DOI: 10.1021/cg900749j
Insight into the Formation Mechanism of One-Dimensional Indium Oxide Wires
2010, Vol. 10 140–145
Alberto Vomiero,* Matteo Ferroni, Elisabetta Comini, Guido Faglia, and Giorgio Sberveglieri INFM-CNR Sensor Lab and Department of Chemistry and Physics for Engineering and Materials, University of Brescia, Via Valotti 9, 25133 Brescia, Italy Received July 2, 2009; Revised Manuscript Received October 26, 2009
ABSTRACT: Controlled growth of oxide nanowires is an emerging research field with many applications in, for example, electronics, energy production and storage, security, and life science. A great number of theoretical and experimental studies have been carried out on the model of growth of semiconducting nanowires, while the systematic analysis of oxide nanowires is still missing. We have examined the morphological and structural features of In2O3 nanowires grown on single crystal substrate, catalyzed by gold clusters, to investigate the basic growth mechanisms of oxide nanowires. These nanowires exhibit a bodycentered cubic structure and grow along the [100] vector of the cubic crystalline cell, thus being an ideal system for modeling nanowire nucleation and growth, without any preferential direction due to cell anisotropy. Bare vapor-solid (VS) or vapor-liquid-solid (VLS), which are the commonly accepted growth mechanisms, cannot account for the complete description of oxide nanowire growth. The experimental findings can be satisfactorily explained under the hypothesis of concurrent direct VS and catalyst-mediated VLS mechanisms during nanowire growth, with the formation of high-index lateral faces which regulates longitudinal elongation at high temperature, according to the periodic bond chain (PBC) theory.
Introduction The increasing scientific interest in quasi one-dimensional (1-D) systems such as nanowires and nanorods has stimulated their functional exploitation, and single-crystalline 1-D nanostructures are nowadays emerging as building blocks for a new generation of electronic1-3 and optoelectronic nanometerscaled devices.4,5 The fabrication techniques of homogeneous 1-D nanostructures have pursued the control over shape, aspect-ratio, and the crystalline arrangement to a considerable degree, and the improvement of the synthesis methods6 has recently achieved the direct integration of functional nanostructures into nanodevices. The shape, dimension, and termination of single crystal nanowires are the main contributors to build up of all the functional properties, including the quantum-size effects, which are being intensively investigated.7 Tailoring of these properties is mandatory for achieving new functionalities rising from quasi 1-D confinement of nanowire growth. Shape, lateral dimensions, and aspect ratio can be controlled in a number of ways, including the application of catalysts to precursors or substrates, modification of the gas carrier and pressure, and variation of the temperature of condensation. In the field of ionic metal oxides, the structural and functional properties of 1-D nanostructures and their great functional potential are attracting almost the same consideration addressed to covalent 1-D nanostructures based on either silicon or group III-V compounds.8,9 Starting from the first paper in 2001,10 which illustrated the fabrication of oxide nanowires using the simple and cheap vapor transport-and-condensation methodology, a lot of works followed. Many different oxide materials and shapes have been obtained. The influence of the shape and morphology on the *Corresponding author: e-mail:
[email protected]; fax: þ39 030 2091 271. pubs.acs.org/crystal
Published on Web 12/14/2009
functional properties and performances of oxide nanowires was clearly demonstrated over a broad range of applications, ranging from subwavelength photonics integration,11 field emitters,12 chemical sensors and nanoscopic electronic nose,13 energy storage for lithium batteries,14 and photovoltaic.15 For this reason, rational control of the condensation process is mandatory. While extensive investigations (both experimental and theoretical) have been carried out to understand the fundamentals of nucleation and growth for semiconducting nanowires, typically grown by chemical vapor deposition techniques, there is a lack of consolidated models to satisfactorily explain all the details of the mechanism of formation of ionic metal oxide nanowires, and only partial descriptions exist.16 Typical models for the vapor-liquid-solid (VLS) growth of semiconducting nanowires17-20 take into account the condensation of the volatile species at various positions of the system (the catalytic tip, the lateral sides of the wire, the substrate), and then diffusion equations are applied, to account for migration of adatoms toward the catalytic tip, at which incorporation occurs. Direct adsorption of volatile species and adatoms at lateral sides is often prevented, and can be neglected in first approximation. Very recent results, however, demonstrated that condensation at lateral side can occur via vapor phase epitaxy (VPE) in InP systems.21 In the peculiar case of GaP/GaAs heterostructures, mass adsorption at the lateral side has been demonstrated to be competitive with wire elongation, resulting in the decrease of the longitudinal growth rate.22 In general, it has been demonstrated that the longitudinal growth rate is dependent on the nanowire diameter, and is determined by the rate limiting process of mass adsorption. Various situations have been observed: (i) increase in the growth rate with the increasing radius, as a consequence of the Gibbs-Thomson effect at the nanowirecatalyst interface caused by the curvature of the nanowire and catalyst surfaces, when the rate limiting process is mass absorption at the nanowire interface;23,24 (ii) faster growth r 2009 American Chemical Society
Article
Crystal Growth & Design, Vol. 10, No. 1, 2010
141
Figure 1. (a) Prospect view of typical In2O3 nanowires showing the preferential growth direction, normal to the sapphire substrate. (b-d) Detailed SEM images of nanowires after 100 condensation at three temperatures: (b) 1315 K; (c) 1290 K; (d) 1235 K. Insets in (b-d): plane view that highlight the presence of the gold catalytic tips and the rectangular cross section of the nanowire. Scale bars: (a) 1 μm; (b-d) 200 nm (main pictures), 100 nm (insets).
of thinner nanowires, when the growth rate is limited by reaction kinetics at the surface of the catalyst or by diffusion of atoms from the substrate.18,20 The purpose of this investigation is to present and explain the nucleation and growth of [100]-oriented nanowires with Ia3 body-centered cubic structure over gold-catalyzed sapphire substrates and to give quantitative description of the condensation processes responsible for crystal growth. The model experimental system composed by In2O3 nanowires grown by Au-particle-assisted transport-and-condensation method was applied. Experimental Section In2O3 single crystalline nanowires have been synthesized via a transport-and-condensation method. The experimental setup for the oxide deposition consists of a high temperature alumina furnace capable of activating decomposition of the oxide precursors and to promote evaporation. Indium oxide precursor powders (purity 99.999% from Sigma-Aldrich) were put in the center of the furnace at a temperature of 1770 K. The controlled pressure (100 mbar) and the gradient of temperature within the furnace allowed condensation and nucleation of the nanostructures downstream from the inert Ar gas flow (100 sccm) in the cooler region of the tube. Alpha-plane single crystalline sapphire substrates were applied to drive nanowire formation via heteroepytaxial growth, due to the very good matching of the lattice plane distances of alpha-plane sapphire (cubic, lattice constant a ≈ 0.476 nm) and In2O3 (cubic, lattice constant a ≈ 1.01 nm) along some specific directions.25,16 For this reason, wire growth with precise orientation (normal, in the specific case) with respect to the substrate surface is expected, as already experimented for similar systems.26 The substrates were preliminarily seeded via drop coating with ∼10 nm gold colloids (BBInternational) to serve as active sites for preferential adsorption of the volatiles. No external oxygen source was applied: residual oxygen inside the furnace guaranteed eventual oxidation of all the condensing species. Condensation products were collected in the temperature range 12001320 K in a series of experiments at different condensation durations (from 10 to 40 min), in order to investigate the kinetics of nanowire nucleation and growth. A reverse Ar flux was applied during transient heating and cooling of the furnace, to prevent uncontrolled condensation. Scanning and transmission electron microscopy (SEM and TEM) have been carried out in order to determine the morphology, the degree of homogeneity, and crystalline arrangement of the nanowires.
The as-prepared structures were observed at low accelerating voltage, in order to avoid the electrostatic charging of the insulating substrate. The optimal voltage for observation was found to be in the 2-3 keV range. The nanostructures have been removed from the substrate used for deposition through dry scratching with a razor blade and dropped over a standard holey carbon film grid for TEM investigation. Standard removal through sonication in alcohol has been avoided to prevent the thin structures from breaking. High-resolution TEM imaging was applied for investigation of the termination of the nanowire lateral sides and apex. Electron diffraction (ED) and analysis of zero-order and higher order Laue-zones were carried out for precise determination of unit cell and space group.
Results and Discussion In Figure 1 we show an overview SEM image of In2O3 nanowires after 10 min condensation at different temperatures (condensation conditions: Ar flux 75 sccm; p=100 mbar). Each nanowire has rectangular cross section and is terminated by a gold catalytic tip (Figure 2a). The higher the temperature is, the smaller the gold tip is (tip radius: rtip, see insets in Figure 1b-d and Figure 2c). A critical temperature Tc exists, which discriminates between two different mechanisms for wire formation, as detailed in the following analysis of the experimental findings. At a temperature above Tc = 1290 K the lateral dimension of the nanowire is significantly larger than the diameter of the tip, while at a temperature below Tc the catalytic tip determines the lateral dimensions of the wire. The evolution of the dimension of the gold tip terminating the nanowire can be summarized as follows (Figure 2c): at T > Tc the gold cluster shrinks immediately at the very beginning of condensation; at T < Tc the gold cluster shrinks quite slowly and reaches the asymptotic value (∼13.0 ( 1.3 nm) in about 40 min condensation. The asymptotic size of the catalyst does not depend on the condensation temperature within the statistical error of the present investigation. No significant change in the size of gold clusters has been detected at temperatures below 1235 K even after 40 min condensation. The time evolution of the cluster size at different temperatures influences the dynamics of wire formation, as discussed below.
142
Crystal Growth & Design, Vol. 10, No. 1, 2010
Figure 2. (a) HRTEM image of the faceted residual catalyst after long-term condensation (40 min). (b) Early nucleation stage of the nanowire, in which formation of the buffer layer and nucleation of nanowire from catalytic seed (white arrows) are evident. (c) Radius of the tip (rtip) as a function of the condensation time at various temperatures. (d) HRTEM image of a nanowire, indicating the growth direction. (e) Fast Fourier transform (FFT) of the HRTEM image indicates that the crystal structure of the nanowire is Ia3 body-centered cubic (spatial group 206). Scale bars: (a) 5 nm; (b) 100 nm; (d) 10 nm.
The nanowires present la3 body-centered cubic In2O3 single crystalline structure (spatial group 206), and grow along the [100] vector of the cubic In2O3 crystalline cell (Figure 2d,e). The parameter radius equivalent (req) is defined to quantitatively describe the lateral dimensions of a single nanowire; req is the radius of the circle, which has the area equal to the cross section of the nanowire. The parameter req is univocally identified for these nanowires, which exhibit parallel lateral sides and uniform cross section. In the case of tapered nanowires, the parameter req is calculated at the base of the nanowire, to take into account the overall condensation process from the very beginning of nanowire nucleation and growth. We note two trends (Figure 3): (i) the mean lateral dimension of the nanowires is smaller at a lower temperature and (ii) the distribution of the lateral dimensions narrows at a lower growth temperature. In Figure 4 we show detail of the termination of In2O3 nanowires grown at different times and temperatures, which can give information on the condensation process leading to nanowire formation, and a pictorial view of the growth mechanism. High index planes are forming at the lateral sides of the pyramidal termination above Tc (Figure 4a,b), while slow shrinking of the tip is visible below Tc (Figure 4c,d). The time evolution of height and lateral dimension of the nanowires is reported in Figure 5. Lateral Growth. Different from most semiconducting nanowires of the III-V group, the diameter of the nanowire is not mainly determined by the diameter of the catalytic tip. The parameter req is linearly dependent upon the time, and the rate of enlargement is higher at a higher temperature (see Figure 5a and Table 1). A nucleation time t0 (∼10 min) exists before which no lateral enlargement of nanowire is recorded,
Vomiero et al.
and a buffer layer forms, which covers the sapphire substrate (see Figure 2b). The source material for lateral enlargement can be either directly absorbed from the vapor, or supplied via adatom flux from the substrate. It is hardly possible discriminating between the two processes: the constant rate of enlargement is compatible with both direct adsorption of vapor precursors, and adatom flux from the substrate, under the hypothesis that the adatom mean free path is larger than the wire length. In both cases, the limiting rate process is the incorporation of the atoms into the crystalline lattice at the nanowire side surface. The lateral condensation is a thermally activated process. In Figure 5c, we show the Arrhenius plots for req at various condensation durations. Different condensation durations result in the same slope, clearly indicating that the condensation is regulated by well-defined activation energy Er. The longer the condensation is, the larger the nanowires are, determining the shift of the Arrhenius plots. We obtained Er =(2.9 ( 0.2) eV. Longitudinal Growth. The height of the nanowires is linearly dependent upon the condensation time only at Tc, at which the growth rate is (128 ( 4) nm/min (Figure 5b). A sublinear trend is found below Tc, while a superlinear trend above Tc. The growth of In2O3 nanowires on polycrystalline substrates mediated by gold active catalytic particles results in very high growth rates, with respect to the present study.27 Concerning nanowire elongation, two phenomena are concurrent with longitudinal growth: (i) VLS condensation at the catalytic tip and (ii) VS condensation at the high-index planes of the pyramidal side. The sublinear trend of nanowire growth at low temperature can be explained as follows. Shrinking of the catalytic tip results in lowered mass flow, and decreases the longitudinal growth rate. One further effect reducing the growth rate at small tip radii is the Gibbs-Thomson effect. As the radius of the Au particle shrinks, the chemical potential of In in Au increases, reducing the driving force for incorporation into Au. The result is a decrease in the longitudinal growth rate.17 Moreover, due to the very low In concentration in the Au-In eutectic catalytic tip,28 shrinking of the catalyst does not result in enhanced growth rate due to In precipitation, as was found for the growth of Si nanowires, instead.19 At high temperature, the effect of mass collection of the catalyst can be readily neglected since it shrinks immediately, and other mechanisms have to be taken into account. We attribute the superlinear longitudinal growth at high temperature to the formation of high index faces of the pyramidal tip. Such a mechanism was never examined for oxide nanowire growth. We briefly recall that in the classical theory of the dynamic growth of the crystals (periodic bond chain (PBC) theory by Hartman)29 the attachment energy (i.e., the bond energy released when one building unit is attached to the surface of a crystal face) is different among the different faces, and is one of the main factors controlling the habit of the crystal. In fact, the higher the attachment energy, the faster the growth velocity is in the direction normal to the corresponding surface; the face tends to disappear quickly and not to contribute to the final habit. In this model, three kinds of faces exist: flat (F), stepped (S), or kinked (K) faces. K and S faces have higher attachment energy with respect to the F faces, which tend to be the most stable boundary for a crystal. All of the lateral sides of the nanowires in the present study are terminated by F faces. At low temperature, the slow lateral enlargement of the wires (from 50 to 200 nm in 40 min
Article
Crystal Growth & Design, Vol. 10, No. 1, 2010
143
Figure 3. Histogram of the radius equivalent for nanowires at six different condensation temperatures, in the range 1215-1315 K, after 40 min condensation. req distribution is a strong function of temperature. At low temperature the distribution of the req is narrower and moves toward a smaller mean lateral dimension of the nanowires.
at 1270 K) and the presence of the catalytic tip at the apex prevent formation of S and K faces. At high temperature, instead, the fast enlargement (from 85 to 490 nm in 40 min at 1315 K) induces formation of high index S faces (Figure 4). Such dynamic behavior is somehow in contradiction with previous observations29 for microcrystals, in which high index surfaces form at the beginning of the nucleation process, and tend to disappear as the dimensions enlarge. At high temperatures, however, the difference in the attachment
energy among F, S, and K faces tends to decrease, and formation of S and K surfaces is allowed. Our results clearly indicate a mechanism for elongation at high temperatures which is not the typical VLS. In fact, in VLS, the growth rate is assumed to be constant when shrinking of the catalyst does not occur, and exhibits a peak as a function of the condensation temperature.30,31 Analysis of the VLS mechanism32,33 indicates that under particular conditions the elongation rate may decrease at
144
Crystal Growth & Design, Vol. 10, No. 1, 2010
Vomiero et al.
Figure 4. (a-d) Termination of the nanowires ad different durations and different temperatures. (a) T = 1315 K, t = 15 min; (b) T = 1315 K, t = 40 min; (c) T = 1260 K, t = 15 min; (d) T = 1260 K, t = 40 min. Scale bars: 200 nm. (e, f) Schematic of nanowire growth at temperature above and below Tc.
Figure 5. (a) Radius equivalent req of the nanowires as a function of the condensation time, at different temperatures (() 1315 K; (blue circle) 1295 K; (green circle) 1280 K; (red up triangle) 1270 K. The straight solid lines are linear fits of the experimental data. The data relative to t = 10 min were not included in the fits. (Inset) Rate of enlargement of the nanowires as a function of the temperature, according to the linear fits. (b) Height h of the nanowires as a function of the time. Three temperatures have been chosen representative of the different situations: (() 1315 K; (blue circles) 1290 K; (red down triangle) 1270 K. The solid lines are a fit of the experimental data. Linear fit has been applied for T = 1290 K, obtaining a growth rate of (128 ( 4) nm/min. Second-order polynomials have been applied for T = 1315 K and T = 1270 K. (c) Arrhenius plot of req at various condensation durations: (blue down triangle) 10 min; (red up triangle) 15 min; (9) 40 min. The straight lines are least-squares linear fits of the experimental data. (Inset): activation energies for radial condensation, as obtained from the Arrhenius plots. The dashed line indicates the average activation energy (Er = 2.9 ( 0.2 eV) obtained from the experimental data. (d) Longitudinal growth rate VS req after 10 min condensation at various temperatures above Tc.
long condensation duration, when surface diffusion from the substrate is the main process of mass delivery and the length
of nanowires becomes comparable with the diffusion length of adatoms. In this study, instead, the growth rate dh/dt is
Article
Crystal Growth & Design, Vol. 10, No. 1, 2010
Table 1. Rates of Lateral and Longitudinal Growth As a Function of the Condensation Temperaturea T (K)
dreq/dt (nm/min)
dh/dt (nm/min)
1270 1280 1290 1295 1315
5.6 ( 0.6 6.57 ( 0.03 10.0 ( 0.2 10.4 ( 0.2 14.8 ( 5
128 ( 4
145
under PRIN2007 Project are acknowledged for partial funding. Mr. A. Pistoni is acknowledged for partial data collection.
References
a
The longitudinal growth rate has been reported only for Tc = 1290 K, at which linear growth occurs.
even increased at high temperatures as condensation proceeds, indicating that adatom mobility from the substrate is not the limiting process for wire elongation, and that formation of high index planes driven by wire enlargement leads to faster growth. One further evidence of the role of high index faces in increasing longitudinal rate is the analysis of the longitudinal growth rate as a function of wire diameter after 10 min condensation at temperatures above Tc, when VS condensation at high index faces takes place (Figure 5d). Lateral enlargement favors high index formation resulting in faster growth of thicker wires and in augmented longitudinal growth rate as the condensation proceeds. The activation energy for longitudinal growth cannot be defined, since the growth rate is not constant. The combination of constant lateral enlargement and wire elongation would result in conical shape of the nanowire. Taking into account the numerical values of dreq/dt and dh/dt a small angle of about 3° should be expected at Tc. At higher temperature, the effect is even lower since the longitudinal growth rate increases with time, while at lower temperatures truncated conical shape is visible, as expected, since the longitudinal growth rate decreases as condensation proceeds due to the gradual shrinking of the catalyst (Figure 4d). Conclusions In summary, we carried out quantitative investigation on the nucleation and growth of indium oxide nanowires on single crystal substrate, mediated by catalytically active gold particles. We demonstrated that two competitive mechanisms concur to nanowire growth, that is, direct VS and catalystmediated VLS adsorption of volatiles. VS mechanism regulates the lateral enlargement of the nanowires, which is a thermally activated process (Er=2.9 ( 0.2 eV), while VLS or VS mechanisms are responsible for wire elongation, depending on the condensation temperature. One future prospect of such knowledge is to determine strategies to control the shape and the lateral and longitudinal dimensions of nanowires, and to tailor their related functional properties in a broad range of application fields. Acknowledgment. INFM-Seed Project 2008, the CARIPLO Foundation under Project ref. 2008.2393 and the MIUR
(1) Cui, Y.; Lieber, C. M. Science 2006, 291, 851–853. (2) Duan, X.; Huang, Y.; Cui, Y.; Wang, J.; Lieber, C. M. Nature 2001, 409, 66. (3) Huang, Y.; Duan, X.; Cui, Y.; Lieber, C. M. Nano Lett. 2002, 2, 101. (4) Lieber, C. M. MRS Bull. 2003, 28, 486. (5) Samuelson, L. Mater. Today 2003, 6, 22. (6) Stellacci, F. Adv. Funct. Mater. 2006, 16, 15–16. (7) Ma, D. D. D.; Lee, C. S.; Au, F. C. K.; Tong, S. Y.; Lee, S. T. Science 2003, 299, 1874. (8) Wan, Q.; Dattoli, E. N.; Fung, W. Y.; Guo, W.; Chen, Y.; Pan, X.; Lu, W. Nano Lett. 2006, 6, 2909. (9) Lu, J. G.; Chang, P.; Fan, Z. Mater. Sci. Eng. R 2006, 52, 49. (10) Pan, Z. W.; Dai, Z. R.; Wang, Z. L. Science 2001, 291, 1947. (11) Law, M.; Sirbuly, D. J.; Johnson, J. C.; Goldberger, J.; Saykally, R. J.; Yang, P. Science 2004, 305, 1269–1273. (12) He, H., Jr; Wu, T. H.; Hsin, C. L.; Li, K. M.; Chen, L. J.; Chueh, Y. L.; Chou, L. J.; Wang, Z. L. Small 2006, 2, 116–120. (13) Sysoev, V. V.; Button, B. K.; Wepsiec, K.; Dmitriev, S.; Kolmakov, A. Nano Lett. 2006, 6, 1584–1588. (14) Kim, D.-W.; Hwang, I.-S.; Kwon, S. J.; Kang, H.-Y.; Park, K.-S.; Choi, Y.-J.; Choi, K.-J; Park, J.-G. Nano Lett. 2007, 7, 3041–3045. (15) Law, M.; Greene, L. E.; Johnson, J. C.; Saykally, R.; Yang, P. Nat. Mater. 2005, 4, 455–459. (16) Nguyen, P.; Ng, H. T.; Yamada, T.; Smith, M. K.; Li, J.; Han, J.; Meyyappan, M. Nano Lett. 2004, 4, 651. (17) Fr€ oberg, L. E.; Seifert, W.; Johansson, J. Phys. Rev. B 2007, 76, 153401. (18) Johansson, J.; Patrik, C.; Svensson, T.; Martensson, T.; Samuelson, L.; Seifert, W. J. Phys. Chem. B 2005, 109, 13567. (19) Kodambaka, S.; Tersoff, J.; Reuter, M. C.; Ross, F. M. Phys. Rev. Lett. 2006, 96, 096105. (20) Johansson, J.; Wacaster, B. A.; Dick, K. A.; Seifert, W. Nanotechnology 2006, 17, S355. (21) Woo, R. L. ; Gao, L ; Goel, N. ; Hudait, M. K. ; Wang, K. L. ; Kodambaka, S. ; Hicks R. F. Nano Lett. 9, 2207-2211. (22) Verheijen, M. A.; Imminl, G.; De Smet, T.; Borgtroem, M. T.; Bakkers, E. P. A. M. J. Am. Chem. Soc. 2006, 128, 1353–1359. (23) Givargizov, E. I. J. Cryst. Growth 1975, 31, 20. (24) Kashchiev, D. Cryst. Growth Des. 2006, 6, 1154. (25) Chen, C.-J.; Xu, W.-L.; Chern, M.-Y. Adv. Mater. 2007, 19, 3012–3015. (26) Wan, Q.; Wei, M.; Zhi, D.; MacManus-Driscoll, J. L.; Blamire, M. G. Adv. Mater. 2006, 18, 234–238. (27) Vomiero, A.; Bianchi, S.; Comini, E.; Faglia, G.; Ferroni, M.; Sberveglieri, G. Cryst. Growth Des. 2007, 7, 2500. (28) Binary Alloy Phase Diagrams; American Society of Metals: Metals Park, OH, 1986; Vol. 1. (29) Hartman, P.; Perdok, W. G. Acta Crystallogr. 1955, 8, 49. Hartman, P.; Perdok, W. G. Acta Crystallogr. 1955, 8, 521. Hartman, P.; Perdok, W. G. Acta Crystallogr. 1955, 8, 525. Hartman, P.; Bennema, P. J. Cryst. Growth 1980, 49, 145. Hartman, P. J. Cryst. Growth 1980, 49, 157. Hartman, P. J. Cryst. Growth 1980, 49, 166. (30) Lew, K.-K.; Redwing, J. M. J. Cryst. Growth 2003, 14, 254. (31) Mohammad, S. N. Nano Lett. 2008, 8, 1532. (32) Harmand, J. C.; Patriarche, G.; Pere-Laperne, N.; Merat-Combes, M.-N.; Travers, L.; Glas, F. Appl. Phys. Lett. 2005, 87, 203101. (33) Dubrovskii, V. G.; Sibirev, N. V.; Cirlin, G. E.; Harmand, J. C.; Ustinov, V. M. Phys. Rev. E 2006, 73, 02163.