Concentrations and size distributions of particle lung-deposited 1 surface area ( LDSA ) in an underground mine 2 3

1 Concentrations and size distributions of particle lung-deposited 1 surface area (LDSA) in an underground mine 2 3 Salo, Laura*; Rönkkö, Topi; Saarikoski, Sanna; Teinilä, Kimmo; Kuula, 4 Joel; Alanen, Jenni; Arffman, Anssi; Timonen, Hilkka; Keskinen, Jorma 5 6 Aerosol Physics Laboratory, Physics Unit, Tampere University, Tampere, 33720, Finland 7 Atmospheric Composition Research, Finnish Meteorological Institute, Helsinki, 00560, Finland 8 Currently employed at Dekati Ltd., Kangasala, 36240, Finland 9 10  Corresponding author. Tel: +358 451122546 11


INTRODUCTION
Particulate matter (PM) in the air is known to be harmful for human health (Burnett et al., 2018; 39 Lelieveld et al., 2015), mostly due to the ability of particles to penetrate and deposit into the human 40 lungs (Pope, 2000;Pope and Dockery, 2006). This lung-deposition, as well as the deposition of 41 particles in other parts of the human respiratory tract, is strongly dependent on the size distribution 42 of particles. Especially ultrafine particles (particle diameter smaller than 100 nm) and to some 43 extent larger fine particles (particle diameter smaller than 2.5 µm) can reach the most vulnerable 44 alveolar areas of human lungs and deposit there (Oberdörster, 2001). These particles can consist of 45 compounds that are toxic for the human body or they can carry the toxic components on their 46 surfaces (Goulaouic et al., 2008). In addition to acute symptoms, the particulate matter of inhaled 47 aim of this article is to report the concentrations of LDSA in various locations in the mine and 113 evaluate the applicability of LDSA as a measurement metric and the applicability of sensor-type 114 measurements of LDSA in a mine environment. Five measurement sites were chosen for this study, each to represent an aerosol source or a 126 location of personnel exposure (or both). Three sites-the dumping area, crushing station and 127 maintenance area-each had workers operating in the area, although in the dumping area most 128 workers were inside vehicles. The maintenance area also had vehicle traffic, and many people were 129

A C C E P T E D M A N U S C R I P T
7 also walking in the area, as offices and the cafeteria were located on this level. The two remaining 130 sites were the blasting area and conveyor belt. The blasting area did not have workers during the 131 blasting for safety reasons, but trucks arrived shortly afterwards to transport rocks. The conveyor 132 belt location did not have much human activity during our measurements, although the belt itself 133 was in operation. We expected vehicle emissions to be present in all locations, but especially in the 134 dumping area and in the blasting area (after the blasting, which were scheduled to be at 14:00 and 135 22:00), where heavy-duty traffic was frequent. The maintenance area was frequented by passenger 136 vehicles, taking people to the offices and cafeteria, so some traffic exhaust was expected to 137 originate from there as well. We expected to see coarser particles at the crushing station as well as 138 the blast site. The measurement locations along with the mobile laboratory setup are described in 139 more detail by Saarikoski et al. (2019), where other aspects (particle mass, number, and chemical 140 composition) of this measurement campaign are reported. 141 Measurements were conducted primarily using ELPI+ (Electrical Low-Pressure Impactor, 142 Dekati Ltd.) but, in addition, the applicability and performance of five smaller, sensor-type 143 instruments was tested in the mine environment. 144 ELPI+ is a 14-stage low-pressure cascade impactor. Unlike in traditional impactors, the particles 145 are measured electrically in real time (Keskinen et al., 1992). For the particles to be detected, they 146 are first charged with positive ions from a diffusion charger. The raw data from the instrument is 147

A C C E P T E D M A N U S C R I P T
8 the current measured from each of the 14 stages. The lowest stage is a filter, which collects 148 essentially all particles, but due to weakened charging of very small particles, the lowest detectable 149 size is 6 nm. Before the highest stage, there is an impactor to remove particles over 10 µm in 150 aerodynamic size. When handling the data, the currents were first corrected using zero 151 measurement data (current measured when a HEPA filter was placed at the inlet) and then adjusted 152 to account for secondary collection by diffusion (Järvinen et al., 2014). The LDSA distribution was 153 then calculated from the current distribution, using the method described by Lepistö et al. (2020). 154 Total LDSA was calculated for different size ranges, based on the impactor cut-off sizes. In this 155 article we have rounded the actual cut-off sizes to nearest round values, true cut-offs can be found 156 in Table S1. 157 The chemical composition of submicron (<1 µm) particles was determined by a Soot Particle 158 Aerosol Mass Spectrometer (Aerodyne Research Inc.). The SP-AMS was equipped with both laser 159 and tungsten vaporizers, and therefore, it was able to measure both non-refractory material 160 (organics, sulfate, nitrate, ammonium and chloride), and refractory material (i.e., the refractory BC 161 and metals) in particles. The details of the SP-AMS analysis can be found in Saarikoski et al. (2019). 162 Additionally, the black carbon concentrations were measured with a dualspot Aethalometer (AE33; 163 Magee Scientific). 164

A C C E P T E D M A N U S C R I P T
9 The five electrical particle sensors included in this study were two Partectors (Naneos), one 165 DiSCmini (Testo) and two AQ Indoors (Pegasor). The sensors employ a diffusion charger to charge 166 incoming particles and the electrical current generated by the particles is measured. The raw current 167 is transformed into LDSA using an internal calibration factor. The calibration factor relies on the 168 linear relationship between diffusion charging and LDSA, but this linear relationship is only true 169 for a rather small size range of approximately 20-300 nm. If a sensor is calibrated with 100 nm 170 particles, as is the case with the Partector, then the LDSA of 300 nm particles will be over- this method is that the particles do not need to be collected. The DiSCmini has two-stage collection, 177 which allows for it to estimate the mean size of the particle population. The first stage is a diffusion 178 collector, collecting only the smallest particles, and the second stage is a filter, collecting what is 179 left over after the first stage. The AQ Indoor employs a particle trap with a cycling voltage to 180 remove some of the particle population, also allowing for an estimation of the particle size. The 181 larger the voltage, the larger the particles which can be removed. To ensure coarse particles and 182

A C C E P T E D M A N U S C R I P T
10 dirt do not get inside the sensors, the Partector uses a wire mesh at the inlet, the DiSCmini has an 183 impactor which can be inserted at the inlet, and AQ Indoor has a cyclone. 184 The ELPI+ was located inside a mobile measurement laboratory along with various other air 185 quality instruments. The sensors were located initially in the Maintenance area of the mine and 186 connected to each other with Tygon tubing, to make sure they sampled the same air for the duration 187 of a sensor comparison. Later, the AQ Indoor sensors were moved to different locations in the mine 188 (AQ Indoor A to the dumping area and AQ Indoor B to the crushing station), for the sensor network 189 type measurement. The Partector and DiSCmini instruments began reporting faults after just a few 190 days of measurements, due to over-loading, and were not employed further.  Fig. 3 shows the time-averaged particle LDSA size distributions. In the figure, each line 224 corresponds to one of the color plots in Fig. 2. Most locations had the largest particle mode around 225 50-200 nm, the Blasting area being the main exception, with a mode size around 500 nm. Several 226 locations had two particle modes. Looking back at Fig. 2, at the blasting area and the crushing 227 station these different modes occur independently, meaning they must be from different sources. 228 In contrast, at the dumping area and conveyor belt, both modes appear at the same time. The 229 particles are either from the same source, or from different sources which have activity at the same 230 time. From Fig. 3 it may seem as though the conveyor belt and dumping area had less activity 231 (plots are smooth); however, the smooth appearance is due to the shorter time periods measured. 232 The dumping area was also measured with a sensor for a longer duration (Fig. S1), which shows 233 dramatic changes in the total LDSA. 234 Fig. S2 shows the chemical composition of particles by mass fraction, plotted against particle 235 size. As the data is by mass, it is difficult to directly compare with the LDSA size distributions in 236  Table 1 contains a summary of the numerical results for LDSA for each measurement location 245 (columns) and separated into particle size ranges (rows). The size ranges have been selected to 246 relate to the conventions in mass concentration measurement (PM10, PM2.5, PM1, PM0.1); 247 additionally, diffusion-charging based sensors are often calibrated for sizes between 30-300 nm. 248 The table shows that a large portion of particles (34-70 %) is measured incorrectly with these 249 sensors, due to the particles being larger than the size range that the sensors are calibrated for. On 250 the other hand, less than 13 % of LDSA is contributed from particles over 2.5 µm. 251  Fig. 4 shows a time series (top panel) of the sensor comparison performed at the beginning of 255 the measurement campaign (see Fig. 1). Five different diffusion-charging sensors were used, and 256 the 5-minute averaged values were compared to ELPI+. The lower plot shows the correlation 257 between ELPI+ and each sensor, along with fitted lines. The best correlations were with AQ Indoor 258 the slightly different measurement locations (the ELPI+ measurement point was approximately 5 260 meters away from the sensors). The largest discrepancy between sensors and ELPI+ was at the end 261 of the sensor comparison, where all sensors overestimated the LDSA concentration. The 262 overestimation did not persist after the sensors were moved to new locations (Fig. S4, Fig. S5). In most locations, the majority of the total LDSA concentration was contributed by sub 300 nm 280 particles; however, a significant portion (34 %-70 %) of LDSA was in larger particle sizes. 281 Despite the large size range of particles in the mine, electrical sensors were able to measure total 282 LDSA with a good correlation to ELPI+ (R 2 =0.53-0.59). Somewhat surprisingly, discrepancies 283 between sensors and ELPI+ measurements in the sensor comparison were unrelated to the presence 284 of large particles, and perhaps slightly effected by sub 30 nm particles. The relationship was 285 investigated by plotting the percentage difference of each sensor vs. ELPI+ on the y-axis, and the 286 portion of particles over 300 nm on the x-axis (Fig. S3 a), and similarly for particles smaller than 287 30 nm (Fig. S3 b). A slight correlation was observed only in the latter case. A previous laboratory 288 study showed that sub 20 nm particles are overestimated by LDSA sensors, while particles over 289 400 nm are underestimated (Todea et al., 2015), which is also explained by the relationship of the 290 diffusion charger Pn-curve (particle penetration efficiency multiplied by the number of unit 291 charges) compared to the deposition probability function (Fierz et al., 2014). We did find some 292 evidence of LDSA from large particles being underestimated by the AQ Indoor sensor from a 293 an inlet cyclone removing coarse particles is a necessity for long-term measurement. An optical 296 particle sensor would be a useful addition for locations with coarse particles, as they are better 297 suited for particle diameters over 300 nm. 298 Another recent study (Barrett et al., 2019) also examined particle sensors for underground mine 299 applications; however, they focused on measuring the mass concentration of elemental or black 300 carbon. They found good potential in a prototype black carbon sensor, which had good correlation 301 with the reference method during field testing (R 2 =0.85). Barrett  Longer measurements with the two AQ Indoor sensors indicated that particle populations in the 306 mine mix and spread quickly, at least in these two locations (Fig. S1). This means that to improve 307 air quality at a specific location requires changes in ventilation, in addition to controlling the 308 location-specific sources. Based on total LDSA time series data, the air exchange rate at the 309 dumping area was approximately 15 (1/h) (Fig S7 a) and 22 (1/h) in the crushing station ( Fig. S7  highest exposure levels had an overall mean exposure of 106 µm 2 cm -3 . In another exposure study, 324 airport taxiway personnel were found to operate in mean LDSA concentrations of 59 to 174 325 µm 2 cm -3 (Marcias et al., 2019). Our study was not a personal exposure study and cannot be directly 326 compared to these values; however, the potential for similar or even higher exposure exists. 327 The previous paper covering other measurements from this same campaign shows average PM10 328 was highest at the blasting area (355 µg m -3 ) and lowest in the maintenance area (49 µg m -3 ) 329 M A N U S C R I P T 21 (Saarikoski et al., 2019). While the LDSA is also higher in the blasting area than in the maintenance 330 area, the average LDSA to mass ratios for these locations are 0.9 and 1.6 (units µm 2 cm -3 µg -1 m 3 ), 331 respectively. This indicates that one unit of particle mass in the blasting area is likely to be less 332 toxic than one unit of mass in the maintenance area, based on the lung-deposition. Basing 333 guidelines or regulation purely on mass concentration misses this effect.