Iron Speciation in PM 2.5 From Urban, Agriculture, and Mixed Environments in Colorado, USA

Atmospheric iron solubility varies depending on whether the particles are collected in rural or urban areas, with urban areas showing increased iron solubility. In this study, we investigate if the iron species present in different environments affects its ultimate solubility. Field data are presented from the Platte River Air Pollution and Photochemistry Experiment (PRAPPE), aimed at understanding the interactions between organic carbon and trace elements in atmospheric particulate matter (PM). 24 ‐ hr PM 2.5 samples were collected during the summer and winter (2016 – 2017), at three different sites on the Eastern Colorado plains: an urban, agricultural, and a mixed site. Downtown Denver had an average total and water ‐ soluble iron air concentration of 181.2 and 7.7 ng m − 3 , respectively. Platteville, the mixed site, had an average of total iron of 76.1 ng m − 3 , with average water ‐ soluble iron concentration of 9.1 ng m − 3 . Jackson State Park (rural/agricultural) had the lowest total iron average of 31.5 ng m − 3 and the lowest water ‐ soluble iron average, 1.3 ng m − 3 . The iron oxidation state and chemical speciation of 97 samples across all sites and seasons was probed by X ‐ ray absorption near edge structure (XANES) spectroscopy. The most common iron phases observed were almandine (Fe ₃ Al ₂ Si ₃ O ₁₂ ) (Denver 21%, Platteville 16%, Jackson 24%), magnetite (Fe 3 O 4 ) (Denver 9%, Platteville 4%, Jackson 5%) and Fe (III)dextran (Denver 5%, Platteville 13%, Jackson 5%), a surrogate for Fe ‐ organic complexes. Additionally, native iron [Fe(0)] was found in signi ﬁ cant amounts at all sites. No correlation was observed between iron solubility and iron oxidation state or chemical speciation. being related diesel vehicles in Platteville and Jackson resulting lower of magnetite.


Introduction
The presence of iron in the ecosystem has major implications on various geochemical cycles (carbon and sulfur), redox environments, and nutrient availability (Le & Ricard, 1999;Mahowald et al., 2005;Wang & Cappellen, 1996). Iron serves as a necessary nutrient for primary producers, such as algae, and acts as a limiting nutrient in roughly half of the world's oceans (Moore et al., 2013). In both biological and environmental media, the water-soluble iron fraction has been demonstrated to result in the creation of reactive oxygen species (ROS) (Faiola et al., 2011;Hamad et al., 2016;Park et al., 2006), likely generated through Fenton chemistry (Faiola et al., 2011;Kuang et al., 2017;Park et al., 2006). In biological systems, following the inhalation of PM 2.5 (Brauer et al., 2001), the production of ROS results in oxidative stress to the respiratory system and is suspected to be related to pulmonary inflammation, DNA damage, and the oxidation of proteins and lipids (Landreman et al., 2008;Park et al., 2006;Verma et al., 2014). Given iron's importance from both a health and ecological standpoint, and that these processes are dependent on the availability of iron, the primary driver of this study is to better understand what controls atmospheric iron's solubility.
Iron is the most abundant transition element in the atmosphere and results from windblown dust and urban air pollution Mahowald et al., 2009). Typically, water-soluble iron from crustal sources is less than 1% of the total iron (Paris et al., 2010). This greatly differs from anthropogenic sources where water-soluble iron can range from 5-50% of total iron (Majestic et al., 2007;Petroselli et al., 2018;Salazar et al., 2020;Sedwick et al., 2007;Sholkovitz et al., 2012). The majority of anthropogenic water-soluble iron is estimated to result from combustion sources (Luo et al., 2008). The resulting iron water-solubility is hypothesized to result from an interaction between organics and iron (III) (Haynes & Majestic, 2020;Paris & Desboeufs, 2013;Salazar et al., 2020). ©2020. The Authors. This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited.
In addition to giving insight into organic-iron interactions, the iron speciation may be an important predictor of its solubility. Synchrotron-based X-ray absorption near edge structure (XANES) spectroscopy has been used to investigate the Fe oxidation state and chemical speciation (iron's surrounding ligands) in ambient PM. Prior reports making use of XANES have attempted to describe the relationship between water-soluble iron and its chemical speciation Oakes et al., 2012;Takahashi et al., 2011). Several ambient studies observed that Fe (III) is the principal oxidation state in PM 2.5 , primarily as Fe-Al and Fe-Al-Si complexes, originating from windblown dust and soils (Fittschen et al., 2008;Oakes et al., 2012).
In the present campaign, the primary purpose is to better understand how iron speciation and oxidation state can affect iron water solubility across geographical areas and seasons. To achieve this goal, PM 2.5 was collected from three sites of multiple types: urban, agricultural, and mixed, and in both the winter and the summer. Samples were collected from each site and 97 Fe K-edge XANES spectra in total (31 from Denver [urban], 35 from Platteville [mixed], and 31 Jackson [agricultural]) were analyzed along with total and water-soluble iron. The results were examined to understand if any patterns in season and source influence iron speciation, water-solubility, and oxidation state.

Materials Preparation
All vessel cleaning and analytical preparation was performed under a laminar flow hood with incoming air passing through a high-efficiency particulate air (HEPA) filter (NuAire, Plymouth, MN). All water used was purified to 18.2 MΩ-cm (Milli-Q Thermo-Fisher Nanopore). An acid cleaning process was used to prepare the following: 15-and 50-ml plastic centrifuge vials, Petri dishes (Fisher), Teflon forceps (Fisher), syringes (Fisher), and 47 mm Teflon filters. For the plastic centrifuge vials, Petri dishes, Teflon forceps, syringe, and syringe cases this involved 24-hr soaks in a 10% reagent grade nitric acid bath, followed by 10% reagent grade hydrochloric bath, then a 3% trace metal grade nitric acid (Fisher) resting bath with 18.2 MΩ-cm water rinses before, after, and between each step. Syringe filters (0.45-micron) (Whatman, Marlborough, MA) were prepared with 10% trace-metal grade hydrochloric acid, MQ water, and 5% nitric acid rinse. All materials were handled with powder-free nitrile gloves (Fisher), double-bagged.

Sample Collection
PM 2.5 was collected at three sites in Colorado's eastern plains: (1) the Colorado Air Monitoring Program (CAMP) site in downtown Denver, an urban setting, (2) Jackson Lake State Park, a rural and agricultural setting, and (3) Platteville Elementary School in Platteville, CO, a mixed agricultural and urban setting that is geographically situated between the other sites ( Figure 1).
Air was passed through a cyclone (URG 2000−30 ENB) at 92 L min −1 , resulting in a PM 2.5 size cut. Each site had two sampling lines. One line collected PM 2.5 on a water cleaned Teflon filter for ion quantification and water-soluble organic experiments. The other sampling line was further split into two lines: one of the lines was collected onto a prebaked quartz filter, and the other line was collected onto an acid-cleaned Teflon filter. The acid-cleaned Teflon filter was used for overall mass determination and trace metal measurements. The total flow was measured for each filter by a flow totalizer (Honeywell BK-G4 Meter).
One campaign was conducted in winter 2016-2017 and one in summer 2017. Samples were collected in 24hr intervals (midnight to midnight), automated using a 7-day timer (INTERMATIC Electromechanical Timer, 7-Day, SPDT, 21A) every other day in December 2016, February 2017, and August 2017. Exposed Teflon filters were stored in Petri dishes sealed with Teflon tape, double-bagged, and stored in a freezer (−20°C) until analysis.

Water-Soluble Element Preparation
Water-soluble iron was extracted from the Teflon filters for 12 hr on a shaker table in 15 ml of 18.2 MΩ-cm water. The water extract was filtered with 0.45 μm pore size polyethersulfone (PES) syringe filters. This filtered water-soluble extract was acidified to 5% trace-metal grade nitric acid and 2.5% trace-metal grade hydrochloric acid and analyzed by inductively coupled plasma mass spectrometry (ICP-MS, Agilent 7,700).

Elemental Analysis
The polymethylpentene ring was removed from the Teflon filters and about 4% (measured exactly) of the filter was removed and used for Fe XANES spectroscopy. The remaining (~96%) Teflon filter for each sample was placed into a microwave digestion vessel. In each digestion vessel, 750 μL of concentrated trace metal grade nitric acid, 250 μL of concentrated trace metal grade hydrochloric acid, 100 μL of concentrated trace metal grade hydrofluoric acid, and 100 μL of 30% hydrogen peroxide were added. The filters with the collected PM 2.5 were digested (Ethos EZ, Milestone Inc.) according to the following temperature program: 15-min ramp to 200°C, then held at 200°C for 25 min, and a 60-min cooling period . Finally, the solution was removed and diluted to 15 mL with MQ water and analyzed via ICP-MS.
Blank filters and standard reference materials (SRMs) were digested alongside the collected PM 2.5 using the same digestion process described above. Two SRMs were used to address the recoveries of Fe in the digestion process: urban particulate matter (1648a, NIST) and San Joaquin Soil (2709a, NIST). The recoveries of the SRMs were between 80% and 120%. Indium (~1 ppb) was used as an internal standard and a He collision cell was used to remove polyatomic interferences.

XANES Spectroscopy
Fe K-edge X-ray absorption near-edge structure (XANES) and X-ray fluorescence (XRF) mapping data for 39 samples were collected at the Advanced Light Source Microprobe beamline (10.3.2), at Lawrence Berkeley National Laboratory, Berkeley, CA (Marcus et al., 2004). A broad XRF map of each sample was acquired at 10 keV with 20 μm by 20 μm pixel size and 80 ms dwell time per pixel to locate the concentrated iron spots on the filter. Where iron was concentrated, a fine map was collected with 8 μm by 8 μm pixel size and 80 ms dwell time per pixel. From this map, 2-3 iron spots per filter were chosen for Fe K-edge extended XANES analysis (totaling 97 spectra). All data were recorded using a seven-element solid state Ge detector (Canberra, ON) in fluorescence mode by continuously scanning the Si (111) monochromator (Quick XAS mode) in the range of 7,010-7,415 eV. The spectra were then deadtime corrected, deglitched, calibrated, pre-edge background subtracted and post-edge normalized using a suite of LabVIEW custom programs (Marcus et al., 2008) available at the beamline. Spectra were calibrated using an iron foil with first

10.1029/2020EA001262
Earth and Space Science derivative maximum set at 7,110.75 eV (Kraft et al., 1996). Least squares linear combination fitting (LCF) was performed in the range 7,090 to 7,365 eV to identify oxidation states and iron speciation following methods previously described in (Westphal et al., 2009) using a large updated 10.3.2 XAS database of Fe bearing compounds and custom LabVIEW software available at the beamline. Determination of the mineral groups and in some cases mineral identity can be obtained (+/− 10% error estimate). Fe valence scatter plots were generated from Fe XANES data, producing a two-dimensional plot where each datapoint represents a XANES spectrum, following procedures described elsewhere (Marcus et al., 2008). Note that this technique needs to be applied in conjunction with LCF to confirm Fe valence.

Water-Soluble Iron and Total Iron
Total and water-soluble iron from each region (urban, agricultural, and mixed) differed in averages and from day to day, and the results for all sites are presented in Figure 2. Across all seasons, downtown Denver had an The extent of seasonal variations differs for the total and the water-soluble iron (t test, p < 0.05). Across all sites, winter had lower water-soluble iron with the average being 2.3 ng m −3 (range; 0.17-5.6 ng m −3 ), while summer average of 8.6 ng m −3 (range; 0.6-27.8 ng m −3 ). For total iron, however, winter was greater than the summer, on average. The winter average was 107.3 ng m −3 (range; 9.6-380.5 ng m −3 ), and the summer average was 86.8 ng m −3 (range; 7.9-247.2 ng m −3 ). Across all sites, iron fractional solubility for the winter was 2.1% and for the summer, it was 9.9%.

Iron Oxidation States
An initial valence state scatter plot was generated using iron's K edge normalized XANES data to quickly classify particles according to iron oxidation states (Figure 4). The oxidation states observed were Fe(0), Fe (II), Fe (III), and mixed [a mixture of both Fe (II) and Fe (III)]. On a relative basis, the Jackson site observed the most Fe(0) and Fe (II) with less of the mixed and Fe (III). The Platteville site showed a large amount of Fe (II) and mixed iron phases. Denver had most samples fall in the mixed region followed by Fe (III) and then Fe (II).
A mixture of multiple iron mineral groups can be found for the iron that falls into the mixed oxidation state region in the valence plot. Using LCF, the fraction of Fe (II) and Fe (III) that are in the mixed area of the valence plot were obtained, adding more detail to the chemical speciation of iron in this region ( Figure 5). Figure 5 shows two example Fe-XANES plots showing the weighted iron standards involved in the LCF of the data. These percentages were obtained for each by summing the percentages for each oxidation state. Fe with mixed oxidation state were labeled as "mixed oxidation state." The average oxidation states for each iron sample per site are as follows: Denver Fe (III) 37%, Denver Fe (II) 39%, Platteville Fe (III) 44%, Platteville Fe (II) 32%, Jackson Fe (III) 27%, and Jackson Fe (II) 34%. Fe(0) was ubiquitous at all three sites, but to varying degrees: Jackson 27%, Platteville 13%, and Denver 3%. Fe (II) and Fe (III) were also compared between winter and summer for all sites. Fe (II) was shown to be 31% of all particles tested in the winter. In the summer, Fe (II) accounted for 39% of the total iron tested. Fe (III) was found to be 39% in the winter of iron and in the summer Fe (III) shown to be 30% of iron.
As Fe (II) is generally more water-soluble than Fe (III), the particle-bound oxidation states were compared to the percent water-soluble iron. Fe (II) and water-soluble iron by day showed no significant relationship

10.1029/2020EA001262
Earth and Space Science SALAZAR ET AL.
(R 2 = 0.03). Fe (III) was tested against water-soluble iron and no relationship was found between Fe (III) and water-soluble iron (R 2 = 0.00). Generally, no relationship between iron solubility and particle-bound oxidation state was observed for any iron species.

Iron Mineralogy
As no correlation was found between the iron oxidation state and its water-solubility, we probed deeper by investigating how iron speciation (i.e., ligands and immediate chemical environment) may affect the water solubility. From the XANES spectra and the LCF, the percent of iron species were defined for each iron spot, then averaged together to look at the most common iron species for each site ( Table 1). The most common iron species for all three sites was the mineral almandine (Fe₃Al₂Si₃O₁₂), which is an iron-aluminum-silicate complex. Almandine accounted for 21% of iron particles in Denver, 16% of iron particles in Platteville, and 24% of iron particles in Jackson. The second most common iron species was magnetite (Fe 3 O 4 ) with 9% in Denver, 4% in Platteville, and 5% in Jackson. The third most common iron species was Fe (III)dextran, a surrogate for an iron-organic complex, which accounted for 5% of iron in Denver and Jackson and 13% of iron in Platteville.

Total and Water-Soluble Iron
The Colorado Coarse Rural-Urban Sources and Health (CCRUSH) study found that, in residential neighborhoods in Denver, total iron in PM 2.5 was 89 ng m −3 at Alsup Elementary and 56 ng m −3 Edison Elementary (Clements et al., 2014). These locations were in different parts of Denver at distances less than 11 km from the CAMP sampling site. These concentrations were less than the average 181.2 ng m −3 measured in this study. In 2014, another study done in the same Denver location as the current study showed an average iron in PM 2.5 of 190 ng m −3 (Clements et al., 2014). This result illustrates the difference in concentrations in downtown versus residential neighborhoods within a major metropolitan area. Water-soluble iron was also measured in Denver in 2012 January to February averaging 29 ng m −3 which is more than the 7.7 ng m −3 measured in this study. However, those samples were extracted in a 0.5 mM acetate buffer (pH = 4.25) compared to the pure 18.2 MΩ-cm water used here, and the acetate buffer is known to give a higher iron solubility compared to pure water Majestic et al., 2006).
Iron water-solubility and oxidation state were examined for summer and winter. There was no change for iron oxidation state in winter and summer. Iron water-solubility was 3.8 times higher in the summer (8.6 ng m −3 ) than in the winter (2.3 ng m −3 ). Water-soluble iron was not correlated to either particle-bound Fe (II) or Fe (III). This evidence indicates that iron oxidation state is not a factor in iron solubility. In the summer, we observed increased iron solubility, with a decreased total iron load, relative to winter. This scenario is consistent with the iron being derived from a pyrogenic source (Sholkovitz et al., 2012), probably from the record-breaking (at the time) wild fires prevalent across the Western United States, which historically impact the air quality at this study's test sites (Creamean et al., 2016).

Mineralogy
A surprisingly high fraction of the Fe found in Jackson (27%) and Platteville (13%) is in the native Fe(0) oxidation state. In solution, Fe(0) produces the most ROS compared with Fe (II) and Fe (III) (Phenrat et al., 2009). To our knowledge, this is the first study to measure native iron in the atmosphere in any significant amount.
Fe (II) aluminosilicates were found to be the most common form of iron at all three sites. Because of the association with aluminum and silicate, this shows that at least 20% of iron found in PM 2.5 in eastern Colorado is from crustal sources. Fe (II) aluminum silicate complexes have been previously found in other geographical areas (Oakes et al., 2012;Petroselli et al., 2018). Petroselli et al. collected a PM 2.5 sample near a steel plant in Terni, Italy and measured Fe (III) as the predominate species in this sample. Oakes et al. collected four samples in both rural and urban sites and, similar to the results presented here, found a large amount of Fe-aluminosilicates in all sites and did not show a variation in iron speciation from the two sites, an effect attributed to low sample size. Here, we had a larger sample size and observed no seasonal variation with respect to Fe-aluminosilicates, which helps to confirm the results from Oakes et al.
Magnetite, a mixed Fe (II,III) oxide has been found to originate from diesel vehicles and is also present in crustal sources (Abdul-Razzaq & Gautam, 2001;Oakes et al., 2012;Petroselli et al., 2018). Here we identified magnetite in 9% of iron tested in Denver, 5% in Platteville, and 4% in Jackson (Table 1). The increase of magnetite in Denver suggests that it can originate from vehicular activity in Denver. There is less evidence for Fe (III)dextran is an iron-sugar complex. While likely not directly found in the atmosphere, Fe (III)dextran is used as a surrogate for iron complexed to oxidized organic species. Water-soluble Fe has been hypothesized to primarily result from biomass burning, whose emissions contain sugars like levoglucosan, mannosan, and galactosan (Scaramboni et al., 2015). Concentration of sugars from biomass burning can range from 6 ng m −3 to 12.5 μg m −3 (Scaramboni et al., 2015;Theodosi et al., 2018), and they are highly water-soluble (Barbaro et al., 2019;Simoneit et al., 2004). Water-solubilization of iron by theses sugar complexes in biomass burning fumes could be a large contributor to total water-soluble iron. However, since these organic biomass sugar compounds were not measured in this study, this could not be directly tested here.

Conclusion
Atmospheric iron was probed from three different environments in the eastern plains of Colorado. The most common mineral found was almandine, which accounted for 20% of total iron tested. The large prevalence of the Si-Fe complexes suggests a large contribution from natural, crustal sources, even within urban areas. Fe (III)dextran, an organic iron complex and magnetite were also abundant. Iron oxidation states were compared seasonally, where no change in oxidation state was observed; however, an increase in iron water-solubility was present in the summer. Finally, water-soluble iron was compared to Fe (II) and Fe (III), and results indicated there is no relationship between iron oxidation state and water-soluble iron.
There was no observed relationship between atmospheric iron solubility and the chemical speciation of iron at any of the sites. This result suggests that overall atmospheric iron is dependent on other factors aside from its oxidation state and chemical speciation. Previous studies have shown that the pH of the extract solution is an important parameter in addressing iron solubility (Stumm & Morgan, 1996). Therefore, nonferrous acidbase active species may be important contributors to the solubility in aqueous extracts. Additionally, iron may be solubilized by interactions with organic compounds that are present in the aerosol and, thus, the extract solution. Recent studies have shown that aerosols rich in specific combustion-derived organic species lead to increased aqueous iron solubility (Ito et al., 2019;Salazar et al., 2020). To fully understand atmospheric iron solubility, it is becoming clearer that researchers must look at contributing species, such as organics and pH-active compounds, and not solely at the iron.

Data Availability Statement
Data archiving is provided at the Digital Commons @ DU (https://digitalcommons.du.edu), with this data set residing at this site (https://digitalcommons.du.edu/chemistrydatasets/1/). These data sets are publicly available and contain the raw ICPMS and XANES data, along with the least squares fitting results.