Plasma diagnostics as a tool for process optimization: the case of microcrystalline silicon deposition

Properties of thin films such as the crystallinity of silicon deposited from SiH4 - H-2 discharges are governed by the plasma composition. Therefore, it is crucial to measure the plasma composition in order to understand and optimize the deposition rate, deposition efficiency and material quality of thin film deposition, such as for microcrystalline silicon, which is a key material for silicon thin film photovoltaic solar cells. This task can be performed by using Fourier transform infrared absorption spectroscopy or optical emission spectroscopy. This work compares these two techniques and shows their range of applicability and their limitations.

Investigations of radio-frequency, capacitively-coupled large area industrial reactor

Benjamin Strahm

The constant energy consumption and world-wide demography expansion add to the potential risks of ecological and human disaster associated with global warming. This makes it a necessity to develop renewable energy technologies such as photovoltaic energy. These technologies already exist, but their cost has to be reduced in order to compete with well-established energies based on natural resources such as oil, coal or natural gas. A first step has been performed successfully in recent years in the field of low cost photovoltaic (PV) solar cells. Manufacturing equipment for large area (> 1 m2) amorphous silicon thin film solar cell production is now available. Even if this type of PV cell has lower energy conversion efficiencies (≈ 9 – 10 %) than other types of cells such as crystalline silicon cells (≈ 25 %), it has lower financial and ecological costs. Nevertheless, the next generation of large area silicon thin film PV cells promise higher conversion efficiencies (≈ 12 %) and stability under light exposure. This new type of PV cell is based on microcrystalline silicon grown on large glass substrates by plasma-enhanced chemical vapor deposition from silane (SiH4) and hydrogen (H2) gas, as for the previous amorphous silicon generation. Amorphous/microcrystalline silicon PV multi-junction cells require a thick (≈ 2 µm) microcrystalline intrinsic light absorber layer because of the need to fit the photo-generated current of the two stacked cells. This increases the cost of the final product since the deposition rate of microcrystalline silicon achieved nowadays is limited to a few Å/s, making the processing time very long. The enhancement of the deposition rate while maintaining a good material quality, i.e. at the boundary between amorphous and microcrystalline growth, is difficult because the phenomena involved in the deposition of microcrystalline silicon are not well understood, and the optimization is then generally performed empirically. The plasma composition is measured using Fourier transform infrared absorption spectroscopy and optical emission spectroscopy. It is shown that the deposited films can be classified into three categories (amorphous, transitional and microcrystalline) as a function of silane concentration in the plasma, while it is impossible to do so as a function of all other process parameters (silane input concentration, RF power, pressure, etc...) if they are all varied simultaneously. This means that the common way to deposit microcrystalline silicon by strongly diluting the silane with hydrogen (< 5 %) is not unique. This is because the plasma composition does not depend only on the gas composition, but also on the fraction of silane depleted in the plasma. Analytical and numerical plasma chemistry modeling show that this is because the silane concentration in the plasma determines the species flux towards the growing film surface. Hence, in agreement with existing phenomenological models of microcrystalline growth, the lower the silane concentration in the plasma, the higher the H to SiHx flux ratio towards the surface and the higher the crystallinity. This is used to demonstrate the feasibility of the growth of microcrystalline silicon in large area reactors using radio-frequency excitation (40 MHz) even from pure silane gas. Moreover, it is shown that the optimum in terms of deposition rate and deposition efficiency tends towards pure silane and not to the common H2-diluted regime. Following this conclusion, an optimization strategy is constructed by varying only the hydrogen flow rate and the working pressure. Guidelines for the selection of the initial process parameters are given in order to achieve high deposition rate (> 10 Å/s) and high gas utilization efficiency (> 80 %) of good quality microcrystalline silicon with a high input silane concentration (> 10 – 20 %) by using the optimization strategy. Furthermore, it is shown by using time-resolved plasma composition measurement and modeling that the time necessary to reach chemical steady-state is about 1 second in large area Plasma-Box™ reactors. This time is much shorter than times reported for small laboratory reactors, which are typically of about 1 minute. It is demonstrated by using two-dimensional modeling that this is due to back-diffusion of the silane molecules from the vacuum chamber to the plasma zone in laboratory reactors, whereas the plasma fills the whole volume in large area Plasma-Box™ reactors, making the plasma composition quasi-instantaneously uniform at plasma ignition. This is of importance for the film microstructure uniformity across the thickness of the layer. In large area Plasma-Box™ reactors, it is not necessary to use strategies such as H2-dilution profiling in order to eliminate the plasma-induced amorphous incubation layer at plasma ignition, because plasma composition suitable to grow microcrystalline silicon is reached directly from ignition. Finally, it is shown that the Telegraph effect and the standing-wave, which are the two necessary and sufficient distinct electromagnetic modes to determine the electromagnetic fields in a RF reactor, affects the film thickness and microstructure uniformity. It is shown that the film thickness variation at the reactor edges due to the Telegraph effect can be eliminated by symmetrizing the two electrodes. It is also shown that the standin-gwave affects significantly the film microstructure of microcrystalline silicon deposited from discharges with a silane input concentration about 10 %, whereas for low input silane concentrations (< 4 – 5 %) the microstructure is more or less uniform. This means that the higher the silane concentration, the better has to be the design of the lens-shaped electrode to compensate for the standing-wave in large area reactors.

EPFL2007

Design, characterisation and modelling of a high current DC arc plasma source for silicon and silicon carbide processing at low pressure

Lukas Derendinger

In the frame of this thesis, two similar high current DC arc (HCDCA) plasma sources were investigated in a low gas pressure regime (10-3-10-2 mbar). One of them was initially designed for the epitaxial growth of silicon and silicon germanium (LEP), the other for the industrial deposition of diamond (BAI). The LEP source was analysed using pure argon plasmas. Measurements of the ion saturation current were performed with a custom-built multi-Langmuir probe to analyse plasma density homogeneity. Plasma instabilities at 50 Hz were observed and studied by different means. One source of the instability is the AC current used for the filament heating, whereas lower frequency instabilities are due to the use of a ring shaped anode. A sensitive Hall sensor was used to measure the magnetic field induced by the discharge current. It was found that depending on the plasma parameters gas pressure and external magnetic field the current tends to attach at different points on the anode, leading to a complete loss of reactor symmetry. The installation of an additional cusp field around the reactor chamber lead to an increase of the overall homogeneity of the plasma density, but it could not resolve the problems of current attachment. In the following the ring anode was replaced by a point anode and the maximum external magnetic field strength was increased by a factor of ten. With a novel multi-Hall probe the current density of the now columnar shaped plasma was measured, showing a strongly peaked current density profile, when an external magnetic field is applied. Together with ion saturation current measurements made with a Langmuir probe, the electron temperature inside the plasma column was estimated to be about 4 eV. In earlier works made on the BAI reactor the high dissociation efficiency of the HCDCA plasma source has already been shown. Optical emission spectra were compared between RF plasmas, low pressure (1.5mbar) HCDCA plasmas and our very low pressure (10-3-10-2 mbar) HCDCA plasmas. The dominating species found in RF plasma spectra are molecular, the spectra of low pressure HCDCA plasmas are dominated by atomic species and the spectra of very low pressure HCDCA plasmas are dominated by ions, emphasising the high dissociation efficiency of this system. Silane and Methane was used as precursor gases in the BAI reactor for the deposition of silicon carbide films. Promising high deposition rates up to 9 nm/s were found, but FTIR spectroscopy showed high hydrogen and oxygen concentrations in the porous films, making the not optimised deposited material useless for the application of wear-resistant coatings. Microcrystalline hydrogenated silicon (µc-Si:H) is viewed as a cost- and energy-effective alternative to crystalline silicon for the production of solar cells. A detailed analysis of the deposition rate and the Raman crystallinity of µc-Si:H films deposited in the BAI reactor showed deposition rates up to 6.5 nm/s and a wide range of crystallinity from 0-80%. Film thickness inhomogeneity in silicon solar cells has to be less than 5%. To meet this standard a first substantial improvement was achieved with the installation of a linear gas injection along the plasma column. Tests on large surface glasses (47 × 37 cm2) revealed strong diffusive effects, which could be reproduced with a simple gas diffusion model. The model showed the necessity to reduce the dead volume around the plasma and to set the substrates as close as possible to the plasma column in order to minimise film thickness inhomogeneity due to diffusion. Deposition rate measurements made in these conditions assured the results of the model. The development of a method to estimate the dissociation efficiency of the plasma by simple pressure measurements showed also an important increase of the silane dissociation from 75-92% when the plasma is confined in a smaller volume. Therefore, compared to the reactor with a large dead volume, only a third of the initial silane is lost to the pumps.

EPFL2008

Non-intrusive plasma diagnostics for deposition of large area thin film silicon

Christoph Hollenstein, Alan Howling, Benjamin Strahm

Plasma diagnostics for large area, industrial RF parallel-plate reactors can be useful for process optimization and monitoring, provided that their implementation is practical and non-intrusive. For instance, Fourier transform infrared (FTIR) absorption spectroscopy and/or time-resolved optical emission spectroscopy (OES) can easily be retro-fitted into the pumping line of a reactor. Both techniques were used to measure the fractional depletion of silane in silane/hydrogen plasmas. By means of a simple analytical plasma chemistry model, it is shown that the silane depletion is related to the silicon thin film properties such as microcrystallinity. Uses of the diagnostics are demonstrated by two examples: (i) the optimal plasma parameters for high deposition rate of microcrystalline silicon, along with efficient gas utilization, are shown to be high input concentration and strong depletion of silane; and (ii) the optimal reactor design, in terms of fast equilibration of the plasma chemistry, is shown to be a closed, directly-pumped showerhead reactor containing a uniform plasma. (C) 2009 Elsevier B.V. All rights reserved.