Hostname: page-component-586b7cd67f-r5fsc Total loading time: 0 Render date: 2024-11-23T19:18:50.112Z Has data issue: false hasContentIssue false

Schatten class composition operators on the Hardy space

Published online by Cambridge University Press:  24 July 2023

Wenwan Yang
Affiliation:
School of Mathematics and Statistics, Guangdong University of Technology, Guangzhou, Guangdong 510520, China ([email protected], [email protected])
Cheng Yuan
Affiliation:
School of Mathematics and Statistics, Guangdong University of Technology, Guangzhou, Guangdong 510520, China ([email protected], [email protected])
Rights & Permissions [Opens in a new window]

Abstract

Suppose $2< p<\infty$ and $\varphi$ is a holomorphic self-map of the open unit disk $\mathbb {D}$. We show the following assertions:

  1. (1) If $\varphi$ has bounded valence and0.1

    \begin{equation} \int_{\mathbb{D}} \left(\frac{1-|z|^2}{1-|\varphi(z)|^2}\right)^{p/2}\frac{\mathrm{d} A(z)}{(1-|z|^2)^2}<\infty, \end{equation}
    then $C_{\varphi }$ is in the Schatten $p$-class of the Hardy space $H^2$.

  2. (2) There exists a holomorphic self-map $\varphi$ (which is, of course, not of bounded valence) such that the inequality (0.1) holds and $C_{\varphi }: H^2\to H^2$ does not belong to the Schatten $p$-class.

Type
Research Article
Copyright
Copyright © The Author(s), 2023. Published by Cambridge University Press on behalf of The Royal Society of Edinburgh

1. Introduction and main results

1.1 Backgrounds and motivations

Let $\mathbb {D}=\{z\in \mathbb {C} :|z|<1\}$ be the unit disk of the complex plane $\mathbb {C}$. Let $H(\mathbb {D})$ be the space of holomorphic functions on $\mathbb {D}$ and let $\varphi$ be a holomorphic function on $\mathbb {D}$ with $\varphi (\mathbb {D})\subset \mathbb {D}$. For $f\in H(\mathbb {D})$, the composition operator $C_{\varphi }$ is a linear operator defined by $C_{\varphi }(f)=f\circ \varphi$.

Recall that a positive $T$ on a separable Hilbert space $H$ is in the trace class if

\[ \mathrm{tr}(T)=\sum_{n=0}^\infty\langle Te_n,e_n\rangle_H<{+}\infty \]

for some (or all) orthonormal basis $\{e_n\}$ of $H$. For any $0< p<\infty$, the Schatten $p$-class $\mathcal {S}_p(H)$ of $H$ consists of bounded linear operators $T:H\to H$ such that $(T^\ast T)^{p/2}$ belongs to the trace class. In particular, $\mathcal {S}_1(H)$ is the trace class of $H$, and $\mathcal {S}_2(H)$ is called the Hilbert–Schmidt class. It is easy to check that $T\in \mathcal {S}_p(H)$ if and only if $T^*\in \mathcal {S}_p(H)$. For more details about Schatten $p$-class operators, we refer the readers to Zhu [Reference Zhu16].

The Hardy space $H^2$ is a Hilbert space of analytic functions $f$ on $\mathbb {D}$ such that

\[ \|f\|_{H^2}^2=\sup_{0< r<1}\int_0^{2\pi}|f(r e^{i\theta})|^2\frac{\mathrm{d}\theta}{2\pi} <\infty. \]

For $\alpha >-1$, the weighted Bergman space $A_\alpha ^2$ consists of holomorphic functions $f$ on $\mathbb {D}$ satisfying

\[ \|f\|_{A_\alpha^2}^2= \int_{\mathbb{D}}|f(z)|^2 {\mathrm{d} A_\alpha(z)} <\infty, \]

where $\mathrm {d} A_\alpha (z)=(\alpha +1)(1-|z|^2)^\alpha \mathrm {d} A(z)$ and $\mathrm {d} A(z)$ is the normalized area measure on $\mathbb {D}$. When $\alpha =0$, the space $A_0^2$ is usually denoted by $A^2$. Properties of composition operator on $A_\alpha ^2$ and $H^2$ has been widely investigated for decades, see e.g. [Reference Cowen and MacCluer3, Reference Shapiro8, Reference Zhu16]. In particular, conditions for $C_{\varphi }$ that belong to $\mathcal {S}_p(A_\alpha ^2)$ and $\mathcal {S}_p(H^2)$ are also characterized, see [Reference Benazzouz, El-Fallah, Kellay and Mahzouli1, Reference Bendaoud, Korrichi, Merghni and Yagoub2, Reference Luecking4Reference Pau and Pérez7, Reference Wirths and Xiao9, Reference Xia10, Reference Yuan and Zhou12, Reference Zhu14].

It is well known (see e.g. Zhu [Reference Zhu15]) that $H^2$ can be viewed as the limit case of $A^2_{\alpha }$ as $\alpha \to -1^+$ in some sense. It is also known that for $0< p<\infty$, $C_{\varphi }\in \mathcal {S}_p(H^2)$ if and only if

\[ \int_{\mathbb{D}}\left(\frac{N_\varphi(z)}{\log\frac1{|z|}}\right)^{p/2}\mathrm{d} \lambda(z)<\infty, \]

where

\[ \mathrm{d}\lambda(z)=(1-|z|^2)^{{-}2}\mathrm{d} A(z) \]

is the Möbius invariant measure on $\mathbb {D}$, and

\[ N_\varphi(z)=\sum_{w\in \varphi^{{-}1}(z)}\log\frac1{|w|} \]

is the Nevanlinna counting function of $\varphi$. Similarly, $C_{\varphi }\in \mathcal {S}_p(A_\alpha ^2)$ if and only if

\[ \int_{\mathbb{D}}\left(\frac{N_{\varphi,\alpha+2}(z)}{(\log \frac1{|z|})^{\alpha+2}}\right)^{p/2}\mathrm{d} \lambda(z)<\infty, \]

where $N_{\varphi,\alpha +2}(z)$ is a generalized Nevanlinna counting function of $\varphi$ given by

\[ N_{\varphi,\alpha+2}(z)=\sum_{w\in \varphi^{{-}1}(z)}\left(\log\frac1{|w|}\right)^{\alpha+2}. \]

See Luecking-Zhu [Reference Luecking and Zhu5].

1.2 Main results

A holomorphic map $\varphi :\mathbb {D}\to \mathbb {D}$ is of bounded valence if there is a positive integer $N$ such that for each $z\in \mathbb {D}$, the set $\varphi ^{-1}(z)$ contains at most $N$ points. Zhu [Reference Zhu14] shows that if $\alpha >-1$, $2\le p<\infty$ and $\varphi :\mathbb {D}\to \mathbb {D}$ is an analytic function of bounded valence, then $C_{\varphi }$ is in the Schatten class $\mathcal {S}_p$ of $A_\alpha ^2$ if and only if

\[ \int_{\mathbb{D}} \left(\frac{1-|z|^2}{1-|\varphi(z)|^2}\right)^{p(\alpha+2)/2} {\mathrm{d} \lambda(z)} <\infty. \]

Meanwhile, Zhu [Reference Zhu16, Exercise 11.6.7] says that if $p>2$ and $C_{\varphi }\in \mathcal {S}_p(H^2)$, then

\[ \int_{\mathbb{D}} \left(\frac{1-|z|^2}{1-|\varphi(z)|^2}\right)^{p/2} {\mathrm{d} \lambda(z)} <\infty. \]

These observations hint us to give the following result.

Theorem 1.1 If $2< p<\infty,$ $\varphi$ has bounded valence and

(1.1)\begin{equation} \int_{\mathbb{D}} \left(\frac{1-|z|^2}{1-|\varphi(z)|^2}\right)^{p/2} \mathrm{d}\lambda(z)<\infty, \end{equation}

then $C_{\varphi }\in \mathcal {S}_p(H^2)$.

For $p>2$, Xia [Reference Xia10] constructs a holomorphic map $\varphi :\mathbb {D}\to \mathbb {D}$ such that

\[ \int_{\mathbb{D}} \left(\frac{1-|z|^2}{1-|\varphi(z)|^2}\right)^p \mathrm{d}\lambda(z)<\infty \]

and such that $C_{\varphi }: A^2\to A^2$ does not belong to the Schatten class $\mathcal {S}_p(A^2)$. Motivated by Xia [Reference Xia10], we prove the following theorem:

Theorem 1.2 For any $2< p<\infty,$ there exists a holomorphic function $\varphi :\mathbb {D}\to \mathbb {D}$ such that

(1.2)\begin{equation} \int_{\mathbb{D}} \left(\frac{1-|z|^2}{1-|\varphi(z)|^2}\right)^{p/2} \mathrm{d}\lambda(z)<\infty, \end{equation}

but $C_{\varphi }:H^2\to H^2$ does not belong to the Schatten class $\mathcal {S}_p(H^2)$.

The proof of theorem 1.1 is based on Wirths-Xiao [Reference Wirths and Xiao9] and Zhu [Reference Zhu14]. The proof of theorem 1.2 is modified from Xia [Reference Xia10]. Although the idea of the proof of theorem 1.2 is coming from [Reference Xia10], there are several technical barriers we need to overcome. Thus, we need to adapt Xia's construction for our situation.

Notation. Throughout this paper, we only write $U\lesssim V$ (or $V\gtrsim U$) for $U\le c V$ for a positive constant $c$, and moreover $U\approx V$ for both $U\lesssim V$ and $V\lesssim U$.

2. Preliminaries

For $\alpha >-1$, the Dirichlet-type space is a space of holomorphic functions $f$ on $\mathbb {D}$ for which

\[ \|f\|_{ \alpha}^2=|f(0)|^2+\|f'\|^2_{ A_\alpha^2}<\infty. \]

It is easy to check that $A_\alpha ^2=\mathcal {D}_{\alpha +2}$ and $H^2=\mathcal {D}_1$ with equivalent norms.

The following lemma is contained in [Reference Wirths and Xiao9, Theorem 3.2].

Lemma 2.1 Let $\alpha >-1$ and $0< p<\infty$. Suppose $\varphi :\mathbb {D}\to \mathbb {D}$ is holomorphic. Then $C_{\varphi }\in \mathcal {S}_p(\mathcal {D}_\alpha )$ if and only if

(2.1)\begin{equation} \int_{\mathbb{D}}\left(\int_{\mathbb{D}}\left(\frac{(1-|w|^2)^\varepsilon}{|1- \bar w \varphi(z)|^{1+\varepsilon}}\right)^{2+\alpha}|\varphi'(z)|^2 (1-|z|^2)^\alpha\mathrm{d} A(z)\right)^{p/2} \mathrm{d}\lambda(w)<\infty \end{equation}

for some (any) $\varepsilon >\max \{1/(2+\alpha ),\, 2/(2p+p\alpha )\}$.

For fixed $\alpha >0$, $f,\,g\in \mathcal {D}_\alpha$ with

\[ f(z)=\sum_{n=0}^\infty a_nz^n\quad\hbox{ and }\quad g(z)=\sum_{n=0}^\infty b^n z^n, \]

let

\[ \langle f,g\rangle_{\mathcal{D}_\alpha}= \sum_{n=0}^\infty \frac{ n!\Gamma(\alpha)}{\Gamma(n+\alpha)} a_n\overline{b_n}. \]

Then the reproducing kernel of $\mathcal {D}_\alpha$ associated with the inner product $\langle \cdot,\,\cdot \rangle _{\mathcal {D}_\alpha }$ is given by

\[ K_{\alpha,w}(z)=K_{\alpha}(z,w)=\frac1{(1- \bar w z)^\alpha},\quad z,w\in\mathbb{D}. \]

This means that for each $f\in \mathcal {D}_\alpha$,

\[ f(w)=\langle f,K_{\alpha,w}\rangle_{\mathcal{D}_\alpha} \quad w\in\mathbb{D}. \]

Meanwhile, if we write

\[ J_{\alpha,w}(z)= J_{\alpha}(z,w)=\frac{\partial }{\partial\bar w}K_\alpha(z,w)=\frac{\alpha z}{(1-\bar w z)^{\alpha+1}}, \]

then

(2.2)\begin{equation} f'(w)=\langle f,J_{\alpha,w}\rangle_{\mathcal{D}_\alpha}. \end{equation}

Let

\[ \|f\|_{\mathcal{D}_\alpha}^2=\langle f,f\rangle_{\mathcal{D}_\alpha}. \]

Then

\[ \|K_{\alpha,w}\|_{\mathcal{D}_\alpha}^2=\frac1{(1-|w|^2)^\alpha} \]

and

(2.3)\begin{equation} \|J_{\alpha,w}\|_{\mathcal{D}_\alpha}^2=\langle J_{\alpha,w},J_{\alpha,w}\rangle_{\mathcal{D}_\alpha}=J_{\alpha,w}'(w) =\frac{\alpha(1+\alpha|w|^2)}{(1-|w|^2)^{\alpha+2}}\approx \frac1 {(1-|w|^2)^{\alpha+2}}. \end{equation}

Let

\[ k_{\alpha,w}(z)=\frac{K_{\alpha,w}(z)}{\left\|K_{\alpha,w} \right\|_{\mathcal{D}_\alpha}} \quad \hbox{ and }\quad j_{\alpha,w}(z)=\frac{J_{\alpha,w}(z)}{\left\|J_{\alpha,w} \right\|_{\mathcal{D}_\alpha}} . \]

The following lemma comes from [Reference Yang and Liu11, Lemma 10].

Lemma 2.2 Suppose $\alpha >0$ and $T: \mathcal {D}_\alpha \to \mathcal {D}_\alpha$ is a positive operator. Let

\[ \widehat{ T}^{ \alpha,t}(w)=\langle T j_{\alpha,w} , j_{\alpha,w} \rangle_{\mathcal{D}_\alpha},\quad w\in\mathbb{D}. \]
  1. (1) Let $0< p\le 1$. If $\,\widehat { T}^{ \alpha,t}\in L^p (\mathbb {D},\,\mathrm {d} \lambda ),$ then $T$ is in $\mathcal {S}_p(\mathcal {D}_\alpha )$.

  2. (2) Let $1\le p<\infty$. If $\, T$ is in $\mathcal {S}_p(\mathcal {D}_\alpha ),$ then $\widehat { T}^{ \alpha,t}\in L^p (\mathbb {D},\,\mathrm {d} \lambda )$.

Immediately, we have the following theorem.

Theorem 2.3 Suppose $\alpha >0$ and $\varphi :\mathbb {D}\to \mathbb {D}$ is a holomorphic function.

  1. (1) If $0< p\le 2$ and

    (2.4)\begin{equation} \int_{\mathbb{D}} \left(\frac{(1-|z|^2)^{\alpha+2}|\varphi '(z)|^2}{(1-|\varphi(z)|^2)^{\alpha+2}} \right)^{p/2}\mathrm{d} \lambda(z)<\infty, \end{equation}
    then $C_{\varphi }$ is in $\mathcal {S}_p$ of $\,\mathcal {D}_\alpha$.
  2. (2) If $2\le p<\infty$ and $C_{\varphi }$ is in $\mathcal {S}_p$ of $\,\mathcal {D}_\alpha,$ then (2.4) holds.

Proof. Write $S=C_{\varphi }C_{\varphi }^\ast$, then $S:\mathcal {D}_\alpha \to \mathcal {D}_\alpha$ is a positive operator. We have

\begin{align*} \widehat{ S}^{ \alpha,t}(w)=\langle S j_{\alpha,w} , j_{\alpha,w} \rangle_{\mathcal{D}_\alpha}& = \langle C_{\varphi}^\ast j_{\alpha,w} ,C_{\varphi}^\ast j_{\alpha,w} \rangle_{\mathcal{D}_\alpha}\\& = \frac{\langle C_{\varphi}^\ast J_{\alpha,w} ,C_{\varphi}^\ast J_{\alpha,w} \rangle_{\mathcal{D}_\alpha}}{\|J_{\alpha,w}\|^2 _{\mathcal{D}_\alpha}} = \frac{\| C_{\varphi}^\ast J_{\alpha,w} \|^2_{\mathcal{D}_\alpha}}{\|J_{\alpha,w}\|^2 _{\mathcal{D}_\alpha}}. \end{align*}

For each $f\in \mathcal {D}_\alpha$, (2.2) implies that

\begin{align*} \langle f,C_{\varphi}^\ast J_{\alpha,w}\rangle_{\mathcal{D}_\alpha}& = \langle C_{\varphi} f,J_{\alpha,w} \rangle_{\mathcal{D}_\alpha}=f'(\varphi(w))\varphi'(w)\\ & = \varphi'(w)\langle f,J_{\alpha,\varphi(w)} \rangle_{\mathcal{D}_\alpha}= \langle f,\overline{\varphi'(w)}J_{\alpha,\varphi(w)} \rangle_{\mathcal{D}_\alpha}. \end{align*}

Thus,

\[ C_{\varphi}^\ast J_{\alpha,w} =\overline{\varphi'(w)}J_{\alpha,\varphi(w)} . \]

Then (2.3) implies that

\[ \| C_{\varphi}^\ast J_{\alpha,w}\|_{\mathcal{D}_\alpha}^2\approx \frac{|\varphi'(w)|^2}{(1-|\varphi(w)|^2)^{2+\alpha}} . \]

This gives that

\[ \langle C_{\varphi} C_{\varphi}^\ast j_{\alpha,w} , j_{\alpha,w} \rangle_{\mathcal{D}_\alpha} = \frac{\langle C_{\varphi}^\ast J_{\alpha,w} ,C_{\varphi}^\ast J_{\alpha,w} \rangle_{\mathcal{D}_\alpha}}{\|J_{\alpha,w}\|^2 _{\mathcal{D}_\alpha}} \approx \frac{(1-|w|^2)^{2+\alpha}|\varphi'(w)|^2}{(1-|\varphi(w)|^2)^{2+\alpha}} . \]

An application of lemma 2.2 gives the desired assertions.

By letting $p=2$ in theorem 2.3, we have the following corollary.

Corollary 2.4 Suppose $\alpha >0$ and $\varphi :\mathbb {D}\to \mathbb {D}$ is a holomorphic function. Then $C_{\varphi }$ is in the Hilbert–Schmidt class of $\mathcal {D}_\alpha$ if and only if

\[ \int_{\mathbb{D}} \frac{(1-|z|^2)^{\alpha }|\varphi '(z)|^2}{(1-|\varphi(z)|^2)^{\alpha+2}} \, \mathrm{d} A(z)<\infty. \]

There are several well-known characterizations of the Hilbert–Schmidt compositions on $H^2$ and $A_\alpha ^2$, see e.g. [Reference Cowen and MacCluer3, Reference Yuan and Zhou13, Reference Zhu16]. Combine these characterizations with corollary 2.4, we have the following corollaries.

Corollary 2.5 Suppose $\varphi :\mathbb {D}\to \mathbb {D}$ is holomorphic. Then the following statements are equivalent:

  1. (1) $C_{\varphi }\in \mathcal {S}_2(H^2)$.

  2. (2) The following inequality holds:

    \[ \int_{\mathbb{D}} \frac{(1-|z|^2) |\varphi '(z)|^2}{(1-|\varphi(z)|^2)^{3}} \, \mathrm{d} A(z)<\infty. \]
  3. (3) The following inequality holds:

    \[ \int_{\mathbb{D}} \frac{N_\varphi(z)}{\log\frac1{|z|}} \, \mathrm{d} \lambda(z)<\infty. \]
  4. (4) The following inequality holds:

    \[ \int_{0}^{2\pi} \frac{\mathrm{d}\theta }{(1-|\varphi(\mathrm{e}^{\mathrm{i}\theta})|^2) } <\infty. \]

Corollary 2.6 Suppose $\alpha >-1$ and $\varphi :\mathbb {D}\to \mathbb {D}$ is holomorphic. Then the following statements are equivalent:

  1. (1) $C_{\varphi }\in \mathcal {S}_2(A^2_\alpha )$.

  2. (2) The following inequality holds:

    \[ \int_{\mathbb{D}} \frac{(1-|z|^2)^{\alpha+2} |\varphi '(z)|^2}{(1-|\varphi(z)|^2)^{\alpha+4}} \, \mathrm{d} A(z)<\infty. \]
  3. (3) The following inequality holds:

    \[ \int_{\mathbb{D}} \frac{N_{\varphi,\alpha+2}(z)}{(\log\frac1{|z|})^{\alpha+2}} \, \mathrm{d} \lambda(z)<\infty. \]
  4. (4) The following inequality holds:

    \[ \int_{\mathbb{D}} \frac{(1-|z|^2)^{\alpha }}{(1-|\varphi(z)|^2)^{2+\alpha} }\,\mathrm{d} A(z) <\infty. \]

3. Proof of theorem 1.1

Theorem 1.1 is just the case $\alpha =1$ of the following proposition.

Proposition 3.1 Suppose $\alpha >0,$ $2\le p<\infty$ and $p\alpha >2$. Let $\varphi :\mathbb {D}\to \mathbb {D}$ is a holomorphic function which has bounded valence and

(3.1)\begin{equation} \int_{\mathbb{D}} \left(\frac{1-|z|^2}{1-|\varphi(z)|^2}\right)^{p\alpha/2} \mathrm{d}\lambda(z)<\infty, \end{equation}

then $C_{\varphi }$ is in the Schatten class $\mathcal {S}_p$ of $\,\mathcal {D}_\alpha$.

The condition $p\alpha > 2$ in the above proposition is necessary. Indeed, if $0 < p\alpha \le 2$, then the involved integral is trivially divergent.

Proof. When $p=2$, the condition $p\alpha >2$ implies that $\alpha >1$. Notice that in this case $\mathcal {D}_\alpha =A_{\alpha -2}^2$. According to [Reference Zhu14], the condition (3.1) implies that $C_{\varphi }\in \mathcal {S}_p(A_{\alpha -2}^2)$.

Now we suppose $2< p<\infty$. According to lemma 2.1, if we can check the inequality (2.1) for some $\varepsilon >\max \{1/(2+\alpha ),\, 2/(2p+p\alpha )\}$, then we have $C_{\varphi }\in \mathcal {S}_p(\mathcal {D}_\alpha )$. Write $q=p/2$, then $q>1$. Let

\[ F(w)=\int_{\mathbb{D}} \frac{(1-|w|^2)^{(2+\alpha)\varepsilon} }{|1- \bar w \varphi(z)|^{({2+\alpha})(1+\varepsilon)}} |\varphi'(z)|^2 (1-|z|^2)^\alpha\mathrm{d} A(z). \]

Then it is sufficient to check that $F\in L^q(\mathbb {D},\,\mathrm {d} \lambda )$.

Let

\[ H(w,z)=\frac{(1-|w|^2)^{(\alpha+2)\varepsilon}(1-|\varphi(z)|^2)^{\alpha }(1-|z|^2)^2 |\varphi'(z)|^2}{|1-\bar w\varphi(z)|^{(2+ \alpha)(1+\varepsilon)}} \]

and

\[ h( z)=\left(\frac{(1-|z|^2) }{(1-|\varphi(z)|^{2})}\right)^\alpha. \]

Then,

\[ F(w)=\int_{\mathbb{D}} H(w,z) h(z)\mathrm{d} \lambda(z). \]

Recall that $\varphi :\mathbb {D}\to \mathbb {D}$ is holomorphic. Schwarz's lemma implies that

(3.2)\begin{equation} \frac{ (1-|z|^2)^2 |\varphi'(z)|^2}{(1-|\varphi(z)|^2)^{2 }}\le1. \end{equation}

Then, for each $\varepsilon >1/(2+\alpha )$, Forelli–Rudin's estimate implies that

(3.3)\begin{align} \int_{\mathbb{D}} H(w,z)\mathrm{d} \lambda(w) & =(1-|\varphi(z)|^2)^{\alpha }(1-|z|^2)^2 |\varphi'(z)|^2\int_{\mathbb{D}} \frac{(1-|w|^2)^{(\alpha+2)\varepsilon-2}\mathrm{d} A(w)}{|1-\bar w\varphi(z)|^{(2+ \alpha)(1+\varepsilon)}}\nonumber\\ & \lesssim \frac{(1-|\varphi(z)|^2)^{\alpha }(1-|z|^2)^2 |\varphi'(z)|^2}{(1-|\varphi(z)|^2)^{2+\alpha }}\nonumber\\ & \le1. \end{align}

Meanwhile, recall that $\varphi$ is of bounded valence. Let $n_\varphi (z)$ be the number of points in $\varphi ^{-1}(z)$. Then,

\[ \sup_{z\in\mathbb{D}} n_\varphi(z)<\infty \]

and

(3.4)\begin{align} \int_{\mathbb{D}} H(w,z)\mathrm{d} \lambda(z)& =\int_{\mathbb{D}} \frac{(1-|w|^2)^{(\alpha+2)\varepsilon}(1-|\varphi(z)|^2)^{\alpha } |\varphi'(z)|^2}{|1-\bar w\varphi(z)|^{(2+ \alpha)(1+\varepsilon)}} \,\mathrm{d} A(z) \nonumber\\ & =(1-|w|^2)^{(\alpha+2)\varepsilon}\int_{\mathbb{D}} \frac{n_\varphi(z) (1-| z|^2)^{\alpha } }{|1-\bar w z|^{(2+ \alpha)(1+\varepsilon)}} \,\mathrm{d} A(z)\nonumber\\ & \lesssim 1. \end{align}

Put (3.3) and (3.4) together. Application of Schur's test tells us that the integral operator with kernel $H(w,\,z)$ is bounded on $L^q(\mathbb {D},\,\mathrm {d}\lambda )$. Recall that condition (3.1) implies that $h\in L^q(\mathbb {D},\,\mathrm {d} \lambda )$. This gives that $F\in L^q(\mathbb {D},\,\mathrm {d} \lambda )$ as desired.

4. Proof of theorem 1.2

4.1 Construction of $\varphi$

The construction is modified from Xia [Reference Xia10]. We adapt some parameters for our argument. For $n=1,\,2,\,\dots$, let

\[ T_n=\left(2^{-(n+1)},2^{{-}n}\right]\quad \hbox{ and }\quad S_n=\left(( 4/3) 2^{-(n+1)}, ( 5/3) 2^{-(n+1)}\right]. \]

That is, $S_n$ is the middle third of $T_n$. Let $t_n= ( 4/3) 2^{-(n+1)}$ be the left end-point of $S_n$.

For fixed $p\in (2,\,\infty )$, let $\varepsilon$ be a fixed rational number such that

\[ 0<\varepsilon<\frac2p<1. \]

We can choose a strictly increasing sequence $k(1)<\dots < k(n)<\dots$ of positive integers such that

\[ 2^{-(\frac2p+\varepsilon)k(n)} \cdot 2\cdot 2^{ \varepsilon k(n) }=2^{- \frac2p k(n)+1} \le (1/3) 2^{-(n+1)}=|S_n| \]

for all $n$ and such that every $\varepsilon k(n)$ is an integer.

For integers $n\ge 1$ and $1\le j\le 2^{ \varepsilon k(n)}$, recall that $t_n$ is the left end-point of $S_n$. Define the intervals

\[ J_{n,j}=(a_{n,j},c_{n,j})=\left(t_n+2^{-(\frac2p+\varepsilon)k(n)} \cdot2\cdot (j-1), t_n+2^{-(\frac2p+\varepsilon)k(n)} \cdot2\cdot j\right) \]

and

\[ I_{n,j}=(a_{n,j},b_{n,j})=\left(t_n+2^{-(\frac2p+\varepsilon)k(n)} \cdot2\cdot (j-1), t_n+2^{-(\frac2p+\varepsilon)k(n)} \cdot (2j-1)\right). \]

It is easy to check that $I_{n,j}$ is the left half of $J_{n,j}$, $J_{n,j}$'s are pairwise disjoint,

\[ \bigcup_{j=1}^{2^{ \varepsilon k(n)}}J_{n,j}\subset S_n, \]

and the length of the interval $I_{n,j}$ is denoted by $\rho _n$, that is

(4.1)\begin{equation} \rho_n=|I_{n,j}|=b_{n,j}- a_{n,j}=2^{-(\frac2p+\varepsilon)k(n)}. \end{equation}

We now define a measurable function $u$ on the unit circle $\mathbb {T}=\{w\in \mathbb {C}:|w|=1\}$ as follows:

\begin{align*} u(\mathrm{e}^{\mathrm{i} t}) & =2^{{-}k(n)} \quad \hbox{ if } t\in \bigcup_{j=1}^{2^{ \varepsilon k(n)}}I_{n,j},n\ge1 ,\\ u(\mathrm{e}^{\mathrm{i} t}) & =1 \quad \hbox{ if } t\in (-\pi,\pi]\setminus\left(\bigcup_{n=1}^\infty \bigcup_{j=1}^{2^{ \varepsilon k(n)}}I_{n,j}\right). \end{align*}

The harmonic extension of $u$ to $\mathbb {D}$ is also denoted by $u$. Let

\[ h(z)=\frac1{2\pi}\int_{-\pi}^\pi\frac{\mathrm{e}^{\mathrm{i} t}+z}{\mathrm{e}^{\mathrm{i} t}-z} \,u(\mathrm{e}^{\mathrm{i} t})\,\mathrm{d} t \]

and

(4.2)\begin{equation} \varphi(z)=\exp({-}h(z)) \end{equation}

for all $z\in \mathbb {D}$. Then, $\mathrm {Re} (h(z))=u(z)>0$ for each $z\in \mathbb {D}$, and thus,

\[ |\varphi(z)|=\mathrm{e}^{\mathrm{Re} (h(z))}=\mathrm{e}^{{-}u(z)}<1. \]

This implies $\varphi (\mathbb {D})\subset \mathbb {D}$. We will need the fact that $\varphi \in H^2$ with

(4.3)\begin{equation} \|\varphi\|_{H^2}^2=\sup_{0< r<1}\frac1{2\pi}\int_{-\pi}^\pi\left| \varphi(r\mathrm{e}^{ \mathrm{i}\theta})\right|^2\mathrm{d}\theta= \frac1{2\pi}\int_{-\pi}^\pi\left| \varphi( \mathrm{e}^{ \mathrm{i}\theta})\right|^2\mathrm{d}\theta. \end{equation}

4.2 Estimates

For $z\in \mathbb {D}$ and $\mathrm {e}^{\mathrm {i} t}\in \mathbb {T}$, let

\[ P(z,\mathrm{e}^{\mathrm{i} t})=\frac{1-|z|^2}{|\mathrm{e}^{\mathrm{i} t}-z|^2} \]

be the Poisson kernel. It is shown in [Reference Xia10, p. 2508] that if $1/2\le r<1$ and $|\theta -t|\le 5$, then there exist constants $0<\alpha <\beta <\infty$ such that

(4.4)\begin{equation} \frac{\alpha(1-r)}{(1-r)^2+(\theta-t)^2}\le \frac1{2\pi} P(r\mathrm{e}^{\mathrm{i}\theta},\mathrm{e}^{\mathrm{i} t})\le \frac{\beta(1-r)}{(1-r)^2+(\theta-t)^2} . \end{equation}

We have the following lemma modified from [Reference Xia10, Lemma 4].

Lemma 4.1 For any positive integer $n$ and $1\le j\le 2^{ \varepsilon k(n)}$, let $G_{n,j}$ be the Carleson box based on $I_{n,j}$, i.e.

(4.5)\begin{equation} G_{n,j}=\left\{r\mathrm{e}^{\mathrm{i}\theta}: \theta\in I_{n,j},0<1-r\le\rho_n\right\}\!. \end{equation}

Then there is a constant $C_1$ independent of $n,\,j$ such that

(4.6)\begin{equation} \int_{G_{n,j}}\left(\frac{1-|z|}{1-|\varphi(z)|}\right)^{p/2}\mathrm{d} \lambda(z)\le C_1 2^{-\frac{p\varepsilon}2 k(n) }. \end{equation}

Proof. Given such a pair of $n,\,j$, we write

\[ G_{n,j}=\bigcup_{\nu=0}^{k(n)} G_{n,j}^\nu, \]

where

\[ G_{n,j}^0=\left\{r\mathrm{e}^{\mathrm{i}\theta}: \theta\in I_{n,j}, 0<1-r\le\rho_n\cdot 2^{{-}k(n)}\right\}\!, \]

and

\[ G_{n,j}^\nu=\left\{r\mathrm{e}^{\mathrm{i}\theta}: \theta\in I_{n,j}, \rho_n\cdot 2^{{-}k(n)}\cdot 2^{\nu-1}<1-r\le\rho_n\cdot 2^{{-}k(n)}\cdot 2^{\nu }\right\}\!, \]

for $1\le \nu \le k(n)$.

It is shown in [Reference Xia10, p. 2509] that there is a constant $0< c<1$ independent of $n,\,j$ such that

\[ 1-|\varphi(z)|=1-\mathrm{e}^{{-}u(z)}\ge 1-\exp({-}c 2^{{-}k(n)+\nu}) \]

if $z\in G_{n,j}^\nu$ and $0\le \nu \le k(n)$. Let $\delta =\inf _{0< x\le 1}x^{-1}(1-\mathrm {e}^{-x})$. Then,

(4.7)\begin{align} \inf_{z\in G_{n,j}^\nu}\left(1-|\varphi(z)|\right)^{p/2}\ge (\delta c)^{p/2}\cdot 2^{{-}p/2k(n)}\cdot 2^{p/2\nu}, \quad 0\le\nu \le k(n). \end{align}

This implies that

(4.8)\begin{align} & \int_{G_{n,j}}\left(\frac{1-|z|}{1-|\varphi(z)|}\right)^{p/2}\mathrm{d} \lambda(z)\nonumber\\& \quad=\int_{G^0_{n,j}}\left(\frac{1-|z|}{1-|\varphi(z)|}\right)^{p/2}\mathrm{d} \lambda(z)+\sum_{\nu=1}^{k(n)} \int_{G^\nu_{n,j}}\left(\frac{1-|z|}{1-|\varphi(z)|}\right)^{p/2}\mathrm{d} \lambda(z)\nonumber\\ & \quad\le\frac{2^{p/2k(n)}}{(\delta c)^{p/2}}\int_{G^0_{n,j}}(1-|z|^2)^{p/2-2}\mathrm{d} A(z)\nonumber\\& \qquad+ \sum_{\nu=1}^{k(n)}\frac{2^{p/2k(n)}}{(\delta c)^{p/2} \cdot 2^{p/2\nu}} \int_{G^\nu_{n,j}} (1-|z| )^{p/2-2}\mathrm{d} A(z). \end{align}

Notice that $p/2-2>-1$. Straightforward computation shows that

(4.9)\begin{align} \int_{G^0_{n,j}}(1-|z|^2)^{p/2-2}\mathrm{d} A(z) & =\frac1\pi \int_{I_{n,j}} \mathrm{d} \theta\int_{1-\rho_n\cdot2^{{-}k(n)}}^1(1-r^2)^{p/2-2} r\, \mathrm{d} r\nonumber\\& \le C_2 \rho_n^{p/2}\cdot 2^{-(p/2-1)k(n)} \end{align}

for some $C_2>0$, and

(4.10)\begin{align} \int_{G^\nu_{n,j}}(1-|z| )^{p/2-2}\mathrm{d} A(z) & =\frac1\pi \int_{I_{n,j}} \mathrm{d} \theta\int_{1-\rho_n\cdot2^{{-}k(n)}\cdot 2^\nu}^{1-\rho_n\cdot 2^{{-}k(n)}\cdot 2^{\nu-1}}(1-r )^{p/2-2}r\, \mathrm{d} r\nonumber\\& \le C_3 \rho_n^{p/2}\cdot 2^{-(p/2-1)k(n)} \cdot 2^{(p/2-1)\nu} \end{align}

for some $C_3>0$. Put (4.8), (4.9) and (4.10) together, we have

\begin{align*} & \int_{G_{n,j}}\left(\frac{1-|z|}{1-|\varphi(z)|}\right)^{p/2}\mathrm{d} \lambda(z)\\& \quad \le\frac{C_2 \cdot 2^{ k(n)} \cdot \rho_n^{p/2}}{(\delta c)^{p/2}} + \sum_{\nu=1}^{k(n)}\frac{2^{p/2k(n)}\cdot C_3 \rho_n^{p/2}\cdot 2^{-(p/2-1)k(n)} \cdot 2^{(p/2-1)\nu}}{(\delta c)^{p/2} \cdot 2^{p/2\nu}}\\ & \quad= 2^{ k(n)} \cdot \rho_n^{p/2}\cdot\left( \frac{C_2 }{(\delta c)^{p/2}}+ \frac{C_3 }{(\delta c)^{p/2}} \sum_{\nu=1}^{k(n)} 2^{-\nu}\right)\!. \end{align*}

Recall the inequality (4.1), we get the desired inequality (4.6) by letting

\[ C_1= \frac{C_2 }{(\delta c)^{p/2}}+ \frac{C_3 }{(\delta c)^{p/2}} \sum_{\nu=1}^{\infty} 2^{-\nu}= \frac{C_2 +C_3 }{(\delta c)^{p/2}}. \]

The following lemma is quoted from [Reference Xia10, Lemma 7].

Lemma 4.2 There is a $C_4>0$ such that

\[ u(z)\ge C_4 \quad\hbox{ for every }\quad z\in\mathbb{D}\setminus\left( \bigcup_{n=1}^\infty \bigcup_{j=1}^{2^{ \varepsilon k(n)}} G_{n,j}\right)\!, \]

where $G_{n,j}$ is defined by (4.5).

4.3 Proof of theorem 1.2

Let $\varphi$ be the holomorphic self-map of $\mathbb {D}$ given by (4.2). It is sufficient to check the inequality (1.2) for this $\varphi$, and $C_{\varphi }\notin \mathcal {S}_p(H^2)$.

Let

\[ G=\bigcup_{n=1}^\infty \bigcup_{j=1}^{2^{ \varepsilon k(n)}} G_{n,j}, \]

where $G_{n,j}$ is given by (4.5). For $z\in \mathbb {D}\setminus G$, lemma 4.2 implies that

\[ |\varphi(z)|=\mathrm{e}^{-\mathrm{Re}(h(z))}=\mathrm{e}^{{-}u(z)}\le \mathrm{e}^{{-}C_4}. \]

Since $p/2-2>-1$, we have

(4.11)\begin{align} \int_{\mathbb{D}\setminus G}\left(\frac{1-|z|^2}{1-|\varphi(z)|^2}\right)^{p/2}\mathrm{d} \lambda(z)& \le \frac1{(1-\mathrm{e}^{{-}C_4})^{p/2}} \int_{\mathbb{D}\setminus G} (1-|z|^2)^{p/2-2}\mathrm{d} A(z) \nonumber\\ & \le \frac1{(1-\mathrm{e}^{{-}C_4})^{p/2}} \int_{\mathbb{D}} (1-|z|^2)^{p/2-2}\mathrm{d} A(z)<\infty. \end{align}

Meanwhile, lemma 4.1 implies that

(4.12)\begin{align} \int_G \left(\frac{1-|z|^2}{1-|\varphi(z)|^2}\right)^{p/2}\mathrm{d} \lambda(z)& \approx \int_G \frac{(1-|z|)^{p/2-2} }{(1-|\varphi(z)|)^{p/2} }\, \mathrm{d} A(z)\nonumber\\& =\sum_{n=1}^\infty \sum_{j=1}^{2^{\varepsilon k(n)}}\int_{G_{n,j}} \frac{(1-|z|)^{p/2-2} }{(1-|\varphi(z)|)^{p/2} }\, \mathrm{d} A(z) \nonumber\\ & \le C_1 \sum_{n=1}^\infty 2^{\varepsilon k(n)}\cdot 2^{-\frac{p\varepsilon}2 k(n) } \le C_1 \sum_{n=1}^\infty 2^{-(p/2-1)\varepsilon k(n)}<\infty, \end{align}

where the last inequality is following from the fact that $p/2-1>0$. Now (1.2) follows from (4.11) and (4.12) easily.

It remains to check that $C_{\varphi }\notin \mathcal {S}_p(H^2)$, or equivalently, $\mathrm {tr}((C_{\varphi }^\ast C_{\varphi })^{\frac p2})=\infty$. Let $e_\ell (z)=z^\ell$, $\ell =0,\,1,\,2,\,\dots$. It is well known that $\{e_\ell :\ell \ge 0\}$ is an orthonormal basis for $H^2$. Since $p/2>1$, we have

\begin{align*} & \left\langle \left(C_{\varphi}^\ast C_{\varphi}\right)^{p/2} e_\ell,e_\ell\right \rangle_{H^2}\ge \left(\left\langle C_{\varphi}^\ast C_{\varphi} e_\ell, e_\ell\right\rangle_{H^2}\right)^{p/2}\\ & \quad =\left\| C_{\varphi} e_\ell\right\|_{H^2}^p= \left\| \varphi^l \right\|_{H^2}^p=\left(\frac1{2\pi}\int_{-\pi}^\pi \left|\varphi(\mathrm{e}^{ \mathrm{i}\theta})\right|^{2\ell}\mathrm{d}\theta\right)^{p/2}. \end{align*}

Write

\[ I_n=\bigcup_{j=1}^{2^{\varepsilon k(n)}}I_{n,j}. \]

Then,

\[ |I_n|=2^{\varepsilon k(n)}\rho_n= 2^{- \frac2p k(n)}, \]

and

\[ \left|\varphi(\mathrm{e}^{\mathrm{i}\theta})\right|= \exp({-u(\mathrm{e}^{\mathrm{i}\theta})}) =\exp({-}2^{{-}k(n)}) \]

for almost every $\theta \in I_n$. Thus,

\[ \int_{-\pi}^\pi \left|\varphi(\mathrm{e}^{ \mathrm{i}\theta})\right|^{2\ell}\mathrm{d}\theta \ge\sum_{n=1}^\infty \int_{I_{n }} \left|\varphi(\mathrm{e}^{ \mathrm{i}\theta})\right|^{2\ell}\mathrm{d} \theta= \sum_{n=1}^\infty \mathrm{e}^{{-}2\ell\cdot 2^{{-}k(n)}}\cdot 2^{- \frac2p k(n)}. \]

Notice that

\[ \left(\sum_n a_n\right)^s\ge \sum_n a_n^s \]

if $s\ge 1$ and $a_n\ge 0$. We get

\[ \left( \int_{-\pi}^\pi \left|\varphi(\mathrm{e}^{ \mathrm{i}\theta})\right|^{2\ell}\mathrm{d}\theta \right)^{p/2}\ge \left(\sum_{n=1}^\infty \mathrm{e}^{{-}2\ell\cdot 2^{{-}k(n)}}\cdot 2^{- \frac2p k(n)}\right)^{p/2}\ge \sum_{n=1}^\infty \mathrm{e}^{{-}p\ell\cdot 2^{{-}k(n)}}\cdot 2^{- k(n)}. \]

This gives that

\begin{align*} \mathrm{tr}\left((C_{\varphi}^\ast C_{\varphi})^{p/2}\right)& =\sum_{\ell=0}^\infty \left\langle (C_{\varphi}^\ast C_{\varphi})^{p/2} e_\ell,e_\ell\right\rangle_{H^2}\ge \sum_{\ell=0}^\infty \left(\frac1{2\pi}\int_{-\pi}^\pi \left|\varphi(\mathrm{e}^{ \mathrm{i}\theta})\right|^{2\ell}\mathrm{d}\theta\right)^{p/2}\\ & \ge\frac1{(2\pi)^{p/2}} \sum_{\ell=0}^\infty \sum_{n=1}^\infty \mathrm{e}^{{-}p\ell\cdot 2^{{-}k(n)}}\cdot 2^{- k(n)}\\ & =\frac1{(2\pi)^{p/2}}\sum_{n=1}^\infty \left( 2^{- k(n)} \sum_{\ell=0}^\infty \mathrm{e}^{{-}p\ell\cdot 2^{{-}k(n)}}\right)\\ & = \frac1{(2\pi)^{p/2}}\sum_{n=1}^\infty 2^{- k(n)} \cdot\frac1{1-\mathrm{e}^{{-}p \cdot 2^{{-}k(n)}}}. \end{align*}

Since

\[ \sup_{x>0}\frac{1-\mathrm{e}^{{-}x}}x\le1. \]

We have

\[ \frac1{1-\mathrm{e}^{{-}p \cdot 2^{{-}k(n)}}}\ge \frac1{p \cdot 2^{{-}k(n)}}. \]

Then,

\[ \sum_{n=1}^\infty 2^{- k(n)} \cdot\frac1{1-\mathrm{e}^{{-}p \cdot 2^{{-}k(n)}}}\ge \sum_{n=1}^\infty 2^{- k(n)} \cdot\frac1{p \cdot 2^{{-}k(n)}}=\sum_{n=1}^\infty\frac1p=\infty. \]

This implies that $C_{\varphi }\notin \mathcal {S}_p(H^2)$ and the proof is complete.

Acknowledgements

This work is supported by Guangdong Basic and Applied Basic Research Foundation (2022A1515010358).

References

Benazzouz, H., El-Fallah, O., Kellay, K. and Mahzouli, H.. Contact points and Schatten composition operators. Math. Z. 279 (2015), 407422.CrossRefGoogle Scholar
Bendaoud, Z., Korrichi, F., Merghni, L. and Yagoub, A.. Contact points and Schatten class of composition operators. Indian J. Pure Appl. Math. 49 (2018), 651661.CrossRefGoogle Scholar
Cowen, C. C. and MacCluer, B. D.. Composition Operators on Spaces of Analytic Functions. Studies in Advanced Mathematics (CRC Press, Boca Raton, FL, 1995).Google Scholar
Luecking, D. H.. Trace ideal criteria for Toeplitz operators. J. Funct. Anal. 73 (1987), 345368.CrossRefGoogle Scholar
Luecking, D. and Zhu, K.. Composition operators belonging to the Scatten ideals. Amer. J. Math. 114 (1992), 11271145.CrossRefGoogle Scholar
Pau, J. and Perälä, A.. A Toeplitz-type operator on Hardy spaces in the unit ball. Trans. Amer. Math. Soc. 373 (2020), 30313062.CrossRefGoogle Scholar
Pau, J. and Pérez, P.. Composition operators acting on weighted Dirichlet spaces. J. Math. Anal. Appl. 401 (2013), 682694.CrossRefGoogle Scholar
Shapiro, J. H.. Composition Operators and Classical Function Theory. Springer (New York, 1993).CrossRefGoogle Scholar
Wirths, K.-J. and Xiao, J.. Global integral criteria for composition operators. J. Math. Anal. Appl. 269 (2002), 702715.CrossRefGoogle Scholar
Xia, J.. On a proposed characterization of Schatten-class composition operators. Proc. Amer. Math. Soc. 131 (2003), 25052514.CrossRefGoogle Scholar
Yang, W. and Liu, J.. Schatten classes of Toeplitz operators on Bergman-Besov Hilbert spaces in the unit ball. J. Math. Anal. Appl. 526 (2023), 127257.CrossRefGoogle Scholar
Yuan, C. and Zhou, Z.. Composition operators belonging to the Schatten class $\mathcal {S}_p$. Bull. Aust. Math. Soc. 81 (2010), 465472.CrossRefGoogle Scholar
Yuan, C. and Zhou, Z.. The Hilbert-Schmidt norm of a composition operator on the Bergman space. Bull. Aust. Math. Soc. 95 (2017), 250259.CrossRefGoogle Scholar
Zhu, K.. Schatten class composition operators on weighted Bergman spaces of the disk. J. Operator Theory 46 (2001), 173181.Google Scholar
Zhu, K.. Translating certain inequalities between Hardy and Bergman spaces. Amer. Math. Monthly 111 (2004), 520525.CrossRefGoogle Scholar
Zhu, K.. Operator Theory in Function Spaces (American Mathematical Society, Providence, RI, 2007).CrossRefGoogle Scholar