Regularity of stochastic convolution with distributed delay
-
摘要: 本文讨论了一类线性随机泛函微分方程的随机卷积的正则性,这类带分布延迟的随机泛函微分方程是由分数布朗运动驱动的。利用基本解的估计和分数布朗运动的性质得到了随机卷积的正则性。特别地,在方程的系数算子符合一些特定条件时,给出了满足随机卷积的正则性的条件。Abstract: This paper discusses the regularity of stochastic convolution of a class of linear stochastic functional differential equations with distributed delay driven by fractional Brownian motion. In this paper, the regularity of stochastic convolution is obtained by using the estimation of fundamental solutions and the properties of fractional Brownian motion. In particular, when the coefficient operator of the equation conforms to some specific conditions, this paper gives a condition satisfying the regularity of stochastic convolution.
-
Keywords:
- regularity /
- fractional Brown motion /
- stochastic convolution /
- fundamental solution
-
0. 引言
随机微分方程是在常微分方程的基础上加入噪声项得到的,被广泛应用于生物、物理和金融[1-3]等领域。泛函微分方程与时间延迟有关,方程的性质不仅取决于当前状态,还取决于过去的状态,为反映实际问题,需要引入延迟量。近年来,随机泛函微分方程在数学及其应用中引起了广泛的关注,如随机泛函微分方程稳定性[4],无穷时滞脉冲随机泛函微分方程一般衰减意义下p阶矩稳定性[5],由分数布朗运动驱动的随机泛函微分方程解的存在唯一性、可行性、全局吸收集和指数衰减[6]等。
分数布朗运动出现在许多复杂现象模型的应用中。当分数布朗运动的Hurst参数$ \hbar$ = 1/2时,分数布朗运动成为标准布朗运动。分数布朗运动与标准布朗运动的不同之处在于分数布朗运动缺少半鞅性、马尔可夫性和其他经典的随机过程理论[7]。分数布朗运动作为一个中心高斯过程,其特点是增量的平稳性和中或长记忆的性质。长记忆特性使分数布朗运动成为许多领域建模的候选者[8-12]。对于与分数布朗运动有关的数学模型是近年来的研究热点。如Boufoussi和Hajji[13]研究了由分数布朗运动驱动的中立随机泛函微分方程解的存在唯一性和指数渐近行为;Caraballo等[14]研究了具有分数布朗运动驱动的随机延迟演化方程解的存在性和指数行为;Li和Yan[15]讨论了在希尔伯特空间上由分数布朗运动驱动的一类中立随机泛函微分方程的遍历性和平稳解;Xu等[16]讨论了由分数布朗运动驱动的二阶中立型随机泛函微分方程的全局吸引集和指数衰减。
现阶段对于随机泛函微分方程的随机卷积的正则性的研究少之又少,Liu[17-19]研究了由布朗运动驱动的带分布延迟和点延迟的随机卷积的正则性。
对随机卷积的正则性的研究还处于发展阶段,因此,本文研究了一类线性随机泛函微分方程的随机卷积的正则性。这类方程是由Hurst参数为$\hbar(1 / 2 <\hbar <1) $的分数布朗运动驱动的。正则性是指随机卷积的轨道是Hölder连续的。
1. 预备知识
设H是一个可分的希尔伯特空间,在此空间中定义范数为||·||H,内积为$ \langle\cdot\rangle_H$。定义$\mathcal{L}(H) $为H → H的有界线性算子所组成的空间。考虑确定性的线性随机泛函微分方程
$$ \left\{\begin{array}{l} \mathrm{d} y(t)=A y(t) \mathrm{d} t+\int_{-r}^0 a(\theta) A_1 y(t+\theta) \mathrm{d} \theta \mathrm{d} t+f(t) \mathrm{d} t, t \geqslant 0, \\ y(0)=\phi_0, y(\theta)=\phi_1(\theta), \quad \theta \in[-r, 0], \phi=\left(\phi_0, \phi_1\right), \end{array}\right. $$ (1) 式中:r > 0是引起系统延迟的某个常数;a (·) ∈ L2([-r,0],R);ϕ = (ϕ0,ϕ1) 为恰当的初值;A:Dom (A) ⊂ H → H是解析半群{etA,t ≥ 0} 的无穷小生成元;A1是一个闭的线性算子;定义域Dom (A1) ⊃ Dom (A) 并且f (·) 是一个在H中取值的连续函数。假设C0 - 半群etA满足以下2个条件:$\left\|\mathrm{e}^{t A}\right\| \leqslant M \mathrm{e}^{-\mu t}, M \geqslant 1, \mu>0 ;\left\|A \mathrm{e}^{t A}\right\| \leqslant M / t, M \geqslant 1 $。方程的基本解G (·):R → L (H) 是一个算子值函数,在H中是强连续的并且满足积分形式[19]
$$ G(t)= \begin{cases}\mathrm{e}^{t A}+\int_0^t \int_{-r}^0 a(\theta) \mathrm{e}^{(t-s) A} A_1 G(s+\theta) \mathrm{d} \theta \mathrm{d} s, & t \geqslant 0, \\ O, & t <0, \end{cases} $$ 式中O为H中的零算子。
引理1.1[19] 若a (·) ∈ L∞ ([-r,0],R) 且存在ν ∈ (0,1),使得Dom (- A)ν ⊂ Dom (A),对于基本解有以下估计:
(ⅰ)对于任意的γ ∈ [ν,1],0 < t ≤ nr有
$$ \left\|(-A)^\gamma G(t)\right\| \leqslant \frac{C_{n, \gamma}}{t^\gamma}, $$ (2) 式中Cn,γ > 0,n ∈ 1,2,…,是一个依赖于n,γ的常数。
(ⅱ)对于任意的γ ∈ [ν,1],0 < β < 1 - γ,
$$ \left\|(-A)^\gamma(G(t)-G(s))\right\| \leqslant C_{1, \beta, \gamma} \frac{(t-s)^\beta}{s^{\beta+\gamma}}, 0 <s \leqslant r, $$ (3) $$ \left\|(-A)^\gamma(G(t)-G(s))\right\| \leqslant C_{n, \beta, \gamma} \frac{(t-s)^\beta}{(s-r)^\beta}, r <s \leqslant n r, $$ (4) 式中:Cn,β,γ > 0,n ∈ 1,2,…,是一个依赖于n,β,γ的常数;C1,β,γ > 0,是一个依赖于β,γ的常数。
定义1.1[16] 设$Q \in \mathcal{L}(H) $是一个正的自共轭算子。在概率空间$ (\varOmega, \mathcal{F}, \mathrm{P})$中,一个H - 值高斯过程$\left(B_Q^\hbar(t), t \geqslant 0\right) $如果满足以下条件:
(ⅰ)对于任意的t ≥ 0,映射$B_Q^\hbar(t): H \rightarrow L^2\left(\varOmega ; \mathbf{R}^1\right) $是线性的;
(ⅱ)对于任意的$h \in H, t \geqslant 0, B_Q^\hbar(t) h $是一个实高斯过程,并且$\mathrm{E} B_Q^\hbar(t) h=0 $;
(ⅲ)对于任意的对于任意的s,t ∈ R+ = [0,∞),h1,h2∈ H有
$$ \mathrm{E}\left(B_Q^\hbar(t) h_1 \cdot B_Q^\hbar(t) h_2\right)=\frac{1}{2}\left(t^{2 \hbar}+s^{2 \hbar}-|t-s|^{2 \hbar}\right)\left\langle Q h_1, h_2\right\rangle_H $$ 则$\left(B_Q^\hbar(t), t \geqslant 0\right) $是一个Hurst参数为$\hbar \in(0, 1) $的分数Q −布朗运动。特别地,如果TrQ < ∞,则$ B_Q^\hbar(t)$被叫做Hurst参数为$\hbar \in(0, 1) $的(真)分数布朗运动;如果Q为单位算子,这个过程被叫做Hurst参数为$\hbar \in(0, 1) $的圆柱形分数布朗运动。当$\hbar=1 / 2 $时,分数布朗运动成为标准的布朗运动。
假设存在一个序列{ λk},k ∈ N使得Qek = λkek,其中{ek,k ∈ N}是一个在H中的标准正交基。设对于任意的n ∈ N,t ≥ 0,有$\beta_k^\hbar(t)=\frac{1}{\sqrt{\lambda_k}} B_Q^\hbar(t), t \geqslant 0 $,其中$\left\{\beta_k^\hbar, k \in \mathbf{N}\right\} $是一个相互独立的分数布朗运动序列,并且$B_Q^\hbar(t) $可以统一地表示为
$$ B_Q^\hbar(t)=\sum\limits_{k=1}^{\infty} \sqrt{\lambda_k} \beta_k^\hbar(t) \boldsymbol{e}_k, t \geqslant 0 \text { 。 } $$ 如果Q是一个迹类算子,即TrQ < ∞,则BQℏ (t) 是一个H −值随机过程。
引理1.2[16] 令HQ = Ran (Q1/2) ⊂ H,在此空间定义内积为
$$ \langle u, v\rangle_{H_Q}=\left\langle Q^{-1 / 2} u, Q^{-1 / 2} v\right\rangle_H, \forall u, v \in H。 $$ 设$\mathcal{L}_2\left(H_Q, H\right) $表示含有所有从HQ到H的希尔伯特施密特算子,则$\mathcal{L}_2\left(H_Q, H\right) $是一个可分的希尔伯特空间,对于任意的$ \varPsi \in \mathcal{L}_2\left(H_Q, H\right)$有$\|\varPsi\|_{\mathcal{L}_2\left(H_Q, H\right)}^2= \left\|\varPsi Q^{1 / 2}\right\|_{H_Q}=\operatorname{Tr}\left(\varPsi Q \varPsi^*\right)$。
2. 随机卷积
本文考虑线性随机泛函微分方程
$$ \left\{\begin{aligned} \mathrm{d} y(t)= & A y(t) \mathrm{d} t+\int_{-r}^0 a(\theta) A_1 y(t+\theta) \mathrm{d} \theta \mathrm{d} t+B \mathrm{d} B_Q^h(t), \\ & t \geqslant 0, \\ y(0)= & \phi_0, y(\theta)=\phi_1(\theta), \theta \in[-r, 0], \phi=\left(\phi_0, \phi_1\right), \end{aligned}\right. $$ (5) 式中: $B \in \mathcal{L}_2\left(H_Q, H\right) $; BQℏ (t) 表示Hurst参数为ℏ (1/2 < ℏ < 1) 的分数布朗运动。
方程中唯一的mild解可表示为[16]
$$ y(t)=\int_{-r}^0 U_t(\theta) \phi_1(\theta) \mathrm{d} \theta+\int_0^t G(t-s) B \mathrm{d} B_Q^\hbar(t), t \geqslant 0, $$ 式中$ U_t(\theta)=\int_{-r}^\theta G(t-\theta+\tau) a(\tau) \mathrm{d} \tau, \theta \in[-r, 0]$。当ϕ = (0,0)时,方程的唯一解叫做随机卷积过程
$$ y(t):=B_G(t)=\int_0^t G(t-s) B \mathrm{d} B_Q^\hbar(t), t \geqslant 0 \text { 。 } $$ 引理2.1[14] 对于任意的$ \varphi:[0, T] \rightarrow \mathcal{L}_2\left(H_Q, H\right)$使得$ \sum\limits_{n=1}^{\infty}\left\|\varphi Q^{1 / 2} \boldsymbol{e}_n\right\|_{L^{1 / \hbar}([0, T] ; H)}<\infty$,以及α,β ∈ [0,T],α > β,有
$$ \begin{gathered} \mathrm{E}\left\|\int_\alpha^\beta \varphi(\mathrm{s}) \mathrm{d} B_Q^\hbar(s)\right\|_H^2 \leqslant \\ C_h \hbar(2 \hbar-1)(\alpha-\beta)^{2 \hbar-1} \sum\limits_{n=1}^{\infty} \int_\alpha^\beta\left\|\varphi(s) Q^{1 / 2} \boldsymbol{e}_n\right\|_H^2 \mathrm{d} s 。 \end{gathered} $$ (6) 若任意的$ t \in[0, T], \sum\limits_{n=1}^{\infty}\left\|\varphi Q^{1 / 2} \boldsymbol{e}_n\right\|_H$是一致有界的,则
$$ \begin{gathered} \mathrm{E}\left\|\int_a^\beta \varphi(s) \mathrm{d} B_Q^\hbar(s)\right\|_H^2 \leqslant \\ C_h \hbar(2 \hbar-1)(\alpha-\beta)^{2 \hbar-1} \int_\alpha^\beta\|\varphi(s)\|_{\mathcal{L}_2\left(H_Q, H\right)}^2 \mathrm{d} s 。 \end{gathered} $$ (7) 定理2.1 设T > 0并且TrQ < ∞,假设引理1.1的条件成立,即
$$ \sum\limits_{k=1}^{\infty} \lambda_k\left|\frac{C_{n, \gamma}}{t^\gamma}\right|<\infty, $$ 并且$\sum\limits_{k=1}^{\infty} \lambda_k\left|\frac{C_{n, \gamma}}{t^\gamma}\right| $关于t ∈ [0,T]是一致有界的,其中Cn,γ > 0,n ∈ 1,2,…,是一个依赖于n,γ的常数,则对于任意的$0<\gamma<\frac{1}{2}, 0<\beta<\frac{1}{2}-\gamma, \frac{1}{2}<\hbar<1 $,BG的轨道在$ C^{\beta-}\left([0, T), \operatorname{Dom}(-A)^\gamma\right)$中,其中
$$ C^{\beta-}:=\bigcap\limits_{m \geqslant 1} C^{\beta-\frac{1}{2 m}} 。 $$ 证明 任意的0 ≤ s < t ≤ T,根据BG (t) 的定义,可以得到
$$ \begin{gathered} (-A)^\gamma B_G(t)-(-A)^\gamma B_G(s)= \\ \int_0^t(-A)^\gamma G(t-u) B \mathrm{d} B_Q^\hbar(u)- \\ \int_0^s(-A)^\gamma G(s-u) B \mathrm{d} B_Q^\hbar(u)= \\ \int_s^t(-A)^\gamma G(t-u) B \mathrm{d} B_Q^\hbar(u)+ \\ \int_0^s(-A)^\gamma(G(t-u)-G(s-u)) B \mathrm{d} B_Q^\hbar(u), \end{gathered} $$ 则
$$ \begin{gathered} \mathrm{E}\left\|(-A)^\gamma B_G(t)-(-A)^\gamma B_G(s)\right\|^2 \leqslant \\ 2 \mathrm{E}\left\|\int_s^t(-A)^\gamma G(t-u) B \mathrm{d} B_Q^\hbar(u)\right\|^2+ \\ 2 \mathrm{E}\left\|\int_0^s(-A)^\gamma(G(t-u)-G(s-u)) B \mathrm{d} B_Q^\hbar(u)\right\|^2:= \\ I_1(t, s)+I_2(t, s)。 \end{gathered} $$ (8) 首先,利用公式和估计I1 (t,s),
$$ \begin{gathered} I_1(t, s) \leqslant \\ 2 C_{\hbar} \hbar(2 \hbar-1)(t-s)^{2 \hbar-1} \int_s^t\left\|(-A)^\gamma G(t-u) B\right\|_{L_2\left(H_Q, H\right)}^2 \mathrm{d} u \leqslant \\ 2 C_h \hbar(2 \hbar-1)(t-s)^{2 \hbar-1} \int_s^t\left\|(-A)^\gamma G(t-u)\right\|_H^2\|B\|_{L_2\left(H_Q, H\right)}^2 \mathrm{d} u \leqslant \\ 2 M C_{\hbar} \hbar(2 \hbar-1)(t-s)^{2 \hbar-1} \int_0^{t-s}\left\|(-A)^\gamma G(v)\right\|_H^2 \mathrm{d} v \leqslant \\ 2 M C_{\hbar} \hbar(2 \hbar-1)(t-s)^{2 \hbar-1} \int_0^{t-s} \frac{C_{0, \gamma}^2}{v^{2 \gamma}} \mathrm{d} v= \\ \frac{2 M C_{\hbar} \hbar(2 \hbar-1) C_{0, \gamma}^2}{1-2 \gamma}(t-s)^{2 \hbar-1}(t-s)^{1-2 \gamma}= \\ C_{0, \gamma, \hbar}(t-s)^{2(\hbar-\gamma)}, \end{gathered} $$ 式中$C_{0, \gamma, \hbar}>0 $为一个依赖于γ,$ \hbar$的常数。
其次,利用公式和估计$ {I}_{2}(t, s) $,当$ s\le r $时,
$$ \begin{array}{c}{I}_{2}\left(t, s\right)\le 2{C}_{\hslash }'\hslash (2\hslash -\left.1\right){s}^{2\hslash -1}{\int }_{0}^{s}{‖{(-A)}^{\gamma }\left(G\right(t-u)-G(s-u\left)\right)B‖}_{{L}_{2}({H}_{Q}, H)}^{2}\mathrm{d} u\le \\ 2{C}_{\hslash }'\hslash (2\hslash -\left.1\right){s}^{2\hslash -1}{\int }_{0}^{s}{‖{(-A)}^{\gamma }\left(G\right(t-u)-G(s-u\left)\right)B‖}_{H}^{2}{‖B‖}_{{L}_{2}({H}_{Q}, H)}^{2}\mathrm{d} u\le \\ 2M{C}_{\hslash }'\hslash (2\hslash -\left.1\right){s}^{2\hslash -1}{\int }_{0}^{s}{‖{(-A)}^{\gamma }\left(G\right(t-s+v)-G(v\left)\right)‖}_{H}^{2}\mathrm{d} v\le \\ 2M{C}_{\hslash }'\hslash (2\hslash -\left.1\right){s}^{2\hslash -1}{\int }_{0}^{s}\frac{{C}_{1, \beta , \gamma }^{2}{(t-s)}^{2\beta }}{{v}^{2\beta +2\gamma }}\mathrm{d} v\le \\ \frac{2M{C}_{\hslash }'\hslash (2\hslash -\left.1\right){C}_{1, \beta , \gamma }^{2}}{1-2\beta -2\gamma }{s}^{2\hslash -1}{s}^{1-2\beta -2\gamma }{(t-s)}^{2\beta }\le \\ \frac{2M{C}_{\hslash }'\hslash (2\hslash -\left.1\right){C}_{1, \beta , \gamma }^{2}}{1-2\beta -2\gamma }{r}^{2\hslash -2\beta -2\gamma }{(t-s)}^{2\beta }\text { := }\\ {C}_{\hslash , r, \beta , \gamma }{(t-s)}^{2\beta }, \end{array} $$ 式中$ 1-2\beta -2\gamma > 0 $,$ {C}_{\hslash , r, \beta , \gamma } > 0 $是一个依赖于$ \hslash , r, \beta , \gamma $的常数。当$ s > r $时,
$$ \begin{array}{c}{I}_{2}(t, s)\le 2{C}_{\hslash }'\hslash (2\hslash -\left.1\right){s}^{2\hslash -1}{\int }_{0}^{s}{‖{(-A)}^{\gamma }\left(G\right(t-u)-G(s-u\left)\right)B‖}_{{L}_{2}({H}_{Q}, H)}^{2}\mathrm{d} u\le \\ 2{C}_{\hslash }'\hslash (2\hslash -\left.1\right){s}^{2\hslash -1}{\int }_{0}^{s}{‖{(-A)}^{\gamma }\left(G\right(t-u)-G(s-u\left)\right)B‖}_{H}^{2}{‖B‖}_{{L}_{2}({H}_{Q}, H)}^{2}\mathrm{d} u\le \\ 2M{C}_{\hslash }'\hslash (2\hslash -\left.1\right){s}^{2\hslash -1}{\int }_{0}^{s}{‖{(-A)}^{\gamma }\left(G\right(t-s+v)-G(v\left)\right)‖}_{H}^{2}\mathrm{d} v\le \\ 2M{C}_{\hslash }'\hslash (2\hslash -\left.1\right){s}^{2\hslash -1}\left[{\int }_{0}^{r}{‖{(-A)}^{\gamma }\left(G\right(t-s+v)-G(v\left)\right)‖}_{H}^{2}\mathrm{d} v\right.+\left.{\int }_{r}^{s}{‖{(-A)}^{\gamma }\left(G\right(t-s+v)-G(v\left)\right)‖}_{H}^{2}\mathrm{d} v\right]\le \\ 2M{C}_{\hslash }'\hslash \left(2\hslash -1\right){s}^{2\hslash -1}\left[\frac{{C}_{1, \beta , \gamma }^{2}{r}^{1-2\beta -2\gamma }}{1-2\beta -2\gamma }+\frac{{C}_{n, \beta , \gamma }^{2}{r}^{1-2\beta }}{1-2\beta }\right]{(t-s)}^{2\beta }\le \\ 2M{C}_{\hslash }'\hslash \left(2\hslash -1\right){T}^{2ħ-1}\left[\frac{{C}_{1, \beta , \gamma }^{2}{r}^{1-2\beta -2\gamma }}{1-2\beta -2\gamma }+\frac{{C}_{n, \beta , \gamma }^{2}{r}^{1-2\beta }}{1-2\beta }\right]{(t-s)}^{2\beta }\text { := }{C}_{\hslash , r, \beta , \gamma , T}{(t-s)}^{2\beta }, \end{array} $$ 式中$ {C}_{\hslash , r, \beta , \gamma , T} > 0 $是一个依赖于$ \hslash , r, \beta , \gamma , T $的常数。
根据公式,对于任意的$ 0\le s < t\le T $,有
$$ \begin{array}{c}E{‖{(-A)}^{\gamma }{B}_{G}\left(t\right)-{(-A)}^{\gamma }{B}_{G}\left(s\right)‖}^{2}\le \\ {C}_{0, \gamma , \hslash }{(t-s)}^{2(\hslash -\gamma )}+{C}_{\hslash , r, \beta , \gamma , T}{(t-s)}^{2\beta }=\\ {C}_{0, \gamma , \hslash }{(t-s)}^{2(\hslash -\gamma )-2\beta }{(t-s)}^{2\beta }+{C}_{\hslash , r, \beta , \gamma , T}{(t-s)}^{2\beta }\le \\ {C}_{0, \gamma , \hslash }{T}^{2(\hslash -\gamma )-2\beta }{(t-s)}^{2\beta }+{C}_{\hslash , r, \beta , \gamma , T}{(t-s)}^{2\beta }\text { := }\\ {{C}'}_{\hslash , r, \beta , \gamma , T}{(t-s)}^{2\beta }, \end{array} $$ 式中$ {C}_{\hslash , r, \beta , \gamma , T}' > 0 $是一个依赖于$ \hslash , r, \beta , \gamma , T $的常数。
因为$ {(-A)}^{\gamma }{B}_{G}\left(t\right)-{(-A)}^{\gamma }{B}_{G}\left(s\right) $是高斯过程,则对于任意的$ m\ge 1 $,存在常数$ {C}_{m} > 0 $使得
$$ E{‖{(-A)}^{\gamma }{B}_{G}\left(t\right)-{(-A)}^{\gamma }{B}_{G}\left(s\right)‖}^{2m}\le {C}_{m}({{C}'}_{\hslash , r, \beta , \gamma , T}{)}^{m}{(t-s)}^{2m\beta }, $$ 利用文献[20]中定理3.3的柯尔莫哥洛夫检验,$ {(-A)}^{\gamma }{B}_{G}(·) $是$ \alpha - $Hölder连续的,阶为
$$ \alpha =\frac{2m\beta -1}{2m}=\beta -\frac{1}{2m}, $$ 令$ m\to \infty $,则$ \alpha \to \beta $,所以过程$ {(-A)}^{\gamma }{B}_{G} $的轨道在$ {C}^{\beta -}\left(\right[0, T], H) $中,其中$ {C}^{\beta -}\text { := }\underset{m\ge 1}{\cap }{C}^{\beta -\frac{1}{2m}} $,即$ {B}_{G} $的轨道在$ {C}^{\beta -}\left(\right[0, T], \operatorname{Dom}{(-A)}^{\gamma }) $中。
3. 随机卷积的正则性
本节假设在方程中,存在$ {\mu }_{n} > 0 $,$ \bar{\mu}_n \in \mathbf{R} $,$ k \in \mathbf{N}$,使得$ A \boldsymbol{e}_k=-\mu_k \boldsymbol{e}_k $,$ A_1 \boldsymbol{e}_n=\bar{\mu}_n \boldsymbol{e}_n $。对于任意的$ \alpha , \beta \in \mathbf{R} $,设$ g\left(t\right) $,$ t\ge 0 $表示以下方程的唯一解
$$ \left\{\begin{array}{l}g'\left(t\right)+\alpha g\left(t\right)=\beta {\int }_{-r}^{0}a\left(\theta \right)g(t+\theta )\mathrm{d} \left(\theta \right), t\ge 0, \\ g\left(0\right)=1, g\left(\theta \right)=0, \theta \in [-r, 0], \end{array}\right. $$ 则基本解算子$ G $可表示为[19]
$$ G\left(t\right){\boldsymbol{e}}_{k}=g({\mu }_{k}, t){\boldsymbol{e}}_{k}。 $$ 引理3.1[19] 对于任意的常数$ \gamma \in \left(\mathrm{0, 1}\right) $,$ 0\le s < t\le nr $,$ n\in \mathbf{N} $,$ k\in \mathbf{N} $,有
$$ {\int }_{s}^{t}{g}^{2}({\mu }_{k}, u)\mathrm{d} u\le {M}_{n, \gamma }·{{\mu }_{k}}^{-\gamma }{(t-s)}^{1-\gamma }, $$ (9) 式中$ {M}_{n, \gamma } > 0 $是一个依赖于$ n, \gamma $的常数。
引理3.2[19] 设$ \gamma \in \left(\mathrm{0, 1}\right) $并且$ 0 < \beta < 1-\gamma $,则对于任意的$ 0\le s < t\le nr $,$ n\in \mathbf{N} $,$ k\in \mathbf{N} $,有
$$ {\int }_{0}^{s}\left[g\right({\mu }_{k}, t-u)-g({\mu }_{k}{, s-u\left)\right]}^{2}\mathrm{d} u\le {M}_{n, \gamma , \beta }·{{\mu }_{k}}^{-\gamma }{(t-s)}^{\beta }, $$ (10) 式中$ {M}_{n, \gamma , \beta } > 0 $是一个依赖于$ n, \gamma , \beta $的常数。
定理3.1 假设$ \gamma \in \left(\mathrm{0, 1}\right) $,$ \sum\limits_{k=1}^{\infty }{\lambda }_{k}\left|\frac{{C}_{n, \gamma }}{{t}^{\gamma }}\right| < \infty $,其中$ {C}_{n, \gamma } > 0 $,$ n\in \mathrm{1, 2}, \dots $,是一个依赖于$ n, \gamma $的常数,并且
$$ \sum\limits_{k=1}^{\infty }\frac{{\lambda }_{k}}{{\mu }_{k}^{\gamma }} < \infty 。 $$ (11) 则$ H- $值过程$ {B}_{G}\left(t\right) $,$ t\ge 0 $的轨道是几乎处处$ \alpha - $ Hölder连续的,$ 0 < \alpha < \frac{1-\gamma }{2} $。
证明 任意的$ 0\le s < t\le nr $,$ n\in \mathbf{N} $,$ k\in \mathbf{N} $,有
$$ \begin{array}{c}\mathrm{E}{‖{B}_{G}\left(t\right) -{B}_{G}\left(s\right)‖}^{2}=\\ \mathrm{E}{‖{\int }_{s}^{t}G\left(t-u\right)B\mathrm{d} {B}_{Q}^{\hslash }\left(u\right) +{\int }_{0}^{s}\left(G\left(t-u\right)-G\left(s-u\right)\right)B\mathrm{d} {B}_{Q}^{\hslash }\left(u\right)‖}^{2}\le \\ 2\mathrm{E}{‖{\int }_{s}^{t}G\left(t-u\right)B\mathrm{d} {B}_{Q}^{\hslash }\left(u\right)‖}^{2}+ 2\mathrm{E}‖{\int }_{0}^{s}\left(G\left(t-u\right)-\right.{\stackrel{}{\underset{}{B\mathrm{d} {B}_{Q}^{\hslash }\left(u\right)}}‖}^{2}\text { := }\\ {J}_{1}(s, t) +{J}_{2}(s, t)。\end{array} $$ 首先,根据公式和估计$ {J}_{1}(s, t) $,
$$ \begin{array}{c}{J}_{1}\left(s, t\right)\le 2{C}_{\hslash }\hslash (2\hslash -\left.1\right){(t-s)}^{2\hslash -1}·\\ \sum\limits_{k=1}^{\infty }{\int }_{s}^{t}{‖G\left(t-u\right)B{Q}^{1/2}{\boldsymbol{e}}_{k}‖}_{H}^{2}\mathrm{d} u\le \\ 2{C}_{\hslash }\hslash (2\hslash -\left.1\right){(t-s)}^{2\hslash -1}\\ \sum\limits_{k=1}^{\infty }{\int }_{s}^{t}{‖G\left(t-u\right){Q}^{1/2}{\boldsymbol{e}}_{k}‖}_{H}^{2}{‖B‖}_{{L}_{2}({H}_{Q}, H)}^{2}\mathrm{d} u\le \\ 2M{C}_{\hslash }\hslash (2\hslash -\left.1\right){(t-s)}^{2\hslash -1}·\sum\limits_{k=1}^{\infty }{\lambda }_{k}{\int }_{s}^{t}{‖G\left(t-u\right){\boldsymbol{e}}_{k}‖}_{H}^{2}\mathrm{d} u\le \\ 2M{C}_{\hslash }\hslash (2\hslash -\left.1\right){(t-s)}^{2\hslash -1}\sum\limits_{k=1}^{\infty }{\lambda }_{k}{\int }_{0}^{t-s}{‖G\left(v\right){\boldsymbol{e}}_{k}‖}^{2}\mathrm{d} v\le \\ 2M{C}_{\hslash }\hslash (2\hslash -\left.1\right){(t-s)}^{2\hslash -1}\sum\limits_{k=1}^{\infty }{\lambda }_{k}{\int }_{0}^{t-s}{‖g\left({\mu }_{k}, v\right){\boldsymbol{e}}_{k}‖}^{2}\mathrm{d} v\le \\ 2M{C}_{\hslash }\hslash \left(2\hslash -1\right){\left(t-s\right)}^{2\hslash -1}{C}_{n, \gamma }\left(\sum\limits_{k=1}^{\infty }\frac{{\lambda }_{k}}{{\mu }_{k}^{\gamma }}\right){\left(t-s\right)}^{1-\gamma }=\\ 2M{C}_{\hslash }\hslash (2\hslash -\left.1\right){C}_{n, \gamma }\left(\sum\limits_{k=1}^{\infty }\frac{{\lambda }_{k}}{{\mu }_{k}^{\gamma }}\right){(t-s)}^{2\hslash -\gamma }\le \\ 2M{C}_{\hslash }\hslash (2\hslash -\left.1\right){C}_{n, \gamma }\left(\sum\limits_{k=1}^{\infty }\frac{{\lambda }_{k}}{{\mu }_{k}^{\gamma }}\right){\left(nr\right)}^{2\hslash -\gamma -\beta }{(t-s)}^{\beta }\text { := }\\ {M}_{n, \beta , \gamma , \hslash }{(t-s)}^{\beta }, \end{array} $$ 式中$ {M}_{n, \beta , \gamma , \hslash } > 0 $是一个依赖于$ n, \beta , \gamma , \hslash $的常数,$ 0 < \beta < 1-\gamma $。
其次,根据式(6和10)估计$ {J}_{2}(s, t) $,
$$ \begin{array}{c}{J}_{2}(s, t)\le 2{C}_{\hslash }\hslash (2\hslash -\left.1\right){s}^{2\hslash -1}·\\ \sum\limits_{k=1}^{\infty }{\int }_{0}^{s}{‖\left[G\left(t-u\right)-G\left(s-u\right)\right]B{Q}^{1/2}{\boldsymbol{e}}_{k}‖}_{H}^{2}\mathrm{d} u\le \\ 2{C}_{\hslash }\hslash (2\hslash -\left.1\right){s}^{2\hslash -1}·\\ \sum\limits_{k=1}^{\infty }{\int }_{0}^{s}{‖\left[G\left(t-u\right)-G\left(s-u\right)\right]{Q}^{1/2}{\boldsymbol{e}}_{k}‖}_{H}^{2}{‖B‖}_{{L}_{2}({H}_{Q}, H)}^{2}\mathrm{d} u\le \\ 2{C}_{\hslash }\hslash (2\hslash -\left.1\right){s}^{2\hslash -1}·\\ \sum\limits_{k=1}^{\infty }{\lambda }_{k}{\int }_{0}^{s}{‖\left[G\left(t-u\right)-G\left(s-u\right)\right]{\boldsymbol{e}}_{k}‖}^{2}\mathrm{d} u=\\ 2{C}_{\hslash }\hslash (2\hslash -\left.1\right){s}^{2\hslash -1}·\\ \sum\limits_{k=1}^{\infty }{\lambda }_{k}{\int }_{0}^{s}{‖\left[g\left({\mu }_{k}, t-u\right)-g\left({\mu }_{k}, s-u\right)\right]{\boldsymbol{e}}_{k}‖}^{2}\mathrm{d} u\le \\ 2{C}_{\hslash }\hslash (2\hslash -\left.1\right){s}^{2\hslash -1}{C}_{n, \beta , \gamma }\left(\sum\limits_{k=1}^{\infty }\frac{{\lambda }_{k}}{{\mu }_{k}^{\gamma }}\right){(t-s)}^{\beta }\le \\ 2{C}_{\hslash }\hslash \left(2\hslash -1\right){\left(nr\right)}^{2\hslash -1}{C}_{n, \beta , \gamma }\left(\sum\limits_{k=1}^{\infty }\frac{{\lambda }_{k}}{{\mu }_{k}^{\gamma }}\right){(t-s)}^{\beta }\text { := }\\ {M}_{n, \beta , \gamma , \hslash }'{(t-s)}^{\beta }, \end{array} $$ 式中$ {M}_{n, \beta , \gamma , \hslash }' > 0 $是一个依赖于$ n, \beta , \gamma , \hslash $的常数,$ 0 < \beta < 1-\gamma $。
因为$ {B}_{G}\left(t\right)-{B}_{G}\left(s\right) $是高斯过程,则对于任意的$ m\ge 1 $,存在常数$ {C}_{m}' > 0 $使得
$$ \mathrm{E}{‖\left({B}_{G}\right(t)-{B}_{G}(s)‖}^{2m}\le {C}_{m}'({M}_{n, \beta , \gamma , \hslash }+{M}_{n, \beta , \gamma , \hslash }'{)}^{m}{(t-s)}^{m\beta }, $$ 利用文献[20]定理3.3的柯尔莫哥洛夫检验,$ {B}_{G}(·) $在$ H $中是$ \alpha - $Hölder连续的,并且阶为
$$ \alpha =\frac{m\beta -1}{2m}=\frac{\beta }{2}-\frac{1}{2m}, $$ 令$ m\to \infty $,则$ \alpha \to \frac{\beta }{2} $,所以$ {B}_{G}\left(t\right) $的轨道是几乎处处意义下$ \alpha - $Hölder连续的。当$ \beta \to 0 $时,$ \alpha \to 0 $,当$ \beta \to 1-\gamma $时,$ \alpha \to \frac{1-\gamma }{2} $,则$ 0 < \alpha < \frac{1-\gamma }{2} $。过程$ {B}_{G} $的轨道在$ {C}^{\frac{\beta }{2}-}\left(\right[0, T], H) $中,其中$ {C}^{\frac{\beta }{2}-}\text { := }\underset{m\ge 1}{\cap }{C}^{\frac{\beta }{2}-\frac{1}{2m}} $。
4. 结束语
本文研究了一类由分数布朗运动驱动的带分布延迟的线性随机泛函微分方程的随机卷积的正则性,证明了在一定条件下带分布延迟的随机卷积的轨道在一个连续空间中,随机卷积的轨道具有Hölder连续正则性。本文的研究方法可以推广到由分数布朗运动驱动的中立型随机泛函微分方程,利用基本解的估计建立其随机卷积的正则性。
-
[1] HAN X, KLOEDEN P E. Random ordinary differential equations and their numerical solution[M]. Singapore: Springer, 2017. [2] SOBCZYK K. Stochastic differential equations with applications to physics and engineering[M]. Germamy: Springer Science and Business Media, 2001. [3] MAO X. Stochastic differential equations and applications[M]. London: Elsevier, 2007. [4] LIU K. Stability in distribution for stochastic differential equations with memory driven by positive semigroups and Lévy processes[J]. Applied Mathematics and Computation, 2019, 362: 124580. doi: 10.1016/j.amc.2019.124580 [5] 余国胜. 无穷时滞脉冲随机泛函微分方程一般衰减意义下P阶矩稳定性[J]. 湖北大学学报(自然科学版), 2020, 42(4): 359-362. [6] 徐丽平. 分数布朗运动驱动的随机泛函微分方程的相关研究[D]. 广州: 广州大学, 2019. [7] DUNCAN T E, PASIK-DUNCAN B, MASLOWSKI B. Fractional Brownian motion and stochastic equations in Hilbert spaces[J]. Stochastics and Dynamics, 2002, 2(2): 225-250. doi: 10.1142/S0219493702000340 [8] COMTE F, RENAULT E. Long memory continuous time models[J]. Journal of Econometrics, 1996, 73(1): 101-149. doi: 10.1016/0304-4076(95)01735-6 [9] BOUDRAHEM S, ROUGIER P R. Relation between postural control assessment with eyes open and centre of pressure visual feed back effffects in healthy individuals[J]. Experimental Brain Research, 2009, 195(1) : 145-152. doi: 10.1007/s00221-009-1761-1 [10] RYPDAL M, RYPDAL K. Testing hypotheses about sunclimate complexity linking[J]. Physical Review Letters, 2010, 104(12): 128-151. [11] SIMONSEN I. Measuring anti-correlations in the nordic electricity spot market by wavelets[J]. Physica A, 2003, 322: 597-606. doi: 10.1016/S0378-4371(02)01938-6 [12] BAHUGUNA D, SAKTHIVEL R, CHADHA A. Asymptotic stability of fractional impulsive neutral stochastic partial integro-difffferential equations with infifinite delay[J]. Stochastic Analysis and Applications, 2017, 35(1): 63-68. doi: 10.1080/07362994.2016.1249285 [13] BOUFOUSSI B, HAJJI S. Neutral stochastic functional difffferential equations driven by a fractional Brownian motion in a Hilbert space[J]. Statistics and Probability Letters, 2012, 82(8): 1549-1558. doi: 10.1016/j.spl.2012.04.013 [14] CARABALLO T, GARRIDO-ATIENZA M J, TANIGUCHI T. The existence and exponential behavior of solutions to stochastic delay evolution equations with a fractional Brownian motion[J]. Nonlinear Analysis, 2011, 74(11): 3671-3684. doi: 10.1016/j.na.2011.02.047 [15] LI Z, YAN L. Ergodicity and stationary solution for stochastic neutral retarded partial differential equations driven by fractional Brownian motion[J]. Journal of Theoretical Probability, 2019, 32(3): 1399-1419. doi: 10.1007/s10959-018-0810-8 [16] XU L, LI Z, LUO J. Global attracting set and exponential decay of second-order neutral stochastic functional differential equations driven by fBm[J]. Advances in Difference Equations, 2017, 2017(1): 1-16. doi: 10.1186/s13662-016-1057-2 [17] LIU K. On regularity of stochastic convolutions of functional linear differential equations with memory[J]. Discrete & Continuous Dynamical Systems-B, 2019, 25(4): 2019220. [18] LIU K. Some regularity results on stochastic convolutions in point delay differential equations perturbed by noise[J]. Statistics and Probability Letters, 2020, 156: 108624. doi: 10.1016/j.spl.2019.108624 [19] LIU K. Hölder continuous regularity of stochastic convolutions with distributed delay[J]. Potential Analysis, 2022, 57(1): 29-53. doi: 10.1007/s11118-021-09904-5 [20] DAPRATO G, ZABCZYK J. Stochastic equations in infinite dimensions[M]. Cambridge: Cambridge University Press, 2014: 67-78.