InAs Core–Shell Nanowires

Oct 3, 2012 - Giant Magnetoconductance Oscillations in Hybrid Superconductor−Semiconductor Core/Shell Nanowire Devices. Ö. Gül , H. Y. Günel , H. Lüth...
5 downloads 0 Views 1MB Size
Letter pubs.acs.org/NanoLett

Molecular Beam Epitaxy Growth of GaAs/InAs Core−Shell Nanowires and Fabrication of InAs Nanotubes Torsten Rieger,*,†,∥ Martina Luysberg,‡,§ Thomas Schap̈ ers,†,∥,⊥ Detlev Grützmacher,†,∥ and Mihail Ion Lepsa†,∥ †

Peter Grünberg Institute 9, ‡Peter Grünberg Institute 5, and §Ernst Ruska-Centre (ER-C) for Microscopy and Spectroscopy with Electrons, Forschungszentrum Jülich, 52425 Jülich, Germany ∥ Jülich Aachen Research Alliance for Fundamentals of Future Information Technology (JARA-FIT), Germany ⊥ II. Institute of Physics, RWTH Aachen University, 52074 Aachen, Germany S Supporting Information *

ABSTRACT: We present results about the growth of GaAs/ InAs core−shell nanowires (NWs) using molecular beam epitaxy. The core is grown via the Ga droplet-assisted growth mechanism. For a homogeneous growth of the InAs shell, the As4 flux and substrate temperature are critical. The shell growth starts with InAs islands along the NW core, which increase in time and merge giving finally a continuous and smooth layer. At the top of the NWs, a small part of the core is free of InAs indicating a crystal phase selective growth. This allows a precise measurement of the shell thickness and the fabrication of InAs nanotubes by selective etching. The strain relaxation in the shell occurs mainly via the formation of misfit dislocations and saturates at ∼80%. Additionally, other types of defects are observed, namely stacking faults transferred from the core or formed in the shell, and threading dislocations. KEYWORDS: GaAs/InAs core−shell nanowire, misfit dislocation, InAs nanotube, MBE

N

be much more difficult.10 Using Au-assisted NW growth, core− shell structures suffer from an additional axial growth through the Au-droplet.11−13 Therefore, self-catalyzed NWs are superior for core−shell structures since either there is no droplet14,15 or it can be consumed after the growth of the core.16 The lattice mismatch between bulk GaAs and bulk InAs is 7%. In the case of layer growth, the relaxation depends on the orientation: on GaAs (001) relaxation occurs via the formation of quantum dots17 while misfit dislocations are created on GaAs (110),18 (111)B,19 and (111)A.20 For NWs, calculations have shown that there is a critical core radius below which coherent core−shell NWs can be grown independent of their shell thickness.21 So, Ge/Si core−shell NWs with a lattice mismatch of 4.2% were grown without dislocations.22 However, for GaAs/InAs core−shell NWs, this critical radius would be in the range of few atoms.21 Apart from misfit dislocations, core−shell NWs can show also a roughening as a result of strain relaxation.22,23 In this work, we present a detailed study about the growth of GaAs/InAs core−shell NWs using molecular beam epitaxy (MBE). Defects in the InAs shell as a result of the lattice

anowires (NWs) offer new possibilities for the growth of heterostructures. On one hand, due to the large surface to volume ratio, it is expected that highly lattice-mismatched semiconductor material systems can be combined. On the other hand, the NW geometry permits the growth of either axial or radial heterostructures. Axial heterostructures have been used to position a quantum dot inside a NW1 or to obtain controlled potential barriers for the electronic transport.2 Coaxially grown shells have been used to passivate the NW3,4 or to form prismatic quantum heterostructures.5 In the simple core−shell geometry, the GaAs core passivation with AlGaAs leads to an enhanced photoluminescence intensity3 while the passivation of an InAs core with InP increases the electron field effect mobility.4 However, the core can also be used as material support for a highly conductive shell, that is, a GaAs core with an InAs shell. So far, this material combination and geometry were not well investigated. Paladugu et al. have obtained InAs rings around GaAs NWs having {211} side facets.6 These rings grew site-selective in concave regions along the NWs corresponding to the positions of the rotational twins. The reverse arrangement, that is InAs/GaAs core−shell, was investigated using Au-catalyzed wurtzite InAs NWs.7,8 Recently, Uccelli et al. have grown InAs quantum dots on the side facets of GaAs NWs using a thin interlayer of AlAs.9 However, the growth of uniform InAs shells around GaAs NWs was found to © 2012 American Chemical Society

Received: July 6, 2012 Revised: September 21, 2012 Published: October 3, 2012 5559

dx.doi.org/10.1021/nl302502b | Nano Lett. 2012, 12, 5559−5564

Nano Letters

Letter

Figure 1. SEM micrographs showing the growth of the InAs shell at different substrate temperatures and As4 BEPs. For constant As4 BEP of 10−5 Torr, the growth temperature was decreased from 530 (a) to 490 (b), 460 (c), and 430 °C (d). Keeping the substrate temperature constant at 490 °C, the As4 BEP was varied between 2 × 10−5 Torr (e) and 0.6 × 10−6 Torr (h,i). The best morphology of the core−shell NWs is shown in panel g. The inset focus on the top segment of the NWs with the uncovered part. The scale bar is 200 nm except for the inset where it is 100 nm.

shell growth investigations, the In rate was 0.1 μm/h and the shell growth time was 15 min, unless otherwise mentioned. SEM micrographs of NWs obtained in different growth conditions are shown in Figure 1. As seen, at high As4 BEPs of around 1 × 10−5 Torr, the NWs either break along the growth axis for 530 °C substrate temperature (see Figure 1a) or the growth takes place only in few positions along the NW for 460 and 430 °C substrate temperatures (see Figures 1c,d). However, a continuous and longer segment of InAs shell has been obtained when the substrate temperature was 490 °C. Considering this growth temperature, the As4 BEP was varied between 2 × 10−5 and 6 × 10−7 Torr. From Figure 1e−i, it can be seen that the best morphology, namely a continuous and smooth InAs layer along the whole NW, is achieved for an As4 BEP of 1 × 10−6 Torr, corresponding to a V/III ratio of 2 (see Figure 1g). At higher As4 BEP, 5 × 10−6 and 2 × 10−5 Torr, the core is not entirely covered by the shell (see Figure 1e,f) and at lower As4 BEP, 6 × 10−7 Torr, the shell surface is rough or incomplete (see Figure 1h,i). The optimum As4 BEP is the same as that used for the growth of the GaAs core. Therefore, only the substrate temperature had to be changed for the shell growth. The core−shell structure is confirmed by an EDX profile shown in the Supporting Information, Section 5. Summarizing, the best shell morphology was achieved at a substrate temperature of 490 °C, In rate of 0.1 μm/h and an As4 BEP of 10−6 Torr (see Supporting Information for a detailed list of the optimum growth parameters). Using these conditions, apart from the core−shell NWs, few thin and very short InAs NWs are found on the sample (see Figure 1g). At the top of the core−shell NWs, a characteristic shape can be seen (inset in Figure 1g and Supporting Information, Figure S1). Here, the lateral end part of the GaAs core is not covered by InAs. This was confirmed by EDX (see Supporting Information, Section 2). The top facet of the core is again covered with InAs, which has finally a triangular prismatic shape (see Inset in Figure 1g). The fact that InAs does not grow on the small end part of the core can be attributed to different crystallographic arrangements in the core: almost the entire GaAs NW has ZB crystal structure with only few rotational twins only at the top, a small part is WZ.24 This crystal phase

mismatch and the growth mechanism are described. Information about the strain relaxation in the shell is also reported. Additionally, we show a way for the fabrication of InAs nanotubes. The growth of our NWs has been done in a solid-source Varian Gen II MBE system using an Au-free approach. First, self-catalyzed GaAs NWs were grown on GaAs (111)B substrates covered with ∼6 nm of SiOx. The SiOx was obtained by thermal treatment at 300 °C for 10 min of spin-coated hydrogen silsesquoxiane (HSQ) diluted with methylisobutylketone (MIBK).24 No etching of the oxide was performed prior to the growth. The GaAs cores were grown at 590 °C with a Ga rate of 0.075 μm/h and an As4 beam equivalent pressure (BEP) of 10−6 Torr for 45 min. Using these growth conditions, the NWs are around 1.2 μm in length and 60−70 nm in diameter. The SiOx thickness of 6 nm results in a NW density of around 0.1 NW/μm2. The NWs are perpendicular on the substrate and have a hexagonal prism morphology with {110} side facets. The crystal structure is zinc blende (ZB) with only very few rotational twins along the whole NW except a short segment at the NW top where it is wurtzite (WZ).24 The WZ structure of the top is related to the consumption of the droplet at the end of the growth.24,25 The NW tapering is negligible. After the GaAs NW growth, the Ga shutter was closed while the arsenic and the substrate temperature were kept at the same value. As a result, the catalyzing Ga droplet was consumed. Subsequently, the substrate temperature was decreased for the InAs shell growth. The morphology of the as-grown core−shell NWs was analyzed by scanning electron microscopy (SEM) (Zeiss Gemini 1550 operated at 20 kV). For the analyses of the crystal structure, the NWs were examined by transmission electron microscopy (TEM). Additionally, the atomic composition was evaluated by energy dispersive X-ray spectroscopy (EDX). Single EDX spectra and low-resolution images were acquired with a Philips CM200 TEM, EDX profiles with a FEI Tecnai G2 F20 TEM, and high-resolution images were taken with a FEI Titan 80−300 TEM. We have found that the growth of the InAs shell is very sensitive to the substrate temperature and the As4 BEP. In our 5560

dx.doi.org/10.1021/nl302502b | Nano Lett. 2012, 12, 5559−5564

Nano Letters

Letter

without an AlAs interlayer.9 This growth mode starting from islands, which afterward coalesce giving a closed layer, limits the lowest thickness of the InAs shell to few nm. Because of the high lattice mismatch between GaAs and InAs, the strain relaxation is expected to take place mainly by misfit dislocations.10,21 Dislocations are known to cause the relaxation in the growth of InAs on GaAs (110) substrates18,28 and they were already observed in InAs islands on GaAs NWs,9 InAs rings around GaAs NWs6 as well as InAs/GaAs core−shell NWs.7,8 When the GaAs core is covered by the InAs shell, Moiré fringes from the overlapping of the two crystal lattices can be seen in TEM micrographs (see Figure 3a). Only close to

selective growth results in a conducting shell with only ZB crystal structure as WZ core regions are not covered. A more detailed description of the crystal phase selective growth of InAs on GaAs is beyond the scope of this report and will be published elsewhere. However, it should be mentioned that differences in the radial growth on ZB and WZ core material were already reported.26 This characteristic of the NW top enables to measure the thickness of the shell with high accuracy already in the SEM and for every individual NW. Additionally, the core can be etched selectively resulting in InAs nanotubes (see below). Regarding the time evolution of the shell thickness, we have grown five samples with shell growth times ranging from 1 to 15 min. The shell thickness as a function of the growth time is plotted in Figure 2. On the same figure, the insets show SEM

Figure 2. Shell thickness as a function of the shell growth time. The solid line is a guide to the eye. The insets illustrate the evolution of the InAs shell in time: (a) 1, (b) 3, (c) 5, and (d) 15 min of shell growth time. The scale bars are 200 nm.

micrographs after 1 (a), 3 (b), 5 (c), and 15 min (d). The thickness of the shell increases almost linearly with time, the resulted shell growth rate being around 90 nm/h. This is much higher than it is expected from direct impingement on the side facets (∼20 nm/h) considering a planar growth rate of 100 nm/h. Since the oxide on the substrate is not covered by a parasitic GaAs or InAs layer, the difference is explained by In diffusion from the substrate as well as the secondary adsorption. Here, the secondary adsorption means that In adatoms impinging on the substrate (SiOx) desorb from it and are readsorbed on the NW side facets.24,27 Especially for low NW densities, this process can yield high additional fluxes. The SEM micrographs in the inset of Figure 2 show that the shell growth does not proceed in the layer-by-layer mode. Only when the shell growth time is 5 min or higher, the surface becomes smooth. Before, it is rough and several islands are observed along the NW (see Figure 2a). Subsequently, the islands coalesce between 3 and 5 min shell growth time (see Figure 2a,b). The islands are found along the entire NW and independent of rotational twins which are present in the crystal structure of the core (see Supporting Information, section 3). Thus, the shell nucleation does not start in concave regions as it is reported for GaAs core NWs with {211} side facets.6 Such InAs islands on GaAs NWs were already observed by Uccelli et al. when InAs quantum dots were grown on GaAs NWs

Figure 3. (a) TEM micrograph of a GaAs/InAs core−shell NW and (b) the corresponding selective area electron diffraction (SAED) pattern. The stripes in the TEM image are due to Moiré fringes. (c) Degree of axial strain relaxation for different shell/core ratio. The dashed line is a best fit to the data. Axial relaxation saturates at about 80%.

the top, these fringes are not observed since there is only pure GaAs. The two lattices are also identified by the double spots in the diffraction pattern in Figure 3b. The Moiré fringe period D allows the determination of the degree of relaxation. D is given by D = (d2d1)/(d2 − d1) where d1 and d2 are the distances between the (111) planes of the GaAs core and the InAs shell, respectively. Here, we assume that the core lattice is completely relaxed while the shell is strained. In this way, we have calculated the degree of relaxation as a function of the shell and core geometry, which is shown in Figure 3c. As expected, the relaxation increases with the shell thickness and depends on the shell/core ratio. The relaxation saturates at about 80%, which is comparable to results from InAs/GaAs core−shell NWs7 as well as InAs layers grown on GaAs (111)A.29 5561

dx.doi.org/10.1021/nl302502b | Nano Lett. 2012, 12, 5559−5564

Nano Letters

Letter

Figure 4. (a) TEM micrograph showing two types of stacking faults/rotational twins in GaAs/InAs core−shell NWs. The red arrows show stacking faults/rotational twins that are adopted from the core while the blue and green arrows point at stacking faults in the InAs shell due to misfit dislocations at the interface. Green arrows indicate loop dislocations. The blue rectangles mark discontinuous shifts in the Moiré fringes coming from the stacking faults in the shell. (b) HRTEM micrograph showing misfit dislocations at the GaAs/InAs interface. (c) Illustration of a perfect dislocation that does not give a stacking fault in the shell. (d) Illustration of a Frank partial dislocation that gives a stacking fault in the shell.

has been observed already in the small islands at the beginning of the shell growth (this investigation and ref 9). To explain this discrepancy, we have to consider that not only partial dislocations but also perfect dislocations are created. Indeed, the upper dislocation in Figure 4b is a perfect dislocation having the Burgers vector b = (a/2)[110]. Obviously, the crystal structure on both sides of this dislocation is perfect ZB which means that it does not create any stacking fault. This is also schematically illustrated in Figure 4c. Apart from that, correlated to the fact that almost no loop dislocations are observed, the lower density of stacking faults should be related also to the growth mechanism via islands and their coalescence. Our TEM analysis revealed also the presence of other kind of crystal defects, which are illustrated with TEM micrographs shown in Figure S5 from the Supporting Information, Section 4. Thus, we have observed stacking faults perpendicular to other ⟨111⟩ directions (see Supporting Information, Figure S5a). They are the result of Shockley partial dislocations with Burgers vector b = a/6 ⟨112⟩ and also contribute to the strain relaxation.32,33 Similar stacking faults were also found in InAs quantum dots on GaAs NWs,32 InAs grown on GaAs (110) substrates28 as well as Germanium islands on Si NWs.33 Moreover, careful examination of the Moiré fringes show also stacking faults not being perpendicular to any ⟨111⟩ direction (islands boundaries) as well as threading dislocations (see Supporting Information, Figure S5b). The stacking faults not being perpendicular to ⟨111⟩ directions are seen by discontinuous shifts in the Moiré fringes (inset (i) in Supporting Information, Figure S5b) while the threading dislocations are visible through terminating Moiré fringes (inset (ii) in Supporting Information, Figure S5b). Both kind of

The observation of the Moiré fringes using TEM already indicates the presence of dislocations.30,31 Analyzing the fringes and HRTEM micrographs we determined important information regarding the different defects in the InAs shell. As illustrated by the TEM micrograph in Figure 4a, two different types of stacking faults in the shell can be distinguished: stacking faults or twins adopted from the GaAs core (marked with red arrows) and stacking faults due to misfit dislocations (marked with blue and green arrows) at the core−shell interface, which are present only in the shell. In a simple picture, creating a dislocation at the interface means the removal of one atomic plane in the InAs lattice with respect to the GaAs lattice. Consequently, a stacking fault is created in the shell (see Figure 4d). Such a dislocation is a partial dislocation and it is identified in Figure 4b with the second one. Its Burgers vector b = (a/3)[111], as determined by a Burgers circuit, indicates a Frank partial dislocation. These Frank partial dislocations cannot glide; therefore, their formation hinders further strain relaxations and thus explains the strain saturation depicted in Figure 3d. However, as seen in Figure 4a, these stacking faults/partial dislocations are in general not around the entire NW, that is, they are not complete dislocation loops as they are proposed for both ZB and WZ core−shell NWs in refs 10 and 21. In the TEM micrograph depicted in Figure 4a, only two such loop dislocations are found and are pointed with green arrows. The average distance between these partial dislocations (more than 8 nm) is much larger than required for the observed degree of relaxation. Considering the bulk lattice constants of GaAs and InAs, a misfit dislocation at every 4−5 nm is expected.28 Additionally, a high density of misfit dislocations,