Skip to main content

Full text of "Symplectic Resolutions for Coverings of Nilpotent Orbits"

See other formats


Symplectic Resolutions for Coverings of 
Nilpotent Orbits 

Baohua FU 
February 1, 2008 

Abstract 

Let O he a nilpotent orbit in a semisimple complex Lie algebra 
g. Denote by G the simply connected Lie group with Lie algebra g. 
For a G-homogeneous covering M ^ O, let X be the normalization 
of O in the function field of M. In this note, we study the existence 
of symplectic resolutions for such coverings X. 

1 Introduction 

Let G be a simply connected semi-simple complex Lie group with Lie algebra 
0. Let C be a nilpotent adjoint orbit in q. The fundamental group 7ii{0) 
of O is finite and in general non-zero (see corollary 6.1.6 and section 8.4 of 
Let p : M — > C be a G-homogeneous covering of degree d. We denote 



by X the normalization of O in the function field of M, which will be called 
a covering of the nilpotent orbit O. Then proposition 1.2 of ||BK|] says that 



X contains M as a Zariski open subset and the map p extends to a finite 
surjective G-equivariant morphism p : X ^ O. Furthermore G has finitely 
many orbits on X and X — M is of codimension at least 2 in X. Recall that 
a regular 2-form on a smooth algebraic variety is symplectic if it is closed 
and non-degenerate at every point. The Kostant-Kirillov form on the orbit 
O gives a symplectic form u on M. 

Definition 1. A symplectic resolution for X is a projective resolution of 
singularities it : Z ^ X such that the 2-form tt*{uj) defined a priori on 
TT~^{M) extends to a symplectic form Q on the whole of Z . 



1 



1 INTRODUCTION 



2 



As shown in ||l^'u|| , a resolution is symplectic if and only if it is crepant. 
The purpose of this note is to consider the existence of symplectic resolutions, 
i.e. for which coverings of a nilpotent adjoint orbit there exists a symplec- 
tic resolution. It turns out that this is a difficult problem. The following 
proposition gives some examples of symplectic resolutions. 

Proposition 1. Let P be a parabolic subgroup of G and M the unique open 
G -orbit in T*[G/P). Let O be the orbit of a Richardson element in the 
nilradical u of the Lie algebra of P. Then the G-equivariant desingularization 
T*{G/P) ^ X is a symplectic resolution forX. 



For the proof, see proposition 7.4 of ||BK||. It should be pointed out that 



there exist some symplectic resolutions for coverings of nilpotent orbits which 
are not of the above form. An example is the double covering of the minimal 
nilpotent orbit in sp(2n, C) which is X = C^" Omin- However this is the 
only example where X is non-singular (see theorem 4.6 ||BK|| ) . The purpose 



of this note is to prove the converse of this proposition under some additional 
hypotheses. 

Recall that O and O are both stable under the scaling action of C* on 
0. This induces the Euler action of C* on O and O. Unfortunately this 
action does not lift to X in general. An example is the double covering of 
the minimal nilpotent orbit Omin in sp(2n, C). However we have (see lemma 
1.3 and proposition 1.4 ||BK|]) 



Proposition 2. For any covering X of O, there exists a C* -action on X 
which lifts the square of the Euler action of C* on O, i. e. for any A G C* 
and X & X , p{X ■ x) = X^plx). Furthermore there exists a unique point o G X 
such that p{o) = 0. This a is the unique G-fixed point in X and also the 
unique C* -fixed point in X . 

For this C*-action, we have X*lj = X'^uj for any A G C*, which is different 
to our situation in ||I''u| | , where the C*-action satisfies A*^^ = Xuj. This makes 
the situation more complicated here. However, under the hypothesis that 
the degree of the covering is odd, we can prove the following 

Theorem 1. Let Q be a semisimple complex Lie algebra and X a covering 
of a nilpotent adjoint orbit O in q. Suppose that the degree d of the covering 
p : X O is odd. Then for any symplectic resolution it : Z ^ X , Z is 
isomorphic to T*{G/P) for some parabolic subgroup P of G. Furthermore, 



2 OUTLINE OF THE PROOF 



3 



under this isomorphism, the map p o n becomes 

T*{G/P) :^ G u ^ 0, {g,u)^ Ad{g)u, 

where u is the nilradical of p = Lie{P). 

This theorem generahzes our main theorem in [ [Fu| , where we considered 
the case d = 1. 

2 Outline of the proof 

Lemma 1. The Euler action of C* on O lifts to X and for this action 
X*uj = Xuj. 

Proof. Let R be the regular functions ring of X. For A; G Z, let R[k] = {0 G 
R\\ ■ (j) = X'^cp, A G C*}. Proposition 1.4 of pK|] implies that if the degree of 



the covering X — > O is odd, then R[k] = for A; odd, i.e. R = R['^k]. 
This gives that the Euler action of C* (not only its square!) on O lifts to a 
C*-action on X. For this action, we have X*uj = Xuj. □ 

Lemma 2. The action of G x C* on X lifts to Z such that it : Z ^ X 
and pore: Z—*OisGx C*-equivariant. For the C*-action on Z , we have 
X*n = Xn for any A G C*. 

Note that the action of G (resp. C*) on M lifts to 7i-^{M) C X. Now 
the proof goes along the same line as proposition 3.1 of |[Fu||. Using the two 



lemmas, we can apply our analysis in fFuf to complete the proof of theorem 



3. For reader's convenience, we give an outline of our method. 

Let Z"^ be the fixed points subvariety in Z. Since p o vr is proper and the 
C*-action on O extends to a C-action, the valuative criterion of properness 
shows that there exists an attraction q : Z ^ Z^* . As show by lemma 3.5 
[[l^'ull , the C*-action on the smooth connected component Zq C Z""* containing 
g(7r~"^(M)) is definite, i.e. for z G Zq,T^Z = {v E TzZ\X^v = X^v} is zero 
if < 0. Now the equation X*Q = XQ implies a duality between T^Z and 
T}-''Z. Thus for z G Zq, we have T^Zq ~ T^Z ~ T^Z and T^Z = if 
k 7^ 0, 1. This shows that Zq is Lagrangian in Z. By a classical result 
of Bialynicki-Birula, the attraction q : g~^(Zo) Zq is C*-equivariantly 
isomorphic to T*Zq Zq. 

Now we study the G-action on Zq, which has an open-dense orbit, namely 
g(7r-^(M)). In fact this orbit is the whole of Zq. Note that Zq C Z^* C 



3 SOME COROLLARIES AND EXAMPLES 



4 



7r^^(o) is projective, so Zq = G / P for some parabolic subgroup P of G. To 
complete the proof, we need to show that Zq = Z'^* . Note that t\^^{6) is 
connected, so we need only to show that Zq is a connected component of 
7r~^(o) or equivalently a connected component of (p o 7r)~^(0), this is proved 
by using the explicit formula for the map T*{G / P) — > O, ((7, u) i— > Ad{g)u. 



3 Some corollaries and examples 

Corollary 1. Let X be a covering of odd degree of a nilpotent orbit O. If O 
is not a Richardson orbit, then X admits no symplectic resolution. 

This follows directly from the above theorem and proposition 2. Con- 
sider the nilpotent orbits 2A2, 2A2 + Ai, A^, E^lai), Eq in the exceptional Lie 
algebra of type -Eg; here we use notations from section 8.4 of |(J1VI|] . All 



these nilpotent orbits have Z/3 as fundamental group, but none of them is 
a Richardson orbit. The above corollary implies that the three- fold covering 
of any of the above orbits admits no symplectic resolution. 

Corollary 2. A covering X of odd degree d > 1 of a nilpotent orbit O in 
s[(n, C) admits no symplectic resolution. 

This follows from the above theorem and theorem 3.3 |[He^] , which says 
that for any polarization P oi O C sl(n, C), the morphism T*{G/P) O is 
birational. Recall that every nilpotent orbit in 5l{n, C) admits a symplectic 
resolution ( [pu| ), however the above corollary shows that their odd higher 
degree coverings admit no symplectic resolution. 

As an example, let O be the principal nilpotent orbit in sl(3, C), then 
7ri(0) = Z/3. Let X be the three-fold covering of O. Then by the above 
corollary, X does not admit any symplectic resolution. In fact, X is a so- 
called shared orbit in ||BK|| , more precisely X is nothing but the closure of 



the minimal nilpotent Omin in the exceptional Lie algebra 02, which is known 
to admit no symplectic resolution ||l^'u| . 



In the list of [pK|| , shared orbits do not admit any symplectic resolution 
except in the following case: let O be the nilpotent orbit in sp(2n, C) cor- 
responding to the partition [2,2, 1, ■ ■ ■ ,1]. Then tti{0) = Z/2. By results 
of ||Fu|| , the nilpotent orbit O does not admit any symplectic resolution. Let 
X be the double covering of O. Then X is isomorphic to Omin in sl{2n, C), 
which admits a symplectic resolution. 



4 A CONJECTURE 



5 



As shown by the above examples, there exist some nilpotent orbits which 
admit some symplectic resolutions, but not their coverings, and there exist 
some nilpotent orbits which do not admit any symplectic resolution, while 
some of their coverings do admit some symplectic resolutions. This indicates 
that the problem to determine which covering admits a symplectic resolution 
might be difficult and interesting. 



4 A conjecture 

From now on, we suppose that g is a simple complex Lie algebra. Let O he a 
nilpotent adjoint orbit in q and X a covering of O. Recall that R[k] = {0 G 
R\X ■ (j) = G C*}. Let g' = -R[2], which is also a simple Lie algebra 

(see theorem 4.2 of | |BK| ] ) . Let G' be the simply connected Lie group with 
Lie algebra g', which is the maximal connected Lie group of holomorphic 
symplectic automorphisms of X (see corollary 7.3 [BK]). By theorem 3.1 



K] , X contains a Zariski open subset M' such that M C M', and there 



exists a finite G'-covering M' — > (9' to a nilpotent orbit O' in g'. We propose 
the following conjecture on symplectic resolutions for X: 

Conjecture 1. Let p : X ^ O be a covering of a nilpotent orbit in a simple 
Lie algebra g such that R[l] = 0. Suppose that ti : Z ^ X is a symplectic 
resolution. Then there exists some parabolic sub-group P' of G' such that Z 
is isomorphic to T*{G'/P') and the map it becomes 

T*{G'/P') X. 

Remark 1. It is proved in (theorem 4-6) that R[l] 7^ z/ and only if 

X is the double covering of Omin C sp(2n, C). In this case, X = C^". 

Remark 2. By theorem 7.5 ^BI^J , ifT*{G/P) ^ X is a symplectic resolution 
for X , then there exists some parabolic subgroup P' of G' such that G/P = 
G'/P'. 

Remark 3. // the covering p : X O is of odd degree, then by our theorem 
the conjecture is true. 

Acknowledgements: I want to thank A. Beauville and M. Brion for 
helpful discussions and suggestions. I am grateful to the referee for the 
careful reading and pertinent remarks. 



REFERENCES 



6 



References 

[BK] R. Brylinski and B. Kostant, Nilpotent orbits, normality, and Hamil- 
tonian group actions, J. Am. Math. Soc. 7(1994), 269-298. 

[CM] D. Collingwood and W. Mc Govern, Nilpotent Orbits in Semi-simple 
Lie Algebras, Van Nostrand Reinhold Co., New York 1993. 

[Fu] B. Fu, Symplectic resolutions for nilpotent orbits, Invent. Math. 
151(2003), 167-186. 



[Hes] W. HcsscUnk, Polarizations in the classical groups. Math. Z. 
160(1978), 217-234.