Low-temperature time-resolved phosphorescence excitation emission matrices for the analysis of phenanthro-thiophenes in chromatographic fractions of complex environmental extracts

Abstract

The present study investigates the analytical potential of low-temperature photoluminescence spectroscopy for the analysis of seven phenanthrothiophenes with molecular mass 234 g mol⁻¹. The studied PASHs include Phenanthro [1,2-b]thiophene, Phenanthro [2,1-b]thiophene, Phenanthro [2,3-b]thiophene, Phenanthro [3,2-b] thiophene, Phenanthro [3,4-b]thiophene, Phenanthro [4,3-b]thiophene and Phenanthro [9,10-b]thiophene. Excitation and emission spectra recorded from n-alkane solutions at room temperature, 77 K and 4.2 K show phosphorescence emission from all the studied isomers at cryogenic temperatures. The analytical figures of merit obtained under steady state (fluorescence) and time-resolved (phosphorescence) conditions provide limits of detection at the parts-per-billion (ng mL⁻¹) concentration levels. Processing 77 K and 4.2 K phosphorescence data with parallel factor analysis showed to be a robust approach to the determination of phenanthro-thiophenes in complex fluorophore mixtures.

1. Introduction

Polycyclic aromatic sulfur heterocycles (PASHs) comprise a complex class of condensed multi-ring benzenoid compounds containing a sulfur atom in one of their heterocyclic rings. Originating from numerous natural and anthropogenic combustion processes, PASHs have been found in a variety of environmental samples, including air particulate matter [1,2], sediments [3], coal liquids [4], crude oils [5–9] and asphalt samples [10]. Health risks associated with human exposure to PASHs have prompted the inclusion of a few representatives in the priority pollutants lists of the European Union [11], the National Oceanic and Atmospheric Administration (NOAA) [12] and the US Environmental Protection Agency (EPA) [13]. These include three-ring PASHs (dibenzothiophene; 4-methyldibenzothiophene; 4,6-dimethyldibenzothiophene; 2,4,6-trimethyldibenzothiophene; and 1,2,3,4-tetramethyldibenzothiophene) and four-ring PASHs (benzo[b] naphtho[1,2-b]thiophene; benzo[b]naphtho[2,1-b]thiophene; and benzo[b]naphtho[2,3-b]thiophene.

Since the mutagenic and carcinogenic properties of PASHs depend on their molecular structures [14–17], the ability to detect trace concentration levels of toxic PASHs in a complex sample is highly valued. Analytical approaches for the determination of PASHs rely on high-performance liquid chromatography (HPLC) coupled to absorption or fluorescence detectors [18,19] and gas chromatography-mass spectrometry (GC/MS) [20,21]. Unfortunately, the HPLC detection schemes are not specific enough to differentiate co-eluting PASHs with overlapping absorption and fluorescence spectra. Therefore, the unambiguous determination of PASHs in complex environmental extracts is best accomplished with the GC/MS analysis of HPLC fractions. In addition to co-eluting PASHs, analysts face the challenge of potential interference from polycyclic aromatic hydrocarbons (PAHs) and their alkylated homologues (A-PAHs) with strong absorption and fluorescence emission. GC/MS analysis of normal-phase liquid chromatography (NPLC) fractions provides three qualitative parameters for PASHs identification; namely NPLC and GC retention times and mass fragmentation patterns [22,23].

Particularly relevant to the present studies is the direct determination – i.e. no chromatographic separation – of targeted compounds with the combination of multidimensional spectroscopic data and second (or higher) order multivariate calibration methods [24–29]. For the specific case of PASHs, we investigated the photoluminescence properties of anthra-thiophenes and benzonaphtho-thiophenes at room temperature (RT), 77 K and 4.2 K. These included anthra[1,2-b]thiophene, anthra[2,1-b]thiophene, anthra[2,3-b]thiophene, benzo[b]naphtho [1,2-b]thiophene (BbN12T), benzo[b]naphtho[2,1-b]thiophene (BbN21T), and benzo[b]naphtho[2,3-b]thiophene [30].

When dissolved in n-octane, anthra- and benzo-thiophenes emitted fluorescence at RT, 77 K and 4.2 K. The emission of phosphorescence was only observed at 77 K and 4.2 K. Independent of the cryogenic temperature and the type of photoluminescence, the limits of detection (LODs) were at the parts-per-billion concentration levels (ng.mL⁻¹) for all the studied PASHs. The main advantage of phosphorescence over steady-state fluorescence measurements was the possibility to time resolve fluorescence interference with commercial instrumentation. In addition to avoiding the interference of fluorescence concomitants usually present in NPLC fractions, we could resolve the strong overlapping between BbN12T and BbN21T - in both the wavelength and the time domains - by processing 77 K phosphorescence excitation-emission matrices (P-EEMs) with parallel factor analysis (PARAFAC) [30]. This approach allowed for the identification – i.e.; qualitative analysis - of BbN21T in NPLC fractions of a complex coal tar sample; namely Standard Reference Material (SRM) 1597a.

The present study investigates the photoluminescence properties of phenanthro-thiophenes (PTs), a group of four-ring PASHs with molecular mass 234 g mol⁻¹. These include phenanthro [1,2-b] thiophene (P12T), phenanthro [2,1-b] thiophene (P21T), phenanthro [2,3-b] thiophene (P23T), phenanthro [3,2-b] thiophene (P32T), phenanthro [3,4-b]thiophene (P34T), phenanthro [4,3-b]thiophene (P43T) and phenanthro [9,10-b]thiophene (P910T). Their molecular structures are shown in Fig. 1.

Fig. 1.

Molecular structures of phenanthrothiophenes.

To the extent of our literature search, there are no reports on the photoluminescence properties of the studied PT. Analytical figures of merit based on the fluorescence and phosphorescence emitted at 77 K and 4.2 K show the feasibility to detect these compounds at the ng.mL⁻¹ concentration levels. Previous NPLC analysis of combustion-related SRMs 1597, 1991 and 1597a showed the co-elution of P12T, P34T, P43T, and P910T in fraction 6 and the co-elution of P21T, P23T, and P32T in fraction 7 of the NPLC procedure. Individual identification of the co-eluted compounds was only obtained after GC/MS analysis of the NPLC fractions. Herein, we present an alternative method for the identification of phenanthro-thiophenes in NPLC fractions based on 77 K P-EEMs and PARAFAC. We demonstrate the potential of this approach for the identification of P43T in the presence of co-eluted phenanthro-thiophenes with strong phosphorescence emission and co-eluted PAHs with strong fluorescence emission. In addition to extending the applicability of low-temperature time-resolved phosphorescence excitation emission matrices - PARAFAC from benzonaphtho-thiophenes [30] to phenanthro-thiophenes, the present studies demonstrate its ability to quantify PASHs isomers in NPLC fractions with diverse chemical compositions.

2. Experimental

2.1. Chemicals

All chemicals were of analytical-reagent grade and utilized without further purification. HPLC grade n-octane and n-heptane were acquired from Acros Organics. P12T, P21T, P23T, P32T, P34T, P43T, and P910T were obtained from the National Institute of Technology (NIST) and used as received. Daily instrumental performance for photoluminescence measurements was monitored with a commercial standard from Photon Technology International (PTI) consisting of a single crystal of dysprosium-activated yttrium aluminum garnet mounted in a cuvette-sized holder with well-characterized quasi-line excitation and emission spectra.

2.2. Preparation of stock solutions

Stock solutions of PASHs prepared in n-alkanes were kept in the dark at 4 °C until use. Possible PASHs degradation was monitored via room-temperature fluorescence spectroscopy. PASHs working solutions were prepared prior to data collection using serial dilution of stock solutions with the appropriate n-alkane solvent.

2.3. Instrumentation

Photoluminescence spectra and signal intensities were recorded with a FluoroMax – P (Horiba Jobin-Yvon) equipped with a continuous xenon arc source for fluorescence measurements and a pulsed xenon arc source for phosphorescence measurements. The range of flash rate for the pulsed source was 0.05–25 Hz 3 μs was the duration of the pulse at full-width half maximum. The 1200 grooves/mm gratings in the single excitation and emission monochromators were blazed at 330 nm and 500 nm, respectively. Their reciprocal linear dispersion was equal to 4.25 nm/mm. The uncooled photomultiplier tube (Hamamatsu, Model R928) detector was operated in the photon-counting mode. For time-resolved measurements, the gating circuitry provided delays after the flash ranging from 50 μs to 10 s (in increments of 1 μs) and sample windows (gate times) ranging from 10 μs to 10s (in increments of 1 μs). Commercial software (DataMax version 2.20, Horiba Jobin-Yvon) was used for automated scanning and fluorescence data acquisition. Origin software (version 5, Micronal Software, Inc.) was used for curve fitting of phosphorescence decays. Fitted decay curves were obtained by fixing x₀ and y₀ at a value of zero in equation (1).

$$y=y_0+A_1{e}^{[-(x-x_0\left)t_1\right]}$$ (1)

2.4. Fiber optic probe (FOP)

Photoluminescence measurements at 77 K and 4.2 K were made with the aid of a FOP built in house [30]. All fibers were silica-clad silica and coated with polyimide buffer (Polymicro Technologies, Inc.). Their length was 2 m each and their core diameter was 500 μm. The sample end of the FOP consisted of two excitation fibers and eight emission fibers arranged in an eight-around two configuration. At the instrument end, the excitation and the emission fibers were positioned in a slit configuration. Fibers at both ends of the probe were held in place with vacuum epoxy. At the sample end, the copper tubing was flared stopping a phenolic screw cap threaded for a 0.75 mL propylene sample vial. The FOP was coupled to the sample compartment of the spectrofluorimeter with the aid of a commercial fiber optic mount (F-3000, Horiba Jobin Yvon) that used two concave mirrors to optimize the collection efficiency of excitation and emission radiation. Alignment of excitation and emission fibers with their respective focusing mirrors was facilitated by commercially available adapters (Horiba Jobin-Yvon).

2.5. Measurement procedures

RT measurements were made by pouring un-degassed liquid solutions into standard quartz cuvettes (1 cm optical path). Signals and spectra were recorded by using a 90° configuration between the excitation source and the detector. 77 K and 4.2 K measurements were made with the FOP as follows: (1) 100 μL of un-degassed sample solution were pipetted into the sample vial; (2) the sample vial was secured to the sample end of the copper tubing; and (3) tip of the FOP was positioned at a constant depth below the solution surface; and (4) sample freezing was accomplished by lowering the sample vial into the liquid cryogen.

Liquid nitrogen and liquid helium were held in two separate Dewar containers with 5- and 60 L storage capacity, respectively. The 60 L liquid helium volume would typically last three weeks of daily use, averaging 15–20 samples per day. Complete sample freezing took less than 90 s. The ~1 min probe clean up procedure involved removing the sample vial from the cryogen container, melting the frozen matrix and warming the resulting solution to approximately room temperature with a heat gun, rinsing the probe with n-alkane, and drying it with warm air from the heat gun. The entire freeze, thaw, and clean up cycle took less than 5 min.

2.6. Software for data analysis with PARAFAC

MATLAB 7.10 was used for all calculations (The Mathworks Inc., Natick, Massachusetts, USA, 2010) [31] Software processing was facilitated by a MVC3 graphical user interface written in MATLAB previously described by Olivieri et al. (2012) [32].

3. Results and discussion

The spectral resolution of polycyclic aromatic hydrocarbons (PAHs) recorded from n-alkane solutions at low temperature depends on the relative dimensions of the guest (PAH) and host (n-alkane) molecules. If the dimensions of the PAH and the n-alkane match well enough, the small number of crystallographic sites - ideally just one - occupied by guest molecules in the host lattice provide vibrational spectra with fingerprinting information for PAH identification. A criterion often employed for the initial selection of n-alkanes is to match the linear dimensions of the PAH molecule to those of the organic solvent [33]. The same criterion was employed here to investigate the photoluminescence properties of phenanthro-thiophenes.

The molecular dimensions of PTs and n-alkanes were calculated at NIST using commercial modeling programs (PC-Model and MMX, Serena Software, Bloomington) [34]. The best PASH/n-alkane matchings are summarized in Table 1. All compounds were dissolved in the calculated n-alkanes and used without further solvent optimization studies. Spectral characteristics were investigated at RT, 77 K and 4.2 K from pure standards at their maximum excitation and emission wavelengths. When necessary, second order emissions from excitation scattering were removed by placing long-pass cut-off filters at the entrance slit of the emission monochromator.

Table 1.

Shpol’skii solvents for spectral collection at RT, 77 K and 4.2 K.

PASHa	Length of PASH (Å)	Matching Solvent	Length of Solvent (Å)
P12T	13.49	n-octane	13.142
P21T	13.35	n-octane	13.142
P23T	13.43	n-octane	13.142
P32T	13.74	n-octane	13.142
P34T	12.27	n-heptane	11.867
P43T	12.27	n-heptane	11.867
P910T	11.83	n-heptane	11.867

3.1. Steady-state photoluminescence spectroscopy

Steady-state spectra were recorded with the aid of a continuous wave (CW) xenon excitation source. Figs. S1 and S2 show the RT excitation and fluorescence spectra recorded from the studied compounds in n-alkane solutions. Except for P21T, all the other compounds showed some vibrational structure in their excitation and fluorescence spectra. No phosphorescence emission was observed at RT. Considering that all the phenanthro-thiophenes showed phosphorescence at 77 K and 4.2 K, the lack of RT phosphorescence is probably due to the relatively long lifetimes of triplet states that provide many opportunities for collisional deactivations of excited molecules. Since all measurements were made from un-degassed solutions, the contribution of oxygen quenching to the lack of phosphorescence is also possible. The lack of vibrational features in the RT spectra of P21T could be attributed to the 2 nm excitation and emission band-pass used for spectral acquisition. A narrower band-pass could have possibly resolved the vibrational structure of P21T. Its vibrational levels in the excited (S₁) and the ground (S₀) energy states appear to be much closer in energy than the vibrational levels of all the other compounds.

All the studied phenanthro-thiophenes showed fluorescence and phosphorescence at low temperatures. The entire set of 77 K and 4.2 K excitation, fluorescence and phosphorescence spectra are compiled in Fig. S3 – S6. Fig. 2A compares the relative intensities of the fluorescence signals observed at RT, 77 K and 4.2 K. In all cases, lowering the temperature to 77 K or 4.2 K enhanced the fluorescence of the studied compounds. With the exception of P21T, all the other phenanthrothiophenes emitted the strongest fluorescence at 4.2 K. These observations probably result from reducing vibrational relaxation and collisional deactivation of the first singlet states in the frozen matrixes.

Fig. 2.

(A) Relative intensities of fluorescence signals recorded at RT, 77 K and 4.2 K. (B) Relative intensities of phosphorescence signals recorded at 77 K and 4.2 K. The uncertainty listed with each value is the standard deviation for the average normalized intensity value for each set of fluorescence (A) and phosphorescence (B) signals. All measurements were made with the CW lamp using standard concentrations within the linear dynamic ranges of the calibration curves. Standards were prepared with the n-alkane solvents shown in Table 1. Excitation an emission band-pass was 4 nm. RT measurements were made with the aid of 1-cm quartz cuvettes. Low temperature measurements were made with the aid of the fiber optic probe.

Fig. 2B compares the phosphorescence intensities at 77 K and 4.2 K. With the exception of P12T and P21T, all the other compounds showed the strongest phosphorescence at 4.2 K. Since we do not anticipate for the rate of intersystem crossing between the singlet and triplet manifolds to change significantly with the temperature of the frozen matrix, the observed behavior is probably due to reducing vibrational relaxations and collisional deactivations of the first triplet states in the frozen matrixes.

Table S1 summarizes the phosphorescence (I_P) to fluorescence (I_F) intensity ratios of the studied compounds at 77 K and 4.2 K P12T showed the highest I_P/I_F ratio followed by P910T and P43T. While P34T and P21T emitted phosphorescence and fluorescence with similar intensities, P23T and P32T emitted stronger fluorescence than phosphorescence. The observed differences stress the role of the heavy atom (sulfur) position in promoting intersystem crossing between the first singlet excited state and the triplet state manifold of phenanthro-thiophenes.

3.2. Fluorescence analytical figures of merit (AFOMs) at RT, 77 K and 4.2 K

Table 2 summarizes the fluorescence AFOMs of phenanthro-thiophenes recorded with the CW lamp at RT, 77 K, and 4.2 K. Calibration curves were prepared by serial dilutions of stock standard solutions with the solvents in Table 1. All measurements were made at maximum excitation and emission wavelengths. Each calibration curve consisted of a minimum of five linear concentrations. Each concentration was plotted in the calibration graph versus the average intensity of three signal determinations made from three sample aliquots (N = 3). No efforts were made to experimentally obtain the upper concentration limit of the calibration curve. The best linear fitting of each calibration curve was obtained with the least squares method.

Table 2.

Fluorescence analytical figures of merit of phenanthrothiophenes recorded with the CW Lamp at RT, 77 K and 4.2 K.

PASH	Temperature	λex/ema (nm)	LDRb (ng mL−1)	LOQc (ng mL−1)	LODd (ng mL−1)	RSDe (%)
P12T	RT	275/380	58–1250	58	18	1
	77 K	277/381	14–250	14	4	0.2
	4.2 K	277/381	11.5–250	12	3.4	2.7
P21T	RT	320/383	9.9–1250	9.9	2.9	0.4
	77 K	320/374	9.5–500	9.5	3	0.7
	4.2 K	321/374	7–500	7	2.2	2.5
P23T	RT	284/368	6.8–1250	6.8	2	1
	77 K	288/390	7–1000	7	2	0.5
	4.2 K	288/390	12–1000	12	3.5	3
P32T	RT	283/372	5.8–313	5.8	1.7	2.5
	77 K	287/373	4–250	4	1	0.6
	4.2 K	287/373	1–250	1	0.3	2.6
P34T	RT	279/367	121–1000	121	36	0.4
	77 K	281/362	19–500	19	5.6	1
	4.2 K	281/362	24–500	24	7	1.5
P43T	RT	317/363	188–1000	188	56	2.6
	77 K	322/361	54–500	54	16	11
	4.2 K	322/361	20–500	20	5.9	0.6
P910T	RT	291/363	162–1000	162	48	1.6
	77 K	313/370	17–500	17	5	0.8
	4.2 K	313/370	50–500	50	15	0.4

^aExcitation (λ_ex) and emission (λ_em) wavelengths.

^bLDR = linear dynamic range in ng mL⁻¹ extending from the limit of quantification (LOQ) to an arbitrarily chosen upper linear concentration.

^cLOQs were calculated using LOQ = 10S_B/m.

^dLODs were calculated using LOD = 3S_B/m, where S_B is the standard deviation of 16 blank measurements and m is the slope of the calibration curve.

^eRelative standard deviation (RSD) = S_F/I_F x 100, where S_F is the standard deviation of the average calculated from three emission measurements at medium linear PASH concentrations within same day.

The limits of detection (LOD) and the limits of quantitation (LOQ) were calculated as follows: LOD = 3 × S_B/M and LOQ = 10 × S_B/M; where S_B is the standard deviation of 16 blank determinations and M is the slope of the calibration curve. Although considerable fluorescence enhancements were observed by lowering the temperature of all the studied compounds (see Fig. 2A), only P34T, P43T, and P910T experienced significant improvements – i.e. ~ one order of magnitude - of their LODs and LOQs. All the other compounds had LODs and LOQs with ~ the same order of magnitude at the three studied temperatures.

Table S2 compares the slopes (m), blank signals, and standard deviations of the blank signals (S_B) used to calculate the LODs and LOQs in Table 2. Comparison of the slopes at room and low temperatures shows higher slopes at 77 K and 4.2 K in all cases. This trend reflects the fluorescence enhancements observed at lower temperatures from all the studied compounds. Blank signals (Y_B) and standard deviations of the blank signals (S_B) showed a similar trend with the highest values at lower temperatures. Comparison of the slope ratios at 4.2 K (M_4.2K) and room temperature (M_RT) reveals the highest M_4.2K/M_RT values for P34T (78.6), P34T (66.7), P12T (62), and P910T (37.1). Similar comparisons with the blank signals reveal the smallest Y_B^4·2K/Y_B^RT ratios for P34T (4.1), P910T (18.3), and P43T (18.5). The same is true for the standard deviations of the blank signals; i.e. the smallest S_B^4·2K/S_B^RT ratios were observed from P34T (1.4), P43T (6.9), and P910T (10.3). Although the Y_B ± S_B values for P34T, P43T, and P910T increase at the lower temperature, the improvements in the slopes of their calibration curves overcome the higher backgrounds and still provide better LODs and LOQs at 4.2 K. The same is true at 77 K. All the other compounds presented higher blank ratios than slope ratios. The Y_B^4·2K/Y_B^RT and M_4.2K/M_RT ratios were the following: P12T (62 and 95.6), P21T (8.2 and 50), P23T (20 and 68.3), and P32T (6.1 and 58.3). Their S_B^4·2K/S_B^RT ratios varied as follows: P12T (12.3), P21T (6.2), P23T (34.5), and P32T (10.4). The higher blank signals and standard deviations at 4.2 K appear to overcome the slope improvements at the lower temperature to provide LODs and LOQs with the same order of magnitude at RT and 4.2 K. The same trends were observed at 77 K. The higher Y_B ± S_B values at 77 K and 4.2 K can be attributed to the presence of unknown fluorescence impurities in the n-alkane solvents. Although HPLC grade solvents were used throughout, further solvent purification via distillation steps should provide better LODs and LOQs.

3.3. Time-resolved phosphorescence spectroscopy

All the studied phenanthro-thiophenes emitted phosphorescence at 77 K and 4.2 K. Since the spectra presented in Figs. S3–S6 were recorded with the CW excitation source of the spectrofluorimeter, the emission spectra show the contribution of both fluorescence and phosphorescence. Due to the “forbidden” nature of phosphorescence, which results from the transition between two electronic states of different multiplicity (T₁ to S₀), phosphorescence spectroscopic techniques are potentially more selective than fluorescence spectroscopy techniques. Since the time domain of phosphorescence (~10⁻³ s to s) is considerably longer than the “short-lived” fluorescence phenomenon (~10⁻⁹ to 10⁻⁶ s), spectral interference from fluorescence concomitants is often removed with the aid of time-resolution techniques [35–37]. One way to accomplish time-resolved phosphorescence measurements is with the aid of pulsed excitation sources and gated detection systems. Removing the contribution of fluorescence from the total emission spectrum of a sample requires the choice of an appropriate delay time (D), which depends on the full-width at half maximum of the excitation pulse and the duration of the fluorescence decay. To record phosphorescence and still avoid the collection of background noise, a suitable gate time (G) is required that depends on the phosphorescence time decay of the phosphor.

Table S3 summarizes the 77 K and 4.2 K phosphorescence lifetimes of the phenanthro-thiopenes studied herein. Measurements were made from standard solutions prepared in n-alkanes (see Table 1) at medium linear concentrations. Signal intensities were recorded at the maximum excitation and emission wavelengths of each compound using 4 nm excitation and emission band-passes. All phosphorescence decays consisted of well-behaved single exponential decays with no systematic trends.

P23T, P43T, and P910T showed significantly shorter phosphorescence lifetimes at 4.2 K than at 77 K (P = 95%; N₁ = N₂ = 3). The observed difference could be attributed to an enhancement of the radiative rate constant for the deactivation of the first triplet state (T₁). This assumption agrees with the higher phosphorescence intensities observed at 4.2 K. Although the average lifetimes of P12T, P21T, P32T and P34T are also shorter at the lowest temperature (4.2 K), the observed differences are not statistically significant (P = 95%; N₁ = N₂ = 3) fall within the reproducibility of measurements of our experiments.

Since the gates in Table S3 lead to high phosphorescence intensities but unrealistically long spectra collection times, we conducted a study to reach a compromise between phosphorescence intensity and spectral recording time. The time it takes to record phosphorescence spectra with a gated spectrometer depends on the number of excitation pulses per data point, the time between excitation pulses and the number of data points per spectrum. The time between excitation pulses totals the sum of the gate time and the time for data transfer and storage (DTS). Since our spectrometer provides no control over the length of the DTS step, we reduced the gate time and kept constant all the other instrumental parameters used in Table 4; i.e. the same delay times, 10 pulses per data point, 4 nm excitation and emission band-pass and 3 nm monochromator steps.

Table 4.

Core consistencies of P34T obtained with PARAFAC from 77 K time-resolved phosphorescence EEMs^a.

Sample	P34T (μg.mL−1)	P43T (μg.mL−1)	P21T (μg.mL−1)	P910T (μg.mL−1)	Core consistency (%)
Calibration sample 1	1	1	1	1	—b
Calibration sample 2	2	1	1	1	—b
Calibration sample 3	4	1	1	1	—b
Test sample 1	2.5	1	1	1	100
Test sample 2	3.5	1	1	1	100
HPLC fraction 6	NA	NA	NA	NA	100

^aEEMs were recorded using fix delay (40 μs) and gate (9 ms) times.

^bNot applicable.

Among the gate delays we tested, the shortest G that provided LODs and LOQs of the same order of magnitude as those obtained under steady-state conditions was 9ms. Table 3 summarizes the low-temperature time-resolved phosphorescence AFOM obtained with a G of 9ms. The complete set of excitation and time-resolved phosphorescence spectra are shown in Figs. S7–S10. A 9 ms gate time was then used for all the remaining studies of this article.

Table 3.

Time-resolved phosphorescence AFOMs of phenanthrothiophenes recorded with the pulsed lamp.

PASH	Temperature	λex/ema (nm)	LDRb (ng mL−1)	LOQc (ng mL−1)	LODd (ng mL−1)	RSDe (%)
P12T	77 K	275/515	5–500	5	1.5	4
	4.2 K	275/514	6–1000	6	1.8	1
P21T	77 K	262/487	81.9–2500	81.9	24.5	7.5
	4.2 K	262/486	95–2500	95	28.6	2
P23T	77 K	288/546	71–750	71	21	5
	4.2 K	288/546	54.7–750	54.7	16	1.5
P32T	77 K	287/530	10.5–500	10.5	3	0.6
	4.2 K	287/530	11.8–500	11.8	3.5	2.6
P34T	77 K	283/501	6.7–500	6.7	2	0.26
	4.2 K	284/501	7.8–500	7.8	2.3	1
P43T	77 K	274/481	14–250	14	4	0.79
	4.2 K	275/480	5–250	5	1.5	0.56
P910T	77 K	262/482	5.7–500	5.7	1.7	2.9
	4.2 K	262/480	30.5–500	30.5	9	0.5

^aExcitation (λ_ex) and emission (λ_em) wavelengths.

^bLDR = linear dynamic range in ng mL⁻¹ extending from the limit of quantification (LOQ) to an arbitrarily chosen upper linear concentration.

^cLOQs were calculated using LOQ = 10S_B/m.

^dLODs were calculated using LOD = 3S_B/m, where S_B is the standard deviation of 16 blank measurements and m is the slope of the calibration curve.

^eRelative standard deviation (RSD) = S_F/I_F x 100, where S_F is the standard deviation of the average calculated from three emission measurements at medium linear PASH concentrations within same day.

3.4. Analysis of P34T in NPLC fractions from SRM 1597a

Previous work in our lab to circumvent potential interference from matrixes of unknown composition has coupled multidimensional data formats to multi-way calibration algorithms. Multidimensional data formats include total synchronous fluorescence spectra [38], EEMs [39], wavelength-time matrices (WTMs) [40,41], time-resolved EEMs (TREEMs) [42] and time-resolved excitation-emission cubes (TREECs) [43]. Spectral deconvolutions have been achieved with PARAFAC, unfolded-partial least squares/residual bi-linearization (U-PLS/RBL) and unfolded-partial least squares/residual tri-linearization (U-PLS/RTL).

The approach we present here is based on the collection of low-temperature time-resolved phosphorescence EEMs. This data format is inherently different from TREEMs and TREECs. Herein, the term “time-resolved” refers to the time-discrimination of fluorescence interference with the application of a fixed delay and gate time. Since all the EEMs were recorded with the same delay and gate times, the temporal dimension did not contribute to the deconvolution of overlapped phosphorescence spectra.

Under the NPLC conditions optimized previously for the analysis of PASHs in combustion-related SRMs [22,23], P12T, P34T, P43T, and P910T co-elute in fraction 6 of the chromatogram along with numerous strong emitting fluorophores of known and unknown identity. Known fluorophores co-eluting with PTs in fraction 6 include benzo[c]phenanhtrene, naphthacene, 2-methylbenzo[c]phenanthrene, 3-methyl-benzo[c]phenanthrene, 4-methylbenzo[c]phenanthrene, 5-methyl-benzo[c]phenanthrene, 6-methylbenzo[c]phenanthrene, 1-methylbenz [a]anthracene and 11-methylbenz[a]anthracene.

After evaporating the mobile-phase to dryness and re-constituting it with n-heptane, fraction 6 provides a matrix of analysis with negligible fluorescence under the time-resolved parameters of this study; i.e. D = 40 μs and G = 9 ms. Therefore, the successful identification of any given PT in fraction 6 depends on the method’s ability to handle phosphorescence interference from the other three co-eluting PTs and from any other co-eluting phosphorescence species with unknown chemical identity.

P34T was the model compound we chose to investigate the selectivity of this approach. PARAFAC calculations were performed with EEMs recorded from three calibration mixtures, 2 test samples and NPLC fraction 6. All EEMs were recorded at 77 K using a D = 40 μs and a G = 9 ms gate, 4 nm excitation and emission band-passes, 10 pulses per data points, and 3 nm monochromator steps. The wavelength ranges extended from 200 nm to 330 nm (excitation) and from 420 nm to 700 nm (emission).

Table 4 summarizes the results obtained with the calibration set, the two test samples and the unknown (NPLC fraction 6). Figs. 3 and 4 compare the spectral profiles extracted with PARAFAC with the experimental spectra recorded from pure standard solutions of P34T. The core consistency values in Table 4 confirm the visual similarities among predicted and experimental spectra.

Fig. 3.

Summary of results obtained for test sample 2 (see Table 4 for sample composition). Top left: 77 K time-resolved phosphorescence EEM recorded from sample 2 in n-octane; Top right: Phosphorescence (emission) and excitation spectra of P34T predicted with PARAFAC; Bottom: 77 K excitation and phosphorescence spectra recorded from a pure standard solution of P34T in n-octane. EEM and 2D spectra were recorded with the pulsed excitation source and the fiber optic probe using a delay = 40 μs, a gate = 9 ms, and excitation/emission band-pass = 4 nm.

Fig. 4.

Summary of results obtained for HPLC fraction 6. Top left: 77 K time-resolved phosphorescence EEM recorded from fraction 6 in n-octane; Top right: Phosphorescence (emission) and excitation spectra of P34T predicted with PARAFAC; Bottom: 77 K excitation and phosphorescence spectra recorded from a pure standard solution of P34T in n-octane. EEM and 2D spectra were recorded with the pulsed excitation source and the fiber optic probe using a delay = 40 μs, a gate = 9 ms, and excitation/emission band-pass = 4 nm.

4. Conclusion

Due to the complex chemical composition of oil-contaminated sites, the un-doubtful identification of PASHs requires extensive chromatographic procedures based on the GC-MS analysis of NPLC fractions. The studies presented here provide an alternative approach that requires no further separation of NPLC fractions. The presence of the sulfur atom in the molecular structure of PTs promotes intersystem crossing between the first singlet excited state and the triplet excited state manifold. The heavy atom effect associated to the matrix rigidity of the frozen sample (77 K or 4.2 K) provides strong phosphorescence for the determination of PTs at ppb concentration levels. The use of a 40 μs delay after the excitation pulse removes fluorescence interference from co-eluting species in the NPLC fraction. Interference from co-eluting phosphors can be processed with the aid of PARAFAC; i.e. a multivariate algorithm with the ability to handle time-resolved phosphorescence EEMs. These features provide a straightforward procedure based on recording 77 K phosphorescence EEMs from NPLC fractions for further data analysis with PARAFAC.