数学物理学报  2018, Vol. 38 Issue (2): 291-303   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
张淑琴
胡雷
含有变量阶导数及一个参数的非线性分数阶微分方程初值问题
张淑琴1, 胡雷2     
1. 中国矿业大学(北京) 理学院 北京 100083;
2. 山东交通学院理学院 济南 250357
摘要:该文利用Schauder不动点定理研究了含有变量阶导数的非线性分数阶微分方程初值问题解的存在性.
关键词变量阶导数与积分    初值问题    存在性    不动点定理    
Initial Value Problem for Nonlinear Fractional Differential Equations with Variable-order Derivative and a Parameter
Zhang Shuqin1, Hu Lei2     
1. Department of Mathematics, China University of Mining & Technology, Beijing 100083;
2. School of Science, Shandong Jiaotong University, Jinan 250357
Abstract: In this work, we study the existence result of solution to an initial value problem for a differential equation involving with variable-order derivative. Our analysis relies on the Schauder fixed-point theorem and some analysis technique.
Key words: Derivatives and integrals of variable-order     Initial value problem     Existence of solution     Fixed-point theorem    
1 引言

近年来, 分数阶微分方程在物理, 化学, 工程中得到了广泛应用, 分数阶微分方程的理论研究获得了快速的发展.科研工作者利用泛函分析的技巧, 如Banach压缩映像原理, Krasnoselskii's不动点定理, Schauder不动点定理获得了很多分数阶初边值问题解的存在性结果, 如文献[1-3, 18, 23-24].

2008年, Lakshmikantham[19]等, 研究了下列分数阶微分方程初值问题

$ \left\{ \begin{array}{ll} D_{0+}^q x(t)=f(t, x(t)), t\in [0, T], \\ x(0)=x_0, \end{array}\right. $

其中$0 < q\leq1$, $D_{0+}^{q}$表示Riemann-Liouville分数阶导数.作者利用经典的分析方法获得了上述微分方程解的存在性.

2011年, 寇春海[20]研究了下列在右半轴上的分数阶初值问题

$ \left\{ \begin{array}{ll} D_{0+}^\alpha x(t)=f(t, x(t)), t\in (0, +\infty), \\ \lim\limits_\limits{t\rightarrow 0^+}t^{1-\alpha}x(t)=u_0, \end{array}\right. $

其中$0 < \alpha\leq1$, $D_{0+}^{\alpha}$表示Riemann-Liouville分数阶导数, $f:(0, +\infty)\times {\Bbb R} \rightarrow {\Bbb R}$.作者构造了一个新的Banach空间并利用不动点定理获得了上述方程解的存在性结果.

2012年, Baleanu[21]研究了如下分数阶初值问题

$ \left\{ \begin{array}{ll} D_{0+}^\alpha x(t)+ D_{0+}^{\beta} x(t)=f(t, x(t)), \ \ 0<t < 1, \\ x(0)=0, \end{array}\right. $

其中$0 < \alpha < \beta < 1$, $D_{0+}^{\alpha}, D_{0+}^{\beta}$表示Riemann-Liouville分数阶导数, $f:[0, 1]\times [0, +\infty)$是连续的.作者利用锥上的不动点定理得到了解与多个正解的存在性结果.

近年来, 变量阶分数次导数在工程, 物理, 金融, 水文等领域中发挥了越来越重要的作用, 试验证明它能够更好地模拟现实问题.与此同时, 变量阶的分数次微分方程受到了诸多学者的关注, 其中对于含有变量阶导数的初值问题解的存在性研究也取得了一些成果, 如文献[4-16].

2015年, 李秀英[22]研究了如下分数阶泛函边值问题的数值解

$ \left\{ \begin{array}{ll} {{D}^{\alpha (x)}}u+a(x){u}'(x)+b(x)u(\tau (x))=f(x),  0\leq x\leq 1, \\ u(0)={{\mu }_{1}}, ~u(1)={{\mu }_{2}}, \end{array}\right. $

其中$1 < \alpha (x)\leq 2, {\mu }_{1}, {\mu }_{2}$是常数, $D_{0+}^{\alpha (x)}$表示变量阶Caputo导数, $\tau (x)\in C^1[0, 1], a(x), b(x)\in W^1[0, 1], u(x)\in W^4[0, 1], \tau (x)$大于或小于$x$. $D_{0+}^{\alpha (x)}$定义为

$ D_{0+}^{\alpha (x)}u(x)=\frac{1}{\Gamma (2-\alpha (x))}\int_{0}^{x}{{{(x-t)}^{1-\alpha (x)}}{u}''(t){\rm d}t}. $

虽然分数阶微分方程解的存在性的研究结果已经很多, 但鲜有文章涉及变量阶分数次微分方程解的存在性的研究.受前面文献的启发, 此文研究如下带有变量阶导数的分数次微分方程初值问题

$ \left\{ \begin{array}{ll} D_{0+}^\alpha x(t)+\lambda D_{0+}^{p(t)} x(t)=f(t, x(t), D_{0+}^\beta x(t)), \ \ 0 <t\leq T<+\infty, \\ x(0)=0, \end{array}\right. $ (1.1)

这里$0 < \beta < \alpha < 1$, $0 < p(t) < \alpha-\beta$, $0\leq t\leq T < +\infty$, $\lambda\in {\Bbb R}\backslash\{0\}$, 变量阶导数$D_{0+}^{p(t)}$定义为

$ D_{0+}^{p(t)}x(t)=\frac {\rm d}{{\rm d}t}\int_0^t\frac{(t-s)^{-p(s)}} {\Gamma(1-p(s))}x(s){\rm d}s, t>0, $ (1.2)

$ I_{0+}^{1-p(t)}x(t)=\int_0^t\frac{(t-s)^{-p(s)}}{\Gamma(1-p(s))} x(s){\rm d}s, t>0. $ (1.3)

上式表示指标为$1-p(t)$的变量阶积分, $D_{0+}^\alpha, D_{0+}^\beta$是Riemann-Liouville分数次导数.

很显然常微分方程或分数阶微分方程是变量阶微分方程的特殊情形.由(1.2)和(1.3)式知, 当$p(t)$是一个常值函数时, 即$p(t)\equiv q$ ($q$是一个正常数), $I_{0+}^{p(t)}, D_{0+}^{p(t)}$便是通常意义下的Riemann-Liouville分数阶积分和导数[17].

该文安排如下.第二章给出变量阶分数次导数和积分的某些性质.在第三章中, 利用Schauder不动点定理, 我们得到了初值问题(1.1)解的存在性结果.

2 一些结论

首先我们给出分数阶算子的一些基本性质.

命题2.1[17]  如果$f\in L(0, b), 0 < b < +\infty$, 则$I_{0+}^\gamma I_{0+}^\delta f(t)=I_{0+}^{\gamma+\delta}f(t)$, $\gamma>0, \delta>0$.

命题2.2[17]  如果$f\in L(0, b), 0 < b < +\infty$, 则$D_{0+}^\gamma I_{0+}^\gamma f(t)=f(t)$, $\gamma>0$.

命题2.3[17]  设$0 < \alpha\leq 1$, 则微分方程

$ D_{0+}^\alpha u=0 $

有解

$ u(t)=ct^{\alpha-1}, c\in {\Bbb R}. $

命题2.4[17]  若$0 < \alpha\leq 1$, $u(t)\in L(0, b)$, $D_{0+}^\alpha u\in L(0, b)$, 则

$ I_{0+}^\alpha D_{0+}^\alpha u(t)=u(t)+ct^{\alpha-1}, c\in {\Bbb R}. $

命题2.5[17]   Riemann-Liouville分数阶积分算子$I_{0+}^\delta : C[0, b]\rightarrow C[0, b], \delta>0$是有界的.

上述分数阶算子的性质是否仍适用于变量阶分数次算子是值得我们研究的.以命题2.1为例, 我们考虑一种特殊的情形: $f(t)\equiv 1, t\in [0, T]$.接下来, 我们将验证$I_{0+}^{p(t)}I_{0+}^{q(t)}f(t)$$I_{0+}^{p(t)+q(t)}f(t)$是否相等.

首先, 根据(1.3)式, 有

$ \begin{eqnarray*} I_{0+}^{p(t)}I_{0+}^{q(t)}f(t)&=&\int_0^t\frac{(t-s)^{p(s)-1}}{\Gamma(p(s))} \int_0^s\frac{(s-\tau)^{q(\tau)-1}}{\Gamma(q(\tau))}f(\tau){\rm d}\tau {\rm d}s\\ &=&\int_0^t\frac{(t-s)^{p(s)-1}}{\Gamma(p(s))} \int_0^s\frac{(s-\tau)^{q(\tau)-1}}{\Gamma(q(\tau))}{\rm d}\tau{\rm d}s\\ &=&\int_0^t\frac{(t-s)^{p(s)-1}}{\Gamma(p(s))} \int_0^1\frac{s^{q(s\mu)}(1-\mu)^{q(s\mu)-1}}{\Gamma(q(s\mu))}{\rm d}\mu {\rm d}s\\ &=&\int_0^1\int_0^1\frac{t^{p(tr)}(1-r)^{p(tr)-1}}{\Gamma(p(tr))} \frac{(tr)^{q(tr\mu)}(1-\mu)^{q(t\mu r)-1}}{\Gamma(q(t\mu r))}{\rm d}\mu {\rm d}r\\ & =&\int_0^1\int_0^1\frac{t^{p(tr)+q(t\mu r)}(1-r)^{p(tr)-1}}{\Gamma(p(tr))} \frac{r^{q(tr\mu)}(1-\mu)^{q(t\mu r)-1}}{\Gamma(q(t\mu r))}{\rm d}\mu {\rm d}r. \end{eqnarray*} $

另一方面

$ \begin{eqnarray*} I_{0+}^{p(t)+q(t)}f(t)=\int_0^t\frac{(t-s)^{p(s)+q(s)-1}}{\Gamma(p(s)+q(s))}f(s){\rm d}s=\int_0^1 \frac{t^{p(tr)+q(tr)}(1-r)^{p(tr)+q(tr)-1}}{\Gamma(p(tr)+q(tr))}{\rm d}r. \end{eqnarray*} $

从而, 对$f(t)\equiv 1$, 仅当

$ \int_0^1\frac{t^{p(tr)+q(t\mu r)}(1-r)^{p(tr)-1}}{\Gamma(p(tr))} \frac{r^{q(tr\mu)}(1-\mu)^{q(t\mu r)-1}}{\Gamma(q(t\mu r))}{\rm d}\mu =\frac{t^{p(tr)+q(tr)}(1-r)^{p(tr)+q(tr)-1}}{\Gamma(p(tr)+q(tr))}, $

才有下式成立

$ I_{0+}^{p(t)}I_{0+}^{q(t)}f(t)=I_{0+}^{p(t)+q(t)}f(t). $

然而, 我们不能确定上式的正确性.我们仅知

$ \int_0^1\frac{t^{p(tr)+q(t r)}(1-r)^{p(tr)-q(tr)-1}}{\Gamma(p(tr))} \frac{\mu^{p(tr)-1}(1-\mu)^{q(tr)-1}}{\Gamma(q(tr))}{\rm d}\mu =\frac{t^{p(tr)+q(tr)}(1-r)^{p(tr)+q(tr)-1}}{\Gamma(p(tr)+q(tr))}. $

因此, 对于$p(t), q(t), f(t), t\in[0, T]$, 对于$[0, T]$中的某些点, 我们有

$ I_{0+}^{p(t)}I_{0+}^{q(t)}f(t)\neq I_{0+}^{p(t)+q(t)}f(t). $ (2.1)

特别地, 对于一般的函数$0 < p(t) < 1, f(t)$, 对于$[0, T]$中的某些点, 有

$ I_{0+}^{p(t)}I_{0+}^{1-p(t)}f(t)\neq I_{0+}^{p(t)+1-p(t)}f(t)=I_{0+}^1f(t). $

下面, 我们通过下例来说明(2.1)式是正确的.

例2.1  设$p(t)=t, f(t)=1, ~0\leq t\leq 3, $

$ q(t)= \left\{ \begin{array}{ll} 2,&0\leq t\leq 1, \\ t^2+1, \ \&1 \leq t\leq 3. \end{array}\right. $

下面计算$I_{0+}^{p(t)}I_{0+}^{q(t)}f(t)|_{t=1}$$I_{0+}^{p(t)+q(t)}f(t)|_{t=1}$.

首先, 有

$ \begin{eqnarray*} I_{{0+}}^{p(t)}I_{{0+}}^{q(t)}f(t)|_{t=1}&=& \int_0^1\frac{(1-s)^{p(s)-1}}{\Gamma(p(s))}\int_0^s \frac{(s-\tau)^{q(\tau)-1}}{\Gamma(q(\tau))}f(\tau){\rm d}\tau{\rm d}s\\ & =&\int_0^1\frac{(1-s)^{s-1}}{\Gamma(s)} \int_0^s\frac{(s-\tau)^{2-1}}{\Gamma(2)} {\rm d}\tau{\rm d}s\\ &=&\int_0^1\frac{(1-s)^{s-1}s^2}{\Gamma(s)\Gamma(3)}{\rm d}s= \int_0^1\frac{(1-s)^{s-1}s^2}{2\Gamma(s)}{\rm d}s\approx 0.173. \end{eqnarray*} $

另一方面, 有

$ \begin{eqnarray*} I_{{0+}}^{p(t)+q(t)}f(t)|_{t=1}=\int_0^1 \frac{(1-s)^{p(s)+q(s)-1}}{\Gamma(p(s)+q(s))}{\rm d}s =\int_0^1 \frac{(1-s)^{s+1}}{\Gamma(s+2)}{\rm d}s \approx 0.352. \end{eqnarray*} $

因此

$ I_{{0+}}^{p(t)}I_{{0+}}^{q(t)}f(t)|_{t=1}\neq I_{{0+}}^{p(t)+q(t)}f(t)|_{t=1}. $

通过上述分析, 可知命题2.1-2.4对于变量阶分数次算子而言是无效的, 从而我们不能将一个变量阶分数次微分方程转化成与之等价的积分方程.因此, 研究变量阶分数次微分方程及其初边值问题解的存在性, 虽然困难但很有意义.

接下来, 我们将给出关于变量阶分数次积分算子的一个重要结论.

引理2.1  设$x(t), p(t)$是实函数且$I_{0+}^{p(t)}x(t)$存在.若$q>0$, 则

$ I_{0+}^qI_{0+}^{p(t)}x(t)=I_{0+}^{q+p(t)}x(t), \ \ t>0. $ (2.2)

  由(1.3)式, 得

$ \begin{eqnarray*} I_{0+}^qI_{0+}^{p(t)}x(t)&=&\frac 1{\Gamma(q)}\int_0^t(t-s)^{q-1}\int_0^s\frac{(s-\tau)^{p(\tau)-1}}{\Gamma(p(\tau))}x(\tau){\rm d}\tau{\rm d}s\\ & =&\frac 1{\Gamma(q)}\int_0^t\int_\tau^t(t-s)^{q-1}\frac{(s-\tau)^{p(\tau)-1}}{\Gamma(p(\tau))}x(\tau){\rm d}s {\rm d}\tau \\ &=&\frac 1{\Gamma(q)}\int_0^t\int_0^1\frac{(t-\tau)^{q-1+p(\tau)}}{\Gamma(p(\tau))}(1-r)^{q-1}r^{p(\tau)-1}x(\tau){\rm d}r {\rm d}\tau \\ &=&\frac 1{\Gamma(q)}\int_0^t\frac{(t-\tau)^{q-1+p(\tau)}}{\Gamma(p(\tau))}\beta(q, p(\tau))x(\tau){\rm d}\tau \\ & =&\frac 1{\Gamma(q)}\int_0^t\frac{(t-\tau)^{q-1+p(\tau)}}{\Gamma(p(\tau))}x(\tau)\frac{\Gamma(q)\Gamma(p(\tau))}{\Gamma(q+p(\tau))}{\rm d}\tau \\ &=&\int_0^t\frac{(t-\tau)^{q-1+p(\tau)}}{\Gamma(q+p(\tau))}x(\tau){\rm d}\tau \\ &=&I^{q+p(t)}x(t), \end{eqnarray*} $

上式意味着(2.2)式是成立的.

注2.1  设$x(t), p(t)$是实函数且$I_{0+}^{p(t)}x(t)$存在.对于$q>0$, 则我们无法断定下式是否成立

$ I_{0+}^{p(t)}I_{0+}^qx(t)=I_{0+}^{q+p(t)}x(t), \ \ t>0. $ (2.3)

引理2.2  设$q:[0, T]\rightarrow (0, q^{*}]$ ($0 < q^{*} < 1$)是连续的, 则对于$x(t)\in C[0, T]$, 有$I_{0+}^{1-q(t)}x(t)\in C[0, T]$.

  由函数$q(t)$的连续性及Gamma函数, 可知$\Gamma(1-q(t))\in C[0, T]$.从而

$ \mbox{如果}\ \ 0 <T\leq 1, \ \ \mbox{则} \ \ T^{-q(s)}\leq T^{-q^{*}}, \ \ \mbox{如果}\ \ 1<T<+\infty, \ \ \mbox{则} \ \ T^{-q(s)}<1. $ (2.4)

对于$0 < T < +\infty$, 有

$ T^{-q(s)}\leq \max\{1, T^{-q^{*}}\}\doteq T^{*}. $ (2.5)

首先, 我们将证明

$ (I_{0+}^{1-q(t)}x(t))|_{t=0}=0, x\in C[0, T]. $ (2.6)

$x\in C[0, T]$, 有

$ \begin{eqnarray*} |I_{0+}^{1-q(t)}x(t)|&\leq&\int_0^t \bigg|\frac{(t-s)^{-q(s)}}{\Gamma(1-q(s))}\bigg|\cdot|x(s)|{\rm d}s\\ &\leq&\max\limits_{0\leq s\leq T}\bigg|\frac 1{\Gamma(1-q(s))}\bigg|\cdot \int_0^t T^{-q(s)} \bigg(\frac{t-s}T\bigg)^{-q(s)}|x(s)|{\rm d}s\\ &\leq&T^{*}\max\limits_{0\leq s\leq T}\bigg|\frac 1{\Gamma(1-q(s))}\bigg|\cdot\max\limits_{0\leq s\leq T}|x(s)|\int_0^t \bigg(\frac{t-s}T\bigg)^{-q^{*}}{\rm d}s\\ & \leq &\max\limits_{0\leq s\leq T}\bigg|\frac 1{\Gamma(1-q(s))}\bigg|\cdot \max\limits_{0\leq s\leq T}|x(s)|\frac{T^{*}T^{q^{*}}t^{1-q^{*}}}{1-q^{*}}, \end{eqnarray*} $

这意味着(2.6)式是成立的.

同样, 函数

$ k(t)=a^t-b^t, t\in (-1, 0), 0 <a<b<1 $ (2.7)

是递减的.事实上, 由于$\ln a < \ln b < 0$, $0 < b^t < a^t$, 我们有

$ k'(t)=a^t\ln a-b^t\ln b <b^t\ln a-b^t\ln b=b^t(\ln a-\ln b)<0. $

这表明$k(t)$是递减的.类似的

$ l(s)=\Big(\frac{t_1-s}T\Big)^{-q(s)}-\Big(\frac{t_2-s}T\Big)^{-q(s)} $

(这里$0 < \frac{t_1-s}T < \frac{t_2-s}T < 1$)也是递减的.

对于$t_0=0$时, 设$t>t_0=0$, $t-t_0=t\leq T$, 记

$ L=\max\limits_{0\leq t\leq T}\bigg|\frac 1{\Gamma(1-q(t))}\bigg|, \ \ \ M=\max\limits_{0\leq t\leq T}|x(t)|. $

对于$\forall\varepsilon>0$, 取$\delta=\big(\frac{(1-q^{*})\varepsilon}{LMT^{*}T^{q^{*}}}\big)^{\frac 1{1-q^{*}}}$.对$t\in (0, T]$$0 < t < \delta$, 根据(2.4)-(2.7)式, 可得

$ \begin{eqnarray*} \big|I_{0+}^{1-q(t)}x(t)-I_{0+}^{1-q(0)}x(0)\big|&=&\bigg |\int_0^t\frac{(t-s)^{-q(s)}}{\Gamma(1-q(s))}x(s){\rm d}s\bigg|\\ & \leq&\int_0^t\frac{T^{-q(s)}}{\Gamma(1-q(s))}\Big(\frac{t-s}{T}\Big) ^{-q(s)}|x(s)|{\rm d}s\\ &\leq&LM\int_0^tT^{*}\Big(\frac{t-s}{T}\Big)^{-q^{*}}{\rm d}s\\ & =&\frac{LMT^{*}T^{q^{*}}}{1-q^{*}}t^{1-q^{*}}\\ &<&\frac{LMT^{*}T^{q^{*}}}{1-q^{*}}\delta^{1-q^{*}}\\ & =&\varepsilon. \end{eqnarray*} $

$I_{0+}^{1-q(t)}x(t)$$t=0$点连续.

对于$t_0\in (0, T]$, 取$t>t_0$, $t-t_0 < T$.因此, 当$x\in C[0, T]$, 对$\forall\varepsilon>0$, 取$\delta=(\frac{(1-q^{*})\varepsilon}{2LMT^{*}T^{q^{*}}})^{\frac 1{1-q^{*}}}$, $t, t_0\in [0, T]$.不失一般性, 设$t_0 < t$, 当$t-t_0 < \delta$时, 由(2.4)-(2.7)式, 可得

$ \begin{eqnarray*} & &\big|I_{0+}^{1-q(t)}x(t)-I_{0+}^{1-q(t_0)}x(t_0)\big|\\ &=&\bigg|\int_0^t\frac{(t-s)^{-q(s)}}{\Gamma(1-q(s))}x(s){\rm d}s- \int_0^{t_0}\frac{(t_0-s)^{-q(s)}}{\Gamma(1-q(s))}x(s){\rm d}s\bigg|\\ &=&\bigg|\int_0^{t_0}\bigg[\frac{(t-s)^{-q(s)}}{\Gamma(1-q(s))}-\frac{(t_0-s)^{-q(s)}} {\Gamma(1-q(s))}\bigg]x(s){\rm d}s+ \int_{t_0}^{t}\frac{(t-s)^{-q(s)}}{\Gamma(1-q(s))}x(s){\rm d}s\bigg|\\ &\leq&LM\bigg[\int_0^{t_0}|(t-s)^{-q(s)}-(t_0-s)^{-q(s)}|{\rm d}s+ \int_{t_0}^{t}(t-s)^{-q(s)}{\rm d}s\bigg]\\ & =&LM\bigg[\int_0^{t_0}T^{-q(s)}\bigg|\Big(\frac{t-s}{T}\Big)^{-q(s)}- \Big (\frac{t_0-s}{T}\Big)^{-q(s)}\bigg|{\rm d}s+ \int_{t_0}^{t}T^{-q(s)}\Big(\frac{t-s}{T}\Big)^{-q(s)}{\rm d}s\bigg]\\ &\leq&T^{*}LM\bigg[\int_0^{t_0}\Big[ \Big(\frac{t_0-s}{T}\Big)^{-q(s)}- \Big(\frac{t-s}{T}\Big)^{-q(s)}\Big]{\rm d}s+ \int_{t_0}^{t}\Big(\frac{t-s}{T}\Big)^{-q(s)}{\rm d}s\bigg]\\ & \leq&T^{*}LM\bigg[\int_0^{t_0} \Big[\Big(\frac{t_0-s}{T}\Big)^{-q^{*}}-\Big(\frac{t-s}{T}\Big)^{-q^{*}}\Big]{\rm d}s+ \int_{t_0}^{t}\Big(\frac{t-s}{T}\Big)^{-q^{*}}{\rm d}s\bigg]\\ &=&\frac{LMT^{*}T^{q^{*}}}{1-q^{*}}\big[t_0^{1-q^{*}}+2(t-t_0)^{1-q^{*}}-t^{1-q^{*}}\big]\\ &\leq&\frac{LMT^{*}T^{q^{*}}}{1-q^{*}}\big[t^{1-q^{*}}+2(t-t_0)^{1-q^{*}}-t^{1-q^{*}}\big]\\ &=&\frac{2LMT^{*}T^{q^{*}}}{1-q^{*}}(t-t_0)^{1-q^{*}}\\ &<&\frac{2LMT^{*}T^{q^{*}}}{1-q^{*}}\delta^{1-q^{*}}\\ &=& \varepsilon. \end{eqnarray*} $

$I_{0+}^{1-q(t)}x(t)$$t_0$点连续.由$t_0$的任意性知, $I_{0+}^{1-q(t)}x(t)\in C[0, T], x\in C[0, T]$.证毕.

3 主要结果

本章将给出本文的主要结果.现假设如下条件成立

$(H_1)$ 设$0 < \beta < \alpha < 1$, $p:[0, T]\rightarrow (0, \alpha-\beta]$为连续函数, 对于$h\in C[0, T]$, 都有$I_{0+}^{1-p(t)}I_{0+}^\beta h(t)\in AC[0, T]$;

$(H_2)$ $t^rf:[0, T]\times {\Bbb R}\times {\Bbb R}\rightarrow {\Bbb R}$ ($0\leq r < \alpha-\beta$)是连续的;

$(H_3)$ 存在正常数$c_1, c_2, c_3, 0 < u_1 < 1, 0 < u_2 < 1$, 使得

$ t^r|f(t, x, y)|\leq c_1+c_2|x|^{\mu_1}+c_3|y|^{\mu_2}, 0\leq t\leq T, x, y\in {\Bbb R}. $

引理3.1  设

$ E=\{x(t)| x\in C[0, T], D_{0+}^{1-\alpha+\beta}x\in C[0, T] \}, $

其范数定义为

$ \|x\|=\max\big\{\max\limits_{0\leq t\leq T}|x(t)|, \max\limits_{0\leq t\leq T}|D_{0+}^{1-\alpha+ \beta} x(t)|\big\}, $

$(E, \|\cdot\|)$是一个Banach空间.

  此引理的证明类似于文献[2]中的引理2.2.该证明我们此处省略了.

为了得到本文结果, 我们首先给出如下定理.

定理3.1  假设$(H_1), (H_2)$成立, 则

(A)  如果初值问题$(1.1)$有一个解为$x\in C[0, T]$$D_{0+}^\beta x\in C[0, T]$, 则$z(t)=I_{0+}^{1-\alpha}x(t)\in C[0, T]$是如下积分方程的一个解

$ z(t)+\lambda I_{0+}^{1-p(t)}I_{0+}^\beta D_{0+}^{1-\alpha+\beta}z(t)= I_{0+}^1f(t, I_{0+}^\beta D_{0+}^{1-\alpha+\beta}z, D_{0+}^{1-\alpha+\beta}z(t)). $ (3.1)

(B)  如果方程$(3.1)$有一个解为$z\in E$, 则$x(t)=I_{0+}^\beta D_{0+}^{1-\alpha+\beta}z\in C[0, T]$是初值问题$(1.1)$的一个解.

  假设$x\in C[0, T]$$D_{0+}^\beta x\in C[0, T]$是初值问题(1.1)的一个解.根据(1.2)式及Riemann-Liouville分数阶导数的定义, 对方程(1.1)两边应用积分算子$I_{0+}^1$, 得

$ I_{0+}^1D_{0+}^{\alpha}x(t)+\lambda I_{0+}^1 D_{0+}^1I_{0+}^{1-p(t)}x(t)=I_{0+}^1f(t, x, D_{0+}^\beta x). $

由命题2.1, 2.2, 2.4及(2.6)式, 有

$ I_{0+}^{1-\alpha}x(t)+\lambda I_{0+}^{1-p(t)}x(t)=I_{0+}^1f(t, x, D_{0+}^\beta x). $ (3.2)

$D_{0+}^\beta x(t)=y(t)$, 由命题2.4, 得$x(t)=I_{0+}^\beta y(t)$.由$(3.2)$和命题2.1, 我们有

$ I_{0+}^{1-\alpha+\beta}y(t)+\lambda I_{0+}^{1-p(t)}I_{0+}^\beta y(t)= I_{0+}^1f(t, I_{0+}^\beta y, y). $ (3.3)

同样的, 设$I_{0+}^{1-\alpha+\beta}y(t)=z(t)$, 由命题2.2, 得$y(t)=D_{0+}^{1-\alpha+\beta}z(t)$.根据$(3.3)$式, 我们有

$ z(t)+\lambda I_{0+}^{1-p(t)}I_{0+}^\beta D_{0+}^{1-\alpha+\beta}z(t)= I_{0+}^1f(t, I_{0+}^\beta D_{0+}^{1-\alpha+\beta}z, D_{0+}^{1-\alpha+\beta}z(t)), $

$z(t)=I_{0+}^{1-\alpha+\beta}y(t)=I_{0+}^{1-\alpha+\beta}D_{0+}^\beta x(t)=I_{0+}^{1-\alpha}x(t)$是方程$(3.1)$的一个解.

另一方面, 如果积分方程(3.1)有一个解为$z\in E$, 则我们可以得到问题(1.1)的一个解.

事实上, 若对于方程(3.1)的一个解$z\in E$, 设

$ D_{0+}^{1-\alpha+\beta}z(t)=y(t). $ (3.4)

对(3.4)式两边作用算子$I_{0+}^{1-\alpha+\beta}$, 由命题2.4知

$ z(t)=I_{0+}^{1-\alpha+\beta}y(t)+ct^{\beta-\alpha}. $

$\beta < \alpha$$y(t)=D_{0+}^{1-\alpha+\beta}z(t)\in C[0, T]$$c=0$.所以

$ z(t)=I_{0+}^{1-\alpha+\beta}y(t). $ (3.5)

由此, 将$z(t)=I_{0+}^{1-\alpha+\beta}y(t)$代入$(3.1)$式, 得

$ I_{0+}^{1-\alpha+\beta}y(t)+\lambda I_{0+}^{1-p(t)}I_{0+}^\beta y(t)=I_{0+}^1f(t, I_{0+}^\beta y(t), y(t)). $ (3.6)

同理, 对于$(3.6)$式, 可设

$ I_{0+}^\beta y(t)=x(t). $ (3.7)

根据命题2.2, 对$(3.7)$式两边同时应用$D_{0+}^\beta$, 可得

$ y(t)=D_{0+}^\beta x(t). $ (3.8)

$y(t)=D_{0+}^\beta x(t)$带入$(3.6)$式, 由命题2.1, 可得

$ I_{0+}^{1-\alpha}x(t)+\lambda I_{0+}^{1-p(t)}x(t)=I_{0+}^1f(t, x(t), D_{0+}^\beta x). $ (3.9)

根据命题2.5, 知

$ x(t)=I_{0+}^\beta y(t)=I_{0+}^\beta D_{0+}^{1-\alpha+\beta}z(t)\in C[0, T], D_{0+}^\beta x(t)=y(t)=D_{0+}^{1-\alpha+\beta}z(t)\in C[0, T]. $

同样的, 根据命题2.5, $D_{0+}^{1-\alpha+\beta}z\in C[0, T]$$(H_1)$, 有

$ \lambda I_{0+}^{1-p(t)}x=\lambda I_{0+}^{1-p(t)}I_{0+}^\beta D_{0+}^{1-\alpha+\beta}z\in AC[0, T]. $ (3.10)

因此, 根据(3.9)及(3.10)式, 我们有$I_{0+}^{1-\alpha}x\in AC[0, T]$.对$(3.9)$式两边同时求导, 可得

$ D_{0+}^\alpha x(t)+\lambda D_{0+}^{p(t)} x(t)=f\big(t, x(t), D_{0+}^\beta x\big). $

根据$D_{0+}^{1-\alpha+\beta}z\in C[0, T]$我们有$x(0)=I_{0+}^\beta y(t)|_{t=0}=I_{0+}^\beta D_{0+}^{1-\alpha+\beta}z(t)|_{t=0}=0$, 这意味着$x(t)=I_{0+}^\beta y(t)=I_{0+}^\beta D_{0+}^{1-\alpha+\beta}z(t)$是初值问题(1.1)的一个解.

下面给出方程$(1.1)$解的存在性结论.

定理3.2  假设$(H_1)$-$(H_3)$成立且有下面的不等式成立, 则初值问题$(1.1)$至少有一个解, 这里

$ |\lambda|\leq \min\bigg\{\frac {\Gamma(2-q^{*}+\beta)}{4L_pT^{*}T^{1+\beta}\Gamma(1-q^{*})}, \frac {\Gamma(1+\alpha-q^{*})}{4L_{p\alpha\beta}T^{*}\Gamma(\alpha-\beta-q^{*})T^{1+\beta}}\bigg\}, $

其中

$ L_p=\max\limits_\limits{0\leq t\leq T}\frac 1{|\Gamma(1-p(t))|}, \ \ L_{p\alpha\beta}=\max\limits_\limits{0\leq t\leq T}\frac 1{|\Gamma(\alpha-\beta-p(t))|}. $

  由定理$3.2$可知, 我们将讨论方程$(3.1)$解的存在性.定义算子$T:E\rightarrow E$

$ Tz(t)=I_{0+}^1f\big(t, I_{0+}^\beta D_{0+}^{1-\alpha+\beta}z, D_{0+}^{1-\alpha+\beta}z(t)\big)-\lambda I_{0+}^{1-p(t)}I_{0+}^\beta D_{0+}^{1-\alpha+\beta}z(t). $ (3.11)

易知$T:E\rightarrow E$是有意义的.

首先, 我们将证明$Tz(t),  D_{0+}^{1-\alpha+\beta}Tz(t)\in C[0, T], z(t)\in E$.

事实上, 设

$ g\big(s, I_{0+}^\beta D_{0+}^{1-\alpha+\beta}z, D_{0+}^{1-\alpha+\beta}z\big)= s^rf\big(s, I_{0+}^\beta D_{0+}^{1-\alpha+\beta}z, D_{0+}^{1-\alpha+\beta}z\big). $

$(H_2)$, 可知$g:[0, T]\times {\Bbb R}\times {\Bbb R}\rightarrow{\Bbb R}$是连续的.同样, 由命题2.5, 若$D_{0+}^{1-\alpha+\beta}z\in C[0, T]$, 有$I_{0+}^\beta D_{0+}^{1-\alpha+\beta}z\in C[0, T]$.因为$D_{0+}^{1-\alpha+\beta}z(s)$在任意点$t_0\tau\in [0, T]$($\tau\in [0, 1]$)都连续, 从而对于任意的$\varepsilon>0$, 存在$\delta_1>0$, 当$t_0\tau, t\tau\in [0, T], \tau\in [0, 1]$, $|t\tau-t_0\tau| < \delta_1$, 下式恒成立

$ |D_{0+}^{1-\alpha+\beta}z(t\tau)-D_{0+}^{1-\alpha+\beta} z(t_0\tau)|<\varepsilon, $ (3.12)

类似的, $I_{0+}^\beta D_{0+}^{1-\alpha+\beta}z(s)$也在任意点$t_0\tau\in [0, T]$($\tau\in [0, 1]$)连续.对任意的$\varepsilon>0$, 存在$\delta_2>0$, 当$t_0\tau, t\tau\in [0, T], \tau\in [0, 1]$, $|t\tau-t_0\tau| < \delta_2$, 恒有

$ |I_{0+}^\beta D_{0+}^{1-\alpha+\beta}z(t\tau)-I_{0+}^\beta D_{0+}^{1-\alpha+\beta}z(t_0\tau)|<\varepsilon. $ (3.13)

由于$g\big(s, I_{0+}^\beta D_{0+}^{1-\alpha+\beta}z, D_{0+}^{1-\alpha+\beta}z\big)$$(t_0\tau, I_{0+}^\beta D_{0+}^{1-\alpha+\beta}z(t_0\tau), D_{0+}^{1-\alpha+\beta}z(t_0\tau))$ ($t_0\in [0, T], \tau\in [0, 1]$)处连续.从而, 对于任意的$\varepsilon>0$, 存在$\delta>0$, 当$t_0\tau, t\tau\in [0, T], \tau\in [0, 1]$, $|t\tau-t_0\tau| < \delta$, 结合公式(3.12)和(3.13), 我们有

$ |g\big(t\tau, I_{0+}^\beta D_{0+}^{1-\alpha+\beta}z(t\tau), D_{0+}^{1-\alpha+\beta}z(t\tau)\big)-g\big(t_0\tau, I_{0+}^\beta D_{0+}^{1-\alpha+\beta}z(t_0\tau), D_{0+}^{1-\alpha+\beta}z(t_0\tau)\big)|<\varepsilon. $ (3.14)

另一方面, 由Riemann-Liouville分数阶导数和积分的定义, 我们有

$ \begin{eqnarray*} I_{0+}^\beta D_{0+}^{1-\alpha+\beta}z(s)&=&\frac 1{\Gamma(\beta)}\int_0^s(s-\mu)^{\beta-1} D_{0+}^{1-\alpha+\beta}z(\mu){\rm d}\mu\\ &\mathop{=}\limits^{s=t\tau} &\frac 1{\Gamma(\beta)}\int_0^{t\tau}(t\tau-\mu)^{\beta-1} D_{0+}^{1-\alpha+\beta}z(\mu){\rm d}\mu\\ &=&I_{0+}^\beta D_{0+}^{1-\alpha+\beta}z(t\tau) \end{eqnarray*} $

$ \begin{eqnarray*} D_{0+}^{1-\alpha+\beta}z(s)&=&\frac{\rm d}{{\rm d}s}I_{0+}^{\alpha-\beta}z(s)\\ &=&\frac{\rm d}{{\rm d}s}\bigg[\frac 1{\Gamma(\alpha-\beta)}\int_0^s(s-\mu)^{\alpha-\beta-1} z(\mu){\rm d}\mu\bigg]\\ &\mathop{=}\limits^{s=t\tau} &\frac{\rm d}{{\rm d}(t\tau)}\bigg[\frac 1{\Gamma(\alpha-\beta)}\int_0^{t\tau}(t\tau-\mu)^{\alpha-\beta-1} z(\mu){\rm d}\mu\bigg]\\ &=&D_{0+}^{1-\alpha+\beta}z(t\tau). \end{eqnarray*} $

因此, 通过前面的讨论, 对于$t, t_0\in [0, T]$, 使得$t>t_0$, $t-t_0 < T$, 我们有

$ \begin{eqnarray*} &&\bigg |I_{0+}^{\alpha-\beta}f\big(t, I_{0+}^\beta D_{0+}^{1-\alpha+\beta}z(t), D_{0+}^{1-\alpha+\beta}z(t)\big)-I_{0+}^{\alpha-\beta}f\big(t_0, I_{0+}^\beta D_{0+}^{1-\alpha+\beta}z(t_0), D_{0+}^{1-\alpha+\beta}z(t_0)\big)\bigg|\\ & =&\bigg|\frac 1{\Gamma(\alpha-\beta)}\bigg[\int_0^{t}(t-s)^{\alpha-\beta-1}s^{-r}s^rf\big(s, I_{0+}^\beta D_{0+}^{1-\alpha+\beta}z(s), D_{0+}^{1-\alpha+\beta}z(s)\big){\rm d}s\\ &&-\int_0^{t_0}(t_0-s)^{\alpha-\beta-1}s^{-r}s^rf\big(s, I_{0+}^\beta D_{0+}^{1-\alpha+\beta}z(s), D_{0+}^{1-\alpha+\beta}z(s)\big)\bigg]\bigg|\\ &=&\frac 1{\Gamma(\alpha-\beta)} \bigg| t^{\alpha-\beta-r}\int_0^1 (1-\tau)^{\alpha-\beta-1}\tau^{-r}g(t\tau, I_{0+}^\beta D_{0+}^{1-\alpha+\beta}z(t\tau ), D_{0+}^{1-\alpha+\beta}z(t\tau)){\rm d}\tau\\ &&-t_0^{\alpha-\beta-r} \int_0^1(1-\tau)^{\alpha-\beta-1}\tau^{-r}g(t_0\tau, I_{0+}^\beta D_{0+}^{1-\alpha+\beta}z(t_0\tau ), D_{0+}^{1-\alpha+\beta}z(t_0\tau)){\rm d}\tau\bigg| \\ &\leq &\frac {t^{\alpha-\beta-r}-t_0^{\alpha-\beta-r}}{\Gamma(\alpha-\beta)}\int_0^1(1-\tau)^{\alpha-\beta-1} \tau^{-r}|g\big(t\tau, I_{0+}^\beta D_{0+}^{1-\alpha+\beta}z(t\tau ), D_{0+}^{1-\alpha+\beta}z(t\tau)\big)|{\rm d}\tau\\ &&+\frac{t_0^{\alpha-\beta-r}}{\Gamma(\alpha-\beta)}\bigg[\int_0^1(1-\tau)^ {\alpha-\beta-1}\tau^{-r}|g\big(t\tau, I_{0+}^\beta D_{0+}^{1-\alpha+\beta}z(t\tau ), D_{0+}^{1-\alpha+\beta}z(t\tau)\big) \\ &&-g\big(t_0\tau, I_{0+}^\beta D_{0+}^{1-\alpha+\beta}z(t_0\tau ), D_{0+}^{1-\alpha+\beta}z(t_0\tau))|{\rm d}\tau\bigg]. \end{eqnarray*} $

可知$I_{0+}^\beta D_{0+}^{1-\alpha+\beta}z(t\tau)$, $D_{0+}^{1-\alpha+\beta}z(t\tau)$均在$t_0\tau\in [0, T]$ $(\tau\in [0, 1])$处连续, $g\big(t\tau, I_{0+}^\beta D_{0+}^{1-\alpha+\beta}z(t\tau), $ $D_{0+}^{1-\alpha+\beta}z(t\tau)\big)$$(t_0\tau, I_{0+}^\beta D_{0+}^{1-\alpha+\beta}z(t_0\tau), D_{0+}^{1-\alpha+\beta}z(t_0\tau))$ ($t_0\in [0, T]$, $\tau\in [0, 1]$)处连续.易知$t^{\alpha-\beta-r}$$t_0\in [0, T]$处连续.继而我们有

$ |t\tau-t_0\tau|\leq |t-t_0|. $

从而, $I_{0+}^{\alpha-\beta}f\big(t, I_{0+}^\beta D_{0+}^{1-\alpha+\beta}z(t), D_{0+}^{1-\alpha+\beta}z(t)\big)$$t_0$处连续.由于$t_0$是任意的, 故有

$ I_{0+}^{\alpha-\beta}f\big(t, I_{0+}^\beta D_{0+}^{1-\alpha+\beta}z(t), D_{0+}^{1-\alpha+\beta}z(t)\big)\in C[0, T]. $

同样的分析可知

$ \int_0^tf\big(s, I_{0+}^\beta D_{0+}^{1-\alpha+\beta}z, D_{0+}^{1-\alpha+\beta}z\big){\rm d}s\in C[0, T]. $

根据前面的分析和引理2.2得$I_{0+}^{1-p(t)}I_{0+}^\beta D_{0+}^{1-\alpha+\beta}z(t)\in C[0, T]$

$ D_{0+}^{1-\alpha+\beta}I_{0+}^{1-p(t)}I_{0+}^\beta D_{0+}^{1-\alpha+\beta}z(t)=I_{0+}^{\alpha-\beta-p(t)}I_{0+}^\beta D_{0+}^{1-\alpha+\beta}z(t)\in C[0, T]. $

从而$Tz(t)\in C[0, T], ~D_{0+}^{1-\alpha+\beta}Tz(t)\in C[0, T], z\in C[0, T]$.这意味着$T: E\rightarrow E$是有意义的.

另一方面, 易知$T: E\rightarrow E$是全连续的.

$\Omega=\{z\in E; \|z\|\leq R\}$, 其中

$ \begin{eqnarray*} R&>&\bigg\{\frac{4c_1T^{1-r}}{1-r}, \bigg(\frac{4c_2T^{1-r+\beta\mu_1}}{(1-r+\beta\mu_1) \Gamma^{\mu_1}(1+\beta)}\bigg)^{\frac 1{1-\mu_1}}, \bigg(\frac {4c_3T^{1-r}}{1-r}\bigg)^{\frac 1{1-\mu_2}}, \frac{4c_1T^{\alpha-\beta-r}\Gamma(1-r)}{\Gamma(1-r+\alpha-\beta)}, \\ && \bigg(\frac{4c_2T^{\alpha-\beta-r+\beta\mu_1}\Gamma(1-r+\beta\mu_1)}{\Gamma(1-r+\beta\mu_1+\alpha-\beta)\Gamma^{\mu_1}(1+\beta)} \bigg)^{\frac 1{1-\mu_1}}, \bigg(\frac{4c_3T^{\alpha-\beta-r}\Gamma(1-r)} {\Gamma(1-r+\alpha-\beta)}\bigg)^{\frac 1{1-\mu_2}}\bigg\}. \end{eqnarray*} $

$z\in \Omega$, 有

$ \begin{eqnarray*} |Tz(t)|&\leq &\bigg|\int_0^{t}f(s, I_{0+}^\beta D_{0+}^{1-\alpha+\beta}z(s), D_{0+}^{1-\alpha+\beta}z(s)){\rm d}s\bigg|\\ &&+\bigg| \lambda\int_0^{t}\frac{(t-s)^{-p(s)}}{\Gamma(1-p(s))}I_{0+}^\beta D_{0+}^{1-\alpha+\beta}z(s){\rm d}s\bigg|\\ & \leq&\int_0^{t}\big|f(s, I_{0+}^\beta D_{0+}^{1-\alpha+\beta}z(s), D_{0+}^{1-\alpha+\beta}z(s))\big|{\rm d}s\\ &&+L_p|\lambda|\int_0^{t}(t-s)^{-p(s)}I_{0+}^\beta |D_{0+}^{1-\alpha+\beta}z(s)|{\rm d}s\\ & \leq&\int_0^{t}s^{-r}\bigg(c_1+c_2|I_{0+}^\beta D_{0+}^{1-\alpha+\beta}z(s)|^{\mu_1}+c_3|D_{0+}^{1-\alpha+\beta}z(s)|^{\mu_2}\bigg){\rm d}s\\ &&+L_p|\lambda|\int_0^{t}(t-s)^{-p(s)}I_{0+}^\beta \|z\|{\rm d}s\\ &\leq&\int_0^{t}s^{-r}\bigg(c_1+\frac{c_2s^{\beta\mu_1}}{\Gamma^{\mu_1}(1+\beta)}\|z\|^{\mu_1} +c_3\|z\|^{\mu_2}\bigg){\rm d}s\\ &&+\frac {L_p|\lambda|}{\Gamma(1+\beta)}\int_0^{t}T^{-p(s)} \Big(\frac{t-s}T\Big)^{-p(s)}s^{\beta}\|z\|{\rm d}s\\ & \leq&\frac{c_1T^{1-r}}{1-r}+ \frac{c_2T^{1-r+\beta\mu_1}}{(1-r+\beta\mu_1)\Gamma^{\mu_1}(1+\beta)}\|z\|^{\mu_1} +\frac{c_3T^{1-r}}{1-r}\|z\|^{\mu_2}\\ &&+\frac {L_p|\lambda|T^{*}}{\Gamma(1+\beta)}\int_0^{t}\Big(\frac{t-s}T\Big)^{-q^{*}}s^{\beta}\|z\|{\rm d}s\\ & \leq&\frac{c_1T^{1-r}}{1-r}+\frac{c_2T^{1-r+\beta\mu_1}} {(1-r+\beta\mu_1)\Gamma^{\mu_1}(1+\beta)}\|z\|^{\mu_1} +\frac{c_3T^{1-r}}{1-r}\|z\|^{\mu_2}\\ &&+\frac {L_p|\lambda|T^{*}T^{1+\beta}\Gamma(1-q^{*})}{\Gamma(2-q^{*}+\beta)}\|z\|\\ & \leq&\frac{c_1T^{1-r}}{1-r}+\frac{c_2T^{1-r+\beta\mu_1}}{(1-r+\beta\mu_1) \Gamma^{\mu_1}(1+\beta)}R^{\mu_1} +\frac{c_3T^{1-r}}{1-r}R^{\mu_2} \\ &&+\frac {L_p|\lambda|T^{*}T^{1+\beta}\Gamma(1-q^{*})}{\Gamma(2-q^{*}+\beta)}R\\ &=&\frac{c_1T^{1-r}}{1-r}+\frac{c_2T^{1-r+\beta\mu_1}}{(1-r+\beta\mu_1) \Gamma^{\mu_1}(1+\beta)}RR^{\mu_1-1} +\frac{c_3T^{1-r}}{1-r}RR^{\mu_2-1} \\ &&+\frac {L_p|\lambda|T^{*}T^{1+\beta}\Gamma(1-q^{*})}{\Gamma(2-q^{*}+\beta)}R\\ &\leq&\frac R4+\frac R4+\frac R4+\frac R4=R. \end{eqnarray*} $

类似的, 对$z\in E$, 有

$ \begin{eqnarray*} &&|D_{0+}^{1-\alpha+\beta}Tz(t)|\\ &=& \bigg|I_{0+}^{\alpha-\beta}f\big(t, I_{0+}^\beta D_{0+}^{1-\alpha+\beta}z(t), D_{0+}^{1-\alpha+\beta}z(t)\big)-\lambda I_{0+}^{\alpha-\beta-p(t)}I_{0+}^\beta D_{0+}^{1-\alpha+\beta}z(t)\bigg|\\ & \leq&\frac 1{\Gamma(\alpha-\beta)} \int_0^{t}(t-s)^{\alpha-\beta-1}s^{-r}\big(c_1+c_2|I_{0+}^\beta D_{0+}^{1-\alpha+\beta}z(s)|^{\mu_1}+c_3|D_{0+}^{1-\alpha+\beta}z(s)|^{\mu_2}\big){\rm d}s\\ &&+|\lambda|\int_0^{t}\frac{(t-s)^{\alpha-\beta-p(s)-1}} {|\Gamma(\alpha-\beta-p(s))|}I_{0+}^\beta \|z\|{\rm d}s\\ & \leq&\frac 1{\Gamma(\alpha-\beta)} \int_0^{t}(t-s)^{\alpha-\beta-1}s^{-r}\bigg(c_1+\frac{c_2s^{\beta\mu_1}}{\Gamma^{\mu_1}(1+\beta)}\|z\|^{\mu_1} +c_3\|z\|^{\mu_2}\bigg){\rm d}s\\ &&+\frac {L_{p\alpha\beta}|\lambda|}{\Gamma(1+\beta)}\int_0^{t}T^{\alpha-\beta-p(s)-1} \Big(\frac{t-s}T\Big)^{\alpha-\beta-p(s)-1}s^{\beta}\|z\|{\rm d}s\\ &\leq&\frac{c_1T^{\alpha-\beta-r}\Gamma(1-r)}{\Gamma(1-r+\alpha-\beta)}+ \frac{c_2T^{\alpha-\beta-r+\beta\mu_1}\Gamma(1-r+\beta\mu_1)} {\Gamma(1-r+\beta\mu_1+\alpha-\beta)\Gamma^{\mu_1}(1+\beta)}RR^{\mu_1-1} \\ &&+\frac{c_3T^{\alpha-\beta-r}\Gamma(1-r)}{\Gamma(1-r+\alpha-\beta)}RR^{\mu_2-1}+\frac {L_{p\alpha\beta}|\lambda|T^{*}}{\Gamma(1+\beta)}\int_0^{t} \Big(\frac{t-s}T\Big)^{\alpha-\beta -q^{*}-1}s^{\beta}R{\rm d}s \\ &\leq&\frac{c_1T^{\alpha-\beta-r}\Gamma(1-r)}{\Gamma(1-r+\alpha-\beta)}+ \frac{c_2T^{\alpha-\beta-r+\beta\mu_1}\Gamma(1-r+\beta\mu_1)} {\Gamma(1-r+\beta\mu_1+\alpha-\beta)\Gamma^{\mu_1}(1+\beta)}RR^{\mu_1-1} \\ &&+\frac{c_3T^{\alpha-\beta-r}\Gamma(1-r)}{\Gamma(1-r+\alpha-\beta)}RR^{\mu_2-1}+\frac {L_{p\alpha\beta}|\lambda|T^{*}\Gamma(\alpha-\beta-q^{*}) T^{1+\beta}}{\Gamma(1+\alpha-q^{*})}R\\ & \leq&\frac R4+\frac R4+\frac R4+\frac R4=R. \end{eqnarray*} $

这意味着$T(\Omega)\subseteq \Omega$.根据Schauder不动点定理知$T$$z\in E$中至少有一个不动点, 即方程$(3.1)$至少有一个解.证毕.

注3.1  在$(H_3)$中, 若$\mu_1, \mu_2\geq 1$, 则适当的增加对$c_1, c_2, c_3$的约束条件, 利用类似于定理2.1的方法, 我们同样可以得到初值问题$(1.1)$的解.

参考文献
[1] Agarwal R P, Benchohra M, Hamani S. Boundary value problems for fractional differential equations. Georgian Math J, 2009, 16: 401–411.
[2] Su X. Solutions to boundary value problem of fractional order on unbounded domains in Banach space. Nonlinear Anal, 2011, 74: 2844–2852. DOI:10.1016/j.na.2011.01.006
[3] Zhang S. The uniqueness result of solutions to initial value problems of differential equations of variable-order. Rev R Acad Cienc Exactas Fís Nat Ser A Mat, 2017. DOI:10.1007/s13398-017-0389-4
[4] Sun H, Chen W, Chen Y. Variable-order fractional differential operators in anomalous diffusion modeling. Phys A, 2009, 388: 4586–4592. DOI:10.1016/j.physa.2009.07.024
[5] Zhang S. Existence of solutions for two-point boundary-value problems with singular differential equations of variable order. Electron J Differential Equations, 2013, 2013: 1–16.
[6] Sun H, Chen W, Wei H, Chen Y. A comparative study of constant-order and variable-order fractional models in characterizing memory property of systems. Eur Phys J Special Topics, 2011, 193: 185–192. DOI:10.1140/epjst/e2011-01390-6
[7] Chan C, Shu J, Yang R. Iterative design of variable fractional-order ⅡR differintegrators. Signal Process, 2010, 90: 670–678. DOI:10.1016/j.sigpro.2009.08.006
[8] Ross B. Fractional integration operator of variable-order in the Hölder space Hλ(x). Internat J Math and Math Sci, 1995, 18: 777–788. DOI:10.1155/S0161171295001001
[9] Valério D, Costa J S. Variable-order fractional derivatives and their numerical approximations. Signal Process, 2011, 91: 470–483. DOI:10.1016/j.sigpro.2010.04.006
[10] Razminia A, Dizaji A F, Majd V J. Solution existence for non-autonomous variable-order fractional differential equations. Math Comput Model, 2012, 55: 1106–1117. DOI:10.1016/j.mcm.2011.09.034
[11] Lin R, Liu F, Anh V, Turner I. Stability and convergence of a new explicit finite-difference approximation for the variable-order nonlinear fractional diffusion equation. Appl Math Comput, 2009, 2: 435–445.
[12] Sierociuk D, Malesza W, Macias M. Derivation, interpretation, and analog modelling of fractional variable order derivative definition. Appl Math Model, 2015, 39: 3876–3888. DOI:10.1016/j.apm.2014.12.009
[13] Umarov S, Steinber S. Variable order differential equations and diffusion processes with changing modes. Math Phys, 2009, 2009: 1–20.
[14] Zhao X, Sun Z, EmKarniadakis G. Second-order approximations for variable order fractional derivatives:Algorithms and applications. J Comput Phys, 2015, 293: 184–200. DOI:10.1016/j.jcp.2014.08.015
[15] Atangana A. On the stability and convergence of the time-fractional variable order telegraph equation. J Comput Phys, 2015, 293: 104–114. DOI:10.1016/j.jcp.2014.12.043
[16] Li X, Wu B. A numerical technique for variable fractional functional boundary value problems. Appl Math Lett, 2015, 43: 108–113. DOI:10.1016/j.aml.2014.12.012
[17] Kilbas A A, Srivastava H M, Trujillo J J. Theory and applications of fractional differential equations. Amsterdam: Elsevier B V, 2006.
[18] Dreher M, Jüngel A. Compact families of piecewise constant functions in Lp (0, T; B). Nonlinear Anal, 2012, 75: 3072–3077. DOI:10.1016/j.na.2011.12.004
[19] Lakshmikanthama V, Vatsala A S. Basic theory of fractional differential equations. Nonlinear Anal, 2008, 69: 2677–2682. DOI:10.1016/j.na.2007.08.042
[20] Kou C, Zhou H, Yan Y. Existence of solutions of initial value problems for nonlinear fractional differential equations on the half-axis. Nonlinear Anal, 2011, 74: 5975–5986. DOI:10.1016/j.na.2011.05.074
[21] Baleanu D, Mohammadi H, Rezapour S. Positive solutions of an initial value problem for nonlinear fractional differential equations. Abstr Appl Anal, 2012, 2012: 1–7.
[22] Li X, Wu B. A numerical technique for variable fractional functional boundary value problems. Appl Math Lett, 2015, 43: 108–113. DOI:10.1016/j.aml.2014.12.012
[23] 姚庆六. 一个非线性分数微分方程奇异解的存在性与逐次迭代方法. 数学物理学报, 2016, 36A: 287–296.
Yao Q. Existence and successively iterative method of singular solution to a nonlinear fractional differential equation. Acta Math Sci, 2016, 36A: 287–296. DOI:10.3969/j.issn.1003-3998.2016.02.007
[24] 邓继勤, 邓子明. 分数阶微分方程非局部柯西问题解的存在和唯一性. 数学物理学报, 2016, 36A: 1157–1164.
Deng J, Deng Z. Existence and uniqueness of solutions for nonlocal Cauchy problem for fractional evolution equations. Acta Math Sci, 2016, 36A: 1157–1164. DOI:10.3969/j.issn.1003-3998.2016.06.014