A Study on the Effect of Ion Implantation Process Variations on the Small-Signal MESFET Model Parameters

With increasing complexity of manufacturing technology and semiconductor devices it becomes extremely important to make a connection between process parameters and characteristics of integrated components. Over the past 10 years, a lot of research has been done on a novel approach to the semiconductor manufacturing process control called virtual metrology. Virtual metrology allows to estimate the output process parameters, which ought to be under control, using the input process variables, such as sensor data, collected from the manufacturing equipment without using expensive metrology tools. In this research we utilized the basic concepts of virtual metrology to define interrelations between the input process variables and the model parameters of the semiconductor device. Particularly, we developed several predictive models, trained by the most common machine learning techniques, which are capable to predict values of small-signal MESFET model capacitances using the parameters of implant doping for MESFET channel and contact region formation.