Modeling of Sidewall Bottom Etching Using a Neural Network

The discrepancy in the sidewall bottom etch rate with respect to the center etch rate, referred to as DSE, should be minimized to achieve a uniform surface etching. A neural network model was constructed to investigate the etching characteristics of DSE. The experimental data were collected during a via formation by using a magnetically enhanced reactive ion etch system. To systematically characterize the etch process, we employed a statistical experimental design. To evaluate the contributions from polymer deposition, chemical etching, or physical ion bombardment, we collected important radicals (such as CF or F) by using optical emission spectroscopy and the dc bias. For decreasing CF4 flow rates, the smaller DSE was attributed to enhanced polymer deposition. Both polymer deposition and ion bombardment were determined to be the main contributors to the smaller center etch rate at higher power. The model suggest that a large dc bias should be induced to achieve a smaller DSE. KCI Citation Count: 5