SNES solver diverged with wired SNES Function norm

General
1

Dear all,

I was solving a nonlinear variational form using SNES solver in Fenics. However the SNES solver diverged and the generated SNES Function norm is wired:

......
494 SNES Function norm 1.225656496452e-04 
495 SNES Function norm 2.796984171885e-01 
496 SNES Function norm 1.225656496172e-04 
497 SNES Function norm 2.796984171884e-01 
498 SNES Function norm 1.225656496065e-04 
499 SNES Function norm 2.796984171884e-01 
500 SNES Function norm 1.225656494447e-04 
  *** Warning: PETSc SNES solver diverged in 500 iterations with divergence reason DIVERGED_MAX_IT.

As we can see, the SNES Function norm is 1.22e-04, then 2.79e-01, and then 1.22e-04 again. Does anyone know why the SNES Function norm is repeating and how to solve this issue?
Below is the parameters of the SNES solver I used:

prm = {"nonlinear_solver": "snes",
                "snes_solver": {
                "linear_solver": "lu",       
                "absolute_tolerance": 1e-4,
                "relative_tolerance": 1e-4,
                "maximum_iterations": 500,
                "report": True,
                "error_on_nonconvergence": True}
                }

I appreciate any help.