Abstract
In this paper, we study a class of semilinear parabolic equation and its perturbed system driven by a random force. Such driving noise is assumed to be a regular approximation to the white noise and satisfy certain properties. We show that the \(C^1\) inertial manifold structure is persisted under such perturbation in the sense that inertial manifolds of the perturbed system are converging to those of the original system as the perturbation tends to zero.
Similar content being viewed by others
References
Arnold, L.: Random Dynamical Systems. Springer, New York (1998)
Boutet De Monvel, L., Chueshov, I.D., Rezounenko, A.V.: Inertial manifolds for retarded semilinear parabolic equations. Nonlin. Anal. 34, 907–925 (1998)
Chueshov, I.D.: Introduction to the Theory of Inertial Manifolds (Lecture Notes). Kharkov University Press, Kharkov (1992). ((in Russian))
Chueshov, I.D.: Approximate inertial manifolds of exponential order for semilinear parabolic equations subjected to additive white noise. J. Dynam. Differ. Equ. 7, 549–566 (1995)
Chueshov, I.: Monotone Random Dynamical Systems Theory and Applications. Springer-Verlag, Berlin (2002)
Chueshov, I.D., Scheutzow, M.: Inertial manifolds and forms for stochastically perturbed retarded semilinear parabolic equations. J. Dynam. Differ. Equ. 13, 355–380 (2001)
Chueshov, I.D., Scheutzow, M., Schmalfuss, B.: Continuity properties of inertial manifolds for stochastic retarded semilinear parabolic equations, pp. 355–377. Springer, Berlin Heidelberg, Interacting Stochastic Systems (2005)
Caraballo, T., Kloeden, P.E., Schmalfuß, B.: Exponentially stable stationary solutions for stochastic evolution equations and their perturbation. Appl. Math. Optim. 50, 183–207 (2004)
Castaing, C., Valadier, M.: Convex Analysis and Measurable Multifunctions. Lecture Notes in Mathematics, vol. 580. Springer, Berlin-New York (1977)
Chow, S.-N., Lu, K.: Smoothness of Inertial Manifolds. J. Math. Anal. Appl. 169, 283–312 (1992)
Constantin, P., Foias, C., Nicolaenko, B., Temam, R.: Integral Manifolds and Inertial Manifolds for Dissipative Partial Differential Equations. Springer-Verlag, New York (1989)
Da Prato, G., Debussche, A.: Construction of stochastic inertial manifolds using backward integration. Stoch. Int. J. Probability Stoch. Process. 59, 305–324 (1996)
Duan, J., Lu, K., Schmalfuß, B.: Invariant manifolds for stochastic partial differential equations. Ann. Probab. 31, 2109–2135 (2003)
Foias, C., Sell, G.R., Temam, R.: Inertial manifolds for nonlinear evolutionary equations. J. Differ. Equ. 73, 309–353 (1988)
Gu, A.: Asymptotic behavior of random lattice dynamical systems and their Wong-Zakai approximations. Disc. Cont. Dyn. Sys. 24, 5737–5767 (2019)
Gu, A., Wang, B.: Asymptotic behavior of random FitzHugh-Nagumo systems driven by colored noise. Disc. Cont. Dyn. Sys.-Series B 23, 1689–1720 (2018)
Guo, S.: On the mollifier approximation for solutions of stochastic differential equations. J. Math. Kyoto Univ. 22, 243–254 (1982)
Henry, D.: Geometric Theory of Semilinear Parabolic Equations. Lecture Notes in Mathematics, vol. 840. Springer-Verlag, New York (1981)
Ikeda, N., Watanabe, S.: Stochastic Differential Equations and Diffusion Processes, 2nd edn. Noth-Holland (1989)
Jiang, T., Liu, X., Duan, J.: A Wong-Zakai approximation for random invariant manifolds. J. Math. Phys. 58, 122701 (2017)
Koksch, N., Siegmund, S.: Pullback attracting inertial manifolds for nonautonomous dynamical systems. J. Dynam. Differ. Equ. 14, 889–941 (2002)
Kunita, H.: Stochastic Flows and Stochastic Differential Equations. Cambridge University Press, Cambridge (1990)
Lu, K., Wang, Q.: Chaotic behavior in differential equations driven by a Brownian motion. J. Differ. Equ. 251, 2853–2895 (2011)
Lu, K., Wang, B.: Wong-Zakai approximations and long term behavior of stochastic partial differential equations. J. Dynam. Differ. Equ., 1–31 (2017)
Malliavin, P.: Stochastic calculus of variation and hypoelliptic operators, Proc. Int. Symp. on S.D.E. Kyoto 1976 (ed. by K. Ito), 195-263, Kinokuniya, (1978)
Pazy, A.: Semigroups of Linear Operators and Applications to Partial Differential Equations. Springer-Verlag, New York (1983)
Shen, J., Lu, K.: Wong-Zakai Approximations and center manifolds of stochastic differential equations. J. Differ. 263, 4929–4977 (2017)
Shen, J., Lu, K., Wang, B.: Convergence and center manifolds for differential equations driven by colored noise, Disc. Cont. Dyn. Sys.-Series A 2019, 4797–4840
Shen, J., Lu, K., Zhang, W.: Heteroclinic chaotic behavior driven by a Brownian motion. J. Differ. Equ. 255, 4185–4225 (2013)
Shen, J., Zhao, J., Lu, K., Wang, B.: The Wong-Zakai approximations of invariant manifolds and foliations for stochastic evolution equation. J. Differ. Equ. 266, 4568–4623 (2019)
Temam, R.: Infinite-dimensional dynamical systems in mechanics and physics, Infinite-dimensional dynamical systems in mechanics and physics (2000)
Wang, X., Lu, K., Wang, B.: Wong-Zakai approximations and attractors for stochastic reaction-diffusion equations on unbounded domains. J. Differ. Equ. 264, 378–424 (2018)
Author information
Authors and Affiliations
Corresponding author
Additional information
Dedicated to the memory of Pavol Brunovsky.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
This work was supported by NSFC (11831012, 11971330, 12001449, 12090013, 12071317), the Fundamental Research Funds for the Central Universities (A0920502052101-86) and Sichuan Science and Technology Program (2020YJ0328). All correspondences should be addressed to Jun Shen.
Appendix: Some Proofs
Appendix: Some Proofs
We give two proofs related to the hypotheses on the noise given in the Introduction. First we show that (1.8)
holds for \(\zeta _\delta (\theta _t\omega )={\mathcal {G}}_{\delta }(\theta _t \omega )=\frac{1}{\delta }(\omega (t+\delta )-\omega (t))\).
Proof
We observe that
It then follows that for all \(\delta \in (0,1]\),
where \(t^*=t^*(\omega ) \in (t, t+\delta )\). Along with (1.4), for any \(\zeta >0\), there exists a \(T_1=T_1(\omega )>0\) such that for every \(t<-T_1\), we have that
which yields that for all \(t \le -T_1-2\) and \(\delta \in (0,1]\), we have that
On the other hand, there exists a \(T_2=T_2(\omega )>1\) such that for all \(t \le -T_2\) and \(\delta \in (0,1]\),
Taking \(T_3=\max \{T_1+2, T_2\}\), for \(t \le -T_3\) and \(\delta \in (0,1]\) we get that
which implies the desired result. \(\square \)
Secondly, based on the fact that there exists a \(\theta _t\)-invariant subset of full measure \(\varOmega _1\) such that
we justify the properties of the stationary solution to the random differential equation
As we stated earlier, here we consider the probability space \((\varOmega _1, {\mathcal {F}}_1, {\mathbb {P}})\), where \({\mathcal {F}}_1\) is the trace algebra of \(\varOmega _1\). For simplicity, this space is still denoted by \((\varOmega , {\mathcal {F}}, {\mathbb {P}}).\)
Proposition A.1
For any \(T_1, T_2 \in {\mathbb {R}}\) with \(T_1<T_2\) and \(\delta >0\), the following statements hold:
-
(1)
The random variable
$$\begin{aligned} x_{\delta }^*(\omega )=\int _{-\infty }^{0} e^r \varPhi _\delta (\theta _r \omega )d r, \quad \omega \in \varOmega , \end{aligned}$$exists and generates a stationary solution of (A.2) given by
$$\begin{aligned} \varOmega \times {\mathbb {R}}\ni (\omega ,t)\rightarrow x_{\delta }^*(\theta _t\omega ) =\int _{-\infty }^{0} e^r \varPhi _\delta (\theta _{r+t} \omega )d r. \end{aligned}$$The mapping \(t\rightarrow x_\delta ^*(\theta _t\omega )\) is continuous.
-
(2)
On \(\varOmega \) we have
$$\begin{aligned} \lim _{t\rightarrow \pm \infty }\frac{|x_{\delta }^*(\theta _t\omega )|}{|t|}= & {} 0, \quad \lim _{t\rightarrow \pm \infty }\frac{1}{t}\int _0^t x_{\delta }^*(\theta _r \omega )\;d r=0, \end{aligned}$$uniformly with respect to \(\delta \in (0, 1]\).
-
(3)
For every \(\omega \in \varOmega \),
$$\begin{aligned} \lim _{\delta \rightarrow 0^+} \sup _{t\in [T_1, T_2]} |x_\delta ^*(\theta _{t} \omega )-x(\theta _{t} \omega )|=0, \end{aligned}$$where \(x(\theta _t \omega )=-\int _{-\infty }^0 e^r \theta _t \omega (r) dr\) is a stationary solution of stochastic differential equation
$$\begin{aligned} dx=-xdt+dW. \end{aligned}$$(A.3)
Proof
Let \(\omega \in \varOmega \). We first show (1) holds. Recalling that for each \(\delta >0\),
by (A.1) we have \( x_{\delta }^*(\omega )=\alpha \int _{-\infty }^{0} e^r \varPhi _\delta (\theta _r \omega )d r \) is well-defined. Note that \(x_\delta ^*(\theta _t\omega )\) can be written as
Then, \(x_\delta ^*(\theta _t\omega )\) satisfies equation (A.2) and is a stationary process.
(2) We observe that
Along with Fubini’s theorem, we get that
Taking the integral transformation \(u=r+s\) and \(v=\frac{s}{\delta }\) successively, we have that
By integration by parts, we further get that
We first show that \(\frac{|x_{\delta }^*(\theta _t\omega )|}{|t|}\rightarrow 0\) as \(t \rightarrow -\infty \), uniformly for \(\delta \in (0, 1]\). (A.1) implies that for any \(\varepsilon >0\), there exists a positive constant \({\hat{T}}_0={\hat{T}}(\omega , \varepsilon )\) such that for all \(|s| \ge {\hat{T}}_0\),
Then we have for \(t\le -{\hat{T}}_0-1\) and \(\delta \in (0, 1]\),
where (A.5) is used in the second inequality. Clearly, there exists a positive constant \({\hat{T}}_1\) such that as \(|t| \ge {\hat{T}}_1\), \(\frac{1}{|t|} < 1.\) Choosing \({\hat{T}}_2=\max \{{\hat{T}}_0, {\hat{T}}_1\}\), as \(t \le -{\hat{T}}_2\), for all \(\delta \in (0,1]\), we have that
Hence, \(\frac{|x_{\delta }^*(\theta _t\omega )|}{|t|}\rightarrow 0\) as \(t \rightarrow -\infty \), uniformly for \(\delta \in (0, 1]\).
Next we show that \(\frac{|x_{\delta }^*(\theta _t\omega )|}{|t|}\rightarrow 0\) as \(t \rightarrow +\infty \), uniformly for \(\delta \in (0, 1]\). Using (A.1) again, we find that there exists \({\hat{T}}_3 = {\hat{T}}_3(\omega )>0\) such that for \(|t| \ge {\hat{T}}_3\),
Meanwhile, by the continuity of \(\omega \in \varOmega \), there exists \(M(\omega )>0\) such that for \(|t| \le {\hat{T}}_3\)
Then we get for all \(t \in {\mathbb {R}}\),
By (A.4), we have
where \(T_*\) is a positive number to be specified later. For the first integral on the right hand side, we observe that for all \(\delta \in (0,1]\),
We choose a sufficiently large \(T_*>0\) such that \(2 e^{-T_*}<\varepsilon \) and note that there exists a \({\hat{T}}_4={\hat{T}}_4(\omega , \varepsilon )>0\) such that as \(|t|>{\hat{T}}_4\), \(\frac{2 M(\omega )+3}{|t|}<\varepsilon \). Thus as \(t>{\hat{T}}_4\), for all \(\delta \in (0,1]\) we have that
On the other hand, we note that for \(\delta \in (0, 1]\), \(|t+\delta v|>|t|-|\delta v|>|t|-1\) and \(|u+\delta v+t|>|t|-|u+\delta v|>|t|-(T_*+1)\) on \(u \in [-T_*, 0]\). It then follows that for \(t>T_*+1+{\hat{T}}_0\) and \(\delta \in (0, 1]\),
Combing (A.7) with (A.8), as \(t>\max \{{\hat{T}}_4,T_*+1+{\hat{T}}_0, {\hat{T}}_1 \}\), we find that for all \(\delta \in (0, 1]\),
This implies that \(\frac{|x_{\delta }^*(\theta _t\omega )|}{|t|}\rightarrow 0\) as \(t \rightarrow +\infty \), uniformly for \(\delta \in (0, 1]\).
We now show that \(\frac{1}{t}\int _0^t x_{\delta }^*(\theta _r \omega )d r\rightarrow 0\) as \(t \rightarrow -\infty \), uniformly for \(\delta \in (0, 1]\). By using (A.4) and Fubini’s theorem, we have that
Here we first restrict \(t<-1\). By applying Fubini’s theorem again, we find that
Then, as \(t<-1\) we have that
By using (A.5) and (A.6), for \(t\le -{\hat{T}}_0-1\) and \(\delta \in (0, 1]\) we have that
Clearly, as \(t<\max \{-{\hat{T}}_1, -{\hat{T}}_4, -{\hat{T}}_0-1\}\), for all \(\delta \in (0, 1]\), we find that \(|\frac{1}{t}\int _0^t x_{\delta }^*(\theta _r \omega )d r|<4\varepsilon \) and hence \(\frac{1}{t}\int _0^t x_{\delta }^*(\theta _r \omega )\;d r\rightarrow 0\) as \(t \rightarrow -\infty \), uniformly for \(\delta \in (0, 1]\).
In order to show \(\frac{1}{t}\int _0^t x_{\delta }^*(\theta _r \omega )d r\rightarrow 0\) as \(t \rightarrow +\infty \), uniformly for \(\delta \in (0, 1]\), we rewrite (A.9) as
By applying Fubini’s theorem, for \(t>0\) we investigate that
Then, for \(t>0\) we have that
Recall that we have chosen a \(T_{*}>0\) such that \(2 e^{-T_*}<\varepsilon \). Applying (A.6), as \(t>{\hat{T}}_4\), for all \(\delta \in (0, 1]\) we have that
Moreover, since \(|u+t+\delta v|>|t|-|u+\delta v|>|t|-(T_*+1)\) for all \(u \in [-T_*, 0]\) and \(\delta \in (0, 1]\), by (A.4) we have that for all \(t>\max \{{\hat{T}}_0+T_*+1, {\hat{T}}_1, {\hat{T}}_4\}\) and \(\delta \in (0, 1]\),
Combing (A.10) with (A.11), for all \(t \ge \max \{{\hat{T}}_0+T_*+1, {\hat{T}}_1, {\hat{T}}_4 \}\) and \(\delta \in (0, 1]\), we have
It follows that \(\frac{1}{t}\int _0^t x_{\delta }^*(\theta _r \omega )d r\rightarrow 0\) as \(t \rightarrow +\infty \), uniformly for \(\delta \in (0, 1]\).
(3) By (A.4) we investigate that
Along with \(x(\omega )=-\int _{-\infty }^0e^r\omega (r)d r\), we have
Note that \(|e^{u} \big (\omega (\delta v)-\omega (u+\delta v)+\omega (u) \big ) | \le e^{u} (M(\omega )+2+2|u|)\) for \(\delta \in (0, 1]\) and \(v \in [0,1]\), and \(\int _{-\infty }^0 e^{u} (M(\omega )+2+2|u|) du<+\infty \). Then, by using the Dominated Convergence Theorem, we get that
Recall that \(x_\delta ^*(\theta _t \omega )\) and \(x(\theta _t \omega )\) are solutions of equations (A.2) and (A.3), respectively. Then we can rewrite them as
which together with Gronwall’s inequality implies that
for all \(t \in [0, T]\) and
for all \(t \in [-T, 0]\), where T is a positive constant such that \([T_1, T_2] \subset [-T, T]\). It then follows from (A.12) and (1.7) that
This completes the proof of this proposition. \(\square \)
Rights and permissions
About this article
Cite this article
Zhao, J., Shen, J. & Lu, K. Persistence of \(C^1\) Inertial Manifolds Under Small Random Perturbations. J Dyn Diff Equat 36 (Suppl 1), 333–385 (2024). https://doi.org/10.1007/s10884-021-10103-4
Received:
Revised:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10884-021-10103-4