Nonlinear Fokker-Planck equations driven by Gaussian linear multiplicative noise (Q1669800)

From MaRDI portal
scientific article
Language Label Description Also known as
English
Nonlinear Fokker-Planck equations driven by Gaussian linear multiplicative noise
scientific article

    Statements

    Nonlinear Fokker-Planck equations driven by Gaussian linear multiplicative noise (English)
    0 references
    0 references
    0 references
    4 September 2018
    0 references
    The authors consider the following stochastic partial differential equation \[ \begin{cases} dX-\text{div}(DX)dt-\Delta\beta(X)dt=XdW &\text{ in }(0,T)\times\mathbb{R}^d,\;T>0,\\ X(0,\xi)=x(\xi),&\xi\in\mathbb{R}^d,\;1\leq d<\infty,\end{cases}\tag{1} \] where \(W\) is a Wiener process in \(H^{-1}(\mathbb{R}^d)\) over a stochastic basis \((\Omega,\mathcal{F},(\mathcal{F}_t)_{t\geq0},\mathbb{P})\) with normal filtration \((\mathcal{F}_t)_{t\geq0}\) of the form \(W=\sum_{j=1}^N \mu_j e_j \beta_j\) with an orthonormal system \(\{e_1,\dots,e_n\}\) in \(H^{-1}(\mathbb{R}^d)\) belonging to \(C^2_b(\mathbb{R}^d)\cap W^{1,2}(\mathbb{R}^d)\), \(\mu_j\in\mathbb{R}\) and \(\{\beta_j\}_{j=1}^N\) are independent \(\mathcal{F}_t\)-Brownian motions on \((\Omega,\mathcal{F},\mathbb{P})\). Moreover, it is assumed that functions \(D\) and \(\beta\) in equation (1) satisfy the following conditions: {\parindent=8mm \begin{itemize}\item[(i)] \(D\in C^1_b(\mathbb{R}^d;\mathbb{R}^d)\), \(|D|\in L^1(\mathbb{R}^d)\) and \(\text{div}D\in L^2(\mathbb{R}^d)\). \item[(ii)] \(\beta\in C(\mathbb{R})\cap C^2(\mathbb{R}\setminus\{0\})\) is monotonically nondecreasing, \(\beta(0)=0\) and there are \(m\in[0,1]\), \(a_i\in(0,\infty)\), \(i=1,2\), such that \[ \begin{aligned} &|\beta(r)|\leq a_1|r|^m,\quad\text{ for all }\, r\in\mathbb{R},\\ & |\beta''(r)r^2|+\beta'(r)|r|\leq a_2|\beta(r)|,\quad \text{ for all }\,r\in\mathbb{R}\setminus\{0\},\\ & \beta'(r)\neq0\quad \text{ and }\quad\text{sign }r\beta''(r)\leq0,\quad\text{ for all }\,r\in\mathbb{R}\setminus\{0\}. \end{aligned} \] \item[(iii)] There exists a decreasing function \(\varphi:(0,1]\to(0,\infty)\) such that \[ \beta'(\lambda r)\leq \varphi(\lambda)\beta'(r),\quad \text{ for all }\,r\in\mathbb{R}\setminus\{0\},\quad\text{ for all }\,\lambda\in(0,1]. \] \end{itemize}} By the transformation \(X(t)=e^{W(t)}y(t)\), \(t\geq0\), with \(\mu:=\sum_{j=1}^N\mu_j^2e_j^2\), equation (1) reduces to the random differential equation \[ \begin{cases} \frac{\partial y}{\partial t}-e^{-W}\text{div}(e^WDy)-e^{-W}\Delta\beta(e^Wy)+\frac12\mu y=0 &\text{ in }(0,T)\times\mathbb{R}^d,\\ y(0,\xi)=x(\xi), &\xi\in \mathbb{R}^d.\end{cases}\tag{2} \] The paper is devoted to the proof of the following results: {Theorem 2.2:} Under hypotheses (i)--(iii), for each \(x\in D_0\), equation (2) has, for each \(\omega\in\Omega\), at least one strong solution \[ y\in W^{1,2}([0,T]; H^{-1})\cap L^\infty((0,T)\times\mathbb{R}^d)\cap L^\infty([0,T];L^1),\tag{3} \] \[ y\in L^2([0,T];H^1),\tag{4} \] \[ \beta(e^Wy)\in L^2([0,T];H^1),\tag{5} \] where \[ D_0:=\left\{x\in L^1\cap L^\infty\cap H^1:\beta(x)\in H^1,\;\Delta x\in L^1,\;\Delta\beta(x)\in L^1 \right\} \] is dense in \(H^{-1}\) and in \(L^p\), \(p\in[1,\infty)\). Moreover, if \(x\geq0\) a.e. on \(\mathbb{R}^d\) then \(y\geq0\) a.e. on \((0,T)\times\mathbb{R}^d\). If \(\beta\) is locally Lipschitz on \(\mathbb{R}\) and assumptions (i)--(iii) hold, then there is a unique strong solution to (2). This solution is \(\mathcal{F}_t\)-adapted, the map \(D_0\ni x\to y(t,x)\) is Lipschitz from \(H^{-1}\) to \(C([0,T]; H^{-1})\) on balls in \(L^1\cap L^\infty\) and \(y\) extends by density to a strong solution to (2), satisfying (3) and (5), for all \(x\in L^1\cap L^\infty\). {Theorem 2.3:} If \(\beta\) is locally Lipschitz on \(\mathbb{R}\) and assumptions (i)--(iii) hold then, for every \(x\in D_0\), equation (1) has a unique strong solution \(X=e^Wy\) which satisfies \[ Xe^{-W}\in W^{1,2}([0,T]; H^{-1})\quad\mathbb{P}\text{-a.s.} \] And \(X\geq0\) a.e. on \((0,T)\times\mathbb{R}^d\times\Omega\) if \(x\geq0\) a.e. on \(\mathbb{R}^d\). Moreover, a strong solution exists for all \(x\) in \(L^1\cap L^\infty\) and the map \(x\mapsto X(t,x)\) is \(H^{-1}\)-Lipschitz from balls in \(L^1\cap L^\infty\) to \(C([0,T]; H^{-1})\). As it is mentioned by the authors, the general theory for the nonlinear accretive Cauchy problems in a Banach space is not directly applicable to equation (1). So, the first step of the proofs was to approximate \(W\) by a family of smooth random functions and so the equation (2) too by a family of nonlinear evolution equations with smooth time-dependent coefficients. Further, sharp \(H^{-1}\)-energetic and \(L^1\)-techniques were combined to obtain the existence of a strong solution to (2). The authors consider also a version of equation (1) with nonlinear drift term \(\text{div}(a(X))dt\), where \(a\) is Lipschitz and \(a(0)=0\), and obtain the corresponding analogue of Theorem 2.3.
    0 references
    nonlinear SPDEs
    0 references
    Wiener process
    0 references
    Fokker-Planck equation
    0 references
    random differential equation
    0 references
    \(m\)-accretive operator
    0 references
    0 references

    Identifiers

    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references