Abstract
Plasma processing plays a crucial role in modern semiconductor manufacturing. The plasma processing results are determined by the plasma properties. Nowadays, the trend in the semiconductor industry to process smaller geometries as well as results in recipes being run at lower rf output power into lower pressures. Both factors result in the increase of probability of instability in a plasma source driven by rf power. In addition to the attachment-induced plasma relaxation oscillation occurring within this operating region, interactions between the power-dependent plasma impedance and load impedance-dependent rf delivery system also can introduce a notable system instability. This study focuses on the later one. In this study, two diagnostic tools for characterizing the dynamic temporal behaviors of plasmas were developed. One is the rf impedance meter for measuring the rf power delivered to the inductive coil and the rf impedance of a ICP source. The other is the novel transmission-line microwave interferometer (TLMI) for plasma electron density measurements. However, the temporal variations of electron density are too subtle to detect by the TLMI when instability occurs at low power and low pressure plasma. Thus, in the experimental measurements of Ar instability, a rf-compensated Langmuir probe is used to carry out the measurements of plasma density. In the second phase of this study, the performance of rf impedance meter is tested through the measurements of temporal electric characteristics of a pulsed Ar plasma. The time-resolved measurement results reveal that the magnitude of rf voltage and rf current increase as the duty cycles decrease at high modulation frequencies. A spike of the real part rf impedance due to the transition from capacitive to inductive coupling (E mode to H mode) is observed in the beginning of the modulation pulse. After the transition, the real part of the coil impedance increases and the imaginary part of the coil impedance decreases as plasma density rises. And the time dependences of ion saturation current follow the rf power closely. For plasmas with longer pulse periods, it needs about 0.7 ms to reach steady state at the setting of match network in this experiment. In the study of Ar plasma instability, the measurement results reveal that the behaviors of instabilities will be distinct according to their matching settings. For the high frequency instability, the series capacitor (CL) position is fixed at where a perfect matching condition is achieved. The oscillation frequency is between 10 to 20 kHz and increases with the increasing of chamber pressure and rf power setting and the decreasing of parallel capacitor (CT) position. For the low frequency instability, the CL position is tuned at higher than the perfect matching point and the oscillation frequency is independent of operation condition and fixed at 150 Hz. In addition, oscillations of instability only occur as the CT position is higher than the perfect matching. From the disturbance simulation of the electric circuit model, a negative feedback mechanism between the electron density and the rf delivered power is obtained when a mismatch locates at the inductive reactance side of the Smith chart. As CT position increases from perfect matching point, the input impedance of system moves into the capacitive reactance region. The relation between the electron density and the rf delivered power becomes positive feedback. It leads system into unstable.