Master’s degree thesis in Chemistry: 15 ECTS
“Source characterization of soils contaminated with
Polycyclic Aromatic Compounds (PACs) by use of Partial
Least Squares Discriminant Analysis (PLS-DA)”
Tim Sinioja 2017-03-20 Supervisors: Maria Larsson and Hans F. Grahn
2
Abstract
Polycyclic aromatic compounds (PACs) are organic compounds that include several sub-groups of toxic, persistent and carcinogenic environmental pollutants consisting of two or more non-substituted or non-substituted aromatic rings. Due to the complexity of PAC-mixtures found in the environment it can be challenging and time-consuming to track the sources of contamination. In the present study, multivariate data analysis (MVDA) models, such as principal component analysis (PCA) and partial least squares discriminant analysis (PLS-DA) were applied to track sources of PACs at contaminated sites. Based on the chemical profile of 78 PACs obtained in GC-MS analysis of soils, 26 observations were classified according to their petrogenic, pyrogenic or urban background soil origin. Two soil samples of unknown origin collected at a contaminated site in Mjölby, Sweden, were successfully fitted to the validated PLS-DA model and their origins were determined as petrogenic. The study shows that validated PLS-DA models can be applied to predict the petrogenic, pyrogenic and urban background soil origins of samples collected at PAC contaminated sites, thus to track the sources of contamination. It is also concluded that 16 U.S. Environmental Protection Agency’s (EPA) priority polycyclic aromatic hydrocarbons (PAHs) are not sufficient to predict the origin of contamination with PCA or PLS-DA.
3
Table of content
ABSTRACT ... 2 TABLE OF CONTENT ... 3 LIST OF ABBREVIATIONS ... 4 1. INTRODUCTION ... 51.1 PROPERTIES, SOURCES AND FATE OF PACS IN THE ENVIRONMENT ... 5
1.2 PRESSURIZED LIQUID EXTRACTION AND GAS CHROMATOGRAPHY MASS SPECTROMETRY ... 9
1.3 MULTIVARIATE DATA ANALYSIS ... 10
1.4 OBJECTIVES ... 11
2. METHOD AND MATERIALS ... 12
2.1CHEMICALS ... 12 2.2ORIGIN OF SAMPLES ... 12 2.3INSTRUMENTATION ... 13 2.3.1 PLE ... 13 2.3.2 GC-MS ... 13 2.4MVDA ... 14
3. RESULTS AND DISCUSSION... 15
3.1PRINCIPAL COMPONENT ANALYSIS ... 15
3.2PARTIAL LEAST SQUARES DISCRIMINANT ANALYSIS ... 17
CONCLUSION... 25
ACKNOWLEDGEMENTS ... 26
REFERENCES ... 27
APPENDIX A – GC-MS PARAMETERS ... 32
APPENDIX B – ADDITIONAL FIGURES ... 33
APPENDIX C – ADDITIONAL TABLES ... 38
4
List of abbreviations
Alkyl-PAH – Alkylated Polycyclic Aromatic Hydrocarbon EPA – Environmental Protection Agency
GC-MS – Gas Chromatography Mass Spectrometry
HMW PAH – High Molecular Weight Polycyclic Aromatic Hydrocarbon LMW PAH – Low Molecular Weight Polycyclic Aromatic Hydrocarbon LOD – Limit of Detection
LOQ – Limit of Quantification
MVDA – Multivariable Data Analysis
Oxy-PAH – Oxygenated Polycyclic Aromatic Hydrocarbon PAC – Polycyclic Aromatic Compound
PAH – Polycyclic Aromatic Hydrocarbon PCA – Principal Component Analysis PLE – Pressurized Liquid Extraction
PLS-DA – Partial Least Squares Discriminant Analysis RRF – Relative Response Factor
5
1. Introduction
1.1 Properties, sources and fate of PACs in the environment
Polycyclic aromatic compounds (PACs) are organic compounds that include several sub-groups of toxic, persistent and carcinogenic environmental pollutants consisting of two or more non-substituted or non-substituted aromatic rings. The subgroups of PACs such as polycyclic aromatic hydrocarbons (PAHs), oxygenated PAHs (oxy-PAHs), alkylated PAHs, nitrogen containing aromatic compounds (azaarenes) etc., comprise thousands individual compounds. Exposure to PACs entails a significantly increased risk of the developing lung, skin, bladder or other types of cancer (Mastrangelo, et.al., 1996). PACs can also induce immunotoxic effects, that is inhibition of blood monocytes differentiation into macrophages, which leads to the immune system suppression in humans (van Grevenynghe, et.al., 2003).
The major anthropogenic sources of PACs are petroleum processing during incineration, coke production, auto-engines and other combustion processes of hydrocarbons (Baek, et.al., 1991). Thus, PACs are commonly found at coke production sites and coal gas manufacturing sites, such as gasworks (Howsam & Jones, 1998). In addition to incomplete combustion and pyrolytic processes in gas- and coke works, PACs are also found in creosote, a complex mixture of carbonaceous compounds that has been widely used as a preservative in wood treatment processes. For example, a coal tar creosote, i.e. creosote derived from the distillation of a coal tar, can contain up to 85% PAHs (Murphy & Brown, 2005) together with up to 5% oxy-PAHs and azaarenes (Mueller, et.al., 1989). The major preservative applications of creosote are impregnation of wooden railway cross ties (sleepers), bridge timber, utility and fence poles, marine pilings and wood for other exterior uses. A previous study showed that a single creosote treated railway sleeper can emit around 5kg creosote during its lifespan of more than 30 years (Kohler & Künniger, 2003). Because of high concentrations of PACs in creosote and its wide use, the substance has been classified in the European Union, Australia, Canada, United States and other countries as a hazardous and probable human carcinogen. As a result, used railway sleepers are classified as an environmentally hazardous material if the creosote concentration exceeds 1g per kg dry weight (94/67/EEC, 1994).
During the risk assessment of contaminated sites, a small subgroup of non-substituted PACs is generally monitored. This subset includes 16 non-substituted PAHs identified as priority pollutants by the United States Environmental Protection Agency (EPA).
6 Nevertheless, it has been recognized that other classes of PACs are present as well at contaminated sites and may contribute significantly to the toxicity, and thus increase the environmental risks posed by polluted sites (Lundstedt, et.al., 2003). For example, oxy-PAHs, alkylated PAHs and azaarenes are of a high relevance in this context (Andersson & Achten, 2015).
PAHs are PAHs with one or more oxygen atoms attached to the aromatic ring. Oxy-PAHs may also contain other chemical groups, e.g. a hydroxyl group in
4-hydroxy-9-fluorenone and alkyl groups in 2-methyl-9-4-hydroxy-9-fluorenone and 2-methylanthracenedione (Figure 1).
7 Because oxy-PAHs are formed during an incomplete combustion of organic material, they are usually emitted from the same sources as PAHs (Stenberg & Alsberg, 1981). Oxygenated PAHs may also be formed through photo- or chemical oxidation and biological transformation of earlier emitted to the environment PAHs (Cerniglia, 1992; Kochany & Maguire, 1994; Yu, 2002). Previous studies showed that oxy-PAHs are generally more persistent than other transformation products, therefore they can potentially accumulate in the environment as PAHs are being degraded (Andersson, et.al., 2003; Lundstedt, et.al., 2003; Lundstedt, et.al., 2006). Petroleum PAHs are abundantly substituted with alkyl groups on their ring structures which are referred as alkyl or alkylated PAHs (Irwin, et.al., 1997). These alkyl groups generally consist of one to four saturated carbon atoms attached to one or several carbons in the aromatic rings. For example, many of alkylated isomers of the same parent compound can be present in petroleum products (Figure 2).
Figure 2. Molecular structures of C2-substituted isomers of naphthalene.
Another polycyclic aromatic compound in petroleum and petroleum products is dibenzothiophene and its alkylated isomers (Ho, 2004). Because dibenzothiophenes are organosulfur compounds, their abundance in petroleum products depends on the amount of
8 reduced sulfur, i.e. level of the oxygen depletion, in the source rock in which petroleum was formed (Peters, et.al., 2005). Many of alkylated PAHs and dibenzothiophenes with formation temperatures below 150 °C are well-preserved in petroleum products (Mitra, et.al., 1999). Furthermore, with a crude maturation, the proportion of thermally stable isomers, e.g. alkylated isomers of phenanthrene, increases (Stout, et.al., 2002). As a result, petrogenic sources, usually contain higher concentrations of alkylated PAHs and dibenzothiophenes compared to the pyrogenic sources, and this phenomenon enables the characterization of the contamination source by recognition of PACs chemical profile (Stogiannidis and Laane, 2015).
Azaarenes, i.e. nitrogen-containing PACs (Figure 3), are another group of polycyclic aromatic compounds that have been shown to be highly mutagenic (Hashimoto, et.al., 1979) and highly carcinogenic in experimental animals (Matsuoka, et.al., 1982). As well as for other PACs, an emission source of azaarenes to the environment is combustion of crude oil products, coal or coal tar (Chen & Preston, 1998). In addition, azaarenes are used as pesticides, antioxidants and in the pharmaceutical industry (Bollag and Kaiser, 1991).
Figure 3. Molecular structures of selected azaarenes.
PACs, both unsubstituted and substituted, enter the environment in various ways. They are found in the air (Chen & Preston, 1998), in lake and river sediments (Kozin, et.al., 1997), in soil (Bucheli, et.al., 2004; Larsson, 2013), in groundwater (Pereira, et.al., 1987) and in sewage sludge (Tyrpień, et.al., 1995). Low molecular weight PACs are generally more volatile, thus their mobility in the environment decreases as the number of aromatic rings (i.e. molecular weight) increases (Iqbal, et.al., 2008). Because PACs are hydrophobic, they adsorb to the organic matter in soils and sediments, rather than dissolving in water or vaporizing in the air (Bertilsson & Widenfalk, 2002; Pavlova & Ivanova, 2003). Therefore, the major environmental sink of PACs is the organic fraction of sediments and soils (Stark, et.al., 2003).
9 PACs can be divided in three sub-groups depending on their origin: pyrogenic PACs, originate from pyrolysis processes of fossil fuels and biomass; petrogenic PACs, originate from petroleum and petroleum-related sources, including fuels, crude oils and their byproducts; and natural PACs, originate from diagenetic or biogenic processes (Mitra, et.al., 1999; Boehm, et.al., 2007). Previous studies revealed that combustion byproducts as well as petroleum and its products contain very complex mixtures of PACs (Farrington, et.al., 1977; Windsor & Hites, 1979). Nevertheless, concentration of different PACs varied between pyrogenic and petrogenic contamination sources (Grimmer, et.al., 1981; Grimmer, et.al., 1983; Laflamme & Hites, 1978). Degradation processes, such as microbial transformation, evaporation and dissolution, change the composition of PACs mixtures at contaminated areas depending on the contamination type, microbiological activity and other environmental conditions (Short, et.al., 2007; Sicre, et.al., 1987; Wang, et.al., 1999). Because petrogenic PACs associate less with organic material after discharge (e.g. oil spills) and are more bioavailable, they degrade at faster rates compared to pyrogenic PACs (Zakaria, et.al., 2002). In addition, due to a higher volatility and solubility, low molecular weight PACs are impacted by evaporation and dissolution earlier, therefore, both petrogenic and pyrogenic PACs mixtures will be dominated with time by four-, five- and six-ringed aromatic compounds (Stout, et.al., 2003). The complexity of these PACs mixtures at contaminated sites leads to a more complicated, time consuming and expensive processes of chemical and toxicological analyses, risk assessments and remediation procedures. But, if PACs at contaminated sites could be source characterized, i.e. linked with their sources, then the parties that caused the contamination would be determined and, potentially, proven to be financially responsible for a remediation or a clean-up of the contaminated area according to the polluter pays principle. Furthermore, the source tracking and more accurate determination of PACs chemical profile, may potentially make the remediation processes of contaminated sites more effective and secure.
1.2 Pressurized Liquid Extraction and Gas Chromatography Mass Spectrometry
Pressurized Liquid Extraction (PLE) is a sample extraction technique that utilizes principles similar to the Soxhlet extraction (Richter, et.al., 1996). Prior to the PLE, samples are mixed and homogenized with anhydrous sodium sulfate, loaded in stainless steel extraction cells and capped with end fittings. Organic solvents are then pumped into the extraction cells and samples are heated and pressurized (70-200°C, 1000-3000 PSI). As a result, the more efficient extractions are performed as the solvent inside the extraction cell approaches the supercritical region.
10 Because there are many PAC isomers, their separation only by mass-to-charge ratio would not be sufficient and chromatographic separation is usually performed in chemical analysis of PACs. For instance, gas chromatography mass spectrometry (GC-MS) is a common analytical technique for PAC analysis that provides good selectivity, sensitivity and resolution. In order to obtain good separation, a PAC selective column with siloxane stationary phase was used in this study.
1.3 Multivariate Data Analysis
Because there are thousands of various PACs and contaminated sites contain PACs of different petrogenic and pyrogenic origins, the Multivariate Data Analysis (MVDA) can be a good tool to obtain insight into the profiles of PAHs, alkylated PAHs, dibenzothiophenes, oxy-PAHs and azaarenes. For example, MVDA can be used to group samples of similar origin according to the similarities in the chemical profile of PACs. In other words, a pattern recognition of similarities and dissimilarities among the variables (PACs concentrations) can help to differentiate petrogenic contamination sources from pyrogenic, and vice versa.
Principal component analysis (PCA) is a multivariate projection method for the extraction of the systematic variation from the data (Eriksson, et.al., 2013). It is formulated as “finding lines and planes of closest fit to systems of points in space” (Jackson, 1991). PCA helps to represent a data as a low (2 to 5) dimensional plane, i.e. gives a graphical overview of the obtained variables. Prior to PCA, variables are scaled and mean-centered (Figure 4). Scaling of data is performed to set data into similar ranges and variance before calculating a PCA model. Thus, the variables with a greater range, influence the model with equal importance as variables with a smaller range. Mean-centering is performed by subtracting the average value of each variable from the data to improve the interpretability of the model.
11 After scaling and mean-centering, the principal components (scores- and loading-vectors) are computed and can be plotted against each other. For example, samples of the pyrogenic origin would ideally be plotted near each other and samples of petrogenic origin at a distance from them, but close to each other.
Another major multivariate data analytical method to relate two data matrices to each other is by using partial least squares projections to latent structures, PLS (Wold, et.al., 1984). PLS is based on a perturbation theory of a multivariable system and its projection models can approximate any data with a certain degree similarity between observations (Eriksson, et.al., 2013). Thus, approximation of data is more accurate the larger the number of model components used and the greater the similarity between the observations. In contrast to PCA, the knowledge related to a class membership (e.g. sample’s petrogenic or pyrogenic origin) can be used to find the location of principal components in PLS discriminant analysis, PLS-DA (Worley, et.al., 2013). As a result, PLS-DA provides a model for class separation of observations based on corresponding X-variables. For instance, PLS-DA was successfully applied to classify over five hundred environmentally occurring chemicals into reactive, less inert, inert and specifically acting compounds (Nouwen, et.al., 1997).
1.4 Objectives
The aims of this master’s degree thesis are:
- To analyze and quantify 78 PACs in 29 soil samples collected from different PAC contaminated industrial sites and city parks
- To derive, interpret and validate a MVDA model based on the chemical profiles of 78 PACs in 27 soil samples and group those according to their petrogenic, pyrogenic or urban background soil origins
- To predict the origin of two soil samples from a site contaminated with PACs with the derived PCA or PLS-DA model
12
2. Method and materials
2.1 Chemicals
All solvents used during this study were of analytical grade quality. To prevent sample contamination, ethanol (Solveco), n-hexane (SupraSolv) and dichloromethane (Fluka Analytical) were used to rinse all glassware. Silica gel 60 was purchased from Merck AB. A standard containing 36 PAHs (SRM 2260a) was purchased from National Institute of Standards and Technology (NIST). Alkylated PAHs and dibenzothiophenes mixture (S-4406-200-2T) containing 20 analytes was purchased from Chiron AS. Standards of 1-methylchrysene (99.1%), 2-methylchrysene (99.3%), 3-methylchrysene (99.3%), 7-methylbenzo[a]pyrene (98%), 7-methylbenz[a]anthracene (n/a), 7,12-dimethylbenz[a]anthracene (99.8%) were purchased from Sigma-Aldrich; dinaphtho[1,2-b;1´,2´-d]furan (>96%), 9-methylacridine (99%) and 11H-benzo[a]carbazole (99.8%) were purchased from Chiron AS; and benzo[a]fluorene (98%) was purchased from Ultra Scientific. A standard mixture contained oxy-PAHs and azaarenes (97-99.8%) purchased from Alfa Aesar, Institute for Reference Materials and Measurements, Ultra Scientific and LGC standards. An internal standard mixture with 16 deuterated PAHs (“PAH-mix 9”) purchased from Labor Dr.Ehrenstrofer-Schäfers, two deuterium-labelled alkyl-PAHs (1-methylnaphthalene-d10 and 9-methylanthracene-d12), dibenzothiophene-d8, anthraquinone-d8 and two azaarenes (acridine-d9 and carbazole-d8) purchased from Chiron AS were used as internal standards to correct losses during sample preparations. A standard solution with perylene-d12 (Sigma-Aldrich) was used to obtain the recovery of all internal standards.
To avoid a cross-contamination, Hamilton syringes that were used to prepare standards and dilute and spike samples, were rinsed with n-hexane and toluene (Fluka Analytical). Prior to the analyses, samples were stored at -18 °C.
2.2 Origin of samples
In this study, twenty nine samples, all collected in Sweden, with following origins were analyzed with GC-MS and obtained data were studied with MVDA. Eight urban background soil samples were collected in city parks in Norrköping, Stockholm and Örebro, at a distance from known contaminated with PACs industrial areas. Four samples were collected at an abandoned gas house (gaswork) that had been in operation for 37 years until 1954 in Visby. Two samples were collected from an abondoned gaswork site in Norrköping that had been used
13 for more than hundred years until 1988. Four samples were collected in Rydal at a site contaminated with gas oil. One sample was collected near to a canal lock in Örebro, an area previously used for oil handling. Two samples were collected from a site contaminated with oil from a leaking boiler central. One sample was collected from a former wood impregnation facility contaminated with creosote in Hultsfred’s industrial area. Three samples were collected at a railroad yard in Nässjö where railway sleepers were treated with creosote and stored over a longer period. Two samples were collected in Åsbro at a site contaminated with creosote used for wood impregnation. Two samples were collected at a railroad yard in Mjölby, with an unknown origin. Thus, out of 27 soil samples, 13 were petrogenic, six – pyrogenic, and eight had urban background origin.
2.3 Instrumentation
All collected soil samples were homogenized, sieved through a 2-mm sifter, extracted with Pressurized Liquid Extraction (PLE, Fluid Managements Systems, Inc.) and analyzed with Gas Chromatography Mass Spectrometry (GC-MS, Agilent Technologies).
2.3.1 PLE
Extractions were performed according to Larsson et.al. (2013) with smaller modifications. Extractions with in-cell clean up were implemented in 44-mL stainless steel extraction cells with n-hexane/dichloromethane (9:1 v/v) solvent mixture. Extraction cells were packed in four layers: i) anhydrous sodium sulfate (Na2SO4), ii) soil sample, homogenized with anhydrous
Na2SO4 (1:5 w/w), iii) thin layer of anhydrous Na2SO4, iv) four grams of basic silica.
Extractions were performed in two static cycles at 120 °C and 12 MPa for 10 min. Upon the evaporation of the samples under gentle nitrogen flow to approximately 1 mL volume, the precipitates were observed in five samples. Therefore, the soil extract of those samples were cleaned up on an open column with 15 mm inner diameter. Columns were packed with a glass wool at the bottom, followed by 5 g basic silica and 1 g anhydrous Na2SO4 at the top. The
elution was completed with 30 mL dichloromethane. Prior to the GC-MS analyses, sample solvent was changed to toluene.
2.3.2 GC-MS
To identify and quantify PACs in the soil sample extracts, GC-MS analyses were carried out on an Agilent 7890A gas chromatograph coupled to an Agilent 5975 mass spectrometer. Helium with a constant flow of 2 mL/min was used as a carrier gas. Samples with a volume 1µL were
14 injected in splitless mode at 250 °C injector’s temperature. Select PAH column (30m x 0.25mm, df=0.15µm; Agilent Technologies) and the temperature program described in the Table AI were used to separate compounds of interest. Selected Ion Monitoring (SIM) mode was used for the detection of the PACs. To detect all PACs of interest, two different runs were implemented for each sample, one for the analysis of PAHs, alkylated PAHs and dibenzothiophenes, and the second to detect oxy-PAHs and azaarenes. Quantification of analyzed 78 PACs was performed in Mass Lynx software (Waters Corporation) with the internal standard method using deuterium-labelled standards. When a respective labelled standard was lacking, compounds’ relative response factor (RRF) values were calculated using the labelled standard with the nearest retention time. Concentrations of PACs were quantified using three to four point calibration curves. Samples with PACs concentrations above the range of the calibration curve were diluted and reanalyzed. Procedure blanks were included and analyzed in all batches. The limit of detection (LOD) was determined as a mean concentration in blanks plus three standard deviations. Consequently, 27 soil samples were analyzed with GC-MS to obtain a concentration of 36 PAHs, 18 alkylated PAHs, four dibenzothiophenes (including three alkylated derivatives), 13 oxy-PAHs and seven azaarenes.
2.4 MVDA
To derive the MVDA-model, eight urban background soil samples collected in three different cities, 13 samples with petrogenic origin from five different contaminated sites, and six samples with pyrogenic origin from three other polluted locations were used.
Multivariate data analysis was performed with two different projection methods, namely, PCA and PLS-DA. Before the model derivation, raw data was evaluated and pre-processed. Concentration values of all PACs were normalized through transformation into PACs chemical profile by dividing the concentration of each PAC by the total concentration of all PACs (see Section 3.1). Pre-processing continued with unit variance scaling and mean-centering using SIMCA-P+ software (MKS Data Analytics Solutions).
Overview of the data was performed with PCA score vector plots of the complete dataset. Outliers, i.e. observations that did not fit the PCA model, were indicated using Hotelling’s T2 distribution plots and “Distance to the model in X space (DModX)” graphs. Groups of observations were studied in each class using separate scatter plots of scores and loadings. To further resolve observed groupings, PLS discriminant analysis was used. A matrix of three dummy Y-variables expressing petrogenic, pyrogenic and background origin of the samples
15 was created. To understand variables’ contribution to the separation of observed groups, the PLS weight plot was studied. Variable importance plot (VIP) was also studied to revise the importance of the variables to explain the observed grouping. Classification and discrimination properties of the derived model were checked by studying Cooman’s plots, classification lists and misclassification tables. Prediction ability of the derived model was controlled by plotting model validation graphs for each group of observations. The derived and validated PLS-DA model was then used to classify two samples with an unknown origin.
3. Results and Discussion
Recoveries of the analyzed PACs were in a range of 70%-130%, except for naphthalene, which had the minimum recovery of 63% in one of the samples. To control the repeatability of the GC-MS analysis, a quantification standard was run every tenth injection, RSD values of PAC concentrations in these standards were below 10%, thus indicating good repeatability of the method.
Total PAC concentrations in city park soils ranged from 0.16 to 7.1 mg/kg dry weight soil and from 3.4 to 3240 mg/kg dry weight soil in soils collected at PAC-contaminated industrial sites (Tables D1-D6). The proportion of native PAHs differed between samples of different origins: 84-91% in background samples, 53-92% in petrogenic samples, and 71-93% in pyrogenic samples. Consequently, the proportion of alkylated PAHs and dibenzothiophenes were higher in the petrogenic samples 47%) compared to background (3-4%) and pyrogenic samples (4-20%).
During the raw data evaluation two variables were omitted. The first variable, alkylated PAH 7,12-dimethylbenz[a]anthracene, was excluded due to the probable compound’s peak interference in GC-MS analysis, i.e. concentration in blanks were higher than in nine analyzed samples. The second variable, oxy-PAH 1,4-chrysenequinone was omitted because its concentration was below detection limit in all analyzed samples.
3.1 Principal Component Analysis
The PCA-model derivation was first performed using PAC concentrations as variables (Tables D1-D6). However, the high difference between the PAC concentrations (0.16 - 3242 ppm) resulted in that observations belonging to the samples collected at highly contaminated sites became strong outliers and therefore could not be used in the model derivation. The logarithmic
16 values of PAC concentrations with respect to bases 2, 5 and 10 were also tested as variables, without significant improvements of the PCA-model. Consequently, PAC concentrations were transformed into PACs chemical profile by dividing the concentration of each PAC by the total concentration of all PACs. If the aim of the MVDA model derivation in this study had been to classify samples of petrogenic and pyrogenic origins only, i.e. background samples with low concentrations were not taken into consideration, then the MVDA model derivation based on PAC concentrations instead of the chemical profile would probably be a better choice. It can be explained by the fact that concentration differences between samples are ignored while deriving the model based on the chemical profile, thus an important part of a statistical information is ignored as well.
After studying the score and Hotelling’s T2 distribution plots of the PCA model containing all classes of observations (background, petrogenic and pyrogenic origins) with 76 variables, it was concluded that one of the samples collected at the site, contaminated with oil from leaking boiler central, was a strong outlier (Figure B1). Thus, this observation did not fit the PCA model well and therefore was excluded from further MVDA modelling. In PCA, two major parameters are used to evaluate models, namely R2, which indicates how well the variation of a variable is explained and Q2, which is estimated by cross validation and indicates how well a variable can be predicted. Therefore, after evaluation of 76 variables with X overview plot of R2 and Q2, 19 variables with values of explained variation R2<0.45 and predicted variation Q2<0.25 were
excluded from the PCA modelling. The derived PCA model using three principal components (cumulative R2=0.76 and cumulative Q2=0.61) was used to plot a score scatter graph (Figure
17 Figure 5. PCA score plot of the second PC (t2) versus the first PC (t1). The first component explains 53% of the
variation and the second component 14%. Background samples colored in green, samples with petrogenic origin in black and samples with pyrogenic origin in red.
After PCA modelling, the following groupings were observed: i) three samples with pyrogenic origin from Nässjö, ii) three samples with pyrogenic origin from Åsbro and Hultsfred; iii) eight background samples, iv) four samples with petrogenic origin from contaminated by gaswork site in Rydal and v) three samples from gaswork in Visby grouped together with the sample from the surrounding area of a leaking boiler central. Two samples from Norrköping gaswork, a sample from Visby gaswork and a sample collected near to a canal lock in Örebro were not grouped properly. To better classify the groups of observations with different origin, a PLS discriminant analysis was performed.
3.2 Partial Least Squares Discriminant Analysis
Four principal components were used in the PLS-DA model with cumulative values of R2X=0.79, R2Y=0.86 and Q2=0.71. Partial least squares discriminant analysis resulted in better resolved groups of observations of the same origin (Figure 6).
18 Figure 6. Above: PLS-DA score plot of the second PC (t2) versus the first PC (t1). The first component explains
51% of the variation and the second component 15%. Below: PLS-DA score plot of the third PC (t3) versus the
second PC (t2). The third component explains 6% of the variation. Background samples colored in green, samples
with petrogenic origin in black and samples with pyrogenic origin in red.
The Variable Importance for the Projection (VIP) plot indicated that all variables had VIP values above 0.5; therefore, no additional variables were excluded in PLS-DA modelling (Figure B2). Hotelling’s T2 distribution and PLS-DA score plots indicated no outliers, therefore
19 26 observations were used for the classification of samples according to their origin (Figure B3). The PLS-DA model was validated with a validation plot for all three classes of observations, based on 100 permutations of the y-vector (Figure B4). The regression lines of Q2 points intersected the vertical axis below zero and all Q2 and R2 values were lower than the original points, thus the validation plots indicated that the PLS-DA model derived on 26 observations of three different classes and 57 variables was valid.
Samples with unknown origin were analyzed with the validated PLS-DA model and their origin was predicted. By observing the probability plot, it was concluded that samples from Mjölby are closely related to the samples with petrogenic origin and at the same time do not indicate any similarities with neither samples with pyrogenic origin nor background samples (Figure 7).
Figure 7. Probability plot based on values from classification list (Table C1). Columns’ color indicates the probability of class membership of each observation; green – background samples; black – petrogenic origin; red – pyrogenic origin. Samples with unknown origin marked with red rectangle.
Score scatter plots were plotted and Mjölby samples were observed near to the group of samples with petrogenic origin, thus confirming the similarities between PACs chemical profile of the named observations (Figure 8).
20 Figure 8. Above: PLS-DA score plot of the second PC (t2) versus the first PC (t1). The first component explains
50% of the variation and the second component 15%. Below: PLS-DA score plot of the third PC (t3) versus the
second PC (t2). The third component explains 6% of the variation. Background samples colored in green, samples
with petrogenic origin in black, samples with pyrogenic origin in red and samples with unknown origin in blue. A PLS weight plot corresponding to the PLS-DA score plot of the second versus the first principal, was plotted and studied (Figure 9). It was observed that all alkyl-PAHs except for 1-methylfluoranthene were close to the petrogenic dummy variable; most of the high molecular
21 weight PAHs (HMW) with four or more aromatic rings, as well as oxy-PAHs and azaarenes were located close to the background dummy variable; and two dibenzothiophenes together with the HMW PAHs pyrene, fluoranthene and benzo[ghi]fluoranthene were found near the pyrogenic dummy variable.
Figure 9. PLS-DA weight plot corresponding to the score plot of the second versus the first principal component. PACs that are close to the labelled dummy variables contribute strongly to the class separation. Priority PAHs are pink, non-priority low molecular weight PAHs (2-3 rings) are green, non-priority high molecular weight PAHs (4-7 rings) are orange, alkyl-PAHs are dark grey, dibenzothiophenes are light grey, oxy-PAHs and azaarenes are turquoise.
Thus, the contributions of PACs to the class separation could be explained to some extend by results from previous studies that showed a higher alkyl-PAH concentration in petroleum and petroleum products (Mitra, et.al., 1999; Stout, et.al., 2002); higher abundance of HMW PAHs compared to LMW PAHs in pyrogenic sources (Stout, et.al., 2015); accumulation of high molecular weight PAHs and oxy-PAHs in background samples (Stout, et.al., 2003; Andersson, et.al., 2003; Lundstedt, et.al., 2003) and possible photo- and chemical oxidation, as well as biotransformation of PAHs into oxy-PAHs in background samples (Cerniglia, 1992; Kochany & Maguire, 1994; Yu, 2002).
The PLS-DA model derivation, interpretation and validation showed that the model was valid and could be successfully applied to predict the origin of two unknown samples. However, the
22 number of variables used in the model can be discussed further. In other words, can it be possible to derive a valid MVDA model with good prediction properties using fewer variables? When all oxy-PAHs and azaarenes were excluded from the PLS-DA model, it was still possible to predict the petrogenic origin of Mjölby samples. However, the cumulative values of R2 and Q2 decreased to R2X=0.76, R2Y=0.82 and Q2=0.62, thus losing its predictive properties with 15%. In addition, three out of four observations belonging to Visby gaswork samples showed lower probability of belonging to a petrogenic class
,
now with range of 5-10%, instead of “greater than 10%” (Table C2). PLS-DA model’s usability, even without oxy-PAHs and azaarenes, can probably be explained by the fact that concentrations of those in the samples and thus contribution to PAC profile was in the range of 0.14 - 11%. Omitting alkyl-PAHs and dibenzothiophenes from MVDA modelling caused less grouping of petrogenic and background samples in PCA with three principal components (Figure B5) and declined cumulative R2X=0.65 and Q2=0.41. Thus, the predictive power of the model decreased by nearly 50%. Yet, one observation belonging to a creosote contaminated site with pyrogenic origin turned into a moderate outlier and didn’t fit the model. Therefore, no further attempts were made to decrease amount of variables based on the PAC sub-group.Out of 57 variables that were used to derive the PLS-DA model and classify samples of different origins, 22 had VIP value above one (Figure 10) and therefore can be considered as important variables (Eriksson, et.al., 2013).
23 Figure 10. VIP plot of the derived PLS-DA model with sorted by importance variables. Priority PAHs colored in pink, non-priority LMW PAHs are green, non-priority HMW PAHs are orange, alkyl-PAHs are dark grey, dibenzothiophenes are light grey, oxy-PAHs are turquoise and azaarenes are blue.
These important 22 variables can be divided into classes depending on the PACs sub-group: 11 PAHs including one LMW PAH and 10 HMW PAHs; seven alkylated PAHs, one oxy-PAH and three azaarenes. It can be also noticed that out of 11 PAHs only six are on EPA’s priority list. A PLS-DA model, based on the chemical profile of 16 priority PAHs and 16 abovementioned PACs, was derived and validated using four principal components (Figure 11).
24 Figure 11. PLS-DA loadings Bi plot corresponding to the score plot of the second versus the first principal component. Circles are scores, there priority PAHs colored in pink, non-priority HMW PAHs are orange, alkyl-PAHs are dark grey, oxy-PAH and azaarenes are turquoise. Observations are hexagons with following colors: background samples colored in green, samples with petrogenic origin in black, samples with pyrogenic origin in red and samples with unknown origin in violet.
Cumulative values of R2X=0.81, R2Y=0.87 and Q2=0.77 were only marginally increased compared to the PLS-DA model based on 57 variables (R2X=0.79, R2Y=0.86 and Q2=0.71), which can be explained by the fact that only important variables were used and thus, the noise of variations was decreased. The source of two unknown soil samples was predicted as petrogenic, which corresponded well to the results obtained after prediction using the PLS-DA model based on chemical profile of 57 PACs.
In this study, the PLS-DA model was derived using PAC chemical profile of 27 samples with known origins. Nevertheless, the model could be further validated and improved if more samples collected at other PAC contaminated sites with known origin were fitted to the derived model and their origin re-confirmed. Thus, the larger number of samples with known origin is
25 used in the MVDA model classification, the more accurate classification of the samples with unknown origin can be expected.
The results of this study shows that it would not be possible to derive a valid PLS-DA model analyzing only 16 priority PAHs. The following substances can be recommended for the analysis of PAC contaminated samples to predict a contamination source with a PLS-DA model: five additional HMW PAHs – coronene, dibenz[a,c]anthracene, benzo[ghi]fluor-anthene, benzo[c]phenanthrene and triphenylene; seven alkylated PAHs – 1,2,6-trimethylphen-anthrene, 1-methylchrysene, 1,2,8-trimethylphen1,2,6-trimethylphen-anthrene, 2,3-dimethylanthracene, 2-methylphenanthrene, 1,6-dimethylnaphthalene and 7-methylbenzo[a]pyrene; one oxy-PAH 4H-cyclopenta[d,e,f]phenanthrenone; and three azaarenes – dibenz[a,h]acridine, carbazole and 9-methylacridine.
Conclusion
The validated PLS-DA model was used for the classification of samples with background, petrogenic and pyrogenic origins, based on their PACs profiles, including PAHs, alkylated PAHs, dibenzothiophenes, oxy-PAHs and azaarenes. Twenty-six samples with known sources of contamination were class specified and two samples with an unknown source were fitted to the derived PLS-DA model to predict their origin. The derived PLS-DA model has a potential to predict the petrogenic, pyrogenic and background origins of samples collected at PAC contaminated sites. To predict the sources of contamination at PAC contaminated sites the following 16 PACs, besides 16 priority PAHs, are recommended to be included in the chemical analysis: five HMW PAHs – coronene, dibenz[a,c]anthracene, benzo[ghi]fluoranthene, benzo[c]phenanthrene and triphenylene; seven alkylated PAHs – 1,2,6-trimethylphenanthrene, 1-methylchrysene, 1,2,8-trimethylphenanthrene, 2,3-dimethylanthracene, 2-methylphen-anthrene, 1,6-dimethylnaphthalene and 7-methylbenzo[a]pyrene; one oxy-PAH – 4H-cyclo-penta[d,e,f]phenanthrenone; and three azaarenes – dibenz[a,h]acridine, carbazole and 9-methylacridine.
26
Acknowledgements
I would like to thank my academic advisor Maria Larsson for her priceless advices and encouragement during this master’s thesis. I had your constant support but at the same time I always felt free to make my own decisions, which I’m very grateful for. I wish to express my gratitude to my second academic advisor, Hans F. Grahn, for his well-timed help with
multivariate data analysis. I would also like to thank Monika Lam for her help with
everything at the lab. And sincerely thank my wife for taking care of so many practical things at home during this thesis, for her patience and of course for her unconditional love.
27
References
94/67/EEC (1994) Council directive on the incineration of hazardous waste. The Council of the European Communities.
Andersson, B., Lundstedt, S., Tornberg, K., Schnurer, Y., Oberg, L. & Mattiasson, B. (2003) "Incomplete degradation of polycyclic aromatic hydrocarbons in soil inoculated with wood-rotting fungi and their effect on the indigenous soil bacteria", Environmental Toxicology and Chemistry, 22(6):1238-1243.
Andersson, J.T. & Achten, C. (2015) "Time to Say Goodbye to the 16 EPA PAHs? Toward an Up-to-Date Use of PACs for Environmental Purposes", Polycyclic Aromatic Compounds, 35(2-4):330-354.
Baek, S.O., Field, R.A., Goldstone, M.E., Kirk, P.W., Lester, J.N. & Perry, R. (1991) “A review of Polycyclic Aromatic Hydrocarbons - sources, fate and behaviour”, Water, Air and Soil Pollution, 60:279-300.
Bertilsson, S. & Widenfalk, A. (2002) “Photochemical degradation of PAHs in freshwaters and their impact on bacterial growth – influence of water chemistry”, Hydrobiologia, 469:23– 32.
Bjerrum, J.T., Nielsen, O.H., Wang, Y.L. & Olsen, J. (2008) “Technology Insight: metabonomics in gastroenterology—basic principles and potential clinical applications”, Nature Clinical Practice Gastroenterology and Hepatology, 5(6):332-43.
Boehm, P.D., Neff, J.M. & Page, D.S. (2007) “Assessment of polycyclic aromatic
hydrocarbon exposure in the waters of Prince William Sound after Exxon Valdez oil spill: 1989–2005”, Marine Pollution Bulletin, 54:339–367.
Bollag, J.M. & Kaiser, J.P. (1991) "The transformation of heterocyclic aromatic compounds and their derivatives under anaerobic conditions", Critical Reviews in Environmental Control, 21:297-329.
Bucheli, T.D., Blum, F., Desaules, A. & Gustafsson, O. (2004) “Polycyclic aromatic
hydrocarbons, black carbon, and molecular markers in soils of Switzerland”, Chemosphere, 56:1061–1076
Cerniglia, C.E. (1992) “Biodegradation of polycyclic aromatic hydrocarbons”, Biodegradation, 3:351-368.
Chen, H. & Preston, M.R. (1998) "Azaarenes in the aerosol of an urban atmosphere", Environmental Science and Technology, 32(5):577-583.
Eriksson, L., Byrne, T., Johansson, E., Trygg, J. & Vikström C. (2013) ”Multi- and Megavariate Data Analysis. Basic Principles and Applications”, 3rd revised ed., MKS Umetrics AB, Malmö, Sweden. ISBN 9789197373050
Farrington, J.W., Frew, N.M., Gschwend, P.M. & Tripp, B.W. (1977) “Hydrocarbons in cores of Northwestern Atlantic coastal and continental margin sediments”, Estuarine, Coastal and Marine Science, 5:793–808.
28 Grimmer, G., Jacob, J. & Naujack, K.W. (1981) “Profile of the polycyclic aromatic
hydrocarbons from lubricating oils inventory by GCGC/MS - PAH in environmental materials, Part 1”, Fresenius’ Journal of Analytical Chemistry, 306:347–355.
Grimmer, G., Jacob, J. & Naujack, K.W. (1983) “Profile of the polycyclic aromatic
compounds from crude oils inventory by GCGC/MS - PAH in environmental materials, Part 3”, Fresenius’ Journal of Analytical Chemistry, 314:29–36.
Hashimoto, T., Negishi, T., Namba, T., Hayakawa, S. & Hayatsu, H. (1979) "Mutagenicity of Quinoline Derivatives and Analogs – Quinoxaline 1,4-Dioxide is a Potent Mutagen",
Chemical and Pharmaceutical Bulletin, 27(8):1954-1956.
Ho, T.C. (2004) "Deep HDS of diesel fuel: chemistry and catalysis", Catalysis Today, 98(1):3-18.
Howsam, M. & Jones, K.C. (1998) “Sources of PAHs in the environment”. The Handbook of Environmental Chemistry, Vol. 3, Anthropogenic compounds. Part I. PAHs and Related Compounds. Neilson. A.H. (ed). Springer-Verlag. Berlin, 137-174.
Irwin, R.J., Van Mouwerik, M., Stevens, L., Seese, M.D. & Basham, W. (1997) “Entry on alkyl PAHs”. Environmental contaminants encyclopedia. National Park Service, Water Resources Division, Fort Collins, Colorado.
Iqbal, J., Overton, E.B. & Gisclair, D. (2008) “Polycyclic aromatic hydrocarbons in Louisiana Rivers and coastal environments: source fingerprinting and forensic analysis”. Environmental Forensics, 9:63–74.
Jackson, J.E. (1991) “A User’s Guide to Principal Components”, John Wiley, New York. ISBN 0-471-62267-2.
Kochany, J. & Maguire, R.J. (1994) “Abiotic transformations of polynuclear aromatic hydrocarbons and polynuclear aromatic nitrogen heterocycles in aquatic environments”, Science of the Total Environment, 144:17-31.
Kohler M. & Künniger T. (2003) “Emissions of polycyclic aromatic hydrocarbons (PAH) from creosoted railroad ties and their relevance for life cycle assessment (LCA)”. Holz als Roh- und Werkstoff 61:117-124.
Kozin, I.S., A. Larsen, O.F., de Voogt, P., Gooijer, C. & Velthorst, N.H. (1997) "Isomer-specific detection of azaarenes in environmental samples by Shpol'skii luminescence spectroscopy", Analytica Chimica Acta, 354(1):181-187.
Laflamme, R.E. & Hites, R.A. (1978) “The global distribution of polycyclic aromatic hydrocarbons in recent sediments”, Geochimica et Cosmochimica Acta, 42:289–303. Larsson, M., Hagberg, J., Rotander, A., van Bavel, B. & Engwall, M. (2013) ”Chemical and bioanalytical characterisation of PAHs in risk assessment of remediated PAH-contaminated soils”, Environmental Science and Pollution Research International, 20(12):8511-8520. Lundstedt, S., Haglund, P. & Öberg, L.G. (2003) ”Degradation and formation of polycyclic aromatic compounds during bio-slurry treatment of an aged gasworks soil”, Environmental Toxicology and Chemistry, 22:1413-1420.
29 Lundstedt, S., Persson, Y., Öberg, L. (2006) "Transformation of PAHs during ethanol-Fenton treatment of an aged gasworks’ soil", Chemosphere, 65(8):1288-1294.
Lundstedt, S., White, P.A., Lemieux, C.L., Lynes, K.D., Lambert, I.B., Öberg, L., Haglund, P. & Tysklind, M. (2007) "Sources, Fate, and Toxic Hazards of Oxygenated Polycyclic
Aromatic Hydrocarbons (PAHs) at PAH-contaminated Sites", AMBIO, 36(6):475-485. Mastrangelo, G., Fadda, E. & Marzia, V. (1996) “Polycyclic aromatic hydrocarbons and cancer in man”, Environmental Health Perspectives, 104(11):1166-1170.
Matsuoka, A., Shudo, K., Saito, Y., Sofuni, T. & Ishidate, M. Jr. (1982) "Clastogenic potential of heavy oil extracts and some aza-arenes in Chinese hamster cells in culture", Mutation Research/Genetic Toxicology, 102(3):275-283.
Mitra, S., Dellapenna, T.M. & Dickhut, R.M. (1999) “Polycyclic aromatic hydrocarbon distribution within Lower Hudson River estuarine sediments: physical mixing vs sediment geochemistry”, Estuarine, Coastal and Shelf Science, 49:311–326.
Mueller, J., Chapman, P. & Pritchard, P. (1989) "Creosote-contaminated sites - Their potential for bioremidiation", Environmental Science & Technology, 23(10):1197-1201. Murphy, B.L. & Brown, J. (2005) "Environmental Forensics Aspects of PAHs from Wood Treatment with Creosote Compounds", Environmental Forensics, 6(2):151-159.
Nouwen, J., Lindgren, F., Hansen, B. & Karcher, W. (1997) ”Classification of
Environmentally Occurring Chemicals Using Structural Fragments and PLS Discriminant Analysis”, Environmental Science and Technology, 31(8):2313-2318.
Pavlova, A. & Ivanova, R. (2003) “Determination of petroleum hydrocarbons and polycyclic aromatic hydrocarbons in sludge from wastewater treatment basins”, Journal of
Environmental Monitoring, 5:319–323.
Pereira, W.E., Rostad, C.E., Updegraff, D.M. & Bennett, J.L. (1987) "Fate and movement of azaarenes and their anaerobic biotransformation products in an aquifer contaminated by wood-treatment chemicals", Environmental Toxicology and Chemistry, 6(3):163-176. Peters, K.E., Walters, C.C. & Moldowan, M. (2005) “The biomarker guide. Biomarkers and isotopes in the environment and human history”, vol 1, 2nd edn., Cambridge University Press, Cambridge, 296–312. ISBN 9780521781589.
Richter, B.E., Jones, B.A., Ezzell, J.L., Porter, N.L., Avdalovic, N. & Pohl, C. (1996) “Accelerated solvent extraction: a technique for sample preparation”, Analytical Chemistry, 68:1033–1039.
Saber, D., Mauro, D. & Sirivedhin, T. (2006) “Environmental forensic investigation in sediments near a former manufactured gas plant site”, Environmental Forensics, 7:65–75.
30 Short, J.W., Irvine, G.V., Mann, D.H., Maselko, J.M., Pella, J.J., Lindeberg, M.R., Payne, J.R., Driskell, W.B. & Rice, S.D. (2007) “Slightly weathered Exxon Valdez oil persists in gulf of Alaska beach sediments after 16 years”, Environmental Science & Technology, 41:1245– 1250.
Sicre, M.A., Marty, J.C., Saliot, A., Aparicio, X., Grimalt, J. & Albaiges, J. (1987) “Aliphatic and aromatic hydrocarbons in different sized aerosols over the Mediterranean Sea: occurrence and origin”, Atmospheric Environment, 21:2247–2259.
Stark, A., Abrajano, T., Hellou, J. & Metcalf-Smith, J.L. (2003) “Molecular and isotopic characterization of polycyclic aromatic hydrocarbon distribution and sources at the international segment of the St. Lawrence River”, Organic Geochemestry, 34:225–237. Stenberg, U.R. & Alsberg, T.E. (1981) "Vacuum sublimation and solvent extraction of polycyclic aromatic compounds adsorbed on carbonaceous materials", Analytical Chemistry, 53(13):2067-2072.
Stogiannidis, E. & Laane, R. (2015) "Source Characterization of Polycyclic Aromatic Hydrocarbons by Using Their Molecular Indices: An Overview of Possibilities", Reviews of Environmental Contamination and Toxicology, 234:49-133.
Stout, S.A., Magar, V.X., Uhler, R.M., McCarthy, K.J. & Emsbo-Mattingly, S. (2002) “Chemical fingerprinting of hydrocarbons”, in Murphy, B.L. & Morrison, R.D. (eds). Introduction to environmental forensics, Academic, New York, NY, 137–260. ISBN 9780124046962.
Stout, S.A., Leather, J.M. & Corl, W.E. (2003) “A user’s guide for determining the sources of contaminants in sediments”, Technical report 1907. Spawar Systems Center, San Diego, CA, 1–85.
Stout, S.A., Emsbo-Mattingly, S.D., Douglas, G.S., Uhler, A.D. & McCarthy, K.J. (2015) "Beyond 16 Priority Pollutant PAHs: A Review of PACs used in Environmental Forensic Chemistry", Polycyclic Aromatic Compounds, 35(2-4):285-315.
Tyrpień, K., Bodzek, D. & Janoszka, B. (1995) “Application of TLC and GC-MS to the identification of azaarenes in sewage sludges”, Journal of Planar Chromatography, 8:75-78. Van Grevenynghe, J., Rion, S., Le Ferrec, E., Le Vee, M., Amiot, L., Fauchet, R. & Fardel, O. (2003) “Polycyclic aromatic hydrocarbons inhibit differentiation of human monocytes into macrophages”, Journal of Immunology, 170(5):2374-2381.
Wang, Z., Fingas, M. & Page, D.S. (1999) “Oil spill identification”, Journal of Chromatography A, 843:369–411.
Windsor, J.G. & Hites, R.A. (1979) “Polycyclic aromatic hydrocarbons in Gulf of Maine sediments and Nova Scotia soils”, Geochimica et Cosmochimica Acta, 43:27–33.
31 Wold, S., Albano, C., Dunn, W.J., Edlund, U., Esbensen, K., Geladi, P., Hellber, S.,
Johansson, E., Lindberg, W. & Sjöström, M. (1984) ”Multivariate Data Analysis in
Chemistry”, In: B.R. Kowalski (ed.), Chemometrics: Mathematics and Statistics in Chemstry, D. Reidel Publishing Company, Dordrecht, Holland. ISBN 9789048184071.
Worley, B., Halouska, S. & Powers, R. (2013) “Utilities for quantifying separation in PCA/PLS-DA score plots”
Yu, H. (2002) "Environmental carcinogenic polycyclic aromatic hydrocarbons:
Photochemistry and phototoxicity", Journal of Environmental Science and Health, 20(2):149-183.
Zakaria, M.P., Takada, H., Tsutsumi, S., Ohno, K., Yamada, J., Kound, E. & Kumata, H. (2002) “Distribution of polycyclic aromatic hydrocarbons (PAHs) in rivers and estuaries in Malaysia: a widespread input of petrogenic PAHs”, Environmental Science & Technology, 36:1907–1918.
32
Appendix A – GC-MS parameters
Table A1. Parameters and conditions for the GC-MS analyses. Conditions
Column Select PAH, 30m x 0.25mm, df 0.15µm
Injection volume 1µL
Injector 250 °C, 1 min at 50 mL/min, splitless mode
Carrier gas Helium at constant flow 2 mL/min
Temperature 70 °C (0 min), 8°C/min to 205 °C (2 min), 8 °C/min to 250 °C, 3 °C/min to 270 °C (2 min), 9 °C/min to 279 °C, 1 °C/min to
280 °C (3 min), 5 °C/min to 325 °C (5 min) SIM Parameters
(Mass, Dwell time)
oxy-PAHs and azaarenes:
Group 1: ( 129.00, 30 ) ( 132.00, 30 ) ( 167.00, 30 ) ( 175.00, 30 ) ( 179.00, 30 ) ( 180.00, 30 ) ( 188.00, 30 ) ( 193.00, 30 ) ( 195.00, 30 ) ( 204.00, 30 ) ( 208.00, 30 ) ( 216.00, 30 ) ( 217.00, 30 ) ( 218.00, 30 ) ( 222.00, 30 ) ( 230.00, 30 ) Group 2: ( 193.00, 30 ) ( 195.00, 30 ) ( 217.00, 30 ) ( 218.00, 30 ) ( 222.00, 30 ) ( 230.00, 30 ) ( 236.00, 30 ) ( 254.00, 30 ) ( 258.00, 30 ) ( 264.00, 30 ) ( 268.00, 30 ) ( 270.00, 30 ) ( 279.00, 30 )
PAHs, alkylated PAHs and dibenzothiophenes
Group 1: ( 128.00, 30 ) ( 136.00, 30 ) ( 142.00, 30 ) ( 152.00, 30 ) ( 154.00, 30 ) ( 156.00, 30 ) ( 160.00, 30 ) ( 164.00, 30 ) ( 166.00, 30 ) ( 170.00, 30 ) ( 176.00, 30 ) Group 2: ( 178.00, 30 ) ( 184.00, 30 ) ( 188.00, 30 ) ( 190.00, 30 ) ( 192.00, 30 ) ( 198.00, 30 ) ( 202.00, 30 ) ( 204.00, 30 ) ( 206.00, 30 ) ( 212.00, 30 ) ( 214.00, 30 ) ( 216.00, 30 ) ( 220.00, 30 ) ( 226.00, 30 ) ( 228.00, 30 ) ( 230.00, 30 ) Group 3: ( 226.00, 30 ) ( 228.00, 30 ) ( 240.00, 30 ) ( 242.00, 30 ) ( 252.00, 30 ) ( 256.00, 30 ) ( 264.00, 30 ) ( 266.00, 30 ) ( 276.00, 30 ) ( 278.00, 30 ) ( 288.00, 30 ) ( 292.00, 30 ) ( 300.00, 30 ) ( 302.00, 30 )
33
Appendix B – Additional Figures
Figure B1. Observation of the sample collected at the site contaminated with oil from a leaking boiler central labelled “OilRigs_2” on the score scatter plot (above) is a strong outlier, which confirms with a Hotelling’s T2
34 Figure B2. VIP plot of the derived PLS-DA model with sorted by importance variables. PAHs colored in orange, alkyl-PAHs in dark grey, and oxy-PAHs and azaarenes in turquoise.
35 Figure B3. Hotelling’s T2 distribution plot of 26 observations indicating that no outliers are detected in the
PLS-DA model. Background samples colored in green, samples with petrogenic origin in black and samples with pyrogenic origin in red.
36 Figure B4. Validation plots based on 100 permutations of the y-vector for background samples (above), samples with petrogenic origin (middle) and samples with pyrogenic samples (below).
37 Figure B5. PCA score plot of the second PC (t2) versus the first PC (t1). The first component explains 45% of the
variation and the second component 11%. After excluding alkyl-PAHs from PCA, one observation (labelled with ”CreÅs_1”) of sample collected at contaminated by creosote site turned into outlier. Background samples colored in green, samples with petrogenic origin in black and samples with pyrogenic origin in red.
38
Appendix C – Additional Tables
Table C1. The classification list for 28 observations with probability of class membership in each column based on 57 variables. The color behind the values indicates the probability of belonging to a model: white – below 5%, orange – between 5% and 10%, green – greater than 10%.
Sample Origin Background Petrogenic Pyrogenic
Oskarsparken Background 0.75 0.14 0.11 Sveaparken Background 0.85 -0.01 0.16 Karlaparken Background 0.76 0.15 0.09 Skytteparken Background 1.18 -0.45 0.27 Norrkoping_1 Background 0.93 0.22 -0.15 Norrkoping_2 Background 0.94 0.14 -0.08 Hummlegarden Background 0.85 0.24 -0.09 Sundbyberg Background 0.67 0.27 0.07 GasVisby_1-1 Petrogenic 0.13 0.93 -0.07 GasVisby_1-2 Petrogenic 0.21 0.86 -0.08 GasVisby_2 Petrogenic 0.17 0.69 0.15 GasVisby_3 Petrogenic -0.10 0.79 0.31 GasRydal_1-1 Petrogenic -0.02 1.02 0.00 GasRydal_1-2 Petrogenic 0.07 0.95 -0.03 GasRydal_1-3 Petrogenic -0.28 1.17 0.11 GasRydal_2 Petrogenic 0.04 0.91 0.05 GasNorr_1 Petrogenic 0.04 0.99 -0.03 GasNorr_2 Petrogenic 0.14 0.97 -0.11 OilRgs_1 Petrogenic 0.01 0.94 0.05 OilÖre_1 Petrogenic 0.28 0.95 -0.23 CreNäs_1-1 Pyrogenic -0.11 0.13 0.99 CreNäs_1-2 Pyrogenic 0.17 0.03 0.79 CreNäs_2 Pyrogenic -0.32 0.19 1.14 CreHult_1 Pyrogenic 0.08 -0.02 0.94 CreÅs_1 Pyrogenic 0.00 0.13 0.87 CreÅs_2 Pyrogenic 0.57 -0.33 0.76 UnMjöl_1 Unknown -0.08 1.19 -0.11 UnMjöl_2 Unknown 0.01 0.89 0.10
39 Table C2. The classification list for 28 observations with probability of class membership in each column based on profile of PAHs, alkyl PAHs and dibenzothiophenes. The color behind the values indicates the probability of belonging to a model: white – below 5%, orange – between 5% and 10%, green – greater than 10%.
Sample Origin Background Petrogenic Pyrogenic
Oskarsparken Background 0.78 0.10 0.13 Sveaparken Background 0.78 0.12 0.10 Karlaparken Background 0.73 0.21 0.06 Skytteparken Background 1.02 -0.26 0.24 Norrkoping_1 Background 0.88 0.35 -0.22 Norrkoping_2 Background 0.88 0.29 -0.16 Hummlegarden Background 0.95 0.12 -0.07 Sundbyberg Background 0.72 0.27 0.01 GasVisby_1-1 Petrogenic 0.32 0.61 0.06 GasVisby_1-2 Petrogenic 0.32 0.74 -0.06 GasVisby_2 Petrogenic 0.28 0.62 0.11 GasVisby_3 Petrogenic 0.16 0.48 0.36 GasRydal_1-1 Petrogenic 0.02 1.00 -0.02 GasRydal_1-2 Petrogenic 0.10 0.93 -0.02 GasRydal_1-3 Petrogenic -0.22 1.22 0.00 GasRydal_2 Petrogenic 0.04 0.91 0.05 GasNorr_1 Petrogenic -0.20 1.15 0.05 GasNorr_2 Petrogenic 0.26 0.80 -0.06 OilRgs_1 Petrogenic -0.19 1.17 0.02 OilÖre_1 Petrogenic 0.30 0.86 -0.17 CreNäs_1-1 Pyrogenic -0.14 0.24 0.90 CreNäs_1-2 Pyrogenic 0.16 -0.04 0.88 CreNäs_2 Pyrogenic -0.37 0.33 1.04 CreHult_1 Pyrogenic 0.08 0.01 0.91 CreÅs_1 Pyrogenic -0.08 0.09 0.99 CreÅs_2 Pyrogenic 0.45 -0.32 0.87 UnMjöl_1 Unknown 0.08 0.98 -0.06 UnMjöl_2 Unknown 0.10 0.73 0.18
Appendix D – Raw data
Table D1. PAH concentrations in background samples, ng/g soil dry weight (n.d.=not detected).
Substance Oskarsparken Sveaparken Karlaparken Skytteparken Norrkoping_1 Norrkoping_2 Hummlegarden Sundbyberg
Naphtalene 2.2 0.5 2.3 0.4 3.3 7.0 1.6 0.6 Biphenyl 0.5 0.2 0.5 0.2 1.1 1.5 0.5 0.4 Acenaphthylene 10.5 4.4 18.0 0.9 16.2 15.0 7.2 3.1 Acenaphthene 1.2 0.2 1.8 0.0 11.1 9.8 1.4 2.1 Fluorene 2.9 0.8 3.8 0.1 10.5 10.3 1.4 3.3 Phenanthrene 54.0 17.0 81.4 4.5 202.4 189.9 39.5 43.9 Anthracene 6.0 2.7 10.7 0.5 33.4 30.6 5.5 7.0 4H-Cyclopenta[def]phenanthrene 9.3 3.3 15.3 0.9 30.8 27.0 7.2 10.7 Fluoranthene 144.2 59.9 236.6 13.3 473.4 445.5 110.6 149.9 Pyrene 138.0 57.7 226.2 11.7 476.8 449.9 107.7 145.7 Benzo(a)fluorene 22.4 9.3 36.9 2.5 86.3 74.4 18.2 43.0 Benzo[ghi]fluoranthene 5.5 2.3 8.3 0.3 20.7 19.4 4.1 6.4 Cyclopenta[cd]pyrene 21.5 6.9 42.7 1.8 71.0 51.6 15.3 7.9 Benzo[c]phenanthrene 11.6 5.0 17.7 1.4 43.1 39.2 9.0 14.9 Benzo(a)anthracene 80.1 35.3 137.6 8.8 392.5 354.6 70.0 125.6 Chrysene 55.6 25.1 89.7 6.5 217.9 199.2 45.5 70.1 Triphenylene 21.7 9.3 36.9 2.9 90.5 83.6 21.3 32.4 Benzo[b]fluoranthene 98.6 45.2 164.0 12.2 459.5 417.8 93.2 120.7 Benzo[j]fluoranthene 45.8 21.7 68.0 5.5 202.0 185.2 40.8 51.6 Benzo[k]fluoranthene 75.4 31.9 128.1 8.1 386.1 336.6 66.0 101.2 Benzo[a]fluoranthene 25.1 10.6 44.0 3.1 161.2 138.2 24.1 40.4 Benzo[e]pyrene 72.0 32.0 115.3 7.6 354.8 323.3 67.0 86.5 Benzo[a]pyrene 95.3 40.9 160.1 10.3 539.5 481.1 89.1 119.7 Perylene 23.5 9.6 43.3 2.8 150.7 131.4 25.0 34.0 Indeno[1,2,3-cd]pyrene 140.8 65.8 236.3 17.1 823.2 736.1 148.5 154.7 Benzo[g,h,i]perylene 62.2 25.8 109.1 3.8 365.4 344.1 68.6 68.0 Anthanthrene 16.9 6.2 31.0 1.6 131.8 101.4 24.9 24.1 Dibenz[a,h]anthracene 7.9 4.0 14.1 0.9 53.4 46.3 9.1 13.0 Dibenz[a,c]anthracene 15.5 6.8 25.9 2.4 87.6 72.4 16.5 23.3 Dibenz[a,j]anthracene 8.4 3.6 13.1 0.8 38.2 34.0 7.8 10.0 Picene 19.8 8.0 28.6 2.2 107.1 100.7 20.8 22.8 Benzo[b]chrysene 12.2 5.3 23.1 1.4 106.5 91.8 15.1 24.2 Coronene 18.2 8.2 28.1 2.7 95.5 86.7 23.8 21.5 Dibenzo[b,k]fluoranthene 16.9 7.2 34.6 n.d. 136.2 111.0 25.0 32.4 Dibenzo[a,e]pyrene 11.9 6.6 23.9 n.d. 94.3 81.3 19.0 21.0 naphtho[2,3-a]pyrene n.d. n.d. n.d. n.d. 13.9 9.2 n.d. n.d.
41 Table D2. Alkylated PAH, dibenzothiophene, oxy-PAH and azaarene concentrations in background samples, ng/g soil dry weight (n.d.=not detected).
Substance Oskarsparken Sveaparken Karlaparken Skytteparken Norrkoping_1 Norrkoping_2 Hummlegarden Sundbyberg
2-Methylnaphthalene 1.3 0.1 1.0 0.4 2.6 3.2 0.5 0.6 1-Methylnaphthalene 0.9 0.1 0.6 0.2 1.9 2.2 0.3 0.4 1,6-Dimethylnaphthalene 1.8 0.6 1.4 0.5 3.8 3.9 1.0 1.3 2,3,5-Trimethylnaphthalene 0.4 0.1 0.7 n.d. 2.1 1.6 0.3 0.5 Dibenzothiophene 2.3 0.7 3.7 0.3 10.7 8.9 1.8 2.0 2-Methyldibenzothiophene 0.6 0.2 1.7 n.d. 3.0 3.0 0.9 1.2 2-Methylphenanthrene 14.1 4.9 19.4 1.6 40.1 34.4 10.2 15.8 2-Methylanthracene 3.2 1.2 4.3 0.4 10.4 9.0 2.2 4.8 2,8-Dimethyldibenzothiophene n.d. n.d. 0.3 n.d. 0.7 0.4 n.d. n.d. 2,4-Dimethylphenanthrene n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. 2,4,7-Trimethyldibenzothiophene n.d. 0.0 0.2 n.d. 0.4 0.4 0.1 0.3 2,3-Dimethylanthracene 1.1 0.4 1.6 0.2 3.6 3.1 0.8 1.6 1-Methylfluoranthene 10.0 4.2 13.6 1.2 28.5 26.1 7.9 12.6 1,2,8-Trimethylphenanthrene 0.6 0.4 1.4 n.d. 3.2 2.7 0.5 1.4 1,2,6-Trimethylphenanthrene 0.9 0.3 1.1 n.d. 2.3 2.0 0.5 0.9 7-Methylbenz(a)anthracene 1.4 0.7 3.1 n.d. 12.0 9.8 1.7 4.3 3-Methylchrysene 5.4 2.7 7.8 0.9 19.0 15.4 4.1 8.9 2-Methylchrysene 9.8 4.1 13.6 1.2 30.2 25.9 7.5 14.1 1-Methylchrysene 5.1 2.1 7.1 0.5 19.8 17.1 4.0 7.6 6-Ethylchrysene n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. 7-Methylbenzo(a)pyrene n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. Quinoline n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. 1-Indanone n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. Carbazole 8.0 2.6 12.0 0.7 40.8 35.0 5.9 4.0 Benzo[h]quinoline 0.3 0.0 0.2 0.0 1.1 0.8 0.3 0.1 Acridine 0.2 n.d. 0.3 0.1 0.6 n.d. 0.3 0.3 9-Fluorenone 3.8 1.1 4.0 0.6 6.7 6.5 1.8 1.6 4H-Cyclopenta[def]phenanthrenone 7.2 3.1 10.0 0.9 15.5 15.2 4.2 8.2 Anthracene-9,10-dione 19.8 7.1 25.9 2.7 26.0 25.1 14.1 8.8 9-Methylacridine n.d. n.d. 0.4 n.d. 0.6 0.6 0.3 0.3 11H-Benzo[a]carbazole 16.9 6.3 23.1 1.6 80.7 68.7 15.2 15.3 2-Methylanthracene-9,10-dione 4.9 1.9 6.4 0.8 7.8 8.0 3.5 2.9 Benzo[a]fluorenone 19.4 9.0 29.0 2.2 53.6 54.7 13.3 22.6 7H-Benz[de]anthracen-7-dione 23.4 8.7 32.9 1.4 67.2 58.2 16.0 16.7 6-HBenzo[cd]pyren-6-one 22.6 9.0 38.5 4.8 94.4 80.5 22.5 16.3 Benz[a]anthracene-7,12-dione 14.8 6.1 19.4 1.8 25.1 24.0 9.4 9.2 Naphthacene-5,12-dione 8.8 3.7 12.3 0.6 34.7 30.1 7.3 16.3 Dinaphtho[1,2-b;1,2d]furan 1.0 0.5 1.8 0.1 4.4 3.6 0.9 1.2 9,10-dihydrobenzo[a]pyren-(8H)-none n.d. n.d. n.d. n.d. 0.5 n.d. n.d. n.d. dibenz[ah]acridine 1.6 0.3 1.8 n.d. 6.4 3.7 3.1 4.2
42 Table D3. PAH concentrations in petrogenic samples, ng/g soil dry weight (n.d.=not detected).
Substance GasVisby_1-1 GasVisby_1-2 GasVisby_2 GasVisby_3 GasRydal_1-1 GasRydal_1-2 GasRydal_1-3 GasRydal_2 GasNorr_1 GasNorr_2 OilRgs_1 OilRgs_2 OilÖre_1
Naphtalene 28 325 69 4988 11567 16198 47 73665 282 2 6 3 6 Biphenyl 10 51 6 727 14602 45260 156 21967 58 3 2 5 1 Acenaphthylene 79 349 144 4958 54764 158243 931 75968 560 97 54 67 32 Acenaphthene 10 222 53 3032 17958 49727 355 24877 165 33 5 38 1 Fluorene 41 406 149 7752 36537 91425 934 48514 563 29 31 58 4 Phenanthrene 689 5781 2074 72047 249064 591087 8733 364453 5245 795 385 432 62 Anthracene 147 1523 801 17594 34173 75248 1372 46014 1546 383 57 59 25 4H-Cyclopenta[def]phenanthrene 126 949 613 8013 43123 110317 1771 59880 833 158 75 126 20 Fluoranthene 1178 12819 7590 117390 40488 93882 1635 65144 7035 2783 818 936 341 Pyrene 964 10052 5695 91189 51702 115880 2286 79284 5692 2221 672 876 271 Benzo(a)fluorene 175 1930 1139 6133 13111 30827 544 18468 1048 419 92 130 52 Benzo[ghi]fluoranthene 44 391 227 2169 1386 2895 49 1998 266 106 24 31 11 Cyclopenta[cd]pyrene 72 483 190 6072 6438 14524 323 8664 1623 111 99 103 68 Benzo[c]phenanthrene 73 797 477 3736 4595 9857 163 6471 621 235 48 63 23 Benzo(a)anthracene 573 7891 4570 36677 30698 64751 1068 43745 4467 2030 303 348 206 Chrysene 416 4183 2904 21031 22816 48704 1039 31730 2574 1581 264 341 146 Triphenylene 168 1383 819 5661 1091 2757 46 2108 889 285 85 163 41 Benzo[b]fluoranthene 504 6592 3622 30528 6411 14345 267 10439 3375 1959 290 335 210 Benzo[j]fluoranthene 251 3112 1656 12580 6677 13929 219 9248 2486 972 149 160 99 Benzo[k]fluoranthene 460 6188 3388 28367 7118 17059 305 12289 3196 1806 271 311 158 Benzo[a]fluoranthene 177 2552 1114 7334 2548 5679 120 4072 1149 395 89 94 56 Benzo[e]pyrene 482 4977 2617 20332 8481 17514 345 12412 3127 1474 241 300 154 Benzo[a]pyrene 534 7381 3719 30963 14944 31026 624 21859 3106 1897 307 346 249 Perylene 211 2210 1067 9755 2858 5410 107 3799 895 416 96 134 84 Indeno[1,2,3-cd]pyrene 704 10537 4381 58417 11757 27072 349 19327 2836 1788 397 431 196 Benzo[g,h,i]perylene 422 5410 2104 22659 3943 8861 173 5967 2193 1214 184 209 127 Anthanthrene 100 1157 516 10224 2345 5022 42 3464 393 360 44 49 66 Dibenz[a,h]anthracene 58 875 395 831 1945 4497 60 2872 389 207 31 36 23 Dibenz[a,c]anthracene 100 1157 516 1706 1211 3127 42 3038 580 127 44 49 13 Dibenz[a,j]anthracene 42 608 283 1015 1760 3749 59 2274 360 144 25 26 15 Picene 110 1355 531 2255 5532 12685 161 7586 755 505 63 60 46 Benzo[b]chrysene 84 1243 492 3434 2490 6360 113 4203 754 319 41 42 30 Coronene 98 834 258 3677 500 1036 37 650 663 258 25 38 13 Dibenzo[b,k]fluoranthene 173 1350 561 1443 2355 4607 128 3149 1281 606 74 76 62 Dibenzo[a,e]pyrene 80 1221 518 3454 1368 2950 75 2223 741 379 51 56 39 naphtho[2,3-a]pyrene n.d. 263 86 n.d. n.d. n.d. 42 n.d. n.d. n.d. n.d. n.d. 0
43 Table D4. Alkylated PAH, dibenzothiophene, oxy-PAH and azaarene concentrations in petrogenic samples, ng/g soil dry weight (n.d.=not detected).
Substance GasVisby_1-1 GasVisby_1-2 GasVisby_2 GasVisby_3 GasRydal_1-1 GasRydal_1-2 GasRydal_1-3 GasRydal_2 GasNorr_1 GasNorr_2 OilRgs_1 OilRgs_2 OilÖre_1
2-Methylnaphthalene 34 186 17 1060 85010 295973 242 108459 302 4 7 6 2 1-Methylnaphthalene 30 130 12 954 52673 173949 498 66270 262 3 4 11 2 1,6-Dimethylnaphthalene 50 164 27 955 51124 144443 979 64691 436 4 11 330 3 2,3,5-Trimethylnaphthalene 17 96 20 350 4252 11108 123 5435 110 3 12 370 2 Dibenzothiophene 23 209 71 3102 43 116 3 62 131 49 18 103 4 2-Methyldibenzothiophene 8 51 27 287 n.d. n.d. n.d. n.d. 75 6 9 169 1 2-Methylphenanthrene 127 1139 489 7247 279172 610069 5334 371759 4574 147 76 298 16 2-Methylanthracene 65 617 269 3278 42126 91690 865 53888 1984 167 29 46 13 2,8-Dimethyldibenzothiophene n.d. 9 5 40 n.d. n.d. n.d. n.d. 32 3 4 65 0 2,4-Dimethylphenanthrene n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. 35 n.d. n.d. 4 2,4,7-Trimethyldibenzothiophene 3 9 4 15 n.d. n.d. n.d. n.d. 12 2 38 414 0 2,3-Dimethylanthracene 15 169 38 331 11226 27320 229 14936 336 26 15 132 3 1-Methylfluoranthene 66 588 270 3432 13168 30795 284 20388 2264 162 43 87 22 1,2,8-Trimethylphenanthrene 22 162 27 214 11707 31911 267 16356 375 15 35 295 2 1,2,6-Trimethylphenanthrene 24 133 18 131 4205 11375 94 5614 212 13 36 324 2 7-Methylbenz(a)anthracene 20 452 115 839 2551 6157 70 3953 516 79 11 30 8 3-Methylchrysene 76 662 241 1591 8114 15803 165 10585 1593 135 33 101 17 2-Methylchrysene 130 996 329 3056 9731 20535 220 13981 2504 195 49 154 25 1-Methylchrysene 77 676 156 1091 22587 49871 473 30806 1342 82 31 105 10 6-Ethylchrysene n.d. n.d. n.d. 20 46 173 n.d. 80 n.d. 1 n.d. n.d. n.d. 7-Methylbenzo(a)pyrene n.d. n.d. n.d. 606 4545 9640 103 5378 n.d. 0 n.d. n.d. n.d. Quinoline n.d. 11 n.d. 51 154 277 7 80 n.d. n.d. n.d. n.d. n.d. 1-Indanone n.d. 4 2 13 21 n.d. 74 19 32 0 2 24 0 Carbazole 96 1009 172 5376 207 388 19 219 347 230 40 23 10 Benzo[h]quinoline 5 61 7 575 n.d. n.d. n.d. n.d. 39 5 5 18 0 Acridine 9 217 11 457 n.d. n.d. n.d. n.d. 23 24 n.d. n.d. 2 9-Fluorenone 24 227 54 988 473 305 109 156 330 13 21 29 2 4H-Cyclopenta[def]phenanthrenone 28 285 242 1735 106 39 36 18 341 87 39 48 10 Anthracene-9,10-dione 47 360 187 1733 290 411 59 417 647 121 110 226 15 9-Methylacridine n.d. 43 12 82 602 1164 29 493 104 n.d. 4 35 1 11H-Benzo[a]carbazole 166 1228 620 3949 26 35 2 22 704 330 77 113 24 2-Methylanthracene-9,10-dione 16 111 61 221 n.d. n.d. 49 n.d. 407 38 26 38 5 Benzo[a]fluorenone 125 914 541 2352 1495 719 91 528 1778 325 96 203 26 7H-Benz[de]anthracen-7-dione 96 1225 625 3939 947 728 68 594 987 220 101 98 61 6-HBenzo[cd]pyren-6-one 170 1003 473 6369 423 250 15 244 1528 204 101 106 83 Benz[a]anthracene-7,12-dione 140 279 112 392 159 89 10 64 601 111 80 190 11 Naphthacene-5,12-dione 89 511 168 629 54 35 4 57 383 165 37 55 13 Dinaphtho[1,2-b;1,2d]furan 10 69 24 145 153 74 6 58 138 18 4 4 2 9,10-dihydrobenzo[a]pyren-(8H)-none n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. 2 dibenz[ah]acridine 26 235 103 851 n.d. n.d. n.d. n.d. 155 59 9 9 6
