| Low temperature plasma technology plays an important role in the development of the semiconductor industry. Capacitively coupled plasmas (CCPs) are commonly used as etching and deposition devices in microelectronic manufacturing, due to the simple geometry structure and the ability of producing large area and uniformity plasmas. In the past several decades, the microelectronics industry has developed very rapidly, and the critical dimensions of semiconductor devices shrink continuously, hence this requires strict control of the CCP in the various processes. Motivated by this need, CCPs are evolving all the time, and many new plasma sources with special features are generated, such as the dual-frequency operation of CCPs, which can, to a large degree, independently control the plasma density and ion energy; the direct current/radio frequency (dc/rf) CCPs, which can alleviate the charging of the bottom of high aspect ratio features etched in insulators; and the electrically asymmetric CCPs, which can control the ion energy in a flexible way. In the development of these plasma sources, numerical simulations play an essential role in the fundamental researches for CCPs. Numerical simulations can not only give information on trends of the plasma behavior for different conditions, but also explore the physical mechanisms in CCPs. The most commonly used simulation methods include fluid models and PIC/MCC (Particle-in-cell/Monte Carlo collision) models. Although a fluid model runs with good stability, it cannot consider the non-local behavior in CCPs, and it can not acquire the energy distribution of the charged particles. A PIC/MCC model is a fully kinetic method based on the first-principles, so it does not have any of the above shortcomings. However, a PIC/MCC model is very time-consuming, since it needs to track a large number of particles and the time and space steps need to be very small to reach convergence. Nevertheless, with the rapid development of computer technology, computers become faster and faster. Thus, PIC/MCC models are getting more and more applications.The main purpose of the presented work is studying the characteristics of two new operations of CCPs, i.e. a dc/rf CCP and an electrically asymmetric CCP. A dc/rf CCP is mainly used to alleviate the charging of the bottom of high aspect ratio features etched in insulators, by introducing an extra dc source on the electrode opposite to the substrate electrode. An electrically asymmetric CCP is driving the discharge by a voltage waveform, that contains a fundamental frequency and (at least) one even harmonic. By tuning the phase angle between the driving frequencies, a self-bias can be generated on the powered electrode, and flexible control of the ion energy can be thus achieved. Although these two new CCPs have been verified experimentally and good processes can be obtained, the physical processes occurring in the plasma are still not completely clear. We will study in detail the influence of different discharge parameters on the plasma density, electric potential and heating rate in these two types of CCPs by a PIC/MCC model.In chapter1, we introduce the application of plasma technology in microelectronic manufacturing, as well as several commonly used plasma sources. Subsequently, we pay particular attention to the key issues in dry etching processes, and to the research progress of dc/rf CCPs and electrically asymmetric CCPs.In chapter2, we present the algorithms of the PIC/MCC model and of external circuit model. According to the flow of the PIC model, the weighting scheme, electrostatic field solving and particle pushing are outlined, respectively. In the MCC part, we mainly introduce the null-collision method and the method for calculating the post-collision velocity of different collision types, such as elastic, excitation, ionization, recombination and endothermic reactions. In comparison with the normal collision method, the computational saving can be quite significant by using the null-collision method. Furthermore, several external circuit models, including the dielectric equivalent-circuit model, the physical circuit model and a general two-dimensional circuit model, are given.In chapter3, firstly, the surface-charging effect in dc/rf CCPs is investigated by incorporating the equivalent-circuit module to the PIC/MCC model. A large number of high-energy electrons are propelled to the substrate, and a self-bias is built up under the effect of the negative dc voltage. However, with increasing the dc voltage, the plasma bulk is largely compressed, which results in a monotonous decrease of plasma density. When increasing the thickness of the substrate, the self-bias increases and the net dc voltage between the two electrodes decreases. Secondly, we study the heating mechanism of electrons in dc/rf CCPs. In a single frequency discharge, the superposed dc source always weakens the overall heating rate, so that the plasma density drops by reducing the plasma bulk length. In the dual frequency (DF) discharge, there exist clearly two competitive effects due to the coupling of dc and DF source:the addition of a dc bias weakens the heating rate in the first LF half-period, while it can drastically enhance the stochastic heating in the second LF half-period. Finally, we investigate the heating mode transition in a dc/DF capacitively coupled CF4discharge. When applying a superposed dc voltage, the plasma density first increases, then decreases, and finally increases again, which is in good agreement with experiments. This trend can be explained by the transition between the four main heating modes, i.e. DF coupling, dc and DF coupling, dc source dominant heating, and secondary electron dominant heating.In chapter4, firstly, simulations based on the one-dimensional PIC/MCC model are performed to study the electrical asymmetry effect on electronegative oxygen discharges, and the results are compared with those of electropositive argon discharges at the same conditions. The self-bias voltage, in both electropositive and electronegative discharges, increases almost linearly with the increase in the phase angle θ between the two driving voltages in the range00≤θ≤90°, and the maximum ion energy varies by a factor of3. However, relative to the small variation of ion flux in an argon discharge (i.e.±5%at30mTorr,±12%at103mTorr), the ion flux varies with θ by±12%at30mTorr and by±15%at103mTorr in the oxygen discharge. This may place a limitation for achieving separate control of ion energy and flux in electronegative plasma. Secondly, using a combined experimental and numerical approach, we investigate the control of plasma properties via the electrical asymmetry effect. It is shown that the self-bias decreases with increasing pressure, and the density profiles and electric potential distribution have a clear dependence on θ, while the peak densities stay rather constant (within±10%). At low pressure, the plasma series resonance (PSR) is self-excited, which results in high frequency oscillations in the temporal evolution of the conducting current. The amplitude of the PSR is just too small to cause a reasonably enhanced electron heating. Thirdly, a two-dimensional PIC/MCC model is used to investigate both the geometrical and electrical asymmetry effects in CCP. When changing the ratio of the top and bottom electrode surface areas and the phase shift between the two driving frequencies, the induced self-bias, power deposition and parameter optimization were found to develop separately. Moreover, it is shown that both the axial and radial plasma density distributions can be modulated via the electrical asymmetry effect, by adjusting θ and the ratio between the high and low harmonic amplitudes. The modulation of the radial plasma density by the electrical asymmetry effect is found for the first time.The influence of the dc bias on the properties and internal physical mechanisms of electropositive and electronegative plasmas are studied in this work. However, in most applications, a mixture of gases revealing complex chemical and physical features is employed, and the uniformity of the radial density distribution greatly affects the surface processes. Thus, in future work, a two-dimensional model based on a mixture of gases should be adopted for more accurate simulations of a dc/rf CCPs. |
PDF Full Text Request