Experimental investigation of the effect of inlet particle properties on the capture efficiency in an exhaust particulate filter

Journal of Aerosol Science 113 (2017) 250–264

Contents lists available at ScienceDirect

Journal of Aerosol Science journal homepage: www.elsevier.com/locate/jaerosci

Experimental investigation of the effect of inlet particle properties on the capture efficiency in an exhaust particulate filter

MARK

⁎

Sandeep Viswanathana, , David Rothamera, Alla Zelenyukb, Mark Stewartb, David Bellb a b

Engine Research Centre, University of Wisconsin-Madison, United States Pacific Northwest National Laboratory, United States

AR TI CLE I NF O

AB S T R A CT

Keywords: Particulate matter (PM) Soot agglomerates Particle shape factor Gasoline particulate filter (GPF) Integrated particle size distribution (IPSD) method Spark ignition direct injection (SIDI) engine Gasoline direct injection (GDI) engine

The impact of inlet particle properties on the filtration performance of clean and particulate matter (PM) laden cordierite filter samples was evaluated using PM generated by a spark-ignition direct-injection (SIDI) engine fuelled with tier II EEE certification gasoline. Prior to the filtration experiments, an advanced aerosol characterization system that comprised of a scanning mobility particle spectrometer (SMPS), centrifugal particle mass analyzer (CPMA), a differential mobility analyzer (DMA), and a single particle mass spectrometer (SPLAT II) was used to obtain a wide range of information on the SIDI PM emissions including particle size distribution (PSD), composition, mass, and dynamic shape factors (DSFs) in the transition ( χt ) and free-molecular ( χv ) flow regimes. During the filtration experiments, real-time measurements of PSDs upstream and downstream of the filter sample were used to estimate the filtration performance and the total trapped mass within the filter using an integrated particle size distribution method. The filter loading process was paused multiple times to evaluate the filtration performance in the partially loaded state. The change in vacuum aerodynamic diameter (dva ) distribution of mass-selected particles was examined for flow through the filter to identify whether preferential capture of particles of certain shapes occurred in the filter. The filter was also probed using different inlet PSDs. Pausing the filter loading process and subsequently performing the filter probing experiments did not impact the overall evolution of filtration performance. Within the present distribution of particle sizes, filter efficiency was independent of particle shape potentially due to the diffusion-dominant filtration process. Particle mobility diameter and trapped mass within the filter appeared to be the dominant parameters that impacted filter performance.

1. Introduction The stratified nature of combustion in modern spark-ignition direct-injection (SIDI) engines, combined with wetting of piston and wall surfaces, results in a significant increase in emission of nanoparticles from SIDI engines compared to traditional port fuel-

Abbreviations: APM, Aerosol particle mass analyzer; CAD, Crank angle degrees; CPC, Condensation particle counter; CPMA, Centrifugal particle mass analyzer; CT, Computerized tomography; DMA, Differential mobility analyzer; DSF, Dynamic shape factor; EEPS, Engine exhaust particle spectrometer; EFA, Exhaust filtration analysis system; EOI, End of injection timing; EPA, Environmental protection agency; FE, Filtration efficiency (%); FV, Filtration velocity (cm/s); FWHM, Full-width at half-maximum; GCI, Gasoline compression ignition; HL, Heavy load engine operation; IPSD, Integrated particle size distribution method; MBT, Maximum brake torque; MPPS, Most penetrating particle size (nm); PFD, Partial flow diluter; PM, Particulate matter; PSD, Particle size distribution; SIDI, Spark-ignited direct-injection; SMPS, Scanning mobility particle spectrometer; SPLAT, Single-particle laser-ablation time-of-flight mass spectrometer ⁎ Corresponding author. E-mail address: [email protected] (S. Viswanathan). http://dx.doi.org/10.1016/j.jaerosci.2017.08.002 Received 20 February 2017; Received in revised form 9 July 2017; Accepted 8 August 2017 Available online 11 August 2017 0021-8502/ © 2017 Elsevier Ltd. All rights reserved.

Journal of Aerosol Science 113 (2017) 250–264

S. Viswanathan et al.

Nomenclature

Df Dfm dR da dm d va d ve mp ndm

vs ρ0 ρeff ρp χt χv

Fractal dimension Fractal dimension from mass-mobility data Interception equivalent diameter (nm) Aerodynamic diameter (nm) Mobility diameter (nm) Vacuum aerodynamic diameter (nm) Volume equivalent diameter (nm) Particle mass (fg) Size-resolved number concentration (1/cm3)

Δp M ′′′ k

Superficial velocity (cm/s) Unit material density (1000kg /m3 ) Effective density (g/cm3) Particle density (g/cm3) Dynamic shape factor in transition flow regime Dynamic shape factor in free-molecular flow regime Pressure drop (kPa) IPSD mass concentration (g/cm3) Permeability (m2)

injected gasoline engines (Chan et al., 2012; Hall & Dickens, 1999; Stojkovic, Fansler, Drake, & Sick, 2005; Zhao, Lai, & Harrington, 1999). The harmful nature of these nanoparticle emissions (Ibald-Mulli, Wichmann, Kreyling, & Peters, 2002; Peters, Wichmann, Tuch, Heinrich, & Heyder, 1997; Valavanidis, Fiotakis, & Vlachogianni, 2008) has led to number-based emissions regulations for particulate matter (PM) from SIDI engines in the European Union since 2014, with more stringent regulations going into effect in 2017. Particle number (PN)-based regulations for gasoline vehicles have been proposed in other parts of the world following Europe’s lead. India will be implementing Euro 6 PN requirements for gasoline vehicles in 2020, while the fuel-neutral China 6 PN-based standards will go into effect in 2019. Gasoline particulate filters (GPF) have been demonstrated to be effective means of reducing tailpipe PM mass and number emissions (Johnson, 2013; Mamakos, Martini, Marotta, & Manfredi, 2013), but large gaps in the fundamental understanding of filtration in GPFs exists. Typically, exhaust particulate filters for gasoline and diesel engine applications are exposed to PM with sizes ranging from a few nanometers to micron-sized particles, depending on the engine operating conditions. The majority of these particles are fractal soot agglomerates comprised of primary spherules (Park, Cao, Kittelson, & McMurry, 2002). For the range of particle sizes trapped by automotive exhaust filters, diffusion and interception have been shown to be the dominant capture mechanisms (Konstandopoulos & Johnson, 1989; Viswanathan et al., 2015). The net capture efficiency by these mechanisms is commonly associated with dimensionless parameters that are dependent on both the particle morphology and filter structure (Hinds, 1999). The length scales associated with filtration can span multiple orders of magnitude based on the clean filter properties and the evolution of filter structure due to trapped PM. The current study focuses on understanding the impact of inlet particle properties on capture efficiency within an exhaust particulate filter under different levels of PM loading in the filter wall. To the best of the authors’ knowledge, this is the first study attempting to isolate the impact of the combustion generated particle morphology (shape and size) on filtration capture efficiency in clean and partially loaded ceramic particulate filters. 2. Background Particle morphology plays an important role in determining a particle’s behavior in a filter medium. The complex nature of agglomerated PM requires multiple parameters to effectively define their morphology and behavior. For instance, different equivalent diameters, particle electrical mobility (dm ), aerodynamic (da ), and vacuum aerodynamic (d va ) diameters, are used to describe particle behavior in different flow regimes under the influence of various forces (DeCarlo, Slowik, Worsnop, Davidovits, & Jimenez, 2004; Zelenyuk, Cai, & Imre, 2006). To account for the fact that the behavior of an aspherical particle depends on its shape, a correction factor known as a dynamic shape factor (DSF or χ) is used. The DSF is defined as the ratio of the drag force on an aspherical particle to the drag on a sphere of the same volume-equivalent diameter (d ve ) moving with the same velocity. The volume equivalent diameter is the diameter that the particle would have if it were melted to form a droplet while preserving any internal void spaces (DeCarlo et al., 2004). Note that DSF is not an intrinsic particle property but a measure of its behavior. The DSF depends on gas pressure and particle size, hence on the flow regime, and particle orientation. Here, we use DSFs in the transition (χt) and free molecular (χv) flow regimes to characterize particle shapes. Zelenyuk et al. (2014) utilized different sequential combinations of an aerosol particle mass analyzer (APM) (or centrifugal particle mass analyzer (CPMA)), a differential mobility analyzer (DMA), and a single particle mass spectrometer (SPLAT II) to extensively characterize PM emissions from a gasoline compression ignition (GCI) engine operated under different load conditions. Direct simultaneous measurements of mp , dm , d va , and composition of GCI exhaust particles were subsequently used to calculate several other particle properties, including particle density ( ρp ), mass-mobility power-law exponent (Dfm ), dva-dm fractal dimension, χv , χt , average primary spherule diameter, number of primary spherules for a given particle mass, and void fraction (Shapiro et al., 2012; Zelenyuk et al., 2014). Moreover, Beranek, Imre, and Zelenyuk (2012) demonstrated that by combining all three instruments (APM/DMA/SPLAT system), it is possible to select and characterize particles with one charge and a narrow distribution of mass and mobility, and thus, shapes. The impact of multiply charged particles was also discussed in significant detail and was demonstrated to be insignificant while using the combined system. The information obtained using these advanced characterization techniques, enables the fundamental filtration experiments discussed in this paper focused on understanding the relative importance of morphological parameters on the capture efficiency of agglomerated particles within a filter. Despite the fractal nature of combustion generated PM, many filtration models operate under the assumption of spherical particle morphology to evaluate the impact of different capture mechanisms. This can lead to inherent errors in estimation of capture 251

Journal of Aerosol Science 113 (2017) 250–264

S. Viswanathan et al.

efficiency especially if the capture mechanisms are sensitive to the larger surface area to volume ratio of the agglomerates and the possibility of flow-dependent particle orientation. Experiments studying the filtration efficiency (FE) of fibrous filters using agglomerated particles (Boskovic, Altman, Braddock, & Agranovski, 2009; Kim, Wang, Emery, Shin, Mulholland & Pui, 2009; Lange, Fissan, & Schmidt-Ott, 2000) indicated that, for agglomerates of the same electrical mobility diameter, the FE increased with increasing projected diameter/area of the agglomerate. Experiments by Boskovic et al. (2009) showed that the impact of shape on FE could be as high as 20% points for MgO particles in the size range of 50–200 nm at a filtration velocity (FV) of 20 cm/s. The discrepancy between experimental and theoretical FE increased to 35% points when the FV was reduced to 10 cm/s. The predicted FE was observed to show better agreement with the experimental results when corrections were applied for the additional total and projected particle surface areas due to non-spherical particle shape. Variation in the agglomerate orientation between different experimental runs was believed to be responsible for the measured uncertainty in the FE results (∼ ± 10% points). However the orientation effect is expected to be low (< 3%) for lower filtration velocities (Cheng, Xie, Fu, & Shaw, 1991). Lange et al. (2000) used an interception equivalent diameter (dR ) to account for the additional capture efficiency observed for cluster-cluster grown, 40–900 nm carbon agglomerates in wire-screen filters. The dR was determined by comparing the penetration of the carbon agglomerates with spherical wax particles of identical mobility diameters selected using a DMA. For particles smaller than 300 nm the dR was estimated to be nearly twice the measured dm . However, diffusion was reported to be the dominant capture mechanism within a clean filter for particles smaller than 300 nm at FV < 5 cm/s. The difference in penetration between the agglomerated and spherical particles was therefore relatively small (< 5% points). Kim et al. (2009) studied the impact of agglomerate morphology for 50–300 nm silver nanoparticles on their penetration through a glass fiber filter at a FV of ∼5 cm/s. The nanoparticles generated using a furnace, were subsequently allowed to form agglomerates (Dfm = 2.07, χt ∼ 3.5) in a separate chamber. By sintering the agglomerates at different temperatures, the authors simultaneously (but not independently) varied both the Dfm and the DSF at any mobility dimeter to understand its impact on penetration. The penetration for 200-nm particles sintered at 600 °C (Dfm = 2.95, χt ∼ 1) was a factor of 1.7 lower compared to the un-sintered 200nm particles. The authors showed that model calculations using the maximum projected length of the particles, instead of the electrical mobility diameter, gave better prediction of collection efficiency. It is important to note that sintering of the particles changes the initial fractal agglomerates into approximately spherical particles which results in drastically different effective densities and hence aerodynamic diameters between the sintered and pre-sintered particles. The large change in aerodynamic diameter can significantly impact capture efficiency as was observed in the study. In addition to particle morphology, the filter structure also plays an important role in particle capture by both the diffusion and interception mechanisms. The length scales within a full-scale particulate filter can span several orders of magnitude ranging from a few millimeters for flow through a channel, to a few microns inside the pores of a clean filter, down to several nanometers when a soot cake is formed (Konstandopoulos, Kostoglou, Vlachos, & Kladopoulou, 2007). In order to isolate the impact of just the filter walls, an exhaust filtration analysis system (EFA) was developed at the University of Wisconsin-Madison to perform small-scale filtration experiments on pieces of the filter wall referred to as wafers (Viswanathan, Sakai, & Rothamer, 2014; Wirojsakunchai, Kolodziej, Yapaulo, & Foster, 2008). The EFA has been previously used to understand the impacts of filter loading conditions and trapped ash on the evolution of filtration performance of cordierite wafers with properties representative of diesel particulate filters (DPFs) (Rakovec, Viswanathan, & Foster, 2011; Viswanathan, Rakovec, & Foster, 2012; Viswanathan et al., 2014). Size-resolved filtration efficiency measurements showed the presence of a most penetrating particle size (MPPS) which was observed to decrease from ~200 nm for low mass loadings (< 0.01 g/m2) to ~100 nm as the filter approached cake filtration (~ 0.3 g/m2) (Viswanathan et al., 2016). This is believed to be the effect of improvement in interception capture efficiency with increasing trapped mass within a filter. Since the interception mechanism is influenced by particle morphology the relative impact of different particle shapes on capture efficiency is of significant interest. In the current study, the CPMA, DMA, and SPLAT II instruments were used to characterize PM emissions from a wide range of SIDI operating conditions to understand the differences in particle morphologies from these conditions. In addition, a sequential combination of the instruments was used to select and characterize particles with the same mass and charge, but with different shapes (i.e. DSFs). Subsequently the EFA was used alongside these instruments to perform filtration experiments to identify the impacts of inlet particle morphology and size distribution on the filter capture efficiency at different levels of PM loading within the filter sample.

3. Theoretical background Shapiro et al. (2012) showed that virtually all properties of a particle can be determined with the knowledge of its mp , dm , d va , and ρp , all of which were experimentally evaluated in the current study. Both mp and d ve are intrinsic particle properties and are influenced by the particle composition (and thus ρp ). On the other hand, dm , which is arguably the most commonly evaluated particle property for engine emission studies, is not influenced by the particle composition, but significantly affected by particle shape. A brief review of the relationships between the experimentally derived properties and other properties of interest used in this study is presented in this section. A more detailed description of these parameters can be found elsewhere (DeCarlo et al., 2004; Shapiro et al., 2012; Zelenyuk et al., 2006). Under the assumption of uniform material density and negligible internal voids, the volume equivalent diameter (d ve ) corresponding to a given particle mass (mp ) was calculated using (1).

252

Journal of Aerosol Science 113 (2017) 250–264

S. Viswanathan et al.

6mp

d ve =

πρp

(1)

D

mp ∝ dm fm

(2)

For fractal agglomerates, the arrangement of the primary particles was described using the mass-mobility exponent (Dfm ) evaluated by applying a power-law expression to relate mp and dm of the particles as shown in (2) (Shapiro et al., 2012). The relationship between Dfm and the geometric fractal dimension (Df ) is described elsewhere (Park et al., 2002; Sorensen, 2011). To account for the effect of additional drag force on the fractal particles the DSFs in the transition ( χt ) and free-molecular regimes ( χv ) were calculated using (3) and (4), respectively.

χv =

ρp d ve ρ0 d va

(3)

χt =

dm Cc (d ve ) d ve Cc (dm)

(4)

where, ρ0 is the unit material density (1000kg /m3 ) and Cc is the Cunningham slip correction factor which accounts for the effect of flow regime on the drag on the particle. For a given mass or d ve , a larger DSF implies a larger dm , but a smaller d va . The relationship between mp and dm was used to determine the mobility size-resolved effective density ( ρeff , dm ) using (5). The ρeff , dm along with the SMPS-measured PSD were subsequently used to estimate the PM mass concentration in the exhaust (M ′′′) using an integrated particle size distribution (IPSD) method as shown in (6) (Li et al., 2014; Viswanathan et al., 2016). The IPSD method has been shown to agree well with more traditional gravimetric measurements when artifacts associated with absorption of volatile organic compounds onto the filter medium are minimal (Li et al., 2014; Viswanathan et al., 2016).

ρeff , dm =

6 mdm π dm3

(5)

π M ′′′ = ∑ ρeff , dm . . dm3 . ndm 6

TrappedMass (t ) =

∫0

t

(6)

′ ′′ ) vs Af dt (Min′ ′′ − Mout

(7)

During filtration experiments, the instantaneous trapped mass in the filter was estimated by using the IPSD method on the PSD ′ ′′ ) . The instantaneous trapped mass was integrated over the duration measured both upstream and downstream of the filter (Min′ ′′ & Mout of the experiments to estimate the total trapped mass in the filter as a function of loading duration (t) using (7), where vs is the flow velocity incident on the filter surface area ( Af ), referred to throughout as the superficial velocity or face velocity. For the SIDI particles used in the current study, the organic content was previously reported to be tightly bound to the elemental carbon (Gaddam & Vander Wal, 2013; Matthias et al., 2011; Stewart, 2013) and not impacted by additional sample conditioning using either a thermodenuder or an evaporative chamber (Matthias et al., 2011). However, in the presence of significant unburnt hydrocarbons in the raw exhaust, as is the case with rich SIDI operation, the IPSD method is expected to underestimate the true PM mass trapped in the filter (Li et al., 2014; Symonds, Reavell, Olfert, Campbell, & Swift, 2007; Viswanathan et al., 2016).

Fig. 1. Schematic of experimental layout used in the current study. Red lines indicate heated sections to minimize thermophoretic particle losses.

253

Journal of Aerosol Science 113 (2017) 250–264

S. Viswanathan et al.

4. Experimental setup The overall experimental layout used in the current study utilized an SIDI engine to generate distinct PSDs, a range of characterization equipment to understand the nature of the PM in the SIDI exhaust, and a wall-scale filtration system to evaluate the behavior of the particulate within the walls of an exhaust particulate filter. A schematic of the overall layout is presented in Fig. 1.

4.1. Engine & gaseous emissions measurements The SIDI engine was a single-cylinder, spray-guided setup adapted from the GM 2.2 L Ecotec and had a displacement volume of 549 cm3 with a geometric compression ratio of 11.97. The engine was fuelled with EPA Tier II EEE certification gasoline using a single-hole pressure-swirl injector with a 70° spray angle, oriented at 45° relative to the horizontal piston surface. A pair of intake and exhaust surge tanks were used to dampen any pressure pulsations caused by the single-cylinder setup. The major gaseous constituents in the undiluted exhaust stream were measured using a five-gas emissions bench (Horiba), which was capable of measuring total hydrocarbons (HC), CO, CO2, NOx, and O2 concentrations in the raw exhaust, as well as CO2 in the diluted sample during characterization and filtration. The emissions measurements were used to determine oxygen- and carbon-based air-fuel ratios, and the total dilution ratio achieved by the dilution system.

4.2. PM characterization Several instruments were used to characterize the particulate emissions from the different engine operating conditions, post dilution (Fig. 1). A two-stage partial-flow diluter (PFD) was used to sample raw exhaust downstream of the exhaust surge tank. The temperature at the sampling location varied between 260 and 400 °C depending on the engine operating condition and duration of the experiment. The first dilution stage in the PFD used heated, dry, filtered air at 235 °C to minimize artifacts due to particle losses and condensation. The second stage used filtered air at room temperature to minimize further nucleation. A TSI 3081 long DMA built into a TSI SMPS was used to measure the mobility size distribution in the diluted exhaust stream during the initial characterization. The DMA was controlled using a TSI 3080 electrostatic classifier while a TSI model 3010 condensation particle counter (CPC) was used to count particles in the mono-disperse sample from the DMA. The SMPS was used to measure concentrations of particles in the mobility diameter range from 7 to 290 nm. Detailed characterization of the SIDI PM upstream and downstream of the filter was performed with a CPMA, a DMA (second SMPS), and a single-particle mass spectrometer (SPLAT II). The CPMA isolates particles with a narrow mass-to-charge ratio distribution by establishing a balance between electrostatic and centrifugal forces acting on the particles. SPLAT II uses an aerodynamic lens to form a collimated particle beam in which a size-dependent velocity is imparted to each particle. Once a particle is detected and its velocity, hence vacuum aerodynamic diameter, is determined, a CO2 laser pulse is used to heat the particle and desorb the semivolatile fraction. Subsequently, a time-delayed ultraviolet excimer laser pulse is used to ionize the evaporated plume and ablate the remaining non-volatile fraction. Individual particle mass spectra are acquired using an angular reflector, time-of-flight mass spectrometer. In the current study, material density was determined based on the particle mass spectra, assuming volume additivity and using densities of 2.26 g/cm3 for elemental carbon, 1.0 g/cm3 for partially oxidized organics, and 1.35 g/cm3 for poly-aromatic hydrocarbons. This approach results in material densities of 1.7 – 1.8 g/cm3 for fractal soot particles generated by different engines, including GCI, GDI, diesel (China et al., 2015; Zelenyuk et al., 2014, 2016) and other combustion sources (propane, mini-CAST), which is in good agreement with literature (Johnson, Devillers, & Thomson, 2013). More information on the SPLAT II and its diverse capabilities are available in (Matthias et al., 2011; Zelenyuk et al., 2014; Zelenyuk, Yang, Choi, & Imre, 2009). The layout of the different instruments which constitute the CPMA/DMA/SPLAT system is shown in Fig. 1 while Table 1 outlines the different arrangements of the available instruments used during various stages of the PM characterization study. The different layouts allowed characterization of various properties, both through direct measurements and using the relationships discussed in Section 3. The measured and calculated properties are also summarized in Table 1. Additional information on these measurement techniques can be found in Beranek et al. (2012) and Zelenyuk et al. (2014). Table 1 Different layouts used during the PM characterization study and the resultant measured and calculated particle properties. Id

Layout

Measured

A

PFD → SMPS

PSD (n dm vs.

B

PFD → CPMA → SMPS

mp vs. dm

Calculated

dm )

M ′′′ (6) Dfm (2) ρeff (5)

χt (4) C D

PFD → SPLAT II PFD → CPMA → SPLAT II

dva distribution of all particles mp -resolved mass spectra & dva distributions

E

PFD → CPMA → DMA → SPLAT II

Shape-resolved and mass-resolved mass spectra & dva distributions.

254

ρp (from mass spectra) χv (3) dve (1)

Journal of Aerosol Science 113 (2017) 250–264

S. Viswanathan et al.

4.3. Filtration setup The filtration experiments were performed using an exhaust filtration analysis system. The EFA uses small pieces of the filter wall called wafers. An image of a typical wafer is shown in the bottom right corner of Fig. 1. The flow areas at the top and bottom faces of the filter samples were restricted using silicone O-rings with 1-in. inner diameter. The exposed filter face area was used to estimate the superficial velocity based on the exhaust flow rate through the filter samples. During the filtration experiment, the wafers were housed in a stainless-steel holder and heated to the desired temperature in an oven. Flow through the filter was regulated using a bypass valve, which provided an alternate path for exhaust flow around the filter sample. The pressure drop (∆p ) across the filter holder was continuously recorded using a differential pressure transducer (1 psi range). An ejector diluter was used to dilute the exhaust stream downstream of the oven before subsequent characterization of the particles that made it through the filter. 5. Filter characterization results Different techniques were used to understand the microstructure of the filter samples used in this study. A representative filter sample was characterized using mercury intrusion porosimetry. The distribution of pore sizes obtained using the intrusion measurements is shown in Fig. 2. The pore diameter was evaluated assuming a contact angle of 130° between the substrate and mercury. The estimated porosity and median pore diameter based on these measurements agreed well with porosimetry results provided by the filter manufacturer. The median pore diameter was evaluated as the diameter corresponding to 50% cumulative intrusion volume. The full width at half maximum (FWHM) for the pore distribution was ∼10 µm. A different filter sample from the same batch was analyzed using 3D X-ray CT analysis to characterize the distribution of filter porosity as a function of filter depth. The results of this analysis are also presented in Fig. 2. For the samples used in the current study, measurable asymmetry (~5–15%) was observed between the top and bottom surfaces in terms of the local porosity. The current experiments were performed with the exhaust flow entering the high porosity side of the filter samples (right hand axis of Fig. 2). To quantify filter sample-to-sample variability, the wall permeability of each wafer used was measured using the EFA system. The filter permeability (k ) is a measure of its resistance to flow and is given by (8).

k = μL

vs ∆p

(8)

where, μ is the viscosity of the fluid, L is the filter thickness, and ∆p is the pressure drop across the filter. Filtered air at room temperature was passed through the clean filter samples at different superficial velocities ranging from 0.5 to 4 cm/s, while the corresponding pressure drops were recorded to evaluate k (Wirojsakunchai et al., 2008). A summary of the filter properties evaluated using these methods is presented in Table 2. 6. PM characterization results Before the filtration experiments, a detailed characterization study was performed to understand the nature of PM from the different operating conditions. The layout Id shown in Table 1 is used to indicate the respective layouts used to obtain the various characterization results discussed in this section 6.1. Particle size distributions PM from five distinct engine operating conditions were used in this study. The EOI 280 condition had an end-of-injection (EOI) timing of −280 crank angle degrees (CAD) resulting in minimal piston wetting and significant time for fuel-air mixing (0 CAD

Fig. 2. Distribution of pore volume obtained using mercury intrusion porosimetry (solid black) and porosity profile along filter depth obtained using X-ray CT (dashed red).

255

Journal of Aerosol Science 113 (2017) 250–264

S. Viswanathan et al.

Table 2 Properties of filter samples used in the current study. Material

Cordierite

Filter thickness Mean porosity Median pore diameter Permeability Catalyst loading

0.98 mm 43% 14 µm 6.8 ± 0.1 *1e−13 m2 –

corresponds to top dead center (TDC) of compression and negative values are before TDC of compression). A slightly retarded injection timing (−220 CAD) was used in the EOI 220 condition to obtain more stratification of the air-fuel mixture at the start of combustion. The MBT-10 condition was identical to the EOI 220 condition but had a retarded spark timing (by 10 °) to delay the start of combustion. The last two conditions were used to achieve rich and heavy-load SIDI operation. More information on these operating conditions is provided in Table 3. Results from SMPS measurements of PM in the SIDI exhaust obtained using layout A (Table 1) are shown in Fig. 3(a). The dashed lines indicate the 95% confidence interval bounds using a Student’s t-distribution from 7 consecutive SMPS scans under steady-state engine operation. The near homogeneous conditions for EOI 280 resulted in the least PM emissions among the different operating modes with a large fraction of the particles having mobility diameters < 150 nm. The additional stratification for the EOI 220 condition resulted in higher PM emissions, though the shape of the PSD is similar to the EOI 280 condition. The longer mixing time due to the retarded spark and possibly the higher late cycle temperatures for the MBT-10 condition resulted in a noticeable reduction in the concentration of large agglomerates compared to the EOI 220 condition. Both rich and heavy-load (HL) operation resulted in a higher fraction of large agglomerates compared to the other conditions. 6.2. Mass-mobility relationship The relation between particle mass and mobility diameter was obtained for a wide range of particle sizes by measuring the mobility distribution of mass-selected particles using layout B (Table 1). The results from these measurements performed on different engine operating conditions are presented in Fig. 3(b). In the current study, the Dfm was evaluated for fractal particles defined by dm > 50 nm. The choice of 50 nm as the cut-off diameter for evaluating the mass-mobility exponent was based on the gradual deviation in slope (in the mass-mobility log-log plot) observed for particles smaller than 50 nm in Fig. 3(b) and in previous measurements performed on the same engine (Sandeep Viswanathan et al., 2016). For the range of particle sizes evaluated in this study, the mass-mobility relationship was found to be only weakly dependent on the engine operating mode. The resultant mass-mobility exponent (Dfm ) was evaluated from the slope of the mass mobility data for the fractal particles and found to lie within a narrow range of 2.4 ± 0.1, consistent with relevant studies using light-duty diesel and SIDI / GDI PM (Harris & Maricq, 2001; Momenimovahed & Olfert, 2015; Quiros et al., 2015). Based on the shift in MPPS from ~200 nm to ~100 nm observed during previous experiments (Viswanathan et al., 2016), the interception capture mechanism and consequently the particle shape was expected to impact particles larger than 100 nm. Engine-out PSD measurements seen in Fig. 3(a) showed a rapid reduction in number concentration for particles larger than 200 nm for all SIDI conditions. Therefore, the current study focuses primarily on particles in the size range of 100–200 nm (or mass range of 0.3–2 fg). 6.3. Particle morphology SPLAT II was initially used to measure the d va distributions of all particles in the SIDI exhaust using layout C (Table 1). The CPMA was subsequently used to select particles with a narrow range of mp ; while the SPLAT II was used to measure the d va distribution of the mass-selected particles (layout D). Fig. 4 compares the dva distribution of mass-selected, 0.7 fg particles, from the rich and heavy load SIDI operating conditions. These conditions were picked due to their large concentration of accumulation mode particles, which can exhibit different shapes due to their agglomerated structure. As is evident from Fig. 4, HL operation resulted in a wider dva Table 3 Details of engine operating conditions used in the current study. Condition

Unit

EOI 280

EOI 220

MBT-10

Rich

Heavy Load

Load/IMEP COV of IMEP Speed Fuel injection pressure End of injection Air/ fuel ratio Intake manifold pressure Exhaust back pressure

Bar % rpm bar CAD – kPa-abs kPa-abs

3.5 1.2 2100 110 −280 15 35 102

3.35 4 2100 110 −220 15 35 102

3.1 8 2100 110 −220 15 35 102

3 2.5 2100 110 −220 13 31 102

6.5 2.5 2100 110 −220 15 60 102

256

Journal of Aerosol Science 113 (2017) 250–264

S. Viswanathan et al.

Fig. 3. (a) Particle size distributions measured using the SMPS from different engine operating conditions. (b) Mass-mobility relationship for different SIDI engine operating conditions.

Fig. 4. Normalized dva distributions of 0.7 fg particles from the rich and heavy load operating conditions measured using layout D (Table 1).

distribution for the mass-selected particles. Single particle mass spectra analysis clearly indicated that the wide dva distribution of mass-selected HL particles was not caused by a difference in particle densities. Additionally, the dm distributions of mass-selected HL soot particles did not indicate observable contribution of doubly-charged particles, rendering the impact of multiple charges insignificant. A broader relative width of the dva

Fig. 5. (a) Normalized dva distributions of mass-selected and all particles (shaded) from the heavy-load condition measured using SPLAT II. (b) Normalized dva distribution of mass and mobility classified (shape-selected) 0.7 fg particles from the heavy-load condition measured using SPLAT II. Mobility diameter is given after the mass in the legend.

257

Journal of Aerosol Science 113 (2017) 250–264

S. Viswanathan et al.

distributions therefore indicates a wider range in DSFs due to the presence of different particle shapes and orientations, potentially due to the longer injection durations during HL operation, resulting in more in-cylinder fuel stratification or due to pressure influences on the soot formation. The wider range of particle shapes, made the HL condition suitable for the fundamental filtration experiments discussed in Section 8. Fig. 5(a) shows the normalized d va distribution for all particles generated during HL operation (shaded) along with the distributions for mass-selected particles. The FWHM of the dva size distributions for all HL particles shown in Fig. 5(a) was 54%. The FWHM of the dva size distributions for mass-selected particles were 31%, 40%, 46%, 47% and 47% for the 0.2, 0.4, 0.7, 1 and 2 fg particles, respectively. For comparison, the FWHM of mass-selected spherical particles is typically smaller than 5% (Beranek et al., 2012). An increase in particle mass by a factor of 10 (0.2–2 fg) resulted only in a small increase in particle dva (59–80 nm) compared to the corresponding change in dm (70–200 nm). The average diameter of primary spherules that comprise soot agglomerates generated under HL condition was estimated to be 19 ± 1 nm which was in good agreement with previous TEM images (Gaddam & Vander Wal, 2013). Other attributes, including number of spherules, void fraction, etc., were also derived using the mass-dva relationship, but are not discussed here since they were outside the scope of the current study. The distribution of shapes associated with the mass-selected particles, was further characterized by placing a DMA downstream of the CPMA to isolate particles with a narrow χt and subsequently measuring the d va distribution of the mass- and shape-selected particles using the SPLAT II (layout E, Table 1). A schematic of the setup is shown in Fig. 6 below. The normalized d va distribution of dm -selected 0.7-fg particles from the HL condition are shown in Fig. 5(b). The three distinct peaks correspond to 0.7-fg particles with mobility diameters of 112, 131 and 147 nm. The larger dm was seen to correspond with a smaller d va and result in narrower peaks with FWHM of ~25%. Note that a larger mobility diameter for a given mass (or d ve ) implies a less compact agglomerate and therefore a larger shape factor. Fig. 5(b) demonstrates the ability of the CPMA/DMA/SPLAT system to separate particles based on their shape. The dynamic shape factors in the transition and free molecular regimes (χv and χt) were calculated for particles in the mass range from 0.2 to 2 fg using (3) and (4). These results are summarized in Fig. 7(a) and (b) where the bars indicate the FWHM from each mass-selected dm-distribution. In both plots the open symbols indicate χ values of specific shape-selected particles obtained using CPMA, DMA, and SPLAT II (layout E) while the shaded bands indicate the range of DSFs measured for mass-selected particles using just the CPMA and SPLAT II or SMPS (layout D or B). We chose to probe the filter with 0.7-fg and 1-fg samples because they corresponded to a mobility range from 110 to 175 nm, where the particle concentrations were relatively high and shape was expected to have a bigger impact on filtration due to the interception mechanism. Fig. 7(a) and (b) also shows that χv is larger than χt, especially for larger particles. These differences result from the size-dependent void fractions of soot particles and how a particle’s shape affects its drag in different flow regimes. For the 0.7-fg particles with mobility diameters of 112, 131, and 147 nm, the corresponding χt values were 1.42, 1.85, and 2.24, respectively, while χv values were 1.97, 2.35, and 2.75. The FWHM for both the shape factors was ~0.5 as shown in Fig. 7. The demonstrated ability to identify the presence of different shapes motivated further fundamental filtration experiments to understand the impact of particle shape on capture efficiency in a filter. The shape factors observed in the transition regime were believed to be more relevant to the filtration experiments considering the conditions seen by the particles within the filter bed. The filtration experiments were performed under the premise that a noticeable difference in the d va distribution of mass-selected particles, upstream, and downstream of a filter, can be attributed to the preferential capture of particles of certain shapes within the filter. Measurement of χt during the entire wall loading stage was expected to help identify the critical trapped mass where the impact of particle shape becomes significant. 7. Procedure for filtration experiments The filtration experiments performed in the current study can be broadly classified as “filter loading” and “filter probing” experiments. Filter loading refers to the process of deliberately modifying the filter structure by passing PM laden exhaust through the

Fig. 6. Schematic of experimental setup used to isolate and characterize shape selected particles using a CPMA, DMA and SPLAT II.

258

Journal of Aerosol Science 113 (2017) 250–264

S. Viswanathan et al.

Fig. 7. Range of shape factors in the (a) transition and (b) free molecular regimes associated with mass-selected particles from the heavy load SIDI operating condition. Red triangles and black squares show average DSFs for a given mass, shaded areas depict FWHM of the DSF distributions. Also shown are DSFs that were probed using CPMA/DMA/SPLAT system.

filter sample for significant durations of time (30–60 min). Filter probing refers to the process of evaluating the filter performance at any stage of PM loading, while ensuring that the change in trapped mass during the probing process was small enough to not have a noticeable impact on the filter performance. Filter probing was performed over a 2–5-min duration to reduce measurement uncertainties. All filtration experiments were performed at a filtration velocity of 2.7 ± 0.15 cm/s. A schematic of the experimental layout used during the filter loading process is shown in Fig. 8(a). Raw exhaust was sampled just downstream of the exhaust surge tank and sent through the EFA which housed the filter sample. The filter temperature was maintained at 125 ± 10 °C to avoid any condensation of water vapor or potential particle oxidation. The downstream PSD was measured using an SMPS post dilution by the ejector diluter in the EFA (Fig. 1). The upstream PSD was measured real-time using a TSI engine exhaust particle spectrometer (EEPS) post dilution using the PFD. The dilution ratios were estimated using the CO2 concentrations measured in the raw and diluted exhaust streams. The measured size distributions were corrected for dilution and systematic errors due to differences in operating principles of the EEPS and SMPS. More information on the data processing procedure is presented in Viswanathan et al. (2014) and Viswanathan et al. (2015). The filter loading process was paused intermittently to probe the filter with different inlet PSDs and mass-selected particles. The experimental setup used during filter probing with the different inlet PSDs was identical to the one used during filter loading (Fig. 8(a)). The filter was successively loaded with PM from the different conditions in order of increasing IPSD mass concentration in the exhaust (EOI 280 → MBT-10 → EOI 220 → Heavy Load → Rich). When switching between conditions, the exhaust flow was diverted around the filter sample using a bypass valve to avoid artifacts due to rich transients. The inlet PSDs for probing were selected such that the IPSD mass in the exhaust from the probing condition was significantly smaller than the IPSD mass in the exhaust from the most recent loading condition. A second set of probing experiments were performed using mass-selected particles obtained by placing the CPMA downstream of the PFD, as shown in Fig. 8(b). The mass-selected particles were passed through the EFA, and SPLAT II was used to measure the d va distribution downstream of the EFA. Within the EFA, the flow was initially diverted around the filter sample (through the bypass valve) to measure the d va distribution at the filter inlet. Subsequently, the bypass valve was shut and flow was forced through the

Fig. 8. (a) Experimental layout used during filter loading process. (b) Experimental layout used during filter probing with mass-selected particles.

259

Journal of Aerosol Science 113 (2017) 250–264

S. Viswanathan et al.

filter sample to measure the filtration effect on the d va distribution of the mass-selected particles. Note that, based on SPLAT measurements, the different pathways between engine-out and EFA by-pass valve PM sampling, resulted in very minor changes in particle mass spectra, fractal dimension, effective density, and dva distributions of mass- and mobility selected particles. 8. Results and discussion The clean filter was initially probed to understand the impacts of particle shape and inlet PSD on the filtration efficiency. Next, the filter was loaded with SIDI PM and probed after different loading durations. This was done to determine whether the change in filtration length scales in the partially loaded filter had a measurable impact on the capture efficiency of different particle shapes. 8.1. Clean filter probing Fig. 9(a) shows the size-resolved FE for flow through a clean filter sample using PSDs from four different loading conditions. The dashed lines indicate the 95% uncertainty bounds for the measured FE from at least 5 SMPS scans. Significant noise was observed for particle sizes > 200 nm due to the low concentration of these particles observed in the raw exhaust (see Fig. 3(a)). The size-resolved FE was seen to exhibit consistent overlap for the four different inlet PSDs. Any discrepancies observed were within the measurement uncertainty for all particle sizes. The MPPS was observed to be > 200 nm based on the size-resolved FE curves. The FE at MPPS was relatively high (> 70%), despite the clean state of the filter, due to the relatively large filter thickness (See Table 2). Fig. 9(b) shows the normalized dva distributions of 0.7-fg mass-selected particles measured using SPLAT II for flow through the PFD (Layout C, Table 1), through the EFA bypass valve, and through the clean filter (Fig. 8(b)). No noticeable change in shape or position of the d va distribution was observed for flow through the filter; suggesting that, particles of specific shapes were not preferentially captured in this size range. Classical filtration theory attributes the presence of an MPPS due to competing capture mechanisms which are sensitive to particle size (Hinds, 1999). In general particles smaller than the MPPS are believed to be captured predominantly by the diffusion while larger particles are captured by a combination of diffusion, interception and impaction mechanisms. Based on the MPPS of approximately 200 nm seen in Fig. 9(a), diffusion is expected to be the dominant capture mechanism for the 0.7-fg (~130 nm) particles within the clean filter. The corresponding FE of the 0.7-fg particles within the clean filter was measured to be 80 ± 6%. The results seen in Fig. 9(b) suggest that the range of shapes observed in the current study did not have a significant impact on the diffusion capture mechanism within the clean filter. 8.2. PM laden filter probing 8.2.1. Effect of inlet PSDs After probing the clean filter, exhaust from different SIDI conditions was sent through the filter sample. Fig. 10(a) shows the change in penetration of dm = 50, 90, 150 and 225 nm particles as a function of trapped mass during filter loading with PM from four different SIDI conditions. Utilizing the different SIDI conditions in order of increasing mass concentration allowed for the evaluation of filtration performance over four orders of magnitude of trapped mass (Viswanathan et al., 2016). The penetration results showed continuity between the operating conditions despite the differences in PM observed during the characterization study. The filtration performance appeared to be unchanged for trapped mass < 1e-3 g/m2. When switching between the SIDI loading conditions, the filter was probed with different inlet PSDs to evaluate filtration performance of the partially loaded filter and its potential sensitivity to PSD. The regions where the probing experiments were performed are indicated by the vertical dashed lines between the different loading steps. The continuity observed in Fig. 10(a) (on either side of the probing experiments) suggested that probing the filter with the different inlet PSDs did not alter the filter structure enough to

Fig. 9. (a) Size-resolved filtration efficiency for different inlet PSDs for a clean filter. (b) dva distribution of mass-selected 0.7 fg particles for flow through the PFD, bypass valve, and clean filter.

260

Journal of Aerosol Science 113 (2017) 250–264

S. Viswanathan et al.

Fig. 10. (a) Change in penetration of 50, 90, 150, and 225 nm particles with IPSD trapped mass during filter loading for PM from the four different SIDI conditions specified in the legend. Size-resolved FE from different SIDI operating conditions at filter mass loadings of (b) 4e-3 g/m2 and (c) 1.25e-2 g/m2.

change its performance. Fig. 10(b) and (c) show the size-resolved FE for the probing experiments with the different inlet PSDs at estimated filter mass loadings of 4e-3 g/m2 and 1.25e-2 g/m2. The FE curves were observed to become relatively flat beyond dm = 150 nm in Fig. 10(b) and dm = 80 nm in Fig. 10(c) at the same time the corresponding minimum FE increased from 83 ± 6% to 95 ± 4%. As seen with the clean filter (Fig. 9(a)), the size-resolved FE curves were observed to be insensitive to the engine operating condition and inlet PSD and only sensitive to the trapped mass within the filter.

Fig. 11. (a) Change in penetration of 50, 90, 150, and 225 nm particles with IPSD trapped mass during filter loading with PM from the heavy-load condition. (b) Change in size-resolved FE with increasing trapped mass in the filter. The legend shows the estimated IPSD trapped mass in the filter corresponding to each FE curve. The dashed arrow shows the shift in most penetrating particle size.

261

Journal of Aerosol Science 113 (2017) 250–264

S. Viswanathan et al.

8.2.2. Effect of particle shape Due to the insensitivity to inlet PSD demonstrated during the loading and probing experiments in Section 8.2; probing experiments with mass-selected particles were performed with PM from just the HL condition. Fig. 11(a) shows the change in penetration for 4 different particle sizes (dm = 50, 90, 150, and 225 nm) as function of trapped mass in the filter. As is evident from the penetration curves, the probing experiments did not measurably alter the filter structure. Fig. 11(b) shows the change in size-resolved FE with increasing trapped mass (estimated trapped mass is given in the legend). As in the previous experiments, the flattening of the FE curve beyond the MPPS suggests that particles in this size range were being captured by a combination of both diffusion and interception capture mechanisms. It is also possible that these mechanisms alone cannot be used to explain the trends observed here. Fig. 12(a) and (b) show the normalized d va distributions for 0.7-fg particles for flow through the PFD, through the bypass valve, and through the filter at two different IPSD mass loadings (3e-3 g/m2 and 3e-1 g/m2). The corresponding filtration efficiencies for the 0.7-fg particles were 84 ± 4% and 98 ± 0.5%. Despite the significant increase in capture efficiency within the filter, the d va distributions appeared to be relatively unaffected for the range of shapes observed in the heavy-load PM. This can in part be explained by the relatively flat FE curves observed in Fig. 11(b) for a broad range of particle sizes beyond the MPPS. Based on this observation, the FE can also possibly be expected to be insensitive to the range of particle shapes observed under the current experimental conditions. This conclusion is based on the relationship between dm and χt shown in (4). For a given volume equivalent diameter (or particle mass), the shape factor is solely a function of the mobility diameter. If the filtration efficiency is approximately independent of dm for dm greater than the MPPS (in the present size range), then the filtration efficiency for a given particle mass or volume equivalent diameter must also be independent of the particle shape as represented by χt . 9. Conclusion The effects of inlet particle properties and inlet particle size distribution (PSD) on filtration performance were evaluated using particulate matter (PM) from a spark-ignition direct-injection (SIDI) engine for exhaust flow through nearly identical cordierite filter samples. These samples had micro-structural properties representative of diesel particulate filters with a porosity of 43% and a median pore diameter of 14 µm. The SIDI engine was operated under different steady-state conditions to generate distinct PSDs for the filtration experiments. A wide range of particle characterization instruments were used to understand the properties of the PM emissions from the different SIDI conditions and evaluate the filtration performance at different stages of filter loading. The stoichiometric SIDI conditions generated PSDs with a significant fraction of particles with mobility diameter (dm ) smaller than 70 nm with heavy-load operation resulting in the largest fraction of agglomerate particles (dm > 50 nm). Rich operation resulted in a larger fraction of particles in the size range of dm = 70–100 nm. Despite these differences, the relationship between particle mass (mp ) and dm was insensitive to the engine operating condition. The resultant mass-mobility exponent was 2.4 ± 0.1. The CPMA/ DMA/SPLAT system was used to identify the range of particle masses and shapes from the different operating conditions. Heavy-load SIDI operation resulted in the widest range of particle shapes in the 130 nm mobility diameter size range. The mass-dependent average dynamic shape factors were characterized in the transition and free-molecular flow regimes and were found to lie between 1.65–2 and 1.8–2.8, respectively. Starting from a clean filter, the size-resolved filtration efficiency (FE) remained unchanged for trapped masses < 1e-3 g/m2 and increased to > 99% at a trapped mass of ~ 0.1 g/m2 indicating transition to cake filtration. The most penetrating particle size (MPPS)

Fig. 12. Normalized dva distribution of 0.7 fg particles for flow through the PFD, bypass valve and the filter at two different IPSD mass loadings of (a) 3e-3 g/m2 and (b) 3e-1 g/m2. The corresponding filtration efficiencies for the 0.7-fg particles were 84 ± 4% and 98 ± 0.5%.

262

Journal of Aerosol Science 113 (2017) 250–264

S. Viswanathan et al.

was observed to reduce from ~200 nm for a clean filter to ~100 nm during transition to cake filtration irrespective of the inlet PSD. The filtration performance of the clean and partially loaded filter samples was found to be relatively insensitive to the inlet PSD. No discernible change in the d va -distribution of mass-selected 0.7 fg (130 nm) particles was observed upstream and downstream of the filter despite an increase in capture efficiency from 80 ± 6 to > 99%. These results suggest that the FE was also insensitive to the range of particle shapes explored in the current study. This is believed to be the result of the diffusion dominated capture mechanism for the particle sizes in the current study. Within the present distribution of particle sizes, the size-resolved FE was observed to become relatively insensitive to mobility diameter beyond the MPPS with increasing trapped mass, consistent with the lack of sensitivity to particle shape. The trapped mass within the filter appeared to play the most significant role in impacting the size-resolved capture efficiency within the walls of the exhaust particulate filter. Notes The authors declare no competing financial interest. Acknowledgements This work was funded by General Motors through the Collaborative Research Laboratory at the University of Wisconsin-Madison, Engine Research Center and the US Department of Energy (DOE), Office of Energy Efficiency and Renewable Energy, Vehicle Technology Office. Development of SPLAT II was supported by the Office of Basic Energy Sciences, Division of Chemical Sciences, Geosciences, and Biosciences and EMSL, a national scientific user facility by the US Department of Energy's Office of Biological and Environmental Research and located at Pacific Northwest National Laboratory. References Beranek, J., Imre, D., & Zelenyuk, A. (2012). Real-time shape-based particle separation and detailed in situ particle shape characterization. Analytical Chemistry, 84(3), 1459–1465. http://dx.doi.org/10.1021/ac202235z. Boskovic, L., Altman, I. S., Braddock, R. D., & Agranovski, I. E. (2009). Removal of elongated particle aggregates on fibrous filters. CLEAN – Soil, Air, Water, 37(11), 843–849. http://dx.doi.org/10.1002/clen.200900175. Chan, T. W., Meloche, E., Kubsh, J., Rosenblatt, D., Brezny, R., & Rideout, G. (2012). Evaluation of a gasoline particulate filter to reduce particle emissions from a gasoline direct injection vehicle. SAE International Journal of Fuels and Lubricants, 5(3), 1277–1290. http://dx.doi.org/10.4271/2012-01-1727. Cheng, M.-T., Xie, G.-W., Fu, T.-H., & Shaw, D. T. (1991). Filtration of ultrafine chain aggregate aerosols by nuclepore filters. Aerosol Science and Technology, 15(1), 30–35. http://dx.doi.org/10.1080/02786829108959510. DeCarlo, P. F., Slowik, J. G., Worsnop, D. R., Davidovits, P., & Jimenez, J. L. (2004). Particle morphology and density characterization by combined mobility and aerodynamic diameter measurements. Part 1: Theory. Aerosol Science and Technology, 38(12), 1185–1205. http://dx.doi.org/10.1080/027868290903907. Gaddam, C. K., & Vander Wal, R. L. (2013G). Physical and chemical characterization of SIDI engine particulates. Combustion and Flame, 160(11), 2517–2528. http:// dx.doi.org/10.1016/j.combustflame.2013.05.025. Hall, D. E., & Dickens, C. J. (1999). Measurement of the number and size distribution of particles emitted from a gasoline direct injection vehicle. SAE Technical Paper. http://dx.doi.org/10.4271/1999-01-3530 1999-01-3530. Harris, S. J., & Maricq, M. M. (2001). Signature size distributions for diesel and gasoline engine exhaust particulate matter. Journal of Aerosol Science, 32(6), 749–764. http://dx.doi.org/10.1016/S0021-8502(00)00111-7. Hinds, W. C. (1999). Aerosol technology: Properties, behavior, and measurement of airborne particles. John Wiley & Sons, Inc (ISBN: 978-0-471-19410-1). Ibald-Mulli, A., Wichmann, H., Kreyling, W., & Peters, A. (2002). Epidemiological evidence on health effects of ultrafine particles. Journal of Aerosol Medicine, 15(2), 189–201. http://dx.doi.org/10.1089/089426802320282310. Johnson, M. R., Devillers, R. W., & Thomson, K. A. (2013). A generalized Sky-LOSA method to Quantify Soot/Black Carbon emission rates in atmospheric plumes of gas flares. Aerosol Science and Technology, 47(9), 1017–1029. http://dx.doi.org/10.1080/02786826.2013.809401. Johnson, T. (2013). Vehicular emissions in review. SAE International Journal of Engines, 6(2), 699–715. http://dx.doi.org/10.4271/2013-01-0538. Kim, S. C., Wang, J., Emery, M. S., Shin, W. G., Mulholland, G. W., & Pui, D. Y. H. (2009). Structural property effect of nanoparticle agglomerates on particle penetration through fibrous filter. Aerosol Science and Technology, 43(4), 344–355. http://dx.doi.org/10.1080/02786820802653763. Konstandopoulos, A., & Johnson, J. (1989). Wall-Flow Diesel Particulate Filters-Their Pressure Drop and Collection Efficiency. Paper presented at the SAE technical paper 890405. 〈http://dx.doi.org/:10.4271/890405〉. Konstandopoulos, A. G., Kostoglou, M., Vlachos, N., & Kladopoulou, E. (2007). Advances in the science and technology of diesel particulate filter simulation. Advances in chemical engineering: Vol. 33, (pp. 213–294). Academic Press. http://dx.doi.org/10.1016/S0065-2377(07)33004-4. Lange, R., Fissan, H., & Schmidt-Ott, A. (2000). A new method for the determination of the interception equivalent diameter. Journal of Aerosol Science, 31(2), 221–231. http://dx.doi.org/10.1016/S0021-8502(99)00050-6. Li, Y., Xue, J., Johnson, K., Durbin, T., Villela, M., Pham, L., . . . Karavalakis, G. (2014). Determination of suspended exhaust PM mass for light-duty vehicles. Paper presented at the SAE Technical Paper 2014-01-1594. 〈http://dx.doi.org/10.4271/2014-01-1594〉. Mamakos, A., Martini, G., Marotta, A., & Manfredi, U. (2013). Assessment of different technical options in reducing particle emissions from gasoline direct injection vehicles. Journal of Aerosol Science, 63, 115–125. http://dx.doi.org/10.1016/j.jaerosci.2013.05.004. Matthias, N., Farron, C., Foster, D. E., Andrie, M., Krieger, R., Najt, P., ... Zelenyuk, A. (2011). Particulate matter sampling and volatile organic compound removal for characterization of spark ignited direct injection engine emissions. SAE International Journal of Fuels and Lubricants, 5(1), 399–409. http://dx.doi.org/10.4271/ 2011-01-2100. Momenimovahed, A., & Olfert, J. S. (2015). Effective density and volatility of particles emitted from gasoline direct injection vehicles and implications for particle mass measurement. Aerosol Science and Technology, 49(11), 1051–1062. http://dx.doi.org/10.1080/02786826.2015.1094181. Park, K., Cao, F., Kittelson, D. B., & McMurry, P. H. (2002). Relationship between particle mass and mobility for diesel exhaust particles. Environmental Science & Technology, 37(3), 577–583. http://dx.doi.org/10.1021/es025960v. Peters, A., Wichmann, H., Tuch, T., Heinrich, J., & Heyder, J. (1997). Respiratory effects are associated with the number of ultrafine particles. American Journal of Respiratory and Critical Care Medicine, 155(4), 1376–1383. http://dx.doi.org/10.1164/ajrccm.155.4.9105082. Quiros, D. C., Hu, S., Hu, S., Lee, E. S., Sardar, S., Wang, X., ... Huai, T. (2015). Particle effective density and mass during steady-state operation of GDI, PFI, and diesel passenger cars. Journal of Aerosol Science, 83, 39–54. http://dx.doi.org/10.1016/j.jaerosci.2014.12.004. Rakovec, N., Viswanathan, S., & Foster, D. E. (2011). Micro-scale study of DPF permeability as a function of PM loading. Paper presented at the SAE technical paper 2011-

263

Journal of Aerosol Science 113 (2017) 250–264

S. Viswanathan et al.

01-0815. 〈http://dx.doi.org/10.4271/2011-01-0815〉. Shapiro, M., Vainshtein, P., Dutcher, D., Emery, M., Stolzenburg, M., Kittelson, D., McMurry, P., et al. (2012). Characterization of agglomerates by simultaneous measurement of mobility, vacuum aerodynamic diameter and mass. Journal of Aerosol Science, 44, 24–45. http://dx.doi.org/10.1016/j.jaerosci.2011.08.004. Sorensen, C. M. (2011). The mobility of fractal aggregates: A review. Aerosol Science and Technology, 45(7), 765–779. http://dx.doi.org/10.1080/02786826.2011. 560909. Stewart, M. (2013). Fuel-neutral studies of particulate matter transport emissions. Retrieved from Washington DC. Stojkovic, B. D., Fansler, T. D., Drake, M. C., & Sick, V. (2005). High-speed imaging of OH* and soot temperature and concentration in a stratified-charge directinjection gasoline engine. Proceedings of the Combustion Institute, 30, 2657–2665. http://dx.doi.org/10.1016/j.proci.2004.08.021. China, S., Kulkarni, G., Scarnato, B. V., Sharma, N., Pekour, M., Shilling, J. E., ... Liu, S. (2015). Morphology of diesel soot residuals from supercooled water droplets and ice crystals: Implications for optical properties. Environmental Research Letters, 10(11), 114010. http://iopscience.iop.org/1748-9326/10/11/114010. Symonds, J. P., Reavell, K. S. J., Olfert, J. S., Campbell, B. W., & Swift, S. J. (2007). Diesel soot mass calculation in real-time with a differential mobility spectrometer. Journal of Aerosol Science, 38(1), 52–68. http://dx.doi.org/10.1016/j.jaerosci.2006.10.001. Valavanidis, A., Fiotakis, K., & Vlachogianni, T. (2008). Airborne particulate matter and human health: Toxicological assessment and importance of size and composition of particles for oxidative damage and carcinogenic mechanisms. Journal of Environmental Science and Health Part C Environmental Carcinogenesis & Ecotoxicology Reviews, 26(4), 339–362. http://dx.doi.org/10.1080/10590500802494538. Viswanathan, S., Rakovec, N., & Foster, D. E. (2012). Microscale study of ash accumulation process in DPF walls using the diesel exhaust filtration analysis (DEFA) system. ASME 2012 internal combustion engine division fall technical conference (pp. 537–549). 〈http://dx.doi.org/10.1115/ICEF2012-92104〉. Viswanathan, S., Rothamer, D., Sakai, S., Hageman, M., Foster, D., Fansler, T., et al. (2015). Effect of particle size distribution on the deep-bed capture efficiency of an exhaust particulate filter. Journal of Engineering for Gas Turbines and Power, 137(10), 11. http://dx.doi.org/10.1115/1.4030098. Viswanathan, S., Rothamer, D. A., Foster, D. E., Fansler, T. D., Zelenyuk, A., Stewart, M. L., et al. (2016). Evolution of deep-bed filtration of engine exhaust particulates with trapped mass. International Journal of Engine Research. http://dx.doi.org/10.1177/1468087416675094. Viswanathan, S., Sakai, S., & Rothamer, D. (2014). Design & evaluation on an Exhaust Filtration Analysis (EFA) system. Paper presented at the SAE technical paper 201401-1558. Detroit, MI. 〈http://dx.doi.org/10.4271/2014-01-1558〉. Viswanathan, S., Sakai, S. S., Hageman, M., Foster, D. E., Fansler, T., Andrie, M., et al. (2014). Effect of particle size distribution on the deep-bed capture efficiency of an exhaust particulate filter. Paper presented at the ASME 2014 internal combustion engine division fall technical conference . 〈http://dx.doi.org/10.1115/1.4030098〉. Wirojsakunchai, E., Kolodziej, C., Yapaulo, R., & Foster, D. (2008). Development of the Diesel Exhaust Filtration Analysis System (DEFA). Paper presented at the SAE technical paper 2008-01-0486. 〈http://dx.doi.org/10.4271/2008-01-0486〉. Zelenyuk, A., Cai, Y., & Imre, D. (2006). From agglomerates of spheres to irregularly shaped particles: Determination of dynamic shape factors from measurements of mobility and vacuum aerodynamic diameters. Aerosol Science and Technology, 40(3), 197–217. http://dx.doi.org/10.1080/02786820500529406. Zelenyuk, A., Reitz, P., Stewart, M., Imre, D., Loeper, P., Adams, C., ... Solomon, A. (2014). Detailed characterization of particulates emitted by pre-commercial singlecylinder gasoline compression ignition engine. Combustion and Flame, 161(8), 2151–2164. http://dx.doi.org/10.1016/j.combustflame.2014.01.011. Zelenyuk, A., Wilson, J., Imre, D., Stewart, M., Muntean, G., Storey, J., ... Parks, J. (2016). Detailed characterization of particulate matter emitted by lean-burn gasoline direct injection engine. International Journal of Engine Research. http://dx.doi.org/10.1177/1468087416675708 1468087416675708. Zelenyuk, A., Yang, J., Choi, E., & Imre, D. (2009). SPLAT II: An aircraft compatible, ultra-sensitive, high precision instrument for in-situ characterization of the size and composition of fine and ultrafine particles. Aerosol Science and Technology, 43(5), 411–424. http://dx.doi.org/10.1080/02786820802709243. Zhao, F., Lai, M. C., & Harrington, D. L. (1999). Automotive spark-ignited direct-injection gasoline engines. Progress in Energy and Combustion Science, 25(5), 437–562. http://dx.doi.org/10.1016/S0360-1285(99)00004-0.

264