Statistical Correlation Between SERS Intensity and NanoparticleCluster SizeConor P. Shaw,† Meikun Fan,‡,§ Chelsey Lane,‡ Garrett Barry,‡ Andrew I. Jirasek,*,†

and Alexandre G. Brolo*,‡

†Department of Physics and Astronomy, University of Victoria, Victoria, British Columbia, V8P 5C2 Canada‡Department of Chemistry, University of Victoria, P.O. Box 3065, Victoria, British Columbia, V8W 3V6 Canada§Faculty of Geosciences and Environmental Engineering, Southwest Jiaotong University, 610031 China

*S Supporting Information

ABSTRACT: Surface enhanced Raman scattering (SERS) mapping ofbiomarkers has shown great promise in determining the distribution ofproteins of interest in cells and tissues. Metallic nanoparticle (NP) probesare generally used in such mapping. Since SERS intensities from NPs aredependent on size, shape, and interparticle distance/distribution, it isunclear if this method can provide reliable biomarker quantification. Toaddress this problem, we investigated a statistical approach to thequantification of SERS from SERS probe clusters. The investigation beganby considering multiple biotinylated surfaces that had been exposed topegylated NPs (designed for biological SERS mapping) functionalizedwith streptavidin (defined as SERS probes). The surfaces were imagedwith a scanning electron microscope and SERS-mapped with a Ramanmicroscope. Statistical distributions of the SERS probe clusters andmapped SERS intensities on the surfaces were developed. It was found that there was a smooth polynomial relationship betweenSERS intensity and probe cluster size. Our result is in contrast to the sharp, highly variable intensity increases observed in studiesof unmodified NPs. Based on the polynomial relationship found, it is clear that pegylated NP SERS probes might be useful forquantification in the SERS mapping of biological material, as the SERS intensity can be potentially related back to the number ofprobes at the acquisition point.

■ INTRODUCTION

Surface enhanced Raman scattering (SERS) has proven to be avery useful technique in the field of cellular imaging andbioanalytics due to its ability to detect low-level concentrationsof molecules with a high degree of specificity.1−12 SERS-basedimmunoassays have been used for the detection of biomarkersin samples such as cells,13−20 tissue,21−23 and human bloodserum.24−29 Metallic nanoparticles (NPs; typically silver orgold) are the SERS-enabling platform in those assays. Typically,the NPs are coated with a Raman reporter molecule andfunctionalized with an antibody for a particular biomarker ofinterest to impart selectivity to the assays.14,16,21−23,28 The NP-reporter-antibody platforms used in these types of applicationsare known as “SERS probes”. In order to be comparable tomore traditional immunoassay techniques, such as the enzyme-linked immunosorbent assay (ELISA), quantification of thedetected biomolecules is necessary, and has been attempted inSERS with promising results in many cases.1,5,6,8,11,14−16,26,28−48

However, a well-known difficulty with SERS is the highlyvariable signal intensity, due to its dependence on the structureof the SERS-active substrate.45,47,49−60 In the case of metalnanoparticles, the size, shape, and interparticle distance at the

probed area all have a strong influence on the observed SERSintensity.45,51−53,56,57

While modern techniques allow for a great degree of controlover the size and shape of NPs,55,59 the random degree ofaggregation of SERS probes during an assay can potentially leadto wild variations in SERS signal. If the SERS probes areintroduced to a sample that is in solution and assumed to havea relatively uniform target molecule (analyte) concentration,the variations in measured SERS intensity can be addressed bysimply performing spectral averaging throughout the sam-ple.8,14,22,28,30,32,34−37,39 Normalization using an internal stand-ard to correct for changes in the sample properties, SERSsubstrate, or laser beam intensity during the acquisition periodis also used for quantification.26,29,33,41

However, signal averaging might not be appropriate inattempts to use SERS probes to map an analyte that isnonuniformly distributed over a surface, as in the case of aplanar assay platform or in the in vitro detection of cellmembrane proteins. When mapping a surface, the SERS

Received: April 29, 2013Revised: July 21, 2013Published: July 23, 2013

Article

pubs.acs.org/JPCC

© 2013 American Chemical Society 16596 dx.doi.org/10.1021/jp404250q | J. Phys. Chem. C 2013, 117, 16596−16605

intensity at each acquisition point depends on the number ofNPs at that point and how they are grouped (aggregated)relative to each other. In order to quantify the analyte at a givenpoint in a sample, the measured SERS spectrum has to berelated back to the number of SERS probes at the measurementpoint. It is then possible to determine the amount of analyte,assuming a linear relationship between the number of targetmolecules and the number of adsorbed SERS probes. However,determining the number of NPs that resulted in a particularSERS intensity is not straightforward and requires a specificunderstanding of the variability in SERS efficiency provided bydiffering groupings of clusters.The variation in SERS intensity with NP aggregation is well-

known, and previous studies have monitored changes inintensity caused by the chemically induced aggregation ofnanorods61 and aggregation due to the deposition of multipleNP monolayers.62−64 In general, it was found that an increasein particle aggregation leads to an increase in the average SERSintensity. A few groups have studied the SERS enhancementprovided by individual NP clusters of varying size.51−53,57 Forinstance, Wustholz et al.53 achieved a degree of control over theNP aggregations by using field flow fractionation65 andstabilized the resulting clusters by individually encasing themin SiO2 shells. After identifying individual clusters withtransmission electron microscopy (TEM) and acquiring SERSand localized surface plasmon resonance (LSPR) measure-ments, they discovered that there is not necessarily an increasein SERS enhancement with the cluster size.53 The resultsshowed that a NP dimer can provide SERS intensities as high asthat of a NP heptamer, with the enhancement much moredependent on the size of the interparticle gap. Crevices formedin the junction between adjacent NPs act as “hot spots”dominating the overall enhancement provided by the clusterand can result in signal over 108 times larger than that ofnormal Raman scattering.In practice, the strong dependence on the interparticle gap

can be very problematic when attempting to quantify the resultsof SERS mapping. However, a popular type of NP probetypically used in SERS imaging of cells and tissues has a NPcore protected by a coating of polyethylene glycol (PEG),17

which should control the gap distance between the Au-NPcores upon aggregation, leading to smaller variations in SERSintensities. The potential of these SERS probes to reduce SERSintensity variation warrants further investigation. Furthermore,despite the success of Wustholz et al.53 in isolating andanalyzing a variety of NP aggregates, their sample size of 89individual nanoclusters represents only a small fraction of theNP cluster sizes that could be encountered in, for instance, acell imaging experiment. In order to more completelycharacterize the relationship between SERS intensity and NPcluster size, we have developed a statistical approach thatenables sampling of thousands of NP aggregates. Also, thisapproach helps to mitigate the time-consuming efforts ofattempting to obtain images of the probe clusters (taken viascanning electron microscope (SEM),20 atomic force micros-copy (AFM),52 or TEM53) and SERS data from the exact sameregions of the sample. Through the analysis of SEM images andSERS maps acquired from randomly chosen regions of a goldsurface coated with SERS probe clusters of varying size, arelationship between SERS intensity and cluster size isdetermined.By implementing this statistical approach, a large sample size

is achieved with as many as 20 000 SERS intensities and 1500

individual clusters included in the analysis for a singleincubation time, and four different NP incubation timesconsidered in all. Furthermore, the use of pegylated SERSprobes, instead of uncoated Au NPs, provides information onhow SERS intensity varies for probes that are specificallydesigned for target applications involving SERS imaging ofbiological cells. As a result, analysis of the SERS intensities ofclusters formed by these SERS probes will provide informationthat could be directly applicable in vitro, allowing, for example,the identification and quantification of protein cancer-markerswithin cells.17

■ EXPERIMENTAL METHODSChemicals. Ultrapure water (UP-H2O) was used through-

out the experiment (Barnstead Nanopure ultrapure waterpurification system, 18.2 MOhm-cm resistivity; ThermoScientific, Ottawa, Ontario, Canada). Gold(III) chloridetrihydrate (HAuCl4·3H20; ≥99.9+% trace metals basis), sodiumcitrate dehydrate (≥99%, FG), sodium azide (ReagentPlus,≥99.5%), Nile Blue A perchlorate (dye content 95%),streptavidin (essentially salt-free, lyophilized powder, ≥13units/mg protein), bovine serum albumin (BSA; ≥98%(agarose gel electrophoresis), lyophilized powder), Polysorbat20 (Tween20), and acetone (ACS reagent, ≥99.5%) werepurchased from Sigma Aldrich (Saint Louis, MO, USA).Polyethylene glycols (PEG) were obtained from RappPolymere (Tuebingen, Germany; HS-PEG-COOH, molecularweight = 3000 Da) and Jenkem Technology (Allen, Texas,USA; Methoxy PEG-Thiol, molecular weight = 5000 Da). Ethyldimethylaminopropyl carbodiimide (EDC) and sulfo-N-hydroxysuccinimide (sulfo-NHS) were purchased from Pro-teoChem (Cheyenne, Wyoming, USA). Anhydrous ethylalcohol was purchased from Commercial Alcohols (Brampton,Ontario, Canada); HPLC grade methanol from CaledonLaboratory Chemicals (Caledon Laboratories Ltd., Georgetown(Halton Hills), Ontario, Canada). Biotinylated tri(ethyleneglycol) undecane thiol was purchased from WhitesidesMonothiol (Nanoscience Instruments, Inc., Phoenix, AZ,USA).

Synthesis of Au NPs. Colloidal gold nanoparticles weresynthesized by the reduction of HAuCl4 with sodium citrate ina glass beaker.62,66−68 To begin, 1 mL of a 1% solution ofHAuCl4·3H20 was added to 99 mL of UP-H2O, and theresulting solution was brought to a boil while stirring. Onceboiling, 1 mL of a 1% trisodium citrate dihydrate solution wasadded dropwise to the mixture. After boiling for another 5 min,the heat was turned off, and 1 mL of a 5% sodium azidesolution was added dropwise to the beaker. The mixturecontinued to stir as it was allowed to cool for about 30 min,after which point it could be stored in the refrigerator.

Preparation of SERS Probes. SERS probes were producedby modifying the NPs of the preceding section according to anamended version of the method described by Qian et al.17 Thenanoparticles were first coated with NBA dye, followed by aPEG shell to which a targeting molecule was attached (see theschematic in Figure 1a). The PEG coating acts to reduce thenonspecific binding of the SERS probes to sites other than thetarget.To begin, 250−300 μL of a 5 μM NBA solution was slowly

added, dropwise, to 2 mL of a vigorously stirring NPsuspension. The dye was allowed to mix with the NPs for 15min, after which time 320 μL of a mixed PEG solution wasadded to the rapidly stirring mixture. The mixed PEG solution

The Journal of Physical Chemistry C Article

dx.doi.org/10.1021/jp404250q | J. Phys. Chem. C 2013, 117, 16596−1660516597

was comprised of 0.5 mL of a 200 μM solution of HS-PEG-COOH (MW = 3000 Da) and 0.25 mL of a 200 μM methoxylPEG-Thiol solution (PEG-SH, MW = 5000 Da), diluted to 1mL with UP-H2O. The heterofunctional linker HS-PEG-COOH was included to attach to the streptavidin (targetingmolecule), while the PEG-SH was to coat any areas of the goldNPs not covered by the linker.After stirring for 20 min, the NP/NBA/PEG suspension was

centrifuged twice at 13 000 rpm for 7 min each time andresuspended in 2 mL of UP-H2O. To activate the COOHgroups on the NP surfaces, 40 μL of a 20 mg/mL solution ofEDC was added, rapidly followed by 40 μL of a 55 mg/mLsolution of sulfo-NHS. Following 40 min of stirring, 0.5 mL of aPBS buffer containing 1% Tween20 (PBST) was added to theNP mixture to improve its solubility, and the suspension wasstirred for a further 5 min. Excess reagent was then removed bytwo rounds of centrifugation (13 000 rpm for 7 min) andresuspension in 2 mL of PBST. At this point, 140 μL of a 0.5mg/mL streptavidin solution was added to the nanoparticlesuspension, resulting in a protein concentration of approx-imately 35 μg/mL. The mixture was left to react overnight inthe refrigerator at 4 °C.After 24 h, 0.5 mL of PBST was added to the suspension

before centrifuging (13 000 rpm for 7 min) and resuspending in2 mL of a PBS buffer with 0.1% Tween20 and 100 μg/mL BSA.

The purpose of the BSA was to cap any unreacted activatedCOOH groups. After reacting for 20 min, the solution wascentrifuged once more and resuspended in 2 mL of PBST.

Characterization of the SERS Probes. Particle character-ization is vital to guarantee that samples from different batchesused in the experiments had comparable physical and chemicalproperties. In fact, several characterization techniques must beemployed in this type of research to also ensure that enoughinformation about the samples are available for eventualassessment of the results by other laboratories. The SERSprobes were characterized at diverse stages during the synthesisby dynamic light scattering (DLS), ultraviolet/visible (UV/vis),and surface enhanced Raman spectroscopy (SERS). Thesemeasurements were performed at four stages during the probeproduction: on unmodified Au NPs, NPs after the addition ofNBA and PEG, NPs with NBA/PEG/streptavidin, and NPswith NBA/PEG/streptavidin/BSA. Results of the character-ization process are available as Supporting Information.Particle diameter was measured by DLS using a 90Plus

Particle Size Analyzer (Brookhaven Instruments Corporation,Holtsville, NY). The DLS technique yields the hydrodynamicdiameter, which included contributions from moleculesattached to the NP surface, so it was possible to observeincreases in the effective probe diameter as the NBA, PEG, andBSA were attached to the NP core.The stability of the NPs during synthesis was monitored by

UV/vis spectroscopy, performed on the probe sample at eachstage, using a Varian Cary 50 scan spectrophotometer (AgilentTechnologies Canada Inc., Mississauga, Ontario). UnmodifiedNPs exhibited a characteristic spectral peak at approximately530 nm. The probes were discarded at any stage if the 530 nmpeak presented large red-shifts or broadening, since these areindicative of particle aggregation.SERS of the probe suspension was carried out in all stages

after the addition of the NBA to ensure that the spectralcharacteristics of the dye, including a strong peak at 600 cm−1,were always present. Raman spectroscopy was performed usingan InVia Renishaw microscope (Renishaw Inc., HoffmanEstates, IL) with a 5× dry objective (Leica Microsystems,Wetzlar, Germany) and 1200 lines/mm diffraction grating. A633 nm Helium Neon Laser (Renishaw plc, TransducerSystems Division, Gloucestershire, UK)) was used for sampleexcitation and was focused on a drop of the SERS probesuspension centered on a glass slide. Spectra were acquired witha 5 s exposure and a laser power of ∼1 mW at the sample. Laserpower through the objective was measured using a CoherentFieldMax-TOP Laser Power/Energy Meter (Coherent Inc.,Portland, Oregon, USA).

Biotinylation of Gold Slides and SERS ProbeIncubation. Glass slides coated with 100 nm gold filmsthrough a 5 nm chromium (Cr) adhesion layer (EMFCorporation, Ithaca, NY, USA) were chemically modifiedwith biotin in order to provide a surface to which thestreptavidin-modified SERS probes could attach. Each slide wasfirst annealed, thoroughly rinsed with acetone and ethanol, andplaced in a clean 50 mL glass beaker. A 1 mM solution ofbiotinylated tri(ethylene glycol) undecane thiol was thenprepared in 2 mL of ethanol and added to the beaker. Thebeaker was sealed with Parafilm M (Bemis Company, Inc.,Neenah, WI, USA) as a monolayer self-assembled on the Ausurface for 12 h.The slide was then removed from the biotin solution, rinsed

thoroughly with ethanol and UP-H2O, and placed in a sealed

Figure 1. Process outline showing the (a) SERS probes bound to thesurface and a schematic of the SERS probes; (b) SEM and SERS dataacquisition (M1 = map 1, etc.); (c) statistical distributions generatedfrom the SERS and SEM data; (d) general equation for the total SERSintensity, generated using the distributions from panel c; (e)relationship between SERS intensity and SERS probe cluster size.

The Journal of Physical Chemistry C Article

dx.doi.org/10.1021/jp404250q | J. Phys. Chem. C 2013, 117, 16596−1660516598

beaker containing the SERS probe suspension for either 1, 2, 3,or 12 h. After the incubation, the surface was washedthoroughly with PBST and UP-H20 before being left to air-dry.SERS Mapping and SEM Imaging. SERS Microscopy.

Raman spectroscopy was performed with the InVia Renishawmicroscope using a 100× dry objective (N.A. = 0.9; LeicaMicrosystems, Wetzlar, Germany), 1200 lines/mm diffractiongrating, and 633 nm helium neon laser. Maps of three randomlychosen regions were acquired on the gold slide correspondingto each SERS probe incubation time. To ensure rapid scanningwithout compromising signal-to-noise ratio, Streamline PlusRaman imaging (Renishaw Inc., Hoffman Estates, IL) was used.The resolution of the maps was 1 μm × 0.5 μm, with each pointexposed to ∼1 mW of laser power for 3 s. Each map covered anarea of about 4000 μm2 and took approximately 15 min tocomplete. Two-dimensional SERS map data was presented witheach pixel corresponding to the area under the NBA dye peakat 600 cm−1. The three SERS maps for each slide were alsocombined and presented as a histogram of SERS intensity. Inthese histograms, SERS intensity refers to the area of the NBAdye peak at 600 cm−1.SEM Imaging. A Hitachi S-4800 field emission scanning

election microscope (FESEM; Hitachi High-TechnologiesCanada, Inc., Toronto, Ontario) was used to acquire highresolution, high magnification images of the SERS probes onthe surface of the gold slides. All images were obtained using anacceleration voltage of 1.0 kV and a magnification of 20 000×,covering an area of about 30 μm2. A sampling of images of theSERS probes was acquired on the slide corresponding to eachincubation time. At least five images were taken of each slide,with any differences in the imaged area of the slides accountedfor in the analysis.Data Analysis and Statistical Approach. An overview of

the statistical approach is provided in Figure 1. A detaileddescription of each step is provided as Supporting Information.The SERS probes were pegylated gold NPs,17 coated withSERS-active NBA dye, and functionalized with streptavidin inorder to target a biotin monolayer deposited on a gold slide(Figure 1a). Several different incubation times were used toallow for the production of SERS probe clusters of a greatvariety of sizes, from single probes to clusters as large as 15probes or more. The clusters were identified and categorizedthrough the analysis of SEM images and the SERS intensitieswere determined from two-dimensional SERS maps of multipleregions on the slide (Figure 1b). Previous investigations in thecorrelation between SERS intensities and NP cluster sizes reliedon using the same region for imaging (AFM,52 TEM,53 orSEM20) and SERS and a direct, individualized, comparisonbetween SERS intensities and cluster sizes. In contrast, wedeveloped an approach where the correlation between theSERS intensity mappings and the SEM images was based onthe distributions of SERS intensities under each laserilluminated spot and the distribution of SERS probe clustersizes under a similar area. Representative histograms of theSERS probe cluster size and SERS intensity for a givenincubation time were generated (Figure 1c), and the twodistributions were related through a series of equationsdescribing the total SERS intensity acquired at a given point(Figure 1d). Solutions of the empirical formulas provided arelationship between SERS probe cluster size and the averageSERS intensity resulting when the cluster was probed with theRaman microscope (Figure 1e).

■ RESULTSSERS Mapping and Histograms. Figure 2 shows the

SERS mapping of the gold slides coated with SERS probes for

four different incubations times (1−12 h). The SERS probes(illustrated in Figure 1a) included streptavidin proteins thatspecifically attached to the gold surface decorated with biotin.The SERS mapping at low incubation times (Figure 2a) showsa large spatial variation in SERS intensities, due to the sparsecoverage of the gold surface by SERS probes. The intensitymaps become more uniform as the incubation time increases(Figure 2, panels b−d). In those cases, the overall SERS probesurface coverage increases, as does the number of particlesilluminated in each mapped spot. These observations arecorroborated in Figure 3, where the distributions of SERSintensities corresponding to each incubation time arerepresented as histograms. Each count in the histogramcorresponds to the SERS intensity (integrated SERS peak)obtained under a laser-illuminated area during mapping. Thehistograms were generated by combining the data from threelarge area (76 × 53 μm2) maps obtained from randompositions in every slide, leading to a large sample size of SERSintensities. Zero intensity events (defined by a cutoff) werediscarded in the construction of the histograms; hence, thesample sizes varied from approximately 7000 for the 1hr data toover 23 000 for the 12 h data. Following a common treatmentof error on histograms, the measured number of counts in eachbin (Nbin) is assumed to be the mean of a Poisson distribution,with the error bars corresponding to ±Nbin

1/2 (1σ).69,70

The results for the shortest incubation time (Figure 3a) showa tailed distribution, with a skewness of 10.1 (skewness valuescan provide a quantitative assessment regarding the symmetryof the distribution71), and a high number of low intensityevents. These correspond to illuminated areas that containedlow efficiency probes that produced a Raman signal near thedetection limit of the system. Even at this low coverage regime,however, a small number of events presented a relatively highSERS intensity (∼1000 counts). The distribution clearly spreadto higher intensities as the incubation time increased (Figure3b) and the number of regions with zero-intensity events

Figure 2. SERS map of the gold slide surface after a SERS probeincubation of (a) 1, (b) 2, (c) 3, and (d) 12 h.

The Journal of Physical Chemistry C Article

dx.doi.org/10.1021/jp404250q | J. Phys. Chem. C 2013, 117, 16596−1660516599

decreased. When the incubation time reached 3 h (Figure 3c),the number of low signal events became small, and a peakclearly develops in the distribution. This result is consistentwith a larger coverage and a higher probability of SERS probesbeing found under each laser illuminated area during mapping.Finally, in Figure 3d, the least skewed distribution, a peak formsat higher intensities than in Figure 3c. The distribution inFigure 3d can be said to be more symmetric than the others(Figure 3, panels a−c), based on its low skewness value,appearing almost Gaussian in shape.71 This result is consistentwith a higher surface coverage of SERS probes and,consequently, a larger probability of several NPs beinginterrogated by the laser in each particular mapped spot.SEM Imaging and Histograms. Sample SEM images of

the gold slide incubated with SERS probes for different timesare shown in Figure 4 (SEM images of individual clusters areprovided in the Supporting Information). The scale of theimages in Figure 4 was chosen so as to provide the largestsurface area possible, displaying a large number of clusters whilestill allowing for the identification of individual probes. SingleSERS probes were identified, and a histogram of their pixelareas (using data from all SEM images) was generated and fitwith a Gaussian. From the Gaussian, the average size of a singleSERS probe was obtained, and using this value it was thenpossible to produce a histogram, presented in Figure 5, of theSERS probe cluster size distribution for the slides correspond-ing to each incubation time. The histograms in Figure 5 wereobtained using a different number of SEM images for eachincubation time. Therefore, the number of clusters wasnormalized using the imaged area. The histograms in Figure5 clearly show an increase in the number of SERS probes in theaggregated clusters with the incubation time, although some

large clusters, with more than 10 SERS probes, were alsoobserved in the 1 h incubation experiment. Notice that thenumber of single particles and small clusters (less than fiveSERS probes) still dominates the distribution even after 12 h.This is a strong indication that the PEG-layer efficientlyprotects the SERS probes against aggregation; meaning that the

Figure 3. Histogram of integrated SERS peak of Nile Blue-A (∼600cm−1) measured on the slide surface incubated with SERS probes for(a) 1, (b) 2, (c) 3, and (d) 12 h. Error bars correspond to ±1σ(N1/2).

Figure 4. Sample SEM images of the gold slide surface after SERSprobe incubation of (a) 1, (b) 2, (c) 3, and (d) 12 h.

Figure 5. Histogram showing the distribution of cluster sizes(measured in the number of SERS probes per cluster) for the regionof the slide incubated with SERS probes for (a) 1, (b) 2, (c) 3, and (d)12 h. The histograms are normalized by the area of the slide imagedwith the SEM (ASEM). Error bars indicate ±1σ((1/ASEM)N

1/2).

The Journal of Physical Chemistry C Article

dx.doi.org/10.1021/jp404250q | J. Phys. Chem. C 2013, 117, 16596−1660516600

SERS probe adsorption is possibly driven by the specificinteraction with the surface, rather than by particle-to-particleattraction.Our final goal is to correlate the SERS intensity from each

illuminated spot obtained from the SERS map histograms(Figure 3) with the number of SERS probe clusters at that spot.In order to achieve that, a new set of histograms, shown inFigure 6 were obtained from the SEM images. In the case of

Figure 6, the histograms considered the number of SERSprobes clusters illuminated per acquisition (i.e., the number ofSERS probe clusters in a 1 × 1 μm2 SEM imaged area). Asexpected, the histograms in Figure 6 shows that each laser-illuminated area is mostly populated with either 0, 1, or 2 SERSprobe clusters for the lower incubation time (1 h). On theother hand, most of the mapped SERS intensities weregenerated by illuminating two to five SERS probe clusterswhen the incubation was 12 h.Finally, by sampling the histogram in Figure 3, a SERS

intensity from a random 1 × 1 μm2 spot is obtained. Theprobability that the intensity is from a certain number ofclusters is generated from Figure 6, and the possible size ofthese clusters (number of SERS probes per cluster) is revealedby sampling the distribution in Figure 5. Therefore, using thetotal SERS intensities of Figure 3 and the histograms of Figures5 and 6, a system of equations was then producedcorresponding to the relationship between the SERS intensityand the number of clusters. This approach allowed the use ofthousands of sampled values from all histograms, generating astatistically robust result. A more detailed description of this

procedure, including examples, is available as SupportingInformation.

Solving Systems of Equations for the SERS IntensityRatios. As discussed in the previous section, the SERS intensityat each illuminated area was correlated to the number of SERSprobe clusters and their size in that same area. In other words,the total SERS intensity from a particular spot was treated as alinear sum of SERS intensity contributions from each cluster.The individual intensities from each cluster were unknown.Therefore, for each incubation time, the process was repeatedto generate enough equations that would allow the unknownSERS intensities, attributed to each SERS probe cluster size, tobe evaluated. The system of equations for each SERS data setwas then an m × m matrix, with m being the maximum clustersize for a particular set. Due to the fact that very large SERSprobe clusters (>15 probes) were rarely detected, particularlyfor the shorter probe incubation times (see Figure 5), it wasdecided that the total SERS intensity distributions could bedescribed by only considering SERS probe clusters up to amaximum cutoff value, mmax. For each incubation time, adifferent value for mmax was chosen, based on the cluster sizedistributions of Figure 5.Starting with the 1 h incubation time, a value of mmax = 6 was

chosen (based on where the histogram begins to approach zeroin Figure 5a), and a 6 × 6 equation matrix was generated. Thisprocedure is illustrated in Figure 7. The solutions to the 6 × 6matrix provided values for I1 to I6, corresponding to thecontributions to the total SERS intensity from one SERS probe(I1) and a cluster containing six SERS probes (I6), respectively.To account for variability in these solutions, 500 6 × 6 equationmatrices were generated for the 1 h data set and solved, giving500 independent sets of solutions (Figure 7b). All solutions forI1 to I6 were then averaged and the uncertainties on the meanvalues were calculated (Figure 7c). To provide an idea of thevariability in the values of these intensities, seven independentcalculations of I1 to I6 for the 1 h incubation experiment havebeen overlaid in Figure 8. In Figure 8, the solution for I6 is seento have a high uncertainty compared to the first five SERSintensity solutions due to the fact that it contains contributionsfrom the relatively small number of NP clusters larger than sixthat form after a 1 h incubation. As a result, I6 was recalculatedfrom the equations for the 2 h incubation, while I1−I5 weretreated as known values.For the 2 h incubation, a system of equations was generated

with an mmax set to be equal to nine (based on Figure 5b), andusing the intensity solutions for clusters of one to five SERSprobes as known values, solutions for clusters up to nine probeslarge were found (Figure 7d). The process was again repeated500 times, as discussed in Figure 7e. The 3 h probe incubationdata was then similarly used to find intensity solutions forclusters up to 12 probes large (this time with I1 to I8 as knownvalues), and the 12 h probe incubation data was used to findsolutions for SERS intensities produced by clusters up to 15probes large (with I1 to I11 as known values). The final resultsof the calculations are found in Figure 9, where the averageSERS intensities normalized to the intensity due to a singleprobe (I/I1) are plotted against the number of SERS probes percluster.High reproducibility of the calculated SERS intensities of

Figure 9 is suggested by the strong agreement (p < 0.05)between the seven sets of independent calculations of the SERSintensities for the first six cluster sizes shown in Figure 8.

Figure 6. Histograms showing the distribution of illuminated SERSprobe clusters per acquisition for the data corresponding to the (a) 1,(b) 2, (c) 3, and (d) 12 h probe incubation times. Error barscorrespond to ±1σ(N1/2).

The Journal of Physical Chemistry C Article

dx.doi.org/10.1021/jp404250q | J. Phys. Chem. C 2013, 117, 16596−1660516601

■ DISCUSSIONWhen using NPs as a substrate in SERS, the specificenhancement to the Raman signal of the sample provided bythe NPs can be highly variable and difficult to control. Ingeneral, it is known that an irregular surface allows for thegreatest enhancement and that a collection of NPs providesgreater enhancement than a single NP.60 Studies that observethe change in the SERS intensity as a function of random NPaggregation have shown an increase in intensity withaggregation,61−64 and the polynomial relationship shown inFigure 9 supports this idea. However, the gradual increase inSERS intensity with SERS probe cluster size is somewhatunexpected considering that studies of multiple depositions ofNP monolayers show a more rapid increase in average SERSintensity as the number of layers (and thus NP aggregation)increases, eventually reaching a plateau.62−64 Similarly, studies

of individual NP clusters52,53 found a very sharp increase inintensity (as much as 4 orders of magnitude) between singleand dimer NPs. Although the results recently reported byPazos-Peres et al.57 showed a smaller enhancement of the dimerrelative to the single particle, the importance of interparticlespacing when it comes to SERS enhancement was also evident.However, an increase in the cluster size beyond the dimer doesnot necessarily result in a much larger increase in SERSefficienty. In fact, Wustholz et al.53 suggest that a pair of NPswith a particular spacing can produce a “hot spot” which canpotentially result in a SERS enhancement as large as thatprovided by bigger clusters.The slowly rising curve in Figure 9 seems to suggest that the

type of “hot spot” that would result in a 4 orders of magnitudeincrease in SERS intensity between a single and dimer SERSprobe was not observed in this work or at least did not occur

Figure 7. Chart describing the method used to solve the system of equations for the SERS intensities due to each SERS probe cluster size.

Figure 8. Seven independent calculations of the SERS intensities dueto the first six cluster sizes. Figure 9. SERS intensity ratio solutions due to clusters from 1 to 15

SERS probes in size. The data set used to calculate each of the values isindicated. A polynomial fit of the first 14 intensity ratios is included.

The Journal of Physical Chemistry C Article

dx.doi.org/10.1021/jp404250q | J. Phys. Chem. C 2013, 117, 16596−1660516602

frequently enough to influence the average SERS intensities foreach cluster size. Likely, this is due to the fact that the distancebetween the Au-NP cores of the SERS probes used to generateFigure 9 was controlled by the PEG coating. The pegylation ofthe Au-NPs and their functionalization with streptavidin toform the SERS probes provided a relatively “thick” coating thatlimited the distance of the closest approach between the Au-NPcores which was not present on the Au-NPs used in the work ofWustholz et al.53 Due to this fundamental difference, we cannotexpect that our work will compare directly with that ofWustholz et al.53 However, the systematic attempt to correlateSERS intensities to NPs cluster sizes performed by Wustholz etal.53 does mirror the study described here. In our case, themolecular coating provided by the PEG seems to haveprevented the conditions necessary for the formation of thetype of “hot-spots” formed in between the NPs that providedsuch significant rises in SERS intensity53,57 and likely had aneffect on the rapid increase in SERS intensity found in thestudies of multiple depositions of NP monolayers62−64 and ofnanorod aggregation.61 Indeed, in high magnification SEMimages of clusters of the SERS probes (as can be seen in theSupporting Information), there is a clear separation of 10−20nm between many of the probes within the clusters. It ispossible that the separation between the NPs in the SERSprobe clusters causes the probes to provide surface-enhance-ment somewhat independently of each other, and the intensitydue to a cluster is the cumulative result of the intensity due toeach probe in the cluster. However, this would suggest a linearrelationship, which is not what is seen in Figure 9. While theexact reason for the nonlinear nature of the relationship inFigure 9 is unclear, it does imply that there may be some degreeof “hot-spot” interaction between the probes in the cluster.Similarly, it is possible that the SERS intensity produced by theclusters is affected by a relationship between the gold slidesurface and the SERS probes, whether that be due to a “hot-spot” interaction, or simply reflection of Raman scatteredphotons toward the objective that would otherwise have beenlost. Nonetheless, the polynomial relationship between SERSprobe clusters and their resultant SERS intensity is extremelywell-defined (R2 = 0.99812) and reproducible (see Figure 8),and a nonlinear component to the relationship is entirelypossible.Given confidence in the reproducibility and accuracy of the

results shown in Figure 9, the polynomial relationship that thefigure displays will prove to be useful in experiments that usePEG NPs as SERS probes. The exhibited relationship betweencluster size and the resultant SERS intensity is unique to thePEG NP probes described in this work, and these probes, basedon those in the paper by Qian et al.17 are specifically designedfor targeted SERS imaging, such as in the identification ofproteins within cells and tissues. As discussed in theintroduction, determining the quantity and distribution of thetarget within the SERS maps of a sample remains challenging.However, considering the simple relationship between clustersize and SERS intensity shown in Figure 9, PEG NP SERSprobes seem ideally suited for quantifiable SERS mapping. Forexample, using a calibration curve similar to Figure 9, the SERSintensity at a given point would be known to be a multiple ofthe intensity due to an individual SERS probe. It would then besimple to determine the number of clusters contributing to thetotal SERS intensity at the given point and thus the number oftargeted sites could be determined at that location.

■ CONCLUSIONThis work provides a novel technique that relates 2D SERSintensity maps to SEM images of SERS probes on abiotinylated gold surface, allowing for the calculation of SERSintensities due to SERS probe clusters of varying size. In orderto examine a variety of cluster sizes, targeted SERS probes wereincubated with biotinylated surfaces for several time periodsprior to the acquisition of SEM and SERS images. Using theSEM images and SERS maps, statistical descriptions of both theSERS probe cluster distribution and SERS intensities on thesurface were developed. Based on the statistical relationships,systems of equations for the total SERS intensity measured at agiven point were devised. Solutions to the equations providedthe SERS intensities resulting from enhancement due toclusters from 1 to 14 SERS probes in size.A simple polynomial relationship between SERS probe

cluster size and the resultant SERS intensity was found,suggesting the utility of pegylated SERS probes in targetedSERS mapping. By simply controlling the probe incubationtime to limit the distribution of cluster sizes formed in atargeted SERS sample, quantification of targeted sites within asample would be relatively straightforward. The ability toquantify targeted sites in a sample would make targeted SERSmapping a viable technique for imaging protein distributionwithin cells and tissues, providing a quantifiable, high-resolutionalternative to methods such as immunohistochemistry.

■ ASSOCIATED CONTENT*S Supporting InformationInformation regarding the characterization of the SERS probes;high magnification SEM images of individual probe clusters; adetailed description of the data analysis. This material isavailable free of charge via the Internet at http://pubs.acs.org.

■ AUTHOR INFORMATIONCorresponding Author*E-mail: jirasek@uvic.ca. Phone: 1 250 721 7704 (A.I.J.). E-mail: agbrolo@uvic.ca. Phone: 1 250 721 7167 (A.G.B.).NotesThe authors declare no competing financial interest.

■ ACKNOWLEDGMENTSThis work was supported by operating grants from NSERC andby equipment grants from the Canadian Foundation forInnovation, the British Columbia Knowledge DevelopmentFund, and the Western Economic Diversification Fund. Theauthors thank Dr. Elaine Humphrey and Adam Schuetze fortheir assistance with the SEM imaging and Dr. Quinn Matthewsand Samantha Harder for the laser power measurements.

■ REFERENCES(1) Hudson, S. D.; Chumanov, G. Bioanalytical Applications of SERS(Surface-Enhanced Raman Spectroscopy). Anal. Bioanal. Chem. 2009,394, 679−686.(2) Smith, W. E. Practical Understanding and Use of SurfaceEnhanced Raman Scattering/Surface Enhanced Resonance RamanScattering in Chemical and Biological Analysis. Chem. Soc. Rev. 2008,37, 955−964.(3) Zhang, Y.; Hong, H.; Myklejord, D. V.; Cai, W. MolecularImaging with SERS-Active Nanoparticles. Small 2011, 7, 3261−3269.(4) Huh, Y. S.; Chung, A. J.; Erickson, D. Surface Enhanced RamanSpectroscopy and its Application to Molecular and Cellular Analysis.Microfluid. Nanofluid. 2009, 6, 285−297.

The Journal of Physical Chemistry C Article

dx.doi.org/10.1021/jp404250q | J. Phys. Chem. C 2013, 117, 16596−1660516603

(5) Kneipp, K.; Kneipp, H.; Kneipp, J. Surface-Enhanced RamanScattering in Local Optical Fields of Silver and Gold Nanoaggregates-From Single-Molecule Raman Spectroscopy to Ultrasensitive Probingin Live Cells. Acc. Chem. Res. 2006, 39, 443−450.(6) Hering, K.; Cialla, D.; Ackermann, K.; Dorfer, T.; Moller, R.;Schneidewind, H.; Mattheis, R.; Fritzsche, W.; Rosch, P.; Popp, J.SERS: A Versatile Tool in Chemical and Biochemical Diagnostics.Anal. Bioanal. Chem. 2008, 390, 113−124.(7) Qian, X.-M.; Nie, S. M. Single-Molecule and Single-NanoparticleSERS: from Fundamental Mechanisms to Biomedical Applications.Chem. Soc. Rev. 2008, 37, 912−920.(8) Han, X. X.; Zhao, B.; Ozaki, Y. Surface-Enhanced RamanScattering for Protein Detection. Anal. Bioanal. Chem. 2009, 394,1719−1727.(9) Ryder, A. G. Surface Enhanced Raman Scattering for NarcoticDetection and Applications to Chemical Biology. Curr. Opin. Chem.Biol. 2005, 9, 489−493.(10) Willets, K. A. Surface-Enhanced Raman Scattering (SERS) forProbing Internal Cellular Structure and Dynamics. Anal. Bioanal.Chem. 2009, 394, 85−94.(11) Porter, M. D.; Lipert, R. J.; Siperko, L. M.; Wang, G.;Narayanan, R. SERS as a Bioassay Platform: Fundamentals, Design,and Applications. Chem. Soc. Rev. 2008, 37, 1001−1011.(12) Doering, W. E.; Piotti, M. E.; Natan, M. J.; Freeman, R. G. SERSas a Foundation for Nanoscale, Optically Detected Biological Labels.Adv. Mater. (Weinheim, Ger.) 2007, 19, 3100−3108.(13) Eliasson, C.; Loren, A.; Engelbrektsson, J.; Josefson, M.;Abrahamsson, J.; Abrahamsson, K. Surface-Enhanced Raman Scatter-ing Imaging of Single Living Lymphocytes with MultivariateEvaluation. Spectrochim. Acta. Part A 2005, 61, 755−760.(14) Kim, J.-H.; Kim, J.-S.; Choi, H.; Lee, S.-M.; Jun, B.-H.; Yu, K.-N.; Kuk, E.; Kim, Y.-K.; Jeong, D. H.; Cho, M.-H.; Lee, Y.-S.Nanoparticle Probes with Surface Enhanced Raman SpectroscopicTags for Cellular Cancer Targeting. Anal. Chem. 2006, 78, 6967−6973.(15) Sun, L.; Irudayaraj, J. Quantitative Surface-Enhanced Raman forGene Expression Estimation. Biophys. J. 2009, 96, 4709−4716.(16) Pallaoro, A.; Braun, G. B.; Moskovits, M. QuantitativeRatiometric Discrimination Between Noncancerous and CancerousProstate Cells Based on Neuropilin-1 Overexpression. Proc. Natl. Acad.Sci. U.S.A. 2011, 108, 16559−16564.(17) Qian, X.; Peng, X.-H.; Ansari, D. O.; Yin-Goen, Q.; Chen, G. Z.;Shin, D. M.; Yang, L.; Young, A. N.; Wang, M. D.; Nie, S. In vivoTumor Targeting and Spectroscopic Detection with Surface-EnhancedRaman Nanoparticle Tags. Nat. Biotechnol. 2008, 26, 83−90.(18) Nithipatikom, K.; McCoy, M. J.; Hawi, S. R.; Nakamoto, K.;Adar, F.; Campbell, W. B. Characterization and Application of RamanLabels for Confocal Raman Microspectroscopic Detection of CellularProteins in Single Cells. Anal. Biochem. 2003, 322, 198−207.(19) Lee, S.; Kim, S.; Choo, J.; Shin, S. Y.; Lee, Y. H.; Choi, H. Y.;Ha, S.; Kang, K.; Oh, C. H. Biological Imaging of HEK293 CellsExpressing PLCgamma1 Using Surface-Enhanced Raman Microscopy.Anal. Chem. 2007, 79, 916−922.(20) Hu, Q.; Tay, L.-L.; Noestheden, M.; Pezacki, J. P. MammalianCell Surface Imaging with Nitrile-Functionalized Nanoprobes:Biophysical Characterization of Aggregation and Polarization Aniso-tropy in SERS Imaging. J. Am. Chem. Soc. 2007, 129, 14−15.(21) Schlucker, S.; Kustner, B.; Punge, A.; Bonfig, R.; Marx, A.;Strobel, P. Immuno-Raman Microspectroscopy: In Situ Detection ofAntigens in Tissue Specimens by Surface-Enhanced Raman Scattering.J. Raman Spectrosc. 2006, 37, 719−721.(22) Sun, L.; Sung, K.-B.; Dentinger, C.; Lutz, B.; Nguyen, L.; Zhang,J.; Qin, H.; Yamakawa, M.; Cao, M.; Lu, Y.; Chmura, A. J.; Zhu, J.; Su,X.; Berlin, A. A.; Chan, S.; Knudsen, B. Composite Organic-InorganicNanoparticles as Raman Labels for Tissue Analysis. Nano Lett. 2007, 7,351−356.(23) Lutz, B.; Dentinger, C.; Sun, L.; Nguyen, L.; Zhang, J.; Chmura,A.; Allen, A.; Chan, S.; Knudsen, B. Raman Nanoparticle Probes forAntibody-Based Protein Detection in Tissues. J. Histochem. Cytochem.2008, 56, 371−379.

(24) Feng, S.; Chen, R.; Lin, J.; Pan, J.; Chen, G.; Li, Y.; Cheng, M.;Huang, Z.; Chen, J.; Zeng, H. Nasopharyngeal Cancer DetectionBased on Blood Plasma Surface-Enhanced Raman Spectroscopy andMultivariate Analysis. Biosens. Bioelectron. 2010, 25, 2414−2419.(25) Eliasson, C.; Loren, A.; Murty, K. V.; Josefson, M.; Kall, M.;Abrahamsson, J.; Abrahamsson, K. Multivariate Evaluation ofDoxorubicin Surface-Enhanced Raman Spectra. Spectrochim. Acta.Part A 2001, 57, 1907−1915.(26) Zakel, S.; Rienitz, O.; Guttler, B.; Stosch, R. Double IsotopeDilution Surface-Enhanced Raman Scattering as a ReferenceProcedure for the Quantification of Biomarkers in Human Serum.Analyst 2011, 136, 3956−3961.(27) Lin, D.; Feng, S.; Pan, J.; Chen, Y.; Lin, J.; Chen, G.; Xie, S.;Zeng, H.; Chen, R. Colorectal Cancer Detection by Gold NanoparticleBased Surface-Enhanced Raman Spectroscopy of Blood Serum andStatistical Analysis. Opt. Express 2011, 19, 13565−13577.(28) Grubisha, D. S.; Lipert, R. J.; Park, H.-Y.; Driskell, J.; Porter, M.D. Femtomolar Detection of Prostate-Specific Antigen: An Immuno-assay Based on Surface-Enhanced Raman Scattering and ImmunogoldLabels. Anal. Chem. 2003, 75, 5936−5943.(29) Stosch, R.; Henrion, A.; Schiel, D.; Guttler, B. Surface-EnhancedRaman Scattering Based Approach for Quantitative Determination ofCreatinine in Human Serum. Anal. Chem. 2005, 77, 7386−7392.(30) O’neal, P. D.; Cote, G. L.; Motamedi, M.; Chen, J.; Lin, W.-C.Feasibility Study Using Surface-Enhanced Raman Spectroscopy for theQuantitative Detection of Excitatory Amino Acids. J. Biomed. Opt.2003, 8, 33−39.(31) Luo, X. L.; Buckhout-White, S.; Bentley, W. E.; Rubloff, G. W.Biofabrication of Chitosan-Silver Composite SERS Substrates EnablingQuantification of Adenine by a Spectroscopic Shift. Biofabrication2011, 3, 034108.(32) Driskell, J. D.; Kwarta, K. M.; Lipert, R. J.; Porter, M. D.; Neill, J.D.; Ridpath, J. F. Low-Level Detection of Viral Pathogens by a Surface-Enhanced Raman Scattering Based Immunoassay. Anal. Chem. 2005,77, 6147−6154.(33) Yaghobian, F.; Weimann, T.; Guttler, B.; Stosch, R. On-chipApproach for Traceable Quantification of Biomarkers Based onIsotope-Dilution Surface-Enhanced Raman Scattering (IDSERS). Lab.Chip 2011, 11, 2955−2960.(34) Yazgan, N. N.; Boyaci, I. H.; Temur, E.; Tamer, U.; Topcu, A. AHigh Sensitive Assay Platform Based on Surface-Enhanced RamanScattering for Quantification of Protease Activity. Talanta 2010, 82,631−639.(35) Guven, B.; Basaran-Akgul, N.; Temur, E.; Tamer, U.; Boyacı, I.H. SERS-Based Sandwich Immunoassay Using Antibody CoatedMagnetic Nanoparticles for Escherichia Coli Enumeration. Analyst2011, 136, 740−748.(36) Cheng, H.-W.; Luo, W.-Q.; Wen, G.-L.; Huan, S.-Y.; Shen, G.-L.; Yu, R.-Q. Surface-Enhanced Raman Scattering Based Detection ofBacterial Biomarker and Potential Surface Reaction Species. Analyst2010, 135, 2993−3001.(37) Lin, C.-C.; Yang, Y.-M.; Chen, Y.-F.; Yang, T.-S.; Chang, H.-C.A New Protein A Assay Based on Raman Reporter LabeledImmunogold Nanoparticles. Biosens. Bioelectron. 2008, 24, 178−183.(38) Knauer, M.; Ivleva, N. P.; Niessner, R.; Haisch, C. A Flow-Through Microarray Cell for the Online SERS Detection of Antibody-Captured E. Coli Bacteria. Anal. Bioanal. Chem. 2012, 402, 2663−2667.(39) Strehle, K. R.; Cialla, D.; Rosch, P.; Henkel, T.; Kohler, M.;Popp, J. A Reproducible Surface-Enhanced Raman SpectroscopyApproach. Online SERS Measurements in a Segmented MicrofluidicSystem. Anal. Chem. 2007, 79, 1542−1547.(40) Sun, L.; Irudayaraj, J. PCR-Free Quantification of MultipleSplice Variants in a Cancer Gene by Surface-Enhanced RamanSpectroscopy. J. Phys. Chem. B 2009, 113, 14021−14025.(41) Yin, P.-G.; Jiang, L.; Lang, X.-F.; Guo, L.; Yang, S. QuantitativeAnalysis of Mononucleotides by Isotopic Labeling Surface-EnhancedRaman Scattering Spectroscopy. Biosens. Bioelectron. 2011, 26, 4828−4831.

The Journal of Physical Chemistry C Article

dx.doi.org/10.1021/jp404250q | J. Phys. Chem. C 2013, 117, 16596−1660516604

(42) Schlucker, S. SERS Microscopy: Nanoparticle Probes andBiomedical Applications. Chem. Phys. Chem. 2009, 10, 1344−1354.(43) Tsoutsi, D.; Montenegro, J. M.; Dommershausen, F.; Koert, U.;Liz-Marzan, L. M.; Parak, W. J.; Alvarez-Puebla, R. A. QuantitativeSurface-Enhanced Raman Scattering Ultradetection of AtomicInorganic Ions: The Case of Chloride. ACS Nano 2011, 5, 7539−7546.(44) Clarke, S. J.; Littleford, R. E.; Smith, W. E.; Goodacre, R. RapidMonitoring of Antibiotics Using Raman and Surface Enhanced RamanSpectroscopy. Analyst 2005, 130, 1019−1026.(45) Driskell, J. D.; Lipert, R. J.; Porter, M. D. Labeled GoldNanoparticles Immobilized at Smooth Metallic Substrates: SystematicInvestigation of Surface Plasmon Resonance and Surface-EnhancedRaman Scattering. J. Phys. Chem. B 2006, 110, 17444−17451.(46) Lin, M.; He, L.; Awika, J.; Yang, L.; Ledoux, D. R.; Li, H.;Mustapha, A. Detection of Melamine in Gluten, Chicken Feed, andProcessed Foods Using Surface Enhanced Raman Spectroscopy andHPLC. J. Food Sci. 2008, 73, T129−T134.(47) Tourwe, E.; Hubin, A. Preparation of SERS-active ElectrodesVia Ex Situ Electrocrystallization of Silver in a Halide Free Electrolyte.Vib. Spectrosc. 2006, 41, 59−67.(48) Abalde-Cela, S.; Hermida-Ramon, J. M.; Contreras-Carballada,P.; De Cola, L.; Guerrero-Martínez, A.; Alvarez-Puebla, R. A.; Liz-Marzan, L. M. SERS Chiral Recognition and Quantification ofEnantiomers Through Cyclodextrin Supramolecular Complexation.ChemPhysChem 2010, 12, 1529−1535.(49) Kneipp, J.; Kneipp, H.; Kneipp, K. SERS–A Single-Molecule andNanoscale Tool for Bioanalytics. Chem. Soc. Rev. 2008, 37, 1052−1060.(50) Beermann, J.; Novikov, S. M.; Leosson, K.; Bozhevolnyi, S. I.Surface Enhanced Raman Imaging: Periodic Arrays and IndividualMetal Nanoparticles. Opt. Express 2009, 17, 12698−12705.(51) McMahon, J. M.; Henry, A.-I.; Wustholz, K. L.; Natan, M. J.;Freeman, R. G.; Van Duyne, R. P.; Schatz, G. C. Gold NanoparticleDimer Plasmonics: Finite Element Method Calculations of theElectromagnetic Enhancement to Surface-Enhanced Raman Spectros-copy. Anal. Bioanal. Chem. 2009, 394, 1819−1825.(52) Talley, C. E.; Jackson, J. B.; Oubre, C.; Grady, N. K.; Hollars, C.W.; Lane, S. M.; Huser, T. R.; Nordlander, P.; Halas, N. J. Surface-Enhanced Raman Scattering from Individual Au Nanoparticles andNanoparticle Dimer Substrates. Nano Lett. 2005, 5, 1569−1574.(53) Wustholz, K. L.; Henry, A.-I.; McMahon, J. M.; Freeman, R. G.;Valley, N.; Piotti, M. E.; Natan, M. J.; Schatz, G. C.; Van Duyne, R. P.Structure-Activity Relationships in Gold Nanoparticle Dimers andTrimers for Surface-Enhanced Raman Spectroscopy. J. Am. Chem. Soc.2010, 132, 10903−10910.(54) Aroca, R. F.; Alvarez-Puebla, R. A.; Pieczonka, N.; Sanchez-Cortez, S.; Garcia-Ramos, J. V. Surface-Enhanced Raman Scattering onColloidal Nanostructures. Adv. Colloid Interface Sci. 2005, 116, 45−61.(55) Lin, X.-M.; Cui, Y.; Xu, Y.-H.; Ren, B.; Tian, Z.-Q. Surface-Enhanced Raman Spectroscopy: Substrate-Related Issues. Anal.Bioanal. Chem. 2009, 394, 1729−1745.(56) Kim, K.; Shin, K. S. Surface-Enhanced Raman Scattering: APowerful Tool for Chemical Identification. Anal. Sci. 2011, 27, 775−783.(57) Pazos-Perez, N.; Wagner, C. S.; Romo-Herrera, J. M.; Liz-Marzan, L. M.; García de Abajo, F. J.; Wittemann, A.; Fery, A.; Alvarez-Puebla, R. A. Organized Plasmonic Clusters with High CoordinationNumber and Extraordinary Enhancement in Surface-Enhanced RamanScattering (SERS). Angew. Chem. Int. Ed. 2012, 51, 12688−12693.(58) Mahajan, S.; Baumberg, J. J.; Russell, A. E.; Bartlett, P. N.Reproducible SERRS from Structured Gold Surfaces. Phys. Chem.Chem. Phys. 2007, 9, 6016−6020.(59) Banholzer, M. J.; Millstone, J. E.; Qin, L.; Mirkin, C. A.Rationally Designed Nanostructures for Surface-Enhanced RamanSpectroscopy. Chem. Soc. Rev. 2008, 37, 885−897.(60) Moskovits, M. Surface-Enhanced Raman Spectroscopy: A BriefRetrospective. J. Raman Spectrosc. 2005, 36, 485−496.(61) Lee, A.; Andrade, G. F. S.; Ahmed, A.; Souza, M. L.; Coombs,N.; Tumarkin, E.; Liu, K.; Gordon, R.; Brolo, A. G.; Kumacheva, E.

Probing Dynamic Generation of Hot-Spots in Self-Assembled Chainsof Gold Nanorods by Surface-Enhanced Raman Scattering. J. Am.Chem. Soc. 2011, 133, 7563−7570.(62) Fan, M.; Brolo, A. G. Self-Assembled Au Nanoparticles asSubstrates for Surface-Enhanced Vibrational Spectroscopy: Optimiza-tion and Electrochemical Stability. Chem. Phys. Chem. 2008, 9, 1899−1907.(63) Fan, M.; Brolo, A. G. Silver Nanoparticles Self Assembly asSERS Substrates with Near Single Molecule Detection Limit. Phys.Chem. Chem. Phys. 2009, 11, 7381−7389.(64) Addison, C. J.; Brolo, A. G. Nanoparticle-Containing Structuresas a Substrate for Surface-Enhanced Raman Scattering. Langmuir 2006,22, 8696−8702.(65) Giddings, J. C.; Yang, F. J.; Myers, M. N. Flow Field-FlowFractionation: A Versatile New Separation Method. Science 1976, 193,1244−1245.(66) Grabar, K. C.; Freeman, R. G.; Hommer, M. B.; Natan, M. J.Preparation and Characterization of Au Colloid Monolayers. Anal.Chem. 1995, 67, 735−743.(67) Frens, G. Controlled Nucleation for the Regulation of ParticleSize in Monodisperse Gold Suspensions. Nat. Phys. Sci. 1973, 241,20−22.(68) Sutherland, W. S.; Winefordner, J. D. Colloid Filtration: ANovel Substrate Preparation Method for Surface-Enhanced RamanSpectroscopy. J. Colloid Interface Sci. 148, 129−141.(69) Agresti, A.; Coull, B. A. Approximate is Better than “Exact” forInterval Estimation of Binomial Proportions. Am. Stat. 1998, 52, 119−126.(70) Brown, L. D.; Cai, T. T.; DasGupta, A. Interval Estimation for aBinomial Proportion. Stat. Sci. 2001, 16, 101−133.(71) Doane, D. P.; Seward, L. E. Measuring Skewness: A ForgottenStatistic? J. Stat. Educ. 2011, 19, 1−18.

The Journal of Physical Chemistry C Article

dx.doi.org/10.1021/jp404250q | J. Phys. Chem. C 2013, 117, 16596−1660516605