A sensor array for the discrimination of polycyclic aromatic hydrocarbons using conjugated polymers and the inner filter effect

The inner filter effect and multivariate array sensing using conjugated polymers are combined for the detection and challenging discrimination of closely related polycyclic aromatic hydrocarbons.


Introduction
Polycyclic aromatic hydrocarbons (PAHs) are a ubiquitous and prominent class of organic compounds comprised of fused aromatic rings containing only carbon and hydrogen. Well over 120 years of research has intricately connected these compounds with their natural and anthropogenic origins. 1 While natural sources include those such as fossil fuels, open burning, and volcanic activity; pyrogenic and petrogenic sources such as the combustion of these fossil fuels, industrial manufacturing, and dispersed sources (i.e. automotive emission, residential heating, food preparation, etc.) predominate. 2 These activities result in the production of PAHs that are pervasive environmental pollutants with toxic, mutagenic, and carcinogenic properties. 3 For these reasons, research efforts remain unabated toward the detection and discrimination of these compounds; however, this continues to represent a major technological hurdle owing to their uncharged and nonpolar nature, similar and relatively featureless structures, lack of heteroatoms or substituents, and low active concentrations. 4 Current methods for PAH detection necessitate sample collection, transport, protracted multi-stage extraction, preconcentration, and separation procedures. Subsequently, specialized analytical instrumentation sensitive to subtle chemical differences and employing multiple detectors is applied for analysis. Examples include high-performance liquid chromatography (HPLC) coupled to uorescence or ultraviolet detection, 5 gas chromatography coupled to mass spectrometry (GC-MS) or ame ionization detection (FID), 6 and capillary electrophoresis coupled to uorescence or ultraviolet detection. 7 While highly sensitive, these methods require trained personnel, long analysis time, may lead to low analytical precision and analyte bias, are intended for a specic PAH type or property, and cannot discriminate between closely related compounds such as benzo [a]pyrene and dibenz [a,h]anthracene; isomers differing only in the connectivity of an aromatic ring and classied as carcinogens by numerous international agencies. 3 Immunoassay-based approaches utilize antibodyantigen interactions, which demonstrate improved sensitivities in platforms that offer reduced costs, rapid detection, portability, and are amenable to high-throughput. 4,8 Despite these advantages, similarities in the molecular and electronic structure of PAHs preclude the development of specic antibodies, leading to cross-reactivity and the inability to differentiate between closely related compounds. 9 In a similar manner, various nanomaterial-based sensor platforms utilizing quantum dots, graphene, carbon nanotubes, and polymer composites have been advanced as alternative approaches for the detection of PAHs, but are fundamentally limited in their discriminatory power in multi-analyte detection. 4c,10 As such, important questions regarding source attribution, transport and fate in the environment, bioaccumulation in the food chain, and the toxicology of these ubiquitous pollutants remain largely unresolved. 11 Consequently, there remains an urgent need for the qualitative and quantitative assessment of complex mixtures of PAHs in a rapid, portable, high-throughput and cost-effective manner. Fluorescent chemosensory devices remain a leading signal transduction method owing to their high sensitivity, ease of operation, and broad applicability. 12 When compared to small molecule uorophores, signicant gains in sensitivity are achieved using conjugated polymers (CPs) since their delocalized electronic structure leads to large extinction coefficients, strong uorescence emission, efficient excited state (or exciton) migration, and collective properties that are sensitive to minor perturbations. 13 In general, uorescent sensors based on CPs operate through analyte-induced energy transfer, various aggregation phenomena, or conformational rearrangements that serve to manipulate mobile excitons and modulate the uorescence in the form of spectral shis, quenching, or unquenching of the emission. 12, 14 These mechanisms are distance-dependent and require strong CPanalyte interactions that are typically facilitated through the integration of molecular recognition elements (receptors) within or extended from the CP backbone. 15 Nonspecic receptors for PAHs based on cation/p, relatively weak p/p and C-H/p interactions have been reported, but only show affinity toward particular PAHs. 16 Supramolecular host-guest systems based on phenylenebridged 4,4 0 -bipyridinium cations have demonstrated signicant enhancements in binding affinities for PAHs, however, complex molecular topologies are required to form selective inclusion complexes and tailored chemistries for distinct PAHs are currently unavailable. 17 Furthermore, the challenge of electronically coupling these analyte-receptor interactions into transducible optical responses further complicates the development of optical sensing platforms capable of proling the distinct molecular signatures of diverse PAHs in complex mixtures.
More recently, array-based sensing has been used to prole combinations of structurally and chemically similar analytes through multivariate pattern recognition. 18 Subtle structural differences between nonspecic CP-based sensing elements allow for differential interactions with analytes that establish unique and identifying optical response patterns. 19 This creates a "chemical ngerprint" which can be used to discriminate similar compounds using linear discriminant analysis (LDA) and principal component analysis (PCA), pattern recognition algorithms which highlight and summarize distinguishing features in large data sets to provide information leading to chemical differentiation. 20 Still, these methods require spatially distinct sensor units, each with its own recognition element, to build a diagnostic pattern that can be used to rapidly identify individual analytes. The inner lter effect (IFE) results from the absorption of light by a chromophore in solution, preventing photons from reaching a uorophore, creating an observed decrease in uorescence emission. 21 Here, we demonstrate that the IFE in combination with CP-based array sensing offers a straightforward approach for the quantitative detection and qualitative discrimination of PAHs. While previous reports have demonstrated the utility of the IFE for the detection of picric acid and Sudan dyes using CPs, 22 we report the use of differential quenching and pattern recognition to discriminate chemically and structurally similar PAHs, which could not be achieved using the IFE or CPs independently. To obtain the desired differential interactions required for an effective array sensor, we synthesized a series of similar but structurally distinct uorescent CPs based on uorene copolymer scaffolds with 2-phenylbenzimidazole optical modiers. These CPs provide spectral overlap in regions of maximum absorption for many PAHs allowing for an IFE, with the PAH acting as optically dense absorbers. The reported system thus takes advantage of the intrinsic optical properties of individual PAHs, circumventing the need for tailored host-guest interactions. The unique response of each polymer allowed for the discrimination of 16 PAHs listed by the EPA as priority pollutants that are hazardous to human health.

Polymer design, synthesis, and optical characterization
Our studies began with the synthesis of poly [2,7-(9,9- (Fig. 1a) using a Suzuki cross-coupling polymerization. 23 We utilized the same approach for the synthesis of P1-P4 and P6, while P5 was synthesized using a microwave mediated Stille cross-coupling polymerization (Fig. 1a). These approaches resulted in facile access to an array of structurally distinct polymers, that were targeted to maintain solubility, high uorescence emission, and sizes beyond the exciton diffusion length. 24 General protocols regarding monomer and polymer synthesis are included in the Experimental section with full details in the ESI. † P2 exhibits an absorption maximum (l max ) centered at 374 nm, providing spectral overlap in regions of maximum absorption for many PAHs ( Fig. 1 and 2). Each PAH displays a characteristic absorption prole in the 250-500 nm region ( Fig. 2 and S2-S17 †). The tunable nature of these CPs allows for the incorporation of optical modiers which provide greater spectral overlap between the absorption of the polymer and each PAH, enabling efficient uorescence quenching through the IFE. Peripheral 2-phenylbenzimidazole substituents in P2 impart an extra absorption band with l max ¼ 290 nm, which affords the greater spectral overlap required for the detection of PAHs through the IFE (Fig. S18 †). Fig. 1b illustrates the signicant spectral overlap between P2 and anthracene allowing for an IFE, with the PAH acting as a "chemical lter." Upon excitation at 374 nm, P2 exhibits a strong emission between 400-500 nm, with maximum intensity at 415 nm. Addition of anthracene causes apparent quenching of the uorescence through the IFE (Fig. 1c).   2 illustrates the distinct optical proles of the 16 PAH compounds identied as priority PAH pollutants by the US Environmental Protection Agency (EPA). Each PAH demonstrates varying spectral overlap with P2 ( Fig. 2a-c), and distinct wavelength dependencies of the molar absorptivity (3), providing the basis for differential quenching through the IFE (Fig. 2d-f). To obtain the desired diversity of interactions required for an effective array sensor, a series of similar but structurally distinct CPs were synthesized, which demonstrate uorescence quenching by PAHs through the IFE (Fig. 1a). In addition to P2, the polymer array included poly [2,7-(9,9- imidazole))hexyl)-uorene)-alt-2,5,-thiophene] (P5), and poly [2,7-(9,9-di((undec(ethylene glycol) P1-P4 demonstrate a systematic lengthening of the methylene bridge (-CH 2 -) x between the 2-phenylbenzimidazole optical modier and the CP backbone, with x ¼ 3, 4, 6, and 8 units, respectively (Fig. 1a). The structural changes in the polymer series cause subtle but distinct differences in their optical spectra and molar absorptivity (Fig. S1 †). To establish that these slight modications could manifest into discernible responses, several PAHs were titrated into solutions of P1-P4 to study their effect on the uorescence emission of each polymer. Three PAHs, anthracene, acenaphthylene, and pyrene were chosen to represent a range of ring fusion and optical properties (e.g. intrinsic absorption) in the PAH family. The resulting titration proles for P1-P4 are summarized in Fig. 3a, in which quenching of polymer emission is caused by the selected PAHs through the IFE. Small but noticeable differences in quenching were observed between P1-P4 and each PAH, demonstrating that even subtle structural modications affected the spectral response. More dramatic differential responses were shown between the PAHs, which can be explained by the distinct dependence of molar absorptivity for each PAH at a given wavelength (Fig. 3b). At an excitation wavelength (l exc ) of 374 nm, anthracene has the greatest molar extinction coefficient (3 ¼ 0.41 Â 10 4 M À1 cm À1 ) and was the most efficient uorescence quencher of each polymer through the IFE. As a representative example, the detection limit of anthracene using P2 was calculated to be 2.4 mM, demonstrating the low limit of detection (LOD) for the array (Fig. S19 †). A uorene copolymer with a thiophene structural unit in the backbone (P5) was incorporated into the array to provide distinctive quenching behavior from the other copolymers. P5 shows a red-shied absorption (l max ¼ 420 nm) and allows for differential quenching through the IFE when compared to the other copolymers of the array. A uorene-co-phenylene copolymer without 2phenylbenzimidazole optical modiers (P6) and incorporated oligo(ethylene glycol) side chains was synthesized. P6 lacks the extra band in the absorption spectrum (l max ¼ 290 nm) seen in P1-P5, providing another source of differential interaction through the IFE. Minor structural variations between each polymer, in combination with the unique relationship of molar absorptivity and wavelength for each PAH, provide a library of differential responses which can be used for discrimination.

Conjugated polymer-based sensor array for polycyclic aromatic hydrocarbons via pattern recognition
Solutions of P1-P6 in DMF (15 mg L À1 ) were arranged on a 384well plate and exposed to 500 mM solutions of each PAH (Fig. 2g). 25 Each PAH-polymer combination was prepared in 12 replicates and multiple spectroscopic measurements were collected on a microwell plate reader, corresponding to the regions of spectral overlap between the polymers and each PAH, including 21 absorbance measurements from 280-700 nm, and uorescence measurements using the following lter combinations (l exc /l em ): 330/460 nm, 330/485 nm, 330/516 nm, 380/ 460 nm, 380/485 nm, and 380/516 nm. The raw instrumental response pattern for each measurement is tabulated and visually summarized as a heat map in the ESI (Fig. S20-S40 †). Superuous instrumental variables introducing experimental noise were removed from the original data set whose contribution to PAH discrimination was negligible; these included absorbance measurements above 500 nm as PAHs lack absorption in this region. Aer removal of these absorbance measurements, the optical responses of P1-P6 to each PAH were analyzed through linear discriminant analysis (LDA), a wellestablished multivariate patterning algorithm. 26 Complete differentiation of 16 PAHs was observed using the rst two factors obtained from LDA analysis, while retaining 78.0% of the total information content that was present in the raw dataset. Fig. 4a displays the corresponding two-dimensional LDA scores plot. Replicates of the same PAH sample were found to cluster tightly, whereas clusters of replicates from different samples were well-separated. Tight intra-cluster spacing indicates excellent reproducibility, while large intercluster spacing indicates strong discriminatory power of the polymer-based array.
The most important contributors to the differentiation were absorption signals in the range of 420-480 nm, as shown in the corresponding LDA loadings plot (Fig. 4c), which arise from absorption bands displayed only by benzo[k]uoranthene and indeno[1,2,3-cd]pyrene. The unique features of these two PAHs between 420-480 nm are overrepresented in the discrimination    Fig. 4a and are therefore assigned a disproportionally high weighting in the LDA analysis, thus differentiating these two very well from the other 14 PAHs, but providing very little discriminatory power for the other 14, thus drastically reducing the array's effectiveness for analytical applications. An overview of the quality of the information conveyed by each instrumental measurement is presented visually in the ESI (Fig. S42 †). The absorbance measurements for each polymer between 420-480 nm contain little coherent signal and are dominated by noise, relative to the absorption at lower wavelengths (<420 nm) and were therefore removed from the dataset. The LDA analysis was then repeated on this reduced dataset.
Aer removal of the information-poor absorbance measurements between 420-480 nm, subsequent LDA analysis and data reduction provided a better performing differentiation, representing a signicant improvement in the overall discriminatory power of the array (Fig. 5a). The corresponding LDA loadings plot reects a reduced importance of absorption signals to the differentiation, with a more diverse group of variables such as the uorescence measurements acting as the most important contributors to the discrimination (Fig. 5b). Increased reliance on uorescence signals is desirable as the high photoluminescence quantum efficiencies of uorenebased copolymers should offer considerably lower limits of differentiation for PAHs compared to an array relying primarily on the absorption properties of PAHs. Despite the improvement in separation of all 16 PAHs, only half of the total information was retained in the rst two factors (53.4%). By including a third factor, the data could be displayed as a three-dimensional plot, while preserving a larger portion of the total information. As shown in the ESI (Fig. S43 †), the three-dimensional scores plot now retains 74.7% of the total information contained in the original dataset.
The dataset was also analyzed by a similar multivariate technique, principal component analysis (PCA), as PCA provides an unsupervised representation of the variances in a given dataset. The two-dimensional PCA plot is presented in Fig. S44 in the ESI. † PCA analysis also gives complete separation of all 16 PAHs with very small intra-cluster distances, while retaining 70.8% of the total information from the original dataset within the rst two components.

Role of the inner lter effect in polycyclic aromatic hydrocarbon discrimination
The role of P1-P6 in the discrimination was investigated by analyzing a dataset containing optical measurements from only the PAHs, in the absence of polymers. LDA analysis of this dataset showed that complete separation of all 16 PAHs was not achieved (Fig. 6a). Moreover, the corresponding loadings for the rst two factors (Fig. 6b) are, once again, mostly reliant upon the absorbance measurements of the PAHs: not only is this detrimental for the sensitivity of the assay, it also prevents PAHs with similar absorption features such as naphthalene, uorene, acenaphthene, and phenanthrene from being differentiated (Fig. 6b, inset). The modulation of P1-P6 uorescence was therefore found to be critical in the discrimination of all 16 PAHs, as demonstrated by the large contribution of uorescence measurements in the factor loadings (Fig. 5b).
To elucidate the mechanism of uorescence quenching, steady-state absorption, uorescence lifetime, and uorescence anisotropy measurements were performed. As illustrated in Fig. 2, the PAH absorption spectra span 250-480 nm, overlapping with the absorption and emission spectra of P1-P6. The spectral overlap indicates the possibility of uorescence resonance energy transfer (FRET), 27 which may take place in the presence of overlapped emission spectrum of a uorophore (CP) with the absorption spectrum of a quencher (PAH), or the IFE. The uorescence lifetime was measured in the absence and presence of PAH, where P2 and anthracene were chosen as a representative example. It is evident from Fig. 7 that the uorescence lifetime of P2 does not display any signicant change upon the addition of anthracene, which rules out a dynamic quenching mechanism. 22 Furthermore, the fact that the absorption spectrum of a mixture between polymer and PAH is completely additive rules out static quenching due to aggregation and formation of a ground-state complex between P2 and anthracene (Fig. 7, inset). Fluorescence anisotropy experiments are commonly used to investigate molecular interactions between a small uorophore and a macromolecule, where a strong increase in anisotropy indicates a slower rotational diffusion rate of the uorophore due to binding. 27,28 For these experiments, P5 was titrated into a solution of anthracene ([anthracene] ¼ 100 mM in DMF). Polymer P5 was chosen because its emission spectrum is well-separated from that of anthracene, allowing the emission of anthracene and changes in its tumbling rate in solution to be directly observed. However, no increase in uorescence anisotropy was observed (Fig. S45 †), suggesting that no binding occurred between anthracene and P5. When all these results are considered, the IFE is the most probable mechanism responsible for quenching of the polymers.

Conclusions
In summary, we have demonstrated the detection of 16 priority PAHs through a six-membered sensor array. A new set of uorescent CPs was readily prepared, incorporating side chain optical modiers, which were strongly and yet non-selectively affected by the presence of diverse PAHs through an IFE. Multivariate pattern recognition strategies were used to generate two-dimensional score plots, which can operate as effective calibration plots for the differentiation of unknown PAHs using simple, common, and cost-effective instrumentation (UV-vis absorption and uorescence spectroscopies). The reported platform is highly tunable, owing to facile modication of the polymer structure and ease with which optical modiers can be designed and integrated. The discrimination of similar analytes that lack well-dened recognition elements was achieved, widening the scope of the IFE in CPbased sensors. These results lay the groundwork for an extension into qualitative and quantitative detection and discrimination of molecular species with high optical densities that are otherwise inaccessible through energy transfer and aggregation-based mechanisms traditionally utilized in CP-based array sensors.

Spectroscopic methods
All spectra were recorded at ambient temperature, unless otherwise stated. UV-vis absorbance measurements were performed on a Hewlett-Packard 8452a diode array UV-vis spectrophotometer. Benchtop steady-state uorescence measurements were carried out with an ISS PC1 spectrouorimeter. Excitation was carried out using a broad-spectrum high-pressure xenon lamp (CERMAX, 300W). Excitation correction was performed through a rhodamine B quantum counter with a dedicated detector. Detection was through a Hamamatsu red-sensitive PMT. High-aperture Glan-Thompson calcite polarizers were used in the excitation and emission channels to measure steady-state uorescence anisotropy. Experimental temperature (25 C) was controlled by an external circulating water bath.
All optical spectroscopy and binding experiments were performed in DMF. Polymer and PAH stock solutions were prepared separately, then (2 mL) of the polymer solutions were inserted into a quartz cuvette, and absorption and uorescence spectra were collected for P1-P6. Fluorescence emission spectra were collected by exciting the polymer at their most red-shied spectral maximum. Binding titrations were performed by adding aliquots (1-100 mL per addition) of PAH solution to the polymers. Fluorescence decay proles were recorded using a Horiba PPD850 time-correlated single-photon counting (TCSPC) detector with a Fianium WhiteLase SC-400 laser excitation source at a repetition rate of 2 MHz.
Multivariate data was acquired on a BioTek Synergy II multimode microwell plate reader, capable of measuring absorption spectra through a monochromator and steady-state uorescence intensity measurements through a set of bandpass lters. The sample compartment in this instrument was electrically thermostatted to 25 C. Experiments were laid out by hand using Eppendorf Research multichannel pipettors and disposable plastic tips into Aurora microwell plates with clear bottoms for UV absorption and uorescence spectroscopy in a 384-well conguration. The plates were made of non-treated cyclo-olen polymer (COP) with clear at bottoms. Each well was lled with (100 mL) of the sample solution. Plates were read on a multimode plate reader immediately aer preparation. Further details regarding multivariate data analysis are available in the ESI. †

Conflicts of interest
There are no conicts to declare.