The first calibrated in situ measurements of hydrogen cyanide (HCN) concentrations in the upper troposphere/lower stratosphere were made during the March 2000 deployment of the SAGE III—Ozone Loss and Validation Experiment (SOLVE) campaign. The measurements were made from the NASA DC-8 aircraft flying at cruise altitudes between 10 and 12.5 km in the northern polar region. The average HCN volume mixing ratio was found to be 280 ± 48 (1σ) pptv. No concentration gradient was observed between tropospheric and stratospheric air, and no HCN emissions from the Mount Hekla volcano were found. A very slight increase in mixing ratio with altitude was detected, although it is within the uncertainty in the data. The altitude dependence agrees with recent model calculations. The measurements agree with the most recent concentrations derived from total column densities but are higher than earlier measurements.
 In contrast, few studies have reported in situ measurements of HCN. The only works published in the upper troposphere/lower stratosphere are the chemical ionization mass spectrometer (CIMS) measurements of Arnold and coworkers from the Max Planck Institute in Heidelberg [Schneider et al., 1997; Spreng and Arnold, 1994]. They found mixing ratios in the lower stratosphere to be about 164 pptv, a value consistent with the older column density measurements but lower than the most recent ones. A very small increase with altitude was found, possibly due to conversion of CH3CN to HCN by ion-ion recombination.
 This paper reports the first calibrated in situ measurements of HCN in the upper troposphere/lower stratosphere. The measurements were made with the Air Force Research Laboratory's CIMS on the DC-8 during the NASA Sage III Ozone Loss and Validation Experiment (SOLVE), mostly in the Arctic. The measurements were made at cruise altitudes and cover the 10–12.5 km range. The present results are from the third deployment, which took place from 27 February to 15 March 2000.
 The instrument has been described in detail elsewhere [Hunton et al., 2000; Miller et al., 2000; D. Hunton et al., unpublished manuscript, 2001] and this paper only focuses on details relating to HCN measurements. The main goal for the CIMS instrument during SOLVE was to measure HNO3 concentrations and a secondary goal was to measure SO2 concentrations. Therefore most of the instrument time was devoted to those gases. However, HCN was detected by the same source ion and approximately 5% of the measurement time during the third deployment was devoted to HCN. The overnight transit flight from Kiruna began on 27 February 2000. Six flights to Kiruna concentrated on the polar region and the plane returned to Dryden on 15 March. Calibrated HCN measurements were made during all of these flights.
Figure 1 shows a typical CIMS mass spectrum obtained during the SOLVE mission. Note that the signal intensity scale is logarithmic. The CO3− primary ion (60 amu) is saturated in intensity, but the 13C isotope at 61 amu is on scale. CO3−(HCN) is found at 87 amu, where the spectrum is uncongested except for a peak at 88 amu, namely CO3−(N2). Masses 112 and 123 amu are SO5− and CO3−(HNO3), which are used to detect SO2 and HNO3.
 Because of the high gas volumes used in our system, it was not possible to do an in−flight zero air check to see if there was a background at mass 87 due to a species other than HCN. However, several pieces of evidence suggest that this cannot be an appreciable error (<10%). (1) Laboratory tests in dry nitrogen and zero air show no peak at mass 87. This is the most compelling evidence. (2) The ion source gas is ultrapure oxygen. Therefore producing HCN in the ion source requires all three atoms to be present as impurities; laboratory tests show this does not happen. (3) Most reactions of CO3− involve either O− transfer or clustering. There is no obvious choice for a species other than HCN, which could react with CO3− to make an ion of mass 87 [Ikezoe et al., 1987]. For instance, the only other species listed in the NIST Webbook [Mallard and Linstrom, 2001] containing CO3− is C3H3O3−, an extremely unlikely impurity. (4) Calibrations both in the laboratory and in flight indicated that HCN does not stick to the flow tube wall and the HCN calibration gas cannot be a source of background signals when the calibration gas is turned off. For all these reasons, we believe that the only background for HCN is the noise of the electron multiplier, equivalent to <1 counts per second or less than 1% of the HCN signal.
 The bond strength of the CO3−(HCN) cluster has not been measured, although it is obviously sufficient to allow detection of HCN, and the rates of reactions (1) and (2) are unknown. The Heidelberg group estimated that reactions (1) and (2) proceeded at the collision rate with rate constants equal to the similar reactions with HNO3 [Schneider et al., 1997; Spreng and Arnold, 1994]. Even if this assumption is correct, two problems may occur. Because CO3−(HCN) is a weakly bound cluster, collisional dissociation of CO3−(HCN), either thermally in collisions with the bath gas or in sampling, may occur. In addition, the Heidelberg group has shown that large amounts of H2O may cause the reverse of reaction (2) to occur, evidence that CO3−(HCN) is a weakly bonded cluster. At the low H2O concentrations present during SOLVE this was not a problem. Both of these effects might change the detection efficiency for HCN even if the rate constants were collisional and therefore we feel it is vital to calibrate the instrument under flight conditions.
 Calibration was accomplished in the same way as for HNO3 and SO2. A known quantity of HCN was injected at the upstream end of the CIMS flow tube and the response of the instrument to the known concentration was measured under conditions identical to the atmospheric measurement. The flows were derived from permeation tubes contained in a temperature, pressure, and flow controlled vessel. The calibration gases were carried into the flow tube by O2. The emission rate of the HCN permeation tube was monitored from the weight loss of the tube during the campaign. Figure 2 shows a plot of the CO3−(HCN), SO5−, and CO3−(HNO3) ion signals versus time during a typical calibration. Five-second averages are plotted and the relative time spent on each signal is apparent in the noise in the signal. Both the SO5− and CO3−(HCN) signals show sharp onsets and offsets, indicating that HCN did not interact strongly with the flow tube walls, which were heated to 80°C. Laboratory experience indicates that wall temperatures below about 10–20°C may allow adsorption of calibration gases and slow degassing afterward, i.e., the instrument would have a very slow time response. However, the shape of the calibration curve indicates that HCN did not interact with the walls and that the background was low. Note the CO3−(HNO3) signal has longer time constants.
 The sensitivity of the instrument to HCN is shown in Figure 3. The sensitivity was found to vary significantly with altitude, ranging from 2700 counts/ppbv-s at 10 km to 1200 counts/ppbv-s at 12.5 km. Assuming that 100 counts are needed to obtain 10% statistics, these sensitivities are equivalent to detection limits of 37 pptv and 83 pptv in 1 second of time spent measuring 87 amu. During SOLVE 1 s of integration time for HCN occurred every 20–30 s. Faster time responses are obviously possible if the measurements concentrated on HCN. The increasing sensitivity at lower altitudes is probably a complicated function of flow dynamics, pressure and temperature in the ion source region.
 Also shown in Figure 3 are the sensitivity to HNO3 and the ratio of the two sensitivities. At low altitude, the instrument was more sensitive to HNO3 by a factor of about 2.5. This increased to a factor of 3.5 at high altitude. This contrasts to the assumption made by Arnold and coworkers that the sensitivities should be similar because the collision rates are similar [Schneider et al., 1997; Spreng and Arnold, 1994]. However, the ratios of sensitivities presented here are not directly applicable to the previous measurements of the Heidelberg group. Several differences in the instrumental setups are apparent. Both CO3− and NO3− ions were used as reagent ions in the Heidelberg experiments. While the chemistry for NO3− ions is similar to that for CO3−, rate constants for neither ion reacting with HCN have been measured. The flow tube temperatures and pressures are different and the amount of hydration may vary. In any case, the present measurements call into question the previous practice. The increase in the sensitivity ratio with altitude may reflect that at lower pressures HCN does not cluster as efficiently as does HNO3.
 The total error is estimated as follows. The parameters used to derive concentrations and their uncertainties are, HCN permeation rate (10%), neutral flow velocity (15%), temperature (2%), pressure (3%), and ion signal stability (5%). Propagation of error gives a total error of 19%.
3. Results and Discussion
Figure 4 displays 1-min averages of HCN concentration versus time during the 8 March 2000 flight. The results on this flight are typical. Only small variations in the HCN concentration are observed and the average value is 250 ± 36 pptv. Altitude and HNO3 concentrations from the CIMS instrument are shown in the lower panel of Figure 4. High HNO3 mixing ratios indicate that the airplane was in lower stratospheric air and low mixing ratios are representative of upper tropospheric air (D. Hunton et al., unpublished manuscript, 2001). No correlation between HNO3 and HCN is found indicating that the HCN concentration does not change appreciably between the troposphere and stratosphere.
 During the course of the deployment, the DC-8 encountered the Mount Hekla volcanic plume on numerous occasions. The clearest indication of the plume came from simultaneous increases in SO2 (from our CIMS instrument) and particles. The most intense volcanic encounter was on the 27–28 February 2000 transit flight when SO2 concentrations approaching 1 ppmv were found (D. E. Hunton et al., unpublished data, 2000) No enhancement in HCN was observed during the volcanic encounters.
 Calibrated HCN data were obtained only over a limited altitude range during these flights, because no HCN calibration data were taken during ascent and descent. To test for any altitude dependence, average values of HCN concentrations were calculated for each constant altitude flight segment for all flights. The results are plotted in Figure 5 where the solid points represent the 8 March flight and the open points are from all other flights. The average value for all flights is 280 pptv with a standard deviation of 48 pptv, as shown by the crosshatched area.
 The data from 8 March show a slight altitude dependence of about 26 pptv/km. Most other flights also show a small positive altitude dependence, with values ranging from −4 to 58 pptv/km. The slope of all the data in Figure 5 is 17 pptv/km. Averaging the values for each individual flight gives 23 pptv/km. The dashed line in Figure 5 has this slope and is centered on the average value. The line is within the scatter of the data, i.e., in the crosshatched area. Therefore, the data suggest but do not conclusively show, there is a small positive altitude dependence. Two previous studies indicate that a small altitude dependence may be expected. Schneider et al.  observed an altitude dependence of this order, but we have already questioned the accuracy of those measurements. They speculated that the reason for the dependence may be conversion of CH3CN into HCN by ion recombination, i.e., H+CH3CN + NO3− → HCN + CH3NO3. In practice, both the positive and negative ions in this reaction would have additional ligands [Viggiano and Arnold, 1995]. This chemistry also explained a measured decrease in the CH3CN concentration in this altitude range. A three-dimensional (3-D) modeling study by Li et al.  shows HCN concentrations that increase with altitude in the Arctic region due to outflow of tropical air (with high HCN from biomass burning) in the upper troposphere and lower stratosphere.
 The only previous the in situ measurements are those of Arnold and colleagues [Schneider et al., 1997; Spreng and Arnold, 1994], who found a mean value of only 164 pptv. Those measurements also used a CIMS with similar chemistry to that used here but with the sensitivity assumed to be equal to that for HNO3. As shown earlier, the present measurements are less sensitive to HCN than to HNO3 and that the sensitivity ratio depends on altitude. Multiplying the Heidelberg measurements by the sensitivity ratios shown in Figure 3 results in values that are considerably higher those found here and elsewhere. This indicates that the Heidelberg instrument sensitivity is probably between their assumed sensitivity and that found in the present measurements. The Heidelberg CIMS operated at higher pressures and lower temperatures than the present one. Both of these conditions would lead to reaction (1) being faster than in the AFRL CIMS.
Rinsland et al.  have derived HCN profiles in the stratosphere from solar occultation spectra. The measurements were made from ATMOS/ATLAS 3 during November 1994 in the tropics. In the 12–15 km range, they found values that varied from just over 200 pptv to about 940 pptv. About half of the measurements were identified as having elevated HCN levels and the data set was divided into two groups, i.e., normal and high. For the 15 km or lower data, the mean and median of the normal set are 263 and 271 respectively. This compares very well with the present values. The high data points have a median of 451, well above the present values. The authors indicated that these were due to some unknown pollution event. For the data at 15 km and lower, which were all in the troposphere, no altitude dependence was observed.
 There have been numerous measurements of HCN column densities and several studies of HCN in the upper atmosphere by remote sensing. The most recent study is that of Zhao et al. [2000, 2002] who measured total column densities of HCN and other gases over Japan from 1995 to 2000. They found a strong seasonal dependence and essentially no yearly dependence except for an enhancement in 1998. The enhancement is due to a large number of Asian fires in 1998. Converting column densities into concentrations yields a yearly average value of 261 pptv and a typical yearly minimum and maximum of 195 and 333 pptv. The average value agrees very well with our average value of 280 pptv ± 48 pptv. If the model of Li et al.  is correct, higher than average concentrations are expected at the altitudes of the present measurements.
Notholt et al.  measured total column densities above a ship cruise and found fairly large variations with latitude. For instance, at 20–25 N the 0–12 km average density was 109 pptv, while at 25–30 N a value of 220 pptv was found. No measurements further north were made and a maximum was found at 10–15 S. Rinsland et al.  made measurements in Australia during 1997–1998. They found a weak altitude dependence with an average value of 200 pptv during February to July, a period without an appreciable biomass source. From August to December, they found increasing values up to about 12 km. The peak was about 400 pptv. During this period correlations with other gases showed a strong biomass source.
 Previous column measurements [Mahieu et al., 1995; Rinsland et al., 1999, 1982] converted to average densities have found widely varying results, although they are generally lower than 200 pptv. This is lower than the most recent results. Remote sounding measurements of the stratospheric concentration also have large variation and sometimes uncertainties [Abbas et al., 1987; Carli and Park, 1988; Coffey et al., 1981; Rinsland et al., 1998, 1982; Zander et al., 1988]. Thus the present measurements agree best with the recent column density measurements over Japan and the remote sensing measurements of Rinsland et al. . It is not known whether the disagreement with the older measurements is due to variations in the season or position of the measurements or whether the concentrations of HCN have actually increased by ∼100 pptv in the last twenty years.
 The 3-D model calculations [Li et al., 2000] assume biomass burning as the only source of HCN and the oceans as the main sink. Northern hemispheric values vary from about 300 pptv at low altitudes in the tropics to about 220 pptv at 10–15 km over much of the Northern Hemisphere. As mentioned earlier, a small positive altitude dependence is predicted in the high latitude region. This is due to the distance from the source. These values are slightly lower than the present values as well, but in recent conversations with the authors, it was estimated that if biofuels were included as a source in that model, HCN concentrations would increase about 25%. The magnitude of the ocean sink is adjustable as well; a weaker ocean sink would give higher HCN (D. Jabob and Q. Li, private communication, 2001). This would bring the predictions in good agreement with the present measurements. However, maximum HCN source strengths for biomass burning were also used.
 In summary the present measurements show that that technique originally proposed by Arnold and colleagues for measuring atmospheric HCN concentrations is valid. However, it is shown that calibration is necessary for accurate measurements and that the assumption used to estimate sensitivity in previous in situ measurements is not valid. The technique works best under low water concentrations but should be useful under all conditions as long as the calibration is performed under the same conditions. Use of an isotopically labeled calibrant (H13C15N) would make continuous measurements possible under all conditions. The double-labeled species is necessary to avoid a mass coincidence with CO3−(N2).
 The present measurements report the first accurate in situ measurements of HCN in the upper troposphere/lower stratosphere. The average value of 280 pptv agrees well with the yearly average values over Japan derived from column density measurements, and the remote sensing values from ATMOS/ATLAS 3, but less well with other determinations. Much of this deviation is probably due to variations with season, altitude, and position as predicted by the 3-D model results for HCN. This model includes biomass burning as the major source and the ocean as the major sink. A slight positive altitude dependence is hinted at in the data and is in the direction predicted by the model. To fully understand the HCN budget additional coordinated measurements are needed.
 We would like to thank Willard Thorn, John Borghetti, and Frank Federico for technical support. We also thank Daniel Jacob and Qinbin Li for helpful discussions. This research has been supported the NASA SOLVE project and by the Air Force Office of Scientific Research under Project 2303EP4. T.M.M. was supported through Visidyne contract number F19628-99-C-0069.