Hexagonal silicon−germanium nanowire branches with tunable composition

Silicon (Si) and germanium (Ge) normally crystallize in the diamond cubic crystal structure, but are expected to exhibit interesting properties when grown in the hexagonal 2H structure. It has been predicted that hexagonal Si_{1 − x} Ge_x has a direct band gap at the Γ-point which is tunable in the range 0.3−0.7 eV [1–7], for 1.0 > x > 0.65. This emission range is interesting for a broad range of applications, including telecommunication, LiDAR, and chemical sensing. In addition, hexagonal SiGe in the shape of nanowires is expected to have beneficial thermoelectric properties; calculations predict that the thermal conductivity of the hexagonal 2H phase of Silicon is almost a factor of 2 lower than that of cubic Si [8].

Hexagonal Si_{1 − x} Ge_x is a metastable material and does therefore not occur naturally. Although it is a challenge to fabricate this material synthetically, it can be forced in the hexagonal phase in nanosized materials [9–19], and one of the outstanding challenges is to increase the volume of the material. Single-crystalline hexagonal Si, Ge, and SiGe alloys have recently been reported in a nanowire (NW) core/shell geometry using a wurtzite III–V core wire as an epitaxial template [20–24]. This method has been successful for demonstrating the optical properties [24] but also has some drawbacks: (1) the group V element from the core material can be incorporated in the SiGe shell during growth [25] (2) I₃ planar stacking faults have recently been observed in SiGe shells [26], (3) and due to the large dimensions, lattice matching requirements are similar as for bulk materials, limiting the possiblities of growing SiGe heterostructures.

III–V nanowires with the wurtzite or cubic crystal structure can be grown directly from a Au catalyst particle, but this does not work for hexagonal SiGe when grown in the 〈111〉_C/〈0001〉_H direction, since the system always crystallizes in the cubic phase. In this work, we report on the growth of single crystalline hexagonal SiGe nanowires with full compositional control along a 〈1–100〉 growth direction. We utilize the crystal structure transfer method [20] which has been employed successfully in the past to make hexagonal Si and Si_{1 −x} Ge_x shells [20–24]. But, in this case, we add an additional step by positioning Au particles on the sidewalls of the hexagonal core/shell NW and commence vapor−liquid−solid (VLS) growth of NWs, which resemble multiple branches growing out of a tree trunk. Branched nanowire growth has already been reported before [27–32]. In our case, the epitaxial growth on the wurtzite core wire enforces all the SiGe branches to grow in the hexagonal phase and an arbitrarily long NW can be produced. Due to the small dimensions of these branch wires, strain relaxation allows the fabrication of SiGe heterostructures. Moreover, optoelectronic or thermoelectric devices are more easily fabricated with the simple wire geometry as compared to the thick core/shell wires [20–24]. We note that hexagonal SiGe wires can, in principle, also be grown from a planar hexagonal substrate by the VLS mechanism; the challenge is to identify a suitable substrate.

In this work we use wurtzite GaP (GaAs) core wires as epitaxial templates, which are almost lattice matched to Si (Ge). We illustrate the general principle of the branch growth process using a wurtzite GaP core in figure 1. We note here that the growth process is independent of the SiGe composition and the use of GaP or GaAs as trunk wires. Vertical hexagonal GaP NW trunks are grown from an array of 6 nm thick Au islands in 100 nm diameter mask openings defined by nanoimprint lithography (figure 1(a)) [33, 34]. From x-ray diffraction and transmission electron microscopy (TEM) studies we know that the GaP NWs have the hexagonal crystal structure and are basically free of defects (less than 1 SF μm⁻¹) [33, 34]. On the GaP NW sidewalls a second generation of Au catalyst particles is deposited by thermal evaporation under a 45° incidence angle (figure 1(b)). The side facets are then utilized as a substrate in order to epitaxially grow hexagonal Si_{1 −x} Ge_x (0 ≤ x ≤ 1) branches (figure 1(c)). In figure 1(d) a top-view scanning electron microscopy (SEM) image of the sample after growth of Si_0.64Ge_0.36 branches is shown. All Si:Ge ratios mentioned in the text have been determined using quantified STEM-EDX spectra. The absolute quantification of this ratio was calibrated against atom probe tomography in an earlier paper [24]. The cross section of the GaP NW exhibits a hexagonal shape with $\left\{11\bar{2}0\right\}$ side facets. The SiGe branches grow in six equivalent 〈$1\bar{1}00$〉 azimuthal directions parallel to the facets. From the TEM image in figure 1(e) it is clear that all branches grow perpendicular to the long axis of the GaP nanowire trunk. In addition, the absence of diffraction contrast in the wires shows that these branches are single crystalline. In figure 1(f), we present a high-resolution scanning high-angle annular dark field scanning TEM (HAADF STEM) image showing the perfect epitaxial connection between the trunk and the branch. The characteristic 'dumbbells' of the hexagonal 2H crystal structure are evident, no defects are visible, while the contrast between GaP and SiGe is well discernable. The perfect epitaxy displays that the branch-growth method enables a defect-free interface, where the limited diameter of the branches ensures elastic strain relaxation at the base of the branch.

Zoom In Zoom Out Reset image size

We now continue to study the structural properties of the branches in more detail by high resolution TEM of SiGe branches in figure 2. The HRTEM image in figure 2(a) is taken along the $\left\langle 11\bar{2}0\right\rangle$ zone axis which allows to discriminate between the cubic and hexagonal crystal phase [19], and shows a tapered branch. The core of the branch, which was grown by the VLS mechanism, displays ABAB stacking, as highlighted in figure 2(b), which signifies the hexagonal crystal structure, without planar defects. For a branch growing in the $\left\langle 1\bar{1}00\right\rangle$ direction, i.e. orthogonal to the ABAB stacking sequence, stacking faults cannot be formed perpendicular to the growth direction, and the crystal structure cannot be switched to the cubic phase. The observed tapering is partly due to simultaneous vapor−solid (VS) growth on the $\left\{0002\right\}$ side facets on top and on the bottom of the branch as will be discussed below using the in situ TEM studies. This partial shell displays the diamond 3C crystal structure, visualized by ABCABC stacking of the crystal planes in figure 2(b). Crystal defects in the shell are recognizable close to the base of the branch, at the point where the cubic layers of the shell meet the hexagonal layers of the trunk. This implies that the shell growth originates from the nucleation of new layers on the {0002} side facets themselves, rather than by extending shell layers from the hexagonal trunk.

Zoom In Zoom Out Reset image size

To substantiate the single-crystalline nature of the hexagonal SiGe branches, we have prepared cross-sectional cuts of the branches by using a focused ion beam (FIB). In figure 2(c) a HRTEM image of a cross-section of a branch is shown. There are no stacking faults or other defects visible within the structure. The cross-sectional shape of the branches, shown in figure 2(c), is rectangular, resulting from the 2-fold crystal symmetry around the $\left\langle 1\bar{1}00\right\rangle$ crystal axis. The core of the branch has the hexagonal crystal structure, which is extended to the $\left\{11\bar{2}0\right\}$ side facets of the branch by VS growth; we note that in the $\left\langle 11\bar{2}0\right\rangle$ direction, planar stacking faults cannot be formed and therefore the crystal structure of the shell adopts that of the core NW. As mentioned before, VS growth in the 〈0001〉 directions (top and bottom) results in the formation of the cubic crystal structure shell. Upon imaging along the $\left\langle 1\bar{1}00\right\rangle$ zone axis these two different crystal structures cannot be distinguished [19].

We have investigated the growth of Ge branches on wurtzite GaAs trunk nanowires using in situ TEM. The video (see supplementary information Video1 (available online at stacks.iop.org/NANO/34/015601/mmedia)) shows a Ge branch with the hexagonal crystal structure, imaged along a $\left\langle 11\bar{2}0\right\rangle$ zone axis, which grows epitaxially on the GaAs trunk nanowire by the Au-catalyzed VLS mechanism in the $\left\langle 1\bar{1}00\right\rangle$ direction, similar to the ex situ experiments. Snapshots of the main features of the growth are shown in figures 3 and 4. One must note that GaAs trunks have {1100}-type sidewalls, and in that case, branches grow perpendicular to the side facets, conversely to the previous GaP case where branches grow parallel to the {11–20} side facets. At the beginning of the growth, the liquid droplet on the branch tip has a large wetting angle with respect to the (1−100) interface (approximately 105° on the video and a corresponding diameter of around 18.5 nm). This contact angle stabilizes around 96°–100° (see figure 3(a) inset) and is maintained during the growth (the diameter of 17.2 nm does not evolve from 98 s (figure 3(a)) to 175 s (figure 3(b))). The hexagonal crystal structure in the branch is formed by precipitation at the (1–100) liquid/solid interface by a layer-by-layer growth mode. At this moment, it is not clear if nucleation always occurs at the triple phase boundary (TPB) or at the liquid/solid interface. This question will be investigated in more detail in future studies. Any planar defect present in the GaAs template is replicated along the branch as pointed in figure 3. As observed in the ex situ studies, a cubic (3C) lateral Ge layer is grown on the top and bottom facets of the branch on the {0002} sides with several defects in this shell formed during the growth. We note that the axial (hexagonal) growth rate corresponds to 0.3 monolayers s⁻¹ and the radial (cubic) rate is 0.05 monolayers s⁻¹.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

From the video (see supplementary information Video2) and the corresponding enlarged snapshots shown in figures 4(a) and (b) it is clear that at least one monolayer of the cubic phase, highlighted in blue, nucleates at the TPB by the VLS mechanism. The cubic layer expands outwards by a VS mechanism in a layer-by-layer growth mode. The layers progress from the concave corner at the trunk-branch interface and then propagate towards the catalyst particle. The movies, and figures 4(c)–(f), also show that the liquid/solid interface between the catalyst and the branch is atomically flat in the centre, but truncated facets close to the TPB appear periodically as highlighted in blue in snapshots of the figures 4(c)–(f) with variable angle of truncation. We note that the catalyst contact angle (defined as the angle between the droplet edge and the extrapolated planar solid-liquid interface to the edge) is 90° when there is no truncation and larger in case of a truncation. This particular point needs to be analysed in future experiments to understand the possible relationship between the truncation, the wetting angle, and the formation of the cubic layer at the TPB. An oscillatory truncation at the TPB has been observed before during the growth of Ge, and Si in the cubic 〈111〉 direction and depends on the level of supersaturation in the catalyst [35, 36]. Here, using atomic resolution, we observe in the particular $\left\langle 1\bar{1}00\right\rangle$ growth direction that each truncation period corresponds to the formation of one Ge monolayer at the (1–100) liquid/solid interface. During the formation of the truncation, part of the already grown nanowire is dissolved in the catalyst particle and the dissolved Ge in the catalyst may serve to reach the required supersaturation for the formation of a new monolayer.

By developing a route for the growth of hexagonal Ge and Si NWs, we have laid the groundwork to further investigate the controllable growth of hexagonal Si_{1 −x} Ge_x compound nanowires, which is of interest since for x > 0.65 they should exhibit a direct band gap. These compound NWs have been grown at 620 °C, and x is controlled by the ratio of the input pressures of the precursors. In figure 5(a), we present an energy-dispersive x-ray (EDX) spectroscopy map to show the elemental composition of a number of SiGe branches, and measure a composition of Si_0.46Ge_0.54, which is close to the input ratio. The map further clarifies that the composition is uniform along the axis of the branch NW, but a Ge-rich tapered shell can be observed. In order to investigate the composition in the viewing direction that is parallel to the axis of the branch NW, we prepared a cross-sections of a branch. The cross-sectional EDX map in figure 5(b) shows that the composition of the branch NWs is uniform in the center, whereas there is a Ge rich outer shell of the hexagonal SiGe, surrounded by a SiO₂ outer shell. The Ge-enriched sub-surface layer could be explained by segregation due to preferential oxidation of Si at the surface [37], or by the different growth dynamics of VS compared to VLS growth; the non-uniform thickness of the Ge-enriched layer in figure 5(a) suggests the latter. We note that the formation of a ${\rm{Si}}{{\rm{O}}}_{x}$ surface layer, eliminating dangling bonds and creating a wide bandgap capping, is probably beneficial for the optoelectronic properties of the branch NWs. In figure 5(c), we have plotted the Ge content incorporated into the SiGe branch, as determined by EDX spectroscopy, as a function of the input ratio of the Ge and Si precursors. With this fabrication methodology, it is possible to tune the composition of the hexagonal ${{\rm{Si}}}_{1-x}{{\rm{Ge}}}_{x}$ branch NWs in the full range. The flexibility that this fabrication method provides is further investigated by growing more complex heterostructures in the same branch NW. In figure 5(d), A HAADF image shows a number of branch wires grown from the GaP trunk. A quantitative compositional line profile along a 0.5 μm long branch shows the variation in composition along the branch: starting with Si_0.80Ge_0.20 at the base of the branch and pure Si at the tip.

Zoom In Zoom Out Reset image size

In conclusion, we have fabricated ${{\rm{Si}}}_{1-x}{{\rm{Ge}}}_{x}$ nanowires with the hexagonal crystal structure by growing them as epitaxial branches on a suitable substrate trunk NW. We show that it is possible to fabricate ${{\rm{Si}}}_{1-x}{{\rm{Ge}}}_{x}$ NWs throughout the whole stoichiometric range and even simple SiGe heterostructures. Future steps are to realize optically active quantum dots and simple opto-electronic devices of hexagonal Si and SiGe nanowires defined in the axial direction.

We would like to thank P J van Veldhoven for technical support at the MOVPE reactor, A Cavalli, L Gagliano, and M Y Swinkels for nanoimprint preparation, M J Brett for valuable discussions, and G Badawy for designing the figures. This work was supported by the Dutch Organization for Scientific Research (NWO-VICI 700.10.441), and the European Union's Horizon 2020 research and innovation program under grant agreement number 964191 (Opto Silicon). Solliance and the province of Noord-Brabant are acknowledged for funding the TEM facility. This project has received funding from the French National Research Agency (ANR) under the grant ANR-17-CE030–0014–01 (HEXSIGE). We acknowledge the ANR for funding the NANOMAX ETEM through the TEMPOS grant (10-EQPX-0050)). We acknowledge Odile Stephan leader of TEMPOS and Jean Luc Maurice manager of NANOMAX. We wish to particularly acknowledge Ileana Florea and Federico Panceira for the technical assistance during experiments. Thanks are due to the CIMEX at École polytechnique (Palaiseau, France) for hosting the NANOMAX microscope.

The data that support the findings of this study will be openly available following an embargo at the following URL/DOI: https://doi.org/10.5281/zenodo.6664593. Data will be available from 19 July 2022.

The wurtzite trunk GaP wires have been grown following references [31, 32] in a close-coupled showerhead reactor by Aixtron. The branches have been grown in the same MOVPE reactor at a total pressure of 50 mbar, and a total gas flow of 6200 sccm. After deposition of the Au particles on the GaP trunk wires, the samples are annealed at 710 °C for 1 min. For the growth of the branches the samples are cooled to 650 °C and exposed to 1% GeH₄/H₂ and 1% Si₂H₆/H₂. The total group IV input flow is between 190 and 300 sccm.The input flux ratio of the precursors determines the chemical composition of the branches. The branches are grown for 6 or 6 and a half minutes.

For the TEM measurements, two different sample preparation methods were used. In the standard side-view analysis, branched nanowires are mechanically transferred to a holey carbon TEM grid. For the cross-sectional TEM studies, nanowires are prepared using a FEI Nova Nanolab 600i dual beam system. In both cases, high-resolution TEM and scanning TEM (STEM) analyses are conducted using a JEM ARM200F probe-corrected TEM operated at 200 kV. Energy dispersive x-ray (EDX) spectroscopy measurements are carried out using the same microscope equipped with a 100 mm² EDX silicon drift detector.

For in situ dynamical HRTEM observations, we have utilized Au-catalyzed wurtzite GaAs nanowires as a template for the growth of Ge-2H branches. Observations are carried out using a Cs-corrected Titan environmental TEM (ETEM) operated at 300 kV. Movies are recorded using a Gatan US1000 camera at a rate of 4 fps. The images in the article are snapshots extracted from the videos. We use a Protochips FUSION sample holder with heating chips featuring a SiC membrane with nine holes of 10 μm diameter for the observation and providing uniform heating in the observation area. The growth of GaAs NWs is done using trimethylgallium (TMGa) and tertiarybutylarsine (TBAs) and catalysed Ge branches are grown using digermane (Ge₂H₆) diluted in H₂ (10% by volume). The mixture is then delivered to the microscope column by an injector connected near the pole piece.

For the growth of GaAs nanowires, Au nanoparticles are deposited by evaporation on the FUSION heating e-chips and then in situ annealed to generate Au nanodroplets. The growth is preformed at around 450 °C. The zinc blende to wurtzite phase transition is tuned by changing the V/III ratio during the growth with a pressure from 2 × 10⁻³ mbar to 1.10⁻² mbar. After the growth of GaAs NWs, organometallics are simultaneously stopped. The tip Au droplets are destabilised by increasing the temperature rapidly. After the Au droplets have moved to the sidewall of the GaAs-w NW, diluted Ge₂H₆ is injected at 3 sccm and a pressure of around 8 × 10⁻³ mbar and VLS growth of the Ge branch is observed at around 330 °C.

RTT and EPAMB designed the ex situ experiments, LV and CR designed and conducted the in situ experiments. AL and HITH have grown the NWs ex situ, RTT and AL performed the NW sidewall Au deposition, MAV performed the ex situ TEM analysis of the NWs. AL, HITH, RTT, MAV and EPAMB contributed to the interpretation of the ex situ data, and LV analysed the data from in situ experiments. HITH, EPAMB, and LV wrote the manuscript. All authors discussed the results and commented on the manuscript. LV and EPAMB supervised the project, and acquired the research funding that supported this study.

The authors declare no competing financial interest.