Convergence of the spectral method for stochastic Ginzburg-Landau equation driven by space-time white noise (Q2583444)

Consider a stochastic Ginzburg-Landau equation on \([0,L]\) driven by a space-time white noise \[ u_t=\Delta u +\mu u- \lambda u^3 + \dot w, \quad u(0)= u_0\in C_0[0,L], \] with boundary condition \(u(t,0)=u(t,L)=0\), \( t\geq 0\), where \(\mu,\lambda>0\). \(w(t,x)=\sum_kw_k(t)e_k(x)\), where \(w_k\) are independent standard Brownian motions and \(e_k\), \(k=1,2,\dots\), is the orthonormal basis in \({\mathbb L}^2 _0[0,L]\) of eigenvectors of \(\Delta\); let \(-\lambda_k\) be the corresponding eigenvalues. Let \(e^{\Delta t}\) be the semigroup generated by \(\Delta\). Denote \[ W(t,x)=\int_0^te^{\Delta(t-s)}dw(s)=\sum_ke_k(x)\int_0^te^{-\lambda_k(t-s)}dw_k(s). \] Then the solution of the equation is defined in the following mild form: \[ u(t)=e^{\Delta t }u_0+\int_0^te^{\Delta (t-s)}F(u(s))ds+W(t), \] where \(F(x)=\mu x-\lambda x^3\). Let \(P_N\) be the projection of \({\mathbb L}_0^2\) onto the span of \(e_1,\dots,e_N\) and \(W_N=P_NW\). Let \(u_N\) be the solution of \[ u_N=e^{\Delta t}P_Nu_0+\int_0^te^{\Delta(t-s)}P_NF(u_N(s))ds+W_N(t). \] Denote by \(| \cdot| _{{\mathbb H}_0^r}\) the usual Sobolev norm, i.e., \(| \sum_ka_ke_k| ^2_{{\mathbb H}_0^r}=\sum_k\lambda_k^ra_k^2\). The main result of the paper states that for any \(r\in [0,1/2)\) and \(\alpha\in (0,1-2r)\) there are positive constants \(\beta, C, c\) and a random variable \(\xi\) such that for \(N\) large \[ {\mathbb P} \Bigl\{\sup_{t\in[0,T]}| u(t)-u_M(t)| _{{\mathbb H}_0^r}\leq\xi(| u_0-P_Mu_0| _{{\mathbb H}_0^r} +1/M^{\alpha/2}),\;M\geq N\Bigr\}\geq 1-Ce^{-cN^{2\beta}}. \] Moreover, also the \({\mathbb L}^p(\Omega)\) norms of \(\sup_{t\in[0,T]}| u(t)-u_N(t)| _{{\mathbb H}_0^r}\) are estimated by \(| u_0-P_Nu_0| _{{\mathbb H}_0^r}+1/N^{\alpha/2}\) up to constants (\(p\geq 1\)). Finally, by the Sobolev embedding theorem, in these inequalities the \(| \cdot| _{{\mathbb H}_0^r}\) norms can be replaced by the appropriate \({\mathbb L}^q[0,T]\) norms.