Modeling of electron temperature and DC bias voltage in an inductively-coupled Cl2/Ar plasma using neural network

We proposed a novel neural network (NN) model for the inductively-coupled Cl2/Ar plasma (ICP). Plasma experiments were performed in a planar inductively coupled plasma reactor. The plasma parameters and composition were determined by a combination of plasma diagnostics carried out with a double Langmuir probe (Plasmart DLP2000). The input parameters of the plasma model were gas mixing ratio, source power, and bias power, which were varied within the ranges of 0–100% Cl2/Ar, 500–800 W, and 50–300 W, respectively. The voltage–current curves were treated to obtain the electron temperature (Te), DC bias voltage, and ion saturation current (Ji) by using the software supplied by the equipment manufacturer. For the outputs and the input parameters, we developed a back propagation neural network (BPNN) model and extracted its features. Then, we designed a pre-processor using the features to optimize the BPNN model. The experimental results showed that the proposed method was very effective for the optimization of the BPNN model with respect to model precision and learning speed.