Dense Arrays of Nanohelices: Raman Scattering from Achiral Molecules Reveals the Near‐Field Enhancements at Chiral Metasurfaces

Against the background of the current healthcare and climate emergencies, surface enhanced Raman scattering (SERS) is becoming a highly topical technique for identifying and fingerprinting molecules, e.g., within viruses, bacteria, drugs, and atmospheric aerosols. Crucial for SERS is the need for substrates with strong and reproducible enhancements of the Raman signal over large areas and with a low fabrication cost. Here, dense arrays of plasmonic nanohelices (≈100 nm in length), which are of interest for many advanced nanophotonics applications, are investigated, and they are shown to present excellent SERS properties. As an illustration, two new ways to probe near‐field enhancement generated with circular polarization at chiral metasurfaces are presented, first using the Raman spectra of achiral molecules (crystal violet) and second using a single, element‐specific, achiral molecular vibrational mode (i.e., a single Raman peak). The nanohelices can be fabricated over large areas at a low cost and they provide strong, robust and uniform Raman enhancement. It is anticipated that these advanced materials will find broad applications in surface enhanced Raman spectroscopies and material science.


Introduction
Currently, surface enhanced Raman scattering (SERS) spectroscopy is of growing interest as an ultrasensitive chemical analysis technique, with high specificity, in minute quantities of material. [1,2] SERS relies on nanostructured substrates to enhance the light-matter interaction. [3] Such substrates have been developed to address many 21st century challenges, including the detection of pesticides, [4] pharmaceuticals, [5] antibiotics, [6] heavy metals, [7] pathogens [8] and bacteria. [9] through the biological chain posing a threat to human health, [14] animal welfare, and wildlife. [15] Generally, complex mixtures of various analytes with indistinguishable background signals render the detection of trace targets difficult. [16] The unique vibrational fingerprints and high signal enhancement enable SERS to selectively detect these trace pollutants from complex samples. For instance, Zhang et al. [17][18][19] identified the chemical components of a single aerosol via different SERS substrates, enabling simultaneous chemical imaging. All these advances drive a need for high-performance SERS materials.
The typical SERS substrates are metal nanomaterials. [20] Such materials have emerged largely due to the advances in nanoscale fabrication techniques and nanotechnology. [21] Initially, they had basic shapes [22,23] but, as technology advanced, more intricate designs became available to direct and confine light at the nanoscale, [24][25][26] often taking inspiration from the shapes of electromagnetic antennas. [27] One of the most successful antennas is the helical antenna, invented by Kraus. [28] Upon decreasing the dimensions of helical antennas, numerous applications became apparent. For example, negative refractive index materials have been demonstrated for metallic helices with lengths in the tens of millimeters. [29,30] Helices with a total length in the 1-5 µm range exhibit strong (40%) optical activity in the spectral range 0.8 µm to 1.5 µm. [31] In addition, microhelices have been shown to possess potential as ultrabroadband circular polarizers [32] and tailorable chiroptical sensors. [33][34][35][36] At length of the order 100 nm, nanohelices attract scientific interest due to their optical and magnetochiral properties. [37] Applications such as hydrogen sensors, [38] quantumoptics, [39] biosensing, [40,41] and nonlinear chiroptical spectroscopy [42,43] have been identified. When configured in arrays, helical antennas form coupled hot-volumes (regions of enhanced electric field) that are both uniform and reproducible. [44,45] As such, they have a remarkable potential for SERS. [46][47][48] However, most helical SERS substrates that have been considered so far are rather large, with helix lengths >0.5 µm. [49,50] Here, we demonstrate the potential of dense arrays of metallic nanohelices for SERS with large enhancement factors and high reproducibility. These nanohelical arrays are fabricated by nano glancing-angle deposition (nanoGLAD); as such, they are tailorable, reproducible and are free from surface residue contaminants. [34,51] To the best of our knowledge, these are the smallest nanohelices investigated for SERS-each nanohelix has a total length ≈100 nm and wire thickness of ≈25 nm. As an illustration of the high SERS performance that can be achieved with these nanohelices, we perform two new experiments that reveal near-field enhancement generated with circularly polarized light (CPL). First, we use the Raman response of achiral molecules as a probe. Specifically, using a standard achiral analyte (crystal violet molecules), upon illuminating at the popular (for SERS) wavelength of 532 nm, [52,53] with leftand right-hand CPL, we report a clear difference in the circular intensity difference (CID) of SERS spectra that is attributed to the CPL sensitivity of the nanohelices. Second, we perform a finer characterization of the metasurfaces, this time based on a single, element-specific, achiral molecular vibrational mode. Finite-difference time-domain simulations show that the nanohelices have a distinctive hot-spot distribution that originates from electromagnetic coupling with left-and right-handed CPL, which supports the experimental results. Altogether, these nanohelices are highly attractive for SERS applications in environmental science and beyond; for example, for application in tailorable plasmonic metasurfaces that can manipulate CPL.

Results
The experiment is schematically illustrated in Figure 1. A solution of crystal violet in ethanol was prepared ( Figure 1a) and used to spin-coat the SERS substrates (Figure 1b). Crystal violet was chosen as it has well-established Raman band assignments. Raman spectra were acquired (Figure 1c) with 532 nm laser excitation. For all our substrates, the number of crystal violet molecules in the laser spot region (spot diameter of ≈1.38 µm) was estimated at the order of 10 5 . Further details of the substrate fabrication and the data acquisition parameters can be found in the Experimental Section.
Figure 1d-f summarize two novel ways to reveal the nearfield enhancement generated with CPL at metasurfaces, using the Raman response of crystal violet through SERS. Figure 1d shows illumination of the Ag nanohelices samples with leftand right-hand circularly polarized light (LCP and RCP, respectively). Please note that, in the literature, there are conflicting definitions for the handedness of LCP and RCP. [54][55][56] To avoid ambiguity, here, we use the definition stipulated in our previous research publications (see ref. [42]), which is also stated in the Experimental Section. For LCP and RCP illumination, the intensity of the Raman spectra is recorded (I L and I R , respectively). Then, upon plotting the CID (defined as I R − I L ) spectra (see Figure 1e), a CPL-sensitive response is revealed by the sign of the CID spectrum. Furthermore, the sign and the intensity of single Raman peaks can also be used to reveal the CPL sensitivity of the metasurfaces. To clearly demonstrate this effect, we chose the Raman vibrational mode at 1177 cm −1 , because it is well isolated from potential interference with neighboring peaks. Figure 1f shows the corresponding Raman intensity (in units of counts) for LCP and RCP (i.e., I L and I R ) of crystal violet. There is a clear difference in the intensity measured (i.e., CID at 1177 cm −1 ). Moreover, this difference reverses with the chirality of the Ag nanohelices. Therefore, crystal violet serves as a reporter molecule that reveals the interaction between CPL and the chiral metasurfaces. Figure 2 shows the materials used for this study.
The metasurfaces that we used to investigate SERS are shown in Figure 2, characterized using scanning electron microscopy (SEM), transmission electron microscopy (TEM) and atomic force microscopy (AFM). Figure 2a,b show SEM and TEM images of dense arrays of left-and right-handed Au-based nanohelices, respectively. The Au-based nanohelices are made by co-deposition of Au and Cu (ratio 4:1) with total length = 122 nm, pitch = 51 nm, loop diameter = 50 nm, and wire diameter = 25 nm. Figure 2c,d show SEM and TEM images of dense arrays of left-and right-handed Ag nanohelices, respectively. The Ag nanohelices are alloyed with ≈3% Ti to mitigate oxidation and have total length = 110 nm, pitch = 55 nm, loop diameter = 50 nm, and wire diameter = 25 nm. In Figure S1, Supporting Information, top-down SEM images of the nanohelices are presented; in addition, we demonstrate the reproducibility of the substrates between fabrication batches with Ag nanohelices in Figures S1 and S2, Supporting Information. A sample with G-shaped Au nanostructures was prepared as a representative sample geometry with sharp corners (where the electromagnetic lightning rod effect takes place) and to illustrate a 2D versus a 3D helical geometry, (Figure 2e,g); [57] with dimensions: side length 600 nm, strip width 100 nm, nominal height 25 nm, and separation 200 nm. Importantly, to establish a benchmark for the SERS performance of each substrate, we tested a commercially available VSParticle SERS substrate: Au conglomerate nanoparticles (Au-CNPs) with nominal nanoparticle µL of the crystal violet-ethanol solution was spin-coated onto the SERS substrate; spin-coat deposition time was 10 s at 1000 rpm; spin-coat dry time was 60 s at 1000 rpm. c) Raman optical activity setup; a modified inVia Raman microscope. The light is polarized prior to the Rayleigh filter; an achromatic λ/4 waveplate converts the incident light to circularly polarized light; the analyzer was oriented such that the episcattered Raman light with the same handedness as the incident light was measured. d) Diagram of illuminating left-and right-handed Ag nanohelices with left-and right-handed circularly polarized light (LCP and RCP, respectively) at 532 nm. e) The circular intensity difference (CID, defined as I R − I L ) spectra of crystal violet (achiral molecules) for dense arrays of left-handed (top) and right-handed (bottom) Ag nanohelices; irradiance 1.7 kW cm −2 ; average spectrum from three pairs of uniform square grids each covering a 60 µm × 60 µm square with 13 × 13 surface probe points; integration time = 1 s at each point; vertical axes = counts in 1 s averaged from each point with no cosmic ray removal. f) Peak intensity of the Raman vibrational mode at 1177 cm −1 , for LCP and RCP illumination. diameter = 20 nm (Figure 2f,h). The height profiles of the Au G-shaped nanostructures and the Au-CNPs were established using AFM (Figures 2g-i to 2h-ii). Figure 3 presents the circular intensity sum (CIS, defined as I R + I L ) and CID (defined as I R − I L ) of crystal violet Raman spectra from six sample geometries. For all six, the CIS profile is consistent with the Raman spectrum of crystal violet. By definition, the CIS is not sensitive to the CPL sensitive effects, whereas the CID can be sensitive to intrinsic chirality, extrinsic chirality, [58] linear dichroism, circular dichroism, false chirality, [59] pseudo chirality [60] and angular distribution of photoelectrons; [61][62][63] all of which can be used to manipulate CPL, for example, for applications in spintronics [64] and quantum optical computing. [65] The CID spectra for these six sample geometries are clearly sensitive to the nature of the metasurface.
For the G-shaped nanostructures, the Raman peaks are hardly distinguishable in the CID spectrum. This is because, although the G-shaped Au nanostructures are chiral, their chirality is limited in terms of tridimensionality (top-left of Figure 3). Moreover, because the G-shaped nanostructures are fourfold symmetric, they do not exhibit any anisotropy related effect (e.g., linear dichroism or extrinsic chirality). For the Au-CNPs, the Raman peaks can be both positive and negative in the CID spectrum. This is because these samples are intrinsically random in geometric form, as indicated by the SEM and AFM data presented in Figure 2f  signal varies between locations on the sample surface. Hence, due to the random nature of the sample, in the small regions of illumination (1.38 µm spot diameter), local electromagnetic hot-spots can form with both left-and right-handedness and the overall signal we measure is stochastic. [66,67] Moreover, the Raman peak position also varies randomly (by up to 1 cm −1 ). Even small shifts in the peak position can lead to visible CID peaks. These peaks can be positive or negative, as shows in Figure 3 of the main manuscript. The Raman response of these samples is dominated by contributions from the electromagnetic hot-spots. These hot-spots result from the random assembly that can occasionally bring particles to distances of 1 nm and smaller from each other. The resulting gap plasmons provide the highest possible electromagnetic enhancement, just before entering the quantum tunneling regime. [68] Less intense but more uniform enhancements are provided from the interstitial spaces that surround the regions where nanoparticles touch. To illustrate the importance of such hot-spots, Figures  S3 and S4, Supporting Information, compare and contrast simulated electric field distributions of the G-shaped Au nanostructured SERS substrate and the Au-CNPs with linearly polarized light at 532 nm (and 785 nm, as an example of a different wavelength) based on the AFM data presented in Figure 2g,h. Figure  For all nanohelices, the CID data in Figure 3 show a correlation of the Raman spectra with chirality; for dense arrays of left-handed nanohelices, the CID Raman spectra are positive (indicating more SERS signal for RCP illumination) and for right-handed nanohelices, the CID Raman spectra are negative (indicating more SERS signal for LCP illumination). The effect is more pronounced in Ag than in Au-based nanohelices, due to the plasmon resonance at 532 nm (see Figure S7, Supporting Information, which shows the scattering cross-section of a single Ag and a single Au-based nanohelix on Si with linearly polarized light). [46] The laser power can have a dramatic effect on our results, which we carefully took into account. In Figure 3, the irradiance of 17 kW cm −2 is just enough to resolve the CID of dense  arrays of Au-based nanohelices, but it is slightly too much for the dense arrays of Ag nanohelices, which results in quenching effects. The effects of quenching and fluorescence, particularly at 532 nm at higher irradiance can be stark. To illustrate this point, Figures S8-S13, Supporting Information, show Raman spectra for crystal violet on each SERS substrate and plain Si with increasing laser power (linearly polarized 532 nm). Fluorescence and quenching become increasingly conspicuous, and the Raman features diminish into the fluorescence background as the laser power increases. Laser damage threshold studies are presented in Figures S14 and S15, Supporting Information. For LCP and RCP illumination, quenching effects were found to be a significant competing process with the Raman effect. Figures S16 and S17, Supporting Information, show how quenching effects can dominate the CID Raman spectra if the irradiance is too high, leading to large random errors. Hence, CIS and CID Raman spectra for the Ag nanohelices with a tenth of the irradiance (1.7 kW cm −2 ) are presented in Figure  S18, Supporting Information. At this lower irradiance, the CID Raman spectra for dense arrays of left-and right-handed Ag nanohelices appear well resolved.
SERS spectroscopies are widely employed for molecular fingerprinting and identification. A significant further step in sensitivity can be taken by focussing on a single molecular, element-specific vibration mode. Therefore, we next show that our nanohelices are both highly reproducible as SERS substrates and that (upon averaging) the CPL-dependent effects of metamaterials can be revealed by probing single, achiral, molecular vibration modes. To this purpose, the height of the Raman band at 1177 cm −1 (the CH in plane bending mode in crystal violet) [69] was extracted from the Raman spectra for three pairs of Raman maps-three with RCP and three with LCP. Figure 4 presents the height of the Raman band at 1177 cm −1 from each Raman measurement for the data presented in Figure 3 and Figure S18, Supporting Information.
In Figure 4, it is immediately apparent that the SERS signals from all samples are reproducible. The Raman intensity from the dense arrays of Ag nanohelices dominates.
Remarkably, upon averaging the 1177 cm -1 peak intensities for each sample, depending on the direction of circularly polarized illumination signals (see red numbers in bold), the optical behavior of the metasurfaces can again be revealed. For the dense arrays of left-and right-handed Au-based nanohelices, the averaged CID is +36 counts per second and −45 counts per second, respectively. At the same irradiance of 17 kW cm −2 , for the dense arrays of left-and right-handed Ag nanohelices, the CID is +1846 and −337 counts per second, respectively. As we have seen, for this irradiance, quenching can occur in the Ag nanohelices; therefore, measurements were also performed with a lower irradiance of 1.7 kW cm −2 . Figure 4 shows that the corresponding 1177 cm -1 peak intensities, for left-and righthanded Ag nanohelices are +16 and −25 counts per second, respectively, confirming the behavior. We note that the dependence of the CID sign on the handedness of the metasurfaces and on the direction of CPL matches the CID presented in Figure 3 and Figure S18, Supporting Information, for full Raman spectra.
The dense arrays of Ag and Au-based left-handed nanohelices display clear Raman reproducibility, depending on the direction of CPL illumination (<4% standard deviation). The data for right-handed nanohelices on the other hand are more noisy (standard deviation ≈10% in right-handed Au-based nanohelices, standard deviation ≈30% in the right-handed Ag nanohelices at 17 kW cm −2 ). These large standard deviations though are due to using too high-power (the power level here was set for consistency between all samples). Indeed, for Ag nanohelices at 1.7 kW cm −2 irradiance, the standard deviations are ≈3% and ≈7%, for left-and right-handed Ag nanohelices, respectively. The higher standard deviation in the case of the right-handed nanohelices can be attributed to the quality of fabrication and/or spin-coating of this sample. In Figure 2, the TEM and SEM pictures are representative of the samples at study. Some variation in dimensions is clearly present between individual nanohelices. There is also a clear difference in total size of the nanostructures between left-and right-handed nanohelices. These physical differences contribute to the standard deviation values we have obtained. Surface roughness can lead to noisy SERS and stochastic quenching effects due to the lightning rod effect. [70] In order to shed light on the origin of the observed results, we performed numerical simulations of the metal nanostructures upon illumination with CPL. Figure 5 presents the simulated electric-field distributions in the cross-section of a dense array of left-handed Au-based and Ag nanohelices for both LCP and RCP. Further simulations of single isolated nanohelices with and without the Si substrate are presented in Figure S19, Supporting Information. The maximum electric-field strength was also extracted from the distributions, and it is presented in Figure S20, Supporting Information. In the case of Ag nanohelices, Figure 5 shows very clear differences between the electric field distributions for LCP and RCP. By contrast, the differences in electric field distribution, for LCP and RCP illumination of the Au-based nanohelices are more subtle. This pronounced contrast of behaviors can be attributed to the different plasmon resonances; while for Ag nanohelices, a broad plasmon resonance occurs at 475 nm and overlaps with 532 nm, for Au-based nanohelices, the plasmon resonance at 658 nm does not overlap with 532 nm, see Figure S7, Supporting Information. The plasmon resonances in Au nanoparticles are redshifted with respect to those in Ag nanoparticles. Matching the plasmon resonance with the excitation wavelength typically leads to enhanced SERS signals. [46] Accordingly, at 532 nm, our Ag nanohelices provide a stronger SERS signal than the Au nanohelices.

Discussion
Fluorescence and quenching of Raman bands can be a significant factor in selecting the excitation wavelength for Raman spectroscopy, particularly when detecting organic materials and microplastics using SERS. [71] The local optical field responsible for SERS also enhances the excitation and emission fields responsible for fluorescence. [9] Typically, fluorescence can be mitigated by using longer excitation wavelengths. [72] However, except in the case of specific resonances, the Raman crosssection significantly decreases with increasing excitation wavelength. [46] Here, we use the popular excitation laser wavelength of 532 nm. Fluorescence and quenching was easily mitigated by reducing the laser irradiance on the sample and by taking Raman data from large area maps.
In the case of the G-shaped Au nanostructures, hypothetically increasing the thickness from 25 nm to ≈100 nm could increase the SERS intensity, due to the increase in interaction volume and surrounding void. To illustrate this point, Figure S21, Supporting Information, shows the simulated effect of increasing the thickness of the G-shaped Au nanostructures. In the case of 532 nm, the maximum electric field strength increases and eventually exceeds that of the Au-based nanohelices. However, in reality, the G-shaped Au nanostructures were fabricated with electron beam lithography, where the larger thicknesses would result in tapered edges around the nanostructures and lead to significantly more irregularities in the uniformity. Given the significant cost associated with electron beam lithography, the nanohelices (which are far cheaper to produce over large areas) are a preferable substrate for SERS applications.  . Reproducible CPL-sensitive SERS from metasurfaces upon probing a single, element-specific, achiral molecular vibration mode. Illumination wavelength: 532 nm. a) Example Raman spectrum of crystal violet with left-handed circularly polarized (LCP) light and right-handed nanohelices SERS substrate. SERS for each substrate was quantified using the peak height in crystal violet at 1177 cm −1 as indicated in red. In each case, the peak height was measured with (solid) and without (hatched pattern) applying fluorescence background removal. The spectrum in (a) was with irradiance 17 kW cm −2 ; average spectrum from three uniform square grids each covering a 60 µm × 60 µm square with 13 × 13 surface probe points; integration time = 2 s at each point; vertical axes = counts in 2 s averaged from each point with no cosmic ray removal. b) Measured crystal violet SERS peak height at 1177 cm −1 with LCP and right-handed circularly polarized light (RCP); incident laser wavelength = 532 nm; each column is averaged from a single uniform square grid, each covering a 60 µm × 60 µm square with 13 × 13 surface probe points on the same sample surface; integration time = 2 s at each point and irradiance = 17 kW cm −2 (left of dashed line) and 1 s at each point and irradiance = 1.7 kW cm −2 (right of dashed line); no cosmic ray removal. Each color represents the different types of SERS substrates-see legend below. The red numbers above the columns are the average Raman peak height at 1177 cm −1 for LCP and RCP, respectively. Hatched patterns indicate peak height without fluorescence background removal; solid fill indicates peak height after fluorescence background removal; left-and right-slanting hatch patterns are LCP and RCP, respectively.
The CID data we present here are not limited to intrinsic chirality. CPL-sensitivity in metasurfaces can be due to the anisotropy of the samples. [73] In our case, the arrays of nanohelices are strongly anisotropic, with a sixfold symmetric local order and with all helices sharing the same orientation. This latter order could also be responsible for extrinsic chirality. Previously, we have shown that, at large angles of incidence, these anisotropy effects can enhance and tune the intrinsic chiroptical effects, thereby modulating the CLP-sensitivity of the metasurfaces. [74,75] Although in the present experiments the illumination is at normal incidence, which minimizes the effects of anisotropy, our results are general. For reference, Figure S22, Supporting Information, presents the far-field ellipticity spectra of the samples, at normal incidence.
The CID also has its limitations. Since CID is an indirect method that relies on reporter molecules, any effect that affects the molecules represent a potential limitation. Too much laser power denatures the molecules. An increase in sample temperature (e.g., by plasmonic absorption) could also diminish and shift the Raman peaks. Depending on the choice of reporter molecules, chemical side reactions could occur.

Conclusion
Densely packed arrays of plasmonic nanohelices (length ≈ 100 nm) fulfil the need for reproducible SERS templates. To the best of our knowledge, these are the smallest nanohelices investigated for SERS. The SERS performance of Ag and Au-based nanohelices was assessed alongside a commercial Au SERS substrate and a planer counterpart-nanospirals (G-shaped Au nanostructures) with the same thickness as the diameter of the nanohelices. The dense arrays of Ag nanohelices, clearly outperform all substrates studied here. Finite-difference timedomain simulations revealed that the nanohelices present hot-spot patterns that depend on the direction of circularly polarized illumination. Because of their 3D structure and corresponding large surface area, the nanohelices offer extended regions of enhanced electromagnetic fields-hot-regions. Increasing the size of the nanohelices increases the interaction volume with the analyte, however there is a tradeoff-the plasmon wavelength shifts toward the IR. [40,45] The order within the arrays enables uniform and reproducible SERS enhancement factors, which are crucial for SERS applications. Due to their chiral shape, the nanohelices couple preferentially to CPL. Because of their high performance as SERS substrates, this preference becomes apparent in the Raman spectra of crystal violet. Remarkably, it can even be detected in the intensity of single Raman peaks (upon averaging several sample areas). The nanohelices can be prepared over large areas, at a low cost and they provide strong, robust and uniform Raman enhancement. In future studies it will also be interesting to evaluate the dependence of the SERS signal on the analyte concentration. Evaluating the SERS response at other wavelengths will also be of interest.
We present two new ways of probing the CPL-sensitivity of metasurfaces. Densely packed arrays of chiral plasmonic nanohelices show a clear CID in Raman scattering. The observed behavior was largest in the Ag nanohelices. We see that quenching effects can impact the reproducibility but they are effectively mitigated by reducing irradiance and by scanning large sample areas. Our experimental results are supported  by numerical simulations that show a distinctive difference in electromagnetic coupling between the nanohelices and CPL. Crucially, we used achiral molecules as CID reporters situated in the near-field of the chiral SERS substrates. Our methods complement existing near-field probes for CPL-sensitive metasurfaces, such as scanning near-field optical microscopy (SNOM), [76,77] where, however, the influence of the tip can be a complicating factor. [78] Our methods can be further amplified and endowed with imaging capabilities by using tip-enhanced Raman spectroscopy on the chiral metasurfaces. Naturally, our methods could also be generalized, to study other chiral metasurfaces. Therefore, we anticipate that both the materials and the methods presented here will have a far-reaching impact in the realm of advanced materials and metasurface science.

Experimental Section
Dense Arrays of Ag-and Au-Based Nanohelices SERS Substrate Fabrication: The 3D nanohelices were grown on Si (001) wafer (5 nm SiO 2 layer) using oblique angle deposition [79] via the nano glancingangle deposition (nanoGLAD) technique, which is now an established state-of-the-art nanofabrication method. [42,[80][81][82] Block-copolymer micelle nanolithography (see ref. [83]) was employed to arrange Au dots (10 nm diameter for the Ag nanohelices and 14 nm for the Au-based nanohelices) in a hexagonal array that acted as a seed for the nanoGLAD process. For the Ag nanohelices, the seed spacing was 90 nm; for the Au-based nanohelices, the seed spacing was 79 nm. The nanohelices were then grown from the Au dots using a nanoGLAD system that allowed alloying by co-deposition of two metallic elements from two electron beam evaporators at a base pressure ≈10 −6 mbar. First, the substrates were cooled to 90 K with liquid nitrogen for 1 h and tilted to 87° with respect to the direction of flux. A quartz crystal microbalance was used to monitor the deposition rates of each evaporator independently allowing control over the alloy ratio. The Ag nanohelices had ≈3% Ti added during the nanoGLAD process whereas the Au-based nanohelices had a 4:1 ratio of Au to Cu. The samples were then rotated about the normal axis; the rate of rotation controls the resulting pitch of the nanohelices. The direction of rotation determines the handedness of the nanohelices.
G-Shaped Au Nanostructured SERS Substrate Fabrication: The G-shaped Au nanostructured SERS substrate was fabricated on a Si (001) wafer with a thermally grown 100 nm SiO 2 top layer. The wafer was cleaned first in H 2 SO 4 /H 2 O 2 and then in NH 4 OH/H 2 O 2 /DIW mixtures; it was then rinsed in pure deionized water. The substrate was then covered by a double resist layer of poly(methyl methacrylate) (PMMA)/co-PMMA; the thickness was 200 nm and 250 nm, respectively, and was applied by spin-coating. High resolution electron beam lithography (VB6 system from Leica Microsystems Lithography) was employed to create the stencil for the G-shaped nanostructure pattern by removing the top layer of the resist. The bottom layer of the resist (co-PMMA) was dissolved by immersing the substrate in a mixture of isopropyl alcohol and methyl isobutyl ketone. The mask was finalized by annealing. A thin layer of Ti was applied by evaporation to assist adhesion of the Au that was applied by evaporation to a thickness of 25 nm. The remaining resist layer was then removed by immersion in warm acetone or dichloromethane for a few minutes followed by an ultrasonic bath, for a few seconds. The sample was then removed, rinsed in iso-propanol and dried under the flow of nitrogen gas. Further details can be found in ref. [84].
Au Conglomerate Nanoparticles (Au-CNPs) SERS Substrate Fabrication: Au nanoparticles (NP) were prepared by spark ablation using a VSPARTICLE nanoparticle generator (VSP-G1) and deposited onto Si substrates using the VSPARTICLE nanostructured material printer (VSP-P1). With spark ablation, pure nanoparticles of any (semi)conducting material can be produced. The particles were created inside a carrier gas and were free of surfactants or other organic impurities. The primary particle size can be tuned by varying the carrier gas flow or the spark power. In the VSP-P1, the NP aerosol from the VSP-G1 passes through a nozzle into a vacuum chamber (≈1 mbar), equipped with XYZ stage, where the particles can be printed on a substrate by raster scanning. Particles produced by the VSP-G1 generally have a primary particle size below 20 nm. Larger sizes can be achieved by adding an in-flight sintering step before particle deposition in the VSP-P1. The Au NPs were produced in a VSP-G1 nanoparticle generator (VSPARTICLE, the Netherlands) fitted with Au electrodes (4N, ∅3 mm, VSPARTICLE) in a flow of 1 L min −1 argon (5N, Linde gas) as carrier gas and a spark power of 13 W (1.3 kV, 10 mA). The aerosol was passed through a furnace at 200 °C, to increase primary particle size by in-flight sintering. The resultant nanoparticle aerosol was fed into the VSP-P1, where a 4 mm by 4 mm SERS-active layer was deposited on Si wafer by printing parallel lines at a rate of 3 mm 2 min −1 . Samples were stored under ambient conditions before measurement.
SERS Substrate Characterization: The substrates were characterized using SEM, TEM, and AFM. The SEM micrographs in Figure 2, and Figures S1 and S2, Supporting Information, were acquired with a JEOL JSM-7900F Schottky Field Emission SEM. The AFM micrographs in Figure 2 and Figures S3-S5, Supporting Information, were acquired using a Multimode Scanning Probe Microscope (VEECO) operating in contact mode. The TEM images in Figure 2 were acquired using a JEOL JSM-2100PLUS. For TEM, a small square (≈4 mm × 3 mm) of nanohelices on Si wafer was cut and sonicated in a 0.7 mL of solvent for 20 min before deposition (few microliters) onto TEM grids-Au-based nanohelices: chloroform, Formvar TEM grids; Ag nanohelices: ethanol, carbon coated Cu TEM grids.
Simulations: Finite-difference time-domain simulations were performed in ANSYS-Lumerical to illustrate the electric field distributions, revealing the nature of the local field enhancements or hotspots. The simulation domain had periodic boundary conditions applied in the x and y directions to the edges of the unit cell (see Figures S5 and S6, Supporting Information, for dimensions). The domain in the vertical axis spanned −1.5 µm to 3 µm depending on the size of the substrate model and had perfectly matched layer boundary conditions. The Eulerian mesh in the regions of interest was 5 Å for the nanohelices, 2.5 nm for the Au-CNPs substrate, and 2.5 nm for the G-Shaped Au nanostructured substrate. The optical properties of the Si wafer and SiO 2 layer were modelled with an empirical based material model from Palik. [85] The nanohelix substrates had a 5 nm layer of SiO 2 ; likewise, 2 nm for the Au-CNPs; and 100 nm for the G-Shaped Au nanostructures. The optical properties of the Au-based nanohelices were modelled using a 4:1 linear combination of the CRC [86] material models for Au and Cu. CRC based material models were also used for the optical properties of the Ag nanohelices and substrates with Au nanostructures. For linearly polarized light simulations, a pulsed plane wave source of light was incident on the models from 1 µm above the surface; the light was polarized parallel to the x-axis and had an amplitude of 1 V m −1 . For simulations with CPL, two orthogonally polarized plane-wave sources were superimposed with a 90° phase difference. The simulated wavelength range was 250 nm to 2.5 µm. The electric-field distributions were extracted at 532 and 785 nm from the cross-sectional planes indicated in Figures S3-S5 and S6a, Supporting Information. The scattering cross-section spectrum ( Figure S7, Supporting Information) was obtained using a Mie source in Lumerical with a scattering analysis function which computed the scattering cross-section based in the energy flux through the bounding box.
To perform numerical smulations of the sample geometries obtained by the AFM data (geometries also shown in Figure 2d,e), the AFM data for the G-shaped Au nanostructures and Au-CNPs were exported into three column text data files (.txt) using Gwyddion. These x, y, and z coordinates were then reshaped into a grid in Python and imported as a surface into ANSYS Lumerical. The Au surface topography was then superimposed onto the relevant Si/SiO 2 layered substrate model. The models shown in Figures S5c,d and S6a, Supporting Information, were generated using Autodesk Inventor based on dimensions extracted from the SEM micrographs presented in Figure 2.
Crystal Violet: Analytical standard crystal violet was purchased from Sigma Aldrich. A 100 mL solution (concentration 0.3 × 10 −3 m) of crystal violet dissolved in ethanol was prepared. For data shown in Figures S8-S15, Supporting Information, ≈50 µL of the crystal violet solution was drop cast onto the substrates and allowed to dry naturally. For all data presented in Figures 3 and 4, and Figures S16-S18, Supporting Information, the substrates were spin-coated using a Laurell spin-coater (model: WS-650MZ-23NPPB). To spin-coat the substrates, 200 µL of crystal violet solution was drop-cast within for the first 10 s of spinning and then allowed to continue spinning for 60 s at 1000 rpm.
Definition of Left-and Right-Handed Circularly Polarized Light: Lefthanded circularly polarized light (LCP) is defined as follows: looking from the point of view of the source, along the direction of propagation, the electric field of LCP traces a helix in the space that curls anti-clockwise.
Raman Spectroscopy-Linearly Polarized Light: Raman spectra were acquired using a Renishaw inVia Raman microscope. The incident light source for 532 nm was a polarized continuous wave narrow bandwidth laser (Cobolt RL532-08; 50 mW). The irradiated light and episcattered Raman light were focused and collected through an N-plan 50× objective with a numerical aperture of 0.75. Spectra in Figures S8-S13, Supporting Information, were averaged from a 40 µm × 40 µm square grid of 5 × 5 (25) uniformly distributed points; each separated by 10 µm. At each sample point, the spectrum acquisition was a total 10 s with an integration time of 1 s. The spectral resolution was 1.6 cm −1 .
For experiments with the 532 nm continuous wave Cobolt laser, the laser power at the sample was varied between 80 µW and 25 mW using neutral density filters. The irradiance was computed by taking the measured power under the objective and dividing by the area of the spot size of the laser. In focus, the measured laser spot diameter was 1.38 ± 0.09 µm.
Laser Power, Spot Diameter, and Irradiance: The irradiance for the spectra was calculated using the measured laser spot-diameter and power. The laser power at the sample was measured using a Thorlabs S175C -Microscope Slide Thermal Power Sensor. To measure the laser spot diameter at the image plane, a Raman line scan was acquired across a sharp (approximately nanometers scale) Si-to-metal layer edge using a Si wafer with a portion of metallic film (≈100 nm aluminum). Then, by interpolating a cubic spline function (scripted in Python) through the Raman peak height of the O Γ -point phonon in Si (at 520 cm −1 ) as a function of position across the edge, the diameter could be determined from the full-width at half-maximum of the first derivative of the intensity profile. This is illustrated in Figures S16a, S17a, and S23, Supporting Information. For the 50× objective, the separation between measurement points in the line scan was 0.1 µm; for the 5× objective, the separation was 1 µm.
Raman Spectroscopy-Circularly Polarized Light: Raman data were acquired using a modified Renishaw inVia Raman microscope (see Figure 1). The incident light source for 532 nm was a polarized continuous wave narrow bandwidth laser (Cobolt RL532-08; 50 mW). A Glan-laser polarizer was used prior to the Rayleigh filter and a λ/2plate (not shown in Figure 1) was placed at the output of the laser to optimize power throughput. An achromatic λ/4-plate was placed after the Rayleigh filter to compensate for the polarization effects for the Rayleigh filter. The orientation of the λ/4-plate was coarsely optimized to mitigate ellipticity at the sample using a zero-order λ/4plate, a wire-grid polarizer was then placed above a power meter. The analyzer was a wire-grid achromatic polarizer with an achromatic λ/2-plate in tandem, to optimize for the polarization sensitivity of the spectrometer; the orientation of the λ/2-plate was optimized with a Si sample using the linearly polarized light (no λ/4-plate). With these optics in, the orientation of the λ/4-plate was finely optimized using a piece of polycrystalline ZnSe, an N-Plan 5× (NA: 0.12) objective and a wide spectrometer slit (150 µm) to ensure parity between left-handed and right-handed circularly polarized light; in all other measurements, the spectrometer slit width was 65 µm. The optimization data can be accessed through the University of Bath data archive (see Data Availability Statement).
CPL-sensitivity SERS experiments were performed using an N-plan 50× objective (numerical aperture of 0.75); the data presented in Figure S17, Supporting Information, were collected using the N-plan 5× objective (NA: 0.12). The circular intensity sum and difference spectra in Figure 3 and Figure S18, Supporting Information, were averaged from three pairs of 60 µm × 60 µm square grids (13 × 13 = 169) uniformly distributed points; each point separated by 5 µm and each grid separated by ≈100 µm. The CID measurement can be sensitive to minute differences in analyte concentration. Hence, the areas from which LCP and RCP was measured were very close to eachother (≈100 µm separation). To minimize the variation in analyte concentration on the surface, the CV was spin-coated from an ethanol solution. At each sample point, the integration time was a total of 2 s for data at 17 kW cm −2 and 1 s for data at 1.7 kW cm −2 . The spectral resolution was 1.6 cm −1 for the spectra with 532 nm excitation.
To establish the peak height relative to the baseline of the spectra in Figure 4, the fluorescence background was removed using Renishaw's built-in 11th-order polynomial Intelligent Fitting algorithm ("subtract baseline" tool) in WiRE-version 5.3. For all Raman data, the irradiance was computed by taking the measured power under the objective and dividing by the area of the measured spot size.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.