A systematic modeling analysis of reactive ion etching of silicon and silicon dioxide using
CF4/O2
and
CF4/H2
was performed. A parallel plate reactor was employed to obtain relationships among manipulated variables (power, pressure, etchant composition, flow rate), measured process variables ([F],
false[CF2false]
,
false[COxfalse]
, [H], dc bias), and performance variables (Si and
SiO2
etch rates, selectivity, anisotropy, uniformity). Using a fractional factorial experimental design, empirical models were fitted in order to facilitate comparison with experimental trends noted by other investigators, as well as analysis of variable interactions. This is the first paper in the literature that considers the full range of multivariable interactions, covering four manipulated variables, four process variables, and seven performance variables. In addition, this work demonstrates the importance of performing an analysis of variance in order to determine goodness of fit of the empirical model. A high value of the squared multiple correlation coefficient
R2
does not guarantee an adequate model and may lead to erroneous conclusions.