Existence of the solutions of backward-forward SDE's with continuous monotone coefficients (Q2497820): Difference between revisions

The authors consider backward-forward stochastic differential equations (BFSDEs) for scalar \(X_t\) and \(Y_t\) of the form \[ X_t = x + \int_0^t b(s,X_s,Y_s)ds + \int_0^t \sigma(s,X_s)dB_s,\quad Y_t = \Gamma + \int_t^T f(s,X_s,Y_s,Z_s)ds - \int_t^T Z_s dB_s. \] Here, \(B\) is a Brownian motion on a suitable probability space and the coefficient functions satisfy certain measurability and continuity assumptions. Further, the following set of conditions is assumed: \(b\) is increasing in \(y\), \(f\) increasing in \(x\) and for some constant \(K>0\) \[ | b(s,x,y)| \leq K(1+| x| +| y| ),\quad | f(s,x,y,z)| \leq K(1+| y| +| z| ), \] \[ | \sigma(s,x)| \leq K(1+| x| ),\quad | \sigma(s,x) -\sigma(s,x')| \leq K| x-x'| \] for all \(s\in [0,T]\), \(\omega \in \Omega\) and \(x,x',y,z \in R\). Under these conditions, which are weaker than the conditions assumed previously in the literature, the authors can show the existence of a solution \((X,Y,Z)\) of the above BFSDE. This solution is a minimal solution, that is, if \((X',Y',Z')\) is another solution, then \(P\)-a.s. for any \(t\leq T\) it holds that \(X_t\leq X'_t\) and \(Y_t\leq Y'_t\). The starting point of the proof is a standard one-dimensional backward equation; further techniques in the proof include iteration procedures and comparison results. Under the above assumption uniqueness of the solution can not be established; an example is given that illustrates this fact.