ARTICLE https://doi.org/10.1038/s41467-022-31778-3 OPEN Sub-nanometer mapping of strain-induced band structure variations in planar nanowire core-shell heterostructures Sara Martí-Sánchez1,7, Marc Botifoll1,7, Eitan Oksenberg 2, Christian Koch1, Carla Borja1, Maria Chiara Spadaro 1, Valerio Di Giulio 3, Quentin Ramasse 4,5, F. Javier García de Abajo3,6, Ernesto Joselevich 2 & Jordi Arbiol 1,6 ✉ 1234567890():,; Strain relaxation mechanisms during epitaxial growth of core-shell nanostructures play a key role in determining their morphologies, crystal structure and properties. To unveil those mechanisms, we perform atomic-scale aberration-corrected scanning transmission electron microscopy studies on planar core-shell ZnSe@ZnTe nanowires on α-Al2O3 substrates. The core morphology affects the shell structure involving plane bending and the formation of low- angle polar boundaries. The origin of this phenomenon and its consequences on the elec- tronic band structure are discussed. We further use monochromated valence electron energy-loss spectroscopy to obtain spatially resolved band-gap maps of the heterostructure with sub-nanometer spatial resolution. A decrease in band-gap energy at highly strained core-shell interfacial regions is found, along with a switch from direct to indirect band-gap. These findings represent an advance in the sub-nanometer-scale understanding of the interplay between structure and electronic properties associated with highly mismatched semiconductor heterostructures, especially with those related to the planar growth of het- erostructured nanowire networks. 1 Catalan Institute of Nanoscience and Nanotechnology (ICN2), CSIC and BIST, Campus UAB, Bellaterra, 08193 Barcelona, Catalonia, Spain. 2 Department of Molecular Chemistry and Materials Science, Weizmann Institute of Science, Rehovot 76100, Israel. 3 ICFO-Institut de Ciencies Fotoniques, The Barcelona Institute of Science and Technology, 08860Castelldefels, Barcelona, Spain. 4 SuperSTEM Laboratory, STFC Daresbury Campus, Daresbury WA4 4AD, UK. 5 School of Chemical and Process Engineering & School of Physics and Astronomy, University of Leeds, Leeds LS2 9JT, UK. 6 ICREA, Passeig Lluís Companys 23, 08010 Barcelona, Catalonia, Spain. 7These authors contributed equally: Sara Martí-Sánchez, Marc Botifoll. ✉email: arbiol@icrea.cat NATURE COMMUNICATIONS | (2022)13:4089 | https://doi.org/10.1038/s41467-022-31778-3 | www.nature.com/naturecommunications 1 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-022-31778-3 O ver the last three decades, heterostructured semiconductor resulting from the increase in the interaction area between the nanowires (NWs)1 have been in the spotlight because of interfaced materials20,28. Additionally, along these large interfaces, their potential as nanoengineered building blocks for a new complex strain relief mechanisms, in which both plastic and elastic generation of electronic2–5, photonic6 and plasmonic7,8, devices for deformations occur simultaneously, can take place at the quantum9–12, sensing13,14, and energy-harvesting15–19 applications. nanoscale20,21. The arguments above illustrate that an in-depth The nanometer-scale lateral dimensions and resulting quasi one- examination of the epitaxially induced strain fields at the atomic- dimensional morphology of the NWs facilitate the release of the scale in these systems is essential to understand and gain control inherent epitaxial strain, preserving high crystal quality even when over their (opto)electronic properties. interfacing highly mismatched materials20. In NWs, plastic and elastic relaxation mechanisms can combine to release strain during Results and discussion axial growth while maintaining a monocrystalline material Materials description: structure, composition, and strain ana- heterostructure20,21, thus introducing a broad variety of material lyses. In this work, we study the local crystal structure and strain combinations in NWs that were not feasible at bulk scales, and relaxation at the atomic level in horizontal arrays of ZnSe@ZnTe further opening up brand new possibilities for electronic band core-shell nanowires grown on α-Al2O3 (sapphire) substrates by engineering and control over the desired properties22–28. In this guided growth, as sketched in Fig. 1a. ZnSe and ZnTe are both context, free-standing vertical nanostructures, such as hetero- direct band-gap semiconductors with room temperature band- structured NWs, have been used for three decades as design model gap values of 2.67 eV47 and 2.26 eV48,49, respectively, which lie in systems to study basic physical and quantum phenomena at the the visible range50,51. These materials form radial type-II p-n nanoscale, offering a high degree of detail and precision9–11,15,23. heterojunctions when interfaced, which are highly advantageous Nonetheless, difficulties arise when trying to interconnect for photovoltaic applications52. The reason for such core-shell different hybrid NWs to form circuits or networks. Most NW combination arises from the fact that ZnTe can be deposited on systems to date are grown vertically on planar substrates via the ZnSe at a lower temperature, making this combination feasible vapor-liquid-solid (VLS) growth mechanism, using a metal liquid given the growth conditions of both materials46. catalyst seed29–33. Positioning and contacting the grown NWs in The sapphire substrates were cut in two different orientations circuits/networks is challenging because the NW must often be (A-plane and C-plane, i.e., α-Al2O3 (1120) and α-Al2O3 (0001), cut and arranged horizontally, and subsequently contacted to respectively) in order to evaluate different growth directions and form a circuit. Several attempts have been made to interconnect plane interactions in this study. n-type ZnSe NWs were grown45 VLS-grown NWs after or even during the growth process34,35, first and then employed as an epitaxial template (acting as cores) but reproducibility and scalability to industry standards remain for p-type ZnTe shell growth44. The arrays of parallel core-shell elusive. NWs follow six equivalent <1100> directions in the case of To overcome the scalability issues of VLS while maintaining C-sapphire substrates, whereas they follow the two <1100> plus perfect epitaxy and gaining in reproducibility and robustness, four <1101> directions in the case of A-sapphire. Details on the semiconductor NW growth has undergone a dramatic evolution growth directions are offered in Supplementary Fig. 1. Our in recent years. The development of new planar growth meth- scanning transmission electron microscopy (STEM) analyses odologies, such as guided growth (GG)36,37 and template-assisted obtained on NW cross-sectional focus ion beam (FIB) lamellae selective epitaxy (TASE)38,39, as well as the more recent appli- show that substrate orientation induces a variation in core cation of selected area growth (SAG)40,41, allow for the direct morphologies. Moreover, we find that atomic steps morphologies growth of horizontal nanowires on top of a selected substrate with are the main cause for strain field modifications in the shell high crystal quality together with the possibility of obtaining crystal. To reveal, in a sub-nanometer scale, the influence of the seamless intersections, consequently enabling the design of core morphology on the electronic band alignment of the 1D complex circuits and networks. These new planar growth stra- core-shell heterostructure, we perform a detailed scanning tegies are therefore suitable for the large-scale fabrication of tai- transmission electron microscopy (STEM) study. Furthermore, lored networks. In particular, guided growth exploits the epitaxial we have developed an iterative methodology based on electron or graphoepitaxial relationship of the material with the substrate energy-loss spectroscopy (EELS) and used it to map the band-gap in order to guide the VLS growth of nanowires along specific evolution in the complex core-shell structure with high spatial substrate crystal directions or relief features, respectively, hence accuracy (including a sub-nanometer pixel size map). The creating self-assembled arrays of well aligned nanowires. During obtained results highlight the nanoscale electronic band-gap the last decade, the guided growth of several types of II-VI NWs modulation due to the high strain at the core and shell interface (such as ZnO42, CdSe43, ZnTe44, or ZnSe45) was achieved, region, including a switch from direct to indirect band gap in the eventually leading to the successful assembly of guided core-shell ZnTe interface region. A similar strain-induced switch was heterostructure arrays involving these materials46. reported in wurtzite GaAs53, however, here we report that such This paradigm based on horizontally grown NWs demands a strain-induced effect is observed with sub-nanometer resolution closer look at the strong interactions between the substrate and precision in a confined region of a heterostructure. With the different materials along the whole length of the grown nanos- present study, we show the importance of detailed atomic-scale tructures. The high strain induced by the lattice mismatch at the crystal evaluation for controlling and tuning optoelectronic interfaces of these horizontal NWs is not as easily released as in properties in horizontal 1-D arrays, which is of great relevance vertical ones20. The resulting strain profiles can substantially in the fields related to electronics, photonics, quantum applica- influence the interface properties of these materials and thus also tions, and photovoltaics among others. their final emerging electronic, photonic or quantum performance. Therefore, to obtain designated material properties, it is important to monitor in detail the structure of these interfaces and correlate Crystalline structure. Most II-IV binary compounds crystallize in them to the local changes of physical properties. two main polytypes: cubic Zinc Blende (ZB) and hexagonal In particular, crystal lattice deformations caused by strain have Wurtzite (WZ)54,55, which differ in their respective stacking been exploited for tuning the electronic band alignment in such sequences when projected through the <110>ZB or <1120> WZ zone material heterostructures27,28. These effects are accentuated in axes but present an identical projected appearance when observed radial core-shell nanostructures compared to axial heterostructures through the <112>ZB or <1100> WZ direction. For this reason, we 2 NATURE COMMUNICATIONS | (2022)13:4089 | https://doi.org/10.1038/s41467-022-31778-3 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-022-31778-3 ARTICLE a d Zn X (3 ) 111 ( ) ( ) 2 nm [1 1] ZnX (ZB) e b Zn Se Te 100 nm Zn X 111 c 1 020 C-plane A- plane 50 nm 20 nm 2 nm [ 01] ZnX (ZB) Fig. 1 Structure, composition and crystallographic orientations of sapphire-supported core-shell NWs. a Schematics of NW configuration, ZnX crystal directions (X = Se/Te) and core-shell band alignment. b EELS-based elemental maps for Zn (blue), Te (green) and Se (red). c Low-magnification HAADF- STEM micrograph of NW transversal cross-sections on A-plane and C-plane sapphire. d Atomic-resolution HAADF-STEM detail and its corresponding power spectrum (FFT) of the transversal cross-section shown in (c). The inset reveals the atomic arrangement in more detail for precise chemical identification through the atomic-number contrast (Z-contrast). e Low-magnification HAADF STEM micrograph of a longitudinal cut on a NW grown on C-plane sapphire, and higher magnification detail (inset) with its corresponding FFT of the area highlighted with a blue square. used Aberration Corrected High Angle Annular Dark Field (AC- In this regard, Geometric Phase Analysis (GPA) has been applied HAADF) STEM imaging in two perpendicular directions of the as an analytical technique to obtain atomic scale information by studied NWs for determining and evaluating the crystal structure examining the dilatation and rotation of crystal planes, allowing for and quality of the core-shell structures grown on two different high spatial resolution in strain analyses59,60. While the vapor-solid substrates (A- and C-plane sapphire). FIB lamellae were obtained (VS) growth of the shell leads to a fixed geometry formed by two by longitudinally cutting a single NW or transversally cutting lateral {202} non-polar planes, a large top (111)A plane and two several NWs of the array. We evaluated the elemental distribution minor {113}A planes, different core morphologies for each orien- by acquiring EELS-based composition maps for Zn, Te, and Se tation can be observed in the case of the VLS growth of the core (Fig. 1b). These maps reveal a sharp interface between core and (Figs. 1a–c and 2a, b). The analyses presented in Fig. 2 were per- shell with no anionic diffusion at the nanometer scale. To ascertain formed with atomic resolution micrographs obtained from trans- the crystal structure and orientation, an atomic resolution AC- versal cross-sections of VLS-grown samples on both A- and HAADF STEM micrograph was obtained in the squared region C-plane sapphire. On C-plane sapphire, core morphologies tend to (Fig. 1d), allowing us to identify the [121] crystal direction as the present highly faceted morphologies that can range from planar growth direction. From an additional longitudinal cross-section of belts (Fig. 2a) with a predominant (111) top facet to faceted the NW shown in Fig. 1e, the cubic ZB phase imaged through the structures with the development of lateral {202} facets and some [101] zone axis can be unequivocally identified. We determine that minor {113} corners, resembling the faceting of the ZnTe shell the overall growth direction of the NW is a < 112> direction (for (Supplementary Fig. 2). In contrast, on A-plane sapphire, core both A- and C-plane substrates) while the growth front corre- faceting is barely present and core morphologies become either sponds to a < 111> direction. This second structural projection also truncated cylinders in the most extreme case (Fig. 2b) or elliptic shows the lack of stacking defects along the length of the nanowire, cylinders presenting different curvatures and some residual top which we find for both substrates used in this study. facets. These core morphologies with rounded cross-section pre- With closer inspection of the high magnification micrographs sent a clear suppression of the lateral {202} facets and contact (Fig. 1d, e), the Z-contrast arising in HAADF STEM imaging angles with the substrate are higher than 90° (Supplementary allows for atomic column identification. The nanowires grow Fig. 2). We attribute the formation of different morphologies to the following a B-polar (anion-polar) <111> direction, which is the different surface energies of C and A sapphire planes, which modify typical growth direction observed for most III-V and II-VI the contact angle of the catalyst and thus influence the final mor- semiconductor binary compound materials24,26. phology of the deposited material. Fast Fourier Transforms (FFTs) of the HAADF STEM images Strain analyses. ZnTe and ZnSe present lattice constants of show a clear difference in ZnTe crystal planes for NWs on A-plane 6.09 Å56 and 5.67 Å57, respectively, which correspond to a lattice and C-plane sapphire. In every crystalline reflection we expect to mismatch of 7.4%. This high mismatch creates strain fields that can see two distinct spots, arising from an epitaxial core-shell structure produce misfit dislocations at the heterostructure interface due to presenting the same crystal orientation but with different lattice the high accumulation of elastic energy. As mentioned above, strain constants that are indicative of the different materials, as clearly fields and dislocations are critical since they play a role in the (opto) observed in Fig. 2c. The inner spots correspond to the ZnTe lattice electronic behavior of semiconductor materials58. Furthermore, and the outer spots to the ZnSe lattice. However, in A-plane complex geometries in highly mismatched planar configurations sapphire we found a triple spot where the outer spots correspond to are responsible for the formation of intricate lattice deformations. the ZnSe core and the inner ZnTe shell spots split into two NATURE COMMUNICATIONS | (2022)13:4089 | https://doi.org/10.1038/s41467-022-31778-3 | www.nature.com/naturecommunications 3 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-022-31778-3 a b The four A-plane sapphire NW shell sections differ in the plane rotation but retain the same interplanar distance, matching that of relaxed ZnTe (Supplementary Fig. 4). More information can be ascertained on this anomalous rotation phenomenon when analyzing the rotation maps obtained by GPA. In the rotation maps of (111) planes we observe that part of the strain is released 50 nm C-sapphire 20 nm A-sapphire by the creation of misfit dislocations, which are especially well visible in Fig. 2h. Atomic resolution details of the misfit dislocations c d in both NWs are shown in Supplementary Figs. 11 and 12. (111) ( ) (111) ( ) However, we also find elastic deformation of the shell in the form of ( ) a plane rotation variation on both sides of the NW which merge ( ) forming a sharp ±2° bent boundary. In the other nanowires grown in A-plane sapphire, which do not show a perfectly truncated circular cross-section, some bending is also found (Supplementary [1 1] ZnX [1 1] ZnX Fig. 3). However, the maximum rotation achieved there lies in the range of ±1–1.5°, and instead of featuring sharp boundaries, the e f change in rotation is gradual. This phenomenon is not observed in the flat or faceted nanowires grown on C-plane sapphire, where we could not measure any plane rotation, neither in the flat nor in the partially faceted NWs, as summarized in Fig. 2i and detailed in Supplementary Figs. 3e–h and 5. The nature of strain relaxation in the core-shell heterojunction has been determined through the experimental measurement of plane dilatations in two orthogonal directions, [111] and [202], to g R (111) h R (111) obtain the strain tensor components (εij) with respect to the core, in which the relaxed ZnSe lattice constant is measured as reference. The high Z2 average difference between ZnSe and ZnTe in comparison to sapphire (Al2O3) saturates the contrast and makes sapphire invisible in HAADF STEM images, the technique of choice for our GPA analyses as it provides low environmental noise. Therefore, we could not simultaneously image both materials in +10 -10 +10 -10 HAADF STEM. However, by using HRTEM images, we measured Rotation [º] Rotation [º] the lattice constants of the core in comparison to the substrate in i several NWs, and we found that the ZnSe core lattice parameters 2 A-Circular matched with its relaxed value, so we could take this value as a C-Planar A-Intermediate reference in our GPA maps and measure the distortions caused in R(111) (º) C-Intermediate the shell with respect to the core reference. Details on these 0 calculations can be found in Supplementary Note 3. The overall dilatation of ZnTe planes with respect to the ZnSe lattice are compatible with a relaxed shell within the level of experimental -2 accuracy. This result, in addition to the misfit dislocation array revealed in the rotational maps and the direct atomic resolution 0 10 20 30 40 50 60 HAADF STEM images (Fig. 2g, h and in more detail in Position (nm) Supplementary Figs. 11 and 12), show that the core-shell heterostructure releases part of the strain plastically by the creation Fig. 2 Study of the shell rotation. Core-shell ZnSe@ZnTe NWs on of misfit dislocations and then achieves relaxation by releasing the C-plane sapphire (left, a, c, e, g) and A-plane sapphire NWs (right, b, d, f, h). remaining strain by compressing the ZnTe lattice in the 3–7 nm a, b Low-magnification atomically resolved HAADF STEM image showing a region closer to the interface (Supplementary Figs. 4 and 5). The representative NW cross-section. c, d Plane-indexed FFTs corresponding same behavior is observed for the NWs grown on both A- and to (a) and (b), highlighting ZnSe core planes (red) and ZnTe shell planes C-plane sapphire substrates, so they only differ in the appearance of (green/blue to differentiate orientations) in these structures. e, f Frequency- the different sections in the NW formed by plane bending. filtered RGB structural maps (IFFT) with color codes as indicated in (c) and We now elucidate the origin of this plane rotation. Considering a (d). The insets schematically show dumbbell orientations in the central vertical interface between core and shell, a plane rotation β in the part of the ZnTe shell. g, h GPA (111) plane rotation (R) maps relative to the shell planes modifies the effective interplanar distance of the shell NW core planes, which are taken as reference. Color bars represent the eff geometrical phase analysis calculated (111) planes rotation in degrees (°). planes in contact with the core planes as dshell = dshell =cosβ > d shell Black arrows in (h) indicate dislocation positions. i (111) plane rotation (Supplementary Fig. 6). This leads to tilted epitaxy when profiles across the center of the NW shell (following the directions indicated dcore > dshell since the effective mismatch between the two structures by the red and blue arrows in (g) and (h), respectively) for relevant core is reduced. The result is a more stable configuration in terms of morphologies in both C-plane and A-plane substrates. elastic energy61. Nonetheless, when dcore < dshell, the tilting of shell planes would instead increase the effective mismatch, and thus, the elastic energy of the system. Therefore, the origin of plane bending reflections that differ by 4° (Fig. 2d). Filtering each distinct set of is related to the specific morphology of the core. To explore this planes (Fig. 2e, f) reveals the existence of four well-defined hypothesis, atomic models reproducing the A-plane sapphire NWs structural sections in A-plane sapphire NW shells, while a simple were created, their corresponding HAADF STEM images were monocrystalline shell structure is observed for C-plane sapphire. simulated62,63, and GPA rotation maps were obtained. Exploiting 4 NATURE COMMUNICATIONS | (2022)13:4089 | https://doi.org/10.1038/s41467-022-31778-3 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-022-31778-3 ARTICLE Model HAADF D(111) R(111) Simulated (0º) 10 nm Simulated (2º) 10 nm Experimental 10 nm -15 +15 -10 +10 Dilatation [%] Rotation [º] Fig. 3 Study of the shell relaxation mechanism. (Two-upper-rows) (Model column) Atomic modeling of a NW top-right section reproducing the experimental morphology (ZnTe presenting a 2° rotation around the1–21 axis), along with a hypothetical configuration with ZnTe planes completely parallel to ZnSe ones (0°). (HAADF Column) HAADF-STEM micrograph simulations of both configurations along with GPA dilatation (D) and rotational (R) maps for the (111) structure planes of the simulated models in comparison with the equivalent experimental analysis in the same NW section (bottom-row). The left color bar represents the geometrical phase analysis calculated (111) planes percentage dilatation (%), while the right color bar represents the geometrical phase analysis calculated (111) planes rotation in degrees (°). the symmetry of the structure, only the top-right section of the Band-gap measurements through Valence EELS. Low-loss nanowire was modeled since the results can be extrapolated to the electron energy-loss spectra contain information about the other half of the NW. Different shell rotations (from 0° to 5° with allowed optical and electronic transitions of the material, steps of 1°) were modeled (see Supplementary Figs. 7–9). including collective excitations such as phonons and plasmons, as We found that the core and shell structures present the best well as single-electron transitions64–66. In the low-energy region matching for a 2° rotation, resulting in a minimization of the of the loss spectra, valence EELS (VEELS, 0–10 eV) reflects the elastic energy accumulation at the interface, in good agreement joint density of states (JDOS), whose dependence on the energy with experimental results (Fig. 3). This is reflected by the lower losses is (E-Eg)1/2 in a direct band-gap material, and (E-Eg)3/2 in number of misfit dislocations at the core-shell interface, as well as an indirect band-gap material67, where E is the energy lost by an higher matching with planes that are tangential to the interface. electron inelastically scattered by the specimen, and Eg is the local More details on the strain relaxation mechanisms together with a band-gap energy at the electron-sample interaction site. Conse- shell rotation study can be found in Supplementary Note 5. quently, VEELS is a useful technique for band-gap evaluation Due to the polar nature of the bonds in II-VI and III-V with the advantage of offering high spatial resolution. However, semiconductors, the dumbbell orientation plays a key role in the complexities of the required data analysis, in addition to the determining the materials’ properties25. In fact, the presence of presence of parasitic effects such as Cherenkov radiation losses, polar defects such as anti-phase boundaries (APBs) or inversion delocalization effects and signatures from excitons and surface- domains has been reported as detrimental to the electronic state excitations, limit the applicability of this technique for band- behavior of binary semiconductors, since they create local electric gap determination. These limitations are especially dramatic in fields and potential barriers that critically decrease their semiconductor materials that present a strong Cherenkov con- conductance24,26. In the pristine ZnTe structure, the vertical tribution, often occurring in those with high refractive <111> direction exhibits a polarity due to the dumbbell (dipole) indices48,49. As described earlier, the band gap of radial p-n orientation while the horizontal <202> direction remains non- junctions is of special interest because it determines the main polar. However, the presence of plane bending around the growth emission lines. Particularly, ZnTe and ZnSe present band gaps axis direction in A-plane NWs (Fig. 2f and corresponding inset) lying in the visible range (2.26 eV47,68 and 2.67 eV69, respectively, creates a configuration similar to APBs or para-twins as, at the both at room temperature). These band-gap energies can be center of the NW shell, the Zn extremes of the dipoles are facing susceptible to modulations due to the atomic arrangements and each other in the low-angle boundary, breaking the horizontal strain distribution at the core/shell interfaces. We, therefore, polar neutrality across the NW shell vertical axis and leading to evaluated the band-gap properties of the ZnSe@ZnTe system the possible creation of an electrostatic potential barrier at the despite the experimental difficulties. Moreover, in spite of pos- boundary, as suggested elsewhere24,26. Therefore, controlling core sible additional contributions such as surface modes, it was morphologies is essential to optimize the shell epitaxy and avoid relatively straightforward to determine a gap value of 8.8 eV for morphology induced low angle polar boundaries. the sapphire substrate70, providing a strong incentive to establish NATURE COMMUNICATIONS | (2022)13:4089 | https://doi.org/10.1038/s41467-022-31778-3 | www.nature.com/naturecommunications 5 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-022-31778-3 a methodology to do the same with the lower gap materials that reach atomic resolution, it allows us to convolute the atomistic constitute the NW. effect of dislocations. Quantum confinement effects are safely In order to overcome these limitations, one can perform ignored because of the large length scale (~100 nm) involved in the VEELS simulations to compute the spatial and energetic nanowire structure. Our simulations assume perfect crystals, dependence of the parasitic signal and compare it to the total although complementary simulations demonstrate the robustness signal, in which the local DOS information lies71,72. With this of the method with respect to strain and a band-gap change smaller approach, we can subtract the parasitic contributions in the than ±1 eV (Supplementary Notes 7 and 8). experimental EELS data and perform meaningful fittings to Based on the information provided by the simulated retarded and extract physically representative information. The details on the non-retarded spectral images, we correct the experimental data to applied methodology are described next, while further details can subtract parasitic contributions associated with Cherenkov radia- be found in Supplementary Note 7. tion and excitation of surface modes. By geometrically correlating We employed several approaches in order to distinguish the the simulated device shape with the experimental one, a correction different regions of interest across the spatially resolved EELS function is pixelwise applied to the experimental spectral image to data, or spectral image, and ensure an adaptative response of the only retain the experimental loss probability corresponding to the fitting routine. For the methodology developed here, we used the bulk excitations of interest. The correction function, which depends main plasmon peaks of both ZnTe (centered at 15.3 eV) and ZnSe both on position (pixel) and on energy (spectrum channel), is the (centered at 16.9 eV), as well as the one of Al2O3 (centered at ratio between the non-retarded bulk loss (ΓNR Bulk ) and the total 25.4 eV) to map the local distribution of these materials, and ΓNR threshold them to ensure they coincide with the domains retarded loss (δðr0; ωÞ ¼ Γ Bulk , where ΓTotal ¼ ΓRBulk þ ΓSurf ). Each of Total observed in the spectral image, avoiding the effects of plasmon the terms involved in the correction factor depends linearly on the delocalization. Unsupervised machine learning has been used by lamella thickness, making the process robust against thickness means of Multivariate Statistical Analysis (MSA)73,74, to obtain modulations or inhomogeneities. In our case, the lamella thickness component spatial maps of the interface pixels between ZnTe and was estimated to be 45 nm thick and homogenous. The entire ZnSe. The spatial distribution of the components is in agreement workflow of data cleaning and processing assisted by the with the medium angle ADF (MAADF) STEM image presented simulations is described in Fig. 4. The detailed algorithmics behind in Fig. 5b, which highlights the regions containing strain these corrections are discussed in Supplementary Notes 7 and 8). accumulation (e.g., near interfaces and volumetric defects)75,76. The so-cleaned experimental data (free of interfering Cher- The presented results rely on the ability of the fitting routine to enkov and surface-related signals) are ready to perform a capture the spatial distribution of both these components and the meaningful analysis. We proceed by implementing adaptive and plasmonic signatures of every material. These routines show the material-dependent curve fittings in a pixelwise fashion to potential of extending this idea further, for instance, when a determine the band-gap structure (direct or indirect) that better material composition modulation is expected (e.g., elemental represents the nature of the observed spectra67. Our method is diffusion or alloying). In this latter case, simultaneous core-loss not limited to solely determining the band-gap value; it can also EELS and VEELS (in a dual-EELS system) or simultaneous EDX disclose its nature by discriminating between direct and indirect and VEELS could be useful to correlate the exact composition band-gap materials in complex heterostructures, which is (obtained by core-loss EELS or EDX chemical maps) of the achieved by minimizing a reliability factor r2 when fitting a material in every pixel with the electronic structure we are model function to the experimentally measured gap onset (see interested in (collecting the inelastically scattered electrons, i.e., more details in Supplementary Notes 7 and 8, and particularly in VEELS). Therefore, a precise material identification based on Supplementary Figs. 13–18). This process allows us to extract a core-loss compositional quantification could be done, allowing physically reliable band-gap map (Fig. 5c), which we have split for more accurate VEELS simulations to eliminate parasitic into two independent maps that highlight those pixels that contributions in the experimental EELS signal. In addition, represent either a higher probability of fitting a direct band machine and deep learning could improve the methodology with structure (Fig. 5d), or an indirect band structure (Fig. 5e). routines for automatic image segmentation77,78. The band-gap values remain close to the ones expected for the The ability to spatially identify the pixels corresponding to each relaxed materials (i.e., the following average band-gap values from of the materials of interest allows the subsequent tailored-data the full extension of each material and region with the corresponding processing assisted by STEM-VEELS simulations of the device. standard error of the mean: “bulk” ZnTe pixels = 2.3162 ± 0.0012 These simulations are based on classical electrodynamics, with the eV; ‘bulk’ ZnSe = 2.7733 ± 0.0012 eV; α-Al2O3 = 8.271 ± 0.016 eV). materials described by their frequency-dependent, local dielectric However, we observe a significant band-gap energy reduction functions. The energy-loss probability is directly separated as the (statistical significance tests are provided in Supplementary Note 6) sum of a bulk contribution, which is independent of geometry and in regions where strain is accumulated, as inferred from the proportional to the path length traveled inside each material, and a MAADF STEM images, with average band-gap values from ZnTe’s surface term determined by the interface morphology70. We interface pixels: 2.28 ± 0.01 eV; and from ZnSe interface pixels: perform both retarded and non-retarded (i.e., assuming an infinite 2.755 ± 0.011 eV). The strained regions in the MAADF STEM image speed of light) calculations and compare them to assess the role of matching with the locations where the band gap shows lower energy Cherenkov radiation emitted by the fast electrons in their values indicates the modulation of the heterojunction at the active interaction with the bulk materials. Given the translational area, the interface between both materials (Fig. 5f). The sharpness of invariance of our system along the electron-beam direction, the the transition of the type II recombination expected from the p-n bulk contribution (ΓRBulk ) is the same as for an electron moving in an junction is justified by the high-quality epitaxy and sharp interface infinitely extended material, while the interface corrections (ΓSurf ) and the consequent possibility to adapt the fitting whether the pixels are determined by solving Maxwell’s equations with the boundary- correspond to ZnTe or ZnSe, indistinctively. element method71,72,79. We find converged results when the In the unstrained ZnTe and ZnSe areas, according to our GPA interfaces are parametrized with points separated by 0.3–0.5 nm, analyses and MAADF-STEM images, our VEELS measurements which incidentally corresponds to the spatial resolution in our reveal a direct band-gap type in those ZnTe and ZnSe areas that experimental spectral images. Although this precision does not are shown as relaxed in our GPA analyses and MAADF STEM 6 NATURE COMMUNICATIONS | (2022)13:4089 | https://doi.org/10.1038/s41467-022-31778-3 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-022-31778-3 ARTICLE Simulated Spectral-Image Corrected experimental curves fitted 180 160 Experimental signal Vacuum Corrected signal Loss probability (Counts) 140 Pt Direct band gap fit 120 100 80 60 40 20 0 Simulated curves 2,0 2,5 3,0 3,5 4,0 Energy (eV) 50 nm 120 Γ Experimental signal 100 Corrected signal δ , Loss probability (Counts) Direct band gap fit Γ 80 60 40 Experimental Spectral-Image 20 0 2,0 2,5 3,0 3,5 4,0 Energy (eV) 300 Experimental signal Corrected signal 250 Indirect band gap fit Loss probability (Counts) 200 150 100 50 0 Γ , , , 2 4 6 8 10 Energy (eV) Fig. 4 Workflow of the developed method for correcting the measured low-loss Electron Energy-Loss Spectroscopy (EELS) data affected by Cherenkov radiation and surface modes. Classical electrodynamics simulations are used to produce a spectral image of both the non-retarded and retarded loss probabilities, which are by construction decomposed into bulk and surface contributions. Next, a correction function that compares the simulated non- retarded and retarded losses is used to map the effect of parasitic radiation and to clean the experimental spectral image. Finally, the corrected experimental signal is used to fit either a direct or indirect band-gap structure depending on the material and spatial position, and to extract a band-gap value from such fit. images, while a clear indirect band-gap behavior is found for the catalyst wetting control affect the core curvature and final cross- relaxed α-Al2O3 substrate (Supplementary Fig. 14)80. Interest- section morphology during horizontally guided NW growth. Core ingly, in the 3 nm-thick ZnTe shell region close to the core curvature has been shown to be the cause of induced plane interface, the fitting obtained with a direct band-gap function rotations in the shell crystal structure. These rotations tune the produces noisy results with both lower goodness of fit and higher overall electronic band structure of the system by a phenomenon error. Thus, we tried the fitting for the VEELS corresponding directly arising from atomic steps edging nanowire cores. pixels in this region with an indirect band-gap function. The A further developed VEELS methodology has allowed us to indirect fitting offers a better performance in terms of meaningful obtain a full system band-gap map with sub-nanometer scale resulting physical information and in goodness of fit, significantly accuracy that offers a clear experimental fingerprint of a band gap surpassing the direct fitting with a confidence level higher than energy decrease together with a direct to indirect band-gap 99.9995% (see “Significance tests” in the Supplementary Informa- transition, induced by the epitaxial strain accumulated at the tion). Examining this area in the GPA calculations and MAADF- 3 nm-thick ZnTe shell region close to the core interface. Our STEM contrast images suggests that a high degree elastic strain study provides accurate and interesting tools for the electronic remains there. The details on this evaluation can be found in characterization of nanostructured semiconductors in the frame- Supplementary Note 8. This difference in fitting performance and work of electron microscopy that are crucial for understanding obtained average band-gap values per region suggests a band type and designing proper complex electronic, photonic and quantum transition from direct to indirect of ZnTe under compressive devices, especially in the recently developed field of horizontally strain (see Fig. 5f and further details in Supplementary Note 8) grown NW networks and circuits. and supports what was predicted theoretically by DFT calcula- tions in the literature81. From this theoretical study, we conclude that the direct interband transition at the Γ point would change Methods from valence Γ to conduction Z (indirect), due to a decrease in the (S)TEM imaging and spectroscopy. The transversal and longitudinal cross sec- tions were prepared by a Focused Ion Beam (FIB) FEI Helios 650 and HAADF conduction band minimum at this Z point. With its high local imaging was performed in a FEI TITAN Low-Base operated at 300 kV. Lamellae accuracy (0.9 nm pixel size) and the tolerance toward Cherenkov were prepared using Ga-ion FIB operated at voltages ranging between 30 kV in the radiation, our measurements constitute, to the best of our first stages to 1 kV in the last steps. This procedure reduced FIB-induced damage knowledge, the first evidence in which high sensitivity band-gap and Ga implantation to a minimum, also resulting in an amorphized layer thick- ness below 1 nm, which had a negligible impact on the imaging and spectroscopy structure characteristics have been reported experimentally. analyses82,83. EELS elemental maps were obtained in a FEI TECNAI F20 TEM The guided growth of core-shell NW horizontal arrays involves microscope operated at 200 kV. VEELS measurements were performed in a strong epitaxial interactions between materials that directly affect monochromated Nion UltraSTEMTM 100MC “HERMES” corrected to the 5th and modulate the electronic behavior of the effective heterojunction. order and operated at 60 kV. The probe convergence semi-angle was 31mrad, with the monochromator slit adjusted to provide an energy resolution of approximately An accurate sub-nanometer scale study of the local strain- 150 meV (as estimated from the full-width at half maximum of the zero-loss peak), induced electronic band modulation has been performed in the offering sufficient resolution for the gap values expected while maintaining high present work. Substrate orientation and an appropriate substrate- probe currents. Given the dispersion used and the detector’s point spread function, NATURE COMMUNICATIONS | (2022)13:4089 | https://doi.org/10.1038/s41467-022-31778-3 | www.nature.com/naturecommunications 7 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-022-31778-3 a HAADF-STEM c 10 9 8 50 nm 7 b MAADF-STEM 6 Eg (eV) 5 50 nm 3 4 d 3 e 1 f ∝ E-Eg I∝ 2 2.8 I∝ E-Eg 2 ZnTe 2.6 3 2.4 2.2 2 ZnSe 2 Eg (eV) Eg 2,2 2,4 2,6 2,8 (eV) 1 Fig. 5 Measured band-gap maps. a HAADF-STEM and b MAADF-STEM micrographs of a ZnSe-ZnTe core-shell nanowire grown on C-plane sapphire to which the band-gap mapping methodology has been applied. c VEELS-band-gap map of the system. The dotted purple square indicates the zoomed-in area displayed in (d), in which the 2–3 eV scale allows for a better qualitative representation of the theoretically predicted decrease in the band-gap at the strained ZnSe-ZnTe interface. This tendency is quantitatively observed in the black curve superimposed in (d), which has been horizontally averaged over the pixels included in the black dotted box and along the black arrow direction in (c). VEELS-band-gap map decomposition into the regions in which we observe either (e) a direct band structure (all ZnSe pixels and ZnTe “bulk” pixels), or (f) an indirect band-gap (ZnTe shell pixels at the core interface, and sapphire pixels). The pixels corresponding to the metallic polycrystalline Pt protective layer have been manually set to 0 eV (metallic behavior). Notice that the color bar represents the band-gap energy in eV. the zero loss peak (ZLP) was Point-spread-function-limited at ~150 meV wide. Received: 12 January 2022; Accepted: 1 July 2022; High-angle and medium angle annular field image detectors’ angular ranges are 95–190 mrad and 55–95 mrad, respectively. EELS was acquired using a Gatan Enfinium ERS spectrometer, with an entrance aperture semi-angle of 44 mrad. The atomic resolution HAADF STEM images shown in Fig. 1d, e have been filtered by using a Wiener filter plus a beam deconvolution in STEM-CELL software62,63. References (S)TEM data analyses. Analyses were carried out using Gatan DigitalMicrograph 1. Arbiol, J. & Xiong, Q. Semiconductor nanowires. materials, synthesis, software. Geometric Phase Analysis is performed using GPA v4.10 plugin within characterization and applications. ISBN: 9781782422532, Woodhead DigitalMicrograph. Analyses were performed on (111) and (20-2) core-shell Publishing (Elsevier), 550 pages (2015). reflections using a cosine mask type. 2. Kastner, M. A. The single-electron transistor. Rev. Mod. Phys. 64, 849–859 (1992). Atomic modeling and HAADF simulation. Atomic models were generated with 3. De Franceschi, S. et al. Single-electron tunneling in InP nanowires. Appl. Phys. Rhodius software from Universidad de Cadiz84,85, considering the experimental Lett. 83, 344–346 (2003). cross-section dimensions and 10 nm thick supercell. HAADF simulations were 4. Mukherjee, S. et al. Phonon engineering in isotopically disordered silicon performed in STEM-CELL software62,63 using a linear approximation by nanowires. Nano Lett. 15, 3885–3893 (2015). employing 300 kV energy, C ap 15 µm, defocus 0 nm. 5. Ketterer, B. et al. Untangling the electronic band structure of wurtzite GaAs nanowires by resonant raman spectroscopy. ACS Nano 5, 7585–7592 (2011). 6. Heiss, M. et al. Self-assembled quantum dots in a nanowire system for VEELS modeling. Boundary-Element Method simulations were performed in quantum photonics. Nat. Mater. 12, 439–444 (2013). order to calculate the contribution of interface terms to EEL spectral-images, using 7. Oulton, R. F. et al. Plasmon lasers at deep subwavelength scale. Nature 461, a two-dimensional implementation that takes advantage of translational invariance 629–632 (2009). in the structures71,72,79. 8. Ketterer, B., Arbiol, J. & Fontcuberta, I. Morral Phonon confinement and plasmon-phonon interaction in nanowire-based quantum wells. Phys. Rev. B 83, 245327 (2011). Data availability 9. Nadj-Perge, S., Frolov, S. M., Bakkers, E. P. A. M. & Kouwenhoven, L. P. All raw and processed data (including images and spectra) that support the findings of Spin–orbit qubit in a semiconductor nanowire. Nature 468, 1084–1087 this study have been deposited in GitHub with the primary accession codes (2010). “mbotifollmoral/ZnTe-ZnSe_Strain_Modulated_BG_Mapping (https://github.com/ 10. Mourik, V. et al. Signatures of majorana fermions in hybrid superconductor- mbotifollmoral/ZnTe-ZnSe_Strain_Modulated_BG_Mapping.git)”. In addition, raw data semiconductor nanowire devices. Science 336, 1003–1007 (2012). were generated at the Catalan Institute of Nanoscience and Nanotechnology (ICN2), the 11. Boukai, A. I. et al. Silicon nanowires as efficient thermoelectric materials. SuperSTEM Laboratory and the LMA-INMA large-scale facilities. Nature 451, 168–171 (2008). 12. de la Mata, M. et al. Twin-induced InSb nanosails: a convenient high mobility quantum system. Nano Lett. 16, 825–833 (2016). 13. Lähnemann, J. et al. UV photosensing characteristics of nanowire based GaN/ Code availability AlN superlattices. Nano Lett. 16, 3260–3267 (2016). The codes and algorithms used to obtain the results of this study have been 14. Neuderth, P. et al. Optical analysis of oxygen self-diffusion in ultrathin CeO2 deposited in GitHub with the primary accession codes “mbotifollmoral/ZnTe- layers at low temperatures. Adv. Energy Mater. 8, 1802120 (2018). ZnSe_Strain_Modulated_BG_Mapping (https://github.com/mbotifollmoral/ZnTe-ZnSe_ 15. Krogstrup, P. et al. Single-nanowire solar cells beyond the Shockley–Queisser Strain_Modulated_BG_Mapping.git)”. limit. Nat. Photonics 7, 306–310 (2013). 8 NATURE COMMUNICATIONS | (2022)13:4089 | https://doi.org/10.1038/s41467-022-31778-3 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-022-31778-3 ARTICLE 16. van Dam, D. et al. High-efficiency nanowire solar cells with omnidirectionally 48. Keller, D. et al. Local band gap measurements by VEELS of thin film solar enhanced absorption due to self-aligned indium–tin–oxide mie scatterers. ACS cells. Microsc. Microanal. 20, 1246–1253 (2014). Nano 10, 11414–11419 (2016). 49. Zamani, R. R. et al. Unraveling electronic band structure of narrow-bandgap 17. Tang, P. et al. Enhanced photoelectrochemical water splitting of hematite p–n nanojunctions in heterostructured nanowires. Phys. Chem. Chem. Phys. multilayer nanowires photoanode with tuning surface state via bottom-up 23, 25019–25023 (2021). interfacial engineering. Energy Environ. Sci. 10, 2124–2136 (2017). 50. Jeon, H. et al. Blue-green injection laser diodes in (Zn,Cd)Se/ZnSe quantum 18. Tang, P. Y. & Arbiol, J. Engineering surface states of hematite based wells. Appl. Phys. Lett. 59, 3619–3621 (1991). photoanodes for boosting photoelectrochemical water splitting. Nanoscale 51. Wu, Q., Litz, M. & Zhang, X. C. Broadband detection capability of ZnTe Horiz. 4, 1256–1276 (2019). electro-optic field detectors. Appl. Phys. Lett. 68, 2924–2926 (1996). 19. Tang, P. Y. et al. Boosting photoelectrochemical water oxidation of hematite 52. Bang, J. et al. ZnTe/ZnSe (Core/Shell) Type-II quantum dots: their optical and in acidic electrolytes by surface state modification. Adv. Energy Mater. 9, photovoltaic properties. Chem. Mater. 22, 233–240 (2010). 1901836 (2019). 53. Signorello, G. et al. Inducing a direct-to-pseudodirect bandgap transition in 20. de la Mata, M., Magen, C., Caroff, P. & Arbiol, J. Atomic scale strain wurtzite GaAs nanowires with uniaxial stress. Nat. Commun. 5, 3655 (2014). relaxation in axial semiconductor III-V nanowire heterostructures. Nano Lett. 54. Arbiol, J. et al. Influence of the (111) twinning on the formation of diamond 14, 6614–6620 (2014). cubic/diamond hexagonal heterostructures in Cu-catalyzed Si nanowires. 21. Conesa-Boj, S. et al. Plastic and elastic strain fields in GaAs/Si core−shell J. Appl. Phys. 104, 064312 (2008). nanowires. Nano Lett. 14, 1859–1864 (2014). 55. Spirkoska, D. et al. Structural and optical properties of high quality zinc- 22. Arbiol, J., de la Mata, M., Eickhoff, M., Fontcuberta, I. & Morral, A. Band gap blende/wurtzite GaAs nanowire heterostructures. Phys. Rev. B 80, 245325 engineering in a nanowire: self-assembled 0, 1 and 2D quantum structures. (2009). Mater. Today 16, 213–219 (2013). 56. Karzel, H. et al. Lattice dynamics and hyperfine interactions in ZnO and ZnSe 23. de la Mata, M. et al. A review of MBE grown 0D, 1D and 2D quantum at high external pressures. Phys. Rev. B 53, 11425 (1996). structures in a nanowire. J. Mater. Chem. C. 1, 4300–4312 (2013). 57. Magalingham, T., John, V. S., Ravi, G. & Sebastian, P. J. Microstructural 24. de la Mata, M. et al. Polarity assignment in ZnTe, GaAs, ZnO and characterization of electrosynthesized ZnTe thin films. Cryst. Res. Technol. 37, GaN-AlN nanowires from direct dumbbell analysis. Nano Lett. 12, 329 (2002). 2579–2586 (2012). 58. Krizek, F. et al. Selective-area-grown semiconductor-superconductor hybrids: 25. Zamani, M. et al. Optimizing the yield of A-polar GaAs nanowires to achieve A basis for topological networks. Phys. Rev. Lett. 121, 147701 (2018). defect-free zinc blende structure and enhanced optical functionality. 59. Hytch, M. J., Snoeck, E. & Kilaas, R. Ultramicroscopy 74, 131–146 (1998). Nanoscale 10, 17080–17091 (2018). 60. Hÿtch, M. J., Putaux, J.-L. & Pénisson, J.-M. Measurement of the displacement 26. de la Mata, M. et al. The role of polarity in nonplanar semiconductor field of dislocations to 0.03 Å by electron microscopy. Nature 23, 270–273 nanostructures. Nano Lett. 19, 3396–3408 (2019). (2003). 27. Signorello, G. et al. Tuning the light emission from GaAs nanowires over 61. Liu, Y. et al. Semiconductor–ferromagnetic insulator–superconductor 290 meV with uniaxial strain. Nano Lett. 13, 917–924 (2013). nanowires: Stray field and exchange field. Nano Lett. 20, 456–462 (2020). 28. Hong, K. H., Kim, J., Lee, S.-H. & Shin, J. K. Strain-Driven Electronic Band 62. Grillo, V. The effect of surface strain relaxation on HAADF imaging. Structure Modulation of Si Nanowires. Nano Lett. 8, 1335–1340 (2008). Ultramicroscopy 109, 1453–1464 (2009). 29. Wagner, R. S. & Ellis, W. C. Vapor-liquid-solid mechanism of single crystal 63. Grillo, V. & Rotunno, E. STEM_CELL: A software tool for electron growth. Appl. Phys. Lett. 4, 89 (1964). microscopy: Part I—simulations. Ultramicroscopy 125, 97–111 (2013). 30. Sivaram, S. V. et al. Surface hydrogen enables subeutectic vapor–liquid–solid 64. Egerton, R. F. Electron Energy-Loss Spectroscopy in the Electron Microscope semiconductor nanowire growth. Nano Lett. 16, 6717–6723 (2016). (Springer US, Boston, MA, ISBN: 978-1-4615-6887-2 (2011). 31. Hui, H. Y., de la Mata, M., Arbiol, J. & Filler, M. A. Low-temperature growth 65. Genç, A. et al. Tuning the plasmonic response up: Hollow cuboid metal of axial Si/Ge nanowire heterostructures enabled by trisilane. Chem. Mater. nanostructures. ACS Photonics 3, 770–779 (2016). 29, 3397–3402 (2017). 66. Hage, F. S., Kepaptsoglou, D. M., Ramasse, Q. M. & Allen, L. J. Phonon 32. Russo-Averchi, E. et al. High yield of GaAs nanowire arrays on Si mediated by spectroscopy at atomic resolution. Phys. Rev. Lett. 122, 016103 (2019). the pinning and contact angle of Ga. Nano Lett. 15, 2869–2874 (2015). 67. Rafferty, B. & Brown, L. M. Direct and indirect transitions in the region of the 33. Zamani, M. et al. 3D ordering at the liquid–solid polar interface of nanowires. band gap using electron-energy-loss spectroscopy. Phys. Rev. B 58, 10326 Adv. Mater. 32, 2001030 (2020). (1998). 34. Storm, K. et al. Spatially resolved hall effect measurement in a single 68. Van Stryland, E. W., Woodall, M. A., Vanherzeele, H. & Soileau, M. J. Energy semiconductor nanowire. Nat. Nanotechnol. 7, 718–722 (2012). band-gap dependence of two-photon absorption. Opt. Lett. 10, 490–492 35. Gazibegovic, S. et al. Epitaxy of advanced nanowire quantum devices. Nature (1985). 548, 434–438 (2017). 69. Morkoç, H. et al. Large‐band‐gap SiC, III‐V nitride, and II‐VI ZnSe‐based 36. Tsivion, D. et al. Guided growth of millimeter-long horizontal nanowires with semiconductor device technologies. J. Appl. Phys. 76, 1363 (1994). controlled orientations. Science 333, 1003–1007 (2011). 70. Schattschneidera, P., Stoger, M., Hebert, C. & Jouffrey, B. The separation of 37. Zi, Y., Jung, K., Zakharov, D. & Yang, C. Understanding self-aligned planar surface and bulk contributions in ELNES spectra. Ultramicroscopy 93, 91–97 growth of InAs nanowires. Nano Lett. 13, 2786–2791 (2013). (2002). 38. Schmid, H. et al. Template-assisted selective epitaxy of III–V nanoscale 71. García de Abajo, F. J. & Howie, A. Retarded field calculation of devices for co-planar heterogeneous integration with Si. Appl. Phys. Lett. 106, electron energy loss in inhomogeneous dielectrics. Phys. Rev. B 65, 115418 233101 (2015). (2002). 39. Knoedler, M. et al. Observation of twin-free GaAs nanowire growth using 72. García de Abajo, F. J. & Howie, A. Relativistic electron energy loss and template-assisted selective epitaxy. Cryst. Growth Des. 17, 6297–6302 (2017). electron-induced photon emission in inhomogeneous dielectrics. Phys. Rev. 40. Friedl, M. et al. Template-assisted scalable nanowire networks. Nano Lett. 18, Lett. 80, 5180 (1998). 2666–2671 (2018). 73. Trebbia, P. & Bonnet, N. EELS elemental mapping with unconventional 41. Vaitiekėnas, S. et al. Selective-area-grown semiconductor- methods I. Theoretical basis: image analysis with multivariate statistics and superconductor hybrids: a basis for topological networks. Phys. Rev. Lett. 121, entropy concepts. Ultramicroscopy 34, 165–178 (1990). 147701 (2018). 74. Bosman, M., Watanabe, M., Alexander, D. T. L. & Keast, V. J. Mapping 42. Tsivion, D., Schvartzman, M., Popovitz-Biro, R. & Joselevich, E. Guided chemical and bonding information using multivariate analysis of electron growth of horizontal ZnO nanowires with controlled orientations on flat and energy-loss spectrum images. Ultramicroscopy 106, 1024–1032 (2006). faceted sapphire surfaces. ACS Nano 6, 6433–6445 (2012). 75. Perovic, D. D., Rossouw, C. J. & Howie, A. Imaging elastic strains in high- 43. Shalev, E. et al. Guided CdSe nanowires parallelly integrated into fast visible- angle annular dark field scanning transmission electron microscopy. range photodetectors. ACS Nano 11, 213–220 (2017). Ultramicroscopy 52, 353–359 (1993). 44. Oksenberg, E., Popovitz-Biro, R., Rechav, K. & Joselevich, E. Guided growth of 76. Phillips, P. J. et al. Atomic-resolution defect contrast in low angle annular horizontal ZnSe nanowires and their integration into high-performance blue- dark-field STEM. Ultramicroscopy 116, 47–55 (2012). UV photodetectors. Adv. Mater. 27, 3999–4005 (2015). 77. Madsen, J. et al. A deep learning approach to identify local structures in 45. Reut, G. et al. Guided growth of horizontal P-Type ZnTe nanowires. J. Phys. atomic-resolution transmission electron microscopy images. Adv. Theory Chem. C. 120, 17087–17100 (2016). Simul. 1, 1800037 (2018). 46. Oksenberg, E. et al. Surface-guided core-shell ZnSe@ZnTe nanowires as radial 78. Ziatdinov, M. et al. Deep learning of atomically resolved scanning p-n heterojunctions with photovoltaic behavior. ACS Nano 11, 6155–6166 transmission electron microscopy images: chemical identification and tracking (2017). local transformations. ACS Nano 11, 12742–12752 (2017). 47. Mahalingam, T., John, V. S., Rajendran, S. & Sebastian, P. J. Electrochemical 79. García de Abajo, F. J. Optical excitations in electron microscopy. Rev. Mod. deposition of ZnTe thin films. Semicond. Sci. Technol. 17, 465–470 (2002). Phys. 82, 209–275 (2010). NATURE COMMUNICATIONS | (2022)13:4089 | https://doi.org/10.1038/s41467-022-31778-3 | www.nature.com/naturecommunications 9 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-022-31778-3 80. French, R. Electronic band structure of Al2O3 with comparison to alon and Author contributions AlN. J. Am. Ceram. Soc. 13, 471–489 (1990). S.M.-S., M.B., and J.A. conceived and initiated the project. J.A. supervised the project and 81. Peköz, R. & Raty, J. Y. Band structure modulation of ZnSe/ZnTe nanowires led the collaboration efforts. S.M.-S., M.B., and J.A. designed the experiments. E.O. and under strain. Phys. Rev. B. 84, 165444 (2011). E.J. took care of the samples growth. S.M.-S., M.B., C.K., C.B., M.C.S., Q.R., and J.A. 82. McCaffrey, J. P., Phaneuf, M. W. & Madsen, L. D. Surface damage formation obtained and analyzed the TEM, STEM, cross-sectional STEM measurements and during ion-beam thinning of samples for transmission electron microscopy. EELS spectra. M.B., V.D.G., and F.J.G.A performed the theoretical analyses and models. Ultramicroscopy 87, 97–104 (2001). S.M.-S., M.B., and J.A. wrote the paper. All authors discussed the results and commented 83. Mayer, J., Giannuzzi, L. A., Kamino, T. & Michael, J. TEM sample preparation on the manuscript. and FIB-induced damage. MRS Bull. 32, 400–407 (2007). 84. Bernal, S. et al. The interpretation of HREM images of supported metal catalysts using image simulation: profile view images. Ultramicroscopy 72, Competing interests The authors declare no competing interests. 135–164 (1998). 85. Arbiol, J. et al. Optimization of tin oxide nanosticks faceting for the improvement of palladium nanoclusters epitaxy. Appl. Phys. Lett. 80, 329–331 Additional information (2002). Supplementary information The online version contains supplementary material available at https://doi.org/10.1038/s41467-022-31778-3. Acknowledgements Correspondence and requests for materials should be addressed to Jordi Arbiol. J.A., S.M.-S., M.B., and M.C.S. were supported by the Severo Ochoa program from Spanish MCIN/AEI /10.13039/501100011033/ (Grant No. SEV-2017-0706) and are Peer review information Nature Communications thanks Slawomir Kret and the other, funded by the CERCA Program / Generalitat de Catalunya. Part of the present work has anonymous, reviewer(s) for their contribution to the peer review of this work. Peer been performed in the framework of Universitat Autònoma de Barcelona Materials reviewer reports are available. Science Ph.D. program. J.A., S.M.-S., M.B., and M.C.S. acknowledge funding from Generalitat de Catalunya 2017SGR327. S.M.S. acknowledges funding from ICN2 SO Reprints and permission information is available at http://www.nature.com/reprints EXEM project. The HAADF-STEM microscopy was conducted in the Laboratorio de Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in Microscopias Avanzadas at Instituto de Nanociencia de Aragon-Universidad de Zar- published maps and institutional affiliations. agoza. Additional imaging and electron energy loss spectroscopy was carried out at SuperSTEM, the UK National Research Facility for Advanced Electron Microscopy, supported by the Engineering and Physical Sciences Research Council (EPSRC). Authors acknowledge the LMA-INMA for offering access to their instruments and expertise. We Open Access This article is licensed under a Creative Commons acknowledge support from CSIC Research Platform on Quantum Technologies PTI-001 Attribution 4.0 International License, which permits use, sharing, (QTEP). M.B. acknowledges support from SUR Generalitat de Catalunya and the EU adaptation, distribution and reproduction in any medium or format, as long as you give Social Fund; project ref. 2020 FI 00103. C.K. acknowledges funding from InPhINIT appropriate credit to the original author(s) and the source, provide a link to the Creative Scholarships LaCaixa Foundation. M.C.S. have received funding from the European Commons license, and indicate if changes were made. The images or other third party Union’s Horizon 2020 research and innovation program under the Marie Sklodowska- material in this article are included in the article’s Creative Commons license, unless Curie grant agreement No 754510 (PROBIST). J.G.d.A. has been supported by European indicated otherwise in a credit line to the material. If material is not included in the Research Council (Advanced Grant 789104-eNANO). V.D.G. and J.G.d.A have received article’s Creative Commons license and your intended use is not permitted by statutory funding from Spanish MCINN (PID2020-112625GB-I00 and CEX2019-000910-S), regulation or exceeds the permitted use, you will need to obtain permission directly from Fundació Cellex, Fundació Mir-Puig, and Generalitat de Catalunya (CERCA, AGAUR). the copyright holder. To view a copy of this license, visit http://creativecommons.org/ E.J. acknowledges support from the Israel Science Foundation (Grant No. 2444/19). J.A., licenses/by/4.0/. S.M.-S., M.B. and M.C.S. were supported by MCIN with funding from European Union NextGenerationEU (PRTR-C17.I1), by Generalitat de Catalunya and by “ERDF A way of making Europe”, by the “European Union”. © The Author(s) 2022 10 NATURE COMMUNICATIONS | (2022)13:4089 | https://doi.org/10.1038/s41467-022-31778-3 | www.nature.com/naturecommunications