Particle Formation by Ion Nucleation in the Upper Troposphere and Lower Stratosphere

See allHide authors and affiliations

Science  26 Sep 2003:
Vol. 301, Issue 5641, pp. 1886-1889
DOI: 10.1126/science.1087236


Unexpectedly high concentrations of ultrafine particles were observed over a wide range of latitudes in the upper troposphere and lower stratosphere. Particle number concentrations and size distributions simulated by a numerical model of ion-induced nucleation, constrained by measured thermodynamic data and observed atmospheric key species, were consistent with the observations. These findings indicate that, at typical upper troposphere and lower stratosphere conditions, particles are formed by this nucleation process and grow to measurable sizes with sufficient sun exposure and low preexisting aerosol surface area. Ion-induced nucleation is thus a globally important source of aerosol particles, potentially affecting cloud formation and radiative transfer.

Atmospheric aerosols affect climate directly by altering the radiative balance of the Earth (1) and indirectly by acting as cloud condensation nuclei (CCN) (2), which in turn change the number and size of cloud droplets and the cloud albedo. Homogeneous nucleation (HN) (formation of solid or liquid particles directly from the gas phase) is an important source of new particles in the atmosphere (3, 4), but the process is poorly understood and alone is unable to explain the observed particle formation. Homogeneous nucleation includes binary homogeneous nucleation (BHN) of sulfuric acid–water (H2SO4-H2O) (3, 4) and ternary homogeneous nucleation (THN) (or multicomponent homogeneous nucleation) of sulfuric acid–ammonia–water (H2SO4-NH3-H2O) (5, 6). BHN has been used to explain nuclei-mode (∼5- to 100-nm particle diameter, Dp) aerosol concentrations in the tropical upper troposphere (7). In certain locations, such as the marine boundary layer and continental areas, particle formation rates are observed to be much higher than that predicted by BHN (810), and some of these discrepancies have been explained by THN (6). Because ammonia acts to stabilize the critical embryos (∼1 to 2 nm), the nucleation rate of THN is higher than the BHN rate. However, there are large uncertainties in the predictions of classical HN theory because it is on the basis of the liquid drop model (11), which is not valid for small molecular clusters. In addition, classical theory and parameterizations for HN have not been tested experimentally at the conditions of the upper troposphere and lower stratosphere (UT-LS).

Nucleation involving background gas-phase ions generated by galactic cosmic rays (GCRs) is another potentially important atmospheric nucleation process (12, 13). The thermodynamic advantage of ion-induced nucleation (IIN) is an enhancement in the stability of electrically charged clusters and higher particle growth rates because of electrostatic forces (14). Model studies of IIN have been made (1315), but these calculations have used poorly constrained parameters and contained considerable uncertainties because of a lack of data on the thermodynamic properties of charged molecular clusters. Several factors favoring IIN exist in the UT-LS, including relatively high ion-production rates by GCRs (up to ∼20 to 30 pairs of ions cm–3 s–1) (16), low temperatures, and relatively low surface areas of preexisting aerosol. However, there are few direct measurements supporting aerosol formation by IIN in the UT-LS. Recent measurements by an airborne ion mass spectrometer in the upper troposphere showed large positive-ion clusters (17), and other atmospheric measurements indicated that negative bisulfate core ions are abundant in the stratosphere and troposphere (18, 19).

Here, we show aerosol measurements that confirm IIN model predictions of new particle formation. Size distributions with Dp from 4 to 2000 nm and meteorological variables were measured on 56 flights of NASA aircraft DC-8, ER-2, and WB-57F during 1998 through 2000 at altitudes from 7 to 21 km and latitudes from 10° to 90°N (20). OH and SO2 were also measured on the DC-8. Backward trajectory calculations provided meteorological histories and fractional sun exposure (21) for sampled air parcels. The IIN model incorporates measured thermodynamic data for sulfuric acid–water ion clusters (20, 22) and is constrained by observational data for key species and parameters (20): OH and SO2 concentrations, surface area of preexisting aerosol, temperature, relative humidity (RH), and calculated fractional sun exposure.

High concentrations of ultrafine particles (Dp < 9 nm) were observed at altitudes from 7 to 13 km at mid- and high latitudes in November 1999 through March 2000 on the DC-8 (Fig. 1). Overall, 16% of the size distributions measured on 20 flights had statistically significant evidence of newly formed particles. Cases of recent new particle formation were identified by two factors: (i) number concentrations with Dp from 4 to 6 nm exceeded those with Dp from 6 to 9 nm, and (ii) number concentrations at Dp of 4 to 9 nm (N4-9) exceeded 1 cm–3. These particles were formed over 2 to 5 days, as discussed below. In the identified cases of recent particle formation, N4-9 on average contributed 3.3 × 102 cm–3 to the total concentration (N4-2000) of 5.2 × 102 cm–3. The observed ultrafine particles were located in the UT-LS, according to the measured tropopause heights. The other 84% of the measurements that showed no evidence of recent new particle formation had N4-2000 of 2.4 × 102 cm–3 and N4-9 of 60 cm–3. These size distributions were usually bimodal, with peaks at Dp from 10 to 20 nm and from 50 to 200 nm (Fig. 1). The peak at 50 to 200 nm is the typical aerosol mode observed in well-aged stratospheric air. The peak at Dp of 9 to 20 nm indicates relatively aged nuclei-mode particles formed over 5 to 7 days.

Fig. 1.

Mean size distributions for cases satisfying the criteria for recent new particle formation: mid- and high-latitude UT/LS (7 to 13 km), tropical troposphere (7 to 17 km) and high-latitude stratosphere (17 to 21 km). Results from a simulation of the IIN model after 2-day nucleation evolution are shown for a comparison with the mid- and high-latitude UT/LS case. The model uses ∼80% of the measured peak noontime PH2SO4 and the other average conditions observed for samples showing the feature of new particle formation (table S1). (Inset) The average size distribution at the mid- and high latitudes for samples showing no recent particle formation.

We present here two case studies of measurements of high and low number concentrations of nanometer particles and compare each measurement with numerical simulations of the IIN model (23) using the observed atmospheric conditions (table S1). The first case study focuses on the particle-size distributions measured on 25 January 2000 at midlatitudes (Fig. 2A). On this day, extremely high number concentrations of particles with Dp < 9 nm were observed. The averaged N4-2000 exceeded 1.3 × 103 cm–3, with a maximum at Dp < 9 nm and a much smaller peak at Dp ∼ 40 nm. The average N4-9 was 9.8 × 102 cm–3. The predicted evolution of newly formed particles by IIN as a function of time for a 5-day period is illustrated in Fig. 2A. The measured particle concentrations with Dp from 4 to 6 nm match best the simulated size distributions after 2 to 3 days, and particles with Dp from 6 to 9 nm match those simulated after 4 to 5 days. A 7-day backward trajectory calculation shows that the air mass containing the recently formed particles traveled from the polluted North American continent and stayed until very recently at midlatitudes (fig. S1). Because several intensive rainfall events occurred 4 days prior, the air parcel had very low surface area (<1 μm2 cm–3) due to precipitation scavenging. Furthermore, the air parcel was uplifted from 3.5 to 5 km to ∼10 km and the temperature dropped from 270 K to <240 K, triggering nucleation 2.5 days before the time of measurement. This is consistent with the model predictions that these particles with Dp from 4 to 6 nm were formed 2 to 3 days prior. Calculations of IIN at 270 K and 60% RH, however, simulated no significant nucleation. These results indicate that low surface area and low temperature are important factors required for production of ultrafine particles by IIN, depending on PH2SO4 (H2SO4 production rate) as shown below.

Fig. 2.

Comparison of measured and simulated particle-size distributions for two cases: high and low ultrafine particle production. (A) Particle number-size distributions measured over 18 minutes on 25 January 2000 at 11.2 km, latitudes from 59°N to 60°N, and longitudes from 4°E to 6°E (blue circles). Particle size distributions as a function of time as simulated by the IIN model (black curves). The model uses a peak noontime PH2SO4 of 300 cm–3 s–1, corresponding to [OH] of two-thirds of the measured value and a fractional sun exposure of 0.25. Other input parameters, including a background particle mode, were as measured in flight (table S1). The [H2SO4] derived from the model is ∼1 × 106 cm–3. (B) Particle-size distributions measured over a 12-minute period on 10 December 1999 at 12.5 km, latitudes from 67°N to 70°N, and longitudes from 19°E to 22°E (red triangles). Particle-size distributions as a function of time as simulated by the IIN model, initialized with parameters measured aboard the aircraft (table S1), (black curves).

The second case study was conducted for a period when N4-9 was very low, on 10 December 1999 in the Arctic (Fig. 2B). The average N4-2000 was 40 cm–3, with modes at Dp from 10 to 20 nm and from 50 to 200 nm. Backward trajectory calculations show that the air parcel had traveled across the Canadian Arctic with little exposure to sunlight over the previous 7 days (fig. S2). The measured [SO2], surface area, and temperature were comparable to the case of 25 January 2000, but the RH was very low, <2% (table S1). Simulations of IIN for these conditions produced a very low number of ultrafine particles even after 7 days (Fig. 2B), consistent with the measurements. The absence of ultrafine particles is apparently because of the low sun exposure (and accordingly low PH2SO4) and low RH.

The occurrence of recently formed particles with Dp from 4 to 6 nm (N4-6) is a sensitive function of sun exposure and the competition from scavenging by preexisting particles (Fig. 3). When the ratio of sun exposure to the surface area (Rss) was high, substantial N4-6 was observed. On the other hand, with insufficient sun exposure and/or high surface area (small Rss), N4-6 was measured to be low, often near the detection limit (∼0.01 cm–3). This behavior is quantitatively reproduced by the IIN model simulations (Fig. 3). The correlation of ultrafine particle concentrations with Rss illustrates that sunlight and surface area are critical factors controlling new particle formation. Most of our measurements were made at values of Rss from 0.1 to 0.2, and within this range the measured N4-6 varied over four orders of magnitude, depending on PH2SO4, temperature, and RH, as simulated by the model. Model simulations using the range of measured surface area, fractional sun exposure, and prescribed peak noontime PH2SO4 values ranging from 50 to 1600 cm–3 s–1 show a trend and range very similar to the measurements. This consistency between the observations and the simulation strongly suggests that IIN contributes substantially to new particle formation in the UT-LS. The model simulation also illustrates several important features. First, at low PH2SO4 [such as 50 cm–3 s–1, equivalent to 25 parts per trillion by volume (pptv) of [SO2] and 0.04 pptv of midday [OH], for example], ultrafine particles are formed by IIN only when Rss > 1, corresponding to long sun exposure and low preexisting particle surface area. Moreover, when PH2SO4 increases, the required minimum sun exposure decreases, and the allowable preexisting surface area increases for production of the same number of new particles.

Fig. 3.

Concentrations of measured (gray dots) and simulated (symbols) particles with Dp from 4 to 6 nm as a function of the ratio of average fractional sun exposure to particle surface area (Rss). In this data set, which incorporates 5300 individual measurements in the mid- and high-latitude stratosphere (potential temperature >330 K) taken on the DC-8, the average fractional sun exposure of air parcels during the prior 5 days was 0.31 ± 0.19 (1σ, error bars), or about 7 hours of daylight time per day. The particle surface area was 3.4 ± 1.7 μm2 cm–3. The black vertical bars show approximate uncertainties in statistical counting at different particle-number concentrations. Simulated conditions were 207 K, 162 torr, 20% RH, with a range of 3 to 9 daylight hours per day and 0.2 to 8.3 μm2 cm–3 preexisting aerosol surface area. Peak noontime PH2SO4 values shown in the legend are consistent with the measured range of [OH] and [SO2] (table S1).

The frequency and magnitude of new particle formation was higher in the tropics than in high latitudes and was higher in the UT than in the LS (Fig. 4). Measurements in the tropical troposphere in 1998 and 1999 showed high concentrations of recently formed particles (Fig. 1); the average N4-2000 was 2.1 × 103 cm–3, with N4-9 contributing 1.2 × 103 cm–3. Even in the high-latitude stratosphere at altitudes of 17 to 20 km (>∼6 km above the tropopause), new particle formation events were observed in 6% of the particle-size distributions measured in January through March 2000. In these cases, the measured N4-9 was ∼3 cm–3, which is notable compared to the average N4-2000 of ∼15 cm–3 (Fig. 1). The simulated size distribution by IIN was consistent with those measured in the high-latitude stratosphere, within the uncertainties of the model output, indicating that particle formation by IIN takes place in both the UT and LS.

Fig. 4.

Statistics of samples in which criteria identifying recent new particle formation were satisfied, measured at latitudes of 10°N to 90°N and altitudes of 7 to 21 km from 1998 to 2000 on the WB-57F, DC-8, and ER-2. (A) Tropics, extratropics and midlatitudes, and high latitudes. (B) Overall UT and LS. Of all measured size distributions, 20, 25, and 11% of those in the tropics, extratropics and midlatitudes, and high latitudes, respectively, satisfied the criteria; the average N4-9 were 1.1 × 103, 2.8 × 102, and 2.0 × 102 cm–3, respectively. Of all measured size distributions, 35 and 14% of those in the UT and LS, respectively, satisfied the criteria; the average N4-9 in this subset were 4.3 × 102 and 70 cm–3, respectively.

The overall frequency of occurrence of samples indicating recent new particle formation at latitudes from 10°N to 90°N and from 7 to 21 km ranged from 11 to 35%, whereas N4-9 varied from 70 × 103 to 1.2 × 103 cm–3 (Fig. 4). For the remaining measurements, there were still consistently high concentrations of particles with Dp from 9 to 20 nm (N9-20) (Fig. 1), ranging from 60 × 103 to 1.1 × 103 cm–3. These particles have a high potential to grow to CCN sizes (∼50 to 100 nm). Our observations are also consistent with other studies that have measured high number concentrations of particles with Dp from 4 to 12 nm (24) or from 7 to 18 nm (25) at altitudes from 8 to 11 km in the tropics and the midlatitudes. The IIN calculations, using the mid- and high-latitude UT-LS conditions and having a new particle-formation feature (table S1), simulated similar particle-size distributions as those measured (Fig. 1) (26). Although we are aware of no size-distribution measurements for ultrafine particles in the Southern Hemisphere, from the general conditions required for IIN (Fig. 3) we speculate that new particle formation is a common phenomenon occurring on a global scale.

The simulations of IIN using constrained thermodynamic data and observations of key atmospheric species are in excellent agreement with measurements over a wide range of UT-LS conditions. However, we do not rule out the possibility of HN; if [H2SO4] is sufficient, given the very low temperatures in the UT-LS, BHN may be effective (4, 27). It has also been suggested that atmospheric dynamics can affect BHN rates in the upper atmosphere (28). To make a quantitative evaluation of HN, however, thermodynamic data of HN under UT-LS conditions are required. For THN, the growth of clusters depends strongly on the availability of condensable ternary vapors (such as ammonia and oxidized organics) (6); contribution of THN in the mid- and high-latitude UT/LS will not be important because of the low concentrations of these species (29, 30). These compounds are more abundant in the tropical troposphere (31); however, because ultrafine particle concentrations measured in the tropical troposphere are consistent with the prediction of the IIN model, we believe that IIN is efficient in the tropics as well. It is unlikely that organic ions will influence the negativeion chemistry, but there is a potential for ion-ion recombination of large organic ions (12). Because newly formed particles in the tropical UT are a likely source of stratospheric particles (7), the IIN process plays an important role in the identification of the origin of stratospheric aerosol. However, our study also indicates that IIN provides an important local source of new particles in the UT-LS from the tropics to the polar regions throughout the year when there is sufficient sun exposure and low aerosol surface area. It has also been suggested that ions produced by cosmic rays can induce nucleation of particles that may grow into CCN and modify cloud properties (32). New particle formation by IIN in the UT-LS thus has important implications for correctly assessing the indirect effects of aerosol particles on the global climate.

Supporting Online Material

Materials and Methods

Figs. S1 to S3

Table S1


References and Notes

View Abstract

Navigate This Article