ИСТИНА

Neutral beam etching (NBE) is an alternative approach to conventional reactive ion etching (RIE) that avoids structure charging during the etching process. The main advantages of NBE are etching in a (V)UV free environment, the possibility to extract and control the energy of positive (CFx +) and/or negative ions (F-), from the same discharge, and neutralize them before arriving to the substrate, which resides in a volume with lower pressure than that of the discharge. For sub-10 nm CMOS technology ultra-low-k dielectrics are required to eliminate RC-delay, dynamic power consumption and crosstalk. SiO2-based porous organo-silicate glass materials (p-OSG), with high porosity (30% to 50%) and k-values 1.8 – 2.2 are currently the main candidates for the future ULSI production. At such porosity level, pores with typical radius 1-2 nm, are mostly interconnected forming an open structure, rendering the notion of bulk irrelevant. The pores of these dielectrics are covered with methyl groups, making the material hydrophobic, i.e. avoiding moisture adsorption. SiO2-based dielectrics are typically etched in fluorocarbons containing plasmas. P-OSG’s are very sensitive to active radicals and high-energetic plasma photons in the vacuum ultra-violet (VUV) wavelength range. Interactions between p-OSGs and radicals (F*, O*, H*) leads to the degradation of methyl bonds, pore collapse and matrix re-organization leading to -OH groups uptake and a dramatic increase of the k-value [1]. VUV is also known to cause acute low-k degradation by Si-C scission, leading to C depletion and having the same consequences as radicals [2]. Therefore, p-OSG materials are the ideal vehicle for testing the validity of the NBE concept, taking advantage, in this configuration, of the absence of VUV and reduced radical partial pressure. It also presents an opportunity to investigate the interactions of low-k materials with high –energetic fluorine atoms, since a similar study has been conducted on thermal atoms [3] so far. The setup, shown in figure 1, consists of 13.56 MHz Inductively coupled plasma (ICP) source with a top perforated electrode that can be biased in continuous DC/RF or pulsed modes (to control the polarity and energy of the accelerated ions), an extractor electrode (bottom electrode), that separates and the processing region. The extractor, which was grounded, can be either a fine metallic mesh, to extract ions, or a neutralizer used to extract either energetic neutrals (neutralizer). Two types of neutralizers were used in these experiments: graphite and GaAs. The former being easy to manufacture, and the latter being less sensitive to F* erosion, allowing a higher flux (reduced loss) and better beam collimation. The top electrode bias allows to vary the energy distribution of ions/neutrals and ion/neutral beam composition. During processing, the pressure in the sample zone was at least two orders of magnitude lower than in the plasma region. Ion energy distribution and ion flux were measured using a commercial retarding field energy analyser (RFEA) Semion pDC. When one of the neutralisers is placed between the plasma source and the sample area, information about the neutral beam energy distribution can deduced from the RFEA measurements under the same plasma conditions but with the metallic mesh placed between the sample and the plasma regions taking into account the energy shift due to the ion neutralisation process. Low-k p-OSG films with k = 2.0 and porosity ~ 44% were used in the experiments. The low-k damage was investigated by FTIR measurements (Nicolet 6700), Spectroscopic ellipsometry (KLA Tencor F5 SCD). The k-value was measured by mercury probe. The plasma conditions were as follows: CF4/Ar = 1:1, 30 mTorr, 400 W and SF6/Ar = 1:1, 20 mTorr, 400 W. The plasma source was used in CW mode. An example of the results of low-k exposure in full plasma (10 s. exposure) and neutral beam (90 s. exposure) configurations in SF6/Ar are presented in Fig. 2. The full plasma exposure time was chosen to result in similar thickness for the samples exposed in both configurations. The figure shows the equivalent damage layer (EDL), integrated FTIR water band (3100 – 3800 cm-1) and refractive index. The EDL is calculated by the following formula: and gives the thickness of the damaged layer as if all the damage would be concentrated on the top of the film. The thickness of the samples was retrieved from changes in integrated Si-O-Si peak in FTIR spectra. The thickness measured by ellipsometry was systematically lower than the one deduced from the FTIR spectra. This difference is caused by densification of the top of the low-k film during the treatment. The change in thickness was negligible giving a low etch/damage ratio for the studied conditions. The observed 20 nm - 30 nm thickness reduction in the neutral beam case is caused by sputtering and the onset of the etching. The phenomenon observed is similar to the one found by exposing low-k to thermal fluorine radicals, as the thickness reduction starts from some specific radical dose that is required to etch away the material between the topmost pores [5]. In comparison, in full plasma exposure, the etching rate is more than an order of magnitude larger. From Fig.2, after NBE etch, significant damage is found as in terms of EDL (up to 50% of methyl groups were modified in neutral beam configuration) as in integrated water peak. The EDL for full plasma exposure is significantly lower than for full plasma configuration due to much higher etch rate. The correlation between the EDL and moisture band indicates that the damage occurs not only by H abstraction reactions but through the whole methyl groups removal. In, summary, the NBE of low-k, in the described condition, is highly damaging, contrary to expectations. Further investigations & improvements to the technique ae needed, as well as modelling effort looking at the interactions between highly porous low-k dielectrics and high energetic F* radicals.