Development of a Non-Targeted Method to Study Petroleum Polyaromatic Hydrocarbons in Soil by Ultrahigh Resolution Mass Spectrometry Using Multiple Ionization Methods

Abstract

A non-targeted method for analyzing petroleum hydrocarbons in soil was established by extracting a wide range of PAC compounds using a model system. Here, sand was spiked with a heavy crude oil, which simulated a hydrocarbon contaminated soil of low soil organic matter (SOM). Soxhlet and supercritical fluid extraction (SFE) was optimized for PAXHs analysis in soil. A concept of using multidimensional ionization for ultrahigh resolution mass spectrometry was employed for detection using electrospray ionization (ESI), atmospheric pressure chemical ionization (APCI) and atmospheric pressure photo ionization (APPI) for the analysis of polycyclic aromatic hydrocarbons (PAHs) and polycyclic aromatic heterocycles. Fourier transform mass spectrometry (FT-MS) with its ultrahigh mass resolution and high mass accuracy enables an unambiguous assignment of the detected ions obtained from different ionization methods, which provides a secure and comprehensive view of contaminants in soil on a molecular level. Over 95% of the spiked crude oil could be recovered by Soxhlet extraction using toluene, dichloromethane and acetone:n-hexane (1:1, v:v). In comparison, when using SFE with supercritical CO₂ the recovery was only about 50% but showed more polar compounds. Detailed analyses showed that besides 16 PAHs, a wide range with more than 10,000 PAXHs could also be successfully extracted.

Keywords:

• Atmospheric pressure ionization methods
• Fourier transform mass spectrometry
• polycyclic aromatic hydrocarbons
• Soxhlet extraction
• supercritical fluid extraction

Introduction

Polycyclic aromatic hydrocarbons (PAHs) originated from petrogenic, pyrogenic, or biogenic sources are omnipresent contaminants in soils.¹ Alone in Europe over half a million contamination sites have been identified as results of hundred years of industrialization. Annual clean-up cost ranging from 2 to 17 billion were estimated for these contamination sites.² Among others, PAHs produced from coking and manufactured gas plants are of great concern because of their carcinogenic, mutagenic and ecotoxicological effects.³

PAHs in soil are mostly analyzed in a targeted approach, where only 16 PAHs were selected, which belong to the priority pollutant list from the U.S. Environmental Protection Agency (EPA), and monitored for over four decades.⁴ Concerns have been arisen for an extension of this list of PAHs, since high-molecular-weight PAHs, alkylated derivatives (alkyl-PAHs) and polycyclic aromatic heterocycles containing N, S, or O (PAXH, X = N, S, O), which co-occur with those 16 PAHs as contaminants and might have higher toxicity, are not included.⁵

Recently targeted analyses of selected alkyl-PAHs and PAXHs have been reported using a combination of gas chromatography and mass spectrometry (GC-MS).^6–8 Inductively coupled plasma mass spectrometry (ICP-MS) is another powerful tool for the quantification of priority compounds or emerging contaminants, such as sulfur compounds in crude oil.^9,10 In addition, a non-targeted approach for PAH contaminated soils was enabled by coupling time-of-flight mass spectrometry (ToF MS) with high performance liquid chromatography (HPLC) or two-dimensional gas chromatography (GC × GC).^11–13 However, GC related techniques are restricted to volatile and semi-volatile compounds. Furthermore, using HPLC-ToF MS both the separation and mass resolving power could be challenged as the sample complexity and molecular weight of analytes increases.

Fourier transform ion cyclotron resonance (FT-ICR) and Orbitrap mass spectrometry with their ultrahigh mass resolution and mass accuracy ensure an unambiguous assignment of each detected signal in complex mixtures such as crude oil, bio oils, tar oils or lignin derivatives.^14–20 A combination of various atmospheric pressure ionization (API) methods, such as electrospray ionization (ESI), atmospheric pressure chemical and photo ionization (APCI/APPI), enables a comprehensive view of the sample.²¹ Negative mode ESI coupled to FT MS opens the possibility for detailed characterization of acidic compounds especially oxygenated classes. This combination is already applied for the analysis of naphtenic acids from environmental samples,^22–24 natural organic matter (NOM),^25–28 oil-sand-derived organic material,²⁹ and soil organic matters (SOM).^30–32 In contrast to these, basic compounds for instance pyridine-containing molecules are favored under positive mode ESI.³³ Both APCI and APPI are suitable for ionizing nonpolar compounds especially PAXHs, which provide complementary data to the ESI measurements for the same sample.^21,34

In the environmental analysis sample preparation especially extraction is the crucial step.³⁵ Around the beginning of this century various extraction methods such as Soxhlet extraction,³⁶ mechanic shaking,³⁷ ultrasonic extraction,³⁸ accelerated solvent extraction³⁹ or supercritical fluid extraction (SFE),⁴⁰ for extracting PAHs from soil were investigated and compared.^41–44 Despite the long extraction duration and large solvent consumption, the conventional Soxhlet extraction with its continuous and inexpensive character still serves as a standard method for PAHs extraction.⁴⁵ Its extraction efficiency was usually compared with other newer techniques. Among them SFE using inert, reusable CO₂ with/without addition of few organic solvents as modifier offers a more environmental friendly way of extracting contaminants in soil.^46–48

Analyzing different chemical compounds in a matrix such as soil is not easy. Different parameters play a role, such as how the compounds can be fully removed from the soil for a whole determination and characterization. This is followed by answers to the question which analytical method gives the best information. And doing this for an extremely complex analyte mixture such as tar oils makes matters worse. In this study, we perform a non-target analysis of a spiked sand sample with heavy crude oil, which simulates a soil sample contaminated with petroleum hydrocarbons and low SOM to better understand and optimize the extraction procedures and ionization conditions. FT Orbitrap MS coupled with different API methods including ESI, APCI, and APPI in both polarities were applied in order to gain a panorama of the contaminants in the sample. Afterwards the extraction efficiency of Soxhlet extraction with different solvents for petroleum hydrocarbons in this model sample was investigated. Finally, SFE using CO₂ was compared to Soxhlet extraction in detail on a molecular level.

Materials and methods

Model sample

Pure sand (50–70 mesh, Sigma-Aldrich, Germany) was spiked with a heavy crude oil (5%, w/w), which was diluted in dichloromethane. Afterwards the remaining dichloromethane was rotary-evaporated to obtain the spiked sand sample.

Sample preparation

The Soxhlet extractions were performed using 5.00 g of spiked sand and additionally 5.00 g of anhydrous sodium sulfate (99.5%, Acros Organics) with 100 mL of toluene (Tol), dichloromethane (DCM), acetone:n-hexane (1:1, v:v, AceHex) or methanol (MeOH) at the boiling point of the solvents for around 300 cycles. All solvents used were HPLC-grade and purchased from Sigma-Aldrich (Germany) or Acros Organics (Germany). The extracts were concentrated in a rotary evaporator and dried under gentle argon stream.

The SFE experiments were performed using a lab-built SFE system (for detailed description see Figure S1). 5.00 g of spiked sand were added into a stainless steel column and heated to 40 °C. Then samples were extracted at 180 bar and 40 °C with supercritical CO₂ (99.995%, Air Liquide, Germany) at a flow rate of 2 mL min⁻¹ for 4 h. The SFE extracts were rotary-evaporated and dried under gentle argon stream. The SFE residues were Soxhlet extracted using dichloromethane with the procedure described above.

Mass spectrometric measurements and data analysis

Mass analysis was performed on a research type linear quadrupole ion-trap (LTQ) FT Orbitrap MS (Thermo Fisher, Bremen, Germany) with LTQ Tune Plus 2.7.0 data processing system (Thermo Fisher, Bremen, Germany). The mass spectra were collected from 200 to 1200 Da using the spectral stitching method with 30 Da mass windows with 5 Da overlap.^15,49,50 Additionally positive/negative mode ESI or positive mode APCI/APPI with a mass resolution of 960,000 at m/z 400 was applied. For ESI measurements samples were diluted to 250 μg mL⁻¹ in toluene:methanol (1:1, v:v), and injected at a flow rate of 5 μL min⁻¹ with a spray voltage of 4/−3.5 kV for the positive/negative mode. Sheath, auxiliary and sweep gases were set to 5, 2, and 1 arbitrary unit for both polarities. Samples for APCI or APPI were diluted to 250 μg mL⁻¹ in toluene and ionized under the following conditions: flow rate = 20 μL min⁻¹, vaporizer temperature = 350 °C, sheath gas = 20 arbitrary unit, auxiliary gas = 5, and sweep gas = 2. The corona discharge for the APCI was set to 5 μA and a Krypton VUV lamp (10.0 and 10.6 eV, Syagen, Tustin, CA, U.S.A.) was used for the APPI.

Data were exported into Composer64 v1.5.3. (Sierra Analytics, Modesto, CA, U.S.A.) and processed with the following chemical constraints: 0 < C < 200, 0 < H < 1000, 0 < N < 4, 0 < S < 5, 0 < O < 11, 0 < double bond equivalent (DBE) < 11. A maximum mass error of 1 ppm was accepted and only classes with a relative abundance higher than 1% were compared and discussed.

The DBE value for a given elemental composition C_cH_hN_nS_sO_o was calculated with the following equation: $$\text{DBE}\mathrm{}=\text{c}–\mathrm{}\mathrm{h}/2\mathrm{}+\mathrm{}\mathrm{n}/2\mathrm{}+\mathrm{}1$$

Results and discussion

Recovery of crude oil in spiked sand

A non-targeted analysis of petroleum hydrocarbons in soil is an adventure due to the high complexity of compounds present in such a sophisticated system. As one of the most important sample preparation steps, the extraction of petroleum based contaminants from soil needs to be investigated in detail.

Therefore, pure sand spiked with a heavy crude oil served as a model sample for a better understanding of the extraction step. Here, we compared the traditional Soxhlet extraction using different solvents with supercritical CO₂ extraction. Toluene, dichloromethane and a 1:1 (v:v) mixture of acetone and n-hexane are commonly used solvents for the Soxhlet extraction of 16 EPA PAHs in soils on a routine scale.^{36,41–44,51}

Using these three solvent systems a recovery of over 95% for crude oil in the spiked sand was achieved (see Figure 1), where dichloromethane with 97% ± 1% recovery delivered the highest extraction efficiency. Methanol was applied as a mild extraction solvent to study the solubility in polar solvents and therefore, the bioavailability of hydrocarbons in soil.⁵² Here, a lower recovery of 73% ± 4% compared to other Soxhlet extraction solvents was obtained. This implies the complex character of the heavy crude oil.

Figure 1. Recovery of heavy crude oil in the spiked sand using Soxhlet extraction (in green) with toluene (Tol), dichloromethane (DCM), acetone:n-hexane (1:1, v:v) (AceHex), methanol (MeOH), and supercritical CO₂ extraction (extract in blue, residue in brown).

When using supercritical CO₂ extraction 47% ± 1% of petroleum components were extracted. Reports in the literature showed a wide range of results for the extraction using SFE. A good recovery of over 80% was reported for lighter jet-diesel oil and fuel oil from spiked sand after the SFE at 65 °C and 400 bar.⁴⁶ However, Hawthorne and Miller showed that at a comparable extraction temperature of 50 °C and accelerated pressure of 405 bar, lower extraction efficiency was found for higher molecular weight and heteroatom containing PAHs.⁴⁷ This explained the lower extraction efficiency gained in our study, thus the heavy crude oil contained to a proper portion high molecular weight and heteroatom containing PAXHs.

Although SFE is more environmental friendly and time saving, it cannot extract all contaminants from the soil, especially compounds with higher molecular weight. In contrast to that, the traditionally Soxhlet extraction with dichloromethane offers a simple and feasible way to extract petroleum hydrocarbons from the soil for the non-targeted analysis, albeit the large volume and long extraction time needed.

Mass spectra

Because of the sample complexity, the selectivity of different ionization methods and polarities can be used to provide a comprehensive view of the sample.²¹ This is presented in the recombined mass spectra, which was recorded using spectral stitching method,^15,49,50 for the original crude oil (Figure 2, left column). For all ionization methods applied ions were detected throughout the whole mass range starting from 200 to 1200 Da. In the zoomed in mass spectra between m/z 400.00 and 400.50 (Figure 2, right column) up to 42 elemental compositions were detected, which was depended on the selected ionization method. Compounds containing basic nitrogen dominated in the ESI(+) spectrum. Acidic oxygen classes and neutral nitrogen containing components are well ionized in ESI(−).

Figure 2. Recombined (left column) and zoomed in mass spectra at m/z 400 (right column) for the original crude oil sample, analyzed by FT Orbitrap MS using positive mode ESI (red), negative mode ESI (blue), positive mode APCI (green), and APPI (orange), respectively.

Nonpolar classes such as pure hydrocarbons and sulfur containing PAHs appear in both APCI and APPI measurements. To resolve isobaric peaks of C₂₆¹³CH₄₃S⁺ (m/z 400.31135) and C₃₀H₄₀^.+ (m/z 400.31245) with a Δm = 0.00110 Da, a resolving power higher than 400,000 is needed. The result shows the necessity of combining different ionization methods with ultrahigh resolution MS for a non-targeted analysis of complex petroleum hydrocarbons.

In Figure 3, the zoomed in positive mode APPI mass spectra between m/z 400.00 and 400.50 for the original crude oil, Soxhlet extract of the spiked sand using dichloromethane, SFE extract and residue of the spiked sand were compared. In total, 42 elemental compositions were found at this mass range for the original crude oil (Table S1). Here, radical cations of pure hydrocarbons and PASHs dominated the mass spectra. C₃₀H₄₀ and C₂₇H₄₄S with a double bond equivalent (DBE) of 11 and 6 were the first and second highest signal detected in the mass spectra for the original crude oil, Soxhlet and SFE extracts. By decreasing m/z value the DBE value of isobaric pure hydrocarbon ions (C₂₉H₅₂⁺, C₃₀H₄₀⁺, C₃₁H₂₈⁺, C₃₂H₁₆⁺, highlighted in red from right to left in Figure 3), which have a Δm = 0.09390 Da between each other, increases from 4 to 25. The extraction efficiency of SFE decreases the higher the aromaticity is. As can be seen in the spectra, C₂₉H₅₂⁺ (DBE = 4) and C₃₀H₄₀⁺ (DBE = 11) were to a great degree extracted using SFE, while C₃₁H₂₈⁺ and C₃₂H₁₆⁺ with a DBE of 18 and 25 remained mainly in the residue. This is also applicable for S class compositions shown in Figure 3.

Figure 3. Zoomed in mass spectra at m/z 400 for the original crude oil (black), Soxhlet extract using dichloromethane (green), SFE extract (blue), and SFE residue (brown) using positive mode APPI FT Orbitrap MS. Some signals are highlighted with calculated elemental compositions and double bond equivalent (DBE) values, separated with a comma.

Furthermore, positive and negative mode ESI mass spectra for the original crude oil, the extract and residue after SFE reveal that basic and acidic crude oil related contaminants could also be extracted from the spiked sand (Figures S2–S4 are documented in the supporting information). Compounds with lower DBE were preferentially extracted than compounds with higher DBE. Using API-FT MS a detailed comparison between Soxhlet and SFE on a molecular level can be fulfilled.

Class distribution

Due to the high data load obtained from the non-targeted analysis using ultrahigh resolution FT MS, a suitable, concise and explicit data interpretation is obligatory. This can be done by sorting each signal from the mass spectrum into elemental compositions, which then can be categorized into classes according to their counts of heteroatoms. Different classes such as pure hydrocarbon or heteroatom (N, S, and O) containing classes from different ionization techniques or samples can be compared in a bar chart with their corresponding relative abundances or numbers of assigned compositions.

In Figure 4, a summary of intensity based class distributions for the Soxhlet extract using dichloromethane is shown. Protonated basic nitrogen containing classes such as N, N₂, NS, and NS₂ dominate the ESI(+) spectrum with few compounds form radical ions as well as oxygenated sulfur containing class ionized with sodium adduct. In ESI(−) the mass spectrum consists of mainly oxygenated classes especially O₂ and O₄, which can correspond to molecules contain one or two carboxylic acids.⁵³ Other oxidized heteroatom containing classes such as NO_x and SO_x were also present with a lower intensity. The selective ionization of acidic components in the sample makes it a perfect method for examining polar transformation products, which are created during the chemical/biological oxidation of PAHs contaminated soil.^54–56

Figure 4. Intensity based class distributions for the Soxhlet extracted sample using dichloromethane, analyzed by positive/negative mode ESI, positive mode APCI/APPI (top to bottom) FT Orbitrap MS with protonated/deprotonated molecules in red, radical cations in blue, and molecules with sodium adduct in green.

Both APCI and APPI form radical cations [M]⁺ and protonated molecules [M + H]⁺ of predominantly nonpolar pure hydrocarbons (HC), S_x, and N classes. In particular APPI by exciting the double bonds in a molecule which suits the best for the targeted and non-targeted analyses of pure PAHs and PAXHs (X = S, N, O).^21,57 Results show how complex such a system, which is simulating a petroleum hydrocarbon contaminated soil, can be. Additionally, when analyzing petroleum hydrocarbons in soil in a non-targeted way, ion suppression and discrimination effects can occur.⁵⁸ This consolidates the need of multidimensional ionization techniques for the deep and comprehensive analysis of petroleum hydrocarbon contaminated soils.

For the targeted analysis, only the 16 EPA PAHs are of interest. But by using positive mode APPI FT MS other high molecular weight and heteroatom containing PAHs present as contaminants can be efficiently ionized and analyzed as well.²¹ For the original crude oil the highest number of assigned compositions was detected by APPI(+) (25,943), followed by APCI(+) (22,802), ESI(+) (15,264), and finally ESI(−) (12,221) (Figures 5–7). Results confirmed the nonpolar character of the spiked contamination source. For a comparison the extraction efficiency of Soxhlet extraction using different solvents and SFE is summarized in Figure 5. Here, both the relative abundance and the number of assigned compositions for each detected class were considered.

Figure 5. Intensity (top) and population (bottom) based class distributions for the original crude oil, spiked sand Soxhlet extracted using Tol, DCM, AceHex, or MeOH and its SFE extract (SFE E) and residue (SFE R), analyzed by positive mode APPI FT Orbitrap MS. Protonated molecules and radical cations are denoted in red and blue, respectively.

Figure 6. Intensity (top) and population (bottom) based class distributions for the original crude oil, spiked sand Soxhlet extract using Tol, DCM, AceHex, or MeOH and its SFE extract (SFE E) and residue (SFE R), analyzed by positive mode-APCI FT Orbitrap MS.

Figure 7. Intensity (top) and population (bottom) based class distributions for the original crude oil, spiked sand Soxhlet extract using Tol, DCM, AceHex, or MeOH and its SFE extract (SFE E) and residue (SFE R), analyzed by positive (left) and negative (right) mode-ESI FT Orbitrap MS. Protonated molecules and radical cations are denoted in red and blue, respectively.

Among these 25,943 assigned compositions by APPI(+) around 4000 were detected as either protonated or radical hydrocarbons in the mass spectrum, which display a relative intensity of around 35%. With an increase in polarity from toluene to methanol, the number of assigned compositions in the HC class decreases from 4082 to 3629. For the polycyclic aromatic heterocycles, such as compounds in S, S₂, and N classes, over 95% of the assigned compositions from the original crude oil could be found in the Soxhlet extracts using toluene, dichloromethane, and acetone-n-hexane solvent mixture. For methanol as extraction solvent a recovery of over 90% of the assigned compositions was observed.

In comparison, less than 67% (17,367 out of 25,943 compositions from the original crude oil) were present in the SFE extract, which consisted of predominantly pure hydrocarbons (41%), followed by S (32%), S₂ (6%), O (5%), and N (4%) classes. As a result of extraction, the relative intensities of HC and S classes found in the SFE residue decreased dramatically to 21% and 14%, respectively.

However, the numbers of assigned compositions between the original crude oil and SFE residue did not differ too much, which indicates lower extraction efficiency for the complex sample when SFE was applied. The slightly higher amount of HC class compounds and new classes found in the SFE residue could be explained, as there was discrimination and ion suppression occurred during ionization for the original crude oil.⁵⁸ Additionally, SFE separated the sample into two fractions which led to an increasing analysis depth.⁵⁹ Similar results were observed by APCI(+) studies (see Figure 6). There, compositions from pure hydrocarbon und heteroatom containing classes such as S, S₂, and N could be extracted well by Soxhlet extraction using toluene, dichloromethane and acetone/n-hexane solvent mixture. Only a portion (≈70%) of the compositions in the above mentions classes could be extracted using SFE.

Results from APPI(+) and APCI(+) were mutually supportive and implied a better extraction efficiency for nonpolar compounds by Soxhlet extraction using toluene, dichloromethane and acetone-n-hexane solvent mixture than SFE. In ESI(+) the N class with the highest intensity of 60% in the original crude oil consists of mainly basic pyridine-like azaarenes (see Figure 7, left).¹³ Out of 2701 assigned compositions as either protonated or radical cations for the N class in original crude oil 1951, 2002, 1569, and 2259 compositions were recovered using toluene, dichloromethane and acetone-n-hexane solvent mixture and methanol, respectively. For the second and third highest intensity classes in the original crude oil, namely NS and N₂, same trend was observed. As the most polar solvent investigated methanol could extract over 80% of the assigned compositions for these two classes, whereas for other solvent systems the recovery was between 60% and 70%. The appearance of [SO + Na]⁺ class with relative high intensity for the Soxhlet extracts was probably a result of sodium salt added for the removal of water.⁶⁰ Compared to the Soxhlet extraction also a good extraction of polar N class compounds from the spiked sand was achieved when supercritical CO₂ was used. Here, a decrease of relative abundance from 74% (SFE E) to 14% (SFE R) was observed.

Compared to ESI(+) with 15,264 assignments, in ESI(−) 12,221 compositions were detected for the original crude oil, which comprise polar compounds containing acidic functional groups (Figure 7, right). Unlike in the positive mode the assigned compositions in ESI(−) were distributed to O_2–4, N_0–2, and S_2–4 quite equally. A comparable high recovery was achieved by Soxhlet extraction using dichloromethane and methanol. In comparison to this, SFE could only partially extract polar compositions in each class.

These findings obtained from each ionization method showed, Soxhlet extraction is an easy and efficient method, which can be applied prior to the non-targeted analysis of crude oil related contaminants in soil and gives the best representation of the sample.

DBE distribution

The absolute intensity within a given DBE was plotted against compound classes for the APPI(+) measurement of the samples (Figure 8). The radical cation and protonated molecule classes are denoted as M and M[H], respectively.

Figure 8. Intensity based DBE vs. class distributions for the original crude oil, Soxhlet extract using DCM, SFE extract and residue (from top to bottom) using positive mode APPI FT Orbitrap MS (radical cations: M, protonated molecules: M[H]).

A DBE range of 1–50 was observed for the pure hydrocarbon and heteroatom containing classes, such as S_1–3 and N, found in the original crude oil, Soxhlet extract and SFE residue. However, in the SFE extract compounds with DBE only up to 37 were detected. Considering a benzene ring has a DBE value of 4 and by adding another aromatic ring into the system to form naphthalene, anthracene, and so on, the DBE increases by 3 every time with up to 12 aromatic rings. Similar results were observed using APCI(+) as shown in Figure S5. Since predominant compounds with lower DBE could be extracted using SFE, this reduces the sample complexity and increases the intensity of compounds with lower abundance (Figure 5),⁵⁹ which explains the additional classes found in the SFE residue.

As summarized in Table 1, the mean m/z value for the SFE extract was shifted to the lower mass range compared to the original crude oil sample. Besides, the DBE value decreased from 19 to 15 while H/C increased from 1.23 to 1.38. Accordingly for the SFE residue the mean DBE value increased to 20 while H/C decreased to 1.17. These indicate that SFE favors low aromatized hydrocarbons, which is in agreement with previous studies.⁴⁰

Table 1. Calculated mean values of m/z, DBE, number of carbon atoms and X/C for the samples, analyzed using APPI(+) FT Orbitrap MS.

Samples	Mean values
	m/z	DBE	#C	H/C	N/C	O/C	S/C
Orig. crude oil	703	19	50	1.23	0.005	0.008	0.019
Soxhlet_DCM	700	19	50	1.23	0.004	0.008	0.019
SFE_extract	677	15	48	1.38	0.004	0.010	0.016
SFE_residue	703	20	50	1.17	0.006	0.012	0.019

Moreover, polar compounds such as basic pyridine-like or neutral pyrrole-like nitrogen containing compounds and oxygenated classes, analyzed using ESI(+/−), could also be extracted with a maximum DBE of 35 (Figures S6 and S7).

DBE vs. carbon count/intensity plot

A closer look into the radical HC class on a molecular level is shown in left column of Figure 9 using the so-called Kendrick plot.¹⁴ Here, each detected composition was visualized by its DBE value and number of carbon atoms per molecule. In addition, the total monoisotopic abundance as a sum of all compositions within one given DBE was displayed in line.

Figure 9. DBE vs. carbon count/intensity (Kendrick) plots for the radical HC (left column), [N + H]⁺ (middle column) and [N − H]⁻ (right column) classes in the original crude oil, Soxhlet extract using DCM, SFE extract and residue (from top to bottom) using positive mode APPI or positive/negative mode ESI FT Orbitrap MS. Upper abscissa for the absolute intensity in a given DBE and lower abscissa for the number of carbon atoms in a assigned composition.

2107 and 1515 out of in total 2173 radical cations found in the original crude oil could be fully or partially extracted using Soxhlet or SFE, respectively. For the original crude oil, Soxhlet and SFE extracts pure hydrocarbons with DBE of 12 exhibited the highest intensity, which correspond to PAHs containing 4 rings, such as pyrene or fluoranthene with different length of side chains. Instead, in the SFE residue the maximum intensity of DBE shifted to 15, indicating PAHs involving 5 aromatic rings, such as benzo[a]anthracene, benzo[b]fluoranthene or benzo[k]fluoranthene, were dominant in the mass spectrum. Furthermore, the maximum intensity at DBE 15 in the SFE residue (10⁵) was reduced by one order of magnitude compared to the value in the extract (10⁶).

The PASHs in the original crude oil, Soxhlet, and SFE extracts showed a maximum intensity at DBE of 9, which correspond to mostly dibenzothiophene and its alkylated derivatives (Figure S8, left column).⁶¹ However, in the SFE residue compounds with DBE = 19 gained the highest intensity. Also for PANHs a shift of the maximum intensity from DBE = 12 to DBE = 17 was observed after SFE (Figure S8, right column). Overall, the results from APPI(+) measurements showed that SFE as compared to Soxhlet extracts only contains nonpolar PAHs and PAXHs with DBE below 35 and carbon number below 90.

As an example for ESI(+) measurements, basic nitrogen containing compounds are shown in the middle column of Figure 9. From 2134 polar pyridine-like compounds 1538 and 1101 could be extracted using Soxhlet and SFE, which contains mainly 1–5 aromatic rings. However, as the length of alkyl side chains increases, the extraction efficiency decreases. The neutral pyrrole-like nitrogen containing compounds were analyzed using ESI(−) (Figure 9, right column). Since in the original crude oil N class compounds with only DBE up to 36 were detected, they can all be extracted to a certain extent using both methods. Comparing the SFE extract with its corresponding residue, neutral PANHs could be largely (about 90%) extracted through SFE.

Conclusion

In this work, different organic solvents, which are commonly used for the quantitative analysis of 16 EPA PAHs by Soxhlet extraction, were extended to extract other PAHs and PAXHs containing N or S and their alkylated or oxygenated derivatives. The analysis on a molecular level showed that nonpolar compounds could be extracted quite well by the solvents applied, whereas for polar compounds the sample extracted using dichloromethane and methanol exhibited the highest recovery. Results demonstrated that prior to mass spectrometric analysis the Soxhlet extraction could be used for extracting a large variety of hydrocarbon related contaminants from the soil, which are co-occurring with these 16 PAHs.

Additionally, SFE using supercritical CO₂, which is usually referred to as a green extraction method, was studied in detail for extracting petroleum hydrocarbons from spiked sand. Compared to the standard Soxhlet extraction SFE could extract a range of compounds from nonpolar to polar hydrocarbons with a DBE and carbon count lower than 35 and 90, respectively.

In our study, a non-targeted method was developed to examine oil contamination in soil by using ultrahigh resolution mass spectrometry with different ionization methods. A heavy crude soil sample was spiked into pure sand to produce a simple model simulating a hydrocarbon contaminated soil sample with low organic content. Results showed that the selectivity of ionization methods applied offers sometime comparable but more importantly complementary data, each of which reflects only a part of the real sample. Therefore, for the non-targeted analysis of petroleum hydrocarbons in soil, a multidimensional ionization is indispensable. In addition, the coupling to FT MS instrument after ionization ensures a distinct assignment of the signals for such a complex sample.

Acknowledgments

The authors thank Niklas Fuhrmann Max-Planck Institut für Kohlenforschung for assistance with the SFE system. The authors thank Dr. David Stranz (Sierra Analytics, Modesto, CA, USA) for access to new data handling software.

Funding by the ZIM Project KF 3121303SK4 of the Bundesministerium für Wissenschaft und Energie is gratefully acknowledged.