Introduction

Lubricants are essential to the successful operation of automotive engines and industrial machinery [1]. Their primary goal is to reduce friction between two adjacent surfaces, thus reducing the heat generated and extending the lifetime of the moving parts. Additionally, lubricants help to seal and cool adjacent surfaces and remove contaminants from these areas [2]. Increasing number of industrial facilities, more frequent ground transportation, and more extensive commercial goods’ transportation present a continuing demand for high-quality lubricants.

The desirable physical and chemical characteristics of a lubricant include a high boiling point and low freezing point, resistance to oxidation and other reactions, thermal stability, and a high viscosity index [3]. These properties depend on the chemical composition of the lubricant. Therefore, the ability to rapidly and accurately identify and quantitate compounds in lubricant base oils is important to more effectively evaluate and modify the refinery methods used to convert crude oil to lubricant base oils [4].

Liquid lubricants in industrial machinery and automotive engines are complex mixtures of chemicals containing 90% of base oil (by volume) and 10% of additives (detergents, antioxidants, antiwear and antifoam agents) [2]. Base oils are obtained from the heavy saturated hydrocarbon fraction of crude oil via various refining processes, including hydrotreating, hydroisomerization, and solvent extraction [5]. They are comprised of linear alkanes (n-alkanes), branched alkanes (isoalkanes), and one- to six-ring cyclic alkanes (naphthenes). The exact chemical composition of each compound class and their relative amounts in the base oil will influence the physical and chemical performance of the final lubricant product.

Heavy linear alkanes (with more than 15 carbon atoms) are detrimental to the flow properties of lubricants due to their high pour points, which leads to wax precipitation [6]. Hence, refining processes have been designed to convert large linear alkanes in the heavy saturated hydrocarbon fraction of crude oil to branched and cyclic alkanes, which have superior fluidity at low temperatures. The ability to identify and quantify linear alkanes in base oils and lubricants is necessary for an improved evaluation of the success of the refining processes and the anticipated physical and chemical properties of the base oil product.

Gas chromatography (GC) has been commonly utilized for the separation and analysis of mixtures of volatile hydrocarbons [7,8,9,10,11,12,13]. Successful separation of all the mixture components largely depends on the complexity of the sample; the greater the number of similar compounds in the mixture, the more difficult it is to separate and identify them [14]. In particular, examination of base oil chromatograms collected by GC has revealed many unresolved compounds due to the presence of multiple isomers of alkanes and cycloalkanes in the base oils [15]. Therefore, a more powerful method is needed to completely separate the compounds in these samples for accurate identification and quantitation. The use of two consecutive GC columns (GC×GC) instead of just one has been demonstrated to be substantially more powerful for base oil analysis, leading to improved separation and an increase in the number of compounds detected and improved sensitivity [15]. The superior attributes of GC×GC arise from the use of two columns of different phases (i.e., nonpolar and polar), and a modulating system that cryofocuses the effluent coming off the first column and injects it quickly into the second column. Recent literature shows many examples where GC×GC has enabled superior separation of linear alkanes from each other and from other compounds in crude oil samples [14, 16,17,18,19,20]. However, to the best of our knowledge, only a handful of studies have used GC×GC specifically for chemical characterization of base oils and semi-quantitation of the compounds in them [15, 21, 22] Currently, GC×GC with a flame ionization detector (FID) is commonly employed for the quantitation of individual compounds in complex mixtures as it provides signals for organic compounds that are proportional to their concentrations, which is not always true for mass spectrometry. However, GC×GC/FID cannot be used to unambiguously identify the eluting compounds [23]. The use of a high-resolution time-of-flight mass spectrometer (TOF MS) with an electron ionization (EI) source coupled to GC×GC often enables compound identification even in cases where two or more compounds have similar retention times [16]. EI mass spectra can be used to distinguish linear alkanes as a class from individual branched and cyclic alkanes due to their unique fragmentation patterns upon EI. However, identification of large linear alkanes requires the measurement of both the EI mass spectra and the retention times in both columns. Although this approach has mostly been used for the identification of compounds in complex mixtures, several studies have demonstrated successful quantitation of mixtures by using a set of appropriate model compounds to obtain compound type–specific response factors [16, 21, 24, 25]. The major concern here is an ionization bias that may exist for certain compounds in the mixture. Although alkane model compounds have been used to determine response factors for quantitation of alkanes in base oils by using GC×GC/EI TOF MS [21], this approach has not been validated. Unfortunately, most petroleum companies will not utilize new methods unless they have been validated for a specific task due to the relatively large cost of mass spectrometry instrumentation [26].

In this study, an accurate and precise GC×GC/EI TOF MS method was developed for the identification and quantitation of linear alkanes in base oil samples. The method was used to identify and quantitate the linear alkanes present in two base oil samples. For comparison, a GC×GC/FID instrument was also used to quantitate the linear alkanes in the same base oils.

Experimental

Chemicals

Linear butylbenzene (≥ 99% purity), hexane (≥ 97%), isooctane (≥ 99.8% purity), and other alkane model compounds ranging from tetradecane (14 carbon atoms) to heptacosane (27 carbon atoms) (≥ 99%) were purchased from Sigma-Aldrich and used as received to prepare standard solutions. Two group II base oils, Base Oil A and Base Oil B, were studied.

GC×GC/EI TOF Mass Spectrometry

GC×GC/MS analysis was carried out by using an Agilent 7890A GC×GC coupled to a Pegasus HRT 4D (EI) TOF mass spectrometer (Leco Co., St. Joseph, MI, USA). Ultrapure helium (99.9999%) was used as the carrier gas for GC×GC at a constant flow rate of 1.25 mL/min. Separation of the base oil components dissolved in n-hexane was achieved by using a reverse column configuration comprised of a primary polar 60-m capillary column and a secondary nonpolar 2-m capillary column. A quad-jet dual-stage thermal modulator positioned between the primary and secondary columns was used to cryofocus the incoming eluent off the first column and inject it quickly into the second column. The modulator time was 5 s with a hot jet pulse time of 0.75 s. The primary oven was set at an initial temperature of 65 °C and held there for 1 min. After that, the temperature was ramped at a rate of 1 °C per minute until reaching a final temperature of 270 °C, which was held for 6 min. The secondary column temperature was offset by + 10 °C from the primary column temperature, and the modulator temperature was offset by + 80 °C from the secondary column temperature. Table S1 displays a detailed overview of both GC columns and the separation method.

Compounds eluting from the GC×GC entered the electron ionization (EI) source region through a transfer line that was held at 300 °C. The compounds were ionized by EI at 70 eV electron energy and at 10−4 Torr pressure. The EI source ionizes incoming compounds by bombarding them with a beam of high-energy electrons. Collisions between the electrons and molecules result in ejection of an electron from the molecule, generating a molecular radical cation (molecular ion). The molecular ions generally fragment due to the large amount of internal energy imparted into them upon 70 eV EI. The analyte and fragment ions were accelerated and focused into the TOF MS (20-m flight path) for high-resolution measurements (25,000 for ion of m/z 219). An acquisition delay of 350 s at the beginning of the mass spectrometry analysis was employed to extend the life of the filament and prevent ionization of the solvent, linear hexane. Perfluorotributylamine (PFTBA) was used to calibrate and tune the instrument daily. The mass spectral acquisition rate was set to 200 mass spectra/s. The detected mass spectral range was m/z 15–500. LECO Visual Basic Scripting (VBS) software, ChromaTOF Version 1.90.60.0.43266, was used to process and analyze the data.

All model compound mixtures were prepared by performing serial dilutions of a stock solution. The stock solution was prepared by dissolving 10.0 mg of each linear alkane model compound in 10.0 mL of hexane. The final stock solution contained 4550 ppm (by weight) of each model compound. Six more diluted solutions with mass fractions of 4.55, 15.2, 25.8, 36.4, 47.0, and 57.7 ppm (by weight) were prepared from the stock solution. The total volume of each diluted solution studied was 1.00 mL.

Each model compound mixture was doped with 20.0 μL of a 3030 ppm (by weight) stock solution of linear butylbenzene in linear hexane, which served as the internal standard. The ideal mass fraction for analysis of each base oil solution in linear hexane was experimentally determined by analyzing 2D chromatograms of base oil solutions prepared at different mass fractions. The criteria for an ideal mass fraction were the measurement of a signal-to-noise ratio above 3:1 for the total ion current signals for the linear alkanes in each base oil and the observation that the GC×GC peak areas for the linear alkanes were below the saturation limit of the detector (1 × 107 arbitrary units). The peak area is the volume underneath the peak that is generated by both the x (time compound elutes off of the primary column) and y (time compound elutes off of the secondary column) retention times for a specific compound. The mass fraction of Base Oil A was chosen to be 4550 ppm and the mass fraction of Base Oil B was chosen to be 11,400 ppm in 1 mL of linear hexane. A 0.5 μL aliquot of each base oil solution was injected into a split/splitless injection inlet held at 260 °C, with a 1:20 split ratio, by an Agilent G4513A auto injector.

Accuracy and Precision Determination

Validation of the GC×GC/EI TOF MS method for quantitation of linear alkanes was based on the criteria detailed by Shah et al. [27]. Based on their work, acceptable accuracy percentages for a new validated method for quantitation must be greater than 85% and precision percentages less than 15% except at the lower limit of quantitation (LLOQ), where the accuracy must be greater than 80% and the precision less than 20%. Accuracy and precision determinations were conducted to evaluate the calibration plots made for each linear alkane model compound. Six mass fractions of each linear alkane were selected to generate the calibration plots wherein the y-axis was the peak area of the linear alkane divided by the peak area of the internal standard, and the x-axis was the mass fraction of the linear alkane expressed as ppm (weight/weight). The six different linear alkane solutions were examined on three separate days and the y-axis values were averaged. The R2 values for all five linear alkane calibration plots were greater than 0.996.

Four different mass fractions of each linear alkane that were not used to generate the calibration plots were used to validate the calibration plots. The GC×GC peak areas measured for both columns for these four mass fractions of each linear alkane were assigned as the theoretical peak areas (TP). The experimental peak areas (EP) for the four samples were determined by the calibration plots. Percentage accuracy was calculated using the formula 100 − (\( \frac{\left|\mathrm{EP}-\mathrm{TP}\right|}{\mathrm{TP}}\Big) \) × 100 [28]. Within-day and between-day accuracy and total precision values were derived from nine sets of measurements at each of the four mass fractions, which were carried out over 3 days. Within-day and between-day precision values were obtained using Eqs. (1) and (2), respectively:

$$ \mathrm{Within}\ \mathrm{Day}\ \mathrm{Precision}\ \left(\mathrm{W}\right)=\left({\mathrm{STDEV}}_{\mathrm{AUC}}\times 100\right)/{\mathrm{AVG}}_{\mathrm{AUC}} $$
(1)
$$ \mathrm{Between}\ \mathrm{Day}\ \mathrm{Precision}\ \left(\mathrm{B}\right)={\left(\frac{{\mathrm{SRA}}^2}{\mathrm{GA}}-\frac{\mathrm{Average}\ {\mathrm{W}}^2}{N}\right)}^{\frac{1}{2}} $$
(2)

where AUC (in Eq. 1) is the area of the GC×GC chromatogram peak for both columns, SRA (in Eq. 2) is the standard deviation of the 3-day averages of each linear alkane peak area divided by the internal standard peak area, GA is the average of each linear alkane peak area divided by the internal standard peak area of all nine experiments, and N is the total number of experiments (nine) [29]. When determining the total precision, the within-day (W) and between-day (B) precision values were considered, as shown in Eq. (3):

$$ \mathrm{Total}\ \mathrm{Precision}={\left({B}^2+{W}^2\right)}^{\frac{1}{2}} $$
(3)

Total Linear Alkane Content Determinations

Base Oil A and Base Oil B were analyzed once a day on the GC×GC/EI TOF MS for 2 days. The mass fractions of the different types of linear alkanes present in each base oil sample were determined by using the calibration plots of each linear alkane. The mass fractions were added together to get a sum for the total linear alkane content in each base oil. The sum was then divided by the mass fraction of the base oil that was injected into the GC×GC/EI TOF MS (4550 ppm for Base Oil A and 11,400 ppm for Base Oil B) to obtain the percentage of linear alkanes present in the entire base oil.

Results and Discussion

Determination of the Range of the Number of Carbon Atoms in Compounds in Lubricant Base Oil Samples

In order to prepare proper standard alkane mixtures for characterization of base oil samples, some prior knowledge on their chemical compositions is valuable. However, this is not required as the identification of linear alkanes in base oil samples by using GC×GC/EI TOF MS can be achieved by considering the EI mass spectra to verify that the compounds are linear alkanes, followed by a comparison of their retention times in both columns with those measured for standard linear alkanes in an artificial mixture. However, without any prior knowledge on the base oil sample, a mixture of linear alkane standards with 10–40 carbon atoms (typical for base oils) has to be examined. For consideration of time and cost, the carbon atom range of the linear alkanes was first investigated using APCI-LQIT MS. A literature method [30] based on APCI (O2, isooctane)/linear quadrupole ion trap (LQIT) MS (for details, see supplemental material) was utilized to obtain information on the sizes of the linear alkanes in the two base oil samples [30] prior to GC×GC/EI TOF MS analysis. The APCI (O2, isooctane)/LQIT MS method generates predominantly the [M–H]+ ion for each alkane [30]. Figure 1 displays the APCI mass spectra measured for Base Oil A and Base Oil B. A bimodal distribution of ions was observed for both base oils. Based on literature [30], the first distribution of ions represents the fragment ion region (and is denoted as such in Figure 1), and the second distribution of ions represents the [M–H]+ ions presumably formed upon hydride abstraction from the analyte molecules by the tert-butyl cation (the reagent ion formed from isooctane upon APCI) [30]. Based on Figure 1, the compounds in Base Oil A and Base Oil B contain 14–21 and 15–27 carbon atoms, respectively. It is important to emphasize here that the utilization of the APCI (O2, isooctane)/LQIT MS method is not a necessary step for the identification and quantitation of linear alkanes in base oils by using GC×GC/EI TOF MS. It was used here to quickly determine the number of carbon atoms in the alkanes in each base oil in order.

Figure 1
figure 1

Positive ion mode APCI/O2/isooctane LQIT mass spectra measured Base Oil A (top) and Base Oil B (bottom) The ion of m/z 391 derived from PEEK tubing is an unavoidable contaminant. It does not guide the selection of linear alkanes.

The APCI-MS method cannot be used for linear alkane quantitation due to the technique’s inability to separate linear alkanes from branched alkanes of the same number of carbon atoms.

Identification of the Linear Alkanes in Base Oils

Based on the above results, a standard mixture of linear alkanes with 14–27 carbon atoms was prepared at a mass fraction of 37.9 ppm in hexane (total volume 1 mL) and examined using GC×GC/EI TOF MS. While the 70 eV EI mass spectra of small alkanes were different from each other, those of large alkanes were found to be similar to one another due to the absence of molecular ions and extensive fragmentation, which makes identification of these compounds challenging based on EI MS alone. For example, supplemental Figs. S1 and S2 show the EI mass spectra measured for tetradecane and pentadecane and the software-matched NIST library EI mass spectra. In both cases, the software match was incorrect. Further, supplemental Fig. S3 shows that the best software match for the mass spectrum measured for tetradecane is pentadecane in the NIST library, while supplemental Fig. S4 shows that the best software match for the mass spectrum measured for pentadecane is heptadecane in the NIST library. Hence, while EI MS enables the identification of a hydrocarbon as having a linear structure (as opposed to cyclic or branched structure), it cannot be used to differentiate larger linear alkanes from each other due to extensive fragmentation. Because of the above difficulties, the retention times measured for the compounds in the standard mixture in both columns were used to aid in the identification of the n-alkanes in the base oil samples. The GC×GC/EI TOF MS total ion chromatograms measured for Base Oil A and Base Oil B (Figure 2), each doped with 30.3 ppm of linear butylbenzene as the internal standard, revealed hundreds of compounds in the two base oils. Each dot in Figure 2 represents an individual compound.

Figure 2
figure 2

2D gas chromatograms of Base Oil A (left) and Base Oil B (right) at mass fractions of 4550 ppm and 11,400 pm in hexane respectively, obtained using GC×GC/EI TOF MS. Each black square represents a different n-alkane, each green dot represents a different isoalkane, each red dot represents a different mono-, di-, or tricyclic alkane, and each blue dot represents a different tetra- or pentacyclic alkane

The characteristic GC×GC retention times for different types of alkanes in base oils, including linear, branched, and cyclic alkanes, have been determined previously by using model compounds [21]. Based on these retention times, the different classes of compounds are differentiated in Figure 2: black squares correspond to linear alkanes; green dots correspond to isoalkanes; red dots correspond to mono-, di-, and tricyclic alkanes; and blue dots correspond to tetra- and pentacyclic alkanes. Table 1 shows the retention times for the linear alkane model compounds. The compounds in the base oils with the same retention times as the model compounds yielded similar EI mass spectra as the model compounds. Large linear alkanes with 19 up to 27 carbon atoms were not detected in either sample. The base oil samples were analyzed using various mass fractions to ensure that even linear alkanes with low mass fractions were detected. Based on the retention times, linear alkanes with 14–18 carbon atoms were identified in Base Oil A and linear alkanes with 15–18 carbon atoms in Base Oil B.

Table 1 Retention Times in the First (polar) Column and the Second (nonpolar) Column for Standard Linear Alkanes that Have Similar Retention Times as Compounds in Base Oil A and Base Oil B. N/A Denotes “Not Applicable” as a Compound with This Retention Time Was Not Found in the Case Oil Samples

Quantitation of Linear-Alkanes by GC×GC/EI TOF MS

Mixtures of standard linear alkanes with 14–18 carbon atoms were prepared at different mass fractions to create calibration plots. 4.55 ppm was identified as the lowest limit of detection of the method as this was the lowest mass fraction for all linear alkanes that demonstrated a signal-to-noise ratio greater than 3:1. 57.7 ppm was selected as the highest mass fraction for the calibration plot as this was the highest mass fraction for all linear alkanes that did not saturate the detector. The linear dynamic range is about one order of magnitude, which limits the ppm range for accurate quantitation of linear alkanes in complex samples. However, this dynamic range is comparable with what has been reported in literature [31]. Calibration plots for all five linear alkanes in the standard mixtures were constructed based on the average of three measurements for each mass fraction (Fig. S3). All five plots demonstrate excellent linearity with R2 values greater than 0.9967, with 0.9993 as the highest R2 achieved for the linear alkane containing 16 carbon atoms.

Four mass fractions of the standard alkane mixture that were not used to prepare the calibration plots were chosen to test the accuracy and precision of GC×GC/EI TOF MS quantitation of linear alkanes: 12.1, 30.3, 42.5, and 53.1 ppm. Table 2 shows the average accuracy, within-day and between-day precisions, and total precision determined based on 9 days of measurements. The larger error observed for the between-day precision values compared with the within-day precision values is most likely due to differences in daily tuning of the voltages of the instrument. Measurements for all the linear alkane mass fractions had accuracy and precision values better than the acceptable limits of 85% and 15%, except for the 14 carbon linear alkane at a mass fraction of 12.1 ppm. This mass fraction yielded the poorest values for each of these four categories; the average accuracy was 83.8%, within-day precision was 6.0%, between-day precision was 16.1%, and the total precision was 17.2%. According to the IUPAC definition of the LLOQ, the LLOQ for tetradecane was determined to be 10 ppm. Since 12.1 ppm is close to the LLOQ, the method was successfully validated for it achieved better than the acceptable minimum of 80% and the total precision achieved was lower than the acceptable maximum of 20% [27].

Table 2 Average Accuracy, Average within Day Precision, Average Between Day Precision and Total Precision for GC×GC/EI TOF MS Quantitation of Standard Equimolar Mixtures of Linear Alkanes with 14–18 Carbon Atoms at Four Different Mass Fractions

Determining the Percentage of Total Linear Alkane Content in Base Oils

All the mass fractions measured for the linear alkanes present in the base oils were within the range of the calibration plots except for the linear alkane with 15 carbon atoms in Base Oil B (supplemental Table S1). The average total linear alkane contents measured for Base Oil A and Base Oil B (two measurements each) were 146 _+ 8 ppm and 49.2 _+ 21 ppm, respectively. The total linear alkane mass fractions were divided by the total mass fractions of the diluted base oil samples to calculate the percentage of linear alkanes present. The average total linear alkane percentage in Base Oil A was 3.7% _+ 0.5% (by weight percent) and in Base Oil B, 0.8% _+ 0.2% (Table 3).

Table 3 Total Contents (by Weight Percent) of the Linear Alkanes Identified in Each Base Oil Sample, as Determined by GC×GC/EI TOF MS and GC×GC/FID

The main concern when determining the amount of a specific substance in a complex mixture by using EI MS is a possible ionization bias for certain classes of compounds over others, which is inherent to the technique. To validate the quantitation obtained using GC×GC/EI TOF MS, a GC×GC/FID possessing the same column configuration as the GC×GC/EI TOF MS was used to determine the linear alkane mass fractions in both base oil samples. More information on the GC×GC/FID operational conditions, sample preparation, and GC×GC/FID 2D total ion chromatograms can be found in the supplemental information. Each base oil sample was analyzed twice. The average linear alkane mass fraction in Base Oil A was determined to be 4.3% _+ 0.8% and in Base Oil B, 0.7% _+ 0.1% (Table 3). These values agree closely with the results obtained using the GC×GC/EI TOF MS (Table 3), suggesting that this experimental approach can be successfully used to quantitate linear alkanes in base oils. Although the GC×GC/FID is a cheaper instrument than the GC×GC/EI TOF MS, the latter instrument can be used to not only quantitate but also identify compounds in base oils.

Conclusions

A GC×GC/EI TOF MS method was validated for the identification and quantitation of linear alkanes and applied to the characterization of linear in alkanes two base oil samples. Identification of the linear alkanes in the base oils was performed by comparing the retention times and EI mass spectra of model linear alkanes with those of the compounds in the base oils. Calibration plots with appropriate standard linear alkanes were created to quantitate the linear alkanes in the base oils. The accuracy and precision of the measurement of the mass fractions of linear alkane model compound mixtures surpassed the minimum requirements for validation of this method for quantitative purposes [27]. This work expands the benefits of GC×GC/EI TOF MS beyond structural characterization of components of complex mixtures as it also can be used to quantitate alkanes in mixtures.