Probabilistic approximation and inviscid limits for one-dimensional fractional conservation laws (Q2496942)

The authors consider the following one-dimensional conservation law with fractional viscosity \[ \partial_tu(t,x) = \nu D^\alpha u(t,x) - \partial_x A(u(t,x)),\tag{1} \] \(u(0,x)=u_0(x)\), where \((t,x) \in \mathbb R_+\times \mathbb R\), \(\nu > 0\) and \(A: \mathbb R\rightarrow \mathbb R\) is a \(C^1\)-function. The operator \(D^\alpha\), \(\alpha \in (1,2)\), denotes the symmetric fractional derivative (fractional Laplacian) of order \(\alpha\), which is given either via the Fourier transform \(\mathcal F\) as \(D^\alpha g(x) ={\mathcal F}^{-1}(| \xi| ^\alpha {\mathcal F}(g)(\xi))(x)\) or, equivalently, by its singular integral representation \(D^\alpha g(x) = \int_{\mathbb R}(g(x+y)-g(x)- 1_{\{| y| \leq 1\}}g'(x)y ) | y| ^{\alpha-1}\,dy\) for a suitable positive constant \(K\). The first aim of the authors is to give a probabilistic interpretation to (1). In order to obtain this they consider the gradient \(v(t,x)=\partial_x u(t,x)\) which satisfies the evolution equation \[ \partial_t v=\nu D^\alpha v- \partial_x (A'(H*v)v),\tag{2} \] \(v(0,.)=m\), where \(H(y)\) denotes the unit step function \(1_{\{y\geq 1\}}\), and \(m\) is a non-zero, bounded signed measure on \(\mathbb R\) with its total mass \(\| m\| =1\). For this equation they define a nonlinear martingale problem and provide conditions under which it admits a unique solution. Via this solution of (2) they then obtain a unique bounded weak solution of (1). Then the authors develop a Monte-Carlo method for simulating the fractional conservation law. In particular, they construct a sequence of interacting particle systems of size \(n\) whose weighted cumulative empirical distribution functions converge to the unique bounded weak solution of (1) as \(n \rightarrow \infty\). Further, the vanishing viscosity limit is considered. Thus a sequence \(\nu_n\) of positive numbers is introduced which replace the constant \(\nu\) in front of the fractional viscosity term in the interacting particle systems. It is shown that the empirical cumulative distribution functions of these systems converge to the unique entropy solution of the inviscid (\(\nu = 0\)) conservation law (1).