GaN Nanowire Heterostructures

Article pubs.acs.org/crystal

Indium Incorporation in InxGa1−xN/GaN Nanowire Heterostructures Investigated by Line-of-Sight Quadrupole Mass Spectrometry M. Wölz,* S. Fernández-Garrido, C. Hauswald, O. Brandt, F. Limbach, L. Geelhaar, and H. Riechert Paul-Drude-Institut für Festkörperelektronik, Hausvogteiplatz 5−7, 10117 Berlin, Germany ABSTRACT: We investigate the formation of the ternary alloy InxGa1−xN in the growth of self-induced nanowires by plasmaassisted molecular beam epitaxy. A series of samples grown at different temperatures were analyzed in situ by line-of-sight quadrupole mass spectrometry, and the predicted composition was verified by X-ray diffraction on as-grown nanowire ensembles. The InxGa1−xN composition is determined by the thermally activated loss of In due to InN decomposition and In desorption, similar to the composition of planar layers. The convolution of decomposition, reincorporation, and desorption is described by an apparent activation energy of about 2.5 eV. We show the feasibility of in situ control of the InxGa1−xN nanowire composition by line-of-sight quadrupole mass spectrometry.

■

INTRODUCTION The ternary alloy InxGa1−xN is widely used to produce lightemitting diodes (LEDs) in the green and blue spectral ranges. One aspect of their further development is increasing the In content without compromising crystal quality, which could lead to LEDs in the red spectral range that do not contain phosphorus.1 However, because of the lattice mismatch between InxGa1−xN and GaN, the defect density is typically high. Bottom-up grown GaN nanowires (NWs) with axial InxGa1−xN quantum wells (QWs) are investigated as an alternative to planar layers because interface strain can relax laterally at the free sidewall surfaces and the NW heterostructure can be grown defect-free.2,3 The self-induced bottomup growth of GaN NWs4 with InxGa1−xN QWs has been achieved by molecular beam epitaxy (MBE), and NW LEDs have been fabricated with emission wavelengths reaching far into the infrared.5−7 The feasibility of InxGa1−xN NW synthesis over the entire compositional range has been shown.8 Little is known, however, about the mechanism of ternary alloy formation in the growth of III−V NWs in general.9−11 Similar to the growth of planar InxGa1−xN layers, which has been extensively studied, for InxGa1−xN NWs, simultaneous InN decomposition and In desorption occurs at elevated substrate temperatures and affects the alloy composition.9 NW growth by MBE offers the advantage of an in situ analysis of In desorption by line-of-sight quadrupole mass spectrometry (QMS). 12 In this article, we perform a quantitative analysis of the In loss during InxGa1−xN NW growth as a function of the substrate temperature. We measured the In desorption in situ and deduced the composition x. In a series of samples, the change of the composition with growth temperature was verified ex situ by Xray diffraction. We show that the temperature dependence of the In loss can be described in the same terms as planar growth © 2012 American Chemical Society

and that the InxGa1−xN composition can be controlled in situ by line-of-sight QMS.

■

SAMPLE PREPARATION AND IMPACT OF COMPOSITION ON INXGA1−XN NW LEDS

The samples under study consist of GaN NW ensembles with axial InxGa1−xN QWs embedded in the upper part of the NWs. The NWs were grown self-induced on 2 in. Si(111) substrates by plasma-assisted MBE. The impinging fluxes ΦGa and ΦN are given according to the growth rate of planar GaN layers, and ΦIn according to planar InN.12,13 During the entire growth experiment, the samples were exposed to a fixed flux ΦN = 13 nm/min. Prior to growth, the substrate was nitridated under N plasma for 5 min at 780 °C. In this way, the competition between nitridation of Si and nucleation of GaN is avoided and a controlled SixNy interlayer is formed.14 The GaN base NWs were then grown at the same temperature under N-rich conditions to a length of about 500 nm. This self-induced growth yields NWs with a typical diameter between 50 and 100 nm and a density of about 1010 cm−2. After the growth of the base NWs, the Ga flux was interrupted and the substrate temperature T was lowered for the growth of the active region. The temperature was monitored with a pyrometer operating at 900 nm, a wavelength at which the GaN NWs are transparent to the thermal radiation from the Si substrate. We give the pyrometer reading recorded immediately before opening the effusion cell shutters. A series of samples were produced with a variation in the growth temperature of the active region between 588 and 631 °C. All other parameters were kept identical, with ΦIn = 0.8 nm/min and ΦGa = 0.8 nm/min, yielding a V/III flux ratio of 8.1 for the growth of the active region. Scanning electron micrographs of one sample are shown in Figure 1 in cross-sectional (a) and plan (b) views. Received: August 15, 2012 Revised: October 4, 2012 Published: October 23, 2012 5686

dx.doi.org/10.1021/cg301181b | Cryst. Growth Des. 2012, 12, 5686−5692

Crystal Growth & Design

Article

Ga is incorporated preferably because the Ga−N bond strength is higher than the In−N one.17 The maximum amount of In incorporation at low growth temperatures is hence given by the difference in the N and Ga supply. Consequently, excess In segregates on the surface.18 Two additional factors are the thermal decomposition of InN,19 resulting in new In adatoms, and the desorption of In adatoms from the substrate surface.20 Both processes compete with the incorporation of In adatoms originating from the vapor phase and the reincorporation of In adatoms originating from the crystal. All of these atomic processes are sketched in Figure 3a.

Figure 1. Scanning electron micrograph of GaN NWs on Si(111) with axial InxGa1−xN/GaN QWs embedded at the tip, in cross-sectional (a) and plan (b) views. Another set of two additional samples with the active region grown at 603 and 619 °C was used to fabricate LEDs. Only for this NW LED series, the active region was embedded in a Si-doped NW base and a Mg-doped top segment. The as-grown NW ensemble was then processed with an insulator between the NWs, and top and bottom contacts were added.7 Before we analyze the InxGa1−xN growth in detail, we show the effect of T on the emission wavelength of the NW LEDs. Figure 2

Figure 2. Electroluminescence acquired at 300 K from InxGa1−xN NW LEDs. The dominant emission wavelength is lower for the sample with the active region grown at 619 °C than for the one grown at 603 °C.

Figure 3. Atomic processes occurring during InxGa1−xN growth. (a) Impingement, incorporation, decomposition, and desorption from a planar InxGa1−xN film via the formation of adatoms. Full spheres depict In atoms, open spheres the former position of In atoms. (b) Impingement, diffusion, and desorption of adatoms on the top facet and sidewall of a NW.

displays room-temperature electroluminescence spectra of the two NW LED samples that differ only in the growth temperature T of the active region. The intensities are normalized. The sample with the active region grown at 619 °C shows an emission spectrum centered at 551 nm. The sample grown at 603 °C exhibits a longer dominant emission wavelength of 588 nm, which can be explained by a higher In content of the QWs due to a lower loss of In during the growth.6 It should be noted that the quantitative correlation of the emission energy with the In composition is not straightforward.15 In view of the sensitive change in emission color, it becomes obvious, however, that the control of the InxGa1−xN composition during the growth is essential for device manufacturing.

The decomposition of InxGa1−xN can be described as a thermally activated process ΦInN,dec = C x exp( −Eadec /kBT )

(1)

Edec a

is the activation energy for InN decomposition at substrate temperature T, and C is a constant. To describe the resulting composition x as a function of the impinging fluxes ΦN, ΦGa, and ΦIn, two situations must be distinguished: metal-rich (Nlimited) growth, and N-rich growth. In the case of metal-rich growth, where ΦGa + ΦIn > ΦN, the reduction of incorporated N due to InN decomposition yields an effective N flux ΦN,eff = ΦN − ΦInN,dec, and x is given by

■

PROCESSES RELEVANT FOR INXGA1−XN NANOWIRE GROWTH Previous studies showed that the composition x of InxGa1−xN NWs depends on the impinging fluxes and T,9,15,16 similar to planar layers. Before we turn to our own data, we discuss the microscopic processes that affect x. As a starting point, we describe the well-understood growth of planar InxGa1−xN by MBE. Atoms from the vapor phase are not directly incorporated into the crystal but adsorb first on the substrate surface. When Ga and In adatoms compete for incorporation into the crystal,

x=

ΦN,eff − ΦGa ΦN,eff

=1−

ΦGa ΦN − ΦInN,dec

(metal‐rich) (2)

Therefore, x does not depend on the impinging In flux. Equations 1 and 2 were shown by Averbeck and Riechert21 to describe planar InxGa1−xN growth by MBE under metal-rich 5687



Article

background pressure of 10−5 Torr during the growth changes the QMS response. Second, the directional characteristic of desorption from the substrate depends on the exact NW geometry. To quantify ΦIn,des with the observed partial pressure, therefore, requires a calibration for an identical NW geometry and background pressure. This procedure is shown in Figure 4a. After the growth of the self-induced GaN NW base, the Ga

conditions on the Ga-polar C-plane surface and thus constitute a quantitative growth model. This approach was also confirmed for planar growth on the N-polar22 surface and on offorientations.23 Nitride NWs grow in the N-polar C-direction in MBE.24 The literature value of Edec for InN decomposition a from the N-polar surface is 1.2 eV.25 The growth of self-induced InxGa1−xN NWs by MBE requires N-rich conditions (ΦN > ΦGa + ΦIn). In incorporation is not limited by the N supply, but by the decomposition of InN and the desorption of In adatoms ΦIn,des. If all the In atoms that are not incorporated ultimately desorb, the composition is given by x=

ΦIn − ΦIn,des ΦGa + ΦIn − ΦIn,des

(N‐rich) (3)

During InxGa1−xN growth decomposition, reincorporation and desorption occur simultaneously and determine the total In loss. The temperature dependence of these single processes is unknown, and In loss under N-rich conditions has not been quantified. In the growth of NWs, in addition, mass transport and inhomogeneous strain have to be considered. First, mass transport is more complex than that for planar layers. Growth occurs predominantly at the tip of the NW, and adatoms arrive by direct impingement as well as diffusion on the substrate and the sidewall, as depicted in Figure 3b.26 Impingement from the vapor is, in turn, subject to shadowing by other NWs.27 Moreover, it is known that the In adatoms enhance the diffusivity of Ga atoms on the substrate28 and on the NW sidewall.29 One consequence of these mass transport effects is that In incorporation depends on the size and arrangement of the NWs.10 It is unclear so far how these mass transport effects contribute to the final alloy composition. For example, In diffusion along the NW sidewalls increases with T; that is, at higher T, more In should be supplied to the top facet. At the same time, within our experimental range of temperatures, In desorption from the sidewalls might also decrease the diffusion length for increasing T and lead to a nonmonotonic dependence of the In supply on T. Second, during the growth of the InxGa1−xN/GaN heterostructures, strain relaxes elastically at the free NW sidewalls, and hence the strain state changes along the NW radius. This nonuniform interface strain affects the In incorporation, which is thought to be favored in the center of the NW top facet.30

Figure 4. In desorption from the sample surface during the growth of InxGa1−xN/GaN NWs by MBE. (a) Calibration of full desorption and background pressure at 800 °C, and growth of a segment at 604 °C. In addition to the In partial pressure measured by line-of-sight QMS, the status of the In shutter and the substrate temperature are indicated. (b) Comparison of InxGa1−xN QWs and extended NW segment growth at the same temperature. The amount of desorbing In is given relative to the full desorption as described in (a). (c) In desorption during InxGa1−xN QW growth for experiments at different substrate temperatures.

■

IN SITU GROWTH CONTROL BY LINE-OF-SIGHT QUADRUPOLE MASS SPECTROMETRY Equation 3 shows that, for a quantitative description of In incorporation, it is essential to measure In desorption. In MBE, the mean free path of atoms that have escaped the sample surface is long enough to assess desorbing fluxes by line-of-sight quadrupole mass spectrometry.14,31 Here, we describe our experimental procedure. Our QMS (SRS RGA200) is mounted in an effusion cell port at an angle of 21° to the surface normal. Atoms desorbing from the substrate can reach the QMS through an aperture that defines a circular collection area with a diameter of about 3 cm centered on the 2 in. wafer. The In partial pressure, detected by the QMS with a sensitivity of 10−12 Torr, scales linearly with the desorbing In flux ΦIn,des from the substrate surface. We have observed that the partial pressure reading is significantly influenced by two factors. First, the high N

flux is stopped at time t = 0 while the supply of active N continues and the substrate is held at about 800 °C. The In shutter is opened for 2 min. At this high temperature, all In arriving at the substrate desorbs, and we assume that the In partial pressure detected by QMS corresponds to 100% of the incident In flux. This level is reached immediately after opening the In shutter. When the In shutter is closed, the In partial pressure instantaneously drops to the background level, that is, 0% of the incident In flux. Between t ≈ 300 s and t ≈ 900 s, T is stabilized at 604 °C in preparation of the growth of a long InxGa1−xN segment. Figure 4a shows that, upon opening the In shutter again for several minutes, the In desorption does not 5688



Article

attain the level of the impinging flux, with the remaining In being incorporated into the NWs. In this particular case, 52% of the incident In flux desorbs, and hence 48% is incorporated. For the samples under investigation, InxGa1−xN QWs were introduced by opening the In shutter for 54 s. In Figure 4b, the In desorption during the growth of two consecutive QWs is shown by the blue line. The experiment was carried out at the same T as the growth of the longer InxGa1−xN segment in Figure 4a, of which the relative In desorption profile is repeated for comparison. A steady state with an In desorption of 53% of the incident In flux is reached in the last 20 s of the QW growth. This is very close to the In desorption observed during the segment growth. Figure 4c compares the In desorption during the growth of the second QW, with the arrows indicating the average reading over the last 20 s of the shutter opening time. As expected, more In desorbs at higher T. We attribute the fact that, in Figure 4c, the indicated desorption at 625 °C is higher than that at 631 °C to an experimental error of ±8% in the determination of ΦIn,des.

experimentally observed increase of In desorption with T follows ΦIn,des = C′ x exp( −Eades /kBT )

(4)

as the Arrhenius-like plot in Figure 5b indicates. The dependence of ΦIn,des on x is written in analogy to eq 1, because decomposition of InN is one of the processes involved in In loss, and the resulting desorbing flux certainly depends on x. The fit result Edes a = 2.3 ± 0.4 eV is an “apparent activation energy” for the convolution of decomposition, reincorporation, dec and desorption. Edes a is higher than the published value of Ea = 25 1.2 eV for InN decomposition from the N-polar surface. This contrast to the growth of planar layers under metal-rich conditions reveals the importance of In reincorporation and desorption during the growth of InxGa1−xN NWs under N-rich conditions. Our value is similar to the activation energy of 2.5 eV that is given in ref 25 for In desorption from a liquid adlayer. However, this cannot be the sole origin of our apparent activation energy, because desorption from such an adlayer would not depend on x, as is the case in eq 4. Analogous to the growth model for InxGa1−xN under metalrich conditions which we cited above,21 one can formulate a quantitative model for the growth of InxGa1−xN NWs by MBE. Equation 4 can be substituted for ΦIn,des into eq 3. Solving the quadratic equation yields

■

QUANTIFICATION OF IN LOSS Now we analyze the data of Figure 4c on the basis of the atomic processes described previously. To this end, we employ the calibration of the impinging In flux and convert the relative desorbing In flux values into absolute fluxes in equivalent growth rate units. The significant benefit of QMS is that, with the data for the impinging and desorbing In fluxes, we can directly calculate the InxGa1−xN composition via eq 3. The predicted x is shown in Figure 5a by the box symbols and error

x=

ΦGa + ΦIn + A −

(ΦGa + ΦIn + A)2 − 4A ΦIn 2A (5)

exp(−Eades/kBT).

where A stands for C′ Evaluating this expression with the fit result from the Arrhenius-like plot in Figure 5b delivers the InxGa1−xN composition shown by the line in Figure 5a. The agreement with the predicted composition for the individual samples (boxes and error bars) is good because we obtained the fit parameters C′ and Edes a from the same set of experiments. These parameters are not universal constants, and the model is, therefore, heuristic and has to be calibrated for each experimental setup.

■

EX SITU VERIFICATION OF THE GROWTH MODEL FOR INXGA1−XN NANOWIRES To verify the magnitude of the change in InxGa1−xN composition as a function of T with an established ex situ tool, we performed symmetric Bragg X-ray diffraction (XRD) experiments. The profiles shown in Figure 6 were obtained in a Panalytical X’Pert system with a Ge(220) hybrid monochromator and a Ge(220) analyzer crystal. The arrows indicate the zeroth- and ±first-order superlattice satellite peak positions. This interference arises from the diffraction at the periodic stack of InxGa1−xN QWs and GaN barriers along the NW axis. Because of the fluctuation in layer thicknesses, which is always associated with self-induced NW growth, the satellites are broad and of low intensity compared to XRD on planar superlattices.29 Still, the zeroth-order superlattice peak position ω0 can be determined with reasonable accuracy, and the average lattice constant of the superlattice is then given by cavg = 2d0002 = λ/sin ω0, where λ is the wavelength of the X-rays. Vegard’s law provides the average composition xavg in the axial NW superlattice: xavg = (cavg − cGaN)/(cInN − cGaN), where cGaN and cInN are the lattice parameters of relaxed GaN and InN, respectively.32 In the inset in Figure 6, xavg is shown as a

Figure 5. (a) InxGa1−xN QW composition x as predicted from desorbing In flux by eq 3, shown as boxes and error bars. The solid line represents the heuristic model for the In loss with the fit parameters as determined from (b). (b) Arrhenius-like plot of the experimental ΦIn,des from line-of-sight QMS. The solid line shows the fit with Edes a = 2.3 ± 0.4 eV.

bars. A possible systematic error in the calibration of ΦIn and ΦGa is not taken into account. Within the QMS measurement uncertainty, the In content decreases with increasing substrate temperature, as it is also the case for planar layers.19,21,22 The fairly large range of observed In concentrations demonstrates the crucial role of the substrate temperature. Very importantly, the QMS monitoring allows the in situ control of the ternary alloy formation. As the next step, we analyze the dependence of the In loss on T. The growth of the NW samples was N-rich. Therefore, InN decomposition, In reincorporation, and In desorption determine the total In loss, which is accessible to the measurement as In desorbing from the substrate during the growth. The 5689



Article

Figure 7. Estimates of x from XRD for the InxGa1−xN/GaN NW samples. The values were obtained for assumed QW thicknesses of 2.4 nm (squares), 3.3 nm (triangles), and 5.0 nm (diamonds). The lines represent the prediction for x based on XRD data (blue) and in situ QMS data (green and thick). Figure 6. Symmetric Bragg XRD scans of InxGa1−xN/GaN NW superlattice samples. The inset shows the average superlattice composition xavg derived from the zeroth-order peak position ω0.

50%, which, in turn, means that no desorption occurs. We clearly observe desorption in QMS, however, and therefore, sensible estimates of dwell should be higher. If, on the other hand, dwell = 5.0 nm is assumed, there is another contradiction to the QMS data. The resulting x from XRD is then much lower than the prediction from QMS, and one would be forced to conclude that QMS detects only a very small fraction of the In lost during QW growth. In view of the careful QMS calibration described above, this seems unlikely. A QW thickness of dwell = 3.3 nm is a reasonable estimate, giving compositions in the middle between these boundary cases. For the three estimates of dwell, ΦIn,des is obtained by eq 3, and the values for Edes a when fitted to eq 4 range between 4.1 ± 1.0 and 2.2 ± 0.2 eV. (The error increases with the experimental error in x. As a result, the error in Edes a , if obtained from XRD data, is large for small dwell, and vice versa.) For dwell = 3.3 nm, Edes a = 2.7 ± 0.3 eV matches and confirms the value obtained from QMS. The dependence of x on T according to the model in eq 5 with the fit parameters as derived from XRD is given by the thin blue line in Figure 7. The thick green line shows the higher prediction of x with the fit parameters from QMS. The deviation is due to the systematically different estimate of the composition between XRD and QMS, as discussed above. The heuristic model expressed in eq 5 can be used in both cases, however, and the apparent activation energy as a fit parameter is equal in XRD and QMS within the errors.

function of T. The average composition xavg decreases by a factor of 3 between the samples grown at 588 and 631 °C. The prediction for x from the QMS data in Figure 5a decreases only by a factor of 2 between the same samples. Two factors may contribute to this deviation. First, XRD probes a single spot on the sample, which is much smaller than the collection area of QMS. The substrate temperature is inhomogeneous, and, if the collection area contains slightly cooler regions, the In desorption at high growth temperatures is systematically underestimated by averaging over the collection area. x is then overestimated from QMS data. Second, any parasitic incorporation of In on the NW sidewalls would reduce the desorbing In flux, and again, x would be overestimated. A meaningful confirmation of the activation energy for the In loss that we obtained from the QMS data in Figure 4c requires the extraction of the In content x in the QWs from the XRD data. The composition x can be derived from dwellx = dSLxavg if the contribution of the QW thickness dwell to the superlattice period dSL is known. The period dSL can be determined from the superlattice peak separation in the XRD profiles in Figure 6 by dSL = λ/[2(sin ωn − sin ωn−1)]. For this series of samples, dSL amounts to 10.4 nm with a standard deviation of σd,SL = 0.5 nm between samples. However, we can give only a rough estimate of dwell between 2.4 and 5.0 nm, based on calibrated axial growth rates. The QW thickness in self-induced NW ensembles fluctuates randomly by as much as σd,well = 2 nm within one sample,29 and the estimate of dwell for the ensemble of NWs can, therefore, not be verified within a microscopic analysis. As a result, the In loss can be judged from the XRD data only with a considerable error. This further highlights the importance of the QMS data. Figure 7 shows estimates of the In content in the QWs as obtained from XRD. The values for x are given for different estimates of dwell as 2.4, 3.3, and 5.0 nm. Assuming that dwell = 2.4 nm implies that x, for the sample grown at 588 °C, is almost

■

CONCLUSION The present data demonstrate that the formation of the ternary alloy InxGa1−xN in NWs can be treated in the same framework as planar layers: the thermally activated In loss determines the alloy composition. The resulting composition as a function of growth temperature can be modeled with an Arrhenius-like behavior of the loss from the substrate surface. For the quantitative description of the In loss, it has not been necessary to invoke effects of mass transport. Thus, we speculate that, in our experiments, mass transport did not limit In incorporation. There are two limitations to our analysis. First, any elastic effect 5690



Article

(11) (a) Kim, Y.; Joyce, H. J.; Gao, O.; Tan, H. H.; Jagadish, C.; Paladugu, M.; Zou, J.; Suvorova, A. A. Nano Lett. 2006, 6, 599−604. (b) Borg, B. M.; Dick, K. A.; Eymery, J.; Wernersson, L. E. Appl. Phys. Lett. 2011, 98, 113104. (12) Gallinat, C. S.; Koblmüller, G.; Brown, J. S.; Speck, J. S. J. Appl. Phys. 2007, 102, 064907. (13) Heying, B.; Averbeck, R.; Chen, L. F.; Haus, E.; Riechert, H.; Speck, J. S. J. Appl. Phys. 2000, 88, 1855−1860. (14) Chèze, C.; Geelhaar, L.; Trampert, A.; Riechert, H. Appl. Phys. Lett. 2010, 97, 043101. (15) Wölz, M.; Lähnemann, J.; Brandt, O.; Kaganer, V. M.; Ramsteiner, M.; Pfüller, C.; Hauswald, C.; Huang, C. N.; Geelhaar, L.; Riechert, H. Nanotechnology 2012, 23, 455203. (16) (a) Vajpeyi, A. P.; Ajagunna, A. O.; Tsagaraki, K.; Androulidaki, M.; Georgakilas, A. Nanotechnology 2009, 20, 325605. (b) Albert, S.; Bengoechea-Encabo, A.; Lefebvre, P.; Sánchez-García, M. A.; Calleja, E.; Jahn, U.; Trampert, A. Appl. Phys. Lett. 2011, 99, 131108. (c) Keating, S.; Urquhart, M. G.; McLaughlin, D. V. P.; Pearce, J. M. Cryst. Growth Des. 2011, 11, 565−568. (d) Tourbot, G.; Bougerol, C.; Grenier, A.; Hertog, M. D.; Sam-Giao, D.; Cooper, D.; Gilet, P.; Gayral, B.; Daudin, B. Nanotechnology 2011, 22, 075601. (17) Böttcher, T.; Einfeldt, S.; Kirchner, V.; Figge, S.; Heinke, H.; Hommel, D.; Selke, H.; Ryder, P. L. Appl. Phys. Lett. 1998, 73, 3232− 3234. (18) (a) Waltereit, P.; Brandt, O.; Ploog, K. H.; Tagliente, M. A.; Tapfer, L. Phys. Rev. B 2002, 66, 165322. (b) Dussaigne, A.; Damilano, B.; Grandjean, N.; Massies, J. J. Cryst. Growth 2003, 251, 471−475. (19) Adelmann, C.; Langer, R.; Feuillet, G.; Daudin, B. Appl. Phys. Lett. 1999, 75, 3518−3520. (20) Dimakis, E.; Iliopoulos, E.; Tsagaraki, K.; Kehagias, T.; Komninou, P.; Georgakilas, A. J. Appl. Phys. 2005, 97, 113520. (21) Averbeck, R.; Riechert, H. Phys. Status Solidi A 1999, 176, 301− 305. (22) Nath, D. N.; Gur, E.; Ringel, S. A.; Rajan, S. Appl. Phys. Lett. 2010, 97, 071903. (23) Turski, H.; Siekacz, M.; Sawicka, M.; Cywiński, G.; Kryśko, M.; Grzanka, S.; Smalc-Koziorowska, J.; Grzegory, I.; Porowski, S.; Wasilewski, Z. R.; Skierbiszewski, C. J. Vac. Sci. Technol., B 2011, 29, 03C136. (24) (a) Kong, X.; Ristić, J.; Sánchez-García, M. A.; Calleja, E.; Trampert, A. Nanotechnology 2011, 22, 415701. (b) Hestroffer, K.; Leclere, C.; Bougerol, C.; Renevier, H.; Daudin, B. Phys. Rev. B 2011, 84, 245302. (25) Koblmüller, G.; Gallinat, C. S.; Speck, J. S. J. Appl. Phys. 2007, 101, 083516. (26) (a) Debnath, R. K.; Meijers, R.; Richter, T.; Stoica, T.; Calarco, R.; Lüth, H. Appl. Phys. Lett. 2007, 90, 123117. (b) Bertness, K.; Roshko, A.; Mansfield, L.; Harvey, T.; Sanford, N. J. Cryst. Growth 2008, 310, 3154−3158. (c) Ristić, J.; Calleja, E.; Fernández-Garrido, S.; Cerutti, L.; Trampert, A.; Jahn, U.; Ploog, K. H. J. Cryst. Growth 2008, 310, 4035−4045. (d) Consonni, V.; Dubrovskii, V. G.; Trampert, A.; Geelhaar, L.; Riechert, H. Phys. Rev. B 2012, 85, 155313. (27) Gotschke, T.; Schumann, T.; Limbach, F.; Stoica, T.; Calarco, R. Appl. Phys. Lett. 2011, 98, 103102. (28) Landré, O.; Songmuang, R.; Renard, J.; Bellet-Amalric, E.; Renevier, H.; Daudin, B. Appl. Phys. Lett. 2008, 93, 183109. (29) Wölz, M.; Kaganer, V. M.; Brandt, O.; Geelhaar, L.; Riechert, H. Appl. Phys. Lett. 2011, 98, 261907. (30) (a) Knelangen, M.; Hanke, M.; Luna, E.; Schrottke, L.; Brandt, O.; Trampert, A. Nanotechnology 2011, 22, 365703. (b) Niu, X.; Stringfellow, G. B.; Liu, F. Appl. Phys. Lett. 2011, 99, 213102. (c) Tourbot, G.; Bougerol, C.; Glas, F.; Zagonel, L. F.; Mahfoud, Z.; Meuret, S.; Gilet, P.; Kociak, M.; Gayral, B.; Daudin, B. Nanotechnology 2012, 23, 135703. (31) Koblmüller, G.; Pongratz, P.; Averbeck, R.; Riechert, H. Appl. Phys. Lett. 2002, 80, 2281−2283. (32) Wölz, M.; Kaganer, V. M.; Brandt, O.; Geelhaar, L.; Riechert, H. Appl. Phys. Lett. 2012, 100, 179902.

on the InxGa1−xN composition during the growth would lead to a nonuniform distribution of In within the NW.30 We are limited to a description of the average composition, because the characterization methods employed here, QMS and XRD, probe the entire InxGa1−xN NW ensemble. The second limitation is that the present data for N-rich growth do not permit us to disentangle the activation energies for the processes of decomposition, desorption, and reincorporation. The discussion of the atomic processes implies that growth experiments under metal-rich conditions could help to quantify the individual activation energies, but such conditions would prevent the growth of NWs. Our observations have important consequences for the development of growth strategies. The alloy formation can be regulated via the substrate temperature and controlled in situ by QMS.

■

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS We would like to thank A. Giussani for a critical reading of the manuscript. We thank C. Herrmann, G. Jaschke, and H.-P. Schönherr for their dedicated maintenance of the MBE system; A.-K. Bluhm for the scanning electron micrographs; and W. Anders, B. Drescher, and W. Seidel for the fabrication of the NW LED devices. This work has been partly funded by the German government BMBF project MONALISA (Contract no. 01BL0810).

■

REFERENCES

(1) Morkoç, H. Handbook of Nitride Semiconductors and Devices; WILEY-VCH Verlag: Weinheim, Germany, 2008; Vol. 1 (Material Properties, Physics and Growth). (2) Björk, M. T.; Ohlsson, B. J.; Sass, T.; Persson, A. I.; Thelander, C.; Magnusson, M. H.; Deppert, K.; Wallenberg, L. R.; Samuelson, L. Appl. Phys. Lett. 2002, 80, 1058−1060. (3) Ertekin, E.; Greaney, P. A.; Chrzan, D. C.; Sands, T. D. J. Appl. Phys. 2005, 97, 114325. (4) Geelhaar, L.; et al. IEEE J. Sel. Top. Quantum Electron. 2011, 17, 878. (5) (a) Kikuchi, A.; Kawai, M.; Tada, M.; Kishino, K. Jpn. J. Appl. Phys., Part 1 2004, 43, L1524−L1526. (b) Bavencove, A.-L.; Tourbot, G.; Pougeoise, E.; Garcia, J.; Gilet, P.; Levy, F.; André, B.; Feuillet, G.; Gayral, B.; Daudin, B.; Dang, L. S. Phys. Status Solidi A 2010, 207, 1425−1427. (c) Guo, W.; Zhang, M.; Banerjee, A.; Bhattacharya, P. Nano Lett. 2010, 10, 3355. (d) Armitage, R.; Tsubaki, K. Nanotechnology 2010, 21, 195202. (e) Lin, H. W.; Lu, Y. J.; Chen, H. Y.; Lee, H. M.; Gwo, S. Appl. Phys. Lett. 2010, 97, 073101. (f) Kishino, K.; Kamimura, J.; Kamiyama, K. Appl. Phys. Express 2012, 5, 031001. (6) Nguyen, H. P. T.; Cui, K.; Zhang, S.; Fathololoumi, S.; Mi, Z. Nanotechnology 2011, 22, 445202. (7) Limbach, F.; Hauswald, C.; Lähnemann, J.; Wölz, M.; Brandt, O.; Jahn, U.; Trampert, A.; Calarco, R.; Geelhaar, L.; Riechert, H. Nanotechnology 2012, 23, 465301. (8) Kuykendall, T.; Ulrich, P.; Aloni, S.; Yang, P. Nat. Mater. 2007, 6, 951−956. (9) Calleja, E.; Ristić, J.; Fernández-Garrido, S.; Cerutti, L.; SánchezGarcía, M. A.; Grandal, J.; Trampert, A.; Jahn, U.; Sánchez, G.; Griol, A.; Sánchez, B. Phys. Status Solidi B 2007, 244, 2816−2837. (10) Sekiguchi, H.; Kishino, K.; Kikuchi, A. Appl. Phys. Lett. 2010, 96, 231104. 5691



■

Article

NOTE ADDED AFTER ASAP PUBLICATION This paper was published on the Web on October 23, 2012, with incomplete information for Reference 7. The corrected version was reposted October 25, 2012.

5692


GaN Nanowire Heterostructures

Recommend Documents