Microelectronic Engineering 90 (2012) 108–111
Contents lists available at ScienceDirect
Microelectronic Engineering journal homepage: www.elsevier.com/locate/mee
Structural properties of semipolar InGaN/GaN quantum dot superlattices grown by plasma-assisted MBE A. Lotsari a, G.P. Dimitrakopulos a,⇑, Th. Kehagias a, A. Das b, E. Monroy b, Ph. Komninou a a b
Physics Department, Aristotle University of Thessaloniki, GR-54124 Thessaloniki, Greece CEA-CNRS Group ‘‘NanoPhysique et SemiConducteurs,’’ INAC/SP2M/NPSC, CEA-Grenoble, 17 rue des Martyrs, 38054 Grenoble Cedex 9, France
a r t i c l e
i n f o
Article history: Available online 23 March 2011 Keywords: InGaN Quantum dots Semipolar III-Nitrides MBE Electron microscopy
a b s t r a c t The nanoscale properties of self-assembled semipolar InGaN/GaN quantum dot (QD) superlattices, grown 2Þ GaN template, were investigated by by plasma-assisted molecular beam epitaxy (PAMBE) on ð1 1 2 transmission electron microscopy (TEM) techniques. Preferential QD nucleation on crystal planes inclined 2Þ plane was observed. Nominal ð1 1 2 2Þ QDs were lenticular-shaped at small angles relative to the ð1 1 2 but the QD size and faceting increased when nucleation occurred on the inclined planes. Strain and interaction with the threading dislocations introduced fluctuations in the indium concentration. Lattice strain analysis along the growth direction was correlated to the average indium content. Ó 2011 Elsevier B.V. All rights reserved.
1. Introduction Semipolar growth orientations of III-Nitride semiconductors are showing good perspective for reducing the piezoelectric polarization caused by strain in optoelectronic device active regions [1]. In this manner higher internal quantum efficiencies can be attained for higher (In, Al)GaN alloy contents since the exciton 2Þ orientation appears recombination rate is increased. The ð1 1 2 promising in this regard [1]. Quantum dots (QDs) can be exploited in such orientations in order to provide three-dimensional (3D) quantum confinement and to shield the carriers from migrating to the dislocations. The high defect density is still an important problem in semipolar and nonpolar heteroepitaxial films [1–3]. In a recent paper we have studied 2Þ self-assembled semipolar GaN QD superlattices grown in ð1 1 2 AlN by plasma-assisted molecular beam epitaxy (PAMBE), and we have shown that perturbations in the growth orientation of the QDs can be associated to the influence of the strain field of the threading dislocations (TDs) [4]. Recently Das et al. [5] reported 2Þ GaN by PAMBE and their emissive InGaN QDs grown in ð1 1 2 characteristics were studied. A blue-shift in the photoluminescence (PL) peak energies was identified compared to those of corresponding polar (0 0 0 1) QDs grown at same temperature. This was attributed to reduced indium incorporation efficiency for PAMBE growth along this orientation [6]. ⇑ Corresponding author. Tel.: +30 2310 998562. E-mail address:
[email protected] (G.P. Dimitrakopulos). 0167-9317/$ - see front matter Ó 2011 Elsevier B.V. All rights reserved. doi:10.1016/j.mee.2011.03.017
In the present contribution we undertake to elucidate the nano 2Þ InGaN/ scale structural properties of the PAMBE-grown ð1 1 2 GaN QD superlattices since the polarization behaviour of QDs is related to their size and morphology [4,7]. An additional important issue concerns the influence of TDs on the QD layers since the InGaN system is sensitive to strain-induced clustering [8]. Furthermore, using lattice strain measurements we compare the indium content to that of corresponding polar samples grown under identical conditions.
2. Experimental Self-assembled 10-period InxGa1xN/GaN QD superlattices were 2Þ GaN template deposited grown by PAMBE on 1 lm thick ð1 1 2 by metal-organic vapour phase epitaxy (MOVPE) on m-plane sapphire [9]. For comparison, similar samples were grown along the polar (0 0 0 1) orientation under the same conditions. For the samples in this study, the growth temperature of 560 oC was employed and a nominal amount of 10 ternary InxGa1xN monolayers (ML) was deposited. In order to induce the 3D transition, the indium cell temperature was reduced about 30 °C below the In accumulation limit and the Ga flux was fixed at 30% the stoichiometric value. The GaN barriers were deposited with a Ga flux slightly over the Ga stoichiometry, without using indium as surfactant. Samples for TEM analysis were prepared in cross-sectional geometry using tripod polishing, followed by ion-milling in the GATAN PIPS. Conventional and high-resolution TEM (CTEM-HRTEM) experiments
A. Lotsari et al. / Microelectronic Engineering 90 (2012) 108–111
109
were performed using a 200 kV JEOL 2011 (Cs = 0.5 mm, point resolution 0.194 nm) microscope. 3. Results and discussion 0 0 Fig. 1 is a bright field (BF) CTEM micrograph along the ½1 1 zone axis showing the InxGa1xN/GaN superlattice on top of the 2Þ MOVPE GaN template, and the defect structure of the samð1 1 2 ple. The defect content of the film comprised TDs lying on the (0 0 0 2) basal plane with a measured density of 3 1010 cm2. The film surface exhibited a root-mean-square roughness of 3 nm attributed to the growth regime as well as to surface relaxation of the strain fields of emerging TDs. QDs were found to nucleate preferentially at sites where the 2Þ orientation. nucleating plane deviated from the exact ð1 1 2 These local deviations are attributed (a) to the facetted surface 2Þ oriented GaN, roughness associated with the growth of ð1 1 2 and (b) to the effect of the local strain field of the TDs. The HRTEM 3 direction, image of Fig. 2a illustrates QDs viewed along the ½1 1 2 and Fig. 2b shows QDs viewed along the ½1 1 0 0 direction. The latter image shows wetting layers and QDs nucleated on inclined planes in the vicinity of TDs. The QDs were elongated along their base with an average width of 15 nm and their height was 2 nm. The GaN barriers had a height of 5 nm. QDs that nucleated 2Þ plane or on depressions of the GaN barriers exhibon the ð1 1 2 ited a lenticular morphology. On the contrary, QDs nucleated on inclined planes were more facetted and generally larger in height. Fig. 2c is an enlarged HRTEM image of a facetted inclined QD. The preferential nucleation on inclined planes and surface depres 2Þ sions is similar to what has been observed for the case of ð1 1 2 1g can be favourGaN QDs in AlN [4]. Inclined planes such as f1 0 1 able sites of indium adsorption [10]. 2Þ InOther than the clearly resolved QD formations, the ð1 1 2 Ga N layers exhibited corrugated morphology (Fig. 3a). Such x 1x layers were strained and comprised indium fluctuations leading to self-assembled narrow QDs of diameter 2 nm as shown in Fig. 3b [11]. Although the influence of the electron beam cannot be excluded as a cause of indium clustering during TEM observation [12], precautions were taken in order to minimize beam exposure as much as possible. Strain relaxation constitutes another possible mechanism causing the indium fluctuations since the Stranski–Krastanow transition is induced by the lattice mismatch [5]. Enhanced clustering is observed at the points of intersection of the layers with the TDs (Fig. 3a) in agreement with recent molecular dynamics simulations [8]. The existence of strain in the layers is consistent with the observation of upward oriented TD semi-loops as in Fig. 3a. Such semiloops could be introduced by the Matthews–Blakeslee mechanism
0 0 zone axis showing the InxGa1xN/GaN Fig. 1. BF CTEM image along the ½1 1 superlattice on top of the MOVPE GaN template. The inclined threading dislocations lie on the basal plane.
3 zone axis showing QDs nucleated on Fig. 2. (a) HRTEM image along the ½1 1 2 2Þ and at depressions of the GaN barriers. Arrows indicate some of the larger ð1 1 2 0 0 zone axis, QDs as well as an ascending TD. (b) HRTEM image along the ½1 1 2Þ as well as on inclined planes where by arrows indicate QDs nucleated on ð1 1 2 due to the distortion caused by TDs. The dashed line follows an ascending TD. (c) Enlargement of an inclined QD showing a more faceted shape in contrast to the 2Þ-grown QDs. The average orientations of QD facets are indicated by white ð1 1 2 lines.
[13], since a critical resolved shear stress is present on the basal plane in the case of semipolar growth, in contrast to the polar case. TD introduction would be promoted when the overall elastic strain in the InxGa1xN/GaN superlattice exceeds a critical value. Another possible mechanism is the upward climb of threading arms connected to misfit dislocation segments [14]. Misfit dislocation formation may be promoted by the roughening of the growth front [15]. The strain in the corrugated layers was measured at specimen areas close to the TEM foil edge, where the sample thickness was of the order of 5 nm. Image negatives were digitized at 2400 dpi using a special purpose scanner. Using geometrical phase analysis (GPA) [16], the two-dimensional projection of the strain 3 direction. For this purpose field was generated along the ½1 1 2 0 0 spatial frequency and one spatial frequency of type g the g 1 1 1 were employed. The utilized minimum spatial resolution 101 of the analysis was 1.1 nm. Fig. 4 illustrates a strain map superimposed on the HRTEM image, for two successive InxGa1xN layers. The map illustrates the component of the lattice strain along the 2Þ growth direction, i.e. the reduced relative variation of the ð1 1 2 interplanar spacing, d, using as reference the GaN barriers [i.e.
110
A. Lotsari et al. / Microelectronic Engineering 90 (2012) 108–111
exhibited significant deviations from the average that could correspond to indium fluctuations. As discussed in the introduction, the PL blue-shift in the case of semipolar growth compared to polar has been attributed to reduced indium incorporation efficiency [5]. This conclusion was based on previous measurements performed on InxGa1xN films using Rutherford backscattering [6]. It is interesting to verify if this is also valid for the InxGa1xN/GaN superlattice. For this purpose we assumed that the InxGa1xN layers were under an anisotropic biaxial strain state consistent with the expressions of Romanov et al. [17], which were then employed in order to estimate the dependence of the strain on the chemical composition based on Vegard’s law. For this purpose we used a linear extrapolation of the elastic constants using the values given by Wright [18]. The employed lattice constants for InN were as given in Ref. [19]. The normal lattice strain component along the growth direction is given by the expression
elz0 z0 ¼
1 2 3 zone axis, showing InxGa1xN layers with Fig. 3. (a) HRTEM image along the ½1 corrugated morphology. TDs indicated by arrows are observed to induce clustering at the points of their intersection with the layers (dotted circles). (b) Magnified 1 2 3 showing that the layers break apart into small QDs HRTEM image along ½1 (arrows).
dInGaN ð1 þ ez0 z0 Þ 1; dGaN
ð1Þ
2Þ, el 0 0 dewhere z0 is the growth direction along the normal to ð1 1 2 zz notes lattice strain, ez0 z0 is the elastic strain, and dInGaN ¼ xdInN þ ð1 xÞdGaN . For our measured lattice strain in the InxGa1xN layers, the indium content was estimated at x = 0.15 ± 0.04. In contrast, similar HRTEM measurements (not shown) from polar multilayers grown under identical conditions gave an average of elzz = 4.3 ± 0.6% (where z is the [0 0 0 1] direction) obtained from 16 measured layers. The resulting indium percentage is x = 0.24 ± 0.04 for the polar case. Hence the indium incorporation is higher in the polar superlattice in agreement with previous work [5]. 4. Conclusions
(d dGaN)/dGaN]. Overall 10 layers were measured using line profiles averaging the strain along 20 nm of layer length. The overall average value of the lattice strain in the InxGa1xN layers was 3.2 ± 0.8%. Fig. 4 shows that local regions within the layers
The structural properties of semipolar grown InxGa1xN/GaN QD nanostructures were studied by TEM techniques. The QDs were lenticular but presented faceting and increased size in the case of nucleation on inclined planes introduced by the threading
Fig. 4. GPA map of lattice strain distribution in corrugated InxGa1xN layers. The strain map is shown superimposed on the HRTEM image. The inset shows the strain profile averaged over 20 nm along the layers.
A. Lotsari et al. / Microelectronic Engineering 90 (2012) 108–111
dislocations and the roughness of the growth front. QDs nucleated preferentially at inclined facets and surface depressions. Formation of nano-dots at the InxGa1xN layers was also observed and was attributed to the influence of strain. Due to the elastic strain build-up in the superlattice, semi-loops of threading dislocations were identified. Correlation between strain and indium composition in the layers, under biaxial strain approximation, was consistent with reduced indium content compared to PAMBE polar growth under identical conditions. Acknowledgements Work supported under the Greek-EU co-funded program ‘‘HRAKLEITOS II’’ and the EU 7th European Framework Project DOTSENSE (Grant No STREP 224212). Thanks are due to G. Nataf and P. De Mierry for providing the semipolar GaN templates used as substrates. References [1] H. Masui, S. Nakamura, S.P. DenBaars, U.K. Mishra, IEEE Trans. Electron Dev. 57 (2010) 88. [2] J. Smalc-Koziorowska, G. Tsiakatouras, A. Lotsari, A. Georgakilas, G.P. Dimitrakopulos, J. Appl. Phys. 107 (2010) 073525.
111
[3] Th. Kehagias, L. Lahourcade, A. Lotsari, E. Monroy, G.P. Dimitrakopulos, Ph. Komninou, Phys. Status Solidi B 247 (2010) 1637. [4] G.P. Dimitrakopulos, E. Kalesaki, J. Kioseoglou, Th. Kehagias, A. Lotsari, L. Lahourcade, E. Monroy, I. Häusler, H. Kirmse, W. Neumann, G. Jurczak, T.D. _ Young, P. Dłuzewski, Ph. Komninou, Th. Karakostas, J. Appl. Phys. 108 (2010) 104304. [5] A. Das, P. Sinha, Y. Kotsar, P. K. Kandaswamy, G. P. Dimitrakopulos, Th. Kehagias, Ph. Komninou, G. Nataf, P. DeMierry, E. Monroy, J. Crystal Growth, (2011), doi:10.1016/j.jcrysgro.2010.10.130. [6] A. Das, S. Magalhães, Y. Kotsar, P.K. Kandaswamy, B. Gayral, K. Lorenz, E. Alves, P. Ruterana, E. Monroy, Appl. Phys. Lett. 96 (2010) 181907. [7] B. Daudin, J. Phys.: Cond. Matter 20 (2008) 473201. [8] H. Lei, J. Chen, P. Ruterana, J. Appl. Phys. 108 (2010) 103503. [9] P. De Mierry, T. Guehne, M. Nemoz, S. Chenot, E. Beraudo, G. Nataf, Jpn. J. Appl. Phys. 48 (2009) 31002. [10] J.E. Northrup, L.T. Romano, J. Neugebauer, Appl. Phys. Lett. 74 (1999) 2319. [11] S.O. Usov, A.F. Tsatsulnikov, V.V. Lundin, A.V. Sakharov, E.E. Zavarin, D.S. Sizov, Yu G. Musikhin, N.A. Bert, E.M. Arakcheeva, N.N. Ledentsov, Semicond. Sci. Technol. 22 (2007) 528. [12] C.J. Humphreys, Phil. Mag. 87 (2007) 1971. [13] J.W. Matthews, A.E. Blakeslee, J. Cryst. Growth 27 (1974) 118. [14] H.P. Strunk, Solid State Phenom. 131 (2008) 39. [15] J. Mei, R. Liu, F.A. Ponce, H. Omiya, T. Mukai, Appl. Phys. Lett. 90 (2007) 171922. [16] M.J. Hÿtch, E. Snoeck, R. Kilaas, Ultramicroscopy 74 (1998) 131. [17] A.E. Romanov, T.J. Baker, S. Nakamura, J.S. Speck, J. Appl. Phys. 100 (2006) 023522. [18] A.F. Wright, J. Appl. Phys. 82 (1997) 2833. [19] E. Dimakis, E. Iliopoulos, K. Tsagaraki, A. Adikimenakis, A. Georgakilas, Appl. Phys. Lett. 88 (2006) 191918.
