Characterization of the MISG soot generator with an atmospheric simulation chamber 1

10 The performance of a Mini-Inverted Soot Generator (MISG) has been investigated at ChAMBRe (Chamber 11 for Aerosol Modelling and Bio-aerosol Research) by studying the properties of soot particles generated by 12 ethylene and propane combustion. 13 Starting from an extensive classification of combustion conditions and resulting flame shapes, the MISG 14 exhaust was characterized in terms of concentration of emitted particles and gases, particle size distribution 15 and optical properties. Soot particles were also collected on quartz fibre filters and then analysed by optical 16 and thermal-optical techniques, to measure the spectral dependence of the absorption coefficient b_abs, and 17 their composition in terms of Elemental and Organic Carbon (EC and OC). Significant differences could be 18 observed when the MISG is fuelled with ethylene and propane both in terms of particle size and optical 19 behaviour (i.e., absorption coefficient). Values of the Mass Absorption Coefficient (MAC) and of the 20 Angstrom Absorption Exponent (AAE) turned out to be compatible with the literature, even if with some 21 specific difference. 22 The comprehensive characterization of the MISG soot particles is an important piece of information to 23 design and perform experiments in atmospheric simulation chambers. 24 25

"Soot" refers to combustion-generated carbonaceous particles that are a by-product of incomplete 27 combustion of fossil fuels and/or biomass burning (Nordmann et  The MISG can be operated with different fuels: ethylene , propane (Moallemi 51 et al., 2019), and theoretically also with ethane or fuel blends with methane and nitrogen, even if, to our 52 knowledge, no literature is available on such configurations. The air to fuel flow ratio can be adjusted to control 53 concentration and size of the generated particles. The maximum reachable concentration is about 10 7 particles 54 cm -3 (https://www.argonautscientific.com/), while particle size ranges from few tens to few hundreds of nm. 55 The behaviour of soot particles can be efficiently studied in/by atmospheric simulation chambers (ASCs): 56 these are exploratory platforms which allow to study atmospheric processes under controlled conditions, that 57 can be maintained for periods long enough to reproduce realistic environments and to study interactions among 58 their constituents (Finlayson -Pitts and Pitts, 2000; Becker, 2006). Recent examples concern the investigation 59 of the optical properties of mineral dust (Caponi et al., 2017) and wood-burning exhausts (Kumar et al., 2018).

60
Coupling the MISG to an ASC makes possible systematic experiments on the properties of soot particles 61 exposed and maintained in different conditions. In this work we mainly investigated the differences between 62 MISG exhausts produced by ethylene and propane burning. 63 64

.1 Mini-Inverted Soot Generator 67
The MISG, introduced by Kazemimanesh (2019), is a combustion-based soot generator working as an 68 inverted-flame burner (Stipe et al., 2005) where air and fuel flow in an opposite way to the buoyancy force of 69 the hot exhaust gases. This results in a co-flow diffusion flame and leads to a better flame stability by reducing 70 flame tip flickering (Kirchstetter & Novakov, 2007;Stipe et al., 2005) and consequently to a more stable soot 71 particle generation.

72
The MISG is fed with air and fuel supplied by specific cylinders: we used both ethylene and propane, two 73 fuels with a well-known capability of producing soot. Air and fuel flow rates are controlled by two mass flow 74 controllers (MFCs, Bronkhorst High-Tech B.V., Ruurlo, Netherlands, Models F-201CV-10K-MGD-22-V and 75 FG-201CV-MGD-22-V-AA-000, respectively) operated via a home-made National Instruments Labview 76 code. The air and fuel flows can be controlled in the range 0-12 lpm and 0-200 mlpm, respectively. Differently 77 from other commercial generators, the MISG does not require a third gas (i.e., N2) used as a carrier and the air 78 flow is internally split between combustion and carriage operations. This implies that the ratio of comburent 79 and carrier gas is not controllable, and the user can only adjust the comburent to fuel ratio. 80 The efficiency of the combustion process can be given in terms of the global equivalence ratio, starting 81 from the air-to-fuel ratio ( nm) while fuel-rich flames lead to an additional mode in the nucleation size range (i.e., 10-30 nm). Finally, 105 Mamakos (2013) reported that low fuel-to-air ratios (i.e., ϕ < 1) generate particles with a large fraction of EC 106 while semi-volatile organics are generated by high fuel-to-air ratios (i.e., ϕ > 1). In this work, fuel-lean 107 conditions were investigated only. 108 Since the combustion process can produce flame shapes having different characteristics, we first explored 109 the range of combustion flows from 2 to 10 lpm, in 0.5 lpm steps, and from 30 to 100 mlpm, in 5 mlpm steps, -Closed tip flame ( Fig. 1.a), which generates low concentrations of soot particles (i.e., around 10 3 # cm -3 ), 113 generally forming particle aggregates at the nozzle of the MISG.

114
-Partially Open tip flame ( Fig. 1.b), the transition between Open and Closed tip.

116
-Asymmetric flame, which shows a large variability (very short, flickering, etc) and can form particle 117 aggregates at the MISG nozzle.

118
-Curled Base flame ( Fig. 1.d), a particular shape of the asymmetric flames that can also form particles 119 aggregates at the MISG nozzle.   Control And Data Acquisition).

154
The layout of the experimental configuration adopted for the MISG characterization is shown in Fig. 2.

155
The MISG was warmed for about 45 minutes before injecting soot particles inside the chamber. Injection set to measure particles with mobility diameter from 34 nm to 649 nm; aerosol sample and sheath airflow rates 172 were fixed at 0.17 lpm and 1.60 lpm, respectively, while the scanning period for each cycle was 70 s. The 173 DMA unit integrates an impactor with an orifice of 0.0508 cm, resulting in cut-off capability at 50 % of 940 174 nm, useful to exclude all the particles larger than this size to enter in the column. Frequent cleaning of this part 175 was necessary to ensure proper operation and avoid clogging; at the end of each experiment, the whole 176 impactor system was cleaned using compressed air and isopropyl alcohol. 177 We corrected diffusion losses in the instrument by using the option included in the instrument software; 178 size distributions were as well corrected for multiple charges effects through the TSI proprietary software 179 (Aerosol Instrument Manager, Version 11-0-1). 180 Among the other chamber instruments, an Optical Particle Sizer (OPS, TSI Inc., Shoreview, MN, USA, 181 Model 3330) was used for short times to spot the particle size distribution in the range 0.3-10 µm. we disregarded this issue. The PAXs had been calibrated by the manufacturer. 197 In some experiments, soot concentration inside the chamber was too high to be measured directly by PAXs; 198 and a diluter (eDiluter Pro, Dekati Ltd., Kangasala, Finland) was deployed. Dry air from a cylinder was merged 199 prior to the PAXs inlet with dilution factor 1:100. Tests performed with and without the diluter demonstrated 200 a substantial reproducibility of the optical properties measured by the PAXs when the proper dilution factor is 201 considered. 202 203

Offline analysis 204
Soot particles were also collected on pre-fired 47 mm diameter quartz fibre filters (Pallflex Tissuquartz 205 2500 QAO-UP) held in a stainless-steel filter holder to allow additional offline analysis. The sampling started 206 when stable gas and particle concentration values were reached inside the chamber (i.e., about 3 minutes -207 corresponding to the chamber mixing time -after the MISG switching off): for each working condition three 208 filters with different loadings were obtained by a low-volume sampler (TECORA -Charlie HV) working at a 209 fixed sampling flow (i.e., 10 lpm during experiments without cyclone and 13.67 lpm during experiments with 210 cyclone). 211 For each sample, the EC and OC mass concentration was determined by thermal-optical transmittance 212 analysis (TOT) using a Sunlab Sunset EC/OC analyzer and the NIOSH5040 protocol (NIOSH, 1999), 213 corrected for temperature offsets. 214 Prior to EC/OC determination, particle-loaded filters were analyzed by the Multi-Wavelength Absorbance 215 Analyzer  Table 4. It is noteworthy that no 237 correlation could be found between the global equivalence ratio (ϕ) and the shape of the corresponding flame.

238
This means that the fundamental parameter of the combustion process can not be used to predict the flame 239 shape.

240
The reproducibility and stability of the MISG emissions were investigated, in terms of number 241 concentration and size distribution of the generated soot particles. Different combustion conditions were 242 selected, and four experiments were performed for each combination of air and fuel flows. We chose to keep 243 fixed the air flow to observe the differences produced by different fuel flows that correspond to different flame 244 shapes (i.e., Partially Open tip or Open tip). In each test, we recorded the values of total particle number 245 concentration, peak concentration, and mode diameter. The reproducibility was calculated as the percentage 246 ratio between standard deviation and mean value of each series of repeated experiments. With propane, mode 247 reproducibility turned out to be 6 %, while total concentration and peak concentration showed a 16 % 248 reproducibility. With ethylene, the reproducibility was 4 % and 10 %, respectively for mode and total/peak 249 concentration. In addition, we monitored the combustion gases: CO2 and NO concentration varied by about 2 250 % and 3 %, respectively with propane and ethylene. 251 252

Size distribution 268
To compare different experiments, particle concentration values were normalized to the maximum recorded 269 in the whole set of tests and therefore varied in the 0-1 range. Fig. 3 shows the result for the total particle 270 number concentration, we can notice that:

271
-At fixed air flow, the particle number concentration increases with the fuel flow (i.e., with the global 272 equivalence ratio).

273
-In the same combustion conditions (i.e., same air flow and same global equivalence ratio), ethylene generates 274 more particles than propane.

275
-With ethylene and at fixed fuel flow, the particle number concentration increases with the air flow. The same 276 holds in some cases with propane but with much smaller variations.

283
A similar comparison is shown in Fig. 4 for the particle mode diameter: while the values are basically constant 284 for ethylene, the mode diameter with propane slightly increases with air flow (at fixed fuel flow). Furthermore, 285 at each ϕ value, propane generated particles bigger than ethylene.  diameters < 200 nm, but this can be due to the specific combustion conditions (i.e., lower global equivalence 299 ratios resulting from higher air flow or lower fuel flow). 300 The mean size distributions observed at ChAMBre are given in Fig. 5, for all the selected operative 301 conditions. All the curves are normalized to the same injection time (i.e., 3 min).

306
Significant differences between the two fuels emerge when considering the particle mass concentration 307 (extended to 10 μm, including the data collected by the OPS): ethylene combustion produced a limited number 308 of big particles, likely super-aggregates, formed directly at the MISG exhaust. Kazemimanesh (2019) also 309 observed the formation of aggregates, even with smaller dimensions (i.e., about 2 μm). We calculated the 310 super-micrometric fraction of the total measured by the OPS with both the fuel (Fig. 6): this resulted to be 311 about 3% with ethylene and 0.2% with propane. Particles larger than 4 μm were about 2% with ethylene, with 312 a peak at 8 μm, and totally negligible with propane. Considering the particle volume distribution, the latter 313 difference is obviously enhanced: the super-micrometric fraction is about 99% of the total concentration with 314 ethylene and 9% only with propane. Particles larger than 4 μm contribute to the total volume (and hence to the 315 soot concentration) for about 98% and 1%, respectively with ethylene and propane.

EC/OC quantification 339
The OC/EC composition was quantified by thermal-optical analysis of samples collected on quartz fibre 340 filters during each experiment. EC:TC concentration ratios resulted to be around 0.7 and 0.9 with propane and 341 ethylene, respectively. In addition, the EC:TC concentration ratios increased with the global equivalence ratio. 342 All the results are given in Fig. 9 and 10, for experiments without and with cyclone respectively, adopting the 343 same normalization already introduced in Fig

359
The OC:EC ratio varies from 0.31 for propane to 0.19 and 0.10 for ethylene, with and without cyclone 360 respectively. In each series of experiments (i.e., air flow rate 7 or 8 lpm, ethylene or propane) the OC fraction 361 turned out to be inversely proportional to the fuel flow with a minimum at the lowest fuel flow (i.e., 70 lpm 362 with propane and 118 lpm with ethylene). This is likely due to the shape of the flame: flames generated by the 363 lowest fuel flow conditions are Partially Open tip, with less capability to generate soot particles and hence EC; 364 so that the EC:OC ratio results lower.

365
We also performed some tests adding a backup filter during the sampling to catch the volatile fraction of 366 OC. The OC concentration values measured on backup filters showed high variability, but they were 367 compatible with those on not-sampled filters. We analysed 13 blank filters from different bunches and the 368 https://doi.org/10.5194/amt-2021-345 Preprint. Discussion started: 18 November 2021 c Author(s) 2021. CC BY 4.0 License. average concentration of OC resulted <OC> = 0.5 ± 0.2 µg cm -2 while OC concentration on backup filters was 369 <OCBF> = 0.6 ± 0.2 µg cm -2 (average OC concentration on the corresponding main filters was 1.4 ± 0.7 µg 370 cm -2 ). A relationship between OC concentration on the backup filter and the global equivalence ratio was 371 instead reported in (Kazemimanesh et al., 2019). Actually, in that study the range of investigated global 372 equivalence ratio values was 0.129 < ϕ < 0.186 to be compared ϕ > 0.210 adopted in this work. 373 374

Optical properties 375
The optical properties of the MISG aerosol were determined in terms of the absorption coefficient (b_abs; 376 i.e. the absorbance per unit length) . The b_abs definition applies both to 377 measurements directly performed on the aerosol dispersed in the atmosphere (by PAXs, in this work) and to 378 off-line analyses on aerosol sampled on filters (by MWAA, in this work), provided a proper data reduction is 379 adopted (Massabò and Prati, 2021; and references therein). 380 The measured b_abs values were normalized to the total particle concentration inside ChAMBRe reached 381 in each single experiment. Absorption coefficients measured at three wavelengths by the PAXs and with the 382 cyclone mounted upstream, are shown in Fig. 11. Similar results were obtained even for experiments without  MAC values, likely due to a higher sensitivity to particle size than filter based MWAA analysis.   (i.e., 635 and 375 nm) too, is given in  are directly compared in Fig. 15, merging all the data collected by the two setups (i.e., with and without the 463 cyclone) and for the two fuels. The agreement between the two analyses turned out within 25 % and 7 %, 464 respectively without and with the cyclone.

Conclusion 489
A Mini-Inverted Soot Generator (MISG) was coupled with an atmospheric simulation chamber 490 (ChAMBRe) to compare the emissions of two fuels, ethylene, and propane. Different combustion conditions 491 (i.e., air and fuel flow, global equivalence ratio) were characterized in terms of size distribution, particle and 492 gas composition, optical properties, and EC concentration in the exhaust. 493 The MISG turned out to be a stable and reproducible soot particles source, suitable for experiments in 494 atmospheric simulation chambers. In addition, properties of emitted soot particles can be modulated by varying 495 the combustion conditions i.e., tuning the global equivalence ratio and/or varying the fuel used for combustion.

496
With equal conditions, ethylene combustion produces particles with higher number concentration and 497 smaller diameter than propane but is prone to generation of super-aggregates. These are likely formed directly 498 in the exhaust line where particles density is very high.

499
The carbonaceous compounds produced by propane are generally characterized by higher EC to TC ratios 500 than ethylene. 501 From the optical point of view, particles generated by propane turned out to be more light absorbing than 502 those formed by ethylene, although burning conditions (in terms of global equivalence ratio) were the same.

503
The values of the MAC parameter show a substantial agreement except those retrieved from the data collected 504 in the ethylene-no cyclone experiments. The latter resulted in lower MAC values, probably due to the presence 505 of super-aggregates in the chamber. 506 This work opens to new and more complex experiments. Well-characterized soot particles could be used to 507 investigate the effects that atmospheric parameters such as temperature and relative humidity can have on soot 508 particles, and also to study the interactions between soot particles and gaseous pollutants, solar radiation or 509 bio-aerosol.