , yz 5 , z, z 2 , z 3 , z 4
e1 , v8 , v9
1, y, yz , yz , yz 2 , z, z 2 , z 3 , z 4 , zy , zy , zy
e1 , v7 , v8
1, y, yz , yz , yz 2 , yz 2 , yz 3 , yz 4 , yz 4 , yz 5 , z, z 2
e1 , v6 , v7
1, y, z, z 2 , zy , zy , zy2 , yz 2 , zy2 , yz 3 , yz 3 , yz 4
e1 , v5 , v6
1, y, y 2, y 3 , z, z 2 , zy , zy , yz 2 , yz 2 , yz 2 , yz 3
e1 , v4 , v5
1, y, y 2, y 3, y 4, y 5 , yz , yz 2 , yz 2 , z, z 2 , zy
e1 , v3 , v4
1, y, y 2, y 3, y 4, y 5 , z, z 2 , zy , yz 2 , yz 3 , yz 3
10
xz 5 y 3 z 10
, ,
y2 z 9 y2
xy y 4 z 9
, ,
z 4 z 8 y3
xz 4 y 5 z 8
, ,
y3 z 7 y4
xy 2 y 6 z 7
, ,
z 3 z 6 y5
xy 2 y 7 z 6
, ,
z 3 z 5 y6
xz 2 y 8 z 5
, ,
y5 z 4 y7
xy 3 y 9 z 4
, ,
z 2 z 3 y8
e1 , v2 , v3
1, y, y 2, y 3 , y 4, y 5, y 6, yz , yz , yz , yz , z
xz y 10 z 3
, ,
y6 z 2 y9
11
e1 , v1 , v2
xy 4 y 11 z 2
, z 1 , y10
z
12
e1 , e3 , v1
1, y, y 2, y 3, y 4 , y 5, y 6, y 7, y 8 , y 9, y 10, y 11
x
, y 12, yz11
y7
e2 , v9 , v11
1, x, xz, xz 2 , xz 3 , xz 4 , x2 , x2 z, z, z 2 , z 3 , z 4
5
x3
, y ,z
z 3 x2 z x
v9 , v10 , v11
1, x, z, xz, z 2 , xz 2 , z 3 , xz 3 , z 4 , xz 4 , y, yz
v7 , v8 , v9
1, x, xy, xy
, xz, xz 2 , y, yz , yz , yz 2 , z, z 2
z
e2 , v7 , v9
1, x, x2 , x3 , x4 , xz, xz 2 , x2 z, x3 z, x4 z, z, z 2
v4 , v6 , v7
v4 , v5 , v6
3
x2 y 5
, , z
y 2 xz 2 xy 2
v2 , v3 , v4
1, x, xy, xy 2, xy 3 , xz, y, y 2, y 3, y 4 , y 5, z
x2 y 6 z 2
, ,
y 2 xz xy 3
v2 , v4 , v7
x5 y 2 z 2
, ,
z x2 x3 y
e3 , v1 , v2
1, x, xy, xy 2, xy 3, xy 4 , y, y 2, y 3, y 4 , y 5, y 6
x2 y 7
, , z
y 2 x xy 4
10
e3 , v2 , v7
1, x, xy, x2 , x2 y, x3 , x3 y, x4, x4 y, x5 , x6 , y
x7 y 2 z
, ,
y x2 x5
e2 , e3 , v7
1, x, x2 , x3 , x4 , x5 , x6 , x7 , x8 , x9 , x10 , x11
x2 z y 3 z 4
, xz 4 , xy
y
1
(1, 7, 5)
12
x2 z y 2 z 5
, xz 5 , x
y
3
x5
, y ,z
z x2 z x3
x3 y y 2 z 3
, ,
z 2 x2 xy
37
References
[1] T. Bridgeland, A. King, M. Reid, The McKay correspondence as an equivalence of derived categories, J. Amer. Math. Soc. 14 (2001), no. 3, 535554.
[2] A. Craw, A. Ishii, Flops of G-Hilb and equivalences of derived categories by
variation of GIT quotient, Duke Math. J. 124 (2004), no. 2, 259307.
[3] A. Craw, D. Maclagan, R. R. Thomas, Moduli of McKay quiver representations I: The coherent component, Proc. Lond. Math. Soc. (3) 95 (2007), no.
1, 179198.
[4] A. Craw, D. Maclagan, R. R. Thomas, Moduli of McKay quiver representations II: Gr
obner basis techniques, J. Algebra 316 (2007), no. 2, 514535
[5] V. Danilov, Birational geometry of three-dimensional toric varieties, Izv.
Akad. Nauk SSSR Ser. Mat. 46 (1982), no. 5, 971982, 1135; English translation: Math. USSR-Izv. 21 (1983), no. 2, 269280.
[6] S. Davis, T. Logvinenko, M. Reid, How to calculate A-Hilb Cn for
1
r (a, b, 1, . . . , 1), preprint.
[7] Y. Ito, H. Nakajima, McKay correspondence and Hilbert schemes in dimension three, Topology 39 (2000), no. 6, 11551191.
[8] S.-J. Jung, McKay Quivers and Terminal Quotient Singularities in Dimension 3, PhD thesis, University of Warwick, 2014.
[9] S.-J. Jung, Note on GIT chambers for G-constellations, in preparation.
[10] O. Kedzierski,
Cohomology of the G-Hilbert scheme for 1r (1, 1, r 1), Serdica