Elliptic self similar stochastic processes (Q4495243)

A generalized stochastic process \(X=\{X(x),x\in R^d\}\) is said to be elliptic if there exist a stochastic measure \(M\) on \(R^d\) and a pseudo-differential elliptic operator \(L\) such that \(LX=M\) a.s. For \(\alpha\in (1,2]\), \(H \in(0,1)\) and a real-valued measurable function \(S(\xi)\) on the unit sphere \(\Sigma_{d-1}\), satisfying \(s(\xi)= s(-\xi)\), \(0<c\leq s(\xi)\leq c^{-1}\), \(\xi\in \Sigma_{d-1}\), an operator \(L^{(H)}_{\alpha,s}\) is defined as the pseudo-differential elliptic operator with the symbol \(|\xi|^{d/ \alpha+H} s(\xi/ |\xi|)\). Assumptions on \(M\) are found under which the equations \(L^{(H)}_{\alpha^*,s}X=M\), \(X(0)=0\), and \(L^{(H)}_{\alpha,s} Z=\widehat M\), \(Z(0) =0\), are uniquely solvable in the suitable spaces of generalized stochastic processes, where \(\alpha^{-1}+ (\alpha^*)^{-1} =1\) and \(\widehat M\) denotes the Fourier transform of \(M\). It is investigated the inverse scaling invariance property with index \(H\) for \((X,Z)\) if for each \(\lambda>0\) \[ \text{Law}\bigl( \lambda^{-H}X(\lambda \cdot),\lambda^H Z(\lambda^{-1}\cdot) \bigr)=\text{Law} \bigl(X(\cdot), Z(\cdot)\bigr). \] Wavelet decomposition and existence of continuous modification for \(X\) are also considered. For the entire collection see [Zbl 0936.00006].