The Effect of Varying Engine Conditions on Unregulated VOC Diesel Exhaust Emissions

An extensive set of measurements were performed to investigate the effect of different engine conditions ( i.e . load, speed, temperature, ‘driving scenarios’) and emission control devices (with/without diesel oxidative catalyst, DOC) on the 15 composition and abundance of unregulated exhaust gas emissions from a light-duty diesel engine. Exhaust emissions were introduced into an atmospheric chamber and measured using thermal desorption comprehensive two-dimensional gas chromatography coupled to a flame ionisation detector (TD-GC×GC-FID). In total, 16 individual and 8 groups of volatile organic compounds (VOCs) were measured in the exhaust gas, ranging from volatile to intermediate volatility. The total speciated VOC (∑SpVOC) emission rates varied significantly with different engine conditions, ranging from 70 to 9268 20 milligrams of VOC mass per kilogram of fuel burnt (mg kg -1 ). ∑SpVOC emission rates generally decreased with increasing engine load and temperature, and to a lesser degree, engine speed. The exhaust gas composition changed as a result of two main influencing factors, the DOC hydrocarbon (HC) removal efficiency and engine combustion efficiency. Increased DOC HC removal efficiency and engine combustion efficiency resulted in a greater percentage contribution of the C 7 to C 12 branched aliphatics and C 7 to C 12 n- alkanes, respectively, to the ∑SpVOC emission

experiment 14, is the result of a slight difference in the fuel composition between batches A and B. Comprehensive twodimensional gas chromatography coupled to a time-of-flight mass spectrometer was used to further investigate any compositional differences between the fuel batches. The experimental method used for the analysis of the liquid diesel fuel is shown below in section 2.1. An extensive analysis of the liquid diesel fuel was not performed. The aim of this analysis was to investigate whether there were any apparent differences in the fuel composition that would prevent a direct comparison of the 5 emission rates from fuel batches A and B. An extracted ion chromatogram for m/z 57 (dominant aliphatic fragment ion) from fuel batch A and B are shown in Figure S6 A and B, respectively. Both chromatograms have been normalised to the total peak area to allow direct comparison of peak intensity. The highlighted region in Figure S6 displays straight-chain and branched aliphatics with a carbon number range of approximately C7 to C12. The carbon number range was determined using the NIST library. From Figure S6, it can be observed that the peak intensity in the chromatograms from fuel batches A and B are largely 10 comparable, except for the highlighted region, where a slightly lower peak intensity is observed in Figure S6B (fuel batch B).
As a result, the emission rates from experiments where two different fuel batches were used, have not been directly compared.

GC×GC-TOFMS experimental method
Liquid fuel samples were analysed using comprehensive two-dimensional gas chromatography (model 6890N, Agilent Technologies, UK) coupled to a time-of-flight mass spectrometer (Pegasus 4D, Leco, MI, USA) (GC×GC-TOFMS). 15 Compound separation was achieved using a primary 15 meter 5% phenyl polysilphenylene-siloxane (BPX5, SGE, Ringwood, Australia), column with a 0.25 mm film thickness and 0.25 mm internal diameter, and a secondary 2 meter 50% phenyl polysilphenylene-siloxane (BPX50, SGE, Ringwood, Australia) column with a 0.25 mm film thickness and 0.25 mm internal diameter. Samples were introduced into the GC×GC-TOFMS using a Gerstel multipurpose sampler (MPS 2, Gerstel, USA) with dedicated controller (model C506, Gerstel, USA). A 1 µL injection volume was used with a split ratio of 20 100:1. The transfer line was set to 270°C. Cryo-jet modulation cooling was used to achieve comprehensive two-dimensional separation. Helium (CP grade, BOC, UK) was used as the carrier gas with a constant flow rate of 1.5 ml min -1 . The oven starting temperature was set to 65°C with a 0.2 minute hold, followed by a temperature ramp of 4°C min -1 to 240°C, with a further 10 minute hold. The modulator and secondary oven temperature was set to 15°C and 20°C above the oven temperature, respectively. The TOFMS acquisition rate was 50 spectra per second, with a scan range of m/z 35 to 500. The data was analysed 25 using Leco ChromaTOF software version 4.51.6 (Leco, MI, USA). Compounds were identified using the National Institute of Standard and Technology (NIST) standard reference database (version 11).

Calculation of emission factors
The mixing ratios of the individual and grouped VOCs in the exhaust emissions were determined using either a NPL gas standard or the relative response factors (RRF) of liquid standards. The NPL gas standard consisted of 30 VOCs ranging from 30 C2 to C8 with mixing ratios of 3 to 5 ppbv. In total, 11 VOCs in the NPL standard were used for quantification. A list of the speciated VOCs, the calibration method and the compounds used for quantification are shown in the SI, Table S1. The response of an FID is assumed to be proportional to the number of carbon atoms present in a compound and is termed 'effective carbon number' (IOFI, 2011). The effective carbon was used to quantify the VOC groupings, allowing multiple isomers in each group to be calibrated using one compound with the same number of carbon atoms. For example, the mixing ratio of the C7 branched 5 aliphatics was determined using heptane in NPL standard. The mixing ratio of styrene was also determined using the effective carbon number approach. The peak area of styrene was not direct measured but calculated by subtracting the peak area of the aromatic grouping with two carbon substitutions, from the sum of ethylbenzene, m/p-xylene and o-xylene, to give the peak area of the only other remaining compound in this group, styrene. The mixing ratio of styrene was determined using o-xylene.
The mixing ratios of n-nonane to n-tridecane were determined using the RRFs from liquid standards. The RRF is an internal 10 standardisation method commonly used with FIDs to determine an unknown concentration of a compound based on the peak area and concentration of an internal standard or reference compound (e.g. (IOFI, 2011;Tissot et al., 2012)). Liquid standards were prepared consisting of toluene, nonane, decane, undecane, dodecane and tridecane at known concentrations. Toluene was used as the reference compound. The RRF was calculated as shown in Eq. 1 (IOFI, 2011); where A is the peak area of the reference compound (rc) (i.e. toluene) or the analyte (a) (e.g. nonane) and M is the concentration. Once the RRF had been 15 determined, the unknown concentration of the analyte (e.g. nonane) or the VOC grouping (using the effective carbon number approach) in the exhaust emissions were calculated using Eq.2. The mixing ratios of the individual and grouped VOCs were converted from ppbv to mg m -3 , accounting for the molecular weight of the compound or grouping, and the average chamber temperature during the sampling period. The measured VOC mass (mg) was determined by dividing the mixing ratio of the individual and grouped compounds in mg m -3 by the chamber volume (18 m 3 ). Finally, the emission rates were calculated by 20 dividing the measured VOC mass by the amount of fuel burnt (mg kg -1 ) (corrected for exhaust dilution, see Whitehead et al.

Uncertainty in emission rates
A propagation of errors was performed to determine the uncertainty in the measured VOC emissions rates. The propagation of errors included; (i) the standard deviation in the replicate measurements of the calibration standard and the reported uncertainty in the standard VOC mixing ratios, (ii) standard deviation of the replicate measurements of the liquid standards used for the calculation of the RRF (where applicable), and (iii) a 5% standard deviation in the chamber volume. An additional 5 20% error was also included for the emission rates obtained from three-dimensional integration using GC Image software.
This additional error was included to account for the inability of the automated peak integration software to distinguish closely eluting peaks. The software was observed to draw a straight line through two closely eluting peaks, rather than following the peak curvature, effecting the measured volume. The variability in the emission rates between one-and three-dimensional integration was estimated by measuring the emission rate of toluene in the exhaust samples using both integration methods. 10 Toluene was selected due to its importance in the RRF calculation (reference compound) and because it was observed to elute near to an unknown compound in some experiments (i.e. model compound). The average variability in the emission rate of toluene between the two integration methods was determined to be 20.7%. Overall, the uncertainty in the measured emission rates of the individual and grouped VOCs ranged from 6 to 50%, with an average of 22%.       increasing from bottom-to-top). Colour scale represents peak intensity, increasing from blue to red. Chromatograms have been normalised to allow direct comparison of peak intensity. Dashed box highlights an approximate carbon number range of C7 to C12, determined from the library identification of individual compounds.