数学杂志  2018, Vol. 38 Issue (6): 1107-1118   PDF    
扩展功能
加入收藏夹
复制引文信息
加入引用管理器
Email Alert
RSS
本文作者相关文章
张石梅
龙见仁
吴秀碧
非线性微分-差分方程的解
张石梅1, 龙见仁1,2, 吴秀碧1,3    
1. 贵州师范大学数学科学学院, 贵州 贵阳 550001;
2. 北京邮电大学计算机学院; 理学院, 北京 100876;
3. 贵州大学数学与统计学院, 贵州 贵阳 550025
摘要:本文研究了非线性微分-差分方程fzn+an-1fzn-1+…+a1fz)+qzeQzfkz+c)=Pz)的有穷级非零整函数解的增长性和零点分布问题.利用微分-差分Nevanlinna值分布的方法,获得了当方程的系数满足一定条件时,方程解的增长性估计和零点分类.特别地,当n=2,a1≠0指数多项式解满足某些条件时,获得了解具有特别的形式.该结果推广了先前文献[12]的结果.
关键词微分-差分方程    指数多项式    有穷级    
SOLUTIONS OF NONLINEAR DIFFERENTIAL-DIFFERENCE EQUATIONS
ZHANG Shi-mei1, LONG Jian-ren1,2, WU Xiu-bi1,3    
1. School of Mathematical Science, Guizhou Normal University, Guiyang 550001, China;
2. School of Computer Science; School of Science, Beijing University of Posts and Telecommunications, Beijing 100876, China;
3. School of Mathematics and Statistics, Guizhou University, Guiyang 550025, China
Abstract: In this paper, we study the growth and distribution of zeros of entire solutions of finite order of nonlinear differential-difference equation f(z)n+an-1f(z)n-1+…+a1f(z)+q(z)eQ(z)f(k)(z+c)=P (z). By using the differential-difference Nevanlinna values distribution theory, we obtain an estimation of the growth and the distribution of zeros of solutions of the differential-difference equation if there are some attached condiction on the coefficients. Particularly, when n=2 and a1≠0, we obtain that exponential polynomial solutions satisfying some condictions must reduce to rather specific forms, which improves the results of [1, 2].
Keywords: differential-difference equation     exponential polynomial     finite order    
1 引言及结果

本文中, 亚纯函数指复平面上的亚纯函数, 假定读者熟悉Nevanlinna值分布理论的标准记号及主要结果[3-6].例如$m(r, f), ~N(r, f), ~T(r, f)$等.为方便起见, 用$S(r, f)$表示一个量使得$ \frac{S(r, f)}{T(r, f)}\rightarrow 0~(r\rightarrow 0)$, 除去一个对数测度为有穷的集合; 并用${\rm card}(X)$表示集合$X$中的元素个数.

最近, 许多专家学者利用差分Nevanlinna理论, 尤其是差分对数导数引理[7-9]来研究复差分, 复微分-差分方程.刘凯和他的合作者考虑了Fertmat型微分-差分方程的解[10-13].在文献[13, 定理2.6], Yang和Laine研究了非线性微分-差分$ f(z)^{n}+L(z, f)=h(z)~(n\geq2)$的有穷级整函数解, 其中$ L(z, f)$是关于$ f(z)$的线性微分-差分多项式, $ h(z)$是亚纯函数.

Yang和Laine (参见文献[13, 定理2.4])得到方程

$ f(z)^{2}+q(z)f(z+1)=P(z) $ (1.1)

没有有穷级的超越整函数解, 其中$ q(z), ~P(z)$是多项式.

温志涛等人[14]讨论了方程

$ f(z)^{n}+q(z)e^{Q(z)}f(z+c)=P(z) $ (1.2)

的有穷级非零整函数解, 其中$q(z), ~ Q(z), ~P(z)$是多项式, $n\geq2$是正整数, $c\in\mathbb{C}\backslash\{0\}$, 并得到了定理1.1.

定理1.1 [14]$q(z), ~ Q(z), ~P(z)$是多项式且$Q(z)$非常数, $q(z)\not\equiv 0, ~n\geq2, ~c\in\mathbb{C}\backslash\{0\}$, 那么方程(1.2)的有穷级超越整函数解满足

(ⅰ)每一个解满足$\sigma(f)=\deg(Q(z))$$f(z)$是正规型的.

(ⅱ)每一个解满足$\lambda(f)=\sigma(f)$的充要条件是$P(z)\not\equiv 0 $.

(ⅲ) $f\in\Gamma_{0}$当且仅当$P(z)\equiv 0.$特别地, $n\geq3$时成立.

(ⅳ)如果解$f, g\in\Gamma_{0}$, 那么$f=\eta g$$\eta^{n-1}=1$.

(ⅴ)如果$f(z)$是(1.5)形式的指数多项式解, 那么$f(z)\in\Gamma_{1}$.更进一步地, 如果$f\in \Gamma_{1}\backslash \Gamma_{0}, $那么$\sigma (f)=1.$

刘凯[2]研究了方程

$ f(z)^{n}+q(z)e^{Q(z)}f^{(k)}(z+c)=P(z) $ (1.3)

的有穷级非零整函数解, 其中$q(z), ~ Q(z), ~P(z)$是多项式, $n\geq2$是正整数, $c\in\mathbb{C}\backslash\{0\}$, 并得到了定理1.2.

定理1.2 [2]$q(z), ~ Q(z), ~P(z)$是多项式且$Q(z)$非常数, $q(z)\not\equiv 0, ~k\geq1$$n\geq2, ~c\in\mathbb{C}\backslash\{0\}$, 那么方程(1.3)的有穷级超越整函数解满足

(ⅰ)每一个解满足$\sigma(f)=\deg(Q(z))$$f(z)$是正规型的.

(ⅱ)每一个解满足$\lambda(f)=\sigma(f)$的充要条件是$P(z)\not\equiv 0 $.

(ⅲ) $f\in\Gamma'_{0}$当且仅当$P(z)\equiv 0. $特别地, $n\geq3$时成立.

(ⅳ)如果解$f, ~g\in\Gamma'_{0}$, 那么$f=\eta g$$\eta^{n-1}=1$.

(ⅴ)如果$f(z)$是(1.5)形式的指数多项式解, 那么$f(z)\in\Gamma'_{1} $.

李楠等[1]研究了下列更为一般形式方程的解

$ f(z)^{n}+a_{n-1}f(z)^{n-1}+\cdots+a_{1}f(z)+q(z)e^{Q(z)}f(z+c)=P(z) , $ (1.4)

其中$q(z), ~Q(z), ~P(z)$是多项式, $n\geq2$是正整数, $c\in\mathbb{C}\backslash\{0\}$, 并得到了定理1.3.

定理1.3 [1]$q(z), ~ Q(z), ~P(z)$是多项式且$Q(z)$非常数, $q(z)\not\equiv 0$$n\geq2, ~c\in\mathbb{C}\backslash\{0\}$, 那么方程(1.4)的有穷级非零整函数解满足

(a) 每一个解满足$\sigma(f)=\deg(Q(z))$$f(z)$是正规型的.

(b) 如果0是$f(z)$的Borel例外值, 那么$a_{n-1}=\cdots=a_{1}=0\equiv P(z)$.

(c) 如果$P(z)\equiv 0$, 那么有$z^{n-1}+a_{n-1}z^{n-2}+\cdots +a_{1}=(z+\frac{a_{n-1}}{n})^{n-1}$.更进一步地, 如果存在$i_{0}\in \{1, 2, \cdots, n-1\} $使得$a_{i_{0}}=0$, 那么所有的$a_{j}=0~(j=1, 2, \cdots, n-1)$且有$\lambda(f)< \sigma(f)$; 否则, 有$\lambda(f)=\sigma(f)$.

(d) $f\in\Gamma_{0}$当且仅当$P(z)\equiv 0$且存在$i_{0}\in\{1, 2, \cdots, n-1\}$使得$ a_{i_{0}}=0$.

(e) 当$n\geq2$时, 如果存在$ i_{0}\in\{1, 2\cdots, n-1\}$使得$a_{i_{0}}=0$${\rm card}\{z:P_{1}(z)=P'_{1}(z)=P''_{1}(z)=0\}\geq 1$或者${\rm card}\{z:P_{1}(z)=P'_{1}(z)=0\}\geq 2$, 其中$P_{1}(z)=z^{n}+a_{n-1}z^{n-1}+\cdots+a_{1}, $那么$f(z)\in\Gamma_{0}$$P(z)\equiv 0=a_{1}(z)=\cdots=a_{n-1} $.

(f) 如果解$f, ~g\in\Gamma_{0}$, 那么$f=\eta g$$\eta^{n-1}=1$.

注1  方程(1.4)存在无穷级的超越城函数解.例如$f(z)=e^{z}e^{e^{z}}+1$是方程$f^{2}(z)-2f(z)-\frac{1}{2}f(z+\log2)=1$的无穷级超越整函数解.

为方便读者, 回忆下列形式指数多项式的定义

$ f(z)=p_{1}(z)e^{Q_{1}(z)}+\cdots+p_{k}(z)e^{Q_{k}(z)}, $ (1.5)

其中$p_{j}(z), ~ Q_{j}(z)~(j=1, 2, \cdots, k)$是关于$z$的多项式.令$q=\max\{ \deg(Q_{j}(z)), Q_{j}(z)\not\equiv0~(j=1, 2, \cdots, k)\}$, $ w_{j}~(j=1, 2, \cdots, m)$是多项式$ Q_{j}(z)(j=1, 2, \cdots, k)$的最高次数为$q$的主导系数.因此(1.5)式可以改写成下列形式

$ f(z)=H_{0}(z)+H_{1}(z)e^{w_{1}z^{q}}+\cdots+H_{m}(z)e^{w_{m}z^{q}}, $ (1.6)

其中$ H_{j}~(j=0, 1, \cdots, m)$或者是次数小于$q$的指数多项式, 或者是关于$z$的一般多项式, 令

$ \begin{eqnarray}&&\Gamma_{1}=\{h=e^{\alpha(z)} +d;d\in\mathbb{C}\} , ~\Gamma_{0}=\{h=e^{\alpha(z)} \}, \\ &&\Gamma'_{1}=\{p(z)e^{\alpha(z)} +h(z)\}, ~\Gamma'_{0}=\{p(z)e^{\alpha(z)}\}.\end{eqnarray}$

上述$p(z), ~h(z), ~\alpha(z)$是多项式, 其中$\alpha(z)$非常数.

受定理1.2和定理1.3的启发, 本文考虑了下述方程

$ f(z)^{n}+a_{n-1}f(z)^{n-1}+\cdots+a_{1}f(z)+q(z)e^{Q(z)}f^{(k)}(z+c)=P(z), $ (1.7)

其中$q(z), ~Q(z), ~P(z)$是多项式, $n\geq2$是正整数, $c\in\mathbb{C}\backslash\{0\}, ~a_{i}\in\mathbb{C}~(1, 2, \cdots, n-1)$, 并得到了以下结果.

定理1.4  设$q(z), ~ Q(z), ~P(z)$是多项式且$Q(z)$非常数, $q(z)\not\equiv 0$$n\geq2, ~ k\geq1, ~c\in\mathbb{C}\backslash\{0\}$, 那么方程(1.7)的有穷级非零整函数解满足

(ⅰ)每一个解满足$\sigma(f)=\deg(Q(z))$$f(z)$是正规型的.

(ⅱ)如果有穷复数$a$$f(z)$的Borel例外值, 那么有$a=-\frac{a_{n-1}}{n}$.特别地, 当$a=0$时, 有$a_{n-1}=\cdots=a_{1}=0\equiv P(z)$.

(ⅲ)当$P(z)\equiv 0$时, 如果对任意的$j\in \{1, 2, \cdots, n-1\}$都有$ a_{j}\neq 0~ (j=1, 2, \cdots, n-1)$, 那么有$\overline{\lambda}(f)=\sigma(f)$; 否则, 对所有的$j\in \{1, 2, \cdots, n-1\}$$ a_{j}= 0~ (j=1, 2, \cdots, n-1)$$\lambda(f)< \sigma(f)$.

(ⅳ) $f\in\Gamma'_{0}$当且仅当$P(z)\equiv 0$且存在$i_{0}\in\{1, 2\cdots, n-1\}$使得$ a_{i_{0}}=0$.

(ⅴ)当$n\geq3$时, 如果存在$i_{0}\in\{1, 2\cdots, n-1\}$使得$a_{i_{0}}=0$, 且$ {\rm card}\{z:P_{1}(z)=P'_{1}(z)=P''_{1}(z)=0\}\geq 1$或者${\rm card}\{z:P_{1}(z)=P'_{1}(z)=0\}\geq 2$, 其中, $P_{1}(z)=z^{n}+a_{n-1}z^{n-1}+\cdots+a_{1}(z), $那么$f(z)\in\Gamma'_{0}$.

(vi) 如果解$f, g\in\Gamma'_{0}$, 那么$f=\eta g$$\eta^{n-1}=1$.

注2  方程(1.7)存在有穷级整函数解.例如$f(z)=e^{z}+1$是方程$f^{2}(z)-2f(z)-3e^{z}f'(z-\log3)=-1$的有穷级非零整函数解. $1$$f(z)$的Borel例外值, 满足$a=-\frac{a_{n-1}}{n}$.

注3  当$n\geq2$, 方程(1.7)也存在无穷级超越整函数解.例如$f(z)=e^{e^{z}}-e^{-e^{z}}$是方程$f^{2}(z)-2f(z)-\frac{1}{2}e^{-z}f'(z+\log2)=4$的无穷级整函数解.

$n=2, ~a_{1}\neq 0$时, 方程(1.7)变形

$ f^{2}(z)+a_{1}(z)+q(z)e^{Q(z)}f^{(k)}(z+c)=P(z). $ (1.8)

定理1.5  设$q(z), ~ Q(z), ~P(z)$是多项式且$Q(z)$非常数, $q(z)\not\equiv 0$$n\geq2, ~k\geq1, ~ a_{1}, c\in\mathbb{C}\backslash\{0\}$, 如果$f(z)$方程(1.8)的具有(1.6)式形式的指数多项式解, 那么下列结论成立

(a) 当$n\geq2$时, 存在$ i_{0}, j_{0}\in\{1, 2, \cdots, m\}$使得$w_{i_{0}}=2w_{j_{0}}$.

(b) 当$m=1$时, $f\in\Gamma'_{1}$.

$n=1$时, 方程(1.7)退化为

$ f(z)+q(z)e^{Q(z)}f^{(k)}(z+c)=P(z). $ (1.9)

我们也得到了相应的结果.

定理1.6  设$c\in\mathbb{C}\backslash\{0\}, ~q(z), ~P(z)$是多项式, $Q(z)$非常数多项式且$q(z)\not\equiv 0, ~k\geq1 $, 那么方程(1.9)的每一个有穷级非零整函数解满足

(ⅰ) $\sigma(f)\geq \deg(Q(z))$.

(ⅱ)如果$P(z)\not\equiv 0$, 则有$\lambda(f)=\sigma(f)$.

(ⅲ)如果$P(z)\equiv 0$, 那么解不是具有(1.6)式形式的指数多项式解, 其中$q=\deg(Q(z))$.

2 引理

引理2.1 [1]$T:[0, +\infty)\rightarrow [0, +\infty)$是一个连续非减函数, $ s\in(0, +\infty)$, 如果

$ \limsup\limits_{r\rightarrow\infty}\frac{\log\log T(r)}{\log r}=\zeta<1$

$\delta\in(0, 1-\zeta), $那么

$ T(r+s)=T(r)+o(\frac{T(r)}{r^{\delta}})(r\rightarrow\infty), $

除去一个对数测度为有穷的集合.

差分对数导数引理参看文献[7-9, 15-17]在复差分方程, 差分Nevanlinna理论方面起着非常重要的作用.下面的引理是文献[17, 引理2.2]的特殊情形.

引理2.2  假设$f(z)$是一个非常数亚纯函数, $c, ~h$是两个不相等的复数.如果$\sigma_{2}(f)<1$, 那么

$ m(r, \frac{f(z+h)}{f(z+c)})=S(r, f)$

对于所有的$r$成立, 除去一个对数测度为有穷的集合.

显然, 对任意的$c\neq 0, $$r\rightarrow\infty$时, 对一般亚纯函数$f(z)$下列不等式成立

$ (1+o(1))N(r-|c|, \frac{1}{f(z)})\leq N(r, \frac{1}{f(z+c)})\leq (1+o(1))N(r+|c|, \frac{1}{f(z)}).$

结合引理2.1, 对于计数函数有下列关系.

引理2.3 [1]假设$f(z)$是超级小于1的亚纯函数, $c\in\mathbb{C}\backslash\{0\}$, 那么

$ N(r, \frac{1}{f(z+c)})= N(r, \frac{1}{f(z)})+S(r, f(z)).$

引理2.4 [17]假设$f(z)$是超级小于1的亚纯函数, $c\in\mathbb{C}\backslash\{0\}$, 那么有

$ T(r, f(z+c))=T(r, f)+S(r, f).$

引理2.5 [3, 7]假定$f(z)$是一个亚纯函数, $ \Psi(z):=a_{n}f^{n}(z)+\cdots+a_{0}(z)$满足$a_{n}\neq 0, T(r, a_{j})=S(r, f)$.更进一步地, 假设$\overline{N}(r, \frac{1}{\Psi})+\overline{N}(r, f)=S(r, f), $那么有$\Psi=a_{n}(f+\frac{a_{n-1}}{na_{n}})^{n}.$

引理2.6 [2]$q(z)$是多项式, $L(r, f)$是关于$f(z)$或者它的导数, 变换的线性微分-差分多项式, 那么方程$f(z)^{2}+q(z)L(r, f)=0$没有有穷级的超越整函数解.

引理2.7 [14]假定$q$是一个正整数, $a_{0}(z), \cdots, a_{n}(z)$或者是度小于$q$的指数多项式, 或者是关于$z$的一般多项式, $b_{1}, \cdots, b_{n}\in\mathbb{C}\backslash\{0\}$是互不相等的常数, 那么

$\sum\limits_{j=1}^{n}a_{j}(z)e^{b_{j}z^{q}}=a_{0}(z)$

当且仅当$a_{0}(z)\equiv\cdots\equiv a_{n}(z) \equiv 0$.

$ W\subseteq \mathbb{C}$, 所有包含$W$的凸集的交定义为$W$的凸包, 记为co($W$).如果$W$包含有穷多个元素, 那么co($W$)可以直接通过有穷多个闭半平面的交得到.因此, co($W$)或者是一个紧polygon集, 或者是一条线段.用$C$(co$(W)$)表示co($W$)的直径.如果co($W$)是一条线段, 那么$C$(co$(W)$)是这条线段长度的2倍.在本文中记$W=\{\overline{w}_{1}, \cdots, \overline{w}_{m}\}$, $W_{0}=\{0, \overline{w}_{1}, \cdots, \overline{w}_{m}\}$.

引理2.8 [18]假定$f(z)$具有(1.6)式形式, 那么

$ T(r, f)=C({\rm co}(W_{0}))\frac{r^{q}}{2\pi}+o(r^{q}).$

注4  假设$f(z)$是具有(1.6)形式的指数多项式解且$m\geq1$, 那么由引理2.8可知, 当$\rightarrow\infty$时,

$ \frac{S(r, f)}{r^{q}}=\frac{S(r, f)}{ T(r, f)} \cdot \frac{ T(r, f)}{r^{q}}\rightarrow 0$

$ \frac{o(r^{q})}{T(r, f)}=\frac{o(r^{q})}{ r^{q}} \cdot \frac{ r^{q}}{T(r, f)}\rightarrow 0.$

换句话说$S(r, f)=o(r^{q}), ~o(r^{q})=S(r, f)$.

引理2.9 [18]假定$f(z)$具有(1.6)式形式, 如果$H_{0}(z)\not\equiv0$, 那么$ m(r, \frac{1}{f})=o(r^{q}).$如果$H_{0}(z)\equiv 0$, 那么

$ N(r, \frac{1}{f})= C({\rm co}(W_{0}))\frac{r^{q}}{2\pi}+o(r^{q}).$

利用文献[2, 推论2.6, 引理2.7]完全类似的方法, 可得下述引理2.10和引理2.11.

引理2.10  假定$f(z)$具有(1.6)式形式, 对任意的$ i\neq j, ~w_{i}\neq 2w_{j}$$f(z)$是方程(1.8)的解.如果点$0, w_{i}, \cdots, w_{n}$是线性的, 那么$m=1$.

引理2.11  当$m\geq2$时, 如果对任意的$ i\neq j, ~w_{i}\neq 2w_{j}$, 那么具有(1.6)式形式的$f(z)$不是方程(1.8)的解.

引理2.12  假定$f(z)$具有(1.6)式形式, 其中$m=1$.如果$f(z)$是方程(1.8)的解, 那么$f\in \Gamma'_{1}$.

  令$f(z)=H_{0}(z)+H_{1}(z)e^{w_{1}z^{q}}$.要证$H_{0}(z), ~H_{1}(z)$是多项式, 把$f(z)$代入方程(1.8)得

$ \begin{eqnarray} &&P(z)-H^{2}_{0}(z)-a_{1}H_{0}(z) =-q(z)e^{Q(z)}f^{(k)}(z+c)\nonumber\\ &=&H^{2}_{1}(z)e^{2w_{1}z^{q}}+(2H_{0}(z)+a_{1})H_{1}(z)e^{w_{1}z^{q}}+q(z)e^{Q_{0}(z)}H^{(k)}_{0}(z+c)e^{b_{q}z^{q}} \nonumber\\ &&+q(z)H^T_{1}(z+c)e^{Q_{0}(z)+P_{1}(z)}e^{(b_{q}+w_{1})z^{q}}, \end{eqnarray} $ (2.1)

其中$Q_{0}(z)=Q(z)-b_{q}z^{q}, ~P_{1}(z)=w_{1}(z+c)^{q}$是次数$\leq q-1$的多项式, $ H^T_{1}(z+c)$是关于$ H_{1}(z+c), ~w_{1}(z+c)^{q}$和他们导数的微分多项式.分三种情况来讨论.

情况1  假设$b_{q}\neq \pm w_{1}$, 对(2.1)式由引理2.7得

$ q(z)H^T_{1}(z+c)e^{Q_{0}(z)+P_{1}(z)}e^{(b_{q}+w_{1})z^{q}} \equiv 0. $

$ H^T_{1}(z+c)\equiv 0$, 而$ H^T_{1}(z+c)$是关于$ H_{1}(z+c), ~w_{1}(z+c)^{q}$和其导数的微分多项式.所以$H_{1}(z)\equiv 0$, 矛盾.

情况2  假设$b_{q}= - w_{1}$, 同样对(2.1)式由引理2.7得$H_{1}(z)\equiv 0$, 矛盾.

情况3  $b_{q}= w_{1}$, 那么(2.1)式可改写为

$ \begin{eqnarray}P(z)-H^{2}_{0}(z)-a_{1}H_{0}(z)&=&(H^{2}_{1}(z)+q(z)H^T_{1}(z+c)e^{Q_{0}(z)+P_{1}(z)})e^{2w_{1}z^{q}}\\ &&+((2H_{0}(z)+a_{1})H_{1}(z)+q(z)e^{Q_{0}(z)}H^{(k)}_{0}(z+c))e^{w_{1}z^{q}}.\end{eqnarray}$

如果$\deg(H_{0}(z))\geq k$, 那么由(2.1)式和由引理2.7得

$ \begin{eqnarray}&& H^{2}_{1}(z)+q(z)H^T_{1}(z+c)e^{Q_{0}(z)+P_{1}(z)}\equiv 0, \end{eqnarray} $ (2.2)
$ \begin{eqnarray} (2H_{0}(z)+a_{1})H_{1}(z)+q(z)e^{Q_{0}(z)}H^{(k)}_{0}(z+c)\equiv 0, \end{eqnarray} $ (2.3)
$ \begin{eqnarray} P(z)-H^{2}_{0}(z)-a_{1}H_{0}(z)\equiv 0. \end{eqnarray} $ (2.4)

由(2.4)式知$H_{0}(z)$是一个多项式.下证$H_{1}(z)$也是多项式.

(a) 当$\deg(H_{0}(z))\geq k$时, 则$ H^{(k)}_{0}(z+c)\not\equiv0 $.如果$\deg(Q_{0}(z)+P_{1}(z))=0$, 反证设$H_{1}(z)$是一个超越整函数.由差分对数导数引理及文献[16, 定理1.24 ]知

$ N(r, \frac{1}{f^{(k)}(z+c)})\leq N(r, \frac{1}{f(z+c)})+k\overline{N}(r, f(z+c))+S(r, f(z+c)).$

再由上式, (2.2)式和引理2.4得

$ \begin{align*} 2T(r, H_{1}(z))\leq &T(r, H^T_{1}(z+c))+T(r, q(z)e^{Q_{0}(z)+P_{1}(z)})\\ \leq{}&m(r, H^T_{1}(z+c))+N(r, H^T_{1}(z+c)) +S(r, H_{1}(z))\\ \leq{}& T(r, H_{1}(z)) +S(r, H_{1}(z)). \end{align*} $

上式与假设矛盾, 所以$H_{1}(z)$是一个多项式.由(2.3)式知$ Q_{0}(z)$是一个常数.因此, $ P_{1}(z)$也是一个常数.可得$q=1, f\in \Gamma'_{1}$.

如果$\deg(Q_{0}(z)+P_{1}(z))\geq 1$, 由定理1.2 (ⅲ)知$H_{1}(z)\in\Gamma'_{0}$, 又因为$H_{0}(z)$是多项式, 所以$f\in \Gamma'_{1}$.

(b) 当$\deg(H_{0}(z))< k$时, 则$ H^{(k)}_{0}(z+c)\equiv 0 $.由(2.3)式知$(2H_{0}(z)+a_{1})H_{1}(z)= 0.$如果$H_{1}(z)= 0$, 那么$f(z)$是一个多项式, 与已知矛盾.如果$ 2H_{0}(z)+a_{1}= 0 $, 则$ H_{0}(z)$是常数.如果$\deg(Q_{0}(z)+P_{1}(z))= 0$, 由引理2.6知$H_{1}(z)$是一个多项式.如果$\deg(Q_{0}(z)+P_{1}(z))\geq 1$, 由定理1.2 (ⅲ)知$H_{1}(z)\in\Gamma'_{0}$, 所以$f\in \Gamma'_{1}$.

3 定理的证明

定理1.4的证明   (ⅰ)假设$f(z)$是方程(1.7)的有穷级非零整函数解, 由(1.7)式和引理2.7知$f(z)$是超越的.

一方面, 由引理2.2和(1.7)式及差分对数导数引理有

$ \begin{align*} nT(r, f)+S(r, f)&=m(r, f^{n}+\cdots+a_{1} f)~=m(r, P(z)-q(z)e^{Q(z)}f^{(k)}(z+c)) \\&\leq m(r, P(z))+m(r, q(z))+m(r, e^{Q(z)})+m(r, f^{(k)}(z+c))+O(1) \\&\leq m(r, e^{Q(z)})+m(r, \frac{f^{(k)}(z+c)}{f(z+c)})+m(r, f(z+c))+S(r, f) \\&\leq T(r, e^{Q(z)})+T(r, f(z+c)+S(r, f) \\&\leq T(r, e^{Q(z)})+T(r, f)+S(r, f). \end{align*} $

所以

$ \begin{eqnarray} (n-1)T(r, f)\leq T(r, e^{Q(z)})+S(r, f). \end{eqnarray}$ (3.1)

如果$\sigma(f)< \deg(Q(z))$, 则与方程(1.7)矛盾, 所以有$ \sigma(f)=\deg(Q(z))~~(n\geq 2)$.由型的定义得

$ \tau (f)=\limsup\limits_{r\rightarrow \infty}\frac{T(r, f)}{r^{\deg(Q(z))}}\in (0, \infty).$

所以$f(z)$是正规型的.

(ⅱ)假设$f(z)$是方程(1.7)的有穷级非零整函数解, 由定理1.4(ⅰ)知, $f(z)$是超越的.下面要证:如果有穷复数$a$$f(z)$的Borel例外值, 那么有$a=-\frac{a_{n-1}}{n}$.由于$\lambda(f-a)<\sigma (f), ~f(z)$是整函数, 则$f(z)$是正规增长.令$\lambda(f-a)<\alpha<\beta<\sigma (f), $即有

$ \limsup\limits_{r\rightarrow\infty} \frac{\log^{+}N(r, \frac{1}{f-a})}{\log r}<\alpha<\beta<\lim\limits_{r\rightarrow\infty}\frac{\log^{+}T(r, f)}{\log r}=\sigma(f).$

从而有

$ \frac{N(r, \frac{1}{f-a})}{T(r, f)}\leq\frac{r^{\alpha}}{r^{\beta}}\rightarrow 0~~~(r\rightarrow \infty).$

因此$N(r, \frac{1}{f-a})=S(r, f)$. (1.7)式可改写为

$ \begin{eqnarray}G(z)=f(z)^{n}+a_{n-1}f(z)^{n-1}+\cdots+a_{1}f(z)-P(z)= -q(z)e^{Q(z)}f^{(k)}(z+c). \end{eqnarray}$ (3.2)

由引理2.3和文献[6, 定理1.24]得

$ \begin{eqnarray} N(r, \frac{1}{G(z)})&=&N(r, \frac{1}{q(z)f^{(k)}(z+c) }) \nonumber\\&\leq& N(r, \frac{1}{q(z)}) +N(r, \frac{1}{f^{(k)}(z+c) }) \nonumber\\&=& N(r, \frac{1}{q(z)}) +N(r, \frac{1}{[f(z+c)-a ]^{(k)} }) \nonumber\\&\leq& N(r, \frac{1}{f(z+c)-a}) +k\overline{N}(r, f(z+c)-a)+S(r, f(z+c)-a)+S(r, f) \nonumber\\&\leq& N(r, \frac{1}{f(z+c)-a}) +S(r, f(z+c)) \nonumber\\&\leq& N(r, \frac{1}{f(z)-a}) +S(r, f) = S(r, f). \end{eqnarray}$ (3.3)

由于$f(z)$是整函数, 由引理2.5得

$ G(z)=(f+\frac{a_{n-1}}{n})^{n}=f(z)^{n}+a_{n-1}f(z)^{n-1}+\cdots+a_{1}f(z)-P(z), $ (3.4)

并且有

$ a_{i+1}=C^{i}_{n-1}(\frac{a_{n-1}}{n-1})^{n-1-i}, i=0, 1, \cdots, n-2~;~~~~ C^{i}_{n-1}=\frac{(n-1)!}{i!(n-1-i)!} .$ (3.5)

反证:假如$a\neq -\frac{a_{n-1}}{n}$.那么由(3.4)和(3.5)式得

$ \overline{N}(r, \frac{1}{f+\frac{a_{n-1}}{n}})=\overline{N}(r, \frac{1}{G(z)})=S(r, f).$

又由Nevanlinna第二基本定理得

$ T(r, f)\leq \overline{N}(r, \frac{1}{f+\frac{a_{n-1}}{n}})+\overline{N}(r, \frac{1}{f(z)-a})=S(r, f), $

与已知矛盾, 所以$a=-\frac{a_{n-1}}{n}$.特别地, 当$a=0$时, 有$a_{n-1}=0$.又有(3.5)式得$a_{n-1}=\cdots=a_{1}=0\equiv P(z)$.

(ⅲ)假设$f(z)$是方程(1.7)的有穷级非零整函数解, 由定理1.4 (ⅰ)知, $f(z)$是超越的.由$ P(z)\equiv 0 $

$ H(z):=f(z)^{n-1}+a_{n-1}f(z)^{n-2}+\cdots+a_{1}= -q(z)e^{Q(z)}\frac{f^{(k)}(z+c)}{f(z)}. $ (3.6)

由引理2.2和差分对数导数引理得

$ m(r, \frac{f^{(k)}(z+c)}{f(z)})=S(r, f), $ (3.7)

又由(3.6)式得

$ N(r, \frac{f^{(k)}(z+c)}{f(z)})\leq N(r, \frac{1}{q(z)})=S(r, f), $ (3.8)

即有

$ T(r, \frac{f^{(k)}(z+c)}{f(z)})=S(r, f). $ (3.9)

因此, 由(3.6)式和(3.9)式得

$ \begin{align*} \overline{N}(r, \frac{1}{ H(z)})+\overline{N}(r, f(z))& \leq N(r, \frac{1}{\frac{f^{(k)}(z+c)}{f(z)}})+ N(r, \frac{1}{q(z)})+N(r, f) \\ &\leq T(r, \frac{f^{(k)}(z+c)}{f(z)})+S(r, f)= S(r, f) . \end{align*} $

因此, 由引理2.5, 得到(3.6)式可变形为

$ (f+\frac{a_{n-1}}{n-1})^{n-1}=-q(z)e^{Q(z)}\frac{f^{(k)}(z+c)}{f(z)}, $ (3.10)

并且有

$ a_{i+1}=C^{i}_{n-1}(\frac{a_{n-1}}{n-1})^{n-1-i}, i=0, 1, \cdots, n-2 ~;~~~~~C^{i}_{n-1}=\frac{(n-1)!}{i!(n-1-i)!} .$ (3.11)

情形1  如果对任意的$j\in \{1, 2, \cdots, n-1\}$都有$ a_{j}\neq 0~ (j=1, 2, \cdots, n-1)$.

一方面

$ \begin{eqnarray} &&\overline{N}(r, \frac{1}{f+\frac{a_{n-1}}{n-1}})\leq N(r, \frac{1}{\frac{f^{(k)}(z+c)}{f(z)}})+ N(r, \frac{1}{q(z)}), \\ && T(r, \frac{f^{(k)}(z+c)}{f(z)})+ N(r, \frac{1}{q(z)}) = S(r, f).\end{eqnarray}$

另一方面, 由Nevanlinna第二基本定理

$ \begin{align*} T(r, f)& \leq \overline{N}(r, \frac{1}{f+\frac{a_{n-1}}{n-1}})+\overline{N}(r, \frac{1}{f(z)})+\overline{N}(r, f(z)) \\ &=\overline{N}(r, \frac{1}{f(z)})+ S(r, f) . \end{align*} $

因此 $\overline{\lambda}(f)= \sigma (f).$

情形2  如果存在$ i_{0}\in \{1, 2, \cdots, n-1\}$使得$a_{i_{0}}=0$.由(3.11)式得$a_{i_{0}}=a_{n-1}=\cdots=a_{i_{0}+1}=a_{i_{0}-1}=\cdots=a_{1}=0$.再由定理1.2(ⅱ)知$\lambda (f)<\sigma(f)$.

(ⅳ)假设$f(z)$是方程(1.7)的有穷级非零整函数解.

充分性  假设$P(z)\equiv 0 $且存在$ i_{0}\in \{1, 2, \cdots, n-1\} $使得$ a_{i_{0}}=0$.由定理1.2 (ⅲ)知$f\in \Gamma'_{0}$.

必要性  由于$f\in \Gamma'_{0}$, 可设$f(z)=A(z)e^{\alpha(z)}, A(z), \alpha(z)$是多项式且满足$\sigma(A(z))=\lambda(f)< \sigma(f)=\deg(Q(z))$.则$0$$f(z)$的一个Borel例外值, 由定理1.4 (ⅱ)知$a_{n-1}=\cdots=a_{1}=0\equiv P(z)$.

(ⅴ)假设$f(z)$是方程(1.7)的有穷级非零整函数解, 由定理1.4 (ⅰ)知$f$是超越的.由已知必有$P(z)\equiv 0$.否则, 如果$P(z)\not\equiv 0$, 由引理2.3和已知${\rm card}\{z:P_{1}(z)=P'_{1}(z)=P''_{1}(z)=0\}\geq 1$或者${\rm card}\{z:P_{1}(z)=P'_{1}(z)=0\}\geq 2, $有不等式

$ \begin{align*} nT(r, f)=&T(r, f^{n}+a_{n-1}f^{n-1}+\cdots+a_{1}f) \\\leq& \overline{N}(r, \frac{1}{f^{n}+a_{n-1}f^{n-1}+\cdots+a_{1}f-P(z)})+\overline{N}(r, \frac{1}{f^{n}+a_{n-1}f^{n-1}+\cdots+a_{1}f}) \\&+\overline{N}(r, f^{n}+a_{n-1}f^{n-1}+\cdots+a_{1}f) \\\leq& N(r, \frac{1}{q(z){f^{(k)}(z+c)}})+(n-2)T(r, f) \\\leq& N(r, \frac{1}{q(z)})+N(r, \frac{1}{{f^{(k)}(z+c)}})+(n-2)T(r, f) \\\leq& N(r, \frac{1}{{f(z+c)}})+k\overline{N}(r, f(z+c))+S(r, f(z+c))+(n-2)T(r, f)+S(r, f) \\\leq& N(r, \frac{1}{{f(z+c)}})+S(r, f(z+c))+(n-2)T(r, f) \\\leq& N(r, \frac{1}{{f(z)}})+S(r, f)+(n-2)T(r, f) \\\leq& (n-1)T(r, f)+S(r, f), \end{align*} $

$T(r, f)=S(r, f)$, 矛盾, 所以$P(z)\equiv 0$.又因为存在$ i_{0}\in \{1, 2, \cdots, n-1\}$使得$a_{i_{0}}=0$.由定理1.4 (ⅳ)可得$P(z)\equiv 0=a_{1}(z)=\cdots=a_{n-1} $$ f(z)\in\Gamma'_{0}$.

(vi) 如果$f, g \in\Gamma'_{0}$且是方程(1.7)的有穷级整函数解, 由定理1.4 (ⅳ)得$P(z)\equiv 0$$a_{n-1}=\cdots=a_{1}=0$.又由定理1.2 (ⅳ)知$f=\eta g$$\eta^{n-1}=1$.

定理1.5的证明

由引理2.10, 引理2.11, 引理2.12可得定理1.5的证明.

定理1.6的证明

(ⅰ)假设$f(z)$是方程(1.9)的有穷级非零整函数解, 由(1.9)式得

$ \frac{f(z)-P(z)}{f^{(k)}(z+c)}=-q(z)e^{Q(z)}.$

由增长级的性质和引理2.4得$ \deg(Q(z))\leq \max\{\sigma(f(z)-P(z)), \sigma(f^{(k)}(z+c))\}=\sigma(f).$

(ⅱ)由于$\sigma(f)\geq \deg(Q(z))$$Q(z)$是非常数多项式, 可得$f(z)$是超越的.否则有$\lambda(f)<\sigma(f)$.因此$f(z)$是正规增长的, 有

$ N(r, \frac{1}{f(z)})=S(r, f). $ (3.12)

又(1.9)式可变形为$ f(z)-P(z)=-q(z)e^{Q(z)}f^{(k)}(z+c).$由引理2.3, (3.12)式和文献[6, 定理1.24]得

$ \begin{eqnarray} N(r, \frac{1}{f(z)-P(z)})&=&N(r, \frac{1}{q(z)f^{(k)}(z+c)})\nonumber\\ &\leq& N(r, \frac{1}{q(z)})+N(r, \frac{1}{f^{(k)}(z+c)})\nonumber\\ &\leq& N(r, \frac{1}{f(z+c)})+k\overline{N}(r, f(z+c))+S(r, f(z+c))+S(r, f)\nonumber\\ &\leq& N(r, \frac{1}{f(z+c)})+S(r, f(z+c))+S(r, f)\nonumber\\ &\leq& N(r, \frac{1}{f(z)})+S(r, f)=S(r, f). \end{eqnarray} $ (3.13)

如果$P(z)\not\equiv 0$, 由(3.12), (3.13)式和Nevanlinna第二基本定理得

$ T(r, f)\leq \overline{N}(r, \frac{1}{f(z)})+\overline{N}(r, \frac{1}{f(z)-P(z)})+S(r, f)=S(r, f), $

与假设矛盾, 所以有$\lambda(f)=\sigma(f)$.

(ⅲ)反证法:假设$f(z)$是方程(1.9)的具(1.6)形式的指数多项式解, 满足$q=\deg(Q(z)).$把(1.9)式改写为

$ \frac{1}{q(z)}\cdot \frac{f(z)}{f^{(k)}(z+c)}=-e^{Q(z)}. $ (3.14)

由(3.14)式, 引理2.8, 注4, 差分对数导数引理得

$ \begin{align*} &\frac{|b_{q}|}{\pi}+o(r^{q}) =T(r, e^{Q(z)})= m(r, e^{Q(z)})\\ \leq& m(r, \frac{1}{q(z)})+m(r, \frac{f(z)}{f^{(k)}(z+c)}) =S(r, f)=o(r^{q}), \end{align*} $

矛盾.

参考文献
[1] Li N, Yang L Z. Solutions of nonlinear difference equations[J]. J. Math. Anal. Appl., 2017, 452: 1128–1144. DOI:10.1016/j.jmaa.2017.03.047
[2] Liu K. Exponential polynomials as solutions of differential-difference equations of certain types[J]. J. Med. Math., 2016, 13: 3015–3027.
[3] Hayman W K. Meromorphic functions[M]. Oxford: Clarendon Press, 1964.
[4] Laine I. Nevanlinna theory and complex differential equations[M]. Berlin: Walter de Gruyter, 1993.
[5] Yang L. Value distribution theory[M]. Berlin: Springer-Verlag, 1993.
[6] Yi H X, Yang C C. Uniqueness theory of meromorphic functions[M]. Dutch: Kluwer Academic Publishers, 2003.
[7] Chiang Y M, Feng S J. On the growth of logarithmic differences, difference quotients and logarithmic derivatives of meromorphic functions[J]. Trans. Amer. Math. Soc., 2009, 361(7): 3767–3791. DOI:10.1090/S0002-9947-09-04663-7
[8] Halburd R G, Korhonen R J. Difference analogue of the lemma on the logarithmic derivative with applications to difference equations[J]. J. Math. Anal. Appl., 2006, 314(2): 477–487. DOI:10.1016/j.jmaa.2005.04.010
[9] Halburd R G, Korhonen R J. Nevanlinna theory for the difference operator[J]. Ann. Acad. Sci. Fenn. Math., 2006, 31: 463–478.
[10] Liu K, Cao T, Cao H. Entire solutions of Fermat type differential-difference equations[J]. Arch. Math. (Basel), 2012, 99(2): 147–155. DOI:10.1007/s00013-012-0408-9
[11] Liu K, Yang L Z. On entire solutions of some complex differential-difference equations[J]. Comput. Meth. Funct. Theory., 2013, 13: 433–447. DOI:10.1007/s40315-013-0030-2
[12] Liu K, Dong X J. Some results related to complex differential-difference equations of certain types[J]. Bull. Korean. Math. Soc., 2014, 51(5): 1453–1467. DOI:10.4134/BKMS.2014.51.5.1453
[13] Yang C C, Laine I. On analogies between nonlinear difference and differential equations[J]. Proc. Japan. Acad. Ser. A. Math. Sci., 2010, 86(1): 10–14. DOI:10.3792/pjaa.86.10
[14] Wen Z T, Heittokangas J, Laine I. Exponential polynomials as solutions of certain nonlinear differential equations[J]. Acta. Math. Sin., 2012, 28(7): 1295–1306. DOI:10.1007/s10114-012-1484-2
[15] Halburd R G, Korhonen R J, Tohge K. Holomorphic curves with shift-invariant hyperplane preimages[J]. Trans. Amer. Math. Soc., 2014, 366: 4267–4298. DOI:10.1090/tran/2014-366-08
[16] Chen Z X. Complex difference and difference equations[M]. Beijing: Science Press, 2014.
[17] Korhonen R J. An extension of Picard theorem for meromorphic functions of small hyper-order[J]. J. Math. Anal. Appl., 2009, 357: 244–253. DOI:10.1016/j.jmaa.2009.04.011
[18] Steinmetz N. Zur werverteilung von exponential polynomal[J]. Manuscripta. Math., 1978, 26(1-2): 155–167. DOI:10.1007/BF01167971