Response of black carbon and aerosol absorption measuring instruments to laboratory-generated soot coated with controlled amounts of secondary organic matter

We report on an inter-comparison of black carbon and aerosol absorption measuring instruments with laboratorygenerated soot particles coated with controlled amounts of secondary organic matter (SOM). The aerosol generation setup consisted of a miniCAST 5201 Type BC burner for the generation of soot particles and a new automated oxidation flow reactor based on the micro smog chamber (MSC) for the generation of SOM from the ozonolysis of α-pinene. A series of test 20 aerosols were generated with elemental to total carbon (EC/TC) mass fraction ranging from about 90 % down to 10 % and single scattering albedo (SSA) from almost 0 to about 0.7. A dual-spot aethalometer AE33, a photoacoustic extinctiometer (PAX, 870 nm), a multi-angle absorption photometer (MAAP), a prototype photoacoustic instrument and two prototype photo-thermal interferometers (PTAAM-2λ and MSPTI) were exposed to the test aerosols in parallel. Significant deviations in the response of the instruments were observed depending on the amount of secondary organic coating. We believe that the 25 setup and methodology described in this study can easily be standardized and provide a straightforward and reproducible procedure for the inter-comparison and characterisation of both filter-based and in situ BC-measuring instruments based on realistic test aerosols.

ambient air are usually internally mixed with organic and/or inorganic species, which may cause absorption enhancement through the so-called "lensing effect" (Liu et al., 2015;Cappa et al., 2012).
Despite a plethora of commercially available BC-monitoring instruments based on different measurement techniques, 35 quantification of BC mass concentration remains to this day a challenge. Deviations between 15 % and 30 % among instruments of the same type (Müller et al., 2011a;Cuesta-Mosquera et al., 2021) and up to 50-60 % for instruments of different measurement principle (Chirico et al., 2010;Slowik et al., 2007) have been reported. Among all commercial BC monitors, filter-based absorption photometers, such as the aethalometer, are the most widely used at air quality monitoring stations thanks to their robust design. At the same time, these instruments are the most prone to measurement artefacts due to 40 the use of filters for collecting the particles. Even though correction algorithms have been proposed for minimizing measurement biases (see (Collaud Coen et al., 2010) and references therein; (Drinovec et al., 2015) for a measurement of the loading bias), no satisfactory solution has been found for quantifying the absorption coefficient or for determining siteindependent equivalent mass BC (eBC) concentrations.
To compare the performance of different instruments or to investigate unit-to-unit variability, several field and laboratory-45 based inter-comparisons of BC-monitoring instruments have been conducted in the past. Slowik et al. compared a single particle soot photometer (SP2), a multi-angle absorption photometer (MAAP), and photoacoustic spectrometer (PAS) with uncoated soot generated by a McKenna burner and soot coated with organic material, such as oleic acid and anthracene . In another study, soot generated by a McKenna burner was coated with sulphuric acid and dioctyl sebacate (DOS), and the effect of non-absorbing coatings on the response of filter-based and in situ BC-measuring 50 instruments was determined (Cross et al., 2010). Holder et al. compared an SP2, a three-wavelength photoacoustic soot spectrometer (PASS-3) and an aethalometer (AE-42) during on-road and near-road measurements (Holder et al., 2014) while Tasoglou et al. compared six commercially available BC-measuring instruments using aerosols from biomass burning (Tasoglou et al., 2018). Moreover, two workshops with a large set of aerosol absorption photometers were conducted in 2005 and 2007, revealing a large variation in the response to absorbing aerosol particles for different types of instruments (Müller 55 et al., 2011a). More recently, an inter-comparison of 23 aethalometers was carried out with synthetic particles (soot generated by a miniCAST burner, nigrosin particles) and ambient air to investigate the individual performance of the instruments and their comparability (Cuesta-Mosquera et al., 2021).
Experiments in large-scale smog chambers are also conducted to simulate atmospheric ageing of soot particles and investigate the response of the instruments to secondary organic coating (Weingartner et al., 2003;Cappa et al., 2008;60 Chirico et al., 2010). Whilst smog-chamber studies allow for controlled laboratory experiments with realistic test aerosols, they are time-consuming, with each measurement ranging up to a few days (Weingartner et al., 2003). Consequently, such experiments are typically restricted to the generation of a single or a limited number of test aerosol types. A reliable inter- Recently, a compact and user-friendly setup based on a miniCAST combustion generator and an oxidation flow reactor (OFR) known as micro smog chamber (MSC) was proposed for the controlled generation of fresh and aged soot particles in the laboratory (Ess et al., 2021a). A series of test aerosols simulating a wide range of optical properties and elemental to total carbon (EC/TC) mass fraction could be generated within a few hours as opposed to a few days with conventional smog chambers. Compared to other OFRs reported in the literature (Kang et al., 2007;George et al., 2007), the MSC is designed to 70 operate at much higher aerosol loads which can subsequently be diluted, thus generating aerosols at high flow rates but still sufficiently high number concentrations to simultaneously feed multiple devices.
This study moves beyond the work by Ess et al. (Ess et al., 2021a) by demonstrating in practice how soot particles coated with controlled amounts of secondary organic matter (SOM) from the ozonolysis of α-pinene can be used to challenge a large number of BC-measuring instruments in parallel. More specifically, a dual-spot aethalometer, a photoacoustic 75 extinctiometer (PAX, 870 nm), a MAAP, a prototype photoacoustic instrument (PAS) and two prototype photo-thermal interferometers (PTAAM-2λ and MSPTI) were exposed to a series of aerosols with EC/TC mass fraction ranging from > 90 % down to 10 % and single scattering albedo (SSA) from almost 0 to about 0.7. The PTAAM-2λ is now commercially available (Haze Instruments, 2021) and this is the first time it has been compared to a range of established instruments. We believe that the setup and methodology described in this study can easily be standardized and provide a straightforward and 80 reproducible procedure for the inter-comparison and characterisation of both filter-based and in situ BC-measuring instruments based on realistic test aerosols.

Aerosol generation
Soot particles were generated by a miniCAST 5201 Type BC (Jing Ltd., Switzerland), hereafter referred to simply as 85 miniCAST BC, as previously described in Ess et al. (Ess et al., 2021b;Ess and Vasilatou, 2019). Two operation points in the "premixed flame mode" were used, both resulting in particles of roughly 90 nm mobility diameter (see Table S1 for gas flows if the operation points). The sample flow was dried using a diffusion dryer (Silicagel orange Perlen, Dry & Safe GmbH, Switzerland).
A novel "organic coating unit" (OCU, FHNW, Switzerland, (Keller et al., 2021a, b)) was used to coat soot particles with 90 secondary organic matter. The process is described in (Ess et al., 2021a), however, the OCU combines an optional humidifier (not used in this study), a dosing system for up to two volatile organic compounds (VOC1 and VOC2, see were held at constant concentration using the integrated dosing system and built-in photo-ionisation detector (PID-A1 Rev 2, Alphasense Ltd, UK). The PID sensor was regularly calibrated using a 100 ppm isobutylene-air mixture. The OCU was only 95 used for the coated operation points, for the uncoated operation points an identical setup without the OCU was used.
Two variations of the setup were used in this study. For "Setup 1" the soot aerosol generated by the miniCAST BC was delivered undiluted to the OCU while for "Setup 0.1" the aerosol was diluted at a 1:10 ratio with dry air (VKL 10 dilution unit, Palas GmbH, Germany) as shown in Fig. 1. The aerosol relative humidity before coating was about 25 % and <5 % for Setup 1 and 0.1, respectively. 100 After generation, the aerosol was further diluted by a rotational diluter (MD19, Matter Engineering AG, Switzerland) using different dilutions depending on the experiment. As the sample flow after the diluter (9.5 L min -1 ) was not enough for all the instruments under test, an additional dilution stage was built using dry filtered air provided through an MFC and a mixing volume. The aerosol was then first split up to the high-volume instruments (MAAP and nephelometer) before using a second 19-port flow splitter for all the other instruments (see Fig. 1 for a schematic overview). The design of the custom-made flow 105 splitter is shown in Fig. S1 of the supplementary information. The splitter bias was determined as per (ISO 27891, 2015) and found to be around 1 %.

BC-and aerosol-absorption-measuring instruments
The dual-spot aethalometer (AE33 aethalometer, Magee Scientific, Berkeley, USA) is a filter-based absorption photometer (Drinovec et al., 2015). It measures at seven different wavelengths (370-950 nm). To correct for filter-loading artifacts, the device measures the change of light attenuation at two distinct filter spots loaded at different flow rates. A standard multiple-115 scattering parameter C=1.39 (provided by the manufacturer) was applied to obtain the absorption coefficient from the https://doi.org/10.5194/amt-2021-214 Preprint. Discussion started: 21 July 2021 c Author(s) 2021. CC BY 4.0 License. measured attenuation coefficient. The aethalometer was operated at a sample flow of 2 L min -1 and a temporal resolution of 1 min. Absorption Ångström exponents (AAE) were calculated using absorption coefficients measured by the aethalometer over all wavelengths (Drinovec et al., 2015).

120
The Thermo Scientific Model 5012 Multi-Angle Absorption Photometer, MAAP is a filter-based instrument that measures aerosol absorption at a nominal wavelength of 670 nm. The filter loading-related artifacts affecting the determination of absorption coefficient are taken into account in the design of the instrument. This is done by incorporating light transmittance and reflectance measurements at multiple angles and by implementing a radiative transfer calculation in the internal programming of the instrument. The absorption coefficient babs for MAAP has been derived throughout the 125 manuscript by where MBC is the mass concentration of black carbon, QBC is the specific absorption coefficient of 6.6 m 2 g -1 of MAAP, and 130 1.05 is a factor to correct the absorption coefficient to the true wavelength of the instrument light source, 637 nm (Müller et al., 2011a).
Periodic variations in the babs measurements performed with the MAAP were observed especially at high black carbon concentrations during the campaign (see Fig. S2 for an example). While the MAAP is known to suffer from a measurement 135 artefact occurring at high concentrations (Hyvärinen et al., 2013), the observed variations were unrelated to it. While the exact reason for the variations is not known, they occur mid-range of the MAAP spot collection duration, and thus seem to be instrument-dependent.
A photoacoustic extinctiometer PAX (PAX 870 nm, Droplet Measurement Technologies Inc., Boulder, Colorado, USA) was 140 also used. The PAX measures absorption and extinction in parallel by combining a photoacoustic cell with an integrating nephelometer (Arnott et al., 1999). The PAX was operated at a sample flow of 1 L min -1 and an averaging time of 1 min with a 1 min zero measurement every 5 min.
The prototype photoacoustic sensor (PAS) from the FHNW group uses three different wavelengths (445 nm, 520 nm, 145 638 nm, ~300 mW each) for in-situ light absorption measurements. The diode-lasers are guided into a metallic resonator with elliptical cross-section along a focal point and are modulated at ultrasonic frequencies (~23.7 kHz) sequentially every minute for each wavelength. The modulation frequency adapts every 5 to 10 min (thermal drift) to match the resonance frequency of the resonator cell. The resulting standing wave is measured with a digital microphone placed in the middle of the resonator cell (long axis) at the other focal point of the ellipse. The signal is then preamplified and demodulated with a 150 Stanford SR850 Lock-In Amplifier (Stanford Research Systems, Sunnyvale, California, USA). The instrument was calibrated using nitrogen dioxide (NO2) and operated at 1 L min -1 .
The photo-thermal interferometer PTAAM-2λ is based on a folded Mach-Zender interferometer design (similar to (Moosmüller and Arnott, 1996;Sedlacek, 2006;Visser et al., 2020)). The He-Ne probe laser beam is split into the sample 155 chamber and reference beams. Pump lasers at 532 and 1064 nm are modulated at different frequencies and focused in the sample chamber using an axicon for concurrent measurement of the same sample. The quadrature point is maintained using a pressure cell. The interferometer signal is detected by two photodiodes and resolved by a dual-channel lock-in amplifier measuring at the two respective frequencies. The green channel is calibrated using NO2. The calibration is transferred to the infrared (IR) channel using aerosolized nigrosin and its relative green-to-infrared absorption ratio, determined using a Mie 160 calculation based on size distribution measurements. The verification at 532 nm shows a 6 % difference between the Mie calculation and the calibrated measurements of the absorption coefficient (Drinovec et al., , 2021. The photo-thermal interferometer MSPTI is an improved version of the instrument presented in . Briefly, the instrument design is similar to a Mach-Zehnder interferometer (Moosmüller and Arnott, 1996;Sedlacek, 2006), with the 165 optical elements in the interferometer consisting of a combined beam splitter and mirror block and a retroreflector. In contrast to the PTAAM-2 and other photo-thermal interferometers, the MSPTI operates with only a single modulated laser (Nd:YAG, 532 nm), which is employed as both the pump and probe beam. This beam is split 50-50 and one of the resulting beams is sent through the sample chamber, whereas the other traverses the reference chamber. The beams are recombined at the beam splitter, resulting in interference patterns. In these experiments the filtered sample (HEPA grade absolute filter) is 170 employed as the "zero" sample in the reference arm of the interferometer. Phase quadrature is maintained via an improved version of the pressure cell from . The MSPTI is calibrated using NO2 and was operated at a flow rate of 0.25 L min -1 .

Additional aerosol characterization
Mobility size distribution and number concentration were measured using a scanning mobility particle sizer SMPS 175 47 mm QR-100 Quartz fibre filters (Advantec, Japan) were prebaked at 500 °C for 1.5 hours before filter collection.
Aerosols were sampled on three sets of two superimposed filters for each measurement point according to path b) in Fig. 1. 195 For the coated samples, the aerosol was passed through an activated charcoal denuder first. During sampling the filters placed in a metallic filter holder (Merck Millipore, Germany). Punches of 1.5 cm 2 were later used for thermal-optical analysis.
Thermal-optical analysis was performed on the filter punches in order to establish the composition of the soot and the 200 amount of coating using a Lab OC-EC Aerosol Analyser (Sunset Laboratory Inc., Hillsborough, USA). This instrument distinguishes carbonaceous material into EC and OC (elemental and organic carbon) after being calibrated against solutions of glucose at different concentrations. The EUSAAR2-protocol (Cavalli et al., 2010) was slightly modified by extending the last temperature step to ensure that the evolution of carbon is complete (Ess and Vasilatou, 2019). Two superimposed filters were used. OC and EC masses were determined from the upper filter, with OC masses then corrected by subtracting the mass 205 of OC from the lower filter, consisting of the absorbed gas phase (Mader et al., 2003;Moallemi et al., 2019). The results of the thermal-optical analysis were then used to calculate EC/TC and OC/TC ratios, where TC = OC + EC.

Results and discussion
Two series of test aerosols were generated as summarized in Table 1 Table S1 in the supporting information. The decrease in GMDmob, despite the considerable amount of OC condensed on the soot particles, is not surprising. As explained in Ess et al., this is due to: i) the decrease in dynamic shape 215 factor that dominates over the increase of volume equivalent diameter and/or ii) a restructuring of the soot core during SOM condensation (see (Ess et al., 2021a) and references therein). The uncertainty of the corrected OC and EC masses is based on the uncertainties given by the instrument's software, calculated as the detection limit of 0.2 µg C cm -2 plus 5 % of the carbon mass determined in the analysis for each carbon 225 fraction. The uncertainties due to the determination of the split point were not taken into account as they could not be quantified. 4 Measured right after the mixing volume.
On the contrary, with Setup 0.1, i.e., including a dilution of factor 10 upstream of the OCU, the GMDmob of the soot particles 230 increased from 88 nm to 126 nm while the EC/TC mass fraction dropped from ~85 % to ~10 % and the SSA increased up to ~0.7. Due to the lower concentration of soot particles by an order of magnitude, the α-pinene/eBCPAX mass ratio rapidly increased to ~500 (see Table S1). As a result, the increase in volume equivalent diameter due to the high amount of condensed SOM dominated over the decrease of shape factor. The mobility size distributions of the test aerosols are displayed in Fig. 2b.  based on measurements with soot particle-aerosol mass spectrometry (Cappa et al., 2012;Liu et al., 2015). To facilitate 245 comparison with previous literature, the total mass to BC mass ratio (Mtotal/MBC) as measured by the TEOM is shown on the secondary x-axis. All babs values have been converted to a wavelength of 532 nm using the absorption Ångström exponents determined from the fit over all babs values from the aethalometer. The absorption enhancement at 532 nm (Eabs, 532) is shown in Fig. 3c, and is equal to babs of the coated soot divided by that of the uncoated soot. The EC/TC mass fraction and SSA of the test aerosols are displayed in Fig. 3a (main and secondary y-axis, respectively) while GMDmob is shown as label on the 250 data points.
As shown in Fig. 3b, significant deviations in the response of the different BC-and absorption-measuring instruments are observed even for the uncoated soot aerosol. Instruments based on photoacoustic spectroscopy and interferometry report a babs in the range 20 to 50 Mm -1 while the MAAP and AE33 report ~70 Mm -1 and ~110 Mm -1 , respectively. The largest 255 deviation is observed between the PAS and the AE33, with the AE33 overestimating babs by a factor of 4 to 5 compared to the PAS. In general, babs increases with increasing SOM coating, apart from the PAX which is insensitive to coating and the PAS which shows a rather erratic behaviour. The deviation between the AE33 and PAS increase with increasing SOM coating up to a factor of 6 to 7 for the "thickest" coating (SSA ≈ 0.7, see Table 1 and Fig. 3a).
In the visible and near UV region of the spectrum, the values of Eabs include effects of both "lensing" and absorption by 260 SOM. Instruments measuring in the wavelength region 520-637 nm all recorded an increase in Eabs, 532 as a function of RBC https://doi.org/10.5194/amt-2021-214 Preprint. Discussion started: 21 July 2021 c Author(s) 2021. CC BY 4.0 License. (Fig. 3c). At RBC ≈ 3.4, corresponding to an EC/TC mass fraction of 10 % and an SSA of about 0.7, an absorption enhancement in the range 1.1 (PAS 520 nm) to ~ 2 (MSPTI 532 nm) was observed.
As expected based on the babs measurements, no absorption enhancement was reported by the PAX at 870 nm. This is in good agreement with Cappa et al., who reported that BC emitted from large to medium-sized urban centres (dominated by 265 fossil fuel emissions) does not exhibit a substantial absorption enhancement when internally mixed with non-BC material (Cappa et al., 2012). Eabs during both field campaigns exhibited minimal dependence on RBC, with Eabs, 532nm remaining close to 1 (absorption was measured by photoacoustic spectroscopy). These results suggest that the absorption enhancement observed by the AE33, MAAP, PAS, PTAAM and MSPTI could be due to light absorption by SOM.   Table S3. The error bars correspond to one standard deviation of the mean (k=1; 68 % confidence interval; n=100-160).
280 Figure 4 shows the response of the BC-and absorption-measuring instruments to the test aerosols generated by Setup 1. In this case, coating is more moderate and, as explained above, the GMDmob of the soot particles decreases slightly upon coating. AE33 overestimates babs by up to a factor of 2 to 3 compared to the other instruments as shown in Fig. 4b. The 285 instruments report no or only a weak absorption enhancement as a function of SOM coating (Figure 4c Table S3. The error bars correspond to one standard deviation of the mean (k=1; 68 % confidence interval; n=120-180).

300
Two photo-thermal instruments based on different designs (MSPTI and PTAAM) were operated in parallel to measure the aerosol absorption coefficient. Here, we compare the measurements performed at 532 nm. The response of the PTAAM was regularly tested during the campaign and showed average variation of 3% for the 532 nm channel (Supplement S5). Testing of the MSPTI response showed larger variability at the end of the measurement campaign as the laser became more unstable.
This especially affected the measurements of the uncoated particles, which were performed at the end; due to this fact the 305 MSPTI to PTAAM ratio for the uncoated particles is more uncertain (Fig. S4a). Two additional one-day experiments were performed comparing different coating treatments ( Figure S4b). These measurements show the opposite behaviour for the uncoated particles compared to Figure S4a. Comparing the experiments one can conclude that the average response of both instruments agree well within the measurement uncertainty, thus showing similar absorption enhancement at 532 nm for both instruments. 310 To decouple a possible "lensing" effect from the light absorption by SOM, the absorption enhancement in the near infrared (NIR) region Eabs, 950 is plotted as a function of RBC in Fig. 5. SOM does not absorb in the NIR region; thus, any absorption enhancement would be due to the "lensing" effect. In this study, the only instruments measuring in the NIR were the AE33 (950 nm), the PAX (870 nm) and the PTAAM (1064 nm). Prior to the calculation of Eabs, all babs values had been converted 315 to a wavelength of 950 nm using the absorption Ångström exponents determined from the pair of babs values at 880 nm and 950 nm reported by the aethalometer. As shown in Fig. 5, the measurements by the PTAAM and PAX agree very well and both instruments yield an Eabs, 950 close to 1. This is in agreement with the findings of Nakayama et al. (Nakayama et al., 2014) who compared the absorption coefficient at 781 nm of ambient aerosols before and after passing through a thermodenuder. The authors found that the increase in BC light absorption due to the coating of non-refractory materials (i.e., the 320 lensing effect) was small (on average, 10 %) in summertime and negligible during wintertime.
On the contrary, the AE33 reports an absorption enhancement as a function of the organic coating, with Eabs, 950 ≈ 1.5 at RBC ≈ 3.4. It is known that the multiple-scattering parameter C of the aethalometer depends on the SSA and possibly size of the aerosol (Yus-Díez et al., 2021). As mentioned earlier, this variation was not taken into account but, instead, a fixed C value of 1.39 (provided by the manufacturer) was applied throughout this study to obtain the absorption coefficient from the 325 measured attenuation coefficient. We believe that the absorption enhancement reported by the aethalometer is an artefact arising from keeping the C value fixed. Under this assumption, it is possible to calculate new values of C (Bernardoni et al., 2021) as a function of SSA using the mean babs value of the PTAAM and PAX as a reference (Bernardoni et al., 2021):   Table S4. The error bars correspond to one standard deviation of the mean (k=1; 68 % confidence interval; n=100-160).

Conclusions
A series of test aerosols were produced using a miniCAST BC generator and a novel organic coating unit, comprising of a micro smog chamber and an integrated dosing system for VOC. Both uncoated soot particles and soot particles coated with 340 varying amounts of α-pinene derived SOM were generated covering a wide range of particle sizes (83-126 nm), EC/TC mass fractions (10-91 %) and optical properties (SSA almost 0 to 0.7).
Several BC-and aerosol-absorption-measuring instruments were compared using these aerosols: A dual-spot aethalometer, a photoacoustic extinctiometer (PAX, 870 nm), a MAAP, a prototype photoacoustic instrument and two prototype photothermal interferometers (PTAAM-2λ and MSPTI). This is the first time that the PTAAM-2λ, which is now commercially available, has been compared to other absorption-measuring instruments. In general, the filter-based instruments (AE33 and MAAP) overestimated babs compared to in situ measuring instruments. The bias is systematic and increases rapidly with increasing SSA. The absorption enhancement is equally highest for the filter-based instruments. On the other hand, the PAX and the NIR channel of the PTAAM measured almost no enhancement, indicating that any observed absorption enhancement was most probably caused by light absorption by the SOM coating rather than being due to any 350 "lensing" effect.
The setup of miniCAST combined with the novel organic coating unit and the methodology described in this study provide a straightforward and reproducible procedure for the inter-comparison and characterisation of both filter-based and in situ BCmeasuring instruments. The system is very robust, compact, relatively inexpensive and allows to generate realistic test 355 aerosols in a reproducible and standardized manner. Additionally, in comparison with smog chambers, stability of the aerosols is reached within minutes after changing operation points, allowing for several measurements within a day.

Data availability
Measurement data will be made available for the final publication at https://zenodo.org/communities/aerotox/ .

Author contribution 360
DMK and KV designed the study and wrote the manuscript with contributions from all authors. DMK generated the model aerosols and analysed the data of the AE33 and PAX, GM and LD operated and analysed the data of the PTAAM, BV and JR operated and analysed the data of the MSPTI, MO operated and analysed the data of the PAS, APH analysed the data of the MAAP and nephelometer. All authors contributed to the interpretation of the results.

Funding
This work has received funding from the EMPIR 18HLT02 AeroTox and 16ENV02 Black Carbon projects. EMPIR is cofinanced by the Participating States and from the European Union's Horizon 2020 research and innovation programme. This research has also been supported by the Swiss National Science Foundation (grant no. 200021_172649) and the EUROSTARS programme (IMALA, grant no. 11386