The fundamental ideals of the proof are those used by J. H. Shpairo [8] to obtain the essential norm of a composition operator on Hilbert spaces of analytic functions (Hardy and weighted Bergman spaces) in terms of natural counting functions associated with $\phi$  .

This paper generalizes the result on the Bloch space in [10] to the Bloch-type space in polydisk.

Throughout the remainder of this paper $C$  will denote a positive constant, the exact value of which will vary from one appearance to the next.

Our main results are the following:

Theorem 1 Let $\phi =({\phi }_1,{\phi }_2,\cdots ,{\phi }_n)$  be a holomorphic self-map of $U^n$  and $\parallel C_{\phi }{\parallel }_e$  the essential norm of a bounded composition operator $C_{\phi }:$  ${\mathcal{ℬ}}^p\left(U^n\right)$  ( ${\mathcal{ℬ}}_0^p\left(U^n\right)$  or ${\mathcal{ℬ}}_{0*}^p\left(U^n\right)$  ) $\to$  ${\mathcal{ℬ}}^q\left(U^n\right)$  ( ${\mathcal{ℬ}}_0^q\left(U^n\right)$  or ${\mathcal{ℬ}}_{0*}^q\left(U^n\right)$  ) , then

$$\begin{array}{ccc}& & \frac{1}{n}{lim}_{\delta \to 0}{sup}_{dist\left(\phi \right(z),\partial U^n)<\delta }{\sum }_{k,l=1}^n\left|\frac{\partial {\phi }_l}{\partial z_k}\left(z\right)\right|\frac{(1-|z_k{|}^2{)}^q}{(1-|{\phi }_l\left(z\right){|}^2{)}^p}\end{array}$$

$$\begin{array}{ccc}& & \le \parallel C_{\phi }{\parallel }_e\le 2{lim}_{\delta \to 0}{sup}_{dist\left(\phi \right(z),\partial U^n)<\delta }{\sum }_{k,l=1}^n\left|\frac{\partial {\phi }_l}{\partial z_k}\left(z\right)\right|\frac{(1-|z_k{|}^2{)}^q}{(1-|{\phi }_l\left(z\right){|}^2{)}^p}.\end{array}$$

(3)

By Theorem 1 and the fact that $C_{\phi }:$  ${\mathcal{ℬ}}^p\left(U^n\right)$  (or ${\mathcal{ℬ}}_0^p\left(U^n\right)$  or ${\mathcal{ℬ}}_{0*}^p\left(U^n\right)$  ) $\to$  ${\mathcal{ℬ}}^q\left(U^n\right)$  (or ${\mathcal{ℬ}}_0^q\left(U^n\right)$  or ${\mathcal{ℬ}}_{0*}^q\left(U^n\right)$  ) is compact if and only if $\parallel C_{\phi }{\parallel }_e=0$  , we obtain Theorem 2 at once.

Theorem 2 Let $\phi =({\phi }_1,\dots ,{\phi }_n)$  be a holomorphic self-map of $U^n.$  Then the bounded composition operator $C_{\phi }:$  ${\mathcal{ℬ}}^p\left(U^n\right)$  ( ${\mathcal{ℬ}}_0^p\left(U^n\right)$  or ${\mathcal{ℬ}}_{0*}^p\left(U^n\right)$  ) $\to$  ${\mathcal{ℬ}}^q\left(U^n\right)$  ( ${\mathcal{ℬ}}_0^q\left(U^n\right)$  or ${\mathcal{ℬ}}_{0*}^q\left(U^n\right)$  ) is compact if and only if for any $\varepsilon >0,$  there exists a $\delta$  with $0<\delta <1,$  such that

$$\begin{array}{c}{sup}_{dist\left(\phi \right(z),\partial U^n)<\delta }{\sum }_{k,l=1}^n\left|\frac{\partial {\phi }_l}{\partial z_k}\left(z\right)\right|\frac{(1-|z_k{|}^2{)}^q}{(1-|{\phi }_l\left(z\right){|}^2{)}^p}<\varepsilon .\end{array}$$

(4)

When $n=1,$  on $\mathcal{ℬ}\left(U\right)$  we obtain Theorem 2 in [3]. Since $\partial U={\partial }^{*}U,$  ${\mathcal{ℬ}}_0\left(U\right)={\mathcal{ℬ}}_{0*}\left(U\right),$  we can also obtain Theorem 1 in [3].

By Theorem 2 and Lemmas 3, 4 and 5 in next part, we can get the following three Corollaries.

Corollary 1 Let $\phi =({\phi }_1,\dots ,{\phi }_n)$  be a holomorphic self-map of $U^n.$  Then $C_{\phi }:{\mathcal{ℬ}}^p\left(U^n\right)$  ( ${\mathcal{ℬ}}_0^p\left(U^n\right)$  or ${\mathcal{ℬ}}_{0*}^p\left(U^n\right)$  ) $\to {\mathcal{ℬ}}^q\left(U^n\right)$  is compact if and only if $${\sum }_{k,l=1}^n\left|\frac{\partial {\phi }_l}{\partial z_k}\left(z\right)\right|\frac{(1-|z_k{|}^2{)}^q}{(1-|{\phi }_l\left(z\right){|}^2{)}^p}\le C$$  for all $z\in U^n$  and ( 4 ) holds.

Proof By Lemma 3 in next part, we know $C_{\phi }:{\mathcal{ℬ}}^p\left(U^n\right)$  ( ${\mathcal{ℬ}}_0^p\left(U^n\right)$  or ${\mathcal{ℬ}}_{0*}^p\left(U^n\right)$  ) $\to {\mathcal{ℬ}}^q\left(U^n\right)$  is bounded. It follows from Theorem 2 that $C_{\phi }:{\mathcal{ℬ}}^p\left(U^n\right)$  ( ${\mathcal{ℬ}}_0^p\left(U^n\right)$  or ${\mathcal{ℬ}}_{0*}^p\left(U^n\right)$  ) $\to {\mathcal{ℬ}}^q\left(U^n\right)$  is compact.

Conversely, if $C_{\phi }:{\mathcal{ℬ}}^p\left(U^n\right)$  ( ${\mathcal{ℬ}}_0^p\left(U^n\right)$  or ${\mathcal{ℬ}}_{0*}^p\left(U^n\right)$  ) $\to {\mathcal{ℬ}}^q\left(U^n\right)$  is compact, it is clear that $C_{\phi }:{\mathcal{ℬ}}^p\left(U^n\right)$  ( ${\mathcal{ℬ}}_0^p\left(U^n\right)$  or ${\mathcal{ℬ}}_{0*}^p\left(U^n\right)$  ) $\to {\mathcal{ℬ}}^q\left(U^n\right)$  is bounded, by Theorem 2, ( 4 ) holds.

Corollary 2 Let $\phi =({\phi }_1,\dots ,{\phi }_n)$  be a holomorphic self-map of $U^n.$  Then $C_{\phi }:$  ${\mathcal{ℬ}}_{0*}^p\left(U^n\right)$  ( ${\mathcal{ℬ}}_0^p\left(U^n\right)$  ) $\to {\mathcal{ℬ}}_{0*}^q\left(U^n\right)$  is compact if and only if ${\phi }_l\in {\mathcal{ℬ}}_{0*}^q\left(U^n\right)$  for every $l=1,2,\cdots ,n$  and ( 4 ) holds.

Proof Note that Lemma 4 in next part, similar to the proof of Corollary 1, the Corollary follows.

Corollary 3 Let $\phi =({\phi }_1,\dots ,{\phi }_n)$  be a holomorphic self-map of $U^n.$  Then $C_{\phi }:$  ${\mathcal{ℬ}}_0^p\left(U^n\right)\to {\mathcal{ℬ}}_0^q\left(U^n\right)$  is compact if and only if ${\phi }_l\in {\mathcal{ℬ}}_0^q\left(U^n\right)$  for every $l=1,2,\cdots ,n$  and ( 4 ) holds.

Proof Note that Lemma 5 in next part, similar to the proof of Corollary 1, the Corollary follows.

2 Some Lemmas

In order to prove Theorem 1, we need some Lemmas.

Lemma 1 Let $f\in {\mathcal{ℬ}}^p\left(U^n\right),$  then (1) If $0\le p<1,$  then $\parallel f\left(z\right)|\le |f\left(0\right)|+\frac{n}{1-p}\parallel f{\parallel }_p;$  (2) If $p=1,$  then $\left|f\right(z\left)\right|\le \left(1+\frac{1}{nln2}\right)\left({\sum }_{k=1}^{n}ln\frac{2}{1-|z_k{|}^2}\right)\parallel f{\parallel }_p.$  (3) If $p>1,$  then $\left|f\right(z\left)\right|\le \left(\frac{1}{n}+\frac{2^{p-1}}{p-1}\right){\sum }_{k=1}^n\frac{1}{(1-|z_k{|}^2{)}^{p-1}}\parallel f{\parallel }_p.$ 

Proof This Lemma can be proved by some integral estimates (if necessary, the proof can be omitted).

By the definition of $\parallel .{\parallel }_p$  , $$\left|f\right(0\left)\right|\le \parallel f{\parallel }_p,\left|\frac{\partial f\left(z\right)}{\partial z_l}\right|\le \frac{\parallel f{\parallel }_p}{(1-|z_l{|}^2{)}^p}(l\in \{1,2,\cdots ,n\left\}\right)$$  and

$$\begin{array}{ccc}& & f\left(z\right)-f\left(0\right)={\int }_0^1\frac{df\left(tz\right)}{dt}dt={\sum }_{l=1}^n{\int }_0^1{z}_l\frac{\partial f}{\partial {\zeta }_l}\left(tz\right)dt,\end{array}$$

So

$$\begin{array}{ccc}& & \left|f\right(z\left)\right|\le \left|f\right(0\left)\right|+{\sum }_{l=1}^n\left|z_l\right|{\int }_0^1\frac{\parallel f{\parallel }_p}{{\left(1-t^2|z_l{|}^2\right)}^p}dt\end{array}$$

$$\begin{array}{ccc}& & \le \parallel f{\parallel }_p+\parallel f{\parallel }_p{\sum }_{l=1}^n{\int }_0^{\left|z_l\right|}\frac{1}{{\left(1-t^2\right)}^p}dt.\end{array}$$

(5)

If $p=1,$ 

$$\begin{array}{c}{\int }_0^{\left|z_l\right|}\frac{1}{{\left(1-t^2\right)}^p}dt=\frac{1}{2}ln\frac{1+\left|z_l\right|}{1-\left|z_l\right|}\le \frac{1}{2}ln\frac{4}{1-|z_l{|}^2}.\end{array}$$

(6)

It is clear that $ln\frac{4}{1-|z_l{|}^2}>ln4=2ln2,$  so

$$\begin{array}{c}1\le \frac{1}{2ln2}ln\frac{4}{1-|z_l{|}^2}\le \frac{1}{2nln2}{\sum }_{l=1}^{n}ln\frac{4}{1-|z_l{|}^2}.\end{array}$$

(7)

Combining ( 5 ),( 6 ) and ( 7 ), we get $$\left|f\right(z\left)\right|\le \left(\frac{1}{2}+\frac{1}{2nln2}\right)\left({\sum }_{l=1}^{n}ln\frac{4}{1-|z_l{|}^2}\right)\parallel f{\parallel }_p.$$  If $p\ne 1,$ 

$$\begin{array}{ccc}& & {\int }_0^{\left|z_l\right|}\frac{1}{{\left(1-t^2\right)}^p}dt={\int }_0^{\left|z_l\right|}\frac{1}{(1-t{)}^p}\cdot \frac{1}{(1+t{)}^p}dt\end{array}$$

$$\begin{array}{ccc}& & \le {\int }_0^{\left|z_l\right|}\frac{1}{(1-t{)}^p}dt=\frac{1-(1-|z_l|{)}^{-p+1}}{1-p}.\end{array}$$

(8)

If $0<p<1,$  ( 8 ) gives that ${\int }_0^{\left|z_l\right|}\frac{1}{{\left(1-t^2\right)}^p}dt\le \frac{1}{1-p},$  it follows from ( 5 ) that $\left|f\right(z\left)\right|\le \left(1+\frac{n}{1-p}\right)\parallel f{\parallel }_p.$  If $p>1,$  ( 8 ) gives that $${\int }_0^{\left|z_l\right|}\frac{1}{{\left(1-t^2\right)}^p}dt\le \frac{1-(1-|z_l{|}^{p-1})}{(p-1)(1-|z_l|{)}^{p-1}}\le \frac{2^{p-1}}{(p-1)(1-|z_l^2|{)}^{p-1}},$$  it follows from ( 5 ) that

$$\begin{array}{ccc}\left|f\right(z\left)\right|& \le & \parallel f{\parallel }_p+\frac{2^{p-1}}{p-1}\left({\sum }_{l=1}^n\frac{1}{(1-|z_l{|}^2{)}^{p-1}}\right)\parallel f{\parallel }_p\end{array}$$

$$\begin{array}{ccc}& \le & \left(\frac{1}{n}+\frac{2^{p-1}}{p-1}\right)\left({\sum }_{l=1}^n\frac{1}{(1-|z_l{|}^2{)}^{p-1}}\right)\parallel f{\parallel }_p.\end{array}$$

Now the Lemma is proved.

Lemma 2 Set $$f_w\left(z\right)={\int }_0^{z_l}\frac{dt}{(1-\overline{w}t{)}^p},$$  where $w\in U.$  Then $f\in {\mathcal{ℬ}}_0^p\left(U^n\right)\subset {\mathcal{ℬ}}_{0*}^p\left(U^n\right)\subset {\mathcal{ℬ}}^p\left(U^n\right).$ 

Proof Since $$\frac{\partial f_w}{\partial z_l}={\left(1-\overline{w}t\right)}^{-p},\frac{\partial f_w}{\partial z_i}=0,(i\ne l),$$  it follows that $$\left|f\right(0\left)\right|+{\sum }_{k=1}^n\left|\frac{\partial f_w}{\partial z_k}\left(z\right)\right|(1-|z_k{|}^2{)}^p=\frac{(1-|z_l{|}^2{)}^p}{|1-\overline{w}t{|}^p}\le (1+|z_l{|}^2{)}^p\le 2^p.$$  Hence $f_w\in {\mathcal{ℬ}}^p\left(U^n\right).$  Now we prove that $f_w\in {\mathcal{ℬ}}_0^p\left(U^n\right).$  Using the asymptotic formula $$(1-\overline{w}t{)}^{-p}={\sum }_{k=0}^{+\infty }\frac{p(p+1)\cdots (p+k-1)}{k!}(\overline{w}{)}^k{t}^k,$$  we obtain $$f_w\left(z\right)={\sum }_{k=0}^{+\infty }\frac{p(p+1)\cdots (p+k-1)}{k!}(\overline{w}{)}^k{\int }_0^{z_l}{t}^{k}dt.$$  Denote $P_n\left(z\right)={\sum }_{k=0}^n\frac{p(p+1)\cdots (p+k-1)}{k!}(\overline{w}{)}^k{\int }_0^{z_l}{t}^{k}dt,$  it is easy to see that $$f_w\left(z\right)-P_n\left(z\right)={\sum }_{k=n+1}^{+\infty }\frac{p(p+1)\cdots (p+k-1)}{k!}(\overline{w}{)}^k{\int }_0^{z_l}{t}^{k}dt,$$  $$\left|\frac{\partial (f_w-P_n)}{\partial z_l}\right|\le {\sum }_{k=n+1}^{+\infty }\frac{p(p+1)\cdots (p+k-1)}{k!}|w{|}^k\to 0,\text{as}n\to \infty ,$$ 

$$\begin{array}{ccc}\parallel f_w-P_n{\parallel }_p& =& \left|f_w\right(0)-P_n(0\left)\right|+{sup}_{z\in U^n}\left|\frac{\partial (f_w-P_n)}{\partial z_l}\right|(1-|z_l{|}^2{)}^p\end{array}$$

$$\begin{array}{ccc}& \le & {sup}_{z\in U^n}\left|\frac{\partial (f_w-P_n)}{\partial z_l}\right|\to 0,\end{array}$$

it shows that $f_w\in {\mathcal{ℬ}}_0^p\left(U^n\right).$  So $f\in {\mathcal{ℬ}}_0^p\left(U^n\right)\subset {\mathcal{ℬ}}_{0*}^p\left(U^n\right)\subset {\mathcal{ℬ}}^p\left(U^n\right).$ 

Lemma 3 Let $\phi =({\phi }_1,\dots ,{\phi }_n)$  be a holomorphic self-map of $U^n$  , $p,q>0.$  Then $C_{\phi }:{\mathcal{ℬ}}^p\left(U^n\right)\left({\mathcal{ℬ}}_0^p\right(U^n)$  or ${\mathcal{ℬ}}_{0*}^p\left(U^n\right))⟶{\mathcal{ℬ}}^q(U^n)$  is bounded if and only if there exists a constant $C$  such that

$$\begin{array}{c}{\sum }_{k,l=1}^n\left|\frac{\partial {\phi }_l}{\partial z_k}\left(z\right)\right|\frac{{\left(1-|z_k{|}^2\right)}^q}{{\left(1-\left|{\phi }_l\right(z){|}^2\right)}^p}\le C,\end{array}$$

(9)

for all $z\in U^n.$ 

Proof First assume that condition ( 9 ) holds. Let $f\in {\mathcal{ℬ}}^p\left(U^n\right)\left({\mathcal{ℬ}}_0^p\right(U^n)$  or ${\mathcal{ℬ}}_{0*}^p\left(U^n\right)),$  by Lemma 1, we know the evaluation at $\phi \left(0\right)$  is a bounded linear functional on ${\mathcal{ℬ}}^p\left(U^n\right),$  so $\left|f\right(\phi \left(0\right)\left)\right|\le C\parallel f{\parallel }_p.$  On the other hand we have

$$\begin{array}{ccc}& & {\sum }_{k=1}^n\left|\frac{\partial \left(C_{\phi }f\left(z\right)\right)}{\partial z_k}\right|(1-|z_k{|}^2{)}^q={\sum }_{k=1}^n\left|{\sum }_{l=1}^n\frac{\partial f}{\partial {\phi }_l}\left(\phi \right(z\left)\right)\frac{\partial {\phi }_l}{\partial z_k}\left(z\right)\right|(1-|z_k{|}^2{)}^q\end{array}$$

$$\begin{array}{ccc}& & \le {\sum }_{k,l=1}^n\left|\frac{\partial f}{\partial {\phi }_l}\left(\phi \right(z\left)\right)\frac{\partial {\phi }_l}{\partial z_k}\left(z\right)\right|(1-|z_k{|}^2{)}^q\end{array}$$

$$\begin{array}{ccc}& & \le {\sum }_{l=1}^n\left|\frac{\partial f}{\partial {\phi }_l}\left(\phi \right(z\left)\right)\right|{\left(1-\left|{\phi }_l\right(z){|}^2\right)}^p{\sum }_{k,l=1}^n\left|\frac{\partial {\phi }_l}{\partial z_k}\left(z\right)\right|\frac{{\left(1-|z_k{|}^2\right)}^q}{{\left(1-\left|{\phi }_l\right(z){|}^2\right)}^p}\end{array}$$

(10)

$$\begin{array}{ccc}& & \le \parallel f{\parallel }_p{\sum }_{k,l=1}^n\left|\frac{\partial {\phi }_l}{\partial z_k}\left(z\right)\right|\frac{{\left(1-|z_k{|}^2\right)}^q}{{\left(1-\left|{\phi }_l\right(z){|}^2\right)}^p}.\end{array}$$

(11)

From ( 11 ) it follows that $${\sum }_{k=1}^n\left|\frac{\partial \left(C_{\phi }f\left(z\right)\right)}{\partial z_k}\right|(1-|z_k{|}^2{)}^q\le C\parallel f{\parallel }_p.$$  So $C_{\phi }:{\mathcal{ℬ}}^p\left(U^n\right)\to {\mathcal{ℬ}}^q\left(U^n\right)$  is bounded.

For the converse, assume that $C_{\phi }:{\mathcal{ℬ}}^p\left(U^n\right)\left({\mathcal{ℬ}}_0^p\right(U^n)$  or ${\mathcal{ℬ}}_{0*}^p\left(U^n\right))⟶{\mathcal{ℬ}}^q(U^n)$  is bounded, with

$$\begin{array}{c}\parallel C_{\phi }f{\parallel }_q\le C\parallel f{\parallel }_p\end{array}$$

(12)

for all $f\in {\mathcal{ℬ}}^p\left(U^n\right)\left({\mathcal{ℬ}}_0^p\right(U^n)$  or ${\mathcal{ℬ}}_{0*}^p\left(U^n\right)).$  For fixed $l(1\le l\le n),$  we will make use of a family of test functions $\{f_w:w\in \text{C},|w|<1\}$  in $\mathcal{ℬ}\left(U^n\right)$  defined as follows: If $p>0$  , let $$f_w\left(z\right)={\int }_0^{z_l}{\left(1-\overline{w}{z}_l\right)}^{-p}dt.$$  It follows from Lemma 2 that $$f_w\in {\mathcal{ℬ}}_0^p\left(U^n\right)\subset {\mathcal{ℬ}}_{0*}^p\left(U^n\right)\subset {\mathcal{ℬ}}^p\left(U^n\right).$$  For $z\in U^n,$  it follows from ( 12 ) that

$$\begin{array}{c}{\sum }_{k=1}^n\left|{\sum }_{l=1}^n\frac{\partial f_w\left(\phi \right(z\left)\right)}{\partial {\phi }_l}\frac{\partial {\phi }_l}{\partial z_k}\left(z\right)\right|(1-|z_k{|}^2{)}^q\le C.\end{array}$$

(13)

Let $w={\phi }_l\left(z\right),$  then $${\sum }_{k=1}^n\left|\frac{\partial {\phi }_l}{\partial z_k}\left(z\right)\right|\frac{{\left(1-|z_k{|}^2\right)}^q}{{\left(1-\left|{\phi }_l\right(z){|}^2\right)}^p}\le C.$$  Now the proof of Lemma 3 is completed.

Lemma 4 Let $\phi =({\phi }_1,{\phi }_2,\cdots ,{\phi }_n)$  be a holomorphic self-map of $U^n.$  Then $C_{\phi }:{\mathcal{ℬ}}_{0*}^p\left(U^n\right)\left({\mathcal{ℬ}}_0^p\right(U^n\left)\right)\to {\mathcal{ℬ}}_{0*}^q\left(U^n\right)$  is bounded if and only if ${\phi }_l\in {\mathcal{ℬ}}_{0*}^q\left(U^n\right)$  for every $l=1,2,\cdots ,n$  and ( 9 ) holds.

Proof If $C_{\phi }:{\mathcal{ℬ}}_{0*}^p\left(U^n\right)\left({\mathcal{ℬ}}_0^p\right(U^n\left)\right)\to {\mathcal{ℬ}}_{0*}^q\left(U^n\right)$  is bounded , it is clear that, for every $l=1,2,\cdots ,n$  , $f_l\left(z\right)=z_l\in {\mathcal{ℬ}}_0^p\left(U^n\right)\subset {\mathcal{ℬ}}_{0*}^q\left(U^n\right),$  so $C_{\phi }{f}_l={\phi }_l\in {\mathcal{ℬ}}_{0*}^q\left(U^n\right).$  In the proof of Lemma 3, note that the test functions $f_w\in {\mathcal{ℬ}}_0^p\left(U^n\right)\subset {\mathcal{ℬ}}_{0*}^p\left(U^n\right),$  we know  9  $\left(\right)$  holds.

In order to prove the Converse, we first prove that if ${\phi }_l\in {\mathcal{ℬ}}_{0*}^q\left(U^n\right)$  for every $l=1,2,\cdots ,n.,$  then $f\circ \phi \in {\mathcal{ℬ}}_{0*}^q\left(U^n\right)$  for any $f\in {\mathcal{ℬ}}_{0*}^p\left(U^n\right).$  Without loss of generality, we prove this result when $n=2.$  For any sequence $\{z^j=(z_1^j,z_2^j\left)\right\}\subset U^n$  with $z^j\to {\partial }^{*}{U}^n$  as $j\to \infty ,$  then $$\left|z_1^j\right|\to 1,\left|z_2^j\right|\to 1.$$  Since $\left|{\phi }_1\right(z^j\left)\right|<1$  and $\left|{\phi }_2\right(z^j\left)\right|<1,$  there exists a subsequence $\left\{z^{j_s}\right\}$  in $\left\{z^j\right\}$  such that $$\left|{\phi }_1\right(z^{j_s}\left)\right|\to {\rho }_1,\left|{\phi }_2\right(z^{j_s}\left)\right|\to {\rho }_2,$$  as $s\to \infty .$  It is clear that $0\le {\rho }_1,{\rho }_2\le 1.$ 

$$\begin{array}{ccc}& & \left|\frac{\partial (f\circ \phi )}{\partial z_k}\left(z^{j_s}\right)\right|(1-|z_k^{j_s}{|}^2{)}^q\end{array}$$

$$\begin{array}{ccc}& & \le \left|\frac{\partial f}{\partial w_1}\left(\phi \right(z^{j_s}\left)\right)\right|\left|\frac{\partial {\phi }_1}{\partial z_k}\left(z^{j_s}\right)\right|(1-|z_k^{j_s}{|}^2{)}^q+\left|\frac{\partial f}{\partial w_2}\left(\phi \right(z^{j_s}\left)\right)\right|\left|\frac{\partial {\phi }_2}{\partial z_k}\left(z^{j_s}\right)\right|(1-|z_k^{j_s}{|}^2{)}^q\end{array}$$

$$\begin{array}{ccc}& & =\left|\frac{\partial f}{\partial w_1}\left(\phi \right(z^{j_s}\left)\right)\right|(1-|{\phi }_1\left(z^{j_s}\right){|}^2{)}^p\left|\frac{\partial {\phi }_1}{\partial z_k}\left(z^{j_s}\right)\right|\frac{(1-|z_k^{j_s}{|}^2{)}^q}{(1-|{\phi }_1\left(z^{j_s}\right){|}^2{)}^p}\end{array}$$

$$\begin{array}{ccc}& & +\left|\frac{\partial f}{\partial w_2}\left(\phi \right(z^{j_s}\left)\right)\right|(1-|{\phi }_2\left(z^{j_s}\right){|}^2{)}^p\left|\frac{\partial {\phi }_2}{\partial z_k}\left(z^{j_s}\right)\right|\frac{(1-|z_k^{j_s}{|}^2{)}^q}{(1-|{\phi }_2\left(z^{j_s}\right){|}^2{)}^p},\end{array}$$

(14)

$k=1,2.$  Now we prove the left of  14  $\left(\right)\to 0$  as $s\to \infty$  according to four cases.

Case 1. If ${\rho }_1<1$  and ${\rho }_2<1.$  It is clear that there exist $r_1$  and $r_2$  such that ${\rho }_1<r_1<1$  and ${\rho }_2<r_2<1,$  so as $j$  is large enough, $\left|{\phi }_1\right(z^{j_s}\left)\right|\le r_1$  and $\left|{\phi }_2\right(z^{j_s}\left)\right|\le r_2.$  By ${\phi }_1,{\phi }_2\in {\mathcal{ℬ}}_{0*}^q\left(U^n\right)$  and ( 14 ), we get

$$\begin{array}{ccc}\left|\frac{\partial (f\circ \phi )}{\partial z_k}\left(z^{j_s}\right)\right|(1-|z_k^{j_s}{|}^2{)}^q& \le & \parallel f{\parallel }_p\frac{1}{(1-r_1^2{)}^p}\left|\frac{\partial {\phi }_1}{\partial z_k}\left(z^{j_s}\right)\right|(1-|z_k^{j_s}{|}^2{)}^q\end{array}$$

$$\begin{array}{ccc}& & +\parallel f{\parallel }_p\frac{1}{(1-r_2^2{)}^p}\left|\frac{\partial {\phi }_2}{\partial z_k}\left(z^{j_s}\right)\right|(1-|z_k^{j_s}{|}^2{)}^q\end{array}$$

$$\begin{array}{ccc}& & \to 0\end{array}$$

as $s\to \infty .$  Case 2. If ${\rho }_1=1$  and ${\rho }_2=1.$  Then $\phi \left(z^{j_s}\right)\to {\partial }^{*}{U}^n,$  by ( 9 ) and $f\in {\mathcal{ℬ}}_{0*}^p\left(U^n\right)$  , ( 14 ) gives that

$$\begin{array}{ccc}& & C\left|\frac{\partial (f\circ \phi )}{\partial z_k}\left(z^{j_s}\right)\right|(1-|z_k^{j_s}{|}^2{)}^q\end{array}$$

$$\begin{array}{ccc}& & C\le \left|\frac{\partial f}{\partial w_1}\left(\phi \right(z^{j_s}\left)\right)\right|(1-|{\phi }_1\left(z^{j_s}\right){|}^2{)}^p+\left|\frac{\partial f}{\partial w_2}\left(\phi \right(z^{j_s}\left)\right)\right|(1-|{\phi }_2\left(z^{j_s}\right){|}^2{)}^p\to 0\end{array}$$

as $s\to \infty .$  Case 3. If ${\rho }_1<1$  and ${\rho }_2=1.$  Similar to Case 1, we can prove that

$$\begin{array}{ccc}& & \left|\frac{\partial f}{\partial w_1}\left(\phi \right(z^{j_s}\left)\right)\right|(1-|{\phi }_1\left(z^{j_s}\right){|}^2{)}^p\left|\frac{\partial {\phi }_1}{\partial z_k}\left(z^{j_s}\right)\right|\frac{(1-|z_k^{j_s}{|}^2{)}^q}{(1-|{\phi }_1\left(z^{j_s}\right){|}^2{)}^p}\end{array}$$

$$\begin{array}{ccc}& & \le \parallel f{\parallel }_p\frac{1}{(1-r_1^2{)}^p}\left|\frac{\partial {\phi }_1}{\partial z_k}\left(z^{j_s}\right)\right|\frac{(1-|z_k^{j_s}{|}^2{)}^q}{(1-|{\phi }_1\left(z^{j_s}\right){|}^2{)}^p}\to 0\end{array}$$

(15)

as $s\to \infty .$  On the other hand, for fixed $s,$  let $w_2^{j_s}={\phi }_2\left(z^{j_s}\right),$  then $\left|w_2^{j_s}\right|<1.$  Denote $$F\left(w_1\right)=\frac{\partial f}{\partial w_2}(w_1,w_2^{j_s}).$$  It is clear that $F\left(w_1\right)$  is holomorphic on $\left|w_1\right|<1,$  choose $R_{j_s}\to 1$  with $r_1\le R_{j_s}<1.$  $\left|{\phi }_1\right(z^{j_s}\left)\right|\le r_1,$  so $$\left|F\right({\phi }_1\left(z^{j_s}\right)\left)\right|\le {max}_{\left|w_1\right|\le r_1}\left|F\right(w_1\left)\right|\le {max}_{\left|w_1\right|\le R_{j_s}}\left|F\right(w_1\left)\right|={max}_{\left|w_1\right|=R_{j_s}}\left|F\right(w_1\left)\right|=\left|F\right(w_1^{j_s}\left)\right|,$$  where $\left|w_1^{j_s}\right|=R_{j_s}\to 1.$  This means that $\left|\frac{\partial f}{\partial w_2}\left({\phi }_1\right(z^{j_s}),{\phi }_2(z^{j_s}\left)\right)\right|\le \left|\frac{\partial f}{\partial w_2}(w_1^{j_s},w_2^{j_s})\right|.$  Since $\left|w_1^{j_s}\right|\to 1,\left|w_2^{j_s}\right|\to {\rho }_2=1$  and $f\in {\mathcal{ℬ}}_{0*}^p\left(U^n\right),$  $$\left|\frac{\partial f}{\partial w_2}(w_1^{j_s},w_2^{j_s})\right|(1-|w_2^{j_s}{|}^2{)}^p\to 0$$  as $s\to \infty ,$  so by ( 9 ),

$$\begin{array}{ccc}& & \left|\frac{\partial f}{\partial w_2}\left(\phi \right(z^{j_s}\left)\right)\right|(1-|{\phi }_2\left(z^{j_s}\right){|}^2{)}^p\left|\frac{\partial {\phi }_2}{\partial z_k}\left(z^{j_s}\right)\right|\frac{(1-|z_k^{j_s}{|}^2{)}^q}{(1-|{\phi }_2\left(z^{j_s}\right){|}^2{)}^p}\end{array}$$

$$\begin{array}{ccc}& & \le C\left|\frac{\partial f}{\partial w_2}(w_1^{j_s},w_2^{j_s})\right|(1-|w_2^{j_s}{|}^2{)}^p\to 0\end{array}$$

(16)

as $s\to \infty .$  By ( 15 ) and ( 16 ), ( 14 ) gives $$\left|\frac{\partial (f\circ \phi )}{\partial z_k}\left(z^{j_s}\right)\right|(1-|z_k^{j_s}{|}^2{)}^q\to 0,$$  as $s\to \infty .$  Case 4. If ${\rho }_1=1$  and ${\rho }_2<1.$  Similar to Case 3, we can prove $$\left|\frac{\partial (f\circ \phi )}{\partial z_k}\left(z^{j_s}\right)\right|(1-|z_k^{j_s}{|}^2{)}^q\to 0,$$  as $s\to \infty .$  Combining Case 1, Case 2, Case 3 and Case 4, we know there exists a subsequence $\left\{z^{j_s}\right\}$  in $\left\{z^j\right\}$  such that $$\left|\frac{\partial (f\circ \phi )}{\partial z_k}\left(z^{j_s}\right)\right|(1-|z_k^{j_s}{|}^2{)}^q\to 0,$$  as $s\to \infty$  for $k=1,2.$  We claim that $$\left|\frac{\partial (f\circ \phi )}{\partial z_k}\left(z^j\right)\right|(1-|z_k^j{|}^2{)}^q\to 0,$$  as $j\to \infty .$  In fact, if it fails, then there exists a subsequence $\left\{z^{j_s}\right\}$  such that

$$\begin{array}{c}\left|\frac{\partial (f\circ \phi )}{\partial z_k}\left(z^{j_s}\right)\right|(1-|z_k^{j_s}{|}^2{)}^q\to \varepsilon >0\end{array}$$

(17)

for $k=1$  or $2$  . But from the above discussion, we can find a subsequence in $\left\{z^{j_s}\right\}$  we still write $\left\{z^{j_s}\right\}$  with $$\left|\frac{\partial (f\circ \phi )}{\partial z_k}\left(z^{j_s}\right)\right|(1-|z_k^{j_s}{|}^2{)}^q\to 0,$$  it contradicts with ( 17 ).

So for any sequence $\left\{z^j\right\}\subset U^n$  with $z^j\to {\partial }^{*}{U}^n$  as $j\to \infty ,$  we have $$\left|\frac{\partial (f\circ \phi )}{\partial z_k}\left(z^j\right)\right|(1-|z_k^j{|}^2{)}^q\to 0$$  for $k=1,2.$  By ( 9 ) and Lemma 3, it is clear that $f\circ \phi \in {\mathcal{ℬ}}^q\left(U^n\right),$  so $f\circ \phi \in {\mathcal{ℬ}}_{0*}^q\left(U^n\right).$  For any $f\in {\mathcal{ℬ}}_0^p\left(U^n\right)).$  Since ${\mathcal{ℬ}}_0^p\left(U^n\right))\subset {\mathcal{ℬ}}_{0*}^p(U^n\left)\right),$  then $f\circ \phi \in {\mathcal{ℬ}}_{0*}^q\left(U^n\right).$  By closed graph theorem we known that $$C_{\phi }:{\mathcal{ℬ}}_{0*}^p\left(U^n\right)\left({\mathcal{ℬ}}_0^p\right(U^n\left)\right)\to {\mathcal{ℬ}}_{0*}^q\left(U^n\right)$$  is bounded. This ends the proof of Lemma 4.

Remark 1 For the case $C_{\phi }:{\mathcal{ℬ}}^p\left(U^n\right)\to {\mathcal{ℬ}}_{0*}^q\left(U^n\right)$  , the necessity is also true, but we can't guaranty that the sufficiency is true because we can't sure that $C_{\phi }f\in {\mathcal{ℬ}}_{0*}^q\left(U^n\right)$  for all $f\in {\mathcal{ℬ}}^p(U^n$  .

Lemma 5 Let $\phi =({\phi }_1,{\phi }_2,\cdots ,{\phi }_n)$  be a holomorphic self-map of $U^n.$  Then $$C_{\phi }:{\mathcal{ℬ}}_0^p\left(U^n\right)\to {\mathcal{ℬ}}_0^q\left(U^n\right)$$  is bounded if and only if if and only if ${\phi }^{\gamma }\in {\mathcal{ℬ}}_0^q\left(U^n\right)$  for every multi-index $\gamma$  , and ( 9 ) holds.

Proof Sufficiency. From ( 9 ) and by Theorem 1 we know that $C_{\phi }:{\mathcal{ℬ}}^p\left(U^n\right)\to {\mathcal{ℬ}}^q\left(U^n\right)$  is bounded, in particular $$\parallel C_{\phi }f{\parallel }_q\le \parallel C_{\phi }{\parallel }_{{\mathcal{ℬ}}^p\left(U^n\right)\to {\mathcal{ℬ}}^q\left(U^n\right)}\parallel f{\parallel }_p,\text{for all}f\in {\mathcal{ℬ}}_0^p\left(U^n\right).$$  The boundedness of $C_{\phi }:{\mathcal{ℬ}}_0^p\left(U^n\right)\to {\mathcal{ℬ}}_0^q\left(U^n\right)$  directly follows, if we prove $C_{\phi }f\in {\mathcal{ℬ}}_0^q\left(U^n\right)$  whenever $f\in {\mathcal{ℬ}}_0^p\left(U^n\right).$  So, let $f\in {\mathcal{ℬ}}_0^p\left(U^n\right).$  By the definition of ${\mathcal{ℬ}}_0^p\left(U^n\right)$  it follows that for every $\varepsilon >0$  there is a polynomial $p_{\varepsilon }$  such that $\parallel f-p_{\varepsilon }{\parallel }_p<\varepsilon .$  Hence

$$\begin{array}{c}\parallel C_{\phi }f-C_{\phi }{p}_{\varepsilon }{\parallel }_q\le \parallel C_{\phi }{\parallel }_{{\mathcal{ℬ}}^p\left(U^n\right)\to {\mathcal{ℬ}}^q\left(U^n\right)}\parallel f-p_{\varepsilon }{\parallel }_p<\varepsilon \parallel C_{\phi }{\parallel }_{{\mathcal{ℬ}}^p\left(U^n\right)\to {\mathcal{ℬ}}^q\left(U^n\right)}.\end{array}$$

(18)

Since ${\phi }^{\gamma }\in {\mathcal{ℬ}}_0^q\left(U^n\right)$  for every multi-index $\gamma ,$  we obtain $C_{\phi }{p}_{\varepsilon }\in {\mathcal{ℬ}}_0^q\left(U^n\right).$  From this and ( 18 ) the result follows.

If $C_{\phi }:{\mathcal{ℬ}}_0^p\left(U^n\right)\to {\mathcal{ℬ}}_0^q\left(U^n\right)$  is bounded, then ( 9 ) can be proved as in Lemma 3, since the test functions appearing there belong to ${\mathcal{ℬ}}_0^p\left(U^n\right).$  Since the polynomials $z^{\gamma }\in {\mathcal{ℬ}}_0^p\left(U^n\right)$  for every multi-index $\gamma ,$  we get $C_{\phi }{z}^{\gamma }\in {\mathcal{ℬ}}_0^q\left(U^n\right),$  as desired.

Remark 2 For the case $C_{\phi }:{\mathcal{ℬ}}^p\left(U^n\right)\left({\mathcal{ℬ}}_{0*}^p\right(U^n\left)\right)\to {\mathcal{ℬ}}_0^q\left(U^n\right)$  , similar to Remark 1, the necessity is also true, but we can't guaranty that the sufficiency is true.

Lemma 6 If $\left\{f_k\right\}$  is a bounded sequence in ${\mathcal{ℬ}}^p\left(U^n\right)$  , then there exists a subsequence $\left\{f_{k_l}\right\}$  of $\left\{f_k\right\}$  which converges uniformly on compact subsets of $U^n$  to a holomorphic function $f\in {\mathcal{ℬ}}^p\left(U^n\right)$  .

Proof Let $\left\{f_k\right\}$  be a bounded sequence in ${\mathcal{ℬ}}^p\left(U^n\right)$  with $\parallel f_k{\parallel }_p\le C.$  By Lemma 1, $\left\{f_j\right\}$  is uniformly bounded on compact subsets of $U^n$  and hence normal by Montel's theorem.

Hence we may extract subsequence $\left\{f_{j_k}\right\}$  which converges uniformly on compact subsects of $U^n$  to a holomorphic function $f$  . It follows that $\frac{\partial f_{j_k}}{\partial z_l}\to \frac{\partial f}{\partial z_l}$  for each $l\in \{1,2,\cdots ,n\}$  , so $${\sum }_{l=1}^n\left|\frac{\partial f}{\partial z_l}\right|(1-|z_l{|}^2{)}^p={lim}_{k\to \infty }{\sum }_{l=1}^n\left|\frac{\partial f_{j_k}}{\partial z_l}\right|(1-|z_l{|}^2{)}^p=\le {sup}_k\parallel f_{j_k}{\parallel }_p\le C,$$  which implies $f\in {\mathcal{ℬ}}^p\left(U^n\right)$  . The Lemma is proved.

Lemma 7 Let $\Omega$  be a domain in ${\text{C}}^n,$  $f\in H(\Omega ).$  If a compact set $K$  and its neighborhood $G$  satisfy $K\subset G\subset \subset \Omega$  and $\rho =dist(K,\partial G)>0,$  then $${sup}_{z\in K}\left|\frac{\partial f}{\partial z_j}\left(z\right)\right|\le \frac{\sqrt{n}}{\rho }{sup}_{z\in G}\left|f\right(z\left)\right|.$$ 

Proof Since $\rho =dist(K,\partial G)>0,$  for any $a\in K,$  the polydisc $$P_a=\left\{(z_1,\cdots ,z_n)\in {\text{C}}^n:|z_j-a_j|<\frac{\rho }{\sqrt{n}},j=1,\cdots ,n\right\}$$  is contained in $G.$  By Cauchy's inequality, $$\left|\frac{\partial f}{\partial z_j}\left(a\right)\right|\le \frac{\sqrt{n}}{\rho }{sup}_{z\in {\partial }^{*}{P}_a}\left|f\right(z\left)\right|\le \frac{\sqrt{n}}{\rho }{sup}_{z\in G}\left|f\right(z\left)\right|.$$  Taking the supremum for $a$  over $K$  gives the desired inequality.

3 The Proof of Theorem 1

Now we turn to the proof of Theorem 1. The lower estimate. It is clear that $\left\{m^{p-1}{z}_1^m\right\}\subset {\mathcal{ℬ}}_0^p\left(U^n\right)\subset {\mathcal{ℬ}}_{0*}\left(U^n\right)\subset \mathcal{ℬ}\left(U^n\right)$  for $m=1,2,\cdots ,$  and this sequence converges to zero uniformly on compact subsets of the unit polydisc $U^n.$ 

$$\begin{array}{c}\parallel m^{p-1}{z}_1^m{\parallel }_p={sup}_{z\in U^n}(1-|z_1{|}^2{)}^p\left|m^p{z}_1^{m-1}\right|.\end{array}$$

(19)

Let $p\left(x\right)=m^p(1-x^2{)}^p{x}^{m-1},$  then $$p^{\prime }\left(x\right)=-m^p{x}^{m-2}(1-x^2{)}^{p-1}\left[(2p+m-1)x^2-(m-1)\right],$$  so $p^{\prime }\left(x\right)\le 0$  for $x\in \left[\sqrt{\frac{m-1}{2p+m-1}},1\right],$  and $p^{\prime }\left(x\right)\ge 0$  for $x\in \left[0,\sqrt{\frac{m-1}{2p+m-1}}\right].$  That is, $p\left(x\right)$  is a decreasing function for $x\in \left[\sqrt{\frac{m-1}{2p+m-1}},1\right]$  and $p\left(x\right)$  is a increasing function for $x\in \left[0,\sqrt{\frac{m-1}{2p+m-1}}\right].$  Hence $${max}_{x\in [0,1]}p\left(x\right)=p\left(\sqrt{\frac{m-1}{2p+m-1}}\right).$$  It follows from ( 19 ) that $$\parallel m^{p-1}{z}_1^m{\parallel }_p=p\left(\sqrt{\frac{m-1}{2p+m-1}}\right)={\left(\frac{2p}{2p+m-1}\right)}^p{m}^p{\left(\frac{m-1}{2p+m-1}\right)}^{\frac{m-1}{2}}\to {\left(\frac{2p}{e}\right)}^p,$$  as $m\to \infty .$  Therefore, the sequence $\{m^{p-1}{z}_1^m{\}}_{m\ge 2}$  is bounded away from zero. Now we consider the normalized sequence $\{f_m=\frac{m^{p-1}{z}_1^m}{\parallel m^{p-1}{z}_1^m{\parallel }_p}\}$  which also tends to zero uniformly on compact subsets of $U^n.$  For each $m\ge 2,$  we define $$A_m=\{z=(z_1,\dots ,z_n)\in U^n:r_m\le |z_1|\le r_{m+1}\},$$  where $r_m=\sqrt{\frac{m-1}{2p+m-1}}.$  So

$$\begin{array}{ccc}& & {min}_{A_m}{\sum }_{l=1}^n\left\{\left|\frac{\partial f_m}{\partial z_l}\left(z\right)\right|(1-|z_l{|}^2{)}^p\right\}={min}_{A_m}\left|\frac{\partial f_m}{\partial z_1}(1-|z_1{|}^2{)}^p\right|\end{array}$$

$$\begin{array}{ccc}& & =\frac{(1-r_{m+1}^2{)}^p|m^p{r}_{m+1}^{m-1}|}{\parallel m^{p-1}{z}_1^m{\parallel }_p}=\left(\frac{2p+m-1}{2p+m}\right){\left(\frac{m(2p+m-1)}{(m-1)(2p+m)}\right)}^{\frac{m-1}{2}}=c_m.\end{array}$$

It is easy to show that $c_m$  tends to 1 as $m\to \infty$  . For the moment fix any compact operator $K:{\mathcal{ℬ}}^p\left(U^n\right){\mathcal{ℬ}}^p\left(U^n\right)\left({\mathcal{ℬ}}_0^p\right(U^n)$  or ${\mathcal{ℬ}}_{0*}^p\left(U^n\right))⟶{\mathcal{ℬ}}^q(U^n)$  $\left({\mathcal{ℬ}}_0^q\right(U^n)$  or ${\mathcal{ℬ}}_{0*}^q\left(U^n\right)).$  The uniform convergence on compact subsets of the sequence $\left\{f_m\right\}$  to zero and the compactness of $K$  imply that $\parallel K{f}_m{\parallel }_q\to 0.$  It is easy to show that if a bounded sequence that is contained in ${\mathcal{ℬ}}_{0*}^p\left(U^n\right)$  converges uniformly on compact subsets of $U^n,$  then it also converges weakly to zero in ${\mathcal{ℬ}}_{0*}^p\left(U^n\right)$  as well as in ${\mathcal{ℬ}}^p\left(U^n\right).$  Since $\parallel f_m{\parallel }_p=1$  , we have

$$\begin{array}{ccc}& & \parallel C_{\phi }-K\parallel \ge {limsup}_m\parallel (C_{\phi }-K)f_m{\parallel }_q\end{array}$$

$$\begin{array}{ccc}& & \ge {limsup}_m\left(\parallel C_{\phi }{f}_m{\parallel }_q-\parallel K{f}_m{\parallel }_q\right)={limsup}_m\parallel C_{\phi }{f}_m{\parallel }_q\end{array}$$

$$\begin{array}{ccc}& & \ge {limsup}_m{sup}_{z\in U^n}{\sum }_{k=1}^n\left\{\left|\frac{\partial (f_m\circ \phi )}{\partial z_k}\right|(1-|z_k{|}^2{)}^q\right\}\end{array}$$

$$\begin{array}{ccc}& & ={limsup}_m{sup}_{z\in U^n}{\sum }_{k=1}^n\left|\frac{\partial f_m}{\partial w_1}\left(\phi \right(z\left)\right)\right|\left|\frac{\partial {\phi }_1}{\partial z_k}\left(z\right)\right|(1-|z_k{|}^2{)}^q\end{array}$$

$$\begin{array}{ccc}& & ={limsup}_m{sup}_{z\in U^n}{\sum }_{k=1}^n\left|\frac{\partial {\phi }_1}{\partial z_k}\left(z\right)\right|\frac{(1-|z_k{|}^2{)}^q}{(1-|{\phi }_1\left(z\right){|}^2{)}^p}\left|\frac{\partial f_m}{\partial w_1}\left(\phi \right(z\left)\right)\right|(1-|{\phi }_1\left(z\right){|}^2{)}^p\end{array}$$

$$\begin{array}{ccc}& & \ge {limsup}_m{sup}_{\phi \left(z\right)\in A_m}{\sum }_{k=1}^n\left|\frac{\partial {\phi }_1}{\partial z_k}\left(z\right)\right|\frac{(1-|z_k{|}^2{)}^q}{(1-|{\phi }_1\left(z\right){|}^2{)}^p}\left|\frac{\partial f_m}{\partial w_1}\left(\phi \right(z\left)\right)\right|(1-|{\phi }_1\left(z\right){|}^2{)}^p\end{array}$$

$$\begin{array}{ccc}& & \ge {limsup}_m{sup}_{\phi \left(z\right)\in A_m}{\sum }_{k=1}^n\left|\frac{\partial {\phi }_1}{\partial z_k}\left(z\right)\right|\frac{(1-|z_k{|}^2{)}^q}{(1-|{\phi }_1\left(z\right){|}^2{)}^p}\end{array}$$

$$\begin{array}{ccc}& & \times {liminf}_m{min}_{\phi \left(z\right)\in A_m}\left|\frac{\partial f_m}{\partial w_1}\left(\phi \right(z\left)\right)\right|(1-|{\phi }_1\left(z\right){|}^2{)}^p\end{array}$$

$$\begin{array}{ccc}& & \ge {limsup}_m{sup}_{\phi \left(z\right)\in A_m}{\sum }_{k=1}^n\left|\frac{\partial {\phi }_1}{\partial z_k}\left(z\right)\right|\frac{(1-|z_k{|}^2{)}^q}{(1-|{\phi }_1\left(z\right){|}^2{)}^p}{liminf}_m{c}_m\end{array}$$

$$\begin{array}{ccc}& & \ge {limsup}_m{sup}_{\phi \left(z\right)\in A_m}{\sum }_{k=1}^n\left|\frac{\partial {\phi }_1}{\partial z_k}\left(z\right)\right|\frac{(1-|z_k{|}^2{)}^q}{(1-|{\phi }_1\left(z\right){|}^2{)}^p}.\end{array}$$

So

$$\begin{array}{ccc}\parallel C_{\phi }{\parallel }_e& =& inf\{\parallel C_{\phi }-K\parallel :K\text{is compact}\}\end{array}$$

$$\begin{array}{ccc}& \ge & {limsup}_m{sup}_{\phi \left(z\right)\in A_m}{\sum }_{k=1}^n\left|\frac{\partial {\phi }_1}{\partial z_k}\left(z\right)\right|\frac{(1-|z_k{|}^2{)}^q}{(1-|{\phi }_1\left(z\right){|}^2{)}^p}.\end{array}$$

(20)

For each $l=1,2,\cdots ,n,$  define

$$\begin{array}{c}{a}_l={lim}_{\delta \to 0}{sup}_{dist\left(\phi \right(z),\partial U^n)<\delta }{\sum }_{k=1}^n\left|\frac{\partial {\phi }_l}{\partial z_k}\left(z\right)\right|\frac{(1-|z_k{|}^2{)}^q}{(1-|{\phi }_l\left(z\right){|}^2{)}^p}.\end{array}$$

(21)

For any $\varepsilon >0,$  ( 21 ) shows that there exists a ${\delta }_0$  with $0<{\delta }_0<1,$  such that

$$\begin{array}{c}{\sum }_{k=1}^n\left|\frac{\partial {\phi }_l}{\partial z_k}\left(z\right)\right|\frac{(1-|z_k{|}^2{)}^q}{(1-|{\phi }_l\left(z\right){|}^2{)}^p}>a_l-\varepsilon ,\end{array}$$

(22)

whenever $dist\left(\phi \right(z),\partial U^n)<{\delta }_0$  and $l=1,2,\cdots ,n.$  Since $r_m\to 1$  as $m\to \infty ,$  so as $m$  is large enough, $r_m>1-{\delta }_0.$  If $\phi \left(z\right)\in A_m,$  $r_m\le \left|{\phi }_1\right(z\left)\right|\le r_{m+1},$  so $1-r_{m+1}<1-\left|{\phi }_1\right(z\left)\right|<1-r_m<{\delta }_0,$  $dist\left({\phi }_1\right(z),\partial U)<{\delta }_0.$  There exists $w_1$  with $\left|w_1\right|=1$  such that $dist\left({\phi }_1\right(z),w_1)=dist\left({\phi }_1\right(z),\partial U)<{\delta }_0.$  Let $w=(w_1,{\phi }_2(z),\dots ,{\phi }_n(z\left)\right),$  $w\in \partial U^n$  , then $$dist\left(\phi \right(z),\partial U)\le dist\left(\phi \right(z),w)=dist\left({\phi }_1\right(z),w_1)<{\delta }_0.$$  By ( 22 ), ( 20 ) implies that $$\parallel C_{\phi }{\parallel }_e\ge a_1-\varepsilon .$$  Similarly, if we choose $g_m\left(z\right)=\frac{m^{p-1}{z}_l^m}{\parallel m^{p-1}{z}_l^m\parallel }$  , we have $$\parallel C_{\phi }{\parallel }_e\ge a_l-\varepsilon ,$$  for every $l=2\cdots ,n.$  So

$$\begin{array}{ccc}\parallel C_{\phi }{\parallel }_e& \ge & \frac{1}{n}{\sum }_{l=1}^n(a_l-\varepsilon )\end{array}$$

$$\begin{array}{ccc}& =& \frac{1}{n}{\sum }_{l=1}^n({lim}_{\delta \to 0}{sup}_{dist\left(\phi \right(z),\partial U^n)<\delta }{\sum }_{k=1}^n\left|\frac{\partial {\phi }_l}{\partial z_k}\left(z\right)\right|\frac{(1-|z_k{|}^2{)}^q}{(1-|{\phi }_l\left(z\right){|}^2{)}^p}-\varepsilon )\end{array}$$

$$\begin{array}{ccc}& \ge & \frac{1}{n}{lim}_{\delta \to 0}{sup}_{dist\left(\phi \right(z),\partial U^n)<\delta }{\sum }_{k,l=1}^n\left|\frac{\partial {\phi }_l}{\partial z_k}\left(z\right)\right|\frac{(1-|z_k{|}^2{)}^q}{(1-|{\phi }_l\left(z\right){|}^2{)}^p}-\varepsilon .\end{array}$$

Let $\varepsilon \to 0,$  the low estimate follows.

The upper estimate. To obtain the upper estimate we first prove the following proposition.

Proposition 1 Let $\phi =({\phi }_1,\dots ,{\phi }_n)$  a holomorphic self-map of $U^n.$  The operators $K_m$  ( $m\ge 2$  ) as follows:

$$K_{m}f\left(z\right)=f\left(\frac{m-1}{m}z\right),$$  for $f\in H\left(U^n\right).$  Then the operators $K_m$  have the following properties:

(i) For any $f\in H\left(U^n\right),$  $K_{m}f\in {\mathcal{ℬ}}_0^p\left(U^n\right)\subset {\mathcal{ℬ}}_{0*}^p\left(U^n\right)\subset {\mathcal{ℬ}}^p\left(U^n\right).$  (ii) If $C_{\phi }:$  ${\mathcal{ℬ}}^p\left(U^n\right)$  ( ${\mathcal{ℬ}}_0^p\left(U^n\right)$  or ${\mathcal{ℬ}}_{0*}^p\left(U^n\right)$  ) $\to$  ${\mathcal{ℬ}}^q\left(U^n\right)$  ( ${\mathcal{ℬ}}_0^q\left(U^n\right)$  or ${\mathcal{ℬ}}_{0*}^q\left(U^n\right)$  ) is bounded, then $C_{\phi }{K}_{m}f\in {\mathcal{ℬ}}^q\left(U^n\right)$  ( ${\mathcal{ℬ}}_0^q\left(U^n\right)$  or ${\mathcal{ℬ}}_{0*}^q\left(U^n\right)$  ) for all $f\in H\left(U^n\right).$  (iii) For fixed $m$  , the operator $K_m$  is compact on ${\mathcal{ℬ}}^p\left(U^n\right)$  ( ${\mathcal{ℬ}}_0^p\left(U^n\right)$  or ${\mathcal{ℬ}}_{0*}^p\left(U^n\right)$  ).

(iv) If $C_{\phi }:$  ${\mathcal{ℬ}}^p\left(U^n\right)$  ( ${\mathcal{ℬ}}_0^p\left(U^n\right)$  or ${\mathcal{ℬ}}_{0*}^p\left(U^n\right)$  ) $\to$  ${\mathcal{ℬ}}^q\left(U^n\right)$  ( ${\mathcal{ℬ}}_0^q\left(U^n\right)$  or ${\mathcal{ℬ}}_{0*}^q\left(U^n\right)$  ) is bounded, then $C_{\phi }{K}_{m}f\in {\mathcal{ℬ}}^q\left(U^n\right)$  ( ${\mathcal{ℬ}}_0^q\left(U^n\right)$  or ${\mathcal{ℬ}}_{0*}^q\left(U^n\right))$  is compact.

(v) $\parallel I-K_m\parallel \le 2.$  (vi) $(I-K_m)f$  converges uniformly to zero on compact subset of $U^n$  .

Proof (i) Let $f\in H\left(U^n\right),$  $r_m=\frac{m-1}{m},(0<r_m<1)$  and $f_m\left(z\right)=K_{m}f\left(z\right)=f\left(r_{m}z\right).$  First note that

$$\begin{array}{ccc}\parallel f_m{\parallel }_p& =& \left|f\right(0\left)\right|+{sup}_{z\in U^n}{\sum }_{k=1}^n{r}_m\left|\frac{\partial f}{\partial z_k}\left(r_{m}z\right)\right|{\left(1-|z_k{|}^2\right)}^p\end{array}$$

$$\begin{array}{ccc}& \le & \left|f\right(0\left)\right|+{sup}_{z\in U^n}{\sum }_{k=1}^n\left|\frac{\partial f}{\partial z_k}\left(r_{m}z\right)\right|{\left(1-|r_m{z}_k{|}^2\right)}^p\le \parallel f{\parallel }_p.\end{array}$$

(23)

On the other hand, $f_m\in H\left(\frac{1}{r_m}{U}^n\right).$  $0<\frac{2}{1+r_m}<\frac{1}{r_m},$  $\frac{2}{1+r_m}\overline{U^n}\subset \frac{1}{r_m}{U}^n.$  which implies that for fixed $m,$  and $\varepsilon =\frac{1}{j},j=1,2,\cdots ,$  there is a polynomial $P_m^{\left(j\right)}$  such that $${sup}_{z\in \frac{2}{1+r_m}\overline{U^n}}\left|f_m\right(z)-P_m^{\left(j\right)}(z\left)\right|<(1-r_m{)}^2\frac{1}{j}.$$  Let $K=\overline{U^n},$  $G=\frac{2}{1+r_m}{U}^n,$  $\Omega =\frac{1}{r_m}{U}^n,$  then $K\subset G\subset \subset \Omega$  , and $\rho =dist(K,\partial G)=\frac{1-r_m}{1+r_m}>0$  , so $\forall w\in U^n$  , $k\in \{1,\cdots ,n\}$  , it follows from Lemma 7 that

$$\begin{array}{ccc}& & \left|\frac{\partial (f_m-P_m^{\left(j\right)})}{\partial w_k}\right(w\left)\right|\le {sup}_{w\in K}\left|\frac{\partial (f_m-P_m^{\left(j\right)})}{\partial w_k}\right(w\left)\right|\end{array}$$

$$\begin{array}{ccc}& \le & \frac{\sqrt{n}(1+r_m)}{1-r_m}{sup}_{w\in G}\left|f_m\right(w)-P_m^{\left(j\right)}(w\left)\right|\end{array}$$

$$\begin{array}{ccc}& \le & \frac{\sqrt{n}(1+r_m)}{1-r_m}(1-r_m^2)\frac{1}{j}\le 4\sqrt{n}\frac{1}{j}.\end{array}$$

Therefore $${\sum }_{k=1}^n\left|\frac{\partial (f_m-P_m^{\left(j\right)})}{\partial w_k}\right(w\left)\right|(1-|w_k{|}^p{)}^p\le 4n\sqrt{n}\frac{1}{j}\to 0$$  as $j\to \infty .$  that is, $$\left|\right|f_m-P_m^{\left(j\right)}|{|}_{{\mathcal{ℬ}}^p}=|f_m\left(0\right)-P_m^{\left(j\right)}\left(0\right)|+{sup}_{w\in U^n}{\sum }_{k=1}^n|\frac{\partial (f_m-P_m^{\left(j\right)})}{\partial w_k}\left(w\right)\left|\right(1-|w_k{|}^p{)}^p\to 0.$$  $P_m^{\left(j\right)}\left(w\right)\in {\mathcal{ℬ}}_0^p\left(U^n\right)$  implies that $f_m\in {\mathcal{ℬ}}_0^p\left(U^n\right)$  .

(ii) By (i), as desired.

(iii) For any sequence $\left\{f_j\right\}\subset {\mathcal{ℬ}}^p\left(U^n\right)$  ( ${\mathcal{ℬ}}_0^p\left(U^n\right)$  or ${\mathcal{ℬ}}_{0*}^p\left(U^n\right)$  ) with $\parallel f_j{\parallel }_p\le M,$  by (i), $\left\{K_m{f}_j\right\}\in {\mathcal{ℬ}}_0^p\left(U^n\right).$  By Lemma 6, there is a subsequence $\left\{f_{j_s}\right\}$  of $\left\{f_j\right\}$  which converges uniformly on compact subsets of $U^n$  to a holomorphic function $f\in {\mathcal{ℬ}}^p\left(U^n\right)$  and $\parallel f{\parallel }_p\le M.$  $\left\{\frac{\partial f_{j_s}}{\partial z_i}\right\},i=1,2,\cdots ,n,$  also converges uniformly on compact subsets of $U^n$  to the holomorphic function $\frac{\partial f}{\partial z_i}.$  So as $s$  is large enough, for any $w\in E=\{\frac{m-1}{m}z:z\in \overline{U^n}\}\subset U^n$ 

$$\begin{array}{c}\left|\frac{\partial (f_{j_s}-f)}{\partial w_l}\left(w\right)\right|<\varepsilon ,\end{array}$$

(24)

for every $l=1,2,\cdots ,n.$  So

$$\begin{array}{ccc}& & {\parallel K_m{f}_{j_s}-K_{m}f\parallel }_p={\Vert f_{j_s}\left(\frac{m-1}{m}z\right)-f\left(\frac{m-1}{m}z\right)\Vert }_p\end{array}$$

$$\begin{array}{ccc}& & ={sup}_{z\in U^n}{\sum }_{k=1}^n\left\{\left|\frac{\partial \left[(f_{j_s}-f)\left(\frac{m-1}{m}z\right)\right]}{\partial z_k}\right|(1-|z_k{|}^2{)}^p\right\}+\left|f_{j_s}\right(0)-f(0\left)\right|\end{array}$$

$$\begin{array}{ccc}& & \le {sup}_{z\in U^n}{\sum }_{k=1}^n{\sum }_{l=1}^n\left|\frac{\partial (f_{j_s}-f)}{\partial w_l}\left(\frac{m-1}{m}z\right)\right|\frac{m-1}{m}+\left|f_{j_s}\right(0)-f(0\left)\right|\end{array}$$

$$\begin{array}{ccc}& & \le n{sup}_{w\in E_1}\frac{m-1}{m}{\sum }_{l=1}^n\left|\frac{\partial (f_{j_s}-f)}{\partial w_l}\left(w\right)\right|+\left|f_{j_s}\right(0)-f(0\left)\right|\to 0,\end{array}$$

(25)

as $s\to \infty .$  This shows that $\left\{K_m{f}_{j_s}\right\}$  converges to $g=K_{m}f\in {\mathcal{ℬ}}_0^p\left(U^n\right)\subset {\mathcal{ℬ}}_{0*}^p\left(U^n\right)\subset {\mathcal{ℬ}}^p\left(U^n\right).$  So $K_m$  is compact on ${\mathcal{ℬ}}^p\left(U^n\right)$  ( ${\mathcal{ℬ}}_0^p\left(U^n\right)$  or ${\mathcal{ℬ}}_{0*}^p\left(U^n\right)).$  (iv) By (i) and (iii), the result is obvious.

(v) In fact, for any $f\in {\mathcal{ℬ}}^p\left(U^n\right)\left({\mathcal{ℬ}}_0^p\right(U^n)$  or ${\mathcal{ℬ}}_{0*}^p\left(U^n\right))$  , note that $(I-K_m)f\left(0\right)=0$  , so

$$\begin{array}{ccc}& & \parallel (I-K_m)f{\parallel }_p={sup}_{z\in U^n}{\sum }_{k=1}^n\left|\frac{\partial (I-K_m)f}{\partial z_k}\left(z\right)\right|(1-|z_k{|}^2)\end{array}$$

$$\begin{array}{ccc}& & =n{sup}_{z\in U^n}{max}_{1\le k\le n}\left|\frac{\partial f}{\partial z_k}\left(z\right)-(1-\frac{1}{m})\frac{\partial f}{\partial z_k}\left(\right(1-\frac{1}{m}\left)z\right)\right|(1-|z_k{|}^2{)}^p\end{array}$$

$$\begin{array}{ccc}& & \le {sup}_{z\in U^n}{\sum }_{k=1}^n\left|\frac{\partial f}{\partial z_k}\left(z\right)\right|(1-|z_k{|}^2{)}^p\end{array}$$

$$\begin{array}{ccc}& & +(1-\frac{1}{m}){sup}_{z\in U^n}{\sum }_{k=1}^n\left|\frac{\partial f}{\partial z_k}\left(\right(1-\frac{1}{m}\left)z\right)\right|(1-|(1-\frac{1}{m})z_k{|}^2{)}^p\end{array}$$

$$\begin{array}{ccc}& & \le \parallel f{\parallel }_p+\parallel f{\parallel }_p=2\parallel f{\parallel }_p,\end{array}$$

so $\parallel I-K_m\parallel \le 2.$  (vi) For any compact subset $E\subset U^n$  , $\exists r,$  $0<r<1$  such that $E\subset r{U}^n\subset \subset U^n$  . For $\forall z\in E$  ,

$$\begin{array}{ccc}\left|\right(I-K_m\left)f\right(z\left)\right|& =& \left|f\right(z)-f_m(z\left)\right|=\left|f\right(z)-f(r_{m}z\left)\right|\end{array}$$

$$\begin{array}{ccc}& =& \left|{\int }_{r_m}^1\frac{d}{dt}\left(f\right(tz\left)\right)dt\right|=\left|{\int }_{r_m}^1{\sum }_{k=1}^n\frac{\partial f}{\partial w_k}\left(tz\right)\cdot z_{k}dt\right|\end{array}$$

$$\begin{array}{ccc}& \le & {\sum }_{k=1}^n{\int }_{r_m}^1\left|\frac{\partial f}{\partial w_k}\left(tz\right)\right|dt.\end{array}$$

$t\in [r_m,1]$  , $\forall z\in U^n,\left|t{z}_k\right|=t\left|z_k\right|<\left|z_k\right|<r,$  so $\frac{\partial f}{\partial w_k}\left(w\right)$  is bounded in $r\overline{U^n}$  , i.e., $\forall z\in E,$  $\left|\frac{\partial f}{\partial w_k}\left(tz\right)\right|\le M$  . So $$\left|\right(I-K_m\left)f\right(z\left)\right|\le nM(1-r_m)\to 0$$  as $m\to \infty$  , the results follows.

Now return to the upper estimate. For the convenience, we denote $\parallel f\parallel =\parallel f{\parallel }_p.$ 

$$\begin{array}{ccc}& & \parallel C_{\phi }{\parallel }_e\le \parallel C_{\phi }-C_{\phi }{K}_m\parallel =\parallel C_{\phi }(I-K_m)\parallel ={sup}_{\parallel f\parallel =1}\parallel C_{\phi }(I-K_m)f{\parallel }_q\end{array}$$

$$\begin{array}{ccc}& & ={sup}_{\parallel f\parallel =1}\left({sup}_{z\in U^n}{\sum }_{k=1}^n\left\{\left|\frac{\partial (I-K_m)(f\circ \phi )}{\partial z_k}\right|(1-|z_k{|}^2{)}^q\right\}+\left|(I-K_m)f\left(\phi \right(0\left)\right)\right|\right)\end{array}$$

$$\begin{array}{ccc}& & \le {sup}_{\parallel f\parallel =1}{sup}_{z\in U^n}{\sum }_{k=1}^n{\sum }_{l=1}^n\left|\frac{\partial (I-K_m)f}{\partial w_l}\left(\phi \right(z\left)\right)\right|\left|\frac{\partial {\phi }_l}{\partial z_k}\left(z\right)\right|(1-|z_k{|}^2{)}^q\end{array}$$

$$\begin{array}{ccc}& & +{sup}_{\parallel f\parallel =1}\left|f\left(\phi \right(0\left)\right)-f\left(\frac{m-1}{m}\phi \right(0\left)\right)\right|\end{array}$$

$$\begin{array}{ccc}& & \le {sup}_{\parallel f\parallel =1}{sup}_{z\in U^n}{\sum }_{k,l=1}^n\left|\frac{\partial {\phi }_l}{\partial z_k}\left(z\right)\right|\frac{(1-|z_k{|}^2{)}^q}{(1-|{\phi }_l\left(z\right){|}^2{)}^p}\left|\frac{\partial (I-K_m)f}{\partial w_l}\left(\phi \right(z\left)\right)\right|(1-|{\phi }_l\left(z\right){|}^2{)}^p\end{array}$$

$$\begin{array}{ccc}& & +{sup}_{\parallel f\parallel =1}\left|f\left(\phi \right(0\left)\right)-f\left(\frac{m-1}{m}\phi \right(0\left)\right)\right|.\end{array}$$

(26)

Denote $G_1=\{z\in U^n:dist(\phi \left(z\right),\partial U^n)<\delta \},$  $G_2=\{z\in U^n:dist(\phi \left(z\right),\partial U^n)\ge \delta \},$  $G=\{w\in U^n:dist(w,\partial U^n)\ge \delta \}$  , where $G$  is a compact subset of ${\text{C}}^n.$  Then by Lemma 3, Lemma 4 and Lemma 5, condition (9) holds, so

$$\begin{array}{ccc}\parallel C_{\phi }{\parallel }_e& \le & {sup}_{\parallel f\parallel =1}{sup}_{z\in G_1}{\sum }_{k,l=1}^n\left|\frac{\partial {\phi }_l}{\partial z_k}\left(z\right)\right|\frac{(1-|z_k{|}^2{)}^q}{(1-|{\phi }_l\left(z\right){|}^2{)}^p}\left|\frac{\partial (I-K_m)f}{\partial w_l}\left(\phi \right(z\left)\right)\right|(1-|{\phi }_l\left(z\right){|}^2{)}^q\end{array}$$

$$\begin{array}{ccc}& & +C{sup}_{\parallel f\parallel =1}{sup}_{z\in G_2}{\sum }_{l=1}^n(1-|{\phi }_l\left(z\right){|}^2{)}^p\left|\frac{\partial (I-K_m)f}{\partial w_l}\left(\phi \right(z\left)\right)\right|\end{array}$$

$$\begin{array}{ccc}& & +{sup}_{\parallel f\parallel =1}\left|f\left(\phi \right(0\left)\right)-f\left(\frac{m-1}{m}\phi \right(0\left)\right)\right|\end{array}$$

$$\begin{array}{ccc}& \le & \parallel I-K_m\parallel {sup}_{z\in G_1}{\sum }_{k,l=1}^n\left|\frac{\partial {\phi }_l}{\partial z_k}\left(z\right)\right|\frac{(1-|z_k{|}^2{)}^q}{(1-|{\phi }_l\left(z\right){|}^2{)}^p}\end{array}$$

$$\begin{array}{ccc}& & +C{sup}_{\parallel f\parallel =1}{sup}_{z\in G_2}{\sum }_{l=1}^n(1-|{\phi }_l\left(z\right){|}^2{)}^p\left|\frac{\partial (I-K_m)f}{\partial w_l}\left(\phi \right(z\left)\right)\right|\end{array}$$

$$\begin{array}{ccc}& & +{sup}_{\parallel f\parallel =1}\left|f\left(\phi \right(0\left)\right)-f\left(\frac{m-1}{m}\phi \right(0\left)\right)\right|\end{array}$$

$$\begin{array}{ccc}& \le & 2{sup}_{z\in G_1}{\sum }_{k,l=1}^n\left|\frac{\partial {\phi }_l}{\partial z_k}\left(z\right)\right|\frac{(1-|z_k{|}^2{)}^q}{(1-|{\phi }_l\left(z\right){|}^2{)}^p}\end{array}$$

$$\begin{array}{ccc}& & +C{sup}_{\parallel f\parallel =1}{sup}_{z\in G_2}{\sum }_{l=1}^n(1-|{\phi }_l\left(z\right){|}^2{)}^p\left|\frac{\partial (I-K_m)f}{\partial w_l}\left(\phi \right(z\left)\right)\right|\end{array}$$

$$\begin{array}{ccc}& & +{sup}_{\parallel f\parallel =1}\left|f\left(\phi \right(0\left)\right)-f\left(\frac{m-1}{m}\phi \right(0\left)\right)\right|.\end{array}$$

(27)

Denote the second term and third term of the right hand side of ( 27 ) by $I_1$  and $I_2$  .

Then Theorem 1 is proved if we can prove $${lim}_{m\to \infty }{I}_1=0\text{and}{lim}_{m\to \infty }{I}_2=0.$$  To do this, let $z\in G_2$  and $w=\phi \left(z\right),$  then $w\in G$ 

$$\begin{array}{ccc}{I}_1& \le & C{sup}_{\parallel f\parallel =1}{sup}_{w\in G}{\sum }_{l=1}^n(1-|w_l{|}^2{)}^p\left|\frac{\partial f}{\partial w_l}\left(w\right)-(1-\frac{1}{m})\frac{\partial f}{\partial w_l}\left(\right(1-\frac{1}{m}\left)w\right)\right|\end{array}$$

$$\begin{array}{ccc}& \le & C{sup}_{\parallel f\parallel =1}{sup}_{w\in G}{\sum }_{l=1}^n(1-|w_l{|}^2{)}^p\left|\frac{\partial f}{\partial w_l}\left(w\right)-\frac{\partial f}{\partial w_l}\left(\right(1-\frac{1}{m}\left)w\right)\right|\end{array}$$

$$\begin{array}{ccc}& & +\frac{C}{m}{sup}_{\parallel f\parallel =1}{sup}_{w\in G}{\sum }_{l=1}^n(1-|w_l{|}^2{)}^p\left|\frac{\partial f}{\partial w_l}\left(\right(1-\frac{1}{m}\left)w\right)\right|\end{array}$$

$$\begin{array}{ccc}& \le & C{sup}_{\parallel f\parallel =1}{sup}_{w\in G}{\sum }_{l=1}^n(1-|w_l{|}^2{)}^p\left|\frac{\partial f}{\partial w_l}\left(w\right)-\frac{\partial f}{\partial w_l}\left(\right(1-\frac{1}{m}\left)w\right)\right|+\frac{C}{m}.\end{array}$$

(28)

Let $w=(w_1,w_2,\cdots ,w_{n-1},w_n),$  for $m$  large enough, we have

$$\begin{array}{ccc}& & \left|\frac{\partial f}{\partial w_l}\left(w\right)-\frac{\partial f}{\partial w_l}\left(\right(1-\frac{1}{m}\left)w\right)\right|\end{array}$$

$$\begin{array}{ccc}& & \le {\sum }_{j=1}^n|\frac{\partial f}{\partial w_l}\left((1-\frac{1}{m})w_1,\cdots ,(1-\frac{1}{m})w_{j-1},w_j,\cdots ,w_n\right)\end{array}$$

$$\begin{array}{ccc}& & -\frac{\partial f}{\partial w_l}\left((1-\frac{1}{m})w_1,\cdots ,(1-\frac{1}{m})w_j,w_{j+1},\cdots ,w_n\right)|\end{array}$$

$$\begin{array}{ccc}& & ={\sum }_{j=1}^n\left|{\int }_{(1-\frac{1}{m})w_j}^{w_j}\frac{{\partial }^{2}f}{\partial w_l\partial w_j}\left((1-\frac{1}{m})w_1,\cdots ,(1-\frac{1}{m})w_{j-1},\zeta ,w_{j+1},\cdots ,w_n\right)d\zeta \right|\end{array}$$

$$\begin{array}{ccc}& & \le \frac{1}{m}{\sum }_{j=1}^n{sup}_{w\in G}\left|\frac{{\partial }^{2}f}{\partial w_l\partial w_j}\left(w\right)\right|.\end{array}$$

(29)

Denote $G_3=\left\{w\in U^n:dist(w,\partial U^n)>\frac{\delta }{2}\right\},$  then $G\subset G_3\subset \subset U^n.$  Since $dist(G,\partial G_3)=\frac{\delta }{2},$  then by Lemma 7, ( 29 ) gives

$$\begin{array}{c}\left|\frac{\partial f}{\partial w_l}\left(w\right)-\frac{\partial f}{\partial w_l}\left(\right(1-\frac{1}{m}\left)w\right)\right|\le \frac{2n\sqrt{n}}{m\delta }{max}_{z\in G_3}\left|\frac{\partial f}{\partial w_l}\left(w\right)\right|.\end{array}$$

(30)

On the other hand, on the unit ball of ${\mathcal{ℬ}}^p\left(U^n\right)$  , we have $${sup}_{z\in G_3}(1-|w_l{|}^2{)}^p\left|\frac{\partial f}{\partial w_l}\left(w\right)\right|={sup}_{dist(w,\partial U^n)>\frac{\delta }{2}}(1-|w_l{|}^2{)}^p\left|\frac{\partial f}{\partial w_l}\left(w\right)\right|\le \parallel f{\parallel }_p=1,$$  namely

$$\begin{array}{c}{sup}_{z\in G_3}\left|\frac{\partial f}{\partial w_l}\left(w\right)\right|\le \frac{1}{1-{\left(\frac{\delta }{2}\right)}^2}=\frac{4}{4-{\delta }^2}.\end{array}$$

(31)

Combining ( 28 ), ( 30 ) and ( 31 ), imply $$I_1\le \frac{2n\sqrt{n}C}{m\delta }\frac{4}{4-{\delta }^2}+\frac{C}{m}$$  and ${lim}_{m\to \infty }{I}_1=0.$  Now we can prove ${lim}_{m\to \infty }{I}_2=0$  . In fact,

$$\begin{array}{ccc}& & f\left(\phi \right(0\left)\right)-f\left(\frac{m-1}{m}\phi \right(0\left)\right)={\int }_{\frac{m-1}{m}}^1\frac{df\left(t\phi \right(0\left)\right)}{dt}dt={\sum }_{l=1}^n{\int }_{\frac{m-1}{m}}^1{\phi }_l\left(0\right)\frac{\partial f}{\partial {\zeta }_l}\left(t\phi \right(0\left)\right)dt.\end{array}$$

By Lemma 1, it follows that for any compact subset $K\subset U^n$  , $\left|f\right(z\left)\right|\le C_K\parallel f{\parallel }_p=C_K.$  Let $K=\{z\in U^n:|z_i|\le |{\phi }_i\left(0\right)\left|\right\},$  So

$$\begin{array}{ccc}& & \left|f\right(\phi \left(0\right))-f(\frac{m-1}{m}\phi \left(0\right)\left)\right|\le {\sum }_{l=1}^n\left|{\phi }_l\right(0\left)\right|{\int }_{\frac{m-1}{m}}^1{C}_{K}dt\le n{C}_K(1-\frac{m-1}{m})=\frac{n{C}_K}{m},\end{array}$$

so $I_2\le \frac{n{C}_K}{m}\to 0.$  Thus let first $m\to \infty ,$  then $\delta \to 0$  in ( 27 ), we get the upper estimate of $\parallel C_{\phi }{\parallel }_e$  :

$$\parallel C_{\phi }{\parallel }_e\le 2{lim}_{\delta \to 0}{sup}_{dist\left(\phi \right(z),\partial U^n)<\delta }{\sum }_{k,l=1}^n\left|\frac{\partial {\phi }_l}{\partial z_k}\left(z\right)\right|\frac{(1-|z_k{|}^2{)}^q}{(1-|{\phi }_l\left(z\right){|}^2{)}^p}.$$  Now the proof of Theorem 1 is finished.

References

1. R. Timoney, Bloch function in several complex variables, I, Bull. London Math. Soc., 1980, 12(37): 241-267.
2. R. Timoney, Bloch function in several complex variables, II, J. Reine Angew. Math., 1980, 319: 1-22.
3. K. Madigan and A. Matheson, Compact composition operators on the Bloch space. Trans. Amer. Math. Soc.,1995, 347 (7) : 2679-2687.
4. Z. H. Zhou and J. H. Shi, Compact composition operators on the Bloch spaces in polydiscs, Science in China (Series A), 2001, 44 (3):286-291.
5. Z. H. Zhou and J. H. Shi, Composition operators on the Bloch space in polydiscs, Complex Variables, 2001, 46(1): 73-88.
6. Z. H. Zhou and J. H. Shi, Compactness of composition operators on the Bloch space in classical bounded symmetric domains. The Michigan Mathematical Journal, 2002, 50(2):381-405.
7. Z. H. Zhou, Composition Operators on the Lipschitz Spaces in Polydiscs. Science in China $A$  , 2003, 46(1):33-38.
8. J. H. Shapiro, The essential norm of a composition operator, Annals of Math., 1987, 125: 375-404.
9. J. H. Shapiro, Composition operators and Classical Function Theory, Spring-Verlag, 1993.
10. Alsonso Montes-Rodriguez, The essential normal of a composition operator on the Bloch space. Pacific Journal of Mathematics, 1999, 188(2): 339-351.

^$1$  Supported in part by National Natural Science Foundation of China (Grand Nos. 10371091, 10001030), and LiuHui Center for Applied Mathematics, Nankai University & Tianjin University