Introduction

Hydrothermal synthesis of inorganic nanomaterials1,2,3,4 offers wide range of applications5,6,7,8,9,10,11 due to the scalability and the compatibility with various substrates and material processes12,13,14 originating from its low process temperature which is not available by vapor-phase synthesis15,16,17,18. The functions of inorganic nanomaterials strongly depend on their geometry. Therefore, most of previous researches have been devoted to establish predictable models and strategies for controlling the morphology of nanomaterials. As for the hydrothermally synthesized zinc oxide (ZnO) nanowires, the strategies based on the electrostatic interaction between metal ions and crystals19, the organic ligand based surface capping20, the ligand-exchange effect21, and the difference in critical nucleation concentrations on crystal planes so-called “concentration window”22,23,24 have been demonstrated and successfully manipulated the nanowire morphology.

On the other hand, elemental doping is a typical approach for aggressively tuning the functions of inorganic nanomaterials. The elemental doping to nanowires has been intensively investigated in vapor-phase synthesis25,26. The feasibility of elemental doping for hydrothermal ZnO nanowires has been also demonstrated together with successful modifications of electrical, optical, and surface chemical properties27,28,29,30,31. However, there still remains much room for investigating the elemental doping in hydrothermal nanowire synthesis. For example, the doping often accompanies the variations of anisotropic crystal growth of ZnO nanowires32,33,34,35 and in other case, the metal impurities solely serve as counter ions rather than being incorporated into the nanowires19,36. These results plausibly originate from the variety of chemical reactions and electrostatic effects in the hydrothermal system19,37,38,39, which are not seen in vapor-phase synthesis. Furthermore, the dopant distribution in an individual hydrothermal nanowire has not been discussed so far, while its importance has been highlighted in a vapor–liquid–solid (VLS) nanowire25. As such, the elemental doping in hydrothermal synthesis is far from comprehensive understanding and controlling over the morphology and the dopant distribution of hydrothermal nanowire is a challenging issue.

In the present research, we report the unusual face-selectivity of tungstate ion on growth direction and dopant incorporation of hydrothermal ZnO nanowire. Systematic experiments and simulations reveal that tungstate ions (WO42−) face-selectively act as surfactant and dopant on different crystal planes according to their coordination structure. Furthermore, the dopant distribution in an individual hydrothermal ZnO nanowire is unveiled by means of scanning transmission electron microscopy (STEM) observation with energy dispersive x-ray spectroscopy (EDS) and three-dimensional atom probe (3DAP) analysis. Based on above results, a predictable model and a rational strategy to control over the morphology and the dopant distribution of hydrothermal ZnO nanowires are demonstrated.

Results and discussion

Tungsten is an appropriate dopant species for ZnO in terms of the ionic radius (Zn2+ 60 pm, W6+ 42 pm) and the coordination number (Zn2+(IV) and W6+(IV)). In fact, the improvement of molecular sensing property of ZnO thin film was previously demonstrated by W doping40,41. The nanowire growth was conducted by a seed-assisted hydrothermal approach in aqueous solution containing Zn(NO3)2 (Zn precursor), hexamethylenetetramine (HMTA), polyethyleneimine (PEI), and Na2WO4 (W precursor). A 50-nm-thick ZnO seed layer coated on Si/SiO2 substrate was utilized for the growth. The temperature and the time for the growth were 95 °C and 5 h, respectively. For all experiments, the concentrations of HMTA 25 mM and PEI 1.25 mM were kept constant (the role of PEI is shown in Supplementary Figs. 1, 2 and Supplementary Notes 1, 2). Figure 1a shows side view and top view scanning electron microscopy (SEM) images of ZnO nanowires when varying Na2WO4 concentration (CW) 0–0.25 mM under the constant Zn(NO3)2 concentration (CZn) 25 mM. When increasing CW, the nanowire tends to be a platelet form. The statistic nanowire length and diameter in Fig. 1b show that the diameter exceeds the length above CW 0.25 mM. This indicates that the preferential growth direction of ZnO nanowires changes from [0001] to \(\left[ {10\bar 10} \right]\) just by adding 1% of impurity in the growth solution. Figure 1c shows the low and high magnification transmission electron microscopy (TEM) images and the electron diffraction patterns for both nanowire (CZn 25 mM, CW 0 mM) and nanoplatelet (CZn 25 mM, CW 0.125 mM) structures. Both samples show single crystallinity and hexagonal wurtzite structure, indicating that the crystallographic features of ZnO nanostructure are not varied by the W addition. Figure 1d, e shows the x-ray photoelectron spectroscopy (XPS) spectra (Zn 2p and W 4f) and the estimated W concentration (at%) of ZnO nanostructures synthesized with CZn 25 mM and various CW. When increasing CW, W 4f peaks tend to be larger while Zn 2p peaks remain almost unchanged, leading to the systematic increase of W concentration. Figure 1f, g shows the x-ray diffraction (XRD) patterns and the calculated d values of ZnO nanostructures synthesized with CZn 25 mM and various CW. The ZnO (0004) peak shifts to be higher angle and the d value decreases with increasing CW, which is plausibly due to the substitution of Zn2+ with W6+. The results of XPS and XRD show that W is incorporated into ZnO nanostructures. Alternatively, these results represent that controlling morphology and W doping of hydrothermal ZnO nanowires is quite difficult.

Fig. 1: Morphology, composition, and crystal structure of nanowires with varying CW.
figure 1

a SEM images and b statistic length and diameter data of ZnO nanowires grown with varying CW. CZn is kept to be 25 mM. The error bars are defined by standard deviation values. c TEM images and selected area electron diffraction (SAED) patterns of both nanowire and nanoplatelet grown with CW 0 mM and CW 0.125 mM, respectively. d Zn 2p and W 4f XPS spectra and e W concentration of ZnO nanowires grown with various CW. f XRD patterns and g d value of ZnO nanowires grown with various CW.

Full size image

For exploring a way to overcome above dilemma between the dopant incorporation and the morphological variation, we first consider why the preferential growth direction of nanowires is changed by W addition. In previous studies, the formation of nanoplatelet structure was interpreted in terms of (i) the growth inhibition by the electrostatic adsorption of counter ions19 and/or the chemisorption of inorganic and organic ligands20,36 on ZnO (0001) plane, (ii) the interfacial energy reduction of ZnO (0001) plane via chemical cappings42, and (iii) the dislocation-driven growth by lowering the supersaturation degree43,44,45,46 (For the detailed review, see Supplementary Table 1). In order to specify the possible metal counter ion species, we performed thermodynamic calculation of equilibrium concentration of ionic species under the given growth condition. We found that WO42− ion is the most plausible adsorption species for the growth inhibition (Supplementary Fig. 3 and Supplementary Note 3). To understand the contribution of WO42− ions for the nucleation phenomenon, next we examine the Zn precursor concentration dependence on ZnO nanowire growth with various CW. Previous our study demonstrated that there exist two threshold concentrations corresponding to the critical concentrations for nucleation on (0001) plane and \(\left( {10\bar 10} \right)\) plane22,23,24. By evaluating the shifts of critical concentrations with varying CW, the influence of WO42− ions for the nucleation events on ZnO crystal planes can be identified. Figure 2a, b shows the Zn precursor concentration dependences on length and diameter of ZnO nanowires with CW 0–0.25 mM. The SEM images for the series of CZn are shown in Supplementary Fig. 4. The triangle marks in the graphs represent the critical concentration for nucleation. These results are summarized in Fig. 2c. The critical concentration on \(\left( {10\bar 10} \right)\) plane (i.e., sidewall) tends to be lower while that on (0001) plane (i.e., nanowire top) is almost unchanged with increasing CW. This indicates that WO42− ions rather enhance the nucleation on \(\left( {10\bar 10} \right)\) plane than suppress the nucleation on (0001) plane. Note that the growth inhibition on (0001) plane seems to be negligible at lower CW range, but it becomes significant by increasing CW to 0.25 mM. We found that the nucleation on (0001) plane would be enhanced with decreasing CZn below the critical concentration on \(\left( {10\bar 10} \right)\) plane although the W/Zn ratio in precursor becomes higher. Thus, above results indicate that the variation of preferential growth direction of ZnO nanowires by W addition cannot be explained by the existing models and rather suggest the selective nucleation enhancement on \(\left( {10\bar 10} \right)\) plane.

Fig. 2: Morphology, composition, and crystal structure of nanowires with varying CZn.
figure 2

Zn concentration dependences on a length and b diameter of ZnO nanowires grown with various CW. c Critical Zn concentration for nucleation on (0001) plane and \(\left( {10\bar 10} \right)\) plane as a function of CW. d W 4f XPS spectra and e W concentration of ZnO nanowires when varying CZn. f XRD patterns and g d value of ZnO nanowires when varying CZn.

Full size image

In general, dopant incorporation event is accompanied with crystal growth. Therefore we assume that W incorporation mainly occurs through \(\left( {10\bar 10} \right)\) plane. In fact, the results of XPS analysis in Fig. 2d, e support this model, i.e., the W concentration of ZnO nanowires increases with increasing CZn above the critical concentration for nucleation on \(\left( {10\bar 10} \right)\) plane. However, the following XRD results show more complicated situation for the W incorporation. Figure 2f, g shows the XRD pattern of ZnO (0002) peak and the d value for the W-doped ZnO nanowires when varying CZn with constant CW 0.25 mM. We found that there are two components in (0002) peak and among the one is consistent with bulk ZnO, implying the coexistence of non-doped part and W-doped part in an individual ZnO nanowire. Furthermore, there are two inconsistencies between the results of XPS and XRD. First, the XRD peak associated with W-doped ZnO still remains when completely suppressing the nucleation on \(\left( {10\bar 10} \right)\) plane. Second, in XRD result, W incorporation seems to be promoted by decreasing CZn as can be seen in the decreasing trend of d value for W-doped ZnO in Fig. 2g. These two features are contradictive to the results of XPS. Considering the difference of surface-sensitive XPS and bulk-sensitive XRD, such contradictive features can be interpreted by that ZnO nanowires are composed of W-doped core and non-dope shell. Also this suggests that W incorporation unexpectedly occurs on (0001) plane even though W addition enhances the nucleation on \(\left( {10\bar 10} \right)\) plane.

In order to clarify the dopant incorporation pathway, we examine the feasibility of W incorporation via the crystal growth on \(\left( {10\bar 10} \right)\)-oriented and (0001)-oriented single-crystalline ZnO substrates as shown in Fig. 3a, b. In these experiments, CZn and CW are 25 and 0.25 mM. The W concentration of ZnO structures grown on (0001)-oriented and \(\left( {10\bar 10} \right)\)-oriented substrates are 1.03 and 0.04 at%, respectively, supporting that W is mainly incorporated through (0001) plane. This is further confirmed via the compositional analysis using two-step grown nanowires (see Supplementary Fig. 5 and Supplementary Note 4). According to the face-selective W incorporation on (0001) plane, the result in Fig. 2e can be interpreted in terms of an expansion of (0001) plane size followed by the sidewall growth. Therefore, we perform the TEM observation and analyze the correlation between the (0001) plane size and the W concentration of ZnO nanowires. Figure 3c shows the TEM images of ZnO nanowires grown with various CZn (10–25 mM) and CW (0–0.125 mM). The (0001) plane tends to be larger with increasing CZn and/or CW, which is associated with the nucleation enhancement on \(\left( {10\bar 10} \right)\) plane discussed above. The relationship between (0001) plane width and W concentration is summarized in Fig. 3d. There is a strong correlation between them, and the W concentration is irrelevant to W/Zn ratio in precursors, clearly evidencing that W incorporation selectively occurs on (0001) plane. The TEM observation also reveals the presence of \(\left( {10\bar 11} \right)\) plane at the tip of nanowires. Note that the \(\left( {10\bar 11} \right)\) plane exists even suppressing the sidewall growth. The W concentration becomes almost zero when the \(\left( {10\bar 11} \right)\) plane dominates the nanowire tip with decreasing the diameter. This suggests that W is not incorporated on \(\left( {10\bar 11} \right)\) plane.

Fig. 3: Evaluations of dopant incorporation pathway.
figure 3

a SEM images and b W concentration of ZnO nanowires grown on (0001)-oriented and \(\left( {10\bar 10} \right)\)-oriented single-crystalline ZnO substrates. The nanowires are grown by CZn 25 mM and CW 0.25 mM. c TEM images of ZnO nanowires grown with various CZn and CW. d W concentration of ZnO nanowires as a function of averaged (0001) plane width. The width of (0001) plane is evaluated by TEM images.

Full size image

To confirm the hypothesis as to the face-selective dopant incorporation, we directly identify the dopant distribution in an individual ZnO nanowire. It should be described that the dopant distribution in hydrothermal ZnO nanowire has not been discussed before unlike that in VLS nanowire16,17,18,25. This might be due to that the impinging rate of impurity metal ions on each crystal plane of nanowire is equal in hydrothermal synthesis, but those on liquid–solid interface and vapor–solid interface in VLS synthesis are supposed to be different. In high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) image of Fig. 4a, the clear contrast with darker core and brighter shell is seen in the nanowire. For this experiment, the nanowires were grown with CZn 15 mM and CW 0.125 mM (W concentration in precursor: 0.83%) and we chose a nanowire with sidewall growth. We found that the darker area, which is associated with W-doped part, tends to be larger along the nanowire growth direction. This is well consistent with the model of face-selective W incorporation on (0001) plane discussed above. The qualitative analysis of dopant distribution by EDS shows that the darker area has higher W concentration and in this area dopant seems to be homogeneously distributed as shown and discussed in Supplementary Fig. 6 and Supplementary Note 5. In addition, the dopant distribution in an individual ZnO nanowire is quantitatively analyzed by 3DAP measurement as details can be seen in Supplementary Fig. 7. Figure 4b shows 3D atom maps for Zn, O (left), and W (right) taken around the projected area in Fig. 4a. W atoms are more densely distributed in core side than shell side. This is also confirmed by 2D concentration plot shown at the bottom of 3D atom maps and line profile in Fig. 4c. Note that averaged W concentrations in core area are 0.66 at%, indicating that 80% of WO42− ions impinged on (0001) plane are incorporated into nanowires. On the other hand, W concentration in shell area is limited to 0.11 at%, which corresponds to 13% of WO42− ions impinged on \(\left( {10\bar 10} \right)\) plane. These results highlight the inhomogeneous dopant distribution in an individual hydrothermal ZnO nanowire, resulting from the face-selective dopant incorporation.

Fig. 4: Spatial composition analyses of nanowires.
figure 4

a HAADF-STEM image of ZnO nanowire grown with CZn 15 mM and CW 0.125 mM. b 3D atom maps for Zn, O (left), and W (right) taken around the projected area in a. 2D W concentration plot in z direction is shown at the bottom of 3D W atom map. c Line profile of W concentration along the dotted line in 2D W concetration plot of b. The dotted line is W concentration in precursor.

Full size image

To gain the theoretical insight as to the conflicted face-selective behaviors of WO42− ions on nucleation and dopant incorporation, we performed density functional theory (DFT) calculations. The details of calculation are shown in Supplementary Fig. 8. Figure 5a shows the most stable coordination structure of WO42− ion on ZnO \(\left( {10\bar 10} \right)\) plane, (0001) plane, and \(\left( {10\bar 11} \right)\) plane. WO42− ion is coordinated with tripodal bonds on (0001) and \(\left( {10\bar 11} \right)\) planes, and with unipodal or bipodal bonds on \(\left( {10\bar 10} \right)\) plane. Here we discuss the WO42− ion-induced nucleation enhancement on \(\left( {10\bar 10} \right)\) plane by examining the various coordination of WO42− ion and Zn species (Zn2+, [ZnOH]+, and Zn(OH)2) on \(\left( {10\bar 10} \right)\) plane in the presence of water solvent (Fig. 5b). The calculated interaction energies between WO42− ion and various Zn species are summarized in Table 1. The results demonstrate that the sufficient energy gain is obtained for all the examined coordination structures. The growth system is further stabilized by incorporating Zn species into ZnO crystal which is shown as a relative energy in Fig. 5c, d. Since three-terminated oxygen site of WO42− ions in step 4 of Fig. 5c is not stable by considering the valency of oxygen, WO42− ions are not incorporated into ZnO crystal and form new coordination structure with other Zn species in the solution, highlighting the surfactant role of WO42− ions for enhancing the nucleation. Such surfactant effect of WO42− ion must be more significant on \(\left( {10\bar 10} \right)\) plane than on (0001) plane in terms of the number of unsaturated bonds of WO42− ion for coordinating with Zn ions (i.e., two or three bonds on \(\left( {10\bar 10} \right)\) plane and one bond on (0001) plane). Thus WO42− ions selectively enhance the nucleation on \(\left( {10\bar 10} \right)\) plane as experimentally shown above. Next we consider the face-selective dopant incorporation on (0001) plane, which must be correlated with the stability of coordinated WO42− ions. We calculated the adsorption energies of WO42− ions on ZnO \(\left( {10\bar 10} \right)\) plane, (0001) plane, and \(\left( {10\bar 11} \right)\) plane, as summarized in Table 2. Note that in the simulation, the WO42− ions are not adsorbed on O-terminated (0001) and \(\left( {10\bar 11} \right)\) planes, therefore, we calculated the adsorption energies on Zn-terminated planes. The adsorption energies of WO42− ions on ZnO crystal planes become higher in the following order \(\left( {10\bar 11} \right)\) < \(\left( {10\bar 10} \right)\) < (0001). This indicates that the WO42− ion is preferentially adsorbed on (0001) plane followed by being incorporated into nanowire. Although the stable adsorption of WO42− ions with tripodal bonds is also available on \(\left( {10\bar 11} \right)\) plane, it is destabilized by the hydration effect. Although the hydration effects on (0001) plane and \(\left( {10\bar 11} \right)\) plane are similar, the weaker adsorption of WO42− ions on \(\left( {10\bar 11} \right)\) plane originating from its distorted coordination structure leads to the negative adsorption energy in water. Thus our calculation clearly explain that the coordination structure of WO42− ion on ZnO crystal surface critically determines the morphology and the dopant distribution of hydrothermal ZnO nanowire.

Fig. 5: Computational analyses of Zn ions and WO42− ions on ZnO crystal planes.
figure 5

a The coordination structures of adsorbed WO42− ion on ZnO \(\left( {10\bar 10} \right)\) plane, (0001) plane, and \(\left( {10\bar 11} \right)\) plane, respectively. b The coordination structures of WO42− ion and various Zn species (Zn2+, [ZnOH]+, and Zn(OH)2) on ZnO \(\left( {10\bar 10} \right)\) plane. c The coodination structures and d relative energies of WO42−-Zn2+ ions on ZnO \(\left( {10\bar 10} \right)\) plane during the Zn incorporation process. The coordination structures are simulated by DFT calculation in the presence of water solvent.

Full size image
Table 1 Calculated interaction energy between Zn ion species (Zn2+, [ZnOH]+, Zn(OH)2) and WO42− ions adsorbed on ZnO \(\left( {10\bar 10} \right)\) plane in kcal/mol.
Full size table
Table 2 Calculated adsorption energy of WO42− ion on ZnO \(({10\bar 10} )\) plane, (0001) plane, and \(( {10\bar 11} )\) plane in vacuum and water conditions.
Full size table

On the basis of suggested model, finally we control over the morphology and the elemental doping of ZnO nanowires. To the best of our knowledge, the well-defined synthesis of W-doped hydrothermal ZnO nanowire has not been achieved so far because the W addition seriously causes the nanowire morphology34,35. In principle, there are two possible approaches for modulating the dopant incorporation, i.e., a control of W precursor concentration and a control of dopant adsorption on nanowire surface. On the other hand, the morphology can be manipulated by controlling CZn according to the critical concentration for nucleation on ZnO crystal planes, which is so-called as “concentration window” principle22,23,24. However, as shown in Fig. 2c, W addition substantially narrows the concentration window for enhancing the anisotropic nanowire growth. Therefore further increase of CW is not practical for enhancing W incorporation. Here, we control the adsorption of WO42− ions on nanowire surface by varying the thermodynamic parameter. The DFT calculation revealed that the resident time of WO42− ions on (0001) plane becomes longer by decreasing the temperature, as shown in Table 3. Supplementary Fig. 9a, b shows the TEM images and XRD patterns of W-doped ZnO nanowires (CZn 12 mM and CW 0.25 mM) grown at 95 and 75 °C, respectively. In this experiment, the sidewall growth i.e., nucleation on \(\left( {10\bar 10} \right)\) plane of each sample is perfectly suppressed by controlling CZn below the critical nucleation concentration. We found that (0002) peak of W-doped ZnO shifts to higher angle (i.e., smaller d value) by decreasing the growth temperature, meaning that more W is incorporated into nanowires. We successfully demonstrated the controlled morphology and elemental doping of W-doped ZnO nanowires for the first time by the predictable model based on coordination structure of impurity metal ions. The controllable nanowire morphology and elemental doping in hydrothermal synthesis enable to design the property of various ZnO nanostructure-based applications such as photovoltaic devices, photocatalysis, and chemical sensing47,48,49,50.

Table 3 The average surface resident time of WO42− ion on (0001) plane at 75 and 95 °C, estimated from the Frenkel equation using a typical molecular vibrational period 1.0 × 10−13 s.
Full size table

In conclusion, we demonstrated the unusual face-selectivity of WO42− ions on growth direction and dopant incorporation in hydrothermal ZnO nanowire synthesis. Introduction of WO42− ions significantly enhanced nucleation at sidewall \(\left( {10\bar 10} \right)\) plane and led to a formation of nanoplatelet structures while the dopant incorporation occurred only at (0001) plane, which cannot be explained by the existing models. These conflicting face-selective behaviors of WO42− ions led to inhomogeneous dopant distribution in an individual hydrothermal ZnO nanowire. By STEM-EDS observation and 3DAP analysis, we evidently demonstrated the inhomogeneity of dopant distribution in hydrothermal ZnO nanowire. DFT calculations revealed that the conflicting face-selective behaviors of WO42− ions can be interpreted in terms of the coordination structure of WO42− ions on each ZnO crystal plane. WO42− ions coordinated on \(\left( {10\bar 10} \right)\) plane with unipodal or bipodal bonds serve as surfactant for attracting Zn ion species, whereas those coordinated on (0001) plane with tripodal bonds are easily incorporated into nanowire as dopant due to the different stability of their coordination structure. Based on this predictable model, we successfully demonstrated the rational strategy for controlling over the morphology and the elemental doping of W-doped ZnO nanowire via modulating nucleation events and impurity adsorption with “concentration window” principle. This study clearly highlights the essential importance to understand the coordination structure of metal ions for designing the hydrothermal metal oxide nanowire synthesis and the resultant functions of metal oxide nanowire-based device applications.

Methods

W-doped ZnO nanostructures/nanowires were synthesized by a seed layer-assisted hydrothermal method under atmospheric pressure. Prior to the growth, a 50-nm-thick ZnO seed layer was deposited onto a 100-nm-thick SiO2 coated Si (100) substrate with a 1-nm-thick Ti adhesion layer by radio frequency (RF) sputtering (Ar pressure 0.3 Pa, RF power 50 W). The growth solution was prepared at room temperature by mixing zinc nitrate hexahydrate (Zn(NO3)2·6H2O, Wako 99.0 %), HMTA ((CH2)6N4 Wako 99.0 %), PEI (number average m.w. 1800, ALDRICH 50 wt% in H2O), and sodium tungstate dehydrate (Na2WO4·2H2O, ALDRICH ≥ 99%) in deionized (DI) water with varying their concentrations. The concentrations of each reagent was varied in the range of 8–25 mM for Zn(NO3)2, 0–3 mM for PEI, and 0–0.5 mM for Na2WO4. For all experiments, the concentration of HMTA 25 mM was constant. Then the hydrothermal reaction was conducted at 95 and 75 °C for 5 h by immersing the seed layer-coated substrate into the growth solution in manner of upside down. In two-step growth for growing non-doped core ZnO nanowires and W-doped ZnO shell, 1st step process and 2nd step process were conducted with Zn(NO3)2 25 mM and Na2WO4 0 mM for 5 h and with Zn(NO3)2 25 mM and Na2WO4 0.25 mM for 5 h, respectively. After the growth, the samples were rinsed in DI water and dried by N2 blow at room temperature. The equilibrium concentrations of ionic species in the growth solution was calculated by Visual MINTEQ 3.1. at constant temperature of 95 °C. The DFT calculations for the coordinated structures of WO42− ion and Zn ion species on ZnO surface were performed by VASP 5.4.4 using implicit water solvation model. The resident time of WO42− ion on ZnO crystals was estimated by using the Frenkel equation. The morphology, composition, and crystal structure of synthesized nanostructures were characterized by field effect scanning electron microscopy (FESEM, JEOL JSM-7610F) equipped with EDS at accelerating voltage of 15–30 kV, TEM (JEOL JEM-2100F) at accelerating voltage of 200 kV, XPS (Kratos AXIS-ULTRA) with Al Kα radiation source (12 kV, 5 mA, monochromator), and XRD (Philips X’Pert MRD) with Cu Kα radiation source (45 kV, 40 mA, monochromator), respectively. The binding energies were corrected at C 1s peak. HAADF-STEM and EDS analyses in Fig. 5 and Supplementary Fig. 7 were performed by FEI Titan G2 80-200 at accelerating voltage of 200 kV. 3DAP analysis was carried out using CAMECA LEAP5000XS in a laser mode at specimen temperature of 30 K and the laser pulse energy and pulse rate were 30 pJ and 250 kHz, respectively. For samples preparation of HAADF-STEM and 3DAP, Ni protection layer with 200-nm thickness was first deposited to support nanowires. Then, a sliced sample and a needle-shaped sample composed of a single nanowire, were prepared for HAADF-STEM and 3DAP, respectively, by focused ion beam (FIB) technique (FEI Helios G4UX) using the standard lift-out method at accelerating voltage of 30 kV. Low accelerating voltage of 2 kV was used in the final cleaning steps for reducing the damages caused by 30-kV FIB.