Characterization of polycyclic aromatichydrocarbons using Raman and surface-enhanced Raman spectroscopyJing Chen,a,b* Yao-Wen Huanga,b and Yiping Zhaob,c

Polycyclic aromatic hydrocarbons (PAHs) are ubiquitous atmospheric pollutants and food contaminants, which exhibit potent car-cinogenicity, mutagenicity, and teratogenicity. Vibrational spectroscopy techniques, especially Raman spectroscopy and surface-enhanced Raman spectroscopy (SERS), can be potentially used as an alternative technique to liquid and gas chromatography inPAH analysis. However, there is limited information on the intrinsic Raman and SERS fingerprints of PAHs. In this study, we haveacquired the Raman and SERS spectra of seven PAH compounds and compared their experimental spectra with theoretical Ramanspectra calculated by density function theory (DFT). The vibrational modes corresponding to the Raman peaks have also beenassigned using DFT. Characteristic Raman and SERS peaks have been identified for five PAH compounds, and the limits of detec-tion were estimated. Such information could be useful for developing SERS assays for simple and rapid PAH identification.Copyright © 2014 John Wiley & Sons, Ltd.

Additional supporting information may be found in the online version of this article at the publisher’s web site.

Keywords: polycyclic aromatic hydrocarbons; surface-enhanced Raman spectroscopy; silver nanorod arrays; density function theory

Introduction

Polycyclic aromatic hydrocarbons (PAHs) are a group of com-pounds with fused aromatic rings but do not contain heteroatomsor substitution groups. PAHs are formed during the incompletecombustion of fossil fuels and other carbon-containing fuels suchas wood and charcoal at high temperatures (500–700 °C), althoughformation can also occur at lower temperatures (100–150 °C) over aperiod on the geological timescale.[1] PAHs can most frequently befound in soil and sediments for their lipophilic nature but are alsoconsidered widespread organic pollutants in the atmosphere. Hu-man exposure to PAHs is mainly through inhalation of pollutedair and ingestion of contaminated food, particularly those preparedat high temperatures (e.g. smoked foods) and seafood during oilspills. Once entering the human body, PAHs could act as carcino-gens or carcinogenic synergists. Some PAHs also bind to geneticmaterials and exhibit mutagenic and teratogenic effects.[2]

Traditionally, detection of PAHs has mostly been relying on chro-matography techniques. The current gold standards for PAH identi-fication are liquid chromatography using fluorescence detectors(LC/FLDs) or gas chromatography coupled with mass spectrometry(GC/MS).[1,3] The detection limits could typically reach sub parts perbillion levels. However, lengthy and laborious sample preparation isoften required in GC and LC, especially with detection from high fatcontent matrices.[3] More recently, alternative techniques havebeen proposed for PAH identification, including a capillary zoneelectrophoresis method,[4] and optical and spectroscopic methodssuch as surface-enhanced Raman spectroscopy (SERS).[5–8]

SERS takes advantage of the enhanced electromagnetic fieldnear nanostructured metal surfaces (i.e. SERS substrates) to en-hance the Raman scattering signal of target analytes, thereby pro-viding molecular fingerprints at trace levels.[9–11] In the literature,

the spectra of several PAH compounds (e.g. naphthene, anthra-cene, fluorene, pyrene, phenanthrene, tetracene, and chrysene)have been documented in the 1970s and 1980s using conventionalRaman, Fourier transform Raman, coherent anti-Stokes Raman, andresonance Raman techniques.[12,13] However, direct SERS detectionof PAHs is limited because of the poor adsorption of PAHs onto theSERS active substrates. As a result, a significant portion of researcheffort has been directed towards functionalizing the SERS sub-strates for improved PAH adsorption.[14–21] For example, SERS hotspots were created using viologens,[21] alkanethiols,[16] humicacids,[20] calixarenes,[22] pNIPAM,[23] and C18,[24] so that the PAHsadsorption rates on the modified metal surface were improvedand lower detection limits were achieved. However, during sub-strate surface modification, the intrinsic SERS fingerprints of PAHswere distorted. On the other hand, themolecular vibrational modesassociated with the PAH Raman or SERS peaks have not been wellassigned in previous studies.

In order to reveal the inherent PAH Raman and SERS fingerprintsand provide a foundation for future PAH analysis and detection, wehave obtained the Raman and SERS spectra of seven PAH

* Correspondence to: Jing Chen, Department of Food Science and Technology,University of Georgia, Athens, GA 30602, USA.E-mail: jingchen@uga.edu

a Department of Food Science and Technology, University of Georgia, Athens, GA,30602, USA

b Nanoscale Science and Engineering Center, University of Georgia, Athens, GA,30602, USA

c Department of Physics and Astronomy, University of Georgia, Athens, GA, 30602,USA

J. Raman Spectrosc. 2015, 46, 64–69 Copyright © 2014 John Wiley & Sons, Ltd.

Research articleReceived: 19 August 2014 Revised: 14 October 2014 Accepted: 14 October 2014 Published online in Wiley Online Library: 14 November 2014

(wileyonlinelibrary.com) DOI 10.1002/jrs.4612

64

compounds and compared the experimental data with the theoret-ical Raman spectra predicted by density function theory (DFT). De-tailed band assignments have been provided. The PAH compoundshave been differentiated based on their specific peaks as well as prin-ciple component analysis (PCA), and the SERS detection limits havealso been determined on silver nanorod array (AgNR) substrates.

Materials and methods

Materials

PAHs used in this study, as shown in Table 1, were all obtained fromSigma-Aldrich (St. Louis, MO). Methanol was purchased from J. T.Baker (Phillipsburg, NJ). Silver (99.999%) and titanium (99.995%)were acquired from Kurt L. Lesker (Clairton, PA).

Density function theory (DFT) calculation

The theoretical Raman spectra of seven PAH compounds were cal-culated using the Gaussian 03WDFT package. The DFT calculationswere based on Becke’s three-parameter exchange function (B3)[25]

with the dynamic correlation function of Lee, Yang, and Parr(LYP)[26]. The molecular geometries of the PAHs were optimizedusing the hybrid B3 (exchange) and the LYP (correlation) function(B3LYP) in conjunction with a modest 6-311g** basis set.[27]

Fabrication of AgNR substrates

SERS-active silver nanorod (AgNR) array substrates were fabricatedusing oblique angle deposition (OAD) in a custom-built electronbeam evaporator as described previously.[28,29] Briefly, glass slideswere cleaned with Piranha solution (80% sulfuric acid, 20% hydro-gen peroxide), rinsed with deionized water, dried with nitrogen,and loaded into the deposition chamber above the sourcematerial.

Under high vacuum (<10!6 Torr), 20 nm of titanium and 200nm ofsilver were deposited onto the glass slides at a normal incidence an-gle at a deposition rate of 0.2 and 0.3 nm/s, respectively. Then, thesubstrate surface normal was rotated to 86° with respect to the in-cident vapor direction, and silver continued to be deposited at arate of 0.3 nm/s. The last OAD step yielded a film of aligned nano-rods ~900nm in length, ~100nm in rod diameter, with a tilting an-gle of approximately 73° with respect to the substrate normal.[28,30]

To remove the organic contaminants accumulated during fabrica-tion and storage, before SERS measurements, the as-depositedAgNR substrates were cleaned for 2min in a custom-built induc-tively coupled radio frequency plasma chamber, which was oper-ated at 30W under a constant flow of ultra-pure argon with achamber pressure of ~600mTorr.[31]

Bulk Raman spectra

The Raman spectra of five PAHs (ACP, BaA, BaP, F, and P) were col-lected directly from the powder using a portable Raman analyzer,Enwave ProRaman 785A2 (Enwave Optronics, Irvine, CA) equippedwith a 785nm diode laser at a power of 60mW and a spectral ac-quisition time of 10 s.

SERS spectra

For each SERS measurement, 0.1μl aliquots of PAHs at 200μg/ml inmethanol were applied to the AgNR substrates and dried under am-bient conditions, and spectrawere collected through a 10× objectivelens at an excitation power of 60mW and a collection time of 10 s.

Limits of detection (LODs)

A series of PAH dilutions were prepared in methanol, with the finalconcentration ranging from 50 to 1000ng/ml. After the dilutions

Table 1. PAH compounds used in this study and their corresponding SERS limits of detection (LODs) and lowest detectable mass (LDMs)

Chemical name Abbreviation Molecular structure LOD (μg/ml) LDM (g)

Acenaphthene ACP >1000 >1.1× 10!10

Acenaphthylene ACY >1000 >1.1× 10!10

Anthracene ANT 50 5.7× 10!12

Benz(a)anthracene BaA 100 1.1× 10!11

Benzo(a)pyrene BaP 50 5.7× 10!12

Fluorene F 100 1.1× 10!11

Pyrene P 10 1.1× 10!12

Raman, SERS, and theoretically calculated spectra of polycyclic aromatic hydrocarbons

J. Raman Spectrosc. 2015, 46, 64–69 Copyright © 2014 John Wiley & Sons, Ltd. wileyonlinelibrary.com/journal/jrs

65

were applied to the AgNRs and dried, SERS spectra were acquiredat 60mW, 10 s using a 10× objective lens. At least three spectrawere collected from randomly selected locations on the samesubstrate, and substrates fabricated from two independent batcheswere used to account for batch-to-batch variations. The lower LODwas determined using the 3σmethod,[32] in which the intensities ofcharacteristic SERS peaks were compared with a threshold valuedetermined by three times standard deviation of the spectral inten-sity fluctuation at a featureless spectral region (1750–1800 cm!1).

Principle component analysis (PCA)

To differentiate between selected PAHs, PCA was performedusing PLS_Toolbox (Eigenvector, Wenatchee, WA) running underthe Matlab 2011b (Mathworks, Denver, CO) environment. Thespectra were derivatized, normalized, and mean centered prior toPCA analysis.

Results and discussion

DFT-calculated Raman spectra of PAHs

The DFT-calculated Raman spectra of seven PAH compounds be-tween 300 and 1800 cm!1 are shown in Fig. 1, with the correspond-ing 3D molecular structures and vibrational mode assignmentslisted in Fig. S1 and Tables S1–S7 (Supporting Information).According to the DFT prediction, an abundance of Raman peaksappear near 1200–1600 cm!1. Several distinct peaks are also foundat lower wavenumbers (300–1000 cm!1) with lower intensities.The peaks in the low-wavenumber region are mainly attributedto the C–C bending modes, while those appearing at the

mid-wavenumber regions tend to result from the C–H bendingmodes. Because of the fused aromatic ring structure, all the carbonatoms in the PAH molecules are coplanar (refer to Fig. S1–S3 inSupporting Information); hence, it is not surprising that themajorityof these bending modes are restrained within this plane. Neverthe-less, out-of-plane C–C and C–H bending modes that contribute tovery intense Raman peaks are also identified.

It is worth noting that the vibrational mode assignments onlyconvey the most dominant oscillations, but in most cases, the pre-dicted peaks are rather a result of the collective vibrations of multi-ple or all of the atoms and bonds within the molecule. Influencedby other oscillations within the same molecule, peaks of similar vi-brational modes are often observed at different wavenumbers fordifferent compounds. This leads to a unique Raman fingerprintfor each PAH compound as shown in Fig. 1.

In addition, DFT can also estimate the magnitude of the pre-dicted vibrational modes, as expressed by their Raman activity(Fig. 1). It appears that the C–H bending modes contribute to moreintense Raman peaks near the mid-wavenumber range, as a resultof the higher polarizability of the C–H bending modes comparedwith the C–C bendingmodes. Consequently, the PAHs with a largernumber of C–H groups, such as BaP, BaA, and ANT, tend to havehigher predicted Raman activity compared with PAHs with fewerC–H groups (e.g. ACP and ACY). Furthermore, PAHs withmore com-plex and asymmetric structures, such as BaP and BaA, tend to yieldmore Raman peaks that are closely packed within a short spectralrange, while fewer and more resolved peaks are found in simplerand more symmetric molecules like ANT.

Bulk Raman spectra of selected PAHs

The bulk Raman spectra were also obtained from five PAHs thatwere supplied in the powder form (ACP, BaA, BaP, F, and P). Asshown in Fig. 2, sharp and well-resolved peaks can be identifiedin the spectra of ACP, F, and P. The Raman spectrum of P obtainedin this study is consistent with those acquired in previous reports[13]

with minor differences in relative peak intensities, which is likely tobe caused by the use of different excitation wavelengths. Overall,despite the structural similarities, each PAH displays its distinctRaman peaks. As discussed in the previous section, the peak shiftsare mainly caused by influence of other vibrational modes withinthe conjugated ring system.

Because non-resonance Raman signals are generally very weak,and fluorescence has a longer excited state lifetime compared withRaman scattering, Raman signals can be easily overwhelmed byfluorescence signals, causing a problem in identifying Raman peaksfrom pronounced fluorescence backgrounds.[33] Among the testedPAHs, BaP and BaA have the lowest fluorescence excitationenergies,[34] making their Raman signal more susceptible to fluores-cence interference. Hence, less resolved peaks with much lower in-tensities are found in the BaA and BaP spectra as a result of thestrong fluorescent background, which also agrees with the litera-ture findings.[35]

Some differences between the experimentally acquired Ramanspectra and the DFT-calculated Raman spectra are quite obvious.The observed discrepancies are threefold: emergence and absenceof peaks (e.g. the Δν=447, 806, and 896 cm!1 peaks in the DFTspectrum of F are absent in the corresponding Raman spectra),shifts in peak position (e.g. the Δν=507, 815, and 1074 cm!1 peaksin the DFT spectrum of P have shifted to 503, 805, and 1064 cm!1,respectively), and changes in Raman signal intensity (e.g. theΔν=1021 cm!1 peak in the DFT spectrum of F is extremely weak,Figure 1. DFT-calculated Raman spectra of seven PAH compounds.

J. Chen, Y.-W. and Y. Zhao

wileyonlinelibrary.com/journal/jrs Copyright © 2014 John Wiley & Sons, Ltd. J. Raman Spectrosc. 2015, 46, 64–69

66

but in the experimental Raman spectrum, the peak atΔν=1019 cm!1 is prominent). In general, fewer peaks are observedin the experimental spectra compared with the DFT Raman spectra,as indicated in Tables S1–S7 (Supporting Information) (note: Forbrevity of the tables, Raman peaks with low-predicted intensitiesare not listed). This is because the DFT algorithm has an inherenttendency to overestimate the vibrational modes.[36] The probabili-ties for the vibrational modes are not all equal, and some predictedoscillations only contribute negligibly to the overall Raman scatter-ing. Thus, the signal of these weak vibrational modes is difficult tobe picked up by current Raman techniques. On the other hand, ad-jacent peaks tend to merge into single peaks in the experimentalspectra because of the limited spectral resolution provided by theinstrument. Absence of certain peaks in the experimental spectrais rare but also observed, for example, the Δν=1646 cm!1 peak inthe DFT spectrum of F was not present in the corresponding exper-imental spectrum. A reasonable explanation is that because DFT es-timation only presents the vibrational modes for the ‘optimized’structure (i.e. the molecular configuration that results in the lowestoverall energy state), the vibrational modes that are not in thefavored configuration tend to be underestimated or ignored.Moreover, DFT has a limitation in its ability to optimize molecularconfigurations, and the results are also dependent on the input pa-rameters and functions used. Therefore, the optimization results donot necessarily represent the true configuration of the moleculesunder experimental conditions. Shifts in peak position are anothercommonly observed phenomenon with DFT calculation as a resultof the aforementioned reasons. The largest discrepancy betweenthe theoretical and experimental data is the peak intensity. In pre-vious studies, it is reported that DFT tends to overestimate certain

vibrational modes while underestimating others.[36] For example,the experimental spectra do not agree with the DFT prediction thatpeak intensities near the 1200–1600 cm!1 regions are higher thanother spectral regions; instead, an even distribution of peak intensi-ties across the entire scanned region is observed.

SERS spectra of selected PAHs

The obtained SERS spectra of seven PAHs are shown in Fig. 3.Although SERS peaks are identified in all PAH spectra, those ofACP and ACY are strikingly similar to the negative control (metha-nol on blank substrates) peaks. Distinguishable peaks are observedin the control spectra even after Ar+ plasma cleaning, indicatingthat a small portion of organic residues still persisted on the sub-strate surface. As methanol is applied to the substrate and thenevaporates, the nanorods undergo slight deformation in responseto the surface tension of the evaporating solvent, which leads todecreased distance between the rod tips.[37] This so-called‘nanobundling’ effect is known to facilitate the formation of SERShot spots, so that the Raman signal of the molecules near the hotspots can be dramatically enhanced. In the methanol control sam-ple, although no PAH is to remain on the substrate surface after sol-vent evaporation, the surface contaminant residues can stillproduce pronounced SERS signal that interferes with target detec-tion. For molecules with large Raman cross sections (e.g. Raman re-porters) and high adsorption rates, signal from the target can oftenoutcompete this interference, so negligible contamination peaksare observed in the SERS spectra. For most analytes, signals fromboth the target and contaminant residues can be found in thespectra. A typical example is theΔν=690 cm!1 contamination peak

Figure 3. SERS spectra of seven PAH compounds on the AgNR substratesand the solvent (methanol) control. The spectra have been normalizedand vertically offset for clarity.

Figure 2. Bulk Raman spectra of BaA, BaP, F, ACP, and P after vectornormalization. The intensity of BaA and BaP was multiplied by 10 and 5times, respectively, and the spectra were vertically offset for clarity.

Raman, SERS, and theoretically calculated spectra of polycyclic aromatic hydrocarbons

J. Raman Spectrosc. 2015, 46, 64–69 Copyright © 2014 John Wiley & Sons, Ltd. wileyonlinelibrary.com/journal/jrs

67

found in all the SERS spectra in Fig. 3, regardless of the presence ofdistinct PAH peaks. The interference is particularly detrimental tothe detection of molecules with relatively small Raman cross sec-tions, as in the cases of ACP and ACY. As indicated in Fig. 1, theDFT-calculated spectra of ACP and ACY showweak Raman activitiescompared with other PAHs. This is in concert with the observationon the SERS spectra of ACP and ACY, in which no obvious PAH sig-nal could be identified, as they are most susceptible to the interfer-ence from contamination.Distinct target peaks can be found in the SERS spectra of all other

PAHs. Tables S1–S7 (Supporting Information) providemore detailedinformation on the observed peaks and corresponding vibrationalmodes. The spectra are all normalized to their vector lengths. EachPAH exhibits its characteristic features in the SERS spectrum. Somepeaks are shared by two or more PAHs, indicative of similar vibra-tional modes. For instance, the Δν=~1406 cm!1 peak found inboth ANT and P is attributed to C–H and C–C in-plane bendingmodes (Tables S3 and S7 in Supporting Information). Althoughthe compositions of all PAHs are very similar (all consisting of onlycarbon and hydrogen atoms at similar ratios), the fingerprint region(300–1600 cm!1) contributed mostly by C–C stretching and bend-ing vibrations can still provide invaluable information in differenti-ating the PAHs. Using the SERS spectra acquired on the AgNRs,differentiation can be achieved by identifying characteristic peaks(except ACP and ACY) without the aid of multivariate statisticalanalyses. Nevertheless, PCA was performed to provide bettervisualization of the differences between PAHs. According to Fig.S2a (Supporting Information), BaA, BaP, F, and P can be groupedinto four distinct clusters, whereas ACP and the solvent controlare closely grouped as one cluster. ACY and ANT showmore differ-ence from the control. However, because of the relatively large var-iations within these two groups, they may also be consideredidentical to themethanol control sample. The loading scores shownin Fig. S2b (Supporting Information) indicate that themajor spectralvariations originate from a broad range of spectral region, and themost variation is contributed by peaks at Δν=717–728, 1007–1029,1144–1160, 1230, and 1368–1397 cm!1, which correspond to spe-cific SERS peaks of F, BaA, P, and BaP.Overall, the SERS spectra of ANT, BaP, P, and F agree well with

previous reports. However, some obvious discrepancies have beenobserved. For example, the 1419 and 1634 cm!1 peaks of BaP re-ported by Bao et al.[6] have shifted to 1428 and 1619 cm!1, respec-tively. The 415, 742, and 1611 cm!1 of F are found at 429, 729, and1600 cm!1 in this study, and the previously reported 1401 cm!1

peak of F and 492 cm!1 peak of ANT cannot be identified in our ex-periments. However, such disagreements are expected, consider-ing that different types of SERS substrates and detection formatswere used. Because SERS substrates are involved, the interstudy dis-crepancies of the SERS data can be much greater than those foundin the bulk Raman spectra.A glimpse at the SERS spectra of BaA, BaP, F, and P suggests that

spectral deviation from their bulk Raman spectra are a commonphenomenon. Not surprisingly, the SERS spectra exhibit morepeaks compared with the bulk Raman spectra. This is because theweak vibrational modes that cannot be detected by Ramanspectroscopy may be revealed because of the signal enhancementprovided by SERS. Meanwhile, the interaction between the PAHmolecule and the silver substrate can also introduce new vibra-tional modes or alter the molecular configuration that leads tochanges in the detected SERS signal. Interaction between silverand the analyte molecule can also impact the SERS spectra in theform of peak shifting, and if a vibrational mode is suppressed

because of adsorption to silver, the corresponding Raman peakcan also disappear in the SERS spectra.[38] As discussed before, itis also possible for the interfering contamination peaks to emergein the PAH spectra, if the peaks do not already exist in the spectra.

Consistent with the DFT calculation, the SERS peak intensities(particularly those of BaP and BaA) appear to be higher in the spec-tral region between 1000 and 1600 cm!1. This suggests that thehigher wavenumber modes might exert greater impact on thespectra. However, this phenomenon is confounded by the fact thatthe Raman instrument used in this study is reported to be ultra-sensitive at this spectral range.

Overall, the SERS and bulk Raman spectra of PAHs show high re-semblance despite some anticipated discrepancies. The majority ofthe observed PAH SERS peaks can be assigned to one or morevibrational modes based on DFT calculation.[38]

Limits of detection (LODs) of PAHs in methanol solution

Concentration-dependent PAH spectra and correspondingcharacteristic peak intensities are shown in Figs. S3 through S9(Supporting Information). In general, at low PAH concentrations,peaks from surface contaminants dominate the SERS spectra,completely masking signal from the PAHs. As the concentration in-creases to above a certain threshold (i.e. the LOD), weak PAH peaksbegin to emerge. The LODs for ANT, BaA, BaP, F, and P are deter-mined to be 50, 100, 50, 100, and 10μg/ml, respectively. The LODsfor ACP and ACY are above the highest available concentration as aresult of their small scattering cross sections. For other PAHs, theLODs are also high, which is consistent with previous literature find-ings where low sensitivity is reported when PAHs were directly de-tected on SERS-active metal surfaces. The low detection sensitivityhas been widely attributed to the low affinity of PAHs to metallicsurfaces.[15,39] Lower LODs are only achieved on chemically func-tionalized substrates or on substrates with ultra-high enhancementfactors enabled by specific geometries. Surface functionalizationenhances the detection sensitivity either by improving the adsorp-tion rate of PAHs or by providing interparticle junctions that act asSERS hot spots. Hexanethiol,[40] decanethiol,[16] calix[4]arene,[19]

cyclodextrin derivatives,[15] per-6-deoxy-(6-thio)-β-cyclodextrin(CD-SH),[41] and humic acids[17,20] have been reported, with theLODs ranging from 10!5 to 10!6μg/ml. Novel SERS substrates withultra-high enhancement factors, such as Au on TiO2 nanotubearrays[42] and AuNPs on alginate gel,[6] have been proposed inPAH detection as well, and LODs could reach as low as 0.365nMfor BaP (equivalent to 9.2× 10!5μg/ml). However, in these studies,SERS substrates were often soaked in the PAH solution for a certainperiod of time (e.g. 1 h) to allow the PAHs to partition to the sub-strate surface. This equilibration step negatively impacts the overalldetection rapidity andmay introduce additional errors that can fur-ther compromise assay reproducibility.

In addition, the LODs may also be limited by the extremely smallsample volume (0.1μl) used in this study. The vast majority ofreported low LODs methods involve sampling as much as 10mlof PAH solution. This ‘fishing for target’ strategy helps attract andconcentrate the analyte molecules to the sensing surface, therebyincreasing the apparent LOD. In this study, the analyte solutionswere directly applied to the SERS-active surface without pre-concentration. As a result, the LODs obtained using a smaller testvolume are higher compared with the large volume tests. In fact,the LOD of a detectionmethod is a result ofmultiple factors, includ-ing the area on the substrate that liquid samples spread into, thearea of the incident laser spot, and the true mass sensitivity of the

J. Chen, Y.-W. and Y. Zhao

wileyonlinelibrary.com/journal/jrs Copyright © 2014 John Wiley & Sons, Ltd. J. Raman Spectrosc. 2015, 46, 64–69

68

sensing platform. The true mass sensitivity of the SERS platformcan be expressed as the lowest amount of PAHs that could bedetected, that is, lowest detectable mass (LDM) of the analytes, ascalculated by

LDM ¼ CVAA0

(1)

where C is the concentration of the PAH, V is the volume of samplesolution (V=0.1μl), A is the area of the Raman laser spot(A=0.0143mm2), andA0 is the area of the circular spot that the sam-ple has spread into (A0=12.6mm2). According to our experimentalresults, the LDMs for ANT, BaA, BaP, F, and P are ~5.7× 10!12,1.1×10!11, 5.7×10!12, 1.1×10!11, and 1.1×10!12 g, respectively,demonstrating a high inherent sensitivity of the sensing platform.

In the cases of BaP and P, the peak intensities undergo a de-crease when the concentration increases to >500μg/ml (Figs. S7and S9 in Supporting Information). Based on the AgNR geometry,the actual surface area of the 0.1μl sample spot is approximately6.1× 1013 nm2, and the number of molecules within the spot is2.4× 1015–3× 1015. Therefore, the average surface coverage of theBaP or P molecules is between 39 and 49 molecules/nm2, whichis very close to what is sufficient to form a monolayer if a uniformcoverage is assumed. Hence, the surface will be packed by BaPand P at ~500μg/ml. Because previous studies have reported a re-duction in SERS intensity as a result of adsorbate excited statequenching,[43] it is not surprising that the peak intensities of BaPand P have experienced a decline at 1000μg/ml. Interestingly, suchsignal reduction is not observed in other PAHs at similar concentra-tions. This is mainly because the varying adsorption rates andRaman cross sections among the PAHs can lead to different thresh-olds for this signal saturation effect.

Conclusions

In this study, we have characterized seven PAH compounds withtheir Raman and SERS Raman spectra and compared the experi-mentally acquired spectra with those predicted by DFT. DFT calcu-lation facilitated the spectral band assignment and also providedinsights into the observed differences in SERS intensity. Characteris-tic SERS peaks were identified for each PAH compound except ACPand ACY, whose signal intensities were below the detection thresh-old. The tested PAHs could be differentiated based upon character-istic SERS peaks and confirmed by PCA analysis of the SERS spectra.The LDMs indicate that inherent sensitivity of the SERS platformwas high. However, the determined LODs of PAHs were relativelyhigh as a result of poor molecular adsorption rates and smallsample volume. Future work will be directed to SERS substrateimprovement and developing sample preparation protocols to im-prove the LODs of PAHs from real samples.

Acknowledgements

This work was supported by National Science Foundation undercontract number CBET-1064228 and UGA College of Agriculturaland Environmental Sciences Experimental Station. The authorsthank Dr. Xibo Li for setting up DFT calculations.

References[1] P. Plaza-Bolanos, A. G. Frenich, J. L. M. Vidal, J. Chromatogr. A 2010,

1217, 6303.

[2] Scientific Committee on Food, European Commission, Opinion of thescientific committee on food on the risks to human health of polycyclicaromatic hydrocarbons in food, European Commission (EC), Brussels,Belgium, 2002.

[3] S. Moret, L. S. Conte, J. Chromatogr. A 2000, 882, 245.[4] L. Ferey, N. Delaunay, D. N. Rutledge, C. B. Y. Cordella, H. This, A. Huertas,

Y. Raoul, P. Gareil, Talanta 2005.[5] AOAC INTERNATIONAL, Official Methods of Analysis of AOAC INTERNA-

TIONAL, Official Method 973.30, Gaithersburg, 2005.[6] L. Bao, P. T. Sheng, J. Li, S. Y.Wu, Q. Y. Cai, S. Z. Yao, Analyst2012, 137, 4010.[7] H. Schmidt, N. B. Ha, J. Pfannkuche, H. Amann, H. D. Kronfeldt,

G. Kowalewska, Mar. Pollut. Bull. 2004, 49, 229.[8] S. D. Stewart, P. M. Fredericks, J. Raman Spectros. 1995, 26, 629.[9] M. Fleischmann, P. J. Hendra, A. J.McQuillan,Chem. Phys. Lett.1974, 26, 163.

[10] K. Kneipp, H. Kneipp, S. Abdali, R. W. Berg, H. Bohr, Spectros. Int. J. 2004,18, 433.

[11] S. M. Nie, S. R. Emery, Science 1997, 275, 1102.[12] S. A. Asher, Anal. Chem. 1984, 56, 720.[13] W. Maddams, I. Royaud, Spectrochim. Acta Mol. Spectros. 1990, 46, 309.[14] Y. F. Xie, X. Wang, X. X. Han, W. Song, W. D. Ruan, J. Q. Liu, B. Zhao,

Y. Ozaki, J. Raman Spectros. 2011, 42, 945.[15] Y. F. Xie, X. Wang, X. X. Han, X. X. Xue, W. Ji, Z. H. Qi, J. Q. Liu, B. Zhao,

Y. Ozaki, Analyst 2010, 135, 1389.[16] C. L. Jones, K. C. Bantz, C. L. Haynes, Anal. Bioanal. Chem. 2009, 394, 303.[17] L. L. Qu, Y. T. Li, D. W. Li, J. Q. Xue, J. S. Fossey, Y. T. Long, Analyst 2013,

138, 1523.[18] L. Guerrini, J. V. Garcia-Ramos, C. Domingo, S. Sanchez-Cortes, J. Phys.

Chem. C 2008, 112, 7527.[19] L. Guerrini, J. V. Garcia-Ramos, C. Domingo, S. Sanchez-Cortes, Anal.

Chem. 2009, 81, 1418.[20] P. Leyton, I. Cordova, P. A. Lizama-Vergara, J. S. Gomez-Jeria, A. E. Aliaga,

M. M. Campos-Vallette, E. Clavijo, J. V. Garcia-Ramos, S. Sanchez-Cortes,Vib. Spectros. 2008, 46, 77.

[21] A. I. Alajtal, H. G. M. Edwards, I. J. Scowen, J. Raman Spectros. 2011, 42, 179.[22] P. Leyton, S. Sanchez-Cortes, J. V. Garcia-Ramos, C. Domingo,

M. Campos-Vallette, C. Saitz, R. E. Clavijo, J. Phys. Chem. B 2004, 108,17484.

[23] M. Mueller, M. Tebbe, D. V. Andreeva, M. Karg, R. A. Alvarez-Puebla,N. Pazos-Perez, A. Fery, Langmuir 2012, 28, 9168.

[24] L. G. Olson, R. H. Uibel, J. M. Harris, Appl. Spectrosc. 2004, 58, 1394.[25] A. D. Becke, J. Chem. Phys. 1993, 98, 5648.[26] C. Lee, W. Yang, R. G. Parr, Phys. Rev. B Condens. Matter 1988, 37, 785.[27] X. Wu, S. Gao, J.-S. Wang, H. Wang, Y.-W. Huang, Y. Zhao, Analyst 2012,

137, 4226.[28] Y. J. Liu, H. Y. Chu, Y. P. Zhao, J. Phys. Chem. C 2010, 114, 8176.[29] J. X. Fu, A. Collins, Y. P. Zhao, J. Phys. Chem. C 2008, 112, 16784.[30] J. D. Driskell, S. Shanmukh, Y. Liu, S. B. Chaney, X. J. Tang, Y. P. Zhao,

R. A. Dluhy, J. Phys. Chem. C 2008, 112, 895.[31] P. Negri, N. E. Marotta, L. A. Bottomley, R. A. Dluhy, Appl. Spectrosc.

2011, 65, 66.[32] W. B. Knighton, E. P. Grimsrud, Anal. Chem. 1983, 55, 713.[33] P. Matousek, M. Towrie, A. Parker, J. Raman Spectros. 2002, 33, 238.[34] G. I. Romanovskaya, N. A. Lebedeva, J. Anal. Chem. 1993, 48, 1400.[35] L. Colangeli, V. Mennella, G. Baratta, E. Bussoletti, G. Strazzulla,

Astrophys. J. 1992, 396, 369.[36] X. M. Wu, S. M. Gao, J. S. Wang, H. Y. Wang, Y. W. Huang, Y. P. Zhao,

Analyst 2012, 137, 4226.[37] J. L. Abell, J. D. Driskell, Y. P. Zhao, Chem. Commun. 2014, 50, 106.[38] E. Le Ru, P. Etchegoin, Principles of surface-enhanced raman spectroscopy

and related plasmonic effects, Elsevier, Amsterdam, Boston, 2009.[39] Y. H. Kwon, K. Sowoidnich, H. Schmidt, H. D. Kronfeldt, J. Raman

Spectros. 2012, 43, 1003.[40] X. Jiang, Y. Lai, M. Yang, H. Yang,W. Jiang, J. Zhan,Analyst2012, 137, 3995.[41] Z. Y.Wang, H. Wu, C. L. Wang, S. H. Xu, Y. P. Cui, J. Mater. Chem. 2011, 21,

4307.[42] P. T. Sheng, S. Y. Wu, L. Bao, X. Wang, Z. Chen, Q. Y. Cai, New J. Chem.

2012, 36, 2501.[43] E. Zeman, K. Carron, G. Schatz, R. Van Duyne, J. Chem. Phys. 1987, 87,

4189.

Supporting information

Additional supporting information may be found in the online ver-sion of this article at the publisher’s web site.

Raman, SERS, and theoretically calculated spectra of polycyclic aromatic hydrocarbons

J. Raman Spectrosc. 2015, 46, 64–69 Copyright © 2014 John Wiley & Sons, Ltd. wileyonlinelibrary.com/journal/jrs

69