Galvanically Replaced Au/Ag Nanostructures as a SERS‐Active Substrate with Progressive Interior Hotspots

Interior hotspots in surface‐enhanced Raman spectroscopy (SERS) platforms have attracted intensive attention because they enable facile methodologies and exhibit excellent sensing behavior. Molecules surrounded by plasmonic materials exhibit dual functions as field‐confined regions and analytes, and their domain consistency eventually triggers the amplification of SERS signals. In this study, to conveniently realize interior hotspots, hollow regions such as voids and interstitials are strategically introduced via a galvanic reaction (GR) as a result of the difference in reduction potential between Au and Ag. The imbalanced stoichiometric ratio and diffusion fluxes induce the Kirkendall effect in conjunction with the GR between Au and Ag. SERS platforms with narrow and densified interior hotspots are optimized by controlling the reaction time. The activation of interior hotspots is confirmed using the finite‐difference time‐domain method, which indicates a theoretical enhancement factor of 1.07 × 107 based on a fourth‐power approximation. The spontaneous GR reaction enables sensing operation with high reproducibility (relative standard deviation of <10%). The proposed bimetallic platforms are used to trace methylene blue and rhodamine 6G dyes until their concentrations reach 50 nm. Therefore, the GR methodology for interior hotspot engineering demonstrates the potential for enabling the fabrication of SERS platforms with practical applications.


Introduction
Surface-enhanced Raman spectroscopy (SERS) has been used extensively as a powerful tool to validate chemical and biomedical DOI: 10.1002/adsr.202300090analytes because the unique spectral features correspond to the vibrational energy states of individual molecules. [1,2]ompared with other optical sensing methods (e.g., fluorescence, fiber Bragg grating, and Förster resonance energy transfer), SERS offers the particular benefits of contactless and label-free diagnosis with negligible interferences (e.g., photobleaching and aqueous solvents).The amplification of spectral intensity arises from both electromagnetic and chemical enhancement. [3,4]he SERS enhancement factor is dominantly determined by the electric field confined in hotspots between plasmonic nanostructures, referred to as the localized surface plasmon resonance (LSPR) effect. [5,6]Because the field strength is inversely and exponentially proportional to the hotspot scale, elaborated SERS substrates have demonstrated singlemolecule sensitivity. [7,8]Accordingly, hotspot engineering techniques strongly influence SERS activities.[11][12] However, the application of SERS in practical fields is still limited.To achieve coherent electron oscillations, highly periodic nanostructures prepared by complicated semiconductor fabrication processes (e.g., lithography and vacuums) are required.Nanoimprinting techniques have been adopted to prepare sensitive SERS platforms owing to the convenient formation of nanostructured replicas on various secondary substrates, but their limited hotspot regions remain a challenge. [13,14]In addition, the molecules dissolved in solvents diffuse in random directions because of the thermodynamic energy (i.e., Brownian motion), which results in the molecules adsorbing onto various substrate sites, including sites other than hotspots.Consequently, the sensing performance for target analytes is degraded, especially at lower analyte concentrations.
To improve the efficiency of capturing molecules in the intended areas, researchers have investigated SERS platforms based on conjugation systems. [15,16]For example, aptamers are universally used in research and clinical fields because of their robust and stable characteristics.Aptamers have a high affinity to covalently bind with the specific sequences in the target specimen.Accordingly, the analytes become anchored (i.e., immobilized) near hotspots and their locations can be adjusted through modification of the aptamers' chain length.Dyes (e.g., cyanine 5 and streptavidin-biotin) can be further introduced into this system to enhance sensitivity. [17,18]However, this approach requires multiple and complex pre-treatment steps with prolonged reaction times (e.g., incubation, blocking, modification, and activation).Precisely controlling the site-selective conjugation and aptamer orientation, which strongly influences hybridization efficiency as well as device performance, is also difficult.Therefore, a novel methodological strategy is required to enable the development of facile, rapid, sensitive, and reproducible SERS substrates as alternatives to conventional platforms.
[21] Au shell-1 nm gap-Au core nanoparticles filled with conjugated probes were recently reported to demonstrate femtomolar sensitivity. [19]This result implies that the molecules surrounded by plasmonic materials could be activated as both a dielectric medium (i.e., interior hotspots with confined electric fields (E-fields)) and an analyte.The match of their domains provided the potential to prevail over the molecular loss (i.e., out-of-hotspot adsorption) that occurs with traditional SERS substrates.Porous Au-Ag nanospheres prepared by a dealloying treatment were used to trace the fingerprint spectral features of 4-aminothiophenol probes at a sub-nanomole level. [20]The hotspot density was sufficiently increased as a result of abundant nanoscale vacancies, improving the sensitivity and reproducibility of the SERS substrate.These research trends suggest a paradigm shift in SERS analysis toward practical applications.
Interior hotspots have been realized through various electrochemical methods, including catalytic reduction, [19] dealloying with acid, [20] electric reduction by an applied potential, [22,23] and galvanic reaction (GR). [21,24,25]In particular, GR provides the benefits of a facile, rapid, and easily tunable process.28] When the reaction is predominantly determined by the differences in the reduction potentials of two or more materials, the replacement process is observed.In this case, internal hollow regions, which potentially function as interior hotspots, can be formed according to the stoichiometric ratio.However, other factors, such as reducing agents, chemical residues, the solution environment (e.g., acid or base), and temperature, also affect the reaction associated with the deposition process.[31] Among these alloys, Au/Ag bimetallic nanostructures have attracted intensive attention as SERS platforms because i) both Au and Ag have strong intrinsic plasmonic properties, ii) a replacement reaction spontaneously occurs, iii) the inertness Au compensates for the poor stability of Ag, and iv) an unequal stoichiometric ratio of Ag to Au (1:3) results in voids and interstitials.Therefore, a comprehensive investigation of GRinduced interior hotspots is necessary.
In the present study, Au/Ag bimetallic nanostructures were prepared on glass substrates.A 3 nm-thick Au layer was first deposited onto the glass substrate to prevent light loss (i.e., transmission).The sacrificial Ag layer was synthesized using a mirror reaction and then galvanically replaced by Au.The replacement process occurred via a reaction without chemical additives or the input of thermal energy.The influence of the Au/Ag bimetal-lic nanostructures' morphology on the SERS signals of analytes was investigated.The activation of interior hotspots was also experimentally and numerically verified.The Raman active dyes (i.e., methylene blue (MB) and rhodamine 6G (R6G)) were used to evaluate the reproducibility and quantitative sensitivity of the Au/Ag/glass SERS platforms.

Strategy of the GR-Synthesized Au/Ag/glass SERS Platforms
The enhancement of Raman signals is determined by several factors, including the number of hotspots, the confined field strength, and the molecular distributions.Thus, for the sensitive detection of analytes at low concentration ranges, SERS platforms should have highly populated nanoscale hotspots per unit volume.In the present study, such plasmonic nanostructures were realized with interior hotspots as a result of an atomic exchange between two different metals (Scheme 1).Before bimetals were formed, a thin (≈3 nm) Au layer was deposited to prevent optical loss (i.e., transmission) via the transparent glass substrate.The Ag sacrificial layers were prepared by a mirror reaction.The Ag nuclei rapidly grew into grains as the reaction proceeded.The GR process was then applied to form the interior hotspots in the Au/Ag bimetallic nanostructure.When a HAuCl 4 solution is involved, the Ag atoms are oxidized and the AuCl 4 − ions are reduced because of the difference in their standard hydrogen reduction potentials (i.e., AuCl 4 − /Au: 1.5 V; Ag + /Ag: 0.8 V). [32,33] According to the stoichiometric ratio (Ag:Au = 3:1), the undercoordinated Ag atoms are dissociated, resulting in internal hollow regions.Solid Au is subsequently deposited onto the surfaces.With increasing GR time, the scales of both the hollow regions and the entire domains increase.The target analytes were distributed on the proposed platforms using a drop-and-dry method, and the SERS signals were measured under laser exposure at 785 nm.

Optimization of the Au/Ag/Glass Platforms
During the formation of Au/Ag bimetal alloys via the GR process, Ag is spontaneously replaced by Au because of the difference in their reduction potentials.On the basis of the stoichiometric relationship between Ag and Au, the outward flux of Ag is greater than the inward flux of Au.This behavior is known as the Kirkendall effect and, in conjunction with the GR, leads to the formation of hollow regions (e.g., voids and interstitials) in the Au/Ag bimetallic nanostructures.During the dynamic redox process, some oxidized Ag + ions can be reduced to Ag(0) at the outer surfaces, followed by the GR, resulting in an increase in the number of structural domains and in surface roughness.Because the performance of the proposed platforms is determined by the coupling effect of interior and environmental hotspots, as demonstrated in our previous studies, [17,34] such a phenomenon strongly influences the sensing efficiency.Therefore, we investigated the morphological characteristics of the Au/Ag/glass substrates as a function of the GR time (Figure 1a-e).The Ag layers with the grains were formed as a result of the rapid mirror reaction.When the AuCl 4 − ions were injected, the surface roughness increased with increasing GR time.This behavior is attributed to the thin Ag areas (especially at grain boundaries) being nearly dissociated at the early stages while the thick Ag areas were transformed into the Au/Ag bimetallic nanostructures.
To determine the optimal SERS performance, we measured the 1 μm MB signals using the Au/Ag/glass platforms prepared using various GR times (Figure 1f).For all the platforms, the spectral features of MB were clearly observed.Compared with the spectrum of MB deposited onto the Au/Ag/glass platform  (MB-Au/Ag/glass), that of MB deposited onto the Ag/glass platform (MB-Ag/glass) was negligible because of the weak field enhancement provided by the dominant particle modes of Ag grains.Notably, prominent MB spectra were attained with the Au/Ag/glass platforms, indicating an excellent signal-to-noise ratio.
To investigate the plasmonic resonant modes, the reflectance, transmittance, and absorption spectra at each step (i.e., glass, 3 nm Au/glass, Ag/glass, and Au/Ag/glass) were investigated by using the UV-vis spectrometer (Figure S1, Supporting Information).It is known that the isolated or a few plasmonic nanostructures exhibit sharp resonant peaks.When these structures form arrays with imperfect symmetry, the absorption spectra become broadened due to the multiple resonant modes. [35,36]From the measured data, the glass exhibited high transparency (≈91%) and low reflectance (≈7%) according to the Fresnel equation at a normal incidence, implying negligible absorption due to the absence of plasmonic nanostructures.After the 3 nm Au deposition on the glass, the small Au islands (Figure S2, Supporting Information) played a role to decrease transmittance and to produce the resonant features.The Ag/glass reflected most of the incident light because the large Ag grains were insufficient for collective electron oscillations.After the GR, the reflectance was decreased while the transmission of light was not allowed, indicating the multiple resonant modes (i.e., broad absorption spectrum) originating from the interior hotspots in the arrays.This result could be direct evidence of the activation of interior hotspots.
The SERS efficiency of the Au/Ag/glass platforms is related to the following factors.First, the interior hotspots and their scales increase with increasing GR time.Second, although the environmental hotspots were formed on the textured surfaces, the sufficiently large and irregular Au/Ag bimetals (i.e., the bimetals obtained when the GR time was ≥60 s) contributed to the inherent plasmonic modes as well as the performance degradation.Third, the partial replacement of Ag with Au compensated for the poor stability of Ag. [37] In the spectrum corresponding to the Au/Ag/glass chip prepared using GR for 30 s, the MB peak at 1617 cm −1 is 38.9 times more intense than the corresponding peak in the spectrum of MB adsorbed onto an Ag/glass chip.Herein, the analytic peak at 1617 cm −1 is assigned as a ring stretching vibration of C−C. [38,39]Accordingly, the optimum GR time was set to 30 s.

Interior Hotspot Analysis of the Au/Ag/Glass Platforms
To directly observe the internal hollow regions, we analyzed the Ag/glass and Au/Ag/glass platforms by scanning transmission electron microscopy (STEM).Before the GR process, the layers (average thickness of 176.1 nm) with Ag were fully retained (Figure 2a).The elemental mapping analysis confirmed the superior Ag content.The particle mode of the Ag nanostructures became dominant; however, the platform's ability to enhance the molecular absorption cross-section was poor because i) the smooth surfaces were inappropriate to confine the E-fields and ii) sufficiently red-shifted plasmonic bands that originated from the large-scale Ag grains were not properly matched to the incident laser wavelength of 785 nm.After the GR process, the entire structure was expanded (average height of 280.4 nm) and internal hollow regions were observed (Figure 2b).At the bottom, large voids (height of 170.7 nm) created within 30 s were observed, indicating the rapid dissociation of Ag by Au.Moreover, bimetals with low Au contents were also revealed by the elemental mapping analysis.These results provide direct evidence of the Kirkendall effect during the GR process.The outwardly diffusing Ag + ions were also reduced by electron transfer, leading to the generation of additional GR sites at the top regions.Therefore, the desired SERS behaviors could be attained with the Au/Ag/glass platforms because of i) the functionalization of voids and tials for the interior hotspots and molecular diffusion pathways, ii) the E-field confinement in the environmental hotspots (i.e., roughened surfaces), and iii) their resonance coupling effect.
We characterized the atomic distributions to investigate the Au/Ag bimetal alloys.Notably, Au and Ag have similar crystalline characteristics, including the same space group (i.e., cubic Fm̅ 3m) and similar lattice constants (4.079 Å for Au; 4.086 Å for Ag). [33,40]These similarities led to excellent miscibility and epitaxial growth of the Au/Ag alloys during the one-pot GR synthesis, as shown in the mapping images in Figure 2.

Numerical Calculation of E-Field Distributions
To theoretically verify the activation of interior hotspots formed by the GR, the E-field distributions of the Ag/glass and Au/Ag/glass platforms prepared at a GR time of 30 s were calculated using the finite-difference time-domain (FDTD) method (Figure 3).The modeling structures were designed on the basis of the geometric parameters extracted from the STEM images.In the simulation, the following assumptions were made: i) the Ag/glass and Au/Ag/glass were highly symmetrically distributed, ii) the Ag grains at the top of Ag/glass were ellipsoidal, and iii) the voids in the Au/Ag/glass were ellipsoidal.The interior hotspots were constructed with empty spaces in the mixtures of Au and Ag nanoparticles.The ratio of Au nanoparticles to Ag nanoparticles was set to 1:3 and 1:9 for the top and bottom Au/Ag layers, respectively, according to the elemental mapping analyses.The Ag/glass demonstrated intense fields in the hotspots between adjacent Ag grains according to the periodic boundary condition, which corresponds to conventional hotspots.Light penetration was not allowed because of the Ag electric conductor.In the case of the Au/Ag/glass, the presence of E-fields was observed not only in the interstitials (i.e., interior hotspots) but also in the voids.Therefore, the 3 nm-thick Au layer on the glass substrate reflected the penetrated light, leading to light trapping as well as an increase in the optical path length.E-field distributions in the Ag/glass and Au/Ag/glass platforms confirmed that the GR process enabled the formation of interior hotspots with enhanced field confinement and density within 30 s.The Au/Ag/glass platforms with the narrow and densified hotspots showed potential for sensitive detection.Cracks in the platforms were also able to be functionalized as molecular diffusion paths, possibly leading to a high rate of plasmonic sites occupied with molecules.The maximum E-field intensity of the Au/Ag/glass platform was evaluated to be 57.2.Therefore, the theoretical enhancement factor was calculated to be 1.07 × 10 7 on the basis of a fourth-power approximation.

Reproducibility of the Au/Ag/Glass Platforms
For practical sensing applications, reliable device operations based on reproducible and uniform signals from multiple measurement points are necessary.Therefore, prior to investigating the sensitivity, we estimated the reproducibility of the MB-Au/Ag/glass platforms using a Raman mapping system (Figure 4).A 5 μL droplet of MB solution was dropped onto a platform and then dried under the air environment.The mapping was carried out for the SERS-active areas with a square of 50 × 50 μm 2 .The interval between the spots was determined to be 5 μm to prevent the side effect originating from the regional overlap by a laser spot diameter of 2.4 μm.The spectra were recorded for 10 μm MB at an optical power of 1.5 mW and an acquisition time of 1 s.The fingerprint MB peaks were clearly observed in the spectra acquired at all the spots, without substantial degradation (Figure 4a).The signals were uniform irrespective of the peak positions.For easy understanding, the spectral visualization is presented as a 2D plot (Figure 4b).To calculate the relative standard deviation (RSD) values, the intensity profiles at the analytic peaks of 447, 1398, and 1617 cm −1 were acquired (Figure 4c-e).Herein, the MB peaks at 447 and 1398 cm −1 correspond to the inplane vibration of C−S−C skeletal deformation and the symmetrical stretching of C−N, respectively. [38,39]The RSD values were ≈9.5%, 8.6%, and 9.0% for the peaks at 447, 1398, and 1617 cm −1 , respectively.The reliable signal acquisition (i.e., RSD < 10%) was attributed to i) the formation of interior hotspots with high density and ii) the spontaneous GR synthesis at room temperature.To investigate the large-scale reproducibility, the SERS mapping was performed over the areas of 1000 × 1000 μm 2 with an interval of 50 μm (Figure S3, Supporting Information), and subsequently, the RSD values of the square areas from the center were evaluated (Table S1, Supporting Information).As the estimated areas were expanded from the center, the RSD values of the 5 μL droplet were more significantly changed as compared with those of the 10 μL droplet due to the different surface coverage.This was attributed to the inhomogeneous molecular distributions, which is well known as the coffee ring effect.In the case of a water droplet, contact lines are pinned at edges and suspended analytes are driven by a capillary force, resulting in the of molecules in outer areas. [41,42]Consequently, the SERS intensity is varied according to the coordinates.Despite the significant coffee ring effect, it is noted that the RSD < 9% was observed from the central areas in both cases, which were reasonably matched with the mapping data in Figure 4.This result could be evidence of i) the high reproducibility of the proposed Au/Ag/glass platforms and ii) the significant influence of molecular distributions on SERS signals.
To study the effect of Ag oxidation on signal variations, the Ag/glass and Au/Ag/glass platforms were used to measure 1 μm MB signals every day for 2 weeks (Figure S4, Supporting Information).Both platforms were kept in a desiccator after the synthesis.In the case of Ag/glass substrates, the MB signals were remarkedly degraded within a few days and then eventually almost disappeared because the oxidized Ag sublayers (e.g., Ag 2 O) inhibited sensitive molecular detection based on electromagnetic enhancement. [43,44]To prevent such an issue, chemically inert Au is used for Au/Ag metallic nanostructures. [45]In the case of Au/Ag/glass platforms, the MB signals were not significantly degraded (i.e., intensity decrement of 30.3%) over the tested period.Therefore, the Au/Ag bimetallic nanostructures prepared by the GR technique were stably and reliably operational against the Ag oxidation issue.

Quantitative Sensitivity of the Au/Ag/Glass Platforms
For the trace analysis of target molecules, it is important to clarify the correlation between the concentration/mass of the target molecules and the SERS intensity.For example, the number of biomedical analytes varies depending on the infection period, such as during the incubation, acute, and recovery phases.Moreover, the use of hazardous materials (e.g., pesticides and plasticizers) is restricted according to their mass-based critical levels.Therefore, an accurate quantitative investigation is required for field applications, including epidemiological surveys and industrial safety.
To explore the sensitivity of the developed platforms, SERS measurements were conducted under different molecular concentration ranges from 50 nm to 50 μm.In the case of MB, its unique signals were obtained at all the investigated concentrations (Figure 5a).The intensity gradually decreased with decreasing concentration.The spectrum of 50 nm MB-Au/Ag/glass was distinguishable from that of Au/Ag/glass (i.e., a bare platform with no dye), indicating an excellent signal-to-noise ratio (Figure S5, Supporting Information).The intensity profiles (n = 3) at the analytic peaks were plotted to explore the quantitative relationship (Figure 5b).In the logarithmic regression, high linearity was obtained, with a correlation coefficient (R 2 ) of ≥0.93.The LOD indicates the detectable range for the target analytes by using the sensing platforms.In general, the measured data includes the positive, control, false positive, and false negative.In SERS research, the error probability (i.e., false negative and false positive) becomes negligible because fingerprint spectra are obtained when molecules are present in hotspots.For accurate evaluation of LOD, therefore, the signal-to-noise ratio approach is used with the three standard deviations equation [46][47][48] of where S is the standard deviation of blank measurements (n = 5) at the analytic peak and M is the slope of the calibration curve within the linear range.Accordingly, the average LOD value of MB-Au/Ag/glass platforms was estimated to be 12.0 nm.To evaluate the feasibility of multiplex assays, the Raman active dye R6G was also used (Figure 5c).The characteristic R6G peaks were prominently detected in the desired concentration ranges.The analytic peaks at 610, 1358, and 1506 cm -1 were selected for further investigation.The peak at 610 cm -1 was assigned to the in-plane vibration of the C−C−C ring, whereas those at 1358 and 1506 cm -1 were assigned to the symmetric modes of in-plane C−C stretching vibrations. [49]The LOD of the R6G−Au/Ag/glass platforms was determined to be 17.9 nm (Figure 5d).These results demonstrate that the methodological strategy with GR enables the recognition of small molecules in a simple, rapid, sensitive, and reliable manner.

Conclusion
We presented a methodological strategy based on the GR process to prepare Au/Ag/glass SERS platforms with interior hotspots.The sacrificial Ag and subsequent Au/Ag bimetal nanostructures were prepared via facile and rapid chemical synthesis.During the GR, internal voids and interstitials were formed as a result of the imbalanced atomic exchanges between Au and Ag (i.e., the Kirkendall effect).The GR reactions were spontaneously performed without chemical additives at room temperature, leading to reproducible (i.e., RSD < 10%) and reliable sensing operations.The hollow regions functioned as molecular channels and interior hotspot regions.MB and R6G probe dyes were successfully traced using the Au/Ag/glass SERS platforms until their concentration was as low as 50 nM.The proposed platforms also demonstrated feasibility for use in epidemiological and multiplex investigations.Therefore, the galvanically engineered interior hotspots proposed in this work show strong potential for the facile, rapid, sensitive, and reproducible recognition of small molecules in various fields.Cl) were purchased from Sigma-Aldrich (St. Louis, MO, USA).A potassium hydroxide (KOH) solution (1 n) was obtained from Samchun (Gangnam, Korea).Ammonia solution (NH 3 , 28%) was purchased from Junsei (Tokyo, Japan).Glass substrates were acquired from AMGtech (Uiwang, Korea).Au pellets (99.99%) for vacuum deposition were obtained from Taewon Scientific (Seoul, Korea).

Experimental Section
Synthesis of Au/Ag/Glass SERS Platforms: Glass substrates were prepared by removal of their protective film and then loaded into a thermal evaporation system (LAT, Osan, Korea).The chamber was evacuated to a base pressure of 9.6 × 10 −6 Torr was achieved.A 3 nm-thick Au layer was deposited onto the substrate at a vapor flux rate of 2.0 Å s −1 and an RF power of 100 W, followed by an annealing process at 200 °C.Ag layers were then prepared using a mirror reaction.Tollens' reagent (i.e., a mixture of 0.5 m AgNO 3 (5 mL), NH 3 (680 μL), and 0.8 m KOH (660 μL)) and a 0.5 m glucose solution (660 μL) were sequentially applied to the substrates.The reaction solution was mechanically stirred at 100 rpm for 3 min.The obtained Ag/glass platform was carefully rinsed with deionized water three times and then dried using blown air.To create plasmonic hotspots, a GR process was carried out by immersing the substrates into a 2 mm HAuCl 4 solution.After the platforms were rinsed with deionized water three times and air-drying, the final Au/Ag/glass platforms were achieved.
SERS Measurements: To investigate the performance of the proposed platforms, Raman probe molecules (i.e., MB and R6G) dissolved in deionized water at concentrations ranging from 10 nm to 100 μm were used.A 5 μL droplet was dropped onto a sensing chip and then dried under the atmospheric environment.An Ocean Optics portable probe spectrometer (UQEPRO-Raman) was used to collect SERS spectra with an acquisition time of 30 s.A 785 nm laser with an optical power of 49 mW was incident onto the substrates in the normal direction.Raman mapping was carried out using a Nanoscope spectrometer system (NS-200) equipped with a three-axis stage.
Theoretical Analysis: The E-field distributions of the Ag/glass and Au/Ag/glass platforms during GR for 30 s were calculated on the basis of the FDTD method using a commercial program (Ansys Lumerical 2021 R1.2).The 3 nm-thick Au-deposited glass was used as the supporting substrate.The Ag grains are depicted with a combination of a bottom Ag layer (thickness: 115.4 nm) and an ellipsoid (minor radius: 73.8 nm) at the top.In the case of the optimal Au/Ag/glass platform, Au/Ag bimetallic layers with interior hotspots were realized using mixtures of Au and Ag nanoparticles.The void (i.e., empty space) was placed in the bottom Au/Ag layer (thickness: 173.8 nm).The ellipsoidal Au/Ag features (major radius: 83.4 nm and minor radius: 60.4 nm) were placed at the top.The top Au/Ag layer was composed of mixtures of Au and Ag nanoparticles with a ratio of 1:3, whereas the bottom layer was composed of 1:9 Au/Ag mixtures according to the elemental mapping analysis conducted during the STEM observations.Under the assumption of symmetric distributions of plasmonic nanostructures, periodic boundary conditions were used for the xand y-axes.A perfectly matched layer was applied at the top and bottom layers.An x-polarized plane wave (785 nm) was incident to the constructed nanostructures in a normal direction.A fixed mesh scale of 0.5 nm was applied.For the optical parameters, at 785 nm, the dielectric constants of Au and Ag were used:  Au = −21.64+ i0.74 and  Ag = −30.87+ i2.99, respectively.The refractive index of glass was set to be 1.50.
Characterization: The morphological properties of the proposed nanostructures were observed by field-emission scanning electron microscopy (FE-SEM; JSM-6700F, Jeol, Tokyo, Japan) and STEM (JEM-ARM200F, Jeol, Tokyo, Japan).The optical characteristics of the substrates at each step were measured by a Cary Series UV-Vis-NIR spectrophotometer (Cary 5000, Agilent Technologies).

Scheme 1 .
Scheme 1. Schematic diagram of the procedure for preparing Au/Ag/glass platforms via a mirror reaction and GR reaction.

Figure 1 .
Figure 1.Morphological and optical properties.SEM images of the Au/Ag/glass platforms at GR times of a) 0, b) 10, c) 30, d) 60, and e) 120 s. f) MB spectra obtained from the Au/Ag/glass platforms prepared using different GR times.

Figure 2 .
Figure 2. Analysis of interior hotspots.STEM and elemental mapping images of a) Ag/glass and b) Au/Ag/glass platforms.

Figure 3 .
Figure 3. FDTD simulation.E-field profiles of the a) Ag/glass and b) Au/Ag/glass platforms.