Dynamic simulation of process control of the reactive sputter process and experimental results

The control of reactive sputter processes has been dynamically simulated by integrating the Larsson differential equations. This was done by employing a fast Runge-Kutta step control algorithm, allowing us to simulate sputtering with more than 20-fold real time speed on a pentium 166 Mhz. A simple proportional integral differential (PID) algorithm was implemented to simulate (i) the partial pressure control via reactive gas flow at a fixed current and (ii) the partial pressure control via current at a fixed reactive gas flow. The control cycle time was varied with respect to real life process control. These simulations show that arbitrary setpoints on the stationary s curve resulting from the steady state Larsson equations can be stabilized. However, the cycle time of the PID controller has to be small enough, e.g., less than 600 ms, for a reliable control. The setpoints in the transition mode are highly unstable, so that the process drifts immediately into one of the two corresponding stable steady states (typically within about 3-15 s) after freezing the control. In addition these computations were compared with experimental control results of reactively sputtered TiO2 and Nb2O5 films deposited by the midfrequency technique. In both cases the total s curve was stabilized at a constant oxygen flow. The process stabilization was performed at power densities of up to 5 W/cm(2), limited by the generator output. For the oxygen partial pressure measurements a lambda -probe with optimized speed was used. (C) 2001 American Institute of Physics.