Fourier analysis applied to SPDEs (Q678372): Difference between revisions
From MaRDI portal
Removed claim: author (P16): Item:Q188378 |
Changed an Item |
||
Property / author | |||
Property / author: Douglas Blount / rank | |||
Normal rank |
Revision as of 16:53, 10 February 2024
scientific article
Language | Label | Description | Also known as |
---|---|---|---|
English | Fourier analysis applied to SPDEs |
scientific article |
Statements
Fourier analysis applied to SPDEs (English)
0 references
17 April 1997
0 references
Two particular stochastic parabolic equations are identified as weak limits of \(N\)-dimensional systems of stochastic differential equations (that, in turn, arise as diffusion limits of stochastic population models) as \(N\) tends to infinity. To be more precise, let us describe one of the models: Denote \(N^{-1}\mathbf Z\cap [0,1[\) by \(\mathbf S_{N}\) and the quotient space \(\mathbf R/\mathbf Z\) by \(\mathbf S\). Let \(R\) be a polynomial with a negative leading coefficient, \(R(x) = \sum ^{p}_{i=0} c_{i}x^{i}\), \(x\in \mathbf R\), \(c_{0} \geq 0\), \(c_{p}<0\). Let \(X_{N}\) be the solution of the stochastic differential equation \[ \begin{aligned} X_{N}(t,r) = X_{N}(0,r) &+ \int ^{t}_{0}\left \{ \Delta _{N} X_{N}(s,r) + R(X_{N}(s,r))\right \} ds \\ &+ \int ^{t}_{0} \sqrt {\gamma _{N}(r)X_{N}(s,r)} dW_{N}(s,r), \quad r\in \mathbf S_{N}, \end{aligned} \] where \(\gamma _{N}>0\), \(\{N^{-d/2}W_{N} (t,r), \;r\in \mathbf S_{N}\}\) are independent standard Wiener processes and \(\Delta _{N}\) stands for the discrete Laplacian, \(\Delta _{N}f(r) = N^{2}[f(r+N^{-1})+f(r-N^{-1})-2f(r)]\). The weak existence and uniqueness of nonnegative solutions to this system is established by means of the theory of measure-valued processes. Let us consider \(\gamma _{N}\), \(X_{N}(t,\cdot )\) as piecewise constant functions on \([0,1[\); suppose that \(\gamma _{N} \to \gamma >0\) on \([0,1[\) and that \(X_{N}(0)\) converges weakly to an \(L^{2}(\mathbf S)\)-valued random variable \(\xi \). Under some integrability assumptions on \(X_{N}(0)\), the processes \(X_{N}\) are shown to converge weakly in \(C([0,\infty [; L^{2}(\mathbf S))\) to the mild solution \(X\) of the equation \[ dX(t) = \left \{\frac {\partial ^{2}X(t)}{\partial x^{2}} + R(X(t))\right \}dt + dZ(t), \quad X(0)=\xi , \] where \(Z\) is an \(W^{-\alpha ,2}\)-valued martingale with continuous paths such that the quadratic variation of \(\langle Z(t), f\rangle \) is \(\int ^{t}_{0} \langle \gamma X(s),f^{2}\rangle ds\) for any \(f\in W^{\alpha ,2}\), \(\alpha >\frac 12\) (\(W^{s,2}\) denotes the usual Sobolev-Slobodetskij space on \(\mathbf S\)).
0 references
stochastic partial differential equations
0 references
diffusion limits
0 references