Increasing number of snes iterations, I get convergence. So it is a problem of initial guess being too far from the solution of the nonlinear equation.
Solution can be seen here https://github.com/cpraveen/fenics/blob/master/1d/cosmic_ray/cosmic_ray.ipynb Green curve is solution after two time steps. It took about 100 snes iterations in first time step and about 50 in second time step. I use exact Jacobian and direct LU solve. Thanks praveen
