Plasma processes have been used in semiconductor processes for etching and chemical vapor deposition processes. In a plasma reactor, the generated reactive ions and radicals react on a wafer surface to remove target materials or form a thin film. To improve the results of these reaction processes, it is necessary to control the key process parameters such as gas flow rate, pressure, and applied electrical source/bias power that are related to reaction rate. However, as the size of semiconductor devices has been reduced, it is becoming difficult to understand the plasma processes. If the process parameters are not maintained as intended and are off the process window, the unexpected process failure will be found. To solve these problems, the sensitivity-enhanced real-time plasma monitoring techniques are needed. There are various non-invasive plasma diagnostic tools for responding to the change of plasma condition immediately. Commonly, optical emission spectroscopy (OES) is widely applied to monitor the changes of chemical elements in the plasma phase, and commercially used for etching endpoint detection because OES can detect the presence of active species without any perturbation. Optical emission spectra are measured through a viewport, and a few sensitive single wavelength signals, which are related to the reaction reactants and/or products, are chosen for real-time plasma monitoring. However, this method has some critical limitations: A few selected wavelength signals are analyzed only, and the others are ignored even though the non-analyzed wavelengths?which are more than 99 % of raw data?have hidden information related to reactions in the reactor. The viewport for gathering optical signals can be contaminated due to the by-product polymer deposition. This polymer layer absorbs the light emitted from plasma, and the intensity of target optical signals can be weakened. The RF impedance monitoring system, which is the origin of VI probe, was reported for plasma etching endpoint detection in 1979, and this system has been developed for detecting changes in reactor, plasma and wafer conditions without affecting plasma processes at low cost. VI probe, which is called plasma impedance monitor (PIM) or I-V monitor, is a non-invasive sensor installed between matching network and electrode for monitoring the voltage, current and phase angle. This tool can be applied for measuring the loss of electrical power and detecting the endpoint of etching and cleaning processes if the plasma reactor system is regarded as an equivalent circuit. By means of the equivalent circuit theory, a plasma etcher system can be modeled with a combination of conductors, resistors, capacitors, and inductors. Therefore, it is possible to detect the change of plasma or reactor conditions from VI probe signals. However, the target of VI probe analysis is not only the plasma phase but also the reactor and wafer condition, and the meaning of the voltage, current, and phase angle harmonics have not been identified clearly. This problem disturbs to find out the most sensitive electrical signal for monitoring plasma, and the signal of VI probe should be analyzed empirically. To solve these problems, endpoint and fault detection with multivariate analysis are demonstrated for small area dielectric etching by plasma. The endpoint is determined by the optical and impedance harmonic signals variation from OES, self-plasma optical emission spectroscopy (SPOES) and VI probe. Moreover, modified multivariate analysis techniques are applied to extract information from the enormous amount of correlated data for enhancing the sensitivity of real-time small area etching processes. The results indicate that modified multivariate analysis techniques can enhance the sensitivity than before, and the sensitivity is enhanced. These techniques can be applied to plasma etching processes as a sensitive process monitoring tool.
ContentsList of FiguresList of TablesAbstractChapter 1. Introduction1.1. Research Motivation1.2. Research ObjectivesChapter 2. Plasma Monitoring Techniques2.1. Plasma Diagnosis for Plasma Processes2.2. Optical Emission Spectroscopy (OES)2.3. Self-Plasma Optical Emission Spectroscopy (SPOES)2.4. Plasma Impedance Monitoring (VI Probe)Chapter 3. Multivariate Analysis Techniques3.1. History of Multivariate Analysis for Plasma Processes3.2. Modified Principal Component Analysis (mPCA)3.3. Modified K-means cluster analysis (mKmCA)3.4. Time-series Cluster Analysis (TsCA)Chapter 4. Endpoint Detections4.1. SiO2 Etching on Silicon: OES and VI probe with mPCA4.2. SiNx Etching on Silicon: OES with mPCA and mKmCA4.3. SiO2 Etching on Photo-resist coated Silicon: OES with mPCA, mKmCA, and TsCA4.4. Multi-layers Etching: OES, SPOES and VI probe with mPCA4.5. Patterned Multi-layers Etching: OES with PCAChapter 5. Sensitivity Enhanced Plasma Equipment Monitoring5.1. Experimental Condition5.2. Source/Bias Power and Gas Flow Monitoring5.3. Case Study5.3.1. Abnormal Power Delivery5.3.2. MFC errorChapter 6. ConclusionReferences