Research partially supported by NSF grant DMV-0400909.
<ph f="ptmb7t">Interpolation and Sampling Hypersurfaces for weighted Bergman spaces on the unit ball</ph>

### Dror Varolin

center Department of Mathematics, University of Illinois, Urbana, IL 61801, forgacs@uiuc.edu dror@uiuc.edu

1 Introduction

Recall that the Bergman metric on the unit ball $B=\left\{z\in {\mathbb{C}}^{n};|z{|}^{2}<1\right\}$  is the Kähler metric whose associated $\left(1,1\right)$  -form is ${\omega }_{B}=-\left(n+1\right)d{d}^{c}\lambda ,$  where $\lambda =log\left(1-|z{|}^{2}\right)-\frac{n}{n+1}log\left(n+1\right)$  and in our convention ${d}^{c}=\frac{\sqrt{-1}}{2}\left(\overline{\partial }-\partial \right).$  The weighted Bergman spaces on the Bergman ball are ${\mathsf{ℌ}}^{2}\left(B,\kappa \right):=\left\{F\in \mathfrak{O}\left(B\right);{\int }_{B}|F{|}^{2}{e}^{-\kappa }{\omega }_{B}^{n}<+\infty \right\}.$  In this paper we shall make the minor assumption that ${e}^{-\kappa }$  is ${\mathcal{C}}^{2}$  . (This assumption can be reduced by regularization to the case where ${e}^{-\kappa }$  and $\kappa$  are ${L}_{\ell oc}^{1}$  .) The case $\kappa =-\left(n+1\right)log\left(1-|z{|}^{2}\right)$  corresponds to the classical Bergman space of holomorphic functions that are square integrable with respect to Lebesgue measure.
Given a smooth closed hypersurface $W\subset B$  , we let ${ℌ}^{2}\left(W,\kappa \right):=\left\{f\in \mathfrak{O}\left(W\right);{\int }_{W}|f{|}^{2}{e}^{-\kappa }{\omega }_{B}^{n-1}<+\infty \right\}.$
Definition 1.1.
• (a) We say that $W$  is an interpolation hypersurface if for each $f\in {ℌ}^{2}\left(W,\kappa \right)$  there exists $F\in {\mathsf{ℌ}}^{2}\left(B,\kappa \right)$  such that $F|W=f$  .
• (b) We say that $W$  is a sampling hypersurface if there exists a constant $A>1$  such that for every $F\in {\mathsf{ℌ}}^{2}\left(B,\kappa \right)$  ,  $\begin{array}{ccc}\frac{1}{A}{\int }_{B}|F{|}^{2}{e}^{-\kappa }{\omega }_{B}^{n}\le {\int }_{W}|F{|}^{2}{e}^{-\kappa }{\omega }_{B}^{n-1}\le A{\int }_{B}|F{|}^{2}{e}^{-\kappa }{\omega }_{B}^{n}.& & \end{array}$ (1)
Let ${F}_{a}$  denote a holomorphic involution of $B$  sending $0$  to $a$  (see Section  2 ). We define the total density tensor of $W$  in the ball of radius $r$  to be the $\left(1,1\right)$  -form ${\Upsilon }_{r}^{W}\left(z\right)=\frac{1}{{V}_{n}\left(r\right)}\left({\int }_{B\left(0,r\right)}\frac{{\partial }^{2}log|T\left({F}_{z}\left(\zeta \right)\right){|}^{2}}{\partial {z}^{i}\partial {\overline{z}}^{j}}{\omega }_{B}^{n}\right)\sqrt{-1}d{z}^{i}\wedge d{\overline{z}}^{j}.$  Here ${V}_{n}\left(r\right)={\int }_{B\left(0,r\right)}{\omega }_{B}^{n}.$  Let us denote by $\left[W\right]$  the current of integration along $W$  , and by $\left[W{\right]}_{\varepsilon }$  the average (see Section  6 ) of $\left[W\right]$  over the Bergman-Green ball of radius $\varepsilon$  .
Definition 1.2.
• (I) Let ${\mathfrak{P}}_{W}\left(B\right)$  denote the set of $\left(n-1,n-1\right)$  -forms $\theta$  on $B$  with the following properties.
• (a) $\theta \wedge {\omega }_{B}\ge c{e}^{n\lambda }{\omega }_{B}^{n}$  for some constant $c>0$  .
• (b) For each $\varepsilon >0$  there exists $C>0$  such that $\left[W{\right]}_{\varepsilon }\wedge \theta \le C\left[W{\right]}_{\varepsilon }\wedge {\omega }_{B}^{n-1}$  .
• (c) $d{d}^{c}\theta =0$  .
• (II) For $\theta \in {\mathfrak{P}}_{W}\left(B\right)$  , let ${\mathfrak{D}}_{B}^{+}\left(W,\kappa \right)\left[\theta \right]={limsup}_{r\to 1}{sup}_{z\in B}\frac{\left({\Upsilon }_{r}^{W}+\frac{n}{n+1}{\omega }_{B}\right)\wedge \theta }{\sqrt{-1}\partial \overline{\partial }\kappa \wedge \theta }$  and ${\mathfrak{D}}_{B}^{-}\left(W,\kappa \right)\left[\theta \right]={liminf}_{r\to 1}{inf}_{z\in B}\frac{\left({\Upsilon }_{r}^{W}+\frac{n}{n+1}{\omega }_{B}\right)\wedge \theta }{\sqrt{-1}\partial \overline{\partial }\kappa \wedge \theta }$
• (III) The upper and lower densities of $W$  are ${\mathfrak{D}}_{B}^{+}\left(W,\kappa \right)={sup}_{\theta \in {\mathfrak{P}}_{W}\left(B\right)}{\mathfrak{D}}_{B}^{+}\left(W,\kappa \right)\left[\theta \right]$  and ${\mathfrak{D}}_{B}^{-}\left(W,\kappa \right)={sup}_{\theta \in {\mathfrak{P}}_{W}\left(B\right)}{\mathfrak{D}}_{B}^{-}\left(W,\kappa \right)\left[\theta \right]$
From here on out we assume that $W$  is uniformly flat (see Section  4 ) and that $0<\frac{1}{C}{\omega }_{B}\le \sqrt{-1}\partial \overline{\partial }\kappa \le C{\omega }_{B}.$  Our main results are the following two theorems.
Theorem 1. If ${\mathfrak{D}}_{B}^{+}\left(W,\kappa \right)<1$  , then $W$  is an interpolation hypersurface.
Theorem 2. If ${\mathfrak{D}}_{B}^{-}\left(W,\kappa \right)>1$  , then $W$  is a sampling hypersurface.
By now Theorems  1 and  2 carry with them a rich history. Most recently, results analogous to Theorems  1 and  2 have been established for the case of ${\mathbb{C}}^{n}$  in the paper [OSV, which we refer to for further historical remarks regarding interpolation and sampling problems for Bergman spaces.
There are similarities and differences between our approach and the approach of [OSV.
The most obvious difference comes about from the fact that the Bergman ball is negatively curved. Thus for interpolation, we need to solve a $\overline{\partial }$  -problem that requires weaker curvature conditions than those needed in Hörmander's Theorem. Such a $\overline{\partial }$  theorem was proved by Ohsawa. If one considers submanifolds $W$  having less than optimal density, then an adaptation of the method of Ohsawa-Takegoshi can be used to extend holomorphic functions from $W$  to the ball $B$  , and thus establish the interpolation property for $W$  . The method has the aesthetic advantage that a function is extended at once, instead of being extended locally and then patched back together. This approach is interesting in its own right. Surprisingly, we have not been able to apply the Ohsawa-Takegoshi approach to the case of optimal density conditions. Vaguely speaking, the issue can be described as follows. In order to do extension, the positivity required to patch together local solutions must be contained in the Ohsawa-Takegoshi procedure. On the other hand, the negativity of the Ricci curvature of the Bergman metric must also be compensated for by this procedure. It is precisely the combination of these two positivity contributions that we have been unable to realize with the right a priori (twisted) estimate. It would be interesting to see if the appropriate a priori estimate is possible.
By contrast with [OSV, our approach to sampling is closer in spirit to the technique that has been used in the one-variable case in [BO-95. We feel that this approach is more intuitive, as well as more elementary than the Beurling-inspired approach used in [OSV.
The paper is organized as follows.

Contents

Acknowledgment. We are grateful to Jeff McNeal and Quim Ortega-Cerdà for many stimulating discussions.

2 Rapid review of Bergman geometry

Bergman geometry is one of the oldest and most studied areas of complex geometry.
Therefore we content ourselves with stating facts, and provide few proofs.

Bergman metric

As already mentioned, the Bergman metric is ${\omega }_{B}=-\left(n+1\right)d{d}^{c}\lambda .$  It is easy to see that, with ${\omega }_{E}=d{d}^{c}|z{|}^{2}$  denoting the Euclidean metric, ${{\omega }_{B}|}_{z=0}=\left(n+1\right){{\omega }_{E}|}_{z=0}and{\omega }_{B}^{n}={e}^{-\left(n+1\right)\lambda }{\omega }_{E}^{n},$  and in particular, $Ricci\left({\omega }_{B}\right)=-{\omega }_{B}.$

Basics of Aut(B)

For the reader's convenience, we recall that Aut $\left(B\right)$  contains the involutions ${F}_{a}\left(z\right)=\frac{a-{P}_{a}z-{s}_{a}{Q}_{a}z}{1-〈z,a〉},a\in B,$  where ${P}_{a}=|a{|}^{-2}a{a}^{†}$  , ${Q}_{a}=I-{P}_{a}$  and ${s}_{a}=\sqrt{1-|a{|}^{2}}$  . Moreover, the Schwarz Lemma shows that any automorphism of $B$  is of the form $U{F}_{a}$  or ${F}_{a}U$  for some unitary $U$  . Note that ${F}_{a}\left(0\right)=a$  and $1-|{F}_{a}\left(z\right){|}^{2}=\frac{\left(1-|z{|}^{2}\right)\left(1-|a{|}^{2}\right)}{|1-〈z,a〉{|}^{2}}.$  Thus $Aut\left(B\right)$  acts transitively on the ball and ${\omega }_{B}$  is Aut $\left(B\right)$  -invariant. (For much more detail on this and the next paragraph, the reader is referred to [R-80or [St-94.)

Basic potential theory of the Bergman metric

Recall that the Bergman Laplacian ${\Delta }_{B}$  associated to ${\omega }_{B}$  is the ${\omega }_{B}$  -trace of $d{d}^{c}$  :
$\left({\Delta }_{B}g\right){\omega }_{B}^{n}=d{d}^{c}g\wedge {\omega }_{B}^{n-1}.$
Definition 2.1. The Green's function with pole at $a\in B$  is the function ${G}_{B}\left(z,a\right)$  satisfying ${\Delta }_{B}\left({G}_{B}\left(\cdot ,a\right)\right)={\delta }_{a}and{G}_{B}\left(\cdot ,a\right)|\partial B=0.$
Using Aut $\left(B\right)$  -invariance, it is easily seen that $G\left(z,a\right)=G\left({F}_{a}\left(z\right),0\right)$  and that $n\left(n+1\right)\left({\Delta }_{B}g\right)\left(a\right)=trace\left(D{F}_{a}\left(0{\right)}^{†}{D}^{1,1}g\left(a\right)D{F}_{a}\left(0\right)\right).$  Here ${D}^{1,1}g$  is the matrix of the $\left(1,1\right)$  -form $\sqrt{-1}\partial \overline{\partial }g$  in Euclidean coordinates. Setting ${\gamma }_{B}={G}_{B}\left(\cdot ,0\right),$  we see from unitary invariance that ${\gamma }_{B}\left(z\right)=f\left(|z{|}^{2}\right)$  for some function $f$  .
Substitution into the Bergman-Laplace equation and solving the resulting ODE shows that $f\left(r\right)=-{C}_{n}{\int }_{r}^{1}\frac{\left(1-u{\right)}^{n-1}}{{u}^{n}}du,$  where ${C}_{n}=\left(2\pi {\right)}^{-n}\left(n+1{\right)}^{-\left(n-1\right)}.$  By using the Green-Stokes identity
 $\begin{array}{ccc}{\int }_{\partial D}\left({g}_{1}{d}^{c}{g}_{2}-{g}_{2}{d}^{c}{g}_{1}\right)\wedge {\omega }^{n-1}={\int }_{D}\left({g}_{1}d{d}^{c}{g}_{2}-{g}_{2}d{d}^{c}{g}_{1}\right)\wedge {\omega }^{n-1},& & \end{array}$ (2)
where $\left(D,\omega \right)$  is an $n$  dimensional Kähler manifold with boundary and ${g}_{1},{g}_{2}:D\to \mathbb{C}$  are functions, we obtain the following Lemma.
Lemma 2.2. Let $h$  be a function such that ${\Delta }_{B}h\ge 0$  . Then
 $\begin{array}{ccc}h\left(0\right)\le \frac{1}{\left(2\pi {\right)}^{n}}{\int }_{\partial B}h\left(rz\right){d}^{c}|z{|}^{2}\wedge {\omega }_{E}^{n-1}\left(z\right).& & \end{array}$ (3)
Moreover, equality holds when ${\Delta }_{B}h\equiv 0$  .
• Proof. Apply ( 2 ) with $D=B\left(0,r\right)$  , $r<1$  , ${g}_{1}=h$  and ${g}_{2}={\gamma }_{r}\left(z\right)={\gamma }_{B}\left(z\right)+{C}_{n}{\int }_{{r}^{2}}^{1}\frac{\left(1-t{\right)}^{n-1}}{{t}^{n}}dt,$  observing that ${\gamma }_{r}|\partial B\left(0,r\right)\equiv 0$  and $d{d}^{c}{\gamma }_{r}\wedge {\omega }_{B}^{n-1}={\delta }_{0}.$  The result now follows from direct computation.
Corollary 2.3. Let $f$  be a function such that ${\Delta }_{B}f\ge 0$  . Then for all $r<1$  ,
 $\begin{array}{ccc}f\left(0\right)\le \frac{1}{{V}_{n}\left(r\right)}{\int }_{B\left(0,r\right)}f{\omega }_{B}^{n}.& & \end{array}$ (4)
Moreover, equality holds in ( 4 ) when ${\Delta }_{B}f\equiv 0$  .

3 The density conditions again

3.1 Reformulation of the density conditions

It will be useful to rewrite the positivity of the upper and lower densities in terms of the positivity of certain associated differential forms.
Lemma 3.1. Let the notation be as above.
• 1. If ${\mathfrak{D}}_{B}^{+}\left(W,\kappa \right)<1$  , then there is a positive constant $c$  such that $\sqrt{-1}\partial \overline{\partial }\kappa -\frac{n}{n+1}{\omega }_{B}-{\Upsilon }_{r}^{W}\ge c\sqrt{-1}\partial \overline{\partial }\kappa$  in the sense of currents.
• 2. If ${\mathfrak{D}}_{B}^{-}\left(W,\kappa \right)>1$  , then there exists $\theta \in {\mathfrak{P}}_{W}\left(B\right)$  and $c>0$  such that $\left({\Upsilon }_{r}^{W}+\frac{n}{n+1}{\omega }_{B}-\sqrt{-1}\partial \overline{\partial }\kappa \right)\wedge \theta \ge c{e}^{n\lambda }{\omega }_{B}^{n}.$
• Proof. After using condition (a) in the definition of ${\mathfrak{P}}_{W}\left(B\right)$  , assertion 2 is trivially true from the definition of the lower density.
To see assertion 1, choose any $v\in {T}_{B,p}$  . Choose holomorphic Euclidean coordinates ${x}^{1},...,{x}^{n}$  in ${\mathbb{C}}^{n}$  (where the ball lies) such that $v=c\frac{\partial }{\partial {x}^{1}}$  . (These are just obtained from the coordinates we started with by a unitary transformation of ${\mathbb{C}}^{n}$  .) Consider the $\left(n-1,n-1\right)$  -form $\theta ={\theta }_{v}:=\left(\sqrt{-1}{\right)}^{n-1}d{x}^{2}\wedge d{\overline{x}}^{2}\wedge \cdots \wedge d{x}^{n}\wedge d{\overline{x}}^{n}.$  We claim that $\theta \in {\mathfrak{P}}_{W}\left(B\right)$  . Indeed, $\sqrt{-1}\partial \overline{\partial }\theta =0$  so condition (c) in the definition of ${\mathfrak{P}}_{W}\left(B\right)$  holds. Condition (b) is trivial. Condition (a) can be seen as follows:  $\begin{array}{ccc}\theta \wedge {\omega }_{B}& =& C\theta \wedge \left(\frac{{\omega }_{E}}{1-|x{|}^{2}}+\frac{\sqrt{-1}\partial |x{|}^{2}\wedge \overline{\partial }|x{|}^{2}}{\left(1-|x{|}^{2}{\right)}^{2}}\right)\end{array}$
 $\begin{array}{ccc}& \ge & C{e}^{-\lambda }{\omega }_{E}^{n}\end{array}$
 $\begin{array}{ccc}& =& C{e}^{n\lambda }{\omega }_{B}^{n}.\end{array}$
By the density condition there exists $\delta >0$  such that  $\begin{array}{ccc}1-\delta & \ge & \frac{\left({\Upsilon }_{r}^{W}+\frac{n}{n+1}{\omega }_{B}\right)\wedge \theta }{\sqrt{-1}\partial \overline{\partial }\kappa \wedge \theta }\end{array}$
 $\begin{array}{ccc}& =& \frac{\left({\Upsilon }_{r}^{W}\left(v,v\right)+\frac{n}{n+1}{\omega }_{B}\left(v,v\right)\right)}{\sqrt{-1}\partial \overline{\partial }\kappa \left(v,v\right)}.\end{array}$
 $\begin{array}{ccc}& & \end{array}$
This completes the proof.

3.2 A seemingly better notion of density

In the paper [OSV, the approach taken to solve the interpolation and sampling problems was quite different. To begin with, a different notion of density was used. As a consequence, the method of proof was very different from the one used in the present paper. The purpose of this section is to demonstrate the equivalence of these two notions of density.
Let us define the Bergman ball analogues of the densities used in [OSV. One first sets ${D}_{z,r}\left(W,\kappa \right):={sup}_{v\ne 0}\frac{{\Upsilon }_{r}^{W}\left(v,v\right)+\frac{n}{n+1}{\omega }_{B}\left(v,v\right)}{\sqrt{-1}\partial \overline{\partial }\kappa \left(v,v\right)}.$  Then one takes ${D}^{+}\left(W,\kappa \right):={limsup}_{r\to 1}{sup}_{z\in B}{D}_{z,r}\left(W,\kappa \right)$  and ${D}^{-}\left(W,\kappa \right):={liminf}_{r\to 1}{inf}_{z\in B}{D}_{z,r}\left(W,\kappa \right)$  Note that ${D}_{z,r}\left(W,\kappa \right)$  is the maximum eigenvalue of the $\left(1,1\right)$  -form ${\Upsilon }_{r}^{W}+\frac{n}{n+1}{\omega }_{B}$  with respect to the positive $\left(1,1\right)$  -form $\sqrt{-1}\partial \overline{\partial }\kappa$  .
Lemma 3.2. Let $\alpha$  be a non-negative $\left(n-1,n-1\right)$  form on a Hermitian manifold $M$  of dimension $n$  and Hermitian form $\omega$  . Then for each $p$  there exists a vector $v\in {T}_{M,p}^{1,0}$  such that for any real $\left(1,1\right)$  -form $\beta$  , one has $\alpha \wedge {\beta }_{p}={\beta }_{p}\left(v,\overline{v}\right){\omega }^{n}.$
The Lemma says that the mapping $\sqrt{-1}v\wedge \overline{v}↦{\theta }_{v}$  , with ${\theta }_{v}$  as in the proof of Lemma  3.1 , is a pointwise isomorphism.
• Proof. We shall use linear algebra on ${T}_{M,p}$  . To this end, choose a unitary basis ${e}^{1},...,{e}^{n}$  for $\left({T}_{M,p}^{*}{\right)}^{1,0}$  and ${e}_{1},...,{e}_{n}$  its dual basis. Let ${\alpha }^{i\overline{j}}$  be a basis for ${\Lambda }^{n-1,n-1}\left({T}_{M,p}^{*}\right)$  such that $\sqrt{-1}{e}^{k}\wedge {\overline{e}}^{\ell }\wedge {\alpha }^{i\overline{j}}={\delta }^{ik}{\delta }^{\overline{j}\overline{\ell }}\frac{{\omega }^{n}}{n!}.$  Let $A$  (resp. $B$  ) be the Hermitian matrix with entries ${a}_{i\overline{j}}$  (resp. ${b}_{i\overline{j}}$  ) such that at the point $p$  , $\alpha ={a}_{i\overline{j}}{\alpha }^{i\overline{j}}\left(\text{resp.}\beta ={b}_{i\overline{j}}\sqrt{-1}{e}^{i}\wedge {\overline{e}}^{j}\right).$  Then $\alpha \wedge {\beta }_{p}=\text{Trace}\left(A{B}^{†}\right)\frac{{\omega }^{n}}{n!}\text{and}{\beta }_{p}\left(v,\overline{v}\right)={v}^{†}Bv.$  After a unitary rotation, we may assume that the basis ${e}^{1},...,{e}^{n}$  diagonalizes $A$  . Thus, since $\alpha$  is positive, there exist non-negative numbers ${\lambda }_{1},...,{\lambda }_{n}$  such that $\text{Trace}\left(A{B}^{†}\right)={\sum }_{k=1}^{n}{\lambda }_{k}{b}_{kk}.$  Taking $v={\sum }_{k=1}^{n}\sqrt{{\lambda }_{k}}{e}_{k}$  completes the proof.
Proposition 3.3.
• (a) ${\mathfrak{D}}^{+}\left(W,\kappa \right)={D}^{+}\left(W,\kappa \right)$  .
• (b) ${\mathfrak{D}}^{-}\left(W,\kappa \right)\le {D}^{-}\left(W,\kappa \right)$  .
• Proof. (a) Fix $z\in B$  and $r\in \left[0,1\right)$  . By definition of ${D}^{+}\left(W,\kappa \right)$  , we have that for any $\theta \in {\mathfrak{P}}_{W}\left(B\right)$  ,  $\begin{array}{ccc}{D}^{+}\left(W,\kappa \right)& \ge & {D}_{z,r}\left(W,\kappa \right)\end{array}$
 $\begin{array}{ccc}& \ge & \frac{{\Upsilon }_{r}^{W}\wedge \theta \left(z\right)+\frac{n}{n+1}{\omega }_{B}\wedge \theta \left(z\right)}{\sqrt{-1}\partial \overline{\partial }\kappa \wedge \theta \left(z\right)}.\end{array}$
(In the second inequality we have used Lemma  3.2 .) Taking the supremum over $z$  and them the $limsup$  as $r\to 1$  , we see that ${D}^{+}\left(W,\kappa \right)\ge {\mathfrak{D}}^{+}\left(W,\kappa \right)\left[\theta \right].$  Finally, taking the supremum of the right hand side over all $\theta \in {\mathfrak{P}}_{W}\left(B\right)$  shows that ${D}^{+}\left(W,\kappa \right)\ge {\mathfrak{D}}^{+}\left(W,\kappa \right)$  .
To obtain the reverse inequality, fix $\varepsilon >0$  . Then there exist $z\in B$  , $r\in \left[0,1\right)$  and $v\in {\mathbb{C}}^{n}$  such that  $\begin{array}{ccc}{D}^{+}\left(W,\kappa \right)-\varepsilon & \le & \frac{{\Upsilon }_{r}^{W}\left(v,v\right)+\frac{n}{n+1}{\omega }_{B}\left(v,v\right)}{\sqrt{-1}\partial \overline{\partial }\kappa \left(v,v\right)}\end{array}$
 $\begin{array}{ccc}& =& \frac{{\Upsilon }_{r}^{W}\wedge {\theta }_{v}\left(z\right)+\frac{n}{n+1}{\omega }_{B}\wedge {\theta }_{v}\left(z\right)}{\sqrt{-1}\partial \overline{\partial }\kappa \wedge {\theta }_{v}\left(z\right)}\end{array}$
 $\begin{array}{ccc}& \le & {\mathfrak{D}}^{+}\left(W,\kappa \right)\left[{\theta }_{v}\right]\le {\mathfrak{D}}^{+}\left(W,\kappa \right),\end{array}$
where ${\theta }_{v}$  is defined as in the proof of Lemma  3.1 . Since $\varepsilon$  is arbitrary, 1 is proved.
(b) Fix $\varepsilon >0$  . By definition of ${\mathfrak{D}}^{-}\left(W,\kappa \right)$  , there exists $\theta \in {\mathfrak{P}}_{W}\left(B\right)$  such that ${\mathfrak{D}}^{-}\left(W,\kappa \right)\le {\mathfrak{D}}^{-}\left(W,\kappa \right)\left[\theta \right]+\frac{\varepsilon }{2}.$  Moreover, by the definition of ${\mathfrak{D}}^{-}\left(W,\kappa \right)\left[\theta \right]$  , we have that for all $z\in B$  , and all $r\in \left[0,1\right)$  sufficiently large, ${\mathfrak{D}}^{-}\left(W,\kappa \right)\left[\theta \right]\le \frac{{\Upsilon }_{r}^{W}\wedge \theta \left(z\right)+\frac{n}{n+1}{\omega }_{B}\wedge \theta \left(z\right)}{\sqrt{-1}\partial \overline{\partial }\kappa \wedge \theta \left(z\right)}+\frac{\varepsilon }{2}.$  But by Lemma  3.2 and definition of ${D}_{z,r}\left(W,\kappa \right)$  , $\frac{{\Upsilon }_{r}^{W}\wedge \theta \left(z\right)+\frac{n}{n+1}{\omega }_{B}\wedge \theta \left(z\right)}{\sqrt{-1}\partial \overline{\partial }\kappa \wedge \theta \left(z\right)}\le {D}_{z,r}\left(W,\kappa \right).$  This proves (b).
Theorem 3. ${\mathfrak{D}}^{-}\left(W,\kappa \right)\ge {D}^{-}\left(W,\kappa \right).$
• Proof. We introduce the notation ${\Omega }_{\delta }:={\Upsilon }_{r}^{W}+\frac{n}{n+1}{\omega }_{B}-\left({D}^{-}\left(W,\kappa \right)-\delta \right)\sqrt{-1}\partial \overline{\partial }\kappa .$  Let $\delta >0$  be given. For $r>>0$  we are going to construct a form $\theta \in {\mathfrak{P}}_{W}\left(B\right)$  such that ${\Omega }_{\delta }\wedge \theta \ge 0.$  If this is done, the proof is complete.
By definition of ${D}^{-}\left(W,\kappa \right)$  , there exists a locally finite open cover ${U}_{j}$  of $B$  and constant $\left(n-1,n-1\right)$  -forms (i.e., forms of the type ${\theta }_{v}$  defined in the proof of Lemma  3.1 ) ${\theta }_{j}$  on ${U}_{j}$  such that ${\Omega }_{\delta /2}\wedge {\theta }_{j}\ge 0\text{on}{U}_{j}.$  By the uniform flatness of $W$  , we may choose the cover $\left\{{U}_{j}\right\}$  such that any point of $B$  is contained in some finite number of neighborhoods, this number depending only on the dimension. Moreover, by the continuity of the forms ${\Omega }_{\varepsilon }$  we may choose the forms ${\theta }_{j}$  so that if ${U}_{j}\cap {U}_{k}\ne \varnothing$  then ${\theta }_{j}-{\theta }_{k}$  is as small as we like. In fact, by elementary anti-differentiation we may take forms ${\mu }_{j}$  depending quadratically on the (global Euclidean) coordinates in $B$  such that if ${U}_{j}\cap {U}_{k}\ne \varnothing$  then $||{\mu }_{j}-{\mu }_{k}|{|}_{{\mathsf{ℭ}}^{2}\left({U}_{j}\cap {U}_{k}\right)}$  is as small as we like, where $||\cdot |{|}_{{\mathsf{ℭ}}^{2}}$  denotes ${\mathsf{ℭ}}^{2}$  -norm.
The argument we present here requires a little more precision. Later we will have to control the size of the neighborhoods ${U}_{j}$  in order to make the ${\theta }_{j}-{\theta }_{k}$  small enough. To this end, we choose the ${U}_{j}$  to be balls (or polydisks) of diameter $\varepsilon$  , measured with respect to the Bergman-Green distance ${d}_{B}\left(z,w\right)=|{F}_{w}\left(z\right)|$  . We indicate this dependence on $\varepsilon$  by writing ${U}_{j,\varepsilon }$  , ${\mu }_{j,\varepsilon }$  and ${\theta }_{j,\varepsilon }$  . Observe that if we take ${\mu }_{j,\varepsilon }$  to be quadratic and use coordinates emanating from the origin of ${U}_{j,\varepsilon }$  , then the uniform estimates for ${\mu }_{j,\varepsilon }$  scale by ${\varepsilon }^{2}$  , those for $D{\mu }_{j,\varepsilon }$  by $\varepsilon$  , and those from ${\theta }_{j,\varepsilon }$  are invariant with respect to $\varepsilon$  .
Let $\left\{{\psi }_{j,\varepsilon }\right\}$  be a partition of unity subordinate to the cover $\left\{{U}_{j,\varepsilon }\right\}$  . We may choose this partition so that ${\sum }_{j}||{\psi }_{j,\varepsilon }{\mu }_{j,\varepsilon }|{|}_{{\mathsf{ℭ}}^{2}}\le C$  for some constant $C$  independent of $\varepsilon$  . Indeed, as the neighborhoods ${U}_{j,\varepsilon }$  scale by $\varepsilon$  , the estimates for $D{\psi }_{j,\varepsilon }$  scale by ${\varepsilon }^{-1}$  while those for ${D}^{2}{\psi }_{j,\varepsilon }$  scale by ${\varepsilon }^{-2}$  . Thus the desired estimate follows from the product rule ${D}^{2}\left({\psi }_{j,\varepsilon }{\mu }_{j,\varepsilon }\right)={\mu }_{j,\varepsilon }{D}^{2}{\psi }_{j,\varepsilon }+\left(D{\mu }_{j,\varepsilon }\right)\left(D{\psi }_{j,\varepsilon }\right)+{\psi }_{j,\varepsilon }{D}^{2}{\mu }_{j,\varepsilon }.$  Thus is is clear that we have scale invariant estimates. To simplify the exposition, we shall drop the notational dependence on $\varepsilon$  .
We would like to correct the local forms ${\theta }_{j}$  so that they can be pieced together to give us an element of ${\mathfrak{P}}_{W}\left(B\right)$  with the desired density. We shall use cocycles to do this. To this end, the obstruction to the ${\theta }_{j}$  piecing together to give a global form is carried by the 1-cocycle ${\alpha }_{jk}={\theta }_{j}-{\theta }_{k}=\sqrt{-1}\partial \overline{\partial }\left({\mu }_{j}-{\mu }_{k}\right)$  supported on ${U}_{j}\cap {U}_{k}$  . By our choice of the ${\theta }_{j}$  , the ${\alpha }_{jk}$  are small in ${\mathsf{ℭ}}^{0}$  -norm. We now define ${\eta }_{j}=\sqrt{-1}\partial \overline{\partial }\left({\sum }_{k}{\psi }_{k}\left({\mu }_{j}-{\mu }_{k}\right)|{U}_{j}\cap {U}_{k}\right).$  By modifying our choices of the ${\mu }_{j}$  we may make the ${\eta }_{j}$  as small as we like. Moreover, $\sqrt{-1}\partial \overline{\partial }{\eta }_{j}=0$  and  $\begin{array}{ccc}{\eta }_{j}-{\eta }_{\ell }& =& \sqrt{-1}\partial \overline{\partial }{\sum }_{k}{\psi }_{k}\left({\mu }_{j}-{\mu }_{k}+{\mu }_{k}-{\mu }_{\ell }\right)\end{array}$
 $\begin{array}{ccc}& =& \sqrt{-1}\partial \overline{\partial }{\sum }_{k}{\psi }_{k}\left({\mu }_{j}-{\mu }_{\ell }\right)\end{array}$
 $\begin{array}{ccc}& =& {\alpha }_{j\ell }.\end{array}$
It follows that $\theta ={\theta }_{j}-{\eta }_{j}\text{on}{U}_{j}$  is well defined and belongs to ${\mathfrak{P}}_{W}\left(B\right)$  . Moreover, by choosing the ${\mu }_{j}-{\mu }_{k}$  even smaller if necessary, we see that ${\Omega }_{\delta }\wedge \theta \ge 0,$  as desired.

4 Uniform flatness

In [OSVa notion of uniform flatness was developed for closed smooth hypersurfaces in ${\mathbb{C}}^{n}$  . Here we define the analogous notion for the ball with its Bergman geometry.
Let ${N}_{\varepsilon }^{B}\left(W\right)=\left\{z\in B|{inf}_{w\in W}|{F}_{z}\left(w\right)|<\varepsilon \right\}$
Definition 4.1. We define a smooth divisor $W$  in $B$  to be uniformly flat if there exists an ${\varepsilon }_{0}>0$  such that ${N}_{{\varepsilon }_{0}}^{B}\left(W\right)$  is a tubular neighborhood of $W$  .
The following consequence of uniform flatness is useful.
Lemma 4.2. If a closed hypersurface $W\subset B$  is uniformly flat, then there exist ${\varepsilon }_{0}>0$  and $C>0$  such that for each $z\in W$  the set ${F}_{z}\left(W\right)\cap B\left(0,{\varepsilon }_{0}\right)$  is a graph, over the Euclidean ${\varepsilon }_{0}$  -neighborhood of the origin in the tangent space ${T}_{{F}_{z}\left(W\right),0}=d{F}_{z}\left({T}_{W,z}\right)$  , of some function $f$  such that $|f\left(x\right)|\le C|x{|}^{2},|x|<{\varepsilon }_{0}.$
The proof is left to the reader.

5 Interpolation

5.1 A negative function singular along a hypersurface

Recall that ${V}_{n}\left(r\right):={\int }_{B\left(0,r\right)}{\omega }_{B}^{n}.$  As the Bergman metric is invariant under automorphisms, one sees that for each $a\in B$  , ${V}_{n}\left(r\right)$  is also the Bergman volume of the set $E\left(a,r\right):={F}_{a}\left(B\left(0,r\right)\right).$  In view of the form of the Green's function for the Bergman Laplacian, we call the sets $E\left(a,r\right)$  Bergman-Green balls.
Let
 $\begin{array}{ccc}{\Gamma }_{r}\left(z,\zeta \right)={G}_{B}\left(z,\zeta \right)-\frac{1}{{V}_{n}\left(r\right)}{\int }_{E\left(z,r\right)}{G}_{B}\left(x,\zeta \right){\omega }_{B}^{n}\left(x\right)& & \end{array}$
Since ${\Delta }_{B}G\left(\cdot ,\zeta \right)\equiv 0$  on $B-\left\{\zeta \right\}$  , we see from Corollary  2.3 that ${\Gamma }_{r}$  is non-negative and supported on the neighborhood $\left\{\left(z,\zeta \right)\in B×B||{F}_{z}\left(\zeta \right)|  of the diagonal in $B×B$  .
We define the function
 $\begin{array}{ccc}{s}_{r}\left(z\right)& :=& {\int }_{B}{\Gamma }_{r}\left(z,\zeta \right){\omega }_{B}^{n-1}\left(\zeta \right)\wedge d{d}^{c}log|T{|}^{2}\left(\zeta \right)\end{array}$
 $\begin{array}{ccc}& =& {\int }_{\left\{\zeta ;|{F}_{z}\left(\zeta \right)|
By the Lelong-Poincaré identity,
${s}_{r}\left(z\right)=2\pi {\int }_{{W}_{z,r}}\left({G}_{B}\left(z,\zeta \right)-\frac{1}{V\left(r\right)}{\int }_{E\left(z,r\right)}{G}_{B}\left(x,\zeta \right){\omega }_{B}^{n}\left(x\right)\right){\omega }_{B}^{n-1}\left(\zeta \right),$
where
${W}_{z,r}=W\cap \left\{\zeta ;|{F}_{z}\left(\zeta \right)|
Proposition 5.1. Let $T\in \mathcal{O}\left(B\right)$  be a holomorphic function so that $W={T}^{-1}\left(0\right)$  and $dT$  is nowhere zero on $W$  . Then ${s}_{r}\left(z\right)=log|T\left(z\right){|}^{2}-\frac{1}{{V}_{n}\left(r\right)}{\int }_{E\left(z,r\right)}log|T\left(\zeta \right){|}^{2}{\omega }_{B}^{n}\left(\zeta \right).$
• Proof. Let $\alpha :\left[0,\infty \right)\to \left[0,1\right]$  be a smooth function with compact support which is identically 1 on $\left[0,1\right]$  . Then for $R>r$   $\begin{array}{ccc}{s}_{r}\left(z\right)& =& {\int }_{B}\Gamma \left(z,\zeta \right){\omega }_{B}^{n-1}\left(\zeta \right)\wedge d{d}^{c}log|T{|}^{2}\left(\zeta \right)=\end{array}$
 $\begin{array}{ccc}& =& {\int }_{B}\alpha \left(\frac{|{F}_{\zeta }\left(z\right)|}{R}\right)\Gamma \left(z,\zeta \right){\omega }_{B}^{n-1}\left(\zeta \right)\wedge d{d}^{c}log|T{|}^{2}\left(\zeta \right)\end{array}$
the latter equality following from the fact that ${\Gamma }_{r}\left(\cdot ,\zeta \right)$  is supported on $\left\{z;|{F}_{z}\left(\zeta \right)|  .
An application of the Green-Stokes Identity ( 2 ) then gives ${s}_{r}\left(z\right)={\int }_{B}d{d}^{c}\left(\alpha \left(\frac{|{F}_{\zeta }\left(z\right)|}{R}\right)\Gamma \left(z,\zeta \right)\right)\wedge log|T\left(\zeta \right){|}^{2}\wedge {\omega }_{B}^{n-1},$  where the boundary terms have disappeared because the forms are compactly supported.
Letting $R\to \infty$  we obtain ${s}_{r}\left(z\right)={\int }_{B}log|T\left(\zeta \right){|}^{2}d{d}^{c}{\Gamma }_{r}\left(z,\zeta \right)\wedge {\omega }_{B}^{n-1},$  and the proof is complete.
Lemma 5.2. The function ${s}_{r}\left(z\right)$  has the following properties:
• 1. It is non-positive
• 2. For each $r,\epsilon >0$  there exist a constant ${C}_{r,\epsilon }$  such that if ${\delta }_{B}\left(z,W\right)>\epsilon$  then ${s}_{r}\left(z\right)>-{C}_{r,\epsilon }$  .
• 3. The function ${e}^{-{s}_{r}}$  is not locally integrable at any point of W.
Here ${\delta }_{B}\left(z,W\right)=inf\left\{|{F}_{z}\left(w\right)|;w\in W\right\}$  .
• Proof. By Corollary  2.3 and the fact that ${\Delta }_{B}G\left(\cdot ,\zeta \right)\equiv 0$  on $B-\left\{\zeta \right\}$  , ${\Gamma }_{r}\le 0$  and 1 follows. Moreover, 3 is an immediate consequence of Proposition  5.1 .
To see 2, we first note that since ${\delta }_{B}\left(z,W\right)>\varepsilon$  , ${G}_{B}\left(z,\zeta \right)>{A}_{\varepsilon }$  . Thus it suffices to obtain an estimate $-{\int }_{E\left(z,r\right)}{G}_{B}\left(x,\zeta \right){\omega }_{B}^{n}\left(x\right)=-{\int }_{B\left(0,r\right)}{G}_{B}\left(x,y\right){\omega }_{B}^{n}\left(x\right)\le {D}_{r}$  for some ${D}_{r}>0$  and all $y={F}_{z}\left(\zeta \right)\in B\left(0,r\right)$  . To do this, it is enough to estimate the integral $I\left(r\right):=-{\int }_{B\left(0,\left(r+1\right)/2\right)}{G}_{B}\left(x,y\right){\omega }_{B}^{n}\left(x\right).$  Fix $y\in B\left(0,r\right)$  . Let $\rho >0$  be the largest number such that $B\left(y,\rho \right)\subset B\left(0,\left(r+1\right)/2\right).$  One has ${n}_{r}\le \rho \le \frac{r+1}{2}$  for some ${n}_{r}>0$  depending on $r$  but not on $y$  .
Write $I\left(r\right)={I}_{1}\left(r\right)+{I}_{2}\left(r\right),$  where ${I}_{1}\left(r\right):=-{\int }_{E\left(y,\rho \right)}{G}_{B}\left(x,y\right){\omega }_{B}^{n}\left(x\right)$  and ${I}_{2}\left(r\right):=-{\int }_{B\left(0,\left(r+1\right)/2\right)-E\left(y,\rho \right)}{G}_{B}\left(x,y\right){\omega }_{B}^{n}\left(x\right).$  Now ${I}_{1}\left(r\right)=-{\int }_{B\left(0,\rho \right)}{\gamma }_{B}\left(x\right){\omega }_{B}^{n}\left(x\right)$  is clearly bounded by a constant independent of $y$  .
Next, note that for $x\in B\left(0,\left(r+1\right)/2\right)-E\left(y,\rho \right)$  one has the estimate $|{F}_{y}\left(x\right)|\ge \rho \ge {n}_{r}.$  It follows that for such $x$  , $G\left(x,y\right)\ge -{N}_{r}$  for some ${N}_{r}\in \mathbb{R}$  independent of $y$  . Thus  $\begin{array}{ccc}{I}_{2}\left(r\right)& \le & {N}_{r}{\int }_{B\left(0,\left(r+1\right)/2\right)-E\left(y,\rho \right)}{\omega }_{B}^{n}\left(x\right)\end{array}$
 $\begin{array}{ccc}& \le & {N}_{r}{\int }_{B\left(0,\left(r+1\right)/2\right)}{\omega }_{B}^{n}\left(x\right),\end{array}$
and the latter is independent of $y$  . Thus 2 follows.
Remark. There is a direct proof of Lemma  5.2 .3 that does not use the formula of Proposition  5.1 . Since we will make use of the calculation needed, we present this proof now.
We may assume that $W$  is the coordinate hyperplane ${z}_{n}=0$  and $z={z}^{n}{e}_{n}$  for $|\lambda |\le \varepsilon$  with $\varepsilon$  sufficiently small. By the proof of Lemma  5.2 .2 and the form of the Green's function, we see that
 $\begin{array}{ccc}{s}_{r}\left(z\right)& =& 2\pi {\int }_{U\cap W}{\Gamma }_{r}\left(z,\zeta \right){\omega }_{B}^{n-1}\left(\zeta \right)+O\left(1\right)\end{array}$
 $\begin{array}{ccc}& =& 2\pi {C}_{n}\left(n+1{\right)}^{n-1}{A}_{n-1}{\int }_{0}^{\alpha }-\frac{2{r}^{2n-3}dr}{{\left({r}^{2}+|{z}^{n}{|}^{2}\right)}^{n-1}}+O\left(1\right)\end{array}$
 $\begin{array}{ccc}& =& log|{z}^{n}{|}^{2}+O\left(1\right)\end{array}$
where $U$  is some neighborhood of the origin and $\alpha >0$  .

5.2 One-time extension for hypersurfaces with smaller density.

In this subsection we prove that hypersurfaces whose density is slightly smaller than the optimal density are interpolating. As mentioned in the introduction, this result is weaker than the one we will prove in the next subsection. However, the method is interesting in its own right, and it would be an advance to be able to modify this approach to work for those hypersurfaces whose density is optimal.
To be specific, we will prove the following result.
Theorem 4. Let $W\subset B$  be a properly embedded, codimension-1 complex submanifold. Assume that for some $c>0$  ,
 $\begin{array}{ccc}{\Upsilon }_{r}^{W}+{\omega }_{B}\le \left(1-c\right)\sqrt{-1}\partial \overline{\partial }\kappa .& & \end{array}$ (5)
Then $W$  is an interpolation hypersurface.

Tubular limits

For each $\Omega \subset \subset B$  , let ${\Omega }_{\varepsilon }:=\Omega \cap \left\{{s}_{r}
Lemma 5.3. Let $W\subset B$  be uniformly flat. Then there exists a positive constant $C>0$  such that for all $\Omega \subset \subset B$  and all $f$  holomorphic in $\overline{\Omega }$  , ${limsup}_{\varepsilon \to 0}\frac{1}{{\varepsilon }^{2}}{\int }_{{\Omega }_{\varepsilon }}|f{|}^{2}{e}^{-\kappa }{\omega }_{B}^{n}\le C{\int }_{\Omega \cap W}|f{|}^{2}{e}^{-\kappa }{\omega }_{B}^{n-1}.$
• Sketch of proof. We may assume the right hand side is finite. Moreover, we can take $\Omega =E\left(a,\varepsilon \right)$  for some $a\in W$  , with $\varepsilon$  so small that $W\cap \Omega$  is the graph of a quadratic hypersurface. By uniform flatness, $\varepsilon$  can be taken independent of $a$  .
Consider first the case $a=0$  . Then $\Omega =B\left(0,\varepsilon \right)$  , and the result follows after an elementary analysis of the properties of ${s}_{r}$  as in the proof of (2) and (3) in Lemma  5.2 .
If we now apply the automorphism ${F}_{a}$  to $B\left(0,\varepsilon \right)$  , then Lemma  4.2 and the Aut $\left(B\right)$  -invariance of ${\omega }_{B}$  show that the same estimates hold on $E\left(a,\varepsilon \right)$  .

The twisted Bochner-Kodaira Technique

We fix a smoothly bounded pseudoconvex domain $\Omega \subset \subset B$  . Let us denote by ${\overline{\partial }}_{\nu }^{*}$  the formal adjoint of $\overline{\partial }$  in the Hilbert space of $\left(0,1\right)$  -forms on $\Omega$  , square integrable with respect to a weight ${e}^{-\nu }{\omega }_{B}^{n}$  . For a $\left(0,1\right)$  -form $u={u}_{\overline{\alpha }}d{\overline{z}}^{\alpha }$  , one has ${\overline{\partial }}_{\nu }^{*}u=-{e}^{\nu +\left(n+1\right)\lambda }{\partial }_{\alpha }\left({e}^{-\left(\nu +\left(n+1\right)\lambda \right)}{u}^{\alpha }\right).$  Recall that for $\left(0,1\right)$  -forms $u$  the Bochner-Kodaira Identity, often also referred to as the basic estimate, is
 $\begin{array}{ccc}& & {\int }_{\Omega }{|{\overline{\partial }}_{\nu }^{*}u|}^{2}{e}^{-\nu }{\omega }_{B}^{n}+{\int }_{\Omega }{|\overline{\partial }u|}^{2}{e}^{-\nu }{\omega }_{B}^{n}\end{array}$ (6)
 $\begin{array}{ccc}& & ={\int }_{\Omega }\left(\left({\partial }_{\alpha }{\partial }_{\overline{\beta }}\left(\nu +\left(n+1\right)\lambda \right)\right){u}^{\alpha }\overline{{u}^{\beta }}\right){e}^{-\nu }{\omega }_{B}^{n}+{\int }_{\Omega }{|\overline{\nabla }u|}^{2}{e}^{-\nu }{\omega }_{B}^{n}\end{array}$
 $\begin{array}{ccc}& & +{\int }_{\partial \Omega }\left({\partial }_{\alpha }{\partial }_{\overline{\beta }}\rho \right){u}^{\alpha }\overline{{u}^{\beta }}{e}^{-\nu }{d}^{c}\left(-\left(n+1\right)\lambda \right)\wedge {\omega }_{B}^{n-1},\end{array}$
where $\rho$  is a defining function for $\Omega$  such that $|d\rho |\equiv 1$  on $\partial \Omega$  . (See, for example, [Siu-82.) The term $\left(n+1\right)\lambda$  comes from the Ricci curvature of ${\omega }_{B}$  . Writing ${e}^{-\psi }=\frac{{e}^{-\nu }}{\tau }$  we obtain ${\overline{\partial }}_{\nu }^{*}u={\overline{\partial }}_{\psi }^{*}u-\frac{\left({\partial }_{\alpha }\tau \right){u}^{\alpha }}{\tau }and{\partial }_{\alpha }{\partial }_{\overline{\beta }}\psi ={\partial }_{\alpha }{\partial }_{\overline{\beta }}\nu +\frac{{\partial }_{\alpha }{\partial }_{\overline{\beta }}\tau }{\tau }-\frac{\left({\partial }_{\alpha }\tau \right)\left({\partial }_{\overline{\beta }}\tau \right)}{{\tau }^{2}}.$  Substitution into ( 6 ), followed by some simple manipulation, gives the so-called Twisted Bochner-Kodaira Identity for $\left(0,1\right)$  -forms: If $u$  is a $\left(0,1\right)$  -form in the domain of ${\overline{\partial }}^{*}$  , then
 $\begin{array}{ccc}& & {\int }_{\Omega }\tau {|{\overline{\partial }}_{\psi }^{*}u|}^{2}{e}^{-\psi }{\omega }_{B}^{n}+{\int }_{\Omega }\tau {|\overline{\partial }u|}^{2}{e}^{-\psi }{\omega }_{B}^{n}\end{array}$ (7)
 $\begin{array}{ccc}& =& {\int }_{\Omega }\left(\tau \left({\partial }_{\alpha }{\partial }_{\overline{\beta }}\left(\psi +\left(n+1\right)\lambda \right)\right){u}^{\alpha }\overline{{u}^{\beta }}-\left({u}^{\alpha }\overline{{u}^{\beta }}{\partial }_{\alpha }{\partial }_{\overline{\beta }}\tau \right)\end{array}$
 $\begin{array}{ccc}& & +2Re\left(\left({\partial }_{\alpha }\tau \right){u}^{\alpha }\overline{{\overline{\partial }}_{\psi }^{*}u}\right)\right){e}^{-\psi }{\omega }_{B}^{n}+{\int }_{\Omega }\tau {|\overline{\nabla }u|}^{2}{e}^{-\psi }{\omega }_{B}^{n}\end{array}$
 $\begin{array}{ccc}& & +{\int }_{\partial \Omega }\tau \left({\partial }_{\alpha }{\partial }_{\overline{\beta }}\rho \right){u}^{\alpha }\overline{{u}^{\beta }}{e}^{-\psi }{d}^{c}\left(-\left(n+1\right)\lambda \right)\wedge {\omega }_{B}^{n-1}.\end{array}$
We now use positivity of the last two integrals on the right hand side, together with the Cauchy-Schwarz inequality applied to the first term in the third line, to obtain the so-called Twisted basic estimate: If $u$  is a $\left(0,1\right)$  -form in the domain of ${\overline{\partial }}^{*}$  , then
 $\begin{array}{ccc}& & {\int }_{\Omega }\left(\tau +A\right){|{\overline{\partial }}_{\psi }^{*}u|}^{2}{e}^{-\psi }{\omega }_{B}^{n}+{\int }_{\Omega }\tau {|\overline{\partial }u|}^{2}{e}^{-\psi }{\omega }_{B}^{n}\end{array}$ (8)
 $\begin{array}{ccc}& & \ge {\int }_{\Omega }\left(\tau \left({\partial }_{\alpha }{\partial }_{\overline{\beta }}\left(\psi +\left(n+1\right)\lambda \right)\right){u}^{\alpha }\overline{{u}^{\beta }}\end{array}$
 $\begin{array}{ccc}& & -{\partial }_{\alpha }{\partial }_{\overline{\beta }}\tau {u}^{\alpha }\overline{{u}^{\beta }}-\frac{1}{A}{|\left({\partial }_{\alpha }\tau \right){u}^{\alpha }|}^{2}\right){e}^{-\psi }{\omega }_{B}^{n}.\end{array}$

Choice of $\psi$  , $\tau$  and $A$

From the very beginning, we choose $\psi =\kappa +{s}_{r}.$  From the density hypothesis (via Lemma  3.1 .1) one has
 $\begin{array}{ccc}\sqrt{-1}\partial \overline{\partial }\left(\psi +\left(n+1\right)\lambda \right)& =& \sqrt{-1}\partial \overline{\partial }\left(\kappa +\left(n+1\right)\lambda +{s}_{r}\right)\end{array}$
 $\begin{array}{ccc}& \ge & c\sqrt{-1}\partial \overline{\partial }\kappa \end{array}$
Next, fix $\gamma >1$  . We define $\xi =log\left({e}^{{s}_{r}}+{\varepsilon }^{2}\right),$  with $\varepsilon >0$  so small that $\gamma -\xi \ge 1$  . One has
 $\begin{array}{ccc}\sqrt{-1}\partial \overline{\partial }\xi & =& \sqrt{-1}\partial \left(\frac{{e}^{{s}_{r}}}{{e}^{{s}_{r}}+{\varepsilon }^{2}}\overline{\partial }{s}_{r}\right)\end{array}$
 $\begin{array}{ccc}& =& \frac{{e}^{{s}_{r}}}{{e}^{{s}_{r}}+{\varepsilon }^{2}}\sqrt{-1}\partial \overline{\partial }{s}_{r}+\frac{{\varepsilon }^{2}}{\left({e}^{{s}_{r}}+{\varepsilon }^{2}{\right)}^{2}}{e}^{{s}_{r}}\sqrt{-1}\partial {s}_{r}\wedge \overline{\partial }{s}_{r}\end{array}$
 $\begin{array}{ccc}& =& \frac{-{e}^{{s}_{r}}}{{e}^{{s}_{r}}+{\varepsilon }^{2}}{\Upsilon }_{r}^{W}+\frac{4{\varepsilon }^{2}}{\left({e}^{{s}_{r}}+{\varepsilon }^{2}{\right)}^{2}}\sqrt{-1}\partial \left({e}^{\frac{1}{2}{s}_{r}}\right)\wedge \overline{\partial }\left({e}^{\frac{1}{2}{s}_{r}}\right),\end{array}$
where the last equality follows since $\sqrt{-1}\partial \overline{\partial }{s}_{r}=\left[W\right]-{\Upsilon }_{r}^{W}$  and ${e}^{{s}_{r}}|W\equiv 0$  .
Let $0<\alpha <<1$  and set $a=\gamma -\alpha \xi .$  Observe that $a\ge 1.$  Moreover, we have
 $\begin{array}{ccc}-\sqrt{-1}\partial \overline{\partial }a& =& \alpha \sqrt{-1}\partial \overline{\partial }\xi \end{array}$
 $\begin{array}{ccc}& \ge & -\alpha {\Upsilon }_{r}^{W}+\alpha \frac{4{\varepsilon }^{2}}{\left({e}^{{s}_{r}}+{\varepsilon }^{2}{\right)}^{2}}\sqrt{-1}\partial \left({e}^{\frac{1}{2}{s}_{r}}\right)\wedge \overline{\partial }\left({e}^{\frac{1}{2}{s}_{r}}\right)\end{array}$
Now let $\tau =a+loga\text{and}A=\left(1+a{\right)}^{2}.$  Then $\tau \ge 1$  and we have $\partial \tau =\left(1+\frac{1}{a}\right)\partial a\text{and}\sqrt{-1}\partial \overline{\partial }\tau =\left(1+\frac{1}{a}\right)\sqrt{-1}\partial \overline{\partial }a-\frac{1}{{a}^{2}}\sqrt{-1}\partial a\wedge \overline{\partial }a,$  and thus $-\sqrt{-1}\partial \overline{\partial }\tau -\frac{\sqrt{-1}\partial \tau \wedge \overline{\partial }\tau }{A}=\left(1+\frac{1}{a}\right)\left(-\sqrt{-1}\partial \overline{\partial }a\right)\ge -\sqrt{-1}\partial \overline{\partial }a.$  It follows that
 $\begin{array}{ccc}& & \tau \sqrt{-1}\partial \overline{\partial }\left(\psi +\left(n+1\right)\lambda \right)-\sqrt{-1}\partial \overline{\partial }\tau -\frac{|\partial \tau {|}^{2}}{A}\end{array}$
 $\begin{array}{ccc}& \ge & c\sqrt{-1}\partial \overline{\partial }\kappa -\alpha {\Upsilon }_{r}^{W}+\alpha \frac{4{\varepsilon }^{2}}{\left({e}^{{s}_{r}}+{\varepsilon }^{2}{\right)}^{2}}\sqrt{-1}\partial \left({e}^{\frac{1}{2}{s}_{r}}\right)\wedge \overline{\partial }\left({e}^{\frac{1}{2}{s}_{r}}\right)\end{array}$
 $\begin{array}{ccc}& \ge & \alpha \frac{4{\varepsilon }^{2}}{\left({e}^{{s}_{r}}+{\varepsilon }^{2}{\right)}^{2}}\sqrt{-1}\partial \left({e}^{\frac{1}{2}{s}_{r}}\right)\wedge \overline{\partial }\left({e}^{\frac{1}{2}{s}_{r}}\right),\end{array}$
 $\begin{array}{ccc}& & \end{array}$
provided we take $\alpha$  sufficiently small. (For example, by the density hypothesis ( 5 ) we may take $\alpha =c$  .) Substituting into the twisted basic estimate ( 8 ), we obtain the following lemma.
Lemma 5.4. If $u$  is a $\left(0,1\right)$  -form in the domain of ${\overline{\partial }}^{*}$  , then
 $\begin{array}{ccc}{\int }_{\Omega }\left(\tau +A\right){|{\overline{\partial }}_{\psi }^{*}u|}^{2}{e}^{-\psi }{\omega }_{B}^{n}+{\int }_{\Omega }\tau {|\overline{\partial }u|}^{2}{e}^{-\psi }{\omega }_{B}^{n}& & \end{array}$
 $\begin{array}{ccc}\ge c{\int }_{\Omega }\frac{4{\varepsilon }^{2}}{\left({e}^{{s}_{r}}+{\varepsilon }^{2}{\right)}^{2}}{|\partial \left({e}^{\frac{1}{2}{s}_{r}}\right)\left(u\right)|}^{2}{e}^{-\psi }{\omega }_{B}^{n}.& & \end{array}$

An a priori estimate

We write ${\Omega }_{j}=B\left(0,\frac{j}{1+j}\right)$  . Suppose given $f\in {ℌ}^{2}\left(W,\kappa \right)$  . Since $W$  is a closed submanifold of $B$  , there exists a holomorphic extension $\stackrel{~}{f}$  of $f$  to $B$  . We write ${W}_{j}=W\cap {\Omega }_{j},{f}_{j}=f|{W}_{j}and{\stackrel{~}{f}}_{j}=\stackrel{~}{f}|{\Omega }_{j}.$  Observe that ${\int }_{{W}_{j}}|{f}_{j}{|}^{2}{e}^{-\kappa }{\omega }_{B}^{n-1}\le {\int }_{W}|f{|}^{2}{e}^{-\kappa }{\omega }_{B}^{n-1}<+\infty .$  Let $\chi \in {\mathcal{C}}_{0}^{\infty }\left(\left[0,1\right)\right)$  be such that $0\le \chi \le 1,\chi |\left[0,1/3\right]\equiv 1and{sup}_{\left[0,1\right)}|{\chi }^{\prime }|\le 2.$  We set ${\chi }_{\varepsilon }=\chi \left(\frac{{e}^{{\sigma }_{r}}}{{\varepsilon }^{2}}\right)$  and define the 1-forms ${\alpha }_{\varepsilon ,j}$  on ${\Omega }_{j}$  by ${\alpha }_{\varepsilon ,j}=\overline{\partial }{\chi }_{\varepsilon }{\stackrel{~}{f}}_{j}.$  Observe that ${\alpha }_{\varepsilon ,j}$  is supported on the tubular neighborhood ${\Omega }_{\varepsilon ,j}:={\Omega }_{j}\cap \left\{{e}^{\frac{1}{2}{s}_{r}}\le \varepsilon \right\}$  of ${W}_{j}$  in ${\Omega }_{j}$  . Thus, for a $\left(0,1\right)$  -form $u$  with compact support on ${\Omega }_{j}$  , we have
 $\begin{array}{ccc}{|\left({\alpha }_{\varepsilon ,j},u\right)|}^{2}& =& {|{\int }_{{\Omega }_{j}}〈{\chi }^{\prime }\left(\frac{{e}^{{s}_{r}}}{{\varepsilon }^{2}}\right)\frac{\overline{\partial }\left({e}^{{s}_{r}}\right)}{{\varepsilon }^{2}}{\stackrel{~}{f}}_{j},u〉{e}^{-\psi }{\omega }_{B}^{n}|}^{2}\end{array}$
 $\begin{array}{ccc}& \le & {\left(\frac{2}{{\varepsilon }^{2}}{\int }_{{\Omega }_{j}}|{\chi }^{\prime }\left(\frac{{e}^{{s}_{r}}}{{\varepsilon }^{2}}\right)||\overline{\partial }\left({e}^{\frac{1}{2}{s}_{r}}\right)\left(u\right)||{\stackrel{~}{f}}_{j}|{e}^{-\frac{1}{2}{s}_{r}}{e}^{-\kappa }{\omega }_{B}^{n}\right)}^{2}\end{array}$
 $\begin{array}{ccc}& \le & \frac{16}{{\varepsilon }^{4}}\left({\int }_{{\Omega }_{\varepsilon ,j}}|{\stackrel{~}{f}}_{j}{|}^{2}\frac{\left({e}^{{s}_{r}}+{\varepsilon }^{2}{\right)}^{2}}{4{\varepsilon }^{2}}{e}^{-\kappa }{\omega }_{B}^{n}\right)\end{array}$
 $\begin{array}{ccc}& & ×{\int }_{{\Omega }_{j}}\frac{4{\varepsilon }^{2}}{\left({e}^{{s}_{r}}+{\varepsilon }^{2}{\right)}^{2}}{|\partial \left({e}^{\frac{1}{2}{s}_{r}}\right)\left(u\right)|}^{2}{e}^{-\psi }{\omega }_{B}^{n}\end{array}$
 $\begin{array}{ccc}& \le & \frac{16}{c}{C}_{\varepsilon ,j}\left(||{T}^{*}u|{|}^{2}+||Su|{|}^{2}\right),\end{array}$
where $Tu=\overline{\partial }\left(\sqrt{\tau +A}u\right)andSu=\sqrt{\tau }\left(\overline{\partial }u\right),$  and ${C}_{\varepsilon ,j}=\frac{1}{{\varepsilon }^{2}}{\int }_{{\Omega }_{\varepsilon ,j}}|{\stackrel{~}{f}}_{j}{|}^{2}{e}^{-\phi }{\omega }_{B}^{n}.$  Thus the last inequality follows from Lemma  5.4 .
By the standard ${L}^{2}$  theory, we have the following ${L}^{2}$  twisted- $\overline{\partial }$  theorem.
Theorem 5. There exists a function ${h}_{j,\varepsilon }$  on ${\Omega }_{j}$  such that $T{h}_{\varepsilon ,j}={\alpha }_{\varepsilon ,j}and{\int }_{{\Omega }_{j}}|{h}_{\varepsilon ,j}{|}^{2}{e}^{-\psi }{\omega }_{B}^{n}\le \frac{16}{c}{C}_{\varepsilon ,j}.$  In particular, ${h}_{\varepsilon ,j}|W\equiv 0.$
• Proof. It remains only to prove the last assertion. But by Lemma  5.2 .3, ${e}^{-\psi }$  is not locally integrable at any point of $W$  , and thus the vanishing of ${h}_{\varepsilon ,j}|W$  follows.

Conclusion of the proof of Theorem  4

Observe first that by Lemma  5.3 there exists a constant $C>0$  such that ${limsup}_{\varepsilon \to 0}{C}_{\varepsilon ,j}\le C{\int }_{W}|f{|}^{2}{e}^{-\kappa }{\omega }_{B}^{n-1}.$  We set ${F}_{j,\varepsilon }={\chi }_{\varepsilon }{\stackrel{~}{f}}_{j}-\sqrt{\left(\tau +A\right)}{h}_{\varepsilon ,j}on{\Omega }_{j}.$  By Theorem  5 , ${F}_{\varepsilon ,j}$  is holomorphic on ${\Omega }_{j}$  and ${F}_{\varepsilon ,j}|{W}_{j}-{f}_{j}\equiv 0$  . Moreover there exists a constant $M$  such that ${\int }_{{\Omega }_{j}}|{F}_{\varepsilon ,j}{|}^{2}{e}^{-\kappa }{\omega }_{B}^{n}\le M\left(o\left(1\right)+{\int }_{W}|f{|}^{2}{e}^{-\kappa }{\omega }_{B}^{n-1}\right),\varepsilon \sim 0.$  Indeed, the integral ${\int }_{{\Omega }_{j}}|{\chi }_{\varepsilon }\stackrel{~}{f}{|}^{2}{e}^{-\kappa }{\omega }_{B}^{n}$  is negligible for small $\varepsilon$  , since the integrand is locally integrable and supported on a set of arbitrarily small measure. On the other hand,
 $\begin{array}{ccc}{\int }_{{\Omega }_{j}}\left(\tau +A\right)|{h}_{\varepsilon ,j}{|}^{2}{e}^{-\kappa }{\omega }_{B}^{n}& =& {\int }_{{\Omega }_{j}}{e}^{{s}_{r}}\left(\tau +A\right)|{h}_{\varepsilon ,j}{|}^{2}{e}^{-\psi }{\omega }_{B}^{n}\end{array}$
 $\begin{array}{ccc}& \le & \left({sup}_{{\Omega }_{j}}{e}^{{s}_{r}}\left(\tau +A\right)\right){\int }_{{\Omega }_{j}}|{h}_{\varepsilon ,j}{|}^{2}{e}^{-\psi }{\omega }_{B}^{n}\end{array}$
 $\begin{array}{ccc}& \le & {C}_{\varepsilon ,j}{e}^{\frac{\gamma }{\alpha }}\left({sup}_{{\Omega }_{j}}{e}^{-\frac{1}{\alpha }a}\left(a+log\left(a\right)+\left(1+a{\right)}^{2}\right)\right)\end{array}$
 $\begin{array}{ccc}& \le & K{C}_{\varepsilon ,j}\end{array}$
for some universal constant $K$  depending only on the density of $W$  . The last estimate holds since $a\ge 1$  .
By Corollary  2.3  ${F}_{j}={lim}_{\varepsilon \to 0}{F}_{\varepsilon ,j}$  exists, in the compact open topology, for each fixed $j$  . Moreover, since ${F}_{\varepsilon ,j}=f$  on ${W}_{j}$  and ${F}_{\varepsilon ,j}\to {F}_{j}$  pointwise, the same is true of ${F}_{j}$  . We thus have a sequence of holomorphic functions ${F}_{j}$  such that ${F}_{j}|W=f$  and ${\int }_{{\Omega }_{j}}|{F}_{j}{|}^{2}{e}^{-\phi }{\omega }^{n}\le C{\int }_{W}|f{|}^{2}{e}^{-\phi }{\omega }^{n-1}.$  Moreover, the constant $C$  does not depend on $j$  . Letting $j\to \infty$  , we obtain, for the same reasons as above, a holomorphic function $F$  that also agrees with $f$  on $W$  , and furthermore satisfies ${\int }_{B}|F{|}^{2}{e}^{-\kappa }{\omega }_{B}^{n}\le C{\int }_{W}|f{|}^{2}{e}^{-\kappa }.$  This completes the proof of Theorem  1

5.3 Sharper results using Cousin I

In this subsection we prove Theorem  1 by using the solution of Cousin I problems with ${L}^{2}$  -bounds, together with Ohsawa's $\overline{\partial }$  -theorem in the ball.

Local extensions

Let $\kappa$  be a plurisubharmonic function in $B$  such that $\sqrt{-1}\partial \overline{\partial }\kappa \le K{\omega }_{B}$  for some constant $K$  . denote by $H$  the hyperplane ${z}^{n}=0$  in ${\mathbb{C}}^{n}$  . Let $\Omega \subset \subset B$  be a domain containing the origin, and write ${H}_{\Omega }={P}_{H}\left(\Omega \right),$  where ${P}_{H}$  denotes the (Euclidean) orthogonal projection onto $H$  . Assume $\Omega$  is sufficiently small that there exists a complex disk ${D}_{\Omega }$  so that ${H}_{\Omega }×{D}_{\Omega }\subset \subset B.$  Let $R<1$  be such that ${H}_{\Omega }×{D}_{\Omega }\subset B\left(0,R\right).$  We have the following Lemma.
Lemma 5.5. There is a constant $C>0$  , depending only on $K$  and $R$  , such that for any holomorphic function $f\in \mathcal{O}\left({H}_{\Omega }\right)$  there is a function $F\in \mathcal{O}\left(\Omega \right)$  such that $F|H\cap \Omega =f$  and ${\int }_{\Omega }|F{|}^{2}{e}^{-\kappa }{\omega }_{B}^{n}\le C{\int }_{{H}_{\Omega }}|f{|}^{2}{e}^{-\kappa }{\omega }_{B}^{n-1},$  provided the right hand side is finite.
• Proof. In fact, we will construct $F\in \mathcal{O}\left({H}_{\Omega }×{D}_{\Omega }\right)$  . To this end, a result of N. Lindholm [L-01,Lemma6provides us with a function $u$  on $B\left(0,R\right)$  such that
• (1) $u$  is bounded in $B\left(0,R\right)$  by a constant depending only on $K$  and $R$  , and
• (2) $\sqrt{-1}\partial \overline{\partial }u=\sqrt{-1}\partial \overline{\partial }\kappa .$
Define $h=\kappa -u$  . Since $h$  is pluriharmonic there is a function $g\in \mathcal{O}\left(B\left(0,R\right)\right)$  such that $Reg=h$  . Writing $z=\left({z}^{\prime },{z}^{n}\right)\in {\mathbb{C}}^{n-1}×\mathbb{C}$  , let $F\left({z}^{\prime },{z}^{n}\right):=f\left({z}^{\prime }\right){e}^{\frac{1}{2}\left(g\left({z}^{\prime },{z}^{n}\right)-g\left({z}^{\prime },0\right)\right)}.$  Then $F\in \mathcal{O}\left({H}_{\Omega }×{D}_{\Omega }\right)$  , $F|{H}_{\Omega }=f$  , and we have  $\begin{array}{ccc}|F\left(z\right){|}^{2}{e}^{-\kappa \left(z\right)}& =& |f\left({z}^{\prime }\right){|}^{2}exp\left(h\left({z}^{\prime },{z}_{n}\right)-h\left({z}^{\prime },0\right)\right){e}^{-\kappa \left(z\right)}\end{array}$
 $\begin{array}{ccc}& =& |f\left({z}^{\prime }\right){|}^{2}exp\left(-u\left(z\right)+u\left({z}^{\prime },0\right)-\kappa \left({z}^{\prime },0\right)\right)\end{array}$
 $\begin{array}{ccc}& \le & M|f\left({z}^{\prime }\right){|}^{2}{e}^{-\kappa \left({z}^{\prime },0\right)}.\end{array}$
The result follows.
Take $\varepsilon >0$  sufficiently small so that ${D}_{\varepsilon }={B}^{n-1}\left(0,\varepsilon \right)×\mathbb{D}\left(0,\varepsilon \right)\subset B.$  Define $D\left(a,\varepsilon \right)={F}_{a}\left({D}_{\varepsilon }\right).$  Then we have the following Lemma.
Lemma 5.6. Let $\varepsilon <{\varepsilon }_{0}$  , where ${\varepsilon }_{0}$  is as in Definition  4.1 . There exists a constant $C$  , depending only on $\varepsilon$  and $K$  such that the following holds.
For all $a\in B$  and all $f\in \mathcal{O}\left({F}_{a}\left({B}^{n-1}\left(0,\varepsilon \right)\right)\right)$  with ${\int }_{{F}_{a}\left({B}^{n-1}\left(0,\varepsilon \right)\right)}|f{|}^{2}{e}^{-\kappa }{\omega }_{B}^{n}<+\infty ,$  there exists $F\in \mathcal{O}\left(D\left(a,\varepsilon \right)\right)$  such that $F|{F}_{a}\left({B}^{n-1}\left(0,\varepsilon \right)=fand{\int }_{D\left(a,\varepsilon \right)}|F{|}^{2}{e}^{-\kappa }{\omega }_{B}^{n}\le C{\int }_{{F}_{a}\left({B}^{n-1}\left(0,\varepsilon \right)\right)}|f{|}^{2}{e}^{-\kappa }{\omega }_{B}^{n}.$
• Proof. By uniform flatness and group invariance, it suffices to prove the result when $a=0$  , and then it is a consequence of Lemmas  5.5 and  4.2 .

Ohsawa's $\overline{\partial }$  theorem

We are going to solve a Cousin problem with ${L}^{2}$  estimates. However, for the estimates we need, the usual Hörmander Theorem will not suffice. Instead, we shall need the following Theorem of Ohsawa.
Theorem 6. (Ohsawa) There exists a constant ${C}_{0}>0$  , depending only on the density of $W$  , such that the following holds. For any $\overline{\partial }$  -closed $\left(0,1\right)$  -form $\alpha$  such that $I\left(\alpha \right):={\int }_{B}|\alpha {|}^{2}{e}^{-\left(\kappa +{s}_{r}\right)}{\omega }_{B}^{n}<+\infty$  there exists a function $U$  such that $\overline{\partial }U=\alpha and{\int }_{B}|U{|}^{2}{e}^{-\left(\kappa +{s}_{r}\right)}{\omega }_{B}^{n}\le {C}_{0}I\left(\alpha \right).$  In particular, $U|W\equiv 0$  .
Remark. We remind the reader that the norm $|\alpha |$  is calculated with respect to ${\omega }_{B}$  . Since $\alpha$  is a $1$  -form, this norm involves the inverse of the metric.
• Proof of Ohsawa's Theorem. With the work done in subsection  5.2 , it is now quite straightforward to give the proof of Ohsawa's Theorem. Indeed, Let $\tau ={e}^{\lambda },A=\frac{\tau }{\delta }and\psi =\kappa +{s}_{r}-\lambda .$  Then some straight-forward manipulation followed by substitution into ( 8 ) gives  $\begin{array}{ccc}& & {\int }_{\Omega }\left(1+\frac{1}{\delta }\right)\tau {|{\overline{\partial }}_{\psi }^{*}u|}^{2}{e}^{-\psi }{\omega }_{B}^{n}+{\int }_{\Omega }\tau {|\overline{\partial }u|}^{2}{e}^{-\psi }{\omega }_{B}^{n}\end{array}$
 $\begin{array}{ccc}& & \ge {\int }_{\Omega }\tau \left({\partial }_{\alpha }{\partial }_{\overline{\beta }}\left(\kappa +\left(n+\delta \right)\lambda +{s}_{r}\right)\right){u}^{\alpha }\overline{{u}^{\beta }}{e}^{-\psi }{\omega }_{B}^{n}\end{array}$
 $\begin{array}{ccc}& & \ge c{\int }_{\Omega }\tau |u{|}^{2}{e}^{-\psi }{\omega }_{B}^{n}\end{array}$
for some $c>0$  sufficiently small. The second inequality follows from the density hypothesis and taking $\delta >0$  sufficiently small.
Now, one has $|\left(\alpha ,u\right){|}^{2}\le {||\frac{1}{\sqrt{\tau }}\alpha ||}^{2}{||\sqrt{\tau }u||}^{2},$  and thus we can solve the equation $\overline{\partial }\left(\tau h\right)=\alpha$  with estimate $\int |h{|}^{2}{e}^{-\psi }{\omega }_{B}^{n}\le {C}_{\delta }{\int }_{\Omega }\frac{1}{\tau }|\alpha {|}^{2}{e}^{-\psi }{\omega }_{B}^{n}={C}_{\delta }{\int }_{\Omega }|\alpha {|}^{2}{e}^{-\left(\kappa +{s}_{r}\right)}{\omega }_{B}^{n}.$  Letting $U=\tau h$  completes the proof.

Construction of the interpolating function

Fix an $f\in {ℌ}^{2}\left(W,\kappa \right)$  , and let $\varepsilon <\frac{{\varepsilon }_{0}}{2}$  . Take a sequence $\left\{{w}_{j};j=1,2,...\right\}\subset W$  of distinct points such that ${N}_{\varepsilon }^{B}\left(W\right){\subset }^{\infty }{\bigcup }_{j=1}E\left({w}_{j},\varepsilon \right)$  and such that each point of $W$  is contained in at most a fixed, finite number of the sets $E\left({w}_{j},2\varepsilon \right)$  . (We say that the cover is uniformly locally finite.) For ease of exposition, write ${E}_{j}=E\left({w}_{j},\varepsilon \right)$  . Let ${E}_{0}=B-{N}_{\frac{\varepsilon }{2}}^{B}\left(W\right)$  so that $\left\{{E}_{j};j\ge 0\right\}$  is a uniformly locally finite open cover of the ball. Let $\left\{{\chi }_{j}\right\}$  be a partition of unity subordinate to this cover. We can also assume that ${\sum }_{j}|d{\chi }_{j}|\le C$  for some $C>0$  , where $|\cdot |$  is the ${\omega }_{B}$  -norm.
For $j\ge 1$  , let ${F}_{j}$  denote the extension to ${E}_{j}$  of the function $f|W\cap E\left({w}_{j},2\varepsilon \right)$  , given by Lemma  5.6 . We take ${F}_{0}=0$  . Since the covering $\left\{{E}_{j}\right\}$  is locally uniformly finite, we have ${\int }_{B}\sum {1}_{j}|{F}_{j}{|}^{2}{e}^{-\kappa }{\omega }_{B}^{n}\le {\int }_{W}|f{|}^{2}{e}^{-\kappa }{\omega }_{B}^{n-1},$  where ${1}_{j}$  denotes the characteristic function of ${E}_{j}$  .
Define ${G}_{ij}={F}_{i}-{F}_{j}.$  Then, with ${E}_{ij}={E}_{i}\cap {E}_{j}$  , we obtain ${G}_{ij}|W\cap {B}_{ij}\equiv 0,{G}_{ij}+{G}_{jk}+{G}_{ki}\equiv 0in{E}_{i}\cap {E}_{j}\cap {E}_{k}$  and ${\int }_{B}{\sum }_{ij}{1}_{i}{1}_{j}|{G}_{ij}{|}^{2}{e}^{-\kappa }{\omega }_{B}^{n}\lesssim {\int }_{W}|f{|}^{2}{e}^{-\kappa }{\omega }_{B}^{n-1}.$  Let ${\stackrel{~}{G}}_{i}={\sum }_{j}{\chi }_{j}{G}_{ij}\in {\mathcal{C}}^{\infty }\left({E}_{i}\right).$  Then ${\stackrel{~}{G}}_{i}|W\cap {E}_{i}\equiv 0$  and $\overline{\partial }\left({\stackrel{~}{G}}_{i}-{\stackrel{~}{G}}_{k}\right)={\sum }_{j}\overline{\partial }{\chi }_{j}\left({G}_{ij}-{G}_{kj}\right)={G}_{ik}\overline{\partial }\left({\sum }_{j}{\chi }_{j}\right)=0.$  Thus the $\overline{\partial }$  -closed 1-form $\alpha =\overline{\partial }{\stackrel{~}{G}}_{i}on{E}_{i}$  is well defined, and has norm ${\int }_{B}|\alpha {|}^{2}{e}^{-\kappa }{\omega }_{B}^{n}\le C{\int }_{W}|f{|}^{2}{e}^{-\kappa }{\omega }_{B}^{n-1}.$  But in fact, more holds.
Lemma 5.7. The form $\alpha$  satisfies the following estimate.
${\int }_{B}|\alpha {|}^{2}{e}^{-\left(\kappa +{s}_{r}\right)}{\omega }_{B}^{n}\le C{\int }_{W}|f{|}^{2}{e}^{-\kappa }{\omega }_{B}^{n-1}.$
Given Lemma  5.7 , Ohsawa's Theorem  6 provides us with a function $U$  such that ${\int }_{B}|U{|}^{2}{e}^{-\left(\kappa +{s}_{r}\right)}{\omega }_{B}^{n}\le C{\int }_{W}|f{|}^{2}{e}^{-\kappa }{\omega }_{B}^{n-1}.$  The finiteness of these integrals implies that $U|W\equiv 0$  . It follows that ${G}_{i}={\stackrel{~}{G}}_{i}-U$  is a holomorphic function satisfying ${G}_{ij}={G}_{i}-{G}_{j}and{G}_{i}|W\equiv 0.$  Thus the function $\Phi ={F}_{i}-{G}_{i}on{E}_{i}$  satisfies $\Phi |W=fand{\int }_{B}|\Phi {|}^{2}{e}^{-\kappa }{\omega }_{B}^{n}\lesssim {\int }_{W}|f{|}^{2}{e}^{-\kappa }{\omega }_{B}^{n-1}.$  To establish Lemma  5.7 , and thus complete the proof of Theorem  1 we shall need the following result, which will also be used later on.
Lemma 5.8. Let $\phi$  be a subharmonic function on the unit disk $\mathbb{D}$  . Then there exist a positive constant $K$  and a holomorphic function on $G\in \mathcal{O}\left(\mathbb{D}\left(0,\frac{1}{2}\right)\right)$  such that $G\left(0\right)=0$  and ${sup}_{\mathbb{D}\left(0,\frac{1}{2}\right)}|\phi -\phi \left(0\right)-2ReG|\le K.$  Moreover, if $\phi$  depends smoothly on a parameter, then so does $G$  .
The proof of this lemma can be found in [BO-95. It follows from Lemma  5.8 that if $g$  is holomorphic on the unit disk $\mathbb{D}$  and $\psi$  is a weight such that $\Delta \psi \le K$  , then ${\int }_{\mathbb{D}}|g{|}^{2}{e}^{-\psi }\le C{\int }_{\frac{1}{2}<|z|<1}|g{|}^{2}{e}^{-\psi }.$  Indeed, the result is an elementary application of Bergman's inequality (i.e., the ${L}^{2}$  Cauchy estimates) when $\psi =0$  . In the general case, Lemma  5.8 reduces us to this case, since we may replace $g$  by $gh$  , and $|gh{|}^{2}\simeq |g{|}^{2}{e}^{-\psi }.$  Of course, this also applies to $g$  that depends on a parameter.
• Proof of Lemma  5.7 . We claim that  $\begin{array}{ccc}{\int }_{{E}_{ij}}|{G}_{ij}{|}^{2}{e}^{-\left(\kappa +{s}_{r}\right)}{\omega }_{B}^{n}& \lesssim & {\int }_{{E}_{ij}-{N}_{\frac{1}{2}\varepsilon }\left(W\right)}|{G}_{ij}{|}^{2}{e}^{-\left(\kappa +{s}_{r}\right)}{\omega }_{B}^{n}\end{array}$
 $\begin{array}{ccc}& \simeq & {\int }_{{E}_{ij}-{N}_{\frac{1}{2}\varepsilon }\left(W\right)}|{G}_{ij}{|}^{2}{e}^{-\kappa }{\omega }_{B}^{n}\end{array}$
 $\begin{array}{ccc}& \lesssim & {\int }_{{E}_{ij}\cap W}|f{|}^{2}{e}^{-\kappa }{\omega }_{B}^{n-1}.\end{array}$
The only inequality requiring proof is the first. To see how it follows, note that by Lemma  5.1  ${s}_{r}=log|T{|}^{2}-{\lambda }_{r}$  , where ${\lambda }_{r}\left(z\right)=\frac{1}{{V}_{n}\left(r\right)}{\int }_{E\left(z,r\right)}log|T{|}^{2}{\omega }_{B}^{n}.$  Letting ${\psi }_{r}=\kappa -{\lambda }_{r},$  we have $|{G}_{ij}{|}^{2}{e}^{-\left(\kappa +{s}_{r}\right)}={|\frac{{G}_{ij}}{T}|}^{2}{e}^{-{\psi }_{r}}$  and $\sqrt{-1}\partial \overline{\partial }{\psi }_{r}\simeq {\omega }_{B}.$  Now, ${G}_{ij}/T$  is holomorphic in ${E}_{ij}$  . We let ${W}_{ij}={E}_{ij}\cap W$  , and observe that ${E}_{ij}\simeq {W}_{ij}×\mathbb{D}\left(0,\varepsilon \right)$  , which allows us to integrate along the fibers, where we can apply the one dimensional calculations above.

6 Sampling

6.1 Restriction from tubes and the upper inequality

Proposition 6.1. Let $W$  be a uniformly flat smooth hypersurface. Then there exists a constant $C>0$  such that for all $\varepsilon >0$  sufficiently small and all $\Phi \in {H}^{2}\left({N}_{\varepsilon }^{B}\left(W\right),\kappa \right)$  one has $C{\varepsilon }^{2}{\int }_{W}|\Phi {|}^{2}{e}^{-\kappa }{\omega }_{B}^{n-1}\le {\int }_{{N}_{\varepsilon }^{B}\left(W\right)}|\Phi {|}^{2}{e}^{-\kappa }{\omega }_{B}^{n}.$
• Proof. Let $D\left(0,\varepsilon \right)=\left\{\left({z}^{\prime },{z}^{n}\right)\in {\mathbb{C}}^{n-1}×\mathbb{C};|{z}^{\prime }|\le \varepsilon ,|{z}^{n}|<\varepsilon \right\}.$  By uniform flatness, it suffices to prove that for some $\varepsilon >0$  and all $a\in W$  , $C{\varepsilon }^{2}{\int }_{D\left(0,\varepsilon \right)\cap {F}_{a}\left(W\right)}|\Phi {|}^{2}{e}^{-\kappa }{\omega }_{B}^{n}\le {\int }_{D\left(0,\varepsilon \right)}|\Phi {|}^{2}{e}^{-\kappa }.$  After a change of variables, we may assume ${F}_{a}\left(W\right)\subset {\mathbb{C}}^{n-1}×\left\{0\right\}$  .
Now, by Lemma  5.8 there exists a function $G$  , holomorphic in ${z}^{n}$  , such that $G\left({z}^{\prime },0\right)|\equiv 0and{e}^{-\kappa \left({z}^{\prime },0\right)+2ReG\left({z}^{\prime },{z}^{n}\right)}\le c{e}^{-\kappa \left({z}^{\prime },{z}^{n}\right)}$  for some $c>0$  . We then have  $\begin{array}{ccc}{\int }_{B\left(0,\varepsilon \right)\cap {\mathbb{C}}^{n-1}×\left\{0\right\}}|\Phi {|}^{2}{e}^{-\kappa }{\omega }_{B}^{n}& =& {\int }_{B\left(0,\varepsilon \right)\cap {\mathbb{C}}^{n-1}×\left\{0\right\}}|\Phi {e}^{G}{|}^{2}{e}^{-\kappa }{\omega }_{B}^{n}\end{array}$
 $\begin{array}{ccc}& \le & \frac{1-{\varepsilon }^{2}}{2\pi n{\varepsilon }^{2}}{\int }_{D\left(0,\varepsilon \right)}|\Phi {e}^{G}{|}^{2}{e}^{-\kappa \left({z}^{\prime },0\right)}{\omega }_{B}^{n}\end{array}$
 $\begin{array}{ccc}& =& \frac{1-{\varepsilon }^{2}}{2\pi n{\varepsilon }^{2}}{\int }_{D\left(0,\varepsilon \right)}|\Phi {|}^{2}{e}^{-\kappa \left({z}^{\prime },0\right)+2ReG}{\omega }_{B}^{n}\end{array}$
 $\begin{array}{ccc}& \le & \frac{1}{C{\varepsilon }^{2}}{\int }_{D\left(0,\varepsilon \right)}|\Phi {|}^{2}{e}^{-\kappa \left({z}^{\prime },{z}^{n}\right)}{\omega }_{B}^{n}\end{array}$
 $\begin{array}{ccc}& & \end{array}$
The first inequality follows from the sub-mean value property for radial measures in the disk (see also Corollary  4 ) together with ${\omega }_{B}^{n}\left(z\right)=n{\omega }_{B}^{n-1}\left({z}^{\prime }\right)\otimes {\omega }_{B}\left({z}^{n}\right)and{\int }_{\mathbb{D}\left(0,\varepsilon \right)}{\omega }_{B}=\frac{2\pi {\varepsilon }^{2}}{1-{\varepsilon }^{2}}.$  This completes the proof.
Corollary 6.2. If $W$  is a uniformly flat hypersurface then there exists a constant $M$  such that for all $\Phi \in {H}^{2}\left(B,\kappa \right)$  , ${\int }_{W}|\Phi {|}^{2}{e}^{-\kappa }{\omega }_{B}^{n-1}\le M{\int }_{B}|\Phi {|}^{2}{e}^{-\kappa }{\omega }_{B}^{n}.$

6.2 Regularization of the singular function ${s}_{r}$

Consider the function ${s}_{r,\varepsilon }\left(z\right):=\frac{1}{{V}_{n}\left(\varepsilon \right)}{\int }_{E\left(z,\varepsilon \right)}{s}_{r}{\omega }_{B}^{n}.$  In this section we prove the following result.
Lemma 6.3. The function ${s}_{r,\varepsilon }$  enjoys the following properties.
• (1) ${lim}_{\varepsilon \to 0}\sqrt{-1}\partial \overline{\partial }{s}_{r,\varepsilon }=\left[W\right]-{\Upsilon }_{r}^{W}.$
• (2) For each $r$  there exists a constant ${C}_{r}$  such that if $0<\varepsilon \le {\varepsilon }_{1}<<1$  and $dist\left(z,W\right)<\varepsilon$  , then $log{\varepsilon }^{2}-{C}_{r}\le {s}_{r,\varepsilon }\le 0.$
• Proof. Property 1 is a standard consequence of regularization. Property 2 may be established locally, and using group invariance and uniform flatness, we need only consider the case $z=0$  . But then by the calculation in the proof of 3 of Lemma  5.2 we may assume that ${s}_{r}=log|{\zeta }^{n}{|}^{2}$  , and thus 2 follows by simple integration .

6.3 The proof of Theorem  2

A positivity lemma

The key idea behind the proof of the lower sampling inequality is a certain positivity lemma, which we now state and prove.
Lemma 6.4. Let $\theta$  be a positive $\left(n-1,n-1\right)$  -form in $B$  such that for some weight $\psi$  and each $h\in {\mathsf{ℌ}}^{2}\left(B,\psi \right)$  , ${\int }_{B}|h{|}^{2}{e}^{-\psi }\sqrt{-1}\partial \overline{\partial }\theta <+\infty .$  Then ${\int }_{B}|h{|}^{2}{e}^{-\psi }\sqrt{-1}\partial \overline{\partial }\psi \wedge \theta \ge -{\int }_{B}|h{|}^{2}{e}^{-\psi }\sqrt{-1}\partial \overline{\partial }\theta .$
• Proof. Letting $S=|h{|}^{2}{e}^{-\psi }$  , one calculates that $\frac{\sqrt{-1}\partial \overline{\partial }S}{S}=\frac{\sqrt{-1}\partial S\wedge \overline{\partial }S}{{S}^{2}}+\sqrt{-1}\partial \overline{\partial }log|h{|}^{2}-\sqrt{-1}\partial \overline{\partial }\psi ,$  and thus $\sqrt{-1}\partial \overline{\partial }S\wedge \theta \ge -S\sqrt{-1}\partial \overline{\partial }\psi \wedge \theta .$  Let $f:\mathbb{R}\to \left[0,1\right]$  be a smooth function supported on $\left(-\infty ,3/4\right]$  such that $f|\left(-\infty ,0\right]\equiv 1$  . Consider the function ${\chi }_{a}\left(z\right)=f\left(\frac{1-|z{|}^{2}}{-a}+1\right),a>0.$  Then  $\begin{array}{ccc}{\int }_{B}\sqrt{-1}\partial \overline{\partial }S\wedge \theta & =& {lim}_{a\to 0+}{\int }_{B}{\chi }_{a}\left(z\right)\sqrt{-1}\partial \overline{\partial }S\wedge \theta \end{array}$
 $\begin{array}{ccc}& =& {lim}_{a\to 0+}{\int }_{B}S\sqrt{-1}\partial \overline{\partial }\left({\chi }_{a}\left(z\right)\wedge \theta \right)\end{array}$
 $\begin{array}{ccc}& =& {lim}_{a\to 0+}\left({\int }_{B}{\chi }_{a}\left(z\right)S\sqrt{-1}\partial \overline{\partial }\theta +O\left(a\right)\right)\end{array}$
 $\begin{array}{ccc}& =& {\int }_{B}S\sqrt{-1}\partial \overline{\partial }\theta ,\end{array}$
where the second equality follows from the Green-Stokes identity ( 2 ).

Conclusion of the proof of Theorem  2

Let $\psi =\kappa +n\lambda +\alpha {s}_{r,\varepsilon }.$  In view of Lemmas  6.3 and  3.1 .2, for some $0<<\alpha <1$  , $c>0$  and $\theta \in {\mathfrak{P}}_{W}\left(B\right)$  we have $\sqrt{-1}\partial \overline{\partial }\psi \wedge \theta \le -c{e}^{n\lambda }{\omega }_{B}^{n}+C\left[W{\right]}_{\varepsilon }\wedge {\omega }_{B}^{n-1},$  where $\left[W{\right]}_{\varepsilon }$  denotes the regularization of the current $\left[W\right]$  in the manner of Lemma  6.3 .
Let $F\in {\mathsf{ℌ}}^{2}\left(B,\kappa \right)$  . Then
 $\begin{array}{ccc}{\int }_{B}|F{|}^{2}{e}^{-\kappa }{\omega }_{B}^{n}& \le & {\int }_{B}|F{|}^{2}{e}^{-\psi }{\omega }_{B}^{n}\end{array}$
 $\begin{array}{ccc}& \le & C{\int }_{B}|F{|}^{2}{e}^{-\psi }\left[W{\right]}_{\varepsilon }\wedge {\omega }_{B}^{n-1}-C{\int }_{B}|F{|}^{2}{e}^{-\psi }\sqrt{-1}\partial \overline{\partial }\psi \wedge \theta \end{array}$
 $\begin{array}{ccc}& \le & C{\int }_{B}|F{|}^{2}{e}^{-\psi }\left[W{\right]}_{\varepsilon }\wedge {\omega }_{B}^{n-1},\end{array}$
where the last inequality follows from Lemma  6.4 and the definition of ${\mathfrak{P}}_{W}\left(B\right)$  . Thus we have
 $\begin{array}{ccc}{\int }_{B}|F{|}^{2}{e}^{-\kappa }{\omega }_{B}^{n}& \le & C{\int }_{B}|F{|}^{2}{e}^{-\psi }\left[W{\right]}_{\varepsilon }\wedge {\omega }_{B}^{n-1}\end{array}$
 $\begin{array}{ccc}& \le & \frac{C}{{\varepsilon }^{2}}{\int }_{{N}_{\varepsilon }\left(W\right)}|F{|}^{2}{e}^{-\psi }{\omega }_{B}^{n}\end{array}$
 $\begin{array}{ccc}& \le & \frac{C}{{\varepsilon }^{2+2\alpha }}{\int }_{{N}_{\varepsilon }\left(W\right)}|F{|}^{2}{e}^{-\kappa }{\omega }_{B}^{n}.\end{array}$
 $\begin{array}{ccc}& & \end{array}$
Our next task is to compare ${\int }_{{N}_{\varepsilon }\left(W\right)}|F{|}^{2}{e}^{-\kappa }{\omega }_{B}^{n}with{\int }_{W}|F{|}^{2}{e}^{-\kappa }{\omega }_{B}^{n-1}.$  To do this, we cover $W$  by Bergman-Green balls $\left\{E\left(p,{\varepsilon }_{p}\right){\right\}}_{p\in \mathfrak{W}},$  where $\mathfrak{W}\subset W$  is a discrete set that is uniformly separated with respect to the Bergman-Green distance. We now employ Lemma  5.8 once more to obtain a function ${N}_{\varepsilon }\left(W\right)\cap E\left(p,{\varepsilon }_{p}\right)\sim =\left(W\cap E\left(p,{\varepsilon }_{p}\right)\right)×\mathbb{D}\left(0,\varepsilon \right)\ni \left(x,t\right)↦{H}_{p}\left(x,t\right)\in \mathbb{C}$  that is holomorphic in $t$  and satisfies ${H}_{p}\left(x,0\right)=0and|2Re\left({H}_{p}\left(x,t\right)\right)+\kappa \left(x,0\right)-\kappa \left(x,t\right)|\le C$  where $C$  is an absolute constant depending only on $\sqrt{-1}\partial \overline{\partial }\kappa .$  Let ${F}_{p}=F{e}^{-{H}_{p}}$  . By Taylor's Theorem, for each $x$  we have
 $\begin{array}{ccc}& & |{F}_{p}\left(x,t\right){|}^{2}\le C|F\left(x,0\right){|}^{2}+{\varepsilon }^{2}{sup}_{|t|\le \varepsilon }{|\frac{\partial {F}_{p}}{\partial t}|}^{2}\end{array}$
We then obtain
 $\begin{array}{ccc}& & {\int }_{{N}_{\varepsilon }\left(W\right)\cap E\left(p,{\varepsilon }_{p}\right)}|F{|}^{2}{e}^{-\kappa }{\omega }_{B}^{n}\end{array}$
 $\begin{array}{ccc}& \le & {\int }_{{N}_{\varepsilon }\left(W\right)\cap E\left(p,{\varepsilon }_{p}\right)}|{F}_{p}{|}^{2}{e}^{-\kappa \left(x,0\right)}{\omega }_{B}^{n}\end{array}$
 $\begin{array}{ccc}& \le & {C}_{1}{\varepsilon }^{2}{\int }_{W\cap E\left(p,{\varepsilon }_{p}\right)}|F{|}^{2}{e}^{-\kappa }{\omega }_{B}^{n-1}+{\varepsilon }^{2}{\int }_{{N}_{\varepsilon }\left(W\right)\cap E\left(p,{\varepsilon }_{p}\right)}{sup}_{|t|\le \varepsilon }{|\frac{\partial {F}_{p}}{\partial t}|}^{2}{e}^{-\kappa \left(x,0\right)}{\omega }_{B}^{n}\end{array}$
 $\begin{array}{ccc}& \le & {C}_{1}{\varepsilon }^{2}{\int }_{W\cap E\left(p,{\varepsilon }_{p}\right)}|F{|}^{2}{e}^{-\kappa }{\omega }_{B}^{n-1}+{\varepsilon }^{4}{\int }_{W\cap E\left(p,{\varepsilon }_{p}\right)}{sup}_{|t|\le \varepsilon }{|\frac{\partial {F}_{p}}{\partial t}|}^{2}{e}^{-\kappa \left(x,0\right)}{\omega }_{B}^{n-1}\end{array}$
 $\begin{array}{ccc}& \le & {C}_{1}{\varepsilon }^{2}{\int }_{W\cap E\left(p,{\varepsilon }_{p}\right)}|F{|}^{2}{e}^{-\kappa }{\omega }_{B}^{n-1}+C{\varepsilon }^{4}{\int }_{{N}_{{\varepsilon }_{o}}\left(W\right)\cap E\left(p,{\varepsilon }_{p}\right)}|{F}_{p}{|}^{2}{e}^{-\kappa \left(x,0\right)}{\omega }_{B}^{n}\end{array}$
 $\begin{array}{ccc}& \le & {\varepsilon }^{2}{\int }_{W\cap E\left(p,{\varepsilon }_{p}\right)}|F{|}^{2}{e}^{-\kappa }{\omega }^{n-1}+C{\varepsilon }^{4}{\int }_{{N}_{{\varepsilon }_{o}}\left(W\right)\cap E\left(p,{\varepsilon }_{p}\right)}|F{|}^{2}{e}^{-\kappa }{\omega }_{B}^{n},\end{array}$
where $\varepsilon <{\varepsilon }_{o}/2$  and we have used the Cauchy estimates in the penultimate inequality.
Combining all of this, and summing over $p\in \mathfrak{W}$  , we obtain ${\int }_{B}|F{|}^{2}{e}^{-\kappa }{\omega }_{B}^{n}\le \frac{C}{{\varepsilon }^{2\alpha }}{\int }_{W}|F{|}^{2}{e}^{-\kappa }{\omega }_{B}^{n-1}+{C}^{\prime }{\varepsilon }^{2-2\alpha }{\int }_{B}|F{|}^{2}{e}^{-\kappa }{\omega }_{B}^{n},$  which establishes the left inequality in ( 1 ) as soon as we take $\varepsilon$  small enough. The right inequality was already established in Corollary  6.2 , and thus the proof of Theorem  2 is complete. References

1. Berndtsson, B.; Ortega Cerdà, J., On interpolation and sampling in Hilbert spaces of analytic functions. J. Reine Angew. Math. 464 (1995), 109–128.
2. Lindholm, Niklas, Sampling in weighted ${L}^{p}$  spaces of entire functions in ${\mathbb{C}}^{n}$  and estimates of the Bergman kernel. J. Funct. Anal. 182 (2001), no. 2, 390–426.
3. McNeal, J., On large values of ${L}^{2}$  holomorphic functions. Math. Res. Lett. 3 (1996), no. 2, 247–259.
4. Ohsawa, T., On the extension of ${L}^{2}$  holomorphic functions. V. Effects of generalization. Nagoya Math. J. 161 (2001) 1-21.
5. Ortega-Cerdà, J., Schuster, A., Varolin, D., Interpolation and Sampling hypersurfaces for the Bargmann-Fock Space on ${\mathbb{C}}^{n}$  , Preprint, 2004.
6. Rudin, W., Function Theory in the Unit Ball of ${\mathbb{C}}^{n}$  . Springer-Verlag, 1980
7. Siu, Y.-T., Complex-analyticity of harmonic maps, vanishing and Lefschetz theorems. J. Diff. Geom. 17 (1982), no. 1, 55–138.
8. Siu, Y.-T., The Fujita conjecture and the extension theorem of Ohsawa-Takegoshi. Geometric complex analysis (Hayama, 1995), 577–592, World Sci. Publishing, River Edge, NJ, 1996.
9. Stoll, M. Invariant potential theory in the unit ball of ${C}^{n}$  . London Mathematical Society Lecture Note Series, 199. Cambridge University Press, Cambridge, 1994.

center Department of Mathematics, University of Illinois, Urbana, IL 61801, forgacs@uiuc.edu dror@uiuc.edu