Perchlorate in stratospheric aerosol particles

Significance

Perchlorate chemistry connects a very wide range of environmental issues. It is produced in the stratosphere and is a health hazard when in drinking water. Measurements in the stratosphere show that almost none of the perchlorate is in the majority sulfuric acid particles and almost all is in less-common particles such as those from biomass burning. This raises the possibility that intensifying fires with climate change or the intentional modification of stratospheric aerosols could significantly increase the amount of perchlorate in the atmosphere.

Abstract

Perchlorate is a toxic, regulated contaminant in drinking water. According to previous isotopic studies, much of the perchlorate deposited to the Earth’s surface is formed in the atmosphere, with ³⁶Cl suggesting a large contribution from the stratosphere. Here, we present measurements of perchlorate in stratospheric aerosol particles and confirm that the stratosphere is an important source of perchlorate, whereas we did not observe production in the troposphere. Mass mixing ratios of aerosol perchlorate in the stratosphere were 1 to 10 parts per trillion by mass (pptm), with the highest concentrations observed in summer and in the Southern Hemisphere. Almost all of the perchlorate is in biomass burning and nitrogen-rich particles, despite those types contributing only a few percent of the aerosol particles. Such particles are less acidic than the majority of sulfuric acid particles. If the formation of perchlorate is sensitive to acidity, then the injection of some materials for solar radiation modification might significantly increase the global production of perchlorate.

Perchlorates are chemicals containing the ClO₄⁻ group. Perchlorates in drinking water are toxic to humans because of inhibition of iodine uptake and related hormones (1, 2). Perchlorate in drinking water has been regulated at the state level (3), and the US Environmental Protection Agency (EPA) has committed to issue a proposed National Primary Drinking Water Regulation for perchlorate (4).

Isotopic studies, especially of ³⁶Cl formed by cosmic rays, show that perchlorate formed in the stratosphere is a major component of groundwater perchlorate (5). Mass-independent oxygen isotope fractionation in perchlorate also supports an atmospheric source involving ozone (6, 7). A lower surface deposition rate in the tropics than at higher latitudes (8) is also consistent with a stratospheric source because most downward transport from the stratosphere is outside of the tropics. Single particle mass spectrometry has definitively identified the presence of perchlorate in stratospheric aerosol particles (9). In contrast to the formation in the stratosphere, significant amounts of perchlorate are not formed from the much larger amount of chlorine released from sea-salt particles at low altitudes (8).

Although industrial contamination is important for localized perchlorate in groundwater, on a continental scale, much of the perchlorate in groundwater originates in the atmosphere (10, 11). Another major source of perchlorate in the United States is former use of Chilean nitrate fertilizer containing natural perchlorate, possibly from long-ago stratospheric production (11). Once in groundwater, perchlorates have extremely long lifetimes, especially in arid regions, where lifetimes can exceed 10,000 y (12). Perchlorate formed during the Pleistocene epoch is still significant for groundwater in the southwestern United States (13).

The perchlorate concentration in both Greenland and Antarctic snow scaled with the chlorine content of the stratosphere as chlorofluorocarbon (CFC) concentrations increased from the 1960s to the 1990s (14, 15). Together with the long lifetime of perchlorate once it reaches the surface, the implication is that increases in stratospheric chlorine content have had slow, but nearly permanent, impacts on groundwater pollution. This pulse of stratospheric perchlorate could be the longest-lasting legacy of CFC emissions, remaining long after their atmospheric lifetime is over. It is not widely recognized that controls on CFCs, implemented because of ozone destruction, probably also prevented some perchlorate groundwater pollution.

Perchlorate is less than 1% of the reactive chlorine budget in the stratosphere so its abundance is limited by efficiency of its formation, not the amount of available chlorine. The formation chemistry of perchlorate in the atmosphere is still the subject of considerable uncertainty. There is a lack of definitive studies that show the fundamental steps (rate coefficients, mechanisms) that are involved in the conversion of chloride to HClO₄ in the gas phase or perchlorate in the condensed phase.

A number of studies have observed small amounts of ClO₄⁻ being formed when Cl species react with O₃ with or without the presence of liquid water (16). A recent modeling study by Chan et al. (17) proposed a gas phase mechanism by which HClO₄ could be formed in the stratosphere via the reaction of OClO with O₃ to form ClO₃ followed by reaction with OH radicals to form HClO₄. To explain observations of very low levels (<1 × 10⁶ cm⁻³) of gas-phase HClO₃ and HClO₄ in the Artic boundary layer, Tham et al. (18) postulated a mechanism in which BrO_x, ClO_x, and HO_x radicals react to form HClO_x compounds in the gas phase, with corresponding O₃ loss through BrO_x/ClO_x pathways. These chemical cycles have the necessary feature that they produce perchlorate with a large excess of ¹⁷O. However, that modeled excess is larger than those in perchlorate measurements in remote regions where it is thought that perchlorate is from natural sources. This suggests that there is at least one step, possibly in the condensed phase, involving water in the stratosphere, because water has a much lower ¹⁷O-excess relative to O₃ (19).

Here, we present single-particle observations in the stratosphere that place constraints on the processes controlling perchlorate. Notably, almost none of the perchlorate in the stratosphere is found on the most common sulfuric acid particles there. Much of the perchlorate is on biomass burning particles that have been transported to the stratosphere. A significant amount of perchlorate is also on nitrogen-rich particles that are similar in composition to those found in the Asian tropopause aerosol layer (20). A typical biomass burning or nitrogen-rich particle in the stratosphere contains more than 200 times as much perchlorate as a similar particle in the upper troposphere. Therefore, the perchlorate is added in the stratosphere. Our data indicate much more perchlorate in the lowermost stratosphere in the Southern Hemisphere than in the Northern Hemisphere.

Results

Particle Analysis by Laser Mass Spectrometry (PALMS) instruments have been flown on the NASA WB-57 aircraft to measure the composition of single aerosol particles in the stratosphere (Methods). A new PALMS-NG version flown since 2021 measures both positive and negative ions from each particle. Positive ion spectra have more information about a broad range of species and are usually more useful than negative ions for determining the source of a particle. Negative ion spectra have more information about anions such as sulfate, nitrate, and, for this work, perchlorate.

Data from three aircraft missions are presented here. The primary dataset is the PALMS-NG instrument (21) on the NASA WB-57 aircraft during the 2023 Stratospheric Aerosol processes, Budget and Radiative Effects (SABRE) mission. These data are mostly from flights from Fairbanks, Alaska. Test flights from Houston, Texas are excluded. The WB-57 reaches altitudes of up to about 19 km and the SABRE flights accessed air in the stratospheric polar vortex as it distorted and broke up in late winter. About 450,000 mass spectra were obtained in the stratosphere during the SABRE mission. A nearly identical PALMS-NG instrument flew on the NASA ER2 aircraft during the 2022 Dynamics and Chemistry of the Summer Stratosphere (DCOTSS) mission. To avoid cloud contamination, DCOTSS data are filtered to exclude relative humidity over ice of more than 50% or air influenced by recent convection. Data from an older, single polarity, PALMS instrument (22) on the NASA DC–8 during the 2016–2018 Atmospheric Tomography (ATom) mission are included to provide a comparison to the Southern Hemisphere. The flights on the DC–8 do not extend as high into the stratosphere as SABRE or DCOTSS, but stratospheric air with O₃ values of more than 200 ppbv were sampled at high latitudes and in tropopause folds.

Particles Containing Perchlorate.

Fig. 1 shows two examples chosen to represent the most common types of positive ion spectra associated with large perchlorate signal: biomass burning particles and nitrogen-rich particles with a large NO⁺ peak. In laser ionization of particles, NO⁺ can be formed from many nitrogen-containing species, including ammonium, and is not specific for nitrate. Biomass burning particles are defined by a combination of potassium signals, organic peaks, and the absence of other markers (23, 24). In the stratosphere, many biomass burning particles coagulate with sulfuric acid particles containing meteoric metals. To include these, Fig. 2 also shows a looser definition of modified biomass burning particles, defined simply as a ³⁹K⁺ signal larger than 10% of the positive ion signal. Such a simple definition would not work in the troposphere but in the stratospheric SABRE data there were essentially no other potassium-rich particles such as sea-salt or mineral dust. The examples shown in Fig. 1 have NO⁺ and ³⁹K⁺ peaks about twice as large as the thresholds used in subsequent figures.

Fig. 1.

Sample mass spectra of single stratospheric particles containing perchlorate. The carbon and potassium peaks identify the particle in (A) as originally from biomass burning. The large NO⁺ peak in (B) indicates nitrate and/or ammonium. Above m/z 115, the positive ion spectra are shown multiplied by 100 and offset by 0.2 to show small peaks from iodine and a potassium perchlorate cluster ion.

Fig. 2.

Quasivertical distributions of (A) the frequency of relevant types of particles and (B) the amount of perchlorate in those particles. The tropopause is near the bottom of the vertical scale and the lowest N₂O mixing ratios encountered during the flights were inside the polar vortex. The inset shows mass fractions of the biomass burning and nitrogen-rich particles near the tropopause on a logarithmic scale. The two points less than 10⁻⁴ are below the detection limit.

Fig. 2A shows the abundance of various types of particles as a function of the N₂O concentration, with lower N₂O values indicating air that is higher (deeper) into the stratosphere and has spent more time there. Using N₂O concentrations has been shown to be an effective way of averaging aerosol data in the stratosphere (25). Biomass burning particles and nitrogen-rich particles are most abundant at the bottom of the stratosphere. Modified biomass burning particles occur in older (lower N₂O) air where there has been time for coagulation with stratospheric particles. All three types, along with organic-rich particles from the troposphere, are uncommon in air with less than about 250 ppbv N₂O, where the vast majority of particles are sulfuric acid with or without meteoric metals.

The biomass burning and nitrogen-rich particles acquire substantial amounts of perchlorate. Fig. 2B shows the average mass fraction of perchlorate for each particle type. Mass fractions are calculated using laboratory calibrations for the sensitivity of PALMS to perchlorate along with the organic content of each particle type (Methods). For N₂O less than about 220 ppbv, almost all nitrogen-rich or biomass burning particles contain perchlorate (SI Appendix, Fig. S1).

PALMS has measured many biomass burning particles in the upper troposphere and many nitrogen-rich particles near the Asian tropopause during the Asian Summer Monsoon Chemical & CLimate Impact Project (ACCLIP) mission. There was no detectable perchlorate on the fresh particles in the troposphere; the perchlorate is added to such particles in the stratosphere. The presence of very rare particles in the troposphere with perchlorate is consistent with a stratospheric source and a shorter aerosol residence time in the troposphere than in the stratosphere. Comparison of the mass spectra to those from the space shuttle (26) excludes solid rocket exhaust as the source of the stratospheric particles containing perchlorate.

Almost all Perchlorate Is in Rare Particle Types.

Fig. 3A shows an average volume size distribution from optical particle counters for an N₂O concentration range that is well into the stratosphere. The shaded regions are the size distributions for three types of particles obtained by multiplying the size distribution by the fraction of particles at each size that PALMS identifies as belonging to a type. As expected from Fig. 2A, the nitrogen-rich, biomass burning, and modified biomass burning particles contribute only a tiny fraction of the aerosol volume. Fig. 3B shows a similar size distribution, except now multiplied by the perchlorate ion signal for each particle type as a function of particle size and converted to mass fraction using laboratory calibrations. In contrast to the overall volume (Fig. 3A), almost all of the perchlorate is in biomass burning, modified biomass burning, and nitrogen-rich particles. The remaining unshaded area is mostly from particles that do not quite meet the criteria for nitrogen-rich or biomass burning. Sulfuric acid particles with or without meteoric metals, which are almost all of the volume in Fig. 3A, contribute almost nothing to the perchlorate signal.

Fig. 3.

Size distributions of all particles and of perchlorate during the 2023 SABRE Alaska flights. This 175 to 225 ppbv range of N₂O corresponds to the peak perchlorate mixing ratio (Fig. 4). Panel (A) shows the volume-weighted size distribution from optical particle counters. Volumes are at standard conditions (1 atmosphere and 273 K). The shaded areas are the volumes multiplied by the fraction of specific particle types identified by PALMS at those sizes. The solid line in (B) shows the aerosol volume in the upper panel multiplied by the fraction of perchlorate ions in the negative ion spectra at each size. The shaded areas again allocate the perchlorate signal by the type of particle. Nearly all of the perchlorate signal was from nitrogen-rich particles, biomass burning particles, and modified biomass burning particles.

Much of the perchlorate is in biomass burning particles about 0.5 µm diameter. A significant amount of perchlorate is also in nitrogen-rich particles, most of which are smaller than the biomass burning particles. This size difference is consistent with the size of the particles when they enter the stratosphere. Nitrogen-rich particles are less than 200 nm diameter when they enter the stratosphere through the Asian tropopause aerosol layer (20), whereas the volume median diameter of biomass smoke is slightly larger (27), especially for biomass burning particles injected into the stratosphere from pyrocumulus (28). Growth from the addition of perchlorate may also contribute to the observed sizes.

There are at least two important ways that both biomass burning and nitrogen-rich particles differ from stratospheric sulfuric acid particles. They have more organic content and are not as acidic as the sulfuric acid particles. Within each type of particle, the perchlorate content was not well correlated with organic content. For example, biomass burning particles with high organic content had similar perchlorate signals to biomass burning particles with lower organic content. In contrast, the perchlorate signals were correlated with the K⁺ signal within the biomass burning category and with the NO⁺ signal within the nitrogen-rich category. Particles with high organic content but without biomass burning or nitrogen markers contained smaller amounts of perchlorate (Fig. 2b). These correlations suggest that (less) acidity may be more important than organic content.

Perchlorate Concentrations in the Stratosphere.

Fig. 4 shows the mass mixing ratio of perchlorate in stratospheric aerosols calculated by integrating over size distributions such as Fig. 3B. A small amount of perchlorate may be in particles less than 120 nm diameter and not included in the integral over particle sizes. The highest perchlorate concentrations are at about 200 ppbv N₂O. The contributions of the various particle types are shaded. Throughout the entire vertical profile, most of the perchlorate is on either biomass burning or nitrogen-rich particles. As a reference, 200 ppbv N₂O corresponds to a mean age in the stratosphere of about 4 y (29), although the biomass burning and nitrogen-rich particles may be associated with the younger portion of the spectrum of ages.

Fig. 4.

A quasivertical profile of the perchlorate mixing ratio during the 2023 SABRE Alaska flights integrated over the volume size range, as shown in Fig. 3. The highest N₂O mixing ratios were at the tropopause and the lowest at higher altitudes in remnants of the polar vortex.

During the ATom mission, perchlorate concentrations in the Southern Hemisphere were several times those in the Northern Hemisphere at similar ozone concentrations. Fig. 5 shows profiles of the perchlorate signal integrated over the size distribution. This quantity is proportional to the perchlorate mass and can be more easily calculated for the older PALMS instrument flown during ATom. Perchlorate concentrations in the Southern Hemisphere were several times those in the Northern Hemisphere at similar ozone concentrations. The higher concentrations in the Southern Hemisphere stratosphere are consistent with a surface flux of perchlorate to Antarctic snow that is up to 10 times the flux to Arctic snow (30).

Fig. 5.

The integrated perchlorate signal as a function of ozone. The ATom curves in the two panels are colored by season rather than by the deployment number. The earlier, unipolar, version of PALMS used during ATom does not allow allocating the perchlorate signal to types of particles as shown in Figs. 2 through 4. However, if the same average types found in SABRE are assumed to apply to ATom, a value in this figure of 0.008 would be between 7 and 8 pptm of perchlorate.

In both hemispheres, the highest perchlorate concentrations during ATom were during local summer and the lowest concentrations were in winter. The SABRE and DCOTSS data show the same pattern to higher ozone concentrations.

Most of the perchlorate in the lowermost stratosphere is in aerosols rather than the gas phase. An iodide chemical ionization mass spectrometer (CIMS) flown on the later ATom deployments (31) detected peaks that are consistent with gas-phase perchloric acid in the stratosphere. Because of concern about formation of perchloric acid on the inlet wall from O₃ reacting with other chlorine compounds, this detection should tentatively be treated as an upper limit. For the very lowermost stratosphere sampled during ATom, that upper limit was about 0.1 pptv, which is much less than the amount of aerosol perchlorate.

Properties of Particles Containing Perchlorate.

Some information on other chlorine species can be obtained from the ratios of peaks in the negative ion mass spectra. The sizes of the ClO⁻, ClO₂⁻, and ClO₃⁻ ion peaks closely followed ClO₄⁻ and were consistent with the amount observed from fragmentation of perchlorate in the laboratory calibrations. Thus, there is no evidence for significant amounts of chlorate or other oxidized chlorine species other than perchlorate in the aerosol particles. In contrast, the Cl⁻ ion had a different vertical profile than ClO₄⁻. At less than 280 ppbv N₂O, the Cl⁻ ions were consistent with the majority coming from fragmentation of ClO₄⁻. Closer to the tropopause, there was an additional Cl^- ion signal from particles, especially organic-rich particles, that often did not contain any perchlorate. This is consistent with uptake of HCl by organic-rich particles (32). There was also a very small Cl^- signal in vortex air with less than 180 ppbv N₂O, associated with metal-rich particles.

It is likely that the nitrogen-rich perchlorate particles in the stratosphere are solid. In PALMS, the light scattered from the continuous laser beams by each particle can be compared to the aerodynamic diameter to give an effective density. Nonspherical particles have effective densities lower than their true densities (33). The nitrogen-rich particles containing perchlorate had an effective density of less than 1.3. That is too low for their composition but is consistent with crystallization of mixed ammonium nitrate and ammonium sulfate particles characteristic of the Asian tropopause aerosol layer (34). The deliquescence relative humidity for ammonium perchlorate is over 94% at room temperature (35) and over 84% for potassium perchlorate (36). The phase of the biomass burning particles with perchlorate is unclear. Their effective density, although low, was usually larger than 1.3 and could be explained by either nonsphericity or a low-density organic component.

Discussion

The Budget of Perchlorate in the Stratosphere.

The flux of perchlorate down from the stratosphere can be estimated by ratioing it to stratospheric sulfuric acid. Although the perchlorate and sulfuric acid are on different particles, they are subject to the same air motions. Gravitational sedimentation will be similar because the particle sizes are similar (Fig. 3). That creates a strong correlation between the amount of perchlorate and sulfuric acid in the lowermost stratosphere. Feinberg et al. (37) estimated the flux of stratospheric sulfur into the troposphere as 121 Gg y⁻¹, or 370 Gg y⁻¹ of sulfuric acid. There are some opposing caveats to applying this to the SABRE data. The Feinberg et al. model has too much gas-phase SO₂ at the tropopause compared to observations (38), which would lead to an overestimate of the sulfur flux. On the other hand, the SABRE data in 2023 were influenced by some sulfate formed after the Hunga-Tonga eruptions about a year earlier, which would increase the sulfur flux.

Using the ratio of perchlorate to sulfuric acid aerosols from the SABRE data and a ratio of two to three times that for the Southern Hemisphere (Fig. 5) yields a global downward flux of stratospheric perchlorate of about 2.5 Gg y⁻¹. Uncertainties in both the PALMS data and the stratospheric sulfur budget imply a range of about 1.2 to 6 Gg y⁻¹. The central estimate of 2.5 Gg y⁻¹ may be low because the SABRE data were in winter and the ATom and DCOTSS data suggest that perchlorate concentrations are higher in summer.

Chan et al. (17) estimated a gas-phase global stratospheric source of 1.7 Gg y⁻¹. This is something of an upper limit because their base case did not include photolysis of ClO₃, which they estimated might drastically reduce gas-phase perchlorate formation. Even so, the gas-phase stratospheric chemistry did not generate enough perchlorate to match observed surface fluxes, especially in the Southern Hemisphere.

An analysis of samples from the National Acid Deposition Program (39) estimated a wet deposition flux of 6.5 ± 3 µg m⁻² y⁻¹ to the conterminous United States. Two North American glaciers had recent deposition fluxes of 1 to 5 µg m⁻² y⁻¹ (40). For comparison, a global stratospheric source of 2.5 Gg y⁻¹ corresponds to 7.5 µg m⁻² y⁻¹ over the Earth poleward of 20° where most of the downward transport from the stratosphere occurs (41, 42) and, since we found more perchlorate in the Southern Hemisphere, about 5 µg m⁻² y⁻¹ in the extratropical Northern Hemisphere. Although a uniform deposition is too simple, this comparison indicates that the stratospheric source could account for much or even all of the perchlorate in precipitation. A May-to-July peak in perchlorate deposition to the United States (39) is also reasonably consistent with a spring maximum in the downward transport of ozone in the Northern Hemisphere (41, 42) along with a summer maximum of perchlorate relative to ozone (Fig. 5).

Although Figs. 4 and 5 show that the stratosphere is a major source of perchlorate in aerosol particles and show no evidence of perchlorate production in the upper troposphere, it is difficult to set a stringent upper limit on a possible tropospheric contribution because the shorter residence time of tropospheric aerosols amplifies the possible surface flux from even very small amounts of perchlorate per particle. To set an upper limit on the perchlorate content of tropospheric particles, we averaged the raw mass spectra of biomass burning particles in the upper troposphere during SABRE. The averaging largely eliminates electronic noise and reaches a limit set by the variability of unusual organic fragment ions at almost every unit mass. The resulting best estimate for the perchlorate mass fraction in upper tropospheric biomass burning particles is near zero with an upper limit of about 0.0003, compared to an average mass fraction of about 0.1 in the stratosphere (Fig. 1). Even with more biomass burning particles and a shorter residence time in the upper troposphere, the downward flux of perchlorate formed in the upper troposphere is probably less than the input from the stratosphere. Indeed, a significant fraction of the average tropospheric signal at one of the least cluttered perchlorate fragment peaks (³⁵ClO₂⁻) came from just 5 particles (out of over 9,000) that had mass spectra consistent with particles mixed down from the stratosphere. In the boundary layer, particle residence times are so short that there could be a perchlorate flux from particles with amounts too tiny for PALMS to detect. Jiang et al. (30) reported a lack of a correlation between perchlorate and sea-salt chloride.

Implications for Perchlorate Chemistry.

Our data provide the striking observation that almost all of the perchlorate in stratospheric aerosol particles is present only in a very small percentage of particles with special compositions. There are different implications for perchlorate chemistry depending on the explanation.

A Henry’s Law calculation by Jaeglé et al. (43) estimated that sulfuric acid particles at stratospheric conditions can contain less than 0.01% perchlorate, which is qualitatively consistent with the data reported here. Perchloric acid formed in the gas phase or even on the surface of the more abundant sulfuric acid would end up in biomass burning and nitrogen-rich particles where it could be stabilized by forming compounds such as potassium perchlorate and ammonium perchlorate. If there is migration of perchlorate from aerosol particles where it is less stable to particles where it is more stable, that is probably a slow process. At HClO₄ gas-phase concentrations of less than 0.1 pptv suggested by the CIMS data during the ATom mission, adding 20% perchlorate to the mass of a 0.5 µm particle would take months or more.

A second possible explanation for the presence of perchlorate in nitrogen-rich and biomass burning particles is that the perchlorate forms in those particles. If so, an important corollary of perchlorate formation in certain particles would be that adding more particles other than sulfuric acid to the stratosphere would result in the production of more perchlorate. In particular, major pyrocumulus injections such as the 2019–2020 Australian fires (44) could cause more perchlorate to form. This hypothesis may be testable with Antarctic snow samples. There were more biomass burning particles and higher perchlorate concentrations during DCOTSS than during SABRE (Fig. 5). Long-term trends in ammonium nitrate particles in the Asian tropopause aerosol layer may also be affecting stratospheric perchlorate concentrations. Increases in tropopause-overshooting convection with climate change (45) could bring more nonacidic particles into the lower stratosphere and provide more sites for perchlorate.

Indirect support for a heterogeneous mechanism comes from spikes of perchlorate in ice cores corresponding to spikes in sulfate from major volcanic eruptions (46). Smaller eruptions that did not reach the stratosphere did not produce perchlorate signals, showing that the perchlorate is not a direct volcanic emission but instead was formed in the stratosphere in response to the volcanic perturbation. The perchlorate observed after volcanic eruptions could simply be from the extra surface area of sulfuric acid particles at those times. However, the Southern Hemisphere during the 2016 ATom–1 deployment contained extra sulfuric acid from the 2015 Calbuco eruption (47), but the perchlorate was not significantly enhanced (Fig. 5). If perchlorate formation is enhanced by alkalinity, the association in ice cores with volcanic eruptions might be due to reactions on ash particles rather than sulfuric acid. Production of perchlorate in the troposphere may occur in Arctic conditions (18), but in our data, particles in the troposphere contained insignificant amounts of perchlorate. Although the troposphere usually contains some partially neutralized aerosol particles, it has less ClO_x chemistry than the stratosphere. Studies of the mechanisms of the formation of perchlorate on Mars (7, 48) may also provide information about perchlorate formation in the stratosphere, and vice versa.

Lower oxides such as ClO and OClO may be steps on the way to making perchlorate (49). The data here do not provide a clear picture of the role of these oxides. The Southern Hemisphere ozone hole chemistry is stronger, with more of these oxides of chlorine present than in the Northern Hemisphere. This might be the explanation for the higher perchlorate concentrations in the Southern Hemisphere (Fig. 5). The highest concentrations were sampled during local summer, which includes air mixed out after the polar vortex broke up. But if ClO or other chlorine oxides from polar ozone chemistry are a precursor to perchlorate, then it is not clear why the Northern Hemisphere measurements in remnants of the polar vortex (the lowest N₂O mixing ratios in Fig. 4) have less perchlorate than air outside the polar vortex. It may be that perchlorate formation requires air with both chlorine oxides and less acidic particles, which are especially rare in the polar vortex.

Whatever the mechanism for forming perchlorate, the association with aerosols after volcanoes (46) raises the concern that deliberate injection of material for solar radiation modification might produce more perchlorate. This could happen either if the production is simply dependent on surface area or if it depends on special chemistry in the particles. In this context, it is concerning that, although calcium-rich particles were very rare (≪0.1% of particles) in the stratospheric SABRE data, those that were present often contained perchlorate. Injection of climatically important amounts of calcium carbonate (50) would produce a massive reduction in the acidity of stratospheric aerosol particles. This could change the concentration of perchlorate if, as suggested by these measurements, its formation is sensitive to acidity. Perchlorate is a reminder that not all of the chlorine chemistry in the stratosphere is fully understood and there could be unexpected consequences of modifying the stratosphere.

Methods

Single-particle mass spectra were obtained with the PALMS instrument (21). This instrument determines the composition of individual aerosol particles by hitting them with a pulsed 193 nm laser beam and analyzing the resulting ions in a time-of-flight mass spectrometer. The recent next-generation instrument (PALMS-NG) has a bipolar mass spectrometer that analyzes both positive and negative ions from each particle. The measurements include particles between about 0.11 to 3 µm diameter. The earlier, unipolar, version used during the ATom mission had a lower size limit of about 0.14 µm. PALMS also measures the aerodynamic diameter of individual particles so that each mass spectrum can be associated with the size of the particle. Geometric diameters are derived from the aerodynamic diameters and densities calculated from the composition (33). The PALMS instrument flies on an aircraft and particles spend much less than 1 s between ambient conditions and analysis, minimizing changes in composition during sampling. Sampling locations are shown in SI Appendix, Fig. S2.

Perchlorate produces a characteristic negative ion signal in the PALMS instrument (SI Appendix, Fig. S2). For the analysis here, the perchlorate signal is defined as a sum of ClO⁻, ClO₂⁻, ClO₃⁻, and ClO₄⁻ ions. In each case, the peaks must be larger than nearby peaks so that occasional spectra with organic fragments at nearly every mass are not counted as perchlorate. Each pair of peaks must be consistent with the ³⁵Cl/³⁷Cl isotopic ratio. Because ³⁵ClO₄⁻ has the same integer mass as H³⁴SO₄⁻, the m/z 99 peak is corrected for sulfate based on the larger H³²SO₄⁻ peak. Small excesses of m/z 99 were not counted as perchlorate if the other ClO_x⁻ peaks were absent. Because the total ion yield can vary widely depending on how the pulsed laser hits a particle, perchlorate signals are normalized to the total negative ion signal for each particle.

Sulfuric acid particles are almost uniquely hard to ionize (51) and have a higher percentage of spectra with too few or no negative ions from which to compute a perchlorate signal. Spectra without negative ions probably come from particles near the edge of the ionization laser beam. Sulfuric acid particles that did ionize well contained almost no perchlorate. The perchlorate ion fraction was therefore set to zero for spectra with no negative ions except for a few (<0.1% of all spectra) with enough perchlorate that it was evident in the positive ions. Setting the perchlorate signal to zero for spectra with few or no negative ions only slightly underestimates the amount of perchlorate because almost all particles with perchlorate are of types that ionize well. In contrast, excluding spectra with few negative ions from the calculations would bias the fraction of particles with perchlorate high.

PALMS data are normalized to independently measured particle size distributions (33). Most of the PALMS size range was measured using an Ultra-High Sensitivity Aerosol Spectrometer (UHSAS) optical particle counter that was modified for aircraft operation (52). N₂O was measured with a gas chromatograph (53) and interpolated onto the sampling times of each PALMS particle. We utilize a product from Eric Ray that uses fast-response ozone measurements to intelligently interpolate between the less frequent N₂O points. Ozone is used in Fig. 5 because during ATom and DCOTSS (54), ozone had better data coverage than N₂O. Using N₂O or O₃ as an independent variable is common for stratospheric studies (25) and accounts for most of the meteorology and location of the sampling.

PALMS was calibrated for perchlorate by atomizing solutions of sodium perchlorate and ammonium sulfate and measuring the relative negative ion signals. The same equations for summing ClO_x⁻ peaks were used for the calibration and ambient data. The calibration was linear with a sensitivity to perchlorate similar to sulfate (SI Appendix, Figs. S3 and S4). The calibration provides the ratio of perchlorate to sulfate. Computing the perchlorate mass also requires an estimate of the organic fraction from the positive ion spectra and an estimated density. Propagating uncertainties for the calibrated sensitivity to perchlorate, correction for the organic fraction, particle density, and the normalization to the size distribution results in an overall uncertainty of about ±40%, mostly from the estimate of the organic fraction (33). However, the overall uncertainty could be dominated by how well the calibration particles mimic those in the atmosphere. In particular, the relative humidity during stratospheric sampling is less than 1% after the air warms to the temperature in the nose of the WB57. The laboratory particles, at more moderate relative humidity, may contain more water than those sampled in-flight. Although this does not nominally affect the calibrated ratio of perchlorate to sulfate, there is a chance that the presence of water might affect the ionization process. Based on experience with other species such as bromine and iodine (55), we subjectively estimate the overall uncertainty as less than a factor of two.

The iodide CIMS was deployed on ATom-3 and ATom-4 and has been described in detail elsewhere (31, 56). Gas phase perchloric acid (HClO₄) was detected via proton transfer to produce hydroiodic acid and the perchlorate anion (ClO₄⁻; m/z: 98.949 and 100.946). This reaction channel was confirmed in the laboratory by introducing gas-phase perchloric acid into the CIMS. At the time of this work, quantification of this signal was not possible; however, due to the gas phase acidity of perchloric acid, we expect a fast formation rate of ClO₄⁻. Therefore, we use the high sensitivity for N₂O₅ (35 ncps/pptv) as a proxy.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

SABRE data have been deposited in https://csl.noaa.gov/projects/sabre/data.html and https://csl.noaa.gov/groups/csl7/datasets/data/palms/ (57). DCOTSS data have been deposited in https://www-air.larc.nasa.gov/cgi-bin/ArcView/dcotss.2022 (58). ATom data have been deposited in http://dx.doi.org/10.5067/Aircraft/ATom/TraceGas_Aerosol_Global_Distribution (59).