Evaluation of Multichannel Annular Denuders for a Newly Developed Ultrafine Particle Sampling System

Laboratory studies evaluated the gas collection efficiencies and particle transmission efficiencies of 1-, 3-, 5-, and 8-channel denuders under simulated atmospheric conditions with the goal of combining an annular denuder–filter pack (AD-FP) system with a newly developed inertial classification. The multichannel denuders having large surface areas were considered for this AD-FP system since the inertial classification system operates at a flow rate of 40 L/min. Gas collection efficiency was measured with phosphorous acid–coated denuders targeted on ammonia gas. Particle transmission efficiency was measured for ammonium sulfate particles. The denuder-coating procedure and the method for extracting the multichannel denuders are described in detail. Gas collection efficiencies for the 1-, 3-, 5-, and 8-channel denuders were 96.3, 98.6, 98.2, and 98.3% (1-ch < 5-ch, 8-ch, and 3ch), respectively. The extraction efficiency was greater than 96% for all but the 8-channel denuder (92%). The particle transmission efficiencies for the 1-, 3-, 5-, and 8-channel denuders at 40 L/min were 97.3, 97.4, 94.7, and 93.4% (8-ch < 5-ch < 1-ch and 3-ch), respectively. These results indicate that a 1or 3-channel denuder can be effectively used in this AD-FP system at a flow rate of 40 L/min.


INTRODUCTION
* Corresponding author.Tel./Fax: +81-48-858-9542 E-mail address: kseki@mail.saitamau.ac.jp The formation and growth of new aerosol particles is a growing interest due to their climatic and health effects.Atmospheric formation from gaseous precursors is a significant source of new aerosol particles into the global atmosphere (Kulmala et al., 2004).
Several studies suggest that ultrafine particles (< 0.1 μm in diameter; UFPs) disproportionately induce oxidative stress in cells and are more toxic than larger particles of similar composition (Li et al., 2003;Nel, 2005;Nel et al., 2006).Although UFPs do not contribute large quantities to PM 2.5 or PM 10 mass, they dominate the number concentration and most of the surface area (Chow and Watson, 2007).Atmospheric oxidation of gasphase primary exhaust species can produce low-vapor-pressure compounds that readily condense onto existing particles and produce secondary mass.Particle composition and size can also evolve due to interaction and reaction of gas-and liquid-or solid-phase species at the particle surface or in the bulk solution, as well as through coagulation of existing particles (Moore et al., 2007).Therefore, measurement of gaseous compounds participating in the nucleation and growth of particles and determination of the chemical composition of nucleated particles, as well as their other properties, are very important to our understanding of the nucleation process itself.The use of annular denuders is a promising approach for simultaneously sampling gaseous and particulate-phase aerosols.Since the appearance of the first article on annular denuders (Possanzini et al. 1983), this device has been increasingly used for ambient air monitoring.The attractiveness of annular denuders comes from the order of magnitude gain in gas collection efficiency for a given length/flow rate ratio (Ye et al., 1991).Particles, which diffuse about 1000 times more slowly than gaseous compounds (Febo et al. 1999), are theoretically not trapped on the tube walls but on a filter downstream from the denuder.
The combination of an annular denuder and a filter pack system (AD-FP) has been used at flow rates of 10 or 16.7 L/min in a cyclone system with a 2.5 μm cut-point diameter (Possanzini et al., 1983;Ye et al., 1991;Baron and Willeke, 2005), to collect both particulate and gaseous phases.However, to obtain sufficient quantitative accuracy and high time resolution, the sampling flow rate must be relatively high when the method is applied to UFPs or nanoparticles.Recently, a new technique for the inertial classification of UFPs with fibrous filters was developed (Otani et al., 2007).With this new technique, pressure drop within the sampling system is lower than existing sampling systems such as low pressure impactor (LPI), and the technique can be used to separate particles smaller than 50 nm or 100 nm.Furthermore, particle masses sufficient for chemical analysis can be collected at a high sampling flow rate (40 L/min), and the composition changes in particles can be measured with high time resolution.
An annular denuder consists of two or more coaxial cylinders so that air is forced to pass through the annular space.An annular denuder can achieve the highest possible efficiency and capacity.The sampling behaviour of a denuder is characterized by the diffusion of gases to the tube walls, which are considered to act as perfect sinks, from a laminar air stream.The condition for a laminar flow is a Reynolds number (Re) < 2000 (Possanzini et al., 1983).However, the flow regime should be considered for each multichannel denuder to ensure that the flow is laminar at 40 L/min and if the annular denuder technique is to be applied to the newly developed UFPs sampling system.The calculated Reynolds number for a multichannel denuder was all below 2000 at 40 L/min with given geometry of annular denuders in this study.Therefore the diffusion of gases to the tube walls can be considered to act as a perfect sink from a laminar air stream.Although many researchers are currently using multichannel annular denuders, particle losses, gas collection efficiency, the denuder-coating procedure, and the sample extraction method for multichannel denuders have not been studied in detail.Therefore, the particle losses and gas collection efficiencies for multichannel denuders, along with the denuder-coating procedure and extraction method were also studied and introduced here.Furthermore, the adverse health effect of UFPs is much higher than that of fine particles due to their large surface area with higher particle number concentration.Thus, it is very important to evaluate the particle transmission efficiency in denuders to analyze gaseous and particulate atmospheric UFPs in field sampling since the loss of smaller particles (< 0.1 μm) could be greater than that of larger particles (Ye et al., 1991;Roussel et al., 2004).
In this study, the possibilities of AD-FP system in combination with the newly developed inertial classification system were estimated.Because the AD-FP system was designed to be combined with a cyclone system operated at either 10 or 16.7 L/min, it was necessary to assess the gas collection efficiency and particle loss at 40 L/min within a 1-channel annular denuder.Also, multichannel denuders (3-, 5-, and 8channel) were considered for the high sampling flow rate since they have much larger surface areas compared to a 1-channel denuder.Ammonia gas was used to measure gas collection efficiency at 40 L/min for 1 h and 6 h, and ammonium sulfate was used to generate particles for particle loss measurement by means of a scanning mobility particle sizer.Both gas collection efficiency and particle loss by diffusion were also theoretically estimated, and the theoretical values were compared with the corresponding experimental values.Our results should assist users in selecting the appropriate number of denuder channels, as well as the amounts of coating and extraction solutions for the channels.

Multichannel Annular Denuders
Annular denuders with the following

Experimental Set-up for Determining Gas Collection Efficiency
Ammonia gas was supplied at 0.1-0.2L/min from a standard ammonia gas cylinder (6.46 ppm, Takachiho Trading, Tokyo, Japan), and the gas was diluted at the glass chamber with purified air supplied from a compressor (Fig. 1).
The mixed air was introduced into the denuder section.The denuder section consisted of two denuders connected in series: the number of channels in the first denuder varied from 1 to 8, and the second denuder was a 1-channel denuder, which was used to estimate the collection efficiency of each denuder.The total flow rate was checked by means of a mass flow meter located in the downstream of the denuders.Neither relative humidity nor temperature was controlled, but both were monitored by means of a thermohygrometer (HI 8564, Hanna Instruments, USA).The temperature was 23 C (± 2.0 C), and the relative humidity was below 25%.The sampling durations were 1 h and 6 h.6 hour sampling duration was chosen to see whether there is breakthrough or not and for field application.

Measurement
Ammonium sulfate particles were generated by an atomizer (Constant output atomizer, TSI 3076, MN, USA) from an aqueous ammonium sulfate solution (0.003 mol/L).The generated particles were introduced into the glass chamber and diluted with purified air supplied from the compressor at 40 L/min (Fig. 2. a and b).The generated particle size ranged from 0.02 to 0.3 μm.
Particle loss in the annular denuder under test was determined by measuring particle number concentration at the upstream and downstream locations of the test system, and comparing the particle loss with the test system present to the loss with the test system removed.Upstream and downstream concentrations (N in and N out ) were measured with a differential mobility analyzer (DMA) to measure the particle size distribution (TSI 3080, USA) and an ultrafine condensation particle counter (UCPC) to measure the particle number distribution (TSI 3025A, USA).Particle loss at 20 L/min was also measured to determine the effect of sampling flow rate on electrostatic forces and particle diffusion.The fraction of particle losses ( L ) was estimated from the following equation:

Denuder Cleaning and Coating Procedure
Citric acid has been widely used to coat the walls of denuders used to collect ammonia gas.However, some researchers recently suggested that using citric acid coating for ammonia gas suffers from insufficient strength of the bond between collected ammonia and the coating layer, which causes a release of the collected ammonia both towards the active sites of the denuder glass (Perrino and Gherardi, 1999;McCulloch and Shendrikar, 2000).They studied three coating layers, citric and phosphorous acid and oxalic acid, and found that phosphorous acid proved to be more suitable for ammonia determination.Many studies have used phosphorous acid to collect ammonia gas since its capability was proved (Perrino et al., 2002;Yu et al., 2006;Edgerton et al., 2007).
Denuders were cleaned as follows.Each denuder was cleaned prior to use.First, the denuder was rinsed with methanol for 20 s with manual rotation, as described elsewhere (Winberry et al., 1990;Winberry, 1992).
Second, the denuder was rinsed with deionised water for 20 s twice with manual rotation.The denuder coating was prepared by dissolving 1% phosphorous acid in a 1:9 water-methanol solution (10 g phosphorous acid, 100 mL deionised water, and 900 mL methanol) as described by Yu et al.   (2006).A total of approximately 18 mL of coating solution was applied to the denuder wall for the 1-channel denuder (three 6-mL portions), 30 mL was applied for the 3channel denuder (three 10-mL portions), 42 mL was applied for the 5-channel denuder (three 14-mL portions), and 60 mL was applied for the 8-channel denuder (three 20-mL portions) to coat the denuder surfaces efficiently.Each coating procedure was carried out for 20 s with manual rotation to distribute the coating solution evenly.The coated denuders were dried with under a flow of ultrapure nitrogen gas (1 L/min) for several minutes; then both ends were capped with clean caps, and the denuders were soon used for gas collection.

Sample Extraction
Sample extraction had to be considered since multichannel denuders were used in this study.The main method used for sample extraction is the U.S. EPA method for an annular denuder with only 1 channel (Winberry et al., 1990;Winberry, 1992).Although many researchers are currently using multichannel annular denuders, there have been no detailed studies of sample extraction methods specifically for multichannel annular denuders.When extraction is necessary, the efficiency of the extraction method must be considered.In this study, we used different volumes of extract solution for each multichannel denuder because the denuders have different surface areas.The extraction efficiency of each denuder was calculated to optimize the extraction.Although the U.S. EPA method recommends 10-mL total volume of extract solution for 1-channel denuder, 15 mL total volume of extract solution was applied for 1-, 3-channel denuders for the effectiveness.
The denuders were extracted with deionised water immediately after sampling.The applied extract solution was determined on the basis of pre-experiments.Different volumes (total 30 mL) of extract solution were used for the 5-and 8-channel denuders for the pre-experiments, but the extraction efficiencies were lower than 90% for the 5and 8-channel denuders.In contrast, extraction efficiencies greater than 95% were obtained for the 1-and 3-channel denuders with a 15-mL total volume of extract solution.Therefore, 15 mL of deionised water was applied for the 1-and 3-channel denuders, and 45 mL of deionised water was applied for the 5-and 8-channel denuders in this study.The total volume of extract solution was divided into three portions to improve the extraction efficiency (three 5-mL portions for the 1-and 3-ch; three 15-mL portions for the 5-and 8-ch).Each extraction was carried out for 1 min with manual rotation.

Sample Analysis
The collected ammonia gas was analyzed by ion chromatography (DX-100 Ion Chromatograph, Dionex, CA, USA).The ion chromatograph was calibrated prior to the analysis by injection of aqueous standards prepared in our laboratory from analytical-grade salts.

Extraction Efficiency
The extracted denuders were extracted once more to measure the extraction efficiency, which was calculated by dividing the amount of gas obtained from the first extraction by the sum of amounts obtained from the first and second extractions.The extraction efficiencies were 96.4,98.4, 96.7, and 92.0% for the 1-, 3-, 5-, and 8-channel denuders, respectively.The extraction efficiencies of the 1-, 3-, and 5-channel denuders were acceptable, but the efficiency of the 8-channel denuder was somewhat lower than the others, owing to the lack of extraction solution compared to its surface area.

Gas Collection Efficiency
The theoretical gas collection efficiency of a denuder can be determined from the following equations (Possanzini et al., 1983): (2) 3.The relative standard deviations were less than 1% and 1.5% for the gas collection efficiency (except for the 5-ch denuder) and the extraction efficiency, respectively.
4. The theoretical gas collection efficiency was calculated with an NH 3 diffusion coefficient (D) of 0.227 cm 2 /sec (Andrew, 1955) and at a flow rate of 40 L/min.

Particle Loss
Although the gas collection efficiency is high, particle loss can occur inside of the denuder.Because annular denuders are designed to have a narrow annular space, there could be particle losses, especially for UFPs (< 0.1 μm in diameter) mainly due to the diffusion of particles and electrostatic forces in the annular space.The particle loss by diffusion was evaluated using the equation developed by Hinds (1982).This equation was first established for the diffusion and loss of particles in a rectangular cross section.Ye et al. (1991) later adapted it for annular denuders.The fraction of particles lost by diffusion ( L ) in a denuder channel can be estimated from the following equations: where D(d) is the diffusion coefficient of a particle diameter, L is the tube length, W is the mean perimeter of the channel under study, Q is the flow rate, and h is the width of the channel.The particle transmission efficiency can be calculated from the following equation: The theoretical particle transmission efficiencies and the corresponding experimental values with particle sizes are presented in Table 3 (at 20 L/min) and 3 (at 40 L/min).
The standard deviations seem large; we attributed the large standard deviations mainly due to alterations in the electrostatic surface charge caused by various factors, including changes that occurred during the L = 1 -0.910e -7.54μ -0.0531e -85.7μ -0.0153e -249μ (5) drying and assembling procedure.Ye et al. (1991) reported the electrostatic surface charge problems in an annular denuder previously.Therefore, the large standard deviations, particularly for the 5-and 8channel denuders, may have been due to the fact that these multichannel denuders have many annulus spaces providing more places to be altered by the electrostatic forces.
As shown in Tables 3 and 4, particle loss increased as the flow rate decreased and the number of channels increased.Because of the vertical position of the denuder and the laminar flow conditions (Re < 2000) in the channel of the denuder, the main mechanisms responsible for the losses of the small particles (< 0.2 μm in diameter) were electrostatic forces and diffusion (Febo et al., 1999).Previous studies (Ye et al., 1991;Roussel et al., 2004) showed that particle loss is negligible for bigger particles (> 0.1 μm in diameter) and significant for smaller particles (< 0.1 μm in diameter).Particle transmission efficiency for all denuder channels at 20 L/min was much lower than the theoretically expected particle transmission efficiency, especially for the particles smaller than 0.04 μm and for the 5and 8-channel denuders, whereas the particle transmission efficiency at 40 L/min was close to the expected transmission efficiency, especially for the 1-and 3-channel denuders, which indicate that the electrostatic forces were minimized.Particle transmission efficiencies for different numbers of channels at the two flow rates are plotted in Unlike the particles studied by Ye et al. (1999) the particles generated from the atomizer in our study could be affected not only by particle diffusion but also by the electrostatic field in the denuder walls because the generated particles were not neutralized before being introduced to the denuder section.The theoretical particle transmission efficiencies presented in Tables 3 and 4 were only considered particle losses by diffusion using Eq. ( 5) and ( 6), as described in Section Particle loss.Therefore, the particle losses that occurred at 20 L/min were larger than the expected values mainly because of the local electrostatic field in the denuder channels.For particles smaller than 0.04 μm in diameter, electrostatic forces cannot be considered as negligible (Roussel et al., 2004).Madler and Friedlander (2007) studied on transport of nanoparticles, and one of their conclusions was that the surface of nanoparticles is large enough to carry one or more charges which greatly influence their transport in electrical fields.
Furthermore, it was observed that the electrostatic forces could be reduced when the flow rate increased, because the decrease of transmission efficiency was not significant at the high flow rate (40 L/min), whereas the decrease in transmission efficiency was significant at the low flow rate (20 L/min).This lower influence for the highest flow rate is the consequence of the shorter residence time of the particle and minimized electrostatic forces in the denuder channels.As a matter of fact, the residence times at 20 L/min were 0.05, 0.09, 0.28, and 0.54 s for the 1-, 3-, 5-, and 8channel denuders, respectively whereas the residence times at 40 L/min were 0.03, 0.05, 0.14, and 0.27 s for the 1-, 3-, 5-, and 8channel denuders, respectively.A similar effect of electrostatic forces and shorter residence time was also observed by Roussel et al. (2004) and theoretically considered by Ye et al. (1991).The results of particle transmission efficiencies observed in previous studies (Ye et al., 1991;Roussel et al., 2004) were compared with those observed in this study as shown in Table 5.
The annular denuders used in the previous studies were manufactured by URG and have the same geometry (1-and 8-channel denuders).
The particle transmission efficiencies at 10 L/min with 1-channel denuder obtained by Ye et al. (1991) and at 20 L/min with 1channel denuder estimated in this study are compared in Table 4.As shown in this table, the particle transmission efficiency of this study was a little lower than that obtained by Ye et al. even   c: Particle transmission efficiency for an 8-channel annular denuder coated with XAD-4 for diesel exhausts (Roussel et al., 2004).
higher.This result could be explained in terms of differences between generated particles.Ye et al. (1991) used monodisperse and neutralized particles or Boltzmann charged particles to estimate the effects of particle diffusion and electrostatic forces separately.However, in the atmosphere, particles are polydisperse, and both neutral and charged particles can be present; hence particle losses could be larger.In contrast, the particle transmission efficiencies at 34 L/min with 8-channel denuder obtained by Roussel et al. (2004) and by us at 40 L/min were quite similar.Roussel et al. (2004) used diesel vehicle exhaust to estimate the particle transmission efficiency with XAD-4 coated 8-channel denuders.Therefore, they could estimate the particle transmission efficiency under real atmospheric conditions.Because the UFPs and nanoparticles can be lost in denuder systems, it is very important to assess these small particles (< 0.1 μm) in denuders for an AD-FP system.However, it was found that the loss of particles with diameters < 0.1 μm could be minimized when the flow rate is higher; the particle losses in the important 0.02-0.2-diameterrange were less than 3% for the 1-and 3-channel denuders at 40 L/min.
Andrew, S.P.S. (1955) The 3-channel denuder would be the best denuder to use in an AD-FP system combined with the inertial classification system, in terms of the collection efficiency and particle transmission efficiency (for particles with diameters < 0.1 μm).However, the 1-channel denuder could also be used with the new classification system because the 1-channel denuder is competitively priced, coating the denuder wall is easier, and the 1-channel denuder is easy to carry for field sampling.Furthermore, gas collection efficiency for denuder is generally acceptable when it is over 95%.
(a) Overall experimental set-up.(b) Experimental set-up in detail.
Figs. 3. a and b.The figure clearly shows the effect of flow rate and the number of channels on the size-segregated particle losses in the denuders.
where E is the collection efficiency, C 0 is the inlet pollutant concentration, C is the exit pollutant concentration, L is the length of the denuder, d 1 is the inner diameter of the annulus, d 2 is the outer diameter of the annulus, F is the air flow rate, and D is the

Table 3 .
Theoretical and experimental particle transmission efficiencies at 20 L/min for each denuder.Each value is given with the standard deviation obtained over 16 replicate tests.2. Theo, theoretical particle transmission efficiency; Ex, experimental particle transmission efficiency.

Table 4 .
Theoretical and experimental particle transmission efficiencies at 40 L/min for each denuder.

Table 5 .
(Ye et al., 1991rticle transmission efficiencies obtained in previously reported studies and in this study.Particle transmission efficiency for an uncoated 1-channel annular denuder for Boltzmann charged particles(Ye et al., 1991; the particle transmission efficiency for 0.04 μm particles is the average of the particle loss values for 0.03 and 0.05 μm particles).