Cicely Rathmell

Surface enhanced Raman spectroscopy (SERS), although popular in research, has not yet been adopted in industrial applications. This is largely because of the challenge of manufacturing cost-effective, reproducible SERS substrates in volume that can be read by a compact reader for practical deployment in the field. Here we demonstrate and characterize a new dual-nanostructured SERS substrate approach that could be amenable to large-scale production. This dual-nanostructured surface combines long-range ordering of silver nanowires (AgNWs) with dense non-ordered decoration via silver nanoparticles (AgNPs) to improve on the SERS response characteristics of the individual structures, using fabrication methods that could be scaled for deposition via roll-to-roll Mayer rod coating. Characterization of the substrates’ SERS response using 638 nm Raman spectroscopy and ATP as the reporter molecule found the dual-nanostructured SERS substrate to offer both greater spatial uniformity and a consistently higher SERS enhancement factor (EF) than the nanowires alone.

Surface-enhanced Raman spectroscopy (SERS) is a technique whereby the Raman effect is enhanced by a factor of 10

 or more through interaction of the analyte with noble metal nanostructures (1). This increases the limit of detection of Raman significantly for both quantitative and qualitative applications, and reduces performance demands on the associated instrumentation to allow greater deployment in the field. In SERS, the Raman signal enhancement originates from a combination of electromagnetic (EM) field enhancement because of the interaction of light with the nanostructures and the chemical enhancement (CE) that occurs when the analyte is chemisorbed on a SERS-active surface.

Common SERS-active structures include nanoparticles, nanowires, and nanostructured fabricated surfaces, with the greatest EM enhancement occurring in SERS hotspots, junctions between two nanostructures (1). Although both colloidal and solid SERS structures have been explored, solid substrates offer the greatest potential for commercial use, provided that they possess a sufficient number of SERS hotspots uniformly distributed within the spot size of the reader, in order to produce a reproducible, concentration-dependent signal while retaining high sensitivity. SERS substrates also allow convenient solution sampling, with the analyte becoming concentrated on the surface as the solvent evaporates to further increase interaction between the analyte and signal-enhancing substrate within the laser focus.

Both silver nanoparticles (AgNPs) and silver nanowires (AgNWs) are known to produce significant SERS signals, and can be synthesized relatively easily in the laboratory. However, the current fabrication methods utilize reaction conditions that pose challenges for large-scale production, from the need for meticulous controls over synthesis parameters to slow reaction times. This study demonstrates a SERS substrate where AgNWs and AgNPs are used as complementary nanostructures to improve on the SERS response characteristics of AgNWs alone, and where both structures are fabricated via easy, scalable synthetic methods.

For the AgNWs, a simple one-pot synthesis with a fast reaction rate and low temperature is used, based on the popular “polyol” method. The polyol method utilizes hot ethylene glycol (EG) to reduce silver ions in the presence of polyvinylpyrrolidone (PVP, Figure 1a). The method is popular due to the use of environmentally-friendly reactants and relatively mild reaction conditions compared to other thermal solvolysis methods. Here, we employ FeCl

 to catalyze the reduction of silver and create shorter, thicker wires (2), as well as halide ions to promote wire formation.

For the AgNPs, a silver ion molecular complex is used as an alternative to traditional AgNP fabrication, in which a solid silver surface is rapidly produced when a diamminesilver(I) complex is exposed to an aldehyde functional group. The silver complex used here is μ-oxalatobis(ethylenediamine)silver(I), shown in Figure 1b. By utilizing a carboxylic acid as a ligand, gaseous CO

 is produced upon decomposition, and the decomposition reaction is slower. Because the reaction produces all gaseous organic species, a clean silver surface is produced absent of background interference from the capping agents often used to ensure colloidal stability (3), or additives required to avoid the coffee stain effect common in SERS substrates (uneven distribution of AgNPs resulting in concentrated rings on the surface) (4). Though the silver complex alone has previously failed to yield the morphology required for SERS response (5), the consistent formation of nanoparticles and known shelf stability (6) makes it a promising choice for incorporation with AgNWs into a dual-nanostructured SERS substrate.

FIGURE 1: (a) The structure of polyvinylpyrrolidone, used with hot ethylene glycol EG) to reduce Ag

 to form silver nanowires; (b) The structure of the silver complex μ-oxalatobis (ethylenediamine)silver(I) used to synthesize silver nanoparticles; (c), Schematic of the proposed dual-nanostructured substrate, composed of AgNWs decorated with AgNPs.

Additionally, by drawing on established industrial processing techniques like use of a whirlpool tank to isolate the AgNWs and roll-to-roll Mayer rod coating to deposit the SERS-active materials, a pathway is proposed for a SERS substrate which could be both manufacturable and cost-effective in volume.

Both AgNW substrates and AgNW substrates coated with AgNPs were fabricated using a variety of reaction parameters and deposition methods. These were evaluated to explore whether the easily produced AgNW coating could be utilized as a substrate to produce a uniform AgNP surface through the decomposition of an aqueous ink of the Ag complex on the AgNW substrate without the need of any additives, thus creating a dual-nanostructured surface. Results presented herein represent the optimized conditions relevant to demonstrating the dual-nanostructured surface; full details of the optimization process are presented elsewhere (7).

A synthetic method was developed to produce AgNWs based on the “polyol” method, incorporating a transition metal cation catalyst, and utilizing a selective precipitation purification method.

To synthesize the AgNWs, PVP (~150 mg) with a molecular weight of ~55,000 was dissolved in EG (20 mL ) with vigorous stirring, protected from light exposure. Silver nitrate was dissolved directly into the same reaction vessel, with a mole ratio of PVP monomer to silver of 2.0 to help direct one-dimensional growth of nanowires. An 11 mM stock solution of FeCl3 in EG (100 μL) was added with gentle swirling. The solution was heated to 100 °C in an oven for 45 min before increasing the temperature to 140 °C for 70 min, then cooled to room temperature.

The reaction products were dissolved in water (40 mL ) to create a homogenous dispersion. Acetone was then added slowly, dropwise with swirling, until a visible granularity was observed, and the mixture was allowed to settle for 15 min. The supernatant was removed, and water (~20 mL ) was added to resuspend the AgNWs. The selective precipitation was repeated 3–5 times, as necessary. Residual acetone and water were removed by centrifuging the solution at 3000 rpm for 3 min, removing the supernatant, redispersing in ethanol, and repeating the procedure four times before suspending the purified wires in a stock solution of ethanol (~5 mL ).

Qualitative analysis of the wire size distribution and purity was determined by bright- field optical microscopy using an Olympus BX60 microscope equipped with a DP73 color camera and polarization filters. SEM images were taken using a Hitachi SU8000 UHR FE-SEM. Composition and structure were characterized using UV-vis spectroscopy, thermal gravimetric analysis (TGA), and scanning electron microscopy (SEM), reported elsewhere (7).

The silver complex μ-oxalato-bis (ethylenediamine) silver(I) was synthesized as reported in literature (6). First, 0.5 M aqueous solutions of silver nitrate and oxalic acid were combined in an Ag

:oxalate volume ratio of 5:3. The silver oxalate precipitate was collected by filtration, washed with water, and dried at room temperature under vacuum. The silver oxalate was then dissolved in small increments, while stirring, in an aqueous solution of ethylene diamine (20% v/v), followed by precipitation with an excess of isopropanol, and subsequently filtered. The white precipitate was washed with cold isopropanol and vacuum dried at room temperature for 6 h prior to storage at -20 °C. The resulting silver complex was prepared as an ink for substrate fabrication by dissolving the silver complex in water (11 mg/mL). Care was taken to avoid light exposure during all synthetic steps and storage. The structure of the silver complex was confirmed via infrared (IR) and 

C nuclear magnetic resonance (NMR) spectroscopy, as well as TGA reported elsewhere (7).

Nanowire and Dual-Nanostructure Substrate Fabrication

Mayer rod coating was used to produce AgNW surfaces for direct use as a SERS substrate, herein referred to as a nanowire substrate (NWS), and for the fabrication of dual-nanostructured surfaces (DNS). To produce a NWS, the stock AgNW solution was pipetted (25 μL) in a line on glass slides and spread by hand using a #10 Mayer rod. The slides were dried at room temperature for ~30 min.

SERS quality DNS were produced by first preheating a NWS slide to 125 °C, then 25 μL of the Ag complex ink was drop-cast in the center of one half of the slide, and allowed to decompose for 5 min before cooling to room temperature, when it was then gently rinsed with ethanol and allowed to dry.

The uniformity of the fabricated surfaces was examined using an optical microscope. The morphology of the nanostructures on the surfaces were characterized using a Veeco Dimension 3100 AFM, operating in tapping mode.

SERS Characterization and Instrumentation

Aminothiolphenol (ATP) was used as the SERS reporter molecule for characterization of the substrates, drop-casted in small aliquots onto both the NWS and DNS sides of the SERS substrates, and air dried at room temperature for 2 min. Raman spectra were acquired with a fully integrated Wasatch Photonics 638 nm Raman spectroscopy system, utilizing a backscattering sampling geometry and spot size of ~60 μm, with detection via a fixed holographic transmission grating and a thermal electric-cooled (TEC) charge-coupled device (CCD) detector. SERS measurements were acquired using a laser power of 5 mW, taking 1 s per scan. Slide positioning was achieved using a Thorlabs MT3 XYZ stage controller equipped with micrometers to control translation (XY) and focus (Z) of the laser on the slide surfaces.

Nanowire synthesis produced a narrow distribution of diameters of 80–120 nm, with lengths of 10–30 μm, as verified by optical microscopy, yielding an aspect ratio of ~200.

Notably, the reaction conditions employed required no control over stirring, injection rate, or other factors that would be difficult or expensive to incorporate into industrial scale reactions. The mechanical strength of the ~100 nm wires is enough to remain mostly straight with only a small amount of curvature, which bodes well for processability of the wires for substrate manufacture. The selective precipitation successfully isolates the desired AgNWs in a high purity and is readily adaptable to large scale execution in a whirlpool tank.

Deposition of AgNWs onto a slide via Mayer rod coating resulted in uniform coating and areas of linear structure with a high density of wire intersections to generate SERS-active hotspots, as imaged using microscopy (Figure 2a). To assess the uniformity of the SERS response in relation to spatial variations in sampling, SERS spectra of a 10 μL sample of 0.5 μM ATP solution were taken every 500 μm along a 1 cm linear path on the NWS, averaging four scans per location (Figure 2b). The SERS response of the NWS was found to be highly dependent on the position sampled, dropping to zero at some locations. The inconsistent spatial response of the NWS is a consequence of the AgNWs’ length being comparable to the spot size of the laser, which limits the number of SERS hotspots, as well as to gaps in deposition, that could be significantly improved in an industrial setting through the use of process automation.

FIGURE 2: (a) A microscope image of a typical nanowire substrate (NWS) produced by Mayer rod coating, showing both linear regions and gaps. (b) The SERS spectra acquired on the NWS over a spatial distance of 1 cm.

Because the AgNWs retain a surface layer of PVP as a byproduct of synthesis, it was hypothesized that the NWS could provide a guiding template for the deposition of a silver complex to populate the NWS with AgNPs in the interstitial space between and on the AgNWs, thus creating a dual-nanostructured surface with improved SERS characteristics, without the use of additives.

The NWS was then utilized as a substrate to produce a uniform AgNP surface via dropcasting and subsequent thermal decomposition of the aqueous ink of the Ag complex as the solvent evaporated. The outer edge of the deposited ink spot experienced a concentration of silver deposition due to the coffee stain effect, and thus all characterization was performed on the central area. While dropcasting was chosen to optimize SERS response for proof of concept, a pathway exists to adapt this step to Mayer rod coating for volume production (7).

Optical microscopy showed craters interspersed with the AgNWs due to the elimination of the gaseous decomposition products (Figure 3a), while AFM revealed clusters of AgNPs at wire intersections, confirming the dual-nanostructured surface (Figure 3a, inset). The particles exhibit extremely small sizes and exist on a large range of Z heights. The same spatial uniformity assessment was applied to the DNS substrate, sampling SERS spectra of ATP every 500 μm along a 1 cm linear path for comparison to the NWS, averaging four scans per location (Figure 3b). The SERS response of the DNS retains some spatial fluctuations in the intensity of the acquired spectra, as expected due to variation in the underlying AgNW layer, but provides a minimum SERS response at every sampled location, as well as a consistently higher SERS response overall when compared to the NWS.

FIGURE 3: (a) A microscope image of the dual-nanostructured substrate (DNS) produced by drop casting the silver complex ink onto an NWS, with tapping mode AFM image (inset) showing nanoparticles filling the interstitial spaces between nanowires. (b) The SERS spectra acquired on the DNS over a spatial distance of 1 cm.

Augmenting the NWS with dense, nonordered AgNPs produced from a silver complex with good shelf stability (Figure 1b) resulted in an increase in the number of SERS hotspots. This approach improved the magnitude of the SERS response, while decreasing the spatial variability of the signal obtained from the AgNWs alone.

To quantify and compare enhancement factors (EF), the SERS spectra of a 5 μL sample of 423.5 μM ATP solution were taken at multiple locations of both a NWS and DNS fabricated on the same slide. Due to the spatial variability in the NWS, a total of 15 positions with local SERS signal max- ima were sampled, a process which took roughly 30 s per location. No optimization was needed for the DNS, and thus 10 random locations were sampled twice, at both 

 s, to capture the initial response, and to approximate the laser exposure time seen in the NWS sampling. Spectra are shown in Figure 4.

FIGURE 4: The SERS spectra of ATP for the nanowire substrate sampled at 15 locations and averaged (NWS), and for the dual-nanostructured substrate sampled at 10 locations, both initially (DNS-1) and after 30 sec (DNS-30).

 are due to vibrations of the aromatic ring of ATP, with contribution from the dimerization of ATP to 4,4’-dimercaptoazobenzene (DMAB) at 1084 cm

 (8). The effects of ATP concentration and laser exposure on dimerization for these substrates were explored in more detail elsewhere (7). The calculated EFs for these peaks are shown in Table I. Peaks at 1150 cm

 were also observed, largely attributed to the selective enhancement of the 

 vibrational modes of ATP during SERS analysis (9,10).

Although the contribution from DMAB to the spectrum makes it difficult to reliably quantify the increase in EF for the DNS vs. the NWS (7), all substrates demonstrate an EF of 105 or greater, with the dual-nanostructured substrate showing a clear improvement in EF over the nanowire substrate.

A nanowire substrate utilizing AgNWs and a dual-nanostructured SERS surface that also incorporated AgNPs were fabricated using synthesis methods and production processes amenable to cost-effective volume production. The AgNWs synthesized were of high purity and narrow size distribution, ~10 μm in length and ~100 nm in diameter, and were deposited using Mayer rod coating to form a SERS-active substrate. Although high spatial variation in the number of SERS hotspots relative to the spot size of the laser limits its applicability as a quantitative SERS device, the AgNWs were shown to be an effective substrate to facilitate subsequent dropcasting with an aqueous solution of a silver complex to populate the interstitial spaces between the AgNWs and improve the SERS response using an additive-free process.

The resulting DNS exhibited both improved spatial uniformity in the SERS response compared to the NWS alone, as well as a significant increase in intensity. Through further process development and adaptation of AgNP deposition to a Mayer rod coating method, the dual-nanostructured SERS approach offers a pathway to industrial fabrication of a low-cost, SERS-active substrate.

(1) C.L. Haynes, A.D. McFarland, and R.P. Van Duyne, 

(2) K. Zhan, R. Su, S. Bai, Z. Yu, N. Cheng, C. Wang, S. Xu, W. Liu, S. Guo, and X.-Z. Zhao, 

(4) A.T. Fafarman, S.-H. Hong, S.J. Oh, H. Caglayan, X. Ye, B.T. Diroll, N. Engheta, C.B. Murray, and C.R. Kagan, 

(5) X. Liu, C. Zong, K. Ai, W. He, and L. Lu, 

(6) K.R. Zope, D. Cormier, and S.A. Williams, 

(7) K.B. Castro, M.Sc. dissertation, Old Dominion University, Norfolk, Virginia (2019). 

(8) Y.-F. Huang, H.-P. Zhu, G.-K. Liu, D.-Y. Wu, B. Ren, and Z.-Q. Tian, 

(9) K. Uetsuki, P. Verma, T.-a. Yano, Y. Saito, T. Ichimura, and S. Kawata, 

(10) Y.-C. Chen, J.-H. Hsu, and Y.-K. Hsu, 

 are with the Department of Chemistry at Old Dominion University, in Norfolk, Virginia. 

 is Vice President of Marketing at Wasatch Photonics, in Morrisville, North Carolina. Direct correspondence to: 

Articles

SERS can amplify Raman signals, but to make the technique practical for industrial use, large quantities of substrate are needed. The approach described here could enable cost-effective, reproducible manufacturing of SERS substrates at large scale.

A Dual Nanostructured Approach to SERS Substrates Amenable to Large-Scale Production

Microplastic particles in aqueous ecosystems constitute a worldwide environmental problem. Raman spectroscopy is uniquely positioned to detect and classify even the smallest, and most abundant, particle sizes, down to micrometers in size. However, the current “filter and scan” method used to concentrate and detect dilute microplastics in water is slow, and prone to contamination from handling and interference from the filter substrate. Here, we present a simpler approach to concentrate the particles in situ, without any sample preparation, in the focus of a Raman probe. The technique uses an ultrasonic resonator with a 2 MHz standing acoustic wave to capture 3.4 µm poly(methyl methacrylate) (PMMA) microspheres in the sample solution for direct detection. When tested with a 90 ppm solution, ultrasonic capture for 5 min enabled a clear PMMA Raman signal to be distinguished from the background. The signal enhancement provided by ultrasonic capture was estimated to be greater than 1500x.

	
	The majority of plastics, like the 78 million tons of plastic packaging manufactured every year, is produced for single-use items, approximately a third of which ends up in the ocean (at the rate of one garbage truck every minute). Engulfed by these vast waters, it gradually breaks apart into progressively smaller particles, and eventually enters the food chain as microplastics, which are defined as plastic particles smaller than 5 mm in size. Slowly, and unbeknownst to the world until recently, microplastic particles have spread to every continent, ocean (1–3), the air (4), and most food and beverages (5–8).

	Studies into the nature and distribution of microplastics across the world have found both bottled and ocean water to contain concentrations on the order of 100 particles per liter of microplastics 5 µm or larger in size, and several 1000s of microplastic particles greater than or equal to 1 µm in size. The most common microplastic particles found are polypropylene (PP), polyethylene (PE), polyethylene terephthalate (PET) in drinking water, and PE, PP, and polystyrene (PS) in the ocean.

	Currently, the most common approach to detecting microplastics requires filtering large sample volumes of water to concentrate particles for detection, then scanning the filter substrate for any particles (9), and chemically identifying the particles with Fourier-transform infrared spectroscopy (FT-IR) or Raman spectroscopy (10–11). However, this approach is very time-consuming (12), and can be prone to contamination of the filter and to interference from the filter material itself (13). This approach has shown variable sensitivity to the most abundant, but smallest, particles (14). Furthermore, the lack of a unified protocol for filtering and detection leads to inconsistent results, and thus compromises the collection and comparison of quantitative data for global monitoring (15–19).

	Raman microscopy on filter substrates is currently the only tool capable of detecting and identifying the smallest, and most numerous, particles of 20 µm in size or less, as its short wavelength, tight laser focus, and ability to raster systematically across the sample is sensitive to even small particles at low density. However, Raman microscopy in the present form uses large bench top-sized instruments, scanning slowly across the membrane used for the filtration of a water sample. As one review put it: “There is an urgent need for a fast and easily implementable monitoring tool capable of detecting small microplastics (20).”

	Instead of filtration, a different approach to capture and bring these microparticles to the point of measurement is proposed here as a proof of concept investigation (21). It employs a strong ultrasonic standing wave of about 2 MHz frequency in a resonator composed of a piezo transducer on one end and a Raman probe on the other. Small particles are trapped in the nodes of the standing wave field, where they are held by acoustic radiation forces (22). If one of the nodal planes of the ultrasonic trap is colocated with the laser focus of a Raman immersion probe, the particles can immediately be studied in situ. A trap of this design can be employed directly in the sample stream as collected, bypassing all preparation steps. This reduces the analysis time from hours to minutes, prevents any contamination of the sample from handling, and eliminates the Raman background commonly seen when concentrating samples onto a filter in the currently accepted method.

	The Raman system consisted of a compact Wasatch Photonics WP 785 Raman spectrometer with an integrated 785-nm laser set to 350-mW, equipped with a temperature-regulated CCD detector, and a 50 µm slit leading to a resolution of 10 cm

. The spectrometer was chosen for its f/1.3 input aperture and transmissive optical design, as the application requires a high degree of sensitivity in a portable design to enable eventual deployment in the field. The laser and spectrometer were fiber-coupled to a Raman probe with a 1/2” immersion probe tip terminated by a sapphire ball lens with a very short working distance.

	The ultrasonic resonator used for particle capture, soniccatch by usePAT, has been described in detail elsewhere (23). Briefly, a piezo-electric transducer generates the outgoing ultrasonic wave with an approximate frequency of 2 MHz, while the Raman probe ball lens forms the reflector for the resonator at a distance of about 3 mm from the surface of the transducer. Superposition of the incoming and reflected waves generates a standing wave with a small number of nodes, as shown in Figure 1.

	The radiation forces imposed by the ultrasonic wave are predominantly dependent on the particle size, but also depend on their mass density and compressibility relative to the medium. The forces exerted are due to the scattering of the wave at the particles’ surfaces, and lead to agglomerations in the nodal or anti-nodal regions of the standing wave (23).

	The resonator assembly consisting of the ultrasonic transducer and the Raman probe was lowered directly into a stirred 90-ppm suspension of 3.4 µm poly(methyl methacrylate) (PMMA) microspheres in water, in which the microspheres had been dispersed with a drop of detergent. PMMA is a lightweight, transparent plastic most commonly known as acrylic or plexiglass, and is also used as microbeads in personal care products.

	The integration time of the spectrometer was set to 2.5 s, and 500 individual non-averaged spectra were continuously recorded for each experiment. Five repeat runs were performed, four in which ultrasonic capture was activated at the beginning of each run, and one static reference run (no capture).

	Comparing the Raman spectra with and without ultrasonic capture of the PMMA microspheres, as shown in the bottom panel of Figure 2, we identified three main contributions to the signal arising from the sample and setup. The first contribution comes from a series of bands at ~414 cm

, present in all spectra, which do not change with ultrasonic capture. These were attributed to the sapphire of the ball probe lens, an effect also reported in Raman imaging using sapphire fiber (24). The second contribution, present in all spectra with ultrasonic capture in the steady state, agrees well with a bulk PMMA signal shown for comparison in the top panel of Figure 2, with key peaks at ~596 cm

. The third contribution, a broader peak at roughly ~1640 cm

, was present without ultrasonic capture, but disappeared over time once capture was activated. This more elusive signal can be attributed to the front window of the transducer. As particles are trapped, the Raman laser no longer reaches the surface of the transducer, but instead gets scattered by the trapped particles, leading to the marked reduction in signal and to the enhancement of the signal of the particles of interest. A direct comparison of the Raman spectra with and without ultrasonic capture to bulk PMMA in Figure 2 indicates that trapping leads to a large enhancement of the PMMA signal. The difference is distinct, from no discernible signal without ultrasonic capture to a very large and clean signal after ultrasonic trapping is activated.

	To study the ultrasonic capture process in greater detail, we recorded the Raman signal as a time series after activating the ultrasonic transducer and were able to observe the continuous evolution of the PMMA trapping in time as illustrated in Figure 3. Here, we have bound the region associated with the main PMMA Raman lines using vertical lines just beyond 750 cm

. In this region, we subtracted the baseline to obtain a clearer illustration of the growth of the PMMA signal in time (Figure 4). The asymmetric least squares (ALS) method (25) (26) was used for baseline correction.

	With background subtraction applied, the PMMA signal was clearly seen to rise after an initial induction period with no discernible signal change, during which we saw no indication of any PMMA signal in the baseline to indicate ultrasonic capture. Once initiated, the signal grew and reached a maximum quickly.

	For a quantitative analysis, we separated the signal contributions to the spectra set using multivariate curve resolution (MCR). MCR is an iterative method that can be used to identify a potential separation of the spectra set into fixed component spectra and time-varying contributions for each component. This separation is “ill-defined,” and does not lead to a unique solution, that is, the results depend on the initial guess for the spectra. As we already expected three independent contributions (dip probe, PMMA, and transducer), we limited the decomposition to three components and seeded the analysis with the first three loadings from the principal component analysis (PCA) of the same spectra, setting those three component spectra as our initial guesses for the MCR spectra.

	In addition, we enforced additional non-negative constraints on both the component spectra and the component contributions of the separation, an approach known as the ALS (alternative least squares) variant of MCR (27). This matches the expectations of actual spectra and “concentrations”, each of which would be expected to contribute additively to the observed spectra.

	Figure 5 shows the component spectra resulting from the MCR-ALS analysis. We assigned components 1 and 3 to the Raman signal arising from the dip probe and the transducer, and component 2 to the Raman signal from the PMMA microspheres in solution. Comparison to the Raman spectrum of neat PMMA (Figure 2, top panel) shows very good agreement; the PMMA component spectrum shows little interference from the signal caused by the setup (sapphire ball lens or transducer material).

	The MCR-ALS result for the time-varying contribution of component 2, which we assigned to PMMA, can be interpreted as the time evolution of the PMMA signal, as shown in Figure 6 for four independent repeat runs of the same experiment. The figure presents a similar growth curve of the signal for each run, with an induction period of approximately 250 s, and a full, steady state PMMA signal after about 1000 s. The rapid signal evolution after the initial induction period may be due to a feedback effect in which captured PMMA particles slow down other particles near the trapping nodes of the standing wave.

	To determine the Raman signal enhancement for PMMA attributable to ultrasonic capture in the 90 ppm dispersion, we averaged the MCR-ALS result for the contribution of the component assigned to PMMA with and without ultrasonic trapping activated. The average for the signal with trapping is formed across all four runs for all measurements past the 1000 second mark (steady state capture), leading to an average signal of about 23,100 (arbitrary units).

	For the reference signal, we used a separate experiment with the same sample in the same setup, but without trapping activated, and determined the MCR-ALS signal contributions using the same decomposition as used in the four runs with trapping activated. Averaging over the entire run of 100 spectra, we found a reference signal of 15±20 (arbitrary units) with the uncertainty given by the standard deviation across the 100 spectra. Even at “one sigma,” this signal is not significantly different from zero, limiting the possibilities of standard approaches for calculation of a lower limit for the enhancement factor. Though the true magnitude of the enhancement factor for ultrasonic capture cannot be fully quantified, a minimum enhancement of 1500x can safely be reported using this reference signal value.

	Growing concern over the increasing presence of microplastics in our environment and their potential environmental impact has heightened the need for a simple, robust, and portable measurement method. Capture of PMMA microspheres with an ultrasonic standing wave for direct detection using a compact Raman system is one possible alternative to current filtration-based methods. Here, we demonstrated a more than 1500-fold enhancement of the Raman signal using ultrasonic capture and a dip probe from a 90 ppm suspension in water. The captured microplastic sample amount increased in time up to the capacity of the resonator trap, reaching the full enhancement in about 500 s. While still experimental, the method shows a possible route to microplastic measurements in water requiring little to no sample preparation, and which can be performed using compact, portable instrumentation in the field.

N. P. Ivleva, A.C. Wiesheu, and R. Niessner. 2017, 

L. Van Cauwenberghe, A. Vanreusel, J. Mees, and C.R. Janssen, 

E. Gaston, M. Woo, C. Steele, S. Sukumaran, and S. Anderson, 

D. Schymanski, C. Goldbeck, H.-U. Humpf, and P. Fürst, 

T.M. Karlssona, A.D. Vethaaka, B. Carney Almrothd , F. Ariesee, M. van Velzena, M.Hassellöv, and H.A. Leslie. 

B.E. Oßmann, G. Sarau, H. Holtmannspotter, M. Pischetsrieder, Silke H. Christiansen, and W. Dicke, 

A.C. Wiesheu, P.M. Anger, T. Baumann, R. Niessner, and N.P. Ivleva, 

G. Sarau, L. Kling, B.E. Ossmann, A-K. Unger, F. Vogler, and S.H. Christiansen, 

R.Lenz, K. Enders, C.A. Stedmon, D. M.A. Mackenzie, and T.G. Nielsen, 

L. Vethaak, G.H. Tinnevelt, J.J. Jansen. 

L. Frère, I. Paul-Pont, J. Moreau, P. Soudant, C. Lambert, A. Huvet, and E. Rinnert, 

B.E. Oßmann, G. Sarau, S.W. Schmitt, H. Holtmannspötter, S.H. Christiansen, and W. Dicke, 

A.M. Elert, R. Becker, E. Duemichen, P. Eisentraut, J. Falkenhagen, H. Sturm, and U. Braun, 

S. Primpke, R.K. Cross, S.M. Mintenig, M. Simon, A. Vianello, G. Gerdts, and J. Vollertsen. 

S. Primpke, S.H. Christiansen, W. Cowger, H. De Frond, A. Deshpande, M. Fischer, E.B. Holland, M. Meyns, B.A. O’Donnell, B.E. Ossmann, M. Pittroff, G. Sarau, B.M. Scholz-Böttcher, and K. J. Wiggin. 

C. Thaysen, K. Munno, L. Hermabessiere, C.M. Rochman. 

L. Mai, L-J. Bao, L. Shi, C.S. Wong, and E.Y. Zeng, 

Microplastics in Aquatic Systems – Monitoring Methods and Biological Consequences

 YOUMARES 8 -Oceans Across Boundaries: Learning from each other Proceedings of the 2017 conference for YOUng MARine RESearchers.

 C.F. Araujo, M.M. Nolasco, A.M.P. Ribeiro, and P.J.A. Ribeiro-Claro, 

 K. Wieland, S. Tauber, C. Gasser, L. A. Rettenbacher, L. Lux, S. Radel, and B. Lendl, 

 P.H.C. Eilers and H.F.M. Boelens, “Baseline Correction with Asymmetric Least Squares Smoothing,” Leiden University Medical Centre, Leiden, The Netherlands, 2005.

 M. Garrido, F.X. Rius, and M.S. Larrechi, 

are with usePAT GmbH, in Vienna, Austria. 

are with Wasatch Photonics, in Morrisville, North Carolina. Direct correspondence to: 

Of the 78 million tons of plastic packaging manufactured every year, approximately one-third ends up in the ocean, the air, and most foods and beverages. To monitor the proliferation of these plastics, an ultrasonic capture method is demonstrated that produces a 1500-fold enhancement of Raman signals of microplastics in water.

In Situ Enhancement of Microplastic Raman Signals in Water Using Ultrasonic Capture

Antibiotic resistance rates are rising for many common pathogens, putting added pressure on healthcare providers to rapidly, and accurately, diagnose the cause of bacterial infections. New diagnostic methods must be sensitive, selective, and capable of multiplexed detection, as well as rapid, dependable, and portable enough to operate in a variety of settings. A new application of surface-enhanced Raman spectroscopy (SERS) shows great promise to meet these criteria. In this novel assay, bacteria are captured and isolated using functionalized metal nanoparticles for rapid optical identification via SERS. Initial tests with a portable SERS system validated the ability to identify the presence of 

bacteria, quantifying concentrations down to 10 colony-forming units per mL (cfu/mL), with high reproducibility. The system was also able to discriminate between the bacteria within the same sample matrix at clinically relevant concentrations (10

cfu/mL), demonstrating its potential as a rapid, reliable on-site diagnostic for bacterial pathogens. 

	The proliferation of multidrug-resistant bacteria is a silent but growing crisis, acknowledged by public health authorities and regulatory agencies as an emergent global malady (1). Although resistance to antibiotic drugs may develop naturally via mutation, or by transfer of mobile genetic elements encoded with resistance genes, it is the selective pressure exerted through the widespread use, and misuse, of antimicrobial agents that causes these bacteria to selectively flourish, relative to weaker strains.

	Both the absolute number and proportion of antimicrobial-resistant pathogens have increased this century. Although the prevalence of methicillin-resistant 

(commonly known as MRSA) is stabilizing or decreasing in Europe (2), infections due to antimicrobial-resistant 

) are increasing across the continent. In the United States, 

have been identified as the most urgent threats in a list of 18 drug-resistant pathogens (3). The rise of multi-drug resistance in 

 is particularly concerning, as it is the most common Gram-negative pathogen in humans, causing a variety of maladies, ranging from general digestive upset to urinary tract infections and life-threatening bloodstream infections. Unlike MRSA, which is often acquired in a healthcare setting, most multidrug-resistant strains of 

originate in the general community. In one Icelandic study, increased prevalence of quinolone-resistant 

has even been traced back through the food chain to its presence in chicken feed (4).

	As the costs to our healthcare systems grow, and the points of origin increase, so does the need for rapid, specific, and sensitive pathogen detection. Given that infections due to antimicrobial-resistant pathogens may stem from the food chain, the environment, or the clinic, detection methods must be flexible, and portable enough to be deployed in a wide variety of field conditions and by nonexperts. The vast majority of technologies currently used for the task fail on these counts, due to long analysis times (most approaches require culture prior to further analysis to increase bacteria concentration), the need for specialized equipment or staff, high cost, or additional sample preparation steps; these include staining, culturing and biochemical assays, real-time polymerase chain reaction (RT-PCR), and microarray assays like enzyme-linked immunosorbent assay (ELISA). Lab-on-a-chip technology seeks to streamline sample handling and concentration steps to reduce human error, and improve accuracy in a portable format, but the technology still lacks the simplicity of our technique.

	Surface enhanced Raman spectroscopy (SERS) offers a rapid and portable alternative to the existing methods for bacterial detection. SERS leverages the specificity of Raman signals with their sharp, well-defined peaks to provide a fingerprint unique to each analyte, with enhanced sensitivity due to interaction with silver or gold nanostructures. When in close proximity, surface plasmon resonance in the noble metal generates an amplified electromagnetic field that enhances the Raman signal. If the laser frequency used is tuned to the absorbance maxima of the analyte, additional enhancement can be observed, an effect known as 

surface enhanced resonance Raman scattering

 (SERRS). The combined enhancement can yield up to ~10

in total signal amplification. Using silver-coated gold nanorods, SERRS has been used by Wang and associates to differentiate between strains of carbapenem-resistant 

with almost 100% accuracy, using orthogonal partial least squares (OPLS) discriminant analysis of their spectra (5).

	SERS using functionalized nanoparticles (NPs) tagged with resonant Raman reporter molecules has proven ideal for multiplexed detection of bacteria, offering excellent selectivity, high photostability, and lack of spectral crosstalk. The addition of a "magnetic separation" step to isolate bacteria from the host matrix has been shown to improve detection limits to within clinically relevant levels (10

 colony-forming units [cfu]). This approach enabled Guven and colleagues to report detection limits of eight cfu/mL in water for 

cfu/mL in apple juice (7). Wang and colleagues were able to simultaneously detect 

 in spiked spinach solution and peanut butter at concentrations of 10

cfu/mL, demonstrating the potential of the approach in complex matrices (8). Although these studies meet the threshold for speed and minimum detection limit, they leave opportunities for further improvement in sensitivity and ease of sample preparation.

	To this end, we developed a highly sensitive, efficient, and easy-to-use sandwich SERS assay for the isolation and multiplexed detection of bacteria (Figure 1). Lectin- functionalized magnetic nanoparticles (MNPs) are used to capture and isolate bacteria within the sample, while SERS-active silver nanoparticles (AgNPs), functionalized with bacteria-specific antibodies and Raman reporter molecules, bind to each specific strain for interrogation.


	Figure 1: Schematic illustrating the sandwich SERS assay used for: (a) single-pathogen and (b) multiplexed pathogen detection. (c) Lectin-functionalized silver-coated magnetic nanoparticles facilitate magnetic separation of all bacterial strains from the sample matrix, while SERS active conjugates composed of silver nanoparticles functionalized with strain-specific antibodies and unique Raman reporter molecules facilitate optical detection of each pathogen via SERS.

	Lectins were chosen as the functional group for the magnetic nanoparticles, because of their affinity for the sugars expressed on the surface of bacteria; the MNPs then act as a handle by which magnetic separation can be used to capture bacteria indiscriminately from the sample matrix. This approach allows any unbound material to be washed away, concentrating the sample for increased signal, less background, and elimination of false positives.

	SERS-active conjugates facilitate optical detection of the bacterial pathogens. Each conjugate type consists of silver nanoparticles functionalized with strain-specific bacterial antibodies and a unique Raman reporter molecule, providing each bacterial strain under interrogation with a bespoke optical fingerprint when investigated using SERS. The small size of the AgNPs, relative to the bacteria surface, permits multiple binding sites, allowing the AgNPs to come in close proximity to one another and form SERS hot spots for increased Raman signal. Unbound AgNPs are washed away during the concentration stage, further benefiting signal to noise of the detection technique.

	The use of lectin-functionalized MNPs makes this separation step novel and efficient in capturing and isolating bacteria from a sample matrix, while concurrent addition of bacteria-specific SERS conjugates simplifies and speeds sample preparation. Previous work by our group has shown this detection assay to be effective in isolating and detecting 

 cfu/mL individually, using a benchtop Raman microscope system for SERS detection. We have also demonstrated triplex detection, successfully identifying each bacterial strain within the same sample matrix (9). The present study establishes transferability of the technique to a more compact, cost-effective, and portable system, without compromise to detection limit or multiplexing ability, greatly expanding its potential for practical use.

	Synthesis of silver nanoparticles and silver coated magnetic nanoparticles, and the preparation of the biomolecule-AgNP conjugates that were used in these studies have previously been reported by our group (9). Detailed protocols for the preparation and characterization of these colloidal suspensions are provided in the supporting information (ESI) of this paper, and previously published in the journal 

 (methicillin resistant) ATCC BAA-1766 were used in this study. 

and MRSA were grown on Luria-Bertani (LB) Miller agar in an aerobic atmosphere for 24 h at 37 

C. Harvested cells were suspended in LB broth to obtain an optical (OD

) of 0.6. Cells were then plated onto LB agar plates as triplicates, and incubated for 24 hours, as described previously. All the strains were grown using the same batch of culturing plates to reduce any potential unwanted phenotypic variation.

	Bacterial slurries were prepared by harvesting the biomass from the surface of each plate using sterile inoculating loops, and resuspending in physiological saline solution (1 mL; 0.9% NaCl). The prepared bacterial slurries were washed by centrifugation (

650 g for 5 min; MRSA, 1600 g for 5 min), and the pellets resuspended in 1 mL saline. The wash step was repeated a further twice with the final re-suspension being in deionized water (1 mL; d.H

was recorded for all samples, and bacterial concentrations were obtained by serial dilutions based on plate-counting results. All samples were stored at -80 °C until further analysis.

	The novel method for bacterial detection described previously by Kearns and colleagues (9) was modified slightly for use in this study.

	For each of the bacterial pathogens, the following procedure was carried out: SERS active antibody functionalized silver nanoparticles (200 µL; Ab-AgNP) were added together with lectin functionalized silver coated magnetic nanoparticles (200 µL; Con A-Ag@MNP), and a specific bacterial strain (50 µL, 10

 cfu/mL). Note for the control sample, the bacteria was replaced with d.H

O (50 µL) only. The sample was mixed thoroughly for 30 min before being placed in a magnetic rack for a further 30 min to allow the sample to collect. The clear supernatant was removed and the sample re-suspended in d.H

O (600 µL) ready for analysis. The same procedure was followed for the concentration study of 

except the bacteria concentration was varied from 10

 cfu/mL. It should be noted that each sample was prepared in triplicate.

	The two sets of antibody functionalized silver nanoparticles (100 µL of each conjugate) were added together with lectin-functionalized silver-coated magnetic nanoparticles (200 µL); plus the two bacterial strains (50 µL of each pathogen; 10

O (100 µL) only. The same procedure as described above (for the single pathogen detection) was employed.

	Raman reporters used were malachite green isothiocyanate (MGITC) and 4-(1H-pyrazol-4-yl)-pyridine (PPY) and these were used to detect 

	The samples were analyzed immediately after preparation using a WP 532 compact Raman spectrometer (50 µm slit, temperature-regulated detector set at 15 °C) and fiber optic probe (Wasatch Photonics, Morrisville, North Carolina), with 532 nm laser excitation (InPhotonics, Norwood, Massachusetts). The probe was focused directly into a 1 mL glass vial containing 600 µL of the bacteria–NP conjugate solution. All the measurements had a 1 s acquisition time, and a laser power operating at 7.7 mW. For all Raman spectra, data handling was carried out in Excel software, and no preprocessing was required. The peak intensities were obtained by scanning three replicate samples five times, and in all plots the error bars represent one standard deviation. The peak heights were calculated by averaging the peak intensities acquired from the 15 scans, and then subtracting the maximum intensity at 1620 cm

 (MRSA) from the base of the peak. Following this, the signal enhancements were calculated by dividing the peak heights of the sample by the peak height of the control. Furthermore, the relative standard deviations were calculated (% RSD) by dividing the average standard deviation by the mean peak intensity, and multiplying by 100.

	All multivariate statistical analysis was carried out in MATLAB software (The MathWorks, Natick, Massachusetts). Prior to conducting principal component analysis (PCA), the data was scaled. PCA was performed on three different data sets, two consisting of spectra obtained from single pathogen detection experiments, and one data set obtained for the multiplex detection of the two pathogens (10).

	To validate the SERS assay for simplex detection of each individual bacterial pathogen, we compared spectra of bacterial samples treated with the 

assay (SERS active conjugates + MNPs) to control samples. The total analysis time was ~1 h. The distinct Raman spectrum seen for 10

samples treated with the assay is due to the Raman reporter MGITC (selected for use with 

 antibodies), and thus confirms the presence of 

 in the sample (Figure 2a). In contrast, the control (water + assay) showed only very weak signal attributable to unbound residual SERS conjugates not washed away during the magnetic separation step. Comparing the characteristic MGITC peak at 1620 cm

, a signal enhancement of 14.5 times was observed for the 

 bacteria over the control (Figure 2b), matching the 10 times enhancement observed previously using a Raman microscope for detection (9).


	Figure 2: SERS spectra obtained from single pathogen detection using the sandwich SERS assay. (a) SERS spectrum of MGITC observed when detecting E. coli (green) and the control spectrum representing when no bacteria was present (red); (b) SERS peak intensities at 1620 cm

 for assay and control when detecting E. coli; (c) SERS spectrum of PPY observed when detecting MRSA (blue) and the control spectrum (red); (d) SERS peak intensities at 960 cm

 for assay and control when detecting MRSA. Raman intensity is given in arbitrary units (a.u.).

 cfu/mL MRSA samples treated with the MRSA assay containing the Raman reporter PPY clearly confirmed presence of MRSA in the sample versus the control (Figure 2c). Examining the characteristic PPY peak at 960 cm

, a signal enhancement of 12.5 times was observed for the MRSA bacteria over the control (Figure 2d), matching the 11 times enhancement observed previously using a Raman microscope. Transfer of the method to detection using a portable Raman system did not compromise the level of discrimination which could be achieved, and may have increased it somewhat. We believe the sampling geometry with the portable system may be more conducive to this type of assay, because it accessed a larger optical area and therefore more conjugates. Further work will directly compare the two detection methods using the same sample set.

	Having established the ability of our portable bionanosensor system to clearly identify pathogens within a matrix, we characterized the detection limit for 

 cfu/mL. Spectra for concentrations up to 10

 cfu/mL showed a monotonic increase in signal, with clear discrimination at all concentrations versus the control (Figure 3a). Comparing signal intensity for the characteristic MGITC peak at 1620 cm

 cfu/mL, at which concentration a signal enhancement of 3 times over the control was observed (Figure 3b). Above 10

cfu/mL, the signal decreased; we believe it was dampened due to self-absorbance of Raman light when large numbers of conjugated bacteria are present. If a high concentration of bacteria is present, the sample could be diluted and reanalyzed for accurate signal response.


	Figure 3: Concentration series for E. coli showing: (a) Comparison of SERS spectra for concentrations 10

 cfu/mL versus the control, and (b) peak intensity at 1620 cm-

 as a function of concentration. The dashed line gives visual clarification of the SERS peak intensity of the control. Raman intensity is given in arbitrary units (a.u.).

	Comparing the relative standard deviation (% RSD) of the portable system to results from the earlier microscope system study, the portable system showed much lower variance at all concentrations (Table I). This result indicates improvement in reproducibility, which may be due to the larger sampling area and increased laser power, and should be explored in further work.

	For a bionanosensor to be of practical use in healthcare, food safety, or environmental testing, it must be capable of multiplexed pathogen detection, both Gram-negative and Gram-positive. To this end, we tested the performance of our assay when both 

(Gram-negative) and MRSA (Gram-positive) were present at a clinically relevant concentration, 10

 cfu/mL. SERS conjugates for both pathogens were included in the assay. The resulting duplex sample spectrum shared features of both the 

and MRSA reference spectra, and clear discrimination versus the control (Figure 4a). Given that the peak previously used to identify 

 overlaps a SERS peak for the MRSA Raman reporter (1620 cm

, shown with a green star), we instead used the peak at 1180 cm

. Both peaks were clearly present in the duplex sample. Furthermore, comparison of the duplex sample to the control using the 1620 cm

 peak present in both Raman reporters yielded a signal enhancement of 8.4 times, clearly demonstrating a unique biorecognition interaction between the SERS conjugates and the two bacterial strains (Figure 4b).


	Figure 4: Duplex pathogen detection using the SERS assay for E. coli and MRSA at 10

 cfu/mL. (a) Stacked SERS spectra showing the spectra obtained from the detection of both bacterial pathogens simultaneously (black) versus each pathogen separately and a control sample (red). MGITC spectrum (green) represents E. coli and PPY spectrum (blue) represents MRSA. The black dotted lines show peaks unique to each Raman reporter, used to identify the presence of the bacterial targets. (b) Comparative peak intensities at 1620 cm

 for assay and control. (c) PCA scores plot showing the relationship between the multiplex spectra (black) and the E. coli (green) and MRSA (blue) single pathogen spectra. Raman intensity is given in arbitrary units (a.u.).

	To confirm detection of each pathogen, principal component analysis (PCA) was performed on the multiplex spectrum and two single pathogen spectra (Figure 4c). The tight clustering observed for the fifteen scans taken for each individual pathogen, as well as the duplex sample, shows the excellent reproducibility and discrimination in the identification of pathogens that is possible with a portable system.

	A compact SERS system providing all the requirements for a field-deployable assay (sensitivity, specificity, reproducibility, cost, portability, ease of use, and speed of analysis) has been demonstrated. This novel approach employs a sandwich-type nanoparticle assay in which magnetic separation is used to capture SERS active bacteria complexes for isolation and concentration from the sample matrix. Using this detection assay, cell concentrations as low as 10 cfu/mL were readily detected in single pathogen tests for 

and MRSA using a compact, portable Raman detection system. Duplex detection clearly and simultaneously identified the presence of both pathogens, with discrimination being made using both PCA and SERS analysis.

	The ability of this portable bionanosensor to detect bacterial concentrations below the requirements for clinical diagnosis, and to discriminate between bacterial pathogens within a duplex system in ~1 h with minimal sample preparation, demonstrates its potential as a field-deployable technique for use in healthcare, food, and environmental testing. As sample preparation time and system size are further reduced, its potential to provide rapid, sensitive detection in a field-deployable format will surely play a role in global efforts to curb the spread of antimicrobial-resistant pathogens.

	(1) I. Roca, M. Akova, F. Baquero, J. Carlet, M. Cavaleri, S. Coenen, J. Cohen, D. Findlay, I. Gyssens, O.E. Heure, G. Kahlmeter, H. Kruse, R. Laxminarayan, E. Liébana, L. López-Cerero, A. MacGowan, M. Martins, J. Rodríguez-Bañ o, J.-M. Rolain, C. Segovia, B. Sigauque, E. Taconelli, E. Wellington, and J. Vila, 

	(2) European Antimicrobial Resistance Surveillance Network, 

Surveillance of Antimicrobial Resistance in Europe (

A new application of surface-enhanced Raman spectroscopy (SERS) is described for quantifying low concentrations of pathogens with high reproducibility. In this novel assay, bacteria are captured and isolated using functionalized metal nanoparticles for rapid optical identification via SERS. Initial tests with a portable SERS system validated the ability to identify the presence of Escherichia coli and methicillin-resistant Staphylococcus aureus bacteria.

Cicely Rathmell

Articles

A Dual Nanostructured Approach to SERS Substrates Amenable to Large-Scale Production

In Situ Enhancement of Microplastic Raman Signals in Water Using Ultrasonic Capture

Rapid, Portable Pathogen Detection with Multiplexed SERS-based Nanosensors

Data-Driven Raman Spectroscopy in Oil and Gas: Rapid Online Analysis of Complex Gas Mixtures

Stray Light in Absorbance Measurements

SERS for Label-Free Biosensing

Raman Spectroscopy as a Quantitative Tool for Industry