Silicon etching in NF3/O2 remote microwave plasmas
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
Silicon etching in NF3 /O2 remote microwave plasmas P. J. Matsuo,a) B. E. E. Kastenmeier, and G. S. Oehrleinb) Department of Physics, University at Albany, State University of New York, Albany, New York 12222 J. G. Langanc) Air Products and Chemicals, Inc., Allentown, Pennsylvania 18195 共Received 21 April 1999; accepted 18 June 1999兲 The etching of silicon in remote microwave discharges fed with NF3 /O2 has been investigated. In situ ellipsometry and x-ray photoelectron spectroscopy 共XPS兲 were used to monitor surface effects, while mass spectrometry was used to monitor the gas phase dynamics. Varying the microwave power from 600 to 1400 W has little effect, due to the near complete dissociation of the NF3, even at lower powers. For discharges containing pure NF3, the poly-Si etch rate increases linearly with NF3 flow. When a low proportion of O2 (O2 /NF3⫽0.1) is added to the discharge, the etch rate increases quickly to its maximum of ⬃700 nm/min. With further O2 addition, this etch rate decreases below that observed for pure NF3 processing. The fluorine concentration in the processing region decreases for all O2 additions by a dilution effect. For pure NF3 discharges, XPS measurements reveal 1–2 nm thick, highly fluorinated reaction layers with a gradual loss of fluorine content as the NF3 flow is increased. Specimens processed with both NF3 and O2 show much less surface fluorination that decreases with increasing O2 content in the feed gas. At the etch rate maximum, the observed N (1s) signal is also maximized. The reaction layer thickness increases with added O2 and continues to more than 10 nm at O2 :NF3 ratios greater than unity. We discuss the enhanced reactivity of the modified Si surface and compare our results with the role of admixed N2 into the CF4 /O2 system. We also injected NO directly into the effluent of NF3 and CF4 /O2 discharges. For fluorine rich discharges, NO removes the modified surface layer on Si and provides for an enhanced etch rate. In the oxygen rich regime, NO injection can increase both the etch rate and the reaction layer thickness. We will present a mechanism for the enhanced etching of Si in the presence of fluorine, oxygen and the NO molecule. © 1999 American Vacuum Society. 关S0734-2101共99兲10205-7兴 I. INTRODUCTION The resulting maximum in the fluorine concentration corre- lates with the etching behavior of Si.6,7 Adding O2 to a low Halogen based plasmas have been utilized extensively to density NF3 discharge can similarly enhance the fluorine produce reactive atomic species for the etching of silicon and density.8,9 In this work we used high flows 共300 sccm NF3兲 certain dielectrics. Of these systems, CF4 and CF4 /O2 plas- and high microwave powers 共600–1400 W兲 and saw no en- mas are the best understood. In our previous work, we have hancement in fluorine densities, but rather a reduction due to investigated the role of N2 addition to these discharges.1,2 a dilution effect. The etch rates, however, still displayed a Here, we will build upon this work and compare this system maximum after O2 addition. with the mechanisms of Si etching in the downstream efflu- ent of a NF3 /O2 microwave plasma. There has been a previous study of the dissociation levels In comparison to CF4, the use of NF3 as a substitute of NF3 as compared to CF4. 3 In a rf discharge, NF3 has been source of reactive fluorine is attractive in three ways: 共1兲 the shown to dissociate much more efficiently than CF4 共by more enhanced dissociation of the parent molecule,3 共2兲 the elimi- than an order of magnitude兲. In our microwave discharge, we nation of carbonaceous residues on the etched surface and observed near complete dissociation of the NF3, even at our 共3兲 as an alternate source of fluorine due to the increasing lowest power setting. demands for more environmentally compatible process Due to the strong evidence for a nitric oxide assisted etch gases. It is primarily as this alternate fluorine source and mechanism, and the role that both O2 addition into a NF3 specifically for its development into a reliable chamber discharge and N2 addition into a CF4 /O2 discharge play in cleaning process for amorphous silicon deposition that NF3 the formation of NO, we sought to support our conclusions has been studied.4,5 by the direct downstream injection of NO into the processing It is well known that the addition of small amounts of O2 chamber with both NF3 and CF4 /O2 gas mixtures fed into the to a CF4 plasma enhances the atomic fluorine concentration discharge. Similar experiments performed while etching by limiting recombination through oxidation of CFx radicals. Si3N4 have shown the NO molecule to be central to a mecha- a兲 nism for nitrogen removal.10–12 When a low flow of NO is Electronic mail: pm1747@csc.albany.edu b兲 Electronic mail: oehrlein@csc.albany.edu injected downstream from a NF3 discharge (NO/NF3⫽0.17), c兲 Current address: Air Products and Chemicals, Inc., Santa Clara, CA 95054. mass spectrometry results indicate that surface oxides are 2431 J. Vac. Sci. Technol. A 17„5…, Sep/Oct 1999 0734-2101/99/17„5…/2431/7/$15.00 ©1999 American Vacuum Society 2431
2432 Matsuo et al.: Silicon etching in NF3 /O2 remote microwave plasmas 2432 FIG. 2. Etch rates of poly-Si vs NF3 flow at varied microwave powers. FIG. 1. Schematic of the remote microwave plasma reactor used in this study. discharge was maintained which provided the best window for a stable NF3 discharge across a varying pressure range. removed which leads to enhanced Si etching. When CF4 /O2 For the etching and surface analysis experiments, 2.5 cm chemistries are used and when high levels of fluorine are by 2.5 cm square samples were prepared from 125 mm wa- available for the Si etch 共e.g., O2 :CF4⫽0.15兲, injection of fers covered with 580 nm polysilicon on 1000 nm SiO2 on Si NO (NO/CF4⫽0.125) into the effluent of a CF4 /O2 dis- and 125-mm-diam single-crystal silicon wafers. These charge results in a strong enhancement in the poly-Si etch samples were HF dipped to remove the native oxide layer rate. Again, we attribute this to the removal of the SiFx Oy and were then mounted on 125 mm silicon carrier wafers reaction layer formed without NO. If a high O2 :CF4 ratio is using thermal glue. The mounted samples were then placed used, the NO has both the effect of increasing the poly-Si on an electrostatic chuck 共dc bias of 600 V兲 with a helium etch rate and the thickness of the modified surface layer. We backside pressure of 5 Torr. The electrostatic chuck tempera- postulate that this effect is possibly due to the formation of a ture was set to 10 °C. The wafer temperature was measured highly dynamic 共liquid-like兲 reaction layer on the etching with a fluoroptic probe which contacts the backside of the surface. Even though at any instant, this layer is quite thick, wafer. During processing, the silicon wafers heated up by in its steady state composition, this layer is highly reactive, approximately 2 °C. allowing for the rapid attack of the Si by fluorine and the In situ ellipsometry using a rotating compensator ellip- desorption of the etch products. someter in the polarizer-compensator-sample-analyzer 共PCSA兲 configuration with a 632.8 nm He/Ne laser source II. EXPERIMENTAL SETUP beam was employed to monitor real time film etching and In Fig. 1, the ultrahigh vacuum-compatible microwave deposition. Optical emission spectroscopy 共OES兲 and mass based etching reactor used in this work is shown. A micro- spectrometry were used for real-time plasma and chamber wave plasma is produced in a modified ASTeX DPA-38 mi- gas phase diagnostics. The etching chamber is also con- crowave plasma applicator. This device is equipped with a nected via an ultrahigh vacuum 共UHV兲 central wafer handler 38-mm-o.d. wall water cooled sapphire tube to which the to both a load-lock and a multi-technique surface analysis process gases are fed. The plasma is separated from the pro- chamber. Processed specimens were transported under cessing chamber by tubing of variable length and lining ma- vacuum to the multitechnique surface analysis system for terial. For the current work, this lining material was polytet- x-ray photoelectron spectroscopy 共XPS兲 measurements. rafluoroethylene 共PTFE兲 and the lengths studied were 0 and 75 cm. A gas line connected directly to the processing cham- III. RESULTS ber allows for the direct injection of process gases, bypassing A. Etch rates the plasma excitation region. The etching experiments were performed in a stainless The silicon etch rate versus flow for a pure NF3 discharge steel cylindrical processing chamber 共inner diameter 27 cm, is shown in Fig. 2. These rates display a linear enhancement height 30 cm兲 with water cooled walls. The system is with increasing NF3 flow. This is explained by the near com- pumped using a Balzers 2200 l/s corrosive service turbomo- plete dissociation of the NF3 in the microwave discharge. lecular pump backed by a roots blower and vane pump. The Figure 3 shows this in terms of two sample mass spectrom- pressure was measured with a MKS Baratron capacitance etry analog spectra. Panel 共a兲 representing the plasma off manometer. For the experiments described here, the pressure condition and panel 共b兲 with a 1400 W microwave discharge. was kept at 1.0 mTorr. A throttle valve in the bypass line When the discharge is ignited, we observe the complete loss was used to control the pumping speed and maintain this of the 71 amu NF3 peak, the 52 amu NF2 peak and the 33 pressure in the chamber. Although, in general, the poly-Si amu NF peak as well as the formation of the 38 amu F2 etch rate is independent of the microwave power, a 1400 W signal and the 19 amu F peak. Also, the nitrogen signal is J. Vac. Sci. Technol. A, Vol. 17, No. 5, Sep/Oct 1999
2433 Matsuo et al.: Silicon etching in NF3 /O2 remote microwave plasmas 2433 FIG. 5. Mass spectral data for NO and N2O vs O2 content in the feed gas. to compete for the removal of this oxygen. XPS data pre- FIG. 3. Analog mass spectra illustrating the complete dissociation of NF3 in our discharge. sented later will show that the surface is slightly oxidized in the ‘‘zero oxygen’’ experiments. The gas phase data of Fig. 5 show that there is a high level of N2O at this zero oxygen dramatically increased beyond the residual magnitude seen level. It is known that N2O is a good oxidizer of Si. Only in panel 共a兲. We proceeded in spite of the background N2 when O2 is injected into the discharge does the NO signal contamination, justifying that in these chemistries N2 will appear, and then at a local maximum. always exist in a concentration several orders above the more We also acknowledge that nitrosyl fluoride species should reactive species. Blain, Jarecki, and Simonson13 have ob- be present when O2 is added to the feed gas.14 One might served, through Cl titration experiments, that the transfer of speculate that these species are a more reactive fluorine the atomic fluorine from the discharge to the reactor can be source for the etching reaction. Without O2 addition, the ef- almost 100% efficient. We suspect that a substantial amount ficiency of the heterogeneous recombination of atomic fluo- of fluorine recombines to F2, on our reactor walls. There was rine on the reactor walls results in a high density of F2, little dependence of the etch rate on the microwave power presumably a less efficient source.14 This may account for used. In fact, for the range of 600–1400 W, both the etch the initial etch rate enhancement. In any case, we were not rates and the complete dissociation of the NF3 remain un- able to detect these species with our mass spectrometer, as it changed. When small amounts of O2 were added to the feed dissociates before ionization. gas, an increase of the etch rate was observed. This increase quickly reached its maximum at an O2 :NF3 ratio of 0.1. B. Surface analysis These results are presented in Fig. 4. As the O2 concentration After processing, specimens were transported under UHV is further increased, the etch rate is suppressed to a magni- conditions to the surface analysis chamber for XPS measure- tude below that observed for the pure NF3 discharge. This ments. Figures 6共a兲 and 6共b兲 present these results for the suppression of the etch rate is easily accounted for by the samples processed with pure NF3 discharges of varying passivation of the reactive surface by an oxide-like film. The flows. Panel 共a兲 shows the Si 2p and O 1s spectra, while 共b兲 mechanism behind the initial increase is not as clear. The shows the N 1s and F 1s core levels. From these spectra, one SiO2 chamber viewport acts as an oxygen source, however can see that the modified silicon surface contains silicon, the NO formation takes place only when the oxygen source fluorine, oxygen, and nitrogen after processing. Our previous is in the discharge region. This viewport-born oxygen serves work has shown that after CF4 /O2 /N2 processing, the modi- to passivate the sample surface, while the NO is unavailable fied surface layers contain only silicon, fluorine and oxygen, while the nitrogen and carbon signals are not present. As the flow of NF3 is increased from 100 to 300 sccm, both the F 1s and the N 1s peak areas decrease. When the flow is further increased to 500 sccm, however, this decrease is less pro- nounced. Also apparent in Fig. 6 are the asymmetric peak shapes in the F 1s, O 1s, and N 1s spectra. Again, we attribute the oxygen contamination to the SiO2 chamber viewport. The Si 2p peaks of Fig. 6 were fitted and the results were used to determine the apparent overlayer thickness by a method described elsewhere.15,16 These results are presented in Fig. 7. XPS spectra were also taken in response to varying O2 flows added to the NF3 feed gas. In Fig. 8共a兲, the Si 2p and O 1s core levels are shown. As the panels progress from left FIG. 4. Poly-Si etch rates vs O2 content in a NF3 discharge. to right, from an untreated reference sample to one treated at JVST A - Vacuum, Surfaces, and Films
2434 Matsuo et al.: Silicon etching in NF3 /O2 remote microwave plasmas 2434 FIG. 6. 共a兲 Si 2p and O 1s XPS spectra for samples processed in pure NF3 FIG. 8. 共a兲 Si 2p and O 1s XPS spectra for samples processed in O2 /NF3 at at varied flows. 共b兲 N 1s and F 1s XPS spectra for samples processed in varied ratios. 共b兲 N 1s and F 1s XPS spectra for samples processed in pure NF3 at varied flows. O2 /NF3 at varied ratios. an O2NF3 ratio of 2:1, the level of oxidized Si monotonically nm when a high portion of O2 is injected. The apparent increases. The O 1s spectra of the upper panels do not cor- thickness obtained from the Si 2p core levels of the XPS respond to this change. As the O2 content increases from spectra was used to obtain the film thickness corresponding 10% to 50%, the O 1s peak area remains constant, while in to 10% O2, because surface roughening makes the ellipso- the Si 2p spectra, the bulk is no longer visible. In Fig. 8共b兲, metrically determined thickness unreliable. In the case of where the N 1s and F 1s spectra are shown, there is a strong pure NF3 processing, the surface modifications are due to the maximum in the N 1s signal at 10% O2 addition. These development of a thin overlayer. In the case of NF3 with trends suggest that the NO molecule may account for a sig- 10% O2 addition, the sample surface is roughened, rather nificant portion of the oxygen signal coming from the sample than modified by the growth of a film. This roughening is surface. characteristic of small oxide islands left on the Si surface During processing, the sample surface was monitored in which serve as micro-etch masks. When higher flows of O2 real time with a He–Ne ellipsometer. This allowed us to are added, roughening is not an issue and thick layers form determine the film thickness in situ. These film thicknesses on the sample surface. versus O2 content in the feed gas are shown in Fig. 9. Thick- nesses range from 1 nm for pure NF3 processing to above 12 C. Downstream NO injection 1. NF 3 plasma The surface analysis of NF3 /O2 treated samples and the mass spectral results have suggested that a NO assisted mechanism, where NO reacts with the surface to remove the FIG. 7. Thickness of the modified surface layer after exposure to the effluent of a pure NF3 plasma. FIG. 9. Thickness of the modified surface layer for O2 /NF3 processing. J. Vac. Sci. Technol. A, Vol. 17, No. 5, Sep/Oct 1999
2435 Matsuo et al.: Silicon etching in NF3 /O2 remote microwave plasmas 2435 FIG. 10. Comparison of etch products 共left兲 and reactants 共right兲 for Si etching. NO was injected downstream of the NF3 discharge. passivating oxide, plays a large role in the etching reaction. FIG. 12. Schematic of the mechanistic role NO plays in etching Si in the presence of fluorine and oxygen. The importance of NO in the etching of Si3N4 in similar chemistries has been established. In the case of Si3N4, the NO acts to remove nitrogen from the surface.10–12 We sought slow with NO, even in response to pure NO injection. In Fig. to determine the products of such a reaction with Si by using 11, which depicts the oxidation reactions, the 7% drop in the a Leybold Transpector mass spectrometer modified so that 28 amu N2 and 14 amu N signals may be in response to the the orifice could sample the gas phase immediately above the 10% increase in the 44 amu N2O signal. This is consistent sample surface. The entire unit is mounted on a vertical lin- with the N2O oxidizing the Si surface. Finally, the fluorine ear motion drive so that the orifice can be lifted without etch reactants and the lower density NFx reactants are clearly changing the orifice-ionization chamber distance. More in- presented. A schematic of these etching reactions is pre- formation on this apparatus can be found elsewhere.17 The sented in Fig. 12. In reaction I, fluorine attacks the Si to form results of a sampling taken during the Si etch are presented the volatile etch product SiF4. Reaction II is similar, though in Figs. 10 and 11. For these experiments, the microwave the fluorine source contains nitrogen. Some of this nitrogen power was 900 W and the transport tube was set adjacent to may be left on the sample surface. Reaction III represents the the processing region 共0 cm setting兲. A total of 300 sccm of role of the NO molecule, that is to remove oxidation in the NF3 was fed into the discharge, while 30 sccm of NO was form of NO2. Reactions IV and V are both oxidation mecha- injected directly into the processing chamber, bypassing the nisms competing with the etching reaction. The surface discharge region. For the first 40 scans, the orifice was just analysis of the NF3 /O2 system presented earlier indicated a 0.5 mm from the sample surface. After the fortieth scan, the maximum of the N 1s signal at the etch rate maximum. We orifice-sample distance was quickly increased to 20 cm. In postulate that this corresponds to reaction III, except where this way, the products of the etching reaction are clearly the main product is O2. In the surface region, a lot of oxygen distinguishable from the etch reactants. In Fig. 10, these is lost to oxidation reactions. Even in conditions where products are displayed on the left while the reactants are nearly all of the oxidation is then removed, a significant por- displayed on the right. As one would expect, the dominant tion is not in the same form as when it reacted 共e.g., NO2 product is SiF4 as indicated by the 78% drop in the 85 amu instead of O2兲. Under such consideration, one would expect SiF⫹3 peak after scan number 40. The loss of fluorine to the that the ‘‘orifice down’’ O2 signal would be higher, even etch reaction is clearly visible in the 5% increase of the sig- though O2 is coming off as a product of the passivation layer nals when the orifice is moved away from the surface. This etching. This can explain the slight increase in the O⫹ 2 signal same behavior was observed for low intensity NFx species. of Fig. 11, after the orifice has been lifted from the sample The 10% drop in the 46 amu NO2 signal nicely supports the surface. previous hypothesis that NO acts to remove oxidation free- ing Si bonding sites for fluorination. The rise of the 30 amu NO signal after 40 scans is perfectly consistent with this. The 2. CF4 /O2 plasmas time constant for the mass spectrometer response is always We also investigated direct NO injection in conjunction with the effluent from a CF4 /O2 discharge. Some changes in experimental conditions must be noted for these experi- ments, namely the applicator to which the microwave power is applied was quartz, not sapphire. Again, the 75 cm PTFE transport tubing was removed, and the discharge was located adjacent to the processing chamber, i.e., the discharge edge to sample distance was 15 cm. When the NO was injected into the effluent of a pure CF4 discharge, there is an increase in the silicon etch rate. This increase is more pronounced than when nitrogen alone is FIG. 11. Comparison of the oxidation products 共left兲 and reactants 共right兲 injected into the discharge.1 When the discharge also con- during Si etching. NO was injected downstream of the NF3 discharge. tains admixed oxygen, so that the effluent is fluorine rich JVST A - Vacuum, Surfaces, and Films
2436 Matsuo et al.: Silicon etching in NF3 /O2 remote microwave plasmas 2436 FIG. 13. Poly-Si etch rates vs downstream NO injection in the effluent of CF4 /O2 plasmas. FIG. 15. Loaded SiF⫹ ⫹ ⫹ 3 , NO2 , and O2 mass spectral intensities vs NO (O2 :CF4⫽0.15), the etch rates are much greater. These re- injection in the effluent of fluorine rich CF4 /O2 plasmas. Loaded refers to Si sults are shown in Fig. 13. When the NO is passed into the loaded chamber 共plasma on–plasma off兲 minus a CFx loaded chamber effluent of an oxygen rich discharge, both the etch rate and 共plasma on–plasma off兲. the overlayer thickness increase. This fast etching through thick overlayers is presented in Fig. 14 and suggests a very reactive and highly dynamic film. Mass spectrometry loading evidence that a fraction of the NO arrives at the Si surface, experiments were performed where the chamber was loaded donates its oxygen to the removal of a surface oxygen in the with a 125 mm silicon wafer or a 125 mm fluorocarbon film form of O2, and leaves the nitrogen atom behind. coated wafer. Figure 15 shows these results for fluorine rich Real-time ellipsometry measurements taken while 50 processing for the 85 amu SiF3, the 46 amu NO2 and the 32 sccm NO is injected, and then removed from this processing amu O2 signals. ‘‘Loaded intensity’’ refers to the plasma on recipe are shown in Fig. 16. The top panel displays the el- minus plasma off intensities for the silicon loaded chamber lipsometric variable delta versus time, while the bottom minus the CFx loaded chamber. If we focus on the fluorine panel shows the evolution of psi. At event marker 共a兲 is the rich regime of the CF4 /O2 discharge at a ratio of ignition and stabilization of the CF4 /O2 discharge, after O2 :CF4⫽0.15, we find a strong correlation between the which a thick reaction layer is formed. At event marker 共b兲, ‘‘loaded’’ SiF3 signal of Fig. 15 and the loaded NO2 and O2 50 sccm of NO is injected directly into the processing cham- signals. This indicates that both of these species are products ber. This results in the removal of the overlayer back to the of the etching reaction. We attribute the initial negative value bare silicon point. A similar effect has been seen with the in the O2 signal to the oxidation of the Si surface. Such a injection of nitrogen directly into the dicharge.1 When the correlation suggests that the etch rate enhancement with NO NO is removed, the modified surface layer again begins to addition is due to the removal of oxygen from the Si surface, grow. This process is repeated and the removal of the over- freeing bonds for fluorine attack. Such a mechanism is con- layer is again observed. The mirror image of psi in the bot- sistent with the data presented for CF4 /O2 /N2 processing1 tom panel for the first growth/removal stage is indicative of and also the NF3 /O2 processing presented earlier. A return to the growth and removal of a smooth film. As the second the XPS spectra of Fig. 8共b兲 shows the N 1s intensity maxi- growth and removal stage proceeds, the psi evolution be- mum coincides with that of the etch rate. This supports the comes less pronounced, and in some instances follows the FIG. 16. Real-time ellipsometry data showing the surface effect of down- FIG. 14. 共a兲 Poly-Si etch rates and 共b兲 surface film thicknesses vs down- stream NO injection. 共a兲 CF4 /O2 discharge only 共b兲 with downstream NO stream NO injection in the effluent of oxygen rich CF4 /O2 plasmas. injection 共c兲 discharge extinguished. J. Vac. Sci. Technol. A, Vol. 17, No. 5, Sep/Oct 1999
2437 Matsuo et al.: Silicon etching in NF3 /O2 remote microwave plasmas 2437 delta evolution in trend. This is due to the formation of a In the presence of fluorine, Si etch rates are higher in the rough surface. Such roughness is consistent with the removal presence of NO. 共4兲 Reaction layers composed of SiOx Fy of the overlayer in the first NO treatment. Although this first species are thinned in the presence of NO. 共5兲 NO2 and O2 injection was efficient at oxygen removal, some islands re- are observed as a product of the Si etching reaction, while mained, and these islands served as micro-etch masks lead- NO is lost in the surface region. ing to roughening. At event marker 共c兲, the plasma is extin- guished, and silicon overlayer quickly increases in thickness; ACKNOWLEDGMENTS without the source of fluorine, the etching component of the The authors would like to thank M. G. Blain for helpful steady state film dynamics dies. NO serves to enhance the Si discussions. They also thank R. E. Ellefson and L. C. Frees reaction component of the steady state film dynamics by for their mass spectrometry assistance. opening bonding sites through the removal of oxygen. In 1 P. J. Matsuo, B. E. E. Kastenmeier, J. J. Beulens, and G. S. Oehrlein, J. chemistries with abundant fluorine, an etch rate enhancement Vac. Sci. Technol. A 15, 1801 共1997兲. 2 will occur. B. E. E. Kastenmeier, P. J. Matsuo, J. J. Beulens, and G. S. Oehrlein, J. Vac. Sci. Technol. A 14, 2802 共1996兲. 3 K. E. Greenberg and J. T. Verdeyen, J. Appl. Phys. 57, 1596 共1985兲. IV. CONCLUSIONS 4 G. Bruno, P. Capezzuto, G. Cicala, and P. Manodoro, J. Vac. Sci. Tech- We have discussed the role of nitrogen in Si etching, nol. A 12, 690 共1994兲. 5 B. A. Huling, G. M. Engle, C. A. Schneider, R. J. Gibson, and R. G. when using chemistries including fluorine and oxygen. Ridgeway, Solid State Technol. 37, 73 共1994兲. NO, and possibly one of the nitrosyl fluoride isomers, 6 C. J. Mogab, A. C. Adams, and D. L. Flamm, J. Appl. Phys. 49, 3796 serves to enhance the Si reaction component of the steady 共1978兲. state film dynamics by opening bonding sites through the 7 L. M. Lowenstein, J. Vac. Sci. Technol. A 6, 1984 共1988兲. 8 K. J. Nordheden and J. T. Verdeyen, J. Electrochem. Soc. 133, 2168 removal of oxygen. If the gas phase is rich in fluorine, the 共1986兲. etch rate will increase. If the gas phase is oxygen rich, it will 9 A. Nagata, H. Ichihashi, Y. Kusunoki, and Y. Horiike, Jpn. J. Appl. not. An exception to this was observed at high NO densities, Phys., Part 1 28, 2368 共1989兲. 10 where both the reaction layer thickness and the etch rate B. E. E. Kastenmeier, P. J. Matsuo, G. S. Oehrlein, and J. G. Langan, J. Vac. Sci. Technol. A 16, 2047 共1998兲. increased. In fact, fast etching was observed through 20-nm- 11 M. G. Blain, T. L. Meisenheimer, and J. E. Stevens, J. Vac. Sci. Technol. thick overlayers. A 14, 2151 共1996兲. In particular, NO reacts with the oxygen on the surface to 12 M. G. Blain, J. Vac. Sci. Technol. A 17, 665 共1999兲. 13 form NO2 or O2 which escape. In the latter reaction, nitrogen M. G. Blain, R. L. Jarecki, and R. J. Simonson, J. Vac. Sci. Technol. A 16, 2115 共1998兲. can be left behind on the silicon surface. This mechanism has 14 I. L. Alberts, N. C. Handy, and P. Palmiere, Chem. Phys. Lett. 129, 176 appeared in both CF4 /O2 /N2 chemistries and NF3 /O2 chem- 共1986兲. istries. This notion is supported by several observations: 共1兲 15 D. Briggs and M. P. Seah, Practical Surface Analysis 共Wiley, New York, 1990兲, p. 202. the fluorine densities alone cannot explain the etching behav- 16 G. S. Oehrlein, J. Vac. Sci. Technol. A 11, 34 共1993兲. ior 共i.e., nonfluorine limited etching兲. 共2兲 There is a correla- 17 B. E. E. Kastenmeier, P. J. Matsuo, G. S. Oehrlein, R. E. Ellefson, and L. tion of etch rates and NO densities for both chemistries. 共3兲 C. Frees, Appl. Phys. Lett. 共submitted兲. JVST A - Vacuum, Surfaces, and Films
You can also read