1150
Novel Multilayered Ti/TiN Diffusion Barrier for Al Metallization
WEN-FA WU,1,4KOU-CHIANG TSAI,1,2CHUEN-GUANG CHAO,2JEN-CHUNG CHEN,2and KENG-LIANG OU 3
1.—National Nano Device Laboratories, Hsinchu 300, Taiwan, Republic of China. 2.—Department of Materials Science and Engineering, National Chiao Tung University, Hsinchu 300, Taiwan, Republic of China. 3.—Graduate Institute of Oral Sciences, Taipei Medical University, Taipei, Taiwan, Republic of China. 4.—E-mail: [email protected]
A novel, multilayered Ti/TiN diffusion barrier is proposed and successfully applied for Al metallization. The multilayered Ti/TiN structure is effective in enhancing the barrier properties since the very thin Ti layer inserted into ti-tanium nitride (TiN) barrier can cause disruption of the TiN columnar growth and reduction of open grain boundaries resulting in retarded interdiffusion be-tween metal and silicon. Multilayered Ti/TiN films are deposited sequentially by sputtering without breaking vacuum. It is found that TiN grain boundaries are discontinuous when a Ti layer is inserted into TiN. Multilayered Ti/TiN has a better barrier performance than single-layer TiN in Al metallization. However, the barrier performance is related to the number and thickness of the inserted Ti layers, because increasing titanium will enhance chemical reactions between Al and barrier layers, and produce more titanium-aluminum compounds. The total thickness of introduced Ti layers should be reduced to improve barrier performance.
Key words: Diffusion barrier, titanium nitride, multilayer, aluminum
(Received November 2, 2004; accepted February 28, 2005) INTRODUCTION
An interconnection system with a barrier metal layer is indispensable in submicron ultra-large-scale integrated devices to realize high reliability such as resistance against electromigration, hillocks, stress-induced voids, and Si precipitation at silicon/ metal interfaces.1,2Many diffusion-barrier materials
have been studied for the Al-Si contact system. Reac-tively sputtered titanium nitride (TiN) film has been widely used as a diffusion barrier layer between the aluminum and the silicon substrate due to its high thermal and chemical stability, low electrical resis-tivity, and excellent mechanical properties.3–5 The
stoichiometric TiN has a NaCl-type face-centered cubic structure with a lattice constant of 4.24 Å. The phase is stable over a broad composition range concerning the nitrogen concentration.6 In general,
resistivity of TiN film depends on the deposition conditions, and minimum resistivity has been found at a composition corresponding to stoichiometric TiN.7
Devices with Al interconnects require some ther-mal processes to improve interface property and achieve low contact resistance. It is reported that sputtered TiN film has a columnar grain structure with both inter- and intracolumnar voids.8–10 The dominant failure of TiN diffusion barrier is attrib-uted to diffusion via fast diffusion paths in columnar grains.11,12 When TiN film is used as the diffusion
barrier in Al metallization, Al and Si interdiffuse through the grain boundaries of the TiN film during annealing at elevated temperature, resulting in degradation of the electrical characteristics and de-vice failure. Barrier performance of TiN layer can be improved by in-situ plasma treatment and stuffing grain boundaries with a thin Al interlayer.1,13,14
The purpose of this work is to demonstrate a novel, multilayered Ti/TiN diffusion barrier for Al metal-lization. Al-contacted systems with high thermal stability are obtained. It is found that fast diffusion paths in columnar TiN grains are destroyed when a thin Ti layer is inserted into TiN. Additionally, effects of number, thickness, and distribution of inserted Ti layer on barrier properties of multilayered Ti/TiN films are investigated.
Novel Multilayered Ti/TiN Diffusion Barrier for Al Metallization 1151 EXPERIMENTAL PROCEDURE
The structure of the Al/barrier/Ti/n-p junction diode was used to investigate the barrier capabilities of multilayered Ti/TiN films in this work. The starting materials were p-type (100)-oriented silicon wafers with resistivity of 15–25 Ω cm. The wafers were ad-ministered a local oxidation of silicon process to define active regions after RCA cleaning. The n-p junctions were formed by As implantation at 60 keV with a dose of 5 1015cm2and subsequent rapid thermal
annealing in N2ambient at 1,050°C for 30 sec. The Al
alloy (Al-Si-Cu), Ti, and TiN films were deposited by sputtering in a multichamber cluster system without breaking vacuum. The wafers were dipped in a dilute HF solution (HF: H2O 1:50) to remove native oxide
prior to loading into the system. The Ti layer, 40-nm thick, was deposited onto the Si substrate first and employed as a conventional contact metal layer to im-prove contact property. Reactively sputtered TiN film, 100-nm thick, was used as the standard (STD) barrier layer. Multilayered Ti/TiN films were prepared by inserting thin Ti layers into the TiN barrier. Various multilayered Ti/TiN films, as summarized in Table I, were employed to investigate the effects of number, thickness, and distribution of the inserted Ti layer on barrier properties. The total thickness of multilayered Ti/TiN barrier layer was 100 nm. Al-Si-Cu film was deposited on top of the barrier layer at 15 kW and 200°C. All samples were alloyed in forming gas ambi-ent at 400°C for 30 min to improve contact properties. The samples were further subjected to a cumulative furnace annealing in forming gas ambient at 500°C for 30 min to investigate thermal stability and barrier performance.
The microstructure and grain size of the film were examined using transmission electron microscopy (TEM). Structure and crystalline orientation of the as-deposited and annealed samples were analyzed by an x-ray diffractometer with Cu Kαradiation op-erated at 50 kV and 250 mA. Surface roughness and morphology of the film were observed using a Digital Instruments Nanoscope II model atomic force micro-scope (Vecco, Cleveland, OH) with a 0.5 Hz scanning speed in air ambient. The surface morphology was studied with a field emission scanning electron mi-croscope. Compositional depth profiles analyzed by secondary ion mass spectrometry (SIMS) were used
to investigate the Al diffusion after annealing. A four-point probe system was employed to measure sheet resistance. Contact resistance was analyzed using the four-terminal Kelvin structure. Leakage currents of diodes were measured at a reverse bias of 5V by a HP 4156 semiconductor parameter analyzer (Hewlett-Packard, Palo Alto, CA).
RESULTS AND DISCUSSION
Figure 1 shows cross-sectional TEM micrographs of the typical TiN and multilayered Ti/TiN films. Typical columnar grain structure is observed for con-ventionally sputtered TiN barrier layer, as shown in Fig. 1a. A sawtooth profile is exhibited in a projected two-dimensional view as is expected for films with
Table I. Contact Systems with Various Multilayered Ti/TiN Barriers Used in the Study
Thickness of Ti Barrier Type Thickness of Ti/TiN (nm) Contact Layer (nm)
TiN-1 TiN (100) 40
TiN-2 TiN (47.5)/Ti (5)/TiN (47.5) 40
TiN-3 TiN (30)/Ti (5)/TiN (30)/Ti (5)/TiN (30) 40
TiN-4 TiN (21.3)/Ti (5)/TiN (21.2)/Ti (5)/TiN (21.3)/Ti (5)/TiN (21.2) 40
TiN-5 TiN (45)/Ti (10)/TiN (45) 40
TiN-6 TiN (42.5)/Ti (15)/TiN (42.5) 40
TiN-7 TiN (25)/Ti (5)/TiN (75) 40
TiN-8 TiN (75)/Ti (5)/TiN (25) 40
a
b
Fig. 1. Cross-sectional TEM images of (a) typical TiN film on the SiO2/Si substrate and (b) multilayered Ti/TiN (TiN-3) film on the Ti/Si
substrate. The thickness of typical TiN and TiN-3 barriers is 100 nm.
voided grain boundaries for which growth proceeds in a three-dimensional mode. The trace of the surface facets is along the [022] direction that is con-sistent with the facets being {111} planes. Discontin-uous columnar grains are successfully formed as a very thin Ti layer is inserted into the TiN barrier, as shown in Fig. 1b. The thickness of the Ti layer in-serted into TiN to form multilayered Ti/TiN films is 5 nm. The very thin Ti interlayer will cause disrup-tion of the TiN columnar growth. Figure 2 shows the statistical distributions of leakage currents of Al/ barrier/Ti/Si junction diodes. Al-contacted junction
diodes with multilayered Ti/TiN (TiN-2) barriers exhibit lower leakage currents than those with STD TiN (TiN-1) barriers after cumulative annealing at 500°C for 30 min, indicating that multilayered Ti/TiN structure has successfully impeded the inter-diffusion of Al and Si. The SIMS process is further employed to investigate the barrier capability of multilayered Ti/TiN film against diffusion. Figure 3 plots SIMS depth profiles of Al and Si for the Al/bar-rier/Si system after annealing at 500°C for 30 min. It is also found that Al is less diffused in the multi-layered Ti/TiN sample than that in the STD sample. Figure 4 schematically shows the possible brief reactions of the Al/Ti/Si, Al/TiN/Ti/Si, and Al/TiN/
Fig. 2. Histograms showing statistical distributions of reversed-biased leakage currents of Al/TiN (100 nm)/Ti/Si and Al/TiN (47.5 nm)/Ti (5 nm)/TiN (47.5 nm)/Ti/Si junction diodes after annealing at 500°C for 30 min.
Fig. 3. The SIMS depth profiles of Al/TiN-1/Si and Al/TiN-2/Si contact systems after annealing at 500°C for 30 min.
Fig. 4. Schematic illustrations of the possible brief reactions of the Al/Ti/Si, Al/TiN/Ti/Si, and Al/TiN/Ti/TiN/Ti/Si contact systems before and after high-temperature annealing.
Ti/TiN/Ti/Si contact systems before and after high-temperature annealing. It is reported that the Ti layer between Al and Si behaves as a sacrificial barrier because it reacts with Al to form TiAl3
com-pounds at temperatures above 400°C. The Ti is a good diffusion barrier as long as the Ti is not com-pletely consumed. Once the Ti has comcom-pletely re-acted to form TiAl3, its diffusion barrier properties
are lost. It is expected that TiAl3, TiSix, and Al
spik-ing will be formed for the Al/Ti/Si contact system after high-temperature annealing, as shown in Fig. 4a. The TiN is also an attractive barrier mater-ial because it is chemically and thermodynamically stable and behaves as a passive barrier. However, it is typical to form TiN films with columnar grains, as mentioned previously. It is expected that Al and Si would interdiffuse through the grain boundaries of the typical columnar TiN barrier during annealing, which results in spiking in the Si substrate, as shown in Fig. 4b. Discontinuous columnar grains are formed for the multilayered Ti/TiN barrier, as indicated in the TEM image in Fig. 1b. The diffusion has been relatively reduced, as shown in Fig. 3, because the thin Ti interlayer has successfully resulted in disruption of the TiN columnar growth and reduced open grain boundaries of TiN film. Fur-thermore, the inserting Ti interlayer creates a bind-ing spot for Al. Short-circuit diffusion paths through the TiN would be plugged by the Ti or Ti-Al com-pound. In order to further explore effects of very thin Ti interlayer on barrier capabilities and ther-mal stabilities, multilayered Ti/TiN films with vari-ous numbers, thicknesses, and distributions of Ti interlayer were prepared and used in this work. Effects of the Number of Ti Interlayer
Figure 5 shows the x-ray diffraction (XRD) spec-tra of as-deposited STD TiN and multilayered Ti/TiN barriers with various numbers of the Ti interlayer on Ti/Si substrates. All the XRD patterns show three peaks of TiN (111), Ti (10–11), and Ti (0002). Ti (10–11) and Ti (0002) peaks are believed to result from 40-nm Ti contact layer on the Si sub-strate because it is much thicker than inserting Ti interlayer. The high (0002) orientation of underlying Ti layer will enhance the development of TiN (111) preferred orientation since they have similar atomic arrangement patterns. The (111) planes of TiN are built of alternate Ti and N layers and are the most closely packed planes.15The relative intensity of the
TiN (111) peak in the TiN-2 sample is lower and the peak shape becomes broader than that in the TiN-1 (STD TiN) sample, indicating that grain size de-creases in the TiN-2 sample. Moreover, the relative intensity of the TiN (111) peak in the TiN-3 sample is much lower and the peak shape becomes broader than that in the TiN-2 sample. That is, the more the Ti-inserting layer, the lower the relative inten-sity of TiN (111) and the broader the peak shape. It is reported that the first 10-nm or 20-nm TiN films grow in an amorphous-like or nanograined
Novel Multilayered Ti/TiN Diffusion Barrier for Al Metallization 1153
structure followed by columnar grain growth. The columns in the TiN film increase in size with an increase in distance from the film/substrate inter-face.16The nanocrystalline barrier is more effective
than the polycrystalline barrier since the nano-crystalline film can slow interdiffusion.13,17–21 The
nanocrystal TiN grains in the multilayered Ti/TiN barrier will enhance the effectiveness of the barrier. Furthermore, the Ti interlayer will turn into an alu-minide, causing a step change in Al composition, and thus a lower concentration gradient for Al below the Ti layer. The step change will reduce the gradi-ent of Al diffusion flux. The diffusion barrier perfor-mance of TiN film is significantly improved by inserting a thin Ti interlayer into TiN, as shown in Figs. 2 and 3.
Figure 6 shows the variation percentage in sheet resistance of the Al/barrier/Ti/Si sample as a function of the annealing temperature for various multilay-ered Ti/TiN barrier layers. The samples are subjected to a cumulative furnace annealing in forming gas environment from 400°C to 550°C for 30 min. Sheet resistance increases with increasing annealing tem-perature. The increase in sheet resistance is attrib-uted to formation of the compounds. The increasing rate in sheet resistance of the STD-TiN (TiN-1) sample is higher than those in multilayered Ti/TiN (TiN-2, TiN-3, and TiN-4) samples, as indicated in Fig. 6. Low increasing rate for multilayered Ti/TiN
Fig. 5. XRD spectra of as-deposited TiN-based barriers on Ti/Si substrates.
samples is believed to result from discontinuous columnar grains in multilayered Ti/TiN films, which alleviate the interdiffusion and formation of the com-pound. It is also seen in Fig. 6 that the increasing rate in sheet resistance of the TiN-3 sample is higher than that in the TiN-2 sample, and the increasing rate in sheet resistance of the TiN-4 sample is higher than that in the TiN-3 sample. That is, variation per-centage in sheet resistance increases with increasing number of Ti interlayer. X-ray diffraction is employed to analyze annealed Al/barrier/Ti/Si samples to more distinctly display the differences among barrier lay-ers. It is found that the relative intensity of the TiAl3
phase increases with increasing Ti interlayer. As mentioned previously, the Ti layer between Al and Si behaves as a sacrificial barrier because Al and Ti start to react with each other at 400°C and above. The limited efficiency is probably caused by the fact that increasing Ti interlayer reduces total effective thickness of TiN in the multilayered Ti/TiN barrier. Barrier performance is determined by a competition among the disruption effect of columnar growth, gradient of Al diffusion flux, and total effective thick-ness of TiN. A thin Ti interlayer will result in discon-tinuity or shift in columnar TiN grain and create a binding spot for Al, thus alleviating interdiffusion and improving barrier performance. However, more Ti interlayers will reduce the total effective thickness of TiN, causing barrier performance to be degraded. Effects of the Thickness of Ti Interlayer
Effects of thickness of the inserting Ti interlayer on the barrier properties of multilayered Ti/TiN films are further investigated. Figure 7 shows the statistical distributions of reverse-biased leakage currents of the diodes with the TiN-1, TiN-2, TiN-5, and TiN-6 barriers. The distributions of biased leakage currents are measured at a reverse-biased voltage of 5 V after cumulative annealing at 400°C and 500°C for 30 min. Diodes with Al/bar-rier/Ti/Si exhibit low leakage currents of the order of 1011–108 amperes after cumulative annealing at 400°C and 500°C for 30 min. The diodes with TiN-2
barriers show lower leakage currents than those with STD TiN (TiN-1) barriers, indicating that mul-tilayered Ti/TiN structure has successfully impeded the interdiffusion of Al and Si. However, barrier per-formance is degraded as the thickness of inserting Ti interlayer increases. The diodes with TiN-5 and TiN-6 barriers show higher leakage currents than those with TiN-1 (STD TiN) and TiN-2 barriers. It is felt that increasing the thickness of the Ti interlayer reduces the total effective thickness of TiN in multi-layered Ti/TiN barrier since the total thickness of the TiN-based barrier layer is 100 nm. The multi-layered Ti/TiN structure is effective in enhancing the barrier performance and retarding Al/Si inter-diffusion, but the thick inserting Ti interlayer in multilayered Ti/TiN barrier is not suggested due to resulting degradation in barrier performance.
Figure 8 presents the contact resistance (Rc) of Al/barrier/Ti/Si contact systems, which include multi-layered Ti/TiN barriers with various thicknesses of inserting Ti interlayer. It can be seen that the contact resistance in the TiN-2 contact system is lower than that in the TiN-1 (STD TiN) contact system, and the contact resistance of the Al/barrier/Ti/Si contact system increases as the thickness of the Ti inter-layer increases. High contact resistance is found after
Fig. 6. Variation percentage in sheet resistance of Al/barrier/Ti/Si contact system as a function of annealing temperature.
Fig. 7. Histograms showing statistical distributions of reversed-biased leakage currents of Al/barrier/Ti/Si junction diodes after annealing at 500°C for 30 min.
Fig. 8. Contact resistance of Al/barrier/Ti/Si system as a function of thickness of inserting Ti interlayer.
Novel Multilayered Ti/TiN Diffusion Barrier for Al Metallization 1155 annealing at 500°C because high-temperature
an-nealing will enhance interdiffusion of Al and Si via microscopic defects (i.e., grain boundaries) and for-mation of the compounds. Low contact resistance in the TiN-2 contact system is attributed to the fact that the additional Ti interlayer has successfully alle-viated the interdiffusion of Al and Si. It is important to note that the formation of TiAl3is more likely to
occur in the Al-Ti system than in the Al-TiN system after thermal processing, since thermal stability be-tween Al and Ti layers is lower than that bebe-tween Al and TiN layers.22–24Increased thickness of the
insert-ing Ti interlayer will enhance the formation of TiAl3
compounds, resulting in increased contact resistance. Effects of the Distribution of Ti Interlayer
Figure 9 shows the statistical distributions of reverse-biased leakage currents of the diodes with 1, 2, 7, and 8 barriers. 2, TiN-7, and TiN-8 barriers have different inserting posi-tions of Ti interlayers in TiN, as summarized in Table I, to explore effects of distribution of inserting Ti interlayer on barrier properties of multilayered Ti/TiN barriers. Superiority of multilayered Ti/TiN as a barrier can be gauged from the leakage current of the diode. Low leakage currents are obtained for multilayered Ti/TiN films compared to the TiN-1 (STD-TiN) sample. Reduction in leakage current of the TiN-7 sample is apparent. Though the detailed mechanisms need to be further investigated, the gradient of Al diffusion flux is believed to be one of the key factors on TiN barrier performance. As men-tioned previously, all short-circuit diffusion paths through the TiN would be plugged by the Ti or Ti-Al compounds. If there is a fixed Al concentration on the substrate side of the Ti interlayer, it would be expected that pulling the Ti barrier further away from the Ti contact metal would reduce Al diffusion flux by reducing the gradient. Thin TiN film (25 nm) on the inserting Ti interlayer in the TiN-7 sample leads to better barrier capability due to the reduc-tion of Al diffusion flux.
CONCLUSIONS
A novel method to improve TiN-based diffusion barriers using multilayered Ti/TiN structure has been proposed. Multilayered Ti/TiN barriers can alleviate interdiffusion via fast diffusion paths in columnar TiN barriers, because the thin Ti inter-layer results in disruption or shift in TiN columnar grains and creates a binding spot for Al. Less inter-diffusion is observed in multilayered Ti/TiN barriers from SIMS analyses. The sheet resistance of the Al/TiN/Ti/TiN/Ti/Si contact system is relatively stable compared to the Al/TiN/Ti/Si (STD) contact system. The Al/barrier/Ti/Si junction diodes with multilayered Ti/TiN barriers show relatively low leakage currents compared to those with STD TiN barriers. However, an increase in the number and thickness of the inserting Ti interlayer will reduce efficiency of the TiN-based barrier because Ti is a sacrificial barrier. More titanium layers or increased thickness in the multilayered Ti/TiN film enhances the chemical reactions between Ti and Al. This re-duces the total effective thickness of TiN, causing a degradation of barrier performance. As the thick-ness of the inserting Ti interlayer in TiN increases from 5 nm to 15 nm, degradation of the multilayered Ti/TiN barrier and enhanced formation of TiAl3are
found. The total effective thickness of the inserting Ti interlayer in TiN should be reduced to improve barrier performance. Barrier performance of the multilayered Ti/TiN barrier is also affected by the distribution of the inserting Ti interlayer in TiN. Thin TiN film on the inserting Ti interlayer shows better barrier performance against Al diffusion due to reduced Al diffusion flux.
ACKNOWLEDGEMENTS
This work was financially supported by the Na-tional Science Council of the Republic of China (Contract No. NSC 93-2215-E-492-005) and sup-ported, in part, by the Ministry of Economic Affairs of the Republic of China (Contract No. 91-EC-17-A-08-S1-0003). Technical support from the National Nano Device Laboratories is greatly acknowledged.
REFERENCES
1. V. Fortin, S.C. Gujrathi, G. Gagnon, R. Gauvin, J.F. Currie, L. Ouellet, and Y. Tremblay, J. Vac. Sci. Technol. B 17, 423 (1999).
2. H. Xiao, Introduction to Semiconductor Manufacturing
Technology (Upper Saddle River, NJ: Prentice-Hall Inc.,
2001), pp. 452–457.
3. C.Y. Ting and M. Wittmer, Thin Solid Films 96, 327 (1982). 4. M. Wittmer, Appl. Phys. Lett. 37, 540 (1980).
5. C.Y. Ting, J. Vac. Sci. Technol. 21, 14 (1982).
6. B. Pécz, N. Frangis, S. Logothetidis, I. Alexandrou, P.B. Barna, and J. Stoemenos, Thin Solid Films 268, 57 (1995).
7. B.O. Johansson, J.E. Sundgren, J.E. Greene, A. Rockett, and S.A. Barnett, J. Vac. Sci. Technol. A 3, 303 (1985). 8. S.K. Rha, W.J. Lee, S.Y. Lee, Y.S. Hwang, Y.J. Lee, D.I.
Kim, D.W. Kim, S.S. Chun, and C.O. Park, Thin Solid
Films 320, 134 (1998).
Fig. 9. Histograms showing statistical distributions of reversed-biased leakage currents of Al/barrier/Ti/Si junction diodes after annealing at 500°C for 30 min.
9. J.E. Greene, J.E. Sundgren, L. Hultman, I. Petrov, and D.B. Bergstrom, Appl. Phys. Lett. 67, 2928 (1995).
10. L. Hultman, J.E. Sundgren, J.E. Greene, D.B. Bergstrom, and I. Petrov, J. Appl. Phys. 78, 5395 (1995).
11. J.S. Chun, P. Desjardins, C. Lavoie, I. Petrov, C. Cabral, Jr., and J.E. Greene, J. Vac. Sci. Technol. A 19, 2207 (2001). 12. K.Y. Ahn, M. Wittmer, and C.Y. Ting, Thin Solid Films 107,
45 (1983).
13. W.F. Wu, K.L. Ou, C.P. Chou, and J.L. Hsu, Electrochem.
Solid-State Lett. 6, G27 (2003).
14. K.T. Nam, A. Datta, S.H. Kim, and K.B. Kim, Appl. Phys.
Lett. 79, 2549 (2001).
15. W.F. Wu, C.C. Lin, C.C. Huang, H.C. Lin, T.C. Chang, R.P. Yang, and T.Y. Huang, Electrochem. Solid-State Lett. 2, 342 (1999).
16. N. Kumar, J.T. McGinn, K. Pourrezaei, B. Lee, and E.C. Douglas, J. Vac. Sci. Technol. A 6, 1602 (1988).
17. D.J. Kim, Y.T. Kim, and J.W. Park, J. Appl. Phys. 82, 4847 (1997).
18. K.L. Ou, W.F. Wu, C.P. Chou, S.Y. Chiou, and C.C. Wu,
J. Vac. Sci. Technol. B 20, 2154 (2002).
19. W.F. Wu, K.L. Ou, C.P. Chou, and C.C. Wu, J. Electrochem.
Soc. 150, G83 (2003).
20. K.C. Tsai, W.F. Wu, J.C. Chen, T.J. Pan, and C.G. Chao,
J. Vac. Sci. Technol. B 22, 993 (2004).
21. K.C. Tsai, W.F. Wu, J.C. Chen, T.J. Pan, and C.G. Chao,
J. Electrochem. Soc. 152, G83 (2005).
22. J.L. Murray, Phase Diagrams of Binary Titanium Alloys (Metals Park, OH: ASM, 1987), pp. 13–25.
23. R. Beyers, R. Sinclair, and M.E. Thomas, J. Vac. Sci.
Tech-nol. B 2, 781 (1984).
24. M. Caron, G. Gagnon, V. Fortin, J.F. Currie, L. Ouellet, Y. Tremblay, M. Biberger, and R. Reynolds, J. Appl. Phys. 79, 4468 (1996).
