Coefficient estimates and Bohr phenomenon for pluriharmonic mappings in the polydisc

Let f=h+\overline g\in \mathscr{P}_{\mathcal{H}_n^0}(\alpha). Then

for all z=(z_1,z_2,\ldots,z_n)\in \mathbb{P} \Delta(0_n;1_n). It follows from (?) that

for all z\in \mathbb{P} \Delta(0_n;1_n), where

are homogeneous polynomials of degree m\geq 2 in z\in \mathbb{P} \Delta(0_n;1_n). We know that the number of terms in \sum\limits_{|\beta|=m} is \binom{|\beta|+n-1}{n-1}. A simple computation using (Proof) shows that

Using (Proof) together with (Proof), we get

By applying Cauchy’s integral formula to \frac{\partial g(z)}{\partial z_k}, it follows from (Proof) that

where 0<r_j<1 for j=1,2,\ldots,n. We set z=\left(re^{\iota\theta_1},re^{\iota\theta_2},\ldots,re^{\iota\theta_n}\right), where 0\leq \theta_j\leq 2\pi, j=1,2,\ldots,n. Now from (Proof), we have

and so

for all k=1,2,\ldots,n. Consequently from (Proof), we obtain

However, we observe that

Using (Proof) together with (Proof), we get

Moreover, we know that

For any multi-index \nu=(\nu_1,\nu_2,\ldots,\nu_n), we have

Using (Proof) together with (Proof), we deduce that

and so we get

Consequently

In view of (Proof) and (Proof) and using (Proof), we obtain

Letting r\to 1^-, we obtain

if m\geq 2.

To show that the bound is sharp, we consider the following function

where b_{\beta}=\frac{n(1-\alpha)}{m} for all multi-index \beta=(\beta_1,\beta_2,\ldots,\beta_n) such that |\beta|=m. A direct computation shows that f_1\in \mathscr{P}_{\mathcal{H}_n^0}(\alpha), and

Let f=h+\overline g\in \mathscr{P}_{\mathcal{H}_n^0}(\alpha). Note that

for all \varepsilon\in\mathbb{C} such that |\varepsilon|=1. Therefore from (Proof), we deduce that

for all z=(z_1,z_2,\ldots,z_n)\in \mathbb{P} \Delta(0_n;1_n). Let F_{\varepsilon}(z)=h(z)+\varepsilon g(z). Clearly

for all z\in \mathbb{P} \Delta(0_n;1_n), where P_m(z) and Q_m(z) are defined in (Proof). Note that

where c_{\beta}=a_{\beta}+\varepsilon b_{\beta}. Consequently

for all z=(z_1,z_2,\ldots,z_n)\in \mathbb{P} \Delta(0_n;1_n). Then from (Proof), we see that

Therefore, there exists a function

such that

for all z\in \mathbb{P}\Delta(0_n;1_n), where P(z) is holomorphic in \mathbb{P}\Delta(0_n;1_n) and \Re(P(z))>0 in \mathbb{P}\Delta(0_n;1_n).

Let \omega_l=e^{\frac{2\pi \iota}{k}l}, where k\geq 1 is an integer. We see that \sum_{l=1}^k \omega_l^s=0 if s\leq k-1. Let

The function \tilde g is clearly holomorphic in \mathbb{P}\Delta(0_n;1_n) such that \Re\{\tilde g(z)\} > 0 throughout the domain. Furthermore, we see that \tilde g(0) =1 and \tilde g has the series expansion

for all z\in \mathbb{P}\Delta(0_n;1_n). Let

for all z\in \mathbb{P}\Delta(0_n;1_n). One can easily verify that |\phi(z)|<1 for all z\in \mathbb{P}\Delta(0_n;1_n) and k is the zero-multiplicity of \phi at 0. We expand \phi(z) to a Taylor series with multi-index \beta,

By Lemma 2.1, we have |Q_{\beta}|\leq 1

for multi-index \beta. Now from (Proof) and (Proof), we deduce that

It then follows that

holds for all z\in \mathbb{P}\Delta(0_n;1_n). Comparing the homogeneous terms of degree k in (Proof), we obtain

Since |Q_{\beta}|\leq 1 for multi-index \beta, from (Proof), we conclude that

where \beta=(\beta_1,\beta_2,\ldots,\beta_n) such that |\beta|=k.

On the other hand, from (Proof) and (Proof), we obtain

where \beta=(\beta_1,\beta_2,\ldots,\beta_n) such that |\beta|=m\geq 2. Putting z_1=z_2=\ldots=z_n=z on both sides of (Proof) and then comparing coefficients, we get

and so from (Proof), we get

for all \beta=(\beta_1,\beta_2,\ldots,\beta_n) such that |\beta|=m\geq 2 and for all \varepsilon such that |\varepsilon|=1.

Suppose A_{\beta}=\sum\limits_{|\beta|=m}a_\beta\neq 0 and B_{\beta}=\sum\limits_{|\beta|=m}b_\beta\neq 0. If we choose

then |\varepsilon|=1 and so from (Proof), we have

for \beta=(\beta_1,\beta_2,\ldots,\beta_n) such that |\beta|=m\geq 2. If either A_{\beta}=0 or B_{\beta}=0, then (Proof) also holds. Therefore, we have

for \beta=(\beta_1,\beta_2,\ldots,\beta_n) such that |\beta|=m\geq 2.

On the other hand, we have

for \beta=(\beta_1,\beta_2,\ldots,\beta_n) such that |\beta|=m\geq 2. Consequently

for \beta=(\beta_1,\beta_2,\ldots,\beta_n) such that |\beta|=m\geq 2.

It is clear from (Proof) that

where \beta=(\beta_1,\beta_2,\ldots,\beta_n) such that |\beta|=m\geq 2.

To show all the inequalities are sharp, we consider the following function

where

for all multi-index \beta=(\beta_1,\beta_2,\ldots,\beta_n) such that |\beta|=m. A direct computation shows that f_2\in \mathscr{P}_{\mathcal{H}_n^0}(\alpha), and

Let f=h+\overline g\in \mathscr{P}_{\mathcal{H}_n^0}(\alpha). Let F_{\varepsilon}=h+\varepsilon g. Then from (Proof), we see that

for all z=(z_1,z_2,\ldots,z_n)\in \mathbb{P} \Delta(0_n;1_n). Then there exists a holomorphic function \omega(z) with \omega(0)=0 and |\omega(z)|<1 for all z=(z_1,z_2,\ldots,z_n)\in \mathbb{P} \Delta(0_n;1_n) such that

or equivalently

for all z=(z_1,z_2,\ldots,z_n)\in \mathbb{P} \Delta(0_n;1_n). Also by Lemma 6.1.28 [Graham-Kohr], we have

for all z=(z_1,z_2,\ldots,z_n)\in \mathbb{P} \Delta(0_n;1_n).

Let z\in \mathbb{P}\Delta(0_n;1_n) be fixed in such a way that z\neq 0. Obviously \|z\|_{\infty}<1.

We define

and

where h(z) and g(z) are defined by (Proof) and t\in\mathbb{C} such that |t|<1. For a fixed z\in \mathbb{P}\Delta(0_n;1_n), we can say that \tilde g(t) and \tilde h(t) are analytic in |t|<1. Obviously for a fixed z\in \mathbb{P}\Delta(0_n;1_n), the following function

is analytic in |t|<1, where \tilde h(t) and \tilde g(t) are defined by (Proof) and (Proof). Also from (Proof) and (Proof), we see that

We define

which is analytic in |t|<1. Consequently from (Proof), we have

In terms of one variable t, we deduce from (Proof) that

Note that

Now using (Proof) and (Proof) into (Proof), we obtain

and so from (Proof), we get

which also holds for z=0.

Again using (Proof) and (Proof) into (Proof), we obtain

and so from (Proof), we get

which also holds for z=0.

Since |\sum_{j=1}^nz_j|\leq n\|z\|_{\infty}, from (Proof) and (Proof), we get

Finally since \varepsilon\; (|\varepsilon|=1) is arbitrary and F_{\varepsilon}(z)=h(z)+\varepsilon g(z), for each 0\leq \alpha<1, we have

Thus the desired inequalities are established.

To show the left inequality is sharp, we consider the function f_3(z) defined by

where

for \beta=(\beta_1,\ldots,\beta_n) such that |\beta|=m\geq 2. It is easy to verify that f_3\in \mathscr{P}_{\mathcal{H}_n^0}(\alpha).

Now for the point z=(r,r,\ldots,r), where 0<r<1, we find that

Therefore the left inequality is sharp.

Next to show the right inequality is sharp, we consider the function f_4(z) defined by

where

for \beta=(\beta_1,\ldots,\beta_n) such that |\beta|=m\geq 2. It is easy to verify that f_4\in \mathscr{P}_{\mathcal{H}_n^0}(\alpha).

For the point z=(r,r,\ldots,r), where r<1, we have

which shows that the right inequality is sharp.

Let f\in \mathscr{P}_{\mathcal{H}_n^0}(\alpha). Then by Theorem 3.1, we have

for all z\in \mathbb{P}\Delta(0;1/n). By taking \liminf as (|z_1|,|z_2|,\ldots,|z_n|)\to (1/n,1/n,\ldots,1/n) on both sides of (Proof), we obtain

The Euclidean distance between f(0) and the boundary of f(\mathbb{P}\Delta(0;1/n)) is given by

Since f(0)=0, from (Proof) and (Proof), we get

Let H_1:[0,1)\to \mathbb{R} be defined by

Note that

for all \alpha\in [0,1) and. On the other hand, we see that H_1(n{\bf{r}})\to +\infty as n{\bf{r}}\to 1 and

for n{\bf{r}}\in (0,1).

This shows that H_1(n{\bf{r}}) is strictly increasing on (0,1).

Since H_1(0)<0 and H_1(n{\bf{r}})\to +\infty as n{\bf{r}}\to 1, the monotonicity of H_1(n{\bf{r}}) implies that H_1(n{\bf{r}}) has exactly one zero in (0,1). Let n{\bf{r}}_f be the unique root of H_1(n{\bf{r}}) in (0,1). Then H_1(n{\bf{r}}_f)=0 implies that

For 0\leq n{\bf{r}}\leq n{\bf{r}}_f, we deduce from (Proof) that

Let z\in \mathbb{P}\Delta(0;1/n) such that r=(|z_1|,|z_2|,\ldots,|z_n|)=(r_1,r_2,\ldots,r_n) and {\bf{r}}=\|z\|_{\infty}. Using Theorem 2.2 and the inequalities (Proof) and (Proof), for 0\le n\|z\|_{\infty}=n{\bf{r}}\leq n{\bf{r}}_f, we deduce that

To show that the constant {\bf{r}}_f is the best possible, we consider the function f_5(z) defined by

for \beta=(\beta_1,\ldots,\beta_n) such that |\beta|=m\geq 2.

Let z=(r,r,\ldots,r), where nr<1. Now using (Proof), we deduce that

and hence the radius {\bf{r}}_f is the best possible.

Let f=h+\overline g\in \mathscr{P}_{\mathcal{H}_n^0}(\alpha), where h and g are given by (?). Note that

Also we see that

and so

On the other hand, from (Proof) and (Proof), we deduce that

Using Theorem 2.1 to (Proof), we obtain

Therefore S_{1,q}(f)\in \mathscr{P}_{\mathcal{H}_n^0}(\alpha).