In the semiconductor manufacturing, which consists of significantly precise and diverse unit processes, minute defects can cause significantly large risk, which is directly related to the yield. Through fault detection and classification (FDC), the equipment status is monitored, and the potential causes of faults can be investigated. In the mass production process, unbalanced data problems are also important, including preprocessing methods for data analysis in real time. This study proposes a stepwise FDC method with a process fault detection (FD) and faulty equipment part classification. Fault detection (FD) is proposed using a one-class support vector machine (OC-SVM) to determine anomalies that occur during a process, and fault classification (FC) is followed by the importance between variables that determine whether a fault exists is extracted using extreme gradient boosting (XGBoost). Variables whose importance has been confirmed, are reclassified to a part-level based on the variable name, and defects are notified to the part-level level. An empirical study to validate the proposed data-based framework for fault detection and diagnosis was performed under the scenario of unexpected failure of two ${\mathbf {\mathrm {SF}}}_{\mathbf {6}}/{\mathbf {\mathrm {O}}}_{\mathbf {2}}$ mass flow controllers (MFCs). The experimental results confirmed that the application-oriented proposed framework performed well in FDC operations and showed that it can provide part-level notification to engineers.

Coating the inner surfaces of high-powered plasma processing equipment has become crucial for reducing maintenance costs, process drift, and contaminants. The conventionally preferred alumina (Al₂O₃) coating has been replaced with yttria (Y₂O₃) due to the long-standing endurance achieved by fluorine-based etching; however, the continuous increase in radio frequency (RF) power necessitates the use of alternative coating materials to reduce process shift in a series of high-powered semiconductor manufacturing environments. In this study, we investigated the fluorine-based etching resistance of atmospheric pressure-sprayed alumina, yttria, yttrium aluminum garnet (YAG), and yttrium oxyfluoride (YOF). The prepared ceramic-coated samples were directly exposed to silicon oxide etching, and the surfaces of the plasma-exposed samples were characterized by scanning electron microscopy, energy-dispersive X-ray spectroscopy, and X-ray photoelectron spectroscopy. We found that an ideal coating material must demonstrate high plasma-induced structure distortion by the fluorine atom from the radical. For endurance to fluorine-based plasma exposure, the bonding structure with fluoride was shown to be more effective than oxide-based ceramics. Thus, fluoride-based ceramic materials can be promising candidates for chamber coating materials.

For advanced equipment control, two schemata of real-time fault detection were performed using machine learning algorithms in silicon etching in SF ₆ /O ₂ /Ar plasma. Fault detection and classification is investigated with the plasma state information with optical emission spectroscopy (OES) data to find the root cause of the anomaly in the process parameters. Fault detection and control is also demonstrated to predict the shift of the process parameter along the amount of process gas flow rate injected into the chamber, considering a fault. Especially, plasma information (PI), such as electron temperature and electron density, was derived from OES data into equation-based corona model. These were utilized to evaluate which process parameter is the most significantly affecting on the performance of the established model through Shapley value in fault detection and control. By the combination of isolation forest algorithm for finding the plasma abnormalities in real time and Adaboost algorithm for classifying root causes of faults, the suggested algorithm could accurately detect the root cause. DeepSHAP algorithm helped not only the prediction of gas flow rate, but PI was identified as critical parameter, interpreting the model through Shapley value. We propose a new multi-function integrated algorithm by the ensemble algorithms.

Semiconductor manufacturing comprises hundreds of consecutive unit processes. A single misprocess could jeopardize the whole manufacturing process. In current manufacturing environments, data monitoring of equipment condition, wafer metrology, and inspection, etc., are used to probe any anomaly during the manufacturing process that could affect the final chip performance and quality. The purpose of investigation is fault detection and classification (FDC). Various methods, such as statistical or data mining methods with machine learning algorithms, have been employed for FDC. In this paper, we propose an artificial immune system (AIS), which is a biologically inspired computing algorithm, for FDC regarding semiconductor equipment. Process shifts caused by parts and modules aging over time are main processes of failure cause. We employ state variable identification (SVID) data, which contain current equipment operating condition, and optical emission spectroscopy (OES) data, which represent plasma process information obtained from faulty process scenario with intentional modification of the gas flow rate in a semiconductor fabrication process. We achieved a modeling prediction accuracy of modeling of 94.69% with selected SVID and OES and an accuracy of 93.68% with OES data alone. To conclude, the possibility of using an AIS in the field of semiconductor process decision making is proposed.

Silicon nitride (Si₃N₄) etching using CF₄/O₂ mixed with N₂ has become very popular in 3D NAND flash structures. However, studies on Si₃N₄ dry etching based on optical emission spectroscopy (OES) are lacking; in particular, no study has reported the use of OES for analyzing N₂-mixed CF₄/O₂ plasma. Thus, this study demonstrates an OES-based approach for analyzing a mixed-gas plasma for etching Si₃N₄ thin films. The state of each single gas plasma of CF₄, O₂, and N₂ as well as that of mixed plasmas of heterogeneous gases CF₄/O₂, CF₄/N₂, and O₂/N₂ was investigated to analyze the mixed-gas plasma. Furthermore, the amount of N₂ in the CF₄/O₂ plasma varied from 0 to 8 sccm. The relationship between the OES analysis results and the Si₃N₄ etch rate was subsequently established using Si₃N₄ film etching, and the explanation was verified through a chemical reaction modeling and surface reaction. Therefore, our study confirmed the alteration in chemical species and quantity that occurred when N₂ was added to CF₄/O₂ plasma and the effect of the alteration on Si₃N₄ etch.