Si (110) substrates

Current Applied Physics 14 (2014) S29eS33

Contents lists available at ScienceDirect

Current Applied Physics journal homepage: www.elsevier.com/locate/cap

Optical and crystal properties of ammonia MBE-grown GaN layers on plasma-assisted MBE-grown AlN/Si (110) substrates Young-Kyun Noh a, b, Chul-Hyun Park a, Sang-Tae Lee c, Kyung-Jin Kim c, Moon-Deock Kim c, Jae-Eung Oh a, * a b c

Department of Electrical and Communication Engineering, Hanyang University, Ansan, Republic of Korea IV Works Co., Ltd., Ansan, Republic of Korea Department of Physics, Chungnam National University, Daejeon, Republic of Korea

a r t i c l e i n f o

a b s t r a c t

Article history: Received 15 August 2013 Received in revised form 30 September 2013 Accepted 29 November 2013 Available online 19 December 2013

GaN layers were grown by ammonia molecular beam epitaxy (NH3 MBE) on rf plasma MBE (rf-MBE) AlN grown on (110) Si substrates. The surface morphology of GaN epitaxial films is sensitive to the V/III ratio with the RHEED transition from 2D to 3D as NH3 beam equivalent pressure (BEP) increases. The measured FWHMs of X-ray rocking curve for slightly N-rich sample of 0.8 mm thick are 665 and 961 arcsec for (0002) and ð1012Þ peaks, respectively. Based on transmission electron microscopy studies, the reduction in rocking curve width is attributed to enhanced annihilation of dislocations during the initial stage of growth, which agrees with much higher luminescence intensity in room-temperature cathodoluminescence measurements. A kinetic growth model based on the reference [jae.] is used to explain the growth behavior of GaN layers with different NH3 BEP. Ó 2013 Elsevier B.V. All rights reserved.

Keywords: Gallium nitride Molecular beam epitaxy V/III ratio Defect annihilation

The hetero-epitaxial growth of nitride semiconductors on Si has attracted significant research effort for more than two decades. The most ambitious target of the work on (001) Si substrates, concerns the monolithic integration of IIIeV devices with Si CMOS electronic circuits. Because of two-terrace problems [1,2] of nitride semiconductor growth on (001) Si substrates, until now, the (111) Si surface, providing a hexagonal surface symmetry, has been preferred for GaN-based devices grown on silicon because it is considered more suitable for epitaxial growth of the wurtzite phase. Indeed, several reports concerning the performance of high electron mobility transistor (HEMT) and light emitting diode (LED) structures have demonstrated the objectivity of this approach [3,4]. Although GaN-based devices on (111) Si are demonstrating amazing performances, as far as the integration with Si electronics is concerned, technically more suitable Si substrate orientations are required. A candidate for this purpose is (110) Si substrates which are compatible with MOSFET realization and have been studied in view of enhancing the performances of CMOS circuits [5e7]. This has been demonstrated by the realization of light emitting diodes [8] and high-electron mobility transistors [9].

* Corresponding author. E-mail addresses: [email protected], [email protected] (J.-E. Oh). 1567-1739/$ e see front matter Ó 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.cap.2013.11.047

The emergence of nitrides grown on silicon substrates is essentially based on the result of numerous understandings and breakthroughs mainly in the area of material growth technology. Metal-organic chemical vapor deposition (MOCVD) is closely associated with IIIeV nitride development, making it the leading growth technique. More recently, molecular beam epitaxy (MBE) has proved to be a viable technique for growing IIIeV nitrides due to much lower growth temperature than that typically used in MOCVD. Two different methods are available for producing nitrogen radicals in MBE: (i) the use of nitrogen plasma (PA-MBE) and (ii) of ammonia which is thermally cracked at the heated surface (NH3 MBE). For the former case, it has been largely reported that Ga-rich conditions are necessary to obtain high-quality GaN layers, and that N-rich conditions results in three-dimensional (3-D) growth, columnar growth, and very low electron mobility [10]. As a consequence, the rf-plasma MBE growth of GaN must be performed as close as possible to the stoichiometric point. In contrast to PAMBE, for which optimal growth has consistently been reported over a relatively narrow range of temperatures and V/III flux ratios resulting in pit-free step-flow growth, reported parameters for NH3 MBE growth of GaN have varied widely. It is clear that several stepflow regimes exist for NH3 MBE of GaN e both Ga-rich and N-rich. Most studies have focused primarily on N-rich growth regimes since a much larger temperature and flux window for step-flow

S30

Y.-K. Noh et al. / Current Applied Physics 14 (2014) S29eS33

growth is observed within the N-rich regime [11,12]. However, significantly more work would be required to take a potential advantage of NH3 MBE over PA-MBE, which is, as of yet, much less well understood. In this paper, we report on the NH3 MBE growth of GaN layers on rf-MBE grown AlN/(110) Si substrates. The emphasis is made near the stoichiometric point of V/III ratio rather than typical N-rich conditions used in NH3 MBE growth. The surface morphology, structural and optical properties of the GaN films were investigated using reflection high-energy electron diffraction (RHEED), atomic force microscopy (AFM), double-crystal X-ray diffraction (DC-XRD), scanning electron microscopy (SEM), transmission electron microscopy (TEM), and cathodoluminescence (CL) measurements. Growth mechanism of GaN using NH3 MBE is identified based on resulting film properties and a kinetic model proposed by Karpov et al. [13]. Growth was performed on 3 inch diameter (110) Si wafers in Varian Gen-II MBE system equipped with both a nitrogen RF plasma source (SVT RF-4.5) and ammonia source. Reactive nitrogen and ammonia of 99.99995% purity were further purified using separate getter filters. AlN layers were grown using nitrogen rf plasma source. A procedure for the surface oxide desorption is used to prepare reconstructed clean silicon surfaces, consisting of temperature ramping from room temperature to 700  C at which the RHEED transition from 16  2 to 1  1 is occurred. Then, a stacked buffer layer consisting of thin b-Si3N4 and 50-nm thick AlN seed layers were prepared and the procedure is following: First, the silicon surface is exposed to nitrogen plasma for 20 s at 700  C followed by a rapid annealing at 820  C for 10 s in order to form bSi3N4 ðabSi3 N4 ¼ 7:586  AÞ. The temperature is then decreased to 650  C and approximately 1 monolayer (ML) of aluminum is deposited and converted to AlN (0001). A streaky RHEED pattern under conditions in Al-intermediate regime is observed all along the growth of subsequent 50 nm thick AlN layers. The diffraction lines along the ½110Si azimuth appear with half-order fractional diffraction line whereas additional lines are visible along the [001]Si azimuth, representing the 2  6 reconstruction surface, as shown in Fig. 1(a) and (b). For the growth of GaN, ammonia was used as a nitrogen precursor and introduced through a home-made injector. Ga was evaporated from a standard solid source effusion cell. The Ga beam equivalent pressure (BEP) used during the present study was of 2.2  106 Torr. Four GaN samples (AeD) were grown with the ammonia BEP varying from 1  105 to 7  105 Torr. The growth temperature and the growth rate saturated in N-rich condition were 750  C and 0.8 mm/h, respectively. In-situ reflection highenergy electron diffraction (RHEED), transmission electron microscopy (TEM), high-resolution X-ray diffraction (HR-XRD), and cathodoluminescence measurements were used to characterize the growth and material properties of the GaN films. Fig. 2 shows SEM images and corresponding RHEED patterns at the end of growth from a series of samples grown under different NH3 BEP and at growth temperature of 750  C, revealing a large variation in surface morphology and RHEED pattern among the different growth regimes. The films grown with the lowest NH3 BEP of 1  105 Torr (Fig. 2(a)) show no pit features and hexagonal shape Ga droplets. The surface of GaN layers grown with the intermediate NH3 BEP of 3  105 Torr has a uniform, atomically flat surface without Ga droplets. The AFM image corresponding to Fig. 2(b) shows spiral growth hillocks with an average diameter of approximately 2 mm, leading to the very low root mean square (rms) roughness of 1.9 nm over 5 mm  5 mm. The RHEED patterns of this film are streaky with 2  2 reconstruction pattern. The film grown with high NH3 BEP has a very rough, cratered morphology, as shown in Fig. 2(c), characterized by “V”-shape pits bounded by

Fig. 1. RHEED patterns correspond to 50 nm thick AlN seed layers grown on (110) Si surfaces along the (a) Si ½110 azimuth and (b) Si [001] azimuth, showing 2  6 reconstruction surface. Fundamental and fractional order reconstruction rods are marked.

f1012g and f1013g-type planes with a density of w1  109 cm2. Spotty transmission patterns embedded into the weak streak pattern were observed due to the surface pits and roughness in the corresponding RHEED pattern. Fig. 3 shows cross-sectional TEM images from samples grown under Ga-rich near stoichiometric point condition (sample B) and N-rich condition (sample C) corresponding to growth conditions shown in Fig. 2(b) and (c), respectively. Distinct difference between Ga-rich and N-rich films is observed. A high density of threading dislocations and flat merged surface is the characteristics of Ga-rich sample. As shown in Fig. 3(a), in Ga-rich GaN film, the threading dislocations initiated at the interface with AlN seed layer propagates through the entire film and atomically flat surface is observed. In contrast to Ga-rich film, the N-rich film shown in Fig. 3(b) exhibits significant bending of threading dislocations observed throughout the first 200 nm of growth. Many of these dislocations intersect others where they can annihilate if the dislocations have Burgers vectors with opposite sign. The arrows in Fig. 3(b) indicate a few of the closed dislocation loops seen in this film. The dislocation loops seen in the first phase of growth are consistent with the coalescence of 3D islands. In accord with the finding of Lee et al. [10] for MBE growth of GaN on SiC substrates, it is found that a roughened surface morphology in N-rich regime aids the reduction of threading dislocations. These results indicate that NH3 MBE grown GaN under N-rich regime promotes better crystal quality and rough surface. The impact of V/III ratio conditions on the structural quality of GaN layers was investigated using HR-XRD measurements for both symmetric (0002) and asymmetric ð1012Þ. Fig. 4(a) and (b) show uscan X-ray rocking spectra of the samples for each diffraction of

Y.-K. Noh et al. / Current Applied Physics 14 (2014) S29eS33

S31

Fig. 2. SEM micrographs (left) and corresponding RHEED patterns (right) along Si ½110 azimuth of GaN layers: (a) sample A (NH3 BEP ¼ 1  105 Torr), (b) sample B (NH3 BEP ¼ 3  105 Torr), (c) sample C (NH3 BEP ¼ 5  105 Torr), and sample D (NH3 BEP ¼ 7  105 Torr).

symmetric (0002) and asymmetric ð1012Þ, respectively. The broadening of the off-axis ð1012Þ GaN reflections with skew angle of 43 was slightly bigger than the on-axis (0002) reflections. Compared to those (1050 arc-sec and 1700 arc-sec for (0002) and ð1012Þ peaks, respectively) of Ga-rich GaN film grown with NH3 BEP of 3  105 Torr, the FWHM (full-width at half-maximum) values of symmetric and asymmetric diffraction peaks of N-rich GaN sample with NH3 BEP 5  105 Torr are much narrower with values of 665 arc-sec and 961 arc-sec, respectively. Using a reciprocal-space model convolving contributions due to the tilt variance, the twist variance, and the coherence length [14], the edge and screw dislocation densities can be, with relatively high accuracy, estimated from the dependence of the broadened Bragg peak-widths produced by rocking curve results. Variation in the TD density with GaN surface morphologies could be contributed to variation in the initial dislocation density set by the nucleation conditions or to dislocation fusion and annihilation reactions during growth. Therefore, the optimization of the V/III ratio of GaN

layers is essential in order to improve the structural quality of the GaN on (110) Si substrates. To investigate the consequences of V/III ratio on the optical properties of GaN films, Fig. 5 displays the room temperature (RT) cathodoluminescence (CL) spectra of the samples discussed above. For all samples, the PL band-edge corresponds to the dominant contributions of the donor-bound exciton (DoX), and yellow luminescence is observed near 560 nm. A drastic effect is observed on the integrated band-edge PL intensity, which is a good figure of merit. Indeed, a ratio of 10 between Ga-rich and N-rich samples is measured at room temperature, which is clearly observed in CL mapping of these samples. This is, at least in part, related to the ratio of dislocation densities between these samples. For NH3 MBE growth, the active nitrogen species are provided by the thermal cracking of the ammonia molecules incident on the growing surface. In order to understand the growth mechanism with the NH3 MBE technique, it is crucial to understand the cracking behavior of ammonia molecules and the growth kinetics.

Y.-K. Noh et al. / Current Applied Physics 14 (2014) S29eS33

2.0x106

Intensity [arb. units]

It is known that the NH3 cracking efficiency becomes significant above 450  C and reaches about 3.8% or slightly more for temperatures above 700  C. Karpov et al. [13] also developed a detailed surface kinetic model for interpretation of available experimental and growth data, under the assumption of an adsorption layer consisting of gallium adatoms and adsorbed NHx radicals. The efficiency of NH3 can be estimated by studying the growth rate variation as a function of NH3 beam equivalent pressure (BEP), as shown in Fig. 6, and compared them with the calculated ones from Karpov’s kinetic model simulated with parameters of a cracking efficiency of 4% and a growth temperature of 750  C. The characteristics of either N-limited (Ga-rich) or Ga-limited (N-rich) growth are clearly distinguished by changing NH3 BEP over a large range while keeping the Ga flux constant. We have studied the NH3-MBE grown GaN surface, optical, and structural properties of GaN grown on PA-MBE grown AlN seed layers on (110) Si substrates. We found that the material properties are strongly dependent on the V/III ratio. We observe the transition of 2D-to-3D growth as NH3 BEP increases. In Ga-rich regime, the atomically flat surface is obtained with relatively high threading dislocations identified from X-ray rocking curve widths and crosssectional TEM measurements. On the other hand, in N-rich samples, much higher luminescence efficiency is observed due to the

(a)

NH3 BEP 1x10-5 Torr NH3 BEP 3x10-5 Torr NH3 BEP 5x10-5 Torr

6

1.5x10

NH3 BEP 7x10-5 Torr

1.0x106 5.0x105 0.0 -3000 -2000 -1000

0

1000

2000

3000

ω − scan [arcsec] 1.6x104

Intensity [arb. units]

S32

1.4x104 4

1.2x10

(b)

NH3 BEP 1x10-5 Torr NH3 BEP 3x10-5 Torr

NH3 BEP 5x10-5 Torr

NH3 BEP 7x10-5 Torr

1.0x104 8.0x103 6.0x103 4.0x103 2.0x103 0.0 -3000 -2000 -1000

0

1000

2000

3000

ω − scan [arcsec] Fig. 4. Comparison of X-ray rocking curves of samples AeD around (a) (0002) GaN and (b) ð1012Þ reflections, respectively. The measured FWHM (full-width at halfmaximum) values of symmetric and asymmetric diffraction peaks of N-rich GaN sample grown with NH3 BEP 5  105 Torr are 665 arc-sec and 961 arc-sec, respectively.

reduction of threading dislocation density of the film. It is also found that the growth behavior of GaN layers on (110) Si substrates is well explained by the kinetic model proposed by Karpov et al. [13].

6

Intensity [arb. units]

10

5

10

-5

NH3 BEP 3x10 Torr NH3 BEP 5x10 -5 Torr

104 103 102 101 300

350

400

450

500

550

600

Wavelength [nm] Fig. 3. Cross-sectional transmission electron micrographs of (a) sample B (NH3 BEP ¼ 3  105 Torr) and (b) sample C (NH3 BEP ¼ 5  105 Torr). White arrows indicate the bending and coalescence of threading dislocations.

Fig. 5. Room temperature cathodoluminescence (CL) spectra of samples B and C. The insets show the surface CL integrated intensity mapping of each sample.

Y.-K. Noh et al. / Current Applied Physics 14 (2014) S29eS33

Growth Rate [μm/hr]

N Coverage

0.6

10-1 0.4

Growth Rate

Ga Coverage

0.2

0.0

0.0

of Korea (NRF) funded by the Ministry of Education, Science and Technology (2012035274).

0

10-2

Surface Coverage [ML]

10

0.8

S33

1.0x10-5 2.0x10-5 3.0x10-5 4.0x10-5 5.0x10-5 6.0x10-5 7.0x10-5 8.0x10-5

NH3 BEP [Torr] Fig. 6. Measured growth rates of samples AeD and the comparison with the calculated one from Karpov’s kinetic model as a function of NH3 BEP. In simulation, the ammonia cracking of 4% and the growth temperature of 750  C are used. The calculated surface coverage of Ga and N from the kinetic model is also shown.

Acknowledgments This research was supported by Nano Material Technology Development Program through the National Research Foundation

References [1] F. Schulze, A. Dadgar, J. Blasing, A. Krost, Appl. Phys. Lett. 84 (2004) 4747. [2] S. Joblot, F. Semond, Y. Cordier, P. Lorenzini, J. Massies, Appl. Phys. Lett. 87 (2005) 133505. [3] P. Chyurlia, F. Semond, T. Lester, J.A. Bardwell, S. Rolfe, Y. Cordier, N. Baron, J.C. Moreno, N.G. Tarr, Phys. Stat. Sol. A 206 (2009) 371. [4] A. Dadgar, C. Hums, A. Diez, F. Schulze, J. Blasing, A. Krost, Advanced LEDs for solid state lighting, Proc. SPIE 6355 (2006) 63550R. [5] Y. Cordier, J.-C. Moreno, N. Baron, E. Frayssinet, J.-M. Chauveau, M. Nemoz, S. Chenot, B. Damilano, F. Semond, J. Cryst. Growth 312 (2010) 2683. [6] O.E. Contreras, F. Ruiz-Zepeda, A. Dadgar, A. Krost, F.A. Ponce, Appl. Phys. Exp. 1 (2008) 061104. [7] X.Q. Shen, T. Takahashi, H. Kawashima, T. Ide, M. Shimizu, Appl. Phys. Lett. 101 (2012) 031912. [8] D. Marti, C.R. Bolognesi, Y. Cordier, M. Chmielowska, M. Ramdani, Appl. Phys. Exp. 4 (2011) 064105. [9] B. Damilano, F. Natali, J. Brault, T. Huault, D. Lefebvre, R. Tauk, E. Frayssinet, J.C. Moreno, Y. Cordier, F. Semond, S. Chenot, J. Massies, Appl. Phys. Exp. 1 (2008) 121101. [10] C.D. Lee, A. Sagar, R.M. Feenstra, C.K. Inoki, T.S. Kuan, W.L. Samey, L. Salamanca-Riba, Appl. Phys. Lett. 79 (2001) 3428. [11] N. Grandjean, M. Leroux, J. Massies, M. Mesrine, M. Laugt, Jpn. J. Appl. Phys. 38 (1999) 618. [12] A.L. Corrion, F. Wu, J.S. Speck, J. Appl. Phys. 112 (2012) 054903. [13] S.Yu. Karpov, R.A. Talalaev, Yu.N. Makarov, N. Grandjean, J. Massies, B. Damilano, Surf. Sci. 450 (2000) 191. [14] S.R. Lee, A.M. West, A.A. Allerman, K.E. Waldrip, D.M. Follstaedt, P.P. Provencio, D.D. Koleske, Appl. Phys. Lett. 86 (2005) 241904.