Secondary Organic Aerosol Formation from Camphene Oxidation: Measurements and Modeling

While camphene is one of the dominant monoterpenes measured in biogenic and biomass burning emission samples, oxidation of camphene has not been well-studied in environmental chambers and very little is known about 10 its potential to form secondary organic aerosol (SOA). The lack of chamber-derived SOA data for camphene may lead to significant uncertainties in predictions of SOA from oxidation of monoterpenes using existing parameterizations when camphene is a significant contributor to total monoterpenes. Therefore, to advance the understanding of camphene oxidation and SOA formation, and to improve representation of camphene in air quality models, a series of experiments were performed in the University of California Riverside environmental chamber to explore camphene 15 SOA yields and properties across a range of chemical conditions at atmospherically relevant OH concentrations. The experimental results were compared with modeling simulations obtained using two chemically detailed box models, Statewide Air Pollution Research Center (SAPRC) and Generator for Explicit Chemistry and Kinetics of Organics in the Atmosphere (GECKO-A). SOA parameterizations were derived from the chamber data using both the two-product and volatility basis set (VBS) approaches. Experiments performed with added nitrogen oxides (NOx) resulted in higher 20 SOA yields (up to 64%) than experiments performed without added NOx (up to 28%). In addition, camphene SOA yields increased with SOA mass (Mo) at lower mass loadings, but a threshold was reached at higher mass loadings in which the SOA yields no longer increased with Mo. SAPRC modeling of the chamber studies suggested that the higher SOA yields at higher initial NOx levels were primarily due to higher production of peroxy radicals (RO2) and the generation of highly oxygenated organic molecules (HOMs) formed through unimolecular RO2 reactions. SAPRC 25 predicted that in the presence of NOx, camphene RO2 reacts with NO and the resultant RO2 undergo hydrogen (H)shift isomerization reactions; as has been documented previously, such reactions rapidly add oxygen and lead to products with very low volatility (i.e., HOMs). The end products formed in the presence of NOx have significantly lower volatilities, and higher O:C ratios, than those formed by initial camphene RO2 reacting with hydroperoxyl radicals (HO2) or other RO2. Moreover, particle densities were found to decrease from 1.47 to 1.30 g cm-3 as 30 [HC]0/[NOx]0 increased and O:C decreased. The observed differences in SOA yields were largely explained by the gas-phase RO2 chemistry and the competition between RO2 + HO2, RO2 + NO, RO2 + RO2, and RO2 unimolecular reactions. https://doi.org/10.5194/acp-2021-587 Preprint. Discussion started: 15 July 2021 c © Author(s) 2021. CC BY 4.0 License.


Introduction 35
On a global scale, biogenic monoterpene emissions are estimated to contribute 14% of the total reactive volatile organic compound (VOC) flux (Tg C) (Guenther, 1995). Camphene is an ubiquitous monoterpene emitted from biogenic sources (Geron et al., 2000;Hayward et al., 2001;Ludley et al., 2009;Maleknia et al., 2007;White et al., 2008) and pyrogenic sources ( Akagi et al., 2013;Gilman et al., 2015;Hatch et al., 2015). Many studies have reported camphene as a top contributor by mass in measured biogenic and pyrogenic monoterpene emissions (Benelli et al., 40 2018;Hatch et al., 2019;Komenda, 2002;Mazza & Cottrell, 1999;Moukhtar et al., 2006). For example, in measurements of laboratory and prescribed fires reported by Hatch et al. (2019), camphene was among the top two monoterpenes emitted from subalpine and Douglas fir fires based on emission factors (mass of compound emitted/mass of fuel burned). When emitted to the atmosphere, monoterpenes form oxygenated compounds through reactions with oxidants such as hydroxyl radicals (OH), ozone (O 3 ) and nitrate radicals (NO 3 ); compounds with sufficiently low volatility can then condense to form secondary organic aerosol (SOA). Figure 1 shows the chemical structure of camphene and its 50 reaction rate constants with major atmospheric oxidants. The SOA formation potential of individual monoterpenes can vary greatly based on their molecular structure, atmospheric lifetimes, and the volatility of their oxidation products (Atkinson and Arey, 2003;Griffin et al., 1999;Ng et al., 2007a;Zhang et al., 1992). Previous experimental studies of other monoterpenes (such as α-pinene, β-pinene, d-limonene, etc.) have reported SOA mass yields from ~10% to 50% through OH oxidation and from ~ 0 to 65% through NO 3 oxidation; among the studied monoterpenes, d-limonene 55 often has the highest reported yields (Mutzel et al., 2016;Griffin et al., 1999;Ng et al., 2007b;Fry et al., 2014). Few studies have been published regarding camphene SOA formation.
Past experimental studies of camphene largely have been focused on gas-phase reactivity with OH, NO 3 , and/or 4 basis set (VBS) modeling approaches (Donahue et al., 2006;Donahue et al., 2009). Two chemically detailed box models, Statewide Air Pollution Research Center (SAPRC) and Generator for Explicit Chemistry and Kinetics of Organics in the Atmosphere (GECKO-A), were used to provide mechanistic insights into the chamber observations and to elucidate the connections between the fate of RO 2 , HOM forming mechanisms, and camphene SOA formation.

Chamber Facility and Instrumentation
The camphene photooxidation experiments were conducted in the University of California, Riverside (UCR) dual indoor environmental chamber. Chamber characterization and features have been previously described in detail (Carter et al., 2005). Briefly, the UCR environmental chamber consists of two 90 m 3 collapsible Teflon reactors (2MIL (0.0508 mm) FEP film) kept at a positive pressure differential  to the enclosure where the reactors are 110 located to minimize contamination during experiments. The enclosure is relative humidity controlled (<0.1%), temperature controlled (300 ± 1 K), and continuously flushed with dry purified air (dew point < -40 °C). Prior to and between experiments, reactors were collapsed to a volume < 20 m 3 for cleaning. The cycle of filling-purging the reactors was repeated until particle number concentrations were < 5 cm -3 and NO x mixing ratios were < 1 ppb. The reactors were then flushed with dry purified air and filled up to 90 m 3 overnight. The filling-purging of the reactors is 115 controlled by an "elevator" program in LabView.
NO and NO 2 mixing ratios were monitored by a Thermo Environmental Instruments Model 42C chemiluminescence NO x analyzer. O 3 mixing ratios were monitored by a Dasibi Environmental Corp. Model 1003-AH O 3 analyzer. An Agilent 6890 gas chromatograph with flame ionization detector (GC-FID) was used to measure the camphene levels during the experiments. 120 Multiple instruments were used for particle-phase monitoring. Each reactor was equipped with a scanning mobility particle sizer (SMPS), including a TSI 3081 differential mobility analyzer (DMA), to measure the particle mass concentration. Particle effective density was directly measured by an Aerosol Particle Mass Analyzer (APM, Kanomax) with a SMPS built in house (Malloy et al., 2009). Chemical composition of SOA was measured using HR-ToF-AMS (DeCarlo et al., 2006) and analyzed to obtain O:C and H:C ratios by applying the method of Canagaratna 125 et al. (2015). Data processing was performed using the ToF-AMS Analysis Toolkit 1.57 and PIKA 1.16 on Igor Pro 6.36. A prior characterization of this UCR chamber system (Li et al., 2016) reported an experimental uncertainty in SOA yields of < 6.65%.
Particle wall loss corrections were performed following the method described in Cocker et al. (2001). Vapor wall loss of organics has been reported in multiple chambers (e.g., Zhang et al., 2015Zhang et al., , 2014Schwantes et al., 2019); and 130 has been modeled as a function of the mass and volatility of the condensing compounds, condensation sink, and characteristics of the chamber (e.g., La et al., 2016;Zhang et al., 2014;Ye et al., 2016). The extent to which these observations and modeling simulations are relevant in the UCR chamber is unclear, given the significant difference in chamber sizes. The UCR chamber is 4.5 times larger (90 m 3 ) than the largest referenced chamber in these studies (20 m 3 ) and most are ~10 m 3 . In the UCR chamber, the role of vapor wall loss has been investigated in SOA 135 experiments using various precursor compounds (including α-pinene and m-xylene) under seed and no seed conditions (Clark et al., 2016;L. Li et al., 2015). There has been no evidence of non-negligible vapor wall loss in those experiments, including no measurable differences in SOA formation in experiments with and without seed. In this work, stability tests on camphene demonstrated negligible vapor wall loss of the parent compound. Thus without evidence to suggest otherwise, negligible vapor wall loss was assumed for these experiments. This assumption is 140 further discussed where it may affect the major conclusions regarding the role of gas-phase chemistry on SOA formation.

Experimental Conditions
A series of 13 camphene photooxidation experiments were carried out under varying levels of camphene and NO x (Table 1). Due to the relatively high melting point of camphene (51 °C), camphene (Sigma-Aldrich, purity > 96 %, 145 FG) was injected into a glass manifold (heated to 50 °C by heating tape) using a preheated (~50-55 °C) microliter syringe. As camphene evaporated it was carried to the reactors by dry purified compressed air flowing through a glass manifold at 8 LPM for 15 mins. Injection lines from the glass manifold to the reactors were also heated to reduce losses of camphene. H 2 O 2 (Sigma Aldrich, 50 wt.% in H 2 O) was injected by adding 200 μl onto glass wool in glass tubing and then placing the tubing in a 56 °C oven with 10 LPM of dry purified compressed air flowing through the 150 tubing for 15 mins and into the reactors. An inert tracer, perfluorohexane (Sigma-Aldrich, 99 %) or perfluorobutane (Sigma-Aldrich, 99 %), was injected to the reactors through the heated glass manifold by a carrier gas of 50 °C pure N 2 . NO (Matheson, UHP) at known volume and pressure was transferred and injected through the same glass manifold as the inert tracer. When gaseous injection of camphene, H 2 O 2 , inert tracer, and NO (when used) was completed, the reactors were internally mixed using built-in blowers to ensure uniform distribution of chemicals, and then irradiated 155 using UV black lights (115w Sylvania 350BL) to start photooxidation. No seed aerosol was used in this study. All experiments were conducted under dry conditions (relative humidity < 0.1 %) at 300 K. The initial conditions of the experiments are summarized in Table 1. mixing ratio was targeted at 1ppm but corrected based on tracer (perfluorohexane or perfluorobutane) concentration to offset initial reactor volume bias. Corrected H 2 O 2 mixing ratios were used in SAPRC modeling.

Model Configurations and Conditions
The chamber experiments were modeled using two different box models, SAPRC and GECKO-A. The SAPRC model was chosen because it has been designed to evaluate gas-phase chemistry in the UCR chamber. The GECKO-A model 165 was chosen because of the ability to predict both gas and particle phase composition, and the prior work of Afreh et al. (2020), in which GECKO-A was used to study SOA formation from camphene. The initial conditions of the simulations are summarized in Table 1.

SAPRC
A gas-phase oxidation mechanism was derived using the SAPRC mechanism generation system (MechGen) with 170 modified initial rate constants (camphene with OH, NO 3 and O 3 ) based on published literature data (Atkinson and Arey, 2003). MechGen, described elsewhere (Carter, 2021;Carter, 2020b;Jiang et al., 2020), is capable of generating fully explicit mechanisms for the atmospheric reactions of many types of organic compounds and the intermediate radicals they form. MechGen uses experimentally derived rate constants and branching ratios if data are available and otherwise uses estimated rate constants and branching ratios based on group additivity and other estimation methods. 175 This system was used to derive reactions of explicit and lumped organic compounds and products in the development of the SAPRC-18 mechanism (Carter, 2020a) and a detailed SAPRC furans mechanism (Jiang et al., 2020).
The MechGen-derived camphene mechanism was implemented into the SAPRC box model to simulate chamber experiments under the same chemical conditions as the chamber experiments, where the initial hydrocarbon concentrations and NO x levels were as defined in Table 1. The SAPRC box model system has been used for chemical 7 mechanism development, evaluation, and box modeling applications since the mid-1970s (Carter, 1990(Carter, , 1994(Carter, , 2000(Carter, , 2010a(Carter, , 2010b(Carter, , 2020a. The initial conditions and relevant chemical parameters for environmental chamber experiments are required inputs; simulations can be performed using multiple versions of the SAPRC gas-phase chemical mechanism. In this work, the recently published version, SAPRC-18 (Carter, 2020a), was selected as the base mechanism because it represents the current state of the science and includes the most up-to-date model species 185 and explicit representation of RO 2 chemistry.

GECKO-A
GECKO-A is a nearly explicit mechanism generator and SOA box model. GECKO-A relies on experimental data and structure-activity relationships (SARs) to generate detailed oxidation reaction schemes for organic compounds. The generated reaction schemes are applied in the SOA box model to simulate SOA formation based on the absorptive 190 gas/particle partitioning model of Pankow (1994), where thermodynamic equilibrium between the gas and an ideal particle phase is assumed. Detailed descriptions of GECKO-A, including mechanism generation and SOA formation, are provided by Aumont et al. (2005) and Camredon et al. (2007). GECKO-A has been used to predict SOA in a number of studies (e.g., Aumont et al., 2012;Lannuque et al., 2018;McVay et al., 2016), including camphene (Afreh et al., 2020). Details of the camphene mechanism and SOA box modeling were described in Afreh et al. (2020). 195 Briefly, the camphene mechanism includes 1.3 × 10 6 reactions and 1.8 × 10 5 oxidation products; vapor pressures of products were calculated based on the Nannoolal method (Nannoolal et al., 2008).
The GECKO-A simulations were performed for a predefined set of conditions, prior to the chamber experiments, and thus in some cases differ from the experimental conditions. GECKO-A simulations were performed under two NO x conditions, with 80 ppb of NO x and without NO x (Table 1). For both NO x conditions, the initial hydrocarbon 200 mixing ratios were set at 10, 25, 50, 100, and 150 ppb. All simulations were run under the following initial conditions: 1000 ppb of H 2 O 2 , 1 µg m -3 of organic seed with molecular weight of 250 g mol -1 , 298 K temperature, 1% relative humidity, and 50 o solar zenith angle (required to compute the photolysis frequencies). Simulation results for camphene were compared with chamber data including SOA mass yields, precursor decay rates, and oxidant levels.  Fig. S6. 215 ** Peak d p refers to the diameter of particles at the peak of the size distribution plot at the end of the experiment. The uncertainty of peak d p values is less than 5%.  Fig. S1. Higher camphene decay rates and higher OH levels (0.15-0.88 ppt with added NO x ; 0.05-0.29 ppt without added NO x ) were observed and predicted for experiments with added NO x than without; likely due to the fast recycling of OH when NO x was present (Fig. 2). For all experiments, the β values changed as a function of time due to changing chemical conditions. Note that due to off-gassing of NO x from the 225 Teflon reactor (Carter et al., 2005), β values simulated here were larger than 0 even for experiments without added  (Carter and Lurmann, 1990;Carter, 1999;Carter, 2009;Carter, 2020). quantify how well constrained the other RO 2 reaction rates and product yields are without corresponding measurements, which are not available. In this case, the SAPRC model was largely used to probe the mechanism (diagnostic) and not to predict yields (prognostic).

SOA Mass and Yield
Measured SOA mass yields are shown in Fig. 3 (Krechmer et al., 2020). The observed trends in SOA mass yields were unexpected based on prior chamber studies of SOA formation from monoterpenes, such as OH oxidation studies of α-and β-pinene, in which SOA mass yields were reported to be suppressed under high-NO x conditions (Eddingsaas et al., 2012;Pullinen et al., 2020;Sarrafzadeh et al., 2016).  [OH] (1) where [OH] atm was assumed to be 2 × 10 6 molecule cm -3 . Figure 4 shows the measured SOA mass yields as a function have had higher SOA mass yields at longer irradiation times. However, even at the same aging time (Fig. S8), the SOA yields were higher in the experiments with added NO x . The higher SOA mass yields in experiments with added NO x may partially be attributed to the difference in [OH] levels and extents of aging. Similar NO x effects have been reported in many previous studies (e.g., Ng et al., 2007a;Sarrafzadeh et al., 2016). Sarrafzadeh et al. (2016) proposed that in a study of β-pinene the OH level was the main factor that accounted for differences in SOA mass yields under 280 varying [NO x ] 0 . In the camphene experiments presented herein, the aging effects were determined to be less important than RO 2 chemistry, since the SOA mass yield curves as a function of photochemical aging already plateau or nearly plateau by the end of experiments (Fig. 4) and the shapes of the growth curves ( Fig. S9 and Fig. S10) indicate different kinetics and contributions from oxidation products that form slowly among and between experiments with and without added NO x . 285   (Fig. 7), suggesting more oxygenated RO 2 s were formed by NO pathway than others, which is consistent with the formation of HOMs with added NO x . There is a general trend of increasing SOA mass yield with decreasing [HC] 0 /[NO x ] 0 ( Fig. 5 and Fig. 7 experiments would increase the measured SOA, but would not affect the following discussion or conclusions regarding the role of RO 2 chemistry. Table S1. By assuming the gas-phase chemistry and product distribution were similar when RO 2 + NO accounted for more than 80% of the total RO 2 consumption and when RO 2 + HO 2 accounted for more than 80% of the total RO 2 consumption, experiments with (W1-3, 5-6) and without (WO1-4) added NO x were grouped and used to derive SOA parameters using the two-product (Odum et al., 1996) and VBS approaches (Donahue et al., 2006;Donahue et al., 335 2009). The resultant parameters are shown in Table 3 (two-product) and Table 4 (VBS).

Discussion
The reaction rate constant of camphene with O 3 is relatively low compared to OH, and thus it is expected that OH is the dominant oxidant in the photooxidation of camphene under chamber conditions, especially with the high initial H 2 O 2 (~1 ppm) concentrations. This is supported by SAPRC simulation results (see Fig. S3 in SI), in which O 3 accounts for 0-3% and NO 3 for 0-16% of camphene oxidation, demonstrating the important role of OH oxidation in 345 these studies. Figure 8 shows the MechGen predicted reactions and products of OH-initiated oxidation of camphene in the presence of NO x through one major pathway, which had a yield of 0.83 (a more detailed reaction mechanism schematic is presented in Fig. S4). The reaction starts with OH addition to the CH 2 =(C) position to form a ring-retaining alkyl 350 radical, which further reacts with O 2 to form the camphene peroxy radical, RO 2 -a. RO 2 -a can react with oxidants (NO, NO 3 , HO 2 , and/or other RO 2 ) to create an alkoxy radical, RO-a, with NO to NO 2 conversion; or form stable products such as organic nitrate (NO3CAMP1), hydroperoxide (HO2CAMP1), and alcohol (RO2CAMP1) compounds. The cyclic alkoxy radical RO-a can undergo prompt beta (β)-scission ring-opening reaction, and then O 2 addition to form another peroxy radical, RO 2 -b. In the presence of NO x , rapid β-scission decomposition, or ring-opening reactions of 355 the camphene alkoxy radicals (RO-b and RO-c) occur through the RO 2 + NO pathway, leading to the generation of the peroxy radical RO 2 -d with lower carbon number and higher O:C ratio (increases from 0.30 for RO 2 -a to 0.71 for RO 2 -d). presence of ~ 100 ppb NO x . Subsequent rapid addition of O 2 can form a new peroxy radical RO 2 -d* which can undergo 1,7 H-shift isomerization and form the peroxy radical RO 2 -d ** . RO 2 -d ** can participate in termination reactions with NO and HO 2 to form organic nitrate (NO3CAMP4) and hydroperoxide (HO2CAMP4) products, which are known as highly oxygenated organic molecules (HOMs). In the presence of NO x , RO 2 -d ** can also react with NO to form the alkoxy radical RO-d that can undergo 1,4 H-shift isomerization and then O 2 addition to form the new peroxy radical 365 RO 2 -e which will also lead to the formation of HOMs such as NO3CAMP5, HO2CAMP5, and UNICAMP. A recent SOA study by Mehra et al. (2020)  conditions. Based on their observations and analysis, the average molecular formula of the camphene SOA was C 7.26 H 9.85 O 4.03 for low NO x and C 6.63 H 9.7 N 0.12 O 4.21 for the medium NO x conditions, which also suggest the occurrence 370 of ring-opening and decomposition reactions during camphene photooxidation, as predicted by MechGen. Table 5. Log 10 C* value for selected 1 st generation of stable end products formed from camphene reactions with OH.

Species
Atom O:C log 10 C* Species Atom O:C log 10 C*  Table 5 lists the log C* values and O:C ratios for the major camphene products predicted; vapor pressures of products were calculated based on the Nannoolal method (Nannoolal et al., 2008). HOMs have much lower volatilities than the earlier terminal products such as NO3CAMP1, HO2CAMP1, and RO2CAMP1. HOMs formed by 380 autoxidation steps in camphene radical chain reactions are mediated by the H-shift isomerization of RO 2 -d and RO-d. Table 6 shows the SAPRC predicted fate of RO 2 -a for all chamber runs; the fate of summed RO 2 is shown in Table   S1, which includes RO 2 -a~d and all the RO 2 radicals formed from other minor pathways. For the experiments without added NO x (WO1-6), once the initial peroxy radical RO 2 -a was formed, a large fraction of RO 2 -a (0.54-0.98) quickly reacted with HO 2 to form the terminal product HO2CAMP1, while only 2-3% of RO 2 -a reacted through the NO 385 pathway and led to the generation of HOMs. For the experiments with added NO x (W1-7), much higher RO 2 -a + NO fractions (0.65-1.00) were predicted by SAPRC. The fates of summed RO 2 also suggested that not only RO 2 -a, but also the other RO 2 radical intermediates would tend to favor further reactions through the NO reaction chain to form lower volatility products.
Based on the predicted fate of RO 2 in SAPRC simulations, the higher SOA mass yields in experiments with NO x 390 were due to the formation of HOMs through autoxidation in the presence of NO x . In general, faster RO 2 reaction with NO, HO 2 or other RO 2 limits HOM formation by autoxidation (Bianchi et al., 2019). In previous monoterpene SOA studies, HOM formation was often observed when NO x was absent or under lower NO x conditions (Pye et al., 2019;Schervish and Donahue, 2020;Zhao et al., 2018). For example, Zhao et al. (2018) demonstrated that autoxidation for some RO 2 is competitive with RO 2 + NO at ppb levels of NO for O 3 -initiated α-pinene oxidation. They also reported 395 that HOM formation decreased as the initial NO concentration increased from 0 ppb to 30 ppb. In the camphene experiments presented herein, the reverse trend was observed (see experiments WO4, W4 and W5 conducted with ~50 ppb camphene at different NO x levels). This was due to the key RO 2 species, RO 2 -d, which was predicted to form only in the presence of NO x and had a fast enough autoxidation rate constant to effectively compete with bimolecular reactions. 400 While the decreasing SOA mass yields at high [HC]  to 0.54 (WO6) and the fraction of RO 2 -a + RO 2 increased by a factor of five, from 0.08 to 0.41. Moreover, this shift from bimolecular reactions with HO 2 to RO 2 as [HC] 0 increased also occurred in the context of the total RO 2 (Table   S1). Generally, products that were predicted to form from one RO 2 reacting with another RO 2 in the absence of NO x , 410 had relatively higher volatility than those formed from that RO 2 reacting with HO 2 ; for example, RO2CAMP1 formed from RO 2 -a + RO 2 was more volatile than HO2CAMP1 formed from RO 2 -a + HO 2 ( Table 5). The increasing fraction of self-and cross-reactions of RO 2 therefore is one likely explanation for the decreasing SOA mass yields at high ΔHC and M o in the experiments without NO x .
The relatively low SOA mass yields in experiments W1 and W2 (0.36 and 0.33), also can be explained due to 415 differences in product distribution. An underestimation of the SOA mass yields in these experiments due to the assumption of negligible wall loss is not sufficient to explain these relatively low yields. A comparison of the product distributions between W1, W2, W3 and W5 suggested similar yields of NO3CAMP1-5 and NOCAMP1-2, but major differences in yields of UNICAMP and HO2CAMP1-5 (Fig. S5). Experiments W3 and W5 were selected for comparison because of their closest total RO 2 fractional reaction distribution (approximately 90% RO 2 + NO and 10% 420 RO 2 + HO 2 ) to W2 (98% RO 2 + NO and 2% RO 2 + HO 2 ) and W1 (96% RO 2 + NO and 4% RO 2 + HO 2 ) but higher SOA mass yield (0.64 and 0.6). W1 and W2 were predicted to have much smaller SOA mass yield than W3 and W5 in the low volatility products HO2CAMP1-5 (especially product HO2CAMP5, the lowest volatility among all listed products in Table 5, log 10 C* = -4.3) and UNICAMP (the second lowest volatility shown in Table 5, log 10 C * = -3.9), which can contribute to the lower SOA mass yield. Further analysis of W1 and W2 revealed a likely cause for the 425 different yields of HO2CAMP1-5 and UNICAMP. W1 and W2 were predicted to have delayed peaks of [OH] (after 3-4 hours of irradiation) which likely was due to the high NO x concentrations (Fig. S1b)

SOA Mass and Yield
The comparison of gas-and particle-phase species between chamber experiments and GECKO-A model simulations are shown in Fig. S1a and Fig. S1b. Without added NO x , GECKO-A predicts much smaller camphene consumption 435 rates and no O 3 formation, while both the chamber data and SAPRC simulations suggest a final O 3 mixing ratio of ~10 ppb (Fig. S1a). This may be due to an underrepresentation of data and relevant pathways for low to no NO x conditions in the GECKO-A mechanism generation system, and the incomplete treatment of wall effects in the application of the GECKO-A box model. The without added NO x simulations therefore are not further discussed. With added NO x , GECKO-A shows good agreement with the experimental data and SAPRC simulations in the context of 440 camphene consumption, O 3 , and OH levels.  Figure 9 shows the predicted SOA mass yields based on GECKO-A. For simulations with added NO x , while the model predicted higher SOA mass yields (0.64-0.93) than were observed (0.33-0.64), the trends in the SOA mass 445 yields were consistent between chamber observation and simulations. The simulated SOA mass yield increased with SOA mass for SOA mass < 260µg m -3 , plateaued for SOA mass between 260 and 524 µg m -3 , and then decreased for SOA mass > 524 µg m -3 .
The predicted O:C ratio and average carbon number (Fig. 10), defined as the mole-weighted averaged carbon number for the main products (~95% by mass), were consistent with the plateauing/decreasing SOA yields at higher 450 [HC] 0 (Fig. 9). The average carbon number was calculated using equation (2): where nC i , M o,i , and MW i are the carbon number, mass, and molecular weight of species i, respectively. With added NO x , the average carbon number of both the gas and particle phases increased as [HC] 0 increased, while the O:C ratio decreased. These trends indicate there is a significant fraction of higher volatility compounds formed that contribute 455 to SOA at higher [HC] 0 (or M o ) , resulting in lower SOA mass yields. In addition, only at the highest two [HC] 0 were non-negligible fractions of precursor predicted to react with O 3 and NO 3 (Fig. S7)  A negative correlation was also observed between measured particle density and [HC] 0 /[NO x ] 0 . The final density 470 of particles decreased from 1.47 g cm -3 to 1.30g cm -3 as [HC] 0 /[NO x ] 0 increased from 0.08 to 120 (Fig.11b). The change in O:C ratio could account for the change in density. O:C and H:C have been used in semi-empirical SOA density parameterizations (Nakao et al., 2013;Kuwata et al., 2012), in which O:C plays a dominant role in determining organic particle density compared to H:C. Consistent with the semi-empirical formulations, the density of particles formed from oxidation of camphene increased as O:C increased (from 0.39 to 1.21), while H:C varies over a smaller 475 range (from 1.42 to 1.79). The change in density supports the proposed explanation that more oxygenated products were formed under lower [HC] 0 /[NO x ] 0 . The wide range in final density and the correlation with [HC] 0 /[NO x ] 0 shown here has not been previously reported. The SOA mass of each experiment in this study was calculated with its own density of SOA, instead of applying an averaged density. A list of particle densities used in this study can be found in Table 2. 480

Conclusions
The first SOA mass yields from oxidation of camphene based on experiments performed in UCR environmental chamber with varying [NO x ] 0 are presented herein. Higher SOA mass yields were measured with added NO x (0.33-0.64) than without added NO x (0.08-0.26) at atmospherically relevant OH concentrations. SOA formation from the oxidation of camphene showed different NO x dependence than what has previously been reported for other 485 monoterpenes (e.g., α-pinene, d-limonene) and n-alkanes (carbon≤ 10), in which higher SOA mass yields were measured when [NO x ] was lower (Nøjgaard et al., 2006;Ng et al., 2007b). For camphene oxidation, higher Δ [HC] and lower [HC] 0 /[NO x ] 0 (within 0.5-200) generally led to higher SOA mass yields. Similar NO x dependence has been observed for two sesquiterpenes (longifolene and aromadendrene) but was attributed to the production of nonvolatile organic nitrates with no detailed mechanistic analysis provided at that time (Ng et al., 2007b). 490 Although [HC] 0 /[NO x ] 0 shows clear correlation with SOA mass yield, this quantity cannot completely represent the underlying RO 2 chemistry. The RO 2 chemistry and the competition between varying bimolecular RO 2 and unimolecular RO 2 reaction pathways, explored using SAPRC MechGen, can be used to explain the dependence of SOA mass yields on HC and NO x . The RO 2 + NO pathway favored in experiments with added NO x formed HOMs with much lower volatilities than products formed in other pathways. In addition to the regular NO x regime introduced 495 above ([HC] 0 /[NO x ] 0 > 0.5), the results suggested an extreme NO x regime where high [NO x ] may suppress SOA mass yield. High NO x levels may suppress HO 2 levels at the beginning of the experiments, causing a subsequent reduction in the yields of low volatility products such as UNICAMP and HO2CAMP5. This suggests that if the reactions happened in NO x -rich environments with extremely high ratios of NO to HO 2 (NO/HO 2 ), the SOA mass yield from oxidation of camphene might be significantly suppressed. As demonstrated here, simulations with chemically detailed 500 box models such as SAPRC are useful for identifying SOA formation regimes.
Overall, SOA formation from oxidation of camphene may be larger in polluted environments (e.g., urban environments) than NO x -free environments. This reveals a possible underestimation of SOA formed from oxidation of camphene and potentially other VOCs that are assumed to have lower SOA mass yields at higher NO x levels.
Further chamber and modeling studies of other understudied VOCs will be important for identifying other systems in 505 which moderate NO x levels can promote HOM formation.

Data Availability
The experimental and modeling data is available upon request from the corresponding authors.