Stable Isotopes as a Tool to Apportion Atmospheric Iron - American

May 7, 2009 - BRIAN J. MAJESTIC,* , †. ARIEL D. ANBAR, †,‡. AND. PIERRE HERCKES †. Department of Chemistry and Biochemistry, and School of...
0 downloads 0 Views 532KB Size
Environ. Sci. Technol. 2009, 43, 4327–4333

Stable Isotopes as a Tool to Apportion Atmospheric Iron B R I A N J . M A J E S T I C , * ,† A R I E L D . A N B A R , †,‡ A N D PIERRE HERCKES† Department of Chemistry and Biochemistry, and School of Earth and Space Exploration, Arizona State University, PO Box 871604, Tempe, Arizona 85287-1604

Received January 5, 2009. Revised manuscript received April 13, 2009. Accepted April 15, 2009.

Identification of atmospheric iron is a key parameter to understanding the source of iron in urban and remote areas. Atmospheric deposition of desert dust, which also can include an anthropogenic component, is a primary nutrient source for most of the open ocean. To better assess particulate matter (PM) sources specific to iron, we measured the iron isotopic composition of aerosols in two size fractions: PM with aerodynamic diameters less than 2.5 µm and less than 10 µm (PM2.5 and PM10, respectively). Using colocated samplers, atmospheric aerosol samples were collected in the U.S. desert Southwest at a mixed suburban/agricultural site near Phoenix, AZ. The measurements are presented as δ56Fe relative to the IRMM014 (Institute for Reference Materials and Measurements) standard. Using multiple collector inductively coupled plasma mass spectrometry, we found differences in iron isotopic composition within the PM10 aerosol. Half of the PM10 samples had an iron isotopic signature similar to crustal material (+0.03 ‰), which implicates wind-blown soil-dust as the primary source. The other PM10 samples showed a lighter iron isotopic composition, centered at -0.18 ‰. Further analysis showed that the lighter iron was associated with winds originating from the southwest. This strongly suggests that there is a different PM10 source in this direction, with a distinct iron isotopic composition. The iron in the PM2.5 samples was usually substantially lighter than the corresponding PM10 samples, which is consistent with coarse and fine particles having different sources, again with distinctively different isotopic compositions. The magnitude of the iron isotopic difference between the PM10 and the PM2.5 size fractions (δ56FePM10 - δ56FePM2.5) correlated with the PM2.5 concentrations of elements known to be emitted from industrial sources (Pb, Cd, As, V, and Cr). This observation implies that the isotopically light iron is created or emitted alongside industrial processes. Our data demonstrate that iron isotope composition can be a valuable tool in the sourceapportionment of iron in atmospheric particles.

Introduction Iron, the most abundant transition metal in the Earth’s crust, is central to many environmental and biological processes (1, 2). In particular, iron is a limiting nutrient to large parts of the open ocean (3). Atmospheric deposition is a crucial * Corresponding author phone: (480) 204-9515; [email protected]. † Department of Chemistry and Biochemistry. ‡ School of Earth and Space Exploration. 10.1021/es900023w CCC: $40.75

Published on Web 05/07/2009

e-mail:

 2009 American Chemical Society

source of iron to these environments (4), leading to interest in tracing both natural and anthropogenic sources of iron in the atmosphere (5, 6). Human activities have been shown to significantly impact the bioavailable fraction of atmospheric iron (7, 8). Therefore, in order to assess the human impact on ecosystems, it would be important to differentiate anthropogenic from natural iron sources. Generally, source-apportionment studies focus on identifying the sources of particulate matter using either organic molecular markers or elemental data (9, 10). Sourcing of a single element (e.g., iron) can be performed using positive matrix factorization (PMF), where the source of iron (and other elements) is determined by measuring the association of iron with other components of the aerosol (11). A more powerful approach to sourcing iron is to directly measure the chemical characteristics of iron in aerosols, and to trace those back to different sources. For example, microscopy and X-ray absorption techniques have been successful in identifying different chemical forms of iron in atmospheric particles (12, 13). Stable isotopes have also recently been used for source-identification of various metals in atmospheric aerosols (14-16). Here, we assess the potential of stable iron isotope variations as a method to identify pools of iron from different sources in ambient atmospheric aerosol samples. Variations in iron isotopic composition are known to arise from physical, chemical, and biological processes that fractionate stable (i.e., nonradioactive) iron isotopes (1, 17-19). These variations can be measured with multiple collector inductively coupled plasma-mass spectrometry (MC-ICPMS). In environmental samples, deviations in iron isotopic composition on the order of (3 ‰ have been measured, greatly exceeding the analytical uncertainties ((0.08 ‰) (20). To date, the iron isotopic composition of atmospheric aerosols has only been reported in a few studies (21-23). Only one study has focused on tropospheric urban aerosols (22). In this study, we examine iron isotopes in aerosols and potential source materials in the U.S. desert Southwest, at a suburban site about 30 km southeast of Phoenix, AZ. Iron in this region is mostly crustally derived, but anthropogenic iron sources such as automobile brake wear, power generation, and industrial sources (smelters and foundries) are also important sources (24). This extreme diversity of iron sources, with potentially distinct iron isotope compositions, presents an opportunity to assess the use of iron isotopes to trace the sources of iron in atmospheric particulate matter.

Materials and Methods Plastics Cleaning. Unless otherwise noted, all acids used were either trace metal grade (Fisher) or reagent grade acid (Fisher) distilled at Arizona State University using trace metal clean Teflon stills in a Class 10 clean hood. Hydrogen peroxide was ULTREX Ultrapure II grade (J.T. Baker). All aerosol and soil samples were handled in Class 10 trace metal clean hoods under ULPA filters. Teflon vials (Savillex) were cleaned in sub-boiling 50% reagent grade nitric acid for at least 36 h. The vessels were then moved to a hot 50% reagent grade hydrochloric acid bath for 36 h. Finally, the vessels were placed in a 2% distilled HCl acid bath at room temperature for at least 24 h, rinsed three times with >18.0 MΩ water and dried under an ULPA filter. All analysis vessels (polypropylene or low density polyethylene) were cleaned at room temperature in 20% nitric acid for one week, followed by one week in 20% hydrochloric acid and rinsed three times with high purity water. VOL. 43, NO. 12, 2009 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

4327

FIGURE 1. Elemental composition of the PM2.5 and PM10 aerosols. Most of the metal-mass is crustal in origin. The very low values during the last two days of sampling are a result of a rain storm during these sampling periods. Aerosol Sampling. Colocated PM2.5 and PM10 samples were collected at a site maintained by the Maricopa County Air Quality Department, located in a mixed suburban/ agricultural area about 30 km southeast of Phoenix, AZ (AQS Code: 04-013-4006). The Phoenix area is currently the fifth most populated metropolitan area in the United States, so there is substantial vehicular traffic. The region of study is in the U.S. desert and is therefore arid, with substantial amounts of PM resulting from wind-blown dust. Particulate matter sampling was achieved using two Partisol-Plus model 2025 air samplers (Thermo Scientific) operating for eight 48-h periods at 16.7 Lpm. Teflon filter membranes (47 mm) with polypropylene support rings (Whatman) were used as 4328

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 43, NO. 12, 2009

collection substrates. Simultaneous collection of PM2.5 and PM10 aerosols allowed for a direct comparison of iron isotopic composition between two different size fractions. Elemental Analysis. After removing the polypropylene support ring from the filter membrane with a ceramic blade, the aerosols were digested in 22 mL precleaned Teflon vials. Initially, 1 mL of HF and 4 mL of HNO3 were added to the Teflon vial and heated for 10 h on a metal-free hot plate (surface temperature ) 140 °C). The HF/HNO3 mixture was evaporated and 5 mL of a 1:4 mixture of H2O2/HNO3 was added to the vial. This mixture (with the aerosol residue) was heated in the same fashion and then evaporated. Finally, a mixture of HCl and HNO3 was added to the aerosol residue,

FIGURE 2. δ56FePM2.5 values for the PM2.5 (triangles) and PM10 (filled circles) aerosols. Dates marked with a “*” indicate that δ56FePM2.5 measurements were not made on this date, due to low concentrations of iron. Note that two different sources of PM10 are likely, based on this data. Dates are in 2008. heated for 10 h and evaporated, where the now clean Teflon filter was rinsed with concentrated HNO3 and discarded. Another mixture of HCl and HNO3 was added to the Teflon vessel, transferred to a precleaned LDPE bottle and diluted to 30 mL with >18.0 MΩ water. The resulting solution was 2% HNO3 and 0.5% HCl. Elemental data were obtained using quadrupole ICPMS (X-Series, Thermo Electron Corp.). To minimize isobaric interferences, collision cell gas (7% H2 and 93% He) was utilized for analysis of 54Fe, 56Fe, 57Fe, 75As, and 78 Se. Indium, germanium, and bismuth were used as internal standards. Elements were quantified by using matrix-matched calibration curves from diluted individual ICP standards (Alfa Aesar). The elements analyzed include Na, Mg, Al, P, K, Ca, Ti, V, Cr, Mn, Fe, Co, Ni, Cu, Zn, As, Se, Rb, Sr, Mo, Cd, Sb, Cs, Ba, W, Pb, and U. Recovery of trace elements was verified by digesting a standard reference material (NIST SRM 2709, San Joaquin Soil) under the same conditions as the PM. Elements with a recovery of worse than (20% were discarded. Iron Isotope Analysis. To minimize isobaric interferences (e.g., 54Cr for 54Fe or 40Ca16O for 56Fe), column chromatography was employed to remove the other metals in the digested sample (25). For each sample to be analyzed, a minimum of 4 µg (maximum 25 µg) of iron was transferred from the digested sample and evaporated in a clean 7 mL Teflon vessel. Due to their low iron content, the iron isotope composition for samples containing 95% of the iron was recovered for each sample (18). Finally, the Fe-containing eluent was evaporated and the residue brought up in distilled 0.32 M HNO3 for MC-ICP-MS analysis.

Iron isotope composition was determined using a MCICP-MS (Neptune, Thermo Electron Corp.). Because of the 56 Fe and 57Fe interferences with compounds formed from the argon plasma (40Ar16O and 40Ar16O1H, respectively), analyses were performed in either medium (4-8 µg Fe) or high mass resolution mode (>8 µg Fe). Prior to analysis, samples were diluted to 0.8-1.0 ppm Fe (medium resolution) or 2.5 ppm Fe (high resolution). Sample introduction at a flow rate of 50 or 100 µL min-1 was achieved using a temperature controlled nebulizer and a heated spraychamber (Apex, Elemental Scientific Inc.). Sample-standard bracketing was used to determine the iron isotopic composition of the aerosol and soil samples (28). To correct for instrumental bias, copper standard solution was added to each sample. More details of the MC-ICP-MS measurement can be found in other sources (29, 30). Data are reported in δ notation (per mil, ‰), relative to the IRMM-014 standard (Institute for Reference Materials and Measurements, Geel, Belgium). δ56Fe )

[

(56Fe/ 54Fe)sample (56Fe/ 54Fe)IRMM-014

]

- 1 × 1000

(1)

The performance of each run was monitored by the analysis of a gravimetric standard (δ56Fe ) -1.00‰) and an in-house marine sediment standard (δ56Fe ) -0.86‰). Satisfactory data was obtained if the standards were measured within (0.03‰ of the actual value. Reproducibility was determined by analyzing samples on different days and under different instrument conditions. As seen in Figure 2, the precision was (0.08‰(1σ) or better.

Results and Discussion The PM2.5 and PM10 elemental and total mass concentration data for each 48-h sampling period can be seen in Figure 1. There is a 3-week delay between the first sampling period and the subsequent periods due to a sampler malfunction. We have displayed the major elements in Figure 1. When combined, these elements make up >10% of the total PM10 aerosol mass. The detailed results (in ng m-3) for all of the elemental data can be found in the Supporting Information (Table S1). The elemental concentrations at this desert site were also compared with dust collected in Dunhuang, China, near the Gobi Desert. The relative total elemental signature of the PM10 fraction is similar to that of the total suspended particles (TSP) collected in Dunhuang, China during the ACEAsia campaign. Assuming similarities between these desert sites, this suggests that a majority of the PM10 collected in this study is composed of desert dust. However, the absolute concentrations are somewhat lower during the current study (7). One major difference between the two desert sites is that the relative iron concentration at the Higley site is much greater than that at Dunhuang, compared to the other crustal elements (Na, Ca, Al, and Mg). The aerosol composition observed here is also similar to the crustal signature assigned by positive matrix factorization (PMF) studies performed in Phoenix, AZ (11, 24, 31). In addition, the relative elemental composition of the PM10 fraction of the resuspended soil (Figure S1) collected at the sampling site is identical to most of the PM10 aerosol samples (Figure 1). These modeling and experimental observations support the notion that windblown dust is a major source of PM10 to this area. Iron in this region makes up a significant fraction of the overall PM10 mass (2.5-3.5%), with air concentrations reaching almost 2.5 µg Fe m-3 in some samples. We also observe elevated air concentrations of Na, Ca, Mg, Al, and K, all elements which are crustally derived (24). The key fact that much of the coarse particulate matter is primarily crustal in character highlights the difficulty in this region to VOL. 43, NO. 12, 2009 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

4329

FIGURE 3. δ56Fe versus IRMM-014 in the PM10 fraction as a function of wind direction. For simplification, wind directions are given as the 1-h maximum sustained wind over the sampling period. The light iron is present when strong winds originate from the south-southwest. differentiate industrial and automotive sources from the high crustal background. Significant differences in mass and elemental concentrations are also observed in PM collected during the two sampling periods on May 24-27, 2008. These differences are due to a steady rain during those periods which preferentially washed much of the PM10-2.5 from the air. Figure 2 summarizes the iron isotopic composition of the PM10 (filled circles) and the PM2.5 (triangles) fractions collected at the site and is presented in tabular form in Table S2. Focusing on the PM10 data, two different groups of iron appear to be present. One subset of samples, which is centered around +0.03 ‰, has values similar to those seen in past studies of desert dust collected downwind of the Gobi Desert during the ACE-Asia campaign. Both the PM10 subset (+0.03‰) and the ACE-Asia samples have iron isotopic compositions similar to the bulk silicate crust (21), consistent with wind-blown surface dust. However, we also observe a lighter group of samples in the PM10 fraction for which δ56Fe is centered at -0.18‰. The difference of 0.21‰ within the PM10 fraction is greater than the estimated external precision of the measurement technique (99% confidence, t test). To better differentiate the two PM10 sources, we utilized back trajectory analysis using the HYSPLIT back trajectory model. In this case, no correlations between air mass origin and iron isotopic composition were observed. Therefore, we conclude that the iron isotope difference in the PM10 fraction must be locally derived. Thus, hourly wind direction and wind speed were considered. During some sampling periods, the winds were highly advective, (up to 30 km hr-1), but they were usually relatively calmer (around 20 km hr-1). The lighter PM10 samples (δ56Fe ≈ -0.18 ‰) were always associated with strong components of wind originating from the south-southwest direction. To illustrate, Figure 3 portrays the direction of the 1-h maximum sustained wind direction and its corresponding iron isotopic composition (as δ56Fe). One-hour sustained wind data were used because, with the exception of one sampling day, these data approximated the general wind direction for the entire sampling period. Figure 3 shows that, when the winds are not originating from the southwesterly direction, the iron appears to be of crustal origin (δ56Fe ) +0.03 ‰). A parallel wind diagram for the PM2.5 fraction (Figure S2) does not show any such dependence of iron isotope composition on wind direction. As a result, we can infer that that the iron in the coarse fraction (PM10-2.5) is responsible for the two different subsets within 4330

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 43, NO. 12, 2009

the PM10 aerosol. Also, no dependence was found with wind speed, only with wind direction. Further attempts to determine the source of the light iron in the PM10 fraction are shown in Table 1. Six soil samples were collected from areas in and around the site and resuspended into PM2.5 and PM10 fractions: (1) road dust (two samples), (2) agricultural soil (three samples), and (3) the site itself (one sample). According to Table 1, there is very little deviation from 0.00‰ in most of these fractions. The only differences appear in the two PM2.5 road dust samples, in which we measure iron isotopic compositions of -0.10‰ and +0.14‰, relative to the IRMM-014 standard. From these analyses, we infer that automobile emissions (using road dust as a proxy) and wind-blown agricultural soils are not the source of the anomalously light iron in the PM10 fraction. Past studies have shown that the iron isotopic composition in many agricultural crops tends toward light iron (32). Agricultural samples of vegetation around the sampling site were not collected as part of this study, and it is possible that the light iron originating from southwest of the site comes from vegetative detritus from the surrounding agricultural areas. There were three major crops growing in the area during the course of the sampling study (corn, alfalfa, and grains). However it is not clear if intensity of each crop was a function of direction from the site. Figure 2 also summarizes the iron isotopic composition of the PM2.5 fraction in all samples analyzed. It is apparent that the iron present in the PM2.5 fraction is isotopically lighter than that found in the PM10 fraction. Unlike with the PM10 aerosol, however, no correlation was observed in the PM2.5 fraction between iron isotopic composition and wind direction or wind speed (See Figure S2). We hypothesize that the difference in iron isotopic composition between the two size fractions is primarily the result of different sources of the PM2.5 aerosols. This possibility is explored in more detail below. Principal component analysis (PCA) was used to assess the sources of material to the PM2.5 aerosol (SPSS version 15.0, Varimax rotation method). The results are shown in Table 2. Although PCA is subject to substantial errors with small sample sets such as this one, this analysis yielded some useful insights, primarily from Factors 1 and 2. The analysis resulted in five factors and, for clarity, elements with correlations >0.50 are shown in bold type. One factor, which contributed to 0.75), (ii) elements with intermediate slopes (0.30-0.60), and (iii) elements with very low slopes (