Pontryagin's maximum principle is used for computing the optimum control function u(t) for a given plant and a given performance criterion. If u(t) is bounded, the control is of the bang-bang type in many cases. If u(t) is expressed as the function of the state variables, that means, u(t) = sgn f(xi), the equation f(xi) = 0 determines the switching surface in the state space. In general these surfaces are not given by simple analytic functions, in particular not if the transfer function of the plant contains complex poles. If the desired final state is given by error and error derivatives being zero, this surface goes through the origin of the phase space.