Atmospheric particles were sampled during the EUCFe cruise (August–October 2006, R/V Kilo Moana, chief scientist James W. Murray). This cruise was carried out to study the iron cycle, including atmospheric deposition, in the equatorial and tropical Pacific (Slemons et al., 2009, 2010, 2012; Radic et al., 2011; Labatut et al., 2014). Samples were collected along the cruise track with a small volume collector equipped with 1 µm porosity, 47 mm diameter PTFE membranes, placed in a Millipore^® polycarbonate filter holder. The membranes were pre-cleaned in ultrapure HNO₃ for 2 d and stored in clean plastic Petri dishes. The collector was located on the top deck and equipped with a control system to stop pumping when the wind came from a direction greater than 60° from the bow to prevent ship smoke collection. To protect the samples from rain, the filter support was angled downward and covered with a plastic protector. A flow meter provided information on the pumped airflow: 8 L min⁻¹ for A281 and A284 samples and 28 L min⁻¹ for the eight other samples. Each sample was collected over a duration of 3 d on average for sample size ranging between 9 and 93 m³ (from coastal to open-ocean areas). The sampling locations are reported in Fig. 1. The sampling area is more than 8000 km wide.

Three samples previously published close to the Bismarck Sea and in the equatorial Pacific are reported to enrich the discussion: A269, A266, and A259 (Fig. 1 and Table 3) (Labatut et al., 2014).

The elemental concentrations and iron isotopic compositions were measured at the LEGOS laboratory (Observatoire Midi-Pyrénées, Toulouse, France) in the years 2009 to 2012. The analytical procedure was described by Labatut et al. (2014) and is summarized here. A trace metal clean laboratory, an ISO4 laminar flow hood, high-purity reagents, and acid-cleaned labware were used for all chemical procedures. The particles were totally digested using a mixture of 5 M HCl, 2.1 M HNO₃, and 0.6 M HF at 130 °C. To check that the procedure was quantitative, some filters were digested twice, and no particulate Fe was detected in the second leach. A ⁵⁷Fe–⁵⁸Fe double spike was added to the leachates. 2 % aliquots were taken for multi-elemental concentration determination on a ThermoScientific Element-XR HR-ICPMS. Na, Mg, Al, Ca, Ti, Fe, V, Rb, Sr, Ba, and Pb concentrations were quantified. Fe was purified from the remaining 98 % with an AG^® 1-X4 anionic resin, and its isotopic composition and concentration were measured on a ThermoScientific Neptune MC-ICPMS.

Throughout this article, uncertainties are given at a 95 % confidence level. For the Fe concentration and isotope measurements on the Neptune MC-ICPMS, the total procedural recovery was 93 ± 25 %. The total procedural blank, including contamination from the sampling filter, was 3.0 ng, which was 3.8 % and 14.7 % of the average and smallest sample, respectively. Repeatability was not determined on aerosol samples (due to limited sample sizes) but was quantified during the same measurement sessions from duplicate analyses, including distinct chemical treatments, of four seawater suspended particle samples. It was 4 % and 0.04 ‰ for concentration and isotopic composition, respectively. This repeatability for δ56Fe is better than the long-term external precision of 0.07 ‰ of our measurements, determined from repeated analysis of a secondary isotopic standard (an in-house “ETH-Hematite” standard). The uncertainties characterizing our Fe isotope data are therefore 0.07 ‰ or the internal measurement uncertainty (2 standard errors), when the latter is larger. The iron isotope protocol at LEGOS has been validated through intercalibration and intercomparison exercises (Boyle et al., 2012; Conway et al., 2016) and described in Lacan et al. (2008, 2010, 2021). The in-house “ETH-Hematite” standard displayed an isotopic composition of +0.52 ± 0.08 ‰ (2 SD, n= 81), which was perfectly consistent with the recommended value of +0.53 ± 0.06 ‰ (2 SD, n= 6) (Lacan et al., 2010). We also measured the sediment geostandard GBW 07315 with δ56Fe = +0.04 ± 0.046 ‰. Unfortunately, it is not certified for Fe isotopes, and we could not find Fe isotope values reported in the literature. We still report it here as it could be useful in the future. The trueness of concentrations determined by HR-ICPMS analysis was verified using certified SLRS-5 river water material and GBW 07315 sediment material. The accuracy (trueness and repeatability) of our HR-ICPMS concentration determination was also validated through intercalibration exercises (Yeghicheyan et al., 2013, 2019). Blanks were quantified for Fe only. Based on the latter and assuming a crustal composition, they were estimated for the other elements. This assumption is supported by the lack of contamination discussed in Sect. 3.1 below. This leads to blank levels always lower than 15 % of each sample and all elements, except for Ca, for which it was 11.8 % on average and 35.7 % at maximum.

To identify the origin of sampled aerosols, air mass back trajectories were calculated using the NOAA Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) model (Stein et al., 2015; Rolph et al., 2017). The meteorological data selected were from the Global Data Assimilation System (GDAS). Trajectories were computed at 50 m above ground level with a 7.5 d run time. Aerosol samplings were conducted between 22 August and 12 October 2006. In order to represent spatial and temporal variabilities and to present a synthetic overview, we divided the cruise track into four areas (Fig. 2).

For each area from which back trajectories are simulated, the starting points of back trajectories were chosen as a grid for representativity and clarity purposes. The grid points are not precisely sampling locations, but they are close to them. The starting times were chosen as the central dates between the sampling period of each area (Table 1).

Aerosol iron concentrations ranged from 0.38 ± 0.02 to 7.22 ± 0.28 ng m⁻³ (Table 2). Excluding aerosol sample A266 close to the Bismarck Sea (5.56 ± 0.22 ng m⁻³), concentrations vary from low values (< 1 ng m⁻³) between 140° E and 160° W along the Equator to large values (> 1.5 and < 8 ng m⁻³) in the tropical North Pacific region and between 160 and 140° W along the Equator. There was no correlation between distance from land and concentration. A major volcanic eruption of Tavurvur (Papua New Guinea) occurred on 7 October 2006 (Wunderman, 2006). Samples A233 to A269 were collected prior to this event and are therefore unaffected. While it is theoretically possible that samples A281 and A284 could have been influenced by the eruption, they were collected over 1500 km away from the volcano. A simulation of the forward trajectory of air masses confirms that samples A281 and A284 were not affected by the eruption (Figs. A1 and A2). Additionally, their concentrations are consistent with those of samples collected before the eruption, confirming that they were not impacted.

Aerosol Fe concentrations in EUCFe samples are consistent with the literature in the central equatorial Pacific for particulate Fe: 2.01 ± 1.56 ng m⁻³ (2 SD, n= 11) (GEOTRACES GP15 cruise: between 20° N and 20° S and along the 152° W meridian) (Marsay et al., 2022) and 5.60 ± 5.65 ng m⁻³ (2 SD, n= 8) (P16 cruise of the CLIVAR/CO₂ Repeat Hydrography Program: between 9° N and 2° S and along the 151° W meridian) (Landing et al., 2013). The range of EUCFe values was also similar to concentrations in Alaskan coastal and pelagic regions in the subarctic North Pacific, in the North Pacific, and in the South Pacific (Buck et al., 2019; Kurisu et al., 2021, 2024; Marsay et al., 2022; Sakata et al., 2022). EUCFe data are lower than aerosol iron concentrations reported in the coastal northwest Pacific, closer to industrialized areas (Kurisu et al., 2021; Sakata et al., 2022).

The concentrations of the major elements of seawater (Na, Mg, Ca, Sr) depend on the height of sampling, wave height, and wind intensity (Bruch et al., 2021; Madawala et al., 2024). Therefore, comparing Na, Mg, Ca, and Sr concentrations in EUCFe samples with those measured in other Pacific samples is not meaningful. However, we can compare Al, Ti, V, and Pb elements with the literature. Their concentrations are in the same order of magnitude as those found previously in the atmosphere over the North Pacific (Kurisu et al., 2021, 2024). To the best of our knowledge, the EUCFe Rb and Ba concentrations are the first measurements over the Pacific Ocean. Their concentrations are similar to those of aerosols over the Atlantic Ocean (Landing and Shelley, 2014; Shelley et al., 2017). Given that V can be used as a tracer of the ship's exhaust (Duce and Hoffman, 1976), the lack of correlation between V concentrations and δ56Fe (Fig. B2) ruled out the possibility of contamination from the ship's exhaust.

Overall, these comparisons are consistent with previous values for these elements and validate the analytical procedure, from sampling to final concentrations.

EUCFe aerosols have Fe isotopic ratios ranging from -0.16 ‰ to +0.47 ‰ (Table 3, Figs. 1 and 3). Those sampled along the Equator and near the Bismarck Sea exhibit similar, slightly heavy signatures, ranging from +0.26 ‰ to +0.43 ‰. Those sampled in the tropical North Pacific present more variable signatures but are still positive from +0.14 ‰ to +0.47 ‰. One sample, the southeasternmost one (A238), differed significantly from the others in the equatorial Pacific with the lightest value, -0.16 ‰.

δ56Fe aerosol values from the EUCFe cruise can be compared with three other cruises in the Pacific: KH-13-7 and KH-14-3 in the North Pacific (Kurisu et al., 2021) and GP02 in the subarctic North Pacific (Kurisu et al., 2024) (Fig. 3). In these previous studies, all δ56Fe values below 0 ‰ were measured in samples taken less than 1500 km from the Japanese and Alaskan coasts (Fig. 3). In the open ocean, they also reported positive δ56Fe values as for EUCFe samples (apart from sample A238). South of the Tropic of Cancer, Kurisu et al. (2021) reported bulk aerosol heavy δ56Fe values, between +0.04 ‰ and +0.42 ‰, with a mean value of +0.27 ± 0.26 ‰ (2 SD, n= 7). In the subarctic North Pacific, the pelagic and Alaskan areas have δ56Fe values between -0.07 ‰ and +0.45 ‰ (Kurisu et al., 2024). Overall, EUCFe δ56Fe values are in excellent agreement with these previous works.

First, we will discuss the possibility that aerosol signatures correspond to unmodified source signatures. We will explore three hypotheses: contributions (i) from sea spray, (ii) from crustal sources, and (iii) from anthropogenic sources.

A first hypothesis is a contribution from seawater, i.e., sea spray. Based on the assumptions that all Na in EUCFe samples comes from seawater and that the chemical composition of sea spray is that of North Pacific seawater (Nozaki, 1997), the contribution of sea spray to our samples can be estimated with the following equation (Eq. 2): 2EIseaspray=NasampleEISW-refNaSW-ref, where EI is the element of interest (Fe for instance) and SW-ref is the seawater used as a reference (Nozaki, 1997) for Na and EI (Table 2).

This leads to insignificant contributions from seawater to the Fe content of all our samples (lower than 10⁻⁵ % of the total Fe content) (Table C1). On the other hand, the estimated sea spray contribution for Mg, Ca, and Sr was > 89 % for all samples.

A second hypothesis is a source from the erosion products of crustal rocks. The crustal signature, δ56Fe =+ 0.07 ‰, has been characterized in granites (Poitrasson, 2006), but other materials, such as volcanic rocks, exhibit similar isotopic composition. Desert dust, e.g., of Saharan origin (Beard et al., 2003; Waeles et al., 2007; Mead et al., 2013; Conway et al., 2019), and basalts (Poitrasson, 2006; Craddock et al., 2013; Teng et al., 2013) display the same signature. Accordingly, runoff water collected from the flanks of volcano Rabaul in the Bismarck area has been characterized by δ56Fe = +0.07 ± 0.03 ‰ (2 SD, n= 2) (Labatut et al., 2014). Therefore, EUCFe aerosol sample isotopic signatures, whether those in the group of nine samples slightly enriched in heavy isotopes or that of the A238 sample slightly enriched in light isotopes, do not directly reflect a crustal source.

A third hypothesis is an anthropogenic origin. Human activities emit aerosols within a wide range of δ56Fe. On the one hand, vehicle exhaust, steel manufacturing, and solid waste incineration have been characterized by negative δ56Fe signatures (Kurisu et al., 2016a). On the other hand, coal fly ash, metallic brake dust, and steel manufacturing have been characterized by positive δ56Fe signatures (Flament et al., 2008; Majestic et al., 2009; Mead et al., 2013; Li et al., 2022). Biomass burning can be characterized by both negative δ56Fe signatures (Mead et al., 2013) and positive δ56Fe signatures, with the latter due to the presence of suspended soil particles (Kurisu and Takahashi, 2019).

Sample A238 (δ56Fe = -0.16 ‰) is located in the southern part of the Pacific around 140° W (Figs. 1 and 3, Table 3). The air mass back trajectories (Fig. 2) suggest that aerosols collected in this area originated from the South Pacific or the South American coast. As stated above, several anthropogenic sources, biomass burning, vehicle exhaust, steel manufacturing, and solid waste incineration have been characterized by negative signatures (Mead et al., 2013; Kurisu et al., 2016a; Kurisu and Takahashi, 2019). Combustion processes from South America are therefore a potential explanation for the A238 sample.

The remaining of the discussion will focus on the group of nine samples, characterized by slightly heavy Fe isotopic composition (δ56Fe = +0.31 ± 0.21 ‰, 2 SD, n= 9; Figs. 1 and 3, Table 3). From a purely isotopic signature point of view, anthropogenic sources, e.g., coal combustion and steel manufacturing, possibly mixed with crustal sources, could explain these slightly heavy signatures (Wei et al., 2024). Nevertheless, there are several arguments contradicting this hypothesis: demography, modeled atmospheric back trajectories, aerosol size (> 1 µm), and elemental ratios such as Fe/Ti. While discussing similar slightly heavy aerosol isotopic signatures in the Bismarck Sea, a possible anthropogenic pollution contribution was excluded (Labatut et al., 2014) given the very low demography of the surrounding lands such as Papua New Guinea (Brunskill, 2004). Back trajectories presented in Fig. 2 reveal that the sampled air masses had a wide variety of geographic origins. The fact that aerosols have variable sources but similar isotope signatures does not support the hypothesis of an anthropogenic source such as coal fly ash, metallic brake dust, and steel manufacturing, which are not expected to be widely and homogeneously distributed around our study area. The separation between fine and coarse aerosol particles is 2 to 2.5 µm (Whitby, 1978; Seinfeld and Pandis, 2006). Nevertheless, fine particles do not ordinarily grow larger than 1 µm (Whitby, 1978). The EUCFe samples are mainly coarse aerosols, a size fraction associated with crustal sources (Mead et al., 2013). The Fe fractional solubility of aerosols was not measured. While this would have been interesting, it is not critical, as this information alone is not necessarily indicative of aerosol sources (Baker and Jickells, 2006; Conway et al., 2015).

The enrichment factor (EF) in an element of interest relative to the crust (Zoller et al., 1974) can be defined as (Eq. 3): 3enrichmentfactorEF=elementofinterestlithogenictracersampleelementofinterestlithogenictracerUCC. UCC stands for upper continental crust (Rudnick and Gao, 2014) (Table 2). Ti and Al are often used as lithogenic tracers (Dammshäuser, 2012). Because one sample (A252) is suspected to be contaminated in Al (Table 2), we chose Ti to calculate the EF relative to the crust in the following. Average UCC concentrations are often used as a reference (Rudnick and Gao, 2014). Nevertheless, the UCC exhibits variability in its elemental concentrations, which accounts for the range of Fe/Ti ratios depicted as a gray band in Fig. 4 (Hu and Gao, 2008). This range reflects eight types of rocks (n= 40), offering a non-exhaustive but more representative overview of the UCC.

Eight of the EUCFe samples fall within the UCC range (Fig. 4). However, two samples exhibit slightly lower (A259) and higher (A266) ratios. Their concentrations of anthropogenic tracers (Pb, V) do not suggest stronger anthropogenic contributions than in the other samples. The Fe/Ti ratios, which fall slightly outside the classical range (Hu and Gao, 2008) in samples A259 and A266, can nonetheless be explained by ultramafic rocks (e.g., pyroxenites), volcanic rocks (e.g., basalts and andesites), metamorphic rocks (e.g., gneiss), or plutonic rocks (e.g., diorite) (Turekian and Wedepohl, 1961; Canil and Lacourse, 2011). These rock types are present around the study area, notably the widespread volcanic rocks (Nusantara, 2000; Neall and Trewick, 2008; Ramos, 2009; Canil and Lacourse, 2011). Thus, despite the variable Fe/Ti ratios in our ten samples, they are all consistent with a crustal origin. Although it is common practice to use Pb or V enrichment factors relative to lithogenic tracers (such as Al or Ti) to trace anthropogenic sources, we chose not to do so because anthropogenic enrichments in Pb or V do not necessarily imply a significant anthropogenic enrichment in Fe (Table D1). Their use may therefore be misleading when studying the Fe cycle specifically.

Note that while the Fe/Ti A238 ratio is consistent with a crustal origin, it is also consistent with, for example, biomass burning (Zhai et al., 2021).

Based on the assumptions that all Ti in EUCFe samples comes from the UCC and that the chemical composition of crustal aerosol is that of UCC (Rudnick and Gao, 2014) (Table 2), the lithogenic contribution to our samples can be estimated (adjusting Eq. 2 to the case of a lithogenic source). For Fe, this leads to high lithogenic contributions (123 % on average). The fact that this calculation leads to contributions larger than 100 % likely reflects source ratios which differ from that chosen above (UCC) and/or Fe removal during transport.

These arguments suggest that the slightly heavy iron isotopic compositions are unlikely to be explained by anthropogenic sources but mainly by crustal ones. We will discuss below if our observations (δ56Fe_average = +0.31 ‰) can be explained by aerosols of crustal origin (+0.07 ‰) whose isotopic signature has been modified by isotopic fractionation during atmospheric transport.

A major process influencing aerosol chemistry, during atmospheric transport, is partial dissolution during condensation/evaporation cycles in clouds (Lelieveld and Crutzen, 1991; Desboeufs, 2001). Atmospheric aerosol Fe dissolution is mainly due to dissolution by low-pH cloud water and effects of solar irradiation. Different dissolution mechanisms exist, including proton-promoted (Chapman et al., 2009; Kiczka et al., 2010), ligand-promoted (Chapman et al., 2009; Kiczka et al., 2010; Mulholland et al., 2021; Maters et al., 2022), and reductive ligand-promoted dissolution (Mulholland et al., 2021; Maters et al., 2022). These processes fractionate iron isotopes (Mulholland et al., 2021; Maters et al., 2022). In most studies, light iron isotopes are preferentially dissolved, and the isotopic composition of the remaining particulate iron becomes gradually heavier (Maters et al., 2022) (Fig. 5). In the following, the notation Δ56Fe_{solution − particle} is used to denote the isotopic fractionation characterizing a given dissolution process, also named the initial fractionation step or the enrichment factor (also note ε, Wiederhold et al., 2006; Maters et al., 2022).

The magnitude of the isotope fractionations, Δ56Fe_{solution − particle}, were found between -0.3 ‰ and -2.0 ‰ for biotite and chlorite mineral dissolution (Kiczka et al., 2010) and at -1.95 ‰ for granite dissolution by hydrochloric acid (Chapman et al., 2009). An experiment dissolving anthropogenic aerosols with a synthetic cloud water solution showed a preferential release of light isotopes with Δ56Fe_{solution − particle} = -1 ‰ (Mulholland et al., 2021). Another experiment of mineral dust and industrial ash dissolution in simulated cloud water also showed an enrichment in light Fe isotopes in solution, with an isotope fractionation Δ56Fe_{solution − particle} of -1.8 ‰ for ash and dust (Maters et al., 2022). Thus, mineral dissolution appears to favor light isotopes, thereby enriching the remaining solid fraction in heavy isotopes. Therefore, we will assess whether partial dissolution during clouds transport can produce aerosols with a heavier iron isotopic composition. Some authors have suggested that the observed isotopic compositions may be partly due to isotopic fractionation during transport (Kurisu et al., 2021, 2024; Wang et al., 2022).

Considering that the leachate is isolated from the solid fraction of the aerosol, the system can be modeled as a Rayleigh distillation. The isotope composition of the solid fraction of the aerosol is calculated according to Eqs. (4) and (5): 4(δ56Feparticle)f≈(δ56Feparticle)f=1+Δ56Fesolution-particleln(f), where the particle is the solid fraction of the aerosol, the solution is the leached solution, and f is the remaining fraction of Fe_particle (when f= 1 all Fe is in the particle; no Fe has been leached).

For the particle value, we assume an initial crustal signature for EUCFe aerosols, (δ56Feparticle)f=1 = +0.07 ‰ (Poitrasson, 2006). For the isotopic fractionation, Δ56Fe_{solution − particle}, although the experiments described above document values ranging between -2.0 ‰ and -0.3 ‰, we choose -1.8 ‰ (Maters et al., 2022). This value was measured during a laboratory experiment on dust with simulated cloud water, i.e., a similar situation to the EUCFe field study (Maters et al., 2022). Equation (5) (derived from Eq. 4) allows us to estimate the fractions of the particles that have to be dissolved (1-f) in order to reach the slightly heavy isotope composition measured. 51-f=1-e(δ56Feparticle)f=1-(δ56Feparticle)fΔ56Fesolution-particle Based on these calculations, we estimate Fe dissolution percentages varying from 4 % to 20 % with an average value of 13 % (Table 4). This is the first estimate of this kind to our knowledge. A comparison can be made with Fe fractional solubility of aerosols measured during seawater or ultrapure deionized water leaching experiments (Sholkovitz et al., 2012; Buck et al., 2013; Shelley et al., 2018; Kurisu et al., 2021, 2024; Desboeufs et al., 2024), keeping in mind that clouds are slightly acidic with a pH around 5 in the equatorial Pacific (Shah et al., 2020). Locally, Fe fractional solubility can reach 23 % in the northwestern Pacific (Kurisu et al., 2021) and 29 % in the Pacific Ocean (3 cruises) (Buck et al., 2013) during leaching experiments with ultrapure deionized water. Mean Fe fractional solubility has been reported as the highest in the world in the equatorial Pacific, with mean values ranging from 12 % to 20 % (Hamilton et al., 2019). Fe fractional solubility depends on numerous factors such as aerosol size and origin, as well as atmospheric processes (pH, solar irradiation, composition of the solution). Crustal aerosols collected during dust events in coastal Namibia (aerosols < 10 µm) can reach high Fe fractional solubilities of 20 % (Desboeufs et al., 2024). Therefore, a 13 % dissolution is a realistic value for crustal aerosols.

An isotopic fractionation by partial dissolution of crustal origin aerosols could therefore explain the slightly heavy signatures observed (Fig. 5). This would require that the leached fraction, enriched in light isotopes, is separated from the solid fraction. In the absence of separation, the effect of isotope fractionation would not have been measured in our samples. This process has not yet been demonstrated, but the hypothesis has already been proposed in two publications (Kurisu et al., 2021, 2024). The processes that could lead to such separation are difficult to identify. They are, however, necessary to explain our observations, provided that aerosol original signatures were crustal. Shattering or ice breaking are two ways to separate the leached fraction and the residual particle of the aerosols. Their occurrence is understudied, especially regarding shattering process. The enrichment of light isotopes in the leached fraction was not observed in this study. This is likely due to the presence of this fraction in aerosols smaller than 1 µm produced by ice-breaking or shattering processes (or its removal by wet deposition), which were not sampled during the EUCFe cruise. The Fe isotopic composition of fine aerosols, often negative, is mostly attributed to anthropogenic sources (Conway et al., 2019; Kurisu et al., 2021). However, this study proposes a new possible cause for the light Fe isotopic composition of aerosols smaller than 1 µm: the residual leached fractions of crustal aerosols. In the above model, isotopic fractionation applies to the whole particle (its entire mass), whereas in reality it is a surface process that only affects a peripheral layer. Taking into account surface processes would lead to smaller isotopic effects (Wiederhold et al., 2006). Our approximation led to an overestimation of the effect of isotope fractionation and therefore an underestimation of the leached fraction.