Surface etching mechanism of silicon nitride in fluorine and nitric oxide containing plasmas B. E. E. Kastenmeier IBM Semiconductor Research and Development Center, 2070 Route 52, Hopewell Junction, New York 12533 P. J. Matsuo Department of Physics, The University at Albany, State University of New York, 1400 Washington Avenue, Albany, New York 12222 G. S. Oehrlein Department of Materials and Nuclear Engineering and Institute for Plasma Research, University of Maryland, College Park, Maryland 20742-2115 R. E. Ellefson and L. C. Frees Leybold Inficon Inc., East Syracuse, New York 13057-9714 共Received 27 August 1999; accepted 2 October 2000兲 The etch rate of silicon nitride (Si3N4 ) in the afterglow of fluorine-containing plasmas is strongly enhanced when both nitrogen and oxygen are added to the remote discharge. This effect is attributed to the formation of nitric oxide 共NO兲, which we identify as a highly reactive precursor for the etching of Si3N4 . The Si3N4 etch rate, surface oxidation, and the depletion of the surface of N atoms show a linear dependence on the NO density. In order to determine the products of the NO reaction at the Si3N4 surface, mass spectrometry was performed in immediate proximity to the surface with a specially designed movable sampling orifice. Both SiF4 and N2 are identified as primary etch products, but a smaller amount of N2O was also detected. Based on our results, we suggest that NO enhances the removal of N from the Si3N4 surface by the formation of gaseous N2 , and leaving behind an O atom, while the overall surface oxidation remains very low, and the reactive layers are very thin. This modified surface reacts more readily with F atoms than the Si3N4 surface. © 2001 American Vacuum Society. 关DOI: 10.1116/1.1329118兴 I. INTRODUCTION In remote plasma chemical dry etching 共CDE兲, the synergistic effect of direct ion bombardment on the removal rate of materials is not available. Instead, all etching reactions are purely chemical. The absence of ion bombardment on the surfaces allows for high etch rate ratios of different materials and processes free of ion- or charge-induced damages. However, in many cases removal rates are suppressed due to the absence of sufficient activation energy. Small changes in the feed-gas composition of the remote discharge can significantly alter the rate of removal of certain materials due to the formation of reactive intermediates. The new species can open up additional pathways for etching reactions. In this paper, we describe the effect of nitric oxide 共NO兲 addition to a F and O containing gas phase on the etching of silicon nitride (Si3N4 ). The changes in this system upon injection of NO are significant enough to allow experimental characterization of the new reaction channels. We first describe the experimental setup. Subsequently the results allowing the characterization of the Si3N4 etch mechanism by NO and F are presented. Etch rates of Si3N4 and their relation to the NO partial pressure are shown. The effects of NO on the chemical composition of the reactive layer are demonstrated by x-ray photoelectron spectroscopy measurements. The main products of the chemical reaction of NO with Si3N4 are determined by mass spectrometry, al25 J. Vac. Sci. Technol. A 19„1…, JanÕFeb 2001 lowing the determination of the mechanism by which NO enhances the etching of Si3N4 . II. EXPERIMENTAL SETUP The remote plasma chemical dry etching 共CDE兲 system used in this investigation is shown in Fig. 1 共see Refs. 1–3 for a more detailed description兲. A 2.45 GHz microwave discharge is ignited in a sapphire applicator. The species effluent from the plasma travel through tubing of variable length and lining material into the reaction chamber. For some experiments, nitric oxide 共NO兲 is injected directly into the reaction chamber downstream from the discharge. The sample is placed on the center of an electrostatic chuck. Etch rates are measured in situ by ellipsometry. Two orifice designs were employed for the mass spectrometry measurements reported here. For gas phase sampling, an orifice with a diameter of 70 m is placed at the chamber wall. Additionally, for sampling the gas phase in immediate proximity to the sample, a cone-shaped 70 m sampling orifice is welded onto a 0.25-in.-o.d. stainless steel tube. The cone shape allows for uninhibited flow of species to and from the surface. The mass spectrometer is mounted on top of the reaction chamber by means of a bellows and a drive gear. This allows the mass spectrometer and orifice setup to move vertically within a distance of 5 cm from the sample surface. This setup allows one to sample mass spectrometric data in 0734-2101Õ2001Õ19„1…Õ25Õ6Õ$18.00 ©2001 American Vacuum Society 25 26 Kastenmeier et al.: Surface etching mechanism FIG. 1. Schematic of the chemical downstream etcher used in this investigation. The gases are fed into the sapphire applicator, where a microwave discharge is ignited. The species effluent from the plasma travel through tubing of variable length and lining material to the reaction chamber. The sample is placed on the center of an electrostatic chuck. A quadrupole mass spectrometer is mounted on the chamber on top of the sample, and monochromatic ellipsometry is used to determine etch rates. NO is fed directly into the chamber downstream from the discharge. two different modes: 共1兲 At constant sample-orifice distance, a multilayer stack is etched. The intensities of etch products are expected to show changes at transitions between materials. 共2兲 During the etching of one material, the sample-orifice distance is varied, thus scanning the concentration profile of products. Experiments were performed with Si3N4 /SiO2 /c-Si multilayer samples. The mass spectrometry experiments designed to sample the etch products are repeated with a sample covered with a Teflon™-like fluorocarbon film for comparison. This film was grown using CHF3 in an inductively coupled high-density plasma reactor.4 Teflon-like films are not attacked by the chemistries at the wafer temperatures employed here, therefore the amount of etch products desorbing from that surface is small compared to a Si3N4 surface. All samples were placed on a silicon wafer covered with the same inert fluorocarbon film in order to minimize the generation of etch products from surfaces other than that of the sample. 26 FIG. 2. Normalized Si3N4 etch rates and NO partial pressures for two distinctively different plasma chemistries. The data shown in the top panel are taken in the afterglow of a CF4 /O2 discharge, with either 0 or 20 sccm of N2 added to the discharge. The bottom panel shows the same set of data, measured in the afterglow of a NF3 /O2 discharge. For both cases, the Si3N4 etch rates and the NO partial pressure show the same trend as a function of the O2 /CF4 ratio. Data are taken from Refs. 1 and 2. bottom panel shows the normalized etch rates of Si3N4 in the afterglow of a NF3 /O2 discharge and the NO partial pressure in the reaction chamber. Again, the etch rates and the NO partial pressure follow the same trend. The close correlation between Si3N4 etch rates and NO partial pressure allowed this group to identify NO as a chemically active species in Si3N4 etching,7 which was later verified by Blain et al.8 The chemical composition of the reactive layer during Si3N4 etching is changed by the presence of NO. This is shown in Fig. 3, in which photoemission spectra from Si(2p), O(1s), and N(1s) electrons are plotted for experi- III. RESULTS The starting point of this investigation are Si3N4 etch rates in the afterglow of various plasma chemistries. These etch rates are shown for CF4 /O2 and NF3 /O2 discharges in Fig. 2, normalized to their respective maximum values 共data taken from Refs. 1 and 2兲. The CF4 /O2 etch rates 共top panel兲 show a significant enhancement by a factor of 7 or more when 20 sccm of N2 are added upstream to the discharge as compared to the case when only CF4 and O2 are used. The etch rates of Si and SiO2 are not significantly changed by the upstream addition of N2 . 5,6 Also included in the top panel is the normalized partial pressure of nitric oxide 共NO兲 in the reaction chamber for the case of 20 sccm of N2 added upstream, as determined by mass spectrometry. Both the NO partial pressure and the Si3N4 etch rates follow the same trend. The J. Vac. Sci. Technol. A, Vol. 19, No. 1, JanÕFeb 2001 FIG. 3. Photoemission spectra of Si(2p), O(1s), and N(1s) electrons. The spectra were taken from Si3N4 samples processed with CF4 /O2 400/20 sccm and NO addition of 0 sccm 共top row兲 and 10 sccm 共bottom row兲. The addition of NO enhances the surface oxidation and aides the removal of N atoms from the surface. 27 Kastenmeier et al.: Surface etching mechanism 27 Two reactions of the NO molecule on the Si3N4 surface have been proposed to explain the observed etch rate enhancement.1,8 The first produces N2 : Nsurface⫹NOgas→N2⫹Osurface . 共1兲 The O atom left on the surface by this reaction must desorb as the etching progresses, most likely as O2 . The net reaction then becomes 2Nsurface⫹2NOgas⫹→2N2gas⫹O2gas . The direct formation of O2 from two adsorbed NO molecules without the O atoms adsorbing on the surface is second order in both the NO density and the surface site density and is therefore much less likely to occur. The second equation produces N2O: Nsurface⫹NOgas→N2Ogas . FIG. 4. Normalized Si3N4 etch rate and oxidation during NF3 /O2 etching as a function of normalized NO density. For these experiments, the flow of NF3 was kept constant, and O2 was added to the discharge 共see Fig. 2兲. Thus, the NO density, determined in the reaction chamber with a quadrupole mass spectrometer, varied in a nonobvious fashion 共Ref. 2兲. The surface oxidation is determined using x-ray photoelectron spectroscopy. Both etch rate and oxidation depend on the NO density in the same fashion. ments where samples were processed with a CF4 /O2 remote discharge. The effects of NO were investigated by adding a small amount 共10 sccm兲 of NO gas directly into the reaction chamber downstream. After processing, the samples were transferred in vacuum to the surface analysis chamber. The spectra were taken with an emission angle of 15° with respect to the surface to enhance surface sensitivity. The top row shows spectra taken from a sample that was processed with the remote CF4 /O2 discharge only. In the bottom row, spectra from a sample processed with the remote discharge and 10 sccm of NO are shown. The Si(2p) peak, shown in the left column, shows a shoulder at high binding energies when 10 sccm of NO is injected downstream 共bottom panel兲. This corresponds to an increased number of Si atoms bonded to O atoms during etching as compared to the case where no NO is used 共top panel兲. Correspondingly, the O(1s) signal, shown in the center column, is significantly higher for the case in which NO is used. In the right column of Fig. 3, the N(1s) peak is shown. The peak intensity is reduced significantly for the case in which NO is added downstream. The correlation between the NO partial pressure in the reaction chamber, the Si3N4 etch rate, and the surface oxidation is summarized in Fig. 4. The data are acquired under NF3 /O2 processing conditions. The etch rate and NO partial pressure data are the same as in the bottom panel of Fig. 2, normalized to their respective maximum values. Samples for x-ray photoelectron spectroscopy measurements were prepared under the same experimental conditions and then transferred to the surface analysis chamber. The O(1s)/Si(2p) photoemission intensity ratio, normalized to its maximum value, is used as a measure for the surface oxidation. Within experimental error, both the Si3N4 etch rate and the degree of surface oxidation as expressed by the O/Si ratio are proportional to the partial pressure of NO in the reaction chamber. JVST A - Vacuum, Surfaces, and Films 共2兲 Although the above-mentioned results and those shown in Fig. 4 favor reaction 共1兲, contributions from Eq. 共2兲 cannot be ruled out based on these data. In fact, in a recent publication,9 Blain monitors the N2O partial pressure during Si3N4 etching in the afterglow of a NF3 /Ar discharge. His data show a decline of the N2O partial pressure after the Si3N4 is removed. The N2 produced by reaction 共1兲 on the Si3N4 surface, however, cannot be distinguished from the background N2 produced by the NF3 /Ar discharge employed in his work. In order to clarify the relative importance of reactions 共1兲 and 共2兲, mass spectrometry experiments were carried out with the orifice placed in proximity to the sample surface. The partial pressure of etch products is higher close to the sample surface as compared to a recessed wall position. The experiments with this orifice setup were conducted in two modes. In the first mode, the sample–orifice distance was kept constant, and a Si3N4 /SiO2 stack was etched. Changes of the signal intensities can be attributed to the amount of species produced during the etching of either Si3N4 or SiO2 . In the second mode, the sample–orifice distance was increased in steps. The intensity of product species is expected to decrease with increasing distance in this case. For the experiments reported here, sulfurhexafluoride (SF6 ) with a small amount of O2 was injected into the sapphire microwave applicator. This gas mixture was used, because it does not yield discharge products with mass m/e ⫽28. Other gases like CF4 /O2 and NF3 /O2 produce significant amounts of CO and N2 , respectively. The possible contribution of Si as a cracking product of SiF4 to the mass peak at m/e⫽28 could be ruled out based on specifically designed experiments in which the SiF4 partial pressure showed significant changes, but the mass peak at m/e⫽28 remained constant at background levels. For these experiments, polycrystalline Si on SiO2 was etched in SF6/02 discharges. The intensity of the sif3⫹ peak declined significantly as the poly-Si was removed and SiO2 etching started at a much lower rate. The intensity of the peak at m/e⫽28, however, remained unchanged. By using SF6 , it is thus possible to detect N2 as an etch product against a relatively small background at 28 u. For the SF6 /O2 system, the downstream injection of NO has the same etch rate enhancing effect as for CF4 /O2 gases. This is demonstrated in Fig. 5, where Si3N4 28 Kastenmeier et al.: Surface etching mechanism 28 FIG. 5. Etch rates of Si3N4 as a function of downstream NO addition. A similar behavior is observed for SF6 /O2 and CF4 /O2 as discharge gases. Note that the absolute etch rate values differ significantly 共see the text兲. etch rates are shown for both gas mixtures as a function of downstream NO addition. The etch rates are normalized to their respective maximum values. The normalized etch rates show identical trends as a function of NO addition. The absolute etch rates for SF6 /O2 are greater by a factor of 5 than those for CF4 /O2 , however. Typical etch rate values are 20 nm/min for CF4 /O2 and 100 nm/min for SF6 /O2 at 10 sccm of NO. For all experiments reported here, the flow of SF6 FIG. 6. Mass spectrometric signal intensities during the etching of a Si3N4 /SiO2 stack 共left column兲 and a Teflon-like fluorocarbon film 共right column兲. The numbers in the panels indicate the magnitude of the ion current detected in the mass spectrometer. J. Vac. Sci. Technol. A, Vol. 19, No. 1, JanÕFeb 2001 ⫹ ⫹ FIG. 7. Mass spectrometric intensities for SiF⫹ 3 , N2 , and N2O during the etching of Si3N4 and SiO2 in the afterglow of a SF6 /O2 discharge and NO gas injected directly into the reaction chamber. At point A, NO gas is injected downstream, increasing the Si3N4 etch rate and the intensities of the product species in the mass spectrometer. At points B – F, the sample– orifice distance was increased in steps from 0.03 to 4 cm. The decrease in the mass spectrometer intensities indicate that these species are produced by the etching of the Si3N4 . At point G, the Si3N4 film is completely removed, and SiO2 etching starts. At point H, the sample–orifice distance is again reduced to 0.03 cm, causing the signal intensity of etch products to increase. and O2 was kept constant at 300 and 50 sccm, respectively. The microwave power was 700 W and the pressure 500 mTorr. Figure 6 shows the mass spectrometric signal intensities of etch products during SF6 /O2 etching. The orifice was placed at a fixed position 0.3 mm above the sample surface, corresponding to approximately two mean free path lengths at the operating pressure of 500 mTorr. In the left column the intensities as recorded during the etching of a Si3N4 /SiO2 stack are plotted. All curves of that column are obtained during the same run, and the material that is being etched at a certain time is indicated by the arrows at the top of the column. Initially, the Si3N4 is being etched and only SF6 and O2 is injected. At the time labeled ‘‘a,’’ 20 sccm of NO is injected directly into the reaction chamber. This increases the Si3N4 etch rate of by a factor of 6 共see Fig. 5兲 and thus the signal intensity of reaction products increases. At point ‘‘b,’’ the Si3N4 is removed completely and etching of the underlying SiO2 starts. No other parameters are changed. At this point, the signal intensities of Si3N4 etch products are expected to decline. At point ‘‘c,’’ the discharge is switched off and the flow of gases is stopped. The right column shows the intensities of the same peaks while a fluorocarbon film is etched. The ordinates have the same scale as on the left-hand side for comparison. At point ‘‘d,’’ NO is added downstream of the SF6 /O2 discharge, and the flow of NO is stopped at ‘‘e.’’ In the top row, the intensity of the SiF⫹ 3 peak at 85 u is 29 Kastenmeier et al.: Surface etching mechanism 29 TABLE I. The various heats of formation occurring in the two processes. The values are taken from Ref 10. Reaction Nsurface⫹NOgas→N2⫹Osurface Nsurface⫹NOgas→N2Ogas Bond energy 共eV兲 Formation of N–N in N2 : ⫺9.79 Formation of Osurface –Si: ⫺8.27 Breaking of N–O: 6.53 Breaking of Nsurface –Si: 4.54 Formation of N–N in N2O: ⫺4.50a Formation of Osurface –Si: ⫺8.27 Weakening of N–O: 4.80b Breaking of Nsurface –Si: 4.54 Sum ⫺6.99 ⫺3.43 a Values taken from Ref. 11. Difference between the bond energies of the N–O bond in NO 共6.53 eV兲 and the N–O bond in N2O 共1.73 eV兲. b plotted. The intensity follows the increase of the Si3N4 etch rate at point ‘‘a,’’ and then declines at point ‘‘b,’’ since SiO2 is removed at a rate of only approximately 10 nm/min at these conditions. For the Teflon-like sample, the SiF⫹ 3 peak remains at much lower levels. The background signal is probably due to etching of unprotected areas of the Si carrier wafer and the quartz windows. The next two rows show the intensities of the N⫹ 2 and N2O⫹ products. The N2 intensity increases significantly and abruptly at point ‘‘a.’’ When the Si3N4 is completely removed at ‘‘b,’’ the signal intensity decreases. The changes of the N2O⫹ signal are much weaker by comparison, but there is a small decrease upon completion of the Si3N4 etching, indicative of the formation of N2O. The changes during etching of the fluorocarbon film are more subtle for both species. Finally, in the bottom row the intensity of the NF⫹ 2 peak at 52 u is plotted. This ion is a cracking product of NF3 in the ionization region of the mass spectrometer. The peak intensity shows the typical behavior of a Si3N4 etch product, while no changes are observed when no Si3N4 is present 共see the right column兲. The overall intensity of this peak, however, is smaller than that of other etch products by approximately two orders of magnitude. In Fig. 7, peak intensity profiles of etch products are shown for an experiment in which the sample–orifice distance was increased in steps while the Si3N4 surface was etched in the afterglow of a SF6 /O2 discharge. The concentration profile as a function of distance from the sample surface allows one to distinguish the role of a given species in the etch mechanism. An etchant species is expected to increase in intensity as the sampling orifice is removed from the surface, which acts as a drain for that particular species. A product species is expected to decrease in intensity. At a location very close to the sample, the flux of product species is spread out over an area equal to the sample area. At a recessed orifice position, the same flux of species is spread out over a larger area, resulting in a lower partial pressure. At point A, 20 sccm of NO is injected directly into the reaction chamber, leading to an increased Si3N4 etch rate. Correspondingly, the SiF⫹ 3 signal increases by a factor of 8. Subsequently, the orifice is moved in steps from its starting position at 0.03 to 4 cm above the sample surface, indicated ⫹ by letters B–F. The decline of the SiF⫹ 3 and N2 signal indicates that SiF4 and N2 are produced at the Si3N4 surface. The N2O⫹ signal in the bottom panel of Fig. 7 shows only a slight increase upon NO addition to the reaction chamber, JVST A - Vacuum, Surfaces, and Films and a smaller relative change as the sample–orifice distance is varied. IV. CONCLUSIONS It can be concluded from these data that N2 , N2O, and NF3 are reaction products of the NO reaction on Si3N4 surfaces in the presence of atomic F. The signal intensities were calibrated to partial pressure values by using pure N2 and N2O gases. At point ‘‘b’’ in Fig. 6, the N2 signal declines by 15%, corresponding to a partial pressure decline of 60 ⫻10⫺6 Torr. The N2O signal declines by 5% or 13 ⫻10⫺6 Torr. The data indicate that reaction 共1兲 is the predominant pathway for N removal from the surface. From an energetic point of view, the N2 producing reaction 共1兲 is preferred over reaction 共2兲. In Table I, the bond energies relevant for the two reaction pathways are listed, together with the total energy released by the reactions. Both are exothermic, but the energy released by the N2 producing reaction is almost twice as high as that of the N2O producing reaction. The difference stems, for the biggest part, from the formation of the extremely stable N2 molecule. We have demonstrated how the introduction of a new chemical reaction pathway can increase the overall reaction rate. The chemically active species was obtained in a nonobvious fashion, i.e., not through electron impact of the parent gases, but through chemical reactions in the gas phase 共for CF4 /O2 /N2 and NF3 /O2 processes兲 or through direct downstream injection. This approach to increase the reaction rate can be of importance in other systems in which the amount of activation energy supplied to the surface is limited. ACKNOWLEDGMENTS This work was supported by Sandia National Laboratories, Leybold Inficon, and the New York State Science and Technology Foundation. We also thank Matthew Blain and John Langan for stimulating discussions. 1 B. E. E. Kastenmeier, P. J. Matsuo, J. J. Beulens, and G. S. Oehrlein, J. Vac. Sci. Technol. A 14, 2802 共1996兲. 2 B. E. E. Kastenmeier, P. J. Matsuo, G. S. Oehrlein, and J. G. Langan, J. Vac. Sci. Technol. A 16, 2047 共1998兲. 3 P. J. Matsuo, B. E. E. Kastenmeier, J. J. Beulens, and G. S. Oehrlein, J. Vac. Sci. Technol. A 15, 1801 共1997兲. 30 Kastenmeier et al.: Surface etching mechanism 4 N. R. Rueger, J. J. Beulens, M. Schaepkens, M. F. Doemling, J. M. Mirza, T. E. F. M. Standaert, and G. S. Oehrlein, J. Vac. Sci. Technol. A 15, 1881 共1997兲. 5 N. Hayasaka, H. Okano, and Y. Horiike, Solid State Technol. 31, 127 共1988兲. 6 J. J. Beulens, B. E. E. Kastenmeier, P. J. Matsuo, and G. S. Oehrlein, Appl. Phys. Lett. 22, 2634 共1995兲. 7 S. Lakeman, Semicond. Int. 18, 127 共1995兲. J. Vac. Sci. Technol. A, Vol. 19, No. 1, JanÕFeb 2001 30 8 M. G. Blain, T. L. Meisenheimer, and J. E. Stevens, J. Vac. Sci. Technol. A 14, 2151 共1996兲. 9 M. G. Blain, J. Vac. Sci. Technol. A 17, 665 共1999兲. 10 CRC Handbook of Chemistry and Physics, 67th ed., edited by R. C. Weast, M. J. Astle, and W. H. Beyer 共Chemical Rubber, Boca Raton, FL, 1986兲. 11 R. K. Curran and R. E. Fox, J. Chem. Phys. 34, 1590 共1961兲.