Surface reconstruction patterns of GaN grown by molecular beam epitaxy on GaN bulk crystals

Surface reconstruction patterns of GaN grown by molecular beam epitaxy on GaN bulk crystals

Journal of Crystal Growth 207 (1999) 1}7 Surface reconstruction patterns of GaN grown by molecular beam epitaxy on GaN bulk crystals C.T. Foxon!,*, T...

198KB Sizes 2 Downloads 50 Views

Journal of Crystal Growth 207 (1999) 1}7

Surface reconstruction patterns of GaN grown by molecular beam epitaxy on GaN bulk crystals C.T. Foxon!,*, T.S. Cheng!, S.V. Novikov!,", D. Korakakis!, N.J. Je!s!, I. Grzegory#, S. Porowski# !School of Physics and Astronomy, University of Nottingham, University of Park, Nottingham NG7 2RD, UK "Iowe Physical-Technical Institute, St. Petersburg, Russia #High Pressure Research Center UNIPRESS, Warsaw, Poland Received 26 November 1998; accepted 22 June 1999 Comminicated by J.B. Mullin

Abstract An investigation of the conditions giving rise to surface reconstruction for GaN grown by molecular beam epitaxy on (0 0 0 11 ) bulk GaN substrates has been carried out. The results of di!erent surface reconstructions were investigated by supplying Ga and active nitrogen separately to GaN surfaces. A (2]2) reconstruction on (0 0 0 11 )GaN substrates and for GaN layers grown on such substrates was observed. The (2]2) surface reconstruction is stable in the presence of an active nitrogen #ux at high temperature, but it disappears on cooling the GaN sample below &4003C. The (2]2) reconstruction is considered to be due to an additional amount of relatively tightly bound Ga on this (1]1) surface. Adding further Ga to the (2]2) reconstructed surface resulted in a return to a (1]1) reconstruction. ( 1999 Elsevier Science B.V. All rights reserved. PACS: 81.15.Hi; 81.05.Ea Keywords: Gallium nitride; Molecular beam epitaxy; Surface reconstruction

1. Introduction The group-III nitrides are now widely used as novel materials for light-emitting devices. This class of materials also has potential applications in the

* Corresponding author. Tel.: #44-115-951-5164; fax: #44115-951-5184. E-mail address: [email protected] (C.T. Foxon)

high-power, high-temperature electronic device area. At present, the precise growth mechanisms responsible for the growth of the III-nitrides is still not well understood. Amongst the growth methods for III-nitrides, molecular beam epitaxy (MBE) has so far provided the most insight into the growth kinetics because of its unique ability to provide in situ analytical measurement by re#ection high-energy electron di!raction (RHEED). Most of the nitride growths have been performed on sapphire or silicon carbide substrates because of

0022-0248/99/$ - see front matter ( 1999 Elsevier Science B.V. All rights reserved. PII: S 0 0 2 2 - 0 2 4 8 ( 9 9 ) 0 0 3 5 4 - 1

2

C.T. Foxon et al. / Journal of Crystal Growth 207 (1999) 1}7

a lack of suitable size of bulk GaN crystals. Due to the absence of inversion symmetry, the GaN substrate has a distinct polarity [1,2] which in most cases is maintained during epitaxy. Wurtzite GaN grown on GaN substrates, therefore, has two distinctly di!erent crystal directions * [0 0 0 1] and [0 0 0 11 ] corresponding to the Ga and N polarity, respectively. A variety of RHEED patterns have been seen and reported in Refs. [3}24], these include (1]1), (2]1), (2]2), (2]3), (3]2), (3]3), (4]6) and (5]5). The un-reconstructed (1]1) pattern has been shown to correspond to a monolayer of Ga which is tightly bound to the GaN [3}6]. Smith et al. envisaged a &Ga adlayer' on the GaN surface [3}6]. On top of this relatively stable Ga adlayer there are mobile Ga adatoms which give rise to the surface reconstructions. Smith et al. reported that di!erent reconstruction patterns can be produced by otherwise identical MBE growth conditions on the two di!erent polarities [3}6]. From the di!erent RHEED patterns they conclude that the N- and Ga-polarities have characteristic reconstructions. For the N-polarity, they observed (1]1), (3]3), (6]6) and c(6]12), whilst for the Ga-polarity they reported (2]2), (1]2), (5]5) and (6]4) patterns. Smooth GaN layers grown by metal-organic vapour phase epitaxy (MOVPE) on sapphire are believed to have Ga-polarity [3}6]. It was shown recently by coaxial impact collision ion scattering spectroscopy (CAICISS) for the (0 0 0 11 )GaN "lms grown by plasma-assisted MBE on nitrided sapphire substrates that the surfaces of GaN "lms grown under both N- and Ga-rich conditions are terminated with Ga atoms [8]. Recently, we have obtained atomically #at GaN layers by MBE at high growth temperatures ('7303C) on pressure-grown GaN bulk substrates [25]. RHEED studies during the MBE on the bulk GaN show intense RHEED patterns with pronounced streaks and strong Kikuchi features [25]. During growth the GaN surface is unreconstructed, but on cooling to about 400}5003C, a (2]2) reconstruction appears followed by a (4]4) at lower temperatures. Atomic force microscopy (AFM) studies con"rmed that MBE-grown GaN surfaces are atomically #at over large areas ('2]2 lm2) for growth on (0 0 0 11 ), the chemically active side of bulk GaN substrates [25].

In this paper, we report on an investigation of the growth mechanism of MBE homoepitaxial GaN layers on bulk GaN substrates by studying the RHEED reconstruction patterns. We show that all the reconstructed RHEED patterns observed are related to additional Ga adatoms on the GaN surface. Adding additional Ga to this surface results initially in a (2]2) reconstruction, which converts to a (1]1) un-reconstructed surface with additional Ga.

2. Experimental details The experiments were performed in an MBE reactor specially designed for RHEED measurements [25]. The MBE system has an Oxford Applied Research (OAR) CARS25 RF activated plasma source to provide the atomic nitrogen species required for the growth. In addition the group-III source provides elemental gallium. The growth rate of the GaN "lms used in this study was 0.25}0.30 lm/h. The nitrogen plasma source was operated at 200}450 W with a nitrogen #ow rate of a few standard cubic centimetres per minute (sccm) resulting in a system pressure of 1]10~5}2] 10~5 Torr. The Ga and N #uxes were initially adjusted to establish growth under stoichiometric conditions on a sapphire substrate at di!erent growth temperatures. The temperature of the substrate was measured by a pyrometer through a direct sight optical window and monitored by the thermocouple in the substrate holder. In order to study the surface reconstructions, thin GaN "lms were grown on pressure-grown GaN substrates. The substrates were mounted on molybdenum holders with gallium}tin alloy [26]. RHEED patterns prior to, during and after growth were monitored using a 12 kV VG LEG110 electron gun. The RHEED images were displayed on a monitor, accessed using a video capture card and stored on a PC system. The high-quality GaN substrates used in this study were grown from the Ga solutions at nitrogen pressures of 12}15 kbar and at temperatures of 1500}16003C. These pressure-grown substrates were hexagonal platelets 6}8 mm in size, highly conductive if grown from solutions in Ga or

C.T. Foxon et al. / Journal of Crystal Growth 207 (1999) 1}7

semi-insulating if doped with Mg during growth. The (0 0 0 11 ) hexagonal face corresponding to the N-polarity is chemically active, and therefore, it could be prepared for epitaxy by mechano-chemical polishing [27] to give atomically #at surfaces without sub-surface damage. All the studies reported in our paper were obtained on this polarity.

3. Results and discussion When the GaN bulk crystals were mounted on the molybdenum holder with the nitrogen polarity surface exposed for growth, a strong unreconstructed (1]1)RHEED pattern was observed immediately after loading the samples into the MBE chamber at room temperature with the electron beam along either the [2 11 11 0] or the [1 11 0 0] direction. This indicates that the surface was not covered in a thick oxide as is commonly observed in other III}V compounds. This (1]1) pattern does not change after thermal cycling in vacuum up to temperatures of &8003C. We have grown GaN layers on the bulk GaN crystals at &7503C, under slightly Ga-rich conditions. During growth, we still observed the streaky (1]1) pattern which became even stronger after the "rst minutes of epitaxy. Closing simultaneously the Ga and N shutters to terminate growth did not change the (1]1) pattern immediately. However, a few seconds after termination of the growth, a (2]1) pattern emerged and it became stronger with time. The time taken for this pattern to appear varied from experiment to experiment, but was in the range 15}50 s. This (2]1) pattern was very stable and can be maintained at &7503C for several hours. On cooling, the GaN layers grown under Ga-rich conditions showed a clear (2]2) reconstruction pattern. It could always be observed below &3003C. However, during growth at high temperature, under similar conditions it is also sometimes possible to see a (2]1) pattern. The (2]1) pattern which we observe cannot truly represent a valid reconstruction for the hexagonal lattice. We believe that this is a poorly ordered (2]2) reconstructed surface and that the two-fold periodicity is more easily seen with the electron beam oriented along the [2 11 11 0] direction. With

3

the electron beam in the [1 11 0 0] direction the surface reconstruction is more di$cult to see. This could be due in part to the narrower spacing of the di!racting planes, but may more reasonably be associated with a di!erent degree of ordering in the two directions. Closing the Ga shutter after growth at &7503C and so supplying only an active nitrogen #ux to the surface also produced a (2]1) pattern almost immediately. This pattern was stable and could be maintained at &7503C under an active nitrogen #ux for a few hours. However, during this exposure to active nitrogen, the previously streaky RHEED pattern broadened and became weaker with time. The (2]1) pattern changed to a (1]1) pattern upon cooling the GaN sample under N plasma #ux at temperatures below &4003C. On the other hand, closing only the N shutter during GaN growth at &7503C led to a slow decrease in intensity of the (1]1) pattern due to excessive coverage of the surface with gallium. The decrease in intensity is slow due to the partial re-evaporation of Ga from the surface at this temperature. As in other III}V compounds, above 6503C the re-evaporation rate of Ga becomes signi"cant. To study the mechanism responsible for the reconstruction, we have investigated the e!ects of separately supplying Ga and active nitrogen to the GaN surface, both to the GaN substrate prior to growth and to the GaN epitaxial layers. Adding a few monolayers of Ga (up to &15 ML) to the surface of the GaN bulk crystal prior to growth in vacuum without nitrogen at &6503C, did not cause any change in the (1]1)RHEED reconstruction, however, the overall intensity decreased. Then if the GaN crystal was heated to a temperature of about 7503C to evaporate the excess Ga from the surface, the (1]1) pattern remained. However, upon cooling the substrate to about 3003C, a (2]2) reconstruction was observed and further cooling resulted in a (4]4) pattern being observed as previously reported after epitaxial growth of GaN on bulk GaN [25]. We carried out similar experiments on GaN epitaxial "lms immediately after growth by MBE. As noted above, such "lms show a (2]1)RHEED pattern. Supplying additional Ga, in vacuum, without

4

C.T. Foxon et al. / Journal of Crystal Growth 207 (1999) 1}7

any nitrogen in the chamber, induced a change in the reconstruction. At &6503C, if we added less than 1 ML of Ga, the (2]1) pattern became weaker, but was still visible. After adding more than 1 ML of Ga to the surface, the (2]1) pattern changed immediately to a (1]1) pattern and this process could not be recovered at 6503C. The (2]1) pattern could, however, be restored by heating to a higher temperature (&7503C). On cooling the GaN layers, a clear (2]2) reconstruction pattern appeared (Fig. 1) and could always be observed below &3003C. Similar experiments were performed on GaN layers grown under N-rich conditions. During the growth we also observed a (1]1) pattern, however, this became spotty after the "rst few minutes of growth and remained spotty till the end of the deposition. Unlike the situation for Ga-rich growth, on closing together both the Ga and N shutters the RHEED pattern did not change and did not change upon cooling. A weak (2]1) pattern could be obtained on the surface of GaN layers grown under nitrogen-rich conditions, by supplying an additional few monolayers of Ga (up to &15 ML) without nitrogen at &6503C and then heating the sample to &7503C to evaporate the excess Ga from the surface and "nally cooling to &3003C. However, this reconstruction exists only over a narrow temperature range and if the sample is cooled to room temperature, a (1]1) pattern is observed. When interpreting the experimental data one has to remember that with the RF nitrogen plasma

source operating, even when the nitrogen shutter is closed, there is a "nite leakage of active nitrogen which reaches the substrate surface. This small nitrogen #ux can in#uence the surface reconstruction if the Ga shutter is closed; because under these conditions, the atomic nitrogen #ux reaching the substrate surface is very low. This has been studied by supplying a "xed amount of Ga to the surface of GaN layers grown under Ga-rich conditions and observing the subsequent changes in the RHEED pattern. At &6503C, as previously observed when less than 1 ML of Ga was added, the (2]1) pattern became weaker, but in this case the RHEED pattern recovered afterwards to the original (2]1) intensity. Adding 1 ML of Ga to the surface immediately changed the (2]1) pattern to a (1]1) pattern, but after &30 s the (2]1) pattern reappeared. Increasing the substrate temperature up to &7503C did not signi"cantly change the recovery time of the (2]1) pattern. To investigate this further the recovery time has been studied as a function of the amount of Ga deposited on the surface. Fig. 2 shows that the time depends linearly upon the amount of Ga deposited. To check this point we also carried out experiments with nitrogen #owing through the plasma source, but with the plasma o!. In this case, the time for recovering the (2]1) pattern after deposition of 1 ML of Ga at &7503C increased up to &10 min, in comparison with 30 s with the plasma running and the Nshutter closed. At a temperature of &6503C the (2]1) pattern did not recover even after a few hours following deposition of 1 ML of Ga, but the

Fig. 1. The (2]2)RHEED patterns observed with the electron beam along (a) the [2 11 11 0] and (b) [1 11 0 0] direction.

C.T. Foxon et al. / Journal of Crystal Growth 207 (1999) 1}7

5

Fig. 2. The relationship between the time taken for the (2]1)RHEED pattern to reappear at &7503C (recovery time) and the amount of the Ga deposited on the surface at a "xed rate (deposition time).

pattern can be recovered to the (2]1) structure after heating to higher temperatures. From the above data we can conclude that the streaky (2]1) pattern observed on the surface of MBE GaN layers is very stable at high temperatures in the range &650}7503C and can be maintained under a #ux of active nitrogen for a long period of time, even up to a few hours. The (2]1) pattern can be recovered after Ga deposition experiments by two main mechanisms, both of which lead to the removal of the excess Ga from the surface. The "rst one is a slow process of Ga thermal desorption from the surface, which depends strongly on the substrate temperature. The second fast process, which exists when the RF nitrogen plasma source is running but with the N shutter closed; is the growth of GaN from the excess Ga on the surface reacting with a small quantity of active nitrogen coming from the RF source around the closed N shutter. This is possible because once formed, the nitrogen atom is very long lived, with a lifetime of the order of seconds. Nitrogen atoms are expected to require a third body, usually the

walls of the chamber, in order to interact and form nitrogen molecules [28,29]. We suggest that the (1]1) surface reconstruction is associated with either an N-terminated surface, or with a strongly bonded Ga monolayer on the (0 0 0 11 ) surface. We have demonstrated that the (2]2) reconstruction is due to an additional amount of relatively tightly bound Ga on this (1]1) surface. Adding further Ga to the (2]2) reconstructed surface results in a return to a (1]1) reconstruction. However, at present we cannot distinguish between the two growth models, but further experiments are in progress to resolve this issue.

4. Summary and conclusions Homoepitaxial growth of GaN by MBE has been studied using a nitrogen RF plasma source for active nitrogen on GaN bulk substrates of nitrogen polarity. Both Ga- and N-rich MBE growth conditions have been studied. The mechanisms giving

6

C.T. Foxon et al. / Journal of Crystal Growth 207 (1999) 1}7

rise to surface reconstruction on this (0 0 0 11 ) surface have been investigated. To study the mechanism responsible for the reconstruction, the e!ects of separately supplying Ga and active nitrogen to the GaN surface, both to the GaN substrate prior to growth and to the GaN epitaxial layers have been examined. The (2]2) reconstruction was observed on the GaN substrates having nitrogen polarity and on GaN layers grown on (0 0 0 11 )GaN substrates with an equivalent polarity. The (2]2) reconstruction can be observed during growth under appropriate conditions and can also be obtained after terminating growth. We have demonstrated that the (2]2) reconstruction was stable and could be maintained at &7503C under an active nitrogen #ux for a few hours. However, the (2]2) pattern disappeared upon cooling the GaN sample under an N plasma #ux at temperatures below &4003C. We have demonstrated that a small fraction of the atomic nitrogen from the RF plasma source arrives at the surface even when the shutter in front of the source is closed. We suggest that the initial (1]1) surface reconstruction is associated with either an N-terminated surface, or with a strongly bonded Ga terminated monolayer on the (0 0 0 11 ) surface and we have demonstrated that the (2]2) reconstruction is due to an additional amount of relatively tightly bound Ga on this (1]1) surface.

Acknowledgements This work was supported by grants from the LAQUANI (ESPRIT project no. 20968), ANISET Programme (BE-3384), EPSRC (GR/L77157), Royal Society, INTAS (96-1031) and NATO (HTECH.LG971309).

[3]

[4]

[5]

[6] [7] [8] [9] [10]

[11] [12] [13] [14] [15] [16]

[17] [18] [19] [20]

[21]

References

[22] [23]

[1] M. Seelmann-Eggebert, J.L. Weyher, H. Obloh, H. Zimmermann, A. Rar, S. Porowski, Appl. Phys. Lett. 71 (1997) 18. [2] Z. Liliental-Weber, S. Ruvimov, Ch. Kisielowski, Y. Chen, W. Swider, J. Washborn, N. Newman, A. Gassmann, X. Liu, L. Schloss, E.R. Weber, I. Grzegory, M. Bockowski,

[24]

J. Jun, T. Suski, K. Pakula, J. Baranowski, S. Porowski, H. Amano, I. Akasaki, Mater. Res. Soc. Symp. Proc. 395 (1996) 351. A.R. Smith, V. Ramachandran, R.M. Feenstra, D.W. Greve, M.S. Shin, M. Skowronski, J. Neugebauer, J.E. Northrup, J. Vac. Sci. Technol. A 16 (1998) 1641. A.R. Smith, R.M. Feenstra, D.W. Greve, M.S. Shin, M. Skowronski, J. Neugebauer, J.E. Northrup, Appl. Phys. Lett. 72 (1998) 2114. A.R. Smith, V. Ramachandran, R.M. Feenstra, D.W. Greve, A. Ptak, T. Myers, W. Sarney, L. Salamanca-Riba, M. Shin, M. Skowronski, MRS Internet J. Nitride Semicond. Res. 3 (1998) 12. A.R. Smith, R.M. Feenstra, D.W. Greve, J. Neugebauer, J.E. Northrup, Phys. Rev. Lett. 79 (1997) 3934. J. Neugebauer, T. Zywietz, M. Sche%er, J.E. Northrup, C.G. VandeWalle, Phys. Rev. Lett. 80 (1998) 3097. S. Shimizu, Y. Suzuki, T. Nishihara, S. Hayashi, M. Shinohara, Jpn. J. Appl. Phys. 37 (1998) L703. B. Daudin, F. Widmann, J. Crystal Growth 182 (1997) 1. M.A. SanchezGarcia, E. Calleja, F.J. Sanchez, F. Calle, E. Monroy, D. Basak, E. Munoz, C. Villar, A. SanzHervas, M. Aguilar, J.J. Serrano, J.M. Blanco, J. Electron. Mater. 27 (1998) 276. M. Wassermeier, H. Yang, O. Brandt, A. Yamada, J. Behrend, K.H. Ploog, Appl. Surf. Sci. 123 (1998) 181. J.M. Van Hove, G. Carpenter, E. Nelson, A. Wowchak, P.P. Chow, J. Crystal Growth 164 (1996) 154. K. Rapcewicz, M.B. Nardelli, J. Bernholc, Phys. Rev. B 56 (1997) 12725. M.B. Nardelli, K. Rapcewicz, J. Bernholc, J. Vac. Sci. Technol. B 15 (1997) 1144. P. Hacke, G. Feuillet, H. Okumura, S. Yoshida, J. Crystal Growth 175 (1997) 94. H. Okumura, K. Ohta, G. Feuillet, K. Balakrishnan, S. Chichibu, H. Hamaguchi, P. Hacke, S. Yoshida, J. Crystal Growth 178 (1997) 113. G. Feuillet, H. Hamaguchi, K. Ohta, P. Hacke, H. Okumura, S. Yoshida, Appl. Phys. Lett. 70 (1997) 1025. P. Hacke, G. Feuillet, H. Okumura, S. Yoshida, Appl. Phys. Lett. 69 (1996) 2507. A.D. Bykhovski, M.S. Shur, Appl. Phys. Lett. 69 (1996) 2397. D. Korakakis, H.M. Ng, M. Misra, W. Grieshaber, T.D. Moustakas, MRS Internet J. Nitride Semicond. Res. 1 (1996) 10. R.F. Davis, S. Tanaka, L.B. Rowland, R.S. Kern, Z. Sitar, S.K. Ailey, C. Wang, J. Crystal Growth 164 (1996) 132. K. Iwata, H. Asahi, S.J. Yu, K. Asami, H. Fujita, M. Fushida, S. Gonda, Jpn. J. Appl. Phys. 35 (1996) L289. M.A.L. Johnson, S. Fujita, W.H. Rowland, K.A. Bowers, W.C. Hughes, Y.W. He, N.A. El-Masry, J.W. Cook, J.F. Schetzina, J. Ren, J.A. Edmond, J. Vac. Sci. Technol. B 14 (1996) 2349. W.C. Hughes, W.H. Rowland, M.A.L. Johnson, S. Fujita, J.W. Cook, J.F. Schetzina, J. Ren, J.A. Edmond, J. Vac. Sci. Technol. B 13 (1995) 1571.

C.T. Foxon et al. / Journal of Crystal Growth 207 (1999) 1}7 [25] D. Korakakis, T.S. Cheng, I. Grzegory, S. Porowski, R.P. Campion, S.V. Novikov, V.B. Lebedev, N.J. Je!s, O.H. Hughes, C.T. Foxon, in preparation. [26] R. Held, S.M. Seutter, B.E. Ishaug, A. Parkhomovsky, A.M. Dabiran, P.I. Cohen, G. Nowak, I. Grzegory, S. Porowski, in preparation.

7

[27] J.L. Weyher, S. Muller, I. Grzegory, S. Porowski, J. Crystal Growth 182 (1997) 17. [28] M.A.L. Johnson, N.A. El-Masry, J.W. Cook, J.F. Schetzina, Mater. Sci. Forum 264}268 (1998) 1161. [29] H. Morkoc, IEEE J. Selected Topics Quantum Electron. 4 (1998) 537.