Energetic Au ion beam implantation of ZnO nanopillars for optical response modulation

ZnO nanopillars were grown on Si wafers using a hydrothermal approach [33]. Polished silicon wafers 1 × 1 cm in size from Siegert Wafer (Germany) with the orientation of <100> ± 5° were immersed for 10 min in a 75 °C solution of deionised H₂O (0.055 μS cm⁻¹, Evoqua, USA), NH₃ (30%, Chemsolute, Germany), H₂O₂ (25%, Chemsolute, Germany) in the ratios of 5:1:1, followed by rinsing in deionised H₂O and a further cleaning two times in an ultrasonic bath. Afterwards, they were dipped for 20 s in a solution of 0.005 mol l⁻¹ zinc acetate (Alfa Aesar, UK) in isopropanol (Chemsolute, Germany) and rinsed immediately with isopropanol. This procedure was repeated four times to assure a continuous coverage of zinc acetate. The samples were heated for 20 min at 350 °C to get a thin layer of calcined ZnO seeds on the silicon substrate. The samples were placed almost vertically in 50 ml tubes, with the top side facing slightly downwards. An equimolar solution of 0.05 mol l⁻¹ zinc nitrate (Alfa Aesar, UK) and hexamethylenetetramine (Alfa Aesar, UK) in deionised H₂O was filled in the tubes and used as deposition solution. Finally, the samples were cleaned five times in an ultrasonic bath with deionised H₂O. ZnO nanopillars were implanted with the Au-400 keV ions using the fluences of 1 × 10¹⁵ cm⁻², 5 × 10¹⁵ cm⁻² and 1 × 10¹⁶ cm⁻² using an implantor at the Ion Beam Center, HZDR Dresden-Rossendorf, Germany. The implanted samples were subsequently annealed at ambient atmosphere at 750 °C for 15 min in a conventional furnace in order to reduce the implantation damage and to support Au NP aggregation.

XRD was utilised for the study of the crystal-structure modification of the pristine ZnO nanopillars as well as those with Au NPs. We used a Rigaku SMARTLAB 9 kW rotating-anode generator and CuKα₁ characteristic radiation; the diffractometer was equipped with a primary parabolic multi-layered mirror and a 2 × 220 Ge channel-cut monochromator. In the middle-resolution (MR) set-up we used a linear detector on the secondary side, while for high resolution (HR) we applied a secondary 2 × 220 Ge channel-cut analyser and a point detector. The MR set-up was used for long-range 2Θ-scans keeping the incidence angle constant (${\alpha _i} = 0.4^\circ$). Almost symmetric 2Θ/ω scans were measured in the HR set-up; in these scans we avoided very intense diffraction peaks of the Si substrate by introducing the angular offset of $\Theta - \omega = 5^\circ .$

RBS was employed for characterisation of ZnO nanopillars using a beam of 2.0 MeV He⁺ ions from a Tandetron accelerator using the detector placed at scattering angle 170°. Simultaneously, RBS was used for the implanted Au depth profile analysis.

SEM was realised using a TESCAN MAIA 3 instrument with a field emission gun (TESCAN, Czech Republic). All samples were attached to SEM stubs using double-sided conductive carbon tape. For side-view imaging, a small piece of Si substrate was cleaved to expose the transition between the substrate and ZnO nanopillars. For all samples, imaging was performed in high vacuum at 5 kV accelerating voltage using an in-beam detector.

The optical emission properties of ZnO nanopillars were investigated by PL measurements carried out at 10 K by employing a closed-cycle He refrigerator system and using a 325 nm wavelength He–Cd laser as an excitation source (power density ∼20 W cm²). The PL spectra were recorded by fiber-optic spectrometers (Ocean Optics HR4000/USB4000) with a spectral resolution below 2 nm. The optical scattering and absorption properties were investigated at room temperature by means of diffuse-reflectance spectroscopy (DRS) using EVO-600 (Thermo Fisher Scientific, Inc.) UV–Vis spectrophotometer.

The spectra of the ellipsometric parameters were recorded by a VASE ellipsometer (JA Woollam Co.) in the VIS spectral range for incidence angles of 50°‒80°. Reflectance measurements were carried out by the same instrument and were numerically treated simultaneously along with the ellipsometry with the help of WVASE commercial software. In the investigated spectral range, the light beam reflected from ZnO nanopillars showed predominantly specular reflection with a negligible diffused component. Ellipsometric spectroscopy, as an indirect characterisation technique, requires a sample model and materials optical constants in the form of table data or parameterised dielectric functions. The sample model consisted of crystalline Si substrate with a SiO₂ native oxide over-layer (2.5 nm) covered by ZnO nanopillars. The ZnO nanopillar layer was modelled as a layer with a gradient profile of refractive index. The non-uniform thickness of the ZnO nanopillar layer was also accounted for and was found to be closely related to the recorded depolarisation factor. The optical constants of crystalline Si were taken from the WVASE software database. The ZnO optical constants were parameterised by Cauchy dispersion relation extended with the short-wavelength exponential absorption edge [34]. The upper (surface) part of the ZnO layer, where the implantation of Au ions was expected, was further described by an additional Lorentz harmonic oscillator to account for possible Au local surface plasmon resonance (SPR) (see supplementary material for details available online at stacks.iop.org/JPhysD/55/215101/mmedia).

For a better understanding of the electronic and nuclear stopping interplay in the implanted ZnO, SRIM calculation [35] of energy stopping was performed. Additionally, we calculated the projected range using SRIM for the Au-400 keV ions as R_P = 65 nm with the standard deviation ΔR_P = 19 nm. SRIM calculation of the Au-ion 400 keV depth profiles, O and Zn vacancies creation (displacement energies for Zn and O used in SRIM calculations were referred to in [15, 16]) and the energy stopping in ZnO nanopillars has been provided. However, it must be emphasised that SRIM calculation is not capable of taking into account any morphological aspects, so a 500 nm thick ZnO bulk layer with ZnO bulk density is taken into account in the calculation in figures 1(a) and (b). Moreover, we provided additional simulation with the effective density calculated by using the known ZnO nanopillar layer thickness as it was determined from SEM cross-section measurement on the pristine samples (see section 3.3). The projected height of the ZnO nanopillars was averaged over several SEM measurements being about 380 nm (taken as a perpendicular thickness of the nanopillar layer facing the penetrating ion implantation beam). Additionally, based on the RBS analysis, the pristine sample exhibited an areal atomic density of the ZnO nanopilar layer 1626 × 10¹⁵ atoms cm⁻² taking into account Zn and O atoms (see section 3.2, table 1).

Zoom In Zoom Out Reset image size

The known areal atomic density of the ZnO layer and the independently determined average ZnO layer thickness from SEM analysis were used to determine the effective density of ZnO nanopillar layer (3.0 g cm⁻³) differing significantly from the density of bulk ZnO (5.61 g cm⁻³). SRIM calculation of the Au-400 keV projected ranges with the modified ZnO layer density is presented for comparison in figure 1(a) (dashed line) and electronic and nuclear stopping powers are shown in figure 1(b) (dashed lines). It is clear, that ZnO nanopillars would suffer with higher uncertainty of the Au-ion projected ranges due to significant ion-energy straggling during the penetration through voids in ZnO layer. The projected range and standard deviation for the ZnO layer with the effective density 3.0 g cm⁻³ was calculated by SRIM as R_P = 120 nm, ΔR_P = 36 nm.

The results published recently in [16, 32, 36] give us a picture of the distinct radiation damage growth, Au NP coalescence and optical response in ZnO single crystalline orientations, which should be taken into account in the case of the experiment with ZnO nanopillars implanted by Au-400 keV ions. Au NP coalescence in such ZnO nanopillars and their resulting optical response might differ from ZnO bulk as radiation damage, accumulated in polar and non-polar planes, taking place at the same time and radiation defects would cumulate distinctly due to the high effective ZnO surface acting as a sink for defects.

RBS analysis was used to follow, at least qualitatively, the Au distribution in ZnO nanopillar layer. The pristine ZnO nanopillars were not fully oriented along c-axis <0001>, so structural analysis based on the ion channelling effect is not possible. Moreover, the depth scale calculation is strongly affected by the complex morphology and high roughness of ZnO nanopillar layer. Besides, the density cannot be taken as the bulk ZnO density.

However, we can qualitatively follow the RBS spectra of ZnO nanopillars deposited on Si, where we saw the modification of the Zn and O signals from the ZnO nanopillars layer. The spectra of the pristine and the as-implanted ZnO nanopillars are presented in figure 2(a). The Au depth profiles are presented in figure 2(b). In general, the Au depth profiles measured by RBS differ from the SRIM simulation for bulk ZnO layer (with the bulk density 5.61 g cm⁻³). RBS profiles are broader and exhibit an asymmetric Au-distribution. Implantation into the nanopillars differ from implantation into compact layers and suffered from a strong scattering of the projected ranges of the implanted ions, so that the Au ion distribution was much broader than it was theoretically calculated. However, it can be suggested that the Au coalescence can take a place within the major part of the nanopillars. RBS did not evidence Au content in the Si substrate.

Zoom In Zoom Out Reset image size

Additionally, RBS analysis is sensitive to a surface morphology, where a high roughness and porosity of the ZnO surface can be mimicked by the typical in-depth tales at Au-depth profiles. A roughness effect exhibited in the RBS depth profiles can be simulated properly in case of the surfaces with regular and simple morphology, as it was investigated e.g. in [37, 38]. We can conclude that RBS analysis determined the depth of the Au maximal concentration. However, the in-depth tail of the Au-peak in the RBS spectra (see figure 2(a)) could be connected to the surface roughness effects [37, 38]. The effective density of the ZnO layer was used for converting the depth scale of the Au-depth profile evaluation to nm.

We observed the Au-concentration maximum shift to the surface in the ZnO nanopillar layer with increasing Au-ion fluence in the implanted samples (see figure 2(b)). It can be connected to the deterioration of the ZnO nanopillar layer in the upper part facing to the ion implantation beam.

We performed the detailed analysis of the ZnO thicknesses in atoms cm⁻² from RBS spectra showing the total number of Zn and O atoms in the layer. The results are summarized in table 1. The ZnO thicknesses in atoms cm⁻² decreased for fluences of 5 × 10¹⁵ and 1 × 10¹⁶ cm⁻². It showed us ZnO nanopillar layer degradation under the Au-ion bombardment. Zn depletion in RBS spectra can be connected to nanopillar erosion as it was observed also in [39] after Cu-ion implantation of ZnO nanorods. ZnO surface morphology modification can cause the Zn-signal asymmetry in the RBS spectrum [38, 39].

This effect could be related also to the compaction of ZnO nanopillars layer at the bottom part, while sputtering effects would degrade the ZnO nanopillar upper parts. The annealing slightly influenced the Au depth profile shapes (see figure 2(b)) without a clear correlation to the Au-ion fluence, however the Au-integral amount did not change (see table 1). We did not see significant changes of the Zn peak in the RBS spectra of the annealed samples compared to the implanted samples (see table 1 showing ZnO thicknesses after the annealing); thus, the annealing modifies the internal structure rather than the ZnO nanopillar layer thickness.

XRD analysis of ZnO nanopillars was carried out by combining the MR 2Θ scans with the HR 2Θ/ω scans. The former experiments yielded much higher diffracted intensity, which made it possible to determine the ZnO lattice parameters from a large number of diffraction maxima with sufficient accuracy. Moreover, these measurements are suitable for searching for possible (very weak) diffraction peaks of possible Au NPs. On the other hand, the high diffracted intensity of the 2Θ scans is achieved by a much worse angular resolution, so that these scans cannot be used for the determination of the size of the coherent domains in nanopillars.

Figure 3 displays the 2Θ scans measured before (blue) and after (red) annealing. All the ZnO diffraction maxima could be identified in the measured curves. Most probably, additional non-identified peaks (denoted by question marks in figure 3) correspond to long tails of diffraction maxima of the Si substrate (crystal truncation rods). These peaks very sensitively depend on the azimuthal sample orientation and this is the reason why they do not appear in all diffraction curves. The ZnO peak heights roughly correspond to the calculated intensities; therefore, the ZnO layer is polycrystalline with no strong texture. Au diffraction peaks were fully not resolved, only in the curve of the annealed sample implanted with the highest Au fluence (figure 3(d), the red curve) not fully resolved Au reflections 111 and 200 might be suspected.

Zoom In Zoom Out Reset image size

We fitted the individual ZnO diffraction maxima to Voight functions and from the positions of the diffraction maxima we determined the ZnO lattice parameters (see table 2). For this procedure, we used the well-known Cohen‒Wagner approach modified for hexagonal crystals [40].

Table 2. The lattice parameters and radii of ellipsoidal coherent domains in ZnO nanopillars as implanted with Au-400 keV ions and as-annealed at the various ion-fluences are compared to the pristine ones.

Au-ion fluence (ions−2)	ZnO nanopillars—as-implanted; lattice parameters (Å)	ZnO nanopillars—as-annealed; lattice parameters (Å)	ZnO nanopillars—as-implanted; domain radii (nm)	ZnO nanopillars—as-annealed; domain radii (nm)
Pristine	a = 3.244 ± 0.002 c = 5.198 ± 0.001	a = 3.242 ± 0.002 c = 5.192 ± 0.001	${R_{\text{B}}} = \left( {21 \pm 3} \right)$ ${R_{\text{V}}} = \left( {59 \pm 1} \right)$	${R_{\text{B}}} = \left( {20 \pm 3} \right)$ ${R_{\text{V}}} = \left( {65 \pm 1} \right)$
1 × 1015	a = 3.243 ± 0.002 c = 5.192 ± 0.001	a = 3.241 ± 0.002 c = 5.189 ± 0.001	${R_{\text{B}}} = \left( {13 \pm 3} \right)$ ${R_{\text{V}}} = \left( {23 \pm 1} \right)$	${R_{\text{B}}} = \left( {18 \pm 3} \right)$ ${R_{\text{V}}} = \left( {28 \pm 1} \right)$
5 × 1015	a = 3.241 ± 0.002 c = 5.196 ± 0.001	a = 3.243 ± 0.002 c = 5.192 ± 0.001	${R_{\text{B}}} = \left( {14 \pm 3} \right)$ ${R_{\text{V}}} = \left( {11 \pm 1} \right)$	${R_{\text{B}}} = \left( {20 \pm 3} \right)$ ${R_{\text{V}}} = \left( {17 \pm 1} \right)$
1 × 1016	a = 3.243 ± 0.002 c = 5.198 ± 0.001	a = 3.245 ± 0.002 c = 5.194 ± 0.001	${R_{\text{B}}} = \left( {15 \pm 3} \right)$ ${R_{\text{V}}} = \left( {11 \pm 1} \right)$	${R_{\text{B}}} = \left( {22 \pm 3} \right)$ ${R_{\text{V}}} = \left( {17 \pm 1} \right)$

The HR offset scans are displayed in figure 4. From the integral widths ${w_j}\,$ of the HR diffraction maxima, we determined the mean sizes of the coherent domains in the ZnO nanopillars. We assumed that the domains are uniaxial ellipsoids oriented along the [0001], where the mean radii along and across [0001] are denoted R_B and R_V, respectively. We reformulated the Williamson‒Hall approach [41] for ellipsoidal domains. The domain radii are determined from the minimum of the following residuum function.

$$\begin{equation}S = \sum\limits_j {{{\left[ {{w_j}\cos {\theta _j} - {A_j}\left( {{R_{\text{B}}},{R_{\text{V}}}} \right) - B\sin {\theta _j}} \right]}^2}} ,\end{equation} \tag{ 1 }$$

Zoom In Zoom Out Reset image size

where the sum runs over individual diffraction peaks with diffraction angles ${\theta _j}$, the constant $B$ accounts for a possible microstrain in the nanopillars (assumed isotropic for simplicity), and:

$$\begin{equation}{A_j}\left( {{R_{\text{B}}},{R_{\text{V}}}} \right) = \lambda K\frac{{\sqrt {R_{\text{B}}^2 + \left( {R_{\text{V}}^{\text{2}} - R_{\text{B}}^{\text{2}}} \right){\text{si}}{{\text{n}}^2}\left( {{\gamma _j}} \right)} }}{{{R_{\text{B}}}{R_{\text{V}}}}}.\end{equation} \tag{ 2 }$$

Here we denoted $\lambda = 1.5405$ Å as the x-ray wavelength, $K \approx 0.29$ is the shape factor, and ${\gamma _j}$ is the angle between the ellipsoid axis (assumed to be parallel to [0001]) and the $j$th diffraction vector. The resulting values of R_B,V are summarised in table 2 as well.

Table 2 shows that the lattice parameters do not significantly change in the ZnO nanopillars after implantation or annealing. This indicated that Au dopant preferably would not enter any substitutional position, as was already seen in [42], but rather creates Au interstitials and clusters. However, we saw the modification in the basal and vertical domain radii decreasing with increasing Au-400 keV ion fluence. After the annealing, we observed the partial restoration of the domain sizes mainly in the basal direction and much less in the vertical direction. However, we can conclude from figure 3 that there is no full restoration of the diffraction peaks, keeping the broader FWHM and the lower intensity compared to the pristine sample. Thus the persisting damaged structure is evidenced in the ZnO nanopillars.

Figure 5 shows the surface morphology and cross-sectional SEM images of the ZnO nanopillars grown on silicon substrates. Medium density and uniformly distributed nanopillars with the upper defined hexagonal facets were grown vertically on the substrates. However, some bending and not perpendicularly oriented nanopillars appeared in the pristine sample (see figure 5(a)). The diameters of the nanopillars were analysed from SEM images, being in the broad range from 77 to 139 nm and the thickness of the ZnO nanopillar layer was deduced from the cross-sectional images, being about 360‒400 nm. The morphologies of the nanopillars before and after the Au-400 keV ion implantation are presented with the ascending ion implantation fluence in figures 5(b)–(d), respectively.

Zoom In Zoom Out Reset image size

We observed degradation of the ZnO nanopillars in figure 5 for the higher Au-ion fluences as a result of radiation damage. This degradation is exhibited by sharpening of ZnO nanopillar tips, as well as interconnecting of the bottom ZnO nanopillar parts exhibited at the highest Au-ion implantation fluence (compare figures 5(a)–(d)). As the implantation fluence increased, the surface of ZnO nanopillars could be destroyed. Small pits were evidenced on ZnO nanorod surfaces implanted with Cu-ions by HR TEM in [39]. The authors suggested a pit aggregation into voids mainly after annealing. SEM images of the ZnO nanopillars after the annealing are presented in figure 6 for the higher Au-ion fluences of 5 ×10¹⁵ and 1 ×10¹⁶ cm⁻² in figures 6(a) and (b), respectively. A very similar ZnO nanopillar morphology is observed after the annealing.

Zoom In Zoom Out Reset image size

Optical emission and absorption properties of ZnO nanopillars addressed by PL and DRS techniques are summarized in figure 7. Note that these measurements were performed only on Au-implanted and annealed ZnO nanopillars along with the pristine (non-implanted) sample as a reference. The evolution of PL spectra obtained at 10 K as a function of Au-ion implantation fluence is shown in figure 7(a), where all curves are normalized with regard to the deep-level emission (DLE), the common feature dominating in all spectra. This way of presenting the data highlights the implantation-induced spectral changes within the broad DLE band, which is comprised of several partially overlapping and thus hardly distinguishable components. The original non-normalized DLE as well as the total luminescence yield appear as a rapidly decreasing function of implantation fluence (not shown), similarly to ion irradiated ZnO nanorods reported in [43].

Zoom In Zoom Out Reset image size

The pristine sample exhibits PL features typical to high-crystallinity ZnO—a narrow NBE peak at 368 nm (3.37 eV) associated with exciton recombination and a broad DLE band centered at around 600 nm (2 eV) originating from luminescent deep defect sites (V_O—green luminescence, the transition between Zn_i and O_i levels—orange-red luminescence, etc) [32, 44]. By contrast, Au-implanted and post-annealed samples demonstrate generally suppressed NBE emission and apparent developments of DLE as function of the Au-ion fluence, particularly within the GL region centered at around 520 nm (2.4 eV). While the absence of NBE recovery after thermal annealing would be rather unusual for ZnO, in the present case of Au-implanted ZnO nanopillars this is accompanied with GL enhancement and thus might be indicative of an efficient charge transfer to plasmonic Au NPs. The similar effect of the charge transfer modification due to the increasing Au NP load in ZnO leading to NBE decrease and emerging of emission at 500 nm was observed in [45]. The spectral region of GL enhancement (highlighted by the shaded area in figure 7(a)) falls well within the reported PL activity range of Au NPs [46]. Indeed, SPR is a complex effect that modifies light coupling with a medium containing embedded Au NPs. The excitation of SPR in Au NPs results in an extreme concentration of light and an enhanced electric field around the nanostructures, leading to the significant enhancement in absorption and scattering efficiencies for photons at the resonant energy [47].

The light scattering phenomenon was addressed independently by means the DRS, where the detected signal, being directly proportional to the scattering efficiency, has provided a measure of SPR effect for different implantation fluencies. The differential DRS spectra presented in figure 7(b) are built by dividing each DRS by that of the pristine (non-implanted) ZnO nanopillars, and in this way directly exposing Au-implantation induced changes at different fluences. The well matching spectral regions of the enhanced scattering and GL emission in figures 7(a) and (b) strongly support their common plasmonic nature.

The absorption properties of ZnO nanopillars assessed by DRS indicate certain regions of the absorbance upturn (550–750)nm and downturn (450–600)nm with reference to the pristine (non-implanted) ZnO nanopillars (not shown). While DRS offers only qualitative assessments via Kubelka–Munk (K–M) function ascribed by both absorbance and scattering (sum of these parameters is extinction), the ellipsometry deals with extinction in a quantitative way, i.e. provides the exact extinction coefficient values and also offers independent measurements for comparison as described in the following section.

Optical ellipsometry was employed to evaluate optical properties of ZnO pristine, irradiated and annealed nanopillars. Sample structure models tuned by ellipsometry fitting yielded refractive index gradient profiles across the layer (cf supplemental information). It was found that the refraction index in the pristine (figure 8(a)—left), as well as in the implanted ZnO nanopillars (figure 8(a)—right, (e) and (f)), exhibited a gradual step-like vertical depth profile. This corresponds with increasing void fraction towards the top of the nanopillar structure and it is consistent with the SEM observations. The refractive index variation is plotted for the selected wavelength of 500 nm in ZnO nanopillar layers in figure 8. It is worth noting that the substrate side with compact ZnO reads refractive index values close to bulk ZnO which is 2.1, and that the upper-most part of ZnO layer with low density of nanopillars is approaching ambient refractive index value of 1.0, as it is expected. This provides a degree of confidence to our models and fits (see the detailed fits of elipsometry spectra in supplement figure S3). The Au-ion implantation induced radiation damage. The implanted (figures 8(d)–(f)) and post-annealed samples (figures 8(b) and (c)) showed refractive index enhancement in the bottom part of the nanopillar layer (in figure 8 the retrograde depth scale is used, 0 nm meant Si/ZnO interface).

Zoom In Zoom Out Reset image size

The RBS evidenced the Au-content distributed within the layer about 50–200 nm below the surface of the ZnO nanopillar layer. In the upper part of the nanopillar layer, a decreased refractive index was observed, whereas the refractive index in the layer near the Si substrate increases after the implantation and annealing. A sharper interface appeared between the various values of the refractive index in the ZnO nanopillar upper part enriched with the Au NPs and the bottom part close to Si substrate (see figures 8(b)–(f)).

The optical activity of Au NPs is presented for the implanted and annealed samples in figure 9. The absorption maximum ascribed to Au NPs or clusters is evidenced from 500 to 650 nm. We observe the more pronounced intensity of the absorption peak (SPR) ascribed to the Au NPs with the increased Au-ion fluence. Additionally, the peak maximum is shifting to the higher wavelenghts with the increased fluence (see figure 9). The annealing caused the absorption peak to shift to the higher wavelengths and increased the intensity for the lower fluences. For the highest Au-ion implantation fluence of 1 × 10¹⁶ cm⁻², we do not see this effect.

Zoom In Zoom Out Reset image size

In the literature [48], the Au NPs decorating ZnO nanowires were analysed by optical spectroscopy. The extinction spectrum contained a distinctive peak at 520 nm which was ascribed to the absorption of visible light by Au NPs with a small degree of agglomeration. However, the Au NP shape, size, and complexity influenced the absorption maximum being evidenced from 520 to 600 nm in the literature [14, 48, 49]. We can find the SPR observation in a range of wavelengths of 530 ∼ 585 nm in ZnO–Au composites, where the SPR red shift has been ascribed to the agglomeration of Au NPs [48, 49]. The ellipsometric spectra obtained from analysis of pure Au island-like films were fitted using the contributions of the Drude function and Lorentz oscillators in [49], where the SPR positioned around 600 nm has been ascribed to Au NPs 15 nm in size. We could suggest from the SE results that rather non-spherical Au NPs or clusters are present in the ZnO nanopillars (SE data fits are presented in supplementary material).