Filtration Efficiency and Pressure Drop of Miniature Diesel Particulate Filters

Transcription

1 Aerosol Science and Technology ISSN: (Print) (Online) Journal homepage: Filtration Efficiency and Pressure Drop of Miniature Diesel Particulate Filters J. Swanson, W. Watts, D. Kittelson, R. Newman & R. Ziebarth To cite this article: J. Swanson, W. Watts, D. Kittelson, R. Newman & R. Ziebarth (2013) Filtration Efficiency and Pressure Drop of Miniature Diesel Particulate Filters, Aerosol Science and Technology, 47:4, , DOI: / To link to this article: Accepted author version posted online: 04 Jan Published online: 04 Jan Submit your article to this journal Article views: 1127 View related articles Citing articles: 10 View citing articles Full Terms & Conditions of access and use can be found at

2 Aerosol Science and Technology, 47: , 2013 Copyright C American Association for Aerosol Research ISSN: print / online DOI: / Filtration Efficiency and Pressure Drop of Miniature Diesel Particulate Filters J. Swanson, 1,2 W. Watts, 1 D. Kittelson, 1 R. Newman, 3 and R. Ziebarth 3 1 Department of Mechanical Engineering, University of Minnesota, Minneapolis, Minnesota, USA 2 Department of Engineering, University of Cambridge, Cambridge, UK 3 Core R&D, The Dow Chemical Company, Midland, Michigan, USA A method was developed to evaluate miniature diesel particulate filters (DPFs). To validate the performance of the instrumentation and test apparatus, measurements were made using silicon carbide (SiC) and cordierite miniature filters with representative microstructures. Filtration efficiency (FE), the most penetrating particle size (MPPS), and pressure drop were measured for catalyzed and uncatalyzed advanced ceramic material (ACM) acicular mullite and representative commercial filters to determine the impact of substrate morphology, the formation of a soot cake, and the presence of a catalyst coating on filtration properties. FE measurements demonstrated that filter geometry and microstructure significantly influence initial filtration performance. ACM filters had high initial FE and the MPPS near 200 nm. Reduction of the ACM pore size in the absence of a reduction in porosity increased initial FE even more, but its influence on MPPS was not resolvable. The presence of a catalyst and washcoat on the ACM increased the pressure drop but increased initial FE and reduced MPPS to <100 nm. The addition of a washcoat allowed the rapid buildup of a soot cake, which resulted in a more rapid rate of increase in FE compared to uncatalyzed ACM. The similarity in the ACM and cordierite soot cakes after a long loading time is consistent with theory that suggests the formation of the soot cake depends primarily on the Péclet (Pe) number, which is influenced only by macroscopic filter geometry and prevailing test conditions. INTRODUCTION Diesel engines are widely used in automotive, transportation, construction equipment, power generators, marine engines, and off-road applications. In 2007, the US Environmental Protection Agency (EPA) reduced the diesel particulate matter emis- Received 18 March 2012; accepted 11 December The authors thank the anonymous reviewers for providing their keen insights into the filtration behavior of ACM and other suggestions. David Gladis (UMN graduate student) assisted with data analysis. DOW Core R&D provided funding and technical support for this work. Address correspondence to J. Swanson, Department of Mechanical Engineering, University of Minnesota, 111 Church St. SE, Minneapolis, MN 55455, USA. jswanson@me.umn.edu 452 sion (DPM) standard for heavy-duty diesel engines used onroad to g/kwh (0.01 g/bhp-h) (DieselNet 2011). This represents a tenfold reduction over the previous standard. As a consequence, diesel engine manufacturers have made significant improvements in diesel engine technology. To meet the on- and off-road DPM standards, manufacturers may rely on exhaust after treatment devices such as diesel particulate filters (DPFs). Both the US EPA (US EPA 2011) and the California Environmental Protection Agency Air Resources Board (CARB 2011) certify exhaust emission control devices for use in the United States. DPFs are typically wall-flow filters composed of cordierite or silicon carbide (SiC) in which alternate channels are blocked forcing filtration to take place as the exhaust gases pass through the channel walls while the DPM is retained in the filter (Konstandopoulos and Johnson 1989; Konstandopoulos and Papaioannou 2008). Collected DPM is continuously (passive) or periodically (active) removed from the filter by oxidation (regeneration) to prevent increased engine back pressure. Active regeneration relies on external measures such as postinjection of fuel into an upstream oxidation catalyst whereas passive regeneration relies on the direct or indirect use of catalyst technologies. Detailed reviews of emission control technology are found elsewhere (Johnson 2004, 2008, 2010). DPFs are also made from an advanced ceramic material (ACM) known as acicular mullite. The approach used to make ACM is based on a gas-to-solid catalyzed reaction to form acicular mullite by low-temperature decomposition of fluorotopaz (Al 2 SiO 4 F 2 ) (Pyzik et al. 2008). The unique feature of this process is a resulting highly elongated grain structure in the mullite. The formation and growth of grains interlocks the microstructure and result in retained high porosity that is stable to C. The ACM manufacturing process is controlled so that the needle size, porosity, and the pore size distribution can be varied. For example, the needle length can vary from 3 to 50 μm with an aspect ratio of about 20. Flexibility in the manufacturing process enables the production of wall-flow monoliths for different exhaust filtration applications and substrates with high porosity microstructures that are required for catalyzed

3 FE AND dp OF MINI DPFS 453 applications (Pyzik and Li 2005, 2008; Li and Pyzik 2009). ACM DPFs have demonstrated high filtration efficiency (FE), low-pressure drop, high-temperature survivability, and excellent mechanical integrity at a porosity of 60% or higher (Pyzik and Li 2005; Li et al. 2008). The focus of this work is the evaluation of small-scale filters. Many years of study have been devoted to understanding the catalytic performance of emission control systems through laboratory evaluations of small-scale devices using real exhaust (Kalogirou et al. 2007; Tan et al. 2011) and simulated challenge gases/aerosols (Otto et al. 1974; Konstandopoulos et al. 2006; Schrade et al. 2012). The use of small-scale devices is now a standard approach in the catalyst industry. More recent work has extended this idea to the evaluation of particle filtration performance but there are additional obstacles to overcome due to the relative difficulty of particle measurements in high-temperature systems. One example of small-scale evaluation is the use of single channels cut from full-size filters. Results from loading dry, spherical ammonium sulfate particles at room temperature on single, cordierite, and SiC channel walls have shown that depth and cake filtration regimes can be reproduced by a model aerosol (Yang et al. 2009). Similar single-walled DPF studies have showcased this technique as a means for visualizing DPM deposition within the microstructure (Dillon et al. 2007b). On the other hand, single-wall studies conducted with ACM have concluded that experimental errors were too large to allow quantitative conclusions to be drawn regarding the impact of ACM microstructure on pressure drop (Dillon et al. 2007a). Other work has focused on measurements using wafers where a slice of DPF material with a circular filtration area 25.4 mm in diameter is placed in a traditional DPM sampling filter holder (Wirojsakunchai et al. 2008). These experiments have enabled pressure drop models to be precisely verified because conditions are well known and well defined, for example, face velocity profile. Careful measurements of the effect of velocity on pressure drop and cake permeability using this method (Yapaulo et al. 2009, Rakovec et al. 2011) provide further insight into the nature of soot cake formation that is useful for DPF design optimization. The first objective of the present work is to develop an alternative test methodology to simultaneously evaluate the initial and transient FE and pressure drop characteristics of miniature, rectangular, wall-flow DPFs. The rationale for the use of this geometry is its similarity to commercial filtration systems. The use of miniature, multichannel rectangular monolith filters helps to bridge the gap between single-walled DPF studies that provide insight into fundamental processes but are experimentally tenuous and therefore not always able to provide realistic or precise filtration and pressure drop information and traditional tests where full-sized emission control systems are evaluated. The method enables filters to be evaluated rapidly and consistently with independent control of temperature, face velocity, and engine operating condition. Thus, it is useful for the evaluation and screening of prototype DPFs. Simultaneous measurement of catalytic performance and filtration are reported on separately (Swanson et al. 2013). To validate the performance of the instrumentation and test apparatus, measurements were made using miniature SiC and cordierite filters with representative microstructures. Such measurements were necessary to ensure miniature filter results that were quantitatively similar to fullsize DPF results. The second objective is to measure the FE, the most penetrating particle size (MPPS), and the pressure drop for miniature catalyzed and uncatalyzed ACM, SiC, and cordierite filters to determine the impact of substrate morphology, the formation of a soot cake, and the presence of a catalyst coating on filtration properties. Our measurement approach required that semi-volatile particles, which typically form during dilution and cooling, be eliminated before FE was calculated. This ensures that the particles being characterized are the same as the solid particles that are being filtered. Measurements of sizeresolved solid particle FE are required to understand how differences in microstructure influence the MPPS, which represents the particle size that is least influenced by filtration mechanisms (diffusion, impaction, and interception) and its overall effect on FE (Hinds 1999). Transient FE measurements during loading were made using a scanning mobility particle sizer (SMPS) operated with short-scan times. The SMPS is preferred to the use of size spectrometers that rely on unipolar charging due to their tendency to undersize aggregates (Wang et al. 2009), therefore providing less reliable MPPS information. Short SMPS scan times enabled changes in FE and MPPS to be determined during the brief initial DPF loading period. EXPERIMENTAL Miniature Filters Figure 1 shows micrographs of the microstructure of SiC (a) and the cordierite (b) compared to the ACM (c) and (d). Figure 1e shows examples of multichannel cm square prism miniature filters that were contained in the stainless steel (SS) holder shown in Figure 1f. The rectangular SS filter holder is 287 mm 25 mm 25 mm in size including the flange and connectors. Filters were wrapped with an intumescent mat and pressed into the holders to provide a gas-tight fit. A graphite gasket was used to seal the flange bolted to the outlet. Table 1 lists selected physical properties of these filters. Some filters were catalyzed using a standard formulation NO x trap catalyst. The impact of coating on pore size and porosity was not measured but both will decrease with coating (Tsuneyoshi et al. 2011). Commercial SiC DPFs are almost always composed of individually extruded longitudinal segments that can be scaled in size whereas ACM or cordierite would typically have to be cut into rectangular monoliths to replicate the miniaturized sampling apparatus because production pieces are cylindrical or oval. ACM filters described here were extruded into rectangular filters rather than cut from production-sized DPFs.

4 454 J. SWANSON ET AL. TABLE 1 Miniature DPF identification and physical properties. Catalyst loading refers to both the precious metal and washcoat mass; the precious metal mass is similar for catalyst coating 1 and 2. The Péclet (Pe) number is defined in Equation (1) Cell density Wall thickness Filtration Mean pore Catalyst Sample (cell/in. 2 ) (μm) area (m 2 ) Pe size (μm) Porosity loading (g/l) ACM A n/a ACM B <15 65 n/a Cordierite n/a Silicon carbide n/a ACM A, catalyst <15 < ACM A, catalyst <15 < ACM B, catalyst <15 < ACM B is a modified version of the standard ACM A filter that was processed further using a standard method known to reduce the mean pore size while negligibly reducing porosity (<5% change). Wall thicknesses are estimated. Test Apparatus The test engine was a model year 2005, John Deere 4045H, four-cylinder, 4.5 L engine rated at 129 kw (at 2400 rpm). The engine is an EPA Tier 2 certified off-highway engine that is turbocharged and aftercooled with common rail fuel injection. Exhaust gas recirculation was not used and the engine was fueled with 100% ultralow sulfur diesel fuel containing <15 ppm sulfur. All tests were conducted at one steady-state condition (1400 rpm, 250 Nm). Figure 2 shows a schematic of the test apparatus. A sample of exhaust was drawn from the exhaust pipe using a transfer line that was approximately 2 m in length with a 6 mm inner diameter. Sample exhaust was maintained at 200 C in the transfer line. Exhaust flow was controlled using a mass flow controller (MFC) and vacuum pump. The pressure drop across the filter was measured using a wet differential pressure transducer. Samples of the exhaust aerosol were taken upstream and downstream of the filter and diluted 5:1 to reduce concentrations to the measurable range prior to passage through a catalytic stripper (CS). The dilution ratio (DR) is defined as the sum of the sample and dilution air mass flowrates divided by the sample mass flowrate. Exhaust was diluted with a porous tube dilutor (Mikkanen 2001; Dekati 2011; Swanson et al. 2011), where dilution air is introduced under slight pressure into a cylindrical SS chamber through the walls of a porous sintered metal tube. One advantage of this technique is that it allows the use of flow FIG. 1. Micrographs of silicon carbide (a), cordierite (b), acicular mullite (c), a magnified view of acicular mullite showing needle-like structure (d), an example of the square prism sample geometry that is common to all filters (e), and the sample holder (f). Figure 1a is reprinted with permission from Elsevier.

5 FE AND dp OF MINI DPFS 455 experiments. After each experiment, the DPF was regenerated at 600 C in a furnace for a few hours using a heating rate of 200 C/h and a cooling rate of 100 C/h. FIG. 2. Schematic of the miniature DPF test apparatus. Flowrates indicated correspond to a dilution ratio of 5:1. As shown, the flowrate though the filter is 20 L/min, but only 18 L/min is controlled using the exhaust MFC when a downstream sample is extracted for analysis. When an upstream sample is extracted, the exhaust MFC is set to 20 L/min. controllers to measure and control the DR, reducing the need for additional gas analyzers. The dilution air and excess flowrates were controlled using MFCs that have a full-scale range of 0 to 20 L/min, 1% full-scale accuracy, and repeatability of 0.15%. Low flowrate uncertainties are important because the difference in the upstream and downstream flowrates was used to determine the DR in situ. Accurate measurement of DR is required to calculate loaded mass. Gas analyzers and pressure transducers were zeroed and spanned with National Institute of Standards and Technology (NIST) traceable calibration gases and an electronic pressure meter. Flowmeters were factory calibrated and checked frequently using a factory-calibrated Gilian Gilibrator flowmeter. All pressure, temperature, gas concentration, and flowrate data were recorded continuously using National Instruments Lab- VIEW hardware and software at a data acquisition rate of 1 Hz. Each of the filters identified in Table 1 was evaluated in quintuplicate using a nominal exhaust flowrate of 20 standard L/min and an exhaust temperature of 200 C. This corresponds to a space velocity of 72,000/h. The actual flowrate was slightly higher (<5%) than the nominal flowrate because it was not corrected for the unknown fraction of water vapor removed by the condensate trap. Experimental results are shown as quantity versus time or loaded mass, depending on the context. Loading time was varied with each experiment and filter formulation as required to capture the full range of the characteristic response. For example, all filters reached >99% FE within 10 min for a loading rate of 0.14 mg/min, corresponding to 1.4 mg loaded mass. This illustrates the need for fast SMPS scans to determine FE and MPPS. On the other hand, pressure drop measurements required more time and this time varied with DPF formulation because the characteristic filtration regimes took more or less time to develop. Typically, 1 h or longer was required for these Aerosol Instrumentation Catalytic Stripper After the sample was diluted, it passed through a CS to remove semi-volatile material (Abdul-Khalek and Kittelson 1995; Swanson and Kittelson 2010). This enabled measurement of the solid particle number concentration and size distribution. The principle of operation of the CS is to evaporate particle phase semi-volatile material and to remove the resulting gas phase compounds by oxidation. Inorganic compounds such as sulfate are stored by a sulfur trap. The CS was heated to 300 C. The principal mechanisms by which particles are lost in the CS are thermophoresis and diffusion. Penetration was approximately 70% for particles above 100 nm (where most of the mass is found) and decreased to about 25% at 10 nm due to diffusion losses. Mass-based calculations, like loaded mass that require an absolute number, were corrected for losses in the CS. Particle Sizing and Counting During loading, a TSI SMPS, consisting of a TSI 3080L Electrostatic Classifier with a long differential mobility analyzer (DMA) and a 3010 Condensation Particle Counter (CPC), was used to determine the solid particle number size distribution. The SMPS was configured to cover the size range of nm over a period of 45 s using a 30 s upscan and 15 s downscan and a 15 s interval between scans. This relatively fast scan time is consistent with recent studies that demonstrate upscans of even 5 or 10 s are possible with fast response CPCs (Erickson and Quant 2011). The DMA was operated using a 4 L/min sheath airflow and 1 L/min aerosol flow. Data were processed in 16 channels per decade mode. This configuration reduces particle size resolution but increases the number of particles in each channel, improving counting statistics. Total number concentration was calculated by integrating the size distribution and number FE was calculated as (1 number upstream/number downstream) 100. To estimate mass concentration from a number distribution, effective particle density must be known. The effective density inversion described by Park et al. (2003) was used to estimate the total mass concentration of solid particles upstream of the DPF. Total loaded mass was calculated by multiplying mass concentration by flowrate through the DPF and the time sampled. The upstream mass concentration was measured before and after each DPF evaluation, and the average was used for calculations. The maximum difference observed in before and after measurements was 10%. RESULTS The test condition of 1400 rpm and 250 Nm load produced a particle size distribution that was lognormally distributed with a number concentration of part/cm 3 withamean

6 456 J. SWANSON ET AL. the transient filtration stage and the higher Pe during loading. Finally, soot cake filtration occurs when filtration and pressure drop are primarily due to the soot aggregates that have accumulated on the surface. In this regime, there is a linear relationship between pressure drop and accumulated mass. The relationship between soot cake formation and filter microstructure is examined further using results from prior modeling efforts. The pressure drop of the soot cake depends on its geometric structure that includes the soot packing density, the cake thickness, and the permeability (Konstandopoulos and Johnson 1989; Konstandopoulos et al. 2000, 2002). The formation of the soot cake is influenced by the Péclet (Pe) number that is shown in Equation (1): FIG. 3. Raw exhaust solid particle number and mass distributions. diameter of 60 nm and a geometric standard deviation of 1.6, as shown in Figure 3. The calculated mass concentration was 6.8 ± 1mg/m 3. This value is consistent with the elemental carbon mass concentration (6.76 ± 0.82 mg/m 3 ) measured using the same engine and test condition that was reported in previous work (Kittelson et al. 2010), which suggests a representative exhaust sample is used and that the effective density method used to calculate solid particle mass is robust. Filter Loading Curves Miniature Apparatus Validation Typical commercial DPFs exhibit characteristic filtration regimes as filtration occurs first in the deep-bed or pores of the DPF and then transitions into a surface or soot cake filtration phenomenon (Tien 1989; Murtagh et al. 1994; Suresh et al. 2000; Wirojsakunchai et al. 2005). During the pore filling, regime filtration occurs on the internal porous wall structure of the substrate. As a result, the pressure drop increases slowly as the area becomes restricted. As the pores begin to close the pressure drop rapidly increases, resulting in a transition from DPF filtering to soot cake filtering once the pores are fully closed. Figure 4 shows that these regimes are also evident in the miniature cordierite and SiC filter loading curves. Previous results suggest the depth filtration capacity of a typical DPF porous wall is between 0.5 and 2 g/m 2 of soot (Konstandopoulos and Papaioannou 2008). Data in Figure 4 suggest that the transition from pore filling to soot cake filtration occurs near 5 mg of loaded mass for the cordierite and 3 mgforthesic,corresponding to depth capacities of 0.25 g/m 2 and 0.1 g/m 2, respectively. These values are slightly lower than reported in the literature as typical. A possible explanation is that loading at very low Pe yields very high cake porosities, which reduce the soot cake mass in a given volume. The relatively higher capacity of cordierite than SiC is consistent with microstructure images that reveal that cordierite contains very large pores on the surface, which are last to be completely filled, thereby prolonging Pe = u o D pp D, [1] where u o is filtration velocity, D pp is primary particle size (mode size measured to be 25 nm), and D is aggregate particle diffusion coefficient, which is calculated based on the geometric mean of the size distribution. In these experiments, the difference between ACM, cordierite, and SiC filters is cell density. SiC s higher cell density reduces the filtration velocity by 30% and thus the Pe (as shown in Table 1). During high Pe conditions (Pe > 0.8), packing density, porosity, and permeability are at their asymptotic values (Konstandopoulos et al. 2002). On the other hand, soot cake porosity varies from 0.98 to 0.93 for Pe between 0.3 and 0.8. Thus, in the very low Pe regime of these experiments, packing density, porosity, and permeability are strong functions of Pe. Packing density and permeability are computed from porosity as shown in Equations (2) and (3): ρ soot = ρ solid (1 ε), [2] FIG. 4. Cordierite and SiC DPF pressure drop loading curves showing the characteristic loading regimes. The loaded mass of 8 mg corresponds to a sampling time of 1 h, a loaded area concentration of 0.4 g/m 2 (cordierite) and 0.29 g/m 2 (SiC), and a loaded volume concentration of 0.28 g/l.

7 FE AND dp OF MINI DPFS 457 k soot = f (ε) D 2 pp C c, [3] where ρ soot is packing density, ρ solid is bulk density, ε is porosity, k soot is permeability, f(ε) is Kuwabara hydrodynamic function (Konstandopoulos and Johnson 1989), and C c is the slip correction factor. Soot cake porosity is estimated based on Pe (Konstandopoulos et al. 2002). The primary contribution of the soot cake layer to the total pressure drop is shown in Equation (4): P = μ Q [ 2 V (α + w s) 2 1 ln 2 k soot ( )] α α 2 w [4] where μ is exhaust viscosity, Q is flowrate, V is filter volume, α is substrate cell length, w s is wall thickness, and w is soot cake thickness. Equation (4) is used to compute the slope of the soot cake pressure drops in the linear loading regime (between 30 and 35 mg loaded mass or 1.6 g/m 2 ) to examine the trends observed between ACM (Pe = 0.36, porosity = 0.96), cordierite (Pe = 0.36, porosity = 0.96), and SiC (Pe = 0.26, porosity = 0.98). These results are compared with experimental measurements in Figure 5. Equivalency of the experimental ACM and cordierite pressure drop slopes suggests that the soot cakes are nearly identical even though the underlying substrate microstructure is quite different. On the other hand, the SiC soot cake pressure drop slope is nearly 50% lower than ACM. The modeling result confirms that this is due to the difference in the soot cake microstructure, which can be described by the loading Pe. However, only in the low Pe regime are differences in ACM (or cordierite) and SiC slopes predicted due to the asymptotic nature of the parameters that describe soot cake microstructure. Overall, the similarity in representative miniature filter loading characteristics and theoretical calculations further validates the FIG. 5. Slope of the soot cake in the linear loading regime (ratio of pressure drop to loaded mass) for ACM, cordierite, and SiC compared to modeling calculations. FIG. 6. ACM A pressure drop results with dashed straight lines drawn to aid the eye. The loaded mass of 40 mg corresponds to a sampling time of 5 h, 2g/m 2, and 1.4 g/l. use of the miniature substrates to obtain representative results for ACM performance. ACM Loading Figure 6 shows the average of ACM A loading measurements. The repeatability of these evaluations, which include effects of the multiple regenerations, is very good. For example, at 3 mg of loaded mass the average ACM pressure drop is 490 Pa with a standard deviation of 30 Pa, resulting in a coefficient of variation of 6%. Figure 6 illustrates that the nature of the soot loading and cake formation on ACM is different from that with cordierite and SiC since it first forms in a lower porosity upper region of the wall and is supported by protruding needles. This unique geometry gives rise to two distinct loading regimes. The first regime shows that the transition to soot cake filtration occurs near 8 mgor 0.4 g/m 2. This is consistent with the idea that the highly porous ACM walls can retain more soot than the lower porosity filters, enabling more depth filtration. The open structure and graded density of the ACM discourage pore bridging and the formation of soot islands that block openings. After this brief linear regime of pressure increase, the loading curve inflects into a second, steeper curve due to a decrease in soot layer permeability caused by depth penetration of particles within the incipient cake growing off the needles, as well as deposition on top of the soot layer. In this second regime, the properties of the soot cake are now mostly independent of microstructure and can be readily modeled using Pe, as described by Equations (1) (5). The initial pressure drops of uncatalyzed ACM A and B are virtually identical. However, after loading 2 mg of soot, the uncatalyzed ACM B pressure drop was 7% lower than uncatalyzed ACM A. A slightly higher pressure drop was expected with the narrower pore size (Mizutani et al. 2007). However, this difference of 7% is close to the sample variability of 6%.

8 458 J. SWANSON ET AL. in the capacity for depth filtration. This is consistent with other results that show the addition of a washcoat is known to narrow and block pores in SiC filters (Tsuneyoshi et al. 2011). Results for ACM A catalyst 2 and ACM B catalyst 2 (same catalyst loading, different pore size) indicate that the pressure drop for catalyzed ACM A is 17% lower at 2 mg loaded soot mass than catalyzed ACM B. Uncatalyzed ACM B (smaller mean pore size) is a nominally better bare filter but clearly is not as good when combined with a catalyst, compared with ACM A. Thus, some ACM microstructures are more favorable for catalyst loading, although there are also FE effects to consider. FIG. 7. Summary of pressure drop results for catalyzed and uncatalyzed ACM. The loaded mass of 2 mg corresponds to a sampling time of 0.25 h, 0.1 g/m 2, and0.07g/l. Influence of Catalyst Coating on Loading Figure 7 shows the results from the ACM filters with and without catalyst coatings. These results indicate that both loaded catalyst mass and substrate porosity impact soot loading capacity. ACM A catalysts 1 and 2 have the same base substrate properties, but catalyst 2 has 60% more catalyst loading as shown in Table 1. The increased catalyst loading qualitatively and quantitatively influences the loading behavior. For example, there is evidence of a transition between depth filtration to surface filtration for ACM A catalyst 1 near 1.2 mg of loaded mass. These results contrast slightly with those of ACM A catalyst 2 (same bare filter, higher catalyst loading), where the initial pore filling regime is much less evident and the transition to soot cake filtration occurs very rapidly. The depth capacity of ACM A catalyst 1 is 0.06 g/m 2 compared to 0.4 g/m 2 for the uncatalyzed ACM. This suggests that the additional washcoat significantly altered the substrate microstructure, which results in a decrease Total Number Filtration Efficiency Figure 8 shows the average number-based FE for each miniature filter as a function of time. Error bars are calculated as the standard deviation of quintuplicate tests. All filters exhibited an increase in number-based FE as they loaded, reaching 99% after 10 min, but there were clear differences between the ACM, catalyzed ACM, SiC, and cordierite filters. Uncatalyzed ACM A and ACM B filters were >95% efficient even without loading and their FE very gradually increases to >99% after 10 min. Close inspection reveals that the initial FE of ACM B is higher than ACM A, as generally expected from the smaller pore size (Mizutani et al. 2007). These results compare favorably to the uncatalyzed SiC and cordierite filters, which show relatively low initial FE of 60 and 80%, respectively. However, the rate of FE increase was much higher for both SiC and cordierite filters in comparison to ACM. The rapid increase demonstrates that in general even a very small amount of soot deposition can significantly aid filtration. Additionally, because ACM has a higher porosity the early stage of soot loading is not as useful from a filtration perspective compared to cordierite. Catalyzed ACM FE performance is different from the uncatalyzed ACM filters in two ways. First, unlike the regular ACM the initial FE is quite low, near 75%. Second, the rate of FE increase is much higher for the filters with catalyst. Higher catalyst loadings ( catalyst 2 ) are associated with the fastest rates FIG. 8. Number-based FE for ACM, catalyst coated ACM, SiC, and cordierite. The loading time of 9 min corresponds to a loaded mass of 1.25 mg, g/m 2 (ACM/cordierite) and g/m 2 (SiC), and 0.04 g/l.

9 FE AND dp OF MINI DPFS 459 FIG. 9. Size-resolved FE for ACM B, SiC, and ACM B with catalyst 2. of FE increase. One explanation for the rapid increase in the catalyzed ACM FE is the filtration benefit of the soot cake that forms more readily for the catalyzed filters. This is similar to the soot cake hypothesis used to explain the SiC and cordierite results. However, it is not yet clear why the initial catalyzed ACM FE is very low compared to bare ACM. Size-Resolved Filtration Efficiency Figure 9 shows the averaged, size-resolved FE curves for uncatalyzed ACM B, SiC, and ACM B with catalyst 2. Each filter exhibits high FE at small and large particle sizes with a minimum FE at an intermediate size that is referred to as the MPPS. Total FE is linked to the MPPS and the challenge particle size distribution. The FE of filters with a characteristic MPPS near the geometric mean particle size of the particles in raw FIG. 10. Most penetrating particle size for uncatalyzed and catalyzed ACM, SiC, and cordierite filters. The MPPS was very similar for ACM A and ACM B, so these results were averaged and shown with standard deviation error bars. Loading statistics are similar to those in Figure 8. exhaust (60 nm) will be lower compared to those with a larger MPPS. Figure 10 shows that the initial MPPS depends on DPF formulation and that for all DPFs, the MPPS decreases with loading time. The initial MPPS depends on microstructure and prevailing conditions. The largest initial MPPS is associated with SiC, which experienced the lowest filtration velocity, thus favoring small particle capture due to diffusion. On the other hand, the catalyzed filters have a smaller initial MPPS than uncatalyzed ACM of the same microstructure. The smaller MPPS is a result of the smaller mean pore size of the catalyzed DPF microstructure that leads to higher interception efficiencies. Local filtration velocities are higher due to narrower pores, which reduce the collection efficiency across the entire size range of interest and reduced diffusion capture, also shift the MPPS toward lower sizes. Thus, the addition of a washcoat on bare ACM reduces the initial FE because higher velocities result in lower efficiencies for all particle sizes and the MPPS shifts toward the peak of the challenge exhaust aerosol size distribution. For highly loaded filters, the soot cake is primarily responsible for filtration. Earlier analysis suggested similarity in the ACM and cordierite soot cake microstructures. This would imply similar MPPS values. Unfortunately in these experiments, FE is so high after just a few minutes, the MPPS is not resolvable and curves are truncated in Figure 10 after the MPPS is no longer discernible. MPPS curves appear to be approaching nm but this is not certain. Additionally, the soot cake MPPS for the SiC should be different from that of ACM and cordierite because of the difference in microstructure properties of the soot cake (as a result of loading at different Pe), but again the FE increases too quickly for that measurement to be possible. CONCLUSIONS A test methodology was developed that allowed the measurement of the pressure drop and solid particle size-resolved FE of

10 460 J. SWANSON ET AL. miniature filters in near real-time. The method enabled miniature filters to be screened quickly for filtration performance, which may offer economic advantages over testing full-size DPFs. The impact of loading on pressure drop was found to be consistent with well-tested theories developed to describe full-size filtration systems (Konstandopoulos and Johnson 1989). For example, while ACM and cordierite microstructures were very different, the pressure drops of the soot cakes that formed on their surfaces were virtually identical. This was consistent with previous findings that formation of the soot cake depends primarily on Pe, which is influenced only by macroscopic DPF geometry and test conditions. Use of the SMPS in a relatively fast scanning mode (30 s up scans) enabled size-resolved FE to be characterized up to 99.9%. FE results shown were very repeatable (total number relative standard deviation <10%) for filters that were regenerated five times. Additional insights into the pressure drop and filtration behavior of miniature catalyzed and uncatalyzed ACM, SiC, and cordierite filters were gained. Results showed that the graded density of uncatalyzed ACM gave rise to two distinct loading regimes. These regimes were observable only during pressure drop measurements because FE reached nominally 100% even before the transition into the first regime. Pressure drop loading curves were used to estimate depth capacities, which were higher for the highly porous uncatalyzed ACM compared to SiC and cordierite filters. At the same time, the high porosity ACM microstructure yielded higher initial FE than the lower porosity microstructures, which differs from conventional wisdom. Thus, porosity cannot be used as the only metric for predicting the FE of new microstructures for all cases. The total number-based FE is related to the MPPS, which depends on the macro (such as wall thickness and cell density) and microscopic (such as microstructure, the presence of catalyst and/or soot cake) geometrical features of the filter. Current filtration theory has not advanced to the point where MPPS can be predicted apriori. The SiC filter had the largest initial MPPS of 400 nm, while uncatalyzed ACM filters had initial MPPSs near 200 nm. For ACM filters, the catalyst coating reduced the MPPS to <100 nm. The coating adversely affected initial FE due to a change in the parameters that effect filtration (mainly increased velocities due to pore restriction) and as a result of the relationship between the MPPS and the challenge size distribution. On the other hand, the washcoat enabled the rapid buildup of a soot cake, compared to that of uncatalyzed ACM, and the soot cake is a very efficient particle filter. REFERENCES Abdul-Khalek, I. S., and Kittelson, D. B. (1995). Real Time Measurement of Volatile and Solid Exhaust Particles Using a Catalytic Stripper. SAE Tech. Paper Ser., California Environmental Protection Agency Air Resources Board (CARB). (2011). Verification Procedure -Currently Verified. diesel/verdev/vt/cvt.htm. Dekati. (2011). FPS-4000 Dekati Fine Particle Sampler. fps Dieselnet. (2011). Emission Standards Summary of Worldwide Diesel Emission Standards. Dillon, H., Maupin, G., Carlson, S., Saenz, N., and Gallant, T. (2007a). Visualization Techniques for Single Channel DPF Systems. SAE Tech. Paper Ser., Dillon, H. E., Stewart, M. L., Maupin, G. D., Gallant, T. R., Li, C., Mao, F. H., et al. (2007b). Optimizing the Advanced Ceramic Material for Diesel Particulate Filter Applications. SAE Tech. Paper Ser., Erickson, K., and Quant, F. (2011). Investigation of Fast Scanning SMPS Measurements: 16s and Below. Proceedings of 30th AAAR Conference, October 3 7, Orlando, Florida, USA. Hinds, C. W. (1999). Aerosol Technology, Properties, Behavior, and Measurement of Airborne Particles. John Wiley & Sons, New York. Johnson, T. V. (2004). Diesel Emission Control Technology 2003 in Review. SAE Tech. Paper Ser., Johnson, T. V. (2008). Diesel Engine Emissions and Their Control. Platinum Met. Rev., 52: Johnson, T. V. (2010). SAE 2009 World Congress Key Developments in Diesel Emission Control and Catalysts. Platinum Met. Rev., 54: Kalogirou, M., Katsaounis, D., Koltsakis, G., and Samaras, Z. (2007). Measurements of Diesel Soot Oxidation Kinetics in an Isothermal Flow Reactor Catalytic Effects Using Pt Based Coatings. Top. Catal., 42 43: Kittelson, D. B., Watts, W. F., Johnson, J. P., and Ragatz, A. C. (2010). New Method for the Real-Time Measurement of Diesel Aerosol. Contract Final Report for NIOSH Grant no., R01 OH Available online at: Konstandopoulos, A. G., and Johnson, J. H. (1989). Wall-flow Diesel Particulate Filters-Their Pressure Drop and Collection Efficiency. SAE Tech. Paper Ser., Konstandopoulos, A. G., Kostoglou, M., Skaperdas, E., Papaioannou, E., Zarvalis, D., and Kladopoulou, E. (2000). Fundamental Studies of Diesel Particulate Filters: Transient Loading, Regeneration and Aging. SAE Tech. Paper Ser., Konstandopoulos, A. G., and Papaioannou, E. (2008). Update on the Science and Technology of Diesel Particulate Filters. KONA Part. Powder, 26: Konstandopoulos, A. G., Skaperdas, E., and Masoudi, M. (2002). Microstructural Properties of Soot Deposits in Diesel Particulate. SAE Tech. Paper Ser., Konstandopoulos, A., Zarvalis, D., Kladopoulou, E., and Dolios, I. (2006). A Multi-Reactor Assembly for Screening of Diesel Particulate Filters. SAE Tech. Paper Ser., Li, C. G., Koelman, H., Ramanathan, R., Baretzky, U., Forbriger, G., and Meunier, T. (2008). Particulate Filter Design for High Performance Diesel Engine Application. SAE Tech. Paper Ser., Li, C. G., and Pyzik, A. J. (2009). Application of Porous Acicular Mullite for Filtration of Diesel Nano Particulates, in Developments in Porous, Biological and Geopolymer Ceramics: Ceramic Engineering and Science Proceedings, M. Brito, E. Case, W. M. Kriven, J. Salem, and D. Zhu, eds., John Wiley & Sons, Hoboken, USA, p. 27. Mikkanen, P. (2001). Characterization of Exhaust Particulate Emissions from Road Vehicles Deliverable 4: Prototype Dilution Sampling System. Particulates project sponsored by: European Commission Directorate General Transport and Environment. Particulates D4.pdf. Mizutani, T., Kaneda, A., Ichikawa, S., Miyairi, Y., Ohara, E., Takahashi, A., et al. (2007). Filtration Behavior of Diesel Particulate Filters (2). SAE Tech. Paper Ser., Murtagh, M. J., Sherwood, D. L., and Socha, L. S., Jr. (1994). Development of Diesel Particulate Filter Composition and Its Effect on Thermal Durability and Filtration Performance. SAE Tech. Paper Ser., Otto, K., Betta, R. A., and Yao, H. C. (1974). A Laboratory Method for the Simulation of Automobile Exhaust and Studies of Catalyst Poisoning. J. Air Pollut. Contr. Assoc., 24(6):

11 FE AND dp OF MINI DPFS 461 Park, K., Cao, F., Kittelson, D. B., and McMurry, P. (2003). Relationship Between Particle Mass and Mobility for Diesel Exhaust Particles. Environ. Sci. Technol., 37: Pyzik, A. J., and Li, C. G. (2005). New Design of a Ceramic Filter for Diesel Emission Control Applications. Int. J. Appl. Ceram. Technol., 2: Pyzik, A. J., Todd, C., and Han, S. C. (2008). Formation Mechanism and Microstructure Development in Acicular Mullite Ceramics Fabricated by Controlled Decomposition of Fluorotopaz. J. Eur. Ceram. Soc.,28: Rakovec, N., Viswanathan, S., and Foster, D. (2011). Micro-scale Study of DPF Permeability as a Function of PM Loading. SAE Tech. Paper Ser., Schrade, F., Brammer, M., Schaeffner, J., Langeheinecke, K., and Kraemer, L. (2012). Physico-Chemical Modeling of an Integrated SCR on DPF (SCR/DPF) System. SAE Int. J. Engines, 5(3): Suresh, A., Khan, A., and Johnson, J. H. (2000). Pressure Drop and Permeability of Clean and Particulate Loaded Filters. SAE Tech. Paper Ser., Swanson, J., and Kittelson, D. (2010). Evaluation of Thermal Denuder and Catalytic Stripper Methods for Solid Particle Measurements. J. Aerosol Sci., 41: Swanson, J. J., Watts, W., and Kittelson. D. B. (2011). Diesel Exhaust Aerosol Measurements Using Air-Ejector and Porous Wall Dilution Techniques. SAE Tech. Paper Ser., Swanson, J. J., Kittelson, D., Watts, W., Newman, R., and Ziebarth, R. (2013). Simultaneous Reduction of Particulate Matter and NO x Emissions using 4-Way Catalyzed Filtration Systems. Submitted. Tan, J., Solbrig, C., and Schmieg, S. (2011). The Development of Advanced 2-Way SCR/DPF Systems to Meet Future Heavy-Duty Diesel Emissions. SAE Tech. Paper Ser., Tien, C. (1989). Granular Filtration of Aerosols and Hydrosols. Butterworths, New York. Tsuneyoshi, K., Takagi, O., and Yamamoto, K. (2011). Effects of Washcoat on Initial PM Filtration Efficiency and Pressure Drop in SiC DPF. SAE Tech. Paper Ser., U.S. Environmental Protection Agency. (2011). National Clean Diesel Campaign (NCDC). Verification Verified Technologies List. cleandiesel/verification/verif-list.htm. Wang, X., Grose, M., Caldow, R., Swanson, J., Watts, W., and Kittelson (2009). Improvement of Engine Exhaust Particle Sizer Spectrometer (EEPS) for Engine Emissions Measurement. Proceedings of 28th AAAR Conference, October 26 30, Minneapolis, MN, USA. Wirojsakunchai, E., Kolodziej, C., Yapaulo, R., and Foster, D. E. (2008). Development of the Diesel Exhaust Filtration Analysis System (DEFA). SAE Tech. Paper Ser., Wirojsakunchai, E., Schroeder, E., Kolodziej, C., Foster, D., Schmidt, N., Root, T., et al. (2007). Detailed Diesel Exhaust Particulate Characterization and Real-Time DPF Filtration Efficiency Measurements During PM Filling Process. SAE Tech. Paper Ser., Yang, J., Stewart, M., Maupin, G. D., Herling, D. R., and Zelenyuk, A. (2009). Single Wall Diesel Particulate Filter (DPF) Filtration Efficiency Studies Using Laboratory Generated Particles. Chem. Eng. Sci., 64(8): Yapaulo, R. A., Wirojsakunchai, E., Orita, T., Foster, D. E., Akard, M., Walker, L. R., et al. (2009). Impact of Filtration Velocities and Particulate Matter Characteristics on Diesel Particulate Filter Wall Loading. Intl. J. Engine Res., 10(5):