Measurement of Formic and Acetic Acid in Air by Chemical Ionization Mass Spectroscopy: Airborne Method Development

The goals of this study were to determine whether formic and acetic acid could be quantified from the Deep Convective Clouds and Chemistry Experiment (DC3) through post-mission calibration and analysis and to optimize a reagent gas mix with CH3I, CO2, and O2 that allows quantitative cluster ion formation with hydroperoxides and organic acids suitable for use in future field measurements. There is a scarcity of organic acid measurements in the upper troposphere under stormy conditions with which to compare and assess photochemical and transport theory. DC3 observations made in May and June 2012 extending from the surface to 13 km over the central United States during convective storm conditions would be a significant addition to our knowledge base. Organic acid measurements were made with a chemical ionization mass spectrometer (CIMS) aboard the NCAR Gulfstream-V platform in DC3. The primary objective of the CIMS was to observe hydrogen peroxide and methyl hydroperoxide though it recorded signals attributed to iodide clusters of formic and acetic acid at 173 and 187 townsend, respectively. Both organic acids were targets of opportunity as the reagent ion (I ) needed for clustering was not specifically used in the field. However an iodide source gas (iodomethane, CH3I), was used during premission experiments and a residual amount adsorbed onto the inlet surfaces was found to be bleeding off the plumbing in the field. Post-mission laboratory experiments were performed to determine the CIMS instrument’s sensitivity to these organic acids under iodomethane outgassing and the varying water vapor and sample flow conditions encountered during DC3 flights. Laboratory calibration experiments with varying iodomethane concentrations, inlet pressures, and water vapor mixing ratios were performed. Calibrations for hydrogen peroxide and methyl hydroperoxide were used to tune the reagent gas mixture composed of CH3I, CO2, and O2 to best match the observed sensitivities for hydrogen peroxide and methyl hydroperoxide in DC3. Formic and acetic calibration functions were fit to the water vapor and sample flow results. Formic acid displays both a water vapor and inlet pressure dependence, whereas, acetic acid sensitivity was found to be independent of water vapor and inlet pressure within the range of conditions encountered in DC3 and the laboratory. These laboratory calibrations were further evaluated by comparison to in-flight calibrations from the Front Range Air Pollution and Photochemistry Experiment (FRAPPE) to confirm the reliability of the fits and laboratory methodology. The laboratory and field calibration work showed very good correspondence for hydrogen peroxide, methyl hydroperoxide, and acetic acid. However, the apparent formic acid sensitivity fits gave a factor of two greater sensitivity than observed in the FRAPPE formic acid field calibrations. Further comparison to ground and prior airborne measurements suggest the calibration problem lies in the FRAPPE field calibration and the laboratory work was valid for formic acid as it was for DC3 and FRAPPE peroxides and FRAPPE acetic acid. Subsequently, the laboratory calibration fits were applied to quantify DC3 formic and acetic acid. A DC3 research flight (RF 03) case study is presented to illustrate formic and acetic acid behavior when air from a high biogenic source region with identifiable convective storm outflow was sampled.

were performed. Calibrations for hydrogen peroxide and methyl hydroperoxide were used to tune the reagent gas mixture composed of CH 3 I, CO 2 , and O 2 to best match the observed sensitivities for hydrogen peroxide and methyl hydroperoxide in DC3.
Formic and acetic calibration functions were fit to the water vapor and sample flow results. Formic acid displays both a water vapor and inlet pressure dependence, whereas, acetic acid sensitivity was found to be independent of water vapor and inlet pressure within the range of conditions encountered in DC3 and the laboratory. These laboratory calibrations were further evaluated by comparison to in-flight calibrations from the Front Range Air Pollution and Photochemistry Experiment (FRAPPE) to confirm the reliability of the fits and laboratory methodology. The laboratory and field calibration work showed very good correspondence for hydrogen peroxide, methyl hydroperoxide, and acetic acid. However, the apparent formic acid sensitivity fits gave a factor of two greater sensitivity than observed in the FRAPPE formic acid field calibrations. Further comparison to ground and prior airborne measurements suggest the calibration problem lies in the FRAPPE field calibration and the laboratory work was valid for formic acid as it was for DC3 and FRAPPE peroxides and FRAPPE acetic acid. Subsequently, the laboratory calibration fits were applied to quantify DC3 formic and acetic acid. A DC3 research flight (RF 03) case study is presented to illustrate formic and acetic acid behavior when air from a high biogenic source region with identifiable convective storm outflow was sampled. iv ACKNOWLEDGMENTS I would like to thank Brian Heikes who has supported me through this whole process and provided much needed guidance through the many unanticipated roadblocks we encountered. You have opened me up to a world I didn't even know existed and as frustrating as "out of the beaker" chemistry is I wouldn't change it for anything. Thanks to Dan O'Sullivan for your support and for being my pseudo second advisor. I would also like to thank my committee for their insight and support as well through this process. None of this would be possible without the fellow scientists and the "mechs and techs" from both DC3 and FRAPPE. Atmospheric chemistry is truly a group adventure and without each person it would like trying to solve a puzzle without all the pieces. I would also like to thank the many friends I have made in Rhode Island. I could not have made it through this without your support, distractions, advice, coffee runs, Netflix binges, and Brickley's ice cream trips in the summer. I have been lucky to meet each of you and now you are stuck with me forever.
Finally, I would like to thank my family. They have stood by me from the beginning and supported my choice to move across the country to do science. My parents have provided insight and guidance in the way that only parents can and have kept me tethered to reality by reminding me what is really important. Last, but never least, to my brother who no matter how far apart we are it is like no time passed when we are together and I know you always have my back. Here is an official in writing promise to one day go to SDCC. v

INTRODUCTION
Formic acid (HFo) and acetic acid (HAc) have both natural and anthropogenic sources. A representation of the biogeochemical cycle of HFo and HAc is shown in Figure 1 and some of the major sources are discussed here. Primary emissions for both include biomass burning, biogenic sources, and motor vehicles (Khare et al. 1999;Paulot et al. 2011). Secondary production is a significant source for both acids especially from biogenic precursors, secondary organic aerosols, and photochemical production from volatile organic compounds (VOCs) and oxygenated volatile organic compounds (OVOCs) (Khare et al. 1999;Paulot et al. 2011). For example, an indirect VOC source of HFo is the ozonolysis of isoprene which is released naturally by vegetation (Khare et al. 1999). Models, laboratory work, and field measurements agree that the ozonolysis of isoprene is a significant source of HFo (Khare et al. 1999).
The oxidation of other biogenic compounds will also release both acids.
Unfortunately, a great deal of uncertainty remains concerning the yield and mechanisms of these in situ chemical reactions although the critical component is the source strength and fate of the Criegee biradical from alkene ozonolysis RC=CH 2 + O 3 →RCHO+HCOO - (Millet et al. 2015).
Biomass burning provides an important primary and secondary source of both acids to the atmosphere. HAc has a larger direct source from biomass burning than HFo (Paulot et al. 2011). Most HFo and HAc produced in biomass plumes occur via secondary production from VOC precursors. Both organic acids have been measured as primary emissions in motor vehicular exhaust with greater HAc relative to HFo (Kawamura, Ng, and Kaplan 1985). It is hypothesized that both organic acids are released as a result of incomplete combustion which is supported by the high organic acid concentration measured in used oil (Kawamura, Ng, and Kaplan 1985).
Secondary organic aerosols are another potential source for HFo and HAc, though this is a poorly quantified production pathway. The role of organic acids in secondary organic aerosol (SOA) aging has become more widely recognized and studied in the last ten years. SOAs are formed by the chemical transformation of atmospheric organic compounds through oxidation of gas-phase species (Fuzzi and Andreae 2006). Photochemical aging alters the physical and chemical properties of aerosols in the atmosphere (Fuzzi and Andreae 2006). Field studies (INTEX-B and ARCPAC 2008) generated data that support a positive correlation between HFo and HAc production and organic aerosol aging in the atmosphere. It is hypothesized that SOA aging releases both acids into the atmosphere. Current models underestimate the level of HFo and HAc especially when there are no large biogenic or anthropogenic sources nearby (Paulot et al. 2011). For example, HFo and HAc measured in Asian air masses sampled during INTEX-B were expected to have lower concentrations due to scavenging; however, both acids were present at high concentrations which indicates that there must have been a secondary source for both acids (Paulot et al. 2011;Dunlea et al. 2009). SOA production of these acids could account for the gap between modeled and observed concentrations. Paulot et al. (2011) remark that there is a positive correlation between the concentration of HFo and the level of aerosols present. These data help support the hypothesis that organic acids are released or coformed during the aerosol formation and aging process. Similar evidence was collected during ARCPAC 2008, which showed the relationship between HAc and black carbon along with aerosol organic content. This work indicated a potential missing precursor for HFo and HAc coupled with biomass burning (Paulot et al. 2011). Until the recent advancement in measurement technology, it was difficult to measure these aging processes due to the short time scale over which they occur.
Current atmospheric models cannot explain the magnitude and evolution of atmospheric SOAs. Paulot et al.'s (2011) comparison of model and field measurements found that models may be underestimating the concentrations of both organic acids by up to 50%. More accurate measurements of HFo and HAc in the atmosphere would contribute to our understanding of SOA aging processes and improve atmospheric models.
HFo and HAc have the potential to be used as convective transport tracers. The water solubility for both acids is dependent on the pH due to dissociation. Below pH 4, HAc is more soluble than HFo (Pandis and Seinfield 2006). Above pH 4, HFo is more soluble because it has a more efficient dissociation (Pandis and Seinfield 2006). Barth et al. (2007) modeled various mid-latitude storms for peroxides and HFo. Hydrogen peroxide (HP) and methyl hydroperoxide (MHP) are used as deep convective transport tracers (Barth et al. 2007;Snow et al. 2007). HP has a higher Henry's Law constant (H = C aq /P) than MHP (O'Sullivan et al., 1996); therefore, HP will be removed in a storm cloud preferentially via wet deposition. Very little MHP was scavenged in Barth et al.'s (2007) modeled storms: about 7%. Barth et al. (2007) found that HFo in modeled convective outflow was small when assuming the initial aqueous and gasphase concentrations were zero. However, the amount of HFo present depends on the storm type (affecting aqueous phase HFo production) and cloud and rainwater pH (Barth et al. 2007). The majority of HFo formed in clouds is in the aqueous phase. Barth et al. (2007) determined that it may be possible to use HFo to detect cloudprocessed air though it is highly dependent on cloud conditions and the initial concentration of HFo. Unfortunately, there is minimal observational literature using HFo and HAc as deep convective system transport tracers.
Measurements reported here were obtained using a newer analytical technique for atmospheric measurements known as chemical ionization mass spectrometry (CIMS). The major advantages of using CIMS are the rapid response time and increased sensitivity (Crounse et al. 2006;Messer et al. 2000). CIMS directly measures chemical species of interest based upon the molecule's mass-to-charge ratio (m/z) and can rapidly monitor and alternate between several mass-to-charge ratios. A faster sampling rate makes it possible to detect reactions on the storm scale (10s to 100s of meters). The CIMS individual m/z sampling rate is about 0.1 s -1 but it takes about 1 s to sample all selected m/z values. These rates correspond to distances of 0.02 -0.2 km for an aircraft traveling at 200 m/s, such as the G-V used in DC3, and the horizontal scale of a deep convective storm is on the order of 10 km. While CIMS instruments have been used to measure many trace atmospheric species, there have been limited measurements of peroxides and organic acids.
A recent example of airborne measurements for HFo using CIMS was performed by Le Breton et al. (2012). They reported one flight day worth of data and measured HFo mixing ratios between 34.4 pptv and 358 pptv with an average of 142.4 pptv (Le Breton et al. 2012). This was an informative study because they were able to distinguish three different urban plumes with elevated levels of HFo. This work highlights the ability of CIMS to perform rapid, sensitive measurements in-flight. This thesis expands Le Breton's iodide reagent chemistry work to both HFo and HAc.
HP and MHP were target species of interest measured using CIMS during the Deep Convective Clouds and Chemistry (DC3) campaign in 2012. DC3 focused on studying atmospheric chemistry and transport under deep convective storm conditions in the central and eastern United States. Prior to the DC3 mission, an initial mixture of CH 3 I in nitrogen (unknown mixing ratio) was used in the peroxide CIMS instrument.
This reagent mixture did not produce an ion for the measurement of MHP and was discontinued before DC3. A second mixture composed of CO 2 in pure air and blended with pure nitrogen was tested. This reagent mix yielded ion signals for HP

Instrumental Configuration
Continuous gas analysis was performed using a chemical ionization mass spectrometer (THS Instruments, Inc., Atlanta, GA). The chemical ionization mass spectrometer (CIMS) is adapted for peroxides and organic acids and is used in a negative ion mode. This thesis focuses on HFo and HAc, which form cluster ions with the iodide reagent ion, I -, at 173 (I -•HCOOH) and 187 (I -•CH 3 COOH) townsend, respectively. The CIMS normally operates in a "hop" mode, whereby specific massto-charge ratios are sampled sequentially. The sample and analytical system is depicted schematically in Figure 2. Critical system elements include: gas delivery inlet with calibration system, the CIMS, exhaust lines, and a data acquisition and control system. The CIMS is further composed of an ion generation and reaction system, ion selection (declustering ion guide and quadrupole), multi-ion counting detector, controlling electronics, and vacuum pump system.
Ambient, or sample, air enters the CIMS system through a PFA Teflon inlet. In the laboratory, synthetic air mixtures are delivered to the inlet. In airborne field work, the inlet is external to the fuselage and aerodynamically designed to minimize the collection, volatilization, and subsequent analysis of large aerosol and droplet material. The inlet is heated to minimize artifacts by the adsorption of the target gases onto the inlet walls. Calibrations are performed by standard addition to the sample air stream. The calibration gas flows continuously and a "draw-back" design is employed to modulate between sample air and sample air with calibration gas added. In the "draw-back" mode a 2-way valve is opened and a pump, regulated by a mass flow controller (MFC), pulls the calibration gas flow and a small excess flow of sample air to waste. In this configuration sample air without standard is measured. When the valve is closed, the "draw-back" is off and calibration gas is added to the sample air stream. A detailed discussion of the different calibration set-ups is given in Section 2.3. Analytical blanks are determined by passing the sample air stream, with or without calibration gas, through an organic acid trap. A three-way solenoid valve is used to select between the sample air stream or sample air after passing through the trap. A description of the different traps tested is in Section 2.4. Last the inlet flow is split with a portion being pumped to waste and the remainder is introduced to the CIMS system. The excess sent to waste is used to increase the sample volume relative to the plumbing surface area which minimizes instrumental material effects on the sample.
In the CIMS, the sample air passes through a series of chambers to form, select, and quantify the organic acid ion clusters. The first chamber is the ion-sample reaction cell, RXN in Figure 2. In the reaction cell, the sample air is mixed with a reagent ion stream (discussed fully in Section 2.5) of which the bulk is pure nitrogen and controlled by MFCs. The reagent gas passes through a commercial electrostatic eliminator (NRD, Inc., Grand Island, NY) which contains 20 mCi of 210 Po, an alpha emitter, and thus develops the requisite reagent ion stream. Sample air is introduced into the reaction cell through a critical orifice. The sample-reagent ion stream is pumped out of the RXN cell through a second critical orifice into the collisional dissociation chamber, CDC in Figure 2, which breaks apart weak ion clusters in order to lower background signal counts. The remaining ion clusters are mass filtered and counted by a quadrupole mass spectrometer and multi-channel ion detector.
In this work both the sample-in and exit critical orifices were of fixed diameter. The gas volume flow rate across an orifice depends upon the area of the orifice and the pressure differential across the orifice. Upstream of the inlet orifice the pressure is at ambient level. Downstream of the outlet orifice the pressure is However in airborne operations, the inlet pressure decreases with altitude as does the sample flow rate into the RXN chamber, and it is necessary to increase the reagent nitrogen flow rate to maintain a constant pressure within the CIMS flow sections. The effect of sample inlet pressure or sample flow rate on instrument response was examined in the laboratory (Section 2.2). For future airborne work, the sample fixed size critical orifice will be replaced by a variable size orifice which will maintain constant sample and reagent gas flow rates.

Laboratory Set-up
In the laboratory different field conditions are simulated by varying the water vapor and/or the inlet pressure of the sample air stream as depicted in Figure 3. A zero air generator (Aadco Instruments Inc., Cleves, OH) supplies the sample air stream to prevent the addition of organics and excess water into the system. This air stream is split between dry and humidified lines. The dry line comes directly from the Aadco.
The water concentration in the humidified line is controlled with two gas washing bottles and a gas-water equilibration coil immersed in a water bath kept routinely at 288 K ( Figure 4). By changing the ratio of air flow through the dry and humidified lines, it is possible to alter the overall water vapor in the air stream entering the CIMS.
Most laboratory work was done with the coil water bath at 288 K which gives a maximum reaction cell water vapor mixing ratio of 4500 ppm. Normal room temperature is controlled at 293 K. Higher humidity work was performed to expand the reaction cell water vapor mixing ratio range to 7800 ppm for comparison to literature data (Lee et al. 2014;Baasandorj et al. 2014). The temperature in the coil water bath was raised to 298 K and the laboratory room temperature was increased to 303 K to prevent the water vapor from condensing on the tube walls (Thanks to Dallas Hazard).
The inlet pressure is controlled after the humidification with a needle valve (P, Figure 3) and a pressure transducer. The needle valve is able to approximate the atmospheric altitude conditions (sea level to 14 km, approximately 120 hPa) experienced in the field. The sample flow rate changes as a function of altitude (ambient pressure) which will impact the signal response and sensitivity.

Calibration Configuration
The HFo and HAc standards (HCOOH, > 95% and CH 3 COOH, 99.9%, respectively) were obtained from Sigma-Aldrich. There is a <1% HAc impurity in the HFo standard and 2.0-2.5% water for stabilization. Laboratory prepared microfluidic and coil aqueous standards were not standardized against a primary reference material.
In FRAPPE microfluidic aqueous standards were verified by titration against sodium hydroxide (Santa Monica College 2014). The percent errors between the theoretical and titrated concentrations for FRAPPE were 1.00% and 1.51% for HFo and HAc, respectively. The sensitivities are reported with three significant figures (Table 2) for all work as a result of this small error. The FRAPPE aqueous standards were titrated post-mission (2 months later) as well. The standard solutions did not decay substantially with time. There was a 1.91% and 3.67% percent decrease in concentration by the end of the campaign for both HFo and HAc.
Calibrations were performed with coil or microfluidic injection. The laboratory coil calibration system ( Figure 5a) allowed the organic acids to be added in the sample air stream prior to entering the CIMS. Zero air (200 or 400 sccm) was passed through a glass coil in a water bath (nominally 288 K) along with the liquid organic acid standard. HFo (0.3 mM) and HAc (0.9 mM), were acidified (0.02 N H 2 SO 4 ) to keep each acid in its protonated form and thereby ensure partitioning into the gas phase according to each acid's Henry's Law constant. The calibration gas was added to the sample air stream after humidification. The higher humidity work was done with the coil calibration system; therefore, the calibration coil temperature was increased to 298 K. The Johnson et al. (1996) temperature dependent Henry's Law constants were chosen because they provided the closest relationship between the microfluidic and coil work. The Henry's Law constants for HFo are 17.87 and 8.79 M/hPa at 288 and 298 K, respectively (Johnson, Betterton, and Craig 1996). The Henry's Law constants for HAc are 8.43 and 4.05 M/hPa at 288 and 298 K, respectively (Johnson, Betterton, and Craig 1996).
Two versions of the microfluidic system were used to inject the liquid standard into the CIMS via nitrogen gas line. For the first set-up ( Figure 5b) the standard, an aqueous HFo and HAc mixture (1 mM each), contained in a Hamilton glass syringe was injected using a single syringe pump (1 x 10 -6 L/min aqueous flow rate, KD Scientific Inc., Holliston, MA). The liquid standard was vaporized in a heating block (328 K) into a gaseous N 2 stream (0.4 slpm). The disadvantage of this system is that it can only calibrate for peroxides or organic acids. In DC3, a single syringe was used and the priority measurements were the peroxides. After DC3, a second system ( Figure 5c) was set up which allows for calibration of peroxides and organic acids.
Both peroxide and organic acid standards are evaporated into a N 2 gas stream (0.4 slpm) via mixing-Ts and equilibrated in a mixing jar. This system uses two syringes; therefore, the flow rate is halved (500 x 10 -9 L/min) and the aqueous concentrations of the organic acids are doubled (2 mM) to prevent water saturation of the N 2 gas while maintaining a similar reaction cell sensitivity factor (Section 2.8). Both microfluidic standard addition systems can be contained within the CIMS instrument rack. In-flight calibrations were done as part of the FRAPPE program (Section 2.7) in the summer of 2014 with the second microfluidic set-up. These standard additions along with laboratory calibrations are used to reduce organic acid data for DC3.

Blank Configuration
The laboratory blank work used either a Carulite-200® trap or Aadco zero air. This trap was tested with and without the NaOH trap. Unfortunately, there was an outgassing of compounds which interfered at the masses used to measure HAc and HP; possibly from contaminants in the glycerol or methanol used to prepare the trap or from in situ chemical reactions leading to HAc and/or HP production on the Na 2 CO 3 trap. One example is which could lead to the formation of HAc in the trap (Hermans, Jacobs, and Peeters 2007). It is also possible that an isomer was formed such as glycolaldehyde which has the same molecular weight as HAc.

Reagent Gas
The iodomethane (CH 3 I) reagent gas laboratory calibrations were performed using two different mixing ratios of CH 3 I in N 2 . The initial iodomethane mixture (0.33% in N 2 ) was prepared similarly to Le Breton et al.'s work (2012) with the exception water vapor was not added. The dynamic range of the iodomethane MFC was insufficient to reduce the iodomethane concentration of the "Le Breton" mixture enough to represent DC3 as O 2 ion signals were too small. The 0.33% iodomethane mixture was then serially diluted twice with N 2 gasfirst to 0.025% and then to 5 ppm. The latter dilution was used with a MFC to prepare custom reagent gas blends of CH 3 I, CO 2 , O 2 , and N 2 sufficient to yield responses for both Iand O 2 cluster ions which overlapped those observed in DC3. A mixing cross was used to blend the CO 2 in air (400 ppm), the CH 3 I in N 2 (5 ppm), and the pure N 2 reagent gases (Table 1).
Each gas flow was controlled via MFCs.
Calibrations at the different inlet pressures and specific humidities were performed. The CO 2 in air mass flow rate was held constant at 0.080 slpm. The 5 ppm CH 3 I mixture was added at 5 different flow rates ranging from 0.0005 to 0.0025 slpm in 0.0005 slpm increments. Table 1 lists the CH 3 I, CO 2 , and O 2 mixing ratios in N 2 of the reagent gas blend as a function of CH 3 I mass flow rate and inlet pressure (recall N 2 flow varies with inlet pressure to maintain a constant reaction cell pressure).
This made it possible to explore how changing CH 3 I will impact the signal response of both peroxides and both organic acids. During the variable CH 3 I experiments, blanks were performed using the Aadco system. For each inlet pressure (120, 180, 360, 600, and 1013 hPa), the different CH 3 I flow rates were all performed in one day to avoid daily instrumental variance.

Deep Convective Clouds and Chemistry Experiment
The Deep Convective Clouds and Chemistry (DC3)  using the single syringe microfluidic system described in Section 2.3. The solution standards were 5.7 mM for HP and 4.8 mM for MHP and the syringe flow rate was 1 x 10 -6 L/min.

Front Range Air Pollution and Photochemistry Experiment
The provide the best combination of peroxide and organic acid blanks.

Data Analysis
The organic acid reaction cell mixing ratios (sensitivity factors) were calculated for both the peroxides and organic acids. The reaction cell pressure was kept constant at 22 hPa. The cell temperature was between 298 and 308 K as randomly checked during flight and in the laboratory. These mixing ratios or concentrations are directly proportional and mixing ratio was simply used as the analyte measure of choice. For standard addition using the coil equilibration sources, the reaction cell mixing ratio is given by: where C is the aqueous concentration of the standard in the coil (molar, M), H is the Henry's Law constant (M / hPa), P 1013 is the coil reference pressure (1013 hPa), F c is the mass flow rate of air through the coil (slpm), F Aadco is the total flow (slpm) through the inlet prior to the inlet pressure valve, F sample is the mass flow rate across the reaction cell critical orifice (slpm), and F CIMS is the mass flow rate across the reaction cell outlet critical orifice (4.68 slpm). F sample is calculated by difference by subtracting the reagent gas and N 2 mass flow rates from F CIMS .
Alternatively when using the microfluidic addition system, the standard gas mixing ratio in the reaction cell is calculated by: where c is the aqueous concentration of the standard within the syringe (M), F syr is the syringe aqueous flow rate (L min -1 ), 22.4 (L mole -1 ) is a conversion factor for standard liters of gas per mole of substance at STP (273.15 K and 1013.25 hPa), F std is the mass flow rate of the standard carrier gas, F excess is the inlet excess (slpm) removed from the CIMS to increase the sample size (Section 2.1), and F CIMS and F sample are the same as above.
The peroxide and organic acid sensitivity factors are impacted by the inlet pressure because changing the pressure affects the sample flow (F sample ), the inlet excess flow (F excess ), and the reagent N 2 flow (F N2 ). As ambient pressure decreases the sample and excess flow rates decrease and F N2 increases. Table 2 shows the calculated sensitivity factors used for both organic acids with respect to the inlet pressure for the three calibration systems: coil, one-syringe, and two-syringe. As discussed above, only the FRAPPE aqueous standards were standardized against a primary reference material. There is a small percent error so it is assumed that both laboratory and FRAPPE standards are accurate to three significant figures.
The water vapor mixing ratio in the reaction cell is used to develop sensitivity factors as a function of specific humidity. It was calculated assuming water equilibration in the humidified line at the set point temperature of the coil water bath (normally 288 K and occasionally 298 K; see above): where P sat (hPa) is the water saturation pressure at 288 K (or 298 K), F H2O is the air flow rate (slpm) through the humidified line and P 1013 , F Aadco , F sample , and F CIMS are the same as above.

RESULTS
The following results establish the calibration process used to quantify the DC3 organic acids. The coil and one-syringe microfluidic work was with the "Le Breton" iodomethane regent gas mixture (Section 3.1). After previous work with HP and MHP found that sensitivity is water and pressure dependent, organic acid work with the Le Breton mixture was used to check the published method. There was a pressure dependency for both organic acids for the one-syringe calibration system though it was not the same relationship as with the diluted Le Breton mixture. The diluted Le Breton iodomethane mixture (Section 3.2) was prepared because the original mixture was too concentrated to accurately tune to DC3. The water and pressure dependencies were explored in greater detail for this diluted mixture and laboratory fits were derived for both organic acids (Section 3.3). HFo was fit as a function of water and pressure and HAc was fit with a constant value. The laboratory calibration technique (of altering water vapor and inlet pressure) was verified by comparing laboratory and FRAPPE sensitivities (Section 3.4). HFo laboratory calibrations were a factor of two greater than FRAPPE. HAc had good agreement between laboratory and FRAPPE as did the HP and MHP oxygen clusters though they are not reported here. The diluted Le Breton mixture was assessed at five different flow rates to find a balance between Iand O 2 ion clusters. Different iodomethane flow rates were also tested to find the most appropriate iodomethane mixing ratio for DC3. Laboratory calibrations for the HP iodide cluster were compared to DC3 HP calibrations (Section 3.5) to determine the appropriate iodide flow rate. Finally, the study regions for DC3 are described and vertical profiles for both organic acids in each region are presented (Section 3.6).

Le Breton Mixture
The laboratory calibration sensitivities for the coil (open circles) and a onesyringe microfluidic standard addition system (dots) were compared as a function of reaction cell water vapor for HFo and HAc. The results are shown in Figures 6 and 7, respectively. In this comparison, the 0.33% iodomethane reagent gas mixture (the "Le Breton" mixture) was used. The organic acid sensitivities at five different inlet pressures were determined (120, 180, 360, 600, and 1013 hPa) and these are identified in Figures 6 and 7 as well. Only experiments at the two highest pressures (600 and 1013 hPa) were conducted using the coil calibration system. The laboratory coil sensitivities are a factor of two to three greater than the microfluidic sensitivities for HFo and comparable for HAc using the Johnson et al. (1996) Henry's Law constants.
The error bars represent one standard deviations and the calculations are described in  (1996) values. This indicates that Johnson et al. (1996) accurately quantified the HAc gas-phase partitioning. The laboratory calculated Henry's Law constants for HFo are lower than the Johnson et al. (1996) values. The difference in Henry's Law constants could be due to a higher gas-phase partitioning through the coil system than measured by Johnson et al. (1996). Therefore, the HFo mixing ratio in the reaction cell is greater than expected which results in an overestimation of the sensitivity of the CIMS towards HFo. This overestimation explains the difference between the coil and microfluidic system. As with the Le Breton calibration work above, the HFo sensitivity increased as a function of water and there is no trend for HAc outside of the instrumental variance. The lowest HFo and HAc sensitivities are at 1013 hPa which is unexpected because the most water and Iis present in the reaction cell at ambient pressure. As a result, it was hypothesized that the highest sensitivity (especially for HFo) would be at ambient pressure. Since this was not the case, the dynamics between iodomethane, pressure, and water for both acids were investigated.
The impact of iodomethane on sensitivity was examined using the maximum sensitivity at each inlet pressure for both acids as a function of the iodomethane flow rate. This is illustrated in Figures 11 and 12 for HFo and HAc, respectively. Both acids are fit with a linear and quadratic fit and 95% confidence intervals. Three outlier points for HFo at 600 hPA were removed from the data set as they fell well outside the 95% confidence interval for both fits. One day of HAc work at 1013 hPa was markedly below the rest of the work at that pressure so that day was removed as well from any subsequent work. While it is possible that this low day is due to human error, there is not a similar decrease in sensitivity for HFo or the peroxides. Visually both fits represent the data (with the exception of 600 hPa for HFo) so the F-test was used to determine the best fit. It was found that there is no significant difference between the fits for both acids at all the pressures within the tested flow rate range. The sixth panel in Figures 11 and 12 show maximum sensitivity regardless of pressure as a function of iodomethane flow rate. The standard deviation bars represent the impact of pressure on sensitivity. The sensitivity for both organic acids tripled from the lowest (0.0005 slpm) to the highest (0.0025 slpm) iodomethane flow rate.

Water and Pressure Dependencies and Calibration Fits
Water and pressure dependencies were studied using the ratio of the sensitivity to the maximum sensitivity. Only the work at the 0.0005 slpm iodomethane flow rate is reported. This ratio removed the bias of daily variation in the instrument. An average of this sensitivity ratio for each water vapor value was used to combine the different experimental days. These ratios as a function of the reaction cell water vapor are shown in Figures 13 and 14 for HFo and HAc, respectively. The sixth panel shows all the pressures together as a function of the water mixing ratio. For each pressure, the maximum sensitivity (ratio of one) for HFo was at the highest water vapor mixing ratio. The HAc maximum sensitivity was not at the same water mixing ratio for the different pressures or between days. This is why the ratio is always less than one. The lowest HAc sensitivity was always within 50% of the maximum sensitivity regardless of pressure. This supports what was seen above that the HAc-water relationship is small and not a critical determining factor to the sensitivity.
These ratios were not sufficient to study pressure dependencies. The sensitivities at each pressure were calculated using the sensitivity ratios and the averaged maximum sensitivity for each pressure. This calculated sensitivity will be referred to as the "cleaned-up sensitivity". The cleaned-up sensitivities were calculated for each iodomethane flow rate though only the sensitivities at 0.0005 slpm are reported. The cleaned-up sensitivities (and one standard deviation) for HFo at each pressure as a function of the reaction cell water vapor are in Figure 15. Even with the cleaned-up sensitivity there is still a noticeable difference between the pressures as seen in the sixth panel. Therefore the HFo fit must be a function of water and pressure. The HFo sensitivity initially increases with water vapor and reaches an asymptote at the highest water mixing ratiosespecially at higher pressure which also has the largest water mixing ratio. A Michaelis-Menten fit y = Ax B+x +C (4) is used to represent this increase and subsequent plateau in sensitivity at higher water.
The water dependency coefficients for HFo also change as a function of pressure. As mentioned above the highest pressure (1013 hPa) sensitivity was unexpectedly low.
Therefore, the three coefficients were determined without 1013 hPa. The equations for the three coefficients are Several different fits as a function of water and/or pressure were considered but none of them accurately reproduced the laboratory or field sensitivities or contributed enough explanation of variance to warrant additional fitting terms. It was decided to use an average constant value (752.5 cts/ppb) for HAc sensitivity. The laboratory data and fits are compared to field data to confirm the accuracy of the fits.

FRAPPE Organic Acid Calibrations
In-flight calibrations from FRAPPE are compared to the cleaned-up laboratory sensitivities for HFo and HAc (Figures 17 and 18). The laboratory data are the same as The laboratory fits were tested using the FRAPPE calibration data and reasonably represented the in-flight calibrations.

DC3 Hydrogen Peroxide Calibrations
The final calibration step was to determine which iodomethane flow rate of the 5 ppm in N 2 mixture best represented DC3 conditions. This was done using HP because it clusters with iodide and was calibrated for during DC3. The laboratory sensitivity for the HP and iodide cluster at 161 townsend (I -•H 2 O 2 ) as a function of reaction cell water vapor is in Figure 19 for the five different iodomethane flow rates. HFo and HAc using the 0.0005 slpm iodomethane flow rate will be used to quantify both organic acids during DC3. Each study region has higher HAc mixing ratios compared to HFo. In general, all three profiles have a decrease in HFo up to 6 km followed by an increase either back to boundary layer mixing ratio values or higher (a so called "C" shape). This profile is most pronounced in the Eastern region. The Eastern region also has the highest altitude measurements and the HFo sensitivity starts to decrease again above 12 km. The highest mixing ratios of both organic acids in the Oklahoma-Texas region were measured at 2 km. The Colorado HFo profile has more HFo at the top of the profile than in the boundary layer. The HAc altitude trend is not as strong in any of the study regions though the mixing ratio decreases up to 6 km. The Eastern region has the biggest difference between both acids at high altitude. The largest range of mixing ratios (represented by the 10 th -90 th percentile) is in the Oklahoma-Texas region and is reflected in the peroxide and organic acid profiles.   (Lee et al, 2014). Lee et al (2014) found the sensitivity plateaus and declines when the reaction cell water was above 2200 ppm. Unlike Lee et al. (2014), our HFo sensitivity did not decrease towards the higher water mixing ratios though it did plateau -especially the ambient pressure work (1013 hPa). The water mixing ratio in the reaction cell could not be increased enough to achieve a decline in sensitivity. The maximum water mixing ratio in the reaction cell during laboratory experiments was 7800 ppm. The FRAPPE in-flight calibrations covered a larger water mixing ratio yet there was still no decline in sensitivity. It is likely that variations in the CIMS configurations cause the differences in response as a function of water. Lee et al. (2014) used a permeation tube and not a gaseous mixture to add the iodide. It is possible that a difference in the total amount of iodide available impacts clustering preference. Their standard addition system differed from ours as well as they had no water present in their driest conditions. We never had a completely dry system as the standards were prepared in aqueous solutions. Lee et al. (2014) also had a higher reaction cell pressure (90 hPa) which could impact reaction cell chemistry. Le Breton et al. (2012) did not discuss a decrease in sensitivity as a function of water but they did note that water was needed to promote clustering. They There is minimal other literature data available for the HAc and iodide cluster. Acetate is a commonly used reagent gas for CIMS systems; therefore, HAc is not detected.

Ionization Chemistry and Water Sensitivity Dependence
HAc has been measured using proton transfer reaction mass spectroscopy (PTR-MS) and HAc sensitivity decreased with the addition of water (Baasandorj et al. 2014).
Unfortunately, a direct comparison is not possible as this a different instrumental setup and reagent chemistry.

Ionization Chemistry and Iodide Concentration Dependence
The goal of the dual reagent gas system was to monitor HP, MHP, HFo, and HAc. This requires a balance between the iodide and oxygen chemistry. In general, as the proportion of iodomethane increases the sensitivity of the CO 2 and O 2 clusters decreases. The negative impact on sensitivity is greater for MHP than HP as discussed earlier. MHP is an important target species and it is critical to find a reagent ion balance that does not reduce the sensitivity significantly. This CIMS is not as sensitive to HAc as HFo thus a decrease in iodide results in a larger reduction of HAc sensitivity. Finding a balance between the two reagent gases ultimately depends on a balance between MHP and HAc. The lowest iodomethane flow rate was chosen for FRAPPE as there was a quantifiable HAc addition with a minimal impact on MHP.
Based on the HP and iodide cluster (at 161 townsend) this iodomethane flow rate also represents the iodide mixing ratio during DC3. It was fortuitous that there was enough iodide present during DC3 to promote organic acid clustering without significantly impairing the MHP sensitivity.

Ionization Chemistry and Pressure Dependency
It was originally thought that the HFo maximum sensitivity would be at the highest pressure since that is the highest water mixing ratio and HFo sensitivity increases as a function of water. However, the highest pressure actually had the lowest sensitivity for both organic acids. At the 0.0005 slpm iodomethane flow rate, the maximum sensitivities for HFo and HAc were at 600 and 360 hPa, respectively.
Unfortunately, at this time it is not possible to fully understand the underlying cause.
Pressure dependency will not be an issue for future work with this CIMS as a variable orifice will be added that maintains constant sample and reagent gas flow rates. It is possible that this pressure dependency is a function of the iodide-oxygen relationship in the reaction cell. As discussed above, the addition of iodide results in a decrease in sensitivity of the oxygen dependent clusters (i.e. MHP). However, it had not been explored that the oxygen could negativity impact the iodide clusters. A chemical computational diagnosis to test the strength of the ion clusters is not accessible at this time. This limits the extent to which this problem can be investigated. Even though this situation will be resolved for future field work it still needs to be explored for past missions. A potential reason for the unexpected pressure relationship is discussed here.
The ratio of iodide to oxygen through the ion source is constant regardless of the inlet pressure because the only source of oxygen is from the CO 2 •O 2 reagent gas.
The iodide to oxygen ratio varies inside the reaction cell due to the oxygen in the sample flow. The changing sample flow rate results in a dilution of iodide with respect to oxygen as pressure is increased. Therefore, if there is a negative impact on the iodide chemistry it would be at the greatest sample flow rate (1013 hPa) as this represents the greatest oxygen addition to the system which is seen for both organic acids. However, if this was a linear relationship then the greatest sensitivity should be at 120 hPa. As this was not the case it is likely that there is a threshold for the iodide sensitivity and not a linear dilution factor. It is possible that at 600 hPa HFo is still able to efficiently cluster with iodide though greater than 600 hPa the proportion of oxygen becomes too large to support iodide clustering. As mentioned above, the CIMS is not as sensitive to HAc which could explain why the maximum sensitivity is at an even lower inlet pressure (360 hPa) than HFo. Therefore, the original thought process that the highest sample flow rate would result in the highest HFo sensitivity is flawed. Using calibration fits as a function of inlet pressure eliminates the need to fully comprehend the mechanism of the iodide loss in sensitivity. The HFo sensitivities showed the greatest dependency on inlet pressure and are fit as a function of water and pressure. HAc also had a pressure dependency though the addition of a pressure dependency term does not add significant improvement to the HAc fit and is not used.

Ethanol Interference
Concentrated ethanol responds at 173 townsend (I -•C 2 H 6 OH) and can be used as a CIMS system leak check. The CIMS sensitivity towards ethanol was tested to confirm that ethanol could serve as a leak check without significant ambient signal interference. Ethanol was added via the 2-syringe microfluidic and coil standard addition systems. The microfluidic standard (~ 6mM) was prepared to be equivalent to HFo standard during FRAPPE. The coil standard (0.2 mM) was prepared at a similar concentration to MHP due to the similarities in the Henry's law constant. The Henry's Law constant (0.20 M/hPa) used for the coil system was from Sander (1999) and is temperature independent. Similar to HFo, ethanol's sensitivity increased with the addition of water though ethanol's sensitivity was significantly lower than HFo at both the minimum (100x lower) and maximum (30x lower) water concentrations used in the laboratory. Baasandorj et al (2014) performed similar work with a PTR-MS and found HFo's sensitivity to be 5-20x higher than ethanol. We strongly recommend verifying that the ethanol measured in the study region is low (background mixing ratios).

DC3 Case Study: Research Flight 03
The DC8 and GV sampled convection in the Alabama radar region during Research Flight 3 (RF 03) on May 21, 2012 and the flight tracks for both aircraft are in Figure 25. The GV sampled convective outflow at several altitudes along the Alabama-Tennessee border (19:00 to 20:45 UTC) and then executed a spiral descent to 0.9 km before returning back to base. HFo and HAc measurements as a function of time and altitude are in Figure 25. The altitude (m) is also given as a function of time for reference. The maximum HFo and HAc mixing ratios were 720 ppt and 1900 ppt, respectively. The largest HFo mixing ratios were between 8 and 10 km during the storm sampling with the peak at 8.6 km. The largest HAc mixing ratios sampled were at the lowest altitude. The ratio of HFo to HAc is less than one for the majority of the flight reflecting the substantial HAc sampled regardless of altitude. Higher HAc relative to HFo was measured during the whole campaign in each of the study regions ( Figures 22, 23 and 24).
Previous field measurements have reported varied results about the proportion of HFo to HAc. Reiner et al. (1999) and Talbot et al. (1996)  Convection in Alabama was the weakest of the three study regions (Barth et al. 2014). Convective outflow is characterized by a relative increase of MHP to HP.
Convective transport is identified above 8 km with the strongest outflow above 10 km. This is confirmed by Barth et al. (2014) using other soluble and boundary layer tracers in conjunction with peroxides. There is an enhancement of HAc relative to HFo at the highest altitude (above 10 km) during RF 03 though at this time it is not enough evidence that organic acids can be used as convective transport tracers. Peroxides are a more reliable convective transport tracer because there is a clear reversal in the relationship between the two from the surface (HP>MHP) to the storm outflow (MHP>HP). The solubilities of organic acids are more complicated due to the pH dependency and thus not as reliable of a tracer. HFo and HAc could be used as a secondary tracer especially in a biogenically dominated region. Using organic acids as convective transport tracers needs to be investigated further.
There was a descent to 0.9 km towards the end of the flight over Tennessee.
There was an increase in HFo and the highest measured HAc for this flight. As mentioned above, Tennessee is known for high biogenic emissions. An air quality cancer study in Tennessee found elevated levels of acetaldehyde in the summer primarily as the result of secondary formation from biogenic VOCs (Díaz-Robles, Fu, and Reed 2013). The OH oxidation of acetaldehyde forms HAc from the peroxy acetyl radical. Acetaldehyde, also measured with the TOGA, is shown as a function of altitude in Figure 26. Acetaldehyde increased from 100 ppt at 10 km to 800 ppt near the surface. The lower HFo response near the surface could be because HFo was scavenged in the storm. HFo is more soluble than HAc at typical rain pH values.
Unfortunately, there is no low altitude work during the other flights in the region for comparison. There are higher altitude flights in this region that see the elevated HAc relative to HFo though the mixing ratios are higher for RF 03 than the other flights.
There were only two days classified as Alabama storm sampling and RF 03 was the stronger case (Barth et al. 2014). The other flight had less than 500 ppt for both acids when sampling in a similar region. It is possible the stronger storm lofted more organic acid precursors than during the other flight. The other difference is that there should be more iodomethane earlier on in the campaign which could result in a difference of sensitivities. However, flights in other regions later in the campaign still saw substantial responses for both acids.

CONCLUSIONS
This work successfully replicated Le Breton's iodomethane reagent gas mixture for the measurement of HFo and HAc. Based on microfluidic and coil experiments it was possible to calculate Henry's Law constants for both organic acids which are comparable to literature values. A blended reagent gas mixture (CO 2 in air and CH 3 I) was developed and deployed in the field during FRAPPE measuring HP, MHP, HFo, and HAc. Calibration fits using the blended reagent gas were determined for HFo (as a function of inlet pressure and water) and HAc (an average value independent of inlet pressure and water). These calibration fits made it possible to quantify both organic acids measured during DC3. All three DC3 study regions were characterized by greater HAc relative to HFo throughout the altitude profile. Both organic acids had a "C" shaped altitude profile for the majority of the flights. A test case, RF 03, was highlighted which represents a deep convective storm influenced by biogenic tracers.    Figure 1: Representation of the biogeochemical cycle for formic and acetic acid with an emphasis on the sources (blue) and sinks (purple).   Formic Acid (m/z 173) Maximum Sensitivity as a function of CH 3 I Figure 12: Acetic acid maximum sensitivity at each iodomethane flow rate for the 5 ppm mixture regardless of reaction cell water vapor. The black line is the linear fit and 95% confidence interval and the green line is the quadratic fit and 95% confidence interval. The sixth panel (bottom right) is the average maximum sensitivity and one standard deviation regardless of pressures at each flow rate.         where F N2 is the nitrogen flow rate and F reag is the reagent flow rate.

FIGURES
The coil system sensitivity (Eqn 1) error must account for the error associated with the sample flow rate (F sample ) and the calibration flow (F m = F c /F Aadco ). It is assumed that the error in the standard preparation is much smaller than the other sources of error and is neglected. It is also assumed that the error in Henry's Law constants are negligible. The coil system reaction cell (Eqn 1) error ( X rxncell, coil ) is where F T is the total flow (Aadco: dry and humidified and calibration (Fc)).
The microfluidic sensitivity (Eqn 2) error (δX rxncell,syr ) needs to account for the sample and excess flow rates. It is assumed that the error in the syringe flow rate is negligible and F CIMS is treated as a "true value". for F B which is F sample + F excess . The two-syringe microfluidic system will be slightly different because there are two reagent gases but the general formula is the same.