Facility for generation of ambient-like model aerosols in the 1 laboratory: application in the intercomparison of automated 2 PM monitors with the reference gravimetric method

. A new facility has been developed which allows for a stable and reproducible generation of ambient-like 15 aerosols in the laboratory. The setup consists of multiple aerosol generators, a custom-made flow tube homogeniser, 16 isokinetic sampling probes and a system to control aerosol temperature and humidity. Model aerosols containing 17 elemental carbon, secondary organic matter from the photo-oxidation of α-pinene, inorganic salts such as 18 ammonium sulphate and ammonium nitrate, mineral dust particles and water were generated at different 19 environmental conditions and different number and mass concentrations. The aerosol physical and chemical 20 properties were characterised with an array of experimental methods, including scanning mobility particle sizing, 21 ion chromatography, total reflection X-ray fluorescence spectroscopy, and thermo-optical analysis. The facility is 22 very versatile and can find applications in the calibration and performance characterisation of aerosol instruments 23 monitoring ambient air. In this study, we performed, as proof of concept, an intercomparison of three different 24 commercial PM (particulate matter) monitors (TEOM 1405, DustTrak DRX 8533 and Fidas Frog) with the 25 gravimetric reference method under three simulated environmental scenarios. The results are presented and 26 compared to previous field studies. We believe that the laboratory-based method for simulating ambient aerosols 27 presented here could provide in the future a useful alternative to time-consuming and expensive field campaigns, 28 which are often required for instrument certification and calibration. 29

Ambient PM is not uniform with respect to chemical composition, particle size and shape. In most cases, PM does 57 not refer to a single pollutant with a distinct chemical signature, but rather to a highly variable mixture of To date, automated PM instruments which are used for regulatory purposes (e.g. at national air quality monitoring 61 stations) are tested for equivalence with the manual gravimetric reference method in monitoring sites using real 62 (Red-y MFC, Vögtlin, Switzerland) by splitting and directing part of the main primary aerosol flow to the exhaust. 144 A filter (HEPA capsule, Pall Corporation, USA) was placed upstream of each MFC to remove the particles from the 145 air flow. All four MFCs were connected to the same aerosol pump (VTE8, Thomas, Germany) as shown in Fig. 1.  146 The mobility diameter and number concentration of the soot and salt particles were determined with a scanning Germany) and a high-resolution optical particle counter LAS-X II (Particle Measuring Systems, USA). 153

Aerosol homogenisation and sampling 154
The homogenizer is a 2.3-m-long custom-made stainless steel tube with an inner diameter of 16.4 cm, placed 155 vertically. The design is based on a previous study, but has been significantly improved and the facility has been 156 shortened (Horender et al., 2019). The tube is equipped with five identical inlets, placed at the very top as shown in 157 Fig. 1 and 2(a). Dilution air (filtered, humidity and temperature controlled) is delivered to each one of the inlets at a 158 flowrate of 24 L/min. The air is conditioned in two steps (Niedermeier et al., 2020) in such a way that the 159 humidified air is particle free: First, the dew point is adjusted by passing the air through a Nafion humidifier (Series 160 FC125-240-10MP, PermaPure, USA) filled with water (ultra-analytic grade, Purelab ultra, ELGA, Switzerland) at a 161 preselected water temperature, adjusted between 3 °C and 30 °C with a cryostat/thermostat (LAUDA Ecoline 162 Staredition RE 306, Lauda DR. R. Wobser GmbH & Co. KG, Germany). After the Nafion humidifier, the air is fully 163 saturated with water. Subsequently, the air is guided through a heated hose (Series T-7000, Thermocoax Isopad 164 GmbH, Germany), where the temperature can be adjusted up to 100 °C. The temperature and RH of the aerosol were 165 monitored in the homogeniser at the height of the sampling probes with digital sensors (FHAD 46 series/Almemo 166 D6, Ahlborn, Germany). 167 The primary aerosols are injected in the middle of the tube through separate ports located 50 cm downstream as 168 shown in Fig. 2(b). The dilution air sweeps the particles down the tube, where they are further mixed by three 169 turbulent jets of air. The three air-jet injection tubes (flow rate 20 L/min each) are placed symmetrically around the 170 homogenizer tube pointing 60° downwards ( Fig. 2(b)). The total flow rate of the homogenised aerosol is hence 171 equal to 180 L/min plus the flows of the four primary aerosols (in total less than 10 L/min). The temperature and 172 relative humidity of the air-jets are adjusted as described above for the dilution air. Finally, the homogeniser is 173 surrounded by copper tubes with flowing water in order to maintain the stainless-steel tube at the same temperature 174 as the aerosol. The temperature of water is adjusted by a flow-type cooler (AS-160 Green Line, Lindr, Czech simulated in the laboratory; even though the aerosol entering the homogeniser can be preconditioned at a 179 temperature down to about 5 °C, the aerosol temperature at the outlet of the homogeniser will always be ≥10 °C. 180 The sampling zone is located 1.25 m downstream of the injection position and accommodates isokinetic sampling 181 probes (funnels) placed at the bottom end of the homogenizer as illustrated in Fig. 2(c). Isokinetic conditions are 182 necessary when sampling with instruments operating at different flow rates to ensure representative sampling, e.g. 183 by minimizing sampling artefacts of larger particles. Several custom-made sampling probes with different cross 184 sections have been therefore designed to match the flow rate of the various automated PM monitors, which typically 185 ranges between 0.2 L/min and 20 L/min. It is worth noting that the sampling system is highly adaptable; the lower 186 end (outlet) of each sampling probe has custom-made threads so that it can be screwed in and out of the bottom 187 metallic plate of the homogeniser. This ensures that the sampling probes can be readily exchanged before each 188 experiment depending on the specifications of the PM monitors under test. Finally, the excess aerosol flow exits the 189 homogeniser through an exhaust outlet connected to a vacuum line as illustrated schematically in Fig. 1. 190 To characterise the aerosol homogeneity in the flow tube as a function of particle size, sodium chloride (NaCl) 191 particles with a geometric mean mobility diameter of 50 nm and mineral dust particles with aerodynamic diameter in 192 the lower µm range (ISO A2 dust) were generated with a nebuliser and a rotating-brush generator, respectively, as The tests were performed with NaCl and mineral dust particles separately. In both cases the aerosol spatial 203 homogeneity was found to be well within 3 % in number concentration as shown in Fig. 3(a) and (b), respectively, 204 indicating that the particle mixing characteristics do not depend on particle size in the tested range (i.e. from lower 205 nm to lower µm range). A final test was performed by mixing NaCl and dust particles to investigate whether the 206 particle mixing properties are affected when two primary aerosols are introduced into the homogeniser 207 simultaneously. It was confirmed that the aerosol homogeneity remains well within ±3 % (measurements not 208 shown), indicating that the simultaneous injection of primary aerosols into the homogeniser through separate ports 209 (see Fig. 2(b)) does not compromise particle mixing in any way. 210 By calculating the standard deviation of all 28 measured data points, the spatial inhomogeneity of the aerosol in 211 terms of number concentration was found to be 1.3 % for coverage factor k=1 or 2.6 % for k=2. This is used as an 212 estimate for the uncertainty of the aerosol spatial homogeneity (see 4th row of Table 1). This is a crucial 213 parameter which had not been evaluated so rigorously, if at all, in previous chamber studies (Hogrefe et al. inert, electrically conducting rubber material and was kept as short as possible (≈ 5 cm) without bends to minimize 235 deposition losses of particulate matter by kinetic processes as well as losses due to thermal, chemical or electrostatic 236 processes. Finally, the laboratory temperature and pressure were kept constant at (21 ± 1) °C and (950 ± 20) hPa, 237 respectively. 238 Before sampling, the filters were conditioned and weighed at NPL and shipped in individual plastic containers to 239 METAS. After sampling, the filter samples were placed in Petri dishes, wrapped tightly in plastic cover and stored at 240 4 °C for about a week. They were then shipped to NPL for conditioning and weighing. NPL use a Measurement 241 Technology Laboratories robotic filter weighing system that comprises an environmental chamber (20 °C ± 1 °C and 242 47.5 % ± 2.5 % relative humidity), an autohandler system and a Mettler Toledo XP2U balance. The filters are 243 conditioned in the chamber for 48 hours before weighing. The filters are weighed, then the system pauses for 24 244 hours before reweighing the filters to identify any time-variation in filter mass. Numerous QA/QC checks are made 245 before each set of weighings. 246

Uncertainty budget for the laboratory-based calibration of PM monitors 247
The reference mass concentration, , , is given by the equation , = , where is the aerosol 248 homogeneity in the flow tube, m is the particulate mass collected on the filter and V is the sampled volume. V is 249 given by the aerosol flow through the filter, Q, multiplied by the time duration of the measurement t.
is defined 250 https://doi.org/10.5194/amt-2020-362 Preprint. Discussion started: 7 October 2020 c Author(s) 2020. CC BY 4.0 License. as the relative particle penetration, = , where and is the penetration through the sampling 251 probe and connecting tube of the device under test (DUT) and the reference method, respectively. The associated 252 uncertainties are listed in Table 1. 253 Since sampling is carried out with isokinetic sampling probes and the tubes leading to the filter holder and the DUT 254 are kept straight and as short as possible, particle losses are minimised. Penetration was set to 1, however, an 255 uncertainty of 2 % was assigned to account for the higher impaction losses of supermicrometre particles in the 256 sampling funnel of the reference method due to the higher sampling flow (von der Weiden et al., 2009). These losses 257 are to some extent counteracted by the lower diffusion losses of submicrometre particles, which decrease with 258 increasing sampling flow. Here, we followed a rather conservative approach and kept the uncertainty of at 2 %. 259 3 Chemical characterisation of model aerosols The elemental composition of the model aerosols was characterised by combining a cascade impactor for PM 275 sampling with Total Reflection X-ray Fluorescence Spectroscopy (TXRF, Bruker TStar S4™, Germany) (Osán et 276 al., 2020). A 13 stage low pressure cascade impactor (Dekati DLPI 10™, Finland) with particle size range from 30 277 nm to 10 µm was modified to sample at a rate of 10 L/min on smooth and clean commercial-grade acrylic discs with 278 30 mm diameter, suitable for TXRF. In TXRF, the incident X-ray beam hits the disc's surface at the total reflection 279 angle. The fluorescence spectrum is detected perpendicular to the surface and is dominated by the contributions 280 from the deposit, i.e. the sampled particles. This allows for the detection of element masses as low as ≈10 to 100 pg 281 and thus short sampling periods. The measured element quantities, combined with the sampled air volume, provide 282 the particle size-selected element mass concentrations in the aerosol. The discs were prepared with a 50 ng Yttrium 283 standard for TXRF calibration. 284 https://doi.org/10.5194/amt-2020-362 Preprint. Discussion started: 7 October 2020 c Author(s) 2020. CC BY 4.0 License.
As example, the TXRF analysis of model aerosol 1 is shown in Fig. 4. The analysis revealed that the mineral dust 285 particles contain primarily the elements Si and Al and it was assumed that these are present as oxides SiO2 and 286 Al2O3. The mass-based aerodynamic distribution of the SiO2 particles exhibits a maximum in the range 1−2 µm 287 while the Al2O3 particles are larger (≈7 µm). Sulphur (i.e. in the form of sulphate ions) appears predominantly in the 288 submicrometre range (aerodynamic diameter of 30 nm−1 µm) but a second weaker mode is visible at ≈4−7 µm, thus 289 simulating the aerodynamic size distribution of sulphates in ambient air (Wall et al., 1988;Zhuang et al., 1999) 290 reasonably well. The coarse mode arises most probably from internal mixing of sulphate ions with mineral dust 291 particles. Since nitrates and sulphates were generated with the same method, nitrates are expected to exhibit a 292 similar bimodal size distribution but this could not be experimentally confirmed since nitrogen is difficult to detect 293 with TXRF spectroscopy. Finally, K + and Clions appear in the micrometre range (>2 µm). It is reasonable to expect 294 that Na + ions appear also in this size range, however, this could not be investigated by TXRF. By comparing the 295 results of ion chromatography with those of TXRF spectroscopy, there is no evidence of insoluble potassium. 296 The results of the chemical analysis of the model aerosols with ion chromatography, EC/OC analysis and TXRF 297 spectroscopy are summarised in Table 2  DustTrak DRX 8533 and the Fidas Frog aerosol monitors are, unlike TEOM, portable and more cost efficient. These 304 do not measure particle mass directly but record instead the particle number concentration and size distribution 305 using optical techniques, from which they calculate the mass concentration using built-in algorithms. 306 The PM monitors were exposed to three different model aerosols, which were generated in the laboratory with the 307 facility described in Sect. 2. All three model aerosols were ambient-like mixtures, i.e. they contained inorganic salts, 308 elemental carbon (soot), secondary organic matter, mineral dust and water. The aerosol composition was analysed 309 with the methods described in Sect. 3. The chemical composition of the model aerosols and the environmental 310 conditions during each experiment are listed in Table 2  aerosol decreased during measurement, the best way to assess the performance of the TEOM 1405 with respect to 336 the reference method is to calculate the 4-h-average mass concentration. This amounts to 41.6 µg/m 3 (see Table 3), 337 only 3.7 % lower than the reference measurement (43.2 µg/m 3 ). 338 The fresh soot particles consist mainly of EC and have a geometric mean mobility diameter of about 120 nm, i.e. 339 below the cut-off limit of the Fidas Frog. Indeed, experiments with miniCAST soot showed that the Fidas Frog and 340 DustTrak 8533 failed to detect soot particles of this size. This explains why the Fidas Frog reported a constant mass 341 concentration over the whole measurement period. In Table 3 The results obtained with model aerosol 2 are displayed in Fig. 6(b). Here, the concentration of the aerosol remained The results obtained in the case of model aerosol 3 are illustrated in Fig. 6(c). With an average PM10 mass 372 concentration of 19.2 µg/m 3 , the TEOM 1405 exhibits an excellent agreement with the reference method (19.3 373 µg/m 3 , see Table 2). The DustTrak 8533 overestimates the mass concentration by approx. 33 %, and thus performs 374 slightly better than in the case of model aerosol 2. Fidas Frog underestimates the mass concentration by about 23 %, 375 or ≈15 % after correction for the undetected mass of fresh soot, in agreement with the findings of the experiment 376 with model aerosol 2. As mentioned above, PM monitors based on light scattering, such as the Fidas Frog and the 377 DustTrak, measure particle number concentration and convert this into mass concentration by using a size-378 dependent particle density function. This function is integrated into the software of the instrument. Deviations may 379 occur if the built-in functions differ substantially from the real density function of the aerosol. More experiments 380 with ambient-like model aerosols under low and high relative humidity would be needed to define a comprehensive 381 set of calibration factors for these instruments. 382

383
In this study, we present the first steps towards the generation of ambient-like aerosols in the laboratory.      Table 2 for a discussion on all three model aerosols).   https://doi.org/10.5194/amt-2020-362 Preprint. Discussion started: 7 October 2020 c Author(s) 2020. CC BY 4.0 License.