Equivalence of the sharp effectiveness results of strong openness property

Abstract.

In this paper, we show the equivalence of the sharp effectiveness results of the strong openness property of multiplier ideal sheaves obtained in [2] using $\xi-$Bergman kernels and in [8] using minimal $L^{2}$ integrals.

1. Introduction

There are two distinct sharp effectiveness results of the strong openness property of multiplier ideal sheaves, proven in [8] by Guan and in [2] by Bao-Guan, using minimal $L^{2}$ integrals and $\xi$-Bergman kernels respectively. In the present paper, we will prove that these two effectiveness results are equivalent.

1.1. Notations and conventions

Before we discuss the background and motivation of the present paper, we recall some notations and conventions.

1.1.1. $\xi-$Bergman kernel

We recall the $\xi-$Bergman kernel (a version of generalized Bergman kernel) defined in [2] in the following.

Denote

$$\ell_{1}^{(n)}:=\Big{\{}(\xi_{\alpha})_{\alpha\in\mathbb{N}^{n}}\colon\xi_{\alpha}\in\mathbb{C},\ \sum_{\alpha\in\mathbb{N}^{n}}|\xi_{\alpha}|\rho^{|\alpha|}<+\infty,\ \forall\rho>0\Big{\}}.$$

We also denote

$$\ell_{0}^{(n)}:=\Big{\{}(\xi_{\alpha})_{\alpha\in\mathbb{N}^{n}}\colon\xi_{\alpha}\in\mathbb{C},\ \exists k\in\mathbb{N},\ \text{s.t.}\ \xi_{\alpha}=0,\ \forall\alpha\ \text{with}\ |\alpha|>k\Big{\}}\subseteq\ell_{1}^{(n)},$$

in the present paper. $\ell_{1}^{(n)}$ can be seen as $\mathcal{O}(\mathbb{C}^{n})$, the space of entire functions on $\mathbb{C}^{n}$, while $\ell^{(n)}_{0}$ can be seen as $\mathbb{C}[z_{1},\ldots,z_{n}]$, the space of polynomials on $\mathbb{C}^{n}$.

Let $D$ be a domain in $\mathbb{C}^{n}$. For any $z_{0}\in D$, $\xi=(\xi_{\alpha})\in\ell_{1}^{(n)}$, and $f(z)=\sum_{\alpha\in\mathbb{N}^{n}}c_{\alpha}(z-z_{0})^{\alpha}\in\mathcal{O}_{z_{0}}$, denote

$$(\xi\cdot f)(z_{0}):=\sum_{\alpha\in\mathbb{N}^{n}}\xi_{\alpha}\frac{(D^{\alpha}f)(z_{0})}{\alpha!}=\sum_{\alpha\in\mathbb{N}^{n}}\xi_{\alpha}c_{\alpha}.$$

Definition 1.1 (see [2]).

Let $\xi\in\ell_{1}^{(n)}$. The Bergman kernel with respect to $\xi$ on $D$ is defined as:

$$K_{\xi,D}(z):=\sup\big{\{}|(\xi\cdot f)(z)|^{2}\colon f\in A^{2}(D),\ \|f\|_{D}\leq 1\big{\}},$$

where $A^{2}(D)=L^{2}(D)\cap\mathcal{O}(D)$ is the Bergman space, and $\|f\|^{2}_{D}=\int_{D}|f|^{2}$.

It can be checked that $K_{\xi,D}(z)<+\infty$ (see [2, Lemma 2.3]).

Suppose $I$ is an ideal of $\mathcal{O}_{o}$, the ring of holomorhpic function germs on the origin $o\in\mathbb{C}^{n}$. Denote

$$\tilde{\ell}_{I}:=\big{\{}\xi\in\ell_{1}^{(n)}:(\xi\cdot f)(o)=0,\ \forall(f,o)\in I\big{\}},$$

and $\ell_{I}:=\tilde{\ell}_{I}\setminus\{0\}$. $\tilde{\ell}_{I}$ is a linear subspace of $\ell^{(n)}_{1}$, where it is not trivial when $I$ is a proper ideal of $\mathcal{O}_{o}$.

Now let $D$ be a domain in $\mathbb{C}^{n}$ containing the origin $o$, $I$ a proper ideal of $\mathcal{O}_{o}$, and $(F,o)\in\mathcal{O}_{o}$. Recall the notation

$$B(F,I,D):=\sup_{\xi\in\ell_{I}}\frac{|(\xi\cdot F)(o)|^{2}}{K_{\xi,D}(o)}.$$

Proposition 1.2 (Proposition 4.6 in [2]).

$B(F,I,D)=0$ if and only if $(F,o)\in I$.

This proposition was proved in [2] by the separation theorem in the functional analysis theory.

In the present paper, we also denote

$$B^{\circ}(F,I,D):=\sup_{\xi\in\ell_{I}\cap\ell_{0}^{(n)}}\frac{|(\xi\cdot F)(o)|^{2}}{K_{\xi,D}(o)}.$$

Clearly, we have $B^{\circ}(F,I,D)\leq B(F,I,D)$.

1.1.2. Minimal $L^{2}$ integral

The second author of the present paper introduced the minimal $L^{2}$ integrals in [8] to establish a sharp effectiveness result of the strong openness property. We recall the definition and notation of the minimal $L^{2}$ integral below.

Again let $D$ be a domain in $\mathbb{C}^{n}$ containing the origin $o$, $I$ a proper ideal of $\mathcal{O}_{o}$, and $(F,o)\notin I$. Recall the notation of minimal $L^{2}$ integral ([8])

$$C_{F,I}(D):=\inf\big{\{}\|\tilde{F}\|_{D}^{2}:\tilde{F}\in A^{2}(D),\ (\tilde{F}-F,o)\in I\big{\}},$$

and if the set in the right hand is empty, then denote $C_{F,I}(D)=+\infty$. By Montel’s theorem, if $C_{F,I}(D)<+\infty$, there exists a unique $G\in A^{2}(D)$ with $\|G\|_{D}^{2}=C_{F,I}(D)$ and $(G-F,o)\in I$. One can also note that $C_{F,I}(D)=0$ if and only if $(F,o)\in I$ (see [8]).

Denote

$$A^{2}(D,I):=\big{\{}f\in A^{2}(D):(f,o)\in I\big{\}},$$

which is a closed subspace of $A^{2}(D)$. Then there exists a unique closed subspace $A^{2}(D,I)^{\perp}$ of $A^{2}(D)$, which is the orthogonal complement of $A^{2}(D,I)$ in $A^{2}(D)$. Moreover, for any $F\in A^{2}(D)\setminus A^{2}(D,I)$, the unique $G\in A^{2}(D)$ with $\|G\|_{D}^{2}=C_{F,I}(D)$ and $(G-F,o)\in I$ is actually the image of $F$ under the orthogonal projection from $A^{2}(D)$ to $A^{2}(D,I)^{\perp}$.

With simple computations, one can check that $B(F,I,D)\leq C_{F,I}(D)$ holds. More precisely, assume that $\tilde{F}\in A^{2}(D)$ satisfying $(\tilde{F}-F,o)\in I$ and $\|\tilde{F}\|_{D}^{2}=C_{F,I}(D)<+\infty$. Then $(\xi\cdot F)(o)=(\xi\cdot\tilde{F})(o)$ for any $\xi\in\ell_{I}$, and thus

$$B(F,I,D)=\sup_{\xi\in\ell_{I}}\frac{|(\xi\cdot F)(o)|^{2}}{K_{\xi,D}(o)}\leq\sup_{\begin{subarray}{c}\xi\in\ell_{I},\\ (\xi\cdot F)(o)\neq 0\end{subarray}}\frac{|(\xi\cdot F)(o)|^{2}}{|(\xi\cdot\tilde{F})(o)|^{2}/\|\tilde{F}\|_{D}^{2}}=\|\tilde{F}\|_{D}^{2}=C_{F,I}(D).$$

1.2. Background and motivation

Let $D$ be a pseudoconvex domain in $\mathbb{C}^{n}$ containing the origin $o$, $\varphi$ a negative plurisubharmonic function on $D$ with $\varphi(o)=-\infty$, and $F$ a holomorphic function on $D$.

Recall the jumping number

$$c_{o}^{F}(\varphi):=\sup\big{\{}c\geq 0:(F,o)\in\mathcal{I}(c\varphi)_{o}\big{\}},$$

and the multiplier ideal

$$\mathcal{I}(c\varphi)_{o}:=\big{\{}(G,o)\in\mathcal{O}_{o}:|G|^{2}e^{-c\varphi}\ \text{is\ }L^{1}\ \text{integrable near\ }o\big{\}},\ \forall c>0.$$

The strong openness property of multiplier ideal sheaves ([10]) shows that: if $\int_{D}|F|^{2}e^{-\varphi}<+\infty$, then there must exist some $p>1$ such that $(F,o)\in\mathcal{I}(p\varphi)_{o}$, or equivalently, $(F,o)\notin\mathcal{I}(c_{o}^{F}(\varphi)\varphi)_{o}$. The effectiveness of the strong openness property was established in [11]. Moreover, [8] gave the following sharp effectiveness result.

Theorem 1.3 ([8]).

Suppose $\int_{D}|F|^{2}e^{-\varphi}<+\infty$. Then for any $p>1$ satisfying

$$\frac{p}{p-1}>\frac{\int_{D}|F|^{2}e^{-\varphi}}{C_{F,\mathcal{I}_{+}(c_{o}^{F}(\varphi)\varphi)_{o}}(D)},$$

we have $(F,o)\in\mathcal{I}(p\varphi)_{o}$.

Afterwards, [2] proved the following result, which gave the optimal $L^{2}$ extension approach to the effectiveness result of the strong openness property.

Theorem 1.4 ([2]).

Suppose $\int_{D}|F|^{2}e^{-\varphi}<+\infty$. Then for any $p>1$ satisfying

$$\frac{p}{p-1}>\frac{\int_{D}|F|^{2}e^{-\varphi}}{B(F,\mathcal{I}_{+}(c_{o}^{F}(\varphi)\varphi)_{o},D)},$$

we have $(F,o)\in\mathcal{I}(p\varphi)_{o}$.

In [2], Theorem 1.4 was proved for bounded pseudoconvex domains $D$, but it is easy to generalize the result to all pseudoconvex domains.

These two effectiveness results of the strong openness property were both shown to be sharp in [8] and [2], respectively. Theorem 1.3 was established by estimating the minimal $L^{2}$ integrals on the sublevel sets of plurisubharmonic functions in [8], while Theorem 1.4 was proved using the log-plurisubharmonicity of fiberwise $\xi$-Bergman kernels in [2]. There are differences between minimal $L^{2}$ integrals and $\xi$-Bergman kernels. For example, in [6], Bao-Guan-Yuan showed that the log-convexity of the minimal $L^{2}$ integrals on the sublevel sets of plurisubharmonic functions does not hold generally, but this property is true for $\xi$-Bergman kernels. This leads naturally to the following question:

Question 1.5.

Can one show that effectiveness results of the strong openness property in Theorem 1.4 and Theorem 1.3 are equivalent, completing the optimal $L^{2}$ extension approach to Theorem 1.3? Moreover, for any ideal $I$ of $\mathcal{O}_{o}$, do we have

$$B(F,I,D)=C_{F,I}(D)?$$

We provide an affirmative answer to Question 1.5 in the present paper, which completes the optimal $L^{2}$ extension approach to Theorem 1.3.

Additionally, the proof of Proposition 1.2 (a key step in Theorem 1.4) in [2] relies on the separation theorem from functional analysis, which only guarantees the existence of a desired functional $\xi$ without providing a method to find it. Therefore, it is valuable to find a proof that does not rely on the separation theorem. In this paper, we will present a method for finding the functional in the proof of the main results.

1.3. Main results and applications

Now we show the main results of the present paper, with several applications.

1.3.1. Main results

The following theorem is the main theorem of the present paper, which provides the affirmative answer to Question 1.5.

Theorem 1.6.

Let $D$ be a domain in $\mathbb{C}^{n}$ containing the origin $o$, $I$ a proper ideal of $\mathcal{O}_{o}$, and $(F,o)\in\mathcal{O}_{o}$. Then

$$B^{\circ}(F,I,D)=B(F,I,D)=C_{F,I}(D).$$

This theorem indicates that Theorem 1.4 (the effectiveness result of the strong openness property obtained in [2]) is equivalent to Theorem 1.3 (obtained in [8]).

Corollary 1.7.

Theorem 1.4 is equivalent to Theorem 1.3.

1.3.2. Applications of the main results

For any $\xi\in\ell_{1}^{(n)}$, the functional

$$\begin{split}A^{2}(D)&\longrightarrow\mathbb{C}\\ f&\longmapsto(\xi\cdot f)(o),\end{split}$$

is linear and continuous. By Riesz’s representation theorem, there exists some $T(\xi)\in A^{2}(D)$ such that $\langle f,T(\xi)\rangle=(\xi\cdot f)(o)$ for any $f\in A^{2}(D)$, where $\langle\cdot,\cdot\rangle$ is the inner product on the Hilbert space $A^{2}(D)$. It is known that $T(\ell_{0}^{(n)})$ is dense in $A^{2}(D)$ (see [4, Appendix]).

Theorem 1.6 implies

Corollary 1.8.

For any ideal $I$ of $\mathcal{O}_{o}$, $T\big{(}\tilde{\ell}_{I}\cap\ell_{0}^{(n)}\big{)}$ is dense in $A^{2}(D,I)^{\perp}$.

Let $D$ be a bounded pseudoconvex domain in $\mathbb{C}^{n}$ containing the origin $o$, $\varphi$ a negative plurisubharmonic function on $D$ with $\varphi(o)=-\infty$, and $(F,o)\in\mathcal{O}_{o}$.

For any $\xi\in\ell_{1}^{(n)}$, the following limit

$$\gamma_{\xi}(\varphi):=\lim_{t\to+\infty}\frac{\log K_{\xi,\{\varphi<-t\}\cap D}(o)}{t}\in[0,+\infty]\cup\{-\infty\},$$

is called the $\xi-$cse (a generalization of complex singularity exponent, see [5]) of $\varphi$, where it is proved in [5] that the limit always exists (since $\log K_{\xi,\{\varphi<-t\}\cap D}(o)$ is convex in $t$), and independent of the choice of the bounded pseudoconvex domain $D$ by the fact: if $\xi\neq 0$,

$$\gamma_{\xi}(\varphi)=\inf\big{\{}c\geq 0:\xi\in\ell_{\mathcal{I}(c\varphi)_{o}}\big{\}}.$$ (1.1)

Theorem 1.6 also implies

Corollary 1.9.

If $c_{o}^{F}(\varphi)<+\infty$, then

$$c_{o}^{F}(\varphi)=\min_{\xi\in\ell_{0}^{(n)},\ (\xi\cdot F)(o)\neq 0}\gamma_{\xi}(\varphi).$$

Corollary 1.9 generalizes [5, Corollary 3.2].

2. Preliminaries

In this section, we recall some preliminaries which will be used in the proofs.

2.1. Krull’s lemma

The famous Krull’s lemma for Noetherian local rings is the key ingredient of the proof of Theorem 1.6.

Lemma 2.1 (Krull’s lemma, see [1]).

Let $R$ be a Noetherian local ring with the unique maximal ideal $\mathfrak{m}$. Then for any ideal $I$ of $R$,

$$\bigcap_{k\in\mathbb{N}_{+}}(I+\mathfrak{m}^{k})=I.$$

It is well-known that the ring $\mathcal{O}_{o}$ is a Noetherian local ring, with the unique maximal ideal $\mathfrak{m}=(z_{1},\ldots,z_{n})$.

2.2. Orthonormal basis of the Bergman space

First, we give an order for $\mathbb{N}^{n}$.

Let $\alpha,\beta\in\mathbb{N}^{n}$ be two multi-indices, where $\alpha=(\alpha_{1},\ldots,\alpha_{n})$, $\beta=(\beta_{1},\ldots,\beta_{n})$. Denote $|\alpha|:=\alpha_{1}+\cdots+\alpha_{n}$ (and the same for $|\beta|$). We write $\alpha\prec\beta$, if and only if

(1) $|\alpha|<|\beta|$; or

(2) $|\alpha|=|\beta|$, and there exists $j\in\{1,\ldots,n\}$ such that $\alpha_{n}=\beta_{n}$, $\ldots$, $\alpha_{j+1}=\beta_{j+1}$, and $\alpha_{j}<\beta_{j}$.

Let $D\subseteq\mathbb{C}^{n}$ be a domain with $o\in D$, and $\phi$ a plurisubharmonic function on $D$. Denote

$$A_{\phi}^{2}(D):=\left\{f\in\mathcal{O}(D)\colon\|f\|_{D,\phi}^{2}<+\infty\right\},$$

where $\|f\|_{D,\phi}^{2}:=\int_{D}|f|^{2}e^{-\phi}$.

The following lemma shows that we can have an orthonormal basis of $A_{\phi}^{2}(D)$ satisfying some specific conditions.

For any $\xi\in\ell_{1}^{(n)}$, the functional

$$\begin{split}A_{\phi}^{2}(D)&\longrightarrow\mathbb{C}\\ f&\longmapsto(\xi\cdot f)(o),\end{split}$$

is linear and continuous. Thus, by Riesz representation theorem, there exists some $T_{\phi}(\xi)\in A_{\phi}^{2}(D)$ such that $\langle f,T_{\phi}(\xi)\rangle_{\phi}=(\xi\cdot f)(o)$ for any $f\in A_{\phi}^{2}(D)$, where $\langle\cdot,\cdot\rangle_{\phi}$ is the inner product on the Hilbert space $A_{\phi}^{2}(D)$. $T_{\phi}$ gives an operator from $\ell_{1}^{(n)}$ to $A_{\phi}^{2}(D)$.

Lemma 2.2 (see the Appendix of [4]).

There exist a subset $\mathbf{E}$ of $\mathbb{N}^{n}$, a sequence $\xi[\alpha]=(\xi[\alpha]_{\gamma})_{\gamma}\in\ell_{0}^{(n)}$, and an orthonormal basis $\{\sigma_{\alpha}\}_{\alpha\in\mathbf{E}}$ of $A_{\phi}^{2}(D)$, such that for any $\alpha\in\mathbf{E}$,

(1) $\sigma_{\alpha}=T_{\varphi}(\xi[\alpha])$;

(2) $\max\big{\{}\gamma\in\mathbb{N}^{n}:\xi[\alpha]_{\gamma}\neq 0\big{\}}=\alpha$;

(3) $\min\big{\{}\gamma\in\mathbb{N}^{n}:D^{\gamma}(\sigma_{\alpha})(o)\neq 0\big{\}}=\alpha$,

where the maximum and minimum in (2) and (3) are w.r.t. ‘$\prec$’.

Here in fact, $\mathbf{E}$ is the set containing all $\alpha\in\mathbb{N}^{n}$ such that there exists an $f\in A_{\phi}^{2}(D)$ with $D^{\alpha}f(o)\neq 0$ and $D^{\beta}f(o)=0$ for all $\beta\prec\alpha$. For example, if $D$ is a bounded domain and $\mathcal{I}(\phi)_{o}=\mathcal{O}_{o}$, then $\mathbf{E}=\mathbb{N}^{n}$.

3. Proof of Theorem 1.6

Now we prove Theorem 1.6. For any $\xi\in\ell_{0}^{(n)}$, we denote

$$\mathrm{ord}_{o}(\xi):=\max\big{\{}k\in\mathbb{N}:\exists\alpha\in\mathbb{N}^{n},\ \text{s.t.}\ |\alpha|=k,\ \xi_{\alpha}\neq 0\big{\}},$$

which coincides the order of $\xi$ seen as an element in $\mathbb{C}[z_{1},\ldots,z_{n}]$. We first prove a special case of Theorem 1.6.

Lemma 3.1.

Let $D$ be a bounded domain in $\mathbb{C}^{n}$, $I$ a proper ideal of $\mathcal{O}_{o}$ and $(F,o)\in\mathcal{O}_{o}$. Assume that $\mathfrak{m}^{j_{0}}\subseteq I$ for some $j_{0}\in\mathbb{N}_{+}$, where $\mathfrak{m}$ is the maximal ideal of $\mathcal{O}_{o}$. Then there exists $\eta\in\ell_{I}\cap\ell_{0}^{(n)}$ with $\mathrm{ord}_{o}(\eta)<j_{0}$ such that

$$\frac{|(\eta\cdot F)(o)|^{2}}{K_{\eta,D}(o)}=C_{F,I}(D).$$

Proof.

The case $(F,o)\in I$ is trivial, where we can choose $\eta=(1,0,\ldots,0,\ldots)$ for example. In the following, we assume $(F,o)\notin I$.

First, we can actually assume that $F\in A^{2}(D)$. For the reason, if we take the order$\leq j_{0}$ part of the Taylor expansion of $F$ at $o$, denoted by $\tilde{F}$, then $(F-\tilde{F},o)\in\mathfrak{m}^{j_{0}}\subseteq I$, and $\tilde{F}\in A^{2}(D)$ since it is a polynomial and $D$ is bounded. Now for any $\xi\in\ell_{I}$, it follows that $(\xi\cdot F)(o)=(\xi\cdot\tilde{F})(o)$. Replacing $F$ by $\tilde{F}$, it is enough to assume $F\in A^{2}(D)$.

Since $F\in A^{2}(D)$ with $(F,o)\notin I$, $C_{F,I}(D)\leq\|F\|_{D}^{2}<+\infty$, then there exists a unique $G\in A^{2}(D,I)^{\perp}\setminus\{0\}$ such that $\|G\|^{2}_{D}=C_{F,I}(D)$ and $(G-F,o)\in I$. We can write that

$$G=\sum_{\alpha\in\mathbb{N}^{n}}a_{\alpha}\sigma_{\alpha},\ a_{\alpha}\in\mathbb{C},$$

where $\{\sigma_{\alpha}\}_{\alpha\in\mathbb{N}^{n}}$ is the orthonormal basis of $A^{2}(D)$ in Lemma 2.2. Note that $D^{\beta}(\sigma_{\alpha})(o)=0$ for any $\beta\prec\alpha$, then $\sigma_{\alpha}\in A^{2}(D,\mathfrak{m}^{j_{0}})\subseteq A^{2}(D,I)$ for any $\alpha$ with $|\alpha|\geq j_{0}$. Thus, $\langle G,\sigma_{\alpha}\rangle_{D}=0$ if $|\alpha|\geq j_{0}$, where $\langle\cdot,\cdot\rangle_{D}$ is the inner product on $A^{2}(D)$. It follows that we have

$$G=\sum_{|\alpha|\leq j_{0}-1}a_{\alpha}\sigma_{\alpha},\ b_{\alpha}\in\mathbb{C},$$

which is a finite summation. Set

$$\eta=\sum_{|\alpha|\leq j_{0}-1}\overline{a_{\alpha}}\cdot\xi[\alpha]\in\ell_{0}^{(n)},$$

where $\xi[\alpha]\in\ell^{(n)}_{0}$ satisfying $T(\xi[\alpha])=\sigma_{\alpha}$ such as which in Lemma 2.2. Then $\mathrm{ord}_{o}(\eta)<j_{0}$, and $T(\eta)=G$. In addition, we have

$$K_{\eta,D}(o)=\sup_{f\in A^{2}(D)}\frac{|(\eta\cdot f)(o)|^{2}}{\|f\|_{D}^{2}}=\sup_{f\in A^{2}(D)}\frac{|\langle f,G\rangle_{D}|^{2}}{\|f\|_{D}^{2}}=\|G\|_{D}^{2},$$

yielding that

$$\frac{|(\eta\cdot F)(o)|^{2}}{K_{\eta,D}(o)}=\frac{|\langle F,G\rangle_{D}|^{2}}{\|G\|_{D}^{2}}=\frac{\Big{|}\|G\|_{D}^{2}+\langle F-G,G\rangle_{D}\Big{|}^{2}}{\|G\|_{D}^{2}}=\|G\|_{D}^{2}=C_{F,I}(D).$$

Now, we only need to verify $\eta\in\ell_{I}$. For any $(f,o)\in I$, by Taylor’s expansion, for any $j\geq j_{0}$, we can write that

$$f=f_{j}+\tilde{f}_{j},$$

where $f_{j}\in\mathbb{C}[z_{1},\cdots,z_{n}]$ with $\mathrm{deg\ }f_{j}\leq j$, and $\tilde{f}_{j}\in\mathfrak{m}^{j}\subseteq I$. Since $D$ is bounded, $f_{j}\in A^{2}(D)$. We also have $(f_{j},o)=(f-\tilde{f}_{j},o)\in I$, thus $f_{j}\in A^{2}(D,I)$. Then by that $T(\eta)=G\in A^{2}(D,I)^{\perp}$, it holds that

$$(\eta\cdot f_{j})(o)=\langle f_{j},G\rangle_{D}=0.$$

Since $\tilde{f}_{j}\in\mathfrak{m}^{j}$ and $\mathrm{ord}_{o}(\eta)<j_{0}\leq j$, we have $(\eta\cdot\tilde{f}_{j})(o)=0$. Thus, $(\eta\cdot f)(o)=0$, for all $(f,o)\in I$. The proof of $\eta\in\ell_{I}$ is completed, so the desired result holds. ∎

Using Krull’s lemma, we give an approximation result for the minimal $L^{2}$ integrals.

Lemma 3.2.

Let $D$ be a bounded domain in $\mathbb{C}^{n}$, $I$ a proper ideal of $\mathcal{O}_{o}$, and $(F,o)\in\mathcal{O}_{o}$. Then

$$\lim_{j\to\infty}C_{F,I+\mathfrak{m}^{k}}(D)=C_{F,I}(D).$$ (3.1)

Proof.

Clearly, $C_{F,I+\mathfrak{m}^{k}}(D)$ is an increasing sequence in $k$, and $C_{F,I+\mathfrak{m}^{k}}(D)\leq C_{F,I}(D)$ for any $k$. Without loss of generality, we assume $\lim_{k\to\infty}C_{F,I+\mathfrak{m}^{k}}(D)<+\infty$. Then there exists $G_{k}\in A^{2}(D)$ with

$$(G_{k}-F,o)\in I+\mathfrak{m}^{k},\ \|G_{k}\|_{D}^{2}=C_{F,I+\mathfrak{m}^{k}}(D),\ \forall k\in\mathbb{N}_{+}.$$

Since $\lim_{k\to\infty}C_{F,I+\mathfrak{m}^{k}}(D)<+\infty$, by Montel’s theorem, we can extract a subsequence of $G_{k}$ compactly convergent to some $G\in A^{2}(D)$. Fatou’s lemma shows $\|G\|_{D}^{2}\leq\lim_{k\to\infty}C_{F,I+\mathfrak{m}^{k}}(D)$. We also have $(G-F,o)\in I+\mathfrak{m}^{k}$ for any $k$ (see [7, Chapter 2.2.3]). Krull’s lemma indicates

$$\bigcap_{k\geq 1}(I+\mathfrak{m}^{k})=I.$$

Then $(G-F,o)\in I$, which implies $\|G\|_{D}^{2}\geq C_{F,I}(D)$. This confirms (3.1). ∎

For an unnecessarily bounded domain, we will use the following lemma to approximate the minimal $L^{2}$ integral.

Lemma 3.3.

Let $D$ be a domain in $\mathbb{C}^{n}$ containing $o$, $I$ a proper ideal of $\mathcal{O}_{o}$, and $(F,o)\in\mathcal{O}_{o}$. Let $\{D_{i}\}_{i=1}^{\infty}$ be a sequence of domains such that $D_{i}\subseteq D_{i+1}$, $\bigcup_{i\geq 1}D_{i}=D$, and $o\in D_{1}$. Then

$$\lim_{i\to\infty}C_{F,I}(D_{i})=C_{F,I}(D).$$

Proof.

Note that $C_{F,I}(D_{i})$ is increasing in $i$ and $C_{F,I}(D_{i})\leq C_{F,I}(D)$ for any $i$, then without loss of generality, we assume $\lim_{i\to\infty}C_{F,I}(D_{i})<+\infty$. For any $i$, let $h_{i}\in A^{2}(D_{i})$ with $(h_{i}-F,o)\in I$ and $\|h_{i}\|_{D_{i}}^{2}=C_{F,I}(D_{i})$. Then Montel’s theorem implies that there exists a subsequence of $\{h_{i}\}$ which converges to some $h\in\mathcal{O}(D)$ on any compact subset of $D$. According to Fatou’s lemma, we have $\|h\|_{D}^{2}\leq\lim_{i\to\infty}C_{F,I}(D_{i})$. We also have $(h-F,o)\in I$, which implies $\|h\|_{D}^{2}\geq C_{F,I}(D)$. Then the proof is completed. ∎

Now we give the proof of Theorem 1.6.

Proof of Theorem 1.6.

We only need to prove $B^{\circ}(F,I,D)\geq C_{F,I}(D)$, where $(F,o)\notin I$.

First, we assume $D$ is bounded. For the ideal $I$ of $\mathcal{O}_{o}$, since $\mathfrak{m}^{k}\subset I+\mathfrak{m}^{k}$, Lemma 3.1 implies

$$B^{\circ}(F,I+\mathfrak{m}^{k},D)=C_{F,I+\mathfrak{m}^{k}}(D),\ \forall k\in\mathbb{N}_{+},$$

yielding that

$$B^{\circ}(F,I,D)\geq B^{\circ}(F,I+\mathfrak{m}^{k},D)=C_{F,I+\mathfrak{m}^{k}}(D),\ \forall k\in\mathbb{N}_{+}.$$ (3.2)

Combining with Lemma 3.2, we get

$$B^{\circ}(F,I,D)\geq C_{F,I}(D).$$

If $D$ is an unbounded domain, we choose an increasing sequence of bounded domains $\{D_{i}\}_{i=1}^{\infty}$ such that $\bigcup_{i\geq 1}D_{i}=D$, and $o\in D_{1}$, then Lemma 3.3 implies

$$B^{\circ}(F,I,D)\geq\limsup_{i\to\infty}B^{\circ}(F,I,D_{i})=\lim_{i\to\infty}C_{F,I}(D_{i})=C_{F,I}(D).$$

At last, the case $(F,o)\in I$ is trivial.

The proof of Theorem 1.6 is completed. ∎

We give some remarks here.

Remark 3.4.

We can reprove Proposition 1.2 by Theorem 1.6 without using the separation theorem in functional analysis theory:

for any proper ideal $I$ of $\mathcal{O}_{o}$, and any $(F,o)\in\mathcal{O}_{o}$, there exists some $\xi\in\ell_{I}\cap\ell^{(n)}_{0}$, such that $(\xi\cdot F)(o)\neq 0$.

This is due to $B^{\circ}(F,I,D)=C_{F,I}(D)\in(0,+\infty)$ for a domain $D$ containing $o$ and $F\in A^{2}(D)$.

Remark 3.5.

From the proof of Theorem 1.6, we can see

$$C_{F,I}(D)=\sup_{k\geq 1}\sup_{\xi\in\ell_{I+\mathfrak{m}^{k}}}\frac{|(\xi\cdot F)(o)|^{2}}{K_{\xi,D}(o)}=\sup_{k\geq 1}\min_{\xi\in\ell_{I+\mathfrak{m}^{k}}}\frac{|(\xi\cdot F)(o)|^{2}}{K_{\xi,D}(o)},$$

where $\ell_{I+\mathfrak{m}^{k}}$ can be seen as a finite-dimensional linear subspace of $\mathbb{C}[z_{1},\ldots,z_{n}]$.

4. Proofs of Corollary 1.8 and Corollary 1.9

We prove Corollary 1.8 and Corollary 1.9 in this section.

Proof of Corollary 1.8.

It is clear that $T(\ell_{I})\subseteq A^{2}(D,I)^{\perp}$.

Now fix any $F\in A^{2}(D,I)^{\perp}\setminus\{0\}$. Then by the definition, $C_{F,I}(D)=\|F\|_{D}^{2}$. According to Theorem 1.6, we can find a sequence $\{\xi_{j}\}_{j\in\mathbb{N}_{+}}\subset\ell_{I}\cap\ell_{0}^{(n)}$ such that

$$\lim_{j\to+\infty}\frac{|(\xi_{j}\cdot F)(o)|^{2}}{K_{\xi_{j},D}(o)}=C_{F,I}(D)=\|F\|_{D}^{2}.$$

Denote $T(\xi_{j})=g_{j}\in A^{2}(D,I)^{\perp}$. It follows that

$$K_{\xi_{j},D}(o)=\sup_{f\in A^{2}(D)}\frac{|\langle f,g_{j}\rangle_{D}|^{2}}{\|f\|_{D}^{2}}=\|g_{j}\|_{D}^{2},\ \forall j\in\mathbb{N}_{+},$$

which verifies that

$$\|F\|_{D}^{2}=\lim_{j\to+\infty}\frac{|\langle F,g_{j}\rangle_{D}|^{2}}{K_{\xi_{j},D}(o)}=\lim_{j\to+\infty}\frac{|\langle F,g_{j}\rangle_{D}|^{2}}{\|g_{j}\|_{D}^{2}}.$$

Set

$$\eta_{j}:=e^{\sqrt{-1}\theta_{j}}\frac{\|F\|_{D}}{\|g_{j}\|_{D}}\cdot\xi_{j}\in\ell_{I}\cap\ell_{0}^{(n)},\ \forall j\in\mathbb{N}_{+},$$

and $G_{j}=T(\eta_{j})\in A^{2}(D,I)^{\perp}$, where $\theta_{j}\in\mathbb{R}$ such that $\langle F,G_{j}\rangle_{D}\in\mathbb{R}_{\geq 0}$. Then $\|G_{j}\|_{D}^{2}=\|F\|_{D}^{2}$, and

$$\lim_{j\to+\infty}\langle F,G_{j}\rangle_{D}=\lim_{j\to+\infty}\mathrm{Re}\langle F,G_{j}\rangle_{D}=\|F\|_{D}^{2}.$$

Now we deduce that

$$\begin{split}\|F-G_{j}\|_{D}^{2}=\|F\|_{D}^{2}+\|G_{j}\|_{D}^{2}-2\mathrm{Re}\langle F,G_{j}\rangle_{D}\to 0,\ j\to+\infty,\end{split}$$

which implies $T(\eta_{j})=G_{j}\to F$ as $j\to+\infty$ in the topology of $A^{2}(D)$.

The proof is completed. ∎

Proof of Corollary 1.9.

It is proved in [5] that $c_{o}^{F}(\varphi)\leq\gamma_{\xi}(\varphi)$ for any $\xi\in\ell_{0}^{(n)}$ with $(\xi\cdot F)(o)\neq 0$ (one can deduce this from the equality (1.1)).

In addition, for any $\xi\in\ell_{\mathcal{I}_{+}(c_{o}^{F}(\varphi)\varphi)_{o}}$ with $(\xi\cdot F)(o)\neq 0$, by (1.1), it holds that $\gamma_{\xi}(\varphi)=c_{o}^{F}(\varphi)$. Since $(F,o)\notin\mathcal{I}_{+}(c_{o}^{F}(\varphi)\varphi)_{o}$, according to Remark 3.4, there exists some $\eta\in\ell_{0}^{(n)}\cap\ell_{\mathcal{I}_{+}(c_{o}^{F}(\varphi)\varphi)_{o}}$ satisfying $(\eta\cdot F)(o)\neq 0$, which implies $\gamma_{\eta}(\varphi)=c_{o}^{F}(\varphi)$. Thus, we get

$$c_{o}^{F}(\varphi)=\min_{\xi\in\ell_{0}^{(n)},\ (\xi\cdot F)(o)\neq 0}\gamma_{\xi}(\varphi).$$

∎

5. A weighted version

A weighted version of generalized Bergman kernels was introduced in [3], and used to obtain a (sharp) effectiveness result for the $L^{p}$ strong openness property, which reproved some results in [9].

For the purpose of comparing the results in [3] and [9], we demonstrate a weighted version of Theorem 1.6 in this section. Let $D$ be a domain in $\mathbb{C}^{n}$ containing $o$, and $\psi$ a plurisubharmonic function on $D$. Denote by

$$A_{\psi}^{2}(D):=\Big{\{}f\in\mathcal{O}(D):\int_{D}|f|^{2}e^{-\psi}<+\infty\Big{\}}$$

the weighted Bergman space and $\|f\|_{D,\psi}:=\left(\int_{D}|f|^{2}e^{-\psi}\right)^{1/2}$ the norm on the space.

For any $\xi\in\ell_{1}^{(n)}$, the weighted Bergman kernel with respect to $\xi$ on $D$ is denoted by (see [3]):

$$K^{\psi}_{\xi,D}(z):=\sup\big{\{}|(\xi\cdot f)(z)|^{2}\colon f\in A^{2}_{\psi}(D),\ \|f\|_{D,\psi}^{2}\leq 1\big{\}},\ \forall z\in D.$$

For any proper ideal $I$ of $\mathcal{O}_{o}$ satisfying $I\not\supseteq\mathcal{I}(\psi)_{o}$ and $(F,o)\in\mathcal{O}_{o}$, set

$$B_{\psi}(F,I,D):=\sup_{\xi\in\ell_{I}}\frac{|(\xi\cdot F)(o)|^{2}}{K^{\psi}_{\xi,D}(o)},$$

and

$$B^{\circ}_{\psi}(F,I,D):=\sup_{\xi\in\ell_{I}\cap\ell_{0}^{(n)}}\frac{|(\xi\cdot F)(o)|^{2}}{K^{\psi}_{\xi,D}(o)}.$$

In addition, recall the minimal $L^{2}$ integral:

$$C_{F,\psi,I}(D):=\inf\big{\{}\|\tilde{F}\|^{2}_{D,\psi}\colon\tilde{F}\in A^{2}_{\psi}(D),\ (\tilde{F}-F,o)\in I\big{\}}.$$

One can establish the following weighted version of Theorem 1.6, where the detail of the proof will be omitted here.

Theorem 5.1.

The following equality holds:

$$B^{\circ}_{\psi}(F,I,D)=B_{\psi}(F,I,D)=C_{F,\psi,I}(D).$$

Sketch of the proof.

Let $\psi_{j}:=\max\{\psi,-j\}$ for each $j\in\mathbb{N}_{+}$. Since each $\psi_{j}$ does not have singularity, we can repeat the process of the proof of Theorem 1.6, and obtain

$$B^{\circ}_{\psi_{j}}(F,I,D)=B_{\psi_{j}}(F,I,D)=C_{F,\psi_{j},I}(D).$$

Then by approximation, the same equality for $\psi$ can be verified. ∎

Acknowledgements. We thank Dr. Zheng Yuan and Dr. Zhitong Mi for checking this paper. The second named author was supported by National Key R&D Program of China 2021YFA1003100, NSFC-11825101 and NSFC-12425101.