Introduction

The microphysical and radiative properties of marine stratocumulus clouds are very susceptible to changes in CCN availability, rendering it one of the most important aspects of the aerosol-cloud-climate system1,2,3,4. New particle formations and subsequent growth, therefore, have a huge impact on the radiative budget in the marine boundary layer (MBL)5. Recent studies have advanced our knowledge of the new particle formation in the MBL6,7,8,9,10,11,12. However, the growth of these new particles into CCN sizes remains rather unclear and unresolved, despite such growth being observed in various oceanic regions including the North Atlantic and Southern Ocean13,14,15,16. It has proven challenging to elucidate the dominant mechanism and species driving the growth mainly because of such small amounts of mass at such small sizes17,18,19. Sulfur-based and organic sulfur-based acids (e.g. sulphuric acid and methanesulphonic acid (MSA), both derivatives of dimethyl sulfide derived from phytoplankton released products) were considered the major species evolving in the particle growth in the MBL20,21,22. There is a growing body of research suggesting that sulfur-based species are insufficient to account for the observed abundance of particles, stressing the importance of organic vapours as more likely condensing species14,23,24,25.

Indirect measurements such as volatility26 and hygroscopicity27,28,29, have provided valuable information on the properties of some of the condensable gases and suggested the potential contribution of organic vapours during such events. However, conclusive and direct evidence is still lacking in terms of what is driving the growth. Direct measurement of Aitken mode particles over particle growth events in MBL is rare.

This study aims to fill this knowledge gap by deploying high-resolution time-of-flight aerosol mass spectrometry (HR-ToF-AMS), humidified tandem differential mobility analyser (HTDMA), and scanning mobility particle sizer (SMPS) to track the change of chemical composition of Aitken mode particles. We focus on ultrafine SMA growth events in the clean and pristine marine air mass that originated from the open ocean to reveal the chemical composition of the condensing gases. Our results show that the condensation of organic vapour and sulphuric acid from biogenic activities is the major driver of ultrafine SMA growth in a clean marine atmosphere.

Results and Discussion

Direct measurement of secondary marine aerosol growth events

Long-term ambient measurements of aerosol physical and chemical properties were conducted from 2008 to 2018 at Mace Head Atmospheric Research Station (MHD), on the west coast of Ireland30,31,32. Using an unsupervised machine-learning-based clustering technique, we categorized the hourly aerosol number size distribution at MHD from 2008 to 2018 into four classes (See Supplementary Fig. 1 and Methods). The product of ultrafine SMA growth product (ultra-fine particle growth16), was found to account for over 60% of the summertime clean marine air masses at MHD, stressing the great importance of such SMA growth process. Such ultra-fine particle growth events represent different stages of a growing nucleation mode of the order of 10–15 nm into a developed Aitken mode up to 50 nm in diameter16. The source of such particles is likely to be in the open ocean off shore assuming a constant growth rate14.

More than a hundred such events were identified over a 10-year period (Supplementary Table 1). These events were characterised by a steady growth of newly formed Aitken mode particles from sizes as small as 10–30 nm to 40–80 nm, with a median growth rate of 1 nm hour−1. The duration of these observed growth events ranged from 6 h to 48 h, defining the spatial scale as regional phenomenon spanning scales of 1000 km to 2000 km. The events were observed in air masses advected from different marine air mass sectors15 but most frequently in marine polar air masses, which were considered as a biological active oceanic region33.

To closely examine the ultrafine SMA growth mechanism, the evolution of aerosol number size distribution observed on 9th June 2012, as presented in Fig. 1 is a typical example. During the event, the number size distributions of aerosol were typical marine bimodal distribution. The growth event was observed with an Aitken modal diameter of 35 nm. The modal diameter increased near linearly over 24 h to about 60 nm, corresponding to a growth rate of 1.0 ± 0.1 nm hour−1 (Fig. 1a). In fact, the number size distribution extracted at different growth stages (start, middle, and end) demonstrated that the Aitken and accumulation modes were growing simultaneously (Fig. 1b). During the event, the average black carbon (BC) mass concentration was 6.3 ± 5.2 ng m−3 (mean ± standard deviation) and the averaged bulk PM1 concentration was 0.72 ± 0.14 μg m−3. The bulk submicron aerosol mass was dominated by non-sea-salt sulfate (nss-SO42−), which accounted for 60 ± 7% of the total mass, followed by organic matter (Org, 20 ± 3%). The methanesulfonic acid accounted for a small fraction of the mass (4 ± 1%). The degree of neutralization was 0.22 ± 0.04, confirming the acidic nature of the aerosol, and sulfate mainly existed as sulphuric acid (about 60%) and ammonium bisulfate (about 40%) based on the simplified ion-pairing scheme34.

Fig. 1: An example of ultrafine secondary marine aerosol (SMA) condensational growth event at MHD, 9th June 2012.
figure 1

a The time series of aerosol number size distributions, the arrows represent the defined growth stages: start, middle, and end phases; the grey line represents linear regression of Aitken mode diameter against growth time; b the average aerosol number size distribution for different growth stages; c the particle mass size distribution of different species at various stages, the bars represent raw data, and the lines represent two modal lognormal distribution fitting, the fitted Aitken modal diameter were labelled for each panel; d the mass fraction of chemical species contributing to the growth of Aitken (left) and accumulation (right); e the averaged volume size distribution retrieved by SMPS (black dashed line) and HR-ToF-AMS (coloured area), the grey area represent 95th confidence interval of SMPS volume size distribution.

Full size image

The particle time-of-flight (pToF) mode of HR-ToF-AMS provides direct quantification of the evolution of the chemical composition of the Aitken mode. The averaged mass size distributions for each stage are shown in Fig. 1c. At the start of the event, the bimodal mass size distribution consisted of the Aitken and the accumulation mode which were separated. The modal diameters of the Aitken mode particles ranged from 49 to 55 nm for all the species. The similarity in modal diameters between different chemical species suggests the internal state of mixing35,36. Along with the particle growth to the middle and end phases, the Aitken mode position of both organics and sulfate grew to about 64 to 70 nm. The changes in modal diameters for all the species have confirmed that methanesulfonic acid, organics, and partially-neutralised sulphate participated in the Aitken mode growth simultaneously. The volumes of Aitken mode of each chemical species increased from the start to end phase, indicating that the condensation of gases, rather than coagulation, was driving the growth.

By comparing the mass size distributions of the start and end stages, the contribution of organics, methanesulfonic acid, NH4+, and nss-SO42− to Aitken mode growth was 60%, 9%, 6%, and 25%, respectively (Fig. 1d). The larger mass fraction suggested the important role of organics in condensational growth. The agreement between SMPS and pToF-derived size-resolved volume distribution (Fig. 1e) lends further confidence that the condensation of organic vapour, sulphuric acid with the smaller contribution of ammonium, and methanesulfonic acid explains most of the SMA growth.

We expected the initial clusters originated from the open ocean because of the following reasons: (1) the heights of air mass backward trajectories mostly within the marine boundary layer. (2) During the events, the black carbon concentrations were very low, indicates extremely clean air mass arriving at Mace Head. (3) Aerosol growth over several hours normally represents regional phenomenon17. (4) the minor contribution from coastal alage-derived new particle formation. The impact of coastal algae-derived iodine at Mace Head is expected to be minor during this growth event because (1) coastal nucleation typically occurs rapidly over much shorter timescales and is readily detected with much higher particle number concentrations (over 104 cm−3) of sub 10 nm particles (Supplementary Fig. 1); (2) the uniform growth rate of this event was modest at 1 nm hour−1, which is similar to a growth event observed during a Northeast Atlantic Ocean cruise37; (3) coastal nucleation is typically observed during low tide, and this event has no tidal pattern; (4) the continuous SMA particle growth over many hours is usually attributed to regional-scale particle formation, which is the open ocean in this case. In summary, the coastal events are readily distinguishable from the open ocean. The contribution of sea spray at the beginning of the events is considered to be negligible, as no strong sea-salt hygroscopicity mode was found at the start and end of the growth events. (Supplementary Fig. 2).

As indicated by Fig. 1d, the contribution of chemical species differs slightly between the Aitken and accumulation mode particles. Furthermore, accumulation mode particles are also subjected to cloud processing which affects their chemical composition. Therefore it is important to evaluate the chemical composition of the Aitken mode directly. We further analysed 15 ultrafine SMA growth events with distinguishable pToF data signals between the start and end stage of the event (Fig. 2a and presented individually in Supplementary Fig. 3). We found that on average organic vapours and sulphuric acid accounted for 55% and 35% of the mass, respectively. The contribution of organic vapour is significantly higher in Aitken mode than accumulation mode (Supplementary Fig. 4), highlighting the value of direct measurement of Aitken mode chemical composition with high temporal resolution.

Fig. 2: The statistical summary of the contributions of chemical species to ultrafine secondary marine aerosol (SMA) growth events.
figure 2

a Distribution of relative mass contribution to Aitken mode resolved by the pToF mode of HR-ToF-AMS. The vertical lines represent median, the boxes represent 25th and 75th quantile, and the whiskers represent 1.5 inter-quarter ranges; b the distribution of hygroscopicity parameter of the condensable gas (κchange) derived by HTDMA, the grey bars represent the number of events of each κchange bins and the curves represent probability density functions; c the relationship between κchange and the contribution of sulphuric acid and organic vapour under different NH4+ and methanesulfonic acid (MSA) scenarios based on volume-mixing rule(See Methods), the colours represent calculated κchange, the size of markers represent the number of observations with specific κchange bins.

Full size image

However, it should be noted that the measurement of Aitken mode particles using HR-ToF-AMS is subjected to a small signal-to-noise ratio (Methods), and usually requires large and strong Aitken mode particles at sizes smaller than 100 nm. We further calculated the hygroscopicity parameter change of Aitken mode induced by condensed species by comparing the measured hygroscopicity before and after the growth28. Briefly, the change of Aitken mode hygroscopicity was assumed to be the sole result of condensational growth. The change of Aitken mode particle volume was attributed to the condensational gases. An example of estimating κchange is given in Supplementary Fig. 5. For all the SMA growth events, the κchange was dominated by values ranging between 0.4 and 0.7 (Fig. 2b) with a median value of 0.52, which is a typical value of organic and sulphuric acid mixture considering the ammonia-poor nature of the North-east Atlantic.

To assess the consistency between the chemical composition of Aitken mode particles from pToF and the hygroscopicity parameters derived from HTDMA measurements, we conducted a closure test. Since the hygroscopicity measurements cannot reveal the contribution of MSA and NH4+, we considered three scenarios. As shown in Fig. 2c, when MSA and NH4+ are assumed to have no contribution, sulphuric acid accounts for 50–70% of the mass changes, with the remainder coming from the organic vapours. As the contribution of MSA and NH4+ increase to 5% and 2.5%, respectively (within the range obtained by HR-ToF-AMS Aitken mode calculation) the contribution of organic vapour increased up to 50%, in good agreement with HR-ToF-AMS derived organic contributions. This successful closure further supports the importance of condensation of organic vapours in SMA growth.

This analysis supports the previous inference that the condensation of the organics and sulfate dominates the ultrafine SMA production26,28, but further identifies the contribution of NH4+ and methanesulfonic acid, which could not be resolved by hygroscopicity measurements alone. The results also agree with a recent study that observed multiple species, including sulfate, methanesulfonic acid, and secondary organics, contributing to Aitken mode composition during intensive aerosol formation events in the Arctic38. Furthermore, the median contribution of organic vapour, summarised by the AMS Aitken mode measurement, is somewhat higher than the previous mesocosm experiment39 conducted in an artificial ocean-atmosphere simulator over the course of phytoplankton bloom with oxidative flow reactor, suggesting a better understanding of particle formation and growth over the open ocean necessitates more observational and theoretical studies.

The organic vapour contributing to condensational growth is likely of marine origin rather than terrestrial as the data were filtered by black carbon (<15 ng m−3) and backward trajectories (no land contact for three days, Fig. 3a, b). Most of the SMA growth events were observed from April to September with the highest frequency in June (Supplementary Figs. 6–7). The monthly distribution of these events shared a similar trend with Northeast Atlantic biological activities (Supplementary Fig. 7b) based on the commonly used proxy of chlorophyll-a concentration over the North Atlantic. The backward trajectories (Fig. 3a, b) suggest that these SMA particles were likely to be formed within the MBL rather than being entrained from the free troposphere, as indicated by the height of trajectories (over 95.4% of air mass is lower than 500 m, which is a typical height of the marine boundary layer (500 m), this is similar to the previously reported marine new particle formation observed in the Southern Ocean13.

Fig. 3: Geographical distribution and meteorological conditions of the SMA growth events.
figure 3

a Relative frequency of air mass backward trajectory height; b relative frequency of air mass backward trajectory path before arriving at MHD during SMA growth events; c evolution of particle number size distribution as a function of the exposure to chlorophyll-a.

Full size image

We analysed summertime clean marine aerosol number concentration from 2008 to 2018 and their relationship with air mass exposure to chlorophyll-a, which is a commonly used proxy for marine biological activities. The air mass exposure to chlorophyll-a was obtained by overlaying the air mass backward trajectories and satellite-derived chlorophyll-a concentration over the oceans. For all the data collected in the clean marine sector, as shown in Fig. 3c, the number concentration, particularly of Aitken mode particles, increased with the chlorophyll exposure over the air mass history, pointing to the impact of phytoplankton to the aerosol formation. This is also true when comparing the exposure to chlorophyll-a between ultrafine SMA growth events and climatological values (calculated from all data in the clean marine sector), the values for growth events were significantly higher (p-value < 0.001).

Implications for aerosol-cloud interactions

The formation and growth of SMA represent a potentially essential source of CCN in the MBL35. The impact of a typical growth event on CCN concentration was evaluated. While the CCN concentration at the lowest supersaturations (0.1% and 0.25%) was relatively insensitive to condensational growth, the CCN number at higher supersaturations increased with four-fold increases over 8 h of condensational growth(e.g. from ca. 100 to 400 cm−3 at supersaturation of 0.5%, Fig. 4). During all the growth events, the CCN number concentration increased by 25 cm−3 at SS from 0.2% to 0.3%, corresponding to a 32% to 65% increase compared to the background (Supplementary Figs. 8, 9). Demonstrating the particle formed from growth events in the North Atlantic can contribute to marine cloud droplet numbers at a very modest SS of 0.25%. This effect is even more pronounced if higher SS levels are considered, when CCN concentration increased by 53 cm−3 at 0.3% to 0.5%, corresponding to a 72% increase in concentration compared to the start stage. The analysis indicates that such SMA growth is an important pathway to replenish the CCN in the remote marine environment; therefore, it plays a key role in regulating the aerosol-cloud interactions and the radiative budget in the MBL.

Fig. 4: Enhancement of CCN number during SMA growth event.
figure 4

a The time series of aerosol number size distribution, the circle markers represent the fitted modal diameter; b the time series of CCN number concentration enhancement (ΔCCN) during the SMA growth events, coloured by the level of supersaturation.

Full size image

The growth mechanism of SMA has been elusive since the proposal of the marine-aerosol-cloud feedback hypothesis20. This study provides direct evidence that the growth of SMA in the clean marine atmosphere occurs as a result of the condensation of organic and inorganic vapours. This is also in line with previous observations that summertime marine aerosol is internally mixed by organics and sulfate. A better appreciation of the SMA growth and production is likely to improve our understanding of the aerosol-cloud interactions in the marine atmosphere.

Methods

Mace Head station

Measurements were conducted at Mace Head Atmospheric Research Station (MHD) on the West coast of Ireland (www.macehead.org, last access 19th April 2023) from 2008 to 2018. MHD is exposed to marine air masses considered among the cleanest in the Northern Hemisphere with the least anthropogenic influences after several days of advection over the open ocean32.

In situ instrumentation

A scanning mobility particles sizer (SMPS, TSI) and a nano-SMPS were used to measure aerosol number size distribution ranging from 20 to 500 nm and from 3.5 to 20 nm, respectively. A triple-mode lognormal distributions was used to fit the aerosol number size distribution, comprising nucleation, Aitken, and accumulation modes.

The bulk PM1 chemical composition and mass size distributions including organic matter, ammonium, non-sea-salt sulfate, nitrate, and methanesulfonic acid were measured using an HR-ToF-AMS (Aerodyne Research Inc.). The size-resolved chemical compositions in vacuum aerodynamic diameter were obtained in Particle Time-of-Flight (pToF) mode, during which the instrument performed particle sizing based on particle time of flight with the aid of a chopper40. Operational details of MHD HR-ToF-AMS can be found in previous papers41. The HR-ToF-AMS data were processed using the PiKa 1.31 package based on Igor Pro 7.08 software (Wavemetrics). The composition-dependent collection efficiencies were applied42.

Standard AMS data processing approaches were used to integrate the ion signals for unit mass resolution data (V-mode and pToF mode) and high mass resolution data (W-mode). In pToF mode, the baseline for each m/z channel is established by averaging the ion signal in two defined time regions (DC markers) at the very beginning and the end of the chopper cycle. These correspond to particle velocities beyond the upper and lower transmission capabilities of the instrument and therefore represent background ion contributions.

For m/z channels susceptible to gas-phase species interference within the first DC marker region (e.g. m/z 15 or m/z 44), only the second DC marker was utilized for establishing the PToF baseline. In this work, the DC markers spanning m/z 12–150 were evaluated on an individual basis for each growth event to determine the optimal DC marker pair for calculating the baseline across all m/z channels. The integrated particle mass concentration of per PToF mode run was calculated by normalising to the V-mode concentration of the same time step. The particle size distribution obtained during regular particle-free time periods (ambient air sampled through HEPA-filter bypass) was subtracted from the ambient size distribution obtained during the growth events.

In this study, the lognormal peak fitting was fitted to each event-average AMS mass size distribution employing Multipeak Fit V2 algorithm in Igor Pro (Wavemetrics) using manual assigned initial guesses on peak position, height width, and standard deviation. The uncertainty of mass size distribution is size-bin dependent due to the cubic relationship between particle mass and particle diameter and the improvement in signal-to-noise with increasing particle size.

The mass concentration of black carbon (BC), used as an anthropogenic tracer, was measured by a multi-angle absorption photometer (MAAP, Thermo Fisher 5012). An HTDMA43,44 was used to measure aerosol hygroscopic growth at a fixed relative humidity of 90% for aerosol with selected dried sizes of 35, 50, 75, 110, and 165 nm. The HTDMA used in this study has been described in great detail in previous studies36,45. The growth factors measured by HTDMA were inverted using a piecewise linear function45 and converted to hygroscopicity parameter κ46, assuming the surface tension of water-air interface. CCN number concentration was measured using a continuous-flow streamwise thermal gradient CCN counter (CCN-100, Droplet Measurement Technologies, USA)47. The CCN-100 operates at a flow rate of 0.5 L min−1, which is then separated into a wetted zero-air sheath flow and sample flow at a ratio of 10. Aerosol is drawn into a temperature-regulated wetted vertical column, which creates supersaturated humidity conditions through its centreline proportional to the applied vertical temperature gradient along the column wall. Supersaturation levels were calibrated using ammonium sulphate particles48.

Data analysis and processing

All measurement data were averaged hourly. Data were filtered to meet MHD clean sector criteria representing the pristine clean marine environment32. Only the data with BC less than 15 ng m−3, and wind direction ranging from 190° to 300° were used(32, 42). The 96-hour air mass backward trajectories were calculated every three hours for 100 m above ground level using the Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) model49 based on Global Data Assimilation System (GDAS) with a one-degree resolution. Downward radiative flux is also retrieved from GDAS. Only the events lasting longer than 6 h with valid HTDMA or HR-ToF-AMS and SMPS data were selected for further analysis.

The aerosol growth rate (GR) was determined by the rate of change in the geometric mean diameter of the growing mode by using the following equation50:

$${GR}=\frac{d{P}_{p}}{{dt}}$$
(1)

where the dPp is the change in the geometric mean diameter of the growing Aitken mode, and the dt is the duration of the growth event.

The estimation of the hygroscopicity parameter change induced by condensable gas (κchange) followed the procedure28. The hygroscopicity of the condensing gas was tracked by the change in particle hygroscopicity. The condensable gaseous species were inferred by taking advantage of the varying hygroscopicity of candidate gases. Briefly, the change in particle diameter was assumed to be driven by the condensational growth only, and the aerosol hygroscopicity followed the volume-weighted mixing law. The κchange is then given by:

$${\kappa }_{1}=\left({\kappa }_{{change}}-{\kappa }_{0}\right)\times V{F}_{{gas}}+{\kappa }_{0}$$
(2)

where the κ0 and the κ1 are the measured aerosol κ values before and after the condensational growth, and VFgas is the volume fraction of condensable gas. The κchange is then obtained as the sum of slope and intercept of the linear regression fitting between κ1 and VFgas.

The condensable gaseous species were inferred by taking advantage of the varying hygroscopicity of candidate gases. The mass contribution of organic vapour and sulphuric acid using the derived kappa change is based on the Zdanovskii–Stokes–Robinson mixing rule51 and simplified ion-pairing scheme34:

$${\kappa }_{{change}}={\sum}_{i}{\varepsilon }_{i}{\kappa }_{i}$$
(3)

Where the ki and εi are the hygroscopicity parameter and volume ratio of the ith component of the aerosol, respectively. We assumed the mass contributed of methanosulfonic acid and NH4+ under three different scenarios: 0% and 0%; 5% and 2.5%, 10% and 5%. The presence form of sulfate was based on the amount of NH4+. The density and hygroscopicity parameters can be found52.

Aerosol number size distribution clustering

The aerosol number size distributions were clustered by using the unsupervised machine learning k-means clustering technique. We assigned 12 cluster centers and further segregated them into four categories: ultrafine particle growth event, coastal nucleation, marine background, and anthropogenic emissions15. The clustering was done via “kmeans” function in R programming software53.

Event selection

SMA growth events were selected manually by looking at the evolution of aerosol number size distributions14. To ensure the representativeness of the event to clean marine environments, we used multiple criteria to screen the events. In brief, aerosol started at a diameter of 10 −30 nm and continuously and gradually grew to 40–70 nm in size depending on the duration of the event. Only the events with continuous growth over 6 h were included in this analysis.

Cloud supersaturation calculation

We use Hoppel minimum54 to infer effective cloud supersaturation, following the approach reported55. The Hoppel minimum was identified as the particle diameter corresponding to the local minimum of the number concentration between ~40 and 120 nm, it represents the average size threshold above which particle activated into cloud droplets. The critical supersaturation for Hoppel minimum diameter to be activated, calculated via kappa-Köhler theory46, is considered as the average effective maximum supersaturation in clouds. We analysed aerosol number size distribution data over summer time clean marine air mass arrived at Mace Head. The distribution of Hoppel minimum is about 80 (with 62 to 91, 25th to 75th quantile), which corresponds to maximum cloud supersaturation of 0.25 (0.19 to 0.32, 25th to 75th percentile). The results largely agree with a recent study that simulated the supersaturation at the cloud base below 2 km during summer at the upstream area of Mace Head to be around 0.2% to 0.3%55.

Uncertainty

The uncertainty in the measurement of aerosol size distribution and CCN number concentration is about 10%48,56. The uncertainty in the particle electrical mobility is below 5% for sizes between 40 and 150 nm. The uncertainty in the AMS bulk measurement was about 30–40%, which was mainly arising from the species-dependent collection efficiency. The accuracy of GF–PDF is sensitive to the total number; therefore, GF–PDF uncertainty was 20% for marine cases45. The AMS chemical mass size distributions were averaged over approximately 6–10 h, leading to an improvement in the detection limit. The detection limit of organic matter was about 1 ng m−3 for events averaging several hours35. Uncertainty related the relative ionization of organic matter was well summarized57 The relative ionization efficiency of ambient organic matter used in AMS data processing is 1.4. The uncertainty related to Aitken mode chemical composition measurement is about 10–50% and 1–10% for accumulation mode particles58.

To investigate the impact of PToF uncertainty on the Aitken and accumulation modes, we performed a Monte Carlo simulation (N = 100,000) by adding random noise within measurement uncertainty (50% for Aitken mode and 10% for accumulation mode) to get a robust estimate of the difference between Aitken and accumulation mode with t-test. The distribution of p values obtained from the Monte Carlo and t-test simulations are shown in the Supplementary Fig. 11 below. As shown in Supplementary Fig. 11, the differences in the organic compounds (Org) and sulfate (SO4) between the Aitken mode and accumulation mode are significant (p < 0.05). However, the differences for MSA and NH4 are indeed not significant.