3D nanoplasmonic structure for ultrahigh enhanced SERS with less variability, polarization independence, and multimodal sensing applied to picric acid detection

3D nanoplasmonic structure for ultrahigh enhanced SERS with less variability, polarization independence, and multimodal sensing applied to picric acid detection† Anand M. Shrivastav, ‡ ab Mohammad Abutoama ‡ acd and Ibrahim Abdulhalim * a Surface-enhanced Raman scattering (SERS) is recognized as a powerful analytical method. However, its efficacy is hindered by considerable signal variability stemming from factors like surface irregularities, temporal instability of the substrate, interference with substrate signal, polarization sensitivity and uneven molecular distribution. To address these challenges, a new strategy is employed to enhance the reproducibility of SERS signals. Initially, a periodic 3D metallic structure is utilized to achieve polarizationindependent ultrahigh enhancement. Additionally, signal averaging over multiple points and normalization are implemented. The integration of these techniques enables multimodal sensing (SERS, SEF, SPR) using a plasmonic chip, demonstrating ultrahigh enhancement through the interaction of extended and localized plasmons alongside nanoantenna-type resonances. The chip comprises a periodic silver 2D grating adorned with Au nanocubes, behaving as a 3D metasurface to amplify plasmonic local fields, thus facilitating SERS. Its uniformity and polarization independence together with signal averaging and normalization mitigate signal variability. Fabricated via electron beam lithography, the chip's performance is evaluated for surface-enhanced fluorescence (SEF) and SERS using Rhodamine 6G as the target molecule. Results exhibit two orders of magnitude enhancement factor for SEF and 2.5 ×10 7 for SERS. For chemical sensing, the chip is tested for picric acid detection across a concentration range from nanomolar to millimolar, demonstrating a detection limit of approximately 3 nM. 1. Introduction Raman Spectroscopy has demonstrated its potency as a formidable tool for molecular ngerprint analysis, with diverse applications across elds such as security, pharmaceuticals, biomedical research, and agriculture. Nevertheless, its efficacy is curtailed by weak inelastic scattering, resulting in the generation of only one Raman-scattered photon for every 10 7 incident photons. This limitation constrains its broader applicability, 1–3 which can be improved by enhancing the intensity of Raman signals in what is called surface enhanced Raman scattering (SERS). Traditionally, SERS is accomplished when the target material is situated in a highly localized manner over a rough plasmonic nanostructured surface, with the corresponding plasmonic resonance wavelength overlapping both the laser excitation wavelength and the vibrational Stokes regime. 4–6 Extensive efforts have been undertaken to design and create innovative SERS substrates, yielding high levels of enhancement factors ranging typically from 10 3 to 10 10 . Typically, the enhancement factor in a SERS substrate is attributed to two primary components: electromagnetic enhancement and chemical enhancement. The electromagnetic enhancement arises from the heightened localized optical eld resulting from the presence of plasmonic nanomaterial near the Raman-active molecule, leading to a remarkable improvement in the SERS signal. 7,8 However, this enhancement strongly relies on various parameters, including the material, shape, and structure of the plasmonic nanomaterial, as well as the surrounding environment. The second major component is chemical enhancement, involving a charge transfer mechanism between the plasmonic surface and the Raman molecule. This a Department of Electro-Optics and Photonics Engineering, School of Electrical and Computer Engineering and The Ilse-Katz Institute for Nano-scale Science and Technology, Ben Gurion University, Beer Sheva, 84105, Israel. E-mail: abdulhlm@ bgu.ac.il b Department of Physics and Nanotechnology, College of Engineering and Technology, SRM Institute of Science and Technology, Kattankulathur, Chennai, Tamil Nadu, 603203, India. E-mail: [email protected] c DTU Electro, Technical University of Denmark, Ørsteds Plads, Building 343, 2800 Kgs, Lyngby, Denmark. E-mail: moa[email protected] d NanoPhoton –Center for Nanonphotonics, Ørsteds Plads, Building 345A, 2800 Kgs, Lyngby, Denmark †Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d4na00387j ‡Equal contribution. Cite this: Nanoscale Adv.,2024,6, 5681 Received 9th May 2024 Accepted 16th September 2024 DOI: 10.1039/d4na00387j rsc.li/nanoscale-advances © 2024 The Author(s). Published by the Royal Society of Chemistry Nanoscale Adv.,2024,6,5681–5693 | 5681 Nanoscale Advances PAPER process induces a change in molecular polarizability, resulting in a signicantly enhanced Raman signal, oen reaching orders of magnitude higher. 9,10 Harnessing the substantial electromagnetic enhancement, coupled with molecular ngerprint technology and advanced nanofabrication techniques, SERSbased sensors represent a rapidly advancing research domain, emerging as a predominant method for biosensing applications, even down to the level of single molecules. 11,12 To date, a wide range of SERS substrates have been reported that achieve high local eld enhancements including Au nanoparticles, 13 Au nanorings, 14 Au nonorods, 15,16 Au nanodimers, 17 Ag nano sculptured thin lms, 18–20 Ag nanoclusters, 21 and many others. To further enhance the SERS intensity the coupling between localised and extended plasmon congurations was proposed and found very efficient. 4,22–24 The ESP (extended surface plasmons) causes some local eld enhancement factor F esp , it excites the LSP (localized surface plasmons) which has enhancement factor of F lsp , however the nal eld enhancement factor was found to be even larger than the simple multiplication of the two enhancement factors F esp F lsp . All these structures have demonstrated enhancement factors of up to the orders of 10 5 –10 10 . Nevertheless, although these structures offer substantial enhancement factors, they still pose limitations for industrial applications due to notable variability in the SERS signal across the substrate area. This variability stems from factors like surface irregularities, uneven molecular adsorption on the surface, uctuations in laser light, and polarization dependence, temporal instability of the substrate 25–27 and interference with substrate signal 28 etc. To overcome these challenges, we have developed a polarization-independent twodimensional subwavelength periodic metallic 3D structure composed of silver boxes decorated with gold nanocubes as a SERS substrate, ensuring both low variability and high local eld enhancement due to effective ESP–LSP interaction. We noticed the excitation of nanoantenna type resonances on top of the silver boxes when the ESP is excited as their lateral dimensions are larger than the extended plasmon wavelength. The primary rationale for choosing metallic gratings lies in their uniformity across a substantial surface area, contrasting with nanostructures that may be grown or deposited on a surface. This uniformity plays a crucial role in minimizing signal variability for SERS. Additionally, the planar nature of metallic gratings presents the added benet of requiring minimal optical components, such as prisms or optical bers, for plasmonic excitations. 4,23,24 It is important to note that the designed metallic gratings have a period smaller than the wavelength of the incident light, which prevents diffraction and ensures that the obtained spectra are solely due to the contribution of optical resonant excitations. 29 These types of gratings are referred to as subwavelength gratings, and fall under the category of plasmonic metasurfaces. In our previous work, we designed one-dimensional subwavelength metallic gratings and demonstrated their application in refractive index sensing, as well as in SERS and surface-enhanced uorescence (SEF) based applications using the same substrate. 30 Additionally, several subwavelength grating congurations can be used to excite various resonance phenomena. Examples include: (a) designing the grating over a thin metallic lm to achieve resonance excitation of ESPs and cavity modes, 31 (b) using a thin dielectric grating over a metallic thin lm to provide ESP excitation and guided mode resonance (GMR), 32–34 (c) employing metallic thin nano slits on a dielectric substrate for enhanced optical transmission (EOT), 35–37 and (d) utilizing thin dielectric gratings with or without a waveguide layer to exhibit GMR. 38–40 Furthermore, as mentioned 2D grating can give ultrahigh enhancement of SERS through the ESP–LSP coupling 23 basically achieved through decorating plasmonic nanostructures over the ESP supported platform, which can be grating or metallic thin lm. For SERS applications, 1D metallic gratings are particularly useful for SERS, however, they still face limitations due to variability issues, as mentioned in the previous paragraph, which stems from factors such as surface uniformity, and polarization dependence. In this research, we present a novel approach to mitigate variability arising from nonuniformity and light polarization by designing two-dimensional silver gratings decorated with gold nanocubes for ultrahigh SERS enhancement. These gratings are specically craed to overlap optical resonances corresponding to both TE (transverse electric) and TM (transverse magnetic) polarizations. The selection of gold nano-cubes as the structure that supports LSPs is motivated by their eight-cornered conguration, generating a higher concentration of enhanced eld hot-spots across the sensor surface. The nal proposed substrate is fabricated using the e-beam lithography method and characterized through reectance spectrum analysis. Subsequently, the grating chip undergoes characterization using surface-enhanced uorescence (SEF) and surfaceenhanced Raman spectroscopy (SERS) methods, employing Rhodamine 6G molecules adsorbed on the chip surface, thus demonstrating its use for multimodal sensing. Finally, the substrate is applied to explosive detection, with picric acid (PA) as the targeted sensing material. 2. Numerical simulations and chip design The two-dimensional metallic grating was meticulously designed using COMSOL Multiphysics, involving multiple iterations. The nalized design of the 3D plasmonic structure is illustrated in Fig. 1(a). The thickness (d) of the continuous metal lm below the 2D grating is set at 100 nm to avoid transmission into the silicon substrate and to operate in reection mode. The 2D grating has a period (L) of 746 nm, line width (w) of 671 nm, and grating height (h) of 40 nm. All these parameters underwent optimization through the numerical study conducted in COMSOL. The optimization was basically performed to have the main resonance at ∼785 nm (the 1st order of the extended surface plasmon mode) to match with the wavelength of the laser to be used for the Raman measurements. Furthermore, Au nano-cubes (50 nm ×50 nm ×50 nm) were introduced onto the grating's surface to facilitate the interactions of extended and localized plasmons, resulting in a heightened local electromagnetic eld at the hot spots. This strategic addition 5682 |Nanoscale Adv.,2024,6,5681–5693 © 2024 The Author(s). Published by the Royal Society of Chemistry Nanoscale Advances Paper enhances the overall performance of the chip. 22,23 Although the chip is designed to have the nanocubes centered on the Ag boxes, due to fabrication limitations, the obtained chip has offcentered gold nanostructures along with partial changes in the numbers, as shown in the SEM image below (Fig. 1(b)). The offdiagonal shiwas found not to affect the performance, which can be seen in Fig. S1†where the resonances are obtained at 791 nm for off-diagonal and 785 nm centered Au nanocubes respectively. The fabrication process of the chip will be discussed in the experimental section of the manuscript. Hence, to obtain realistic results, the calculations are repeated keeping all the real parameters found in the fabricated ones like for electric eld calculations, the size of Au boxes is chosen as (60 nm × 60 nm ×60 nm). Additionally, Fig. 1(c) represents the EDX spectroscopic data for the fabricated substrate which has been further discussed in the experimental section. The eld distribution pattern in x–zand x–yplanes was calculated using COMSOL Multiphysics, which is based on the nite element analysis (FEA) method. The incident light beam is modeled as a plane wave propagating along the z-axis coming from the air domain (see Fig. 2). Perfectly Matched Layer (PML) boundary condition is imposed for z-coordinate above and below the superstrate (air) and substrate (Si) domains. Field monitor is used at a xed zposition above the nanostructure to detect the beam reection as a function of incidence wavelength. Adaptive mesh is used over the different domains to achieve reliable calculations. Normal incident light is used in all the simulations. The dielectric functions of the materials used in the simulations were interpolated based on the data taken from the Sopra database (http://sspectra.com/). Fig. 3(a) and (b) depicts the electric eld amplitude for the proposed grating structure corresponding to TE and TM polarized light respectively without the Au nanocubes. From the gure, the almost similar electric eld amplitude is obtained for both the polarizations providing the opportunity to minimize the polarization dependence for SERS. In addition, Fig. 3(c) and (d) corresponds to the electric eld intensities for the structures having Au nanocubes on the top of grating. An improvement by Fig. 1 (a) The targeted design: grating substrate comprising two-dimensional Ag grating supported by Ag thin film over Si substrate along with additional Au nanocubes centred on top of the Ag boxes, (b) SEM image and (c) EDX data of the fabricated structure. Fig. 2 Structure and meshing of the structure used for the simulations. © 2024 The Author(s). Published by the Royal Society of Chemistry Nanoscale Adv.,2024,6,5681–5693 | 5683 Paper Nanoscale Advances factor ×6 is obtained as seen in Fig. 3 where maximum electric eld is compared with/without Au nanocubes over the Ag grating. This is because of the coupling of extended surface plasmons (ESPs, due to the grating) with localized surface plasmons (LSPs, due to plasmonic nanocubes) providing the higher enhancement in the local eld, compared to the bare grating structure. Here, the simulations done on the fabricated structure with the nanocubes offcentred. Note that when the ESP is excited two maxima of the local eld appear on the surfaces of the Ag boxes which have lateral dimension of 671 nm. When the ESP is excited by the periodic structure it propagates, hits the edges of the Ag box and partially reects back. As a result, a standing wave is formed, which is a nanoantenna type resonance causing enhanced localization of the eld at the hot spots. Note also that the offcentring of the nanocubes causes a difference between the TE and TM eld distribution (Fig. 3(c) and (d)) which can be understood due to symmetry breaking. It may also be noted that 2D designed grating plays an important role as it eliminates the effect of the polarization of input light on the orientation of the substrate. A similar study was reported by Xiao et al. in 2018 to obtain polarization independence. However, in their works the 2D sinusoidal grating was prepared by using two-beam interference fabrication process and they have successfully demonstrated the polarization independent nature of the grating structure. 41 In the similar manner, if one refers to Fig. 3(c) and (d), for the incident TE and TM polarization of the input light, a nearly similar electromagnetic behaviour is obtained, however in the case of 1D grating one can simply observe the effect of polarization as the extended plasmons can be excited only with TM polarization (along the grating vector). When the structure is polarization sensitive then any small changes of the structure manifest in large changes in the signal, for example due to polarization change on the surface. Also, when a polarizer is used then polarization instabilities might affect the variability of the signal. 3. Experimental analysis 3.1. Materials Rodhamine 6G and picric acid (PA) were procured from SigmaAldrich Pvt. Ltd. Ethanol (99.9% pure) was obtained from ROMICAL chemicals Ltd, Israel. The chip was fabricated at nanofabrication facility at BGU, with 99% purity of gold (Au) and silver (Ag). All the materials were used in the similar form as received from the vendor, without any further modications. 3.2. Chip fabrication and characterizations The proposed chip underwent fabrication at the BGU nanofabrication center utilizing electron beam lithography. Initially, a 100 nm thick silver lm was deposited on the Si wafer, followed by the thin lm deposition of photoresist. Subsequently, based on the dened structure (as illustrated in Fig. 1(a)), the photoresist was hardened only in regions where the presence of Ag was unnecessary for grating fabrication, covering the entire surface. Consequently, the photoresist was conned to areas of Ag in the grating surface. Next, a 40 nm thick Ag lm was deposited across the entire surface, followed by surface cleaning to eliminate the hardened photoresist. This selective process resulted in Ag deposition exclusively in the line width part of the grating, completing the 2D grating fabrication. The same procedure was then repeated for the Au nanocubes over the grating surface. As displayed in Fig. 1(b) the scanning electron microscope (SEM) image of the fabricated chip, showcasing a high degree of uniformity along the surface. The bright spot on the surface corresponds to Au nanocubes. Some variation in the fabricated sensing chip occurred compared to the proposed design, attributed to experimental limitations. Specically, the Fig. 3 The electric field distribution for the designed SERS substrate. (a and b) Bare Ag 2D grating on Ag film for TE and TM polarizations respectively at the resonance wavelength 786 nm, while (c and d) with the Au Nano cubes on top of the Ag boxes for TE and TM waves respectively at the resonance wavelength 791 nm. The field enhancement at the hot spots increased by factor ×6 after adding the nanocubes. 5684 |Nanoscale Adv.,2024,6,5681–5693 © 2024 The Author(s). Published by the Royal Society of Chemistry Nanoscale Advances Paper nanocubes were obtained off-centered and the decentring slightly varied across the surface. Although from the SEM image we could not conrm precisely that prepared Au nanocubes are in cubical shapes as the light get saturated at higher beam intensity and limited resolution of the instrument. For the elemental analysis of the fabricated chip, we have also performed the EDX spectroscopy providing the conrmation of Ag and Au on the surface as shown in Fig. 1(c). The elemental data corresponding to chip fabrication is given in Table 1. It may be noted carbon peak is obtained in EDX due the photoresist remnants used in the fabrication process. 3.3. Reectance measurements Aer the chip fabrication, the reectance measurements were taken using a broadband source and a spectrometer using the setup shown in Fig. 4(a). The white light was incident normally to the substrate and corresponding reected beam was transferred to a spectrometer using beam splitter. The polarization of input light can also be tuned using the rotatable polarizer aer the collimation of the input light. Fig. 4(b) represents the output re- ected spectra for both TE and TM polarized light, depicting that the ESP resonance is obtained at around 785 nm wavelength for both the polarizations. It should be noted that this wavelength is very important since it overlaps with the laser wavelength used for the Raman excitation, in agreement with the design (Fig. 4(c)). This conrms the polarization independent response of the designed chip. Although, there is a small dip obtained at 762 nm for TM polarized light but to obtain the polarization independent nature in SERS, one needs to focus on the resonance near the laser wavelength (785 nm), where for the two polarizations a similar resonance is obtained. The change in depth of the experimental resonance dips is due to the losses and as well as experimental/fabrication limitations such as the use of etching which causes some roughness, well known to affect strongly surface plasmons. The roughness can be seen in the SEM image. 3.4. Surface enhanced uorescence (SEF) characterisations While measuring the reectance spectra for the fabricated chip, shown in Fig. 5(a), it was found that the chip possesses one Table 1 EDX data for the fabricated SERS chip Material wt% Error % Ag (silver) 87.5 0.9 Si (silicon) 9.2 0.4 Carbon (C) 2.6 0.5 Gold (Au) 0.7 0.2 Fig. 4 (a) Experimental setup for the reflectance measurements, (b) experimentally recorded reflected spectrum in the neighbourhood of the Raman excitation wavelength of 785 nm for TE and TM polarized light of the bare fabricated chip, and (c) numerically calculated spectrum of reflection for the TE/TM polarized light for the same geometry in (b). © 2024 The Author(s). Published by the Royal Society of Chemistry Nanoscale Adv.,2024,6,5681–5693 | 5685 Paper Nanoscale Advances more resonance dip around 500–520 nm. Using the SPR excitation formulae from gratings possible to verify that it is indeed a 2nd order SPR resonance: lSP zReL jffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 3m3a 3mþ3a r(1) where 3 m =3 mr +i3 mi is the metal dielectric constant and 3 a is the dielectric constant of the analyte or ambient, Ldenotes the pitch period, l SP corresponds to the plasmonic wavelength while jstands for the order of the resonance. Eqn (1) is valid when the grating role is just to excite the SPR, however it is used as well for estimating the SPR location for thin metal lines as the thickness of the metal has negligible effect on the k-vector of the plasmon. The period obtained from the SEM images is around 770 nm which when used in eqn (1) gives a resonance at 785 nm for the 1st order resonance j=1, and 504 nm for the 2nd order j=2, thus conrming the simulations and the experimental results. The period is calculated by measuring the distance between rst edge of start of a grating line to rst edge of the next grating line as shown in Fig. 1(b). Further, it is worth mentioning that reections obtained through experiments at nearly 504 nm and 785 nm are matching with the theoretical simulations, as shown in Fig. 5(a) and (b) respectively. Since, one of the resonances obtained falls within the excitation uorescence window of Rhodamine 6G (R6G), it was used as a uorescence tag molecule by spin coating of 1 mg ml −1 R6G over the fabricated chip surface. Fluorescence measurements were taken using Olympus uorescence microscope with an Hg arc-lamp as excitation light. The green Hg line at 546 nm was used for excitation and the emission was detected using a red lter at 590 nm. The detection was done using a high sensitivity cooled CCD camera with a controlled exposure time. Fig. 5(b) Fig. 5 (a) Reflected spectrum of the fabricated chip with a broader wavelength range, showing the broad 2nd order SPR dip around 504 nm, (b) numerical results for the validation of the same showing also the 1st order SPR at 785 nm. (c) Chip surface image when excited by the 546 nm Hg line and captured at the emission mode (using 590 nm edge filter to cut the excitation light wavelength), (d) fluorescence spectra for R6G molecules over Ag grating chip surface and Ag thin film surface depicting eight times signal enhancement taken from area of about 20 mm diameter on the surface. 5686 |Nanoscale Adv.,2024,6,5681–5693 © 2024 The Author(s). Published by the Royal Society of Chemistry Nanoscale Advances Paper represents the theoretical reection results of the proposed chip claiming the similar nature of both TE and TM polarized light. Fig. 5(c) shows the excitation and emission images of the chip surface. To evaluate the enhanced SEF, the ber coupled spectrometer was added to one of the eyepiece channels of the uorescence microscope to obtain the uorescence spectrum. Two different kinds of chips were placed under the microscope including the fabricated 3D plasmonic chip along with the bare Ag coated chip with R6G. For more quantitative evaluation the corresponding uorescence counts were collected by a spectrometer connected to the microscope in a conjugate plane to the camera. 42 Fig. 5(d) shows around an 8 times enhanced excitation using the 3D chip compared to bare Ag coated chip and the main reason is the high localized eld near the 2D grating surface. As it is known that the at Ag surface also enhances the uorescence by about factor ×10, hence possible to conclude that the overall SEF enhancement of the chip is close to ×80. Considering the fact that the SPR is at 504 nm, and the excitation is at 546 nm, even higher SEF enhancement factor is expected when they coincide. It can be noted that the unsmoothness and irregularity in the chip surface are partially due to the etching during the fabrication process and partially due to the ununiform immobilization of R6G molecules. The immobilization is done through the drop casting method by dropping 1 mg ml −1 solution of R6G. Because of this, at some places the clusters of R6G are formed providing irregular surface. However, it can easily be removed by sosonicating in ethanol ensuring reproducibility of the substrate. Irregularities of R6G can affect the variability over the surface but the irregularities due to the fabrication process are random and are believed not to affect it much such the measurement if from an area of many periods. To further elucidate the origin of the different resonances and the observed signals enhancement, the extinction and absorption cross-section of the fabricated structure were calculated as shown in Fig. 6(a) providing a broad spectrum around 550–650 nm. It also justies the enhanced uorescence when the green laser around 532 nm is used as what observed in Fig. 5(c) and (d) respectively. In addition, the electric eld distribution of the structure for two different side slice crosssection and as well as top cross-section is represented in Fig. 6(b) and (c) respectively. 3.5. SERS measurements 3.5.1. Setup. Fig. 7 depicts pictorial illustration of the experimental setup of a ber optic Raman spectrometer. The Fig. 6 (a) Numerical calculation of the extinction and absorption cross-section of the structure having the same parameters as the fabricated structure with Au nanocubes at the corners of the Ag boxes, (b and c) show normalized electric field distributions from the edge and top crosssections respectively at 532 nm wavelength. © 2024 The Author(s). Published by the Royal Society of Chemistry Nanoscale Adv.,2024,6,5681–5693 | 5687 Paper Nanoscale Advances setup consists of a semiconductor laser operating at a wavelength of 785 nm with a power of 600 mW. The laser light is coupled to a Raman probe (as shown in the inset of the gure) and precisely directed to the desired location on the SERS-active R6G immobilized 3D grating substrate using a 3-D translation stage, ensuring a tightly focused spot. The scattered Stokes Raman lines are collected through the same aperture of the SERS probe and transmitted through a specically designed optical conguration integrated with an optical ber. The ber guides the collected light to the Raman spectrometer, which is connected to a computer for signal recording. The inset of Fig. 7 provides an expanded view of the SERS probe. The laser light is collimated and passes through a sharp bandpass lter to ensure a single wavelength. Subsequently, it is tightly focused to a 300 mm diameter spot onto the substrate using another convex lens. The same lens is used to capture the reected and scattered light, which is then directed through a dichroic lter to separate the incident and collected light. The collected light is further transmitted through a long-pass lter to eliminate the Rayleigh scattering, the laser reected beam, and the anti-Stokes Raman scattered lines. Finally, another lens is employed to focus the Raman scattered light onto the input end of the collecting optical ber, which is then interfaced with the spectrometer. 3.5.2. SERS characterizations. To evaluate the SERS performance of the proposed chip, 1 ml of 1 mg ml −1 R6G in ethanol solution was dropped on the chip surface and equally spread using the spin coating method. The chip was then placed under the Raman probe to measure the Raman signal at various positions. A set of 10 measurements were recorded with 10 seconds integration time at different positions of the surface, then the average of all the curves is plotted in Fig. 8(a). From the gure, certain peaks corresponding to the R6G vibrational frequencies are observed. The R6G dye exhibits vibrational bands ranging from 1800 to 600 cm −1 . The observed vibrational bands in the R6G spectra between 1313 and 1649 cm −1 are attributed to the stretching of aromatic C–C bonds. Additionally, there is a weak band at 1130 cm −1 resulting from the bending of C–H bonds (in-plane), a band at 775 cm −1 arising from the bending of C–H bonds (out-of-plane), and a band at 612 cm −1 due to the bending of C–C–C bonds (in-plane). 43 As mentioned, the variability in the Raman signal is a crucial issue limiting the detection limit of sensors based on SERS, which is calculated by taking the ratio between the standard deviation in SERS signal and the average while measuring the SERS spectra at ten different locations. This originates from several factors: laser stability, variability of plasmonic structure, variability of the analyte concentration on the surface. As can be seen from Fig. 8(c) there are two sources of noise in the laser intensity, the low frequency one which is nearly 15% peak-topeak, and the high frequency one which is around 5% peakto-peak. Upon averaging the high frequency one can be reduced drastically, however the low frequency one which not completely random cannot be cancelled out completely by averaging, however this its time scale is in minutes, its effect may be minimized by normalization. The SERS signal variability was assessed by recording 10 measurements at different locations on the sensor chip, and the ratio of the standard deviation to the average was plotted for each peak in Fig. 8(b). The gure clearly shows that the variability falls within the range of 0.06– 0.1 (or 6–10%). In contrast, in typical cases, the signal variation is usually around 15–30% or even higher. 44–46 The improvement in signal reproducibility is attributed to the presence of a homogeneous sensor surface achieved through the use of a 2D grating, which effectively eliminates changes in the enhancement factor and the polarization dependency of the signal, as well as to the normalization of the signal to the average. Another important factor is the variations in the concentration of the analyte molecules on the surface. This can be improved using more sophisticated techniques of Fig. 7 Pictorial representation of SERS setup. 5688 |Nanoscale Adv.,2024,6,5681–5693 © 2024 The Author(s). Published by the Royal Society of Chemistry Nanoscale Advances Paper depositing the analyte molecules on the surface such as using nano-injectors. Normalization to a constant Raman peak that does not change with the analyte concentration is another approach to minimize the effect of the variations in the analyte concentration. 3.5.2.1 Calculation of enhancement factor. The SERS enhancement factor is a crucial parameter for evaluating the performance of a sensor surface. It is determined by comparing the intensity of the SERS signal obtained from each adsorbed molecule on the surface with the Raman signal intensity of each molecule in the bulk medium, as expressed below: 47,48 E:F:¼ISERS=Nads IRaman=Nbulk Here, I Raman and I SERS represent the peak intensities of the Raman measurement under normal and SERS conditions, respectively. N bulk and N ads refer to the number of R6G molecules in the scattering volume for the normal Raman measurement and SERS measurement, respectively. In our case, the concentration of R6G molecules (conc. 1 mg ml −1 )ina1ml volume is estimated to be approximately 1.255 ×10 16 , considering the molecular weight of R6G as 479.02. Studies have shown that the surface area covered by a single R6G molecule is approximately 1 nm 2 (molecular area), 49 while the substrate has an area of 1 mm 2 . This suggests that the maximum number of molecules that can be accommodated within a single layer is approximately 10 12 . It is important to note that SERS can be observed for molecules within a few tens of nm distance from the surface according to the eld distribution observed in Fig. 3, which limits the total number of molecules taking part of SERS to be within a few to ten monolayers of R6G. Considering laser parameters such as a spot diameter of 0.3 mm and a depth of eld of 1 mm, the number of molecules eligible for SERS (N SERS ) would be 2.826 ×10 11 (considering 10 monolayers), while the number of molecules for Raman scattering (N Raman ) would be approximately 3.55 ×10 17 , since 1 mg of R6G is required to collect the Raman spectra of bulk sample. To determine the enhancement factor, Raman spectra for bulk Fig. 8 Recorded SERS spectra for R6G sample over Au nanocube decorating 2D Ag grating substrate. (b) The signal variability (standard deviation/average) of 10 measurements, (c) the laser intensity fluctuation playing an important role for higher variability, and (d) Raman spectra of R6G sample over bare Si substrate under the same condition as in (a). © 2024 The Author(s). Published by the Royal Society of Chemistry Nanoscale Adv.,2024,6,5681–5693 | 5689 Paper Nanoscale Advances