Increasing gold nanostars SERS response with silver shells: a surface-based seed-growth approach

A straightforward method to prepare surface enhanced Raman spectroscopy (SERS) chips containing a monolayer of silver coated gold nanostars (GNS@Ag) grafted on a glass surface is introduced. The synthetic approach is based on a seed growth method performed directly on surface, using GNS as seeds, and involving a green pathway, which only uses silver nitate, ascorbic acid and water, to grow the silver shell. The preparation was optimized to maximize signals obtaining a SERS response of one order of magnitude greater than that from the original GNS based chips, offering in the meantime good homogeneity and acceptable reproducibility. The proposed GNS@Ag SERS chips are able to detect pesticide thiram down to 20 ppb.


Introduction
Since its discovery a few decades ago, surface enhanced Raman spectroscopy (SERS) has now become a well-developed method providing high sensitivity for the detection of a multitude of chemical and biological targets.Strongest SERS effect occurs by the excitation of localized surface plasmonic resonances (LSPR) typically generated in nanostructured noble metals surfaces, which can lead to huge enhancement of the Raman light scattered by molecules adsorbed or located close to the metal surface.This feature allows to overcome the main drawback of Raman spectroscopy (RS) i.e. the very weak scattering cross-section of the effect itself, and at the same time to retain the main advantages of RS, which can be summarized in specificity (with the recognition of the vibrational fingerprint of the molecule), speed and flexibility [1,2].
Amongst the factors influencing the extent of the enhancement, the shape of the metal nanostructure plays an essential role, together with the metal's nature.Nano-objects having anisotropic shapes and presenting edges, spikes, protruding branches can concentrate the electromagnetic field near these points, the so-called intrinsic 'hot spots' [3,4], exploiting the 'Lightning-rod effect'.In this frame, gold nanostars (GNS) have been introduced as a successful SERS substrate [5][6][7][8][9][10][11][12].Indeed, thanks to the sharpness of the GNS branches, a remarkably large enhancement of the electromagnetic field, and thus of the Raman signal, can be obtained in correspondence of the tips [13].
An effective trick to further boost the SERS effect is to add a shell of silver to GNS, a procedure introduced by Vo-Dinh a few years ago, exploiting the high plasmonic effects given by Ag when a proper thickness of its layer is used [14,15].Original content from this work may be used under the terms of the Creative Commons Attribution 4.0 licence.Any further distribution of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.
Since then, use of GNS@Ag as powerful SERS substrates has been attracting an increasing interest, and in the last few years encouraging results were obtained with this approach, allowing to reach low detection limits for the detection of analytes of environmental and biological interest [16][17][18][19].In all these studies, a common finding is that the highest Enhancement Factor (EF) is obtained when silver shell includes most of the GNS core, while the top of tips protrudes from the Ag shell.These results are also supported by theoretical calculations and modeling [14].
Recently, it was demonstrated that a precise optimization of the thickness of the Ag layer in GNS@Ag allows to gain a 7-fold signal amplification compared to the value obtained with pristine GNS prepared by a seed-growth method based on Triton-X-100 [20,21].This result was also corroborated by boundary element method calculations [22].The optimized GNS@Ag were used to prepare homogeneous and reproducible SERS chips, able to detect thiram, a pesticide with a well-known toxicity [23], down to 0.1 ppm.
The field of GNS@Ag has been explored by several recent investigations, using different synthetic approaches [18,24].Anyway, in all the reported examples of GNS@Ag based systems, the silver layer was grown on anisotropic gold objects, obtained by a seed-growth or one-pot method, adding to their colloidal suspensions proper mixtures of a silver salt precursor (usually silver nitrate) and ascorbic acid in various proportions.In all cases, the final SERS probes were obtained through a multistep procedure, often involving several purification and centrifugation steps of GNS and GNS@Ag, and further steps were required in the case of realization of SERS chips or for implementation on optical fibers [25,26].
In our continuous effort to produce trustable SERS chips we introduce here a more straightforward and clean preparation, lowering the number of steps and the use of reactants and solvents to obtain reproducible samples.The growth of silver shells is obtained directly on a monolayer of GNS grafted on a glass surface, as depicted in scheme 1.
We started with the synthesis of GNS that were subsequently grafted on glass chips with a standard procedure [6,27].After this, we proceeded with the immersion of glass chips bearing a monolayer of grafted GNS in a growing solution containing only silver nitrate and ascorbic acid.After each step, removal of excess of reactants, surfactants and byproducts can be obtained by simply washing the chips with water, eliminating further costly and time-consuming procedures (e.g.dialysis, centrifugationK).The SERS chips preparation was optimized in order to maximize their performances by carefully characterizing their properties by means of scanning electron microscopy (SEM), UV-vis-NiR and Raman spectroscopy.This new simple synthetic pathway allows to obtain SERS chips characterized by higher EFs with respect to GNS and by satisfying homogeneity and reproducibility, both tested using well known Raman reporters as described in the experimental section.
SERS chips were then tested for thiram detection.The peculiar molecular structure of thiram render its detection by SERS quite easy to perform, as strong covalent interaction takes place between S atom contained in the pesticide and noble metal atoms.Various SERS substrates for thiram detection based on core-shell Au@Ag plasmonic objects have been proposed [28,29], in some cases using GNS@Ag, allowing to recognize its presence below the ppm value [18,22] in other cases reaching ppb [16] values.In a recent work, attomolar detection limits were reached using multi spiked silver stars [30].

Instruments
UV-vis-NIR absorption spectra of colloidal suspensions and functionalized glasses were taken with a Varian Cary 6000i spectrophotometer in the 300-1800 nm range.Transmission Electron Microscopy (TEM): images of GNS samples were collected on a Jeol JEM-1200 EX II instrument on 1:100 diluted solutions.10 μl of this sample were dropped on nickel grids, 300 mesh, with a coating of a Parlodion membrane.
Dynamic Light Scattering measurements (DLS): Zeta potential measurements of GNS were performed on 1 ml of colloidal suspension with a Zetasizer Nano-ZS90 (source: polarized He-Ne laser, 30 mW output power, vertically polarized).
Scanning Electron Microscopy (SEM): SEM images were taken from a Tescan Mira XMU variable pressure Field Emission Scanning Electron Microscope-FEG SEM (Tescan USA Inc., USA), located at the Arvedi Laboratory, CISRiC, Pavia.Slides were fitted onto aluminium stubs, using double sided carbon adhesive tape.Then, they were made electrically conductive through the deposition of a thin layer of Pt/Pd (nm) while in vacuum.To have higher spatial resolution, images were gathered at 25 kV using In-Beam Secondary electron detector.
Raman and SERS: Raman and SERS measurements were performed at room temperature using an XploRa Plus (HORIBA Scientific) integrated confocal spectrometer equipped with an Olympus microscope BX43 with a motorized xy stage, acting as sample holder.The spectral resolution is about 1 cm −1 and the light signals are revealed by an Open Electrode CCD camera cooled down to −60 °C in a multistage Peltier air-cooling system.A 638 nm laser light (90 mW) was used as excitation source.Neutral filters with different optical density were used to set the proper incident laser power.
ICP-OES: ICP-OES measurements were performed using a ICP-OES OPTIMA 3000, PerkinElmer, USA.The total amounts of Au and Ag on GNS@Ag functionalized glass slides were determined by dipping each sample in 5 ml of diluted aqua regia (4:25 with double-distilled water) in a beaker, shaking them for 24 h on reciprocating shaker.After the removal of the glass, the resulting solution was diluted again (1:3, always with double-distilled water) and sent to ICP-OES measurement.

Synthesis of GNSs
GNS were prepared following the seed growth procedure previously described [20].Seeds were prepared in a vial, by adding 5 ml of Triton-X-100 aqueous solution (0.2 M), 5 ml of HAuCl 4 aqueous solution (4.5 × 10 −4 M) and then 600 μl of an iced-cooled aqueous solution of NaBH 4 (0.01 M).NaBH 4 was quickly added to the pale-yellow solution of AuCl 4 -prepared in the previous step.The resulting solution had a brown-orange colour and was stored in an ice bath.
Seeds should be used within 3 h.
The growth solution was prepared starting from 50 ml of a water solution of Triton-X-100 (0.2 M) and adding, under magnetic stirring, 2500 μl of AgNO 3 in water (0.004 M), 50 ml of aqueous HAuCl 4 (4.5 × 10 −4 M), 1700 μl of L-Ascorbic Acid in water (0.0788 M) and 120 μl of the seed solution previously prepared.After this last addition, the suspension turns from pink to purple and blue, finally giving a grey/dark green colloid.The mixing was then stopped.

Preparation of SERS-chips
In a typical preparation, 7 glass slides were prepared at the same time in a 9-place Hellendahl staining jar with glass lid.
• Functionalization of slides with APTES: samples were prepared according to a reported method [22,27,[31][32][33].Glass substrates were soaked for 30 min in freshly prepared Piranha solution , then washed three times with double distilled water (ddH 2 O) in a sonic bath, for three minutes each time, and then oven-dried at 120 °C.Thereafter, the slides were plunged again for 5 min in a 10% (v/v) solution of APTES in ethanol at 60 °C, washed two times with ethanol and one time with ddH 2 O in a sonic bath, for three minutes each time.At the end, the glass substrates were moved to a clean Hellendahl staining jar, free of silanes, and then they were blow-dried with nitrogen.• Functionalization of slides with a SAM of GNS: APTESfunctionalized glass slides were fully immersed in a GNS colloidal suspension for 15 h, keeping them under mild agitation on a roll and tilt stirrer.The samples were then washed three times, for three minutes each time, in ddH 2 O without sonication, again on a roll and tilt stirrer.After this step, samples were carefully dried under N 2 flux.
The functionalization of glass slides with GNS takes place thanks to an electrostatic interaction between our nanoparticles, with a negatively charged surface (ζpotential = −6(1) mV), and the terminal amino groups of APTES, that are positively charged.The amino moieties of APTES are positively charged in our work condition with GNS colloidal suspension at pH 3.
• Silver coating of Gold Nanostars (GNS@Ag) grafted on glass chips: relying on previous work on seed-growth synthesis performed on surface [34], once the GNS glass slides were dry, we proceeded with the formation of silver shells upon them.The growth of the silver shells was obtained by dipping the GNS glass samples in a growth solution of silver nitrate (AgNO 3 ) and L-ascorbic acid (total volume of 40 ml).Four final different concentrations of silver nitrate (AgNO 3 ) and L-ascorbic acid were tested, keeping their ratio always at the value of 2:1.Concentrations of reactants in this step are reported in table 1. Preliminarily, for the 'high' concentration of AgNO 3 and L-ascorbic acid, we tried out different immersion times (2 h, 4 h, 8 h and 24 h).

SERS testing of GNS and GNS@Ag glass slides
Relying on previous work [22], SERS response of GNS and GNS@Ag glass slides was tested using Rhodamine 6G (R6G), a well-known Raman reporter, as probe molecule: 80 μl of an aqueous solution of R6G (10 −5 M) were deposited on GNS and GNS@Ag glass slides.The SERS measurements were performed right after the deposition in order to prevent the drops to dry.The spectra were collected with a power density of 2 × 10 3 W cm −2 , a 50x magnification objective, an exposure time of 10 s and a number of accumulation equal to 5. Each sample was sampled in 10 different areas.

Preparation of SERS-chips for SERS substrate homogeneity test
The homogeneity of the SERS substrates was performed using 7-mercapto-4-methylcoumarine (MMC) as Raman reporter.GNS and GNS@Ag glass slides were immersed in a standard solution of MMC (10 −5 M) in ethanol for 1 h, washed three times with ethanol, without sonication on a roll and tilt stirrer, to remove any excess from the surface.Previously to the SERS homogeneity tests, they were dried under N 2 flux.SERS mapping was performed on both small and large scale.Automatic scanning acquisitions were used to collect spectra over areas of (70 × 50) μm 2 and (700 × 480) μm 2 by means of a 50x and 10x magnification objectives, thus with a spot size of about 4 μm 2 and 100 μm 2 , respectively.The power density was about 2 × 10 4 with the 50x magnification, while ≈9 × 10 3 W cm −2 for 10x.In both cases, a 2 s integration time and 5 accumulations were used.

Detection of thiram in ethanol
Thiram stock at 3.12 × 10 −5 mol l −1 (7.5 ppm) was prepared in ethanol and then diluted to the desired concentrations.SERS chips of GNS@Ag were immersed in ethanolic solutions of thiram for 1 h, washed copiously with ethanol for three times, without sonication, and dried under N 2 flux.SERS spectra were collected with a 100x magnification objective, a laser density power of 4.5 × 10 4 W cm −2 , an integration time of 10 s and a number of accumulations equal to 10.The collection was performed by automatic scanning along a line of about 30 μm and 10 spectra were recorded.The reported spectra for each sample are the average of these 10 acquisitions.

Results and discussion
3.1.Preparation of a GNS monolayer on glass GNS were prepared using a seed-growth approach using triton X-100 according to the described method.Synthetic details, characterization and features were widely described in previous works [7,20,22,[45][46][47][48][49][50][51][52][53][54].Here, we characterized the GNS colloidal solution with TEM and UV-vis-NIR spectra.In figure S1, TEM image shows the typical morphology of GNS, and the corresponding UV-vis-NIR spectra show the expected plasmonic feature close to 900 nm, that is relative to a longitudinal resonance along one single branch.A second one is located around 1500 nm and is due to longitudinal resonance across the whole objects, involving a couple collinear branches.
Grafting of a GNS monolayer on coverslip glass samples is obtained through formation of electrostatic interactions.Coverslips were functionalized with a monolayer of (3-aminopropyl)trimethoxysilane (APTES) using a described procedure [22,27,[31][32][33] and APTES terminated coverslips were then immersed in a colloidal suspension of GNS, which were used without any purification step.GNS in the colloidal suspension at native pH 3 have a Z-potential of −6 (1) mV, while at the same pH the amino groups on glass are protonated, hence the electrostatic interaction bring to GNS monolayer formation, which can be visually perceived as the coverslides assume a blue-grey colour.If one excludes the washing with bidistilled water and drying under nitrogen, no tedious or expensive purification steps are needed.GNS chips were characterized by UV-vis spectroscopy, SEM and SERS spectroscopy.
UV-vis spectroscopy (figure 1(a)) of GNS coated glass chips in the 300-1800 nm range showed the expected two LSPR features, with a sensible blue-shift compared to the same GNS in colloidal suspension (figure S1(a)), due to change of refractive index moving from water to air/glass interface, and probably also influenced by plasmonic conjugation between close GNS.
The enhancing capability of these GNS substrates were tested by using as probe molecule Rhodamine 6G, as described in detail in the previous section.An average EF of 4.0 × 10 6 was derived (see [55] for the detailed procedure of EF derivation), stating a moderate SERS activity.Figure 1(b) shows the comparison between the SERS spectrum of the 10 −5 M R6G solution and the Raman spectrum obtained with a 10 −3 M solution of R6G spread on blank glass, to evidence qualitatively the enhancement given by the GNS monolayer.Figures 1(c) and (d) show two SEM images at different magnification, demonstrating the good homogeneity of GNS monolayers and the conservation of shape and size of GNS after the loading on APTES-terminated chips.
The relative standard deviation (RSD) value associated to the height of the 610 cm −1 R6G mode is about 13%, in accordance with the RSD obtained in the homogeneity tests performed with MMC molecule.Indeed, both small-and large-scale mappings were carried out, showing an RSD value of about 10% in both cases.As an example, figure 2(a) reports  the MMC spectrum obtained as the mean of the multiple acquisitions within the small-scale map with its RSD of 10.8%, while in figure 2(b) the relative color map reporting the height of the MMC mode at 1170 cm −1 is shown.The results of the GNS large-scale map are reported in the supplementary data (figure S2(a)).It has been demonstrated [33,56], that with 10 −5 M concentration of MMC one can obtain a monolayer of molecules fully covering the available surface of GNS/GNS@Ag that is the appropriate situation for homogeneity tests.
The preparation is highly reproducible, as can be deduced from routinely performed spectra (taken in the 300-1100 nm range) for several preparations, which are showed in figure S3.

Growth of silver shells on GNS glass chips: preparation and optimization of GNS@Ag chips
To investigate the possibility of the formation of silver shells upon GNS grafted on glass chips, we started taking inspiration from previous works on seed-growth synthesis performed on surface [34] and on growing of a silver shell around GNS [22].In this case, our approach was to keep the preparation the simplest possible, with immersion for a given time of GNS chips in a growth solution containing a certain quantity of silver precursor (AgNO 3 ) and ascorbic acid (reductant) in a fixed ratio (2:1).
In a first experiment, four different immersion times were used, keeping constant the growing solution composition (1.0 × 10 −4 M for AgNO 3 , 5.0 × 10 −5 M for ascorbic acid).The resulting GNS@Ag chips were then characterized by UV-vis spectroscopy and SEM.
LSPR spectra show (see figure 3(a)) progressive blueshift of the 'long' resonance band, which moves from 1350 nm to 950 nm, and the growth of a wide feature in the 400-600 nm range.As already explained by boundary element method results on GNS@Ag in colloidal suspension, this is consistent with an increasing thickness of silver shells around the GNS core [22].This is confirmed by SEM images (see supplementary material, figure S4), which show that an increasing number of GNS is surrounded by silver shells of increasing thickness as the immersion time increases.
Quantification of thickness by morphological analysis is extremely difficult due to the variability of nano-objects features.Thus, we evaluated silver thickness by measuring the amounts of silver brought to the chips by the growing processes thanks to ICP-OES measurements, after digestion of the GNS@Ag chips in diluted aqua regia (see supplementary data table S3 and figure S5).It is worth of note that the pristine GNS monolayers contain a defined amount of silver, as silver is a fundamental reactant in GNS preparation.For each immersion time, the measurements in ICP-OES were done on three different samples, and then in table S3 and figure S5 we reported the mean value of Ag, Au and Ag/Au and their standard deviations.
Indeed, for long immersion times one can observe a situation in which, for the majority of the objects, only the final part of the branches is left outside the silver shell, a morphology which is known to maximize the SERS effects: the SEM image of the 24 h' growth sample is reported in figure 3(b).This behavior is consistent to what is reported in the previous literature, which has assessed how the growth of silver shells starts from the gold objects cores, to an extent which can be modulated leaving the desired part of the branches exposed [14,22,57].
The obtained chips are stable for weeks.As it was repeatedly demonstrated, when silver is present as shell or layer upon gold nano-objects, its stability to oxidation is greatly enhanced.In some cases, the introduction of silver layer is able to further stabilize the gold nanoparticles [25,58,59].
SERS characterization with the R6G 10 −5 M solution gave the expected results, with the R6G signals increasing with immersion time (figure 3(c)), a trend which follows silver shell thickness evidenced by SEM.An immersion time of 24 h allows reaching a plateau in silver shell growth and SERS signal increase, as evidenced by figure 3(d).
Having assessed a useful growth timescale, we investigated the effect of the precursor/reductant couple concentration in the growth solution, keeping their ratio always at the value of 2:1.We considered four different concentrations (named respectively 'low', 'medium', 'high' and 'very high') of reactants in the growth solution, ranging between 10 −5 M and 5.0 × 10 −4 M for AgNO 3 (and consequently between 5.0 × 10 −6 M and 2.5 × 10 −4 M for ascorbic acid), using an immersion time of 24 h.It is worth of note that during the preparation with the 'very high' silver/ascorbic acid concentration we saw quite immediately the appearance of turbidity in the growth solution, slowly disappearing during the 24 h' immersion.
For all samples, SEM images were taken, together with UV-vis-NIR spectra.Figure S6 (supplementary data) shows the UV-vis-NIR spectra obtained for four chips coming from the four different preparations at different concentration of growth solutions.As can be clearly observed, the variation of LSPR features is consistent with the formation of a silver shell of increasing dimensions when the concentration of silver/ ascorbic acid is raised up to 'high' (1.0 × 10 −4 M/5.0 × 10 −5 M, black line), as reported above.When the concentration is raised further to the 'very high' one (red line), the trend is interrupted, with a spectrum which does not show any further blue-shift of the so called 'long' band but just a broadening of NIR absorption.Comparing the SEM images for all the samples obtained at different growth solution concentrations (figure S7 in the supplementary data), one can observe a progressively increasing amount and thickness of silver around GNS, increasing concentrations up to the 'high' value (figures S7(a)-(d)), while no silver shell can be apparently observed for the 'low' investigated concentration (figure S6(b)).It also qualitatively suggests that increase in Ag+/ascorbic acid concentration at the 'very high' value does not produce a significant variation in the extent of the silver shell and the general morphology of GNS@Ag, at the same time denouncing the presence of silver nanoprisms randomly distributed (200 nm edges length, see images in figures S7(e) and (f)).We can tentatively explain these findings with the presence of nucleation and growth of large silver nanoprisms in the growing solution due to the too high concentration of silver ion precursor and ascorbic reductant.This also gives an explanation of the reported turbidity in the growing solution, since large objects forming in the growth solution can reasonably act as scattering centers for visible light.As described above, this turbidity slowly disappears during the immersion, probably because of the sedimentation in the growing jars and/or deposition on the chips (as showed by SEM images S7(e) and (f)).The GNS@Ag chips, as described above, were then used to register SERS spectra of 10 −5 M R6G solutions.Representative spectra as a function of concentrations used in the growth solution are reported in figure 4(a).Again we evaluated, for all the samples, the quantity of silver and gold on chips with ICP-OES (see table S4 in supplementary data), and calculated the Ag/Au ratio as an indication of the mean silver shell extent.
Figure 4(b) shows the Ag/Au ratio from ICP-OES (the exact values and the EFs calculated from SERS spectra reported in figure 4(a) (red circles), as a function of the concentration of silver ions in the growth solutions used in the chips' synthesis with 24 h of immersion.The increased concentration of silver ion/ascorbic acid in the growth solution produces an increase in silver shell size, demonstrated by the increased amount of silver revealed by ICP on chips surface, which correlates nicely with the increase in EF.ICP data for the 'very high' value of silver ion/ascorbic acid in growth solution only show a slight increase in silver surface concentration when compared to the 'high' samples, confirming what already observed in SEM images.This strong increase in silver ion/ascorbic acid up to the 'very high' quantities in the growth solution does not produce an increase on silver shell size, but, probably, only the formation of plate objects, which occasionally can deposit on chips.Measurements of EF with the 'very high' chips gave completely unreproducible results (thus not reported in figure 4(b)), due to the random presence of silver nanoplates of incontrollable size and distribution.
It is evident, comparing all these results, that the preparation obtained with concentration of silver/ascorbic acid fixed at 10 −4 M/5 × 10 −5 M and with 24 h of immersion yields the best performing GNS@Ag chips, in which, in affordable immersion time, the silver shell is maximized, still leaving the GNS branches tips outside, as evidenced by SEM.This peculiar situation, as expected, brings to a boost in the EFs, which move from the value of 4.0 × 10 6 observed with GNS chips, to the value of 4.3 × 10 7 of optimized GNS@Ag chips, gaining one order of magnitude in SERS response.As stated above, higher concentrations of silver/ascorbic acid seem to produce silver shells of similar extent, but with uncontrollable formation and deposition of silver nanoprisms.On the other side, the lowest concentrations do not allow an appreciable silver shell formation, with no increase in SERS signal and EF.
The increasing SERS performances of appropriately silver coated GNS can be at least qualitatively understood considering the distribution of E field on a single GNS, coated by a growing silver amount, as calculated in [55] using the boundary element method applied to a six-branched GNS surrounded by a silver shell of increasing thickness and immersed in water.The silver coating leads to a huge blue-shift of the main dipolar resonance towards the =600-1000 nm range and to the appearance of typical dipolar and quadrupolar resonance of the spherical silver shell around =400 nm.The complex structure is able to sustain higher order resonances likely in the high energy end of the spectrum contributing to the creation of the double peak feature observed in the silver rich samples.
Having assessed the optimized preparation of GNS@Ag chips with a boost in EF of a tenfold compared to pristine GNS chips, we moved to investigate homogeneity and reproducibility of the SERS response, again using MMC as a reporter, with the methodology described above.
The obtained results again point out an RSD around 10% on large-scale investigation, while slightly higher values on the small-one, demonstrating that the performed Ag coating did not perturb the pristine GNS layer distribution.The slightly higher RSD value derived for small-scale sampling is likely related to the different spatial resolution.Indeed, a 100 μm 2 spot size, used for the large-scale mapping, intrinsically averages the local distribution of hot spots that the Ag coating seems to have increased.On the contrary, a smaller spot size (4 μm 2 for small scale) is more sensitive to local changes.
The results of the map acquisition for the large-scale area are shown in figure 5(a), with an RSD value equal to 6.7%.To verify the reproducibility of the SERS response the MMC tests were performed also on different chips; as an example, the comparison between the mean spectra of two different GNS@Ag chips that for simplicity will be identified as A and B, is reported in figure 5(b) for the mode at 1170 cm −1 used to reproduce the color maps.It is possible to note a good accordance of the spectra within their RSD, thus confirming that the performed methodology of Ag covering leads to reproducible SERS substrates.Further map acquisitions on small-and large-scale are reported in the supporting data for  Comparison between the mean SERS spectra from large-scale maps collected on A and B GNS@Ag substrates.

Testing thiram detection
After this detailed characterization we moved to demonstrate the usability of the optimized SERS chips for pollutants detection using thiram as test molecule, since it is a quite diffused pesticide which represents a potential threat for human health [60,61].Thiram can be usually found on fruit peels and the simplest sampling consists in washing the fruit surface with ethanol [34].To simulate this situation, we simply prepared ethanolic solutions of thiram at sub-ppm concentrations and we immersed our as-prepared optimized GNS@Ag chips in these solutions for 1 h.Thiram molecular structure is based on a dithiocarbamate moiety, thus it is expected to give a strong S-Ag or S-Au covalent bond when interacting with the nano-objects.After immersion, samples were rinsed with EtOH to remove any unbound molecule, and after drying in air the SERS spectra were collected on the chip.
Figure 6 shows the SERS spectra obtained using EtOH thiram solutions with concentrations ranging between 20 and 250 ppb.The expected thiram spectra, dominated by modes at 550 and 1375 cm −1 , can be clearly observed down to 20 ppb, a value which could be useful for its detection with the cited procedure, as in the European Union and in the United States the limit range of thiram in fruits and vegetables is placed between 0.1 and 15 ppm.
For comparison with works already published, SERS substrates based on GNSs placed on cotton fabric allowed to detect thiram down to 100 μM [62], while properly optimized hybrid SERS substrates, based on femtosecond laser patterned silicon embedded with GNSs, allowed to detect a minimum concentration of 12 ppb [63].

Conclusions
A controlled and convenient synthetic method was developed and optimized in order to obtain SERS chips characterized by high EF, good homogeneity and acceptable reproducibility.The procedure involves a one step, straightforward growth of a silver shell on GNS grafted as a monolayer on glass, with no need of time/reagents consuming purification steps.The optimized preparation route leads to a situation in which tips of GNS are still exposed to the outside of the grown silver shell, a situation which was recognized as the best to gain high EF.These chips allow to gain one order of magnitude in the enhanced signal if compared to the pristine GNS ones.The practicality of these SERS chips was demonstrated since in all the described experiments, no tedious or time-consuming sample preparations were needed.SERS spectra of tested molecules (R6G, MMC and thiram) were obtained by the simple deposition of a drop of solution or by immersion of the chip in the target solution, indicating that they could also have great potentials in on-site detection of analytes.

Scheme 1 .
Scheme 1.The proposed strategy to a straightforward preparation of GNS@Ag SERS chips.
AgNO 3 and ascorbic acid concentrations in the four different growth solutions.Entry AgNO 3 (mol l −1 ) Ascorbic acid (mol l −1 )

Figure 1 .
Figure 1.(a) UV-vis-NIR spectra of a GNS coated glass chip; (b) Raman spectra of a 10 −3 M solution of R6G on a blank glass chip (black line) compared with the SERS spectra of a 10 −5 M R6G solution on a GNS chip (red line); (c) and (d) SEM images of a GNS chip at different magnifications.

Figure 2 .
Figure 2. (a) Average SERS spectrum of 10 −5 M MMC relative to mapping scan on a small-scale area of GNS glass chip.The surrounding shadow represents the standard deviation; (b) Corresponding color map showing the height of the mode at 1170 cm −1 as a function of the spot's position.

Figure 3 .
Figure 3. (a) UV-vis-NIR spectra of GNS@Ag chips obtained using a growth solution containing 1.0 × 10 −4 M AgNO 3 and 5.0 × 10 −5 M ascorbic acid, increasing growth times from zero (pristine GNS, black line) to 24 h (blue line); (b) SEM image of GNS@Ag chips after 24 h growing in the same growth solution; (c) SERS spectra of 10 −5 M R6G solutions using GNS@Ag chips obtained with increasing growth times from zero (pristine GNS, black line) to 24 h (blue line) and using a growth solution containing 1.0 × 10 −4 M AgNO 3 and 5.0 × 10 −5 M ascorbic acid; (d) Enhancement Factors calculated from R6G mode at 610 cm −1 of spectra in (c) plotted as a function of growth time (the red dotted line is a guide for the eye).

Figure 5 .
Figure 5. (a) Colour map of the height of the 1170 cm −1 MMC mode of spectra collected on a large-scale map for the A sample; (b) Comparison between the mean SERS spectra from large-scale maps collected on A and B GNS@Ag substrates.

Figure 6 .
Figure 6.Spectra of optimized GNS@Ag SERS chips after incubation in EtOH solutions of thiram with concentration of (from top to bottom) 250 ppb, 100 ppb, 50 ppb, 20 ppb.Typical modes at 550 and 1375 cm −1 are evidenced.