AlN as an encapsulate for annealing SiC

Transcription

1 JOURNAL OF APPLIED PHYSICS VOLUME 83, NUMBER JUNE 1998 AlN as an encapsulate for annealing SiC K. A. Jones, K. Xie, D. W. Eckart, and M. C. Wood Army Research Lab, SEDD, Adelphi, Maryland V. Talyansky, R. D. Vispute, and T. Venkatesan Physics Department, University of Maryland, College Park, Maryland K. Wongchotigul and M. Spencer MSRCE, School of Engineering, Howard University Washington, District of Columbia Received 17 November 1997; accepted for publication 16 March 1998 AlN films grown by either organometallic vapor phase epitaxy OMVPE or pulsed laser deposition PLD can be used to encapsulate SiC when heated in an argon atmosphere at temperatures at least as high as 1600 C for times at least as long as 30 min. The coverage of the AlN remains complete and the AlN/SiC interface remains abrupt as determined by Auger electron spectroscopy. However, considerable atomic movement occurs in the AlN at 1600 C, and holes can form in it as the film agglomerates if there are large variations in the film thickness. Also, the SiC polytype near the surface can in some instances be changed possibly by the stress generated by the epitaxial AlN film. Using x-ray diffraction measurements, we also found that, during the 1600 C anneal, grains with nonbasal plane orientations tended to grow at the expense of those with basal plane orientations in the OMVPE films, whereas grains with only the basal plane orientation tended to grow in the PLD films. However, there is no indication that the type of grain growth that is dominant affects the film s ability to act as an encapsulate American Institute of Physics. S I. INTRODUCTION Considerable interest exists in being able to ion implant dopants such as N, 1 7 Al, 8 11 B, 8,10,12 or Ga 13 into SiC for controlled doping because the rate of dopant diffusion in it is extremely low even at temperatures as high as 1600 C. 14 Moreover, being able to create field-effect transistor channels, 15 as is often done with Si and GaAs, would have considerable economic benefits, and being able to heavily dope regions such as source and drain regions 16 before forming Ohmic contacts can considerably reduce the contact resistance. This is particularly true for devices such as gate turn-off thyristors where the layer that is contacted is often lightly doped. 17 Numerous studies involving ion implantation have been made, 1 13 but in some instances there have been conflicting results. For example, by annealing some researchers were able to activate Al implanted at room temperature, 10 while others were not. 8,18 In one instance the sheet resistance decreased continuously with the annealing time, 7 whereas in another study it reached a minimum and then increased. 6 Also, in one case the coimplantation of C with Al appeared to have no effect, 9 but in another study, it appeared to have a profound effect. 11 One important reason why ion implant technology has not been fully developed for SiC and, in some instances, conflicting results have been produced is that very high temperatures are required to activate the implant, and at these temperatures a multitude of interrelated processes can be activated. From their electrical measurements, Kawase et al. 19 showed that an annealing temperature of 1600 C is required for a significant amount of activation to occur while others have shown that a temperature of or 1500 C Refs. 2 and 7 is required for the activation process to begin. However, at these temperatures the preferential evaporation rate of Si is high enough to destroy the surface quality. 1,18,20,21 Moreover, the Si vacancies created by this process can enhance diffusion rates near the surface. This effect, in fact, has been used to explain the different results on the effects of annealing time. 6 To reduce the problem of silicon evaporation, samples have been annealed in a SiC boat 4,8,10 or in the presence of SiC powder. 11 However, even this precaution cannot prevent the occurrence of surface degradation so severe that electrical measurements cannot be made on samples annealed at 1650 C. 8 Moreover, if there is any oxygen in the atmosphere, it could etch the surface by forming volatile oxides. 22,23 A second SiC wafer could be used to cover the implanted surface, but this is expensive, and some surface degradation probably occurs because a hermetic seal is not formed. The usual encapsulates, SiO 2 and Si 3 N 4, cannot be used at this temperature because their vapor pressures are too high. 24,25 With a stoichiometric vapor pressure that is almost two orders of magnitude less at 1600 C, AlN 26 could possibly be used as an encapsulate. In this article we show that under certain conditions it can be when the AlN is deposited by organometallic vapor phase epitaxy OMVPE or pulsed laser deposition PLD. OMVPE films were used because it has been demonstrated that good quality films can be grown, 27 and PLD films were used because they can be deposited at lower temperatures without being grossly contaminated with unreacted species such as carbon. 28 The AlN films /98/83(12)/8010/6/$ American Institute of Physics

2 J. Appl. Phys., Vol. 83, No. 12, 15 June 1998 Jones et al FIG. 1. SEM micrographs of the surface of AlN films grown by OMVPE and a deposited on SiC, b deposited on SiC and annealed at 1600 C for 30 min, c deposited on Al 2 O 3, and d deposited on Al 2 O 3 and annealed at 1600 C for 30 min. are examined by observing the surface morphology using scanning electron microscopy SEM, recording variations in the film thickness using energy dispersive x-ray EDX analysis, determining the abruptness of the AlN/SiC junction from the Auger electron spectroscopy AES depth profiles, and studying changes in the orientations of the AlN film and changes in the structure of the SiC near the interface by x-ray diffraction XRD. Films grown on sapphire, single crystal Al 2 O 3, were examined also to better determine if the stability of AlN was an inherent property of the AlN or if it was strongly influenced by the fact it was grown epitaxially on the SiC to which it is quite closely lattice matched. The lattice match with Al 2 O 3 is poor, and Al 2 O 3 is very inert. Moreover, it is relatively easy to deposit amorphous AlN on it so that any stability garnered by vapor phase epitaxial growth could be eliminated. Also, the AlN could be stable at the elevated temperatures and still be adversely affected by the high silicon vapor pressure in SiC, whereas it would be unaffected on the Al 2 O 3. II. PROCEDURE 4H SiC substrates used for the OMVPE growth were chemically cleaned in an HF:H 2 O 1:10 solution and then thermally cleaned in situ by a 900 C anneal for 3 min. The Al 2 O 3 substrates were chemically cleaned in an acetone/methanol/h 2 O solution and thermally cleaned in H 2 as the temperature was ramped up. The films grown on SiC were deposited at 1300 C, and the films grown on Al 2 O 3 were deposited at 1180 C because it has been determined previously that these temperatures produced films with the best morphology; both types of films had a nominal thickness of 200 nm. The 6H SiC substrates for PLD growth were also chemically cleaned in a 10% HF solution, while the Al 2 O 3 substrates were thermally cleaned during the heating up process. Both types of samples were prepared in a base pressure of Torr using a KrF excimer laser operating at 248 nm with an energy density of 2 J/cm 2.To try to deposit amorphous films, the growth temperature for the SiC substrates was 800 C, and the growth temperature for the Al 2 O 3 substrates was 900 C. Lower temperatures could not be used because the film morphology deteriorated; both types of films were nominally 100 nm thick. The samples were annealed in a resistively heated furnace in an argon atmosphere at temperatures of 1200, 1400, or 1600 C for 30 min, and then were cooled down at an initial rate of 20 C/min. The film and substrate were then examined by SEM, EDX, AES, and XRD. III. RESULTS For the OMVPE samples no changes in the morphology were observed for the films annealed at 1200 and 1400 C. However, as seen in Fig. 1, significant morphological changes occur for growth on both types of substrates when the samples are annealed at 1600 C indicating a considerable atomic motion. Similar observations were made for the PLD films. As seen in Fig. 2, there was so much atomic movement in the PLD film grown on SiC and annealed at 1600 C that the AlN has agglomerated in some regions and left holes in others. The film thickness varied from point to point in this PLD film as was demonstrated by comparing the EDX peak heights of Si and Al in the as-grown sample. This was also confirmed by observing cleaved cross sections in the SEM. All of the as-grown OMVPE films showed little variation in thickness, and this, too, was verified with cleaved cross sections. The ratio of the Si peak height to the Al peak height was much smaller in the OMVPE samples confirming they were thicker than the PLD films. This, too, was seen in the cleaved cross sections. The films grown on Al 2 O 3 also show no discernible changes when annealed at 1200 and 1400 C, but considerable atomic movement occurs after the 1600 C anneal. However, both the OMVPE and PLD films retain their complete coverage of the sub-

3 8012 J. Appl. Phys., Vol. 83, No. 12, 15 June 1998 Jones et al. FIG. 2. SEM micrographs of the surface of AlN films grown by PLD and a deposited on SiC, b deposited on SiC and annealed at 1600 C for 30 min, c deposited on Al 2 O 3, and d deposited on Al 2 O 3 and annealed at 1600 C for 30 min. strate, and cleaved cross sections show that there is little variation in the thickness of the as grown films. In the AES depth profiles in Fig. 3 for the PLD film with large thickness variations and a typical OMVPE film both of which are deposited on SiC and annealed at 1600 C, one sees that the profile is very abrupt for the OMVPE sample. However, the PLD sample has a gradual profile which could indicate a considerable in-diffusion. Most of the x-ray spectra were traditional rocking curves for the substrate and film, but occasionally they were unusual. Only the latter are discussed. The x-ray diffraction pattern for an OMVPE as-deposited film shown in Fig. 4 a has a second basal plane reflection suggesting that a polytype other than the original 4H polytype is present. In most of the OMVPE 1600 C annealed samples the 002 AlN peak is greatly reduced compared to the peak for the as-grown sample, while at the same time the 100 and 101 peaks have increased in intensity. However, this considerable amount of grain growth did not appear to affect the film s ability to act as an encapsulate. As is shown in Figs. 4 c and 4 d some AlN grains are deposited on Al 2 O 3 with the 100 and 101 orientations, but their peaks are not increased in intensity by the 1600 C anneal as they were for the AlN film grown on SiC. The diffraction pattern for a PLD asdeposited AlN film grown on SiC shown in Fig. 5 a has no 002 AlN peak because it was amorphous. This was probably caused by the low deposition temperature. On annealing at 1600 C this film shows only grains with a preferred 001 orientation as shown in Fig. 5 b. However, it also shows that some of the 6H SiC has been converted to another polytype as there are two basal plane SiC peaks. For all of the PLD films deposited on Al 2 O 3 both the as-grown and the annealed films show only the 002 peak. IV. DISCUSSION FIG. 3. AES depth profiles for AlN films annealed at 1600 C that were grown on SiC substrates by a OMVPE and b PLD. The results suggest that AlN is robust at temperatures as high as 1600 C, and its stability does not depend upon the type of substrate it is deposited on. Thus, with a few caveats it appears that AlN can be used as an encapsulate for SiC. One concern is that the relatively high vapor pressure of silicon in SiC could cause adhesion problems or could out diffuse into the AlN with its increased mobility. There is no evidence of adhesion problems even when the deposited film is amorphous, and the Auger results in Fig. 3 a show little evidence of diffusion across the interface when the AlN film completely covers the SiC. However, when there are openings in the AlN film, considerable aluminum and nitrogen diffusion appears to occur. It is possible that the enhanced Al

4 J. Appl. Phys., Vol. 83, No. 12, 15 June 1998 Jones et al FIG. 4. X-ray spectra of AlN films grown by OMVPE and a deposited on SiC, b deposited on SiC and annealed at 1600 C for 30 min, c deposited on Al 2 O 3, and d deposited on Al 2 O 3 and annealed at 1600 C for 30 min. FIG. 5. X-ray spectra of AlN films grown by PLD and a deposited on SiC b deposited on SiC and annealed at 1600 C for 30 min, c deposited on Al 2 O 3, and d deposited on Al 2 O 3 and annealed at 1600 C for 30 min.

5 8014 J. Appl. Phys., Vol. 83, No. 12, 15 June 1998 Jones et al. and N in diffusion is due to Si preferentially evaporating 20,21 through the openings in the film leaving vacancies behind that assist in the in-diffusion process. It is also possible that the gradual profile in the PLD film is due to the film having pinholes as the AES profile could have a tail simply because the cross section of the AES electron beam contains pinholes as well the AlN film with a finite thickness. Holes are formed in the film that was amorphous when it was deposited because the atoms have sufficient mobility to agglomerate. Work with the other samples suggests that the agglomeration is not associated with the fact the film was originally amorphous or with the deposition method. Rather, it is a result of the large variation in the thickness of the AlN layer that primarily results from its low deposition temperature. It is possible that the holes would not appear in a thicker sample, but the film would be more likely to crack. The morphological changes for samples annealed at 1600 C shown in Figs. 1 and 2 also show there is considerable atomic movement for films deposited on either substrate or grown by either method suggesting that this movement is a bulk phenomenon. This atomic motion also accounts for the growth of the 100 and 101 crystallites in the OMVPE films. Given the strong tendency for wurtzite crystals to grow in the 001 direction, it is surprising that in this case strong c-axis growth did not occur. What makes it more surprising is that there is no apparent grain growth for the OMVPE films grown on Al 2 O 3 suggesting the 100 and 101 crystallite growth on SiC is a substrate effect. One possibility is strain reduction as there could be more strain in the films grown on SiC with its more perfect epitaxy due to the better lattice match. That epitaxy could be playing a role is supported by the fact that all the crystalline PLD films showed only grains with the 001 orientation. The epitaxy in PLD films is not as good as it is in OMVPE films due to their different growth mechanisms. Clearly more in-depth studies are required to test these suppositions. Under certain conditions stress also appears to affect the structure of the SiC as it can account for the appearance of a second polytype in the 4H SiC substrate with the asdeposited OMVPE film grown on it. This could possibly have been caused by stress created by the lattice and/or thermal coefficient mismatch. It is not likely that the transformation was due only to the high annealing temperature as it has been shown that thermal transformations do not occur until the temperature is in the vicinity of 2000 C. 29 The second polytype observed in the 6H SiC with the annealed PLD film grown on it as represented in Fig. 5 b can also be attributed to stress as much atomic motion must occur when the film changes from an amorphous to a crystalline state. V. CONCLUSIONS AlN films deposited on SiC by either OMVPE or PLD can act as an encapsulate for temperatures at least as high as 1600 C for 30 min annealing times in that they can retain their coverage of the substrate. With the complete coverage there is little in diffusion from the AlN into the substrate as there are only a limited number of Si vacancies since the Si cannot preferentially evaporate. However, the film should have a uniform thickness as there is considerable movement of the atoms at this temperature that can lead to the formation of holes if the thickness varies. The AlN films have the disadvantage that they can create SiC with a polytype different than the original polytype. This is probably due to stress created by the lattice and/or thermal mismatch between the epitaxial film and the substrate, but because the second polytype was created only some of the time under similar annealing conditions, it was not possible to define the conditions under which it will form. Even if it does not have a bearing on the quality of the encapsulate, it is interesting to note that the grains in the OMVPE grown AlN films have some grains with an off-axis orientation that continue to grow at the expense of others when the films are annealed. However, all of the grains in the crystalline PLD AlN films appear to have the 001 preferred orientation. These observations were attributed to a higher degree of strain in the OMVPE films grown on SiC due to its more perfect epitaxy. This explanation is, however, speculative and more in-depth studies will be required. ACKNOWLEDGMENTS We would like to acknowledge Honghua Du and Kiyoshi Tone for their help in the annealing experiments, Michael Derenge for his assistance in the preparation of the manuscript, and the Army Research Laboratory s Microelectronics Research Collaborative Program for its financial support. 1 M. Ghezzo, D. M. Brown, E. Downey, J. Kretchmer, W. Hennessy, D. L. Polla, and H. Bakhru, IEEE Electron Device Lett. 13, S. Yaguchi, T. Kimoto, N. Ohyama, and H. Matsunami, Jpn. J. Appl. Phys., Part 1 34, Y. Hirano and T. Inada, J. Appl. Phys. 77, T. Kimoto, A. Itoh, H. Matsunami, T. Nakata, and M. Watanabe, J. Electron. Mater. 24, J. Gardner, M. V. Rao, O. W. Holland, G. Kelner, D. S. Simons, P. H. Chi, J. M. Andrews, J. Kretchmer, and M. Ghezzo, J. Electron. Mater. 25, J. N. Pan, J. A. Cooper, Jr., and M. R. Melloch, J. Electron. Mater. 26, N. Inoue, A. Itoh, T. Kimoto, H. Matsunami, T. Nakata, and M. Inoue, J. Electron. Mater. 26, M. V. Rao, P. Griffiths, O. W. Holland, G. Kelner, J. A. Freitas, Jr., D. S. Simons, P. H. Chi, and M. Ghezzo, J. Appl. Phys. 77, M. V. Rao, P. Griffiths, J. Gardner, O. W. Holland, M. Ghezzo, J. Kretchmer, G. Kelner, and J. A. Freitas, Jr., J. Electron. Mater. 25, T. Kimoto, A. Itoh, H. Matsunami, T. Nakata, and M. Watanabe, J. Electron. Mater. 25, J. H. Zhao, K. Tone, S. R. Weiner, M. Caleca, H. Du, and S. P. Withrow, IEEE Electron Device Lett. 18, M. Ghezzo, D. M. Brown, E. Downey, J. Kretchmer, and J. J. Kopanski, Appl. Phys. Lett. 63, K. K. Burdel, A. V. Suvorov, and N. G. Chechenin, Sov. Phys. Solid State 32, R. F. Davis, G. Kelner, M. Shur, J. W. Palmour, and J. A. Edmond, Proc. IEEE 79, J. N. Pan, J. A. Cooper, and M. R. Melloch, IEEE Electron Device Lett. 18, J. Shenoy, J. A. Cooper, and M. R. Melloch, IEEE Electron Device Lett. 18, K. Xie, J. H. Zhao, J. R. Flemish, T. Burke, W. R. Buchwald, G. Lorenzo, and H. Singh, IEEE Electron Device Lett. 17, V. Heera and W. Skoupa, Mater. Res. Soc. Symp. Proc. 438, D. Kawase, T. Ohno, T. Iwasaki, and T. Yatsuo, Inst. Phys. Conf. Ser. 142,

6 J. Appl. Phys., Vol. 83, No. 12, 15 June 1998 Jones et al S. Adachi, M. Mohri, and T. Yamachina, Surf. Sci. 161, L. Muehlhoff, W. J. Choyke, M. J. Bozac, and J. T. Yates, Jr., J. Appl. Phys. 60, W. L. Vaughn and H. G. Maahs, J. Am. Ceram. Soc. 73, T. Narushima, T. Goto, Y. Iguchi, and T. Hirai, J. Am. Ceram. Soc. 74, J. Phys. Chem. Ref. Data 14, ; J. Phys. Chem. Ref. Data 14, J. Phys. Chem. Ref. Data 14, ; J. Phys. Chem. Ref. Data 14, G. A. Slack and T. F. McNelly, J. Cryst. Growth 34, C.-M. Zetterling, M. Ostling, K. Wongchotigul, M. G. Spencer, C. I. Harris, N. Nordell, and S. S. Wong, J. Appl. Phys. 82, V. Talyansky, R. D. Vispute, R. Ramesh, R. P. Sharmap, S. Choopun, M. Downes, T. Venkatesan, Y. X. Li, L. G. Salamanca-Riba, A. A. Iliadis, K. A. Jones, and M. C. Wood, Thin Solid Films to be published. 29 W. S. Yoo and H. Matsunami, Jpn. J. Appl. Phys., Part 1 30,