Enhanced Iron Solubility at Low pH in Global Aerosols

: The composition and oxidation state of aerosol iron were examined using synchrotron-based iron near-edge X-ray absorption spectroscopy. By combining synchrotron-based techniques with water leachate analysis, impacts of oxidation state and mineralogy on aerosol iron solubility were assessed for samples taken from multiple locations in the Southern and the Atlantic Oceans; and also from Noida (India), Bermuda, and the Eastern Mediterranean (Crete). These sampling locations capture iron-containing aerosols from different source regions with varying marine, mineral dust, and anthropogenic inﬂuences. Across all locations, pH had the dominating inﬂuence on aerosol iron solubility. When aerosol samples were approximately neutral pH, iron solubility was on average 3.4%; when samples were below pH 4, the iron solubility increased to 35%. This observed aerosol iron solubility proﬁle is consistent with thermodynamic predictions for the solubility of Fe(III) oxides, the major iron containing phase in the aerosol samples. Source regions and transport paths were also important factors affecting iron solubility, as samples originating from or passing over populated regions tended to contain more soluble iron. Although the acidity appears to affect aerosol iron solubility globally, a direct relationship for all samples is confounded by factors such anthropogenic inﬂuence, aerosol buffer Generally, samples with the highest total iron low fractional solubility of iron. these compounds are collectively labeled as secondary phases because these are phases that may form through secondary reactions that occur in the atmosphere. However, such secondary phases may also be initially present in aerosols derived from coal ﬂy ash or volcanic materials.

. Global map with sampling locations indicated. Precise latitude and longitude for each sampling location can be found in Table S3.

HYSPLIT
To determine the geographic origin of the sampled air masses, the GDAS half-degree meteorological archive was utilized for the Hybrid Single Particle Lagrangian Integrated Trajectory Model (HYSPLIT) back trajectories [50,51]. Approximate air mass locations five days (ca. atmospheric lifetime of aerosol) before sample collection for all sampling locations were calculated using HYSPLIT back trajectories at 3000 m above ground level. The 3000 m height was chosen for the HYSPLIT back trajectories to establish the origin of dust layers transported from distant regions.

Total and Soluble Iron
For the purposes of this article, soluble iron refers to water soluble iron. The ferrozine technique [52] was used to measure soluble iron for all samples. All filters were handled under a HEPA-filtered laminar flow hood. Approximately 1/4 of the filter was extracted into 10 mL of ultrahigh purity water (>18 MΩ) in a centrifuge tube free of trace metals using an ultrasonic bath for 30 minutes. Then, insoluble particles were removed from the extract by filtering through a syringe filter with a 0.2 µm pore size. A 40 µM solution of hydroxylamine hydrochloride was added in a 1:100 ratio (reagent volume to filter extract volume) to reduce any dissolved Fe(III) in the sample to Fe(II). This was then set overnight to allow the samples to react with the hydroxylamine hydrochloride. In the morning, a 5 mM ferrozine solution was added in a 1:100 ratio (reagent volume to filter extract volume) to form a color specific complex with dissolved Fe(II). After allowing the samples to react for 10 min, they were measured  Table S3.
India: Daily sampling of particulate matter with an aerodynamic diameter of <10 µm (PM 10 ) was completed at Amity University in Noida, India (28.5 • N, 77.3 • E). Samples were collected for 17 h with a high-volume sampler where air was pulled through a size-fractionated cyclone inlet with a flow rate of 1000 L min −1 . Samples were collected on Whatman GF/A filters (p/n 1820-866). Fourteen samples were collected in late September 2014, when mineral dust is most prevalent in aerosol loadings [47].
Bermuda: Once a week for a year, samples were collected at the Bermuda Institute of Ocean Sciences (BIOS: 32.24 • N, 64.87 • W), which receives aerosols with marine, North African, and North American influences. Total suspended particle sampling began in June 2011 and concluded in June 2012. A representative subset of this sample collection was analyzed. A Tisch 5170V-BL high-volume Total Suspended Particle aerosol sampler with Whatman-41 filters, a flow rate of~1200 L min −1 and a face velocity of~100 cm s −1 was used to collect these samples.
Crete: The Finokalia Research Station (35.32 • N, 25.67 • E) is located 70 km from the nearest major city on the island of Crete (Eastern Mediterranean), Greece. This location isolates it from both local and regional interferences [48]. PM 10 samples were collected on Teflon filters using a virtual impactor with Atmosphere 2018, 9,201 4 of 17 an operational flow rate of 16.7 L min −1 . Fourteen samples were collected at the Finokalia site over a span ranging from one to three days from 2009 to 2011. These samples were dominated by either European or North African air masses.
Atlantic Ocean: Twenty-four hour particle collections occurred as part of the GEOTRACES Atlantic transect in 2011 [49]. Nine samples in total from different source regions were analyzed from the GEOTRACES collection: three marine, three North African, and three North American samples. Total suspended particle samples were collected using Whatman-41 filters and a Tisch 5170V-BL high-volume sampler averaging~1200 L min −1 with a face velocity of approximately 100 cm s −1 [49]. To prevent contamination from the ship's exhaust, the aerosol sampler was controlled for wind speed (>0.5 m s −1 ) and sector (+/−60 deg from the bow).
Southern Ocean: Three samples from the Southern Ocean were collected along a transect from southwest Australia to the Australian Antarctic Casey Station [34]. Shipboard aerosol sampling was conducted from November 2010 to March 2011. Forty-seven millimeter diameter, 1.0 µm pore size Teflon filters were used to collect bulk aerosol particles under a sampling flow rate of 15 L min −1 for two to three days. A wind sector and speed control system regulated the sampling instruments to prevent contamination from the ship's exhaust.
Until analysis, all aerosol samples were protected from light and stored at −20 • C. More detailed sampling information is available in Tables S1-S3.

HYSPLIT
To determine the geographic origin of the sampled air masses, the GDAS half-degree meteorological archive was utilized for the Hybrid Single Particle Lagrangian Integrated Trajectory Model (HYSPLIT) back trajectories [50,51]. Approximate air mass locations five days (ca. atmospheric lifetime of aerosol) before sample collection for all sampling locations were calculated using HYSPLIT back trajectories at 3000 m above ground level. The 3000 m height was chosen for the HYSPLIT back trajectories to establish the origin of dust layers transported from distant regions.

Total and Soluble Iron
For the purposes of this article, soluble iron refers to water soluble iron. The ferrozine technique [52] was used to measure soluble iron for all samples. All filters were handled under a HEPA-filtered laminar flow hood. Approximately 1/4 of the filter was extracted into 10 mL of ultrahigh purity water (>18 MΩ) in a centrifuge tube free of trace metals using an ultrasonic bath for 30 min. Then, insoluble particles were removed from the extract by filtering through a syringe filter with a 0.2 µm pore size. A 40 µM solution of hydroxylamine hydrochloride was added in a 1:100 ratio (reagent volume to filter extract volume) to reduce any dissolved Fe(III) in the sample to Fe(II). This was then set overnight to allow the samples to react with the hydroxylamine hydrochloride. In the morning, a 5 mM ferrozine solution was added in a 1:100 ratio (reagent volume to filter extract volume) to form a color specific complex with dissolved Fe(II). After allowing the samples to react for 10 min, they were measured spectrophotometrically at 562 nm [31]. Total iron was determined for Southern Ocean, GEOTRACES, and BIOS samples using established protocols, as discussed in literature [34,49,53,54]. Total iron was also measured for Bermuda Institute of Ocean Science and GEOTRACES samples using elemental densities collected from X-ray fluorescence maps. Oakes et al. [44] demonstrated that synchrotron-based techniques could dependably measure total iron. Elemental densities obtained from X-ray fluorescence maps were used to determine total iron for samples from India and the Finokalia Research Station [43].

Soluble Ions and pH Modelling
The main soluble ions for samples from Finokalia Research Station and a subset of samples from the GEOTRACES transect and BIOS collection were measured using ion chromatography [55]. Approximately 1/8 of each filter was extracted into 10 mL of ultrahigh purity water (>18 MΩ, Barnstead Atmosphere 2018, 9, 201 5 of 17 Nanopure) by ultrasonication for 45 min. Insoluble species were removed from the filter extracts using syringe filters (PALL IC Acrodisc (PES), 0.45 µm, 13 mm). These new filtered solutions were then run to detect anions (Cl − , Br − , NO 3 − , SO 4 2− , HPO 4 2− , and C 2 O 4 2− ) and cations (NH 4 + , K + , Na + , Mg 2+ , and Ca 2+ ). Anions were measured with a Dionex-500 ion chromatograph equipped with an Ion Pac AS4A-SC column and an AG4A-SC pre-column, with an ASRS-300 suppressor. Anions were separated by isocratic elution of NaHCO 3 (3.4 mM)/Na 2 CO 3 (3.6 mM) with a flow of 1.5 mL min −1 . An Ion Pac CS12A column and a CG12A guard column with a CSRS-300 suppressor separated the cations under isocratic elution of 20 mM MSA (methanesulfonic acid) at a 1.0 mL min −1 flow rate. The reproducibility of the measurements exceeded 2% with a 1-5 ppb range for the detection limit of the major anions and cations. Models using ISORROPIA-II were run to calculate the aerosol acidity for these samples [56,57]. This thermodynamic model provides the concentrations of species in the solid and liquid aerosol phases, pH, and aerosol water. The size fractions of aerosols collected in these samples conform to the equilibrium assumption of ISORROPIA-II; therefore, this program is limited to model the samples as a range from acidic to circum neutral. To calculate the precise pH, relative humidity of the marine boundary layer and ambient temperature were assumed (80% relative humidity; 293 K).

Synchrotron Analysis
The mineralogy of aerosol iron was measured with the micro X-ray fluorescence capabilities at Station 2-ID-D at the Advanced Photon Source at Argonne National Laboratory. Utilizing X-ray fluorescence microscopy and X-ray Absorption Near Edge Structure (XANES) spectroscopy, the iron in the aerosol samples was examined. These techniques provide incident X-rays with enough energy to bombard atoms, ejecting the electrons out of the innermost electron shell. Following this, an outer shell electron moves into the vacated position, emitting a specific fluorescence signal. As the ejected electron interacts with the adjacent atoms, the oxidation state, type, and structural arrangement of atoms in a particle are revealed, which is illustrated in XANES spectra [58]. This is how XANES spectra can provide data for both the oxidation state and the composition associated with the element of interest, in this case, iron.
A Vortex EM, an energy dispersive Si-drift detector with a 50 mm 2 sensitive area, and a 12.5 µm Be window from SII NanoTechnology in Northridge CA) measured the samples' X-ray fluorescence. Samples were measured in a helium filled chamber to minimize absorption and fluorescence caused by beam interactions with air. The effect of iron oxidation state changes due to the X-ray beam was also examined in selected samples by repeated XANES measurements yielding a cumulative dwell time of 10 s for each energy step. This cumulative dwell time was far longer than any typical sample analysis. No spectral changes were observed in the repeated XANES measurements. An area of each filter sample, ca. 0.5 cm 2 , selected at random, was mounted over a slot on an aluminum sample holder for analysis. The sample was initially analyzed in microscopic X-ray fluorescence (XRF) mode to identify iron-containing particles on the filter. In this mode, the X-ray beam with a diameter of approximately 350 nm was scanned over a randomly selected filter area (usually between 40 to 50 µm 2 with a 0.4 µm step size and a dwell time of half a second to provide an elemental distribution map of the filter. These XRF maps were used to determine the precise location of aerosol particles (>1 µm diameter) which contained iron for individual particle Fe-XANES spectroscopy.
Fe-XANES spectroscopy was performed in both individual particle and bulk modes. In bulk particle mode, the X-ray zone plate and order sorting aperture were removed to permit characterization of iron present on filter areas of approximately 0.4 mm 2 . In the individual particle Fe-XANES configuration, particles containing iron, observed in elemental distribution maps, were characterized with a focused beam with a 350 nm spot size. In both individual particle and bulk modes, spectroscopy scans were collected in half eV steps between 7100 to 7180 eV with a dwell time of 0.5−3 s per step.

Data Reduction and Analysis
For the bulk and individual particle spectra, oxidation state was determined by measuring the pre-edge centroid position of the Fe-XANES spectra. The pre-edge centroid position has been demonstrated to be a more reliable indicator of iron oxidation state than the absorption edge position [59]. Percent Fe(II) of aerosol samples was determined by interpolation of their pre-edge position relative to standards of known iron oxidation state [44,59,60].
Spectral data was normalized to the incoming X-ray flux and processed using the Athena software package [61]. Because of the large number of spectra collected, data reduction techniques were employed. A combination of principle component analysis with k-means clustering were utilized to group similar spectra [62]. Principle components were determined using the algorithm built into Athena [61]. To create an average spectra representing a cluster, similar spectra were merged, and linear combination fitting was then completed for each cluster's Fe-XANES spectra using Athena [61]. Dividing the spectra into bulk and individual particle Fe-XANES, these data reduction techniques were used for each sampling site.
Using a database of known iron mineral standards, the linear combination fitting algorithm in Athena determines the phases contributing to an unknown Fe-XANES spectra. In this study, the database of mineral standards was run on the same beamline [63]. An iterative process was used to refine the standard database used to model the spectra. First, exotic iron minerals which are not likely to be significant contributors to aerosol iron were excluded from the database [44]. Then, the database was refined by eliminating standards with low contribution (e.g., under 10%) or a poor fitting factor during initial linear combination fitting. Spectra of individual particles in each sample were also helpful in the selection of standards to include in the modeling of bulk spectra. Additionally, insufficient quantities of a specific mineral or compound in a sample can also lead to underestimation of the specific compound during linear combination fitting [64]. Therefore, iron composition was generalized to express broad chemical classes, specifically: Fe(III) oxides, Fe(III) sulfates, Fe(II) sulfates, Fe(III) phosphates, and Fe(II) silicates. Athena utilizes a nonlinear, least squares minimization method to fit unknown spectra with standard spectra, producing the fitting factor which quantifies the goodness of a fit.

Iron Solubility
The total solubility of iron varied widely across the five different sampling locations (Tables S4  and S5). The highest soluble iron concentrations were found in samples from India, which contained 90 ± 26 ng/m 3 of soluble iron on average ( Figure 2). North African influenced samples from Finokalia Research Station exhibited the second highest levels of soluble iron (averaging 32 ± 52 ng/m 3 ) and contain on average four times more soluble iron than the European influenced samples (7.4 ± 6.3 ng/m 3 ) collected at the same sampling location. Similar to Finokalia Research Station, North African influenced samples from the GEOTRACES cruise contained the highest levels of soluble iron (on average 6.0 ± 4.1 ng/m 3 ). Samples with primarily marine influences-the GEOTRACES and Southern Ocean samples-averaged 0.18 ± 0.13 ng/m 3 and 1.2 ± 1.4 ng/m 3 of soluble iron, respectively. North-American-influenced samples from the GEOTRACES collection averaged the same magnitude of soluble iron (1.3 ± 0.78 ng/m 3 ) as the year-long collection of BIOS samples (3.8 ± 7.9 ng/m 3 ).
However, when the fractional solubility of iron was considered (percentage of the total iron that is water solubilized) in lieu of total solubility, trends begin to emerge. European influenced samples from Finokalia Research Station had the most fractional soluble iron, followed by samples dominated by marine influences. In the samples for which North African mineral dust was mixed with anthropogenically influenced North American air masses, such as at the Bermuda site, aerosol iron often had a higher fractional iron solubility than samples that contained a stronger North African year-long collection of BIOS samples (3.8 ± 7.9 ng/m 3 ).
However, when the fractional solubility of iron was considered (percentage of the total iron that is water solubilized) in lieu of total solubility, trends begin to emerge. European influenced samples from Finokalia Research Station had the most fractional soluble iron, followed by samples dominated by marine influences. In the samples for which North African mineral dust was mixed with anthropogenically influenced North American air masses, such as at the Bermuda site, aerosol iron often had a higher fractional iron solubility than samples that contained a stronger North African mineral dust influence ( Figure 2). Fractional iron solubility in the Indian samples is also high, generally >5%, possibly indicating anthropogenic contributions of soluble iron.

Oxidation State
Samples from India contained primarily Fe(III), with only one bulk sample showing an Fe(II) signal of 2.1% (reflecting that 2.1% of the Fe in the sample is in this oxidation state) (Tables S4 and S5). This was similar to the findings for Southern Ocean samples, where two of the three samples contained only Fe(III), and the final sample contained about 90% Fe(III). At Finokalia Research Station, samples with a European influence tended to contain more reduced iron (13.5% on average) than samples with a North African  However, when the fractional solubility of iron was considered (percentage of the total iron that is water solubilized) in lieu of total solubility, trends begin to emerge. European influenced samples from Finokalia Research Station had the most fractional soluble iron, followed by samples dominated by marine influences. In the samples for which North African mineral dust was mixed with anthropogenically influenced North American air masses, such as at the Bermuda site, aerosol iron often had a higher fractional iron solubility than samples that contained a stronger North African mineral dust influence ( Figure 2). Fractional iron solubility in the Indian samples is also high, generally >5%, possibly indicating anthropogenic contributions of soluble iron.

Oxidation State
Samples from India contained primarily Fe(III), with only one bulk sample showing an Fe(II) signal of 2.1% (reflecting that 2.1% of the Fe in the sample is in this oxidation state) (Tables S4 and S5). This was similar to the findings for Southern Ocean samples, where two of the three samples contained only Fe(III), and the final sample contained about 90% Fe(III). At Finokalia Research Station, samples with a European influence tended to contain more reduced iron (13.5% on average) than samples with a North African , BIOS Saharan air masses However, when the fractional solubility of iron was considered (percentage of the total iron that is water solubilized) in lieu of total solubility, trends begin to emerge. European influenced samples from Finokalia Research Station had the most fractional soluble iron, followed by samples dominated by marine influences. In the samples for which North African mineral dust was mixed with anthropogenically influenced North American air masses, such as at the Bermuda site, aerosol iron often had a higher fractional iron solubility than samples that contained a stronger North African mineral dust influence ( Figure 2). Fractional iron solubility in the Indian samples is also high, generally >5%, possibly indicating anthropogenic contributions of soluble iron.

Oxidation State
Samples from India contained primarily Fe(III), with only one bulk sample showing an Fe(II) signal of 2.1% (reflecting that 2.1% of the Fe in the sample is in this oxidation state) (Tables S4 and S5). This was similar to the findings for Southern Ocean samples, where two of the three samples contained only Fe(III), and the final sample contained about 90% Fe(III). At Finokalia Research Station, samples with a European influence tended to contain more reduced iron (13.5% on average) than samples with a North African , BIOS mixed air masses However, when the fractional solubility of iron was considered (percentage of the total iron that is water solubilized) in lieu of total solubility, trends begin to emerge. European influenced samples from Finokalia Research Station had the most fractional soluble iron, followed by samples dominated by marine influences. In the samples for which North African mineral dust was mixed with anthropogenically influenced North American air masses, such as at the Bermuda site, aerosol iron often had a higher fractional iron solubility than samples that contained a stronger North African mineral dust influence ( Figure 2). Fractional iron solubility in the Indian samples is also high, generally >5%, possibly indicating anthropogenic contributions of soluble iron.

Oxidation State
Samples from India contained primarily Fe(III), with only one bulk sample showing an Fe(II) signal of 2.1% (reflecting that 2.1% of the Fe in the sample is in this oxidation state) (Tables S4 and S5). This was similar to the findings for Southern Ocean samples, where two of the three samples contained only Fe(III), and the final sample contained about 90% Fe(III). At Finokalia Research Station, samples with a European influence tended to contain more reduced iron (13.5% on average) than samples with a North African , BIOS North

American air masses
year-long collection of BIOS samples (3.8 ± 7.9 ng/m ).
However, when the fractional solubility of iron was considered (percentage of the total iron that is water solubilized) in lieu of total solubility, trends begin to emerge. European influenced samples from Finokalia Research Station had the most fractional soluble iron, followed by samples dominated by marine influences. In the samples for which North African mineral dust was mixed with anthropogenically influenced North American air masses, such as at the Bermuda site, aerosol iron often had a higher fractional iron solubility than samples that contained a stronger North African mineral dust influence ( Figure 2). Fractional iron solubility in the Indian samples is also high, generally >5%, possibly indicating anthropogenic contributions of soluble iron.

Oxidation State
Samples from India contained primarily Fe(III), with only one bulk sample showing an Fe(II) signal of 2.1% (reflecting that 2.1% of the Fe in the sample is in this oxidation state) (Tables S4 and S5). This was similar to the findings for Southern Ocean samples, where two of the three samples contained only Fe(III), and the final sample contained about 90% Fe(III). At Finokalia Research Station, samples with a European influence tended to contain more reduced iron (13.5% on average) than samples with a North African
However, when the fractional solubility of iron was considered (percentage of the total iron that is water solubilized) in lieu of total solubility, trends begin to emerge. European influenced samples from Finokalia Research Station had the most fractional soluble iron, followed by samples dominated by marine influences. In the samples for which North African mineral dust was mixed with anthropogenically influenced North American air masses, such as at the Bermuda site, aerosol iron often had a higher fractional iron solubility than samples that contained a stronger North African mineral dust influence ( Figure 2). Fractional iron solubility in the Indian samples is also high, generally >5%, possibly indicating anthropogenic contributions of soluble iron.

Oxidation State
Samples from India contained primarily Fe(III), with only one bulk sample showing an Fe(II) signal of 2.1% (reflecting that 2.1% of the Fe in the sample is in this oxidation state) (Tables S4 and S5) year-long collection of BIOS samples (3.8 ± 7.9 ng/m 3 ). However, when the fractional solubility of iron was considered (percentage of the total iron that is water solubilized) in lieu of total solubility, trends begin to emerge. European influenced samples from Finokalia Research Station had the most fractional soluble iron, followed by samples dominated by marine influences. In the samples for which North African mineral dust was mixed with anthropogenically influenced North American air masses, such as at the Bermuda site, aerosol iron often had a higher fractional iron solubility than samples that contained a stronger North African mineral dust influence ( Figure 2). Fractional iron solubility in the Indian samples is also high, generally >5%, possibly indicating anthropogenic contributions of soluble iron.

Oxidation State
Samples from India contained primarily Fe(III), with only one bulk sample showing an Fe(II) signal of 2.1% (reflecting that 2.1% of the Fe in the sample is in this oxidation state) (Tables S4 and S5). This was similar to the findings for Southern Ocean samples, where two of the three samples contained only Fe(III), and the final sample contained about 90% Fe(III). At Finokalia Research Station, samples with a European influence tended to contain more reduced iron (13.5% on average) than samples with a North African  year-long collection of BIOS samples (3.8 ± 7.9 ng/m 3 ). However, when the fractional solubility of iron was considered (percentage of the total iron that is water solubilized) in lieu of total solubility, trends begin to emerge. European influenced samples from Finokalia Research Station had the most fractional soluble iron, followed by samples dominated by marine influences. In the samples for which North African mineral dust was mixed with anthropogenically influenced North American air masses, such as at the Bermuda site, aerosol iron often had a higher fractional iron solubility than samples that contained a stronger North African mineral dust influence ( Figure 2). Fractional iron solubility in the Indian samples is also high, generally >5%, possibly indicating anthropogenic contributions of soluble iron.

Oxidation State
Samples from India contained primarily Fe(III), with only one bulk sample showing an Fe(II) signal of 2.1% (reflecting that 2.1% of the Fe in the sample is in this oxidation state) (Tables S4 and S5). This was similar to the findings for Southern Ocean samples, where two of the three samples contained only Fe(III), and the final sample contained about 90% Fe(III). At Finokalia Research Station, samples with a European influence tended to contain more reduced iron (13.5% on average) than samples with a North African . Generally, samples with the highest total iron have low fractional solubility of iron.

Oxidation State
Samples from India contained primarily Fe(III), with only one bulk sample showing an Fe(II) signal of 2.1% (reflecting that 2.1% of the Fe in the sample is in this oxidation state) (Tables S4 and S5). This was similar to the findings for Southern Ocean samples, where two of the three samples contained only Fe(III), and the final sample contained about 90% Fe(III). At Finokalia Research Station, samples with a European influence tended to contain more reduced iron (13.5% on average) than samples with a North African influence (3.5% on average). In contrast, GEOTRACES samples contained at least 30% Fe(II) at the bulk level and could be dominated by Fe(II) (>95%). BIOS samples all contained at least 30% Fe(II) with Fe(II) reaching levels of about 75% in some bulk samples. Fe(II) is generally considered more soluble than Fe(III), but this relationship was not seen in our data [43]. When all the data was considered, there was a wide range of fractional iron solubility at each oxidation state, as represented by the position of the pre-edge centroid ( Figure S1), indicating the absence of a direct relationship between iron oxidation state and the fractional solubility of aerosol iron.

Discussion
There are several hypothesized controlling factors for aerosol iron solubility, including composition and mineralogy, oxidation state, cloud processing, acidic reactions, organic ligand dissociation, and photoreduction [13,16,18,[65][66][67][68]. In ambient aerosol samples, finding direct evidence of these mechanisms is challenging. Here, we focus primarily on the roles of composition, acidic reactions, oxidation state, and by extension photoreduction, which may influence the solubility of aerosol iron. Photoreduction pathways have been shown to play a key role in iron transformation in laboratory based experiments [69,70]; however, nitric acid suppresses the fraction of dissolved Fe(II) at low pH [26]. This suggests that these particle aging mechanisms further confounds the identification of a particular mechanism in ambient aerosol samples.
When data from India, the Southern Ocean, Bermuda, GEOTRACES, and Finokalia Research Station were considered together, there were no clear relationships between traditional factors that are typically associated with iron solubility. Composition (as a function of total iron content, R 2 = 0.027) and oxidation state of iron (R 2 = 0.058) are both not correlated with the fractional solubility of iron, suggesting that solubility of iron may not be linked to a specific iron phase or oxidation state ( Figure S1). This is in contrast to evidence suggesting that photoreduction of iron occurs during the long-range transport of marine aerosol [43,65]. While our samples cover a wide range of transport distances, and presumably transport times, the absence of a direct relationship between iron solubility and oxidation state may stem from insufficient transport times for a large portion of our sample library, as detailed in Table S2. Photoreduction mechanisms working to solubilize mineral dust have been suggested to need more than 10 days of transport to show significant solubilization effects [71]. Composition can also dictate how photochemical mechanisms will proceed. Even within a general class of minerals, such as Fe(III) oxides, various minerals have variable levels of susceptibility to photochemical processing [35].
Correlation seen between inorganic acids, sulfuric acid, and nitric acid, and the solubility of aerosol iron had been used as evidence for acidic reactions promoting the solubility of iron in mineral dust A total of 157 individual particles were examined for iron mineralogy across the five different sample sets. Individual particles were generally more diverse in mineralogy than the bulk mineralogy (Table S7). Individual particles sampled from India belong to three mineralogical groups and were either dominated by Fe(III) oxides or Fe(II) silicates. Individual particles sampled from the Southern Ocean contained a combination of Fe(III) oxides and Fe(II) sulfates or were dominated by Fe(III) phosphates. Individual particles from Finokalia Research Station contained largely iron oxides. Some particles also contained minor quantities of ilmenite, an iron oxide containing titanium. Individual particles sampled from Finokalia Research Station also contained ilmenite but were largely dominated by Fe(III) oxides. Individual particles from Bermuda were largely a mixture of Fe(III) oxides and Fe(II) silicates. Individual particles sampled from the GEOTRACES transect had a strong presence of Fe(II and III) sulfates and Fe(III) phosphates with only minor contributions from Fe(III) oxides. The general dominance of iron oxides and sulfates was also seen in these individual particle analyses. In general, the phases identified in individual particles were consistent with the ones observed in bulk samples. In contrast to bulk spectra, which often have contributions from several different iron minerals, individual particle spectra were typically easier to fit with Athena [61]. Identification of phases in the individual particles supports their choice when modeling bulk spectra. Bulk spectra should be viewed as more representative of the overall iron mineralogy of a sample because they access particles across the entire spectrum of size ranges present in a sample.

Discussion
There are several hypothesized controlling factors for aerosol iron solubility, including composition and mineralogy, oxidation state, cloud processing, acidic reactions, organic ligand dissociation, and photoreduction [13,16,18,[65][66][67][68]. In ambient aerosol samples, finding direct evidence of these mechanisms is challenging. Here, we focus primarily on the roles of composition, acidic reactions, oxidation state, and by extension photoreduction, which may influence the solubility of aerosol iron. Photoreduction pathways have been shown to play a key role in iron transformation in laboratory based experiments [69,70]; however, nitric acid suppresses the fraction of dissolved Fe(II) at low pH [26]. This suggests that these particle aging mechanisms further confounds the identification of a particular mechanism in ambient aerosol samples.
When data from India, the Southern Ocean, Bermuda, GEOTRACES, and Finokalia Research Station were considered together, there were no clear relationships between traditional factors that are typically associated with iron solubility. Composition (as a function of total iron content, R 2 = 0.027) and oxidation state of iron (R 2 = 0.058) are both not correlated with the fractional solubility of iron, suggesting that solubility of iron may not be linked to a specific iron phase or oxidation state ( Figure S1). This is in contrast to evidence suggesting that photoreduction of iron occurs during the long-range transport of marine aerosol [43,65]. While our samples cover a wide range of transport distances, and presumably transport times, the absence of a direct relationship between iron solubility and oxidation state may stem from insufficient transport times for a large portion of our sample library, as detailed in Table S2. Photoreduction mechanisms working to solubilize mineral dust have been suggested to need more than 10 days of transport to show significant solubilization effects [71]. Composition can also dictate how photochemical mechanisms will proceed. Even within a general class of minerals, such as Fe(III) oxides, various minerals have variable levels of susceptibility to photochemical processing [35].
Correlation seen between inorganic acids, sulfuric acid, and nitric acid, and the solubility of aerosol iron had been used as evidence for acidic reactions promoting the solubility of iron in mineral dust [47,72,73]. Here, there was no relationship between total inorganic acidity (sum of sulfuric and nitric acids) (R 2 = 0.0016), nitrate (R 2 = 7 × 10 −6 ), or non-sea-salt sulfate (R 2 = 0.17) and the fractional solubility of iron. However, when pH was modelled with ISORROPIA [56,57], a trend indicates one possible explanation (Figure 4). When the pH is approximately neutral, a range from approximately 4 to 8, the solubility of iron is relatively low, but as the pH drops below 4, the fractional solubility of iron rapidly increases, up to nearly 35%. This behavior is consistent with the speciation of Fe(III), determined using the standard selection in the Mineql + database to create an Fe(III) speciation diagram [74], where below a pH of approximately 3, the abundance of dissolved Fe(III) sharply increases [75][76][77]. This suggests that dissolved Fe(III) production is a consequence of acidic reactions in the atmosphere as would be expected for acidic species acting on a largely Fe(III) oxide composition. So only in the acidic range, when sufficient acidic species have aggregated on a particle, can the protons act to solubilize iron.
The mineralogy of aerosol iron is also key to understanding the acidic reactions that are likely at work. Heterogeneous mineralogy of aerosol iron can lead to iron with different levels of susceptibility to acid mobilization that act at different time scales [28,30,78]. Acidification of mineral dust has shown that iron will first undergo rapid iron dissolution, followed by a continued iron release at a much slower rate [78]. These differences in the rate of iron dissolution were linked to mineralogy, where the rapidly solubilizing iron is likely ferrihydrite, an amorphous iron oxide. The iron phases that undergo slow dissolution under acidic conditions are likely crystalline iron oxide phases and clay minerals [78]. Generally, amorphous iron oxides, like ferrihydrite, account for only 2% of aerosol iron, while more crystalline iron minerals and clays can account for up to 40% of the total aerosol iron [78]. The dominance of Fe(III) oxides in our aerosols (Figure 3), is consistent with the observed low to medium levels of fractional iron solubility.
increases, up to nearly 35%. This behavior is consistent with the speciation of Fe(III), determined using the standard selection in the Mineql + database to create an Fe(III) speciation diagram [74], where below a pH of approximately 3, the abundance of dissolved Fe(III) sharply increases [75][76][77]. This suggests that dissolved Fe(III) production is a consequence of acidic reactions in the atmosphere as would be expected for acidic species acting on a largely Fe(III) oxide composition. So only in the acidic range, when sufficient acidic species have aggregated on a particle, can the protons act to solubilize iron. The mineralogy of aerosol iron is also key to understanding the acidic reactions that are likely at work. Heterogeneous mineralogy of aerosol iron can lead to iron with different levels of susceptibility to acid mobilization that act at different time scales [28,30,78]. Acidification of mineral dust has shown that iron will first undergo rapid iron dissolution, followed by a continued iron release at a much slower rate [78]. These differences in the rate of iron dissolution were linked to mineralogy, where the rapidly solubilizing iron is likely ferrihydrite, an amorphous iron oxide. The iron phases that undergo slow dissolution under acidic conditions are likely crystalline iron oxide phases and clay minerals [78]. Generally, amorphous iron oxides, like ferrihydrite, account for only 2% of aerosol iron, while more crystalline iron minerals and clays can account for up to 40% of the total aerosol iron [78]. The dominance of Fe(III) oxides in our aerosols (Figure 3), is consistent with the observed low to medium levels of fractional iron solubility.
Alteration of clay mineral phases to produce soluble iron minerals, such as ferrihydrite, has been shown to occur as a result of chemical reactions during atmospheric transport [79]. Traditional indicators of aerosol acidity, such as pH, suggest that acidic reactions are likely solubilizing iron, and our observations of iron phosphates and Fe(II and III) sulfates provide further support that acidic reactions are occurring. Although iron phosphates and sulfates may be found locally at emission sources, they do not typically occur in natural environments at high concentrations. To account for the concentrations of Alteration of clay mineral phases to produce soluble iron minerals, such as ferrihydrite, has been shown to occur as a result of chemical reactions during atmospheric transport [79]. Traditional indicators of aerosol acidity, such as pH, suggest that acidic reactions are likely solubilizing iron, and our observations of iron phosphates and Fe(II and III) sulfates provide further support that acidic reactions are occurring. Although iron phosphates and sulfates may be found locally at emission sources, they do not typically occur in natural environments at high concentrations. To account for the concentrations of these phases seen in aerosol samples across all sampling locations, secondary atmospheric processes are likely at work. Iron phosphates and sulfates may form as a result of acidic processes that solubilize iron and phosphorus minerals through reactions with sulfuric and nitric acid species. One likely source of phosphorus mobilized under acidic conditions is the mineral apatite which has been identified as a common constituent in aerosols [80]. Such acidic processes can create an aqueous solution rich in dissolved iron, phosphorus, and sulfur which favor the precipitation of secondary mineral phases [23]. However, acids can be buffered by reactions associated with carbonate minerals commonly found in mineral dust [66,81]. Furthermore, aerosol pH undergoes diel fluctuations [82] that will cause the process of acidic transformation to be intermittent and slow. Three to five days are necessary to solubilize 1-2% of iron under similar acidic conditions of a deliquescent mineral dust particle [66,83]. Varying levels of reactive carbonate minerals in dust combined with the time-dependent impact of diel fluctuations on pH would lead to the observed variations in the concentrations of secondary phosphates, sulfates, and Fe(III) oxides in particles.
Aerosol source region may also influence the solubility of iron [84,85]. In this study, three sampling locations received air masses from different source regions: Finokalia Research Station, BIOS, and the GEOTRACES transect. At Finokalia Research Station, two distinct air masses were collected-anthropogenically influenced European air masses and mineral-dust laden North African air masses. North African air masses exhibit low fractional iron solubility and high pH, while samples from European air masses have higher fractional solubility and lower pH, possibly due to the anthropogenic contributions of nitrates and sulfates, affecting pH in these air masses [84,85]. At the Bermuda location, no correlations between aerosol iron solubility and source region, North African vs. North American, were present. However, the effects of source region on aerosol iron solubility could be subtly seen, where the average solubility of iron from North American air masses (6.3 ± 2.3%) was 1.5 times higher than the average iron solubility of North African air masses (3.8 ± 2.8%). Higher solubility in anthropogenically influenced aerosols may in part be influenced by their co-emission with acidic gases or their higher solubility at emission, in the case of biomass burning [31,73,[85][86][87]. Samples collected on the GEOTRACES cruise also showed trends in iron solubility that were not apparent in the larger dataset. The composition (R 2 = 0.53) and oxidation state (R 2 = 0.64) were correlated with iron solubility ( Figure 5). The higher levels of Fe(II) present in marine aerosol samples corresponded to increased iron solubility, which could be evidence of a reductive process, such as photoreduction. Moreover, these marine samples also had the highest quantities of Fe(II and III) sulfates and Fe(III) phosphates, phases that can be formed as a result of secondary processes that occur in the atmosphere.
Atmosphere 2018, 9, x FOR PEER REVIEW 12 of 17 Figure 5. In samples from the GEOTRACES collection, the fractional solubility of iron was higher when more Fe(II) (black line) was present or when more secondary phases (grey bars) were present. Secondary phases include Fe(II and III) sulfates and iron phosphates; these compounds are collectively labeled as secondary phases because these are phases that may form through secondary reactions that occur in the atmosphere. However, such secondary phases may also be initially present in aerosols derived from coal fly ash or volcanic materials.

Conclusions
This study presents an analysis of aerosol soluble iron characteristics from five different locations, on both a bulk-mass and individual-particle basis. In general, the total and soluble-iron trends seen in this data compilation are similar to those seen in other aerosol iron data sets for specific regions [33,34,47,73,89,92,93]. For aerosol iron, samples that are high in total iron have low fractional iron solubility, and samples with low total iron have higher fractional iron solubility [86]. This could reflect the influence of aerosol source region, where aerosols with high total iron tend to come from mineral dust sources, which can have significant buffer capacity. This contrasts with aerosols from anthropogenically influenced air masses that generally have lower buffer capacity and more acidic species or contribution from combustion sources with high iron solubility [94].
In the natural environment, the various chemical transformations of aerosols occurring during atmospheric transport may not only occur on different time scales, but the mechanisms can also have complex interactions with each other. For example, one study noted marine samples to be 5-17 times more soluble than samples containing significant anthropogenic influence when acidified to the same pH, and this increased solubility was attributed to conversion of Fe(III) to Fe(II) via photoreduction [65]. While in our data, acidification appears to be a key mechanism that affects aerosol iron regardless of source region, the findings of Zhuang et al. [17] demonstrate that one factor alone cannot be considered to control the solubility of aerosol iron. The observed variations in composition, oxidation state, pH, and solubility of aerosol iron shown here suggest that chemical transformations occur during atmospheric transport of aerosols originating from diverse source regions. Acidification and source region appear to play a significant role in determining the solubility of aerosol iron, but these data also show that one universal rule of iron solubility should not be applied to estimate the solubility of aerosol iron on a global scale. Better quantification of iron speciation and better understanding of the evolution of iron solubility in aerosols is needed. This in turn emphasizes the need to conduct further studies on environmental Fe Solubility (%) Figure 5. In samples from the GEOTRACES collection, the fractional solubility of iron was higher when more Fe(II) (black line) was present or when more secondary phases (grey bars) were present. Secondary phases include Fe(II and III) sulfates and iron phosphates; these compounds are collectively labeled as secondary phases because these are phases that may form through secondary reactions that occur in the atmosphere. However, such secondary phases may also be initially present in aerosols derived from coal fly ash or volcanic materials.
Studies of mineral dust have also suggested a physical control over the solubility of aerosol iron, where larger particles are removed during atmospheric transport [88]. Compositional differences or changes in surface area to volume ratio in different particle size classes could have a potentially strong influence iron solubility [88,89]. Differences in sample collection techniques for the sample sets from different regions in this study resulted in differences in the size fractions collected. However, no clear trends in solubility emerged when comparing samples of different size fractions in our data set [88]. Another potential control of aerosol iron solubility is the presence of ligands that are thought to stabilize Fe(II) in solution and can also participate in ligand mediated photochemical reactions [90,91]. These physical processes along with ligand promoted iron dissolution are mechanisms that were not evaluated in this study. Variations in the impact of these physical and chemical processes in ambient aerosol could help explain the lack of direct overall correspondence between aerosol iron solubility, composition, and oxidation state in the global dataset.

Conclusions
This study presents an analysis of aerosol soluble iron characteristics from five different locations, on both a bulk-mass and individual-particle basis. In general, the total and soluble-iron trends seen in this data compilation are similar to those seen in other aerosol iron data sets for specific regions [33,34,47,73,89,92,93]. For aerosol iron, samples that are high in total iron have low fractional iron solubility, and samples with low total iron have higher fractional iron solubility [86]. This could reflect the influence of aerosol source region, where aerosols with high total iron tend to come from mineral dust sources, which can have significant buffer capacity. This contrasts with aerosols from anthropogenically influenced air masses that generally have lower buffer capacity and more acidic species or contribution from combustion sources with high iron solubility [94].
In the natural environment, the various chemical transformations of aerosols occurring during atmospheric transport may not only occur on different time scales, but the mechanisms can also have complex interactions with each other. For example, one study noted marine samples to be 5-17 times more soluble than samples containing significant anthropogenic influence when acidified to the same pH, and this increased solubility was attributed to conversion of Fe(III) to Fe(II) via photoreduction [65]. While in our data, acidification appears to be a key mechanism that affects aerosol iron regardless of source region, the findings of Zhuang et al. [17] demonstrate that one factor alone cannot be considered to control the solubility of aerosol iron. The observed variations in composition, oxidation state, pH, and solubility of aerosol iron shown here suggest that chemical transformations occur during atmospheric transport of aerosols originating from diverse source regions. Acidification and source region appear to play a significant role in determining the solubility of aerosol iron, but these data also show that one universal rule of iron solubility should not be applied to estimate the solubility of aerosol iron on a global scale. Better quantification of iron speciation and better understanding of the evolution of iron solubility in aerosols is needed. This in turn emphasizes the need to conduct further studies on environmental samples, both in situ and under controlled conditions, to quantify physical processes and chemical transformations that happen to aerosols during atmospheric transport and their effects on iron solubility.

Supplementary Materials:
The following are available online at http://www.mdpi.com/2073-4433/9/5/201/s1, Figure S1: Pre-edge Centroid vs Fe Solubility, Table S1: Sample collection information for India, Southern Ocean, Finokalia Research Station and GEOTRACES samples, Table S2: Sample collection information for samples collected at BIOS, Table S3: Sample Collection Locations, Table S4: BIOS Aerosol Data, Table S5: India, GEOTRACES, Southern Ocean and Finokalia Research Station Aerosol Data, Table S6: Correlation values for iron solubility, Table S7: Composition of bulk aerosol and individual aerosol particle samples