Dielectric Nanoparticles Coated upon Silver Hollow Nanosphere as an Integrated Design to Reinforce SERS Detection of Trace Ampicillin in Milk Solution

: Surface-enhanced Raman scattering (SERS) technique is competent to trace detection of target species, down to the single molecule level. The detection sensitivity is presumably degraded by the presence of non-speciﬁc binding molecules that occupy a SERS-active site (or hot spot) on the substrate surface. In this study, a silver hollow nano-sphere (Ag HNS) with cavity has been particularly designed, followed by depositing dielectric nanoparticles (Di NPs) upon Ag HNS. In the integrated nanostructures, Di NPs / Ag HNS were furthermore conﬁrmed by cutting through the cross sections using the Focused Ion Beam (FIB) technique, which provides the Scanning Electron Microscope (SEM) with Energy-dispersive Spectroscope (EDS) mode for identifying the distribution of Di NPs upon Ag HNS. The results have indicated that Di NPs / Ag HNS exhibits small diameter of cavity, and among Di NPs in this study, Al 2 O 3 with lower dielectric constant provides a much higher SERS enhancement factor (e.g., ~6.2 × 10 7 ). In this study, to detect trace amounts (e.g., 0.01 ppm) of Ampicillin in water or milk solution, Al 2 O 3 NPs / Ag HNS was found to be more e ﬃ cient and less inﬂuenced by non-speciﬁc binding molecules in milk. A substrate with integrated plasmonic and dielectric components was designed to increase the adsorption of target species and to repulse non-speciﬁc binding molecules from SERS-active sites. with the synergy from the characteristic of Ag as a SERS-active surface, an enhanced plasmonic e ﬀ ect inside Ag shell, and the contribution of dielectric property at Ag shell and Di NPs interface when a Raman laser with an appropriate wavelength is induced. The results demonstrate that a cavity with small diameter and Di NPs with low dielectric constant will result in a higher EF. In a practical application for trace detection of ampicillin in water, SERS intensities increase with the addition of ampicillin for all these three substrates, while Al 2 O 3 / Ag HNS exhibits sensitivity with the increased concentration of ampicillin, in consistence with its low dielectric constant. For the case of detecting ampicillin in milk solution, Di NPs / Ag HNS is competent to achieve trace detection of Ampicillin as low as 0.01 ppm. However, for detecting Ampicillin on the respective Al 2 O 3 / , HfO 2 / , and TiO 2 / Ag HNS in milk solution, SERS intensity at the main peak of 1007 cm − 1 decrease up to 60.3% with respect to that measured in water. The presence of NSB molecules tends to lengthen the distance of Ampicillin with SERS-active sites and thereafter relax the e ﬀ ect of SERS for its di ﬀ erent characteristic peaks. A possible SERS mechanism of Di NPs / Ag HNS for trace detection of Ampicillin in water and in milk solution is proposed: when a Raman laser with an appropriate wavelength is induced, the main electromagnetic e ﬀ ect is particularly resulted at the interface Di NPs and Ag shells where electron transfers the contacts Di NPs and Ag surface distribution “hot spots” along Ag shells with cavities. For trace detection milk solution, NSB molecules, prior NPs / Ag HNS, may signiﬁcantly reduce the measured SERS intensity of ampicillin upon SERS-active sites. An e ﬀ ective method to shorten the distance between target molecules and hot-spot or to reduce NSB molecules is therefore


Introduction
The detection of contaminants or hazardous substances by a precise, sensitive-to-target species, and convenient device is of great interest for the fields such as health care, food preparation, and environmental sampling [1][2][3]. Among them, ampicillin residue in food and to the environment provokes great public concern since it could pose potential health risks not only to the consumers but also the Earth [4,5]. Various analytical methods, including gas chromatography [6,7] and high-performance liquid chromatography [8], have been applied for trace detection of ampicillin in a solution. However, these approaches usually need expensive laboratory facilities and involve time-consuming and costly sample preparation [9]. Thus, alternative research has been conducted to develop advanced detection

Fabrications of PS, Ag HNS, and Di NPs/Ag HNS
Uniform polystyrene beads (PS_Bs), with a diameter of 130, 240, and 260 nm that were synthesized through surfactant-free emulsion polymerization, were first deposited onto clean 2 × 2 cm 2 Si substrates, by drawing out the Si substrates that were briefly immersed in PS_Bs suspended in water, as illustrated in Figure 1a(i-v). The PS-coated Si was then subjected to slight annealing at 50 • C for 10 min [44], ensuring that all the solvents have evaporated and that the PS_Bs are fixed onto the surface of the Si. As shown in Figure 1a(vi), Ag thin film was then deposited onto PS_Bs/Si by an e-beam evaporator (VT1-10CE, ULVAC, Chigasaki, Japan), followed by annealing at 400 • C for 45 min to remove the PS_Bs. The heat-treated Ag thin film without PS_Bs then became a crystalline Ag shell with a cavity; the as-formed structure is described as Ag HNS. Based on the sizes of the as-prepared PS_Bs, 130, 240, and 260 nm, three Ag HNSs: Ag HNS_130, _240, and _260 were employed, as shown in Figure 1a(i-vii). Three different Di NPs: HfO 2 , TiO 2 , and Al 2 O 3 , were then deposited on Ag HNS through an electron beam evaporator (VT1-10CE, ULVAC Inc., Tokyo, Japan) with a thickness of~1.5 nm and at a deposition rate of 0.1 Å/s. Three Di NPs/Ag HNSs were fabricated: HfO 2 /, TiO 2 /, and Al 2 O 3 /Ag HNS, were used for the study, as shown in Figure 1b(i,ii). The Di NPs were loosely attached upon Ag HNS, which results in a broad distribution over the SERS-active surface. It is conjecture that if there is a film-like Di NPs upon Ag HNS, the incident laser tends to interact with only Di NPs, rather than with the inclusion of Di NPs/Ag interface and Ag HNS, the effect of SERS is presumably reduced.

Fabrications of PS, Ag HNS, and Di NPs/Ag HNS
Uniform polystyrene beads (PS_Bs), with a diameter of 130, 240, and 260 nm that were synthesized through surfactant-free emulsion polymerization, were first deposited onto clean 2 × 2 cm 2 Si substrates, by drawing out the Si substrates that were briefly immersed in PS_Bs suspended in water, as illustrated in Figure 1a(i-v). The PS-coated Si was then subjected to slight annealing at 50 °C for 10 min [44], ensuring that all the solvents have evaporated and that the PS_Bs are fixed onto the surface of the Si. As shown in Figure 1a(vi), Ag thin film was then deposited onto PS_Bs/Si by an e-beam evaporator (VT1-10CE, ULVAC, Chigasaki, Japan), followed by annealing at 400 °C for 45 min to remove the PS_Bs. The heat-treated Ag thin film without PS_Bs then became a crystalline Ag shell with a cavity; the as-formed structure is described as Ag HNS. Based on the sizes of the as-prepared PS_Bs, 130, 240, and 260 nm, three Ag HNSs: Ag HNS_130, _240, and _260 were employed, as shown in Figure 1a(i-vii). Three different Di NPs: HfO2, TiO2, and Al2O3, were then deposited on Ag HNS through an electron beam evaporator (VT1-10CE, ULVAC Inc., Tokyo, Japan) with a thickness of ~1.5 nm and at a deposition rate of 0.1 Å/s. Three Di NPs/Ag HNSs were fabricated: HfO2/, TiO2/, and Al2O3/Ag HNS, were used for the study, as shown in Figure 1b(i,ii). The Di NPs were loosely attached upon Ag HNS, which results in a broad distribution over the SERS-active surface. It is conjecture that if there is a film-like Di NPs upon Ag HNS, the incident laser tends to interact with only Di NPs, rather than with the inclusion of Di NPs/Ag interface and Ag HNS, the effect of SERS is presumably reduced. Figure 1. (a) Ag HNS template fabrication procedures: (i) PS_Bs suspension is pipetted onto an angled glass substrate and moved onto water/air interface; (ii) PS_Bs start assembling on water/air interface; (iii) as more suspensions join PS_Bs monolayer, a full coverage over the interface is obtained; (iv) a plate is then entered a water phase at a shallow angle of ∼10° to transfer the monolayer; (v) PS_Bs of e. g., 130 nm are formed as a monolayer on Si substrate by the growth-at-interface method; (vi) subsequently e-beam evaporation of ~20 nm Ag NPs is deposited on PS_Bs monolayer, and (vii) followed by heat treatment (400 °C, 45 min) to release PS_Bs, Ag HNS is thereafter formed. (b) An Figure 1. (a) Ag HNS template fabrication procedures: (i) PS_Bs suspension is pipetted onto an angled glass substrate and moved onto water/air interface; (ii) PS_Bs start assembling on water/air interface; (iii) as more suspensions join PS_Bs monolayer, a full coverage over the interface is obtained; (iv) a plate is then entered a water phase at a shallow angle of ∼10 • to transfer the monolayer; (v) PS_Bs of e.g., 130 nm are formed as a monolayer on Si substrate by the growth-at-interface method; (vi) subsequently e-beam evaporation of~20 nm Ag NPs is deposited on PS_Bs monolayer, and (vii) followed by heat treatment (400 • C, 45 min) to release PS_Bs, Ag HNS is thereafter formed. (b) An optimization of Di NPs upon Ag HNS: (i) E-beam evaporation of Di NPs; (ii) three as formed samples: HfO 2 /, TiO 2 /, and Al 2 O 3 /Ag HNS; (iii) the test molecule, R6G for the evaluation of EFs.

Structural and Morphological Characterization
The surfaces of Di NPs/Ag HNS were characterized using high-resolution thermal FESEM (JSM-7000, JEOL, Tokyo, Japan), which was operated at an accelerating voltage of 10 kV. The compositions on the surface of Di NPs/Ag HNS were analyzed by a focused module of EDS. The cross-sectioning and imaging of Ag HNS were performed by a dual-focused ion and electron beam FIB/SEM (FEI Nova-200 NanoLab Compatible, Hillsboro, OR, USA) system.

SERS Property of Ag HNS and Di NPs/Ag HNS
Experiments in evaluating SERS properties were done with 5 replicates; as-prepared Ag HNS and Di NPs/Ag HNS were examined using Raman spectrometer (Renishaw, Wotton-under-Edge, UK) with an incident power of 3 mW and an air-cooled charge-coupled device as the detector. An amount of 2 µL of R6G was dropped onto the substrate and left to dry in air at room temperature before acquiring the Raman spectra at an integration time of 30 s.
Raman spectrometer is operated with the 50× objective of optical microscope, the spot size of Raman laser at a diameter of 1 µm, and the diode and He-Ne lasers at excitation wavelengths of 633 nm with gratings at 1800 lines/nm. The calibration of the spectrum was done first before taking any measurements with the use of a standard Si and at a laser power set at 100%. The obtained Raman spectra were processed after adjusting background fluorescent signals and noise that may be induced by the use of an objective lens to focus the laser spot. Baseline correction and smoothing were performed in conjunction with each other to obtain accurately defined signals.

Trace Detection of Ampicillin in Water and in Milk Solution
The Di NPs/Ag HNS_130 was employed for trace detection of ampicillin in water and in milk solution. An appropriate amount of ampicillin (EMD Millipore Corp., Temecula, CA, USA) was added to 10 mL of deionized water to prepare standard solutions with the concentrations of 100, 1, and 0.01 ppm for the studies. Milk solution was prepared as follows: 10 mL of commercially-available pasteurized milk (Uni-President Enterprises Corp., Tainan, Taiwan) was added to 500 mL of distilled water under vigorous stirring at room temperature to obtain a homogeneous solution. Note that the "raw milk" was commercially processed, ready for consumption, and thus pasteurized. This milk referred to it being "raw" as it was not mixed with anything else before diluting it with water or adding the sulfuric acid to remove the lipids. A lipid-removal procedure was then performed and lipids were pre-aggregated in milk by adding 5 µL sulfuric acid per 1 mL milk solution at room temperature, at which point the lipids were aggregated, allowing the solids to be easily removed using a 0.2-µm syringe filter, yielding a clear filtrate. Ampicillin was then added to the as-prepared homogeneous milk solution at a concentration of 0.01 ppm.

Enhancement Factor Evaluation and Calculation
To assess the SERS effect of the substrate, a normal Raman scattering is required as a benchmark for comparative analysis, in which each substrate could lead to a distinguished effect. The quantification effect value is called the EF and is calculated in the Formula (1): where N SERS and N Bulk are the numbers of probe molecules contributing to the SERS and non-SERS signal, respectively; and I SERS and I Bulk are the intensities of the selected scattering bands in the SERS and non-SERS spectra, respectively. The probed molecules are assumed to distribute uniformly on the substrates. The number of probe molecules located in the focus volume within the bulk sample N Bulk can be estimated by Formula (2): where A ls is the laser spot area (diameter: 1 µm); h is the focus length (19 µm); ρ is the density of solid R6G (1.26 g/cm 3 ); and m is molecular weight of R6G (479 g/mole). The number of R6G molecules adsorbed on the SERS-active substrate surface N SERS can be estimated by Formula (3): where A c , A sc , and A m represent the area of a circle with a diameter equal with that of one Ag HNS, the area of a spherical calotte, and the area occupied by a single R6G molecule (59.9 Å 2 ), respectively.

Size and Dimension of Ag HNS
In Figure 2a-f, surface morphologies of the as-prepared Ag and Ag HNSs with different sizes and dimensions were verified using Field Emission SEM (FESEM). PS_Bs of 130, 240, and 260 nm in diameter were successfully prepared and employed as the templates to assist the formation of the substrates. Ag_x and Ag HNS_x denote the substrates before and after the PS_Bs template removal; x represents the diameter of PS_Bs template used, which in the cases of this study are 130, 240, and 260 nm. Three types of substrates with cavities, according to PS_Bs template sizes, were thus produced: Ag HNS_130, Ag HNS_240, and Ag HNS_260. The inner structure in Ag HNS was furthermore studied by cutting through cavities using a Focused Ion Beam (FIB) with an equipped SEM (FIB/SEM) under a magnification of 100,000×. Owing to e-beam evaporation of Ag toward PS_Bs, followed by annealing to remove them, it is likely that the as-formed cavities s are slightly smaller than the original size and dimension of PS_Bs [37]. The diameters of Ag HNSs were then measured by SEM as 130, 235, and 255 nm for Ag HNS_130, _240, _260, respectively. The as-measured size and dimension of PS_Bs and cavities did not show significant difference. It is most probably contributed by the rigidity of Ag shells and an appropriate control of released PS_Bs [29,45].
In Figure 2g, the cross-section of Ag HNS_130 is demonstrated. The rectangular dark area marked as #1 contains the cross-sectioned Ag HNSs with cavities, which are distributed between the annealed Ag shells and Si substrate. To have a better description on cavities, the larger rectangular area marked as #2 in Figure 2g was enlarged and illustrated in Figure 2h. On the left side of Figure 2h, two complete Ag HNSs were shown, whereas on the right side, two distorted Ag HNSs caused by an unparalleled direction of FIB/SEM manipulation with respect to the order of Ag HNSs were also demonstrated. The result reveals a clear formation of HNS between the annealed Ag shells and Si substrate. In our further applications, the well-ordered Ag HNSs with cavities are hence presumed. The magnification is 25,000×. To examine the cross section of (g) Ag HNSs_130, the sample was cut by FIB/SEM. A magnification of 100,000× was shown. The HNS structures between Si substrate and Ag shell were cut and shown (boxed in red and marked as #1) and (h) a closer examination of (g) with that of 150,000× (blue circles marked as #2).

Di NPs upon Ag HNS
In Figure 3a-c, FESEM photo-images for the attachment of Di NPs: HfO2, TiO2, and Al2O3 upon Ag HNSs were shown under a magnification of 90,000×. Apparently, their original sizes and dimensions of Ag HNSs still remained. To examine the composition and the distribution of Di NPs upon Ag HNS, a focused module of Energy Dispersive X-ray Spectrometer (EDS) was applied. In Figure 2d-f, taking the images from Figure 3a-c, single HfO2/, TiO2/, and Al2O3/Ag HNSs were respectively illustrated and enlarged on their top-right columns. In these EDS mapping images, the blue and green dots represent the background of Ag shell and O element, while the red dots represent Hf in 3d, Ti in 3e, and Al in 3f. The result reveals that the attachment of HfO2, TiO2 or Al2O3 upon Ag HNS is well confirmed through the EDS mapping process. Presumably, e-beam evaporation of Di NPs may result in random scattering of non-aggregated NPs, the possibility of such NPs upon Ag HNS is expected to be equal. Based on this assumption, a good distribution of Di NPs over the surface of Ag HNS can be presumably achieved. The magnification is 25,000×. To examine the cross section of (g) Ag HNSs_130, the sample was cut by FIB/SEM. A magnification of 100,000× was shown. The HNS structures between Si substrate and Ag shell were cut and shown (boxed in red and marked as #1) and (h) a closer examination of (g) with that of 150,000× (blue circles marked as #2).

Di NPs upon Ag HNS
In Figure 3a-c, FESEM photo-images for the attachment of Di NPs: HfO 2 , TiO 2 , and Al 2 O 3 upon Ag HNSs were shown under a magnification of 90,000×. Apparently, their original sizes and dimensions of Ag HNSs still remained. To examine the composition and the distribution of Di NPs upon Ag HNS, a focused module of Energy Dispersive X-ray Spectrometer (EDS) was applied.
In Figure 2d-f, taking the images from Figure 3a-c, single HfO 2 /, TiO 2 /, and Al 2 O 3 /Ag HNSs were respectively illustrated and enlarged on their top-right columns. In these EDS mapping images, the blue and green dots represent the background of Ag shell and O element, while the red dots represent Hf in 3d, Ti in 3e, and Al in 3f. The result reveals that the attachment of HfO 2 , TiO 2 or Al 2 O 3 upon Ag HNS is well confirmed through the EDS mapping process. Presumably, e-beam evaporation of Di NPs may result in random scattering of non-aggregated NPs, the possibility of such NPs upon Ag HNS is expected to be equal. Based on this assumption, a good distribution of Di NPs over the surface of Ag HNS can be presumably achieved.

SERS Property of Di NPs/Ag HNS
The concept of designing Di NPs/Ag HNS is to obtain an integrated SERS property with the synergy from the characteristic of Ag as SERS-active surface, an enhanced plasmonic effect inside Ag shell and the contribution of dielectric property at the interface between Ag shell and Di NPs when a Raman laser with an appropriate wavelength is induced. To verify their SERS property, a molecular probe R6G was firstly used to test the respective substrates, i.e., Ag HNS, Al2O3/, HfO2/, and TiO2/Ag HNS. In Figure 4, the characteristic peaks of R6G molecule are marked at: 611 cm −1 (C-C-C ring in plane bend), 773 cm −1 (C-C-C ring in plane bend), 1186 cm −1 (C-H in plane bend), 1309 cm −1 (C-O-C stretching), 1361 cm −1 (arom. C-C stretching), 1509 cm −1 (arom. C-C stretching), 1574 cm −1 (arom. C-C stretching), and 1649 cm −1 (arom. C-C stretching).

SERS Property of Di NPs/Ag HNS
The concept of designing Di NPs/Ag HNS is to obtain an integrated SERS property with the synergy from the characteristic of Ag as SERS-active surface, an enhanced plasmonic effect inside Ag shell and the contribution of dielectric property at the interface between Ag shell and Di NPs when a Raman laser with an appropriate wavelength is induced. To verify their SERS property, a molecular probe R6G was firstly used to test the respective substrates, i.e., Ag HNS, Al 2 O 3 /, HfO 2 /, and TiO 2 /Ag HNS. In Figure 4, the characteristic peaks of R6G molecule are marked at: 611 cm −1 (C-C-C ring in plane bend), 773 cm −1 (C-C-C ring in plane bend), 1186 cm −1 (C-H in plane bend), 1309 cm −1 (C-O-C stretching), 1361 cm −1 (arom. C-C stretching), 1509 cm −1 (arom. C-C stretching), 1574 cm −1 (arom. C-C stretching), and 1649 cm −1 (arom. C-C stretching).

SERS Property of Di NPs/Ag HNS
The concept of designing Di NPs/Ag HNS is to obtain an integrated SERS property with the synergy from the characteristic of Ag as SERS-active surface, an enhanced plasmonic effect inside Ag shell and the contribution of dielectric property at the interface between Ag shell and Di NPs when a Raman laser with an appropriate wavelength is induced. To verify their SERS property, a molecular probe R6G was firstly used to test the respective substrates, i.e., Ag HNS, Al2O3/, HfO2/, and TiO2/Ag HNS. In Figure 4, the characteristic peaks of R6G molecule are marked at: 611 cm −1 (C-C-C ring in plane bend), 773 cm −1 (C-C-C ring in plane bend), 1186 cm −1 (C-H in plane bend), 1309 cm −1 (C-O-C stretching), 1361 cm −1 (arom. C-C stretching), 1509 cm −1 (arom. C-C stretching), 1574 cm −1 (arom. C-C stretching), and 1649 cm −1 (arom. C-C stretching).  By taking the characteristic peak at 1361 cm −1 as the reference, Raman intensities, as compared to Di NPs/Ag HNS with Ag HNSs, significantly increased, as shown in Table 1. For examples, the EFs of Di NPs/Ag HNS_130 are 3.2-3.8 times higher than that of Ag HNS_130. Based on the equations (2.1), (2.2), and (2.3), I SERS is the Raman signal intensity at 1361 cm −1 for each substrate, I Bulk is 25 after Raman spectral analysis, N Bulk is estimated around 3.9 × 10 10 molecules, and N SERS is estimated as a value of 2.6 × 10 6 . The results indicate that (1) the EF of Ag HNS shows slightly higher as the size and dimension of cavity is smaller [16]; (2) all the Di NPs/Ag HNS_130 have EFs significantly higher than Ag HNSs and greater than 10 7 ; (3) Al 2 O 3 /Ag HNS_130 has the largest EF with a value of 6.2 × 10 7 , which also shows that Ag HNS with the attachment of lower dielectric constant NPs results in a high EF [44]. Note that the values of dielectric constant for Al 2 O 3 , HfO 2 , and TiO 2 / are 9-10, 20-30, and 10-85, respectively [46,47]. In general, for a SERS-active substrate with an EF of 10 7 to 10 8 or higher, it can provide the SERS property with a capability of single-molecule detection [48]. In any case, by increasing the specific surface area of the SERS-active substrate, more hot spots are likely formed that simultaneously increase the resulting Raman signal. If the bead size is decreased, the specific surface area tends to be augmented, consequently the Raman signal is enhanced owing to the presence of a high number of hot spots. In addition, the dielectric NPs not only increase the specific surface area but also induce chemical enhancement mechanism to occur owing to the involvement of charge transfers.

Di NPs/Ag HNS for Trace Detection of Antibiotic Residue in Water
As indicated in Section 3.3, a smaller diameter of cavity tends to increase the enhancement of plasmonic effect inside Ag shell. Thus, in the following applications, Di NPs with Ag HNS_130 (shortened as Ag HNS) are employed for trace detection of antibiotic, e.g., ampicillin, residue in different solutions.
In Figure 5a-c, SERS spectra for ampicillin in water with the respective concentrations of 100, 1, and 0.01 ppm, using HfO 2 /, Al 2 O 3 /, and TiO 2 /Ag HNS were demonstrated. Three characteristic peaks, at 1007, 1115, and 1447 cm −1 , for Ampicillin molecule are broadly marked in the spectra [46]. Among them, the peak at 1007 cm −1 is the most prominent characteristic peak, which is derived from the benzene ring vibration, shown on the top-left side in Figure 5d. Note that the peak at 1594 cm −1 is another significant peak of ampicillin from C=C stretching; it is usually affected by background (or fluorescent interference) while in detection, making its intensity hard to compare [49]. On the other hand, the peaks at 852 cm −1 and a broad one at around 1300-1400 cm −1 are most probably attributed to Ag-O associated effect, which indicates, in this study, the interactions between Di NPs and Ag or Ag-O [50]. In Figure 5d, by taking the peak at 1007 cm −1 as the reference, the relation of ampicillin's concentrations with respect to SERS intensities detected by three SERS-active substrates was illustrated. The results indicate that (1) SERS intensities increase with the addition of ampicillin for all these three substrates; (2) in consistence with dielectric constant studied in Section 3.3, Al 2 O 3 /Ag HNS exhibits sensitive with the increase of ampicillin's concentration in water, while TiO 2 /Ag HNS is less sensitive; (3) in general, Di NPs/Ag HNSs are all competent for trace detection of ampicillin in water. However, their increased enhancements at these three characteristic peaks differ from the substrates and concentrations of ampicillin.
It is very likely that the adsorption mechanism of ampicillin on each substrate is varied owing to the electronic structures formed by the combined SERS effect of target molecule upon different Di NPs/Ag HNS. For example, as found in Figure 5a-c, there are some peaks on Di NPs/Ag HNS, and the peaks' intensity obviously increases on Al 2 O 3 NPs/Ag HNS than on TiO 2 NPs/Ag HNS. Thus, a strong interaction between Al 2 O 3 and Ag leads to a modification of the electronic structure, while the significant characteristic peak (1007 cm −1 ) for three kinds of substrates is all derived from the benzene ring. The benzene ring molecules are weakly bonded to SERS substrate, and with excitation by laser light, the charge transfer occurs from the highest occupied molecular orbital region (HOMO) of ampicillin molecule to the metal's energy level, and then jumps to the lowest unfilled molecular orbital region (LUMO) of ampicillin molecule, finally returns to the ground state [50]. The result shown in Figure 5d is thus consistent with the correlation of dielectric property with Ag shell when Raman laser is applied to detect Ampicillin molecules upon Di NPs/Ag HNS. It also corresponds to the detection of R6G molecule that Al 2 O 3 with the lowest dielectric constant upon the similar structure of Ag HNS has the highest SERS intensity and EF.
Coatings 2020, 10, x FOR PEER REVIEW 9 of 15 the peaks' intensity obviously increases on Al2O3 NPs/Ag HNS than on TiO2 NPs/Ag HNS. Thus, a strong interaction between Al2O3 and Ag leads to a modification of the electronic structure, while the significant characteristic peak (1007 cm −1 ) for three kinds of substrates is all derived from the benzene ring. The benzene ring molecules are weakly bonded to SERS substrate, and with excitation by laser light, the charge transfer occurs from the highest occupied molecular orbital region (HOMO) of ampicillin molecule to the metal's energy level, and then jumps to the lowest unfilled molecular orbital region (LUMO) of ampicillin molecule, finally returns to the ground state [50]. The result shown in Figure 5d is thus consistent with the correlation of dielectric property with Ag shell when Raman laser is applied to detect Ampicillin molecules upon Di NPs/Ag HNS. It also corresponds to the detection of R6G molecule that Al2O3 with the lowest dielectric constant upon the similar structure of Ag HNS has the highest SERS intensity and EF.

Di NPs/Ag HNS for Trace Detection of Ampicillin in Milk Solution
In Figure 6a, SERS spectra for ampicillin in milk solution with low concentration of 0.01 ppm (e. g., to meet a standard for the residue regulation), using Al2O3/, HfO2/, and TiO2/Ag HNS were demonstrated. Characteristic peaks at 1007, 1115, 1447, and 1594 cm −1 for ampicillin were still distinguishable [39]. It was also observed that there are peaks at 1300-1400 cm −1 due to the interactions between Di NPs and Ag or Ag-O surface, as previously discussed in Section 3.4 (i.e., Figure 5a-c). The result shows that Di NPs/Ag HNSs are competent to achieve trace detection of ampicillin molecule as low as 0.01 ppm in milk solution. Among them, the peak at 1007 cm −1 is the most prominent peak. Note that milk is a complex fluid with an emulsion of suspended lipid droplets

Di NPs/Ag HNS for Trace Detection of Ampicillin in Milk Solution
In Figure 6a, SERS spectra for ampicillin in milk solution with low concentration of 0.01 ppm (e.g., to meet a standard for the residue regulation), using Al 2 O 3 /, HfO 2 /, and TiO 2 /Ag HNS were demonstrated. Characteristic peaks at 1007, 1115, 1447, and 1594 cm −1 for ampicillin were still distinguishable [39]. It was also observed that there are peaks at 1300-1400 cm −1 due to the interactions between Di NPs and Ag or Ag-O surface, as previously discussed in Section 3.4 (i.e., Figure 5a-c). The result shows that Di NPs/Ag HNSs are competent to achieve trace detection of ampicillin molecule as low as 0.01 ppm in milk solution. Among them, the peak at 1007 cm −1 is the most prominent peak. Note that milk is a complex fluid with an emulsion of suspended lipid droplets and a variety of sugars and proteins [39]. A lipid-removal procedure, as described in Experimental section, was thus necessarily prior to the sample into the detection [50,51]. The peak found at 1035 cm −1 presumably resulted from the surface of Al 2 O 3 NPs/Ag HNS, which is likely the background thar contributed by the core shell structure of Ag HNS [52,53] and Al 2 O 3 NPs upon Ag HNS, in the case of ampicillin in milk solution [44].
In Figure 6b, by taking these three peaks and low concentration (i.e., 0.01 ppm) of ampicillin in water and milk solution for a comparison, the results show the influence of ampicillin in different solutions. For the peak at 1007 cm −1 , SERS intensities decrease as ampicillin is detected in milk solution; in general, for Al 2 O 3 /, HfO 2 /, and TiO 2 /Ag HNS, their intensities decrease 60.3%, 26.3%, and 25.9%, respectively, with respect to ampicillin as measured in water. Presumably the milk solution results in much interference from other NSB molecules to hinder the absorption of ampicillin and leads to a decrease in the amount of target molecules upon SERS-active sites that may cause a reduction of Raman intensities. For the peaks at 1115 and 1447 cm −1 , under a comparable condition, a similar trend with less sensitive changes can be found. It thus implies NSB molecules in milk solution are very influential to the peak of ampicillin at 1007 cm −1 (benzene ring vibration).
Coatings 2020, 10, x FOR PEER REVIEW 10 of 15 and a variety of sugars and proteins [39]. A lipid-removal procedure, as described in Experimental section, was thus necessarily prior to the sample into the detection [50,51]. The peak found at 1035 cm −1 presumably resulted from the surface of Al2O3 NPs/Ag HNS, which is likely the background thar contributed by the core shell structure of Ag HNS [52,53] and Al2O3 NPs upon Ag HNS, in the case of ampicillin in milk solution [44]. In Figure 6b, by taking these three peaks and low concentration (i.e., 0.01 ppm) of ampicillin in water and milk solution for a comparison, the results show the influence of ampicillin in different solutions. For the peak at 1007 cm −1 , SERS intensities decrease as ampicillin is detected in milk solution; in general, for Al2O3/, HfO2/, and TiO2/Ag HNS, their intensities decrease 60.3%, 26.3%, and 25.9%, respectively, with respect to ampicillin as measured in water. Presumably the milk solution results in much interference from other NSB molecules to hinder the absorption of ampicillin and leads to a decrease in the amount of target molecules upon SERS-active sites that may cause a reduction of Raman intensities. For the peaks at 1115 and 1447 cm −1 , under a comparable condition, a similar trend with less sensitive changes can be found. It thus implies NSB molecules in milk solution are very influential to the peak of ampicillin at 1007 cm −1 (benzene ring vibration).

Schematic Description of Ampicillin Detection in Water and in Milk Solution
In Figure 7a,b, possible SERS mechanisms of Di NPs/Ag HNS are respectively illustrated for trace detections of residual ampicillin in water and in milk solution. In Figure 7a, the arrows represent the incident light rays. The direction of the light is scattered by HNSs that increases the probability of interactions with adsorbates. The main electromagnetic effect area is anticipated to happen: (i) in surface plasmon resonances of Ag HNS owing to the incident lights and electron transfers from the

Schematic Description of Ampicillin Detection in Water and in Milk Solution
In Figure 7a,b, possible SERS mechanisms of Di NPs/Ag HNS are respectively illustrated for trace detections of residual ampicillin in water and in milk solution. In Figure 7a, the arrows represent the incident light rays. The direction of the light is scattered by HNSs that increases the probability of interactions with adsorbates. The main electromagnetic effect area is anticipated to happen: (i) in surface plasmon resonances of Ag HNS owing to the incident lights and electron transfers from the contact of Di NPs and Ag surface and (ii) upon Raman-active sites, which are distributed among Ag shells with cavities where "hot spots" are formed when target molecules are attached and a Raman laser with an appropriate wavelength induced. In addition, the light on NPs induces the conduction electrons to oscillate collectively with a resonant frequency that depends on the nanoparticles' size, shape, composition, inter-particle distance, and environment (e.g., dielectric properties such as the inclusion of the dielectric edges may create additional hotspots). In the particular case of Ag NPs, the LSPR yields exceptionally high absorption coefficients and scattering properties within the UV/visible wavelength range. The SPR of Ag NPs (~420 nm, depending on their size) is obvious due to their spherical nature. The dielectric materials like Al 2 O 3 , TiO 2 , or HfO 2 coating can shift the band to higher wavelengths.
Coatings 2020, 10, x FOR PEER REVIEW 11 of 15 laser with an appropriate wavelength induced. In addition, the light on NPs induces the conduction electrons to oscillate collectively with a resonant frequency that depends on the nanoparticles' size, shape, composition, inter-particle distance, and environment (e. g., dielectric properties such as the inclusion of the dielectric edges may create additional hotspots). In the particular case of Ag NPs, the LSPR yields exceptionally high absorption coefficients and scattering properties within the UV/visible wavelength range. The SPR of Ag NPs (~420 nm, depending on their size) is obvious due to their spherical nature. The dielectric materials like Al2O3, TiO2, or HfO2 coating can shift the band to higher wavelengths. In the practical cases, a milk solution contains many NSB molecules that may hinder the absorption of ampicillin to be measured on SERS-active sites and therefore decrease Raman signals, in particular for the main peak at 1007 cm −1 . In Figure 7b, the incident laser interacts with ampicillin on Di NPs/Ag HNS in water and milk solution is illustrated. Ampicillin is expected to attach upon Di NPs/Ag HNS in water, making its SERS-active signals enhanced. In milk solution, non-target or NSB molecules such as the residues of sugar and protein are also adsorbed upon Di NPs/Ag HNS, Figure 7. Schematic diagrams of (a) the main electromagnetic enhancement areas: when the incident lights are appropriately introduced, (i) electron transfers from the contacts at the interface between Di NPs and Ag surface occur, while (ii) upon Raman-active sites, neighboring Ag shells with cavities form "hot spots" and (b) SERS mechanism of Di NPs/Ag HNS with the inclusion of ampicillin in water and in milk solution. Ampicillin molecules in water attach upon Di NPs/Ag HNS, making its SERS-active signals enhanced. In milk solution, NSB molecules are also adsorbed upon Di NPs/Ag HNS, making only a part of Ampicillin molecules attach to the surfaces of Di NPs/Ag HNS.
In the practical cases, a milk solution contains many NSB molecules that may hinder the absorption of ampicillin to be measured on SERS-active sites and therefore decrease Raman signals, in particular for the main peak at 1007 cm −1 . In Figure 7b, the incident laser interacts with ampicillin on Di NPs/Ag HNS in water and milk solution is illustrated. Ampicillin is expected to attach upon Di NPs/Ag HNS in water, making its SERS-active signals enhanced. In milk solution, non-target or NSB molecules such as the residues of sugar and protein are also adsorbed upon Di NPs/Ag HNS, making a part of ampicillin molecules indirectly attached to the surfaces of Di NPs/Ag HNS. The presence of NSB molecules tends to lengthen the distance of ampicillin with SERS-active sites and thereafter relax the effect of SERS for its different characteristic peaks. It is likely that Di NPs may act as a dielectric spacer for Ag HNS that reduce the chemisorption of NSB molecules on Ag shell surface; further study on the influence of overall SERS property is required.
As described in Figure 7a,b, Raman fingerprints of ampicillin can be clearly defined using SERS-active substrates, HfO 2 /, TiO 2 /, and Al 2 O 3 /Ag HNS; their detection limit can be as low as 0.01 ppm. In the case of milk solution, the interference from NSB molecules in milk solution may occur that significantly reduces the measured intensity of ampicillin molecules on SERS-active surface. A physical method to shorten the distance of target molecules with hot-spot areas by e.g., activating the substrate for improving high coverage rate of target molecules or reducing the absorption of NSB molecules is therefore suggested.

Conclusions
In this study, the removal of PS_B is subsequently used as the template for forming HNS with a cavity in Ag HNS, followed by depositing Di NPs and constructed as Di NPs/Ag HNS. The as-formed Di NPs/Ag HNS obtains an integrated SERS property with the synergy from the characteristic of Ag as a SERS-active surface, an enhanced plasmonic effect inside Ag shell, and the contribution of dielectric property at Ag shell and Di NPs interface when a Raman laser with an appropriate wavelength is induced. The results demonstrate that a cavity with small diameter and Di NPs with low dielectric constant will result in a higher EF. In a practical application for trace detection of ampicillin in water, SERS intensities increase with the addition of ampicillin for all these three substrates, while Al 2 O 3 /Ag HNS exhibits sensitivity with the increased concentration of ampicillin, in consistence with its low dielectric constant. For the case of detecting ampicillin in milk solution, Di NPs/Ag HNS is competent to achieve trace detection of Ampicillin as low as 0.01 ppm. However, for detecting Ampicillin on the respective Al 2 O 3 /, HfO 2 /, and TiO 2 /Ag HNS in milk solution, SERS intensity at the main peak of 1007 cm −1 decrease up to 60.3% with respect to that measured in water. The presence of NSB molecules tends to lengthen the distance of Ampicillin with SERS-active sites and thereafter relax the effect of SERS for its different characteristic peaks. A possible SERS mechanism of Di NPs/Ag HNS for trace detection of Ampicillin in water and in milk solution is proposed: when a Raman laser with an appropriate wavelength is induced, the main electromagnetic effect is particularly resulted at the interface of Di NPs and Ag shells where electron transfers from the contacts of Di NPs and Ag surface occur and the distribution of "hot spots" arises along Ag shells with cavities. For trace detection of ampicillin in milk solution, as compared with that in water, additional NSB molecules, prior to the attachment of ampicillin upon Di NPs/Ag HNS, may significantly reduce the measured SERS intensity of ampicillin upon SERS-active sites. An effective method to shorten the distance between target molecules and hot-spot areas or to reduce the absorption of NSB molecules is therefore proposed.

Highlights
• Ag HNS is prepared by a template-assisted method, followed by e-beam evaporated Di NPs, and formed as Di NPs/Ag HNS. • Al 2 O 3 /Ag HNS with low dielectric constant and smaller diameter of HNS obtains higher SERS enhancement factor.

•
The main electromagnetic effect of Di NPs/Ag HNS occurs at the interfaces of Di NPs, the surface of Ag shell, and the space of cavity. • For all Di NPs/Ag HNS, trace detection of ampicillin in water is efficient, SERS intensities increase with the addition of ampicillin.

•
Ampicillin in milk solution is influenced by non-specific binding molecules that lengthen the distance of ampicillin with SERS-active sites.