Specifies the number of extra nonlinear solves and updates to apply after convergence is detected. The default is 0.
Newton algorithms have a second-order convergence rate when sufficiently close to the solution. Taking a few extra nonlinear steps can often produce a solution to the discretized nonlinear system that is accurate to machine precision. This is not to say that there would be no discretization error. Discretization error is related to the timestep, element size, spatial interpolation, and the operators in the equations.