Porous Silicon Photonic Crystals Coated with Ag Nanoparticles as Efficient Substrates for Detecting Trace Explosives Using SERS

Picric acid (PA) is an organic substance widely used in industry and military, which poses a great threat to the environment and security due to its unstable, toxic, and explosive properties. Trace detection of PA is also a challenging task because of its highly acidic and anionic character. In this work, silver nanoparticles (AgNPs)-decorated porous silicon photonic crystals (PS PCs) were controllably prepared as surface-enhanced Raman scattering (SERS) substrates using the immersion plating solution. PA and Rhodamine 6G dye (R6G) were used as the analyte to explore the detection performance. As compared with single layer porous silicon, the enhancement factor of PS PCs substrates is increased to 3.58 times at the concentration of 10−6 mol/L (R6G). This additional enhancement was greatly beneficial to the trace-amount-detection of target molecules. Under the optimized assay condition, the platform shows a distinguished sensitivity with the limit of detection of PA as low as 10−8 mol/L, the linear range from 10−4 to 10−7 mol/L, and a decent reproducibility with a relative standard deviation (RSD) of ca. 8%. These results show that the AgNPs-modified PS PCs substrates could also find further applications in biomedical and environmental sensing.


Introduction
In recent years, ultra-sensitive detection of trace amounts of explosive chemicals has attracted widespread research attention due to the importance of these chemicals in many practical applications, including terrorist attacks, homeland security, and environmental protection [1][2][3][4][5]. Most explosives have low room-temperature vapor pressure; for example, the saturated vapor pressure of 2,4,6-trinitrotoluene (TNT), 2,4-dinitrotoluene (DNT), and picric acid (PA) are 9, 411, and 0.97 ppb, respectively [6]. Therefore, it is difficult to detect explosives quickly and accurately using gas sensors. Ion mobility spectrometry technology has distinct advantages in detecting volatile compounds and is often used to detect trace amounts of some explosives quickly and accurately [7]. Nevertheless, the poor selectivity and poor quantitative analysis hinder further applications of this technology. There has been a consequent demand for a simple, selective, and sensitive quantitative analysis method for the detection of trace explosives [8]. Surface-enhanced Raman spectroscopy (SERS), as a powerful chemical and biological analytical technique, has received much attention from researchers owing to its high surface sensitivity and molecular specificity [9,10]. Some researchers have used this technology to detect trace explosives [11][12][13][14].
Over the last few decades, owing to its unique optical and physical characteristics, porous Si has been widely used as sensors, such as biological sensors [15], electrochemical sensors [16],

Materials and Reagents
AgNO 3 (99.8%), hydrofluoric acid, PA, and alcohol were purchased from Sinopharm Chemical Reagent Co., Ltd. (Shanghai, China) Rhodamine 6G was purchased from Shanghai Aladdin Bio-Chem Technology Co., Ltd. (Shanghai, China) Monocrystalline Si was purchased from the Tianjin Institute of Semiconductor Technology (Tianjin, China). All the chemicals were of analytical grade and were used without further purification. Wahaha pure water (deionized water) (TDS ≈ 0) was used for all the preparations.

Preparation of Porous Silicon Photonic Crystals
The highly-doped crystalline silicon (c-Si) p-type wafers with a resistivity of 0.02 to 0.06 Ω·cm were first cleaned ultrasonically several times in ternate baths of alcohol and acetone, before being dried in a nitrogen environment. Porous Si was prepared using an anodic electrochemical etching method, as reported in our previous papers [31]. In brief, the process was conducted in a homemade double cell etch tank, as illustrated in Scheme 1. The electrolyte was a mixture of aqueous HF (48% by weight) and alcohol (99.7% by volume) in a concentration ratio of 1:1. Current densities of 230 mA/cm 2 for 1 s and 130 mA/cm 2 for 1.5 s were applied alternately for the fabrication of multilayer porous silicon. The total etching time was 12 cycles. After etching, the samples were repeatedly washed with pure alcohol and distilled water immediately. Then, the porous Si samples were placed in an alcohol solution for subsequent use.

In Situ Synthesis of Silver Nanoparticles on Porous Silicon Photonic Crystals
Ag deposition on the PS PC samples was performed by using an immersion plating solution. Due to the abundant Si-H bonds on the porous silicon surface, Ag + ions can be easily reduced to AgNPs. Thus, the AgNPs obtained without any chemical contamination present a great advantage due to the simplicity of operation and the fact that no toxic chemical reagents are introduced. Therefore, the immersion plating method is a simple, popular way to produce AgNPs on a substrate. Since the thickness of AgNPs can be adjusted by setting the immersion plating time, the longer the immersing time, the greater the yield of AgNPs is. AgNP arrays were prepared by using immersion plating solutions, according to the procedure reported by Zeiri et al. [32]. In brief, the 1 mM AgNO 3 precursor solution was prepared in an alcohol solvent. The PS PC samples were completely immersed in the AgNO 3 solution and then dried in air. In this study, five types of samples were obtained at five different immersion plating times: 30, 45, 60, 75, and 90 s. The AgNP-decorated PS PCs were washed with Wahaha pure water before SERS applications.

Characterization
Morphological and structural investigations of PS PCs were respectively performed using a field-emission scanning electron microscope (FESEM, Hitachi S-4800, Hitachi, Tokyo, Japan) and a Bruker D8 Advance Diffraction diffractometer (Bruker AXS, Karlsruhe, Germany) in the 2θ range from 10 • to 80 • , with Cu Kα radiation (λ = 0.15405 nm) at 40 kV and 40 mA. The reflectance spectra of the PS PC samples were obtained at room temperature by using a Hitachi U-4100 spectrometer (Hitachi, Tokyo, Japan).

Raman Measurement
SERS spectra were obtained using a HORIBA micro-Raman spectrometer (HORIBA Jobin Yvon, Kyoto, Japan) equipped with 50-mW Nd:YAG laser of 532-nm light and He-Ne laser of 633-nm light as excitation radiation. Low laser power of 1 mW was used to avoid sample heating effects, and the laser spot focused on the sample surface was about 2 µm in diameter. The SERS spectra of R6G were accomplished with an excitation wavelength of 633 nm, while the SERS spectra of PA were accomplished with an excitation wavelength of 532 nm. The data acquisition time was 3 s for each accumulation. A drop of (10 µL) of the R6G or PA alcoholic solution (analyte) was pipetted out onto the PS PC substrate and was dispersed to a circular shape area~0.5 cm 2 . The same sample preparation method was used for all concentrations of R6G and PA. The reproducibility was evaluated at 5 randomly chosen spots on the PS PC substrate. The Si-Si Raman peak of 521 cm −1 was used to calibrate the spectrometer. Figure 1 shows the SEM images of the surface and cross-section of the PS PCs. The surface structures of the freshly prepared PS PCs are shown in Figure 1a. Top-view observation revealed that the PS PCs have a uniform distribution of pores with a diameter of 20 to 50 nm, estimated from the SEM images. The cross-section SEM images, as shown in Figure 1b, showed the hierarchy of 12 cycles and that the thickness of porous Si was~2 µm. The layers with high and low porosities were formed by anodic electrochemical etching using large and small current densities, respectively. Figure 1c-f shows the AgNP-decorated porous Si obtained with different soaking times (30-75 s). These images revealed that with increasing soaking time, the number of AgNPs increased and the gap between the nanoparticles reduced. After soaking for more than 75 s, many overlapping AgNPs were observed, as shown in Figure 1f.

Reflectance and Raman Spectrum
Samples previously analyzed by FESEM were subjected to specular spectral reflectance measurements to check the optical response of the Ag-PS nanostructures. We have presented in Figure 2b the frequency evolution corresponding to Ag deposition time ranging from 30 to 90 s. With increasing Ag deposition time, the photonic bandgap gradually disappeared. This may be

Reflectance and Raman Spectrum
Samples previously analyzed by FESEM were subjected to specular spectral reflectance measurements to check the optical response of the Ag-PS nanostructures. We have presented in Figure 2b the frequency evolution corresponding to Ag deposition time ranging from 30 to 90 s. With increasing Ag deposition time, the photonic bandgap gradually disappeared. This may be attributed to the increase in AgNP concentration and the surface becoming rougher. This observation also reveals the presence of penetration of the porous silicon. However, the existence of interference fringes was observed up to an Ag deposition time of 75 s. When the deposition time was 90 s, almost no interference effect was observed. These results are in good agreement with the SEM results. Figure 2a shows the Raman spectra of the PS PC samples with and without silver nanoparticles coating. It can be seen that both types of samples showed a Raman peak at 520 cm −1 . The symmetrical Raman peak indicated almost no stress change in the as-prepared PS PCs. However, the AgNP-decorated samples show an asymmetry in the low-frequency side, compared to crystalline Si; this downshift of the phonon frequency was related to the width of the pores [20]. Redshift and broadening of absorption peaks were also observed. The shifts in absorption band and asymmetry can be attributed to a quantity-size effect and the surface effect [33], while the peak broadening was attributed to an increase in local temperature [34]. These results indicate that the local temperature is more likely to increase on AgNP-decorated porous Si.

Raman Scattering
As is well known, SERS is a powerful fingerprint analytical technique to provide real-time onsite detection for biological and chemical sensing owing to its advantageous non-destructive and non-invasive nature [7]. Here, we use R6G and PA as Raman probes to investigate the SERS-active substrate, whose Raman-active modes are summarized in Table 1.

Raman Scattering
As is well known, SERS is a powerful fingerprint analytical technique to provide real-time onsite detection for biological and chemical sensing owing to its advantageous non-destructive and non-invasive nature [7]. Here, we use R6G and PA as Raman probes to investigate the SERS-active substrate, whose Raman-active modes are summarized in Table 1.
It is known that AgNP size depends on the deposition time. Thus, the optimal immersing time can be obtained experimentally. Figure 3 illustrates the comparative SERS intensities of R6G molecules (10 −5 mol/L) obtained on PS PC samples. The typical bands of R6G were observed at 612, 772, 1087, 1127, 1189, 1312, 1363, 1509, and 1651 cm −1 in the first four cases, which were well indexed to the C-C in-plane bending in xanthene rings, C-H out-of-plane bending, C-H in-plane flexural stretching vibration, C-H out-of-plane vibration in xanthene rings, hybrid-mode associated with the NHC 2 H 5 group and xanthene rings, and C-C ring stretching to xanthene rings, respectively [19,37]. When the deposition time/immersing time exceeded 90 s, it was difficult to observe the Raman peaks of R6G because the layer of AgNPs was too thick, which hindered the electromagnetic interaction between Si and AgNPs [38]. The SEM image, shown in Figure 1f, confirms this result. The results show that the sample prepared with an immersing time of 60 s has the largest Raman enhancement effect.   Figure 4 shows the SERS signals of R6G (concentration 10 −3 -10 −10 mol/L) adsorbed on the AgNP-decorated PS PC samples, upon excitation at 633 nm. Although the Raman intensity largely decreased upon reducing the R6G solution concentration, the main characteristics of R6G molecules can be identified even when the solution concentration was as low as 10 −10 mol/L. A 10-μL reference sample with 10 −5 mol/L R6G deposited on a glass slide was measured. The constructive Raman band at 1512 cm −1 was selected to estimate the SERS enhancement factor (EF) based on the following Equation 1 [39]: where ISERS is the SERS intensity of R6G on PS PCs and INR is the normal Raman intensity of R6G at the same band. NNR and NSERS represent the corresponding numbers of molecules in the focused incident beam on the substrate, which can be estimated using a reported method [40]. In our experiments, the focus area of the laser beam is ~4 μm 2 . A 10-μL drop of R6G (10 −5 mol/L) contains 6 × 10 6 molecules. Thereby, the EF was calculated to be 4.3 × 10 6 . The EF for AgNPs decorated on monolayer porous silicon (PS) is 1.2 × 10 6 . SERS results demonstrated that nanoparticles with a molecular-dimension gap between the particles led to significant enhancement in Raman scattering of the linked molecules and nanoparticles. A linear dependence was observed between the logarithmic concentrations of R6G and the intensities of the fingerprint peaks (1512 cm −1 ), see Figure  4b, as per the following equation: where y is the peak intensity of the SERS spectra of R6G and C is the R6G concentration. Therefore, it can be concluded that AgNP-decorated PS PCs can be applied as sensitive SERS substrates for the quantitative detection of trace R6G. It can also be seen that the results show excellent linearity (R 2 = 0.9903) over a wide concentration range.
where I SERS is the SERS intensity of R6G on PS PCs and I NR is the normal Raman intensity of R6G at the same band. N NR and N SERS represent the corresponding numbers of molecules in the focused incident beam on the substrate, which can be estimated using a reported method [40]. In our experiments, the focus area of the laser beam is~4 µm 2 . A 10-µL drop of R6G (10 −5 mol/L) contains 6 × 10 6 molecules. Thereby, the EF was calculated to be 4.3 × 10 6 . The EF for AgNPs decorated on monolayer porous silicon (PS) is 1.2 × 10 6 . SERS results demonstrated that nanoparticles with a molecular-dimension gap between the particles led to significant enhancement in Raman scattering of the linked molecules and nanoparticles. A linear dependence was observed between the logarithmic concentrations of R6G and the intensities of the fingerprint peaks (1512 cm −1 ), see Figure 4b, as per the following equation: y =2133 +184lgC (2) where y is the peak intensity of the SERS spectra of R6G and C is the R6G concentration. Therefore, it can be concluded that AgNP-decorated PS PCs can be applied as sensitive SERS substrates for the quantitative detection of trace R6G. It can also be seen that the results show excellent linearity (R 2 = 0.9903) over a wide concentration range. The AgNP-decorated PS PC nanostructures were used as SERS substrates to explore their application in quantitative analysis of trace PA. As is well known, PA is a common explosive that is widely used in the manufacture of acid dyes, photographic drugs, explosives, and pesticides, etc. Thus, PA seriously pollutes the environment through water and soil [41]. Here, PA was used as the target analyte to effectively verify the performance of the PS PC substrate. A 10-μL solution of a particular concentration was drop-casted on the PS PC substrate. With regard to the significant issue of reproducibility, we first observed the uniformity of the substrate. Figure 5 shows that the SERS signals were all of comparable intensity, indicating that the PS PC substrates exhibited uniform SERS enhancements over their entire surface. For the intensities of the characteristic peaks at 938, 1087, and 1393 cm −1 of PA in these five spectra, which were obtained at randomly selected spots under identical experimental conditions, the relative standard deviations (RSD) were 8.16, 8.09, and 8.05%, respectively. This is probably due to the uniform coating of AgNPs on PS PC nanopore structures.  Figure 6a shows the SERS spectra for PA detection. The Raman intensity decreased upon reducing the PA concentration from 10 −4 to 10 −7 mol/L. The characteristic peaks, see Table 1, of PA at 786 cm −1 , 938 cm −1 (ring breathing), and 1085 cm −1 (phenolic C-O stretching) were observed at all the concentrations tested. Compared with the Raman characteristic peaks of PA itself, the Raman peaks The AgNP-decorated PS PC nanostructures were used as SERS substrates to explore their application in quantitative analysis of trace PA. As is well known, PA is a common explosive that is widely used in the manufacture of acid dyes, photographic drugs, explosives, and pesticides, etc. Thus, PA seriously pollutes the environment through water and soil [41]. Here, PA was used as the target analyte to effectively verify the performance of the PS PC substrate. A 10-µL solution of a particular concentration was drop-casted on the PS PC substrate. With regard to the significant issue of reproducibility, we first observed the uniformity of the substrate. Figure 5 shows that the SERS signals were all of comparable intensity, indicating that the PS PC substrates exhibited uniform SERS enhancements over their entire surface. For the intensities of the characteristic peaks at 938, 1087, and 1393 cm −1 of PA in these five spectra, which were obtained at randomly selected spots under identical experimental conditions, the relative standard deviations (RSD) were 8.16, 8.09, and 8.05%, respectively. This is probably due to the uniform coating of AgNPs on PS PC nanopore structures. The AgNP-decorated PS PC nanostructures were used as SERS substrates to explore their application in quantitative analysis of trace PA. As is well known, PA is a common explosive that is widely used in the manufacture of acid dyes, photographic drugs, explosives, and pesticides, etc. Thus, PA seriously pollutes the environment through water and soil [41]. Here, PA was used as the target analyte to effectively verify the performance of the PS PC substrate. A 10-μL solution of a particular concentration was drop-casted on the PS PC substrate. With regard to the significant issue of reproducibility, we first observed the uniformity of the substrate. Figure 5 shows that the SERS signals were all of comparable intensity, indicating that the PS PC substrates exhibited uniform SERS enhancements over their entire surface. For the intensities of the characteristic peaks at 938, 1087, and 1393 cm −1 of PA in these five spectra, which were obtained at randomly selected spots under identical experimental conditions, the relative standard deviations (RSD) were 8.16, 8.09, and 8.05%, respectively. This is probably due to the uniform coating of AgNPs on PS PC nanopore structures.  Figure 6a shows the SERS spectra for PA detection. The Raman intensity decreased upon reducing the PA concentration from 10 −4 to 10 −7 mol/L. The characteristic peaks, see Table 1, of PA at 786 cm −1 , 938 cm −1 (ring breathing), and 1085 cm −1 (phenolic C-O stretching) were observed at all the concentrations tested. Compared with the Raman characteristic peaks of PA itself, the Raman peaks of the PS PC substrates showed some redshift, attributed to the stress on the contact surface between  Figure 6a shows the SERS spectra for PA detection. The Raman intensity decreased upon reducing the PA concentration from 10 −4 to 10 −7 mol/L. The characteristic peaks, see Table 1, of PA at 786 cm −1 , 938 cm −1 (ring breathing), and 1085 cm −1 (phenolic C-O stretching) were observed at all the concentrations tested. Compared with the Raman characteristic peaks of PA itself, the Raman peaks of the PS PC substrates showed some redshift, attributed to the stress on the contact surface between silicon and silver [42].   Table 2, the results obtained herein indicated that PS PCs could be used as a potential substrate for SERS-quantitative detection. This SERS platform showed high sensitivity, low detection limit, and wide linear range, which can be ascribed to the PS PC substrate. In general, SERS enhancement in the Raman signal from AgNP-decorated PS PCs is mainly correlated to the electromagnetic enhancement mechanism [43]. However, in this case, chemical enhancement may have played an important role in SERS enhancement. Porous Si is a new functional material with a sponge-like structure, based on nanocrystalline Si particles. Thus, this nano-dimensional semiconductor structure is very conducive to charge transfer between semiconductor nanoparticles and the target molecules [44]. The Si nanoparticles in the present AgNP-decorated PS PCs nanostructure may lead to enhanced electron charge transfer from the AgNPs to the target molecules, which in turn affect the SERS enhancement. For the PS PC nanostructure, the internal reflections in multilayers can enhance the Raman signals [45]. Compared to monolayer porous Si or glass substrate, PS PC substrates have an extraordinary scattering effect in response to the exciting light, and thus, it has a stronger photoelectric coupling effect. As shown in Figure 6, the AgNP-decorated PS PC substrates have a wider linear dynamic range (LDR) owing to them having a higher surface-to-volume ratio than porous silicon [46]. As is known, SERS intensity depends on hotspots. The high surface-to-volume ratio implied that the substrate has a larger surface area, which greatly increased the hotspot area and led to wide LDR.   Table 2, the results obtained herein indicated that PS PCs could be used as a potential substrate for SERS-quantitative detection. This SERS platform showed high sensitivity, low detection limit, and wide linear range, which can be ascribed to the PS PC substrate. In general, SERS enhancement in the Raman signal from AgNP-decorated PS PCs is mainly correlated to the electromagnetic enhancement mechanism [43]. However, in this case, chemical enhancement may have played an important role in SERS enhancement. Porous Si is a new functional material with a sponge-like structure, based on nanocrystalline Si particles. Thus, this nano-dimensional semiconductor structure is very conducive to charge transfer between semiconductor nanoparticles and the target molecules [44]. The Si nanoparticles in the present AgNP-decorated PS PCs nanostructure may lead to enhanced electron charge transfer from the AgNPs to the target molecules, which in turn affect the SERS enhancement. For the PS PC nanostructure, the internal reflections in multilayers can enhance the Raman signals [45]. Compared to monolayer porous Si or glass substrate, PS PC substrates have an extraordinary scattering effect in response to the exciting light, and thus, it has a stronger photoelectric coupling effect. As shown in Figure 6, the AgNP-decorated PS PC substrates have a wider linear dynamic range (LDR) owing to them having a higher surface-to-volume ratio than porous silicon [46]. As is known, SERS intensity depends on hotspots. The high surface-to-volume ratio implied that the substrate has a larger surface area, which greatly increased the hotspot area and led to wide LDR.

Conclusions
In summary, we have presented new PS PC SERS substrates with a molecule concentrating effect by integrating AgNPs into the micropores of PS PCs. The AgNP-decorated PS PCs were prepared by an in-situ method by immersion in an AgNO 3 solution. The experimental results showed that this substrate has a sufficiently high detection sensitivity for the easy detection of R6G at concentrations as low as 10 −10 mol/L by a simple dipping method, and its enhancement factor reaches 1.2 × 10 6 at the concentration of 10 −6 mol/L. The detection of a biomolecule, e.g., PA, was also explored. A good linear response was obtained for PA concentrations ranging from 10 −4 to 10 −7 mol/L and the detection limit was as low as 10 −8 mol/L. More importantly, the RSD values of the Raman signals were~8%. Thus, the as-prepared nanostructures can be used to detect trace compounds with minimal analyte consumption using PS PC substrates.