Gas and fine particle (PM2.5) phase formic acid concentrations were measured with online instrumentation during separate one-month studies in the summer of 2010 in Los Angeles (LA), CA, and Atlanta, GA. In both urban environments, median gas phase concentrations were on the order of a few ppbv (LA 1.6 ppbv, Atlanta 2.3 ppbv) and median particle phase concentrations were approximately tens of ng/m3 (LA 49 ng/m3, Atlanta 39 ng/m3). LA formic acid gas and particle concentrations had consistent temporal patterns; both peaked in the early afternoon and generally followed the trends in photochemical secondary gases. Atlanta diurnal trends were more irregular, but the mean diurnal profile had similar afternoon peaks in both gas and particle concentrations, suggesting a photochemical source in both cities. LA formic acid particle/gas (p/g) ratios ranged between 0.01 and 12%, with a median of 1.3%. No clear evidence that LA formic acid preferentially partitioned to particle water was observed, except on three overcast periods of suppressed photochemical activity. Application of Henry's Law to predict partitioning during these periods greatly under-predicted particle phase formate concentrations based on bulk aerosol liquid water content (LWC) and pH estimated from thermodynamic models. In contrast to LA, formic acid partitioning in Atlanta appeared to be more consistently associated with elevated relative humidity (i.e., aerosol LWC), although p/g ratios were somewhat lower, ranging from 0.20 to 5.8%, with a median of 0.8%. Differences in formic acid gas absorbing phase preferences between these two cities are consistent with that of bulk water-soluble organic carbon reported in a companion paper.
 In highly polluted urban environments, typical concentrations of gaseous formic acid (HCOOH) can be up to several ppbv [Khare et al., 1999; Veres et al., 2011], while concentrations in the condensed phase, measured as the formate anion (HCOO−), can reach tens of ng/m3 [Chebbi and Carlier, 1996; Khare et al., 1999]. Aerosol formate is typically about two orders of magnitude lower than formic acid gas phase concentrations [Baboukas et al., 2000]. Most previous work on formic acid gas-particle partitioning was based on gas phase measurements using scrubbers (e.g., mist chambers) and particle phase measurements by integrated filter collection. These studies have found a higher fraction of formic acid in the condensed phase in winter than summer, apparently due to reduced vapor pressures at lower temperatures [Chebbi and Carlier, 1996; Khare et al., 1999]. Filter sampling, however, constrains partitioning studies by only providing highly time-integrated data (e.g., 12 to 24 h averages) and limits investigating the response of formic acid to diurnally changing meteorological parameters. In this study formic acid ambient concentrations and particle/gas (p/g) ratios are investigated using online measurement techniques in two contrasting cities, Los Angeles (LA), CA and Atlanta, GA. Los Angeles is an urban environment dominated by anthropogenic emissions, whereas Atlanta air quality is influenced by a mix of anthropogenic and biogenic emissions.
2. Experimental Methods
2.1. Sampling Sites
 Ground-based measurements were made in Pasadena, CA on the California Institute of Technology campus as part of the 2010 CalNex (Research at the Nexus of Air Quality and Climate Change) experiment from 15 May to 15 June 2010. Located within the Los Angeles Basin (here-after referred to as LA) approximately 16 km downwind of central LA, the site (34.140582 N, 118.122455 W) was regularly impacted by emissions transported from the source-rich downtown LA area [Hersey et al., 2011; Veres et al., 2011]. Following CalNex, a subset of the instruments was moved to Atlanta for continuous sampling between 3 August and 10 September 2010. Identical measurements between the two studies included gas and particulate formic acid and water-soluble organic carbon (WSOC), PM2.5 inorganic anions, and PM2.5 organic and elemental carbon (OC and EC, respectively). The Atlanta measurements were conducted from the roof of the Environmental Science and Technology Building (33.778427 N, 84.396181 W) on the Georgia Institute of Technology (GIT) campus (∼30–40 m above ground level). A summary of instruments/sampling setups for both the LA and Atlanta studies is given in Table 1.
Table 1. Summary of Instrumentation and Approximate Sampling Setups for the Various Species Reported
Inlet Height Above Ground Level (m)
Inlet Length (m)
Sampling was from a penthouse rooftop laboratory located roughly 30 m above ground level with the inlet 1 to 4 m above the roof surface.
 A suite of organic and inorganic acids was measured by negative ion-proton transfer-chemical ionization mass spectrometry (NI-PT-CIMS) through proton transfer with acetate ions and detection of the conjugate anion with a quadrupole mass spectrometer [Veres et al., 2008]. Two NI-PT-CIMS were deployed in LA while only the Georgia Institute of Technology (GIT) CIMS was deployed in Atlanta. In LA, Formic acid was measured by both GIT and the National Oceanic and Atmospheric Association (NOAA) CIMS, which were in reasonable agreement with a slope of 0.95 and r2 of 0.80 (auxiliary material Figure S1). NOAA CIMS measured nitric acid only in LA; no nitric acid data is available for the Atlanta study. Details of the sampling configurations are given in Table 1. The limit of detection (LOD) and measurement uncertainty is estimated at 0.02 ppbv and 25% for formic acid, respectively, and 0.08 ppbv and 37% for nitric acid by NOAA CIMS during CalNex-LA campaign, while the LOD of GIT CIMS was 0.02 ppbv in LA and 0.12 ppbv in ATL with an uncertainty of 30%.
 Gas-phase water-soluble organic carbon (WSOCg) was measured with a mist chamber [Hennigan et al., 2008]. Air was sampled at roughly 21 ± 1 L/min and was first passed through a Teflon filter for particle removal and then scrubbed of soluble gases (gases with a Henry's Law constant greater than 103 M/atm) [Spaulding et al., 2002] in a glass mist chamber [Cofer and Edahl, 1986] initially filled with 10 mL of ultra-pure deionized water (>18.0 MΩ-cm). WSOCg was then quantified with a Sievers Total Organic Carbon (TOC) analyzer (Model 800, GE Analytical Instruments; Boulder, CO) that was shared between the gas and particle WSOC instruments. Data were corrected by blank measurements performed prior to every mist chamber sampling cycle by loading the chamber with pure water, without sample airflow. The LOD and measurement uncertainty is estimated at 0.83 μgC/m3 and 8%.
 Along with the measurements of WSOC and formic acid, the mixing ratios for a variety of Volatile Organic Compounds (VOCs) were determined with a NOAA Gas Chromatography Mass Spectrometry (GC-MS) [Gilman et al., 2009] during the CalNex-LA campaign. In both studies, meteorological data were provided by weather stations located on the respective campuses.
 PM2.5 soluble nitrate (NO3−) and formate (HCOO−) concentrations, along with a range of other soluble anions, were determined using a Particle-Into-Liquid-Sampler coupled to an Ion Chromatograph (PILS-IC), similar to that described byOrsini et al. . Sample air was aspirated at 16.7 ± 0.7 L/min through a URG PM2.5 cyclone, a parallel plate carbon denuder [Eatough et al., 1993] and honeycomb acid (citric acid) and base (sodium carbonate) coated denuders, then mixed with water vapor near 100°C generated from deionized water. The resulting droplets were collected by impaction and produced a liquid sample that was analyzed online by a Dionex Ion Chromatograph (UTAC-ULP1 concentrator column, AG11guard column and AS11 anion column) using a gradient NaOH eluent procedure lasting 20 min. Systematic blank measurements were made throughout the study period by diverting sample flow through a HEPA filter downstream of the denuder and upstream of the PILS. A linear interpolation of blank data was performed and subtracted from the ambient data. During the last several days of the study (12 June–15 June, 2010), the inlet cyclone was alternated between cyclones of PM1 and PM2.5cut sizes (URG) during daytime periods to investigate size-dependent concentration differences. (This was done for anions, OC, EC, and particle WSOC). Sample heating due to the PILS particle condensation growth collection method may cause potential collection losses of semi-volatile components [Sorooshian et al., 2006]. Of the measured compounds, the most susceptible is thought to be formate (HCOO−) because it was the most volatile of the anions measured in this study. Sample heating may also promote some particulate organic constituents (e.g., glyoxylic acid) and oligomers to decompose and form small organic compounds prior to IC analysis, which might cause an over-estimation in formate [Ervens et al., 2008]. Post study comparisons involving simultaneous ambient air sampling with a manually operated Mist Chamber, which does not heat the sample (in this case the Mist Chamber inlet filter was replaced with gas denuders) and PILS indicated that PILS formate concentrations can be systematically under-estimated by up to 30%. Data were not corrected for the potential systematic error. Random measurement uncertainty based on a quadrature sum of squares that included the precision of standards, variability in sample airflow rate, liquid flow rate, and blanks (all one standard deviation), was estimated at 13% for nitrate and 15% for formate. LODs are 0.03μg/m3 for nitrate and 0.02 μg/m3 for formate, respectively.
 A second PILS coupled to a TOC analyzer provided online measurements of PM2.5water-soluble organic carbon (WSOCp) following methods implemented in previous studies [Sullivan et al., 2004]. PILS-TOC included a URG PM2.5 cyclone inlet and parallel plate carbon denuder upstream. An automated valve that directed sample flow through a Teflon filter located immediately upstream of the PILS and downstream of the denuders performed dynamic blanks twice daily. A linear interpolation between successive blanks was used to blank correct the online measurements. Instrument LOD and measurement uncertainty are estimated at 0.20 μgC/m3 and 7%, respectively. More details on this measurement and discussion of WSOC partitioning results can be found in Zhang et al. [2012a].
 A high-resolution time-of-flight aerosol mass spectrometer (HR-ToF-AMS, Aerodyne Research Inc., hereinafter “AMS” for short) was deployed in LA and provided measurements of non-refractory inorganic and organic aerosol components (P. L. Hayes et al., Aerosol composition in Los Angeles during the 2010 CalNex Campaign studied by high resolution aerosol mass spectrometry, submitted toJournal of Geophysical Research, 2012). Ambient air was dried prior to sampling by the HR-ToF-AMS. The AMS measures submicron non-refractory species [Canagaratna et al., 2007]. Differences in AMS and PILS upper size measurement limits and measurement of non-refractory versus soluble aerosol components likely account for the significant difference in nitrate concentrations measured by these two instruments while sampling during CalNex, a site influenced by marine air masses. The linear regression (ODR fit) of AMS to PILS nitrate slope was 0.66 (r2 = 0.86).. In the following analysis we interpret AMS measurements to be dominated by volatile nitrate (i.e., ammonium nitrate), which is most important in the submicron mode, and the PILS both volatile and nonvolatile nitrate (e.g., ammonium nitrate plus sodium nitrate), the latter being more important in the 1–2.5 μm PM range (Hayes et al., submitted manuscript, 2012). AMS measurement uncertainty is estimated at 30% [Middlebrook et al., 2012]. AMS/PILS sulfate were in good agreement, with regression slope of 1.01, and r2 of 0.84.
 PM2.5organic carbon and elemental carbon (OC and EC) were measured semi-continuously with a Sunset Labs OC/EC analyzer (Model 3F, Forest Grove, OR) following theNational Institute for Occupational Safety and Health (NIOSH)  method. Optical EC is used in this analysis since it was less noisy than thermal EC. A parallel plate carbon denuder [Eatough et al., 1993] upstream of the instrument was installed to reduce positive sampling artifacts. Measurements were blank corrected based on systematic blank measurements made throughout the study period by installing a Teflon filter (47 mm dia., 2.0 μm pore size, Pall Life Sciences) on the cyclone inlet. A linear change in blank concentrations was assumed between consecutive blank measurements and was subtracted from the ambient data. OC blanks ranged from 1.3 to 2.6 μgC/m3 and EC blanks varied from below LOD to 0.03 μgC/m3. Uncertainty due to random measurement errors is estimated to be 28%. Systematic errors due to particle volatility are possible [Grover et al., 2009]. Significant discrepancies were observed between the AMS OA (organic aerosol mass, i.e., OM) and Sunset Labs OC, with a study average OM/OC slope of 3.28 and r2 = 0.76, where a slope closer to 2 is more realistic [Turpin and Lim, 2001; Hayes et al., submitted manuscript, 2012]. Possible causes include loss of semi-volatile compounds from the collection filter media within the OC/EC analyzer, or due to the presence of the denuder [Grover et al., 2009]. Other unidentified systematic errors associated with the measurements cannot be ruled out.
3.1. Los Angeles Diurnal Variation in Air Pollutant Concentrations
 As a coastal city with clean upwind sector (Pacific Ocean) and a basin-shaped topography, LA has been observed to have a regular and distinct diurnal pattern in air pollutants. Emissions from the western regions of the LA Basin were transported across the basin in a northeastward direction by land-sea breezes on a daily basis [Veres et al., 2011]. The CalNex-LA sampling site at Pasadena, CA, located ∼16 km NE of downtown LA, is a receptor of gaseous and particulate pollutants transported from downtown LA. The arrival of transported emissions is readily observed in the diurnal profiles of primary emissions (Figure 1). Acetylene and PM2.5EC reached daily maximum concentrations in Pasadena between 12:00 and 13:00 PDT after being transported from downtown LA over the course of a few hours. Only minor enhancements were observed between 6:00 and 9:00 PDT from local morning rush-hour traffic.
 Secondary gas and aerosol components followed behind the arrival of primary species, coupled to both transport and photochemical processes. Diurnal maximum concentrations of gaseous photochemical products, such as acetaldehyde, formic acid, and WSOCg, occurred at 13:30 PDT, which is near the peak for solar insolation and after primary emissions reached their peak (Figure 1). Secondary organic aerosol components, including oxalate and WSOCp, then followed the secondary gases, with daily maximum concentrations approximately 1 h later than the gases, indicating a gas-to-particle conversion process (Figure 1).
 The complete study time series shows that gas and particle phase formic acid had consistent and clear diurnal trends (Figure 2). Both gas and particle phase formic acid followed other photochemical gas/particle products with peak concentrations in the early afternoon (Figure 1). However, the diurnal trends in both gas and particle phase formic acid were generally broader than other secondary gas and particle species (e.g., WSOC), and particle formic acid had a daily concentration peak that was closer in time to the peak of the gas phase components (13:30 PDT), compared to other SOA components (WSOCp, oxalate) that peaked at ∼14:30 PDT (Figure 1).
 For the entire CalNex-LA study, gaseous formic acid was in the range of 0.37 ppbv to 5.86 ppbv and had an average concentration of 1.89 ± 0.59 ppbv (mean ±1σ, n = 731) (see Table 2 for a summary). The data are consistent with previous studies in urban environments [Kawamura and Kaplan, 1985; Grosjean, 1988]. For the aerosol phase, the range of PM2.5 formate concentrations observed during the entire campaign was from the LOD (0.02 μg/m3) to 0.21 μg/m3 and the mean formate concentration was 0.054 μg/m3. Most formate was associated with particles below 1 μm since little difference was observed between PM1 and PM2.5 concentrations (PM1/PM2.5 mean ∼104% ± 17%, n = 18), when PM1 and PM2.5 cyclones were interchanged during daylight periods of the final week of sampling (Figure 2).
Table 2. Summary of Gas and Particulate Phase Formic Acid Concentrations and Partitioning Ratio (p/g, Given in %), for Los Angles (LA) and Atlanta (ATL)
LOD = 0.02 μg/m3.
Sampling times are in PDT (Pacific Daylight Time).
Sampling times are in EDT (Eastern Daylight Time).
 Formic acid was predominately found in the gas phase with particle-to-gas (p/g) ratios typically between 1 to 2% throughout the study (Table 2). Ratios from previous studies generally range from 0.8 to 10% [Chebbi and Carlier, 1996; Baboukas et al., 2000; Löflund et al., 2001]. The high volatility and resulting minor partitioning to the particle phase may partially explain why the average diurnal trend of particle formic acid tended to track the gas phase species, instead of other SOA components that probably undergo further partitioning processes or are formed within the particle (or aqueous) phase, such as oxalate and WSOCp (WSOC p/g ratios are closer to 10 to 20% [Zhang et al., 2012a]).
 The LA study mean ± standard deviation formic acid p/g ratio was 1.6% ± 1.3%, but at times the ratio reached up to 10%. Highest partitioning to the particle phase observed in LA occurred during a few unique periods when partitioning to liquid water was most apparently observed. Factors that may contribute to this variability are investigated.
3.2. Los Angeles Formic Acid Partitioning
 Gas-phase formic acid could partition to the particle phase through absorption to the preexisting organic phase, or undergo dissolution into aerosol water, or some combination of both. A schematic of possible partitioning routes for formic acid is shown inFigure 3. Due to its high solubility (Henry's law constant at 20°C is ∼4 × 103 M/atm [Chameides, 1984; Lelieveld and Crutzen, 1991]), partitioning to the aqueous fraction of aerosols may be expected if particle water is present. Chang et al.  performed a model simulation of overall SOA formation in the Los Angeles basin and found that SOA partitioning to both organic and aqueous phases are important and that the aqueous phase becomes more important during periods of low SOA concentrations. Similar behavior was observed for formic acid in this study.
 The regular diurnal temperature and relative humidity (RH) trends during the LA study are shown in Figure 4. During CalNex, RH often reached over 90% by midnight and remained at or near saturation levels until sunrise. RH rarely fell below 60% at its lowest point in mid-afternoon (Figure 4). Despite the expected higher nighttime particle water concentrations and formic acid's relatively high solubility, observed formic acid p/g ratios generally did not have a strong dependence on time of day (Figure 5a), or RH (Figure 6a), although p/g ratios were slightly more variable at night when RH was highest (Figure 5a). Part of this variability in the average diurnal profile occurred during unique isolated periods of overcast conditions when formic acid p/g ratios did follow RH trends. The median p/g value ranged from 1.33 to 1.77% at night (20:00–08:00 PDT) and 1.15–1.43% during day (08:00–20:00 PDT). However, during the three overcast periods shown in Figure 4 (14:30 May 17 to 9:30 May 18, 22:30 May 27 to 15:30 May 28, and 12:30 June 10 to 12:30 June 11, 2010, PDT), the p/g ratio was substantially higher than the study mean values (Figures 5b and 6b). Apart from these unique episodes, the formic acid p/g ratio seemed largely independent of RH (e.g., particle water) suggesting partitioning to some other phase, possibly the organic aerosol phase.
 In the following analysis, formic acid partitioning is investigated for the two types of periods discussed: (1) when formic acid appeared to be related to RH (partitioning to liquid water) and (2) when the dependence on liquid water was not as evident.
3.2.1. Los Angeles Formic Acid Partitioning to Particle Water
 The overcast periods, when formic acid tracked RH, were associated with much lower daytime concentrations of photo-chemically generated species. For example, during the same time of day (11:30–17:30 PDT), HNO3concentrations were one-fifth, WSOCgapproximately one-half and gaseous formic acid less than one-third of the respective study average concentrations. Daytime temperatures were slightly lower than non-overcast conditions (Figure 4), which could further favor partitioning to the particle phase. HYSPLIT (R. R. Draxler and G. D. Rolph, HYSPLIT (Hybrid Single - Particle Lagrangian Integrated Tracker) Model, 2012,http://ready.arl.noaa.gov/HYSPLIT.php; G. D. Rolph, Real-time Environmental Applications and Display sYstem (READY) Website, 2012,http://ready.arl.noaa.gov) back-trajectory calculations do not indicate a common air mass source region during these three overcast periods (auxiliary material Figure S2).
 Uptake of gas phase formic acid by aerosol water can be investigated most simply by assuming the gas and particle are in equilibrium and comparing measured p/g ratios to those predicted by Henry's law, assuming ideal solution behavior. Under these assumptions the expected formic acid p/g ratio is:
where LWC is the particle liquid water concentration, KH the Henry's law constant for formic acid, Ka the formic acid dissociation constant, and [H+] the hydrogen ion concentration (mol/L). Both KH and Ka depend on temperature, and were calculated using ambient temperatures by:
 In the following analysis we apply pH predicted by R. A. Ellis et al. (Gas-particle partitioning of ammonia at the CalNex-LA ground site and the influence of aerosol pH, manuscript in preparation, 2012) determined from the Extended-Aerosol Inorganic Model (www.aim.env.uea.ac.uk/aim/aim.php) [Wexler and Seinfeld, 1991; Wexler and Clegg, 2002] into equation (1). Although predicting aerosol pH can be highly uncertain [Yao et al., 2006], Ellis et al. (manuscript in preparation, 2012) show that pH is better constrained through measurements of both aerosol ions and gas phase ammonia, as was done during CalNex. Aerosol water (LWC) was predicted by thermodynamic models (ISORROPIA-II or E-AIM, results were similar). Measurements and thermodynamic models show that aerosol water concentrations are largely determined by inorganic salt concentrations, with studies showing a deviation smaller than ∼15% among inorganic models [Fountoukis et al., 2009, and references therein]. Other studies suggest not including organic species can produce errors in predicted LWC of up to ∼20% [Speer et al., 2003; Aggarwal et al., 2007]. (Note, including measured concentrations of organic acids in the E-AIM calculations had little effect on predicted LWC; <3% in this study).
18.104.22.168. Measured Formic Acid Partitioning Compared to Henry's Law
 The three periods showing the clearest formic acid p/g dependence on RH were all associated with a clear diurnal trend in RH, weak photochemistry due to low solar radiation, and significant variation in formic p/g ratios. RH was mostly above 70% during the three periods, prompting the assumption that particles were often deliquesced and thus Henry's law is applicable.
 The formic acid p/g ratio predicted by equation (1) has a RH dependence similar to the observed ratio (Figure 6b), but is roughly 7 orders of magnitude lower, suggesting essentially no partitioning to the particle phase. For these periods the predicted LWC values ranged from 5 to 40 μg/m3 and particle pH 3.24–4.57 (mean ± std pH = 3.88 ± 0.34). In contrast, based on the predicted LWC and observed formic acid p/g ratio, a pH of 9.31 ± 0.60 is predicted for formic acid to conform to Henry's law, which is not reasonable since most previous studies showed acidified aerosol particles in urban environments impacted by marine aerosols [Keene et al., 2004; Yao et al., 2006].
 Under-prediction of PM2.5 formate concentrations by Henry's law may be due to several reasons. Formic acid p/g ratios are highly sensitive to pH, and uncertainty in pH of the particle fractions that the gas phase is partitioning to, is potentially a large source of error. Past studies have noted that pH can also vary with aerosol size [Keene et al., 2004]. There are uncertainties related to the LWC prediction due to the role of organics, but as noted above these are likely small. Other uncertainties associated with using bulk measurements and unknown particle mixing state are possible, including variability in particle water between different particles and other components that may interact with formic acid but are not included in the models (e.g., amines, carbonates). The assumption that formic acid partitions to an ideal solution is also likely not true at all times during the overcast periods, given that many inorganic ions are present. However, considering non-ideal solution behavior does not significantly change the large discrepancy; E-AIM estimated formic acid activity coefficients could increase the predicted formic acid p/g ratio by at most roughly a factor of 10 under the lowest RH of the overcast periods. Finally, production of formate in the aerosol phase through heterogeneous processes followed by a process that inhibits evaporation, such as the formation of oligomers, could also account for the higher than expected formate p/g ratios.
 Similar formic acid discrepancies from an aqueous equilibrium state have also been reported [Keene et al., 2004; Healy et al., 2008] based on filter sampling methods. Keene et al. found that formic acid was expected to partition almost exclusively to the gas phase. Significant particle concentrations were, however, detected. They argued that this discrepancy might have been due to positive artifacts related to filter sampling. Our online measurements, which may be less susceptible to artifacts, are consistent with these filter-based studies. A similar enhanced partitioning has been observed for other soluble atmospheric components, including hydrogen peroxide [Arellanes et al., 2006], carbonyls [Matsunaga et al., 2005], dicarbonyls [Healy et al., 2008; Volkamer et al., 2009], and bulk water-soluble organic carbon (WSOC) [Hennigan et al., 2008].
3.2.2. Los Angeles Formic Acid Partitioning During Periods of No Observed RH Dependence
 Highest formic acid concentrations for both gas and particulate phases occurred during daytime hours, peaking in the early afternoon when other photo-chemically derived compounds (including nitric acid, several organic acids and WSOCg) were also at their highest concentrations [Veres et al., 2011], indicating that photochemistry contributed to formation of both gas and particle phases. Further evidence for photochemical sources for formic acid is indicated by a significant weekend enhancement in both, relative to weekdays. Gas phase formic acid mixing ratios were 1.2 times higher on the 5 weekends (10 days) during the study relative to weekdays, and similarly PM2.5 formate levels were 1.3 times higher. The difference is thought to be due to higher weekend oxidant concentrations (e.g., O3 was 1.4 times higher on weekends) resulting from a significant reduction in NOx emissions on weekends relative to weekdays [Bahreini et al., 2012; Pollack et al., 2012].
 Although concentrations of gas and particle formic acid were highest during the early afternoon peak in photochemical production (Figure 1), p/g ratios were lower (1.38% mean) and less variable (0.87% stdev, see Table 2) at these times compared to nighttime or overcast periods. During these afternoon photochemical periods, RH was typically at the daytime minimum of 60–70% (mean ± stdev = 62 ± 3%), and particle liquid water concentrations are generally expected to be the lowest (8.3 ± 2.8 μg/m3), in contrast to the high LWC values of greater than 40 μg/m3 when RH exceeded 95% (e.g., see Figure 7).
 If formic acid has multiple routes for partitioning (i.e., water or OA), it is likely that during overcast periods when OA is low (average OA concentration = 3.68 μg/m3, 54.0% of study average) the particle water route may be relatively more important; in contrast, during the photo-chemically active afternoon periods the opposite conditions occur: OA concentrations are at a maximum and RH (LWC) at a minimum, conditions most conducive for partitioning to OA. A rough estimation based on measured OA concentration and formic acid p/g ratio gives a partitioning coefficient (Kp) ranging from 0.00003 to 0.024 (average ± stdev = 0.0036 ± 0.0038) m3/μg, which are roughly three orders of magnitude higher than the theoretical value (2.04 × 10−6 m3/μg). However, the specific component of the aerosol the formic acid partitioned to is not completely clear; no correlation was found between formic acid p/g ratio and any other aerosol components, including OA, or AMS PMF factors such as oxygenated organic aerosol (OOA), semi-volatile oxygenated organic aerosol (SV-OOA), low volatility oxygenated organic aerosol (LV-OOA) factors), or hydrocarbon-like organic aerosol (HOA) (auxiliary material Figure S3).
3.2.3. LA Nitric Acid Partitioning and RH as a Contrast to Formic Acid
 Like formate, aerosol nitrate is also semi-volatile (excluding nonvolatile forms, such as NaNO3), but nitric acid forms salts and has a much higher effective KHupon dissociation. As noted above, both gas phase formic and nitric acids had similar diurnal profiles, peaking at mid-day along with other photo-chemically generated gaseous species (Figure 1 versus Figure 8). However, while particle phase formate tended to peak in the afternoon, both AMS and PILS nitrate were usually highest in early morning (Figure 8). Note that AMS PM1non-refractory nitrate (e.g., dominated by semi-volatile ammonium nitrate (Hayes et al., submitted manuscript, 2012)) concentrations dropped off relative to PM2.5 soluble nitrate (e.g., NH4NO3 + NaNO3) as the morning progressed, consistent with the more volatile NH4NO3 responding to increasing daytime temperatures.
 The contrasts in the particle phase formate and nitrate diurnal profiles are consistent with differences in formic and nitric p/g ratios relative to RH (particle water). Whereas the formic acid p/g ratio generally did not track RH, nitric acid p/g ratios did follow RH (Figures 5b and 7). This contrast between the relationship of formic acid and nitric acid p/g ratio with RH further supports the idea that the partitioning of formic acid is not dominated by partitioning to aerosol water; other absorption media possibly exist, such as the organic aerosol fraction. Although there was no evidence showing the partitioning of formic acid was correlated to OA, WSOC was observed to correlate with both RH (e.g., particle LWC) and OA [Zhang et al., 2012a].
3.3. Atlanta Formic Acid Concentrations and Partitioning
 Observed gas and particle phase formic acid concentrations in Atlanta are shown in Table 2. Gas phase formic acid mixing ratios were higher in Atlanta (LA median of 1.61 ppbv, Atlanta median of 2.29 ppbv) and particle phase formate levels were slightly lower in Atlanta (LA median of 49 ng/m3 versus Atlanta median of 39 ng/m3). The time series of both gas and particle phase formic acid were not nearly as regular in Atlanta as LA (Figure 9), which is expected due to the significantly different meteorological and topographical differences between these two sites: LA is bordered by the Pacific Ocean to the West and San Gabriel Mountains to the Northeast, whereas Atlanta is a continental setting lacking nearby major geological features. In general average diurnal profiles show peak gas and particle formic acid concentrations later in the afternoon (15:30 Eastern Daylight Time (EDT)), suggesting a link to photochemical sources, but roughly three hours later than the peak in solar insolation (12:30 EDT) (Figure 10). Curiously, this is the same time that oxalate concentrations reached a maximum (although a small peak relative to a large background) and both may be at least partially due to transport from aloft [Zhang et al., 2012a]. These organic acid peaks are not well synchronized with the overall WSOCg or WSOCp, except that the gaseous phase formic acid and WSOCg began to increase together in the early morning (07:30 EDT), while formate and WSOCp increased together later in the morning (09:30 EDT). Like WSOCg, gas phase formic acid maintained fairly high concentrations until evening (18:00 EDT) and then dropped off more rapidly in the morning, whereas formate concentrations dropped off in the afternoon and maintained fairly consistent levels until the morning increase. The result is that the formic acid p/g ratio tended to be higher in the early morning hours (02:00–08:00), a period of high RH. The same was observed for WSOC, but the extent of WSOC partitioning to the particle phase in the early morning was even stronger [Zhang et al., 2012a].
 The study average formic acid p/g ratio in Atlanta was on the order of 1% (Table 2), slightly lower than LA, but with a more pronounced diurnal variation and more of the extremes biased to higher p/g ratios (Figure 11). Gas phase formic acid concentrations were on average higher in Atlanta (Atlanta mean ± stdev = 2.79 ± 1.01 ppbv, LA = 1.89 ± 0.59 ppbv) while particle phase concentrations were slightly lower (Table 2). In contrast to LA, Atlanta formic acid partitioning generally had a more consistent relationship to RH (although not as clear following sunset), and hence apparently to particle water. This can be seen in both the average diurnal profile (Figure 11) and the binned p/g ratio versus ambient RH (Figure 12), although less data at high RH in Atlanta makes this comparison less robust.
3.4. Formic Acid Partitioning Absorbing Phases: LA Versus Atlanta
 A difference in RH-partitioning dependence between LA and Atlanta was also observed for the bulk WSOC measured during these studies [Zhang et al., 2012a]. In LA, WSOC partitioning appeared to be linked to an absorbing phase correlated with OC as opposed to RH, whereas in Atlanta, WSOC partitioning was linked to RH, but not OC, when RH was greater than ∼70% (i.e., when LWC is sensitive to RH). When the particles were drier (i.e., RH less than ∼70%), there was evidence that WSOC partitioning correlated with OC in Atlanta. Thus, in a general sense, WSOC and formic acid may partition to similar absorbing phases in both locations: more likely the organic components in LA (or some fraction within), but in Atlanta partitioning to particle water seems more prevalent.
 Causes for the differences in formic acid absorbing phases between these two cities may be due to a number of factors.
 1. Differences in aerosol liquid water content. The extent of absorption to particle water or organic mass is sensitive to the mass of the absorbing phase available. The LA experiment was in May/June, whereas the Atlanta experiment was in August/September with generally higher temperatures and lower RH's compared with LA (Figure 13). Zhang et al. [2012a]used the inorganic species concentrations and found that ISORROPIA-II predicted similar LWC levels at two sites for RH < 70%, but higher LWC in LA at RH > 70%. By this calculation, there was typically more liquid water available in LA, while observation suggests a greater preference of Atlanta WSOC and formic acid to partitioning to particle water, relative to LA.
 2. Differences in OA concentrations.The theory of equilibrium absorptive partitioning to the organic phase predicts that the particle-to-gas ratio will depend on the concentration of the absorbing phase, in this case possibly OA [Pankow, 1994], or some component of the OA (e.g., polar organic compounds). In LA, during overcast periods when formic acid partitioned to water, OA was significantly lower than other periods of the study (3.68 μg/m3 versus study average of 6.81 μg/m3) making the aqueous phase a more favorable route for partitioning. However, the OA concentration differences between LA and Atlanta are not significant; during these two studies LA AMS OA was on average 6.81 μg/m3, while ATL OA was at ∼5.09 μg/m3 (estimated by OC × 1.6 for urban aerosols [Turpin and Lim, 2001]). Differences in polar organic concentrations, which may be a more likely absorbing phase for formic acid, or the relative polarity of organic aerosols between these two cities, may be a more likely explanation for partitioning preferences of formic acid.
 3. Differences in liquid water pH. As a weak acid, formic acid partitioning is greatly influenced by the solution pH (equation (1)). One characteristic of LA ambient air is the high levels of nitric acid that could serve as a strong acid reservoir readily available to partition to water as particle LWC responds to changing RH. During the overcast periods of low photochemistry in LA, nitric acid levels were one-fifth the study average and so a droplet pH closer to neutral is expected and would favor formic acid absorption to liquid water at that time. However, the calculated pH for the overcast periods was not significantly closer to neutral compared to other times, but predicted pHs are uncertain so the possible role of nitric acid is not clear. High concentrations of nitric acid in LA could generally suppress formic acid dissolution and account for little variability in diurnal trends in formic acid p/g ratio. Significantly lower levels of nitric acid in Atlanta may account for a greater partitioning preference to water. Based on measured values, average nitric acid was ∼2.5 ppbv during CalNex and ∼1.5 ppbv in Atlanta during the 1999 supersite. Whether this is a viable explanation for differences in formic acid partition is not clear, but the effect of pH would only influence a fraction of the WSOC. Given that roughly half the WSOCp is acidic [Sullivan and Weber, 2006], it is doubtful that pH alone explains differences in partitioning preferences of WSOC in these two cities.
 4. Differences in organic aerosol composition.Aerosol composition can influence partitioning by affecting the physical properties of the absorbing organic phase, and to a lesser extent the particle LWC. Aerosol composition is linked to differences in emissions, dispersion and transport, and length of time for atmospheric aging processes. Differences in all these categories exist between LA and Atlanta. In terms of organic components, LA emissions are largely anthropogenic, whereas Atlanta is a mix of biogenic and anthropogenic. LA has a clean upwind sector (Pacific Ocean), a regular land-sea breeze and limited dispersion due to local topology. Atlanta is situated in a mid-continental region with significant biogenic emissions and prevailing winds. Thus, LA has a substantial buildup of fresh anthropogenic SOA (with ΔWSOC = 1.23μgC/m3, where ΔWSOC is a proxy for fresh SOA and Δ implies day minus night), whereas in Atlanta greater dispersion and a large regional background results in only a minor increase in fresh SOA (ΔWSOC = 0.56 μgC/m3) that rides on the large regional signal [Zhang et al., 2012b] of aged organics mainly from biogenic emissions [Weber et al., 2007]. This is also evident in the WSOC/OC ratios, where in LA (with OC based on AMS data) the WSOC/OC ratio was 37%, in contrast to the Atlanta ratio of 62%. Thus it is expected that the Atlanta accumulation mode aerosol would be more oxygenated [Simon et al., 2011]. Differences in concentrations of inorganic components may also have some influence. LA is dominated by NH4+, NO3− and SO42−, whereas Atlanta is dominated by NH4+ and SO42−. Complex interactions between inorganic/organic (polar/non-polar) particle components as a function of O/C ratios have been shown to impact particle interactions with water, resulting in phase separations and transitions between various phases as a function of RH [Bertram et al., 2011]. These processes could play some role in explaining the differences in formic acid partitioning between LA and Atlanta.
 Gas and particle concentrations of formic acid, and factors affecting formic acid particle/gas ratios, are presented based on summer measurements made during separate one-month studies in Los Angeles, CA, and Atlanta, GA, USA. In both studies gas phase formic acid was measured via NI-PT-CIMS and particle formate with a PILS-IC setup to provide continuous data at a 20-min duty cycle or faster.
 Gas phase formic acid concentrations were moderately higher in Atlanta (LA median 1.61 ppbv; Atlanta median 2.29 ppbv), whereas particle phase concentrations were higher in LA (LA median 49 ng/m3, Atlanta median 39 ng/m3). Similar diurnal patterns of higher daytime and lower nighttime gas and particle concentrations were observed in both cities, but in LA the gaseous and particle phase formic acid daily trends closely followed most other secondary gas species, whereas in Atlanta they were not as synchronized with other secondary compounds. In LA, photochemistry is clearly a major contributor to formic acid formation; in Atlanta transport (e.g., variation in air mass origin) may play a larger role in the diurnal trends, along with photochemistry.
 Formic acid p/g ratios were on the order of a few percent (1–2%) in both cities. In LA, for the most part, no specific component of the aerosol was correlated with p/g ratios, so the exact nature of the absorbing phase remains unclear; although based on overall WSOC partitioning behavior [Zhang et al., 2012a], partitioning to OA, or some component of the OA, may be possible route. Formic acid p/g ratios in LA depended little on RH, except on three brief occasions that were all characterized by overcast conditions with little daytime photochemical production. Henry's Law predicts very little partitioning to water, based on particle LWC and pH inferred from thermodynamic models. Including non-ideal solution behavior still significantly under-predicts the particle phase concentrations. In contrast to formic acid, nitric acid partitioning is linked to ambient RH throughout the CalNex-LA study.
 In contrast to practically no RH-dependence in LA, formic acid had a clearer trend of partitioning to liquid water in Atlanta based on correlations with ambient RH, despite lower RHs and lower calculated liquid water in Atlanta. This contrast is similar to differences in bulk WSOC partitioning also measured in these two cities during these studies [Zhang et al., 2012a]; WSOC preferentially partitioned to OA in LA, whereas in Atlanta when particles were expected to have deliquesced (RH > 70%), WSOC partitioning was related to RH. Causes for the observed differences in partitioning preferences for formic acid and WSOC between LA and Atlanta are not known, but one of the biggest differences between these two cities is likely the composition of the secondary organic aerosol resulting from differences in VOC emissions.
 Georgia Tech researchers were funded through National Science Foundation grants ATM-0931492 and ATM-0802237. The authors are grateful for logistical support from John Seinfeld, Rick Flagan, and the California Institute of Technology. P.L.H. and J.L.J. were supported by CARB-319 and DOE (BER, ASR program) DE-SC0006035, and P.L.H. acknowledges a CIRES Visiting Postdoctoral Fellowship. The NOAA work was supported by NOAA's Climate Change and Health of the Atmosphere programs. We are also appreciative of the California Air Resources Board for supplying sampling structures and electrical services during the CalNex field mission.