RETRACTED ARTICLE: An augmented Riesz decomposition method for sharp estimates of certain boundary value problem

Abstract

In this paper, by using an augmented Riesz decomposition method, we obtain sharp estimates of harmonic functions with certain boundary integral condition, which provide explicit lower bounds of functions harmonic in a cone. The results given here can be used as tools in the study of integral equations.

Introduction

Let ${\mathbf{R}}^{n}$ be the n-dimensional Euclidean space, where $n\geq 2$. Let $V=(X,y)$ be a point in ${\mathbf{R}}^{n}$, where $X=(x_{1},x_{2},\ldots,x_{n-1})$. Let E be a set in ${\mathbf{R}}^{n}$, the boundary and the closure of it are denoted by ∂E and E̅, respectively.

For $P=(X, y)\in{\mathbf{R}}^{n}$, it can be re-expressed in spherical coordinates $(l,\Lambda)$, $\Lambda=(\theta_{1}, \theta_{2}, \ldots ,\theta_{n})$ via the following transforms:

$$x_{1}=l\prod^{n-1}_{j=1}\sin \theta_{j} \quad (n\geq2), \qquad y=l\cos\theta_{1} $$

and, if $n\geq3$,

$$x_{n-k+1}=l\cos\theta_{k}\prod^{k-1}_{j=1} \sin\theta_{j}\quad (2\leq k\leq n-1), $$

where $0\leq l<+\infty$, $0\leq\theta_{j}\leq\pi$ ($1\leq j\leq n-2$; $n\geq3$), and $-\frac{\pi}{2}\leq\theta_{n-1}\leq\frac{3\pi}{2}$ ($n\geq2$).

The unit sphere in ${\mathbf{R}}^{n}$ is denoted by ${\mathbf{S}}^{n-1}$. Let $\Gamma\subset{\mathbf{S}}^{n-1}$. A point $(1,\Lambda)$ on ${\mathbf{S}}^{n-1}$ and the set $\{ \Lambda; (1,\Lambda)\in\Gamma\}$ are often identified with Λ and Γ, respectively. By $\Xi\times\Gamma$ we denote the set $\{(l,\Lambda )\in{\mathbf{R}}^{n}; l\in\Xi,(1,\Lambda)\in\Gamma\}$, where $\Xi\subset{\mathbf{R}}_{+}$. The set ${\mathbf{R}}_{+}\times\Gamma$ is denoted by $\mathcal{T}_{n}(\Gamma)$, which is called a cone. We denote the sets $I\times\Gamma$ and $I\times\partial{\Gamma}$ by $\mathcal {T}_{n}(\Gamma;I)$ and $\mathcal{S}_{n}(\Gamma;I)$, respectively, where $I\subset\mathbf{R}$. The two sets $\mathcal{T}_{n}(\Gamma)\cap S_{l}$ and $\mathcal{S}_{n}(\Gamma; (0,+\infty))$ are denoted by $\mathcal {S}_{n}(\Gamma; l)$ and $\mathcal{S}_{n}(\Gamma)$, respectively.

If the Green’s function on $\mathcal{T}_{n}(\Gamma)$ is denoted by $\mathrm{G}_{\Gamma}(V,W)$ ($P, Q\in\mathcal{T}_{n}(\Gamma)$), then the Poisson kernel on $\mathcal{T}_{n}(\Gamma)$ is defined by

$$c_{n}\operatorname{PI}_{\Gamma}(V,W)=\frac{\partial\mathrm{G}_{\Gamma }(V,W)}{\partial n_{W}}, $$

where

$$c_{n}=\left \{ \textstyle\begin{array}{l@{\quad}l} 2\pi & \mbox{if } n=2, \\ (n-2)w_{n} & \mbox{if } n\geq3, \end{array}\displaystyle \right . $$

and ${\partial}/{\partial n_{W}}$ denotes the differentiation at W along the inward normal into $\mathcal{T}_{n}(\Gamma)$.

Consider the boundary value problem (see [1])

$$\begin{aligned}& \bigl(\Xi^{*}+\iota\bigr)\eta=0 \quad \text{on } \Gamma, \end{aligned}$$

(1)

$$\begin{aligned}& \eta=0 \quad \text{on } \partial{\Gamma}, \end{aligned}$$

(2)

where $\Xi^{*}$ is the spherical Laplace operator and Γ ($\subset{\mathbf{S}}^{n-1}$) has a twice smooth boundary. The least positive eigenvalue of (1) and (2) is denoted by ι. By $\eta(\Lambda)$ we denote the normalized eigenfunction corresponding to ι. Define

$$2\varrho^{\pm}=-n+2\pm\sqrt{(n-2)^{2}+4\iota}, $$

$\varrho^{+}-\varrho^{-}$ will be denoted by λ.

We denote $f^{+}=\max\{f,0\}$ and $f^{-}=-\min\{f,0\}$, where f is a function defined on $\mathcal {T}_{n}(\Gamma)$. Throughout this paper, let A denote various constants independent of the variables in questions, which may be different from line to line. Let $\sigma(t)$ be a nondecreasing real valued function on $[1,+\infty)$ satisfying $\sigma(t)> \varrho^{+}$ for any $t\geq1$.

In a recent paper, Li and Zhang (see [2], Theorem 1) solved boundary behavior problems for functions harmonic on $\mathcal {T}_{n}(\Gamma)$, which admit some lower bounds.

Theorem A

Let $h(V)$be a harmonic function on $\mathcal{T}_{n}(\Gamma)$and a continuous function on $\overline{\mathcal{T}_{n}(\Gamma)}$, where $V=(R,\Lambda)$. If

$$h(V)\leq KR^{\sigma(R)} $$

for any $V=(R,\Lambda)\in\mathcal{T}_{n}(\Gamma;(1,+\infty))$and

$$-K\leq h(V) $$

for any $V=(R,\Lambda) \in \overline{\mathcal{T}_{n}(\Gamma;(0,1])}$. Then we have

$$h(V)\geq-MK\bigl(1+\sigma(R)R^{\sigma(R)}\bigr)\eta^{1-n}(\Lambda), $$

where $V\in\mathcal{T}_{n}(\Gamma)$, Kis a constant andMdenotes a constant independent ofR, K, and the two functions $h(V)$and $\eta(\Lambda)$.

Main results

Now we state our main results in this paper.

By using a modified Carleman formula and an augmented Riesz decomposition method, we obtain sharper estimates of harmonic functions with certain boundary integral conditions. Compared with the original proof in [2], the new one is more easily applied.

Theorem 1

Let $h(V)$be a function harmonic on $\mathcal{T}_{n}(\Gamma)$and continuous on $\overline{\mathcal{T}_{n}(\Gamma)}$, where $V=(R,\Lambda)$. Suppose that the two conditions (I) and (II) hold:

• (I) For any $V=(R,\Lambda)\in\mathcal{T}_{n}(\Gamma;(1,\infty))$, we have

  $$ \int_{\mathcal{S}_{n}(\Gamma;(1,R))}h^{-}t^{\varrho^{-}}{\partial \eta}/{ \partial n}\, d\sigma_{W} \leq MK\sigma(dR)R^{\sigma(dR)-\varrho^{-}} $$

  (3)

  and

  $$ \lambda \int_{\mathcal{S}_{n}(\Gamma ;R)}h^{+}R^{\varrho^{-}-1}\eta \, d S_{R} \leq MKR^{\sigma(dR)-\varrho^{-}}. $$

  (4)

• (II) For any $V=(R,\Lambda)\in\mathcal{T}_{n}(\Gamma;(0,1])$, we have

  $$ h(V)\geq-K. $$

  (5)

Then

$$h(V)\geq-MK\bigl(1+\sigma(dR)R^{\sigma(dR)}\bigr)\eta^{1-n}(\Lambda), $$

where $V\in\mathcal{T}_{n}(\Gamma)$, Kis a constant, $0< d\leq1$, andMdenotes a constant independent ofR, K, and the two functions $h(V)$and $\eta(\Lambda)$.

Remark 1

By virtue of Theorem 1, we easily see that Theorem 1(I) is weaker than corresponding condition in Theorem A in the case $d\equiv1$.

Theorem 2

The conclusion of Theorem  1remains valid if Theorem  1(I) is replaced by

$$ h(V)\leq KR^{\sigma(dR)}, \quad V=(R,\Lambda)\in\mathcal {T}_{n}\bigl(\Gamma;(1,\infty)\bigr), $$

(6)

where $0< d\leq1$.

Remark 2

In the case $d\equiv1$, Theorem 2 reduces to Theorem A.

Lemmas

The following result is an augmented Riesz decomposition method, which was used to study the boundary behaviors of Poisson integral. For similar results for solutions of the equilibrium equations with angular velocity, we refer the reader to the paper by Wang et al. (see [3]).

Lemma 1

For $W'\in\partial{\mathcal{T}_{n}(\Gamma)}$and $\epsilon>0$, there exist a positive numberRand a neighborhood $B(W')$of $W'$such that

$$ \frac{1}{c_{n}} \int_{S_{n}(\Gamma;(R,\infty))}\bigl\vert g(W)\bigr\vert \bigl\vert \operatorname{PI}_{\Gamma }(V,W)\bigr\vert \, d\sigma_{W}< \epsilon $$

(7)

for any $V=(r,\Lambda)\in\mathcal{T}_{n}(\Gamma)\cap B(W')$, wheregis an upper semi-continuous function. Then

$$ \limsup_{V\in\mathcal{T}_{n}(\Gamma), V\rightarrow W'}\operatorname{PI}_{\Gamma}[g](V) \leq g\bigl(W'\bigr). $$

(8)

Proof

Let $W'=(l',\Phi')$ be any point of $\partial{\mathcal{T}_{n}(\Gamma)}$ and ϵ (>0) be any number. There exists a positive number $R'$ satisfying

$$ \frac{1}{c_{n}} \int_{S_{n}(\Gamma;(R',\infty))}\bigl\vert \operatorname{PI}_{\Gamma }(V,W) \bigr\vert \bigl\vert g(W)\bigr\vert \, d\sigma_{W}\leq \frac{\epsilon}{4} $$

(9)

for any $V=(r,\Lambda)\in\mathcal{T}_{n}(\Gamma)\cap B(W')$ from (7).

Let ϕ be continuous on $\partial{\mathcal{T}_{n}(\Gamma)}$ such that $0\leq\phi\leq1$ and

$$\phi=\left \{ \textstyle\begin{array}{l@{\quad}l} 1 & \mbox{on } S_{n}(\Gamma;(0,R'])\cup\{O\}, \\ 0 & \mbox{on } S_{n}(\Gamma;(2R',\infty)). \end{array}\displaystyle \right . $$

Let $\mathrm{G}_{\mathcal{T}_{n}(\Gamma;(0,j))}$ be a Green’s function on $\mathcal{T}_{n}(\Gamma;(0,j))$, where j is a positive integer. Since $\Gamma_{j}(V,W)=\mathrm{G}_{\mathcal{T}_{n}(\Gamma)}(V,W)-\mathrm {G}_{\mathcal{T}_{n}(\Gamma;(0,j))}(V,W)$ on $\mathcal{T}_{n}(\Gamma;(0,j))$ converges monotonically to 0 as $j\rightarrow\infty$. Then we can find an integer $j'$, $j'>2R'$ such that

$$ \frac{1}{c_{n}} \int_{S_{n}(\Gamma;(0,2R'))}\biggl\vert \frac{\partial }{\partial n_{W}}\Gamma_{j'}(V,W) \biggr\vert \bigl\vert \phi(W)g(W)\bigr\vert \, d\sigma_{W}< \frac{\epsilon}{4} $$

(10)

for any $V=(r,\Lambda)\in\mathcal{T}_{n}(\Gamma)\cap B(W')$.

Then we have from (9) and (10)

$$\begin{aligned}& \frac{1}{c_{n}} \int_{\partial{\mathcal{T}_{n}(\Gamma)}}\operatorname {PI}_{\Gamma}(V,W)g(W)\,d \sigma_{W} \\& \quad \leq \frac{1}{c_{n}} \int_{S_{n}(\Gamma;(0,2R'))}\frac{\partial \mathrm{G}_{\mathcal{T}_{n}(\Gamma;(0,j'))}(V,W) }{\partial n_{W}}\phi(W)g(W)\,d\sigma_{W} \\& \qquad {} + \frac{1}{c_{n}} \int_{S_{n}(\Gamma;(0,2R'))}\bigl\vert \phi(W)g(W)\bigr\vert \biggl\vert \frac {\partial \Gamma_{j'}(V,W)}{\partial n_{W}}\biggr\vert \,d\sigma_{W} \\& \qquad {} + \frac{2}{c_{n}} \int_{S_{n}(\Gamma;(R',\infty))}\bigl\vert \operatorname {PI}_{\Gamma}(V,W) \bigr\vert \bigl\vert g(W)\bigr\vert \,d\sigma_{W} \\& \quad \leq \frac{1}{c_{n}} \int_{S_{n}(\Gamma;(0,2R'))}\frac{\partial \mathrm{G}_{\mathcal{T}_{n}(\Gamma;(0,j'))}(V,W) }{\partial n_{W}}\phi(W)g(W)\,d\sigma_{W}+ \frac{3}{4}\epsilon \end{aligned}$$

(11)

for any $V=(r,\Lambda)\in\mathcal{T}_{n}(\Gamma)\cap B(W')$.

Consider an upper semi-continuous function

$$\eta(W)=\left \{ \textstyle\begin{array}{l@{\quad}l} \phi(W)g(W) & \mbox{on } S_{n}(\Gamma;(0,2R'])\cup\{O\}, \\ 0 & \mbox{on } \partial{\mathcal{T}_{n}(\Gamma;(0,j'))}-S_{n}(\Gamma ;(0,2R'])-\{O\}, \end{array}\displaystyle \right . $$

on $\partial{\mathcal{T}_{n}(\Gamma;[0,j'))}$ and denote the PWB solution of the Dirichlet problem on $\mathcal{T}_{n}(\Gamma;(0,j'))$ by $H_{\eta}(P;\mathcal{T}_{n}(\Gamma ;(0,j')))$ (see, e.g., [4]); we know that

$$ \frac{1}{c_{n}} \int_{S_{n}(\Gamma;(0,2R'))}\frac{\partial \mathrm{G}_{\mathcal{T}_{n}(\Gamma;(0,j'))}(V,W) }{\partial n_{W}}\phi(W)g(W)\, d\sigma_{W}=H_{\eta} \bigl(P;\mathcal{T}_{n}\bigl(\Gamma;\bigl(0,j'\bigr)\bigr) \bigr) $$

(12)

(see [5], Theorem 3). If $\mathcal{T}_{n}(\Gamma;(0,j'))$ is not a Lipschitz domain at O, we can prove (12) by considering a sequence of the Lipschitz domains $\mathcal{T}_{n}(\Gamma;(\frac{1}{m},j'))$ which converges to $\mathcal{T}_{n}(\Gamma;(0,j'))$ as $m\rightarrow\infty$. We also have

$$\limsup_{V\in\mathcal{T}_{n}(\Gamma), V\rightarrow W'}H_{\eta }\bigl(P;\mathcal{T}_{n} \bigl(\Gamma;\bigl(0,j'\bigr)\bigr)\bigr)\leq\limsup _{Q\in S_{n}(\Gamma), Q\rightarrow W'}\eta(W)=g\bigl(W'\bigr) $$

(see, e.g., [4], Lemma 8.20). Hence we know that

$$\limsup_{V\in\mathcal{T}_{n}(\Gamma), V\rightarrow W'}\frac {1}{c_{n}} \int_{S_{n}(\Gamma;(0,2R'))}\phi(W)\frac{\partial \mathrm{G}_{\mathcal{T}_{n}(\Gamma;(0,j'))}(V,W) }{\partial n_{W}}g(W)\, d\sigma\leq g \bigl(W'\bigr). $$

With (11) this gives (8). □

The following growth properties play important roles in our discussions.

Lemma 2

(see [6])

Let $V=(r,\Lambda )\in\mathcal{T}_{n}(\Gamma)$and $W=(t,\Phi)\in S_{n}(\Gamma)$. Then we have

$$\operatorname{PI}_{\Gamma}(V,W)\leq M r^{\varrho^{-}}t^{\varrho^{+}-1} \eta(\Lambda)\quad \biggl(0< \frac{t}{r}\leq \frac{4}{5}\biggr) $$

and

$$\operatorname{PI}_{\Gamma}(V,W)\leq M r^{\varrho^{+}}t^{\varrho^{-}-1} \eta(\Lambda) \quad \biggl(0< \frac{r}{t}\leq \frac{4}{5}\biggr). $$

Let $V=(r,\Lambda)\in\mathcal{T}_{n}(\Gamma)$and $W=(t,\Phi)\in S_{n}(\Gamma; (\frac{4}{5}r,\frac{5}{4}r))$. Then we have

$$\operatorname{PI}_{\Gamma}(V,W)\leq M\frac{\eta(\Lambda)}{t^{n-1}}+M \frac{r\eta(\Lambda)}{|P-Q|^{n}}. $$

Let $\mathrm{G}_{\mathcal{T}_{n}(\Gamma;(t_{1},t_{2}))}$be the Green’s function of $\mathcal{T}_{n}(\Gamma;(t_{1},t_{2}))$. Then we have

$$\frac{\partial\mathrm{G}_{\mathcal{T}_{n}(\Gamma ;(t_{1},t_{2}))}((t_{1},\Phi),(r,\Lambda))}{\partial t}\leq M\biggl(\frac{t_{1}}{r}\biggr)^{-\varrho^{-}} \frac{\eta(\Phi)\eta(\Lambda)}{t_{1}^{n-1}} $$

and

$$-M\biggl(\frac{r}{t_{2}}\biggr)^{\varrho^{+}}\frac{\eta(\Phi)\eta(\Lambda )}{t_{2}^{n-1}}\leq \frac{\partial\mathrm{G}_{\mathcal{T}_{n}(\Gamma ;(t_{1},t_{2}))}((t_{2},\Phi),(r,\Lambda))}{\partial t}, $$

where $0<2t_{1}<r<\frac{1}{2}t_{2}<+\infty$.

Many previous studies (see [7, 8]) focused on the following lemma with respect to the half space and its applications.

Lemma 3

(see [2], Lemma 2)

Ifhis a function harmonic in a domain containing $\mathcal{T}_{n}(\Gamma ;(1,R))$, where $R>1$, then we have

$$\lambda \int_{S_{n}(\Gamma;R)}h\eta R^{\varrho^{-}-1} d S_{R} + \int_{S_{n}(\Gamma;(1,R))}h \bigl(t^{\varrho^{-}}-t^{\varrho ^{+}}R^{-\lambda} \bigr){\partial\eta}/{\partial n} \, d\sigma_{W}+d_{1}+ \frac{d_{2}}{R^{\lambda}}=0, $$

where

$$d_{1}= \int_{S_{n}(\Gamma;1)}\varrho^{-}h\eta-\eta({\partial h}/{\partial n}) \, dS_{1} $$

and

$$d_{2}= \int_{S_{n}(\Gamma;1)}\eta({\partial h}/{\partial n})-\varrho ^{+}h\eta \, dS_{1}. $$

Proof of Theorem 1

By Lemma 1 we have

$$\begin{aligned} -h(V) =& \int_{\mathcal{S}_{n}(\Gamma;(0,R))}\bigl(-h(W)\bigr)\operatorname{PI}_{\Gamma }(V,W) \,d\sigma_{W} \\ &{}+ \int_{\mathcal{S}_{n}(\Gamma;R)}\bigl(-h(W)\bigr)\frac {\partial\mathrm{G}_{\Gamma,R}(V,W)}{\partial R}\, dS_{R} \end{aligned}$$

(13)

for any $V=(l,\Lambda)\in\mathcal{T}_{n}(\Gamma;(0,R))$.

Case 1. $V=(l,\Lambda)\in\mathcal{T}_{n}(\Gamma ;(\frac{5}{4},\infty))$ and $R=\frac{5}{4}l$.

From (13) we know that

$$ -h=\sum_{i=1}^{4} \mathfrak{U}_{i} , $$

(14)

where

$$\begin{aligned}& \mathfrak{U}_{1}(V)= \int_{\mathcal{S}_{n}(\Gamma;(0,1])}\bigl(-h(W)\bigr)\operatorname {PI}_{\Gamma}(V,W)\,d \sigma_{W}, \\& \mathfrak{U}_{2}(V)= \int_{\mathrm{S}_{n}(\Gamma;(1,\frac {4}{5}l])}\bigl(-h(W)\bigr)\operatorname {PI}_{\Gamma}(V,W)\,d \sigma_{W}, \\& \mathfrak{U}_{3}(V)= \int_{\mathcal{S}_{n}(\Gamma;(\frac {4}{5}l,R))}\bigl(-h(W)\bigr)\operatorname {PI}_{\Gamma}(V,W)\,d \sigma_{W}, \end{aligned}$$

and

$$\mathfrak{U}_{4}(V)= \int_{\mathcal{S}_{n}(\Gamma;R)}\bigl(-h(W)\bigr)\operatorname {PI}_{\Gamma}(V,W)\, d \sigma_{W}. $$

We obtain from Lemma 2

$$ \mathfrak{U}_{1}(V)\leq MK\eta(\Lambda) $$

(15)

and

$$ \mathfrak{U}_{2}(V) \leq MK\sigma(dR)R^{\sigma(dR)} \eta(\Lambda). $$

(16)

Put

$$ \mathfrak{U}_{3}(V)\leq \mathfrak{U}_{31}(V)+ \mathfrak{U}_{32}(V), $$

(17)

where

$$\mathfrak{U}_{31}(V)=M \int_{\mathcal{S}_{n}(\Gamma;(\frac {4}{5}l,R))}\bigl(-h(W)\bigr)t^{1-n} \eta(\Lambda) \frac{\partial\phi( \Phi)}{\partial n_{\Phi}}\, d\sigma_{W} $$

and

$$\mathfrak{U}_{32}(V)=Mr\eta(\Lambda) \int_{\mathcal{S}_{n}(\Gamma ;(\frac{4}{5}l,R))}\bigl(-h(W)\bigr)\vert V-W\vert ^{-n} l \eta(\Lambda) \frac{\partial\phi( \Phi)}{\partial n_{\Phi}}\, d\sigma_{W}. $$

From (3) we obtain

$$ \mathfrak{U}_{31}(V) \leq MK\sigma(dR)R^{\sigma(dR)} \eta(\Lambda). $$

(18)

To estimate $\mathfrak{U}_{32}(V)$. There exists a sufficiently small number d satisfying $d>0$ and

$$\mathcal{S}_{n}\biggl(\Gamma;\biggl(\frac{4}{5}l,R\biggr) \biggr)\subset B\biggl(V,\frac{l}{2}\biggr) $$

for $V=(l,\Lambda)\in\Pi(d)$, where

$$\Pi(d)=\Bigl\{ Q=(r,\Lambda)\in \mathcal{T}_{n}(\Gamma); \inf _{(1,z)\in\partial\Gamma}\bigl\vert (1,\Lambda)-(1,z)\bigr\vert < d, 0< l< \infty \Bigr\} . $$

We divide $\mathcal{T}_{n}(\Gamma)$ into the two sets $\Pi(d)$ and $\mathcal{T}_{n}(\Gamma)-\Pi(d)$.

For any $V=(l,\Lambda)\in\mathcal{T}_{n}(\Gamma)-\Pi(d)$, we can find a number $d'$ satisfying $d'>0$ and

$${d}'l\leq|V-W| $$

for $W\in \mathcal{S}_{n}(\Gamma)$, and hence

$$ \mathfrak{U}_{32}(V) \leq MK\sigma(dR)R^{\sigma(dR)} \eta(\Lambda). $$

(19)

If $V=(l,\Lambda)\in\Pi(d)$, then we have

$$H_{i}(V)=\biggl\{ W\in\mathcal{S}_{n}\biggl(\Gamma;\biggl( \frac{4}{5}l,R\biggr)\biggr); 2^{i-1}\xi(V) \leq|V-W|< 2^{i}\xi(V)\biggr\} , $$

where

$$\xi(V)=\inf_{W\in \partial{\mathcal{T}_{n}(\Gamma)}}|V-W|. $$

Since $\{W\in{\mathbf{R}}^{n}: |V-W|< \xi(V)\}\cap\mathcal{S}_{n}(\Gamma)=\varnothing$, we get

$$\mathfrak{U}_{32}(V)=M\sum_{i=1}^{i(V)} \int_{H_{i}(V)}\frac{(-h(W))r\eta(\Lambda )}{|V-W|^{n}}\frac{\partial \eta( \Phi)}{\partial n_{\Phi}}\, d \sigma_{W}, $$

where $l(P)$ is an integer such that $2^{l(P)}\xi(V)\leq r<2^{l(P)+1}\xi(V)$.

Since

$$\eta(\Lambda)\leq Ml^{-1}\xi(V), $$

where $V=(l,\Lambda)\in \mathcal{T}_{n}(\Gamma)$, we have

$$\int_{H_{i}(V)}\bigl(-h(W)\bigr)|V-W|^{-n}r\eta(\Lambda) \frac{\partial \eta( \Phi)}{\partial n_{\Phi}}\, d\sigma_{W}\leq MK\sigma(dR)R^{\sigma(dR)} \eta^{1-n}(\Lambda), $$

where $l=0,1,2,\ldots,l(P)$.

Thus

$$ \mathfrak{U}_{32}(V)\leq MK\sigma(dR)R^{\sigma(dR)} \eta ^{1-n}(\Lambda). $$

(20)

We see that

$$ \mathfrak{U}_{3}(V)\leq MK\sigma(dR)R^{\sigma(dR)} \eta ^{1-n}(\Lambda) $$

(21)

from (17), (18), (19), and (20).

On the other hand, we have from (4)

$$ \mathfrak{U}_{4}(V) \leq MKR^{\sigma(dR)}\eta( \Lambda). $$

(22)

We thus obtain (15), (16), (21), and (22) that

$$ -h(V)\leq MK\bigl(1+\sigma(dR)R^{\sigma(dR)}\bigr) \eta^{1-n}(\Lambda). $$

(23)

Case 2. $V=(l,\Lambda)\in \mathcal{T}_{n}(\Gamma;(\frac{4}{5},\frac{5}{4}])$ and $R=\frac{5}{4}l$.

It follows from (13) that

$$-h= \mathfrak{U}_{1}+\mathfrak{U}_{5}+ \mathfrak{U}_{4}, $$

where $\mathfrak{U}_{1}(V)$ and $\mathfrak{U}_{4}(V)$ were defined in the former case and

$$\mathfrak{U}_{5}(V)= \int_{\mathcal{S}_{n}(\Gamma;(1,R))}\bigl(-h(W)\bigr)\operatorname {PI}_{\Gamma}(V,W)\, d \sigma_{W}. $$

Similarly, we have

$$\mathfrak{U}_{5}(V)\leq MK\sigma(dR)R^{\sigma(dR)}\eta ^{1-n}(\Lambda), $$

which, together with (15) and (22), gives (23).

Case 3. $V=(l,\Lambda)\in \mathcal{T}_{n}(\Gamma;(0,\frac{4}{5}])$.

It is evident from (5) that we have

$$-h\leq K, $$

from which one also obtains (23).

We finally have

$$h(V)\geq -KM\bigl(1+\sigma(dR)R^{\sigma(dR)}\bigr)\eta^{1-n}(\Lambda) $$

from (23), which is required.

Proof of Theorem 2

By applying Lemma 3 to $h=h^{+}-h^{-}$, we obtain

$$\begin{aligned}& \lambda \int_{\mathcal{S}_{n}(\Gamma;R)}h^{+}R^{\varrho^{-}-1}\eta \,d S_{R}+ \int_{\mathcal{S}_{n}(\Gamma;(1,R))}h^{+} \bigl(t^{\varrho ^{-}}-t^{\varrho^{+}}R^{-\lambda} \bigr) {\partial\eta}/{\partial n} \,d\sigma_{W}+d_{1}+d_{2}R^{-\lambda} \\& \quad =\lambda \int_{\mathcal{S}_{n}(\Gamma;R)}h^{-}R^{\varrho^{-}-1}\eta \,d S_{R}+ \int_{\mathcal{S}_{n}(\Gamma;(1,R))}h^{-} \bigl(t^{\varrho ^{-}}-t^{\varrho^{+}}R^{-\lambda} \bigr) {\partial\eta}/{\partial n} \,d\sigma_{W}. \end{aligned}$$

(24)

From (6) we see that

$$ \lambda \int_{\mathcal{S}_{n}(\Gamma;R)}h^{+}R^{\varrho^{-}-1}\eta\, d S_{R} \leq MKR^{\sigma(dR)-\varrho^{+}} $$

(25)

and

$$ \int_{\mathcal{S}_{n}(\Gamma;(1,R))}h^{+} \bigl(t^{\varrho ^{-}}-t^{\varrho^{+}}R^{-\lambda} \bigr){\partial\eta}/{\partial n}\, d\sigma_{W}\leq MKR^{\sigma (dR)-\varrho^{+}}. $$

(26)

Notice that

$$ d_{1}+d_{2}R^{-\lambda} \leq MKR^{\sigma(dR)-\varrho^{+}}. $$

(27)

We have from (24), (25), (26), and (27)

$$ \int_{\mathcal{S}_{n}(\Gamma ;(1,R))}h^{-} \bigl(t^{\varrho^{-}}-t^{\varrho^{+}}R^{-\lambda } \bigr) {\partial\eta}/{\partial n}\, d\sigma_{W} \leq MKR^{\sigma(dR)-\varrho^{+}}. $$

(28)

Hence (28) gives (6), which, together Theorem 1, gives the conclusion of Theorem 2.