PREDICTION MODEL OF ETCHING BIAS BASED ON ARTIFICIAL NEURAL NETWORK

Etching is an important step in the transfer of design layout to wafer results. Its quality directly impacts the performance of chips or integrated circuit. The etch aperture effect and micro-loading effect that affect the etch quality (etch bias) are considerably dependent on layout features such as pattern density, spacing between patterns, etc. Under fixed etching process parameters, a predicative model of etch bias is proposed in this paper. The experimental results show that the absolute error of the predicative value of the model can reach within +/- 2nm, and the relative error can reach less than 10%, respectively, relative to the true etch bias.