The Analogue of the Dedekind Eta Function for CY Manifolds., , , November 27, 2006

Contents

1 Introduction

1.1 General Comments

Explicit formulas for regularized determinants have many important applications in Mathematics as well as Physics. In the case of an elliptic curve the determinant is well known via the Kronecker limit formula. In this paper we begin a series of papers in which we will study the regularized determinant of the Laplacian of a Calabi-Yau manifold. This is very analogous to the elliptic curve case in many ways, so we will first review the computation and the main results about the the regularized determinant of the flat metric for elliptic curves. We plan to describe the determinant's relationship to the discriminant locus and then we will describe our program how to generalize the results in the case of elliptic curves to CY manifolds.

The computation of the regularized determinant in the case of the flat metric on an elliptic curve is based on the Kronecker limit formula. See [28] . It states that if

E (s) = \sum n, m \in Z^{'} {(\frac{1}{2 π})}^{2 s} \frac{1}{| n + m τ |^{2 s}}

where

τ \in C, Im τ > 0

and ' means that the sum is over all pair of integers

(m, n) \neq (0, 0),

then

E (s)

has a meromorphic continuation in

C

with only one pole at

s = 1

and

exp (- \frac{d}{d s} E (s) |_{s = 0}) = {(Im τ)}^{2} {| η |}^{4}

where

η

is the Dedekind eta function. In the case of elliptic curves

{E_{τ} = C / (n + m τ) | Im τ > 0},

the function

E (s)

is the zeta function of the Laplacians of the flat metrics on the elliptic curves

E_{τ} .

The regularized determinant of the Laplacian is

\begin{matrix} det Δ_{τ, 1} : = exp (- \frac{d}{d s} E (s) |_{s = 0}) = {(Im τ)}^{2} {| η |}^{4} \end{matrix}

(2)

where

η^{24}

is equal to the discriminant of the elliptic curve

E_{τ} .

One way to derive formula 2

()

is based on the following observation that

log

det Δ_{τ, 1}

is the potential of the Weil-Petersson metric on the moduli space of elliptic curves.

Notice that

Im τ = - \frac{\sqrt{- 1}}{2} \int_{E} ω_{τ} \land \bar{ω_{τ}} = 〈 ω_{τ}, ω_{τ} 〉,

where

ω_{τ}

are holomorphic one form on the elliptic curves normalized as follows:

\begin{matrix} \int_{γ_{0}} ω_{τ} = 1, \end{matrix}

(3)

where

γ_{0}

is one of the generators of

H_{1} (E, Z) .

This means that:

\begin{matrix} d d^{c} log det Δ_{τ, 1} = d d^{c} log Im τ = d d^{c} log 〈 ω_{τ}, ω_{τ} 〉 . \end{matrix}

(4)

From the relation 4

()

we derive that locally we have

\begin{matrix} det Δ_{τ, 1} = 〈 ω_{τ}, ω_{τ} 〉 {| η (τ) |}^{2} = Im τ {| η (τ) |}^{4}, \end{matrix}

(5)

where

η (τ)

is a holomorphic function. The normalization 3

()

implies that

〈 ω_{τ}, ω_{τ} 〉 = Im τ

is an automorphic form of weight

- 2 .

()

is related to the choice of a maximal unipotent element of the mapping class group. Maximal unipotent elements in the mapping class group. It was proved in [21] that maximal unipotent elements in the mapping class group correspond to a monodromy operator

T

acting on the middle cohomology group of a generic fibre of a family of projective algebraic varieties over the unit disk with only one singular fibre over the origin of the disk. Grothendieck proved that we always have:

(T^{N} - i d)^{n + 1} = 0

where

N

is some positive integer and

n

is the complex dimension of the fibre of the family. Maximal unipotent elements corresponds to Jordan blocks of dimension one. In case of CY manifolds it was proved in [21] that if

T

has maximal index of unipotency then

T

has a unique Jordan block of dimension

n + 1 .

Thus once we choose the unipotent element of the mapping class group we can associate to it a unique up to a sign primitive class of homology in the middle homology group which corresponds to the invariant vanishing cycle.

We will outline how 5

()

implies that

η (τ)

is the Dedekind eta function.

This can be done in several steps. The first step is to prove that

det Δ_{τ, 1}

is a bounded function on the moduli space of the elliptic curves. The next step is to prove that

〈 ω_{τ}, ω_{τ} 〉

has a logarithmic growth near infinity. This fact will imply that

| η (τ) |

must vanish at infinity. It is not difficult to see that

\begin{matrix} S L_{2} (Z) / [S L_{2} (Z), S L_{2} (Z)] ≊ Z / 12 Z . \end{matrix}

(6)

Thus 5

()

and 6

()

imply that

η (τ)^{24}

will be a cusp form of weight 12. This will prove that

η (τ)

is the Dedekind eta function up to a constant. This fact intuitively represents the following observation, when the metric degenerates then the discrete spectrum of the Laplacian in the limit turns into a continuous spectrum which contains the zero. Thus the regularized determinant must vanish at the points of the compactified moduli space that correspond to singular elliptic curves.

It is a classical fact that the Dedekind eta function is related to the algebraic discriminant of the equation that defines the elliptic curve. Thus the arguments that we provided to prove that the absolute value of the holomorphic function that appeared in 5

()

is in fact the analytic analogue of the discriminant.

This follows from the fact that

det Δ_{τ, 1}

can be interpreted as the

L^{2}

norm of a holomorphic section of some power of the line bundle of the holomorphic one forms over the moduli space of elliptic curves. In this and in following publications we will generalized these arguments for CY manifolds.

This paper is the first one in the realization of the above described program.

In it we will compute

d d^{c} log det Δ_{τ, 1} .

We will show that locally in the Teichmüller space of the polarized CY manifolds we have

\begin{matrix} d d^{c} log det Δ_{τ, 1} = d d^{c} log 〈 ω_{τ}, ω_{τ} 〉, \end{matrix}

(7)

where

ω_{τ}

is a family of holomorphic

n

forms. Thus 7

()

implies

\begin{matrix} det Δ_{τ, 1} = 〈 ω_{τ}, ω_{τ} 〉 | φ (τ) |^{2}, \end{matrix}

(8)

where

φ (τ)

is a locally defined holomorphic function on the Teichmüller space.

In fact by proving 8

()

we generalized 5

()

for higher dimensional CY manifolds.

In the case of elliptic curves formula 4

()

implies that

log det Δ_{τ, 1}

is the potential of the Poincare metric on the upper half plane which is the Teichmüller space of elliptic curves. In [32] we proved that

d d^{c} log 〈 ω_{τ}, ω_{τ} 〉

gives the imaginary part of the Weil-Petersson metric on the Teichmüller space of polarized CY manifolds. Thus by establishing formula 7

(),

we proved that the logarithm of the regularized determinant

det Δ_{τ, 1}

forms is the potential of the Weil-Petersson metric.

1.2 Outline of the Main Ideas

The ideas that we used in this paper are similar to the ideas used in [6] to established variational formulas for regularized determinants on vector bundles on Riemann surfaces. We need first to fix the coordinates of the local deformation space of the fix CY manifold M

_{0}

with a fix polarization class. In fact we use the coordinates

τ = (τ^{1}, . . ., τ^{N})

introduced in [32] . These coordinates were later used in string theory. See [2] . The coordinates

(τ^{1}, . . ., τ^{N})

depend on the choice of a basis of harmonic forms

φ_{1}, . . ., φ_{N}

H^{1} (

_{0},

^{1, 0}) .

In these coordinates we have the following local expression for the family of operators

\bar{\partial_{τ}}

on M

_{0}

defined by the complex structures corresponding to

τ = (τ^{1}, . . ., τ^{N}) :

\bar{\partial_{i, τ}} = \frac{\bar{\partial}}{\bar{\partial z^{i}}} - \sum_{k = 1}^{N} τ^{k} φ_{k, \bar{i}}^{j} \frac{\partial}{\partial z^{j}} + O (τ^{2}),

where

(z^{1}, . . ., z^{n})

are local coordinates on M

_{0} .

Thus we get that

\bar{\partial_{τ}}

depends holomorphically on

τ

and

\begin{matrix} \frac{\partial}{\partial τ^{i}} \bar{\partial_{τ}} = - φ_{i} \circ \partial_{0} + O (τ) . \end{matrix}

(9)

From the uniqueness of the solution of Calabi conjecture by Yau we know that once we fix the polarization class then we obtain a family of Calabi Yau metrics

g_{τ}

such that

Im g_{τ} = L .

One of our results is that

Im g_{τ}

is a constant symplectic form. From here and the expression for

{\bar{\partial_{τ}}}^{*}

we obtain that

{\bar{\partial_{τ}}}^{*}

depends anti holomorphically on

τ

This can be viewed as the generalization of the results in [14] and [17] .

After these preliminary results we are ready to compute the Hessian of

log det Δ_{τ, q} .

If we know the expressions for

\frac{\partial^{2}}{\partial τ^{i} \bar{\partial τ^{j}}} log Δ_{τ, q}^{^{”}},

where

Δ_{τ, q}^{^{”}} = Δ_{τ, q} |_{Im {\bar{\partial}}^{*}}

then we know how to compute

\frac{\partial^{2}}{\partial τ^{i} \bar{\partial τ^{j}}} log Δ_{τ, q} .

So we will describe the ideas of the computations of

\frac{\partial^{2}}{\partial τ^{i} \bar{\partial τ^{j}}} log Δ_{τ, q}^{^{”}} .

Since by definition of the regularized determinants of the Laplacians are given by the formula

Δ_{τ, q}^{^{”}} = exp (- \frac{d}{d s} ζ_{Δ_{τ, q}^{^{”}}} (s)) |_{s = 0}

all our computations are based on the following expression

\begin{matrix} ζ_{Δ_{τ, q}^{^{”}}} (s) = \frac{1}{Γ (s)} \int_{0}^{\infty} T r (\frac{1}{4 π t^{n}} exp (- t Δ_{τ, q}^{”})) t^{s - 1} d t . \end{matrix}

(10)

From 9

()

and 10

()

we obtain

\frac{\bar{\partial}}{\bar{\partial τ^{i}}} (ζ_{q, τ}^{^{”}} (s)) |_{τ = 0} =

\begin{matrix} \frac{1}{Γ (s)} \int_{0}^{\infty} T r (exp (- t (△_{0, q}^{^{”}}) \circ △_{0, q}^{^{”}} \circ \partial_{0}^{- 1} \circ ℱ^{^{'}} (q + 1, \bar{φ_{i}}) \circ \bar{\partial_{0}}) t^{s} d t, \end{matrix}

(11)

where

φ_{i} \in H^{1} (

Θ

) are the Kodaira-Spencer classes viewed as bundle maps:

φ_{i} : C^{\infty} (M, Ω_{M}^{1, 0}) \to C^{\infty} (M, Ω_{M}^{0, 1})

and

ℱ (q + 1, φ_{i})

are the map induced by

φ_{i} \land i d_{q - 1} : C^{\infty} (M, Ω_{M}^{0, q}) \to C^{\infty} (M, Ω_{M}^{1, 0} \otimes Ω_{M}^{0, q - 1}) .

C^{\infty} (M, Ω_{M}^{0, q + 1})

and restricted in

Im \bar{\partial} .

From 11

()

we obtain

\frac{\partial^{2}}{\bar{\partial τ^{j}} \partial τ^{i}} (ζ_{τ, q}^{”} (s)) =

\frac{1}{Γ (s)} \int_{0}^{\infty} T r ((\frac{\partial}{\partial τ^{j}} exp (- t (△_{τ, q}^{^{”}})) \circ (△_{τ, q}^{^{”}}) \circ \partial_{τ}^{- 1} \circ F^{^{'}} (q + 1, \bar{φ_{i}}) \circ \bar{\partial_{0}}) t^{s} d t +

\begin{matrix} \frac{1}{Γ (s)} \int_{0}^{\infty} T r (exp (- t (△_{τ, q}^{^{”}}) \circ \frac{\partial}{\partial τ^{j}} (△_{τ, q}^{^{”}}) \circ \partial_{τ}^{- 1} \circ F^{^{'}} (q + 1, \bar{φ_{i}}) \circ \bar{\partial_{0}}) t^{s} d t . \end{matrix}

(12)

Following the ideas in [6] to integrate by parts 12

()

we derive that:

\begin{matrix} \frac{\partial^{2}}{\partial τ^{i} \bar{\partial τ^{j}}} ζ_{τ, q}^{”} (s) = \frac{s (1 - s)}{Γ (s)} \int_{0}^{\infty} T r (exp (- t Δ_{τ, q + 1}^{^{'}}) ℱ (q + 1, φ_{i} \circ \bar{φ_{j}})) t^{s - 1} d t . \end{matrix}

(13)

By using the short term asymptotic expansion of

exp (- t Δ_{τ, q + 1}^{^{'}})

we derive that

\begin{matrix} T r (exp (- t Δ_{τ, q + 1}^{^{'}}) ℱ (q + 1, φ_{i} \circ \bar{φ_{j}})) = \sum_{k = - n}^{1} \frac{α_{q, k}}{t^{k}} + α_{q, 0} + ψ_{q} (t) . \end{matrix}

(14)

Combining 13

()

and 14

()

we get that

\frac{d}{d s} (\frac{\partial^{2}}{\partial τ^{i} \bar{\partial τ^{j}}} ζ_{τ, q}^{”} (s)) |_{s = 0} = \frac{\partial^{2}}{\partial τ^{i} \bar{\partial τ^{j}}} log det Δ_{τ, q}^{^{”}} = α_{0, q} .

In this paper we gave an explicit formula for

α_{0, q} .

By using the expressions for

α_{0, 1}

and

α_{0, n - 1}

we show that

d \circ d^{c} (log det Δ_{τ, 1}) = - Im W P,

where

Im W P

is the imaginary part of the Weil-Petersson metric. This can be viewed as the generalization of the results in [14] and [17] .

1.3 Organization of the Paper

This article is organized as follows.

In Section 2 we introduce some basic notions about zeta functions of Laplacians on Riemannian manifolds. We review the results from [32] .

In Section 3 We review the basic properties of the Weil-Petersson metric on the moduli of CY manifolds. See also [24] .

In Section 4 we review the theory of moduli of CY manifolds following [22] and also metrics with logarithmic singularities on vector bundle following Mumford's article [?] . We prove that the

L^{2}

metric on the dualizing line bundle over the moduli space of CY manifolds is a good metric in the sense of Mumford. This implies that the volumes of the moduli spaces of CY manifolds are rational numbers.

In Section 5 we review some facts about the Hilbert spaces of the

(0, q)

forms and their isospectral identifications which we used in the paper. We also study traces the operators acting on the

L^{2}

sections of some vector bundle induced by some global

C^{\infty}

section its endormorphisms composed with the heat kernel. We study the short term expansions of these operators and especially the constant term of the expansion.

In Section 6 we establish the variational formulas for the zeta functions of the Laplacians and its regularized determinants. We also showed that

d d^{c} log det Δ_{τ, 1} = - Im W . P .

In Section 7 we prove that for Kähler Ricci flat manifold N with

H^{0} (

Ω_{M}^{n}) = 0

we have

d d^{c} log det Δ_{τ, n} = - Im W . P .

Acknowledgement 1 The second author wants to thank G. Moore, J. Jorgenson S. Donaldson, S.-T. Yau, G. Zuckerman, D. Kazhdan, P. Deligne, S. Lang, B. Lian, J. Li, K. Liu, Y. Eliashberg, Dai and R. Donagi for useful comments and support. I want to thank Sinan Unver for his help. Special thanks to The Institute of Mathematical Sciences at CUHK and Hangzhou for their hospitality during the preparation of this article.

Acknowledgement 2 The first author would like to thank Andrey Todorov for his patient guidance during this work. I also thank Harry Nishanian for his help throughout the years.

2 Preliminary Material

2.1 Basic Notions

Let M be a n-dimensional Kähler manifold with a zero canonical class.

Suppose that

H^{k} (

, O_{M}) = 0

for

1 \leq k < n .

Such manifolds are called Calabi-Yau manifolds. A pair (M,

L

) will be called a polarized CY manifold if M is a CY manifold and

L \in H^{2} (

Z

)^$1$ is a fixed class such that it represents the imaginary part of a Kähler metric on M. Yau's celebrated theorem asserts the existence of a unique Ricci flat Kähler metric g on M such that the cohomology class

[I m (g)] = L .

(See [34] .) From now on we will consider polarized CY manifolds of odd dimension. The polarization class

L

determines the CY metric g uniquely. We will denote by

△_{q} = {\bar{\partial}}^{*} \circ \bar{\partial} + \bar{\partial} \circ {\bar{\partial}}^{*}

the associated Laplacians that act on smooth

(0, q)

forms on M for

0 \leq q \leq n

{\bar{\partial}}^{*}

is the adjoint operator of

\bar{\partial}

with respect to the CY metric g.

The regularized determinants are defined as follows: Let (M,g) be an n-dimensional Riemannian manifold. Let

Δ_{q} = d d^{*} + d^{*} d

be the Laplacian acting on the space of q forms on M. We recall that the spectrum of the Laplacian

Δ_{q}

is positive and discrete. Thus the non zero eigen values of

Δ_{q}

are

0 < λ_{1} \leq λ_{2} \leq . . . \leq λ_{n} \leq . . .

We define the zeta function of

Δ_{q}

as follows:

ζ_{q} (s) = \sum_{i = 1}^{\infty} λ_{i}^{- s} .

It is known that

ζ_{q} (s)

is a well defined analytic function for

Re (s) ≫ C,

it has a meromorphic continuation in the complex plane and

0

is not a pole of

ζ_{q} (s) .

Define

det (Δ_{q}) = exp (- \frac{d}{d s} (ζ_{q} (s)) |_{s = 0}) .

The determinant of these operators

△_{q},

defined through zeta function regularization, will be denoted by det

(△_{q}) .

The Hodge decomposition theorem asserts that

Γ (

, Ω_{M}^{0, q}) = Im (\bar{\partial}) \oplus Im ({\bar{\partial}}^{*})

for

1 \leq q \leq {dim}_{C} M - 1 .

The restriction of

△_{q}

Im (\bar{\partial})

will be denoted by

△_{q}^{^{'}}

and

△_{q}^{^{'}} = \bar{\partial} \circ {\bar{\partial}}^{*}

and the restriction of

Δ_{q}

Im ({\bar{\partial}}^{*})

will be denoted by

△_{q}^{”}

and

= {\bar{\partial}}^{*} \circ \bar{\partial} .

Hence we have

T r (exp (- t △_{q}) = T r (exp (- t △_{q}^{^{'}}) + T r (exp (- t △_{q}^{”}) .

This implies that

ζ_{q} (s) = \sum_{k = 1}^{\infty} λ_{k}^{- s} = ζ_{q}^{^{'}} (s) + ζ_{q}^{”} (s),

where

λ_{k} > 0

are the positive eigen values of

△_{q}

and

ζ_{q}^{^{'}} (s)

ζ_{q}^{”} (s)

are the zeta functions of

△_{q}^{^{'}}

and

△_{q}^{”} .

From here and the definition of the regularized determinant we obtain that

log det (△_{q}) = log det (△_{q}^{^{'}}) + log det (△_{q}^{”}) .

It is a well known fact that the action of

△_{q}^{^{”}}

Im {\bar{\partial}}^{*}

is isospectral to the action of

△_{q + 1}^{^{'}}

Im \bar{\partial},

which means that the spectrum of

△_{q}^{^{”}}

is equal to the spectrum of

△_{q + 1}^{^{'}} .

So we have the equality

det (△_{q}^{”}) = det (△_{q + 1}^{^{'}}) .

^$1$ Notice that $H^{1, 1} ($ M, $R) = H^{2} ($ M, $R)$ since $H^{2}$ (M, $O_{M}) = 0$ for CY manifolds.

2.2 Basic Notions about Complex Structures

Let M be an even dimensional C

^{\infty}

manifold. We will say that M has an almost complex structure if there exists a section

I \in C^{\infty} (

, H o m (T^{*}, T^{*})

such that

I^{2} = - i d .

T

is the tangent bundle and

T^{*}

is the cotangent bundle on M. This definition is equivalent to the following one: Let M be an even dimensional C

^{\infty}

manifold. Suppose that there exists a global splitting of the complexified cotangent bundle

T_{M}^{*} \otimes C = Ω_{M}^{1, 0} \oplus Ω_{M}^{0, 1},

where

Ω_{M}^{0, 1} = \bar{Ω_{M}^{1, 0}} .

Then we will say that M has an almost complex structure. We will say that an almost complex structure is an integrable one, if for each point

x \in

M there exists an open set

U \subset

M such that we can find local coordinates

z^{1}, . ., z^{n},

such that

d z^{1}, . ., d z^{n}

are linearly independent in each point

m \in U

and they generate

Ω_{M}^{1, 0} |_{U} .

Definition 3 Let M be a complex manifold. Let

φ \in Γ (

, H o m (Ω_{M}^{1, 0}, Ω_{M}^{0, 1})),

then we will call

φ

a Beltrami differential.

Since

Γ (

, H o m (Ω_{M}^{1, 0}, Ω_{M}^{0, 1})) ⋍ Γ (

, Ω_{M}^{0, 1} \otimes T_{M}^{1, 0}),

we deduce that locally

φ

can be written as follows:

φ |_{U} = \sum φ_{\bar{α}}^{β} {\bar{d z}}^{α} \otimes \frac{\partial}{\partial z^{β}} .

From now on we will denote by

A_{φ}

the following linear operator:

A_{φ} = (\begin{matrix} i d & φ (τ) \\ \bar{φ (τ)} & i d \end{matrix}) .

We will consider only those Beltrami differentials

φ

such that det(

A_{φ}) \neq 0 .

The Beltrami differential

φ

defines an integrable complex structure on M if and only if the following equation holds:

\begin{matrix} \bar{\partial} φ = \frac{1}{2} [φ, φ], \end{matrix}

(15)

where

\begin{matrix} [φ, φ] |_{U} : = \sum_{ν = 1}^{n} \sum_{1 ≦ α < β ≦ n} (\sum_{μ = 1}^{n} (φ_{\bar{α}}^{μ} (\partial_{μ} φ_{\bar{β}}^{ν}) - φ_{\bar{β}}^{μ} (\partial_{ν} φ_{\bar{α}}^{ν}))) {\bar{d z}}^{α} \land {\bar{d z}}^{β} \otimes \frac{\partial}{d z^{ν}} \end{matrix}

(16)

(See [20] .)

2.3 Kuranishi Space and Flat Local Coordinates

Kuranishi proved the following Theorem:

Theorem 4 Let

{φ_{i}}

be a basis of harmonic

(0, 1)

forms of

H^{1} (

, T^{1, 0})

on a Hermitian manifold M. Let

G

be the Green operator and let

φ (τ^{1}, . ., τ^{N})

be defined as follows:

\begin{matrix} φ (τ^{1}, . . ., τ^{N}) = \sum_{i = 1}^{N} φ_{i} τ^{i} + \frac{1}{2} {\bar{\partial}}^{*} G [φ (τ^{1}, . . ., τ^{N}), φ (τ^{1}, . . ., τ^{N})] . \end{matrix}

(17)

There exists

ɛ > 0

such that if

τ = (τ^{1}, . . ., τ^{N})

satisfies

| τ_{i} | < ɛ

then

φ (τ^{1}, . . ., τ^{N})

is a global

C^{\infty}

section of the bundle

Ω_{M}^{(0, 1)} \otimes T^{1, 0}

.(See [20] .)

Based on Theorem 4 , we proved in [32] the following Theorem:

Theorem 5 Let M be a CY manifold and let

{φ_{i}}

be a basis of harmonic

(0, 1)

forms with coefficients in

T^{1, 0},

i.e.

{φ_{i}} \in H^{1} (

, T^{1, 0}),

then the equation 15

()

has a solution in the form:

φ (τ^{1}, . . ., τ^{N}) = \sum_{i = 1}^{N} φ_{i} τ^{i} + \sum_{| I_{N} | ≧ 2} φ_{I_{N}} τ^{I_{N}} =

\sum_{i = 1}^{N} φ_{i} τ^{i} + \frac{1}{2} {\bar{\partial}}^{*} G [φ (τ^{1}, . . ., τ^{N}), φ (τ^{1}, . . ., τ^{N})]

and

{\bar{\partial}}^{*} φ (τ^{1}, . . ., τ^{N}) = 0,

φ_{I_{N}} ⌟ ω_{M} = \partial ψ_{I_{N}}

where

I_{N} = (i_{1}, . . ., i_{N})

is a multi-index,

φ_{I_{N}} \in C^{\infty} (

, Ω_{M}^{0, 1} \otimes T^{1, 0}), τ^{I_{N}} = (τ^{i})^{i_{1}} . . . (τ^{N})^{i_{N}}

and if for some

ɛ > 0

| τ^{i} | < ɛ

then

φ (τ) \in C^{\infty} (

, Ω_{M}^{0, 1} \otimes T^{1, 0})

where

i = 1, . . ., N .

[31 ] [32 ]

(S e e a n d) .

It is a standard fact from Kodaira-Spencer-Kuranishi deformation theory that for each

τ = (τ^{1}, . . ., τ^{N}) \in K

as in Theorem 5 the Beltrami differential

φ (τ^{1}, . . ., τ^{N})

defines a new integrable complex structure on M. This means that the points of

K,

where

K : {τ = (τ^{1}, . . ., τ^{N}) | | τ^{i} | < ɛ}

defines a family of operators

{\bar{\partial}}_{τ}

on the

C^{\infty}

family

K \times

\to

M and

{\bar{\partial}}_{τ}

are integrable in the sense of Newlander-Nirenberg. Moreover it was proved by Kodaira, Spencer and Kuranishi that we get a complex analytic family of CY manifolds

π : X \to K,

where as

C^{\infty}

manifold

X ⋍ K \times

.

The family

\begin{matrix} π : X \to K \end{matrix}

(18)

is called the Kuranishi family. The operators

{\bar{\partial}}_{τ}

are defined as follows:

Definition 6 Let

{U_{i}}

be an open covering of M, with local coordinate system in

U_{i}

given by

{z_{i}^{k}}

with

k = 1, . . ., n =

dim

_{C}

M. Assume that

φ (τ^{1}, . . ., τ^{N}) |_{U_{i}}

is given by:

φ (τ^{1}, . . ., τ^{N}) = \sum_{j, k = 1}^{n} (φ (τ^{1}, . . ., τ^{N}))_{\bar{j}}^{k} d {\bar{z}}^{j} \otimes \frac{\partial}{\partial z^{k}} .

Then we define

\begin{matrix} {(\bar{\partial})}_{τ, \bar{j}} = \frac{\bar{\partial}}{\bar{\partial z^{j}}} - \sum_{k = 1}^{n} (φ (τ^{1}, . . ., τ^{N}))_{\bar{j}}^{k} \frac{\partial}{\partial z^{k}} . \end{matrix}

(19)

Definition 7 The coordinates

τ = (τ^{1}, . . ., τ^{N})

defined in Theorem 5 , will be fixed from now on and will be called the flat coordinate system in

K

2.4 Family of Holomorphic Forms

In [32] the following Theorem is proved:

Theorem 8 There exists a family of holomorphic forms

ω_{τ}

of the Kuranishi family 18

()

such that

〈 [ω_{τ}], [ω_{τ}] 〉 =

1 - \sum_{i, j} 〈 ω_{0} ⌟ φ_{i}, ω_{0} ⌟ φ_{j} 〉 τ^{i} \bar{τ^{j}} + \sum_{i, j} 〈 ω_{0} ⌟ (φ_{i} \land φ_{k}), ω_{0} ⌟ (φ_{j} \land φ_{l}) 〉 τ^{i} \bar{τ^{j}} τ^{k} \bar{τ^{l}} + O (τ^{5}) =

1 - \sum_{i, j} τ^{i} \bar{τ^{j}} + \sum_{i, j} 〈 ω_{0} ⌟ (φ_{i} \land φ_{k}), ω_{0} ⌟ (φ_{j} \land φ_{l}) 〉 τ^{i} \bar{τ^{j}} τ^{k} \bar{τ^{l}} + O (τ^{5}) and

\begin{matrix} 〈 [ω_{τ}], [ω_{τ}] 〉 \leq 〈 [ω_{0}], [ω_{0}] 〉 . \end{matrix}

(20)

3 Weil-Petersson Metric

3.1 Basic Properties

It is a well known fact from Kodaira-Spencer-Kuranishi theory that the tangent space

T_{τ, K}

at a point

τ \in K

can be identified with the space of harmonic (0,1) forms with values in the holomorphic vector fields

H^{1} (

_{τ}, T

). We will view each element

φ \in H^{1} (

_{τ}, T

) as a point wise linear map from

Ω_{M_{τ}}^{(1, 0)}

Ω_{M_{τ}}^{(0, 1)} .

Given

φ_{1}

and

φ_{2} \in H^{1} (

_{τ}, T

)

,

the trace of the map

φ_{1} \circ \bar{φ_{2}} : Ω_{M_{τ}}^{(0, 1)} \to Ω_{M_{τ}}^{(0, 1)}

at the point

m \in

_{τ}

with respect to the metric g is simply:

\begin{matrix} T r (φ_{1} \circ \bar{φ_{2}}) (m) = \sum_{k, l, m = 1}^{n} (φ_{1})_{\bar{l}}^{k} (\bar{φ)_{\bar{k}}^{m}} g^{\bar{l}, k} g_{k, \bar{m}} \end{matrix}

(21)

Definition 9 We will define the Weil-Petersson metric on

K

via the scalar product:

\begin{matrix} 〈 φ_{1}, φ_{2} 〉 = \int_{M} T r (φ_{1} \circ \bar{φ_{2}}) v o l (g) . \end{matrix}

(22)

We proved in [32] that the coordinates

τ = (τ^{1}, . . ., τ^{N})

as defined in Definition 7 are flat in the sense that the Weil-Petersson metric is Kähler and in these coordinates we have that the components

g_{i, \bar{j}}

of the Weil Petersson metric are given by the following formulas:

g_{i, \bar{j}} = δ_{i, \bar{j}} + R_{i, \bar{j}, l, \bar{k}} τ^{l} \bar{τ^{k}} + O (τ^{3}) .

Very detailed treatment of the Weil-Petersson geometry of the moduli space of polarized CY manifolds can be found in [23] and [24] . In those two papers important results are obtained.

3.2 Infinitesimal Deformation of the Imaginary Part of the WP Metric

Theorem 10 Near each point

τ_{0}

of the Kuranishi space

K,

the imaginary part

Im (g)

of the CY metric

g

has the following expansion in the coordinates

τ : = (τ^{1}, . . ., τ^{N})

Im (g) (τ, \bar{τ}) = Im (g) (τ_{0}) + O ((τ - τ_{0})^{2}) .

Proof: Without loss of generality we can assume that

τ_{0} = 0 .

In [32] we proved that the forms

\begin{matrix} θ_{τ}^{k} = d z^{k} + \sum_{l = 1} (φ (τ^{1}, . . ., τ^{N})_{\bar{l}}^{k}) d \bar{z^{l}} \end{matrix}

(23)

for

k = 1, . . ., n

form a basis of

(1, 0)

forms relative to the complex structure defined by

τ \in K

U \subset

M. Let

\begin{matrix} Im (g_{τ}) = \sqrt{- 1} (\sum_{1 \leq k \leq l \leq n} g_{k, \bar{l}} (τ, \bar{τ}) θ_{τ}^{k} \land \bar{θ_{τ}^{l}}) \end{matrix}

(24)

and

\begin{matrix} g_{k, \bar{l}} (τ, \bar{τ}) = g_{k, \bar{l}} (0) + \sum_{i = 1}^{N} ({(g_{k, \bar{l}} (1))}_{i} τ^{i} + {(g_{k, \bar{l}}^{^{'}} (1))}_{\bar{i}} \bar{τ^{i}}) + O (2) . \end{matrix}

(25)

We get the following expression for

Im (g_{τ})

in terms of

d z^{i}

and

\bar{d z^{j}}

, by substituting the expressions for

θ_{τ}^{k}

from 23

()

and the expressions for

g_{k, \bar{l}} (τ, \bar{τ})

from formula 25

()

in the formula 24

()

Im (g_{τ}) = \sqrt{- 1} (\sum_{1 \leq k \leq l \leq n} g_{k, \bar{l}} (τ, \bar{τ}) θ_{τ}^{k} \land \bar{θ_{τ}^{l}}) = \sqrt{- 1} (\sum_{1 \leq k \leq l \leq n} g_{k, \bar{l}} (0) d z^{k} \land \bar{d z^{l}}) +

+ \sqrt{- 1} (\sum_{i = 1}^{N} τ^{i} (\sum_{1 \leq k \leq l \leq n} ({(g_{k, \bar{l}} (1))}_{i} d z^{k} \land \bar{d z^{l}} + \sum_{m = 1}^{n} (g_{k, \bar{m}} \bar{φ_{i, \bar{l}}^{m}} - g_{l, \bar{m}} \bar{φ_{i, \bar{k}}^{m}}) d z^{k} \land d z^{l})))

+ \frac{1}{\sqrt{- 1}} \sum_{i = 1}^{N} \bar{τ^{i}} \bar{(\sum_{1 \leq k \leq l \leq n} ({(g_{k, \bar{l}} (1))}_{i} d z^{k} \land \bar{d z^{l}} + (\sum_{m = 1}^{n} (g_{k, \bar{m}} \bar{φ_{i, \bar{l}}^{m}} - g_{l, \bar{m}} \bar{φ_{i, \bar{k}}^{m}}) d z^{k} \land d z^{l}))} .

On page 332 of [32] the following results is proved:

Lemma 11 Let

φ \in H^{1} (

, T

) be a harmonic form with respect to the CY metric g. Let

φ |_{U} = \sum_{k, l = 1}^{n} φ_{\bar{k}}^{l} {\bar{d z}}^{k} \otimes \frac{\partial}{\partial z^{l}},

then

φ_{\bar{k}, \bar{l}} = \sum_{j = 1}^{n} g_{j, \bar{k}} φ_{\bar{l}}^{j} = \sum_{j = 1}^{n} g_{j, \bar{l}} φ_{\bar{k}}^{j} = φ_{\bar{l}, \bar{k}} .

From Lemma 11 we conclude that

\begin{matrix} \sum_{m = 1}^{n} (g_{k, \bar{m}} \bar{φ_{i, \bar{l}}^{m}} - g_{l, \bar{m}} \bar{φ_{i, \bar{k}}^{m}}) = 0 . \end{matrix}

(26)

From 26

()

we get the following expression for

Im (g_{τ})

Im (g_{τ}) = \sqrt{- 1} (\sum_{1 \leq k \leq l \leq n} g_{k, \bar{l}} (0) d z^{k} \land \bar{d z^{l}}) +

\sqrt{- 1} (\sum_{i = 1}^{N} τ^{i} (\sum_{1 \leq k \leq l \leq n} {(g_{k, \bar{l}} (1))}_{i} d z^{k} \land \bar{d z^{l}})) +

\begin{matrix} + \sqrt{- 1} (\sum_{i = 1}^{N} \bar{τ^{i}} \bar{\sum_{1 \leq k \leq l \leq n} {(g_{k, \bar{l}} (1))}_{i} d z^{k} \land \bar{d z^{l}}}) + O (2) \end{matrix}

(27)

Let us define the (1,1) forms

ψ_{i}

as follows:

\begin{matrix} ψ_{i} = \sqrt{- 1} (\sum_{1 \leq k \leq l \leq n} {(g_{k, \bar{l}} (1))}_{i} d z^{k} \land \bar{d z^{l}}) \end{matrix}

(28)

We derive the following formula, by substituting in the expression 27

()

the expression given by 28

()

\begin{matrix} Im (g_{τ}) = Im (g_{0}) + \sum_{i = 1}^{N} τ^{i} ψ_{i} + \sum_{i = 1}^{N} \bar{τ^{i} ψ_{i}} + O (τ^{2}) \end{matrix}

(29)

From the fact that the class of the cohomology of the imaginary part of the CY metric is fixed, i.e.

[Im (g_{τ})] = [Im (g_{0})] = L,

and 29

()

we deduce that each

ψ_{i}

is an exact form, i.e.

\begin{matrix} ψ_{i} = \sqrt{- 1} \partial \bar{\partial} f_{i}, \end{matrix}

(30)

where

f_{i}

are globally defined functions on M. Our Theorem will follow if we prove that

ψ_{i} = 0 .

Lemma 12

ψ_{i} = 0 .

Proof: In [32] we proved that

\begin{matrix} det (g_{τ}) = \land^{n} Im (g_{τ}) = det (g_{0}) + O (2) \end{matrix}

(31)

in the flat coordinates

(τ^{1}, . . ., τ^{N}) .

We deduce from the expressions 29

()

and 30

()

, by direct computations that:

det (g_{τ}) = det (g_{0}) + \sqrt{- 1} \sum_{i = 1}^{N} τ^{i} (\sum_{k, l} g^{\bar{l}, k} \partial_{k} \bar{\partial_{l}} (f_{i})) +

\begin{matrix} \frac{1}{\sqrt{- 1}} \sum_{i = 1}^{N} \bar{τ^{i}} \bar{(\sum_{k, l} g^{\bar{l}, k} \partial_{k} \bar{\partial_{l}} (f_{i}))} + O (2) . \end{matrix}

(32)

Combining 31

()

and 32

()

we obtain that for each i we have:

\sum_{k, l} g^{\bar{l}, k} \partial_{k} \bar{\partial_{l}} (f_{i}) = △ (f_{i}) = 0,

where

△

is the Laplacian of the metric g. From the maximum principle, we deduce that all

f_{i}

are constants. Formula 30

()

implies that

ψ_{i} = 0 .

Lemma 12 is proved.

■

Theorem 10 follows directly from Lemma 12 . Theorem 10 is proved.

■

Corollary 13 The imaginary part

Im

_{τ}

of the CY metric is a constant symplectic form on the moduli space

M_{L}

(M).

Corollary 14 The following formulas are true:

\begin{matrix} \frac{\partial}{\partial τ_{i}} {(\bar{\partial_{τ}})}^{*} = 0 and \frac{\bar{\partial}}{\bar{\partial τ_{i}}} {(\partial_{τ})}^{*} = 0 . \end{matrix}

(33)

Proof: We know from Kähler geometry that

(\bar{\partial_{τ}})^{*} = [Λ_{τ}, \partial_{τ}],

where

Λ_{τ}

is the contraction with (1,1) vector field

:

\begin{matrix} Λ_{τ} = \frac{\sqrt{- 1}}{2} \sum_{k, l = 1}^{n} g_{τ}^{\bar{k}, l} (θ_{τ}^{l})^{*} \land (\bar{θ_{τ}^{k}})^{*} \end{matrix}

(34)

on M

_{τ}

and

(θ_{τ}^{l})^{*}

is (1,0) vector field on M

_{τ}

dual to the (1,0) form

θ_{τ}^{i} = d z^{i} + \sum_{j = 1}^{N} τ^{j} (\sum_{k = 1}^{n} (φ_{j})_{\bar{k}}^{i} {\bar{d z}}^{k})) .

Corollary 13 implies that

\frac{\partial}{\partial τ_{i}} (Λ_{τ}) = 0 .

On the other hand,

\partial_{τ}

depends antiholomorphically on

τ

, i.e. it depends on

\bar{τ} = (\bar{τ_{1}}, . . ., \bar{τ_{N}}) .

So we deduce that:

\frac{\partial}{\partial τ_{i}} ((\bar{\partial_{τ}})^{*}) = ([\frac{\partial}{\partial τ_{i}} (Λ_{τ}), \partial_{τ}] + [Λ_{τ}, \frac{\partial}{\partial τ_{i}} (\partial_{τ})]) = 0 .

Exactly in the same way we prove that

\frac{\bar{\partial}}{\bar{\partial τ_{i}}} {(\partial_{τ})}^{*} = 0 .

Corollary 14 is proved.

■

4 Moduli of CY Manifolds

4.1 Basic Construction

Definition 15 We will define the Teichmüller space

T

(M) of a CY manifold M as follows:

T (

) : = ℐ (

) / D i f f_{0} (

),

where

ℐ (M) : = {all integrable complex structures on M}

and Diff

_{0}

(M) is the group of diffeomorphisms isotopic to identity. The action of the group Diff(M

_{0})

is defined as follows; Let

φ \in

Diff

_{0}

(M) then

φ

acts on integrable complex structures on M by pull back, i.e. if

I \in C^{\infty} (M, H o m (T (

), T (

)),

then we define

φ (I_{τ}) = φ^{*} (I_{τ}) .

Definition 16 We will call a pair

(

; γ_{1}, . . ., γ_{b_{n}})

a marked CY manifold where M is a CY manifold and

{γ_{1}, . . ., γ_{b_{n}}}

is a basis of

H_{n}

(M,

Z

)/Tor.

Remark 17 Let

K

be the Kuranishi space. It is easy to see that if we choose a basis of

H_{n}

(M,

Z

)/Tor in one of the fibres of the Kuranishi family

π : ℳ \to K

then all the fibres will be marked, since as a

C^{\infty}

manifold

X_{K} ≊

\times K

In [22] the following Theorem was proved:

Theorem 18 There exists a family of marked polarized CY manifolds

\begin{matrix} Z_{L} \to \tilde{T} (M), \end{matrix}

(35)

which possesses the following properties: a) It is effectively parametrized, b) The base has dimension

h^{n - 1, 1},

c) For any marked CY manifold M of fixed topological type for which the polarization class

L

defines an imbedding into a projective space

C P^{N},

there exists an isomorphism of it (as a marked CY manifold) with a fibre M

_{s}

of the family

Z_{L} .

Corollary 19 Let

Y \to

X be any family of marked polarized CY manifolds, then there exists a unique holomorphic map

φ :

\to \tilde{T} (

)

up to a biholomorphic map

ψ

of M which induces the identity map on

H_{n} (

, Z) .

From now on we will denote by

T

(M) the irreducible component of the Teichmüller space that contains our fixed CY manifold M.

Definition 20 We will define the mapping class group

Γ_{1} (

)

of any compact C

^{\infty}

manifold M as follows:

Γ_{1}

) = D i f f_{+} (

) / D i f f_{0} (

),

where

D i f f_{+} (

)

is the group of diffeomorphisms of M preserving the orientation of M and

D i f f_{0} (

)

is the group of diffeomorphisms isotopic to identity.

Definition 21 Let

L \in H^{2} (

, Z)

be the imaginary part of a Kähler metric.

Let

Γ_{2} : = {φ \in Γ_{1} (

) | φ (L) = L} .

It is a well know fact that the moduli space of polarized algebraic manifolds

ℳ_{L} (

) = T (

) / Γ_{2} .

In [22] the following fact was established:

Theorem 22 There exists a subgroup of finite index

Γ_{L}

Γ_{2}

such that

Γ_{L}

acts freely on

T

(M) and

Γ \ T

(M)

= M_{L}

(M) is a non-singular quasi-projective variety. Over

M_{L}

(M) there exists a family of polarized CY manifolds

ℳ \to M_{L} (

) .

Remark 23 Theorem 22 implies that we constructed a family of non-singular CY manifolds

π : X \to M_{L} (

)

over a quasi-projective non-singular variety

M_{L} (

)

. Moreover it is easy to see that

X \subset C P^{N} \times M_{L} (

) .

X

is also quasi-projective. From now on we will work only with this family.

5 Hilbert Spaces and Trace Class Operators

5.1 Preliminary Material

Definition 24 We will denote by

L_{0, q}^{2} (Im ({\bar{\partial}}^{*}))

the Hilbert subspace in

L^{2} (

Ω_{M}^{(0, q)})

which is the

L^{2}

completion of

\bar{\partial^{*}}

exact forms in

C^{\infty} (

Ω_{M}^{(0, q)})

for

q \geq 0 .

In the same manner we will denote by

L_{1, q - 1}^{2} (Im (\partial))

the Hilbert subspace in

L^{2} (

Ω_{M}^{(1, q - 1)})

which is the

L^{2}

competition of the

\partial

exact

(1, q - 1)

forms in

C^{\infty} (

Ω_{M}^{(1, q - 1)})

for

q > 0

. All the completions are with respect to the scalar product on the bundles

Ω_{M}^{p, q}

defined by the CY metric g.

Let

φ (τ^{1}, . ., τ^{N}

) be a solution of the equation 15

()

\bar{\partial} φ (τ^{1}, . . ., τ^{N}) = \frac{1}{2} [φ (τ^{1}, . . ., τ^{N}), φ (τ^{1}, . . ., τ^{N})]

which is guaranteed by Theorem 5 . From Definition 3 of the Beltrami differential, we know that the Beltrami differential

φ (τ^{1}, . ., τ^{N}

) defines a linear fibrewise map

φ (τ^{1}, . . ., τ^{N}) : Ω_{M}^{(1, 0)} \to Ω_{M}^{(0, 1)} .

\begin{matrix} φ (τ^{1}, . . ., τ^{N}) \in C^{\infty} (M, H o m (Ω_{M}^{(1, 0)}, Ω_{M}^{(0, 1)}) . \end{matrix}

(36)

Definition 25 We define the following maps between vector bundles

φ \land i d : Ω_{M}^{(1, q - 1)} \to Ω_{M}^{(0, q)}

φ (d z^{i} \land α) = φ (d z^{i}) \land α

for each

1 \leq q \leq n .

Clearly each fibre wise linear map

φ \land i d_{q - 1}

defines a natural linear operator

F (q, φ) : L^{2} (M, Ω_{M}^{(1, q - 1)}) \to L^{2} (M, Ω_{M}^{(0, q)})

between the Hilbert spaces. The restriction of the linear operator F(q,

φ

) on the subspace

Im (\partial) \subset L^{2} (

, Ω_{M}^{(1, q - 1)}))

Im (\bar{\partial}) \subset L^{2} (

, Ω_{M}^{(0, q)})

will be denoted by

F^{'} (q, φ)

. The restriction of the linear operator F(q,

φ

) on the subspace

Im (\partial^{*}) \subset L^{2} (

, Ω_{M}^{(1, q - 1)}))

Im ({\bar{\partial}}^{*}) \subset L^{2} (

, Ω_{M}^{(0, q)})

will be denoted by

F^{”} (q, φ)

. Let

φ

and

ψ

be two Kodaira Spencer classes and let

φ \circ \bar{ψ} : L^{2} (M, Ω_{M}^{(0, 1)}) \to L^{2} (M, Ω_{M}^{(0, 1)})

be fibrewise linear map given by

\begin{matrix} φ \circ \bar{ψ} |_{U} : = \sum_{α, β = 1}^{n} {(φ \circ \bar{ψ})}_{\bar{β}}^{\bar{α}} \bar{d z^{β}} \otimes \frac{\partial}{\bar{\partial z^{α}}} . \end{matrix}

(37)

We define the fibrewise bundle maps

\begin{matrix} (φ \circ \bar{ψ}) \land i d_{q - 1} : Ω_{M}^{0, q} \to Ω_{M}^{0, q}, \end{matrix}

(38)

as follows:

\begin{matrix} ((φ \circ \bar{ψ}) \land i d_{q - 1}) (ω) : = (\sum_{α, β = 1}^{n} {(φ \circ \bar{ψ})}_{\bar{β}}^{\bar{α}} \bar{d z^{β}} \otimes \frac{\bar{\partial}}{\bar{\partial z^{α}}}) ⌟ ω, \end{matrix}

(39)

where

⌟

means contraction of tensors and

ω

is some global form of type

(0, q)

on M. We will define for the linear operators

\begin{matrix} ℱ^{^{'}} (q, φ \circ \bar{ψ}) : L_{0, q}^{2} (Im (\bar{\partial})) \to L_{0, q}^{2} (Im (\bar{\partial})) \end{matrix}

(40)

and

\begin{matrix} ℱ^{”} (q, φ \circ \bar{ψ}) : L_{0, q}^{2} (Im (\bar{\partial})^{*}) \to L_{0, q}^{2} (Im (\bar{\partial})^{*}) \end{matrix}

(41)

as the restriction of the operators

((φ \circ \bar{ψ}) \land i d_{q - 1})

L_{0, q}^{2} (Im (\bar{\partial}))

and

L_{0, q}^{2} (Im (\bar{\partial})^{*})

respectively.

Remark 26 It is a standard fact that we can choose globally

\bar{\partial}

closed forms

ω_{1}, . . ., ω_{N}

of type

(0, q)

such that at each point

z \in

M they span the fibre

Ω_{M, z}^{0, q} .

We can deduce directly from the definitions of the operators

F^{'} (q, φ), ℱ^{^{'}} (q, φ \circ \bar{ψ})

and

F^{^{'}} (q, \bar{ψ} \circ φ)

and the existence of the forms

ω_{1}, . . ., ω_{N}

that the operators

F^{'} (q, φ), ℱ^{^{'}} (q, φ \circ \bar{ψ})

and

F^{^{'}} (q, \bar{ψ} \circ φ)

pointwise will be represented by matrices of dimensions

(\binom{n}{q}), (\binom{n}{q})

and

n \times (\binom{n}{q - 1}) .

5.2 Trace Class Operators (See [3] )

Let

H

be a Hilbert space with a orthonormal basis

e_{i} .

An operator

A

is a Hilbert-Schmidt operator if

{∥ A ∥}_{H S}^{2} = \sum_{i} {∥ A e_{i} ∥}^{2} = \sum_{i j} {| 〈 A e_{i}, e_{j} 〉 |}^{2} < \infty

is finite. The number

{∥ A ∥}_{H S}^{2}

is called the Hilbert-Schmidt norm of A. If A is a Hilbert-Schmid so is its adjoint

A^{*}

and

{∥ A ∥}_{H S}^{2} = {∥ A^{*} ∥}_{H S}^{2} .

U

is a bounded operator on

H

and

A

is an Hilbert-Schmidt, then

U \circ A

and

A \circ U

are Hilbert-Schmidt operators and

{∥ U \circ A ∥}_{H S} \leq {∥ A \circ U ∥}_{H S} .

In this paper we will consider the Hilbert spaces of the square integrable sections of the bundles

Ω_{M}^{0, q} \otimes {| Λ_{M} |}^{1 / 2}

on M, where

| Λ_{M} |

is the trivial density bundles generated by the volume form of the CY metric.

An operator

K

with square-integrable kernel

k (w, z) \in Γ_{L^{2}} (M \times M, (Ω_{M}^{0, q} \otimes {| Λ_{M} |}^{1 / 2} ⊠ Ω_{M}^{0, q} \otimes {| Λ_{M} |}^{1 / 2}))

is Hilbert-Schmidt, and

\begin{matrix} {∥ K ∥}_{H S}^{2} = \int_{(w, z) \in M \times M} T r (k {(w, z)}^{*} k (w, z)) . \end{matrix}

(42)

Formula 42

()

follows from the definition of the Hilbert-Schmidt norm

{∥ K ∥}_{H S}^{2} = \sum_{i, j} {| 〈 K e_{i}, e_{j} 〉 |}^{2} .

An operator

K

is said to be trace class if it has the form

A \circ B,

where

A

and

B

are Hilbert-Schmidt. For such operators, the sum

T r K = \sum_{i} 〈 K e_{i}, e_{i} 〉

is absolutely summable and

T r K

is independent of the choice of the orthonormal basis in

H

and is called the trace of

K .

5.3 Adiabatic Limits, Heat Kernels and Traces

In this subsection we study the traces of operators which are compositions of the heat kernel with operators induced by endomorphisms of some vector bundle. We will use some of the results from [6] and will adopt them to our situation.

Let

h

be a metric on a vector bundle

E

over M. Let

Δ_{h}

be the Laplacian on

E

. It is a well known fact that the operator

exp (- t △_{h})

can be represented by an integral kernel:

k_{t} (w, z, τ) = \sum_{j} exp (- t λ_{j}) φ_{j} (ω) \otimes φ_{j} (z),

where

λ_{j}

and

φ_{j}

are the eigen values and the eigen sections of the Laplace operator

Δ_{h}

on some vector bundle

E

on M

.

k_{t} (w, z, τ)

is an operator of trace class. We know that the following formula holds for the short term asymptotic expansion of

T r (k_{t} (w, z, τ))

T r (k_{t} (w, z, τ)) = \frac{α_{- n}}{t^{n}} + . . . + \frac{α_{- 1}}{t} + α_{0} + O (t) .

Let

E

be a holomorphic vector bundle over M, let

φ \in C^{\infty} (

M,Hom(

E, E

)). It is easy to see that the operator

exp (- Δ_{h}) \circ φ

is of trace class and its trace has an asymptotic expansion

\begin{matrix} T r (k_{t} (w, z, τ) \circ φ) = \frac{β_{- n} (φ)}{t^{n}} + . . . + \frac{β_{- 1} (φ)}{t} + β_{0} (φ) + O (t) \end{matrix}

(43)

according to [3] . We will study the following problem in this section:

Problem 27 Find an explicit expression for

β_{0} (φ) .

Definition 28 We define the function

k_{τ}^{d} (w, z, t)

in a neighborhood of the diagonal

Δ

in M

\times

M as follows: Let

ρ_{τ}

be the injectivity radius on M

_{τ} .

Let

d_{τ} (w, z)

be the distance between the points

w

and

z

on M

_{τ}

with respect to CY metric g

_{τ} .

We suppose that

| τ | < ɛ .

Let

δ

be such that

δ > ρ_{τ} .

Then we define the function

k_{τ}^{d} (w, z, t)

as a C

^{\infty}

function on M

\times

M using partition of unity by using the functions

\begin{matrix} k_{t}^{d} (w, z, τ) = {\begin{matrix} \frac{1}{(4 π t)^{\frac{n}{2}}} exp (- \frac{d_{τ}^{2} (w, z)}{4 t}) & i f d_{τ} (w, z) < ρ_{τ} \\ 0 & i f d_{τ} (w, z) > δ . \end{matrix} \end{matrix}

(44)

defined on the opened balls around countable points

(w_{k}, z_{k})

on M

\times

M with injectivity radius

ρ_{τ} .

Let

E

be a holomorphic vector bundle on M with a Hermitian metric

h

on it and let

P_{τ} (w, z)

be the parallel transport of the bundle

E

along the minimal geodesic joining the point

w

and

z

with respect to natural connection on

E

induced by the metric

h

E .

It was proved in [3] on page 87 that we can represent the operator

exp (- t △_{h})

by an integral kernel

k_{t} (w, z, τ)

, where

\begin{matrix} k_{t} (w, z, τ) = k_{t}^{d} (w, z, τ) (P_{τ} (w, z) + O (t)) \end{matrix}

(45)

and

Δ_{h} : = {\bar{\partial}}_{h}^{*} \circ \bar{\partial} .

Definition 29 We will define the kernel

k_{t}^{#} (w, z, τ)

as the matrix operator defined as follows

\begin{matrix} k_{t}^{#} (w, z, τ) = k_{t}^{d} (w, z, τ) P_{τ} (w, z) . \end{matrix}

(46)

So we have the following formula

\begin{matrix} k_{t} (w, z, τ) = k_{t}^{#} (w, z, τ) + ɛ_{t} (w, z, τ) . \end{matrix}

(47)

Let us define

\begin{matrix} ϒ_{t} (φ, τ, z) : = \int_{M} T r ({(k_{t / 2}^{d} (w, z, τ))}^{*} \circ (k_{t / 2}^{#} (w, z, τ) \circ φ)) v o l (g)_{w} . \end{matrix}

(48)

Proposition 30 We have

\begin{matrix} lim t \to 0 \int_{M} T r (ɛ_{t} (w, z, 0) \circ φ) v o l (g)_{w} = 0 . \end{matrix}

(49)

Proof: The definition 29 of

k_{τ}^{#} (w, z, t)

and the arguments from [6] on page 260 imply that

ɛ_{0} (w, z, t)

is bounded and tends to zero away from the diagonal, as

t

tends to zero. From here we deduce that

lim t \to 0 \int_{M} T r (ɛ_{t} (w, z, 0) \circ φ) v o l (g)_{w} = 0

uniformly in

z

. Proposition 30 is proved.

■

Lemma 31 Let

E

be a holomorphic vector bundle over M, let

φ \in C^{\infty} (

M,Hom(

E, E

))

,

then

lim t \to 0 ϒ_{t} (φ, τ, z)

exists and

\begin{matrix} lim t \to 0 ϒ_{t} (φ, τ, z) = T r (φ |_{E_{z}}) . \end{matrix}

(50)

Proof: We have:

lim t \to 0 ϒ_{t} (φ, 0, z) = lim t \to 0 \int_{M} T r ({(k_{t / 2}^{d} (w, z, τ))}^{*} \circ (k_{t / 2}^{#} (w, z, τ) \circ φ)) v o l (g)_{w} =

\begin{matrix} lim t \to 0 \int_{M} T r (\frac{1}{(4 π t)^{\frac{n}{2}}} exp (- \frac{d_{0}^{2} (w, z)}{4 t}) \circ P_{0} (w, z) \circ φ) v o l (g)_{w} . \end{matrix}

(51)

Using the facts that

\begin{matrix} lim t \to 0 (\frac{1}{(4 π t)^{\frac{n}{2}}} exp (- \frac{d_{τ}^{2} (w, z)}{4 t})) = δ (z - w), \end{matrix}

(52)

\begin{matrix} lim w \to z P_{0} (w, z) = i d \end{matrix}

(53)

and the explicit formula 51

()

for

ϒ_{t} (φ, τ, z)

we obtain that

\begin{matrix} lim t \to 0 ϒ_{t} (φ, τ, z) = \int_{M} T r (δ (z - w) \circ φ) v o l (g)_{ω} = T r (φ |_{E_{z}}) . \end{matrix}

(54)

Lemma 31 is proved.

■

Theorem 32 Let

φ \in C^{\infty} (M, {(Ω_{M}^{0, q})}^{*} \otimes Ω_{M}^{0, q})

then the operator

exp (- t Δ_{h}) \circ φ

for

t > 0

is of trace class and its trace is given by the formula;

\begin{matrix} T r (exp (- t Δ_{h}) \circ φ) = \int_{M} ϒ_{t} (φ, τ, z) v o l (g)_{z} + Φ (t), \end{matrix}

(55)

where the short term asymptotic of

Φ (t)

is given by

\begin{matrix} Φ (t) = \sum_{k = 1}^{N_{0} > 0} \frac{a_{- k}}{t^{k}} + O (t) . \end{matrix}

(56)

Proof: The proof of Theorem 32 is based on the facts that

\begin{matrix} exp (- t Δ_{h}) \circ φ = exp (- \frac{t}{2} Δ_{h}) \circ exp (- \frac{t}{2} Δ_{h} \circ φ), \end{matrix}

(57)

and the operators

exp (- \frac{t}{2} Δ_{h})

and

exp (- \frac{t}{2} Δ_{h} \circ φ)

can be represented by

C^{\infty}

kernels

k_{1} (z, w, t)

and

k_{φ} (z, w, t) .

As we pointed out the operators defined by the kernels

k_{1} (z, w, t)

and

k_{φ} (z, w, t)

are Hilbert-Schmidt operators. Thus since the operator

exp (- t Δ_{h}) \circ φ

is a product of two Hilbert-Schmidt operators it is of trace class. On the other hand the definition of the trace of the operator

exp (- t Δ_{h}) \circ φ

implies that

T r (exp (- t Δ_{h}) \circ φ) = 〈 exp {(- \frac{t}{2} Δ_{h})}^{*}, exp (- \frac{t}{2} Δ_{h}) \circ φ 〉 =

\begin{matrix} \int_{(z, w) \in M \times M} T r ({(k_{1} (z, w, t))}^{*} \circ k_{φ} (z, w, t)) v o l (g_{z, w}) . \end{matrix}

(58)

From the definitions of the function

ϒ_{t} (φ, τ, z)

and the operator

ɛ_{t} (w, z, τ)

we deduce that

T r (exp (- t Δ_{h}) \circ φ) = 〈 (k_{t / 2} (w, z, τ)), k_{t / 2} (w, z, τ) \circ φ 〉 =

\int_{M} (\int_{M} {(k_{t / 2} (w, z, τ))}^{*} \circ (k_{t / 2} (w, z, τ) \circ φ) v o l (g)_{w}) v o l (g)_{z} =

\int_{M} ϒ_{t} (φ, τ, z) v o l (g)_{z} + \int_{M} (\int_{M} T r ({(ɛ_{t / 2} (w, z, τ))}^{*} \circ k_{t / 2}^{#} (w, z, τ) \circ φ) v o l (g)_{ω}) v o l (g)_{z} +

\begin{matrix} \int_{M} (\int_{M} T r (k_{t / 2}^{#} (w, z, τ) \circ ɛ_{t / 2} (w, z, τ) \circ φ) v o l (g)_{ω}) v o l (g)_{z} . \end{matrix}

(59)

Lemma 33 Let

Φ_{1} (t) : = \int_{M} (\int_{M} T r ({(ɛ_{t / 2} (w, z, τ))}^{*} \circ k_{t / 2}^{#} (w, z, τ) \circ φ) v o l (g)_{ω}) v o l (g)_{z}

and

\begin{matrix} Φ_{2} (t) : = \int_{M} (\int_{M} T r (k_{t / 2}^{#} (w, z, τ) \circ ɛ_{t / 2} (w, z, τ) \circ φ) v o l (g)_{ω}) v o l (g)_{z} \end{matrix}

(60)

then we have

\begin{matrix} Φ_{1} (t) = \sum_{k = 1}^{N_{0} > 0} \frac{b_{- k}}{t^{k}} + O (t) and Φ_{2} (t) = \sum_{k = 1}^{N_{0} > 0} \frac{c_{- k}}{t^{k}} + O (t) . \end{matrix}

(61)

Proof: Let

\begin{matrix} k_{t / 2}^{#} (w, z, τ) = \sum_{k = 1}^{N_{0} > 0} \frac{ℬ_{- k} (w, z)}{t^{k}} + ℬ_{0} (w, z) + \sum_{k = 1} ℬ_{k} (w, z) t^{k} \end{matrix}

(62)

be the short term asymptotic expansion of the operator

k_{t / 2}^{#} (w, z, τ) .

We know that

\begin{matrix} lim t \to 0 ɛ_{t} (w, z, τ) = 0 \end{matrix}

(63)

away from the diagonal

Δ \subset

\times

M. Combining 62

()

and 63

()

with the definitions of operators

k_{t / 2}^{#} (w, z, τ) \circ ɛ_{t / 2} (w, z, τ) \circ φ

and

{(ɛ_{t / 2} (w, z, τ))}^{*} \circ k_{t / 2}^{#} (w, z, τ) \circ φ

we obtain that

k_{t / 2}^{#} (w, z, τ) \circ ɛ_{t / 2} (w, z, τ) \circ φ =

\begin{matrix} \sum_{k = 1}^{N_{0} > 0} \frac{ℬ_{- k} (w, z) \circ ɛ_{t / 2} (w, z, τ) \circ φ}{t^{k}} + ℬ_{0} (w, z) \circ ɛ_{t / 2} (w, z, τ) \circ φ + O (t) \end{matrix}

(64)

and

{(ɛ_{t / 2} (w, z, τ))}^{*} \circ k_{t / 2}^{#} (w, z, τ) \circ φ =

\begin{matrix} \sum_{k = 1}^{N_{0} > 0} \frac{{(ɛ_{t / 2} (w, z, τ))}^{*} \circ ℬ_{- k} (w, z)}{t^{k}} + {(ɛ_{t / 2} (w, z, τ))}^{*} \circ ℬ_{0} (w, z) + O (t) . \end{matrix}

(65)

Combining 64

()

, 65

()

with 63

()

we get that

lim t \to 0 ℬ_{0} (w, z) \circ ɛ_{t / 2} (w, z, τ) \circ φ = lim t \to 0 {(ɛ_{t / 2} (w, z, τ))}^{*} \circ ℬ_{0} (w, z) = 0

away from the diagonal. From here we we obtain that

lim t \to \infty \int_{M} T r (ℬ_{0} (w, z) \circ ɛ_{t / 2} (w, z, τ) \circ φ) v o l (g) = 0

and

lim t \to \infty \int_{M} T r ({(ɛ_{t / 2} (w, z, τ))}^{*} \circ ℬ_{0} (w, z)) v o l (g) = 0 .

Lemma 33 is proved.

■

Theorem 32 follows directly from Lemma 33 and 59

()

■

Theorem 34 We have the following expression for

β_{0} (φ)

from 43

()

\begin{matrix} β_{0} (φ) = lim t \to 0 \int_{M} ϒ_{t} (φ, τ, z) v o l (g)_{z} = \int_{M} T r (φ) v o l (g) . \end{matrix}

(66)

Proof: Theorem 34 follows directly from Theorem 32 , Lemma 31 and the definition of

ϒ_{t} (φ, τ, z)

■

5.4 Explicit Formulas

Theorem 35 Let

ℱ^{^{'}} (q, φ \circ \bar{ψ})

be given by the formula 40

()

. Then for

t > 0

and

q \geq 1

the following equality of the traces of the respective operators holds

T r (exp ((- t △_{q - 1}^{^{”}}) \circ {\bar{\partial}}^{- 1} \circ ℱ^{^{'}} (q, φ \circ \bar{ψ}) \circ \bar{\partial})) =

\begin{matrix} T r (exp (- t △_{q}^{^{^{'}}}) \circ (ℱ^{^{'}} (q, φ \circ \bar{ψ}))) . \end{matrix}

(67)

Proof: From Proposition 2.45 on page 96 in [3] it follows directly that the operators

exp (- t Δ_{q - 1}^{”}) \circ {\bar{\partial}}^{- 1} \circ ℱ^{^{'}} (q, φ \circ \bar{ψ}) \circ \bar{\partial}, exp (- t Δ_{q}^{^{'}}) \circ ℱ^{^{'}} (q, φ \circ \bar{ψ})

are of trace class since the operators

exp (- t Δ_{q - 1}^{”})

have smooth kernels for

q \geq 1

. We know from Proposition 2.45 in [3] that we have the following formula:

\begin{matrix} T r (D K) = T r (D A) \end{matrix}

(68)

where

D

is a differential operator and

A

is an operator with a smooth kernel.

By using 68

()

and the fact that the operators

Δ_{q}

and

\bar{\partial}

commute we derive Theorem 35 . Theorem 35 is proved.

■

Remark 36 From Definition 25 of the operator

ℱ^{^{'}} (q, φ_{i} \circ \bar{φ_{j}})

and Remark 26 we know that it can be represented pointwise by a matrix which we will denote by

ℱ^{^{'}} (q, (φ_{i} \circ \bar{φ_{j}}))

. Since

k_{t}^{#} (z, w, 0)

is also a matrix of the same dimension as the operator

ℱ^{^{'}} (q, φ_{i} \circ \bar{φ_{j}})

we get that the operator

k_{t}^{#} (z, w, 0) \circ ℱ^{^{'}} (q, φ_{i} \circ \bar{φ_{j}})

will be represented pointwise by the product of finite dimensional matrices. So the integral

\int_{M} T r ({(k_{t / 2}^{#} (z, w, 0))}^{*} \circ k_{t / 2}^{#} (z, w, 0) \circ ℱ^{^{'}} (q, φ_{i} \circ \bar{φ_{j}})) v o l (g)

makes sense for

t > 0

Theorem 37 Let

T r (k_{t} (w, z, τ) \circ ℱ^{^{'}} (q, (φ_{i} \circ \bar{φ_{j}}))) =

\begin{matrix} \frac{β_{- n} (φ_{i} \circ \bar{φ_{j}})}{t^{n}} + . . . + \frac{β_{- 1} (φ_{i} \circ \bar{φ_{j}})}{t} + β_{0} (φ_{i} \circ \bar{φ_{j}}) + O (t) \end{matrix}

(69)

be the short term asymptotic. Then the following limit

lim t \to 0 \int_{M} (\int_{M} T r ({(k_{t / 2}^{#} (z, w, 0))}^{*} \circ k_{t / 2}^{#} (z, w, 0) \circ ℱ^{^{'}} (q, φ_{i} \circ \bar{φ_{j}}) v o l (g)_{w}) v o l (g)

exists and

β_{0} (φ_{i} \circ \bar{φ_{j}}) =

lim t \to 0 \int_{M} (\int_{M} T r ({(k_{t / 2}^{#} (z, w, 0))}^{*} \circ k_{t / 2}^{#} (z, w, 0) \circ ℱ^{^{'}} (q, φ_{i} \circ \bar{φ_{j}})) v o l (g)_{w}) v o l (g) =

\begin{matrix} \int_{M} T r (ℱ^{^{'}} (q, φ_{i} \circ \bar{φ_{j}})) v o l (g) < \infty . \end{matrix}

(70)

Proof: Formulas 55

()

and 56

()

in Theorem 32 imply that

β_{0} (φ_{i} \circ \bar{φ_{j}}) = lim t \to 0 \int_{M} ϒ_{t} (ℱ^{^{'}} (q, φ_{i} \circ \bar{φ_{j}}), τ, z) v o l (g)_{z}

Theorem 34 imply that

\begin{matrix} β_{0} (φ_{i} \circ \bar{φ_{j}}) = lim t \to 0 \int_{M} ϒ_{t} (ℱ^{^{'}} (q, φ_{i} \circ \bar{φ_{j}}), τ, z) v o l (g)_{z} = T r (ℱ^{^{'}} (q, φ_{i} \circ \bar{φ_{j}})) |_{z} . \end{matrix}

(71)

Formula 71

()

implies formula 70

() .

Theorem 37 is proved.

■

6 The Variational Formulas

6.1 Preliminary Formulas

Lemma 38 The following formulas are true for

1 \leq q \leq n

\begin{matrix} \frac{\partial}{\partial τ^{i}} (\bar{\partial_{τ}}) |_{τ = 0} = - F (q, φ_{i}) \circ \partial \end{matrix}

(72)

and

\begin{matrix} \frac{\bar{\partial}}{\bar{\partial τ^{i}}} (\partial_{τ}) |_{τ = 0} = - F (q, \bar{φ_{i}}) \circ \bar{\partial} . \end{matrix}

(73)

Proof: From the expression of

{\bar{\partial}}_{τ}

given in Definition 6 :

{\bar{\partial}}_{τ} = \frac{\bar{\partial}}{\bar{\partial z^{j}}} - \sum_{m = 1}^{N} (\sum_{k = 1}^{n} (φ_{m})_{\bar{j}}^{k} \frac{\partial}{\partial z^{k}}) τ^{m} + O (τ^{2})),

we conclude that

\begin{matrix} \frac{\partial}{\partial τ^{i}} (\bar{\partial_{τ}}) |_{τ = 0} = - \sum_{k = 1}^{N} (φ_{i})_{\bar{j}}^{k} \frac{\partial}{\partial z^{k}} . \end{matrix}

(74)

Formula 73

()

is proved in the same way as formula 74

()

. Lemma 38 follows directly from Definition 25 of the linear operators F

^{^{'}} (q, φ)

and F

^{”} (q, φ) .

■

Corollary 39 The following formulas are true for

1 \leq q \leq n

(\frac{\partial}{\partial τ^{i}} (△_{τ, q}^{^{”}}) |_{Im {\bar{\partial_{τ}}}^{*}}) |_{τ = 0} = - Δ_{0, q}^{”} \circ {\bar{\partial_{0}}}^{- 1} \circ F (q, φ_{i}) \circ \partial_{0},

and

\begin{matrix} (\frac{\bar{\partial}}{\bar{\partial τ^{j}}} (△_{τ, q}^{^{”}}) |_{Im {\bar{\partial_{τ}}}^{*}}) |_{τ = 0} = - △_{0, q}^{^{”}} \circ \partial_{0}^{- 1} \circ F^{^{'}} (q + 1, \bar{φ_{i}}) \circ \bar{\partial_{0}} . \end{matrix}

(75)

Proof: From the standard facts of Kähler geometry we obtain that on

Im {\bar{\partial}}^{*}

Ω_{M}^{0, q}

we have

\begin{matrix} △_{τ, q}^{^{^{”}}} |_{Im {\bar{\partial}}^{*}} = Λ_{τ} \circ \partial_{τ} \circ \bar{\partial_{τ}} = Λ_{τ} \circ \bar{\partial_{τ}} \circ \partial_{τ} . \end{matrix}

(76)

We know from 72

()

and 73

()

that

\begin{matrix} \frac{\partial}{\partial τ^{j}} (\bar{\partial_{τ}}) |_{τ = 0} = - F^{^{'}} (q + 1, φ_{j}) \circ \partial_{0}, \frac{\partial}{\partial τ^{j}} (Λ_{τ}) = \frac{\partial}{\partial τ^{j}} (\partial_{τ}) = 0 . \end{matrix}

(77)

Combining 77

()

, 76

()

we obtain:

\frac{\partial}{\partial τ^{j}} (△_{τ, q}^{^{”}}) |_{τ = 0} = \frac{\partial}{\partial τ^{j}} (Λ_{τ} \circ \bar{\partial_{τ}} \circ \bar{\partial_{τ}}) |_{τ = 0} =

\begin{matrix} ({\bar{\partial_{τ}}}^{*} \circ \frac{\partial}{\partial τ^{j}} (\bar{\partial_{τ}})) |_{τ = 0} = - {\bar{\partial_{0}}}^{*} \circ F^{^{'}} (q + 1, φ_{j}) \circ \partial_{0} . \end{matrix}

(78)

Thus on

Im {\bar{\partial}}^{*}

we have

\begin{matrix} \partial_{τ}^{*} = △_{τ, q}^{^{”}} \circ {\bar{\partial_{τ}}}^{- 1} . \end{matrix}

(79)

Substituting 79

()

in 78

()

we obtain the first formula in 75

()

. In the same manner we obtain the second formula in 75

()

. Corollary 39 is proved.

■

6.2 The Computation of the Antiholomorphic Derivative of $ζ_{τ, q - 1}^{”} (s)$

First we will compute the antiholomorphic derivative of

ζ_{τ, q}^{^{”}} (s) .

Theorem 40 The following formula is true for

t > 0

\frac{\bar{\partial}}{\bar{\partial τ^{i}}} (ζ_{q, τ}^{^{”}} (s)) |_{τ = 0} = \frac{1}{Γ (s)} \int_{0}^{\infty} T r (exp (- t (△_{0, q}^{^{”}}) \circ △_{0, q}^{^{”}} \circ \partial_{0}^{- 1} \circ ℱ^{^{'}} (q + 1, \bar{φ_{i}}) \circ \bar{\partial_{0}}) t^{s} d t .

Proof: For the proof of Theorem 40 we will need the following Lemma:

Lemma 41 The following formula is true for

t > 0

and

0 < q < n :

\frac{\bar{\partial}}{\bar{\partial τ^{i}}} (T r (exp (- t △_{τ, q}^{^{”}})) |_{τ = 0} =

\begin{matrix} t T r (exp (- t △_{0, q}^{^{”}}) \circ Δ_{τ, q}^{^{”}} \circ \partial^{- 1} \circ ℱ^{^{'}} (q + 1, \bar{φ_{i}}) \circ \bar{\partial}) |_{τ = 0} . \end{matrix}

(80)

Proof: Direct computations based on Proposition 9.38. on page 304 of the book [6] show that:

\begin{matrix} \frac{\bar{\partial}}{\bar{\partial τ^{i}}} (T r (exp (- t △_{τ, q}^{^{”}})) |_{τ = 0} = - t (exp (- t △_{τ, q}^{^{”}}) \circ \frac{\bar{\partial}}{\bar{\partial τ^{i}}} (△_{τ, q}^{”})) |_{τ = 0} . \end{matrix}

(81)

See also [3] page 98 Theorem 2.48. Formulas 72

()

and 73

()

in Lemma 38 imply that

\begin{matrix} \frac{\bar{\partial}}{\bar{\partial τ^{i}}} (\partial_{τ}) = - F^{^{'}} (q + 1, \bar{φ_{i} (τ)}) \circ \bar{\partial} \end{matrix}

(82)

and on

Im {\bar{\partial}}^{*}

we have

\begin{matrix} \frac{\bar{\partial}}{\bar{\partial τ^{i}}} (\partial_{τ}^{*}) = \frac{\bar{\partial}}{\bar{\partial τ^{i}}} (Λ \circ \bar{\partial_{τ}}) = (\frac{\bar{\partial}}{\bar{\partial τ^{i}}} Λ) \circ \bar{\partial_{τ}} + Λ \circ \frac{\bar{\partial}}{\bar{\partial τ^{i}}} (\bar{\partial_{τ}}) = 0 . \end{matrix}

(83)

The last equality follows from Corollary. 38 and 14 . On Kähler manifolds we know that

\partial^{*} \circ \partial + \partial \circ \partial^{*} = {\bar{\partial}}^{*} \circ \bar{\partial} + \bar{\partial} \circ {\bar{\partial}}^{*} .

So we deduce that

Δ_{τ, q}^{”} = (\partial_{τ}^{*} \circ \partial_{τ} + \partial_{τ} \circ \partial_{τ}^{*}) |_{Im {\bar{\partial_{τ}}}^{*}} = \partial_{τ}^{*} \circ \partial_{τ} |_{Im {\bar{\partial_{τ}}}^{*}} .

Thus from formulas 82

()

and 83

()

it follows

:

\begin{matrix} \frac{\bar{\partial}}{\bar{\partial τ^{i}}} (△_{τ, q}^{^{”}}) = (\partial_{τ}^{*} \circ \frac{\bar{\partial}}{\bar{\partial τ^{i}}} (\partial_{τ})) = - \partial_{τ}^{*} \circ F^{^{'}} (q + 1, \bar{φ_{i} (τ)}) \circ \bar{\partial} . \end{matrix}

(84)

By substituting in 81

()

the expression from 84

()

we obtain:

\begin{matrix} \frac{\bar{\partial}}{\bar{\partial τ^{i}}} (T r (exp (- t △_{τ, q}^{^{”}})) |_{τ = 0} = t T r (exp (- t △_{q}^{^{”}}) \circ \partial_{0}^{*} \circ F^{^{'}} (q + 1, \bar{φ_{i}}) \circ \bar{\partial_{0}}) . \end{matrix}

(85)

The operator

\partial_{τ}^{*}

is well defined on the space of

C^{\infty}

(0,q) forms on M

_{τ} .

So the following formula is true on

Im {\bar{\partial}}_{τ}^{*}

\begin{matrix} \partial_{τ}^{*} = (△_{τ, q}^{^{”}}) \circ {(\partial_{τ})}^{- 1} . \end{matrix}

(86)

Substituting the expression for

\partial_{τ}^{*}

in formula 86

()

in 85

()

, we deduce formula 80

() .

Lemma 41 is proved.

■

The end of the proof of Theorem 40 : The definition of the zeta function implies that

\frac{\bar{\partial}}{\bar{\partial τ^{i}}} (ζ_{Δ_{τ, q}^{”}} (s)) |_{τ = 0} = \frac{\bar{\partial}}{\bar{\partial τ^{i}}} (\frac{1}{Γ (s)} \int_{0}^{\infty} T r (exp (- t (△_{τ, q}^{^{^{”}}})) t^{s - 1} d t) |_{τ = 0} =

\begin{matrix} \frac{1}{Γ (s)} \int_{0}^{\infty} (\frac{\bar{\partial}}{\bar{\partial τ^{i}}} (T r exp (- t (△_{τ, q}^{^{^{”}}}))) |_{τ = 0} t^{s - 1} d t . \end{matrix}

(87)

Substituting in 87

()

the expression for

\frac{\bar{\partial}}{\bar{\partial τ^{i}}} (T r (exp (- t (△_{τ, q}^{”})))

given by 80

()

we obtain:

\frac{\bar{\partial}}{\bar{\partial τ^{i}}} (ζ_{Δ_{τ, q}^{”}} (s)) |_{τ = 0} =

\begin{matrix} \frac{1}{Γ (s)} \int_{0}^{\infty} T r ((exp (- t (△_{0, q}^{”})) \circ (△_{0, q}^{”}) \circ \partial^{- 1} \circ F^{^{'}} (q + 1, \bar{φ_{i}}) \circ \bar{\partial}) t^{s} d t . \end{matrix}

(88)

Theorem 40 is proved.

■

6.3 The Computation of the Hessian of $ζ_{τ, q}^{^{”}} (s)$

Theorem 42 The following formula holds:

\begin{matrix} \frac{\partial^{2}}{\partial τ^{j} \bar{\partial τ^{i}}} (ζ_{τ, q}^{^{”}} (s)) |_{τ = 0} = \frac{(1 - s) s}{Γ (s)} \int_{0}^{\infty} T r ((exp (- t (△_{0, q + 1}^{^{^{^{'}}}})) \circ ℱ^{^{'}} (q + 1, φ_{j} \circ \bar{φ_{i}})) t^{s - 1} d t . \end{matrix}

(89)

Proof: The facts that the operators

\frac{\partial}{\partial τ^{i}} (\bar{\partial_{τ}}) = (- φ_{i} + O (τ)) \circ \partial_{0}

depend holomorphically on

τ

and the operator

\partial_{τ}^{- 1}

depends antiholomorphically imply that the operator

\partial_{τ}^{- 1} \circ F^{^{'}} (q + 1, \bar{φ_{i}}) \circ \bar{\partial_{0}}

depends antiholomorphically on the coordinates

τ = (τ^{1}, . . ., τ^{N}) .

By using the explicit formula 88

()

for the antiholomorphic derivative of

ζ_{τ, q}^{”} (s)

and that

\frac{\partial}{\partial τ^{j}} (\partial_{τ}^{- 1} \circ F^{^{'}} (q + 1, \bar{φ_{i}}) \circ \bar{\partial_{0}}) = 0

we derive

\frac{\partial^{2}}{\bar{\partial τ^{j}} \partial τ^{i}} (ζ_{τ, q}^{”} (s)) =

\frac{1}{Γ (s)} \int_{0}^{\infty} T r ((\frac{\partial}{\partial τ^{j}} exp (- t (△_{τ, q}^{^{”}})) \circ (△_{τ, q}^{^{”}}) \circ \partial_{τ}^{- 1} \circ F^{^{'}} (q + 1, \bar{φ_{i}}) \circ \bar{\partial_{0}}) t^{s} d t +

\begin{matrix} \frac{1}{Γ (s)} \int_{0}^{\infty} T r ((exp (- t (△_{τ, q}^{”})) \circ \frac{\partial}{\partial τ^{j}} (△_{τ, q}^{^{”}}) \circ \partial_{0}^{- 1} \circ F^{^{'}} (q + 1, \bar{φ_{i}}) \circ \bar{\partial_{0}}) t^{s} d t . \end{matrix}

(90)

Lemma 43 We have the following expression:

\frac{1}{Γ (s)} \int_{0}^{\infty} T r ((\frac{\partial}{\partial τ^{j}} (exp (- t (△_{τ, q}^{^{”}}))) \circ (△_{τ, q}^{^{”}}) \circ \partial_{τ}^{- 1} \circ F^{^{'}} (q + 1, \bar{φ_{i}}) \circ \bar{\partial_{0}}) t^{s} d t =

\begin{matrix} \frac{- s}{Γ (s)} \int_{0}^{\infty} T r ((exp (- t △_{0, q}^{^{”}}) \circ \frac{\partial}{\partial τ^{j}} (△_{τ, q}^{^{”}}) \circ \partial_{τ}^{- 1} \circ F^{^{'}} (q + 1, \bar{φ_{i}}) \circ \bar{\partial_{0}}) t^{s} d t . \end{matrix}

(91)

Proof: Direct computations show that

\frac{1}{Γ (s)} (\int_{0}^{\infty} T r ((\frac{\partial}{\partial τ^{j}} (exp (- t (△_{τ, q}^{^{”}}))) \circ (△_{τ, q}^{^{”}}) \circ \partial_{τ}^{- 1} \circ F^{^{'}} (q + 1, \bar{φ_{i}}) \circ \bar{\partial_{0}}) t^{s} d t) |_{τ = 0} =

\begin{matrix} \frac{- 1}{Γ (s)} \int_{0}^{\infty} (\frac{d}{d t} T r ((\frac{\partial}{\partial τ^{j}} (exp (- t (△_{τ, q}^{^{”}}))) |_{τ = 0} \circ \partial_{0}^{- 1} \circ F^{^{'}} (q + 1, \bar{φ_{i}}) \circ \bar{\partial_{0}})) t^{s} d t \end{matrix}

(92)

By integrating by parts the right hand side of formula 92

()

we deduce that:

\frac{- 1}{Γ (s)} \int_{0}^{\infty} (\frac{d}{d t} T r ((\frac{\partial}{\partial τ^{j}} (exp (- t (△_{τ, q}^{^{”}}))) |_{τ = 0} \circ \partial_{0}^{- 1} \circ F^{^{'}} (q + 1, \bar{φ_{i}}) \circ \bar{\partial_{0}})) t^{s} d t =

\begin{matrix} \frac{s}{Γ (s)} \int_{0}^{\infty} T r ((\frac{\partial}{\partial τ^{j}} (exp (- t (△_{τ, q}^{^{”}}))) |_{τ = 0} \circ \partial_{0}^{- 1} \circ F^{^{'}} (q + 1, \bar{φ_{i}}) \circ \bar{\partial_{0}}) t^{s - 1} d t . \end{matrix}

(93)

Direct computations of the right hand side of 93

()

show that:

\frac{s}{Γ (s)} \int_{0}^{\infty} T r ((\frac{\partial}{\partial τ^{j}} (exp (- t △_{τ, q}^{^{”}})) |_{τ = 0} \circ \partial_{0}^{- 1} \circ F^{^{'}} (q + 1, \bar{φ_{i}}) \circ \bar{\partial_{0}}) t^{s - 1} d t =

\frac{- s}{Γ (s)} \int_{0}^{\infty} T r ((exp (- t △_{τ, q}^{^{”}}) \circ \frac{\partial}{\partial τ^{j}} (△_{τ, q}^{^{”}}) |_{τ = 0} \circ \partial_{0}^{- 1} \circ F^{^{'}} (q + 1, \bar{φ_{i}}) \circ \bar{\partial_{0}}) t^{s} d t .

Formula 91

()

is proved.

■

Substituting in 90

()

the expression from 91

()

\frac{1}{Γ (s)} \int_{0}^{\infty} T r ((\frac{\partial}{\partial τ^{j}} (exp (- t (△_{τ, q}^{^{”}}))) \circ (△_{τ, q}^{^{”}}) \circ \partial_{τ}^{- 1} \circ F^{^{'}} (q + 1, \bar{φ_{i}}) \circ \bar{\partial_{0}}) t^{s} d t

we get:

\frac{\partial^{2}}{\bar{\partial τ^{j}} \partial τ^{i}} (ζ_{q - 1, τ}^{”} (s)) |_{τ = 0} =

\begin{matrix} \frac{1 - s}{Γ (s)} \int_{0}^{\infty} T r ((exp (- t (△_{τ, q}^{^{”}})) \circ \frac{\partial}{\partial τ^{j}} (△_{τ, q}^{^{”}}) |_{τ = 0} \circ \partial_{0}^{- 1} \circ F^{^{'}} (q + 1, \bar{φ_{i}}) \circ \bar{\partial_{0}}) t^{s} d t . \end{matrix}

(94)

Lemma 44

\int_{0}^{\infty} T r ((exp (- t (△_{τ, q}^{^{”}})) \circ \frac{\partial}{\partial τ^{j}} (△_{τ, q}^{^{”}}) |_{τ = 0} \circ \partial_{0}^{- 1} \circ F^{^{'}} (q + 1, \bar{φ_{i}}) \circ \bar{\partial_{0}}) t^{s} d t =

\begin{matrix} s \int_{0}^{\infty} T r ((exp (- t (△_{τ, q}^{^{”}})) |_{τ = 0} \circ {(\bar{\partial_{0}})}^{- 1} \circ ℱ^{^{'}} (q + 1, φ_{j} \circ \bar{φ_{i}}) \circ \bar{\partial_{0}}) t^{s - 1} d t . \end{matrix}

(95)

Proof: Substituting the expression of 75

()

for

\frac{\partial}{\partial τ^{j}} (△_{τ, q}^{^{”}}) |_{τ = 0} = - △_{0, q}^{^{”}} \circ {(\bar{\partial_{0}})}^{- 1} \circ F^{^{'}} (q + 1, φ_{j}) \circ \bar{\partial_{0}}

in the expression of

\int_{0}^{\infty} T r ((exp (- t (△_{τ, q}^{^{”}})) \circ \frac{\partial}{\partial τ^{j}} (△_{τ, q}^{^{”}}) |_{τ = 0} \circ \partial_{0}^{- 1} \circ F^{^{'}} (q + 1, \bar{φ_{i}}) \circ \bar{\partial_{0}}) t^{s} d t

we obtain:

\int_{0}^{\infty} T r ((exp (- t (△_{τ, q}^{^{”}})) \circ \frac{\partial}{\partial τ^{j}} (△_{τ, q}^{^{”}}) |_{τ = 0} \circ \partial_{0}^{- 1} \circ F^{^{'}} (q + 1, \bar{φ_{i}}) \circ \bar{\partial_{0}}) t^{s} d t =

\begin{matrix} - \int_{0}^{\infty} T r ((exp (- t (△_{0, q}^{^{”}})) \circ △_{0, q}^{^{”}} \circ \partial_{0}^{- 1} \circ F^{^{'}} (q + 1, \bar{φ_{i}}) \circ F^{^{'}} (q + 1, \bar{φ_{i}}) \circ \bar{\partial_{0}}) t^{s} d t . \end{matrix}

(96)

Simple observations shows that

\int_{0}^{\infty} T r ((exp (- t (△_{0, q}^{^{”}})) \circ △_{0, q}^{^{”}} \circ \partial_{0}^{- 1} \circ F^{^{'}} (q + 1, \bar{φ_{i}}) \circ F^{^{'}} (q + 1, \bar{φ_{i}}) \circ \bar{\partial_{0}}) t^{s} d t =

\begin{matrix} - \int_{0}^{\infty} T r (\frac{d}{d t} exp (- t (△_{0, q}^{^{”}}) \partial_{0}^{- 1} \circ F^{^{'}} (q + 1, \bar{φ_{i}}) \circ F^{^{'}} (q + 1, \bar{φ_{i}}) \circ \bar{\partial_{0}}) t^{s} d t . \end{matrix}

(97)

By integrating by parts the right hand side of 97

()

we obtain:

- \int_{0}^{\infty} T r (\frac{d}{d t} exp (- t (△_{0, q}^{^{”}}) \partial_{0}^{- 1} \circ F^{^{'}} (q + 1, \bar{φ_{i}}) \circ F^{^{'}} (q + 1, \bar{φ_{i}}) \circ \bar{\partial_{0}}) t^{s} d t =

s \int_{0}^{\infty} T r (exp (- t (△_{0, q}^{^{”}}) \partial_{0}^{- 1} \circ F^{^{'}} (q + 1, \bar{φ_{i}}) \circ F^{^{'}} (q + 1, \bar{φ_{i}}) \circ \bar{\partial_{0}}) t^{s - 1} d t .

Thus we derive formula 95

(),

i.e.

\int_{0}^{\infty} T r ((exp (- t (△_{τ, q}^{^{”}})) \circ \frac{\partial}{\partial τ^{j}} (△_{τ, q}^{^{”}}) |_{τ = 0} \circ \partial_{0}^{- 1} \circ F^{^{'}} (q + 1, \bar{φ_{i}}) \circ \bar{\partial_{0}}) t^{s} d t =

s \int_{0}^{\infty} T r ((exp (- t (△_{τ, q}^{^{”}})) |_{τ = 0} \circ {(\bar{\partial_{0}})}^{- 1} \circ ℱ^{^{'}} (q + 1, φ_{j} \circ \bar{φ_{i}}) \circ \bar{\partial_{0}}) t^{s - 1} d t .

Lemma 44 is proved.

■

Substituting the expression of 95

()

in the expression 94

()

we get the following equality:

\frac{\partial^{2}}{\bar{\partial τ^{j}} \partial τ^{i}} (ζ_{q, τ}^{”} (s)) |_{τ = 0} =

\begin{matrix} \frac{(1 - s) s}{Γ (s)} \int_{0}^{\infty} T r ((exp (- t (△_{0, q}^{^{”}})) \circ {\bar{\partial_{0}}}^{- 1} \circ ℱ^{^{'}} (q + 1, φ_{j} \circ \bar{φ_{i}}) \circ \bar{\partial_{0}}) t^{s - 1} d t . \end{matrix}

(98)

Applying Theorem 35 we deduce Theorem 42 .

■

6.4 The Computations of the Hessian of $log det Δ_{τ, q}$

Theorem 45 The following formula is true:

\frac{\partial^{2}}{\bar{\partial τ^{j}} \partial τ^{i}} (log det Δ_{τ, q}^{”}) |_{τ = 0} =

- lim t \to 0 \int_{M} (\int_{M} T r (4 π t)^{- \frac{n}{2}} exp (- \frac{d_{0}^{2} (w, z)}{4 t}) \circ P \circ ℱ^{^{'}} (q + 1, φ_{i} \circ \bar{φ_{j}})) v o l (g)_{w}) v o l (g)_{z} =

\begin{matrix} - \int_{M} T r (ℱ^{^{'}} (q + 1, (φ_{i} \circ \bar{φ_{j}})))) v o l (g) . \end{matrix}

(99)

Proof:

ζ_{τ, q}^{”} (s)

is obtained from the meromorphic continuation of

\frac{1}{Γ (s)} \int_{0}^{\infty} T r (exp (- t Δ_{τ, q}^{”})) t^{s - 1} d t

and it is a meromorphic function on

C

well defined at

0 .

Thus we get that

ζ_{τ, q}^{”} (s) = μ_{0} (τ) + μ_{1} (τ) s + O (s^{2})

and so

\frac{\partial^{2}}{\bar{\partial τ^{j}} \partial τ^{i}} (ζ_{τ, q}^{”} (s)) |_{τ = 0} = \frac{\partial^{2}}{\bar{\partial τ^{j}} \partial τ^{i}} μ_{0} (τ) |_{τ = 0} + (\frac{\partial^{2}}{\bar{\partial τ^{j}} \partial τ^{i}} μ_{1} (τ)) |_{τ = 0} s + O (s^{2}) =

\begin{matrix} α_{0} + α_{1} s + O (s^{2}) . \end{matrix}

(100)

Thus from the definition of the regularized determinant

log det (△_{τ, q}^{”}) = (\frac{d}{d s} (- ζ_{τ, q}^{”} (s))) |_{s = 0}

we see that

\begin{matrix} \frac{\partial^{2}}{\bar{\partial τ^{j}} \partial τ^{i}} (log det Δ_{τ, q}^{”}) |_{τ = 0} = \frac{d}{d s} (\frac{\partial^{2}}{\bar{\partial τ^{j}} \partial τ^{i}} ((- ζ_{τ, q}^{”} (s))) |_{τ = 0}) |_{s = 0} = - α_{1} . \end{matrix}

(101)

Combining formula 89

()

\frac{\partial^{2}}{\partial τ^{j} \bar{\partial τ^{i}}} (ζ_{τ, q}^{^{”}} (s)) |_{τ = 0} = \frac{s}{Γ (s)} \int_{0}^{\infty} T r ((exp (- t (△_{0, q + 1}^{^{'}})) \circ ℱ^{^{'}} (q + 1, φ_{j} \circ \bar{φ_{i}})) t^{s - 1} d t +

- \frac{s^{2}}{Γ (s)} \int_{0}^{\infty} T r ((exp (- t (△_{0, q + 1}^{^{^{^{'}}}})) \circ ℱ^{^{'}} (q + 1, φ_{j} \circ \bar{φ_{i}})) t^{s - 1} d t

with the short term expansion:

\begin{matrix} T r ((exp (- t (△_{0, q + 1}^{^{^{'}}})) \circ ℱ^{^{'}} (q + 1, φ_{j} \circ \bar{φ_{i}})) = \sum_{k = - n}^{1} \frac{ν_{k}}{t^{k}} + ν_{0} + ψ (t) \end{matrix}

(102)

where

ψ (t) = T r ((exp (- t (△_{0, q + 1}^{^{^{'}}})) \circ ℱ^{^{'}} (q + 1, φ_{j} \circ \bar{φ_{i}})) - \sum_{k = - n}^{1} \frac{ν_{k}}{t^{k}} + ν_{0}

we obtain:

\frac{\partial^{2}}{\partial τ^{j} \bar{\partial τ^{i}}} (ζ_{τ, q}^{^{”}} (s)) |_{τ = 0} = \frac{s}{Γ (s)} (\int_{0}^{1} (\sum_{k = - n}^{1} \frac{ν_{k}}{t^{k}}) t^{s - 1} d t + ν_{0} \int_{0}^{1} t^{s - 1} d t + \int_{0}^{1} ψ (t) t^{s - 1} d t) +

\frac{s}{Γ (s)} (\int_{1}^{\infty} T r ((exp (- t (△_{0, q + 1}^{^{^{'}}})) \circ ℱ^{^{'}} (q + 1, φ_{j} \circ \bar{φ_{i}})) t^{s - 1} d t) -

- \frac{s^{2}}{Γ (s)} (\int_{0}^{1} (\sum_{k = - n}^{1} \frac{ν_{k}}{t^{k}}) t^{s - 1} d t + ν_{0} \int_{0}^{1} t^{s - 1} d t + \int_{0}^{1} ψ (t) t^{s - 1} d t) -

\begin{matrix} - \frac{s^{2}}{Γ (s)} (\int_{1}^{\infty} T r ((exp (- t (△_{0, q + 1}^{^{^{'}}})) \circ ℱ^{^{'}} (q + 1, φ_{j} \circ \bar{φ_{i}})) t^{s - 1} d t) . \end{matrix}

(103)

By using formula 103

()

we will prove the following Lemma:

Lemma 46 We have the following formula:

\frac{\partial^{2}}{\bar{\partial τ^{j}} \partial τ^{i}} (log det Δ_{τ, q}^{”}) |_{τ = 0} = - \int_{M} T r ℱ^{^{'}} (q + 1, φ_{i} \circ \bar{φ_{j}}) v o l (g) = - α_{1} .

Proof: Lemma 9.34 on page 300 of [3] or direct computations show that for

| s | < ɛ

we have the following identity:

\frac{1}{Γ (s)} \int_{0}^{\infty} T r ((exp (- t (△_{0, q}^{^{'}})) \circ ℱ^{^{'}} (q + 1, φ_{j} \circ \bar{φ_{i}})) t^{s - 1} d t =

\frac{1}{Γ (s)} (\int_{0}^{1} (\sum_{k = - n}^{1} \frac{ν_{k}}{t^{k}}) t^{s - 1} d t + ν_{0} \int_{0}^{1} t^{s - 1} d t + \int_{0}^{1} ψ (t) t^{s - 1} d t) +

\begin{matrix} + \frac{1}{Γ (s)} \int_{1}^{\infty} T r ((exp (- t (△_{0, q}^{^{^{'}}})) \circ ℱ^{^{'}} (q + 1, φ_{j} \circ \bar{φ_{i}})) t^{s - 1} d t = \frac{ν_{0}}{s} + κ + O (s) . \end{matrix}

(104)

Combining the expression in 104

()

with the following standard fact

\frac{s}{Γ (s)} = s^{2} + O (s^{3})

we obtain from formulas 103

()

and 104

()

that for

| s | < ɛ

\begin{matrix} \frac{\partial^{2}}{\partial τ^{j} \bar{\partial τ^{i}}} (ζ_{τ, q}^{^{”}} (s)) |_{τ = 0} = ν_{0} s + O (s^{2}) . \end{matrix}

(105)

Thus according to 101

()

and 105

()

\begin{matrix} \frac{d}{d s} (\frac{\partial^{2}}{\partial τ^{j} \bar{\partial τ^{i}}} (ζ_{τ, q}^{^{”}} (s)) |_{τ = 0}) |_{s = 0} = ν_{0} = α_{1} . \end{matrix}

(106)

Applying Theorem 34 to formula 106

()

we deduce that

α_{1} = ν_{0} = \int_{M} T r (ℱ^{^{'}} (q + 1, φ_{i} \circ \bar{φ_{j}})) v o l (g) .

Lemma 46 is proved.

■

Lemma 46 implies directly Theorem 45 .

■

6.5 Some Applications of the Variational Formulas

Theorem 47 The following identity holds

d d^{c} (log det Δ_{τ, 1}) = d d^{c} (log det Δ_{τ, 1}^{^{'}} det Δ_{τ, 1}^{”}) = - Im W . P .

Proof: The proof of Theorem 47 is based on the following formulas which holds for Kähler manifolds:

\begin{matrix} Δ_{\partial} = Δ \bar{_{\partial}}, \partial^{*} = - * \bar{\partial} * and {\bar{\partial}}^{*} = - * \partial *, \end{matrix}

(107)

where

*

is the Hodge star operator. See [20] page 95. On CY manifold we have the following duality:

* : Ω_{M}^{0, q} ≊ Ω_{M}^{0, n - q}

induced by the Hodge star operator

*

of a CY metric and the holomorphic

n

form. Using this duality direct check shows that on CY manifolds we have

\begin{matrix} * (Im \bar{\partial_{q}}) = Im ({\bar{\partial_{n - q}}}^{*}) and * (Im {\bar{\partial_{q}}}^{*}) = Im (\bar{\partial_{n - q}}) \end{matrix}

(108)

Formulas 107

()

and 108

()

imply that we have

det Δ_{\bar{\partial}, q}^{^{'}} = det Δ_{\partial, q}^{^{'}} = det Δ_{\partial, n - q}^{”^{'}} = det Δ_{\bar{\partial}, n - q}^{”^{'}} .

Lemma 48 We have the following relations between the operators on a CY manifold

T r (exp (- t Δ_{τ, 1}^{^{'}}) \circ ℱ^{^{'}} (1, φ_{i} \circ \bar{φ_{j}})) = T r (exp (- t Δ_{τ, 1}^{”}) \circ ℱ^{^{”}} (n - 1, φ_{i} \circ \bar{φ_{j}}))

and

\begin{matrix} T r (exp (- t Δ_{τ, 1}^{^{”}}) \circ ℱ^{”} (1, φ_{i} \circ \bar{φ_{j}})) = T r (exp (- t Δ_{τ, 1}^{^{'}}) \circ ℱ^{^{'}} (n - 1, φ_{i} \circ \bar{φ_{j}})) \end{matrix}

(109)

by identifying the Hilbert spaces

Im \bar{\partial} \subset L^{2} (

, Ω_{M}^{0, 1})

and

Im {\bar{\partial}}^{*} \subset L^{2} (M, Ω_{M}^{0, 1})

with

Im \partial^{*} \subset L^{2} (M, Ω_{M}^{0, n - 1})

and

Im \partial \subset L^{2} (M, Ω_{M}^{0, n - 1})

respectively by using 108

() .

Proof: We will need the following Propositions to prove Lemma 48 :

Proposition 49 Let Let

{ω_{i}}

be an orthonormal basis at

Ω_{x}^{1, 0} .

Let

{\bar{ω}}_{n} = \bar{ω_{1}} \land . . . \land \bar{ω_{n}}

be the antiholomorphic volume form. Then for

α \neq β

we have

\begin{matrix} (φ_{i} \circ \bar{φ_{j}} \land i d_{n - 2}) ({\bar{ω}}_{n} ⌟ \frac{\bar{\partial}}{\bar{\partial z^{α}}}) = \sum_{β} {(φ_{i} \circ \bar{φ_{j}})}_{\bar{β}}^{\bar{α}} ({\bar{ω}}_{n} ⌟ \frac{\bar{\partial}}{\bar{\partial z^{α}}} \land \frac{\bar{\partial}}{\bar{\partial z^{β}}}) \land \bar{d z^{α}} . \end{matrix}

(110)

Proof: Formula 110 follows directly from the definition of the linear operator

(φ_{i} \circ \bar{φ_{j}} \land i d_{n - 2}) . ■

Proposition 50 Let us define

{(φ_{i} \circ \bar{φ_{j}} \land i d_{n - 2})}^{*} : Ω_{M}^{0, n - 1} \to Ω_{M}^{0, n - 1}

as follows:

\begin{matrix} {(φ_{i} \circ \bar{φ_{j}} \land i d_{n - 2})}^{*} (* \bar{ω_{k}}) = * ((φ_{i} \circ \bar{φ_{j}}) (\bar{ω_{k}})) \end{matrix}

(111)

Let us denote by

M (\bar{φ_{j}} \circ φ_{i})

and

M ({(φ_{i} \circ \bar{φ_{j}} \land i d_{n - 2})}^{*})

the matrices of the operators

\bar{φ_{j}} \circ φ_{i}

and

{(φ_{i} \circ \bar{φ_{j}} \land i d_{n - 2})}^{*}

in the orthonormal bases with respect to the CY metric. Then fibrewise we have the equality

\begin{matrix} M (\bar{φ_{j}} \circ φ_{i}) = M ({(φ_{i} \circ \bar{φ_{j}} \land i d_{n - 2})}^{*}) . \end{matrix}

(112)

Proof: We need to compute the matrix of the operator

{(φ_{i} \circ \bar{φ_{j}} \land i d_{n - 2})}^{*}

in the orthonormal basis

\bar{ω_{i_{1}}} \land . . . \land \bar{ω_{i_{n - k}}}

and compare it with the matrix of the operator

\bar{φ_{j}} \circ φ_{i}

of the bundle

Ω_{M}^{1, 0}

written in the orthonormal basis

{ω_{i}} .

Let

{ω_{i}}

be an orthonormal basis at

Ω_{x}^{1, 0} .

According to Lemma 11 the operators

φ_{i} : Ω_{x}^{1, 0} \to Ω_{x}^{0, 1}

in the orthonormal basis

{ω_{i}}

and

{\bar{ω_{i}}}

are given by symmetric matrices. From here 112

()

follows directly. Indeed from the relations of the elements of the matrices of the operators

φ_{i}

in an orthonormal basis

φ_{i, \bar{β}}^{α} = φ_{i, \bar{α}}^{β}

we obtain

\begin{matrix} {(φ_{i} \circ \bar{φ_{j}})}_{\bar{α}}^{\bar{β}} = \sum_{μ = 1}^{n} φ_{i, \bar{α}}^{μ} \bar{φ_{j, \bar{μ}}^{β}} = \sum_{μ = 1}^{n} φ_{i, \bar{μ}}^{α} \bar{φ_{j, \bar{β}}^{μ}} = \sum_{μ = 1}^{n} \bar{φ_{j, \bar{β}}^{μ}} φ_{i, \bar{μ}}^{α} = {(\bar{φ_{j}} \circ φ_{i})}_{β}^{α} . \end{matrix}

(113)

From the definition of the operator

{(φ_{i} \circ \bar{φ_{j}} \land i d_{n - 2})}^{*}

given by 111

()

we get:

{(φ_{i} \circ \bar{φ_{j}} \land i d_{n - 2})}^{*} (* \bar{ω_{k}}) = * ((φ_{i} \circ \bar{φ_{j}}) (\bar{ω_{k}})) = * (\sum_{k, l = 1}^{n} {(φ_{i} \circ \bar{φ_{j}})}_{k}^{l} (\bar{ω_{l}})) =

\begin{matrix} \sum_{k, l = 1}^{n} {(φ_{i} \circ \bar{φ_{j}})}_{k}^{l} (* (\bar{ω_{l}})) = \sum_{k, l = 1}^{n} {(φ_{i} \circ \bar{φ_{j}})}_{k}^{l} (\bar{ω_{1}} \land . . . \land \bar{ω_{l - 1}} \land \bar{ω_{l + 1}} \land . . . \land \bar{ω_{n}}) \end{matrix}

(114)

Combining 113

()

and 114

()

we get

{(φ_{i} \circ \bar{φ_{j}} \land i d_{n - 2})}^{*} (* \bar{ω_{k}}) = \sum_{l = 1}^{n} {(\bar{φ_{j}} \circ φ_{i})}_{l}^{k} (\bar{ω_{1}} \land . . . \land \bar{ω_{l - 1}} \land \bar{ω_{l + 1}} \land . . . \land \bar{ω_{n}}) =

\sum_{l = 1}^{n} {(φ_{i} \circ \bar{φ_{j}})}_{k}^{l} (\bar{ω_{1}} \land . . . \land \bar{ω_{k - 1}} \land \bar{ω_{k + 1}} \land . . . \land \bar{ω_{n}}) .

By using the Calabi-Yau metric and the holomorphic volume form we can identify the dual of

Ω_{M}^{0, n - 1}

with

Ω_{M}^{1, 0} .

This identification is given by

\bar{ω_{1}} \land . . . \land \bar{ω_{l - 1}} \land \bar{ω_{l + 1}} \land . . . \land \bar{ω_{n}} ⟶ ω_{l} .

Thus from 110

()

we get

* ((φ_{i} \circ \bar{φ_{j}}) (\bar{ω_{k}})) =

\begin{matrix} {(φ_{i} \circ \bar{φ_{j}} \land i d_{n - 2})}^{*} (* \bar{ω_{k}}) = \sum_{l = 1}^{n} {(\bar{φ_{j}} \circ φ_{i})}_{l}^{k} (* \bar{ω_{k}}) = \sum_{l = 1}^{n} {(\bar{φ_{j}} \circ φ_{i})}_{l}^{k} ω_{k} \end{matrix}

(115)

From 115

()

and 113

()

we conclude Proposition 50 .

■

Corollary 51 Formula 115

()

implies that the composition of the complex conjugation with the Hodge star operator

*

identifies the restriction of the image

(\bar{φ_{j}} \circ φ_{i}) (Im \partial)

Im \partial

C^{\infty} (

Ω_{M}^{1, 0})

with the restriction of the image

(φ_{i} \circ \bar{φ_{j}} \land i d_{n - 2}) (Im {\bar{\partial}}^{*})

Im {\bar{\partial}}^{*}

C^{\infty} (

Ω_{M}^{0, n - 1})

and the restriction of

\bar{φ_{j}} \circ φ_{i} (Im \partial^{*})

Im \partial^{*}

C^{\infty} (

Ω_{M}^{1, 0})

with the restriction of

(φ_{i} \circ \bar{φ_{j}} \land i d_{n - 2}) (Im \bar{\partial})

Im \bar{\partial}

C^{\infty} (

Ω_{M}^{0, n - 1}) .

From 115

()

, 108

(),

and the identification

Ω_{M}^{1, 0}

with

Ω_{M}^{0, n - 1}

we deduce that we can identify

(\bar{φ_{j}} \circ φ_{i}) Im \partial^{*}

C^{\infty} (

Ω_{M}^{1, 0})

with

(φ_{i} \circ \bar{φ_{j}} \land i d_{n - 2}) Im \bar{\partial}

C^{\infty} (

Ω_{M}^{0, n - 1})

and

(\bar{φ_{j}} \circ φ_{i}) Im \partial

C^{\infty} (

Ω_{M}^{1, 0})

with

(φ_{i} \circ \bar{φ_{j}} \land i d_{n - 2}) Im {\bar{\partial}}^{*}

C^{\infty} (

Ω_{M}^{0, n - 1}) .

Since on a Kähler manifold we have that

Δ_{\partial} = Δ_{\bar{\partial}},

109

()

is established

.

Lemma 48 is proved.

■

From 109

()

we deduce that

\begin{matrix} T r (exp (- t Δ_{τ, 1}^{^{'}}) ℱ^{^{^{'}}} (1, φ_{i} \circ \bar{φ_{j}}) + exp (- t Δ_{τ, 1}^{^{”}}) ℱ^{^{”}} (1, φ_{i} \circ \bar{φ_{j}})) = \int_{M} T r (exp (- t Δ_{τ, 1}) (φ_{i} \circ \bar{φ_{j}})) v o l . \end{matrix}

(116)

We know that

\begin{matrix} T r (ℱ^{^{^{'}}} (1, φ_{i} \circ \bar{φ_{j}}) + ℱ^{^{”}} (1, φ_{i} \circ \bar{φ_{j}})) = \int_{M} T r (φ_{i} \circ \bar{φ_{j}}) v o l = W . P . \end{matrix}

(117)

Combining 116

(),

117

()

with Theorem 45 we deduce that

\begin{matrix} d d^{c} (log det Δ_{τ, 1}) = d d^{c} log (Δ_{τ, 1}^{^{'}} \times Δ_{τ, 1}^{^{”}}) = - Im W . P . \end{matrix}

(118)

Theorem 47 is proved.

■

Theorem 52 The relative dualizing sheaf

π_{*} (ω_{X / M_{L} (M)}) : = ℒ

is a trivial C

^{\infty}

line bundle.

Proof: It is well known fact that a complex line bundle on a complex manifold

M

(M) is topologically trivial if and only if its first Chern class is zero. According to [32] the first Chern class of the relative dualizing sheaf

π_{*} (ω_{X / M_{L} (M)}) : = ℒ

is the minus the imaginary part of the Weil-Petersson metric on

\bar{M} (

M). We observe that the regularized determinant of the Laplacian of the CY metric is a well defined function on

M (

M). Theorem 47 implies that

d d^{c} (log det Δ_{τ, 1})

is minus the imaginary part of the Weil-Petersson metric on

\bar{M} (

M). Thus the first Chern class of the relative dualizing sheaf

π_{*} (ω_{X / M_{L} (M)}) : = ℒ

is represented by the zero class of cohomology. This proves that the relative dualizing sheaf

π_{*} (ω_{X / M_{L} (M)}) : = ℒ

is topologically trivial on

M

(M). Theorem 52 is proved.

■

7 The Computation of the Variations of the Regularized Determinant on Kähler Ricci flat Manifold with a Non Trivial Canonical Bundle

Theorem 53 Suppose that M is

n

complex dimensional Kähler Ricci flat Manifold such that

{(K_{M})}^{\otimes n} ≊ O_{M} .

Then

\begin{matrix} d d^{c} (log det Δ_{τ, n}) = - Im W . P . \end{matrix}

(119)

Proof: The proof of Theorem 53 is based on Theorem 45 which shows that we need to compute

\int_{M} T r (ℱ^{^{'}} (n, (φ_{i} \circ \bar{φ_{j}})))) v o l (g) .

Here we are using the fact that

det Δ_{τ, n} = det Δ_{τ, n - 1}^{”} .

Since we assumed that

H^{n} (

Ω_{M}^{n}

)

= 0

we get that

\int_{M} T r (ℱ^{^{'}} (n, (φ_{i} \circ \bar{φ_{j}})))) v o l (g) = \int_{M} T r (((φ_{i} \circ \bar{φ_{j}}) \land i d_{n - 1}))) v o l (g) .

Direct computation using elementary linear algebra shows that if

A : V \to V

is a linear operator acting on an

n

dimensional vector space then the trace of the operator

A \land i d_{n - 1}

acting on

\land^{n} V

is given by the formula:

\begin{matrix} T r (A \land i d_{n - 1}) = T r (A) . \end{matrix}

(120)

From formula 120

()

and the definition of the Weil-Petersson metric given by 22

()

we derive 119

() .

Theorem 53 is proved.

■

References

D. Abramovich, J. -F. Burnol, J. Kramer and C. Soulé, ”Lectures on Arakelov Geometry”, Cambridge Studies In Advanced Mathematics Volume 33, Cambridge University Press, 1992.
M. Bershadsky, S. Cecotti, H. Ooguri and C. Vafa, ”Kodaira-Spencer Theory of Gravity and Exact Results for Quantum String Amplitude”, Comm. Math. Phys. 165 (1994), 311-428.
N. Berline, E. Getzler and M. Vergne, ”Heat Kernels and Dirac Operators”, Springer Verlag, 1991.
E. Calabi, ”Extremal Kähler Metrics”, Seminar on Differential Geometry, ed. S.-T. Yau, Annals of Mathematical Studies, vol. 102, Princeton University Press.
S. Donaldson, ”Infinite Determinants, Stable Bundles and Curvature”, Duke Mathematical Journal, vol. 54, Number 1(1987), 231-248.
S. Donaldson and P. Kronheimer, ”The Geometry of Four Manifolds”, Oxford Mat. Monog., Oxford Science Publications, Oxford University Press, New York 1990.
A. Grothemdieck, ”Éléments de Géométrie Algébrique”, IHES, Publications Mathématiques, vol. 8.
H. Fang and Z. Lu, ”Generalized Hodge Metric and BCOV Torsion on Calabi-Yau Moduli”, DG/0310007.
P. Gilkey, ”Invariance Theory, The Heat Equation, And the Atiyah-Singer Index Theorem”, Mathematics Lecture Series vol. 11, Publish or Parish, Inc. Wilmington, Delaware (USA) 1984.
Ph. Griffiths, ”Periods of Integrals on Algebraic Manifolds” I and II, Amer. Jour. Math. 90 (1968), 568-626 and 805-865.
Ph. Griffiths and J. Harris, ”Principles of Algebraic Geometry”.
J. Harvey and G. Moore, ”Algebras, BPS States, and Strings”, Nucl. Phys. B 463(1996), 315-368, hep-th/9510182 .
J. Harvey and G. Moore, ”On the Algebras of BPS States”, hep-th/9609017.
J. Jorgenson and A. Todorov, ”A Conjectural Analogue of Dedekind Eta Function for K3 Surfaces”, Math. Research Lett. 2(1995) 359-360.
J. Jorgenson and A. Todorov, ”Analytic Discriminant for Manifolds with Zero Canonical Class”, Manifolds and Geometry, ed. P. de Bartolomeis, F. Tircerri and E. Vesantini, Symposia Mathematica 36, (1996) 223-260.
J. Jorgenson and A. Todorov, ”Ample Divisors, Automorphic Forms and Shafarevich's Conjecture”, Mirror Symmetry IV, ed. S-T. Yau and Phong, AMS.
J. Jorgenson and A. Todorov, ”Analytic Discriminant for Polarized Algebraic K3 Surfaces”, Mirror Symmetry III, ed. S-T. Yau and Phong, AMS, p. 211-261.
J. Jorgenson and A. Todorov, ”Enriques Surfaces, Analytic Discriminants, and Borcherd's $Φ$ Function”, Com. Math. Phys. 191, (1998), 249-264.
J. Jorgenson and A. Todorov ”Correction to ”Enriques Surfaces Analytic Discriminants, and Borcherd's $Φ$ Function”, Com. Math. Phys. (1999).
K. Kodaira and Morrow, ”Complex Manifolds”.
B. Lian, A. Todorov and Shing-Tung Yau, Maximal Unipotent Monodromy for Complete Interesction CY Manifolds”, American Journal of Mathematics, 127(2005), 1-50.
K. Liu, A. Todorov, Shing-Tung Yau and K. Zuo, ”The Analogue of Shafarevich Conjecture for CY Manifolds I.”, (to appear)..
Z. Lu, ”On the Curvature Thensor of the Hodge Metric of Moduli of Polarized Calabi-Yau Threefolds”, J. Geom. Analysis, 11(2001), No 4, 633-645.
Z. Lu and X. Sun, ”Weil-Petersson Geometry on Moduli space of Polarized Calabi-Yau Manifolds.” preprint.
Z. Lu and X. Sun, ”On Weil-Petersson Volume and the First Chern Class of the Moduli Space of CY Manifolds.” preprint.
D. Mumford, ”Hirzebruch's Proportionality Principle in the Non-Compact Case”, Inv. Math. 42(1977), 239-272.
J.Roe, ”Elliptic Operators, Topology and Asymptotic Methods” Pitman Research Notes in Mathematics Series 179, Longman Scientific & Technical, 1988.
D. Ray and I. Singer, ”Analytic Torsion for Complex Manifolds”, Ann. Math. 98 (1973) 154-177.
A. Grothendieck, ”Séminaire de Géométrie Algébrique”, IHES, Paris, 1960-1961.
D. Sullivan, ”Infinitesimal Computations in Topology”, Publ. Math. IHES, No 47 (1977), 269-331.
G. Tian, ”Smoothness of the Universal Deformation Space of Calabi-Yau Manifolds and its Petersson-Weil Metric”, Math. Aspects of String Theory, ed. S.-T.Yau, World Scientific (1998), 629-346.
A. Todorov, ”The Weil-Petersson Geometry of Moduli Spaces of SU(n $\geq 3$ ) (Calabi-Yau Manifolds) I”, Comm. Math. Phys. 126 (1989), 325-346.
A. Todorov, ”Weil-Petersson Volumes of the Moduli of CY Manifolds”, preprint.
S. T. Yau, ”On the Ricci Curvature of Compact Kähler Manifolds and Complex Monge-Amper Equation I”, Comm. Pure and App. Math. 31, 339-411(1979).

The Analogue of the Dedekind Eta Function for CY Manifolds.

Jamey Bass University of California Department of Mathematics, Santa Cruz, CA 95064

Andrey Todorov Partially supported by The Institute of Mathematical Sciences of The Chinese University of Hong Kong. University of California Department of Mathematics, Santa Cruz, CA 95064 Bulgarian Academy of Sciences Institute of Mathematics Sofia, Bulgaria

November 27, 2006