PhD Thesis abstract, R.J. Chatfield, 1994

R.J. Chatfield, 'Mass and Optical Spectroscopy of CF4 + O2 Plasmas and their Application to the Etching of Si, Ge and SiGe Alloys', PhD Thesis, 1993.


CF4 + O2 plasmas are widely used to etch Si in VLSI technology. Despite many investigations, much remains to be learnt of the important chemical mechanisms. We have applied Quadrupole Mass Spectrometry (QMS) with capillary sampling and Optical Emission Spectroscopy (OES) simultaneously to study this system and included the etching of Ge and SiGe alloys. The approach used was to fully characterise CF4 + O2 plasmas with no wafer present and then to introduce Si, Ge and SiGe substrates. SiGe alloy has recently come into prominence because it shows promise for the production of fast-switching Heterojunction Bipolar Transistors and novel optoelectrnnic devices.

Our investigations were carried out using a single slice 13.56 MHz Reactive Ion Etcher (Omega, Electrotech) under a variety of conditions of flow rate, power, pressure and substrate area, and also with small additions of O2, CO, CO2 and C2F6 to the process gas. For each set of conditions, detailed measurements of the plasma composition were made. Using QMS, the absolute concentrations of CF4, O2, CO, CO2, C2F6, COF2, F2, SiF4 and GeF4, were determined by applying appropriate correction factors. In addition, OES observations in the range 6200 to 7400 allowed the measurement of emission intensities from F atoms in a variety electronically excited states.

A kinetic model involving 53 gas phase and surface reactions was constructed for the CF4 + O2 system with no wafer present. Using steady state considerations, it was possible to calculate [CF3] and [F] from the QMS results over a wide range of experimental conditions. Calculations of [F] agree with trends observed by OES, but only when wall reactions involving F are included in the model. These wall reactions are shown to be inhibited by the presence of O atoms, because O atoms compete with F atoms for surface sites. The kinetic model was then tested for one specific set of conditions (40 sccm, 100 W and 85 mTorr) to see if it could reproduce the concentrations of species measured experimentally. Only 17 of the original 53 reactions are found to be important. This test also reveals the main chemical pathways, and demonstrates the importance of certain wall reactions, which when included in the model allow product distributions to be reproduced to within a factor of 3.

Si and Ge wafers were etched at various conditions of flow rate, power, pressure and exposed wafer area. Typical etch rates lie in the 1000 to 3000 /min range for 3-inch wafers, with the Ge etch rate always faster than that of Si, typically by a factor of around 1.3. The results indicate that the etching mechanisms for Si and Ge are similar, although the etching of Si relies more on ion bombardment than does the etching of Ge. F atoms are the main etchant species for both Si and Ge, and the only observed products of etching were SiF4 and GeF4. The Si budget is present almost exclusively as SiF4, whereas up to 40% of the Ge budget is present as Ge- containing species adsorbed at the chamber walls. The well-known fall in F atom emission intensity on the introduction of a wafer to the plasma is found not to be due to the direct removal of F atoms by reactions at the wafer surface during the etching process, as was previously thought. The effect is shown instead to be due to a reduction in the effective fluorine-to-carbon ratio in the plasma caused by the etching process. This results in a change in the composition of the plasma, which in turn results in an increase in the removal rate of F atoms via fast reactions in the gas phase and at the chamber walls. A mechanism is proposed which explains qualitatively the changes in the plasma composition observed by QMS when a wafer is introduced to the chamber.

OES studies show that emission from different electronically excited states of F atoms behaves differently when the experimental conditions are altered. Specifically, emission from the two halves of the 3p 2P0 doublet behave differently from each other, and also from all the other states observed which all exhibit the same behaviour. This observation is shown to be linked to changes in [O2] in the gas phase; the emission from the 3p 2P0 states is relatively stronger in the presence of O2. It is proposed that this effect is caused by resonant energy transfer from excited O2 molecules in the 1Ψu state, which has similar energy to the 3p 2P0 states of F. In the presence of O2, this process competes with electron impact processes for the production of these two excited states. The magnitude of the effect is observed to be 1.8+0.3 times stronger for the 3p 2P03/2 state than for the 3p 2P01/2 state. This figure is correctly predicted by calculations.

Studies into the etching of SiGe alloy layers on Si wafers with 33 and 67 mol% Ge content have also been undertaken over a wide range of experimental conditions. The purpose was to see if SiGe alloy material exhibits etching behaviour which is consistent with a simple mixture of its Si and Ge component parts, or whether it shows unexpected behaviour. We find that SiGe alloy shows the expected etching behaviour for all the conditions studied, except at low power (20 W, 85 mTorr, 5 sccm) where it exhibits novel fast etching, with corresponding changes in the gas phase composition of the plasma, viz. lower [F] and higher [C2F6]. The problems of endpoint detection and process control, which are crucial for the fabrication of devices, have also been addressed. It is clear that QMS monitoring can cope with both of these problems. It has also been shown that species suitable for endpoint detection must be found after the process has been optimised, because the changes observed in species.