The users of cognitive radio network are working in a highly time-varying environments. The classic ergodic capacity definition fails to capture the dynamic property of wireless channels. In the paper, we set up a new model to characterize the time-varying property of wireless channels. Firstly, we established a dynamic optimal rate allocation model under time-varying channel capacity. Due to the distributed property of cognitive radio terminals, we extend the model to differential game approach and turn it into a Boundary Value Problem(BVP). We give an analytic open-loop Nash equilibria solution of the differential games model.