Degenerate stochastic differential equations for catalytic branching networks (Q985339)

Let \((V,\mathcal E)\) be a directed graph with vertices \(1, \ldots, d\) and corresponding directed edges \(e_1,\ldots, e_k\). The vertices stand for different types whose growth is under investigation; \((i,j)\in\mathcal E\) means that type \(i\) `catalyzes' the branching of type \(j\). Throughout it is assumed that \((i,i)\not\in\mathcal E\). Let \(C\) be the catalysts, \(R\) the set of reactants (i.e.\ vertices appearing as 2nd element of an edge), \(C_j = \{i : (i,j)\in\mathcal E\}\) and \(R_i = \{j:(i,j)\in \mathcal E\}\). The author studies the weak uniqueness of solutions of the following system of stochastic differential equations driven by a Brownian motion: \[ \begin{aligned} dX_t^{(j)} &= b_j(X_t)\,dt + \sqrt{2\gamma_j(X_t) \sum_{i\in C_j} X_t^{(i)}X_t^{(j)}}\;dB_t^j, \qquad j\in R;\\ dX_t^{(j)} &= b_j(X_t)\,dt + \sqrt{2\gamma_j(X_t) X_t^{(j)}}\;dB_t^j, \qquad j\not\in R. \end{aligned} \] It is assumed that the coefficients \(\gamma_i, b_i : \mathbb R_+^d \to [0,\infty)\) are locally Hölder continuous and that the \(b_i\) satisfy a linear growth condition. Basically this is the setup of the paper [Ill. J. Math. 50, No. 1-4, 323--383 (2006; Zbl 1107.60045] by \textit{D. A. Dawson} and \textit{E. A. Perkins}; note, however, that now the catalysts may depend on each other and that the covariance coefficients may be degenerate. The proof relies on the fact that the system of SDEs has a unique weak solution if, and only if, the associated Martingale problem for the generator \[ Lf(x) = \sum_{j\in R} \gamma_j(x) \sum_{i\in C_j} x_ix_j \partial_j^2 f(x) + \sum_{j\not\in R} \gamma_j(x)x_j\partial_j^2f(x) + \sum_{j\in V} b_j(x)\partial_jf(x), \,\, f\in C_b^2(\mathbb R_+^d) \] is well-posed. To show well-posedness, the author uses an perturbation argument where \(L\) is understood as a `small' perturbation of a certain constant-coefficient generator \(L^0\).