On free stochastic differential equations (Q639339)

For the equation \(dX_{t}=a(X_{t})dt+b(X_{t})(dW_{t})c(X_{t})\), \(X_{0}=\overline{X}\) (and also for the one with \(\sum_{i=1}^{n}b_{i}(X_{t})(dW_{t})c_{i}(X_{t})\) instead of the last term) where \(W_{t}\) is a free Brownian motion and \(a\),\(b\),\(c\) are functions \(f\) satisfying, for every \(A>0\), \(\| f(X)-f(Y)\| \leq K_{A}\| X-Y\|\) for some \(K_{A}\) and all \(\| X\|\), \(\| Y\| \leq A\), the author shows the existence of a \(t_{0}>0\) and of a unique solution in \(t\in [0,t_{0})\) with bounded \(\| X_{t} \|\), \(t \in [ 0,t_{0})\). If \(G(t,z)=(X_{t}-z\cdot 1)^{-1}\), \(g(t,z)=EG(t,z)\), then \[ dg_{t}/dt=-E(a(X_{t})G_{t}^{2})+E(b(X_{t})c(X_{t})G_{t})E(b(X_{t}) c(X_{t})G_{t}^{2}). \] Particular cases of the formula: for \(a\) a polynomial, \(bc\) also, then \(bc=1\). If \(a\) has real coefficients and \(b=c=1\), then, if \(X_{0}^{\ast }=X_{0}\), the spectral distribution of \(X_{t}\) has a density \(p\) satisfying \(dp/dt=-\partial (ap+p\cdot Hp)/\partial x\), \(Hp\) being the Hilbert transform. The following particular cases are solved. \(dX_{t}=\theta X_{t}dt+\sigma dW_{t}\), \(X_{0}=0\), in which \(X_{t}\) has a semicircle spectral distribution on an explicitly expressed interval. \(dX_{t}=\theta X_{t}dt+X_{t}^{1/2}(dW_{t})X_{t}^{1/2}\), \(X_{0}=1\), the other equation is \(z+g^{-1}=e^{(\theta -1- g)t}\) and the density of the spectral distribution of \(X_{t}\) is supported on an explicitly expressed interval. \(dX_{t}=\theta X_{t}dt+X_{t}dW_{t}+(dW_{t})X_{t}\), \(X_{0}=1\), determining \(E(X_{t}^{2})\). \(dX_{t}=kX_{t}(dW_{t})X_{t}\), \(X_{0}=a\cdot 1\), \(t_{0}=(ak)^{-2}\tau\), \(\tau \in (0,1)\) (explosion); explicitly expressed density of the spectral distribution of \(a^{-1}X_{(ak)^2\tau }\), \(\tau \in (0,1)\). The paper contains preliminary information making it easier to read and a 7 page introduction with comments, history of the problem, graphics etc.