2.1 Environmental chambers and experimental design

The data for this study were acquired during a 2-month measurement campaign (SOA15) in October and November 2015 at environmental chambers of the Institute of Meteorology and Climate Research (IMK) at the Karlsruhe Institute of Technology (KIT). The measurement campaign investigated yields, physical properties, and chemical composition of SOA from α-pinene ozonolysis as a function of precursor concentration, temperature, RH, and the ice nucleation abilities of the SOA particles (Wagner et al., 2017). The focus on ice cloud formation allowed for the investigation of the particles' physicochemical properties at temperatures as low as 223 K (representative of conditions in the upper troposphere at 8–12 km altitude at the mid-latitudes), a range where detailed characterization is largely missing. Here, we discuss a subset (Table 1) of the large dataset of the SOA15 campaign that is based on experiments investigating the influence and mutual interaction of particle chemical composition and viscosity shortly after SOA formation and after a residence time of ∼ 3.5 h. Particles were formed at different temperatures (223–296 K) and RH (< 1–61 %) conditions using both environmental chambers available at IMK (see Fig. 1).

Figure 1Simple schematic and conditions for the two types of experiments (modified from Wagner et al., 2017). Both chambers at IMK (APC and AIDA) were used in this study. Instruments are annotated in green, blue, and orange, and precursor gases in red. More detailed information on the instruments and precursor gases are explained in the text.

Download

The AIDA (Aerosol Interaction and Dynamics in the Atmosphere) aerosol and cloud chamber is an 84.3 m³ sized aluminum vessel. It can be operated in the temperature range of 183 to 333 K, pressure range of 1 to 1000 hPa, RH from close to 0 to 200 %, and at different warming and cooling rates (Schnaiter et al., 2016; Möhler et al., 2003; Saathoff et al., 2009).

The APC (Aerosol Preparation and Characterization) chamber (Möhler et al., 2008) is a 3.7 m³ sized stainless steel vessel, situated at a distance of 3 m from AIDA and connected to it with a 7 m stainless steel tube of 24 mm inner diameter. The APC chamber can only be operated at room temperature (296 K) and was used to prepare SOA particles in a reproducible manner (Wagner et al., 2017).

Figure 2(a) Particle mass concentrations derived from SMPS size distributions (blue circles), CHOI mass concentrations measured by CIMS (red triangles), and organic mass concentrations measured by AMS (green circles) representative of type 1 experiments (here CH), (b) representative of type 2 experiments (here WDtoCH). Data were not wall-loss-corrected. t0 and t1 indicate points in time used for comparisons in this study. Averaged size distributions measured by AMS at t0 (c) and (d) t1 for the four experiments.

Download

We present two types of chamber experiments (Fig. 1): for the first type, SOA from α-pinene ozonolysis was directly formed at 223 K in the AIDA chamber. For the second type, SOA was first produced in the APC chamber kept at room temperature and then transferred to the AIDA chamber kept at 223 K. The second type of experiment thus represents a simplified simulation of particle formation in the boundary layer and subsequent uplifting of particles to higher altitudes with lower temperature conditions. We stress here that for both types of experiments, the particles were sampled from the cold AIDA chamber for chemical analysis. The detailed conditions for these two types of experiments are listed in Table 1. During the first type of chamber experiment, SOA was formed by reaction of an excess of O₃ (initially 2.3 ppm generated by silent discharge in pure oxygen, Semozon 030.2 discharge generator, Sorbios GmbH) with α-pinene (initially 0.714 ppm, 99 %, Aldrich) in the dark AIDA chamber at 223 K at 61 % RH (experiment termed “cold humid”, CH) or 6 % RH (experiment termed “cold dry”, CD) conditions. For the second type of chamber experiment, SOA was first formed with an excess of O₃ (initially 1.8 ppm) and 2.2 ppm α-pinene in the dark APC chamber at room temperature (296 K), < 1 % RH (experiment termed “warm dry”, WD) or 21 % RH (experiment termed “warm humid”, WH) conditions. After a residence time of 1–1.5 h in the APC chamber, its pressure was increased by 5 hPa compared to AIDA, and a fraction of the SOA particles was then transferred to the dark AIDA chamber kept at 223 K at 61 % RH (WDtoCH) or 30 % RH (WHtoCH), respectively, resulting in particle number concentrations ranging between 1500 and 2200 cm⁻³ in the AIDA chamber. No OH scavenger was used during SOA formation, and RH was kept constant in AIDA during the course of the experiments. The time series of total particle mass for experiment type 1 (particles formed in situ in AIDA, CH) and experiment type 2 (aerosols formed in APC and transferred to AIDA, WDtoCH) are shown in Fig. 2. The times t0 (right after SOA formation in CD and CH, or SOA transfer in WDtoCH and WHtoCH) and t1 (∼ 3.5 h later) indicate the points in experiment time which were used for the investigation of the physicochemical evolution of α-pinene SOA.

2.2 Temperature and relative humidity measurements

Temperature (T) in the AIDA chamber was measured by in-house thermocouples (NiCrNi) and in-house PT 100 temperature sensors with an accuracy of ±3 %, which are regularly calibrated with reference sensors traceable to standards of the National Institute of Standards and Technology (NIST). Under static conditions, gas temperature in the AIDA chamber deviated by less than 0.3 K in time and in space. Water vapor concentrations in the AIDA chamber were measured by a in-house tunable diode laser (TDL) spectrometer with an accuracy of ±5 % (Fahey et al., 2014; Skrotzki et al., 2013) and by a dew point mirror hygrometer (MBW373LX, MBW Calibration Ltd.) with an accuracy of ±1 % traceable to different national metrology standards including Federal Institution of Physical Technology (PTB), National Physical Laboratory (NPL), Federal Office of Metrology and Surveying (BEV) and NIST. Both instruments agree within ±2 %. RH in the AIDA chamber was calculated using the measured water vapor concentrations and temperature based on the saturation water vapor pressures given by Murphy and Koop (2005), resulting in an accuracy of ±5 %.

2.3 Particle and gas measurements

Number concentrations of SOA particles formed in APC or AIDA were recorded with two condensation particle counters (CPC3022, CPC3010; TSI Inc.) outside the temperature-controlled housing of the chambers via stainless steel tubes extending 35 cm into the AIDA chamber. The absolute uncertainty of the number concentrations is estimated to be ±20 % by comparison of the different CPCs with each other and with an electrometer (3068, TSI Inc.). Particle size distributions were sampled in the same way from both chambers with scanning mobility particle sizers (SMPS; differential mobility analyzer (DMA) 3071 connected to a CPC 3010, TSI Inc.). Mass concentrations were derived from integrated number size distributions and their conversions to mass using their corresponding calculated particle density (1.3–1.5 kg m⁻³). Particle densities were calculated using the ratio of vacuum aerodynamic diameter (d_va) measured by a high-resolution time-of-flight aerosol mass spectrometer (HR-ToF-AMS, hereafter AMS; Aerodyne Research Inc.) and mobility diameter (d_m) measured by the SMPS, assuming particle sphericity (shape factor = 1). O₃ concentrations were measured by an O₃ monitor (O₃ 41M, Environment S.A.). The AMS was connected to the AIDA chamber by a stainless steel tube of 1.35 m length (flow rate 0.1 L min⁻¹, residence time 1.6 s). It was equipped with a high-pressure lens (HPL; Williams et al., 2013) and continuously measured total organic particle mass as a function of size (up to 2.5 µm particle d_va) at a time resolution of 0.5 min. Elemental oxygen-to-carbon (O : C) and hydrogen-to-carbon (H : C) ratios were derived using the EALight_1_06 procedure in the AMS data analysis software package SQUIRREL (version 1.57H; Canagaratna et al., 2015). An AMS collection efficiency (CE) of 0.4–0.5 was used, except for the CH experiment where CE was 0.7, likely due to higher particle water content (Middlebrook et al., 2012). AMS mass concentrations compare well with the total mass derived from SMPS (slopes are between 0.87 and 1.04 except for the slope of 2.2 in the CD experiment, possibly due to the lower transmission efficiency in the aerodynamic lens of the AMS for sub-100 nm particles; Pearson's correlation coefficients are between 0.87 and 0.98 for the experiments presented here). Individual organic compounds in both the gas and particle phase were measured with a Filter Inlet for Gases and AEROsols coupled to a high-resolution time-of-flight chemical ionization mass spectrometer (FIGAERO-HR-ToF-CIMS, Aerodyne Research Inc., hereafter CIMS) deploying iodide ions (I⁻) as reagent ions (Lopez-Hilfiker et al., 2014; Lee et al., 2014). During the gas-phase measurement, gases were sampled via a fluorinated ethylene propylene (FEP) tube of 0.83 m length, while particles were simultaneously collected on a Teflon (Polytetrafluoroethylene, PTFE) filter via a separate sampling port (stainless steel tube of 0.66 m length, flow rate 5 L min⁻¹, residence time 0.9 s). At regular intervals (5–20 min; see Table S1 in the Supplement), the gas-phase measurement was switched off and particles on the filter were desorbed by a flow of ultra-high-purity (99.999 %) nitrogen heated from room temperature to 200 ^∘C over the course of 35 min. The resulting mass spectral signal evolutions as a function of desorption temperature are termed thermograms (Lopez-Hilfiker et al., 2014). Single-mode thermograms of a compound with signal maxima occurring at distinct desorption temperatures (T_max), which correlate with the compound's enthalpy of sublimation, can be used to infer its saturation vapor pressure (Lopez-Hilfiker et al., 2015; Mohr et al., 2017). Multi-mode thermograms indicate contributions from isomers having different vapor pressures, or thermal fragmentation of larger molecules during the heating of the filter (Lopez-Hilfiker et al., 2015). Integration of thermograms of individual compounds yielded their total signal in counts per deposition, which were converted to mass concentrations using a sensitivity of 22 counts s⁻¹ ppt⁻¹ (collisional limit; Lopez-Hilfiker et al., 2016). For each experiment, backgrounds were determined by sampling from the AIDA chamber before adding any precursor gases. For type 2 experiments, backgrounds were negligible with initial particle number concentrations below 1 cm⁻³. For type 1 experiments, we observed a small increase in both gas mixing ratio and particle mass (< 0.01 µg m⁻³) after O₃ addition, which was subtracted from the mass loadings presented here. However, the background and the increase induced by O₃ addition were negligible compared to the increase by the SOA mass (> 1000-fold for particle mass).

All instruments were set up at room temperature, outside the temperature-controlled housing of AIDA. Despite inlet insulation with Armaflex, we calculated a theoretical temperature increase (Fitzer and Fritz, 1989) of ∼ 15 K for the particle inlet of the CIMS (the FIGAERO filter was thus presumably at 238 K during deposition), and cannot entirely rule out partial evaporation of water or semivolatile organic compounds, which is taken into account in our interpretation of results.

3.1 Organic particle mass and size distribution

Figure 2a–b show the time series of total particle mass derived from SMPS size distributions, total organic particle mass measured by AMS, and total mass of particulate oxygenated hydrocarbons (C_x>1H_y>1O_z>1 detected as clustered with I⁻, termed CHOI compounds) measured by CIMS for both types of experiments. Figure 2a depicts the CH experiment, representative of experiment type 1, where particles were directly formed in AIDA. Figure 2b shows experiment type 2, where aerosol was formed in the APC and transferred to AIDA (here the WDtoCH example; see Table 1). Note that the data were not wall-loss-corrected. Gaps in the AMS time series were due to filter measurements. To investigate the evolution of the SOA particles' physicochemical properties with time, we chose two points in time during the experiments, t0 and t1. t0 is the first FIGAERO filter deposition from AIDA after particle formation (experiment type 1) or particle transfer (experiment type 2), while t1 is approximately 3.5 h later. Averaged concentrations of total organics and total CHOI compounds, elemental O : C ratios at t0 and t1, and an overview of the experimental conditions including temperature (T), RH, and added precursor (α-pinene and O₃) concentrations for all experiments discussed here (WDtoCH, WHtoCH, CH, and CD) are listed in Table 1. Particle size distributions measured by AMS for all four experiments at t0 and t1 are shown in Fig. 2c–d.

For SOA formed in AIDA (type 1 experiments), at t0 and t1, mean total organic mass concentrations and mean total concentrations of CHOI compounds were in the range of 67.5–440.1 µg m⁻³ and 97.8–247.6 µg m⁻³, respectively. When particles were transferred from the APC chamber (type 2 experiments), organic and CHOI mass concentrations in AIDA reached values of 48.5–64.2 µg m⁻³ and 23.3–40.7 µg m⁻³, respectively. We stress here that even though particle mass concentrations in AIDA were higher for the experiments of type 1 (particles formed at 223 K directly in AIDA), the α-pinene concentration for the type 2 experiments was higher by a factor of ∼ 3 (Fig. 2a–b and Table 1). This also led to larger particle sizes for the type 2 experiments. Due to additional α-pinene addition between t0 and t1 only for the CH experiment, we observed a step increase of total particle mass for this experiment (Fig. 2a).

The discrepancies between AMS and CIMS concentrations are likely due to the CIMS with I⁻ as reagent ion being more sensitive to more polar oxygenated organic compounds (Lee et al., 2014), and thus only a potential subset of organic compounds are measured by CIMS. Evaporation losses of particulate compounds during filter deposition in the FIGAERO may play a minor role. In addition, by using the collisional limit for the CIMS data, we apply maximum sensitivity and thus present lower limits of CHOI compounds. The differences between the AMS- and SMPS-derived mass concentrations in Fig. 2a are likely due to the lower transmission of sub-100 nm particles in the aerodynamic lens of the AMS used here. The AMS measured lower concentrations than the SMPS at the beginning of the CH experiment (Fig. 2a), when the newly formed particles were much smaller (see Fig. 2c), compared to later in the experiment when they had grown in size (see Fig. 2d). For the WDtoCH experiment (Fig. 2b) with larger particles transferred from the APC to the AIDA chamber, AMS- and SMPS-derived mass concentrations agree very well. The slightly decreasing trend observed during both experiments was due to wall losses (Donahue et al., 2012).

Figure 3CIMS mass spectra (normalized to the sum of signal of all detected CHOI compounds) of experiments WDtoCH and CD (a), WHtoCH and CD (b), and CH and CD (c) at t0. Inserts show enlarged regions of dimers (left) and trimers (right).

Download

3.2 Chemical characterization of SOA particles

3.2.2 CIMS mass spectra

Mass spectra of integrated desorptions from the CIMS are compared for the four experiments and two points in time, t0 and t1. Mass spectra shown were normalized to the sum of signal of all detected CHOI compounds. The corresponding mass loadings and sampling times (particle collection on filter) for the four experiments are listed in Table S1. Figure 3a, b, and c show a comparison of mass spectral patterns for the experiments WDtoCH and CD, for WHtoCH and CD, and for CH and CD, respectively, all at t0 (the same comparisons for t1 are to be found in Fig. S1). Overall, the mass spectral patterns across all experimental conditions and points in time were relatively similar. Monomers (C_mH_yO_z compounds, m ≤ 10), dimers (C_nH_yO_z compounds, $\mathrm{11}\le n\le \mathrm{20}$), and even trimers (C_pH_yO_z compounds, $\mathrm{21}\le p\le \mathrm{30}$) clustered with I⁻ were observed in the mass spectra at t0 and t1 for all occasions.

Monomers dominated the overall signal of detected compounds, with the largest signal at m∕z 327 (mainly C₁₀H₁₆O₄I${}_{\mathrm{1}}^{-}$, likely hydroxy-pinonic acid clustered with I⁻). As we can see from Fig. 3, relatively higher contributions of monomers were measured at t0 for experiments WDtoCH and WHtoCH compared to CD. The difference in relative monomer contributions for experiments CH and CD was less distinct. At the same time, relatively larger contributions from dimers and trimers (inserts in Fig. 3) were observed for the experiment CD (and to a lesser extent for the CH). This was also the case for t1 (Fig. S1).

Figure 4 shows the relative mass contributions of monomers and adducts (this definition includes dimers, trimers, and oligomers in general) for the four experiments at both time points. As already observed in the mass spectral patterns, larger relative mass contributions from monomers were measured for the type 2 experiments (WDtoCH, WHtoCH), and larger relative mass contributions from adducts for the type 1 experiments (CH, CD). There was no significant change for the relative contributions and absolute concentrations of adducts (Fig. S2) between t0 and t1 for type 2 experiments (WDtoCH, WHtoCH). For type 1 experiments (CH and CD), absolute concentrations of monomers and adducts (Fig. S2) increased from t0 to t1 due to the addition of α-pinene after t0 and hence the continuing production of oxidation products and particle mass (compare to Fig. 2). However, the relative contributions of monomers for type 1 experiments increased from t0 to t1, which may be partially influenced by smaller FIGAERO sampling time and thus less evaporation losses of semivolatiles at t1 (see Table S1 and Supplement), but mostly by increased condensation of semivolatiles or lower-molecular-weight products with increasing particle size (compare Fig. 2c–d).

Figure 4Relative mass contributions of monomers and adducts with error bars at t0 (blue) and t1 (red).

Download

Figure 5(a) Average mass-weighted number of carbon atoms (numC) and (b) oxygen atoms (numO) with error bars at t0 (blue) and t1 (red).

Download

Figure 6Thermograms of monomer C₁₀H₁₆O₄ (a) and adduct C₁₇H₂₆O₈ (b), both clustered with I⁻ at t0, and sum thermograms of monomers (c) and adducts (d) at t0. Dashed lines refer to the corresponding T_max.

Download

Figure 5 shows the average mass-weighted number of carbon atoms (numC) and oxygen atoms (numO) for CHOI compounds for the four experiments at t0 and t1. The corresponding average mass-weighted compounds' formulae for SOA generated in APC and AIDA were C₁₀₋₁₂H_yO₆₋₇ and C₁₁₋₁₃H_yO₆₋₇, respectively. Slightly bigger numC were observed for type 1 experiments (CH, CD) than type 2 experiments, with the largest value for experiment CD, followed by CH and WHtoCH. numC was smallest for WDtoCH. There was no obvious trend for numO.

In summary, smaller particles with slightly lower O : C ratios, bigger carbon numbers, and relatively more mass from adducts were observed for type 1 experiments (CH, CD), which had lower α-pinene concentrations and colder formation temperature (223 K) compared to the type 2 experiments. For type 2 experiments (WDtoCH, WHtoCH), higher α-pinene concentrations (by a factor of ∼ 3) and warmer formation temperature (296 K) produced larger particles with slightly higher O : C ratios, smaller carbon numbers, and relatively more mass from monomers. The slightly higher O : C ratio in type 2 experiments is thus not due to bigger oxygen numbers, but due to smaller carbon numbers (Fig. 5), indicating that relatively more small oxygenated molecules were formed for type 2 experiments. This is likely due to higher α-pinene concentrations and faster oxidation at 296 K leading to rapid condensation of monomers, providing enough gaseous oxidation products for the equilibrium of semivolatiles to be shifted into the particle phase. Type 1 experiments, on the other hand, were performed with lower α-pinene concentrations, and particles were formed in situ, favoring higher contributions of larger ELVOC/LVOC compounds, especially at the early stages of particle growth (Tröstl et al., 2016). At the same time, the low-temperature conditions may also have shifted equilibrium to the particle phase and led to condensation of compounds with a relatively lower degree of oxygenation (compared to warm temperature conditions). Overall, the differences observed in mass spectral patterns between the two types of experiments are a consequence of both temperature and precursor concentration differences. They underline the importance of experiment conditions when interpreting laboratory data or using them for modeling.

3.3 Thermograms: variation in T_max of SOA compounds for different experiments

In addition to information on mass spectral patterns and mass loadings when peaks are integrated, the FIGAERO also provides signal curves as a function of desorption temperature (referred to as thermograms). Although T_max can be used to infer the compound's saturation vapor pressure (Lopez-Hilfiker et al., 2015; Mohr et al., 2017), evaporative behavior and inferred volatility of a particle-bound compound are also influenced by the particles' physical phase state, particle-phase diffusivity, and viscosity (Yli-Juuti et al., 2017). Here we show that thermograms may also be used for qualitative information on particle viscosity.

Figure 7T_max distribution for individual CHOI compounds of WDtoCH (a) and CD (b) experiments at t0 according to number of oxygen atoms (numO) vs. number of carbon atoms (numC). Dashed boxes specify the compounds with nominal molecular formulae C₅₋₁₀H_yO₁₋₁₀I⁻ and C₁₅₋₂₅H_yO₁₁₋₂₀I⁻ that had bigger T_max differences.

Download

Thermograms resulting from the thermal desorption of deposited SOA particles from the four experiments CH, CD, WDtoCH, and WHtoCH at both time points t0 and t1 were analyzed. Examples of the thermograms of a monomer (C₁₀H₁₆O₄, molecular formula corresponding to hydroxy-pinonic acid identified by Zhang et al., 2017) and an adduct (C₁₇H₂₆O₈, molecular structure identified in SOA from α-pinene ozonolysis as a cis-pinyl-diaterpenyl ester by Yasmeen et al., 2010; molecular formula identified in SOA from α-pinene ozonolysis by, e.g., Zhang et al., 2015; Mohr et al., 2017), both clustered with I⁻ at t0 are shown in Fig. 6a–b. Figure 6c shows the sum of thermograms of all monomers, Fig. 6d shows the sum of all adduct thermograms at t0. The same plots for t1 can be found in Fig. S3. Thermograms and sums of thermograms were normalized to their maximum values. The corresponding mass loadings and sampling times (particle collection on filter) for the four experiments are listed in Table S1. For experiment CD, the C₁₀H₁₆O₄I${}_{\mathrm{1}}^{-}$ thermograms exhibited a multi-modal shape, indicative of contributions from isomers having different vapor pressures, or thermal decomposition of larger molecules. Different isomeric hydroxypinonic acids were found in α-pinene SOA (Zhang et al., 2017) and the decomposition of cis-pinyl-hydroxypinonyl diester could have a residue of cis-pinic acid and 7-hydroxypinonic acid (Müller et al., 2008). Based on previous FIGAERO data analyses (Lopez-Hilfiker et al., 2015; D'Ambro et al., 2017; Wang et al., 2016), we can safely presume that the first mode corresponds to the monomer.

Figure 6a–b show that T_max of an individual compound varied by up to 20 ^∘C, depending on experimental conditions. It has been shown earlier that thermograms and corresponding T_max are highly reproducible for stable conditions (Lopez-Hilfiker et al., 2014). In our instrument, T_max varied by 2 ^∘C at most for the monomer, C₁₀H₁₆O₄, and for another adduct, C₁₆H₂₄O₆ (molecular formula identified in SOA from α-pinene ozonolysis by, e.g., Zhang et al., 2015) both clustered with I⁻, for six subsequent thermograms under stable conditions (Fig. S4). The variation in T_max as a function of experiment types observed here thus indicates that the shape of a thermogram for a given compound and given FIGAERO configuration is not only defined by the compound's enthalpy of evaporation. For both C₁₀H₁₆O₄I${}_{\mathrm{1}}^{-}$ and C₁₇H₂₆O₈I${}_{\mathrm{1}}^{-}$ thermograms, T_max was highest for experiment CD, followed by WHtoCH, CH, and WDtoCH. Similar trends were observed for all compounds measured by the CIMS, as shown by the sums of thermograms of all monomer compounds (Fig. 6c), and by the sums of thermograms of all adduct compounds (Fig. 6d). Sum T_max of monomers and adducts varied from 46 ^∘C (experiment WDtoCH) to 74 ^∘C (experiment CH) to 93 ^∘C (experiment WHtoCH) to 104 ^∘C (experiment CD).

Variation in T_max of the sum of CHOI compounds was larger for monomers (Fig. 6c) than for adducts (Fig. 6d). Monomers are thus the more important contributors to the shifts in T_max, likely because at the higher temperatures where adducts desorb, particle matrix effects may become less important. Since the sum of thermograms and its T_max is highly influenced by compounds with large signal, we also show a box and whisker diagram of T_max for monomers and adducts (Fig. S5). The median T_max values showed similar variation as the T_max values based on thermogram sums. Examples of the T_max distribution of individual CHOI compounds in numO vs. numC space at t0 are shown in Fig. 7 for the WDtoCH and CD experiments. Points were color-coded by T_max. Compounds with nominal molecular formula C₈₋₁₀H_yO₄₋₆I⁻ were the main contributors to mass concentrations (data not shown), and thus also aggregated T_max values. Generally, T_max for CHOI compounds ranged from 25 to 165 ^∘C, and increased with carbon numbers and oxygen numbers of compounds, as is to be expected given the relationship between enthalpy of evaporation and volatility of a compound (Lopez-Hilfiker et al., 2015; Mohr et al., 2017). The comparison between WDtoCH (Fig. 7a) and CD (Fig. 7b) experiments, however, showed differences in T_max values for most compounds. T_max values, especially of many compounds with nominal molecular formulae C₅₋₁₀H_yO₁₋₁₀I⁻ and C₁₅₋₂₅H_yO₁₁₋₂₀I⁻, were higher for the CD experiment. The similar behavior in the variation of T_max of most compounds measured by CIMS indicates that T_max is not purely a function of a compound's vapor pressure or volatility, but is influenced by diffusion limitations within particles (particle viscosity; Vaden et al., 2011; Yli-Juuti et al., 2017), interactions between particles deposited on the filter (particle matrix), and/or particle mass on the filter. In the following we will discuss these implications in more detail.

Mass transport limitations within SOA particles, often measured or modeled as evaporation rates of specific compounds (Yli-Juuti et al., 2017; Wilson et al., 2015; Roldin et al., 2014), have been related to the particle viscosity (Vaden et al., 2011; Yli-Juuti et al., 2017). Particle viscosity is highly influenced by temperature and RH (Shiraiwa et al., 2017; Kidd et al., 2014), with higher viscosities at cool and/or dry conditions (Shiraiwa et al., 2011). Since the temperature was 223 K in AIDA for all experiments discussed here, the observed differences in T_max, and presumed viscosity, cannot be directly explained by differences in temperature. In addition, during desorption of compounds with the FIGAERO, particles are actively heated (with heat transfer assumed to be immediate), and are not evaporating under equilibrium conditions. Presumed variations in particle viscosity based on observed variations in T_max must therefore be due to variations in particle chemical composition, and/or RH differences.

The biggest T_max difference in Fig. 6 was between WDtoCH and CD experiments, which was in accordance with the largest differences in mass spectra as discussed above (see Figs. 3a and 4). This is indicative of a relationship between T_max in the thermograms and particle chemical composition. It has been shown earlier that the chemical properties of particulate compounds influence particle viscosity (Kidd et al., 2014; Hosny et al., 2016). Viscosity is expected to be higher with higher oligomer content, due to inter-component hydrogen bonding, especially at low RH (Kidd et al., 2014). This is in accordance with our results, which showed highest T_max values for the CD experiment, which also had the highest contribution from adducts.

RH is an additional parameter that greatly influences particle viscosity (Kidd et al., 2014; Hosny et al., 2016; Renbaum-Wolff et al., 2013). Despite the fact that the SOA particles might be dried very quickly by the dry heated nitrogen during particle desorption, we suppose that RH might have a “memory effect” and still influence T_max. RH conditions during the four experiments presented here ranged from 6 % (CD) to 30 % (WHtoCH) to 61 % (WDtoCH and CH). Note that these were the conditions of the measurement time in the AIDA chamber; for WDtoCH and WHtoCH, the RH conditions during SOA formation in the APC chamber were 1 and 21 %, respectively. We thus need to differentiate between RH_formation and RH_measurement. As shown in Fig. 8, there was no trend between RH_formation and T_max, indicating that the RH during particle formation did not play an important role in the observed viscosity variation. However, we observed a negative correlation of RH_measurement and T_max of all monomer compounds at t0, indicating that even under low-temperature conditions of 223 K there is particle water uptake, and an influence of RH on viscosity. Particle water uptake thus seems to influence particle viscosity even at such low temperature and on such short timescales (few hours). To what extent RH and particle water uptake, or chemical properties and adduct content, and their respective influence on water uptake via increased hygroscopicity, contribute to the observed differences in T_max and presumed viscosity, we can only speculate. In the CH and WDtoCH experiments, RH_measurement was ∼ 60 % for both. The adduct mass fraction was only slightly higher for SOA in the CH experiment, and so was T_max and thus potentially particle viscosity. More controlled studies at low temperature are needed to separate these effects.

Figure 8Relationship of RH_formation (gray), RH_measurement (red), and T_max of all CHOI monomer compounds for four experiments at t0.

Download

We also noticed that different mass loadings on the filter due to different sampling times and/or sample concentrations influenced the shape of thermograms and thus T_max. T_max increased as a function of mass loading on the filter, likely due to the increase in heat capacity of the increasing mass of the particle matrix, and potential interactions between the particles. The dependency of T_max on filter mass loading was not linear, and for our FIGAERO, it reached a plateau at mass loadings of 2–4 µg. Our results are therefore not affected by the mass loading effect, but we recommend taking it into account in analyses that involve T_max. A detailed discussion can be found in the Supplement.