Ag Nanorods-Oxide Hybrid Array Substrates: Synthesis, Characterization, and Applications in Surface-Enhanced Raman Scattering

Over the last few decades, benefitting from the sufficient sensitivity, high specificity, nondestructive, and rapid detection capability of the surface-enhanced Raman scattering (SERS) technique, numerous nanostructures have been elaborately designed and successfully synthesized as high-performance SERS substrates, which have been extensively exploited for the identification of chemical and biological analytes. Among these, Ag nanorods coated with thin metal oxide layers (AgNRs-oxide hybrid array substrates) featuring many outstanding advantages have been proposed as fascinating SERS substrates, and are of particular research interest. The present review provides a systematic overview towards the representative achievements of AgNRs-oxide hybrid array substrates for SERS applications from diverse perspectives, so as to promote the realization of real-world SERS sensors. First, various fabrication approaches of AgNRs-oxide nanostructures are introduced, which are followed by a discussion on the novel merits of AgNRs-oxide arrays, such as superior SERS sensitivity and reproducibility, high thermal stability, long-term activity in air, corrosion resistivity, and intense chemisorption of target molecules. Next, we present recent advances of AgNRs-oxide substrates in terms of practical applications. Intriguingly, the recyclability, qualitative and quantitative analyses, as well as vapor-phase molecule sensing have been achieved on these nanocomposites. We further discuss the major challenges and prospects of AgNRs-oxide substrates for future SERS developments, aiming to expand the versatility of SERS technique.


Introduction
Since the discovery of surface-enhanced Raman scattering (SERS) in the 1970's [1], this vibration spectroscopic phenomenon has attracted enormous attention both in experimental study [2][3][4] and theoretical calculation [5][6][7]. Compared with normal Raman signals, the Raman scattering cross-sections of molecules can be enhanced by many orders of magnitude when they are adsorbed on the rough surfaces of noble metal (Au, Ag, and Cu) nanostructures [8][9][10]. This remarkable enhancement is aroused from two types of mechanisms, i.e., electromagnetic (EM) and chemical (CM) enhancements. The EM enhancement originates from the strongly amplified electric field at metal surface that is capable of generating localized surface plasmon resonance (LSPR), which depends significantly on the shape, size, and separation of metallic nanostructures [11][12][13]. The Raman cycles; (C) Transmission electron microscopy (TEM) images of AgNRs coated with Al2O3 layers by 1, 2, 3, and 5 ALD cycles, respectively (Reprinted with permission from [49]); (D) TEM images of AgNRs before and after coating with SiO2 layers for different reaction times (Reprinted with permission from [58]. Copyright (2011) American Chemical Society); (E) TEM image of a single AgNR with TiO2 capping primarily at the nanorod tip (top), and electron diffraction pattern of multiple AgNRs (bottom) (Reprinted with permission from [50]. Copyright (2014) AIP Publishing LLC).

Fabrication of AgNRs-Oxide Hybrid Array Substrates with Different Oxide Layers
Several approaches have been exploited to produce oxide layers over AgNR arrays. Because SERS is a highly localized effect that depends significantly on the distance between metal surfaces and target molecules [62][63][64], the oxide shells should be thin enough so as not to eliminate the SERS enhancement. In this regard, atomic layer deposition (ALD) holds great potential in the oxide formation of AgNRs-oxide hybrid substrates. ALD is a unique thin film growth technique by means of sequential self-limiting surface reactions of gaseous precursors [65][66][67]. It is capable of preparing high-quality films with precise thickness control at the atomic scale, excellent conformality independent of the substrate geometry, low defect density, and large-scale uniformity. As an example, Ma et al. have deposited Al2O3 over AgNRs using trimethylaluminum (TMA) and water as ALD precursors [49]. High purity N2 was adopted as the carrier and purge gas. Typically, one complete reaction cycle consisted of four steps: (1) TMA reactant exposure; (2) N2 gas purging; cycles; (C) Transmission electron microscopy (TEM) images of AgNRs coated with Al 2 O 3 layers by 1, 2, 3, and 5 ALD cycles, respectively (Reprinted with permission from [49]); (D) TEM images of AgNRs before and after coating with SiO 2 layers for different reaction times (Reprinted with permission from [58]. Copyright (2011) American Chemical Society); (E) TEM image of a single AgNR with TiO 2 capping primarily at the nanorod tip (top), and electron diffraction pattern of multiple AgNRs (bottom) (Reprinted with permission from [50]. Copyright (2014) AIP Publishing LLC).

Fabrication of AgNRs-Oxide Hybrid Array Substrates with Different Oxide Layers
Several approaches have been exploited to produce oxide layers over AgNR arrays. Because SERS is a highly localized effect that depends significantly on the distance between metal surfaces and target molecules [62][63][64], the oxide shells should be thin enough so as not to eliminate the SERS enhancement. In this regard, atomic layer deposition (ALD) holds great potential in the oxide formation of AgNRs-oxide hybrid substrates. ALD is a unique thin film growth technique by means of sequential self-limiting surface reactions of gaseous precursors [65][66][67]. It is capable of preparing high-quality films with precise thickness control at the atomic scale, excellent conformality independent of the substrate geometry, low defect density, and large-scale uniformity. As an example, Ma et al. have deposited Al 2 O 3 over AgNRs using trimethylaluminum (TMA) and water as ALD precursors [49]. High purity N 2 was adopted as the carrier and purge gas. Typically, one complete reaction cycle consisted of four steps: (1) TMA reactant exposure; (2) N 2 gas purging; (3) water vapor exposure; and (4) N 2 gas purging. This reaction cycle was repeated for different times so as to control the Al 2 O 3 thickness. TiO 2 and HfO 2 have also been successfully deposited over AgNRs by ALD [57,60], with similar reaction cycles but different precursors and reaction time. It is worth noticing that a low-temperature ALD process at 50 to 80 • C is required to avoid the coarsening and fusion of the underneath AgNR arrays during oxide growth. Besides, based on the hydrolysis reaction of tetraethyl orthosilicate (TEOS), uniform and conformal SiO 2 layers have been coated onto AgNRs to form AgNRs-SiO 2 core-shell nanostructures [58]. The shell growth made porous SiO 2 more compact, and its thickness was tailored by varying the coating time. Moreover, Huang et al. proposed to cover AgNRs with a thin TiO 2 layer directly through OAD fabrication [50]. To be specific, after AgNRs deposition, they changed the evaporation material from Ag to TiO 2 , and did the deposition again at the same substrate orientation. Because this coating approach only requires the OAD system instead of many complicated apparatuses, it is very efficient and cost-effective and thus promising for real-world fabrication. Figure 1B shows the scanning electron microscope (SEM) image of the AgNRs-oxide substrate. Due to the ultrathin feature of oxide layer, AgNRs-oxide arrays reveal no visible morphology variation compared with the pristine AgNRs. To have a better observation of the oxide shells, high-resolution transmission electron microscope (HRTEM) analysis is employed to provide a visual evidence, which is also applied to investigate oxide thickness growth. As illustrated in Figure 1C,D, the oxide layers fabricated by ALD and hydrolysis reaction are amorphous in structure and of different thickness, uniformly and conformally wrapping AgNRs. When the reaction time is very short, the as-prepared oxide layers possess a few pinholes; as the reaction continues, the shells become thicker and more compact. HRTEM results also demonstrate that the growth thickness increases linearly with the ALD cycle/reaction time [49,58], which is beneficial for us to precisely control the layer thickness to sub-nanometer scale. The thickness of Al 2 O 3 , TiO 2 , and HfO 2 shells fabricated by ALD is about 0.6-0.8 nm per ALD cycle [49,57,60], and that of SiO 2 grown by hydrolysis reaction is approximately 0.27 nm/min [58]. For TiO 2 capping prepared by the OAD method in Figure 1E, it is about 5 nm thick and mainly located on the top surfaces of AgNRs, and has both amorphous and crystalline regions [50].

SERS Sensitivity and Reproducibility of AgNRs-Oxide Hybrid Array Substrates
It has been long recognized that the SERS enhancement of metal nanostructures depends strongly on the distance between metal surfaces and adsorbed molecules. We therefore investigate the coating effect on the sensitivity of SERS substrates. The results in Figure 2a clearly present that the SERS efficiency of AgNR arrays decreases to~65% and~50% after~0.7 nm (1 ALD cycle) and~1.5 nm (2 ALD cycles) oxide coating, and declines monotonously with further increasing the oxide thickness [49]. We should note that, due to the ultrathin feature of oxide layers, the strong SERS enhancement of AgNRs is well maintained. All hybrid substrates exhibit satisfactory SERS EFs on the order of 10 7 [49,57,60], confirming the remarkable sensitivity of AgNRs-oxide nanocomposites. A low relative standard deviation (RSD) value with Raman signals of~5% [56] (see Figure 2b) indicates that the AgNRs-oxide substrates are uniform in structure and of good reproducibility for SERS measurements, which is also a prerequisite for quantitative analysis.

Thermal Stability of AgNRs-Oxide Hybrid Array Substrates
High-temperature SERS detection is a vital part for routine applications, which can be utilized for monitoring many in situ reactions, such as thermal crystallization [68], structural variations [69,70], and chemical reactions [52,71] at elevated temperatures. For pristine AgNR arrays, their structure begins to change at a very low temperature of 50 °C, and collapses completely at 100 °C [49,50]. It is thereby highly demanded to improve the thermal stability of AgNRs-based sensors both in morphology robustness and SERS sensitivity. Since the melting point of oxides (1700-2700 °C) is much higher than that of silver (960 °C), covering AgNR arrays with oxide layers might be effective to address this issue. It is illustrated in Figure 3a that the AgNRs substrate coated by 1-cycle Al2O3 (~0.7 nm) are robust in morphology at 200 °C, but melt partly at 300 and 400 °C. For AgNRs coated by 2-cycle Al2O3 of ~1.5 nm thick, no obvious structural change is observed after being heated at 300 and 400 °C [49]. The morphology robustness also leads to the stabilized SERS performance at elevated temperatures, demonstrated in Figure 3b-c. As such, AgNR arrays coated with ~1.6 nm TiO2 or HfO2 shell also sustain their morphology and SERS efficiency at 300-400 °C [57,60]. Moreover, by capping the top surfaces of AgNRs with high melting-temperature TiO2 of ~5 nm, the Ag mass transport from tips to sides is slowed down. They preserve their shapes well at 100 °C, while they coarsen to some extent at 200 °C [50]. These results suggest that the oxide coating/capping functions as a barrier to protect AgNRs both in morphology stiffness and SERS sensitivity against high temperatures, and covering the entire surfaces of AgNRs is especially useful to achieve these goals.

Thermal Stability of AgNRs-Oxide Hybrid Array Substrates
High-temperature SERS detection is a vital part for routine applications, which can be utilized for monitoring many in situ reactions, such as thermal crystallization [68], structural variations [69,70], and chemical reactions [52,71] at elevated temperatures. For pristine AgNR arrays, their structure begins to change at a very low temperature of 50 • C, and collapses completely at 100 • C [49,50]. It is thereby highly demanded to improve the thermal stability of AgNRs-based sensors both in morphology robustness and SERS sensitivity. Since the melting point of oxides (1700-2700 • C) is much higher than that of silver (960 • C), covering AgNR arrays with oxide layers might be effective to address this issue. It is illustrated in Figure 3a that the AgNRs substrate coated by 1-cycle Al 2 O 3 (~0.7 nm) are robust in morphology at 200 • C, but melt partly at 300 and 400 • C. For AgNRs coated by 2-cycle Al 2 O 3 of 1.5 nm thick, no obvious structural change is observed after being heated at 300 and 400 • C [49]. The morphology robustness also leads to the stabilized SERS performance at elevated temperatures, demonstrated in Figure 3b,c. As such, AgNR arrays coated with~1.6 nm TiO 2 or HfO 2 shell also sustain their morphology and SERS efficiency at 300-400 • C [57,60]. Moreover, by capping the top surfaces of AgNRs with high melting-temperature TiO 2 of~5 nm, the Ag mass transport from tips to sides is slowed down. They preserve their shapes well at 100 • C, while they coarsen to some extent at 200 • C [50]. These results suggest that the oxide coating/capping functions as a barrier to protect AgNRs both in morphology stiffness and SERS sensitivity against high temperatures, and covering the entire surfaces of AgNRs is especially useful to achieve these goals.

Thermal Stability of AgNRs-Oxide Hybrid Array Substrates
High-temperature SERS detection is a vital part for routine applications, which can be utilized for monitoring many in situ reactions, such as thermal crystallization [68], structural variations [69,70], and chemical reactions [52,71] at elevated temperatures. For pristine AgNR arrays, their structure begins to change at a very low temperature of 50 °C, and collapses completely at 100 °C [49,50]. It is thereby highly demanded to improve the thermal stability of AgNRs-based sensors both in morphology robustness and SERS sensitivity. Since the melting point of oxides (1700-2700 °C) is much higher than that of silver (960 °C), covering AgNR arrays with oxide layers might be effective to address this issue. It is illustrated in Figure 3a that the AgNRs substrate coated by 1-cycle Al2O3 (~0.7 nm) are robust in morphology at 200 °C, but melt partly at 300 and 400 °C. For AgNRs coated by 2-cycle Al2O3 of ~1.5 nm thick, no obvious structural change is observed after being heated at 300 and 400 °C [49]. The morphology robustness also leads to the stabilized SERS performance at elevated temperatures, demonstrated in Figure 3b-c. As such, AgNR arrays coated with ~1.6 nm TiO2 or HfO2 shell also sustain their morphology and SERS efficiency at 300-400 °C [57,60]. Moreover, by capping the top surfaces of AgNRs with high melting-temperature TiO2 of ~5 nm, the Ag mass transport from tips to sides is slowed down. They preserve their shapes well at 100 °C, while they coarsen to some extent at 200 °C [50]. These results suggest that the oxide coating/capping functions as a barrier to protect AgNRs both in morphology stiffness and SERS sensitivity against high temperatures, and covering the entire surfaces of AgNRs is especially useful to achieve these goals.

Temporal Stability of AgNRs-Oxide Hybrid Array Substrates
The effect of oxide coating on the temporal stability of AgNRs in air is investigated as a function of shelf time, see details in Figure 4 [49]. For pristine AgNR arrays, due to the highly active surfaces, the SERS activity drops drastically in air, which declines by half after 10-day storage time and is about one order smaller after 50 days. On the other hand, when AgNRs are uniformly wrapped with oxides, the protective shells could suppress their surface reactions with air, and accordingly the shelf life is dramatically increased. Specifically, the AgNRs substrate coated with ~0.7 nm Al2O3 presents a slight signal decrease after 50 days, which can be explained by the pinhole-containing feature of the ultrathin oxide shell. While the substrates coated by ~1.5 nm or thicker Al2O3 remain almost constant in SERS response during the whole test period. As a result, a thin but compact coating layer could sufficiently passivate the internal Ag NRs so as to stabilize their SERS activity under atmospheric conditions for a long period. Similar conclusion has been verified on AgNRs-TiO2 substrates [57].

Chemical stability of AgNRs-oxide hybrid array substrates
To function as reliable SERS sensors, the chemical stability is of essential importance especially at erosive environments. Because chloride ions [54,72,73], strong oxidants [54,56], and acidic

Temporal Stability of AgNRs-Oxide Hybrid Array Substrates
The effect of oxide coating on the temporal stability of AgNRs in air is investigated as a function of shelf time, see details in Figure 4 [49]. For pristine AgNR arrays, due to the highly active surfaces, the SERS activity drops drastically in air, which declines by half after 10-day storage time and is about one order smaller after 50 days. On the other hand, when AgNRs are uniformly wrapped with oxides, the protective shells could suppress their surface reactions with air, and accordingly the shelf life is dramatically increased. Specifically, the AgNRs substrate coated with~0.7 nm Al 2 O 3 presents a slight signal decrease after 50 days, which can be explained by the pinhole-containing feature of the ultrathin oxide shell. While the substrates coated by~1.5 nm or thicker Al 2 O 3 remain almost constant in SERS response during the whole test period. As a result, a thin but compact coating layer could sufficiently passivate the internal Ag NRs so as to stabilize their SERS activity under atmospheric conditions for a long period. Similar conclusion has been verified on AgNRs-TiO 2 substrates [57].

Temporal Stability of AgNRs-Oxide Hybrid Array Substrates
The effect of oxide coating on the temporal stability of AgNRs in air is investigated as a function of shelf time, see details in Figure 4 [49]. For pristine AgNR arrays, due to the highly active surfaces, the SERS activity drops drastically in air, which declines by half after 10-day storage time and is about one order smaller after 50 days. On the other hand, when AgNRs are uniformly wrapped with oxides, the protective shells could suppress their surface reactions with air, and accordingly the shelf life is dramatically increased. Specifically, the AgNRs substrate coated with ~0.7 nm Al2O3 presents a slight signal decrease after 50 days, which can be explained by the pinhole-containing feature of the ultrathin oxide shell. While the substrates coated by ~1.5 nm or thicker Al2O3 remain almost constant in SERS response during the whole test period. As a result, a thin but compact coating layer could sufficiently passivate the internal Ag NRs so as to stabilize their SERS activity under atmospheric conditions for a long period. Similar conclusion has been verified on AgNRs-TiO2 substrates [57].

Chemical stability of AgNRs-oxide hybrid array substrates
To function as reliable SERS sensors, the chemical stability is of essential importance especially at erosive environments. Because chloride ions [54,72,73], strong oxidants [54,56], and acidic

Chemical Stability of AgNRs-Oxide Hybrid Array Substrates
To function as reliable SERS sensors, the chemical stability is of essential importance especially at erosive environments. Because chloride ions [54,72,73], strong oxidants [54,56], and acidic solutions [61,74]  can severely etch silver atoms and cause the shape transformation of Ag nanostructures (see Figure 5a as an example), it is crucial to improve the chemical stability of AgNRs against corrosion. We find that after Al 2 O 3 deposition, the chemical stability of coated substrates is substantially improved, even if the oxide shell is sub-nanometer and has a few pinholes [56]. As shown in Figure 5b-d, the chemically inert Al 2 O 3 shell prevents the internal AgNRs from direct contact with external etchants such as NaCl and H 2 O 2 , retaining sufficiently their morphology and SERS efficiency. Additionally, for strong acidic media where Al 2 O 3 will dissolve quickly, ultrathin HfO 2 film has been implemented to protect AgNRs from failure [61]. This kind of substrate possesses the acid-resistant property, making it applicable in acid solutions of practical environments.
Sensors 2017, 17,1985 7 of 17 solutions [61,74] can severely etch silver atoms and cause the shape transformation of Ag nanostructures (see Figure 5a as an example), it is crucial to improve the chemical stability of AgNRs against corrosion. We find that after Al2O3 deposition, the chemical stability of coated substrates is substantially improved, even if the oxide shell is sub-nanometer and has a few pinholes [56]. As shown in Figure 5b-d, the chemically inert Al2O3 shell prevents the internal AgNRs from direct contact with external etchants such as NaCl and H2O2, retaining sufficiently their morphology and SERS efficiency. Additionally, for strong acidic media where Al2O3 will dissolve quickly, ultrathin HfO2 film has been implemented to protect AgNRs from failure [61]. This kind of substrate possesses the acid-resistant property, making it applicable in acid solutions of practical environments.

Reusable SERS Substrates
Given that SERS substrates are generally made of noble metals and are not readily reused, the costly preparation and disposable property seriously hinder the universality of SERS technique. For this reason, it is significant to develop recyclable SERS substrates. One feasible way is the direct degradation of adsorbed molecules from substrate surfaces after SERS identification. For AgNR arrays coated with photocatalytic material TiO2 [75][76][77], the self-cleaning ability is realized through ultraviolet (UV) light-induced decomposition of organic molecules adsorbed on the substrate, i.e., subsequent to SERS measurements, the substrate can be purified by UV illumination and be reused for further SERS analysis [57]. As revealed in Figure 6a, four "detection-UV cleaning" circulations are carried out on the AgNRs substrate coated with ~2 nm TiO2. Strong target signals are observed during

Reusable SERS Substrates
Given that SERS substrates are generally made of noble metals and are not readily reused, the costly preparation and disposable property seriously hinder the universality of SERS technique. For this reason, it is significant to develop recyclable SERS substrates. One feasible way is the direct degradation of adsorbed molecules from substrate surfaces after SERS identification. For AgNR arrays coated with photocatalytic material TiO 2 [75][76][77], the self-cleaning ability is realized through ultraviolet (UV) light-induced decomposition of organic molecules adsorbed on the substrate, i.e., subsequent to SERS measurements, the substrate can be purified by UV illumination and be reused for further SERS analysis [57]. As revealed in Figure 6a, four "detection-UV cleaning" circulations are carried out on the AgNRs substrate coated with~2 nm TiO 2 . Strong target signals are observed during SERS sensing, and they almost vanish after UV irradiation. More importantly, the results from the subsequent three circulations show that the Raman intensities are fully recovered at each detection step, which suggests that the AgNRs-TiO 2 structure is capable of enduring multiple UV irradiations with considerable robustness. Another point to note is that the molecule degradation capability not only comes from the photocatalysis of TiO 2 , it also benefits from the intensive interaction between Ag and TiO 2 that optimizes the separation of photo-excited charge carriers and as a result facilitates the degradation efficiency (see Figure 6b) [78][79][80].
Thermal annealing is another way to detach molecules from adsorbed surfaces, and accordingly it might be helpful to clean and regenerate SERS substrates. By virtue of the high melting-temperature HfO 2 [81], the AgNRs substrate coated by~1.6 nm HfO 2 represent good thermal stability and morphological robustness at temperatures up to 400 • C [60]. After SERS detection, the regeneration of AgNRs-HfO 2 can be achieved by heating the substrate on a hot plate within several seconds. This process leads to the thermal release of adsorbed molecules and refreshes the substrate for subsequent measurements. From Figure 6c,d, one sees that the hybrid substrate maintains its SERS efficiency well during 30 "detection-heating" cycles, demonstrating the remarkable stability and recyclability of AgNRs-HfO 2 substrate. As a consequence, reusability could eliminate the single-use shortcoming of conventional SERS substrates, in which way the high costs for SERS measurements are substantially reduced and the practicability of the SERS technique is extended.
Sensors 2017, 17,1985 8 of 17 SERS sensing, and they almost vanish after UV irradiation. More importantly, the results from the subsequent three circulations show that the Raman intensities are fully recovered at each detection step, which suggests that the AgNRs-TiO2 structure is capable of enduring multiple UV irradiations with considerable robustness. Another point to note is that the molecule degradation capability not only comes from the photocatalysis of TiO2, it also benefits from the intensive interaction between Ag and TiO2 that optimizes the separation of photo-excited charge carriers and as a result facilitates the degradation efficiency (see Figure 6b) [78][79][80]. Thermal annealing is another way to detach molecules from adsorbed surfaces, and accordingly it might be helpful to clean and regenerate SERS substrates. By virtue of the high melting-temperature HfO2 [81], the AgNRs substrate coated by ~1.6 nm HfO2 represent good thermal stability and morphological robustness at temperatures up to 400 °C [60]. After SERS detection, the regeneration of AgNRs-HfO2 can be achieved by heating the substrate on a hot plate within several seconds. This process leads to the thermal release of adsorbed molecules and refreshes the substrate for subsequent measurements. From Figure 6c-d, one sees that the hybrid substrate maintains its SERS efficiency well during 30 "detection-heating" cycles, demonstrating the remarkable stability and recyclability of AgNRs-HfO2 substrate. As a consequence, reusability could eliminate the single-use shortcoming of conventional SERS substrates, in which way the high costs for SERS measurements are substantially reduced and the practicability of the SERS technique is extended. Figure 6. (a) Raman spectra of 5 × 10 −6 M MB molecules adsorbed on the AgNRs substrate coated with ~2 nm TiO2 in four "detection-UV cleaning" cycles; (b) the schematic for the photocatalytic mechanism of AgNRs-TiO2 hybrids (Reprinted with permission from [57]); (c) Raman spectra of 1 × 10 −6 M MB on the AgNRs substrate coated with ~1.6 nm HfO2 measured in multiple "detection-heating" cycles and (d) the 1623 cm −1 peak intensity variations in 30 cycles (Reprinted with permission from [60]. Copyright (2016) American Chemical Society).

Qualitative and Quantitative SERS Analyses
To make SERS technique a practical and reliable analysis tool, both qualitative and quantitative abilities are required for desirable SERS substrates. It has been reported that Al2O3 has high affinity Figure 6. (a) Raman spectra of 5 × 10 −6 M MB molecules adsorbed on the AgNRs substrate coated with 2 nm TiO 2 in four "detection-UV cleaning" cycles; (b) the schematic for the photocatalytic mechanism of AgNRs-TiO 2 hybrids (Reprinted with permission from [57]); (c) Raman spectra of 1 × 10 −6 M MB on the AgNRs substrate coated with~1.6 nm HfO 2 measured in multiple "detection-heating" cycles and (d) the 1623 cm −1 peak intensity variations in 30 cycles (Reprinted with permission from [60]. Copyright (2016) American Chemical Society).

Qualitative and Quantitative SERS Analyses
To make SERS technique a practical and reliable analysis tool, both qualitative and quantitative abilities are required for desirable SERS substrates. It has been reported that Al 2 O 3 has high affinity to carboxyl (-COOH) functional groups, ascribed to the strong polar interaction [82,83]. Therefore, we adopt AgNRs-Al 2 O 3 substrates to enhance the absorbability and correspondingly SERS detection efficiency of carboxylic acids. Dipicolinic acid (DPA) is a commonly used biomarker for the recognition of bacterial spores [83,84], so its sensitive and quantitative probing is particularly important. Figure 7a shows the Raman spectra of DPA by employing AgNRs coated with 1-cycle Al 2 O 3 as the SERS platform, with the limit of detection (LOD) down to 10 −8 M. Meanwhile, the partial least squares regression (PLSR) [85,86] model in Figure 7b exhibits good predictability within the concentration ranging from 1 × 10 −8 to 1 × 10 −5 M, which provides a calibration for the quantification of trace DPA. In addition to analytes that can directly react with Al 2 O 3 surfaces, for those who have no or weak interaction with oxides (such as pyridine, acridine, and cyanide [87][88][89]), the pinhole-containing AgNRs-Al 2 O 3 arrays provide a channel to anchor them directly onto Ag surfaces through Al 2 O 3 pinholes. For instance, the LOD of NaCN on the pinhole-containing AgNRs-Al 2 O 3 substrate is as low as 1 ppb, and quantitative analysis also reveals a satisfactory PLSR predictability [56]. These results suggest that AgNRs-oxide substrates are appropriate to detect a variety of molecules in both qualitative and quantitative manners.
Sensors 2017, 17,1985 9 of 17 to carboxyl (-COOH) functional groups, ascribed to the strong polar interaction [82,83]. Therefore, we adopt AgNRs-Al2O3 substrates to enhance the absorbability and correspondingly SERS detection efficiency of carboxylic acids. Dipicolinic acid (DPA) is a commonly used biomarker for the recognition of bacterial spores [83,84], so its sensitive and quantitative probing is particularly important. Figure 7a shows the Raman spectra of DPA by employing AgNRs coated with 1-cycle Al2O3 as the SERS platform, with the limit of detection (LOD) down to 10 -8 M. Meanwhile, the partial least squares regression (PLSR) [85,86] model in Figure 7b exhibits good predictability within the concentration ranging from 1 × 10 −8 to 1 × 10 −5 M, which provides a calibration for the quantification of trace DPA. In addition to analytes that can directly react with Al2O3 surfaces, for those who have no or weak interaction with oxides (such as pyridine, acridine, and cyanide [87][88][89]), the pinholecontaining AgNRs-Al2O3 arrays provide a channel to anchor them directly onto Ag surfaces through Al2O3 pinholes. For instance, the LOD of NaCN on the pinhole-containing AgNRs-Al2O3 substrate is as low as 1 ppb, and quantitative analysis also reveals a satisfactory PLSR predictability [56]. These results suggest that AgNRs-oxide substrates are appropriate to detect a variety of molecules in both qualitative and quantitative manners. From a more realistic perspective, AgNRs-oxide substrates have been adopted to sense food antiseptics [61]. It is known that food antiseptics with appropriate amount could inhibit bacteria and extend food's shelf life; whereas excessive addition might be harmful for human health [90,91]. Hence, quantitative analysis of antiseptics based on SERS is of great value. Potassium sorbate (PS) and sodium benzoate (SB) are common food antiseptics. Because they only work in acidic media [92,93], the AgNRs-HfO2 substrate with acid resistance is applicable for their identification [61]. The LODs of these two antiseptics are both 300 μg/L, which are much lower than their dosage standard in food. Therefore, the SERS substrate meets the demand of identifying PS and SB in practice. Moreover, the PLSR relationship between the concentrations and SERS spectra of a series of PS solutions performs quite well. As for the mixture of PS and SB, even if some of their characteristic peaks overlap, the two respective PLSR models of them are both accurate and reliable (see Figure 8). That is to say, the Raman peaks corresponding to SB in spectra do not interfere the quantification of PS molecules, and vice versa. From a more realistic perspective, AgNRs-oxide substrates have been adopted to sense food antiseptics [61]. It is known that food antiseptics with appropriate amount could inhibit bacteria and extend food's shelf life; whereas excessive addition might be harmful for human health [90,91]. Hence, quantitative analysis of antiseptics based on SERS is of great value. Potassium sorbate (PS) and sodium benzoate (SB) are common food antiseptics. Because they only work in acidic media [92,93], the AgNRs-HfO 2 substrate with acid resistance is applicable for their identification [61]. The LODs of these two antiseptics are both 300 µg/L, which are much lower than their dosage standard in food. Therefore, the SERS substrate meets the demand of identifying PS and SB in practice. Moreover, the PLSR relationship between the concentrations and SERS spectra of a series of PS solutions performs quite well. As for the mixture of PS and SB, even if some of their characteristic peaks overlap, the two respective PLSR models of them are both accurate and reliable (see Figure 8). That is to say, the Raman peaks corresponding to SB in spectra do not interfere the quantification of PS molecules, and vice versa.

Vapor-Phase Molecule Sensing
Aside from revealing the feasibility of SERS determination in liquids, AgNRs-oxide arrays also provide real-time monitoring of vapor-phase molecules at ultralow concentrations [60]. To capture and sense target gases, the AgNRs-HfO2 substrate is placed in a homemade gas detection system presented in Figure 9a. 2-naphthalenethiol (2-NAT) is selected as the model gas, and high purity N2 is utilized as the carrier gas. During gas detection, N2 is injected into the analyte solution, and target molecules in the vapor phase are carried out together with the N2 flow and are captured by the AgNRs-HfO2 platform. SERS spectra are recorded simultaneously during gas passing. It is shown in Figure 9b that the 2-NAT signals ascend continuously with the gas flowing until the saturation of the substrate surface. The LOD of 2-NAT is down to 20 ppb, verifying the effectiveness of AgNRs-HfO2 substrate for gas detection. Furthermore, we explore the renewability of AgNRs-HfO2 arrays during gas recognition, i.e., after SERS measurement, the substrate is heated on a hot plate at 250 °C for 30 s to desorb molecules. The results in Figure 9c reveal that the Raman intensities of 600 ppb 2-NAT at 1379 cm −1 peak escalate with vapor passing and after each annealing treatment, the substrate is free of 2-NAT while it is totally recovered in the following "vapor exposure-thermal cleaning" cycles. We believe this highly robust and versatile SERS platform could act as a recyclable sensor for in situ monitoring of complex gases from realistic environments, such as air pollutants [94,95], explosives [96,97], volatile organic compounds [98,99], and chemical warfare agents [100,101].

Vapor-Phase Molecule Sensing
Aside from revealing the feasibility of SERS determination in liquids, AgNRs-oxide arrays also provide real-time monitoring of vapor-phase molecules at ultralow concentrations [60]. To capture and sense target gases, the AgNRs-HfO 2 substrate is placed in a homemade gas detection system presented in Figure 9a. 2-naphthalenethiol (2-NAT) is selected as the model gas, and high purity N 2 is utilized as the carrier gas. During gas detection, N 2 is injected into the analyte solution, and target molecules in the vapor phase are carried out together with the N 2 flow and are captured by the AgNRs-HfO 2 platform. SERS spectra are recorded simultaneously during gas passing. It is shown in Figure 9b that the 2-NAT signals ascend continuously with the gas flowing until the saturation of the substrate surface. The LOD of 2-NAT is down to 20 ppb, verifying the effectiveness of AgNRs-HfO 2 substrate for gas detection. Furthermore, we explore the renewability of AgNRs-HfO 2 arrays during gas recognition, i.e., after SERS measurement, the substrate is heated on a hot plate at 250 • C for 30 s to desorb molecules. The results in Figure 9c reveal that the Raman intensities of 600 ppb 2-NAT at 1379 cm −1 peak escalate with vapor passing and after each annealing treatment, the substrate is free of 2-NAT while it is totally recovered in the following "vapor exposure-thermal cleaning" cycles. We believe this highly robust and versatile SERS platform could act as a recyclable sensor for in situ monitoring of complex gases from realistic environments, such as air pollutants [94,95], explosives [96,97], volatile organic compounds [98,99], and chemical warfare agents [100,101].

Conclusions
This review provides insights for the synthesis, characterization, and applications of AgNRsoxide hybrid array substrates as SERS sensing platforms, all in order to emphasize our understanding and utilization of the SERS technique. AgNR arrays are prepared based on the OAD method, and oxide shells are readily coated onto AgNRs by diverse approaches with controllable thickness and excellent uniformity. The characterization of these nanocomposites is detailed in aspects of shell growth, SERS sensitivity, and reproducibility, as well as thermal, temporal, and chemical stability. By virtue of the ultrathin thickness, uniformity, and stability of oxide layers, AgNRs-oxide substrates possess large SERS EFs, outstanding reproducibility, and excellent stability. In addition, the reusability, quantification, and gas sensing have been achieved on the substrates, which hold great potential for the identification of trace analytes in real systems.
Although AgNRs-oxide array substrates have proved their extraordinary advantages, they are still in their infancy for practical SERS applications, and several challenges are yet to be addressed. First, AgNRs-oxide hybrid arrays possess satisfactory reproducibility and reliability exceeding the performance of commonly used nanoparticles, which are crucial for biosensing. The qualitative and quantitative detection of DPA, an anthrax biomarker, has been achieved on the AgNRs-Al2O3 substrate. As for the SERS-based immunoassay and nucleic acid detection, AgNRs-oxide arrays offer potent capture substrates to bind SERS nanotags. The close proximity of the substrate-nanotag

Conclusions
This review provides insights for the synthesis, characterization, and applications of AgNRs-oxide hybrid array substrates as SERS sensing platforms, all in order to emphasize our understanding and utilization of the SERS technique. AgNR arrays are prepared based on the OAD method, and oxide shells are readily coated onto AgNRs by diverse approaches with controllable thickness and excellent uniformity. The characterization of these nanocomposites is detailed in aspects of shell growth, SERS sensitivity, and reproducibility, as well as thermal, temporal, and chemical stability. By virtue of the ultrathin thickness, uniformity, and stability of oxide layers, AgNRs-oxide substrates possess large SERS EFs, outstanding reproducibility, and excellent stability. In addition, the reusability, quantification, and gas sensing have been achieved on the substrates, which hold great potential for the identification of trace analytes in real systems.
Although AgNRs-oxide array substrates have proved their extraordinary advantages, they are still in their infancy for practical SERS applications, and several challenges are yet to be addressed. First, AgNRs-oxide hybrid arrays possess satisfactory reproducibility and reliability exceeding the performance of commonly used nanoparticles, which are crucial for biosensing. The qualitative and quantitative detection of DPA, an anthrax biomarker, has been achieved on the AgNRs-Al 2 O 3 substrate. As for the SERS-based immunoassay and nucleic acid detection, AgNRs-oxide arrays offer potent capture substrates to bind SERS nanotags. The close proximity of the substrate-nanotag interface provides additional plasmonic coupling that further increases the SERS enhancement. It is thus urgent to demonstrate the SERS determination of bacteria, viruses, nucleic acids, and proteins on AgNRs-oxide arrays before the substrates can be routinely applied. Second, surface functionalization [102][103][104] should be realized on AgNRs-oxide substrates to further boost their widespread employments. Third, because different oxide materials possess various unique features according to their chemical or physical properties, further exploration of the diverse functions of AgNRs-oxide nanocomposites is still in great demand, which is essential for developing multifunctional SERS sensors in certain conditions. Moreover, aside from oxides, combining AgNR arrays with other attractive materials, e.g., graphene [105][106][107] and MoS 2 [108,109], is another advisable method for future SERS developments. Overall, along with the growing study, deeper understanding, and better optimization of AgNRs-oxide nanostructures, these substrates hold great promise as affordable and portable SERS sensors, and would open up a new era for practical SERS applications.