Seasonal Analysis of Reduced and Oxidized Nitrogen-Containing 3 Organic Compounds at a Coastal Site

freezing water and potentially clogging the line. We also tested lower

substrates have been used in past work, which often involve solvent extraction instead of thermal 24 desorption, and subsequent analysis via GC-or LC-MS (Chu et al., 2016;Harper, 2000;Ramírez 25 et al., 2010;Ras et al., 2009). This can be more effective when compounds are thermally labile, 26 though extraction into solution can be time consuming, and selecting a safe but effective solvent 27 for extraction of a broad range of analytes presents an additional challenge. 28 Here, samples collected on pre-cleaned, cooled PEEK tubing were desorbed directly into 29 the LC flow path using the LC mobile phase (without any additional sample preparation 30 methods), then trapped/focused on the LC column, separated, and finally analyzed with  Inline LC mobile phase desorption with ESI-MS analysis provides a gentle and efficient 32 mechanism for the desorption and separation/analysis of polar analytes that are typically not GC 33 amenable. This method also avoids a separate extraction (and thus dilution step), allowing for the 34 detection of lower concentration analytes. 35

S1.2. Sampling on Cooled PEEK Collectors 36
We used PEEK tubing as an adsorptive sampler for functionalized gas-phase compounds. 37 To cool the adsorptive sampler, we inserted the PEEK tubing (1/16" OD, 0.04" ID, 2.75" long, 38 resulting in a volume of 0.06 mL) into a piece of aluminum tubing (1/8" OD, 1/16" ID) to jacket 39 the PEEK collector, and placed the ensemble into an aluminum block (2" x 1.25") through a 1/8" 40 hole cut through the center of the block. The PEEK tubing outer diameter and length were 41 selected to match the dimensions of the aluminum cooling block while maintaining a straight 42 flow path. The block was insulated and cooled to 2°C with a Peltier cooler (Custom 43 Thermoelectric, 1.5" x 1.5"). We selected 2°C to maximize trapping efficiency (due to a decrease 44 in analyte vapor pressure, discussed below), to maintain a thin liquid layer of water in the tubing, 45 but to avoiding freezing water and potentially clogging the line. We also tested lower 46 temperatures (e.g. -15°C) in follow-up laboratory experiments to inform future developments of 47 this method. Temperature was monitored by a thermocouple in the aluminum block and 48 controlled by a proportional-integral-derivative (PID) controller (Omega CN742). A pump was 49 connected downstream of the PEEK tubing and flow controller. We set flow through the PEEK 50 tubing with an Alicat mass flow controller, and limited air flow to 22 mL/min to maintain 51 laminar flow and provide sufficient time for gas diffusion to the inner tubing walls. 52 S1.3. Inline LC Mobile Phase Desorption and LC-ESI-MS Analysis 53 After sampling in the field (or in the laboratory), PEEK collectors were installed directly 54 inline with the LC column for analyte desorption and analysis. We used a multi-position valve to 55 pass flow through or to circumvent the PEEK tubing ( Figure 1 in the main manuscript). To 56 desorb analytes, we held the mobile phase composition at 95% water and 5% methanol at 0.05 57 mL/min for 20 minutes, similar to methods employed in other fields that use LC column loading 58 steps to study other forms of complex mixtures (Alvarez-Segura et al., 2016;Greco et al., 2013;59 Pyke et al., 2015). In this step, analytes were desorbed from the PEEK tubing and subsequently 60 trapped and focused on the LC column ( Figure 1Ai, Table S1 line 1); compounds with good 61 affinity to the C18 column stationary phase were successfully trapped and focused (e.g. low-to 62 moderate-polarity functionalized species with larger molecular weight), while very water soluble 63 species were not focused well (Table S2). We used a 20 minute hold time to maximize 64 desorption from the PEEK collector tubing (by allowing the solvent in the PEEK tubing to be 65 exchanged ~20 times), while minimizing early elution from the LC column during this trapping 66 step. This leveraged the very low flow rate through the column and the high water content of the 67 mobile phase, whose eluent strength was not sufficient to allow much elution from the reverse 68 phase LC column. In cases of early elution (i.e. elution from the column during the 20 minute 69 hold period with very low flow rate through the column), peak shape was very poor, and thus 70 peak identification in later analysis was challenging. Thus, when longer hold periods were tested 71 and showed more early elution from the LC column, they were not pursued further either. We 72 also note that we evaluated a shorter hold period of 5 minutes during testing, but this did not 73 show sufficient desorption from the PEEK tubing. 74 At 20 minutes, the multi-position valve was switched to position "b" (Figure 1Aii), to 75 circumvent the PEEK tubing. At this point, the flow rate was increased to 0.3 mL/min and went 76 straight to the LC column, desorbing compounds from the column stationary phase and carrying 77 them to the mass spectrometer following the gradient in Table S1. 78 An Agilent Poroshell 120 EC-C18 column (3.0 x 50 mm, 2.7 Micron) with a C18 guard 79 column (3.0 x 5 mm, 2.7 Micron) were used for analyte focusing and chromatography, in an 80 Agilent 1260 Infinity HPLC system. We note that this is a different column than the SB-Aq 81 column used for particle-phase measurements, though a comparison of analyte retention times 82 showed similar behavior across columns. 83 Ionization was performed with an electrospray source (ESI), and samples were analyzed 84 with an Agilent 6550 Q-TOF in positive ionization mode. Methanol (≥99.9% purity, Sigma-85 Aldrich) and water (Milli-Q, 18.2 MΩ·cm, <3 ppb TOC) were used as LC mobile phases, with 86 0.1% acetic acid (for HPLC, Sigma-Aldrich) as a modifier. ESI and Q-TOF parameters are 87 discussed in detail in past work (Ditto et al., 2018). In brief, the ESI source drying gas was 88 operated at 225°C and 17 L/min, with a fragmentor and capillary voltage of 365 V and 4000 V 89 respectively, and a sheath gas flow of 12 L/min at 400°C. The Q-TOF scanned for ions between 90 m/z 50-1000, at 4 spectra/second. A solution of reference masses from Agilent Technologies (5 91 mM purine and 2.5 mM HP-0921 in 95% acetonitrile (≥99.9% purity, Honeywell) and 5% water 92 (Milli-Q, 18.2 MΩ·cm, <3 ppb TOC)), was introduced to the Q-TOF throughout the entire LC 93 elution time to reduce ion mass drift. 94

S1.4. Method Evaluation 95
To evaluate desorption from the PEEK tubing over the 20 minute desorption time and the 96 effectiveness of trapping/focusing on the LC column, we spiked pieces of PEEK tubing with a 97 range of liquid standards (e.g. 10 ng/μL each of several functionalized species such as carboxylic 98 acids, phthalates, alcohols, and a range of nitrogen-or sulfur-containing compounds, selected to 99 cover a range of atmospherically-relevant functional groups and compound sizes that could be 100 observed in the gas-or particle-phase, see Table S2). This spiked PEEK tubing was then 101 desorbed immediately in the inline desorption system. 102 A comparison of peak areas from spiked PEEK experiments to peak areas from a typical 103 LC run with standard injection (i.e. without PEEK tubing in the flow path) for the same standard 104 suggests that most compounds were effectively desorbed from the PEEK during the 20 minute 105 inline desorption period (71% ± 23% recovery). Also, most compounds from the spiked PEEK 106 experiment showed a 20 minute delay from their expected retention time observed during a 107 typical LC run ( Figure 1B and Table S2). We discuss these recovery results more below. 108 To evaluate compound breakthrough, we cooled pieces of PEEK tubing to 2°C in the 109 Peltier cooler setup discussed above, and challenged them with a series of breakthrough 110 volumes. The standard shown in Table S2 was spiked into the inlet of the cooled PEEK tubing 111 with a glass syringe. Liquid standards were used here instead of gaseous standards since these 112 functionalized standards were not available to mix in a compressed gas cylinder, and 113 volatilization was avoided to limit thermal degradation or other reactive losses during 114 evaporation. Air flow at 22 mL/min was maintained through the 2°C block for 30 minutes, 1 115 hour, 2 hours, 2.5 hours, and 3 hours, and each sample was analyzed on the LC-ESI-MS system 116 with inline analyte desorption from the PEEK tubing. Analyte breakthrough became significant 117 at 2.5-3 hours. After 2 hours of air flow (equivalent to 2.64 L), breakthrough was limited to 20% 118 or less (i.e. breakthrough test peak areas were 80% ± 27% on average compared to those from a 119 typical LC run without PEEK tubing in line, though a range of behaviors was observed, see 120 Table S2). Thus, 2 hours was set as the maximum sampling time for this method. 121 There are several factors that could potentially influence weaker analyte retention or 122 recovery from the PEEK samplers. For example, compounds with higher volatility could 123 evaporate from the PEEK collector prior to LC analysis; this possible behavior is seen in cases 124 where signal was lower for the spiked PEEK test compared to the cooled breakthrough test (e.g. 125 pyrogallol, Table S2). In these cases, higher volatility species likely evaporated from the spiked 126 PEEK, as they were not distributed across the cooled PEEK tubing and thus likely did not adsorb 127 as effectively as they may have during the breakthrough test. Matrix effects with other analytes 128 in the test mixture could have also contributed to these differences. 129 Conversely, compounds with minimal functionality may have adsorbed less effectively to 130 the cooled PEEK surface during breakthrough testing (e.g. benzophenone, Table S2), some 131 compounds may have been irreversibly taken up by the PEEK surface in the absence of solvent, 132 and some compounds may have suffered from chemical incompatibility with PEEK (though 133 PEEK is generally considered to be quite inert and often is used in standard LC system flow 134 paths). These factors may have contributed to compounds with lower breakthrough testing 135 signals, but good recovery from the spiked PEEK experiment without breakthrough testing. 136 Also, after mobile phase desorption from the PEEK tubing, some analytes were poorly 137 retained in the LC column and eluted during the trapping/focusing stage (e.g. pyrogallol, xylitol, 138 mannose). These compounds thus had poor peak shape, since they eluted with very low flow rate 139 through the LC column (0.05 mL/min), and their peaks were therefore challenging to integrate. 140 As such, data from these early-eluting compounds may still be used qualitatively, though a 141 quantitative assessment of their abundances is difficult, and future work may employ other more 142 specialized columns for polar analytes to resolve this. 143 We did not observe many functionalized compounds in the VOC volatility range in our 144 ambient samples, but we expect challenges with trapping high volatility species in the cooled 145 PEEK tubing at 2°C. In past studies investigating the delay time exhibited in various types of 146 tubing by compound volatility, the volatility of individual compounds was inversely related to 147 their delay time in PEEK tubing: higher volatility species exhibited shorter delays (Deming et al.,148 2019; Pagonis et al., 2017). Past characterization of PEEK suggests that lightly functionalized 149 VOCs (they studied ketones with C0 > 10 6 μg/m 3 (Donahue et al., 2011)) are delayed by ~60 150 seconds/meter of tubing at room temperature (~295 K) (Deming et al., 2019), which would 151 translate to a delay of just ~4 seconds in the 7 cm (2.75") of PEEK used here if the PEEK were 152 held at room temperature. While this delay time would be lengthened if the PEEK were held at 153 lower temperature, e.g. the 2°C setpoint that we used in this study, this delay time is still short 154 and thus indicates that VOCs would likely not be effectively trapped in PEEK tubing used as a 155 sampler unless cooled to much lower temperatures. For comparison, by extrapolating from 156 Deming et al.'s analysis of PEEK tubing at room temperature, which showed data for some 157 IVOCs but mostly focused on VOCs, we note that less volatile IVOCs (C0 ~ 300 μg/m 3 ) and 158 SVOCs (0.3 < C0 < 300 μg/m 3 ) (Donahue et al., 2011) may remain adsorbed to tubing walls for 159 1+ hours/meter, and likely longer if cooled due to decreasing volatility with temperature (in 160 combination with possible changes in the adsorptive properties of PEEK, uptake to condensed 161 water, and the more minor effect of changes in gas-phase diffusivity with temperature (Pagonis 162 et al., 2017)). 163 Here, we demonstrate this method as a qualitative approach to probe understudied 164 functionalized gases in the atmosphere and use it to examine the presence of functionalized gas-165 phase organic compounds, but acknowledge that for highly quantitative measurements, future 166 work is required to further optimize this sampling and analysis system for compounds of interest. properties. This method could also be fine-tuned to sample and analyze select compound classes 175 of interest, for example, per-and polyfluoroalkyl substances (PFAS). As an avenue of 176 exploratory method development, we evaluated the system with a challenge mixture of PFAS 177 species to test the system's ability to sample and analyze these highly fluorinated compounds 178 (Table S3). We note that ambient PFAS measurements were outside of the scope of this study, 179 but describe this as an example of other extended applications of this sampling and analytical 180 methodology. 181

S2. Further YCFS Site Characterization 182
The Long Island Sound region is often in non-attainment for O3 in the summer months 183 due to a mix of pollutant transport up the coast from large East Coast metropolitan areas, 184 regional biogenic emissions, and summertime chemical processing. In the summer of 2018 185 during sampling, there were several high O3 events observed. In the summer, O3 mixing ratios 186 showed a strong diurnal variation, with maximum mixing ratios (computed over corresponding 8 187 hour filter sampling periods) reaching 57 ± 20 ppb on average. In contrast, wintertime O3 during 188 the sampling period was more consistent and did not exhibit the same characteristic diurnal 189 patterns driven by photochemistry. Winter maxima were 46 ± 5 ppb and showed some decreases 190 during periods of higher NOx concentrations due to O3 titration. Biogenic VOC emissions and 191 actinic fluxes were reduced in the winter, which are both crucial to O3 formation and which have 192 been also shown to extend wintertime NOx lifetimes in the Northeast U.S. (Kenagy et al., 2018). 193 This decrease in biogenic VOC contributions was observed in the gas-phase adsorbent 194 tube data, where gas-phase CH species (i.e. fully reduced hydrocarbons) played a more important 195 role in summer (24% of detected ion abundance) than in winter (18%). This difference could also 196 be related to larger contributions of fresh emissions from urban cores during the summer. The 197 greater prevalence of fresh emissions in summer led to summertime samples showing typically 198 lower average molecular weight (summer: 156 ± 58 g/mol, vs. winter 200 ± 74 g/mol, p < 0.05) 199 and predictably higher saturation mass concentrations (summer: log(C0) = 5.3± 2.1 μg/m 3 , vs. 200 winter: log(C0) = 3.8 ±2.6 μg/m 3 , p < 0.05) than wintertime samples. However, with more fresh 201 VOC emissions in summer combined with increased photochemistry, gas-phase oxygen-to-202 carbon ratios (O/C) were slightly higher in the summer relative to winter (summer: 0.2 ± 0.2, vs. liquid water using the hygroscopicity parameter approach (Petters and Kreidenweis, 2007), and 226 found a higher organic-derived aerosol liquid water content in summer, consistent with increased 227 hygroscopicty in summer. Estimated organic-derived aerosol liquid water concentrations were on 228 average 0.8 μg/m 3 in summer vs. 0.2 μg/m 3 in winter (p < 0.05). 229 With ammonium sulfate, ammonium nitrate, and sea salt concentrations from IMPROVE 230 data (assuming sea salt was dominated by sodium chloride), we used ISORROPIA to estimate 231 the contribution of inorganic-liquid water similar to past work (Slade et al., 2019). We found no 232 statistically significant difference between summer and winter inorganic-derived aerosol liquid 233 water (on average 1.6 μg/m 3 in summer vs. 1.5 μg/m 3 in winter, p > 0.05). Overall, total 234 estimated aerosol liquid water was higher in summer than in winter in the region (on average 2.4 235 μg/m 3 in summer vs. 1.7 μg/m 3 in winter, p < 0.05). While we observed more noticeable 236 indicators of aqueous-phase processing in winter despite lower calculated aerosol liquid water, 237 this is perhaps because there were fewer competing photochemical processes in winter, as 238 discussed in the main text. 239 240

S3.1. CHONS 242
Similar to the other nitrogen-containing compound classes, we observed significantly 243 more CHONS at this site than at past studied ambient sites (i.e. 20% of detected functionalized 244 organic aerosol ion abundance in summer vs. 21% in winter, Figure 3A-B). CHONS compounds 245 in summer and winter both showed a sizable contribution from sulfonamides, which contain 246 oxygen, nitrogen, and sulfur atoms (and contributed to 24-27% of CHONS species). However, 247 we also observed a wide range of other functional groups that contributed to this compound class 248 containing oxygen, nitrogen, or sulfur, suggesting that once again, this compound class was 249 composed of a combination of different functional groups and structural features ( Figure S4). 250 This is consistent with our past observations of CHONS species, which showed an important 251 contribution from sulfide groups in combination with other heteroatom-containing moieties 252 (Ditto et al., 2021). 253

S3.2. CHNS 254
While this compound class contributed minorly overall at this site (i.e. 1% in summer, vs. 255 2% in winter, Figure 3A  summer and 11% in winter containing a nitrogen atom. We note that given the relative 285 susceptibility of alkanes to fragmentation in the APCI source (Khare et al., 2019), along with the 286 configuration of the adsorbent tubes and GC, which were not optimized for light hydrocarbons 287 (Sheu et al., 2018), the contributions of CH species here were a lower limit estimate and thus not 288 our focus here. 289 Figure S6. Predicted partitioning of compounds observed in gas-phase LC-ESI-MS data. 290 Stacked bars and left axis are the same as in Figure 5A. Red triangle markers corresponding to 291 the right axis were added here, to predicted phase partitioning of the functionalized gases 292 observed from PEEK tubing samples. For these calculations, we used partitioning theory 293 (Donahue et al., 2009) and average PM2.5 concentration during the sampling period. We assumed 294 that these compounds partitioned to the particle phase via condensation onto pre-existing organic 295 aerosol, and did not include dissolution into aerosol liquid water or cloud/fog water due to 296 uncertainty in ambient water concentrations at the site at the time of sampling. This therefore 297 likely represents a lower bound estimate of the expected distribution across phases, since the 298 exact role of water in influencing partitioning at this site is uncertain. We considered two 299 extreme scenarios, where organic aerosol comprised 20% of PM2.5 and 90% of PM2.5 300 concentrations, similar to past established ranges (Jimenez et al., 2009). 301    Table S4. Fraction of volatility bin contents estimated to be in the particle phase at 300 K 320 (average summer sampling period temperature) and at 270 K (average winter sampling period 321 temperature). Volatility bins defined at the 300 K reference temperature (i.e. bins shown in 322 Figure 2) were shifted according to the Clausius Clapeyron equation from 300 K to 270 K, 323 assuming an average enthalpy of vaporization of 100 kJ/mol (Donahue et al., 2006). This 324 resulted in a decrease of approximately two orders of magnitude in saturation mass concentration 325 at 270 K relative to 300 K. Partitioning to a pre-existing condensed phase was estimated 326 according to Donahue et al. (Donahue et al., 2006) using average summertime and wintertime 327 PM2.5 measurements from the site and C* defined at 300 K for summer and 270 K for winter. 328 Note, the volatility bins are held at a reference condition of 300 K for comparison between 329 seasons in Figure 2, and are similarly shown as C* at 300 K here. 330 C* (µg/m 3 at 300 K) Fraction in particle phase at 300 K