The magnetization ripple: a nonlocal stochastic PDE perspective (Q2274102): Difference between revisions

This paper deals with the nonlocal elliptic equation \( \mathcal{L}u+P(u\partial_2Ru)+ (1/2) \partial_2 Ru^2 - (1/2)P(u\partial_1Ru^2)=\sigma P\xi\) in \(\mathbb{R}^2\), with \(u\) representing the transversal magnetization component and \(\xi\) a quenched noise. Here, \(\mathcal{L}:=-\partial_1^2-|\partial_1|^{-1}\partial_2^2\), \(R:=|\partial_1|^{-1}\partial_1\), \(P\) is the \(L^2\)-orthogonal projection onto the set of functions of vanishing average in \(x_1\) (extended to periodic distributions), and \(\sigma\) is a constant such that \(0 \leq\sigma\leq\sigma_0\), for some random positive constant \(\sigma_0\). The derivation of the model is discussed in terms of the connection between ferromagnetism and singular stochastic PDE. The authors establish the existence of a unique smooth and periodic solution \(u^\ell:=w^\ell+\sigma v_\ell\), of vanishing average in \(x_1\), to the approximate problem, which in the right hand side of the elliptic equation under study the noise \(\xi\) is replaced by its convolution \(\xi_\ell= \ell^{-5/2} \phi(x_1/\ell, x_2/\ell^{3/2})*\xi \) with some symmetric Schwartz function \(\phi\) having vanishing average in \(x_1\), and \(v_\ell\) is the mollification of the solution \(v\) to the linear equation \(\mathcal{L}v=P\xi\). For the other hand, it is established the existence of \(u:=w+\sigma v\) solving \( \mathcal{L}w+P(\sigma^2F +\sigma w\partial_2Rv +u\partial_2Rw)+ (1/2) \partial_2 Ru^2 - (1/2)P(u\partial_1Ru^2)=0\) (in a distributional sense), with \(F=\lim_{\ell\downarrow 0} F^\ell\), \(F^\ell:=P(v_\ell \partial_2Rv_\ell)\), and it verifies \(\lim_{\ell\downarrow 0} |w^\ell-w|_{C^{5/4-\epsilon}}=0\), for \(\epsilon \in ]0,1/4[\). The proofs of existence of \(w^\ell\) and \(w\) rely on the Banach's contraction mapping theorem. The passage to the limit as \(\ell\downarrow 0\) follows by estimating the quadratic terms \(w^\ell\partial_2Rv_\ell\) and \(u^\ell \partial_2Rw^\ell\) and the cubic term \(u^\ell\partial_1R(u^\ell)^2\) as the regular product of a function of \(C^\alpha\) and a distribution of \(C^\beta\) with \(-3/2<\beta<0<\alpha<3/2\) being such that \(\beta\not=-1,-1/2\) and \(\alpha+\beta>0\), and that the one-dimensional Hilbert transform \(R\) maps the anisotropic Hölder space \(C^\alpha\) into a slightly larger Hölder space \(C^{\alpha-\epsilon}\) on the two-dimensional torus \([0,1[^2\). The estimates of the stochastic terms, namely \(P\xi_\ell\) and the off-line term \(F^\ell\), are derived from the stochastic (second moment) bounds of the Fourier coefficients and its logarithmic derivative \(\partial/\partial(\ln\ell)= \ell\partial/\partial\ell\) in the convolution length \(\ell\).