A new inlet and instrument have been developed for the rapid measurement of gas phase nitric acid (HNO3) from an airborne platform. The inlet was kept near ambient temperatures with a very short sampling time (100 ms) to minimize desorption of particle nitrates. In addition, inlet surface adsorption problems were minimized by the use of extruded perfluoroalkoxy as a sampling material. Nitric acid was detected by selected ion chemical ionization mass spectrometry using deprontonated methanesulfonic acid as a reagent ion. Laboratory tests showed no interferences from NO, NO2, NO3, and N2O5 under wet (relative humidity (RH) = 100%) or dry (RH = 0%) conditions at levels exceeding those found in the troposphere. The inlet and instrument were flown on the NASA P-3B aircraft as part of the NASA Transport and Chemical Evolution over the Pacific (TRACE-P) field campaign off the coast of Asia during February–April 2001. Nitric acid was measured every 5 s for a 3 s integration period with a limit of detection of ∼10 parts per trillion by volume (pptv). The instrument was calibrated by the addition of isotopically labeled H15NO3 near the front of the ion source on a continual basis. Absolute uncertainties including systematic errors are the limit of detection (10 pptv) plus ±20% for HNO3 > 200 pptv, ±25% for HNO3 100–200 pptv, and ±30% for HNO3 < 100 pptv (±2 σ). Rapid changes in ambient HNO3 were resolved, suggesting minimal influences from instrument surfaces. Finally, the measurements compared favorably with the University of New Hampshire’s mist chamber/ion chromatography instrument flown on board the NASA DC-8 aircraft during two intercomparison flights. The in-flight performance of the instrument is demonstrated under the wide range of conditions observed in TRACE-P.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 Nitric acid, HNO3, plays important roles in both the gas phase and condensed phase chemistry of the troposphere. Gas phase HNO3 is formed by the oxidation of the nitrogen oxides NO and NO2 (NOx), species that play important roles in ozone photochemistry. Tropospheric HNO3 has a relatively long lifetime with respect to photolysis (weeks) and reaction with OH (weeks). The dominant loss mechanism is by removal onto particle surfaces, followed by either wet or dry deposition, with a heterogeneous lifetime of a few days [Liang et al., 1998]. In the condensed phase, HNO3 usually dissociates into particulate nitrate, NO3−. The availability of gas phase HNO3 has been shown to be important in liquid particle growth and composition [Kerminen et al., 1997; Adams et al., 1999]. Furthermore, HNO3 uptake by particles indirectly affects cirrus cloud properties by altering the deliquescence behavior of salt particles [Lin and Tabazadeh, 2002] and by activating cloud condensation nuclei [Laaksonen et al., 1997]. Modeling studies suggest that gas phase HNO3 assists in the development of unhealthy particulate matter in urban areas [Meng et al., 1997]. Because of the above roles of HNO3 as a sink for NOx and its significance to aerosol particle dynamics [Kerminen et al., 1997; Adams et al., 1999], the chemistry of HNO3 is important toward understanding issues such as photochemical smog [Meng et al., 1997], acid deposition [Galloway, 1995], climate change [Tabazadeh and Toon, 1998; Laaksonen et al., 1997; Lin and Tabazadeh, 2002], and human health [Meng et al., 1997].
 Understanding the chemistry of HNO3 has been further complicated by the numerous challenges in its measurement, especially on airborne platforms where conditions change rapidly. Resolving HNO3 concentrations in thin layers of particles, for example, requires an instrument with high sensitivity and short integration times. Furthermore, instrument surfaces should have minimal influence on the sampling of the ambient air. Otherwise, rapid changes in gas phase HNO3 are buffered by adsorption/desorption of HNO3 on instrument and inlet surfaces, and thus the measurements lose their time response. One method used to minimize surface effects is by heating the inlet surfaces [Neuman et al., 2000, 2002]. However, heating induces the possibility that particulate nitrate may desorb into gas phase HNO3 and thereby result in artificially inflated gas phase HNO3 measurements. Thus the characteristics of an ideal HNO3 instrument for aircraft studies are high sensitivity and temporal resolution, short sampling times, inlet surfaces with minimal adsorption problems, and inlets kept as close to ambient conditions as possible.
 The mist chamber/ion chromatography technique [Talbot et al., 1997, 1999] has been used extensively on aircraft for measuring HNO3 as well as a variety of other soluble compounds simultaneously. Briefly, ambient air is pulled into the cabin at very high flow rates (1500–3000 standard liters per minute (sLpm)) and collected on a fine particle mist. The corresponding solutions are analyzed by ion chromatography. The technique is extremely reliable and sensitive (3 parts per trillion by volume (pptv)) but has relatively long integration periods (minutes) for aircraft platforms. Furthermore, because particles less than ∼2.5 μm in aerodynamic diameter are sampled by the inlet, fine nitrate-containing aerosols may complicate the gas phase measurements.
 Airborne chemical ionization mass spectrometer (CIMS) instruments have shown high time resolution (<1 Hz) with excellent sensitivity (>1 count pptv−1 s−1) and selectivity. Neuman et al. [2000, 2002] use SiF5− ion chemistry which is extremely sensitive (1–10 ion counts pptv−1 s−1) and fast (≤1 Hz), but their use of heated perfluoroalkoxy (PFA) inlets at 50°C may complicate the distinction between particulate and gas phase HNO3, particularly in polluted areas. A number of studies have used CO3−(H2O) ion chemistry for sensitive HNO3 measurements, but in all cases the inlets remained well above ambient temperatures [Möhler and Arnold, 1991; Arnold et al., 1992; Reiner et al., 1998; Schneider et al., 1998; Miller et al., 2000; Hanke et al., 2002]. To acquire accurate budgets of HNO3, unambiguous measurements of gas phase HNO3 need to be made near ambient conditions and at high temporal resolutions.
 To this end a new inlet and CIMS instrument design have been developed and characterized to measure HNO3 from an airborne-based platform. The instrument used deprotonated methanesulfonic acid (CH3SO3−, MSA) as a reagent ion for HNO3 detection and a unique choice of inlet design and materials for the sampling of the ambient airstream. The instrument was flown on the NASA P-3B aircraft as part of the NASA Transport and Chemical Evolution over the Pacific (TRACE-P) field experiment off the coast of eastern Asia during February–April 2001 [Jacob et al., 2003]. Results from the field and in the laboratory suggest that the technique has a low limit of detection (10 pptv) and a fast time response (5 s). A description of the behavior and characteristics of the instrument under a wide variety of laboratory and field conditions is presented.
 The detection of HNO3 was conducted using selected ion chemical ionization mass spectrometry with deprotonated methanesulfonic acid (MSA) as the reagent ion. The instrument was part of a four-channel mass spectrometer system also containing components to measure hydroxyl radicals (OH), gas phase sulfuric acid (H2SO4), gas phase MSA [Mauldin et al., 1999], and components for measuring peroxy radicals HO2/RO2 [Cantrell et al., 2003]. Each channel had independent ionization schemes, flow controls, pressure controls, electrostatic lenses, quadrupoles, and electron multipliers, but they shared a common vacuum housing inside the aircraft. The vacuum housing was pumped by three 1000 L s−1 turbomolecular pumps in four stages. The first stage was pumped at a pressure of 10−2 torr, the second stage of lenses and skimmers was pumped at 10−3 torr, the quadrupoles were pumped at ∼6 × 10−5 torr, and the Channeltron electron multipliers in the last stage were at 3 × 10−5 torr. The four-channel system and corresponding inlets were located on the front, port side of the P-3B aircraft. Figure 1 shows a photo of the inlet and its location on the P-3B aircraft, and Figure 2 shows a schematic of the HNO3 inlet and instrument. The inlet/instrument consisted of four parts: (1) a long, shrouded duct to straighten and slow the ambient airflow outside the plane, (2) a transport tube to bring the sampled air toward the airplane and for calibration, (3) an ion-molecule reaction region, and (4) the vacuum housing of ion lenses and quadrupoles inside the plane. For clarity, the general term “inlet” will refer to all parts of the instrument outside the airplane (1–3 above).
 The theory and rationale of design of the shrouded duct originated from Eisele et al. , and thus only a general overview of this part of the HNO3 inlet will be described. The shrouded duct (l = 76.8 cm, 7.6 cm ID, 8.9 cm OD) used in this study was scaled down by a factor of 1.5 from the one described by Eisele et al. . The aluminum duct was tipped 9° away from the aircraft and secured at the back by an aluminum pylon that extended 15 cm from the fuselage of the aircraft (Figure 1). The front, center of the duct was ∼40 cm away from the aircraft surface, or ∼3 times the distance that the boundary layer was calculated to expand out from the aircraft at this station. A shroud of 19.6 cm length (16.0 cm OD, 13.3 cm ID) had an elliptically shaped surface to minimize turbulence effects while straightening, but not slowing, the airflow. The duct was concentric with the shroud, beginning 9.3 cm inside the shroud and continuing to the back of the pylon. The purpose of the duct was to transport the sampled air from a turbulence-free region to the transport tube and to slow the flow by about an order of magnitude relative to the free air speed (110–160 m s−1) via a restricting orifice (r = 1.9 cm) on the back end of the duct. Wind tunnel tests for the inlet in the work of Eisele et al.  indicated that the airflow remained nonturbulent in the center of the duct at angles of attack <17°, and the scaled-down version of inlet used in this study was expected to behave similarly. Although no wind tunnel testing of the inlet occurred, coarse grid modeling of the airflow around the inlets and flow separation were conducted using StarCD software.
 A transport tube (2), located ∼65 cm downstream from the front of the duct, pulled a sample flow of 4–6 sLpm air toward the ion source region (3). The transport tube (1.9 cm ID, 2.2 cm OD, l = 10 cm), composed of extruded PFA, was transverse to the duct, and its top was just below the centerline of the inner duct. The remainder of the flow in the duct (∼104 sLpm depending upon air speed and ambient pressure) was vented out the back through the restricting orifice. On the forward facing side of the transport tube a semicircle notch of r = 1 cm was cut out near the top, and a PFA cap covered the top of the transport tube. In this way the air that entered the tube generally was pulled down toward the ion source. Because the transport tube extended to the middle of the duct, it was unavoidable that turbulence developed in the duct near and downstream of the transport tube. However, air sampled by the transport tube should have had minimal contact with the aluminum surfaces of the duct up to this location. As will be demonstrated later, observational data on the response of the inlet to changes in gas phase HNO3 concentration support this assertion.
 Air sampled into the transport tube underwent an 81° deflection, and subsequent turbulence and contact with transport tube surfaces was unavoidable. Although cartridge heaters and a resistance temperature detector (RTD) were placed in the aluminum housing surrounding the transport tube, these heaters were never activated during flight and were only used to heat the inlet on ground if necessary (e.g., for cleaning). Temperature readings indicated that the transport tube remained within 10 K of the ambient temperature during flight. Because the sampled air readily contacted the transport tube walls at near ambient temperatures and humidities, it was necessary to find a material that was particularly inert to surface adsorption and desorption of HNO3. Neuman et al.  reported that PFA tubing was an optimal choice for inlet materials for HNO3 sampling. However, even PFA had significant adsorption problems below 10°C, and Neuman et al.  concluded that inlets needed to be heated to at least this temperature.
 A very similar inlet to the one described above was flown on the National Science Foundation/National Center for Atmospheric Research (NSF NCAR) C-130 aircraft as part of the Tropospheric Ozone Production about the Spring Equinox (TOPSE) field campaign from February to May 2000 from Colorado northward to the Arctic Ocean. Although no robust HNO3 measurements were made during this campaign due to numerous problems, significant advances were made in inlet design and characterization from in-flight tests. The inlet initially used regular (machined) PFA as a material for a transport tube (length 16 cm, diameter 1.9 cm) with the finding of similar conclusions by Neuman et al. . In other words, whenever the inlet became even slightly cold, the response of the inlet to changes in gas phase HNO3 was very slow (timescale of minutes for a factor of 10 change in concentration). To help mitigate this problem, an extruded PFA tube replaced the machined PFA tube, and it showed excellent response, even under very cold conditions. Presumably, extruded PFA has significantly less surface area and is less porous than machined PFA, thereby giving far superior transmission of gas phase HNO3 down the tube. Finally, it is important to note that the dimensions of the transport tube used in TRACE-P (1.9 cm in diameter and only ∼10 cm long) were kept as small as possible to help minimize the available surface contact with ambient air.
 On the basis of the C-130 flights in TOPSE and additional results from the laboratory, extruded PFA was the material of choice for the transport tube on the NASA P-3B aircraft. Therefore the turbulence and temperature of the sampled air inside the transport tube were no longer a major concern. Three additional flows were added in the transport tube. First, ∼3 cm below the top of the tube, a series of holes 1 mm in diameter (not shown in Figure 2 for clarity) encircled the circumference of the tube. These holes allowed for the addition of zero air into the top of the transport tube for examining the amount of HNO3 adsorbed onto the inlet surfaces downstream of this point. A second hole ∼5 cm down from the top of the transport tube allowed for the introduction of a 1.1 mm OD, 0.6 mm ID tube to flow isotopically labeled H15NO3 for calibration. Finally, a flow of zero air could be added to the ambient air at the end of the transport tube immediately before entering the ion source. This flow was most often used to examine for background signals of HNO3 downstream from the transport tube.
 The ambient air next passed through a removable type 316 stainless steel tube (l = 8.76 cm, ID = 1.22 cm, OD = 1.27 cm) to enter the ion source (3). This tube, although relatively short, was composed of stainless steel in order to remain at ground potential (uncharged). A notched TFE collar around the outside of the stainless steel tube kept it secured in flight and also served as the source of the reagent ion, methanesulfonic acid (MSA, CH3SO3H). MSA was physically applied to a 0.05 cm indentation in the TFE by rubbing a piece of PFA tubing dipped in MSA (J. T. Baker, 99.7%) to ensure a visual even distribution of small droplets. MSA was applied to the TFE collar before every flight, and the entire MSA assembly (TFE collar and stainless steel tube) was manually placed into the ion source. Similarly, the MSA assembly was removed and cleaned immediately after flight to prevent MSA from significantly coating the surfaces of the ion source while on the ground. Any residual MSA on other parts of the ion source were removed by heating the entire ion source (minus the MSA assembly) and transport tube while on the ground.
 A flow of ∼3 sLpm zero air (Matheson, “Zero Gas” purity, H2O < 6 ppmv, CO2 350–400 ppmv), termed “sheath flow,” was distributed around the outside of the stainless steel tube through a showerhead assembly of 0.75 mm holes and fine mesh screens to minimize turbulence. The sheath flow passed over the TFE indentation and picked up MSA vapor. About 1 cm downstream of the MSA source, the MSA-doped sheath flow passed over a 1.1 millicuries (mC) americium-241 foil that emitted alpha particles (5.5 MeV). The radioactive source was a 0.7 cm wide strip of americium-241 in the center of a 1.3 cm × 7.2 cm gold foil. The foil was positioned 0.5 cm behind the end of the stainless steel tube and held on the outside of a 2.28 cm diameter mount. Alpha particles ionized only the outermost, annular region of the reagent flow; direct ionization and associated radical production of the ambient airstream was prevented. Initially, alpha particles predominantly ionized N2 and O2 due to their high abundance (>99%) in the sheath flow, but subsequent collisions resulted in charge transfers from N2 ions to more stable O2+ and O2− species. Positive ions collided with negatively-charged optics shortly after being generated. A probable reaction sequence is noted below
Undoubtedly, other ion/molecule reactions occurred in this region of the ion source. However, because MSA was the most abundant and strongest acidic species present in the ion source, a series of charge transfers resulted in MSA ions and clusters dominating the spectra.
 MSA and other ions joined the flow of the ambient air at the end of the stainless steel tube. Electrostatic lenses pushed the newly formed ions toward the center of the ambient stream where ion-molecule reactions occurred through a drift region of 6 cm (at ambient pressure). Potentials in the drift region increased in four steps from −350 V at the Am-241 source to −50 V at the virtual iris, yielding an average electric field in the drift region of ∼50 V cm−1. The residence time of ions in the drift region under typical flow conditions was ∼50 ms, sufficient time for the ion-molecule species to achieve equilibrium. Nitric acid clustered with MSA in the following equilibrium:
Very recently, Schoon et al.  have investigated the kinetics and thermodynamics of MSA with ions. However, the above cluster reaction was not explicitly studied, and the uncertainties in the derived thermodynamic parameters for MSA anions and HNO3 were too large to make any significant conclusions from their work. By rearranging the equilibrium equation, the concentration of HNO3 can be obtained
where c is a constant calculated by adding a known amount of isotopically labeled H15NO3 to the transport tube. The isotopically labeled H15NO3 ion chemistry was assumed to be indistinguishable from the unlabeled species. Most ions and molecules react at near the collisional rate, and the differences in masses between the labeled and unlabeled species are <2%. No evidence from either field or laboratory data suggested different behaviors between HNO3 and H15NO3.
 In practice, the raw HNO3 signal is the ratio of the ion counts at mass/negative charge (m/e) 158 (HNO3 · CH3SO3−) to the ion counts at m/e 95 (CH3SO3−). Regardless of the change of ion counts for MSA, the observed ratio will remain constant for a given concentration of HNO3, all other things being equal. Although the ratio remained constant for changes in MSA anion abundance, the overall sensitivity (concentration per unit ion ratio) was determined by the total number of MSA monomer counts. Other unknown ion-molecule equilibria (formation of more stable cluster species, for example) may complicate the above reaction scheme, but the near continual use of H15NO3 to calibrate the ambient signal helped to minimize this potential problem.
 The resulting ions were electrostatically forced downstream toward the virtual iris/pinhole plates, while the remaining neutral species were pumped away through annular ports located downstream of the ion-molecule reaction region at a flow of ∼9 sLpm (balanced by ∼3 sLpm of sheath, ∼6 sLpm of ambient air). The ions were directed by electric fields through a flow of 800 cubic centimeters per minute at standard temperature and pressure (cm3 min−1 STP) nitrogen in front of a virtual iris consisting of two coaligned 0.33 mm and 0.20 mm diameter orifices on 0.25 mm thick stainless steel plates separated by 0.76 mm. The flow of dry nitrogen in front of the pinhole helped to minimize ion clustering with water. The ∼85 torr pressure of the interstitial space between the virtual iris plates was kept constant to ensure that the same amount of gas entered the vacuum system at all flight altitudes and pressures [Mauldin et al., 1998a].
 Upon passing through the virtual iris, ions expanded supersonically into a differentially pumped region of ∼10−2 torr (l = 8 cm) where a series of additional lenses applied an electric field of ∼2 V cm−1. The field strength at these relatively low pressures resulted in collisions that broke apart a fraction of weakly bound clusters into their most acidic core ions [Tanner and Eisele, 1995]. Specifically, the electric field in this region was optimized to ensure that the signals from the desired clusters of interest, MSA · HNO3 at m/e 158 (for ambient) and MSA · H15NO3 m/e 159 (for calibration), were maximized and that weaker bound species (e.g., water clusters) were broken apart. The maximum signals of the clusters occurred at identical field strengths (2 V cm−1) for both species, adding evidence that differences in reactivity between labeled and unlabeled HNO3 were insignificant. The ions were focused by three lenses into a skimmer, by four more lenses into the quadrapole mass filters, and by one back lens into a Channeltron electron multiplier. With the exception of the virtual iris and high-vacuum chamber, all of the inlet remained outside the fuselage of the aircraft at ambient pressures and near ambient temperatures.
 The choice of ion chemistry for HNO3 detection was particularly challenging for an airborne atmospheric pressure ionization scheme. Mauldin et al. [1998b] previously described a ground-based CIMS technique using bisulfate (HSO4−) as a reagent ion. Unfortunately, difficulties in maintaining a constant source of reagent ion HSO4− were encountered. The gas phase concentration of sulfuric acid (H2SO4), the reagent ion precursor species, was difficult to control due to the very low vapor pressure of H2SO4 and its efficient ability to cluster with itself. Therefore HSO4− reagent ion chemistry was unreliable for use on an airborne platform where temperatures and relative humidities can change rapidly. However, HSO4− reagent ion chemistry did show excellent sensitivity and selectivity for detection of HNO3, and it was desired to keep these characteristics as much as possible. A compound of similar chemical structure and gas phase acidity was sought but also one with a significantly higher vapor pressure. To this end, MSA, a derivative of sulfuric acid, was used for the first time as a reagent ion to detect HNO3. In comparison to H2SO4, MSA is slightly less acidic (Gibbs free energy (ΔGacid)(MSA) = 1318 ± 8.4 kilojoules (kJ) mol−1; ΔGacid(H2SO4) = 1251 ± 13 kJ mol−1) [Koppel et al., 1994], and its vapor pressure is several orders of magnitude higher [Ayers et al., 1980; Tang and Munkelwitz, 1991]. MSA is comparable in acidity to HNO3 (ΔGacid(HNO3) = 1330 ± 8.4 kJ mol−1) [Koppel et al., 1994]. Although MSA proved more reliable and easier to use than H2SO4, as will be described later, controlling the concentration of MSA in the ion source under flight conditions remained problematic at times.
3.1. Laboratory Studies
 A series of laboratory experiments were performed to examine the sensitivity of the ion chemistry to potential interferences expected in the atmosphere. Specifically, the nitrogen oxides NO, NO2, NO3, and N2O5 were tested under dry and wet conditions (relative humidity with respect to water, RHw, 0.001–100%) at mixing ratios far exceeding those expected in the troposphere. Experiments were conducted at two different mixing ratios of H15NO3 (80 and 800 pptv) to examine if MSA/H15NO3 ion chemistry was perturbed by the addition of these species. No unlabeled HNO3 was added to the system, but the signal at m/e 158 was monitored during the experiments to identify if any unknown ion chemistry would result in a signal at this mass.
 NO was prepared by filling a 5 L bulb with 10 torr of a 0.5% NO/N2 cylinder mixture and 960 torr of N2. A flow of 2 cm3 min−1 STP from the bulb was passed through ∼1 m of 3.2 mm OD, 1.6 mm ID nylon tubing to help remove any residual impurities of HNO3 in the gas source (nylon is known to be an efficient scavenger of gas phase HNO3). The flow was then diluted into 8000 cm3 min−1 STP of zero air over the ion source resulting in a mixing ratio of 13 ppbv NO. No change in the ratio of the H15NO3/MSA (m/e 159) ratio was observed from the addition of the NO, nor was any increase observed for the HNO3/MSA (m/e 158) ratio. These results indicate that NO neither produced HNO3 in the ion source nor did it alter the HNO3/MSA ion chemistry. Tests were done under both dry (0.001% RH) and wet (100% RH) conditions, and no noticeable change was observed in either case.
 N2O5 was synthesized by the method of Davidson et al. . The N2O5 was stored in dry ice and kept in the dark when not in use. The N2O5 was differentially pumped for several minutes prior to use at temperatures as warm as 243 K. A flow of 93 cm3 min−1 STP of N2 was passed over the N2O5 which was kept in a dry ice/ethanol bath at 205 K. The vapor pressure of N2O5 at 205 K was ∼6 mtorr [McDaniel et al., 1988]. It was assumed that the flow of nitrogen was saturated with the vapor pressure of N2O5 based upon the work of Cantrell et al. . The N2O5−doped nitrogen flow passed through nylon tubing in order to ensure that any heterogeneous decomposition of N2O5 into HNO3 would remain on the walls and not in the gas phase. The temperature of the gas handling line was kept at 298 K, and therefore ∼1% thermal decomposition of N2O5 into NO3 and NO2 occurred [Cantrell et al., 1988]. The residence time of N2O5 in the lines was ∼30 s. The flow of N2O5-doped N2 was diluted into 6000 cm3 min−1 STP of zero air. On the basis of the vapor pressure and the flow rates, the concentration of N2O5 was 120 ppbv, while NO2 and NO3 were 1.2 ppbv. These experiments were also conducted under wet and dry conditions, and no systematic differences were noted in either the ratio of the isotopically labeled H15NO3 · MSA or ambient HNO3 · MSA signals.
 Although NO, NO2, NO3, and N2O5 showed no observable interference in the MSA · HNO3 (labeled and unlabeled) signals under either wet or dry conditions, a significant change in sensitivity was observed with gas phase water. The vapor pressure of MSA is impacted by the relative humidity of water over the surface. Figure 3 shows the MSA ion counts for the apparent monomer, dimer, and trimer populations, and the total counts of MSA in these clusters versus relative humidity at 23°C. At low relative humidities a high percentage of MSA is tied up in the trimer (and likely higher clusters), with relatively little MSA in the monomer. As the humidity increases, the number of trimers decreases, while the number of dimers significantly increases. At RH = 30%, the dimer population decreases, and at RH = 65% the monomer becomes the dominant cluster. Overall, as the relative humidity increases, the vapor pressure of MSA decreases, and the cluster distribution moves toward the lower clusters. Therefore one problem of MSA reagent ion chemistry is the changing sensitivity of the instrument as a function of ambient water vapor concentration. A more reliable way to introduce MSA into the ion source and control its cluster distribution is desired, although the use of continual calibrations by the addition of H15NO3 helped to alleviate this problem.
3.2. Field Studies
 The NCAR CIMS HNO3 instrument was flown on the NASA P-3B aircraft as part of the NASA TRACE-P which took place during February–April 2001 off the coast of eastern Asia (based in Hong Kong, China, and Yokota Air Base, Japan) [Jacob et al., 2003]. No data were collected during the transit flights 4–8 due to an ion source “flooded” with MSA vapor, which resulted in little or no ambient and calibration signals due to most MSA being tied up at higher clusters (>3). Cleaning and heating of the ion source before flight 9 remedied this problem, but large eddies in the transport tube removed both the isotopically labeled calibration gas as well as the background zero air in the transport tube, preventing reliable calibrations and background tests. A PFA cap (shown in Figure 2) was placed above transport tube before flight 10, and in-flight tests indicated that it prevented large eddies from removing calibration gas and zero air (background) air flows. Therefore data were archived from flights 10 onward, initially at 20 s resolution (flights 10 and 11) and at 5 s resolution thereafter. The measurement scheme generally involved a 20 s cycle as follows: 2 s for the MSA monomer at m/e 95, 1 s for MSA dimer at m/e 191, 3 s for the ambient signal (m/e 158), 3 s for the calibration signal (m/e 159), 3 s for ambient, 1 s for an electronic noise background at m/e 20 (polarity of the quadrupoles was reversed temporarily), 3 s for the ambient signal, 2 s for a partial mass scan at 0.5 s per atomic mass unit (amu) from 20 to 160 amu, and another 3 s for the ambient measurement. In this way, HNO3 was measured on average once every 5 s. The calibration signal at m/e 159 was only measured once every 20 s because its concentration was well controlled. Likewise, the concentration of reagent ion was fairly constant on a 20 s timescale as well. On occasion, the time of the partial mass scan was modified or eliminated during portions of flights to measure other masses of interest.
 The duct and transport tube were not actively heated in flight, and therefore sampled air was only warmed by adiabatic compression from the slowing of the air inside the inlet and from conduction of heat from inside to outside of the instrument. In combination, temperature readings around the transport tube suggest a warming of <10 K above ambient temperatures. Occasionally, under very cold conditions (T < −15°C), it was necessary to heat the ion source region to raise the vapor pressure of MSA to ensure adequate amounts of reagent ion signal (>500 cts s−1). Because of the very short transit time in the ion source (50 ms), it appears unlikely that the ambient air warmed sufficiently for desorption of particulate nitrates.
 For warm (T > 10°C) and very dry (Td < 0°C) conditions, sufficiently high vapor pressures of MSA resulted in clustering of MSA into the dimer, which does not readily cluster with HNO3. Therefore it was necessary to humidify the sheath flow to increase the population of the monomer MSA ion. A fraction (0–1.4 sLpm) of ∼3 sLpm sheath flow was bubbled through a 300 ml ∼0.01 M aqueous sodium hydroxide (NaOH) solution to tie up any residual NO3− in the water (Aldrich, HPLC grade). The humidified flow was passed into an empty 500 ml container to ensure that any small droplets would settle or collide with the container walls before entering the ion source. Finally, the sheath flow passed through a 0.9 micron nylon/sodium bicarbonate (NaHCO3) filter to remove any small (micron-sized) particles. The use of a humidified sheath flow was incrementally adjusted to keep the signal of the MSA dimer to approximately one third of the monomer. At colder temperatures (<10°C ambient), a low dimer signal (due to lower vapor pressure of MSA) precluded the need for a humidified sheath flow.
 Because the monomer MSA ion signal changed with ambient conditions, the sensitivity of the HNO3 signal (ion ratio m/e 158: m/e 95) varied with ambient conditions. In the laboratory, sensitivities of the MSA · HNO3 cluster were routinely 1–3 counts pptv−1 s−1. In contrast, in-flight operation of the instrument decreased the sensitivity by an order of magnitude relative to the laboratory experiments due to decreased gas flow into the vacuum system in order to accommodate the other two channels (OH/MSA/H2SO4, HO2/RO2) and due to nonoptimal pressures in the collision chamber (again, a compromise of the multichannel shared vacuum housing). Although in-flight sensitivities of the cluster at times approached 1 count pptv−1 s−1, more typical values were ∼0.1–0.4 counts pptv−1 s−1 during TRACE-P. Therefore it was necessary to continuously calibrate with a flow of isotopically labeled H15NO3 at the top of the transport tube.
 Background estimates of HNO3 sticking to instrument surfaces could be experimentally-determined by three different methods. The first method involved adding a large flow of zero air (not shown in Figure 2 for clarity) to the very top of the transport tube, just upstream from the addition of the calibration gas. Although this method worked reasonably well on the ground, in-flight tests showed little effect even at zero air flows several sLpm greater than the flow down the transport tube. Because the transport tube was perpendicular to the main flow in the large inlet, significant eddies formed in the upper region of the transport tube, and it is thought that large (though unquantified) amounts of the zero air were turbulently ejected from the inlet. Thus this method was not used to estimate backgrounds while in flight.
 A second way to measure the background involved the removal of the calibration flow at the top of the upper transport tube. This method estimated the amount of HNO3 adsorbed to the extruded PFA transport tube. Simply turning off the flow of the calibration gas was insufficient to test for a background as any small flows into the transport tube would still emit H15NO3. Therefore it was necessary to reverse the flow of the calibration gas by pumping on the calibration tube ∼1 m downstream of the end of the calibration tube. Figure 4 shows a representative example of the response of ∼515 pptv of the calibration signal upon pumping on the flow at ∼0715:40 UT on flight 14 at 4359 m (577 millibars (mb), −5°C). The data are presented in raw ion signal because the sensitivity (but not the concentration) was steadily changing as shown by the general increase in ion signal before the calibration gas was removed. In the 20 s time resolution of the calibration measurement, the signal decreased to a value indistinguishable from the background level. The background level remained constant for the remaining duration of the test, indicating a constant (over the 2 min timescale of the test) desorption source of HNO3 off walls or a background signal at this mass unrelated to HNO3 adsorption. As will be discussed later, a significant portion of this background can be attributed to factors apparently unrelated to HNO3 adsorption (for the case of Figure 4, 75% of the background, or all but 18 pptv, can be attributed to other factors). Overall, the extruded PFA transport tube showed minimal wall/surface effects from the removal of calibration H15NO3 gas from the upper transport tube. Because the inside surfaces of the long, thin tubing that carried the isotopically labeled H15NO3 flow had to reequilibrate to a new concentration, relatively long periods of time (2–20 min) were needed to ensure the isotopic signal was stable after the test. Additionally, pumping on the calibration gas line pulled ambient HNO3 into the calibration line, complicating the ambient measurement as well as contaminating the calibration gas line. Thus this background measurement was conducted only about once per flight, mainly to ensure that the extruded PFA transport tube remained a minimal sink for gas phase HNO3.
 A third method to estimate background signals involved overfilling the transport tube with a flow of zero air just above the ion source. No change in ion chemistry or sensitivity was noted in zero air versus ambient air. This method effectively examined the adsorption of HNO3 to the metal pieces around and downstream of the ion source, although it could not be used to estimate the amount of HNO3 adsorbed to the extruded PFA upper transport tube. Figure 5 shows the responses of the ambient and calibration signals as zero air was added in front of the ion source during a 52 min segment of flight 24 (Dryden to Wallops transit) at an altitude of 7320 m (368 mb). The zero point on the abscissa is the equivalence point where the flow of gas being pulled into the ion source is balanced by the flow of zero air above the ion source. Data <0 are considered “underfilling” the transport tube, while data >0 are considered “overfilling” the transport tube. As zero air is added in front of the ion source, a smaller flow of ambient air is sampled, and the mean flow velocity in the transport tube decreases. Above the equivalence point, as the inlet is overfilled, the flow of zero air through the transport tube is away from the ion source, and therefore the calibration gas gets pushed outward. Similarly, no ambient air is drawn into the transport tube. Background levels of both the ambient and calibration gas can be recorded under these conditions.
 In the absence of turbulence the expected response of the calibration signal to the zero airflow should remain unchanged until the equivalence point. That is, the concentration of the calibration gas is independent of whether it is diluted by ambient air or by zero air. Under conditions of overfilling the inlet, the calibration gas should be at background levels. For the case of the ambient signal, however, a flow of zero air above the ion source decreases the amount of ambient air pulled into transport tube. Therefore the response of the ambient signal should linearly depend on the overfill flow. The expected curves for each case are denoted in Figure 5 by dotted lines for the calibration and ambient signals.
 In situ field tests indicated that the responses of the signals differed from the expected behavior. Specifically, for both gases, a sigmoid curve is observed, initially flat, showing no response as zero air is added, and then rapidly decreasing when approaching within 2 sLpm of the equivalence point. No significant decrease in the signal was observed from slightly beyond the equivalence point to over 3 sLpm overfill. No physical basis for fitting with a Sigmoid curve is claimed, except that the shape of the curve represents the data well. The calibration data suggests that, as the zero air flow increased and the sampled airflow (and speed) in the transport tube decreased, turbulent eddies partially removed H15NO3 gas from the upper part of the transport tube. The data also show that overfilling the transport tube by at least 2 sLpm greater than the equivalence flow was sufficient to record background conditions of calibration gas.
 In comparison to the isotopically labeled H15NO3, the response of the ambient signal shows less curvature. Presumably, not all of the zero air was drawn into the ion source when underfilling, but some small fraction was “pulled out” of the upper opening of the transport tube. Nonetheless, the amounts were significantly less than for the case of the calibration signal. In addition, the ambient background was nearly constant at flows >1 sLpm overfilling. Therefore all ambient backgrounds measured by this technique were also conducted at overfills at least 2 sLpm greater than equivalence point. Finally, we caution that the ambient concentration of HNO3 may have changed during the in-flight test shown in Figure 5, and therefore the ambient results are complicated by this factor. However, other chemical species such as NOy, NO, O3, and CO indicated a relatively homogeneous air mass for the duration of this test.
 Similar tests on the signal response as a function of overfill flow were conducted on other flights during TRACE-P. Data were qualitatively consistent with the results shown here, whether at low altitude (152 m) or at high altitude (7300 m). A slight dependence on airspeed was observed with higher speeds having greater turbulence in the transport tube. For example, the data in Figure 5 were obtained at 170 m s−1, and signal was halved at a flow of xhalf = −1.68 ± 0.03 sLpm. In contrast, at a slower airspeed of 153 m s−1, the calibration signal was halved at xhalf = −0.97 ± 0.05 sLpm at 674 mb. Although pressure may indeed affect the degree of turbulence in the upper transport tube, the effect appeared to be caused by airspeed. A comparison of similar pressures (554, 516 mb) with different airspeeds (321, 308 knots, respectively) indicated slightly more turbulence (xhalf constants of −1.24 ± 0.13 versus 1.05 ± 0.15) with the higher airspeeds.
Figure 6a shows a representative background when overfilling the inlet with zero-air on flight 14 (Okinawa to Yokota transit) at an altitude of 152 m. Upon overfilling the inlet, both the ambient and calibration signals decreased to the same respective levels of 5–10 pptv. In addition, the ambient HNO3 data, with a temporal resolution of once every 5 s, show that the background signal was achieved within the 5 s measurement time, and no significant decrease was observed for the duration of the overfill. Likewise, the signal responded within the 5 s measurement window when the overfilling was stopped. These results indicate that the adsorption background was a small percentage of the overall signal, and that HNO3 adsorbed to the metal surfaces of the ion source did not significantly affect the measurement.
Figure 6b shows a background of HNO3 at much lower concentrations of ∼80 pptv taken at 5273 m (509 mb, −9°C) on flight 23 (Kona to Dryden transit). Like the previous example, the background mixing ratios for the ambient and calibration signals were 5–10 pptv. Despite the relatively low concentration of HNO3, the response of the instrument to the background measurement remained rapid, on the timescale of the temporal resolution (5 s). Background tests similar to the ones shown in Figure 6 were usually conducted at least once per level flight leg or more frequently when significant changes of HNO3 were encountered.
 Although the timescale to achieve a constant background level was relatively rapid, constant, nonzero ion signals were often observed. For example, the raw ion ratio signal plotted in Figure 4 shows a background signal of ∼14%. The largest sources of background signals at masses 158 and 159 originated from isotopic contributions of significant peaks near m/e 156 and m/e 157. Figure 7 shows a mass spectrum averaged over all of flight 22 (Midway to Kona). The mean ambient mixing ratio for this flight was 79 pptv while the mean calibration signal was 305 pptv. Although the MSA monomer (848 counts s−1) and dimer (671 counts s−1) were the largest peaks in the spectrum, the m/e 156 (77 counts s−1) and 157 (216 counts s−1) peaks were comparable to the ambient (m/e158, 57 counts s−1) or calibration (m/e159, 132 counts s−1) peaks. The 156 and 157 peaks correlated with the availability of MSA reagent ions, giving evidence that these species contained MSA. The remaining part of the cluster most likely resided at m/e 60 and m/e 61, CO3− and HCO3−, and indeed these species were abundant in Figure 7 and in mass spectra taken throughout the experiment. Furthermore, CO3− is expected to react with MSA at the collision rate and a number of products including m/e 156 have been experimentally observed [Schoon et al., 2002]. Therefore it seems reasonable that masses 156 and 157 are [MSA · CO3]− and [MSA · HCO3]−, respectively. Ultimately, tandem mass spectrometry would be needed to help identify these compounds unambiguously.
 At concentrations of HNO3 < 100 pptv, the peak signals at m/e 156 and 157 were often larger (factor of 2–5) than the ambient signal. Although the resolution of the mass spectrometer near these peaks was such that only 1% of adjacent ions were measured (m/Δm = 320), the isotopic abundances of sulfur (33S = 0.76%, 34S = 4.22%), carbon (13C = 1.11%), oxygen (17O = 0.037%, 18O = 0.204%), and nitrogen (15N = 0.37%) resulted in significant mass counts at the ambient and calibration signals (m/e = 158, 159). The signals at m/e 156 and m/e 157 were measured every 20 s during portions of several flights and no obvious dependence on sheath flow, pinhole flow, or zero air was observed. In addition, on all flights these species were measured at least once every 10–15 min during partial but continual mass scans as part of the measurement scheme. No apparent correlation was observed with any factor besides increasing or decreasing with MSA monomer abundance. It remains unclear how or why these species would cluster so readily with MSA or be so abundant in the ion source, but they most likely formed before MSA ions were introduced.
 One possible source of carbonate ions could be the nylon/NaHCO3 filter for the humidified sheath air. Another possibility includes the use of various nylon filters coated with NaHCO3 in the installation and early stages of TRACE-P. Replacement of new tubing leading to the ion source occurred a number of times during TRACE-P, but no decrease in the signal was noted afterward. Furthermore, the ion source was cleaned and sonicated a number of times in baths of water (HPLC grade, Aldrich) and ethanol (>99.5% purity, Fisher), but no reduction the peaks at 156 or 157 were noted afterward. It is possible that permanent segments (30–60 cm) of tubing leading from the inside to the outside of the fuselage were significantly contaminated with bicarbonate.
 Nonetheless, the isotopic and resolution effects were accounted for in the data set. For the m/e 158 signal, background contributions derived from the following sources: 5.4% from m/e 156 (4.2% from 34S, 1.2% from six 18O atoms at 0.2% each), 4.24% from m/e 157 (2.2% from two 13C atoms at 1.1% each, 1% from resolution, 0.8% from 33S, 0.2% from six atoms of 17O at 0.04% each), and 1.4% from m/e 159 (0.4% from impurities in H15NO3, 1% resolution). In a similar fashion, a linear set of equations was solved to determine the background signals at 158 and 159 arising from the effects described above
where v, w, x, y, and z correspond to the modified signals (effects of isotopes removed, adjacent masses removed, etc.) at masses 156, 157, 158,159, and 160 signals, respectively. Generally, the size of this correction was small (<10%) compared to the ambient signal at mixing ratios >100 pptv HNO3 but became increasingly significant at lower mixing ratios. Ultimately, the correction for the relatively large peaks at m/e 156 and m/e 157 accounted for most of the background in the lower concentration data and was the single largest source of error in the measurements at low (<100 pptv) concentrations.
 Additional sources of background include electronic noise (<1 count s−1) in the Channeltron electron multiplier. Noise was determined by the measuring the signal at m/e 20 when the polarity of the quadrupole lenses was reversed. Any production of HNO3 from the high-energy ionization near the americium-241 would also result in a constant background. Overall, however, these sources are estimated to account for less than a few percent of the background.
 Calibrations of the system occurred by the addition of the isotopically labeled H15NO3 in the transport tube. The H15NO3 source contained 99.6% enrichment of 15N nitric acid and was enclosed in a permeation tube (Vici). The tube was initially reported with an emission rate of 57 ng min−1, however permeation tubes are notorious for different emission rates under different conditions [Talbot et al., 1997; Ryerson et al., 1999]. Therefore the permeation tube was held in a heated cavity with a critical orifice of 7 cm of 0.06 mm ID PFA tubing. The permeation tube was always regulated at a constant pressure of 35 psi N2 during the mission (except when exchanging N2 cylinders). An inert flowmeter upstream of the permeation tube measured the flow (∼100 cm3 min−1 STP) of N2 over the permeation tube. In this way, N2 was continually flowing over the permeation source (herein called the permeation flow) at a constant pressure even when the aircraft had no power. The permation flow passed through a manifold where a portion (0–100 cm3 min−1 STP) of the flow could be removed by pumping (measured by a 100 cm3 min−1 STP flow controller), while the remaining amount flowed through a 10 cm3 min−1 STP Teflon™-coated flowmeter (MKS) and joined a carrier flow of ∼100 cm3 min−1 STP N2. The Teflon-coated 10 cm3 min−1 STP flow meter accurately measured low flows of the permeation flow into the carrier flow. In this way, both large and small dilutions of the permeation flow could be achieved. The resulting carrier flow passed through ∼40 cm of 1.1 mm OD, 0.6 mm ID PFA tubing. This tubing was enclosed by 1.3 mm ID stainless steel tubing which was resistively heated to 45°C from just inside the fuselage to the outside edge of the transport tube.
 Temperature control of the permeation tube consisted of an aluminum, thermofoil backing, resistive heater (Minco) and a temperature controller (Watlow, model 241). The permeation tube was kept at a constant temperature of 40°C except when the airplane was powered down. Before flight, the temperature of the permeation cell was raised to 40°C at least 2 hours before measurements commenced, and the carrier/calibration tube was heated at that time as well. Ground-based tests indicated that one hour was sufficient for the tubing walls and emission rate to reequilibrate at 40°C.
 The mass of the permeation cell was measured for 8 months after the end of the deployment, and an average emission rate of 34 ± 1 ng min−1 was obtained from the data shown in Figure 8. The permeation cell was kept at the same pressure and temperature as in the field, except for the 45 days when the instrument was being transported back to the laboratory. No mass measurements were made in the field due to a lack of an accurate and consistent scale and due time spent on the diagnoses of other problems. Because some slight curvature does exist in the data, we estimate that the permeation cell may have emitted as high as 41 ng min−1 during the timeframe of TRACE-P. Although no mass measurements were made in the field, the output of the permeation cell was measured twice in the field by the University of New Hampshire mist chamber/ion chromatography instrument on the NASA DC-8 airplane. The University of New Hampshire (UNH) measurements yielded an average of 29 ± 1 ng min−1 for three measurements before P-3B flight 17, and 32 ± 2 ng min−1 for two measurements before P-3B flight 23. The mass data and the two ion chromatography measurements provide strong evidence that the permeation cell emitted H15NO3 at a rate within 20% of 34 ng min−1 during the entire TRACE-P campaign. The overall accuracy of the permeation cell calibration was the largest source of error for measurements above 200 pptv.
 Overall error (±2σ) of the measurement including systematic and random errors were the limit of detection (10 pptv) plus the following: ±30% pptv at mixing ratios <100 pptv, ±25% from 100–200 pptv, and ±20% for mixing ratios >200 pptv. At high mixing ratios the dominant source of error derives from the uncertainty in the emission of the calibration gas in the permeation cell. At low mixing ratios the uncertainty in the background correction is the largest source of error. Flow controllers and flow meters were checked multiple times throughout the mission for zero readings and for absolute accuracy against a soap bubble meter (Gillibrator). Uncertainties in these parameters were small (less than a few percent).
 Because the calibration gas was almost always flowing during the flight measurements, multipoint calibrations were conducted in flight to examine the linearity and response of the measured H15NO3 signal versus H15NO3 concentration. Typically, multipoint calibrations were conducted at least once per flight for flights 16–24. To conduct these experiments, a known amount of the permeation flow was removed by the manifold assembly before joining the carrier flow which transported the H15NO3 to the inlet. A representative example of an in-flight calibration is shown in Figure 9 for a wide range of altitudes and mixing ratios (50–2500 pptv) during flight 16 (Yokota local 2). Despite the much different environments of the data (ranging from 5182 m to the marine boundary layer), a linear response is clearly noted in the data in Figure 9, suggesting that the continuous single-point calibrations throughout the rest of the flight were valid. Although the absolute calibration factor and sensitivity did change from flight to flight and within a flight (mainly due to the impact of MSA clustering changing the availability of MSA monomer as noted previously), the overall sensitivity was generally around ∼0.1 counts pptv s−1.
 Three brief and informal intercomparisons between instruments onboard the NASA P-3B and NASA DC-8 were conducted during TRACE-P. Typical distances between the airplanes ranged from 0.2–1.0 km with a vertical separation of <100 m. For more information on the intercomparisons, refer to Eisele et al. . During the first intercomparison on the transit to Hong Kong, no data were obtained due to a lack of reagent ion signal and calibration signal (MSA had coated the ion source and reagent monomer MSA signals were extremely small). Data were recorded on the second (P-3B flight 16, DC flight 10) and third (P3-flight 23, DC flight 20) intercomparisons, and the results were compared to the HNO3 measurements from the mist chamber/ion chromatograph instrument from the University of New Hampshire onboard the DC-8 aircraft (UNH). Details of the UNH instrument are described elsewhere [Talbot et al., 1997, 1999, 2003]. It should be noted that each instrument detected HNO3 with a different technique (chemical ionization versus ion chromatography), vastly different inlets, and different permeation standards (though as noted above, within 20% of one another).
Figure 10a shows the data collected on the second intercomparison, lasting ∼24 min at 5182 m. No data were collected in the final 3 min of the intercomparison from NCAR CIMS due to turbulence problems arising from an overfill flow near the equivalence point. The light blue solid line in Figure 10a shows the NCAR CIMS HNO3 data collected every 5 s. The red line in Figure 10a shows the UNH measurements. Finally, the dark blue line in Figure 10a shows the NCAR CIMS measurements averaged within the UNH sampling window. Reasonable agreement is observed between measurements with NCAR CIMS reporting an average of 158 ± 32 pptv (N = 224, 1 σ) in 5 s measurements compared to 185 ± 28 pptv (N = 9) for the UNH samples of 120–180 s. The averages were within the error bars of ±25% for NCAR and ±15% for UNH. On a point by point basis, CIMS was lower than UNH on 6 of 9 points, with a mean deviation of UNH greater than NCAR by 23 pptv and a standard deviation of the mean of 29 pptv. It is unclear why NCAR had lower measurements than UNH, and potential sources of disagreements are discussed later.
Figure 10b shows the results of the third intercomparison over the Pacific Ocean just east of Hawaii. This intercomparison consisted of 20 minutes at 5180 m, followed by a descent to 150 m, and 20 minutes at 150 m. A problem with electrical noise in the OH channel of the four channel system prevented data collection for a 7 min period on the descent for NCAR CIMS. Both techniques measured low values of HNO3 during the high-altitude portion. CIMS measured on average 12 pptv higher than UNH during the high-altitude portion, with the exception of a spike measured only by UNH around 1824 UT. On the descent, both techniques measured a local maximum of 190 pptv HNO3 around 3050 m and subsequent decrease to <100 pptv near the end of the descent. A gradual rise of HNO3 was measured by both instruments during the boundary layer segment. A significant deviation between the data sets did appear near the end of the boundary layer run with CIMS HNO3 measuring 72 pptv higher than the values obtained by UNH. No other tracers identified significant deviations, and therefore a change in air mass is not likely a valid reason for the discrepancy. Overall, the mean deviation between the measurements indicates that CIMS was 15 pptv higher than UNH throughout the third intercomparison.
 One potential reason for the discrepancies between the two measurements could be the sampling of nitrate-containing particles (<2.5 μm aerodynamic diameter) by UNH’s instrument. The NCAR CIMS inlet was not expected to directly sample particles because only gas phase ions (not charged particles) were electrostatically directed into the vacuum system. Sampling of nitrate-containing particles would result in higher measurements of gas phase HNO3 relative to the true amount. Particulate nitrate smaller than 1.3 μm diameter was measured on board the P-3B [Ma et al., 2003; Weber et al., 2001]. The second intercomparison showed steadily increasing levels of particulate nitrate from 22 to 44 pptv, and UNH measurements indeed averaged 27 pptv higher than NCAR CIMS. Particulate nitrate on the third intercomparison were generally below 20 pptv, and yet NCAR CIMS reported higher measurements than UNH. Therefore the differences between the HNO3 measurements could not be explained by the amount of particulate nitrate measured onboard the P-3B. A more thorough analysis on the size distributions of nitrate-containing aerosol particles is ultimately needed, however.
Figure 11 shows a plot of the NCAR measurements versus UNH measurements for all of the data points in the second and third intercomparisons. The associated error bars for the measurements are determined using the stated errors mentioned previously for NCAR CIMS and the uncertainties of UNH stated in the TRACE-P data archive: ±30% for <20 pptv, ±25% for 20–25 pptv, ±20% for 25–100 pptv, and ±15% for >100 pptv. The uncertainties of each measurement technique for the data in Figure 11 differ from the more general ones used by Eisele et al. . The error limits shown in Figure 11 are the square root of the sum of the errors squared for each measurement. The slope of the data is 0.94 ± 0.18 (2σ), showing general agreement between techniques, although an offset of 19 ± 14 pptv (2σ) indicates that potential disagreements may be related to background issues. Overall, 36 out of the 40 data points lie within the expected error bars of each set of measurements. Clearly, the two techniques appear to be in good agreement, although issues such as background determinations and aerosol particle sampling most likely need to be better quantified, especially at low mixing ratios.
 Finally, the response of the inlet and instrument to rapid changes in ambient HNO3 concentration is demonstrated. Figure 12a shows the response of the instrument to ship plumes at 150 m above the South China Sea on flight 14 (Okinawa to Yokota transit). A uniform atmospheric concentration of ∼450 pptv prevailed initially, although individual spikes as high as 1800 pptv occurred during the 5 s measurement resolution time for this flight. A rise in background was observed at the very end of the ship plumes, but it is unclear if the rise in background is truly indicative of adsorption problems. Although NOy remained constant, PAN measurements also showed an increase from 1386 pptv before to 2125 pptv after encountering the ship plumes. In addition, multiple decreases and increases occurred to near the initial background level, showing that the ability of the inlet to resolve rapid HNO3 changes and also return to baseline values. If significant amounts of HNO3 had been adsorbed onto the inlet (e.g., if the “true” value of the peaks were 5 ppbv), one would expect to observe a gradual desorption of HNO3 off the inlet surfaces. Instead, the values between the spikes are generally near the initial background level of ∼450 pptv until three minutes into the ship plumes, followed by a rise only at the very end. The corresponding changes for NOy (1 s resolution) as measured by the University of Tokyo are also shown [Kondo et al., 1997; Koike et al., 2000; Miyazaki et al., 2003]. Tight correlation is observed, indicating very good resolution of the inlet. No attempts have been made to resolve the ∼3 s offset between the measurements.
Figure 12b shows a vertical profile of HNO3 during descent into the marine boundary layer where highly polluted air was encountered. Upon entering the polluted layer, ambient HNO3 increases from around 360 to 1600 pptv in 15 s. Although the NOy instrument was not reporting data for the transition, a pollution tracer of CO is shown for reference [Palmer et al., 2003]. The rapid changes presented in Figure 12, in combination with in-flight tests of the calibration gas, suggest that HNO3 adsorption onto inlet surfaces was not a significant problem on the timescale of the measurement and therefore generally did not interfere with the measurement.
 Another method of estimating the time response of the instrument is through power spectral analysis as described by Murphy . Ideally, uninterrupted data segments well above the limit of detection should be analyzed individually, and the corresponding spectra from all segments should be averaged. Because of relatively frequent background checks and instrumentation problems, long and uninterrupted data segments were not available for analysis. Instead, the power spectrum for all data taken at 5 s (0.2 Hz) resolution (flights 12–24) with mixing ratios >100 pptv (N = 33,909) was analyzed. An estimate of the power spectral density of the data was determined by using a nonparametric, periodogram method with Matlab software (version 6.5.0, release 13). Instrumental white noise has a slope of zero, whereas passive sampling of atmospheric species is expected to show a slope −5/3. Sampling from aircraft platforms may alter the expected slopes, and in these cases, a changing but smooth slope is expected to be observed [Ryerson et al., 1999]. The transition frequency between instrumental noise (zero slope) and real atmospheric variability (nonzero slope) is generally regarded as the instrumental time response, and low pass filtering of the data at this frequency yields the maximum amount of information from the data set [Murphy, 1989]. Power spectrum analysis of the HNO3 data showed a near zero slope at frequencies higher than 0.06 Hz, and a definite negative slope at lower frequencies. Therefore a transition frequency around 0.06 Hz indicates a general time response of the instrument of 15 s, a timescale roughly consistent with the results of the background tests and observations of changes in ambient HNO3 concentrations. A more detailed analysis of the power spectrum is beyond the scope of this research, but it provides another supporting piece of evidence for the fast time response of the instrument.
4. Summary and Future Work
 Nitric acid was detected using a unique choice of inlet design and materials and selective ion chemical ionization mass spectrometry using methanesulfonic acid as a reagent ion. The instrument measured HNO3 every 5 s with a limit of detection of 10 pptv. The inlet was not actively heated to prevent possible desorption of particulate nitrates into gas phase HNO3. The instrument was calibrated continuously using isotipcally labeled H15NO3, and the inlet demonstrated very few problems of surface adsorption. Intercomparisons with a more established HNO3 measurement technique by UNH were promising.
 Future areas of improvement with this instrument include identifying a more reliable way to introduce methanesulfonic acid or selecting a more stable reagent ion. Under low concentrations of reagent ion, the sensitivity of the measurement decreased significantly. One possibility to better control MSA reagent concentration includes using a reduced pressure ion source with extruded PFA surfaces. Although MSA worked favorably for this campaign, future research will examine ways to control the addition of MSA to the ion source (as opposed to manually placing MSA droplets in the ion source). Furthermore, the presence of peaks at m/e 156 and m/e 157 can significantly complicate background determinations, and solving this problem will enhance the sensitivity of the technique as well as reducing uncertainties in background estimates. Nonetheless, the techniques described here are a start toward quantifying and ultimately better understanding the chemistry of HNO3 in the troposphere.
 MAZ gratefully acknowledges support from an NCAR ASP Postdoctoral Fellowship. The authors thank the following people: J. Dibb, E. Scheuer, and R. Talbot for invaluable assistance and advice in the calibration of the permeation standard and for great cooperation in the intercomparisons; D. Hanson for suggesting MSA ion chemistry; J. Orlando and G. Tyndall for the use of their N2O5; J. Fox and the ATD instrument shop for inlet design and construction; J. Vanderpol for shop work; T. Ryerson for permeation cell oven plans; R. Hendershot and W. Bradley for electronics assistance; and K. Hellyer for editing. The assistance of the NASA TRACE-P Science Team and aircraft support personnel at NASA Wallops Flight Facility are greatly appreciated. Finally, we thank J. Smith, I. Faloona, and two anonymous reviewers for their helpful and constructive feedback on the manuscript. This research was supported by the NASA Global Tropospheric Experiment (GTE) program. The National Center for Atmospheric Research is operated by the University Corporation for Atmospheric Research under sponsorship of the National Science Foundation.