J. Phys. Chem. B 1999, 103, 6919-6929
6919
Surface Interactions of NH2 Radicals in NH3 Plasmas Patrick R. McCurdy,† Carmen I. Butoi, Keri L. Williams, and Ellen R. Fisher*,‡ Department of Chemistry, Colorado State UniVersity, Fort Collins, Colorado 80523-1872 ReceiVed: March 18, 1999; In Final Form: June 23, 1999
Using the imaging of radicals interacting with surfaces (IRIS) technique, the scattering of NH2 on a variety of substrates has been measured during NH3, NH3/H2, and NH3/SiH4 plasma processing. In most cases, NH2 surface scattering was greater than unity for 300 K substrates, suggesting that NH2 is produced through surface reactions. Removal of the charged species from the plasma molecular beam results in a significant decrease in scattered NH2 signal. We have also measured velocity distributions and translational temperatures for NH2 radicals scattering from 300 K substrates. Monte Carlo simulation methods were used to model spatially and temporally resolved profiles of scattered molecules. The model assumes an initial Gaussian distribution for radicals across the laser beam and calculates time-dependent changes in the profiles using Maxwell-Boltzmann distributions. For NH2 radicals scattering from a 300 K Si substrate, the translational temperature, ΘTsc, is 400 ( 30 K, significantly higher than the substrate temperature. Removal of the charged species from the plasma molecular beam results in a decrease in translational temperature for scattered NH2 molecules, ΘTsc ) 300 ( 30 K. This suggests ions are important in surface production of NH2 and in the translational temperature of the scattered radicals.
I. Introduction Ammonia plasmas are widely employed in industrial processes as a means of modifying surfaces for increased adhesion between polymer layers in composite materials;1,2 for creating ultrathin nitride passivation layers on semiconducting devices, such as polycrystalline silicon3 and GaAs;4 and extensively (with mixtures of silane) to deposit amorphous, hydrogenated silicon nitride (a-SiNx:H) as a passivation layer on integrated circuits.5,6 Silicon nitride films produced from NH3/SiH4 plasmas are known to have low pinhole densities with good stoichiometry. In the past 10 years, basic studies of NH3 adsorption on Si(111)7,8 and Si(100)9 have been performed in an effort to understand mechanisms for creating thin nitrides through chemical vapor deposition (CVD) or plasma-enhanced CVD (PECVD). In many ammonia-based plasma systems, the NH2 radical is considered an important intermediate.10,11 It has been observed in these systems using optical emission spectroscopy (OES),10,12,13 and in the photosensitized decomposition of SiH4 and NH3 mixtures using laser-induced fluorescence (LIF).14 In NH3/SiH4 plasmas, NH2 radicals are thought to react with SiHx (x ) 1-3) radicals to form tetra- and triaminosilane complexes [Si(NH2)4 and Si(NH2)3], which in turn react at a substrate surface to form a-SiNx:H.10,15 Films grown from these NH3/SiH4 plasmas had no measurable Si-H bonding and exhibited superior dielectric properties.11 Direct measurement of the surface interactions of gas-phase species, including aminosilane, SiHx, and NH2 radicals, during plasma treatment with ammonia-based plasmas, have not, however, been investigated. Our imaging of radicals interacting with surfaces (IRIS) technique directly measures the reactivity of gas-phase species * To whom correspondence should be addressed. † Current address: Pacific Northwest National Labs, WRW EMSL, P.O. Box 999, Mailstop K8-88, Richland, WA 99352. ‡ Camille-Dreyfus Teacher-Scholar.
at a substrate during plasma processing. IRIS combines molecular beam and plasma technologies with LIF to provide spatially resolved 2D images of plasma species. To date, the IRIS method has been used to investigate five different radical species in several plasma systems. The surface reactivities of these species vary from near unity (SiH),16,17 to intermediate (OH),18,19 to low (SiO and NH),20,21 to surface production of the species (CF2).22,23 Here, we investigate the interactions of NH2 with a variety of surfaces, specifically Si, SiO2, Si3N4, polyimide, poly(tetrafluoroethylene) (PTFE), and Pt, using 100% NH3 and SiH4/NH3 plasma molecular beams. With the exception of polyimide, NH2 is generated at the surface of these substrates during processing with a 100% NH3 plasma. With the SiH4/ NH3 plasma, an intermediate reactivity is observed for NH2 during deposition of an a-SiNx:H film on Si substrates. In a previous study, we reported the translational temperature for NH2 molecules in the 100% NH3 plasma molecular beam (ΘTmb) as a function of applied rf power by exploiting the time delay on our detector.24 A linear relationship was observed between ΘTmb and the applied power, P, from 5 to 150 W. In the present study, we have measured the translational temperature for the NH2 radicals scattering from substrates (ΘTsc). For NH2 scattered from a Si substrate, ΘTsc is significantly lower than ΘTmb, but is notably higher than the substrate temperature (Ts). Possible explanations for these observations are discussed along with mechanisms for surface production of NH2. II. Experimental Methods A. IRIS Apparatus. The IRIS method has been described in detail previously.17,19 In a typical IRIS experiment, feed gases enter a cylindrical glass tube reactor, rf power is applied, and a plasma is produced. Expansion of the plasma into a differentially pumped vacuum chamber, and ultimately into a highvacuum region, generates an effusive molecular beam consisting of virtually all species present in the plasma, including the species of interest. A tunable laser beam intersects the molecular
10.1021/jp9909558 CCC: $18.00 © 1999 American Chemical Society Published on Web 07/30/1999
6920 J. Phys. Chem. B, Vol. 103, No. 33, 1999 beam downstream from the plasma source and excites the selected molecule. Spatially resolved LIF signals are collected by an electronically gated, intensified charge coupled device (ICCD) located perpendicular to both the molecular beam and the laser beam, directly above the interaction region.17 A substrate is rotated into the path of the molecular beam, and LIF signals are again collected. Differences between the spatial distributions with the surface in and out of the path of the molecular beam are used to measure the ratio of incident NH2 in the molecular beam to those scattering from the surface. In the present work, the sources of the molecular beam are plasmas created from either 100% NH3 (Air Products, 99.99%), an NH3/H2 (Air Products, 99.999%) mixture, or an NH3/SiH4 (Matheson, 99.998%) mixture. The plasma is produced by the inductive coupling of 13.56 MHz rf power and is tuned by an rf matching network. For these experiments, the applied rf power, P, was kept at 45 W, unless otherwise stated. All data were collected with an ammonia flow rate of 20.0 sccm and pressures of ∼65 mTorr at the plasma source. For NH3/SiH4 plasmas, the silane flow was varied from 1.00 to 6.00 sccm with scatter data being collected at a ratio of 20:3 NH3/SiH4. The plasma molecular beam was collimated by two 1.0 mm wide slits. The slits were mounted on a liquid nitrogen cold shield cooled to below -100 °C, monitored by a K-type thermocouple.24 NH2 has a high-intensity fluorescence band in the region around 600 nm, with the most intense peak at 597.725 nm. Tunable laser light from 596 to 598 nm (∼12 mJ/pulse) was produced using an excimer-pumped (XeCl, 100 mJ, 100 Hz) dye laser system (Rhodamine 6G). Fluorescence excitation spectra were obtained to determine the identity of the fluorescing species and to ensure no unwanted signals were present. Total fluorescence was collected as the dye laser was stepped in 0.005 nm steps with a 10 s ICCD exposure per step from 596 to 598 nm. High-resolution spectra (0.001 nm) of NH2 in the molecular beam and of NH2 scattering from the surface were also taken from 597.5 to 598 nm using a gate delay of 2.0 µs and a 10.0 µs gate width with a 20 s ICCD exposure. The long gate delay and gate width allow for spatial resolution of signals from NH2 radicals in the molecular beam and from NH2 scattering from the surface. Various substrates were placed in the path of the molecular beam, specifically Si, SiO2, Si3N4, poly(tetrafluoroethylene) (PTFE), polyimide, and Pt substrates cleaned by sonicating in a methanol (spectroanalyzed, 99.9%) bath for 5 min before use. Native oxide was removed from p-type silicon substrates by immersion in a solution of 15:10:30 HF(52%)/HNO3(70%)/H2O for 30 min directly prior to use.25 Substrates were mounted on a resistively heated holder which can be translated along the molecular beam axis. Substrate temperature, TS, was controlled and measured using a K-type thermocouple attached to the front of the substrate. A ceramic block was placed on top of the holder to prevent IR light from the resistive heating wire from reaching the ICCD detector, and an IR cutoff filter (>800 nm) was also used to decrease IR radiation from the substrate at Ts > 600 K. LIF signals were collected with the ICCD gated on immediately after the laser light exits the main chamber. While NH2 has a radiative lifetime of ∼10 µs, only a 1.0 µs gate width was used to collect the fluorescence to minimize movement of the molecules during signal collection, thus simplifying our reactivity modeling.26 LIF signals were integrated by the ICCD for five accumulations of 30 000 laser pulses. Fluorescence data were collected in four sets, by alternating with the surface in and out of the molecular beam path. Backgrounds were collected
McCurdy et al. with the laser tuned to an off-resonance wavelength (581 nm). For surface scattering experiments, pixels were binned (4 × 4) to increase signal-to-noise and reduce processing time. For velocity measurements, pixels were unbinned to increase the spatial resolution of our data. For reactivity data, a onedimensional representation of the data image is made by averaging 15-20 columns of pixels (∼5.5-7 mm) along the laser path. For velocity measurements of the NH2 scattering from the substrates, the time the molecule has to travel after excitation and before fluorescence collection, τ, is an important parameter. τ is determined by the time required for the laser to trigger and fire, experimentally determined as 1.1 µs, the gate delay, d, and the gate width, w:
1 τ ) d - 1.1 + w 2
(1)
For the velocity distribution measurements made with scattered NH2 molecules, τ is stepped in 2.0 µs increments from 0.75 to 8.75 µs. The gate width was set at 500 ns to minimize movement of the molecules in time during collection of signal. LIF images taken at each gate delay are a function of time and distance the molecules travel after crossing the laser path. Velocity distribution measurements for NH2 radicals in the molecular beam have been previously reported.24 B. Numerical Simulations. 1. ReactiVity. A complete description of the model we use to simulate our surface reactivity data is given elsewhere.16,19 Briefly, our model calculates the spatial distribution of the radical number density along the laser path for molecules in the molecular beam and scattering from the surface and is based on the geometry of the IRIS experiment. In the present work, the model assumes a cosine distribution about the surface normal along the laser path for the scattered molecules. The generated curve assumes all of the incident molecules scatter from the surface. To determine the actual scattering coefficient, S, of the NH2 radicals, the generated curve is then multiplied by a scaling factor to best fit the experimental data. In most systems, we define surface reactivity R as 1 - S. R can be considered as an effective surface loss coefficient for a given species. Here, we will discuss the surface interactions of NH2 primarily from the scattering coefficient, S. The rotational state and, thus, the rotational temperature, ΘR, of the radicals can also influence the apparent S obtained in the simulation of our IRIS experiments. If the incident NH2 radicals adsorb onto the substrate, thermally equilibrate, and then desorb, the internal energy of the scattered molecules is defined by TS. If TS * ΘR, then the population of a particular rotational state will change upon equilibration with the substrate. This will result in an apparent change in S since S is proportional to the ratio of rotational state populations. The numerical simulation allows input of specific rotational state and TS values so as to correct for differences in rotational state populations, and thus correcting for apparent differences in S. This has been discussed in more detail previously.19,21 Because of the geometry of the IRIS experiment, a Doppler shift between the LIF signal from molecules in the incident beam and the signal from scattered molecules must be considered.21 The loss in LIF signal due to a Doppler shift is related to the line widths of both the laser (0.15 cm-1) and the transition being pumped. The average velocity of NH2 in the molecular beam is 831 ms-1 at an applied power of 45 W.24 This corresponds to a Doppler shift of TS) we measure for desorbing NH2 radicals. This is discussed further below. Our results using increased applied rf power with and without a grounded mesh screen, Table 1, do suggest charged particles play an important role in the scattering mechanism for NH2 from a Si substrate. Increasing the applied plasma power has several effects on plasma composition. For example, the number density of charged species (both electrons and ions) increases with rf power,42 as can the ion energy.30 Thus, if charged species are important in the surface production of NH2, S should increase at higher P. As shown in Table 1, a significant increase in scatter is measured for the 150 W plasmas compared to the 45 W plasmas. Also, if charged species are responsible for the increase in NH2 scatter, selective removal of the charged particles from the plasma molecular beam should result in a decrease in S. As shown by the data in Figure 8 and in Table 1, this is indeed what we observe for both high- and low-power plasmas. Although there is no evidence in the literature to support the idea of ion-enhanced dissociation of NH3(a) or ion-enhanced desorption of NH2(a), there is direct evidence for electroninduced dissociation of NH3(a) on Pt(111)39,43 and on Cu(110).44 Stechel and co-workers showed H atoms desorb from the surface of Pt dosed with NH3 and subsequently irradiated with electrons.43 Two types of NH3(a) were observed, a strongly bonded R state (∼1 eV) and a weakly bonded β state (∼0.4 eV). The R state occurs at low NH3 coverages (e0.25 monolayer), while at higher coverages, β state adsorbates are present. There was no evidence, however, for the desorption of NH2(a) in this study or in others.45 The increase in S after removing the SiO2 layer on the Si(100) substrates indicates a change in the interaction of plasma species with the surface. Etching the native oxide from a Si substrate with an HF solution causes significant changes in the surface. Not only is the native oxide etched away, but surface “dangling bonds” are capped by hydrogen, helping to passivate the surface with respect to air oxidation. Moreover, the etched surface has a much higher static water contact angle (∼10°) than the native oxide surface (∼70°). This suggests that the etched substrate is much more hydrophilic than the unetched substrate, which could also be a factor in the observed scattering coefficients. Surface production of NH2 was observed for all substrates with the exception of polyimide where a net surface loss was observed (S ) 0.80 or R ) 0.20). While the reason for this change is unclear, it is likely due to the polymeric nature of this substrate. In addition, the static water contact angle we measured for polyimide was 72 ( 2°, very similar to that observed for the SiO2 substrate, and substantially lower than that observed for PTFE (106 ( 2°). Other factors such as the porosity of the films could also contribute to the significant decrease in S and the subsequent increase in R. The correlation between S and contact angle for both polymeric and Si-based substrates suggests the hydrophilicity of the substrate material may be important to the scattering mechanisms for NH2. B. Rotational and Translational Energies for NH2. To fully understand the chemistry of low-temperature plasmas, knowledge of the energy partitioning between species must be acquired. While kinetic energy distributions of plasma species are relatively difficult to determine,42 the energetics of species that desorb from a substrate during plasma processing are even more difficult to determine directly in a plasma. The IRIS method allows us to obtain velocity distributions of radicals in the plasma molecular beam as well as for scattered species. In
6928 J. Phys. Chem. B, Vol. 103, No. 33, 1999
McCurdy et al.
Figure 2b, we show the relationship of ΘTmb and P for NH2 radicals in the NH3 molecular beam. Under the conditions used here (P ) 45 W) for our reactivity studies, ΘTmb ) 520 ( 8 K.24 In contrast, the scattered NH2 radicals show a significantly lower value, ΘTsc ) 400 ( 40 K, Figure 11. If the NH2 radicals were adsorbing on the substrate, equilbrating energetically, and desorbing,19 we would expect the desorbing radicals to have internal and translational temperatures characterized by TS. Although ΘTsc is significantly higher than TS, it is lower than ΘTmb. This suggests there may be partial energy accommodation on the surface. The source of the “extra” translational energy found for desorbing NH2 molecules (i.e., ΘTsc > TS), however, is likely other beam species that are bombarding the substrate during the IRIS experiments. The best explanation for these observations lies in the proposed mechanism for the surface generation of NH2 discussed in the previous section. We believe that a significant fraction of the NH2 radicals desorbing from the substrate is formed through ion bombardment of adsorbed NHx (x ) 2 or 3) species. To determine if this is plausible on an energetic basis, the ion kinetic energy is needed. Calculating the ion kinetic energy lost at a surface, i, is not straightforward, however, as it depends on the electron temperature, Te, substrate bias, and sheath voltage, Vs.42 For a nonbiased substrate, however, Vs can be estimated by eq 5,
VS )
Te M ln 2 2πm
( )
(5)
where M is the ion mass and m is the mass of an electron.42 i is then given by the sum of the ion energy entering the sheath, 0.5Te, and VS.42 Assuming Te ≈ 3 eV and M ) 0.01703 kg,46 we estimate i ≈ 14 eV. This is clearly enough energy to break any surface bonds and to produce NH2 (g) or other species. Thus, not only do charged species in the molecular beam contribute to the surface production of NH2 radicals they also contribute to the translational temperature of the desorbed NH2. V. Summary Surface loss coefficients for NH2 radicals have been measured on room-temperature Si, SiO2, Si3N4, polyimide, PTFE, and Pt during plasma processing with a 100% NH3 plasma. All NH2 surface interactions can be modeled with an adsorptiondesorption mechanism. For all substrates except polyimide, S > 1. S values greater than 1 indicate NH2 is being produced through surface reactions. Rf plasma power dependence experiments demonstrate that the amount of NH2 scatter increases with applied rf power. When charged particles are selectively removed from the plasma molecular beam, however, S decreases significantly. Moreover, with the grounded mesh screen in the plasma molecular beam, surface loss of NH2 is observed (i.e., S < 1), suggesting that charged species in the molecular beam are contributing to the surface formation of NH2. Velocity distribution measurements for NH2 radicals scattering from 300 K Si substrates indicate ΘTsc > Ts. Removal of charged species from the molecular beam yields ΘTsc ) TS, again implicating charged species in the mechanism for surface production of NH2. Acknowledgment. Financial support from the National Science Foundation (CHE-951157 and CHE-9812332) is gratefully acknowledged.
References and Notes (1) Badey, J. P.; Espuche, E.; Duc, T. M. Polymer 1996, 37, 1377. (2) Tatoulian, M.; Arefi-Khonsari, F.; Amouroux, J. Int. J. Adhes. Adhes. 1995, 15, 177. (3) Cheng, H.-C.; Wang, F.-S.; Huang, C.-Y. IEEE Trans. Elec. DeVices 1997, 44, 64. (4) Jones, M. E.; Shealy, J. R.; Engstrom, J. R. Appl. Phys. Lett. 1995, 67, 542. (5) Lee, K. R.; Sundaram, K. B.; Malocha, D. C. J. Mater. Sci. 1993, 4, 283. (6) Reif, R.; Kern, W. In Thin Film Processes II; Vossen, J. L., Kern, W., Eds.; Academic: San Diego, CA, 1991; p 537. (7) Tanaka, S.; Onchi, M.; Nishijima, M. Surf. Sci. Lett. 1987, 191, L756. (8) Isu, T.; Fujiwara, K. Solid State Commun. 1982, 42, 477. (9) Kubler, L.; Hlil, E. K.; Bolmont, K.; Gewinner, G. Surf. Sci. 1987, 183, 503. (10) Murley, D. T.; Gibson, R. A. G.; Dunnett, B.; Goodyear, A.; French, I. D. J. Non-Cryst. Solids 1995, 187, 324. (11) Smith, D. L.; Alimonda, A. S.; Chen, C.-C.; Ready, S. E.; Wacker, B. J. Electrochem. Soc. 1990, 137, 614. (12) Hicks, S. E.; Gibson, R. A. G. Plasma Chem. Plasma Process. 1991, 11, 455. (13) Vervloet, M.; Merienne-Lafore, M. F. J. Chem. Phys. 1978, 69, 1257. (14) Fuyuki, T.; Allain, B.; Perrin, J. J. Appl. Phys. 1990, 68, 3322. (15) Kushner, M. J. J. Appl. Phys. 1992, 71, 4173. (16) Ho, P.; Breiland, W. G.; Buss, R. J. J. Chem. Phys. 1989, 91, 2627. (17) McCurdy, P. R.; Bogart, K. H. A.; Dalleska, N. F.; Fisher, E. R. ReV. Sci. Instrum. 1997, 68, 1684. (18) Fisher, E. R.; Ho, P.; Breiland, W. G.; Buss, R. J. J. Chem. Phys. 1993, 97, 10287. (19) Bogart, K. H. A.; Cushing, J. P.; Fisher, E. R. Chem. Phys. Lett. 1997, 267, 377. Bogart, K. H. A.; Cushing, J. P.; Fisher, E. R. J. Phys. Chem. B 1997, 101, 10016. (20) Buss, R. J.; Ho, P.; Weber, M. E. Plasma Chem. Plasma Process. 1993, 13, 61. (21) Fisher, E. R.; Ho, P.; Breiland, W. G.; Buss, R. J. J. Phys. Chem. 1992, 96, 9855. (22) Mackie, N. M.; Venturo, V. A.; Fisher, E. R. J. Phys. Chem. B 1997, 101, 9425. Mackie, N. M.; Capps, N. E.; Fisher, E. R. J. Phys. Chem., submitted for publication. (23) Capps, N. E.; Mackie, N. M.; Fisher, E. R. J. Appl. Phys. 1998, 84, 4736. (24) McCurdy, P. R.; Venturo, V. A.; Fisher, E. R. Chem. Phys. Lett. 1997, 274, 120. (25) This solution has been previously shown to etch SiO2 films. Bogart, K. H. A.; Ramirez, S. K.; Gonzales, L. A.; Bogart, G. R.; Fisher, E. R. J. Vac. Sci. Technol. A 1998, 16, 3175. Alonso, J. C.; Ramirez, S. J.; Garcia, M.; Ortiz, A. J. Vac. Sci. Technol. A 1995, 13, 2924. (26) The movement of an NH2 radical of average velocity at 900 K is less than two 4 × 4 binned pixels for a time delay of 1.0 µs. (27) Dressler, K.; Ramsey, D. A. Philos. Trans. A 1959, 251, 553. (28) Halpern, J. B.; Hancock, G.; Lenzi, M.; Welge, K. H. J. Chem. Phys. 1975, 63, 4808. (29) Actual values for both the translational temperatures and velocity distributions are provided in ref 24. (30) Grill, A. Cold Plasma in Materials Fabrication; IEEE Press: Piscataway, NJ, 1994. (31) Static contact angles for water were measured using the sessile drop method with a contact angle goniometer (Rame´-Hart model 100). (32) Liston, E. M.; Martinu, L.; Wertheimer, M. R. J. Adhes. Sci. Technol. 1993, 7, 109 and references therein. (33) Serway, R. A. Physics for Scientists and Engineers, 2nd ed.; Saunders: New York, 1986. (34) This is based on FTIR analysis of the substrates and on the previous IRIS studies of NH radicals, ref 21. (35) Dresser, M. J.; Taylor, P. A.; Wallace, R. M.; Choyke, W. J.; Yates, J. T., Jr. Surf. Sci. 1989, 218, 75. (36) Johnson, A. L.; Walczak, M. M.; Madey, T. E. Langmuir 1988, 4, 277. (37) Bozso, F.; Avouris, Ph. Phys. ReV. B 1988, 38, 3937. (38) Larsson, C. U. S.; Flodstro¨m, A. S. Surf. Sci. 1991, 241, 353. (39) Sun, Y.-M.; Sloan, D.; Ihm, H.; White, J. M. J. Vac. Sci. Technol. A 1996, 14, 1516. (40) Zemlyanov, D. Y.; Smirnov, M. Y.; Gorodetskii, V. V. Surf. Sci. 1997, 391, 37. (41) Fattal, E.; Radeke, M. R.; Reynolds, G.; Carter, E. A. J. Phys. Chem. B 1997, 101, 8658.
Surface Interactions of NH2 Radicals in NH3 Plasmas (42) Lieberman, M. A.; Lichtenberg, A. J. Principles of Plasma Discharges and Material Processing; Wiley and Sons: New York, 1994. (43) Stechel, E. B.; Burns, A. R.; Jennison, D. R. Surf. Sci. 1995, 340, 71. (44) Mocuta, D.; Ahner, J.; Yates, J. T., Jr. Surf. Sci. 1997, 383, 299.
J. Phys. Chem. B, Vol. 103, No. 33, 1999 6929 (45) Selwyn, G. S.; Fujimoto, G. T.; Lin, M. C. J. Phys. Chem. 1982, 86, 760. (46) This corresponds to the mass of NH3+ ions in the plasma. If the most significant ionic species in the plasma is some other species, the value of i decreases.
