Expert Systems with Applications: An International Journal
Expert Systems with Applications: An International Journal
| Hi-index | 2.88 |
To maintain device yield, plasma status should be stringently monitored. In this study, a new monitoring method for plasma diagnosis is presented. The method was based on a wavelet detection of surface roughness measured by atomic force microscopy. The wavelet enabled us to analyze spatial variations of surface roughness in vertical, lateral, and diagonal components. The sensitivity of spatial variation was examined in terms of a ratio pattern for the variations in process parameters. The ratio patterns were defined by the numerical values characterizing wavelet-decomposed images. The method was applied to the etching of a silicon oxynitride film in a C"2F"6 inductively coupled plasma. The process parameters involved include radiofrequency source power, bias power, pressure, and C"2F"6 flow rate. Applicability of this method was evaluated without or with respect to a fixed reference condition. In the case of a non-reference condition, a particular ratio for the diagonal component demonstrated the highest sensitivity to the variations in all parameters, but C"2F"6 flow rate. It is also noticeable that this ratio showed a strong correlation with actual surface roughness measurements. Meanwhile, in the case of one fixed reference, all ratio sensitivities were much higher than those for actual measurements for each component. A particular ratio for the lateral component yielded the highest sensitivity compared to other components. The proven high sensitivity indicates that the presented method can be effectively used for monitoring plasma conditions.