Axial Growth Characteristics of Optically Active InGaAs Nanowire Heterostructures for Integrated Nanophotonic Devices

Abstract

III–V semiconductor nanowire (NW) heterostructures with axial InGaAs active regions hold large potential for diverse on-chip device applications, including site-selectively integrated quantum light sources, NW lasers with high material gain, as well as resonant tunneling diodes and avalanche photodiodes. Despite various promising efforts toward high-quality single or multiple axial InGaAs heterostacks using noncatalytic growth mechanisms, the important roles of facet-dependent shape evolution, crystal defects, and the applicability to more universal growth schemes have remained elusive. Here, we report the growth of optically active InGaAs axial NW heterostructures via completely catalyst-free, selective-area molecular beam epitaxy directly on silicon (Si) using GaAs(Sb) NW arrays as tunable, high-uniformity growth templates and highlight fundamental relationships between structural, morphological, and optical properties of the InGaAs region. Structural, compositional, and 3D-tomographic characterizations affirm the desired directional growth along the NW axis with no radial growth observed. Clearly distinct luminescence from the InGaAs active region is demonstrated, where tunable array–geometry parameters and In content up to 20% are further investigated. Based on the underlying twin-induced growth mode, we further describe the facet-dependent shape and interface evolution of the InGaAs segment and its direct correlation with emission energy.

1. Introduction

Free-standing III–V semiconductor nanowires (NWs) have evolved into a promising material platform with various potential device applications in electronics, optoelectronics, and integrated photonics.¹⁻¹² Their tiny footprint enables compatibility with lattice-mismatched substrates, such as silicon (Si),^11,13 and the unique one-dimensional (1D) optical cavity structure with a large refractive index contrast facilitates effective confinement and low-loss propagation of optical modes along the NW’s axial direction.^14,15 Additionally, the small diameter and lateral relaxation of strain within the NW enable the formation of axial heterostructures well beyond the critical thickness limit of planar structures.¹⁶ Such platform can, thus, realize various 1D heterostructure devices, such as resonant tunneling diodes,^17,18 ultrafast photodetectors^19,20 and lasers,^21,22 terahertz NW quantum cascade lasers,²³ and single-photon sources based on NW quantum dots (NW-QDs).²⁴⁻³⁰ Particularly, the deterministic nature of single, site-controlled quantum disks or QDs embedded along the NW growth axis offers attractive capabilities beyond those of conventional Stranski–Krastanov growth. Such NW-QDs or -disks have demonstrated great potential as quantum light sources and low-threshold lasers with high material gain due to their intriguing physical properties, including enhanced light–matter interaction and efficient carrier confinement.^8,15,31−35 It has also been demonstrated that NW-QDs grown along the [111] orientation have a vanishing fine-structure splitting due to their symmetry, which can be ideal candidates for generating entangled photon pairs.^36,37 Thus, the realization of high-quality axial heterostructures within III–V NWs, featuring abrupt interfaces and controlled composition, is a crucial step to harnessing the full potential of these nanomaterials for practical applications.

Conventional approaches for the bottom-up growth of axial NW heterostructures mostly rely on the vapor–liquid–solid (VLS) mechanism, where metallic (e.g., Au) or self-catalysts assist in the axial crystal growth.^{9,12,17,18,33,34,38,39} While VLS has proven successful in the synthesis of various NW structures with high growth rates and aspect ratios, it suffers from certain limitations in achieving axial heterostructures with sharp interfaces, a uniform crystal structure, and controlled composition due to the well-known reservoir and size (contact angle) effect of the catalytic droplets.^21,40−42 The metastability of the catalyst droplet during growth can, thus, lead to inhomogeneous composition profiles, crystal polytype switching, and morphological irregularities, adversely affecting the device performance, as shown for many GaAs-based NW heterostructures.^{9,12,39,43,44} To mitigate such detrimental effects, significant efforts are required, such as employing low-temperature growth, or dynamically adjusting growth rates and beam supply.⁴⁵⁻⁴⁸ Moreover, the VLS growth can further limit incorporation dynamics of ternary material systems, as shown, e.g., for self-catalyzed axial InGaAs/GaAs NW heterostructures,⁴⁹ where In incorporation is limited by the solubility criteria of the catalyst droplet.⁵⁰ In contrast, catalyst-free, site-controlled methods using selective area epitaxy (SAE) have been considered advantageous and much simpler due to the absence of catalyst droplets. Over the past decade, significant efforts have been made to realize high-quality axial NW heterostructures using catalyst-free SAE methods, with impressive results demonstrated for, e.g., GaAs/InGaAs-based NW-QDs in active NW lasers, single-photon emitters, and NW avalanche photodiodes for single-photon detection.^8,31,51−55 Remarkably, the SAE growth of GaAs- or InGaAs-based axial NW heterostructures has remained the exclusive domain of the metal–organic chemical vapor deposition (MOCVD) technique for long, whereas other prominent methods such as molecular beam epitaxy (MBE) remained fully underdeveloped despite distinct advantages, including material purity, growth precision, and reproducibility. Likewise, integration on a versatile platform that is not limited to the commonly used III–V substrates still requires to be demonstrated. Only a few years ago, SAE-based MBE growth of catalyst-free GaAs NWs was demonstrated for the first time, using extraordinarily high V/III ratios and specific substrate preparation schemes.⁵⁶ Subsequent work showed that the addition of small traces of antimony (Sb ≈ 1–2%) and Si dopants helped to establish high-uniformity GaAs NW arrays with excellent aspect ratios, phase-pure crystal structures, and well-defined growth facets.⁵⁷⁻⁵⁹ Hence, the droplet-free SAE-MBE growth of GaAs(Sb) NWs with such pristine properties opens up great versatility for the realization of diverse axial heterostructures.

In this work, we report on the growth of optically active InGaAs axial NW heterostructures by entirely noncatalytic SAE-MBE using the GaAs(Sb) NW-host platform and its integration on Si. To assess the quality of the axial growth, we conducted a comprehensive characterization of the structural and compositional properties of individual NWs, using scanning transmission electron microscopy (STEM), energy-dispersive X-ray spectroscopy (EDXS), and associated three-dimensional (3D) tomographic reconstruction methods. These allowed us to further map the interfacial structure and facet-dependent shape evolution of the InGaAs active region. Optical properties of as-grown NW arrays as well as transferred single NWs were further investigated by microphotoluminescence (μPL) and cathodoluminescence (CL) spectroscopy to identify energy- and position-dependent luminescence characteristics of the distinct InGaAs emission within the heterostructure. These characteristics were subsequently explored under increased In supply and their relation to the SAE geometry parameters.

2. Results and Discussion

2.1. Noncatalytic SAE of Axial NW Heterostructures

All NWs were grown using a solid-source Gen-II MBE system equipped with conventional effusion cells for group-III elements (In, Ga, and Al) and Veeco valved cracker cells to provide group-V elements (As and Sb). The group-V elements were supplied in the form of uncracked As₄ and Sb₂ molecules for the As and Sb species, respectively. For the fabrication of the SAE growth templates, commercial 2 in. p-type Si (111) wafers with single-side polish were employed as substrates. The wafer surfaces were covered with a thermally grown layer of SiO₂, approximately 20 nm in thickness, serving as a mask layer. To induce site-selective NW growth, the SiO₂ mask layer was prepatterned using electron beam lithography, reactive ion etching, and wet chemical etching with buffered hydrofluoric acid.^57,59−61 This process involved creating periodically arranged circular hole-openings with varying nominal diameters of d₀ = 30–120 nm and a pitch (p) of 500 nm (unless otherwise noted). Using these prepatterned openings, the growth of noncatalytic vertical NWs with a high yield was initiated by following the procedures reported in our previous work.^56,57,59 First, GaAsSb NWs were grown for 60 min at a temperature of 630 °C, using a Ga flux of 0.35 Å/s, an As BEP (beam equivalent pressure) of 5.5 × 10^–5 mbar, and a Sb BEP of 3 × 10^–7 mbar, respectively. The GaAsSb NWs served as a stem for the subsequent axial heterostructure, which contained only a small Sb molar fraction (≈2–3%), given the selected Sb/As flux ratio. This low Sb content was chosen purposefully for two reasons: (i) as recently reported, small additions of Sb improve the morphological, microstructural, and optical properties of noncatalytically grown GaAs(Sb) NWs, thanks to an underlying Sb surfactant effect;^57,59 (ii) low Sb content inhibits the formation of type-II band-lineup between GaAsSb and InGaAs, which would induce poor spatial overlap of electron–hole pairs and limited radiative efficiency of the InGaAs active region. For the deposition of the axial InGaAs segment on top of the GaAsSb stems, the substrate temperature was cooled to 590 °C while keeping the Ga flux of 0.35 Å/s and As BEP of 5.5 × 10^–5 mbar fixed. InGaAs segments with different In molar fractions were then grown for 200 s (unless otherwise stated) by varying the In flux between 0.15 and 0.30 Å/s. The InGaAs active region was further embedded in a coaxial 5 nm/3 nm thin AlGaAs/GaAs surface passivation layer (growth temperature at 500 °C) to prevent nonradiative surface recombination.⁶²

2.2. Structural Properties of SAE-Grown NW Heterostructure

Figure 1a illustrates schematically the desired axial NW heterostructure grown by SAE on a prepatterned SiO₂/Si(111) substrate. To view the morphology of the as-grown NW arrays, scanning electron microscopy (SEM) was conducted using a NVision 40 FIB-SEM apparatus from Carl Zeiss. Figure 1b shows a typical SEM image of a vertical GaAsSb/InGaAs/AlGaAs NW array grown on a prepatterned Si substrate (p = 500 nm, d₀ = 70 nm), recorded at 45° bird-eye view and tilt-corrected by the SEM software to display the actual dimensions. A top view of the same array is shown in Figure 1c. The NW arrays typically exhibit a high growth yield (≈90%) and uniform morphology without tapering, with an average length of ≈540 ± 30 nm and a diameter of ≈180 ± 10 nm as well as overall flat top facets.

Figure 1.

Structural and chemical composition-related properties of axial GaAsSb/InGaAs/AlGaAs NWs. (a) Schematic of the axial NW heterostructure by SAE-templated growth on a prepatterned SiO₂/Si(111) substrate. (b) SEM image of a field of NW arrays grown on a patterned SiO₂ mask with a pitch (p) of 500 nm and an opening diameter (d₀) of 70 nm. The image was taken at 45° and tilt-corrected to display actual lengths. (c) Top-view SEM image of identical NWs. (d,e) TEM and HAADF-STEM micrographs (top) and associated EDXS maps (bottom) scanned in the top regions of a NW grown with (d) d₀ = 70 nm ([In] ≈12.6%) and (e) d₀ = 30 nm ([In] ≈9.3%).

To investigate the microstructural features of the as-grown heterostructures, NWs were mechanically transferred from the identical SAE field onto carbon-coated copper grids and analyzed by TEM, high-angle annular dark-field STEM (HAADF-STEM), and EDXS and HAADF-STEM tomography using an FEI Titan Themis transmission electron microscope at 300 kV equipped with a probe aberration corrector. For the chosen camera length of 195 mm, the inner and outer collection semiangles of the HAADF-detector were 41 and 200 mrad, respectively. Figure 1d shows a TEM micrograph of an entire NW (left) and a HAADF-STEM image from the top region of the NW (right-top). Since the image contrast in HAADF-STEM strongly correlates to the atomic number (Z), with heavier elements appearing brighter, the AlGaAs-shell containing Al (Z = 13) appears significantly darker than the GaAs capping with pure Ga (Z = 31), while the InGaAs layer containing In (Z = 49) appears brighter. We also discriminate the GaAsSb stem as the image contrast is similar to the GaAs capping layer due to the very small amount of Sb. Growth of the coaxial 5 nm/3 nm thin AlGaAs/GaAs passivation layer also resulted in significant axial growth, forming a clear AlGaAs segment with a thickness of ≈20 nm in the growth direction. This is expected since the arriving group-III flux at the planar NW top facet is about 5× larger compared to the sidewall facets, due to flux geometry reasons in MBE growth⁶² and the limited surface diffusion of Al species.

An associated EDXS elemental map (Figure 1d, right-bottom) exhibits the compositional profiles. For quantification, we used an absorption correction assuming a thickness of 170 nm and an average density of 5.3 g/cm³, with an estimated accuracy of better than 7%. For elemental contents below 1 at %, the additional combined error from low signal-to-noise ratios and peak fitting may be in the range of experimental values themselves. Here, the uniform distributions of In (purple, [In] ≈ 12.6%) and Al (green, [Al] ≈ 43%) clearly evidence the formation of distinct layers, especially the successful growth of an axial InGaAs insertion (≈30 nm thick) on top of the GaAsSb stem. We observe that the radial growth of InGaAs is negligible, since the measured atomic fraction of In is near the detection limit (<0.5%) at the perimeter, and any NW diameter increase with respect to the GaAsSb stem reference arises only from the coaxial AlGaAs/GaAs passivation layer. For details regarding the EDXS analysis of In and Al distributions, we refer to the Supporting Information (Figure S1). For comparison, as shown in Figure 1e, a smaller hole-opening size of d₀ = 30 nm (under otherwise identical growth conditions) leads to an increased axial NW length (≈1100 ± 200 nm) and decreased diameter (≈130 ± 7 nm), a typical characteristic of the noncatalytic growth mode.^{56,57,59,63,64} Likewise, the InGaAs segment also increased in axial size (≈70 nm), containing a slightly smaller amount of [In] ≈ 9.3% as compared to d₀ = 70 nm. In this example, we also observe that in thinner NWs a more pronounced facet structure is present, with extended inclined facets adjacent to the top facet, which obscure to some extent the compositional profile. Further quantitative comparison of the NW dimensions between the GaAsSb/InGaAs heterostructures and the GaAsSb stem reference reveals a significant increase in the axial NW length with varying SiO₂ mask hole opening (d₀ = 30–120 nm, Figure S2, Supporting Information). In contrast, the NW diameters remain constant over the entire range of d₀, which directly supports the dominant axial growth of InGaAs segments, irrespective of the mask opening size.

Since the composition profiles of the axial NW heterostructure, as mapped by conventional STEM-EDXS, appear to be influenced by the facet structure in the top region, HAADF-STEM tomography analysis was performed to resolve these features in three dimensions. A single-axis tilt series ranging from −75 to +75° in 2.5° steps was recorded and aligned. Following the procedure outlined by Wolf et al.,⁶⁵ first the attenuation coefficient μ was calculated for each projection assuming Beer–Lambert’s law. Second, the series was tomographically reconstructed, whereas, finally, artifacts due to the limited tilt range were compensated for by a finite support approach. For reconstruction, a weighted SIRT algorithm⁶⁶ was been used with five iterations. Residual artifacts from the missing tilt range are still expected to be the main limiting factor for the resolution in the z-direction and source of possible artifacts, while the influence of dynamical Bragg scattering is assumed to average out over the tilt series. Particularly, the potentially harmful contrast variations due to changing scattering conditions across the often observed twin boundaries perpendicular to the [1,1,1] growth direction are for the most part avoided by an inclined tilt axis. The conversion of the attenuation coefficient into Z-contrast was done using the average coefficients in two reference areas of the tomogram, identified as pure AlGaAs (assuming a 1:1 ratio of Al and Ga) and pure GaAs (neglecting the low Sb concentration) and is therefore only qualitative.

Figure 2a shows a 3D isosurface rendering of the STEM tomogram (full video in Supporting Information, Video S1) recorded from the tip of a single NW (d₀ = 70 nm). The image illustrates the complex facet structure of the NW, composed of major {1–10} sidewall facets, as well as a (111)B flat top facet that is intersected by three equivalent {−1–10} inclined facets at every 120°. This structure resembling the form of truncated tetrahedra represents the 3-fold symmetric nature of zincblende III–V NWs, as widely reported for the noncatalytic SAE growth mode.^58,63,67 In fact, the shape formation of the top facet region under this growth mode is further governed by the presence of rotational twin defects, as we discuss later.

Figure 2.

STEM tomographic analysis of the GaAsSb/InGaAs/AlGaAs NW heterostructure grown with d₀ = 70 nm. (a) 3D isosurface rendering of the NW tip, illustrating the complex facet structure consisting of six {1–10} sidewall facets perpendicular to the [111] growth direction. (b) Z-contrast volume rendering of the NW. (c) Corresponding EDXS map of the same NW with a comparable color code. (d) Two central cuts parallel to {1–10} sidewall facets (as indicated in (a)), exposing the Z-contrast through the core. (e) Cross sections at different positions along the NW (as indicated in (d) by labels (i–iv)), revealing the Z-contrast in the radial direction from top (GaAs) to bottom (GaAsSb) of the NW heterostructure.

Comparing the volume rendering through the entire NW tip (Figure 2b, additionally, a volume rendering animation of the NW tip rotating along the NW axis is presented in the Supporting Information Video S2), with an EDXS map without absorption correction (Figure 2c, for a modified version of this map with identical color code to the other EDXS maps in this work, see Figure S3, Supporting Information) recorded for the very same NW in the corresponding orientation shows very similar qualitative features. Both images resolve the anticipated compositional structure, i.e., a thin GaAs capping layer on the outside, followed by the AlGaAs passivation layer, and the inner core of InGaAs/GaAsSb. Due to the predominant GaAs in both regions, a relatively low Z-contrast between the InGaAs and GaAsSb regions is observed as expected. Yet, it is still possible to resolve an approximately 50 nm wide region of higher In concentration, following the top facets of the NW core, which strongly resembles the material contrast as mapped with EDXS (Figure 2c). Figure 2d shows cross-sectional cuts across two major {1–10} sidewall facets (indicated in Figure 2a as labels I and II), which directly illustrate the Z-contrast of the axial NW heterostructure through its core, averaging over ≈6 nm.

Figure 2e illustrates cross-sectional cuts (averaging over ≈6 nm) for each different layer indicated in Figure 2d at selected distances below the terminating top facet, whereas an animation, going slice-by-slice through the entire NW tomogram, is shown in the Supporting Information (see Video S3). Clearly, the uppermost section (i) maps the GaAs capping layer and shows the characteristic shape of the top facet when intersected by the inclined and rotated {−1–10} facets upon twin formation.⁵⁸ About 20 nm below the top surface ((ii)—AlGaAs region), the inclined facets seem to have almost vanished; thus, the hexagonal shape of the NW starts to appear. The next cross section taken further below nicely reveals the inner part composed of InGaAs (iii), surrounded by the coaxial AlGaAs/GaAs passivation layer. The inner diameter (i.e., the distance from one facet to its opposing facet) of the hexagonally shaped InGaAs segment is ≈ 135 nm, reflecting the size of the NW at this growth stage. When moving further down to the position of the GaAsSb stem (iv), the overall lateral dimensions of the NW do not change further. Only the AlGaAs layer appears to shrink in size, which is due to the complete disappearance of the inclined {−1–10} facets at such distance. Compositional analysis further verifies that at this position the inner part is entirely composed of GaAsSb and that no InGaAs shell is observed in between the adjacent AlGaAs/GaAs coaxial layers (Figure S3, Supporting Information). Also, another tomographic analysis of a NW grown under the same conditions, but with a reduced hole-opening size of d₀ = 30 nm, reveals very consistent results (Figure S4, Supporting Information). Further analysis of the InGaAs segment, performed on additional NWs grown at increased In flux, is shown further below and confirms the anticipated facet-mediated shape.

2.3. Emission Properties of the Optically Active InGaAs NW Heterostructure

Assessing the emission properties, steady-state photoluminescence (PL) measurements were performed on the as-grown NW arrays from specific SAE fields (here, p = 500 nm and d₀ = 70 nm, same NWs as in Figure 1b,c). For all PL measurements, samples were probed at 10 K using a He-flow cryostat. A mode-locked tunable Ti:sapphire laser (780 nm) with a repetition rate of 80 MHz and a pulse width of ≈100 ps was used for excitation. The focused spot size (NA = 0.42) was ≈1.9 μm, which was expanded to ≈12.4 μm for probing a large area of NW arrays (i.e., sampling ≈600 NWs). For comparison, GaAsSb NWs without InGaAs insertion were also measured as a reference. Under a low-excitation power density (P_exc) of 0.2 μW/μm², as shown in Figure 3a, the PL spectra of the GaAsSb NWs (gray, bottom) and those with InGaAs inserts (purple, top) are very different. While the reference GaAsSb NWs show emission at ≈1.47 eV, the NWs with InGaAs inserts ([In] ≈13%) exhibit a dominant emission at a lower energy of ≈1.35 eV, with a full width at half-maximum (fwhm) of about 50 meV.

Figure 3.

Emission properties of GaAsSb/InGaAs NW heterostructures. (a) Low-temperature (10 K) PL spectra from as-grown GaAsSb NW arrays (p = 500 nm, d₀ = 70 nm) with (top) and without (bottom) InGaAs insertion. (b) SEM image of transferred GaAsSb/InGaAs NWs on Si. (c) P_exc-dependent μPL spectra on a logarithmic scale obtained from a single GaAsSb/InGaAs NW. (d) Hyperspectral CL line scans extracted along the axis of a GaAsSb/InGaAs NW with d₀ = 70 nm, recorded with an SEM acceleration voltage of 2 kV and beam currents of 200 pA (top) and 4 nA (bottom). (e) CL spectral line scan of a GaAsSb/InGaAs NW with a nominal d₀ of 30 nm, recorded at 2 kV and 200 pA.

The strong PL peak at this energy shows that carrier recombination takes place predominantly in the InGaAs segment and that the surrounding AlGaAs passivation layer efficiently prevents nonradiative surface recombination. However, weak emission from the GaAsSb stem is also observed, suggesting that not all photogenerated carriers diffuse and recombine in the InGaAs segment. The fact that the GaAsSb emission occurs at the same peak energy for both NW arrays, with and without the InGaAs insert, proves the absence of any radial growth of the lattice-mismatched InGaAs surrounding the NW core. Additionally, the PL emission intensity and energy from the InGaAs segments on the GaAsSb NW arrays exhibit a slight dependency on the hole-opening sizes (Figure S5a, Supporting Information). This observation may indicate variations in growth and incorporation dynamics that are sensitive to geometrical parameters and will require further investigation to confirm these effects.

To minimize the spectral broadening caused by the ensemble effect of the NW arrays, steady-state confocal μPL measurements were performed on single NWs harvested from the same SAE field (p = 500 nm and d₀ = 70 nm) and transferred onto Si substrates (Figure 3b). Data were recorded with a nonexpanded spot size of ≈1.9 μm and varying P_exc ranging from P₀ = 0.02 μW/μm² to 500P₀. In Figure 3c, it is shown that at the lowest power density (P₀), the emission energy from the InGaAs insert is ≈ 1.30 eV with a fwhm of only about 7.4 meV. This energy matches fairly well with the expected low-temperature band-gap energy of unstrained InGaAs for the EDXS-measured In content of ≈13%.^68,69 Upon increasing the power by an order of magnitude to 10P₀, which is equivalent to the power density in Figure 3a, the fwhm is approximately 28.6 meV, nearly half the width compared to that of the ensembles from identical NW arrays. Overall, with increasing P_exc, the spectrum gradually broadens toward a higher energy, eventually merging with the GaAsSb emission peak, likely due to state-filling effects and carrier escape.

The spatial origin of the different luminescence bands observed in PL was investigated by a CL spectroscopy. The CL hyperspectral line scans were performed at 10 K in a liquid He-cooled stage integrated with a Gatan MonoCL4 system using a liquid-nitrogen-cooled Si charge-coupled device as a detector. Figure 3d exhibits hyperspectral CL line scans extracted along the axis of the identical NW as in Figure 3b,c, which were recorded with an SEM acceleration voltage of 2 kV and beam currents of 200 pA (top) and 4 nA (bottom), respectively. While the CL emission at ≈1.47 eV is definitely attributed to the GaAsSb core NW, it is evident that the spectrally distinct emission of lower energy around ≈1.3 eV, which slightly blue-shifts with increasing excitation density, originates primarily from the top of the NW (axial position of y ≈ 0.35 μm). This observation provides clear evidence of the spatial origin of the luminescence from the InGaAs inserts. For comparison, Figure 3e shows a CL spectral line scan from a NW grown under the same conditions but with a reduced hole-opening size of d₀ = 30 nm, which was recorded at the same acceleration voltage of 2 kV and a beam current of 200 pA. The data confirm the spatially distinct luminescence from the InGaAs segment (y ≈ 1 μm) and further illustrate a slight shift of the InGaAs emission energy toward 1.35 eV for a reduced opening size of 30 nm. In line with Figure 1d, this directly reflects the lower In content of [In] ≈9.3% for d₀ = 30 nm (vs [In] ≈13% at d₀ = 70 nm).

2.4. Growth and Optical Characteristics of the InGaAs Active Region

To achieve tunability of the emission properties, it is desired to further control the In content in the InGaAs segments. Figure 4a shows PL spectra (at T = 10 K, P_exc = 0.2 μW/μm²) from as-grown GaAsSb/InGaAs NW arrays (p = 500 nm, d₀ = 70 nm) with different In contents, as obtained by increasing the In flux from 0.15 Å/s (blue, top) to 0.23 Å/s (green, middle) and to 0.30 Å/s (red, bottom). (All other growth and geometrical conditions are identical.) A clear red-shifted PL peak of up to ≈50 meV is observed toward the highest In flux of 0.30 Å/s. This directly reflects the reduced band-gap energy of the InGaAs segment that arises from the increased In incorporation ([In] ≈17.4–20.0%, according to EDXS measurements). The emission also exhibits an increase in inhomogeneous spectral broadening (fwhm of ≈118 meV from the ensemble, ≈82 meV from a single NW, see Figure S6 in the Supporting Information), which is up to ≈3× broader compared to the lowest In flux of 0.15 Å/s (≈49.5 meV from the ensemble; ≈28.6 meV from a single NW). Both the red shift and broadening of the PL emission are consistently observed over varying geometrical parameters, e.g., hole-opening sizes (Figure S5b,c, Supporting Information). We attribute such emission broadening not only to increased compositional fluctuations resulting from the increased In supply but also to the unique structure of the InGaAs segment.

Figure 4.

Optical and compositional properties of GaAsSb/InGaAs NWs with increased In supply. (a) PL spectra (T = 10 K, P_exc = 0.2 μW/μm²) from as-grown GaAsSb/InGaAs NW arrays under varying In supply. (b) EDXS profile along the center axis of the NW with highest In supply, from the NW tip toward the GaAsSb core. Inset: magnified view of the Sb content recorded at the axial position x = 50–90 nm (top) and schematic description of the (i) InGaAs region with a relatively constant In content and (ii) [In] gradient region along the inclined facets (bottom). (c) Schematic illustration of the InGaAs axial deposition (purple) on top of the GaAsSb NW stem (dark-blue) with {−1–10} inclined facets. (d) EDXS map of the In distribution across the InGaAs segment, of which the bottom part reveals overgrowth on the inclined (−1–10) facet. (e) Associated HAADF-STEM micrograph from the tip of the GaAsSb NW reference with an inclined (−1–10) facet.

Figure 4b–e presents insights into the actual structure of the InGaAs segment by mapping the distribution of each element using EDXS scans along and across the growth direction of a NW grown with the highest In flux of 0.30 Å/s. For EDXS scans performed along the NW axis (see Figure 4b), the elemental profile clearly indicates the presence of the ≈20 nm thick AlGaAs passivation layer with [Al] ≈35% (light-green) and [Ga] ≈65% (light-blue) at position x ≈ 20–40 nm, followed by an InGaAs layer with a relatively constant [In] ≈18% (purple) from x ≈ 40–65 nm (Figure 4b-inset, region (i)). Interestingly, from x ≈ 65 nm onward, a clear gradient toward a reduced In content is observed (Figure 4b inset, region (ii)). This does not stem from a nonabrupt GaAsSb/InGaAs interface but can be attributed to the peculiar facet structure exposed by the underlying GaAsSb NW stem. As schematically depicted in Figure 4c, during the noncatalytic growth mechanism of the GaAsSb NWs along the vertical (111)B direction, three inclined {−1–10} facets emerge from the corners and extend toward the center, where the intersections create a flat triangle on the top (111)B facet.^58,59 The presence of such inclined facets influences the shape formation of the InGaAs segment via the competition of growth on the top and the inclined facets. Usually, the {−1–10} inclined facets are slow-growing facets compared to the (111)B facet; however, they are known to turn into fast-growing ones when a rotational twin defect is formed.^58,67 Note that the presence of twins in the InGaAs segment supporting this mechanism is confirmed by TEM in the Supporting Information (Figure S7). Therefore, the twinned mesa grows out laterally and the space above the original {−1–10} facets is filled up. This behavior is directly seen in the EDXS map of Figure 4d, showing that the InGaAs segment extends outward from the inclined {−1–10} GaAsSb facets. A corresponding HAADF-STEM micrograph of the GaAsSb reference shown in Figure 4e confirms this observation, clearly revealing the presence of inclined {−1–10} facets that extend about 35 nm along the NW. From Figure 4d,e it is further evident that the interface between GaAsSb and InGaAs is relatively sharp without any gradient in In content across the interface. Note that the ≈35 nm wide inclined {−1–10} facets at the GaAsSb interface also agree with the scale of the [In] gradient recorded along the NW axis toward the GaAsSb stem in Figure 4b (region ii, x ≈ 60–95 nm) and, thus, explains the apparent reduction in In content over this region. Essentially, over the length of the inclined facet toward the GaAsSb, the fraction of the InGaAs probe volume decreases continually, while the fraction of the underlying GaAsSb stem increases correspondingly. This thickness gradient is also reflected by the apparent increase in the Sb content from [Sb] ≈ 1.5–3%, as shown in the inset of Figure 4b (orange trace). Due to this, the mean In content mapped over region ii across the inclined facet is only 7.7–8.9% for this sample with the highest In flux (0.30 Å/s). Similarly, a lower mean In content (≈4.4–4.5%) is also observed for InGaAs sections along the {−1–10} facets grown with the lowest In flux (0.15 Å/s) (Figure S8b, Supporting Information). Despite the pronounced thickness effect, we cannot rule out the overall lower In incorporation on the inclined NW facets. Evidence for an overall lower In incorporation on the inclined {−1–10} facets can be suggested from additional growth experiments performed with reduced InGaAs growth time at an equivalent In flux (0.30 Å/s). As shown in the Supporting Information (Figure S8a,c), a 4-fold reduced growth time (50 s) of the InGaAs insertion results in growth taking place almost exclusively on the inclined facets ([In] ≈7.4%), with a PL emission energy that corresponds to this lower In content. This suggests that control of the extension of the inclined facet structure is critical for tuning the shape of the InGaAs active region, especially toward strongly confined size dimensions, while it plays only a minor role for extended InGaAs active regions. Suppression of extended {−1–10} facets could be realized in the future by substantially increasing the twin density in the underlying NW stem, e.g., by employing Si codopants, which can lead to pristine (111)B top facets.⁵⁷ Under such improved conditions, we expect to approach the ideal shape of InGaAs inserts/quantum disks that are uniformly extended in size and composition over the top growth facet, similar to recent reports in MOCVD-grown InGaAs/GaAs and InAsP/InP NW-QD heterostructures.^8,35,54

3. Conclusions

In conclusion, we have demonstrated the growth of axial, optically active InGaAs NW heterostructures by entirely catalyst-free selective-area MBE directly on Si, employing GaAs(Sb) NW arrays as tunable, highly uniform growth stems. Structural and compositional characterizations using STEM, EDXS, and associated tomographic reconstruction methods revealed the successful axial growth of InGaAs segments on the GaAsSb NW stems with a negligible degree of undesired radial growth. The correlation of PL spectroscopy and single-wire CL hyperspectral line scans clearly evidenced the spatial origin of the distinct luminescence arising from the InGaAs segment. A systematic increase of the In supply in the InGaAs insertion resulted in a clear red-shifted PL peak energy with In content up to ≈20%, besides an increase in inhomogeneous broadening. Further analysis could delineate the shape evolution of the InGaAs active region, which is governed by the interplay between incorporation at the (111)B top facet and inclined {−1–10} facets on the tip of the underlying GaAsSb NW stem and its correlation with emission properties. This comprehensive understanding of axial growth properties will enable further control of the InGaAs shape and heterointerface properties, to drive the advancements in axial NW heterostructure devices, such as NW-QDs, and expand their applications in various optoelectronic and silicon photonic technologies. In particular, future directions should aim at increasing the In content in the InGaAs active region to push emission wavelengths toward the Si transparent region. These challenging endeavors need to overcome strain-limiting effects, which will be possible by integrating InGaAs QD emitters into very thin NW resonators where strain relaxation can be inhibited.¹⁶

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsanm.3c05392.

• STEM-EDXS analysis; SEM analysis of GaAsSb NWs with and without InGaAs inserts; STEM-EDXS map of the NW heterostructure; STEM-tomography of the NW heterostructure with d₀ = 30 nm; macro-PL for varying SiO₂ mask hole-opening sizes; μ-PL and CL emissions from InGaAs segments grown with high In flux; rotational twins in the InGaAs segment; and PL and EDXS results influenced by InGaAs deposition characteristics (PDF)

• Iso-surface rendering animation of the NW tip rotating along the NW axis (MPG)

• Volume rendering animation of the NW tip rotating along the NW axis (MPG)

• Tomographic animation slicing through the entire NW (MPG)

Author Contributions

H.W.J. and A.A. contributed equally to this work. H.W.J. wrote the manuscript in consultation with G.K. and with contributions from all authors. A.A., H.W.J., and G.K. conceived this work. H.W.J., N.M., and D.S. fabricated SAE templates. A.A. and H.W.J. performed MBE growth. H.W.J. prepared the samples for experiments. H.W.J. and A.A. conducted SEM measurements. M.D. performed STEM and EDXS measurements. S.S. and R.Z. carried out STEM tomography reconstruction with guidance from K.M.-C. H.W.J. implemented PL measurements with contributions from D.S. and N.M. M.G.R. performed CL measurements with guidance from J.L. G.K. and J.J.F. supervised the project. All authors have given approval to the final version of the manuscript.

European Research Council (ERC project QUANtIC, ID: 771747); Deutsche Forschungsgemeinschaft (DFG) via Germany’s Excellence Strategy—EXC-2111-390814868 (Munich Center for Quantum Science and Technology, MCQST); DFG under contract EXC 2089/1–390776260 (Excellence Cluster e-conversion); Bavarian Hitech Agenda within the EQAP project.

The authors declare no competing financial interest.