-Al2O3)
of (00·1) orientation is commonly used as substrate for MOCVD epitaxy of
AlxGa1-xN.
The influence of plasma and thermal treatments on the structure and composition of sapphire (00·1) surfaces have been studied by hemispherically recorded x-ray photoelectron spectroscopy in view of substrate preparation for the epitaxy of GaN. Producing well-ordered surfaces, O2 plasma based treatments are found to efficiently remove surface contamination. AlN films with good short-range order are obtained by a simple high temperature nitridation step in the MOCVD reactor.
Presently, much effort is directed towards the technological
developments regarding the direct wide band gap material system
AlxGa1-xN and
InxGa1-xN1,2. Despite the relatively large
lattice mismatch of 16% to GaN, sapphire (-Al2O3)
of (00·1) orientation is commonly used as substrate for MOCVD epitaxy of
AlxGa1-xN.
High quality growth of GaN requires the formation of a buffer layer which is typically grown at low temperature and subsequently annealed at high temperature3. Surface conditioning of the sapphire substrate prior to buffer layer growth is found to have an essential influence on the crystalline quality of the final GaN epilayer grown on top of the buffer4. Surface conditioning is typically performed in an initial high temperature step and may involve a simple initial heat treatment at 1100°C to desorb surface contaminants3 or a short exposure to a N2 plasma5 or an NH34,6,7 environment. The treatment with nitrogen has been reported to result in a change of the surface structure tentatively ascribed to the transformation of a surface layer into AlN5. A nitridation by conversion of the substrate potentially bears the advantage to form a smooth initial layer7 of improved lattice match by a quasi-epitaxial chemical reaction. Starting from such a surface epilayer growth is facilitated to proceed in a two-dimensional mode, since film nucleation can start from lattice matched crystal regions of well defined orientation.
Since substrate preparation is known to have an important influence on the quality of the resulting epitaxial layers, we performed a systematic investigation of sapphire surfaces. In view of MOCVD of GaN epilayers the objective of this investigation was to set up processing recipes regarding 1.) the preparation of a clean and well ordered substrate surface and 2.) the "in situ" nitridation of Al2O3 surfaces.
The single crystalline (00.1) oriented sapphire wafers were, as received, polished to an epitaxial grade finish. The samples could be cleaned in a plasma reactor chamber equipped with an ASTEX compact ECR source and transferred under UHV condition (< 1 x 10-7 Pa) to an attached VG ESCALAB MK II surface analysis facility equipped with angle-resolved x-ray photoelectron spectroscopy (ARXPS) capabilities. The measured ARXPS raw data were evaluated in view of x-ray photoelectron diffraction (XPD) effects ("forward-focusing" enhancement and higher order interference) as well as in view of the surface and subsurface composition of the examined samples. The depth profile analysis, in principle, was based on the common concept of a linear relation between the escape depth and the cosine of the polar escape angle of the recorded photoelectrons. Details on the more involved evaluation techniques used for the quantitative assessment of the depth profiles have been published elsewhere8. XPD was used to probe short-range order in a subsurface volume extending from the surface to a depth of a few 10 Å. For the detailed structural analysis a novel direct crystallographic method was employed called CHRISDA8 (Combined Holographic Real space Imaging by Superimposed Dimer functions Algorithm).
Polished sapphire surfaces were typically found to be covered with a contamination layer which had a thickness of 7-10 Å and consisted predominantly of carbonaceous species. Frequently, in addition impurities of Ca and F were present on polished sapphire surfaces, whereas oxygen was generally absent in these contamination layers.
The effects of various surface pretreatments on the surface composition of
sapphire are summarized in Fig. 1 which shows compositional depth profiles
obtained by analysis of the polar angle dependent ARXPS signal
contributions8. On the given exponential scale compositional ARXPS
profiles are quantitative, though (except for special cases) the respective
depth resolution is relatively poor9. In Fig.1, the concentration
profiles of Al (,N) and O have been expressed as Al2O3
(and AlN) mole fractions. The depth unit in the profiles is the escape length
ie (
15-25Å).
Fig. 1 a shows a typical depth profile for a polished sapphire (00·1) surface. Within the experimental error the contamination forms a closed layer on the Al2O3 substrate. The presence of surface roughness can not be unambiguously detected by ARXPS, since in the framework of our simple model the prevailing effect of surface roughness is to change the result in the ARXPS profile in a sense that two actually distinct compositional zones separated by an abrupt interface will erroneously show considerable intermixing. Hence, the profile of Fig. 1a is a demonstration of the remarkable smoothness reported10,11 for sapphire surfaces (00·1). Evidence that polishing does only little damage to the near surface region of sapphire (00·1) was found by the observation of XPD patterns with high anisotropy contrast which indicated a good crystalline order in the uppermost 20 Å of the substrate.
The efficiency of cleaning and the detrimental effects on the crystalline
surface structure upon plasma exposure of the sapphire samples were
investigated for an O2 plasma as well as for an H2/Ar
plasma. Both plasma treatments reduced the hydrocarbon and flourine
contamination below the detection limit (0.1 ML). Calcium could not be
removed by either one of the two plasma treatments.
Fig. 3 : Hemispherical diffraction diagrams of (a) the Al 2p and (b) the O1s
photoelectrons obtained for a clean sapphire substrate with MgK
excitation. To reflect the anisotropy contrast the patterns are normalized with
respect to the averaged polar angle dependent signal.
Differences between the oxygen and the hydrogen plasma based cleaning procedures affecting the resulting surface were observed in two respects, namely, the crystalline order and the surface termination. After O2 exposure (Fig. 1b) the composition was found to be uniform over the entire probed depth region and to correspond to the nominal composition. The XPD pattern of the Al2p and the O1s photoelectrons (Fig.2) showed a good anisotropy contrast indicating that the near-range order in the surface near region was not impaired by the plasma exposure. In contrast, with an H2/Ar plasma the XPD anisotropy contrast was reduced by 40 % upon plasma exposure and Ar2p photoelctrons could be detected. Hence, the observed crystal damage has to be attributed to the implantation of Ar in the sapphire matrix implying that ions accelerated by the plasma potential of the ECR source ( <=20 eV) carry sufficient energy to displace aluminum or oxygen atoms from their regular lattice sites. Upon sputtering with 1 keV Ar ions a complete loss of short-range order was indicated by XPD even for low dose exposure ( <=1015 cm-2) . An amorphous surface region12 extending over more than 50 Å forms under these conditions and surface cleaning was found to be very unefficient owing to considerable intermixing effects in this region.
H2/Ar and O2 plasma treatments result in a different termination of the sapphire surfaces. By interaction with hydrogen, the sapphire surfaces depleted slightly of oxygen and became terminated by Al atoms (Fig. 1c). Very similar depth profiles were found when oxygen cleaned surfaces (Fig. 1b shows an equal O and Al occupation of surface sites) were annealed at 900°C for a period of 30 min (Fig.1d). A surface depletion of oxygen upon heat treatments of sapphire has also been reported by Gautier et.al.12. At 1400°C further oxygen depletion leads to the formation of an ordered double layer of metallic aluminum10,12. The (non-conducting) Al terminated surface of sapphire (00·1) has been also predicted by theory to be the stable one13. The amount of Al found for the depth profile of Fig. 1c and d in the surface layer corresponds only two half a monolayer and possibly indicates the presence of surface vacancies. Harrison14 has pointed out that polar surfaces are likely to be stabilized by surface vacancies to avoid dielectric breakdown by an internal field. To eliminate the internal field, the surface terminating Al layer terminating the (00.1) surface of sapphire has to be filled half by surface vacancies, in agreement with the profiles Fig. 1c and d.
On a first view, nitridation of the sapphire surface to obtain AlN nuclei appears to be impossible. On the contrary, AlN decomposes in contact with oxygen according to
Al2O3 + N2 => 2 AlN+ 3/2
O2 ;  G298=+10.4 eV per formula unit | (1) |
The energetic situation is improved in favor of AlN formation if atomic rather than molecular nitrogen can be employed. For the nitridation reaction
| Al2O3 + 2 N => 2 AlN+ 1/2 O2 ;
G298=+1.0 eV p.f.u. | (2) |
the equilibrium is strongly shifted to the right hand side if a suitable reactant, such as hydrogen, is provided for the freed oxygen
| Al2O3 + 2 N + 3 H2 => 2 AlN+ 3
H2O ; G298= -6.1 eV p.f.u. | (3) |
Hence, an H2/N2 plasma which provides atomic nitrogen may serve as a suitable ambient for the nitridation of sapphire at room temperature. In our experiments, AlN layers of about 6 to 12 Å thickness were formed on sapphire samples upon exposure to an N2/H2 plasma. These layers were relatively uniform but extremely prone to reoxidation (Fig. 1e). By the absence of XPD anisotropy for the N1s signal it was evident that nitride layers obtained by plasma exposure are amorphous. The poor crystallinity of the AlN films and their sensitivity to reoxidation upon exposure to air calls in question if ex situ plasma nitridation is a substrate preparation adequate for GaN epitaxy.
A simple possibility for surface nitridation inside the MOCVD reactor is offered by the exposure of the sapphire surface to NH3 in a high temperature step preceding layer growth. At room temperature no AlN can be formed by the reaction
| Al2O3 + 2 NH3 => 2 AlN+ 3
H2O ; G298= 3.68 eV p.f.u. | (4) |
G of formation is reduced to about 2.0 eV p.f.u. at 1100°C
(kT120 meV). If we assume for the MOCVD reactor under NH3 flow
that the water pressure does not rise above 10-6 bar at the surface
of the sapphire substrate then AlN formation is favored by reaction (5) downto
a NH3 partial pressure as low as 10-5 bar.
The result of the ARXPS analysis demonstrates that nitridation of sapphire by
exposure to NH3 is feasible at high temperature. The sapphire sample
analysed in Fig. 2f had been exposed in the MOCVD reactor to NH3 for
3 min (T=1100°C, partial pressure>40 mbar, carrier gas N2)
and subsequently transferred into the UHV system within 10 minutes. An AlN
layer of thickness 20-30 Å was found to be formed upon this treatment (at
least on a large part of the surface). The respective ARXPS depth profile
analysis (Fig. 2f) shows an AlN layer which is not closed and is consistent
with the presence of AlN islands which cover about one third of the sample.The
formation of AlN islands as a result of a high temperature NH3
exposure has also been reported in the literature7.
Fig.3: Anisotropy contrast patterns of the (a) the Al 2p and (b) the O1s and (c) the N1s photoelectrons recorded after nitridation of a clean sapphire surface. (d) is the pattern reproduction obtained upon CHRISDA analysis of (c).
Fig.4: Occupation probability profiles obtained by CHRISDA analysis of the N1s
pattern of Fig. 3 c. (a) analysis based on the shown emitter configurations
The coincidence of the oxygen sublattice and the AlN layer lattice is
confirmed in Fig 4a by the peak for configuration
Hence, the epitaxial relationship is confimed to be [21.0]AlN ||
[11.0]Al2O3 as previously suggested by Yamamoto
et.al.6. However, the observed termination of the AlN
layer disagrees with the prediction of Kung15.
Using ARXPS as a technique providing depth compositional as well as
structural information of the surface near region we have studied the effect of
plasma treatments on the composition, termination and structure of sapphire
surfaces with (00·1) orientation. O2 as well as
H2/Ar based plasma treatments are shown to adequately remove surface
contamination. Short-range order is preserved in both cases, however, a higher
grade of the surface crystallinity is obtained with the oxygen
treatment.Thermal or H2 plasma treatments of sapphire are found to
result in an Al termination of the sapphire surface.We have shown that sapphire
substrates can be nitrided by either a plasma process or by NH3
exposure at high temperatures in the MOCVD reactor, however, only the latter
process leads to the formation of a crystalline AlN phase. This phase is
identified to be of the wurtzite type. The AlN layer is found to be formed in
registry with the oxygen sublattice and to be terminated by nitrogen atoms. The
AlN has a fully relaxed structure and tends to conglomerate in extended
crystallites rather than forming a closed thin film.
Y. Sugimoto, Appl. Phys. Lett. 68,2105 (1996)
[2] B.W.Lim,Q.C.Chen,J.Y.Yang, M.A.Khan, Appl. Phys. Lett. 68, 3761
(1996)
[3] C.F. Lin, G.C. Chi, M.S. Feng, J.D.Guo, J.S.Tsang, J.Minghuang
Hong,
Appl. Phys. Lett. 68,3758 (1996)
[4] S. Keller, B.P. Kemmer, Y.-F. Wu, B. Heyring, D. Kapolneck, J. S.
Speck, U.K. Mishra, S.P. Denbaars, Appl. Phys. Lett. 68, 1525
(1996)
[5] R.J. Molnar, T.D. Moustakas, J. Appl. Phys. 76,4587
(1994)
[6] A. Yamamoto, M. Tsujino, M. Ohkubo, A. Hashimoto, J. Cryst.
Growth 137, 415 (1994)
[7] K. Uchida, A. Watanabe, F. Yano, M. Kouguchi, T. Tanaka, S.
Minagawa, J. Appl.
Phys. 79, 3487 (1996)
[8] M. Seelmann-Eggebert, G.P. Carey, R. Klauser, H.J. Richter, Surf.
Sci. 287/288,495
(1993); M. Seelmann-Eggebert, accepted for publication in Surf.
Sci.
[9] M. Seelmann-Eggebert, R.C.Keller, Surf. and Interf. Anal. 23,589
(1995)
[10] G. Renaud, B. Vilette, I. Vilfan, A. Bourret, Phys. Rev. Letters
73,1825 (1994)
[11] Yan Yu, R.J. Lad, Mat. Res. Soc. Symp. Proc. 317,583
(1994)
[12] M. Gautier, J.P. Durand, L. Pham Van , M. J. Guittet, Surf. Sci.
250, 71 (1991)
[13] J. Guo, D.E. Ellis, D.J. Lam, Phys. Rev. B 45,13647
(1992)
[14] W.A. Harrison, J. Vac. Sci. Technol. 16, 1492
(1979)
[15] P. Kung, C.J. Sun, A. Saxler, H. Ohsato, M. Razeghi,
J. Appl. Phys. 75,4515 (1994)
© 1997 The Materials Research Society
,
,
. (b) Nominal joint occupation density of an hexagonal
AlN (00·1) surface for emission from nitrogen atoms. The surface is
assumed to be terminated by nitrogen atoms at d=0. The nearest
plane above the N emitters is found for scatterers in top sites at a distance
of about 1.9 Å. Up to a distance of about 6 Å (beyond which a
determination of layers becomes unreliable owing to the noise in the pattern of
Fig. 3c) there is good agreement between the OPPs of Fig. 4a and Fig. 4b,
indicating that the AlN is fully relaxed. In particular, the absence of a
significant response at a distance of 0.6 Å confirms that the AlN is
(predominantly) of nitrogen termination. For d<=5 Å similar layer
images would also be expected if twinned regions of cubic AlN (111) were
analyzed. However, with the OPPs of Fig. 4a the presence of the hexagonal phase
becomes unambiguous by a response at about 5 Å for configuration .
CONCLUSIONS
REFERENCES
[1] S. Nakamura, M. Senoh, S.Nagahama, N. Iwasa, T. Ymada, T.
Matsushita, H. Kiyoku,
last updated Tuesday, April 22, 1997 3:23:53 PM.