Sanghee Kwon

Optimization of optical lens-controlled scanning electron microscopic resolution using generalized regression neural network and genetic algorithm

A scanning electron microscope (SEM) is a sophisticated equipment employed for fine imaging of a variety of surfaces. In this study, prediction models of SEM were constructed by using a generalized regression neural network (GRNN) and genetic algorithm (GA). The SEM components examined include condenser lens 1 and 2 and objective lens (coarse and fine) referred to as CL1, CL2, OL-Coarse, and OL-Fine. For a systematic modeling of SEM resolution (R), a face-centered Box-Wilson experiment was conducted. Two sets of data were collected with or without the adjustment of magnification. Root-mean-squared prediction error of optimized GRNN models are GA 0.481 and 1.96x10^-^1^2 for non-adjusted and adjusted data, respectively. The optimized models demonstrated a much improved prediction over statistical regression models. The optimized models were used to optimize parameters particularly under best tuned SEM environment. For the variations in CL2 and OL-Coarse, the highest R could be achieved at all conditions except a larger CL2 either at smaller or larger OL-Coarse. For the variations in CL1 and CL2, the highest R was obtained at all conditions but larger CL2 and smaller CL1.

Use of a neural network to characterize the charge density of PECVD-silicon nitride films

Silicon nitride (SiN) films were deposited using a plasma-enhanced chemical vapor deposition system (PECVD). The charge density of the SiN films was modeled using a generalized regression neural network (GRNN). The PECVD process was characterized by means of a face-centered Box Wilson experiment. The prediction performance of the GRNN model was optimized using a genetic algorithm (GA). The GA-GRNN model significantly improved the GRNN prediction performance by more than 55%. The optimized GA-GRNN model was used to investigate the effects of various parameters on the charge density. A higher charge density was obtained at higher temperatures (i.e. a lower H concentration). Increasing the pressure increased the charge density at all temperatures levels with a much stronger impact at a lower H concentration. The effects of the SiH4 and N2 (or NH3) flow rates on the charge density were similar in that a higher charge density was achieved at a lower Si−N ratio (N-rich films). A considerable increase in the charge density with a radio frequency power at a lower NH3 flow rate was attributed to the generation of more Si−H than N−H bonds.

Wavelet-coupled backpropagation neural network as a chamber leak detector of plasma processing equipment

Research highlights? We developed a real-time leak detection method by combining neural network and wavelets. ? The wavelets were used as a preprocessor of optical emission spectroscopy (OES) data. ? Auto-correlated time series neural network model was developed by training it with wavelet-filtered OES leak-associated data. ? Wavelet-based neural network models demonstrated high sensitivity to leak occurrence, enabling them to be utilized for detecting chamber leaks in real time. In order to improve equipment throughput and device yield, chamber leaks needs to be strictly monitored. A new technique for leak detection is presented and this was accomplished by combining backpropagation neural network, discrete wavelet transformation (DWT), and continuous transformation (CWT). Different types of BPNN models were constructed with raw, DWT, and CWT data and these are referred to as raw, DWT, and CWT models, respectively. Constructed models were validated with a total of 47 data sets for normal and leaky chamber conditions. The experimental data were in-situ collected by using an optical emission spectroscopy. Both raw and DWT models could detect all abnormal data sets. Worst detection by CWT model was noted. Wider detection margin provided by DWT model was attributed to enhanced sensitivity of model to leaky condition. A modified cumulative control chart was applied to the statistical mean of raw OES spectra as well as to DWT and CWT data. The statistical mean-based CUSUM control chart was unable to detect chamber leaks. In contrast, chamber leaks could be identified by all model-based CUSUM control charts. Of the proposed models, DWT model is identified to be the most appropriate to chamber leak detection.

Neural network characterization of plasma-induced charging damage on thick oxide-based metal-oxide-semiconductor device

Charging damage can critically degrade oxide reliability. Antenna-structured metal-oxide-semiconductor field-effect transistors were fabricated to examine the effect of process parameters on charging damage. Charging damage to threshold voltage (Vth) was investigated experimentally as well as by constructing a neural network model. For a systematic modeling, charging damage process was characterized by means of a face-centered Box–Wilson experiment. The prediction performance of neural network model was optimized by applying genetic algorithm. A radio frequency source power was identified as the most influential factor. This could be more ascertained by the insignificant impact of bias power or gas ratio. Using the model, implications of plasma nonuniformity and polymer deposition were examined under various plasma conditions.

Statistical Characterization of Process-Induced Plasma Damage

During plasma processes, charging damage produces various defects in silicon oxide, thereby deteriorating device performance. Optimizing process-induced charging damage requires a computer model, as well as a quantitative analysis of process parameter effects. In this study, plasma charge damage on threshold voltage of metal-semiconductor field-effect transistors is statistically investigated. This includes the analysis of main and interaction effects of process parameters, as well as the construction of response surface models. Charging damage is characterized by means of a statistical experiment. Four types of statistical regression models are constructed. A model with the largest R-Square (R2) fit of 90.6 is chosen for the response surface analysis. Analysis of the main effects revealed that radio frequency power and gas ratio are the most significant and least significant factors, respectively. Among various interaction terms, only the interaction between radio frequency power and bias is found to be influential. Meanwhile, several conflicting effects are noted as the bias power or gas ratio are varied. An optimized regression model is used to understand parameter effects on plasma charging damage.