ARTICLE IN PRESS
Journal of Crystal Growth 265 (2004) 577–581
Natural growth habit of bulk AlN crystals B.M. Epelbauma,*, C. Seitzb, A. Magerlb, M. Bickermanna, A. Winnackera b
a Department of Materials Science 6, University of Erlangen-Nurnberg, Martensstr. 7, D-91058 Erlangen, Germany . Department of crystallography and structural physics, University of Erlangen-Nurnberg, Bismarckstr. 10, D-91054 Erlangen, Germany .
Received 23 January 2004; accepted 24 February 2004 Communicated by L.F. Schneemeyer
Abstract Growth conditions for self-nucleation and subsequent growth of bulk AlN crystals by sublimation are presented. With increasing growth temperature, the natural habit of AlN crystals changes from needle-like to prismatic and then turns to thick asymmetric platelet. The best-formed platelet crystals up to 14 7 2 mm3 in size exhibit a number of atomically smooth surfaces. Growth morphology and crystal quality were found to be strongly inﬂuenced by the polar nature of AlN. Al-terminated faces produce mirror-like facets and transparent material of high crystalline quality, whereas development of N-terminated faces leads to opaque and defective sectors in grown crystals. It is suggested that the most successful seeded growth of AlN can be achieved along Al-terminated ð0 0 0 1Þ; ð1% 0 1 2Þ and non-polar ð1% 0 1 0Þ faces. r 2004 Elsevier B.V. All rights reserved. PACS: 81.05.Ea; 81.10.Aj; 81.10.Bk Keywords: A1. Crystal morphology; A2. Growth from vapor; B1. Nitrides; B2. Semiconducting aluminum compounds
1. Introduction Breakthrough developments in GaN semiconductor technology have triggered a great number of worldwide research activities on GaN-based light emitting devices. In particular, compact ultraviolet light sources are currently under development for applications in solid-state lighting, short-range communication, and bio-chemical detection. GaAlN-based laser diodes hold promise *Corresponding author. Tel.: +49-9131-8527634; fax: 499131-8528495. E-mail address: [email protected]
for numerous applications such as optical data storage and printing. Bulk single-crystalline AlN is a very attractive substrate material for III-nitride homoepitaxy, superior to sapphire and silicon carbide dominantly used at present. However, the potential of AlN in substrate applications has been disadvantaged by the lack of sufﬁciently large and perfect single crystals. Despite a number of crystal growth studies on AlN [1–4] published since the pioneering work of Slack and McNelly , the conditions necessary for the formation of perfect bulk crystals remain uncertain. The target of our research was the preparation of self-nucleated AlN crystals grown with minimal contact with crucible walls or other crystals and
0022-0248/$ - see front matter r 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.jcrysgro.2004.02.100
ARTICLE IN PRESS 578
B.M. Epelbaum et al. / Journal of Crystal Growth 265 (2004) 577–581
the investigation of their growth habit in relation to growth conditions. The identiﬁcation of the most stable (slowly growing) faces in actual growth conditions is very important since they are favorable for high-quality seeded growth. One prominent example is the development of SiC crystal growth technique from the vapor phase, which was ﬁrst elaborated by Lely  for the preparation of self-standing SiC platelets and later extended into seeded physical vapor transport (PVT) growth of large-diameter crystals .
2. Growth procedures and results Crystal growth was accomplished by sublimation of the AlN charge placed in the hot zone of a tungsten crucible and subsequent condensation of vapor species in a cooler region. Our growth procedure was very similar to that of Slack and MacNelly , except that they were aimed at growing a polycrystalline AlN body inside a conical crucible bottom, relied on grain selection and made use of a considerable temperature gradient between the growing crystal and the source material. Our arrangement, however, was made to facilitate self-nucleation of separate single crystals on the crucible walls and on the perforated diaphragm in between the evaporation and condensation zones heated separately, see Fig. 1. To
Fig. 1. Growth cell used for growth of AlN crystals by sublimation.
achieve this, the proﬁle of the upper heating element was altered in such a way that the temperature gradient over the condensation section of the crucible above the diaphragm was reduced to the lowest possible value. Other, less essential, deviations from the procedure of Slack and MacNelly were (i) vertical positioning of the growth crucible (similar to well-established growth of SiC crystals by sublimation ) and (ii) employment of two heating zones with independent thermal control instead of crucible withdrawal from a single heater. The starting material was AlN powder containing about 0.5 wt% of Al2O3 supplied by Chempur, Germany. Prior to use for crystal growth, the powder was resublimed in the same growth cell as shown in Fig. 1 in an atmosphere of high-purity nitrogen, but without setting up the diaphragm. During resublimation, the powder was kept at about 2300 C. The crucible lid where the material was condensed was kept at 2100 C. The charge of 200 g was normally completely sublimed during 24 h of holding time. Sublimation product collected on the crucible lid comprised two very different portions: a white layer less than 1 mm in thickness produced by small crystals of micrometer size and generated in the beginning of the sublimation transport process, was followed by a brown colored dense polycrystalline mass consisting of imperfect crystals up to 3–5 mm in size. The white product was identiﬁed by powder XRD analysis as a mixture of approximately 10% of aluminum oxinitride Al3O3N (JCPDS 36-0050 standard ﬁle) and AlN (JCPDS 25-1133 standard ﬁle), and the brown product as pure AlN. The white layer was formed at about 1700–1900 C presumably because of decomposition of Al2O3 in nitrogen leading to volatilization of aluminum in the form of Al2O. Condensation of Al2O is excess of nitrogen on the colder crucible lid yields ﬁne-particulated layer of AlN and Al3O3N mixture. After the oxygen in the system is exhausted, this reaction-induced process stops and transport starts again at temperatures above 2000 C already due to evaporation of pure AlN in PVT process. The white layer was mechanically removed from the main portion of resublimed bulk product before it was used for the subsequent growth process.
ARTICLE IN PRESS B.M. Epelbaum et al. / Journal of Crystal Growth 265 (2004) 577–581
The charge of about 200 g was placed in the lower-evaporation section of the tungsten crucible below the diaphragm. Three series of growth experiments were made in which the growth temperature in the condensation section of the crucible was set at 2050 C, 2150 C and 2250 C. In all experiments the source temperature was 75–100 C higher than growth temperature. The accuracy of temperature assessment may be estimated as +/25 C, since direct measurement using a calibrated optical pyrometer was possible only at the bottom of the cooling channel, Fig. 1, and the above values were obtained through the utilization of numeric codes for simulation of thermal ﬁelds [8,9]. The furnace was operated in high-purity N2 gas at pressures below 1000 mbar; holding time was 24–48 h depending on the temperature range. The temperature gradient along crucible walls in the condensation room after installation of the diaphragm was as low as 3–5 C/cm. The low value of the temperature gradient manifested itself in a small thickness (1–3 mm only) of the dense AlN layer deposited in the middle of the crucible lid representing the coldest point in the crucible, Fig. 1. After bringing the system to room temperature, spontaneously nucleated AlN crystals were collected on the crucible walls and on the diaphragm. Fig. 2A is a plot of the typical size and morphology of the crystals grown in each series as a function of growth temperature. Crystals grown at 2050 C were mainly in the form of sixsided prismatic needles 0.1–0.3 mm in diameter
and 5–15 mm in length. Very few elongated plates 0.5–1.0 mm wide and 5–7 mm long having the particular appearance described in detail by Pastrnak and Roskovcova  were also found. All these crystals were transparent and almost colorless with light yellow tint. Total mass yield of needle-like crystals was 1–2 g after 48 h of growth. Crystals grown at 2150 C were typically columnar and only slightly extended along the ½0 0 0 1 axis. The largest crystals grown at 2150 C were up to 3 mm in length and 1.5 mm in diameter. The bestformed, translucent crystals were dark yellow in color and show mainly rhombohedron (presumably f1 0 1% 1g and f1 0 1% 2g) and prismatic f1 0 1% 0g faces. Basal f0 0 0 1g faces were also present, but they were usually smaller than rhombohedron faces. Crystals of needle-like habit typical for lower-temperature growth were also occasionally observed at 2150 C. Some of them had signs of thermal etching. Crystal yield was about 2–3 g after 24 h of growth. The amount of transported material and the size of AlN crystals were found to increase drastically at 2250 . At this growth temperature up to 70 g of dark-amber or brownish crystals were produced within 24 h. The largest crystal clusters were collected in the center of perforated diaphragm, Fig. 2B. Typical thick platelet crystals grown at 2250 are shown in Fig. 3A. They were normally grown with the largest dimension parallel to the direction of heat ﬂow, i.e. perpendicular to the diaphragm or to the crucible wall. In many investigated crystals platelet thickness varied from 1 to 3 mm, but habit facets
Fig. 2. (A) Crystal habit and dimension versus growth temperature, and (B) cluster of AlN crystals up to 15 mm in length, grid spacing is 5 mm.
ARTICLE IN PRESS 580
B.M. Epelbaum et al. / Journal of Crystal Growth 265 (2004) 577–581
Fig. 3. (A) Selected crystals grown at 2250 C showing typical growth habit, grid spacing is 1 mm, and (B) characteristic growth faces of best-formed AlN platelets.
governing the asymmetric appearance shown in Fig. 3A were omnipresent. The largest crystal measured 14 7 2 mm3 on the sides. The largest opposite surfaces of characteristic platelets (parallel to the main platelet plane) were somewhat different, one being ﬂat and another stepped. Elongated sides perpendicular to platelet planes were always strongly different. From one side (left in Fig. 3B), the platelet is bounded by a number of mirror-like facets; from the other side (right in Fig. 3B) only one smooth, slightly curved face was observed. The part of the crystal 0.5–1.0 mm in depth under this face was opaque because of the presence of micrometer-sized inclusions. With the exception of this defective area, all grown crystals were transparent without any visible defects, thus we will deﬁne the facet as ‘imperfect’. Singlecrystalline platelets of this size are already large enough to be employed for seeded PVT growth; because of this reason their orientation and polarity were investigated in detail.
3. Identiﬁcation of faces orientation and polarity First, a backscattering Laue pattern of one major facet was recorded. Indexing was performed with the software OrientExpress . The lattice ( for the a- and parameters used were 3.104 A ( for the c-parameter . As the error in 4.965 A
crystal alignment was below 0.1 the indexing is unambiguous. Indexing gives not only the orientation of the investigated facet, but also the complete crystal orientation matrix. The ½0 0 0 1 axis appears to lay within the platelet plane perpendicular to the elongated platelet side (i.e. to the imperfect facet). In a next step, the crystal was transferred onto an optical two-circle reﬂection goniometer, without demounting it from the goniometer head. The angular error introduced by the transfer is estimated to be less than 1 . Thus the angular positions from the reﬂection goniometer are compatible with the crystal orientation matrix obtained by the Laue pattern. For all crystal facets the angular positions were measured. The facets were classiﬁed into three classes depending on the optical reﬂection quality. Facets assigned with one in Table 1 are mirror-like. A lower quality results from rough or stepped faces. Subsequently, the stereographic projection of the crystal was simulated with OrientExpress using the crystal orientation matrix found in the ﬁrst step by indexing the Laue pattern. The two rotation angles of the goniometer can be identiﬁed as two rotations of the simulated crystal in OrientExpress. Rotating the simulated crystal accordingly to the angles obtained by the reﬂection goniometer results in a stereographic projection of the speciﬁc facet. The results are summarized in Table 1.
ARTICLE IN PRESS B.M. Epelbaum et al. / Journal of Crystal Growth 265 (2004) 577–581 Table 1 Indexing of the most pronounced facets in AlN crystals grown at 2250 C
1 2 3 4 5 6 7 8 9
1 0 1% 0 1 0 1% 1 1 0 1% 2 1 0 1% 3 0001 1% 0 1 3 1% 0 1 2 1% 0 1 0 0 0 0 1%
2 2 1 2 1 2 2 3 3
of AlN. Al-terminated ð0 0 0 1Þ and positive rhombohedron faces produce mirror-like facets and transparent material of high crystalline quality, but the development of N-terminated ð0 0 0 1% Þ face leads to opaque and defective sectors in grown crystal. We suggest that the most successful seeded growth of AlN can be achieved at growth temperatures exceeding 2200 C and along Al-terminated ð0 0 0 1Þ; ð1% 0 1 2Þ and nonpolar ð1% 010Þ faces.
Acknowledgements Crystal polarity was identiﬁed by etching in an eutectic KOH/NaOH mixture at a temperature of 250 C for 3 min. We found that the mirror-like ð0 0 0 1Þ surface as well as adjacent rhombohedron facets are resistant to a 250 C molten NaOH/ KOH etchant, whereas the ‘imperfect’ ð0 0 0 1% Þ surface is chemically very active. After 3 min of etching it appears totally rough composed by hexagonal hillocks 20–50 mm in size. By analogy with GaN , we may assume that the positive direction of ½0 0 0 1 is pointing towards the ‘perfect’ ð0 0 0 1Þ facet, hence this facet is Alterminated, Fig. 3B. This conclusion is in line with polarity studies on AlN crystals published by Schowalter et al. . Faces listed in Table 1 are indexed taking this polarity calibration into account. In conclusion, growth conditions for selfnucleation and subsequent growth of bulk AlN crystals by sublimation have been established. With increasing growth temperature, the natural habit of AlN crystals changes from needle-like to prismatic and then to thick asymmetric platelet. The best-formed platelet crystals up to 14 7 2 mm3 in size exhibit a number of atomically smooth surfaces. At a growth temperature of 2250 C relevant for production of bulk AlN crystals, the growth habit and crystal quality were found to be strongly inﬂuenced by the polar nature
This work has been partially supported by the Deutsche Forschungsgemeinschaft (DFG) under contract Wi-393/13.
References  L.J. Schowalter, J.C. Rojo, N. Yakovlev, Yu. Shusterman, K. Dovidenko, R. Wang, I. Bhat, G.A. Slack, MRS Internet J. Nitride Semicond. Res. 5S1 (2000) W6.7.  J.H. Edgar, L. Liu, B. Liu, D. Zhuang, J. Chaudhuri, M. Kuball, S. Rajasinggam, J. Crystal Growth 246 (2002) 187.  R. Schlesser, R. Dalmau, Z. Sitar, J. Crystal Growth 241 (2002) 416.  M. Tanaka, S. Nakahata, K. Sogabe, H. Nakata, M. Tobioka, Jpn. J. Appl. Phys. 36 (1997) 1062.  G.A. Slack, T.F. McNelly, J. Crystal Growth 34 (1976) 263.  J.A. Lely, Ber. Dtsch. Keram. Ges. 8 (1955) 229.  Yu.M. Tairov, V.F. Tsvetkov, J. Crystal Growth 43 (1978) 209.  M.V. Bogdanov, A.O. Galyukov, S.Yu. Karpov, A.V. Kulik, S.K. Kochuguev, D.Kh. Ofengeim, A.V. Tsiryulnikov, M.S. Ramm, A.I. Zhmakin, Yu.N. Makarov, J. Crystal Growth 225 (2001) 307.  http://servww6.ww.uni-erlangen.de/cgl/crysvun.  J. Pastrnak, L. Roskovcova, Phys. Stat. Sol. 7 (1964) 331.  http://www.ccp14.ac.uk/ccp/web-mirrors/lmgp-laugier-bochu.  J.D.H. Donnay, Crystal Data, American Crystallographic Association, 1963.  J.L. Rouviere, J.L. Weyher, M. Seelmann-Eggebert, S. Porowski, Appl. Phys. Lett. 73 (1998) 668.