Comprehensive LCxGC Characterization of Mineral Oils Using Flame Ionization Detection and Vacuum Ultraviolet Detection

(1)

1

Bachelor Thesis Chemistry

Comprehensive LCxGC Characterization of Mineral Oils Using

Flame Ionization Detection and Vacuum Ultraviolet Detection

by

Joshka Verduin

26 June 2018

Student number

11030194

Research institute

Responsible teacher

Van ’t Hof Institute for Molecular Sciences

Prof. dr. ir. P.J. Schoenmakers

Research group

Mentor

Analytical Chemistry Group

Prof. dr. ir. J.G.M. Janssen

Daily supervisor

(2)

2

Samenvatting

Heel veel alledaagse producten zijn besmet met minerale oliën (MO), omdat de oliën veelal worden toegepast. Zo worden deze oliën gebruikt als smeermiddel in fabrieken, waardoor de geproduceerde artikelen besmet raken met de oliën. Verder worden de oliën gebruikt als oplosmiddel voor drukinkten. Wanneer een product, zoals een pak rijst of pasta, een laagje inkt bevat met MO, dan wordt het voedsel in de verpakking ook besmet. Daarnaast wordt minerale olie veelal gebruikt in cosmetische producten omdat het hydraterend werkt. Ook zijn de oliën aangetroffen in het milieu, waardoor niet alleen de mens maar ook de natuur besmet raakt. Minerale oliën bestaan uit twee klassen stoffen: minerale olie verzadigde koolwaterstoffen (MOSH) en minerale olie aromatische koolwaterstoffen (MOAH). De MOSH hoopt op in het lichaam en is daarmee een van de grootste vervuilers van het lichaam terwijl MOAH kankerverwekkend en mutageen is.

Om deze reden is het doel van dit onderzoek het ontwikkelen van een methode om te achterhalen waar deze oliën uit bestaan. De concentratie van verschillende oliemonsters werd bepaald door middel van het koppelen van gaschromatografie (GC) aan een vlam ionisatie detector (FID). Wanneer je aan je MO-monster een stof (interne standaard) met een bekende concentratie toevoegt, kan aan de hand van de verhoudingen de onbekende concentratie van de minerale olie bepaald worden. Verder werd er in dit project een nieuw soort ultraviolet detector getest en gebruikt: een vacuüm UV-detector (VUV). Terwijl je in normale UV-spectra alleen de geconjugeerde (gekleurde) moleculen kan detecteren, geeft een VUV-detector een uniek signaal voor vrijwel alle stoffen waaronder verzadigde verbindingen. Zo kan aan de hand van het VUV-spectrum bepaald worden of een fractie van een olie uit MOSH of MOAH bestaat. Verder geven 1-ring en 2-ring aromatische verbindingen een andere VUV-absorptie dus dan kan ook het aantal ringen in MOAH bepaald worden.

Wanneer deze twee methoden verder ontwikkeld worden, kunnen ze in de toekomst gebruikt worden om de concentratie van de oliën in echte monsters (bijvoorbeeld rijstkorrels) te bepalen. Verder kan, wanneer de compositie van de oliën in de producten bekend is, de giftigheid van de besmette producten bepaald worden.

(3)

3

Abstract

The quantification and characterization of mineral oil saturated hydrocarbons (MOSH) and mineral oil aromatic hydrocarbons (MOAH) has been a challenge for scientists for many years, due to the complexity of mineral oils (MO). Mineral oils are widely being applied in consumer products, not only contaminating humans but nature as well. Due to the harmful effects of mineral oils, it is important to develop a method for both the quantification and characterization. For the LC-GC-FID quantification of MO, researchers have been using TBB as a marker for MOAH. However, this research has proven that MOAH elutes earlier than TBB, leading to MOAH-concentrations smaller than the actual concentrations. Therefore, the response of the refractive index detector of the LC-system was used as a marker for MOAH, leading to a more accurate quantification. Furthermore, the LC pre-separation was performed using a silver-loaded silica column, leading to not only distinguishing MOSH and MOAH, but also to separate MOAH based on the number of aromatic rings. For the characterization of mineral oils, comprehensive LCxGC-FID was compared with LCxGC-VUV, with VUV being the new vacuum ultraviolet (VUV) detector. Using this new detector, the hypothesis of the composition of mineral oils based on the LCxGC chromatogram can be extended by consulting the corresponding spectra. Each compound is said to give a unique response in the VUV-window, hence the spectra can verify the predicted number of rings and predict the composition of the substituents on the aromatic rings. However, a DCM-gradient was applied in the LC-dimension, influencing the separation of MOAH. Since mineral oils are a complex mixture of many compounds, first pre-separation should be performed with an isocratic gradient. Next, the chromatographic and spectroscopic properties of aromatic hydrocarbons should be analyzed more thoroughly. After that, the addition of a gradient might be applied if it indeed improves the method. Lastly, the newly developed methods could be tested on real samples to determine the toxicity of mineral oils in consumer products.

(4)

4

Glossary

6B hexylbenzene 9B nonylbenzene C11 n-undecane C13 n-tridecane

CCD charged coupled device

Cho 5-α-cholestane

CoV coefficient of variation

DCM dichloromethane

DEHB 1,4-di(2-ethylhexyl) benzene

FID flame ionization detector

GC gas chromatography

HPLC high performance liquid chromatography

LC liquid chromatography

LOD limit of detection

LOQ limit of quantification

MO mineral oil

MOAH mineral oil aromatic hydrocarbon MOSH mineral oil saturated hydrocarbon NPLC normal phase liquid chromatography

PAH polycyclic aromatic hydrocarbon

PDA photo diode array

RID refractive index detector

SPE solid phase extraction

TBB 1,3,5-tri-tertbutylbenzne

(5)

5

1. Introduction

Mineral oils (MO) are suspected of being the biggest contaminant in the human body, making it utterly important to develop a method for both quantitative and qualitative analysis.1_{Mineral oils are complex}

mixtures that consist of mineral oil saturated hydrocarbons (MOSH) and mineral oil aromatic hydrocarbons (MOAH). The latter group is usually alkylated and can possess multiple rings, however, this group should not be confused with polycyclic aromatic hydrocarbons (PAHs).2_{The application of mineral oils is large;}

for instance, these oils are used as solvents for offset printing inks, meaning that if a cardboard packaging (e.g. a box of cereal, rice, pasta) is coated with these inks, the foodstuffs inside can be contaminated.3_This

especially holds true for the packaging of recycled cardboard, because they are said to be the main source of contamination of mineral oils in the human body, which is one of the main disadvantages of recycling paper.4_{The migration of MOAH into foodstuffs from food packaging can be prevented by coating the inside}

of the packaging with a protective layer, such as foil.4_{Furthermore, MOs are frequently applied in industry}

as lubricating oils, causing the fabricated products to be contaminated with these oils.5_{Also, a substantially}

amount of cosmetic products contain mineral oils, such as baby oil, lip sticks, and lotions, due to its moisturizing properties.4_{However, mineral oils are also present in the environment, causing these oils not}

only to contaminate humans but nature as well. Even though mineral oils are being widely used in consumer products, they are by far harmless. The MOSH can accumulate in the body and thereby damage organs and the MOAH is suspected of being carcinogenic and causing DNA-damage.4_{Therefore, it is important to}

develop a method for quantifying and characterizing the mineral oils to elucidate the composition of the oils and to determine the concentration of MOs in foodstuffs.

Scientists have been trying to create a compatible approach for the characterization and quantification of mineral oils, however, due to the complexity of the oil’s composition, the method development remains challenging. One approach for the characterization of mineral oils is comprehensive gas chromatography (GC) with a mass spectrometer (MS): GCxGC-MS6,7_{, but the major drawback of this}

method is that it is time-consuming. A common method for the quantification is, after pre-separation with liquid chromatography (LC), the combination of gas chromatography (GC) with the non-selective flame ionization detector (FID).8,9_{This method is appropriate for quantitative studies, e.g. for determining the}

mass distribution,10_{due to its robustness and decent linearity.}11_{However, problems arise when qualitative}

results are desired or when complex mixtures have to be analyzed.8

To aid the analysis of such complex mixtures, a new type of ultraviolet detector has recently been developed: the vacuum ultraviolet (VUV) detector. The VUV ranges between 115 and 240 nm, which is lower than the regular UV-detector with a range above 190 nm.10_{Whereas only a selection of compounds}

can absorb in the UV-Vis region, virtually every compound absorbs in the VUV-region, leading to unique responses. In the GC-VUV system, the GC-column is connected to the VUV-apparatus (Fig. 1). After a sample, that has been injected, has passed through the GC-column, the particles enter a heated transfer line that is connected to a makeup gas inlet. Next, the particles enter the flow cell that leads them to the right direction. One side of the flow cell is connected to a source module containing a deuterium lamp to create electromagnetic waves, whereas the other side is connected to the detector module; both compartments are connected via a magnesium fluoride window.11_{In the detector module, the particles are first diffracted in a}

grating, after which the diffracted light is detected using a charge coupled device (CCD). The corresponding response is then sent to a computer for data interpretation.

(7)

7

Figure 1 Schematic representation of the GC-VUV system.12_{After a sample has been injected into the GC-column, the particles}

enter a heated transfer line with a flow of makeup gas. Next, the analytes enter the flow cell with on the one side a deuterium lamp and on the other side a grating with a detector. This detector is linked to a PC on which the data can be interpreted.

This study focuses on developing a method for both the characterization and quantification of mineral oils. For the analysis of the mineral oils, a separation of different groups of compounds is needed. For this, high performance liquid chromatography (HPLC) could be deployed. It was demonstrated that normal phase liquid chromatography (NPLC), using silica as a stationary phase, is able to separate the two mineral oil fractions MOSH and MOAH based on their polarity.5_{Besides solely making the distinction}

between the two groups of mineral oils, even more separation could be accomplished by using a silver-loaded silica column.13_{Since the silver-ions interact with the π-bonds in the aromatic rings of the}

MOAH-fraction, this stationary phase can be used to separate the aromatic fraction based on the number of aromatic rings.14_{For the characterization, small fractions from the HPLC-system will be collected, after which they}

will be injected into a GC-VUV-system. The resulting 2D LCxGC-VUV plots will subsequently provide information about the composition of the mineral oils by not only showing a distinction between MOSH and MOAH, but hopefully also between the number of rings in MOAH.

Besides the characterization of the mineral oils, a quantification method should be developed as well to determine the concentration and ratio of MOSH and MOAH in mineral oil. In this case, the MOSH- and MOAH-fractions, collected after a pre-separation by HPLC, will be measured in a GC-FID-system. The FID has the advantage of providing virtually the same response per unit of mass for all the hydrocarbons. This allows to obtain the ratio of MOSH and MOAH inside a mineral oil, as well as the concentration of the compounds when compared with an internal standard.

The aim of this study is thus to develop new methods for characterizing and quantifying mineral oil samples to gain more insight into the chemical composition of mineral oils. In this thesis, first the applied materials and methods will be described. Subsequently, the results will be divided into two sections: the quantification with LC-GC-FID and the characterization using LCxGC-VUV and LCxGC-FID, in which the results will be discussed. Finally, the highlights of this research and suggestions for further research will be given in the conclusion.

(8)

8

2. Materials and methods

2.1 Reagents and standards

Dichloromethane (DCM, ≥99% stabilized, GCR Rectapur®) and n-hexane (≥95%, AnalaR NORMAPUR®) were obtained from VWR Chemicals (France). n-undecane (≥99%, C11), n-tridecane

(≥99%, C13), bicyclohexyl (99%), and 5-α-cholestane (Cho) were used as MOSH-markers and hexylbenzene

(97%, 6B), nonylbenzene (96%, 9B), and 1,3,5-tri-tert-butylbenzene (97%, TBB) were used as MOAH-markers for one-ring aromatics. 1-methylnaphthalene (99%) and biphenyl (99.5%) were used as 2-ring aromatic MOAH-markers. All these marker compounds were obtained from Sigma Aldrich (Zwijnendrecht, the Netherlands). The mineral oil samples were from different stages in the purification process of mineral oils and were obtained from a producer of white oils.

2.2 Validation of the GC-VUV method

GC-FID is a method which has been validated thoroughly in the past. However, since the GC-VUV method is still in its infancy, it is important to validate the method to determine the performance of the method. In this study, a partial validation was performed by determining the linearity, the limit of detection (LOD), the limit of quantification (LOQ), and finally the precision. The Eurachem laboratory guide15_{was followed for}

the (partial) validation of the method. The LOD and LOQ were determined using the following equations (Eq. 1-3):

𝑆𝐷′ =𝑆𝐷 √𝑛

Equation 1: Equation for calculating the LOD and LOQ. SD’ is the standard deviation used for calculating the LOD and LOQ, SD is the estimated standard deviation of all the measurements, and n is the number of samples (which is 3 in this research).

𝐿𝑂𝐷 = 3 ∗ 𝑆𝐷′

Equation 2: The limit of detection is calculated by multiplying SD’ with 3.

𝐿𝑂𝑄 = 𝑘𝑞 ∗ 𝑆𝐷′

Equation 3: The limit of quantification is determined by multiplying SD’ with factor kq of 10.

Mineral oils a, h, 2, and k were used for the calibration of the VUV-detector. Mineral oil k was free of aromatics, MO k only consisted of MOAH after SPE, and mineral oils a and h both contained MOSH and MOAH. For each MO, solutions of 20, 10, 5, 2, 1, and three solutions of 0.5 mg/mL were prepared for creating a calibration curve and to determine the LOD and LOQ. MOSH and MOAH from mineral oil were first separated using solid phase extraction (SPE). A silver-silica packed SPE-column was first washed with DCM (10 mL), after which the column was conditioned with hexane (4 mL). Next, the cartridge was loaded with sample k (0.5 mL). The MOSH fraction was collected after elution with hexane (5 mL), and the MOAH fraction using first a hexane/DCM solution 1:1, (v/v, 6 mL) and lastly with DCM (2 mL). The resulting solution of the aromatic fraction was used to prepare 10, 5, 2, 1, and 0.5 mg/mL dilutions.

For the GC-VUV-analysis, an Agilent G1530 gas chromatograph (sn US00025877) was used along with a J&W Scientific column (15m, 320 μm, 0.1μm, DB5HT) with helium as a carrier gas and atmospheric outlet pressure. The temperature was initially 60°C, and after 3 minutes the temperature increased to 350°C with a speed of 15°C/min. An ATAS GL international OPTIC 3 injector (sn H0303133) was used with a splitless injection and the injector was set at 350°C. The splitless time was 2 minutes with a column flow of

(9)

9

2 mL/min. The vacuum UV-detector was a VGA-101 of VUV Analytics and the flow cell heater, the transfer tube heater, and the stable temperature were set to 350 °C. The GC-VUV analyses were performed using the VGA Controller (v5.05.531) and the peak integration for the calibration curves was performed using the VUV Chromatogram Response Calibration (v5.05.472) software. For the calibration, 1 μL of each concentration was injected three times into the system. The following spectral filters were applied: saturated (125-160 nm), aromatic (190-240), and total mineral oil (125-240 nm).

The results of the partial validation of the GC-VUV method will not be part of the result sections but will be presented in Appendix 8.1.

2.3 Quantification of mineral oils: LC-GC-FID analysis

For the quantification of the mineral oils, two Ag-Phase CP29340 columns (Agilent Chromsep SS 100*4.6 mm (L*ID) SN460175 and Varian Chromsep SS 100*4.6 mm (L*ID) SN375687) were used. A Shimadzu HPLC system was used, including: a UV-VIS detector (SPD-10AVP), a LC-10AD VP and a LC-20AD prominence pump, and a Communication Bus Module (CBM-20A prominence). The HPLC analysis time was 60 minutes and a binary gradient was used. With a flow of 0.5 mL/min, the gradient started with 100% hexane and after 7.00 minutes the system started pumping DCM. The RID (RID-10A) was set at 40.0°. The rinsing and sampling speed were 35 and 15 μL/s respectively and the purge time was 25 min. Furthermore, the LabSolutions software was used.

To quantify the mineral oils A, F, G, E, and H, a sample of each mineral oil was prepared containing mineral oil (2 mg/mL) and internal standards (12 mg/ mL each). These internal standards consisted of C11,

C13, Cho, 6B, 9B, and TBB. 20 μL of each mineral oil sample was injected into the HPLC-system. The

MOSH-fraction was collected between 4.5 and 6.3 minutes, whereas the MOAH-fraction was collected between 6.3 and 35 minutes. Fractions of mineral oils A, F, G, E, and H were also collected according to the literature beginning of MOAH, in which the elution time of TBB is used as the beginning of MOAH. Again, the MOSH-fraction was collected between 4.5 and 6.3 minutes, but MOAH was collected from 9.1 minutes and stopped at 35 minutes.

After the LC fraction collecting, the fractions were evaporated until 0.5 mL of sample was left and 1 μL these remaining samples were introduced onto the GC-FID system. Each sample was measured three times. An Agilent 6890N (G1540N, sn US10731006) system was used, along with an GC PAL CTC ANALYTICS autosampler, an Agilent flame ionization detector, and a J&W Scientific column (15m, 320 μm, 0.1μm, DB5HT). Helium was used as a carrier gas and the injection mode was splitless. The column flow was constantly 2.00 mL/min with a purge time of 120 s. The inlet temperature was set at 350 °C for the entire run. The temperature ramp of the GC started at 60°C and lasted for 3.00 minutes, after which the temperature increased to 350°C with a rate of 15°C. For the FID, nitrogen was used as a makeup gas and the constant makeup and column flow were set at 40. The hydrogen and air flow were 36 and 450 respectively. Furthermore, the temperature of the detector was 350°C. The software used for the FID-integration was LECO® ChromaTOF® optimized for GCxGC FID.

(10)

10

2.4 Characterization of mineral oils: LCxGC-VUV analysis

To separate MOSH from MOAH, two Ag-loaded silica columns (Agilent Chromsep SS 100*4.6 mm (L*ID) were connected in series. A Waters 2695 Separations Module Alliance was used for the characterization of the mineral oils, along with a Waters 996 PDA detector. 40 μL of mineral oils A, E, H, H80, and T22 (50 mg/mL) were injected into the Waters-system A gradient with a flow rate of 0.5 mL/min was applied, which started at 100% hexane and after 12 minutes, the gradient switched to 50% DCM and 50% hexane. The start and end wavelength of the detector were 210 and 400 nm respectively. The fractioning of the mineral oils started at 4.00 minutes and stopped at 24.00 minutes, in which fractions were collected after 20 seconds, resulting in a total of 60 fractions per mineral oil. The total analysis time was 40 minutes per mineral oil. Subsequently, all these fractions were measured into a GC-VUV system, in which the settings are described in section 2.2. Of each fraction 2 μL was injected into the system and three detector filters were applied: saturated (125-160 nm), aromatic (190-240), and total mineral oil (125-240 nm). Data was analyzed using the VUV Data Viewer (v5.05.527) and plots were created with MatLab (R2018a).

(11)

11

3 Results and Discussion

3.1 Quantification using LC-GC-FID

After the LC-GC-FID measurements (Appendix 8.2.2) of mineral oils A, F, G, E, and H (Appendix 8.2.1), the corresponding concentrations and ratios were calculated for the quantification of mineral oils (Table 1).

Table 1: Results of the GC-FID measurements of the LC-fractions for the quantification of mineral oils. The beginning and end of the fractions were based on the LC-RID-chromatogram and its results are shown in the column ‘Avg % MOAH’. ‘TBB %MOAH’ corresponds to the fractions in which TBB was used as a marker for MOAH (the concentration, SD, and CoV of this fraction set is shown in Appendix 8.2.3). Results of earlier performed SPE-separation of MOSH and MOAH (SPE %MOAH) were used to compare the two pre-separation methods.

MO Fraction Avg conc

(mg/mL)

Avg SD Avg CoV Avg % MOAH TBB % MOAH SPE % MOAH

A MOSH 997.30 207.92 21.07 31% 21% 29% MOAH 483.58 59.55 7.42 F MOSH 681.18 88.92 12.01 45% 40% 47% MOAH 566.51 144.05 26.21 G MOSH 990.94 149.82 14.91 37% 30% 34% MOAH 613.24 46.40 8.30 E MOSH 776.64 79.72 10.01 42% 42% 38% MOAH 570.91 73.89 13.25 H MOSH 1103.40 251.03 20.17 28% 24% 28% MOAH 421.19 34.57 8.11

Since the mineral oil samples contained internal standards with known concentrations, it was possible to calculate the concentration of MOSH, MOAH, and subsequently the percentage of MOAH in the mineral oils A, F, G, E, and H. The hypothesis is that mineral oils with a high polyaromatic content are darker than MO’s with long saturated chains. The results (Table 1) show that mineral oil H contains the highest concentration of MOSH and the lowest concentration of MOAH, which was expected since MO H had the lightest color of all the quantified mineral oils. Mineral oils F and E had the darkest colors and thus the expectation was that these oils would have the highest amount of polyaromatics and hence a larger percentage of MOAH. The GC-FID results show indeed that these oils contain the highest percentages of MOAH. The only mineral oil with an unexpected ratio is MO E, the darkest of all quantified mineral oils. Since this mineral oil is darker than MO F, the hypothesis was that this oil should contain the highest percentage of aromatics, however, this can be related to the polyaromatic content; higher content of polyaromatic compounds could give a darker color.

In previous studies, TBB was used as a marker for the beginning of the MOAH-fraction13,16_{, but in}

this study the responses of the refractive index detector were used as the beginning of MOAH. As mentioned earlier, MOAH was collected after a response in the RID was visible, which was earlier than the elution time of TBB, which was also noticed by Biedermann et al.9_{This means that, if TBB is used as a marker, the}

percentage of MOAH in mineral oil would be lower than the actual percentage. To prove this hypothesis, besides measuring solely the mineral oil fractions based on the RID-responses, MO-fractions based on the marker were collected and measured as well. The results of the latter are shown in the last column of Table 1 and show generally a lower amount of MOAH than the actual percentage of aromatics up to a difference of 32% (MO A).

(12)

12

When the results are compared with the different pre-separation method SPE, it is immediately visible that the SPE- and LC-results are very common when the RID marks the beginning of MOAH. Whereas the largest deviation between the RID-method and the TBB method was 32%, the largest deviation for the SPE-method was 10%. As expected, the percentage of MOAH is for the SPE-method also higher than when TBB is used as a MOAH-marker. However, again mineral oil E contains less MOAH than expected, due to the same reason as mentioned in the previous paragraph. Hence, for reliable quantitative results, TBB should not be used as a MOAH-marker. This means that either a new MOAH-marker should be used, e.g. 1,4-di(2-ethylhexyl) benzene (DEHB) as Biedermann et al described it as a reliable new marker.9_{This new marker contains longer substituents but is less branched than TBB and thus elutes earlier}

in the LC-chromatogram. The other method is the one applied in this research: simply looking at the RID-response. When the RID gives a response, MOAH starts eluting.

The values of the coefficient of variances lead to one of the drawbacks of mineral oil analysis using splitless injection on a flame ionization detector: non-repeatable sample introduction of heavy compounds on the GC-column due to sample discrimination.17_{As heavier compounds do not volatilize as easily as}

smaller compounds, not all compounds are introduced on the column at the same time, i.e. smaller, lighter compounds enter the column earlier than heavier compounds. This explains why the responses of the small internal standards are repeatable for each measurement of the same vial, whereas the areas of the mineral oil differ. Another note is that the fractions were collected manually, contributing to the non-repeatable results. Therefore, quantification using the GC-FID method is not perfect and thus more research should be conducted to develop a method with a more repeatable sample introduction.

(13)

13

3.2 Characterization using LCxGC-VUV

3.2.1 Separation in the first dimension: Ag-Si NPLC

For the comprehensive LCxGC-VUV characterization of mineral oils, the first dimension consists of silver loaded silica NPLC. In this dimension, the separation of aromatic species is facilitated by the silver-ions present in the silica. These silver ions interact with the π-electrons in the aromatic rings, meaning that diaromatics will have more retention on the Ag-Si column in comparison with monoaromatic compounds. This concept was first tested by separating a mixture of representative aromatic internal standards: 1,3,5-tri-tert-butylbenzene, nonylbenzene, hexylbenzene, 1-methyl-naphthalene, and biphenyl (Fig. 2). As expected, the monoaromatics TBB, nonylbenzene, and hexylbenzene elute first, since these species have less interaction with the silver-ions than the diaromatic species. Since diaromatics could either consist of condensed rings or uncondensed rings, 1-methyl-naphthalene and biphenyl were chosen as internal standards for the 2-ringed compounds. These results show that the condensed rings have less retention on the silver-column and thus elute earlier than the uncondensed rings. In the LCxGC-VUV chromatogram, it would be expected that the first peak should correspond with monoaromatics, the second with condensed diaromatics and the third with uncondensed diaromatics. If there are polyaromatic species present in the mineral oil sample, those would elute at even higher retention times.

Figure 2 Chromatogram of the gradient separation of a mix of internal standards on a silver-loaded silica column, showing the retention of a selection of aromatic species.

(14)

14

3.2.2 Separation in the second dimension: GC-VUV

The second dimension in the comprehensive characterization of mineral oils is the novel GC-VUV separation. In this dimension, a temperature ramp was used to separate the molecules based on their molecular weight. The gradient started at a lower temperature and increased over time, meaning that the lighter molecules elute earlier than heavier compounds. Since the GC was coupled to a VUV-detector, more information about the composition of the mineral oils could be obtained by looking at the resulting spectra. Almost every compound exhibits unique VUV-absorption, thus the spectra should verify the aromaticity predicted with the LC-separation (Appendix 8.3.3). All VUV-spectra were normalized to easily compare the different spectra

The VUV-spectra of different monoaromatic species are shown in Figure 3. When the VUV-spectra of different monoaromatic species are compared, it can be shown that the simplest monoaromatic, benzene, absorbs at the lowest wavelength, whereas the most substituted TBB elutes at the highest wavelength. The other remarkable observation is that toluene absorbs at lower wavelengths than n-propylbenzene and nonylbenzene. This leads to the conclusion that more alkylated substituents on the monoaromatic ring result in an increase in wavelength.

Figure 3 Normalized vacuum UV Library absorbance spectra of benzene, toluene, TBB, n-propylbenzene, and nonylbenzene.

To gain a better understanding of the spectroscopic properties of diaromatic compounds, the VUV-spectra of 1-methylnaphthalene, 2-methylnaphthalene, and biphenyl were compared (Figure 4). These absorbance spectra show that condensed diaromatics, such as 1-, and 2-methylnaphthalene, absorb at higher wavelengths than uncondensed aromatics (e.g. biphenyl). The spectra also show that uncondensed rings have a broader absorption peak than naphthalene-like species. This means that it should be possible to distinguish between these two classes of 2-ring aromatics using the vacuum UV-spectra.

(15)

15

As mineral oils consist of a very complex mixture of aromatic compounds, it is also important to look at the spectral properties of polyaromatic compounds (Figure 5). The VUV-spectra of these polyaromatics show evidence that, indeed, condensed aromatic rings (e.g. anthracene) absorb at higher wavelengths than uncondensed (e.g. p-terphenyl). It seems to be that the three-ringed anthracene absorbs at higher wavelengths than the five-ringed perylene. However, at the end of the spectrum of perylene there is an increasing signal. Since the VUV-detector only measures wavelengths up to 240 nm, it is probably not possible to see that perylene in fact absorbs above 240 nm.

Figure 5 Normalized vacuum UV Library spectra of anthracene, p-terphenyl, and perylene.

Thus, the absorption of different aromatic species in the vacuum UV-spectrum showed evidence that uncondensed aromatic rings absorb at lower wavelength and have a broader absorption peak than condensed aromatics. The spectra also showed that the longer the aliphatic substituents on the aromatic ring, the higher the wavelength, i.e. the peaks shift towards the right for longer-branched aromatics. Besides the length of the substituent, but also the number of substituents on the ring influences the wavelength at which the compounds absorb. The more substituted and hence the more shielded the aromatic ring, the higher the absorbance as well. These trends will be considered when characterizing the analyzed mineral oils according to the LCxGC-VUV chromatograms in the next subsection.

(16)

16

3.2.3 Comprehensive LCxGC characterization

After the LC- and GC-VUV chromatograms were combined, 2D-chromatograms (Appendix 8.3.2) and the VUV-spectra of mineral oils A, E, H, H80, and T22 (Appendix 8.3.1) were obtained, as shown in Figures 6-15. The x-axis corresponds with the LC time, meaning that fraction 1 corresponds to 4.00 minutes, fraction 2 to 4.33 minutes, and thus fraction 60 with 23.67 minutes. The chromatogram corresponding to each fraction is displayed on the y-axis. These two dimensions combined showed 2D-plots for the characterization of mineral oils: one with the total mineral oil filter (125-240 nm) and one with the aromatic filter (190-240). The latter one was used to show the MOAH-part more detailed. The earlier obtained LCxGC-FID plots looked similar to the LCxGC-VUV plots (Appendix 8.3.5). The plots with the aromatic filter were displayed in this section and the ones with the total mineral oil filter were shown in Appendix 8.3.4.

For mineral oil A, three abundant peaks were visible in the LCxGC-VUV chromatogram (Fig. 6). When comparing the signals with the LC elution order described in Figure 2, the expectation was that the first signal should correspond with monoaromatics, whereas the second peak should be caused by diaromatic compounds. However, the GC-VUV spectra (Fig. 7) showed only absorbance around 195 nm, which is the monoaromatic region. Therefore, it is not completely sure if the second spot is really caused by diaromatic compounds, as expected from the LC-hypothesis. What is certain from the VUV-spectra is that the molecules that eluted at the second spot have different properties than the ones eluting at the first spot. Whereas the fractions in the first spot showed a similar absorption spectrum, the one in the second spot looked very different; the absorbance in the diaromatic region (200-230 nm) is higher and broader. This could indicate that either mineral oil A contains two groups of monoaromatic compound with very differing substituents: the molecules that elute in the first group were one-ring aromatics with saturated aliphatic substituents and the molecules eluting in the second group contained monoaromatics with unsaturated aliphatic chains, leading to a higher absorption in the diaromatic region. However, most probable from the Ag-Si separation (Fig. 2), the second group was composed of two-ring aromatics that are uncondensed, i.e. the two aromatic rings are connected via an aliphatic chain. Also, the response around 165 nm could be caused by two-ring aromatics. The second group could hence correspond with biphenyl-like compounds, maybe with a longer aromatic chain in the middle (e.g. diphenyl ethane).

Figure 6 2D LCxGC-VUV chromatogram of mineral oil A with the aromatic filter (190-240 nm).

Figure 7 Normalized VUV –spectra of fractions 10, 11, 12, and 35 of mineral oil A.

(17)

17

The abundant spot at the end of the LC-dimension of MO A was not caused by the mineral oil and should therefore correspond with a contamination (Appendix 8.3.4, Fig. 39,40). While the MOSH- and MOAH-chromatograms showed a hump of peaks, the chromatograms corresponding with the spot at the end of the 2D-plot consisted of many individual peaks. Only weak aromatic responses were visible in the VUV-spectra. Besides, this signal was not visible in the LCxGC-FID chromatogram (Appendix 8.3.5), leading to the assumption that this signal must be caused by contaminations. These contaminations were probably caused by the septum that was visibly present in the samples and washing solvent.

Mineral oil E showed a similar LCxGC-VUV chromatogram as mineral oil A (Fig. 8). For this mineral oil, the expectation was also that the first group of eluting compounds should be caused by monoaromatics and the second by diaromatics. Yet, the GC-VUV spectra (Fig. 9) showed a peak around 195 nm, indicating that there are only monoaromatic species inside this oil. However, the spectra of the fractions inside the second spot differed from the fractions inside the first spot, meaning that there is again a difference in the properties of these two groups of compounds. The small signal around 165 nm could be an indicator for diaromatic species. The compounds in the first spot showed the peak shifting as described in the previous section (Fig. 3), i.e. that monoaromatics absorbing at lower wavelengths contain shorter or less-branched substituents. Just as mineral oil A, the compounds in the second region had a higher spectroscopic response in the diaromatic region than the ones eluting in the first region. Furthermore, a differing response in the saturated region (125-155 nm) was visible as well. The VUV-spectra showed that the molecules in the second spot have a lower saturated response than the ones eluting earlier, meaning that the earlier eluting compounds contain more saturated parts. Even though the VUV-spectra did not show a clear diaromatic response, there could still be diaromatic compounds present inside the mineral oil which are uncondensed. This is a plausible hypothesis, since uncondensed diaromatics are proven to show a broader VUV-signal, with a lower wavelength than condensed diaromatics. This could explain why the GC-VUV response initially looks monoaromatic, whereas the LC-dimension gives a diaromatic response.

Figure 8 2D LCxGC-VUV chromatogram of mineral oil E with the aromatic filter (190-240 nm).

Figure 9 Normalized VUV –spectra of fractions 13, 15, 17, 36, and 37 of mineral oil E.

(18)

18

Mineral oil H showed again two spots in the LCxGC-VUV chromatogram (Fig. 10) and, following the LC-hypothesis, the first spot again seemed to be caused by one-ring aromatics and the second by two-ring aromatics. Contrary to mineral oils A and E, the GC-VUV spectra (Fig. 11 and Fig 38 (Appendix 8.3.4)) did show clear evidence that this mineral oil contains diaromatic compounds. The spectrum of the fraction in the region at which the second group of compounds eluted showed three peaks: one around 165 nm, one around 200 nm and the last one around 230 nm. The 165 nm peak could again indicate diaromatic species, but the high absorption around 230 nm could also suggest the presence of polyaromatic compounds. It is therefore most probable that mineral oil H not only contains one- and two-ring aromatics, but polyaromatics as well. The high absorbance in the diaromatic region could also mean that mineral oil H contains mostly condensed di- or polyaromatics. Furthermore, another remarkable observation is visible in the spectra of MO H: the earlier eluting fractions absorbed at shorter wavelengths than the ones eluting later. As seen in the spectra of mineral oil E, the phenomenon of peak-shifting towards higher wavelengths was also visible in the spectra of mineral oil H, again indicating that the earlier eluting one-ring aromatics have shorter substituents or are less branched.

Figure 10 2D LCxGC-VUV chromatogram of mineral oil H

with the aromatic filter (190-240 nm). Figure 11 Normalized VUV –spectra of fractions 15, 20, 25, and 46 of mineral oil H.

Similar to mineral oil H, mineral oil H80 also showed a clear signal in the diaromatic region in both the LCxGC-VUV plot and the VUV-spectrum (Fig. 12,13). Drawing the same conclusion as in the previous paragraph, mineral oil H80 most probable consists of mono-, di-, and maybe even polyaromatic compounds. The peak-shifting is again visible for the monoaromatic compounds and could even have caused the slight separation inside the monoaromatic region: the first part probably contains monoaromatics with shorter aliphatic substituents, whereas the monoaromatics with longer or more branched substituents eluted in the second part. Again, for both fractions eluting in the second spot in the LCxGC-VUV chromatogram, the small peak around 165 nm could be considered as a signal for two-ring (or more) aromatics. The VUV-responses of the molecules in fraction 35 can be compared with the VUV-spectrum of anthracene (Fig. 5), since the peak is getting narrower around 230 nm. This could mean that this fraction consists of both two-and three-ring aromatics. The peak in the diaromatic region of 36 was quite broad, meaning that the molecules in this fraction are most probable uncondensed or contain more uncondensed rings in comparison with condensed ones. Remarkably, this absorbance also looked like the absorbance of perylene in the VUV-region (Fig.5), indicating that this fraction might contain hydrocarbons with many aromatic rings, such as 5-rings. Because a silver-loaded silica column was used for the LC-separation, it would also make sense

(19)

19

that MOAH-molecules with fewer rings (fraction 35) would elute earlier than aromatic hydrocarbons with more rings (fraction 36).

Figure 12 LCxGC-VUV chromatogram of mineral oil H80 with the aromatic filter (190-240 nm).

Figure 13 Normalized VUV –spectra of fractions 16, 19, 22, 35, and 36 of mineral oil H80.

For the last mineral oil T22, again two regions were visible in the LCxGC-VUV chromatogram (Fig.14): the first for the monoaromatics and the second for the diaromatics. A separation inside the monoaromatic region seemed to be occurred as well and, by looking at the VUV-spectra (Fig.15), it could be induced by the effect of peak-shifting. Maybe the monoaromatics in the first region could contain shorter alkylated chains or are less substituted than the ones in the second region. For the compounds eluting at the diaromatic region, it is most probable that they are indeed diaromatic but uncondensed and hence showing an absorbance at shorter wavelengths.

Figure 14 2D LCxGC-VUV chromatogram of mineral oil T22 with the aromatic filter (190-240 nm).

Figure 15 Normalized VUV –spectra of fractions 16, 20, 23, 36, and 37 of mineral oil T22.

(20)

20

Another note is that a DCM-gradient was applied after 12 minutes in the LC-dimension. This gradient was applied to reduce the analysis times, but it probably influenced the separation. Before applying the gradient, the separation was probably good, however, the appliance of the gradient probably caused all the remaining compounds to elute all around the same time. This explains why there is only one spot visible in the LCxGC chromatogram around 15 minutes and why the VUV-spectra of mineral oils H and H80 are not similar for different fractions in that region. The reason why the VUV-spectra for mineral oils A, E, and T22 in that specific region do look similar is probably because these oils do not contain polyaromatics. For MO H and H80, however, the molecules in the last region are most probable a mixture of aromatics with different number of rings, leading to unique VUV-responses.

In conclusion, the characterization of mineral oils using comprehensive LCxGC-VUV remains challenging, due to the complexity of the oil’s composition. Mineral oils are a mixture of many compounds, containing not only different number of rings, but different substituents as well. At this point of the research, the application of a gradient is probably too difficult. First, a better understanding about the chromatographic and spectroscopic properties of different kinds of aromatic hydrocarbons should be gained, before drawing definitive conclusions about the LCxGC-VUV characterization of mineral oils.

(21)

21

4 Conclusion

From the LC-GC-FID quantification, it could be concluded that TBB is not an accurate marker for the beginning of MOAH. The RID LC-GC-FID, SPE-GC-FID, and LCxGC-FID results all show higher MOAH percentages for almost every analyzed mineral oil than when TBB is used as a marker for MOAH. Therefore, a new marker for the beginning of MOAH should be used or the responses of the RID, as used in this research, should be considered in determining the beginning of MOAH. If not, the analyzed MOAH-concentrations in real samples (e.g. rice) would be too low, leading to the consequence that some products might be marked as MOAH-free, whereas there is still a harmful concentration of aromatics present inside the products.

For the characterization of mineral oils, the LCxGC-VUV method has shown similar responses as the LCxGC-FID method, yet the VUV-detector gives additional information about the composition of the mineral oil. For instance, the analysis of the VUV-spectra most probable have proven that uncondensed aromatic rings absorb at lower wavelengths and have a broader absorption peak than condensed rings. This leads to the observation that the VUV-spectra of uncondensed diaromatic compounds might seem like monoaromatics at the first glance, whereas they are in fact 2-ring aromatics. Another remarkable observation was the peak shifting in the vacuum UV-spectra: longer aromatic substituents on the aromatic ring absorb at higher wavelengths. The absorbance is also increased when the aromatic ring contains more substituents. Nevertheless, due to the complex composition of mineral oils, first isocratic characterization should be employed before applying a DCM-gradient. Consequently, the spectroscopic and chromatographic properties of different types of aromatic hydrocarbons should first be analyzed before drawing definitive conclusions about the chemical composition of mineral oils.

(22)

22

5 Outlook

As the characterization and quantification of mineral oils remains challenging, even after the insights gained from this research, some suggestions for further research will be offered. In this study, the VUV-detector was compared with the FID in the characterization of mineral oils. It was found that the new detector seems to be a promising alternative to the commonly used flame ionization detector. Since mass spectrometry is known to be a popular detection method, the performance of the GC-VUV should be compared with the GC-MS to conclude whether the VUV can be employed as an alternative to the MS. However, as mentioned earlier, the isocratic LCxGC-VUV method should first be improved, before using a gradient. Additionally, the spectroscopic properties of molecules in the VUV-region should be understood more thoroughly to better characterize these compounds.

Nevertheless, the introduction of the new VUV-detector might eliminate the requirement of a pre-separation step. Since the VUV-detector absorbs for both saturates and aromatics, a ratio of MOSH and MOAH could be determined using a single vacuum UV-spectrum of a mineral oil. When the same gas chromatograph is connected to a FID, the deduced ratio from the VUV-results can be verified and, additionally, the concentration of the saturates and aromatics can be calculated as well. Using this new GC-VUV-FID method, pre-separation would be unnecessary and thus the analysis time will be reduced, making the quantification of mineral oils much faster.

Finally, the newly developed methods should be applied on real samples, such as foodstuffs and cosmetics, to test whether these methods are sufficient for the quantification and characterization of real samples. If the methods are proven to be good, they can be used to quantify the amount of mineral oil inside consumer products and subsequently to determine the toxicity of mineral oils.

(23)

23

6 Acknowledgements

First, I would like to thank H.G. Janssen for providing me with this bachelor project and mentoring me. Next, I want to offer my special thanks to A.R. García Cicourel for the daily supervision and help throughout the project. Furthermore, I would like to extend my thanks to B. Van de Velde for assisting me with the LC fraction collecting and for helping me around at the lab. C. Kukurin is thanked for assisting me with the GC-VUV-system. Also T. Aalbers is thanked for the technical support. Furthermore, I would like to thank B.W.J. Pirok for offering me an internship in the Analytical Chemistry Group during Project Chemistry 1. Finally, I thank group leader P.J. Schoenmakers and the Analytical Chemistry Group for providing me this interesting and challenging internship and the great time I have had here.

(24)

24

7 References

(1) Fiselier, K.; Fiorini, D.; Grob, K. Anal. Chim. Acta 2009, 634 (1), 102–109. (2) Biedermann, M.; Grob, K. J. Sep. Sci. 2009, 32 (21), 3726–3737.

(3) Concin, N.; Hofstetter, G.; Plattner, B.; Tomovski, C.; Fiselier, K.; Gerritzen, K.; Fessler, S.; Windbichler, G.; Zeimet, A.; Ulmer, H.; Siegl, H.; Rieger, K.; Concin, H.; Grob, K. Food Chem. Toxicol. 2008, 46 (2), 544–552.

(4) Scientific Opinion on Mineral Oil Hydrocarbons in Food; Parma, 2012; Vol. 10. (5) Biedermann, M.; Fiselier, K.; Grob, K. J. Agric. Food Chem. 2009, 57 (19), 8711–8721. (6) Biedermann, M.; Grob, K. J. Chromatogr. A 2015, 1375, 146–153.

(7) De Koning, S.; Janssen, H. G.; Brinkman, U. A. T. J. Chromatogr. A 2004, 1058 (1–2), 217–221. (8) Gröger, T.; Gruber, B.; Harrison, D.; Saraji-Bozorgzad, M.; Mthembu, M.; Sutherland, A.;

Zimmermann, R. Anal. Chem. 2016, 88 (6), 3031–3039.

(9) Biedermann, M.; Munoz, C.; Grob, K. J. Chromatogr. A 2017, 1521, 140–149. (10) Weber, B. M.; Walsh, P.; Harynuk, J. J. Anal. Chem. 2016, 88 (11), 5809–5817. (11) Schug, K. A.; McNair, H. M.; Hinshaw, J. V. LCGC North Am. 2015, 33 (1), 1–6.

(12) Schug, K. A.; Sawicki, I.; Carlton, D. D.; Fan, H.; McNair, H. M.; Nimmo, J. P.; Kroll, P.; Smuts, J.; Walsh, P.; Harrison, D. Anal. Chem. 2014, 86 (16), 8329–8335.

(13) Zoccali, M.; Barp, L.; Beccaria, M.; Sciarrone, D.; Purcaro, G.; Mondello, L. J. Sep. Sci. 2016, 39 (3), 623–631.

(14) Lommatzsch, M.; Biedermann, M.; Simat, T. J.; Grob, K. J. Chromatogr. A 2015, 1402, 94–101. (15) Eurachem Guide: The Fitness for Purpose of Analytical Methods - A Laboratory Guide to Method

Validation and Related Topics, 2nd ed.; Magnusson, B., Örnemark, U., Eds.; 2014. (16) Biedermann, M.; Grob, K. J. Chromatogr. A 2012, 1255, 56–75.

(17) Crawford Scientific. Mass Spectrometry, Fundamental LC-MS, Electrospray Ionisation – Instrumentation papers3://publication/uuid/A2FAB40B-0D6E-4466-A544-688E45890121 (accessed May 29, 2018).

(25)

25

8 Appendix

8.1 Partial validation of the GC-VUV method

8.1.1 Linearity

The first part of the validation of the GC-VUV-method is the determination of the linearity of the system. Mineral oils A and H were full mineral oils and MO 2 only contained MOSH. For mineral oil K, MOSH and MOAH were separated using SPE and only MOAH was used for the validation measurements. Hence, the variety of the composition of these oils reflect representative samples for different types of mineral oils. Figures 16 and 17 show the linearity of mineral oils 2, A, H, and K in the saturated window and total mineral oil window and Figure 18 show the linearity of mineral oils A, H, and K in the aromatic window.

Figure 16 Calibration curve of mineral oils 2, A, H, and K in the saturated region (125-160 nm).

Figure 17 Calibration curve of mineral oils 2, A, H, and K in the total mineral oil region (125-240 nm).

Figure 18 Calibration curve of mineral oils A, H, and K in the aromatic region (190-240 nm).

These calibration curves show decent linearity for all used filters. Because of the pre-separation of mineral oil 2 using SPE, only the MOSH-fraction was measured, meaning that there is no calibration curve in the aromatic region for this mineral oil. A calibration with a different aromatic window (185-240 nm) was also performed, but because the results were almost identical to those of the standard aromatic window (190-240 nm), the standard window will be used.

(26)

26

8.1.2 Detection and quantification limits and precision

As a second part of the validation of the GC-VUV-method, the limit of detection, limit of quantification, and the precision were determined according to the Eurachem guide. These results are shown in Table 2.

Table 2: GC-VUV LOD and LOQ for saturated hydrocarbons (125-160 nm), aromatic hydrocarbons (190-240) and total mineral oil (125-240 nm), with background subtraction.

Filter Mineral Oil LOD (mg/mL) LOQ (mg/mL) CoV (%)

Saturates 2 0.0258 0.0860 2.29 A 0.0148 0.0493 -10.62 H 0.0171 0.0569 1.93 K 0.0231 0.0770 7.63 Average 0.0202 0.0673 0.31 Aromatics (190-240 nm) 2 - - - A 0.0015 0.0049 47.60 H 0.0031 0.0103 9.56 K 0.0100 0.0335 55.35 Average 0.0049 0.0162 37.50 Total MO 2 0.0087 0.0520 7.14 A 0.0032 0.0107 9.03 H 0.0074 0.0247 3.72 K 0.0112 0.0372 14.74 Average 0.0076 0.0312 8.66

The limit of detection for the saturated hydrocarbon filter, the aromatic hydrocarbon and the whole mineral oil is 0.03, 0.01 and 0.01 mg/mL respectively. Therefore, a minimum concentration of 0.03 mg/mL is the lowest detectable concentration. Furthermore, the limit of quantification for the saturates, the aromatic, and total mineral oil is 0.09, 0.034, and 0.032 mg/mL respectively, meaning that the lowest concentration for quantitative analysis on the VUV-detector is 0.09 mg/mL.

When the coefficient of variation (CoV) is taken into consideration, it is clearly visible that the variance of the aromatic filter is very large (around 37.5-40.7%). This could be because the fluctuation in the responses are proportionally much larger for the aromatic filter than for the saturated or total mineral oil filter, since amount of MOAH is usually smaller than the amount of MOSH. Using these CoV-values, the detection limits are approximately 0.03, 0.01, and 0.01 mg/mL for the saturated, the two aromatic and the total mineral oil filters respectively. Therefore, the limits of quantification for these filters are 0.09, 0.03, and lastly 0.03 mg/mL. Usually, the mineral oil samples have a concentration of 5 mg/mL. With 0.03 mg/mL as the highest limit of detection for all three filters and a standard injection volume of 1 μL, the lowest percentage of MOAH that can be detected in mineral oil using the GC-VUV-method is 0.6%.

(27)

27

Furthermore, the precision was determined using the CoV-values of the second-lowest, the middle, and the highest concentration of the analyzed mineral oils (Table 3). Since these measurements were only investigated on a single laboratory, the reproducibility cannot be determined, but the repeatability can be. These results show that the CoV is not always below 10%, meaning that the precision of this GC-VUV method is not always good enough. Therefore, the method should be validated more thoroughly to draw more reliable conclusions about the validation of this method.

Table 3: The precision of the GC-VUV-method expressed as the coefficient of variation (CoV) for the second-lowest concentration (1 mg/mL), the middle concentration (5 mg/mL) and the highest concentration (20 mg/mL for MO 2, A, and H and 10 mg/mL for MO K).

Filter Conc. (mg/mL) CoV (%)

MO 2 MO A MO H MO K Saturates 1 2.29 -10.62 1.93 7.63 5 5.29 0.18 1.92 41.17 20 (10 for MO k) 7.88 2.31 3.26 34.65 Aromatic 1 - 47.60 9.56 55.35 5 - 1.67 1.59 6.65 20 (10 for MO k) - 2.32 3.51 0.64 Total MO 1 7.14 9.03 3.72 14.74 5 5.20 0.29 1.79 6.89 20 (10 for MO k) 7.96 2.28 3.30 0.58

(28)

28

8.2 LC-GC-FID Quantification

8.2.1 MO samples A, F, G, E, and H

(29)

29

8.2.2 GC-FID chromatograms

Figure 21 GC-FID chromatograms MO A.

(30)

30

(31)

31

(32)

32

(33)

33

(34)

34

8.2.3 GC-FID calculations

Table 4: Results of the GC-FID measurements of the MOSH and MOAH LC-fractions including: the average concentration of the fractions, the standard deviation (SD), the coefficient of variation (CoV) and the percentage of MOAH in the mineral oil (% MOAH). 1, 2, and 3 correspond to the number of the fraction set (3 fraction sets in total) and lit corresponds to the fractions that were collected when TBB was used as a MOAH-marker.

Mineral Oil Fraction Avg Conc (mg/mL) SD CoV % MOAH

A MOSH (1) 616.396845 105.817496 17.1671053 29% MOSH (2) 751.123576 197.655241 26.3146101 27% MOSH (3) 1624.38845 320.291589 19.7176722 36% MOSH (lit) 1033.21502 181.650645 17.5811077 21% MOAH (1) 246.459125 5.14451018 2.08736851 MOAH (2) 284.793077 5.34944997 1.87836377 MOAH (3) 919.485907 168.150455 18.2874424 MOAH (lit) 280.806976 29.6901997 10.5731703 F MOSH (1) 416.1422 5.036166 1.210203 49% MOSH (2) 380.7388 75.73621 19.89191 40% MOSH (3) 1246.664 185.9989 14.91973 46% MOSH (lit) 625.5601 21.75722 3.478039 40% MOAH (1) 398.4252 173.2001 43.47118 MOAH (2) 258.1972 35.43939 13.72571 MOAH (3) 1042.91 223.5224 21.43256 MOAH (lit) 412.8053 26.68304 6.463833 G MOSH (1) 694.350851 194.673613 28.0367789 35% MOSH (2) 693.049249 7.50740291 1.08324234 33% MOSH (3) 1585.40805 247.265035 15.5963024 42% MOSH (lit) 842.750543 10.1951486 1.20974691 30% MOAH (1) 373.160822 56.167282 15.0517628 MOAH (2) 337.109499 11.9872174 3.55588243 MOAH (3) 1129.46224 71.0514247 6.2907304 MOAH (lit) 364.512808 70.4476465 19.3265216 E MOSH (1) 892.2749 122.4071 13.71854 38% MOSH (2) 714.2819 94.72936132 13.26218153 44% MOSH (3) 723.3701 22.03813 3.046591 45% MOSH (lit) 720.4576 57.80266 8.023047 42% MOAH (1) 541.9519 131.1551 24.2005 MOAH (2) 571.6181 55.900409 9.779328399 MOAH (3) 599.1502 34.61547 5.777429 MOAH (lit) 524.2864 18.96888 3.618037 H MOSH (1) 985.751447 85.4771942 8.67127251 29% MOSH (2) 1309.25253 628.932082 48.0374921 23% MOSH (3) 1015.18589 38.6864093 3.81077099 32% MOSH (lit) 1109.3071 25.3335329 2.28372583 24% MOAH (1) 395.139822 37.0847103 9.38521208 MOAH (2) 394.174631 20.8739238 5.29560305 MOAH (3) 474.246201 45.7552741 9.64800012 MOAH (lit) 346.80844 6.98166036 2.0131172

(35)

35

8.3 LCxGC-VUV characterization

8.3.1 MO samples A, E, H, H80, and T22

Figure 25 Mineral oil samples A, E, H, H80, and T22.

8.3.2 Ag-Si HPLC chromatograms

Figure 26 HPLC chromatogram obtained from the silver silica separation of mineral oil A.

(36)

36

Figure 28 HPLC chromatogram obtained from the silver silica separation of mineral oil H.

Figure 29 HPLC chromatogram obtained from the silver silica separation of mineral oil H80.

(37)

37

8.3.3 Typical MOSH and MOAH GC-VUV-spectra

Figure 31 Typical GC-VUV-spectrum of MOSH, spectra taken from fraction 5 of mineral oil T22.

(38)

38

8.3.4 LCxGC-VUV results

Figure 33 2D LCxGC-VUV chromatogram of mineral oil A with the total mineral oil filter (125-240 nm).

Figure 34 2D LCxGC-VUV chromatogram of mineral oil E with the total mineral oil filter (125-240 nm).

Figure 35 2D LCxGC-VUV chromatogram of mineral oil H with the total mineral oil filter (125-240 nm).

Figure 36 LCxGC-VUV chromatogram of mineral oil H80 with the total mineral oil filter (125-240 nm).

Figure 37 2D LCxGC-VUV chromatogram of mineral oil T22 with the total mineral oil filter (125-240 nm).

Figure 38 VUV –spectra of fractions 15, 20, 25, 45, and 46 of mineral oil H.

(39)

39

Figure 39 VUV-response of mineral oil A, fraction 52.

Figure 4022 response of mineral oil A, fraction 52. The spectrum was obtained from one part of the VUV-chromatogram.

(40)

40

8.3.5 Gradient LCxGC-FID results

Figure 41 2D LCxGC-FID chromatogram of mineral oil A. Figure 42 2D LCxGC-FID chromatogram of mineral oil A.

Figure 43 2D LCxGC-FID chromatogram of mineral oil H. Figure 44 2D LCxGC-FID chromatogram of mineral oil H80.

Comprehensive LCxGC Characterization of Mineral Oils Using Flame Ionization Detection and Vacuum Ultraviolet Detection

Bachelor Thesis Chemistry