Classification of Volatile Engine Particles

Volatile particles cannot be detected at the engine exhaust by an aerosol detector, as they are formed when the exhaust is mixed downstream with the ambient air. The lack of a precise definition of volatile engine particles has been an impediment to engine manufacturers and regulatory agencies involved in the development of effective control strategies. Volatile particles from combustion sources contribute to the atmospheric particulate burden, and this is a critical issue in ongoing research in the areas of air quality and climate change. A new instrument, called a volatile particle separator (VPS), is developed in this work. It utilizes a proprietary microporous metallic membrane to separate particles from vapors. VPS data are used in the development of a two-parameter function to quantitatively classify, for the first time, the volatilization behavior of engine particles. The value of parameter “A” describes the volatilization potential of an aerosol. A nonvolatile particle has a larger A-value than a volatile one. The value of parameter “k,” an effective evaporation energy barrier, is found to be much smaller for small engine particles than for large engine ones. The VPS instrument is not simply a volatile particle remover, as it makes possible the characterization of volatile engine particles in numerical terms.


INTRODUCTION
Particles emitted by combustion engines are composed of a highly complex mixture of organic and inorganic matter.They vary in size, chemical composition, morphology and microstructure.Particles emitted from a modern internal combustion engine consist of condensable and nonvolatile fractions (Cheng et al., 2009).The engine particles are generally ultrafine (Cheng et al., 2008(Cheng et al., , 2009)); i.e., their size is generally smaller than 100 nm, making them more toxic than the larger ones (Donaldson et al., 1998).This is because their large surface area per unit mass enhances their transport capacity for toxic substances such as polyaromatic hydrocarbons.The small size promotes translocation and penetration once they are inhaled into a human lung (Oberdörster and Utell, 2002).
With regard to climate-change, Demirdjian et al. (2007) found that combustion particles consist of two fractions with significantly different microstructures and chemical compositions, making one fraction hydrophobic and the other hydrophilic.The finding suggests that combustion particles from aircraft engines could act as effective cloudcondensation nuclei and thus could directly influence contrail and cirrus formation-important to the atmospheric radiation balance and cloud life cycle.
Vapor-particle partitioning, a key environmental process controlling the complicated behavior of engine particles and environmental particles, is driven by thermal energy.At any given temperature, the vapor-particle partitioning process activates, and the equilibrium concentration of a species in both vapor and particle phases is established.When the temperature changes, a new equilibrium state is established.The cool environmental condition that engine exhaust encounters promotes the condensation of highmolecular-weight (or heavy) species into particles.Some of the condensables will adsorb to nonvolatile soot particles by heterogeneous condensation (Ristimaki et al., 2007); others may form new particles through nucleated condensation (Ronkko et al., 2007).These processes mask the size distribution of the nonvolatile engine particles.Thus the current practice of the Particle Measurement Program (PMP) in evaluating particle concentrations for the Euro 5/6 vehicle emissions legislation of the European Union requires the removal of "volatile components" in the exhaust for measurement of nonvolatile soot particles (Giechaskiel et al., 2010, PMP, 2006).The removal devices are called volatile particle removers (VPRs).
Due to the transient nature of the partition dynamics in the turbulent reactive plume of engine exhaust, representative sampling and accurate measurement of volatile engine particles are major challenges.At present, there is no recommended procedure for sampling volatile particulate matter in the exhaust plume of any combustion engine.
Even the removal of volatile components is problematic for sampling and measurement of nonvolatile soot particles (Swanson and Kittelson, 2010).Inadequate removal of volatile components can lead to biased results (Swanson and Kittelson, 2010) and also create challenges for new engine certification and regulatory agencies in the future.
A number of devices have been developed in the past couple of decades for the study of aerosol volatility.However, most of the devices remain as research tools and have never been widely used as commercial instruments.Commercial thermodenuders have been found to have significant particle loss (low particle transmission) and measurement artifacts (Swanson and Kittelson, 2010).
To improve sampling and measurement of ultrafine (< 100 nm) and nanoscale volatile engine particles, we developed a novel instrument built around the concept of cross-flow filtration, which enables thermal separation of desorbed vapors from nonvolatile particles without the use of an adsorbent.The prototype instrument has been shown to have high-efficiency in laboratory tests (Cheng 2010, Cheng andAllman, 2011).The objectives of the current study are to (1) apply the VPS to real engine particles and (2) use VPS data to quantitatively classify the aerosol volatility of particles emitted from a diesel engine and an aircraft turbine engine.

Engine Operating Conditions and Gaseous Emissions Analyses
The engine used for this study was a four-cylinder, 1.9 L light-duty diesel engine operated by the Oak Ridge National Laboratory (ORNL) Fuels, Engines, and Emissions Research Center (FEERC).Samples were collected during a conventional diesel combustion (CDC) operating mode for the construction of a thermogram.The engine was operated at 1500 RPM and 2.6 bar brake mean effective pressure (BMEP) with a diesel feed rate at 0.5 g/s.Conditioned air was supplied to the engine to maintain a nearly constant air temperature and humidity level.The diesel injection (DI) system, piston geometry, and compression ratio were in manufacturer production form.A National Instrumentsbased control system was used to control the engine in place of the production controller.The injector, engine, and fuel specifications are given in Tables 1, 2, and 3, respectively.
An AVL Smoke meter was used to measure soot content.Gaseous engine emissions were measured using standard analysis techniques.A heated flame ionization detector (FID) The engine is in the middle of the picture in Fig. 1 and is surrounded by the emissions monitoring components.The white structure lying on the floor is the home-built PMP system.The tandem differential mobility analysis (TDMA)-VPS setup is shown in Fig. 2.

VPS Design and Operation
The construction of the VPS was described in detail elsewhere (Cheng and Allman, 2011).We will describe briefly the instrument and its operation in the current paper.The main difference between this prototype instrument and other VPRs, such as thermodenuders and catalytic strippers, is its ability to continuously separate vapors from thermally denuded particles.This minimizes sampling and measurement artifacts that are caused by the presence of volatile species reported in previous studies (e.g., Swanson and Kittelson, 2010).The VPS was found to have very low nonvolatile particle loss (high transmission efficiency) in previous tests (Cheng, 2010;Cheng and Allman, 2011).The transmission efficiency is greater than 95% for particles greater than 20 nm and 99% for 50 nm and larger.
The VPS consists of two interconnected sections; the first section is a straight hollow cylinder made of stainless steel 316, which is enclosed by a radiant heater with a ceramic casing.The second section is the vapor-particle  membrane separator.The temperature of the radiant heater is proportional-integral-derivative (PID) controlled and monitored by three K-type thermocouples positioned close to the entrance, center, and exit of the section.The first section is designed to heat the engine particles.The residence time is greater than 0.3 ms.The section for vapor-particle separation consists of a straight microporous metallic membrane tube (MMT) enclosed by a concentric hollow cylinder.The membrane is a propriety material manufactured by ORNL.For interested readers, a previous study used the membrane for removable of microbes in water purification (Phelps et al., 2008).The nominal pore size used in our study was about ten times smaller than that in the membrane used by Phelps et al. (2008).
Condensable materials are more sensitive to heat than soot.They can be removed or vaporized more easily than soot particles, which are converted directly into the vapor phase through sublimation at high temperature (e.g., 800°C).Since vapor has a higher diffusivity than that of solid particles, volatilized material and condensable gases, including unburned and partially burned fuel species, permeate the membrane and are separated from the particles.At that point, the particles are thermally denuded at a given temperature.Vapors penetrate the membrane and are extracted for removal or gas detection while the nonvolatile (or denuded) particles continue passing straight through the MMT.The opportunity for condensable materials to return to the denuded particles during separation, as reported in the literature (e.g., Swanson and Kittelson, 2010), is substantially reduced if not eliminated entirely.Fig. 3(a The MMT, its enclosure assembly, and the Swagelok fittings coupled to the heating section are shown in Fig. 4. The enclosure assembly is fully insulated with glass fiber material to reduce heat loss.The temperature of the heating section can be varied to investigate aerosol volatility as a function of vaporization temperature.The temperature at the entrance of the membrane section varies with the heating temperature and flow rate.For example, temperature of 75°C was measured at the entrance of the membrane section for a 350°C heating temperature and a 2 L/min sample flow rate.With the insulation jacket, the air temperature profile throughout the MMT section decreases slowly throughout the axial length with the temperature at the exit a few degrees lower than that at the entrance.

Characterization of Volatile Engine Particles with a Thermogram
Since the VPS can accurately elute nonvolatile particles at a given heating temperature, it is possible to deduce the thermal property of engine particles as a function of temperature, which we call a thermogram; therefore, quantitatively classifying the volatile fraction of engine particles without the detailed knowledge of the chemical composition of the engine particles.In the current application, a thermogram for engine particles provides a quantitative means for determining aerosol volatility as a function of temperature.
The fundamental properties that determine a thermogram involve (1) the thermal energy required to drive off condensables and (2) the rates of evaporation of the condensables.The required energy and evaporation rate are dependent on particle size and composition and the thermal conductivity of the gas.The evaporation rate dependence on these properties is shown through the Fuchs, Kelvin, and temperature depression effects (Hinds, 1999).For the complex mixture of particles and gases present in engine exhaust, the thermogram helps to define the nature of the particles as a function of engine conditions and fuel type.Although PMP/VPR systems (Khalek, 2007;Giechaskiel et al., 2008;Matthias et al., 2012) provide a practical means for measuring nonvolatile particles by removing all volatiles or condensables at 350°C, the protocol does not provide a means to characterize volatile engine particles.However, this can be achieved by a thermogram generated using the VPS.
Another advantage of the VPS is that the volatile species driven out from the particles into the vapor phase can be collected for chemical analysis.The thermally conditioned nonvolatile particles pass straight through the membrane tube and can be sampled by an instrument or a filter for physical and/or chemical characterization.This feature enables a means of simultaneous analysis for engine species in both vapor and particulate phase, which could extend the characterization for volatile engine particles.

VPS Evaluation Using TDMA
Monodisperse particles were used to investigate the dependence of VPS transmission and vapor-particle separation efficacy on particle size.There are methods to  generate monodisperse particles; one example is the electrospray technique (Cheng et al., 1995, Cheng, 2004, Park et al., 2008).TDMA, developed in the 1980s (Rader and McMurry, 1986), is a practical technique demonstrated for the production of monodisperse particles and is the technique selected in this work.In our TDMA setup, a nano-DMA (Chen et al., 1996, Hummes et al., 1996) classified (i.e., size-selected) particles from the broadband size distribution.The TDMA setup requires the use of two electrical classifiers (ECs); one is used before the VPS for the generation of monodisperse particles; the other is coupled to a condensation particle counter (CPC) at the exit of the VPS to measure the size distribution.
Fig. 5 illustrates the adopted TDMA configuration in this work.The first EC is equipped with the nano-DMA (TSI model 3085 or DMA1 hereafter).The voltage was manually set on the classifier to the desired particle mobility diameter and was fixed throughout the experiment.The second EC equipped with a long-DMA (TSI model 3081) was connected to an ultrafine CPC (TSI model 3025A) for measurement of particle size distribution.

Thermal Evolution of Diesel Engine Particles under CDC Operation
The thermal behavior of the CDC exhaust particles was investigated using monodisperse samples extracted by DMA1 (see Figs. 6(a), (b), and (c)).Three sizes of particles, 95, 52, and 18 nm, were selected by DMA1 from a broadband engine particle size distribution that peaked at about 52 nm.The geometric standard deviations of the selected particle sizes as shown in the 40°C curves in Fig. 6) were all less than 1.2.
The VPS was programmed to operate at five set temperatures (i.e., 40°C, 100°C, 200°C, 300°C, and 350°C).The temperature in the aerosol stream entering DMA1 was about 40°C, but the first temperature setpoint on the VPS indicated a "no heating" condition corresponding to the original temperature of the aerosol stream.
As shown in Fig. 6(a), the peak size of the 95 nm exhaust particles changed slightly from 40°C to 300°C, and the peak height (after dilution correction) decreased as the heating temperature increased.At 350°C, loss of particles occurred to a substantial extent, and the remaining 95 nm particle population was barely detectable.Previous VPS data from the synthetic particle tests (Cheng and Allman, 2011) indicated that the peak number concentration decreased as the heating temperature increased, but the peak size or location would not change significantly for nonvolatile particles.
Initially, it was suspected that the loss of particles at higher temperature was caused by diffusion.Cheng and Allman (2011) conducted a simple particle penetration test by running sodium chloride particles through the MMT for 2 hours.At the end of the experiment, the MMT was submerged in distilled water and the conductivity of the water was measured.The conductivity showed little change before and after the experiment.Furthermore, the SMPS could not detect a single particle at the vapor exit of the membrane section.If particles were lost and deposited on the membrane in the 2 h continuous experiment, the conductivity would change.The result suggests that particle loss inside the VPS has to be through vaporization rather than diffusion.
The relatively constant peak size at 95 nm (± 2nm) for diesel engine particles with increasing temperature shown in Fig. 6(a) is consistent with our previous observations for nonvolatile particles made of a pure compound (e.g., sodium chloride) (Cheng and Allman, 2011).The decrease of the dilution-corrected concentration value as the VPS temperature increased was also observed for the 52 nm (Fig. 6(b)) and 18-nm (Fig. 6(c)) engine exhaust particles.However, there were significant shifts in peak size toward smaller sizes as temperature increased, as shown in Figs.6(b) and particularly 6(c).This suggests that the 18 and 52 nm particles contain more volatile contents than the 95 nm particles.
As the temperature reached 350°C, all particles were completely vaporized to less than 2% of the baseline concentration at 40°C.For 18 nm particles, the temperature required to reduce the concentration to 2% of the baseline was about 200°C lower than 350°C, indicating that the smaller particles were vaporized much more easily than the larger ones.The results indicate that a large fraction of the small particles probably contained condensable species on a solid core or were in the liquid droplet state as previously reported (Farrell and Van Siclen 2007).
The behavior of the CDC diesel engine particles with respect to the applied thermal energy (or heating temperature) was consistent with that for laboratory-generated particles (Cheng 2010, Cheng andAllman, 2011) and reported in literature (Ristimaki et al., 2007).For nonvolatile particles or large engine particles, the particle number concentrations dramatically decreased as the heating temperature reached a critical point.The number concentration decrease for smaller or volatile particles appears to be continuous as the temperature changed.The rate of particle loss appears to be higher for smaller particles (e.g., 18 nm) than for the larger ones (52 and 95 nm), given the same material (or loosely speaking, the same engine exhaust).

Test of VPS Using T63 Turboshaft Engine-Generated Particles
The VPS was also tested for aircraft engine particles.The test was setup in a configuration described as follows: aircraft exhaust from a turboshaft T63 engine was sampled by a dilution chamber (DC).The diluted exhaust was then sampled and measured using the tandem DMA-VPS system.Fig. 7 describes the DC-VPS setup.The use of DC could change the engine particle population from its condition in the raw exhaust, and it could affect the properties of exhaust particles.However, this is not an issue for our test of VPS because the work reported here is on the performance of the VPS.
Tests of the VPS on the aircraft engine emissions yielded data that are plotted in Fig. 8.The engine was operated at idle condition, and the dilution ratio was 10:1; the number concentration was dilution-corrected.It is known from the experience operating the T63 engine at the Wright-Patterson Air Force Research Laboratory that the volatile fraction is higher in emissions at idle than it is when the engine is operated at cruise or maximum engine power conditions.
The peak location of any of the three particle size distributions shifted to the left (size is reduced) as the heating temperature increased (see Fig. 8(a)).This observation is consistent with those observed in the CDC diesel engine emissions and the synthetic volatile particles.A 7 nm size reduction was clearly observed as the VPS heating temperature increased from the base condition (heating not turned on VPS) to 200°C, as compared to that of 2.6 nm from 200°C to 300°C.The peak height (i.e., the particle number concentration) was reduced as the VPS heating temperature increased.
The size distributions as the temperature change for the same T63 engine operated at a higher power condition, the cruise condition, is shown in Fig. 8(b).The shift in size distribution as a function of temperature for the cruise power condition is in the opposite direction from that in Fig. 8(a) at the idle condition.The increase in the peak height is 10 nm from the unheated condition to 200°C and less than 1 nm from 200°C to 300°C.
The peak size increased as the temperature increased to the cruise condition.If the particles were volatile, we would expect the peak to shift to a smaller size, not to a larger size, as we had seen in the CDC and dioctyl phthalate (DOP) particles (see Fig. 8 in Cheng and Allman, 2011).
The following scenario provides a likely explanation for  the discrepancy.For the cruise condition, both the individual volatile particles and volatile fraction on soot in the total particle population evaporated as the temperature increased.Since these volatile particles were likely to be in the smaller size range, their loss could change the particle population statistics and thus would lead to a shift to the right in the distribution, as shown in Fig. 8(b).
As the heating temperature goes from 200°C to 300°C, most of the volatiles were removed from the particle population, leaving those nonvolatile soot particles.Thus, the peaks of the red and green curves were in the same location as seen in Fig. 8(b), even when the temperature went up another 100°C.It is likely that the "dry" particles emitted by the T63 engine at cruising power were larger than approximately 70 nm and were insensitive to the 100°C increase from 200°C.For the idle condition shown in Fig. 8(a), however, the particles are mostly "wet," meaning that they were bearing volatiles or were themselves evaporable.They could be entirely in a droplet state (similar to that of DOP discussed earlier, for example).Thus, its thermal behavior shown in the curves in Fig. 8(a) was consistent with that of NaCl (shown in Fig. 7 in Cheng and Allman, 2011).

Statistical Analysis of Particle Population as a Function of Temperature
The fractional variation of the evaporated particles is dependent on temperature and particle size as shown in Table 4.As expected, the CDC particles evaporated more as the temperature increased.Also, the evaporated fraction at a given temperature increased as the particle size decreased.Furthermore, as the temperature reaches a critical temperature for a given size, the particles disappeared and became undetectable.For 18 nm particles, this temperature was about 300°C; the temperature was 350°C for both 52 and 95 nm CDC engine particles.Thus, if we followed the PMP criteria, we would conclude that the VPS is promising as a VPR for CDC particles.
Particle statistics as a function of VPS set temperature are shown in Table 5.It is assumed the CDC particles followed a lognormal size distribution.The shift of the 95 nm peak (defined as the mode diameter) was negligible from 25°C to 300°C and had no clear trend (Table 5 and Fig. 6(a)).We assume that the incoming 95 nm CDC particles were mostly soot, which would require a temperature higher than 350°C to vaporize effectively.The data appear to support this hypothesis.
With the 52 nm incoming particles, a consistent shift of the peak to smaller diameters as the temperature increased was observed (see Table 5 and Fig. 6(b)).Similar behavior was observed for the incoming 18 nm CDC engine particles (see Table 5 and Fig. 6(c)).This is also consistent with our previous findings using laboratory-generated particles (Cheng and Allman, 2011).The data at 350°C shown in Table 5 further suggest that the thermal behavior of the CDC engine particles of all three sizes considered are consistent with the data obtained from previous laboratorygenerated particles in that they undergo minor changes until they reach a critical temperature, at which many particles vaporize.In that instance, the population will become extinct; the size and number concentration associated with this change support this conclusion.
It is likely that the larger particles were composed of soot and condensable species that require more thermal energy to vaporize than small particles that consist of mostly condensable or volatile species.Large soot particles have more complex morphology (e.g., fractal structure) and active surface area that could attract condensable molecules strongly.Small particles do not have such microstructure, and it is likely that entire particles might be made of  condensables such as the DOP particles in Cheng and Allman (2011).Thus, the distributions of the 95 nm and 52 nm particles underwent smaller changes than that of the 18 nm particles as the temperature increased until it reached about 350°C (a process similar to that of sodium chloride).
On the other hand, the 18 nm particles could consist of condensable species that were completely vaporized in a process similar to that of a DOP oil droplet, even at 200°C.

Defining Aerosol Volatility Using Thermographic Data from VPS
According to the PMP definition, a diesel engine particle is volatile if it is completely vaporized at 350°C (Crayford et al., 2010).A more quantitative relationship may be needed to improve the operational definition of volatile particles.We found that the VPS data can be useful for quantitatively classifying volatile particles, and we demonstrate their value in that regard in this section.
It is assumed the number of particles vaporized in the VPS is a function of temperature, as shown by the following relation: where N is the number of particles remaining that is measured at the exit of the membrane section, N° is the number of particles entering the VPS, A is a constant (in the same units as that of the number concentration) related to the volatilization potential of the aerosol population, k is the exponent representing the effective evaporation energy barrier of volatile engine particles, and T is temperature.
The units of N and N° in Eq. ( 1) are #/cm 3 , while that for T is °C.A has no units and k has the same unit as that of T. We linearized the equation and used linear regression technique to derive values of A and k.
Table 6 shows the calculated A and k values for the synthetic aerosol particles that we have tested (reported in Cheng and Allman, 2011) as well as the CDC engine particles reported in this paper.To calculate A and k values, the leastsquare regression was performed on the linearized relation shown as follows: where Ln is the natural log operation.It appears that the linearized relation, Eq. ( 2), fits the VPS data well with an R 2 value greater than or equal to 0.80, indicating that the vaporization of particles in the VPS follows the empirical function (Eq.( 1)) reasonably well.A large A value indicate the aerosol has a great potential to evaporate, and a large absolute value of k suggests the aerosol has a high vaporization energy barrier and is difficult to remove by heat.Data for the NaCl aerosols were aggregated for estimating their corresponding A and k value, because the particles were made of single chemical compound.Data for the DOP aerosol were analyzed similarly.However, the data for the CDC engine particles were not aggregated but individually analyzed for each size because of the complex chemical nature of engine particles.The results are shown in Table 6.The value of A for the NaCl particles (nonvolatile) is greater than that for the DOP particles (volatile) suggesting a 67% higher volatilization potential.Furthermore, the k value for the DOP aerosol is larger than that for NaCl indicating a lower evaporation barrier for aerosol DOP particles.The three A-estimates for the three sizes of the CDC particles are all smaller than that of the DOP particles indicating the engine particles have higher volatilization potentials, and the three k-values for the CDC particles also suggest that their evaporation barriers are bounded between those of NaCl and DOP, reflecting the complex chemical nature of the engine particles.Smaller CDC engine particles have much lower evaporation energy barrier than that of the larger ones.
Table 6 shows that for the first time that it might be possible to quantitatively classify volatile engine particles using the thermogram developed from data produced by the VPS instrument.Any engine volatile particles can be characterized by a set of A and k values using the thermographic data taken by the VPS.Thus the set of (A, k) for the same diesel engine running on conventional diesel fuel is expected to be different from the other set from the same engine running bio-diesel, for example.The values would vary across the engine types, fuel types, and operational mode.Therefore, the thermogram could provide a systematic means for quantitatively classifying volatile engine particles.However, several temperature settings are required for constructing a thermogram specific for an aerosol population.This will require an engine to be operated stably for a period of time, which can be difficult during an aircraft emission sampling field campaign.It was unfortunate that no such data were collected for the T63 engine to enable the classification of that particular type of aircraft engine emission, because of the limitation of resources at the time.
The A and k values are fundamental to the material properties that make up the aerosol particles.The VPS thermogram can even be developed on line with further engineering development of the instrument.This capability could eventually provide a rapid quantitative means of defining volatile engine particles.

CONCLUSIONS
Lack of a precise definition of volatile engine particles has been an impediment to engine manufacturers and regulatory agencies involved in the development of an effective control strategy.It is beyond doubt that volatile particles from combustion sources contribute to the atmospheric particulate burden, and the effect of that contribution is a critical issue in the ongoing research in the areas of air quality and climate change.A new instrument, called the volatile particle separator (VPS), utilizes a propriety microporous metallic membrane to separate particles from vapors.Using the VPS data, we developed a two-parameter function to quantitatively classify, for the first time, the volatilization behavior of engine particles.The value of parameter "A" describes the volatilization potential of an aerosol.A nonvolatile particle has a larger A-value than a volatile one.The value of parameter "k," an effective evaporation energy barrier, is found to be much smaller for small engine particles than that for large engine particles.The VPS instrument provides a means beyond just being a volatile particle remover; it enables a numerical definition to characterize volatile engine particles.

Fig. 1 .
Fig. 1.Photo of experimental setup in the National Transportation Research Center (NTRC) at Oak Ridge National Laboratory (ORNL).

Fig. 2 .
Fig. 2. Photo of experimental setup at the ORNL engine laboratory showing the tandem differential mobility analysis (TDMA)-vapor particle separator (VPS) setup toward the left of the figure.
) graphically illustrates the process of membrane separation, Fig. 3(b) shows a scanning electron micrograph of the membrane surface, and Fig. 3(c) is a drawing showing the structure of the cross section of an MMT.

Fig. 3 .
Fig. 3. Cartoon showing working of vapor-particle separation (a), scanning electron microscopic image of the cross-section of the metallic membrane (b), and drawing showing the double-layer granular structure of the membrane (c).

Fig. 4 .
Fig. 4. Assembly of the membrane section of the VPS instrument.

Fig. 5 .
Fig. 5. Experimental setup of TDMA for the VPS testing.

Fig. 6 .
Fig. 6.The mobility particle size distributions of the CDC engine particles as a function of the VPS heating temperature.Monodisperse particle size of 95 nm (a), 52 nm (b), and 18 nm (c).

Fig. 7 .
Fig. 7. Setup of the dilution chamber-VPS for the T63 turboshaft engine experiment at the Wright-Patterson Air Force Research Laboratory.

Fig. 8 .
Fig. 8. Particle size distributions as a function of temperature for idle engine power condition (a) and cruise engine power condition (b).

Table 4 .
Statistical Summary for Particle Number Concentration for CDC Data.

Table 5 .
Statistical Summary of Particle Diameter for CDC Data.

Table 6 .
Aerosol Volatility Defined by VPS.