ИСТИНА |
Войти в систему Регистрация |
|
ИПМех РАН |
||
Nowadays, to provide permanent scaling down of semiconductor industry one has to overcome the big variety of challenges and some of them are related to interconnect dielectrics: RC-delay, dynamic power consumption and crosstalk. That is why dielectrics with low dielectric permittivity (low-k material) have been introduced into the industry. For sub 10nm node technology porous organosilicate glasses (OSG) with terminating hydrophobic methyl groups are supposed to be the main candidates. Unfortunately, traditional plasma etching of such materials in fluorocarbon containing plasmas leads to significant damage caused by methyl groups loss and subsequent hydrophilization and k-value increase [1]. It is supposed that plasma induced damage (PID) is mainly attributed to plasma vacuum ultra-violet (VUV) photons [2, 3]. Replacing conventional CF4 gas in etching plasmas by CF3I reduces UV photons flux in 200nm-300nm spectra range [4] and this substitution was expected to reduce VUV photons flux as well. Spectra in VUV range (30nm – 200nm) together with spectra in visible and UV ranges (200nm -800nm) were collected in CF3I/Ar and CF4/Ar discharges in industrial CCP chamber from LAM Research with different ratios of fluorocarbon gases to Ar. The pressure was 50mTrr and 150mTrr with constant gas flow of 400sccm, forwarded power – 700W at 27MHz. The VUV spectra were obtained by McPherson 234/302 monochromator with sodium salicylate scintillator and photon multiplier tube (PMT) as a detector. The spectrometer was separated from the chamber by tube of 0.5m length. The spectra in 200nm-800nm range were collected by the tool spectrometer and calibrated on the spectrometer sensitivity. The spectra of CF3/Ar plasma at 50mTrr are presented on the figure 1(a). The broad emission of CF2 at 130nm – 180 nm, Ar double peaks (not resolved) at 104.8nm and 106.7 nm, I lines at 135-136nm, 139nm, 142nm, 146nm, 159nm, 164nm, 178nm-179nm,183nm-184nm are clearly seen on the VUV spectra. As can be seen from the spectra increasing of fluorocarbon gas percentage leads to a significant decrease of VUV intensity whilst there is an opposite behavior of I line at 206nm and CF2 band near 250nm. On the base of this discrepancy we suppose observed VUV spectra behavior to be mainly controlled by strong light absorption in the tube between the spectrometer and the chamber. Assuming that VUV intensity reducing is mainly due to absorption, the absorption cross-sections in VUV range were estimated. We believe the absorption to be mainly attributed to CF3I itself [6] and CxFy oligomers [8]. The total concentration of CxFy oligomers can be quite high [5, 7] and these molecules, most likely, have high absorption cross-sections. We presume these oligomer molecules to be responsible for light absorption near 7eV photon energies (~170nm) that can’t be explained by CF3I absorption. We also suppose that Ar double line (104.8nm and 106.7nm) on all the spectra is strongly eliminated by high Ar self-absorption. Normalizing on unabsorbed CF2 emission band at around 250nm and I emission at 206nm and assuming low changes in ratios of excitation rate constants of VUV spectra lines to lines we normalizing on the VUV spectra were corrected. The latter reproduce the spectra wafer is exposed to. The corrected spectra are presented on figure 1(b). It was shown that substitution of CF4 by CF3I results in even increasing of light intensity in the VUV range. While CF3I is highly polymerizing gas its addition decreases etch rate and subsequently increases VUV radiation dose. So, we suppose there is no motivation for applying CF3I to low-k PID reducing.