Journal of Topology 1 (2008) 429445

c 2008 London Mathematical Society

doi:10.1112/jtopol/jtm014

Mirror symmetry formulae for the elliptic genus of complete intersections

Vassily Gorbounov and Serge Ochanine
Abstract
In this paper, we calculate the elliptic genus of certain complete intersections in products of
projective spaces. We show that it is equal to the elliptic genus of the LandauGinzburg models
that are, according to Hori and Vafa, mirror partners of these complete intersections. This
provides additional evidence of the validity of their construction.

Introduction
Understanding the mathematics behind Quantum and in particular, Conformal Field Theories
has been a challenge for more then 25 years. The usual ground for mathematical interpretations
of Quantum Field Theory predictions has been the Topological Quantum Field Theory, which
is a certain reduction of the genuine Quantum Field Theory. There have been signicant
mathematical advances in this area. Mirror symmetry is among the major motivations behind
these advances. Vaguely stated, mirror symmetry is a duality between complex and symplectic
geometry. As was originally discovered by physicists [6], among its specic manifestations is a
striking connection between the number of curves of a given genus in a symplectic manifold
and the periods of a holomorphic form of a dierent complex manifold. Ever since, the explicit
construction of the mirror partner for a given manifold has become the central task for the
mathematicians. A vast amount of work has been done in this direction. In this paper, we
follow the line pioneered by Gepner [8] and developed by Vafa [25], who discovered that the
Conformal Field Theory dened by a manifold, the so-called sigma model, can be identical to
the Conformal Field Theory of a dierent type, the so-called LandauGinzburg model. This
point of view was further developed by Witten [28]. As far as the application of this idea to
mirror symmetry is concerned, the major reference for the purposes of this paper is the work of
Hori and Vafa [15]. They showed that the mirror partner of a large class of manifolds turns out
to be a LandauGinzburg model of some kind, or its orbifold. The proof that these models
form a mirror pair would consist of picking an invariant which is known to be identical for mirror
partners and, by a calculation, showing that it is indeed the same for given hypothetical mirror
partners. In [15] such an invariant is given by the periods of a holomorphic form on a manifold
and the so-called BPS masses in the LandauGinzburg model. Of course, the identities implied
by mirror symmetry in Topological Quantum Field Theory are a reduction of stronger identities
in the original Quantum Field Theory. The rigorous mathematical structure behind Quantum
Field Theory is not known at the present time, so exploring the mathematical consequences
of mirror symmetry at this higher level is dicult. Some time ago, Malikov, Schechtman and
Vaintrob [19] introduced a construction of a mathematical approximation to the structure of
Quantum Field Theory dened by a manifold. It is the so-called Chiral de Rham complex.
Mathematically, these ideas were further developed in [11, 10, 4, 5, 16]. The relevance of
this construction to physics was explained only recently in [30, 17]. Despite being only an

Received 6 July 2007; published online 5 February 2008.

2000 Mathematics Subject Classication 14J32 (primary), 55N34, 14M10.
The rst author is partially supported by the NSF.

430

VASSILY GORBOUNOV AND SERGE OCHANINE

approximation to the genuine Quantum Field Theory, the Chiral de Rham complex carries
features not available in the Topological Quantum Field Theory. One of them is the elliptic
genus. Introduced in mathematics in [21], it was almost immediately connected to quantum
physics [27, 26] and later shown to be identical for mirror partners, thus providing another test
for a pair of manifolds to be mirror partners. Moreover, as explained in [29], there is a physics
counterpart of the elliptic genus in a large class of Conformal Field Theories, in particular, in
LandauGinzburg models and their orbifolds, and this physics elliptic genus coincides with the
one dened in topology for the sigma model. Taking on Vafas and Wittens ideas, a number
of physicists came up with a new type of formulas for the elliptic genus of some classes of
manifolds in terms of their mirror LandauGinzburg partner [18, 2, 3, 7]. For about ten years,
no mathematical proof of these formulas was produced. The rst paper where such a proof
was given in the case of CalabiYau hypersurfaces was [9]. The scope of this paper was much
broader and the result for the elliptic genus fell out as a simple consequence of a much deeper
connection found between the LandauGinzburg model and the Chiral de Rham complex of a
hypersurface.
The purpose of the present paper is to prove, by more or less elementary means, that
the LandauGinzburg mirror partners found in [15] for complete intersection in products of
projective spaces have the physics elliptic genus identical to the topological elliptic genus. It is
interesting to note that our result provides an extra test for the constructions in [15], and as
such, renes the conditions for the existence of a mirror LandauGinzburg theory as a mirror
partner for some complete intersections given in [15]. We are planning to return to the case of
complete intersections in more general toric varieties considered in [15] in a future work.
Although inspired by physics, with the exception of the introductory Section 1, this paper is
mathematically self-contained. It starts with a brief introduction to Jacobi functions and the
corresponding elliptic genera. We then proceed to derive the main formula (Theorem 9) using
the residue theorem for functions of several variables. The formula is derived for elliptic genera
of any level, which includes the case of Hirzebruchs level N genus [13, 14]. In Section 11, the
formula is specialized to the level 2 genus discussed in physics literature. A dierent proof of
the level 2 case is given in Section 12.

1. LandauGinzburg orbifolds and their elliptic genus

In this section, we briey state the results about the elliptic genus of LG theories and their
orbifolds without going into details. In the physics literature, the elliptic genus is dened as
a character of an action of an innite-dimensional Lie algebra, namely the so called N = 2
algebra. For the purposes of this note, LandauGinzburg eld theories are described mathematically by a noncompact manifold and a function W on it, called a superpotential, which has
isolated singularities. These data are sucient for dening and calculating explicitly the elliptic
genus of some LG theories [29] that are relevant for this note. An important class of quantum
eld theories related to LG is dened by orbifolds of LG with respect to a nite group action.
The elliptic genus of the LG orbifolds relevant for us was calculated in a number of papers [2,
3, 7, 18] (see also [9] for a more mathematical approach). Such a genus is either a Jacobi form
or a modular function, depending on the data dening the LG model. In this note, we consider
LG dened by (CN , W ), where the superpotential W is a holomorphic and quasi-homogeneous
function of z1 , z2 , . . . , zN and its orbifolds with respect to the action of some nite group of
symmetries. The conditions on the W are the following: It should be possible to assign some
weights ki Z to the variables zi for i N, and a degree of homogeneity d N to W , so that
W (k i zi ) = d W (zi )
for all i and C. Let qi = ki /d.

ELLIPTIC GENUS OF COMPLETE INTERSECTIONS

431

Suppose that the potential is invariant under a nite abelian group of symmetries G. Denote
by Ri the function on G satisfying g(zi ) = exp(2iRi (g))zi . The invariance of W means, of
course, that for all i and g G,
W (g(zi )) = W (zi ).
The elliptic genus Ell(q, y) of such an LG orbifold dened by the data (W, G) following [2, 3],
is given as follows:
Ell(q, y) =

1
|G|

N


y R i (g 1 )

g 1 ,g 2 G i=1

1 ((1 qi )z + Ri (g1 ) + Ri (g2 ) | )

,
1 (qi z + Ri (g1 ) + Ri (g2 ) | )

where y = exp(2iz) and q = exp(2i ).

Example 1. Consider the following LG data taken from [1]: In the space C9 , introduce
coordinates Xa (a = 0, 1, 2, 3), Yb (b = 0, 1, 2), Zc (c = 0, 1), and dene the potential by the
formula
1


W =
(Xr3 + Xr Yr2 + Yr Zr2 ) + X23 + X2 Y22 + X33 .
r =0

In this case, d = 3 and qi = 1/3. It is easy to check that W is invariant under the following
action of Z/12 with generator :
(Xa , Yb , Zc ) = ( 4 Xa , 2 Yb , Zc ).
Calculating the appropriate numbers Ri () which dene such an action, we obtain
Ra () = 1/3,

Rb () = 1/6,

Rc () = 1/12.

Therefore, the above formula for the elliptic genus of the LG orbifold dened by the data
(W, Z/12) becomes
4

12
i+j
2z
1

j /3 1 ( 3 3 | )
Ell(q, y) =
y

12 i,j =0
1 ( z3 + i+j
3 | )
3 
2

i+j
i+j
2z
2z
j /6 1 ( 3 + 6 | )
j /12 1 ( 3 12 | )
y
.
y

1 ( z3 i+j
1 ( z3 + i+j
6 | )
12 | )
It is easy to check that this formula agrees with our formula for the elliptic genus of the
complete intersection in CP3 CP2 CP1 given by equations
3


a=0

Xa3 = 0,

2

Xb Yb2 = 0,

b=0

1

Yc Zc2 = 0

c=0

(see Section 11).

In Sections 28, k will be an algebraically closed eld of characteristic p 0, G a nite group
of order n, and p
n.

2. Regular representations
Consider the k-vector space V = Map(G, k) of all k-valued functions on G. Clearly, dimk V =
n. The group G acts on V by the formula
(g f )(h) = f (hg)

(g, h G, f V ).

With this action, V is called the regular representation of G.

432

VASSILY GORBOUNOV AND SERGE OCHANINE

We now consider the case when G is abelian and write the group operation in G additively.
 = Hom(G, k ) be the character group of G. The following theorem is well known
Let G
(cf. [23, 2.4]).
 V is a basis of V .
Theorem 1. (i) G
 consists of those f V
(ii) The one-dimensional subspace V of V generated by G
which satisfy
f (u + g) (g)f (u)

for all g G.

(iii) The G-modules V are pairwise nonisomorphic.

(iv) Every irreducible G-module is of degree 1 and is isomorphic to one of the V .
3. Generalized Jacobi functions
Let now E be an elliptic curve over k, let k(E) be the eld of rational functions on E, and
let G E be any nite subgroup of order n. As above, we assume that p
n.
Let be the divisor


(g),
=
g G

and let L(G) k(E) be the associated vector space

L(G) = {f k(E) | div(f )
} .
By the RiemannRoch theorem, dimk L(G) = n. The group G acts on E by translations and
leaves unchanged. Therefore, L(G) is invariant under the induced G-action on k(E). Thus,
L(G) is naturally an n-dimensional representation of G.
Theorem 2.
of G.

As a G-representation, L(G) is isomorphic to the regular representation

Proof. Let u0 E be any point such that nu0 = 0. Dene a G-linear map
: L(G) V = Map(G, k)
by
(f )(g) = f (u0 + g)

(f L(G), g G).

This is well dened, since u0  G and therefore u0 + g is not a pole of f . Since L(G) and V
have the same dimension n, we need only to prove that is injective.
Suppose that f = 0 and (f ) = 0. Then, f (u0 + g) = 0 for all g G. Thus, f has at least
n zeroes. On the other hand, since f L(G), it has at most n poles. Since deg div(f ) = 0, we
conclude that f has a simple pole at each g G and a simple zero at each u0 + g; that is,




(g) +
(u0 + g).
div(f ) =
g G

g G

Thus, the image of div(f ) under Abels map

Div(E) E
is nu0 = 0. The contradiction shows that f = 0.
 there is a nonzero function f L(G) satisfying
Corollary 1. For each G,
f (u + g) (g)f (u)
for all g G. This function is determined uniquely up to a nonzero multiplicative constant.

ELLIPTIC GENUS OF COMPLETE INTERSECTIONS

433


Definition 1. We call f a (generalized) Jacobi function belonging to G.
Remark 1. The classical Jacobi functions correspond to the case where G is a cyclic group
of order 2 (char k = 2).

4. The divisor of a Jacobi function

We now describe explicitly the divisor of a Jacobi function.
Theorem 3. A nonzero function f L(G) is a Jacobi function if and only if div(f ) is
invariant under translations by elements of G.
Proof. If f is a Jacobi function, the formula
f (u + g) (g)f (u)
shows that the functions
u  f (u) and

u  f (u + g)

have the same divisor. Thus, div(f ) is invariant by G.

Conversely, let f be a nonconstant function in L(G), and suppose that div(f ) is invariant
by translation. Then there is a constant (g) k such that
f (u + g) (g)f (u),
for all u E. Since
(g1 + g2 )f (u) = f (u + g1 + g2 ) = (g2 )f (u + g1 ) = (g1 )(g2 )f (u),
 and f is a Jacobi function belonging to .
we have (g1 + g2 ) = (g1 )(g2 ). Thus, G,
It is now easy to describe the divisors of the Jacobi functions. Let En be the group of
n-division points on E. Since n is the order of G, we have G En . For every coset En /G,
let


(r) Div(E).
=
r

Theorem 4. A nonconstant function f L(G) is a Jacobi function if and only if

div(f ) = +
for some nonzero En /G.
Proof. Using Theorem 3, we need only to describe the nonzero principal divisors D satisfying D
and invariant under G. Since the polar part of D is nonzero and invariant under
G, it must be equal to . Since D + is invariant, and is of degree n, it is necessarily of the
form


(r + g)
g G

for some r  G, that is, of the form .

434

VASSILY GORBOUNOV AND SERGE OCHANINE

5. Modulus and conjugation

Let f be a Jacobi function with character and divisor div(f ) = + , En /G. For
each r , the involution u  r u takes div(f ) to div(f ). Therefore, f (u)f (r u) is a
nonzero constant, which we call the modulus of f and designate c(r). Thus, c : k . It is
easy to check that
c(r + g) = (g)c(r)
for all g G.
Similarly, the involution u  u takes + to + . This u  f (u) is also a
Jacobi function with character 1 . In particular, when n = 2, En /G has order 2, and 1 = .
It follows that f (u) = af (u), where a2 = 1. Thus, f is either even or odd. Since f has a pole
of order 1 at O E, it has to be odd.
6. Jacobi functions on the Tate curve
Consider a local eld k complete with respect to a discrete valuation v, and let q k be
any element satisfying v(q) < 1. It is well known (cf. [24, Appendix C, 14]) that E = k /q Z
can be identied with the Tate curve
y 2 + xy = x3 + a4 x + a6 ,
where

a4 =

(5m3 )

m 1

and

qm
1 qm

 5m3 + 7m5  q m
a6 =
.

12
1 qm
m 1

Let G E be a nite subgroup of order n. As usual, we assume that char k
n. Let En /G

be any nontrivial coset. We will construct an explicit Jacobi function for G, whose zeroes are
the points of .
for its image in E = k /q Z . Choose g1 = 1, g2 , . . . , gn k , so
If k , we will write
that
G = {
g1 , g2 , . . . , gn } .

Then choose r1 , r2 , . . . , rn k , so that

= {
r1 , r2 , . . . , rn }
and

ri =

gi .

This can be done as follows. First, choose r k , so that r E represents . Then clearly,
= {
rg1 , rg2 , . . . , rgn } .
Since rn = 1 in E, we have rn = q s for some s Z. Let
r1 = q s r,
Then


i

ri = rgi


ri = q

s n


i

(i 2).


gi


i

gi .

ELLIPTIC GENUS OF COMPLETE INTERSECTIONS

Now, consider the basic Theta function

(u) = (1 u1 )
=


k 1

(1 q u)
k

k 1

This is an analytic function on k

satises

435

(1 q k u)(1 q k u1 )


(1 q k u1 ).

k 0

that has simple zeroes at the points of q Z (cf. [22]) and

(q 1 u) = u(u).

Along with (u), we will consider its translates (u) ( k ) dened by

(u) = (1 u).
The function (u) has simple zeroes at the points of q Z and satises
(q 1 u) = 1 u (u).
Consider
f (u) =

n

r i (u)
.
g i (u)
i=1

Theorem 5. The function f is q-periodic and denes a Jacobi function for G having as
the coset of zeroes.
Proof. It is enough to prove the rst statement, since this would imply that f denes a
function on E with simple poles at the points of G and simple zeroes at the points of . We
have
n
n

r i (q 1 u)  (ri u)r i (u)
=
= f (u).
f (q 1 u) =
g i (q 1 u) i=1 (gi u)g i (u)
i=1

7. A special case
We now assume that q is not a root of 1, and choose an arbitrary nth root q0 of q. For n 1
and a Z, dene


na (u) =
(1 q
u)
(1 q
u1 ).

1,
a(n )

0,
a(n )

With the notation of the previous section, we have (u) = 10 (u).

Theorem 6. The function
f (u) =

n1 (u)
u n0 (u)

denes a Jacobi function on E = k /q Z for the group G E consisting of the images of the
nth roots of 1, and that has simple zeroes at the images of the nth roots of q.
Proof. Let 1 = 1, 2 , . . . , n be the nth roots of 1 in k. Thus,
G = {
1 , 2 , . . . , n } ,
Notice that

= {
1 q0 , 2 q0 , . . . , n q0 } .



i q.
(i q0 ) =

436

VASSILY GORBOUNOV AND SERGE OCHANINE

Thus, we can take

r1 = q 1 q0 ,

(i 2).

ri = i q0

According to the previous theorem, the function

n

r i (u)

i (u)
i=1

has all the required properties. It remains to identify the numerator and denominator explicitly.
We will be using the identities


n
(1 i t) =
(1 1
i t) = 1 t ,
which are easily proved by noticing that the three polynomials in t have same degree n, the
same roots, and the same constant term 1.
We have




i (u) =
(1 i u1 )
(1 q k 1
(1 q k i u1 )
i u)
i

= (1 u

k 1

(1 q

nk

u )

k 1

k 1

(1 q

nk

i
n

k 1

= n0 (u).
Similarly,




1
i q 0 (u) =
(1 i q0 u1 )
(1 q k 1
(1 q k i q0 u1 )
i q0 u)
i

= (1 qu

k 1

(1 q

n k 1 n

u )

k 1

k 1

(1 q

n k +1 n

k 1

= n1 (u).
Finally,

r i (u) =

i q 0 (u)

q 1 q 0 (u)
q 0 (u)

and
q 0 (u) = (q01 u) = (q 1 (qq01 u)) = qq01 u q 1 q 0 (u);
that is,
n1 (u)
u n0 (u)
is a constant multiple of
n

r i (u)

i (u)
i=1

and therefore, has all the required properties.

Remark 2. In the case n = 2, the previous theorem gives
 (1 q 2k 1 u2 )(1 q 2k 1 u2 )
1
f (u) =
u(1 u2 )
(1 q 2k u2 )(1 q 2k u2 )
k 1

 (1 q 2k 1 u2 )(1 q 2k 1 u2 )
1
.
=
u u1
(1 q 2k u2 )(1 q 2k u2 )
k 1

ELLIPTIC GENUS OF COMPLETE INTERSECTIONS

437

The formal substitution u = ez /2 leads to

 (1 q 2k 1 ez )(1 q 2k 1 ez )
z/2
,
zf (ez /2 ) =
sinh(z/2)
(1 q 2k ez )(1 q 2k ez )
k 1

which is a familiar expression for the generating function of the level 2 elliptic genus (see
below).
8. Modulus and normalization on the Tate curve
Continuing with the situation of the preceding section, we rst compute the modulus for the
Jacobi function dened by;
n (u)
f (u) = 1n
;
u 0 (u)
that is, we compute the constant value of f (u)f (ru1 ) for r . We start with r = q0 . The
following formulas can be easily obtained from the denition of na :
n0 (u1 ) = un n0 (u),

na (u1 ) = un na (u)

(a  0(n))

and
n0 (q0 u) = q 1 un n1 (u),

n1 (q0 u) = n0 (u).

It follows that
n0 (q0 u1 ) = q 1 un n1 (u),

n1 (q0 u1 ) = un n0 (u),

and therefore
qun0 (u)
= q0n 1 f (u)1 .
q0 n1 (u)

f (q0 u1 ) =
Thus,

c(q0 ) = q0n 1 .
Notice now that the formula for f does not depend on the choice of q0 . Therefore, we have the
following theorem.
Theorem 7. For the Jacobi function dened by
f (u) =

n1 (u)
,
u n0 (u)

and any nth root r of q in k , we have

c(r) = rn 1 .
We now normalize f by requiring that
Resu =1


f (u)

du
u


=

1
.
n

This normalization is formally equivalent to the requirement that zf (ez /n ) = 1 + o(z). Since
n1 (u)
(u 1)n1 (u)
u1

n
nk n
n k n )
u 0 (u)
u(1 un )
k 1 (1 q u )(1 q u
n 1
n1 (u)
u

,
=

nk n
n k n )
1 + u + u2 + + un 1
k 1 (1 q u )(1 q u

438

VASSILY GORBOUNOV AND SERGE OCHANINE

we see that the normalized Jacobi function is f (u)N , where

nk 2
k 1 (1 q )
.
N=
n1 (1)
The modulus of the normalized function is
c(r) = rn 1 N 2
and satises
c(r)n = q n 1 N 2n .
In particular, if n = 2, we have

  1 q 2k 2
N=
1 q 2k 1
k 1

and
c(r)2 = q

  1 q 2k 8
,
1 q 2k 1

k 1

which is the familiar expression for the modular form for one of the three level 2 elliptic
genera.
9. Elliptic genus
From now on, k = C, E = C/L for some lattice L, G = L0 /L and n = [L0 : L]. Let f
be a Jacobi function with character . Then f is an elliptic function with period lattice L.
We normalize f , so that it has residue Resz =0 (f ) = 1. Then the Taylor expansion of zf (z)
is a formal power series with constant term 1 and denes, via the Hirzebruch formalism, a
multiplicative genus
: U C,
which we refer to as the level n elliptic genus dened by f . The case n = 2 is best known. In
this case, zf (z) is an even series in z and factors through a genus
: S O C.
10. Complete intersections
In this section, we will compute the elliptic genus of complete intersections satisfying a
nondegeneracy condition as a summation over some division points of E E . . . E.
Let M = (mij ) be a l t matrix over Z, and let P be the product
P = CPN 1 1 CPN 2 1 . . . CPN t 1 .
For 1 j t, let j be the pull-back over P of the canonical line bundle of CPN j 1 . Then for
1 i l, let i be the line bundle
i = 1m i 1 2m i 2 . . . tm i t
over P . We will write Hi for the stably almost complex manifold (hypersurface) dual to i , and
let X(M ) be the transverse intersection
X(M ) = H1 H2 . . . Hl .
Writing M z = (1 z, 2 z, . . . , l z), we dene l linear forms i : Ct C; that is,
i z = mi1 z1 + mi2 z2 + . . . + mit zt .

ELLIPTIC GENUS OF COMPLETE INTERSECTIONS

439

Let be the elliptic genus of level n dened by a Jacobi function for G E with character .
The standard computation using Hirzebruchs formalism leads to the following statement.
Theorem 8. (X(M )) is the coecient of z11 z21 zt1 in the Laurent expansion at
z = (0, 0, . . . , 0) of
F (z) = F (z1 , z2 , . . . , zt ) =

f (z1 )N 1 f (z2 )N 2 f (zt )N t

.
f (1 z)f (2 z) f (l z)

In accordance with the notation [m] for the multiplication-by-m map on E, we will write
[M ] : E t E s ,

[i ] : E t E,

[] : E t E

for the maps induced by left multiplication by M , by i , and by = i .
For the rest of this section, we will assume that l = t and that det M = 0. Let Gt = G. . .G
E t and H E t be the inverse image of Gt under [M ]. Clearly, H is a subgroup of E t containing
Gt (since the entries of M are integers), and it is easy to check that [H : Gt ] = (det M )2 .
For each j, 1 j t, choose a zero rj E of f , so that




(g) +
(rj + g),
div(f ) =
g

and let r = (r1 , r2 , . . . , rt ) E . Also, choose a s = (s1 , s2 , . . . , st ) E t satisfying [M ]s = r

(this uses det M = 0).
t

mij Nj mod exp(G), then we have

(1)t+1
([]h)f (s + h)N ,
(X(M )) =
(det M )c(r)
t

Theorem 9. If, for every j,

hH /G

where we use the abbreviations

c(r) = c(r1 )c(r2 ) c(rt )
and
f (s + h)N = f (s1 + h1 )N 1 f (s2 + h2 )N 2 f (st + ht )N t ,
and where the sum runs over representatives of the cosets in H/Gt .
Remark 3. The condition
mod exp(G).

mij Nj mod exp(G) is equivalent to c1 (X(M )) 0

The proof of this theorem is based on the following, slightly modied version of the global
residue theorem for functions of several complex variables, as described in [12, Chapter 5, pp.
655656].
Theorem 10. Let V be a compact complex manifold of dimension t, and let
D1 , D2 , . . . , D
(t ) be eective divisors having the property that the intersection of every t of them,
Dn 1 Dn 2 . . . Dn t , is a nite set of points, whereas the intersection of every t + 1 of them
is empty. Let D = D1 + D2 + . . . + D , and let be a meromorphic t-form on V with polar
divisor D. Then for every P Dn 1 Dn 2 . . . Dn t , the residue ResP is dened and


ResP = 0.
P

440

VASSILY GORBOUNOV AND SERGE OCHANINE

The residue ResP for P Dn 1 Dn 2 . . . Dn t can be dened as follows: choose local

coordinates (z1 , z2 , . . . , zt ) near P and write near P as
=

(z) dz1 dz2 . . . dzt

,
1 (z)2 (z) t (z)

where , 1 , 2 , . . . t are pairwise relatively prime holomorphic functions. Then

1
,
ResP =
(2i)t
where is the suitably oriented real cycle
= {z | 1 (z) = , 2 (z) = , . . . , t (z) = } ,
for a small .
It can be easily veried that if F (z) is a meromorphic function near O = (0, 0, . . . , 0) Ct , then
ResO (F (z) dz1 dz2 . . . dzt ) is the coecient of z11 z21 zt1 in the Laurent expansion of
F (z) at O. We will also use the following formula, generalizing a fact well known for functions
of one variable.
Lemma 1.

If is written near P as
(z) dz1 dz2 . . . dzt
,
=
1 (z)2 (z) t (z)

and det(i /zj )(P ) = 0, then

ResP =

(P )
.
det(i /zj )(P )

We now turn to the proof of Theorem 10. In view of Theorem 9, and the above remarks, we
need to compute ResO , where
=

f (z1 )N 1 f (z2 )N 2 f (zt )N t dz1 dz2 . . . dzt

.
f (1 z)f (2 z) f (l z)

We let V = E t and consider on V . According to Theorem 11, we have

ResP = 0,
P

where P runs through the points P Gt , corresponding to the poles of f (z) (rst kind), and
through the points P s + H, which are the simultaneous zeroes of f ([1 ]z), f ([2 ]z), . . . ,
f ([l ]z) (second kind).
Notice that the condition


mij Nj mod exp(G)
i

guarantees that F (z + g) = F (z) for all g Gt . It follows that the contribution of the points
of the rst kind to the sum of residues is nt ResO .
Turning to the points of the second kind, say P = s + h, we rst notice that dierentiating
f (ri z) =

c(ri )
f (z)

with respect to z and taking the limit as z 0 gives

f  (ri ) = c(ri ).

ELLIPTIC GENUS OF COMPLETE INTERSECTIONS

441

Therefore,
f ([i ]z)
(s + h) = f  ([i ](s + h))mij = f  (ri + [i ]h))mij = ([i ]h)c(ri )mij .
zj
Thus,
det(f ([i ]z)/zj )(s + h) = (1)t c(r)([]h) det M,
and
Ress+h =

(1)t ([]h)f (s + h)N

.
(det M )c(r)

t
Also, notice that this residue
 remains unchanged when h is replaced by h + g with g G
(because of the condition j mij Nj mod exp(G)).
Theorem 10 is now an immediate consequence of the residue Theorem 11.

11. The level 2 case

We now specialize the above formula for the level 2 elliptic genus. Let H = {z
C | (z) > 0}, L0 = Z Z , L = Z Z(2 ), G = {0, }, and let : G {1} be dened by
( ) = 1. The divisor of the corresponding Jacobi function f is
div(f ) = (0) ( ) + (1/2) + (1/2 + ).
By comparing the divisors and the 1/z 2 terms in the Taylor expansions at 0, we easily conclude
that
f (z)2 = (z| ) (1/2 | ) = (z| ) e1 .
We will choose r = 1/2. Since f satises the dierential equation
(f  )2 = f 4 2f 2 +
2

2
(cf. [20]), we see
that c(1/2) = f (1/2) = , and, with an appropriate choice of the square
root, c(1/2) = .
Consider rst, the case of a hypersurface X(m) CPN 1 of degree m. Since we can take
s = 1/2m, we have the following theorem.

Theorem 11. If N and m have same parity, then





1
b
a
1
+
+
(1)a f N
(X(m)) =
.
2m m
m
m
0a,b< m

This is the EguchiJinzenji formula derived in [7]. Notice that only the case where both N
and m are even is of interest, since, for dimensional reasons, the genus vanishes when N is odd.
Turning now to complete intersections, consider the case of X(M ) CP3 CP2 CP1 ,
where

3 0 0
M = 1 2 0 .
0 1 2
If we take r = (1/2, 1/2, 1/2), then one can take s = (1/6, 1/6, 1/6), since

3 0 0
1/6
1/2
1 2 0 1/6 = 1/2 .
0 1 2
1/6
1/2

442
Then, noticing that

0
0
1

VASSILY GORBOUNOV AND SERGE OCHANINE

1 0
3
0 1 1
3 6
0

0
1
0 0
2
0

0
2
1

2 4
1 0 0
1 2 = 0 1 0 ,
0
1
0 0 12

we easily conclude that

1 2
1
0 1
12
0 0

4
0
1/3
2 0 = 1/6 = v
1
1
1/12

generates a cyclic subgroup of order 12 of H/G3 , so that one can take

{(a + b )v | 0 a, b < 12}
as representatives of the cosets of H/G3 . Also, notice that []v = 1, so
([](a + b )v) = (1)b .
We obtain the following theorem.
Theorem 12. For M as above,
1
(X(M )) =
12 3/2

(1) f

0a,b< 12

1 a + b
+
6
3

4
f

1 a + b

6
6

3
f

1 a + b
+
6
12

2
.

The general case can be treated in exactly the same way, since one can always nd matrices
A, B, invertible over Z and such that AM B is diagonal.

12. A dierent approach for the level 2 case

Considering, as before, a hypersurface X(m) CPN 1 of degree m, and keeping the
assumptions of the previous section, we rst note that
div(f ) =

(g) +

(r + g),

so that
div(f (mz)) =

(h) +

(s + h),

and thus the polar divisor of 1/f (mz) is


div

1
f (mz)


=

(s + h).

To calculate the residue of 1/f (mz) at s + h, let t  (t) be a small loop going around 0 in
the positive direction and for h H, let (t) = (t) + s + h. Then, using the substitution

ELLIPTIC GENUS OF COMPLETE INTERSECTIONS

u = z + s + h, we have


Resz =s+h

1
f (mz)

443


dz
1
=
2i f (mz)

du
1
=
2i f (mu + r + mh)

du
(mh)
=
2i
f (mu + r)

(mh)
=
f (mu)du
2ic(r)
(mh)
=
.
mc(r)

Theorem 13. We have

1
1
=
f (mz)
mc(1/2)

0a,b< m

(1)a f  (z)
,
f (z) f (a,b )

where
a,b =

b
a
1
+
+
.
2m m
m

Proof. Specializing the above calculations to the level 2 case, we see that the polar divisor
of 1/f (mz) is given by




1
div
(a,b ),
=
f (mz)
0a< 2m , 0b< m

and that


Res a , b

1
f (mz)

Consider the function

ha,b (z) =


=

(1)a
.
mc(1/2)

f  (z)
,
f (z) f (a,b )

that is, the logarithmic derivative of f (z) f (a,b ). Let be the involution (z) = z. Then,
since f is odd,
f ( z) = f (z) = f (z);
that is, f = f . Also,
b
a
1
b 1 (m a)
1

=
+
+
= m a,b1 ,
2m m
m
2m
m
m
where ma and b1 should be interpreted modulo 2m and m respectively. As a consequence,
we have ha,b = hm a,b1 .
(a,b ) =

The function f (z) f (a,b ) has two simple poles 0 and . Thus, it has two zeroes (counted
with multiplicities), one of which is a,b . Since m is even, b  b1 mod m. Thus, m a,b1
is the other zero of f (z) f (a,b ) and both zeroes are simple. Thus
div (ha,b ) = (0) ( ) (a,b ) (m a,b1 ),
with residue 1 at 0 and , and residue 1 at a,b and m a,b1 . Let now




1
(1)a f  (z)
1
=
F (z) =
mc(1/2)
f (z) f (a,b )
2mc(1/2)
0a< m , 0b< m

0a< 2m , 0b< m

(1)a ha,b (z).

444

VASSILY GORBOUNOV AND SERGE OCHANINE

The possible poles of F are 0, , and a,b with 0 a < 2m, 0 b < m. Let us compute the
residues of F at these points. First, we have


1
(1)a (1) = 0,
Res0 (F ) = Res (F ) =
2mc(1/2)
0a< 2m , 0b< m

since m is even. Then

Res a , b (F ) =

1
(1)a
((1)a + (1)m a ) =
.
2mc(1/2)
mc(1/2)

Thus, 1/f (mz) and F (z) have the same polar part. It follows that
1
= F (z) + C,
f (mz)
for some constant C. Replacing z with (z) = z, we have
1
1
1
1
=
=
=
,
f (m(z))
f (m mz)
f (mz)
f (m(z))
since m is a period of f for m even. Similarly, since f (z) is invariant under and f  ((z)) =
f  (z), we have ha,b ((z)) = ha,b (z) and therefore, F ((z)) = F (z). It follows that C = 0.
As a corollary, we obtain a new proof of Theorem 11. Indeed, according to Theorem 8,

f (z)N dz
1
,
(X(m)) =
2i f (mz)
where is a small circle around 0 traversed counter-clockwise. Thus,


(1)a f (z)N f  (z) dz
1
(X(m)) =
.
2imc(1/2)
f (z) f (a,b )
0a,b< m

Changing the variable to u = f (z), we get

(X(m)) =

1
2imc(1/2)

0a,b< m

(1)a uN , du
,
u f (a,b )

where is a large circle around 0 traversed clockwise. The Cauchy integral formula gives


1
(X(m)) =
(1)a f (a,b )N ,
mc(1/2)
0a,b< m

which is exactly the formula in Theorem 11.

Acknowledgements. The authors beneted from a number of stimulating discussions with
F. Hirzebruch, K. Hori, F. Malikov and D. Zagier. Part of the work was completed when the
rst author was visiting the Max Plank Institut f
ur Mathematik, which provided an excellent
and stimulating research environment. The authors are especially grateful to A. Gerasimov
for bringing to their attention the topological aspects of the Landau-Ginzburg/Calabi-Yau
correspondence.
References
bsch, Classical vs. LandauGinzburg geometry of compactica1. P. Berglund, B. R. Greene and T. Hu
tion, Modern Phys. Lett. A 7 (1992) 18551870.
2. P. Berglund and M. Henningson, LandauGinzburg orbifolds, mirror symmetry and the elliptic genus,
Nuclear Phys. B 433 (1995) 311332.
3. P. Berglund and M. Henningson, On the elliptic genus and mirror symmetry, Mirror symmetry II
(ed. B. Greene and S. T. Yau), AMS/IP Studies in Advanced Mathematics 1 (American Mathematical
Society, Providence, RI, 1997) 115127.
4. L. A. Borisov, Vertex algebras and mirror symmetry, Comm. Math. Phys. 215 (2001) 517557.

ELLIPTIC GENUS OF COMPLETE INTERSECTIONS

445

5. L. A. Borisov and A. Libgober, Elliptic genera of toric varieties and applications to mirror symmetry,
Invent. Math. 140 (2000) 453485.
6. P. Candelas, X. C. De la Ossa, P. S. Green and L. Parkes, An exactly soluble superconformal theory
from a mirror pair of CalabiYau manifolds, Phys. Lett. B 258 (1991) 118126.
7. T. Eguchi and M. Jinzenji, Generalization of CalabiYau/LandauGinzburg correspondence, J. High
Energy Phys. 0002:028 (2000).
8. D. Gepner, Exactly solvable string compactications on manifolds of SU(N ) holonomy, Phys. Lett. B
199 (1987) 380388.
9. V. Gorbounov and F. Malikov, Vertex algebras and the Landau--Ginzburg/CalabiYau correspondence, Mosc. Math. J. 4 (2004) 729779, 784.
10. V. Gorbounov, F. Malikov and V. Schechtman, Gerbes of chiral dierential operators. III, The Orbit
Method in Geometry and Physics, Marseilles, 2000, Progress in Mathematics 213 (Birkha user, Boston,
MA, 2003) 73100.
11. V. Gorbounov, F. Malikov and V. Schechtman, Gerbes of chiral dierential operators. II. Vertex
algebroids, Invent. Math. 155 (2004) 605680.
12. P. Griffiths and J. Harris, Principles of algebraic geometry, Wiley Classics-Library (John Wiley & Sons,
Inc., Hoboken, NJ, 1994).
13. F. Hirzebruch, Elliptic genera of level N for complex manifolds, Dierential Geometrical Methods in
Theoretical Physics (ed. K. Bleuler and M. Werner; Kluwer, Dordrecht, 1988) 3763.
14. F. Hirzebruch, T. Berger and R. Jung, Manifolds and modular forms, Aspects of Mathematics E20
(Friedr. Vieweg & Sohn, Braunschweig, 1992). Appendices by Nils-Peter Skoruppa and by Paul Baum.
15. K. Hori and C. Vafa, Mirror symmetry, Preprint, 2000, arXiv:hep-th/0002222.
16. M. Kapranov and E. Vasserot, Vertex algebras and the formal loop space, Publ. Math. Inst. Hautes

Etudes
Sci. 100 (2004) 209269.
17. A. Kapustin, Chiral de Rham complex and the half-twisted sigma-model, Preprint, 2005, arXiv:hepth/0504078.
18. T. Kawai, Y. Yamada and S.-K. Yang, Elliptic genera and N = 2 superconformal eld theory, Nuclear
Phys. B 414 (1994) (12) 191212.
19. F. Malikov, V. Schechtman and A. Vaintrob, Chiral de Rham complex Comm. Math. Phys. 204
(1999) 439473.
20. S. Ochanine, Elliptic genera, Encyclopdia Mathematica, Suppl. Vol. 1 (Kluwer, Dordrecht, 1997) 236
239.
21. S. Ochanine, Sur les genres multiplicatifs denis par des integrales elliptiques, Topology 26 (1987) 143
151.
22. P. Roquette, Analytic theory of elliptic functions over local elds, Hamburger Mathematische
Einzelschriften 1 (Vandenhoeck & Ruprecht, G
o ttingen, 1970).
23. J.-P. Serre, Linear representations of nite groups, Graduate Texts in Mathematics 42 (Springer, Berlin,
1977).
24. J. H. Silverman, The arithmetic of elliptic curves, Graduate Texts in Mathematics 106 (Springer, Berlin,
1986).
25. C. Vafa, String vacua and orbifoldized LG models, Modern Phys. Lett. A 4 (1989) 11691185.
26. E. Witten, Elliptic genera and quantum eld theory, Comm. Math. Phys. 109 (1987) 525536.
27. E. Witten, The index of the Dirac operator in loop space, Elliptic curves and modular forms in algebraic
topology, Princeton 1986 (ed. P. S. Landweber), Lecture Notes in Mathematics 1326 (Springer, Berlin,
1988) 161181.
28. E. Witten, Phases of N = 2 theories in two dimensions, Nuclear Phys. B 403 (1993) (12) 159222.
29. E. Witten, On the LandauGinzburg description of N = 2 minimal models, Internat. J. Modern Phys.
A 9 (1994) 47834800.
30. E. Witten, Two-dimensional models with (0,2) supersymmetry: Perturbative aspects, Preprint, 2005,
arXiv:hep-th/0504078v3.

Vassily Gorbounov
Department of Mathematical Sciences
Kings College
University of Aberdeen
Aberdeen AB24 3UE
United Kingdom
vgorb@maths.abdn.ac.uk

Serge Ochanine
Department of Mathematics
University of Kentucky
Lexington, KY 40506-0027
USA
ochanine@ms.uky.edu