This is an open access article published under an ACS AuthorChoice License, which permits copying and redistribution of the article or any adaptations for non-commercial purposes. Article www.acsnano.org Probing Lattice Dynamics and Electronic Resonances in Hexagonal Ge and SixGe1−x Alloys in Nanowires by Raman Spectroscopy Diego de Matteis, Marta De Luca, Elham M. T. Fadaly, Marcel A. Verheijen, Miquel López-Suárez, Riccardo Rurali, Erik P. A. M. Bakkers, and Ilaria Zardo* Cite This: ACS Nano 2020, 14, 6845−6856 Read Online ACCESS Metrics & More Article Recommendations * sı Supporting Information ABSTRACT: Recent advances in nanowire synthesis have enabled the realization of crystal phases that in bulk are attainable only under extreme conditions, i.e., high temperature and/or high pressure. For group IV semiconductors this means access to hexagonal-phase SixGe1−x nanostructures (with a 2H type of symmetry), which are predicted to have a direct band gap for x up to 0.5−0.6 and would allow the realization of easily processable optoelectronic devices. Exploiting the quasi-perfect lattice matching between GaAs and Ge, we synthesized hexagonal-phase GaAs-Ge and GaAs-SixGe1−x core−shell nanowires with x up to 0.59. By combining position-, polarization-, and excitation wavelength- dependent μ-Raman spectroscopy studies with first-principles calculations, we explore the full lattice dynamics of these materials. In particular, by obtaining frequency−composition calibration curves for the phonon modes, investigating the dependence of the phononic modes on the position along the nanowire, and exploiting resonant Raman conditions to unveil the coupling between lattice vibrations and electronic transitions, we lay the grounds for a deep understanding of the phononic properties of 2H-SixGe1−x nanostructured alloys and of their relationship with crystal quality, chemical composition, and electronic band structure. KEYWORDS: Raman spectroscopy, nanowires, hexagonal (lonsdaleite) SiGe, resonant Raman, phonons, crystal structure transfer start of the “silicon revolution”.7 From 1960 to 2018 an T he exploitation of unique optoelectronic, mechanical, and thermal properties of nanostructured materials has a long history, even antecedent its deep understanding and systematic investigation. Indeed, the high surface-to- volume ratio of nanostructures, together with the possibility of estimated 13 × 1021 MOSFETs have been fabricated,8 and silicon has undoubtedly gained a central role in integrated circuit technologies and informatics. Nevertheless, Si and Ge hardly breached the field of optoelectronics, as in their bulk quantum confinement, results in interesting behaviors for optoelectronic, thermoelectric, biomedical, and many more crystalline form they exhibit an indirect band gap, resulting in applications.1−3 This diversity and abundance also come from poor light emission properties. On the contrary, as supported the possibility of obtaining materials in crystal structures that by both theoretical and experimental investigations, SixGe1−x are not easy to achieve in bulk forms. For example, GaAs compounds exhibit direct band gaps (for x up to 0.5−0.6) nanowires (NWs) can also be synthesized in the wurtzite when synthesized in the hexagonal form (with a 2H type of (WZ) crystal structure, which can be obtained in bulk only symmetry, also known as lonsdaleite), which can be the under extreme pressures,4−6 while the growth mechanisms starting point for an efficient class of optoelectronic devices.9,10 typical of NWs have provided a powerful route to obtain it. The availability of unusual crystal structures can be relevant even for fields in which there is already a general consensus on Received: January 28, 2020 the optimal materials for applications. This is the case, for Accepted: May 11, 2020 example, for group IV semiconductors. Their use in electronics Published: May 11, 2020 dates back to the discovery of the pn junction and has been so successful that we can now define the invention of the metal- oxide-semiconductor field-effect transistor (MOSFET) as the © 2020 American Chemical Society https://dx.doi.org/10.1021/acsnano.0c00762 6845 ACS Nano 2020, 14, 6845−6856 ACS Nano www.acsnano.org Article Figure 1. Epitaxial 2H-GaAs-SixGe1−x core−shell nanowire structures: (A, B) 30 degree tilted-view SEM images of arrays of 2H-Si0.27Ge0.73 (A) and Si0.59Ge0.41 (B) shells grown around thin (40 nm) WZ-GaAs nanowire cores on a GaAs (111)B substrate in the [0001] crystallographic direction. (C) HAADF-STEM image of the interface of the core−shell structures in (B) acquired in the [1120] zone axis, displaying the ABAB stacking along the [0001] direction, confirming the hexagonal crystal structure. (D, E) EDX elemental maps of cross- sectional lamellae of representative GaAs-Si0.27Ge0.73 and GaAs-Si0.59Ge0.41 core−shell NWs showing the GaAs NW core in blue, Si atoms in green, and Ge atoms in red. (F) Radial line scans along the arrows marked by the dashed boxes in the EDX maps shown in (D) and (E) quantifying the silicon content in both samples. The lack of information on the vibrational properties of material template.12 Here, we utilize single-crystalline WZ these materials and on their dependence on composition calls GaAs NWs as a hexagonal template to epitaxially grow high- for the exploration of their lattice dynamics, as the vibrational quality 2H-Ge and SixGe1−x crystals in a core−shell geometry. spectrum is closely related to the chemical composition, the The choice of GaAs NWs as a core is due to the fact that there crystal quality, and the alloy formation process, and it can also is almost no lattice mismatch with 2H-Ge and a small one with be a probe of the electronic band structure. In this work we Ge-rich SixGe1−x alloys. To further avoid the induced strain present an investigation of such properties performed by due to high Si content in SixGe1−x, we use very thin GaAs NWs means of μ-Raman spectroscopy on single NWs and cores (35−45 nm). The studied NW core−shell structures theoretical calculations based on density functional theory were fabricated in a metal organic vapor phase epitaxy (DFT). Hexagonal core−shell GaAs-SixGe1−x NWs with 0 ≤ x (MOVPE). The GaAs NWs were synthesized via catalyst- ≤ 0.59 were synthesized exploiting the quasi-perfect lattice assisted growth following the vapor liquid solid (VLS) matching between WZ GaAs and lonsdaleite Ge (referred to as mechanism using gold (Au) catalyst seeds and from trimethyl 2H in the following) with the crystal structure transfer gallium (TMGa) and arsine (AsH3) gas precursors. The method.11,12 resulting GaAs NWs are of very high crystal quality with only We were able to uncover the composition dependence of the few stacking faults (on average 2 or 3 stacking faults/μm, as NWs’ phononic spectrum, to analyze the homogeneity of the shown in the extended data of the figures in ref 13). Prior to NWs with spatially resolved measurements, and to probe the the SixGe1−x shell growth, the Au catalyst seeds were electronic band structure of the alloys with resonant Raman chemically etched to avoid Au contamination in the shell. scattering experiments. As a consequence, we could obtain a Then, the SixGe1−x shells were grown epitaxially around GaAs full picture of all the phonon modes of 2H-Ge and SixGe1−x, NW cores using germane (GeH4) and disilane (Si2H6) gas frequency−composition calibration curves for the main precursors. For further details on growth and structural phonon modes of the SixGe1−x alloys, an assessment of the characterization see the Methods section. In the same ref 13, quality and reliability of the NW synthesis process, and finally which shows NWs grown with nearly the same conditions as some insight into the optically active electronic bands’ energy the ones studied in this work, atom probe tomography and symmetry, as well as into their coupling with lattice measurements reveal the presence of As in the shell at a vibrations. concentration of about 1019 atoms/cm3. In this study, a wide range of 2H-SixGe1−x compositions (0 < RESULTS AND DISCUSSION x < 0.59) were studied; for details of NW dimensions and Hexagonal Ge and SixGe1−x Samples. We employ the compositions, see Table S1 and Figure S1 in the Supporting crystal structure transfer method to fabricate 2H-Ge and Information 1. Figure 1 shows the structural characterization SixGe1−x crystals as reported by Fadaly et al., where the for two typical samples of the studied composition range hexagonal crystal structure is adopted from a hexagonal (Si0.27Ge0.73 and Si0.59Ge0.41); pure 2H-Ge was extensively 6846 https://dx.doi.org/10.1021/acsnano.0c00762 ACS Nano 2020, 14, 6845−6856 ACS Nano www.acsnano.org Article described in ref 13. Figure 1A,B display two representative scanning electron microscopy (SEM) images of Si0.27Ge0.73 and Si0.59Ge0.41, respectively. The SEM images of the NW arrays show very small variations in the morphology, implying that they are uniform and the SEM images reflect the average morphology across the samples and, accordingly, the structures measured with Raman spectroscopy. To confirm the purely hexagonal crystal structure, we carried out high-resolution high angular annular darkfield (HAADF) scanning transmission electron microscopy (STEM) measurements. A representative image for GaAs/Si0.59Ge0.41 is shown in Figure 1C. The HAADF-STEM image displays an excellent epitaxial relation- ship between the GaAs and Si0.59Ge0.41 and confirms the ABAB stacking of the 2H crystal structure in the [0001] growth direction. To investigate the composition uniformity across the radial direction of the core−shell structure, a few NWs from the NW ensembles shown in Figure 1A,B were cut open and thin lamellae were extracted and analyzed using energy dispersive X-ray (EDX) spectroscopy as depicted in Figure 1D,E. The EDX elemental maps in Figure 1D,E and their corresponding line scans in Figure 1F confirm the uniform material composition after radial shell growth, resulting in a Figure 2. Computed DFPT Raman spectra in ZZ (black lines) and nominal Si composition of 27 ± 2% and 59 ± 4% for these two YY (red lines) scattering geometry of (from top to bottom) 2H- samples, respectively. We also observe very thin Ge-rich alloy pure Ge, Si0.06Ge0.94, Si0.19Ge0.81, and Si0.5Ge0.5 alloys. The small regions along a sunburst-like geometry, aligned with the peak at 222.3 cm−1 in the spectra of pure Ge YY (see the zoomed- in inset in the first panel) is due to the back-folding of the LA corners between the NW facets, a known effect in alloyed branch. The cartoon in the first panel shows the orientation of the nonplanar materials due to growth kinetic effects.14−16 We also axes with respect to the NW samples. notice a Ge-rich region around the core, an effect that we are investigating with further experiments and theoretical model- ing. The Si percentages in these sunburst regions have nominal The results from DFPT are summarized in Figure 2, compositions of 24.1 ± 0.6% and 58 ± 1% for the Si0.27Ge0.73 together with the calculated spectra of pure 2H-Ge for and Si0.59Ge0.41 samples, respectively, which is within the alloy reference. The spectra are generated by using Lorentzian fluctuation percentage within the core−shell structure (see curves with a full width at half-maximum (fwhm) of 3 cm−1 to Figure S2 in the Supporting Information SI1). approximately reproduce the experimental fwhm. In the 2H- Composition Dependence of the Raman Spectra. We Ge, in ZZ configuration only one mode is allowed: the have computed the Raman spectra within density-functional longitudinal optical mode, with symmetry A1g, at 297.7 cm−1. In the YY configuration, there are three modes allowed: an perturbation theory (DFPT) of SixGe1−x with x = 0.0625, 0.25, intense transverse optical mode with E2g symmetry at 286.1 and 0.5 in a 2 × 2 × 1 supercell of the four-atom 2H primitive cm−1, a low-intense transverse optical E1g mode at 297.4 cm−1, cell. More details are given in the Computational Details and the A1g mode at 297.7 cm−1, also visible in ZZ section. In the case of x = 0.25 and x = 0.5 we have generated 5 configuration. Since E1g and A1g are almost degenerate, they or 6 independent random configurations corresponding to the appear as a single peak, and we will label this mode as the A1g/ target nominal composition and averaged the computed E1g mode from now on (for simplicity, we keep this notation Raman spectra (x = 0.0625 corresponds to one single Si also in the ZZ configuration). All these features enable the atom in a Ge matrix, and therefore averaging is not necessary). distinction between group-IV hexagonal and cubic crystals (in In this way, we partially palliate the limitations deriving from a the latter no E2g mode exists, and the transverse and description of a disordered system by using a computational longitudinal optical modes are degenerate and have an F2g cell that is rather small and can thus be treated efficiently at the symmetry; see the calculated Raman spectra of cubic Ge in first-principles level. Despite these limitations, the theoretical Figure S317). Let us now focus on the 2H-SixGe1−x alloys. As it results are a useful guide for the interpretation of the can be seen in Figure 2, when one Si atom is added (x = experimental measurements discussed below. Polarization- 0.0625), a spectral feature close to 400 cm−1 appears, a clear resolved Raman scattering calculations and experiments were signature of a vibrational mode associated with the Si−Ge performed in the same backscattering geometry. Considering bond. The intensity and frequency of this mode increase with an {XYZ} reference system where the single NW is aligned to the Si content, tending to the computed values (indicated with the Z axis, the incident photon wavevector is antiparallel to the two arrows above the top panel) of an artificial 2H-SiGe lattice, X axis and the scattered photon wavevector is parallel to it (see i.e. a fully ordered crystal system where each Si atom is bonded inset of Figure 2). The excitation and detection light to four Ge atoms and vice versa, like in GaAs or SiC. When the polarization vectors, εi and εs, lie in the YZ plane, which is Si concentration increases further, we also observe the the plane of the sample. The scattering configuration is appearance of higher frequency Raman-active modes between expressed in terms of excitation and detection polarizations in 450 and 500 cm−1, which come from Si−Si bonds and tend to brackets. The calculations are performed either with the the values of pure 2H-Si (also in this case the Raman-active exciting and scattered light both polarized along (ZZ) the NW modes of 2H-Si are indicated with two arrows). Notice that growth axis or with both perpendicular (YY) to it. these spectral features were absent at x = 0.0625, where one 6847 https://dx.doi.org/10.1021/acsnano.0c00762 ACS Nano 2020, 14, 6845−6856 ACS Nano www.acsnano.org Article Figure 3. Comparison of the Raman spectra of five SixGe1−x alloy nanowires recorded in (A) ZZ scattering configuration and (B) YY scattering configuration. From bottom to top x = 0.13; 0.21; 0.33; 0.42; 0.59. Both panels also show the spectrum of a pure 2H-Ge NW acquired in the corresponding scattering configuration (purple line, bottom). The main bands are labeled with their respective oscillator. (C) Frequency of the main bands as a function of composition (symbols) and their fits (lines) extracted from the spectra shown in (A) and (B). The full symbols and solid lines represent the A1g/E1g components of the bands (spectra recorded in the ZZ configuration). The empty symbols and dashed lines represent the E2g components (spectra recorded in the YY configuration). Error bars are taken from the fit results. single Si atom was present and thus no Si−Si bond could be SiGe. Regarding the issue of asymmetric lineshapes, in cubic formed. Averaging over more random configurations or using a SixGe1−x alloys, it was shown that, as alloy disorder increases (x much larger computational cell would make the leading Ge− approaches 0.5), more contributions from phonons at the zone Ge, Si−Ge, and Si−Si peaks stand out more clearly with edge appear, which add low-energy tails to the Raman peaks. respect to the secondary peaks, particularly at x = 0.5, as This effect is stronger for the Si−Ge vibrations, just like in indeed happens in the experimental measurements where Figure 3A and B, where one can see that the Si−Ge band is the much larger volumes are probed. most asymmetric.22 Undoped cubic SixGe1−x alloys inves- We have then performed Raman scattering experiments to tigated in refs 23 and 24 also display lineshapes very similar to be compared to the calculations. We have measured NWs ours. All this considered, we attribute the asymmetric transferred by using a micromanipulator from the growth appearance of the Raman bands to alloy-induced disorder substrate to a SiC substrate. Measurements were performed at and convolution between features arising from slightly different room temperature with the 514 nm excitation wavelength. chemical environments, as described earlier. More details are given in the Methods section. In Figure 3A The experimental spectra measured on 2H-Ge are very and B we display representative Raman spectra collected on similar to the calculated spectra displayed in Figure 2. In the single GaAs-SixGe1−x (x = 0, 0.13, 0.21, 0.33, 0.42, 0.59, from YY scattering configuration we can see the transverse optical bottom to top) core−shell NWs in the ZZ and YY geometries. E2g mode at 287.7 cm−1, with a fwhm of 4 cm−1, and the Similarly to the theoretical spectra, the experimental spectra of degenerate transverse and longitudinal A1g/E1g modes at 299.4 the alloy NWs are characterized by the presence of three cm−1 (fwhm of 3 cm−1), with the first component being much intense features attributed to the Ge−Ge, Si−Ge, and Si−Si more intense than the second. In the ZZ configuration, instead, vibrations in order of increasing Raman shift. These peaks are the A1g/E1g modes are the only ones allowed, and they have broad and asymmetric because they result from the the same frequency and fwhm that they have in the YY convolution with secondary features that we assign to configuration. We stress that the absence of a line shape vibrations of one of the aforementioned atom pairs in various broadening on the phonon modes of 2H-Ge allows us to rule chemical environments. These appear as broad and generally out any major effect of As doping in the determination of the weak features, as they result from the averaged contributions of frequency of the phonon modes. Due to the high signal-to- various atomic arrangements18−20 that locally correspond to noise ratio of the measurements and the high structural quality deviations from the global stoichiometry. In addition to these of the samples, these values can be considered as reference effects we also considered the possible influence of the As values for 2H-Ge. Compared to the computed spectra, the dopants on the Raman spectra, in terms of phonon frequency small differences in the relative intensities of the peaks are due shifts and asymmetric broadening of phonon modes (e.g., Fano to experimental effects that are not taken into account in the lineshapes). In cubic Si and Ge, levels of As doping similar to calculations (i.e., size effects, dielectric mismatch effect, and ours typically induce a shift of the phonon modes of <1 effect of the finite numerical aperture of the microscope cm−1,21 and we can expect a similar effect on hexagonal Ge and objective25). 6848 https://dx.doi.org/10.1021/acsnano.0c00762 ACS Nano 2020, 14, 6845−6856 ACS Nano www.acsnano.org Article We performed a quantitative analysis of the spectra as concentration range. The Ge−Ge Raman shifts decrease deconvolution of multiple Lorentzian curves, which enables us linearly starting from the pure 2H-Ge phonon frequency of to find the minimum number of spectral contributions needed 299 cm−1. The trend of the last three points (x = 0.33, 0.42, to describe the experimental line shape. This analysis allows us 0.59) might suggest a damped dependence, as reported by to discuss the behavior of the main bands as well as the Herman and Magnotta24 and by Grein and Cardona,26 but also assignment of the various features, dividing the frequency linear trends have been reported.23,27 range according to the three atom pairs (an example of the In agreement with ref 28, the different composition performed deconvolution is displayed in Figures S4 and S5 in dependencies of the three modes can be explained by the SI2). For discussing the assignment of the peaks, besides those competition between two effects. One is the effect deriving ascribed to the above-discussed Ge−Ge, Si−Ge, and Si−Si from the fact that for any x other than the extremes the alloy modes, we adopt the notation presented in ref 19, with the lattice parameters are forced to be larger than pure Si and reminder that it was developed for a 1D computational smaller than pure Ge. This leads to an upshift of the description of the possible configurations, while we use it for frequencies as x increases (the lattice is compressed and the the general 3D environment. Therefore, we label the peaks as modes harden). The other effect is the disorder-induced (A−B)A;B, where between parentheses we state the atoms shorter order range that results in frequency downshifts with involved in the vibrating bond and as a superscript we indicate respect to an ideal lattice. See ref 29 for more details; also the prevalent element(s) surrounding them. For simplicity, we notice that disorder for Ge−Ge modes increases with x, while do not use the subscript indicating sensitivity to first or second for Si−Si modes it decreases, because the system approaches neighbors. The necessity of such a labeling stems from the fact pure Si. The strong composition dependence of the Si−Si that, in alloys, each species can be in different environments. Raman shift can come from the two effects acting in the same For example, any Si atom can have from zero to four Ge atoms direction. For Ge−Ge vibrations we can imagine the two as first neighbors. The five arrangements might not be all effects acting against each other, but overall resulting in a distinguishable in the Raman measurements and most slower red-shift. Finally, the Si−Ge vibrational frequencies importantly in the deconvolution: in those cases, we will behave similarly (increasing) to Si−Si for x < 0.5 and to Ge− employ multiple superscripts, separated by semicolons. The Ge (decreasing) for x > 0.5, exhibiting thus a nonmonotonic superscripts “SiGe” indicate that the contribution to the trend. Overall, we find a good qualitative agreement with spectrum comes from an atom oscillating in a mixed Raman shift vs composition trends reported in the environment. literature,19,23,24,27 and we ascribe quantitative discrepancies In the analysis of the spectra, we identify the higher energy to the limited concentration range at our disposal. Finally, we components of the bands (more pronounced in the ZZ stress that eqs 1.1−1.6) and the consequent discussion would scattering configuration) as the A1g/E1g modes and the lower be almost the same in the case of undoped Ge and SiGe, energy components (more pronounced in the YY scattering because As dopants are known to introduce a very small shift configuration) as the E2g modes (see SI2). As a result of this in the phonon frequencies (<1 cm−1, as discussed above). analysis, in Figure 3C we show the frequency of the prevalent In order to discuss the assignation of the spectral features to component of the main bands, according to the different bond + environment units, we divide the Raman shift range in polarization configurations, as a function of the alloy three sections, according to the three possible atom pairs. composition together with their fitting curves. The composi- In the highest energy range, i.e., 460 cm−1 < ω < 490 cm−1, tion-dependent frequencies, as derived from fitting the we find the Si−Si band. It is composed of two peaks that we experimental data, and thus strictly valid only for the measured can assign to (Si−Si)SiGe;Si, i.e., main peak at higher frequency, concentration range, are the following: and (Si−Si)Ge, i.e., lower intensity and lower frequency (Figure A1g/E1g modes (in ZZ configuration): S5). Indeed, the frequency of Si−Si vibrations decreases with ωSi − Si (cm−1) = 461.95 + 50.01x the increase of Ge concentration (see the discussion above); (1.1) that is, the weaker peak comes from the oscillation of Si atoms ωSi − Ge (cm−1) = 392.12 + 48.48x − 42x 2 in Ge-rich environments, while the main peak is the sum of the (1.2) contributions from Si atoms with zero to two Ge neighbors. ωGe − Ge (cm−1) = 299.03 − 16.67x (1.3) The (Si−Si)SiGe;Si peak emerges for x = 0.21, though at this composition it has lower intensity than the (Si−Si)Ge peak. E2g modes (in YY configuration): These considerations are consistent with the observations of Pages et al. in ref 19. ωSi − Si (cm−1) = 461.27 + 45.22x (1.4) The Si−Ge vibrations appear in the intermediate range of 370 cm−1 < ω < 450 cm−1 (Figure S5). The results of the ωSi − Ge (cm−1) = 388.36 + 34.26x − 29x 2 (1.5) deconvolution of the spectra are consistent with the picture of ωGe − Ge (cm−1) = 289.37 − 12.56x four distinguishable environments around a Si−Ge bond. We (1.6) propose a division of the Si−Ge contributions in three The Si−Si band shows the strongest dependency on the branches, and, analogously to what was mentioned before for composition and goes linearly toward ∼514 cm−1, the phonon the Si−Si vibrations, we can assign the branches to (Si−Ge)Ge, frequency of pure nanostructured 2H Si.11 The Si−Ge (Si−Ge)SiGe, and (Si−Ge)Si environments in order of frequencies are the only ones to show a nonlinear dependence increasing Raman shift. Regarding (Si−Ge)Si, we observe its on the composition. Given the general similarity of our trends splitting (highlighted by a circle in Figure 3A), as noted in refs to the ones reported in the literature23,24 for analogous cubic 19 and 26 for x ≲ 0.75. This results in a “pseudodoublet” (as it alloys, in which the frequency downshift is reported in the x > contains far more than two spectral features) and reveals the 0.6 range, we suggest that a quadratic fit is representative, fourth distinguishable environment. We assign the lower despite the second-order coefficient being small for this energy component of the pseudodoublet to Ge-rich environ- 6849 https://dx.doi.org/10.1021/acsnano.0c00762 ACS Nano 2020, 14, 6845−6856 ACS Nano www.acsnano.org Article Figure 4. Polarization dependence of the intensity of the main modes of the x = 0.42 sample with A1g/E1g (A) and E2g (B) symmetry collected with polarization of scattered light along Z and Y, respectively. The dashed lines are sine-squared fits of the experimental data. (C and D) Raman shifts of the A1g/E1g and E2g modes with the polarization of the collected light parallel and perpendicular to the NW long axis, respectively. ments and the higher energy one to Si-rich environments at the the point that the Ge−Ge band is almost identical between ZZ second neighbor scale, whereas we do not discriminate (Si− and YY scattering configurations for the x = 0.59 sample. More Ge)Ge and (Si−Ge)Si according to second neighbors. This is details about this effect are presented in the following the only case in which we have to invoke sensitivity to second discussion and in sections SI2 and SI3. nearest neighbors to explain the wealth of peaks. It is also Information about Raman selection rules can be obtained by worth noting that we observe another exchange of spectral polarization-dependent measurements. A full angular depend- weight, i.e., the equalization of (Si−Ge)Ge and (Si−Ge)SiGe at x ence of the E2g and A1g/E1g modes’ intensities in 2H-Ge NWs ≈ 0.16, but we had no means to investigate the one between is displayed in SI3. The selection rules found for all the modes (Si−Ge)SiGe and (Si−Ge)Si hypothesized for x ≈ 0.84. follow the theoretical expectation for hexagonal crystals, which Nevertheless, we can state that the lower frequency component confirms the high crystal quality of our samples.11,20,21 For x ≤ of the pseudodoublet, which at high Si content corresponds to 0.20 we can still distinguish the two components of the Ge− the only component of (Si−Ge)Si (the splitting takes place for Ge band in the YY scattering configuration, but for Si richer x < 0.75), grows with the increase in Si content. This is compositions the resolution is lost due to the increasing consistent with the possibility that it will eventually become disorder of the system. Namely, in the YY spectra we do not more intense than the main (Si−Ge)SiGe feature. resolve the E2g and A1g/E1g modes anymore, but we observe a The Ge−Ge vibrations appear in the 260 cm−1 < ω < 300 consistent red-shift of all three bands with respect to ZZ cm−1 range (Figure S4). In addition to the main intense resulting from the enhancement of the E2g phonon modes with feature, the deconvolution reveals the presence of a second, respect to the A1g/E1g modes. We also performed a full weaker, one. With arguments similar to those used for Si−Si polarization dependence study on an x = 0.42 NW: Raman modes and derived from the analysis of Figure 3C, we attribute spectra were collected rotating the polarization of the exciting the weak components observed between 260 and 280 cm−1 light in 20° steps. The polarization of the collected light was (depending on the value of x and on the polarization) to (Ge− selected either parallel to the NW long axis (Z, see Figure 2) or Ge)Si, which leaves the main Ge−Ge feature to be (Ge− perpendicular to the NW long axis (Y, see Figure 2). In Figure Ge)SiGe;Ge. The analysis of the lower energy component is, 4 we show the intensity and the Raman shift of the three main however, complicated by the rise of a very broad, possibly modes as a function of the polarization angle in the two disorder-induced, band between 150 and 260−270 cm−1. The detection configurations. The scattering intensities of the three E2g and A1g/E1g modes can only be distinguished by the main spectral components extracted from the parallel detection deconvolution analysis for x > 0.20. It is then important to measurements are highest for polarization parallel to the NW note that the increasing disorder brought by the increasing Si long axis and lowest for perpendicular polarization. Therefore, content progressively relaxes the Raman selection rules up to they follow the selection rules, which instead are partially 6850 https://dx.doi.org/10.1021/acsnano.0c00762 ACS Nano 2020, 14, 6845−6856 ACS Nano www.acsnano.org Article regularly over a shorter distance. For more detail about the Si− Si vibrations in the perpendicular detection experiments, see SI3. Table 1. Polarization Degrees of the Main Modes Compared It is important to note that, differently from the behavior of between the Two Detection Configurationsa the pure Ge sample, in which the two components (E2g and A1g/E1g) under each of the three bands are polarized A1g/E1g modes E2g modes (detection along Z) (detection along Y) perpendicularly to each other, in the x = 0.42 sample, they are polarized along the same direction. As displayed in Figure Ge−Ge 0.78 0.40 4C and D and in greater detail in Figures S7 and S8, the Si−Ge 0.79 0.37 Raman shifts are independent of the excitation polarization Si−Si 0.77 not available a angle as well as of detection polarization configuration. This Regarding the Si−Si mode in the Y detection, see SI3 for the reason confirms the accuracy of the data analysis and proves that no of its exclusion from this analysis. systematic error has been introduced. Therefore, we conclude that the alloy-related disorder has relaxed for the perpendicular detection data. In Table 1 we anisotropic effects on the polarization of the phonon modes report an estimation of the polarization degree of the modes, and on the relaxation of Raman selection rules, which are more Ri, calculated according to evident for the minority chemical species. Ii ,max − Ii ,min Spatially Dependent Raman Maps: Homogeneity of Ri = the Alloy. In order to assess the homogeneity of the crystal Ii ,max + Ii ,min (1.7) structure and alloy composition of the nanowires, we where Ii,max and Ii,min are the maximum and minimum performed spatially dependent Raman measurements. In intensities of the i-mode, respectively, as estimated from the Figure 5 we present the results of spatially resolved Raman sine-squared fit to reduce the error. The polarization degrees of measurements performed on x = 0 and x = 0.18 (see SI1) the A1g/E1g modes (detection parallel to Z) are comparable single SixGe1−x nanowires as 2D maps: the scattering intensity and above 75% (see also Figure 4A). On the other hand, the (false colors) is shown as a function of Raman shift and laser E2g modes have a rather small degree of polarization (see also position. Namely, the maps are obtained by acquiring Raman Figure 4B) that further decreases in vibrations involving Si spectra along the NW length with steps of 150 nm (x = 0 atoms. sample) and 330 nm (x = 0.18 sample). In the x = 0 sample we This is a combined effect of the alloy disorder and the NW find a cubic domain at the “top” of the NW (highlighted by the typical aspect ratio. The length of the NW in the Z direction circles in Figure 5), indicated by the slight blue-shift of the guarantees a degree of translational symmetry such that crystal phonon frequencies and the enhancement of the A1g/E1g mode momentum is a good quantum number and selection rules are in the YY configuration. This mode is allowed only with a low respected. Instead, the radial dimension of the NW is smaller intensity for this configuration in the lonsdaleite phase and is than the spot size, the probed phonon q vector range is larger instead the only component present for the diamond phase, at than in the parallel detection experiments, and selection rules slightly higher frequencies (F2g mode, >300 cm−130). are relaxed. We suggest that vibrations involving Si are affected This observation is in agreement with the structural the most because Si is the minority species in the sample and, characterization by TEM studies that exposed cubic Ge (or therefore, it is the least likely that a Si sublattice is reproduced SixGe1−x) is due to VLS growth of Ge (or SixGe1−x) on top of Figure 5. (A−D) Spatially resolved Raman measurements performed on (A, B) an x = 0 NW (pure Ge) and on (C, D) an x = 0.18 NW in the polarization configurations indicated above. (E, F) Peak positions extracted from fits at selected positions. 6851 https://dx.doi.org/10.1021/acsnano.0c00762 ACS Nano 2020, 14, 6845−6856 ACS Nano www.acsnano.org Article Figure 6. Maximum scattering intensity of the main bands as a function of excitation energy for three SixGe1−x NWs. Top row: data from spectra recorded in the ZZ configuration; bottom row: data from spectra recorded in the YY configuration. The gray areas highlight resonance-related enhancements of the intensity and the letters indicate points of the Brillouin zone at which the corresponding electronic transitions happen.13,31,33 (A, D) x = 0: the resonance profile is peaked at 1.91 and 2.33 eV. These resonances can correspond to electronic transitions at the A and M point, respectively; (B, E) x = 0.21: there is a resonance at 1.91 eV, compatible with an electronic transition at the A point; (C, F) x = 0.59: the resonance profile is peaked at 1.83 eV in YY and at 2.21 eV in ZZ. These maxima agree with electronic transitions at the A and M point, respectively. the GaAs core. The x = 0.18 sample shows a blue-shift of the structure or the composition are not homogeneous. We found Ge−Ge band at the upper end of both scans (ZZ and YY). If similar results when performing Raman scans on the x = 0.59 we had only a change in stoichiometry and not in crystal sample (not shown). structure, we would still see an intense E2g contribution in YY Resonant Raman for Band-Structure Probing. We at about 287 cm−1, whereas we observe one feature at about investigated the electronic band structure of three SixGe1−x 297 cm−1 with a low-energy tail that we assign to the F2g mode samples (x = 0, 0.21, 0.59) by means of resonant Raman. We of the cubic structure. Therefore, we can discard the performed Raman scattering experiments on single NWs with hypothesis of the Ge accumulation at the tip of the NW several excitation wavelengths, ranging from 457 to 681 nm. As (i.e., a segregated phase with different composition than the a reference, a cubic Ge substrate has been used. The cubic Ge body of the NW) and confirm the presence of a cubic domain sample shows only one maximum at 2.18 eV (see Figure S10 in also for x = 0.18. The formation of these domains was however SI5), in agreement with refs 17 and 30. More details on the already accounted for in the synthesis method of the samples, experimental procedure can be found in SI5. for which there is still some axial growth during the radial In Figure 6 we report the normalized maximum scattering growth of the shell, and this extra portion tends to assume the intensities as functions of the excitation energy. Comparing our cubic structure (see also Figure 1A and B). For more details data with electronic band structure calculations, we were able about the cubic segments, see Figure S9 in SI4. With the to link enhancements in the Raman scattering intensity to exception of this cubic domain at the end of the NW growth, electronic transitions at high-symmetry points of the BZ.13,31 the measured alloy composition and crystal quality of the NWs We carried out the assignment of the transitions keeping in are homogeneous. The good control of the synthesis mind that the calculations were performed for a system at 0 K conditions in terms of composition and crystalline quality is and, for x > 0, choosing the electronic bands deriving from the further corroborated by the results of the quantitative analysis most probable atomic arrangements (see refs 13 and 31 for of the spatially dependent Raman studies. In Figure 5E and F details). To address the possible influence of the As dopants on we plot the peak frequency extracted from analysis of the the resonant Raman profiles, we considered that, as shown in spectra at selected positions along the NW length: except for ref 13, the As-related electronic levels merge together in a the discussed blue-shifts, the phonon frequencies stay almost “dopant” band, which in turn merges with the lowest-lying constant throughout the length of the wires. Also, the fwhm of conduction band.32 This is consistent with the monoexponen- the peaks (see Figure S10 in SI4) shows little variation along tial decay of the PL intensity, meaning the presence of the the NW length. The bands would otherwise broaden or they dopants does not open other emissive channels than just direct would shift in energy if at certain positions along the NW the band gap emission. It is reasonable to assume that these laser illuminated structurally different regions, i.e., if the crystal considerations would apply not only to the gap but also to the 6852 https://dx.doi.org/10.1021/acsnano.0c00762 ACS Nano 2020, 14, 6845−6856 ACS Nano www.acsnano.org Article points of the BZ probed in our work. Despite the band energy are presented. The Si−Si peak is the most intense in the merging might result in transition energy changes with respect ZZ scattering configuration throughout the excitation energy to an undoped material, this would be hardly detectable in our range, while the same is generally true for the Si−Ge peak in experiments, for two reasons: this change could be masked by the YY scattering configuration. The peculiarity is the stronger the configurational fluctuations of the alloys; resonance- enhancement of Ge−Ge with respect to Si−Ge found at both induced Raman scattering enhancements are generally broad resonances (Figure 6C and F). We suggest this is due to an in energy. outgoing resonance of the Stokes photons after emission of the The pure 2H-Ge sample exhibits a first resonance around Ge−Ge phonon energy. It might indeed be that the emission 1.9 eV and another one at 2.33 eV in both the ZZ (Figure 6A) of the Ge−Ge phonon energy brings the Stokes photons in and YY (Figure 6D) polarization configurations, compatible better resonance with the electronic transitions, specifically with electronic transitions at the A point and at the M point, enhancing this vibration mode. respectively.13,31 The x = 0.21 sample shows a resonance at 1.91 eV in both polarizations (Figure 6B and E), compatible CONCLUSIONS with a transition at the A point.13,31 We did not recover the We perform Raman scattering measurements on seven GaAs- high-energy resonance observed for x = 0, and unfortunately SixGe1−x core−shell nanowire samples of different composi- the lack of laser sources between 2 and 2.2 eV keeps us from tion, namely, x = 0, 0.13, 0.18, 0.21, 0.33, 0.42, and 0.59. In the clearly identifying other maxima within this energy range. x = 0 sample (pure Ge), we find one main phonon band, However, the increase of signal at 2.18 eV especially in the YY composed by the transverse optical E2g mode at ∼288 cm−1 configuration (Figure 6E) hints at the existence of another (which represents the signature of the hexagonal phase) and by maximum in the Raman resonance profile. For the x = 0.59 Si the degenerate transverse and longitudinal A1g/E1g modes at sample, we found a low-energy resonance around 1.85 eV in ∼299 cm−1. The spectra of x > 0 samples are instead formed the YY scattering configuration (Figure 6F), consistent with a by three prominent bands, attributed to Ge−Ge, Si−Ge, and transition at the A point, and a high-energy resonance at 2.21 Si−Si vibrations in order of increasing Raman shift. Each of eV in the ZZ scattering configuration (Figure 6C), in these bands is composed by an E2g-like mode and A1g/E1g-like agreement with a transition at the M point.13,31 modes, which can be distinguished by polarization-resolved If we now focus on the pure 2H-Ge (x = 0) data, it is studies. We are able to uncover the frequency dependence on important to note the slightly stronger enhancement of the E2g the composition of the two components of each of the three phonon mode in the YY scattering configuration with respect bands, most importantly of the hexagonal signature E2g-like to the A1g/E1g enhancement in the ZZ scattering configuration modes. Furthermore, as the vibrations of Ge−Ge, Si−Ge, and at 1.91 eV. The scattering intensities measured at 2.33 eV, Si−Si atom pairs happen in different chemical environments, instead, show a stronger coupling of the electronic transitions this results in the appearance of additional spectral features, to the A1g/E1g mode. On the basis of the electronic band broader and weaker than the three main bands. We assign each structure calculations we identify these transitions as Γ9v/Γ7v+ feature to the most probable atomic configuration based on → Γ8c/Γ7c at the A point and Γ9v → Γ8c at the M point, qualitative and quantitative considerations about their respectively, taking into account the degeneracy of the bands at frequency and intensity as functions of the alloy composition. the A point. Considering the growth direction of the NWs and We also investigate the Raman selection rules of each their orientation during the measurements, the different phonon mode by carrying out extensive polarization-depend- enhancements according to light polarization lead us to believe ent Raman studies on samples with x = 0 and x = 0.42. that the symmetry product Γ9v ⊗ Γ8c is such that the electronic Regarding the former, we found that the E2g mode is polarized transition at M is favored by light polarized parallel to ⟨0001⟩, perpendicularly to the NW long axis, while the A1g/E1g modes namely, the NW long symmetry axis. For the other resonance, are polarized parallel to it. Regarding the latter instead, we there is a considerable enhancement in both polarization quantitatively estimate how a Si atomic percentage of 42 has configurations with a slight preference for excitation the effect of anisotropically relaxing the Raman selection rules perpendicular to ⟨0001⟩. This is explainable if we consider and how this effect is stronger for the minority chemical that we are detecting the relaxation (via the electron−phonon species (Si-related vibrations). In addition, we also show how, scattering) of carriers not only exactly at the A point, where the contrarily to the x = 0 sample, the E2g and A1g/E1g modes are bands become degenerate, but also in its vicinities, where they polarized along the same direction, namely, the one at which can both express their different symmetry character. The the polarization of the collected light is fixed. proliferation of possible atomic configurations and, therefore, By performing spatially resolved measurements on x = 0, x = of their corresponding electronic bands complicates this 0.18, and x = 0.59 samples, we assess the homogeneity and analysis for the alloy samples. crystalline quality of the NWs. Through the observation of the In the case of the SixGe1−x alloys, it is interesting to look at peak blue-shift and enhancement of forbidden modes we what specific phonon mode is more sensitive to the variation of identify cubic domains at one end of the NWs, which is a excitation energy. For the x = 0.21 sample, the Ge−Ge peaks consequence of the NW growth process. Otherwise, we are the most intense in the spectra due to the very small illustrate how peak frequencies and line widths stay constant percentage of Si. Accordingly, the Ge−Ge mode exhibits the throughout the NWs, which testifies their very good crystal largest enhancement effect, followed by Si−Ge and finally Si− quality. Si, which shows almost no response to the varying excitation The results of the resonant Raman scattering experiments energies. The behavior of the 59% Si sample is not so simple, performed on 2H-SixGe1−x alloys are compared with electronic starting from the change of the dominant peak with respect to band structure calculations and show good agreement between light polarization. One can already appreciate this by excitation energies coinciding with enhancements of the comparing Figure 3A and B, where the ZZ and YY spectra of Raman intensity and possible electronic transitions at various a 59% Si NW recorded with 514 nm (2.41 eV) of excitation points of the Brillouin zone,13,31,33 different than the Γ point, 6853 https://dx.doi.org/10.1021/acsnano.0c00762 ACS Nano 2020, 14, 6845−6856 ACS Nano www.acsnano.org Article thus providing a crucial insight into the electronic band In ∝ |εiR nεs|2 structure of this material. Regarding the x = 0 sample, we infer where εi and εs are the polarization vectors of the incident and the symmetry properties of the electronic bands involved in the scattered light, respectively, and Rn is the Raman susceptibility tensor. transitions by comparing the results obtained in polarization- The spectra were then generated by summing Lorentzian functions resolved experiments, specifically observing different coupling centered at each computed frequency, with a full width at half- of transitions at different points of the BZ with phonons of maximum of 3 cm−1. For a discussion on the choice of the k-point different symmetry. As far as the x = 0.21 and x = 0.59 samples mesh see ref 18. are concerned, we observe different resonance profiles with We used a 2 × 2 × 1 supercell of the four-atom unit cell of 2H respect to the polarization configuration. For the x = 0.59 crystals, randomly populating the lattice sites in order to reproduce sample, we observe a sudden enhancement of the Ge−Ge the target composition of the alloy. In particular we used 1 Si atom and 15 Ge atoms for x = 0.0625, 3 Si atoms and 13 Ge atoms for x = vibrational mode, and we argue that it is due to better match of 0.19, and 8 Si atoms and 8 Ge atoms for x = 0.50. In the case of x = the Stokes photons (after loss of the Ge−Ge energy) with the 0.19 and x = 0.50 we independently generated five and six random electronic transition energy. configurations, respectively, and performed a weighted average of the This work provides a deep investigation of the lattice corresponding polarized Raman spectra, taking as weights the dynamics and of the electronic resonances of Ge and SixGe1−x cohesive energies of each alloy. In this way configurations that are alloys, which can be grown in the hexagonal phase owing to the energetically favored, and thus more likely to occur, contributed more nanowires’ geometry, and prompts to the use of these to the averaged spectrum. No averaging was done for x = 0.0625, nanostructures in optoelectronics and phononic devices. where one single Si atom is added substitutionally to the 2H-Ge lattice. We compared our results with those obtained averaging on more configurations of a smaller unit cell, finding qualitatively similar METHODS results, but somewhat more noisy spectra. For this reason, we based Synthesis of Nanowires. The core−shell nanowire structures our analysis on the larger 16-atom cells. The calculations were were grown in a close coupled showerhead (CCS) MOVPE reactor performed for a system at 0 K. via catalyst-assisted growth following the VLS mechanism utilizing Au Raman Spectroscopy Details. Raman experiments were catalyst seeds. The Au catalyst seeds were deposited in a nanodisk performed by exciting the samples with different lasers: the 633 nm array arrangement on a GaAs (111)B substrate via the electron beam line of a HeNe laser, the different wavelengths of Ar and Ar-Kr lasers, lithography technique. The growth was performed at a reactor flow of the 532 nm of a diode-pumped solid-state laser, the 561 nm of an 8.2 slm (standard liters per minute) and a reactor pressure of 50 mbar. optically pumped semiconductor laser, and the 681 nm wavelength of a wavelength-stabilized diode laser. An incident power of 40 μW was For the GaAs NWs, the growth template was annealed at a set used to avoid heating or damaging effects. The samples were thermocouple temperature of 635 °C under an AsH3 flow set to a illuminated with a high numerical aperture (0.95) 100× objective, molar fraction of χAsH3 = 6.1 × 10−3. Then, the growth was performed which allows a spatial resolution of about 600 nm for the 514.5 nm at a set temperature of 650 °C with TMGa and AsH3 as material wavelength. The backscattered light was collected by a T64000 triple precursors set to molar fractions of χTMGa = 1.9 × 10−5 and χAsH3 = spectrometer in subtractive mode, equipped with 1800 g/mm gratings 4.55 × 10−5, respectively, resulting in a V/III ratio of 2.4. After the and a liquid-nitrogen-cooled CCD detector. The polarization of the growth of the GaAs core NWs, they were chemically treated with a incident laser beam and of the backscattered light was controlled and cyanide-based solution to remove the Au catalyst particles to avoid selected by means of lambda-half-waveplates and linear polarizers. gold contamination in the SixGe1−x shells. Eventually, the GaAs NW core was used as a hexagonal material template and was overgrown ASSOCIATED CONTENT with a SixGe1−x shell by introducing the suitable gas precursors for the * sı Supporting Information shell growth, which are GeH4 and Si2H6. The SixGe1−x shell was The Supporting Information is available free of charge at grown at a set temperature of 650 °C at a molar fraction of χSiGe = https://pubs.acs.org/doi/10.1021/acsnano.0c00762. 1.55 × 10−4 for a certain time according to the desired thickness. Structural Characterization. For the cross-section TEM studies, Further structural characterization of the nanowires, NWs were prepared using focused ion beam (FIB). In the different details on computational and spectroscopic methods and imaging modes, HR-TEM and STEM, analyses were conducted using analysis, and Raman spectra taken on different nano- a JEM ARM200F probe-corrected TEM operated at 200 kV. For the wires, with different scattering geometries and with chemical analysis, EDX measurements were carried out using the different excitation wavelengths (PDF) same microscope equipped with a 100 mm2 EDX silicon drift detector. TEM lamellae were prepared in a FEI Nova Nanolab 600i dual-beam system. For this, the NWs were initially transferred with AUTHOR INFORMATION the aid of a nanomanipulator from the growth substrate to a piece of Corresponding Author Si and then arranged to lie parallel to each other. These NWs were Ilaria Zardo − Departement Physik, Universität Basel, 4056 covered with electron- and ion-beam-induced metal deposition to Basel, Switzerland; orcid.org/0000-0002-8685-2305; protect them during the FIB milling procedure. The lamella was cut Email: ilaria.zardo@unibas.ch out by milling with 30 kV Ga ions and thinned down with subsequent steps of 30, 16, and 5 kV ion milling in order to minimize the Ga- Authors induced damage in the regions imaged with TEM. Diego de Matteis − Departement Physik, Universität Basel, Computational Details. We performed density-functional 4056 Basel, Switzerland calculations with the Abinit code,34,35 using a plane wave cutoff of Marta De Luca − Departement Physik, Universität Basel, 4056 38 Ha, the local density approximation for the exchange−correlation Basel, Switzerland; orcid.org/0000-0001-6061-3261 energy, and norm-conserving pseudopotentials.36 The Raman susceptibility tensor was computed within DFPT37 from the third Elham M. T. Fadaly − Department of Applied Physics, derivative of the total energy, twice with respect to the application of Eindhoven University of Technology, 5612AP Eindhoven, The an electric field (i.e., incident and scattered light polarization vectors) Netherlands; orcid.org/0000-0001-7074-8784 and once with respect to the phonon displacement coordinates, Marcel A. Verheijen − Department of Applied Physics, making use of the 2n+1 theorem.38 The off-resonance Raman Eindhoven University of Technology, 5612AP Eindhoven, The intensity of each mode was calculated as Netherlands 6854 https://dx.doi.org/10.1021/acsnano.0c00762 ACS Nano 2020, 14, 6845−6856 ACS Nano www.acsnano.org Article Miquel López-Suárez − Institut de Ciència de Materials de (7) Feldman, L. C. Introduction. In Fundamental Aspects of Silicon Barcelona (ICMAB−CSIC), 08193 Bellaterra, Barcelona, Oxidation, ed 1; Yves, J. C., Ed.; Springer-Verlag: Berlin Heidelberg, Spain 2001; Vol. 46, pp 1−8. Riccardo Rurali − Institut de Ciència de Materials de Barcelona (8) Laws, D. Computer History Museum, https://computerhistory. (ICMAB−CSIC), 08193 Bellaterra, Barcelona, Spain; org/blog/13-sextillion-counting-the-long-winding-road-to-the-most- frequently-manufactured-human-artifact-in-history (accessed July 20, orcid.org/0000-0002-4086-4191 2019). Erik P. A. M. Bakkers − Department of Applied Physics, (9) Von Den Driesch, N.; Stange, D.; Wirths, S.; Mussler, G.; Eindhoven University of Technology, 5612AP Eindhoven, The Holländer, B.; Ikonic, Z.; Hartmann, J. M.; Stoica, T.; Mantl, S.; Netherlands; orcid.org/0000-0002-8264-6862 Grützmacher, D.; Buca, D. Direct Bandgap Group IV Epitaxy on Si Complete contact information is available at: for Laser Applications. Chem. Mater. 2015, 27, 4693−4702. https://pubs.acs.org/10.1021/acsnano.0c00762 (10) Soref, R.; Buca, D.; Yu, S.-Q. Group IV Photonics: Driving Integrated Optoelectronics. Opt. Photonics News 2016, 27, 32−39. Author Contributions (11) Hauge, H. I. T.; Verheijen, M. A.; Conesa-Boj, S.; Etzelstorfer, T.; Watzinger, M.; Kriegner, D.; Zardo, I.; Fasolato, C.; Capitani, F.; I.Z. conceived the experiment. D.D.M. and M.D.L. performed Postorino, P.; Kölling, S.; Li, A.; Assali, S.; Stangl, J.; Bakkers, E. P. A. the Raman measurements, which were analyzed by D.D.M. M. Hexagonal Silicon Realized. Nano Lett. 2015, 15, 5855−5860. M.L.S. and R.R. performed the theoretical calculations. (12) Algra, R. E.; Hocevar, M.; Verheijen, M. A.; Zardo, I.; Immink, E.M.T.F. and E.P.A.M.B. grew the samples. M.A.V. performed G. G. W.; Van Enckevort, W. J. P.; Abstreiter, G.; Kouwenhoven, L. TEM investigations. D.D.M., M.D.L., and I.Z. wrote the P.; Vlieg, E.; Bakkers, E. P. A. M. Crystal Structure Transfer in Core/ manuscript with contributions of all authors. All authors have Shell Nanowires. Nano Lett. 2011, 11, 1690−1694. given approval to the final version of the manuscript. (13) Fadaly, E. M. T.; Dijkstra, A.; Suckert, J. R.; Ziss, D.; Tilburg, M. A. J. v.; Mao, C.; Ren, Y.; Lange, V. T. v.; Kölling, S.; Verheijen, M. Notes A.; Busse, D.; Rödl, C.; Furthmüller, J.; Bechstedt, F.; Stangl, J.; The authors declare no competing financial interest. Finley, J. J.; Botti, S.; Haverkort, J. E. M.; Bakkers, E. P. A. M. Direct Bandgap Emission from Hexagonal Ge and SiGe Alloys. Nature 2020, ACKNOWLEDGMENTS 580, 205−209. (14) Biasiol, G.; Reinhardt, F.; Gustafsson, A.; Martinet, E.; Kapon, This project has received funding from the European Research E. Structure and Formation Mechanisms of AlGaAs V-Groove Council (ERC) under the European Union’s Horizon 2020 Vertical Quantum Wells Grown by Low Pressure Organometallic research and innovation program (grant agreement no. Chemical Vapor Deposition. Appl. Phys. Lett. 1996, 69, 2710−2712. 756365). M.D.L. acknowledges support from the Swiss (15) Biasiol, G.; Kapon, E. Mechanisms of Self-Ordering of National Science Foundation Ambizione grant (grant no. Quantum Nanostructures Grown on Nonplanar Surfaces. Phys. Rev. PZ00P2_179801). R.R. acknowledges financial support by the Lett. 1998, 81, 2962−2965. Ministerio de Economı ́a, Industria y Competitividad (MINE- (16) Heiss, M.; Fontana, Y.; Gustafsson, A.; Wüst, G.; Magen, C.; CO) under grant FEDER-MAT2017-90024-P and the Severo O’Regan, D. D.; Luo, J. W.; Ketterer, B.; Conesa-Boj, S.; Kuhlmann, Ochoa Centres of Excellence Program under grant SEV-2015- A. V.; Houel, J.; Russo-Averchi, E.; Morante, J. R.; Cantoni, M.; 0496 and by the Generalitat de Catalunya under grant no. Marzari, N.; Arbiol, J.; Zunger, A.; Warburton, R. J.; Fontcuberta, I.; 2017 SGR 1506. E.P.A.M.B. and E.M.T.F. acknowledge Morral, A. Self-Assembled Quantum Dots in a Nanowire System for European Union’s Horizon 2020 research and innovation Quantum Photonics. Nat. Mater. 2013, 12, 439−444. program under grant agreement no. 735008 (SiLAS). (17) Cerdeira, F.; Dreybrodt, W.; Cardona, M. Resonant Raman E.P.A.M.B. and M.A.V. acknowledge Solliance, a solar energy Scattering in Germanium. Solid State Commun. 1972, 10, 591−595. (18) Scarle, S.; Mainwood, A. Modelling of Local Modes in SixGe1‑x RD initiative of ECN, TNO, Holst, TU/e, IMEC, and CxSiyGe1‑x‑y Alloys to Explore the Local Clustering of the Species. Forschungszentrum Jü lich, and the Dutch province of Phys. B 1999, 273−274, 616−619. Noord-Brabant for funding the TEM facility. R.R. thanks (19) Pagès, O.; Souhabi, J.; Torres, V. J. B.; Postnikov, A. V.; Silvana Botti for useful discussions. Rustagi, K. C. Re-Examination of the SiGe Raman Spectra: Percolation/One-Dimensional- Cluster Scheme and Ab Initio REFERENCES Calculations. Phys. Rev. B: Condens. Matter Mater. Phys. 2012, 86, (1) De, M.; Ghosh, P. S.; Rotello, V. M. Applications of 1−19. Nanoparticles in Biology. Adv. Mater. 2008, 20, 4225−4241. (20) Torres, V. J. B.; Hajj Hussein, R.; Pagès, O.; Rayson, M. J. (2) Karkan, S. F.; Mohammadhosseini, M.; Panahi, Y.; Milani, M.; Clustering/Anticlustering Effects on the GeSi Raman Spectra at Zarghami, N.; Akbarzadeh, A.; Abasi, E.; Hosseini, A.; Davaran, S. Moderate (Ge, Si) Contents: Percolation Scheme vs. Ab Initio Magnetic Nanoparticles in Cancer Diagnosis and Treatment: A Calculations. J. Appl. Phys. 2017, 121, 085704. Review. Artif. Cells, Nanomed., Biotechnol. 2017, 45, 1−5. (21) Cerdeira, F.; Cardona, M. Effect of Carrier Concentration on (3) Sosna-Glebska, A.; Szczecinska, N.; Znajdek, K.; Sibinski, M. the Raman Frequencies of Si and Ge. Phys. Rev. B 1972, 5, 1440− Review on Metallic Oxide Nanporticles and their Application in 1454. Optoelectronic Devices. Acta Innovations 2019, 30, 5−15. (22) Renucci, A.; Renucci, J. B.; Cardona, M. Volume Dependence (4) Biermanns, A.; Breuer, S.; Davydok, A.; Geelhaar, L.; Pietsch, U. of the Raman Frequencies of Ge-Si Alloys. Solid State Commun. 1971, Structural Polytypism and Residual Strain in GaAs Nanowires Grown 9, 1651−1654. on Si(111) Probed by Single-Nanowire X-Ray Diffraction. J. Appl. (23) Pezzoli, F.; Bonera, E.; Grilli, E.; Guzzi, M.; Sanguinetti, S.; Crystallogr. 2012, 45, 239−244. Chrastina, D.; Isella, G.; von Känel, H.; Wintersberger, E.; Stangl, J.; (5) Dubrovskii, V. G.; Sibirev, N. V. Growth Thermodynamics of Bauer, G. Raman Spectroscopy Determination of Composition and Nanowires and Its Application to Polytypism of Zinc Blende III-V Strain in Si1‑x Gex/Si Heterostructures. Mater. Sci. Semicond. Process. Nanowires. Phys. Rev. B: Condens. Matter Mater. Phys. 2008, 77, 1−8. 2008, 11, 279−284. (6) Panse, C.; Kriegner, D.; Bechstedt, F. Polytypism of GaAs, InP, (24) Herman, I. P.; Magnotta, F. Ge-Si Alloy Microstructure InAs, and InSb: An Ab Initio Study. Phys. Rev. B: Condens. Matter Fabrication by Direct-Laser Writing with Analysis by Raman Mater. Phys. 2011, 84, 1−9. Microprobe Spectroscopy. J. Appl. Phys. 1987, 61, 5118−5128. 6855 https://dx.doi.org/10.1021/acsnano.0c00762 ACS Nano 2020, 14, 6845−6856 ACS Nano www.acsnano.org Article (25) Fasolato, C.; Zardo, I.; De Luca, M. Addressing Crystal Structure in Semiconductor Nanowires by Polarized Raman Spec- troscopy. In Fundamental Properties of Semiconductor Nanowires, ed. 1; Fukata, N., Rurali, R., Eds.; Springer Nature: Switzerland AG, 2020, in press. (26) Grein, C. H.; Cardona, M. Local Vibrational Modes of Ge-Rich c-Si1‑xGex Alloys. Phys. Rev. B: Condens. Matter Mater. Phys. 1992, 45, 8328−8333. (27) Brya, W. J. Raman Scattering in GeSi Alloys. Solid State Commun. 1973, 12, 253−257. (28) Menéndez, J.; Pinczuk, A.; Bevk, J.; Mannaerts, J. P. Raman Study of Order and Disorder in SiGe Ultrathin Superlattices. J. Vac. Sci. Technol., B: Microelectron. Process. Phenom. 1988, 6, 1306. (29) Jusserand, B.; Paquet, D.; Regreny, A. Optical Phonon Quantum Levels in GaAs/Ga1‑xAlxAs Superlattices. Superlattices Microstruct. 1985, 1, 61−66. (30) Fasolato, C.; De Luca, M.; Djomani, D.; Vincent, L.; Renard, C.; Di Iorio, G.; Paillard, V.; Amato, M.; Rurali, R.; Zardo, I. Crystalline, Phononic, and Electronic Properties of Heterostructured Polytypic Ge Nanowires by Raman Spectroscopy. Nano Lett. 2018, 18, 7075−7084. (31) Rödl, C.; Furthmüller, J.; Suckert, J. R.; Armuzza, V.; Bechstedt, F.; Botti, S. Accurate Electronic and Optical Properties of Hexagonal Germanium for Optoelectronic Applications. Phys. Rev. Mater. 2019, 3, 1−11. (32) Alexander, M. N.; Holcomb, D. F. Semiconductor-to-Metal Transition in n-Type Group IV Semiconductors. Rev. Mod. Phys. 1968, 40, 815−829. (33) Cartoixà, X.; Palummo, M.; Hauge, H. I. T.; Bakkers, E. P. A. M.; Rurali, R. Optical Emission in Hexagonal SiGe Nanowires. Nano Lett. 2017, 17, 4753−4758. (34) Gonze, X.; Amadon, B.; Anglade, P. M.; Beuken, J. M.; Bottin, F.; Boulanger, P.; Bruneval, F.; Caliste, D.; Caracas, R.; Côté, M.; Deutsch, T.; Genovese, L.; Ghosez, P.; Giantomassi, M.; Goedecker, S.; Hamann, D. R.; Hermet, P.; Jollet, F.; Jomard, G.; Leroux, S.; et al. ABINIT: First-Principles Approach to Material and Nanosystem Properties. Comput. Phys. Commun. 2009, 180, 2582−2615. (35) Veithen, M.; Gonze, X.; Ghosez, P. Nonlinear Optical Susceptibilities, Raman Efficiencies, and Electro-Optic Tensors from First-Principles Density Functional Perturbation Theory. Phys. Rev. B: Condens. Matter Mater. Phys. 2005, 71, 1−14. (36) Troullier, N.; Martins, J. L. Efficient Pseudopotentials for Plane-Wave Calculations. Phys. Rev. B: Condens. Matter Mater. Phys. 1991, 43, 1993−2006. (37) Baroni, S.; de Gironcoli, S.; Dal Corso, A.; Giannozzi, P. Phonons and Related Crystal Properties from Density-Functional Perturbation Theory. Rev. Mod. Phys. 2001, 73, 515. (38) Gonze, X.; Vigneron, J. P. Density-Functional Approach to Nonlinear-Response Coefficients of Solids. Phys. Rev. B: Condens. Matter Mater. Phys. 1989, 39, 13120−13128. 6856 https://dx.doi.org/10.1021/acsnano.0c00762 ACS Nano 2020, 14, 6845−6856