Ordered Array of Metal Particles on Semishell Separated with Ultrathin Oxide: Fabrication and SERS Properties

Metal particles in gap cavities provide an interesting system to achieve hybrid local surface plasmon modes for local field enhancement. Here, we demonstrate a relatively simple method to fabricate Ag nanoparticles positioned on Ag semishells separated by a thin (~5 nm) dielectric layer. The obtained structure can provide strong local electric field enhancement for surface-enhanced Raman scattering (SERS). The fabrication of the ordered array structure was realized by nanosphere self-assembly, atomic layer deposition, and metal thin-film dewetting. Numerical simulation proved that, compared to the conventional metal semishell arrays, the additional Ag particles introduce extra hot spots particularly in the valley regions between adjacent Ag semishells. As a result, the SERS enhancement factor of the metal semishell-based plasmonic structure could be further improved by an order of magnitude. The developed novel plasmonic structure also shows good potential for application in plasmon-enhanced solar water-splitting devices.


Introduction
As a powerful spectroscopic technique, surface-enhanced Raman scattering (SERS) has found increased applications in various fields, including analytical chemistry, life sciences, medical sciences, and food safety [1][2][3][4][5][6][7][8][9]. One major challenge in making SERS a general analytical tool lies in the fabrication of noble metal nanostructure substrates with large and reproducible Raman enhancements over a wide sampling area, using ideally less costly, high-throughput methods [4,[10][11][12][13][14][15][16]. Among various nanofabrication techniques for SERS substrates, nanosphere monolayer self-assembly (alternatively referred to as nanosphere lithography) has been widely adopted as a facile, inexpensive, and reliable fabrication method for the production of large-area ordered nanostructures [17][18][19][20][21][22][23][24][25][26]. In particular, metal semishell array substrates have been widely employed in many SERS-based biosensing applications in complex media [27,28]. These semishell array substrates consist of ordered nanogaps between the spheres, which create plasmonic features with good control and tunability compared to random metal structures [29]. Nevertheless, the SERS enhancement factors of these semishell array substrates in previous studies were only around 10 4 -10 5 [29]. This is because in most of these conventional metal array nanostructures, there are only limited number of plasmonic 'hot spots' contributing to the overall SERS signal, and the metal gap size is still generally very large compared to the desirable sub-10 nm regime [29][30][31]. It is therefore desirable to further increase the SERS enhancement through increasing the density of hot spots within the same sample surfaces and/or the local field by engineering the nanostructures.
Based on the well-established nanosphere monolayer assembly, more sophisticated structures have been fabricated in order to increase the enhancement factor, such as structures with particles in cavities [32]. For example, Im et al. found that the SERS of an Ag semishell array can be further improved by an order of magnitude by constructing a nanoring cavity on top, forming sub-10 nm nanogaps [33]. In addition, the particle-in-cavity structures, such as particle-in-void and particle-in-bowl, have been also demonstrated to produce noticeably high-field enhancement in the nanogaps between the metal particles and the metal surface [34,35]. Overall, a further increase of SERS enhancement factors relies essentially on the formation of sub-10 nm plasmonic nanogaps [12]. To meet this challenge, atomic layer deposition (ALD) has shown to be capable of conformal and uniform coating of thin films of atomic thickness on 3D surfaces and has been widely employed in nanofabrication and surface engineering of photonics and energy-conversion materials, such as the ALD-fabricated 3D inverse opal structure photoanodes for photoelectron chemical water splitting [36,37]. Recently, ALD has also been proven useful in the fabrication of novel plasmonic structures, for example, an ultrathin isolated dielectric layer on Au particles for ultra-sensitive SERS [38] and nanogaps in plasmonic structures [39,40].
In this communication, we demonstrate a facile and scalable fabrication process based on nanosphere monolayer assembly and atomic layer deposition to achieve an array structure with metal particles on semishell (PoSS) for SERS application. The metal particles were formed via a simple thermal-induced dewetting of thin film. The local field enhancement of the PoSS structure was verified by both simulation and SERS measurements. This subtle design of the PoSS array generated further SERS enhancement up to 10 times that of conventional metal semishell arrays (although still not higher than those of other SERS structures obtained by less scalable or less reproducible fabrication methods). Our results demonstrate the usefulness of nanosphere self-assembly and ALD in fabricating plasmonic nanostructures for applications in biomedical sciences and surface plasmon-enhanced solar water-splitting devices.

Fabrication of the Metal Semishell (PoSS) Array
The fabrication process for the PoSS array is illustrated in Figure 1. First, a~100 nm-thick Ag layer was deposited on the hollow TiO 2 array by an Edward electron-beam evaporation system, as shown in Figure 1a. The fabrication process of the hollow TiO 2 nanosphere array has been reported elsewhere [31], and its steps include self-assembly of monolayer polystyrene (PS) spheres, reactive ion etching, ALD TiO 2 , and thermal annealing (450 • C for one hour). PS nanospheres were purchased from Polysciences, Inc., Warrington, FL, USA. The concentration of the PS used in this study was 2.61% (w/w), and the size distribution of the PS was 505 ± 10 nm. After Ag deposition, a 5 nm TiO 2 layer was coated on the Ag films using ALD, as shown in Figure 1b,c. ALD was performed on a Beneq TFS-200 system (Espoo, Finland) using TiCl 4 and deionized water as Ti and oxygen precursors, respectively. High-purity N 2 was used as the process gas in our experiment. During the deposition, the reaction chamber was maintained at 1.0 mbar with a steady N 2 steam at 200 sccm. Each ALD cycle consisted of a 250 ms precursor pulse and 5 s purging time with N 2 , and 60 cycles were used to deposit a 5 nm-thick TiO 2 . The deposition temperature was set to a low temperature of 60 • C for two reasons: to avoid oxidation of the Ag film and to simultaneously reduce the crystallization and conductivity of the TiO 2 layer. Finally, a thin Ag layer of variable thickness was sputtered (JFC-1600, JEOL, Tokyo, Japan) followed by thermal annealing at 120 • C in an Ar atmosphere for 30 min. The final heating process converted the top Ag thin film to individual particles via dewetting, as shown in Figure 1d. The coverage of Ag particles was controlled by the sputtering time at a fixed current of 10 mA.

Characterization
The morphology of the PoSS array was characterized by JEOL JSM-6700F field emission scanning electron microscope (FE-SEM). Diffused reflectance spectra were recorded by a Zolix Solar Cell QE/IPCE measurement system (Zolix Instruments Co., Ltd., Beijing, China) equipped with an integrated sphere and a Si diode. To avoid oxidation of Ag, fresh samples were used for SERS measurements.
For SERS measurements, R6G was utilized as the standard probe molecule. The substrates were immersed in the prepared R6G solution at the concentration of 0.5 μM for 30 min, then washed with deionized water, and dried by N2 flow before SERS measurements. The Raman spectra were recorded using a Renishaw Raman Imaging Microscope (Wotton-under-Edge, UK) configured with a chargecoupled device (CCD) array detector and a 50× short focal length objective. A 532 nm laser with the output power of 25 mW was used. The incident power on the sample surface was reduced to 1% of the output. The laser focal spot on the sample surface was ~1 μm in diameter, and the focus depth was ~20 μm. The SERS spectra were collected in the wave number range of 1000-2000 cm −1 with fixed acquisition time of 10 s. Each spectrum shown in this paper is the average of five spectra measured from different locations on the same sample surface.

Results
Since ALD is a cyclic layer-by-layer deposition method, it is capable of conformal and uniform coating of thin films on 3D surfaces. In our study, the ALD TiO2 layer covered the whole surface of the Ag semishell array structure with a controllable thickness down to a few nanometers. The TiO2 layer has two functions: firstly, it serves as a dielectric separation layer of the Ag particles from the Ag film; secondly, it protects the Ag semishells from distortion during the final thermal annealing process. The size of the coated particles can be tuned by the sputtering time, which will be further discussed later. While the metal demonstrated herein used Ag, the fabrication and main conclusion are essentially the same for gold. Figure 2a shows a scanning electron microscope (SEM) image of the fabricated hexagonal closepacked Ag semishell array with PS size of (505 ± 10) nm. The final Ag PoSS structures are shown in Figure  2b-d. The Ag semishells underneath were well maintained during the thermal annealing process at 120 °C, otherwise discontinuous islands would have been observed. The size of the particles located in the valley region between adjacent spheres should be slightly smaller than that of the ones located at the top, due to the fact the Ag film deposited laterally on the spheres' surface could be thinner than that deposited on top, as a consequence of the so-called shadowing effect.

Characterization
The morphology of the PoSS array was characterized by JEOL JSM-6700F field emission scanning electron microscope (FE-SEM). Diffused reflectance spectra were recorded by a Zolix Solar Cell QE/IPCE measurement system (Zolix Instruments Co., Ltd., Beijing, China) equipped with an integrated sphere and a Si diode. To avoid oxidation of Ag, fresh samples were used for SERS measurements.
For SERS measurements, R6G was utilized as the standard probe molecule. The substrates were immersed in the prepared R6G solution at the concentration of 0.5 µM for 30 min, then washed with deionized water, and dried by N 2 flow before SERS measurements. The Raman spectra were recorded using a Renishaw Raman Imaging Microscope (Wotton-under-Edge, UK) configured with a charge-coupled device (CCD) array detector and a 50× short focal length objective. A 532 nm laser with the output power of 25 mW was used. The incident power on the sample surface was reduced to 1% of the output. The laser focal spot on the sample surface was~1 µm in diameter, and the focus depth was~20 µm. The SERS spectra were collected in the wave number range of 1000-2000 cm −1 with fixed acquisition time of 10 s. Each spectrum shown in this paper is the average of five spectra measured from different locations on the same sample surface.

Results
Since ALD is a cyclic layer-by-layer deposition method, it is capable of conformal and uniform coating of thin films on 3D surfaces. In our study, the ALD TiO 2 layer covered the whole surface of the Ag semishell array structure with a controllable thickness down to a few nanometers. The TiO 2 layer has two functions: firstly, it serves as a dielectric separation layer of the Ag particles from the Ag film; secondly, it protects the Ag semishells from distortion during the final thermal annealing process. The size of the coated particles can be tuned by the sputtering time, which will be further discussed later. While the metal demonstrated herein used Ag, the fabrication and main conclusion are essentially the same for gold. Figure 2a shows a scanning electron microscope (SEM) image of the fabricated hexagonal close-packed Ag semishell array with PS size of (505 ± 10) nm. The final Ag PoSS structures are shown in Figure 2b-d. The Ag semishells underneath were well maintained during the thermal annealing process at 120 • C, otherwise discontinuous islands would have been observed. The size of the particles located in the valley region between adjacent spheres should be slightly smaller than that of the ones Coatings 2019, 9, 20 4 of 10 located at the top, due to the fact the Ag film deposited laterally on the spheres' surface could be thinner than that deposited on top, as a consequence of the so-called shadowing effect. Firstly, a numerical estimation of the electromagnetic field distribution in the PoSS structure was carried out using a 3D finite-difference time-domain (FDTD) simulation (FDTD Solution 8.5, Lumerical Inc., Vancouver, BC, Canada). For simplicity, the Ag particles were treated as spheres. The outer diameter of the TiO2 hollow spheres in the simulations was 490 nm, and the shell thickness was 15 nm. The distance of the two TiO2 hollow spheres was set to 505 nm. The ALD TiO2 layer between the Ag particle and Ag semishell was set to 5 nm. The Ag particle size was chosen to be 30 nm, and for close-view simulation, we set two different particles with diameters of 30 and 16 nm. Because of the size distribution and the random location of the Ag nanoparticles on the semishell, the periodic boundary condition of the original Ag semishell array was broken. Moreover, as the Ag particle size (30-16 nm) was relatively small compared to the semishells (~500 nm), the PoSS structure could be treated as the well-known 'metal-particle-on-metal-film' system [41]. Thus, a dimer structure was chosen as the unit structure in the simulation. The Yee cell in our calculation was set to 1.0 nm × 1.0 nm × 1.0 nm [35]. These settings are sufficient for characterizing the features of the given geometry. The dielectric constant of Ag was obtained from Johnson and Christy [42], and that for TiO2 from Palik ('Handbook of Optical Constants of Solids'). The particle in a uniform dielectric medium (n = 1) was illuminated from the top with a 532 nm linear polarized plane wave, as shown in Figure 3a. Figure 3a,b shows the simulation result of the structure from the side view and top view, respectively. The simulation results show that the PoSS structure generated localized hybrid surface plasmons and substantial electromagnetic field enhancement between the Ag particles and the Ag film. These gaps are known as plasmonic hot spots for SERS, provided that they are accessible by analytic molecules. For the commonly studied metal semishells, however, the hot spots are located merely at the edge of the shells between adjacent spheres (see Figure 3d). After particle coating, each Ag particle generated one more hot spot. This means the density of hot spots significantly increased in the PoSS array structure. Another important feature of the PoSS structure is the drastic field enhancement in the valley region. Let us take a closer inspection and compare it to control structures of pure Ag particles without an Ag film underneath as well as to pure Ag semishells. As seen in Figure 3c-e, because of the strong coupling between Ag particles and Ag semishells, the addition of Ag particles, produced new electric fields stronger than those produced by the bare Ag semishell array and Ag particles directly on TiO2 spheres. Although the largest field enhancement was found to be at the touching area between the Ag particles and the TiO2 interlayer (Figure 3e), the non-touching side (indicated with yellow arrows), which had a molecule-accessible surface, also exhibited strong fields. It is believed that these open hot spots contributed dominantly to the SERS, as other hot spots beneath the touching points were not accessible to analytic molecules. Firstly, a numerical estimation of the electromagnetic field distribution in the PoSS structure was carried out using a 3D finite-difference time-domain (FDTD) simulation (FDTD Solution 8.5, Lumerical Inc., Vancouver, BC, Canada). For simplicity, the Ag particles were treated as spheres. The outer diameter of the TiO 2 hollow spheres in the simulations was 490 nm, and the shell thickness was 15 nm. The distance of the two TiO 2 hollow spheres was set to 505 nm. The ALD TiO 2 layer between the Ag particle and Ag semishell was set to 5 nm. The Ag particle size was chosen to be 30 nm, and for close-view simulation, we set two different particles with diameters of 30 and 16 nm. Because of the size distribution and the random location of the Ag nanoparticles on the semishell, the periodic boundary condition of the original Ag semishell array was broken. Moreover, as the Ag particle size (30-16 nm) was relatively small compared to the semishells (~500 nm), the PoSS structure could be treated as the well-known 'metal-particle-on-metal-film' system [41]. Thus, a dimer structure was chosen as the unit structure in the simulation. The Yee cell in our calculation was set to 1.0 nm × 1.0 nm × 1.0 nm [35]. These settings are sufficient for characterizing the features of the given geometry. The dielectric constant of Ag was obtained from Johnson and Christy [42], and that for TiO 2 from Palik ('Handbook of Optical Constants of Solids'). The particle in a uniform dielectric medium (n = 1) was illuminated from the top with a 532 nm linear polarized plane wave, as shown in Figure 3a. Figure 3a,b shows the simulation result of the structure from the side view and top view, respectively. The simulation results show that the PoSS structure generated localized hybrid surface plasmons and substantial electromagnetic field enhancement between the Ag particles and the Ag film. These gaps are known as plasmonic hot spots for SERS, provided that they are accessible by analytic molecules. For the commonly studied metal semishells, however, the hot spots are located merely at the edge of the shells between adjacent spheres (see Figure 3d). After particle coating, each Ag particle generated one more hot spot. This means the density of hot spots significantly increased in the PoSS array structure. Another important feature of the PoSS structure is the drastic field enhancement in the valley region. Let us take a closer inspection and compare it to control structures of pure Ag particles without an Ag film underneath as well as to pure Ag semishells. As seen in Figure 3c-e, because of the strong coupling between Ag particles and Ag semishells, the addition of Ag particles, produced new electric fields stronger than those produced by the bare Ag semishell array and Ag particles directly on TiO 2 spheres. Although the largest field enhancement was found to be at the touching area between  (Figure 3e), the non-touching side (indicated with yellow arrows), which had a molecule-accessible surface, also exhibited strong fields. It is believed that these open hot spots contributed dominantly to the SERS, as other hot spots beneath the touching points were not accessible to analytic molecules. The overall field enhancement could be characterized by SERS measurements. Figure 4 shows the Raman spectra collected from Ag PoSS arrays obtained from 25 s Ag deposition (blue curve) and those from the bare Ag semishell array (red line) and the Ag particles on TiO2 spheres without Ag semishells (black line) for comparison. One can see that the Raman signal from the Ag PoSS array showed the strongest SERS signal, which was nearly 10 times that from the bare Ag semishell array. For the Ag particles coated on TiO2 hollow spheres, the signal intensity was rather low. This can be explained by the small local field enhancement in the structure, as revealed by simulation in Figure 3c. In addition, the relatively large separation between individual Ag particles (as shown in SEM images) made the coupling between localized surface plasmon (LSP) modes of the particles negligible.
The Ag nanoparticle size and the SERS enhancement of the PoSS array could be tuned by modulating the Ag sputtering time. Figure 5 shows the SEM images of the Ag PoSS structure fabricated using different Ag sputtering times. When the sputtering time increased from 15 to 40 s, the Ag particle size increased from ~8 to 30 nm. Diffuse reflection spectra of the samples with different Ag deposition times are shown in Figure 5d. For comparison, the spectrum from the bare Ag semishells on TiO2 hollow sphere arrays is also shown. The absorption edge at 380 nm observed for all samples corresponds to the bulk value (3.2 eV) of the TiO2 band gap. The Ag semishell array substrate exhibited two broad localized surface plasmon resonance (LSPR) peaks, located at ~480 and ~630 nm, which correspond to the axial and transverse modes of the Ag semishell, respectively. This result is consistent with previous reports on the Ag semishell array [29,33]. The minimum reflectance at ~560 nm (indicated by the red star) is due to the hexagonal close-packed structure [31]. The reflection spectrum of the final PoSS structure with 40 s sputtering time showed two plasmon bands at ~520 and ~620 nm, indicated with black arrows in Figure 5d. These two peaks could be due to the plasmon coupling between the particle and the semishell. As the Ag particle size (50 nm) is relatively small compared to the semishells (~500 nm), the PoSS structure can be regarded as the well-known 'metal-particle-on-metal-film' system. In such a particle-on-film system, it is known that there exist two plasmon modes, the horizontal and the vertical modes, as previously described [43,44]. The vertical mode has lower resonance energy than the horizontal mode. The s-and p-polarized incident lights excite the The overall field enhancement could be characterized by SERS measurements. Figure 4 shows the Raman spectra collected from Ag PoSS arrays obtained from 25 s Ag deposition (blue curve) and those from the bare Ag semishell array (red line) and the Ag particles on TiO 2 spheres without Ag semishells (black line) for comparison. One can see that the Raman signal from the Ag PoSS array showed the strongest SERS signal, which was nearly 10 times that from the bare Ag semishell array. For the Ag particles coated on TiO 2 hollow spheres, the signal intensity was rather low. This can be explained by the small local field enhancement in the structure, as revealed by simulation in Figure 3c. In addition, the relatively large separation between individual Ag particles (as shown in SEM images) made the coupling between localized surface plasmon (LSP) modes of the particles negligible.
The Ag nanoparticle size and the SERS enhancement of the PoSS array could be tuned by modulating the Ag sputtering time. Figure 5 shows the SEM images of the Ag PoSS structure fabricated using different Ag sputtering times. When the sputtering time increased from 15 to 40 s, the Ag particle size increased from~8 to 30 nm. Diffuse reflection spectra of the samples with different Ag deposition times are shown in Figure 5d. For comparison, the spectrum from the bare Ag semishells on TiO 2 hollow sphere arrays is also shown. The absorption edge at 380 nm observed for all samples corresponds to the bulk value (3.2 eV) of the TiO 2 band gap. The Ag semishell array substrate exhibited two broad localized surface plasmon resonance (LSPR) peaks, located at~480 and 630 nm, which correspond to the axial and transverse modes of the Ag semishell, respectively. This result is consistent with previous reports on the Ag semishell array [29,33]. The minimum reflectance at~560 nm (indicated by the red star) is due to the hexagonal close-packed structure [31]. The reflection spectrum of the final PoSS structure with 40 s sputtering time showed two plasmon bands at~520 and~620 nm, indicated with black arrows in Figure 5d. These two peaks could be due to the plasmon coupling between the particle and the semishell. As the Ag particle size (50 nm) is relatively small compared to the semishells (~500 nm), the PoSS structure can be regarded as the well-known 'metal-particle-on-metal-film' system. In such a particle-on-film system, it is known that there exist two plasmon modes, the horizontal and the vertical modes, as previously described [43,44]. The vertical mode has lower resonance energy than the horizontal mode. The s-and p-polarized incident lights excite the horizontal and vertical modes, respectively [43]. In our case, as the Ag particles were anchored on a curved semishell surface, both horizontal and vertical modes could be excited under a normal incident condition. With a decrease in the Ag sputtering time, the corresponding plasmon peaks blueshifted because of the reduction of Ag particle size (as seen in Figure 5a-c). Moreover, because the Ag PoSS structure was fabricated on the basis of a hexagonal close-packed PS monolayer template, it had a 60 • rotational symmetry. Thus, the polarization effect would be only within 30 • as a result of the rotational symmetry. Because of these two factors, the polarization effect was greatly weakened in our SERS measurements. incident condition. With a decrease in the Ag sputtering time, the corresponding plasmon peaks blueshifted because of the reduction of Ag particle size (as seen in Figure 5a-c). Moreover, because the Ag PoSS structure was fabricated on the basis of a hexagonal close-packed PS monolayer template, it had a 60° rotational symmetry. Thus, the polarization effect would be only within 30° as a result of the rotational symmetry. Because of these two factors, the polarization effect was greatly weakened in our SERS measurements.   incident condition. With a decrease in the Ag sputtering time, the corresponding plasmon peaks blueshifted because of the reduction of Ag particle size (as seen in Figure 5a-c). Moreover, because the Ag PoSS structure was fabricated on the basis of a hexagonal close-packed PS monolayer template, it had a 60° rotational symmetry. Thus, the polarization effect would be only within 30° as a result of the rotational symmetry. Because of these two factors, the polarization effect was greatly weakened in our SERS measurements.   The size of the particles as well as the LSPR resonance frequency affect the strength of the local field. For the three PoSS structures, it is hypothesized that the drastically low Raman intensity of the 15 s sample originated from the blueshift of the LSPR wavelength away from the excitation wavelength (532 nm), as seen in Figure 5d. The slight difference between the 25 and 40 s sample can be ascribed to a tradeoff between particle size and number of open hot spots within the valley region.
Finally, we estimated the SERS enhancement factor (EF). EF characterizes the sensitivity of a SERS substrate, which is defined by the ratio of the intensity of the SERS signal to that of the normal Raman signal (i.e., not surface-enhanced) or the non-SERS signal [45].
where N 0 and N SERS are the number of probe molecules that contribute to the non-SERS and SERS signals, respectively, and I 0 and I SERS are the intensities of the selected scattering bands in the non-SERS and SERS spectra, respectively. The detected molecules were assumed to have been distributed uniformly on the substrates. Thus, the number of the detected molecules could be estimated by N = (N A CV solution /S sub )S laser (2) where N A is the Avogadro constant, C is the molar concentration of the prepared solution, V solution is the volume of the solution used, S sub is the size of the substrate, and S laser is the size of the laser spot.
To prepare the SERS sample, 2 µL of 0.5 µM R6G solution was dropped on the Ag PoSS structure (Ag sputtering time 25 s) substrate, with a drop diameter of approximately 4 mm. For the reference, 2 µL of 10 mM R6G solution was dropped on a single crystal gold foil for the non-SERS measurements, with a drop diameter of approximately 4 mm. Since the samples for SERS and non-SERS detection were prepared by the same method, the EF could be estimated by the following Equation.
EF =(I SERS /C SERS )/(I 0 /C 0 ) where C SERS and C 0 are the molar concentrations of R6G molecules in the aqueous solutions used for the SERS and non-SERS samples, respectively. A typical comparison of the spectra is shown in Figure 6a. The characteristic band at 1360 cm −1 was chosen for the EF estimation. In this way, the average EF could be as high as~2.5 × 10 6 for the PoSS structure (Ag sputtering time 25 s), nearly 10 times that of the bare Ag semishell array (~2.8 × 10 5 ). In our previous work on Au semishells, a SERS EF of 1.5 × 10 5 was also obtained, which was used to detect 10 −9 M target solution. [31] Since the EF of Ag PoSS SERS substrate was almost 20 times that of the Au semishell array SERS substrate, the as-fabricated PoSS substrate could potentially be used in the detection of a 10 −10 M target molecular solution. It is noted that while this EF is not the highest ever obtained, the main intention of this work was to demonstrate a facile fabrication method of a homogeneous and reproducible SERS structure, for comparison with the solution synthesis of metal nanoparticles or individual nanogaps which can exhibit EFs up to 10 9 . The reproducibility of the SERS signal is another important parameter for the Ag PoSS structure as a SERS substrate. To test the reproducibility of the Ag PoSS SERS substrate, the SERS spectra were collected from 15 randomly selected positions, as shown in Figure 6b. For the peak at 1360 cm −1 , the relative standard deviation (RSD) of the SERS intensity was approximately 11.5%. This low RSD indicated that the Ag PoSS SERS substrates had an excellent reproducibility.
Coatings 2019, 9, 20 8 of 10 the solution synthesis of metal nanoparticles or individual nanogaps which can exhibit EFs up to 10 9 . The reproducibility of the SERS signal is another important parameter for the Ag PoSS structure as a SERS substrate. To test the reproducibility of the Ag PoSS SERS substrate, the SERS spectra were collected from 15 randomly selected positions, as shown in Figure 6b. For the peak at 1360 cm −1 , the relative standard deviation (RSD) of the SERS intensity was approximately 11.5%. This low RSD indicated that the Ag PoSS SERS substrates had an excellent reproducibility.

Conclusions
Assisted by PS sphere self-assembly and ALD, we have demonstrated a controllable fabrication of novel ordered arrays of Ag particles on semishells (PoSS) with a reasonably high SERS effect. The SERS enhancement factor of the PoSS array showed an increase up to 10 times as compared to the commonly studied bare metal semishell arrays. The FDTD simulation revealed that the further SERS enhancement was due to the increase of the local field and the density of plasmonic hot spots in the valley region. The large SERS enhancement of such metal PoSS arrays and their facile, reproducible fabrication process suggest their potential application as functional components in chemical sensors and biosensors. This novel PoSS plasmonic structure can also find application in solar energy conversion, such as plasmon-enhanced solar water-splitting devices.