n-Alkane Profiles of Engine Lubricating Oil and ... - ACS Publications

S. Weaver , Thomas W. Hesterberg , Timothy Larson , Mark Davey and L.-J. Sally Liu ... Dayue Shang , Grace Park , Pamela Brunswick , Graham van Ag...
0 downloads 0 Views 123KB Size
Environ. Sci. Technol. 2007, 41, 3697-3701

n-Alkane Profiles of Engine Lubricating Oil and Particulate Matter by Molecular Sieve Extraction GIANNI A. CARAVAGGIO,† J E A N - P I E R R E C H A R L A N D , * ,† PENNY MACDONALD,† AND LISA GRAHAM‡ CANMET Energy Technology Center-Ottawa, Natural Resources Canada, Ottawa, Ontario, Canada, Emissions Research and Measurement Division, Environment Canada, Ottawa, Ontario, Canada

As part of the Canadian Atmospheric Fine Particle Research Program to obtain reliable primary source emission profiles, a molecular sieve method was developed to reliably determine n-alkanes in lubricating oils, vehicle emissions, and mobile source dominated ambient particulate matter (PM). This work was also initiated to better calculate carbon preference index values (CPI: the ratio of the sums of odd over even n-alkanes), a parameter for estimating anthropogenic versus biogenic contributions in PM. n-Alkanes in lubricating oil and mobile source dominated PM are difficult to identify and quantify by gas chromatography due to the presence of similar components that cannot be fully resolved. This results in a hump, the unresolved complex mixture (UCM) that leads to incorrect n-alkane concentrations and CPI values. The sieve method yielded better chromatography, unambiguous identification of n-alkanes and allowed examination of differences between n-alkane profiles in light (LDV) and heavy duty vehicle (HDV) lubricating oils that would have been otherwise difficult. These profile differences made it possible to relate the LDV profile to that of the PM samples collected during a tunnel study in August 2001 near Vancouver (British Columbia, Canada). The n-alkane PM data revealed that longer sampling times result in a negative artifact, i.e., the desorption of the more volatile n-alkanes from the filters. Furthermore, the sieve procedure yielded n-alkane data that allowed calculation of accurate CPI values for lubricating oils and PM samples. Finally, this method may prove helpful in estimating the respective diesel and gasoline contributions to ambient PM.

Introduction It has been recognized that elevated particulate matter (PM) concentrations in ambient air produce adverse health effects (1). Identification of the various PM sources and assessment of their chemical composition are important steps in the management of air quality. In view of this, the Canadian Atmospheric Fine Particle Research Program was implemented to develop sampling and analytical tools to obtain PM emission profiles. These profiles from various primary * Corresponding author address phone: (613) 995-5751; fax: (613) 996-8646; e-mail: [email protected]. † Natural Resources Canada. ‡ Environment Canada. 10.1021/es062233h CCC: $37.00 Published on Web 04/17/2007

Published 2007 by the Am. Chem. Soc.

PM sources would be used to determine their relative contribution to fine atmospheric particles. The knowledge gathered from the project will be used by the Canadian government to set future regulations and standards for air quality and fuels. As part of this program, Environment Canada initiated the Pacific 2001 study (August 2001, British Columbia) to collect ambient and vehicle PM samples. The data from this study will be used to generate transportation emission profiles for use in source apportionment. In addition, these data can be used to derive the carbon preference index (CPI), a parameter to estimate the contribution of anthropogenic and biogenic sources to ambient PM (2). The CPI is the relative abundance of odd versus even n-alkanes measured in the soluble organic fraction of PM. However, there are cases, e.g., lubricating oil and ambient aerosols with a high contribution from transportation PM, where the CPI values and n-alkane profiles cannot be reliably estimated because GC/MS analysis of the n-alkanes is difficult. This arises from the presence of other similar saturated nonpolar compounds, i.e., branched alkanes, cycloalkanes, and alkylated cycloalkanes, which are difficult to resolve by GC. They are also difficult to separate by mass spectrometry because of their similar fragmentation patterns. The resulting chromatograms contain a large hump commonly described as the unresolved complex mixture (UCM), which interferes with the chromatographic separation of the majority of n-alkanes. To address this issue, different methods are currently used to extract and analyze n-alkanes in samples containing a UCM. One of them, applied to complex mixtures found in sediments, petroleum crude, and petroleum-derived fractions, is urea adduction (3-6). It was originally thought that urea adduction was specific to n-alkanes but many exceptions are now known (7). Recently, Shiping et al. (8) showed that the urea adduction method used mostly for petroleum crudes does not completely separate linear from branched alkanes. There are also many examples reporting another method based on 5 Å molecular sieves to extract n-alkanes from food (9), petroleum crude (10), bitumen (11-13), diesel fuel (14), and air PM (15). In many cases, there was either no investigation on the efficiency of the n-alkane extraction from the sieve or the recoveries reported were low. Some procedures reported were time-consuming and others used hydrofluoric acid (HF) to destroy the sieve to recover the n-alkanes. Both the urea adduction and molecular sieve methods have shortcomings but in our view, based on the literature, the molecular sieve method seems to be more specific to n-alkanes. Therefore, our laboratory focused on the molecular sieve method to adapt and improve it to analyze n-alkanes in lubricating oils and PM samples. We will present in this paper the modifications to this method. Next, n-alkane profiles in lubricating oils and in some Cassiar tunnel (Vancouver, BC) PM samples obtained from our method will be shown. Finally, this paper will discuss preliminary comparisons of the improved profiles of a few PM samples with those of related lubricating oils.

Experimental Section Sample Collection. One of the sampling sites in Vancouver was the Cassiar tunnel where emissions were measured and PM collected from representative traffic (>140 000 vehicles) over 7 days. Using fast license plate identification technology, it was estimated that the vehicle fleet composition ranged from 87 to 98% light duty gasoline vehicles. PM2.5 samples of tunnel atmosphere were collected on 90 mm diameter EMFAB filters (borosilicate microfibers reinVOL. 41, NO. 10, 2007 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

3697

forced with woven glass cloth and bonded with polytetrafluoroethylene) for semivolatile organic compounds at a flow rate of 91 L min-1. For six of the seven sampling days, a single 90 mm filter was collected over three 3 hour sampling intervals. On the seventh day, separate 90 mm filter samples were collected over each of the three 3 hour sampling intervals. The two different sampling intervals were employed to assess the loss of semivolatile compounds from the filters, also called negative artifact. In addition, eight fresh light duty gasoline (LDV) (SAE-10W30) and four fresh heavy duty diesel (HDV) (SAE-15W40) vehicle lubricating oils from different manufacturers were acquired in auto part retail stores and service stations in Vancouver for the study. These lubricating oils were chosen to represent a minimum of 75% market share. Details regarding the lubricating oils described in this paper can be found in Table S1, Supporting Information (SI). Solvents, Reagents and Standards. The molecular sieve (Aldrich, CAS 69912-79-4, 5 Å calcium aluminum silicate Linde type A zeolite) was dried in a muffle furnace at 250 °C for at least 24 h prior to use. Anhydrous sodium sulfate (BDH, granular), isooctane (2, 2, 4-trimethyl pentane, HPLC grade, Omnisolv), hexane-200 distilled in glass (J.T. Baker), cyclohexane-205, distilled in glass (Caledon), hydrochloric acid 36.5-38% (Baker “Instra” analyzed Reagent Grade) were used without further purification. Silica gel (Aldrich, 100-200 mesh, pore size 150 Å, pore 1.2 cm3 g-1, and active surface 320 m2 g-1) was activated prior to use according to the procedure detailed in the SI. An even n-alkane standard mixture (Ultra Scientific, even carbon number C10 to C40, SFL-601, 500 µg mL-1) was diluted to 50, 10, 5, and 1 µg mL-1 for use as external GC/MS calibration standards and as spiking solutions for the n-alkane recovery study. Individual deuterated n-alkane standards, dC12, dC16, dC20, dC24, dC30, dC32, dC36 (Chiron, > 98% purity, 1 mg mL-1) were prepared as a mixture and diluted to 50 µg mL-1 as a stock solution to be used for surrogates as well as for the n-alkane recovery study. PM Sample Extraction. The PM filters collected during the tunnel study were placed in extraction tubes of the DIONEX pressurized solvent extractor ASE 200 (ASE) after being spiked with the recovery standards listed in Table S2, SI. Operating conditions for the ASE are located in the SI. Silica Gel Column Chromatography. A column chromatography scheme was implemented in our laboratory to fractionate PM extracted soluble organics (see SI for description). Previous tests performed in our laboratory demonstrated that all nonpolars are collected in the first fraction and none in the next fractions. This scheme was used for PM soluble organics in this investigation to recover only the nonpolar fraction. This fraction was blown down to 1 mL using the Zymark Turbovap II evaporator and analyzed by GC/MS for n-alkanes. For the lubricating oil samples, a mass of ∼20 mg was dissolved in 1 mL of hexane and fractionated in the same manner as the PM filter soluble organics to collect the nonpolar fraction and to exclude the polar additives and detergents added by the manufacturers. This mass was chosen because it approximates the mass of nonpolar compounds found in the PM filters collected during the study, and prevents overloading the silica gel column during fractionation. The hexane fractions of both PM samples and lubricating oils were spiked with 5 µg mL-1 of the deuterated n-alkane mixture prior to the molecular sieve procedure and GC/MS analysis. Validation Solutions for the Molecular Sieve Procedure. The molecular sieve method was validated with two sets of solutions. Set A is comprised of three stock n-alkane solutions in hexane (10 mL) prepared using the SFL-601 standard and the deuterated mixture at 2.5, 5, and 10 µg mL-1. These 3698

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 41, NO. 10, 2007

solutions were used to evaluate the uptake and extraction of the n-alkanes from the sieve and to test the surrogate corrections. Surrogate corrections were calculated as follows: deuterated n-alkane recoveries were plotted as a function of carbon number and fitted by regression with a quadratic equation. From this equation, recovery factors were calculated for each n-alkane carbon number and used to adjust their respective n-alkane concentrations. Set B is made of three stock solutions of lubricating oils (∼20 mg/mL) prepared in hexane (10 mL). These solutions were cleaned up by eluting the lubricating oil with hexane on a 3 g Florisil cartridge (EXTRA-SEP column 1000 mg, 6 mL, from Lida) to remove the polar additives. The resulting hexane solution was transferred in a 10 mL volumetric flask, spiked with the deuterated n-alkane mixture at the same levels as in set A and made up to volume. Set B was used to assess the extraction of the n-alkanes from the sieve in the presence of a UCM containing matrix. An aliquot (1 mL) of each of the stock solutions from sets A and B was treated with the molecular sieve. Molecular Sieve Procedure. The dried molecular sieve was left to cool to room temperature in a desiccator and was weighed (∼0.5 g) in a single neck 100 mL round-bottom flask. Isooctane (25 mL) was added followed by the stock solution aliquots or the nonpolar fraction of either PM or lubricating oil sample. The flask was fitted to a reflux condenser and capped. The mixture was stirred using a Teflon magnetic stir bar, heated, and refluxed for 4 h with a heating mantle. The flask was then left to cool to room temperature (∼20 °C). The mixture was transferred into a Pasteur pipet containing a small plug of glass wool (∼1 cm, prerinsed with three volumes of hexane) to retain the molecular sieve. The isooctane supernatant and two additional 5 mL isooctane rinses were forced through the pipet using a stream of ultrahigh purity nitrogen. The isooctane supernatant and rinses were collected in a Zymark tube. The isooctane was either discarded or evaporated to 1 mL using a Zymark evaporator and analyzed for n-alkanes by GC/MS (to determine n-alkane uptake in to the sieve). The sieve was left in the pipet to dry overnight and extracted the next day. The molecular sieve was recovered from the Pasteur pipet and put back in the original round-bottom flask. Diluted hydrochloric acid (50 mL, 30% v v-1) and hexane (5 mL) were added to the flask that was next tightly sealed with a Teflon cap. The HCl/hexane/sieve mixture was stirred for 10 min and then sonicated (Branson model type d-150 (20 kHz) sonicator bath) for 4 h. The temperature of the bath was monitored to ensure that it did not exceed 50 °C to minimize evaporation losses. The solution was left to cool to room temperature and transferred in a separatory funnel. The aqueous phase was separated and rinsed twice with hexane (5 mL). The hexane rinses were combined and dried over sodium sulfate (∼1 g). The hexane fraction was concentrated to 1 mL using a Zymark Turbovap II concentration workstation. The samples were analyzed for n-alkanes using GC/ MS. Gas Chromatography/Mass Spectrometry (GC/MS). GC/ MS analysis was carried out in the selective ion mode (m/ z-1: 85 for n-alkanes and 66 for deuterated n-alkanes) with an Agilent 6890 GC equipped with a 5972A mass selective detector and a DB-5, 0.25 µm film, 0.25 mm ID, 30 m column. Full GC/MS parameters can be found in the SI.

Results and Discussion n-Alkane uptakes of more than 95% in the sieve were readily achieved after 4 hour reflux in isooctane. The uptake yield was determined by first analyzing n-alkanes in the isooctane supernatants obtained from the 10 ug mL-1 stock solution of set A. The supernatant concentration values were then

subtracted from the initial n-alkane concentrations of the stock solution. The uptake is the ratio of this value over the initial n-alkane concentration. Notwithstanding the straightforward uptake, the extraction of the n-alkanes adducted into the sieve proved to be more challenging. Different procedures to extract adducted n-alkanes from a molecular sieve were reported in the literature. Some using a Soxhlet apparatus (11) or extraction with an organic solvent (10), (15, 16) have been described. Others using hydrofluoric acid (12, 13) or hydrochloric acid (9) to destroy the sieve followed by hexane extraction of the n-alkanes have also been reported. In many cases, there was either no reported recoveries or those reported were low. Hauser reported overall n-alkane recoveries of 42% (15). Saito et al. reported recoveries of 40% for n-C12 and n-C13, and 80-100% for n-C15-C30 (9). In an effort to obtain a faster n-alkane extraction procedure giving higher yields, an automated solvent extractor (Dionex model 200) was tested here using different operating parameters and solvents (see Table S3, SI). Hexane was initially used under four different conditions but all attempts resulted in low yields (on average less than 40%, even n-C12-C20). Pentane was then tested and gave yields lower than hexane, averaging less than 28% (even n-C12-C20). This was not pursued any further. A Soxhlet extraction using a procedure from the literature (11) was attempted over 24 and 96 h in the anticipation of increasing the yield of n-alkanes. Both attempts resulted in yields averaging less than 38% (even n-C12-C20) and the trials were discontinued since the procedure would have become prohibitively time-consuming. An alternative route to increase yields was to remove the n-alkanes by destroying the sieve using an acid followed by their recovery with an organic solvent. The destruction of the sieve using HF was tested but this procedure resulted in recoveries averaging less than 38% (even n-C12-C40). This was deemed unsatisfactory and was discontinued. Diluted HCl, a better alternative to HF as reported by Saito et al. (9) was tested and yielded similar recoveries for pure mixtures of n-alkanes. However, in the presence of a lubricating oil matrix, we obtained lower recoveries averaging 24% (Table S4, SI). A sonication step was added to help increase the yield of n-alkanes. Sonication increased average deuterated n-alkane recoveries from 24 to 59% (Table S4, SI). The results clearly indicate that the sonication step was necessary but this matrix effect is not fully understood at this time. This phenomenon may be related to the n-alkanes becoming harder to remove because the pores of the sieve may be blocked by branched alkanes in the lubricating oil that can also enter the sieve pores (17). A series of 2-methyl alkanes (>C20) were identified in the chromatograms of the samples treated with the sieve thus supporting the previous statement (See figure S1, SI). Since the methyl group on the alkane chain is in position 2, it is hypothesized that the linear portion of the alkane chain can enter the sieve up to the branched methyl group. The results obtained with sonication (Table S4, SI) suggest that this step helps destroy the sieve structure thus allowing for a more complete recovery of n-alkanes. Finally, the method was further optimized by using a Pasteur pipet filled with a glass wool plug instead of a filter paper to recover the sieve, and yields were further increased to 71%, the average of the three concentrations shown in Table 1. Method Validation. The recoveries of the deuterated and regular n-alkanes obtained for solution set A without the lubricating oil matrix are shown in Table S5 (SI). The recoveries for C10 and C12 are lower on average than those of the longer chain n-alkanes. These lower recoveries are attributed to evaporative losses and solvent competition due to hexane, the solvent used to collect the nonpolar fraction in our column chromatography cleanup step. Tests were performed with cyclohexane to examine the solvent com-

TABLE 1. Deuterated n-Alkane Recoveries Obtained from the Sieve Method Applied to the Solutions in Set B (with the Lubricating Oil Matrix) conc. g mL-1

2.5

5

10

trials

n)8

n)8

n)8

carbon no.

avg. rec. %

std. dev. %

avg. rec. %

std. dev. %

avg. rec. %

std. dev. %

dC12 dC16 dC20 dC24 dC30 dC32 dC36 ave.

51 67 79 74 85 84 66 72

10 7 8 9 6 7 23 10

51 68 81 85 85 94 95 80

15 10 7 11 11 11 12 11

42 56 66 59 63 67 75 61

17 12 10 12 10 11 11 12

petition effect, as cyclohexane cannot enter into the pores of the sieve. Data listed in Table S6 (SI) show higher average recoveries of dC12 in cyclohexane than in hexane suggesting that the latter solvent competes with the shorter n-alkanes. Despite this effect, hexane was kept as the elution solvent in the sample column chromatography cleanup step, as it was optimized with this solvent. The results listed in Table S5 reveal that the recoveries of the deuterated n-alkanes are in the same range as those of the regular n-alkanes confirming that the sieve is equally trapping both types of n-alkanes. In addition, Table S5 (SI) lists the surrogate-corrected recoveries of the regular nalkanes. The results show that, except for C10, all corrected n-alkanes recoveries range from 82 to 108%. This confirms that the deuterated n-alkanes can be used as recovery surrogates. As for C10, its deuterated analog could be used to readjust the recovery, however, the adjusted results were acceptable since PM and lubricating oil samples do not contain much C10. Table 1 shows the recovery of the deuterated n-alkanes obtained from the sieve method on solution set B (with the lubricating oil matrix). Since the lubricating oil already contains n-alkanes, their recoveries cannot be assessed. The recoveries of dC12 are lower than those of the other deuterated n-alkanes. For dC12, the lower recoveries and high standard deviations are once more attributed to evaporative losses and to hexane competing with the lighter n-alkanes for sieve uptake. The high standard deviation observed for dC36 at 2.5 ug mL-1 is associated with peak broadening that occurs during elution of heavier n-alkanes on the GC column. This high standard deviation is due to the low peak height at this concentration. Peak broadening for dC36 at 5 and 10 ug mL-1 is negligible as the dC36 peak is more intense, thus not affecting the standard deviation. In addition, the results indicate that the average recovery at 10 ug mL-1 is lower than that at 2.5 and 5 ug mL-1. Since n-alkane uptake tested at 10 ug mL-1 was higher than 95%, this suggests incomplete destruction of the sieve during the extraction step. Even if recoveries are lower at this concentration, the data can still be successfully corrected using the surrogates. The recoveries and standard deviations listed in Table 1 demonstrate clearly that the sieve method is reproducible and can be used to reliably analyze n-alkanes in the presence of the UCM matrix. Application of the Method to Analysis of n-Alkanes in Lubricating Oils. The chromatograms in Figure 1 clearly show that there are minimal n-alkanes left in the lubricating oil after treatment and that their retention times are shifted in the UCM portion hence making their identification difficult. The shift may be due to coating of the GC column by compounds in the UCM thus changing the GC column properties. In addition, comparison of the chromatograms (b) and (c) in Figure 1 reveals that other components are coeluting with C28 and C30. In the case of C26, a poorly resolved VOL. 41, NO. 10, 2007 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

3699

FIGURE 1. Section of chromatograms (m/z-1 85) of (a) a pure mixture of n-alkanes, (b) the concentrated (1 mL) isooctane supernatant solution containing the lube L1 stripped of n-alkanes by the molecular sieve, and (c) the original L1 lubricating oil spiked with 10 ug mL-1 of even n-alkanes (C10-C40) before sieve treatment. peak doublet is apparent in the chromatogram before sieve treatment, and this prevents unambiguous identification of this n-alkane. Overall, this may lead to erroneous n-alkane profiles. This is important since these profiles are used in ambient air source apportionment. In fact, the same situation arises in PM soluble organics displaying large UCM as observed frequently in PM filter samples collected from vehicle exhaust or in ambient air samples from high-density traffic areas. Figure S2 (SI) shows the chromatograms of the lubricating oil L1 before and after the sieve treatment. Before sieve treatment, n-alkanes can only be easily identified up to C24. The chromatogram after sieve treatment has no detectable UCM, thus eliminating the time shift and coelutions, and allowing easy identification of all n-alkanes. In this study, eight SAE-10W30 light duty vehicle (LDV) and 4 SAE-15W40 heavy duty vehicle (HDV) lubricating oils were examined. These oils were acquired in the greater Vancouver area since they would be the typical lubricating oils used in the vehicles crossing the Cassiar tunnel. Figures S3 and S4 (SI) show the n-alkane profiles of LDV lubricating oils L3 and L6, respectively, before and after sieve treatment. The n-alkane profiles of the sieve treated and untreated lubricating oil L3 (Figure S3, SI) are similar up to C23. However, for n-alkanes greater than C24, the untreated oil shows larger concentrations than in the sieve treated material. The n-alkanes between C24 and C32 are those mostly found in the UCM. Their concentrations have been overestimated in the nontreated lubricating oil because of coeluting components and the difficulty in integrating peaks superimposed on the UCM hump. In contrast, the n-alkane profiles of the sieve treated and untreated lubricating oil L6 (Figure S4 SI) are very similar. In L6, the concentrations in the range C24-C30 are much larger than those found in L3 (L6: C25 ∼830 ug g-1; L3: C25 ∼ 138 ug g-1). Because the concentration of the n-alkanes is larger in L6, the contribution due to coeluting peaks in the unprocessed lubricating oil is negligible and n-alkane concentrations are not overestimated. In ambient PM, the CPI is used as an indication of the relative contribution of anthropogenic and biogenic sources to the PM. For anthropogenic sources, a CPI near unity is expected and for biogenic sources, a CPI larger than one is expected (18). Therefore, the CPI values of lubricating oil should be very close to unity. Table S7 (SI) shows the calculated CPI values of the lubricating oils before and after sieve treatment in addition to the percent deviation from a value of unity. These results indicate that CPI values deduced from data obtained using the molecular sieve method are closer to unity and distributed over a narrower range (before treatment: CPI 0.88-1.28, standard deviation: 11%; after treatment: CPI 0.90-1.06, standard deviation: 5%). This shows also that the sieve procedure allows unequivocal 3700

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 41, NO. 10, 2007

FIGURE 2. Average normalized n-alkane concentration profiles of all (a) LDV (2) and (b) HDV (9) oils before sieve treatment.

FIGURE 3. Average normalized n-alkane concentration profiles of all (a) LDV (2) and (b) HDV (9) oils after sieve treatment. identification of a fully anthropogenic material, lubricating oil, a petroleum-derived product. Comparison of LDV/HDV Lubricating Oils With and Without Sieve Treatment. Figures 2 and 3 show the average normalized (i.e., pro-rated to the sum of all identified n-alkanes) n-alkane distributions of the eight LDV and four HDV lubricating oils before and after sieve treatment, respectively. The n-alkane profiles of the oils that were processed by the sieve method (Figure 3) clearly show a bimodal distribution in the LDV but not in the HDV profile. The difference arises from higher concentrations of C24-C29 n-alkanes in the LDV. This profile difference would not have been observed without the sieve treatment. The higher apparent concentrations of the n-alkanes larger than C24 found in the untreated LDV and HDV oils (Figure 2) is due to other saturated compounds coeluting with the n-alkanes. Comparison of n-Alkane Profiles from Sieve-Treated Lubricating Oil and PM Samples. Figure S5 (SI) shows the average n-alkane profiles of the 3 h PM sample before and after sieve treatment. Despite the fact that they appear similar, the sieve treated profile show differences that will help comparing the LDV, HDV, and PM profiles. Based on Rogge et al. (19), who reported that essentially all PM n-alkanes emitted from engines come from engine lubricating oil, it is expected that a PM profile would closely resembles that of a LDV and/or HDV lubricating oils. The 3 h profile in Figure 4 has a bimodal n-alkane distribution that closely resembles that of the LDV lubricating oil except for the shift in the maximum carbon number of the first peak, centered at C23 (compared to C21 for the LDV), and the overall higher profile. This is due to the loss of semivolatiles resulting in a lower total n-alkane concentration. Thus, the normalization causes the maxima in the 3 hour PM profile to be higher than those of the LDV. In spite of this, the bimodal n-alkane distribution in the 3 hour PM profile (Figure 4(c)) clearly suggests that the Cassiar tunnel traffic is highly dominated by light duty vehicles, which is in agreement with vehicle type identification through license plate tagging that indicated that traffic composition ranged from 87 to 98% light duty vehicles. Figure 4 also shows that the 3 h profile

carbon number in lubricating oils before and after sieve treatment; Average n-alkane profiles PM samples before and after sieve treatment.

Supporting Information Available This material is available free of charge via the Internet at http://pubs.acs.org.

Literature Cited

FIGURE 4. Average normalized n-alkane profiles of all sieve treated (a) LDV (2) oils, (b) HDV (9) oils, (c) 3 h, (B) and (d) 9 h ([) Cassiar tunnel PM samples.

TABLE 2. CPI Values of the Daily Cassiar Tunnel Samplesa date

CPI before molecular sieve treatment

CPI after molecular sieve treatment

Aug 9 Aug 10 Aug 11 Aug 12 Aug 13 Aug 14 Aug 15

1.15 1.42 1.35 1.31 1.14 1.27 1.24

0.94 1.02 1.05 1.09 1.07 1.06 1.04

a All the CPI values are calculated from 9 h samples except for Aug 13th where the average of the three three samples is reported.

has a maximum located at C23 versus C25 for the 9 h profile. This suggests that the longer sampling period, 9 h, induced a negative artifact, i.e., the loss of the lighter semivolatiles, and yielded a profile that is no longer clearly a bimodal distribution. An adsorbent downstream of the filter could capture loss of semivolatiles during sample collection, but would also capture semivolatiles from the vapor phase. Table 2 lists the CPI values before and after the sieve treatment of the daily PM samples. As expected, the values after the sieve treatment are closer to unity confirming unequivocally that the PM sampled in the tunnel is of anthropogenic origin. These results clearly demonstrate the advantage of using the molecular sieve procedure for PM samples. The molecular sieve procedure helped unravel differences between HDV and LDV lubricating oils which would have been otherwise impossible. This has made it possible to relate the LDV profile to that of the PM. Data from the sieve procedure also yielded n-alkane concentrations that allowed calculation of accurate CPI values for lubricating oils and PM samples. The method may prove helpful in estimating the respective diesel and gasoline contributions to ambient PM.

Acknowledgments This work was funded by Natural Resources Canada through the Program for Energy Research and Development (PERD), by Environment Canada, and by the British Columbia Clean Air Research Fund.

Supporting Information Available Procedures for the silica gel and column chromatography procedures and operating parameters for ASE and GC/MS. Tables include lubricating oil codes, list of surrogates, ASE parameters, recoveries of n-alkanes during method optimization, and CPI values of LDV and HDV oils before and after sieve treatment. Figures include chromatograms (m/z 85) of lubricating oils H1, L1, L3 and L6 processed by the molecular sieve procedure; Concentration of n-alkanes as a function of

(1) Zheng, M.; Cass, G. R.; Schauer, J. J. Edgerton, E. S. Source apportionment of PM2.5 in the southeastern United States using solvent-extractable organic compounds as tracers. Environ. Sci. Technol. 2002, 36 (11), 1-14. (2) Leal-Granadillo, I. A.; Garcia Alonso, I. J.; Sanz-Medel, A. Determination of n-alkanes and polycyclic aromatic hydrocarbons in atmospheric particulate and vapour phases in Oviedo, Spain by GC-MS. J. Environ. Monit. 2000, 218-222. (3) Ellis, L. Fincannon, A. L. Analytical improvements in irm-GC/ MS analyses: Advanced techniques in tube furnace design and sample preparation. Org. Geochem. 1998, 29 (5-7), 1101-1117. (4) Liang, Z. Hsu, C. S. Molecular speciation of saturates by on-line liquid chromatography-field ionization mass spectrometry. Energy Fuels 1998, 12 (3), 637-643. (5) El Sabagh, S. M.; Al Dhafeer, M. M. Occurrence and distribution of n-alkanes and n-fatty acids in saudi arabian crude oils. Pet. Sci. Technol. 2000, 18 (7-8), 743-754. (6) Khan, H. U.; Sharma, R. L.; Nautiyal, S. P.; Agrawal, K. M.; Schmidt, P. Structural characterization of solid n-paraffin components derived from mukta crude oil. Pet. Sci. Technol. 2000, 18 (7-8), 889-899. (7) Netzel, D. A.; Rovani, J. F. Direct separation and quantitative determination of (n-, iso-) alkanes in neat asphalt using urea adduction and high-temperature gas chromatography (HTGC). Energy Fuels 2007, 21 (1), 333-338. (8) Shiping, X. Yongge, S. An improved method for the microseparation of straight chain and branched/cyclic alkanes: Urea inclusion paper layer chromatography. Org. geochem. 2007, 36 (9), 1334-1338. (9) Saito, Y.; Takeda, M. Uchiyama, M. Separation of normal paraffins and evaluation of petroleum contamination in foods. J. Assoc. Off. Anal. Chem. 1979, 62 (6), 1327-1332. (10) Nwadinigwe, C. A.; Nwobodo, I. O. Analysis of n-paraffins in light crudessMolecular sieve and urea adduction techniques revisited. Fuel 1994, 73 (5), 779-782. (11) Selucky, M. L.; Chu, Y.; Ruo, T.; Strausz, O. P. Chemical composition of athabasca bitumen. Fuel 1977, 56, 356-381. (12) Holmes, S. A. Raska, K. A. Isolation and f.i.m.s. analysis of saturated hydrocarbons from tar sand bitumens. Fuel 1986, 65 (11), 1539-1545. (13) Chong, S. L.; Cummins, J. J. Robinson, W. E. Fractionation of soluble extracts obtained from kerogen thermal degradation with CO and H2O. Am. Chem. Soc. Div. Fuel Chem. Prepr. 1976, 21 (6), 265-277. (14) Briker, Y.; Ring, Z.; Iacchelli, A.; McLean, N.; Fairbridge, C.; Malhotra, R.; Coggiola, M. A. Young, S. E. Diesel fuel analysis by GC-FIMS: normal paraffins, isoparaffins, and cycloparaffins. Energy Fuels 2001, 15 (4), 996-1002. (15) Hauser, H. R.; Pattison, J. N. Analysis of aliphatic fraction of air particulate matter. Environ. Sci. Technol. 1972, 6, 549-555. (16) O’Connor, J. G.; Norris, M. S. Molecular sieve adsorption. Anal. Chem. 1960, 32, 701-706. (17) Peters, K. E., Walters, C. C., and Moldowan, J. M. In Biomarkers and Isotopes in the Environment and Human History, 2nd ed.; The Press Syndicate of the University of Cambridge: Cambridge, 2005; Vol. 1, chapter 8. (18) Choiseul, V.; Wilson, J. G. Nixon, E. The distribution of hydrocarbons on the east and south-west Irish coasts and in the Liffey Estuary. Biol. Environ.: Proc. R. Ir. Acad. 1998, 98B (2), 75-86. (19) Rogge, W. F.; Hildemann, L. M.; Mazurek, M. A.; Cass, G. R. Simoneit, B. R. T. Sources of fine organic aerosol. 2. Noncatalyst and catalyst-equipped automobiles and heavy-duty diesel trucks. Environ. Sci. Technol. 1993, 27 (4), 636-651.

Received for review September 19, 2006. Revised manuscript received March 7, 2007. Accepted March 8, 2007. ES062233H VOL. 41, NO. 10, 2007 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

3701