Methylsiloxanes from Vehicle Emissions Detected in Aerosol Particles

Methylsiloxanes have gained growing attention as emerging pollutants due to their toxicity to organisms. As man-made chemicals with no natural source, most research to date has focused on volatile methylsiloxanes from personal care or household products and industrial processes. Here, we show that methylsiloxanes can be found in primary aerosol particles emitted by vehicles based on aerosol samples collected in two tunnels in São Paulo, Brazil. The aerosol samples were analyzed with thermal desorption-proton transfer reaction-mass spectrometry (TD-PTR-MS), and methylsiloxanes were identified and quantified in the mass spectra based on the natural abundance of silicon isotopes. Various methylsiloxanes and derivatives were found in aerosol particles from both tunnels. The concentrations of methylsiloxanes and derivatives ranged 37.7–377 ng m–3, and the relative fractions in organic aerosols were 0.78–1.9%. The concentrations of methylsiloxanes exhibited a significant correlation with both unburned lubricating oils and organic aerosol mass. The emission factors of methylsiloxanes averaged 1.16 ± 0.59 mg kg–1 of burned fuel for light-duty vehicles and 1.53 ± 0.37 mg kg–1 for heavy-duty vehicles. Global annual emissions of methylsiloxanes in vehicle-emitted aerosols were estimated to range from 0.0035 to 0.0060 Tg, underscoring the significant yet largely unknown potential for health and climate impacts.


S1. Identification and quantification of methylsiloxanes
The identification and quantification of methylsiloxanes and derivatives were conducted based on a method described in our previous study. 1The methylsiloxane peaks identified in the tunnel samples are shown in Table S1a, including the main peak (also called base peak), main+1, and main+2 isotope peaks.
Most of the peaks were consistent with the theoretical m/z value, validating the presence of various methylsiloxanes.Some peaks deviated slightly from the theoretical m/z value, which might be due to peak identification and integration bias, interference of other compounds with similar m/z values, and instrument calibration.
The theoretical (M+1)/M and (M+2)/M ratios between the main peak and isotope peaks are shown in Table S2a.The identified methylsiloxanes were mainly small cVMS (D3-D15) and their positively charged fragments (D3f-D15f), produced during ionization by loss of one -CH3.These positively charged fragments were added to the mass of the corresponding cVMS for easy understanding.As described in our previous study, (M+1)/M ratios of CHON compounds with similar m/z values are much lower than methylsiloxanes. 1 For example, D5 has m/z = 371.102and (M+1)/M = 0.3641, and other possible compounds are C11H18O12N2H + (371.094),C16H18O10H + (371.098),C15H18O9N2H + (371.109), and C13H22O12H + (371.119) with (M+1)/M ratios ranging from 0.1329 to 0.1789.Based on the uncertainty of the PTR-MS quantification, we consider (M+1)/M ratios that fall within the range of 70% to 130% of the theoretical ratio as indicative methylsiloxanes, since this range does not overlap with that of possible CHON compounds.For verification, the (M+1)/M ratios of D3-D15 and D3f-D15f were checked in the raw data before blank subtraction, and on average 49% and 48% of the (M+1)/M ratios fell within the valid range for Tunnel 1 and Tunnel 2, respectively.Since about half of the large number of ratios met the requirements (26 × 33 × 3 × 6 ratios in total; 26 methylsiloxanes; 20 samples in Tunnel 1 and 13 samples in Tunnel 2; 3 replicates; 6 temperature steps), the presence of methylsiloxanes was confirmed.For D1 and DMSD with only one Si atom in the molecule, the (M+1)/M ratios are not significantly higher than for other CHON structures, so the verification method above does not work effectively.We identified the existence of D1 and DMSD for three reasons: first, the detected m/z values were in good agreement with the theoretical values; second, they are reasonable fragments and products of methylsiloxanes formed in PTR-MS; third, substantial amounts of D1 and DMSD were also found in our previous study on ship emissions. 1e detected median (M+1)/M and (M+2)/M ratios of D1-D15 are shown in Table S2a.The median (M+1)/M ratios of D3-D15 were mostly close the theoretical ratios, but (M+2)/M ratios were sometimes higher than the theoretical ratios.Higher (M+2)/M ratios might be caused by the oxidation products of methylsiloxanes in the PTR-MS, for example, the main+2 peak of D3 has a m/z value of 225.061-225.071,and associated oxidation product D2T OH (C5H16O4Si) has a m/z value of 225.043.These two peaks are sometimes not separated by PTR-MS resolution, which ultimately contribute to a higher detected main+2 peak.In addition, the (M+1)/M and (M+2)/M ratios of minor peaks in some samples (close to the detection limit) can be influenced by background, peak integration, and other compounds with close m/z values.To minimize these interferences, concentration corrections were conducted for each set of methylsiloxane peaks.The main idea of the concentration correction is to include the main and isotope peaks without interferences.The main peak, main+1 isotope peak, and (M+1)/M ratio were used for correction in this study.Specifically, if the (M+1)/M ratio is higher than the theoretical ratio, the main peak is considered less interfered, and interfering compounds are more likely to be present on the main+1 isotope peak.
Otherwise, the main+1 isotope peak is considered less interfered.Then the concentration can be calculated by dividing the peak with less interference by the theoretical fraction of this particular peak in the total amount of the substance (including the main peak and all isotope peaks).The theoretical fractions of the main peaks and isotope peaks in the total amount of the substance are also shown in Table S2a.The quantification of methylsiloxanes in mixtures is inherently difficult, the interference can only be minimized, not eliminated.For an easier understanding, the fragments D3f-D15f were combined with relevant main compounds (D3-D15) in the calculation and presentation of the results.

S4
Hydroxylated methylsiloxanes, where a methyl group (-CH3) is replaced by a hydroxyl group (-OH), have been identified in ship emissions. 1However, in the mass spectra of vehicle samples, the related peaks were either not present or maybe insufficiently resolved from the isotope peaks of cVMS in Table S1a-S2a.For instance, the detected peak at m/z = 225.053was likely a composite signal resulting from the overlap of the main+2 peak of D3 (m/z = 225.061-225.071)and the main peak of D2T OH (m/z = 225.043).
The hydroxylated peaks were likely not fully resolved due to relatively low concentrations, but can tentatively be hypothesized due to the shift of the observed main+2 peak of e.g.D3 to slightly smaller masses (see Table S1b) than expected based on theory.In addition, the (M+2)/M ratios (see Table S2a) were higher than theoretically predicted, indicating an additional compound was present at the M+2 mass.This indirect evidence was consistent with the presence of lower molecular weight hydroxylated methylsiloxanes (D2T OH -D5T OH ) and corresponding fragments (D2T OH f-D5T OH f) as shown in Table S1b-S2b.For cVMS of higher molecular weight, the (M+2)/M values in Table 2a, were much closer to their theoretical (M+2)/M ratios.Thus, hydroxylated compounds were therefore likely not present at methylsiloxanes larger than D5 or in relatively small quantities.PDMS standard samples analyzed by PTR-MS exhibited hydroxylated fragments as well, 1 so the origin of these hydroxylated methylsiloxanes cannot be unequivocally attributed to oxidation in the vehicle engine or the TD-PTR-MS ionization process.
The hydroxylated methylsiloxanes were not well resolved and only accounted for a small proportion in the vehicle samples, but excluding these compounds may underestimate the vehicle emissions.Therefore, these hydroxylated methylsiloxanes were included in the overall emission calculation, analogous to D1 and DMSD.Considering the peak overlapping, the determination of concentration becomes more difficult.The detailed procedure involved initially computing the concentrations of main+2 and main+3 of D3-D6 and D3f-D6f methylsiloxanes according to their theoretical ratios, as presented in Table 1a-2a.
Then these calculated concentrations were subtracted from the measured peaks at the relevant masses, yielding the concentrations of the main and main+1 peaks of D2T OH -D5T OH and D2T OH f-D5T OH f.

S5
Thereafter, the (M+1)/M ratio of D2T OH -D5T OH and D2T OH f-D5T OH f can be utilized to estimate their concentrations as described above.The adoption of a conservative calculation approach guarantees that the calculation is skewed towards underestimation and can effectively mitigate the potential bias emanating from these not well-resolved hydroxylated methylsiloxanes.The deviation of the detected m/z value from the theoretical m/z value is shown Figure S1.The instrument was well calibrated in the m/z < ~800 range and this was considered a valid range.On the other hand, the detected peaks increasingly deviated from the theoretical m/z values in the m/z > 800 range, but we can still identify these peaks as methylsiloxanes based on the isotope peaks and (M+1)/M ratios.According to the PDMS depolymerization described in Figure 2f, the concentrations of cVMS products are in descending order of molecular size, which means that the peaks of large molecular products may become so low that they cannot be distinguished from blanks.Therefore, we set a cutoff point for conservative calculations of concentrations and emission factors, when methylsiloxane peaks were not significantly different from blanks, even though some methylsiloxane peaks were still detected by the PTR-MS.As a result, methylsiloxanes of D3-D10 in tunnel samples were statistically different from the blank (Welch ttest, p = 0.0033), but methylsiloxanes of D11-D15 were not (Welch t-test, p = 0.3070).In the following calculations, only D1-D10 were included for concentrations and emission factors of methylsiloxanes in vehicle emissions.Furthermore, there might still be some large molecular methylsiloxanes that cannot be thermally desorbed (up to 350 °C) by our TD-PTR-MS, so this study was originally a conservative estimation due to the temperature limitation.A considerable amount of D2 (fragments of two Si(CH3)2O units) was found in ship emissions, 1 but no D2 was identified in mass spectra of the tunnel samples.This indicates relatively less fragmentation, which might be related to the concentrations of methylsiloxanes.
Table S3.D1-D10 methylsiloxanes and C23-C38 hydrocarbons from lubricating oils detected in the aerosol samples collected in Tunnel 1. Percentage refers to the fraction of pollutants in organic aerosols (OA) detected by the PTR-MS.there should be some short-chain fragments and oxidized products due to engine combustion, but this part overlapped with the mass spectra of fuel oils, and is therefore not included in the calculation.

S2. Molecular size analysis of methylsiloxanes in particulate vehicle emissions
The correlation between viscosity and the degree of polymerization is important for the estimation of molecular size in this study.Therefore, the Table 1 in Mojsiewicz-Pieńkowska et al. 2 was adapted here for convenience, as shown in Table S5.
Table S5.Dependence of molecular weight and viscosity PDMS on the degree of polymerization.
(adapted from Mojsiewicz-Pieńkowska et al.The molecular sizes of some fragments and original PDMS can be much larger than the detection range of various ordinary mass spectrometry (e.g., GC-MS).Therefore, taking the advantage of thermal desorption and depolymerization, TD-PTR-MS shows certain advantages in the detection of methylsiloxanes with large molecular sizes.

S3
. The relationships between methylsiloxanes, hydrocarbons from lubricating oils, and organic aerosols.
The relationships between D1-D10 methylsiloxanes, C23-C38 hydrocarbons, and organic aerosols (OA) are shown in Figure S9.The positive intercept in Figure S9a (and Figure 3e) indicated other potential sources of methylsiloxanes, which may be related to non-combustion vehicle emissions.The desorbed organic compounds CH, CHO, CHON, CHN, methylsiloxanes, and others accounted for 32%, 51%, 10%, 2.0%, 1.1%, and 4.5% of the organic aerosol (OA) on average, respectively.Since C23-C38 hydrocarbons from lubricating oils accounted for 18% of OA on average, the unburned hydrocarbons were mainly large molecules from lubricating oils rather than fuel oils.The concentration relationships between methylsiloxanes and CH, CHO, CHON, CHN compounds are shown Figure S10.The R 2 values S25 were very close, between 0.91-0.93.This suggested that the methylsiloxanes came from vehicle emissions, but did not point to a specific composition or combustion process.
Furthermore, when investigating the concentration relationship between methylsiloxanes and each desorbed molecule, 565 molecules out of 1116 had R 2 values higher than 0.8, 114 molecules had R 2 values higher than 0.9, and only 7 molecules had an R 2 value higher than 0.93.This again indicated that methylsiloxanes emissions did not originate from a specific source or process, but rather from fuel oils and lubricating oils.
to the carbon intensity of the fuel (g CO2 L −1 ).The factor 44/12 refers to the unit conversion from g C to g CO2.

S4.2 Light-duty vehicles (LDV)
The vehicle type in Tunnel 1 is dominated by light-duty vehicles (gasohol).Therefore, the emission factors of light-duty vehicles can be estimated based on the samples from Tunnel 1, referred as EFP_LDV and EF*P_LDV.The parameters of Tunnel 1 and property parameters of gasohol can be applied into equations ( 1) and ( 2), shown as equations ( S3) and (S4).

S4.3 Heavy-duty vehicles (HDV)
Previous tunnel studies have shown that emissions from light-duty and heavy-duty vehicles have similar CO emission rates per kilometer. 3,6The CO emissions from heavy-duty vehicles (diesel) can therefore be estimated according to the equation (S5).The CO2 emissions from heavy-duty vehicles can be estimated according to the equation (S6).vehicle −1 ).ρ refers to the fuel density (g L −1 ).wc is the carbon mass fraction of the fuel (g C g −1 ).U × ρ can be converted into another fuel consumption rate R in unit (g km −1 vehicle −1 ) by equation (S7).Instead of U × ρ, the average fuel consumption rate R (g km −1 vehicle −1 ) was used in calculation in this study, with RD and RG in Table S9.
As the emissions from Tunnel 1 were dominated by light-duty vehicles (gasohol), the ratio between ∆  Sample Tunnel1_06 and Tunnel1_17 are outliers, which was caused by abnormal ∆[CO2], so these two samples were not included for the calculation of the average of emission factors and further calculation of heavy-duty vehicles in Tunnel 2. Sample Tunnel1_13 is also an outlier of especially high methylsiloxanes emissions, so this sample was not included for the emission factor calculation of methylsiloxanes.

S5 Engine lubrication of vehicles and ships
In the engine cylinder lubrication, the lubricating oil is mainly used to reduce heat and friction between the pistons and the liners, to wash away the generated soot and combustion residues, and to reduce corrosive wear by neutralizing the acid of the combustion products.
In vehicle engines and marine engines for small vessels, cylinder lubrication is a part of the engine lubrication system.The lubricating oil is sprayed onto the piston rings from below, and due to the protective isolation of the piston rings and oil wiper rings, the lubricating oil cannot splash into the combustion chamber.A small amount of lubricating oil creates an oil film between the piston and cylinder wall.
In marine engines for large vessels, the cylinder lubrication system is usually independent from the main lubricating oil system.Lubricating oil is fed into the engine cylinder directly through oil holes, which are in the cylinder wall in conjunction with the vertical movement of the piston.An adequate amount of oil film will be maintained between the piston and cylinder wall for boundary lubrication.Waste lubricating oil is discharged from a drain hole at the bottom of the cylinder.
The differences in lubrication method and engine size leads to a higher degree of lubricating oil combustion in the engine of large marine vessels.In addition, the high speed of the pistons in a car engine means shorter combustion time, while the piston speed of a marine engine is usually much slower, which means sufficient time for oxidation.

Figure S1 .
Figure S1.The deviation of the detected m/z value from the theoretical m/z value.

Figure S2 .
Figure S2.Fractions of D1-D10 in methylsiloxanes from vehicle emissions in two tunnels in Brazil.(a)

Figure S3 .
Figure S3.Full mass spectra of the aerosol samples collected in Tunnel 1, (a) morning, (b) evening, and

Figure S4 .
Figure S4.Full mass spectra of the aerosol samples collected in Tunnel 2, (a) morning and (b) afternoon

∆
[CO] .=  .× ∆[CO] (S5) ∆[CO]D refers to CO concentrations of emissions from heavy-duty vehicles in unit (µg C m −3 ).fD is the number fraction of heavy-duty vehicles, fD = ND / (NG + ND).ND and NG refer to the numbers of heavyduty and light-duty vehicles passing the tunnel at the time t, respectively.

(
∆[P]G/∆[CO]G)Tunnel_1 can be used as the ratio for light-duty vehicles ∆[P]G/∆[CO]G.The emissions of heavy-duty vehicles in Tunnel 2 can be estimated by equation (S8).
S8)Where ∆[P]D is the concentration of a pollutant emitted from heavy-duty vehicles (ng m −3 ).(1− fD) refers to the fraction of ∆[CO] from light-duty vehicles.(∆[P]G/∆[CO]G)Tunnel_1 is the estimated ratio of pollutant over CO concentrations of light-duty vehicles based on results of Tunnel 1.With ∆[CO]D, ∆[CO2]D, and ∆[P]D obtained from equations (S3, S4, S6), the emission factors of heavyduty vehicles can be calculated based on equations (1) and (2), referred as EFP_HDV and EF*P_HDV.In calculations, the parameters of Tunnel 2 and property parameters of diesel should be used in equations (1) and (2), shown as equations (S9) and (S10).

Table S1a .
The theoretical and detected m/z values of the main peak and main+1, main+2 isotope peaks of methylsiloxanes in aerosol samples.

Table S1b .
The theoretical m/z values of the main peak and main+1, main+2 isotope peaks of hydroxylated methylsiloxanes in aerosol samples.

Table S2a .
The theoretical fractions of main peaks and isotope peaks in the total amount of the substance, the theoretical (M+1)/M and (M+2)/M ratios between main peak and isotope peaks, and the detected median (M+1)/M and (M+2)/M ratios in two tunnels of methylsiloxanes in aerosol samples (D1-D3 in the 150 °C fraction and D4f-D15 in the 100 °C fraction).

Table S2b .
The theoretical fractions of main peaks and isotope peaks in the total amount of the substance, the theoretical (M+1)/M and (M+2)/M ratios between main peak and isotope peaks, and the calculated median (M+1)/M ratios in two tunnels of hydroxylated methylsiloxanes in aerosol samples.Calculated median ratios refer to the median ratios estimated after subtracting the contribution of the main+2 isotope peaks of the corresponding cVMS.

Table S4 .
D1-D10 methylsiloxanes and C23-C38 hydrocarbons from lubricating oil detected in the aerosol samples collected in Tunnel 2. Percentage refers to the fraction of pollutants in OA detected by the PTR-

Table S6 .
Thermal desorption of commercially available PDMS standards at elevated temperature steps first in He and then in O2, shown as percentages of total carbon.

Table S7 .
Averaged concentrations of D1-D10 methylsiloxanes in the aerosol samples collected in the tunnels, desorbed at elevated temperature steps.

Table S11 .
Emission factors of light-duty vehicles of methylsiloxanes and hydrocarbons from lubricating oils.