arXiv:math/0111179v1 [math.

AG] 15 Nov 2001

A general Plücker formula

Naichung Conan Leung
October 23, 2018

Abstract
We prove a formula which compares intersection numbers of conormal
varieties of two projective varieties and their dual varieties.
When one of them is linear, we can recover the usual Plücker formula
for the degree of the dual variety.
The basic strategy of the proof is to study a category of Lagrangian
subvarieties in the cotangent bundle of a projective space under a bira-
tional transformation.

1
1 The formula
The conormal variety CS of any subvariety S in Pn is a Lagrangian subvariety in
T ∗ Pn with respect to the canonical holomorphic symplectic form. If S is smooth
then its Euler characteristic χ (S) equals to the intersection number CS · Pn up
to a sign (−1)dim S . In general χ (S) is replaced by the Euler characteristic of
the Euler obstruction EuS defined by MacPherson [Ma], we denote it as χ̄ (S).
More generally CS1 · CS2 is well-defined and equals to the Euler characteristic
of the intersection up to a sign provided that S1 and S2 intersect transversely
along a smooth subvariety in Pn .
In this paper we prove the following formula.

Theorem 1 Suppose S1 and S2 are two subvarieties in Pn which intersect

transversely and the same holds true for their dual varieties S1∨ and S2∨ in Pn∗ .
Then we have

(CS1 · Pn ) (CS2 · Pn ) CS1∨ · Pn∗ CS2∨ · Pn∗
CS1 · CS2 + = CS1∨ · CS2∨ + .
(−1)n+1 (n + 1) (−1)n+1 (n + 1)

This formula arises as we studied in [Le] the Legendre transformation on the

category of Lagrangian subvarieties in a hyperkähler manifold under a birational
transformation of the ambient manifold.

By applying this formula to cases when S2 are linear subspaces in Pn with

different dimensions, we obtain the following three corollaries which determine
the degree, the Euler characteristic and the dimension of the dual variety. We
note that the degree of S1 equals CS1 · CS2 with S2 a linear subspace of com-
plementary dimension.
First we recover the generalized Plücker formula of Parusiński [Pa] and Ern-
ström [Er].

Corollary 2 For any k we have

n−k+1



deg S ∨ = (−1) k χ̄ (S) − (k + 1) χ̄ S 1 + χ̄ S k+1 ,

where S k is the intersection of S with a generic codimensional k linear subspace.

When S is a plane curve in P2 this gives the classical Plücker formula for
the degree of the dual curve S ∨ ,

deg S ∨ = d (d − 1) − 2δ − 3κ,

where d, δ, κ denote the degree, the number of nodes, the number of cusps of
S. This classical formula has another generalization for higher dimensional S

2
with only isolated singularities by Teissier [Te] in the hypersurface case and by
Kleiman [Kl] in general. For subjects closely related to the Plücker formula,
readers can consult [GKZ] and [Kl2].

Second, we can derive the Euler characteristic of the dual variety.

Corollary 3 For any subvariety S in Pn we have

χ̄ (S ∨ ) = nχ̄ (S) − (n + 1) χ̄ S 1 .

Third

we can determine the dimension of the dual variety by comparing
χ S k with a linear function in k. For example, in the hypersurface case we
have,

Corollary 4 Let S be a hypersurface in Pn , then S ∨ has codimension c if and

only if for any k ≤ c we have



χ̄ S k = k χ̄ S 1 + (1 − k) χ̄ (S)

and it becomes a strict inequality when k = c + 1.

There are dimension formulas for dual varieties in terms of Hessian matrices
given by Segre [Se] and Katz [Ka].

When S2 is a smooth quadric hypersurface, its dual variety S2∨ is again a

smooth quadric hypersurface. In this case we have the following corollary.

Corollary 5 Suppose a subvariety S in Pn and its dual variety S ∨ are both

hypersurfaces then
 n  n
1 + (−1) ∨ ∨ ′ 1 + (−1)
χ̄ (S) − χ̂ (S ∩ Q) 1 + = χ̄ (S ) − χ̂ (S ∩ Q ) 1 + ,
2n 2n
n−2
where Q, Q′ are general quadric hypersurfaces and χ̂ (S ∩ Q) = (−1) CS ·CQ .

Remark: We expect that χ̂ (S ∩ Q) is simply equal to χ̄ (S ∩ Q).

Proof: By applying our formula to the case when S1 is a hyperplane and
S2 = Qn−1 is a general quadric hypersurface in Pn , we obtain

χ (Qn ) = n + 1 if n is odd,
χ (Qn ) = n + 2 if n is even.

This can also be obtained by using explicit descriptions of Qn ’s. We apply our
formula again by replacing S1 with the hypersurface S and we obtain the result.


3
2 Proof of the formula
The basic idea of our proof is to study the intersection of two Lagrangian sub-
varieties C1 and C2 in the holomorphic symplectic manifold T ∗ Pn and their
behaviors under the canonical birational transformation from T ∗ Pn to T ∗ Pn∗ :
Recall that we can write Pn = (V \0) /C× and Pn∗ = (V ∗ \0) /C× with V a
vector space of dimension n+1. There is a similar description for their cotangent
bundles, namely

T ∗ Pn = {(x, ξ) ∈ (V \0) × V ∗ : ξ (x) = 0} /C× ,

T ∗ Pn∗ = {(x, ξ) ∈ V × (V ∗ \0) : ξ (x) = 0} /C× .

The zero section P in T ∗ Pn (resp. P ∗ in T ∗ Pn∗ ) is given by {ξ = 0} (resp.

{x = 0}). The identity homomorphism on V × V ∗ descends to a birational
map,

Φ : T ∗ Pn 99K T ∗ Pn∗ ,

which is biregular from T ∗ Pn \P to T ∗ Pn∗ \P ∗ .

Even though T ∗ Pn is incomplete, the intersection of CS1 and CS2 only occurs
along the zero section P in T ∗ Pn when S1 and S2 intersect transversely in Pn .1
This holds true even after we compactify T ∗ Pn to M as follow,

M = P (T ∗ Pn ⊕ OPn ) .

If we denote the closure of Ci ’s in M as C̄i ’s, then we have

C̄1 ∩ C̄2 = C1 ∩ C2 = S1 ∩ S2
∩ ∩ ∩
M ⊃ T ∗ Pn ⊃ Pn .

Therefore we can simply write C̄1 · C̄2 as C1 ·C2 . Similarly we have Ci ·P = C̄i ·P .

Remark: We recall the following useful result (see e.g. [Le]): If C1 and
C2 are Lagrangian subvarieties in T ∗ Pn which intersect cleanly. Suppose their
closures C̄1 and C̄2 have the same intersection, i.e. C1 ∩ C2 = C̄1 ∩ C̄2 , then
dim C1 ∩C2
C1 · C2 = (−1) χ (C1 ∩ C2 ) .
n
In particular we have P · P = (−1) (n + 1).

It is not difficult to see that the above birational map Φ on T ∗ Pn extends to

a birational map between two complete varieties M and M ′ :

ΦM : M 99K M ′ ,
1 For now on we will simply write Ci for CSi , the conormal variety of Si .

4
which is biregular outside P and P ∗ .
Note that M ′ can also be described as a flop as follow:
π f π′
M ← M → M′
∪ ∪j ∪
p p′
P ← Pe → P ∗.

Here Mf is the blow up of M along P . The exceptional locus Pe = π −1 (P )

is a Pn−1 -bundle over P . It admits another Pn−1 -bundle structure over the
f along this second fiber
dual projective space P ∗ and π ′ is the blow down of M
e
structure on P .
In the next section, we will show that the Legendre functor L = π∗′ ◦ π −1
on the derived categories of coherent sheaves on M and M ′ is an equivalence of
categories. This implies that
ExtkOM (S1 , S2 ) ∼
= ExtkOM ′ (L (S1 ) , L (S2 )) ,
for any coherent sheaves S1 , S2 on M . In order to use this equivalence to
prove the theorem 1, we need to compute the Chern characters of L (OP ) and
L (OCi )’s. First we relate the Lagrangian intersection number C1 · C2 with the
category of coherent sheaves on M .
Lemma 6 For any n dimensional subvarieties C̄1 , C̄2 in M we have
P k k

n
k dim (−1) ExtOM OC̄1 , OC̄2 = (−1) C̄1 · C̄2 .

Proof: We recall the Riemann-Roch formula for the global Ext groups: For
any coherent sheaves S1 and S2 on M we have,
Z
P k k
k dim (−1) Ext OM (S 1 , S 2 ) = ch (S1 ) ch (S2 ) T dM
M
k
where ch (S1 ) = Σ (−1) chk (S1 ). For Si = OC̄i the structure sheaf of a subva-
riety C̄i of dimension n, we have


chk OC̄i = 0 for k < n,

 
chn OC̄i = C̄i ,
 
where C̄i denotes the Poincaré dual of the variety C̄i .
Therefore
k

dim (−1) ExtkOM OC̄1 , OC̄2
Z



= ch OC̄1 ch OC̄2 T dM
ZM
n 
 

= (−1) C̄1 + h.o.t. C̄2 + h.o.t. (1 + h.o.t.)
M
Z
n     n
= (−1) C̄1 ∪ C̄2 = (−1) C̄1 · C̄2 .
M

5
Here h.o.t. refers to higher order terms which do not contribute to the outcome
of the integral. Hence the result. 

Lemma 7

ch (L (OP )) = ± [P ∗ ] + h.o.t..

Proof: From the previous lemma, we have

P k
(−1) dim ExtkOM (OP , OP )
n
= (−1) P · P
= χ (P )
= (n + 1) .

On the other hand, the support of L (OP ) must be inside P ∗ ⊂ M ′ because

M and M ′ are isomorphic outside their exceptional loci P and P ∗ . This implies
that ch (L (OP )) = α [P ∗ ] + h.o.t. for some integer α. Therefore,

P k
k(−1) dim ExtkOM ′ (L (OP ) , L (OP ))
Z
= ch (L (OP )) ch (L (OP )) T dM
′
ZM
n
= (−1) α [P ∗ ] · α [P ∗ ]
M′
= (n + 1) α2 .

Now L being an equivalence of categories implies that

P P
(−1)k dim ExtkOM ′ (L (OP ) , L (OP )) = (−1)k dim ExtkOM (OP , OP ) .

This forces α to be ±1. Hence the claim. 

Lemma 8 Suppose C is a n dimensional irreducible subvariety in M not con-

taining P . We consider
Ln+1 n LC·P
S= OC − (−1) OP .

Then
n
ch (L (S)) = (n + 1) [C ∨ ] − (−1) (C ∨ · P ∗ ) [P ∗ ] + h.o.t.

Proof: First we have

X k
(−1) dim ExtkOM (S, OP )
n n
= (−1) [(n + 1) C · P − (−1) (C · P ) (P · P )]
= 0.

6
Therefore, using the equivalence of categories induced by L again, we have
X k
(−1) dim ExtkOM ′ (L (S) , L (OP )) = 0.

Away from P ∗ , the support of L (S) is clearly C ∨ with multiplicity n + 1,

therefore we have

ch (L (S)) = (n + 1) [C ∨ ] + β [P ∗ ] + h.o.t.

for some integer β. On the other hand

X k
(−1) dim ExtkOM ′ (L (S) , L (OP ))
Z
= ch (L (S)) ch (L (OP )) T dM ′
M′
= (−1)n ((n + 1) [C ∨ ] + β [P ∗ ]) · (± [P ∗ ]) .
n
This being zero implies that β = − (−1) (C ∨ · P ∗ ). Therefore

ch (L (S)) = (n + 1) [C ∨ ] − (−1)n (C ∨ · P ∗ ) [P ∗ ] + h.o.t.

Hence the result. 

Now we suppose that C1 and C2 are two n dimensional irreducible subva-

rieties in M which do not include P . We denote the corresponding sheaves
constructed as above by S1 and S2 respectively. We have
X k
(−1) dim ExtkOM (S1 , S2 )
Z
= ch (S1 ) ch (S2 ) T dM
M
! !
2 n (C1 · P ) (C2 · P )
= (n + 1) (−1) C1 + n+1 P · C2 + n+1 P
(−1) (n + 1) (−1) (n + 1)
" #
n+1
2 n (−1)
= (n + 1) (−1) C1 · C2 + (C1 · P ) (C2 · P ) .
n+1

By the above computations of the Chern characters of ch (L (Si ))’s we have

P
a similar formula for (−1)k dim ExtkOM ′ (L (S1 ) , L (S2 )), i.e.
X k
(−1) dim ExtkOM ′ (L (S1 ) , L (S2 ))
" #
n+1
2 n ∨ ∨ (−1) ∨ ∗ ∨ ∗
= (n + 1) (−1) C1 · C2 + (C1 · P ) (C2 · P ) .
n+1

Finally using the equivalence of categories induced by L we have

X k
X k
(−1) dim ExtkOM (S1 , S2 ) = (−1) dim ExtkOM ′ (L (S1 ) , L (S2 )) .

7
Combining these and we obtain



(CS1 · P ) (CS2 · P ) CS1∨ · P ∗ CS2∨ · P ∗
CS1 · CS2 + n+1 = CS1∨ · CS2∨ + n+1 .
(−1) (n + 1) (−1) (n + 1)

Thus we have completed the proof of the main theorem 1 assuming that L is
an equivlance of categories. 

8
3 Equivalence of derived categories
In this section we adapt Bondal and Orlov arguments [BO] to prove that Db (M )
f is the blow up of M
and Db (M ′ ) are equivalent categories. Recall that M
∼ n ′ ∗ ∼ n∗
along P = P , or the blow up of M along P = P . We recall the following
commutative diagram:
π f π′
M ← M → M′
∪ ∪j ∪
p p′
P ← Pe → P ∗.

Proposition 9 In the above situation, we have an equivalence of derived cate-

gories,

L = π∗ π ′∗ : Db (M ′ ) → Db (M ) .

Note e n−1
-bundle over the projective space P (or P ∗ ). Therefore
  that P is a P
P ic Pe = Z+Z. Every line bundle on Pe is isomorphic to p∗ OP (a)⊗p′∗ OP ∗ (b)
for some integers a and
 b,we denote it as OPe (a, b).
We claim that OM e e is isomorphic to O e (−1, −1). By symmetry of M
f P |P P
′
and M  we know it must be of the form OPe (a, a) for some a. For the restriction
of O f Pe to a fiber of p equals OPn−1 (−1) because M
M
f is obtained from M by
blowing up the smooth center P . This implies that a = −1.
As a blown up manifold we have
 
ωM ∗
f = π ωM ⊗ OM
e
f (n − 1) P .

This implies that

ωM
f |P
e = OP
e (1 − n, 1 − n) .

Proof of proposition: The arguments presented here are basically the same
as in section 3 of [BO]. For any A, B ∈ Db (M ′ ) we want to show that

Hom (A, B) ∼
= Hom (π∗ π ′∗ A, π∗ π ′∗ B)
∼
= Hom (π ∗ π∗ π ′∗ A, π ′∗ B) .

On the other hand ∼ ′∗ ′∗ ′∗

 Hom (A, B) = Hom (π A, π A) because the functor π :
Db (M ′ ) → Db M f is full and faithful for any blow up morphism. Therefore
it is suffices to show that


Hom Ā, π ′∗ B = 0,

9
where Ā is defined by the following exact triangle

π ∗ π∗ π ′∗ A → π ′∗ A → Ā.
 
Using π ∗ : Db (M ) → Db M f being full and faithful we can obtain Ā ∈
 
D (M )⊥ ⊂ D M f . An earlier result of Orlov showed that

⊥

D (M ) = D (P )−n+1 , · · · , D (P )−1

as a semiorthogonal decomposition, where D (P )−k is the full subcategories of

 

Db M f given by the image of Db (P ) under j∗ O e (−k) ⊗ p∗ (•) . Since P
P
isomorphic to Pn−1 we have


D (M )⊥ = j∗ OPe (a, b) −n+1≤b≤−1
−1≤a−b≤n−1

⊥
From this description we know that j∗ OPe (a, b) belongs to both D (M ) and
⊥
D (M ′ ) when −n + 1 ≤ a, b ≤ −1. In particular


Hom Ā, j∗ OPe (a, b) = 0 when − n + 1 ≤ a, b ≤ −1.

Because of this Ā ∈ D (M )⊥ has to lie inside the subcategory of D (M )⊥ gen-

′ ⊥
erated by those j∗ OPe (a, b) with a ≥ 0. This implies that Ā ⊗ ωM
f ∈ D (M )
b ′
because of ωM
f |P
e = OP e (1 − n, 1 − n). That is for any B ∈ D (M ) we have


Hom π ′∗ B, Ā ⊗ ωMf =0

which is equivalent to


Hom Ā, π ′∗ B = 0,

by the Serre duality. Hence our proposition. 

Acknowledgments: The author thanks D. Abramovich, T. Bridgeland, A.

Voronov for very helpful discussions. This project is partially supported by
NSF/DMS-0103355.

References
[BO] A. Bondal, D. Orlov, Semiorthogonal decompositions for algebraic vari-
eties, alg-geom/9506012.
[Er] L. Ernström, A Plücker formula for singular projective varieties, Comm.
Algebra, 25 (9), (1997), 2897-2901.

10
[GKZ] I.M. Gelfand, M. Kapranov, A. Zelevinsky, Discriminants, resultants
and multidimensional determinants, Birkhäuser (1994).
[GH] P. Griffiths, J. Harris, Principles of algebraic geometry, Wiley-
Interscience (1978).
[Ka] N. Katz, Pinceaux de Lefschetz: théorèm d’existence, Lecture notes in
Math.,340 (1973), Springer-Verlag, 212-253.
[Kl] S. Kleiman, A generalized Teissier-Plücker formula, Classification of Al-
gebraic Varieties (C. Ciliberto, E.L. Livorni and A.J. SOmmese, eds.).
Comt. Math., vol. 162, A.M.S., Providence, 1994, 249-260.
[Kl2] S. Kleiman, Intersection theory and enumerative geometry: A decade in
review, Proc. Symp. Pure Math., Vol. 46, (1987), 321-370.
[Le] N.C. Leung, Lagrangian submanifolds in hyperkähler manifolds, Legen-
dre transformations, math.SG/0110330.
[Ma] R. MacPherson, Chern classes of singular algebraic gvarieties, Ann. of
Math. 100 (1974), 423-432.
[Pa] A. Parusiński, Multiplicity of the dual variety, Bull. London Math. Soc.
23 (1991), 429-436.
[Se] B. Segre, Bertini forms and Hessian matrices, J. Lond. Math. Soc., 26
(1951), 164-176.
[Te] B. Teissier, Sur diverse conditions numériques d’équisingularité des
familles de courbe, preprint, Certre de Math., École Polytechnique
(1975).

Address: School of Mathematics, University of Minnesota, Minneapolis, MN

55454.
Email: [email protected].

11