- Email: [email protected]
X-ray characterization of bulk AIN single crystals grown by the sublimation technique
Journal of Crystal Growth 250 (2003) 244–250
X-ray characterization of bulk AIN single crystals grown by the sublimation technique B. Raghothamachara,*, M. Dudleya, J.C. Rojob, K. Morganb, L.J. Schowalterb,c a
Department of Materials Science and Engineering, State University of New York at Stony Brook, Stony Brook, NY 11794-2275, USA b Crystal IS, Inc., 25 Cord Drive, Latham, NY 12110, USA c Department of Physics, Applied Physics and Astronomy, Rensselaer Polytechnic Institute, Troy, NY 12180-3590, USA
Abstract Bulk AlN single crystal boules have been grown using the sublimation technique and several substrates have been prepared from them. Microstructural characterization of these substrates has been performed using synchrotron white beam X-ray topography (SWBXT) and high-resolution triple axis X-ray diffraction. Our study has revealed that AlN single crystal boules grown by the sublimation technique can possess a high structural quality with dislocation densities of 800–1000/cm2 and rocking curves with a full-width at half-maximum of less than 10 arcsec. The distribution of dislocations is inhomogeneous with large areas of the wafer free from dislocations. Inclusions are also observed (density of the order of 105/cm3) and their distribution is also inhomogeneous. r 2002 Elsevier Science B.V. All rights reserved. PACS: 61.72.Ff; 81.10.Bk; 81.05.Ea Keywords: A1. X-ray topography; A1. High-resolution X-ray diffraction; A1. Defects; A2. Growth from vapor; A2. Single crystal growth; B1. Aluminum nitride
1. Introduction The development of electronic and opto-electronic devices based upon wide band gap III-nitride technology has received much attention due to their potential applications in several opto-electronic technologies in the areas of short wavelength emission and detection, and high power, highfrequency microwave devices. Due to the lack of large bulk nitride substrates, alternative commer*Corresponding author. Tel.: +1-631-632-8501; fax: +1631-632-8052. E-mail address: [email protected] (B. Raghothamachar).
cially available substrates, such as sapphire and silicon carbide (SiC), have been used to fabricate III-nitride devices. However, the use of non-nitride substrates has been demonstrated to have serious problems associated with severe lattice mismatch as well as chemical incompatibility and disparate thermal expansion coefficients. To overcome these problems, there are many efforts worldwide to grow large crystals of gallium nitride (GaN) and aluminum nitride (AlN). While gallium nitride (GaN) substrates will be preferred in the fabrication of blue/violet laser diodes (LDs), aluminum nitride (AlN) will provide good lattice matching particularly for AlGaN active regions with high Al concentrations as required for deep-UV LEDs
0022-0248/03/$ - see front matter r 2002 Elsevier Science B.V. All rights reserved. doi:10.1016/S0022-0248(02)02253-4
B. Raghothamachar et al. / Journal of Crystal Growth 250 (2003) 244–250
with emission wavelengths below 300 nm. An important application of solid-state UV optical sources is anticipated to be compact and highly sensitive bioagent detection systems for airborne pathogens like anthrax spores. Other potential applications of UV LEDs include solid-state white lighting, sterilization and disinfectant devices, and compact analytical devices for the biotechnology and pharmaceutical markets. Aluminum nitride substrates are also desired, as an alternative to semi-insulating SiC, for high power rf applications where insulating substrates with high thermal conductivity is needed and the nitride semiconductor quality is critical. In the absence of a commercial bulk nitride substrate, several approaches, such as the lateral epitaxial overgrowth (LEO) [1] have been developed to reduce the density of dislocations due to the use of non-nitride substrates (mostly SiC and sapphire) for III-nitride epitaxy. In spite of the very low density of dislocations achieved using the LEO technique, the availability of a low-dislocation-density bulk nitride substrate will substantially simplify the fabrication process and, therefore, the cost of nitride-based devices. While melt growth techniques are precluded for AlN crystal growth because of its extremely high melting point (B35001K) and the considerable nitrogen pressure required at those temperatures [2], alternative growth techniques such as crystallization of AlN from the solution in liquid Al [3,4] and by nitridization of Al in supercritical ammonia [5] at high nitrogen pressures have been demonstrated. Nevertheless, the growth of AlN single crystals using the sublimation–recondensation technique, first developed by Slack and McNelly [6], is currently recognized as the most promising method to produce large bulk AlN single crystals. In this method, a thermal gradient drives the sublimation of a polycrystalline or ceramic AlN starting material and posterior recondensation at the colder part of the crucible as a single crystal. X-ray diffraction rocking curves with full-width at half-maximum (FWHM) of less than 40 arcsec [7] and most recently [8] as low as 25 arcsec have been reported for sublimation-grown AlN crystals. Recently, we have demonstrated that, using the sublimation growth method, it is possible to
245
obtain large diameter (up to 15 mm) bulk AlN single crystals at growth rates between 0.5 and 1.0 mm/h [9]. Structural characterization of wafers cut from these boules and polished was carried out using synchrotron white beam X-ray topography (SWBXT) and high-resolution X-ray diffraction and the results are reported here.
2. Experimental section Substrate wafers were cut from an AlN boule (+ 12 mm) grown by the sublimation–recondensation technique such that each wafer surface was about 101 off the c-plane towards ½1 1 2% 0: Wafers were first mechanically polished followed by chemical mechanical polishing to obtain a final RMS roughness of about 1 nm. However, subsequently handling produced some scratches and surface damage. Peripheral regions containing cracks broke apart during the cutting and polishing processes. Consequently, uneven shaped wafers with final dimensions of less than 10 10 mm2 (see Fig. 1(a)) were obtained. Synchrotron white beam X-ray topography (SWBXT) [10] is a non-destructive technique that can be used to rapidly characterize structural defects such as dislocations, precipitates/inclusions, growth sector boundaries, etc. in large, low defect density crystals that are distorted and/or contain twinned regions. Structural defects and their type are revealed by contrast arising from interaction of the X-ray beam with distortions of the diffracting lattice planes caused by the strain fields surrounding the defects. The broad wavelength range of the white radiation can be used to identify crystallographic orientation relationships through Laue diffraction patterns. SWBXT experiments were carried out at the Stony Brook Synchrotron Topography Station, Beamline X-19C, at the National Synchrotron Light Source, Brookhaven National Laboratory. The samples were thin enough (B300 mm) to permit the use of transmission geometry. Topographs were recorded on 800 1000 Kodak Industrex SR-45-1 high-resolution X-ray film. Al filter was used to remove the higher wavelength components of the synchrotron radiation
246
B. Raghothamachar et al. / Journal of Crystal Growth 250 (2003) 244–250
( showing the different Fig. 1. (a) Optical photograph of a polished AlN wafer, (b) transmission topograph (g ¼ 1% 0 1 0; l ¼ 0:78 A) defects observed (Dg —growth dislocation, Ds —slip dislocation, S—slip band, C—crack, A—surface artifact/scratch, I—inclusion), (c) optical micrograph showing surface artifacts (A), cracks (C) and decorated dislocations (Dg —growth dislocation, Ds —slip dislocation), and (d) high-resolution triple crystal X-ray diffraction rocking curve (FWHM=9.7354 arcsec).
spectrum to prevent possible surface deterioration of the crystals during prolonged exposure. The diffraction conditions for recording the X-ray topographs were such that mto1 (m—X-ray absorption coefficient, t—thickness) indicating that direct image contrast formation mechanism is dominant. High-resolution X-ray diffraction rocking curves were recorded on a Bede D1 high-resolution diffractometer system with Cu Ka1 radiation. A Nikon optiphot microscope was used for optical microscopy.
3. Results A single polished 300 mm AlN substrate wafer was selected for detailed X-ray topography studies. An optical photograph of the polished wafer and an X-ray topograph recorded from it in the transmission geometry are shown in Figs. 1(a) and (b), respectively. Other than the normal distortion due to the recording geometry, the shape of the X-ray topograph corresponds to the actual shape of the substrate indicating that no significant strains, either due to growth or processing, are
B. Raghothamachar et al. / Journal of Crystal Growth 250 (2003) 244–250
present in the wafer. Different types of structural defects are readily identified on the X-ray topograph by the nature of the contrast they produce. Dislocation lines, both straight and curved, running along different line directions, mostly close to the periphery of the wafer, are observed. Dark contrast spots distributed inhomogeneously throughout the wafer are inclusions. They appear to be concentrated mostly in the central section of the wafer. In the peripheral regions, inclusions are observed to lie along or adjacent to dislocation lines. High magnification images reveal that each dark spot is composed of two lobes separated by a line of no contrast perpendicular to the diffraction vector. This contrast feature is typical of inclusions characterized by a radial strain field. On the right edge of the wafer where the boule was presumably detached from the crucible, several cracks are seen. Dark bands running along the /1 1 2% 0S directions emanating from the right edge as well from crack tips are slip bands produced by deformation. Fig. 1(c) shows an assembled montage of a series of optical micrographs recorded in transmission from the same wafer. By correlating features observed on the optical micrograph with those on the X-ray topograph, contrast from scratches, surface artifacts and cracks can be separated from those produced by crystalline defects. In this case, by comparing the X-ray topograph with the optical micrograph of the wafer, surface artifacts and cracks are readily correlated. Additionally, dislocation lines are also revealed on the optical micrograph that correlate with those observed on the X-ray topograph. The dislocation lines have been decorated by some sort of impurity, probably the same type that form the inclusions. The fact that these lines are dislocations and not any other type of defect or artifact is readily verified by noting that on X-ray topographs of higher order reflections, the lines exhibit bimodal contrast, indicative of the nature of strain fields surrounding dislocation lines [11]. Additionally, observation under the optical microscope revealed that the lines run through the thickness of the wafer indicating that they do not terminate in the bulk of the sample. Both curved and straight decorated dislocations are observed with some exhibiting a helical character while others have inclusions
247
segregated to them. The helical dislocations are likely to be screw type as shown by Weertman [12]. Inclusions are not observed on the optical micrograph probably because actual size is small. The X-ray image represents the strain field around the inclusion and can be several times larger than actual inclusion size. The slip bands emanating from the right edge of the wafer are also not observed on the optical micrograph, indicating that the slip dislocations are not decorated. Probably, the decoration process occurred during growth while slip bands were generated during post-growth cooling. Double and triple crystal rocking curves were recorded from this wafer using the Bede D1 highresolution diffractometer with Cu Ka1 radiation. Rocking curve widths (FWHM) recorded from different parts of the wafer ranged from 9 to 12 arcsec. A triple crystal rocking curve recorded from this wafer with an FWHM of 9.7 arcsec is shown in Fig. 1(d). From theoretical calculations, the perfect crystal rocking curve width under recorded diffraction conditions was found to be approximately 8.2 arcsec [13]. This indicates that the AlN crystal is of very high crystalline quality. Detailed examination of various X-ray topographs reveal that the dislocations are decorated by impurity to different extents depending on the type of dislocations. Dislocations inclined to the basal plane are decorated with a higher impurity concentration than basal plane dislocations. Therefore, contrast from inclined dislocations is stronger on optical micrographs and X-ray topographs. The strong contrast due to impurity interferes with dislocation contrast and precludes determination of Burgers vector directions by gb ¼ 0 analysis (g—diffraction vector, b—Burgers vector). However, the less decorated basal plane dislocations displayed zero contrast when gb ¼ 0: Using this criterion, the orientations of the Burgers vectors of these dislocations were determined as shown and explained in Fig. 2. These dislocations are less decorated probably because they were formed after growth at lower temperatures. Burgers vectors were found to lie along the /1 1 2% 0S directions with most dislocation lines having Burgers vector along ½1 1 2% 0: Therefore, these dislocations belong to the ð0 0 0 1Þ/1 1 2% 0S
248
B. Raghothamachar et al. / Journal of Crystal Growth 250 (2003) 244–250
Fig. 2. Transmission topographs from upper right corner of ( (b) g ¼ 0 2% 2 1; l ¼ 0:66 A, ( (c) wafer: (a) g ¼ 0 1 1% 0; l ¼ 0:78 A, ( (d) g ¼ 2% 0 2 1; l ¼ 0:70 A, ( (e) g ¼ g ¼ 1% 0 1 0; l ¼ 0:78 A, ( and (f) g ¼ 2% 2 0 1; l ¼ 0:64 A. ( Dislocations 1% 1 0 0; l ¼ 0:84 A, D1 extinguish in (a) and (b) indicating that their Burgers vector is along ½2 1% 1% 0; dislocations D2 extinguish in (c) and (d) indicating that their Burgers vector is along ½1 2% 1 0; dislocations D3 extinguish in (e) and (f) indicating that their Burgers vector is along ½1 1 2% 0:
slip systems. In Fig. 2, a dislocation of type D1 (i.e. b along ½2 1% 1% 0) is continuous across the crack (C) indicating that basal slip occurred before crack formation. Closer examination of the basal plane dislocation configurations in Fig. 2 shows that the dislocation lines are bowed out between pinning points. The pinning points are inclusions or inclined dislocations that impede the motion of dislocations resulting in the dislocation bowing out. Stresses generated due to differential thermal expansion between crucible and boule are responsible for deformation resulting in nucleation and motion of dislocations. Dislocations likely nucleated at the periphery and propagated toward the interior. When they are pinned by inclusions, they bow out and multiply provided the critical resolved shear stress is exceeded. The ½1 1 2% 0 type dislocation is probably preferentially oriented for
stresses generated from the right edge of the wafer and therefore underwent multiplication by slip. Other /1 1 2% 0S dislocation types exhibit considerable bowing between inclusions but stresses were probably not high enough to promote multiplication. Some dislocations overcome pinning leaving inclusions behind. On the X-ray topographs, the right edge of the wafer that was in contact with the crucible wall is characterized by cracks and slip bands. Slip bands are also observed emanating from the crack tip, nucleated due to stress concentration and also at the lower right tip of wafer. Three sets of slip bands running along the three /1 1 2% 0S directions are observed with most slip bands running along the ½1 1 2% 0 direction. Fig. 3 shows a series of topographs recorded to determine the Burgers vector of the slip dislocations in the different slip planes. The slip bands belong to the f1 1% 0 0g/1 1 2% 0S set of prismatic slip systems with most slip bands belonging to the ð1 1% 0 0Þ½1 1 2% 0 (S3) slip system. Again, the ½1 1 2% 0 type dislocation is preferentially oriented for stresses generated from the right edge of wafer. However, at the crack tip, only slip bands belonging to ð0 1 1% 0Þ½2 1% 1% 0 (S1) and ð1% 0 1 0Þ½1 2% 1 0 (S2) slip systems are observed indicating a different stress field at the crack tip. At the bottom right edge in Fig. 1, slip bands belonging to the ð0 1 1% 0Þ½2 1% 1% 0 (S1) and ð1 1 0 0Þ½1 1 2% 0 (S3) slip systems are observed. Crack formation and slip band nucleation occurred at lower temperatures when higher stresses are generated due to differential thermal expansion between boule and crucible. The splitting of basal plane dislocations (see Fig. 2) by a crack also indicates that crack formation occurred during later stages, i.e. during post-growth cooling. Since the slip bands were formed at lower temperatures, diffusion rates are lower and hence they are not decorated by impurity.
4. Discussion Major defects observed in the AlN wafer studied here are slip bands, individual dislocations decorated by impurity and impurity inclusions.
B. Raghothamachar et al. / Journal of Crystal Growth 250 (2003) 244–250
Fig. 3. Transmission topographs from central part of right ( (b) edge of wafer showing slip bands: (a) g ¼ 0 1 1% 0; l ¼ 0:78 A, ( (c) g ¼ 1% 0 1 0; l ¼ 0:78 A, ( (d) g ¼ g ¼ 0 2% 2 1; l ¼ 0:66 A, ( (e) g ¼ 1% 1 0 0; l ¼ 0:84 A, ( and (f) g ¼ 2% 0 2 1; l ¼ 0:70 A, ( Dislocations in slip band S1 extinguish in (a) 2% 2 0 1; l ¼ 0:64 A. and (b) and belong to the ð0 1 1% 0Þ½2 1 1% 0 prismatic slip system. Dislocations in slip band S2 extinguish in (c) and (d) and belong to the ð1% 0 1 0Þ½1 2% 1 0 prismatic slip system. Dislocations in slip band S3 extinguish in (e) and (f) and belong to the ð1% 1 0 0Þ½1 1 2% 0 prismatic slip system (arrow marks indicate direction of diffraction vector g).
Individual dislocations and inclusions interact with each other to produce some interesting configurations. While intentional decoration has long been utilized to study dislocations [14–16], in this case dislocations have been unintentionally decorated. Impurity decorated dislocations are
249
observed to be of three types—heavily decorated helical screw and straight/nearly straight dislocations and lightly decorated basal plane dislocations. Helical screw and straight/nearly straight dislocation lines are inclined at large angles to the wafer surface (30–601 from (0 0 0 1)). Compared to the overall growth direction of the boule, these dislocation lines are close to the growth axis. This suggests that these dislocations are likely to be growth dislocations. Further, the heavy impurity decoration also indicates that these dislocations were present during early stages of growth when higher temperatures and corresponding higher diffusion rates promote enhanced diffusion of impurity atoms to dislocation lines, which are high energy sites conducive to segregation. This reinforces the inference that these are growth dislocations. The basal plane dislocations are nucleated at the boule wall due to thermal mismatch stresses during early stages of postgrowth cooling and propagate into the interior. The basal plane dislocations interact with growth dislocations and inclusions leading to pinning and bowing out under thermal mismatch stresses. The undecorated f1 1% 0 0g prismatic slip bands are formed at much lower temperatures when diffusion rates are very low. Both basal plane and prismatic slip systems have been widely reported to operate during deformation of AlN over a wide range of temperatures [17–19]. f1 1% 0 0g prismatic slip planes are the easy glide planes with slip observed to occur at temperatures as low as room temperature. Basal plane slip dominates at higher temperatures. This is consistent with our observations. The impurity material responsible for inclusion formation and dislocation decoration is most likely to be oxygen since AlN has a high affinity for oxygen [6]. Further studies are underway to confirm this. Overall dislocation density is about 800–1000/ cm2 and dislocation density in the peripheral regions is slightly higher (1.2–1.4 103/cm2). Inclusion density is about 1–1.2 105/cm3 in the periphery and about 5.8–7 105/cm3 in the central regions. The distribution of dislocations and inclusions is inhomogeneous with nearly 40% of the wafer area free from dislocations and with a very low density (o 103/cm3) of inclusions.
250
B. Raghothamachar et al. / Journal of Crystal Growth 250 (2003) 244–250
5. Conclusions
References
Using the sublimation–recondensation growth technique, high-quality AlN single crystals have been grown. Dislocation densities of the order of 800–1000/cm2 and FWHM widths of 9–12 arcsec have been observed for these crystals. Synchrotron white beam X-ray topography in conjunction with optical microscopy has been successfully used to characterize the defect type and distributions. Impurities (most likely oxygen) decorate dislocations near growth temperatures and also form inclusions. Stresses due to differential thermal expansion of crystal and crucible cause significant deformation chiefly in the peripheral regions. f1 1 0 0g prismatic slip is favored at lower temperatures while basal plane slip systems are dominant at high temperatures. At lower temperatures, these stresses also cause cracking which severely limits the size of AlN substrates obtained from bulk crystals.
[1] K. Hiramatsu, A. Usui, Selective area growth and epitaxial lateral overgrowth of GaN, EMIS Datareview Series No.23. INSPEC, The Institution of Electrical Engineers, London, 1999, p. 440. [2] J.A. Van Vecten, Phys. Rev. B 7 (1973) 1479. [3] I. Grzegory, J. Jun, M. Bo!ckowski, St. Krukowski, M. ! Wroblewski, B. Lucznik, S. Porowski, J. Phys. Chem. Solids 56 (1995) 639. [4] M. Bockowski, M. Wroblewski, B. Lucznik, I. Grzegory, Mater. Sci. Semicond. Process. 4 (2001) 543. [5] R. Dwiliski, R. Doradziski, J. Garczyski, L. Sierzputowski, M. Palczewska, A. Wysmolek, M. Kamiska, MRS Internet J. Nitride Semicond. Res 3 (25) (1998) 1. [6] G.A. Slack, T.F. McNelly, J. Crystal Growth 34 (1976) 263. [7] M. Tanaka, S. Nakahata, K. Sogabe, H. Nakata, M. Tobioka, Jpn. J. Appl. Phys. 36 (1997) L1062. [8] R. Schlesser, R. Dalmau, Z. Sitar, J. Crystal Growth 241 (2002) 416. [9] J.C. Rojo, L.J. Schowalter, G. Slack, K. Morgan, J. Barani, S. Schujman, S. Biswas, B. Raghothamachar, M. Dudley, M. Shur, R. Gaska, N.M. Johnson, M. Kneissel, MRS Symposium Proceedings V 722, 2002, p. K1.1. [10] M. Dudley, in: D. Bloor, R. Brook, M. Flemings, S. Mahajan, R. Cahn (Eds.), Encyclopedia of Advanced Materials, Vol. 4, Pergamon Press, New York, 1994, p. 2950. [11] B.K. Tanner, X-ray Diffraction Topography, Pergamon, Oxford, 1976, p. 82. [12] J. Weertman, Phys. Rev. 107 (1957) 1259. [13] D.K. Bowen, B.K. Tanner, High Resolution X-ray Diffractometry and Topography, Taylor & Francis, London, 1998, pp. 85, 101. [14] W.C. Dash, J. Appl. Phys. 27 (1956) 1193; W.C. Dash, J. Appl. Phys. 29 (1958) 228; W.C. Dash, J. Appl. Phys. 30 (1959) 459. [15] G.D. Miles, J. Appl. Phys. 36 (1965) 1471. [16] H.J. Levinstein, C.D. Capio, J. Appl. Phys. 38 (1967) 2761. [17] A. Seifert, A. Berger, W.F. Muller, . J. Am. Ceram. Soc. 75 (1992) 873. [18] V. Audurier, J.L. Demenet, J. Rabier, Philos. Mag. A 77 (1998) 825. [19] M. Azzaz, J.P. Michel, V. Feregotto, A. George, Mater. Sci. Eng. B 71 (2000) 30.
Acknowledgements This work has been partially supported by the Office of Naval Research and the Missile Defense Agency (formerly BMDO). The authors would gratefully like to acknowledge the support of our contract monitors Dr. C.E.C. Wood (ONR) and Dr. C. Litton (AFRL). Topography was carried out at the Stony Brook Synchrotron Topography Facility, beamline X-19C, at the National Synchrotron Light Source, at Brookhaven National Laboratory, which is supported by the Department of Energy.
X-ray characterization of bulk AIN single crystals grown by the sublimation technique B. Raghothamachara,*, M. Dudleya, J.C. Rojob, K. Morganb, L.J. Schowalterb,c a
Department of Materials Science and Engineering, State University of New York at Stony Brook, Stony Brook, NY 11794-2275, USA b Crystal IS, Inc., 25 Cord Drive, Latham, NY 12110, USA c Department of Physics, Applied Physics and Astronomy, Rensselaer Polytechnic Institute, Troy, NY 12180-3590, USA
Abstract Bulk AlN single crystal boules have been grown using the sublimation technique and several substrates have been prepared from them. Microstructural characterization of these substrates has been performed using synchrotron white beam X-ray topography (SWBXT) and high-resolution triple axis X-ray diffraction. Our study has revealed that AlN single crystal boules grown by the sublimation technique can possess a high structural quality with dislocation densities of 800–1000/cm2 and rocking curves with a full-width at half-maximum of less than 10 arcsec. The distribution of dislocations is inhomogeneous with large areas of the wafer free from dislocations. Inclusions are also observed (density of the order of 105/cm3) and their distribution is also inhomogeneous. r 2002 Elsevier Science B.V. All rights reserved. PACS: 61.72.Ff; 81.10.Bk; 81.05.Ea Keywords: A1. X-ray topography; A1. High-resolution X-ray diffraction; A1. Defects; A2. Growth from vapor; A2. Single crystal growth; B1. Aluminum nitride
1. Introduction The development of electronic and opto-electronic devices based upon wide band gap III-nitride technology has received much attention due to their potential applications in several opto-electronic technologies in the areas of short wavelength emission and detection, and high power, highfrequency microwave devices. Due to the lack of large bulk nitride substrates, alternative commer*Corresponding author. Tel.: +1-631-632-8501; fax: +1631-632-8052. E-mail address: [email protected] (B. Raghothamachar).
cially available substrates, such as sapphire and silicon carbide (SiC), have been used to fabricate III-nitride devices. However, the use of non-nitride substrates has been demonstrated to have serious problems associated with severe lattice mismatch as well as chemical incompatibility and disparate thermal expansion coefficients. To overcome these problems, there are many efforts worldwide to grow large crystals of gallium nitride (GaN) and aluminum nitride (AlN). While gallium nitride (GaN) substrates will be preferred in the fabrication of blue/violet laser diodes (LDs), aluminum nitride (AlN) will provide good lattice matching particularly for AlGaN active regions with high Al concentrations as required for deep-UV LEDs
0022-0248/03/$ - see front matter r 2002 Elsevier Science B.V. All rights reserved. doi:10.1016/S0022-0248(02)02253-4
B. Raghothamachar et al. / Journal of Crystal Growth 250 (2003) 244–250
with emission wavelengths below 300 nm. An important application of solid-state UV optical sources is anticipated to be compact and highly sensitive bioagent detection systems for airborne pathogens like anthrax spores. Other potential applications of UV LEDs include solid-state white lighting, sterilization and disinfectant devices, and compact analytical devices for the biotechnology and pharmaceutical markets. Aluminum nitride substrates are also desired, as an alternative to semi-insulating SiC, for high power rf applications where insulating substrates with high thermal conductivity is needed and the nitride semiconductor quality is critical. In the absence of a commercial bulk nitride substrate, several approaches, such as the lateral epitaxial overgrowth (LEO) [1] have been developed to reduce the density of dislocations due to the use of non-nitride substrates (mostly SiC and sapphire) for III-nitride epitaxy. In spite of the very low density of dislocations achieved using the LEO technique, the availability of a low-dislocation-density bulk nitride substrate will substantially simplify the fabrication process and, therefore, the cost of nitride-based devices. While melt growth techniques are precluded for AlN crystal growth because of its extremely high melting point (B35001K) and the considerable nitrogen pressure required at those temperatures [2], alternative growth techniques such as crystallization of AlN from the solution in liquid Al [3,4] and by nitridization of Al in supercritical ammonia [5] at high nitrogen pressures have been demonstrated. Nevertheless, the growth of AlN single crystals using the sublimation–recondensation technique, first developed by Slack and McNelly [6], is currently recognized as the most promising method to produce large bulk AlN single crystals. In this method, a thermal gradient drives the sublimation of a polycrystalline or ceramic AlN starting material and posterior recondensation at the colder part of the crucible as a single crystal. X-ray diffraction rocking curves with full-width at half-maximum (FWHM) of less than 40 arcsec [7] and most recently [8] as low as 25 arcsec have been reported for sublimation-grown AlN crystals. Recently, we have demonstrated that, using the sublimation growth method, it is possible to
245
obtain large diameter (up to 15 mm) bulk AlN single crystals at growth rates between 0.5 and 1.0 mm/h [9]. Structural characterization of wafers cut from these boules and polished was carried out using synchrotron white beam X-ray topography (SWBXT) and high-resolution X-ray diffraction and the results are reported here.
2. Experimental section Substrate wafers were cut from an AlN boule (+ 12 mm) grown by the sublimation–recondensation technique such that each wafer surface was about 101 off the c-plane towards ½1 1 2% 0: Wafers were first mechanically polished followed by chemical mechanical polishing to obtain a final RMS roughness of about 1 nm. However, subsequently handling produced some scratches and surface damage. Peripheral regions containing cracks broke apart during the cutting and polishing processes. Consequently, uneven shaped wafers with final dimensions of less than 10 10 mm2 (see Fig. 1(a)) were obtained. Synchrotron white beam X-ray topography (SWBXT) [10] is a non-destructive technique that can be used to rapidly characterize structural defects such as dislocations, precipitates/inclusions, growth sector boundaries, etc. in large, low defect density crystals that are distorted and/or contain twinned regions. Structural defects and their type are revealed by contrast arising from interaction of the X-ray beam with distortions of the diffracting lattice planes caused by the strain fields surrounding the defects. The broad wavelength range of the white radiation can be used to identify crystallographic orientation relationships through Laue diffraction patterns. SWBXT experiments were carried out at the Stony Brook Synchrotron Topography Station, Beamline X-19C, at the National Synchrotron Light Source, Brookhaven National Laboratory. The samples were thin enough (B300 mm) to permit the use of transmission geometry. Topographs were recorded on 800 1000 Kodak Industrex SR-45-1 high-resolution X-ray film. Al filter was used to remove the higher wavelength components of the synchrotron radiation
246
B. Raghothamachar et al. / Journal of Crystal Growth 250 (2003) 244–250
( showing the different Fig. 1. (a) Optical photograph of a polished AlN wafer, (b) transmission topograph (g ¼ 1% 0 1 0; l ¼ 0:78 A) defects observed (Dg —growth dislocation, Ds —slip dislocation, S—slip band, C—crack, A—surface artifact/scratch, I—inclusion), (c) optical micrograph showing surface artifacts (A), cracks (C) and decorated dislocations (Dg —growth dislocation, Ds —slip dislocation), and (d) high-resolution triple crystal X-ray diffraction rocking curve (FWHM=9.7354 arcsec).
spectrum to prevent possible surface deterioration of the crystals during prolonged exposure. The diffraction conditions for recording the X-ray topographs were such that mto1 (m—X-ray absorption coefficient, t—thickness) indicating that direct image contrast formation mechanism is dominant. High-resolution X-ray diffraction rocking curves were recorded on a Bede D1 high-resolution diffractometer system with Cu Ka1 radiation. A Nikon optiphot microscope was used for optical microscopy.
3. Results A single polished 300 mm AlN substrate wafer was selected for detailed X-ray topography studies. An optical photograph of the polished wafer and an X-ray topograph recorded from it in the transmission geometry are shown in Figs. 1(a) and (b), respectively. Other than the normal distortion due to the recording geometry, the shape of the X-ray topograph corresponds to the actual shape of the substrate indicating that no significant strains, either due to growth or processing, are
B. Raghothamachar et al. / Journal of Crystal Growth 250 (2003) 244–250
present in the wafer. Different types of structural defects are readily identified on the X-ray topograph by the nature of the contrast they produce. Dislocation lines, both straight and curved, running along different line directions, mostly close to the periphery of the wafer, are observed. Dark contrast spots distributed inhomogeneously throughout the wafer are inclusions. They appear to be concentrated mostly in the central section of the wafer. In the peripheral regions, inclusions are observed to lie along or adjacent to dislocation lines. High magnification images reveal that each dark spot is composed of two lobes separated by a line of no contrast perpendicular to the diffraction vector. This contrast feature is typical of inclusions characterized by a radial strain field. On the right edge of the wafer where the boule was presumably detached from the crucible, several cracks are seen. Dark bands running along the /1 1 2% 0S directions emanating from the right edge as well from crack tips are slip bands produced by deformation. Fig. 1(c) shows an assembled montage of a series of optical micrographs recorded in transmission from the same wafer. By correlating features observed on the optical micrograph with those on the X-ray topograph, contrast from scratches, surface artifacts and cracks can be separated from those produced by crystalline defects. In this case, by comparing the X-ray topograph with the optical micrograph of the wafer, surface artifacts and cracks are readily correlated. Additionally, dislocation lines are also revealed on the optical micrograph that correlate with those observed on the X-ray topograph. The dislocation lines have been decorated by some sort of impurity, probably the same type that form the inclusions. The fact that these lines are dislocations and not any other type of defect or artifact is readily verified by noting that on X-ray topographs of higher order reflections, the lines exhibit bimodal contrast, indicative of the nature of strain fields surrounding dislocation lines [11]. Additionally, observation under the optical microscope revealed that the lines run through the thickness of the wafer indicating that they do not terminate in the bulk of the sample. Both curved and straight decorated dislocations are observed with some exhibiting a helical character while others have inclusions
247
segregated to them. The helical dislocations are likely to be screw type as shown by Weertman [12]. Inclusions are not observed on the optical micrograph probably because actual size is small. The X-ray image represents the strain field around the inclusion and can be several times larger than actual inclusion size. The slip bands emanating from the right edge of the wafer are also not observed on the optical micrograph, indicating that the slip dislocations are not decorated. Probably, the decoration process occurred during growth while slip bands were generated during post-growth cooling. Double and triple crystal rocking curves were recorded from this wafer using the Bede D1 highresolution diffractometer with Cu Ka1 radiation. Rocking curve widths (FWHM) recorded from different parts of the wafer ranged from 9 to 12 arcsec. A triple crystal rocking curve recorded from this wafer with an FWHM of 9.7 arcsec is shown in Fig. 1(d). From theoretical calculations, the perfect crystal rocking curve width under recorded diffraction conditions was found to be approximately 8.2 arcsec [13]. This indicates that the AlN crystal is of very high crystalline quality. Detailed examination of various X-ray topographs reveal that the dislocations are decorated by impurity to different extents depending on the type of dislocations. Dislocations inclined to the basal plane are decorated with a higher impurity concentration than basal plane dislocations. Therefore, contrast from inclined dislocations is stronger on optical micrographs and X-ray topographs. The strong contrast due to impurity interferes with dislocation contrast and precludes determination of Burgers vector directions by gb ¼ 0 analysis (g—diffraction vector, b—Burgers vector). However, the less decorated basal plane dislocations displayed zero contrast when gb ¼ 0: Using this criterion, the orientations of the Burgers vectors of these dislocations were determined as shown and explained in Fig. 2. These dislocations are less decorated probably because they were formed after growth at lower temperatures. Burgers vectors were found to lie along the /1 1 2% 0S directions with most dislocation lines having Burgers vector along ½1 1 2% 0: Therefore, these dislocations belong to the ð0 0 0 1Þ/1 1 2% 0S
248
B. Raghothamachar et al. / Journal of Crystal Growth 250 (2003) 244–250
Fig. 2. Transmission topographs from upper right corner of ( (b) g ¼ 0 2% 2 1; l ¼ 0:66 A, ( (c) wafer: (a) g ¼ 0 1 1% 0; l ¼ 0:78 A, ( (d) g ¼ 2% 0 2 1; l ¼ 0:70 A, ( (e) g ¼ g ¼ 1% 0 1 0; l ¼ 0:78 A, ( and (f) g ¼ 2% 2 0 1; l ¼ 0:64 A. ( Dislocations 1% 1 0 0; l ¼ 0:84 A, D1 extinguish in (a) and (b) indicating that their Burgers vector is along ½2 1% 1% 0; dislocations D2 extinguish in (c) and (d) indicating that their Burgers vector is along ½1 2% 1 0; dislocations D3 extinguish in (e) and (f) indicating that their Burgers vector is along ½1 1 2% 0:
slip systems. In Fig. 2, a dislocation of type D1 (i.e. b along ½2 1% 1% 0) is continuous across the crack (C) indicating that basal slip occurred before crack formation. Closer examination of the basal plane dislocation configurations in Fig. 2 shows that the dislocation lines are bowed out between pinning points. The pinning points are inclusions or inclined dislocations that impede the motion of dislocations resulting in the dislocation bowing out. Stresses generated due to differential thermal expansion between crucible and boule are responsible for deformation resulting in nucleation and motion of dislocations. Dislocations likely nucleated at the periphery and propagated toward the interior. When they are pinned by inclusions, they bow out and multiply provided the critical resolved shear stress is exceeded. The ½1 1 2% 0 type dislocation is probably preferentially oriented for
stresses generated from the right edge of the wafer and therefore underwent multiplication by slip. Other /1 1 2% 0S dislocation types exhibit considerable bowing between inclusions but stresses were probably not high enough to promote multiplication. Some dislocations overcome pinning leaving inclusions behind. On the X-ray topographs, the right edge of the wafer that was in contact with the crucible wall is characterized by cracks and slip bands. Slip bands are also observed emanating from the crack tip, nucleated due to stress concentration and also at the lower right tip of wafer. Three sets of slip bands running along the three /1 1 2% 0S directions are observed with most slip bands running along the ½1 1 2% 0 direction. Fig. 3 shows a series of topographs recorded to determine the Burgers vector of the slip dislocations in the different slip planes. The slip bands belong to the f1 1% 0 0g/1 1 2% 0S set of prismatic slip systems with most slip bands belonging to the ð1 1% 0 0Þ½1 1 2% 0 (S3) slip system. Again, the ½1 1 2% 0 type dislocation is preferentially oriented for stresses generated from the right edge of wafer. However, at the crack tip, only slip bands belonging to ð0 1 1% 0Þ½2 1% 1% 0 (S1) and ð1% 0 1 0Þ½1 2% 1 0 (S2) slip systems are observed indicating a different stress field at the crack tip. At the bottom right edge in Fig. 1, slip bands belonging to the ð0 1 1% 0Þ½2 1% 1% 0 (S1) and ð1 1 0 0Þ½1 1 2% 0 (S3) slip systems are observed. Crack formation and slip band nucleation occurred at lower temperatures when higher stresses are generated due to differential thermal expansion between boule and crucible. The splitting of basal plane dislocations (see Fig. 2) by a crack also indicates that crack formation occurred during later stages, i.e. during post-growth cooling. Since the slip bands were formed at lower temperatures, diffusion rates are lower and hence they are not decorated by impurity.
4. Discussion Major defects observed in the AlN wafer studied here are slip bands, individual dislocations decorated by impurity and impurity inclusions.
B. Raghothamachar et al. / Journal of Crystal Growth 250 (2003) 244–250
Fig. 3. Transmission topographs from central part of right ( (b) edge of wafer showing slip bands: (a) g ¼ 0 1 1% 0; l ¼ 0:78 A, ( (c) g ¼ 1% 0 1 0; l ¼ 0:78 A, ( (d) g ¼ g ¼ 0 2% 2 1; l ¼ 0:66 A, ( (e) g ¼ 1% 1 0 0; l ¼ 0:84 A, ( and (f) g ¼ 2% 0 2 1; l ¼ 0:70 A, ( Dislocations in slip band S1 extinguish in (a) 2% 2 0 1; l ¼ 0:64 A. and (b) and belong to the ð0 1 1% 0Þ½2 1 1% 0 prismatic slip system. Dislocations in slip band S2 extinguish in (c) and (d) and belong to the ð1% 0 1 0Þ½1 2% 1 0 prismatic slip system. Dislocations in slip band S3 extinguish in (e) and (f) and belong to the ð1% 1 0 0Þ½1 1 2% 0 prismatic slip system (arrow marks indicate direction of diffraction vector g).
Individual dislocations and inclusions interact with each other to produce some interesting configurations. While intentional decoration has long been utilized to study dislocations [14–16], in this case dislocations have been unintentionally decorated. Impurity decorated dislocations are
249
observed to be of three types—heavily decorated helical screw and straight/nearly straight dislocations and lightly decorated basal plane dislocations. Helical screw and straight/nearly straight dislocation lines are inclined at large angles to the wafer surface (30–601 from (0 0 0 1)). Compared to the overall growth direction of the boule, these dislocation lines are close to the growth axis. This suggests that these dislocations are likely to be growth dislocations. Further, the heavy impurity decoration also indicates that these dislocations were present during early stages of growth when higher temperatures and corresponding higher diffusion rates promote enhanced diffusion of impurity atoms to dislocation lines, which are high energy sites conducive to segregation. This reinforces the inference that these are growth dislocations. The basal plane dislocations are nucleated at the boule wall due to thermal mismatch stresses during early stages of postgrowth cooling and propagate into the interior. The basal plane dislocations interact with growth dislocations and inclusions leading to pinning and bowing out under thermal mismatch stresses. The undecorated f1 1% 0 0g prismatic slip bands are formed at much lower temperatures when diffusion rates are very low. Both basal plane and prismatic slip systems have been widely reported to operate during deformation of AlN over a wide range of temperatures [17–19]. f1 1% 0 0g prismatic slip planes are the easy glide planes with slip observed to occur at temperatures as low as room temperature. Basal plane slip dominates at higher temperatures. This is consistent with our observations. The impurity material responsible for inclusion formation and dislocation decoration is most likely to be oxygen since AlN has a high affinity for oxygen [6]. Further studies are underway to confirm this. Overall dislocation density is about 800–1000/ cm2 and dislocation density in the peripheral regions is slightly higher (1.2–1.4 103/cm2). Inclusion density is about 1–1.2 105/cm3 in the periphery and about 5.8–7 105/cm3 in the central regions. The distribution of dislocations and inclusions is inhomogeneous with nearly 40% of the wafer area free from dislocations and with a very low density (o 103/cm3) of inclusions.
250
B. Raghothamachar et al. / Journal of Crystal Growth 250 (2003) 244–250
5. Conclusions
References
Using the sublimation–recondensation growth technique, high-quality AlN single crystals have been grown. Dislocation densities of the order of 800–1000/cm2 and FWHM widths of 9–12 arcsec have been observed for these crystals. Synchrotron white beam X-ray topography in conjunction with optical microscopy has been successfully used to characterize the defect type and distributions. Impurities (most likely oxygen) decorate dislocations near growth temperatures and also form inclusions. Stresses due to differential thermal expansion of crystal and crucible cause significant deformation chiefly in the peripheral regions. f1 1 0 0g prismatic slip is favored at lower temperatures while basal plane slip systems are dominant at high temperatures. At lower temperatures, these stresses also cause cracking which severely limits the size of AlN substrates obtained from bulk crystals.
[1] K. Hiramatsu, A. Usui, Selective area growth and epitaxial lateral overgrowth of GaN, EMIS Datareview Series No.23. INSPEC, The Institution of Electrical Engineers, London, 1999, p. 440. [2] J.A. Van Vecten, Phys. Rev. B 7 (1973) 1479. [3] I. Grzegory, J. Jun, M. Bo!ckowski, St. Krukowski, M. ! Wroblewski, B. Lucznik, S. Porowski, J. Phys. Chem. Solids 56 (1995) 639. [4] M. Bockowski, M. Wroblewski, B. Lucznik, I. Grzegory, Mater. Sci. Semicond. Process. 4 (2001) 543. [5] R. Dwiliski, R. Doradziski, J. Garczyski, L. Sierzputowski, M. Palczewska, A. Wysmolek, M. Kamiska, MRS Internet J. Nitride Semicond. Res 3 (25) (1998) 1. [6] G.A. Slack, T.F. McNelly, J. Crystal Growth 34 (1976) 263. [7] M. Tanaka, S. Nakahata, K. Sogabe, H. Nakata, M. Tobioka, Jpn. J. Appl. Phys. 36 (1997) L1062. [8] R. Schlesser, R. Dalmau, Z. Sitar, J. Crystal Growth 241 (2002) 416. [9] J.C. Rojo, L.J. Schowalter, G. Slack, K. Morgan, J. Barani, S. Schujman, S. Biswas, B. Raghothamachar, M. Dudley, M. Shur, R. Gaska, N.M. Johnson, M. Kneissel, MRS Symposium Proceedings V 722, 2002, p. K1.1. [10] M. Dudley, in: D. Bloor, R. Brook, M. Flemings, S. Mahajan, R. Cahn (Eds.), Encyclopedia of Advanced Materials, Vol. 4, Pergamon Press, New York, 1994, p. 2950. [11] B.K. Tanner, X-ray Diffraction Topography, Pergamon, Oxford, 1976, p. 82. [12] J. Weertman, Phys. Rev. 107 (1957) 1259. [13] D.K. Bowen, B.K. Tanner, High Resolution X-ray Diffractometry and Topography, Taylor & Francis, London, 1998, pp. 85, 101. [14] W.C. Dash, J. Appl. Phys. 27 (1956) 1193; W.C. Dash, J. Appl. Phys. 29 (1958) 228; W.C. Dash, J. Appl. Phys. 30 (1959) 459. [15] G.D. Miles, J. Appl. Phys. 36 (1965) 1471. [16] H.J. Levinstein, C.D. Capio, J. Appl. Phys. 38 (1967) 2761. [17] A. Seifert, A. Berger, W.F. Muller, . J. Am. Ceram. Soc. 75 (1992) 873. [18] V. Audurier, J.L. Demenet, J. Rabier, Philos. Mag. A 77 (1998) 825. [19] M. Azzaz, J.P. Michel, V. Feregotto, A. George, Mater. Sci. Eng. B 71 (2000) 30.
Acknowledgements This work has been partially supported by the Office of Naval Research and the Missile Defense Agency (formerly BMDO). The authors would gratefully like to acknowledge the support of our contract monitors Dr. C.E.C. Wood (ONR) and Dr. C. Litton (AFRL). Topography was carried out at the Stony Brook Synchrotron Topography Facility, beamline X-19C, at the National Synchrotron Light Source, at Brookhaven National Laboratory, which is supported by the Department of Energy.