VECTOR BUNDLES INDUCED FROM JET SCHEMES

BAILIN SONG

A BSTRACT. A family of holomorphic vector bundles is constructed on a complex mani-

arXiv:1609.03688v2 [math.DG] 21 Oct 2020

fold X. The space of the holomorphic sections of these bundles are calculated in certain
cases. As an application, if X is an N -dimensional compact Kähler manifold with holo-
nomy group SU (N ), the space of holomorphic vector fields on its jet scheme Jm (X) is
calculated. We also prove that the space of the global sections of the chiral de Rham com-
plex of a K3 surface is the simple N = 4 superconformal vertex algebra with central charge
6.

1. I NTRODUCTION

On a Ricci flat compact Kähler manifold X, the holomorphic sections of the bundle
given by tensors of tangent and cotangent bundles are exactly the parallel sections (page
142 of [15]). So if the holonomy group of X is G, the space of the holomorphic sections
of the bundle is isomorphic to the space of the G invariants of the fibre. In this paper,
on a complex manifold X, given holomorphic vector bundles E and F , we construct a
family of holomorphic vector bundles Am (E, F ), m ∈ Z≥0 . If X is an N-dimensional
compact Hermitian manifold with the holonomy group SU(N), and E and F are sums
of copies of the holomorphic tangent and cotangent bundles, we show that the space
of the holomorphic sections of Am (E, F ) is isomorphic to the space of the Jm (SL(N, C))
invariants of the fibre of the bundle. Here Jm (SL(N, C)) is the mth jet scheme of SL(N, C)
(see Theorem 5.6).
The idea of the construction of Am (E, F ) is from the jet scheme. The mth jet scheme [2]
Jm (X) of an algebraic scheme X over C is determined by its functor of points: for every
C-algebra A, we have a bijection
Hom(Spec(A), Jm (X)) ∼
= Hom(Spec(A[t]/(tm+1 )), X).
There is a canonical projection πm : Jm (X) → X. If X is a projective complex manifold,
E is a vector bundle over X, then Jm (E) is a vector bundle over Jm (X). If X is Ricci
flat, we expect that sections of the bundle given by tensors of Jm (T X) and Jm (T ∗ X) are
Jm (SL(N, C)) invariants of the fibre. Am (E, F ) is the vector bundle over X, such that the
sheaf of sections of Am (E, F ) is the push forward of the sheaf of sections of Sym∗ Jm (E) ⊗
∧∗ Jm (F ) through πm .
There are two applications of Theorem 5.6. The first application is that we can calculate
the holomorphic vector fields on Jm (X),
Theorem 1.1. If X is an N-dimensional compact Kähler manifold with holonomy group SU(N),
then the space of holomorphic vector fields on Jm (X) has dimension m.

Key words and phrases. Jet scheme; vector bundle; holomorphic section; chiral de Rham complex; K3
surface.

1
 The second application is that we can calculate the global sections of the chiral de Rham
complex on any K3 surface.
Theorem 1.2. If X is a K3 surface, the space of global sections of the chiral de Rham complex of
X is the simple N = 4 superconformal vertex algebra1 with central charge 6.

This generalizes our previous result in [12], which calculates the global sections of the
chiral de Rham complex on a Kummer surface. By chiral Poincaré duality in [11] and
the fact that the elliptic genus of a K3 surface is the Jacobi form 2φ0,1 (z; τ ) of weight 0
and index 1, all the graded dimensions of the cohomology of the chiral de Rham com-
plex of a K3 surface can be calculated immediately, which has an application (see [13]) in
Mathieu moonshine. According to [6] , on any Calabi-Yau manifold, the cohomology of
the chiral de Rham complex can be identified with the infinite-volume limit of the half-
twisted sigma model defined by E. Witten. This calculation may help to understand the
half-twisted sigma model on K3 surfaces.
The paper is organized as follows. In section 2, we review some basic facts of com-
plex geometry. In section 3, we construct two holomorphic vector bundles Am (E, F )
and Bm (E, F ). In section 4, we compare the holomorphic structures of Am (E, F ) and
Bm (E, F ), and calculate the mean curvature of Am (E, F ) when the mean curvatures of
E, F and T X vanish. In section 5, we give some results of the holomorphic sections of
Am (E, F ). Finally, in section 6, we calculate the global sections of the chiral de Rham
complex on a K3 surface.

2. T HE C HERN CONNECTIONS AND CURVATURES

For a complex manifold X, let Ωk,l X be the space of smooth forms on X of type (k, l).
Similarly, for a holomorphic vector bundle E on X, let Ωk,lX (E) be the space of all smooth
forms of type (k, l) with values in E. Let hE = (−, −)E be a Hermitian metric on E and
∇E = ∇1,0 ¯ 1,0 0,0 1,0
E + ∂ be the associated Chern connection with ∇E : Ω (E) → Ω (E). Then its
curvature is
¯ ∇E ] = [∂,
ΘE = [∂, ¯ ∇1,0 ] ∈ Ω1,1 (Hom(E, E)).
E

Locally, if e = (e1 , · · · , ep ) is a holomorphic frame of E over U ⊂ X, the connection one

form θE of ∇E is given by
X
∇E ei = θijE ej .
If we define the matrix H E = (HijE ),
HijE = (ei , ej )E ,
then
θijE = ∂HilE HElj .
Here (HElj ) is the inverse matrix of H E . The curvature form of ∇E is
¯ E ∈ Ω1,1 (Hom(E, E)).
ΘE = ∂θ
1
In the literature, there are small and large N = 4 superconformal algebras. In this paper, it is the small
N = 4 superconformal algebra.

2
Covariant derivatives of the curvature. According to [5], the tensor fields Fk = FkE , k ≥ 2
as higher covariant derivatives of the curvature can be defined by induction,

F2 = ΘE ∈ Ω1,1 (Hom(E, E)) = Ω0,1 (Hom(T ⊗ E, E)),

Fk = ∇1,0 0,1
E Fk−1 ∈ Ω (Hom(T
⊗(k−1)
⊗ E, E)), k ≥ 3.

Since ∇E is a Chern connection,

[∇1,0 1,0
E , ∇E ] = 0.

So we have a lemma from [5],

Lemma 2.1.
Fn ∈ Ω0,1 (Hom(Symn−1 T ⊗ E, E)).

Locally, if y = (y1 , · · · , yN ) is a holomorphic coordinate system and e = (e1 , · · · , ep ) is a

l
holomorphic frame of E on U ⊂ X, Fk,i 1 ··· ,in
is denoted by

l ∂ ∂
Fk,i e = Fn+1 (
1 ··· ,in l
,··· , )ek ∈ Ω0,1 (E).
∂yi1 ∂yin

If X is an Hermitian manifold with the Hermitian metric h = (−, −), let M = M E be

the mean curvature of E, i,e,
∂ ∂
M = ΘE ( , )Hij ∈ Hom(E, E).
∂yi ∂ ȳj

Here Hij = (dyi , dyi ). We define the tensor field Mn = MnE by

M2 = M, Mn = ∇1,0 Mn−1 , n ≥ 3.

Then
X ∂ ∂
Mn = Fn ( , )Hij ∈ Hom(Symn−2 T ⊗ E, E).
∂yi ∂ ȳj
If M = C0 Id for a constant C0 , then h is an Einstein-Hermitian metric. We have the
following lemma obviously,

Lemma 2.2. If hE is an Einstein-Hermitian metric, then MnE = 0 for n ≥ 3. If the mean

curvature of hE is zero, then MnE = 0 for n ≥ 2. In particular, if the Ricci curvature of X is zero ,
then MnT = 0 for n ≥ 2.

If X is Kähler, then for the holomorphic tangent bundle T of X,

ΘT ∈ Ω0,1 (Hom(Sym2 T, T )),

we have

Lemma 2.3.
FnT ∈ Ω0,1 (Hom(Symn T, T )).
3
The Chern connections of the vector bundle with different holomorphic structures.
Assume E has another holomorphic structure given by ∂¯′ and ∇ ˜ is the Chern connection
corresponding to ∂¯ . Let K = ∂¯ − ∂¯ ∈ Ω (Hom(E, E)). Let K ∈ Ω1,0 (Hom(E, E)) be the
′ ′ 0,1 ∗

dual of K, i.e. K ∗ is determined by

∂ ∂
(K( )a, b) = (a, K ∗ ( )b).
∂ ȳi ∂yi
We have the following relation between the two connections,
Lemma 2.4.
˜ + K − K∗
∇=∇
˜ can be determined by
Proof. This is because that the Chern connections ∇ and ∇
¯
∇ = ∇1,0 + ∂, ˜ 1,0 + ∂¯′ ,
˜ =∇
∇
and for any smooth sections a, b of E,
˜ b) + (a, ∇b).
d(a, b) = (∇a, b) + (a, ∇b) = (∇a, ˜
Now
˜ + K − K ∗ )a, b) + (a, (∇
((∇ ˜ + K − K ∗ )b) = (∇a,
˜ b) + (a, ∇b)
˜ = d(a, b)
and
˜ + K − K∗ = ∇
∇ ˜ 1,0 + ∂¯′ + K − K ∗ = ∇
˜ 1,0 − K ∗ + ∂¯
˜ 1,0 − K ∗ takes values in Ω1,0 (E). So ∇
Since K ∗ ∈ Ω1,0 (Hom(E, E), ∇ ˜ + K − K ∗ is the Chern
connection on E corresponding to the holomorphic structure given by ∂. ¯ Thus by the
˜
uniqueness of the Chern connection, ∇ = ∇ − K + K . ∗


Let E ∨ be the dual complex vector bundle of E, then E ∨ has two holomorphic structures
induced from the two holomorphic structures on E. The two holomorphic structures are
given by ∂¯ and ∂¯′ , which are determined by
¯ ∨ (b)) = (∂a
∂(a ¯ ∨ )(b) + a∨ (∂b)
¯ and ∂(a
¯ ∨ (b)) = (∂¯′ a∨ )(b) + a∨ (∂¯′ b)
for any section a∨ of E ∨ and b of E. Let ∂¯ = ∂¯′ + K ∨ , then
¯ ∨ (b) = ((∂¯ + K ∨ )a∨ )(b) + a∨ ((∂¯ + K)b).
∂a
Thus K ∨ is determined by
(K ∨ a∨ )(b) = −a∨ (Kb).

Weitzenböck formulas. We list the following Weitzenböck formulas for the compact Her-
mitian manifold here:
(1) If α is a smooth section of E,
(2.1) ∆∂¯α = ∇∗ ∇α − Mα.
(2) If X is Kähler, and α is a smooth section of E = T X, then the Ricci curvature
Ric α = Mα.
And
∆∂¯α = ∇∗ ∇α − Ric α.
4
 3. V ECTOR BUNDLES INDUCED FROM JET SCHEMES

Given holomorphic bundles E and F on a complex manifold X, a family of holomor-

phic vector bundles Am (E, F ), m ∈ Z≥0 will be constructed in this section.

The algebra Rm . For m ∈ Z≥0 , y = (y1 , · · · , yN ), e = (e1 , · · · , ep ) and f = (f1 , · · · , fq ), let

Rm = Rm (y, e, f ) be the algebra
(1) (j) (m) (0) (j) (0) (j)
C[y1 , · · · , yi , · · · , yN ] ⊗ C[e1 , · · · , ei , · · · , ep(m) ] ⊗ ∧C [f1 , · · · , fi , · · · , fq(m) ].
(j) (j) (j) (0) (j) (m)
Here yi , ei , fi are new variables and ∧C [f1 , · · · , fi , · · · , , fq ] is the exterior algebra
(0) (j) (m)
generated by f1 , · · · , fi , · · · , fq over C. There is a natrual embeding Rm (y, e, f ) ֒→
Rm+1 (y, e, f ) for all m ≥ 0. Let
R∞ (y, e, f ) = −
lim
→Rm (y, e, f ).
Let
X (j) ∂ X (j) ∂ X (j) ∂
L= j yi (j)
+ j ei (j)
+ j fi (j)
,
i,j ∂yi i,j ∂ei i,j ∂fi
X (j) ∂ X (j) ∂
Le = ei (j)
, Lf = fi (j)
.
i,j ∂ei i,j ∂fi
Le , Lf , L gives a Z3≥0 grading of Rm (y, e, f ), i.e.
M
Rm (y, e, f ) = Rm (y, e, f )[j, k, l]
j,k,l

with
Le (a) = ja, Lf (a) = ka, L(a) = la for a ∈ Rm (y, e, f )[j, k, l].
It is easy to see that Rm (y, e, f )[j, k, l] is a finite dimensional complex vector space. Let
M
Rm (y, e, f )[j, k] = Rm (y, e, f )[j, k, l].
l

Let D̃ be the derivation on R∞ [y, e, f ] given by

(j) (j+1) (j) (j+1) (j) (j+1)
D̃yi = yi , D̃ei = ei , D̃fi = fi .
D̃ maps Rm (y, e, f )[j, k, l] to Rm+1 (y, e, f )[j, k, l + 1].

Construction of the vector bundles Am (E, F ). Let U be the set which consists of (U, y, e, f )
such that y = (y1 , · · · , yN ) is a coordinate system , e = (e1 , · · · , ep ) is a holomorphic frame
of E and f = (f1 , · · · , fq ) is a holomorphic frame of F on U. Let O(U) be the space of
holomorphic functions on U. For m ∈ Z≥0 ∪ {∞}, let
Am (U, y, e, f ) = O(U) ⊗ Rm (y, e, f )
be the algebra of holomorphic maps from U to Rm (y, e, f ). For any a ∈ Am (U, y, e, f ) , let
a(x) ∈ Rm (y, e, f ) be the image of a at x ∈ U.
Let
X ∂ (1)
(3.1) D= ⊗ yi + 1 ⊗ D̃.
i
∂yi
It is a derivation on A∞ (U, y, e, f ).
5
 Now if (Uα , yα, eα , fα ), (Uβ , yβ , eβ , fβ ) ∈ U with
X X
yα,i = fi (yβ ), eα,i = gij eβ,j , fα,i = hij fβ,j on Uα ∩ Uβ ,
there is a unique O(Uα ∩ Uβ ) algebra isomorphism
rαβ : A∞ (Uα ∩ Uβ , yα , eα , fα ) → A∞ (Uα ∩ Uβ , yβ , eβ , fβ )
with
(l)
rαβ (yα,i ) = D l f (yβ ), rαβ (ϕ) = ϕ, for ϕ ∈ O(Uα ∩ Uβ ),
(l) (0) (l) (0)
X X
rαβ (eα,i ) = D l ( gij eβ,j ), rαβ (fα,i ) = D l ( hij fβ,j ).
It is easy to see
Lemma 3.1. For a ∈ A∞ (Uα ∩ Uβ , yα , eα , fα ),
rαβ (Da) = D rαβ (a).

For any x ∈ Uα ∩ Uβ , we get an isomorphism of C-algebras

x
rαβ : R∞ (yα , eα , fα ) → R∞ (yβ , eβ , fβ )
with
(k) (k) (0)
X
x
(3.2) rαβ (yα,i ) = D k fi (yβ )(x), x
rαβ (eα,i ) = D k ( gij eβ,j )(x),
(k) (0)
X
x
rαβ (fα,i ) = D k ( hij fβ,j )(x).
x
rαβ satisfies
x x
rαβ ◦ rβα = Id, for x ∈ Uα ∩ Uβ ,
x x x
rβγ ◦ rαβ = rαγ , for x ∈ Uα ∩ Uβ ∩ Uγ .
Thus the open cover {Uα ×R∞ (yα , eα , fα ) : (Uα , yα, eα , fα ) ∈ U} and the transition functions
x
rαβ define an algebra bundle A∞ (E, F ) on X.
x
By (3.2), the transition functions rαβ map Rm (yα , eα , fα )[j, k, l] to Rm (yβ , eβ , fβ )[j, k, l].
Let Am (E, F )[j, k, l], Am (E, F )[j, k] and Am (E, F ) be the subbundles of A∞ [E, F ] with fi-
bres Rm [j, k, l], Rm [j, k] and Rm , respectively. We have
M
Am (E, F ) = Am (E, F )[j, k, l],
j,k,l
M
Am (E, F )[j, k] = Am (E, F )[j, k, l].
l
Am (E, F )[j, k, l] are finite dimensional holomorphic vector bundles. Let Om be the algebra
bundle Am (E, F )[0, 0] = Am (0, 0) and Om [l] = Am (E, F )[0, 0, l].
Let Am (E, F ), Am (E, F )[j, k] and Om be the sheaf of holomorphic sections of Am (E, F ),
Am (E, F )[j, k] and Om , respectively. Am (E, F ) are algebras of Om and Am (E, F )[j, k] are
modules of Om .
Remark 3.2. If X is a complex projective manifold, then Am (E, F )[j, k] can be induced from the
mth jet scheme of X. For each integer m ≥ 0, the sheaf Om is the push forward of the structure
sheaf of Jm (X) through πm . Jm (E) and Jm (F ) are vector bundles over Jm (X). Am (E, F )[j, k] is
the push forward of the sheaf of sections of Symj Jm (E) ⊗ ∧k Jm (F ) through πm .
6
 By Lemma 3.1, rαβ commutes with the derivation D. So D is a derivation of the sheaf of
algebras A∞ (E, F ). According to (3.1), D maps Am (E, F )[j, k, l] to Am+1 (E, F )[j, k, l + 1].
Thus D is a derivation of the O∞ module A∞ (E, F )[j, k].
D can be extended to a derivation of Ω0,∗ ¯
U (A∞ (E, F )) by assuming D dyi = 0 and Df =
P ∂f (1)
i ∂yi yi , for any smooth function f on U.

Vector bundles Bm (E, F ). Let Bm (E, F ) be the algebra bundle constructed from the open
cover {Uα × R∞ (yα , eα , fα ) : (Uα , yα , eα , fα ) ∈ U} and the transition functions
x
r̃αβ : R∞ (yα , eα , fα ) → R∞ (yβ , eβ , fβ ).
x
For x ∈ Uα ∩ Uβ , r̃αβ is the isomorphism of algebras given by
(k)
X ∂fi (yβ ) (k) (k)
X (k) (k)
X (k)
x
r̃αβ (yα,i ) = (x)yβ,j , rαβ (eα,i ) = gij (x)eβ,j (x), rαβ (fα,i ) = hij (x)fβ,j .
∂yβ,j
x
r̃αβ maps Rm (yα , eα , fα )[j, k, l]to Rm (yβ , eβ , fβ )[j, k, l]. Let Bm (E, F )[j, k, l], Bm (E, F )[j, k]
and Bm (E, F ) be the subbundles of B∞ [E, F ] with fibres Rm [j, k, l], Rm [j, k] and Rm re-
spectively. Then Bm (E, F )[j, k, l] are finite dimensional holomorphic vector bundles. We
have
M
(3.3) Bm (E, F ) = Bm (E, F )[j, k, l]
j,k,l

= Sym∗ (⊕1≤l≤m T ∗ ) ⊗ Sym∗ (⊕0≤l≤m E) ⊗ ∧∗ (⊕0≤l≤m F ),

M
(3.4) Bm (E, F )[j, k] = Bm (E, F )[j, k, l]
l

= Sym∗ (⊕1≤l≤m T ∗ ) ⊗ Symj (⊕0≤l≤m E) ⊗ ∧k (⊕0≤l≤m F ).

x
It is easy to see that D̃ commutes with rαβ . So D̃ is a derivation of the algebra bundle
Bm (E, F ).
Let Om,1 = ⊕k>0 Om [k] and Om,k = (Om,1 )k , the sub bundle of Om generated by product
of k elements of Om with positive weights. Thus
· · · Om,k ⊂ · · · ⊂ Om,2 ⊂ Om,1 ⊂ Om,0 = Om .
Om,k /Om,k+1 ∼
M M
= Symk (⊕1≤l≤m Tl∗ ).
k≥0 k≥0
In general we have
M
A(E, F )[j, k]Om,l /A(E, F )[j, k]Om,l+1 = Bm (E, F )[j, k].
k≥0

4. H OLOMORPHIC STRUCTURES

Let hT ∗ = (−, −)T ∗ be an Hermitian metric of the cotangent bundle T ∗ of X, and hE =

(−, −)E and hF = (−, −)F be Hermitian metrics of E and F respectively. Let ∇T ∗ , ∇E and
∇F be the Chern connections of T ∗ , E and F respectively. For (U, y, e, f ) ∈ U,
X ∗ X X
∇T dyi = θijT dyj , ∇E ei = θijE ej , ∇F fi = θijF fj .
∗
Let ΘT , ΘE and ΘF be their curvatures.
7
Chern connection on Bm (E, F ). By (3.3), there is a canonical Hermitian metric h = (−, −)
on Bm (E, F ) induced from hT ∗ , hE and hF . Let ∇ ˜ = ∇ ˜ 1,0 + ∂¯′ be the Chern connection
associated with h. For (U, y, e, f ) ∈ U, ∇˜ satisfies

˜ (l) =
X ∗ (l)
˜ (l) = E (l) ˜ (l) = (l)
X X
T
(4.1) ∇y i θ y
ij j , ∇ei θ e
ij j , ∇f i θijF fj ,
˜
∇(ab) ˜
= (∇a)b ˜
+ a(∇b), for a, b ∈ ΩU0,0 (Bm (E, F )).

So its curvature Θ̃ satisfies

(l)
X ∗ (l) (l)
X (l) (l)
X (l)
(4.2) Θ̃yi = ΘTij yj , Θ̃ei = ΘE
ij ej , Θ̃fi = ΘFij fj ,
Θ̃(ab) = (Θ̃a)b + a(Θ̃b), for a, b ∈ ΩU0,0 (Bm (E, F )).

A canonical isomorphism. Am (E, F ) and Bm (E, F ) are different holomorphic vector bun-
dles. But as complex vector bundles, they are isomorphic. Here we construct a canonical
isomorphism from Am (E, F ) to Bm (E, F ).
For (U, y, e, f ) ∈ U, let

Hij = (dyi , dyj )T ∗ , HijE = (ei , ej )E , HijF = (fi , fj )F ,

and let (H ij ), (HEij ) and (HFij ) be the inverse matrices of (Hij ), HijE ) and (HijF ) respectively.
For k ≥ 0, let
(k+1) ji (1) (k) ji (0) (k) (0)
X X X
Y(4.3)
l = H lj D k
(H y i ), E l = H E k
lj D (H E
E i ), F l = HljF D k (HFji Fi ).

Proposition 4.1. There is an isomorphism from Φm : Am (E, F ) → Bm (E, F ) of smooth complex

vector bundles, which locally maps Yl (k) , El (k) and Fl (k) to yl (k) , el (k) and fl (k) respectively. This
isomorphism maps Am (E, F )[j, k, l] to Bm (E, F )[j, k, l] and preserves the algebra structures.

(k+1)
Proof. For any point x ∈ X, let (Uα , yα , eα , fα ) ∈ U with x ∈ Uα . By the definition of Yl ,
(k) (k)
El and Fl in (4.3),
(k+1) (k+1)
Yα,l ≡ yα,l mod Om,2 ;
(k) (k) (k) (k)
Eα,l ≡ eα,l , Fα,l ≡ fα,l mod Am (E, F )Om,1.

So Rm (Yα , Eα , Fα ) are the fibres of Am (E, F ) at x.

Let Φm,x : Am (E, F )|x → Bm (E, F )|x be the isomorphism of C-algebras given locally by
(k+1) (k+1) (k) (k) (k) (k)
Φm,x (Yα,l ) = yα,l , Φm,x (Eα,l ) = eα,l , Φm,x (Fα,l ) = fα,l .
(k+1) (k) (k) (k+1) (k) (k)
Yα,l ,Eα,l and Fα,l have the same grades as yα,l ,eα,l and fα,l respectively. So Φm,x
preserves the Z3≥0 grading of each fibre.
P P P
If (Uβ , yβ , eβ , fβ ) ∈ U with x ∈ Uβ , yα,i = fi (yβ ), eα,i = gij eβ,j , fα,i = hij fβ,j on
Uα ∩ Uβ .
X ∂yβ,i ∂yβ,j
Hβ,ij = (dyβ,i , dyβ,j ) = Hα,kl .
∂yα,k ∂yα,l
8
So
(k+1) (1)
X
(4.4) γαβ (Yα,l ) = Hα,lj D k (Hαji γαβ (yα,i ))
X ∂yα,l ∂yα,j k ∂yβ,u uv ∂yβ,v (1)
= ( Hβ,st )D ( Hβ γαβ (yα,i ))
∂yβ,s ∂yβ,t ∂yα,j ∂yα,i
X ∂yα,l ∂yα,j ∂yβ,u k uv ∂yβ,v (1)
= ( Hβ,st )D (Hβ γαβ (yα,i ))
∂yβ,s ∂yβ,t ∂yα,j ∂yα,i
X ∂yα,l (1)
= Hβ,su D k (Hβuv yβ,v )
∂yβ,s
X ∂yα,l (k+1)
= Y .
∂yβ,s β,s
Similarly,
(k) (k) (k) (k)
X X
(4.5) γαβ (Eα,l ) = gls Eβ,s , γαβ (Fα,l ) = hls Fβ,s .
x x
Thus Φm,x ◦ γαβ = γ̃αβ ◦ Φm,x . Φm,x is independent of the choice of (Uα , yα , eα , fα ) ∈ U.
Φm,x smoothly depends on x, so it gives an isomorphism Φm : Am (E, F ) → Bm (E, F ) of
smooth complex vector bundles. Φm,x preserves the grading, so Φm maps Am (E, F )[j, k, l]
to Bm (E, F )[j, k, l]. Φm,x is an isomorphism of C-algebra, so Φm preserves the algebra
structure. 

Holomorphic structures on Am (E, F ). Through the isomorphism Φm , Bm (E, F ) can be

regarded as the same smooth complex vector bundle as Am (E, F ) with a different holo-
morphic structure. Now the underlying complex vector bundle of Am (E, F ) has a Her-
mitian metric (−, −)m ( from Bm (E, F )) and two holomorphic structures. One is from
Am (E, F ), which is determined by ∂¯ with

(4.6) ¯ (l) = 0,
∂y ¯ (l) = 0, ∂f
∂e ¯ (l) = 0,
i i i
¯ ¯ ¯
∂ab = ∂ab + a∂b, 0,0
a, b ∈ Ω (Am (E, F )).
The other is induced from Bm (E, F ), which is determined by ∂¯′ with
(l) (l) (l)
(4.7) ∂¯′ Yi = 0, ∂¯′ Ei = 0, ∂¯′ Fi = 0,
∂¯′ ab = ∂¯′ ab + a∂¯′ b, a, b ∈ Ω0,0 (Am (E, F )).

Under the fixed Hermitian metric (−, −)m on Am (E, F ), let ∇ = ∇1,0 + ∂¯ and ∇ ˜ =
∇˜ 1,0 + ∂¯′ be the Chern connections associated with ∂¯ and ∂¯′ respectively. By (4.1) and (4.2)
˜ satisfies
and Proposition 4.1, for (U, y, e, f ) ∈ U, ∇
˜ (l) =
X ∗ (l)
˜ (l) = E (l) ˜ (l) = (l)
X X
T
(4.8) ∇Y i θ Y
ij j , ∇E i θ E
ij j , ∇F i θijF Fj ,
˜
∇(ab) ˜
= (∇a)b ˜
+ a(∇b), for a, b ∈ Ω0,0
U (Am (E, F )).

Its curvature Θ̃ satisfies

(l)
X ∗ (l) (l)
X (l) (l)
X (l)
(4.9) Θ̃Yi = ΘTij Yj , Θ̃Ei = ΘE
ij Ej , Θ̃Fi = ΘFij Fj ,
Θ̃(ab) = (Θ̃a)b + a(Θ̃b), for a, b ∈ Ω0,0
U (Am (E, F )).

9
 There is also a derivation D̃ : A∞ (E, F ) → A∞ (E, F ) induced from D̃ on Bm (E, F )
through Φ∞ . By Proposition 4.1, for (U, y, e, f ) ∈ U, D̃ satisfies
(j) (j+1) (j) (j+1) (j) (j+1)
D̃Yi = Yi , D̃Ei = Ei , D̃Fi = Fi .
Lemma 4.2.
(1) ˜
X
D = D̃ + Yt ∇ ∂ .
∂yt

P (1) ˜
Proof. D, D̃ and t Yt ∇ ∂ are globally defined operators. To prove the equation, we
∂yt
only need to prove it locally. For (U, y, e, f ) ∈ U,
(k+1) (1)
X
DYl = D(Hlj D k (H ji yi ))
(1) (1)
X X
= (DHlj )D k (H ji yi ) + Hlj D k+1 (H ji yi )
X (1) ∂Hlj (1) (k+2)
= Yt H js Hsm (D k (H mi yi )) + Yl
∂yt
(k+2)
X (1) ∗ ∂
T
= Yl + Yt θls ( )Y (k+1) .
∂yt s
X (1)
= D̃Y k+1 + Yt ∇ ˜ ∂ Y k+1
l ∂yt
l

Similarly,
(k) (1) ˜
X
DEl = D̃Elk + Yt ∇ ∂ Elk
∂yt

(k) (1) ˜
X
DFl = D̃Flk + Yt ∇ ∂ Flk
∂yt

For a smooth function f on U,

X (1)
˜
X (1) ∂f
D̃f + Yt ∇ ∂ f =0+ Yt = Df.
∂yt ∂yt
(1) ˜ (k+1) (k) (k)
are derivations on Ω0,0
P
Since both D and D̃ + Yt ∇ ∂
U (A∞ (E, F )) and Yl , El , Fl
∂yt
(1) ˜
and smooth functions on U generate Ω0,0
P
U (A∞ (E, F )), D is equal to D̃ + Yt ∇ ∂ . 
∂yt

We can find the relation between ∂¯ and ∂¯′ by the above lemma and the following fact.
Lemma 4.3.
¯ = 0,
[D, ∂] [D̃, ∂¯′ ] = 0

¯ and [D̃, ∂¯′ ] are derivations on the sheaf of smooth sections of A∞ (E, F ). Lo-
Proof. [D, ∂]
cally, for (U, y, e, f ) ∈ U and a smooth function f on U,
¯
[D, ∂]y
(k+1) ¯
= [D, ∂]e
(k) ¯
= [D, ∂]f
(k) ¯ = 0.
= [D, ∂]f
l l l

(k+1) (k) (k)

[D̃, ∂¯′ ]Yl = [D̃, ∂¯′ ]El = [D̃, ∂¯′ ]Fl = [D̃, ∂¯′ ]f = 0.

10
 Let K = ∂¯ − ∂¯′ . For any smooth function f on X, Kf = ∂f
¯ − ∂¯′ f = 0. By (4.6) and (4.7),

K(ab) = (Ka)b + a(Kb), for a, b ∈ Ω0,0 (Am (E, F )).

(k+1) (k) (k)

So K is determined if we know KYl , KEl and KFl . Using the following lemma,
we can calculate them out.

Lemma 4.4.

X (1)
˜ ¯ =
X (1) ∂ X (1)
˜ ∂ , K].
[K, D̃] = Y t [∇ ∂ , ∂] −Yt Θ̃( )+ Y t [∇
∂yt ∂yt ∂yt

¯ (1) = 0.
¯ l (1) = ∂y
Proof. ∂Y l

[K, D̃] = [∂¯ − ∂¯′ , D̃]

¯ D̃]
= [∂, (by Lemma 4.3)
X (1)
¯ D−
= [∂, Yt ∇ ˜ ∂ ] (by Lemma 4.2)
∂yt
X (1)
= Y t [∇ ˜ ∂ , ∂]¯ (by Lemma 4.3)
∂yt
X (1)
= Y t [∇ ˜ ∂ , ∂¯′ + K]
∂yt
X (1) ∂ X (1)
˜ ∂ , K].
= −Yt Θ̃( )+ Yt [∇
∂yt ∂yt

1 (j1 +···+ja )!
Let Cj1 ,··· ,ja = − (a−1)! j1 !j2 !···ja !
, if j1 ≥ 0 and jl ≥ 1 for l ≥ 2, otherwise Cj1 ,··· ,ja = 0.

(k) (k+1) (k) (k) ∗

Lemma 4.5. For k ≥ 0, If Pl is Yl , El or Fl , and F is F T , F E or F F respectively, then

k
(k) (j ) (j ) (j )
X X X
KPl = Cj1 ,··· ,ja Fil1 ,··· ,ia Pi1 1 Yi2 1 · · · Yia a .
a=2 j1 +···+ja =k i1 ,··· ,ia

(k) (k+1) (k) (k)

Proof. Here we give the proof for Pl = Yl . The proofs are similar for Pl = El and
(k) (k)
Pl = Fl .
(1) ¯ (1) = ∂y
¯ (1) = 0. Since
We prove it by induction on k. When n = 0, Yl (1) = yl . So ∂Y l l
(1) (1)
∂¯′ Yl = 0, KYl = 0. the lemma is true. Assume the lemma is true for n = k. If n = k + 1,
11
by Lemma 4.4,
(k+2) (k+1)
KYl = K D̃Yl
(k+1)
X (1) ∂ (k+1)
X (1)
˜ ∂ , K]Y (k+1)
= D̃KYl − Yt Θ̃( )Yl + Yt [∇ l (By Lemma4.4)
∂yt ∂yt

k
(j +1) (j ) (j )
X X X
= D̃( Cj1 ,··· ,ja Fil1 ,··· ,ia Yi1 1 Yi2 1 · · · Yia a )
a=2 j1 +···+ja =k i1 ,··· ,ia
X ∂(1) ∗
− Yt ΘTls (
)Y (k+1)
∂yt s
k
X (1) X (j +1) (j ) (j )
X X ∗
+ Yt Cj1 ,··· ,ja (∇T∂ Fa )li1 ,··· ,ia Yi1 1 Yi2 1 · · · Yia a
∂yt
a=2 j1 +···+ja =k i1 ,··· ,ia
k a
(j +1) (j ) (j )
X X X X
= ( Cj1 ,··· ,jl −1,··· ,ja )Fil1 ,··· ,ia Yi1 1 Yi2 1 · · · Yia a
a=2 j1 +···+ja =k+1 i1 ,··· ,ia l=1
(k+1) (1)
−Fi21 ,i2 Yi1 Y i2
k
X X X a (j +1) (j ) (j ) (1)
+ Cj ,··· ,j ,1 F l Y 1 Yi2 1 · · · Yia a Yia+1
a=2 j1 +···+ja =k i1 ,··· ,ia
k + 1 1 a i1 ,··· ,ia ,ia+1 i1
k+1
(j +1) (j ) (j )
X X X
= Cj1,··· ,ja Fil1 ,··· ,ia Yi1 1 Y i2 1 · · · Y ia a .
a=2 j1 +···+ja =k+1 i1 ,··· ,ia

So the lemma is true for n = k + 1. Thus the lemma is true for any k ≥ 0. 

By Lemma 4.5, we immediately have

(k) (k+1) (k) (k) ∗
Lemma 4.6. For k ≥ 0, If Pl is Yl , El or Fl , and F is F T , F E or F F respectively, then
k
˜ 1,0 , K]P (k) = (j ) (j ) (j )
X X X
[∇ l Cj1 ,··· ,ja dyia+1 ∧ Fil1 ,··· ,ia ,ia+1 Pi1 1 Yi2 1 · · · Yia a .
a=2 j1 +···+ja =k i1 ,··· ,ia+1

The mean curvature of Am (E, F ). By Lemma 2.4, the connections on Am (E, F ) satisfy
˜ + K − K ∗.
∇=∇
So their curvatures satisfy

(4.10) ¯ ∇] = [∂¯′ + K, ∇
Θ = [∂, ˜ 1,0 ] − [∂¯′ , K ∗ ] − [K, K ∗ ].
˜ 1,0 − K ∗ ] = Θ̃ + [K, ∇

Now ∇ ˜ 1,0 and K are derivations on the sheaf of smooth sections of Am (E, F ), so [∇
˜ 1,0 , K]
is a derivation on the sheaf of smooth sections of Am (E, F ).
Lemma 4.7. If hT ∗ , hE and hF are Einstein-Hermitian metrics then

˜ 1,0 , K]( ∂ , ∂ )Hij = 0.

X
[∇
∂yi ∂ ȳj
12
 ˜ 1,0 and K are derivations on the sheaf of smooth sections of Am (E, F ), so [∇
Proof. ∇ ˜ 1,0 , K]
P ˜ 1,0
and [∇ , K]( ∂y∂ i , ∂∂ȳj )Hij are derivations on the sheaf of smooth sections of Am (E, F ).
(k) (k+1) (k) (k) ∗
From Lemma 4.6, if Pl is Yl , El or Fl , and F is F T , F E or F F respectively,
k
˜ (k) ∂ ∂ ∂ (j ) (j ) (j )
X X X
1,0
[∇ , K]Pl ( , )Hij = Cj1 ,··· ,ja Fil1 ,··· ,ia ,i ( )Hij Pi1 1 Yi2 1 · · · Yia a .
∂yi ∂ ȳj a=2 j +···+j
∂ ȳj
1 a =k i1 ,··· ,ia
∗
By Lemma 2.2, if hT ∗ , hE and hF are Einstein-Hermitian metrics, then for n ≥ 3, MnT , MnE
and MnF all vanish. So
X ∂
Fil1 ,··· ,ia ,i ( )Hij = 0.
∂ ȳj
P ˜ 1,0 (k)
Thus [∇ , K]P ( ∂ , ∂ )Hij = 0. For any smooth function f on X,
l ∂yi ∂ ȳj

˜ 1,0 , K]f = ∇
[∇ ˜ 1,0 Kf + K ∇
˜ 1,0 f = K∂f = 0.
P ˜ 1,0
Since [∇ , K]( ∂y∂ i , ∂∂ȳj )Hij is a derivation on the sheaf of smooth sections of Am (E, F ),
it is zero. 

We have the following theorem for the mean curvature of the connection ∇ on Am (E, F ).
Theorem 4.8. If hT ∗ , hE and hF are Einstein-Hermitian metrics then the mean curvature of ∇ is
X ∂ ∂ X ∂ ∂ X ∂ ∂ X ∂ ∂
Θ( , )Hij = Θ̃( , )Hij + K( )K ∗ ( )Hij − K ∗( )K( )Hij .
∂yi ∂ ȳj ∂yi ∂ ȳj ∂ ȳj ∂yi ∂yi ∂ ȳj
In particular, if the mean curvatures of hT ∗ , hE and hF are zero,
X ∂ ∂ X ∂ ∂ X ∂ ∂
Θ( , )Hij = K( )K ∗ ( )Hij − K ∗( )K( )Hij .
∂yi ∂ ȳj ∂ ȳj ∂yi ∂yi ∂ ȳj
˜ 1,0]( ∂ , ∂ )Hij = 0, [∂¯′ , K ∗ ]( ∂ , ∂ )Hij also vanishes since
P P
Proof. By Lemma 4.7, [K, ∇ ∂yi ∂ ȳj ∂yi ∂ ȳj
˜ 1,0 ]( ∂ , ∂ )Hij . This proves the first equation by (4.10). By
P
it is the conjugacy of [K, ∇ ∂yi ∂ ȳj
∂ ∂
P
(4.9), Θ̃( ∂yi , ∂ ȳj )Hij vanishes if the mean curvatures of hT ∗ , hE and hF vanish. The
second equation follows. 

5. H OLOMORPHIC SECTIONS

In this section, we assume X is a compact Hermitian manifold. We calculate the space

of holomorphic sections of Am (E, F ) if the mean curvatures of hT ∗ , hE and hF are zero.
Lemma 5.1. Let E be a holomorphic vector bundle with a Hermitian metric h = (−, −) and its
Chern connection is ∇ = ∇1,0 + ∂. ¯ Assume E has another holomorphic holomorphic structure
determined by ∂¯′ = ∂¯ − K with its Chern connection ∇ ˜ = ∇ ˜ 1,0 + ∂¯′ , such that under this
holomorphic structure,
(1) E = k∈Z E k , E k are holomorphic vector bundles;
L
(2) E k is perpendicular to E l for k 6= l;
(3) the mean curvature of ∇ ˜ is zero.
0,0 0,1
(4) K maps Ω (El ) to Ω (⊕∞ k=l+1 Ek ).
˜ , K]( ∂ , ∂ )Hij = 0.
P 1,0
(5) [∇ ∂yi ∂ ȳj

13
Then a ∈ Ω0,0 ˜
X (E) is holomorphic if and only if ∇a = 0 and Ka = 0.

Proof. By condition (1), any smooth section P a of E can be uniquely written as a finite sum
of smooth sections ak of E k , that is a = k ak where only a P finite number of ak not zero.
By condition (2), any smooth sections a and b of E, (a, b) = k (ak , bk ). For any λ ≥ 0, let
hλ = (−, −)λ be the Hermitian metric of E given by
X
(a, b)λ = λk (ak , bk ).
k

Under this Hermitian metric, Let ∇λ and ∇ ˜ λ are the Chern connections corresponding to
∂¯ and ∂¯′ respectively. Since on E k , hλ is a rescale of h, ∇
˜ λ = ∇.
˜
Let Kk,l ∈ Ω0,1 (Hom(E, E)) with Kk,l a = (Kak )l . By condition (4), K = k<l Kk,l . Let
P
∗
Kk,l ∈ Ω1,0 (Hom(E, E)) be the dual of Kk,l , so

∂ ∂ ∂ ∂
(Kk,l ( )a, b)λ = λl (Kk,l ( )ak , bl ) = λl (ak , Kk,l
∗
( )bl ) = λl−k (a, Kk,l
∗
( )b)λ .
∂ ȳi ∂ ȳi ∂yi ∂yi
Let
X
Kλ∗ = λl−k Kk,l
∗
,
k<l

˜ + K − K ∗ , so its curvature
we have (K( ∂∂ȳi )a, b)λ = (a, Kλ∗ ( ∂y∂ i )b)λ . By Lemma 2.4, ∇λ = ∇ λ
is
¯ ∇] = [∂¯′ + K, ∇
Θλ = [∂, ˜ 1,0 − Kλ∗ ] = [∂¯′ , ∇] ˜ 1,0] − [∂¯′ , Kλ∗ ] − [K, Kλ∗ ].
˜ + [K, ∇

By conditions (3) and (5) and the proof of Theorem 4.8, the mean curvature of ∇λ is
X ∂ ∂ X ∂ ∂ X ∂ ∂
M= Θλ ( , )Hij = K( )Kλ∗ ( )Hij − Kλ∗ ( )K( )Hij .
∂yi ∂ ȳj ∂ ȳj ∂yi ∂yi ∂ ȳj

Apply the Weitzenböck formula 2.1 to a smooth section a ∈ Ω0,0 X (E). We have
Z Z
λ λ
(∆∂¯a , a) = (∇λ∗ ∇λ a − Ma, a)λ
X
ZX
= (∇λ a, ∇λ a)λ − (Kλ∗ a, Kλ∗ a)λ + (Ka, Ka)λ
ZX
= ¯ ∂a)
(∇λ1,0 a, ∇λ1,0 a)λ + (∂a, ¯ λ − (K ∗ a, K ∗ a)λ + (Ka, Ka)λ .
λ λ
X

Now
Z Z
(5.1) ¯ ∂a)
(∂a, ¯ λ ≥ 0, (Ka, Ka)λ ≥ 0.
X X

Let
Z
P (λ) = (∇λ1,0 a, ∇λ1,0 a)λ − (Kλ∗ a, Kλ∗ a)λ .
X

14
 Assume ak = 0 for k < l or k > m, al 6= 0 and am 6= 0.
(∇λ1,0 a, ∇λ1,0 a)λ − (Kλ∗ a, Kλ∗ a)λ
˜ 1,0 − K ∗ )a, (∇
= ((∇ ˜ 1,0 − K ∗ )a)λ − (K ∗ a, K ∗ a)λ
λ λ λ λ
˜ 1,0 ˜ 1,0 λ ˜
= (∇ a, ∇ a) − (∇ a, K a) − (K a, ∇1,0 a)λ .
1,0 ∗ λ ∗ ˜
λ λ
m
˜ 1,0 al , ∇
˜ 1,0 al ) − ˜ 1,0 ak , K ∗ ′ ak′ ) − ˜ 1,0 ak ).
X X ′
X
= λl (∇ λk (∇ k,k
∗
(Kk,k ′ ak ′ , ∇

k=l m≥k ′ >k≥l m≥k ′ >k≥l

λ−l P (λ) is a polynomial of λ with constant term

Z
˜ 1,0 al , ∇
(∇ ˜ 1,0 al ) ≥ 0.
X

If a is a holomorphic section, we have X (∆λ∂¯a, a)λ = 0 for any λ > 0. By (5.1), P (λ) ≤ 0.
R
We must have Z
˜ 1,0 al , ∇
(∇ ˜ 1,0 al ) = 0.
X
So
∇˜ 1,0 al = 0.
Kλ∗ al ∈ Ω1,0 k ˜ 1,0
L
X ( k<l E ), which is perpendicular to ∇ ak , k ≥ l. Thus
Z X m
˜ 1,0 al , ∇
˜ 1,0 al )− ˜ 1,0 ak , K ∗ ′ ak′ )− ˜ 1,0 ak )).
X ′
X
P (λ) = ( λl (∇ λk (∇ k,k
∗
(Kk,k ′ ak ′ , ∇
X k=l+1 m≥k ′ >k≥l+1 m≥k ′ >k≥l+1

Induction on l, we can show that ∇˜ 1,0 a = 0 and P (λ) = 0.

Thus to keep X (∆∂¯a, a)λ = 0, the left side of the inequality (5.1) is zero. So
R

Ka = 0, ¯ = 0.
and ∂a
∂¯′ a = ∂a
¯ − Ka = 0 and ∇a˜ =∇ ˜ 1,0 a + ∂¯′ a = 0.
˜ = 0 and Ka = 0. Since the mean curvature of ∇
On the other hand, suppose ∇a ˜ is zero,
¯′ ¯ ¯′
by the Weitzenböck formula (2.1), ∂ a = 0. Therefore ∂a = ∂ a + Ka = 0, so that a is a
holomorphic section of E. 

Let
M
(5.2) E n = Φ−1
m ( Symj (⊕1≤l≤m T ∗ ) ⊗ Symk (⊕0≤l≤m E ∗ ) ⊗ ∧l (⊕0≤l≤m F ∗ )).
j+k+l=n

k
Am (E, F ) = ⊕∞ ∗ ∗
k=0 E . Let Am (E, F )[i, j, k] be the dual vector bundle of Am (E, F )[i, j, k].
Let
A∗m (E, F ) = ⊕i,j,k A∗m (E, F )[i, j, k].
We have the following theorem:
Theorem 5.2. If X is a compact Hermitian manifold and the mean curvatures of hT ∗ , hE and hF
˜ = 0 and Ka = 0.
are zero, then a smooth section a of Am (E, F ) is holomorphic if and only if ∇a
˜
A smooth section a of Am (E, F ) is holomorphic if and only if ∇a = 0 and K ∨ a = 0.
∗

15
Proof. We use Lemma 5.1 to prove the theorem. Am (E, F ) is a holomorphic bundle with
the Hermitian metric h and its Chern connection is ∇. Am (E, F ) has another holomor-
phic structure coming from Bm (E, F ) by Φm . Its Chern connection is ∇. ˜ E k are holo-
morphic under the holomorphic structure from Bm (E, F ). E k is perpendicular to E l if
k
k 6= l. Am (E, F ) = ⊕∞ k=0 E . Let E
−k
= E k∗ , then A∗m (E, F ) = ⊕0−∞ E k . So Am (E, F ) and
A∗m (E, F ) satisfies condition (1) and (2). Since the mean curvature of hT ∗ , hE and hF van-
ish, by (4.9), A∗m (E, F ) satisfies condition (3). By Lemma 4.5 and Lemma 4.7, A∗m (E, F )
satisfies conditions (4) and (5). Since E −k is the dual of E k , it is easy to see that A∗m (E, F )
satisfies conditions (3), (4) and (5). So by Lemma 5.1, the theorem is true.


When the Hermitian metric of E, F and T ∗ are flat, then K = 0 and a smooth section b
˜ = 0. In particular
of Am (E, F )∗ is holomorphic if and only if ∇a
Corollary 5.3. If X is a flat Kähler torus, E and F are sums of some copies of tangent and cotan-
gent bundles of X, then a smooth section a of Am (E, F ) is holomorphic if and only if ∇a ˜ = 0. So
the space of holomorphic sections of Am (E, F ) is isomorphic to its fibre by restriction.

g[t] invariants. Let g be a Lie algebra. Let

M
g[t] = gtn
n≥0

be the Lie algebra given by

[gi ti , gj tj ] = [gi , gj ]ti+j , for gi , gj ∈ g.
Let gm = g[t]/(tm+1 ). The following lemma is obvious.
Lemma 5.4. if g is simple, then gm is generated by g and an element K of gt. So if R is a
representation of gm , then the invariant subspace of R is
Rgm = {r ∈ Rg |Kr = 0}.

If g is the Lie algebra of a connected algebraic group G, then gm is the Lie algebra of the
Lie group Jm (G). Rgm = RJm (G) .
Let G = SU(N) be the special unitary group on V = CN and g = sl(N, C) be its
complexified Lie algebra (or G = Sp(N) be the symplectic group on V = C2N and
g = sp(2N, C) be its complexified Lie algebra). By the Weyl’s dimension formula for
the finite irreducible representation of g, we have the following lemma about the repre-
sentation of g = sl(N, C) (or sp(2N, C)):
LemmaT5.5. Let V be the fundamental representation of g. Regarding g as a subspace of V ∗ ⊗ V ,
V ∗ ⊗ g Sym2 V ∗ ⊗ V is an irreducible representation of g.

Now we assume X is an N (or 2N) dimensional compact Kähler manifold with the
holonomy group G = SU(N) (or G = Sp(N)). Then the mean curvature of hT ∗ is zero.
Let E and F be sums of some copies of the holomorphic tangent and cotangent bundles
of X. Then the mean curvatures of hE and hF are zeros. The holonomy group of Am (E, F )
˜ is G. By Theorem 5.2, a is a holomorphic section of Am (E, F ) if and only if ∇a
for ∇ ˜ =0
˜ ˜
and Ka = 0. ∇a = 0 means that a is a parallel section of ∇ in Am (E, F ). The space of the
16
 ˜ in Am (E, F ) is isomorphic to (Am (E, F )|x )G , for any point x ∈ X.
parallel sections of ∇
The isomorphism is given by the restriction. So the restriction
rx : Γ(Am (E, F )) → (Am (E, F )|x )G
is injective. Let R = Am (E, F )|x be the fibre of Am (E, F ) at x. Let g be the Lie algebra
sl(N, C)) (or sp(2N, C ). The action of G on R induces the action of g on R, then we have
the action of gm on R given by
gtk (ab) = (gtk a)b + a(gtk b), for a, b ∈ R
(j) (j+1) (j) (j)
and if Pi is Yi ,Ei or Fi ,
(j)
gtk (Pi ) = 0, for j < k
j! (j−k)(j)
gtk (Pi ) =
g(Pi ), for j ≥ k.
(j − k)!
We have the following theorem on the holomorphic sections of Am (E, F ).
Theorem 5.6. Let X be an N (or 2N) dimensional compact Kähler manifold with holonomy
group G = SU(N) (or G = Sp(N)). Let E and F be sums of some copies of the holomorphic
tangent and cotangent bundles of X. Then the image of rx : Γ(Am (E, F )) → (Am (E, F )|x )G is
(Am (E, F )|x )gm . So rx induces an isomorphism from Γ(Am (E, F )) to (Am (E, F )|x )gm .
˜ with rx (a) ∈ (Am (E, F )|x )gm for a point x ∈ X. For
Proof. If a is a parallel section of ∇
any point x ∈ X, let’s prove that rx′ (a) ∈ (Am (E, F )|x′ )gm . Let γ be a path from x to
′

x′ . Through the parallel transformation along the path γ under the connection ∇, ˜ we get
an isomorphism of G representation from Am (E, F )|x to Am (E, F )|x′ . From the definition
of the action gm , we actually get an isomorphism of gm representation. Now since a is a
parallel section, the isomorphism maps a|x to a|x′ . Thus a|x is gm invariant if and only if
a|x′ is gm invariant.
The holonomy group of Am (E, F ) for ∇ ˜ is G. Through the action of G on Am (E, F )|x ,
we can regard g as a subspace of End(Am (E, F )|x, (Am (E, F )|x ). We have Θ̃( ∂y∂ i , ∂∂ȳj )|x ∈ g.
A (E,F )
Let F̃n = Fn m be the higher covariant derivative of Θ̃. Then
∂ ∂ ∂
gi2 ···ia ,j = F̃n ( ,··· , , )|x ∈ g.
∂yi2 ∂yin ∂ ȳj
By the expression of K in Lemma 4.5,
k
∂ (k)
X X X ∂ ∂ ∂ (k−j ) (j ) (j )
K( )P |x = Cj1 ,··· ,ja (F̃n ( ,··· , , )|x Pl 1 )Yi2 2 · · · Yia a
∂ ȳj l a=2
∂yi2 ∂yia ∂ ȳj
j1 +···+ja =k i2 ,··· ,ia
k
X X X 1 (k) (j ) (j )
= (gi2 ···ia ,j tk−j1 Pl )Yi2 2 · · · Yia a
a=2 j1 +···+ja =k i2 ,··· ,ia
(a − 1)!j2 ! · · · ja !

So
k
∂ X X X 1 (j ) (j )
(5.3) K( )|x = Yi2 2 · · · Yia a gi2 ···ia ,j tk−j1 .
∂ ȳi a=2 j +···+j
(a − 1)!j2 ! · · · ja !
1 a =k i2 ,··· ,ia

Since a|x is gm invariant for any x ∈ X, so K(a) = 0. By Theorem 5.2, a ∈ Γ(Am (E, F )).
17
 Pn
On the other hand, if a ∈ Γ(Am (E, F )), a = k=l ak with ak ∈ Ω0,0 (E k ). Let
X Yi(j2 ) ∂ ∂ j2
2
K̃i = Θ̃( , )t .
j ≥1,i
j2 ! ∂y i 2 ∂ ȳ i
2 2

Then
∂
K( )a|x = K̃i al |x + b, b ∈ ⊕k>l E k |x .
∂ ȳi
So Ka = 0 implies K̃i al |x = 0.
˜ preserves E k , the action of the holomomy group G on the fibre of Am (E, F )
Since ∇
preserves the grading. So al |x ∈ (E l |x )G . Since the holonomy group of X is G, the
curvature R of T is not equals to zero. There is some point x ∈ X , and some i0 , R( ∂ ȳ∂i )|p 6=
0
0.
The action of g on Ki0 forms a representation W of g. Every element of W is a derivation
of Ω0,0 (Em ) and it vanishes on al |p .
Now for any g ∈ g.
(j ) (j )
X Y i2 2 ∂ ∂ j2 Y i2 2 ∂ ∂
)]tj2 )


[g, K̃i0 ] = (g )(Θ̃( , )t ) + ([g, Θ̃( ,
j2 ≥1,i2
j2 ! ∂ ȳi0 ∂yi2 j2 ! ∂ ȳi0 ∂yi2
(j ) (j )
X Y 2 ∂ ∂ j2 Y 2 ∂ ∂
= − i2 (Θ̃( ,g )t ) + i2 ([g, Θ̃( , )]tj2 )
j2 ≥1,i2
j2 ! ∂ ȳi0 ∂yi2 j2 ! ∂ ȳi0 ∂yi2

Thus the action of g on Ki0 is the same as the action of g on Θ̃( ∂ ȳ∂i ). Since E and F are
0
direct sums of copies of holomorphic tangent and cotangent bundles. By (4.9), the action
of g on Θ̃( ∂ ȳ∂i ) is the same as its action on R( ∂ ȳ∂i ) ∈ Sym2 Tx∗ ⊗ Tx Tx∗ ⊗ g. By Lemma 5.5,
T
0 T 0
Sym2 Tx∗ ⊗ Tx Tx∗ ⊗ g is an irreducible representation of g. So

W ∼
\
= Sym2 Tx∗ ⊗ Tx Tx∗ ⊗ g.

Let g1 , g2 ∈ g, which are determined by

∂ ∂ ∂ ∂ ∂
g1 ( ) = δi1 , g2 ( ) = δi1 − δis .
∂yi ∂ys ∂yi ∂y1 ∂ys
here s = 2 if g = sl(N, C) and s = N + 1 if g = sp(2N, C). We have [g2 , g1 ] = −2g1 . Then
the two derivations
X Y (j) X Ys(j) X Y (j)
1 1
K1 = g1 tj , K2 = − g1 tj + g2 tj
j≥1
j! j≥1
j! j≥1
j!

are in W . We have K1 al |x = 0 and K2 al |x = 0. By Lemma 5.7, al |x is gm invariant. Then

we have Kal = 0 . Now K(a − al ) = 0. By induction on l, we can show ax is gm invariant.
This proves the theorem.

Lemma 5.7. if a ∈ Am (E, F )|x is g invariant and K1 a = 0, K2 a = 0, then a is gm invariant.
18
Proof. Let Pn be a derivation on Am (E, F )|x defined inductively by
P 1 = K1 , Pn = [Pn−1 , K2 ].
We have Pn a = 0. Now
(j) j−1 (j−s) (s)
Y1 1X Y1 Y1
[ , K2 ] = ,
j! j s=1 (j − s − 1)!) s!
j−1 (s)
j
X Y 1
[g1 t , K2 ] = 2 g1 tj+s .
s=1
s!
By the above two equations, we can show inductively that
(j ) (j )
X
Pn = cj1 ,··· ,ja Y1 1 · · · Y1 n g1 tj1 +···+jn
ji ≥1

with cj1 ,··· ,jn > 0. a has finite weight, when k large enough gtk a = 0 for any g ∈ g. Let L
be the largest number such that a is not gtL a invariant. if L ≥ 1,
(1)
PL a = c1,··· ,1 (Y1 )L g1 tL a.
So g1 tL a = 0. Since a is g invariant and gtL is an irreducible representation of g, a is gtL
invariant. But we assume that a is not gtL invariant. So L ≤ 0. we conclude that a is gm
invariant. 

Now we can prove Theorem 1.1.

Proof of theorem 1.1. Let g = sl(N, C). T Jm (X) is isomorphic to Jm (T X). The sheaf of
sections of Am (T, 0)[1, 0] over X is the sheaf of the push forward of the sheaf of sections
of Jm (T X) through πm : Jm (X) → X. Thus the space of holomorphic vector fields of
Jm (X) is isomorphic to Γ(Am (T, 0)[1, 0]). By Theorem 5.6, Γ(Am (T, 0)) is isomorphic to
(Am (T, 0)|x )gm .
Am (T, 0)|x ∼
(1) (j) (m) ∗(0) ∗(j) ∗(m)
= R(y, y ∗, 0) = C[y1 , · · · , yi , · · · , yN , y1 , · · · , yi , · · · , yN ],
So Γ(Am (T, 0)[1, 0]) is isomorphic to the subspace of (Am (T, 0)|x )gm with the degree of y ∗ is
(1) (1) ∗(0) ∗(0)
one. By Theorem 4.3 of [7], (Am (T, 0)|x )gm is generated by C[y1 , · · · , yN , y1 , · · · , yN ]g ,
P (1) ∗(0)
which is generated by v = yi yi . So
m−1
Γ(Am (T, 0)[1, 0]) ∼
M
= CD̃ k v,
k=0

has dimension m. 

Theorem 5.6 can be generalized to the case when X is a compact Ricci-flat Kähler man-
ifold.
Theorem 5.8. Let X be compact Ricci-flat Kähler manifold with holonomy group G. Let g be
the complexified Lie algebra of G. Let E and F be sums of some copies of the holomorphic tan-
gent and cotangent bundles of X. Then the image of rx : Γ(Am (E, F )) → (Am (E, F )|x )G is
((Am (E, F )|x)gm )G . So rx induces an isomorphism from Γ(Am (E, F )) to ((Am (E, F )|x )gm )G .
19
Proof. It is well known that if X is a compact Ricci-flat Kähler manifold, then it admits a
finite cover p : X̃ → X,
X̃ = T × X1 × · · · × Xk ,
where T is a flat Kähler torus, and Xi has holonomy group SU(mi ) or Sp( m2i ) with dim Xi =
mi . (See for example page 124 of [3].) If E and F are sums of some copies of the holo-
morphic tangent and cotangent bundles of X, then the pullback p∗ E and p∗ F are sums of
some copies of the holomorphic tangent and cotangent bundles of X̃. We have
p∗ Am (E, F ) = Am (p∗ E, p∗ F ).
By Theorem 5.6, Corollary 5.3 and the fact that holonomoy group of T is trivial, rx̃ induces
an isomorphism from Γ(Am (p∗ E, p∗ F )) to (Am (p∗ E, p∗ F )|x̃ )gm .
The pullback gives an imbedding p∗ : Γ(X, A(E, F )) → Γ(X̃, A(p∗ E, p∗ F ). By The-
orem 5.2, any element ã ∈ Γ(X̃, A(p∗ E, p∗ F ) satisfies ∇a˜ = 0. So ã = p∗ a for some
smooth section a of A(E, F ) if and only if a|x̃ is G invariant. Since ã is holomorphic,
a must be holomorphic. Thus rx̃ ◦ p∗ induces an isomorphism from Γ(X, A(E, F )) to
((Am (p∗ E, p∗ F )|x̃ )gm )G , which means rx induces an isomorphism from Γ(Am (E, F )) to
((Am (E, F )|x)gm )G . 

G has a connected component G0 which contain the identity. G0 is a normal subgroup

of G and it acts trivially on (Am (E, F )|x )gm . So ((Am (E, F )|x )gm )G = ((Am (E, F )|x )gm )G/G0

6. T HE GLOBAL SECTIONS OF CHIRAL DE R HAM COMPLEX ON K3 SURFACES

In this paper, we will follow the formalism of the vertex algebra developed in [4]. We
also use the notation in our earlier paper [8] and [12].

Chiral de Rham complex. The chiral de Rham complex [9, 10] is a sheaf of vertex algebras
Ωch
X defined on any smooth manifold X in either the algebraic, complex analytic, or C
∞

categories. In this paper we work exclusively in the complex analytic setting. For the
vertex algebra, we refer [4].
Let ΩN be the tensor product of N copies of the βγ−bc system. It has 2N even generators
β 1 (z), · · · , β N (z), γ 1 (z), · · · , γ N (z) and 2N odd generators b1 (z), · · · , bN (z), c1 (z), · · · , cN (z).
Their nontrivial OPEs are
δji δji
β i (z)γ j (w) ∼ , bi (z)cj (w) ∼ .
z−w z−w
Given a coordinate system (U, γ 1 , · · · γ N ) of X, C[γ 1 , · · · γ N ] ⊂ O(U) can be regarded as a
subspace of ΩN by identifying γ i with γ i (z) ∈ ΩN . As a linear space, ΩN has a C[γ 1 , · · · γ N ]
module structure. Ωch
X (U) is the localization of ΩN on U,

Ωch
X (U) = ΩN ⊗C[γ 1 ,···γ N ] O(U).

Then Ωch i i i
X (U) is the vertex algebra generated by β (z), b (z), c (z) and f (z), f ∈ O(U). These
generators satisfy the nontrivial OPEs
∂f
i ∂γ i
(z) i j
δji
β (z)f (w) ∼ , b (z)c (w) ∼ ,
z−w z−w
20
as well as the normally ordered relations

: f (z)g(z) : = f g(z), for f, g ∈ O(U).

Ωch
X (U) is spanned by the elements
(6.1)
: ∂ k1 β i1 (z) · · · ∂ ks β is (z)∂ l1 bj1 (z) · · · ∂ lt bjt ∂ m1 cr1 (z) · · · ∂ n1 γ s1 (z) · · · f (γ)(z) :, f (γ) ∈ O(U).

Let γ̃ 1 , · · · γ̃ N be another set of coordinates on U, with

γ̃ i = f i (γ 1 , · · · γ N ), γ i = g i (γ̃ 1 , · · · γ̃ N ).

The coordinate transformation equations for the generators are

X ∂f i
∂γ̃ i (z) = : j (z)∂γ j (z) : ,
∂γ
X ∂g j
b̃i (z) = : i
(g(γ))bj : ,
∂γ̃
X ∂f i
(6.2) c̃i (z) = : j (z)cj (z) : ,
∂γ
X ∂g j X ∂ ∂g j
β̃ i (z) = : (g(γ))(z)β j
(z) : + :: ( (g(γ)))(z)ck (z) : bj (z) : .
∂γ̃ i ∂γ k ∂γ̃ i

From [12], Ωch

X has an increasing filtration Qn , n ∈ Z≥0 and its associated graded sheaf
is
M
gr(Q) = Qn /Qn−1 .
n

Locally, Qn (U) is spanned by the elements with only at most n copies of β and b , i.e. the
elements in equation (6.1) with s + t ≤ n.
Then the associated graded object
M
(grQ)(U) = Qn (U)/Qn−1 (U)
n

is a ∂-ring. A Z≥0 graded, associative, super-commutative algebra equipped with a deriva-

tion ∂ of degree zero is called ∂-ring. On (grQ)(U), the product and the derivation ∂ are
induced from the wick product and ∂ from Ωch X (U), respectively.
For each n ≥ 0, let
φn : Qn (U) → Qn (U)/Qn−1 (U).
be the projection. As a ring with a derivation, gr(Q(U)) is generated by

β i = φ1 (β i (z)), bi = φ1 (bi (z)), ci = φ0 (ci (z)), φ0 (f (z)), f ∈ O(U).

Qn (U)/Qn−1 (U) is spanned by

(6.3) a = ∂ k1 β i1 · · · ∂ ks β is ∂ ks+1 bis+1 · · · ∂ kn bin ∂ m1 cr1 · · · ∂ n1 γ s1 · · · ∂ nt γ st f (γ), f (γ) ∈ O(U).

21
 For the sheaf gr(Q), the coordinate transformation equations of β, γ, b, c are
X ∂f i
∂ n γ̃ i = ∂ n−1 ( j ∂γ j ),
∂γ
X ∂g j
∂ n b̃i = ∂ n ( i (g(γ))bj ),
∂γ̃
X ∂f i
(6.4) ∂ n c̃i = ∂ n ( j cj ),
∂γ
X ∂g j X ∂ ∂g j
∂ n β̃ i = ∂ n ( j (g(γ))β i ) + ∂ n ( k ( i (g(γ)))ck bj ).
∂γ̃ ∂γ ∂γ̃
The only difference between these equations and those of the chiral de Rham sheaf is that
the Wick product is replaced by the ordinary product in an associated super commutative
algebra.
gr(Q) has an increasing filtration gr(Q)n,s , 0 ≤ s ≤ n and its associated graded sheaf is
M
gr2 (Q) = gr(Q)n,s /gr(Q)n,s−1 .
n,s

Locally, gr(Q)n,s (U) is spanned by all elements a ∈ Qn (U)/Qn−1 (U) of the form (6.3)
with the number of β less or equal than s.
The associated graded object
M
gr2 (Q)(U) = gr(Q)n,s (U)/gr(Q)n,s−1 (U)
n,s

is a ∂-ring. The product and the derivation ∂ are induced from the product and derivation
of gr(Q)(U).
Let
ψn,s : gr(Q)n,s (U) → gr(Q)n,s (U)/gr(Q)n,s−1 (U)
be the projection as a ring with a derivation,gr2 (Q(U)) is generated by
ψ1,1 (β i ), ψ1,0 (bi ), ψ0,0 (ci ) and ψ0,0 (f (γ)), f (γ) ∈ O(U).
When no confusion arise, the symbols β i , γ i , bi , ci in gr(Q)(U) will also be used to denote
the corresponding elements ψ1,1 (β i ), ψ0,0 (γ i ), ψ1,0 (bi ), ψ0,0 (ci ) in gr2 (Q)(U).
For the sheaf gr2 (Q), the relations of β, γ, b, c under the coordinate transformation are
X ∂f i
∂ n γ̃ i = ∂ n−1 ( j ∂γ j ),
∂γ
X ∂g j
∂ n b̃i = ∂ n ( i (g(γ))bj ),
∂γ̃
X ∂f i
(6.5) ∂ n c̃i = ∂ n ( j cj ),
∂γ
X ∂g j
∂ n β̃ i = ∂ n ( i (g(γ))β j ).
∂γ̃
By these coordinate transformation equations, we have
Proposition 6.1. gr2 (Q) is exactly the sheaf A∞ (T, T ⊕ T ∗ ).
22
 From [12], we have the following reconstruction properties for the holomorphic sec-
tions of Ωch
X.

Lemma 6.2. If ai ∈ Qni (X), for 1 ≤ i ≤ l, such that φni (ai ) generate gr(Q)(X) as a ∂-ring,
then φn is surjective, and it therefore induces an isomorphism
Qn (M)/Qn−1 (X)∼ =Qn /Qn−1 (X).
Furthermore, {ai | 1 ≤ i ≤ l} strongly generates the vertex algebra Ωch
X (M).

Lemma 6.3. If ai ∈ gr(Q)ni ,si for 1 ≤ i ≤ l, such that ψni ,si (ai ) generate gr(Q)(X) as a ∂-ring,
then ψn,s is surjective, and it induces an isomorphism
gr(Q)n,s (X)/gr(Q)n,s−1(X)∼ =gr(Q)n,s /gr(Q)n,s−1(X).
Furthermore, {ai | 1 ≤ i ≤ l} generate gr(Q)(X) as a ∂-ring.
Holomorphic sections of the chiral de Rham complex. If X is a Calabi-Yau manifold,
according to [9], Ωch
X (X) is a topological vertex algebra. There are four global sections
Q(z), L(z), J(z), G(z). Locally, they can be represented by
N
X N
X
i i
Q(z) = : β (z)c (z) :, L(z) = (: β i (z)∂γ i (z) : − : bi (z)∂ci (z) :),
i=1 i=1
N
X N
X
i i
J(z) = − : b (z)c (z) :, G(z) = : bi (z)∂γ i (z) :,
i=1 i=1
According to [1], if ω is a nowhere vanishing holomorphic N form of X, there are two
global sections D(z) and E(Z) of Ωch X can be constructed. Locally, if (U, γ) are coordinate
1 N
of U such that ω = dγ ∧ · · · ∧ dγ , D(z) and E(Z) can be represented by
D(z) =: b1 (z)b2 (z) · · · bN (z) :, E(z) =: c1 (z)c2 (z) · · · cN (z) : .
Let B(z) = Q(z)(0) D(z), C(z) = G(z)(0) E(z). It is easy to see that these eight sections Q(z),
L(z), J(z), G(z), D(z), E(z), B(z) and C(z) always close (nonlinearly) under operator
product expansion. Let VN be the vertex algebra generated by these eight sections. When
N = 2, V2 is an N = 4 superconfomal vertex algebra with central charge 6. (Here the
Virasoro field is L(z) − 21 ∂J(z).) We will show that if X is a K3 surface, these eight sections
strongly generate ΩchX (X) as a vertex algebra.
Let Ω̄N be the subalgebra of the ΩN , which is generated by β i (z), ∂γ i (z), bi (z), ci (z). It
is a tensor product of a system of free bosons and a system of free fermions. V0 is a
subalgebra of Ω̄N . On Ω̄N there is a positive definite Hermitian form (−, −) with the
following property:
xi x′
(6.6) (β(n) A, B) = (A, α(−n)
i
B), for any n ∈ Z, n 6= 0, ∀A, B ∈ Ω̄N ;
x′i
(bx(n)
i
A, B) = (A, c(−n−1) B), for any n ∈ Z, ∀A, B ∈ Ω̄N .
It is easy to check
(6.7) Q∗(n) = G(−n+1) , ∗
J(n) = J(−n) ,
N(N−1)
L∗(n) = L(−n+2) − (n − 1)J(−n+1) , ∗
D(n) = (−1) 2 E(N −2−n) ,
Let L̃ = L − 21 ∂J. L̃ is a Virasoro field with central charge 3N. L̃∗(n) = L̃(−n+2) . We can
conclude
23
Lemma 6.4. Ω̄N is a unitary representation of the Lie algebra generated by
{Q(n) , G(n) , J(n) , L(n) , D(n) , E(n) , B(n) , C(n) }n∈Z
and VN is a simple conformal vertex algebra with central charge 3N.
Assume X is an N dimensional complex manifold with holonomy group SU(N). By
sl(N,C)[t]
Theorem 5.6 and Proposition 6.1, gr2 (Q)(X) is isomorphic to gr2 (Q)|x . gr2 (Q)|x is a
∂-ring, which is isomorphic to
R = C[∂ k β i , ∂ k+1 γ i , ∂ k bi , ∂ k ci ], k ≥ 0, 1 ≤ i ≤ N.
So to calculate the holomorphic sections of gr2 (Q) is to calculate Rsl(N,C)[t] . It is easy to see
that the following eight elements of R are sl(N, C)[t] invariant.
N
X N
X
i i
βc, β i ∂γ i ,
i=1 i=1
N
X N
X
i i
− bc, bi ∂γ i ,
i=1 i=1
1 2 N
(6.8) b b ···b , c c · · · cN ,
1 2

N
X N
X
(−1)i−1 b1 · · · bi−1 β i bi+1 · · · bN , (−1)i−1 c1 · · · ci−1 ∂γ i ci+1 · · · cN .
i=1 i=1

We have the following theorem.

Theorem 6.5. If the holonomy group of the N-dimensional Kähler manifold X is SU(N) and
Rsl(N,C)[t] is generated by above eight elements as a ∂-ring, then the space of global sections of the
chiral de Rham complex of X is a simple vertex algebra strongly generated by
Q(z), L(z), J(z), G(z), B(z), D(z), C(z), E(z).
Proof. The images of the above eight sections under the map of φ∗ and ψ∗,∗ locally are
N
X N
X
ψ1,1 (φ1 (Q(z))) = β i ci , ψ1,1 (φ1 (L(z))) = β i ∂γ i ,
i=1 i=1
N
X N
X
i i
ψ1,0 (φ1 (J(z))) = − bc, ψ1,0 (φ1 (G(z))) = bi ∂γ i ,
i=1 i=1
(6.9)
1 1 2
ψN,0 (φN (D(z))) = b b · · · bN , ψ0,0 (φ0 (E(z))) = f c1 c2 · · · cN ,
f
N
X 1
ψN,1 (φN (B(z))) = (−1)i−1 b1 · · · bi−1 β i bi+1 · · · bN ,
i=1
f
N
X
ψ0,0 (φ0 (C(z))) = (−1)i−1 f c1 · · · ci−1 ∂γ i ci+1 · · · cN .
i=1

They are global sections of gr2 (Q). By assumption, the eight elements of (6.8) gener-
sl(N,C)[t]
ated gr2 (Q)|p as a ∂-ring. Compare (6.8) and (6.9). By Theorem 5.6 the eight global
24
sections in (6.9) generate gr2 (Q)(X) as a ∂-ring. By Lemma 6.2 and 6.3, the eight sections
Q(z), L(z), J(z), G(z), B(z), D(z), C(z), E(z) strongly generate the vertex algebra Ωch
X (X).
By Lemma 6.4, this vertex algebra is simple. 

From [8] and the classical invariant theory [14], when N = 2, Rsl(N,C)[t] is generated by
the eight elements of (6.8), so Theorem 1.2 is a corollary of Theorem 6.5.
Let X be an Enriques surface. Then X is the quotient of a K3 surface X̃ by a fixed
point-free involution σ. σ induce an automorphism of Ωch X̃
(X̃) which maps Q(z), L(z),
J(z), G(z), B(z), D(z), C(z), E(z) to Q(z), L(z), J(z), G(z), −B(z), −D(z), −C(z), −E(z),
respectively. Obviously, Ωch ch σ
X (X) is isomorphic to the subalgebra of ΩX̃ (X̃) consisting the
invariant elements of σ.
Theorem 6.6.
Ωch ch σ
X (X) = ΩX̃ (X̃) .

R EFERENCES
[1] J. Ekstrand, R. Heluani, J. Källén, and M. Zabzine, Chiral de Rham complex on Riemannian manifolds
and special holonomy. Comm. Math. Phys. 318 (2013), no. 3, 575-613.
[2] L. Ein and M. Mustata, Jet schemes and singularities, Algebraic geometry—Seattle 2005. Part 2, 504–
546, Proc. Sympos. Pure Math., 80, Part 2, Amer. Math. Soc., Providence, RI, 2009.
[3] D.D. Joyce, Compact Manifolds With Special Holonomy, Oxford Mathematical Monographs. Oxford
University Press, Oxford, 2000. xii+436 pp.
[4] V. Kac, Vertex Algebras for Beginners. University Lecture Series 10. Providence, RI: American Math-
ematical Society, 1998.
[5] M. Kapranov, Rozansky-Witten invariants via Atiyah classes, Compositio. Math. 115(1999) 71-113.
[6] A. Kapustin, Chiral de Rham complex and the half-twisted sigma-model. arXiv:hep-th/0504074.
[7] A. Linshaw, G. Schwarz, B. Song, Jet schemes and invariant theory, Ann. Inst. Fourier(Grenoble) 65
(2015), no. 6, 2571-2599.
[8] A. Linshaw, G. Schwarz, B. Song, Arc spaces and the vertex algebra commutant problem, Adv. Math.
277 (2015), 338-364.
[9] F. Malikov, V. Schectman, and A. Vaintrob, Chiral de Rham complex, Comm. Math. Phys. 204, (1999)
no. 2, 439-473.
[10] F. Malikov, V. Schectman, Chiral de Rham complex. II, Differential topology, infinite-dimensional
Lie algebras, and applications, 149-188, Amer. Math. Soc. Transl. Ser. 2, 194, Amer. Math. Soc.,
Providence, RI, 1999.
[11] F. Malikov, V. Schectman, Chiral Poincaré duality, Math. Res. Lett. 6, (1999) 533-546
[12] B. Song, The global sections of the chiral de Rham complex on a Kummer surfaces, Int. Math. Res. Not.
IMRN 2016, no.14, 4271-4296.
[13] B. Song, Chiral Hodge cohomology and Mathieu moonshine, arXiv:1705.04060 [math.QA].
[14] H. Weyl, The Classical Groups: Their Invariants and Representations, Princeton University Press,
Princeton, N.J., 1939. xii+302 pp.
[15] K. Yano and S. Bochner, Curvature and Betti Numbers, Annals of Mathemathics Studies, No 32.
Princeton University Press, Princeton, N. J., 1953. ix+190 pp.

S CHOOL OF M ATHEMATICAL S CIENCES , U NIVERSITY OF S CIENCE AND T ECHNOLOGY OF C HINA , H EFEI ,

A NHUI 230026, P.R. C HINA
Email address: [email protected]

25