Initial Nucleation Study and New Technique for Sublimation Growth

Document Sample
Initial Nucleation Study and New Technique for Sublimation Growth Powered By Docstoc
					phys. stat. sol. (a) 188, No. 2, 757–762 (2001)

            Initial Nucleation Study and New Technique
            for Sublimation Growth of AlN on SiC Substrate
            Y. Shi (a), B. Liu1) (a), L. Liu (a), J. H. Edgar (a), H. M. Meyer III (b),
            E. A. Payzant (c), L. R. Walker (c), N. D. Evans (c), J.G. Swadener (c),
            J. Chaudhuri (d), and Joy Chaudhuri (d)

            (a) Kansas State University, Department of Chemical Engineering, Manhattan,
            KS 66506, USA
            (b) Oak Ridge National Laboratory, Engineering Technology Division, Oak Ridge,
            TN 37831, USA
            (c) Oak Ridge National Laboratory, Metals and Ceramics Division, Oak Ridge,
            TN 37831, USA
            (d) Wichita State University, Mechanical Engineering Department, Wichita,
            KS 67260, USA

            (Received June 21, 2001; accepted August 4, 2001)

            Subject classification: 61.10.Nz; 68.37.Hk; 68.37.Ps; 81.05.Ea; 81.15.Gh; S7.14

Single crystal platelets of AlN were successfully grown on 6H-SiC(0001) by a novel technique de-
signed to suppress SiC decomposition, promote two-dimensional growth, and eliminate cracking in
the AlN. X-ray diffractometry and synchrotron white beam X-ray topography demonstrate that the
final AlN single crystal is of high structural quality.

Introduction Despite the rapid progress made in group-III nitride based semiconduc-
tor film growth [1], better substrates are still needed for high quality epitaxial growth.
AlN is attractive as a substrate for group-III nitride epitaxy with GaN due to its small
mismatch in both lattice constant ($2.5% along the a-axis) and thermal expansion coef-
ficient, good thermal stability (melting point >2500  C), and high resistivity [2]. To date,
bulk AlN single crystals have been limited to diameters below 15 mm [3]. From the
study by Balkas et al. [4] using 6H-SiC as seeds for bulk AlN crystal growth, three
problems were evident: SiC decomposes at high crystal growth temperatures, individual
AlN grains form instead of a single crystal, and the AlN cracks due to differences in
the thermal expansion coefficients of AlN and SiC.
   A novel technique for AlN single crystal growth on 6H-SiC(0001) was developed [5, 6]
to remedy these problems. This technique consists of depositing an AlN seeding layer by
metalorganic chemical vapor deposition (MOCVD) at 1100  C, then depositing an
(AlN)x(SiC)1– alloy film by sublimation from a mixture of AlN–SiC powders at 1800  C,
and finally depositing pure AlN by sublimation from a pure AlN source, also at 1800  C.
   In this paper, details about this technique are provided. First, the initial stages of
AlN sublimation growth on bare 6H-SiC were investigated by SEM. Second, the effect
of AlN MOCVD buffer layer and composition variation in the region of
(AlN)x(SiC)1– alloys was studied by SEM and SAM. Third, the surface morphologies
and crystal qualities of samples in different steps were characterized by AFM and

      ) Corresponding author; Tel: 1-785-532-4325; Fax: 1-785-532-7372; e-mail:

               # WILEY-VCH Verlag Berlin GmbH, 13086 Berlin, 2001    0031-8965/01/18811-0757 $ 17.50þ.50/0
758                  Y. Shi et al.: Initial Nucleation Study and Sublimation Growth of AlN on SiC

XRD. Finally, X-ray diffractometry and synchrotron white beam X-ray topography de-
monstrate that the final AlN single crystal has high structural quality.

Experimental Procedures Sublimation growth experiments were conducted in a resis-
tively heated furnace using tungsten wire mesh heating elements. The furnace and the
growth process were described in detail in Ref. [5]. The thickness of long time growth sam-
ples was measured with a micrometer, with an accuracy of about 1mm. The average growth
rate was about 5 mm/h at 1800  C and 500 Torr. Using an (AlN)x(SiC)1– alloy single crystal
as the seed, pure AlN was sublimated to form a high quality AlN single crystal for 100 h.
   The AlN buffer layer was grown by low pressure (76 Torr) MOCVD. Trimethylalu-
minum (TMA) and ammonia (NH3) were the Al and N sources, and Pd-cell purified
H2 was the carrier gas. The substrates were preheated at 1100  C for 10 min in 3 slm of
H2 for surface cleaning. Then the AlN films were deposited at the same temperature
under H2, NH3 and TMA flow rates of 3 slm, 3 slm and 0.6 sccm, respectively. A 2 mm
thick AlN buffer layer was obtained in 4 h.
   The samples were characterized using a Hitachi S-4700 scanning electron microscope
(SEM) with electron backscattered diffraction (EBSD), a Philips XL30FEG SEM with
energy dispersive X-ray spectrometry (EDS), a Physical Electronics PHI680 scanning
Auger microprobe (SAM), a Park autoprobe M5 atomic force microscope (AFM) and a
Scintag PTS four axis X-ray diffractometer (XRD) with Cr-Ka radiation (25 kV Â 8 mA).
Reflection topography was undertaken at Stanford Synchrotron Radiation Laboratory
(SSRL) using Laue diffraction technique and white beam radiation.

Results The initial AlN growth on bare 6H-SiC was studied using high magnification
SEM. Figure 1 shows the sample grown at 1900  C under 500 Torr N2 for 15 min. Some
areas of SiC substrate were covered by large AlN crystals, the other areas were covered
by small hillocks of SiC. AlN nucleated only on the SiC hillocks and the AlN nuclei ap-
peared rotated 30 with respect to the SiC hillocks. The lattice mismatch on the basal
plane between AlN and 6H-SiC is less than 1% at room temperature and 1.3% at 2200 K.
This close match in the basal planes is expected to facilitate the formation of nearly per-
fect junctions, i.e. the relationship between the lattices was expected to be parallel, with
[0002]AlN || [0006]SiC and [1120]AlN || [1120]SiC [7]. However, Kikuchi patterns observed in
EBSD studies showed that [0002]AlN || [0006]SiC but [1120]AlN and [1120]SiC differ by a 30o
rotation. Therefore, this result confirms that there was misorientation between the AlN
                                               crystals and the SiC hillocks in the initial stage
                                               of crystal growth. Further studies will be un-
                                               dertaken to explain the reason for this unex-
                                               pected rotation. The decomposition of 6H-SiC
                                               was discussed in detail in Ref. [8].
                                                  Such problems are avoided by first deposit-
                                               ing an AlN layer by MOCVD. Previous studies
                                               [5, 6] showed that an AlN MOCVD buffer

                                              Fig. 1. SEM image of AlN crystal grown on 6H-
                                              SiC(0001) substrate at 1900  C, 500 Torr and
                                              15 min with magnification 2000Â and tilted 15
phys. stat. sol. (a) 188, No. 2 (2001)                                                            759

                                                   Fig. 2. SEM image of cross section of alloy crystal
  alloy crystal   region                           with 1 mm AlN MOCVD buffer layer

 MOCVD buffer      layer                 layer leads to continuous, single grain growth
                                         mode. In this study, the SEM image of the cross
   SiC substrate                         section of alloy crystal in Fig. 2 clearly shows the
                                         AlN MOCVD buffer layer with a thickness of
                                         1 mm. The composition of the substrate and the
                                         buffer layer were pure SiC and pure AlN respec-
                                         tively, but the alloy layer had the same composi-
                                         tion, approximately (AlN)0.8(SiC)0.2, regardless
of the ratio of AlN to SiC powder in the source. Thus, the AlN MOCVD buffer layer not
only protected the SiC substrate, but also acted as seeding layer for (AlN)0.8(SiC)0.2 alloy
   The AFM images of the samples in the subsequent growth process (i.e., from the as-
received SiC substrate, SiC with AlN MOCVD buffer layer, (AlN)0.8(SiC)0.2 alloys to
final pure AlN crystal) showed that the surface roughness of the obtained AlN is com-
parable with the as-received SiC substrate. As-received (0001)Si 6H-SiC substrates typi-
cally contain randomly oriented polishing scratches as shown in Fig. 3a2). The surface
of the AlN MOCVD buffer layer grown on the substrate indicating a 2D growth mode
in Fig. 3b. After subsequent (AlN)0.8(SiC)0.2 alloy sublimation growth, a very rough and

                           RMS 0.882 nm                        RMS 1.751 nm
100 nm                                    100 nm

                                    (a)                                   (b)

                           RMS 2.507 nm                          RMS 0.621 nm
100 nm                                    100 nm

                                    (c)                                   (d)

Fig. 3 (colour). AFM images of a) as-received (0001)Si 6H-SiC substrate; b) SiC with AlN
MOCVD buffer layer; c) (AlN)0.8(SiC)0.2 alloy; d) final pure AlN crystal

      ) Colour figure is published online (
760                 Y. Shi et al.: Initial Nucleation Study and Sublimation Growth of AlN on SiC

         Ta b l e 1
         Results of X-ray rocking curve of symmetric [0002] asymmetric [112] peaks and lattice
         parameters of SiC substrate, SiC substrate with AlN MOCVD buffer layer, (AlN)0.8(SiC)0.2
         alloy and pure AlN crystal

         sample            [0002]                      [112Š
                                                          2                       a      d value

                           2q ( )      FWHM           2q ( )    FWHM            ( )
                                                                                   A     ( )
                                        (arcsec)                  (arcsec)

         SiC substrate*)    53.75        489           120.93      562            3.080 5.065
         SiC substrate with 53.88        478           120.89      632            3.091 5.047
         AlN MOCVD
         buffer layer
         (AlN)0.8(SiC)0.2   54.32       1141           120.32     1454            3.104 5.016
         pure AlN           54.56        238           120.12      461            3.113 4.996
         *) [0006] and [116] for SiC

irregular surface structure was produced as shown in Fig. 3c. Figure 3d shows the image
of the final pure AlN crystal, which had a smoother and more regular surface.
   The conventional q–2q diffraction patterns of symmetric [0002] and asymmetric
[1122] peaks were measured for the AlN single crystal. The lattice parameters deter-
mined from these two peaks were a ¼ 3:113  and c ¼ 4:996  in agreement with the
                                                  A                 A
repoted lattice parmeters of AlN [9]. The rocking curves FWHM (full width at half
maximum) and peak positions are reported in Table 1. For comparison, data obtained
for the SiC substrate, SiC substrate with AlN MOCVD buffer layer and the
ðAlNÞ0:8 ðSiCÞ0:2 alloy film are also presented in the same table. The AlN single crystal
has [0002] symmetric and [112] asymmetric FWHM as low as 238 and 461 arcsec, respec-
tively. Under the same X-ray analysis condition, the [0002] and [1122] peaks rocking
curve FWHM for the SiC substrate were 489 and 562 arcsec, suggesting the AlN single
crystal was of comparable quality to the original substrate. The alloy crystal had much
wider FWHM values of rocking curve for both symmetric and asymmetric peaks (1141
and 1454 arcsec, respectively) compared to either the as-received SiC substrate and the
pure AlN single crystal. This is not unexpected because of the atomic occupancy disorder
                                    in the alloy and composition variation along the growth
                                    axis as measured by SAM. Apparently, pure AlN subli-
                                    mation growth on the alloy seed improved the crystal
                                    quality significantly. The disparity between the FWHM
                                    of the symmetric and asymmetric peaks is commonly
                                    observed for nitride growth [10–12] and has been attrib-

                                     Fig. 4. Synchrotron white beam X-ray (SWBX) topography
                                     for the final pure AlN crystal in reflection. g ¼ [103]. (A)
                                     Dislocation network structure, (B) cluster of dislocation.
                                     White region indicate absence of dislocations with Burgers
                                     vector [100] type
phys. stat. sol. (a) 188, No. 2 (2001)                                                           761

uted to the presence in the nitride films of significantly more dislocations of edge type
rather than screw type [12, 13].
  Figure 4 shows the synchrotron white beam X-ray (SWBX) reflection topograph for
the final pure AlN crystal in refelction. Since individual dislocations can be identified in
some areas, dislocation density is low in those areas. Estimated dislocation density in
the low dislocation density area is about 5.1 Â 104 cm – .

Conclusions Seeded bulk growth of crack-free AlN single crystal was achieved on SiC
substrate by (i) depositing an AlN buffer layer on the SiC substrate by MOCVD at
1100  C, (ii) forming an (AlN)0.8(SiC)0.2 alloy film by condensing vapors sublimated (at
temperatures above 1700  C) from a source mixture of AlN–SiC powders, followed by
(iii) condensing vapors sublimated from a pure AlN source. This method solves three
problems previously associated with using SiC as seed crystals for AlN growth. First, the
low temperature deposited AlN layer promotes layer-by-layer growth forming a continu-
ous single grain. Second, the intermediate properties of the (AlN)0.8(SiC)0.2 alloy reduces
cracking in the growing crystal. Third, the presence of silicon and carbon in the source
material reduces or eliminates the decomposition of the SiC substrate during sublimation
crystal growth, enabling longer duration crystal growth. By employing this technique, a
high quality AlN single crystal was produced, as determined by X-ray diffraction. Such
crystals are ideal for seeding AlN crystal growth at high temperature and hence high
growth rates, to produce substrates for AlxGa1– N based electronics.

Acknowledgements We are grateful for the support of this research from BMDO
(Contract No: N00014-98-C-0407), ONR (Contract No. N00014-99-1-0104), and the As-
sistant Secretary for Energy Efficiency and Renewable Energy, Office of Transportation
Technologies, as part of the High Temperature Materials Laboratory User Program,
Oak Ridge National Laboratory. Oak Ridge National Laboratory is managed by UT-
Battelle, LLC, for the U.S. Department of Energy under contract number DE-AC05-
00OR22725. The Oak Ridge National Laboratory SHaRE Collaborative Research Cen-
ter was sponsored by the Division of Materials Sciences and Engineering, U.S. Depart-
ment of Energy, under contract DE-AC05-00OR22725 with UT-Battelle, LLC, and
through the SHaRE Program under contract DE-AC05-76OR00033 with Oak Ridge
Associated Universities. J. Chaudhuri and Joy Chaudhuri acknowledge the financial
support by NSF/EPSCoR (grant number EPS-9977776) and portions of the work was
done by Dr. Z. Rek at Stanford Synchrotron Radiation Laboratory, operated by the
U.S. Department of Energy, Office of Basic Energy Sciences.

 [1] S. Nakamura, M. Senoh, N. Iwasa, and S. Nagahama, Jpn. J. Appl. Phys. 34, 797 (1995).
 [2] H. Q. Lu, I. B. Bhat, B. C. Lee, G. A. Slack, and L. J. Schowalter, Mater. Res. Soc. Symp.
     Proc. 482, 277 (1997).
 [3] L. J. Schowalter, Y. Shuterman, R. Wang, I. Bhat, G. Arunmozhi, and G. A. Slack, Appl.
     Phys. Lett. 76, 985 (2000).
 [4] C. M. Balkas, Z. Sitar, T. Zheleva, L. Bergman, R. Nemanich, and R. F. Davis, J. Cryst.
     Growth 179, 363 (1997).
 [5] Y. Shi, Z. Y. Xie, L. Liu, B. Liu, J. H. Edgar, and M. Kuball, J. Cryst. Growth 233, 177 (2001).
 [6] Y. Shi, B. Liu, L. Liu, J. H. Edgar, E. A. Payzant, J. M. Hayes, and M. Kuball, MRS Internet
     J. Nitride Semicond. Res. 6, 5 (2001).
 [7] F. A. Ponce, C. G. Van de Walle, and J. E. Northrup, Phys. Rev. B 53, 7473 (1996).
762                  Y. Shi et al.: Initial Nucleation Study and Sublimation Growth of AlN on SiC

 [8] B. Liu, Y. Shi, L. Liu, J. H. Edgar, and D. N. Braski, Mater. Res. Soc. Symp. Proc. 639, G3.13
 [9] W. J. Meng, in: Properties of Group III Nitrides, Ed. J.H. Edgar, INSPEC, London 1994.
[10] F. A. Ponce, B. S. Krusor, J. S. Major, W. E. Plano, and D. F. Welch, Appl. Phys. Lett. 67, 410
[11] Q. Zhu, A. Botchkarev, W. Kim, Ú. Aktas, B. Sverdlov, H. Morkoc, S.-C. Y. Tsen, and D. J.
     Smith, Appl. Phys. Lett. 68, 1141 (1996).
[12] B. Heying, X. H. Wu, S. Keller, Y. Li, D. Kapolnek, B. P. Keller, S. P. DenBaars, and J. S.
     Speck, Appl. Phys. Lett. 68, 643 (1996).
[13] T. Metzger, R. Stommer, M. Schuster, H. Gobel, R. Hopler, E. Born, T. H. Metzger, S. Chris-
                        ¨                         ¨          ¨
     tiansen, H. P. Strunk, O. Ambacher, and M. Stutzmann, phys. stat. sol. (a) 162, 529 (1997).