Using a Patterned Sapphire Template And The Lift-off Technique To Synthesize ZnO Nanoower Arrays With Enhanced Photoluminescence Properties

In this study, the hydrothermal method was used to synthesize ZnO nanorods and ZnO nanoower arrays. Two different substrates were used to prepare the ZnO seed layer. For the p-type silicon <100> wafer, a prepared ZnO gel was deposited as the seed layer using the spin coating method. When a patterned sapphire recess-type substrate was used as a template, Al lm, with a thickness of 120 nm, and OE-6370HF AB glue were used as a sacricial layer and an imprinting lithography carrier for the ZnO seed layer, respectively. To prepare the array-patterned ZnO seed layers with a protrusion structure, a lift-off technique was used. A 0.2 M solution of zinc acetate dihydrate (Zn(CH 3 COO) 2 -2H 2 O) was used at a synthesis temperature of 90°C and a synthesis time of 10–60 min. Because the ZnO seed layer had a protrusion and matrix structure, the ZnO nanorods grew in the vertical bottom direction to form ZnO nanoower arrays. X-ray diffraction patterns, scanning electron microscopy, and a focused ion beam system were used to analyze and compare the crystal characteristics and the heights and widths of the ZnO nanorods and ZnO nanoower arrays. We found that the photoluminescence properties were enhanced in the ZnO nanoower arrays compared with the ZnO nanorods.


Introduction
Group II-VI metal oxide semiconductors, which include zinc oxide (ZnO), can be used in a wide range of devices. In 1968, researchers grew ZnO single crystals and proved that ZnO had excellent piezoelectric characteristics [1]. At room temperature, ZnO-based materials have a wide direct band gap of about 3.37 eV and a large exciton binding energy of about 60 meV. These materials also have high chemical and thermal stability, a low dielectric constant, high photocatalytic properties, and high transparency in the visible light and infrared ranges. Because ZnO-based lms have many superior properties, various methods to deposit ZnO-based lms, including the chemical vapor transport method, spray pyrolysis method, metal organic chemical vapor deposition method, pulsed laser deposition method, and sol-gel method, have been investigated. The sputtering method of lm deposition is most widely used because it can deposit ZnO-based lms with different dopants by the changes in the composition of the prepared targets [2]. ZnO-based lms can be used in many different applications, including varistors [3], light emitting diodes [4], optical waveguide and photodetector arrays [5], transparent electrodes [6], dyesensitized solar cells [7], and sensors for the detection of different objects [8,9].
Many process technologies have been investigated to fabricate ZnO-based nanomaterials with various crystal shapes. For example, Karthik et al. utilized homogenous precipitation methods with urea and ammonium hydroxide as precipitation agents to synthesize ZnO with structural disarray in the lattice and needle-like microstructures [10]. Zhang et al. utilized and precursors in various solvents under different reaction conditions to synthesize ZnO with various morphologies, including ower-like, prism-like, prickly sphere-like, and rod-like shapes [11]. Chen et al. used the hydrothermal method to synthesize ZnO nanorods (or nanowires), and found that the deposition time and the face direction of the substrates in the hydrothermal bottle were important factors that affected the morphologies of the ZnO nanomaterials [12]. The researchers also found that the ZnO nanostructures had three different morphologies: beautiful chrysanthemum-like nano owers, nanorods, and irregularplate structure lms.
Zhang et al. used a capping-molecule-assisted hydrothermal process to investigate ZnO microcrystals [13]. They found that the synthesized formations of dumbbell-like, disk-like, and ower-like shapes of the ZnO microcrystals depended on the capping molecules that were used. Wang et al. also used the hydrothermal method to synthesize ZnO microcrystals [14]. They found that controlling the pH of the solution, the reaction time, and the concentration of the precursor affected the ower-like rod aggregation, hexagonal pyramid-like rod, and microrod appearances of the ZnO nanomaterials. The hydrothermal method is the most economical, simple, and commonly used method to synthesize ZnO nanostructures.
ZnO nanorods and ZnO nano ower arrays may have applications in various optoelectronic devices and, thus, it is important to develop methods to synthesize ZnO in these con guration. In this study, we used the hydrothermal method to synthesize ZnO nanorods and ZnO nano ower arrays using ZnO seed layers on different substrates. In general, ZnO nanorods that are synthesized on a plane-typed substrate perpendicular to the substrate. When ZnO nanorods are used as sensors for different signals, the rst sites for contact and reaction are at the top surfaces of the nanorods. When ZnO nanorods are grown in a structure of radiating needles, they can synthesize to form ZnO nano owers. When ZnO nano owers are used as gas or optical sensors, they have a greater area for contact and reaction sites and then to enhance the reaction sensitivity and the response speed.
The important novel innovation of this study was the development of a low-cost technology to fabricate a patterned ZnO seed layer. This technology involved the use of patterned sapphire templates and the liftoff technique to synthesize ZnO nano ower arrays. We used synthesis time to analyze the properties of ZnO nanorods synthesized on a p-type silicon (Si) < 100 > wafer, as well as the properties of ZnO nano ower arrays synthesized on a prepared patterned substrate. The investigated ZnO nano owers had different shapes from most reported ZnO nano owers, which typically present with prickly sphere-like, rod-like [11], or sea anemone-like morphologies [15]. A series of experiments were performed to investigate the effect of synthesis time on the physical properties, which included the crystal characteristics, morphology, height (or length), diameter, and aspect ratio, and the optical properties, such as the photoluminescence (PL) spectra of the ZnO nanorods and ZnO nano ower arrays. During the experiments, the synthesis time was varied, but the other parameters were kept constant. The novel analysis of this study was that we investigated and compared the height/diameter (H/D) ratio of the ZnO nanorods and ZnO nano ower arrays with the PL properties. We proved that for the same synthesis time, the PL emission intensities of the ZnO nano ower arrays were enhanced compared with the intensities of the ZnO nanorods.

Preparation of ZnO gel
We used zinc acetate dihydrate (Zn(CH 3 COO) 2 -2H 2 O), 2-aminoethanol (monoethanolamine) (C 2 H 7 NO), and ethylene glycol monomethyl ether (CH 3 OCH 2 CH 2 OH) to prepare a solution of 0.75 M Zn + ions. The solution was heated to 60°C, stirred for 2 h, and left for 48 h. Figure 1(a) shows the transparent ZnO gel solution that was obtained.

Process of synthesis ZnO nanorods
To prepare the Al sacri cial layer, a 2-inch sapphire substrate, provided by Shun Haw Technology Ltd., Taiwan, was used. A schematic diagram and a cross-sectional view of the substrate is shown in Figs. 2(a) and 2(b), respectively. The average bottom width and average height of the substrate were 0.37 µm and 0.48 µm, respectively. First, we used the evaporation method to deposit the Al sacri cial layer on the sapphire substrate at a thickness of approximately 120 nm. The sapphire substrates were then put into a tube furnace and heated to 500°C for annealing. The annealing time was 1 h and the pressure in the furnace was controlled at 1.2 bar; a mixing gas of 90% Ar + 10% H 2 was introduced during the annealing process.

Preparation of the Al sacri cial layer
To prepare the Al sacri cial layer, a 2-inch sapphire substrate, provided by Shun Haw Technology Ltd., Taiwan, was used. A schematic diagram and a cross-sectional view of the substrate is shown in Figs. 1(a) and (b), respectively. The average bottom width and average height of the substrate were 0.37 µm and 0.48 µm, respectively. First, we used the evaporation method to deposit the Al sacri cial layer on the sapphire substrate at a thickness of approximately 120 nm. The sapphire substrates were then put into a tube furnace and heated to 500°C for annealing. The annealing time was 1 h and the pressure in the furnace was controlled at 1.2 bar; a mixing gas of 90% Ar + 10% H 2 was introduced during the annealing process.

Preparation of ZnO-seed layer using spin coating method
The ZnO gel was applied to fully cover the surface of the patterned sapphire template with the Al sacri cial layer. The spinning speed and time were set to 2,000 rpm and 30 s, respectively. After the coating process, the ZnO gel-coated template was baked at 300°C for 10 min to dry and harden the coating lm and to volatilize and remove the organic solvent. The processes of spin coating and baking were repeated six times, after which a ZnO thin layer, or seed layer, was obtained, as shown in Fig. 1(c).

Processes of synthesis ZnO nano ower arrays
Before the ZnO thin layer was lifted off, it was coated with optical OE-6370HF AB glue to act as a carrier, as shown in Fig. 1(d). We then used lift-off technology and imprint lithography the to make patterns on the ZnO thin layer to create the ZnO seed layer A solution of K 3 Fe(CN) 6 :KOH:H 2 O (10 g:1 g:100 ml) was used to etch the Al sacri cial layer. The protrusion and patterned ZnO seed layer was then obtained, as shown in Figs. 1(e) and 1(f). Next, the ZnO seed layer was attached to the p-type Si < 100 > substrate using OE-6370HF AB glue, as shown in Fig. 1(g). Finally, a 0.2-M solution of zinc acetate dihydrate (Zn(CH 3 COO) 2 -2H 2 O was used to synthesize the ZnO nano ower arrays at a syntesis temperature of 90°C and at synthesis times of 10 min, 20 min, 30 min, and 60 min ( Fig. 1(h). After the ZnO nanorods and ZnO nano ower arrays were synthesized, an X-ray diffractometer (XRD) was used to analyze their crystalline phases. A eld emission scanning electron microscope (FESEM) was used to observe the surface morphologies of the ZnO nanorods and ZnO nano ower arrays. A focused ion beam (FIB) system was used to cut the ZnO nanorods, the prepared patterned ZnO seed layer, and the ZnO nano ower arrays to observe their cross-sectional morphologies. The PL emission properties of the ZnO nanorods and ZnO nano ower arrays were measured in the wavelength range of 350 ~ 650 nm at room temperature using a Horiba Jobin Yvon iHR550 uorescence spectrophotometer. A single wavelength laser at a wavelength of 325 nm was used as the excitation light source.

Results And Discussion
The diffraction peaks of the ZnO seed layer located at the 2θ values of 31.9°, 34.6°, 36.48°, 47.64°, and 56.78° corresponded to the diffraction peaks at the (100), (002), (101), (102), and (110) planes, respectively, as shown in Fig. 3. Compared with the Joint Committee on Powder Diffraction Standards (JCPDS) card No. 36-1451, we found that the ZnO seed layer presented with a wurtzite structure that had a structure of hexagonal close-packed structure (HCP). The highest diffraction peak was located at the 2θ value of 34.6°, which corresponded to the diffraction peak of the (002) plane. This result suggested that the ZnO seed layer had a c-axis preferred orientation. The diffraction peak of the (002) plane was combined with the Debye-Scherrer's equation to calculate the crystalline size of the ZnO seed layer [16]: When the full width at half maximum (FWHM) value of the (002) plane was incorporated into Eq. (1), the crystalline size of the ZnO seed layer was calculated to be 27.6 nm.
The surface morphology and the cross-sectional image of the ZnO seed layer on the p-type Si < 100 > wafer are shown in Fig. 4. We used ZnO gel (i.e., as the precursor) and the spin coating method to prepare the ZnO seed layer. Although the deposited lm was annealed at 300°C, the porous state of each particle was distinct at the surface, as shown in Fig. 4(a). The particles were about 30 nm in size, which matched the results obtained using the Debye-Scherrer's equation and the diffraction peak of the (002) plane from  Figure 4(b) shows that the thickness of the ZnO seed layer was about 198 nm, which suggests that the thickness of each coating was approximately 33 nm.
The chemical reactions involved in the fabrication of the ZnO nanorods were divided into metal oxide hydrolysis and dehydration reactions. The chemical reactions are listed below [17,18]: From these chemical reaction equilibriums we can determine the direction of these chemical equilibrium reactions by controlling the concentrations of the reactants, the synthesis temperature, and the synthesis time. In general, the concentrations of the reactants determine the density of one-dimension ZnO nanorods, while the synthesis temperature and synthesis time affect the morphology, including the diameter, height, and aspect ratio, of the ZnO nanorods.
In this study, the synthesis time was used the parameter to investigate the properties of the ZnO nanorods. The XRD patterns of the ZnO nanorods as a function of synthesis time are shown in Fig. 5. We used the XRD patterns to determine that there was an apparent diffraction peak located at the 2θ value of 34.7°, which corresponded to the (002) plane; two weak diffraction peaks were located at the 2θ values of 31.9° and 36.48°, which corresponded to the (100) and (101) planes, respectively. These results suggested that the ZnO nanorods had a c-axis preferred orientation when the synthesis time increased, and that the synthesis rate along the peak of the (002) plane was more sensitive to the synthesis time compared with the peaks of the (100) and (101) planes. However, the diffraction peaks of the (102) and (110) planes located at the 2θ values of approximately 47.64° and 56.78° were not observed. Figure 5 shows that the diffraction intensity of the (002) plane increased with synthesis time, which suggests that the crystal quality of ZnO nanorods in the (002) plane improved with time. Figure 5 also shows that the diffraction intensities of the (100) and (101) planes decreased with time. The surface and cross-sectional morphologies of the ZnO nanorods as a function of synthesis time are shown in Fig. 6 and Fig. 7, respectively. Figure 6 and Fig. 7 show that as the synthesis time increased, both the height (i.e., length) and the diameter of the ZnO nanorods increased; the aspect ratio (height/diameter) also increased with synthesis time. Because the ZnO nanorods push against each other, they grow perpendicular to the substrate and, therefore, the diffraction intensities of the (100) and (101) planes gradually become weaker.
It is clear that the height, diameter, and aspect ratio, which is de ned as the height divided by the diameter, of the ZnO nanorods were closely related to the synthesis time. As Fig. 6 and Fig. 7 show, ZnO nanorods could be observed even when the synthesis time was 10 min. Both the average height and diameter of the ZnO nanorods increased, while the density of ZnO nanorods in a unit area decreased with synthesis time. The height and diameter of the ZnO nanorods were determined from Fig. 6 and Fig. 7. The aspect ratio, which can be used to analyze the relationship between the axial and radial growth of the ZnO nanorods, was calculated using the measured values. Both the height and the diameter of 10 ZnO nanorods were measured to nd their average values for synthesis times of 10 min, 20 min, 30 min, and 60 min. When the thickness of the ZnO seed layer was determined to be 200 nm, the average heights of the ZnO nanorods were 42 nm, 130 nm, 335 nm, and 1300 nm, and the average diameters of the ZnO nanorods were 54 nm, 71 nm, 88 nm, and 112 nm, respectively. Correspondingly, the calculated aspect ratios of the ZnO nanorods were 0.78, 1.83, 3.80, and 11.6 when the synthesis time was 10 min, 20 min, nanorods with a long height, a high aspect ratio, and a well-de ned hexagonal prism shape.
As Fig. 6 shows, the diameters and heights of the ZnO nanorods were uniform. Therefore, the relationships shown in Table 1 were used to calculate the radius, height, and aspect ratio of a cube that had been divided into different numbers of cylinders; the results are shown in Table 2. As Fig. 8 shows, when a cylinder (radius r = a/2), with side length a and volume πa 3 /4, is divided 4 cylinders (r = a/4) or 9 cylinders (r = a/6), and the height of each cylinder is a, although the total volumes of the cylinders are equal, the total surface area increases with the number of cylinders. As Table 1 shows, as the radii are a/2, a/4, and a/6, the total surface area are = 3πa 2 /2, 4 × 2π a/4 (a/4+a) = 5πa 2 /2, and 9 × 2π = 7πa 2 /2, respectively. Therefore, as the diameter of the ZnO nanorods decreases and the total volume of the nanorods in a unit area is unchanged, the density (i.e., quantity) of the cylinders in a unit area increases, while both the total surface area and the surface area to volume ratio increase. When prepared samples have the same total volumes, the total surface area or the surface area to volume ratio affects the characteristic responses of the ZnO nanostructures and have an effect on their use as optical devices or sensors. To calculate the density (i.e., quantity) of ZnO nanorods in a unit area, the surface area was divided into squares with an area of 1 µm 2 , and Fig. 9 shows 5 sample areas. We randomly sampled 8 squares, calculated the density of ZnO nanorods in each area, and determined the average value. The densities of the ZnO nanorods were 107 µm − 2 , 94 µm − 2 , 72 µm − 2 , and 65 µm − 2 when the synthesis time was 10 min, 20 min, 30 min, and 60 min, respectively. The measured heights and diameters, as well as the calculated densities, of the ZnO nanorods were used to determine the total surface area (S, nm 2 ), the total volume (V, nm 3 ), and the S/V ratio; the results are shown in Table 2 as a function of synthesis time. As the synthesis time increased from 10 min to 60 min, the total surface area increased from 1.01 ⋅ 10 6 to 3.04 ⋅ 10 7 nm 2 , the total volume increased from 1.03 ⋅ 10 7 to 8.32 ⋅ 10 8 nm 3 , and the S/V ratio decreased from 9.79 ⋅ 10 − 2 to 3.65 ⋅ 10 − 2 . These results showed that it was possible to manipulate the height, diameter, and aspect ratio of the ZnO nanorods by controlling the synthesis time.
The main reason for the increases in total surface area and total volume was that the height and diameter of the ZnO nanorods increased with synthesis time, while the density of ZnO nanorods decreased with synthesis time. From the results in Table 2, we know that the growth rate of ZnO nanorods in the vertical-axis direction (i.e., perpendicular to the substrate) was higher than that in the radial-axis direction, even though the density in a unit area decreased with time. ZnO-based materials have a stable crystal plane group, including nonpolar surface groups of {101 ̅ 0} and {21 ̅ 1 ̅ 0}, and a polar surface group of {0001}. In ZnO-based materials, the polar surface is caused by negatively charged oxygen ions and positively charged zinc ions, which form a vertical dipole interaction torque on the c-axis plane [19]. Therefore, the growth rate in the vertical-axis direction is higher than that in the radial-axis direction, which causes the ZnO nanorods to have a c-axis preferred orientation with a high aspect ratio at a long synthesis time.
When the hydrothermal method is used to synthesize ZnO nanomaterials, the polar {0001} plane has a better activity response than the nonpolar {101 ̅ 0} and {21 ̅ 1 ̅ 0)} planes. Therefore, the ZnO materials have a c-axis preferred orientation when forming one-dimensional nanomaterials, such as nanorods or nanowires. There are two factors that cause an increase in the diameter and a decrease in the density of ZnO nanorods. First, during the synthesis process, two adjacent nanorods may combine with each other to form one nanorod, which causes the diameter of the nanorod to increase and the density to decrease. Second, because the growth rate of the nanorods is inconsistent, those with a fast growth rate will remove elements they need, which can cause nanorods that have a slow growth rate to synthesize slower or to stop synthesizing. This causes the density of the nanorods to decrease and the diameter to increase. Table 2 The density, total surface area (S), total volume (V), and S/V ratio of the ZnO nanorods as a function of synthesis time. The PL emission spectra were used to analyze the types of defects in the ZnO nanomaterials. The ZnO seed layer and the ZnO nanorods were excited using a laser with an excitation wavelength of 325 nm; the PL emission spectra were recorded in the range of 350 nm to 650 nm (Fig. 10). The PL emission spectra showed that the ZnO nanorods had one sharp and strong emission peak located at approximately 380 nm, which is the near-band-edge emission; one broad emission band was also observed in the range of 420 ~ 575 nm, which was emitted green light. However, our result of 380 nm differed from the wavelengths reported by Zheng et al., which ranged from 392nm to 397 nm [17], and Lin et al., which was 390 nm [18], but was similar to the 377 nm reported by Li et al [20]. The energy band gap of emission in the range of 420 ~ 575 nm corresponded to the transition between the local level and the near-band-edge. It formed because of defects in the ZnO nanorods; its emission intensity was smaller than that of the emission peak at 380 nm. There are many intrinsic defects that exist in undoped ZnO-based nanomaterials, including oxygen vacancy ( It is well accepted that the crystal quality of ZnO nanorods is closely associated with the intensity of the near-band-edge emission, and that the emission intensity in the range of 420 ~ 575 nm is closely related to interstitial oxygen and antisite defects [21]. As Fig. 10 shows, the broad emission intensity of the ZnO seed layer at 420 ~ 575 nm was higher than the intensity for the ZnO nanorods at all synthesis times. This result suggested that the ZnO seed layer had a poor crystal quality and contained many different defects, which could explain the results shown in Fig. 3. Therefore, we can understand that the PL property is dependent on the synthesis time by the fact that the crystal quality of ZnO nanorods can be partly improved with time. The emission of green light is caused by interstitial oxygen (544 nm) and antisite (521 nm) defects at the local level in the ZnO seed layer and ZnO nanorods (Fig. 10) [18]. The greater the antisite defect is in the ZnO nanorods, the stronger the emission intensity is for green light in the broad emission band of 420 ~ 575 nm. In general, the maximum emission intensity of green light can be used to judge the quality of ZnO nanorods because it indicates the presence of defects. Figure 10 shows that the emission peak of the broad band for the ZnO seed layer is located at 485 nm, while the emission peak of the broad band for the ZnO nanorods shifted from 485 nm to 540 nm as the synthesis time increased from 10 min to 60 min. These results suggested that the ZnO nanorods repaired the interstitial zinc defect faster than they repaired the antisite defect as they were synthesized on the ZnO seed layer. The two defects decreased with synthesis time, which indicated that the crystal quality of the ZnO nanorods was enhanced with time. Thus, ZnO nanorods with a high optical quality may be successfully obtained when the synthesis time is extended Table 3 shows the emission intensity values of UV (ultraviolet rays) light (I UV ), green light (I G ), and the I G /I UV ratio of the ZnO nanorods as a function of synthesis time. As the synthesis time increased from 10 min to 60 min, the intensity of the I UV value increased from 224 to 742 (a.u.), while the I G value decreased from 150 to 60.7 and the I G /I UV ratio decreased from 0.670 to 0.082. These results proved that the defects in the ZnO nanorods decreased with synthesis time because the crystal quality of the ZnO nanorods improved over time. Li et al. found that an increase in the intensity of UV light was related to the S/V ratio [20]. When the S/V ratio was high, the I UV value was high and the I G value was low. However, the present research showed a different trend; we found that the I UV of the ZnO nanorods was positively related to the total surface area and the total volume, and was negatively related to the S/V ratio. The increase in the S/V ratio was caused by an increase in the total surface area, which provided more area to absorb the excited light and to emit the emission light. Therefore, we believe that the total surface area enhanced the I UV of the ZnO nanorods.
Vanheusden et al. reported that the near-band-edge emission intensity was closely related to the crystal quality of the prepared ZnO samples [21]. Figure 5 shows that the crystal quality of the ZnO nanorods improved with synthesis time; therefore, the PL properties of the ZnO nanorods were dependent on synthesis time. These results suggested that the decrease in the I G values and the increases in the I UV values and the I G /I UV ratio were caused by the enhancement of the crystal quality of the ZnO nanorods.
Thus, the total surface area and the crystal quality of ZnO nanorods are important factors that affect the PL properties.  Figure 11 shows the SEM images of the novel templates at different magni cations, as well as the FIB cross-sectional view. These images showed that the lift-off technology successfully transferred the patterned ZnO seed layer from the sapphire substrate to the p-type Si < 100 > substrate. Figure 11(b) was used to estimate the average particle size and diameter of the top circular plane of the ZnO seed layer, which were determined to be about 32 nm and 365 nm, respectively. The size of the particles were similar to the particle size of the ZnO seed layer that was deposited on the p-type Si < 100 > wafer. The thickness of the ZnO seed layer was obtained from the FIB image after converting its measurement angle, and the measured result was 206 nm, as Fig. 11(c) shows.
The XRD patterns as a function of synthesis time of the ZnO nano ower arrays are shown in Fig. 12 According to the gure, the diffraction peaks of the (100), (002), (101), (102), and (110) planes corresponded to the peaks located at the 2θ values of 31.9°, 34.6°, 36.48°, 47.64°, and 56.78°, respectively. These results proved that the ZnO nano ower arrays had an HCP array and a wurtzite structure. However, there was a large difference between the results shown in Fig. 12 and the results of the ZnO nanorods shown in Fig. 5. The XRD patterns of the ZnO nanorods did not show diffraction peaks of the (102) and (110) planes; the diffraction intensities of the (100) and (101) planes also decreased with synthesis time. In contrast, the diffraction intensities of the ZnO nano ower arrays in ve directions increased with synthesis time. The ZnO nanorods in the nano ower arrays did not push against each other and, thus, the density did not decrease. This result suggested that that the crystal property can be enhanced with synthesis time. The results also indicated that the ZnO nano ower arrays did not have a caxis preferred orientation. The main growth directions of the ZnO nano ower arrays were not perpendicular to the substrate, but rather spread out like a ower, as shown in Fig. 13.
The surface morphologies of the ZnO nano ower arrays are shown in Fig. 13 as a function of synthesis time. The gure shows that the growth direction of the ZnO nanorods was not fully perpendicular to the substrate; they grew in the radial direction. Because the growth of the ZnO nanorods followed the protrusion direction of the array-patterned ZnO seed layers, the nanorods had a ower-like structure in which each petal had a single rod shape. The radial ZnO nanorods exhibited a uniform morphology because they had the same diameter and height. This result was attributed to the effects of strong polarization and high synthesis selectivity, which caused the ZnO nanorods to form a ZnO nano ower in each independent ZnO seed layer. Because the ZnO seed layer was a hexagonal array with a protruding structure in the center of each hexagon, as shown in Fig. 13, the ZnO nanorods grew along the substrate in a vertical direction to form the ZnO nano ower arrays. The ZnO seed layer arrays were more easily formed using the patterned sapphire template. Thus, this technology played a crucial role in synthesizing ZnO nanorods to form ZnO nano ower arrays. For the same synthesis time, the height of the ZnO nano ower arrays was greater than that of the ZnO nanorods. The reason for this observation is that the rod-like ZnO in the ZnO nano ower arrays grew in the radial direction, which was vertical to the protruding structure in the ZnO seed layer. In addition, the density of the ZnO nano ower arrays in a unit area was smaller than the density of the ZnO nanorods. Therefore, the main role of Zn(OH) 2 in the zinc acetate dehydrate solution was to increase the height of ZnO nano ower arrays rather than enhance the diameter.
To calculate the density of the ZnO nano ower arrays in a unit area, the surface area was divided into squares with an area of 1 µm 2 , as shown in Fig. 14(a). We randomly sampled eight squares to calculate the density of ZnO nanorods in each area, and calculated the average value. As Fig. 13 sows, the ZnO nano ower arrays did not fully cover the ZnO seed layer at a synthesis time of 10 min and 20 min. Therefore, the densities were not determined for these times. The densities of the ZnO nanorods were 26 µm − 2 and 28 µm − 2 for a synthesis time of 30 min and 60 min, respectively. Figure 14(b) shows the crosssectional image that was used to calculate the height of the ZnO nano ower arrays. The heights of the nano owers were determined after the thickness of the ZnO seed layer (206 nm) was subtracted. For a synthesis time of 10 min, 20 min, 30 min, and 60 min, the heights of the ZnO nano ower arrays were determined to be 483 nm, 718 nm, 987 nm, and 1500 nm, and the average diameters were 38 nm, 54 nm, 71 nm, and 90 nm, respectively. Correspondingly, the calculated aspect ratios increased from 12.7, 13.3, 13.9, to 16.7 as the synthesis time increased from 10 min, 20 min, 30 min, and 60 min, respectively. Figure 15 shows the PL spectra of the ZnO nanorods and ZnO nano ower arrays as a function of synthesis time. The PL spectra of the ZnO nano ower arrays had one sharp and strong emission peak located at approximately 380 nm; one broad emission band was also observed in the range of 420 ~ 575 nm. The emission intensities of the 380 nm emission peaks increased with synthesis time, while those of the 420 ~ 575 nm broad emission band decreased with time. Table 4 shows the I UV values, the I G values, and the I G /I UV ratio of the ZnO nano ower arrays as a function of synthesis time. As the synthesis time increased from 10 min to 60 min, the intensity of the I UV value increased from 1156 to 2597 (a.u.), while the I G value decreased from 337 to 163 and the I G /I UV ratio decreased from 0.291 to 0.062. For the same synthesis time, the maximum emission intensities of the 380 nm emission peak and the broad emission band of the ZnO nano ower arrays were larger than those of the ZnO nanorods. These results indicated that the I UV and I G values of the ZnO nano ower arrays were larger than those of the ZnO nanorods and, thus, the I G /I UV ratios of the ZnO nano ower arrays were smaller than those of the ZnO nanorods. These results also proved that for the same synthesis time, there were more defects in the ZnO nano ower arrays, as indicated by the I G value, than in the ZnO nanorods. In addition, the I UV and I G values were stronger than those of the ZnO nanorods. A comparison of Fig. 6 and Fig. 13 showed that the top surface of the ZnO nanorods was the main area that received the excitation laser light and emitted the PL spectra; the sides of the ZnO nanorods received almost no excitation laser light and emitted no UV or visible light.
However, laser light can irradiate all surfaces of the ZnO nano ower arrays. Thus, we proved that the ZnO nano ower arrays have better light characteristics than the ZnO nanorods because they have more area to receive light and emit the PL spectra.

Conclusions
As the synthesis time increased, the XRD patterns showed that the ZnO nanorods had a c-axis preferred orientation, and that their main growth rate was along the (002) plane. The ZnO nano ower arrays had 5 diffraction peaks at the (100), (002), (101), (102), and (110) planes, which suggested that the direction of their growth was not fully perpendicular to the substrate. For both the ZnO nanorods and the ZnO nano ower arrays, the average height, average diameter, and aspect ratio increased with time. In addition, the I UV values and the I G /I UV ratio increased with synthesis time, while the I G values decreased with time.
The most important result of this study was that the I UV values of the ZnO nano ower arrays were stronger than the I UV values of the ZnO nanorods. The results in this study also proved that when the hydrothermal method was used with a patterned sapphire template and the lift-off technique to synthesize ZnO nano ower arrays, the PL properties of the ZnO nanorods were enhanced.
Declarations Figure 1 Schematic diagram of the fabrication of ZnO nano ower arrays.  The XRD pattern of the ZnO seed layer on the p-type Si substrate. The XRD patterns of the ZnO nanorods as a function of synthesis time.

Figure 9
Page 21/25 The surface morphology was used to calculate the density of the ZnO nanorods.

Figure 10
The PL emission spectra of the ZnO nanorods as a function of synthesis time.
Page 22/25 Figure 11 SEM images of the novel templates at (a) a low magni cation and (b) a high magni cation on the surface observations; and (c) the FIB cross-sectional view of the template.

Figure 12
The XRD patterns of ZnO nano ower arrays as a function of synthesis time.