Aerodynamic and Mobility Size Distribution Measurements to Reveal Biodiesel Effects on Diesel Exhaust Aerosol

Abstract

This article examines the effect of biodiesel blends on the exhaust aerosol from a Euro 3 passenger car. Five different feedstock oils (soybean, palm, sunflower, rapeseed, and used frying oil) were used to produce fuels with 10% vol. content in biodiesel (B10). Use of the B10 blends led to a systematic reduction of PM mass emissions in the range of ∼9% (rapeseed) to 23% (used frying oil) on average. The combination of particle size distributions based on the aerodynamic and the mobility diameters led to the estimation of the fractal dimension (DF) for non-volatile particles. This was found to range from 2.52 for the baseline (fossil) fuel to 2.62 for the palm oil blend, suggesting that biodiesel can affect the particle morphology, even at this low blending ratio. The differences were statistically significant. The increase of the DF is translated to more compact particle structure, which in turn denotes lower specific surface area. The volatile fraction of PM lies within a range of 1–9% when fossil diesel fuel is employed. Use of palm, sunflower and rapeseed B10 blends results to PM that contain up to 28% volatile particulate mass. The higher emissions of volatile components together with the lower specific area of non-volatile particles, promotes the formation of volatile particles, especially at high speed conditions. This increases the total particle population under motorway driving by up to three times over the baseline levels.

INTRODUCTION

A number of epidemiologic studies reviewed by the California Air Resources Board (ARB 1998) confirmed the association between ambient airborne particulate matter (PM) and adverse health outcomes, including mortality rates, respiratory related hospital admissions, asthma attacks, and aggravation of chronic diseases. Additionally, animal studies have shown that prolonged exposure of rats to high concentrations of diesel PM (>1 mg/m³) initiated a dose-dependent progression of cellular changes that eventually led toward the development of benign and malicious lung tumours (Valberg and Crouch 1999).

Despite the fact that the association between exhaust particles and health effects is well established, there is still poor understanding of the actual biological interactions of particles with cells. This lack of understanding is, to a large extent, due to the complicated character and the physically active nature of particles. Exhaust aerosol in a vehicle's tailpipe is a complex mixture of internally and externally mixed solid particles (soot) with condensable species (organics and ions). Moreover, particles continue to change in the atmosphere after emission. Coagulation, phase changing (condensation, evaporation), and (photo-)chemical reactions continue to alter their size, morphology and composition. This dynamic character makes difficult to identify which particle dimensions are responsible for inducing health effects.

Stoeger and Reinhard (2006) suggested that the inflammatory effects of PM are better related to their surface area, which in turn means that particle morphology plays an important role to their toxicity characterisation. Cheung et al. (2009) showed that the water soluble organic carbon (WSOC) is highly correlated with the oxidative potential of PM. Therefore, the chemistry of particles is important in understanding toxicity. Finally, deposition mechanisms such as diffusion and gravitational settling, determined by particle size, control the deposition of particles to different lung sections. Since the in-cell dosimetry triggers the finally observed inflammation, the knowledge of the particle size distribution is also of great value. This means that particle mass alone is not a proper metric to correlate with diesel exhaust toxicity. On the contrary, a more detailed characterisation of the exhaust aerosol is required in order to provide representative metrics to assess the toxic potential.

TABLE 1 Properties of the base diesel fuel and the biodiesel 10% vol. (B10) blends. The feedstock used for each B10 blend shown on the first row of the table

Properties	Diesel	Palm blend	Rapeseed blend	Sunflower blend	Used frying oil blend	Soybean blend	EN 590 limits	Test method
Viscosity (40°C, mm^2 s^−1)	3.61	3.81	4.0	3.89	4.21	3.97	2.00–4.50	EN ISO 3104
Density (15°C, g cm^−3)	0.834	0.838	0.841	0.839	0.839	0.838	0.820–0.845	EN ISO 3675
Flash point (°C)	71	67	65	67	64	66	55 min	EN 22719
Sulphur content (μg g^−1)	24	22	19	20	18	21	50 max	EN ISO 20846
Water content (μg g^−1)	52	65	78	62	97	118	200 max	EN ISO 12937
CFPP (°C)	–9	–7	–8	–9	–6	–8	+ 5 max	EN 116
Cetane number	55.5	55.5 ± 0.5 (within the accuracy limits of the measurement)					51 min	EN ISO 5165
Distillation curve								EN ISO 3405
IBP (°C)	179	182	185	183	178	180
10% (°C)	220	225	228	231	219	226
50% (°C)	281	290	298	297	276	289
90% (°C)	344	346	349	348	342	342
FBP (°C)	367	355	360	359	358	361
Heating value (MJ kg^−1)	45.6	44.9	44.6	44.7	44.8	44.8		IP 12

The effect of biodiesel on the toxic potential of diesel exhaust is today of increasing interest (Cheung et al. 2009; Liu et al. 2009; Swanson et al. 2007). Biodiesel is a technical jargon for fatty acid methyl-esters (FAME) produced by the trans-esterification of vegetable oil, which can be used as fossil diesel replacement. Biodiesel is promoted as an environmentally friendly fuel, since its production offers substantial greenhouse gas benefits compared to fossil fuels. According to a review study (Lapuerta et al. 2008), well-to-wheel CO₂ benefits of biodiesel are in the range of 50–80% compared to fossil diesel, depending on the feedstock and the production process. However, concerns on its toxicity arise from the fact that biodiesel has an entirely different chemical composition than fossil diesel, consisting of oxygenated molecules. This may change both the physical and the chemical character of the exhaust aerosol.

This article attempts to shed light on the effect of biodiesel on the physical characteristics of PM. Five different biodiesels in 10% vol. blending ratio (B10) with fossil diesel were used on a Euro 3 vehicle. In a European policy framework, B10 is considered as the maximum foreseeable blending ratio for widespread use in the future. Airborne particle measurements together with PM filter collection have been performed. The particle morphology and the volatile vs. non-volatile ratio were studied, as these two properties have been found relevant in toxicological studies.

METHODS

Test Vehicle, Fuels, and Lubricants

A Euro 3 diesel passenger car (Renault Laguna 1.9 dCi) was employed for the measurements. The vehicle was equipped with a common-rail diesel engine with exhaust gas recirculation (EGR), with a maximum power of 79 kW at 4000 rpm, and a maximum torque of 250 Nm at 2000 rpm. Its exhaust aftertreatment comprised of a closed-coupled pre-catalyst and an under-floor main oxidation catalyst. This vehicle technology is the most widespread currently in the European diesel car stock, as it represents a typical configuration for Euro 3 and Euro 4 diesel cars. The lubricating oil used was of a viscosity and quality grade recommended by the manufacturer. The oil was changed before commencing the measurement series and was conditioned for ∼200 km before starting the actual tests. The oil was not changed during the measurements. This was not deemed necessary as the vehicle was driven for about 2000 km during the tests, i.e., a distance far shorter than the oil change interval recommended by the manufacturer. Due to the relatively short distance, oil dilution due to biodiesel use is not considered to be important.

FIG. 1 Schematic of the setup employed for the measurement of airborne particle properties.

Five different biodiesel blends were tested in total. These were produced by mixing methyl-esters originating from different feedstocks with the same base diesel fuel at a 10% vol. mixing ratio (B10). The base fossil diesel fuel complied with the EN590:2009 specifications. The biodiesels were produced from the most widespread feedstock oils in Europe, i.e., soybean, used frying oils, palm, sunflower, and rapeseed. Tests with the different biodiesels were conducted in that order. The final biodiesel blends also complied with the EN590 specifications and details on their properties are provided in Table 1. More details on the fuels and their impact on gaseous pollutant emissions are provided by Fontaras et al. (2010).

Particle Sampling

Emission measurements were conducted following the European regulations (Directive 70/220/EEC and amendments). The exhaust was led into a dilution tunnel following the regulated constant volume sampling (CVS) technique, where it was primarily diluted and conditioned. The primary dilution ratio in the CVS tunnel ranged from 3:1 to 22:1 depending on the engine exhaust flowrate. A 6 m long corrugated stainless steel line transferred the exhaust from the tailpipe to the CVS dilution tunnel inlet. The line was insulated to minimize heat losses and particle thermophoresis and was clamped onto the vehicle exhaust pipe with a metal-to-metal connection to avoid exposing the hot exhaust gas to any synthetic material connectors. A nominal flowrate of 600 Nm³/h was maintained in the CVS tunnel by a positive displacement pump. The dilution air was filtered through a HEPA class H13/EN1822 filter at the inlet of the dilution tunnel. Particulate matter (PM) mass was collected on 47 mm PTFE-coated fiber filters (Pallflex TX40HI20-WW) which were conditioned for 24 h at a constant temperature (22°C) and humidity (40%) before and after the particulate collection and prior to weighing.

TABLE 2 Details of the aerosol sampling system

Component	Dilution ratio [-]	Residence time [s]	Operating temperature [°C]
CVS	3–22	2.5	Tamb
FPS-4000	12	0.3	Tamb
Ejector diluter	10	0.3	Tamb
Mixing stage	10	0.3	Tamb
Thermodenuder	—	0.3	250
CPC	12000 × CVS		Tamb
SMPS	120 × CVS		Tamb
ELPI	120 × CVS		Tamb + ∼10

Figure 1 presents the particle sampling setup, while Table 2 presents some basic characteristics of the sampling system. More info on the operational characteristics of ejector dilutors are given by Giechaskiel et al. (2009). Aerosol samples were taken from the CVS tunnel through an FPS-4000 dilutor (Dekati Ltd., Finland), operating at a nominal dilution ratio of 12:1 and ambient temperature (∼20°C). The actual FPS dilution ratio was determined daily by means of the real time CO₂ concentration ratio upstream (CVS) and downstream of the FPS. In order to bring the particle concentration within the operating range of the instrumentation employed, the sample was further diluted by a tertiary dilution system. This consisted of two calibrated ejector dilutors connected through a mixing stage, with each of the three dilution steps providing a nominal dilution ratio of 10:1. The mixing stage consists of a HEPA capsule attached to a mixing chamber. The underpressure developed by the ejector diluter at the outlet of the mixing stage forces exhaust sample from the first ejector diluter to enter it. Additionally, air is drawn through the HEPA capsule that dilutes the sample. With this configuration, the total dilution ratio downstream of the CVS was fixed at 12000:1.

A Condensation Particle Counter (CPC, Model 3010, TSI Inc, Shoreview, MN) was employed to record the total particle number concentration during the driving cycles. An Electrical Low Pressure Impactor (ELPI—Dekati Ltd, Finland) provided the aerodynamic size distribution in real time. Oil-soaked sintered plates were used as impactor stages to reduce particle bounce and avoid overloading of the impactor stages (van Gulijk et al. 2003). For the same reason, daily cleaning of the impactor stages was performed. The use of a filter stage extended the lower cut-point to ∼7 nm (Marjamäki et al. 2002). The ELPI was sampling downstream of a thermodenuder (Dekati Ltd., Finland), set at 250°C to remove volatile and semi-volatile particle components. Diffusion and space charge losses in the ELPI (Virtanen et al. 2001) as well as thermophoretic losses inside the thermodenuder (Dekati Ltd., HET 2001; Ntziachristos et al. 2004) were taken into account for the ELPI data reduction. Finally, a scanning mobility particle sizer (SMPS, Model 3936L, TSI Inc., Shoreview, MN) was used during steady-state tests (50, 90, and 120 km/h) to monitor the mobility size distribution. The SMPS operated on a sheath over sample flow rate of 10/1 lpm and a scan time of 90 s. The size distributions were acquired using TSI software version 8.1.0.0, which also corrects for particle losses inside the instrument.

Test Protocol

Particle properties were examined over the European type-approval driving cycle (the New European Driving Cycle- NEDC) but also over the three so-called “Artemis” driving cycles (André 2004), which better represent city (Urban), rural (Road), and highway (Motorway) driving conditions in Europe. The daily measurement protocol included an NEDC, which is a cold-start cycle, followed by the suite of the three Artemis cycles. At the end of each measuring day three steady-state measurements at 50, 90, and 120 km/h were performed. Gaseous pollutants, PM mass and particle number were measured over all four driving cycles (i.e., NEDC and 3 Artemis cycles). One PM filter was collected per driving cycle. Gaseous pollutant measurements were not performed over steady-state tests; only particle number and size distribution measurements were conducted. Each steady-state test consisted of 10 min of conditioning at the respective speed, followed by 15 min of sampling. This sampling duration allowed at least 3 scans to be obtained by the SMPS. The daily measurement protocol was repeated twice per biodiesel blend, while a pair of baseline measurements with fossil diesel was performed twice at the beginning and the end of the measurement campaign.

RESULTS AND DISCUSSION

Effect of Biodiesel on PM Emissions

The grand average PM emission level over all four individual cycles was 47 mg/km when using fossil diesel (baseline), with a coefficient of variation (CV) of 16%. All biodiesel blends tested consistently led to reduced PM emissions. The reductions were equal to 9%, 13%, 15%, 19%, and 23% for the rapeseed, soybean, sunflower, palm, and used frying oil blends respectively. The consistent PM reduction when using biodiesel is also supported by a large number of publications reviewed by Lapuerta et al. (2008). The CV values ranged from 6–10%, except for rapeseed, where CV reached 25%. This was due to a very high emission value over the Artemis Motorway cycle. In fact, while rapeseed emissions were lower than the baseline when seen on average for all driving cycles, these were 23% higher than the baseline over the Motorway cycle. Therefore, the biodiesel blends tested generally led to reduced PM emissions, even at a relatively moderate blending ratio (10%), with one exception over a high speed/high load driving cycle. This binary behavior of biodiesel blend is explained in the following sections by examining the physical characteristics of the aerosol emitted.

Particle Size Distributions

In order to explain biodiesel effects on particle emissions, Figure 2 shows the current-weighted aerodynamic size distributions of non-volatile particles (ELPI) and the number weighted mobility size distributions of the total particle populations (SMPS) for the different fuels examined. The distributions have been normalized with respect to the total current measured and number concentration respectively, in order to highlight the effect of the different fuels on their shape. The ELPI stages beyond Stage 6 (>605 nm) were omitted as they represent less than 5% of the total current measured, while their recordings largely reflect diffusion, and space and image charge signals from ultrafine particles rather than true coarse particle concentrations (Maricq et al. 2006).

FIG. 2 Particle size distributions at 50 km/h. (a) Normalized current-weighted size distributions of non-volatile particles measured by the ELPI. Error bars correspond to min-max of measured range. (b) Number-weighted size distribution of total particles measured by the SMPS.

Figure 2a suggests that some of the blends affect the aerodynamic size distribution of the emitted particles. This is beyond the measurement uncertainty range shown by the error bars, which represent the min–max range per stage. The palm and sunflower blends lead to a local shift of the particle size, from Stage 2 (32.3–55 nm) to Stage 3 (55–106 nm) with variations being much less pronounced for the other size bins. The use of rapeseed derived blend results in a broadening of the size distribution, suggested by the systematically higher current contribution in the lower two (<32 nm) and higher two (>208 nm) stages. On the other hand, the soybean and used frying oil derived blends do not seem to significantly alter the distribution.

Figure 2b depicts the number weighted size distributions according to the mobility diameter of the particles. The inset shows the biodiesel over base fuel ratio for each size bin, with a flat horizontal line (y = 1) corresponding to a distribution identical to the baseline one. The error bars have been omitted in this case to improve the readability of the figure. However, all our remarks refer to observations which are beyond measurement uncertainty. The soybean, used frying oil and palm oil derived blends seem to have a minor effect on the distributions. The normalized values in the inset lie flat (y = 1) for most of the size range and are within the range 0.85–1.15 for the complete size spectrum. On the other hand, the sunflower and the rapeseed blends seem to significantly affect the distribution. In particular, use of sunflower shifts the distribution towards larger mobility diameters. On the contrary, rapeseed monotonically shifts the distribution towards smaller sizes.

Comparison of the aerodynamic and mobility particle size distributions shows that biodiesel use, even at a relatively low blending ratio of 10%, affects the physical characteristics of the particles emitted. The extent of the change depends on the fuel used and is differently demonstrated according to the size expression considered (aerodynamic or mobility). The different effect according to the size expression used suggests that biodiesel affects the morphology of the particles. This is further investigated in the following section.

Particle Morphology

Changes in the morphology of the exhaust particles may be identified either with visualisation methods (microscopy) or with appropriate interpretation of their airborne properties. Both the aerodynamic and the mobility expressions are equivalent sizes of the particle complex morphology. The two expressions may be linked by means of the “effective density” (ρ_eff ), i.e., the density of a sphere which has the same aerodynamic (AD) and mobility (MD) diameters as the real agglomerated particle, as in Equation (1) (Kelly and McMurry 1992):

where ρ₀ is the unit density (1 g/cm³) and C_c(d_a) and C_c(d_b) are the Cunningham correction factors (usually called as slip correction factors) evaluated for the AD (d_a ) and MD (d_b ), respectively.

According to Maricq and Xu (2004), the effective density function with particle size is given by Equation (2),

where DF stands for the fractal dimension of the particles and characterizes the compactness of the soot aggregates. The higher the value of DF, the more compact particle structure is denoted and vice versa. The fractal dimension cannot exceed the value of three, in which case particles are perfect spheres with the effective density equal to the bulk material density.

In order to calculate DF, the approach suggested by Virtanen et al. (2004) was implemented by applying the algorithm developed by Mamakos et al. (2006) on the ELPI and SMPS size distributions at 50 km/h. Following Virtanen et al. (2002) the relationship in Equation (2) is assumed to hold down to a minimum mobility diameter (d_{b, ref} ) below which it becomes constant and equal to ρ_eff,ref . A d_b,ref value of 40 nm and ρ_eff,ref value of 0.85 g/cm³ were employed as they are generally consistent with experimental data from other studies (Mamakos et al. 2006; Maricq et al. 2006; Skillas et al. 1998; Virtanen et al. 2004) and prior work on the specific vehicle by Tzamkiozis et al. (2010). Under these assumptions, the effective density profile is completely specified by the DF, which appears in the power exponent. The fractal dimension is then sought by means of fitting the measured SMPS number-weighted mobility-size distributions to the ELPI responses, through the known ELPI kernel functions (Marjamäki et al. 2005). The fitting procedure was performed for each individual SMPS scan per fuel. By using the same ρ_{eff, ref} and d_b,ref for all fuels, any differences in the particle morphology will be reflected in a single quantity: the DF of the particles, which can be easily interpreted.

It needs to be mentioned that the aforementioned analysis primarily aims at the quantification of the relative effect of the different fuel blends on the structure of the emitted non-volatile particles. The quality of the assumptions for ρ_{eff, ref} and d_b,ref values, as well as small errors in the correction for particle losses in the thermodenuder may affect the absolute level of the DF calculated but will not change the relative effect of the biodiesels, as the error will be the same for all fuels. Furthermore, positioning of the SMPS upstream of the thermodenuder, as opposed to ELPI positioning downstream of it, does not affect the DF estimation. This was proven with a dedicated measurement campaign which showed that the upstream and downstream SMPS distributions at 50 km/h differed only due to the particle losses in the thermodenuder which we correct for. In other words, nucleation and particle growth are limited at 50 km/h.

Following this algorithm, the fitting procedure led to the results presented in Table 3. The table shows the average DF per fuel, together with a min-max range produced by applying the algorithm to the individual SMPS scans. The number of scans (sample size) is also shown in the table. Finally, the table shows the results of a t-test performed to examine the level of confidence at which the average DF per blend is statistically different to the baseline. All blends produce particles that are characterized by a greater DF compared to the baseline fuel at a significance level of 99%, with the only exception being used frying oil, which has the same DF with the diesel fuel. These results suggest that biodiesel can affect the morphology of the particles even at this low mixing ratio, with the increased DF denoting more compact particle morphology than base diesel. This observation is in accordance with the findings of Chung et al. (2008), who suggested that biodiesel particles are compact aggregates without primary particles having such a distinctive appearance as in diesel agglomerates. We may only hypothesize why biodiesel exhaust aerosol has a different structure. The different fuel chemistry (e.g., carbon-to-oxygen bonds and double bonds) may affect the in-cylinder formation by decreasing the length of the carbon chains, which form primary soot particles, while increasing the local flame temperature. Moreover, the difference in physical properties, (e.g., density, viscosity, surface tension) affect the spray characteristics and the combustion profile. Finally, the molecular oxygen content may also have a role to play in post-formation surface oxidation of primary particles. Detailed in-cylinder sampling and visualization techniques will be required to understand the different formation path of particles when using biodiesel.

TABLE 3 Estimated fractal dimension (DF) for the various blends

Fuel	Average DF (± range)	Sample size	t-test (p value)
Baseline	2.52 ± 0.04	11	—
Soybean	2.60 ± 0.01	6	0.00
Used frying oil	2.52 ± 0.02	6	0.71
Palm	2.62 ± 0.01	6	0.00
Sunflower	2.58 ± 0.03	6	0.01
Rapeseed	2.56 ± 0.01	6	0.01

Airborne Particle Number and Mass Emissions

Further insight on the effect of biodiesel blends on exhaust aerosol emissions may be obtained by examining the particle number and mass concentration. The effective density profile is required to obtain an accurate number concentration measurement with the ELPI. In the ELPI particles are charged according to their mobility diameter but they are classified according to their aerodynamic diameter, so the effective density is required to link these two properties and correctly infer the number distributions (Marjamäki et al. 2005). For this correction, one may assume that the effective density profile calculated from size distributions at 50 km/h holds for all driving situations, as Maricq and Xu (2004) demonstrated that DF little changes with engine mode. Hence, substitution of the effective density from Equation (2) into Equation (1) allows to convert aerodynamic into mobility size and to assess the particle number (N_i ) concentration of non-volatile particles measured by the ELPI.

Figure 3 summarizes the particle number emissions over all tests including steady speed cruising. The total particle number (measured by either the SMPS or the CPC) is shown in the vertical axis and the number of non-volatile particles calculated by the corrected ELPI recordings is shown on the horizontal axis. Both magnitudes are expressed as particles per kilometre driven [#/km]. Solid symbols are used to discriminate emissions during high load operating conditions (i.e., Artemis Motorway and 120 km/h) from moderate speed and load driving (open symbols). The emission rate of non-volatile and total particle number emissions matches for moderate load conditions. In fact, non-volatile particle concentration appears somehow higher than total concentration. This is most probably due to the discrete nature of the ELPI signal and the inherent simplification that all particles contributing to the current collected on a given stage are of size equal to the stage midpoint. Despite this overestimation, non-volatile and total particle emission rates are very well correlated (R² = 0.90) in the range from 7 × 10¹³ to 1.5 × 10¹⁴ [#/km] for moderate speed and load driving.

FIG. 3 Particle number emissions over all cycles and steady state tests. Empty symbols correspond to tests with moderate speed/load. Solid symbols represent high speed testing (Artemis Motorway and 120 km/h).

On the other hand, the total number concentration substantially exceeded to the non-volatile one for some biodiesel blends and high speed driving (Artemis Motorway, 120 km/h), with only solid symbols significantly departing from the y  =  x line. This relative increase in total particle number should be attributed to nucleation of volatile particles under these conditions, as the exhaust cools down after emission. The extent of the formation and particle growth will of course depend on the exact sampling and dilution conditions, as has been already established by several studies in the past (Giechaskiel et al. 2005; Mathis et al. 2004).

However, we only examine here the potential of some biodiesel blends to lead to volatile particle nucleation, as opposed to some other blends which do not show this tendency. In particular, soybean and used frying oil derived blends seem not to favor nucleation. In fact one may argue that these blends actually depress nucleation, as they even decrease the tendency of the base fuel to produce nucleation mode particles. On the contrary, palm and rapeseed blends seem to further trigger volatile particle nucleation. Especially regarding rapeseed, total particle number emissions are almost three times higher compared to the other biodiesel blends and two times higher compared to the worst case of the baseline fuel.

It is not straightforward to explain the exact reasons for the trends observed with the various fuels. The different characteristics of the blends compared to the fossil fuel, have certainly a role to play, since they affect critical combustion parameters. One may assume that the differentiated physical properties compared to the base fuel lead to non-optimized combustion and a higher production of unburned hydrocarbons. On the other hand, the presence of oxygen in the fuel molecules may promote oxidation of soot particles and decrease the available surface area. This increases the tendency for homogeneous nucleation of semi-volatile species and the formation of new particles instead of condensation to existing ones. Associating nucleation mode formation with total hydrocarbon emissions may provide some insights on the underlying mechanisms. However, only the heavy fraction of total hydrocarbons may lead to particle nucleation. This heavy fraction is not necessarily proportional to total hydrocarbon emissions therefore one should not expect a one-to-one correlation between total HC and concentration in nucleation mode. Despite this limitation, one can possibly explain some of the observed trends for the biofuels tested. According to the findings of Fontaras et al. (2010), soybean and used frying oil blends lead to lower HC emission levels (on average over hot driving conditions) compared to the baseline ones. This could partially explain their trend to depress the formation of nanoparticles. On the other hand, the rest three blends lead to consistently higher HC emission levels compared to the baseline fuel. The observed nucleation mode triggered by these blends could be the combinational effect of the higher HC emission levels and the more compact particle morphology (higher DF) of these fuels.

The effective density profile and the number concentration per ELPI stage (i = 1…7, with 1 used for the filter stage) may also be used to infer the mass emission rate of non-volatile particles (M_NV ), by means of Equation (3):

Based on this, Figure 4 compares the gravimetrically determined PM mass (vertical axis) with the mass of non-volatile airborne particles emitted over all driving cycles. Steady speed emissions are not included in the figure as no filter samples were collected during these tests. Otherwise, we follow the convention of Figure 3, with solid symbols correspond to the high load tests (Artemis Motorway) and the open symbols corresponding to moderate load ones. The two dashed lines one represent the limits of error for the non-volatile mass concentration, which is expected to be in the order of ±20%, according to Maricq et al. (2006). It is shown that most points systematically lie within the error range, which points towards the fact that emitted PM is mainly carbonaceous. This is mainly the result of the two oxidation catalysts (pre-catalyst and main catalyst) carried by the vehicle which effectively oxidize most of the volatile and non-volatile PM emitted. However, rapeseed, and to a lesser extent, palm oil derived blends exhibit increased volatile (and semi-volatile) PM mass emissions.

FIG. 4 Particle mass emissions over all cycles. Empty symbols correspond to tests with moderate speed/load. Solid symbols correspond to Artemis Motorway.

The ratio of volatile and semi-volatile PM compared to total PM may be expressed by means of the volatile fraction (VF), according to Equation (4).

In Equation (4), PM is the mass measured by the gravimetric filter. The VF values for the different blends/driving cycle combinations are summarized in Table 4. The VF ratio is close to zero in most cases. Some slight negative values were observed (up to –3%) as a result of experimental uncertainty. These have been substituted with zero in Table 2 since they have no physical meaning. The low VF values are expected, given the two-step oxidation aftertreatment of the vehicle which may effectively oxidize most of the volatile PM. Only when using palm, sunflower, and rapeseed blends over the Artemis Motorway cycle, the VF ratio was significantly higher than the baseline. This observation is consistent with the increase in volatile particle number for the same driving cycle and fuel combinations (Figure 3). The increase in volatile vs. non-volatile particle number ratio is proportional to the increase in the VF ratio, maximizing for rapeseed (VF = 0.28 or 28%). In absolute terms, the volatile PM emitted with use of palm and rapeseed blends in the Motorway cycle is 3 and 5 times higher than the baseline, respectively. This will probably have an impact on the toxicity of PM as it has been shown that certain components of the VF, such as the water soluble organic carbon, exhibit much higher toxicity than elemental carbon (Cheung et al. 2009).

TABLE 4 Volatile fraction (VF) of PM emitted with different fuels over the various cycles

	VF (%)
Fuel	NEDC	Urban	Road	Motorway
Baseline	1	9	3	6
Soybean	0	2	5	6
Used frying oil	0	5	0	0
Palm	1	5	0	17
Sunflower	0	1	5	11
Rapeseed	0	5	0	28

The observations of increasing particle number and increasing VF of particulate matter for the particular fuel/driving cycle combinations originate from two independent measurements. The number increase has been calculated by comparing ELPI with CPC number concentrations, while the mass increase has been estimated by comparing ELPI derived mass by PM filter measurement. The ELPI number and mass have been calculated using the effective density profile of non-volatile particles. The consistency in the mass and number observations proves that the determination of fundamental airborne particle properties, such as the fractal dimension and the effective density profile, is a powerful tool in assessing fuel and vehicle effects on particle emissions.

CONCLUSIONS

This study demonstrated that the combination of aerodynamic and mobility particle size measurement is a sensitive method to reveal fuel effects on the morphology of aerosol particles. By testing five different B10 biodiesel blends on a Euro 3 passenger car, it was revealed that four out of the five blends led to fractal dimensions that were higher than the baseline (fossil) diesel. This means that biodiesels, even at this low blending ratio, lead to more dense particles and a lower specific surface concentration than fossil diesel. A blend based on used frying oil as the feedstock was the only one that led to equal fractal dimension to the baseline.

Biodiesel blends were also consistently found to reduce PM emissions. Even at this low blending ratio (10% v/v), the decrease was substantial and ranged between 4–30% on average compared to the baseline levels, over a combination of driving cycles ranging from urban to motorway driving. The reductions were relatively smaller over motorway conditions and in fact a rapeseed based biodiesel blend led to 23% increased emissions over the baseline. Comparison of gravimetric PM measurements with the non-volatile PM mass, determined from combination of effective density and the ELPI size distribution, showed that the small decrease or even overall increase of total PM at high speed is due to volatile PM, which increases with use of most biodiesel blends. This increase is undesirable first because it leads to increased particle number emission rate at high speed driving but also because it may be lead to emission of components which are relatively more toxic than elemental carbon.

Acknowledgments

The authors appreciate the contributions of Athanasios Papaza- charias, Dr. Panayotis Pistikopoulos, and Argirios Tzilvelis for their support in the experimental work.