Development of one-step roll-to-roll system with incorporated vacuum sputtering for large-scale production of plasmonic sensing chips

Surface-enhanced Raman spectroscopy (SERS) has attracted intensive attention in the validation of chemical/biomedical materials because inelastic Raman signals provide unique vibrational information for individual molecules.^1–4 The scattered light is remarkedly amplified by electric fields confined in dielectric spaces (i.e., hotspots) between plasmonic nanostructures, which is known as the localized surface plasmon resonance (LSPR) effect.^5–7 SERS platforms exhibit state-of-the-art enhancement factors, which enable sensitivity at the single-molecule level.^8,9 Therefore, engineering techniques strongly influence SERS detection performance. Top-down approaches, such as high-vacuum and lithographic methods, are used to fabricate the highly symmetric nanostructures (i.e., coherent electron oscillations) that enable ultrasensitive and reproducible SERS performance; however, the complexity, prolonged processing times, and cost-ineffectiveness of these systems still pose challenges.^10–13 Meanwhile, plasmonic nanoparticles can be facilely synthesized using bottom-up techniques, but precisely controlling their size, shape, and size distribution (i.e., non-uniformity) is difficult.^14–17 The trade-off relationship between cost and performance hinders the use of SERS platforms in commercial applications. Therefore, the development of innovative engineering techniques is necessary to satisfy the requirements in practical fields.

Many attempts to fabricate populated and quasi-symmetric plasmonic nanostructures using simplified procedures have been reported.^18–20 Among these methods, maskless plasma etching is extensively used to directly define nanoscale textures over large areas in a single step, which is convenient compared with array patterning processes (e.g., electron beam lithography, extreme-ultraviolet lithography, photolithography, and focused-ion-beam milling).^10,11 In particular, polymers are considered strong candidates for SERS-supporting nanostructures [e.g., nanotunnels, nanowires, nanopillars, nanocavities, and nanodimples (NDs)], whose morphological properties can be tuned by controlling the gas dose, pressure, temperature, and applied power.^21–23 These beneficial properties are attributed to polymeric molecules’ high volatility and low molecular weight, which impart them with surface dynamics that favor, for example, selective etching, surface migration, agglomeration, and coalescence during plasma treatment.²¹ In addition, polymer substrates allow flexibility in fabrication systems and processes.

The roll-to-roll system is the key technology for the large-scale, low-cost, and high-throughput production of micro-/nanostructures on polymer films in a broad range of industries, including flexible electronics, organic photovoltaic devices, and displays.^24–26 Ultraviolet or thermal nanoimprint lithography methods are adopted in conventional systems to transfer patterns from primary molds to secondary substrates.^27–29 Therefore, the fidelity of the replica is strongly influenced by the pressure- or thermo-forming parameters and the moving speed.³⁰ However, these replications suffer from mold contamination and environmental effects (i.e., temperature and humidity). In addition, for the fabrication of SERS platforms, additional systems/steps are required to accommodate plasmonic materials: for example, reactive ion etching for the removal of residual polymers while sputtering and evaporation for the deposition of noble metals.^31–34 To overcome these issues, the development of enclosed roll-to-roll systems with plasma generators is needed.

In the present work, we demonstrate a roll-to-roll process combined with ion beam sputtering (IBS) and sputtering deposition systems for the one-step fabrication of Ag-coated nanodimples (Ag/NDs) on A4-scale (210 × 297 mm²) substrates. The plasmonic hotspots were optimized by adjusting the moving speed and Ag thickness. We investigated the fluctuations in SERS signals from multiple batches and chips to estimate the repeatability and reproducibility of the fabrication method. A quantitative study was conducted to examine the sensitivity and limit-of-detection (LOD) of the Ag/ND SERS platforms. The numerical calculation of electric fields was also performed to verify the hotspot characteristics. Our study bridges the gap between microscale precision and macroscale production. The combination of the IBS for ND formation and subsequent sputtering deposition for Ag coating represents a novel approach in the scalable manufacturing of SERS platforms, leveraging the advantages of vacuum roll-to-roll technology for consistent and high-quality production. This innovative technique ushers in a new era of material conservation and processing efficiency.

For SERS platforms to be used in commercial applications, it is necessary to overcome conventional SERS platforms’ trade-off relationship between cost and performance. To satisfy both requirements, we developed high-throughput systems integrated with vacuum sputtering and roll-to-roll processes for the production of A4-scale SERS platforms in series (Scheme 1). This equipment consisted of three main parts (i.e., an unwinder, a drum, and a rewinder), and the IBS and sputtering deposition systems were incorporated into the drum section. A closed drift-type anode-layer linear ion source with a width of 300 mm was used. For cooling the equipment, water at 20 °C was circulated in the drum. The polyethylene terephthalate (PET) was used as a carrier film to mount and transfer the supporting substrates. The NDs with three-dimensional concave volumes were defined on the polyethylene naphthalate (PEN) films via the IBS process. According to our previous studies, the morphological properties (i.e., the size and degree of symmetry) of the PEN NDs could be tuned by adjusting the gas dose, beam power, pressure, moving speed, and cycle.^35,36 In the present work, we fabricated two types of PEN ND patterns: (i) patterns prepared with the films moving at 0.3 meters per minute (mpm) for 30 cycles (denoted as ND 0.3 mpm) and (ii) patterns prepared with the films moving at 0.6 mpm for 60 cycles (denoted as ND 0.6 mpm). The plasmonic material (herein, Ag) was sputter-deposited onto the as-prepared PEN ND films. In this period, the transport speed of the films was set to be 1 mpm and the deposited Ag thickness was estimated to be 25 nm per pass. Ag with various thicknesses (i.e., 50, 100, 150, 200, and 250 nm) was prepared by adjusting the number of passes (i.e., 2, 4, 6, 8, and 10, respectively). A photograph of a fabricated A4-scale Ag/ND film is shown in Fig. S1. Individual sensing chips with sizes suitable for various applications were obtained via the dicing process. Such a fully automatic and seamless roll-to-roll process enables the facile, large-scale, cost-effective, and reproducible production of Ag/ND SERS substrates, which is highly desirable for the identification of chemical/biomedical molecules in clinical fields.

The formation of plasmonic nanostructures was observed through morphological changes in the supporting PEN films during the roll-to-roll process. Before the IBS treatment, the PEN substrates exhibited negligible surface roughness [Fig. 1(a)]. When the ion beam was incident to the PEN, the NDs were readily formed via the polymeric surface dynamics under the O₂ environment [Fig. 1(b)]. Ag was conformally rendered on the underlying structures, resulting in the successful formation of Ag/ND platforms [Fig. 1(c)]. Using the ImageJ analysis software, we estimated the average distance between the neighboring protrusions for the 150 nm Ag/ND 0.6 mpm platform to be 14.8 nm, demonstrating the strong potential of the platform for near-field enhancement. To elucidate the optimum plasmonic activities, we investigated the influence of the Ag thickness on the optical properties of the Ag/ND platforms. The SERS signals of 10 μM 4-aminothiophenol (4-ATP) were collected from the PEN ND 0.6 mpm with various Ag thicknesses (i.e., 0, 50, 100, 150, 200, and 250 nm) [Fig. 1(d)]. Herein, ethanol solvent was used to ensure homogeneous molecular distributions (i.e., to prevent the coffee-ring effect).^37,38 The PEN ND 0.6 mpm exhibited negligible background interference because of the absence of plasmonic materials, which provided a high signal-to-noise ratio. From the Ag-deposited substrates, the 4-APT signals were observed. For the 4-ATP molecule, the peak at 1078 cm⁻¹, corresponding to C–S stretching, was selected as the analytic criteria. The assignment of peaks in the spectrum of 4-ATP is summarized in Table S1.^39,40 The performance of plasmonic hotspots was indicated by the variation in the intensity of the peak at 1078 cm⁻¹ [Fig. 1(e)]. The results show that the maximum peak intensity was observed from 4-ATP on the 150 nm Ag/ND 0.6 mpm. At lower Ag thicknesses (i.e., 50 and 100 nm), the plasmonic activities became weak because the electric field strength was inverse-exponentially proportional to the hotspot size; that is, the LSPR resonance effect was observed. At greater Ag thicknesses (i.e., 200 and 250 nm), the aggregated structures substantially dissipated the electric fields. Dark-field scattering analysis was conducted to investigate the plasmonic behaviors [Fig. 1(f)]. Under illumination, for all the Ag/ND substrates, Rayleigh scattering was induced by LSPR excitation in the visible range because Ag had no interband transition. Notably, the highest scattering intensity at 785 nm was achieved from the 150 nm Ag/ND 0.6 mpm, consistent with the results in Fig. 1(d). The true color of the scattered light from the substrates differed depending on the Ag thickness (Fig. S2), indicating wavelength-selective scattering.

To explore the effect of the moving speed, we also examined the morphological and optical properties of the Ag/ND 0.3 mpm platforms (Fig. S3). Under exposure to higher-energy ion beams, adjacent nanostructures merged, resulting in PEN NDs with larger scales. Compared with the hotspot scale of the 150 nm Ag/ND 0.6 mpm, that of the 150 nm Ag/ND 0.3 mpm was expanded to 27.2 nm, which led to a relatively weak plasmonic performance. The SERS signals of 4-ATP were consequently degraded. To ensure the formation of quasi-symmetric Ag/NDs via the roll-to-roll process, the SERS mapping was conducted over large areas (i.e., 500 × 500 μm²) with an interval of 50 μm (Fig. S4). The influence of moving speed on the SERS performance of the 150 nm Ag/ND platforms was also investigated (Fig. S5). We fabricated the 150 nm Ag/ND 0.1 mpm (10 cycles) and 150 nm Ag/ND 0.9 mpm (90 cycles). At lower speeds, the SERS signals became weakened due to the hotspots with larger scale and lower densities. Moreover, the intense power of exposed O₂ ion beam would cause the thermal degradation of polymeric nanostructures with low thermal conductivities. At higher moving speeds, more number of cycles would make the differences in surface dynamics of polymeric nanostructures due to the influence of acceleration and deceleration when the winding direction is changed. The 150 nm Ag/ND 0.9 mpm platform exhibited higher 4-ATP signals, but its RSD value was significantly increased up to 16.3%. The density profiles of the 150 nm Ag/ND platforms with different moving speeds were analyzed using the SPIP analysis (Fig. S6). Accordingly, the optimum moving speed and Ag thickness were determined to be 0.6 mpm and 150 nm, respectively. Given the surface oxidation of Ag layers, the stability of the developed platforms was also investigated. The 150 nm Ag/ND 0.6 mpm platforms were kept under an atmospheric environment. The intensity of the peak at 1078 cm⁻¹ was decreased by 20% over a period of three months (Fig. S7). The stability would be further improved when the plasmonic substrates are stored in a desiccator.

The hotspots of the Ag/ND platforms prepared with different moving speeds were further investigated by scanning transmission electron microscopy (STEM). The cross-sectional features of the PEN NDs were also observed, revealing a pore radius of ∼100 nm (Fig. S8). For both of the Ag/ND platforms, we observed that the Ag layer was conformally formed on the PEN NDs because of the excellent step coverage of the sputtering system.⁴¹ The three-dimensional spaces between the Ag nanostructures were functionalized as the plasmonic hotspots according to the conventional LSPR phenomenon. Compared with the Ag/ND 0.3 mpm [Fig. 2(a)], the Ag/ND 0.6 mpm with narrower and shallower hotspots [Fig. 2(b)] could provide more intense plasmonic resonance. Therefore, the Ag/ND 0.6 mpm platforms were expected to enable more sensitive molecular detection, consistent with the tendency of the optical behaviors (Figs. 1 and S3).

To theoretically verify the plasmonic activities of the Ag/ND platforms, we evaluated their electric field distributions using the finite-difference time-domain (FDTD) method. The simulation model was constructed on the basis of field-emission scanning electron microscopy (FE-SEM) and STEM images. To simplify the structures for large-scale calculation, we made three assumptions: (i) the Ag/NDs were considered arrays of the Ag tips, (ii) the Ag tips were arranged in a hexagonal lattice, and (iii) the scale and distribution of the Ag tips were determined on the basis of statistic computations. For both the Ag/ND 0.3 mpm and Ag/ND 0.6 mpm platforms, under 785 nm illumination, electric fields were observed to be confined in the spaces between the adjacent Ag tips [Fig. 2(c)]. Compared with the Ag/ND 0.3 mpm, the Ag/ND 0.6 mpm contained more sites for LSPR excitation as a result of their different hotspot scales. The wavelength dependency of the Au/ND platforms under 633 nm illumination was also studied with the induced electric field (Fig. S9). Both the population and intensity of the LSPR excitation decreased, in agreement with the dark-scattering spectra [Fig. 1(f)].

For the Ag/ND SERS platforms to be used in actual applications, it is necessary to achieve uniform signals from multiple chips and batches. To investigate the stability of the roll-to-roll fabrication process, the SERS spectra of 10 μM 4-ATP were acquired from ten different batches of substrates [Fig. 3(a)]. For all the examined substrates, the 4-ATP signals exhibited similar spectral features. To evaluate the signal uniformity, we extracted the intensity profiles for the peaks at 1078, 1178, 1490, and 1594 cm⁻¹ [Fig. 3(b)]. The fluctuation in intensity was evaluated using the relative standard deviation (RSD) calculated using the equation RSD = S × 100/σ, where S is the standard deviation and σ is the mean value. The obtained RSD values were 4.8%, 9.1%, 7.4%, and 8.2% for the analytic peaks at 1078, 1178, 1490, and 1594 cm⁻¹, respectively. These results indicate that the roll-to-roll process is sufficiently stable for the large-scale production of plasmonic sensing chips because of its fully automatic one-step operation without any side effects originating from contamination and misalignment in multiple processes and equipment.

The reproducibility of the SERS signals was estimated using the large-scale mapping method [Fig. 3(c)]. In the present work, the SERS spectra were collected from 100 different points on a single chip. The mapping data were acquired in a square area of 500 × 500 μm² with an interval of 50 μm. Because the substrates were immersed in ethanol solvent, the molecules were assumed to be uniformly distributed on the chip because of the absence of droplets’ pinning effect at edges (i.e., the coffee-ring effect). 4-ATP signals with high consistency were observed in spectra corresponding to the measured spots, providing direct evidence for the formation of quasi-symmetric Ag/NDs. The intensity profile at 1078 cm⁻¹ was used to evaluate the reproducibility, and the RSD value was 10.5% [Fig. 3(d)]. Large-scale mapping was also carried out on chips from ten different batches (Fig. S10).

The sensing performance (i.e., sensitivity and LOD) of the Ag/ND platforms was estimated via a quantitative study using Raman probe dyes. 4-ATP solutions were prepared with various concentrations from 10⁻¹¹ to 10⁻⁴M. For the investigated concentration range, all the Ag/ND platforms enabled the acquisition of the spectral features of the 4-ATP molecule without substantial degradation or deformation [Fig. 4(a)]. We also observed that the spectral intensity increased with increasing number of molecules according to the Langmuir adsorption model. We extracted the variations in signal intensity at 1078 and 1178 cm⁻¹ to calculate the LOD [Fig. 4(b)]. Using a logarithmic fitting, we obtained a correlation coefficient (R²) of 0.97 at both 1078 and 1178 cm⁻¹, indicating a highly linear relationship between the optical signals and molecular concentration. The LOD was also calculated using the three standard deviation equation of LOD = 3 × S_b/m, where S_b is the standard deviation of a blank (n = 3) and m is the slope of the calibration curve. Accordingly, the average LOD value of 4-ATP was obtained to be 3.2 × 10⁻¹²M. To explore the feasibility of multiplex detection, we used the Ag/ND platforms to trace methylene blue (MB) molecules in the concentration range from 10⁻⁸ to 10⁻⁴M [Fig. 4(c)]. The fingerprint spectra of MB were observed without interferences. For MB, herein, the peaks at 450 and 776 cm⁻¹, which correspond to in-plane C–N–C bending and C–H stretching, respectively, were selected as the analytic peaks.^42,43 In the log–log scale plot, the R² values of the MB–Ag/ND platforms were estimated to be 0.92 and 0.96 for the peaks at 450 and 776 cm⁻¹, respectively; consequently, the average LOD value was obtained as 1.8 × 10⁻⁹M [Fig. 4(d)]. These results show that the Ag/ND platforms demonstrate sensitive, reproducible, and reliable SERS activities.

In the present study, we developed a one-step roll-to-roll process to fabricate plasmonic sensing chips with high throughput for commercial applications. The integrated IBS and sputtering deposition systems enabled the fabrication of plasmonic Ag/NDs on A4-scale PEN films in series. The influence of the fabrication parameters (i.e., moving speed and Ag thickness) was experimentally and theoretically investigated. The optical behaviors were verified on the basis of electric fields calculated using the FDTD method. For 4-ATP molecules, the 150 nm Ag/ND 0.6 mpm platforms demonstrated picomolar sensitivity with excellent reproducibility (i.e., RSD of ∼10%). The SERS signals collected from different batches confirmed the superior repeatability of the developed system. Therefore, the proposed roll-to-roll process under a vacuum environment showed strong potential as an alternative to the current engineering techniques that involve multiple steps and apparatus.

4-ATP (C₆H₇NS) and MB (C₁₆H₁₈ClN₃S) were purchased from Sigma-Aldrich (St. Louis, USA). A 125 μm-thick PEN film was obtained from DuPont Teijin Films (Teonex Q65HA; Chester, USA). A 250 μm-thick PET film was purchased from Toray Advanced Materials Korea (Seoul, Korea). A 20-in. Ag (99.99%) sputtering target was acquired from Taewon Scientific (Seoul, Korea).

The Ag/NDs were defined on an A4-scale PEN film using a vacuum roll-to-roll system (Advanced Technologies) with incorporated IBS and sputtering deposition systems. The PEN film was loaded onto the carrier film (herein, PET) in the chamber, and the base pressure was reduced to 1 × 10⁻⁵ Torr. The IBS treatment was carried out under operating conditions that included an accelerating voltage of 1.2 kV, an ion beam current of 120 mA, and an O₂ flow of 56 sccm. Ag was subsequently deposited onto the as-prepared PEN NDs at a radio frequency power of 1 kW and an Ar flow of 200 sccm. After the process, the A4-scale film was cut into individual plasmonic chips with dimensions of 5 × 5 mm².

The optical performance of the Ag/ND platforms was estimated using a Raman spectroscope (NS200; Nanoscope Systems, Daejeon, Korea). SERS signals with an acquisition time of 2 s were collected from samples under a laser exposure with a wavelength and an optical power of 785 nm and 6 mW, respectively. For the evaluation of SERS activities, two different Raman probe dyes (i.e., 4-ATP and MB) dissolved in ethanol were prepared with different concentrations (tenfold dilution). The Ag/ND platforms were immersed into individual solutions for molecular adsorption. After 2 h, the chips were rinsed with ethanol and then dried under the ambient environment.

The electric fields distributed over the Ag/ND platforms were theoretically calculated on the basis of the FDTD method using a commercial program (Ansys Lumerical 2021 R1.2). To design the simulation model, the geometric parameters (i.e., radius and density) of plasmonic nanostructures were analyzed from the SEM images using the scanning probe image processing software (SPIP; Image Metrology). Simplified Ag tips with a hexagonal arrangement were obtained. The radii of the Ag tips and their distribution were determined under the assumption of a Gaussian distribution and using the Box–Muller transform. The supporting PEN ND was designed on the basis of STEM images. For the computations, the refractive index of the background medium (i.e., air) was set to 1, whereas that of Ag was expressed as a = −19.0051 and b = 2.254 27 (at 785 nm). A 785 nm x-polarized plane wave was incident in the normal direction, and the mesh size was determined to be 1 nm.

The morphological properties were observed using FE-SEM (JSM-7800F, JEOL) and STEM (JEM-ARM200F, Jeol, Tokyo, Japan) analyses. The plasmonic resonance properties were measured by large-area dark-field microscopy (Eclipse LV100ND, Nikon, Japan).

See the supplementary material for additional SEM, TEM, FDTD, and Raman mapping data.

This study was financially supported by the National Research and Development Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Science and ICT (Grant Nos. NRF-2021R1A2C2011048, NRF-2021M3H4A1A02051036, and NRF-2021M3H4A4079520).

The authors have no conflicts to disclose.

Tae Eon Kim: Formal analysis (lead); Investigation (lead); Writing – original draft (lead). Sunghoon Jung: Investigation (supporting); Validation (supporting); Writing – review & editing (supporting). Soo Hyun Lee: Validation (supporting); Writing – review & editing (supporting). ChaeWon Mun: Methodology (supporting). Eun-Yeon Byeon: Methodology (supporting). Jun-Yeong Yang: Investigation (supporting). Jucheol Park: Investigation (supporting). Seunghun Lee: Investigation (supporting). Heemin Kang: Investigation (equal); Writing – review & editing (equal). Sung-Gyu Park: Conceptualization (equal); Funding acquisition (equal); Supervision (equal); Writing – review & editing (equal).

The data that support the findings of this study are available from the corresponding authors upon reasonable request.