Preparation and characterization of Ag-nanostars@Au-nanowires hierarchical nanostructures for highly sensitive surface enhanced Raman spectroscopy

In this work we study the surface enhanced Raman scattering (SERS) produced by hierarchical nanostructures obtained by coupling different anisotropic nanomaterial of two SERS active metals, namely Ag nanostars (AgNSs) and Au nanowires (AuNWs). Ag nanostars (AgNSs) are prepared, by a two-step one-pot synthesis by reduction of AgNO3 with hydroxylamine, trisodium citrate and NaOH. AuNWs are obtained by electroless templated synthesis in track-etched polycarbonate membranes with following etching of the template. The two precursors are bound together by bridging with the bifunctional cysteamine molecule, obtaining AgNS@AuNW hierarchical structures. Benzenethiol (BT) is adsorbed on the nanostructured material and used as SERS probe to study the amplification of Raman signals. Experimental results indicate significantly larger Raman enhancement when BT is adsorbed onto the AgNS@AuNW in comparison to AuNWs alone or decorated with quasi-spherical silver nanoparticles obtaining AgNP@AuNW. Digital simulations performed by the boundary element method agree with the experimental findings, showing higher number of hot spots and significantly higher SERS enhancements for AgNS@AuNW versus AuNWs or AgNSs or AgNP@AuNW.


Introduction
Ultrasensitive detection of Raman-active molecules is increasingly studied with focus on the development of chemical sensors which exploit the surface enhanced Raman scattering (SERS) [1,2]. Many research efforts are directed towards the development of highly sensitive SERS substrates based on metal nanostructures [3,4], in some cases, hierarchically organized [5][6][7]. In particular, it was observed that the Raman enhancement generated by spherical nanoparticles (NPs) can be improved using nanomaterials with different shape, especially nanostructures characterized by anisotropic features [8,9]. For instance, large SERS enhancements have been observed using gold sphere segment void nanostructures [10,11] or arrays of macroporous nanostructures produced on bundles of optical fibers [12]. Very high amplification effects have been observed with anisotropic nanoparticles such as silver nanoflowers [13,14] and star-shaped NPs, including gold [15,16] or silver nanostars [17] (AgNSs). AgNSs appear particularly attractive since they produce important SERS enhancements and can be easily synthesized using quite simple one-pot chemical procedures [18]. Because of their starry shape composed of a central core and many pods, these nanoparticles show interaction with UV-vis-NIR light in particular in two spectral regions, namely in the UV and in the NIR. These factors favor a strong enhancement of the Raman scattering [19,20].
On these bases, in the present work we examine the preparation and application of SERS substrates obtained by combining synergistically the properties of AgNSs and AuNWs to prepare hierarchical nanostructures, here named AgNS@AuNW. The two precursor nanomaterials are prepared separately, by using hard template [35,36] and soft template [18] synthesis for AuNWs and AgNSs, respectively. They are assembled together in the hierarchical structures by cysteamine bridges [37,38]. After careful characterization of the obtained nanostructures by spectroscopic and electron microscopy techniques, the SERS amplification potentialities of AgNS@AuNW are tested with respect to the detection of benzenethiol (BT), used as a reference Raman probe. The SERS effectiveness of AgNS@AuNW is evaluated experimentally by comparing the Raman spectra recorded for BT adsorbed onto AgNS@AuNW versus AuNWs with or without decoration with isotropic silver nanoparticles (AgNPs). Experimental results are discussed and interpreted using digital simulations performed using the boundary element method [39,40] which show that higher number of hot spots and significantly higher SERS enhancements characterize AgNS@AuNW structures versus AuNWs and AgNP@AuNW.

Chemicals and materials
All the chemicals used were analytical grade and purchased from Merck-Sigma-Aldrich. Polycarbonate filter membranes, coated by the producer with the wetting agent polyvinylpyrrolidone, were supplied by SPI-Pore Filter (47 mm filter diameter, filter thickness 6 μm, 80 nm nominal pore diameter, 6×10 8 pores/cm −2 pore density). Double distilled water was used to prepare all the solutions.

Synthesis of Ag nanostars (AgNSs)
AgNSs were synthesized using the method of A. García-Leis et al [18]. Briefly, 500 μl of 0.05 M NaOH solution and 500 μl of the hydroxylamine solution (18 μl of NH 2 OH at 50% w/v in 5 ml of double distilled water) are mixed in a flask and stirred for one minute (magnetic stirrer at 670 rpm). Afterwards, 9 ml of the 10 −3 M AgNO 3 solution are added and stirred for 5 min. Finally, 100 μl of 0.045 M trisodium citrate solution are dropped in the flask, stirring for approximately 15 min, i.e. up to a dark green color is developed. The completion of the growth of the AgNSs is performed by storage of the reaction vessel, without stirring in the dark for 48 h at room temperature. The final dispersion is characterized by a green color. TEM analyses indicate that the appearance of a grey-beige color is indicative of the unsuccessful preparation of a polydisperse system composed both of AgNSs and spherical AgNPs.

Synthesis of Ag nanospheres (AgNPs)
AgNPs were synthesized adapting the method described in the literature [41,42]. Briefly, 1·10 −3 M AgNO 3 solution is heated up to boiling then, 1 ml of 0.045 M trisodium citrate solution is added and the mixture is kept under reflux at around 100°C for one hour. During this procedure, a colloidal dispersion is formed, assuming a matt appearance and a light beige color.

Membrane template preparation of Au nanowires (AuNWs)
Ensembles of gold nanowires embedded in the polycarbonate (PC) template (2DNEEs) were prepared by the electroless membrane template procedure described previously [43][44][45]. Briefly, after dipping the PC membrane in methanol for 4 h, it was sensitized with Sn 2+ in a solution containing 0.026 M SnCl 2 and 0.07 M trifluoroacetic acid in 50:50 methanol:water for 45 min. After rinsing with methanol, the membrane was dipped in 0.026 M Ag[(NH 3 ) 2 ]NO 3 for 10 min. Successively, it was immersed in an Au plating bath containing 7.9×10 −3 M Na 3 Au(SO 3 ) 2 in M Na 2 SO 4 at 0°C. After 20 min, 0.625 M formaldehyde was added to the solution. Electroless deposition was allowed to proceed for 24 h. After this period the golden membrane was washed with water and immersed in 10% HNO 3 solution for 12 h. At the end of the electroless deposition, gold is deposited both within the pores and the outer faces of the membrane. Therefore, to expose the polycarbonate, the outer gold layer was removed from one face of the membrane by peeling with scotch tape (Scotch ® Magic™ by 3 M). In order to remove the polycarbonate of the templating membrane to expose the AuNWs, two different etching processes were applied, one to perform the partial and the other for the total etching. For the partial etching, the AuNWs, 2DNEEs were dipped for 30 s in CH 2 Cl 2 :EtOH 1:9 [44]. The total etching of the polycarbonate [46], was instead achieved by attaching a 2DNEE onto a Teflon support by means of Kapton adhesive tape and pouring pure dichloromethane several times on the tilted sample. Exposed AuNWs were cleaned by residues of PC by using a procedure recently described [47], based on the use of a cold atmospheric plasma jet system, namely, Stylus Plasma Noble produced by NADIR SrL (Italy). For the present application, the following parameters were employed: Ar/O 2 flux=8 l min −1 ; t=60 s; HV=4 W-Gain 40%; RF=15 W pulsed; WD=8 mm.

Preparation of the samples for the Raman analysis
The macro-gold slides and the AuNWs arrays employed for the benzenethiol (BT) detection were prepared as follows: 1. overnight immersion in 10 −2 M cysteamine solution in water; 2. overnight immersion in the colloidal suspension of nanoparticles (AgNPs or AgNSs); 3. overnight immersion in 10 −5 M BT solution in ethanol.
Careful washing with water was performed between the above steps. The obtained hierarchical structures resulted stable. It was observed that omitting step 1, the AgNPs and AgNSs that initially adhered to the AuNWs, were removed by washing with water or ethanol.

SERS and Raman spectroscopy
The Raman spectra were recorded using a Renishaw inVia Raman spectrometer equipped with a diode laser operating at 785 nm (300 mW max power), used at 0.1% of its power, with an acquisition time of 10 s, coupled with an optical microscope (Leica), with objective 20X/0.4. The post-processing of the spectra was performed with the OriginLab software. Because of some inhomogeneity of the examined samples, as usual for deposited samples, preliminary measurements were performed analyzing different areas. The spectra reported and compared in this work are those with the highest signals.

UV-visible-NIR absorption spectroscopy
UV-vis absorbance spectra were recorded with a Perkin-Elmer Lambda 40 spectrophotometer equipped with a Peltier-Elmer PTP6 (Peltier temperature programmer) apparatus. The samples were diluted 1:3 in water for the Ag colloids and in dichloromethane for AuNWs suspensions.

Electron microscopy
A JEOL 3010 high resolution electron microscope (0.17 nm point-to-point resolution at Scherzer defocus), operating at 300 kV, equipped with a Gatan slow-scan CCD camera (model 794) was used for transmission electron microscopy (TEM). Drops of colloidal solutions were deposited on Holey-Carbon Copper grids and let dry at room temperature before analysis.
Field-emission scanning electron microscopy (FE-SEM) analysis were performed using a Sigma-VP Carl Zeiss electron microscope. Solid samples were immobilized on stubs by means of conductive tape in order to prevent charging effects.

Dynamic light scattering (DLS)
DLS measurements were performed at room temperature with a Zetasizer Nano ZS (Malvern Instruments) equipped with a green laser (λ=532 nm) and backscatter detector at 175°.

Boundary element method (BEM) simulations
The digital structures used to simulate the nanomaterials in this study were produced by a 3D design modeling software (Blender 2.79b) and then imported in Matlab environment. The minimum distance between Ag star and Au NW was fixed at 0.5 nm, as it is reported to be the minimum distance avoiding complex coupling effects [48]. The attribution of the physical properties for the nanostructures and the environments, as well the spectra and local fields simulations, were done using the MNPBEM libraries developed by U. Hohenester et al [39,40]. The exciting fields (with excitation source at 785 nm) are considered as propagating along the x, y and z axes and polarized along the two other normal directions for each case. The results are an average of the different field excitations, simulating the situation of a nanostructure at a random orientation with respect of the incoming field. The Enhancement Factors are calculated as the forth power of the local fields calculated at the surface of the simulated structures, given unitary the intensity of the exciting electromagnetic field. This can be done because the frequency of the exciting light and that of the scattered light are very similar, being the difference that of a vibrational frequency.

Characterization of AgNSs, AgNPs, AuNWs precursors
The different nanomaterials used here as precursors for the preparation of the desired hierarchical nanostructures have been synthesized using known procedures (see Experimental), however in order to obtain specifically the morphological and physicochemical information required for this study, they have been carefully characterized in view of the achievement of the present purposes. Figures 1(A) and (C) show the typical TEM images of an individual AgNSs (A) and a group of numerous AgNSs (C), synthesized at pH 8.1. These nanoparticles are characterized by a central core of approximately 50 nm diameter and a variable number of pods arranged in an octahedral structure, reaching an overall dimension of approximately 200 nm. The growth of the anisotropic particle starts from spherical seeds of Ag generated by initial reduction with hydroxylamine, then, in the following 48 h, the pods grow, being directed by the presence of citrate ions [18] and NaOH [49]. For the sake of comparison, figure 1(B) shows the TEM image of a typical AgNP synthesized as described in the experimental section, which displays a quasi-spherical shape with a diameter of approximately 70 nm. Figures 1(C), (D) support the critical role of an accurate control of the pH for the correct growth of AgNSs; it shows that operating at pH>8.1 (e.g. pH=11.4), a mixture of AgNSs+AgNPs is indeed obtained, as discussed previously in the literature [18]. In figure 2, the UV-vis-NIR extinction spectrum UV-vis-NIR of a colloidal suspension of AgNSs is reported and compared with the spectrum of a suspension of AgNPs. The comparison shows that Ag nanospheres present maximum extinction at ≈430 nm [42], followed by a marked decrease up to reaching zero at λ>800 nm. The AgNSs present an extinction maximum at 370 nm. However further absorption extended throughout the NIR region can be observed, reaching extinction values comparable to the maximum at 370 nm.
These features agree with the particular shape of the nanostars [18]: the extinction at 370 nm is caused by the central core while the several pods produce the broad absorption band in the NIR region at λ>780 nm. This extinction is influenced by the number and the geometry of the pods and it can even generate a sharp band in some cases [15].
The average size and polydispersity are evaluated by dynamic light scattering (DLS) measurements [50]. Data reported in figure 3, indicate an average diameter of 200 nm for the AgNSs (figure 3(a)) and 70 nm for the AgNPs. The polydispersity index (PdI) presents a lower value for the nanospheres (0.258) and a higher value for the nanostars (0.504). Both AgNSs and AgNPs exhibit a negative zeta potential, −44 and −33 mV respectively, which can be attributed to the presence of residual citrate. Both the obtained colloidal dispersions of AgNPs and AgNSs, are stable and do not tend to aggregate. Figure 4(A) presents the typical extinction spectrum of AuNWs dispersed in dichloromethane, which is characterized by a maximum at λ = 532 nm, in agreement with previous literature findings for similar nanomaterials [51]. Figure S1 is available online at stacks.iop.org/NANOX/1/020006/mmedia in the Supporting Materials (SM) schematizes the morphology of the ensembles of AuNWs at different steps of their preparation [43][44][45]. Figure 4(B) shows the surface of a 2DNEE after peeling the outer gold layer, with the nanowires still inside the membrane pores. Finally, figure 4(C) presents the AuNWs completely exposed after abundant washing with CH 2 Cl 2 . Eventual traces of residual polycarbonate of the template were removed by using a cold atmospheric plasma jet system [47]. The obtained nanowires results approximately 6 μm long, that is the thickness of the templating membrane. The AuNWs are not perfectly cylindrical, instead, they tend to assume a cigar-like shape, with a maximum diameter around 120-150 nm, since they adapt to the cigar-like shape of the track-etched pores in polycarbonate membranes [44], The high enlargement FE-SEM image in figure 4(C) confirms what discussed in previous studies [45], i.e. that the templated electroless deposition of gold in track-etched membranes starts with the formation of small Au nuclei formed in the inner wall of each nanopore. These nuclei successively grow under the shape of a nanowire formed by stacked nanospheres.

Preparation and FE-SEM characterization of AgNSs@AuNWs hierarchical structures
Cysteamine is used as bridging ligand suitable to bind the silver nanoparticles onto the gold nanowires. Indeed, the sulfhydryl group of the thiol has a great affinity for gold while the amine group binds silver. Note that an excess of cysteamine is used in order to prevent the possibility that both the sulfhydryl and the amine groups of the same molecules could bind only the AuNW surface, so that a compact cysteamine monolayer is obtained with the amine functionalities protruding out of the Au surface [37,38]. Figure 5 reports the FE-SEM analyses performed after treating the ensemble of AuNWs with cysteamine and AgNSs as described in the experimental section. These images show the presence of the AgNSs bound onto the surface of the AuNWs, in part dispersed and isolated and in part as clusters of nanostars. The latter areas are the spots on which the laser beam used for SERS measurements was preferentially focused.

SERS characterization with benzenethiol as Raman probe
The efficiency of several SERS substrates was assessed by employing benzenethiol as reference Raman-probe [12,[52][53][54], whose Raman bands are well known from the literature (see table S1 in SM for detailed information). Figure 6 presents the spectra obtained on ensembles of totally etched AuNWs, without or with AgNPs or AgNSs. On the partially etched AuNWs, BT cannot be detected (data not shown), while on totally etched AuNWs the typical Raman spectrum of this BT probe is clearly revealed. This indicates that AuNWs, fully free from the polycarbonate, already act as good SERS substrate. Signals with similar intensity are detected when the nanowire array is decorated with quasi-spherical Ag nanoparticles (i.e. AgNP@AuNW). On the contrary, figure 6 indicates that a large enhancement in SERS signals is observed when BT is absorbed onto the AgNS@AuNW hierarchical nanostructures.
These evidences clearly indicate the superior SERS performances of AgNSs when assembled over AuNWs.
In order to quantify this effect, SERS enhancement factors (EFs) are calculated from the spectra collected on different SERS substrates [55][56][57]. The number of Raman probe molecules adsorbed by the nanostructures here studied can indeed be very different, depending on the number and dimension of nanomaterial used. Therefore, the calculations of the SERS enhancements factors are performed calculating, at first, the approximate surface area of AuNWs, AgNSs, AgNPs and relevant hierarchical structures, using the geometries and particle density parameters evaluated from FE-SEM and TEM images. Knowing the different surface areas and the laser spot diameter, taking into account that the benzenethiol molecular area is 0.22 nm 2 [57], by using equation (S1) in SM [56], it is possible to estimate the EF values. The Raman spectrum of pure BT liquid is employed as non-SERS reference. A detailed description about the calculation is presented in the SM section and all parameters and results are resumed in table S3. The surface selection rules of SERS may results in different enhancements due to the different orientation of the molecules within hot spots [56]. For this reason, two different bands of BT were used for the calculations, at about 1000 and 1025 cm −1 , and band integrals were used instead of band height in order to overcome little band shifts found through different substrates. The calculated EF, for the 1025 cm −1 band, are in the order of AgNS@AuNW (6.22 · 10 5 ), AgNP@AuNW (1.24·10 5 ) and AuNWs (9.75·10 4 ). These esteems are coherent with the experimental results and give a significant indication from the analytical point of view about the efficiency of the tested SERS substrates. It is important to underline that EFs are average values and they take into account the adsorbed molecules whose Raman effects experience a strong enhancement, especially in the so called hot spots.

Boundary element method simulations
Complex morphology of the hierarchical nanostructure above described can be a source of un-certainty when performing the quantitative evaluation of the exposed surface area.This prevents the precise calculation of the surface concentration of the Raman probe, necessary for the precise evaluation of SERS enhancement factors from experimental spectra. To overcome the limit of the relative evaluation achieved by comparing EFs values, we chose to model and compare the localized intensity of electromagnetic fields in the nanostructures by numerical calculations performed using the boundary element method (BEM). The BEM approach assumes a dielectric environment where bodies with homogeneous and isotropic dielectric functions are separated by abrupt interfaces. It can be employed in the field of plasmonics, in which metallic nanoparticles are embedded in a dielectric background. It presents the advantage that only the boundaries between the different dielectric materials have to be discretized and not the entire volume, performing faster simulations with low memory requirements [39,40]. Figure 7 shows in an arbitrary color scale the SERS enhancement factors (EFs) obtained by BEM for AgNS (figure 7(A)), AuNW (figure 7(B)) and the coupling of the two nanostructures in AgNS@AuNW (figure 7(C)), using an excitation source at λ=785 nm. The EFs are calculated as the forth power of the local fields calculated at the surface of the simulated structures, being unitary the intensity of the exciting electromagnetic field. The enhancements calculated in the AgNS hot spots are of the order of 10 4 , but only in very tiny areas, whereas the one obtained for AuNW is order of magnitude smaller (10 2 ). This shows that the expected enhancements for isolated AgNS and AuNW are small, in particular for the latter. Calculations for the assembled AgNS@AuNW show, on the other hand, that the enhancements increases to a range of 10 5 , in more extended areas. The absence of sharp tips, or edges, in both AgNS and AuNW avoid the local fields to concentrate in single features alone. One can conclude, therefore, that the hierarchical system composed by these two nanomaterials optimizes the enhancement by creating hot spots on large areas due to the proximity of the two. The BEM methodology has been applied also to spherical AgNPs bound to AuNWs in the hierarchical structure AgNP@AuNW. The maximum enhancement (see figure 7(D)) is in the order of 10 4 and confined in very limited areas. Comparison between figures 7(C) and (D) clearly indicates a significantly larger number of hot spots for AgNS@AuNW. The results of the BEM simulations confirm what observed experimentally that is the combination of AgNSs with AuNWs acts synergistically in producing a larger number of hot spots with higher SERS enhancement then for the case of AgNSs, AuNWs or AgNP@AuNW structures.

Conclusions
The results here presented fully confirm the expectation that binding together anisotropic nanostars with high aspect ratio nanowires acts synergistically to obtain nanonostructures able to produce important SERS enhancement, in excess to 10 5 . Interestingly, these enhancements are significantly more relevant than those characterizing the individual components (i.e. AgNSs and AuNWs) of the final hierarchical nanostructures. It should be emphasized that, in the present study, the high SERS enhancements that characterize the AgNS@AuNW nanostructures were verified both experimentally and by model calculations. In particular, BEM calculations put in evidence that the AgNS@AuNW hierarchical structure not only is characterized by an enhancement factor of the order of 10 5 , but also presents a high number of hot spots, making possible the detection of analytes in low concentration. Although the decoration of AuNWs is not even, the obtained results represent a clear proof-of-concept confirming the SERS effectiveness of the AgNS@ AuNW combination. Efforts in obtaining more even decoration are in progress.