Holomorphic Functions of Slow Growth on Coverings of Pseudoconvex Domains in Stein Manifolds, ,

1 . Introduction.

Let

M

be a complex manifold satisfying

\begin{matrix} M \subset \subset \tilde{M} \subset N a n d π_{1} (M) = π_{1} (N) \end{matrix}

(1.1)

where

M

and

\tilde{M}

are open connected subsets of a complex manifold

N

and

\tilde{M}

is Stein. (Here

π_{1} (X)

stands for the fundamental group of

X

.) Condition ( 1.1 ) is valid, e.g., for

M

a strictly pseudoconvex domain or an analytic polyhedra in a Stein manifold. It implies that the group

π_{1} (N)

is finitely generated. In [Br1] we presented a method to construct integral representation formulas for holomorphic functions of slow growth defined on unbranched coverings of

M

. Using such formulas we established that some known results for holomorphic functions on

M

can be extended to similar results for holomorphic functions of slow growth on coverings of

M

. In this paper we continue to study holomorphic functions of slow growth on coverings of

M

and apply the methods developed in [Br1] to extend and strengthen certain results of Gromov, Henkin and Shubin [GHS] on holomorphic

L^{2}

functions on coverings of pseudoconvex manifolds in the case of coverings of Stein manifolds.

The presentation in this paper is focused on several problems and results formulated in [GHS]. To describe them we, first, recall some definitions.

Let

M \subset \subset N

be a domain with a smooth boundary

b M

in an

n

-dimensional complex manifold

N

, that is,

\begin{matrix} M = {z \in N : ρ (z) < 0} \end{matrix}

(1.2)

where

ρ

is a real-valued function of class

C^{2} (Ω)

in a neighbourhood

Ω

of the compact set

\bar{M} : = M \cup b M

such that

\begin{matrix} d ρ (z) \neq 0 f o r a l l z \in b M . \end{matrix}

(1.3)

Let

z_{1}, . . ., z_{n}

be complex local coordinates in

N

near

z \in b M

. Then the tangent space

T_{z} N

z

is identified with

C^{n}

. By

T_{z}^{c} (b M) \subset T_{z} N

we denote the complex tangent space to

b M

z

, i.e.,

\begin{matrix} T_{z}^{c} (b M) = {w = (w_{1}, . . ., w_{n}) \in T_{z} (N) : \sum_{j = 1}^{n} \frac{\partial ρ}{\partial z_{j}} (z) w_{j} = 0} . \end{matrix}

(1.4)

The Levi form of

ρ

z \in b M

is a hermitian form on

T_{z}^{c} (b M)

defined in the local coordinates by the formula

\begin{matrix} L_{z} (w, \bar{w}) = \sum_{j, k = 1}^{n} \frac{\partial^{2} ρ}{\partial z_{j} \partial {\bar{z}}_{k}} (z) w_{j} {\bar{w}}_{k} . \end{matrix}

(1.5)

The manifold

M

is called pseudoconvex if

L_{z} (w, \bar{w}) \geq 0

for all

z \in b M

and

w \in T_{z}^{c} (b M)

. It is called strictly pseudoconvex if

L_{z} (w, \bar{w}) > 0

for all

z \in b M

and all

w \neq 0

w \in T_{z}^{c} (b M)

Equivalently, strictly pseudoconvex manifolds can be described as the ones which locally, in a neighbourhood of any boundary point, can be presented as strictly convex domains in

C^{n}

. It is also known (see [C], [R]) that any strictly pseudoconvex manifold admits a proper holomorphic map with connected fibres onto a normal Stein space.

Without loss of generality we may and will assume that

π_{1} (M) = π_{1} (N)

for

M

as above. Let

r : N_{G} \to N

be the regular covering of

N

with (discrete) transformation group

G

. Then

M_{G} : = r^{- 1} (M)

is a regular covering of

M

(with the same transformation group). It is a domain in

N_{G}

with the smooth boundary

b M_{G} : = r^{- 1} (b M)

. By

{\bar{M}}_{G} : = M_{G} \cup b M_{G}

we denote the closure of

M_{G}

N_{G}

Let

X

be a subspace of the space

O (M_{G})

of all holomorphic functions on

M_{G}

A point

z \in b M_{G}

is called a peak point for

X

if there exists a function

f \in X

such that

f

is unbounded on

M_{G}

but bounded outside

U \cap M_{G}

for any neighbourhood

U

z

N_{G}

A point

z \in b M_{G}

is called a local peak point for

X

if there exists a function

f \in X

such that

f

is unbounded in

U \cap M_{G}

for any neighbourhood

U

z

N_{G}

and there exists a neighbourhood

U

z

N_{G}

such that for any neighbourhood

V

z

N_{G}

the function

f

is bounded on

U \ V

The Oka-Grauert theorem [G1] states that if

M

is strictly pseudoconvex and

b M

is not empty then every

z \in b M

is a peak point for

O (M)

. In general it is not known whether the similar statement is true for boundary points of

M_{G}

with an infinite

G

Let

d V_{M_{G}}

be the Riemannian volume form on

M_{G}

obtained by a Riemannian metric pulled back from

N

. By

H^{2} (M_{G})

we denote the Hilbert space of holomorphic functions

g

M_{G}

with norm

{(\int_{z \in M_{G}} | g (z) |^{2} d V_{M_{G}} (z))}^{1 / 2} .

In [GHS] the von Neumann

G

-dimension

d i m_{G}

was used to measure the space

H^{2} (M_{G})

. In particular, in [GHS, Theorem 0.2] the following result was proved.

Theorem A If

M

is strictly pseudoconvex, then

(a) $d i m_{G} H^{2} (M_{G}) = \infty$ and
(b) each point in $b M_{G}$ is a local peak point for $H^{2} (M_{G})$ .

In [GHS, Theorem 0.5] the similar result was established for a covering

M_{G}

of a pseudoconvex manifold

M

with a strictly plurisubharmonic

G

-invariant function existing in a neighbourhood of

b M_{G}

. Finally, in [GHS, section 4] the following open problems were formulated.

Suppose that

M

is strictly pseudoconvex.

1. Does there exist a finite number of functions in

H^{2} (M_{G}) \cap C ({\bar{M}}_{G})

which separate all points in

b M_{G}

2. Assume that

d i m_{C} M = 2

. Does there exist

f \in H^{2} (M_{G}) \cap C ({\bar{M}}_{G})

such that

f (x) \neq 0

for all

x \in b M_{G}

3. Is it true that for every

C R

-function

f \in L^{2} (b M_{G}) \cap C (b M_{G})

in case

d i m_{C} M_{G} > 1

there exists

f^{'} \in H^{2} (M_{G}) \cap C ({\bar{M}}_{G})

such that

f^{'} |_{b M_{G}} = f

Here

L^{2} (b M_{G})

is defined similar to

H^{2} (M_{G})

with respect to the volume form on

b M_{G}

obtained by a Riemannian metric pulled back from

N

. Also, recall that

f \in C (b M_{G})

is called a

C R

-function if for every smooth

(n, n - 2)

-form

ω

with a compact support one has

\int_{b M_{G}} f \land \bar{\partial} ω = 0 .

f

is smooth this is equivalent to the fact that

f

is a solution of the tangential

C R

-equations:

{\bar{\partial}}_{b} f = 0

(see, e.g., [KR]).

The present paper deals with the above results and problems in the case of coverings of

M

satisfying condition ( 1.1 ).

2 . Formulation of Main Results.

2.1. We start with some results related to Problem 2 of the Introduction. Let

M

be a manifold satisfying condition ( 1.1 ) and

M^{'}

be an unbranched covering of

M

. Condition ( 1.1 ) implies that there is a covering

r : N^{'} \to N

N

such that

M^{'}

is a domain in

N^{'}

(i.e.,

π_{1} (M^{'}) = π_{1} (N^{'})

). As above

\bar{M^{'}}

denotes the closure of

M^{'}

N^{'}

Let

φ : N^{'} \to R

be a function uniformly continuous with respect to the path metric induced by a Riemannian metric pulled back from

N

Theorem 2.1 There exist a function

f_{φ} \in O (M^{'}) \cap C (\bar{M^{'}})

and a constant

C = C (φ, M^{'}, N^{'})

^$1$ such that

| f_{φ} (z) - φ (z) | < C a n d | d f_{φ} (z) | < C f o r a l l z \in M^{'} .

(Here the norm

| ω (z) |

of a differential form

ω

z \in M^{'}

is determined with respect to the Riemannian metric pulled back from

N

.) As a corollary of this result we answer an extended version of Question 2 of the Introduction for coverings of manifolds

M

satisfying condition ( 1.1 ). Namely, let

d

be the path metric on

N^{'}

obtained by the pullback of a Riemannian metric defined on

N

. Fix a point

o \in M^{'}

and set

d_{o} (x) : = d (o, x), x \in N^{'} .

From the triangle inequality it follows that the function

φ (x) : = d_{o} (x)

x \in N^{'}

, satisfies the hypothesis of Theorem 2.1 .

Corollary 2.2 Let

f : = f_{φ}

be the function from Theorem 2.1 for

φ = d_{o}

. Then there exists a constant

α > 0

such that

F = e^{- α f} \in H^{2} (M^{'}) \cap C (\bar{M^{'}})

(Note that here

F (x) \neq 0

for all

x \in M^{'}

and there are no restrictions on

d i m_{C} M

.) 2.2. In this part we formulate our results related to Theorem A of the Introduction. Let

M^{'}

be an unbranched covering of

M

satisfying ( 1.1 ). Let

ψ : M^{'} \to R_{+}

be a continuous function and

d V_{M^{'}}

be the Riemannian volume form on

M^{'}

obtained by a Riemannian metric pulled back from

N

. For an open set

D \subset M

we introduce the Banach space

H_{ψ}^{p} (D^{'})

of holomorphic functions

g

D^{'} : = r^{- 1} (D) \subset M^{'}

with norm

{(\int_{z \in M^{'}} | g (z) |^{p} ψ (z) d V_{M^{'}} (z))}^{1 / p} .

Let

r : N^{'} \to N

be the covering of

N

satisfying ( 1.1 ) such that

π_{1} (N^{'}) = π_{1} (M^{'})

. Then

M^{'} (= r^{- 1} (M))

is a domain in

N^{'}

. Suppose that

ψ : N^{'} \to R_{+}

is such that

log ψ

is uniformly continuous with respect to the path metric induced by a Riemannian metric pulled back from

N

. We set

φ : = log ψ

and consider the holomorphic function

f_{φ}

from Theorem 2.1 . This theorem implies that for

\tilde{C} : = e^{C}

\begin{matrix} \frac{1}{\tilde{C}} ψ (z) \leq | e^{f_{φ} (z)} | \leq \tilde{C} ψ (z), z \in \bar{M^{'}} . \end{matrix}

(2.1)

Therefore the following result holds.

Proposition 2.3 For any open set

D \subset M

and every

p \in [1, \infty)

the map

L_{ψ} : H_{ψ}^{p} (D^{'}) \to H_{1}^{p} (D^{'})

L_{ψ} (g) = g \cdot e^{f_{φ} / p}

, is an isomorphism of Banach spaces.

□

Let us now formulate an extension of Theorem A of the Introduction. Suppose that

M

is a strictly pseudoconvex domain in a complex manifold

N

such that

π_{1} (M) = π_{1} (N)

and

N

is a domain in a Stein manifold. Let

r : N^{'} \to N

be an unbranched covering of

N

and

M^{'} = r^{- 1} (M)

be the corresponding covering of

M

. Let

b M^{'} = r^{- 1} (b M)

be the boundary of

M^{'}

N^{'}

Theorem 2.4 Each point in

b M^{'}

is a peak point for

O (M^{'})

and for every

H_{1}^{p} (M^{'})

1 \leq p < \infty

From Theorem 2.4 and Proposition 2.3 we get (for

ψ

as in Proposition 2.3 )

Corollary 2.5 Each

z \in b M^{'}

is a peak point for

H_{ψ}^{p} (M^{'})

1 \leq p < \infty

□

Remark 2.6 The main ingredient of the proof of Theorem 2.4 is uniform estimates for solutions of certain

\bar{\partial}

-equations on

M^{'}

. In fact similar estimates are valid on coverings of, so-called, non-degenerate pseudoconvex polyhedrons on Stein manifolds (see [SH] and [He] for their definition). This class contains, in particular, piecewise strictly pseudoconvex domains and non-degenerate analytic polyhedrons on Stein manifolds. Also, every

M

from this class satisfies condition ( 1.1 ). Let

M^{'}

be a covering of such

M

and

z \in b M^{'}

be such that

M^{'} \cap U

is strictly pseudoconvex for a neighbourhood

U \subset N^{'}

z

. Then arguing as in the proof of Theorem 2.4 one obtains that

z

is a peak point for

O (M^{'})

and for every

H_{1}^{p} (M^{'})

1 \leq p < \infty

2.3.

In this section we discuss some results related to Problem 3 of the Introduction. Let

r : N^{'} \to N

be a covering of

N

satisfying ( 1.1 ). As before we set

M^{'} = r^{- 1} (M) \subset N^{'}

. Consider a function

ψ : N^{'} \to R_{+}

such that

log ψ

is uniformly continuous with respect to the path metric induced by a Riemannian metric pulled back from

N

. For such

ψ

and every

x \in M

we introduce the Banach space

l_{p, ψ, x} (M^{'})

1 \leq p \leq \infty

, of functions

g

r^{- 1} (x) \subset M^{'}

with norm

\begin{matrix} | g |_{p, ψ, x} : = {(\sum_{y \in r^{- 1} (x)} | g (y) |^{p} ψ (y))}^{1 / p} . \end{matrix}

(2.2)

Next, for an open set

D \subset M

we introduce the Banach space

ℋ_{p, ψ} (D^{'})

1 \leq p \leq \infty

, of functions

f

holomorphic on

D^{'} : = r^{- 1} (D) \subset M^{'}

with norm

\begin{matrix} | f |_{p, ψ}^{D} : = {sup}_{x \in D} | f |_{p, ψ, x} . \end{matrix}

(2.3)

Clearly, one has a continuous embedding

ℋ_{p, ψ} (D^{'}) ↪ H_{ψ}^{p} (D^{'})

. Let

U \subset N

be an open set containing

\bar{D}

and

U^{'} = r^{- 1} (U)

. Then for

ψ

as above using the mean value property for plurisubharmonic functions one can easily show that for each

p \in [1, \infty]

the restriction

f \mapsto f |_{D^{'}}

induces a linear continuous map

H_{ψ}^{p} (U^{'}) \to ℋ_{p, ψ} (D^{'})

. Also, for such

ψ

from the results proved in [Br1] follow that holomorphic functions from

ℋ_{p, ψ} (M^{'})

separate all points in

M^{'}

(for each

p \in [1, \infty]

Let us formulate the main result of this section.

Let

M \subset \subset \tilde{M} \subset N

be manifolds satisfying condition ( 1.1 ) with

d i m_{C} M \geq 2

Suppose that

D \subset \subset M

is an open subset whose boundary

b D

is a connected

C^{k}

submanifold of

M

(1 \leq k \leq \infty)

. For a covering

r : N^{'} \to N

we set

D^{'} = r^{- 1} (D)

and

b D^{'} = r^{- 1} (b D)

Theorem 2.7 For every

C R

-function

f \in C^{s} (b D^{'})

0 \leq s \leq k

, satisfying

f |_{r^{- 1} (x)} \in l_{p, ψ, x} (M^{'}) f o r a l l x \in D a n d {sup}_{x \in b D} | f |_{p, ψ, x} < \infty

there exists a function

f^{'} \in ℋ_{p, ψ} (D^{'}) \cap C^{s} (\bar{D^{'}})

such that

f^{'} |_{b D^{'}} = f

Remark 2.8 (1) The converse to this theorem is always true: the restriction of every

f^{'} \in ℋ_{p, ψ} (D^{'}) \cap C^{s} (\bar{D^{'}})

b D^{'}

is a

C R

-function satisfying the hypotheses of the theorem.

(2) We will prove also (see ( 6.6 )) that for some

c = c (M^{'}, M, ψ, p)

| f^{'} |_{p, ψ}^{D} \leq c {sup}_{x \in b D} | f |_{p, ψ, x} .

As a corollary we obtain an analog of the Hartogs extension theorem. We formulate it for functions of the maximal possible growth for which our method works.

Suppose that

M

satisfies ( 1.1 ). Let

D \subset M

be a domain and

K \subset \subset D

be a compact set such that

U : = D \ K

is connected. Consider a covering

r : M^{'} \to M

and set

D^{'} = r^{- 1} (D)

U^{'} = r^{- 1} (U)

. By

d_{o}

o \in M^{'}

, we denote the distance function on

M^{'}

as in Corollary 2.2 .

Corollary 2.9 There exists a constant

c > 0

such that for every

f \in O (U^{'})

satisfying for some

c_{2} > 0

and

0 < c_{1} < c

the inequality

| f (z) | \leq e^{c_{2} e^{c_{1} d_{o} (z)}}, z \in U^{'},

there is

f^{'} \in O (D^{'})

such that

| f^{'} (z) | \leq e^{c_{3} e^{c_{1} d_{0} (z)}}, z \in D^{'}, a n d f^{'} |_{U^{'}} = f;

where

c_{3}

depends on

c_{2}, c_{1}, c, M, M^{'}

only.

We don't know whether a similar extension result holds for functions

f

growing faster than those of the corollary.

2.4. Finally, we formulate a result related to Problem 1 of the Introduction. First, we recall some definitions of the theory of flat vector bundles (see, e.g., [O]).

Let

X

be a complex manifold and

ρ : π_{1} (X) \to G L_{k} (C)

be a homomorphism of its fundamental group. We set

G : = π_{1} (X) / K e r ρ

. It is well known (see, e.g., Example 3.2 (b) below) that to any such

ρ

corresponds a complex flat vector bundle

E_{ρ}

X

(i.e., a bundle constructed by a locally constant cocycle). We call

E_{ρ}

the bundle associated with

ρ

. Assume that

ρ

is such that

E_{ρ}

is topologically trivial, i.e., is isomorphic in the category of continuous bundles to the bundle

X \times C^{k}

. Every such

ρ

can be obtained as the monodromy of the equation

d F = ω F

X

where

ω

is a matrix-valued

1

-form on

X

satisfying

d ω - ω \land ω = 0

. By

T (X)

we denote the class of quotient groups

G

obtained by representations

ρ

as above.

Now, let

r : M_{G} \to M

be the regular covering of

M

satisfying condition ( 1.1 ) with transformation group

G

. Let

G_{1} \subset G

be a subgroup of a finite index. Then there is a finite covering

r_{1} : M_{1} \to M

whose fibre is the quotient set

G / G_{1}

such that

M_{G}

is also the regular covering of

M_{1}

with transformation group

G_{1}

Theorem 2.10 Assume that

G_{1} \in T (M_{1})

. Then there is a finite number of functions in

ℋ_{2, 1} (M_{G}) \cap C ({\bar{M}}_{G})

which separate all points in

M_{G}

Remark 2.11 (1) We will see from the proof that the functions in Theorem 2.10 can be taken even from

ℋ_{2, ψ} (M_{G})

where

ψ : M_{G} \to R_{+}

has a double exponential growth.

(2) As the group

G

in Theorem 2.10 one can take, e.g., a finitely generated free group (see, e.g., [O]) or a polycyclic group (see, e.g., [Ra]). If

d i m_{C} M = 1

then, since

M

is homotopically equivalent to a one-dimensional

C W

-complex (see, e.g., [GR]), every quotient group

G

obtained by a linear representation

ρ

belongs to

T (M)

3 . Preliminary Results.

3.1. First, we recall some basic facts from the theory of bundles see, e.g., [Hi].

Let

X

be a complex analytic space and

S

be a complex analytic Lie group with unit

e \in S

. Consider an effective holomorphic action of

S

on a complex analytic space

F

. Here holomorphic action means a holomorphic map

S \times F \to F

sending

s \times f \in S \times F

s f \in F

such that

s_{1} (s_{2} f) = (s_{1} s_{2}) f

and

e f = f

for any

f \in F

Efficiency means that the condition

s f = f

for some

s

and any

f

implies that

s = e

Definition 3.1 A complex analytic space

W

together with a holomorphic map (projection)

π : W \to X

is called a holomorphic bundle on

X

with structure group

S

and fibre

F

, if there exists a system of coordinate transformations, i.e., if

(1) there is an open cover $U = {U_{i}}_{i \in I}$ of $X$ and a family of biholomorphisms $h_{i} : π^{- 1} (U_{i}) \to U_{i} \times F$ , that map “fibres” $π^{- 1} (u)$ onto $u \times F$ ;
(2) for any $i, j \in I$ there are elements $s_{i j} \in O (U_{i} \cap U_{j}, S)$ such that $(h_{i} h_{j}^{- 1}) (u \times f) = u \times s_{i j} (u) f f o r a n y u \in U_{i} \cap U_{j}, f \in F .$

A holomorphic bundle

π : W \to X

whose fibre is a Banach space

F

and the structure group is

G L (F)

(the group of linear invertible transformations of

F

) is called a holomorphic Banach vector bundle. A holomorphic section of a holomorphic bundle

π : W \to X

is a holomorphic map

s : X \to W

satisfying

π \circ s = i d

We will use the following construction of holomorphic bundles (see, e.g. [Hi, Ch.1]): Let

S

be a complex analytic Lie group and

U = {U_{i}}_{i \in I}

be an open cover of

X

Z_{O}^{1} (U, S)

we denote the set of holomorphic

S

-valued

U

-cocycles. By definition,

s = {s_{i j}} \in Z_{O}^{1} (U, S)

, where

s_{i j} \in O (U_{i} \cap U_{j}, S)

and

s_{i j} s_{j k} = s_{i k} o n U_{i} \cap U_{j} \cap U_{k}

Consider the disjoint union

⊔_{i \in I} U_{i} \times F

and for any

u \in U_{i} \cap U_{j}

identify the point

u \times f \in U_{j} \times F

with

u \times s_{i j} (u) f \in U_{i} \times F

. We obtain a holomorphic bundle

W_{s}

X

whose projection is induced by the projection

U_{i} \times F \to U_{i}

. Moreover, any holomorphic bundle on

X

with structure group

S

and fibre

F

is isomorphic (in the category of holomorphic bundles) to a bundle

W_{s}

Example 3.2 (a) Let

M

be a complex manifold. For any subgroup

H \subset π_{1} (M)

consider the unbranched covering

r : M (H) \to M

corresponding to

H

. We will describe

M (H)

as a holomorphic bundle on

M

First, assume that

H \subset π_{1} (M)

is a normal subgroup. Then

M (H)

is a regular covering of

M

and the quotient group

G : = π_{1} (M) / H

acts holomorphically on

M (H)

by deck transformations. It is well known that

M (H)

in this case can be thought of as a principle fibre bundle on

M

with fibre

G

(here

G

is equipped with the discrete topology). Namely, let us consider the map

R_{G} (g) : G \to G

g \in G

, defined by the formula

R_{G} (g) (q) = q \cdot g^{- 1}, q \in G .

Then for an open cover

U = {U_{i}}_{i \in I}

M

by sets biholomorphic to open Euclidean balls in some

C^{n}

there is a locally constant cocycle

c = {c_{i j}} \in Z_{O}^{1} (U, G)

such that

M (H)

is biholomorphic to the quotient space of the disjoint union

V = ⊔_{i \in I} U_{i} \times G

by the equivalence relation:

U_{i} \times G ∋ x \times R_{G} (c_{i j}) (q) \sim x \times q \in U_{j} \times G

. The identification space is a holomorphic bundle with projection

r : M (H) \to M

induced by the projections

U_{i} \times G \to U_{i}

. In particular, when

H = e

we obtain the definition of the universal covering

M_{u}

M

Assume now that

H \subset π_{1} (M)

is not necessarily normal. Let

X_{H} = π_{1} (M) / H

be the set of cosets with respect to the (left) action of

H

π_{1} (M)

defined by left multiplications. By

[H q] \in X_{H}

we denote the coset containing

q \in π_{1} (M)

. Let

A (X_{H})

be the group of all homeomorphisms of

X_{H}

(equipped with the discrete topology). We define the homomorphism

τ : π_{1} (M) \to A (X_{H})

by the formula:

τ (g) ([H q]) : = [H q g^{- 1}], q \in π_{1} (M) .

Set

Q (H) : = π_{1} (M) / K e r (τ)

and let

\tilde{g}

be the image of

g \in π_{1} (M)

Q (H)

. By

τ_{Q (H)} : Q (H) \to A (X_{H})

we denote the unique homomorphism whose pullback to

π_{1} (M)

coincides with

τ

. Consider the action of

H

V = ⊔_{i \in I} U_{i} \times π_{1} (M)

induced by the left action of

H

π_{1} (M)

and let

V_{H} = ⊔_{i \in I} U_{i} \times X_{H}

be the corresponding quotient set. Define the equivalence relation

U_{i} \times X_{H} ∋ x \times τ_{Q (H)} ({\tilde{c}}_{i j}) (h) \sim x \times h \in U_{j} \times X_{H}

with the same

{c_{i j}}

as in the definition of

M (e)

. The corresponding quotient space is a holomorphic bundle with fibre

X_{H}

biholomorphic to

M (H)

(b) We retain the notation of example (a). Let

B

be a complex Banach space and

G L (B)

be the group of invertible bounded linear operators

B \to B

. Consider a homomorphism

ρ : G \to G L (B)

. Without loss of generality we assume that

K e r (ρ) = e

, for otherwise we can pass to the corresponding quotient group. The holomorphic Banach vector bundle

E_{ρ} \to M

associated with

ρ

is defined as the quotient of

⊔_{i \in I} U_{i} \times B

by the equivalence relation

U_{i} \times B ∋ x \times ρ (c_{i j}) (w) \sim x \times w \in U_{j} \times B

for any

x \in U_{i} \cap U_{j}

. Let us illustrate this construction by an example.

Let

φ : X_{H} \to R^{+}

(

X_{H} : = π_{1} (M) / H

) be a function satisfying

\begin{matrix} φ (τ (h) (x)) \leq c_{h} φ (x), x \in X_{H}, h \in π_{1} (M), \end{matrix}

(3.1)

where

c_{h}

is a constant depending on

h

. By

l_{p, φ} (X_{H})

1 \leq p \leq \infty

, we denote the Banach space of complex functions

f

X_{H}

with norm

\begin{matrix} | | f | |_{p, φ} : = {(\sum_{g \in X_{H}} | f (g) |^{p} φ (g))}^{1 / p} . \end{matrix}

(3.2)

Then according to ( 3.1 ) the map

ρ

defined by the formula

[ρ (g) (f)] (x) : = f (τ (g) (x))

g \in π_{1} (M)

x \in X_{H}

, is a homomorphism of

π_{1} (M)

into

G L (l_{p, φ} (X_{H}))

. By

E_{p, φ} (X_{H})

we denote the holomorphic Banach vector bundle associated with this

ρ

3.2.

We retain the notation of Example 3.2 . Let

r : M^{'} \to M

be a covering where

M^{'} = M (H)

(i.e.,

π_{1} (M^{'}) = H

). Assume that

M

satisfies condition ( 1.1 ), i.e.,

M \subset \subset N

and

π_{1} (M) = π_{1} (N)

. Then there is an embedding

M (H) ↪ N (H)

. (Without loss of generality we consider

M (H)

as an open subset of

N (H)

.) Let

{V_{i}}_{i \in I}

be a finite acyclic open cover of

\bar{M}

by relatively compact sets. We set

U_{i} : = V_{i} \cap M

and consider the open cover

U = {U_{i}}_{i \in I}

M

. Then as in Example 3.2 (a) we can define

M (H)

by a cocycle

c = {c_{i j}} \in Z_{O}^{1} (U, π_{1} (M))

Further, let

ψ : N (H) \to R_{+}

be a function such that

log ψ

is uniformly continuous with respect to the path metric induced by a Riemannian metric pulled back from

N

. Fix a point

z_{0} \in M

and identify

r^{- 1} (z_{0})

with

z_{0} \times X_{H}

(

X_{H} : = π_{1} (M) / H

). We define the function

φ : X_{H} \to R_{+}

by the formula

φ (x) : = ψ (z_{0}, x), x \in X_{H} .

It was proved in [Br1, Lemma 2.3] that

φ

satisfies inequality ( 3.1 ). Then the bundle

E_{p, φ} (X_{H})

is well defined. By definition, any holomorphic section of this bundle is determined by a family

{f_{i} (z, g)}_{i \in I}

of holomorphic functions on

U_{i}

with values in

l_{p, φ} (X_{H})

satisfying

f_{i} (z, τ (c_{i j}) (h)) = f_{j} (z, h) f o r a n y z \in U_{i} \cap U_{j} .

We introduce the Banach space

B_{p, φ} (X_{H})

of bounded holomorphic sections

f = {f_{i}}_{i \in I}

E_{p, φ} (X_{H})

with norm

\begin{matrix} | f |_{p, φ} : = {sup}_{i \in I, z \in U_{i}} | | f_{i} (z, \cdot) | |_{p, φ} . \end{matrix}

(3.3)

(Here

| | \cdot | |_{p, φ}

is the norm on

l_{p, φ} (X_{H})

, see ( 3.2 ).) Further, let

f \in ℋ_{p, ψ} (M (H))

(see section 2.2 for the definition). We define the family

{f_{i}}_{i \in I}

of functions on

U_{i}

with values in the space of functions on

X_{H}

by the formula

\begin{matrix} f_{i} (z, g) : = f (z, g), z \in U_{i}, i \in I, g \in X_{H} . \end{matrix}

(3.4)

It was established in [Br1, Proposition 2.4] that the correspondence

f \mapsto {f_{i}}_{i \in I}

determines an isomorphism of Banach spaces

D : ℋ_{p, ψ} (M (H)) \to B_{p, φ} (X_{H})

. (Here

D

is an isometry for

ψ \equiv 1

.) Next, suppose that

{x_{n}}_{n \geq 1} \subset M

converges to

x \in M

. Then for sufficiently big

n

we can arrange

r^{- 1} (x_{n})

and

r^{- 1} (x)

in sequences

{y_{i n}}_{i \geq 1}

and

{y_{i}}_{i \geq 1}

such that every

{y_{i n}}

converges to

y_{i}

n \to \infty

. For such

n

we define maps

τ_{n} (x) : r^{- 1} (x) \to r^{- 1} (x_{n})

so that

τ_{n} (y_{i}) = y_{i n}

i \in N

. Below,

τ_{n}^{*}

denotes the transpose map generated by

τ_{n}

on functions defined on

r^{- 1} (x_{n})

and

r^{- 1} (x)

Definition 3.3 Let

X \subset M

be a subset.

We say that a function

f

r^{- 1} (X)

belongs to the class

C_{p, ψ} (r^{- 1} (X))

(1) $f |_{r^{- 1} (x)} \in l_{p, ψ, x} (M^{'})$ for all $x \in X$ and
(2) for any $x \in X$ and any sequence ${x_{n}} \subset X$ converging to $x$ the sequence of functions ${τ_{n}^{*} (f |_{r^{- 1} (x_{n})})}$ converges to $f |_{r^{- 1} (x)}$ in the norm of $l_{p, ψ, x} (M^{'})$ .

C_{p, ψ}^{b} (r^{- 1} (X))

we denote the Banach space of functions

f \in C_{p, ψ} (r^{- 1} (X))

with norm

\begin{matrix} | f |_{p, ψ}^{X} : = {sup}_{x \in X} | f |_{r^{- 1} (x)} |_{p, ψ, x} . \end{matrix}

(3.5)

Note that if

X \subset M

is compact, then

| f |_{p, ψ}^{X} < \infty

for every

f \in C_{p, ψ} (r^{- 1} (X))

Comparing with the above definition of

D

one determines a similar map for

C_{p, ψ}^{b} (r^{- 1} (X))

. This gives an isomorphism

D : C_{p, ψ}^{b} (r^{- 1} (X)) \to C B_{p, φ}^{X} (X_{H})

where

C B_{p, φ}^{X} (X_{H})

is the Banach space of bounded continuous sections of

E_{p, φ} (X_{H}) |_{X}

with norm defined as in ( 3.3 ).

3.3. Most of our proofs are based on Theorem 1.3 of [Br1]. In its proof we use the above isomorphisms

D

and Cartan's A and B theorems for coherent Banach vector sheaves (see [B]). Let us formulate this result.

Suppose that

r : M^{'} \to M

is a covering with

M

satisfying ( 1.1 ). We define

ψ : M^{'} \to R_{+}

as in section 3.2. Also, we define

ℋ_{p, ψ} (M^{'})

and

l_{p, ψ, x} (M^{'})

as in section 2.3. For Banach spaces

E

and

F

ℬ (E, F)

we denote the space of all linear bounded operators

E \to F

with norm

| | \cdot | |

Theorem 3.4 For any

p \in [1, \infty]

there is a family

{L_{z} \in ℬ (l_{p, ψ, z} (M^{'}), ℋ_{p, ψ} (M^{'}))}_{z \in M}

holomorphic in

z

such that

(L_{z} h) (x) = h (x) f o r a n y h \in l_{p, ψ, z} (M^{'}) a n d x \in r^{- 1} (z) .

Moreover,

{sup}_{z \in M} | | L_{z} | | < \infty .

The following facts are simple corollaries of this result.

Suppose that

X \subset M

and

f \in C_{p, ψ}^{b} (r^{- 1} (X))

. We define the function

F

X \times M^{'}

by the formula

\begin{matrix} F (x, z) : = (L_{x} (f |_{r^{- 1} (x)})) (z), x \times z \in X \times M^{'} . \end{matrix}

(3.6)

Then

F

is continuous and

F (x, \cdot) \in ℋ_{p, ψ} (M^{'})

for every

x

. Moreover, if

X

is open and

f \in ℋ_{p, ψ} (r^{- 1} (X))

, then

F \in O (X \times M^{'})

and the map

X \to ℋ_{p, ψ} (M^{'})

x \mapsto F (x, \cdot)

, is holomorphic.

We can also express

F

in local coordinates. Namely, take

x \in X

and let

U \subset M

be a neighbourhood of

x

biholomorphic to an open Euclidean ball. Then

r^{- 1} (U) = ⊔_{y \in r^{- 1} (x)} V_{y}

and there are biholomorphisms

s_{y} : U \to V_{y}

such that

r \circ s_{y} = i d

. Now, the restriction of

f \in C_{p, ψ}^{b} (r^{- 1} (X))

r^{- 1} (U \cap X)

can be written as

\begin{matrix} f (z) = \sum_{y \in r^{- 1} (x)} f (z) χ_{y} (z), z \in r^{- 1} (U \cap X), \end{matrix}

(3.7)

where

χ_{y}

is the characteristic function of

V_{y}

. Let us introduce the functions

{\tilde{f}}_{y}

y \in r^{- 1} (x)

, by the formulas

{\tilde{f}}_{y} (v) = f (s_{y} (v)), v \in X \cap U .

Then we have

\begin{matrix} f (z) = \sum_{y \in r^{- 1} (x)} {\tilde{f}}_{y} (v) χ_{y} (z), v = r (z) \in U \cap X . \end{matrix}

(3.8)

Consider the series

\begin{matrix} \sum_{y \in r^{- 1} (x)} {\tilde{f}}_{y} (v) L_{v} (χ_{y} |_{r^{- 1} (v)}), v \in U \cap X, \end{matrix}

(3.9)

with

L_{v}

as in Theorem 3.4 .

Proposition 3.5 For

p \in [1, \infty)

the series in ( 3.9 ) converges in

ℋ_{p, ψ} (M^{'})

F (v, \cdot) : = L_{v} (f |_{r^{- 1} (v)})

uniformly on every compact subset of

U \cap X

. If

p = \infty

and

f \in C_{1, 1}^{b} (r^{- 1} (X))

then this series also converges in

ℋ_{\infty, ψ} (M^{'})

F (v, \cdot)

uniformly on every compact subset of

U \cap X

Proof. Suppose that

p \in [1, \infty)

and

f \in C_{p, ψ}^{b} (r^{- 1} (X))

. Let

C \subset U \cap X

be a compact subset. By the definition the function

Φ : U \cap X \to l_{p, ψ} (X_{H})

z \mapsto {\tilde{f}}_{\cdot} (z)

, is continuous (here we identify

r^{- 1} (x)

with

X_{H}

). Thus

Φ (C) \subset l_{p, ψ} (X_{H})

is compact. Fix a family

{X_{i}}_{i \in N}

of finite subsets of

X_{H}

such that

X_{i} \subset X_{i + 1}

for any

i

and

\cup_{i = 1}^{\infty} X_{i} = X_{H}

. Let

V_{i} \subset l_{p, ψ} (X_{H})

be a finite-dimensional subspace generated by functions

δ_{z}

X_{H}

with

z \in X_{i}

. Here

δ_{z} (v) = 1

v = z

and

δ_{z} (v) = 0

v \neq z

. Then

\cup_{i = 1}^{\infty} V_{i}

is everywhere dense in

l_{p, ψ} (X_{H})

(since

1 \leq p < \infty

). This and compactness of

Φ (C)

imply that for any

ε > 0

there exists an integer

l

such that

Φ (C) \subset V_{l} + B_{ε}

where

B_{ε}

is the open ball in

l_{p, ψ} (X_{H})

centered at

0

of radius

ε

. By

p_{l} : l_{p, ψ} (X_{H}) \to V_{l}

we denote the projection sending

v = \sum_{x \in X_{H}} v_{x} δ_{x} \in l_{p, ψ} (X_{H})

\sum_{x \in X_{l}} v_{x} δ_{x} \in V_{l}

(here all

v_{x} \in C

). Then for

ε

as above and every

v \in Φ (C)

we have

| | v - p_{l} (v) | |_{p, φ} < ε

. From this by ( 3.8 ), identifying

r^{- 1} (x)

with

X_{H}

, we obtain

\begin{matrix} {sup}_{v \in C} | | f |_{r^{- 1} (v)} - \sum_{y \in X_{l}} {\tilde{f}}_{y} (v) χ_{y} |_{r^{- 1} (v)} | |_{p, φ} < ε . \end{matrix}

(3.10)

Thus by the definition of operators

L_{v}

(see Theorem 3.4 )

\begin{matrix} {sup}_{v \in C} | F (v, \cdot) - \sum_{y \in X_{l}} {\tilde{f}}_{y} (v) L_{v} (χ_{y} |_{r^{- 1} (v)}) |_{p, ψ}^{M} < C ε \end{matrix}

(3.11)

for some constant

C

. This implies the required uniform convergence for

p \in [1, \infty)

For

p = \infty

and

f \in C_{1, 1}^{b} (r^{- 1} (X))

we obtain anew that

Φ (C) \subset l_{1, 1} (X_{H})

is compact. Then in the above notation we easily get

| | v - p_{l} (v) | |_{\infty, φ} < ε

for any

v \in Φ (C)

(because

| | \cdot | |_{\infty, φ} = | | \cdot | |_{\infty, 1} \leq | | \cdot | |_{1, 1}

). Thus ( 3.10 ) is also valid for

p = \infty

This gives ( 3.11 ) with

p = \infty

□

4 . Proofs of Theorem 2.1 and Corollary 2.2 .

Proof of Theorem 2.1 . Let

M \subset \subset \tilde{M} \subset N

be complex manifolds such that

π_{1} (M) = π_{1} (N)

and

\tilde{M}

is Stein. Consider an unbranched covering

r : N^{'} \to N

N

and the corresponding coverings

M^{'} = r^{- 1} (M)

and

{\tilde{M}}^{'} = r^{- 1} (\tilde{M})

M

and

\tilde{M}

According to Example 3.2 (a)

{\tilde{M}}^{'}

is defined on an open cover

U = {U_{i}}_{i \in I}

\tilde{M}

by sets biholomorphic to open Euclidean balls by a locally constant cocycle

{{\tilde{c}}_{i j}} \in Z_{O}^{1} (U, Q (H))

. (Here we retain the notation of Example 3.2 (a) so that

{\tilde{M}}^{'} = \tilde{M} (H)

.) Using this construction we identify

r^{- 1} (U_{i})

with

U_{i} \times X_{H}

(

X_{H} = π_{1} (M) / H

). Also, we choose some points

z_{i} \in U_{i}

and assume that diameters of all

U_{i}

in the path metric on

N

induced by a Riemannian metric are uniformly bounded by a constant.

Let

φ : N^{'} \to R

be a function uniformly continuous with respect to the path metric induced by the Riemannian metric pulled back from

N

. For every

i \in I

we define a function

φ_{i} : r^{- 1} (U_{i}) \to R

by the formula

φ_{i} (z, g) : = φ (z_{i}, g), z \times g \in r^{- 1} (U_{i}) .

Then from the uniform continuity of

φ

and boundedness of

d i a m (U_{i})

for all

i

we obtain that there exists a constant

c

such that

\begin{matrix} | φ (v) - φ_{i} (v) | \leq c f o r e v e r y v \in r^{- 1} (U_{i}), i \in I . \end{matrix}

(4.1)

Define a locally constant cocycle

φ_{i j}

on the open cover

{r^{- 1} (U_{i})}

{\tilde{M}}^{'}

by the formula

φ_{i j} (v) = φ_{i} (v) - φ_{j} (v) f o r v \in r^{- 1} (U_{i} \cap U_{j}) .

Then from ( 4.1 ) by the triangle inequality we get

\begin{matrix} {sup}_{i, j, v} | φ_{i j} (v) | \leq 2 c . \end{matrix}

(4.2)

This inequality implies that rewriting cocycle

{φ_{i j}}

in the coordinates on

\tilde{M}

(i.e., taking its direct image

{r_{*} (φ_{i j})}

with respect to

r

) we can regard it as a holomorphic cocycle on the cover

U

with values in the Banach vector bundle

E_{\infty, 1} (X_{H})

with fibre

l_{\infty, 1} (X_{H})

defined on

\tilde{M}

(see Example 3.2 (b)). This correspondence is described in [Br2, Proposition 2.4]. Since

U

is acyclic and

\tilde{M}

is Stein, from the above construction, a version of Cartan's B theorem for coherent Banach sheaves (see [B]), and the classical Leray theorem we obtain as in [Br2] that there are holomorphic functions

f_{i} \in O (r^{- 1} (U_{i}))

such that

(1) for every compact set $K \subset U_{i}$ , ${sup}_{y \in r^{- 1} (K)} | f_{i} (y) | < \infty;$
(2) $f_{i} (z) - f_{j} (z) = φ_{i j} (z) f o r z \in r^{- 1} (U_{i} \cap U_{j}) .$

Let

V = {V_{j}}_{j \in J}

be a refinement of

U

such that every

V_{j}

is open and relatively compact in some

U_{i_{j}}

. Then condition (1) implies that

\begin{matrix} {sup}_{y \in r^{- 1} (V_{j})} | f_{i_{j}} (y) | < \infty . \end{matrix}

(4.3)

Finally, define a function

\tilde{f} \in O ({\tilde{M}}^{'})

by the formula

\begin{matrix} \tilde{f} (z) : = φ_{i} (z) - f_{i} (z), z \in r^{- 1} (U_{i}) . \end{matrix}

(4.4)

Since

\bar{M} \subset \tilde{M}

is a compact set, there is a finite subcover of

V

that covers

\bar{M}

. From here, ( 4.3 ) and ( 4.1 ) for the restriction

f_{φ} : = \tilde{f} |_{\bar{M^{'}}}

we obtain (for some

C

)

\begin{matrix} | f_{φ} (z) - φ (z) | < C a n d | d f_{φ} (z) | < C f o r a n y z \in M^{'} . \end{matrix}

(4.5)

The proof of the theorem is complete.

□

Proof of Corollary 2.2 . We retain the notation of the proof of Theorem 2.1 .

Let

d_{o} : = d (o, \cdot)

be the distance on

N^{'}

from a fixed point

o \in M^{'}

and let

f : = f_{φ}

be the function from Theorem 2.1 for

φ = d_{o}

. Consider a finite open cover

{U_{i}}_{i = 1}^{l}

\bar{M}

such that every

U_{i} \subset \subset \tilde{M}

is biholomorphic to an open Euclidean ball. As above we identify

r^{- 1} (U_{i}) \subset N^{'}

with

U_{i} \times X_{H}

. Fix an element

e \in X_{H}

and set

o_{i} (z) = (z, e) \in r^{- 1} (U_{i})

for every

z \in U_{i}

1 \leq i \leq l

, and

d_{o_{i} (z)} (v) : = d (o_{i} (z), v)

v \in N^{'}

. Then from compactness of every

\bar{U_{i}}

by the triangle inequality we get

\begin{matrix} | d_{o_{i} (z)} (v) - d_{o} (v) | \leq a, 1 \leq i \leq l, \end{matrix}

(4.6)

for some constant

a

. By

B_{z, i} (R)

we denote the open ball on

r^{- 1} (z)

of radius

R

centered at

o_{i} (z)

with respect to the induced metric

d |_{r^{- 1} (z)}

. Also, by

# A

we denote the number of elements of

A

. Now we prove

Lemma 4.1 There is

k \in N

such that

# B_{z, i} (R) \leq e^{k R}, 1 \leq i \leq l .

Proof. Let

\tilde{r} : N_{u} \to N

be the universal covering of

N

and

r^{'} : N_{u} \to N^{'}

be the intermediate covering, i.e.,

\tilde{r} = r \circ r^{'}

. We equip

N_{u}

with the path metric

\tilde{d}

induced by the Riemannian metric pulled back from

N

, the same as in the definition of the metric

d

N^{'}

. Let

{\tilde{o}}_{i} (z) \in {\tilde{r}}^{- 1} (z)

be such that

r^{'} ({\tilde{o}}_{i} (z)) = o_{i} (z)

. By

{\tilde{B}}_{\tilde{z}, i} (R)

we denote the open ball on

{\tilde{r}}^{- 1} (z)

of radius

R

centered at

{\tilde{o}}_{i} (z)

with respect to the metric

\tilde{d} |_{{\tilde{r}}^{- 1} (z)}

. Let us check that

r^{'} ({\tilde{B}}_{\tilde{z}, i} (R)) = B_{z, i} (R)

Indeed, let

y \in {\tilde{B}}_{\tilde{z}, i} (R)

and

γ_{y}

be a path joining

{\tilde{o}}_{i} (z)

and

y

N_{u}

whose length is less than

R

(such a path exists by the definition of

\tilde{d}

). Then

r^{'} (γ_{y})

is a path joining

o_{i} (z)

and

r^{'} (y)

N^{'}

. By the definition of the metrics on

N_{u}

and

N^{'}

the length of

r^{'} (γ_{y})

does not exceed the length of

γ_{y}

. In particular, it is less than

R

. Thus

d_{o_{i} (z)} (r^{'} (y)) < R

, i.e.,

r^{'} (y) \in B_{z, i} (R)

. Conversely, let

w \in B_{z, i} (R)

and let

γ_{w}

be a path in

N^{'}

joining

o_{i} (z)

and

w

with length less than

R

. By the covering homotopy theorem (see, e.g., [Hu, Chapter III]) there is a path

{\tilde{γ}}_{w} \subset N_{u}

that covers

γ_{w}

and joins

{\tilde{o}}_{i} (z)

with some point

\tilde{w}

such that

r^{'} (\tilde{w}) = w

. Moreover, by the definition, the length of

{\tilde{γ}}_{w}

is the same as the length of

γ_{w}

. In particular, it is less than

R

. Thus

\tilde{w} \in {\tilde{B}}_{\tilde{z}, i} (R)

. This shows that

r^{'} ({\tilde{B}}_{\tilde{z}, i} (R)) = B_{z, i} (R)

. In turn, the latter implies that

\begin{matrix} # B_{z, i} (R) \leq # {\tilde{B}}_{\tilde{z}, i} (R) . \end{matrix}

(4.7)

Next, let

A

be a finite set of generators of

π_{1} (N)

(recall that condition ( 1.1 ) implies that

π_{1} (N)

is finitely generated). By

d_{w}

we denote the word metric on

π_{1} (N)

with respect to

A

. Now, from compactness of every

\bar{U_{i}}

by the Švarc-Milnor lemma (see, e.g., [BH, p.140]) we obtain that there exists a constant

c

such that for any

z \in U_{i}

1 \leq i \leq l

, and

g, h \in π_{1} (N)

\begin{matrix} c^{- 1} d_{w} (g, h) \leq \tilde{d} ((z, g), (z, h)) \leq c d_{w} (g, h) \end{matrix}

(4.8)

(Here we identify

{\tilde{r}}^{- 1} (U_{i})

with

U_{i} \times π_{1} (N)

as in Example 3.2 (a).) Let

B_{R} \subset π_{1} (N)

be the open ball of radius

R

centered at

1

with respect to

d_{w}

. Then there is a natural number

\tilde{k}

such that

\begin{matrix} # B_{R} \leq e^{\tilde{k} R} f o r a n y R \geq 0 . \end{matrix}

(4.9)

From here, ( 4.8 ) and ( 4.7 ) we get for

k : = \tilde{k} c

# B_{z, i} (R) \leq e^{k R}, 1 \leq i \leq l . □

We proceed with the proof of the corollary. Let us define

α : = \frac{k + 1}{2}

and prove that

F = e^{- α f} \in ℋ_{2, 1} (M^{'}) \cap C (\bar{M^{'}})

. Since

ℋ_{2, 1} (M^{'}) ↪ H_{1}^{2} (M^{'})

(see section 2.3), this implies the required statement.

Let

z \in U_{i}

for some

1 \leq i \leq l

. We will estimate

| F |_{2, 1, z}

(see ( 2.2 )). By the definition using inequalities ( 4.5 ), ( 4.6 ) and Lemma 4.1 we obtain

\begin{matrix} | F |_{2, 1, z}^{2} = \sum_{y \in r^{- 1} (z)} | e^{- α f (y)} |^{2} \leq e^{2 α (a + C)} \cdot \sum_{y \in r^{- 1} (z)} e^{- 2 α d_{o_{i} (z)} (y)} \leq \end{matrix}

\begin{matrix}  \end{matrix}

\begin{matrix} e^{2 α (a + C)} \cdot \sum_{R = 0}^{\infty} e^{- 2 α R} \cdot # B_{z, i} (R) \leq e^{2 α (a + C)} \cdot \sum_{R = 0}^{\infty} e^{(- 2 α + k) R} = \frac{e^{2 α (a + C) + 1}}{e - 1} . \end{matrix}

Therefore

| F |_{2, 1}^{M} : = {sup}_{z \in M} | F |_{2, 1, z} \leq {(\frac{e^{2 α (a + C) + 1}}{e - 1})}^{1 / 2} .

This shows that

F \in ℋ_{2, 1} (M^{'}) \cap C (\bar{M^{'}})

□

Remark 4.2 (1) Using some construction from [Br2] one can prove that the constant

C

in Theorem 2.1 for

φ = d_{o}

(see ( 4.5 )) can be chosen independent of the covering

r : M^{'} \to M

. It depends only on

M

\tilde{M}

and the Riemannian metric on

N

(2) Consider the holomorphic map

f : M^{'} \to C

with

f

as in Corollary 2.2 . Then

f (M^{'}) \subset S : = {z \in C : | I m z | < C, - C < R e z < \infty}

where

C

is the constant in Theorem 2.1 for

φ = d_{o}

. Let

B_{t} = {x \in M^{'} : d_{o} (x) < t}

be the open ball in

M^{'}

centered at

o

of radius

t

and

S_{R} : = {z \in S : R e z \geq R}

Then

f^{- 1} (S_{R}) \subset M^{'} \ B_{R - C} f o r R > C a n d f^{- 1} (S \ S_{R}) \subset B_{R + C} .

Using such

f

one can construct holomorphic functions on

M^{'}

decreasing faster than the function

F

from Corollary 2.2 . Actually, let

l : R_{+} \to R_{+}

be a continuous function monotonically increasing for

x \geq R_{0}

. Consider a holomorphic function

g

S

satisfying

\begin{matrix} log | g (x + i y) | \geq l (x) f o r x \geq R_{0} a n d {inf}_{z \in S} | g (z) | > 0 . \end{matrix}

(4.10)

Then one can easily check that the function

G = g \circ f \in O (M^{'}) \cap C (\bar{M^{'}})

satisfies

| G (z) | \geq e^{l (d_{o} (z) - C)} f o r d_{o} (z) \geq R_{0} + C .

In particular,

H = 1 / G \in O (M^{'}) \cap C (\bar{M^{'}})

satisfies (for some

c_{1} > 0

)

\begin{matrix} | H (z) | \leq c_{1} e^{- l (d_{o} (z) - C)}, z \in \bar{M^{'}} \ B_{C} . \end{matrix}

(4.11)

Observe that the Harnack inequality for positive harmonic functions implies for

g

as in ( 4.10 ) (for some positive

{\tilde{c}}_{1}

{\tilde{c}}_{2}

)

l (x) \leq log | g (x) | \leq {\tilde{c}}_{1} e^{{\tilde{c}}_{2} x}, x \geq R_{0} .

This and the properties of

f

impose the following restriction on the decay of

H

\begin{matrix} | H (z) | \geq e^{- {\tilde{c}}_{3} e^{{\tilde{c}}_{2} d_{o} (z)}}, z \in M^{'} . \end{matrix}

(4.12)

Example 4.3 As the function

g

in ( 4.10 ) one can take, e.g.,

g (z) = e^{z^{n}}

for

n \in N

(in this case

l (x) = (1 - ε) x^{n}

for any

ε > 0

), or

g (z) = e^{C_{1} e^{C_{2} z}}

for

C_{1} > 0

and

0 < C_{2} < \frac{π}{2 C}

with

C

as above (in this case

l (x) = C_{1} cos (C_{2} C) e^{C_{2} x}

). For the latter example estimate ( 4.11 ) shows that the lower bound ( 4.12 ) of the decay of

H

is attainable.

5 . Proof of Theorem 2.4 .

Suppose that

M \subset \subset N

are domains in a Stein manifold,

M

is strictly pseudoconvex and

π_{1} (M) = π_{1} (N)

. Using the Remmert embedding theorem (see, e.g., [GR]) we may assume without loss of generality that

N

is a domain in a closed complex submanifold of some

C^{k}

. Let

r : N^{'} \to N

be an unbranched covering of

N

. As usual, we set

M^{'} = r^{- 1} (M)

and

b M^{'} = r^{- 1} (b M)

where

b M

is the boundary of

M

We must show that every point in

b M^{'}

is a peak point for

ℋ_{p, 1} (M^{'})

1 \leq p \leq \infty

. In our proof we use a result on uniform estimates for solutions of certain

\bar{\partial}

-equations on

M^{'}

. To its formulation we first introduce the corresponding class of

(0, 1)

-forms on

M^{'}

Let

{V_{i}}_{i \in I}

be a finite acyclic open cover of

\bar{M}

by relatively compact complex coordinate systems. We set

U_{i} \cap M

and consider the open cover

U = {U_{i}}_{i \in I}

M

Let

X_{H}

be the fibre of

N^{'}

with

H = π_{1} (N^{'})

. Using the construction of Example 3.2 (a) we identify

r^{- 1} (U_{i})

with

U_{i} \times X_{H}

. Let

ω

be a

(0, 1)

-form on

M^{'}

. Then in local coordinates on

r^{- 1} (U_{i})

it is presented as

ω (v, x) = \sum_{j = 1}^{n} a_{j} (v, x) d {\bar{v}}_{j} f o r v \times x \in U_{i} \times X_{H}

where

v = (v_{1}, . . ., v_{n})

are coordinates on

U_{i}

. Consider every

a_{j}

as a function on

U_{i}

with values in the space of functions on

X_{H}

. We assume that for every

i \in I

\begin{matrix} a_{j} \in C^{\infty} (U_{i}, l_{p, 1} (X_{H})), 1 \leq j \leq n . \end{matrix}

(5.1)

Then for such an

ω

its direct image

r_{*} (ω)

is a bounded

C^{\infty}

form with values in the Banach vector bundle

E_{p, 1} (X_{H})

(see section 3.1). Also, we assume that the norm of

ω

defined by the formula

\begin{matrix} | | ω | | : = {sup}_{i \in I} {max}_{1 \leq j \leq n} | a_{j} |_{p, 1}^{U_{i}} \end{matrix}

(5.2)

is finite. (Recall that

| \cdot |_{p, 1}^{U_{i}}

are norms on

C_{p, 1} (r^{- 1} (U_{i}))

, see ( 3.5 ).)

Proposition 5.1 There is a constant

C > 0

and for each

\bar{\partial}

-closed

(0, 1)

-form

ω

satisfying ( 5.1 ) there is a function

f \in C^{\infty} (M^{'}) \cap C_{p, 1}^{b} (M^{'})

such that

\bar{\partial} f = ω a n d | f |_{p, 1}^{M} \leq C | | ω | | .

Proof. We apply the operators

L_{z}

from Theorem 3.4 to

ω

. Namely, let us define a form

\tilde{ω}

M

by the formula

\tilde{ω} (v, z) : = \sum_{j = 1}^{n} (L_{v} a_{j} (v, \cdot)) (z) d {\bar{v}}_{j} f o r v \times z \in U_{i} \times M^{'}, i \in I .

It is readily seen that

\tilde{ω}

is a bounded

\bar{\partial}

-closed

C^{\infty}

form on

M

with values in

ℋ_{p, 1} (M^{'})

. We define the norm of

\tilde{ω}

\begin{matrix} | | \tilde{ω} | | : = {sup}_{i \in I, v \in U_{i}} {max}_{1 \leq j \leq n} | L_{v} a_{j} (v, \cdot) |_{p, 1}^{M} \end{matrix}

(5.3)

where

| \cdot |_{p, 1}^{M}

is norm on

ℋ_{p, 1} (M^{'})

. Then according to Theorem 3.4 there is a constant

c

(independent of

ω

) such that

\begin{matrix} | | \tilde{ω} | | \leq c | | ω | | . \end{matrix}

(5.4)

Further, we use Lemma 1 from [He]. According to this lemma there exist a strictly pseudoconvex domain

W \subset C^{k}

with

C^{2}

boundary such that

W \cap N = M

and a holomorphic map

π

from a neigbourhood

U (\bar{W})

\bar{W}

onto

U (\bar{W}) \cap N

such that

π (W) = M

and

π |_{U (\bar{W}) \cap N}

is the identity map.

Using this result we obtain that the pullback

π^{*} \tilde{ω}

with respect to

π

is a bounded

\bar{\partial}

-closed

C^{\infty}

form on

W

with values in

ℋ_{p, 1} (M^{'})

. Moreover, there is a constant

c^{'}

(depending on

π

and

W

) such that

\begin{matrix} | | π^{*} \tilde{ω} | | \leq c^{'} | | \tilde{ω} | | . \end{matrix}

(5.5)

Here for

π^{*} \tilde{ω} (w, \cdot) = \sum_{j = 1}^{k} {\tilde{a}}_{j} (w, \cdot) d {\bar{w}}_{j}

w = (w_{1}, . . ., w_{k}) \in C^{k}

, we define

| | π^{*} \tilde{ω} | | : = {sup}_{w \in W} {max}_{1 \leq j \leq k} | {\tilde{a}}_{j} (w, \cdot) |_{p, 1}^{M} .

In [SH] uniform estimates for solutions of

\bar{\partial}

-equations on so-called pseudoconvex polyhedra were obtained by means of global integral formulas. This class contains, in particular, strictly pseudoconvex domains with

C^{2}

boundaries. Note that the estimates in [SH] remain valid if one solves Banach-valued

\bar{\partial}

-equations. Therefore from the results of [SH] we obtain that there exists a bounded

C^{\infty}

function

h

W

with values in

ℋ_{p, 1} (M^{'})

such that

\bar{\partial} h = π^{*} \tilde{ω}

. Moreover,

\begin{matrix} | | h | | \leq c^{''} | | π^{*} \tilde{ω} | | \end{matrix}

(5.6)

for some

c^{''}

(depending on

W

only). Here

| | h | | : = {sup}_{w \in W} | h (w, \cdot) |_{p, 1}^{M} .

Finally, define a function

f

M^{'}

by the formula

f (z) : = h (r (z), z), z \in M^{'} .

Using that

r

is holomorphic,

\tilde{ω} (r (z), z) = ω (z)

z \in M^{'}

, and

(π^{*} \tilde{ω}) |_{M} = \tilde{ω}

we easily conclude that

\bar{\partial} f = ω

. By the definition

f \in C^{\infty} (M^{'}) \cap C_{p, 1}^{b} (M^{'})

and from ( 5.4 )-( 5.6 ) we have (for some

C

)

| f |_{p, 1}^{M} \leq C | | ω | | . □

Remark 5.2 (1) An analogous to Proposition 5.1 statement is valid for a similar class of bounded

\bar{\partial}

-closed

(0, q)

-forms on

M^{'}

(2) Using the main result of [He] and the estimates from [SH] one can show that the result of Proposition 5.1 is valid also for coverings of non-degenerate pseudoconvex polyhedrons on Stein manifolds (see [He] and [SH] for the definition).

We pass to the proof of Theorem 2.4 . Take a point

z \in b M^{'}

and set

v = r (z) \in b M

. Let

U \subset \subset N

be a simply connected coordinate neighbourhood of

v

and let

W \subset N^{'}

be the neighbourhood of

z

such that

r : W \to U

is biholomorphic. Since

M

is strictly pseudoconvex,

v

is a peak point for

O (U \cap M)

for a sufficiently small

U

Moreover, for such

U

we can find

\tilde{f} \in O (U \cap M)

with a peak point at

v

such that

\tilde{f} \in L^{q} (U \cap M)

for all

1 \leq q < \infty

(see [GHS, p.575]). Then

f : = (r^{*} \tilde{f}) |_{W \cap M^{'}}

has a peak point at

z

and

f \in L^{q} (W \cap M^{'})

for all

1 \leq q < \infty

. Next, let

\tilde{ρ} \in C^{\infty} (U)

be a cut-off function which equals 1 in a neighbourhood

O \subset \subset U

v

and 0 outside

U

Consider its pullback

ρ : = (r^{*} \tilde{ρ}) |_{W} \in C^{\infty} (W)

. Clearly the

(0, 1)

-form

ω = \bar{\partial} (ρ f)

M^{'}

satisfies conditions of Proposition 5.1 . Then this proposition implies that there exists a function

h \in C^{\infty} (M^{'}) \cap C_{p, 1}^{b} (M^{'})

such that

\bar{\partial} h = ω

. Finally, consider the function

h_{z} : = ρ f - h

. Then

h_{z}

is holomorphic, has a peak point at

z_{0}

and belongs to

H_{1}^{p} (M^{'})

for

1 \leq p < \infty

by the choice of

f

. Also, for

p = \infty

the function

h_{z}

is bounded outside

W

The proof of the theorem is complete.

□

6 . Proofs of Theorem 2.7 and Corollary 2.9 .

Proof of Theorem 2.7 . Let

M \subset \subset \tilde{M} \subset N

be manifolds satisfying condition ( 1.1 ) with

d i m_{C} M \geq 2

. Let

D \subset \subset M

be an open subset whose boundary

b D

is a connected

C^{k}

submanifold of

M

(1 \leq k \leq \infty)

. Consider a covering

r : N^{'} \to N

and set

M^{'} = r^{- 1} (M)

D^{'} = r^{- 1} (D)

and

b D^{'} = r^{- 1} (b D)

. Let

ψ : N^{'} \to R_{+}

be such that

log ψ

is uniformly continuous with respect to the path metric induced by a Riemannian metric pulled back from

N

. Let

f \in C^{s} (b D^{'})

0 \leq s \leq k

, be a

C R

-function satisfying the hypotheses of Theorem 2.7 .

(A) First, we will prove the theorem for

s = 0

under the additional assumption

\begin{matrix} f \in C_{p, ψ} (b D^{'}) f o r p \in [1, \infty) a n d f \in C_{1, 1} (b D^{'}) f o r p = \infty . \end{matrix}

(6.1)

(We use here that

C_{1, 1} (b D^{'}) \subset C_{\infty, 1} (b D^{'}) = C_{\infty, ψ} (b D^{'})

, see Definition 3.3 .) For a

C R

-function

f

satisfying ( 6.1 ) we define a continuous

ℋ_{p, ψ} (M^{'})

-valued function

F

b D^{'}

by the formula

F (v) : = L_{v} (f |_{r^{- 1} (v)}), v \in b D,

where

L_{v}

are operators from Theorem 3.4 , see section 3.3.

Lemma 6.1

F

is a

ℋ_{p, ψ} (M^{'})

-valued continuous

C R

-function.

Proof. Let

U \subset M

be a simply connected coordinate neighbourhood of a point

x \in b D

. It suffices to check that

F |_{U \cap b D}

satisfies the required property. Note that by Proposition 3.5

\begin{matrix} [F (v)] (z) = \sum_{y \in r^{- 1} (x)} {\tilde{f}}_{y} (v) H_{y} (v, z), v \times z \in (U \cap b D) \times M^{'}, \end{matrix}

(6.2)

where

{\tilde{f}}_{y} (v) = f (s_{y} (v))

v \in U \cap b D

, and

s_{y} : U \to V_{y}

is a biholomorphic map onto the connected component

V_{y}

r^{- 1} (U)

containing

y

. By the definition of operators

L_{v}

functions

H_{y}

are restrictions to

b D \times M^{'}

of some holomorphic functions on

U \times M^{'}

Moreover, Proposition 3.5 implies that the series in ( 6.2 ) converges uniformly to

F

on every compact subset of

(U \cap b D) \times M^{'}

. Next, since

f |_{V_{y} \cap b D^{'}}

is a continuous

C R

-function,

{\tilde{f}}_{y}

is a continuous

C R

-function on

U \cap b D

. Also, by the definition of

H_{y}

, for a fixed

z \in M^{'}

every

H_{y} (\cdot, z)

is a continuous

C R

-function on

U \cap b D

Hence,

{\tilde{f}}_{y} \cdot H_{y} (\cdot, z)

is a continuous

C R

-function on

U \cap b D

, as well. Indeed, for each

(n, n - 2)

-form

ω

with a compact support in

U

we have

\int_{U \cap b D} {\tilde{f}}_{y} (v) \cdot H_{y} (v, z) \bar{\partial} ω (v) = \int_{U \cap b D} {\tilde{f}}_{y} (v) \bar{\partial} (H_{y} (v, z) \cdot ω (v)) = 0

because

{\tilde{f}}_{y}

C R

. Since the series in ( 6.2 ) converges uniformly to

[F (\cdot)] (z)

on every compact subset of

(U \cap b D) \times z

, every

[F (\cdot)] (z)

z \in M^{'}

, is a continuous

C R

-function on

U \cap b D

. This implies the required statement.

□

Further, since

[F (v)] (z)

from Lemma 6.1 is holomorphic in

z \in M^{'}

, we can expand it in the Taylor series in a complex coordinate neighbourhood

U_{z}

\begin{matrix} [F (v)] (w) = \sum_{0 \leq | α | < \infty} F_{α} (v) w^{α}, v \in b D . \end{matrix}

(6.3)

Here

α = (α_{1}, . . ., α_{s}) \in (Z_{+})^{s}

| α | = \sum_{i = 1}^{s} α_{i}

w^{α} = w_{1}^{α_{1}} . . . w_{n}^{α_{n}}

and

w = (w_{1}, . . ., w_{n})

are coordinates on

U_{z}

such that

w (z) = 0

. Now, from Lemma 6.1 follows that each

F_{α}

in ( 6.3 ) is a continuous

C R

-function on

b D

. Then by Theorem 3.14 of Harvey [Ha] for every

F_{α}

there exists a function

{\tilde{F}}_{α} \in O (D) \cap C (\bar{D})

such that

{\tilde{F}}_{α} |_{b D} = F_{α}

. Also, for a sufficiently small

U_{z}

using estimates of the Cauchy integrals for derivatives of a holomorphic function and compactness of

b D

we get from ( 6.3 )

M : = {sup}_{α, v \in b D} | F_{α} (v) | < \infty .

Thus by the maximum modulus principle

{sup}_{α, v \in \bar{D}} | {\tilde{F}}_{α} (v) | = M < \infty .

The latter implies that for a sufficiently small

U_{z}

the series

{\tilde{F}}_{z} (v, w) = \sum_{0 \leq | α | < \infty} {\tilde{F}}_{α} (v) w^{α}, v \times w \in \bar{D} \times U_{z},

converges absolutely and uniformly. Hence,

{\tilde{F}}_{z} \in O (D \times U_{z}) \cap C (\bar{D} \times U_{z})

. Further, assume that for

y, z \in M^{'}

we have

U_{y} \cap U_{z} \neq \emptyset

. Then for every

w \in U_{y} \cap U_{z}

and

v \in b D

{\tilde{F}}_{y} (v, w) - {\tilde{F}}_{z} (v, w) = [F (v)] (w) - [F (v)] (w) = 0 .

This leads to the identity

{\tilde{F}}_{y} (\cdot, w) = {\tilde{F}}_{z} (\cdot, w), w \in U_{y} \cap U_{z} .

Thus we can define a function

F \in O (D \times M^{'}) \cap C (\bar{D} \times M^{'})

by the formula

\begin{matrix} \tilde{F} (v, w) : = {\tilde{F}}_{z} (v, w), v \times w \in \bar{D} \times U_{z} . \end{matrix}

(6.4)

Lemma 6.2

\tilde{F} (v, \cdot) \in ℋ_{p, ψ} (M^{'})

for any

v \in D

Proof. Observe that the evaluation at

v \in D

is a linear continuous functional on the Banach space

O (D) \cap C (\bar{D})

equipped with supremum norm. Identifying

O (D) \cap C (\bar{D})

with its trace space on

b D

and using the Hahn-Banach and F. Riesz theorems we have

h (v) = \int_{b D} h (ξ) d μ (ξ), h \in O (D) \cap C (\bar{D}),

where

μ

is a complex regular Borel measure on

b D

with the total variation

V a r μ = 1

. Thus for every fixed

w \in M^{'}

we have

\tilde{F} (v, w) = \int_{b D} [F (ξ)] (w) d μ (ξ) .

Now, by the definition of the norm on

ℋ_{p, ψ} (M^{'})

using the triangle inequality, the identity

\tilde{F} (v, \cdot) = F (v)

v \in b D

, and the fact that

F

is a continuous

ℋ_{p, ψ} (M^{'})

-valued function on

b D

we obtain

\begin{matrix} | \tilde{F} (v, \cdot) |_{p, ψ}^{M} : = {sup}_{z \in M} {(\sum_{y \in r^{- 1} (z)} | \tilde{F} (v, y) |^{p} ψ (y))}^{1 / p} \leq \end{matrix}

\begin{matrix}  \end{matrix}

\begin{matrix} {sup}_{z \in M} {(\sum_{y \in r^{- 1} (z)} {(\int_{b D} | \tilde{F} (ξ, y) | | d μ (ξ) |)}^{p} ψ (y))}^{1 / p} \leq \end{matrix}

\begin{matrix}  \end{matrix}

\begin{matrix} {sup}_{z \in M} (\int_{b D} {(\sum_{y \in r^{- 1} (z)} | [F (ξ)] (y) |^{p} ψ (y))}^{1 / p} | d μ (ξ) |) \leq {sup}_{ξ \in b D} | F (ξ) |_{p, ψ}^{M} < \infty . □ \end{matrix}

Further, set

\begin{matrix} f^{'} (z) : = \tilde{F} (r (z), z), z \in \bar{D^{'}} . \end{matrix}

(6.5)

Then using the inequalities of Lemma 6.2 we get

\begin{matrix} f^{'} \in O (D^{'}) \cap C (\bar{D^{'}}), f^{'} |_{b D^{'}} = F |_{b D^{'}} = f, a n d \end{matrix}

\begin{matrix}  \end{matrix}

\begin{matrix} | f^{'} |_{p, ψ, z} : = {(\sum_{y \in r^{- 1} (z)} | f^{'} (y) |^{p} ψ (y))}^{1 / p} \leq {sup}_{ξ \in b D} | F (ξ) |_{p, ψ}^{M} \leq c | f |_{p, ψ}^{b D}, z \in D, \end{matrix}

see ( 3.5 ) for the definition of

| \cdot |_{p, ψ}^{b D}

. (Here the last inequality follows directly from Theorem 3.4 .) The latter implies that

f^{'} |_{r^{- 1} (z)} \in l_{p, ψ, z} (M^{'})

z \in D

, see section 2.3.

Thus

f^{'} \in ℋ_{p, ψ} (D^{'}) \cap C (\bar{D^{'}})

and

\begin{matrix} | f^{'} |_{p, ψ}^{D} : = {sup}_{z \in D} | f^{'} |_{p, ψ, z} \leq c {sup}_{z \in b D} | f |_{p, ψ, z} (: = c | f |_{p, ψ}^{b D}) . \end{matrix}

(6.6)

This completes the proof of the theorem for

s = 0

under assumption ( 6.1 ).

(B) Let us consider the general case of a continuous

C R

-function

f

b D^{'}

satisfying

\begin{matrix} f |_{r^{- 1} (x)} \in l_{p, ψ, x} (M^{'}) f o r a n y x \in D a n d m : = {sup}_{x \in b D} | f |_{p, ψ, x} < \infty . \end{matrix}

(6.7)

According to Remark 4.2 (2) and Example 4.3 there is a constant

c > 0

such that for any

0 < c_{1} < c

and

c_{2} > 0

there exists a function

F_{c_{1}, c_{2}} \in O (M^{'}) \cap C (\bar{M^{'}})

satisfying

\begin{matrix} e^{- c_{3} e^{c_{1} d_{o} (z)}} \leq | F_{c_{1}, c_{2}} (z) | \leq e^{- c_{2} e^{c_{1} d_{o} (z)}} f o r a l l z \in \bar{M^{'}} \end{matrix}

(6.8)

with

c_{3}

depending on

c_{2}

c_{1}

c

M

M^{'}

such that

c_{3} \to 0

c_{2} \to 0

. (Recall that

d_{o}

o \in M^{'}

, is the distance on

N^{'}

defined as in Corollary 2.2 .) Define a continuous

C R

-function

f_{c_{1}, c_{2}}

by the formula

f_{c_{1}, c_{2}} (z) : = f (z) F_{c_{1}, c_{2}} (z), z \in b D^{'} .

Lemma 6.3

f_{c_{1}, c_{2}}

satisfies assumption ( 6.1 ).

Proof. Note that for any

l \in N

there is a nonnegative

r

such that

e^{- c_{2} e^{c_{1} d_{o} (z)}} < e^{- l d_{o} (z)} f o r d_{o} (z) > r, z \in \bar{M^{'}} .

Let

U \subset \subset M

be a neighbourhood of

\bar{D}

and

U^{'} = r^{- 1} (U) \subset M^{'}

. From the above inequality arguing as in the proof of Corollary 2.2 we obtain that

F_{c_{1}, c_{2}} \in ℋ_{p, 1} (U^{'})

for any

p \in [1, \infty]

. Then from [Br, Proposition 2.4] follows that

F_{c_{1}, c_{2}} |_{b D^{'}}

belongs to

C_{p, 1} (b D^{'})

for all

p

Next, take a point

x \in b D

and prove that

f_{c_{1}, c_{1}}

C_{p, ψ}

-continuous over

x

. Let

U_{x} \subset M

be a complex (simply connected) coordinate neighbourhood of

x

. We will identify

r^{- 1} (U_{x})

with

U_{x} \times X_{H}

where

X_{H}

is the fibre of

r : M^{'} \to M

. Consider a sequence

{x_{n}} \subset U_{x} \cap b D

convergent to

x

. For

g \in X_{H}

put

a_{n} (g) : = f (x_{n}, g)

a (g) : = f (x, g)

b_{n} (g) : = F_{c_{1}, c_{2}} (x_{n}, g)

b (g) : = F_{c_{1}, c_{2}} (x, g)

and

c_{n} : = a_{n} b_{n}

c : = a b

Then we must check that

{lim}_{n \to \infty} | c - c_{n} |_{p, ψ, x} : = {lim}_{n \to \infty} {(\sum_{g \in X_{H}} | c (g) - c_{n} (g) |^{p} ψ (x, g))}^{1 / p} = 0 .

Using the triangle inequality we have

| c - c_{n} |_{p, ψ, x} \leq | (a - a_{n}) b |_{p, ψ, x} + | a_{n} (b - b_{n}) |_{p, ψ, x} : = I + I I .

According to ( 6.8 ) for any

ε > 0

we can decompose

b

in the sum

b^{'} + b^{''}

where

b^{'} = 0

outside a finite subset

S_{ε} \subset X_{H}

and

b^{''} = 0

S_{ε}

such that

| b^{''} (g) | < ε

for all

g

. Note also that

| b^{'} (g) | \leq 1

for all

g

. Also, ( 6.7 ) and uniform continuity of

log ψ

on the compact set

b D

imply that

| a - a_{n} |_{p, ψ, x} \leq k m

for some

k > 0

. Finally, by continuity of

f

b D^{'}

we can find a number

N

such that for any

n \geq N

we have

| (a - a_{n}) χ_{ε} |_{p, ψ, x} < ε

, where

χ_{ε}

is the characteristic function of

S_{ε}

. Using all these facts we get for

n \geq N

I \leq | (a - a_{n}) χ_{ε} b^{'} |_{p, ψ, x} + | (a - a_{n}) b^{''} |_{p, ψ, x} \leq ε + k m ε = (1 + k m) ε .

To estimate

I I

observe that from ( 6.7 ) and uniform continuity of

log ψ

follow that for each

g \in X_{H}

(| a_{n} (g) |^{p} ψ (x, g))^{1 / p} \leq | a_{n} |_{p, ψ, x} \leq k^{'} | a_{n} |_{p, ψ, x_{n}} \leq k^{'} m

(for some

k^{'}

). Moreover, since

F_{c_{1}, c_{2}} |_{b D^{'}} \in C_{p, 1} (b D^{'})

, there is an integer

N^{'}

such that for any

n \geq N^{'}

we have

| b - b_{n} |_{p, 1, x} < ε

. These two inequalities yield for

n \geq N^{'}

I I \leq {sup}_{g \in X_{H}} (| a_{n} (g) |^{p} ψ (x, g))^{1 / p} | b - b_{n} |_{p, 1, x} \leq k^{'} m ε .

Combining the estimates for

I

and

I I

we obtain that

{lim}_{n \to \infty} | c - c_{n} |_{p, ψ, x} = 0 .

This is equivalent to

C_{p, ψ}

-continuity of

f_{c_{1}, c_{2}}

over

x

Similarly one can check that if

p = \infty

then

f_{c_{1}, c_{2}}

belongs to

C_{1, 1} (b D^{'})

. We leave it as an excercise to the readers.

□

Let us finish the proof of the theorem. According to Lemma 6.3 and the case (A) there is a function

f_{c_{1}, c_{2}}^{'} \in O (D^{'}) \cap C (\bar{D^{'}})

such that

f_{c_{1}, c_{2}}^{'} |_{b D^{'}} = f_{c_{1}, c_{2}}

. Note also that since

| F_{c_{1}, c_{2}} | \leq 1

inequality ( 6.6 ) yields

| f_{c_{1}, c_{2}}^{'} |_{p, ψ}^{D} \leq c {sup}_{z \in b D} | f |_{p, ψ, z} : = c m .

Consider

f^{'} : = f_{c_{1}, c_{2}} / F_{c_{1}, c_{2}}

. Then

f^{'} \in O (D^{'}) \cap C (\bar{D^{'}})

and

f^{'} |_{b D^{'}} = f

. The uniqueness property for holomorphic functions implies that

f^{'}

does not depend on

c_{1}

and

c_{2}

. Since

F_{c_{1}, c_{2}}

converges uniformly on compact subsets of

M^{'}

1

c_{2} \to 0

from the last inequality we get

| f^{'} |_{p, ψ}^{D} \leq c {sup}_{z \in b D} | f |_{p, ψ, z} .

Therefore

f^{'} \in ℋ_{p, ψ} (D^{'}) \cap C (\bar{D^{'}})

The proof of the theorem for

s = 0

is complete. If, in addition,

f \in C^{s} (b D)

for

1 \leq s \leq k

, then in fact the extended function

f^{'} \in C^{s} (\bar{D^{'}})

(see, e.g., Theorem 3.14 in [Ha], and the discussion that follows it).

□

Proof of Corollary 2.9 . Let us consider the function

F_{c_{1}, c_{2}}

from ( 6.8 ). Suppose that

f \in O (U^{'})

satisfies the hypotheses of Corollary 2.9 with

c_{1}

c_{2}

and

c

as in the definition of

F_{c_{1}, c_{2}}

. Then

f_{c_{1}, c_{2}} : = f F_{c_{1}, c_{2}} \in H^{\infty} (U^{'})

with the norm bounded by 1. Further, by the hypotheses we can find a connected

C^{\infty}

compact submanifold

b S \subset U

that bounds a domain

S

containing

K

. By Theorem 2.7 (applied to

S^{'} : = r^{- 1} (S)

and

b S^{'} : = r^{- 1} (b S)

) the function

f_{c_{1}, c_{2}}

admits an extension

f_{c_{1}, c_{2}}^{'} \in H^{\infty} (D^{'})

. Since

f_{c_{1}, c_{2}}^{'} = f_{c_{1}, c_{2}}

U^{'}

and

h (z) : = {sup}_{y \in r^{- 1} (z)} | f_{c_{1}, c_{2}}^{'} (y) |

is a continuous plurisubharmonic function on

D

| f_{c_{1}, c_{2}}^{'} |_{\infty}^{D} = {sup}_{z \in U^{'}} | f (z) | \leq 1 .

Then the function

f^{'} : = f_{c_{1}, c_{2}}^{'} / F_{c_{1}, c_{2}} \in O (D^{'})

extends

f

and satisfies

| f^{'} (z) | \leq e^{c_{3} e^{c_{1} d_{o} (z)}}, z \in D^{'},

with

c_{3}

as in ( 6.8 ).

□

7 . Proof of Theorem 2.10 .

Let

r : M_{G} \to M

be the regular covering of

M

satisfying condition ( 1.1 ) (for some

\tilde{M}

and

N

) with transformation group

G

. Let

G_{1} \subset G

be a subgroup of a finite index and let

r_{1} : M_{1} \to M

be the covering with fibre

G / G_{1}

. Then there are coverings

N_{1}

and

{\tilde{M}}_{1}

N

and

\tilde{M}

with fibre

G / G_{1}

such that

M_{1} \subset \subset {\tilde{M}}_{1} \subset N_{1}

Clearly this triple satisfies condition ( 1.1 ), as well. Thus without loss of generality we may assume that

M : = M_{1}

G : = G_{1}

\tilde{M} : = {\tilde{M}}_{1}

and

N : = N_{1}

, and so

G \in T (M)

The latter means that

G

admits a linear representation

ρ

into

G L_{k} (C)

and that the flat vector bundle

E_{ρ}

N

associated with

ρ

is topologically trivial. Then the restriction

E_{ρ} |_{\tilde{M}}

is topologically trivial. Since

\tilde{M}

is Stein, according to the Oka-Grauert principle (see [G2]),

E_{ρ} |_{\tilde{M}}

is holomorphically trivial. In particular,

ρ

can be obtained as the monodromy of an equation

d F = ω F

\tilde{M}

where

ω

is a matrix-valued holomorphic 1-form on

\tilde{M}

satisfying

d ω - ω \land ω = 0

. Let

\tilde{ω} = r^{*} ω

be the pullback of

ω

{\tilde{M}}_{G} = r^{- 1} (\tilde{M})

. (Here

r : N_{G} \to N

is the regular covering of

N

with the transformation group

G

so that

M_{G}

and

{\tilde{M}}_{G}

are domains in

N_{G}

.) Then there exists a function

\tilde{F} \in O ({\tilde{M}}_{G}, G L_{k} (C))

such that

d \tilde{F} = \tilde{ω} \tilde{F}

. This follows from the fact that the monodromy of the last equation is the restriction of

ρ

π_{1} (M_{G})

and so it is trivial (since

π_{1} (M_{G}) \subset K e r ρ

). Note that

\tilde{F}

can be obtained by Picard iteration applied to

\tilde{ω}

. Since

\bar{M}

is a compact subset of

\tilde{M}

(and so

ω |_{M}

is bounded), the Picard iteration produces for some positive

c = c (M, ω)

the estimate

\begin{matrix} | | \tilde{F} (z) | |_{2} \leq e^{c d_{o} (z)}, z \in M_{G}, \end{matrix}

(7.1)

where

| | \cdot | |_{2}

is the

l_{2}

-norm on

G L_{k} (C)

and

o \in M_{G}

. (Here as before

d_{o}

is the distance from

o

in the path metric induced by a Riemannian metric pulled back from

N

.) Moreover, for every

z \in {\tilde{M}}_{G}

there exists a matrix

C_{z} \in G L_{k} (C)

such that

\tilde{F} (g z) = C_{z}^{- 1} ρ (g) C_{z}

for any

g \in G

. These are standard facts of the theory of flat connections. In particular, from the last identity we derive easily that

\tilde{F}

separates points in every orbit of the action of

G

{\bar{M}}_{G}

Next, let

f

be the function from Corollary 2.2 . Then by ( 7.1 ) we get (for some

c_{1} = c_{1} (f, α)

)

| | e^{- α f (z)} \tilde{F} (z) | |_{2} \leq c_{1} e^{(c - α) d_{o} (z)}, z \in M_{G} .

From here arguing as in the proof of Corollary 2.2 we deduce that for a sufficiently big

α

all entries of the matrix

e^{- α f} \cdot \tilde{F}

belong to

ℋ_{2, 1} (M_{G}) \cap C ({\bar{M}}_{G})

. Now the family consisting of these entries and the function

e^{- α f}

separate all points in any orbit of the action of

G

{\bar{M}}_{G}

. For otherwise, there are

x, y \in {\bar{M}}_{G}

y = g x

g \neq 1

g \in G

, such that

e^{- α f (x)} \tilde{F} (x) = e^{- α f (y)} \tilde{F} (y)

and

e^{- α f (x)} = e^{- α f (y)}

. But this implies that

\tilde{F} (x) = \tilde{F} (y)

, a contradiction. Finally, since

M \subset \subset \tilde{M}

and

\tilde{M}

is Stein, by the Remmert embedding theorem there are holomorphic functions

h_{1}, . . ., h_{l}

from

H^{\infty} (M) \cap C (\bar{M})

that separate all points in

\bar{M}

. We set

{\tilde{h}}_{i} = e^{- α f} r^{*} h_{i}

1 \leq i \leq l

Then by the definition

{\tilde{h}}_{i} \in ℋ_{2, 1} (M_{G}) \cap C ({\bar{M}}_{G})

and so the family consisting of all

{\tilde{h}}_{i}

, entries of

e^{- α f} \tilde{F}

and

e^{- α f}

separates all points in

{\bar{M}}_{G}

The proof of Theorem 2.10 is complete.

□

References

L. Bungart, On analytic fibre bundles I. Holomorphic fibre bundles with infinite dimensional fibres. Topology, 7 (1) (1968), 55-68.
M. R. Bridson and A. Haefliger, Metric spaces of non-positive curvature. Springer, 1999.
S. Bochner, Analytic and meromorphic continuation by means of Green's formula. Ann. of Math., 44 (1943), 652-673.
A. Brudnyi, Integral representations of holomorphic functions on coverings of pseudoconvex domains in Stein manifolds. Preprint (2005), 18 pp.
A. Brudnyi, Projections in the space $H^{\infty}$ and the corona theorem for subdomains of coverings of finite bordered Riemann surfaces. Ark. Mat., 42 (2004), 31-59.
H. Cartan, Sur les fonctions de plusieurs variables complexes. Les espaces analytiques. Proc. Intern. Congress Mathematicians Edinbourgh 1958, Cambridge Univ. Press, 1960, 33-52.
H. Grauert, On Levi's problem and the imbedding of real-analytic manifolds. Ann. of Math., 68 (1958), 460-472.
H. Grauert, Analytische Faserungen über holomorph-vollständingen Räumen. Math. Ann., 135, no. 3 (1958), 263-273.
M. Gromov, G. Henkin and M. Shubin, Holomorphic $L^{2}$ functions on coverings of pseudoconvex manifolds. GAFA, Vol. 8 (1998), 552-585.
H. Grauert and R. Remmert, Theorie der Steinschen Räume. Springer, New York, 1977.
R. Harvey, Holomorphic chains and their boundaries. Several complex variables (Proc. Sympos. Pure Math., Vol. XXX, Part 1, Williams Coll., Williamstown, Mass., 1975), 309-382. Amer. Math. Soc., Providence, R. I., 1977.
D. Heunemann, Uniform estimates for the $\bar{\partial}$ -equation on pseudoconvex polyhedra on Stein manifolds. Math. Nachr., 114 (1983), 191-196.
F. Hirzebruch, Topological methods in Algebraic Geometry. Springer-Verlag, New York, 1966.
S.-T. Hu, Homotopy theory. Academic Press, new York, 1959.
J. J. Kohn and H. Rossi, On the extension of holomorphic functions from the boundary of a complex manifold. Ann. of Math. (2), 81 (1965), 451-472.
A. L. Onishchik, Certain concepts and applications of non-Abelian cohomology theory. Trudy Moskov. Mat. Obshch., 17 (1967), 45-88.
M. S. Raghunathan, Discrete subgroups of Lie groups. Ergebnisse der Mathematik und ihrer Grenzgebiete, vol. 68, Springer Verlag, New York, 1972.
R. Remmert, Sur les espaces analytiques holomorphiquement séparables et holomorphiquement convexes. C. R. Acad. Sci. Paris, 243 (1956), 118-121.
A. G. Sergeev and G. M. Henkin, Uniform estimates of the solutions of the $\bar{\partial}$ -equation in pseudoconvex polyhedra. (Russian) Mat. Sb. (N.S.) 112 (154) (1980), no. 4 (8), 522-567.