Nanoscale This article is licensed under a Creative Commons Attribution-NonCommercial 3.0 Unported Licence. View Article Online PAPER View Journal | View Issue GaAs nanoscale membranes: prospects for seamless integration of III–Vs on silicon† Open Access Article. Published on 09 December 2019. Downloaded on 2/11/2021 12:05:15 PM. Cite this: Nanoscale, 2020, 12, 815 Andrés M. Raya, ‡a,b Martin Friedl,‡a Sara Martí-Sánchez, ‡c Vladimir G. Dubrovskii,d Luca Francaviglia,a Benito Alén,b Nicholas Morgan,a Gözde Tütüncüoglu,a Quentin M. Ramasse,e,f David Fuster,b Jose M. Llorens,b Jordi Arbiol c,g and Anna Fontcuberta i Morral *a,h The growth of compound semiconductors on silicon has been widely sought after for decades, but reliable methods for defect-free combination of these materials have remained elusive. Recently, inter- connected GaAs nanoscale membranes have been used as templates for the scalable integration of nano- wire networks on III–V substrates. Here, we demonstrate how GaAs nanoscale membranes can be seam- lessly integrated on silicon by controlling the density of nuclei in the initial stages of growth. We also cor- Received 1st October 2019, relate the absence or presence of defects with the existence of a single or multiple nucleation regime for Accepted 8th December 2019 the single membranes. Certain defects exhibit well-differentiated spectroscopic features that we identify DOI: 10.1039/c9nr08453c with cathodoluminescence and micro-photoluminescence techniques. Overall, this work presents a new rsc.li/nanoscale approach for the seamless integration of compound semiconductors on silicon. 1. Introduction grate cutting-edge optoelectronic and high-speed electronic devices with existing silicon technology.8 III–V nanostructures have gained increasing attention over the Taking GaAs as an example, two important factors make last years due to their prospects in a wide range of applications the growth of GaAs on Si substrates a challenge. First, the such as high-speed electronics,1 optoelectronics,2–4 thermo- 4.2% lattice mismatch between Si and GaAs makes the defect- electrics5 and photovoltaics.6,7 Their small dimensions and tai- free growth of GaAs very difficult. Nonetheless, recent work lored shapes can be used to modify the inherent semi- suggests that such large lattice mismatches can be overcome conductor properties, thus extending their range of appli- through the appropriate interface engineering to form a peri- cation. Yet, high substrate and processing costs mean that III– odic misfit array that plastically relaxes the mismatch V devices are generally only cost-effective in specific, high-per- strain.9–11 The second challenge is the formation of anti-phase formance applications. The synthesis of III–V materials on Si boundarys (APBs).8 When a binary compound semiconductor substrates, however, could be a cost-competitive way to inte- (such as GaAs) is deposited on top of a monoatomic semi- conductor (such as Si), both A-polar and B-polar GaAs islands can nucleate. In the mononuclear regime, where each GaAs a Laboratoire des Matériaux Semiconducteurs, Institute of Materials, Faculty of monolayer is nucleated at a single point, the whole grown Engineering, École Polytechnique Fédérale de Lausanne, EPFL, 1015 Lausanne, crystal will have the same polarity. However, in the polynuclear Switzerland. E-mail: anna.fontcuberta-morral@epfl.ch b regime, where many islands nucleate and coalesce to form the Instituto de Micro y Nanotecnología, IMN-CNM, CSIC (CEI UAM+CSIC), Isaac Newton, 8, 28760 Tres Cantos, Madrid, Spain final crystal, the polarity mismatch between islands creates c Catalan Institute of Nanoscience and Nanotechnology (ICN2), CSIC and BIST, APBs.12,13 These APBs manifest as regions of As–As or Ga–Ga Campus UAB, Bellaterra, 08193 Barcelona, Catalonia, Spain bonds. An additional source for such APBs, as will be dis- d ITMO University, Kronverkskiy pr. 49, 197101 St. Petersburg, Russia cussed, is through multiple twinning processes.14 These APBs e SuperSTEM Laboratory, SciTech Daresbury Campus, Daresbury WA4 4AD, UK f are shown to be optically and electrically detrimental15 and School of Chemical and Process Engineering, and School of Physics, University of Leeds, Leeds LS2 9JT, UK thus it is imperative that their formation is suppressed. g ICREA, Pg. Lluís Companys 23, 08010 Barcelona, Catalonia, Spain To overcome these challenges, many approaches have been h Institute of Physics, Faculty of Basic Sciences, École Polytechnique Fédérale de explored both for III–V thin-film and nanostructure growth on Lausanne, EPFL, 1015 Lausanne, Switzerland silicon. III–V thin films on silicon have been achieved through † Electronic supplementary information (ESI) available: EELS elemental maps strained superlattice growth,16 buffer growth,17,18 thermal and strain analyses of AlGaAs-passivated NMs, detailed results of the pre-growth treatments, PL and TRPL fitting model. See DOI: 10.1039/C9NR08453C annealing13,19,20 and substrate patterning.21–24 Additionally, ‡ Equal contribution. III–V nanostructure growth on silicon25,26 has also been This journal is © The Royal Society of Chemistry 2020 Nanoscale, 2020, 12, 815–824 | 815 View Article Online Paper Nanoscale explored with the growth of vertical nanowires (NWs) on silicon25–29 or in-plane NWs and ridges.22,30 Several studies have also explored the use of aspect-ratio trapping to reduce the vertical propagation of defects formed at the substrate This article is licensed under a Creative Commons Attribution-NonCommercial 3.0 Unported Licence. interface.13,31–33 One particularly interesting approach, which has been used successfully to overcome challenging substrate/nanostructure mismatch conditions in the past, is selective-area epitaxy (SAE) or selective-area growth (SAG). SAE and SAG have proven to be Open Access Article. Published on 09 December 2019. Downloaded on 2/11/2021 12:05:15 PM. effective techniques for the synthesis of III–V nanostructures such as NWs,34,35 tripods,36 and V-shaped nanofins.37,38 SAE is carried out on nanopatterned substrates at high temperature in such a way that the sticking coefficient of the adatoms is zero on the mask and non-zero in the etched openings.39 The growth of GaAs nanomembranes (NMs) on GaAs (111)B sub- strates has been achieved by etching long slits along the 〈112 ˉ〉 family of directions, as reported for both metalorganic chemi- cal vapour deposition (MOCVD)40 and molecular beam epitaxy (MBE).41 The shape of these NMs is primarily driven by the growth kinetics.42 They exhibit a defect-free crystal structure which leads to a high homogeneity in the optical properties of NM ensembles.43 Their high aspect-ratio geometry combined with their pristine crystal quality enables their use as tem- plates for growing highly mismatched materials,44 extending Fig. 1 (a) Schematic of a patterned SiO2/Si substrate ready for MBE their functionality and design opportunities.45 growth. (b) Schematic of the SAE growth of NMs. (c–d) Tilted scanning electron microscopy (SEM) images after 60 min of GaAs NM growth In the present study, our particular interest is in exploring without gallium pre-deposition on Si (111) for long slits (c) and short slits (d), how GaAs NMs can be used to address some of the major chal- with APBs circled in red. (e) Top-view SEM image of NMs grown by 30 s Ga lenges associated with integrating III–Vs on silicon substrates. pre-deposition + 5 min anneal + NM growth showing preferential crystal We demonstrate the growth of GaAs NMs on Si (111) by selec- orientation. (f) Summary of different growth recipes and resulting percen- tive-area MBE in both majority-mononuclear and majority- tage of aligned crystals. (g–h) Top-view SEM images after 30 s of gallium pre-deposition, followed by 2 min and 15 min of GaAs growth, respectively. polynuclear regimes by varying slit length. It is demonstrated that the growth of largely defect-free, single crystal NMs is possible for openings shorter than one micrometer. For SAE openings longer than one micrometer, poly-nucleation and the on GaAs substrates,41 the NMs grown on Si do not completely consequent APB formation significantly limits the growth of fill the openings for longer slits, as seen in Fig. 1(c). Instead, defect-free NMs. We study the influence of different growth we observe several separate NMs merging. This merging often conditions on the crystal structure of these NMs and develop a forms visible defects whose shape in SEM resembles APBs in growth model that explains the data. We then analyse the GaAs thin films on Si.12 While the defect density remains con- optical properties of the NMs by cathodoluminescence (CL) stant, the total number of defects per NM is greatly reduced and μ-photoluminescence (PL) and directly correlate them for shorter slits. We deduce that for shorter slits the growth with localized defects identified by scanning transmission moves towards the mononuclear regime where the merging of electron microscopy (STEM). multiple polarity-mismatched nuclei is strongly reduced. In order to increase the crystalline quality of the NMs, we experimented with various substrate treatments before growth. 2. Results and discussion To foster a single polarity in the seeds, growths were preceded by a soaking period in Ga or As4 flux followed by an annealing GaAs NMs were grown by MBE on patterned Si (111) substrates step. The percentage of preferentially-aligned nuclei was calcu- with 30 nm of thermally grown oxide. Stripes were patterned lated with the help of SEM imaging in a region with short, with widths in the range of 60–80 nm, and lengths of up to 180 nm-long slits, as shown in Fig. 1(e). Here, the NM growth several microns. A scheme of the substrate prior to and after was preceded by a 30 s gallium exposure followed by a 5 min growth is depicted in Fig. 1(a) and (b), respectively. Fig. 1(c) anneal and shows a clear enhancement of the left-pointing and (d) show a tilted SEM image of a typical GaAs NM array NMs, marked in blue. The table shown in Fig. 1(f ) summarizes obtained after 60 min of growth in long and short slits, the effect of various substrate treatments that were performed respectively. We observe that GaAs grows only in the longitudi- and the resulting percentage of aligned NMs. The preferential nal openings in the SiO2 mask with a high degree of selectivity, alignment is given by the number of NMs whose long axis as reported for GaAs substrates.39,41,46 Compared to the growth points towards the 〈112 ˉ〉 direction, divided by the total 816 | Nanoscale, 2020, 12, 815–824 This journal is © The Royal Society of Chemistry 2020 View Article Online Nanoscale Paper number of NMs considered for the analysis. A minimum of As mentioned previously, the coalescence of oppositely- 100 NMs per sample were checked for the quantification to oriented structures, due to the presence of polynucleated ensure a large enough sample size. islands, is what causes the formation of kinks such as those From the table in Fig. 1(f ) we see that without any pre-treat- circled in red in Fig. 1(c) and (d). This article is licensed under a Creative Commons Attribution-NonCommercial 3.0 Unported Licence. ment, the NMs are randomly orientated, split about 50/50 We now provide a model that explains some important fea- between the two directions. Preferential alignment increases tures of nucleation and growth of GaAs NMs on Si with with Ga pre-deposition time, and is further enhanced by gallium pre-deposition. This process consists of two steps, the adding an annealing step after pre-deposition. Interestingly, it nucleation of gallium droplets and subsequent growth (with was observed that As4 pre-deposition also led to preferential continued nucleation) of GaAs NMs starting from these dro- Open Access Article. Published on 09 December 2019. Downloaded on 2/11/2021 12:05:15 PM. alignment. However, due to the vastly different surface plets. We first consider the nucleation of gallium droplets in dynamics of the two growth species, as well as the large body Si/SiO2 slits during the gallium pre-deposition step. The of scientific literature about Ga-catalysed nanostructure growth number of gallium atoms arriving per unit length of the slit from the NW community, in this report we will focus on the per unit time equals IW with I = 0.664 nm2 s−1 as the atomic gallium exposure strategy with the As pre-deposition approach flux of gallium (equivalent to the growth rate of 0.3 Å s−1), and to be addressed in a separate paper. W as the slit width (60–80 nm). In principle, the I value can We further explored if the gallium droplets formed on the also include the indirect gallium flux originating from the surface could drive the nucleation down to the mononuclear adatom diffusion on the oxide, without changing the model growth regime (as in the case of self-catalysed NW growth),27 equations and conclusions. However, due to the high substrate thus avoiding APB formation. SEM micrographs of samples on temperature of 630 °C, the indirect flux is estimated to be only which a 30 s gallium pre-deposition was performed followed about 10% of the direct flux and has thus been neglected in by 2 min and 15 min of GaAs growth, are shown in Fig. 1(g) the numerical estimates.46 Taking into account the rotation of and (h), respectively. Before GaAs growth, it was observed that the sample, one side of the slit receives the incoming flux the slits seem mostly empty except for a few nanoscale droplets during half of the rotation period. We can thus say that one- around the edges, in contact with the SiO2. After 2 min of half of the incoming flux arrives to each side of the slit and GaAs growth, multiple GaAs crystals with an average diameter then diffuses along the side. The coalescence of these of about 50 nm begin to nucleate in the same area. The diffusing adatoms leads to the nucleation of gallium droplets, nucleation of many islands during the early stages of growth is as shown in Fig. 2(a). Assuming that the gallium adatoms have consistent with the polynuclear growth regime. Such a regime a concentration n1 per unit length at a distance λ from the has also been observed in the homoepitaxial growth of NMs droplet surface (at x = r + λ) and zero concentration at the on (111) GaAs substrates.47 droplet surface (at x = r) with r as the radius of the droplet We observe that GaAs islands nucleate preferentially at the base, the adatom concentration at the borders of the slit interface with the SiO2. It suggests that this corner offers the should equal n(x) = (n1/λ)(x − r). Here, x is the coordinate along lowest-energy position for crystal nucleation, which is also the the slit and λ plays the role of the effective diffusion length of case for NWs.27 This is in contrast to the growth on GaAs sub- gallium adatoms at the Si/SiO2 edge. The growth rate of the strates where the nucleation occurs inside the slits, rather droplet equals Ddn/dx|x = r = Dn1/λ, with D as the diffusion than at their edges.47 This confirms that the heteroepitaxial interface has a much higher energy than the homoepitaxial interface and drives nucleation to favour the corners at the edges of the slit. The corner nucleation occurs despite the for- mation of a relatively high-energy GaAs/SiO2 interface, which is not favoured when growing on GaAs substrates. Furthermore, as the gallium droplets observed before growth also seem to favour the edges of the slits, it is possible that GaAs islands nucleate within the gallium droplets. These crystal nuclei then grow relatively quickly, forming small tetra- hedra with their long axis aligned either along the [112 ˉ] or ˉˉ [112] direction after only a few minutes of growth. The orien- tation of the first layers of the crystal can be affected by the for- mation of twins which results in these two possible tetrahedral orientations.48,49 In the absence of twins, the orientation of these tetrahedra is directly related to the crystal structure of the substrate. Once the tetrahedra reach the edges of the open- ings, they grow upwards, developing vertical (110) side Fig. 2 (a) Illustration of the model for nucleation of gallium droplets in facets.40,42 The formation of such vertical facets starts to be the slits; (b–d) experimental size distributions of gallium droplets, all clear after 15 min of growth, as seen in Fig. 1(h). Eventually, GaAs structures after 2 min of growth and GaAs NMs after 15 min of the crystals merge yielding a single, continuous NM in the slit. growth (histograms), fitted by eqn (2) (blue curves). This journal is © The Royal Society of Chemistry 2020 Nanoscale, 2020, 12, 815–824 | 817 View Article Online Paper Nanoscale coefficient of gallium adatoms near the Si/SiO2 corner. The NMs, is higher than for spherical cap droplets. It increases to adatom concentration changes with time t according to dn1/dt 25 585 nm2 after 15 min of GaAs growth, where all nanocrys- ≅ (IW/2 − Dn1N/λ) → 0, with N the droplet density per unit tals acquire the energetically preferred shapes of elongated length of the slit. The n1 must tend to zero for long enough t NMs. Blue curves in Fig. 2(c) and (d) show the fits to the size This article is licensed under a Creative Commons Attribution-NonCommercial 3.0 Unported Licence. due to the balance of the arrival rate and growth of gallium dro- histograms of GaAs nanocrystals and NMs by eqn (2), with a = plets.50 On the other hand, the droplet density changes with 1.9 for t = 2 min and a = 1.2 for t = 15 min. The blue curve in time according to dN/dt = Dn12/λ that their nucleation rate Fig. 2(b) shows that eqn (2) can also be used for fitting the size equals the growth rate times the adatom concentration.50,51 distribution of the initial gallium droplets, with the best fit Solving this equation, we obtain the droplet density ( per obtained at a = 3.2. All these values of a are on the order of Open Access Article. Published on 09 December 2019. Downloaded on 2/11/2021 12:05:15 PM. unit length of the slit) in the form: unity, corresponding to the size distribution maxima at small  2=3   surface areas and long right tails of the distributions. Overall, IW 3λð1 þ δÞ 1=3 1=3 our growth model reveals the kinetic mechanisms that lead to N¼ t ð1Þ 2 D the formation of GaAs NMs in the Si/SiO2 slits. It properly describes the nucleation of droplets at the slit edges, As-con- Here, δ accounts for the fraction of GaAs crystals that form trolled crystallization of GaAs crystals from these droplets, without droplets, with a different effective distance between their further evolution to the elongated NMs and the broad the emerging islands. Clearly, this δ should be minimized to size distributions of all objects. According to our eqn (1), the yield a more mononuclear growth regime. This problem number of nuclei in each slit can be reduced by growing in requires a separate study which is beyond the scope of this thinner slits (smaller W), increasing the gallium diffusion work. Within our model, N increases with the gallium pre- coefficient D by increasing the growth temperature, and depos- deposition time and becomes smaller for lower deposition iting gallium for shorter time t. Scenarios that are more flux, narrower slit and higher diffusivity of gallium adatoms, complex can also be considered, such as two-step growth with which is reasonable. Mononucleation of a single gallium higher temperature for nucleation of gallium droplets and droplet in each slit would be reached when the average dis- lower temperature for growing GaAs. Outcomes of eqn (1) thus tance between the droplets becomes equal to the slit length. In reveal approaches that can be used in the future for minimiz- particular, statistical analysis of the droplet size distribution ation of crystallographic defects resulting from the coalescence after 30 s of gallium pre-deposition reveals the average droplet of the NMs. density N = 0.375 μm−1. Using eqn (1), this gives an estimate We now turn to the description of correlated CL, μ-PL and D/[λ(1 + δ)] = 5 cm s−1. STEM measurements used to gain insight into the properties Size distributions of gallium droplets after 30 s of gallium of GaAs NMs, specifically focusing on the effects of defects pre-deposition, GaAs structures after 2 min of GaAs growth, that arise due to their growth on non-polar Si substrates. Two and GaAs NMs after 15 min of GaAs growth over the slit area side-by-side AlGaAs passivated nanostructures were selected are shown in Fig. 2(b)–(d). This data was obtained by statistical due to their structural dissimilarity, as observed by SEM, analysis of SEM images and it shows that the droplets exhibit shown in Fig. 3(b). NM A was chosen as it represents the a large size inhomogeneity in all cases. To understand this majority of the NMs that are believed to grow in the mono- feature, we note that, at least for GaAs crystals emerging from nuclear regime and appear defect-free. In contrast, NM B was gallium droplets, the volume of the crystals increases due to chosen specifically due to the presence of a kinked surface direct impingement of As4 onto the droplet surface. Hence, suggesting the presence of defects which we interpret to be the number of GaAs pairs in the crystals increases with their due to a statistically unlikely polynucleated growth. The com- size as (a + s), where s = i2/3 is the appropriately normalized parison between their luminescence and structural properties dimensionless surface area and a is the dimensionless nuclea- provide insight into the behaviour of such defects in GaAs tion probability of GaAs crystals at s = 0. We can equivalently nanostructures grown on Si. present the growth rate in terms of s in the form ds/dt = (a + s)/τ, In order to increase the luminescence yield of the NMs, with τ as a characteristic growth time. According to ref. 52–54 their surface was passivated with a 41 nm Al0.3Ga0.7As shell. such a size-linear growth rate leads to the Polya-like distri- The shell growth was performed at 440 °C under a high As4 bution over the surface area of the structures S. flux similar to Heigoldt et al.55 Fig. 3(a) and (b) show SEM  a1   micrographs of the structures before and after AlGaAs shell aa 1 S S f ðS; hSiÞ ¼ C exp a ð2Þ growth, respectively. Interestingly, a new (111)B facet appears ΓðaÞ hSi hSi hSi at the top of one of the NMs, with two small {113} inclined where 〈S〉 is the average surface area occupied by the crystals facets on either side of it. A rough AlGaAs layer covers the in a given sample and C is a constant. whole mask surface as it was deposited at a low temperature The measured average surface area of gallium droplets after for which mask selectivity is lost.46 30 s of gallium pre-deposition is 1053 nm2. The average Low-temperature CL and μ-PL spectroscopy was performed surface area of the GaAs nanocrystals emerging from these to characterize their optical properties. While the lumine- droplets decreases to 992 nm2 after 2 min of GaAs growth, scence of GaAs nanoscale membranes on GaAs substrates is showing that the aspect ratio of the nanocrystals, including consistent with their absolute defect-free nature,43 the lumine- 818 | Nanoscale, 2020, 12, 815–824 This journal is © The Royal Society of Chemistry 2020 View Article Online Nanoscale Paper probably due to the highly excited regime of CL which leads to band filling.56 In contrast, the emission from NM B [green and blue curves in Fig. 3(c)] exhibits both wavelength and intensity variation along its length. Two spatially-resolved and spec- This article is licensed under a Creative Commons Attribution-NonCommercial 3.0 Unported Licence. trally-distinct emission regions are present on each side of the NM and are separated by a weaker region at its centre: one side of NM B emits at 817 nm and the other at 827 nm, with emis- sion energy below the band edge. The asymmetric shape of the green curve reveals the presence of a weaker component at Open Access Article. Published on 09 December 2019. Downloaded on 2/11/2021 12:05:15 PM. lower energies. Both signals from NM B are weaker and shifted towards longer wavelengths with respect to NM A. This could be attributed to a defect-mediated mechanism.57–59 Fig. 3(d) shows the μ-PL spectra corresponding to the maximum PL intensity of each NM, acquired at 6 K. In our μ-PL setup, the spot size of the excitation laser is comparable to the size of the NMs. Consequently, although μ-PL can spatially resolve the signal from the two NMs, each μ-PL spec- trum is the convolution of the total emission from each NM. In NM A, the emission is centered at 824 nm, with a full width at half maximum (FWHM) of 25 nm. The emission energy here is consistent with the band edge emission of GaAs at low temperatures. The FWHM is quite broad, especially compared to the values obtained in homoepitaxial GaAs NMs, which is narrower by approximately one order of magnitude.41,43 This broadening is attributed to the presence of twinning defects, as confirmed by transmission electron microscopy (TEM) measurements. The emission of NM B is centered at 833 nm. This spectrum Fig. 3 (a) Tilted SEM image of GaAs NMs after growth in 800 nm-long can be fitted by three different Gaussian contributions: a slits (b) tilted SEM image of NMs A and B which were passivated with an central peak at 836.0 ± 0.5 nm, surrounded by another at 826 ± AlGaAs shell for CL/PL analysis (c) CL spectra corresponding to different positions on NMs A and B. The inset shows an SEM image of the investi- 2 nm and a less intense one at 851 ± 3 nm. The μ-PL emission gated region and its CL map. The value of the GaAs bandgap at 12 K is is consistent with the sum of the two CL signals. The highly marked by the vertical dashed line. (d) μ-PL spectra obtained at the excited regime of the CL leads to band filling effects which are point of each NM where the intensity was strongest. PL map of the seen here as a blue shift of the CL emission with respect to intensity distribution is shown in the inset. (e) TRPL measurements of that of the PL.56 NM A and B revealing two distinguishable lifetime components. The points represent the experimental data while the solid curves show the Time-resolved photoluminescence (TRPL) measurements fits to the data obtained by the convolution of the instrument response from each NM are shown in Fig. 3(e), revealing two different function with the double exponential functions plotted in the inset. lifetime components. This can be seen more clearly when plot- ting the time evolution on a logarithmic scale, as shown in the inset. The longer-lived components have lifetimes of about scence of GaAs nanoscale membranes on Si exhibits a richer 1065 ps for NM A and 723 ps for NM B, while the shorter ones structure. We show the results on these two structures, which have lifetimes of 41 ps and 30 ps, respectively. Details on the are consistent with measurements on the rest of the mem- fitting procedure and the effective lifetime calculation are given branes of the array. Fig. 3(c) shows CL spectra obtained at in the ESI.† The short-lifetime pathway contributes much more different positions on the two NMs. A SEM image of the ana- to the PL decay in NM B than in NM A. This indicates a more lysed region is shown in the inset along with a CL map of the defective structure in NM B and is consistent with its previously same area, where the emission intensity at each point is corre- discussed red-shifted multi-component luminescence. lated with its position in the SEM image. Each colour in the We turn now to the structural characterization of NMs A map represents the relative intensity detected at the selected and B. After the optical characterization, a focused ion beam emission wavelengths. Red, green and blue colours correspond (FIB) cross-section containing both NMs was prepared. The to the intensity measured at 812 nm, 817 nm and 827 nm, NMs were then analysed by high-resolution transmission elec- respectively. tron microscopy (HR-TEM). The results of these analyses are The emission from NM A [red curve in Fig. 3(c)] is uniform summarized in Fig. 4. The top of the figure shows general view along the whole structure. The spectrum is centred at 812 nm. HR-TEM micrographs. Atomically resolved high-resolution This is slightly blue shifted with respect to the band edge high-angle annular dark-field (HAADF)-STEM image details emission of GaAs at 820 nm (dashed line in Fig. 3(c)), most are shown in the lower part of the figure. This journal is © The Royal Society of Chemistry 2020 Nanoscale, 2020, 12, 815–824 | 819 View Article Online Paper Nanoscale This article is licensed under a Creative Commons Attribution-NonCommercial 3.0 Unported Licence. Open Access Article. Published on 09 December 2019. Downloaded on 2/11/2021 12:05:15 PM. Fig. 4 Microstructural characterization of NMs A and B. (a, b) Low-magnification HR-TEM micrographs revealing the overall defect distribution in both NMs. The insets show the fast Fourier transforms (FFTs) corresponding to the regions marked by the red and violet squares, respectively. Aberration-corrected atomically resolved HAADF-STEM micrographs obtained on NM A showing: (c) twinned layer (twins marked in red), (d) defect- free area, (g) zoom from (d) and (h) intensity profile obtained along the blue arrow marked in (g) for polarity determination. For NM B: (e) twinned layer in contact with the twin-free region, (f ) defective area near the top of the NM, (i) zoomed image from (f ) with superimposed gallium signal obtained by electron energy loss spectroscopy (EELS) (red) and HAADF (green); schematic showing dumbbell composition in the defect. ( j) Coloured structural map showing different crystal orientations with polarities indicated. A and B refers to the polarities in the vertical direction (A = cation on top and B = anion on top). Fig. 4(a) and (b) show low-magnification HR-TEM micro- within the AlGaAs shell. Twinning defects following another graphs of NM A and NM B, respectively. It is immediately seen (inclined) 111 plane are found only at the far-right side of the that NM B has a larger number of both in-plane and out-of- NM. Importantly, they do not propagate from the substrate but plane defects. NM A exhibits a much better crystal quality, rather appear after a few nanometres of GaAs and extend to with relatively few twinning defects parallel and close to the sub- the NM top. Their origin may be related to the overgrowth of strate interface. The density of these twins is about one twin GaAs extending on top of the mask, as observed by every 2.5 nm. The twinned region extends 52 nm above the inter- Tutuncuoglu et al.41 However, the density of these defects is face [Fig. 4(c)], after which NM A has a defect-free zinc blende very low (three twins along the whole NM, located close to the (ZB) crystal structure. This is also revealed in the FFT presented vertex between the base and the long edge of the NM). in Fig. 4(a) where a pure 〈110〉 ZB pattern is present without any In contrast to NM A, NM B presents a relatively high density extra spots. Fig. 4(g) shows the atomically resolved image of this of defects. Starting from the bottom, the NM exhibits a succes- defect-free region, with the corresponding intensity profile given sion of two (3ˉ11) and (13ˉ1) facets, followed by two (1 ˉ01 ˉ)′ and in Fig. 4(h). The difference in the intensity values arises from ˉˉ ˉˉ (011)′ inclined facets, and lastly a (110) facet, where the changes in Z-contrast given by the HAADF detector (I ∼ Z2) and “prime” symbol refers to the system of planes obtained after a reveals B-polar (As-polar) growth.60 This is expected when the rotational twin. Several twins along different {111} planes are first nucleated layer of atoms is Ga (a possible consequence of detected in the whole body of the NM. This is consistent with the gallium pre-treatment). the FFT in Fig. 4(b), where we can see multiple out-of-plane In NM A, an additional set of twins parallel to the (1ˉˉ 1ˉ 1) sub- rotations of the ZB structure caused by multi-twinning. When strate is found in a 25 nm-thick layer close to the top facet two twins around different directions merge, only one of them 820 | Nanoscale, 2020, 12, 815–824 This journal is © The Royal Society of Chemistry 2020 View Article Online Nanoscale Paper propagates further. In the case of non-planar polar semi- nucleated islands to achieve defect-free crystal growth of III–Vs conductor nanostructures the growth front tends to follow the on Si. {111} planes.61 In the presence of an APB, the growth front Overall, correlating the features of the representative TEM- evolves along different angles on both sides of the boundary, measured samples by observation of NM faceting observed in This article is licensed under a Creative Commons Attribution-NonCommercial 3.0 Unported Licence. which explains the formation of kinks at the surface of the SEM, we find two distinct types of NMs on Si (111). The first nanostructure. The presence of complex APBs at the merging type of NMs present uniform faceting and shape, similar to point between polar crystals has been also observed in NM A. These NMs contain a region with twins close to the complex nanostructures such as V-shape nanofins,61 vertical interface, similar to a buffer layer, oriented in the (1ˉˉ1ˉ 1) plane. nanospades62 and close to the nucleation seed of A-polar GaAs The second type of NMs, presenting irregular faceting and NWs.63 In addition, the complex defects appearing in this Open Access Article. Published on 09 December 2019. Downloaded on 2/11/2021 12:05:15 PM. non-uniform shape, contain APBs and other complex defects, region also accumulate strain around them (medium-angle similar to NM B. Close to the Si interface, regions both with annular dark-field (MAADF) images revealing the accumu- and without twins exist suggesting that the growth of NM B lation of strain are shown in the ESI†). Fig. 4(e) and (f ) are was mediated by a polynuclear regime. So far, no APBs were atomic resolution HAADF-STEM images showing the defects observed to originate from the Si–GaAs interface though APBs marked in Fig. 4(b). were found within the bulk of the NM. These APBs are a conse- In order to get more information on this low-angle grain quence of the polynuclear growth combined with independent boundary, atomically resolved EELS gallium elemental multi-twinning processes in each domain, as was shown by mapping was performed. The polarity of the dumbbells was the TEM analysis. This is further supported by the fact that identified by overlapping the integrated intensity of the these APBs were not found in NM A, consistent with a struc- gallium L2,3 edge with a HAADF micrograph of the area. ture originating from a single nucleation event. According to the colour compositional map shown in Fig. 4(i), We have also found that the emission intensity and the gallium atoms are located on top of the dumbbell at both energy of the emission of defected NMs are lower than in the sides of the boundary. Consequently, Ga–Ga pairs form at the single crystalline case. Similarly, the effective lifetime of min- border between the A′ and A″ regions, as detailed in the struc- ority carriers is significantly shorter in the defected structures. tural map presented in Fig. 4( j). At least one other defect invol- We thus suggest that the presence of APBs and other complex ving a polarity inversion is present in a different region of this defects deteriorates the optical properties of the NMs. This is NM (shown in the ESI†). supported by the fact that the CL signal is very weak in the The distribution of defects in NM B could explain the region where these defects are located. inhomogeneous CL emission. From TEM analysis, an APB is It is thus clear that the formation of APBs should be mini- observed separating the two regions of CL emission. APBs have mized for electronic and optoelectronic applications of GaAs been reported to act as potential barriers for carriers.60,61 NMs on silicon. According to our theoretical model, growth Therefore, the APB blocks the diffusion of excitons, creating parameter tuning allows for majority mononuclear growth for two spatially separated regions of CL emission. APBs have also NMs grown in short slits, and hence the absence of APBs orig- been shown to act as recombination centres.15 Recombination inating from twins. However, polynucleation may be un- results in the emission quenching close to the APB and leads avoidable for longer NMs (from several to tens of microns). In to a decrease in intensity between the two regions of emission this case, a different approach is required, one that ensures in the CL map. Similarly, the integrated PL intensity is signifi- that all nuclei have the same polarity in order to avoid APB for- cantly lower in NM B than in NM A, which had no APBs. mation. We saw that a pre-growth treatment consisting of The twinned region of NM B is not continuous along its exposing the substrate to gallium or As4 prior to growth base. Some sections are defect-free while others contain mul- resulted in the preferential orientation of the NMs. More tiple twins. The discontinuity in twining reveals that the detailed investigations into such pre-treatments are necessary growth started from multiple nucleation sites. The example of in order to understand the mechanism by which they are able a region between the twinned and twin-free parts of the NM is to drive the preferential orientation of polar NMs on a non- shown in Fig. 4(e). Here, the left part of the image is defect- polar substrate. Equally interesting is the fact that the prefer- free, while the right side contains multiple rotational twins, ential orientation of the NMs occurs despite the presence of a marked as dashed red lines. When going from left to right, the large number of rotational twins at the base of both NMs that crystal structure is only preserved for one orientation of the were investigated. One might expect that a large twinning twin. In addition, we observe vertical defects between the probability during growth would lead to a randomized distri- defect-free and the twinned region. This is due to the change bution of the NMs orientation which is not the case here. of polarity in the horizontal direction and is quite similar to what is observed in vertical nanospades.62 The presence of this highly defective structure at the base 3. Conclusions of NM B is in stark contrast to the relatively defect-free region observed in NM A. The existence of these two very different We have investigated the SAE growth of GaAs NMs on Si(111) crystal structures in NMs grown side-by-side underline the substrates by MBE. In contrast to homoepitaxial growth, we importance of controlling the phase and coalescence of have observed a higher heterogeneity in the nuclei formed in This journal is © The Royal Society of Chemistry 2020 Nanoscale, 2020, 12, 815–824 | 821 View Article Online Paper Nanoscale the initial stages of growth. We have found that a substrate 4.4. μ-PL pre-treatment with gallium or arsenic prior to growth can Spatially resolved μ-PL characterization was performed at 6 K greatly reduce defect formation and can thus be an important using a diffraction-limited confocal microscope inserted into a element in the defect-free integration of the NMs on Si. We low-vibration, cryogen-free cryostat (Attocube Attodry 1000). A This article is licensed under a Creative Commons Attribution-NonCommercial 3.0 Unported Licence. propose a growth model that fits the experimental data and 780 nm pulsed laser was used to excite the sample whose PL reveals the parameters for growth in the mononuclear regime, signal was dispersed by a 750 mm focal length spectrometer at that is, one nucleus per slit. TEM analysis on an APB-free NM a resolution of ∼30 μeV and detected with a cooled Si CCD. has shown the presence of in-plane twin defects close to the Scanning μ-PL maps were recorded using an XY piezo scanner. GaAs–Si interface while crystal continuity along the base of the Open Access Article. Published on 09 December 2019. Downloaded on 2/11/2021 12:05:15 PM. NM suggests that it grew in the mononuclear regime, which is 4.5. TRPL statistically favoured for slit lengths less than about 1 μm. Finally, CL and PL analyses on a NM whose crystal structure is TRPL spectra were acquired by the so-called time correlated consistent with a polynucleated growth regime suggest that the single photon counting (TCSPC) technique. For this, an ava- presence of APBs and other complex defects deteriorate the lanche photodiode (APD) with a time resolution of ∼25 ps was optical properties of AlGaAs-passivated NMs, as evidenced by used. lower measured minority carrier lifetimes. Overall, the results presented in this work pave the way for monolithic integration 4.6. Cross-sectional STEM of GaAs NMs and related nanostructures on Si substrates. A cross-sectional lamella was cut out of selected NMs using FIB. TEM characterization, including HR-TEM, HAADF STEM and EELS imaging (see the ESI† for details) were performed 4. Experimental using a TECNAI F20 field emission gun microscope operated at 200 kV with a point-to-point resolution of 0.14 nm, coupled 4.1. Substrate preparation to a GATAN Quantum EELS spectrometer. Atomically resolved Samples were prepared by starting with an undoped Si(111) HAADF-STEM images were acquired on a probe-corrected FEI wafer on which a 30 nm thick thermal oxide was grown. The Titan 60–300, equipped with a high-brightness field emission wafer was patterned using e-beam lithography in a Raith gun (XFEG) and a CETCOR corrector from CEOS to produce a EBPG-5000+ with ZEP resist. After developing in n-amyl acetate probe size below 1 Å. The microscope was operated at 300 kV, and isopropyl alcohol (1 min each), the wafer was dry etched with a convergence angle of 25 mrad and an inner collection in an SPTS APS dielectric etching tool using CHF3/SF6 chem- angle of the detector of 58 mrad. Atomically resolved aberra- istry. O2 plasma stripping of the resist was performed before tion corrected HAADF-STEM was used to determine with high dipping the wafer briefly into a buffered HF solution and accuracy the atomic column positions, allowing the detailed quickly loading it into the MBE chamber to prevent oxidation study of polarity60 as well as the final strain analysis by means of the silicon surface. of geometrical phase analysis (GPA).64,65 The HAADF-STEM images were treated with a Wiener filter and beam deconvolu- 4.2. MBE growth tion for clarity. Atomically resolved EELS mapping to deter- Each sample was loaded into a DCA P600 solid-source MBE mine the crystal polarity was performed using a Nion chamber and degassed at 400 °C for 2 h under ultra-high UltraSTEMTM 100MC ‘HERMES’ operated at 60 kV.66 vacuum. The sample was then loaded into the growth chamber and annealed at 750 °C for another 30 min before launching the growth recipe. NM growth was performed at a substrate tempera- ture of 630 °C under a gallium flux yielding a GaAs equivalent Author contributions growth rate of 0.3 Å s−1 and an arsenic beam equivalent pressure A. M. R. and M. F. grew the samples presented in this (BEP) of 5.5 × 10−6 Torr. AlGaAs-capped samples were capped at work. S. M. S., Q. M. R. and J. A. performed the TEM a substrate temperature of 440 °C with 41 nm of Al0.3Ga0.7As fol- analysis. V. G. D. developed the growth model. L. F. and lowed by 7 nm of GaAs to prevent oxidation. Here, the AlAs and A. M. R. performed the CL measurements and GaAs equivalent growth rates were 0.5 Å s−1 and 1.0 Å s−1, analysis. A. M. R. and B. A. performed the μ-PL and TRPL respectively, with an arsenic BEP of 1.1 × 10−6 Torr. characterization. G. T. obtained the first preliminary results. A. M. R., M. F. and A. F. i. M. designed the 4.3. CL experiments. A. M. R., M. F., S. M. S., V. G. D., J. A. and CL measurements were performed at 12 K in a closed cycle A. F. i. M. wrote and edited the manuscript. All authors dis- Attolight Rosa cryostat, using an electron-beam energy of 2.5 cussed the results and commented on the manuscript. keV and current of about 1 nA. The CL signal is collected by a reflective objective (numerical aperture (NA) 0.71) and sent through a 32 cm focal-length spectrometer onto a Peltier- Conflicts of interest cooled Si charge coupled device (CCD) after spectral dispersion by a diffraction grating. There are no conflicts to declare. 822 | Nanoscale, 2020, 12, 815–824 This journal is © The Royal Society of Chemistry 2020 View Article Online Nanoscale Paper Acknowledgements 10 M. Fahed, L. Desplanque, D. Troadec, G. Patriarche and X. Wallart, Nanotechnology, 2016, 27, 505301. LMSC acknowledges funding from Swiss National Science 11 F. Krizek, J. E. Sestoft, P. Aseev, S. Marti-Sanchez, Foundation through the NCCR QSIT and project nr S. Vaitiekėnas, L. Casparis, S. A. Khan, Y. Liu, T. Stankevič, This article is licensed under a Creative Commons Attribution-NonCommercial 3.0 Unported Licence. IZLRZ2_163861. H2020 program through project Indeed. VGD A. M. Whiticar, A. Fursina, F. Boekhout, R. Koops, acknowledges Russian Science Foundation for financial E. Uccelli, L. P. Kouwenhoven, C. M. Marcus, J. Arbiol and support under grant 19-72-30004. IMN acknowledge funding P. Krogstrup, Phys. Rev. Mater., 2018, 2, 093401. from the Spanish MINECO (TEC2015-64189-C3-2-R, EUIN2017- 12 A. Georgakilas, J. Stoemenos, K. Tsagaraki, P. Komninou, 88844, RYC-2017-21995, BES-2013-065340), the Community of N. Flevaris, P. Panayotatos and A. Christou, J. Mater. Res., 1993, 8, 1908–1921. Open Access Article. Published on 09 December 2019. Downloaded on 2/11/2021 12:05:15 PM. Madrid (P2018/EMT-4308), and the European Union FEDER, FSE. EURAMET EMPIR 17FUN06 SIQUST. AMR acknowledges 13 B. Kunert, W. Guo, Y. Mols, B. Tian, Z. Wang, Y. Shi, D. Van the service from the X-SEM Laboratory at IMN, and funding Thourhout, M. Pantouvaki, J. Van Campenhout, R. Langer from MINECO under project CSIC13-4E-1794 with support and K. Barla, Appl. Phys. Lett., 2016, 109, 091101. from EU (FEDER, FSE). SMS acknowledges funding from 14 E. Uccelli, J. Arbiol, C. Magen, P. Krogstrup, E. Russo- “Programa Internacional de Becas ‘la Caixa’-Severo Ochoa”. Averchi, M. Heiss, G. Mugny, F. Morier-Genoud, J. Nygård, ICN2 acknowledge funding from Generalitat de Catalunya J. R. Morante and A. Fontcuberta i Morral, Nano Lett., 2011, 2017 SGR 327. ICN2 is supported by the Severo Ochoa 11, 3827–3832. program from Spanish MINECO (grant no. SEV-2017-0706) 15 R. Alcotte, M. Martin, J. Moeyaert, R. Cipro, S. David, and is funded by the CERCA Programme/Generalitat de F. Bassani, F. Ducroquet, Y. Bogumilowicz, E. Sanchez, Catalunya. Part of the present work has been performed in the Z. Ye, X. Y. Bao, J. B. Pin and T. Baron, APL Mater., 2016, 4, framework of Universitat Autònoma de Barcelona Materials 046101. Science PhD program. Part of the HAADF-STEM microscopy 16 Y. Takagi, H. Yonezu, T. Kawai, K. Hayashida, K. Samonji, was conducted in the Laboratorio de Microscopias Avanzadas N. Ohshima and K. Pak, J. Cryst. Growth, 1995, 150, 677– at Instituto de Nanociencia de Aragon-Universidad de 680. Zaragoza. This work has received funding from the European 17 E. Y. Chang, T. H. Yang, G. Luo and C. Y. Chang, J. Electron. Union’s Horizon 2020 Research and Innovation Programme Mater., 2005, 34, 23–26. under grant agreement no. 654360 NFFA-Europe. We acknowl- 18 J. A. Carlin, S. A. Ringel, E. A. Fitzgerald, M. Bulsara and edge support from CSIC Research Platform on Quantum B. M. Keyes, Appl. Phys. Lett., 2000, 76, 1884–1886. Technologies PTI-001. This project has received funding from 19 J. W. Lee, H. Shichijo, H. L. Tsai and R. J. Matyi, Appl. Phys. the European Union’s Horizon 2020 research and innovation Lett., 1987, 50, 31–33. programme under grant agreement no 823717 – ESTEEM3. 20 M. Yamaguchi, A. Yamamoto, M. Tachikawa, Y. Itoh and M. Sugo, Appl. Phys. Lett., 1988, 53, 2293–2295. 21 A. Krost, R. F. Schnabel, F. Heinrichsdorff, U. Rossow, References D. Bimberg and H. Cerva, J. Cryst. Growth, 1994, 145, 314– 320. 1 X. Miao, K. Chabak, C. Zhang, P. K. Mohseni, D. Walker 22 Q. Li, K. W. Ng and K. M. Lau, Appl. Phys. Lett., 2015, 106, and X. Li, Nano Lett., 2015, 15, 2780–2786. 072105. 2 R. Yan, D. Gargas and P. Yang, Nat. Photonics, 2009, 3, 569– 23 Y. Shi, Z. Wang, J. Van Campenhout, M. Pantouvaki, 576. W. Guo, B. Kunert and D. Van Thourhout, Optica, 2017, 4, 3 Y. Ma, X. Guo, X. Wu, L. Dai and L. Tong, Adv. Opt. 1468. Photonics, 2013, 5, 216. 24 D. Van Thourhout, Y. Shi, M. Baryshnikova, Y. Mols, 4 Y. Zhang, J. Wu, M. Aagesen and H. Liu, J. Phys. D: Appl. N. Kuznetsova, Y. De Koninck, M. Pantouvaki, J. Van Phys., 2015, 48, 463001. Campenhout, R. Langer and B. Kunert, Nano-ridge laser 5 S. Yazji, E. A. Hoffman, D. Ercolani, F. Rossella, A. Pitanti, monolithically grown on (001) Si, Elsevier Inc., 1st edn, A. Cavalli, S. Roddaro, G. Abstreiter, L. Sorba and I. Zardo, 2019, vol. 101, pp. 283–304. Nano Res., 2015, 8, 4048–4060. 25 S. Plissard, K. A. Dick, G. Larrieu, S. Godey, A. Addad, 6 R. R. LaPierre, A. C. E. Chia, S. J. Gibson, X. Wallart and P. Caroff, Nanotechnology, 2010, 21, 385602. C. M. Haapamaki, J. Boulanger, R. Yee, P. Kuyanov, 26 F. Furtmayr, M. Vielemeyer, M. Stutzmann, J. Arbiol, J. Zhang, N. Tajik, N. Jewell and K. M. A. Rahman, Phys. S. Estradé, F. Peirò, J. R. Morante and M. Eickhoff, J. Appl. Status Solidi RRL, 2013, 7, 815–830. Phys., 2008, 104, 034309. 7 G. Otnes and M. T. Borgström, Nano Today, 2017, 12, 31–45. 27 J. Vukajlovic-Plestina, W. Kim, L. Ghisalberti, 8 S. F. Fang, K. Adomi, S. Iyer, H. Morkoç, H. Zabel, C. Choi G. Varnavides, G. Tütüncuoglu, H. Potts, M. Friedl, and N. Otsuka, J. Appl. Phys., 1990, 68, R31–R58. L. Güniat, W. C. Carter, V. G. Dubrovskii and 9 S. H. Huang, G. Balakrishnan, A. Khoshakhlagh, A. Fontcuberta i Morral, Nat. Commun., 2019, 10, 869. A. Jallipalli, L. R. Dawson and D. L. Huffaker, Appl. Phys. 28 L. Güniat, P. Caroff and A. Fontcuberta I Morral, Chem. Lett., 2006, 88, 131911. Rev., 2019, 119, 8958–8971. This journal is © The Royal Society of Chemistry 2020 Nanoscale, 2020, 12, 815–824 | 823 View Article Online Paper Nanoscale 29 P. McIntyre and A. Fontcuberta i Morral, Mater. Today 46 M. Heiß, E. Riedlberger, D. Spirkoska, M. Bichler, G. Abstreiter Nano, 2019, 9, 100058. and A. F. I. Morral, J. Cryst. Growth, 2008, 310, 1049–1056. 30 Z. Zhao, K. Yadavalli, Z. Hao and K. L. Wang, 47 G. Tütüncüoglu, PhD Thesis, EPFL, 2017, 7393. Nanotechnology, 2009, 20, 035304. 48 K. Ikejiri, T. Sato, H. Yoshida, K. Hiruma, J. Motohisa, This article is licensed under a Creative Commons Attribution-NonCommercial 3.0 Unported Licence. 31 J. Z. Li, J. Bai, J. S. Park, B. Adekore, K. Fox, M. Carroll, S. Hara and T. Fukui, Nanotechnology, 2008, 19, 265604. A. Lochtefeld and Z. Shellenbarger, Appl. Phys. Lett., 2007, 49 H. Yoshida, K. Ikejiri, T. Sato, S. Hara, K. Hiruma, 91, 021114. J. Motohisa and T. Fukui, J. Cryst. Growth, 2009, 312, 52–57. 32 J. G. Fiorenza, J. S. Park, J. M. Hydrick, J. Li, J. Z. Li, 50 V. G. Dubrovskii, in Semiconductors and Semimetals, M. Curtin, M. Carroll and A. Lochtefeld, ECS Trans., 2010, Elsevier, 2015, vol. 93, pp. 1–78. Open Access Article. Published on 09 December 2019. Downloaded on 2/11/2021 12:05:15 PM. 963–976. 51 F. Matteini, V. G. Dubrovskii, D. Rüffer, G. Tütüncüoğlu, 33 W. Guo, L. Date, V. Pena, X. Bao, C. Merckling, N. Waldron, Y. Fontana and A. F. I. Morral, Nanotechnology, 2015, 26, N. Collaert, M. Caymax, E. Sanchez, E. Vancoille, K. Barla, 105603. A. Thean, P. Eyben and W. Vandervorst, Appl. Phys. Lett., 52 V. G. Dubrovskii and Y. S. Berdnikov, J. Chem. Phys., 2015, 2014, 105, 062101. 142, 124110. 34 J. Motohisa, J. Noborisaka, J. Takeda, M. Inari and 53 V. G. Dubrovskii, Y. Berdnikov, J. Schmidtbauer, M. Borg, T. Fukui, J. Cryst. Growth, 2004, 272, 180–185. K. Storm, K. Deppert and J. Johansson, Cryst. Growth Des., 35 K. Tomioka, K. Ikejiri, T. Tanaka, J. Motohisa, S. Hara, 2016, 16, 2167–2172. K. Hiruma and T. Fukui, J. Mater. Res., 2011, 26, 2127– 54 V. G. Dubrovskii, Phys. Rev. E, 2019, 99, 012105. 2141. 55 M. Heigoldt, J. Arbiol, D. Spirkoska, J. M. Rebled, 36 F. Schuster, M. Hetzl, S. Weiszer, J. A. Garrido, M. de la S. Conesa-Boj, G. Abstreiter, F. Peiró, J. R. Morante and Mata, C. Magen, J. Arbiol and M. Stutzmann, Nano Lett., A. Fontcuberta I Morral, J. Mater. Chem., 2009, 19, 840–848. 2015, 15, 1773–1779. 56 M. Hetzl, M. Kraut, J. Winnerl, L. Francaviglia, 37 S. Conesa-Boj, E. Russo-Averchi, A. Dalmau-Mallorqui, M. Döblinger, S. Matich, A. FontcubertaMorral and J. Trevino, E. F. Pecora, C. Forestiere, A. Handin, M. Ek, M. Stutzmann, Nano Lett., 2016, 16, 7098–7106. L. Zweifel, L. R. Wallenberg, D. Rüffer, M. Heiss, 57 J. Arbiol, S. Estradé, J. D. Prades, A. Cirera, F. Furtmayr, D. Troadec, L. Dal Negro, P. Caroff and A. Fontcuberta i C. Stark, A. Laufer, M. Stutzmann, M. Eickhoff, M. H. Gass, Morral, ACS Nano, 2012, 6, 10982–10991. A. L. Bleloch, F. Peiró and J. R. Morante, Nanotechnology, 38 E. Russo-Averchi, G. Tütüncüoglu, A. Dalmau-Mallorqui, 2009, 20, 145704. I. Canales Mundet, M. de la Mata, D. Rüffer, J. Arbiol, 58 D. Spirkoska, J. Arbiol, A. Gustafsson, S. Conesa-Boj, S. Conesa-Boj and A. Fontcuberta i Morral, J. Cryst. Growth, F. Glas, I. Zardo, M. Heigoldt, M. H. Gass, A. L. Bleloch, 2015, 420, 47–56. S. Estrade, M. Kaniber, J. Rossler, F. Peiro, J. R. Morante, 39 P. Aseev, A. Fursina, F. Boekhout, F. Krizek, J. E. Sestoft, G. Abstreiter, L. Samuelson and A. Fontcuberta i Morral, F. Borsoi, S. Heedt, G. Wang, L. Binci, S. Martí-Sánchez, Phys. Rev. B: Condens. Matter Mater. Phys., 2009, 80, 245325. T. Swoboda, R. Koops, E. Uccelli, J. Arbiol, P. Krogstrup, 59 M. Heiss, B. Ketterer, E. Uccelli, J. R. Morante, J. Arbiol and L. P. Kouwenhoven and P. Caroff, Nano Lett., 2019, 19, 218– A. Fontcuberta i Morral, Nanotechnology, 2011, 22, 195601. 227. 60 M. de la Mata, C. Magen, J. Gazquez, M. I. B. Utama, M. Heiss, 40 C.-Y. Chi, C.-C. Chang, S. Hu, T.-W. Yeh, S. B. Cronin and S. Lopatin, F. Furtmayr, C. J. Fernández-Rojas, B. Peng, P. D. Dapkus, Nano Lett., 2013, 13, 2506–2515. J. R. Morante, R. Rurali, M. Eickhoff, A. Fontcuberta i Morral, 41 G. Tutuncuoglu, M. de la Mata, D. Deiana, H. Potts, Q. Xiong and J. Arbiol, Nano Lett., 2012, 12, 2579–2586. F. Matteini, J. Arbiol and A. Fontcuberta i Morral, 61 M. de la Mata, R. R. Zamani, S. Martí-Sánchez, M. Eickhoff, Nanoscale, 2015, 7, 19453–19460. Q. Xiong, A. Fontcuberta i Morral, P. Caroff and J. Arbiol, 42 M. Albani, L. Ghisalberti, R. Bergamaschini, M. Friedl, Nano Lett., 2019, 19, 3396–3408. M. Salvalaglio, A. Voigt, F. Montalenti, G. Tütüncüoglu, 62 L. Güniat, S. Martí-Sánchez, O. Garcia, M. Boscardin, A. Fontcuberta i Morral and L. Miglio, Phys. Rev. Mater., D. Vindice, N. Tappy, M. Friedl, W. Kim, M. Zamani, 2018, 2, 093404. L. Francaviglia, A. Balgarkashi, J.-B. Leran, J. Arbiol and 43 Z. Yang, A. Surrente, G. Tutuncuoglu, K. Galkowski, A. Fontcuberta i Morral, ACS Nano, 2019, 13, 5833–5840. M. Cazaban-Carrazé, F. Amaduzzi, P. Leroux, D. K. Maude, 63 M. Zamani, G. Tütüncüoglu, S. Martí-Sánchez, A. Fontcuberta I Morral and P. Plochocka, Nano Lett., 2017, L. Francaviglia, L. Güniat, L. Ghisalberti, H. Potts, M. Friedl, 17, 2979–2984. E. Markov, W. Kim, J.-B. Leran, V. G. Dubrovskii, J. Arbiol and 44 M. Friedl, K. Cerveny, P. Weigele, G. Tütüncüoglu, S. Martí- A. Fontcuberta i Morral, Nanoscale, 2018, 10, 17080–17091. Sánchez, C. Huang, T. Patlatiuk, H. Potts, Z. Sun, 64 M. J. Hÿtch, E. Snoeck and R. Kilaas, Ultramicroscopy, 1998, M. O. Hill, L. Güniat, W. Kim, M. Zamani, V. G. Dubrovskii, 74, 131–146. J. Arbiol, L. J. Lauhon, D. M. Zumbühl and A. Fontcuberta i 65 M. de la Mata, C. Magén, P. Caroff and J. Arbiol, Nano Lett., Morral, Nano Lett., 2018, 18, 2666–2671. 2014, 14, 6614–6620. 45 A. M. Raya, D. Fuster and J. M. Llorens, Nanomaterials, 66 R. R. Zamani, F. S. Hage, S. Lehmann, Q. M. Ramasse and 2019, 9, 856. K. A. Dick, Nano Lett., 2018, 18, 1557–1563. 824 | Nanoscale, 2020, 12, 815–824 This journal is © The Royal Society of Chemistry 2020