What are the driving factors influencing the size distribution of 1 airborne synthetic clay particles emitted from a jet milling process? 2

13 14 In the field of workplace air quality, measuring and analyzing the size distribution of airborne 15 particles to identify their sources and apportion their contribution has become widely accepted, 16 however, the driving factors that influence this parameter, particularly for nanoparticles (< 100 17 nm), have not been thoroughly determined. Identification of driving factors, and in turn, general 18 trends in size distribution of emitted particles would facilitate the prediction of nanoparticles’ 19 emission behavior and significantly contribute to their exposure assessment. In this study, a 20 comprehensive analysis of the particle number size distribution data, with a particular focus on 21 the ultrafine size range of synthetic clay particles emitted from a jet milling machine was 22 conducted using the multi-lognormal fitting method. The results showed relatively high 23 contribution of nanoparticles to the emissions in many of the tested cases, and also, that both 24 surface treatment and feed rate of the machine are significant factors influencing the size 25 distribution of the emitted particles of this size. In particular, applying surface treatments and 26 increasing the machine feed rate have the similar effect of reducing the size of the particles, 27 however, no general trend was found in variations of size distribution across different surface 28 treatments and feed rates. The findings of our study demonstrate that for this process and other 29 activities, where no general trend is found in the size distribution of the emitted airborne particles 30 due to dissimilar effects of the driving factors, each case must be treated separately in terms of workplace exposure assessment and regulations. This study infers the emission behavior of the synthetic clay particles from a jet milling process through analysis of the measured size distribution in a comprehensive particle size range, with a focus on the submicrometer and more particularly, nanoparticles. The suite of instruments included a SMPS for measurements of submicrometer and nanoparticles and an OPC for supermicrometer particles. The effects of the surface treatment of the clay materials and of the feed rate of the milling machine were each studied as potential driving parameters of the size distribution of the emitted particles.


36
There is now a growing consensus over the impact of airborne engineered particles' size on 37 their toxicity and behavior in different environments (Maynard and Kuempel 2005;Donaldson 38 et al. 2006;Asbach et al. 2009). Thus, determining the size distribution of these particles 39 during different stages of their life cycle, particularly in nanotechnology workplaces where the 40 amount of the emitted particles is significant and many people are exposed to them, is of 41 crucial importance. This task requires isolating the particles of interest from any other 42 interfering factors (i.e., background particles and those emitted from other activities in the area 43 where measurement takes place), discriminating them based on their size, and finally 44 quantifying the particles in each size bin (Ramachandran and Cooper 2011). Researchers have 45 employed instruments operating on several different principles and/or size ranges for aerosol 46 size distribution measurements (Maynard and Aitken 2007;Ono-Ogasawara et al. 2009). A 47 the present study, the abovementioned size range was deemed to be appropriate for 123 characterization of nanoparticle emission from this process. The up-scan time was set to 120 s 124 followed by a down-scan of 30 s. The sheath and aerosol air flow of the classifier were set to 6 125 and 0.6 L min -1 . 126 At the beginning of each measurement day, the flow rates of the instruments were checked 127 by a bubble flow meter and their times were synchronized. The sampling frequency of the 128 OPC was set to the shortest time possible (1 s) to measure all the momentary variations during 129 size distribution measurements. The SMPS was calibrated prior to conducting the 130 measurements using monodisperse polystyrene latex (PSL) particles with the nominal diameter 131 of 46 nm. A zero check of all instruments was also performed at the beginning of each day of 132 measurements using a high-efficiency particulate air (HEPA) filter. 133 134

Study Design 135
Firstly, by comparing the instruments' readings in the vicinity of the jet milling machine 136 before and during its operating time, it was confirmed that the source of airborne particle 137 emission was at the connection point of the collection bag to the venturi outlet. A black with its other end connected to an aerosol flow splitter, which was used to feed both the OPC 140 and SMPS. To minimize particle losses, the shortest possible lengths were used for tubing (< 141 20 cm). Furthermore, the effect of particle loss inside the connecting tube due to the diffusion 142 was calculated based on the method presented by (Hinds 2012). The results showed than only 143 less than 4% of particles smaller than 20 nm in diameter were lost in the tube, confirming that 144 particle loss could be assumed negligible in this case. 145 For each of the available 24 samples, SMPS scans (150 s each) were conducted in triplicate. 146 Despite minor discrepancies across them, the scans were fairly similar in terms of the key 147 characteristics of the modes, confirming the reliability of the obtained results. After this 148 period, the machine was turned off, dismantled and cleaned for the next sample. Whilst 149 studying the effect of the surface treatments, the feed rate of the jet milling machine was kept 150 constant. 151 To assess the effect of the feed rate on the size distribution of the emitted particles, three 152 feed rates (7.5, 4.1 and 2.1 g min -1 ) were applied to two different samples . 154 software (R.Core.Team 2013). The background PNSDs (Particle Number Size Distributions) at 158 the emission source, during the operation of the jet milling machine and before adding the 159 product to it, were measured by SMPS and OPC, averaged, and then subtracted from the 160 measured PNSDs during the milling process to represent the product emission. Combining 161 SMPS and OPC data, in view of the differences in their measurement techniques, has been 162 done in some other studies such as the one published by (Park et al. 2011), where the data from 163 these instruments were combined to calculate reference surface area concentration. 164 The OPC data were used to obtain an overview of the size distribution in a wide range by 165 examining the number concentration in each size bin, as well as the number concentration of 166 super-micrometer particles. As mentioned earlier, the major focus of this study was on the 167 nanoparticles. Therefore, the main effort in analyzing the data was dedicated to them. Initially, 168 three scans for each sample were replaced by their mean, leading to a data set of 24 size 169 distributions. The data were then smoothed by fitting a Generalized Additive Model (GAM) in 170 R, based on penalized B-splines (Wood 2003;Eilers and Marx 1996) to eliminate the noises in 171 the PNSD spectra. Using the Multi-peak Fitting package in Igor Pro 6.21 (WaveMetrics), the 172 data for each sample were replaced by the following multi-lognormal fit:  Table 2 presents the individual background and process total PNCs measured by SMPS and 193 OPC, as well as the total concentrations for product and background for each sample. 194 "Background" and "Process" refer to the stages, when the milling machine was running 195 without and with the feed material, respectively. Therefore, the difference of the mean PNCs 196 between process and background yields the concentration of the emitted clay particles, i.e. , 197 product emission. 198 Comparison of the individual SMPS and OPC total number concentrations in Table 2 shows 199 that the emitted clay particles were dominated by those in the sub-micrometer range, 200 particularly d<0.3 µm, which is the SMPS data range used in this study. The mean total PNC 201 of the background was 4.3×10 2 cm -3 with a SD of 2.4×10 2 cm -3 . Similar to the emitted clay 202 particles, background particles were also mainly in the range of d< 0.3 µm. Total PNC of the 203 emitted clay particles were obtained by subtracting the background concentration from the 12 process concentration for both SMPS and OPC and summing the results. It varied in the range 205 of 0.3-26.2×10 3 cm -3 . 206 Among the particle sizes, concentrations of LUC samples obtained by both SMPS and 207 OPC, and consequently, the total concentration were the lowest. On the other end of the 208 spectrum, CLO samples had the highest overall SMPS number concentration and the second 209 highest overall OPC number concentration after H80. Therefore, this particle size exhibited the 210 highest average total concentration. The variations of concentration across different surface 211 treatments within each size were not identical. For instance, whilst for H120 and CLO, the 212 emission of the unmodified samples was significantly higher than others, for H80, MMOD 213 treatment showed the highest emission. 214 This initial assessment highlighted a need to investigate the PNSD of the emitted clay 215 particles more closely, mainly in the SMPS size range, which includes ultrafine particles. 216 217
Smoothing the PNSD spectra 219 Fig. 1 shows the smoothed mean PNSDs of background, process including background, and 220 emitted clay particles for sample H120-ETHO, as an example for the application of GAM fit in 13 this study. As can be seen, background particle concentration is very low. It remained almost 222 unchanged for other tested samples due to the effective ventilation of the experiment location, 223 which provided a fairly clean environment. 224 Together with the total number concentration results in Table 2, Fig. 1 provides solid 225 evidence for the emission of clay particles from this process by showing a significant 226 difference between background and product concentrations. As can be seen in Fig. 1, the 227 PNSD for the emitted clay particles includes multiple modes corresponding to both 228 nanoparticles and particles of bigger sizes (>100 nm). Another noteworthy point regarding 229 Fig. 1 is that correcting the measured PNSD for that of background did not affect the 230 properties of these modes, and their shapes and locations were preserved. 231

3.2.2.
Multi-lognormal fitting 232 In order to quantify the effect of surface treatment and feed rate on the size distribution of 233 the emitted particles, a multi-lognormal fit was applied to the PNSD spectrum of each sample. 234 This section starts with studying the effect of surface treatment and will proceed to evaluation 235 of milling feed rate as another parameter of potential impact. 236 hence, confirming the suitability of the applied multi-lognormal fits.  Table 3 presents the results of this model. 279 According to Table 3, deviation values are mostly negative, indicating the emission of 280 smaller particles for the surface-treated samples compared to the unmodified ones, as was 281 previously shown in Fig. 3. Furthermore, CC and ETHO show the most variations, as they 282 have the highest means of absolute deviations from the baseline GMDs among the surface 283

treatments. 284
Contribution of the nanoparticles to the total emissions: The ratios of the number 285 concentration of three size ranges (nanoparticles, 100 < GMD <300 nm, and GMD > 300 nm) 286 to the total number concentration measured for each sample are shown in Fig. 4. The 287 concentrations of the first two ranges were measured by SMPS and the last one was measured 288 by OPC. 289 Fig. 4 shows that CLO contributed the most to the emissions of nanoparticles, whilst H80 290 had the lowest emission of such particles amongst the tested materials. Across the surface 291 treatments, CMOD had the most contribution to the emissions of nanoparticles, with three 292 dominant modes in this range, i.e., more than 50% contribution to the total concentration. 293 As stated previously, feed rate of the milling machine was also considered as a potential 294 driving factor influencing the particle emissions. Fig. 5 gives an overview of how the number 295 concentrations of the emitted particles in different size ranges were affected by the feed rate of 296 the milling machine. 297 According to Fig. 5, the emissions of particles did not undergo any significant variations 298 due to the different feed rates in any of the studied size ranges, however, H80 was shown to be 299 affected more than LUC as the variations across the feed rates were somewhat higher in all 300 size ranges, particularly in the first one (d < 300 nm). Therefore, the PNSD data obtained by 301 the SMPS were processed and analyzed in the same way as in the previous section, in order to 302 better understand the effect of the feed rate on the emission behavior of the submicrometer 303 particles, with a particular interest in nanoparticles. Fig. 6 summarizes the main properties of 304 the lognormal modes, which constitute each PNSD spectra for both tested materials and for the 305 three tested feed rates. 306 It can be seen from Fig. 6 that all tested cases have two nanoparticle modes. It can also be 307 seen that decreasing the feed rate from 1 to 3 resulted in a shift in the location of nanoparticle 308 modes towards larger particles, however, there are some differences in how each material was 309 affected by the feed rate. For instance, the variations in the locations of nanoparticle modes in 310 LUC are more significant than in H80, particularly in the first mode. The same explanation 311 used for the effect of surface treatment on the location of the nanoparticle modes could also be 312 valid for the feed rate, associating the emission of smaller particles to the higher number of 313 particle collisions due to the increased feed rate. In terms of the contribution of the 314 nanoparticles to the total emissions in each case, the differences are more apparent. While the 315 contribution of LUC nanoparticles were increased consistently by decreasing the feed rate 316 (52%, 53%, and 60% for feed rates 1, 2, and 3, respectively), the contribution of the H80 317 nanoparticles reached its peak by 70% at feed rate 2, but subsequently decreased to 38% at 318 feed rate 3. 319 320 4. CONCLUSIONS 321 322 process through analysis of the measured size distribution in a comprehensive particle size 324 range, with a focus on the submicrometer and more particularly, nanoparticles. The suite of 325 instruments included a SMPS for measurements of submicrometer and nanoparticles and an 326 OPC for supermicrometer particles. The effects of the surface treatment of the clay materials 327 and of the feed rate of the milling machine were each studied as potential driving parameters 328 of the size distribution of the emitted particles. 329 The findings of this study show that the size distribution of the particles emitted from the jet 330 milling process is significantly influenced by the surface treatment applied to the material, as 331 well as by the feed rate of the machine. As the emitted particles are confirmed to be mostly 332 nanoparticles, these effects are more significant within this size range. In terms of the emitted 333 particle size, the general implication of the result is that applying surface treatments leads to 334 the emission of smaller particles, as does increasing the machine feed rate. On the other hand, 335 and regarding the contribution of nanoparticles to the emission, although the results indicated 336 relatively high levels in many of the tested cases, no general trend was observed in the 337 variations across either the surface treatment or the feed rate. The findings of this study signify 338 the importance of comprehensive size distribution measurements and analysis in shedding light nanoparticles, to provide input into the workplace exposure assessment and its regulations. 341 To our knowledge, the present study is the first to comprehensively analyze the emission 342 size distribution of synthetic nanoclays, which is an important and widely-used class of 343 materials, from a very common mechanical process. Unlike most of the similar studies, which 344 merely focused on the effect of adding nanofillers to the reference material on the size 345 distribution of the emitted particles, this study had a much closer look into the driving factors 346 influencing the size distribution. The present study may be considered a step forward in size 347 characterization of emitted nanoparticles from nanotechnology activities. A reason to support 348 this statement is that, in contrast to some similar studies, which had difficulties in interpreting 349 the PNSD data for nanoparticles due to different reasons such as the high levels of background 350 particles (Wohlleben et al. 2011;Koponen et al. 2010) anddiffusion (Schlagenhauf et al. 351 2012), in the present study, the emitted synthetic clay nanoparticles could be differentiated 352 from the background efficiently and correction of the size distribution data for the background 353 did not affect the modal properties of the PNSD spectra. Moreover, not only the findings of 354 this study confirmed that characteristics of the materials and the operational factors have a 355 major role in the size distribution of the emitted particles from a real-world process, the effects 356 of these parameters were also quantified and compared across a wide range of cases. 357