Growth of optical-quality, uniform In-rich InGaN
films using
two-heater metal-organic chemical vapor deposition
S.F. Fu
a, C.Y. Chen
a, F.W. Li
a, C.H. Hsu
a, W.C. Chou
a, W.H. Chang
a, W.K. Chen
a,n, W.C. Ke
b aDepartment of Electrophysics, National Chiao Tung University, Hsinchu 300, Taiwan, ROC
bDepartment of Mechanical Engineering, Yuan Ze University, Chung-Li 320, Taiwan, ROC
a r t i c l e i n f o
Article history: Received 1 May 2012 Received in revised form 19 June 2013
Accepted 22 July 2013 Communicated by M. Weyers Available online 2 September 2013 Keywords:
A1. X-ray diffraction
A2. Metal-organic vapor phase epitaxy B1. Nitrides
B2. Semiconducting indium compounds
a b s t r a c t
Good-optical-quality, thick InxGa1xNfilms with high In content were grown using a homemade
two-heater metal-organic chemical vapor deposition system. By varying the growth temperature, it was found that the In composition of the InGaN epilayer varied from 18 to 59% as the substrate temperature decreased from 750 to 6251C. Our results show that the optical properties in terms of the emission peak wavelength and linewidth are uniformly distributed throughout the entire 2 in. wafer for the x¼0.40 InGaN sample. The resultant mean peak wavelength and FWHM are 80876 nm and 229718 meV, respectively, at 18 K. In addition, for the InGaNfilm grown at 625 1C, a noticeable decrease in the In composition occurred when the ceiling temperature was 4800 1C, indicative of the occurrence of parasitic reactions in the gas phase.
& 2013 Elsevier B.V. All rights reserved.
1. Introduction
Direct-band-gap InxGa1xN alloys have proven to be important
materials because of their unique property of wide spectral tun-ability, which can be adjusted continuously from the ultraviolet through the entire visible region and even extending into the near-infrared region[1–3]. This tunability offers many possibilities in a variety of device applications, including high-brightness visible light emitting diodes (LEDs), laser diodes, full-spectrum multi-junction solar cells and phosphor-free solid-state lighting[1–4]. Despite this peculiar feature, many InGaN studies are presently concerned with Ga-rich alloys, while few attempts have been focused on In-rich InxGa1xN (x40.30) with emission wavelengths in the range from
the red to the near-infrared region, spanning from 650 to 1100 nm
[5–9]. In this region, the light-emission efficiency appears to dete-riorate significantly, primarily because of the intrinsic properties of the material itself, such as the wide miscibility gap, the high thermal instability of InN, and the large lattice mismatch between the InGaN emission layer and the GaN base layer, which causes phase separa-tion and composisepara-tionalfluctuation in the film[10]. The situation is even worse when metal-organic chemical vapor deposition (MOCVD) is employed as the method offilm deposition, mainly caused by the use of NH3 precursor, which is known to have a low cracking
efficiency at nominal InGaN growth temperatures (550–800 1C). Several efforts to grow high-quality InGaNfilms have been put forth
by the MOCVD community, including high-growth-rate [11] and raised-pressure methods[12]via either the trapping of In adatoms or the suppression of InN decomposition to increase the In composition. In this study, another approach is proposed. By using a so-called two-heater MOCVD reactor, good-quality In-rich InGaN films of uniform compositional distribution were realized over an entire 2-inch wafer. For example, a sample with x¼0.40 grown at 650 1C exhibits a mean emission wavelength of 80876 nm and a mean FWHM of 229718 meV at 18 K. This suggests that the phase separation and compositional inhomogeneity that are commonly observed in InGaNfilms in the moderate composition range are suppressed to a great extent at low substrate temperatures by the use of this type of reactor.
2. Experiment
The In-rich InxGa1xN epi-films employed here were grown on
2-inch, 600-nm-thick GaN/(0001) sapphire substrates using a homemade low-pressure horizontal MOCVD reactor equipped with two heating elements, as shown inFig. 1. One is the substrate heater, which is designed to provide the necessary growth temperature forfilm deposition; the other is the cracking heater, which is placed on the upper side of the reactor, opposite to the substrate susceptor, for the purpose of providing the ceiling temperature to enhance the supply of active species, especially nitrogen, during the deposition. This arrangement causes the reactor itself to possess the peculiar feature that it is capable of preparing high-In-content InGaN films at low growth Contents lists available atScienceDirect
journal homepage:www.elsevier.com/locate/jcrysgro
Journal of Crystal Growth
0022-0248/$ - see front matter& 2013 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.jcrysgro.2013.07.030
nCorrespondence to: Department of Electrophysics, # 1001, Ta Hsueh Rd,
Hsinchu 300, Taiwann. Tel.:þ886 3 5712121x56125; fax: þ886 3 5725230. E-mail address:[email protected] (W.K. Chen).
independently because the ceiling temperature alone can heat the substrate susceptor to a certain temperature, which limits the minimum substrate temperature that can be used forfilm deposi-tion. The resultant InGaN layers were approximately 400 nm thick, as determined by scanning electron microscope (SEM) measurements.
The as-grown epilayers were then examined using an X-ray diffractometer (XRD) in theω–2θ scan mode of the (002) reflection peak to estimate the In content in the InxGa1xNfilms, under the
assumption that Vegard's law is valid. Photoluminescence (PL) measurements were conducted primarily at 14 K using the 325-nm line of a He–Cd laser as an excitation source and a photomultiplier tube or a liquid-nitrogen-cooled extended InGaAs photodiode as a detector. Cathodoluminescence and reciprocal space mapping were also employed to verify the origin of the double luminescence peaks that were exhibited by the high-temperature InGaNfilms.
3. Results and discussion
Thefirst series of InGaN films were prepared under fixed flow rates of 5.9 and 8.8μmol/min and 4.8 L/min for TMGa, TMIn and
1C. The resultant ω–2θ full widths at half maxima (FWHMs) are shown in the inset of Fig. 2(a). Further reducing the growth temperature, although it could increase the In content to 0.59, caused In droplets to form (32.91), indicating that the In vapor pressure at the growth interface in such a growth environment was above its saturation pressure.
The measured (0002)ω-scan rocking curves at different growth temperatures and the variation of the FWHM as a function of the In content are shown inFig. 3(a) and (b), respectively. Theω scan is known to represent the orientation spread (crystal mosaicity), which is related to the crystalline quality of thefilm. It is found that the rocking curve broadens gradually with decreasing growth temperature, i.e., with increasing In content, up to 40%, which can be ascribed in large part to the increased threading dislocation density, compositionfluctuation, and large internal strain caused by the difference in interatomic spacing between GaN and InN in thefilms[13–15]. Further decreasing the growth temperature to 6251C caused a marked increase of the FWHM in the rocking curve, suggesting that the presence of In droplets caused thefilm quality to deteriorate significantly.
Fig. 4(a) shows the PL spectra of the above InxGa1xN films
obtained at 14 K. For the samples grown at temperaturesZ700 1C, two emission peaks are observable. The corresponding emission peaks, namely, the high- and low-energy emission peaks, are shifted from 2.94 to 2.58 eV and from 2.44 to 2.07 eV, respectively, as the In content increases from 0.18 to 0.30.
The observation of double luminescence peaks in InGaN epilayers has often been attributed to the presence of phase separation and/or strained-relaxed regions [16,17]. The occurrence of macroscopic phase separation can be excluded for our high-temperature films based on the X-ray (002)ω–2θ diffraction profiles, as no additional peak is observed. To verify whether the strain relaxation dominates
Fig. 1. Cross-sectional sketch of the two-heater MOCVD reactor.
Fig. 2. The XRD (0002)ω–2θ curves of InGaN epilayers grown at various temperatures on (a) linear and (b) logarithmic scales. The corresponding variation in the FWHM as a function of the In content of thefilms is shown in the inset of (a).
the luminescent property, we subsequently carried out high-resolution X-ray reciprocal space mapping (RSM), probed at the asymmetrical (105) direction, and cathodoluminescence (CL) mea-surements on 725 (x¼0.23) and 675 1C (x¼0.38) samples, which exemplify the characteristics of our high- and low-temperature samples, respectively. From the asymmetrical reflections, both the out-of-plane (c) and in-plane (a) lattice constants can be derived. The resultant RSMs are displayed inFig. 5(a) and (b). The calculated fully relaxed (dashed) and pseudomorphic (solid) lines are also marked in thefigures. As seen, the RSM of the 725 1C sample indeed exhibits two distinct diffraction peaks. One corresponds to a relaxed InGaN region, labeled InGaN(R), and the other is located at a position near the pseudomorphic line, labeled InGaN(S), indicative of the existence of portions of the InGaN region that are coherent with the underlying GaN layer. On the other hand, only a relaxed diffraction peak is observed in the 6751C sample.
The cathodoluminescence analyses with depth resolution further establish the connection between photoluminescence and the
structural properties of the films. By varying the electron-probe beam energy, the luminescence spectra at different depths can be resolved. The resultant CL spectra probed with beam energies varying from 6 to 20 keV, which correspond to depths from the surface of 40 to 500 nm, are presented inFig. 5(c) and (d). For the 7251C sample, one can observe fromFig. 5(c) that a low-energy peak at 2.45 eV is promptly excited at the lowest electron energy, while the second high-energy component at 2.82 eV begins to appear at beam energies higher than 8 keV. The depth-dependent CL results clearly indicate that the low-energy component of the 7251C sample originates from the relaxed region near the surface, whereas the high-energy feature originates from a deeper, strained region, close to the GaN underlayer, which is in good agreement with the results reported by Pereira et al.[17]. The above argument also explains the narrower observed linewidths in the high-energy PL spectra, which are attributable to the improved crystalline quality of the strained region. For the 6751C sample (x¼0.38), only one CL peak is observed, independent of the electron-probe energy, as depicted inFig. 5(d).
Fig. 3. (a) The XRD (0002) rocking curves of InGaN layers grown at various temperatures and (b) the variation of the FWHM as a function of the In content in thefilms.
Thisfinding is also consistent with the result of the RSM measure-ment, which suggests that the 6751C film is almost fully relaxed. Such a result could be expected. Because of the increasing lattice misfit, the critical layer thickness of the 675 1C sample (x¼0.38) on GaN becomes so thin (o3 nm) that the deposition tends to be subjected to an immediate strain relief at the initial growth stage, leaving almost none of the epilayer to remain coherent.
It is worth mentioning that the photoluminescence emission spectra for the x¼0.38 and 0.40 samples at 14 K peak at 1.82 and 1.68 eV, respectively, corresponding to wavelengths of 681 and 793 nm (854 nm at RT); the latter is well within the infrared color region, which is rarely addressed for thick InGaN layers with high In content by other MOCVD groups. Furthermore, the variation of the linewidths of luminescence peaks from relaxed regions (low-energy components) in these films, obtained from fits with Gaussian functions, is depicted inFig. 4(b). The resultant linewidth is observed to increase gradually from 200 to 306 meV with increasing In content, and it becomes narrower, 258 meV, for the InGaN layer of x¼0.4.
One important concern in this study is the effect of the ceiling temperature on the InGaN growth; we therefore conducted another series of experiments at 6251C with various ceiling temperatures ranging from 700 to 9001C. During the growth, the flow rates of TMGa, TMIn and NH3were set at 5.9 and 8.8μmol/min and 5.8 L/
min, respectively. No In droplets are observable in this set of samples, which is attributable to the higher NH3 flow that was
employed during the deposition. The In compositions derived from the X-ray data and the PL properties as a function of the ceiling temperature are shown inFig. 6(a) and (b), respectively. The effects of the ceiling temperature on the InGaN growth are evident. As seen inFig. 6(a), a slight decrease in the In composition with increasing ceiling temperature is observed at temperatures of 700–800 1C, followed by a steeper fall-off at high temperatures. The PL emission energy essentially reflects the results observed from the X-ray
measurement. The emission peak energy remains nearly unshifted in the temperature region of 700–800 1C and rises monotonically as the ceiling temperature is further increased.
For InGaN MOCVD growth, there are two major factors respon-sible for the decrease of In incorporation; one is In desorption from the growth surface, and the other is parasitic reactions in the gas phase. The reduced In incorporation at high ceiling tempera-tures can be ascribed to the homogenous gas-phase parasitic reactions that occurred in the high-temperature zone above the susceptor; the substrate temperature of 6251C that was used here was in the mass-transport-limited growth region (o675 1C), where the desorption of In is insignificant. A detailed explanation is given below.
For nitride MOCVD deposition that involves NH3and (CH3)3M
source precursors, where M¼Ga, In or Al, the gas-phase parasitic reactions are generally ascribed to two classes of chemical reac-tions [18–20]: either the polymerization reactions of group III-ammonia intermediates, such as (CH3)3M:NH3, (CH3)2MNH2, and
(CH3)MNH, or the radical recombination reactions of group-III
radicals (MCH3), which are a sequence of dissociation reactions
from the intermediates. Unlike AlN, for which the polymerization reactions are initiated at low temperature and exhibit a weak temperature dependence because of the low activation energy, the parasitic reactions of GaN and InN proceed primarily via radical recombination pathways and display steep temperature character-istics, along with a high onset temperature that is necessary to gain sufficient energy to cross the energy barrier and form nuclei and nanoparticle by-products in the gas phase.
The noticeable drop in the In content of our InGaN films at ceiling temperatures 4800 1C agrees well with a recent work by Creighton et al. [20]. They reported that near 8001C, a large portion of the input TMIn reactants are converted into gas-phase metallic In nanoclusters via radical recombination pathways and are no longer available for InGaN deposition because of the
Fig. 5. (105) reciprocal space map of the InGaN epilayers grown at (a) 7251C and (b) 675 1C; CL spectra acquired at various electron-beam energies for the InGaN epilayers grown at (c) 7251C and (d) 675 1C.
thermophoretic force near the growth interface. In addition, we find that the PL intensity increases exponentially with the ceiling temperature in the 700–800 1C region. Such a finding suggests that the imposition of a ceiling temperature can indeed exert a positive influence in growing InGaN despite the occurrence of parasitic reactions at high ceiling temperatures.
For the mass production of light-emitting devices, perhaps the most important concern regarding the growth of high-In-content, thick InGaN epilayers is the spectral uniformity over the entire wafer under the influence of an external energy excitation source, such as photon excitation and current injection. To our knowledge, no study has yet been reported concerning the uniformity of the luminescence properties of MOCVD-grown InGaN thick epilayers in the moderate composition range. We therefore focus our discussion on the x¼0.40 InGaN sample grown at 650 1C.Fig. 7(a) shows the dependence of the PL peak wavelength at 18 K on the probed position moving from the center toward the periphery of the wafer on a quartered 2-inch wafer. The measured temperature is slightly higher than that used in the PL measurement shown in
Fig. 4simply because of the use of a larger sample size for the sake of uniform measurement. As seen in thefigure, good wavelength uniformity is achieved. The PL peak wavelength along the radius fluctuates between 799 and 812 nm throughout most of the wafer area, except for the 2-mm-wide perimeter, where comparatively longer wavelengths are observed because of wafer-edge effects. The deviation of the peak wavelength here is small and comparable to that of the MOCVD-grown undoped InGaN epilayers that emit at a wavelength of425 nm[21].
The shift toward longer wavelengths along the radius suggests that more In atoms are incorporated in thefilm near the periphery of the wafer than at the center. Because the composition of InGaN is sensitive to the growth temperature, the profile of the peak wavelength is closely correlated to the temperature gradient across the wafer. The above result indicates that there was a high-temperature spot of approximately 5 mm in radius located in
the central area of wafer, which can be ascribed to the geometric design of our susceptor.
In addition to the PL uniformity, the resultant evolution of the FWHM across the wafer is also depicted inFig. 7(b). The corre-sponding PL FWHM varies from 252 to 209 meV, except at the periphery. These values are very narrow compared to the other MOCVD-grown undoped InGaNfilms (320–400 meV) that emit at infrared colors [22,23] and comparable to those of MBE-grown films [24,25]of a similar wavelength region, where better out-comes were often achieved (216–297 meV).
As mentioned earlier, unlike GaAs and InP material systems, the luminescence spectrum of an In-rich InGaN epilayer is the collec-tive result of light emissions from a large number of In-rich clusters with various sizes, compositions and densities, rather than from the matrix [26]. When the sample is pumped, the excited carriers, created by absorption in the matrix, can diffuse toward andfill up the lower-band-gap, In-rich clusters, yielding photons with energies lower than the band gap of the matrix. This results in a large Stokes shift for the InGaN material[27,28]. Thus, the resultant FWHM of the luminescence peak could resemble, to a certain extent, the distribution of these In-rich clusters, particu-larly for those of the highest In content. Usually, a rather broader luminescence FWHM is observed for InxGa1xN samples when x is
close to 0.5, as a significant deviation in composition can occur with a modulation wavelength at a length scale of as little as a few nanometers[14,15]. This is because thefilm exhibits pronounced lateral decomposition induced by the onset of 3D growth and the generation of misfit dislocations because of strain relaxation, which occurs almost immediately when InGaN is deposited on GaN because of the large lattice mismatch between the two materials[29,30].
Thus, the comparatively narrow FWHM of the luminescence peak for the x¼0.40 InGaN sample can be regarded as a signature of the suppression of In aggregation during the deposition. Such a suppression can be attributed primarily to the low deposition
Fig. 6. (a) In composition of the InGaN layer and (b) PL intensity as a function of the ceiling temperature.
species, which reduces the barrier for adatom incorporation. This will, in turn, minimize the aggregation of In adatoms at the growth layer, reduce the size and compositionalfluctuation of In clusters, and thereby lead to a narrower PL linewidth in thefilm[32].
4. Conclusions
In summary, we have demonstrated the possibility of growing good-optical-quality, thick InGaNfilms with high In content using a MOCVD reactor equipped with two heating elements. It is found that by the employment of an additional heater placed on the ceiling of the deposition zone, thick InGaNfilms of fairly good optical quality can be fabricated with emission wavelengths extending into the infrared region. More importantly, the imposi-tion of a ceiling temperature does not appear to adversely affect the uniformity of the optical properties of thefilms. For example, an x¼0.40 InGaN sample grown at 650 1C with a ceiling tempera-ture of 9001C exhibits a mean emission wavelength of 80876 nm and a mean FWHM of 229718 meV at 18 K over the entire 2-inch-diameter substrate. Furthermore, a marked drop-off in In compo-sition for the InGaNfilm grown at 625 1C is observed when the ceiling temperature is higher than 8001C, indicative of the occurrence of parasitic reactions in the gas phase in this tempera-ture regime, which deplete the supply of In reactants on the growth surface.
Acknowledgments
This work is supported in part by the projects of MOE-ATU and the National Science Council of Taiwan under Grant no. NSC100-2112-M-009-020-MY3.
References
[1]J. Wu, W. Walukiewicz, K.M. Yu, J.W. Ager III, E.E. Haller, H. Lu, W.J. Schaff, Applied Physics Letters 80 (2002) 4741.
[9]S.C. Bayliss, P. Demeester, I. Fletcher, R.W. Martin, P.G. Middleton, I. Moerman, K.P. O'Donnell, A. Sapelkin, C. Trager-Cowan, W. Van Der Stricht, C. Young, Materials Science and Engineering B 59 (1999) 292.
[10]I. Ho, G.B. Stringfellow, Applied Physics Letters 69 (1996) 2701. [11] B.N. Pantha, J. Li, H.X. Jiang, Applied Physics Letters 96 (2010) 232105. [12]D. Iida, K. Nagata, T. Makino, M. Iwaya, S. Kamiyama, H. Amano, I. Akasaki,
A. Bandoh, T. Udagawa, Applied Physics Express 3 (2010) 075601.
[13]B.N. Panth, H. Wang, N. Khan, J.Y. Lin, H.X. Jiang, Physical Review B 84 (2011) 075327.
[14]M. Müller, G.D.W. Smith, B. Gault, C.R.M. Grovenor, Acta Materialia 60 (2012) 4277.
[15]T.P. Bartel, P. Specht, J.C. Ho and C. Kisielowski, Philosophical Magazine 87 (2007) 1983.
[16]P. Chen, S.J. Chua, Z.L. Miao, Journal of Applied Physics 93 (2003) 2507. [17] S. Pereira, M.R. Correia, E. Pereira, C. Trager-Cowan, F. Sweeney, K.P. O'Donnell,
E. Alves, N. Franco, A.D. Sequeira, Applied Physics Letters 81 (2002) 1207. [18]J.R. Creighton, G.T. Wang, W.G. Breiland, M.E. Coltrin, Journal of Crystal Growth
261 (2004) 204.
[19]A.Y. Timoshkin, H.F. Schaefer III, Journal of Physical Chemistry C 112 (2008) 13816.
[20] J.R. Creighton, M.E. Coltrin, J.J. Figiel, Applied Physics Letters 93 (2008) 171906. [21]M. Bosi, R. Fornari, S. Scardova, M. Avella, O. Martínez, J. Jimenez,
Semicon-ductor Science and Technology 19 (2004) 147.
[22] K.P. O'Donnell, R.W. Martin, S. Pereira, A. Bangura, M.E. White, W. van Der Stricht, K. Jacobs, Physica Status Solidi (b) 216 (1999) 141.
[23] D. Queren, M. Schillgalies, A. Avramescu, G. Brüderl, A. Laubsch, S. Lutgen, U. Strauß, Journal of Crystal Growth 311 (2009) 2933.
[24] Y. Nanishi, Y. Saito, T. Yamaguchi, Japanese Journal of Applied Physics 42 (2003) 2549.
[25] W. Walukiewicz, J.W. Ager III, K.M. Yu, Z. Liliental-Weber, J. Wu, S.X. Li, R.E. Jones, J.D. Denlinger, Journal of Physics D: Applied Physics 39 (2006) R83. [26] V. Lemos, E. Silveira, J.R. Leite, A. Tabata, R. Trentin, L.M.R. Scolfaro, T. Frey,
D.J. As, D. Schikora, K. Lischka, Physical Review Letters 84 (2000) 3666. [27] R.W. Martin, P.G. Middleton, K.P. O'Donnell, W. Van der Stricht, Applied
Physics Letters 74 (1999) 263.
[28] Y.H. Huang, C.L. Cheng, T.T. Chen, Y.F. Chen, K.T. Tsen, Journal of Applied Physics 101 (2007) 103521.
[29] G.B. Stringfellow, Journal of Crystal Growth 312 (2010) 735.
[30] M. Pristovsek, J. Stellmach, M. Leyer, M. Kneissl, Physica Status Solidi (c) 6 (2009) S565.
[31]R.A. Oliver, M.J. Kappers, C.J. Humphreys, G.A.D. Briggs, Journal of Crystal Growth 272 (2004) 393.
[32] S.B. Che, T. Shinada, T. Mizuno, X. Wang, Y. Ishitani, A. Yoshikawa, Japanese Journal of Applied Physics 45 (2006) L1259.
