Necessary and sufficient condition for comparison theorem of 1-dimensional stochastic differential equations (Q2490066): Difference between revisions

The authors consider two one-dimensional stochastic differential equations of the form \[ X^i_s = x_i+\int_t^s b_i(r,X_r^i)dr + \int_t^s \sigma_i(r,X_r^i)dW_r +\int_t^s\int_Z \gamma_i(r,X^i_{r-},z) \tilde{N}(dz,dr),\quad i=1,2, \] where \(W\) is a Brownian motion and \(\tilde{N}(dz,dr)\) the Lévy compensator for a Poisson process \(N\). The main result provides necessary and sufficient conditions such that \(x_2 \geq x_1\) implies \(X_s^2\geq X_s^1\) for all \(s\in[t,T]\) almost surely for all \(t\leq T\). For the proof they recast the above two equations into one two-dimensional system with solution \(X^{t,x}\in R^2\) and impose the constraint that this solution stays in the closed subspace \(K=\{(y_1,y_2):y_1\leq y_2\}\) of \(R^2\) and rephrase the comparison problem in terms of viscosity solutions of linear parabolic partial differential equations.